From a30d757cfd4ad217bad1f6b6bd9e405506e05fab Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ewa=20Tusie=C5=84?= <ewa.tusien@intel.com>
Date: Mon, 7 Jun 2021 11:58:22 +0200
Subject: [PATCH 001/178] [IE CLDNN] Convert permutation order in MatMul
 (#6005)

---
 inference-engine/src/cldnn_engine/ops/matmul.cpp | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/inference-engine/src/cldnn_engine/ops/matmul.cpp b/inference-engine/src/cldnn_engine/ops/matmul.cpp
index b18e199581cd15..5fe6f1432586c3 100644
--- a/inference-engine/src/cldnn_engine/ops/matmul.cpp
+++ b/inference-engine/src/cldnn_engine/ops/matmul.cpp
@@ -83,10 +83,11 @@ void CreateMatMulOp(Program& p, const std::shared_ptr<ngraph::op::v0::MatMul>& o
             for (auto o = transpose_order.size(); o < 4; o++)
                 transpose_order.push_back((uint16_t)o);
 
+            std::vector<uint16_t> cldnn_permute_order = ConvertPermuteOrder(transpose_order);
             auto permuteName = op->get_friendly_name() + "/transpose_b";
             auto permutePrim = cldnn::permute(permuteName,
                                               weightsName,
-                                              transpose_order);
+                                              cldnn_permute_order);
             p.AddPrimitive(permutePrim);
             p.AddInnerPrimitiveToProfiler(permuteName, layerName, op);
             weightsName = permuteName;
@@ -102,10 +103,11 @@ void CreateMatMulOp(Program& p, const std::shared_ptr<ngraph::op::v0::MatMul>& o
             for (auto o = transpose_order.size(); o < 4; o++)
                 transpose_order.push_back((uint16_t)o);
 
+            std::vector<uint16_t> cldnn_permute_order = ConvertPermuteOrder(transpose_order);
             auto permuteName = op->get_friendly_name() + "/transpose_a";
             auto permutePrim = cldnn::permute(permuteName,
                                               inputName,
-                                              transpose_order);
+                                              cldnn_permute_order);
             p.AddPrimitive(permutePrim);
             p.AddInnerPrimitiveToProfiler(permuteName, layerName, op);
             inputName = permuteName;

From 9fd73b4f1a72bc3910c0a300042ef5f1e52c2450 Mon Sep 17 00:00:00 2001
From: Mikhail Nosov <mikhail.nosov@intel.com>
Date: Mon, 7 Jun 2021 13:43:32 +0300
Subject: [PATCH 002/178] [nGraph] Enable 'clang' for frontend-related unit
 tests (#6024)

* [nGraph] Enable 'clang' for frontend-related unit tests

* Moved frontend sources and headers before 'clang' definitions
---
 ngraph/test/CMakeLists.txt                    |  17 +--
 ngraph/test/frontend/frontend_manager.cpp     |  28 ++--
 .../frontend/shared/include/basic_api.hpp     |   4 +-
 .../frontend/shared/include/partial_shape.hpp |  11 +-
 ngraph/test/frontend/shared/include/utils.hpp |   5 +-
 ngraph/test/frontend/shared/src/basic_api.cpp |  65 +++-------
 .../shared/src/cut_specific_model.cpp         | 120 ++++++++++--------
 .../frontend/shared/src/partial_shape.cpp     |  27 ++--
 8 files changed, 132 insertions(+), 145 deletions(-)

diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index c6ce9d629770b3..52148150e81562 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -468,7 +468,15 @@ if (NGRAPH_ONNX_EDITOR_ENABLE)
         onnx/onnx_import_with_editor.in.cpp)
 endif()
 
-add_clang_format_target(unit-test_clang FOR_SOURCES ${SRC} ${MULTI_TEST_SRC})
+# SOURCE AND HEADERS FOR FRONTEND TESTING
+file(GLOB FRONTEND_TESTS_SRC ${CMAKE_CURRENT_SOURCE_DIR}/frontend/frontend_manager.cpp)
+set(SRC ${FRONTEND_TESTS_SRC} ${SRC})
+
+file(GLOB FRONTEND_SHARED_TESTS_SRC ${CMAKE_CURRENT_SOURCE_DIR}/frontend/shared/src/*.cpp)
+file(GLOB FRONTEND_SHARED_TESTS_HDR ${CMAKE_CURRENT_SOURCE_DIR}/frontend/shared/include/*.hpp)
+set(SRC ${FRONTEND_SHARED_TESTS_SRC} ${SRC})
+
+add_clang_format_target(unit-test_clang FOR_SOURCES ${SRC} ${MULTI_TEST_SRC} ${FRONTEND_SHARED_TESTS_HDR})
 
 foreach(BACKEND_NAME ${ACTIVE_BACKEND_LIST})
     string(TOLOWER ${BACKEND_NAME} BACKEND_DIR)
@@ -485,13 +493,6 @@ foreach(BACKEND_NAME ${ACTIVE_BACKEND_LIST})
     message(STATUS "Adding unit test for backend ${BACKEND_NAME}")
 endforeach()
 
-# SOURCE FOR FRONTEND TESTING
-file(GLOB FRONTEND_TESTS_SRC ${CMAKE_CURRENT_SOURCE_DIR}/frontend/frontend_manager.cpp)
-set(SRC ${FRONTEND_TESTS_SRC} ${SRC})
-
-file(GLOB FRONTEND_SHARED_TESTS_SRC ${CMAKE_CURRENT_SOURCE_DIR}/frontend/shared/src/*.cpp)
-set(SRC ${FRONTEND_SHARED_TESTS_SRC} ${SRC})
-
 add_executable(unit-test ${SRC})
 
 target_include_directories(unit-test PRIVATE ".")
diff --git a/ngraph/test/frontend/frontend_manager.cpp b/ngraph/test/frontend/frontend_manager.cpp
index 4500a2c509dfc6..10e8eac8e60001 100644
--- a/ngraph/test/frontend/frontend_manager.cpp
+++ b/ngraph/test/frontend/frontend_manager.cpp
@@ -2,11 +2,11 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <memory>
 #include <frontend_manager/frontend_manager.hpp>
+#include <memory>
 
-#include "gtest/gtest.h"
 #include "gmock/gmock.h"
+#include "gtest/gtest.h"
 
 #include "backend.hpp"
 #include "ngraph/file_util.hpp"
@@ -33,10 +33,8 @@ static int set_test_env(const char* name, const char* value)
 TEST(FrontEndManagerTest, testAvailableFrontEnds)
 {
     FrontEndManager fem;
-    ASSERT_NO_THROW(fem.register_front_end("mock", [](FrontEndCapFlags fec)
-    {
-        return std::make_shared<FrontEnd>();
-    }));
+    ASSERT_NO_THROW(fem.register_front_end(
+        "mock", [](FrontEndCapFlags fec) { return std::make_shared<FrontEnd>(); }));
     auto frontends = fem.get_available_front_ends();
     ASSERT_NE(std::find(frontends.begin(), frontends.end(), "mock"), frontends.end());
     FrontEnd::Ptr fe;
@@ -53,13 +51,11 @@ TEST(FrontEndManagerTest, testAvailableFrontEnds)
 
 TEST(FrontEndManagerTest, testLoadWithFlags)
 {
-    int expFlags = FrontEndCapabilities::FEC_CUT |
-            FrontEndCapabilities::FEC_WILDCARDS |
-            FrontEndCapabilities::FEC_NAMES;
+    int expFlags = FrontEndCapabilities::FEC_CUT | FrontEndCapabilities::FEC_WILDCARDS |
+                   FrontEndCapabilities::FEC_NAMES;
     int actualFlags = FrontEndCapabilities::FEC_DEFAULT;
     FrontEndManager fem;
-    ASSERT_NO_THROW(fem.register_front_end("mock", [&actualFlags](int fec)
-    {
+    ASSERT_NO_THROW(fem.register_front_end("mock", [&actualFlags](int fec) {
         actualFlags = fec;
         return std::make_shared<FrontEnd>();
     }));
@@ -75,8 +71,8 @@ TEST(FrontEndManagerTest, testLoadWithFlags)
 
 TEST(FrontEndManagerTest, testMockPluginFrontEnd)
 {
-    std::string fePath =
-            ngraph::file_util::get_directory(ngraph::runtime::Backend::get_backend_shared_library_search_directory());
+    std::string fePath = ngraph::file_util::get_directory(
+        ngraph::runtime::Backend::get_backend_shared_library_search_directory());
     fePath = fePath + FrontEndPathSeparator + "someInvalidPath";
     set_test_env("OV_FRONTEND_PATH", fePath.c_str());
 
@@ -91,7 +87,7 @@ TEST(FrontEndManagerTest, testDefaultFrontEnd)
     FrontEndManager fem;
     ASSERT_ANY_THROW(fem.load_by_model(""));
 
-    std::unique_ptr<FrontEnd> fePtr (new FrontEnd()); // to verify base destructor
+    std::unique_ptr<FrontEnd> fePtr(new FrontEnd()); // to verify base destructor
     FrontEnd::Ptr fe = std::make_shared<FrontEnd>();
     ASSERT_ANY_THROW(fe->load_from_file(""));
     ASSERT_ANY_THROW(fe->load_from_files({"", ""}));
@@ -109,7 +105,7 @@ TEST(FrontEndManagerTest, testDefaultFrontEnd)
 
 TEST(FrontEndManagerTest, testDefaultInputModel)
 {
-    std::unique_ptr<InputModel> imPtr (new InputModel()); // to verify base destructor
+    std::unique_ptr<InputModel> imPtr(new InputModel()); // to verify base destructor
     InputModel::Ptr im = std::make_shared<InputModel>();
     ASSERT_ANY_THROW(im->get_inputs());
     ASSERT_ANY_THROW(im->get_outputs());
@@ -139,7 +135,7 @@ TEST(FrontEndManagerTest, testDefaultInputModel)
 
 TEST(FrontEndManagerTest, testDefaultPlace)
 {
-    std::unique_ptr<Place> placePtr (new Place()); // to verify base destructor
+    std::unique_ptr<Place> placePtr(new Place()); // to verify base destructor
     Place::Ptr place = std::make_shared<Place>();
     ASSERT_ANY_THROW(place->get_names());
     ASSERT_ANY_THROW(place->get_consuming_operations());
diff --git a/ngraph/test/frontend/shared/include/basic_api.hpp b/ngraph/test/frontend/shared/include/basic_api.hpp
index cf2f37f13aa0cf..5aaf97a09d6801 100644
--- a/ngraph/test/frontend/shared/include/basic_api.hpp
+++ b/ngraph/test/frontend/shared/include/basic_api.hpp
@@ -9,8 +9,8 @@
 #include <gtest/gtest.h>
 
 using BasicTestParam = std::tuple<std::string,  // FrontEnd name
-        std::string,  // Base path to models
-        std::string>; // Model name
+                                  std::string,  // Base path to models
+                                  std::string>; // Model name
 
 class FrontEndBasicTest : public ::testing::TestWithParam<BasicTestParam>
 {
diff --git a/ngraph/test/frontend/shared/include/partial_shape.hpp b/ngraph/test/frontend/shared/include/partial_shape.hpp
index f06f40cf576655..b3aff5aff5be64 100644
--- a/ngraph/test/frontend/shared/include/partial_shape.hpp
+++ b/ngraph/test/frontend/shared/include/partial_shape.hpp
@@ -10,12 +10,13 @@
 
 struct BaseFEParam
 {
-    BaseFEParam()
-    {}
+    BaseFEParam() {}
 
-    BaseFEParam(const std::string& name, const std::string& path) :
-            m_frontEndName(name), m_modelsPath(path)
-    {}
+    BaseFEParam(const std::string& name, const std::string& path)
+        : m_frontEndName(name)
+        , m_modelsPath(path)
+    {
+    }
 
     std::string m_frontEndName;
     std::string m_modelsPath;
diff --git a/ngraph/test/frontend/shared/include/utils.hpp b/ngraph/test/frontend/shared/include/utils.hpp
index 6bb4106c02e042..4a594c480ff871 100644
--- a/ngraph/test/frontend/shared/include/utils.hpp
+++ b/ngraph/test/frontend/shared/include/utils.hpp
@@ -19,11 +19,12 @@ namespace FrontEndTestUtils
             if (c == '/')
             {
                 c = '_';
-            } else if (c == '.')
+            }
+            else if (c == '.')
             {
                 c = '_';
             }
         }
         return res;
     }
-}
\ No newline at end of file
+} // namespace FrontEndTestUtils
\ No newline at end of file
diff --git a/ngraph/test/frontend/shared/src/basic_api.cpp b/ngraph/test/frontend/shared/src/basic_api.cpp
index 1bce97cb6a6ef2..85688762893fb9 100644
--- a/ngraph/test/frontend/shared/src/basic_api.cpp
+++ b/ngraph/test/frontend/shared/src/basic_api.cpp
@@ -2,7 +2,6 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <regex>
 #include "../include/basic_api.hpp"
 #include "../include/utils.hpp"
 
@@ -50,13 +49,11 @@ TEST_P(FrontEndBasicTest, testInputModel_getInputsOutputs)
     ASSERT_NO_THROW(doLoadFromFile());
 
     using CustomCheck = std::function<void(Place::Ptr)>;
-    auto checkPlaces = [&](const std::vector<Place::Ptr>& places, CustomCheck cb)
-    {
+    auto checkPlaces = [&](const std::vector<Place::Ptr>& places, CustomCheck cb) {
         EXPECT_GT(places.size(), 0);
         std::set<Place::Ptr> placesSet(places.begin(), places.end());
         EXPECT_EQ(placesSet.size(), places.size());
-        std::for_each(places.begin(), places.end(), [&](Place::Ptr place)
-        {
+        std::for_each(places.begin(), places.end(), [&](Place::Ptr place) {
             ASSERT_NE(place, nullptr);
             std::vector<std::string> names;
             ASSERT_NO_THROW(names = place->get_names());
@@ -66,25 +63,18 @@ TEST_P(FrontEndBasicTest, testInputModel_getInputsOutputs)
     };
     std::vector<Place::Ptr> inputs;
     ASSERT_NO_THROW(inputs = m_inputModel->get_inputs());
-    checkPlaces(inputs, [&](Place::Ptr place)
-    {
-        EXPECT_TRUE(place->is_input());
-    });
+    checkPlaces(inputs, [&](Place::Ptr place) { EXPECT_TRUE(place->is_input()); });
 
     std::vector<Place::Ptr> outputs;
     ASSERT_NO_THROW(outputs = m_inputModel->get_outputs());
-    checkPlaces(outputs, [&](Place::Ptr place)
-    {
-        EXPECT_TRUE(place->is_output());
-    });
+    checkPlaces(outputs, [&](Place::Ptr place) { EXPECT_TRUE(place->is_output()); });
 }
 
 TEST_P(FrontEndBasicTest, testInputModel_getPlaceByTensorName)
 {
     ASSERT_NO_THROW(doLoadFromFile());
 
-    auto testGetPlaceByTensorName = [&](const std::vector<Place::Ptr>& places)
-    {
+    auto testGetPlaceByTensorName = [&](const std::vector<Place::Ptr>& places) {
         EXPECT_GT(places.size(), 0);
         for (auto place : places)
         {
@@ -117,8 +107,7 @@ TEST_P(FrontEndBasicTest, testInputModel_overrideAll)
 
     using GetPlaces = std::function<std::vector<Place::Ptr>()>;
     using OverridePlaces = std::function<void(const std::vector<Place::Ptr>&)>;
-    auto verifyOverride = [](GetPlaces getCB, OverridePlaces overrideCB)
-    {
+    auto verifyOverride = [](GetPlaces getCB, OverridePlaces overrideCB) {
         std::vector<Place::Ptr> places;
         ASSERT_NO_THROW(places = getCB());
         std::set<Place::Ptr> placesSet(places.begin(), places.end());
@@ -130,20 +119,16 @@ TEST_P(FrontEndBasicTest, testInputModel_overrideAll)
         EXPECT_GT(places.size(), 0);
         std::set<Place::Ptr> placesSetAfter(places.begin(), places.end());
         EXPECT_EQ(placesSet.size(), placesSet.size());
-        std::for_each(places.begin(), places.end(), [&](Place::Ptr place)
-        {
+        std::for_each(places.begin(), places.end(), [&](Place::Ptr place) {
             EXPECT_GT(placesSet.count(place), 0);
         });
     };
-    verifyOverride([&]()
-                   { return m_inputModel->get_inputs(); },
-                   [&](const std::vector<Place::Ptr>& p)
-                   { m_inputModel->override_all_inputs(p); });
-
-    verifyOverride([&]()
-                   { return m_inputModel->get_outputs(); },
-                   [&](const std::vector<Place::Ptr>& p)
-                   { m_inputModel->override_all_outputs(p); });
+    verifyOverride([&]() { return m_inputModel->get_inputs(); },
+                   [&](const std::vector<Place::Ptr>& p) { m_inputModel->override_all_inputs(p); });
+
+    verifyOverride(
+        [&]() { return m_inputModel->get_outputs(); },
+        [&](const std::vector<Place::Ptr>& p) { m_inputModel->override_all_outputs(p); });
 }
 
 TEST_P(FrontEndBasicTest, testInputModel_overrideAll_empty)
@@ -152,16 +137,14 @@ TEST_P(FrontEndBasicTest, testInputModel_overrideAll_empty)
     using GetPlaces = std::function<std::vector<Place::Ptr>()>;
     using OverrideEmpty = std::function<void(void)>;
     using CustomCheck = std::function<void(std::string)>;
-    auto verifyOverride = [](GetPlaces getCB, OverrideEmpty overrideCB, CustomCheck customCB)
-    {
+    auto verifyOverride = [](GetPlaces getCB, OverrideEmpty overrideCB, CustomCheck customCB) {
         std::vector<Place::Ptr> places;
         std::vector<Place::Ptr> newPlaces;
         ASSERT_NO_THROW(places = getCB());
         ASSERT_NO_THROW(overrideCB());
         ASSERT_NO_THROW(newPlaces = getCB());
         ASSERT_EQ(newPlaces.size(), 0);
-        std::for_each(places.begin(), places.end(), [&](Place::Ptr place)
-        {
+        std::for_each(places.begin(), places.end(), [&](Place::Ptr place) {
             std::vector<std::string> names;
             ASSERT_NO_THROW(names = place->get_names());
             for (auto name : names)
@@ -170,21 +153,15 @@ TEST_P(FrontEndBasicTest, testInputModel_overrideAll_empty)
             }
         });
     };
-    verifyOverride([&]()
-                   { return m_inputModel->get_outputs(); },
-                   [&]()
-                   { m_inputModel->override_all_outputs({}); },
-                   [&](const std::string& name)
-                   {
+    verifyOverride([&]() { return m_inputModel->get_outputs(); },
+                   [&]() { m_inputModel->override_all_outputs({}); },
+                   [&](const std::string& name) {
                        EXPECT_FALSE(m_inputModel->get_place_by_tensor_name(name)->is_output());
                    });
 
-    verifyOverride([&]()
-                   { return m_inputModel->get_inputs(); },
-                   [&]()
-                   { m_inputModel->override_all_inputs({}); },
-                   [&](const std::string& name)
-                   {
+    verifyOverride([&]() { return m_inputModel->get_inputs(); },
+                   [&]() { m_inputModel->override_all_inputs({}); },
+                   [&](const std::string& name) {
                        EXPECT_FALSE(m_inputModel->get_place_by_tensor_name(name)->is_input());
                    });
 }
diff --git a/ngraph/test/frontend/shared/src/cut_specific_model.cpp b/ngraph/test/frontend/shared/src/cut_specific_model.cpp
index d1ecbae8976705..d9fe9ce7e1b5f9 100644
--- a/ngraph/test/frontend/shared/src/cut_specific_model.cpp
+++ b/ngraph/test/frontend/shared/src/cut_specific_model.cpp
@@ -2,8 +2,6 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <regex>
-#include <algorithm>
 #include "../include/cut_specific_model.hpp"
 #include "../include/utils.hpp"
 #include "ngraph/opsets/opset7.hpp"
@@ -14,8 +12,7 @@ using namespace ngraph::frontend;
 static std::string joinStrings(const std::vector<std::string>& strings)
 {
     std::ostringstream res;
-    std::copy(strings.begin(), strings.end(),
-              std::ostream_iterator<std::string>(res, "_"));
+    std::copy(strings.begin(), strings.end(), std::ostream_iterator<std::string>(res, "_"));
     return res.str();
 }
 
@@ -85,7 +82,7 @@ TEST_P(FrontEndCutModelTest, testOverrideInputs)
         std::vector<std::string> names;
         ASSERT_NO_THROW(names = newInput->get_names());
         bool found = false;
-        for (const auto& name: m_param.m_newInputs)
+        for (const auto& name : m_param.m_newInputs)
         {
             if (std::find(names.begin(), names.begin(), name) != names.end())
             {
@@ -110,7 +107,7 @@ TEST_P(FrontEndCutModelTest, testOverrideOutputs)
         std::vector<std::string> names;
         ASSERT_NO_THROW(names = newOutput->get_names());
         bool found = false;
-        for (const auto& name: m_param.m_newOutputs)
+        for (const auto& name : m_param.m_newOutputs)
         {
             if (std::find(names.begin(), names.begin(), name) != names.end())
             {
@@ -132,11 +129,13 @@ TEST_P(FrontEndCutModelTest, testOldInputs)
     // Ensure that it contains expected old inputs
     for (const auto& name : m_param.m_oldInputs)
     {
-        EXPECT_TRUE(std::find_if(ops.begin(), ops.end(),
-                                 [&](const std::shared_ptr<ngraph::Node>& node)
-                                 {
-                                     return node->get_friendly_name().find(name) != std::string::npos;
-                                 }) != ops.end()) << "Name not found:" << name;
+        EXPECT_TRUE(std::find_if(ops.begin(),
+                                 ops.end(),
+                                 [&](const std::shared_ptr<ngraph::Node>& node) {
+                                     return node->get_friendly_name().find(name) !=
+                                            std::string::npos;
+                                 }) != ops.end())
+            << "Name not found:" << name;
     }
 }
 
@@ -149,11 +148,13 @@ TEST_P(FrontEndCutModelTest, testOldOutputs)
     // Ensure that it contains expected old outputs
     for (const auto& name : m_param.m_oldOutputs)
     {
-        EXPECT_TRUE(std::find_if(ops.begin(), ops.end(),
-                                 [&](const std::shared_ptr<ngraph::Node>& node)
-                                 {
-                                     return node->get_friendly_name().find(name) != std::string::npos;
-                                 }) != ops.end()) << "Name not found:" << name;
+        EXPECT_TRUE(std::find_if(ops.begin(),
+                                 ops.end(),
+                                 [&](const std::shared_ptr<ngraph::Node>& node) {
+                                     return node->get_friendly_name().find(name) !=
+                                            std::string::npos;
+                                 }) != ops.end())
+            << "Name not found:" << name;
     }
 }
 
@@ -171,21 +172,25 @@ TEST_P(FrontEndCutModelTest, testNewInputs_func)
     // Ensure that it doesn't contain old inputs
     for (const auto& name : m_param.m_oldInputs)
     {
-        EXPECT_TRUE(std::find_if(ops.begin(), ops.end(),
-                                 [&](const std::shared_ptr<ngraph::Node>& node)
-                                 {
-                                     return node->get_friendly_name().find(name) != std::string::npos;
-                                 }) == ops.end()) << "Name shall not exist:" << name;
+        EXPECT_TRUE(std::find_if(ops.begin(),
+                                 ops.end(),
+                                 [&](const std::shared_ptr<ngraph::Node>& node) {
+                                     return node->get_friendly_name().find(name) !=
+                                            std::string::npos;
+                                 }) == ops.end())
+            << "Name shall not exist:" << name;
     }
 
     // Ensure that it contains expected new inputs
     for (const auto& name : m_param.m_newInputs)
     {
-        EXPECT_TRUE(std::find_if(ops.begin(), ops.end(),
-                                 [&](const std::shared_ptr<ngraph::Node>& node)
-                                 {
-                                     return node->get_friendly_name().find(name) != std::string::npos;
-                                 }) != ops.end()) << "Name not found:" << name;
+        EXPECT_TRUE(std::find_if(ops.begin(),
+                                 ops.end(),
+                                 [&](const std::shared_ptr<ngraph::Node>& node) {
+                                     return node->get_friendly_name().find(name) !=
+                                            std::string::npos;
+                                 }) != ops.end())
+            << "Name not found:" << name;
     }
 }
 
@@ -203,21 +208,25 @@ TEST_P(FrontEndCutModelTest, testNewOutputs_func)
     // Ensure that it doesn't contain old outputs
     for (const auto& name : m_param.m_oldOutputs)
     {
-        EXPECT_TRUE(std::find_if(ops.begin(), ops.end(),
-                                 [&](const std::shared_ptr<ngraph::Node>& node)
-                                 {
-                                     return node->get_friendly_name().find(name) != std::string::npos;
-                                 }) == ops.end()) << "Name shall not exist:" << name;
+        EXPECT_TRUE(std::find_if(ops.begin(),
+                                 ops.end(),
+                                 [&](const std::shared_ptr<ngraph::Node>& node) {
+                                     return node->get_friendly_name().find(name) !=
+                                            std::string::npos;
+                                 }) == ops.end())
+            << "Name shall not exist:" << name;
     }
 
     // Ensure that it contains expected new outputs
     for (const auto& name : m_param.m_newOutputs)
     {
-        EXPECT_TRUE(std::find_if(ops.begin(), ops.end(),
-                                 [&](const std::shared_ptr<ngraph::Node>& node)
-                                 {
-                                     return node->get_friendly_name().find(name) != std::string::npos;
-                                 }) != ops.end()) << "Name not found:" << name;
+        EXPECT_TRUE(std::find_if(ops.begin(),
+                                 ops.end(),
+                                 [&](const std::shared_ptr<ngraph::Node>& node) {
+                                     return node->get_friendly_name().find(name) !=
+                                            std::string::npos;
+                                 }) != ops.end())
+            << "Name not found:" << name;
     }
 }
 
@@ -236,21 +245,25 @@ TEST_P(FrontEndCutModelTest, testExtractSubgraph)
     // Ensure that it doesn't contain expected old outputs
     for (const auto& name : m_param.m_oldOutputs)
     {
-        EXPECT_TRUE(std::find_if(ops.begin(), ops.end(),
-                                 [&](const std::shared_ptr<ngraph::Node>& node)
-                                 {
-                                     return node->get_friendly_name().find(name) != std::string::npos;
-                                 }) == ops.end()) << "Name shall not exist:" << name;
+        EXPECT_TRUE(std::find_if(ops.begin(),
+                                 ops.end(),
+                                 [&](const std::shared_ptr<ngraph::Node>& node) {
+                                     return node->get_friendly_name().find(name) !=
+                                            std::string::npos;
+                                 }) == ops.end())
+            << "Name shall not exist:" << name;
     }
 
     // Ensure that it contains expected new outputs
     for (const auto& name : m_param.m_newOutputs)
     {
-        EXPECT_TRUE(std::find_if(ops.begin(), ops.end(),
-                                 [&](const std::shared_ptr<ngraph::Node>& node)
-                                 {
-                                     return node->get_friendly_name().find(name) != std::string::npos;
-                                 }) != ops.end()) << "Name not found:" << name;
+        EXPECT_TRUE(std::find_if(ops.begin(),
+                                 ops.end(),
+                                 [&](const std::shared_ptr<ngraph::Node>& node) {
+                                     return node->get_friendly_name().find(name) !=
+                                            std::string::npos;
+                                 }) != ops.end())
+            << "Name not found:" << name;
     }
 }
 
@@ -266,13 +279,14 @@ TEST_P(FrontEndCutModelTest, testSetTensorValue)
     auto ops = function->get_ordered_ops();
 
     auto const_name = m_param.m_tensorValueName;
-    auto const_node_it = std::find_if(ops.begin(), ops.end(),
-                                      [&](const std::shared_ptr<ngraph::Node>& node)
-                                      {
-                                          return node->get_friendly_name().find(const_name) != std::string::npos;
-                                      });
+    auto const_node_it =
+        std::find_if(ops.begin(), ops.end(), [&](const std::shared_ptr<ngraph::Node>& node) {
+            return node->get_friendly_name().find(const_name) != std::string::npos;
+        });
     ASSERT_TRUE(const_node_it != ops.end()) << "Name shall exist:" << const_name;
     auto data = std::dynamic_pointer_cast<opset7::Constant>(*const_node_it)->get_vector<float>();
-    EXPECT_EQ(data.size(), m_param.m_tensorValue.size()) << "Data size must be equal to expected size";
-    EXPECT_TRUE(std::equal(data.begin(), data.end(), m_param.m_tensorValue.begin())) << "Data must be equal";
+    EXPECT_EQ(data.size(), m_param.m_tensorValue.size())
+        << "Data size must be equal to expected size";
+    EXPECT_TRUE(std::equal(data.begin(), data.end(), m_param.m_tensorValue.begin()))
+        << "Data must be equal";
 }
diff --git a/ngraph/test/frontend/shared/src/partial_shape.cpp b/ngraph/test/frontend/shared/src/partial_shape.cpp
index e71368a487df55..9f8ce264ce9799 100644
--- a/ngraph/test/frontend/shared/src/partial_shape.cpp
+++ b/ngraph/test/frontend/shared/src/partial_shape.cpp
@@ -2,15 +2,14 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <regex>
-#include <algorithm>
 #include "../include/partial_shape.hpp"
 #include "../include/utils.hpp"
 
 using namespace ngraph;
 using namespace ngraph::frontend;
 
-std::string FrontEndPartialShapeTest::getTestCaseName(const testing::TestParamInfo<PartialShapeParam>& obj)
+std::string
+    FrontEndPartialShapeTest::getTestCaseName(const testing::TestParamInfo<PartialShapeParam>& obj)
 {
     BaseFEParam base;
     PartShape part;
@@ -31,7 +30,8 @@ void FrontEndPartialShapeTest::SetUp()
 void FrontEndPartialShapeTest::initParamTest()
 {
     std::tie(m_baseParam, m_partShape) = GetParam();
-    m_partShape.m_modelName = std::string(TEST_FILES) + m_baseParam.m_modelsPath + m_partShape.m_modelName;
+    m_partShape.m_modelName =
+        std::string(TEST_FILES) + m_baseParam.m_modelsPath + m_partShape.m_modelName;
     std::cout << "Model: " << m_partShape.m_modelName << std::endl;
 }
 
@@ -55,11 +55,9 @@ TEST_P(FrontEndPartialShapeTest, testCheckOldPartialShape)
     std::shared_ptr<ngraph::Function> function;
     ASSERT_NO_THROW(function = m_frontEnd->convert(m_inputModel));
     auto ops = function->get_ordered_ops();
-    auto it = std::find_if(ops.begin(), ops.end(),
-                           [&](const std::shared_ptr<ngraph::Node>& node)
-                           {
-                               return node->get_friendly_name().find(m_partShape.m_tensorName) != std::string::npos;
-                           });
+    auto it = std::find_if(ops.begin(), ops.end(), [&](const std::shared_ptr<ngraph::Node>& node) {
+        return node->get_friendly_name().find(m_partShape.m_tensorName) != std::string::npos;
+    });
     ASSERT_NE(it, ops.end());
     auto shape = (*it)->get_output_partial_shape(0).get_shape();
     ASSERT_EQ(shape.size(), m_partShape.m_oldPartialShape.size());
@@ -75,16 +73,15 @@ TEST_P(FrontEndPartialShapeTest, testSetNewPartialShape)
     Place::Ptr place;
     ASSERT_NO_THROW(place = m_inputModel->get_place_by_tensor_name(m_partShape.m_tensorName));
     ASSERT_NE(place, nullptr);
-    ASSERT_NO_THROW(m_inputModel->set_partial_shape(place, PartialShape{m_partShape.m_newPartialShape}));
+    ASSERT_NO_THROW(
+        m_inputModel->set_partial_shape(place, PartialShape{m_partShape.m_newPartialShape}));
 
     std::shared_ptr<ngraph::Function> function;
     ASSERT_NO_THROW(function = m_frontEnd->convert(m_inputModel));
     auto ops = function->get_ordered_ops();
-    auto it = std::find_if(ops.begin(), ops.end(),
-                           [&](const std::shared_ptr<ngraph::Node>& node)
-                           {
-                               return node->get_friendly_name().find(m_partShape.m_tensorName) != std::string::npos;
-                           });
+    auto it = std::find_if(ops.begin(), ops.end(), [&](const std::shared_ptr<ngraph::Node>& node) {
+        return node->get_friendly_name().find(m_partShape.m_tensorName) != std::string::npos;
+    });
     ASSERT_NE(it, ops.end());
     auto shape = (*it)->get_output_partial_shape(0).get_shape();
     ASSERT_EQ(shape.size(), m_partShape.m_newPartialShape.size());

From f9b27c3714ea987a5c8b8e5644f008496e1255cc Mon Sep 17 00:00:00 2001
From: Tomasz Socha <tomasz.socha@intel.com>
Date: Mon, 7 Jun 2021 13:01:20 +0200
Subject: [PATCH 003/178] =?UTF-8?q?[ONNX]=20=F0=9F=9B=A0=EF=B8=8F?=
 =?UTF-8?q?=F0=9F=95=B8=EF=B8=8F=20Fix=20softmax=20opsets=20(#5959)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 ngraph/frontend/onnx_import/src/op/softmax.cpp   | 4 ++--
 ngraph/frontend/onnx_import/src/op/softmax.hpp   | 4 ++--
 ngraph/frontend/onnx_import/src/ops_bridge.cpp   | 4 +---
 ngraph/python/tests/__init__.py                  | 2 +-
 ngraph/python/tests/test_onnx/test_zoo_models.py | 4 +++-
 5 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/ngraph/frontend/onnx_import/src/op/softmax.cpp b/ngraph/frontend/onnx_import/src/op/softmax.cpp
index 1827799b8e38cc..21b23783d19109 100644
--- a/ngraph/frontend/onnx_import/src/op/softmax.cpp
+++ b/ngraph/frontend/onnx_import/src/op/softmax.cpp
@@ -69,7 +69,7 @@ namespace ngraph
                     return {result};
                 }
             } // namespace set_1
-            namespace set_7
+            namespace set_11
             {
                 OutputVector softmax(const Node& node)
                 {
@@ -109,7 +109,7 @@ namespace ngraph
 
                     return {result};
                 }
-            } // namespace set_7
+            } // namespace set_11
             namespace set_13
             {
                 OutputVector softmax(const Node& node)
diff --git a/ngraph/frontend/onnx_import/src/op/softmax.hpp b/ngraph/frontend/onnx_import/src/op/softmax.hpp
index 5528f3dd259c8f..1878c2045fd412 100644
--- a/ngraph/frontend/onnx_import/src/op/softmax.hpp
+++ b/ngraph/frontend/onnx_import/src/op/softmax.hpp
@@ -19,11 +19,11 @@ namespace ngraph
 
             } // namespace set_1
 
-            namespace set_7
+            namespace set_11
             {
                 OutputVector softmax(const Node& node);
 
-            } // namespace set_7
+            } // namespace set_11
 
             namespace set_13
             {
diff --git a/ngraph/frontend/onnx_import/src/ops_bridge.cpp b/ngraph/frontend/onnx_import/src/ops_bridge.cpp
index 511c059c61badd..b807a31212db64 100644
--- a/ngraph/frontend/onnx_import/src/ops_bridge.cpp
+++ b/ngraph/frontend/onnx_import/src/ops_bridge.cpp
@@ -437,9 +437,7 @@ namespace ngraph
             REGISTER_OPERATOR("Slice", 1, slice);
             REGISTER_OPERATOR("Slice", 10, slice);
             REGISTER_OPERATOR("Softmax", 1, softmax);
-            // Softmax v7 should be in the 11th opset but,
-            // other frameworks(mxnet and onnxruntime) already use for older models.
-            REGISTER_OPERATOR("Softmax", 7, softmax);
+            REGISTER_OPERATOR("Softmax", 11, softmax);
             REGISTER_OPERATOR("Softmax", 13, softmax);
             REGISTER_OPERATOR("Softplus", 1, softplus);
             REGISTER_OPERATOR("Softsign", 1, softsign);
diff --git a/ngraph/python/tests/__init__.py b/ngraph/python/tests/__init__.py
index 5ec9abe7a7bc5a..930be465a632d3 100644
--- a/ngraph/python/tests/__init__.py
+++ b/ngraph/python/tests/__init__.py
@@ -149,5 +149,5 @@ def xfail_test(reason="Mark the test as expected to fail", strict=True):
 xfail_issue_49754 = xfail_test(reason="RuntimeError: Unsupported dynamic ops: v1::TopKIE")
 xfail_issue_52463 = xfail_test(reason="test_operator_add_size1_singleton_broadcast_cpu - "
                                       "Not equal to tolerance")
-
 xfail_issue_45432 = xfail_test(reason="Einsum is not implemented in CPU plugin.")
+xfail_issue_onnx_models_140 = xfail_test(reason="https://github.com/onnx/models/issues/140")
diff --git a/ngraph/python/tests/test_onnx/test_zoo_models.py b/ngraph/python/tests/test_onnx/test_zoo_models.py
index 449ad50457aef0..1954b61e64661d 100644
--- a/ngraph/python/tests/test_onnx/test_zoo_models.py
+++ b/ngraph/python/tests/test_onnx/test_zoo_models.py
@@ -25,7 +25,8 @@
     xfail_issue_47430,
     xfail_issue_47495,
     xfail_issue_48145,
-    xfail_issue_48190)
+    xfail_issue_48190,
+    xfail_issue_onnx_models_140)
 
 MODELS_ROOT_DIR = tests.MODEL_ZOO_DIR
 
@@ -170,6 +171,7 @@ def tinyyolov3_post_processing(outputs : Sequence[Any]) -> Sequence[Any]:
             (xfail_issue_47430, "test_onnx_model_zoo_vision_object_detection_segmentation_fcn_model_fcn_resnet101_11_fcn_resnet101_11_model_cpu"),
             (xfail_issue_48145, "test_onnx_model_zoo_text_machine_comprehension_bert_squad_model_bertsquad_8_download_sample_8_bertsquad8_cpu"),
             (xfail_issue_48190, "test_onnx_model_zoo_text_machine_comprehension_roberta_model_roberta_base_11_roberta_base_11_roberta_base_11_cpu"),
+            (xfail_issue_onnx_models_140, "test_onnx_model_zoo_vision_object_detection_segmentation_duc_model_ResNet101_DUC_7_ResNet101_DUC_HDC_ResNet101_DUC_HDC_cpu"),
 
             # Model MSFT
             (xfail_issue_37973, "test_MSFT_opset7_tf_inception_v2_model_cpu"),

From c1608628d42fbc0d70d94abb7c5cff27c5469a24 Mon Sep 17 00:00:00 2001
From: Ivan Tikhonov <ivan.tikhonov@intel.com>
Date: Mon, 7 Jun 2021 15:13:41 +0300
Subject: [PATCH 004/178] LowLatency v2 ngraph transformation (#5160)

* LowLatency 2.0: transformation and unit tests

* low latency 2.0: unit tests

* documentation and ngraph codestyle

* update CNN Interface of LowLatency transformation

* fix build on Windows

* fix build on Windows

* investigation of a failed build on Win OS

* ngraph codestyle

* fix build (werrors)

* New unit tests, refactoring

* update functional tests for Memory

* update LowLatency functional tests

* extend Memory tests to cover LowLatency v2 transformation

* clean up, code style

* fix unit tests

* update and fix unit tests, add feature to apply LLTv2 after LLTv1

* update docs, refactoring

* add several gna tests to skip config

* fix python api tests

* update python api, rename LowLatency_v2 to LowLatency2

* deprecate LowLatency v1

* Deprecate LowLatency v1 in IE

* fix wrong merge, codestyle

* resolve review comments

* fix python test

* update skip config

* apply online review notes, fix unit tests

* clean up, code style

* fix docs

* Use debug_messages instead of exceptions in llt v2

* fix unit tests

* Resolve review remarks
---
 .../offline_transformations_api.pyx           |   4 +-
 .../offline_transformations_api_impl.cpp      |  11 +-
 .../offline_transformations_api_impl.hpp      |   2 +-
 .../offline_transformations_api_impl_defs.pxd |   5 +-
 .../python/tests/test_offline_api.py          |   2 +-
 .../include/ie_transformations.hpp            |  36 +
 .../inference_engine/ie_transformations.cpp   |  10 +
 .../transformations/low_latency_test.cpp      |  10 +
 .../transformations/low_latency_v2_test.cpp   | 829 ++++++++++++++++++
 .../single_layer_tests/memory.cpp             |   8 +
 .../subgraph_tests/memory_LSTMCell.cpp        |  45 +
 .../subgraph_tests/multiple_LSTMCell.cpp      |  10 +
 .../single_layer_tests/memory.cpp             |  12 +-
 .../skip_tests_config.cpp                     |   8 +
 .../subgraph_tests/memory_LSTMCell.cpp        |   9 +
 .../subgraph_tests/multiple_LSTMCell.cpp      |  10 +
 .../include/subgraph_tests/basic_lstm.hpp     |   2 +-
 .../subgraph_tests/memory_LSTMCell.hpp        |   8 -
 .../subgraph_tests/multiple_LSTMCell.hpp      |   8 -
 .../single_layer/memory.hpp                   |   9 +
 .../subgraph/memory_LSTMCell.hpp              |   8 +-
 .../subgraph/multiple_LSTMCell.hpp            |   8 +-
 .../src/single_layer/memory.cpp               | 116 ++-
 .../src/subgraph/memory_LSTMCell.cpp          | 281 +++---
 .../src/subgraph/multiple_LSTMCell.cpp        | 432 ++++-----
 .../ngraph_functions/utils/ngraph_helpers.hpp |  11 +
 .../src/utils/ngraph_helpers.cpp              |  27 +
 .../mo/back/offline_transformations.py        |   2 +-
 model-optimizer/mo/utils/cli_parser.py        |  17 +-
 .../unit_tests/mo/utils/cli_parser_test.py    |  45 +-
 .../core/include/ngraph/pass/low_latency.hpp  |  44 +-
 ngraph/core/src/op/tensor_iterator.cpp        |   1 -
 ngraph/core/src/pass/low_latency.cpp          | 203 ++++-
 33 files changed, 1806 insertions(+), 427 deletions(-)
 create mode 100644 inference-engine/tests/functional/inference_engine/transformations/low_latency_v2_test.cpp
 create mode 100644 inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp

diff --git a/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api.pyx b/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api.pyx
index bd101280fcbb16..266c1dc94d9475 100644
--- a/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api.pyx
+++ b/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api.pyx
@@ -17,8 +17,8 @@ def ApplyPOTTransformations(IENetwork network, string device):
     C.ApplyPOTTransformations(network.impl, device)
 
 
-def ApplyLowLatencyTransformation(IENetwork network, int64_t num_iterations=1):
-    C.ApplyLowLatencyTransformation(network.impl, num_iterations)
+def ApplyLowLatencyTransformation(IENetwork network, bool use_const_initializer = True):
+    C.ApplyLowLatencyTransformation(network.impl, use_const_initializer)
 
 
 def ApplyPruningTransformation(IENetwork network):
diff --git a/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl.cpp b/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl.cpp
index b9ff879da8c843..183deaccfb388f 100644
--- a/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl.cpp
+++ b/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl.cpp
@@ -26,16 +26,9 @@ void InferenceEnginePython::ApplyPOTTransformations(InferenceEnginePython::IENet
     manager.run_passes(network.actual->getFunction());
 }
 
-void InferenceEnginePython::ApplyLowLatencyTransformation(InferenceEnginePython::IENetwork network, int64_t num_iterations) {
+void InferenceEnginePython::ApplyLowLatencyTransformation(InferenceEnginePython::IENetwork network, bool use_const_initializer) {
     ngraph::pass::Manager manager;
-    // TODO: pass num_iterations to LowLatency
-    manager.register_pass<ngraph::pass::LowLatency>();
-    manager.register_pass<ngraph::pass::UnrollTensorIterator>();
-
-    auto pass_config = manager.get_pass_config();
-    pass_config->set_callback<ngraph::pass::UnrollTensorIterator>([](const std::shared_ptr<const ngraph::Node>& node) -> bool {
-        return node->get_rt_info().count("UNROLL_TI") == 0;
-    });
+    manager.register_pass<ngraph::pass::LowLatency2>(use_const_initializer);
     manager.run_passes(network.actual->getFunction());
 }
 
diff --git a/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl.hpp b/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl.hpp
index 504388e4afc1ad..3941c48a50cfaa 100644
--- a/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl.hpp
+++ b/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl.hpp
@@ -15,7 +15,7 @@ void ApplyMOCTransformations(InferenceEnginePython::IENetwork network, bool cf);
 
 void ApplyPOTTransformations(InferenceEnginePython::IENetwork network, std::string device);
 
-void ApplyLowLatencyTransformation(InferenceEnginePython::IENetwork network, int64_t num_iterations);
+void ApplyLowLatencyTransformation(InferenceEnginePython::IENetwork network, bool use_const_initializer = true);
 
 void ApplyPruningTransformation(InferenceEnginePython::IENetwork network);
 
diff --git a/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl_defs.pxd b/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl_defs.pxd
index 726880e9353f37..551e56c27a8da0 100644
--- a/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl_defs.pxd
+++ b/inference-engine/ie_bridges/python/src/openvino/offline_transformations/offline_transformations_api_impl_defs.pxd
@@ -3,7 +3,6 @@
 
 from libcpp cimport bool
 from libcpp.string cimport string
-from libc.stdint cimport int64_t
 
 from ..inference_engine.ie_api_impl_defs cimport IENetwork
 
@@ -12,10 +11,10 @@ cdef extern from "offline_transformations_api_impl.hpp" namespace "InferenceEngi
 
     cdef void ApplyPOTTransformations(IENetwork network, string device)
 
-    cdef void ApplyLowLatencyTransformation(IENetwork network, int64_t num_iterations)
+    cdef void ApplyLowLatencyTransformation(IENetwork network, bool use_const_initializer)
 
     cdef void ApplyPruningTransformation(IENetwork network)
 
     cdef void GenerateMappingFile(IENetwork network, string path, bool extract_names)
 
-    cdef void CheckAPI()
\ No newline at end of file
+    cdef void CheckAPI()
diff --git a/inference-engine/ie_bridges/python/tests/test_offline_api.py b/inference-engine/ie_bridges/python/tests/test_offline_api.py
index b5565c04bb4d66..0bba0951c27c87 100644
--- a/inference-engine/ie_bridges/python/tests/test_offline_api.py
+++ b/inference-engine/ie_bridges/python/tests/test_offline_api.py
@@ -49,4 +49,4 @@ def test_pruning_transformations():
 
     f = ng.function_from_cnn(net)
     assert f != None
-    assert len(f.get_ops()) == 3
\ No newline at end of file
+    assert len(f.get_ops()) == 3
diff --git a/inference-engine/include/ie_transformations.hpp b/inference-engine/include/ie_transformations.hpp
index 6691fa74daeedd..0867b02972c471 100644
--- a/inference-engine/include/ie_transformations.hpp
+++ b/inference-engine/include/ie_transformations.hpp
@@ -52,5 +52,41 @@ namespace InferenceEngine {
  * @param network A network to apply LowLatency transformation
  * *
  */
+
+INFERENCE_ENGINE_DEPRECATED("This transformation will be removed in 2023.1. "
+                            "Use InferenceEngine::lowLatency2 instead.")
 INFERENCE_ENGINE_API_CPP(void) LowLatency(InferenceEngine::CNNNetwork& network);
+
+
+/**
+ * @brief The transformation finds all TensorIterator/Loop layers in the network,
+ * processes all back edges that describe a connection between Result and Parameter
+ * of the TensorIterator/Loop bodies,and inserts ReadValue and Assign layers at the
+ * input and output corresponding to this back edge.
+ * Supported platforms: CPU, GNA.
+ *
+ * The example below describes the changes made by the transformation
+ *  [] - TensorIterator body
+ *  () - new layer
+ *  BE - back-edge
+ *
+ *  before applying the transformation:
+ *  -> input1[BE_1 -> Parameter -> Layers ... -> Result  -> BE_1 ]output1->
+ *
+ *  after applying the transformation:
+ *  ->(ReadValue)-> input1[BE_1 ->Parameter->Layers ...->Result->BE_1]output1 ->(Assign)
+ *                                                                      \
+ *                                                                       ->...
+ * After applying the transformation, the resulting network can be inferred
+ * step by step, the states will store between inferences.
+ * @param network A network to apply LowLatency transformation
+ * @param use_const_initializer Changes the type of the initializing subgraph for ReadValue operations.
+          If "true", then the transformation inserts Constant before ReadValue operation.
+          If "false, then the transformation leaves existed initializing subgraph for ReadValue operation.
+ * Loop operation by a given number. Does not affect TensorIterators.
+ * *
+ */
+INFERENCE_ENGINE_API_CPP(void) lowLatency2(InferenceEngine::CNNNetwork& network,
+                                           bool use_const_initializer = true);
+
 } // namespace InferenceEngine
diff --git a/inference-engine/src/inference_engine/ie_transformations.cpp b/inference-engine/src/inference_engine/ie_transformations.cpp
index 15360ae97ea09a..2a87671ce25d4f 100644
--- a/inference-engine/src/inference_engine/ie_transformations.cpp
+++ b/inference-engine/src/inference_engine/ie_transformations.cpp
@@ -11,6 +11,16 @@ using namespace InferenceEngine;
 void InferenceEngine::LowLatency(InferenceEngine::CNNNetwork &network) {
     auto function = network.getFunction();
     ngraph::pass::Manager manager;
+    NGRAPH_SUPPRESS_DEPRECATED_START
     manager.register_pass<ngraph::pass::LowLatency>();
+    NGRAPH_SUPPRESS_DEPRECATED_END
+    manager.run_passes(function);
+}
+
+void InferenceEngine::lowLatency2(InferenceEngine::CNNNetwork &network,
+                                  bool use_const_initializer) {
+    auto function = network.getFunction();
+    ngraph::pass::Manager manager;
+    manager.register_pass<ngraph::pass::LowLatency2>(use_const_initializer);
     manager.run_passes(function);
 }
diff --git a/inference-engine/tests/functional/inference_engine/transformations/low_latency_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/low_latency_test.cpp
index 69cb89991a3ed7..2d1594d38411db 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/low_latency_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/low_latency_test.cpp
@@ -68,7 +68,9 @@ TEST(TransformationTests, LowLatencyLSTM) {
 
         ngraph::pass::Manager manager;
         manager.register_pass<ngraph::pass::InitNodeInfo>();
+        NGRAPH_SUPPRESS_DEPRECATED_START
         manager.register_pass<ngraph::pass::LowLatency>();
+        NGRAPH_SUPPRESS_DEPRECATED_END
         manager.register_pass<ngraph::pass::UnrollTensorIterator>();
         manager.run_passes(f);
     }
@@ -149,7 +151,9 @@ TEST(TransformationTests, LowLatencyGRU) {
 
         ngraph::pass::Manager manager;
         manager.register_pass<ngraph::pass::InitNodeInfo>();
+        NGRAPH_SUPPRESS_DEPRECATED_START
         manager.register_pass<ngraph::pass::LowLatency>();
+        NGRAPH_SUPPRESS_DEPRECATED_END
         manager.register_pass<ngraph::pass::UnrollTensorIterator>();
         manager.run_passes(f);
 
@@ -227,7 +231,9 @@ TEST(TransformationTests, LowLatencyRNN) {
 
         ngraph::pass::Manager manager;
         manager.register_pass<ngraph::pass::InitNodeInfo>();
+        NGRAPH_SUPPRESS_DEPRECATED_START
         manager.register_pass<ngraph::pass::LowLatency>();
+        NGRAPH_SUPPRESS_DEPRECATED_END
         manager.register_pass<ngraph::pass::UnrollTensorIterator>();
         manager.run_passes(f);
 
@@ -317,7 +323,9 @@ TEST(TransformationTests, LowLatencyLSTMReshape) {
 
         ngraph::pass::Manager manager;
         manager.register_pass<ngraph::pass::InitNodeInfo>();
+        NGRAPH_SUPPRESS_DEPRECATED_START
         manager.register_pass<ngraph::pass::LowLatency>();
+        NGRAPH_SUPPRESS_DEPRECATED_END
         manager.register_pass<ngraph::pass::UnrollTensorIterator>();
         manager.run_passes(f);
     }
@@ -413,7 +421,9 @@ TEST(TransformationTests, LowLatencyLSTM_Loop) {
 
         ngraph::pass::Manager manager;
         manager.register_pass<ngraph::pass::InitNodeInfo>();
+        NGRAPH_SUPPRESS_DEPRECATED_START
         manager.register_pass<ngraph::pass::LowLatency>();
+        NGRAPH_SUPPRESS_DEPRECATED_END
         manager.register_pass<ngraph::pass::UnrollTensorIterator>();
         manager.run_passes(f);
     }
diff --git a/inference-engine/tests/functional/inference_engine/transformations/low_latency_v2_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/low_latency_v2_test.cpp
new file mode 100644
index 00000000000000..5c66e8109ef688
--- /dev/null
+++ b/inference-engine/tests/functional/inference_engine/transformations/low_latency_v2_test.cpp
@@ -0,0 +1,829 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <gtest/gtest.h>
+
+#include <string>
+#include <memory>
+#include <queue>
+
+#include <ngraph/function.hpp>
+#include <ngraph/opsets/opset7.hpp>
+#include <ngraph/pass/manager.hpp>
+
+#include <transformations/control_flow/unroll_tensor_iterator.hpp>
+#include <transformations/init_node_info.hpp>
+#include <transformations/common_optimizations/low_latency.hpp>
+#include <transformations/serialize.hpp>
+
+#include "common_test_utils/ngraph_test_utils.hpp"
+
+using namespace testing;
+using namespace ngraph;
+using namespace opset7;
+using namespace std;
+
+Output<Node> create_init_subgraph(const Output<Node>& in_node) {
+    auto const_zero = make_shared<Constant>(in_node.get_element_type(), Shape{1}, 0);
+    auto shape_of = make_shared<ShapeOf>(in_node);
+    auto broadcast = make_shared<Broadcast>(const_zero, shape_of);
+    return broadcast->output(0);
+}
+
+Output<Node> insert_identity(const Output<Node>& in_node) {
+    auto axis_1 = Constant::create(element::i64, Shape{1}, {1});
+    auto identity_1 = std::make_shared<Unsqueeze>(in_node, axis_1);
+    return std::make_shared<Squeeze>(identity_1, axis_1);
+}
+
+std::shared_ptr<Function> createLSTMBody(const std::shared_ptr<Parameter>& Xi,
+                                                 const std::shared_ptr<Parameter>& H_t,
+                                                 const std::shared_ptr<Parameter>& C_t,
+                                                 bool is_loop = false) {
+    // Body
+    auto axis = Constant::create(element::i64, Shape{}, {0});
+    auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+    auto w_val = std::vector<float>(512 * 16, 0);
+    auto r_val = std::vector<float>(512 * 128, 0);
+    auto b_val = std::vector<float>(512, 0);
+    auto W = Constant::create(element::f32, Shape{512, 16}, w_val);
+    auto R = Constant::create(element::f32, Shape{512, 128}, r_val);
+    auto B = Constant::create(element::f32, Shape{512}, b_val);
+
+    auto lstm_cell = std::make_shared<LSTMCell>(squeeze, H_t, C_t, W, R, B, 128);
+    auto res_1 = std::make_shared<Result>(lstm_cell->output(0));
+    auto unsqueeze = std::make_shared<Unsqueeze>(lstm_cell->output(0), axis);
+    auto res_2 = std::make_shared<Result>(unsqueeze);
+    auto res_3 = std::make_shared<Result>(lstm_cell->output(1));
+
+    auto func = std::make_shared<Function>(OutputVector{res_1, res_2, res_3},
+                                           ParameterVector{Xi, H_t, C_t});
+    if (is_loop) {
+        auto body_condition = std::make_shared<Constant>(
+                element::boolean, Shape{1}, true);
+        auto cond_res = std::make_shared<Result>(body_condition);
+        func->add_results({cond_res});
+    }
+    return func;
+}
+
+TEST(TransformationTests, LowLatency2_LSTM) {
+    std::shared_ptr<Function> f(nullptr), f_ref(nullptr);
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_init = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_init = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto body = createLSTMBody(Xi, H_t, C_t);
+        auto results = body->get_results();
+
+        auto tensor_iterator = std::make_shared<TensorIterator>();
+        tensor_iterator->set_body(body);
+        tensor_iterator->set_friendly_name("LSTMTensorIterator");
+
+        tensor_iterator->set_merged_input(C_t, C_init, results[2]);
+        tensor_iterator->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        tensor_iterator->set_merged_input(H_t, H_init, results[0]);
+
+        tensor_iterator->get_iter_value(results[0], -1);
+        tensor_iterator->get_concatenated_slices(results[1], 0, 1, 1, -1, 0);
+
+        auto res_ti_1 = std::make_shared<Result>(tensor_iterator->output(1));
+        auto res_ti_2 = std::make_shared<Result>(tensor_iterator->output(0));
+        f = std::make_shared<Function>(NodeVector{res_ti_1, res_ti_2},
+                                               ParameterVector{X, H_init, C_init});
+
+        pass::Manager manager;
+        manager.register_pass<pass::InitNodeInfo>();
+        manager.register_pass<pass::LowLatency2>();
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+    {
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        const std::string variable_name_H("LSTMTensorIterator/variable0");
+        const std::string variable_name_C("LSTMTensorIterator/variable1");
+        auto variable_H = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_H});
+        auto variable_C = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_C});
+        auto read_value_H = std::make_shared<ReadValue>(create_init_subgraph(H_t), variable_H);
+        auto read_value_C = std::make_shared<ReadValue>(create_init_subgraph(C_t), variable_C);
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(512 * 16, 0);
+        auto r_val = std::vector<float>(512 * 128, 0);
+        auto b_val = std::vector<float>(512, 0);
+        auto W = Constant::create(element::f32, Shape{512, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{512, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{512}, b_val);
+
+        auto lstm_cell = std::make_shared<LSTMCell>(squeeze, read_value_H, read_value_C, W, R, B, 128);
+        auto assign_H = std::make_shared<Assign>(lstm_cell->output(0), variable_H);
+        auto assign_C = std::make_shared<Assign>(lstm_cell->output(1), variable_C);
+        auto unsqueeze = std::make_shared<Unsqueeze>(lstm_cell->output(0), axis);
+        auto res_2 = std::make_shared<Result>(insert_identity(unsqueeze));
+        auto res_1 = std::make_shared<Result>(insert_identity(lstm_cell->output(0)));
+        f_ref = std::make_shared<Function>(OutputVector{res_1, res_2}, ParameterVector{Xi, H_t, C_t});
+        f_ref->add_sinks({assign_C, assign_H});
+        assign_H->add_control_dependency(read_value_H);
+        assign_C->add_control_dependency(read_value_C);
+    }
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, LowLatency2_GRU) {
+    std::shared_ptr<Function> f(nullptr), f_ref(nullptr);
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto Y = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto Yi = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(384 * 16, 0);
+        auto r_val = std::vector<float>(384 * 128, 0);
+        auto b_val = std::vector<float>(384, 0);
+        auto W = Constant::create(element::f32, Shape{384, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{384, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{384}, b_val);
+
+        auto gru_cell = std::make_shared<GRUCell>(squeeze, Yi, W, R, B, 128);
+        auto res_1 = std::make_shared<Result>(gru_cell);
+        auto unsqueeze = std::make_shared<Unsqueeze>(gru_cell, axis);
+        auto res_2 = std::make_shared<Result>(unsqueeze);
+        auto body = std::make_shared<Function>(OutputVector{res_1, res_2}, ParameterVector{Xi, Yi});
+
+        auto tensor_iterator = std::make_shared<TensorIterator>();
+        tensor_iterator->set_body(body);
+
+        tensor_iterator->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        tensor_iterator->set_merged_input(Yi, Y, res_1);
+
+        auto out0 = tensor_iterator->get_iter_value(res_1, -1);
+        auto out1 = tensor_iterator->get_concatenated_slices(res_2, 0, 1, 1, -1, 0);
+
+        auto res_ti_1 = std::make_shared<Result>(tensor_iterator->output(1));
+        f = std::make_shared<Function>(NodeVector{res_ti_1}, ParameterVector{X, Y});
+
+        pass::Manager manager;
+        manager.register_pass<pass::InitNodeInfo>();
+        manager.register_pass<pass::LowLatency2>();
+
+        manager.run_passes(f);
+
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+    {
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        const std::string variable_name_H("GRUTensorIterator/variable0");
+        auto variable_H = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_H});
+        auto read_value_H = std::make_shared<ReadValue>(create_init_subgraph(H_t), variable_H);
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(384 * 16, 0);
+        auto r_val = std::vector<float>(384 * 128, 0);
+        auto b_val = std::vector<float>(384, 0);
+        auto W = Constant::create(element::f32, Shape{384, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{384, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{384}, b_val);
+
+        auto rnn_cell = std::make_shared<GRUCell>(squeeze, read_value_H, W, R, B, 128);
+        auto assign_H = std::make_shared<Assign>(rnn_cell->output(0), variable_H);
+        auto res_1 = std::make_shared<Result>(assign_H);
+        auto unsqueeze = std::make_shared<Unsqueeze>(rnn_cell->output(0), axis);
+        auto res_2 = std::make_shared<Result>(insert_identity(unsqueeze));
+        f_ref = std::make_shared<Function>(ResultVector {res_2}, ParameterVector{Xi, H_t});
+        f_ref->add_sinks({assign_H});
+        assign_H->add_control_dependency(read_value_H);
+    }
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, LowLatency2_RNN) {
+    std::shared_ptr<Function> f(nullptr), f_ref(nullptr);
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto Y = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto Yi = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(128 * 16, 0);
+        auto r_val = std::vector<float>(128 * 128, 0);
+        auto b_val = std::vector<float>(128, 0);
+        auto W = Constant::create(element::f32, Shape{128, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{128, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{128}, b_val);
+
+        auto rnn_cell = std::make_shared<RNNCell>(squeeze, Yi, W, R, B, 128);
+        auto res_1 = std::make_shared<Result>(rnn_cell);
+        auto unsqueeze = std::make_shared<Unsqueeze>(rnn_cell, axis);
+        auto res_2 = std::make_shared<Result>(unsqueeze);
+        auto body = std::make_shared<Function>(OutputVector{res_1, res_2}, ParameterVector{Xi,
+                                                                                                   Yi});
+
+        auto tensor_iterator = std::make_shared<TensorIterator>();
+        tensor_iterator->set_body(body);
+
+        tensor_iterator->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        tensor_iterator->set_merged_input(Yi, Y, res_1);
+
+        auto out0 = tensor_iterator->get_iter_value(res_1, -1);
+        auto out1 = tensor_iterator->get_concatenated_slices(res_2, 0, 1, 1, -1, 0);
+
+        auto res_ti_1 = std::make_shared<Result>(tensor_iterator->output(1));
+        f = std::make_shared<Function>(NodeVector{res_ti_1}, ParameterVector{X, Y});
+
+        pass::Manager manager;
+        manager.register_pass<pass::InitNodeInfo>();
+        manager.register_pass<pass::LowLatency2>();
+
+        manager.run_passes(f);
+
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+    {
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        const std::string variable_name_H("RNNTensorIterator/variable0");
+        auto variable_H = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_H});
+        auto read_value_H = std::make_shared<ReadValue>(create_init_subgraph(H_t), variable_H);
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(128 * 16, 0);
+        auto r_val = std::vector<float>(128 * 128, 0);
+        auto b_val = std::vector<float>(128, 0);
+        auto W = Constant::create(element::f32, Shape{128, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{128, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{128}, b_val);
+
+        auto rnn_cell = std::make_shared<RNNCell>(squeeze, read_value_H, W, R, B, 128);
+        auto assign_H = std::make_shared<Assign>(rnn_cell->output(0), variable_H);
+        auto res_1 = std::make_shared<Result>(assign_H);
+        auto unsqueeze = std::make_shared<Unsqueeze>(rnn_cell->output(0), axis);
+        auto res_2 = std::make_shared<Result>(insert_identity(unsqueeze));
+        f_ref = std::make_shared<Function>(ResultVector{res_2}, ParameterVector{Xi, H_t});
+        f_ref->add_sinks({assign_H});
+        assign_H->add_control_dependency(read_value_H);
+    }
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, LowLatency2_LSTMReshape) {
+    std::shared_ptr<Function> f(nullptr), f_ref(nullptr);
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{2, 1, 16});
+        auto H = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto body = createLSTMBody(Xi, H_t, C_t);
+        auto results = body->get_results();
+
+        auto tensor_iterator = std::make_shared<TensorIterator>();
+        tensor_iterator->set_body(body);
+
+        tensor_iterator->set_merged_input(C_t, C, results[2]);
+        tensor_iterator->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        tensor_iterator->set_merged_input(H_t, H, results[0]);
+
+        auto out0 = tensor_iterator->get_iter_value(results[0], -1);
+        auto out1 = tensor_iterator->get_concatenated_slices(results[1], 0, 1, 1, -1, 0);
+
+        auto res_ti_1 = std::make_shared<Result>(tensor_iterator->output(1));
+        auto res_ti_2 = std::make_shared<Result>(tensor_iterator->output(0));
+        f = std::make_shared<Function>(NodeVector{res_ti_1, res_ti_2}, ParameterVector{X, H,
+                                                                                                               C});
+
+        // Reshape
+        // change the number of iteration of TI. 2 -> 1
+        auto new_X = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        f->replace_parameter(0, new_X);
+        f->validate_nodes_and_infer_types();
+
+        pass::Manager manager;
+        manager.register_pass<pass::InitNodeInfo>();
+        manager.register_pass<pass::LowLatency2>();
+
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+    {
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        const std::string variable_name_H("LSTMTensorIterator/variable0");
+        const std::string variable_name_C("LSTMTensorIterator/variable1");
+        auto variable_H = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_H});
+        auto variable_C = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_C});
+        auto read_value_H = std::make_shared<ReadValue>(create_init_subgraph(H_t), variable_H);
+        auto read_value_C = std::make_shared<ReadValue>(create_init_subgraph(C_t), variable_C);
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(512 * 16, 0);
+        auto r_val = std::vector<float>(512 * 128, 0);
+        auto b_val = std::vector<float>(512, 0);
+        auto W = Constant::create(element::f32, Shape{512, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{512, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{512}, b_val);
+
+        auto lstm_cell = std::make_shared<LSTMCell>(squeeze, read_value_H, read_value_C, W, R, B, 128);
+        auto assign_H = std::make_shared<Assign>(lstm_cell->output(0), variable_H);
+        auto assign_C = std::make_shared<Assign>(lstm_cell->output(1), variable_C);
+        auto unsqueeze = std::make_shared<Unsqueeze>(lstm_cell->output(0), axis);
+        auto res_2 = std::make_shared<Result>(insert_identity(unsqueeze));
+        auto res_1 = std::make_shared<Result>(insert_identity(lstm_cell->output(0)));
+        f_ref = std::make_shared<Function>(OutputVector{res_1, res_2}, ParameterVector{Xi, H_t, C_t});
+        f_ref->add_sinks({assign_C, assign_H});
+        assign_H->add_control_dependency(read_value_H);
+        assign_C->add_control_dependency(read_value_C);
+    }
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, LowLatency2_LSTM_Loop) {
+    std::shared_ptr<Function> f(nullptr), f_ref(nullptr);
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_init = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_init = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        // Body
+        auto body = createLSTMBody(Xi, H_t, C_t, true);
+        auto results = body->get_results();
+
+        auto trip_count =
+                std::make_shared<Constant>(element::i64, Shape{}, 1);
+        auto exec_condition =
+                std::make_shared<Constant>(element::boolean, Shape{}, true);
+        auto loop = std::make_shared<Loop>(trip_count, exec_condition);
+        loop->set_special_body_ports({-1, 3});
+        loop->set_function(body);
+        loop->set_friendly_name("LSTMLoop");
+
+        loop->set_merged_input(C_t, C_init, results[2]);
+        loop->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        loop->set_merged_input(H_t, H_init, results[0]);
+
+        auto out0 = loop->get_iter_value(results[0], -1);
+        auto out1 = loop->get_concatenated_slices(results[1], 0, 1, 1, -1, 0);
+
+        auto res_ti_1 = std::make_shared<Result>(loop->output(1));
+        auto res_ti_2 = std::make_shared<Result>(loop->output(0));
+        f = std::make_shared<Function>(NodeVector{res_ti_1, res_ti_2},
+                                               ParameterVector{X, H_init, C_init});
+
+        pass::Manager manager;
+        manager.register_pass<pass::InitNodeInfo>();
+        manager.register_pass<pass::LowLatency2>();
+
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+    {
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        const std::string variable_name_H("LSTMTensorIterator/variable0");
+        const std::string variable_name_C("LSTMTensorIterator/variable1");
+        auto variable_H = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_H});
+        auto variable_C = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_C});
+        auto read_value_H = std::make_shared<ReadValue>(create_init_subgraph(H_t), variable_H);
+        auto read_value_C = std::make_shared<ReadValue>(create_init_subgraph(C_t), variable_C);
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(512 * 16, 0);
+        auto r_val = std::vector<float>(512 * 128, 0);
+        auto b_val = std::vector<float>(512, 0);
+        auto W = Constant::create(element::f32, Shape{512, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{512, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{512}, b_val);
+
+        auto lstm_cell = std::make_shared<LSTMCell>(squeeze, read_value_H, read_value_C, W, R, B, 128);
+        auto assign_H = std::make_shared<Assign>(lstm_cell->output(0), variable_H);
+        auto assign_C = std::make_shared<Assign>(lstm_cell->output(1), variable_C);
+        auto unsqueeze = std::make_shared<Unsqueeze>(lstm_cell->output(0), axis);
+        auto res_2 = std::make_shared<Result>(insert_identity(unsqueeze));
+        auto res_1 = std::make_shared<Result>(insert_identity(lstm_cell->output(0)));
+        f_ref = std::make_shared<Function>(OutputVector{res_1, res_2}, ParameterVector{Xi, H_t, C_t});
+        f_ref->add_sinks({assign_C, assign_H});
+        assign_H->add_control_dependency(read_value_H);
+        assign_C->add_control_dependency(read_value_C);
+    }
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, LowLatency2_LSTM_several_iterations) {
+    constexpr int ITER_CNT = 5;
+    std::shared_ptr<Function> f(nullptr), f_ref(nullptr);
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{ITER_CNT, 1, 16});
+        auto H = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto body = createLSTMBody(Xi, H_t, C_t);
+        auto results = body->get_results();
+
+        auto tensor_iterator = std::make_shared<TensorIterator>();
+        tensor_iterator->set_body(body);
+
+        tensor_iterator->set_merged_input(C_t, C, results[2]);
+        tensor_iterator->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        tensor_iterator->set_merged_input(H_t, H, results[0]);
+
+        auto out0 = tensor_iterator->get_iter_value(results[0], -1);
+        auto out1 = tensor_iterator->get_concatenated_slices(results[1], 0, 1, 1, -1, 0);
+
+        auto res_ti_1 = std::make_shared<Result>(tensor_iterator->output(1));
+        auto res_ti_2 = std::make_shared<Result>(tensor_iterator->output(0));
+        f = std::make_shared<Function>(NodeVector{res_ti_1, res_ti_2}, ParameterVector{X, H,
+                                                                                       C});
+
+        pass::Manager manager;
+        manager.register_pass<pass::InitNodeInfo>();
+        manager.register_pass<pass::LowLatency2>();
+
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+
+    // TensorIterator not unrolled.
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{ITER_CNT, 1, 16});
+        auto H = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        const std::string variable_name_H("LSTMTensorIterator/variable0");
+        const std::string variable_name_C("LSTMTensorIterator/variable1");
+        auto variable_H = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_H});
+        auto variable_C = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_C});
+        auto read_value_H = std::make_shared<ReadValue>(create_init_subgraph(H), variable_H);
+        auto read_value_C = std::make_shared<ReadValue>(create_init_subgraph(C), variable_C);
+
+        // Body
+
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(512 * 16, 0);
+        auto r_val = std::vector<float>(512 * 128, 0);
+        auto b_val = std::vector<float>(512, 0);
+        auto W = Constant::create(element::f32, Shape{512, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{512, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{512}, b_val);
+
+        auto lstm_cell = std::make_shared<LSTMCell>(squeeze, H_t, C_t, W, R, B, 128);
+        auto res_1 = std::make_shared<Result>(lstm_cell->output(0));
+        auto unsqueeze = std::make_shared<Unsqueeze>(lstm_cell, axis);
+        auto res_2 = std::make_shared<Result>(unsqueeze);
+        auto res_3 = std::make_shared<Result>(lstm_cell->output(1));
+        auto body = std::make_shared<Function>(OutputVector{res_1, res_2, res_3},
+                                               ParameterVector{Xi, H_t, C_t});
+
+        auto tensor_iterator = std::make_shared<TensorIterator>();
+        tensor_iterator->set_body(body);
+
+        tensor_iterator->set_merged_input(C_t, read_value_C, res_3);
+        tensor_iterator->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        tensor_iterator->set_merged_input(H_t, read_value_H, res_1);
+
+        auto out0 = tensor_iterator->get_iter_value(res_1, -1);
+        auto out1 = tensor_iterator->get_concatenated_slices(res_2, 0, 1, 1, -1, 0);
+        auto out2 = tensor_iterator->get_iter_value(res_3, -1);
+
+        auto assign_H = std::make_shared<Assign>(out0, variable_H);
+        auto assign_C = std::make_shared<Assign>(out2, variable_C);
+        auto outer_res_2 = std::make_shared<Result>(out1);
+        auto outer_res_1 = std::make_shared<Result>(out0);
+        f_ref = std::make_shared<Function>(OutputVector{outer_res_1, outer_res_2}, ParameterVector{X, H, C});
+        f_ref->add_sinks({assign_C, assign_H});
+        assign_H->add_control_dependency(read_value_H);
+        assign_C->add_control_dependency(read_value_C);
+    }
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, LowLatency2_LSTM_Loop_Reshape) {
+    std::shared_ptr<Function> f(nullptr), f_ref(nullptr);
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{10, 1, 16});
+        auto H_init = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_init = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto body = createLSTMBody(Xi, H_t, C_t, true);
+        auto results = body->get_results();
+
+        auto shape_of = std::make_shared<ShapeOf>(X);
+        const auto trip_count = std::make_shared<Gather>(shape_of, Constant::create(ngraph::element::i64, {1}, {0}),
+                                                         Constant::create(ngraph::element::i64, {1}, {0}));
+        auto exec_condition =
+                std::make_shared<Constant>(element::boolean, Shape{}, true);
+        auto loop = std::make_shared<Loop>(trip_count, exec_condition);
+        loop->set_special_body_ports({-1, 3});
+        loop->set_function(body);
+        loop->set_friendly_name("LSTMLoop");
+
+        loop->set_merged_input(C_t, C_init, results[2]);
+        loop->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        loop->set_merged_input(H_t, H_init, results[0]);
+
+        auto out0 = loop->get_iter_value(results[0], -1);
+        auto out1 = loop->get_concatenated_slices(results[1], 0, 1, 1, -1, 0);
+
+        auto res_ti_1 = std::make_shared<Result>(loop->output(1));
+        auto res_ti_2 = std::make_shared<Result>(loop->output(0));
+        f = std::make_shared<Function>(NodeVector{res_ti_1, res_ti_2},
+                                       ParameterVector{X, H_init, C_init});
+
+        // Reshape
+        // change the number of iteration of Loop. 10 -> 1
+        auto new_X = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        f->replace_parameter(0, new_X);
+        f->validate_nodes_and_infer_types();
+
+        pass::Manager manager;
+        manager.register_pass<pass::InitNodeInfo>();
+        manager.register_pass<pass::LowLatency2>();
+
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+    {
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        const std::string variable_name_H("LSTMTensorIterator/variable0");
+        const std::string variable_name_C("LSTMTensorIterator/variable1");
+        auto variable_H = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_H});
+        auto variable_C = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_C});
+        auto read_value_H = std::make_shared<ReadValue>(create_init_subgraph(H_t), variable_H);
+        auto read_value_C = std::make_shared<ReadValue>(create_init_subgraph(C_t), variable_C);
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(512 * 16, 0);
+        auto r_val = std::vector<float>(512 * 128, 0);
+        auto b_val = std::vector<float>(512, 0);
+        auto W = Constant::create(element::f32, Shape{512, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{512, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{512}, b_val);
+
+        auto lstm_cell = std::make_shared<LSTMCell>(squeeze, read_value_H, read_value_C, W, R, B, 128);
+        auto assign_H = std::make_shared<Assign>(lstm_cell->output(0), variable_H);
+        auto assign_C = std::make_shared<Assign>(lstm_cell->output(1), variable_C);
+        auto unsqueeze = std::make_shared<Unsqueeze>(lstm_cell->output(0), axis);
+        auto res_2 = std::make_shared<Result>(insert_identity(unsqueeze));
+        auto res_1 = std::make_shared<Result>(insert_identity(lstm_cell->output(0)));
+        f_ref = std::make_shared<Function>(OutputVector{res_1, res_2}, ParameterVector{Xi, H_t, C_t});
+        f_ref->add_sinks({assign_C, assign_H});
+        assign_H->add_control_dependency(read_value_H);
+        assign_C->add_control_dependency(read_value_C);
+    }
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+
+TEST(TransformationTests, LowLatency2_LSTM_Loop_several_iterations) {
+    std::shared_ptr<Function> f(nullptr), f_ref(nullptr);
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{10, 1, 16});
+        auto H_init = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_init = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto body = createLSTMBody(Xi, H_t, C_t, true);
+        auto results = body->get_results();
+
+        auto trip_count =
+                std::make_shared<Constant>(element::i64, Shape{}, 10);
+        auto exec_condition =
+                std::make_shared<Constant>(element::boolean, Shape{}, true);
+        auto loop = std::make_shared<Loop>(trip_count, exec_condition);
+        loop->set_special_body_ports({-1, 3});
+        loop->set_function(body);
+        loop->set_friendly_name("LSTMLoop");
+
+        loop->set_merged_input(C_t, C_init, results[2]);
+        loop->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        loop->set_merged_input(H_t, H_init, results[0]);
+
+        auto out0 = loop->get_iter_value(results[0], -1);
+        auto out1 = loop->get_concatenated_slices(results[1], 0, 1, 1, -1, 0);
+
+        auto res_ti_1 = std::make_shared<Result>(loop->output(1));
+        auto res_ti_2 = std::make_shared<Result>(loop->output(0));
+        f = std::make_shared<Function>(NodeVector{res_ti_1, res_ti_2},
+                                       ParameterVector{X, H_init, C_init});
+
+        pass::Manager manager;
+        manager.register_pass<pass::InitNodeInfo>();
+        manager.register_pass<pass::LowLatency2>(true);
+
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{10, 1, 16});
+        auto H = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        const std::string variable_name_H("LSTMTensorIterator/variable0");
+        const std::string variable_name_C("LSTMTensorIterator/variable1");
+        auto variable_H = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_H});
+        auto variable_C = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_C});
+        auto read_value_H = std::make_shared<ReadValue>(create_init_subgraph(H), variable_H);
+        auto read_value_C = std::make_shared<ReadValue>(create_init_subgraph(C), variable_C);
+
+        // Body
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(512 * 16, 0);
+        auto r_val = std::vector<float>(512 * 128, 0);
+        auto b_val = std::vector<float>(512, 0);
+        auto W = Constant::create(element::f32, Shape{512, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{512, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{512}, b_val);
+
+        auto lstm_cell = std::make_shared<LSTMCell>(squeeze, H_t, C_t, W, R, B, 128);
+        auto res_1 = std::make_shared<Result>(lstm_cell->output(0));
+        auto unsqueeze = std::make_shared<Unsqueeze>(lstm_cell->output(0), axis);
+        auto res_2 = std::make_shared<Result>(unsqueeze);
+        auto res_3 = std::make_shared<Result>(lstm_cell->output(1));
+        auto body_condition = std::make_shared<Constant>(
+                element::boolean, Shape{1}, true);
+        auto body = std::make_shared<Function>(OutputVector{res_1, res_2, res_3, body_condition},
+                                               ParameterVector{Xi, H_t, C_t});
+
+        auto trip_count =
+                std::make_shared<Constant>(element::i64, Shape{}, 10);
+        auto exec_condition =
+                std::make_shared<Constant>(element::boolean, Shape{}, true);
+        auto loop = std::make_shared<Loop>(trip_count, exec_condition);
+        loop->set_special_body_ports({-1, 3});
+        loop->set_function(body);
+        loop->set_friendly_name("LSTMLoop");
+
+        loop->set_merged_input(C_t, read_value_C, res_3);
+        loop->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        loop->set_merged_input(H_t, read_value_H, res_1);
+
+        auto out0 = loop->get_iter_value(res_1, -1);
+        auto out1 = loop->get_concatenated_slices(res_2, 0, 1, 1, -1, 0);
+        auto out3 = loop->get_iter_value(res_3, -1);
+
+        auto assign_H = std::make_shared<Assign>(out0, variable_H);
+        auto assign_C = std::make_shared<Assign>(out3, variable_C);
+        auto outer_res_2 = std::make_shared<Result>(out1);
+        auto outer_res_1 = std::make_shared<Result>(out0);
+        f_ref = std::make_shared<Function>(OutputVector{outer_res_1, outer_res_2}, ParameterVector{X, H, C});
+        f_ref->add_sinks({assign_C, assign_H});
+        assign_H->add_control_dependency(read_value_H);
+        assign_C->add_control_dependency(read_value_C);
+    }
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, LowLatencyLSTM_LLTv1_LLTv2) {
+    std::shared_ptr<Function> f(nullptr), f_ref(nullptr);
+    {
+        auto X = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_init = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_init = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        auto Xi = std::make_shared<Parameter>(element::f32, Shape{1, 1, 16});
+        auto H_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+        auto C_t = std::make_shared<Parameter>(element::f32, Shape{1, 128});
+
+        // Body
+        auto axis = Constant::create(element::i64, Shape{}, {0});
+        auto squeeze = std::make_shared<Squeeze>(Xi, axis);
+
+        auto w_val = std::vector<float>(512 * 16, 0);
+        auto r_val = std::vector<float>(512 * 128, 0);
+        auto b_val = std::vector<float>(512, 0);
+        auto W = Constant::create(element::f32, Shape{512, 16}, w_val);
+        auto R = Constant::create(element::f32, Shape{512, 128}, r_val);
+        auto B = Constant::create(element::f32, Shape{512}, b_val);
+
+        auto lstm_cell = std::make_shared<LSTMCell>(squeeze, H_t, C_t, W, R, B, 128);
+        auto res_1 = std::make_shared<Result>(lstm_cell->output(0));
+        auto unsqueeze = std::make_shared<Unsqueeze>(lstm_cell->output(0), axis);
+        auto res_2 = std::make_shared<Result>(unsqueeze);
+        auto res_3 = std::make_shared<Result>(lstm_cell->output(1));
+        auto body = std::make_shared<Function>(OutputVector{res_1, res_2, res_3}, ParameterVector{Xi, H_t, C_t});
+
+        auto tensor_iterator = std::make_shared<TensorIterator>();
+        tensor_iterator->set_body(body);
+        tensor_iterator->set_friendly_name("LSTMTensorIterator");
+
+        tensor_iterator->set_merged_input(C_t, C_init, res_3);
+        tensor_iterator->set_sliced_input(Xi, X, 0, 1, 1, -1, 0);
+        tensor_iterator->set_merged_input(H_t, H_init, res_1);
+
+        auto out0 = tensor_iterator->get_iter_value(res_1, -1);
+        auto out1 = tensor_iterator->get_concatenated_slices(res_2, 0, 1, 1, -1, 0);
+
+        auto res_ti_1 = std::make_shared<Result>(tensor_iterator->output(1));
+        auto res_ti_2 = std::make_shared<Result>(tensor_iterator->output(0));
+        f = std::make_shared<Function>(NodeVector{res_ti_1, res_ti_2},
+                                               ParameterVector{X, H_init, C_init});
+
+        auto f_2 = ngraph::clone_function(*f);
+        pass::Manager manager_2;
+        manager_2.register_pass<pass::InitNodeInfo>();
+        NGRAPH_SUPPRESS_DEPRECATED_START
+        manager_2.register_pass<ngraph::pass::LowLatency>();
+        NGRAPH_SUPPRESS_DEPRECATED_END
+        EXPECT_NO_THROW(manager_2.run_passes(f_2));
+
+        pass::Manager manager;
+        manager.register_pass<pass::InitNodeInfo>();
+        NGRAPH_SUPPRESS_DEPRECATED_START
+        manager.register_pass<ngraph::pass::LowLatency>();
+        NGRAPH_SUPPRESS_DEPRECATED_END
+        // LLT v2 doesn't insert Assign/ReadValue ops, they are already inserted
+        // but unrolls TI/Loop
+        manager.register_pass<pass::LowLatency2>();
+
+        EXPECT_NO_THROW(manager.run_passes(f));
+    }
+}
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp
index 750a0e4af5e430..9ab20c3eda4cf7 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp
@@ -10,6 +10,13 @@ using namespace LayerTestsDefinitions;
 
 namespace {
 
+std::vector<ngraph::helpers::MemoryTransformation> transformation {
+        ngraph::helpers::MemoryTransformation::NONE,
+        ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2,
+        ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_REGULAR_API,
+        ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_ORIGINAL_INIT,
+};
+
 const std::vector<InferenceEngine::SizeVector> inShapes = {
         {3},
         {100, 100},
@@ -27,6 +34,7 @@ const std::vector<int64_t> iterationCount {
 
 INSTANTIATE_TEST_CASE_P(smoke_MemoryTest, MemoryTest,
         ::testing::Combine(
+                ::testing::ValuesIn(transformation),
                 ::testing::ValuesIn(iterationCount),
                 ::testing::ValuesIn(inShapes),
                 ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
new file mode 100644
index 00000000000000..8e9754343474cb
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
@@ -0,0 +1,45 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <subgraph_tests/memory_LSTMCell.hpp>
+#include "common_test_utils/test_constants.hpp"
+
+namespace SubgraphTestsDefinitions {
+    std::vector<ngraph::helpers::MemoryTransformation> transformation {
+            ngraph::helpers::MemoryTransformation::NONE,
+            ngraph::helpers::MemoryTransformation::LOW_LATENCY,
+            ngraph::helpers::MemoryTransformation::LOW_LATENCY_REGULAR_API,
+            ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2,
+            ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_REGULAR_API
+    };
+
+    std::vector<size_t> input_sizes = {
+            80,
+            32,
+            64,
+            100,
+            25
+    };
+
+    std::vector<size_t> hidden_sizes = {
+            128,
+            200,
+            300,
+            24,
+            32,
+    };
+
+    std::map<std::string, std::string> additional_config = {
+    };
+
+    INSTANTIATE_TEST_CASE_P(smoke_MemoryLSTMCellTest, MemoryLSTMCellTest,
+                            ::testing::Combine(
+                                    ::testing::ValuesIn(transformation),
+                                    ::testing::Values(CommonTestUtils::DEVICE_CPU),
+                                    ::testing::Values(InferenceEngine::Precision::FP32),
+                                    ::testing::ValuesIn(input_sizes),
+                                    ::testing::ValuesIn(hidden_sizes),
+                                    ::testing::Values(additional_config)),
+                            MemoryLSTMCellTest::getTestCaseName);
+} // namespace SubgraphTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
index 3468d2b417f744..4c18ee6ea054bb 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
@@ -7,6 +7,15 @@
 
 namespace SubgraphTestsDefinitions {
 namespace {
+
+std::vector<ngraph::helpers::MemoryTransformation> transformation {
+    ngraph::helpers::MemoryTransformation::NONE,
+    ngraph::helpers::MemoryTransformation::LOW_LATENCY,
+    ngraph::helpers::MemoryTransformation::LOW_LATENCY_REGULAR_API,
+    ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2,
+    ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_REGULAR_API
+};
+
 std::vector<size_t> input_sizes = {
     80,
     32,
@@ -28,6 +37,7 @@ std::map<std::string, std::string> additional_config = {
 
 INSTANTIATE_TEST_CASE_P(MultipleLSTMCellTest, MultipleLSTMCellTest,
     ::testing::Combine(
+        ::testing::ValuesIn(transformation),
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(InferenceEngine::Precision::FP32),
         ::testing::ValuesIn(input_sizes),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/memory.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/memory.cpp
index 2ab1357f6748e4..c04b76705ccac0 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/memory.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/memory.cpp
@@ -10,9 +10,17 @@ using namespace LayerTestsDefinitions;
 
 namespace {
 
+std::vector<ngraph::helpers::MemoryTransformation> transformation {
+        ngraph::helpers::MemoryTransformation::NONE,
+        ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2,
+        ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_REGULAR_API,
+        ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_ORIGINAL_INIT
+};
+
 const std::vector<InferenceEngine::SizeVector> inShapes = {
         {1, 1},
-        {1, 2}
+        {1, 2},
+        {1, 10}
 };
 
 const std::vector<InferenceEngine::Precision> inputPrecisions = {
@@ -22,11 +30,13 @@ const std::vector<InferenceEngine::Precision> inputPrecisions = {
 const std::vector<int64_t> iterationCount {
     1,
     3,
+    4,
     10
 };
 
 INSTANTIATE_TEST_CASE_P(smoke_MemoryTest, MemoryTest,
                         ::testing::Combine(
+                                ::testing::ValuesIn(transformation),
                                 ::testing::ValuesIn(iterationCount),
                                 ::testing::ValuesIn(inShapes),
                                 ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/skip_tests_config.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/skip_tests_config.cpp
index 9f2c05ab6d1690..cb4cc459a95e28 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/skip_tests_config.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/skip_tests_config.cpp
@@ -64,5 +64,13 @@ std::vector<std::string> disabledTestPatterns() {
         R"(.*CachingSupport.*_batch2_.*)",
         // TODO: Issue 51525
         R"(.*CachingSupport.*KSOFunction.*)",
+        // TODO: Issue 57363 (Param -> Result subgraphs)
+        R"(.*smoke_MemoryTest.*LOW_LATENCY.*iteration_count=1_.*)",
+        // TODO: Issue 57368 (accuracy)
+        R"(.*smoke_MemoryTest.*LOW_LATENCY.*IS=\(1.10\).*)",
+        R"(.*smoke_MemoryTest.*iteration_count=3.*IS=\(1.10\).*)",
+        R"(.*smoke_MemoryTest.*iteration_count=4.*IS=\(1.10\).*)",
+        R"(.*smoke_MemoryTest.*iteration_count=10.*IS=\(1.10\).*)",
+        R"(.*smoke_MemoryTest.*LOW_LATENCY.*iteration_count=10.*IS=\(1.2\).*)",
     };
 }
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
index a8d651f6c807f6..5818a40d4ecaa2 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
@@ -6,6 +6,14 @@
 #include "common_test_utils/test_constants.hpp"
 
 namespace SubgraphTestsDefinitions {
+    std::vector<ngraph::helpers::MemoryTransformation> transformation {
+            ngraph::helpers::MemoryTransformation::NONE,
+            ngraph::helpers::MemoryTransformation::LOW_LATENCY,
+            ngraph::helpers::MemoryTransformation::LOW_LATENCY_REGULAR_API,
+            ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2,
+            ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_REGULAR_API
+    };
+
     std::vector<size_t> input_sizes = {
         80,
         32,
@@ -30,6 +38,7 @@ namespace SubgraphTestsDefinitions {
 
     INSTANTIATE_TEST_CASE_P(smoke_MemoryLSTMCellTest, MemoryLSTMCellTest,
         ::testing::Combine(
+            ::testing::ValuesIn(transformation),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
             ::testing::Values(InferenceEngine::Precision::FP32),
             ::testing::ValuesIn(input_sizes),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
index 2d94617991d3cd..d59c023773bb38 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
@@ -7,6 +7,15 @@
 
 namespace SubgraphTestsDefinitions {
 namespace {
+
+std::vector<ngraph::helpers::MemoryTransformation> transformation {
+        ngraph::helpers::MemoryTransformation::NONE,
+        ngraph::helpers::MemoryTransformation::LOW_LATENCY,
+        ngraph::helpers::MemoryTransformation::LOW_LATENCY_REGULAR_API,
+        ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2,
+        ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_REGULAR_API
+};
+
 std::vector<size_t> input_sizes = {
     80,
     32,
@@ -31,6 +40,7 @@ std::map<std::string, std::string> additional_config = {
 
 INSTANTIATE_TEST_CASE_P(MultipleLSTMCellTest, MultipleLSTMCellTest,
     ::testing::Combine(
+        ::testing::ValuesIn(transformation),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
         ::testing::Values(InferenceEngine::Precision::FP32),
         ::testing::ValuesIn(input_sizes),
diff --git a/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/basic_lstm.hpp b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/basic_lstm.hpp
index 17bf8d31cf056f..7ae1d21a886e4f 100644
--- a/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/basic_lstm.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/basic_lstm.hpp
@@ -39,7 +39,7 @@ TEST_P(Basic_LSTM_S, CompareWithRefImpl_LowLatencyTransformation) {
 
     // Apply LowLatency and UnrollTensorIterator transformations
     ngraph::pass::Manager manager;
-    manager.register_pass<ngraph::pass::LowLatency>(); // LowLatency enables UnrollTI
+    manager.register_pass<ngraph::pass::LowLatency2>(); // LowLatency enables UnrollTI
     manager.run_passes(function);
     LoadNetwork();
     IE_SUPPRESS_DEPRECATED_START
diff --git a/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/memory_LSTMCell.hpp b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/memory_LSTMCell.hpp
index 947f0fa1e1a515..a6f250301b62a6 100644
--- a/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/memory_LSTMCell.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/memory_LSTMCell.hpp
@@ -12,12 +12,4 @@ TEST_P(MemoryLSTMCellTest, CompareWithRefs) {
     Run();
 };
 
-TEST_P(MemoryLSTMCellTest, CompareWithRefs_LowLatencyTransformation) {
-    RunLowLatency();
-};
-
-TEST_P(MemoryLSTMCellTest, CompareWithRefs_LowLatencyRegularAPITransformation) {
-    RunLowLatency(true);
-};
-
 } // namespace SubgraphTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/multiple_LSTMCell.hpp b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/multiple_LSTMCell.hpp
index 5dd18ff446381e..0135c09e17048d 100644
--- a/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/multiple_LSTMCell.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/multiple_LSTMCell.hpp
@@ -12,12 +12,4 @@ TEST_P(MultipleLSTMCellTest, CompareWithRefs) {
     Run();
 };
 
-TEST_P(MultipleLSTMCellTest, CompareWithRefs_LowLatencyTransformation) {
-    RunLowLatency();
-};
-
-TEST_P(MultipleLSTMCellTest, CompareWithRefs_LowLatencyRegularAPITransformation) {
-    RunLowLatency(true);
-};
-
 } // namespace SubgraphTestsDefinitions
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/memory.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/memory.hpp
index ca16e30148e68f..59ad6c54e5a2fc 100644
--- a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/memory.hpp
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/memory.hpp
@@ -14,6 +14,7 @@
 namespace LayerTestsDefinitions {
 
 using MemoryTestParams = std::tuple<
+        ngraph::helpers::MemoryTransformation,   // Apply Memory transformation
         int64_t,                            // iterationCount
         InferenceEngine::SizeVector,        // inputShape
         InferenceEngine::Precision,         // netPrecision
@@ -28,9 +29,17 @@ class MemoryTest : public testing::WithParamInterface<MemoryTestParams>, virtual
     std::vector<std::pair<ngraph::element::Type, std::vector<std::uint8_t>>> CalculateRefs() override;
     void SetUp() override;
 private:
+    void CreateTIFunc();
+    void CreateCommonFunc();
+    void ApplyLowLatency();
+
     InferenceEngine::Precision netPrecision;
     ngraph::EvaluationContext eval_context;
+    ngraph::helpers::MemoryTransformation transformation;
+
     int64_t iteration_count;
+    ngraph::element::Type ngPrc;
+    InferenceEngine::SizeVector inputShape;
 };
 
 }  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/memory_LSTMCell.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/memory_LSTMCell.hpp
index 489431e2a9b67b..cc0aeb26e2f02e 100644
--- a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/memory_LSTMCell.hpp
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/memory_LSTMCell.hpp
@@ -10,6 +10,7 @@
 
 namespace SubgraphTestsDefinitions {
 typedef std::tuple<
+    ngraph::helpers::MemoryTransformation,   // Apply Memory transformation
     std::string,                        // Target device name
     InferenceEngine::Precision,         // Network precision
     size_t,                             // Input size
@@ -21,9 +22,13 @@ class MemoryLSTMCellTest : public LayerTestsUtils::LayerTestsCommon,
     public testing::WithParamInterface<memoryLSTMCellParams> {
 private:
     // you have to Unroll TI manually and remove memory untill ngraph supports it
+    // since we switching models we need to generate and save weights biases and inputs in SetUp
     void switchToNgraphFriendlyModel();
     void CreatePureTensorIteratorModel();
-    // since we switching models we need to generate and save weights biases and inputs in SetUp
+    void InitMemory();
+    void ApplyLowLatency();
+
+    ngraph::helpers::MemoryTransformation transformation;
     std::vector<float> input_bias;
     std::vector<float> input_weights;
     std::vector<float> hidden_memory_init;
@@ -34,7 +39,6 @@ class MemoryLSTMCellTest : public LayerTestsUtils::LayerTestsCommon,
 protected:
     void SetUp() override;
     void Run() override;
-    void RunLowLatency(bool regular_api = false);
 public:
     static std::string getTestCaseName(const testing::TestParamInfo<memoryLSTMCellParams> &obj);
 };
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/multiple_LSTMCell.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/multiple_LSTMCell.hpp
index 7c1e72a7bb7644..7932ba39a07469 100644
--- a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/multiple_LSTMCell.hpp
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/multiple_LSTMCell.hpp
@@ -10,6 +10,7 @@
 
 namespace SubgraphTestsDefinitions {
 typedef std::tuple<
+    ngraph::helpers::MemoryTransformation,   // Apply Memory transformation
     std::string,                        // Target device name
     InferenceEngine::Precision,         // Network precision
     size_t,                             // Input size
@@ -21,9 +22,12 @@ class MultipleLSTMCellTest : public LayerTestsUtils::LayerTestsCommon,
     public testing::WithParamInterface<multipleLSTMCellParams> {
 private:
     // you have to Unroll TI manually and remove memory untill ngraph supports it
+    // since we switching models we need to generate and save weights biases and inputs in SetUp
     void switchToNgraphFriendlyModel();
     void CreatePureTensorIteratorModel();
-    // since we switching models we need to generate and save weights biases and inputs in SetUp
+    void InitMemory();
+    void ApplyLowLatency();
+
     size_t hiddenSize;
     std::vector<float> input_bias;
     std::vector<float> input_weights;
@@ -33,10 +37,10 @@ class MultipleLSTMCellTest : public LayerTestsUtils::LayerTestsCommon,
     std::vector<float> weights_2_vals;
     std::vector<float> reccurrenceWeights_vals;
     std::vector<float> bias_vals;
+    ngraph::helpers::MemoryTransformation transformation;
 protected:
     void SetUp() override;
     void Run() override;
-    void RunLowLatency(bool regular_api = false);
 public:
     static std::string getTestCaseName(const testing::TestParamInfo<multipleLSTMCellParams> &obj);
 };
diff --git a/inference-engine/tests/functional/shared_test_classes/src/single_layer/memory.cpp b/inference-engine/tests/functional/shared_test_classes/src/single_layer/memory.cpp
index 0984a4dbeab07f..c059768c4a2ad9 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/single_layer/memory.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/single_layer/memory.cpp
@@ -3,10 +3,18 @@
 //
 
 #include <signal.h>
+#include <ie_transformations.hpp>
+#include <transformations/control_flow/unroll_tensor_iterator.hpp>
+#include <transformations/serialize.hpp>
+#include <functional_test_utils/core_config.hpp>
 #include "ngraph/opsets/opset7.hpp"
 #include "ngraph_functions/builders.hpp"
+#include "ngraph/pass/low_latency.hpp"
 #include "shared_test_classes/single_layer/memory.hpp"
 
+using namespace ngraph;
+using namespace opset7;
+
 namespace LayerTestsDefinitions {
 
     std::string MemoryTest::getTestCaseName(const testing::TestParamInfo<MemoryTestParams> &obj) {
@@ -14,9 +22,11 @@ namespace LayerTestsDefinitions {
         InferenceEngine::Precision netPrecision;
         InferenceEngine::SizeVector inputShape;
         std::string targetDevice;
-        std::tie(iteration_count, inputShape, netPrecision, targetDevice) = obj.param;
+        ngraph::helpers::MemoryTransformation transformation;
+        std::tie(transformation, iteration_count, inputShape, netPrecision, targetDevice) = obj.param;
 
         std::ostringstream result;
+        result << "transformation=" << transformation << "_";
         result << "iteration_count=" << iteration_count << "_";
         result << "IS=" << CommonTestUtils::vec2str(inputShape) << "_";
         result << "netPRC=" << netPrecision.name() << "_";
@@ -26,20 +36,17 @@ namespace LayerTestsDefinitions {
     }
 
     void MemoryTest::SetUp() {
-        using namespace ngraph;
-        InferenceEngine::SizeVector inputShape;
-        std::tie(iteration_count, inputShape, netPrecision, targetDevice) = this->GetParam();
-        auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
+        std::tie(transformation, iteration_count, inputShape, netPrecision, targetDevice) = this->GetParam();
+        ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
 
-        auto param = ngraph::builder::makeParams(ngPrc, {inputShape});
-        auto variable = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "v0"});
-        auto read_value = std::make_shared<opset7::ReadValue>(param.at(0), variable);
-        auto add = std::make_shared<opset7::Add>(read_value, param.at(0));
-        auto assign = std::make_shared<opset7::Assign>(add, variable);
-        auto res = std::make_shared<opset7::Result>(add);
-        function = std::make_shared<Function>(ResultVector{res}, SinkVector{assign}, param, "TestMemory");
+        if (transformation == ngraph::helpers::MemoryTransformation::NONE) {
+            CreateCommonFunc();
+        } else {
+            CreateTIFunc();
+            ApplyLowLatency();
+        }
 
-        auto hostTensor = std::make_shared<ngraph::HostTensor>(ngPrc, inputShape);
+        auto hostTensor = std::make_shared<HostTensor>(ngPrc, inputShape);
         auto variable_context = std::make_shared<VariantWrapper<VariableContext>>(VariableContext());
         auto variable_value = std::make_shared<VariableValue>(hostTensor);
         variable_context->get().set_variable_value(function->get_variable_by_id("v0"), variable_value);
@@ -48,6 +55,7 @@ namespace LayerTestsDefinitions {
 
 
     void MemoryTest::Run() {
+        SKIP_IF_CURRENT_TEST_IS_DISABLED()
         using namespace LayerTestsUtils;
         auto crashHandler = [](int errCode) {
             auto &s = Summary::getInstance();
@@ -68,7 +76,13 @@ namespace LayerTestsDefinitions {
         }
 
         try {
-            LoadNetwork();
+            if (transformation != ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_REGULAR_API) {
+                LoadNetwork();
+            } else {
+                CoreConfiguration(this);
+                ConfigureNetwork();
+                executableNetwork = core->LoadNetwork(cnnNetwork, targetDevice, configuration);
+            }
             GenerateInputs();
             for (int64_t i = 0; i < iteration_count; ++i) {
                 Infer();
@@ -88,12 +102,12 @@ namespace LayerTestsDefinitions {
         }
     }
 
-    std::vector<std::pair<ngraph::element::Type, std::vector<std::uint8_t>>> MemoryTest::CalculateRefs() {
+    std::vector<std::pair<element::Type, std::vector<std::uint8_t>>> MemoryTest::CalculateRefs() {
         using namespace ngraph;
         function->validate_nodes_and_infer_types();
 
         auto referenceInputs = std::vector<std::vector<uint8_t>>(inputs.size());
-        auto refInputsTypes = std::vector<ngraph::element::Type>(inputs.size());
+        auto refInputsTypes = std::vector<element::Type>(inputs.size());
         HostTensorVector inputTensors;
         for (auto & input : inputs) {
             const auto &dataSize = input->byteSize();
@@ -104,17 +118,25 @@ namespace LayerTestsDefinitions {
             const auto lockedMemory = memory->wmap();
             const auto buffer = lockedMemory.as<const std::uint8_t *>();
 
-            auto hostTensor = std::make_shared<ngraph::HostTensor>(FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(tensorDesc.getPrecision()),
+            auto hostTensor = std::make_shared<HostTensor>(FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(tensorDesc.getPrecision()),
                                                                    tensorDesc.getDims());
             hostTensor->write(buffer, dataSize);
             inputTensors.push_back(hostTensor);
         }
 
+        // evaluate method is not implemented for TI op.
+        ngraph::pass::Manager manager;
+        manager.register_pass<ngraph::pass::UnrollTensorIterator>();
+        manager.run_passes(function);
+
         const auto &outInfo = executableNetwork.GetOutputsInfo();
-        HostTensorVector outputTensors(outInfo.size(), std::make_shared<ngraph::HostTensor>());
+        HostTensorVector outputTensors(outInfo.size());
+        for (auto& outTensor : outputTensors) {
+            outTensor = std::make_shared<HostTensor>();
+        }
         function->evaluate(outputTensors, inputTensors, eval_context);
 
-        std::vector<std::pair<ngraph::element::Type, std::vector<std::uint8_t>>> outputs(outInfo.size());
+        std::vector<std::pair<element::Type, std::vector<std::uint8_t>>> outputs(outInfo.size());
         for (size_t idx = 0; idx < outInfo.size(); ++idx) {
             outputs[idx].first = outputTensors[idx]->get_element_type();
             outputs[idx].second.resize(outputTensors[idx]->get_size_in_bytes());
@@ -123,5 +145,61 @@ namespace LayerTestsDefinitions {
         return outputs;
     }
 
+    void MemoryTest::CreateTIFunc() {
+        auto param = builder::makeParams(ngPrc, {inputShape}).at(0);
+        std::vector<std::vector<size_t>> shape = {{static_cast<size_t>(iteration_count), 1}};
+        auto iter_count = builder::makeParams(ngPrc, shape).at(0);
+
+        // Body
+        auto X = builder::makeParams(ngPrc, {inputShape}).at(0);
+        auto Y = builder::makeParams(ngPrc, {inputShape}).at(0);
+        auto Iter = builder::makeParams(ngPrc, {Shape{1, 1}}).at(0);
+        auto add = std::make_shared<Add>(X, Y);
+        auto res = std::make_shared<Result>(add);
+        auto Iter_res = std::make_shared<Result>(Iter);
+        auto body = std::make_shared<Function>(OutputVector{res, Iter_res}, ParameterVector {X, Y, Iter});
+
+        // TI construction
+        auto tensor_iterator = std::make_shared<TensorIterator>();
+        tensor_iterator->set_body(body);
+
+        tensor_iterator->set_merged_input(X, param, res);
+        tensor_iterator->set_invariant_input(Y, param);
+        tensor_iterator->set_sliced_input(Iter, iter_count, 0, 1, 1, -1, 0);
+
+        auto output = tensor_iterator->get_iter_value(res, -1);
+        auto output_iter = tensor_iterator->get_concatenated_slices(Iter_res, 0, 1, 1, -1, 0);
+        function = std::make_shared<Function>(OutputVector{output, output_iter},
+                                              ParameterVector{param, iter_count},
+                                              "PureTI");
+    }
+
+    void MemoryTest::CreateCommonFunc() {
+        auto param = builder::makeParams(ngPrc, {inputShape});
+        auto variable = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "v0"});
+        auto read_value = std::make_shared<ReadValue>(param.at(0), variable);
+        auto add = std::make_shared<Add>(read_value, param.at(0));
+        auto assign = std::make_shared<Assign>(add, variable);
+        auto res = std::make_shared<Result>(add);
+        function = std::make_shared<Function>(ResultVector{res}, SinkVector{assign}, param, "TestMemory");
+    }
+
+    void MemoryTest::ApplyLowLatency() {
+       if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2) {
+           function->validate_nodes_and_infer_types();
+           pass::Manager manager;
+           manager.register_pass<pass::LowLatency2>();
+           manager.run_passes(function);
+       } else if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_ORIGINAL_INIT) {
+           function->validate_nodes_and_infer_types();
+           pass::Manager manager;
+           manager.register_pass<pass::LowLatency2>(false);
+           manager.run_passes(function);
+        } else if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_REGULAR_API) {
+            cnnNetwork = InferenceEngine::CNNNetwork{function};
+           InferenceEngine::lowLatency2(cnnNetwork, iteration_count);
+        }
+    }
+
 }  // namespace LayerTestsDefinitions
 
diff --git a/inference-engine/tests/functional/shared_test_classes/src/subgraph/memory_LSTMCell.cpp b/inference-engine/tests/functional/shared_test_classes/src/subgraph/memory_LSTMCell.cpp
index a8d5c067334b53..bac0c293add651 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/subgraph/memory_LSTMCell.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/subgraph/memory_LSTMCell.cpp
@@ -9,6 +9,9 @@
 #include "ngraph_functions/builders.hpp"
 #include "shared_test_classes/subgraph/memory_LSTMCell.hpp"
 
+using namespace ngraph;
+using namespace opset7;
+
 namespace SubgraphTestsDefinitions {
 
     std::string MemoryLSTMCellTest::getTestCaseName(const testing::TestParamInfo<memoryLSTMCellParams> &obj) {
@@ -17,9 +20,11 @@ namespace SubgraphTestsDefinitions {
         size_t inputSize;
         size_t hiddenSize;
         std::map<std::string, std::string> config;
-        std::tie(targetDevice, netPrecision, inputSize, hiddenSize, config) = obj.param;
+        ngraph::helpers::MemoryTransformation transformation;
+        std::tie(transformation, targetDevice, netPrecision, inputSize, hiddenSize, config) = obj.param;
         std::ostringstream result;
 
+        result << "transformation=" << transformation << "_";
         result << "netPrecision=" << netPrecision.name() << "_";
         result << "IS=" << inputSize << "_";
         result << "HS=" << hiddenSize << "_";
@@ -34,7 +39,7 @@ namespace SubgraphTestsDefinitions {
         InferenceEngine::Precision netPrecision;
         std::map<std::string, std::string> config;
         size_t inputSize;
-        std::tie(targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
+        std::tie(transformation, targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
         configuration.insert(config.begin(), config.end());
         auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
 
@@ -51,49 +56,53 @@ namespace SubgraphTestsDefinitions {
         reccurrenceWeights_vals = CommonTestUtils::generate_float_numbers(4 * hiddenSize * hiddenSize, -0.1f, 0.1f);
         bias_vals = CommonTestUtils::generate_float_numbers(4 * hiddenSize, -0.2f, 0.1f);
 
-        auto input_parameter = ngraph::builder::makeParams(ngPrc, {input_dims});
+        auto input_parameter = builder::makeParams(ngPrc, {input_dims});
 
-        auto input_add_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_bias);
-        auto add = ngraph::builder::makeEltwise(input_parameter[0], input_add_const, ngraph::helpers::EltwiseTypes::ADD);
+        auto input_add_const = builder::makeConstant(ngPrc, input_dims, input_bias);
+        auto add = builder::makeEltwise(input_parameter[0], input_add_const, helpers::EltwiseTypes::ADD);
 
-        auto input_mul_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_weights);
-        auto mul = ngraph::builder::makeEltwise(add, input_mul_const, ngraph::helpers::EltwiseTypes::MULTIPLY);
+        auto input_mul_const = builder::makeConstant(ngPrc, input_dims, input_weights);
+        auto mul = builder::makeEltwise(add, input_mul_const, helpers::EltwiseTypes::MULTIPLY);
 
-        auto unsqueeze_input_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-        auto unsqueeze_input = std::make_shared<ngraph::opset5::Unsqueeze>(mul, unsqueeze_input_const);
+        auto unsqueeze_input_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+        auto unsqueeze_input = std::make_shared<Unsqueeze>(mul, unsqueeze_input_const);
 
-        auto permute_in_params = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{3}, ngraph::Shape{{1, 0, 2}});
-        auto permute_in = std::make_shared<ngraph::opset5::Transpose>(unsqueeze_input, permute_in_params);
+        auto permute_in_params = std::make_shared<Constant>(element::i64, Shape{3}, Shape{{1, 0, 2}});
+        auto permute_in = std::make_shared<Transpose>(unsqueeze_input, permute_in_params);
 
-        auto cell_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
-        auto cell_memory_read = std::make_shared<ngraph::opset5::ReadValue>(cell_memory_constant, "cell_memory");
+        auto cell_memory_constant = builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
+        auto var_cell =
+                std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "cell_state_1"});
+        auto var_hidden =
+                std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "hidden_state_1"});
+        auto cell_memory_read = std::make_shared<ReadValue>(cell_memory_constant, var_cell);
 
-        auto hidden_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
-        auto hidden_memory_read = std::make_shared<ngraph::opset5::ReadValue>(hidden_memory_constant, "hidden_memory");
+        auto hidden_memory_constant = builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
+        auto hidden_memory_read = std::make_shared<ReadValue>(hidden_memory_constant, var_hidden);
 
         // Body - inputs
-        auto X = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, 1, inputSize});
-        auto H_t = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
-        auto C_t = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
+        auto X = std::make_shared<Parameter>(ngPrc, Shape{1, 1, inputSize});
+        auto H_t = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
+        auto C_t = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
         // Body - layers
-        auto squeeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-        auto squeeze = std::make_shared<ngraph::opset5::Squeeze>(X, squeeze_const);
+        auto squeeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+        auto squeeze = std::make_shared<Squeeze>(X, squeeze_const);
 
-        auto weightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
-        auto reccurrenceWeightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
-        auto biasNode = ngraph::builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
-        auto lstm = std::make_shared<ngraph::opset5::LSTMCell>(squeeze, H_t, C_t, weightsNode, reccurrenceWeightsNode, biasNode, hiddenSize);
+        auto weightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
+        auto reccurrenceWeightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
+        auto biasNode = builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
+        auto lstm = std::make_shared<LSTMCell>(squeeze, H_t, C_t, weightsNode, reccurrenceWeightsNode, biasNode, hiddenSize);
 
-        auto unsqueeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-        auto unsqueeze = std::make_shared<ngraph::opset5::Unsqueeze>(lstm->output(0), unsqueeze_const);
+        auto unsqueeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+        auto unsqueeze = std::make_shared<Unsqueeze>(lstm->output(0), unsqueeze_const);
         // body - outputs
         auto H_o = lstm->output(0);
         auto C_o = lstm->output(1);
         auto unsqueeze_o = unsqueeze->output(0);
 
-        auto body = std::make_shared<ngraph::Function>(ngraph::OutputVector{unsqueeze_o, H_o, C_o}, ngraph::ParameterVector {X, H_t, C_t});
+        auto body = std::make_shared<Function>(OutputVector{unsqueeze_o, H_o, C_o}, ParameterVector {X, H_t, C_t});
         // TI construction
-        auto tensor_iterator = std::make_shared<ngraph::opset5::TensorIterator>();
+        auto tensor_iterator = std::make_shared<TensorIterator>();
         tensor_iterator->set_body(body);
         tensor_iterator->set_invariant_input(X, permute_in);
         tensor_iterator->set_merged_input(H_t, hidden_memory_read, H_o);
@@ -107,27 +116,27 @@ namespace SubgraphTestsDefinitions {
         out_hidden.get_tensor().set_element_type(ngPrc);
         out_cell.get_tensor().set_element_type(ngPrc);
 
-        auto cell_memory_write = std::make_shared<ngraph::opset5::Assign>(out_cell, "cell_memory");
-        auto hidden_memory_write = std::make_shared<ngraph::opset5::Assign>(out_hidden, "hidden_memory");
+        auto cell_memory_write = std::make_shared<Assign>(out_cell, var_cell);
+        auto hidden_memory_write = std::make_shared<Assign>(out_hidden, var_hidden);
 
-        auto final_reshape_pattern = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{4},
+        auto final_reshape_pattern = std::make_shared<Constant>(element::i64, Shape{4},
                                                                                 std::vector<size_t>({1, 1, 1, hiddenSize}));
-        auto final_reshape = std::make_shared<ngraph::opset5::Reshape>(out_unsqueeze, final_reshape_pattern, false);
+        auto final_reshape = std::make_shared<Reshape>(out_unsqueeze, final_reshape_pattern, false);
 
         cell_memory_write->add_control_dependency(cell_memory_read);
-        final_reshape->add_control_dependency(cell_memory_write);
-
         hidden_memory_write->add_control_dependency(hidden_memory_read);
-        final_reshape->add_control_dependency(hidden_memory_write);
 
-        function = std::make_shared<ngraph::Function>(final_reshape, input_parameter, "TI_with_memory");
+        function = std::make_shared<Function>(OutputVector{final_reshape},
+                                              SinkVector{cell_memory_write, hidden_memory_write},
+                                              input_parameter,
+                                              "TI_with_memory");
     }
 
     void MemoryLSTMCellTest::switchToNgraphFriendlyModel() {
         InferenceEngine::Precision netPrecision;
         std::map<std::string, std::string> config;
         size_t inputSize;
-        std::tie(targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
+        std::tie(transformation, targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
         auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
 
         std::vector<size_t> input_dims { 1, inputSize };
@@ -135,46 +144,46 @@ namespace SubgraphTestsDefinitions {
         std::vector<size_t> hidden_memory_dims {1, hiddenSize};
         std::vector<size_t> cell_memory_dims {1, hiddenSize};
 
-        auto input_parameter = ngraph::builder::makeParams(ngPrc, {input_dims});
+        auto input_parameter = builder::makeParams(ngPrc, {input_dims});
 
-        auto input_add_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_bias);
-        auto add = ngraph::builder::makeEltwise(input_parameter[0], input_add_const, ngraph::helpers::EltwiseTypes::ADD);
+        auto input_add_const = builder::makeConstant(ngPrc, input_dims, input_bias);
+        auto add = builder::makeEltwise(input_parameter[0], input_add_const, helpers::EltwiseTypes::ADD);
 
-        auto input_mul_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_weights);
-        auto mul = ngraph::builder::makeEltwise(add, input_mul_const, ngraph::helpers::EltwiseTypes::MULTIPLY);
+        auto input_mul_const = builder::makeConstant(ngPrc, input_dims, input_weights);
+        auto mul = builder::makeEltwise(add, input_mul_const, helpers::EltwiseTypes::MULTIPLY);
 
-        auto unsqueeze_input_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-        auto unsqueeze_input = std::make_shared<ngraph::opset5::Unsqueeze>(mul, unsqueeze_input_const);
+        auto unsqueeze_input_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+        auto unsqueeze_input = std::make_shared<Unsqueeze>(mul, unsqueeze_input_const);
 
-        auto cell_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
+        auto cell_memory_constant = builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
 
-        auto hidden_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
+        auto hidden_memory_constant = builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
 
         // Body - layers
-        auto squeeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-        auto squeeze = std::make_shared<ngraph::opset5::Squeeze>(unsqueeze_input, squeeze_const);
+        auto squeeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+        auto squeeze = std::make_shared<Squeeze>(unsqueeze_input, squeeze_const);
 
-        auto weightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
-        auto reccurrenceWeightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
-        auto biasNode = ngraph::builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
-        auto lstm = std::make_shared<ngraph::opset5::LSTMCell>(squeeze, hidden_memory_constant, cell_memory_constant, weightsNode,
+        auto weightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
+        auto reccurrenceWeightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
+        auto biasNode = builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
+        auto lstm = std::make_shared<LSTMCell>(squeeze, hidden_memory_constant, cell_memory_constant, weightsNode,
                                                                reccurrenceWeightsNode, biasNode, hiddenSize);
 
-        auto unsqueeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-        auto unsqueeze = std::make_shared<ngraph::opset5::Unsqueeze>(lstm->output(0), unsqueeze_const);
+        auto unsqueeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+        auto unsqueeze = std::make_shared<Unsqueeze>(lstm->output(0), unsqueeze_const);
 
-        auto final_reshape_pattern = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64,
-                                                                            ngraph::Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
-        auto final_reshape = std::make_shared<ngraph::opset5::Reshape>(unsqueeze, final_reshape_pattern, false);
+        auto final_reshape_pattern = std::make_shared<Constant>(element::i64,
+                                                                            Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
+        auto final_reshape = std::make_shared<Reshape>(unsqueeze, final_reshape_pattern, false);
 
-        function = std::make_shared<ngraph::Function>(final_reshape, input_parameter, "TI_unrolled_without_memory");
+        function = std::make_shared<Function>(final_reshape, input_parameter, "TI_unrolled_without_memory");
     }
 
     void MemoryLSTMCellTest::CreatePureTensorIteratorModel() {
         InferenceEngine::Precision netPrecision;
         std::map<std::string, std::string> config;
         size_t inputSize;
-        std::tie(targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
+        std::tie(transformation, targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
         auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
 
         std::vector<size_t> input_dims { 1, inputSize };
@@ -182,49 +191,49 @@ namespace SubgraphTestsDefinitions {
         std::vector<size_t> hidden_memory_dims {1, hiddenSize};
         std::vector<size_t> cell_memory_dims {1, hiddenSize};
 
-        auto input_parameter = ngraph::builder::makeParams(ngPrc, {input_dims});
+        auto input_parameter = builder::makeParams(ngPrc, {input_dims});
 
-        auto input_add_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_bias);
-        auto add = ngraph::builder::makeEltwise(input_parameter[0], input_add_const, ngraph::helpers::EltwiseTypes::ADD);
+        auto input_add_const = builder::makeConstant(ngPrc, input_dims, input_bias);
+        auto add = builder::makeEltwise(input_parameter[0], input_add_const, helpers::EltwiseTypes::ADD);
 
-        auto input_mul_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_weights);
-        auto mul = ngraph::builder::makeEltwise(add, input_mul_const, ngraph::helpers::EltwiseTypes::MULTIPLY);
+        auto input_mul_const = builder::makeConstant(ngPrc, input_dims, input_weights);
+        auto mul = builder::makeEltwise(add, input_mul_const, helpers::EltwiseTypes::MULTIPLY);
 
-        auto unsqueeze_input_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-        auto unsqueeze_input = std::make_shared<ngraph::opset5::Unsqueeze>(mul, unsqueeze_input_const);
+        auto unsqueeze_input_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+        auto unsqueeze_input = std::make_shared<Unsqueeze>(mul, unsqueeze_input_const);
 
-        auto permute_in_params = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{3}, ngraph::Shape{{1, 0, 2}});
-        auto permute_in = std::make_shared<ngraph::opset5::Transpose>(unsqueeze_input, permute_in_params);
+        auto permute_in_params = std::make_shared<Constant>(element::i64, Shape{3}, Shape{{1, 0, 2}});
+        auto permute_in = std::make_shared<Transpose>(unsqueeze_input, permute_in_params);
 
-        auto cell_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
+        auto cell_memory_constant = builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
 
-        auto hidden_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
+        auto hidden_memory_constant = builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
 
         // Body - inputs
-        auto X = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, 1, inputSize});
-        auto H_t = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
-        auto C_t = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
+        auto X = std::make_shared<Parameter>(ngPrc, Shape{1, 1, inputSize});
+        auto H_t = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
+        auto C_t = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
         H_t->set_friendly_name("hidden_state_1");
         C_t->set_friendly_name("cell_state_1");
         // Body - layers
-        auto squeeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-        auto squeeze = std::make_shared<ngraph::opset5::Squeeze>(X, squeeze_const);
+        auto squeeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+        auto squeeze = std::make_shared<Squeeze>(X, squeeze_const);
 
-        auto weightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
-        auto reccurrenceWeightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
-        auto biasNode = ngraph::builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
-        auto lstm = std::make_shared<ngraph::opset5::LSTMCell>(squeeze, H_t, C_t, weightsNode, reccurrenceWeightsNode, biasNode, hiddenSize);
+        auto weightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
+        auto reccurrenceWeightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
+        auto biasNode = builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
+        auto lstm = std::make_shared<LSTMCell>(squeeze, H_t, C_t, weightsNode, reccurrenceWeightsNode, biasNode, hiddenSize);
 
-        auto unsqueeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-        auto unsqueeze = std::make_shared<ngraph::opset5::Unsqueeze>(lstm->output(0), unsqueeze_const);
+        auto unsqueeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+        auto unsqueeze = std::make_shared<Unsqueeze>(lstm->output(0), unsqueeze_const);
         // body - outputs
         auto H_o = lstm->output(0);
         auto C_o = lstm->output(1);
         auto unsqueeze_o = unsqueeze->output(0);
 
-        auto body = std::make_shared<ngraph::Function>(ngraph::OutputVector{unsqueeze_o, H_o, C_o}, ngraph::ParameterVector {X, H_t, C_t});
+        auto body = std::make_shared<Function>(OutputVector{unsqueeze_o, H_o, C_o}, ParameterVector {X, H_t, C_t});
         // TI construction
-        auto tensor_iterator = std::make_shared<ngraph::opset5::TensorIterator>();
+        auto tensor_iterator = std::make_shared<TensorIterator>();
         tensor_iterator->set_body(body);
         tensor_iterator->set_sliced_input(X, permute_in, 0, 1, 1, -1, 0);
         tensor_iterator->set_merged_input(H_t, hidden_memory_constant, H_o);
@@ -237,56 +246,35 @@ namespace SubgraphTestsDefinitions {
         out_hidden.get_tensor().set_element_type(ngPrc);
         out_cell.get_tensor().set_element_type(ngPrc);
 
-        auto final_reshape_pattern = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{4},
+        auto final_reshape_pattern = std::make_shared<Constant>(element::i64, Shape{4},
                                                                                 std::vector<size_t>({1, 1, 1, hiddenSize}));
-        auto final_reshape = std::make_shared<ngraph::opset5::Reshape>(out_unsqueeze, final_reshape_pattern, false);
+        auto final_reshape = std::make_shared<Reshape>(out_unsqueeze, final_reshape_pattern, false);
 
-        function = std::make_shared<ngraph::Function>(final_reshape, input_parameter, "PureTI");
+        function = std::make_shared<Function>(final_reshape, input_parameter, "PureTI");
     }
 
     void MemoryLSTMCellTest::Run() {
         SKIP_IF_CURRENT_TEST_IS_DISABLED()
-
-        IE_SUPPRESS_DEPRECATED_START
-        LoadNetwork();
-        auto states = executableNetwork.QueryState();
-        for (auto& state : states) {
-            auto name = state.GetName();
-            if (name == "cell_memory") {
-                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetLastState()->getTensorDesc(),
-                                                                           cell_memory_init.data(), cell_memory_init.size());
-                state.SetState(blob);
-            } else if (name == "hidden_memory") {
-                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetLastState()->getTensorDesc(),
-                                                                           hidden_memory_init.data(), hidden_memory_init.size());
-                state.SetState(blob);
-            } else {
-                GTEST_FAIL() << "unknown memory state";
-            }
+        if (transformation != ngraph::helpers::MemoryTransformation::NONE) {
+            ApplyLowLatency();
+        } else {
+            LoadNetwork();
         }
-        IE_SUPPRESS_DEPRECATED_END
+
+        InitMemory();
         GenerateInputs();
         Infer();
-        switchToNgraphFriendlyModel();
-        Validate();
-    }
-
-    void MemoryLSTMCellTest::RunLowLatency(bool regular_api) {
-        SKIP_IF_CURRENT_TEST_IS_DISABLED()
 
-        CreatePureTensorIteratorModel();
-        if (regular_api) {
-            cnnNetwork = InferenceEngine::CNNNetwork{function};
-            InferenceEngine::LowLatency(cnnNetwork);
-            ConfigureNetwork();
-            executableNetwork = core->LoadNetwork(static_cast<const InferenceEngine::CNNNetwork>(cnnNetwork), targetDevice, configuration);
+        // Calculate ref values
+        if (transformation == ngraph::helpers::MemoryTransformation::NONE) {
+            switchToNgraphFriendlyModel();
         } else {
-            // Apply LowLatency (insert Assigns/ReadValues) and UnrollTensorIterator
-            ngraph::pass::Manager manager;
-            manager.register_pass<ngraph::pass::LowLatency>(); // LowLatency enables UnrollTI
-            manager.run_passes(function);
-            LoadNetwork();
+            CreatePureTensorIteratorModel();
         }
+        Validate();
+    }
+
+    void MemoryLSTMCellTest::InitMemory() {
         IE_SUPPRESS_DEPRECATED_START
         auto states = executableNetwork.QueryState();
         for (auto& state : states) {
@@ -304,13 +292,52 @@ namespace SubgraphTestsDefinitions {
             }
         }
         IE_SUPPRESS_DEPRECATED_END
-        GenerateInputs();
-        Infer();
+    }
 
+    void MemoryLSTMCellTest::ApplyLowLatency() {
+        // Calculate values after LowLatency transformation
         CreatePureTensorIteratorModel();
-        ngraph::pass::Manager manager_2;
-        manager_2.register_pass<ngraph::pass::UnrollTensorIterator>();
-        manager_2.run_passes(function);
-        Validate();
+        if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY) {
+            function->validate_nodes_and_infer_types();
+            // Apply LowLatency (insert Assigns/ReadValues) and UnrollTensorIterator
+            pass::Manager manager;
+            NGRAPH_SUPPRESS_DEPRECATED_START
+            manager.register_pass<ngraph::pass::LowLatency>();
+            NGRAPH_SUPPRESS_DEPRECATED_END // LowLatency enables UnrollTI
+            manager.run_passes(function);
+            bool ti_found = helpers::is_tensor_iterator_exist(function);
+            EXPECT_EQ(ti_found, true);
+            LoadNetwork();
+        } else if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2) {
+            function->validate_nodes_and_infer_types();
+            // Apply LowLatency (insert Assigns/ReadValues) and UnrollTensorIterator
+
+            pass::Manager manager;
+            manager.register_pass<pass::LowLatency2>();
+            manager.run_passes(function);
+            bool ti_found = helpers::is_tensor_iterator_exist(function);
+            EXPECT_EQ(ti_found, false);
+            LoadNetwork();
+        } else if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY_REGULAR_API) {
+            cnnNetwork = InferenceEngine::CNNNetwork{function};
+            IE_SUPPRESS_DEPRECATED_START
+            InferenceEngine::LowLatency(cnnNetwork);
+            IE_SUPPRESS_DEPRECATED_END
+
+            bool ti_found = helpers::is_tensor_iterator_exist(cnnNetwork.getFunction());
+            EXPECT_EQ(ti_found, true);
+
+            ConfigureNetwork();
+            executableNetwork = core->LoadNetwork(cnnNetwork, targetDevice, configuration);
+        } else if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_REGULAR_API) {
+            cnnNetwork = InferenceEngine::CNNNetwork{function};
+            InferenceEngine::lowLatency2(cnnNetwork);
+
+            bool ti_found = helpers::is_tensor_iterator_exist(cnnNetwork.getFunction());
+            EXPECT_EQ(ti_found, false);
+
+            ConfigureNetwork();
+            executableNetwork = core->LoadNetwork(cnnNetwork, targetDevice, configuration);
+        }
     }
 }  // namespace SubgraphTestsDefinitions
diff --git a/inference-engine/tests/functional/shared_test_classes/src/subgraph/multiple_LSTMCell.cpp b/inference-engine/tests/functional/shared_test_classes/src/subgraph/multiple_LSTMCell.cpp
index d854f7049302a4..09f8020df41220 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/subgraph/multiple_LSTMCell.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/subgraph/multiple_LSTMCell.cpp
@@ -2,16 +2,19 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "ngraph/opsets/opset5.hpp"
-#include "ngraph/pass/low_latency.hpp"
-
 #include "ie_transformations.hpp"
-#include "transformations/control_flow/unroll_tensor_iterator.hpp"
+#include "ngraph/opsets/opset7.hpp"
+#include "ngraph/op/util/variable_context.hpp"
+#include "ngraph/pass/low_latency.hpp"
 
 #include "ngraph_functions/builders.hpp"
+#include "ngraph_functions/utils/ngraph_helpers.hpp"
 
 #include "shared_test_classes/subgraph/multiple_LSTMCell.hpp"
 
+using namespace ngraph;
+using namespace opset7;
+
 namespace SubgraphTestsDefinitions {
 std::string MultipleLSTMCellTest::getTestCaseName(const testing::TestParamInfo<multipleLSTMCellParams> &obj) {
     std::string targetDevice;
@@ -19,9 +22,11 @@ std::string MultipleLSTMCellTest::getTestCaseName(const testing::TestParamInfo<m
     size_t inputSize;
     size_t hiddenSize;
     std::map<std::string, std::string> config;
-    std::tie(targetDevice, netPrecision, inputSize, hiddenSize, config) = obj.param;
+    ngraph::helpers::MemoryTransformation transformation;
+    std::tie(transformation, targetDevice, netPrecision, inputSize, hiddenSize, config) = obj.param;
     std::ostringstream result;
 
+    result << "transformation=" << transformation << "_";
     result << "netPrecision=" << netPrecision.name() << "_";
     result << "IS=" << inputSize << "_";
     result << "HS=" << hiddenSize << "_";
@@ -33,7 +38,7 @@ void MultipleLSTMCellTest::SetUp() {
     InferenceEngine::Precision netPrecision;
     std::map<std::string, std::string> config;
     size_t inputSize;
-    std::tie(targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
+    std::tie(transformation, targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
     configuration.insert(config.begin(), config.end());
     auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
 
@@ -51,51 +56,55 @@ void MultipleLSTMCellTest::SetUp() {
     reccurrenceWeights_vals = CommonTestUtils::generate_float_numbers(4 * hiddenSize * hiddenSize, -0.1f, 0.1f);
     bias_vals = CommonTestUtils::generate_float_numbers(4 * hiddenSize, -0.25f, 0.15f);
 
-    auto input_parameter = ngraph::builder::makeParams(ngPrc, {input_dims});
+    auto input_parameter = builder::makeParams(ngPrc, {input_dims});
 
-    auto input_add_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_bias);
-    auto add = ngraph::builder::makeEltwise(input_parameter[0], input_add_const, ngraph::helpers::EltwiseTypes::ADD);
+    auto input_add_const = builder::makeConstant(ngPrc, input_dims, input_bias);
+    auto add = builder::makeEltwise(input_parameter[0], input_add_const, helpers::EltwiseTypes::ADD);
 
-    auto input_mul_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_weights);
-    auto mul = ngraph::builder::makeEltwise(add, input_mul_const, ngraph::helpers::EltwiseTypes::MULTIPLY);
+    auto input_mul_const = builder::makeConstant(ngPrc, input_dims, input_weights);
+    auto mul = builder::makeEltwise(add, input_mul_const, helpers::EltwiseTypes::MULTIPLY);
 
-    auto unsqueeze_input_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto unsqueeze_input = std::make_shared<ngraph::opset5::Unsqueeze>(mul, unsqueeze_input_const);
+    auto unsqueeze_input_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto unsqueeze_input = std::make_shared<Unsqueeze>(mul, unsqueeze_input_const);
 
-    auto permute_in_params = std::make_shared<ngraph::opset1::Constant>(ngraph::element::i64, ngraph::Shape{3}, ngraph::Shape{{1, 0, 2}});
-    auto permute_in = std::make_shared<ngraph::opset1::Transpose>(unsqueeze_input, permute_in_params);
+    auto permute_in_params = std::make_shared<Constant>(element::i64, Shape{3}, Shape{{1, 0, 2}});
+    auto permute_in = std::make_shared<Transpose>(unsqueeze_input, permute_in_params);
 
-    auto cell_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
-    auto cell_memory_read = std::make_shared<ngraph::opset5::ReadValue>(cell_memory_constant, "cell_memory");
+    auto cell_memory_constant = builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
+    auto var_cell =
+            std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "cell_state_1"});
+    auto var_hidden =
+            std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "hidden_state_1"});
+    auto cell_memory_read = std::make_shared<ReadValue>(cell_memory_constant, var_cell);
     cell_memory_read->set_friendly_name("cell_memory");
 
-    auto hidden_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
-    auto hidden_memory_read = std::make_shared<ngraph::opset5::ReadValue>(hidden_memory_constant, "hidden_memory");
+    auto hidden_memory_constant = builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
+    auto hidden_memory_read = std::make_shared<ReadValue>(hidden_memory_constant, var_hidden);
     hidden_memory_read->set_friendly_name("hidden_memory");
 
     // Body - inputs
-    auto X = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, 1, inputSize});
-    auto H_t = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
-    auto C_t = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
+    auto X = std::make_shared<Parameter>(ngPrc, Shape{1, 1, inputSize});
+    auto H_t = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
+    auto C_t = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
     // Body - layers
-    auto squeeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto squeeze = std::make_shared<ngraph::opset5::Squeeze>(X, squeeze_const);
+    auto squeeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto squeeze = std::make_shared<Squeeze>(X, squeeze_const);
 
-    auto weightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
-    auto reccurrenceWeightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
-    auto biasNode = ngraph::builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
-    auto lstm = std::make_shared<ngraph::opset4::LSTMCell>(squeeze, H_t, C_t, weightsNode, reccurrenceWeightsNode, biasNode, hiddenSize);
+    auto weightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
+    auto reccurrenceWeightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
+    auto biasNode = builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
+    auto lstm = std::make_shared<LSTMCell>(squeeze, H_t, C_t, weightsNode, reccurrenceWeightsNode, biasNode, hiddenSize);
 
-    auto unsqueeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto unsqueeze = std::make_shared<ngraph::opset5::Unsqueeze>(lstm->output(0), unsqueeze_const);
+    auto unsqueeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto unsqueeze = std::make_shared<Unsqueeze>(lstm->output(0), unsqueeze_const);
     // body - outputs
     auto H_o = lstm->output(0);
     auto C_o = lstm->output(1);
     auto unsqueeze_o = unsqueeze->output(0);
 
-    auto body = std::make_shared<ngraph::Function>(ngraph::OutputVector{unsqueeze_o, H_o, C_o}, ngraph::ParameterVector {X, H_t, C_t});
+    auto body = std::make_shared<Function>(OutputVector{unsqueeze_o, H_o, C_o}, ParameterVector {X, H_t, C_t});
     // TI construction
-    auto tensor_iterator = std::make_shared<ngraph::opset5::TensorIterator>();
+    auto tensor_iterator = std::make_shared<TensorIterator>();
     tensor_iterator->set_body(body);
     tensor_iterator->set_invariant_input(X, permute_in);
     tensor_iterator->set_merged_input(H_t, hidden_memory_read, H_o);
@@ -108,49 +117,53 @@ void MultipleLSTMCellTest::SetUp() {
     out_hidden.get_tensor().set_element_type(ngPrc);
     out_cell.get_tensor().set_element_type(ngPrc);
 
-    auto cell_memory_write = std::make_shared<ngraph::opset5::Assign>(out_cell, "cell_memory");
-    auto hidden_memory_write = std::make_shared<ngraph::opset5::Assign>(out_hidden, "hidden_memory");
+    auto cell_memory_write = std::make_shared<Assign>(out_cell, var_cell);
+    auto hidden_memory_write = std::make_shared<Assign>(out_hidden, var_hidden);
 
-    auto first_reshape_pattern = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64,
-                                                                        ngraph::Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
-    auto first_reshape = std::make_shared<ngraph::opset5::Reshape>(out_unsqueeze, first_reshape_pattern, false);
+    auto first_reshape_pattern = std::make_shared<Constant>(element::i64,
+                                                                        Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
+    auto first_reshape = std::make_shared<Reshape>(out_unsqueeze, first_reshape_pattern, false);
     // End of TI 1
 
-    auto inbetween_squeeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto inbetween_squeeze = std::make_shared<ngraph::opset5::Squeeze>(first_reshape, inbetween_squeeze_const);
+    auto inbetween_squeeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto inbetween_squeeze = std::make_shared<Squeeze>(first_reshape, inbetween_squeeze_const);
 
     // Second TI
-    auto cell_memory_2_constant = ngraph::builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
-    auto cell_memory_2_read = std::make_shared<ngraph::opset5::ReadValue>(cell_memory_2_constant, "cell_memory_2");
+    auto var_cell_2 =
+            std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "cell_state_2"});
+    auto var_hidden_2 =
+            std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "hidden_state_2"});
+    auto cell_memory_2_constant = builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
+    auto cell_memory_2_read = std::make_shared<ReadValue>(cell_memory_2_constant, var_cell_2);
     cell_memory_2_read->set_friendly_name("cell_memory_2");
 
-    auto hidden_memory_2_constant = ngraph::builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
-    auto hidden_memory_2_read = std::make_shared<ngraph::opset5::ReadValue>(hidden_memory_2_constant, "hidden_memory_2");
+    auto hidden_memory_2_constant = builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
+    auto hidden_memory_2_read = std::make_shared<ReadValue>(hidden_memory_2_constant, var_hidden_2);
     hidden_memory_2_read->set_friendly_name("hidden_memory_2");
 
     // Body - inputs
-    auto X_2 = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, 1, hiddenSize});
-    auto H_t_2 = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
-    auto C_t_2 = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
+    auto X_2 = std::make_shared<Parameter>(ngPrc, Shape{1, 1, hiddenSize});
+    auto H_t_2 = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
+    auto C_t_2 = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
     // Body - layers
-    auto squeeze_2_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto squeeze_2 = std::make_shared<ngraph::opset5::Squeeze>(X_2, squeeze_2_const);
+    auto squeeze_2_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto squeeze_2 = std::make_shared<Squeeze>(X_2, squeeze_2_const);
 
-    auto weightsNode_2 = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, weights_2_vals);
-    auto reccurrenceWeightsNode_2 = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
-    auto biasNode_2 = ngraph::builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
-    auto lstm_2 = std::make_shared<ngraph::opset4::LSTMCell>(squeeze_2, H_t_2, C_t_2, weightsNode_2, reccurrenceWeightsNode_2, biasNode_2, hiddenSize);
+    auto weightsNode_2 = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, weights_2_vals);
+    auto reccurrenceWeightsNode_2 = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
+    auto biasNode_2 = builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
+    auto lstm_2 = std::make_shared<LSTMCell>(squeeze_2, H_t_2, C_t_2, weightsNode_2, reccurrenceWeightsNode_2, biasNode_2, hiddenSize);
 
-    auto unsqueeze_2_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto unsqueeze_2 = std::make_shared<ngraph::opset5::Unsqueeze>(lstm_2->output(0), unsqueeze_2_const);
+    auto unsqueeze_2_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto unsqueeze_2 = std::make_shared<Unsqueeze>(lstm_2->output(0), unsqueeze_2_const);
     // body - outputs
     auto H_o_2 = lstm_2->output(0);
     auto C_o_2 = lstm_2->output(1);
     auto unsqueeze_o_2 = unsqueeze_2->output(0);
 
-    auto body_2 = std::make_shared<ngraph::Function>(ngraph::OutputVector{unsqueeze_o_2, H_o_2, C_o_2}, ngraph::ParameterVector {X_2, H_t_2, C_t_2});
+    auto body_2 = std::make_shared<Function>(OutputVector{unsqueeze_o_2, H_o_2, C_o_2}, ParameterVector {X_2, H_t_2, C_t_2});
     // TI construction
-    auto tensor_iterator_2 = std::make_shared<ngraph::opset5::TensorIterator>();
+    auto tensor_iterator_2 = std::make_shared<TensorIterator>();
     tensor_iterator_2->set_body(body_2);
     tensor_iterator_2->set_invariant_input(X_2, inbetween_squeeze);
     tensor_iterator_2->set_merged_input(H_t_2, hidden_memory_2_read, H_o_2);
@@ -163,33 +176,28 @@ void MultipleLSTMCellTest::SetUp() {
     out_hidden_2.get_tensor().set_element_type(ngPrc);
     out_cell_2.get_tensor().set_element_type(ngPrc);
 
-    auto cell_memory_2_write = std::make_shared<ngraph::opset5::Assign>(out_cell_2, "cell_memory_2");
-    auto hidden_memory_2_write = std::make_shared<ngraph::opset5::Assign>(out_hidden_2, "hidden_memory_2");
+    auto cell_memory_2_write = std::make_shared<Assign>(out_cell_2, var_cell_2);
+    auto hidden_memory_2_write = std::make_shared<Assign>(out_hidden_2, var_hidden_2);
 
-    auto final_reshape_pattern = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64,
-                                                                        ngraph::Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
-    auto final_reshape = std::make_shared<ngraph::opset5::Reshape>(out_unsqueeze_2, final_reshape_pattern, false);
+    auto final_reshape_pattern = std::make_shared<Constant>(element::i64, Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
+    auto final_reshape = std::make_shared<Reshape>(out_unsqueeze_2, final_reshape_pattern, false);
 
     cell_memory_write->add_control_dependency(cell_memory_read);
-    final_reshape->add_control_dependency(cell_memory_write);
-
     hidden_memory_write->add_control_dependency(hidden_memory_read);
-    final_reshape->add_control_dependency(hidden_memory_write);
-
     cell_memory_2_write->add_control_dependency(cell_memory_2_read);
-    final_reshape->add_control_dependency(cell_memory_2_write);
-
     hidden_memory_2_write->add_control_dependency(hidden_memory_2_read);
-    final_reshape->add_control_dependency(hidden_memory_2_write);
 
-    function = std::make_shared<ngraph::Function>(final_reshape, input_parameter, "TI_with_memory");
+    function = std::make_shared<Function>(OutputVector {final_reshape},
+                                          SinkVector{cell_memory_write, hidden_memory_write, cell_memory_2_write, hidden_memory_2_write},
+                                          input_parameter,
+                                          "TI_with_memory");
 }
 
 void MultipleLSTMCellTest::switchToNgraphFriendlyModel() {
     InferenceEngine::Precision netPrecision;
     std::map<std::string, std::string> config;
     size_t inputSize;
-    std::tie(targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
+    std::tie(transformation, targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
     auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
 
     std::vector<size_t> input_dims { 1, inputSize };
@@ -197,72 +205,72 @@ void MultipleLSTMCellTest::switchToNgraphFriendlyModel() {
     std::vector<size_t> hidden_memory_dims {1, hiddenSize};
     std::vector<size_t> cell_memory_dims {1, hiddenSize};
 
-    auto input_parameter = ngraph::builder::makeParams(ngPrc, {input_dims});
+    auto input_parameter = builder::makeParams(ngPrc, {input_dims});
 
-    auto input_add_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_bias);
-    auto add = ngraph::builder::makeEltwise(input_parameter[0], input_add_const, ngraph::helpers::EltwiseTypes::ADD);
+    auto input_add_const = builder::makeConstant(ngPrc, input_dims, input_bias);
+    auto add = builder::makeEltwise(input_parameter[0], input_add_const, helpers::EltwiseTypes::ADD);
 
-    auto input_mul_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_weights);
-    auto mul = ngraph::builder::makeEltwise(add, input_mul_const, ngraph::helpers::EltwiseTypes::MULTIPLY);
+    auto input_mul_const = builder::makeConstant(ngPrc, input_dims, input_weights);
+    auto mul = builder::makeEltwise(add, input_mul_const, helpers::EltwiseTypes::MULTIPLY);
 
-    auto unsqueeze_input_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto unsqueeze_input = std::make_shared<ngraph::opset5::Unsqueeze>(mul, unsqueeze_input_const);
+    auto unsqueeze_input_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto unsqueeze_input = std::make_shared<Unsqueeze>(mul, unsqueeze_input_const);
 
     // Body 1 - layers
-    auto cell_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
+    auto cell_memory_constant = builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
 
-    auto hidden_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
+    auto hidden_memory_constant = builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
 
-    auto squeeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto squeeze = std::make_shared<ngraph::opset5::Squeeze>(unsqueeze_input, squeeze_const);
+    auto squeeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto squeeze = std::make_shared<Squeeze>(unsqueeze_input, squeeze_const);
 
-    auto weightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
-    auto reccurrenceWeightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
-    auto biasNode = ngraph::builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
-    auto lstm = std::make_shared<ngraph::opset4::LSTMCell>(squeeze, hidden_memory_constant, cell_memory_constant, weightsNode,
+    auto weightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
+    auto reccurrenceWeightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
+    auto biasNode = builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
+    auto lstm = std::make_shared<LSTMCell>(squeeze, hidden_memory_constant, cell_memory_constant, weightsNode,
                                                            reccurrenceWeightsNode, biasNode, hiddenSize);
 
-    auto unsqueeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto unsqueeze = std::make_shared<ngraph::opset5::Unsqueeze>(lstm->output(0), unsqueeze_const);
+    auto unsqueeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto unsqueeze = std::make_shared<Unsqueeze>(lstm->output(0), unsqueeze_const);
 
-    auto first_reshape_pattern = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64,
-                                                                        ngraph::Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
-    auto first_reshape = std::make_shared<ngraph::opset5::Reshape>(unsqueeze, first_reshape_pattern, false);
+    auto first_reshape_pattern = std::make_shared<Constant>(element::i64,
+                                                                        Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
+    auto first_reshape = std::make_shared<Reshape>(unsqueeze, first_reshape_pattern, false);
     // Body 1 - end
 
-    auto inbetween_squeeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto inbetween_squeeze = std::make_shared<ngraph::opset5::Squeeze>(first_reshape, inbetween_squeeze_const);
+    auto inbetween_squeeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto inbetween_squeeze = std::make_shared<Squeeze>(first_reshape, inbetween_squeeze_const);
 
     // Body 2 - layers
-    auto cell_memory_2_constant = ngraph::builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
+    auto cell_memory_2_constant = builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
 
-    auto hidden_memory_2_constant = ngraph::builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
+    auto hidden_memory_2_constant = builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
 
-    auto squeeze_2_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto squeeze_2 = std::make_shared<ngraph::opset5::Squeeze>(inbetween_squeeze, squeeze_2_const);
+    auto squeeze_2_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto squeeze_2 = std::make_shared<Squeeze>(inbetween_squeeze, squeeze_2_const);
 
-    auto weightsNode_2 = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, weights_2_vals);
-    auto reccurrenceWeightsNode_2 = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
-    auto biasNode_2 = ngraph::builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
-    auto lstm_2 = std::make_shared<ngraph::opset4::LSTMCell>(squeeze_2, hidden_memory_2_constant, cell_memory_2_constant, weightsNode_2,
+    auto weightsNode_2 = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, weights_2_vals);
+    auto reccurrenceWeightsNode_2 = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
+    auto biasNode_2 = builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
+    auto lstm_2 = std::make_shared<LSTMCell>(squeeze_2, hidden_memory_2_constant, cell_memory_2_constant, weightsNode_2,
         reccurrenceWeightsNode_2, biasNode_2, hiddenSize);
 
-    auto unsqueeze_2_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto unsqueeze_2 = std::make_shared<ngraph::opset5::Unsqueeze>(lstm_2->output(0), unsqueeze_2_const);
+    auto unsqueeze_2_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto unsqueeze_2 = std::make_shared<Unsqueeze>(lstm_2->output(0), unsqueeze_2_const);
 
-    auto final_reshape_pattern = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64,
-        ngraph::Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
-    auto final_reshape = std::make_shared<ngraph::opset5::Reshape>(unsqueeze_2, final_reshape_pattern, false);
+    auto final_reshape_pattern = std::make_shared<Constant>(element::i64,
+        Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
+    auto final_reshape = std::make_shared<Reshape>(unsqueeze_2, final_reshape_pattern, false);
     // Body 2 - end
 
-    function = std::make_shared<ngraph::Function>(final_reshape, input_parameter, "TI_unrolled_without_memory");
+    function = std::make_shared<Function>(final_reshape, input_parameter, "TI_unrolled_without_memory");
 }
 
 void MultipleLSTMCellTest::CreatePureTensorIteratorModel() {
     InferenceEngine::Precision netPrecision;
     std::map<std::string, std::string> config;
     size_t inputSize;
-    std::tie(targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
+    std::tie(transformation, targetDevice, netPrecision, inputSize, hiddenSize, config) = this->GetParam();
     auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
 
     std::vector<size_t> input_dims { 1, inputSize };
@@ -270,49 +278,49 @@ void MultipleLSTMCellTest::CreatePureTensorIteratorModel() {
     std::vector<size_t> hidden_memory_dims {1, hiddenSize};
     std::vector<size_t> cell_memory_dims {1, hiddenSize};
 
-    auto input_parameter = ngraph::builder::makeParams(ngPrc, {input_dims});
+    auto input_parameter = builder::makeParams(ngPrc, {input_dims});
 
-    auto input_add_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_bias);
-    auto add = ngraph::builder::makeEltwise(input_parameter[0], input_add_const, ngraph::helpers::EltwiseTypes::ADD);
+    auto input_add_const = builder::makeConstant(ngPrc, input_dims, input_bias);
+    auto add = builder::makeEltwise(input_parameter[0], input_add_const, helpers::EltwiseTypes::ADD);
 
-    auto input_mul_const = ngraph::builder::makeConstant(ngPrc, input_dims, input_weights);
-    auto mul = ngraph::builder::makeEltwise(add, input_mul_const, ngraph::helpers::EltwiseTypes::MULTIPLY);
+    auto input_mul_const = builder::makeConstant(ngPrc, input_dims, input_weights);
+    auto mul = builder::makeEltwise(add, input_mul_const, helpers::EltwiseTypes::MULTIPLY);
 
-    auto unsqueeze_input_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto unsqueeze_input = std::make_shared<ngraph::opset5::Unsqueeze>(mul, unsqueeze_input_const);
+    auto unsqueeze_input_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto unsqueeze_input = std::make_shared<Unsqueeze>(mul, unsqueeze_input_const);
 
-    auto permute_in_params = std::make_shared<ngraph::opset1::Constant>(ngraph::element::i64, ngraph::Shape{3}, ngraph::Shape{{1, 0, 2}});
-    auto permute_in = std::make_shared<ngraph::opset1::Transpose>(unsqueeze_input, permute_in_params);
+    auto permute_in_params = std::make_shared<Constant>(element::i64, Shape{3}, Shape{{1, 0, 2}});
+    auto permute_in = std::make_shared<Transpose>(unsqueeze_input, permute_in_params);
 
-    auto cell_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
+    auto cell_memory_constant = builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
 
-    auto hidden_memory_constant = ngraph::builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
+    auto hidden_memory_constant = builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
 
     // Body - inputs
-    auto X = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, 1, inputSize});
-    auto H_t = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
-    auto C_t = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
+    auto X = std::make_shared<Parameter>(ngPrc, Shape{1, 1, inputSize});
+    auto H_t = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
+    auto C_t = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
     H_t->set_friendly_name("hidden_state_1");
     C_t->set_friendly_name("cell_state_1");
     // Body - layers
-    auto squeeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto squeeze = std::make_shared<ngraph::opset5::Squeeze>(X, squeeze_const);
+    auto squeeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto squeeze = std::make_shared<Squeeze>(X, squeeze_const);
 
-    auto weightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
-    auto reccurrenceWeightsNode = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
-    auto biasNode = ngraph::builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
-    auto lstm = std::make_shared<ngraph::opset4::LSTMCell>(squeeze, H_t, C_t, weightsNode, reccurrenceWeightsNode, biasNode, hiddenSize);
+    auto weightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, inputSize }, weights_vals);
+    auto reccurrenceWeightsNode = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
+    auto biasNode = builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
+    auto lstm = std::make_shared<LSTMCell>(squeeze, H_t, C_t, weightsNode, reccurrenceWeightsNode, biasNode, hiddenSize);
 
-    auto unsqueeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto unsqueeze = std::make_shared<ngraph::opset5::Unsqueeze>(lstm->output(0), unsqueeze_const);
+    auto unsqueeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto unsqueeze = std::make_shared<Unsqueeze>(lstm->output(0), unsqueeze_const);
     // body - outputs
     auto H_o = lstm->output(0);
     auto C_o = lstm->output(1);
     auto unsqueeze_o = unsqueeze->output(0);
 
-    auto body = std::make_shared<ngraph::Function>(ngraph::OutputVector{unsqueeze_o, H_o, C_o}, ngraph::ParameterVector {X, H_t, C_t});
+    auto body = std::make_shared<Function>(OutputVector{unsqueeze_o, H_o, C_o}, ParameterVector {X, H_t, C_t});
     // TI construction
-    auto tensor_iterator = std::make_shared<ngraph::opset5::TensorIterator>();
+    auto tensor_iterator = std::make_shared<TensorIterator>();
     tensor_iterator->set_body(body);
     tensor_iterator->set_sliced_input(X, permute_in, 0, 1, 1, -1, 0);
     tensor_iterator->set_merged_input(H_t, hidden_memory_constant, H_o);
@@ -326,44 +334,44 @@ void MultipleLSTMCellTest::CreatePureTensorIteratorModel() {
     out_cell.get_tensor().set_element_type(ngPrc);
     tensor_iterator->validate_and_infer_types();
 
-    auto first_reshape_pattern = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64,
-                                                                        ngraph::Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
-    auto first_reshape = std::make_shared<ngraph::opset5::Reshape>(out_unsqueeze, first_reshape_pattern, false);
+    auto first_reshape_pattern = std::make_shared<Constant>(element::i64,
+                                                                        Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
+    auto first_reshape = std::make_shared<Reshape>(out_unsqueeze, first_reshape_pattern, false);
     // End of TI 1
 
-    auto inbetween_squeeze_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto inbetween_squeeze = std::make_shared<ngraph::opset5::Squeeze>(first_reshape, inbetween_squeeze_const);
+    auto inbetween_squeeze_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto inbetween_squeeze = std::make_shared<Squeeze>(first_reshape, inbetween_squeeze_const);
 
     // Second TI
-    auto cell_memory_2_constant = ngraph::builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
+    auto cell_memory_2_constant = builder::makeConstant<float>(ngPrc, cell_memory_dims, cell_memory_init);
 
-    auto hidden_memory_2_constant = ngraph::builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
+    auto hidden_memory_2_constant = builder::makeConstant<float>(ngPrc, hidden_memory_dims, hidden_memory_init);
 
     // Body - inputs
-    auto X_2 = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, 1, hiddenSize});
-    auto H_t_2 = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
-    auto C_t_2 = std::make_shared<ngraph::opset5::Parameter>(ngPrc, ngraph::Shape{1, hiddenSize});
+    auto X_2 = std::make_shared<Parameter>(ngPrc, Shape{1, 1, hiddenSize});
+    auto H_t_2 = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
+    auto C_t_2 = std::make_shared<Parameter>(ngPrc, Shape{1, hiddenSize});
     H_t_2->set_friendly_name("hidden_state_2");
     C_t_2->set_friendly_name("cell_state_2");
     // Body - layers
-    auto squeeze_2_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto squeeze_2 = std::make_shared<ngraph::opset5::Squeeze>(X_2, squeeze_2_const);
+    auto squeeze_2_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto squeeze_2 = std::make_shared<Squeeze>(X_2, squeeze_2_const);
 
-    auto weightsNode_2 = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, weights_2_vals);
-    auto reccurrenceWeightsNode_2 = ngraph::builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
-    auto biasNode_2 = ngraph::builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
-    auto lstm_2 = std::make_shared<ngraph::opset4::LSTMCell>(squeeze_2, H_t_2, C_t_2, weightsNode_2, reccurrenceWeightsNode_2, biasNode_2, hiddenSize);
+    auto weightsNode_2 = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, weights_2_vals);
+    auto reccurrenceWeightsNode_2 = builder::makeConstant<float>(ngPrc, { 4 * hiddenSize, hiddenSize }, reccurrenceWeights_vals);
+    auto biasNode_2 = builder::makeConstant<float>(ngPrc, {4 * hiddenSize}, bias_vals);
+    auto lstm_2 = std::make_shared<LSTMCell>(squeeze_2, H_t_2, C_t_2, weightsNode_2, reccurrenceWeightsNode_2, biasNode_2, hiddenSize);
 
-    auto unsqueeze_2_const = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64, ngraph::Shape{1}, squeeze_axes);
-    auto unsqueeze_2 = std::make_shared<ngraph::opset5::Unsqueeze>(lstm_2->output(0), unsqueeze_2_const);
+    auto unsqueeze_2_const = std::make_shared<Constant>(element::i64, Shape{1}, squeeze_axes);
+    auto unsqueeze_2 = std::make_shared<Unsqueeze>(lstm_2->output(0), unsqueeze_2_const);
     // body - outputs
     auto H_o_2 = lstm_2->output(0);
     auto C_o_2 = lstm_2->output(1);
     auto unsqueeze_o_2 = unsqueeze_2->output(0);
 
-    auto body_2 = std::make_shared<ngraph::Function>(ngraph::OutputVector{unsqueeze_o_2, H_o_2, C_o_2}, ngraph::ParameterVector {X_2, H_t_2, C_t_2});
+    auto body_2 = std::make_shared<Function>(OutputVector{unsqueeze_o_2, H_o_2, C_o_2}, ParameterVector {X_2, H_t_2, C_t_2});
     // TI construction
-    auto tensor_iterator_2 = std::make_shared<ngraph::opset5::TensorIterator>();
+    auto tensor_iterator_2 = std::make_shared<TensorIterator>();
     tensor_iterator_2->set_body(body_2);
     tensor_iterator_2->set_sliced_input(X_2, inbetween_squeeze, 0, 1, 1, -1, 0);
     tensor_iterator_2->set_merged_input(H_t_2, hidden_memory_2_constant, H_o_2);
@@ -376,103 +384,109 @@ void MultipleLSTMCellTest::CreatePureTensorIteratorModel() {
     out_hidden_2.get_tensor().set_element_type(ngPrc);
     out_cell_2.get_tensor().set_element_type(ngPrc);
     tensor_iterator_2->validate_and_infer_types();
-    auto final_reshape_pattern = std::make_shared<ngraph::opset5::Constant>(ngraph::element::i64,
-                                                                        ngraph::Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
-    auto final_reshape = std::make_shared<ngraph::opset5::Reshape>(out_unsqueeze_2, final_reshape_pattern, false);
+    auto final_reshape_pattern = std::make_shared<Constant>(element::i64,
+                                                                        Shape{4}, std::vector<size_t>({1, 1, 1, hiddenSize}));
+    auto final_reshape = std::make_shared<Reshape>(out_unsqueeze_2, final_reshape_pattern, false);
 
-    function = std::make_shared<ngraph::Function>(final_reshape, input_parameter, "PureTI");
+    function = std::make_shared<Function>(final_reshape, input_parameter, "PureTI");
 }
 
-void MultipleLSTMCellTest::Run() {
-    SKIP_IF_CURRENT_TEST_IS_DISABLED()
+void MultipleLSTMCellTest::InitMemory() {
     InferenceEngine::TensorDesc state_description(InferenceEngine::Precision::FP32,
                                                   InferenceEngine::SizeVector({1, hiddenSize}),
                                                   InferenceEngine::Layout::NC);
-    LoadNetwork();
     IE_SUPPRESS_DEPRECATED_START
     auto states = executableNetwork.QueryState();
     for (auto& state : states) {
         auto name = state.GetName();
-        if (name == "cell_memory") {
+        if (name.find("cell_state_1") != std::string::npos) {
             auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state_description,
                                                                        cell_memory_init.data(), cell_memory_init.size());
             state.SetState(blob);
-        } else if (name == "hidden_memory") {
+        } else if (name.find("hidden_state_1") != std::string::npos) {
             auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state_description,
                                                                        hidden_memory_init.data(), hidden_memory_init.size());
             state.SetState(blob);
-        } else if (name == "cell_memory_2") {
+        } else if (name.find("cell_state_2") != std::string::npos) {
             auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state_description,
-                cell_memory_init.data(), cell_memory_init.size());
+                                                                       cell_memory_init.data(), cell_memory_init.size());
             state.SetState(blob);
-        } else if (name == "hidden_memory_2") {
+        } else if (name.find("hidden_state_2") != std::string::npos) {
             auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state_description,
-                hidden_memory_init.data(), hidden_memory_init.size());
+                                                                       hidden_memory_init.data(), hidden_memory_init.size());
             state.SetState(blob);
         } else {
             GTEST_FAIL() << "unknown memory state";
         }
     }
     IE_SUPPRESS_DEPRECATED_END
-    GenerateInputs();
-    Infer();
-    switchToNgraphFriendlyModel();
-    Validate();
 }
 
-void MultipleLSTMCellTest::RunLowLatency(bool regular_api) {
-    SKIP_IF_CURRENT_TEST_IS_DISABLED()
-    InferenceEngine::TensorDesc state_description(InferenceEngine::Precision::FP32,
-                                                  InferenceEngine::SizeVector({1, hiddenSize}),
-                                                  InferenceEngine::Layout::NC);
+void MultipleLSTMCellTest::ApplyLowLatency() {
     // Calculate values after LowLatency transformation
     CreatePureTensorIteratorModel();
-    if (regular_api) {
+    if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY) {
+        function->validate_nodes_and_infer_types();
+        // Apply LowLatency (insert Assigns/ReadValues) and UnrollTensorIterator
+        pass::Manager manager;
+        NGRAPH_SUPPRESS_DEPRECATED_START
+        manager.register_pass<ngraph::pass::LowLatency>();
+        NGRAPH_SUPPRESS_DEPRECATED_END // LowLatency enables UnrollTI
+        manager.run_passes(function);
+        bool ti_found = helpers::is_tensor_iterator_exist(function);
+        EXPECT_EQ(ti_found, true);
+        LoadNetwork();
+    } else if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2) {
+        function->validate_nodes_and_infer_types();
+        // Apply LowLatency (insert Assigns/ReadValues) and UnrollTensorIterator
+
+        pass::Manager manager;
+        manager.register_pass<pass::LowLatency2>();
+        manager.run_passes(function);
+        bool ti_found = helpers::is_tensor_iterator_exist(function);
+        EXPECT_EQ(ti_found, false);
+        LoadNetwork();
+    } else if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY_REGULAR_API) {
         cnnNetwork = InferenceEngine::CNNNetwork{function};
+        IE_SUPPRESS_DEPRECATED_START
         InferenceEngine::LowLatency(cnnNetwork);
+        IE_SUPPRESS_DEPRECATED_END
+
+        bool ti_found = helpers::is_tensor_iterator_exist(cnnNetwork.getFunction());
+        EXPECT_EQ(ti_found, true);
+
+        ConfigureNetwork();
+        executableNetwork = core->LoadNetwork(cnnNetwork, targetDevice, configuration);
+    } else if (transformation == ngraph::helpers::MemoryTransformation::LOW_LATENCY_V2_REGULAR_API) {
+        cnnNetwork = InferenceEngine::CNNNetwork{function};
+        InferenceEngine::lowLatency2(cnnNetwork);
+
+        bool ti_found = helpers::is_tensor_iterator_exist(cnnNetwork.getFunction());
+        EXPECT_EQ(ti_found, false);
+
         ConfigureNetwork();
         executableNetwork = core->LoadNetwork(cnnNetwork, targetDevice, configuration);
+    }
+}
+
+void MultipleLSTMCellTest::Run() {
+    SKIP_IF_CURRENT_TEST_IS_DISABLED()
+    if (transformation != ngraph::helpers::MemoryTransformation::NONE) {
+        ApplyLowLatency();
     } else {
-        function->validate_nodes_and_infer_types();
-        // Apply LowLatency (insert Assigns/ReadValues) and UnrollTensorIterator
-        ngraph::pass::Manager manager;
-        manager.register_pass<ngraph::pass::LowLatency>(); // LowLatency enables UnrollTI
-        manager.run_passes(function);
         LoadNetwork();
     }
-    IE_SUPPRESS_DEPRECATED_START
-    auto states = executableNetwork.QueryState();
-    for (auto& state : states) {
-        auto name = state.GetName();
-        if (name.find("cell_state_1") != std::string::npos) {
-            auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state_description,
-                                                                       cell_memory_init.data(), cell_memory_init.size());
-            state.SetState(blob);
-        } else if (name.find("hidden_state_1") != std::string::npos) {
-            auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state_description,
-                                                                       hidden_memory_init.data(), hidden_memory_init.size());
-            state.SetState(blob);
-        } else if (name.find("cell_state_2") != std::string::npos) {
-            auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state_description,
-                                                                       cell_memory_init.data(), cell_memory_init.size());
-            state.SetState(blob);
-        } else if (name.find("hidden_state_2") != std::string::npos) {
-            auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state_description,
-                                                                       hidden_memory_init.data(), hidden_memory_init.size());
-            state.SetState(blob);
-        } else {
-            GTEST_FAIL() << "unknown memory state";
-        }
-    }
-    IE_SUPPRESS_DEPRECATED_END
+
+    InitMemory();
     GenerateInputs();
     Infer();
 
-    // Calculate ref values for Unrolled TI
-    CreatePureTensorIteratorModel();
-    ngraph::pass::Manager manager_2;
-    manager_2.register_pass<ngraph::pass::UnrollTensorIterator>();
-    manager_2.run_passes(function);
+    // Calculate ref values
+    if (transformation == ngraph::helpers::MemoryTransformation::NONE) {
+        switchToNgraphFriendlyModel();
+    } else {
+        CreatePureTensorIteratorModel();
+    }
     Validate();
 }
 }  // namespace SubgraphTestsDefinitions
diff --git a/inference-engine/tests/ngraph_helpers/ngraph_functions/include/ngraph_functions/utils/ngraph_helpers.hpp b/inference-engine/tests/ngraph_helpers/ngraph_functions/include/ngraph_functions/utils/ngraph_helpers.hpp
index de015677b4a2b4..de2dbab061289d 100644
--- a/inference-engine/tests/ngraph_helpers/ngraph_functions/include/ngraph_functions/utils/ngraph_helpers.hpp
+++ b/inference-engine/tests/ngraph_helpers/ngraph_functions/include/ngraph_functions/utils/ngraph_helpers.hpp
@@ -214,6 +214,15 @@ enum class SequenceTestsMode {
     CONVERT_TO_TI_RAND_SEQ_LEN_PARAM,
 };
 
+enum class MemoryTransformation {
+    NONE,
+    LOW_LATENCY,
+    LOW_LATENCY_REGULAR_API,
+    LOW_LATENCY_V2,
+    LOW_LATENCY_V2_REGULAR_API,
+    LOW_LATENCY_V2_ORIGINAL_INIT
+};
+
 std::ostream &operator<<(std::ostream &os, const ReductionType &m);
 std::ostream &operator<<(std::ostream &os, const PadMode &m);
 
@@ -297,5 +306,7 @@ std::ostream& operator<<(std::ostream & os, TensorIteratorBody type);
 
 std::ostream& operator<<(std::ostream & os, SequenceTestsMode type);
 
+std::ostream& operator<<(std::ostream & os, MemoryTransformation type);
+
 }  // namespace helpers
 }  // namespace ngraph
diff --git a/inference-engine/tests/ngraph_helpers/ngraph_functions/src/utils/ngraph_helpers.cpp b/inference-engine/tests/ngraph_helpers/ngraph_functions/src/utils/ngraph_helpers.cpp
index 5de50203ba28b3..2c5a07540b044b 100644
--- a/inference-engine/tests/ngraph_helpers/ngraph_functions/src/utils/ngraph_helpers.cpp
+++ b/inference-engine/tests/ngraph_helpers/ngraph_functions/src/utils/ngraph_helpers.cpp
@@ -817,5 +817,32 @@ std::ostream& operator<<(std::ostream & os, SequenceTestsMode type) {
     }
     return os;
 }
+
+std::ostream& operator<<(std::ostream & os, MemoryTransformation type) {
+    switch (type) {
+        case MemoryTransformation::NONE:
+            os << "NONE";
+            break;
+        case MemoryTransformation::LOW_LATENCY_V2:
+            os << "LOW_LATENCY_V2";
+            break;
+        case MemoryTransformation::LOW_LATENCY:
+            os << "LOW_LATENCY";
+            break;
+        case MemoryTransformation::LOW_LATENCY_V2_REGULAR_API:
+            os << "LOW_LATENCY_V2_REGULAR_API";
+            break;
+        case MemoryTransformation::LOW_LATENCY_REGULAR_API:
+            os << "LOW_LATENCY_REGULAR_API";
+            break;
+        case MemoryTransformation::LOW_LATENCY_V2_ORIGINAL_INIT:
+            os << "LOW_LATENCY_V2_ORIGINAL_INIT";
+            break;
+        default:
+            throw std::runtime_error("NOT_SUPPORTED_TYPE");
+    }
+    return os;
+}
+
 }  // namespace helpers
 }  // namespace ngraph
diff --git a/model-optimizer/mo/back/offline_transformations.py b/model-optimizer/mo/back/offline_transformations.py
index a363a1ca250756..1df5f6cb7a0adf 100644
--- a/model-optimizer/mo/back/offline_transformations.py
+++ b/model-optimizer/mo/back/offline_transformations.py
@@ -11,7 +11,7 @@ def get_available_transformations():
     try:
         from openvino.offline_transformations import ApplyLowLatencyTransformation  # pylint: disable=import-error,no-name-in-module
         return {
-            'LowLatency': ApplyLowLatencyTransformation,
+            'LowLatency2': ApplyLowLatencyTransformation,
         }
     except Exception as e:
         return {}
diff --git a/model-optimizer/mo/utils/cli_parser.py b/model-optimizer/mo/utils/cli_parser.py
index 21d2873b54ed2c..e6b2f2d2517f83 100644
--- a/model-optimizer/mo/utils/cli_parser.py
+++ b/model-optimizer/mo/utils/cli_parser.py
@@ -8,6 +8,7 @@
 import re
 from collections import OrderedDict
 from itertools import zip_longest
+from distutils.util import strtobool
 
 import numpy as np
 
@@ -257,9 +258,9 @@ def get_common_cli_parser(parser: argparse.ArgumentParser = None):
                               help='Apply additional transformations. ' +
                                    'Usage: "--transform transformation_name1[args],transformation_name2..." ' +
                                    'where [args] is key=value pairs separated by semicolon. ' +
-                                   'Examples: "--transform LowLatency" or ' +
-                                   '          "--transform LowLatency[num_iterations=2]" ' +
-                                   'Available transformations: "LowLatency"',
+                                   'Examples: "--transform LowLatency2" or ' +
+                                   '          "--transform LowLatency2[use_const_initializer=False]" ' +
+                                   'Available transformations: "LowLatency2"',
                               default="")
     common_group.add_argument('--disable_fusing',
                               help='Turn off fusing of linear operations to Convolution',
@@ -1151,6 +1152,14 @@ def isfloat(value):
         return False
 
 
+def isbool(value):
+    try:
+        strtobool(value)
+        return True
+    except ValueError:
+        return False
+
+
 def convert_string_to_real_type(value: str):
     values = value.split(',')
     for i in range(len(values)):
@@ -1159,6 +1168,8 @@ def convert_string_to_real_type(value: str):
             values[i] = int(value)
         elif isfloat(value):
             values[i] = float(value)
+        elif isbool(value):
+            values[i] = strtobool(value)
 
     return values[0] if len(values) == 1 else values
 
diff --git a/model-optimizer/unit_tests/mo/utils/cli_parser_test.py b/model-optimizer/unit_tests/mo/utils/cli_parser_test.py
index 68a9994d2068c9..c6c1c96e46a66e 100644
--- a/model-optimizer/unit_tests/mo/utils/cli_parser_test.py
+++ b/model-optimizer/unit_tests/mo/utils/cli_parser_test.py
@@ -905,64 +905,65 @@ def test_empty(self):
         self.assertEqual(parse_transform(""), [])
 
     def test_single_pass(self):
-        self.assertEqual(parse_transform("LowLatency"), [("LowLatency", {})])
+        self.assertEqual(parse_transform("LowLatency2"), [("LowLatency2", {})])
 
     def test_single_pass_with_args(self):
-        self.assertEqual(parse_transform("LowLatency[num_iterations=2]"),
-                         [("LowLatency", {"num_iterations": 2})])
+        self.assertEqual(parse_transform("LowLatency2[use_const_initializer=True]"),
+                         [("LowLatency2", {"use_const_initializer": True})])
 
     def test_single_pass_with_multiple_args(self):
-        self.assertEqual(parse_transform("LowLatency[num_iterations=2;dummy_attr=3.14]"),
-                         [("LowLatency", {"num_iterations": 2, "dummy_attr": 3.14})])
+        self.assertEqual(parse_transform("LowLatency2[use_const_initializer=True;dummy_attr=3.14]"),
+                         [("LowLatency2", {"use_const_initializer": True, "dummy_attr": 3.14})])
 
     def test_multiple_passes_with_args(self):
-        self.assertEqual(parse_transform("LowLatency[num_iterations=2],DummyPass[type=ReLU]"),
-                         [("LowLatency", {"num_iterations": 2}),
+        self.assertEqual(parse_transform("LowLatency2[use_const_initializer=True],DummyPass[type=ReLU]"),
+                         [("LowLatency2", {"use_const_initializer": True}),
                           ("DummyPass", {"type": "ReLU"})])
 
     def test_multiple_passes_with_args2(self):
-        self.assertEqual(parse_transform("LowLatency[num_iterations=2,3,4.15],DummyPass1,DummyPass2[types=ReLU,PReLU;values=1,2,3]"),
-                         [("LowLatency",  {"num_iterations": [2,3,4.15]}),
+        self.assertEqual(parse_transform("LowLatency2[use_const_initializer=True,False],DummyPass1,"
+                                         "DummyPass2[types=ReLU,PReLU;values=1,2,3]"),
+                         [("LowLatency2",  {"use_const_initializer": [True, False]}),
                           ("DummyPass1",  {}),
                           ("DummyPass2",  {"types": ["ReLU", "PReLU"], "values": [1,2,3]})])
 
     def test_multiple_passes_no_args(self):
-        self.assertEqual(parse_transform("DummyPass,LowLatency2"),
-                         [("DummyPass", {}), ("LowLatency2", {})])
+        self.assertEqual(parse_transform("DummyPass,LowLatency22"),
+                         [("DummyPass", {}), ("LowLatency22", {})])
 
     def test_single_pass_neg(self):
-        self.assertRaises(Error, parse_transform, "LowLatency!")
+        self.assertRaises(Error, parse_transform, "LowLatency2!")
 
     def test_multiple_passes_neg(self):
-        self.assertRaises(Error, parse_transform, "LowLatency;DummyPass")
+        self.assertRaises(Error, parse_transform, "LowLatency2;DummyPass")
 
     def test_single_pass_with_args_neg1(self):
-        self.assertRaises(Error, parse_transform, "LowLatency[=2]")
+        self.assertRaises(Error, parse_transform, "LowLatency2[=2]")
 
     def test_single_pass_with_args_neg2(self):
-        self.assertRaises(Error, parse_transform, "LowLatency[key=]")
+        self.assertRaises(Error, parse_transform, "LowLatency2[key=]")
 
     def test_single_pass_with_args_neg3(self):
-        self.assertRaises(Error, parse_transform, "LowLatency[]")
+        self.assertRaises(Error, parse_transform, "LowLatency2[]")
 
     def test_single_pass_with_args_neg4(self):
-        self.assertRaises(Error, parse_transform, "LowLatency[key=value;]")
+        self.assertRaises(Error, parse_transform, "LowLatency2[key=value;]")
 
     def test_single_pass_with_args_neg5(self):
-        self.assertRaises(Error, parse_transform, "LowLatency[value]")
+        self.assertRaises(Error, parse_transform, "LowLatency2[value]")
 
     def test_single_pass_with_args_neg6(self):
-        self.assertRaises(Error, parse_transform, "LowLatency[key=value")
+        self.assertRaises(Error, parse_transform, "LowLatency2[key=value")
 
     @patch("mo.back.offline_transformations.get_available_transformations")
     def test_check_low_latency_is_available(self, available_transformations):
-        available_transformations.return_value = {"LowLatency": None}
+        available_transformations.return_value = {"LowLatency2": None}
         try:
-            check_available_transforms([("LowLatency" ,"")], True)
+            check_available_transforms([("LowLatency2", "")], True)
         except Error as e:
             self.assertTrue(False, "Exception \"{}\" is unexpected".format(e))
 
     @patch("mo.back.offline_transformations.get_available_transformations")
     def test_check_dummy_pass_is_available(self, available_transformations):
-        available_transformations.return_value = {"LowLatency": None}
+        available_transformations.return_value = {"LowLatency2": None}
         self.assertRaises(Error, check_available_transforms, [("DummyPass", "")], True)
diff --git a/ngraph/core/include/ngraph/pass/low_latency.hpp b/ngraph/core/include/ngraph/pass/low_latency.hpp
index 86757edb800836..507ffe3a21b145 100644
--- a/ngraph/core/include/ngraph/pass/low_latency.hpp
+++ b/ngraph/core/include/ngraph/pass/low_latency.hpp
@@ -8,13 +8,14 @@
 #include <vector>
 
 #include <ngraph/pass/graph_rewrite.hpp>
+#include <ngraph/pass/pass.hpp>
 
 namespace ngraph
 {
     namespace pass
     {
         /**
-         * @brief The transformation finds all TensorIterator layers in the network,
+         * @brief The transformation finds all TensorIterator/Loop layers in the network,
          * processes all back edges that describe a connection between Result and Parameter
          * of the TensorIterator body,and inserts ReadValue layer between Parameter
          * and the next layers after this Parameter, and Assign layer after the layers
@@ -42,11 +43,50 @@ namespace ngraph
          * by step, the states will store between inferences.
          */
 
-        class NGRAPH_API LowLatency : public ngraph::pass::MatcherPass
+        class NGRAPH_DEPRECATED("Use LowLatency2 instead.") NGRAPH_API LowLatency
+            : public ngraph::pass::MatcherPass
         {
         public:
             NGRAPH_RTTI_DECLARATION;
             LowLatency();
         };
+
+        /**
+         * @brief The transformation finds all TensorIterator/Loop layers in the network,
+         * processes all back edges that describe a connection between Result and Parameter
+         * of the TensorIterator/Loop bodies,and inserts ReadValue and Assign layers at the
+         * input and output corresponding to this back edge.
+         * Supported platforms: CPU, GNA.
+         *
+         * The example below describes the changes made by the transformation
+         *  [] - TensorIterator body
+         *  () - new layer
+         *  BE - back-edge
+         *
+         *  before applying the transformation:
+         *  -> input1[BE_1 -> Parameter -> Layers ... -> Result  -> BE_1 ]output1->
+         *
+         *  after applying the transformation:
+         *  ->(ReadValue)-> input1[BE_1 ->Parameter->Layers ...->Result->BE_1]output1 ->(Assign)
+         *                                                                      \
+         *                                                                       ->...
+         * After applying the transformation, the resulting network can be inferred
+         * step by step, the states will store between inferences.
+         */
+        class NGRAPH_API LowLatency2 : public ngraph::pass::FunctionPass
+        {
+        public:
+            NGRAPH_RTTI_DECLARATION;
+
+            explicit LowLatency2(bool use_const_initializer = true)
+                : m_use_const_initializer(use_const_initializer)
+            {
+            }
+
+            bool run_on_function(std::shared_ptr<ngraph::Function> f) override;
+
+        private:
+            bool m_use_const_initializer;
+        };
     } // namespace pass
 } // namespace ngraph
diff --git a/ngraph/core/src/op/tensor_iterator.cpp b/ngraph/core/src/op/tensor_iterator.cpp
index 5252d2124f24c3..35162ef3cebbfa 100644
--- a/ngraph/core/src/op/tensor_iterator.cpp
+++ b/ngraph/core/src/op/tensor_iterator.cpp
@@ -129,7 +129,6 @@ void op::v0::TensorIterator::validate_and_infer_types()
                 m_body->get_results().at(merged_input_description->m_body_value_index)->input(0);
             ends.push_back(body_value.get_node()->shared_from_this());
 
-            auto body_value_partial_shape = body_value.get_partial_shape();
             auto body_parameter =
                 m_body->get_parameters().at(merged_input_description->m_body_parameter_index);
 
diff --git a/ngraph/core/src/pass/low_latency.cpp b/ngraph/core/src/pass/low_latency.cpp
index ea5effce4abdfb..d290eb14b7e4af 100644
--- a/ngraph/core/src/pass/low_latency.cpp
+++ b/ngraph/core/src/pass/low_latency.cpp
@@ -6,12 +6,29 @@
 
 #include <memory>
 
+#include <ngraph/log.hpp>
 #include <ngraph/opsets/opset6.hpp>
+#include <ngraph/opsets/opset7.hpp>
 #include <ngraph/pattern/op/wrap_type.hpp>
+#include <ngraph/rt_info.hpp>
 #include <ngraph/variant.hpp>
 
+NGRAPH_RTTI_DEFINITION(ngraph::pass::LowLatency2, "LowLatency2", 0);
+
+NGRAPH_SUPPRESS_DEPRECATED_START
 NGRAPH_RTTI_DEFINITION(ngraph::pass::LowLatency, "LowLatency", 0);
 
+using namespace std;
+using namespace ngraph;
+
+namespace
+{
+    string generate_variable_name(const string& op_name, const string& param_name, int variable_idx)
+    {
+        return op_name + "/" + param_name + "/" + "variable_" + to_string(variable_idx);
+    }
+
+} // namespace
 ngraph::pass::LowLatency::LowLatency()
 {
     auto tensor_iterator = ngraph::pattern::wrap_type<opset6::TensorIterator, opset6::Loop>();
@@ -58,11 +75,12 @@ ngraph::pass::LowLatency::LowLatency()
                 const auto& inputs_to = func->get_parameters()
                                             .at(merged_in->m_body_parameter_index)
                                             ->get_output_target_inputs(0);
-                const std::string variable_name(sub_graph_op->get_friendly_name() + "/" +
-                                                func->get_parameters()
-                                                    .at(merged_in->m_body_parameter_index)
-                                                    ->get_friendly_name() +
-                                                "/variable_" + std::to_string(variable_id));
+                const std::string variable_name(
+                    generate_variable_name(sub_graph_op->get_friendly_name(),
+                                           func->get_parameters()
+                                               .at(merged_in->m_body_parameter_index)
+                                               ->get_friendly_name(),
+                                           variable_id));
                 auto variable = std::make_shared<Variable>(
                     VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name});
                 auto read_value = std::make_shared<opset6::ReadValue>(
@@ -90,3 +108,178 @@ ngraph::pass::LowLatency::LowLatency()
     auto m = std::make_shared<ngraph::pattern::Matcher>(tensor_iterator, "LowLatency");
     register_matcher(m, callback);
 }
+NGRAPH_SUPPRESS_DEPRECATED_END
+
+void UnrollSingleIteration(const shared_ptr<op::util::SubGraphOp>& sub_graph_op,
+                           const shared_ptr<Function>& outer_f)
+{
+    using namespace opset7;
+
+    const auto& params = sub_graph_op->get_function()->get_parameters();
+    const auto& results = sub_graph_op->get_function()->get_results();
+
+    // before: Layer1 -> TI [input -> bodyParameter -> Layer2 -> ...]
+    // after:  Layer1 -> Layer2 ->...
+    for (const auto& in : sub_graph_op->get_input_descriptions())
+    {
+        const auto& connect_to = sub_graph_op->get_input_source_output(in->m_input_index);
+        for (auto& output : params.at(in->m_body_parameter_index)->outputs())
+        {
+            output.replace(connect_to);
+        }
+    }
+
+    // before: TI [...-> Layer1 -> Result -> output] -> Layer2 -> ...
+    // after:  ...-> Layer1 -> Layer2 -> ...
+    NodeVector new_ops;
+    for (const auto& out : sub_graph_op->get_output_descriptions())
+    {
+        const auto& connect_to = results.at(out->m_body_value_index)->get_input_source_output(0);
+        for (auto& input_to : sub_graph_op->output(out->m_output_index).get_target_inputs())
+        {
+            // create IE output name
+            std::string out_name = sub_graph_op->get_friendly_name();
+            if (sub_graph_op->get_output_size() != 1)
+                out_name += "." + std::to_string(out->m_output_index);
+
+            // IECompatibility: insert identity (Unsqueeze + Squeeze) to store the TensorIterator
+            // output names
+            auto axis_1 = Constant::create(ngraph::element::i64, ngraph::Shape{1}, {1});
+            auto identity_1 = std::make_shared<Unsqueeze>(connect_to, axis_1);
+            auto identity_2 = std::make_shared<Squeeze>(identity_1, axis_1);
+            identity_2->set_friendly_name(out_name);
+            new_ops.push_back(identity_1);
+            new_ops.push_back(identity_2);
+
+            input_to.replace_source_output(identity_2);
+        }
+    }
+    outer_f->add_sinks(sub_graph_op->get_function()->get_sinks());
+    ngraph::copy_runtime_info(sub_graph_op, sub_graph_op->get_function()->get_ops());
+    ngraph::copy_runtime_info(sub_graph_op, new_ops);
+}
+
+Output<Node> create_init_subgraph(const shared_ptr<op::util::SubGraphOp>& sub_graph_op,
+                                  const Output<Node>& in_node)
+{
+    using namespace opset7;
+
+    auto const_zero = make_shared<Constant>(in_node.get_element_type(), Shape{1}, 0);
+    auto shape_of = make_shared<ShapeOf>(in_node);
+    auto broadcast = make_shared<Broadcast>(const_zero, shape_of);
+    copy_runtime_info(sub_graph_op, {const_zero, shape_of, broadcast});
+    return broadcast->output(0);
+}
+
+bool pass::LowLatency2::run_on_function(shared_ptr<Function> f)
+{
+    using namespace opset7;
+
+    SinkVector assigns;
+    for (const auto& op : f->get_ordered_ops())
+    {
+        if (const auto& sub_graph_op = dynamic_pointer_cast<op::util::SubGraphOp>(op))
+        {
+            int64_t variable_id = 0;
+            const auto& func = sub_graph_op->get_function();
+            const auto& params = func->get_parameters();
+            for (const auto& in : sub_graph_op->get_input_descriptions())
+            {
+                // Process all back edges
+                if (const auto& merged_in =
+                        dynamic_pointer_cast<op::util::SubGraphOp::MergedInputDescription>(in))
+                {
+                    // create new Variable
+                    const string& param_name =
+                        params.at(merged_in->m_body_parameter_index)->get_friendly_name();
+                    const string& var_name = generate_variable_name(
+                        sub_graph_op->get_friendly_name(), param_name, variable_id);
+
+                    const auto& input = sub_graph_op->input(merged_in->m_input_index);
+                    if (std::dynamic_pointer_cast<op::ReadValueBase>(
+                            input.get_source_output().get_node_shared_ptr()) != nullptr)
+                    {
+                        NGRAPH_DEBUG
+                            << "LowLatency2 transformation cannot be applied because the "
+                            << "ReadValue node is already an input to the TensorIterator."
+                            << "LowLatency2 transformation may have already been applied, please "
+                            << "do not call it more then once.";
+                        return false;
+                    }
+
+                    const auto& param = sub_graph_op->get_function()->get_parameters().at(
+                        merged_in->m_body_parameter_index);
+                    for (const auto& in_to : param->output(0).get_target_inputs())
+                    {
+                        if (dynamic_cast<op::ReadValueBase*>(in_to.get_node()) != nullptr)
+                        {
+                            NGRAPH_DEBUG
+                                << "LowLatency2 transformation cannot be applied because the "
+                                << "ReadValue node is already inside the TensorIterator. "
+                                << "LowLatency transformation may have been applied, please do "
+                                << "not call LowLatency2 after LowLatency.";
+                            return false;
+                        }
+                    }
+
+                    VariableInfo var_info{PartialShape::dynamic(), element::dynamic, var_name};
+                    auto variable = make_shared<Variable>(var_info);
+
+                    // insert ReadValue
+                    // Layers -> [new op: ReadValue] -> Subgraph operation
+                    Output<Node> read_value_in = input.get_source_output();
+                    if (m_use_const_initializer)
+                    {
+                        read_value_in = create_init_subgraph(sub_graph_op, read_value_in);
+                    }
+                    auto read_value = make_shared<ReadValue>(read_value_in, variable);
+                    input.replace_source_output(read_value->output(0));
+                    read_value->set_friendly_name(var_name);
+                    ngraph::copy_runtime_info(sub_graph_op, read_value);
+
+                    /* insert Assign
+                    // Subgraph operation -> [new op: Assign]
+                    //                    \
+                    //                      ---> Layers -> ...
+                    */
+                    const auto& out_desc = sub_graph_op->get_output_descriptions();
+                    bool is_output_exist = std::any_of(
+                        out_desc.begin(),
+                        out_desc.end(),
+                        [&merged_in](
+                            const std::shared_ptr<op::util::SubGraphOp::OutputDescription>& out) {
+                            return out->m_body_value_index == merged_in->m_body_value_index;
+                        });
+                    // Create new output if it doesn't exist.
+                    if (!is_output_exist)
+                    {
+                        sub_graph_op->get_iter_value(
+                            func->get_results().at(merged_in->m_body_value_index));
+                    }
+                    for (const auto& out : sub_graph_op->get_output_descriptions())
+                    {
+                        if (out->m_body_value_index == merged_in->m_body_value_index)
+                        {
+                            auto assign = make_shared<Assign>(
+                                sub_graph_op->output(out->m_output_index), variable);
+                            ngraph::copy_runtime_info(sub_graph_op, assign);
+                            // control dependency so that ReadValue is processed before Assign
+                            assign->add_control_dependency(read_value);
+                            assigns.emplace_back(assign);
+                            break;
+                        }
+                    }
+                }
+
+                variable_id++;
+            }
+
+            if (sub_graph_op->get_num_iterations() == 1)
+            {
+                UnrollSingleIteration(sub_graph_op, f);
+            }
+        }
+    }
+    f->add_sinks(assigns);
+    return true;
+}

From 6a18b45337bd64542155416415d53e46627d31b9 Mon Sep 17 00:00:00 2001
From: Yegor Kruglov <yegor.kruglov@intel.com>
Date: Mon, 7 Jun 2021 16:44:16 +0300
Subject: [PATCH 005/178] [MO] Implementation of names uniqueness check (#5651)

* added new transformation to check the uniqueness of nodes names

* added unittest

* remove redundant line

* conversation resolving

* updated unittest

* added new unittest, added check for uniqueness of new node name

* added a description

* added renaming of several results with the same name and unittest for this case

* another implementation, updated unittests

* added a comment

* updated comments

* added comment to the nodes_with_equal_names func

* added a condition

* added a result name check in unittests
---
 model-optimizer/automation/package_BOM.txt    |  1 +
 .../extensions/back/names_uniqueness_check.py | 67 +++++++++++++++++
 .../back/names_uniqueness_check_test.py       | 71 +++++++++++++++++++
 3 files changed, 139 insertions(+)
 create mode 100644 model-optimizer/extensions/back/names_uniqueness_check.py
 create mode 100644 model-optimizer/unit_tests/extensions/back/names_uniqueness_check_test.py

diff --git a/model-optimizer/automation/package_BOM.txt b/model-optimizer/automation/package_BOM.txt
index 128c9c98117216..be0dc1c37e0dd5 100644
--- a/model-optimizer/automation/package_BOM.txt
+++ b/model-optimizer/automation/package_BOM.txt
@@ -37,6 +37,7 @@ extensions/back/LRNToNorm.py
 extensions/back/MarkNodesWithShapeValues.py
 extensions/back/MatMulNormalizer.py
 extensions/back/MaxPool.py
+extensions/back/names_uniqueness_check.py
 extensions/back/NormalizeToNormalizeL2.py
 extensions/back/op_versioning.py
 extensions/back/OptimizeTransposeReshapeSequence.py
diff --git a/model-optimizer/extensions/back/names_uniqueness_check.py b/model-optimizer/extensions/back/names_uniqueness_check.py
new file mode 100644
index 00000000000000..188df9f4427fbd
--- /dev/null
+++ b/model-optimizer/extensions/back/names_uniqueness_check.py
@@ -0,0 +1,67 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+from collections import defaultdict
+from extensions.back.pass_separator import BackFinish
+from mo.back.replacement import BackReplacementPattern
+from mo.graph.graph import Graph, rename_node
+
+
+def nodes_with_equal_names(graph: Graph):
+    """
+    :param graph: Graph to operate on
+    :return: Dictionary with node names as keys and a list of their corresponding nodes as values
+    """
+    names_dict = defaultdict(list)
+    for node in graph.get_op_nodes():
+        node_name = node.soft_get('name', node.id)
+        names_dict[node_name].append(node)
+    return names_dict
+
+
+def make_node_names_unique(nodes: list, node_names: set):
+    """
+    :param nodes: List with nodes matching a specific name
+    :param node_names: Set with all node names contained in the graph
+    :return: None
+
+    Result nodes will be renamed only when it is absolutely necessary(if there are several Result nodes with the same name).
+    Function finds a position of Result nodes in the "nodes" list, take the first and rename all other nodes.
+    If the "nodes" list does not contain Result nodes, then all nodes starting from the second one will be renamed.
+    All new names are added to the "node_names" set.
+    """
+    results_pos = [idx for idx, node in enumerate(nodes) if node.op == 'Result']
+    node_position_to_keep = 0
+    if len(results_pos) != 0:
+        node_position_to_keep = results_pos[0]
+    for idx, node in enumerate(nodes):
+        if idx != node_position_to_keep:
+            new_node_name = node.soft_get('name', node.id) + '_' + str(idx)
+            # preparing a new unique name for the node
+            while new_node_name in node_names:
+                new_node_name += '_' + str(idx)
+            node_names.add(new_node_name)
+            rename_node(node, new_node_name)
+
+
+class NamesUniquenessCheck(BackReplacementPattern):
+    """
+    If there are several layers with the same name in the original model and they are saved in the IR, IE will fail with
+    the invalid IR error. IE checks the uniqueness of the names and, if it is not true, throws an exception. The way how
+    to fix it on the MO side is to rename this nodes (one node will remain with the original name). Since we prefer to
+    save framework names for the output nodes, nodes with op=Result will not be renamed, except the case when there are
+    several Result nodes with the same name.
+    """
+    enabled = True
+
+    def run_after(self):
+        return [BackFinish]
+
+    def run_before(self):
+        return []
+
+    def find_and_replace_pattern(self, graph: Graph):
+        names_to_nodes = nodes_with_equal_names(graph)
+        node_names = set(names_to_nodes.keys())
+        for nodes in names_to_nodes.values():
+            if len(nodes) > 1:
+                make_node_names_unique(nodes, node_names)
diff --git a/model-optimizer/unit_tests/extensions/back/names_uniqueness_check_test.py b/model-optimizer/unit_tests/extensions/back/names_uniqueness_check_test.py
new file mode 100644
index 00000000000000..94f83af1a1b18c
--- /dev/null
+++ b/model-optimizer/unit_tests/extensions/back/names_uniqueness_check_test.py
@@ -0,0 +1,71 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import unittest
+
+from extensions.back.names_uniqueness_check import NamesUniquenessCheck
+from mo.graph.graph import Node
+from unit_tests.utils.graph import build_graph
+
+
+class TestNamesUniquenessCheck(unittest.TestCase):
+
+    def test_1(self):
+        graph = build_graph(
+            nodes_attrs={
+                'input': {'kind': 'op', 'op': 'Parameter', 'name': 'node'},
+                'cast': {'kind': 'op', 'op': 'Cast', 'name': 'node'},
+                'result': {'kind': 'op', 'op': 'Result', 'name': 'node'}
+            },
+            edges=[
+                ('input', 'cast'),
+                ('cast', 'result')
+            ]
+        )
+
+        NamesUniquenessCheck().find_and_replace_pattern(graph)
+        names = [node.name for node in graph.get_op_nodes()]
+        result_name = Node(graph, 'result').name
+
+        self.assertTrue(len(set(names)) == 3)
+        self.assertTrue(result_name == 'node')
+
+    def test_2(self):
+        graph = build_graph(
+            nodes_attrs={
+                'input': {'kind': 'op', 'op': 'Parameter', 'name': 'node'},
+                'cast': {'kind': 'op', 'op': 'Cast', 'name': 'node_0'},
+                'result': {'kind': 'op', 'op': 'Result', 'name': 'node'}
+            },
+            edges=[
+                ('input', 'cast'),
+                ('cast', 'result')
+            ]
+        )
+
+        NamesUniquenessCheck().find_and_replace_pattern(graph)
+        names = [node.name for node in graph.get_op_nodes()]
+        result_name = Node(graph, 'result').name
+
+        self.assertTrue(len(set(names)) == 3)
+        self.assertTrue(result_name == 'node')
+
+    def test_3(self):
+        graph = build_graph(
+            nodes_attrs={
+                'input': {'kind': 'op', 'op': 'Parameter', 'name': 'node_0'},
+                'cast': {'kind': 'op', 'op': 'Cast', 'name': 'node_1'},
+                'result_1': {'kind': 'op', 'op': 'Result', 'name': 'node'},
+                'result_2': {'kind': 'op', 'op': 'Result', 'name': 'node'}
+            },
+            edges=[
+                ('input', 'cast'),
+                ('cast', 'result_1'),
+                ('cast', 'result_2'),
+            ]
+        )
+        NamesUniquenessCheck().find_and_replace_pattern(graph)
+        names = [node.name for node in graph.get_op_nodes()]
+
+        self.assertTrue('node' in names)
+        self.assertTrue(len(set(names)) == 4)

From c80e939474e965437d68c082e2d13e8cb3f36dca Mon Sep 17 00:00:00 2001
From: Alexander Zhogov <alexander.zhogov@intel.com>
Date: Mon, 7 Jun 2021 17:53:06 +0300
Subject: [PATCH 006/178] GitHub CI: non PR author commits are allowed (#6064)

---
 .github/org_control/check_pr.py | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/.github/org_control/check_pr.py b/.github/org_control/check_pr.py
index 882854724ca646..e0b48832ead75c 100644
--- a/.github/org_control/check_pr.py
+++ b/.github/org_control/check_pr.py
@@ -139,14 +139,15 @@ def update_labels(gh_api, pull, non_org_intel_pr_users, non_org_pr_users):
 
 def get_wrong_commits(pull):
     """Returns commits with incorrect user and email"""
-    print("GitHub PR user email:", pull.user.email)
+    pr_author_email = pull.user.email.lower()
+    print("GitHub PR author email:", pr_author_email)
     print("Check commits:")
     wrong_commits = set()
     for commit in pull.get_commits():
         # import pprint; pprint.pprint(commit.raw_data)
         print("Commit SHA:", commit.sha)
         # Use raw data because commit author can be non GitHub user
-        commit_email = commit.raw_data["commit"]["author"]["email"]
+        commit_email = commit.raw_data["commit"]["author"]["email"].lower()
         print("    Commit email:", commit_email)
         if not github_api.is_valid_user(commit.author):
             print(
@@ -159,9 +160,8 @@ def get_wrong_commits(pull):
                 "    WARNING: The commit is not verified. Reason:",
                 commit.raw_data["commit"]["verification"]["reason"],
             )
-            if pull.user.email != commit_email:
-                print("    ERROR: Commit email and GitHub user public email are differnt")
-                wrong_commits.add(commit.sha)
+            if pr_author_email != commit_email:
+                print("    WARNING: Commit email and GitHub PR author public email are differnt")
     return wrong_commits
 
 

From 56ada41bd05f47ceeddd86f1e99f61858e26a2ba Mon Sep 17 00:00:00 2001
From: Mikhail Treskin <mikhail.treskin@intel.com>
Date: Mon, 7 Jun 2021 18:19:52 +0300
Subject: [PATCH 007/178] Add get_constant_from_source capability to pyngraph
 (#6018)

---
 ngraph/python/src/pyngraph/util.cpp           | 19 +++++++++++++
 ngraph/python/tests/test_ngraph/test_utils.py | 28 +++++++++++++++++++
 2 files changed, 47 insertions(+)
 create mode 100644 ngraph/python/tests/test_ngraph/test_utils.py

diff --git a/ngraph/python/src/pyngraph/util.cpp b/ngraph/python/src/pyngraph/util.cpp
index a2eac625babb49..5178e84fe9081b 100644
--- a/ngraph/python/src/pyngraph/util.cpp
+++ b/ngraph/python/src/pyngraph/util.cpp
@@ -5,6 +5,7 @@
 #include <pybind11/numpy.h>
 
 #include "pyngraph/util.hpp"
+#include "ngraph/validation_util.hpp"
 
 namespace py = pybind11;
 
@@ -18,4 +19,22 @@ void regmodule_pyngraph_util(py::module m)
 {
     py::module mod = m.def_submodule("util", "ngraph.impl.util");
     mod.def("numpy_to_c", &numpy_to_c);
+    mod.def("get_constant_from_source",
+            &ngraph::get_constant_from_source,
+            py::arg("output"),
+            R"(
+                    Runs an estimation of source tensor.
+
+                    Parameters
+                    ----------
+                    output : Output
+                        output node
+
+                    Returns
+                    ----------
+                    get_constant_from_source : Constant or None
+                        If it succeeded to calculate both bounds and
+                        they are the same returns Constant operation
+                        from the resulting bound, otherwise Null.
+                )");
 }
diff --git a/ngraph/python/tests/test_ngraph/test_utils.py b/ngraph/python/tests/test_ngraph/test_utils.py
new file mode 100644
index 00000000000000..49b90017305356
--- /dev/null
+++ b/ngraph/python/tests/test_ngraph/test_utils.py
@@ -0,0 +1,28 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import ngraph as ng
+from ngraph.impl import Shape
+
+
+def test_get_constant_from_source_success():
+    dtype = np.int
+    input1 = ng.parameter(Shape([5, 5]), dtype=dtype, name="input_1")
+    input2 = ng.parameter(Shape([25]), dtype=dtype, name="input_2")
+    shape_of = ng.shape_of(input2, name="shape_of")
+    reshape = ng.reshape(input1, shape_of, special_zero=True)
+    folded_const = ng.impl.util.get_constant_from_source(reshape.input(1).get_source_output())
+
+    assert folded_const is not None
+    assert folded_const.get_vector() == [25]
+
+
+def test_get_constant_from_source_failed():
+    dtype = np.int
+    input1 = ng.parameter(Shape([5, 5]), dtype=dtype, name="input_1")
+    input2 = ng.parameter(Shape([1]), dtype=dtype, name="input_2")
+    reshape = ng.reshape(input1, input2, special_zero=True)
+    folded_const = ng.impl.util.get_constant_from_source(reshape.input(1).get_source_output())
+
+    assert folded_const is None

From 64d7a40ae4d783eda53a34d8201b928ebf4e3479 Mon Sep 17 00:00:00 2001
From: Andrei Gorbachev <andrei.gorbachev@intel.com>
Date: Mon, 7 Jun 2021 18:36:38 +0300
Subject: [PATCH 008/178] [IE CLDNN] 54304 fix reduce ops (#5986)

---
 .../src/cldnn_engine/ops/reduce.cpp           | 25 ++++++-
 .../subgraph_tests/reduce_eltwise.cpp         | 47 +++++++++++++
 .../include/subgraph_tests/reduce_eltwise.hpp | 15 +++++
 .../subgraph/reduce_eltwise.hpp               | 36 ++++++++++
 .../src/subgraph/reduce_eltwise.cpp           | 67 +++++++++++++++++++
 5 files changed, 189 insertions(+), 1 deletion(-)
 create mode 100644 inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reduce_eltwise.cpp
 create mode 100644 inference-engine/tests/functional/plugin/shared/include/subgraph_tests/reduce_eltwise.hpp
 create mode 100644 inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/reduce_eltwise.hpp
 create mode 100644 inference-engine/tests/functional/shared_test_classes/src/subgraph/reduce_eltwise.cpp

diff --git a/inference-engine/src/cldnn_engine/ops/reduce.cpp b/inference-engine/src/cldnn_engine/ops/reduce.cpp
index 3331d581b86c3e..26343ffb813129 100644
--- a/inference-engine/src/cldnn_engine/ops/reduce.cpp
+++ b/inference-engine/src/cldnn_engine/ops/reduce.cpp
@@ -18,6 +18,7 @@
 
 #include "api/reduce.hpp"
 #include "api/reorder.hpp"
+#include "api/reshape.hpp"
 
 namespace CLDNNPlugin {
 
@@ -78,6 +79,28 @@ void CreateReduceOp(Program& p, const std::shared_ptr<ngraph::Node>& op, cldnn::
 
     p.AddPrimitive(reducePrim);
 
+    auto resultLayerName = layerName;
+    auto out_dims = op->get_output_shape(0).size();
+    if (out_dims == 3 && !keep_dims && rank >= 4) {
+        resultLayerName = layerName + "_reshape";
+        auto out_shape = op->get_output_shape(0);
+        cldnn::tensor outTensor;
+        switch (rank) {
+            case 6:
+                outTensor = cldnn::tensor(TensorValue(out_shape[0]), TensorValue(out_shape[1]),
+                                          1, TensorValue(out_shape[2]), 1, 1);
+            case 5:
+                outTensor = cldnn::tensor(TensorValue(out_shape[0]), TensorValue(out_shape[1]),
+                                          1, TensorValue(out_shape[2]), 1);
+            case 4:
+                outTensor = cldnn::tensor(TensorValue(out_shape[0]), TensorValue(out_shape[1]),
+                                          1, TensorValue(out_shape[2]));
+        }
+        auto reshape_prim = cldnn::reshape(resultLayerName, layerName, outTensor);
+        p.AddPrimitive(reshape_prim);
+        p.AddPrimitiveToProfiler(op, resultLayerName);
+    }
+
     auto reorderLayerName = layerName + "_reorder";
     cldnn::format out_format = cldnn::format::any;
     auto out_dt = DataTypeFromPrecision(op->get_output_element_type(0));
@@ -89,7 +112,7 @@ void CreateReduceOp(Program& p, const std::shared_ptr<ngraph::Node>& op, cldnn::
         else if (rank - rawAxes.size() <= 4)
             out_format = cldnn::format::bfyx;
 
-        auto reorder_prim = cldnn::reorder(reorderLayerName, layerName, out_format, out_dt);
+        auto reorder_prim = cldnn::reorder(reorderLayerName, resultLayerName, out_format, out_dt);
         p.AddPrimitive(reorder_prim);
         p.AddPrimitiveToProfiler(op, reorderLayerName);
     } else {
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reduce_eltwise.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reduce_eltwise.cpp
new file mode 100644
index 00000000000000..04dbc6e800551e
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reduce_eltwise.cpp
@@ -0,0 +1,47 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <vector>
+
+#include "subgraph_tests/reduce_eltwise.hpp"
+
+using namespace SubgraphTestsDefinitions;
+
+namespace {
+
+const std::vector<InferenceEngine::Precision> netPrecisions = {
+        InferenceEngine::Precision::FP32,
+};
+
+INSTANTIATE_TEST_CASE_P(smoke_ReduceEltwise6D, ReduceEltwiseTest,
+                        testing::Combine(
+                                testing::Values(std::vector<size_t>{2, 3, 4, 5, 6, 7}),
+                                testing::Values(std::vector<int>{2, 3, 4}),
+                                testing::Values(CommonTestUtils::OpType::VECTOR),
+                                testing::Values(false),
+                                testing::ValuesIn(netPrecisions),
+                                testing::Values(CommonTestUtils::DEVICE_GPU)),
+                        ReduceEltwiseTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ReduceEltwise5D, ReduceEltwiseTest,
+                        testing::Combine(
+                                testing::Values(std::vector<size_t>{2, 3, 4, 5, 6}),
+                                testing::Values(std::vector<int>{2, 3}),
+                                testing::Values(CommonTestUtils::OpType::VECTOR),
+                                testing::Values(false),
+                                testing::ValuesIn(netPrecisions),
+                                testing::Values(CommonTestUtils::DEVICE_GPU)),
+                        ReduceEltwiseTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ReduceEltwise4D, ReduceEltwiseTest,
+                        testing::Combine(
+                                testing::Values(std::vector<size_t>{2, 3, 4, 5}),
+                                testing::Values(std::vector<int>{2}),
+                                testing::Values(CommonTestUtils::OpType::VECTOR),
+                                testing::Values(false),
+                                testing::ValuesIn(netPrecisions),
+                                testing::Values(CommonTestUtils::DEVICE_GPU)),
+                        ReduceEltwiseTest::getTestCaseName);
+
+}  // namespace
diff --git a/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/reduce_eltwise.hpp b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/reduce_eltwise.hpp
new file mode 100644
index 00000000000000..0b65a847820e5a
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/reduce_eltwise.hpp
@@ -0,0 +1,15 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "shared_test_classes/subgraph/reduce_eltwise.hpp"
+
+namespace SubgraphTestsDefinitions {
+
+TEST_P(ReduceEltwiseTest, CompareWithRefs) {
+    Run();
+};
+
+}  // namespace SubgraphTestsDefinitions
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/reduce_eltwise.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/reduce_eltwise.hpp
new file mode 100644
index 00000000000000..df04891bf906eb
--- /dev/null
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/subgraph/reduce_eltwise.hpp
@@ -0,0 +1,36 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <tuple>
+#include <string>
+#include <vector>
+#include <memory>
+#include "shared_test_classes/base/layer_test_utils.hpp"
+#include "ngraph_functions/builders.hpp"
+#include "ngraph_functions/utils/ngraph_helpers.hpp"
+#include "common_test_utils/test_constants.hpp"
+
+namespace SubgraphTestsDefinitions {
+
+using ReduceEltwiseParamsTuple = typename std::tuple<
+        std::vector<size_t>,              // Input shapes
+        std::vector<int>,                 // Axis to reduce order
+        CommonTestUtils::OpType,          // Scalar or vector type axis
+        bool,                             // Keep dims
+        InferenceEngine::Precision,       // Network precision
+        std::string>;                     // Device name
+
+class ReduceEltwiseTest:
+        public testing::WithParamInterface<ReduceEltwiseParamsTuple>,
+        public LayerTestsUtils::LayerTestsCommon{
+public:
+    std::shared_ptr<ngraph::Function> fn;
+    static std::string getTestCaseName(const testing::TestParamInfo<ReduceEltwiseParamsTuple> &obj);
+protected:
+    void SetUp() override;
+};
+
+}  // namespace SubgraphTestsDefinitions
diff --git a/inference-engine/tests/functional/shared_test_classes/src/subgraph/reduce_eltwise.cpp b/inference-engine/tests/functional/shared_test_classes/src/subgraph/reduce_eltwise.cpp
new file mode 100644
index 00000000000000..b51ca490090767
--- /dev/null
+++ b/inference-engine/tests/functional/shared_test_classes/src/subgraph/reduce_eltwise.cpp
@@ -0,0 +1,67 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "ngraph_functions/builders.hpp"
+#include "shared_test_classes/subgraph/reduce_eltwise.hpp"
+
+namespace SubgraphTestsDefinitions {
+std::string ReduceEltwiseTest::getTestCaseName(const testing::TestParamInfo<ReduceEltwiseParamsTuple> &obj) {
+    std::vector<size_t> inputShapes;
+    std::vector<int> axes;
+    CommonTestUtils::OpType opType;
+    bool keepDims;
+    InferenceEngine::Precision netPrecision;
+    std::string targetName;
+    std::tie(inputShapes, axes, opType, keepDims, netPrecision, targetName) = obj.param;
+
+    std::ostringstream result;
+    result << "IS=" << CommonTestUtils::vec2str(inputShapes) << "_";
+    result << "axes=" << CommonTestUtils::vec2str(axes) << "_";
+    result << "opType=" << opType << "_";
+    if (keepDims) result << "KeepDims_";
+    result << "netPRC=" << netPrecision.name() << "_";
+    result << "targetDevice=" << targetName;
+    return result.str();
+}
+
+void ReduceEltwiseTest::SetUp() {
+    std::vector<size_t> inputShape;
+    std::vector<int> axes;
+    CommonTestUtils::OpType opType;
+    bool keepDims;
+    InferenceEngine::Precision netPrecision;
+    std::string targetName;
+    std::tie(inputShape, axes, opType, keepDims, netPrecision, targetName) = this->GetParam();
+    auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
+    auto params = ngraph::builder::makeParams(ngPrc, {inputShape});
+    auto paramOuts = ngraph::helpers::convert2OutputVector(
+            ngraph::helpers::castOps2Nodes<ngraph::op::Parameter>(params));
+
+    std::vector<size_t> shapeAxes;
+    switch (opType) {
+        case CommonTestUtils::OpType::SCALAR: {
+            if (axes.size() > 1)
+                FAIL() << "In reduce op if op type is scalar, 'axis' input's must contain 1 element";
+            break;
+        }
+        case CommonTestUtils::OpType::VECTOR: {
+            shapeAxes.push_back(axes.size());
+            break;
+        }
+        default:
+            FAIL() << "Reduce op doesn't support operation type: " << opType;
+    }
+    auto reductionAxesNode = std::dynamic_pointer_cast<ngraph::Node>(
+                             std::make_shared<ngraph::opset3::Constant>(ngraph::element::Type_t::i64, ngraph::Shape(shapeAxes), axes));
+
+    auto reduce = std::make_shared<ngraph::opset3::ReduceSum>(paramOuts[0], reductionAxesNode, keepDims);
+
+    std::vector<size_t> constShape(reduce.get()->get_output_size(), 1);
+    constShape[2] = inputShape.back();
+    auto constant = ngraph::builder::makeConstant<float>(ngPrc, constShape, {}, true);
+    auto eltw = ngraph::builder::makeEltwise(reduce, constant, ngraph::helpers::EltwiseTypes::MULTIPLY);
+    ngraph::ResultVector results{std::make_shared<ngraph::opset3::Result>(eltw)};
+    function = std::make_shared<ngraph::Function>(results, params, "ReduceEltwise");
+}
+} // namespace SubgraphTestsDefinitions

From 6a42b47c2f4bba88456fe39b8b203e6a33211293 Mon Sep 17 00:00:00 2001
From: Jozef Daniecki <jozef.daniecki@intel.com>
Date: Mon, 7 Jun 2021 19:12:57 +0200
Subject: [PATCH 009/178] RegionYolo operation specification refactoring.
 (#5926)

* RegionYolo spec refactored against explicit type indication.

* Improve readability.
---
 docs/ops/detection/RegionYolo_1.md | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/docs/ops/detection/RegionYolo_1.md b/docs/ops/detection/RegionYolo_1.md
index c4eece6ff8b419..ebcc79c23777cb 100644
--- a/docs/ops/detection/RegionYolo_1.md
+++ b/docs/ops/detection/RegionYolo_1.md
@@ -6,7 +6,7 @@
 
 **Short description**: *RegionYolo* computes the coordinates of regions with probability for each class.
 
-**Detailed description**: This operation is directly mapped to the original YOLO layer. [Reference](https://arxiv.org/pdf/1612.08242.pdf)
+**Detailed description**: This operation is directly mapped to the [YOLO9000: Better, Faster, Stronger](https://arxiv.org/pdf/1612.08242.pdf) paper. 
 
 **Attributes**:
 
@@ -78,14 +78,17 @@
 
 **Inputs**:
 
-*   **1**: `data` - 4D input tensor with floating point elements and shape `[N, C, H, W]`. Required.
+*   **1**: `data` - 4D tensor of type `T` and shape `[N, C, H, W]`. **Required.**
 
 **Outputs**:
 
-*   **1**: output tensor of rank 4 or less that codes detected regions. Refer to the original YOLO paper to decode the output as boxes. `anchors` should be used to decode real box coordinates. If `do_softmax` is set to 0, then the output shape is `[N, (classes + coords + 1)*len(mask), H, W]`. If `do_softmax` is set to 1, then output shape is partially flattened and defined in the following way:
+*   **1**: tensor of type `T` and rank 4 or less that codes detected regions. Refer to the [YOLO9000: Better, Faster, Stronger](https://arxiv.org/pdf/1612.08242.pdf) paper to decode the output as boxes. `anchors` should be used to decode real box coordinates. If `do_softmax` is set to `0`, then the output shape is `[N, (classes + coords + 1) * len(mask), H, W]`. If `do_softmax` is set to `1`, then output shape is partially flattened and defined in the following way:
 
-    flat_dim = data.shape[axis] * data.shape[axis+1] * ... * data.shape[end_axis]
-    output.shape = [data.shape[0], ..., data.shape[axis-1], flat_dim, data.shape[end_axis + 1], ...]
+    `flat_dim = data.shape[axis] * data.shape[axis+1] * ... * data.shape[end_axis]`  
+    `output.shape = [data.shape[0], ..., data.shape[axis-1], flat_dim, data.shape[end_axis + 1], ...]`
+
+**Types**
+* *T*: any supported floating point type.
 
 **Example**
 

From 4d9fe14ec62641e7ab480d08cd70be0151aecd62 Mon Sep 17 00:00:00 2001
From: Yegor Kruglov <yegor.kruglov@intel.com>
Date: Mon, 7 Jun 2021 20:22:26 +0300
Subject: [PATCH 010/178] [MO] ConvolutionWithGroupResolver update to enable TF
 DepthwiseConv2dNative with in_channels=1  (#5528)

* changed permutation attribute in conv extractor

* changed conv get_group parameter

* implemented a transformation

* updated BOM

* specified transformation for in_channels 1

* added unittest and comment string

* updated convolution normalizer to convert depthwise convolution with group=1 to group convolution

* renamed function

* updated IR reader

* conversations resolving

* condition change
---
 .../extensions/back/ConvolutionNormalizer.py  | 106 ++++++++----------
 .../mo/utils/ir_reader/layer_to_class.py      |   7 ++
 .../back/ConvolutionNormalizer_test.py        |  72 +++++++++++-
 3 files changed, 123 insertions(+), 62 deletions(-)

diff --git a/model-optimizer/extensions/back/ConvolutionNormalizer.py b/model-optimizer/extensions/back/ConvolutionNormalizer.py
index 04da65631c53fa..0abfbec0e208e0 100644
--- a/model-optimizer/extensions/back/ConvolutionNormalizer.py
+++ b/model-optimizer/extensions/back/ConvolutionNormalizer.py
@@ -8,10 +8,43 @@
 from mo.back.replacement import BackReplacementPattern
 from mo.front.common.partial_infer.utils import int64_array
 from mo.front.tf.graph_utils import create_op_node_with_second_input, create_op_with_const_inputs
-from mo.graph.graph import Graph
+from mo.graph.graph import Graph, Node
 from mo.ops.const import Const
 from mo.ops.reshape import Reshape
 from mo.ops.strided_slice import StridedSlice
+from mo.utils.error import Error
+
+
+def resolve_convolution_with_group(node: Node, group: int, ir_version: str):
+    input_shape = node.in_port(0).data.get_shape()
+    assert len(input_shape) in [3, 4, 5]
+
+    weights_shape = node.in_port(1).data.get_shape()
+    assert weights_shape is not None
+    assert len(weights_shape) in [3, 4, 5]
+    assert weights_shape[0] % group == 0
+
+    assert int64_array(node.output).ndim == 0
+    if ir_version == 'V7':
+        if weights_shape[0] == node.output:
+            # weights are already is in [G*O I X Y] format
+            return
+        new_shape = int64_array([node.output, -1, *weights_shape[2:]])
+
+    elif ir_version == 'V10':
+        I = input_shape[1]
+        new_shape = int64_array([group, node.output / group, I / group, *weights_shape[2:]])
+        assert np.prod(weights_shape) == np.prod(new_shape), \
+            'Initial weights shape {}, grouped weights shape {}'.format(weights_shape, new_shape)
+        del node['group']
+        node['type'] = 'GroupConvolution'
+    else:
+        raise Error("Unknown IR version: {}".format(ir_version))
+
+    reshape = create_op_node_with_second_input(node.graph, Reshape, int64_array(new_shape),
+                                               {'override_output_shape': True})
+
+    node.in_port(1).get_connection().insert_node(reshape)
 
 
 class ConvolutionNormalizer(BackReplacementPattern):
@@ -37,33 +70,12 @@ class V7ConvolutionWithGroupsResolver(BackReplacementPattern):
     """
     enabled = False
 
-    @staticmethod
-    def pattern():
-        return dict(
-            nodes=[
-                ('node', dict(type='Convolution', group=lambda g: g is not None and g != 1))
-            ],
-            edges=[]
-        )
-
-    def replace_pattern(self, graph: Graph, match: dict):
-        node = match['node']
-
-        group = node.group
-        assert group > 1
-
-        weights_shape = node.in_port(1).data.get_shape()
-        assert weights_shape is not None
-        assert weights_shape[0] % group == 0
-
-        if weights_shape[0] == node.output:
-            # weights are already is in [G*O I X Y] format
-            return
-
-        new_shape = int64_array([node.output, -1, *weights_shape[2:]])
-        reshape = create_op_node_with_second_input(graph, Reshape, int64_array(new_shape),
-                                                   {'override_output_shape': True})
-        node.in_port(1).get_connection().insert_node(reshape)
+    def find_and_replace_pattern(self, graph: Graph):
+        for node in graph.get_op_nodes(type='Convolution'):
+            group = node.soft_get('group', None)
+            if group is not None:
+                if group != 1 or node.soft_get('op') == 'DepthwiseConv2dNative':
+                    resolve_convolution_with_group(node, group, ir_version='V7')
 
 
 class V10ConvolutionWithGroupsResolver(BackReplacementPattern):
@@ -73,38 +85,12 @@ class V10ConvolutionWithGroupsResolver(BackReplacementPattern):
     """
     enabled = False
 
-    @staticmethod
-    def pattern():
-        return dict(
-            nodes=[
-                ('node', dict(type='Convolution', group=lambda g: g is not None and g != 1))
-            ],
-            edges=[]
-        )
-
-    def replace_pattern(self, graph: Graph, match: dict):
-        node = match['node']
-
-        group = node.group
-        assert group > 1
-
-        weights_shape = node.in_port(1).data.get_shape()
-        assert weights_shape is not None
-        assert weights_shape[0] % group == 0
-        I = node.in_port(0).data.get_shape()[1]
-
-        new_shape = int64_array([group, node.output / group, I / group, *weights_shape[2:]])
-
-        assert np.prod(weights_shape) == np.prod(new_shape), \
-            'Initial weights shape {}, grouped weights shape {}'.format(weights_shape, new_shape)
-
-        del node['group']
-        node['type'] = 'GroupConvolution'
-
-        reshape = create_op_node_with_second_input(graph, Reshape, int64_array(new_shape),
-                                                   {'override_output_shape': True})
-
-        node.in_port(1).get_connection().insert_node(reshape)
+    def find_and_replace_pattern(self, graph: Graph):
+        for node in graph.get_op_nodes(type='Convolution'):
+            group = node.soft_get('group', None)
+            if group is not None:
+                if group != 1 or node.soft_get('op') == 'DepthwiseConv2dNative':
+                    resolve_convolution_with_group(node, group, ir_version='V10')
 
 
 class ConvolutionWithGroupsResolver(BackReplacementPattern):
diff --git a/model-optimizer/mo/utils/ir_reader/layer_to_class.py b/model-optimizer/mo/utils/ir_reader/layer_to_class.py
index d1afed6918d7bf..f9f5fd9fb1db23 100644
--- a/model-optimizer/mo/utils/ir_reader/layer_to_class.py
+++ b/model-optimizer/mo/utils/ir_reader/layer_to_class.py
@@ -195,6 +195,13 @@ def groupconv_to_conv(op: Node):
             'Weight shape and calculated shape mismatch in GroupConv node {}.'.format(op.name)
     # we need to set this attrs for correct shape infer as convolution
     op['group'] = group
+    # The only way GroupConvolution with 'group' = 1 appears in IR is by converting from TF DepthwiseConv2dNative.
+    # In this case we need to specify 'op' parameter for the
+    # extensions.back.ConvolutionNormalizer.ConvolutionWithGroupsResolver to work properly.
+    # Otherwise  there will be 'Convolution' instead 'GroupConvolution' in restored IR, since 'GroupConvolution' is
+    # extended as node with 'type' = 'Convolution' by IR reader
+    if group == 1:
+        op['op'] = 'DepthwiseConv2dNative'
     op.type = 'Convolution'
 
 
diff --git a/model-optimizer/unit_tests/extensions/back/ConvolutionNormalizer_test.py b/model-optimizer/unit_tests/extensions/back/ConvolutionNormalizer_test.py
index d7a2f4809fcca2..47a7fc9ac283e2 100644
--- a/model-optimizer/unit_tests/extensions/back/ConvolutionNormalizer_test.py
+++ b/model-optimizer/unit_tests/extensions/back/ConvolutionNormalizer_test.py
@@ -102,7 +102,7 @@ def test_reshape_reducing_tensor_rank(self):
 class TestV7ConvolutionWithGroupsResolver(unittest.TestCase):
     def test_v7_group_convolution_resolver(self):
         nodes = {
-            **regular_op_with_shaped_data('input', None, {'type': 'Parameter'}),
+            **regular_op_with_shaped_data('input', [1, 3, 224, 224], {'type': 'Parameter'}),
 
             **valued_const_with_data('weights', np.ones([3, 8, 7, 7])),
 
@@ -133,7 +133,7 @@ def test_v7_group_convolution_resolver(self):
 
     def test_v7_group_convolution_resolver_weight_are_in_the_right_layout(self):
         nodes = {
-            **regular_op_with_shaped_data('input', None, {'type': 'Parameter'}),
+            **regular_op_with_shaped_data('input', [1, 3, 224, 224], {'type': 'Parameter'}),
             **valued_const_with_data('weights', np.ones([24, 1, 7, 7])),
             **regular_op_with_shaped_data('convolution', None, {'type': 'Convolution', 'group': 3, 'output': 24}),
             **result(),
@@ -149,6 +149,38 @@ def test_v7_group_convolution_resolver_weight_are_in_the_right_layout(self):
         (flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
         self.assertTrue(flag, resp)
 
+    def test_v7_group_convolution_resolver_depthwise_conv2d(self):
+        nodes = {
+            **regular_op_with_shaped_data('input', [1, 1, 224, 224], {'type': 'Parameter'}),
+
+            **valued_const_with_data('weights', np.ones([1, 8, 7, 7])),
+
+            **valued_const_with_data('dim', int64_array([8, -1, 7, 7])),
+            **regular_op_with_empty_data('reshape', {'type': 'Reshape'}),
+
+            **regular_op_with_shaped_data('convolution', None, {'type': 'Convolution', 'group': 1, 'output': 8,
+                                                                'op': 'DepthwiseConv2dNative'}),
+
+            **result(),
+        }
+        graph = build_graph(nodes, [
+            *connect('input', '0:convolution'),
+            *connect('weights', '1:convolution'),
+            *connect('convolution', 'output'),
+        ], nodes_with_edges_only=True)
+
+        V7ConvolutionWithGroupsResolver().find_and_replace_pattern(graph)
+        graph_ref = build_graph(nodes, [
+            *connect('input', '0:convolution'),
+            *connect('weights', '0:reshape'),
+            *connect('dim', '1:reshape'),
+            *connect('reshape', '1:convolution'),
+            *connect('convolution', 'output'),
+        ], nodes_with_edges_only=True)
+
+        (flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
+        self.assertTrue(flag, resp)
+
 
 class TestV10ConvolutionWithGroupsResolver(unittest.TestCase):
     def test_v10_group_convolution_resolver(self):
@@ -185,3 +217,39 @@ def test_v10_group_convolution_resolver(self):
 
         (flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
         self.assertTrue(flag, resp)
+
+    def test_v10_group_convolution_resolver_depthwise_conv2d(self):
+        nodes = {
+            **regular_op_with_shaped_data('input', [1, 1, 224, 224], {'type': 'Parameter'}),
+
+            **valued_const_with_data('weights', np.ones([1, 8, 7, 7])),
+
+            **valued_const_with_data('dim', int64_array([1, 8, 1, 7, 7])),
+            **regular_op_with_empty_data('reshape', {'type': 'Reshape'}),
+
+            **regular_op_with_shaped_data('convolution', None, {'type': 'Convolution', 'group': 1, 'output': 8,
+                                                                'op': 'DepthwiseConv2dNative'}),
+
+            **result(),
+        }
+        graph = build_graph(nodes, [
+            *connect('input', '0:convolution'),
+            *connect('weights', '1:convolution'),
+            *connect('convolution', 'output'),
+        ], nodes_with_edges_only=True)
+
+        V10ConvolutionWithGroupsResolver().find_and_replace_pattern(graph)
+
+        nodes['convolution']['type'] = 'GroupConvolution'
+        del nodes['convolution']['group']
+
+        graph_ref = build_graph(nodes, [
+            *connect('input', '0:convolution'),
+            *connect('weights', '0:reshape'),
+            *connect('dim', '1:reshape'),
+            *connect('reshape', '1:convolution'),
+            *connect('convolution', 'output'),
+        ], nodes_with_edges_only=True)
+
+        (flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
+        self.assertTrue(flag, resp)

From bc7f61be241544c14d82d2b66c83174bf5a13281 Mon Sep 17 00:00:00 2001
From: Katarzyna Mitrus <katarzyna.mitrus@intel.com>
Date: Tue, 8 Jun 2021 05:51:41 +0200
Subject: [PATCH 011/178] PRelu reference implementation and ReshapePRelu
 transformation alignment (#5915)

* Apply  ReshapePRelu transformation only to const slope input

* Remove xfail from onnx backend prelu_broadcast test

* Fix and add Prelu SLT

* Update PRelu mkldnn transformation to extend broadcast support

* Fix and update PRelu reference implementation

* ONNX Prelu tests

* Add prelu backend tests

* Update ie tests manifest

* Comments clean up

* Fix STL Fill leakyslope blob

* Code refactor

* Unify layer tests slope input values generation
---
 .../ngraph_transformations/reshape_prelu.cpp  |  12 +-
 .../single_layer_tests/activation.cpp         |  28 +-
 .../skip_tests_config.cpp                     |   2 -
 .../single_layer/activation.hpp               |   4 +-
 .../src/single_layer/activation.cpp           |  92 +-
 .../ngraph/runtime/reference/prelu.hpp        |  17 +-
 ngraph/python/tests/__init__.py               |   2 -
 ngraph/python/tests/test_onnx/test_backend.py |   5 +-
 ngraph/test/CMakeLists.txt                    |   1 +
 ngraph/test/backend/fused_op.in.cpp           |  49 -
 ngraph/test/backend/prelu.in.cpp              | 942 ++++++++++++++++++
 ngraph/test/models/onnx/prelu_1d.prototxt     |  72 ++
 .../test/models/onnx/prelu_batch_nd.prototxt  |  81 ++
 ngraph/test/models/onnx/prelu_c_1_1.prototxt  |  78 ++
 ngraph/test/onnx/onnx_import.in.cpp           | 125 ++-
 ngraph/test/runtime/ie/unit_test.manifest     |   3 -
 16 files changed, 1362 insertions(+), 151 deletions(-)
 create mode 100644 ngraph/test/backend/prelu.in.cpp
 create mode 100644 ngraph/test/models/onnx/prelu_1d.prototxt
 create mode 100644 ngraph/test/models/onnx/prelu_batch_nd.prototxt
 create mode 100644 ngraph/test/models/onnx/prelu_c_1_1.prototxt

diff --git a/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_prelu.cpp b/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_prelu.cpp
index 0cc1a33cbc3283..69fd75ea57a661 100644
--- a/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_prelu.cpp
+++ b/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_prelu.cpp
@@ -20,8 +20,16 @@ MKLDNNPlugin::ReshapePRelu::ReshapePRelu() {
         if (!prelu || ngraph::shape_size(prelu->get_input_shape(1)) == 1 || prelu->get_input_shape(1).size() != 1) {
             return false;
         }
-        ngraph::Shape new_shape(prelu->input_value(0).get_shape().size(), 1);
-        new_shape[new_shape.size() > 1 ? 1 : 0] = prelu->input_value(1).get_shape()[0];
+        const auto prelu_shape = prelu->input_value(0).get_shape();
+        const auto slope_shape = prelu->input_value(1).get_shape();
+        ngraph::Shape new_shape(prelu_shape.size(), 1);
+        const auto slope_dim = slope_shape[0];
+        const auto channel_dim_idx = prelu_shape.size() > 1 ? 1 : 0;
+        if (slope_dim != prelu_shape[channel_dim_idx]) {
+            return false;
+        }
+        new_shape[channel_dim_idx] = slope_dim;
+
         auto slope = ngraph::op::util::reshapeTo(prelu->input_value(1), new_shape);
         auto new_prelu = std::make_shared<ngraph::opset1::PRelu>(prelu->input(0).get_source_output(), slope);
         new_prelu->set_friendly_name(prelu->get_friendly_name());
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
index 76607e55f76b91..6e762ed562c004 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
@@ -69,8 +69,8 @@ const std::map<ActivationTypes, std::vector<std::vector<float>>> intActivationTy
 };
 
 const std::map<ActivationTypes, std::vector<std::vector<float>>> activationParamTypes = {
-    {PReLu, {{-0.01f}}},
-    {LeakyRelu, {{0.01f}}}
+        {PReLu, {{}}}, // Slope will be filled with increasing values from -10 to match slope input shape
+        {LeakyRelu, {{0.01f}}}
 };
 
 std::map<std::vector<size_t>, std::vector<std::vector<size_t>>> basic = {
@@ -81,16 +81,12 @@ std::map<std::vector<size_t>, std::vector<std::vector<size_t>>> basic = {
 std::map<std::vector<size_t>, std::vector<std::vector<size_t>>> preluBasic = {
         {{1, 50}, {{1}, {50}}},
         {{1, 128}, {{1}, {128}}},
-        {{20, 128}, {{128}}},
-        {{1, 20, 128}, {{1}, {20}}},
-        {{1, 20, 128, 128}, {{1}, {20}}},
-        {{1, 20, 20, 128, 128}, {{1}, {20}}}
-        // according to spec second input for PRelu must be 1D and must be broadcastabe per channel
-        // at this moment these cases unsupported
-        // {{20, 128}, {{20}, {20, 128}}},
-        // {{1, 20, 128}, {{128}, {20, 128}}},
-        // {{1, 20, 128, 128}, {{128}, {128, 128}, {20, 128, 128}}},
-        // {{1, 20, 20, 128, 128}, {{128}, {128, 128}, {20, 128, 128}, {20, 20, 128, 128}}},
+
+        // Broadcast check
+        {{3, 2}, {{1}, {2}, {3, 2}}},
+        {{3, 2, 5}, {{1}, {2}, {5}, {2, 5}, {3, 1, 5}, {1, 2, 1}, {1, 1, 5}, {3, 1, 1}, {3, 2, 5}}},
+        {{2, 1, 2}, {{2}, {2, 1, 1}}},
+        {{3, 2, 5, 7}, {{1}, {7}, {2}, {5, 7}, {2, 5, 7}, {2, 1, 1}, {1, 2, 1, 1}, {3, 2, 1, 1}, {3, 2, 5, 7}}},
 };
 
 const auto basicCases = ::testing::Combine(
@@ -127,11 +123,9 @@ const auto basicIntegerOperations = ::testing::Combine(
 );
 
 INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_Integer_Activation_Basic, ActivationLayerTest, basicIntegerOperations, ActivationLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic_Prelu, ActivationLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
-
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationParamLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
-
 INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationDynamicLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_CASE_P(smoke_Integer_Activation_Basic, ActivationLayerTest, basicIntegerOperations, ActivationLayerTest::getTestCaseName);
 
+INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic_Prelu_Const, ActivationLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic_Prelu_Param, ActivationParamLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
index f59f8d954e0f29..b3134e9953e9e5 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
@@ -35,8 +35,6 @@ std::vector<std::string> disabledTestPatterns() {
         R"(.*ReluShapeOfSubgraphTest.*)",
         // TODO: Issue: 34805
         R"(.*ActivationLayerTest.*Ceiling.*)",
-        // TODO: Issue: 32032
-        R"(.*ActivationParamLayerTest.*)",
         // TODO: Issue: 43314
         R"(.*Broadcast.*mode=BIDIRECTIONAL.*inNPrec=BOOL.*)",
         // TODO: Issue 43417 sporadic issue, looks like an issue in test, reproducible only on Windows platform
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/activation.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/activation.hpp
index bdf996cd141065..684942ee184653 100644
--- a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/activation.hpp
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/activation.hpp
@@ -96,13 +96,11 @@ class ActivationLayerTest : public testing::WithParamInterface<activationParams>
 };
 
 class ActivationParamLayerTest : public ActivationLayerTest {
-public:
-    void Infer() override;
-
 protected:
     void SetUp() override;
 
 private:
+    InferenceEngine::Blob::Ptr GenerateInput(const InferenceEngine::InputInfo &info) const override;
     void generateActivationBlob(std::vector<float> constantsValue);
     ngraph::ParameterVector createActivationParams(
         ngraph::element::Type ngPrc, std::vector<size_t> inShape = {});
diff --git a/inference-engine/tests/functional/shared_test_classes/src/single_layer/activation.cpp b/inference-engine/tests/functional/shared_test_classes/src/single_layer/activation.cpp
index 3136c604e7d9ab..5b90cfc2079dd6 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/single_layer/activation.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/single_layer/activation.cpp
@@ -41,6 +41,13 @@ void ActivationLayerTest::SetUp() {
     auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
     auto params = ngraph::builder::makeParams(ngPrc, {shapes.first});
     params[0]->set_friendly_name("Input");
+
+    if (activationType == ngraph::helpers::ActivationTypes::PReLu && constantsValue.empty()) {
+        const auto elemnts_count = ngraph::shape_size(shapes.second);
+        constantsValue.resize(elemnts_count);
+        std::iota(constantsValue.begin(), constantsValue.end(), -10);
+    }
+
     auto activation = ngraph::builder::makeActivation(params[0], ngPrc, activationType, shapes.second, constantsValue);
 
     function = std::make_shared<ngraph::Function>(ngraph::NodeVector{activation}, params);
@@ -163,70 +170,26 @@ ngraph::ParameterVector ActivationParamLayerTest::createActivationParams(ngraph:
     }
 }
 
-void ActivationParamLayerTest::generateActivationBlob(std::vector<float> constantsValue) {
-    switch (activationType) {
-        case ngraph::helpers::ActivationTypes::PReLu: {
-            auto blobNegativeSlope = inferRequest.GetBlob("negativeSlope");
-            float negativeSlope = constantsValue[0];
-            blobNegativeSlope = FuncTestUtils::createAndFillBlobWithFloatArray(blobNegativeSlope->getTensorDesc(), &negativeSlope, 1);
-            inferRequest.SetBlob("negativeSlope", blobNegativeSlope);
-            inputs.push_back(blobNegativeSlope);
-            break;
-        }
-        case ngraph::helpers::ActivationTypes::LeakyRelu: {
-            auto blobLeakySlope = inferRequest.GetBlob("leakySlope");
-            float leakySlope = constantsValue[0];
-            blobLeakySlope = FuncTestUtils::createAndFillBlobWithFloatArray(blobLeakySlope->getTensorDesc(), &leakySlope, 1);
-            inferRequest.SetBlob("leakySlope", blobLeakySlope);
-            inputs.push_back(blobLeakySlope);
-            break;
-        }
-        case ngraph::helpers::ActivationTypes::HardSigmoid: {
-            auto blobHardSigmoidAlpha = inferRequest.GetBlob("alpha");
-            auto blobHardSigmoidBeta = inferRequest.GetBlob("beta");
-            float alpha = constantsValue[0], beta = constantsValue[1];
-            blobHardSigmoidAlpha = FuncTestUtils::createAndFillBlobWithFloatArray(blobHardSigmoidAlpha->getTensorDesc(), &alpha, 1);
-            blobHardSigmoidBeta = FuncTestUtils::createAndFillBlobWithFloatArray(blobHardSigmoidBeta->getTensorDesc(), &beta, 1);
-            inferRequest.SetBlob("alpha", blobHardSigmoidAlpha);
-            inferRequest.SetBlob("beta", blobHardSigmoidBeta);
-            inputs.push_back(blobHardSigmoidAlpha);
-            inputs.push_back(blobHardSigmoidBeta);
-            break;
-        }
-        case ngraph::helpers::ActivationTypes::Selu: {
-            auto blobHardSigmoidAlpha = inferRequest.GetBlob("alpha");
-            auto blobHardSigmoidLambda = inferRequest.GetBlob("lambda");
-            float alpha = constantsValue[0], lambda = constantsValue[1];
-            blobHardSigmoidAlpha = FuncTestUtils::createAndFillBlobWithFloatArray(blobHardSigmoidAlpha->getTensorDesc(), &alpha, 1);
-            blobHardSigmoidLambda = FuncTestUtils::createAndFillBlobWithFloatArray(blobHardSigmoidLambda->getTensorDesc(), &lambda, 1);
-            inferRequest.SetBlob("alpha", blobHardSigmoidAlpha);
-            inferRequest.SetBlob("lambda", blobHardSigmoidLambda);
-            inputs.push_back(blobHardSigmoidAlpha);
-            inputs.push_back(blobHardSigmoidLambda);
-            break;
-        }
-        default:
-            IE_THROW() << "Unsupported activation type for Params test type";
-    }
-}
-
-void ActivationParamLayerTest::Infer() {
-    inferRequest = executableNetwork.CreateInferRequest();
-
-    auto blobInput = inferRequest.GetBlob("Input");
-    blobInput = FuncTestUtils::createAndFillBlobFloat(blobInput->getTensorDesc());
-    inferRequest.SetBlob("Input", blobInput);
-    inputs.push_back(blobInput);
-
-    generateActivationBlob(constantsValue);
-
-    if (configuration.count(InferenceEngine::PluginConfigParams::KEY_DYN_BATCH_ENABLED) &&
-        configuration.count(InferenceEngine::PluginConfigParams::YES)) {
-        auto batchSize = executableNetwork.GetInputsInfo().begin()->second->getTensorDesc().getDims()[0] / 2;
-        inferRequest.SetBatch(batchSize);
+InferenceEngine::Blob::Ptr ActivationParamLayerTest::GenerateInput(const InferenceEngine::InputInfo &info) const {
+    InferenceEngine::Blob::Ptr blobPtr;
+    const std::string& name = info.name();
+    if (name == "negativeSlope") {
+        const auto elemnts_count = ngraph::shape_size(function->get_parameters()[1]->get_shape());
+        std::vector<float> param_data(elemnts_count);
+        std::iota(param_data.begin(), param_data.end(), -10);
+        blobPtr = FuncTestUtils::createAndFillBlobWithFloatArray(info.getTensorDesc(), &param_data[0], elemnts_count);
+    } else if (name == "leakySlope") {
+        const auto elemnts_count = ngraph::shape_size(function->get_parameters()[1]->get_shape());
+        std::vector<float> param_data(elemnts_count, constantsValue[0]);
+        blobPtr = FuncTestUtils::createAndFillBlobWithFloatArray(info.getTensorDesc(), &param_data[0], elemnts_count);
+    } else if (name == "alpha") {
+         blobPtr = FuncTestUtils::createAndFillBlobWithFloatArray(info.getTensorDesc(), &constantsValue[0], 1);
+    } else if (name == "beta" || name == "lambda") {
+        blobPtr = FuncTestUtils::createAndFillBlobWithFloatArray(info.getTensorDesc(), &constantsValue[1], 1);
+    } else {
+        blobPtr = FuncTestUtils::createAndFillBlob(info.getTensorDesc(), 20, -10, 1);
     }
-
-    inferRequest.Infer();
+    return blobPtr;
 }
 
 void ActivationParamLayerTest::SetUp() {
@@ -245,7 +208,8 @@ void ActivationParamLayerTest::SetUp() {
     params.insert(params.end(), activationParams.begin(), activationParams.end());
 
     auto activation = ngraph::builder::makeActivation(params, ngPrc, activationType);
-    function = std::make_shared<ngraph::Function>(ngraph::NodeVector{activation}, params);
+    ngraph::ResultVector results{std::make_shared<ngraph::opset1::Result>(activation)};
+    function = std::make_shared<ngraph::Function>(results, params);
 }
 
 void ActivationDynamicLayerTest::Run() {
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/prelu.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/prelu.hpp
index 34a420399ac620..d546bbd79af87e 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/prelu.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/prelu.hpp
@@ -24,12 +24,21 @@ namespace ngraph
                        const Shape& arg_shape,
                        const Shape& slope_shape)
             {
-                int cnt = 0;
-                for (size_t i = 0; i < shape_size(arg_shape); ++i)
+                Shape slope_shape_tmp = slope_shape;
+                const auto channel_dim_idx = arg_shape.size() > 1 ? 1 : 0;
+                if (slope_shape.size() == 1 && arg_shape[channel_dim_idx] == slope_shape[0])
                 {
-                    out[i] =
-                        arg[i] < T(0) ? T(arg[i] * slope[cnt++ % shape_size(slope_shape)]) : arg[i];
+                    Shape channel_slope_shape(arg_shape.size(), 1);
+                    channel_slope_shape[channel_dim_idx] = slope_shape[0];
+                    std::swap(slope_shape_tmp, channel_slope_shape);
                 }
+                autobroadcast_binop(arg,
+                                    slope,
+                                    out,
+                                    arg_shape,
+                                    slope_shape_tmp,
+                                    ngraph::op::AutoBroadcastType::NUMPY,
+                                    [](T x, T y) -> T { return x < T(0) ? T(x * y) : x; });
             }
         } // namespace reference
     }     // namespace runtime
diff --git a/ngraph/python/tests/__init__.py b/ngraph/python/tests/__init__.py
index 930be465a632d3..c179f312f1e7e8 100644
--- a/ngraph/python/tests/__init__.py
+++ b/ngraph/python/tests/__init__.py
@@ -110,8 +110,6 @@ def xfail_test(reason="Mark the test as expected to fail", strict=True):
 xfail_issue_47323 = xfail_test(reason="RuntimeError: The plugin does not support FP64")
 xfail_issue_47337 = xfail_test(reason="RuntimeError: Unsupported dynamic ops: v1::OneHot")
 xfail_issue_33593 = xfail_test(reason="Current implementation of MaxPool doesn't support indices output")
-xfail_issue_51993 = xfail_test(reason="PRelu supports only 1D tensor for 'slope' input broadcasted"
-                                      "by channel")
 xfail_issue_55760 = xfail_test(reason="RuntimeError: Reversed axis have axes above the source space shape")
 
 # Model MSFT issues:
diff --git a/ngraph/python/tests/test_onnx/test_backend.py b/ngraph/python/tests/test_onnx/test_backend.py
index d411919e97785e..a8ee1cbb065c6e 100644
--- a/ngraph/python/tests/test_onnx/test_backend.py
+++ b/ngraph/python/tests/test_onnx/test_backend.py
@@ -59,7 +59,6 @@
                    xfail_issue_49753,
                    xfail_issue_49754,
                    xfail_issue_52463,
-                   xfail_issue_51993,
                    xfail_issue_55760)
 
 
@@ -369,9 +368,7 @@ def expect_fail(test_case_path, xfail):  # type: (str) -> None
      "OnnxBackendNodeModelTest.test_quantizelinear_cpu"),
     (xfail_issue_33593,
      "OnnxBackendNodeModelTest.test_maxpool_with_argmax_2d_precomputed_strides_cpu",
-     "OnnxBackendNodeModelTest.test_maxpool_with_argmax_2d_precomputed_pads_cpu",),
-    (xfail_issue_51993,
-     "OnnxBackendNodeModelTest.test_prelu_broadcast_cpu",)
+     "OnnxBackendNodeModelTest.test_maxpool_with_argmax_2d_precomputed_pads_cpu",)
 ]
 
 for test_group in tests_expected_to_fail:
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 52148150e81562..ba3c6a9d052b80 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -399,6 +399,7 @@ set(MULTI_TEST_SRC
     backend/pad.in.cpp
     backend/parameter_as_output.in.cpp
     backend/power.in.cpp
+    backend/prelu.in.cpp
     backend/proposal.in.cpp
     backend/psroi_pooling.in.cpp
     backend/range.in.cpp
diff --git a/ngraph/test/backend/fused_op.in.cpp b/ngraph/test/backend/fused_op.in.cpp
index 74117a7fb9f69b..4b6bb15cb5171f 100644
--- a/ngraph/test/backend/fused_op.in.cpp
+++ b/ngraph/test/backend/fused_op.in.cpp
@@ -41,22 +41,6 @@ static string s_manifest = "${MANIFEST}";
 
 using TestEngine = test::ENGINE_CLASS_NAME(${BACKEND_NAME});
 
-NGRAPH_TEST(${BACKEND_NAME}, prelu)
-{
-    Shape shape{3, 2};
-    Shape rshape{3};
-    auto A = make_shared<op::Parameter>(element::f32, shape);
-    auto B = make_shared<op::Parameter>(element::f32, rshape);
-    auto prelu = make_shared<op::PRelu>(A, B);
-    auto f = make_shared<Function>(NodeVector{prelu}, ParameterVector{A, B});
-    std::vector<float> a{-2, 3, -2, 1, -1, 0};
-    std::vector<float> b{0, 0.5, 1};
-
-    auto test_case = test::TestCase<TestEngine>(f);
-    test_case.add_multiple_inputs<float>({a, b});
-    test_case.add_expected_output<float>(vector<float>{0, 3, -1, 1, -1, 0});
-    test_case.run();
-}
 
 NGRAPH_TEST(${BACKEND_NAME}, hardsigmoid)
 {
@@ -94,39 +78,6 @@ NGRAPH_TEST(${BACKEND_NAME}, hardsigmoid)
     test_case.run();
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, prelu_shared_slope)
-{
-    Shape shape{3, 2};
-    Shape rshape{};
-    auto A = make_shared<op::Parameter>(element::f32, shape);
-    auto B = make_shared<op::Parameter>(element::f32, rshape);
-    auto prelu = make_shared<op::PRelu>(A, B);
-    auto f = make_shared<Function>(NodeVector{prelu}, ParameterVector{A, B});
-    std::vector<float> a{-2, 3, -2, 1, -1, 0};
-    std::vector<float> b{0.5};
-
-    auto test_case = test::TestCase<TestEngine>(f);
-    test_case.add_multiple_inputs<float>({a, b});
-    test_case.add_expected_output<float>(vector<float>{-1, 3, -1, 1, -0.5, 0});
-    test_case.run();
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, prelu_negative_slope)
-{
-    Shape shape{3, 2};
-    Shape rshape{};
-    auto A = make_shared<op::Parameter>(element::f32, shape);
-    auto B = make_shared<op::Parameter>(element::f32, rshape);
-    auto prelu = make_shared<op::PRelu>(A, B);
-    auto f = make_shared<Function>(NodeVector{prelu}, ParameterVector{A, B});
-    std::vector<float> a{-2, 3, -2, 1, -1, 0};
-    std::vector<float> b{-0.5};
-
-    auto test_case = test::TestCase<TestEngine>(f);
-    test_case.add_multiple_inputs<float>({a, b});
-    test_case.add_expected_output<float>(vector<float>{1, 3, 1, 1, 0.5, 0});
-    test_case.run();
-}
 
 NGRAPH_TEST(${BACKEND_NAME}, space_to_depth_block_first)
 {
diff --git a/ngraph/test/backend/prelu.in.cpp b/ngraph/test/backend/prelu.in.cpp
new file mode 100644
index 00000000000000..8523aeb39c3568
--- /dev/null
+++ b/ngraph/test/backend/prelu.in.cpp
@@ -0,0 +1,942 @@
+// Co pyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/engine/test_engines.hpp"
+#include "util/test_case.hpp"
+#include "util/test_control.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+using TestEngine = test::ENGINE_CLASS_NAME(${BACKEND_NAME});
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_1d_1_param)
+{
+    Shape shape_a{6};
+    Shape shape_slope{1};
+
+    std::vector<float> a{1, 2, -3, -4, 5, 6};
+    std::vector<float> slope{2};
+    std::vector<float> out{1, 2, -6, -8, 5, 6};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_1d_1_const)
+{
+    Shape shape_a{6};
+    Shape shape_slope{1};
+
+    std::vector<float> a{1, 2, -3, -4, 5, 6};
+    std::vector<float> slope{2};
+    std::vector<float> out{1, 2, -6, -8, 5, 6};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_1d_param)
+{
+    Shape shape_a{6};
+    Shape shape_slope{6};
+
+    std::vector<float> a{1, 2, -3, -4, 5, 6};
+    std::vector<float> slope{2, 3, 4, 5, 6, 7};
+    std::vector<float> out{1, 2, -12, -20, 5, 6};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_1d_6_const)
+{
+    Shape shape_a{6};
+    Shape shape_slope{6};
+
+    std::vector<float> a{1, 2, -3, -4, 5, 6};
+    std::vector<float> slope{2, 3, 4, 5, 6, 7};
+    std::vector<float> out{1, 2, -12, -20, 5, 6};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_2d_2_W_param)
+{
+    Shape shape_a{3, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-2, 3, -2, 1, -1, 0};
+    std::vector<float> slope{0, 1};
+    std::vector<float> out{0, 3, 0, 1, 0, 0};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_2d_2_W_const)
+{
+    Shape shape_a{3, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-2, 3, -2, 1, -1, 0};
+    std::vector<float> slope{0, 1};
+    std::vector<float> out{0, 3, 0, 1, 0, 0};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_2d_6_W_param)
+{
+    Shape shape_a{2, 6};
+    Shape shape_slope{6};
+
+    std::vector<float> a{1, 2, -3, -4, 5, 6,
+                         7, 8, -9, -10, 11, 12};
+    std::vector<float> slope{1, 2, 3, 4, 5, 6};
+    std::vector<float> out{1, 2, -9, -16, 5, 6,
+                            7, 8, -27, -40, 11, 12};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_2d_6_W_const)
+{
+    Shape shape_a{2, 6};
+    Shape shape_slope{6};
+
+    std::vector<float> a{1, 2, -3, -4, 5, 6,
+                         7, 8, -9, -10, 11, 12};
+    std::vector<float> slope{1, 2, 3, 4, 5, 6};
+    std::vector<float> out{1, 2, -9, -16, 5, 6,
+                            7, 8, -27, -40, 11, 12};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_2d_C_2_param)
+{
+    Shape shape_a{3, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-1, -1, -1, -1, -1, -1};
+    std::vector<float> slope{2, 0.5};
+    std::vector<float> out{-2, -0.5, -2, -0.5, -2, -0.5};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_2d_C_2_const)
+{
+    Shape shape_a{3, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-1, -1, -1, -1, -1, -1};
+    std::vector<float> slope{2, 0.5};
+    std::vector<float> out{-2, -0.5, -2, -0.5, -2, -0.5};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_negative_equal_dims_slope_param)
+{
+    Shape shape_a{2, 2, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-0.5, -2, -3, -4, -5, -6, -7, -8};
+    std::vector<float> slope{-0.5, -2};
+    // std::vector<float> out{0.25, 4, 1.5, 8, 2.5, 12, 3.5, 16}; // broadcast (1, 1, 2)
+    std::vector<float> out{0.25, 1, 6, 8, 2.5, 3, 14, 16}; // broadcast (1, 2, 1)
+    // std::vector<float> out{0.25, 1, 1.5, 2, 10, 12, 14, 16}; // broadcast (2, 1, 1)
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_negative_equal_dims_slope_const)
+{
+    Shape shape_a{2, 2, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-0.5, -2, -3, -4, -5, -6, -7, -8};
+    std::vector<float> slope{-0.5, -2};
+    // std::vector<float> out{0.25, 4, 1.5, 8, 2.5, 12, 3.5, 16}; // broadcast (1, 1, 2)
+    std::vector<float> out{0.25, 1, 6, 8, 2.5, 3, 14, 16}; // broadcast (1, 2, 1)
+    // std::vector<float> out{0.25, 1, 1.5, 2, 10, 12, 14, 16}; // broadcast (2, 1, 1)
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_negative_slope_param)
+{
+    Shape shape_a{3, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-2, 3, -2, 1, -1, 0};
+    std::vector<float> slope{-0.5, -1};
+    std::vector<float> out{1, 3, 1, 1, 0.5, 0};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_negative_slope_const)
+{
+    Shape shape_a{3, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-2, 3, -2, 1, -1, 0};
+    std::vector<float> slope{-0.5, -1};
+    std::vector<float> out{1, 3, 1, 1, 0.5, 0};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_2d_same_shape_param)
+{
+    Shape shape_a{2, 6};
+    Shape shape_slope{2, 6};
+
+    std::vector<float> a{1, 2, -3, -4, 5, 6,
+                         1, 2, -3, -4, 5, 6};
+    std::vector<float> slope{2, 2, 2, 2, 2, 2,
+                         1, 1, 4, 2, 1, 1};
+    std::vector<float> out{1, 2, -6, -8, 5, 6, 1, 2, -12, -8, 5, 6};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_2d_same_shape_const)
+{
+    Shape shape_a{2, 6};
+    Shape shape_slope{2, 6};
+
+    std::vector<float> a{1, 2, -3, -4, 5, 6,
+                         1, 2, -3, -4, 5, 6};
+    std::vector<float> slope{2, 2, 2, 2, 2, 2,
+                         1, 1, 4, 2, 1, 1};
+    std::vector<float> out{1, 2, -6, -8, 5, 6, 1, 2, -12, -8, 5, 6};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_C_1_W_slope_param)
+{
+    Shape shape_a{2, 2, 2, 2};
+    Shape shape_slope{2, 1, 2};
+
+    std::vector<float> a{1, 2, -3, -4, 1, 2, -3, -4, 1, 2, -3, -4, 1, 2, -3, -4};
+    std::vector<float> slope{1, 2, 3, 4};
+    std::vector<float> out{1, 2, -3, -8, 1, 2, -9, -16, 1, 2, -3, -8, 1, 2, -9, -16};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_C_1_W_slope_const)
+{
+    Shape shape_a{2, 2, 2, 2};
+    Shape shape_slope{2, 1, 2};
+
+    std::vector<float> a{1, 2, -3, -4, 1, 2, -3, -4, 1, 2, -3, -4, 1, 2, -3, -4};
+    std::vector<float> slope{1, 2, 3, 4};
+    std::vector<float> out{1, 2, -3, -8, 1, 2, -9, -16, 1, 2, -3, -8, 1, 2, -9, -16};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_4d_slope_1_C_1_W_param)
+{
+    Shape shape_a{2, 2, 2, 2};
+    Shape shape_slope{1, 2, 1, 2};
+
+    std::vector<float> a{1, 2, -3, -4, 1, 2, -3, -4, 1, 2, -3, -4, 1, 2, -3, -4};
+    std::vector<float> slope{1, 2, 3, 4};
+    std::vector<float> out{1, 2, -3, -8, 1, 2, -9, -16, 1, 2, -3, -8, 1, 2, -9, -16};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_4d_slope_1_C_1_W_const)
+{
+    Shape shape_a{2, 2, 2, 2};
+    Shape shape_slope{1, 2, 1, 2};
+
+    std::vector<float> a{1, 2, -3, -4, 1, 2, -3, -4, 1, 2, -3, -4, 1, 2, -3, -4};
+    std::vector<float> slope{1, 2, 3, 4};
+    std::vector<float> out{1, 2, -3, -8, 1, 2, -9, -16, 1, 2, -3, -8, 1, 2, -9, -16};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_3d_W_param)
+{
+    Shape shape_a{2, 2, 6};
+    Shape shape_slope{2, 1, 6};
+
+
+    std::vector<float> a{1, 2, -3, -4, -5, 6,
+                         -1, -2, -3, -4, -5, -6,
+                         1, 2, -3, -4, 5, 6,
+                         -2, 4, -6, -8, 10, 12};
+    std::vector<float> slope{2, 1, 3, 4, 1, 7,
+                         1, 2, 3, 4, 5, 6};
+    std::vector<float> out{1, 2, -9, -16, -5, 6,
+                            -2, -2, -9, -16, -5, -42,
+                            1, 2, -9, -16, 5, 6,
+                            -2, 4, -18, -32, 10, 12};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_3d_W_const)
+{
+    Shape shape_a{2, 2, 6};
+    Shape shape_slope{2, 1, 6};
+
+    std::vector<float> a{1, 2, -3, -4, -5, 6,
+                         -1, -2, -3, -4, -5, -6,
+                         1, 2, -3, -4, 5, 6,
+                         -2, 4, -6, -8, 10, 12};
+    std::vector<float> slope{2, 1, 3, 4, 1, 7,
+                         1, 2, 3, 4, 5, 6};
+    std::vector<float> out{1, 2, -9, -16, -5, 6,
+                            -2, -2, -9, -16, -5, -42,
+                            1, 2, -9, -16, 5, 6,
+                            -2, 4, -18, -32, 10, 12};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_3d_same_shape_param)
+{
+    Shape shape_a{2, 3, 2};
+    Shape shape_slope{2, 3, 2};
+
+    std::vector<float> a{1, 2, -3, -4, -5, 6,
+                         -1, -2, -3, -4,-5, -6,};
+    std::vector<float> slope{2, 1, 3, 4, 1, 7,
+                            1, 2, 3, 4, 5, 6};
+    std::vector<float> out{1, 2, -9, -16, -5, 6,
+                            -1, -4, -9, -16, -25, -36};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_3d_same_shape_const)
+{
+    Shape shape_a{2, 3, 2};
+    Shape shape_slope{2, 3, 2};
+
+    std::vector<float> a{1, 2, -3, -4, -5, 6,
+                         -1, -2, -3, -4,-5, -6,};
+    std::vector<float> slope{2, 1, 3, 4, 1, 7,
+                            1, 2, 3, 4, 5, 6};
+    std::vector<float> out{1, 2, -9, -16, -5, 6,
+                            -1, -4, -9, -16, -25, -36};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_3d_broadcast_C_W_slope_param)
+{
+    Shape shape_a{2, 1, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-10, -10, -10, -10};
+    std::vector<float> slope{0.1, 10};
+    std::vector<float> out{-1, -100, -1, -100}; // broadcast (1, 2, 1)
+    // std::vector<float> out{-1, -1, -100, -100}; // broadcast (2, 1, 1)
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_3d_broadcast_C_W_slope_const)
+{
+    Shape shape_a{2, 1, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-10, -10, -10, -10};
+    std::vector<float> slope{0.1, 10};
+    std::vector<float> out{-1, -100, -1, -100}; // broadcast (1, 2, 1)
+    // std::vector<float> out{-1, -1, -100, -100}; // broadcast (2, 1, 1)
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_4d_broadcast_C_W_slope_param)
+{
+    Shape shape_a{1, 2, 1, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-10, -10, -10, -10};
+    std::vector<float> slope{0.1, 10};
+    // std::vector<float> out{-1, -100, -1, -100}; // broadcast (1, 1, 1, 2)
+    std::vector<float> out{-1, -1, -100, -100}; // broadcast (1, 2, 1, 1)
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_4d_broadcast_C_W_slope_const)
+{
+    Shape shape_a{1, 2, 1, 2};
+    Shape shape_slope{2};
+
+    std::vector<float> a{-10, -10, -10, -10};
+    std::vector<float> slope{0.1, 10};
+    // std::vector<float> out{-1, -100, -1, -100}; // broadcast (1, 1, 1, 2)
+    std::vector<float> out{-1, -1, -100, -100}; // broadcast (1, 2, 1, 1)
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_4d_broadcast_C_slope_param)
+{
+    Shape shape_a{1, 5, 1, 1};
+    Shape shape_slope{5};
+
+    std::vector<float> a{-1, 0, -1, -1, -1};
+    std::vector<float> slope{1, 2, 3, 4, 5};
+    std::vector<float> out{-1, 0, -3, -4, -5};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_4d_broadcast_C_slope_const)
+{
+    Shape shape_a{1, 5, 1, 1};
+    Shape shape_slope{5};
+
+    std::vector<float> a{-1, 0, -1, -1, -1};
+    std::vector<float> slope{1, 2, 3, 4, 5};
+    std::vector<float> out{-1, 0, -3, -4, -5};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_batch_nd_elementwise_param)
+{
+    Shape shape_a{2, 3, 4, 5};
+    Shape shape_slope{2, 3, 4, 5};
+
+    std::vector<float> a{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.};
+    std::vector<float> slope(shape_size(shape_slope));
+    std::iota(std::begin(slope), std::end(slope), 0);
+
+    std::vector<float> out{-0.,   -1.,   -2.,   -3.,   -4.,   -5.,   -6.,   -7.,   -8.,
+         -9.,  -10.,  -11.,  -12.,  -13.,  -14.,  -15.,  -16.,  -17.,
+        -18.,  -19.,  -20.,  -21.,  -22.,  -23.,  -24.,  -25.,  -26.,
+        -27.,  -28.,  -29.,  -30.,  -31.,  -32.,  -33.,  -34.,  -35.,
+        -36.,  -37.,  -38.,  -39.,  -40.,  -41.,  -42.,  -43.,  -44.,
+        -45.,  -46.,  -47.,  -48.,  -49.,  -50.,  -51.,  -52.,  -53.,
+        -54.,  -55.,  -56.,  -57.,  -58.,  -59.,  -60.,  -61.,  -62.,
+        -63.,  -64.,  -65.,  -66.,  -67.,  -68.,  -69.,  -70.,  -71.,
+        -72.,  -73.,  -74.,  -75.,  -76.,  -77.,  -78.,  -79.,  -80.,
+        -81.,  -82.,  -83.,  -84.,  -85.,  -86.,  -87.,  -88.,  -89.,
+        -90.,  -91.,  -92.,  -93.,  -94.,  -95.,  -96.,  -97.,  -98.,
+        -99., -100., -101., -102., -103., -104., -105., -106., -107.,
+       -108., -109., -110., -111., -112., -113., -114., -115., -116.,
+       -117., -118., -119.};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_batch_nd_elementwise_const)
+{
+    Shape shape_a{2, 3, 4, 5};
+    Shape shape_slope{2, 3, 4, 5};
+
+    std::vector<float> a{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.};
+    std::vector<float> slope(shape_size(shape_slope));
+    std::iota(std::begin(slope), std::end(slope), 0);
+
+    std::vector<float> out{-0.,   -1.,   -2.,   -3.,   -4.,   -5.,   -6.,   -7.,   -8.,
+         -9.,  -10.,  -11.,  -12.,  -13.,  -14.,  -15.,  -16.,  -17.,
+        -18.,  -19.,  -20.,  -21.,  -22.,  -23.,  -24.,  -25.,  -26.,
+        -27.,  -28.,  -29.,  -30.,  -31.,  -32.,  -33.,  -34.,  -35.,
+        -36.,  -37.,  -38.,  -39.,  -40.,  -41.,  -42.,  -43.,  -44.,
+        -45.,  -46.,  -47.,  -48.,  -49.,  -50.,  -51.,  -52.,  -53.,
+        -54.,  -55.,  -56.,  -57.,  -58.,  -59.,  -60.,  -61.,  -62.,
+        -63.,  -64.,  -65.,  -66.,  -67.,  -68.,  -69.,  -70.,  -71.,
+        -72.,  -73.,  -74.,  -75.,  -76.,  -77.,  -78.,  -79.,  -80.,
+        -81.,  -82.,  -83.,  -84.,  -85.,  -86.,  -87.,  -88.,  -89.,
+        -90.,  -91.,  -92.,  -93.,  -94.,  -95.,  -96.,  -97.,  -98.,
+        -99., -100., -101., -102., -103., -104., -105., -106., -107.,
+       -108., -109., -110., -111., -112., -113., -114., -115., -116.,
+       -117., -118., -119.};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_1d_W_slope_param)
+{
+    Shape shape_a{2, 3, 4, 5};
+    Shape shape_slope{5};
+
+    std::vector<float> a{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.};
+
+    std::vector<float> slope{0, 1, 2, 3, 4};
+
+    std::vector<float> out{-0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2.,
+       -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0.,
+       -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3.,
+       -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1.,
+       -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4.,
+       -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2.,
+       -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0.,
+       -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3.,
+       -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1.,
+       -2., -3., -4.};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_1d_W_slope_const)
+{
+    Shape shape_a{2, 3, 4, 5};
+    Shape shape_slope{5};
+
+    std::vector<float> a{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.};
+
+    std::vector<float> slope{0, 1, 2, 3, 4};
+
+    std::vector<float> out{-0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2.,
+       -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0.,
+       -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3.,
+       -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1.,
+       -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4.,
+       -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2.,
+       -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0.,
+       -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3.,
+       -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1.,
+       -2., -3., -4.};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_1d_C_slope_param)
+{
+    Shape shape_a{2, 3, 4, 5};
+    Shape shape_slope{3};
+
+    std::vector<float> a{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.};
+
+    std::vector<float> slope{0, 1, 2};
+
+    std::vector<float> out{-0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0.,
+       -0., -0., -0., -0., -0., -0., -0., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2.,
+       -2., -2., -2., -2., -2., -2., -2., -2., -0., -0., -0., -0., -0.,
+       -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0.,
+       -0., -0., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -2., -2., -2., -2.,
+       -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2.,
+       -2., -2., -2.};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_1d_C_slope_const)
+{
+    Shape shape_a{2, 3, 4, 5};
+    Shape shape_slope{3};
+
+    std::vector<float> a{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.};
+
+    std::vector<float> slope{0, 1, 2};
+
+    std::vector<float> out{-0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0.,
+       -0., -0., -0., -0., -0., -0., -0., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2.,
+       -2., -2., -2., -2., -2., -2., -2., -2., -0., -0., -0., -0., -0.,
+       -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0.,
+       -0., -0., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -2., -2., -2., -2.,
+       -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2.,
+       -2., -2., -2.};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    const auto f = make_shared<Function>(make_shared<op::v0::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_C_1_1_slope_param)
+{
+    Shape shape_a{2, 3, 4, 5};
+    Shape shape_slope{3, 1, 1};
+
+    std::vector<float> a{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.};
+
+    std::vector<float> slope{0, 1, 2};
+
+    std::vector<float> out{-0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0.,
+       -0., -0., -0., -0., -0., -0., -0., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2.,
+       -2., -2., -2., -2., -2., -2., -2., -2., -0., -0., -0., -0., -0.,
+       -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0.,
+       -0., -0., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -2., -2., -2., -2.,
+       -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2.,
+       -2., -2., -2.};
+
+    const auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    const auto SLOPE = make_shared<op::Parameter>(element::f32, shape_slope);
+    const auto f = make_shared<Function>(make_shared<op::PRelu>(A, SLOPE), ParameterVector{A, SLOPE});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a, slope});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, prelu_C_1_1_slope_const)
+{
+    Shape shape_a{2, 3, 4, 5};
+    Shape shape_slope{3, 1, 1};
+
+    std::vector<float> a{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.};
+
+    std::vector<float> slope{0, 1, 2};
+
+    std::vector<float> out{-0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0.,
+       -0., -0., -0., -0., -0., -0., -0., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2.,
+       -2., -2., -2., -2., -2., -2., -2., -2., -0., -0., -0., -0., -0.,
+       -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0.,
+       -0., -0., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -2., -2., -2., -2.,
+       -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2.,
+       -2., -2., -2.};
+
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    auto SLOPE = make_shared<op::Constant>(element::f32, shape_slope, slope);
+    auto f = make_shared<Function>(make_shared<op::PRelu>(A, SLOPE), ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>({a});
+    test_case.add_expected_output<float>(shape_a, out);
+    test_case.run();
+}
diff --git a/ngraph/test/models/onnx/prelu_1d.prototxt b/ngraph/test/models/onnx/prelu_1d.prototxt
new file mode 100644
index 00000000000000..9034d1d42f6137
--- /dev/null
+++ b/ngraph/test/models/onnx/prelu_1d.prototxt
@@ -0,0 +1,72 @@
+ir_version: 7
+producer_name: "onnx-importer-test"
+graph {
+  node {
+    input: "X"
+    input: "SLOPE"
+    output: "Y"
+    op_type: "PRelu"
+  }
+  name: "test-model-prelu"
+  input {
+    name: "X"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 2
+          }
+          dim {
+            dim_value: 3
+          }
+          dim {
+            dim_value: 4
+          }
+          dim {
+            dim_value: 5
+          }
+        }
+      }
+    }
+  }
+  input {
+    name: "SLOPE"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 5
+          }
+        }
+      }
+    }
+  }
+  output {
+    name: "Y"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 2
+          }
+          dim {
+            dim_value: 3
+          }
+          dim {
+            dim_value: 4
+          }
+          dim {
+            dim_value: 5
+          }
+        }
+      }
+    }
+  }
+}
+opset_import {
+  domain: ""
+  version: 12
+}
diff --git a/ngraph/test/models/onnx/prelu_batch_nd.prototxt b/ngraph/test/models/onnx/prelu_batch_nd.prototxt
new file mode 100644
index 00000000000000..cf693db83c379e
--- /dev/null
+++ b/ngraph/test/models/onnx/prelu_batch_nd.prototxt
@@ -0,0 +1,81 @@
+ir_version: 7
+producer_name: "onnx-importer-test"
+graph {
+  node {
+    input: "X"
+    input: "SLOPE"
+    output: "Y"
+    op_type: "PRelu"
+  }
+  name: "test-model-prelu"
+  input {
+    name: "X"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 2
+          }
+          dim {
+            dim_value: 3
+          }
+          dim {
+            dim_value: 4
+          }
+          dim {
+            dim_value: 5
+          }
+        }
+      }
+    }
+  }
+  input {
+    name: "SLOPE"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 2
+          }
+          dim {
+            dim_value: 3
+          }
+          dim {
+            dim_value: 4
+          }
+          dim {
+            dim_value: 5
+          }
+        }
+      }
+    }
+  }
+  output {
+    name: "Y"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 2
+          }
+          dim {
+            dim_value: 3
+          }
+          dim {
+            dim_value: 4
+          }
+          dim {
+            dim_value: 5
+          }
+        }
+      }
+    }
+  }
+}
+opset_import {
+  domain: ""
+  version: 12
+}
diff --git a/ngraph/test/models/onnx/prelu_c_1_1.prototxt b/ngraph/test/models/onnx/prelu_c_1_1.prototxt
new file mode 100644
index 00000000000000..83f2a748860578
--- /dev/null
+++ b/ngraph/test/models/onnx/prelu_c_1_1.prototxt
@@ -0,0 +1,78 @@
+ir_version: 7
+producer_name: "onnx-importer-test"
+graph {
+  node {
+    input: "X"
+    input: "SLOPE"
+    output: "Y"
+    op_type: "PRelu"
+  }
+  name: "test-model-prelu"
+  input {
+    name: "X"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 2
+          }
+          dim {
+            dim_value: 3
+          }
+          dim {
+            dim_value: 4
+          }
+          dim {
+            dim_value: 5
+          }
+        }
+      }
+    }
+  }
+  input {
+    name: "SLOPE"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 3
+          }
+          dim {
+            dim_value: 1
+          }
+          dim {
+            dim_value: 1
+          }
+        }
+      }
+    }
+  }
+  output {
+    name: "Y"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 2
+          }
+          dim {
+            dim_value: 3
+          }
+          dim {
+            dim_value: 4
+          }
+          dim {
+            dim_value: 5
+          }
+        }
+      }
+    }
+  }
+}
+opset_import {
+  domain: ""
+  version: 12
+}
diff --git a/ngraph/test/onnx/onnx_import.in.cpp b/ngraph/test/onnx/onnx_import.in.cpp
index 6a5ecfdcdf8ef8..a8522c08fa1729 100644
--- a/ngraph/test/onnx/onnx_import.in.cpp
+++ b/ngraph/test/onnx/onnx_import.in.cpp
@@ -2101,7 +2101,7 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_model_leaky_relu)
     test_case.run();
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, onnx_model_prelu)
+NGRAPH_TEST(${BACKEND_NAME}, onnx_model_prelu_nd)
 {
     auto function =
         onnx_import::import_onnx_model(file_util::path_join(SERIALIZED_ZOO, "onnx/prelu.prototxt"));
@@ -2133,6 +2133,129 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_model_prelu)
     test_case.run();
 }
 
+NGRAPH_TEST(${BACKEND_NAME}, onnx_model_prelu_batch_nd_elementwise)
+{
+    auto function =
+        onnx_import::import_onnx_model(file_util::path_join(SERIALIZED_ZOO, "onnx/prelu_batch_nd.prototxt"));
+
+    Inputs inputs;
+    // Shape{2, 3, 4, 5}
+    inputs.emplace_back(std::vector<float>{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.});
+
+    // Shape{2, 3, 4, 5}
+    std::vector<float> slope(shape_size(Shape{2, 3, 4, 5}));
+    std::iota(std::begin(slope), std::end(slope), 0);
+    inputs.emplace_back(slope);
+
+    // Shape{2, 3, 4, 5}
+    auto expected_output = std::vector<float>{-0.,   -1.,   -2.,   -3.,   -4.,   -5.,   -6.,   -7.,   -8.,
+         -9.,  -10.,  -11.,  -12.,  -13.,  -14.,  -15.,  -16.,  -17.,
+        -18.,  -19.,  -20.,  -21.,  -22.,  -23.,  -24.,  -25.,  -26.,
+        -27.,  -28.,  -29.,  -30.,  -31.,  -32.,  -33.,  -34.,  -35.,
+        -36.,  -37.,  -38.,  -39.,  -40.,  -41.,  -42.,  -43.,  -44.,
+        -45.,  -46.,  -47.,  -48.,  -49.,  -50.,  -51.,  -52.,  -53.,
+        -54.,  -55.,  -56.,  -57.,  -58.,  -59.,  -60.,  -61.,  -62.,
+        -63.,  -64.,  -65.,  -66.,  -67.,  -68.,  -69.,  -70.,  -71.,
+        -72.,  -73.,  -74.,  -75.,  -76.,  -77.,  -78.,  -79.,  -80.,
+        -81.,  -82.,  -83.,  -84.,  -85.,  -86.,  -87.,  -88.,  -89.,
+        -90.,  -91.,  -92.,  -93.,  -94.,  -95.,  -96.,  -97.,  -98.,
+        -99., -100., -101., -102., -103., -104., -105., -106., -107.,
+       -108., -109., -110., -111., -112., -113., -114., -115., -116.,
+       -117., -118., -119.};
+
+    auto test_case = test::TestCase<TestEngine>(function);
+    test_case.add_multiple_inputs(inputs);
+    test_case.add_expected_output(expected_output);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, onnx_model_prelu_1d)
+{
+    auto function =
+        onnx_import::import_onnx_model(file_util::path_join(SERIALIZED_ZOO, "onnx/prelu_1d.prototxt"));
+
+    Inputs inputs;
+    // Shape{2, 3, 4, 5}
+    inputs.emplace_back(std::vector<float>{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.});
+
+    // Shape{5}
+    inputs.emplace_back(std::vector<float>{0, 1, 2, 3, 4});
+
+    // Shape{2, 3, 4, 5}
+    auto expected_output = std::vector<float>{-0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2.,
+       -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0.,
+       -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3.,
+       -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1.,
+       -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4.,
+       -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2.,
+       -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0.,
+       -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1., -2., -3.,
+       -4., -0., -1., -2., -3., -4., -0., -1., -2., -3., -4., -0., -1.,
+       -2., -3., -4.};
+
+    auto test_case = test::TestCase<TestEngine>(function);
+    test_case.add_multiple_inputs(inputs);
+    test_case.add_expected_output(expected_output);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, onnx_model_prelu_C_1_1)
+{
+    auto function =
+        onnx_import::import_onnx_model(file_util::path_join(SERIALIZED_ZOO, "onnx/prelu_c_1_1.prototxt"));
+
+    Inputs inputs;
+    // Shape{2, 3, 4, 5}
+    inputs.emplace_back(std::vector<float>{-1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1.});
+
+    // Shape{3, 1, 1}
+    inputs.emplace_back(std::vector<float>{0, 1, 2});
+
+    // Shape{2, 3, 4, 5}
+    auto expected_output = std::vector<float>{-0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0.,
+       -0., -0., -0., -0., -0., -0., -0., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2.,
+       -2., -2., -2., -2., -2., -2., -2., -2., -0., -0., -0., -0., -0.,
+       -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0., -0.,
+       -0., -0., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
+       -1., -1., -1., -1., -1., -1., -1., -1., -1., -2., -2., -2., -2.,
+       -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2., -2.,
+       -2., -2., -2.};
+
+    auto test_case = test::TestCase<TestEngine>(function);
+    test_case.add_multiple_inputs(inputs);
+    test_case.add_expected_output(expected_output);
+    test_case.run();
+}
+
 NGRAPH_TEST(${BACKEND_NAME}, onnx_model_selu)
 {
     auto function =
diff --git a/ngraph/test/runtime/ie/unit_test.manifest b/ngraph/test/runtime/ie/unit_test.manifest
index 9d6ea205f82839..c476f2dfac3137 100644
--- a/ngraph/test/runtime/ie/unit_test.manifest
+++ b/ngraph/test/runtime/ie/unit_test.manifest
@@ -758,9 +758,6 @@ lrn_2d_across_outermost_axis
 dyn_group_convolution_backprop_filters
 
 # Layer <op_name> input port 1 is not connected to any data
-prelu
-prelu_shared_slope
-prelu_negative_slope
 convolution_simple_padding
 
 # Cannot cast ngraph node Stack to CNNLayer!

From 6022df66877219abed3e6f0705055d2edb460d1a Mon Sep 17 00:00:00 2001
From: Maria Kaglinskaya <maria.kaglinskaya@intel.com>
Date: Tue, 8 Jun 2021 09:49:53 +0300
Subject: [PATCH 012/178] Pruning with FQ support (#5925)

* Add FQ, Concat, exted Eltwise support

* Fix tests after rebase + small refactoring

* Added Reshape on GroupConv weights mask propagating

* Added printing of  reduced weights to test transformation

* Turn off pruning for test

* Fixed comments + revert transformation comment

* Fixed last comments
---
 .../include/mask_attribute.hpp                |  80 ++++
 .../include/pruning.hpp                       |  11 +
 .../src/pruning/init_const_mask.cpp           |   2 +-
 .../src/pruning/init_masks.cpp                |  64 +++
 .../src/pruning/propagate_masks.cpp           | 411 +++++++++++++---
 .../src/pruning/pruning.cpp                   |   5 +
 .../src/pruning/shrink_weights.cpp            |   2 +
 .../transformations/pruning_test.cpp          | 444 +++++++++++++++++-
 8 files changed, 944 insertions(+), 75 deletions(-)
 create mode 100644 inference-engine/src/offline_transformations/src/pruning/init_masks.cpp

diff --git a/inference-engine/src/offline_transformations/include/mask_attribute.hpp b/inference-engine/src/offline_transformations/include/mask_attribute.hpp
index 70cce141567192..48c5b4ee9f0864 100644
--- a/inference-engine/src/offline_transformations/include/mask_attribute.hpp
+++ b/inference-engine/src/offline_transformations/include/mask_attribute.hpp
@@ -54,10 +54,90 @@ class Mask : public std::vector<std::set<uint64_t>>,
                            });
     }
 
+    std::vector<size_t> get_not_empty_dims() {
+        std::vector<size_t> not_empty_dims;
+        for (size_t i = 0; i < this->size(); i++) {
+            if (!this->at(i).empty())
+                not_empty_dims.push_back(i);
+        }
+        return not_empty_dims;
+    }
+
     bool is_shape_like() const { return m_is_shape_like; }
 
     void set_shape_like(bool flag) { m_is_shape_like = flag; }
 
+    void copy_value_from_mask(Mask *const mask) {
+        auto cur_mask_iter = begin();
+        auto mask_iter = mask->begin();
+        while (cur_mask_iter != end() && mask_iter != mask->end()) {
+            *cur_mask_iter = *mask_iter;
+
+            cur_mask_iter++;
+            mask_iter++;
+        }
+    }
+
+    void copy_value_from_mask_reversed(Mask *const mask) {
+        auto cur_mask_iter = rbegin();
+        auto mask_iter = mask->rbegin();
+        while (cur_mask_iter != rend() && mask_iter != mask->rend()) {
+            *cur_mask_iter = *mask_iter;
+
+            cur_mask_iter++;
+            mask_iter++;
+        }
+    }
+
+    Mask::Ptr intersect_masks_reversed(Mask *const mask) {
+        auto result_mask = std::make_shared<Mask>(std::max(size(), mask->size()));
+        auto result_iter = result_mask->rbegin();
+        auto mask_1_iter = rbegin();
+        auto mask_2_iter = mask->rbegin();
+
+        while (mask_1_iter != rend() &&
+               mask_2_iter != mask->rend()) {
+            // Merge mask dimension values for both masks
+            // Example: (MaskValue[1,2,3,4], MaskValue[2,3]) -> MaskValue[2,3]
+            for (const auto & value : *mask_1_iter) {
+                if (mask_2_iter->count(value)) {
+                    result_iter->insert(value);
+                }
+            }
+
+            result_iter++;
+            mask_1_iter++;
+            mask_2_iter++;
+        }
+        return result_mask;
+    }
+
+    Mask::Ptr union_masks_reversed(Mask *const mask) {
+        auto result_mask = std::make_shared<Mask>(std::max(size(), mask->size()));
+        auto result_iter = result_mask->rbegin();
+        auto mask_1_iter = rbegin();
+        auto mask_2_iter = mask->rbegin();
+
+        while (mask_1_iter != rend() &&
+               mask_2_iter != mask->rend()) {
+            // Union mask dimension values for both masks
+            // Example: (MaskValue[1,2,3,4], MaskValue[2, 5]) -> MaskValue[1, 2, 3, 4, 5]
+            for (const auto & value : *mask_1_iter) {
+                result_iter->insert(value);
+            }
+            for (const auto & value : *mask_2_iter) {
+                if (!result_iter->count(value)) {
+                    result_iter->insert(value);
+                }
+            }
+
+            result_iter++;
+            mask_1_iter++;
+            mask_2_iter++;
+        }
+        return result_mask;
+    }
+
     void add_callback(const std::function<bool(Mask::Ptr)> & receive_callback, Mask::Ptr mask) {
         m_callbacks[mask.get()] = receive_callback;
         m_dependencies.push_back(mask.get());
diff --git a/inference-engine/src/offline_transformations/include/pruning.hpp b/inference-engine/src/offline_transformations/include/pruning.hpp
index f398ab1713f588..7831ee1424678b 100644
--- a/inference-engine/src/offline_transformations/include/pruning.hpp
+++ b/inference-engine/src/offline_transformations/include/pruning.hpp
@@ -14,6 +14,7 @@ namespace ngraph {
 namespace pass {
 
 class InitConstMask;
+class InitMasks;
 class PropagateMasks;
 class ShrinkWeights;
 
@@ -22,6 +23,16 @@ class Pruning;
 } // namespace pass
 } // namespace ngraph
 
+/**
+ * @ingroup ie_transformation_common_api
+ * @brief Initialising masks for pruned operations
+ */
+class ngraph::pass::InitMasks : public ngraph::pass::GraphRewrite {
+public:
+    NGRAPH_RTTI_DECLARATION;
+    InitMasks();
+};
+
 /**
  * @ingroup ie_transformation_common_api
  * @brief Check Constant operation values by given dimensions and set
diff --git a/inference-engine/src/offline_transformations/src/pruning/init_const_mask.cpp b/inference-engine/src/offline_transformations/src/pruning/init_const_mask.cpp
index 73929487b80d18..01e9520082bbfb 100644
--- a/inference-engine/src/offline_transformations/src/pruning/init_const_mask.cpp
+++ b/inference-engine/src/offline_transformations/src/pruning/init_const_mask.cpp
@@ -17,7 +17,7 @@ NGRAPH_RTTI_DEFINITION(ngraph::pass::InitConstMask, "InitConstMask", 0);
 ngraph::pass::InitConstMask::InitConstMask(const ngraph::AxisSet & dims,
                                            const std::function<bool(const double & value)> & condition) {
     auto constant = pattern::wrap_type<opset6::Constant>(
-            pattern::type_matches_any({element::f16, element::f32, element::f64}));
+            pattern::type_matches_any({element::i8, element::u8, element::f16, element::f32, element::f64}));
 
     matcher_pass_callback callback = [=](pattern::Matcher& m) {
         auto const_node = std::dynamic_pointer_cast<opset6::Constant>(m.get_match_root());
diff --git a/inference-engine/src/offline_transformations/src/pruning/init_masks.cpp b/inference-engine/src/offline_transformations/src/pruning/init_masks.cpp
new file mode 100644
index 00000000000000..2b4394b6c32055
--- /dev/null
+++ b/inference-engine/src/offline_transformations/src/pruning/init_masks.cpp
@@ -0,0 +1,64 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "pruning.hpp"
+#include "mask_attribute.hpp"
+
+#include <ngraph/pattern/op/wrap_type.hpp>
+#include <ngraph/opsets/opset6.hpp>
+#include <ngraph/log.hpp>
+
+NGRAPH_RTTI_DEFINITION(ngraph::pass::InitMasks, "InitMasks", 0);
+
+namespace ngraph {
+namespace pass {
+namespace init_masks {
+
+class InitConvMask;
+
+} // namespace init_masks
+} // namespace pass
+} // namespace ngraph
+
+class ngraph::pass::init_masks::InitConvMask : public MatcherPass {
+public:
+    InitConvMask() {
+        auto input = pattern::any_input();
+        auto weights = pattern::any_input();
+        auto conv = pattern::wrap_type<opset6::Convolution, opset6::GroupConvolution>({input, weights});
+
+        ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher& m) {
+            const auto & pattern_map = m.get_pattern_value_map();
+            const auto & m_output = pattern_map.at(conv);
+
+            // Initializing weights mask:
+            // 1. Looking for Const node with weights
+            NodeVector weights_calculation_nodes;
+            auto cur_node = m_output.get_node()->get_input_node_shared_ptr(1);
+
+            while (!ngraph::is_type<opset6::Constant>(cur_node) && cur_node->inputs().size()) {
+                weights_calculation_nodes.push_back(cur_node);
+                cur_node = cur_node->get_input_node_shared_ptr(0);
+            }
+            if (!ngraph::is_type<opset6::Constant>(cur_node)) {
+                NGRAPH_DEBUG << "Can't find Constant weights for Convolution: " <<
+                m_output.get_node()->get_friendly_name() << std::endl;
+                return false;
+            }
+
+            // 2. Init mask for Const node
+            InitConstMask({0}/* check only output channels dim */).apply(cur_node);
+            return true;
+        };
+
+        auto m = std::make_shared<ngraph::pattern::Matcher>(conv, "ConvolutionInitMask");
+        register_matcher(m, callback);
+    }
+};
+
+
+ngraph::pass::InitMasks::InitMasks() {
+    add_matcher<init_masks::InitConvMask>();
+}
+
diff --git a/inference-engine/src/offline_transformations/src/pruning/propagate_masks.cpp b/inference-engine/src/offline_transformations/src/pruning/propagate_masks.cpp
index ac7a8e8b6859c7..424b6ae958317d 100644
--- a/inference-engine/src/offline_transformations/src/pruning/propagate_masks.cpp
+++ b/inference-engine/src/offline_transformations/src/pruning/propagate_masks.cpp
@@ -7,7 +7,9 @@
 
 #include <ngraph/pattern/op/wrap_type.hpp>
 #include <ngraph/opsets/opset6.hpp>
+#include <ngraph/opsets/opset5.hpp>
 #include <ngraph/log.hpp>
+#include <ngraph/rt_info.hpp>
 
 NGRAPH_RTTI_DEFINITION(ngraph::pass::PropagateMasks, "PropagateMasks", 0);
 
@@ -20,11 +22,23 @@ class GroupConvolution;
 class Elementwise;
 class PassThrough;
 class StopPropagation;
+class FakeQuantize;
+class Concat;
+class Reshape;
 
 } // namespace mask_propagation
 } // namespace pass
 } // namespace ngraph
 
+ngraph::Shape broadcast_shape_to_rank(ngraph::Shape shape_to_broadcast, int64_t dst_rank) {
+    auto initial_rank = static_cast<int64_t>(shape_to_broadcast.size());
+    auto num_of_broadcased_dims = dst_rank - initial_rank;
+    std::vector<size_t> dims(num_of_broadcased_dims, 1);
+    dims.insert(dims.end(), shape_to_broadcast.begin(), shape_to_broadcast.end());
+    auto new_shape = ngraph::Shape(dims);
+    return new_shape;
+}
+
 class ngraph::pass::mask_propagation::Convolution : public MatcherPass {
 public:
     Convolution() {
@@ -38,12 +52,15 @@ class ngraph::pass::mask_propagation::Convolution : public MatcherPass {
             const auto & m_output = pattern_map.at(conv);
             const auto & m_input = pattern_map.at(input);
 
-            // In case if weights are Constant we initialize Mask
-            InitConstMask({0}/* check only output channel */).apply(m_weights.get_node_shared_ptr());
-
             auto weights_mask = getMask(m_weights);
-            // If weights are not a Constant and we didn't set Mask value before we will get nullptr
-            if (!weights_mask) return false;
+
+            // Nullptr in weights-mask means that mask for this node wasn't initialized earlier.
+            // Weights mask for convolution should be initialized in the InitMasks pass (and propagate after it).
+            // If mask isn't initialized - this weights (and hence all convolution) can't be pruned for some reason.
+            if (!weights_mask) {
+                NGRAPH_DEBUG << "No weights mask for " << m_output.get_node()->get_friendly_name() << "\n";
+                return false;
+            }
             auto weights_mask_row = weights_mask.get();
 
             if (auto input_mask = getMask(m_input)) {
@@ -119,9 +136,15 @@ class ngraph::pass::mask_propagation::GroupConvolution : public MatcherPass {
 
             auto weights_mask = getMask(m_weights);
             if (!weights_mask) {
-                // TODO: only if weights are constant
-                weights_mask = std::make_shared<Mask>(weights_shape.size());
-                setMask(m_weights, weights_mask);
+                // Setting mask only if weights are constant
+                if (ngraph::is_type<opset6::Constant>(m_output.get_node_shared_ptr())) {
+                    weights_mask = std::make_shared<Mask>(weights_shape.size());
+                    setMask(m_weights, weights_mask);
+                } else {
+                    NGRAPH_DEBUG << "GroupConvolution: No weights mask and weights aren't constant for " <<
+                    *m_output.get_node() << "\n";
+                    return false;
+                }
             }
             auto weights_mask_row = weights_mask.get();
 
@@ -169,13 +192,85 @@ class ngraph::pass::mask_propagation::GroupConvolution : public MatcherPass {
     }
 };
 
+class ngraph::pass::mask_propagation::Reshape : public MatcherPass {
+public:
+    Reshape() {
+        auto input = pattern::any_input(pattern::has_static_shape());
+        auto shape = pattern::any_input();
+        // Working only for Reshapes on Group Convolution weights
+        auto reshape = pattern::wrap_type<opset6::Reshape>({input, shape}, pattern::consumers_count(1));
+        auto gconv = pattern::wrap_type<opset6::GroupConvolution>({pattern::any_input(), reshape},
+                                                                  pattern::has_static_shape());
+
+        ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher& m) {
+            const auto & pattern_map = m.get_pattern_value_map();
+            const auto & m_shape = pattern_map.at(shape);
+            const auto & m_output = pattern_map.at(reshape);
+            const auto & m_input = pattern_map.at(input);
+
+            auto shape_val = m_shape.get_node_shared_ptr();
+
+            // In Depthwise Convolutions Reshape on weights just add additional dimension for output channels count
+            // (1 in case of the depthwise) of kernel.
+            // Example: Reshape from [G, 1 (I), X, Y, Z] -> [G, 1 (O), 1 (I), X, Y, Z], where G - group numbers,
+            // X, Y, Z -  spartial dimensions (can be only X or X, Y), I, O - number of input/output channels of kernel.
+
+            // Checking that matched Reshape meets this conditions (add 1-d dim on 1 position of shape constant)
+            auto inp_shape = m_input.get_shape();
+            auto out_shape = m_output.get_shape();
+            inp_shape.insert(inp_shape.begin() + 1, 1);
+            if (inp_shape != out_shape) {
+                return false;
+            }
+
+            auto input_mask = getMask(m_input);
+            if (!input_mask) {
+                return false;
+            }
+            auto input_mask_row = input_mask.get();
+            auto output_mask = std::make_shared<Mask>(m_output.get_partial_shape().rank().get_length());
+            auto output_mask_row = output_mask.get();
+
+            // Depthwise Convolution pruned only by input channels (== groups) ->
+            // Propagating mask from Group (0) dim in Reshape input to Group (0) dim in Reshape output and back
+            input_mask->add_callback([output_mask_row](Mask::Ptr cur_mask) -> bool {
+                cur_mask->at(0) = output_mask_row->at(0);
+                return true;
+            }, output_mask);
+            output_mask->add_callback([input_mask_row](Mask::Ptr cur_mask) -> bool {
+                cur_mask->at(0) = input_mask_row->at(0);
+                return true;
+            }, input_mask);
+            input_mask->apply_callback(output_mask);
+
+            // To allow pruning on weights (allow reshape input Group (0) dim changing) replace Reshape Shape constant
+            // [G, 1, 1, X, Y, Z] by [-1, 1, 1, X, Y, Z].
+            auto old_shape_const = std::dynamic_pointer_cast<opset6::Constant>(m_shape.get_node_shared_ptr());
+            auto shape_value = old_shape_const.get()->cast_vector<int64_t>();
+            shape_value[0] = -1;
+            auto new_const = opset6::Constant::create(old_shape_const->get_element_type(),
+                                                      old_shape_const->get_shape(), shape_value);
+            new_const->set_friendly_name(old_shape_const->get_friendly_name());
+            ngraph::copy_runtime_info(old_shape_const, new_const);
+            ngraph::replace_node(old_shape_const, new_const);
+
+            setMask(m_output, output_mask);
+            return true;
+        };
+
+        auto m = std::make_shared<ngraph::pattern::Matcher>(reshape, "ReshapeMaskPropagation");
+        register_matcher(m, callback);
+    }
+};
+
 class ngraph::pass::mask_propagation::Elementwise : public MatcherPass {
 public:
     Elementwise() {
         auto input = pattern::any_input();
         auto weights = pattern::any_input();
-        auto eltwise = pattern::wrap_type<op::util::BinaryElementwiseArithmetic>({input, weights},
-                                                                                 pattern::has_static_rank());
+        auto eltwise = pattern::wrap_type<opset6::Add, opset6::Subtract, opset6::Maximum, opset6::Minimum,
+        opset6::Multiply>({input, weights}, pattern::has_static_rank());
+        // TODO: add Div, Power support
 
         ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher& m) {
             const auto & pattern_map = m.get_pattern_value_map();
@@ -183,82 +278,275 @@ class ngraph::pass::mask_propagation::Elementwise : public MatcherPass {
             const auto & m_output = pattern_map.at(eltwise);
             const auto & m_input = pattern_map.at(input);
 
-            // TODO: implement check that compares input shape ranks
+            // Case when input masks should be united instead of intersection
+            bool union_eltwise_type = ngraph::is_type<opset6::Multiply>(m_output.get_node_shared_ptr());
+
             const auto & input_rank = m_input.get_partial_shape().rank().get_length();
             const auto & weights_rank = m_weights.get_partial_shape().rank().get_length();
+            // Here assuming that masks can be propagated only through 3/4 dimensional tensors
+            // (since channel dim is necessary)
             if (weights_rank < 3 || input_rank < 3) return false;
 
-            // In case if one of the inputs is constant
-            // TODO: need to find channel dimension instead of hardcoded zero
-            const size_t & channel_dim = (input_rank == weights_rank ? 1 : 0);
-            InitConstMask({channel_dim}).apply(m_input.get_node_shared_ptr());
-            InitConstMask({channel_dim}).apply(m_weights.get_node_shared_ptr());
+            // In case if first of the inputs is constant
+            InitConstMask({0, 1/* potential output channel dim */}).apply(m_input.get_node_shared_ptr());
+            auto input_mask = getMask(m_input);
+            if (!input_mask) {
+                NGRAPH_DEBUG << "No input mask for: " << m_output.get_node()->get_friendly_name() << std::endl;
+                return false;
+            }
+
+            InitConstMask({0, 1}).apply(m_weights.get_node_shared_ptr());
 
             auto weights_mask = getMask(m_weights);
+            if (!weights_mask) {
+                NGRAPH_DEBUG << "No weights mask for: " << m_output.get_node()->get_friendly_name() << std::endl;
+                return false;
+            }
+            auto input_mask_row = input_mask.get();
+            auto weights_mask_row = weights_mask.get();
+
+            // Merging masks from two inputs
+            auto output_mask = std::make_shared<Mask>(m_output.get_partial_shape().rank().get_length());
+            auto output_mask_row = output_mask.get();
+
+            auto out_mask_callback = [input_mask_row, weights_mask_row, union_eltwise_type](Mask::Ptr cur_mask) -> bool {
+                Mask::Ptr result_mask;
+                if (union_eltwise_type) {
+                    result_mask = input_mask_row->union_masks_reversed(weights_mask_row);
+                } else {
+                    result_mask = input_mask_row->intersect_masks_reversed(weights_mask_row);
+                }
+                cur_mask->copy_value_from_mask_reversed(result_mask.get());
+                return true;
+            };
+            output_mask->add_callback(out_mask_callback, input_mask);
+
+            input_mask->add_callback([weights_mask_row](Mask::Ptr cur_mask) -> bool {
+                cur_mask->copy_value_from_mask_reversed(weights_mask_row);
+                return true;
+            }, weights_mask);
+            input_mask->add_callback([output_mask_row](Mask::Ptr cur_mask) -> bool {
+                cur_mask->copy_value_from_mask_reversed(output_mask_row);
+                return true;
+            }, output_mask);
+            weights_mask->add_callback([input_mask_row](Mask::Ptr cur_mask) -> bool {
+                cur_mask->copy_value_from_mask_reversed(input_mask_row);
+                return true;
+            }, input_mask);
+
+            output_mask->apply_callback(input_mask);
+            weights_mask->apply_callback(input_mask);
+
+            setMask(m_output, output_mask);
+            return true;
+        };
+
+        auto m = std::make_shared<ngraph::pattern::Matcher>(eltwise, "ElementwiseMaskPropagation");
+        register_matcher(m, callback);
+    }
+};
+
+class ngraph::pass::mask_propagation::FakeQuantize : public MatcherPass{
+public:
+    FakeQuantize(){
+        auto input = pattern::any_input(pattern::has_static_shape());
+        auto input_low = pattern::any_input(pattern::has_static_shape());
+        auto input_high = pattern::any_input(pattern::has_static_shape());
+        auto output_low = pattern::any_input(pattern::has_static_shape());
+        auto output_high = pattern::any_input(pattern::has_static_shape());
+        auto fake_quantize = pattern::wrap_type<opset6::FakeQuantize>({input, input_low, input_high, output_low,
+                                                                            output_high});
+        ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher& m) {
+            const auto & pattern_map = m.get_pattern_value_map();
+            const auto & m_input = pattern_map.at(input);
+            const auto & m_input_low = pattern_map.at(input_low);
+            const auto & m_input_high = pattern_map.at(input_high);
+            const auto & m_output_low = pattern_map.at(output_low);
+            const auto & m_output_high = pattern_map.at(output_high);
+            const auto & m_output = pattern_map.at(fake_quantize);
+
             auto input_mask = getMask(m_input);
 
-            if (!weights_mask || !input_mask) {
-                NGRAPH_DEBUG << "No mask for: " << m_output.get_node()->get_friendly_name() << std::endl;
+            // Input mask is the only source of pruning in FQ
+            if (!input_mask) {
+                NGRAPH_DEBUG << "FakeQuantize: No input mask for " << *m_output.get_node() << "\n";
                 return false;
             }
+
             auto input_mask_row = input_mask.get();
-            auto weights_mask_row = weights_mask.get();
 
-            // Merge masks from two inputs
+            // Propagate input mask to output mask and in the opposite direction
             auto output_mask = std::make_shared<Mask>(m_output.get_partial_shape().rank().get_length());
             auto output_mask_row = output_mask.get();
 
-            auto out_mask_callback = [input_mask_row, weights_mask_row](Mask::Ptr cur_mask) -> bool {
-                auto omask_iter = cur_mask->rbegin();
-                auto imask_iter = input_mask_row->rbegin();
-                auto wmask_iter = weights_mask_row->rbegin();
+            // Output mask is equal to input mask
+            auto output_mask_callback = [input_mask_row](Mask::Ptr cur_mask) -> bool {
+                cur_mask->copy_value_from_mask(input_mask_row);
+                return true;
+            };
 
-                for (auto & item : *cur_mask) {
-                    item.clear();
-                }
+            auto input_mask_callback = [output_mask_row](Mask::Ptr cur_mask) -> bool {
+                cur_mask->copy_value_from_mask(output_mask_row);
+                return true;
+            };
 
-                while (imask_iter != input_mask_row->rend() &&
-                       wmask_iter != weights_mask_row->rend()) {
-                    // Merge mask dimension values for both masks
-                    // Example: (MaskValue[1,2,3,4], MaskValue[2,3]) -> MaskValue[2,3]
-                    for (const auto & value : *imask_iter) {
-                        if (wmask_iter->count(value)) {
-                            omask_iter->insert(value);
-                        }
-                    }
+            output_mask->add_callback(output_mask_callback, input_mask);
+            input_mask->add_callback(input_mask_callback, output_mask);
 
-                    omask_iter++;
-                    imask_iter++;
-                    wmask_iter++;
+            // Calculate output mask
+            output_mask->apply_callback(input_mask);
+            setMask(m_output, output_mask);
+
+            auto input_low_size = shape_size(m_input_low.get_shape());
+            auto input_high_size = shape_size(m_input_high.get_shape());
+            auto output_low_size = shape_size(m_output_low.get_shape());
+            auto output_high_size = shape_size(m_output_high.get_shape());
+
+            // In the per-tensor case FQ params shouldn't be pruned
+            if (input_low_size == 1 && output_low_size == 1 && input_high_size == 1 && output_high_size == 1) {
+                return true;
+            }
+
+            // If input/output ranges in FQ should be broadcasted to input shape -> broadcast this consant values
+            // for the convenience of working with the masks
+            NodeVector fq_params_nodes{m_input_low.get_node_shared_ptr(),
+                                                               m_input_high.get_node_shared_ptr(),
+                                                               m_output_low.get_node_shared_ptr(),
+                                                               m_output_high.get_node_shared_ptr()};
+            auto fq_node = std::dynamic_pointer_cast<op::FakeQuantize>(m_output.get_node_shared_ptr());
+            size_t idx = 0;
+            if (fq_node->get_auto_broadcast() != ngraph::op::AutoBroadcastType::NONE) {
+                for (auto const_node : fq_params_nodes) {
+                    auto new_shape = broadcast_shape_to_rank(const_node->get_shape(),
+                                                             m_input.get_partial_shape().rank().get_length());
+                    auto const_copy = const_node->clone_with_new_inputs(const_node->input_values());
+                    auto new_const = std::dynamic_pointer_cast<op::Constant>(const_copy);
+                    new_const->set_data_shape(new_shape);
+                    new_const->validate_and_infer_types();
+                    new_const->set_friendly_name(const_node->get_friendly_name());
+                    ngraph::copy_runtime_info(const_node, new_const);
+                    ngraph::replace_node(const_node, new_const);
+                    fq_params_nodes[idx++] = new_const;
                 }
+            }
+
+            auto fq_params_mask_callback = [input_mask_row](Mask::Ptr cur_mask) -> bool {
+                cur_mask->at(1/* fq params have same shapes as input */) = input_mask_row->at(1 /* channel dim in data */);
                 return true;
             };
-            output_mask->add_callback(out_mask_callback, input_mask);
-            output_mask->add_callback(out_mask_callback, weights_mask);
-
-            auto callback = [output_mask_row](Mask::Ptr cur_mask) -> bool {
-                auto omask_iter = output_mask_row->rbegin();
-                auto cmask_iter = cur_mask->rbegin();
-                while (omask_iter != output_mask_row->rend() &&
-                       cmask_iter != cur_mask->rend()) {
-                    // TODO: check
-                    *cmask_iter = *omask_iter;
-
-                    omask_iter++;
-                    cmask_iter++;
+
+            for (auto fq_param : fq_params_nodes) {
+                auto mask = std::make_shared<Mask>(fq_param->get_shape().size());
+                mask->add_callback(fq_params_mask_callback, input_mask);
+                input_mask->add_callback([mask](Mask::Ptr cur_mask) -> bool {
+                    return true;
+                }, mask);
+                mask->apply_callback(input_mask);
+                setMask(fq_param->output(0), mask);
+            }
+
+            return true;
+        };
+
+        auto m = std::make_shared<ngraph::pattern::Matcher>(fake_quantize, "FakeQuantizeMaskPropagation");
+        register_matcher(m, callback);
+    }
+};
+
+class ngraph::pass::mask_propagation::Concat : public MatcherPass{
+public:
+    Concat() {
+        auto concat = pattern::wrap_type<opset6::Concat>(pattern::has_static_shape());
+
+        ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher &m) {
+            const auto & pattern_map = m.get_pattern_value_map();
+            const auto & m_output = pattern_map.at(concat);
+            auto concat_ptr = std::dynamic_pointer_cast<opset6::Concat>(m_output.get_node_shared_ptr());
+            auto axis = concat_ptr->get_concatenation_axis();
+
+            auto inputs = concat_ptr->inputs();
+            std::map<int64_t , Mask::Ptr> input_masks;
+            std::map<int64_t , Mask *> input_masks_row;
+            std::vector<int64_t> input_sizes;
+
+            size_t first_input_idx = 0;
+            Mask::Ptr first_input_mask;
+            bool first_initialized = false;
+            for (size_t i=0; i < inputs.size(); i++) {
+                auto input = inputs[i];
+                auto input_mask = getMask(input.get_source_output());
+                if (input_mask) {
+                    input_masks[i] = input_mask;
+                    input_masks_row[i] = input_mask.get();
+
+                    if (!first_initialized) {
+                        first_input_idx = i;
+                        first_input_mask = input_mask;
+                        first_initialized = true;
+                    }
+                }
+                input_sizes.push_back(input.get_shape().at(axis));
+            }
+
+            if (!first_initialized) {
+                return false;
+            }
+
+            auto output_mask = std::make_shared<Mask>(m_output.get_partial_shape().rank().get_length());
+            auto output_mask_row = output_mask.get();
+
+            auto out_mask_callback = [input_masks_row, input_sizes, axis](Mask::Ptr cur_mask) -> bool {
+                int64_t cur_size = 0;
+                cur_mask->at(axis).clear();
+
+                for (size_t i=0; i < input_sizes.size(); ++i) {
+                    if (input_masks_row.count(i)) {
+                        for (auto idx : input_masks_row.at(i)->at(axis)) {
+                            cur_mask->at(axis).insert(idx + cur_size);
+                        }
+                    }
+                    cur_size += input_sizes[i];
                 }
                 return true;
             };
-            input_mask->add_callback(callback, output_mask);
-            weights_mask->add_callback(callback, output_mask);
 
-            // Init output mask
-            output_mask->apply_callback(input_mask);
+            auto create_input_mask_callback_for_idx = [output_mask_row, input_sizes, axis](size_t input_idx){
+                auto input_mask_callback = [output_mask_row, input_sizes, axis, input_idx](Mask::Ptr cur_mask) -> bool {
+                    cur_mask->clean_dim_values();
+                    uint64_t min_val = 0;
+                    for (size_t i = 0; i < input_idx; i++) {
+                        min_val += input_sizes[i];
+                    }
+                    uint64_t max_val = min_val + input_sizes[input_idx];
+                    for (auto idx : output_mask_row->at(axis)) {
+                        if (idx < max_val && idx >= min_val) {
+                            cur_mask->at(axis).insert(idx - min_val);
+                        }
+                    }
+                    return true;
+                };
+                return input_mask_callback;
+            };
+            output_mask->add_callback(out_mask_callback, first_input_mask);
+
+            for (size_t i=0; i < inputs.size(); ++i) {
+                if (input_masks.count(i) && i != first_input_idx) {
+                    auto input_mask = input_masks.at(i);
+                    input_mask->add_callback(create_input_mask_callback_for_idx(i),
+                                             first_input_mask);
+                    first_input_mask->add_callback([](Mask::Ptr cur_mask) -> bool {
+                        return true;
+                    }, input_mask);
+                }
+            }
+            first_input_mask->add_callback(create_input_mask_callback_for_idx(first_input_idx),
+                                     output_mask);
+            output_mask->apply_callback(first_input_mask);
             setMask(m_output, output_mask);
+
             return true;
         };
-
-        auto m = std::make_shared<ngraph::pattern::Matcher>(eltwise, "EltwiseMaskPropagation");
+        auto m = std::make_shared<ngraph::pattern::Matcher>(concat, "ConcatMaskPropagation");
         register_matcher(m, callback);
     }
 };
@@ -266,7 +554,9 @@ class ngraph::pass::mask_propagation::Elementwise : public MatcherPass {
 class ngraph::pass::mask_propagation::PassThrough : public MatcherPass {
 public:
     PassThrough() {
-        auto unary_op = pattern::wrap_type<op::util::UnaryElementwiseArithmetic, opset6::Clamp>();
+        auto unary_op = pattern::wrap_type<op::util::UnaryElementwiseArithmetic, opset6::Clamp,
+                                            opset6::Convert, opset6::ConvertLike, opset6::AvgPool, opset6::MaxPool,
+                                            opset6::ROIPooling, opset6::PSROIPooling, opset6::Pad>();
 
         ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher& m) {
             const auto & pattern_map = m.get_pattern_value_map();
@@ -312,5 +602,8 @@ ngraph::pass::PropagateMasks::PropagateMasks() {
     add_matcher<mask_propagation::GroupConvolution>();
     add_matcher<mask_propagation::Elementwise>();
     add_matcher<mask_propagation::PassThrough>();
+    add_matcher<mask_propagation::FakeQuantize>();
+    add_matcher<mask_propagation::Concat>();
+    add_matcher<mask_propagation::Reshape>();
     add_matcher<mask_propagation::StopPropagation>();
 }
diff --git a/inference-engine/src/offline_transformations/src/pruning/pruning.cpp b/inference-engine/src/offline_transformations/src/pruning/pruning.cpp
index 3159e3db7db166..ad7f410b8f356b 100644
--- a/inference-engine/src/offline_transformations/src/pruning/pruning.cpp
+++ b/inference-engine/src/offline_transformations/src/pruning/pruning.cpp
@@ -15,8 +15,13 @@ NGRAPH_RTTI_DEFINITION(ngraph::pass::Pruning, "Pruning", 0);
 
 bool ngraph::pass::Pruning::run_on_function(std::shared_ptr<Function> f) {
     Manager manager(get_pass_config());
+
+    // Initialize masks only for Convolutions/GroupConvolutions weights (needed to init mask in source Constant of
+    // weights-calculating subgraph). For other node types masks initialized in PropagateMasks pass.
+    manager.register_pass<InitMasks>();
     manager.register_pass<PropagateMasks>();
 
+
 #ifdef NGRAPH_DEBUG_ENABLE
     // VisualizeTree modifier helps to print Masks and mark nodes with masks
     /*
diff --git a/inference-engine/src/offline_transformations/src/pruning/shrink_weights.cpp b/inference-engine/src/offline_transformations/src/pruning/shrink_weights.cpp
index 80c2abbb709eb8..08e7ef152c03cc 100644
--- a/inference-engine/src/offline_transformations/src/pruning/shrink_weights.cpp
+++ b/inference-engine/src/offline_transformations/src/pruning/shrink_weights.cpp
@@ -54,6 +54,8 @@ bool ngraph::pass::ShrinkWeights::run_on_function(std::shared_ptr<ngraph::Functi
             for (size_t dim = 0; dim < mask->size(); ++dim) {
                 const auto &dim_size = mask->at(dim).size();
                 if (dim_size == 0) continue;
+                // Broadcastable 1-size dimension shouldn't be shrank with mask
+                if (const_node->get_shape().at(dim) == 1 && dim_size > 1) continue;
 
                 // Convert dims that we want remove to dims that we need to keep
                 std::vector<int64_t> dims_to_keep;
diff --git a/inference-engine/tests/functional/inference_engine/transformations/pruning_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/pruning_test.cpp
index 0f46a853cef123..82c1fa6c9f2eab 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/pruning_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/pruning_test.cpp
@@ -15,6 +15,7 @@
 #include <transformations/init_node_info.hpp>
 #include <ngraph/coordinate_transform.hpp>
 #include <ngraph/pass/manager.hpp>
+#include <inference_engine.hpp>
 
 using namespace testing;
 using namespace ngraph;
@@ -67,6 +68,23 @@ TEST(TransformationTests, InitMasksOutputChannel) {
     compare_masks(*getMask(weights->output(0)), {{}, {1}, {}, {}});
 }
 
+// TODO: add test init masks with subgraph
+TEST(TransformationTests, TestInitMasks) {
+    Shape weights_shape{6, 3, 3, 3};
+    Shape input_shape{1, 3, 64, 64};
+    auto input = std::make_shared<opset5::Parameter>(element::f32, input_shape);
+    auto weights = create_constant_with_zeros(weights_shape, {{1, 2, 3}, {}, {}, {}});
+    auto conv = std::make_shared<opset5::Convolution>(input, weights, Strides(2, 1),
+                                                      CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto f = std::make_shared<Function>(NodeVector{conv}, ParameterVector{input});
+    pass::Manager m;
+    m.register_pass<pass::InitMasks>();
+    m.run_passes(f);
+
+    compare_masks(*getMask(weights.get_node_shared_ptr()->output(0)), {{1, 2, 3}, {}, {}, {}});
+}
+
 TEST(TransformationTests, InitMasksNegative) {
     Shape weights_shape{6, 3, 3, 3};
     auto weights = opset5::Constant::create(element::f32, weights_shape, {0.5});
@@ -85,6 +103,7 @@ TEST(TransformationTests, PropagateMasksNegative) {
     auto f = std::make_shared<Function>(NodeVector{conv}, ParameterVector{input});
 
     pass::Manager m;
+    m.register_pass<pass::InitMasks>();
     m.register_pass<pass::PropagateMasks>();
     m.run_passes(f);
 
@@ -102,27 +121,35 @@ TEST(TransformationTests, PropagateMasksBasic) {
                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
     auto relu = std::make_shared<opset5::Relu>(conv);
 
+    auto add_const = create_constant_with_zeros(Shape{1, 6, 1, 1}, {{}, {1, 2, 3, 4, 5}, {}, {}});
+    auto add = std::make_shared<opset5::Add>(relu, add_const);
+
     auto sub_const = create_constant_with_zeros(Shape{6, 1, 1}, {{1, 2, 3}, {}, {}});
-    auto sub = std::make_shared<opset5::Subtract>(relu, sub_const);
+    auto sub = std::make_shared<opset5::Subtract>(add, sub_const);
 
-    auto mul_const = create_constant_with_zeros(Shape{6, 1, 1}, {{2}, {}, {}});
-    auto mul = std::make_shared<opset5::Subtract>(sub, mul_const);
+    auto mul_const = create_constant_with_zeros(Shape{1, 6, 1, 1}, {{}, {4}, {}, {}});
+    auto mul = std::make_shared<opset5::Multiply>(sub, mul_const);
 
-    auto weights2 = opset5::Constant::create(element::f32, weights_shape2, {0});
+    auto weights2 = create_constant_with_zeros(weights_shape2, {{1, 2}, {1, 2, 3}, {}, {}});
     auto conv2 = std::make_shared<opset5::Convolution>(mul, weights2, Strides(2, 1),
                                                        CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
     auto f = std::make_shared<Function>(NodeVector{conv2}, ParameterVector{input});
 
     pass::Manager m;
+    m.register_pass<pass::InitMasks>();
     m.register_pass<pass::PropagateMasks>();
     m.run_passes(f);
 
-    compare_masks(*getMask(weights->output(0)),  Mask({{2}, {}, {}, {}}));
-    compare_masks(*getMask(conv->output(0)),     Mask({{}, {2}, {}, {}}));
-    compare_masks(*getMask(relu->output(0)),     Mask({{}, {2}, {}, {}}));
-    compare_masks(*getMask(sub_const), Mask({{2}, {}, {}}));
-    compare_masks(*getMask(mul_const), Mask({{2}, {}, {}}));
-    compare_masks(*getMask(weights2->output(0)), Mask({{}, {2}, {}, {}}));
+    compare_masks(*getMask(weights->output(0)),  Mask({{1, 2, 3, 4}, {}, {}, {}}));
+    compare_masks(*getMask(conv->output(0)),     Mask({{}, {1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(relu->output(0)),     Mask({{}, {1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(add_const), Mask({{}, {1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(sub_const), Mask({{1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(mul_const), Mask({{}, {1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(add->output(0)), Mask({{}, {1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(sub->output(0)), Mask({{}, {1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(mul->output(0)), Mask({{}, {1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(weights2.get_node_shared_ptr()->output(0)), Mask({{}, {1, 2, 3, 4}, {}, {}}));
     compare_masks(*getMask(conv2->output(0)),    Mask({{}, {}, {}, {}}));
 }
 
@@ -148,6 +175,7 @@ TEST(TransformationTests, PropagateMasksDynamicConvolution) {
     auto f = std::make_shared<Function>(NodeVector{conv2}, ParameterVector{input});
 
     pass::Manager m;
+    m.register_pass<pass::InitMasks>();
     m.register_pass<pass::PropagateMasks>();
     m.run_passes(f);
 
@@ -182,6 +210,7 @@ TEST(TransformationTests, PropagateMasksDynamicGroupConvolution) {
     auto f = std::make_shared<Function>(NodeVector{conv2}, ParameterVector{input});
 
     pass::Manager m;
+    m.register_pass<pass::InitMasks>();
     m.register_pass<pass::PropagateMasks>();
     m.run_passes(f);
 }
@@ -199,15 +228,16 @@ TEST(TransformationTests, PropagateMasksEmpty) {
     auto sub_const = create_constant_with_zeros(Shape{6, 1, 1}, {{1, 2, 3}, {}, {}});
     auto sub = std::make_shared<opset5::Subtract>(relu, sub_const);
 
-    auto mul_const = create_constant_with_zeros(Shape{6, 1, 1}, {{1, 2}, {}, {}});
-    auto mul = std::make_shared<opset5::Subtract>(sub, mul_const);
+    auto add_const = create_constant_with_zeros(Shape{6, 1, 1}, {{1, 2}, {}, {}});
+    auto add = std::make_shared<opset5::Subtract>(sub, add_const);
 
     auto weights2 = opset5::Constant::create(element::f32, weights_shape2, {0});
-    auto conv2 = std::make_shared<opset5::Convolution>(mul, weights2, Strides(2, 1),
+    auto conv2 = std::make_shared<opset5::Convolution>(add, weights2, Strides(2, 1),
                                                        CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
     auto f = std::make_shared<Function>(NodeVector{conv2}, ParameterVector{input});
 
     pass::Manager m;
+    m.register_pass<pass::InitMasks>();
     m.register_pass<pass::PropagateMasks>();
     m.run_passes(f);
 
@@ -215,11 +245,55 @@ TEST(TransformationTests, PropagateMasksEmpty) {
     compare_masks(*getMask(conv->output(0)),     Mask({{}, {}, {}, {}}));
     compare_masks(*getMask(relu->output(0)),     Mask({{}, {}, {}, {}}));
     compare_masks(*getMask(sub_const), Mask({{}, {}, {}}));
-    compare_masks(*getMask(mul_const), Mask({{}, {}, {}}));
+    compare_masks(*getMask(add_const), Mask({{}, {}, {}}));
     compare_masks(*getMask(weights2->output(0)), Mask({{}, {}, {}, {}}));
     compare_masks(*getMask(conv2->output(0)),    Mask({{}, {}, {}, {}}));
 }
 
+TEST(TransformationTests, PropagateMaskPassThrough) {
+    Shape input_shape{1, 3, 64, 64};
+    Shape weights_shape{8, 3, 3, 3};
+    Shape weight_shape2{3, 8, 3, 3};
+    auto input = std::make_shared<opset5::Parameter>(element::f32, input_shape);
+    input->set_friendly_name("input");
+    auto weights_const_1 = create_constant_with_zeros(weights_shape, {{1, 2, 3}, {}, {}, {}});
+    weights_const_1.get_node_shared_ptr()->set_friendly_name("weights_1");
+
+    auto conv_1 = std::make_shared<opset5::Convolution>(input, weights_const_1, Strides(2, 1),
+                                                      CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+    conv_1->set_friendly_name("conv_1");
+
+    // Adding a couple of PassThrough operations
+    auto relu = std::make_shared<opset5::Relu>(conv_1);
+    relu->set_friendly_name("relu");
+
+    auto clamp = std::make_shared<opset5::Clamp>(relu, 0, 6);
+    clamp->set_friendly_name("clamp");
+
+    auto pads_begin = opset5::Constant::create(element::i32, Shape{4}, {0, 0, 1, 1});
+    auto pads_end = opset5::Constant::create(element::i32, Shape{4}, {0, 0, 2, 2});
+    auto pad = std::make_shared<opset5::Pad>(clamp, pads_begin, pads_end, op::PadMode::CONSTANT);
+    auto max_pool = std::make_shared<opset5::MaxPool>(pad, Strides{1, 1},
+                                                      Shape{0, 0}, Shape{1, 1}, Shape{4, 4});
+    max_pool->set_friendly_name("max_pool");
+
+    auto weights2 = opset5::Constant::create(element::f32, weight_shape2, {0});
+    auto conv2 = std::make_shared<opset5::Convolution>(max_pool, weights2, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+    auto f = std::make_shared<Function>(NodeVector{conv2}, ParameterVector{input});
+
+    pass::Manager m;
+    m.register_pass<pass::InitMasks>();
+    m.register_pass<pass::PropagateMasks>();
+    m.run_passes(f);
+
+    compare_masks(*getMask(weights_const_1.get_node_shared_ptr()->output(0)),  Mask({{1, 2, 3}, {}, {}, {}}));
+    compare_masks(*getMask(conv_1->output(0)),     Mask({{}, {1, 2, 3}, {}, {}}));
+    compare_masks(*getMask(relu->output(0)),     Mask({{}, {1, 2, 3}, {}, {}}));
+    compare_masks(*getMask(clamp->output(0)),     Mask({{}, {1, 2, 3}, {}, {}}));
+    compare_masks(*getMask(max_pool->output(0)),     Mask({{}, {1, 2, 3}, {}, {}}));
+}
+
 TEST(TransformationTests, PropagateMasksHardDependencies) {
     Shape input_shape{1, 3, 3, 3};
 
@@ -280,4 +354,344 @@ TEST(TransformationTests, PropagateMasksHardDependencies) {
 //    compare_masks(*getMask(relu),     Mask({{}, {0, 1, 2, 3, 4, 5}, {}, {}}));
 //    compare_masks(*getMask(weights2), Mask({{}, {0, 1, 2, 3, 4, 5}, {}, {}}));
 //    compare_masks(*getMask(conv2),    Mask({{}, {}, {}, {}}));
-}
\ No newline at end of file
+}
+
+TEST(TransformationTests, PropagateMasksQuantizedGroupConvolution) {
+    Shape input_shape{1, 3, 64, 64};
+    Shape weights_shape{8, 3, 3, 3};
+    Shape weights_group_shape{8, 1, 3, 3};
+    Shape weight_shape2{3, 8, 3, 3};
+    auto input = std::make_shared<opset5::Parameter>(element::f32, input_shape);
+    input->set_friendly_name("input");
+
+    auto weights1 = create_constant_with_zeros(weights_shape, {{0, 1, 2, 3}, {}, {}, {}});
+    auto conv1 = std::make_shared<opset5::Convolution>(input, weights1, Strides(2, 1),
+                                                      CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+    auto weights_group = opset5::Constant::create(element::i8, weights_group_shape, {0});
+    weights_group->set_friendly_name("weights_group");
+
+    auto convert = std::make_shared<opset5::Convert>(weights_group, element::f32);
+    convert->set_friendly_name("convert");
+
+    auto sub_const = create_constant_with_zeros(Shape{8, 1, 1, 1}, {{0, 1, 2, 3}, {}, {}, {}});
+
+    auto sub = std::make_shared<opset5::Subtract>(convert, sub_const);
+    sub->set_friendly_name("sub");
+
+    auto mul_const = create_constant_with_zeros(Shape{8, 1, 1, 1}, {{0, 1, 2, 3, 4}, {}, {}, {}});
+    auto mul = std::make_shared<opset5::Multiply>(sub, mul_const);
+    mul->set_friendly_name("mul");
+
+    auto reshape = std::make_shared<opset5::Reshape>(mul, opset5::Constant::create(element::i64, Shape{5}, {8, 1, 1, 3, 3}), false);
+
+    auto conv_group = std::make_shared<opset5::GroupConvolution>(conv1, reshape, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto add_const = create_constant_with_zeros(Shape{1, 8, 1, 1}, {{}, {0, 1, 2, 3, 4}, {}, {}});;
+    auto add = std::make_shared<opset5::Add>(conv_group, add_const);
+    add->set_friendly_name("add");
+
+    auto weights_2 = opset5::Constant::create(element::f32, weight_shape2, {0});
+    auto conv2 = std::make_shared<opset5::Convolution>(add, weights_2, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+    auto f = std::make_shared<Function>(NodeVector{conv2}, ParameterVector{input});
+
+    pass::Manager m;
+    m.register_pass<pass::Pruning>();
+    m.run_passes(f);
+
+    compare_masks(*getMask(weights1.get_node_shared_ptr()->output(0)), Mask({{0 , 1, 2, 3}, {}, {}, {}}));
+    compare_masks(*getMask(conv1->output(0)),  Mask({{}, {0 , 1, 2, 3}, {}, {}}));
+
+    compare_masks(*getMask(weights_group->output(0)), Mask({{0 , 1, 2, 3}, {}, {}, {}}));
+    compare_masks(*getMask(sub->output(0)), Mask({{0 , 1, 2, 3}, {}, {}, {}}));
+    compare_masks(*getMask(sub_const.get_node_shared_ptr()->output(0)), Mask({{0 , 1, 2, 3}, {}, {}, {}}));
+    compare_masks(*getMask(mul->output(0)), Mask({{0 , 1, 2, 3}, {}, {}, {}}));
+    compare_masks(*getMask(mul_const.get_node_shared_ptr()->output(0)), Mask({{0 , 1, 2, 3}, {}, {}, {}}));
+
+    compare_masks(*getMask(reshape->output(0)), Mask({{0 , 1, 2, 3}, {}, {}, {}, {}}));
+
+    compare_masks(*getMask(conv_group->output(0)),  Mask({{}, {0 , 1, 2, 3}, {}, {}}));
+
+    compare_masks(*getMask(conv2->output(0)),  Mask({{}, {}, {}, {}}));
+    compare_masks(*getMask(weights_2->output(0)),  Mask({{}, {0, 1, 2, 3}, {}, {}}));
+}
+
+TEST(TransformationTests, PropagateMasksFakeQuantizePerTensor) {
+    Shape input_shape{1, 3, 64, 64};
+    Shape weights_shape{8, 3, 3, 3};
+    Shape weight_shape2{3, 8, 3, 3};
+    auto input = std::make_shared<opset5::Parameter>(element::f32, input_shape);
+    input->set_friendly_name("input");
+    auto weights_1 = opset5::Constant::create(element::i8, weights_shape, {0});
+    weights_1->set_friendly_name("weights_int8_const");
+
+    auto convert = std::make_shared<opset5::Convert>(weights_1, element::f32);
+    convert->set_friendly_name("convert");
+
+    auto sub_const = create_constant_with_zeros(Shape{8, 1, 1, 1}, {{0, 1, 2, 3}, {}, {}, {}});
+
+    auto sub = std::make_shared<opset5::Subtract>(convert, sub_const);
+    sub->set_friendly_name("sub");
+
+    auto mul_const = create_constant_with_zeros(Shape{8, 1, 1, 1}, {{0, 1, 2, 3, 4}, {}, {}, {}});
+    auto mul = std::make_shared<opset5::Multiply>(sub, mul_const);
+    mul->set_friendly_name("mul");
+
+    auto conv1 = std::make_shared<opset5::Convolution>(input, mul, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+    conv1->set_friendly_name("conv1");
+
+    auto add_const = create_constant_with_zeros(Shape{1, 8, 1, 1}, {{}, {0, 1, 2, 3, 4}, {}, {}});;
+    auto add = std::make_shared<opset5::Add>(conv1, add_const);
+    add->set_friendly_name("add");
+
+    auto input_low = opset5::Constant::create(element::f32, Shape{1}, {0});
+    auto input_high = opset5::Constant::create(element::f32, Shape{1, 1, 1, 1}, {20});
+    auto output_low = opset5::Constant::create(element::f32, Shape{}, {1});
+    auto output_high = opset5::Constant::create(element::f32, Shape{}, {10});
+    auto fq = std::make_shared<opset5::FakeQuantize>(add, input_low, input_high, output_low, output_high, 8);
+
+    auto weights_2 = opset5::Constant::create(element::f32, weight_shape2, {0});
+    auto conv2 = std::make_shared<opset5::Convolution>(fq, weights_2, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+    auto f = std::make_shared<Function>(NodeVector{conv2}, ParameterVector{input});
+
+    pass::Manager m;
+    m.register_pass<pass::Pruning>();
+    m.run_passes(f);
+
+    compare_masks(*getMask(weights_1->output(0)), Mask({{0 , 1, 2, 3, 4}, {}, {}, {}}));
+    compare_masks(*getMask(sub_const.get_node_shared_ptr()->output(0)),  Mask({{0 , 1, 2, 3, 4}, {}, {}, {}}));
+    compare_masks(*getMask(sub->output(0)),  Mask({{0 , 1, 2, 3, 4}, {}, {}, {}}));
+
+    compare_masks(*getMask(mul_const.get_node_shared_ptr()->output(0)),  Mask({{0 , 1, 2, 3, 4}, {}, {}, {}}));
+    compare_masks(*getMask(mul->output(0)),  Mask({{0 , 1, 2, 3, 4}, {}, {}, {}}));
+
+    compare_masks(*getMask(conv1->output(0)),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+
+    compare_masks(*getMask(add_const.get_node_shared_ptr()->output(0)),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(add->output(0)),  Mask({{}, {0 , 1, 2, 3, 4},  {}, {}}));
+
+    compare_masks(*getMask(fq->output(0)),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+
+    compare_masks(*getMask(weights_2->output(0)),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(conv2->output(0)),  Mask({{}, {}, {}, {}}));
+}
+
+TEST(TransformationTests, PropagateMasksFakeQuantizePerChannel) {
+    Shape input_shape{1, 3, 64, 64};
+    Shape weights_shape{8, 3, 3, 3};
+    Shape weight_shape2{3, 8, 3, 3};
+    auto input = std::make_shared<opset5::Parameter>(element::f32, input_shape);
+    input->set_friendly_name("input");
+    auto weights_1 = opset5::Constant::create(element::i8, weights_shape, {0});
+    weights_1->set_friendly_name("weights_int8_const");
+
+    auto convert = std::make_shared<opset5::Convert>(weights_1, element::f32);
+    convert->set_friendly_name("convert");
+
+    auto sub_const = create_constant_with_zeros(Shape{8, 1, 1, 1}, {{0, 1, 2, 3}, {}, {}, {}});
+
+    auto sub = std::make_shared<opset5::Subtract>(convert, sub_const);
+    sub->set_friendly_name("sub");
+
+    auto mul_const = create_constant_with_zeros(Shape{8, 1, 1, 1}, {{0, 1, 2, 3, 4}, {}, {}, {}});
+    auto mul = std::make_shared<opset5::Multiply>(sub, mul_const);
+    mul->set_friendly_name("mul");
+
+    auto conv1 = std::make_shared<opset5::Convolution>(input, mul, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+    conv1->set_friendly_name("conv1");
+
+    auto add_const = create_constant_with_zeros(Shape{1, 8, 1, 1}, {{}, {0, 1, 2, 3, 4}, {}, {}});;
+    auto add = std::make_shared<opset5::Add>(conv1, add_const);
+    add->set_friendly_name("add");
+
+    auto input_low = opset5::Constant::create(element::f32, Shape{1, 8, 1, 1}, {0});
+    auto input_high = opset5::Constant::create(element::f32, Shape{1, 8, 1, 1}, {20});
+    auto output_low = opset5::Constant::create(element::f32, Shape{8, 1, 1}, {1});
+    auto output_high = opset5::Constant::create(element::f32, Shape{8, 1, 1}, {10});
+    auto fq = std::make_shared<opset5::FakeQuantize>(add, input_low, input_high, output_low, output_high, 8);
+
+    auto weights_2 = opset5::Constant::create(element::f32, weight_shape2, {0});
+    auto conv2 = std::make_shared<opset5::Convolution>(fq, weights_2, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+    auto f = std::make_shared<Function>(NodeVector{conv2}, ParameterVector{input});
+
+    pass::Manager m;
+    m.register_pass<pass::InitMasks>();
+    m.register_pass<pass::PropagateMasks>();
+    m.run_passes(f);
+
+    compare_masks(*getMask(weights_1->output(0)), Mask({{0 , 1, 2, 3, 4}, {}, {}, {}}));
+    compare_masks(*getMask(sub_const.get_node_shared_ptr()->output(0)),  Mask({{0 , 1, 2, 3, 4}, {}, {}, {}}));
+    compare_masks(*getMask(sub->output(0)),  Mask({{0 , 1, 2, 3, 4}, {}, {}, {}}));
+
+    compare_masks(*getMask(mul_const.get_node_shared_ptr()->output(0)),  Mask({{0 , 1, 2, 3, 4}, {}, {}, {}}));
+    compare_masks(*getMask(mul->output(0)),  Mask({{0 , 1, 2, 3, 4}, {}, {}, {}}));
+
+    compare_masks(*getMask(conv1->output(0)),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+
+    compare_masks(*getMask(add_const.get_node_shared_ptr()->output(0)),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(add->output(0)),  Mask({{}, {0 , 1, 2, 3, 4},  {}, {}}));
+
+    compare_masks(*getMask(fq->output(0)),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+
+    compare_masks(*getMask(weights_2->output(0)),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(conv2->output(0)),  Mask({{}, {}, {}, {}}));
+
+    compare_masks(*getMask(fq->input(1).get_source_output()),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(fq->input(2).get_source_output()),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(fq->input(3).get_source_output()),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+    compare_masks(*getMask(fq->input(4).get_source_output()),  Mask({{}, {0 , 1, 2, 3, 4}, {}, {}}));
+}
+
+TEST(TransformationTests, TestConcatMaskPropagation) {
+    Shape input_shape{1, 3, 64, 64};
+    Shape weights_shape1{8, 3, 3, 3};
+    Shape weights_shape2{16, 3, 3, 3};
+    Shape weights_shape3{8, 3, 3, 3};
+
+    Shape weight_shape_out_conv{3, 32, 3, 3};
+    auto input = std::make_shared<opset5::Parameter>(element::f32, input_shape);
+    auto weights_1 = create_constant_with_zeros(weights_shape1, {{0, 1, 2, 3}, {}, {}, {}});
+    auto conv1 = std::make_shared<opset5::Convolution>(input, weights_1, Strides(2, 1),
+                                                      CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto weights_2 = create_constant_with_zeros(weights_shape2, {{7, 8, 9, 10}, {}, {}, {}});
+    auto conv2 = std::make_shared<opset5::Convolution>(input, weights_2, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto weights_3 = create_constant_with_zeros(weights_shape3, {{4, 5, 6, 7}, {}, {}, {}});
+    auto conv3 = std::make_shared<opset5::Convolution>(input, weights_3, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto concat = std::make_shared<opset5::Concat>(OutputVector{conv1->output(0), conv2->output(0), conv3->output(0)}, 1);
+
+    auto weights_out_conv = create_constant_with_zeros(weight_shape_out_conv, {{}, {}, {}, {}});
+    auto conv_out = std::make_shared<opset5::Convolution>(concat, weights_out_conv, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto f = std::make_shared<Function>(NodeVector{conv_out}, ParameterVector{input});
+
+    pass::Manager m;
+    m.register_pass<pass::InitMasks>();
+    m.register_pass<pass::PropagateMasks>();
+    m.run_passes(f);
+
+    compare_masks(*getMask(weights_1.get_node_shared_ptr()->output(0)),  Mask({{0, 1, 2, 3}, {}, {}, {}}));
+    compare_masks(*getMask(conv1->output(0)),  Mask({{}, {0, 1, 2, 3}, {}, {}}));
+
+    compare_masks(*getMask(weights_2.get_node_shared_ptr()->output(0)),  Mask({{7, 8, 9, 10}, {}, {}, {}}));
+    compare_masks(*getMask(conv2->output(0)),  Mask({{}, {7, 8, 9, 10}, {}, {}}));
+
+    compare_masks(*getMask(weights_3.get_node_shared_ptr()->output(0)),  Mask({{4, 5, 6, 7}, {}, {}, {}}));
+    compare_masks(*getMask(conv3->output(0)),  Mask({{}, {4, 5, 6, 7}, {}, {}}));
+
+    compare_masks(*getMask(concat->output(0)),  Mask({{}, {0, 1, 2, 3, 15, 16, 17, 18, 28, 29, 30, 31}, {}, {}}));
+    compare_masks(*getMask(weights_out_conv.get_node_shared_ptr()->output(0)),  Mask({{}, {0, 1, 2, 3, 15, 16, 17, 18, 28, 29, 30, 31}, {}, {}}));
+}
+
+
+TEST(TransformationTests, TestConcatMaskPropagationUp) {
+    Shape input_shape{1, 3, 64, 64};
+    Shape weights_shape1{8, 3, 3, 3};
+    Shape weights_shape2{16, 3, 3, 3};
+    Shape weights_shape3{8, 3, 3, 3};
+
+    Shape weight_shape_out_conv{3, 32, 3, 3};
+    auto input = std::make_shared<opset5::Parameter>(element::f32, input_shape);
+    auto weights_1 = create_constant_with_zeros(weights_shape1, {{0, 1, 2, 3, 4, 5}, {}, {}, {}});
+    auto conv1 = std::make_shared<opset5::Convolution>(input, weights_1, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto weights_2 = create_constant_with_zeros(weights_shape2, {{7, 8, 9, 10}, {}, {}, {}});
+    auto conv2 = std::make_shared<opset5::Convolution>(input, weights_2, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto weights_3 = create_constant_with_zeros(weights_shape3, {{2, 3, 4, 5, 6, 7}, {}, {}, {}});
+    auto conv3 = std::make_shared<opset5::Convolution>(input, weights_3, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto concat = std::make_shared<opset5::Concat>(OutputVector{conv1->output(0), conv2->output(0), conv3->output(0)}, 1);
+
+    auto add_const = create_constant_with_zeros(Shape{1, 32, 1, 1}, {{}, {0, 1, 2, 3, 15, 16, 17, 18, 28, 29, 30, 31}, {}, {}});
+    auto add = std::make_shared<opset5::Add>(concat, add_const);
+
+    auto weights_out_conv = create_constant_with_zeros(weight_shape_out_conv, {{}, {}, {}, {}});
+    auto conv_out = std::make_shared<opset5::Convolution>(add, weights_out_conv, Strides(2, 1),
+                                                          CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto f = std::make_shared<Function>(NodeVector{conv_out}, ParameterVector{input});
+
+    pass::Manager m;
+    m.register_pass<pass::InitMasks>();
+    m.register_pass<pass::PropagateMasks>();
+    m.run_passes(f);
+
+    compare_masks(*getMask(weights_1.get_node_shared_ptr()->output(0)),  Mask({{0, 1, 2, 3}, {}, {}, {}}));
+    compare_masks(*getMask(conv1->output(0)),  Mask({{}, {0, 1, 2, 3}, {}, {}}));
+
+    compare_masks(*getMask(weights_2.get_node_shared_ptr()->output(0)),  Mask({{7, 8, 9, 10}, {}, {}, {}}));
+    compare_masks(*getMask(conv2->output(0)),  Mask({{}, {7, 8, 9, 10}, {}, {}}));
+
+    compare_masks(*getMask(weights_3.get_node_shared_ptr()->output(0)),  Mask({{4, 5, 6, 7}, {}, {}, {}}));
+    compare_masks(*getMask(conv3->output(0)),  Mask({{}, {4, 5, 6, 7}, {}, {}}));
+
+    compare_masks(*getMask(add_const.get_node_shared_ptr()->output(0)),  Mask({{}, {0, 1, 2, 3, 15, 16, 17, 18, 28, 29, 30, 31}, {}, {}}));
+    compare_masks(*getMask(add->output(0)),  Mask({{}, {0, 1, 2, 3, 15, 16, 17, 18, 28, 29, 30, 31}, {}, {}}));
+
+
+    compare_masks(*getMask(concat->output(0)),  Mask({{}, {0, 1, 2, 3, 15, 16, 17, 18, 28, 29, 30, 31}, {}, {}}));
+    compare_masks(*getMask(weights_out_conv.get_node_shared_ptr()->output(0)),  Mask({{}, {0, 1, 2, 3, 15, 16, 17, 18, 28, 29, 30, 31}, {}, {}}));
+}
+
+
+TEST(TransformationTests, TestConcatMaskPropagationUpEmpty) {
+    Shape input_shape{1, 3, 64, 64};
+    Shape weights_shape1{8, 3, 3, 3};
+    Shape weights_shape2{16, 3, 3, 3};
+    Shape weights_shape3{8, 3, 3, 3};
+
+    Shape weight_shape_out_conv{3, 32, 3, 3};
+    auto input = std::make_shared<opset5::Parameter>(element::f32, input_shape);
+    auto weights_1 = create_constant_with_zeros(weights_shape1, {{0, 1, 2, 3, 4, 5}, {}, {}, {}});
+    auto conv1 = std::make_shared<opset5::Convolution>(input, weights_1, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto weights_2 = create_constant_with_zeros(weights_shape2, {{7, 8, 9, 10}, {}, {}, {}});
+    auto conv2 = std::make_shared<opset5::Convolution>(input, weights_2, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto weights_3 = create_constant_with_zeros(weights_shape3, {{2, 3, 4, 5, 6, 7}, {}, {}, {}});
+    auto conv3 = std::make_shared<opset5::Convolution>(input, weights_3, Strides(2, 1),
+                                                       CoordinateDiff(2, 0), CoordinateDiff(2, 0), Strides(2, 1));
+
+    auto concat = std::make_shared<opset5::Concat>(OutputVector{conv1->output(0), conv2->output(0), conv3->output(0)}, 1);
+
+    auto add_const = create_constant_with_zeros(Shape{1, 32, 1, 1}, {{}, {0, 1, 2, 3, 15, 16, 17, 18, 28, 29, 30, 31}, {}, {}});
+    auto add = std::make_shared<opset5::Add>(concat, add_const);
+
+    auto f = std::make_shared<Function>(NodeVector{add}, ParameterVector{input});
+
+    pass::Manager m;
+    m.register_pass<pass::InitMasks>();
+    m.register_pass<pass::PropagateMasks>();
+    m.run_passes(f);
+
+    compare_masks(*getMask(weights_1.get_node_shared_ptr()->output(0)),  Mask({{}, {}, {}, {}}));
+    compare_masks(*getMask(conv1->output(0)),  Mask({{}, {}, {}, {}}));
+
+    compare_masks(*getMask(weights_2.get_node_shared_ptr()->output(0)),  Mask({{}, {}, {}, {}}));
+    compare_masks(*getMask(conv2->output(0)),  Mask({{}, {}, {}, {}}));
+
+    compare_masks(*getMask(weights_3.get_node_shared_ptr()->output(0)),  Mask({{}, {}, {}, {}}));
+    compare_masks(*getMask(conv3->output(0)),  Mask({{}, {}, {}, {}}));
+
+    compare_masks(*getMask(add_const.get_node_shared_ptr()->output(0)),  Mask({{}, {}, {}, {}}));
+    compare_masks(*getMask(add->output(0)),  Mask({{}, {}, {}, {}}));
+
+
+    compare_masks(*getMask(concat->output(0)),  Mask({{}, {}, {}, {}}));
+}

From 327c3149f62e3eb186d0550e9d558d694e4c7a25 Mon Sep 17 00:00:00 2001
From: Anastasia Popova <anastasia.popova@intel.com>
Date: Tue, 8 Jun 2021 09:53:39 +0300
Subject: [PATCH 013/178] Fixed feature dimension calculation in
 grouped_convolutions_fusing(). (#6054)

* Fixed feature dim calculation in concat_convolutions().

* Fixed feature dim calculation in concat_convolutions().

* Added comments, added unit test.
---
 .../middle/passes/fusing/fuse_grouped_conv.py |   5 +-
 .../passes/fusing/fuse_grouped_conv_test.py   | 106 ++++++++++++++++++
 2 files changed, 110 insertions(+), 1 deletion(-)
 create mode 100644 model-optimizer/unit_tests/mo/middle/passes/fusing/fuse_grouped_conv_test.py

diff --git a/model-optimizer/mo/middle/passes/fusing/fuse_grouped_conv.py b/model-optimizer/mo/middle/passes/fusing/fuse_grouped_conv.py
index afb3de7648f0d7..bc256ca49905f0 100644
--- a/model-optimizer/mo/middle/passes/fusing/fuse_grouped_conv.py
+++ b/model-optimizer/mo/middle/passes/fusing/fuse_grouped_conv.py
@@ -87,7 +87,10 @@ def concat_convolutions(graph: Graph, start_node: Node, last_node: Node):
     weights_value = np.array(weights_node.value)
     bias_value = np.array(bias_node.value) if has_biases else None
 
-    feature_dim = 3 if graph.graph['layout'] == 'NHWC' else 0
+    # gconv.get_weights_permute.perm contains permutation indices
+    # where feature dimension is set to zero position, so 0 value
+    # in gconv.get_weights_permute.inv indicates original feature dimension index
+    feature_dim = np.where(gconv.get_weights_permute.inv == 0)[0][0]
 
     for conv in conv_nodes[1:]:
         weights_value = np.concatenate((weights_value, conv.in_node(1).value), axis=feature_dim)
diff --git a/model-optimizer/unit_tests/mo/middle/passes/fusing/fuse_grouped_conv_test.py b/model-optimizer/unit_tests/mo/middle/passes/fusing/fuse_grouped_conv_test.py
new file mode 100644
index 00000000000000..8e8bc61077a5bd
--- /dev/null
+++ b/model-optimizer/unit_tests/mo/middle/passes/fusing/fuse_grouped_conv_test.py
@@ -0,0 +1,106 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import unittest
+
+import numpy as np
+
+from mo.front.common.partial_infer.utils import int64_array
+from mo.graph.graph import Node
+from mo.middle.passes.fusing.fuse_grouped_conv import grouped_convolutions_fusing
+from mo.ops.op import PermuteAttrs
+from mo.utils.ir_engine.compare_graphs import compare_graphs
+from unit_tests.utils.graph import build_graph, result, connect, regular_op_with_shaped_data, regular_op, shaped_data, \
+    valued_const_with_data, shaped_const_with_data, valued_data
+
+nodes = {
+    **regular_op_with_shaped_data('placeholder1', [1, 16, 10, 10], {'type': 'Parameter'}),
+
+    **valued_const_with_data('split_1_axis', int64_array(1), {'type': 'Const'}),
+    **regular_op('split_1', {'type': 'Split', 'can_be_fused': True}),
+    **shaped_data('split_1_data1', [1, 4, 10, 10]),
+    **shaped_data('split_1_data2', [1, 4, 10, 10]),
+    **shaped_data('split_1_data3', [1, 4, 10, 10]),
+    **shaped_data('split_1_data4', [1, 4, 10, 10]),
+
+    **shaped_const_with_data('split_2_in_const_weights', int64_array([3, 3, 4, 16]), {'type': 'Const'}),
+    **regular_op('split_2', {'type': 'Split'}),
+    **valued_data('split_2_data1', np.zeros([3, 3, 4, 4])),
+    **valued_data('split_2_data2', np.zeros([3, 3, 4, 4])),
+    **valued_data('split_2_data3', np.zeros([3, 3, 4, 4])),
+    **valued_data('split_2_data4', np.zeros([3, 3, 4, 4])),
+
+    **regular_op_with_shaped_data('conv2d_1', [1, 4, 8, 8],
+                                  {'type': 'Convolution', 'channel_dims': np.array([1]), 'pad': np.array([2, 2]),
+                                   'stride': np.array([2, 2]),
+                                   'get_weights_permute': PermuteAttrs.Permutation(perm=int64_array([3, 2, 0, 1]),
+                                                                                   inv=int64_array([2, 3, 1, 0])),
+                                   'group': 1, 'output': 4, 'output_shape': [1, 4, 8, 8], 'can_be_fused': True}),
+    **regular_op_with_shaped_data('conv2d_2', [1, 4, 8, 8],
+                                  {'type': 'Convolution', 'pad': np.array([2, 2]), 'stride': np.array([2, 2]),
+                                   'can_be_fused': True}),
+    **regular_op_with_shaped_data('conv2d_3', [1, 4, 8, 8],
+                                  {'type': 'Convolution', 'pad': np.array([2, 2]), 'stride': np.array([2, 2]),
+                                   'can_be_fused': True}),
+    **regular_op_with_shaped_data('conv2d_4', [1, 4, 8, 8],
+                                  {'type': 'Convolution', 'pad': np.array([2, 2]), 'stride': np.array([2, 2]),
+                                   'can_be_fused': True}),
+
+    **regular_op_with_shaped_data('concat', [1, 16, 8, 8], {'type': 'Concat', 'axis': np.array(1)}),
+
+    **regular_op_with_shaped_data('fused_group_conv', [1, 16, 8, 8],
+                                  {'type': 'Convolution', 'channel_dims': np.array([1]), 'pad': np.array([2, 2]),
+                                   'stride': np.array([2, 2]),
+                                   'get_weights_permute': PermuteAttrs.Permutation(perm=int64_array([3, 2, 0, 1]),
+                                                                                   inv=int64_array([2, 3, 1, 0])),
+                                   'group': 1, 'output': 4, 'output_shape': [1, 4, 8, 8], 'can_be_fused': True}),
+    **shaped_const_with_data('new_weights_const', int64_array([3, 3, 4, 16]), {'type': 'Const'}),
+
+    **result('result')
+}
+
+
+class FuseGroupedConvTest(unittest.TestCase):
+    def test_fuse_grouped_conv(self):
+        graph = build_graph(nodes, [*connect('placeholder1', '0:split_1'), *connect('split_1_axis', '1:split_1'),
+                                    ('split_1', 'split_1_data1', {'out': 0}),
+                                    ('split_1', 'split_1_data2', {'out': 1}),
+                                    ('split_1', 'split_1_data3', {'out': 2}),
+                                    ('split_1', 'split_1_data4', {'out': 3}),
+
+                                    *connect('split_2_in_const_weights', 'split_2'),
+                                    ('split_2', 'split_2_data1', {'out': 0}),
+                                    ('split_2', 'split_2_data2', {'out': 1}),
+                                    ('split_2', 'split_2_data3', {'out': 2}),
+                                    ('split_2', 'split_2_data4', {'out': 3}),
+
+                                    ('split_1_data1', 'conv2d_1', {'in': 0}),
+                                    ('split_1_data2', 'conv2d_2', {'in': 0}),
+                                    ('split_1_data3', 'conv2d_3', {'in': 0}),
+                                    ('split_1_data4', 'conv2d_4', {'in': 0}),
+
+                                    ('split_2_data1', 'conv2d_1', {'in': 1}),
+                                    ('split_2_data2', 'conv2d_2', {'in': 1}),
+                                    ('split_2_data3', 'conv2d_3', {'in': 1}),
+                                    ('split_2_data4', 'conv2d_4', {'in': 1}),
+
+                                    *connect('conv2d_1', '0:concat'),
+                                    *connect('conv2d_2', '1:concat'),
+                                    *connect('conv2d_3', '2:concat'),
+                                    *connect('conv2d_4', '3:concat'),
+
+                                    *connect('concat', 'result')])
+
+        graph_ref = build_graph(nodes, [*connect('placeholder1', '0:fused_group_conv'),
+                                        *connect('new_weights_const', '1:fused_group_conv'),
+                                        *connect('fused_group_conv', 'result')])
+
+        graph.graph['layout'] = 'NCHW'
+        grouped_convolutions_fusing(graph)
+
+        (flag, resp) = compare_graphs(graph, graph_ref, 'result')
+        self.assertTrue(flag, resp)
+
+        group_conv_node = Node(graph, 'conv2d_1')
+        group_conv_weights_shape = group_conv_node.in_node(1).shape
+        self.assertTrue((group_conv_weights_shape == int64_array([3, 3, 4, 16])).all())

From d3beab79b22b84268d489d727295641c0817f289 Mon Sep 17 00:00:00 2001
From: Pavel Esir <pavel.esir@intel.com>
Date: Tue, 8 Jun 2021 10:12:35 +0300
Subject: [PATCH 014/178] [nG] fix strided_slice_to_crop (#6032)

---
 .../convert_strided_slice_to_crop.cpp         | 11 ++--
 .../convert_strided_slice_to_crop_test.cpp    | 52 ++++++++++++++++++-
 2 files changed, 58 insertions(+), 5 deletions(-)

diff --git a/inference-engine/src/legacy_api/src/transformations/convert_opset1_to_legacy/convert_strided_slice_to_crop.cpp b/inference-engine/src/legacy_api/src/transformations/convert_opset1_to_legacy/convert_strided_slice_to_crop.cpp
index a3ec122b9c7ab3..cab07f54a762e6 100644
--- a/inference-engine/src/legacy_api/src/transformations/convert_opset1_to_legacy/convert_strided_slice_to_crop.cpp
+++ b/inference-engine/src/legacy_api/src/transformations/convert_opset1_to_legacy/convert_strided_slice_to_crop.cpp
@@ -9,7 +9,6 @@
 #include <vector>
 
 #include <ngraph/opsets/opset1.hpp>
-
 #include <legacy/ngraph_ops/crop_ie.hpp>
 #include <ngraph/rt_info.hpp>
 
@@ -137,7 +136,6 @@ ngraph::pass::ConvertStridedSliceToCropMatcher::ConvertStridedSliceToCropMatcher
                     lb = std::min(static_cast<int64_t>(input_shape[input_shape_idx]), lb);
                     ub = std::min(static_cast<int64_t>(input_shape[input_shape_idx]), ub);
 
-                    offset.emplace_back(lb);
 
                     // set default value for stride or use given value
                     int64_t stride = 1;
@@ -153,6 +151,7 @@ ngraph::pass::ConvertStridedSliceToCropMatcher::ConvertStridedSliceToCropMatcher
                             ub = -1;
 
                         lb = std::min(lb, static_cast<int64_t>(input_shape[input_shape_idx]) - 1);
+                        offset.emplace_back(lb);
                         lb -= 1;  // we always get 1st element, so we need decrease range
                         if (ub <= lb)
                             dimension = (ub - lb) / stride + 1;
@@ -160,12 +159,16 @@ ngraph::pass::ConvertStridedSliceToCropMatcher::ConvertStridedSliceToCropMatcher
                         // apply masks
                         if (begin_mask.count(axis))
                             lb = 0;
-                        if (end_mask.count(axis))
+                        offset.emplace_back(lb);
+
+                        if (end_mask.count(axis)) {
                             ub = static_cast<int64_t>(input_shape[input_shape_idx]);
+                        }
 
                         lb += 1;  // we always get 1st element, so we need decrease range
-                        if (ub >= lb)
+                        if (ub >= lb) {
                             dimension = (ub - lb) / stride + 1;
+                        }
                     }
 
                     dim.emplace_back(dimension);
diff --git a/inference-engine/tests/functional/inference_engine/transformations/convert_strided_slice_to_crop_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/convert_strided_slice_to_crop_test.cpp
index be5560f367ed94..731d96bbd3a2c9 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/convert_strided_slice_to_crop_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/convert_strided_slice_to_crop_test.cpp
@@ -53,6 +53,7 @@ TEST(TransformationTests, ConvertStridedSliceToCropTests1) {
         manager.register_pass<ngraph::pass::InitNodeInfo>();
         manager.register_pass<ngraph::pass::ConvertStridedSliceToCropMatcher>();
         manager.run_passes(f);
+
         ASSERT_NO_THROW(check_rt_info(f));
     }
 
@@ -238,4 +239,53 @@ TEST(TransformationTests, ConvertStridedSliceToCropNegative2) {
 
     auto res = compare_functions(f, f_ref);
     ASSERT_TRUE(res.first) << res.second;
-}
\ No newline at end of file
+}
+
+
+TEST(TransformationTests, ConvertStridedSliceToCropNoneZeroBeginValuesWithMask) {
+    // when begin_mask/end_mask are present begin/end values should not affect output shape
+    std::shared_ptr<ngraph::Function> f(nullptr), f_ref(nullptr);
+    {
+        auto input        = std::make_shared<ngraph::opset1::Parameter>(ngraph::element::f32, ngraph::Shape{1, 2, 4});
+        auto slice_begin  = ngraph::opset1::Constant::create(ngraph::element::i64, ngraph::Shape{4}, {0, 3, 2, 1});
+        auto slice_end    = ngraph::opset1::Constant::create(ngraph::element::i64, ngraph::Shape{4}, {0, 0, 0, 2});
+        auto slice_stride = ngraph::opset1::Constant::create(ngraph::element::i64, ngraph::Shape{4}, {1, 1, 1, 1});
+
+        std::vector<int64_t> begin_mask       = {1, 0, 1, 1};
+        std::vector<int64_t> end_mask         = {1, 0, 1, 0};
+        std::vector<int64_t> new_axis_mask    = {0, 1, 0, 0};
+        std::vector<int64_t> shrink_axis_mask = {0, 0, 0, 0};
+        std::vector<int64_t> ellipsis_mask    = {0, 0, 0, 0};
+
+        auto sslice = std::make_shared<ngraph::opset1::StridedSlice>(input, slice_begin, slice_end, slice_stride,
+                                                                     begin_mask, end_mask,
+                                                                     new_axis_mask, shrink_axis_mask, ellipsis_mask);
+        sslice->set_friendly_name("strided_slice");
+
+        f = std::make_shared<ngraph::Function>(ngraph::NodeVector{sslice}, ngraph::ParameterVector{input});
+        ngraph::pass::Manager manager;
+        manager.register_pass<ngraph::pass::InitNodeInfo>();
+        manager.register_pass<ngraph::pass::ConvertStridedSliceToCropMatcher>();
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+
+    {
+        auto input = std::make_shared<ngraph::opset1::Parameter>(ngraph::element::f32, ngraph::Shape{1, 2, 4});
+
+        std::vector<int64_t> axes   = {0, 1, 2, 3};
+        std::vector<int64_t> dim    = {1, 1, 2, 2};
+        std::vector<int64_t> offset = {0, 0, 0, 0};
+
+        auto reshape = ngraph::op::util::reshapeTo(input, {1, 1, 2, 4});
+        reshape->set_friendly_name("strided_slice/Reshape_for_Crop");
+
+        auto crop = std::make_shared<ngraph::op::CropIE>(reshape, axes, dim, offset);
+        crop->set_friendly_name("strided_slice");
+
+        f_ref = std::make_shared<ngraph::Function>(ngraph::NodeVector{crop}, ngraph::ParameterVector{input});
+    }
+
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}

From 4409a74dcfd82af8b962d4e6b31f9b92934fff81 Mon Sep 17 00:00:00 2001
From: Anton Romanov <anton.romanov@intel.com>
Date: Tue, 8 Jun 2021 10:16:37 +0300
Subject: [PATCH 015/178] samples: Fixed klocwork issues in speech (#6066)

---
 .../samples/speech_sample/fileutils.cpp       | 41 +++++++++++--------
 thirdparty/cnpy/cnpy.cpp                      | 12 +++---
 2 files changed, 30 insertions(+), 23 deletions(-)

diff --git a/inference-engine/samples/speech_sample/fileutils.cpp b/inference-engine/samples/speech_sample/fileutils.cpp
index f3211a21a4b450..102cca25297444 100644
--- a/inference-engine/samples/speech_sample/fileutils.cpp
+++ b/inference-engine/samples/speech_sample/fileutils.cpp
@@ -108,15 +108,18 @@ void NumpyFile::GetFileInfo(const char* fileName, uint32_t numArrayToFindSize, u
     cnpy::npz_t my_npz1 = cnpy::npz_load(fileName);
     auto it = my_npz1.begin();
     std::advance(it, numArrayToFindSize);
-
-    numArrays = my_npz1.size();
-    cnpy::NpyArray my_npy = it->second;
-    numMemoryBytes = my_npy.data_holder->size();
-
-    if (ptrNumArrays != NULL)
-        *ptrNumArrays = numArrays;
-    if (ptrNumMemoryBytes != NULL)
-        *ptrNumMemoryBytes = numMemoryBytes;
+    if (it != my_npz1.end()) {
+        numArrays = my_npz1.size();
+        cnpy::NpyArray my_npy = it->second;
+        numMemoryBytes = my_npy.data_holder->size();
+
+        if (ptrNumArrays != NULL)
+            *ptrNumArrays = numArrays;
+        if (ptrNumMemoryBytes != NULL)
+            *ptrNumMemoryBytes = numMemoryBytes;
+    } else {
+        throw std::runtime_error(std::string("Failed to get info %s  GetFileInfo()!\n") + fileName);
+    }
 }
 
 void NumpyFile::LoadFile(const char* fileName, uint32_t arrayIndex, std::string& ptrName, std::vector<uint8_t>& memory, uint32_t* ptrNumRows,
@@ -124,16 +127,20 @@ void NumpyFile::LoadFile(const char* fileName, uint32_t arrayIndex, std::string&
     cnpy::npz_t my_npz1 = cnpy::npz_load(fileName);
     auto it = my_npz1.begin();
     std::advance(it, arrayIndex);
-    ptrName = it->first;
-    cnpy::NpyArray my_npy = it->second;
-    *ptrNumRows = my_npy.shape[0];
-    *ptrNumColumns = my_npy.shape[1];
+    if (it != my_npz1.end()) {
+        ptrName = it->first;
+        cnpy::NpyArray my_npy = it->second;
+        *ptrNumRows = my_npy.shape[0];
+        *ptrNumColumns = my_npy.shape[1];
+
+        for (size_t i = 0; i < my_npy.data_holder->size(); i++) {
+            memory.at(i) = my_npy.data_holder->at(i);
+        }
 
-    for (size_t i = 0; i < my_npy.data_holder->size(); i++) {
-        memory.at(i) = my_npy.data_holder->at(i);
+        *ptrNumBytesPerElement = sizeof(float);
+    } else {
+        throw std::runtime_error(std::string("Failed to open %s for reading in LoadFile()!\n") + fileName);
     }
-
-    *ptrNumBytesPerElement = sizeof(float);
 }
 
 void NumpyFile::SaveFile(const char* fileName, bool shouldAppend, std::string name, void* ptrMemory, uint32_t numRows, uint32_t numColumns) {
diff --git a/thirdparty/cnpy/cnpy.cpp b/thirdparty/cnpy/cnpy.cpp
index a3a3e0ef406660..26d0614bca1906 100644
--- a/thirdparty/cnpy/cnpy.cpp
+++ b/thirdparty/cnpy/cnpy.cpp
@@ -183,9 +183,9 @@ void cnpy::parse_zip_footer(FILE* fp, uint16_t& nrecs, size_t& global_header_siz
 }
 
 cnpy::NpyArray load_the_npy_file(FILE* fp) {
-    std::vector<size_t> shape;
-    size_t word_size;
-    bool fortran_order;
+    std::vector<size_t> shape(0);
+    size_t word_size = 0;
+    bool fortran_order = false;
     cnpy::parse_npy_header(fp,word_size,shape,fortran_order);
     if (word_size >= 0 && word_size < ULLONG_MAX) {
         cnpy::NpyArray arr(shape, word_size, fortran_order);
@@ -225,9 +225,9 @@ cnpy::NpyArray load_the_npz_array(FILE* fp, uint32_t compr_bytes, uint32_t uncom
     err = inflate(&d_stream, Z_FINISH);
     err = inflateEnd(&d_stream);
 
-    std::vector<size_t> shape;
-    size_t word_size;
-    bool fortran_order;
+    std::vector<size_t> shape(0);
+    size_t word_size = 0;
+    bool fortran_order = false;
     cnpy::parse_npy_header(&buffer_uncompr[0],word_size,shape,fortran_order);
     if (word_size >= 0 && word_size < ULLONG_MAX) {
         cnpy::NpyArray array(shape, word_size, fortran_order);

From 98f45ffbddc94ea480fa7e182345ba6dcaa7f680 Mon Sep 17 00:00:00 2001
From: Maxim Andronov <maxim.andronov@intel.com>
Date: Tue, 8 Jun 2021 10:19:25 +0300
Subject: [PATCH 016/178] ConvertInterpolate1ToInterpolate4 fixes (#6019)

* half_pixel -> asymmetric and round_prefer_floor -> simple in ConvertInterpolate1ToInterpolate4

* test fix
---
 .../convert_interpolate1_to_interpolate4.cpp         | 10 ++++++++--
 .../convert_interpolate1_to_interpolate4_test.cpp    | 12 ++++++------
 2 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/inference-engine/src/transformations/src/transformations/op_conversions/convert_interpolate1_to_interpolate4.cpp b/inference-engine/src/transformations/src/transformations/op_conversions/convert_interpolate1_to_interpolate4.cpp
index 94173079c6203f..36a58551a68a03 100644
--- a/inference-engine/src/transformations/src/transformations/op_conversions/convert_interpolate1_to_interpolate4.cpp
+++ b/inference-engine/src/transformations/src/transformations/op_conversions/convert_interpolate1_to_interpolate4.cpp
@@ -68,14 +68,20 @@ ngraph::pass::ConvertInterpolate1ToInterpolate4::ConvertInterpolate1ToInterpolat
             return false;
         }
         attrsV4.shape_calculation_mode = ngraph::opset4::Interpolate::ShapeCalcMode::sizes;
-        attrsV4.nearest_mode = ngraph::opset4::Interpolate::NearestMode::round_prefer_floor;
+        attrsV4.nearest_mode = ngraph::opset4::Interpolate::NearestMode::simple;
         attrsV4.pads_begin = attrsV0.pads_begin;
         attrsV4.pads_end = attrsV0.pads_end;
         attrsV4.antialias = attrsV0.antialias;
-        attrsV4.coordinate_transformation_mode = ngraph::opset4::Interpolate::CoordinateTransformMode::half_pixel;
+        attrsV4.coordinate_transformation_mode = ngraph::opset4::Interpolate::CoordinateTransformMode::asymmetric;
         attrsV4.cube_coeff = -0.75f;
         if (attrsV0.align_corners) {
             attrsV4.coordinate_transformation_mode = ngraph::opset4::Interpolate::CoordinateTransformMode::align_corners;
+        } else if ((attrsV4.mode == ngraph::op::v4::Interpolate::InterpolateMode::linear_onnx ||
+                    attrsV4.mode == ngraph::op::v4::Interpolate::InterpolateMode::linear) &&
+                    std::all_of(attrsV4.pads_begin.begin(), attrsV4.pads_begin.end(), [](size_t i){return i == 0;}) &&
+                    std::all_of(attrsV4.pads_end.begin(), attrsV4.pads_end.end(), [](size_t i){return i == 0;}) &&
+                    !(input_shape_rank - 2 == 2 && attrsV0.axes == AxisSet{2, 3})) {
+            attrsV4.coordinate_transformation_mode = ngraph::opset4::Interpolate::CoordinateTransformMode::half_pixel;
         }
 
         auto interpolateV4 = std::make_shared<ngraph::opset4::Interpolate>(interpolationV0->input_value(0), interpolationV0->input_value(1),
diff --git a/inference-engine/tests/functional/inference_engine/transformations/convert_interpolate1_to_interpolate4_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/convert_interpolate1_to_interpolate4_test.cpp
index 9468db9287d855..12177f78cbcd4a 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/convert_interpolate1_to_interpolate4_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/convert_interpolate1_to_interpolate4_test.cpp
@@ -54,7 +54,7 @@ TEST(TransformationTests, ConvertInterpolate1ToInterpolate4) {
 
         auto interpolate4_attr = opset4::Interpolate::InterpolateAttrs(opset4::Interpolate::InterpolateMode::nearest,
             opset4::Interpolate::ShapeCalcMode::sizes, std::vector<size_t>{0, 0, 0, 0}, std::vector<size_t>{0, 0, 0, 0},
-            opset4::Interpolate::CoordinateTransformMode::asymmetric, opset4::Interpolate::NearestMode::floor,
+            opset4::Interpolate::CoordinateTransformMode::asymmetric, opset4::Interpolate::NearestMode::simple,
             false, -0.75);
 
         auto interpolate4 = std::make_shared<opset4::Interpolate>(data_node, out_shape_node, default_scales_node, axes_node, interpolate4_attr);
@@ -62,7 +62,7 @@ TEST(TransformationTests, ConvertInterpolate1ToInterpolate4) {
         f_ref = std::make_shared<Function>(NodeVector{interpolate4}, ParameterVector{data_node});
     }
 
-    auto res = compare_functions(f, f_ref);
+    auto res = compare_functions(f, f_ref, true, false, false, true, true);
     ASSERT_TRUE(res.first) << res.second;
 }
 
@@ -97,16 +97,16 @@ TEST(TransformationTests, ConvertInterpolate1ToInterpolate4_1) {
         auto default_scales_node = opset1::Constant::create(ngraph::element::f32, Shape{2}, {4.0f / 3.0f, 4.0f / 3.0f});
         auto axes_node = opset1::Constant::create(ngraph::element::i64, Shape{2}, {2, 3});
 
-        auto interpolate4_attr = opset4::Interpolate::InterpolateAttrs(opset4::Interpolate::InterpolateMode::linear,
+        auto interpolate4_attr = opset4::Interpolate::InterpolateAttrs(opset4::Interpolate::InterpolateMode::linear_onnx,
             opset4::Interpolate::ShapeCalcMode::sizes, std::vector<size_t>{0, 0, 0, 0}, std::vector<size_t>{0, 0, 0, 0},
-            opset4::Interpolate::CoordinateTransformMode::align_corners, opset4::Interpolate::NearestMode::floor,
-            false, -0.75);
+            opset4::Interpolate::CoordinateTransformMode::asymmetric, opset4::Interpolate::NearestMode::simple,
+            true, -0.75);
 
         auto interpolate4 = std::make_shared<opset4::Interpolate>(data_node, out_shape_node, default_scales_node, axes_node, interpolate4_attr);
 
         f_ref = std::make_shared<Function>(NodeVector{interpolate4}, ParameterVector{data_node});
     }
 
-    auto res = compare_functions(f, f_ref);
+    auto res = compare_functions(f, f_ref, true, false, false, true, true);
     ASSERT_TRUE(res.first) << res.second;
 }

From a99343f4ebd432324eb1e7262db1c77ad4fe7806 Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Tue, 8 Jun 2021 11:00:02 +0300
Subject: [PATCH 017/178] Python tools (#6067)

* Removed useless lines

* Enable find PythonInterp only if we ENABLE_PYTHON

* Enable docs only if python interp is found
---
 cmake/features.cmake                          |  5 +++--
 .../conditional_compilation/CMakeLists.txt    |  5 +----
 tools/CMakeLists.txt                          | 19 ++++++++++---------
 3 files changed, 14 insertions(+), 15 deletions(-)

diff --git a/cmake/features.cmake b/cmake/features.cmake
index fe1b8919b51212..adb1fad25238da 100644
--- a/cmake/features.cmake
+++ b/cmake/features.cmake
@@ -18,8 +18,6 @@ Supported values:\
 
 ie_option (ENABLE_PROFILING_FIRST_INFERENCE "Build with ITT tracing of first inference time." ON)
 
-ie_option (ENABLE_DOCS "Build docs using Doxygen" OFF)
-
 ie_option(ENABLE_TEMPLATE_PLUGIN "Register template plugin into plugins.xml" OFF)
 
 ie_option_enum(SELECTIVE_BUILD "Enable OpenVINO conditional compilation or statistics collection. \
@@ -33,6 +31,9 @@ ie_option(ENABLE_ERROR_HIGHLIGHT "Highlight errors and warnings during compile t
 find_package(PythonLibs 3 QUIET)
 ie_dependent_option (ENABLE_PYTHON "enables ie python bridge build" OFF "PYTHONLIBS_FOUND" OFF)
 
+find_package(PythonInterp 3 QUIET)
+ie_dependent_option (ENABLE_DOCS "Build docs using Doxygen" OFF "PYTHONINTERP_FOUND" OFF)
+
 #
 # enable or disable output from NGRAPH_DEBUG statements
 #
diff --git a/openvino/conditional_compilation/CMakeLists.txt b/openvino/conditional_compilation/CMakeLists.txt
index 4a27ac50b7cb84..1db12d4eefbd0a 100644
--- a/openvino/conditional_compilation/CMakeLists.txt
+++ b/openvino/conditional_compilation/CMakeLists.txt
@@ -19,10 +19,7 @@ elseif(SELECTIVE_BUILD STREQUAL "ON")
         message(FATAL_ERROR "In case SELECTIVE_BUILD is enabled, the SELECTIVE_BUILD_STAT variable should contain the path to the collected InelSEAPI statistics.\
  Usage: -DSELECTIVE_BUILD=ON -DSELECTIVE_BUILD_STAT=/path/*.csv")
     endif()
-    find_package (PythonInterp REQUIRED)
-    if(NOT PYTHON_VERSION_MAJOR EQUAL 3)
-        message(FATAL_ERROR " Python3 wasn't found!")
-    endif()
+    find_package (PythonInterp 3 REQUIRED)
 
     file(GLOB STAT_FILES ${SELECTIVE_BUILD_STAT})
 
diff --git a/tools/CMakeLists.txt b/tools/CMakeLists.txt
index 0a09df95faa8bc..fdc6d9ef11a8f4 100644
--- a/tools/CMakeLists.txt
+++ b/tools/CMakeLists.txt
@@ -1,24 +1,25 @@
 # Copyright (C) 2018-2021 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 cmake_minimum_required(VERSION 3.13)
+
 project(python_tools)
 
 if(NOT DEFINED OpenVINO_MAIN_SOURCE_DIR)
     find_package(InferenceEngineDeveloperPackage QUIET)
 endif()
 
-find_package(PythonInterp 3 REQUIRED)
-set(PYTHON_VERSION python${PYTHON_VERSION_MAJOR}.${PYTHON_VERSION_MINOR})
+if(ENABLE_PYTHON)
+    find_package(PythonInterp 3 REQUIRED)
+    set(PYTHON_VERSION python${PYTHON_VERSION_MAJOR}.${PYTHON_VERSION_MINOR})
 
-set(TARGET_NAME "python_tools")
+    set(TARGET_NAME "python_tools")
 
-if(WIN32)
-    set(PYTHON_BRIDGE_OUTPUT_DIRECTORY ${CMAKE_LIBRARY_OUTPUT_DIRECTORY}/$<CONFIG>/python_api/${PYTHON_VERSION}/openvino)
-else()
-    set(PYTHON_BRIDGE_OUTPUT_DIRECTORY ${CMAKE_LIBRARY_OUTPUT_DIRECTORY}/python_api/${PYTHON_VERSION}/openvino)
-endif()
+    if(WIN32)
+        set(PYTHON_BRIDGE_OUTPUT_DIRECTORY ${CMAKE_LIBRARY_OUTPUT_DIRECTORY}/$<CONFIG>/python_api/${PYTHON_VERSION}/openvino)
+    else()
+        set(PYTHON_BRIDGE_OUTPUT_DIRECTORY ${CMAKE_LIBRARY_OUTPUT_DIRECTORY}/python_api/${PYTHON_VERSION}/openvino)
+    endif()
 
-if(ENABLE_PYTHON)
     # creates a copy inside bin directory for developers to have ability running python benchmark_app
     add_custom_target(${TARGET_NAME} ALL
         COMMAND ${CMAKE_COMMAND} -E make_directory ${PYTHON_BRIDGE_OUTPUT_DIRECTORY}/tools

From 77d8973e91a91bbd3c92316b348599b444545839 Mon Sep 17 00:00:00 2001
From: Maxim Shevtsov <maxim.y.shevtsov@intel.com>
Date: Tue, 8 Jun 2021 11:10:05 +0300
Subject: [PATCH 018/178] Perf Intro page brushing (#5861)

* brushing

* Update Intro_to_Performance.md
---
 docs/IE_DG/Intro_to_Performance.md | 30 ++++++++++++++++--------------
 1 file changed, 16 insertions(+), 14 deletions(-)

diff --git a/docs/IE_DG/Intro_to_Performance.md b/docs/IE_DG/Intro_to_Performance.md
index 66fcf48c34f3c5..78d5c59c417d0f 100644
--- a/docs/IE_DG/Intro_to_Performance.md
+++ b/docs/IE_DG/Intro_to_Performance.md
@@ -1,24 +1,28 @@
 # Introduction to the Performance Topics {#openvino_docs_IE_DG_Intro_to_Performance}
 
 This section is a shorter version of the
-[Optimization Guide](supported_plugins/MULTI.md) for the Intel Deep Learning Deployment Toolkit.
+[Optimization Guide](../optimization_guide/dldt_optimization_guide.md) for the Intel® Distribution of OpenVINO™ Toolkit.
 
 ## Precision
 Inference precision directly affects the performance. 
 
-Model Optimizer can produce an IR with different precision. For example, float16 IR initially targets VPU and GPU devices, while, for example, the CPU can also execute regular float32.
-Also, further device-specific inference precision settings are available, for example, [8-bit integer](Int8Inference.md) or [bfloat16](Bfloat16Inference.md) inference on the CPU.
-Note that for [MULTI device](supported_plugins/MULTI.md) that supports automatic inference on multiple devices in parallel, you can use the FP16 IR.
+Model Optimizer can produce an IR with different precision. For example, an FP16 IR initially targets VPU and GPU devices, while, for example, for the CPU, an FP16 IR is    typically up-scaled to the regular FP32 automatically upon loading. But notice that further device-specific inference precision settings are available, 
+for example, [8-bit integer](Int8Inference.md) or [bfloat16](Bfloat16Inference.md), which is specific to the CPU inference, below.
+Note that for the [MULTI device](supported_plugins/MULTI.md) plugin that supports automatic inference on multiple devices in parallel, you can use an FP16 IR (no need for FP32).
 You can find more information, including preferred data types for specific devices, in the
-[Supported Devices](supported_plugins/Supported_Devices.md) section.
+[Supported Devices](supported_plugins/Supported_Devices.md) document.
 
-## Lowering Inference Precision
-Default optimization is used for CPU and implies that inference is made with lower precision if it is possible on a given platform to reach better performance with acceptable range of accuracy.
-This approach can be used for CPU devices where the platform supports the AVX512_BF16 instruction. In this case, a regular float32 model is converted to [bfloat16](Bfloat16Inference.md) internal representation and inference is provided with bfloat16 layers usage.
-Below is the example command line to disable this feature on the CPU device with the AVX512_BF16 instruction and execute regular float32.
+## Automatic Lowering of the Inference Precision
+By default, plugins enable the optimizations that allow lower precision if the acceptable range of accuracy is preserved.
+For example, for the CPU that supports the AVX512_BF16 instructions, an FP16/FP32 model is converted to a [bfloat16](Bfloat16Inference.md) IR to accelerate inference.
+To compare the associated speedup, run the example command below to disable this feature on the CPU device with the AVX512_BF16 support and get regular FP32 execution:
 ```
 $ benchmark_app -m <model.xml> -enforcebf16=false
  ```
+Notice that for quantized (e.g. INT8) models the bfloat16 calculations (of the layers that remain in FP32) is disabled by default.
+Refer to the [CPU Plugin documentation](supported_plugins/CPU.md) for more details.
+
+Similarly, the GPU device has a dedicated config key to enable FP16 execution of the layers that remain in FP32 in the quantized models (as the quantization is typically performed on the FP32 models), refer to the ENABLE_FP16_FOR_QUANTIZED_MODELS key in the [GPU Plugin documentation](supported_plugins/CL_DNN.md) 
 
 ## Latency vs. Throughput
 One way to increase computational efficiency is batching, which combines many (potentially tens) of
@@ -44,17 +48,17 @@ Below is the example command line that limits the execution to the single socket
 limited to the single socket).
 $ numactl -m 0 --physcpubind 0-27  benchmark_app -m <model.xml> -api sync -nthreads 28
  ```
-Note that if you have more than one input, running as many inference requests as you have NUMA nodes (or sockets)
+Note that if you have more than one input, running as many inference streams as you have NUMA nodes (or sockets)
 usually gives the same best latency as a single request on the single socket, but much higher throughput. Assuming two NUMA nodes machine:
 ```
 $ benchmark_app -m <model.xml> -nstreams 2
  ```
 Number of NUMA nodes on the machine can be queried via 'lscpu'.
-Please see more on the NUMA support in the [Optimization Guide](supported_plugins/MULTI.md).
+Please see more on the NUMA support in the [Optimization Guide](../optimization_guide/dldt_optimization_guide.md).
 
 ## Throughput Mode for CPU
 Unlike most accelerators, CPU is perceived as an inherently latency-oriented device. 
-Since 2018 R5 release, the Inference Engine introduced the "throughput" mode, which allows the Inference Engine to efficiently run multiple inference requests on the CPU simultaneously, greatly improving the throughput.
+OpenVINO™ toolkit provides a "throughput" mode that allows running multiple inference requests on the CPU simultaneously, which greatly improves the throughput.
 
 Internally, the execution resources are split/pinned into execution "streams".
 Using this feature gains much better performance for the networks that originally are not scaled well with a number of threads (for example, lightweight topologies). This is especially pronounced for the many-core server machines.
@@ -62,8 +66,6 @@ Using this feature gains much better performance for the networks that originall
 Run the [Benchmark App](../../inference-engine/samples/benchmark_app/README.md) and play with number of infer requests running in parallel, next section. 
 Try different values of the `-nstreams` argument from `1` to a number of CPU cores and find one that provides the best performance. 
 
-In addition to the number of streams, it is also possible to play with the batch size to find the throughput sweet-spot.
-
 The throughput mode relaxes the requirement to saturate the CPU by using a large batch: running multiple independent inference requests in parallel often gives much better performance, than using a batch only.
 This allows you to simplify the app-logic, as you don't need to combine multiple inputs into a batch to achieve good CPU performance.
 Instead, it is possible to keep a separate infer request per camera or another source of input and process the requests in parallel using Async API.

From 9049dee86243d7677eb00fd173a3860562516ad7 Mon Sep 17 00:00:00 2001
From: Maxim Andronov <maxim.andronov@intel.com>
Date: Tue, 8 Jun 2021 11:10:56 +0300
Subject: [PATCH 019/178] klocwork 2021.4 (#6042)

---
 .../InferenceEngine_network_with_state_infer.cpp     |  6 ++++++
 .../src/convert_function_to_cnn_network.cpp          |  3 +++
 .../ngraph_transformations/reshape_fc_fusion.cpp     |  6 ++++++
 .../reshape_fully_connected.cpp                      |  2 ++
 .../src/mkldnn_plugin/nodes/mkldnn_dft_node.cpp      |  2 +-
 .../nodes/mkldnn_gather_elements_node.h              |  2 +-
 .../nodes/mkldnn_strided_slice_node.cpp              |  8 ++++++--
 .../nodes/mkldnn_tensoriterator_node.cpp             | 12 ++++++++++++
 8 files changed, 37 insertions(+), 4 deletions(-)

diff --git a/docs/snippets/InferenceEngine_network_with_state_infer.cpp b/docs/snippets/InferenceEngine_network_with_state_infer.cpp
index 81a3070ba3b319..7af9c076931169 100644
--- a/docs/snippets/InferenceEngine_network_with_state_infer.cpp
+++ b/docs/snippets/InferenceEngine_network_with_state_infer.cpp
@@ -64,7 +64,13 @@ int main(int argc, char *argv[]) {
             inferRequest.Infer();
             // check states
             auto states = inferRequest.QueryState();
+            if (states.empty()) {
+                throw std::runtime_error("Queried states are empty");
+            }
             auto mstate = as<MemoryBlob>(states[0].GetState());
+            if (mstate == nullptr) {
+                throw std::runtime_error("Can't cast state to MemoryBlob");
+            }
             auto state_buf = mstate->rmap();
             float * state =state_buf.as<float*>(); 
             std::cout << state[0] << "\n";
diff --git a/inference-engine/src/legacy_api/src/convert_function_to_cnn_network.cpp b/inference-engine/src/legacy_api/src/convert_function_to_cnn_network.cpp
index 08b54640ee6e26..7d92c77219d834 100644
--- a/inference-engine/src/legacy_api/src/convert_function_to_cnn_network.cpp
+++ b/inference-engine/src/legacy_api/src/convert_function_to_cnn_network.cpp
@@ -244,6 +244,9 @@ CNNLayer::Ptr createSubGraphLayer(const std::shared_ptr<ngraph::Node>& layer) {
     LayerParams params = {layer->get_friendly_name(), "TensorIterator",
                           details::convertPrecision(layer->get_output_element_type(0))};
     auto res = std::make_shared<InferenceEngine::TensorIterator>(params);
+    if (res == nullptr) {
+        IE_THROW() << "Can't create TensorIterator";
+    }
     res->body = body;
 
     // Port map: outputs
diff --git a/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_fc_fusion.cpp b/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_fc_fusion.cpp
index 09d3e7e05540bd..b850bd98ae2979 100644
--- a/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_fc_fusion.cpp
+++ b/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_fc_fusion.cpp
@@ -22,7 +22,11 @@ MKLDNNPlugin::ReshapeFullyConnectedFusion::ReshapeFullyConnectedFusion() {
 
     ngraph::matcher_pass_callback callback = [this](ngraph::pattern::Matcher &m) {
         auto fc = std::dynamic_pointer_cast<MKLDNNPlugin::FullyConnectedNode>(m.get_match_root());
+        if (!fc)
+            return false;
         auto reshape = std::dynamic_pointer_cast<ngraph::opset1::Reshape>(fc->get_input_node_shared_ptr(0));
+        if (!reshape)
+            return false;
 
         // Check that Reshape reshapes 4D tensor to 2D or input shape = output shape
         auto shape_in = reshape->input_value(0).get_shape();
@@ -67,6 +71,8 @@ MKLDNNPlugin::ReshapeFullyConnectedFusion::ReshapeFullyConnectedFusion() {
                                                                         fc->input_value(2),
                                                                         outShape,
                                                                         fc->output(0).get_element_type());
+        } else {
+            return false;
         }
         new_ops.push_back(new_fc);
         new_fc->set_friendly_name(fc->get_friendly_name());
diff --git a/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_fully_connected.cpp b/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_fully_connected.cpp
index 999d1b958d8d91..f140f44e74e701 100644
--- a/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_fully_connected.cpp
+++ b/inference-engine/src/mkldnn_plugin/ngraph_transformations/reshape_fully_connected.cpp
@@ -60,6 +60,8 @@ MKLDNNPlugin::ReshapeFullyConnected::ReshapeFullyConnected() {
                                                                         fc->input_value(2),
                                                                         output_shape_new,
                                                                         fc->get_output_type());
+        } else {
+            return false;
         }
         new_ops.push_back(fc_new);
 
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_dft_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_dft_node.cpp
index 21fb93728c17fb..b9ef511d010fce 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_dft_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_dft_node.cpp
@@ -49,7 +49,7 @@ MKLDNNDFTNode::MKLDNNDFTNode(const std::shared_ptr<ngraph::Node>& op, const mkld
 
     /* Data */
     inputShape = inDims[DATA_INDEX].ToSizeVector();
-    if (inputShape.size() < 1) {
+    if (inputShape.size() < 2) {
         IE_THROW() << layerErrorPrefix << " has invalid 'data' input tensor with rank: " << inputShape.size();
     }
 
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_elements_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_elements_node.h
index 30d1fda9e9553d..bc19866768dfcf 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_elements_node.h
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_elements_node.h
@@ -32,7 +32,7 @@ class MKLDNNGatherElementsNode : public MKLDNNNode {
     size_t dataTypeSize_;
     int strideAxDst_;
     int dstAxDim_;
-    int strideAx1Diff_;
+    int strideAx1Diff_ = 0;
     std::string errorPrefix_;
 
     template <typename dataType>
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_strided_slice_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_strided_slice_node.cpp
index 07cc72247a5eff..1b70de9f0f8341 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_strided_slice_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_strided_slice_node.cpp
@@ -86,7 +86,7 @@ MKLDNNStridedSliceNode::MKLDNNStridedSliceNode(const std::shared_ptr<ngraph::Nod
 
 void MKLDNNStridedSliceNode::getSupportedDescriptors() {
     auto isConstantNode = [](const MKLDNNNodePtr &node) {
-        return node->isConstant() && node->getType() == Input;
+        return node->getType() == Input && node->isConstant();
     };
 
     params.parametersAreConstant = isConstantNode(getParentEdgesAtPort(BEGIN_ID)[0]->getParent()) &&
@@ -138,7 +138,11 @@ void MKLDNNStridedSliceNode::getSupportedDescriptors() {
 
     if (params.parametersAreConstant) {
         auto fillingInParameters = [&](std::vector<int> &parameter, const size_t type, const size_t size, const int value) {
-            auto blob = std::dynamic_pointer_cast<MKLDNNInputNode>(getParentEdgesAtPort(type)[0]->getParent())->getMemoryPtr();
+            const auto constNode = std::dynamic_pointer_cast<MKLDNNInputNode>(getParentEdgesAtPort(type)[0]->getParent());
+            if (!constNode) {
+                THROW_ERROR << "can't cast node on " << type << " port to MKLDNNInputNode";
+            }
+            auto blob = constNode->getMemoryPtr();
             if (blob->GetDataType() != mkldnn::memory::data_type::s32)
                 THROW_ERROR << "supports only parameters input with precision I32";
             const int *ptr = static_cast<const int*>(blob->GetPtr());
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_tensoriterator_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_tensoriterator_node.cpp
index c9a53c79e07865..d1d80e1b7cba7b 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_tensoriterator_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_tensoriterator_node.cpp
@@ -136,6 +136,9 @@ class IterCountPortHelper : public PortMapHelper {
     void execute(mkldnn::stream strm, int n_iter) override {
         auto mem = mem_holder_dst;
         auto data_ptr = static_cast<uint32_t*>(mem.get_data_handle());
+        if (data_ptr == nullptr) {
+            IE_THROW() << "TensorIterator node has not allocated memory for IterCountPortHelper";
+        }
         *data_ptr = n_iter;
     }
 };
@@ -150,6 +153,9 @@ class asBoolCheck : public PortChecker {
 
     int getStatus() override {
         auto data_ptr = static_cast<uint8_t*>(mem_holder.get_data_handle());
+        if (data_ptr == nullptr) {
+            IE_THROW() << "TensorIterator node has not allocated memory for asBoolCheck";
+        }
         return *data_ptr == static_cast<uint8_t>(0) ? 0 : 1;
     }
 };
@@ -164,6 +170,9 @@ class asIntCheck : public PortChecker {
 
     int getStatus() override {
         auto data_ptr = static_cast<uint32_t*>(mem_holder.get_data_handle());
+        if (data_ptr == nullptr) {
+            IE_THROW() << "TensorIterator node has not allocated memory for asIntCheck";
+        }
         return *data_ptr;
     }
 };
@@ -283,6 +292,9 @@ MKLDNNTensorIteratorNode::MKLDNNTensorIteratorNode(const std::shared_ptr<ngraph:
 
 void MKLDNNTensorIteratorNode::getSupportedDescriptors() {
     auto tiOp = std::dynamic_pointer_cast<ngraph::op::util::SubGraphOp>(ngraphOp);
+    if (tiOp == nullptr) {
+        IE_THROW() << "Can't cast TensorIterator node with name: " << getName() << " to ngraph::op::util::SubGraphOp";
+    }
     const std::shared_ptr<const ngraph::Function> body = tiOp->get_function();
     sub_graph.CreateGraph(body, ext_mng, weightCache);
 

From 713e4e1ebe231821e5da6c3e5461d724cb7490c9 Mon Sep 17 00:00:00 2001
From: Mikhail Kozlov <mikhail.kozlov@intel.com>
Date: Tue, 8 Jun 2021 11:17:04 +0300
Subject: [PATCH 020/178] Fix filling input blob for benchmark_app (#6055)

* Fix filling input blob for benchmark_app

* Fix clang format
---
 .../samples/benchmark_app/inputs_filling.cpp  | 21 +++++++++++++++----
 1 file changed, 17 insertions(+), 4 deletions(-)

diff --git a/inference-engine/samples/benchmark_app/inputs_filling.cpp b/inference-engine/samples/benchmark_app/inputs_filling.cpp
index e12f7656f173c6..ef8a045279a57a 100644
--- a/inference-engine/samples/benchmark_app/inputs_filling.cpp
+++ b/inference-engine/samples/benchmark_app/inputs_filling.cpp
@@ -39,6 +39,7 @@ std::vector<std::string> filterFilesByExtensions(const std::vector<std::string>&
     return filtered;
 }
 
+template <typename T>
 void fillBlobImage(Blob::Ptr& inputBlob, const std::vector<std::string>& filePaths, const size_t& batchSize, const benchmark_app::InputInfo& app_info,
                    const size_t& requestId, const size_t& inputId, const size_t& inputSize) {
     MemoryBlob::Ptr minput = as<MemoryBlob>(inputBlob);
@@ -50,7 +51,7 @@ void fillBlobImage(Blob::Ptr& inputBlob, const std::vector<std::string>& filePat
     // locked memory holder should be alive all time while access to its buffer
     // happens
     auto minputHolder = minput->wmap();
-    auto inputBlobData = minputHolder.as<uint8_t*>();
+    auto inputBlobData = minputHolder.as<T*>();
 
     /** Collect images data ptrs **/
     std::vector<std::shared_ptr<uint8_t>> vreader;
@@ -90,7 +91,7 @@ void fillBlobImage(Blob::Ptr& inputBlob, const std::vector<std::string>& filePat
                     size_t offset = imageId * numChannels * width * height + (((app_info.layout == "NCHW") || (app_info.layout == "CHW"))
                                                                                   ? (ch * width * height + h * width + w)
                                                                                   : (h * width * numChannels + w * numChannels + ch));
-                    inputBlobData[offset] = vreader.at(imageId).get()[h * width * numChannels + w * numChannels + ch];
+                    inputBlobData[offset] = static_cast<T>(vreader.at(imageId).get()[h * width * numChannels + w * numChannels + ch]);
                 }
             }
         }
@@ -142,7 +143,7 @@ using uniformDistribution =
                               typename std::conditional<std::is_integral<T>::value, std::uniform_int_distribution<T>, void>::type>::type;
 
 template <typename T, typename T2>
-void fillBlobRandom(Blob::Ptr& inputBlob, T rand_min = std::numeric_limits<T>::min(), T rand_max = std::numeric_limits<T>::max()) {
+void fillBlobRandom(Blob::Ptr& inputBlob, T rand_min = std::numeric_limits<uint8_t>::min(), T rand_max = std::numeric_limits<uint8_t>::max()) {
     MemoryBlob::Ptr minput = as<MemoryBlob>(inputBlob);
     if (!minput) {
         IE_THROW() << "We expect inputBlob to be inherited from MemoryBlob in "
@@ -270,7 +271,19 @@ void fillBlobs(const std::vector<std::string>& inputFiles, const size_t& batchSi
             if (app_info.isImage()) {
                 if (!imageFiles.empty()) {
                     // Fill with Images
-                    fillBlobImage(inputBlob, imageFiles, batchSize, app_info, requestId, imageInputId++, imageInputCount);
+                    if (precision == InferenceEngine::Precision::FP32) {
+                        fillBlobImage<float>(inputBlob, imageFiles, batchSize, app_info, requestId, imageInputId++, imageInputCount);
+                    } else if (precision == InferenceEngine::Precision::FP16) {
+                        fillBlobImage<short>(inputBlob, imageFiles, batchSize, app_info, requestId, imageInputId++, imageInputCount);
+                    } else if (precision == InferenceEngine::Precision::I32) {
+                        fillBlobImage<int32_t>(inputBlob, imageFiles, batchSize, app_info, requestId, imageInputId++, imageInputCount);
+                    } else if (precision == InferenceEngine::Precision::I64) {
+                        fillBlobImage<int64_t>(inputBlob, imageFiles, batchSize, app_info, requestId, imageInputId++, imageInputCount);
+                    } else if (precision == InferenceEngine::Precision::U8) {
+                        fillBlobImage<uint8_t>(inputBlob, imageFiles, batchSize, app_info, requestId, imageInputId++, imageInputCount);
+                    } else {
+                        IE_THROW() << "Input precision is not supported for " << item.first;
+                    }
                     continue;
                 }
             } else {

From eadf2c4ce09d4bde216fee4ad48eac1c5162b35b Mon Sep 17 00:00:00 2001
From: Kate Generalova <kate.generalova@intel.com>
Date: Tue, 8 Jun 2021 12:03:25 +0300
Subject: [PATCH 021/178] samples: update ngraph README (#6070)

---
 .../ngraph_function_creation_sample/README.md    | 16 ++++++++++------
 .../ngraph_function_creation_sample/README.md    |  2 +-
 2 files changed, 11 insertions(+), 7 deletions(-)

diff --git a/inference-engine/ie_bridges/python/sample/ngraph_function_creation_sample/README.md b/inference-engine/ie_bridges/python/sample/ngraph_function_creation_sample/README.md
index c33d67103c6954..f0701f963aea20 100644
--- a/inference-engine/ie_bridges/python/sample/ngraph_function_creation_sample/README.md
+++ b/inference-engine/ie_bridges/python/sample/ngraph_function_creation_sample/README.md
@@ -1,7 +1,8 @@
 # nGraph Function Creation Python* Sample {#openvino_inference_engine_ie_bridges_python_sample_ngraph_function_creation_sample_README}
 
-This sample demonstrates how to execute an inference using [nGraph function feature](../../../../../docs/nGraph_DG/build_function.md) to create a network that uses weights from LeNet classification network. So you don't need an XML file, the model will be created from the source code on the fly.  
-In addition to regular images, the sample also supports single-channel ubyte images as an input.
+This sample demonstrates how to execute an inference using [nGraph function feature](../../../../../docs/nGraph_DG/build_function.md) to create a network that uses weights from LeNet classification network, which is known to work well on digit classification tasks. So you don't need an XML file, the model will be created from the source code on the fly.  
+
+In addition to regular grayscale images with a digit, the sample also supports single-channel `ubyte` images as an input.
 
 The following Inference Engine Python API is used in the application:
 
@@ -14,6 +15,9 @@ Basic Inference Engine API is covered by [Hello Classification Python* Sample](.
 
 | Options                    | Values                                                                  |
 | :------------------------- | :---------------------------------------------------------------------- |
+| Validated Models           | LeNet (image classification network)                                    |
+| Model Format               | Network weights file (\*.bin)                                           |
+| Validated images           | The sample uses OpenCV\* to [read input grayscale image](https://docs.opencv.org/master/d4/da8/group__imgcodecs.html#ga288b8b3da0892bd651fce07b3bbd3a56) (\*.bmp, \*.png) or single-channel `ubyte` image                                          |
 | Supported devices          | [All](../../../../../docs/IE_DG/supported_plugins/Supported_Devices.md) |
 | Other language realization | [C++](../../../../samples/ngraph_function_creation_sample)              |
 
@@ -72,7 +76,7 @@ To run the sample, you need specify a model weights and image:
 You can do inference of an image using a pre-trained model on a GPU using the following command:
 
 ```sh
-python ngraph_function_creation_sample.py -m <path_to_model>/lenet.bin -i <path_to_image>/3.bmp -d GPU
+python ngraph_function_creation_sample.py -m <path_to_model>/lenet.bin -i <path_to_image>/3.png -d GPU
 ```
 
 ## Sample Output
@@ -84,10 +88,10 @@ The sample application logs each step in a standard output stream and outputs to
 [ INFO ] Loading the network using ngraph function with weights from <path_to_model>/lenet.bin
 [ INFO ] Configuring input and output blobs
 [ INFO ] Loading the model to the plugin
-[ WARNING ] <path_to_image>/3.bmp is inverted to white over black
-[ WARNING ] <path_to_image>/3.bmp is resized from (100, 100) to (28, 28)
+[ WARNING ] <path_to_image>/3.png is inverted to white over black
+[ WARNING ] <path_to_image>/3.png is is resized from (351, 353) to (28, 28)
 [ INFO ] Starting inference in synchronous mode
-[ INFO ] Image path: <path_to_image>/3.bmp
+[ INFO ] Image path: <path_to_image>/3.png
 [ INFO ] Top 10 results:
 [ INFO ] classid probability
 [ INFO ] -------------------
diff --git a/inference-engine/samples/ngraph_function_creation_sample/README.md b/inference-engine/samples/ngraph_function_creation_sample/README.md
index 1410241c3a595b..9f7b4f8d4337ee 100644
--- a/inference-engine/samples/ngraph_function_creation_sample/README.md
+++ b/inference-engine/samples/ngraph_function_creation_sample/README.md
@@ -1,6 +1,6 @@
 # nGraph Function Creation C++ Sample {#openvino_inference_engine_samples_ngraph_function_creation_sample_README}
 
-This sample demonstrates how to execute an synchronous inference using [nGraph function feature](../../../docs/nGraph_DG/build_function.md) to create a network, which uses weights from LeNet classification network.
+This sample demonstrates how to execute an synchronous inference using [nGraph function feature](../../../docs/nGraph_DG/build_function.md) to create a network, which uses weights from LeNet classification network, which is known to work well on digit classification tasks.
 
 The sample supports only single-channel `ubyte` images as an input.
 

From dcf36565b0476a5ef45f678238f7359f41df081c Mon Sep 17 00:00:00 2001
From: Shoujiang Ma <shoujiang.ma@intel.com>
Date: Tue, 8 Jun 2021 17:11:58 +0800
Subject: [PATCH 022/178] [AUTO plugin] AUTO plugin will ignore other plugins'
 configuration (#5979)

* AUTO plugin will ignore other plugins' configuration

Signed-off-by: Shoujiang Ma <shoujiang.ma@intel.com>

* Update tests

Signed-off-by: Shoujiang Ma <shoujiang.ma@intel.com>

* Support PER_COUNT config which is needed in benchmark_app

Signed-off-by: Shoujiang Ma <shoujiang.ma@intel.com>

* Address reviewer comments: check config and throw exception for unsupported, but that begin with "AUTO_" will be ignored

Signed-off-by: Shoujiang Ma <shoujiang.ma@intel.com>

* Fix CI tests issue

Signed-off-by: Shoujiang Ma <shoujiang.ma@intel.com>
---
 .../src/auto_plugin/auto_exec_network.cpp     |  8 +-
 .../src/auto_plugin/auto_exec_network.hpp     |  8 +-
 .../src/auto_plugin/auto_infer_request.cpp    | 16 +++-
 .../src/auto_plugin/auto_infer_request.hpp    |  4 +-
 .../src/auto_plugin/auto_plugin.cpp           | 84 ++++++++++---------
 .../src/auto_plugin/auto_plugin.hpp           | 23 +++--
 .../behavior/config.cpp                       | 44 ++--------
 .../behavior/infer_request_config.cpp         | 15 +---
 .../behavior/infer_request_input.cpp          |  4 +-
 .../behavior/infer_request_output.cpp         |  4 +-
 .../behavior/perf_counters.cpp                | 11 ---
 .../behavior/test_plugin.cpp                  |  4 +-
 .../skip_tests_config.cpp                     |  2 +
 .../behavior/config.cpp                       | 11 ++-
 .../behavior/infer_request_input.cpp          |  4 +-
 .../behavior/infer_request_output.cpp         |  3 +-
 .../behavior/perf_counters.cpp                | 22 -----
 .../behavior/test_plugin.cpp                  | 22 ++---
 .../plugin/shared/include/behavior/config.hpp | 29 ++++---
 .../include/behavior/infer_request_config.hpp | 12 ++-
 20 files changed, 140 insertions(+), 190 deletions(-)

diff --git a/inference-engine/src/auto_plugin/auto_exec_network.cpp b/inference-engine/src/auto_plugin/auto_exec_network.cpp
index 353196a88d4a41..49b0963c04d35e 100644
--- a/inference-engine/src/auto_plugin/auto_exec_network.cpp
+++ b/inference-engine/src/auto_plugin/auto_exec_network.cpp
@@ -3,10 +3,8 @@
 //
 
 #include <string>
-#include <vector>
 #include <memory>
 #include <map>
-#include <unordered_map>
 
 #include "ie_metric_helpers.hpp"
 #include "auto_exec_network.hpp"
@@ -15,8 +13,8 @@
 namespace AutoPlugin {
 using namespace InferenceEngine;
 
-AutoExecutableNetwork::AutoExecutableNetwork(const SoExecutableNetworkInternal& network) :
-    _network(network) {
+AutoExecutableNetwork::AutoExecutableNetwork(const SoExecutableNetworkInternal& network, bool enablePerfCount) :
+    _network(network), _enablePerfCount(enablePerfCount) {
 }
 
 AutoExecutableNetwork::~AutoExecutableNetwork() = default;
@@ -24,7 +22,7 @@ AutoExecutableNetwork::~AutoExecutableNetwork() = default;
 InferenceEngine::IInferRequestInternal::Ptr AutoExecutableNetwork::CreateInferRequestImpl(InputsDataMap networkInputs,
                                                                                           OutputsDataMap networkOutputs) {
     SoIInferRequestInternal inferRequest = {_network, _network->CreateInferRequest()};
-    return std::make_shared<AutoInferRequest>(_networkInputs, _networkOutputs, inferRequest);
+    return std::make_shared<AutoInferRequest>(_networkInputs, _networkOutputs, inferRequest, _enablePerfCount);
 }
 
 void AutoExecutableNetwork::Export(std::ostream& networkModel) {
diff --git a/inference-engine/src/auto_plugin/auto_exec_network.hpp b/inference-engine/src/auto_plugin/auto_exec_network.hpp
index a39478b19a753a..e29970711ebc15 100644
--- a/inference-engine/src/auto_plugin/auto_exec_network.hpp
+++ b/inference-engine/src/auto_plugin/auto_exec_network.hpp
@@ -19,16 +19,11 @@ namespace AutoPlugin {
 
 using DeviceName = std::string;
 
-struct DeviceInformation {
-    DeviceName deviceName;
-    std::map<std::string, std::string> config;
-};
-
 class AutoExecutableNetwork : public InferenceEngine::IExecutableNetworkInternal {
 public:
     using Ptr = std::shared_ptr<AutoExecutableNetwork>;
 
-    explicit AutoExecutableNetwork(const InferenceEngine::SoExecutableNetworkInternal& network);
+    explicit AutoExecutableNetwork(const InferenceEngine::SoExecutableNetworkInternal& network, bool enablePerfCount);
 
     void Export(std::ostream& networkModel) override;
     InferenceEngine::RemoteContext::Ptr GetContext() const override;
@@ -43,6 +38,7 @@ class AutoExecutableNetwork : public InferenceEngine::IExecutableNetworkInternal
 
 private:
     InferenceEngine::SoExecutableNetworkInternal _network;
+    bool _enablePerfCount;
 };
 
 }  // namespace AutoPlugin
diff --git a/inference-engine/src/auto_plugin/auto_infer_request.cpp b/inference-engine/src/auto_plugin/auto_infer_request.cpp
index f0777409830d68..46d603187152f7 100644
--- a/inference-engine/src/auto_plugin/auto_infer_request.cpp
+++ b/inference-engine/src/auto_plugin/auto_infer_request.cpp
@@ -11,13 +11,23 @@ namespace AutoPlugin {
 
 AutoInferRequest::AutoInferRequest(const InputsDataMap&              networkInputs,
                                    const OutputsDataMap&             networkOutputs,
-                                   const SoIInferRequestInternal&    inferRequest)
+                                   const SoIInferRequestInternal&    inferRequest,
+                                   bool                              enablePerfCount)
     : IInferRequestInternal(networkInputs, networkOutputs)
-    , _inferRequest(inferRequest) {
+    , _inferRequest(inferRequest)
+    , _enablePerfCount(enablePerfCount) {
 }
 
 std::map<std::string, InferenceEngine::InferenceEngineProfileInfo> AutoInferRequest::GetPerformanceCounts() const {
-    return _inferRequest->GetPerformanceCounts();
+    if (_enablePerfCount) {
+        try {
+            return _inferRequest->GetPerformanceCounts();
+        } catch (...) {
+            return {};
+        }
+    } else {
+        return {};
+    }
 }
 
 void AutoInferRequest::InferImpl() {
diff --git a/inference-engine/src/auto_plugin/auto_infer_request.hpp b/inference-engine/src/auto_plugin/auto_infer_request.hpp
index 1ccaf0093b27b3..c97b2fa5aedd49 100644
--- a/inference-engine/src/auto_plugin/auto_infer_request.hpp
+++ b/inference-engine/src/auto_plugin/auto_infer_request.hpp
@@ -24,7 +24,8 @@ class AutoInferRequest : public InferenceEngine::IInferRequestInternal {
     using Ptr = std::shared_ptr<AutoInferRequest>;
     explicit AutoInferRequest(const InferenceEngine::InputsDataMap&             networkInputs,
                               const InferenceEngine::OutputsDataMap&            networkOutputs,
-                              const InferenceEngine::SoIInferRequestInternal&   inferRequest);
+                              const InferenceEngine::SoIInferRequestInternal&   inferRequest,
+                              bool                                              enablePerfCount);
     std::map<std::string, InferenceEngine::InferenceEngineProfileInfo> GetPerformanceCounts() const override;
     void InferImpl() override;
     void SetBlob(const std::string& name, const InferenceEngine::Blob::Ptr& data) override;
@@ -37,6 +38,7 @@ class AutoInferRequest : public InferenceEngine::IInferRequestInternal {
 
 private:
     InferenceEngine::SoIInferRequestInternal _inferRequest;
+    bool                                     _enablePerfCount;
 };
 
 }  // namespace AutoPlugin
diff --git a/inference-engine/src/auto_plugin/auto_plugin.cpp b/inference-engine/src/auto_plugin/auto_plugin.cpp
index 1fc20063575c5c..274fa9d224f23f 100644
--- a/inference-engine/src/auto_plugin/auto_plugin.cpp
+++ b/inference-engine/src/auto_plugin/auto_plugin.cpp
@@ -75,11 +75,11 @@ IE::QueryNetworkResult AutoInferencePlugin::QueryNetwork(const IE::CNNNetwork& n
     }
 
     auto fullConfig = mergeConfigs(_config, config);
-    auto metaDevices = GetDeviceChoice(fullConfig);
+    auto metaDevices = GetDeviceList(fullConfig);
     std::unordered_set<std::string> supportedLayers;
     for (auto&& value : metaDevices) {
         try {
-            auto deviceQr = GetCore()->QueryNetwork(network, value.deviceName, value.config);
+            auto deviceQr = GetCore()->QueryNetwork(network, value, {});
             std::unordered_set<std::string> deviceSupportedLayers;
             for (auto &&layerQr : deviceQr.supportedLayersMap) {
                 deviceSupportedLayers.emplace(layerQr.first);
@@ -111,7 +111,19 @@ IE::Parameter AutoInferencePlugin::GetConfig(const std::string& name,
 
 void AutoInferencePlugin::SetConfig(const ConfigType& config) {
     for (auto && kvp : config) {
-        _config[kvp.first] = kvp.second;
+        if (kvp.first.find("AUTO_") == 0) {
+            _config[kvp.first] = kvp.second;
+        } else if (kvp.first == IE::PluginConfigParams::KEY_PERF_COUNT) {
+            if (kvp.second == IE::PluginConfigParams::YES ||
+                kvp.second == IE::PluginConfigParams::NO) {
+                _config[kvp.first] = kvp.second;
+            } else {
+                IE_THROW() << "Unsupported config value: " << kvp.second
+                           << " for key: " << kvp.first;
+            }
+        } else {
+            IE_THROW() << "Unsupported config key: " << kvp.first;
+        }
     }
 }
 
@@ -128,7 +140,10 @@ IE::Parameter AutoInferencePlugin::GetMetric(const std::string& name,
         std::string device_name = {"Inference Engine AUTO device"};
         IE_SET_METRIC_RETURN(FULL_DEVICE_NAME, device_name);
     } else if (name == METRIC_KEY(SUPPORTED_CONFIG_KEYS)) {
-        std::vector<std::string> configKeys;
+        std::vector<std::string> configKeys = {
+            IE::KEY_AUTO_DEVICE_LIST,
+            IE::PluginConfigParams::KEY_PERF_COUNT
+        };
         IE_SET_METRIC_RETURN(SUPPORTED_CONFIG_KEYS, configKeys);
     } else if (name == METRIC_KEY(OPTIMIZATION_CAPABILITIES)) {
         std::vector<std::string> capabilities = GetOptimizationCapabilities(options);
@@ -139,42 +154,21 @@ IE::Parameter AutoInferencePlugin::GetMetric(const std::string& name,
 }
 
 //////////////////////////////////// private & protected functions ///////////////////
-std::vector<AutoPlugin::DeviceInformation> AutoInferencePlugin::GetDeviceChoice(const ConfigType&  config) const {
-    std::vector<DeviceInformation> metaDevices;
-    std::vector<std::string> availableDevices;
+std::vector<DeviceName> AutoInferencePlugin::GetDeviceList(const ConfigType& config) const {
+    std::vector<DeviceName> deviceList;
 
     auto deviceListConfig = config.find(IE::KEY_AUTO_DEVICE_LIST);
     if (deviceListConfig == config.end()) {
-        availableDevices = GetCore()->GetAvailableDevices();
+        deviceList = GetCore()->GetAvailableDevices();
     } else {
-        availableDevices = IE::DeviceIDParser::getHeteroDevices(deviceListConfig->second);
+        deviceList = IE::DeviceIDParser::getHeteroDevices(deviceListConfig->second);
     }
 
-    auto getDeviceConfig = [&] (const DeviceName & deviceWithID) {
-        IE::DeviceIDParser deviceParser(deviceWithID);
-        std::string deviceName = deviceParser.getDeviceName();
-        ConfigType tconfig = config;
-
-        // set device ID if any
-        std::string deviceIDLocal = deviceParser.getDeviceID();
-        if (!deviceIDLocal.empty()) {
-            tconfig[IE::PluginConfigParams::KEY_DEVICE_ID] = deviceIDLocal;
-        }
-
-        return GetSupportedConfig(tconfig, deviceName);
-    };
-
-    for (auto && d : availableDevices) {
-        if (d != _pluginName) {
-            metaDevices.push_back({ d, getDeviceConfig(d)});
-        }
-    }
-
-    if (metaDevices.empty()) {
+    if (deviceList.empty()) {
         IE_THROW() << "Please, check environment due to no supported devices can be used";
     }
 
-    return metaDevices;
+    return deviceList;
 }
 
 std::vector<std::string> AutoInferencePlugin::GetOptimizationCapabilities(const std::map<std::string, IE::Parameter> & options) const {
@@ -215,7 +209,21 @@ ConfigType AutoInferencePlugin::GetSupportedConfig(const ConfigType&  config,
     return supportedConfig;
 }
 
-DeviceInformation AutoInferencePlugin::SelectDevice(const std::vector<DeviceInformation>& metaDevices, const std::string& networkPrecision) {
+void AutoInferencePlugin::CheckConfig(const ConfigType& config) {
+    std::vector<std::string> supportedConfigKeys = GetMetric(METRIC_KEY(SUPPORTED_CONFIG_KEYS), {});
+    for (auto&& c : config) {
+        auto itKey = std::find(supportedConfigKeys.begin(), supportedConfigKeys.end(), c.first);
+        if (supportedConfigKeys.end() == itKey) {
+            // CVS-57233
+            if (c.first.find("AUTO_") == 0) {
+                continue;
+            }
+            IE_THROW() << "AUTO plugin doesn't support config key " << c.first;
+        }
+    }
+}
+
+DeviceName AutoInferencePlugin::SelectDevice(const std::vector<DeviceName>& metaDevices, const std::string& networkPrecision) {
     if (metaDevices.empty()) {
         IE_THROW(NotFound) << "No available device to select in AUTO plugin";
     }
@@ -223,15 +231,15 @@ DeviceInformation AutoInferencePlugin::SelectDevice(const std::vector<DeviceInfo
         return metaDevices.at(0);
     }
 
-    std::vector<DeviceInformation> CPU;
-    std::vector<DeviceInformation> GPU;
+    std::vector<DeviceName> CPU;
+    std::vector<DeviceName> GPU;
 
     for (auto& item : metaDevices) {
-        if (item.deviceName.find("CPU") == 0) {
+        if (item.find("CPU") == 0) {
             CPU.push_back(item);
             continue;
         }
-        if (item.deviceName.find("GPU") == 0) {
+        if (item.find("GPU") == 0) {
             GPU.push_back(item);
             continue;
         }
@@ -242,10 +250,10 @@ DeviceInformation AutoInferencePlugin::SelectDevice(const std::vector<DeviceInfo
     }
 
     // Sort GPU by name: GPU.2 > GPU.1 > GPU.0 > GPU, so we always choose the GPU[0] as best device
-    std::sort(GPU.begin(), GPU.end(), [](const DeviceInformation& a, const DeviceInformation& b)->bool{return b.deviceName < a.deviceName;});
+    std::sort(GPU.begin(), GPU.end(), [](const DeviceName& a, const DeviceName& b)->bool{return b < a;});
 
     for (auto&& item : GPU) {
-        std::vector<std::string> capability = GetCore()->GetMetric(item.deviceName, METRIC_KEY(OPTIMIZATION_CAPABILITIES));
+        std::vector<std::string> capability = GetCore()->GetMetric(item, METRIC_KEY(OPTIMIZATION_CAPABILITIES));
         auto res = std::find(capability.begin(), capability.end(), networkPrecision);
         if (res != capability.end()) {
             return item;
diff --git a/inference-engine/src/auto_plugin/auto_plugin.hpp b/inference-engine/src/auto_plugin/auto_plugin.hpp
index af42e9f0ef7c28..858ee2143fd06f 100644
--- a/inference-engine/src/auto_plugin/auto_plugin.hpp
+++ b/inference-engine/src/auto_plugin/auto_plugin.hpp
@@ -30,10 +30,11 @@ class AutoInferencePlugin : public IE::IInferencePlugin {
     void SetConfig(const ConfigType& config) override;
 
 private:
-    std::vector<AutoPlugin::DeviceInformation> GetDeviceChoice(const ConfigType&  config) const;
+    std::vector<DeviceName> GetDeviceList(const ConfigType&  config) const;
     std::vector<std::string> GetOptimizationCapabilities(const std::map<std::string, IE::Parameter>& options) const;
-    DeviceInformation SelectDevice(const std::vector<DeviceInformation>& metaDevices, const std::string& networkPrecision = METRIC_VALUE(FP32));
-    ConfigType GetSupportedConfig(const ConfigType& config, const AutoPlugin::DeviceName & deviceName) const;
+    DeviceName SelectDevice(const std::vector<DeviceName>& metaDevices, const std::string& networkPrecision = METRIC_VALUE(FP32));
+    ConfigType GetSupportedConfig(const ConfigType& config, const DeviceName & deviceName) const;
+    void CheckConfig(const ConfigType& config);
     static ConfigType mergeConfigs(ConfigType config, const ConfigType& local);
 
     template <typename T>
@@ -41,18 +42,21 @@ class AutoInferencePlugin : public IE::IInferencePlugin {
         if (GetCore() == nullptr) {
             IE_THROW() << "Please, work with AUTO device via InferencEngine::Core object";
         }
+
+        CheckConfig(config);
+
         auto fullConfig = mergeConfigs(_config, config);
-        auto metaDevices = GetDeviceChoice(fullConfig);
-        DeviceInformation selectedDevice;
+        auto metaDevices = GetDeviceList(fullConfig);
+        DeviceName selectedDevice;
         IE::SoExecutableNetworkInternal executableNetwork;
         while (!metaDevices.empty()) {
             selectedDevice = SelectDevice(metaDevices, networkPrecision);
             try {
-                executableNetwork = GetCore()->LoadNetwork(param, selectedDevice.deviceName, selectedDevice.config);
+                executableNetwork = GetCore()->LoadNetwork(param, selectedDevice, {});
                 break;
             } catch (...) {
                 auto eraseDevice = std::find_if(metaDevices.begin(), metaDevices.end(),
-                    [=](const DeviceInformation& d)->bool{return d.deviceName == selectedDevice.deviceName;});
+                    [=](const DeviceName& d)->bool{return d == selectedDevice;});
                 if (eraseDevice == metaDevices.end()) {
                     IE_THROW() << "Didn't find the selected device name";
                 }
@@ -63,7 +67,10 @@ class AutoInferencePlugin : public IE::IInferencePlugin {
         if (!executableNetwork) {
             IE_THROW() << "Failed to load network by AUTO plugin";
         }
-        auto impl = std::make_shared<AutoExecutableNetwork>(executableNetwork);
+
+        bool enablePerfCount = fullConfig.find(IE::PluginConfigParams::KEY_PERF_COUNT) != fullConfig.end();
+
+        auto impl = std::make_shared<AutoExecutableNetwork>(executableNetwork, enablePerfCount);
 
         if (std::is_same<std::string, T>::value) {
             SetExeNetworkInfo(impl, executableNetwork->GetInputsInfo(),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/config.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/config.cpp
index 8dab56fdafa52a..e3f0adb1cb6117 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/config.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/config.cpp
@@ -42,18 +42,7 @@ namespace {
     };
 
     const std::vector<std::map<std::string, std::string>> AutoConfigs = {
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-                    {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::CPU_THROUGHPUT_AUTO}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-                    {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::CPU_THROUGHPUT_NUMA}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-                    {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS, "8"}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-                    {InferenceEngine::PluginConfigParams::KEY_CPU_BIND_THREAD, InferenceEngine::PluginConfigParams::NO}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-                    {InferenceEngine::PluginConfigParams::KEY_CPU_BIND_THREAD, InferenceEngine::PluginConfigParams::YES}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-                    {InferenceEngine::PluginConfigParams::KEY_DYN_BATCH_LIMIT, "10"}}
+            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
     };
 
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigTests,
@@ -93,22 +82,14 @@ namespace {
     };
 
     const std::vector<std::map<std::string, std::string>> autoinconfigs = {
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-                    {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS, "OFF"}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-                    {InferenceEngine::PluginConfigParams::KEY_CPU_BIND_THREAD, "OFF"}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-                    {InferenceEngine::PluginConfigParams::KEY_DYN_BATCH_LIMIT, "NAN"}}
+        {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
+            {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS, "OFF"}}
     };
 
     const std::vector<std::map<std::string, std::string>> multiconf = {
             {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES , CommonTestUtils::DEVICE_CPU}}
     };
 
-    const std::vector<std::map<std::string, std::string>> autoconf = {
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
-    };
-
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigAPITests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
@@ -127,7 +108,7 @@ namespace {
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-            ::testing::ValuesIn(autoconf)),
+            ::testing::ValuesIn(AutoConfigs)),
             CorrectConfigAPITests::getTestCaseName);
 
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigTests,
@@ -144,13 +125,6 @@ namespace {
             ::testing::ValuesIn(multiinconfigs)),
             IncorrectConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, IncorrectConfigTests,
-            ::testing::Combine(
-            ::testing::ValuesIn(netPrecisions),
-            ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-            ::testing::ValuesIn(autoinconfigs)),
-            IncorrectConfigTests::getTestCaseName);
-
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
@@ -166,10 +140,10 @@ namespace {
             IncorrectConfigAPITests::getTestCaseName);
 
     INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, IncorrectConfigAPITests,
-            ::testing::Combine(
-            ::testing::ValuesIn(netPrecisions),
-            ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-            ::testing::ValuesIn(autoinconfigs)),
-            IncorrectConfigAPITests::getTestCaseName);
+                            ::testing::Combine(
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(CommonTestUtils::DEVICE_AUTO),
+                                ::testing::ValuesIn(autoinconfigs)),
+                            IncorrectConfigAPITests::getTestCaseName);
 
 } // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_config.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_config.cpp
index 80914183c39e80..3db325234cf1a6 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_config.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_config.cpp
@@ -51,20 +51,7 @@ namespace {
     };
 
     const std::vector<std::map<std::string, std::string>> AutoConfigs = {
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-             {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS,
-              InferenceEngine::PluginConfigParams::CPU_THROUGHPUT_AUTO}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-             {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS,
-              InferenceEngine::PluginConfigParams::CPU_THROUGHPUT_NUMA}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-             {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS, "8"}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-             {InferenceEngine::PluginConfigParams::KEY_CPU_BIND_THREAD, InferenceEngine::PluginConfigParams::NO}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-             {InferenceEngine::PluginConfigParams::KEY_CPU_BIND_THREAD, InferenceEngine::PluginConfigParams::YES}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-             {InferenceEngine::PluginConfigParams::KEY_DYN_BATCH_LIMIT, "10"}}
+            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
     };
 
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferConfigTests,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_input.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_input.cpp
index 771794774a67c9..eb988c0043a289 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_input.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_input.cpp
@@ -26,9 +26,7 @@ namespace {
     };
 
     const std::vector<std::map<std::string, std::string>> autoConfigs = {
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-              {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::CPU_THROUGHPUT_AUTO}}
+            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
     };
 
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestInputTests,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_output.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_output.cpp
index 78be15133142a3..a77253f8900fa5 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_output.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_output.cpp
@@ -22,9 +22,7 @@ namespace {
     };
 
     const std::vector<std::map<std::string, std::string>> autoConfigs = {
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU},
-             {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::CPU_THROUGHPUT_AUTO}}
+            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
     };
 
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestOutputTests,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/perf_counters.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/perf_counters.cpp
index 2e2aab976c6732..8bf1a8f95ba766 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/perf_counters.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/perf_counters.cpp
@@ -14,10 +14,6 @@ namespace {
             {{ MULTI_CONFIG_KEY(DEVICE_PRIORITIES) , CommonTestUtils::DEVICE_CPU}}
     };
 
-    const std::vector<std::map<std::string, std::string>> Autoconfigs = {
-            {{ AUTO_CONFIG_KEY(DEVICE_LIST) , CommonTestUtils::DEVICE_CPU}}
-    };
-
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PerfCountersTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
@@ -32,11 +28,4 @@ namespace {
                                     ::testing::ValuesIn(Multiconfigs)),
                             PerfCountersTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, PerfCountersTest,
-                            ::testing::Combine(
-                                    ::testing::Values(InferenceEngine::Precision::FP32),
-                                    ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-                                    ::testing::ValuesIn(Autoconfigs)),
-                            PerfCountersTest::getTestCaseName);
-
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/test_plugin.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/test_plugin.cpp
index 95208f0a092c88..da76bc26498b1e 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/test_plugin.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/test_plugin.cpp
@@ -37,9 +37,7 @@ namespace {
     };
 
     const std::vector<std::map<std::string, std::string>> AutoConfigsInputOutput = {
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST, CommonTestUtils::DEVICE_CPU}},
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST, CommonTestUtils::DEVICE_CPU},
-             {InferenceEngine::PluginConfigParams::KEY_CPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::CPU_THROUGHPUT_AUTO}}
+            {{InferenceEngine::KEY_AUTO_DEVICE_LIST, CommonTestUtils::DEVICE_CPU}}
     };
 
     const std::vector<std::map<std::string, std::string>> configsOutput = {
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
index b3134e9953e9e5..f013e544074644 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
@@ -67,6 +67,8 @@ std::vector<std::string> disabledTestPatterns() {
 
         // TODO: 55656 AUTO plugin and QueryNetwork
         R"(.*CoreThreading.*smoke_QueryNetwork.*targetDevice=AUTO_config.*)",
+        // Unsupported config KEY_ENFORCE_BF16 for AUTO plugin
+        R"(.*smoke_SetBlobOfKindAUTO.*SetBlobOfKindTest.CompareWithRefs.*)",
         // reference doesn't cover I8, U8 cases. Issue: 55842
         R"(.*Gather7LayerTest.*netPRC=I8.*)",
     };
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp
index 0b550c568b64e1..e21d610db569cb 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp
@@ -106,6 +106,13 @@ namespace {
                     ::testing::ValuesIn(autoconf)),
             CorrectConfigAPITests::getTestCaseName);
 
+    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, CorrectConfigAPITests,
+                            ::testing::Combine(
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(CommonTestUtils::DEVICE_AUTO),
+                                ::testing::ValuesIn(auto_cpu_gpu_conf)),
+                            CorrectConfigAPITests::getTestCaseName);
+
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
             ::testing::Combine(
                     ::testing::ValuesIn(netPrecisions),
@@ -124,14 +131,14 @@ namespace {
             ::testing::Combine(
                     ::testing::ValuesIn(netPrecisions),
                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-                    ::testing::ValuesIn(autoconf)),
+                    ::testing::ValuesIn(autoinconfigs)),
             IncorrectConfigAPITests::getTestCaseName);
 
     INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, IncorrectConfigAPITests,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-                                ::testing::ValuesIn(auto_cpu_gpu_conf)),
+                                ::testing::ValuesIn(autoinconfigs)),
                             IncorrectConfigAPITests::getTestCaseName);
 
 
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp
index a9a07450f70ceb..e15ea827caa814 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp
@@ -26,9 +26,7 @@ namespace {
     };
 
     const std::vector<std::map<std::string, std::string>> autoConfigs = {
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_GPU},
-                {InferenceEngine::PluginConfigParams::KEY_GPU_THROUGHPUT_STREAMS,
-                    InferenceEngine::PluginConfigParams::GPU_THROUGHPUT_AUTO}}
+            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_GPU}}
     };
 
     const std::vector<std::map<std::string, std::string>> auto_cpu_gpu_conf = {
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp
index 550572077c34ba..6c38f5c841c0b0 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp
@@ -22,8 +22,7 @@ namespace {
     };
 
     const std::vector<std::map<std::string, std::string>> autoConfigs = {
-            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_GPU},
-                {InferenceEngine::PluginConfigParams::KEY_GPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::GPU_THROUGHPUT_AUTO}}
+            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_GPU}}
     };
 
     const std::vector<std::map<std::string, std::string>> auto_cpu_gpu_conf = {
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/perf_counters.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/perf_counters.cpp
index c5a5f695359496..d8a89ef317d425 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/perf_counters.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/perf_counters.cpp
@@ -14,14 +14,6 @@ namespace {
             {{ MULTI_CONFIG_KEY(DEVICE_PRIORITIES) , CommonTestUtils::DEVICE_GPU}}
     };
 
-    const std::vector<std::map<std::string, std::string>> Autoconfigs = {
-            {{ AUTO_CONFIG_KEY(DEVICE_LIST) , CommonTestUtils::DEVICE_GPU}}
-    };
-
-    const std::vector<std::map<std::string, std::string>> auto_cpu_gpu_conf = {
-        {{InferenceEngine::KEY_AUTO_DEVICE_LIST , std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}}
-    };
-
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PerfCountersTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
@@ -36,18 +28,4 @@ namespace {
                                     ::testing::ValuesIn(Multiconfigs)),
                             PerfCountersTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, PerfCountersTest,
-                            ::testing::Combine(
-                                    ::testing::Values(InferenceEngine::Precision::FP32),
-                                    ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-                                    ::testing::ValuesIn(Autoconfigs)),
-                            PerfCountersTest::getTestCaseName);
-
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, PerfCountersTest,
-                            ::testing::Combine(
-                                ::testing::Values(InferenceEngine::Precision::FP32),
-                                ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-                                ::testing::ValuesIn(auto_cpu_gpu_conf)),
-                            PerfCountersTest::getTestCaseName);
-
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp
index bc6507d7905fe4..51979116646939 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp
@@ -28,7 +28,7 @@ namespace {
     };
 
     const std::vector<std::map<std::string, std::string>> auto_cpu_gpu_conf = {
-        {{InferenceEngine::KEY_AUTO_DEVICE_LIST , std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}}
+            {{InferenceEngine::KEY_AUTO_DEVICE_LIST , std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}}
     };
 
     const std::vector<std::map<std::string, std::string>> configsInput = {
@@ -42,18 +42,6 @@ namespace {
              {InferenceEngine::PluginConfigParams::KEY_GPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::GPU_THROUGHPUT_AUTO}}
     };
 
-    const std::vector<std::map<std::string, std::string>> AutoConfigsInputOutput = {
-        {{InferenceEngine::KEY_AUTO_DEVICE_LIST, CommonTestUtils::DEVICE_GPU}},
-        {{InferenceEngine::KEY_AUTO_DEVICE_LIST, CommonTestUtils::DEVICE_GPU},
-            {InferenceEngine::PluginConfigParams::KEY_GPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::GPU_THROUGHPUT_AUTO}}
-    };
-
-    const std::vector<std::map<std::string, std::string>> AutoCGConfigsInputOutput = {
-        {{InferenceEngine::KEY_AUTO_DEVICE_LIST, std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}},
-        {{InferenceEngine::KEY_AUTO_DEVICE_LIST, std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU},
-            {InferenceEngine::PluginConfigParams::KEY_GPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::GPU_THROUGHPUT_AUTO}}
-    };
-
     const std::vector<std::map<std::string, std::string>> configsOutput = {
             {},
             {{InferenceEngine::PluginConfigParams::KEY_GPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::GPU_THROUGHPUT_AUTO}}
@@ -77,14 +65,14 @@ namespace {
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-                                    ::testing::ValuesIn(AutoConfigsInputOutput)),
+                                    ::testing::ValuesIn(AutoConfigs)),
                             BehaviorTestOutput::getTestCaseName);
 
     INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-                                ::testing::ValuesIn(AutoCGConfigsInputOutput)),
+                                ::testing::ValuesIn(auto_cpu_gpu_conf)),
                             BehaviorTestOutput::getTestCaseName);
 
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTests,
@@ -133,14 +121,14 @@ namespace {
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-                                    ::testing::ValuesIn(AutoConfigsInputOutput)),
+                                    ::testing::ValuesIn(AutoConfigs)),
                             BehaviorTestInput::getTestCaseName);
 
     INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
-                                ::testing::ValuesIn(AutoCGConfigsInputOutput)),
+                                ::testing::ValuesIn(auto_cpu_gpu_conf)),
                             BehaviorTestInput::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/shared/include/behavior/config.hpp b/inference-engine/tests/functional/plugin/shared/include/behavior/config.hpp
index b9caf6edacfddb..e13fe679b2a82c 100644
--- a/inference-engine/tests/functional/plugin/shared/include/behavior/config.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/behavior/config.hpp
@@ -57,8 +57,7 @@ namespace BehaviorTestsDefinitions {
         // Create CNNNetwork from ngrpah::Function
         InferenceEngine::CNNNetwork cnnNet(function);
         if (targetDevice.find(CommonTestUtils::DEVICE_MULTI) == std::string::npos &&
-            targetDevice.find(CommonTestUtils::DEVICE_HETERO) == std::string::npos &&
-            targetDevice.find(CommonTestUtils::DEVICE_AUTO) == std::string::npos) {
+            targetDevice.find(CommonTestUtils::DEVICE_HETERO) == std::string::npos) {
             ASSERT_NO_THROW(ie->GetMetric(targetDevice, METRIC_KEY(SUPPORTED_CONFIG_KEYS)));
             ASSERT_THROW(ie->SetConfig(configuration, targetDevice),
                          InferenceEngine::Exception);
@@ -73,8 +72,12 @@ namespace BehaviorTestsDefinitions {
         SKIP_IF_CURRENT_TEST_IS_DISABLED()
         // Create CNNNetwork from ngrpah::Function
         InferenceEngine::CNNNetwork cnnNet(function);
-        ASSERT_THROW(auto execNet = ie->LoadNetwork(cnnNet, targetDevice, configuration),
-                     InferenceEngine::Exception);
+        if (targetDevice.find(CommonTestUtils::DEVICE_AUTO) != std::string::npos) {
+            GTEST_SKIP();
+        } else {
+            ASSERT_THROW(auto execNet = ie->LoadNetwork(cnnNet, targetDevice, configuration),
+                         InferenceEngine::Exception);
+        }
     }
 
     using IncorrectConfigAPITests = BehaviorTestsUtils::BehaviorTestsBasic;
@@ -110,8 +113,10 @@ namespace BehaviorTestsDefinitions {
             ASSERT_NO_THROW(ie->SetConfig(config, targetDevice));
         }
         // Load CNNNetwork to target plugins
-        auto execNet = ie->LoadNetwork(cnnNet, targetDevice, config);
-        execNet.CreateInferRequest();
+        if (targetDevice.find(CommonTestUtils::DEVICE_AUTO) == std::string::npos) {
+            auto execNet = ie->LoadNetwork(cnnNet, targetDevice, config);
+            execNet.CreateInferRequest();
+        }
 
         if ((targetDevice == CommonTestUtils::DEVICE_HDDL) || (targetDevice == CommonTestUtils::DEVICE_GNA)) {
             ASSERT_EQ(0u, InferenceEngine::ExecutorManager::getInstance()->getExecutorsNumber());
@@ -139,8 +144,10 @@ namespace BehaviorTestsDefinitions {
             ASSERT_NO_THROW(ie->SetConfig(config, targetDevice));
         }
         // Load CNNNetwork to target plugins
-        auto execNet = ie->LoadNetwork(cnnNet, targetDevice, config);
-        execNet.CreateInferRequest();
+        if (targetDevice.find(CommonTestUtils::DEVICE_AUTO) == std::string::npos) {
+            auto execNet = ie->LoadNetwork(cnnNet, targetDevice, config);
+            execNet.CreateInferRequest();
+        }
 
         if ((targetDevice == CommonTestUtils::DEVICE_MYRIAD) ||
             (targetDevice == CommonTestUtils::DEVICE_KEEMBAY)) {
@@ -170,8 +177,10 @@ namespace BehaviorTestsDefinitions {
                 ASSERT_NO_THROW(ie->SetConfig(config, targetDevice));
             }
             // Load CNNNetwork to target plugins
-            auto execNet = ie->LoadNetwork(cnnNet, targetDevice, config);
-            execNet.CreateInferRequest();
+            if (targetDevice.find(CommonTestUtils::DEVICE_AUTO) == std::string::npos) {
+                auto execNet = ie->LoadNetwork(cnnNet, targetDevice, config);
+                execNet.CreateInferRequest();
+            }
 
             if ((targetDevice == CommonTestUtils::DEVICE_MYRIAD) ||
                 (targetDevice == CommonTestUtils::DEVICE_KEEMBAY)) {
diff --git a/inference-engine/tests/functional/plugin/shared/include/behavior/infer_request_config.hpp b/inference-engine/tests/functional/plugin/shared/include/behavior/infer_request_config.hpp
index cb364c80f8cabf..c94694011238ab 100644
--- a/inference-engine/tests/functional/plugin/shared/include/behavior/infer_request_config.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/behavior/infer_request_config.hpp
@@ -42,8 +42,10 @@ TEST_P(InferConfigTests, canSetExclusiveAsyncRequests) {
         ASSERT_NO_THROW(ie->SetConfig(config, targetDevice));
     }
     // Load CNNNetwork to target plugins
-    auto execNet = ie->LoadNetwork(cnnNet, targetDevice, config);
-    execNet.CreateInferRequest();
+    if (targetDevice.find(CommonTestUtils::DEVICE_AUTO) == std::string::npos) {
+        auto execNet = ie->LoadNetwork(cnnNet, targetDevice, config);
+        execNet.CreateInferRequest();
+    }
 
     if ((targetDevice == CommonTestUtils::DEVICE_HDDL) || (targetDevice == CommonTestUtils::DEVICE_GNA)) {
         ASSERT_EQ(0u, InferenceEngine::ExecutorManager::getInstance()->getExecutorsNumber());
@@ -71,8 +73,10 @@ TEST_P(InferConfigTests, withoutExclusiveAsyncRequests) {
         ASSERT_NO_THROW(ie->SetConfig(config, targetDevice));
     }
     // Load CNNNetwork to target plugins
-    auto execNet = ie->LoadNetwork(cnnNet, targetDevice, config);
-    execNet.CreateInferRequest();
+    if (targetDevice.find(CommonTestUtils::DEVICE_AUTO) == std::string::npos) {
+        auto execNet = ie->LoadNetwork(cnnNet, targetDevice, config);
+        execNet.CreateInferRequest();
+    }
 
     if ((targetDevice == CommonTestUtils::DEVICE_GNA) || (targetDevice == CommonTestUtils::DEVICE_HDDL)) {
         ASSERT_EQ(0u, InferenceEngine::ExecutorManager::getInstance()->getExecutorsNumber());

From 79e44292c5758b9e736c8440e7c0e9f2364991dc Mon Sep 17 00:00:00 2001
From: Michal Papaj <michal.papaj@intel.com>
Date: Tue, 8 Jun 2021 11:35:26 +0200
Subject: [PATCH 023/178] Update to speech recognition demo version. (#6060)

The speech recognition demo packages were updated to align
OV python packages requirements. This patch updates packages
version and SHA.
---
 inference-engine/cmake/dependencies.cmake | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/inference-engine/cmake/dependencies.cmake b/inference-engine/cmake/dependencies.cmake
index 4ce1ef3136550e..b270c46f2da7cc 100644
--- a/inference-engine/cmake/dependencies.cmake
+++ b/inference-engine/cmake/dependencies.cmake
@@ -295,25 +295,25 @@ if (ENABLE_SPEECH_DEMO)
     if(DEFINED IE_PATH_TO_DEPS)
         if (WIN32 AND X86_64)
             RESOLVE_DEPENDENCY(SPEECH_LIBS_AND_DEMOS
-                    ARCHIVE_WIN "speech_demo_1.0.0.755_windows.zip"
+                    ARCHIVE_WIN "speech_demo_1.0.0.774_windows.zip"
                     VERSION_REGEX ".*_([0-9]+.[0-9]+.[0-9]+.[0-9]+).*"
-                    TARGET_PATH "${TEMP}/speech_demo_1.0.0.755"
-                    SHA256 "58adef14b8a749f70fa83888614cee34b941956e6e958e445e3f48885b3c20a0")
+                    TARGET_PATH "${TEMP}/speech_demo_1.0.0.774"
+                    SHA256 "67b25170be5e89a4f0e90e8b39623b60c9a15b965c30329385e295fcd2edc856")
             debug_message(STATUS "speech_libs_and_demos=" ${SPEECH_LIBS_AND_DEMOS})
         elseif (LINUX AND X86_64)
             if (LINUX_OS_NAME STREQUAL "CentOS 7" OR CMAKE_CXX_COMPILER_VERSION VERSION_LESS "4.9")
                 RESOLVE_DEPENDENCY(SPEECH_LIBS_AND_DEMOS
-                    ARCHIVE_LIN "speech_demo_1.0.0.755_centos.tgz"
+                    ARCHIVE_LIN "speech_demo_1.0.0.774_centos.tgz"
                     VERSION_REGEX ".*_([0-9]+.[0-9]+.[0-9]+.[0-9]+).*"
-                    TARGET_PATH "${TEMP}/speech_demo_1.0.0.755"
-                    SHA256 "716201e377714ac50f3909c445d36d47a089de50a557d8ef65232de040671188")
+                    TARGET_PATH "${TEMP}/speech_demo_1.0.0.774"
+                    SHA256 "5ec3b7be9ae05376aefae5bd5fd4a39b12c274e82817fd3218120b8e8fc8ff5a")
                 debug_message(STATUS "speech_libs_and_demos=" ${SPEECH_LIBS_AND_DEMOS})
             else()
                 RESOLVE_DEPENDENCY(SPEECH_LIBS_AND_DEMOS
-                    ARCHIVE_LIN "speech_demo_1.0.0.755_linux.tgz"
+                    ARCHIVE_LIN "speech_demo_1.0.0.774_linux.tgz"
                     VERSION_REGEX ".*_([0-9]+.[0-9]+.[0-9]+.[0-9]+).*"
-                    TARGET_PATH "${TEMP}/speech_demo_1.0.0.755"
-                    SHA256 "7714b8776ec0183ed73eed6d3d965ee6d5c15d2dc49ee5ae118cc368c89c7a9d")
+                    TARGET_PATH "${TEMP}/speech_demo_1.0.0.774"
+                    SHA256 "f0bbd0a6218b0365e7cfb1f860b34e4ace7e0d47dd60b369cdea8a480329810f")
                 debug_message(STATUS "speech_libs_and_demos=" ${SPEECH_LIBS_AND_DEMOS})
             endif()
         else()

From 9aa338a0afcab0f1754e9cc1ec2e08d92adfdf5a Mon Sep 17 00:00:00 2001
From: Rafal Blaczkowski <rafal.blaczkowski@intel.com>
Date: Tue, 8 Jun 2021 11:37:20 +0200
Subject: [PATCH 024/178] OpenVINO ONNX CI - set more stable proxy (#5945)

* Set more stable proxy

* Update env

* Add missing chars

* Update proxy

* Additonal try to overwrite proxy

* update proxy

* fix style
---
 .ci/openvino-onnx/Jenkinsfile | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.ci/openvino-onnx/Jenkinsfile b/.ci/openvino-onnx/Jenkinsfile
index 48529879ef162e..5fe24928798f3d 100644
--- a/.ci/openvino-onnx/Jenkinsfile
+++ b/.ci/openvino-onnx/Jenkinsfile
@@ -113,8 +113,8 @@ def buildDockerImage(Map configuration, String workdir) {
         --build-arg BUILD_TYPE=${configuration.build_type} \
         --build-arg PROTOBUF_LITE=${configuration.protobuf_lite} \
         --file=.ci/openvino-onnx/Dockerfile \
-        --build-arg http_proxy=http://proxy-chain.intel.com:911/ \
-        --build-arg https_proxy=http://proxy-chain.intel.com:912/ .
+        --build-arg http_proxy=http://proxy-ir.intel.com:911/ \
+        --build-arg https_proxy=http://proxy-ir.intel.com:911/ .
     """
 }
 

From 503d18c80f9f1291b5aa5a7456b074b30bac7fec Mon Sep 17 00:00:00 2001
From: Alexander Shchepetov <alexander.shchepetov@intel.com>
Date: Tue, 8 Jun 2021 12:47:20 +0300
Subject: [PATCH 025/178] Enable MYRIAD in stress tests (#6026)

---
 .../myriad_references_config.xml              |  20 +++
 .../nightly_configs/myriad_test_config.xml    | 115 ++++++++++++++++++
 .../nightly_configs/myriad_test_config.xml    |  19 +++
 .../nightly_configs/myriad_test_config.xml    |  20 +++
 4 files changed, 174 insertions(+)
 create mode 100644 tests/stress_tests/.automation/memcheck_tests/nightly_configs/myriad_references_config.xml
 create mode 100644 tests/stress_tests/.automation/memcheck_tests/nightly_configs/myriad_test_config.xml
 create mode 100644 tests/stress_tests/.automation/memleaks_tests/nightly_configs/myriad_test_config.xml
 create mode 100644 tests/stress_tests/.automation/unittests/nightly_configs/myriad_test_config.xml

diff --git a/tests/stress_tests/.automation/memcheck_tests/nightly_configs/myriad_references_config.xml b/tests/stress_tests/.automation/memcheck_tests/nightly_configs/myriad_references_config.xml
new file mode 100644
index 00000000000000..b4a8be108c386d
--- /dev/null
+++ b/tests/stress_tests/.automation/memcheck_tests/nightly_configs/myriad_references_config.xml
@@ -0,0 +1,20 @@
+<?xml version="1.0"?>
+<attributes>
+    <models>
+        # References were collected from DB with next query: {"commit_id": "aa2ae13c1ee6d700dd287ab809403e7de8c7c5e3", "commit_date": "2021-06-05 14:30:47+00:00"}
+        # and modified on FACTOR = 1.3
+        <!--Models with FP32 precision-->
+        <model path="public/alexnet/FP32/alexnet.xml" precision="FP32" test="create_exenetwork" device="MYRIAD" vmsize="2219614" vmpeak="2807734" vmrss="522615" vmhwm="959597" />
+        <model path="public/efficientnet-b0/FP32/efficientnet-b0.xml" precision="FP32" test="infer_request_inference" device="MYRIAD" vmsize="1939886" vmpeak="1939886" vmrss="138392" vmhwm="145381" />
+        <model path="public/mobilenet-v2-1.4-224/FP32/mobilenet-v2-1.4-224.xml" precision="FP32" test="create_exenetwork" device="MYRIAD" vmsize="1750569" vmpeak="1771416" vmrss="131944" vmhwm="144471" />
+        <model path="public/se-resnet-50/FP32/se-resnet-50.xml" precision="FP32" test="infer_request_inference" device="MYRIAD" vmsize="2239920" vmpeak="2325117" vmrss="434096" vmhwm="471562" />
+        <model path="public/Sphereface/FP32/Sphereface.xml" precision="FP32" test="create_exenetwork" device="MYRIAD" vmsize="1890917" vmpeak="2115952" vmrss="192202" vmhwm="362102" />
+        <model path="public/yolo-v1-tiny-tf/FP32/yolo-v1-tiny-tf.xml" precision="FP32" test="infer_request_inference" device="MYRIAD" vmsize="1935247" vmpeak="2020444" vmrss="142324" vmhwm="258949" />
+        <!--Models with FP16 precision-->
+        <model path="public/ctdet_coco_dlav0_384/FP16/ctdet_coco_dlav0_384.xml" precision="FP16" test="create_exenetwork" device="MYRIAD" vmsize="1904879" vmpeak="1904879" vmrss="207604" vmhwm="207604" />
+        <model path="public/i3d-rgb-tf/FP16/i3d-rgb-tf.xml" precision="FP16" test="create_exenetwork" device="MYRIAD" vmsize="3860984" vmpeak="3860984" vmrss="2147116" vmhwm="2147116" />
+        <model path="public/mobilenet-ssd/FP16/mobilenet-ssd.xml" precision="FP16" test="infer_request_inference" device="MYRIAD" vmsize="1886658" vmpeak="1886658" vmrss="94010" vmhwm="94010" />
+        <model path="public/se-inception/FP16/se-inception.xml" precision="FP16" test="create_exenetwork" device="MYRIAD" vmsize="1839182" vmpeak="1839182" vmrss="142168" vmhwm="142168" />
+        <model path="public/ssd_mobilenet_v2_coco/FP16/ssd_mobilenet_v2_coco.xml" precision="FP16" test="infer_request_inference" device="MYRIAD" vmsize="2013221" vmpeak="2098418" vmrss="219991" vmhwm="219991" />
+    </models>
+</attributes>
diff --git a/tests/stress_tests/.automation/memcheck_tests/nightly_configs/myriad_test_config.xml b/tests/stress_tests/.automation/memcheck_tests/nightly_configs/myriad_test_config.xml
new file mode 100644
index 00000000000000..7409c6287dae8d
--- /dev/null
+++ b/tests/stress_tests/.automation/memcheck_tests/nightly_configs/myriad_test_config.xml
@@ -0,0 +1,115 @@
+<attributes>
+    <devices>
+        <value>MYRIAD</value>
+    </devices>
+    <models>
+        <!--Models with FP32 precision-->
+        <model name="mobilenet-v2-1.4-224" precision="FP32" source="omz" />
+        <model name="brain-tumor-segmentation-0001" precision="FP32" source="omz" />
+        <model name="faster_rcnn_inception_resnet_v2_atrous_coco" precision="FP32" source="omz" />
+        <model name="yolo-v2-tf" precision="FP32" source="omz" />
+        <model name="se-inception" precision="FP32" source="omz" />
+        <model name="efficientnet-b0" precision="FP32" source="omz" />
+        <model name="mobilenet-v1-1.0-224-tf" precision="FP32" source="omz" />
+        <model name="mask_rcnn_resnet101_atrous_coco" precision="FP32" source="omz" />
+        <model name="ssd_mobilenet_v1_coco" precision="FP32" source="omz" />
+        <model name="se-resnet-152" precision="FP32" source="omz" />
+        <model name="octave-resnext-50-0.25" precision="FP32" source="omz" />
+        <model name="googlenet-v3" precision="FP32" source="omz" />
+        <model name="ssd_mobilenet_v2_coco" precision="FP32" source="omz" />
+        <model name="alexnet" precision="FP32" source="omz" />
+        <model name="googlenet-v4-tf" precision="FP32" source="omz" />
+        <model name="ssd300" precision="FP32" source="omz" />
+        <model name="vgg19" precision="FP32" source="omz" />
+        <model name="ctdet_coco_dlav0_384" precision="FP32" source="omz" />
+        <model name="googlenet-v1" precision="FP32" source="omz" />
+        <model name="yolo-v3-tf" precision="FP32" source="omz" />
+        <model name="mtcnn-o" precision="FP32" source="omz" />
+        <model name="yolo-v1-tiny-tf" precision="FP32" source="omz" />
+        <model name="googlenet-v1-tf" precision="FP32" source="omz" />
+        <model name="yolo-v2-tiny-tf" precision="FP32" source="omz" />
+        <model name="ssd512" precision="FP32" source="omz" />
+        <model name="densenet-169" precision="FP32" source="omz" />
+        <model name="brain-tumor-segmentation-0002" precision="FP32" source="omz" />
+        <model name="Sphereface" precision="FP32" source="omz" />
+        <model name="googlenet-v2" precision="FP32" source="omz" />
+        <model name="ctdet_coco_dlav0_512" precision="FP32" source="omz" />
+        <model name="ctpn" precision="FP32" source="omz" />
+        <model name="i3d-rgb-tf" precision="FP32" source="omz" />
+        <model name="mobilenet-v2" precision="FP32" source="omz" />
+        <model name="mobilenet-ssd" precision="FP32" source="omz" />
+        <model name="se-resnext-50" precision="FP32" source="omz" />
+        <model name="caffenet" precision="FP32" source="omz" />
+        <model name="mtcnn-r" precision="FP32" source="omz" />
+        <model name="se-resnet-50" precision="FP32" source="omz" />
+        <model name="densenet-121" precision="FP32" source="omz" />
+        <!--Models with FP16 precision-->
+        <model name="mobilenet-v2-1.4-224" precision="FP16" source="omz" />
+        <model name="brain-tumor-segmentation-0001" precision="FP16" source="omz" />
+        <model name="faster_rcnn_inception_resnet_v2_atrous_coco" precision="FP16" source="omz" />
+        <model name="yolo-v2-tf" precision="FP16" source="omz" />
+        <model name="se-inception" precision="FP16" source="omz" />
+        <model name="efficientnet-b0" precision="FP16" source="omz" />
+        <model name="mobilenet-v1-1.0-224-tf" precision="FP16" source="omz" />
+        <model name="mask_rcnn_resnet101_atrous_coco" precision="FP16" source="omz" />
+        <model name="ssd_mobilenet_v1_coco" precision="FP16" source="omz" />
+        <model name="se-resnet-152" precision="FP16" source="omz" />
+        <model name="octave-resnext-50-0.25" precision="FP16" source="omz" />
+        <model name="googlenet-v3" precision="FP16" source="omz" />
+        <model name="ssd_mobilenet_v2_coco" precision="FP16" source="omz" />
+        <model name="alexnet" precision="FP16" source="omz" />
+        <model name="googlenet-v4-tf" precision="FP16" source="omz" />
+        <model name="ssd300" precision="FP16" source="omz" />
+        <model name="vgg19" precision="FP16" source="omz" />
+        <model name="ctdet_coco_dlav0_384" precision="FP16" source="omz" />
+        <model name="googlenet-v1" precision="FP16" source="omz" />
+        <model name="yolo-v3-tf" precision="FP16" source="omz" />
+        <model name="mtcnn-o" precision="FP16" source="omz" />
+        <model name="yolo-v1-tiny-tf" precision="FP16" source="omz" />
+        <model name="googlenet-v1-tf" precision="FP16" source="omz" />
+        <model name="yolo-v2-tiny-tf" precision="FP16" source="omz" />
+        <model name="ssd512" precision="FP16" source="omz" />
+        <model name="densenet-169" precision="FP16" source="omz" />
+        <model name="brain-tumor-segmentation-0002" precision="FP16" source="omz" />
+        <model name="Sphereface" precision="FP16" source="omz" />
+        <model name="googlenet-v2" precision="FP16" source="omz" />
+        <model name="ctdet_coco_dlav0_512" precision="FP16" source="omz" />
+        <model name="ctpn" precision="FP16" source="omz" />
+        <model name="i3d-rgb-tf" precision="FP16" source="omz" />
+        <model name="mobilenet-v2" precision="FP16" source="omz" />
+        <model name="mobilenet-ssd" precision="FP16" source="omz" />
+        <model name="se-resnext-50" precision="FP16" source="omz" />
+        <model name="caffenet" precision="FP16" source="omz" />
+        <model name="mtcnn-r" precision="FP16" source="omz" />
+        <model name="se-resnet-50" precision="FP16" source="omz" />
+        <model name="densenet-121" precision="FP16" source="omz" />
+        <!--Models with FP16-INT8 precision-->
+        <model name="vehicle-attributes-recognition-barrier-0039" precision="FP16-INT8" source="omz" />
+        <model name="person-detection-action-recognition-0006" precision="FP16-INT8" source="omz" />
+        <model name="person-vehicle-bike-detection-crossroad-1016" precision="FP16-INT8" source="omz" />
+        <model name="age-gender-recognition-retail-0013" precision="FP16-INT8" source="omz" />
+        <model name="vehicle-detection-adas-0002" precision="FP16-INT8" source="omz" />
+        <model name="image-retrieval-0001" precision="FP16-INT8" source="omz" />
+        <model name="person-detection-retail-0002" precision="FP16-INT8" source="omz" />
+        <model name="person-attributes-recognition-crossroad-0230" precision="FP16-INT8" source="omz" />
+        <model name="action-recognition-0001-decoder" precision="FP16-INT8" source="omz" />
+        <model name="person-detection-action-recognition-teacher-0002" precision="FP16-INT8" source="omz" />
+        <model name="person-vehicle-bike-detection-crossroad-0078" precision="FP16-INT8" source="omz" />
+        <model name="face-detection-adas-0001" precision="FP16-INT8" source="omz" />
+        <model name="unet-camvid-onnx-0001" precision="FP16-INT8" source="omz" />
+        <model name="human-pose-estimation-0001" precision="FP16-INT8" source="omz" />
+        <model name="faster-rcnn-resnet101-coco-sparse-60-0001" precision="FP16-INT8" source="omz" />
+        <model name="action-recognition-0001-encoder" precision="FP16-INT8" source="omz" />
+        <model name="yolo-v2-ava-sparse-35-0001" precision="FP16-INT8" source="omz" />
+        <model name="person-detection-raisinghand-recognition-0001" precision="FP16-INT8" source="omz" />
+        <model name="person-detection-asl-0001" precision="FP16-INT8" source="omz" />
+        <model name="yolo-v2-tiny-ava-0001" precision="FP16-INT8" source="omz" />
+        <model name="license-plate-recognition-barrier-0001" precision="FP16-INT8" source="omz" />
+        <model name="person-detection-retail-0013" precision="FP16-INT8" source="omz" />
+        <model name="single-image-super-resolution-1032" precision="FP16-INT8" source="omz" />
+        <model name="landmarks-regression-retail-0009" precision="FP16-INT8" source="omz" />
+        <model name="driver-action-recognition-adas-0002-decoder" precision="FP16-INT8" source="omz" />
+        <model name="yolo-v2-ava-0001" precision="FP16-INT8" source="omz" />
+        <model name="person-detection-action-recognition-0005" precision="FP16-INT8" source="omz" />
+    </models>
+</attributes>
\ No newline at end of file
diff --git a/tests/stress_tests/.automation/memleaks_tests/nightly_configs/myriad_test_config.xml b/tests/stress_tests/.automation/memleaks_tests/nightly_configs/myriad_test_config.xml
new file mode 100644
index 00000000000000..3718751b253d7d
--- /dev/null
+++ b/tests/stress_tests/.automation/memleaks_tests/nightly_configs/myriad_test_config.xml
@@ -0,0 +1,19 @@
+<attributes>
+    <processes>
+        <value>1</value>
+    </processes>
+    <threads>
+        <value>1</value>
+    </threads>
+    <iterations>
+        <value>1000</value>
+    </iterations>
+    <devices>
+        <value>MYRIAD</value>
+    </devices>
+    <models>
+        <model name="alexnet" precision="FP32" source="omz" />
+        <model name="mobilenet-ssd" precision="FP32" source="omz" />
+        <model name="mtcnn-r" precision="FP32" source="omz" />
+    </models>
+</attributes>
\ No newline at end of file
diff --git a/tests/stress_tests/.automation/unittests/nightly_configs/myriad_test_config.xml b/tests/stress_tests/.automation/unittests/nightly_configs/myriad_test_config.xml
new file mode 100644
index 00000000000000..979194a991ac26
--- /dev/null
+++ b/tests/stress_tests/.automation/unittests/nightly_configs/myriad_test_config.xml
@@ -0,0 +1,20 @@
+<attributes>
+    <processes>
+        <value>1</value>
+    </processes>
+    <threads>
+        <value>1</value>
+        <value>2</value>
+    </threads>
+    <iterations>
+        <value>100</value>
+    </iterations>
+    <devices>
+        <value>MYRIAD</value>
+    </devices>
+    <models>
+        <model name="alexnet" precision="FP32" source="omz" />
+        <model name="mobilenet-ssd" precision="FP32" source="omz" />
+        <model name="mtcnn-r" precision="FP32" source="omz" />
+    </models>
+</attributes>

From ae5608534ed76572ed4fcaf412291656959091e2 Mon Sep 17 00:00:00 2001
From: iliya mironov <iliya.mironov@intel.com>
Date: Tue, 8 Jun 2021 13:15:54 +0300
Subject: [PATCH 026/178] Add support resize with 2 inputs (#5927)

* Add support resize with 2 inputs

* Add unit tests

* Hot fix

* Change resize check from port count to connected num port conditions

* Fix conditions

* Refactoring code according to review

* Fix according to review

* Change onnresize11 input condition
---
 .../middle/ONNXResize11ToInterpolate.py       | 13 ++--
 .../extensions/ops/ONNXResize11.py            | 13 ++--
 .../extensions/ops/ONNXResize11_test.py       | 68 ++++++++++++++++++-
 3 files changed, 80 insertions(+), 14 deletions(-)

diff --git a/model-optimizer/extensions/middle/ONNXResize11ToInterpolate.py b/model-optimizer/extensions/middle/ONNXResize11ToInterpolate.py
index ffd3c27eebdba6..c462d45285faa5 100644
--- a/model-optimizer/extensions/middle/ONNXResize11ToInterpolate.py
+++ b/model-optimizer/extensions/middle/ONNXResize11ToInterpolate.py
@@ -34,12 +34,13 @@ def replace_resize(graph: Graph, resize: Node):
         log.warning('The input shape is not 4D or 5D for op with name {}'.format(resize_name))
         return
 
-    num_of_inputs = len([port for port in resize.in_ports().values() if not port.disconnected()])
-    assert num_of_inputs in {3, 4}, \
-        "Number of inputs of ONNXResize (with name {}) should be equal to 3 or 4".format(resize_name)
+    assert (resize.is_in_port_connected(0) and (resize.is_in_port_connected(2) or resize.is_in_port_connected(3))), \
+        "Scales or sizes inputs must be connected to Node {} with op {}.".format(resize.soft_get("name", resize.id),
+                                                                                 resize.op)
 
     assert resize.soft_get('coordinate_transformation_mode') != 'tf_crop_and_resize', \
-        'Mode tf_crop_and_resize is not supported for op {} with name {}'.format(resize.op, resize_name)
+        'Mode tf_crop_and_resize is not supported for op {} with name {}'.format(resize.op,
+                                                                                 resize.soft_get("name", resize.id))
 
     layout = graph.graph['layout']
 
@@ -74,7 +75,7 @@ def replace_resize(graph: Graph, resize: Node):
                       {'name': resize_name + '/axis',
                        'value': int64_array(np.arange(begin_dim, end_dim))}).create_node()
 
-    shape_calculation_mode = 'scales' if num_of_inputs == 3 else 'sizes'
+    shape_calculation_mode = 'sizes' if resize.is_in_port_connected(3) else 'scales'
 
     interpolate_node = Interpolate(graph, {'version': 'opset4',
                                            'mode': convert_mode(resize.mode),
@@ -96,7 +97,7 @@ def replace_resize(graph: Graph, resize: Node):
 
     dst_dtype = np.float32  # even if data_type=FP16 use float32 for shape values
 
-    if num_of_inputs == 3:
+    if not resize.is_in_port_connected(3):
         cast_shape_to_float = Cast(graph, {'dst_type': dst_dtype}).create_node()
         mul_node = Mul(graph, {'name': resize_name + '/Mul'}).create_node()
         shape_of.out_port(0).connect(cast_shape_to_float.in_port(0))
diff --git a/model-optimizer/extensions/ops/ONNXResize11.py b/model-optimizer/extensions/ops/ONNXResize11.py
index 5ef8d7f6ac88c2..5476087a3d404d 100644
--- a/model-optimizer/extensions/ops/ONNXResize11.py
+++ b/model-optimizer/extensions/ops/ONNXResize11.py
@@ -35,14 +35,15 @@ def onnx_resize_infer(node: Node):
         if input_shape is None:
             return
 
-        num_of_in_nodes = len(node.in_nodes())
-        assert num_of_in_nodes in {3, 4}, \
-            "Node {} with op {} number of inputs must be equal to 3 or 4.".format(node.name, node.op)
+        assert (node.is_in_port_connected(0) and (node.is_in_port_connected(2) or node.is_in_port_connected(3))), \
+            "One of the scales or sizes inputs must be connected to Node {} with op {}.".format(node.soft_get("name", node.id),
+                                                                                                node.op)
 
         assert node.coordinate_transformation_mode != 'tf_crop_and_resize', \
-            'Mode tf_crop_and_resize is not supported for op {} with name {}'.format(node.op, node.name)
+            'Mode tf_crop_and_resize is not supported for op {} with name {}'.format(node.op,
+                                                                                     node.soft_get("name", node.id))
 
-        if num_of_in_nodes == 3:
+        if not node.is_in_port_connected(3):
             # i.e. input 'sizes' is not given
             input2_value = node.in_port(2).data.get_value()
             assert input2_value is not None, \
@@ -53,7 +54,7 @@ def onnx_resize_infer(node: Node):
             # i.e. input 'sizes' is given
             sizes = node.in_port(3).data.get_value()
             assert sizes is not None, \
-                "Node {} with op {} has no value in input port 3".format(node.name, node.op)
+                "Node {} with op {} has no value in input port 3".format(node.soft_get("name", node.id), node.op)
             output_shape = input_shape.copy()
             spatial_dimension_indices = range(2, len(input_shape))
             output_shape[spatial_dimension_indices] = int64_array(sizes)[2:]
diff --git a/model-optimizer/unit_tests/extensions/ops/ONNXResize11_test.py b/model-optimizer/unit_tests/extensions/ops/ONNXResize11_test.py
index b38773b82798c2..aac234c56ba50e 100644
--- a/model-optimizer/unit_tests/extensions/ops/ONNXResize11_test.py
+++ b/model-optimizer/unit_tests/extensions/ops/ONNXResize11_test.py
@@ -33,10 +33,8 @@
     ('input', 'input_data'),
     ('roi', 'roi_data'),
     ('sizes', 'sizes_data'),
-    ('scales', 'scales_data'),
     ('input_data', 'onnx_resize11', {'in': 0}),
     ('roi_data', 'onnx_resize11', {'in': 1}),
-    ('scales_data', 'onnx_resize11', {'in': 2}),
     ('sizes_data', 'onnx_resize11', {'in': 3}),
     ('onnx_resize11', 'onnx_resize11_data'),
     ('onnx_resize11_data', 'op_output'),
@@ -125,3 +123,69 @@ def test_onnx_resize_using_scales(self, input_shape, output_shape, scales):
 
         self.assertTrue(np.array_equal(graph.node['onnx_resize11_data']['shape'], int64_array(output_shape)),
                         msg.format(scales, output_shape, graph.node['onnx_resize11_data']['shape']))
+
+    @generate(*[([1, 260, 100, 150], [1, 260, 200, 350], [1, 260, 200, 350], [1.0, 1.0, 1.0, 1.0]),
+                ([1, 260, 100, 150], [1, 260, 200, 350], [1, 1, 200, 350], [1.0, 1.0, 1.0, 1.0]),
+                ([5, 14, 300, 40], [5, 14, 140, 280], [1, 1, 140, 280], [1.0, 1.0, 1.0, 1.0]),
+                ([5, 14, 300, 40], [5, 14, 140, 280], [5, 14, 140, 280], [1.0, 1.0, 1.0, 1.0]),
+                ([1, 3, 260, 100, 150], [1, 3, 780, 200, 350], [1, 3, 780, 200, 350], [1.0, 1.0, 1.0, 1.0, 1.0]),
+                ([1, 3, 450, 100, 150], [1, 3, 260, 200, 350], [1, 3, 260, 200, 350], [1.0, 1.0, 1.0, 1.0, 1.0]),
+                ([5, 14, 1000, 300, 40], [5, 14, 500, 140, 280], [1, 1, 500, 140, 280], [1.0, 1.0, 1.0, 1.0, 1.0]),
+                ([5, 14, 1000, 300, 40], [5, 14, 500, 140, 280], [5, 14, 500, 140, 280], [1.0, 1.0, 1.0, 1.0, 1.0])])
+    def test_onnx_resize11_using_sizes_without_roi_input(self, input_shape, output_shape, sizes, scales):
+        np_scales = np.array(scales)
+        np_sizes = int64_array(sizes)
+        graph = build_graph(nodes_attrs=graph_node_attrs_sizes,
+                            edges=[('input', 'input_data'),
+                                   ('sizes', 'sizes_data'),
+                                   ('input_data', 'onnx_resize11', {'in': 0}),
+                                   ('sizes_data', 'onnx_resize11', {'in': 3}),
+                                   ('onnx_resize11', 'onnx_resize11_data'),
+                                   ('onnx_resize11_data', 'op_output'),
+                                ],
+                            update_attributes={
+                                'input_data': {'shape': int64_array(input_shape)},
+                                'scales': {'shape': int64_array(np_scales.shape), 'value': np_scales},
+                                'scales_data': {'shape': int64_array(np_scales.shape), 'value': np_scales},
+                                'sizes': {'shape': int64_array(np_sizes.shape), 'value': np_sizes},
+                                'sizes_data': {'shape': int64_array(np_sizes.shape), 'value': np_sizes},
+                            })
+        node = Node(graph, 'onnx_resize11')
+        ONNXResize11Op.onnx_resize_infer(node)
+
+        msg = "ONNXResize11 infer failed for case: sizes={}, scales={}, expected_shape={}, actual_shape={}"
+
+        self.assertTrue(np.array_equal(graph.node['onnx_resize11_data']['shape'], int64_array(output_shape)),
+                        msg.format(sizes, scales, output_shape, graph.node['onnx_resize11_data']['shape']))
+
+    @generate(*[([1, 260, 100, 150], [1, 260, 200, 350], [1.0, 1.0, 2.0, 350 / 150]),
+                ([1, 3, 100, 200], [1, 3, 350, 150], [1.0, 1.0, 3.5, 150 / 200]),
+                ([5, 14, 300, 40], [5, 14, 140, 280], [1.0, 1.0, 140 / 300, 7.0]),
+                ([5, 14, 300, 40], [5, 14, 140, 560], [1.0, 1.0, 140 / 300, 14.0]),
+                ([1, 3, 260, 100, 150], [1, 3, 780, 200, 350], [1.0, 1.0, 3.0, 2.0, 350 / 150]),
+                ([1, 3, 450, 100, 150], [1, 3, 260, 200, 350], [1.0, 1.0, 260 / 450, 2.0, 350 / 150]),
+                ([5, 14, 1000, 300, 40], [5, 14, 500, 140, 280], [1.0, 1.0, 0.5, 140 / 300, 7.0]),
+                ([4, 3, 180, 1340], [4, 3, 60, 804], [1.0, 1.0, 0.33333334, 0.6]),
+                ([4, 3, 500, 180, 1340], [4, 3, 750, 60, 804], [1.0, 1.0, 1.5, 0.33333334, 0.6])])
+    def test_onnx_resize_using_scales_without_roi(self, input_shape, output_shape, scales):
+        np_scales = np.array(scales)
+        graph = build_graph(nodes_attrs=graph_node_attrs_scales,
+                            edges=[('input', 'input_data'),
+                                   ('scales', 'scales_data'),
+                                   ('input_data', 'onnx_resize11', {'in': 0}),
+                                   ('scales_data', 'onnx_resize11', {'in': 2}),
+                                   ('onnx_resize11', 'onnx_resize11_data'),
+                                   ('onnx_resize11_data', 'op_output'),
+                            ],
+                            update_attributes={
+                                'input_data': {'shape': int64_array(input_shape)},
+                                'scales': {'shape': int64_array(np_scales.shape), 'value': np_scales},
+                                'scales_data': {'shape': int64_array(np_scales.shape), 'value': np_scales},
+                            })
+        node = Node(graph, 'onnx_resize11')
+        ONNXResize11Op.onnx_resize_infer(node)
+
+        msg = "ONNXResize11 infer failed for case: scales={}, expected_shape={}, actual_shape={}"
+
+        self.assertTrue(np.array_equal(graph.node['onnx_resize11_data']['shape'], int64_array(output_shape)),
+                        msg.format(scales, output_shape, graph.node['onnx_resize11_data']['shape']))

From 76313a0ffa2896e76d01180f0843e7c860d9a719 Mon Sep 17 00:00:00 2001
From: Andrey Zaytsev <andrey.zaytsev@intel.com>
Date: Tue, 8 Jun 2021 13:39:35 +0300
Subject: [PATCH 027/178] Feature/azaytsev/update license aggrements link
 (#6068)

* Added info on DockerHub CI Framework

* Feature/azaytsev/change layout (#3295)

* Changes according to feedback comments

* Replaced @ref's with html links

* Fixed links, added a title page for installing from repos and images, fixed formatting issues

* Added links

* minor fix

* Added DL Streamer to the list of components installed by default

* Link fixes

* Link fixes

* ovms doc fix (#2988)

* added OpenVINO Model Server

* ovms doc fixes

Co-authored-by: Trawinski, Dariusz <dariusz.trawinski@intel.com>

* Updated openvino_docs.xml

* Updated the link to software license agreements

* Revert "Updated the link to software license agreements"

This reverts commit 706dac500e764bd7534f7005ac6197f827d68cb5.

* Updated the link to software license agreements

Co-authored-by: Trawinski, Dariusz <dariusz.trawinski@intel.com>
---
 docs/install_guides/installing-openvino-apt.md | 2 +-
 docs/install_guides/installing-openvino-yum.md | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/install_guides/installing-openvino-apt.md b/docs/install_guides/installing-openvino-apt.md
index 665186969912da..1bd734bd856969 100644
--- a/docs/install_guides/installing-openvino-apt.md
+++ b/docs/install_guides/installing-openvino-apt.md
@@ -2,7 +2,7 @@
 
 This guide provides installation steps for Intel® Distribution of OpenVINO™ toolkit for Linux* distributed through the APT repository.
 
-> **IMPORTANT**: By downloading and using this container and the included software, you agree to the terms and conditions of the [software license agreements](https://software.intel.com/en-us/license/eula-for-intel-software-development-products). Please, review the content inside the `<openvino_install_root>/licensing` folder for more details.
+> **IMPORTANT**: By downloading and using this container and the included software, you agree to the terms and conditions of the [software license agreements](https://software.intel.com/content/dam/develop/external/us/en/documents/intel-openvino-license-agreements.pdf). Please, review the content inside the `<openvino_install_root>/licensing` folder for more details.
 
 > **NOTE**: Intel® Graphics Compute Runtime for OpenCL™ is not a part of OpenVINO™ APT distribution. You can install it from the [Intel® Graphics Compute Runtime for OpenCL™ GitHub repo](https://github.com/intel/compute-runtime). 
 
diff --git a/docs/install_guides/installing-openvino-yum.md b/docs/install_guides/installing-openvino-yum.md
index 27e464d1b84bd5..c326cb93a0f6c1 100644
--- a/docs/install_guides/installing-openvino-yum.md
+++ b/docs/install_guides/installing-openvino-yum.md
@@ -2,7 +2,7 @@
 
 This guide provides installation steps for the Intel® Distribution of OpenVINO™ toolkit for Linux* distributed through the YUM repository.
 
-> **IMPORTANT**: By downloading and using this container and the included software, you agree to the terms and conditions of the [software license agreements](https://software.intel.com/en-us/license/eula-for-intel-software-development-products). Please, review the content inside the `<openvino_install_root>/licensing` folder for more details.
+> **IMPORTANT**: By downloading and using this container and the included software, you agree to the terms and conditions of the [software license agreements](https://software.intel.com/content/dam/develop/external/us/en/documents/intel-openvino-license-agreements.pdf). Please, review the content inside the `<openvino_install_root>/licensing` folder for more details.
 
 > **NOTE**: Intel® Graphics Compute Runtime for OpenCL™ is not a part of OpenVINO™ YUM distribution. You can install it from the [Intel® Graphics Compute Runtime for OpenCL™ GitHub repo](https://github.com/intel/compute-runtime).
 

From 440d3ee0ba59700c9ced758877f4033e91c93b8c Mon Sep 17 00:00:00 2001
From: Andrey Zaytsev <andrey.zaytsev@intel.com>
Date: Tue, 8 Jun 2021 13:49:01 +0300
Subject: [PATCH 028/178] Feature/azaytsev/update gpu driver link (#6062)

* Added info on DockerHub CI Framework

* Feature/azaytsev/change layout (#3295)

* Changes according to feedback comments

* Replaced @ref's with html links

* Fixed links, added a title page for installing from repos and images, fixed formatting issues

* Added links

* minor fix

* Added DL Streamer to the list of components installed by default

* Link fixes

* Link fixes

* ovms doc fix (#2988)

* added OpenVINO Model Server

* ovms doc fixes

Co-authored-by: Trawinski, Dariusz <dariusz.trawinski@intel.com>

* Updated openvino_docs.xml

* Updated link

* Update installing-openvino-windows.md

Co-authored-by: Trawinski, Dariusz <dariusz.trawinski@intel.com>
---
 docs/install_guides/installing-openvino-windows.md | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/docs/install_guides/installing-openvino-windows.md b/docs/install_guides/installing-openvino-windows.md
index 1a1a31a07c61fe..054950292b6337 100644
--- a/docs/install_guides/installing-openvino-windows.md
+++ b/docs/install_guides/installing-openvino-windows.md
@@ -248,8 +248,8 @@ Or proceed to the <a href="#get-started">Get Started</a> to get started with run
 
 > **NOTE**: These steps are required only if you want to use an Intel® integrated GPU.
 
-If your applications offload computation to **Intel® Integrated Graphics**, you must have the latest version of Intel Graphics Driver for Windows installed for your hardware. 
-[Download and install a higher version](http://downloadcenter.intel.com/product/80939/Graphics-Drivers). 
+If your applications offload computation to **Intel® Integrated Graphics**, you must have the Intel Graphics Driver for Windows installed for your hardware. 
+[Download and install the recommended version](https://downloadcenter.intel.com/download/30079/Intel-Graphics-Windows-10-DCH-Drivers). 
 
 To check if you have this driver installed:
 
@@ -265,8 +265,6 @@ To check if you have this driver installed:
 
    ![](../img/DeviceDriverVersion.PNG)
 
-> **NOTE**: To use the **Intel® Iris® Xe MAX Graphics**, see the [Drivers & Software](https://downloadcenter.intel.com/download/29993/Intel-Iris-Xe-MAX-Dedicated-Graphics-Drivers?product=80939) page for driver downloads and installation instructions.  
-
 You are done updating your device driver and are ready to use your GPU. Proceed to the <a href="#get-started">Get Started</a> to get started with running code samples and demo applications.
 
 ### <a name="hddl-myriad"></a> Optional: Additional Installation Steps for the Intel® Vision Accelerator Design with Intel® Movidius™ VPUs

From 9214fa72e2809cf7a353082a7107092f25770508 Mon Sep 17 00:00:00 2001
From: Alexandra Sidorova <alexandra.sidorova@intel.com>
Date: Tue, 8 Jun 2021 14:04:31 +0300
Subject: [PATCH 029/178] [CPU] Fixed AvgPooling and FQ fusing (#5994)

---
 .../plugin/cpu/single_layer_tests/pooling.cpp | 82 ++++++++++++++++---
 .../cpu/test_utils/fusing_test_utils.hpp      |  7 ++
 inference-engine/thirdparty/mkl-dnn           |  2 +-
 3 files changed, 80 insertions(+), 11 deletions(-)

diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp
index 72dbe0d5e12e60..0df9c464c4f5fd 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp
@@ -5,6 +5,7 @@
 #include "ngraph_functions/builders.hpp"
 #include "test_utils/cpu_test_utils.hpp"
 #include "shared_test_classes/single_layer/pooling.hpp"
+#include "test_utils/fusing_test_utils.hpp"
 
 using namespace InferenceEngine;
 using namespace CPUTestUtils;
@@ -13,21 +14,24 @@ using namespace LayerTestsDefinitions;
 namespace CPULayerTestsDefinitions {
 typedef std::tuple<
         poolLayerTestParamsSet,
-        CPUSpecificParams
+        CPUSpecificParams,
+        fusingSpecificParams
 > poolLayerCpuTestParamsSet;
 
 class PoolingLayerCPUTest : public testing::WithParamInterface<poolLayerCpuTestParamsSet>,
-                            virtual public LayerTestsUtils::LayerTestsCommon, public CPUTestsBase {
+                            virtual public LayerTestsUtils::LayerTestsCommon, public CpuTestWithFusing {
 public:
     static std::string getTestCaseName(const testing::TestParamInfo<poolLayerCpuTestParamsSet>& obj) {
         poolLayerTestParamsSet basicParamsSet;
         CPUSpecificParams cpuParams;
-        std::tie(basicParamsSet, cpuParams) = obj.param;
+        fusingSpecificParams fusingParams;
+        std::tie(basicParamsSet, cpuParams, fusingParams) = obj.param;
 
         std::ostringstream result;
         result << PoolingLayerTest::getTestCaseName(testing::TestParamInfo<poolLayerTestParamsSet>(
                 basicParamsSet, 0));
         result << CPUTestsBase::getTestCaseName(cpuParams);
+        result << CpuTestWithFusing::getTestCaseName(fusingParams);
 
         return result.str();
     }
@@ -36,7 +40,8 @@ class PoolingLayerCPUTest : public testing::WithParamInterface<poolLayerCpuTestP
     void SetUp() {
         poolLayerTestParamsSet basicParamsSet;
         CPUSpecificParams cpuParams;
-        std::tie(basicParamsSet, cpuParams) = this->GetParam();
+        fusingSpecificParams fusingParams;
+        std::tie(basicParamsSet, cpuParams, fusingParams) = this->GetParam();
 
         poolSpecificParams poolParams;
         std::vector<size_t> inputShape;
@@ -48,6 +53,7 @@ class PoolingLayerCPUTest : public testing::WithParamInterface<poolLayerCpuTestP
         }
 
         std::tie(inFmts, outFmts, priority, selectedType) = cpuParams;
+        std::tie(postOpMgrPtr, fusedOps) = fusingParams;
 
         if (selectedType.empty()) {
             selectedType = getPrimitiveType();
@@ -133,7 +139,8 @@ INSTANTIATE_TEST_CASE_P(smoke_MaxPool_CPU_4D, PoolingLayerCPUTest,
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::Values(std::vector<size_t >({1, 3, 64, 64})),
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
-                        ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs))),
+                        ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs)),
+                        ::testing::Values(emptyFusingSpec)),
                         PoolingLayerCPUTest::getTestCaseName);
 
 INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D, PoolingLayerCPUTest,
@@ -147,7 +154,8 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D, PoolingLayerCPUTest,
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(std::vector<size_t >({1, 4, 64, 64})),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
-                                ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs))),
+                                ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs)),
+                                ::testing::Values(emptyFusingSpec)),
                         PoolingLayerCPUTest::getTestCaseName);
 
 INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D_NotOptimized, PoolingLayerCPUTest,
@@ -161,7 +169,8 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D_NotOptimized, PoolingLayerCPUTest,
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(std::vector<size_t >({1, 4, 64, 64})),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
-                                ::testing::Values(ref)),
+                                ::testing::Values(ref),
+                                ::testing::Values(emptyFusingSpec)),
                         PoolingLayerCPUTest::getTestCaseName);
 
 const std::vector<poolSpecificParams> paramsMax5D = {
@@ -200,7 +209,8 @@ INSTANTIATE_TEST_CASE_P(smoke_MaxPool_CPU_5D, PoolingLayerCPUTest,
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(std::vector<size_t >({1, 3, 16, 32, 32})),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
-                                ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs))),
+                                ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs)),
+                                ::testing::Values(emptyFusingSpec)),
                         PoolingLayerCPUTest::getTestCaseName);
 
 INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D, PoolingLayerCPUTest,
@@ -214,7 +224,8 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D, PoolingLayerCPUTest,
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(std::vector<size_t >({1, 4, 32, 32, 32})),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
-                                ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs))),
+                                ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs)),
+                                ::testing::Values(emptyFusingSpec)),
                         PoolingLayerCPUTest::getTestCaseName);
 
 INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D_NotOptimized, PoolingLayerCPUTest,
@@ -228,7 +239,58 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D_NotOptimized, PoolingLayerCPUTest,
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(std::vector<size_t >({1, 4, 16, 16, 16})),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
-                                ::testing::Values(ref)),
+                                ::testing::Values(ref),
+                                ::testing::Values(emptyFusingSpec)),
+                        PoolingLayerCPUTest::getTestCaseName);
+
+/* === Fusing === */
+
+const auto avx512_nhwc = CPUSpecificParams{{nhwc}, {nhwc}, {"jit_avx512"}, "jit_avx512"};
+const auto avx512_ndhwc = CPUSpecificParams{{ndhwc}, {ndhwc}, {"jit_avx512"}, "jit_avx512"};
+
+const auto avx2_nhwc = CPUSpecificParams{{nhwc}, {nhwc}, {"jit_avx2"}, "jit_avx2"};
+const auto avx2_ndhwc = CPUSpecificParams{{ndhwc}, {ndhwc}, {"jit_avx2"}, "jit_avx2"};
+
+const auto sse42_nhwc = CPUSpecificParams{{nhwc}, {nhwc}, {"jit_sse42"}, "jit_sse42"};
+const auto sse42_ndhwc = CPUSpecificParams{{ndhwc}, {ndhwc}, {"jit_sse42"}, "jit_sse42"};
+
+const std::vector<CPUSpecificParams> vecCpuConfigsFusing_4D = {sse42_nhwc, avx2_nhwc, avx512_nhwc};
+const std::vector<CPUSpecificParams> vecCpuConfigsFusing_5D = {sse42_ndhwc, avx2_ndhwc, avx512_ndhwc};
+
+std::vector<fusingSpecificParams> fusingParamsSet {
+    emptyFusingSpec,
+    fusingFakeQuantizePerTensor,
+    fusingFakeQuantizePerChannel,
+};
+
+INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D_I8, PoolingLayerCPUTest,
+                        ::testing::Combine(
+                                ::testing::Combine(
+                                        ::testing::ValuesIn(paramsAvg4D),
+                                        ::testing::Values(Precision::FP32),
+                                        ::testing::Values(Precision::I8),
+                                        ::testing::Values(Precision::FP32),
+                                        ::testing::Values(InferenceEngine::Layout::ANY),
+                                        ::testing::Values(InferenceEngine::Layout::ANY),
+                                        ::testing::Values(std::vector<size_t >({1, 4, 64, 64})),
+                                        ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                                ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigsFusing_4D)),
+                                ::testing::ValuesIn(fusingParamsSet)),
+                        PoolingLayerCPUTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D_I8, PoolingLayerCPUTest,
+                        ::testing::Combine(
+                                ::testing::Combine(
+                                        ::testing::ValuesIn(paramsAvg5D),
+                                        ::testing::Values(Precision::FP32),
+                                        ::testing::Values(Precision::I8),
+                                        ::testing::Values(Precision::FP32),
+                                        ::testing::Values(InferenceEngine::Layout::ANY),
+                                        ::testing::Values(InferenceEngine::Layout::ANY),
+                                        ::testing::Values(std::vector<size_t >({1, 4, 16, 16, 16})),
+                                        ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                                ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigsFusing_5D)),
+                                ::testing::ValuesIn(fusingParamsSet)),
                         PoolingLayerCPUTest::getTestCaseName);
 
 } // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/test_utils/fusing_test_utils.hpp b/inference-engine/tests/functional/plugin/cpu/test_utils/fusing_test_utils.hpp
index 0e5fc43d366f72..d55f2a988593b2 100644
--- a/inference-engine/tests/functional/plugin/cpu/test_utils/fusing_test_utils.hpp
+++ b/inference-engine/tests/functional/plugin/cpu/test_utils/fusing_test_utils.hpp
@@ -216,6 +216,13 @@ const auto fusingScaleShift = fusingSpecificParams{ std::make_shared<postNodesMg
                 return std::make_shared<ngraph::opset1::Add>(inpNode, constNode);
             }, "Add(PerChannel)"}}), {"Add"} };
 
+const auto fusingFakeQuantizePerTensor = fusingSpecificParams{ std::make_shared<postNodesMgr>(std::vector<postNodeBuilder>{
+            {[](std::shared_ptr<ngraph::Node> inpNode, const ngraph::element::Type& ngPrc, ngraph::ParameterVector& params){
+                auto localPrc = inpNode->get_element_type();
+                ngraph::Shape newShape(inpNode->get_shape().size(), 1);
+                return ngraph::builder::makeFakeQuantize(inpNode, localPrc, 256, newShape);
+            }, "FakeQuantize(PerTensor)"}}), {"FakeQuantize"} };
+
 const auto fusingFakeQuantizePerChannel = fusingSpecificParams{std::make_shared<postNodesMgr>(std::vector<postNodeBuilder>{
             {[](std::shared_ptr<ngraph::Node> inpNode, const ngraph::element::Type& ngPrc, ngraph::ParameterVector& params){
                 auto localPrc = inpNode->get_element_type();
diff --git a/inference-engine/thirdparty/mkl-dnn b/inference-engine/thirdparty/mkl-dnn
index aa47fcd2a03ee5..1cb9d0615aaf51 160000
--- a/inference-engine/thirdparty/mkl-dnn
+++ b/inference-engine/thirdparty/mkl-dnn
@@ -1 +1 @@
-Subproject commit aa47fcd2a03ee5caac119b6417bc66abe3154aab
+Subproject commit 1cb9d0615aaf511b51b8f8fc3c3ff8805ad9be6c

From a7a9364b417b35464d352622254517f4ce602848 Mon Sep 17 00:00:00 2001
From: Irina Efode <irina.efode@intel.com>
Date: Tue, 8 Jun 2021 14:23:45 +0300
Subject: [PATCH 030/178] [IE TESTS] Add local_cache arg to the subgraphDumper
 (#6063)

---
 .../conformance/subgraphs_dumper/README.md    |  1 +
 .../subgraphs_dumper/include/gflag_config.hpp |  5 ++-
 .../conformance/subgraphs_dumper/src/main.cpp | 38 +++++++++++++------
 3 files changed, 32 insertions(+), 12 deletions(-)

diff --git a/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/README.md b/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/README.md
index 1f21dd1c07ff49..07b50de8409749 100644
--- a/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/README.md
+++ b/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/README.md
@@ -17,6 +17,7 @@ Outcome of a build is a `subgrpahsDumper` binary located in building artifacts f
 ## Running
 The tool takes two command line parameters:    
 * `--input_folders` - Required. Comma separated paths to the input folders with IRs
+* `--local_cache` - Optional. Comma separated paths to the local cache folders with IRs.
 * `--output_folder` - Required. Path to the output folders where to serialize IRs
 * `--path_regex` - Optional. regular expression to be applied in input folders recursive discovery
 * `--constants_size_threshold` - Optional. Maximum size of constant in megabytes to be serialized.
diff --git a/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/include/gflag_config.hpp b/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/include/gflag_config.hpp
index c37e45c445be24..13ad0007de73b5 100644
--- a/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/include/gflag_config.hpp
+++ b/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/include/gflag_config.hpp
@@ -9,6 +9,7 @@
 
 static const char help_message[] = "Print a usage message.";
 static const char input_folders_message[] = "Required. Comma separated paths to the input folders with IRs";
+static const char local_cache_message[] = "Optional. Comma separated paths to the local cache folders with IRs";
 static const char output_folder_message[] = "Required. Path to the output folders where to serialize IRs";
 static const char path_regex_message[] = "Optional. regular expression to be applied in input "
                                          "folders recursive discovery";
@@ -21,7 +22,8 @@ static const char eliminate_dynamism_message[] = "Optional. If specified dynamic
                                                  "and replaced by propagated upper bound values (if possible)";
 
 DEFINE_bool(h, false, help_message);
-DEFINE_string(input_folders, ".", input_folders_message);
+DEFINE_string(input_folders, "", local_cache_message);
+DEFINE_string(local_cache, ".", input_folders_message);
 DEFINE_string(output_folder, "output", output_folder_message);
 DEFINE_string(path_regex, ".*", output_folder_message);
 DEFINE_double(constants_size_threshold, 1., constants_size_threshold_message);
@@ -37,6 +39,7 @@ static void showUsage() {
     std::cout << "\n";
     std::cout << "    -h                                     " << help_message << "\n";
     std::cout << "    --input_folders \"<path>\"             " << input_folders_message << "\n";
+    std::cout << "    --local_cache \"<path>\"               " << input_folders_message << "\n";
     std::cout << "    --output_folder \"<path>\"             " << output_folder_message << "\n";
     std::cout << "    --path_regex \"<path>\"                " << path_regex_message << "\n";
     std::cout << "    --constants_size_threshold \"<value>\" " << constants_size_threshold_message << "\n";
diff --git a/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/src/main.cpp b/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/src/main.cpp
index 0bd42bf169df2b..16f8f55a98b4b6 100644
--- a/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/src/main.cpp
+++ b/inference-engine/tests/functional/plugin/conformance/subgraphs_dumper/src/main.cpp
@@ -19,17 +19,8 @@
 #include <stdlib.h>
 #include <string.h>
 
-int main(int argc, char *argv[]) {
-    uint8_t ret_code = 0;
-
-    gflags::ParseCommandLineNonHelpFlags(&argc, &argv, true);
-    if (FLAGS_h) {
-        showUsage();
-        return 0;
-    }
-    SubgraphsDumper::ClonersMap::constant_size_threshold_mb = FLAGS_constants_size_threshold;
+std::vector<SubgraphsDumper::Model> findModelsInDirs(const std::vector<std::string> &dirs) {
     std::vector<std::string> input_folder_content;
-    std::vector<std::string> dirs = CommonTestUtils::splitStringByDelimiter(FLAGS_input_folders);
     for (const auto &dir : dirs) {
         if (!CommonTestUtils::directoryExists(dir)) {
             std::string msg = "Input directory (" + dir + ") doesn't not exist!";
@@ -51,9 +42,13 @@ int main(int argc, char *argv[]) {
         std::string msg = "Output directory (" + FLAGS_output_folder + ") doesn't not exist!";
         throw std::runtime_error(msg);
     }
+    return models;
+}
 
+void cacheModels(std::unique_ptr<SubgraphsDumper::OPCache> &cache,
+                 uint8_t& ret_code,
+                 const std::vector<SubgraphsDumper::Model>& models) {
     auto ie = InferenceEngine::Core();
-    auto cache = SubgraphsDumper::OPCache::make_cache();
     time_t rawtime;
     struct tm *timeinfo;
     char buffer[20];
@@ -92,6 +87,27 @@ int main(int argc, char *argv[]) {
             }
         }
     }
+}
+
+
+int main(int argc, char *argv[]) {
+    uint8_t ret_code = 0;
+
+    gflags::ParseCommandLineNonHelpFlags(&argc, &argv, true);
+    if (FLAGS_h) {
+        showUsage();
+        return 0;
+    }
+    SubgraphsDumper::ClonersMap::constant_size_threshold_mb = FLAGS_constants_size_threshold;
+
+    std::vector<std::string> local_cache_dirs = CommonTestUtils::splitStringByDelimiter(FLAGS_local_cache);
+    std::vector<std::string> dirs = CommonTestUtils::splitStringByDelimiter(FLAGS_input_folders);
+    auto cachedOps = findModelsInDirs(local_cache_dirs);
+    auto models = findModelsInDirs(dirs);
+
+    auto cache = SubgraphsDumper::OPCache::make_cache();
+    cacheModels(cache, ret_code, cachedOps);
+    cacheModels(cache, ret_code, models);
     cache->serialize_cached_ops(FLAGS_output_folder);
 
     return ret_code;

From 9e34622ac1f899b3c6ddd151aadb3696ffc961eb Mon Sep 17 00:00:00 2001
From: Edward Shogulin <edward.shogulin@intel.com>
Date: Tue, 8 Jun 2021 15:29:30 +0300
Subject: [PATCH 031/178] [LPT] Concat precision selection fix (#6069)

---
 .../src/concat.cpp                            |  38 ++-
 .../src/concat_multi_channels.cpp             |  15 +-
 ...ate_precision_selection_transformation.cpp | 317 ++++++++++++++++++
 .../simple_low_precision_transformer.cpp      |  36 +-
 .../simple_low_precision_transformer.hpp      |  16 +-
 .../lpt_ngraph_functions/concat_function.hpp  |  19 ++
 .../src/concat_function.cpp                   | 123 +++++++
 7 files changed, 541 insertions(+), 23 deletions(-)
 create mode 100644 inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_precision_selection_transformation.cpp

diff --git a/inference-engine/src/low_precision_transformations/src/concat.cpp b/inference-engine/src/low_precision_transformations/src/concat.cpp
index 4988e29b1e289a..f6d860ed17225c 100644
--- a/inference-engine/src/low_precision_transformations/src/concat.cpp
+++ b/inference-engine/src/low_precision_transformations/src/concat.cpp
@@ -43,19 +43,21 @@ bool ConcatTransformation::transform(TransformationContext& context, ngraph::pat
         return false;
     }
 
-    // precisions can be different
+    // Concat operations precision is defined:
+    // 1. consumers after Concat
+    // 2. FakeQuantize precisions without zero point
     ngraph::Node& quantizationLayer = *subgraph.quantizationLayers[0];
     std::shared_ptr<ngraph::opset1::FakeQuantize> fq = ngraph::as_type_ptr<ngraph::opset1::FakeQuantize>(quantizationLayer.shared_from_this());
     if (!NetworkHelper::isQuantizeSupported(fq)) {
         return false;
     }
-
-    std::vector<element::Type> concatParentsChildrensPrecisions = precisionsOnActivations;
-    fillAvailablePrecisions(subgraph.quantizationLayers[0], concatParentsChildrensPrecisions);
-    if (concatParentsChildrensPrecisions.empty()) {
+    DataPrecision dataPrecision = getDataPrecision(fq, QuantizationDetails::getDetails(fq), false);
+    if (dataPrecision.precision == ngraph::element::undefined) {
         return false;
     }
 
+    std::vector<element::Type> concatChildrenPrecisions = precisionsOnActivations;
+
     for (size_t i = 0; i < subgraph.quantizationLayers.size(); ++i) {
         fq = ngraph::as_type_ptr<ngraph::opset1::FakeQuantize>(subgraph.quantizationLayers[i]);
         if (fq == nullptr) {
@@ -72,20 +74,28 @@ bool ConcatTransformation::transform(TransformationContext& context, ngraph::pat
         if (quantizationDetails.inputHighValues.size() != 1ul) {
             return false;
         }
-        std::vector<element::Type> fqChildrensPrecisions = precisionsOnActivations;
-        fillAvailablePrecisions(subgraph.quantizationLayers[i], fqChildrensPrecisions);
-        concatParentsChildrensPrecisions = NetworkHelper::precisionIntersection(concatParentsChildrensPrecisions, fqChildrensPrecisions);
 
-        if (concatParentsChildrensPrecisions.empty()) {
+        // define concatenation operation consumers precisions
+        std::vector<element::Type> fqChildrenPrecisions = precisionsOnActivations;
+        fillAvailablePrecisions(subgraph.quantizationLayers[i], fqChildrenPrecisions);
+        concatChildrenPrecisions = NetworkHelper::precisionIntersection(concatChildrenPrecisions, fqChildrenPrecisions);
+        if (concatChildrenPrecisions.empty()) {
+            return false;
+        }
+
+        // define FakeQuantize precisions without zero point
+        const DataPrecision dataPrecision2 = getDataPrecision(subgraph.quantizationLayers[i]->shared_from_this(), quantizationDetails, false);
+        if (dataPrecision2.precision == ngraph::element::undefined) {
             return false;
         }
+
+        if (dataPrecision.precision != dataPrecision2.precision) {
+            dataPrecision = dataPrecision.precision.is_signed() ? dataPrecision : dataPrecision2;
+        }
     }
 
-    DataPrecision dataPrecision;
-    if (std::find(concatParentsChildrensPrecisions.begin(), concatParentsChildrensPrecisions.end(), element::i8) != concatParentsChildrensPrecisions.end()) {
-        dataPrecision = DataPrecision(element::i8);
-    } else {
-        dataPrecision = DataPrecision(concatParentsChildrensPrecisions[0]);
+    if (std::find(concatChildrenPrecisions.begin(), concatChildrenPrecisions.end(), dataPrecision.precision) == concatChildrenPrecisions.end()) {
+        dataPrecision = DataPrecision(concatChildrenPrecisions[0]);
     }
 
     std::vector<QuantizationDetails> quantizationLayersDetails;
diff --git a/inference-engine/src/low_precision_transformations/src/concat_multi_channels.cpp b/inference-engine/src/low_precision_transformations/src/concat_multi_channels.cpp
index dc81d51cd717de..e36c2b5aa74528 100644
--- a/inference-engine/src/low_precision_transformations/src/concat_multi_channels.cpp
+++ b/inference-engine/src/low_precision_transformations/src/concat_multi_channels.cpp
@@ -64,14 +64,23 @@ bool ConcatMultiChannelsTransformation::transform(TransformationContext& context
 
     DataPrecision dataPrecision;
     {
+        std::vector<element::Type> concatChildrenPrecisions = precisionsOnActivations;
         for (auto quantizationLayer : subgraph.quantizationLayers) {
             std::shared_ptr<ngraph::opset1::FakeQuantize> fq = ngraph::as_type_ptr<ngraph::opset1::FakeQuantize>(quantizationLayer->shared_from_this());
             if (!NetworkHelper::isQuantizeSupported(fq)) {
                 return false;
             }
 
-            const DataPrecision tmp = getDataPrecision(fq, QuantizationDetails::getDetails(fq), false);
+            // define concatenation operation consumers precisions
+            std::vector<element::Type> fqChildrenPrecisions = precisionsOnActivations;
+            fillAvailablePrecisions(quantizationLayer, fqChildrenPrecisions);
+            concatChildrenPrecisions = NetworkHelper::precisionIntersection(concatChildrenPrecisions, fqChildrenPrecisions);
+            if (concatChildrenPrecisions.empty()) {
+                return false;
+            }
 
+            // define FakeQuantize precisions without zero point
+            const DataPrecision tmp = getDataPrecision(fq, QuantizationDetails::getDetails(fq), false);
             if (dataPrecision.precision == ngraph::element::undefined) {
                 dataPrecision = tmp;
                 continue;
@@ -81,6 +90,10 @@ bool ConcatMultiChannelsTransformation::transform(TransformationContext& context
                 dataPrecision = tmp;
             }
         }
+
+        if (std::find(concatChildrenPrecisions.begin(), concatChildrenPrecisions.end(), dataPrecision.precision) == concatChildrenPrecisions.end()) {
+            dataPrecision = DataPrecision(concatChildrenPrecisions[0]);
+        }
     }
 
     for (size_t i = 0; i < subgraph.quantizationLayers.size(); ++i) {
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_precision_selection_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_precision_selection_transformation.cpp
new file mode 100644
index 00000000000000..0d6b29d5fe55d4
--- /dev/null
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_precision_selection_transformation.cpp
@@ -0,0 +1,317 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "layer_transformation.hpp"
+
+#include <string>
+#include <sstream>
+#include <memory>
+
+#include <gtest/gtest.h>
+
+#include <transformations/utils/utils.hpp>
+#include <transformations/init_node_info.hpp>
+#include <low_precision/transformer.hpp>
+#include <low_precision/avg_pool.hpp>
+#include <low_precision/concat.hpp>
+#include <low_precision/concat_multi_channels.hpp>
+#include <low_precision/max_pool.hpp>
+
+#include "common_test_utils/ngraph_test_utils.hpp"
+#include "lpt_ngraph_functions/concat_function.hpp"
+#include "lpt_ngraph_functions/common/fake_quantize_on_data.hpp"
+#include "simple_low_precision_transformer.hpp"
+
+using namespace testing;
+using namespace ngraph;
+using namespace ngraph::pass;
+
+namespace {
+
+class ConcatTransformationActualValues {
+public:
+    ngraph::builder::subgraph::FakeQuantizeOnData fakeQuantize1;
+    ngraph::builder::subgraph::FakeQuantizeOnData fakeQuantize2;
+};
+
+inline std::ostream& operator<<(std::ostream& out, const ConcatTransformationActualValues& values) {
+    return out << "_" << values.fakeQuantize1 << "_" << values.fakeQuantize2;
+}
+
+class ConcatTransformationResultValues {
+public:
+    ngraph::builder::subgraph::FakeQuantizeOnData fakeQuantize1;
+    ngraph::builder::subgraph::FakeQuantizeOnData fakeQuantize2;
+    ngraph::element::Type precisionBeforeOp;
+    ngraph::builder::subgraph::DequantizationOperations dequantizationBefore1;
+    ngraph::builder::subgraph::DequantizationOperations dequantizationBefore2;
+    ngraph::element::Type precisionAfterOperation;
+    ngraph::builder::subgraph::DequantizationOperations dequantizationAfter1;
+    ngraph::builder::subgraph::DequantizationOperations dequantizationAfter2;
+};
+
+inline std::ostream& operator<<(std::ostream& out, const ConcatTransformationResultValues& values) {
+    return out << "_" <<
+        values.fakeQuantize1 << "_" <<
+        values.fakeQuantize2 << "_" <<
+        values.dequantizationAfter1 << "_" <<
+        values.dequantizationAfter2;
+}
+
+class ConcatTransformationTestValues {
+public:
+    ngraph::pass::low_precision::LayerTransformation::Params params;
+    bool multiChannels;
+    ConcatTransformationActualValues actual;
+    ConcatTransformationResultValues result;
+};
+
+inline std::ostream& operator<<(std::ostream& out, const ConcatTransformationTestValues& values) {
+    return out << "_" << values.multiChannels << "_" << values.actual << "_" << values.result;
+}
+
+typedef std::tuple <
+    ngraph::element::Type,
+    ngraph::Shape,
+    ConcatTransformationTestValues
+> ConcatTransformationParams;
+
+class ConcatWithIntermediatePrecisionSelectionTransformation : public LayerTransformation, public testing::WithParamInterface<ConcatTransformationParams> {
+public:
+    void SetUp() override {
+        const ngraph::element::Type precision = std::get<0>(GetParam());
+        const ngraph::Shape shape = std::get<1>(GetParam());
+        ConcatTransformationTestValues testValues = std::get<2>(GetParam());
+
+        actualFunction = ngraph::builder::subgraph::ConcatFunction::getOriginalWithIntermediateAvgPool(
+            precision,
+            shape,
+            testValues.actual.fakeQuantize1,
+            testValues.actual.fakeQuantize2);
+
+        SimpleLowPrecisionTransformer transform;
+        if (testValues.multiChannels) {
+            transform.addBranchSpecific<ngraph::pass::low_precision::ConcatMultiChannelsTransformation, ngraph::opset1::Concat>(testValues.params);
+        } else {
+            transform.addBranchSpecific<ngraph::pass::low_precision::ConcatTransformation, ngraph::opset1::Concat>(testValues.params);
+        }
+        transform.add<ngraph::pass::low_precision::MaxPoolTransformation, ngraph::opset1::MaxPool>(testValues.params);
+        transform.add<ngraph::pass::low_precision::AvgPoolTransformation, ngraph::opset1::AvgPool>(testValues.params);
+        transform.transform(actualFunction);
+
+        referenceFunction = ngraph::builder::subgraph::ConcatFunction::getReferenceWithIntermediateAvgPool(
+            precision,
+            shape,
+            testValues.result.fakeQuantize1,
+            testValues.result.fakeQuantize2,
+            testValues.result.precisionBeforeOp,
+            testValues.result.dequantizationBefore1,
+            testValues.result.dequantizationBefore2,
+            testValues.result.precisionAfterOperation,
+            testValues.result.dequantizationAfter1,
+            testValues.result.dequantizationAfter2);
+    }
+
+    static std::string getTestCaseName(testing::TestParamInfo<ConcatTransformationParams> obj) {
+        const ngraph::element::Type precision = std::get<0>(obj.param);
+        const ngraph::Shape shape = std::get<1>(obj.param);
+        const ConcatTransformationTestValues testValues = std::get<2>(obj.param);
+
+        std::ostringstream result;
+        result <<
+            LayerTransformation::getTestCaseNameByParams(precision, shape, testValues.params) << "_" <<
+            (testValues.multiChannels ? "multiChannels_" : "notMultiChannels_") <<
+            testValues.actual << "_" <<
+            testValues.result << "_";
+        return result.str();
+    }
+};
+
+TEST_P(ConcatWithIntermediatePrecisionSelectionTransformation, CompareFunctions) {
+    actualFunction->validate_nodes_and_infer_types();
+    auto res = compare_functions(referenceFunction, actualFunction, true, false, true);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+const std::vector<ngraph::element::Type> precisions = {
+    ngraph::element::f32,
+    // ngraph::element::f16
+};
+
+const std::vector<ConcatTransformationTestValues> testValues = {
+    // Concat: FakeQuantize operations with signed intervals but consumer requires U8
+    {
+        LayerTransformation::createParamsU8I8(),
+        false,
+        {
+            { 256ul, ngraph::Shape({}), {-1.28f}, {1.27f}, {-1.28f}, {1.27f} },
+            { 256ul, ngraph::Shape({}), {-1.28f / 2.f}, {1.27f / 2.f}, {-1.28f / 2.f}, {1.27f / 2.f} }
+        },
+        {
+            { 256ul, ngraph::Shape({}), {-1.28f}, {1.27f}, {0.f}, {255.f} },
+            { 256ul, ngraph::Shape({}), {-1.28f / 2.f}, {1.27f / 2.f}, {64.f}, {192.f} },
+            ngraph::element::u8,
+            {{}, {}, {}},
+            {{}, {}, {}},
+            ngraph::element::u8,
+            { ngraph::element::f32, { 128.f }, { 0.01f } },
+            { {}, { 128.f }, { 0.01f } }
+        }
+    },
+
+    // Concat: FakeQuantize operations with unsigned intervals but consumer requires I8
+    {
+        LayerTransformation::createParamsI8I8(),
+        false,
+        {
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f}, {0.f}, {2.55f} },
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f / 2.f}, {0.f}, {2.55f / 2.f} }
+        },
+        {
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f}, {-128.f}, {127.f} },
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f / 2.f}, {-128.f}, { -0.f} },
+            ngraph::element::i8,
+            {{}, {}, {}},
+            {{}, {}, {}},
+            ngraph::element::i8,
+            { ngraph::element::f32, { -128.f }, { 0.01f } },
+            { {}, { -128.f }, { 0.01f } }
+        }
+    },
+
+    // ConcatMultichannel: FakeQuantize operations with signed intervals but consumer requires U8
+    {
+        LayerTransformation::createParamsU8I8(),
+        true,
+        {
+            { 256ul, ngraph::Shape({}), {-1.28f}, {1.27f}, {-1.28f}, {1.27f} },
+            { 256ul, ngraph::Shape({}), {-1.28f / 2.f}, {1.27f / 2.f}, {-1.28f / 2.f}, {1.27f / 2.f} }
+        },
+        {
+            { 256ul, ngraph::Shape({}), {-1.28f}, {1.27f}, {0.f}, {255.f} },
+            { 256ul, ngraph::Shape({}), {-1.28f / 2.f}, {1.27f / 2.f}, {0.f}, { 255.f} },
+            ngraph::element::u8,
+            {},
+            {},
+            ngraph::element::u8,
+            { ngraph::element::f32, { 128.f }, {{ 0.01f, 0.01f, 0.01f, 0.005f, 0.005f, 0.005f }} },
+            { {}, { 128.f }, { 0.005f } }
+        }
+    },
+
+    // ConcatMultichannel: FakeQuantize operations with unsigned intervals but consumer requires I8
+    {
+        LayerTransformation::createParamsI8I8(),
+        true,
+        {
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f}, {0.f}, {2.55f} },
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f / 2.f}, {0.f}, {2.55f / 2.f} }
+        },
+        {
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f}, {-128.f}, {127.f} },
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f / 2.f}, {-128.f}, { 127.f} },
+            ngraph::element::i8,
+            {{}, {}, {}},
+            {{}, {}, {}},
+            ngraph::element::i8,
+            { ngraph::element::f32, { -128.f }, {{ 0.01f, 0.01f, 0.01f, 0.005f, 0.005f, 0.005f }} },
+            { {}, { -128.f }, { 0.005f } }
+        }
+    },
+
+    // Concat: FakeQuantize operations with unsigned intervals, no consumer limitations: FQ were decomposed to U8 precision
+    {
+        LayerTransformation::createParamsU8I8AndI8(),
+        false,
+        {
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f}, {0.f}, {2.55f} },
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f / 2.f}, {0.f}, {2.55f / 2.f} }
+        },
+        {
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f}, {0.f}, {255.f} },
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f / 2.f}, {0.f}, { 128.f} },
+            ngraph::element::u8,
+            {{}, {}, {}},
+            {{}, {}, {}},
+            ngraph::element::u8,
+            { ngraph::element::f32, {}, { 0.01f } },
+            { {}, {}, { 0.01f } }
+        }
+    },
+
+    // Concat: FakeQuantize operations with signed intervals, no consumer limitations: FQ were decomposed to I8 precision
+    {
+        LayerTransformation::createParamsU8I8AndI8(),
+        false,
+        {
+            { 256ul, ngraph::Shape({}), {-1.28f}, {1.27f}, {-1.28f}, {1.27f} },
+            { 256ul, ngraph::Shape({}), {-1.28f / 2.f}, {1.27f / 2.f}, {-1.28f / 2.f}, {1.27f / 2.f} }
+        },
+        {
+            { 256ul, ngraph::Shape({}), {-1.28f}, {1.27f}, {-128.f}, {127.f} },
+            { 256ul, ngraph::Shape({}), {-1.28f / 2.f}, {1.27f / 2.f}, {-64.f}, {64.f} },
+            ngraph::element::i8,
+            {{}, {}, {}},
+            {{}, {}, {}},
+            ngraph::element::i8,
+            { ngraph::element::f32, {}, { 0.01f } },
+            { {}, {}, { 0.01f } }
+        }
+    },
+
+    // ConcatMultichannel: FakeQuantize operations with unsigned intervals, no consumer limitations: FQ were decomposed to U8 precision
+    {
+        LayerTransformation::createParamsU8I8AndI8(),
+        true,
+        {
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f}, {0.f}, {2.55f} },
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f / 2.f}, {0.f}, {2.55f / 2.f} }
+        },
+        {
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f}, {0.f}, {255.f} },
+            { 256ul, ngraph::Shape({}), {0.f}, {2.55f / 2.f}, {0.f}, {255.f} },
+            ngraph::element::u8,
+            {{}, {}, {}},
+            {{}, {}, {}},
+            ngraph::element::u8,
+            { ngraph::element::f32, {}, {{ 0.01f, 0.01f, 0.01f, 0.005f, 0.005f, 0.005f }} },
+            { {}, {}, { 0.005f } }
+        }
+    },
+
+    // ConcatMultichannel: FakeQuantize operations with signed intervals, no consumer limitations: FQ were decomposed to I8 precision
+    {
+        LayerTransformation::createParamsU8I8AndI8(),
+        true,
+        {
+            { 256ul, ngraph::Shape({}), {-1.28f}, {1.27f}, {-1.28f}, {1.27f} },
+            { 256ul, ngraph::Shape({}), {-1.28f / 2.f}, {1.27f / 2.f}, {-1.28f / 2.f}, {1.27f / 2.f} }
+        },
+        {
+            { 256ul, ngraph::Shape({}), {-1.28f}, {1.27f}, {-128.f}, {127.f} },
+            { 256ul, ngraph::Shape({}), {-1.28f / 2.f}, {1.27f / 2.f}, {-128.f}, {127.f} },
+            ngraph::element::i8,
+            {{}, {}, {}},
+            {{}, {}, {}},
+            ngraph::element::i8,
+            { ngraph::element::f32, {}, {{ 0.01f, 0.01f, 0.01f, 0.005f, 0.005f, 0.005f }} },
+            { {}, {}, { 0.005f } }
+        }
+    }
+};
+
+const std::vector<ngraph::Shape> shapes = {
+    { 1, 3, 9, 9 },
+    { 4, 3, 9, 9 }
+};
+
+INSTANTIATE_TEST_CASE_P(
+    smoke_LPT,
+    ConcatWithIntermediatePrecisionSelectionTransformation,
+    ::testing::Combine(
+        ::testing::ValuesIn(precisions),
+        ::testing::ValuesIn(shapes),
+        ::testing::ValuesIn(testValues)),
+    ConcatWithIntermediatePrecisionSelectionTransformation::getTestCaseName);
+}  // namespace
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/simple_low_precision_transformer.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/simple_low_precision_transformer.cpp
index 8ee17c8e39b966..3c48d56be5b099 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/simple_low_precision_transformer.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/simple_low_precision_transformer.cpp
@@ -49,19 +49,41 @@ bool SimpleLowPrecisionTransformer::isPrecisionPreserved(const std::shared_ptr<n
 }
 
 void SimpleLowPrecisionTransformer::transform(std::shared_ptr<ngraph::Function>& function) {
+    // initialization
+    for (auto it : branchSpecificTransformations) {
+        ngraph::pass::low_precision::LayerTransformationPtr transformation = it.second;
+        transformation->setParamsManager(this);
+        transformation->setLayerTransformationsManager(this);
+    }
+
+    for (auto it : transformations) {
+        ngraph::pass::low_precision::LayerTransformationPtr transformation = it.second;
+        transformation->setParamsManager(this);
+        transformation->setLayerTransformationsManager(this);
+    }
+
+    // transformation
     {
         ngraph::pass::low_precision::TypeRelaxedReplacer pass;
         pass.run_on_function(function);
     }
 
     ngraph::pass::low_precision::TransformationContext context(function);
-    GraphRewrite pass;
-    for (auto it : transformations) {
-        ngraph::pass::low_precision::LayerTransformationPtr transformation = it.second;
+    {
+        GraphRewrite pass;
+        for (auto it : branchSpecificTransformations) {
+            ngraph::pass::low_precision::LayerTransformationPtr transformation = it.second;
+            transformation->registerMatcherIn(pass, context);
+        }
+        pass.run_on_function(function);
+    }
 
-        transformation->setParamsManager(this);
-        transformation->setLayerTransformationsManager(this);
-        transformation->registerMatcherIn(pass, context);
+    {
+        GraphRewrite pass;
+        for (auto it : transformations) {
+            ngraph::pass::low_precision::LayerTransformationPtr transformation = it.second;
+            transformation->registerMatcherIn(pass, context);
+        }
+        pass.run_on_function(function);
     }
-    pass.run_on_function(function);
 }
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/simple_low_precision_transformer.hpp b/inference-engine/tests/functional/inference_engine/lp_transformations/simple_low_precision_transformer.hpp
index b4bf3a9c9787a7..c9582adf0f0115 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/simple_low_precision_transformer.hpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/simple_low_precision_transformer.hpp
@@ -28,9 +28,22 @@ class SimpleLowPrecisionTransformer : public
     bool isQuantized(const std::shared_ptr<ngraph::Node>& layer) const noexcept override;
     bool isPrecisionPreserved(const std::shared_ptr<ngraph::Node>& layer) const noexcept override;
 
+    template <class T, class Operation>
+    ngraph::pass::low_precision::LayerTransformationPtr addBranchSpecific(const ngraph::pass::low_precision::LayerTransformation::Params& params) {
+        const std::string typeName = ngraph::pass::low_precision::LowPrecisionTransformations::getType<Operation>();
+
+        const auto it = branchSpecificTransformations.find(typeName);
+        if (it != branchSpecificTransformations.end()) {
+            branchSpecificTransformations.erase(it);
+        }
+
+        auto transformation = std::make_shared<T>(params);
+        branchSpecificTransformations.emplace(typeName, transformation);
+        return transformation;
+    }
+
     template <class T, class Operation>
     ngraph::pass::low_precision::LayerTransformationPtr add(const ngraph::pass::low_precision::LayerTransformation::Params& params) {
-        // const std::string typeName = typeid(ngraph::op::TypeRelaxed<Operation>).name();
         const std::string typeName = ngraph::pass::low_precision::LowPrecisionTransformations::getType<Operation>();
 
         const auto it = transformations.find(typeName);
@@ -46,5 +59,6 @@ class SimpleLowPrecisionTransformer : public
     void transform(std::shared_ptr<ngraph::Function>& function);
 
 private:
+    std::map<std::string, ngraph::pass::low_precision::LayerTransformationPtr> branchSpecificTransformations;
     std::map<std::string, ngraph::pass::low_precision::LayerTransformationPtr> transformations;
 };
diff --git a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/include/lpt_ngraph_functions/concat_function.hpp b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/include/lpt_ngraph_functions/concat_function.hpp
index c0c1686ca5521c..f70f653efe2f8c 100644
--- a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/include/lpt_ngraph_functions/concat_function.hpp
+++ b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/include/lpt_ngraph_functions/concat_function.hpp
@@ -51,6 +51,12 @@ class ConcatFunction {
         const FakeQuantizeOnData& fqOnData1,
         const FakeQuantizeOnData& fqOnData2);
 
+    static std::shared_ptr<ngraph::Function> getOriginalWithIntermediateAvgPool(
+        const ngraph::element::Type precision,
+        const ngraph::Shape& inputShape,
+        const FakeQuantizeOnData& fqOnData1,
+        const FakeQuantizeOnData& fqOnData2);
+
     static std::shared_ptr<ngraph::Function> getOriginalWithSplitedIntermediate(
         const ngraph::element::Type precision,
         const ngraph::Shape& inputShape,
@@ -134,6 +140,7 @@ class ConcatFunction {
         const std::string& neighborType,
         const std::string& additionalLayer);
 
+    // TODO: refactor: dequantizationBefore2 <=> dequantizationOperations2
     static std::shared_ptr<ngraph::Function> getReferenceWithIntermediate(
         const ngraph::element::Type precision,
         const ngraph::Shape& inputShape,
@@ -142,6 +149,18 @@ class ConcatFunction {
         const FakeQuantizeOnData& fqOnData2,
         const ngraph::element::Type precisionBeforeOp,
         const DequantizationOperations& dequantizationBefore1,
+        const DequantizationOperations& dequantizationOperations2,
+        const ngraph::element::Type precisionAfterOperation,
+        const DequantizationOperations& dequantizationOperations1,
+        const DequantizationOperations& dequantizationBefore2);
+
+    static std::shared_ptr<ngraph::Function> getReferenceWithIntermediateAvgPool(
+        const ngraph::element::Type precision,
+        const ngraph::Shape& inputShape,
+        const FakeQuantizeOnData& fqOnData1,
+        const FakeQuantizeOnData& fqOnData2,
+        const ngraph::element::Type precisionBeforeOp,
+        const DequantizationOperations& dequantizationBefore1,
         const DequantizationOperations& dequantizationBefore2,
         const ngraph::element::Type precisionAfterOperation,
         const DequantizationOperations& dequantizationOperations1,
diff --git a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/concat_function.cpp b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/concat_function.cpp
index 15108abb73e3c8..37387977eb7308 100644
--- a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/concat_function.cpp
+++ b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/concat_function.cpp
@@ -272,6 +272,58 @@ std::shared_ptr<ngraph::Function> ConcatFunction::getOriginalWithIntermediate(
     return function;
 }
 
+std::shared_ptr<ngraph::Function> ConcatFunction::getOriginalWithIntermediateAvgPool(
+    const ngraph::element::Type precision,
+    const ngraph::Shape& inputShape,
+    const FakeQuantizeOnData& fqOnData1,
+    const FakeQuantizeOnData& fqOnData2) {
+    const std::vector<size_t> inputShape1 = { inputShape[0], inputShape[1], inputShape[2] - 2, inputShape[3] - 2 };
+
+    const auto input1 = std::make_shared<ngraph::opset1::Parameter>(precision, ngraph::Shape(inputShape1));
+    input1->set_friendly_name("input1");
+    const auto fakeQuantize1 = makeFakeQuantize(input1, precision, fqOnData1);
+    fakeQuantize1->set_friendly_name("fakeQuantize1");
+
+    const std::vector<size_t> inputShape2 = { inputShape[0], inputShape[1], inputShape[2], inputShape[3] };
+    const auto input2 = std::make_shared<ngraph::opset1::Parameter>(precision, ngraph::Shape(inputShape2));
+    input2->set_friendly_name("input2");
+
+    const auto fakeQuantize2 = makeFakeQuantize(input2, precision, fqOnData2);
+    fakeQuantize2->set_friendly_name("fakeQuantize2");
+
+    std::shared_ptr<Node> intermediateOp = makeMaxPool(fakeQuantize2->output(0), { 3, 3 });
+    intermediateOp->set_friendly_name("intermediate");
+
+    const std::shared_ptr<ngraph::opset1::Concat> concat = std::make_shared<ngraph::opset1::Concat>(
+        ngraph::OutputVector{ fakeQuantize1->output(0), intermediateOp->output(0) }, 1);
+    concat->set_friendly_name("concat");
+
+    auto& rtInfo = concat->get_rt_info();
+    rtInfo["Variant::std::string"] = std::make_shared<VariantWrapper<std::string>>("concat");
+
+    std::shared_ptr<Node> parent2 = std::make_shared<ngraph::opset1::AvgPool>(
+        intermediateOp,
+        Strides{ 1, 1 },
+        Shape{ 1, 1 },
+        Shape{ 0, 0 },
+        Shape{ 2, 2 },
+        true,
+        op::RoundingType::FLOOR);
+    parent2->set_friendly_name("avgPool");
+
+    ngraph::ResultVector results {
+        std::make_shared<ngraph::opset1::Result>(concat),
+        std::make_shared<ngraph::opset1::Result>(parent2)
+    };
+
+    std::shared_ptr<ngraph::Function> function = std::make_shared<ngraph::Function>(
+        results,
+        ngraph::ParameterVector{ input1, input2 },
+        "ConcatWithIntermediateTransformation");
+
+    return function;
+}
+
 std::shared_ptr<ngraph::Function> ConcatFunction::getOriginalWithSplitedIntermediate(
     const ngraph::element::Type precision,
     const ngraph::Shape& inputShape,
@@ -1056,6 +1108,77 @@ std::shared_ptr<ngraph::Function> ConcatFunction::getReferenceWithIntermediate(
     return function;
 }
 
+std::shared_ptr<ngraph::Function> ConcatFunction::getReferenceWithIntermediateAvgPool(
+    const ngraph::element::Type precision,
+    const ngraph::Shape& inputShape,
+    const FakeQuantizeOnData& fqOnData1,
+    const FakeQuantizeOnData& fqOnData2,
+    const ngraph::element::Type precisionBeforeOp,
+    const DequantizationOperations& dequantizationBefore1,
+    const DequantizationOperations& dequantizationBefore2,
+    const ngraph::element::Type precisionAfterOperation,
+    const DequantizationOperations& dequantizationAfter1,
+    const DequantizationOperations& dequantizationAfter2) {
+    const std::vector<size_t> inputShape1 = { inputShape[0], inputShape[1], inputShape[2] - 2, inputShape[3] - 2};
+    const auto input1 = std::make_shared<ngraph::opset1::Parameter>(precision, ngraph::Shape(inputShape1));
+    input1->set_friendly_name("input1");
+
+    const auto fakeQuantize1 = makeFakeQuantizeTypeRelaxed(input1, precision, fqOnData1);
+    low_precision::NetworkHelper::setOutDataPrecisionForTypeRelaxed(fakeQuantize1, precisionBeforeOp);
+    fakeQuantize1->set_friendly_name("fakeQuantize1");
+    const auto deqBefore1 = makeDequantization(fakeQuantize1, dequantizationBefore1);
+
+    const std::vector<size_t> inputShape2 = { inputShape[0], inputShape[1], inputShape[2], inputShape[3] };
+    const auto input2 = std::make_shared<ngraph::opset1::Parameter>(precision, ngraph::Shape(inputShape2));
+    input2->set_friendly_name("input2");
+
+    const auto fakeQuantize2 = makeFakeQuantizeTypeRelaxed(input2, precision, fqOnData2);
+    low_precision::NetworkHelper::setOutDataPrecisionForTypeRelaxed(fakeQuantize2, precisionBeforeOp);
+    fakeQuantize2->set_friendly_name("fakeQuantize2");
+    const auto deqBefore2 = makeDequantization(fakeQuantize2, dequantizationBefore2);
+
+    std::shared_ptr<Node> intermediateOp  = makeMaxPool(deqBefore2, { 3, 3 });
+    intermediateOp->set_friendly_name("intermediate");
+
+    const std::shared_ptr<ngraph::opset1::Concat> concat = std::make_shared<ngraph::opset1::Concat>(
+        ngraph::OutputVector { deqBefore1, intermediateOp },
+        1);
+    concat->set_friendly_name("concat");
+    low_precision::NetworkHelper::setOutDataPrecision(concat, precisionAfterOperation);
+
+    auto& rtInfo = concat->get_rt_info();
+    rtInfo["Variant::std::string"] = std::make_shared<VariantWrapper<std::string>>("concat");
+
+    const std::shared_ptr<ngraph::Node> parent1 = makeDequantization(concat, dequantizationAfter1);
+    parent1->set_friendly_name("concat");
+
+    std::shared_ptr<Node> parent2 = std::make_shared<ngraph::op::TypeRelaxed<ngraph::opset1::AvgPool>>(
+        std::vector<ngraph::element::Type>{ element::f32, element::f32 },
+        std::vector<ngraph::element::Type>{ element::f32 },
+        ngraph::op::TemporaryReplaceOutputType(intermediateOp, element::f32).get(),
+        Strides{ 1, 1 },
+        Shape{ 1, 1 },
+        Shape{ 0, 0 },
+        Shape{ 2, 2 },
+        true,
+        op::RoundingType::FLOOR);
+    parent2->set_friendly_name("avgPool");
+
+    parent2 = makeDequantization(parent2, dequantizationAfter2);
+
+    ngraph::ResultVector results {
+        std::make_shared<ngraph::opset1::Result>(parent1),
+        std::make_shared<ngraph::opset1::Result>(parent2)
+    };
+
+    std::shared_ptr<ngraph::Function> function = std::make_shared<ngraph::Function>(
+        results,
+        ngraph::ParameterVector{ input1, input2 },
+        "ConcatWithIntermediateTransformation");
+
+    return function;
+}
+
 std::shared_ptr<ngraph::Function> ConcatFunction::getReferenceWithSplitedIntermediate(
     const ngraph::element::Type precision,
     const ngraph::Shape& inputShape,

From de2a163363eb885dbfe7367d4c6349204d1641de Mon Sep 17 00:00:00 2001
From: Svetlana Dolinina <svetlana.a.dolinina@intel.com>
Date: Tue, 8 Jun 2021 16:09:04 +0300
Subject: [PATCH 032/178] Error during Caffe model conversion with Python3.8
 (#6056)

* added one more possible reason for Caffe error during caffemodel parser in code + add error description to FAQ
Also added MxNet error to FAQ based on Kate Generalova wording

* review fixes

* wording polishing

* wording polishing

* review fixes

* review fixes

* Update docs/MO_DG/prepare_model/Model_Optimizer_FAQ.md

Co-authored-by: Tatiana Savina <tatiana.savina@intel.com>

* Update docs/MO_DG/prepare_model/Model_Optimizer_FAQ.md

Co-authored-by: Tatiana Savina <tatiana.savina@intel.com>

* Update docs/MO_DG/prepare_model/Model_Optimizer_FAQ.md

Co-authored-by: Tatiana Savina <tatiana.savina@intel.com>

* Update docs/MO_DG/prepare_model/Model_Optimizer_FAQ.md

Co-authored-by: Tatiana Savina <tatiana.savina@intel.com>

* review fixes

Co-authored-by: Tatiana Savina <tatiana.savina@intel.com>
---
 docs/MO_DG/prepare_model/Model_Optimizer_FAQ.md | 14 +++++++++++++-
 model-optimizer/mo/front/caffe/loader.py        |  8 +++++++-
 2 files changed, 20 insertions(+), 2 deletions(-)

diff --git a/docs/MO_DG/prepare_model/Model_Optimizer_FAQ.md b/docs/MO_DG/prepare_model/Model_Optimizer_FAQ.md
index f9aef04a0a9561..bb599cf93b5632 100644
--- a/docs/MO_DG/prepare_model/Model_Optimizer_FAQ.md
+++ b/docs/MO_DG/prepare_model/Model_Optimizer_FAQ.md
@@ -627,4 +627,16 @@ It means that you trying to convert the topology which contains '_contrib_box_nm
   });
 </script>
 
-\endhtmlonly
\ No newline at end of file
+\endhtmlonly
+
+#### 103. What does the message "ModelOptimizer is not able to parse *.caffemodel" mean? <a name="question-103"></a>
+
+If a '*.caffemodel' file exists and it is correct, the error possibly occured due to the use of Python protobuf implementation. In some cases, it shows error message during model parsing, for example: "'utf-8' codec can't decode byte 0xe0 in position 4: invalid continuation byte in field: mo_caffe.SpatialTransformerParameter.transform_type". You can either use Python 3.6/3.7 or build 'cpp' implementation of protobuf yourself for your version of Python. For the complete instructions about building `protobuf` from sources, see the appropriate section in [Converting a Model to Intermediate Representation](Config_Model_Optimizer.md).
+
+#### 104. What does the message "SyntaxError: 'yield' inside list comprehension" during MxNet\* model conversion mean? <a name="question-104"></a>
+
+The issue "SyntaxError: 'yield' inside list comprehension" might occur during converting MXNet\* models (mobilefacedet-v1-mxnet, brain-tumor-segmentation-0001) on Windows* platform with Python* 3.8 environment. This issue is caused by API changes for `yield expression` in Python 3.8.
+The following workarounds are suggested to resolve this issue:
+1. Use Python 3.6/3.7 to convert MXNet\* models on Windows
+2. Update MXNet: pip install mxnet=1.7.0.post2
+Note that you might have conflicts between previously installed PyPI dependencies.
\ No newline at end of file
diff --git a/model-optimizer/mo/front/caffe/loader.py b/model-optimizer/mo/front/caffe/loader.py
index 2ffca364fb6213..14497c6108d9f7 100644
--- a/model-optimizer/mo/front/caffe/loader.py
+++ b/model-optimizer/mo/front/caffe/loader.py
@@ -130,10 +130,16 @@ def load_caffe_proto_model(caffe_pb2, proto_path: str, model_path: [str, None] =
                 map = mmap.mmap(infile.fileno(), 0, access=mmap.ACCESS_READ)
                 model.MergeFromString(map)
     except Exception as e:
+        third_point = ''
+        if api_implementation._implementation_type == 'python':
+            third_point = '      3. Python protobuf implementation was used. Some models can\'t be converted ' + \
+                          ' in this configuration. Please, use Python version with existing cpp implementation of ' + \
+                          'protobuf library or build it by yourself\n' + refer_to_faq_msg(103)
         log.error('Exception message: {}\n\n'.format(e) +
                   '    Possible reasons:\n' +
                   '      1. {} does not exist\n'.format(model_path) +
-                  '      2. {} does not have a valid structure\n'.format(model_path), extra={'framework_error': True})
+                  '      2. {} does not have a valid structure\n'.format(model_path) + third_point,
+                  extra={'framework_error': True})
         raise FrameworkError('Model Optimizer is not able to parse {}'.format(model_path)) from e
 
     return proto, model

From a36d6a0f0678fd0c0d637df9ea3f1b34682643cc Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Tue, 8 Jun 2021 19:33:51 +0300
Subject: [PATCH 033/178] Used setMeanImageForChannel (#6076)

---
 .../src/readers/ir_reader/ie_ir_parser.cpp            | 11 +++++------
 .../plugin/shared/include/behavior/set_preprocess.hpp | 11 ++++++-----
 2 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/inference-engine/src/readers/ir_reader/ie_ir_parser.cpp b/inference-engine/src/readers/ir_reader/ie_ir_parser.cpp
index d26132abf2ea22..892c2e6bae4088 100644
--- a/inference-engine/src/readers/ir_reader/ie_ir_parser.cpp
+++ b/inference-engine/src/readers/ir_reader/ie_ir_parser.cpp
@@ -1011,8 +1011,6 @@ void V10Parser::parsePreProcess(
     if (!meanSegmentPrecision || meanSegmentPrecision == Precision::MIXED)
         IE_THROW() << "mean blob defined without specifying precision.";
 
-    InferenceEngine::PreProcessChannel::Ptr preProcessChannel;
-
     int lastChanNo = -1;
     std::unordered_set<int> idsForMeanImage;
 
@@ -1022,7 +1020,6 @@ void V10Parser::parsePreProcess(
             IE_THROW() << "Pre-process channel id invalid: " << chanNo;
         }
         lastChanNo = chanNo;
-        preProcessChannel = pp[chanNo];
 
         auto meanNode = chan.child("mean");
         if (!meanNode.empty()) {
@@ -1038,13 +1035,15 @@ void V10Parser::parsePreProcess(
                                        << " extpecting " << width << " x " << height << " x "
                                        << meanSegmentPrecision.size();
                 }
-                preProcessChannel->meanData = make_blob_with_precision(
+                auto meanData = make_blob_with_precision(
                     TensorDesc(meanSegmentPrecision, {height, width}, Layout::HW));
-                preProcessChannel->meanData->allocate();
-                auto lockedMem = preProcessChannel->meanData->buffer();
+                meanData->allocate();
+                auto lockedMem = meanData->buffer();
                 char* data = lockedMem.as<char*>();
                 uint8_t* src_data = weights->cbuffer().as<uint8_t*>() + offset;
                 memcpy(data, src_data, size);
+
+                pp.setMeanImageForChannel(meanData, chanNo);
             }
         }
     }
diff --git a/inference-engine/tests/functional/plugin/shared/include/behavior/set_preprocess.hpp b/inference-engine/tests/functional/plugin/shared/include/behavior/set_preprocess.hpp
index de442f12c215a8..ff2948668588b4 100644
--- a/inference-engine/tests/functional/plugin/shared/include/behavior/set_preprocess.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/behavior/set_preprocess.hpp
@@ -157,16 +157,17 @@ TEST_P(PreprocessTest, SetMeanImagePreProcessSetBlob) {
     auto &preProcess = cnnNet.getInputsInfo().begin()->second->getPreProcess();
     preProcess.init(3);
     for (size_t i = 0; i < 3; i++) {
-        preProcess[i]->meanData = make_blob_with_precision(InferenceEngine::TensorDesc(InferenceEngine::Precision::FP32,
-                                                                                       {10, 10},
-                                                                                       InferenceEngine::Layout::HW));
-        preProcess[i]->meanData->allocate();
-        auto lockedMem = preProcess[i]->meanData->buffer();
+        auto meanData = make_blob_with_precision(
+            InferenceEngine::TensorDesc(InferenceEngine::Precision::FP32,  {10, 10},
+            InferenceEngine::Layout::HW));
+        meanData->allocate();
+        auto lockedMem = meanData->buffer();
         auto* data = lockedMem.as<float *>();
         for (size_t j = 0; j < 100; j++) {
             data[j] = 0;
             data[j] -= i * 100 + j;
         }
+        ASSERT_NO_THROW(preProcess.setMeanImageForChannel(meanData, i));
     }
     preProcess.setVariant(InferenceEngine::MEAN_IMAGE);
     // Load CNNNetwork to target plugins

From d171b5c4b76151b98d7935193e61549b75f4796b Mon Sep 17 00:00:00 2001
From: Ilya Churaev <ilya.churaev@intel.com>
Date: Tue, 8 Jun 2021 19:55:36 +0300
Subject: [PATCH 034/178] Removed version.in.hpp from nGraph package (#6081)

---
 ngraph/core/CMakeLists.txt | 1 +
 1 file changed, 1 insertion(+)

diff --git a/ngraph/core/CMakeLists.txt b/ngraph/core/CMakeLists.txt
index fa3a91c34f8200..72272046b01e9d 100644
--- a/ngraph/core/CMakeLists.txt
+++ b/ngraph/core/CMakeLists.txt
@@ -105,6 +105,7 @@ install(DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/include/
     FILES_MATCHING
         PATTERN "*.hpp"
         PATTERN "*.h"
+        PATTERN "*version.in.hpp" EXCLUDE
 )
 install(FILES ${CMAKE_CURRENT_BINARY_DIR}/include/ngraph/version.hpp
     DESTINATION ${NGRAPH_INSTALL_INCLUDE}/ngraph

From 246932a5d358364104b934b8c07f7939407f5319 Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Tue, 8 Jun 2021 20:24:11 +0300
Subject: [PATCH 035/178] Py code style (#6086)

* GitHub CI: Enable Python for Code style

* Update formatting

* Fix

* Add cython

* Add upgrade pip

* Update to src/requirements-dev.txt

* Fixed clang code style

Co-authored-by: azhogov <alexander.zhogov@intel.com>
---
 .github/org_control/check_pr.py                              | 2 +-
 .github/workflows/code_style.yml                             | 5 ++++-
 .../python/src/openvino/inference_engine/CMakeLists.txt      | 3 ++-
 .../src/openvino/offline_transformations/CMakeLists.txt      | 3 ++-
 .../ie_bridges/python/src/openvino/test_utils/CMakeLists.txt | 3 ++-
 ngraph/python/src/pyngraph/util.cpp                          | 2 +-
 6 files changed, 12 insertions(+), 6 deletions(-)

diff --git a/.github/org_control/check_pr.py b/.github/org_control/check_pr.py
index e0b48832ead75c..85e5b4e34c00f4 100644
--- a/.github/org_control/check_pr.py
+++ b/.github/org_control/check_pr.py
@@ -229,7 +229,7 @@ def main():
     if wrong_pulls:
         for pull_number, wrong_commits in wrong_pulls.items():
             print(
-                f"\nERROR: Remove or replace wrong commits in the PR {pull_number}:\n    ",
+                f"\nERROR: Remove or replace wrong commits in the PR {pull_number}:\n   ",
                 "\n    ".join(wrong_commits),
             )
         print(
diff --git a/.github/workflows/code_style.yml b/.github/workflows/code_style.yml
index b538a1793397ef..607fe2cb64ae1a 100644
--- a/.github/workflows/code_style.yml
+++ b/.github/workflows/code_style.yml
@@ -15,14 +15,17 @@ jobs:
       - name: Install dependencies
         run: |
           sudo apt --assume-yes install libusb-1.0-0-dev
+          python3 -m pip install --upgrade pip
           python3 -m pip install -r ./inference-engine/ie_bridges/python/requirements.txt
+          # Add for -DENABLE_PYTHON=ON, no cython
+          python3 -m pip install -r ./inference-engine/ie_bridges/python/src/requirements-dev.txt
 
       # Run cmake with -DENABLE_PROFILING_ITT=ON -DSELECTIVE_BUILD=COLLECT in order to enable codestyle check for ITT collector
       - name: CMake
         run: |
           mkdir build
           cd build
-          cmake -DENABLE_PROFILING_ITT=ON -DSELECTIVE_BUILD=COLLECT ..
+          cmake -DENABLE_PYTHON=ON -DENABLE_PROFILING_ITT=ON -DSELECTIVE_BUILD=COLLECT ..
 
       - name: Check code style
         run: cmake --build build --target clang_format_check_all
diff --git a/inference-engine/ie_bridges/python/src/openvino/inference_engine/CMakeLists.txt b/inference-engine/ie_bridges/python/src/openvino/inference_engine/CMakeLists.txt
index 1f623fb48336c0..17b8bf5b9b513f 100644
--- a/inference-engine/ie_bridges/python/src/openvino/inference_engine/CMakeLists.txt
+++ b/inference-engine/ie_bridges/python/src/openvino/inference_engine/CMakeLists.txt
@@ -77,4 +77,5 @@ install(PROGRAMS __init__.py
         DESTINATION ${PYTHON_BRIDGE_CPACK_PATH}/${PYTHON_VERSION}/openvino/inference_engine
         COMPONENT ${PYTHON_VERSION})
 
-add_clang_format_target(${TARGET_NAME}_clang FOR_TARGETS ${TARGET_NAME})
\ No newline at end of file
+add_clang_format_target(${TARGET_NAME}_clang FOR_TARGETS ${TARGET_NAME}
+                        EXCLUDE_PATTERNS ".*\\.cxx;.*\\.pxd;.*\\.pyx")
diff --git a/inference-engine/ie_bridges/python/src/openvino/offline_transformations/CMakeLists.txt b/inference-engine/ie_bridges/python/src/openvino/offline_transformations/CMakeLists.txt
index c6315336ba2929..27c9e7bf898257 100644
--- a/inference-engine/ie_bridges/python/src/openvino/offline_transformations/CMakeLists.txt
+++ b/inference-engine/ie_bridges/python/src/openvino/offline_transformations/CMakeLists.txt
@@ -42,7 +42,8 @@ if(CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
     target_compile_options(${TARGET_NAME} PRIVATE "-Wno-error=register")
 endif()
 
-add_clang_format_target(${TARGET_NAME}_clang FOR_TARGETS ${TARGET_NAME})
+add_clang_format_target(${TARGET_NAME}_clang FOR_TARGETS ${TARGET_NAME}
+                        EXCLUDE_PATTERNS ".*\\.cxx;.*\\.pxd;.*\\.pyx")
 
 # perform copy
 add_custom_command(TARGET ${TARGET_NAME}
diff --git a/inference-engine/ie_bridges/python/src/openvino/test_utils/CMakeLists.txt b/inference-engine/ie_bridges/python/src/openvino/test_utils/CMakeLists.txt
index 504125d9823c1a..8367f941d9f793 100644
--- a/inference-engine/ie_bridges/python/src/openvino/test_utils/CMakeLists.txt
+++ b/inference-engine/ie_bridges/python/src/openvino/test_utils/CMakeLists.txt
@@ -48,4 +48,5 @@ add_custom_command(TARGET ${TARGET_NAME}
     COMMAND ${CMAKE_COMMAND} -E copy ${PYTHON_BRIDGE_SRC_ROOT}/src/openvino/test_utils/__init__.py ${CMAKE_LIBRARY_OUTPUT_DIRECTORY}/__init__.py
 )
 
-add_clang_format_target(${TARGET_NAME}_clang FOR_TARGETS ${TARGET_NAME})
\ No newline at end of file
+add_clang_format_target(${TARGET_NAME}_clang FOR_TARGETS ${TARGET_NAME}
+                        EXCLUDE_PATTERNS ".*\\.cxx;.*\\.pxd;.*\\.pyx")
\ No newline at end of file
diff --git a/ngraph/python/src/pyngraph/util.cpp b/ngraph/python/src/pyngraph/util.cpp
index 5178e84fe9081b..69c0df89dcd646 100644
--- a/ngraph/python/src/pyngraph/util.cpp
+++ b/ngraph/python/src/pyngraph/util.cpp
@@ -4,8 +4,8 @@
 
 #include <pybind11/numpy.h>
 
-#include "pyngraph/util.hpp"
 #include "ngraph/validation_util.hpp"
+#include "pyngraph/util.hpp"
 
 namespace py = pybind11;
 

From 57850f0a876a8dfdab0a1fec575c948ca4e3bf09 Mon Sep 17 00:00:00 2001
From: Egor Duplensky <egor.duplenskii@intel.com>
Date: Tue, 8 Jun 2021 22:39:08 +0300
Subject: [PATCH 036/178] [CPU] Allow exec graph serialization to cout or xml
 via env variable (#5967)

---
 inference-engine/src/mkldnn_plugin/config.h   |  8 +++-
 .../src/mkldnn_plugin/mkldnn_edge.cpp         | 14 ++++--
 .../src/mkldnn_plugin/mkldnn_edge.h           | 10 ++--
 .../src/mkldnn_plugin/mkldnn_graph.cpp        | 31 ++++--------
 .../src/mkldnn_plugin/mkldnn_graph.h          |  7 ++-
 .../src/mkldnn_plugin/mkldnn_graph_dumper.cpp | 47 +++++++++++++++++++
 .../src/mkldnn_plugin/mkldnn_graph_dumper.h   |  5 +-
 .../src/mkldnn_plugin/utils/README.md         | 19 ++++++++
 .../mkldnn_plugin/utils/debug_capabilities.h  | 45 ++++++++++++++++--
 .../src/mkldnn_plugin/utils/node_dumper.cpp   | 34 +++++++-------
 .../src/mkldnn_plugin/utils/node_dumper.h     |  7 ++-
 11 files changed, 168 insertions(+), 59 deletions(-)

diff --git a/inference-engine/src/mkldnn_plugin/config.h b/inference-engine/src/mkldnn_plugin/config.h
index 01eb0e23c5ee1a..54336d58495276 100644
--- a/inference-engine/src/mkldnn_plugin/config.h
+++ b/inference-engine/src/mkldnn_plugin/config.h
@@ -4,9 +4,11 @@
 
 #pragma once
 
+#include <threading/ie_istreams_executor.hpp>
+#include "utils/debug_capabilities.h"
+
 #include <string>
 #include <map>
-#include <threading/ie_istreams_executor.hpp>
 
 namespace MKLDNNPlugin {
 
@@ -35,6 +37,10 @@ struct Config {
     bool manualEnforceBF16 = false;
 #endif
 
+#ifdef CPU_DEBUG_CAPS
+    DebugCaps::Config debugCaps;
+#endif
+
     void readProperties(const std::map<std::string, std::string> &config);
     void updateProperties();
     std::map<std::string, std::string> _config;
diff --git a/inference-engine/src/mkldnn_plugin/mkldnn_edge.cpp b/inference-engine/src/mkldnn_plugin/mkldnn_edge.cpp
index b5ff60efed0af9..1415dc1ae95e20 100644
--- a/inference-engine/src/mkldnn_plugin/mkldnn_edge.cpp
+++ b/inference-engine/src/mkldnn_plugin/mkldnn_edge.cpp
@@ -32,7 +32,7 @@ bool MKLDNNEdge::isUseExternalMemory() const {
     return externalMemoryPtr;
 }
 
-bool MKLDNNEdge::isDropped() {
+bool MKLDNNEdge::isDropped() const {
     bool not_in_parent = true;
     bool not_in_child = true;
 
@@ -124,6 +124,10 @@ void MKLDNNEdge::reuse(MKLDNNMemoryPtr ptr) {
     status = Status::Allocated;
 }
 
+const InferenceEngine::TensorDesc& MKLDNNEdge::getInputDescRO() const {
+    return inputDesc;
+}
+
 InferenceEngine::TensorDesc MKLDNNEdge::getInputDesc() {
     if (inputDesc.getLayout() == InferenceEngine::Layout::ANY) {
         inputDesc = getSpecifiedInputDesc({});
@@ -131,6 +135,10 @@ InferenceEngine::TensorDesc MKLDNNEdge::getInputDesc() {
     return inputDesc;
 }
 
+const InferenceEngine::TensorDesc& MKLDNNEdge::getOutputDescRO() const {
+    return outputDesc;
+}
+
 InferenceEngine::TensorDesc MKLDNNEdge::getOutputDesc() {
     if (outputDesc.getLayout() == InferenceEngine::Layout::ANY) {
         outputDesc = getSpecifiedOutputDesc({});
@@ -145,11 +153,11 @@ InferenceEngine::TensorDesc MKLDNNEdge::getDesc() {
     return getInputDesc();
 }
 
-int MKLDNNEdge::getInputNum() {
+int MKLDNNEdge::getInputNum() const {
     return parent_port;
 }
 
-int MKLDNNEdge::getOutputNum() {
+int MKLDNNEdge::getOutputNum() const {
     return child_port;
 }
 
diff --git a/inference-engine/src/mkldnn_plugin/mkldnn_edge.h b/inference-engine/src/mkldnn_plugin/mkldnn_edge.h
index c9884caf56e823..63e2a16414d94f 100644
--- a/inference-engine/src/mkldnn_plugin/mkldnn_edge.h
+++ b/inference-engine/src/mkldnn_plugin/mkldnn_edge.h
@@ -61,11 +61,11 @@ class MKLDNNEdge {
     MKLDNNMemoryPtr& getMemoryPtr();
 
     bool needReorder();
-    bool isDropped();
+    bool isDropped() const;
     bool isUseExternalMemory() const;
 
-    int getInputNum();
-    int getOutputNum();
+    int getInputNum() const;
+    int getOutputNum() const;
 
     void setChildPort(const size_t port) { child_port = port; }
 
@@ -73,10 +73,12 @@ class MKLDNNEdge {
     MKLDNNEdgePtr getSharedEdge() const;
     MKLDNNEdgePtr getSharedEdge(std::nothrow_t) const;
 
+    const InferenceEngine::TensorDesc& getInputDescRO() const;
+    const InferenceEngine::TensorDesc& getOutputDescRO() const;
+
 private:
     std::string name();
 
-private:
     std::weak_ptr<MKLDNNNode> parent;
     std::weak_ptr<MKLDNNNode> child;
     int parent_port;
diff --git a/inference-engine/src/mkldnn_plugin/mkldnn_graph.cpp b/inference-engine/src/mkldnn_plugin/mkldnn_graph.cpp
index b92afb8a9f0d4f..efc99bddb84ae3 100644
--- a/inference-engine/src/mkldnn_plugin/mkldnn_graph.cpp
+++ b/inference-engine/src/mkldnn_plugin/mkldnn_graph.cpp
@@ -78,7 +78,10 @@ void MKLDNNGraph::CreateGraph(NET &net, const MKLDNNExtensionManager::Ptr& extMg
 
     Replicate(net, extMgr);
     InitGraph();
+
     status = Ready;
+
+    ENABLE_CPU_DEBUG_CAP(serialize(*this));
 }
 
 template void MKLDNNGraph::CreateGraph(const std::shared_ptr<const ngraph::Function>&,
@@ -344,10 +347,6 @@ void MKLDNNGraph::InitGraph() {
         graphNode->cleanup();
     }
 #endif
-
-#if !defined(NDEBUG) && defined(PRINT_GRAPH_INFO)
-    printGraphInfo();
-#endif
     ExecuteConstantNodesOnly();
 }
 
@@ -809,7 +808,7 @@ void MKLDNNGraph::Infer(MKLDNNInferRequest* request, int batch) {
 
     mkldnn::stream stream(eng);
 
-    ENABLE_CPU_DEBUG_CAP(NodeDumper nd(infer_count));
+    ENABLE_CPU_DEBUG_CAP(NodeDumper nd(config.debugCaps, infer_count));
 
     for (int i = 0; i < graphNodes.size(); i++) {
         if (request != nullptr) {
@@ -954,6 +953,10 @@ void MKLDNNGraph::setConfig(const Config &cfg) {
     config = cfg;
 }
 
+const Config& MKLDNNGraph::getConfig() const {
+    return config;
+}
+
 void MKLDNNGraph::setProperty(const std::map<std::string, std::string>& properties) {
     config.readProperties(properties);
 }
@@ -1217,21 +1220,3 @@ void MKLDNNGraph::EnforceBF16() {
 InferenceEngine::CNNNetwork MKLDNNGraph::dump() const {
     return dump_graph_as_ie_ngraph_net(*this);
 }
-
-void MKLDNNGraph::printGraphInfo() const {
-    for (auto &graphNode : graphNodes) {
-        std::cout << "name: " << graphNode->getName() << " [ ";
-        if (graphNode->parentEdges.size() > 0) {
-            auto prnt_out_desc = graphNode->parentEdges[0].lock()->getOutputDesc();
-            std::cout << "in: " << prnt_out_desc.getPrecision().name()
-                      << "/l=" << prnt_out_desc.getLayout()
-                      << "; ";
-        }
-        if (graphNode->childEdges.size() > 0) {
-            auto chld_in_desc = graphNode->childEdges[0].lock()->getInputDesc();
-            std::cout << "out: " << chld_in_desc.getPrecision().name()
-                      << "/l=" << chld_in_desc.getLayout();
-        }
-        std::cout << " ]"  << std::endl;
-    }
-}
diff --git a/inference-engine/src/mkldnn_plugin/mkldnn_graph.h b/inference-engine/src/mkldnn_plugin/mkldnn_graph.h
index c3fcb0d5c9c635..1b54f71e88c1cd 100644
--- a/inference-engine/src/mkldnn_plugin/mkldnn_graph.h
+++ b/inference-engine/src/mkldnn_plugin/mkldnn_graph.h
@@ -39,6 +39,8 @@ class MKLDNNGraph {
     }
 
     void setConfig(const Config &cfg);
+    const Config& getConfig() const;
+
     void setProperty(const std::map<std::string, std::string> &properties);
     Config getProperty() const;
 
@@ -59,6 +61,10 @@ class MKLDNNGraph {
 
     void Infer(MKLDNNInferRequest* request = nullptr, int batch = -1);
 
+    const std::vector<MKLDNNNodePtr>& GetNodes() const {
+        return graphNodes;
+    }
+
     std::vector<MKLDNNNodePtr>& GetNodes() {
         return graphNodes;
     }
@@ -219,7 +225,6 @@ class MKLDNNGraph {
 
 private:
     void EnforceBF16();
-    void printGraphInfo() const;
 };
 
 }  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/mkldnn_graph_dumper.cpp b/inference-engine/src/mkldnn_plugin/mkldnn_graph_dumper.cpp
index 14d2f6a28ae7e8..ac4bfff6b6d2f0 100644
--- a/inference-engine/src/mkldnn_plugin/mkldnn_graph_dumper.cpp
+++ b/inference-engine/src/mkldnn_plugin/mkldnn_graph_dumper.cpp
@@ -5,9 +5,11 @@
 #include "mkldnn_graph_dumper.h"
 #include <ie_ngraph_utils.hpp>
 #include "exec_graph_info.hpp"
+#include "ie_common.h"
 #include "mkldnn_debug.h"
 #include <ngraph/variant.hpp>
 #include "ngraph/ngraph.hpp"
+#include "utils/debug_capabilities.h"
 
 #include <vector>
 #include <string>
@@ -18,6 +20,9 @@ using namespace InferenceEngine;
 
 namespace MKLDNNPlugin {
 
+void serializeToCout(const MKLDNNGraph &graph);
+void serializeToXML(const MKLDNNGraph &graph, const std::string& path);
+
 namespace {
 
 std::map<std::string, std::string> extract_node_metadata(const MKLDNNNodePtr &node) {
@@ -207,4 +212,46 @@ InferenceEngine::CNNNetwork dump_graph_as_ie_ngraph_net(const MKLDNNGraph &graph
     return net;
 }
 
+#ifdef CPU_DEBUG_CAPS
+void serialize(const MKLDNNGraph &graph) {
+    const std::string& path = graph.getConfig().debugCaps.execGraphPath;
+
+    if (path.empty())
+        return;
+
+    if (path == "cout")
+        serializeToCout(graph);
+    else if (!path.compare(path.size() - 4, 4, ".xml"))
+        serializeToXML(graph, path);
+    else
+        IE_THROW() << "Unknown serialize format. Should be either 'cout' or '*.xml'. Got " << path;
+}
+
+void serializeToXML(const MKLDNNGraph &graph, const std::string& path) {
+    if (path.empty())
+        return;
+
+    graph.dump().serialize(path);
+}
+
+void serializeToCout(const MKLDNNGraph &graph) {
+    for (const auto& node : graph.GetNodes()) {
+        std::cout << "name: " << node->getName() << " [ ";
+        if (!node->getParentEdges().empty()) {
+            const auto& parentEdge = *(node->getParentEdges()[0].lock());
+            const auto& prnt_out_desc = parentEdge.getOutputDescRO();
+            std::cout << "in: " << prnt_out_desc.getPrecision().name()
+                      << "/l=" << prnt_out_desc.getLayout()
+                      << "; ";
+        }
+        if (!node->getChildEdges().empty()) {
+            const auto& childEdge = *(node->getChildEdges()[0].lock());
+            const auto& chld_in_desc = childEdge.getInputDescRO();
+            std::cout << "out: " << chld_in_desc.getPrecision().name()
+                      << "/l=" << chld_in_desc.getLayout();
+        }
+        std::cout << " ]"  << std::endl;
+    }
+}
+#endif
 }  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/mkldnn_graph_dumper.h b/inference-engine/src/mkldnn_plugin/mkldnn_graph_dumper.h
index d954695baaa050..597568224f38f0 100644
--- a/inference-engine/src/mkldnn_plugin/mkldnn_graph_dumper.h
+++ b/inference-engine/src/mkldnn_plugin/mkldnn_graph_dumper.h
@@ -6,11 +6,14 @@
 
 #include "cpp/ie_cnn_network.h"
 #include "mkldnn_graph.h"
+#include "utils/debug_capabilities.h"
 
 #include <memory>
 
 namespace MKLDNNPlugin {
 
 InferenceEngine::CNNNetwork dump_graph_as_ie_ngraph_net(const MKLDNNGraph &graph);
-
+#ifdef CPU_DEBUG_CAPS
+void serialize(const MKLDNNGraph &graph);
+#endif // CPU_DEBUG_CAPS
 }  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/utils/README.md b/inference-engine/src/mkldnn_plugin/utils/README.md
index af50c8d5015fdc..d3b98f1cb48f2f 100644
--- a/inference-engine/src/mkldnn_plugin/utils/README.md
+++ b/inference-engine/src/mkldnn_plugin/utils/README.md
@@ -71,3 +71,22 @@ Example:
 ```sh
     OV_CPU_BLOB_DUMP_NODE_NAME=".+" binary ...
 ```
+
+## Graph serialization
+The functionality allows to serialize execution graph using environment variable:
+```sh
+    OV_CPU_EXEC_GRAPH_PATH=<path> binary ...
+```
+
+Possible serialization options:
+* cout
+
+    Serialize to console output
+* \<path\>.xml
+
+    Serialize graph into .xml and .bin files. Can be opened using, for example, *netron* app
+* \<path\>.dot
+
+    TBD. Serialize graph into .dot file. Can be inspected using, for example, *graphviz* tools.
+
+
diff --git a/inference-engine/src/mkldnn_plugin/utils/debug_capabilities.h b/inference-engine/src/mkldnn_plugin/utils/debug_capabilities.h
index 64af835064d8c7..be6e7a830c2c63 100644
--- a/inference-engine/src/mkldnn_plugin/utils/debug_capabilities.h
+++ b/inference-engine/src/mkldnn_plugin/utils/debug_capabilities.h
@@ -4,7 +4,44 @@
 #pragma once
 
 #ifdef CPU_DEBUG_CAPS
-#   define ENABLE_CPU_DEBUG_CAP(_x) _x;
-#else
-#   define ENABLE_CPU_DEBUG_CAP(_x)
-#endif
+
+#include <map>
+#include <string>
+#include <vector>
+
+#define ENABLE_CPU_DEBUG_CAP(_x) _x;
+
+namespace MKLDNNPlugin {
+namespace DebugCaps {
+
+class Config {
+public:
+    Config() {
+        readParam(blobDumpDir, "OV_CPU_BLOB_DUMP_DIR");
+        readParam(blobDumpFormat, "OV_CPU_BLOB_DUMP_FORMAT");
+        readParam(blobDumpNodeExecId, "OV_CPU_BLOB_DUMP_NODE_EXEC_ID");
+        readParam(blobDumpNodeType, "OV_CPU_BLOB_DUMP_NODE_TYPE");
+        readParam(blobDumpNodeName, "OV_CPU_BLOB_DUMP_NODE_NAME");
+        readParam(execGraphPath, "OV_CPU_EXEC_GRAPH_PATH");
+    }
+
+    std::string blobDumpDir;
+    std::string blobDumpFormat;
+    std::string blobDumpNodeExecId;
+    std::string blobDumpNodeType;
+    std::string blobDumpNodeName;
+    std::string execGraphPath;
+
+private:
+    void readParam(std::string& param, const char* envVar) {
+        if (const char* envValue = std::getenv(envVar))
+            param = envValue;
+    }
+};
+
+} // namespace DebugCaps
+} // namespace MKLDNNPlugin
+
+#else // !CPU_DEBUG_CAPS
+#define ENABLE_CPU_DEBUG_CAP(_x)
+#endif // CPU_DEBUG_CAPS
diff --git a/inference-engine/src/mkldnn_plugin/utils/node_dumper.cpp b/inference-engine/src/mkldnn_plugin/utils/node_dumper.cpp
index f69551159dc1ea..9f3af44a66a26a 100644
--- a/inference-engine/src/mkldnn_plugin/utils/node_dumper.cpp
+++ b/inference-engine/src/mkldnn_plugin/utils/node_dumper.cpp
@@ -6,9 +6,10 @@
 #include "node_dumper.h"
 
 #include "mkldnn_node.h"
+#include "ie_common.h"
 #include "utils/blob_dump.h"
+#include "utils/debug_capabilities.h"
 
-#include "ie_common.h"
 #include <array>
 #include <regex>
 #include <sstream>
@@ -18,27 +19,24 @@ using namespace InferenceEngine;
 
 namespace MKLDNNPlugin {
 
-NodeDumper::NodeDumper(int _count):
-    count(_count), dumpFormat(DUMP_FORMAT::BIN) {
-    const char* dumpDirEnv = std::getenv("OV_CPU_BLOB_DUMP_DIR");
-    if (dumpDirEnv)
-        dumpDirName = dumpDirEnv;
+NodeDumper::NodeDumper(const DebugCaps::Config& config, const int _count)
+    : dumpFormat(DUMP_FORMAT::BIN)
+    , dumpDirName("mkldnn_dump")
+    , count(_count) {
+    if (!config.blobDumpDir.empty())
+        dumpDirName = config.blobDumpDir;
 
-    const char* dumpFormatEnv = std::getenv("OV_CPU_BLOB_DUMP_FORMAT");
-    if (dumpFormatEnv)
-        dumpFormat = parseDumpFormat(dumpFormatEnv);
+    if (!config.blobDumpFormat.empty())
+        dumpFormat = parseDumpFormat(config.blobDumpFormat);
 
-    const char* filter = std::getenv("OV_CPU_BLOB_DUMP_NODE_EXEC_ID");
-    if (filter)
-        dumpFilters[FILTER::BY_EXEC_ID] = filter;
+    if (!config.blobDumpNodeExecId.empty())
+        dumpFilters[FILTER::BY_EXEC_ID] = config.blobDumpNodeExecId;
 
-    filter = std::getenv("OV_CPU_BLOB_DUMP_NODE_TYPE");
-    if (filter)
-        dumpFilters[FILTER::BY_TYPE] = filter;
+    if (!config.blobDumpNodeType.empty())
+        dumpFilters[FILTER::BY_TYPE] = config.blobDumpNodeType;
 
-    filter = std::getenv("OV_CPU_BLOB_DUMP_NODE_NAME");
-    if (filter)
-        dumpFilters[FILTER::BY_NAME] = filter;
+    if (!config.blobDumpNodeName.empty())
+        dumpFilters[FILTER::BY_NAME] = config.blobDumpNodeName;
 }
 
 void NodeDumper::dumpInputBlobs(const MKLDNNNodePtr& node) const {
diff --git a/inference-engine/src/mkldnn_plugin/utils/node_dumper.h b/inference-engine/src/mkldnn_plugin/utils/node_dumper.h
index 7dd1ac1f0c651e..0580bee4731b91 100644
--- a/inference-engine/src/mkldnn_plugin/utils/node_dumper.h
+++ b/inference-engine/src/mkldnn_plugin/utils/node_dumper.h
@@ -6,6 +6,7 @@
 
 #include "mkldnn_node.h"
 #include "utils/blob_dump.h"
+#include "utils/debug_capabilities.h"
 
 #include <unordered_map>
 #include <string>
@@ -22,7 +23,7 @@ namespace MKLDNNPlugin {
  */
 class NodeDumper {
 public:
-    NodeDumper(int _count);
+    NodeDumper(const DebugCaps::Config& config, const int _count);
 
     void dumpInputBlobs(const MKLDNNNodePtr &node) const;
     void dumpOutputBlobs(const MKLDNNNodePtr &node) const;
@@ -41,11 +42,9 @@ class NodeDumper {
     void formatNodeName(std::string& name) const;
 
     DUMP_FORMAT dumpFormat;
-
+    std::string dumpDirName;
     int count;
 
-    std::string dumpDirName = "mkldnn_dump";
-
     enum FILTER {
         BY_EXEC_ID,
         BY_TYPE,

From f0cf1dffa76dd7c93c2b7e8d8b4b5cf2c39c0f6a Mon Sep 17 00:00:00 2001
From: Andrey Somsikov <andrey.somsikov@intel.com>
Date: Wed, 9 Jun 2021 08:21:03 +0300
Subject: [PATCH 037/178] Add ENABLE_STRICT_DEPENDENCIES for faster parallel
 builds (#5473)

---
 cmake/features.cmake       |  2 ++
 ngraph/test/CMakeLists.txt | 12 ++++++++++--
 2 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/cmake/features.cmake b/cmake/features.cmake
index adb1fad25238da..7518c99c86838c 100644
--- a/cmake/features.cmake
+++ b/cmake/features.cmake
@@ -6,6 +6,8 @@ ie_dependent_option (ENABLE_MKL_DNN "MKL-DNN plugin for inference engine" ON "X8
 
 ie_option (ENABLE_TESTS "unit, behavior and functional tests" OFF)
 
+ie_option (ENABLE_STRICT_DEPENDENCIES "Skip configuring \"convinient\" dependencies for efficient parallel builds" OFF)
+
 ie_dependent_option (ENABLE_CLDNN "clDnn based plugin for inference engine" ON "X86_64;NOT APPLE;NOT MINGW;NOT WINDOWS_STORE;NOT WINDOWS_PHONE" OFF)
 
 ie_option (ENABLE_PROFILING_ITT "Build with ITT tracing. Optionally configure pre-built ittnotify library though INTEL_VTUNE_DIR variable." OFF)
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index ba3c6a9d052b80..97f85a3128f926 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -286,13 +286,21 @@ set_source_files_properties(includes.cpp PROPERTIES COMPILE_DEFINITIONS
 if (ENABLE_MKL_DNN)
     message(STATUS "NGRAPH_TESTS: IE:CPU enabled")
     set(ACTIVE_BACKEND_LIST ${ACTIVE_BACKEND_LIST} "IE:CPU")
-    list(APPEND UNIT_TESTS_DEPENDENCIES MKLDNNPlugin)
+    if (NOT ENABLE_STRICT_DEPENDENCIES)
+        # For convinience add a runtime dependency to build along with this target.
+        # Warning: Parallel build with -GNinja may not be efficient.
+        list(APPEND UNIT_TESTS_DEPENDENCIES MKLDNNPlugin)
+    endif()
 endif()
 
 if (ENABLE_CLDNN)
     message(STATUS "NGRAPH_TESTS: IE:GPU enabled")
     set(ACTIVE_BACKEND_LIST ${ACTIVE_BACKEND_LIST} "IE:GPU")
-    list(APPEND UNIT_TESTS_DEPENDENCIES clDNNPlugin)
+    if (NOT ENABLE_STRICT_DEPENDENCIES)
+        # For convinience add a runtime dependency to build along with this target.
+        # Warning: Parallel build with -GNinja may not be efficient.
+        list(APPEND UNIT_TESTS_DEPENDENCIES clDNNPlugin)
+    endif()
 endif()
 
 if (NGRAPH_INTERPRETER_ENABLE)

From aa4a18dda1ad0dfc55412dbe776be99dc0c0c964 Mon Sep 17 00:00:00 2001
From: Vladimir Paramuzov <vladimir.paramuzov@intel.com>
Date: Wed, 9 Jun 2021 09:02:25 +0300
Subject: [PATCH 038/178] [IE CLDNN] Updated GPU device config (#6040)

---
 docs/IE_DG/API_Changes.md                     |  15 ++-
 docs/IE_DG/Extensibility_DG/GPU_Kernel.md     |  16 ---
 docs/IE_DG/GPU_Kernels_Tuning.md              |  39 ------
 docs/IE_DG/Intro_to_Performance.md            |  20 +--
 .../supported_plugins/{CL_DNN.md => GPU.md}   |  45 +++----
 .../supported_plugins/Supported_Devices.md    |  10 +-
 docs/doxygen/ie_docs.xml                      |   3 +-
 docs/model_server/README.md                   |  46 +++----
 .../dldt_optimization_guide.md                |  62 ++++-----
 docs/snippets/GPU_Kernel.cpp                  |   5 -
 docs/snippets/GPU_Kernels_Tuning.cpp          |  14 --
 docs/snippets/GPU_RemoteBlob_API2.cpp         |   4 +-
 .../include/cldnn/cldnn_config.hpp            |  80 +++---------
 inference-engine/include/gpu/gpu_config.hpp   | 120 ++++++++++++++++++
 .../samples/benchmark_app/main.cpp            |   4 +-
 .../samples/hello_query_device/README.md      |   6 +-
 .../src/cldnn_engine/cldnn_config.cpp         |  29 +++--
 .../src/cldnn_engine/cldnn_engine.cpp         |   2 +-
 .../cldnn_engine/cldnn_executable_network.cpp |   1 -
 .../src/cldnn_engine/cldnn_graph.cpp          |   1 -
 .../cldnn_remote_blob_tests.cpp               |   4 +-
 .../behavior/config.cpp                       |  35 ++++-
 .../behavior/core_integration.cpp             |   2 +-
 .../behavior/infer_request_input.cpp          |   2 +-
 .../behavior/infer_request_output.cpp         |   2 +-
 .../behavior/test_plugin.cpp                  |   2 +-
 .../multi/gpu_remote_blob_tests.cpp           |   2 +-
 .../single_layer_tests/tensor_iterator.cpp    |   6 +-
 tools/benchmark/main.py                       |   2 +-
 29 files changed, 301 insertions(+), 278 deletions(-)
 delete mode 100644 docs/IE_DG/GPU_Kernels_Tuning.md
 rename docs/IE_DG/supported_plugins/{CL_DNN.md => GPU.md} (62%)
 delete mode 100644 docs/snippets/GPU_Kernels_Tuning.cpp
 create mode 100644 inference-engine/include/gpu/gpu_config.hpp

diff --git a/docs/IE_DG/API_Changes.md b/docs/IE_DG/API_Changes.md
index a234471c13e550..2534a4a6c3856a 100644
--- a/docs/IE_DG/API_Changes.md
+++ b/docs/IE_DG/API_Changes.md
@@ -14,6 +14,15 @@ The sections below contain detailed list of changes made to the Inference Engine
  * InferenceEngine::Parameter(std::shared_ptr<ngraph::Variant>& var)
  * std::shared_ptr<ngraph::Variant> InferenceEngine::Parameter::asVariant() const
  * InferenceEngine::Parameter::operator std::shared_ptr<ngraph::Variant>() const
+ * KEY_CLDNN_NV12_TWO_INPUTS GPU plugin option. Use KEY_GPU_NV12_TWO_INPUTS instead
+ * KEY_CLDNN_PLUGIN_PRIORITY GPU plugin option. Use KEY_GPU_PLUGIN_PRIORITY instead
+ * KEY_CLDNN_PLUGIN_THROTTLE GPU plugin option. Use KEY_GPU_PLUGIN_THROTTLE instead
+ * KEY_CLDNN_MEM_POOL GPU plugin option
+ * KEY_CLDNN_GRAPH_DUMPS_DIR GPU plugin option
+ * KEY_CLDNN_SOURCES_DUMPS_DIR GPU plugin option
+ * KEY_DUMP_KERNELS GPU plugin option
+ * KEY_TUNING_MODE GPU plugin option
+ * KEY_TUNING_FILE GPU plugin option
 
 ## 2021.3
 
@@ -528,7 +537,7 @@ The sections below contain detailed list of changes made to the Inference Engine
  * DLIA_CONFIG_KEY(ENABLE_STREAMING) config key
 
 ### Removed API
- 
+
  * InferenceEngine::EltwiseLayer::Select from InferenceEngine::EltwiseLayer::eOperation enumeration
 
 ## 2019 R2
@@ -577,7 +586,7 @@ The sections below contain detailed list of changes made to the Inference Engine
  * DLIA_CONFIG_KEY(IO_TRANSFORMATIONS_NATIVE) config key
  * DLIA_CONFIG_KEY(DUMP_SUPPORTED_LAYERS_INFORMATION) config key
  * GNA_CONFIG_VALUE(SW_FP32) config value for GNA_CONFIG_KEY(DEVICE_MODE) key
- * MULTI_CONFIG_KEY(DEVICE_PRIORITIES) config key for `MULTI` device 
+ * MULTI_CONFIG_KEY(DEVICE_PRIORITIES) config key for `MULTI` device
  * InferenceEngine::CNNNetReader::ReadNetwork(const std::wstring &filepath) new method
  * InferenceEngine::CNNNetReader::ReadWeights(const std::wstring &filepath) new method
  * InferenceEngine::ExecutableNetwork::ExecutableNetwork(IExecutableNetwork::Ptr actual, InferenceEnginePluginPtr plg) constructor with additional `plg` parameter
@@ -593,7 +602,7 @@ The sections below contain detailed list of changes made to the Inference Engine
  * InferenceEngine::EltwiseLayer::Logical_NOT, InferenceEngine::EltwiseLayer::Mean, InferenceEngine::EltwiseLayer::Select extensions to InferenceEngine::EltwiseLayer::eOperation enumeration
  * InferenceEngine::OneHotLayer new class
  * InferenceEngine::SelectLayer new class
- * InferenceEngine::BroadcastLayer new class 
+ * InferenceEngine::BroadcastLayer new class
  * InferenceEngine::MathLayer new class
  * InferenceEngine::ReduceLayer new class
  * InferenceEngine::TopKLayer new class
diff --git a/docs/IE_DG/Extensibility_DG/GPU_Kernel.md b/docs/IE_DG/Extensibility_DG/GPU_Kernel.md
index 09ace6f0a2942f..d9fd809f8e4227 100644
--- a/docs/IE_DG/Extensibility_DG/GPU_Kernel.md
+++ b/docs/IE_DG/Extensibility_DG/GPU_Kernel.md
@@ -219,22 +219,6 @@ __kernel void example_relu_kernel(
 
 ## Debugging Tips<a name="debugging-tips"></a>
 
-* **Dumping the Resulting Kernels**.
-It is recommended to get a dump of the kernel with all of
-the values set by the Inference Engine, such as tensor sizes,
-floating-point, and integer kernel parameters. To get the dump, add the
-following line to your code that configures the GPU plugin to output the
-custom kernels:
-
-@snippet snippets/GPU_Kernel.cpp part1
-
-When the Inference Engine compiles the kernels for the specific network,
-it also outputs the resulting code for the custom kernels. In the
-directory of your executable, find files like
-`clDNN_program0.cl`, `clDNN_program1.cl`. There are as many files as
-distinct sets of parameters for your custom kernel: different input
-tensor sizes and kernel parameters.
-
 * **Using `printf` in the OpenCL™ Kernels**.
 To debug the specific values, you can use `printf` in your kernels.
 However, be careful: for instance, do not output excessively
diff --git a/docs/IE_DG/GPU_Kernels_Tuning.md b/docs/IE_DG/GPU_Kernels_Tuning.md
deleted file mode 100644
index 5bb6a8334b2372..00000000000000
--- a/docs/IE_DG/GPU_Kernels_Tuning.md
+++ /dev/null
@@ -1,39 +0,0 @@
-Using GPU Kernels Tuning {#openvino_docs_IE_DG_GPU_Kernels_Tuning}
-======================
-
-GPU Kernels Tuning allows you to tune models, so the heavy computational layers are configured to fit better into
-hardware, which the tuning was done on. It is required to achieve best performance on GPU.
-> **NOTE** Currently only convolution and fully connected layers undergo tuning process. It means that the performance boost depends on the amount of that layers in the model.
-
-OpenVINO™ releases include the `<INSTALL_DIR>/inference_engine/bin/intel64/Release/cache.json` file with pretuned data for current state of the art models. It is highly recommended to do the
-tuning for new kind of models, hardwares or drivers.
-
-## Tuned data
-
-GPU tuning data is saved in JSON format. The file is composed of 2 types of attributes and 1 type of value:
-* Execution units number (attribute): splits the content into different EU sections
-* Hash (attribute): hashed tuned kernel data
-* Key (value): Array with kernel name and kernel's mode index
-
-## Usage
-
----
-
-You can activate Kernels Tuning process by setting `KEY_TUNING_MODE` flag to `TUNING_CREATE` and `KEY_TUNING_FILE` to `<"filename">` in a configuration map that is
-passed to the plugin while loading a network.
-This configuration modifies the behavior of the `ExecutableNetwork` object. Instead of standard network compilation, it will run the tuning process.
-Please keep in mind that the tuning can be very time consuming. The bigger the network, the longer it will take.
-File with tuned data is the result of this step.
-
-> **NOTE** If a filename passed to `KEY_TUNING_FILE` points to existing tuned data and you are tuning a new model, then this file will be extended by new data. This allows you to extend existing `cache.json` provided in the OpenVINO™ release package. 
-
-The example below shows how to set and use the key files:
-
-@snippet snippets/GPU_Kernels_Tuning.cpp part0
-
----
-
-You can activate the inference with tuned data by setting `KEY_TUNING_MODE` flag to `TUNING_USE_EXISTING` and
-`KEY_TUNING_FILE` flag to `<"filename">`. 
-
-GPU backend will process the content of the file during network compilation to configure the OpenCL kernels for the best performance.
diff --git a/docs/IE_DG/Intro_to_Performance.md b/docs/IE_DG/Intro_to_Performance.md
index 78d5c59c417d0f..94d0173dbbe2b4 100644
--- a/docs/IE_DG/Intro_to_Performance.md
+++ b/docs/IE_DG/Intro_to_Performance.md
@@ -22,7 +22,7 @@ $ benchmark_app -m <model.xml> -enforcebf16=false
 Notice that for quantized (e.g. INT8) models the bfloat16 calculations (of the layers that remain in FP32) is disabled by default.
 Refer to the [CPU Plugin documentation](supported_plugins/CPU.md) for more details.
 
-Similarly, the GPU device has a dedicated config key to enable FP16 execution of the layers that remain in FP32 in the quantized models (as the quantization is typically performed on the FP32 models), refer to the ENABLE_FP16_FOR_QUANTIZED_MODELS key in the [GPU Plugin documentation](supported_plugins/CL_DNN.md) 
+Similarly, the GPU device has a dedicated config key to enable FP16 execution of the layers that remain in FP32 in the quantized models (as the quantization is typically performed on the FP32 models), refer to the ENABLE_FP16_FOR_QUANTIZED_MODELS key in the [GPU Plugin documentation](supported_plugins/GPU.md)
 
 ## Latency vs. Throughput
 One way to increase computational efficiency is batching, which combines many (potentially tens) of
@@ -72,30 +72,20 @@ Instead, it is possible to keep a separate infer request per camera or another s
 
 ## Benchmark App
 [Benchmark App](../../inference-engine/samples/benchmark_app/README.md) sample is the best performance reference.
-It has a lot of device-specific knobs, but the primary usage is as simple as: 
+It has a lot of device-specific knobs, but the primary usage is as simple as:
 ```bash
 $ ./benchmark_app –d GPU –m <model> -i <input>
 ```
-to measure the performance of the model on the GPU. 
+to measure the performance of the model on the GPU.
 Or
 ```bash
 $ ./benchmark_app –d CPU –m <model> -i <input>
 ```
 to execute on the CPU instead.
 
-For example, for the CPU throughput mode from the previous section, you can play with number of streams (`-nstreams` command-line param). 
-Try different values of the `-nstreams` argument from `1` to a number of CPU cores and find one that provides the best performance. For example, on a 8-core CPU, compare the `-nstreams 1` (which is a latency-oriented scenario) to the `2`, `4` and `8` streams. Notice that `benchmark_app` automatically queries/creates/runs number of requests required to saturate the given number of streams. 
+For example, for the CPU throughput mode from the previous section, you can play with number of streams (`-nstreams` command-line param).
+Try different values of the `-nstreams` argument from `1` to a number of CPU cores and find one that provides the best performance. For example, on a 8-core CPU, compare the `-nstreams 1` (which is a latency-oriented scenario) to the `2`, `4` and `8` streams. Notice that `benchmark_app` automatically queries/creates/runs number of requests required to saturate the given number of streams.
 
 Finally, notice that when you don't specify number of streams with `-nstreams`, "AUTO" value for the streams is used, e.g. for the CPU this is [CPU_THROUGHPUT_AUTO](supported_plugins/CPU.md). You can spot the actual value behind "AUTO" for your machine in the application output.
 Notice that the "AUTO" number is not necessarily most optimal, so it is generally recommended to play either with the benchmark_app's "-nstreams" as described above, or via  [new Workbench tool](@ref workbench_docs_Workbench_DG_Introduction).This allows you to simplify the app-logic, as you don't need to combine multiple inputs into a batch to achieve good CPU performance.
 Instead, it is possible to keep a separate infer request per camera or another source of input and process the requests in parallel using Async API.
-
-## Kernels Tuning for GPU
-
-GPU backend comes with a feature, that allows models tuning, so the workload is configured to fit better into hardware.
-
-Tuning is time consuming process, which internally execute every layer several (or even hundreds) times to find most performant configuration.
-
-This configuration is saved into json-formatted file, whose name can be passed as plugin param to network. GPU backend will process this data to configure kernels for the best performance.
-
-For more details about Kernels Tuning and How-To please refer to [GPU Kernels Tuning](GPU_Kernels_Tuning.md). 
diff --git a/docs/IE_DG/supported_plugins/CL_DNN.md b/docs/IE_DG/supported_plugins/GPU.md
similarity index 62%
rename from docs/IE_DG/supported_plugins/CL_DNN.md
rename to docs/IE_DG/supported_plugins/GPU.md
index 0216ae71d0dd36..cc12be98a121e1 100644
--- a/docs/IE_DG/supported_plugins/CL_DNN.md
+++ b/docs/IE_DG/supported_plugins/GPU.md
@@ -1,4 +1,4 @@
-GPU Plugin {#openvino_docs_IE_DG_supported_plugins_CL_DNN}
+GPU Plugin {#openvino_docs_IE_DG_supported_plugins_GPU}
 =======
 
 The GPU plugin uses the Intel® Compute Library for Deep Neural Networks (clDNN) to infer deep neural networks.
@@ -89,13 +89,10 @@ Some layers are executed during the load time, not during the inference. One of
 
 The following layers are not accelerated on the GPU and executed on the host CPU instead:
 * Proposal
-* SimplerNMS
+* NonMaxSuppression
 * PriorBox
 * DetectionOutput
 
-## Known Layers Limitations
-* ROIPooling is supported for 'max' value of 'method' attribute.
-
 ## Supported Configuration Parameters
 
 The plugin supports the configuration parameters listed below.
@@ -107,31 +104,21 @@ When specifying key values as raw strings (that is, when using Python API), omit
 | `KEY_CACHE_DIR`      | `"<cache_dir>"`                    | `""`              | Specifies a directory where compiled OCL binaries can be cached. First model loading generates the cache, and all subsequent LoadNetwork calls use precompiled kernels which significantly improves load time. If empty - caching is disabled             |
 | `KEY_PERF_COUNT`      | `YES` / `NO`                    | `NO`              | Collect performance counters during inference             |
 | `KEY_CONFIG_FILE`     | `"<file1> [<file2> ...]"`         | `""`              | Load custom layer configuration files                     |
-| `KEY_DUMP_KERNELS`    | `YES` / `NO`                    | `NO`              | Dump the final kernels used for custom layers             |
-| `KEY_TUNING_MODE`     | `TUNING_DISABLED` <br /> `TUNING_CREATE` <br />  `TUNING_USE_EXISTING`            | `TUNING_DISABLED` | Disable inference kernel tuning     <br /> Create tuning file (expect much longer runtime)  <br />         Use an existing tuning file              |
-| `KEY_TUNING_FILE`     | `"<filename>"`                  | `""`              | Tuning file to create / use                               |
-| `KEY_CLDNN_PLUGIN_PRIORITY` | `<0-3>`                       | `0`               | OpenCL queue priority (before usage, make sure your OpenCL driver supports appropriate extension)<br> Higher value means higher priority for clDNN OpenCL queue. 0 disables the setting. |
-| `KEY_CLDNN_PLUGIN_THROTTLE` | `<0-3>`                       | `0`               | OpenCL queue throttling (before usage, make sure your OpenCL driver supports appropriate extension)<br> Lower value means lower driver thread priority and longer sleep time for it. 0 disables the setting. |
-| `KEY_CLDNN_GRAPH_DUMPS_DIR` | `"<dump_dir>"`                       | `""`               | clDNN graph optimizer stages dump output directory (in GraphViz format)                                     |
-| `KEY_CLDNN_SOURCES_DUMPS_DIR` | `"<dump_dir>"`                       | `""`               | Final optimized clDNN OpenCL sources dump output directory                                   |
-| `KEY_GPU_THROUGHPUT_STREAMS`  | `KEY_GPU_THROUGHPUT_AUTO`, or positive integer| 1 | Specifies a number of GPU "execution" streams for the throughput mode (upper bound for a number of inference requests that can be executed simultaneously).<br>This option is can be used to decrease GPU stall time by providing more effective load from several streams. Increasing the number of streams usually is more effective for smaller topologies or smaller input sizes. Note that your application should provide enough parallel slack (e.g. running many inference requests) to leverage full GPU bandwidth. Additional streams consume several times more GPU memory, so make sure the system has enough memory available to suit parallel stream execution. Multiple streams might also put additional load on CPU. If CPU load increases, it can be regulated by setting an appropriate `KEY_CLDNN_PLUGIN_THROTTLE` option value (see above). If your target system has relatively weak CPU, keep throttling low. <br>The default value is 1, which implies latency-oriented behavior.<br>`KEY_GPU_THROUGHPUT_AUTO` creates bare minimum of streams to improve the performance; this is the most portable option if you are not sure how many resources your target machine has (and what would be the optimal number of streams). <br> A positive integer value creates the requested number of streams. |
+| `KEY_GPU_PLUGIN_PRIORITY` | `<0-3>`                       | `0`               | OpenCL queue priority (before usage, make sure your OpenCL driver supports appropriate extension)<br> Higher value means higher priority for OpenCL queue. 0 disables the setting. |
+| `KEY_GPU_PLUGIN_THROTTLE` | `<0-3>`                       | `0`               | OpenCL queue throttling (before usage, make sure your OpenCL driver supports appropriate extension)<br> Lower value means lower driver thread priority and longer sleep time for it. 0 disables the setting. |
+| `KEY_CLDNN_ENABLE_FP16_FOR_QUANTIZED_MODELS` | `YES` / `NO`                       | `YES`               | Allows using FP16+INT8 mixed precision mode, so non-quantized parts of a model will be executed in FP16 precision for FP16 IR. Does not affect quantized FP32 IRs |
+| `KEY_GPU_NV12_TWO_INPUTS` | `YES` / `NO`                       | `NO`               | Controls preprocessing logic for nv12 input. If it's set to YES, then device graph will expect that user will set biplanar nv12 blob as input wich will be directly passed to device execution graph. Otherwise, preprocessing via GAPI is used to convert NV12->BGR, thus GPU graph have to expect single input |
+| `KEY_GPU_THROUGHPUT_STREAMS`  | `KEY_GPU_THROUGHPUT_AUTO`, or positive integer| 1 | Specifies a number of GPU "execution" streams for the throughput mode (upper bound for a number of inference requests that can be executed simultaneously).<br>This option is can be used to decrease GPU stall time by providing more effective load from several streams. Increasing the number of streams usually is more effective for smaller topologies or smaller input sizes. Note that your application should provide enough parallel slack (e.g. running many inference requests) to leverage full GPU bandwidth. Additional streams consume several times more GPU memory, so make sure the system has enough memory available to suit parallel stream execution. Multiple streams might also put additional load on CPU. If CPU load increases, it can be regulated by setting an appropriate `KEY_GPU_PLUGIN_THROTTLE` option value (see above). If your target system has relatively weak CPU, keep throttling low. <br>The default value is 1, which implies latency-oriented behavior.<br>`KEY_GPU_THROUGHPUT_AUTO` creates bare minimum of streams to improve the performance; this is the most portable option if you are not sure how many resources your target machine has (and what would be the optimal number of streams). <br> A positive integer value creates the requested number of streams. |
 | `KEY_EXCLUSIVE_ASYNC_REQUESTS` | `YES` / `NO`                | `NO`              | Forces async requests (also from different executable networks) to execute serially.|
-| `KEY_CLDNN_MAX_NUM_THREADS` | `integer value` | `maximum # of HW threads available in host environment` |  Specifies the number of CPU threads that can be used for clDNN engine, e.g, JIT compilation of clDNN kernels or clDNN cpu kernel processing. The default value is set as the number of maximum available threads in host environment to minimize the time for LoadNetwork, where the clDNN kernel build time occupies a large portion. Note that if the specified value is larger than the maximum available # of threads or less than zero, it is set as maximum available # of threads. It can be specified with a smaller number than the available HW threads according to the usage scenario, e.g., when the user wants to assign more CPU threads while clDNN plugin is running. Note that setting this value with lower number will affect not only the network loading time but also the cpu layers of clDNN networks that are optimized with multi-threading. |
-| `KEY_CLDNN_ENABLE_LOOP_UNROLLING` | `YES` / `NO`             | `YES`             | Enables recurrent layers such as TensorIterator or Loop with fixed iteration count to be unrolled. It is turned on by default. Turning this key on will achieve better inference performance for loops with not too many iteration counts (less than 16, as a rule of thumb). Turning this key off will achieve better performance for both graph loading time and inference time with many iteration counts (greater than 16). Note that turning this key on will increase the graph loading time in proportion to the iteration counts. Thus, this key should be turned off if graph loading time is considered to be most important target to optimize. |
-
-## Note on Debug Capabilities of the GPU Plugin
-
-Inference Engine GPU plugin provides possibility to dump the user custom OpenCL&trade; kernels to a file to allow you to properly debug compilation issues in your custom kernels.
-
-The application can use the <code>SetConfig()</code> function with the key <code>PluginConfigParams::KEY_DUMP_KERNELS</code> and value: <code>PluginConfigParams::YES</code>. Then during network loading, all custom layers will print their OpenCL kernels with the JIT instrumentation added by the plugin.
-The kernels will be stored in the working directory under files named the following way: <code>clDNN_program0.cl</code>, <code>clDNN_program1.cl</code>.
-
-This option is disabled by default. Additionally, the application can call the <code>SetConfig()</code> function with the key <code>PluginConfigParams::KEY_DUMP_KERNELS</code> and value: <code>PluginConfigParams::NO</code> before network loading.
-
-How to verify that this option is disabled:
-1.  Delete all <code>clDNN_program*.cl</code> files from the current directory
-2.  Run your application to load a network
-3.  Examine the working directory for the presence of any kernel file (for example, <code>clDNN_program0.cl</code>)
+| `KEY_GPU_MAX_NUM_THREADS` | `integer value` | `maximum # of HW threads available in host environment` |  Specifies the number of CPU threads that can be used for GPU engine, e.g, JIT compilation of GPU kernels or cpu kernel processing within GPU plugin. The default value is set as the number of maximum available threads in host environment to minimize the time for LoadNetwork, where the GPU kernel build time occupies a large portion. Note that if the specified value is larger than the maximum available # of threads or less than zero, it is set as maximum available # of threads. It can be specified with a smaller number than the available HW threads according to the usage scenario, e.g., when the user wants to assign more CPU threads while GPU plugin is running. Note that setting this value with lower number will affect not only the network loading time but also the cpu layers of GPU networks that are optimized with multi-threading. |
+| `KEY_GPU_ENABLE_LOOP_UNROLLING` | `YES` / `NO`             | `YES`             | Enables recurrent layers such as TensorIterator or Loop with fixed iteration count to be unrolled. It is turned on by default. Turning this key on will achieve better inference performance for loops with not too many iteration counts (less than 16, as a rule of thumb). Turning this key off will achieve better performance for both graph loading time and inference time with many iteration counts (greater than 16). Note that turning this key on will increase the graph loading time in proportion to the iteration counts. Thus, this key should be turned off if graph loading time is considered to be most important target to optimize. |
+| `KEY_CLDNN_PLUGIN_PRIORITY` | `<0-3>`                       | `0`               | OpenCL queue priority (before usage, make sure your OpenCL driver supports appropriate extension)<br> Higher value means higher priority for OpenCL queue. 0 disables the setting. **Deprecated**. Please use KEY_GPU_PLUGIN_PRIORITY |
+| `KEY_CLDNN_PLUGIN_THROTTLE` | `<0-3>`                       | `0`               | OpenCL queue throttling (before usage, make sure your OpenCL driver supports appropriate extension)<br> Lower value means lower driver thread priority and longer sleep time for it. 0 disables the setting. **Deprecated**. Please use KEY_GPU_PLUGIN_THROTTLE |
+| `KEY_CLDNN_GRAPH_DUMPS_DIR` | `"<dump_dir>"`                       | `""`               | clDNN graph optimizer stages dump output directory (in GraphViz format) **Deprecated**. Will be removed in the next release                                     |
+| `KEY_CLDNN_SOURCES_DUMPS_DIR` | `"<dump_dir>"`                       | `""`               | Final optimized clDNN OpenCL sources dump output directory. **Deprecated**. Will be removed in the next release                                   |
+| `KEY_DUMP_KERNELS`    | `YES` / `NO`                    | `NO`              | Dump the final kernels used for custom layers. **Deprecated**. Will be removed in the next release             |
+| `KEY_TUNING_MODE`     | `TUNING_DISABLED` <br /> `TUNING_CREATE` <br />  `TUNING_USE_EXISTING`            | `TUNING_DISABLED` | Disable inference kernel tuning     <br /> Create tuning file (expect much longer runtime)  <br />         Use an existing tuning file. **Deprecated**. Will be removed in the next release |
+| `KEY_TUNING_FILE`     | `"<filename>"`                  | `""`              | Tuning file to create / use. **Deprecated**. Will be removed in the next release |
 
 ## GPU Context and Video Memory Sharing RemoteBlob API
 
diff --git a/docs/IE_DG/supported_plugins/Supported_Devices.md b/docs/IE_DG/supported_plugins/Supported_Devices.md
index ed8cabec076f03..e1140ae4b74cae 100644
--- a/docs/IE_DG/supported_plugins/Supported_Devices.md
+++ b/docs/IE_DG/supported_plugins/Supported_Devices.md
@@ -9,11 +9,11 @@ The Inference Engine provides unique capabilities to infer deep learning models
 
 | Plugin                                   | Device types                                                                                                                                                |
 |------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------|
-|[GPU plugin](CL_DNN.md)            |Intel&reg; Processor Graphics, including Intel&reg; HD Graphics and Intel&reg; Iris&reg; Graphics                                                            |
+|[GPU plugin](GPU.md)            |Intel&reg; Processor Graphics, including Intel&reg; HD Graphics and Intel&reg; Iris&reg; Graphics                                                            |
 |[CPU plugin](CPU.md)              |Intel&reg; Xeon&reg; with Intel® Advanced Vector Extensions 2 (Intel® AVX2), Intel® Advanced Vector Extensions 512 (Intel® AVX-512), and AVX512_BF16, Intel&reg; Core&trade; Processors with Intel&reg; AVX2, Intel&reg; Atom&reg; Processors with Intel® Streaming SIMD Extensions (Intel® SSE) |
 |[VPU plugins](VPU.md) (available in the Intel® Distribution of OpenVINO™ toolkit)            |Intel® Neural Compute Stick 2 powered by the Intel® Movidius™ Myriad™ X, Intel® Vision Accelerator Design with Intel® Movidius™ VPUs                                                                                           |
 |[GNA plugin](GNA.md) (available in the Intel® Distribution of OpenVINO™ toolkit)              |Intel&reg; Speech Enabling Developer Kit, Amazon Alexa* Premium Far-Field Developer Kit, Intel&reg; Pentium&reg; Silver J5005 Processor, Intel&reg; Pentium&reg; Silver N5000 Processor, Intel&reg; Celeron&reg; J4005 Processor, Intel&reg; Celeron&reg; J4105 Processor, Intel&reg; Celeron&reg; Processor N4100, Intel&reg; Celeron&reg; Processor N4000, Intel&reg; Core&trade; i3-8121U Processor, Intel&reg; Core&trade; i7-1065G7 Processor, Intel&reg; Core&trade; i7-1060G7 Processor, Intel&reg; Core&trade; i5-1035G4 Processor, Intel&reg; Core&trade; i5-1035G7 Processor, Intel&reg; Core&trade; i5-1035G1 Processor, Intel&reg; Core&trade; i5-1030G7 Processor, Intel&reg; Core&trade; i5-1030G4 Processor, Intel&reg; Core&trade; i3-1005G1 Processor, Intel&reg; Core&trade; i3-1000G1 Processor, Intel&reg; Core&trade; i3-1000G4 Processor|
-|[Multi-Device plugin](MULTI.md) |Multi-Device plugin enables simultaneous inference of the same network on several Intel&reg; devices in parallel    |   
+|[Multi-Device plugin](MULTI.md) |Multi-Device plugin enables simultaneous inference of the same network on several Intel&reg; devices in parallel    |
 |[Heterogeneous plugin](HETERO.md) |Heterogeneous plugin enables automatic inference splitting between several Intel&reg; devices (for example if a device doesn't [support certain layers](#supported-layers)).                                                           |
 
 Devices similar to the ones we have used for benchmarking can be accessed using [Intel® DevCloud for the Edge](https://devcloud.intel.com/edge/), a remote development environment with access to Intel® hardware and the latest versions of the Intel® Distribution of the OpenVINO™ Toolkit. [Learn more](https://devcloud.intel.com/edge/get_started/devcloud/) or [Register here](https://inteliot.force.com/DevcloudForEdge/s/).
@@ -60,7 +60,7 @@ For example, the CHW value at index (c,h,w) is physically located at index (c\*H
 |GNA plugin    |Supported               |Supported               |Not supported           |
 <br>\* - currently, only limited set of topologies might benefit from enabling I8 model on GPU<br>
 For [Multi-Device](MULTI.md) and [Heterogeneous](HETERO.md) execution
-the supported models formats depends on the actual underlying devices. _Generally, FP16 is preferable as it is most ubiquitous and performant_.  
+the supported models formats depends on the actual underlying devices. _Generally, FP16 is preferable as it is most ubiquitous and performant_.
 
 ### Supported Input Precision
 
@@ -73,7 +73,7 @@ the supported models formats depends on the actual underlying devices. _Generall
 
 <br>\* - Supported via `SetBlob` only, `GetBlob` returns FP32<br>
 For [Multi-Device](MULTI.md) and [Heterogeneous](HETERO.md) execution
-the supported input precision  depends on the actual underlying devices. _Generally, U8 is preferable as it is most ubiquitous_.  
+the supported input precision  depends on the actual underlying devices. _Generally, U8 is preferable as it is most ubiquitous_.
 
 ### Supported Output Precision
 
@@ -84,7 +84,7 @@ the supported input precision  depends on the actual underlying devices. _Genera
 |VPU plugins   |Supported |Supported     |
 |GNA plugin    |Supported |Not supported |
 For [Multi-Device](MULTI.md) and [Heterogeneous](HETERO.md) execution
-the supported output precision  depends on the actual underlying devices. _Generally, FP32 is preferable as it is most ubiquitous_. 
+the supported output precision  depends on the actual underlying devices. _Generally, FP32 is preferable as it is most ubiquitous_.
 
 ### Supported Input Layout
 
diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index f287487913d56a..bb006c9f01c630 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -293,7 +293,6 @@ limitations under the License.
                 <tab type="user" title="[DEPRECATED] Import an ONNX model" url="@ref openvino_docs_IE_DG_OnnxImporterTutorial"/>
                 <tab type="user" title="Using Dynamic Batching Feature" url="@ref openvino_docs_IE_DG_DynamicBatching"/>
                 <tab type="user" title="Using Static Shape Infer Feature" url="@ref openvino_docs_IE_DG_ShapeInference"/>
-                <tab type="user" title="Using GPU kernels tuning" url="@ref openvino_docs_IE_DG_GPU_Kernels_Tuning"/>
                 <tab type="usergroup" title="Using Bfloat16 Inference" url="@ref openvino_docs_IE_DG_Bfloat16Inference">
                 </tab>
                 <tab type="usergroup" title="Using Low-Precision 8-bit Integer Inference" url="@ref openvino_docs_IE_DG_Int8Inference">
@@ -303,7 +302,7 @@ limitations under the License.
                 </tab>
                 <tab type="user" title="Introduction to OpenVINO state API" url="@ref openvino_docs_IE_DG_network_state_intro"/>
                 <tab type="usergroup" title="Supported Devices" url="@ref openvino_docs_IE_DG_supported_plugins_Supported_Devices">
-                    <tab type="usergroup" title="GPU Plugin" url="@ref openvino_docs_IE_DG_supported_plugins_CL_DNN">
+                    <tab type="usergroup" title="GPU Plugin" url="@ref openvino_docs_IE_DG_supported_plugins_GPU">
                         <tab type="user" title="RemoteBlob API of GPU Plugin" url="@ref openvino_docs_IE_DG_supported_plugins_GPU_RemoteBlob_API"/>
                     </tab>
                     <tab type="user" title="CPU Plugin" url="@ref openvino_docs_IE_DG_supported_plugins_CPU"/>
diff --git a/docs/model_server/README.md b/docs/model_server/README.md
index ae5d03914ab347..e6c7144f3cb6c9 100644
--- a/docs/model_server/README.md
+++ b/docs/model_server/README.md
@@ -1,29 +1,29 @@
 # OpenVINO&trade; Model Server {#openvino_docs_ovms}
 
-OpenVINO&trade; Model Server (OVMS) is a scalable, high-performance solution for serving machine learning models optimized for Intel&reg; architectures. 
-The server provides an inference service via gRPC or REST API - making it easy to deploy new algorithms and AI experiments using the same 
-architecture as [TensorFlow* Serving](https://github.com/tensorflow/serving) for any models trained in a framework that is supported 
-by [OpenVINO](https://software.intel.com/en-us/openvino-toolkit). 
+OpenVINO&trade; Model Server (OVMS) is a scalable, high-performance solution for serving machine learning models optimized for Intel&reg; architectures.
+The server provides an inference service via gRPC or REST API - making it easy to deploy new algorithms and AI experiments using the same
+architecture as [TensorFlow* Serving](https://github.com/tensorflow/serving) for any models trained in a framework that is supported
+by [OpenVINO](https://software.intel.com/en-us/openvino-toolkit).
 
 The server implements gRPC and REST API framework with data serialization and deserialization using TensorFlow Serving API,
  and OpenVINO&trade; as the inference execution provider. Model repositories may reside on a locally accessible file system (for example, NFS),
   Google Cloud Storage\* (GCS), Amazon S3\*, MinIO\*, or Azure Blob Storage\*.
-  
+
 OVMS is now implemented in C++ and provides much higher scalability compared to its predecessor in the Python version.
 You can take advantage of all the power of Xeon® CPU capabilities or AI accelerators and expose it over the network interface.
 Read the [release notes](https://github.com/openvinotoolkit/model_server/releases) to find out what's new in the C++ version.
 
 Review the [Architecture Concept](https://github.com/openvinotoolkit/model_server/blob/main/docs/architecture.md) document for more details.
 
-A few key features: 
+A few key features:
 - Support for multiple frameworks. Serve models trained in popular formats such as Caffe\*, TensorFlow\*, MXNet\*, and ONNX*.
 - Deploy new [model versions](https://github.com/openvinotoolkit/model_server/blob/main/docs/docker_container.md#model-version-policy) without changing client code.
-- Support for AI accelerators including [Intel Movidius Myriad VPUs](../IE_DG/supported_plugins/VPU), 
-[GPU](../IE_DG/supported_plugins/CL_DNN), and [HDDL](../IE_DG/supported_plugins/HDDL). 
+- Support for AI accelerators including [Intel Movidius Myriad VPUs](../IE_DG/supported_plugins/VPU.md),
+[GPU](../IE_DG/supported_plugins/GPU.md), and [HDDL](../IE_DG/supported_plugins/HDDL.md).
 - The server can be enabled both on [Bare Metal Hosts](https://github.com/openvinotoolkit/model_server/blob/main/docs/host.md) or in
 [Docker* containers](https://github.com/openvinotoolkit/model_server/blob/main/docs/docker_container.md).
-- [Kubernetes deployments](https://github.com/openvinotoolkit/model_server/blob/main/deploy). The server can be deployed in a Kubernetes cluster allowing the inference service to scale horizontally and ensure high availability.  
-- [Model reshaping](https://github.com/openvinotoolkit/model_server/blob/main/docs/docker_container.md#model-reshaping). The server supports reshaping models in runtime. 
+- [Kubernetes deployments](https://github.com/openvinotoolkit/model_server/blob/main/deploy). The server can be deployed in a Kubernetes cluster allowing the inference service to scale horizontally and ensure high availability.
+- [Model reshaping](https://github.com/openvinotoolkit/model_server/blob/main/docs/docker_container.md#model-reshaping). The server supports reshaping models in runtime.
 - [Model ensemble](https://github.com/openvinotoolkit/model_server/blob/main/docs/ensemble_scheduler.md) (preview). Connect multiple models to deploy complex processing solutions and reduce overhead of sending data back and forth.
 
 > **NOTE**: OVMS has been tested on CentOS\* and Ubuntu\*. Publicly released [Docker images](https://hub.docker.com/r/openvino/model_server) are based on CentOS.
@@ -68,30 +68,30 @@ For more detailed guides on using the Model Server in various scenarios, visit t
 
 ## API Documentation
 
-### GRPC 
+### GRPC
 
-OpenVINO&trade; Model Server gRPC API is documented in the proto buffer files in [tensorflow_serving_api](https://github.com/tensorflow/serving/tree/r2.2/tensorflow_serving/apis). 
+OpenVINO&trade; Model Server gRPC API is documented in the proto buffer files in [tensorflow_serving_api](https://github.com/tensorflow/serving/tree/r2.2/tensorflow_serving/apis).
 
-> **NOTE:** The implementations for `Predict`, `GetModelMetadata`, and `GetModelStatus` function calls are currently available. 
+> **NOTE:** The implementations for `Predict`, `GetModelMetadata`, and `GetModelStatus` function calls are currently available.
 > These are the most generic function calls and should address most of the usage scenarios.
 
-[Predict proto](https://github.com/tensorflow/serving/blob/r2.2/tensorflow_serving/apis/predict.proto) defines two message specifications: `PredictRequest` and `PredictResponse` used while calling Prediction endpoint.  
-* `PredictRequest` specifies information about the model spec, that is name and version, and a map of input data serialized via 
+[Predict proto](https://github.com/tensorflow/serving/blob/r2.2/tensorflow_serving/apis/predict.proto) defines two message specifications: `PredictRequest` and `PredictResponse` used while calling Prediction endpoint.
+* `PredictRequest` specifies information about the model spec, that is name and version, and a map of input data serialized via
 [TensorProto](https://github.com/tensorflow/tensorflow/blob/r2.2/tensorflow/core/framework/tensor.proto) to a string format.
-* `PredictResponse` includes a map of outputs serialized by 
+* `PredictResponse` includes a map of outputs serialized by
 [TensorProto](https://github.com/tensorflow/tensorflow/blob/r2.2/tensorflow/core/framework/tensor.proto) and information about the used model spec.
- 
+
 [Get Model Metadata proto](https://github.com/tensorflow/serving/blob/r2.2/tensorflow_serving/apis/get_model_metadata.proto) defines three message definitions used while calling Metadata endpoint:
  `SignatureDefMap`, `GetModelMetadataRequest`, `GetModelMetadataResponse`.
 
  A function call `GetModelMetadata` accepts model spec information as input and returns Signature Definition content in the format similar to TensorFlow Serving.
 
 [Get Model Status proto](https://github.com/tensorflow/serving/blob/r2.2/tensorflow_serving/apis/get_model_status.proto) defines three message definitions used while calling Status endpoint:
- `GetModelStatusRequest`, `ModelVersionStatus`, `GetModelStatusResponse` that report all exposed versions including their state in their lifecycle. 
+ `GetModelStatusRequest`, `ModelVersionStatus`, `GetModelStatusResponse` that report all exposed versions including their state in their lifecycle.
 
 Refer to the [example client code](https://github.com/openvinotoolkit/model_server/blob/main/example_client) to learn how to use this API and submit the requests using the gRPC interface.
 
-Using the gRPC interface is recommended for optimal performance due to its faster implementation of input data deserialization. It enables you to achieve lower latency, especially with larger input messages like images. 
+Using the gRPC interface is recommended for optimal performance due to its faster implementation of input data deserialization. It enables you to achieve lower latency, especially with larger input messages like images.
 
 ### REST
 
@@ -99,9 +99,9 @@ OpenVINO&trade; Model Server RESTful API follows the documentation from the [Ten
 
 Both row and column format of the requests are implemented.
 
-> **NOTE**: Just like with gRPC, only the implementations for `Predict`, `GetModelMetadata`, and `GetModelStatus` function calls are currently available. 
+> **NOTE**: Just like with gRPC, only the implementations for `Predict`, `GetModelMetadata`, and `GetModelStatus` function calls are currently available.
 
-Only the numerical data types are supported. 
+Only the numerical data types are supported.
 
 Review the exemplary clients below to find out more how to connect and run inference requests.
 
@@ -110,9 +110,9 @@ REST API is recommended when the primary goal is in reducing the number of clien
 
 ## Known Limitations
 
-* Currently, `Predict`, `GetModelMetadata`, and `GetModelStatus` calls are implemented using the TensorFlow Serving API. 
+* Currently, `Predict`, `GetModelMetadata`, and `GetModelStatus` calls are implemented using the TensorFlow Serving API.
 * `Classify`, `Regress`, and `MultiInference` are not included.
-* `Output_filter` is not effective in the `Predict` call. All outputs defined in the model are returned to the clients. 
+* `Output_filter` is not effective in the `Predict` call. All outputs defined in the model are returned to the clients.
 
 ## OpenVINO Model Server Contribution Policy
 
diff --git a/docs/optimization_guide/dldt_optimization_guide.md b/docs/optimization_guide/dldt_optimization_guide.md
index e70c0365a4165c..9ece7fec93a628 100644
--- a/docs/optimization_guide/dldt_optimization_guide.md
+++ b/docs/optimization_guide/dldt_optimization_guide.md
@@ -2,13 +2,13 @@
 
 ## Introduction
 
-The purpose of this document is to give you performance-related insights to every step of the network deployment process.  
+The purpose of this document is to give you performance-related insights to every step of the network deployment process.
 
 For information on the general workflow, refer to the documentation in <a href="#see-also">See Also</a>. For an example Inference Engine API snippet, see <a href="#new-request-based-api">Request-Based API and “GetBlob” Idiom</a>.
 
 ### Deep Learning Inference Engine Overview <a name="dldt-overview"></a>
 
-Deep Learning Inference Engine is a part of Intel&reg; Deep Learning Deployment Toolkit (Intel&reg; DL Deployment Toolkit) and OpenVINO&trade; toolkit. Inference Engine facilitates deployment of deep learning solutions by delivering a unified, device-agnostic API.  
+Deep Learning Inference Engine is a part of Intel&reg; Deep Learning Deployment Toolkit (Intel&reg; DL Deployment Toolkit) and OpenVINO&trade; toolkit. Inference Engine facilitates deployment of deep learning solutions by delivering a unified, device-agnostic API.
 
 Below, there are the three main steps of the deployment process:
 
@@ -50,7 +50,7 @@ When evaluating performance of your model with the Inference Engine, you must me
 
 ### Latency vs. Throughput <a name="latency-vs-throughput"></a>
 
-In the asynchronous case (see <a href="#new-request-based-api">Request-Based API and “GetBlob” Idiom</a>), the performance of an individual infer request is usually of less concern. Instead, you typically execute multiple requests asynchronously and measure the throughput in images per second by dividing the number of images that were processed by the processing time. 
+In the asynchronous case (see <a href="#new-request-based-api">Request-Based API and “GetBlob” Idiom</a>), the performance of an individual infer request is usually of less concern. Instead, you typically execute multiple requests asynchronously and measure the throughput in images per second by dividing the number of images that were processed by the processing time.
 In contrast, for the latency-oriented tasks, the time to a single frame is more important.
 
 Refer to the [Benchmark App](../../inference-engine/samples/benchmark_app/README.md) sample, which allows latency vs. throughput measuring.
@@ -114,23 +114,23 @@ The resulting IR precision, for instance, `FP16` or `FP32`, directly affects per
 
 ## Multi-Device Execution <a name="multi-device-optimizations"></a>
 OpenVINO&trade; toolkit supports automatic multi-device execution, please see [MULTI-Device plugin description](../IE_DG/supported_plugins/MULTI.md).
-In the next chapter you can find the device-specific tips, while this section covers few recommendations 
+In the next chapter you can find the device-specific tips, while this section covers few recommendations
 for the multi-device execution:
--	MULTI usually performs best when the fastest device is specified first in the list of the devices. 
-    This is particularly important when the parallelism is not sufficient 
+-	MULTI usually performs best when the fastest device is specified first in the list of the devices.
+    This is particularly important when the parallelism is not sufficient
     (e.g. the number of request in the flight is not enough to saturate all devices).
-- It is highly recommended to query the optimal number of inference requests directly from the instance of the ExecutionNetwork 
-  (resulted from the LoadNetwork call with the specific multi-device configuration as a parameter). 
-Please refer to the code of the [Benchmark App](../../inference-engine/samples/benchmark_app/README.md) sample for details.    
--   Notice that for example CPU+GPU execution performs better with certain knobs 
+- It is highly recommended to query the optimal number of inference requests directly from the instance of the ExecutionNetwork
+  (resulted from the LoadNetwork call with the specific multi-device configuration as a parameter).
+Please refer to the code of the [Benchmark App](../../inference-engine/samples/benchmark_app/README.md) sample for details.
+-   Notice that for example CPU+GPU execution performs better with certain knobs
     which you can find in the code of the same [Benchmark App](../../inference-engine/samples/benchmark_app/README.md) sample.
-    One specific example is disabling GPU driver polling, which in turn requires multiple GPU streams (which is already a default for the GPU) to amortize slower 
+    One specific example is disabling GPU driver polling, which in turn requires multiple GPU streams (which is already a default for the GPU) to amortize slower
     inference completion from the device to the host.
--	Multi-device logic always attempts to save on the (e.g. inputs) data copies between device-agnostic, user-facing inference requests 
-    and device-specific 'worker' requests that are being actually scheduled behind the scene. 
-    To facilitate the copy savings, it is recommended to start the requests in the order that they were created 
+-	Multi-device logic always attempts to save on the (e.g. inputs) data copies between device-agnostic, user-facing inference requests
+    and device-specific 'worker' requests that are being actually scheduled behind the scene.
+    To facilitate the copy savings, it is recommended to start the requests in the order that they were created
     (with ExecutableNetwork's CreateInferRequest).
-  
+
 
 ## Device-Specific Optimizations <a name="device-specific-optimizations"></a>
 
@@ -171,7 +171,7 @@ Notice that on a multi-socket machine, the bare minimum of streams for a latency
 
 In addition, you can play with the batch size to find the throughput sweet spot.
 
-If your application is hard or impossible to change in accordance with the multiple-requests logic, consider the "multiple-instance" trick to improve the throughput:  
+If your application is hard or impossible to change in accordance with the multiple-requests logic, consider the "multiple-instance" trick to improve the throughput:
 -   For multi-socket execution, it is recommended to set   [`KEY_CPU_THREADS_NUM`](../IE_DG/supported_plugins/CPU.md) to the number of cores per socket, and run as many instances of the application as you have sockets.
 -   Similarly, for extremely lightweight networks (running faster than 1ms) and/or many-core machines (16+ cores), try limiting the number of CPU inference threads to just `#&zwj;phys` cores and further, while trying to saturate the machine with running multiple instances of the application.
 
@@ -186,15 +186,15 @@ Inference Engine relies on the [Compute Library for Deep Neural Networks (clDNN)
 -	If your application is simultaneously using the inference on the CPU or otherwise loads the host heavily, make sure that the OpenCL driver threads do not starve. You can use [CPU configuration options](../IE_DG/supported_plugins/CPU.md) to limit number of inference threads for the CPU plugin.
 -	In the GPU-only scenario, a GPU driver might occupy a CPU core with spin-looped polling for completion. If the _CPU_ utilization is a concern, consider the `KEY_CLDND_PLUGIN_THROTTLE` configuration option.
 
-> **NOTE**: See the [Benchmark App Sample](../../inference-engine/samples/benchmark_app/README.md) code for a usage example. 
-Notice that while disabling the polling, this option might reduce the GPU performance, so usually this option is used with multiple [GPU streams](../IE_DG/supported_plugins/CL_DNN.md). 
+> **NOTE**: See the [Benchmark App Sample](../../inference-engine/samples/benchmark_app/README.md) code for a usage example.
+Notice that while disabling the polling, this option might reduce the GPU performance, so usually this option is used with multiple [GPU streams](../IE_DG/supported_plugins/GPU.md).
 
 
 ### Intel&reg; Movidius&trade; Myriad&trade; X Visual Processing Unit and Intel&reg; Vision Accelerator Design with Intel&reg; Movidius&trade; VPUs  <a name="myriad"></a>
 
 Since Intel&reg; Movidius&trade; Myriad&trade; X Visual Processing Unit (Intel&reg; Movidius&trade; Myriad&trade; 2 VPU) communicates with the host over USB, minimum four infer requests in flight are recommended to hide the data transfer costs. See <a href="#new-request-based-api">Request-Based API and “GetBlob” Idiom</a> and [Benchmark App Sample](../../inference-engine/samples/benchmark_app/README.md) for more information.
 
-Intel&reg; Vision Accelerator Design with Intel&reg; Movidius&trade; VPUs requires to keep at least 32 inference requests in flight to fully saturate the device.  
+Intel&reg; Vision Accelerator Design with Intel&reg; Movidius&trade; VPUs requires to keep at least 32 inference requests in flight to fully saturate the device.
 
 ### FPGA <a name="fpga"></a>
 
@@ -274,7 +274,7 @@ The following tips are provided to give general guidance on optimizing execution
 
 -	Generally, GPU performance is better on heavy kernels (like Convolutions) and large inputs. So if the network inference time is already too small (~1ms of execution time), using the GPU would unlikely give a boost.
 
--	A typical strategy to start with is to test the CPU-only and GPU-only scenarios first (with samples this is plain `-d CPU` or `-d GPU`). If there are specific kernels that are not supported by the GPU, the best option to try is the `HETERO:GPU,CPU` that automatically applies default splitting (based on the plugins layers support). Then, you can play with the manual affinity settings (for example, to further minimize the number of subgraphs).  
+-	A typical strategy to start with is to test the CPU-only and GPU-only scenarios first (with samples this is plain `-d CPU` or `-d GPU`). If there are specific kernels that are not supported by the GPU, the best option to try is the `HETERO:GPU,CPU` that automatically applies default splitting (based on the plugins layers support). Then, you can play with the manual affinity settings (for example, to further minimize the number of subgraphs).
 
 -	The general affinity “rule of thumb” is to keep computationally-intensive kernels on the accelerator, and "glue" (or helper) kernels on the CPU. Notice that this includes the granularity considerations. For example, running some (custom) activation on the CPU would result in too many conversions.
 
@@ -337,7 +337,7 @@ For inference on the CPU there are multiple threads binding options, see
 
 If you are building an app-level pipeline with third-party components like GStreamer*, the general guidance for NUMA machines is as follows:
 - Whenever possible, use at least one instance of the pipeline per NUMA node:
-   - Pin the _entire_ pipeline instance to the specific NUMA node at the outer-most level (for example, use Kubernetes* and/or `numactl` command with proper settings before  actual GStreamer commands). 
+   - Pin the _entire_ pipeline instance to the specific NUMA node at the outer-most level (for example, use Kubernetes* and/or `numactl` command with proper settings before  actual GStreamer commands).
    - Disable any individual pinning by the pipeline components (e.g. set [CPU_BIND_THREADS to 'NO'](../IE_DG/supported_plugins/CPU.md)).
    - Limit each instance with respect to number of inference threads. Use  [CPU_THREADS_NUM](../IE_DG/supported_plugins/CPU.md) or  or other means (e.g. virtualization, Kubernetes*, etc), to avoid oversubscription.
 - If pinning instancing/pinning of the entire pipeline is not possible or desirable, relax the inference threads pinning to just 'NUMA'.
@@ -416,7 +416,7 @@ If your application simultaneously executes multiple infer requests:
 
 -	For FPGA and GPU, the actual work is serialized by a plugin and/or a driver anyway.
 
-- 	Finally, for <a href="#myriad">any VPU flavor</a>, using multiple requests is a must for achieving good throughput. 
+- 	Finally, for <a href="#myriad">any VPU flavor</a>, using multiple requests is a must for achieving good throughput.
 
 In the Inference Engine, there is no notion of requests priorities. It is left to the user side (for example, not queuing the low priority infer request, until another higher priority is waiting). Notice that it would require additional logic to synchronize between executable networks (queues) in your application code.
 
@@ -470,12 +470,12 @@ Example of Inference Engine calls:
 	Notice that `Task_runNOThrow` is an Async API wrapper and it is executed in a different thread and triggers the Intel MKL-DNN execution:
 
 	![](../img/vtune_timeline.png)
-	
+
 -	In the Intel VTune Amplifier **Top-down view**, grouped by the **Task Domain**.
 	Notice the `Task_runNoThrow` and `MKLDNN _INFER` that are bracketing the actual Intel MKL-DNN kernels execution:
-	
+
 	![](../img/vtune_topdown_view.jpg)
-	
+
 Similarly, you can use any GPU analysis in the Intel VTune Amplifier and get general correlation with Inference Engine API as well as the execution breakdown for OpenCL kernels.
 
 Just like with regular native application, further drill down in the counters is possible, however, this is mostly useful for <a href="#optimizing-custom-kernels">optimizing custom kernels</a>. Finally, with the Intel VTune Amplifier, the profiling is not limited to your user-level code (see the [corresponding section in the Intel&reg; VTune&trade; Amplifier User's Guide](https://software.intel.com/en-us/vtune-amplifier-help-analyze-performance)).
@@ -513,12 +513,12 @@ Since FPGA execution does not separate individual kernels, only bulk execution/d
 
 ```
 subgraph1: 1. input preprocessing (mean data/FPGA):EXECUTED   layerType: preprocessing   realTime: 129     cpu: 129
-subgraph1: 2. input transfer to DDR:EXECUTED       layerType:                    realTime: 201        cpu: 0              
-subgraph1: 3. FPGA execute time:EXECUTED           layerType:                    realTime: 3808       cpu: 0              subgraph1: 4. output transfer from DDR:EXECUTED    layerType:                    realTime: 55         cpu: 0              
-subgraph1: 5. FPGA output postprocessing:EXECUTED  layerType:                    realTime: 7          cpu: 7              
-subgraph1: 6. softmax/copy:   EXECUTED       layerType:                    realTime: 2          cpu: 2              
-subgraph2: out_prob:          NOT_RUN        layerType: Output             realTime: 0          cpu: 0              
-subgraph2: prob:              EXECUTED       layerType: SoftMax            realTime: 10         cpu: 10             
+subgraph1: 2. input transfer to DDR:EXECUTED       layerType:                    realTime: 201        cpu: 0
+subgraph1: 3. FPGA execute time:EXECUTED           layerType:                    realTime: 3808       cpu: 0              subgraph1: 4. output transfer from DDR:EXECUTED    layerType:                    realTime: 55         cpu: 0
+subgraph1: 5. FPGA output postprocessing:EXECUTED  layerType:                    realTime: 7          cpu: 7
+subgraph1: 6. softmax/copy:   EXECUTED       layerType:                    realTime: 2          cpu: 2
+subgraph2: out_prob:          NOT_RUN        layerType: Output             realTime: 0          cpu: 0
+subgraph2: prob:              EXECUTED       layerType: SoftMax            realTime: 10         cpu: 10
 Total time: 4212     microseconds
 ```
 
diff --git a/docs/snippets/GPU_Kernel.cpp b/docs/snippets/GPU_Kernel.cpp
index 5f849eb6a6a6a9..8b21a79dfe27dd 100644
--- a/docs/snippets/GPU_Kernel.cpp
+++ b/docs/snippets/GPU_Kernel.cpp
@@ -1,5 +1,4 @@
 #include <ie_core.hpp>
-#include "cldnn/cldnn_config.hpp"
 
 int main() {
 using namespace InferenceEngine;
@@ -9,9 +8,5 @@ InferenceEngine::Core core;
 core.SetConfig({ { InferenceEngine::PluginConfigParams::KEY_CONFIG_FILE, "<path_to_the_xml_file>" } }, "GPU");
 //! [part0]
 
-//! [part1]
-core.SetConfig({ { PluginConfigParams::KEY_DUMP_KERNELS, PluginConfigParams::YES } }, "GPU");
-//! [part1]
-
 return 0;
 }
diff --git a/docs/snippets/GPU_Kernels_Tuning.cpp b/docs/snippets/GPU_Kernels_Tuning.cpp
deleted file mode 100644
index 25daeec5e2a263..00000000000000
--- a/docs/snippets/GPU_Kernels_Tuning.cpp
+++ /dev/null
@@ -1,14 +0,0 @@
-#include <ie_core.hpp>
-#include "cldnn/cldnn_config.hpp"
-
-int main() {
-using namespace InferenceEngine;
-//! [part0]
-Core ie;          
-  ie.SetConfig({{ CONFIG_KEY(TUNING_MODE), CONFIG_VALUE(TUNING_CREATE) }}, "GPU");
-  ie.SetConfig({{ CONFIG_KEY(TUNING_FILE), "/path/to/tuning/file.json" }}, "GPU");
-  // Further LoadNetwork calls will use the specified tuning parameters
-//! [part0]
-
-return 0;
-}
diff --git a/docs/snippets/GPU_RemoteBlob_API2.cpp b/docs/snippets/GPU_RemoteBlob_API2.cpp
index 1bb00c17e03e94..13597ae45617ba 100644
--- a/docs/snippets/GPU_RemoteBlob_API2.cpp
+++ b/docs/snippets/GPU_RemoteBlob_API2.cpp
@@ -1,6 +1,6 @@
 #include <ie_core.hpp>
 #include <gpu/gpu_context_api_va.hpp>
-#include <cldnn/cldnn_config.hpp>
+#include <gpu/gpu_config.hpp>
 
 
 int main() {
@@ -28,7 +28,7 @@ auto shared_va_context = gpu::make_shared_context(ie, "GPU", disp);
 // compile network within a shared context
 ExecutableNetwork executable_network = ie.LoadNetwork(network,
                                                       shared_va_context,
-                                                      { { CLDNNConfigParams::KEY_CLDNN_NV12_TWO_INPUTS,
+                                                      { { GPUConfigParams::KEY_GPU_NV12_TWO_INPUTS,
                                                           PluginConfigParams::YES } });
 
 
diff --git a/inference-engine/include/cldnn/cldnn_config.hpp b/inference-engine/include/cldnn/cldnn_config.hpp
index cbc2aef0242101..3e5dc4cfb126f6 100644
--- a/inference-engine/include/cldnn/cldnn_config.hpp
+++ b/inference-engine/include/cldnn/cldnn_config.hpp
@@ -11,47 +11,11 @@
 #pragma once
 
 #include "ie_plugin_config.hpp"
+#include "ie_api.h"
+#include "gpu/gpu_config.hpp"
 
 namespace InferenceEngine {
 
-namespace Metrics {
-
-/**
- * @def GPU_METRIC_KEY(name)
- * @brief shortcut for defining GPU plugin metrics
- */
-#define GPU_METRIC_KEY(name) METRIC_KEY(GPU_##name)
-#define DECLARE_GPU_METRIC_KEY(name, ...) DECLARE_METRIC_KEY(GPU_##name, __VA_ARGS__)
-
-/**
- * @def DECLARE_GPU_METRIC_VALUE(name)
- * @brief shortcut for defining gpu metric values
- */
-#define DECLARE_GPU_METRIC_VALUE(name) DECLARE_METRIC_VALUE(GPU_##name)
-
-/**
- * @brief Metric which defines size of memory in bytes available for the device. For iGPU it returns host memory size, for dGPU - dedicated gpu memory size
- */
-DECLARE_GPU_METRIC_KEY(DEVICE_TOTAL_MEM_SIZE, uint64_t);
-
-/**
- * @brief Metric to get microarchitecture identifier in major.minor.revision format
- */
-DECLARE_GPU_METRIC_KEY(UARCH_VERSION, std::string);
-
-/**
- * @brief Metric to get count of execution units for current GPU
- */
-DECLARE_GPU_METRIC_KEY(EXECUTION_UNITS_COUNT, int);
-
-/**
- * @brief Possible return value for OPTIMIZATION_CAPABILITIES metric
- *  - "HW_MATMUL" - Defines if device has hardware block for matrix multiplication
- */
-DECLARE_GPU_METRIC_VALUE(HW_MATMUL);
-
-}  // namespace Metrics
-
 /**
  * @brief GPU plugin configuration
  */
@@ -70,6 +34,7 @@ namespace CLDNNConfigParams {
  * this option should be used with an unsigned integer value (1 is lowest priority)
  * 0 means no priority hint is set and default queue is created.
  */
+INFERENCE_ENGINE_DEPRECATED("Use InferenceEngine::GPUConfigParams::GPU_PLUGIN_PRIORITY instead")
 DECLARE_CLDNN_CONFIG_KEY(PLUGIN_PRIORITY);
 
 /**
@@ -78,22 +43,26 @@ DECLARE_CLDNN_CONFIG_KEY(PLUGIN_PRIORITY);
  * chapter 9.19. This option should be used with an unsigned integer value (1 is lowest energy consumption)
  * 0 means no throttle hint is set and default queue created.
  */
+INFERENCE_ENGINE_DEPRECATED("Use InferenceEngine::GPUConfigParams::GPU_PLUGIN_THROTTLE instead")
 DECLARE_CLDNN_CONFIG_KEY(PLUGIN_THROTTLE);
 
 /**
  * @brief This key controls clDNN memory pool optimization.
  * Turned off by default.
  */
+INFERENCE_ENGINE_DEPRECATED("The config key will be removed")
 DECLARE_CLDNN_CONFIG_KEY(MEM_POOL);
 
 /**
  * @brief This key defines the directory name to which clDNN graph visualization will be dumped.
  */
+INFERENCE_ENGINE_DEPRECATED("The config key will be removed")
 DECLARE_CLDNN_CONFIG_KEY(GRAPH_DUMPS_DIR);
 
 /**
  * @brief This key defines the directory name to which full program sources will be dumped.
  */
+INFERENCE_ENGINE_DEPRECATED("The config key will be removed")
 DECLARE_CLDNN_CONFIG_KEY(SOURCES_DUMPS_DIR);
 
 /**
@@ -108,43 +77,19 @@ DECLARE_CLDNN_CONFIG_KEY(ENABLE_FP16_FOR_QUANTIZED_MODELS);
  * @brief This key should be set to correctly handle NV12 input without pre-processing.
  * Turned off by default.
  */
+INFERENCE_ENGINE_DEPRECATED("Use InferenceEngine::GPUConfigParams::GPU_NV12_TWO_INPUTS instead")
 DECLARE_CLDNN_CONFIG_KEY(NV12_TWO_INPUTS);
 
-/**
- * @brief This key sets the max number of host threads that can be used by GPU plugin on model loading.
- * Default value is maximum number of threads available in the environment.
- */
-DECLARE_CLDNN_CONFIG_KEY(MAX_NUM_THREADS);
-
-/**
- * @brief Turning on this key enables to unroll recurrent layers such as TensorIterator or Loop with fixed iteration count.
- * This key is turned on by default. Turning this key on will achieve better inference performance for loops with not too many iteration counts (less than 16, as a rule of thumb).
- * Turning this key off will achieve better performance for both graph loading time and inference time with many iteration counts (greater than 16).
- * Note that turning this key on will increase the graph loading time in proportion to the iteration counts.
- * Thus, this key should be turned off if graph loading time is considered to be most important target to optimize.*/
-DECLARE_CLDNN_CONFIG_KEY(ENABLE_LOOP_UNROLLING);
-
 }  // namespace CLDNNConfigParams
 
 namespace PluginConfigParams {
 
-/**
- * @brief Optimize GPU plugin execution to maximize throughput.
- *
- * It is passed to Core::SetConfig(), this option should be used with values:
- * - KEY_GPU_THROUGHPUT_AUTO creates bare minimum of streams that might improve performance in some cases,
- *   this option allows to enable throttle hint for opencl queue thus reduce CPU load without significant performance
- * drop
- * - a positive integer value creates the requested number of streams
- */
-DECLARE_CONFIG_VALUE(GPU_THROUGHPUT_AUTO);
-DECLARE_CONFIG_KEY(GPU_THROUGHPUT_STREAMS);
-
 /**
  * @brief This key enables dumping of the kernels used by the plugin for custom layers.
  *
  * This option should be used with values: PluginConfigParams::YES or PluginConfigParams::NO (default)
  */
+INFERENCE_ENGINE_DEPRECATED("The config key will be removed")
 DECLARE_CONFIG_KEY(DUMP_KERNELS);
 
 /**
@@ -159,17 +104,24 @@ DECLARE_CONFIG_KEY(DUMP_KERNELS);
  *
  * For values TUNING_CREATE and TUNING_RETUNE the file will be created if it does not exist.
  */
+INFERENCE_ENGINE_DEPRECATED("The config key will be removed")
 DECLARE_CONFIG_KEY(TUNING_MODE);
 
+INFERENCE_ENGINE_DEPRECATED("The config value will be removed")
 DECLARE_CONFIG_VALUE(TUNING_CREATE);
+INFERENCE_ENGINE_DEPRECATED("The config value will be removed")
 DECLARE_CONFIG_VALUE(TUNING_USE_EXISTING);
+INFERENCE_ENGINE_DEPRECATED("The config value will be removed")
 DECLARE_CONFIG_VALUE(TUNING_DISABLED);
+INFERENCE_ENGINE_DEPRECATED("The config value will be removed")
 DECLARE_CONFIG_VALUE(TUNING_UPDATE);
+INFERENCE_ENGINE_DEPRECATED("The config value will be removed")
 DECLARE_CONFIG_VALUE(TUNING_RETUNE);
 
 /**
  * @brief This key defines the tuning data filename to be created/used
  */
+INFERENCE_ENGINE_DEPRECATED("The config key will be removed")
 DECLARE_CONFIG_KEY(TUNING_FILE);
 
 }  // namespace PluginConfigParams
diff --git a/inference-engine/include/gpu/gpu_config.hpp b/inference-engine/include/gpu/gpu_config.hpp
new file mode 100644
index 00000000000000..96f8754ac8660a
--- /dev/null
+++ b/inference-engine/include/gpu/gpu_config.hpp
@@ -0,0 +1,120 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+/**
+ * @brief A header for advanced hardware related properties for GPU plugin
+ *        To use in SetConfig() method of plugins
+ *
+ * @file gpu_config.hpp
+ */
+#pragma once
+
+#include "ie_plugin_config.hpp"
+
+namespace InferenceEngine {
+
+namespace Metrics {
+
+/**
+ * @def GPU_METRIC_KEY(name)
+ * @brief shortcut for defining GPU plugin metrics
+ */
+#define GPU_METRIC_KEY(name) METRIC_KEY(GPU_##name)
+#define DECLARE_GPU_METRIC_KEY(name, ...) DECLARE_METRIC_KEY(GPU_##name, __VA_ARGS__)
+
+/**
+ * @def DECLARE_GPU_METRIC_VALUE(name)
+ * @brief shortcut for defining gpu metric values
+ */
+#define DECLARE_GPU_METRIC_VALUE(name) DECLARE_METRIC_VALUE(GPU_##name)
+
+/**
+ * @brief Metric which defines size of memory in bytes available for the device. For iGPU it returns host memory size, for dGPU - dedicated gpu memory size
+ */
+DECLARE_GPU_METRIC_KEY(DEVICE_TOTAL_MEM_SIZE, uint64_t);
+
+/**
+ * @brief Metric to get microarchitecture identifier in major.minor.revision format
+ */
+DECLARE_GPU_METRIC_KEY(UARCH_VERSION, std::string);
+
+/**
+ * @brief Metric to get count of execution units for current GPU
+ */
+DECLARE_GPU_METRIC_KEY(EXECUTION_UNITS_COUNT, int);
+
+/**
+ * @brief Possible return value for OPTIMIZATION_CAPABILITIES metric
+ *  - "HW_MATMUL" - Defines if device has hardware block for matrix multiplication
+ */
+DECLARE_GPU_METRIC_VALUE(HW_MATMUL);
+
+}  // namespace Metrics
+
+/**
+ * @brief GPU plugin configuration
+ */
+namespace GPUConfigParams {
+
+/**
+ * @brief shortcut for defining configuration keys
+ */
+#define GPU_CONFIG_KEY(name) InferenceEngine::GPUConfigParams::_CONFIG_KEY(GPU_##name)
+#define DECLARE_GPU_CONFIG_KEY(name) DECLARE_CONFIG_KEY(GPU_##name)
+#define DECLARE_GPU_CONFIG_VALUE(name) DECLARE_CONFIG_VALUE(GPU_##name)
+
+/**
+ * @brief This key instructs the GPU plugin to use the OpenCL queue priority hint
+ * as defined in https://www.khronos.org/registry/OpenCL/specs/opencl-2.1-extensions.pdf
+ * this option should be used with an unsigned integer value (1 is lowest priority)
+ * 0 means no priority hint is set and default queue is created.
+ */
+DECLARE_GPU_CONFIG_KEY(PLUGIN_PRIORITY);
+
+/**
+ * @brief This key instructs the GPU plugin to use throttle hints the OpenCL queue throttle hint
+ * as defined in https://www.khronos.org/registry/OpenCL/specs/opencl-2.1-extensions.pdf,
+ * chapter 9.19. This option should be used with an unsigned integer value (1 is lowest energy consumption)
+ * 0 means no throttle hint is set and default queue created.
+ */
+DECLARE_GPU_CONFIG_KEY(PLUGIN_THROTTLE);
+
+/**
+ * @brief This key should be set to correctly handle NV12 input without pre-processing.
+ * Turned off by default.
+ */
+DECLARE_GPU_CONFIG_KEY(NV12_TWO_INPUTS);
+
+/**
+ * @brief This key sets the max number of host threads that can be used by GPU plugin on model loading.
+ * Default value is maximum number of threads available in the environment.
+ */
+DECLARE_GPU_CONFIG_KEY(MAX_NUM_THREADS);
+
+/**
+ * @brief Turning on this key enables to unroll recurrent layers such as TensorIterator or Loop with fixed iteration count.
+ * This key is turned on by default. Turning this key on will achieve better inference performance for loops with not too many iteration counts (less than 16, as a rule of thumb).
+ * Turning this key off will achieve better performance for both graph loading time and inference time with many iteration counts (greater than 16).
+ * Note that turning this key on will increase the graph loading time in proportion to the iteration counts.
+ * Thus, this key should be turned off if graph loading time is considered to be most important target to optimize.*/
+DECLARE_GPU_CONFIG_KEY(ENABLE_LOOP_UNROLLING);
+
+}  // namespace GPUConfigParams
+
+namespace PluginConfigParams {
+
+/**
+ * @brief Optimize GPU plugin execution to maximize throughput.
+ *
+ * It is passed to Core::SetConfig(), this option should be used with values:
+ * - KEY_GPU_THROUGHPUT_AUTO creates bare minimum of streams that might improve performance in some cases,
+ *   this option allows to enable throttle hint for opencl queue thus reduce CPU load without significant performance
+ * drop
+ * - a positive integer value creates the requested number of streams
+ */
+DECLARE_CONFIG_VALUE(GPU_THROUGHPUT_AUTO);
+DECLARE_CONFIG_KEY(GPU_THROUGHPUT_STREAMS);
+}  // namespace PluginConfigParams
+
+}  // namespace InferenceEngine
diff --git a/inference-engine/samples/benchmark_app/main.cpp b/inference-engine/samples/benchmark_app/main.cpp
index 849dc05ad3344a..cd7ddc641dc256 100644
--- a/inference-engine/samples/benchmark_app/main.cpp
+++ b/inference-engine/samples/benchmark_app/main.cpp
@@ -4,8 +4,8 @@
 
 #include <algorithm>
 #include <chrono>
-#include <cldnn/cldnn_config.hpp>
 #include <gna/gna_config.hpp>
+#include <gpu/gpu_config.hpp>
 #include <inference_engine.hpp>
 #include <map>
 #include <memory>
@@ -282,7 +282,7 @@ int main(int argc, char* argv[]) {
                                << "which releases another CPU thread (that is otherwise "
                                   "used by the GPU driver for active polling)"
                                << slog::endl;
-                    device_config[CLDNN_CONFIG_KEY(PLUGIN_THROTTLE)] = "1";
+                    device_config[GPU_CONFIG_KEY(PLUGIN_THROTTLE)] = "1";
                 }
             } else if (device == "MYRIAD") {
                 device_config[CONFIG_KEY(LOG_LEVEL)] = CONFIG_VALUE(LOG_WARNING);
diff --git a/inference-engine/samples/hello_query_device/README.md b/inference-engine/samples/hello_query_device/README.md
index a185147f8ec50e..059077c48ad6b4 100644
--- a/inference-engine/samples/hello_query_device/README.md
+++ b/inference-engine/samples/hello_query_device/README.md
@@ -63,20 +63,20 @@ Available devices:
                 SUPPORTED_METRICS : [ AVAILABLE_DEVICES SUPPORTED_METRICS FULL_DEVICE_NAME OPTIMIZATION_CAPABILITIES SUPPORTED_CONFIG_KEYS RANGE_FOR_ASYNC_INFER_REQUESTS RANGE_FOR_STREAMS ]
                 FULL_DEVICE_NAME : Intel(R) UHD Graphics 620 (iGPU)
                 OPTIMIZATION_CAPABILITIES : [ FP32 BIN FP16 ]
-                SUPPORTED_CONFIG_KEYS : [ CACHE_DIR CLDNN_ENABLE_FP16_FOR_QUANTIZED_MODELS CLDNN_GRAPH_DUMPS_DIR CLDNN_MAX_NUM_THREADS CLDNN_MEM_POOL CLDNN_NV12_TWO_INPUTS CLDNN_PLUGIN_PRIORITY CLDNN_PLUGIN_THROTTLE CLDNN_SOURCES_DUMPS_DIR CLDNN_ENABLE_LOOP_UNROLLING CONFIG_FILE DEVICE_ID DUMP_KERNELS DYN_BATCH_ENABLED EXCLUSIVE_ASYNC_REQUESTS GPU_THROUGHPUT_STREAMS PERF_COUNT TUNING_FILE TUNING_MODE ]
+                SUPPORTED_CONFIG_KEYS : [ CACHE_DIR CLDNN_ENABLE_FP16_FOR_QUANTIZED_MODELS CLDNN_GRAPH_DUMPS_DIR GPU_MAX_NUM_THREADS CLDNN_MEM_POOL CLDNN_NV12_TWO_INPUTS CLDNN_PLUGIN_PRIORITY CLDNN_PLUGIN_THROTTLE CLDNN_SOURCES_DUMPS_DIR GPU_ENABLE_LOOP_UNROLLING CONFIG_FILE DEVICE_ID DUMP_KERNELS DYN_BATCH_ENABLED EXCLUSIVE_ASYNC_REQUESTS GPU_THROUGHPUT_STREAMS PERF_COUNT TUNING_FILE TUNING_MODE ]
                 RANGE_FOR_ASYNC_INFER_REQUESTS : { 1, 2, 1 }
                 RANGE_FOR_STREAMS : { 1, 2 }
         Default values for device configuration keys:
                 CACHE_DIR : ""
                 CLDNN_ENABLE_FP16_FOR_QUANTIZED_MODELS : YES
                 CLDNN_GRAPH_DUMPS_DIR : ""
-                CLDNN_MAX_NUM_THREADS : 8
                 CLDNN_MEM_POOL : YES
                 CLDNN_NV12_TWO_INPUTS : NO
                 CLDNN_PLUGIN_PRIORITY : 0
                 CLDNN_PLUGIN_THROTTLE : 0
                 CLDNN_SOURCES_DUMPS_DIR : ""
-                CLDNN_ENABLE_LOOP_UNROLLING : YES
+                GPU_MAX_NUM_THREADS : 8
+                GPU_ENABLE_LOOP_UNROLLING : YES
                 CONFIG_FILE : ""
                 DEVICE_ID : ""
                 DUMP_KERNELS : NO
diff --git a/inference-engine/src/cldnn_engine/cldnn_config.cpp b/inference-engine/src/cldnn_engine/cldnn_config.cpp
index ff5d96935224cb..3de19bdff87dcc 100644
--- a/inference-engine/src/cldnn_engine/cldnn_config.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_config.cpp
@@ -5,6 +5,7 @@
 #include <sys/stat.h>
 
 #include <cldnn/cldnn_config.hpp>
+#include <gpu/gpu_config.hpp>
 #include "cldnn_config.h"
 #include "cpp_interfaces/interface/ie_internal_plugin_config.hpp"
 #include "ie_api.h"
@@ -39,6 +40,7 @@ static void createDirectory(std::string _path) {
     }
 }
 
+IE_SUPPRESS_DEPRECATED_START
 void Config::UpdateFromMap(const std::map<std::string, std::string>& configMap) {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNNPlugin, "Config::UpdateFromMap");
     for (auto& kvp : configMap) {
@@ -69,7 +71,8 @@ void Config::UpdateFromMap(const std::map<std::string, std::string>& configMap)
             } else {
                 IE_THROW(NotFound) << "Unsupported property value by plugin: " << val;
             }
-        } else if (key.compare(CLDNNConfigParams::KEY_CLDNN_PLUGIN_PRIORITY) == 0) {
+        } else if (key.compare(GPUConfigParams::KEY_GPU_PLUGIN_PRIORITY) == 0 ||
+                   key.compare(CLDNNConfigParams::KEY_CLDNN_PLUGIN_PRIORITY) == 0) {
             std::stringstream ss(val);
             uint32_t uVal(0);
             ss >> uVal;
@@ -93,7 +96,8 @@ void Config::UpdateFromMap(const std::map<std::string, std::string>& configMap)
                     IE_THROW(ParameterMismatch) << "Unsupported queue priority value: " << uVal;
             }
 
-        } else if (key.compare(CLDNNConfigParams::KEY_CLDNN_PLUGIN_THROTTLE) == 0) {
+        } else if (key.compare(GPUConfigParams::KEY_GPU_PLUGIN_THROTTLE) == 0 ||
+                   key.compare(CLDNNConfigParams::KEY_CLDNN_PLUGIN_THROTTLE) == 0) {
             std::stringstream ss(val);
             uint32_t uVal(0);
             ss >> uVal;
@@ -205,7 +209,8 @@ void Config::UpdateFromMap(const std::map<std::string, std::string>& configMap)
             } else {
                 IE_THROW(NotFound) << "Unsupported property value by plugin: " << val;
             }
-        } else if (key.compare(CLDNNConfigParams::KEY_CLDNN_NV12_TWO_INPUTS) == 0) {
+        } else if (key.compare(GPUConfigParams::KEY_GPU_NV12_TWO_INPUTS) == 0 ||
+                   key.compare(CLDNNConfigParams::KEY_CLDNN_NV12_TWO_INPUTS) == 0) {
             if (val.compare(PluginConfigParams::YES) == 0) {
                 nv12_two_inputs = true;
             } else if (val.compare(PluginConfigParams::NO) == 0) {
@@ -221,7 +226,7 @@ void Config::UpdateFromMap(const std::map<std::string, std::string>& configMap)
             } else {
                 IE_THROW(NotFound) << "Unsupported KEY_CLDNN_ENABLE_FP16_FOR_QUANTIZED_MODELS flag value: " << val;
             }
-        } else if (key.compare(CLDNNConfigParams::KEY_CLDNN_MAX_NUM_THREADS) == 0) {
+        } else if (key.compare(GPUConfigParams::KEY_GPU_MAX_NUM_THREADS) == 0) {
             int max_threads = std::max(1, static_cast<int>(std::thread::hardware_concurrency()));
             try {
                 int val_i = std::stoi(val);
@@ -231,17 +236,17 @@ void Config::UpdateFromMap(const std::map<std::string, std::string>& configMap)
                     n_threads = val_i;
                 }
             } catch (const std::exception&) {
-                IE_THROW() << "Wrong value for property key " << CLDNNConfigParams::KEY_CLDNN_MAX_NUM_THREADS << ": " << val
+                IE_THROW() << "Wrong value for property key " << GPUConfigParams::KEY_GPU_MAX_NUM_THREADS << ": " << val
                                    << "\nSpecify the number of threads use for build as an integer."
                                    << "\nOut of range value will be set as a default value, maximum concurrent threads.";
             }
-        } else if (key.compare(CLDNNConfigParams::KEY_CLDNN_ENABLE_LOOP_UNROLLING) == 0) {
+        } else if (key.compare(GPUConfigParams::KEY_GPU_ENABLE_LOOP_UNROLLING) == 0) {
             if (val.compare(PluginConfigParams::YES) == 0) {
                 enable_loop_unrolling = true;
             } else if (val.compare(PluginConfigParams::NO) == 0) {
                 enable_loop_unrolling = false;
             } else {
-                IE_THROW(ParameterMismatch) << "Unsupported KEY_CLDNN_ENABLE_LOOP_UNROLLING flag value: " << val;
+                IE_THROW(ParameterMismatch) << "Unsupported KEY_GPU_ENABLE_LOOP_UNROLLING flag value: " << val;
             }
         } else {
             IE_THROW(NotFound) << "Unsupported property key by plugin: " << key;
@@ -297,6 +302,7 @@ void Config::adjustKeyMapValues() {
         default: break;
         }
         key_config_map[CLDNNConfigParams::KEY_CLDNN_PLUGIN_PRIORITY] = qp;
+        key_config_map[GPUConfigParams::KEY_GPU_PLUGIN_PRIORITY] = qp;
     }
     {
         std::string qt = "0";
@@ -307,6 +313,7 @@ void Config::adjustKeyMapValues() {
         default: break;
         }
         key_config_map[CLDNNConfigParams::KEY_CLDNN_PLUGIN_THROTTLE] = qt;
+        key_config_map[GPUConfigParams::KEY_GPU_PLUGIN_THROTTLE] = qt;
     }
     {
         std::string tm = PluginConfigParams::TUNING_DISABLED;
@@ -328,11 +335,13 @@ void Config::adjustKeyMapValues() {
     key_config_map[PluginConfigParams::KEY_GPU_THROUGHPUT_STREAMS] = std::to_string(throughput_streams);
     key_config_map[PluginConfigParams::KEY_DEVICE_ID] = device_id;
     key_config_map[PluginConfigParams::KEY_CONFIG_FILE] = "";
-    key_config_map[CLDNNConfigParams::KEY_CLDNN_MAX_NUM_THREADS] = std::to_string(n_threads);
+    key_config_map[GPUConfigParams::KEY_GPU_MAX_NUM_THREADS] = std::to_string(n_threads);
 
     if (enable_loop_unrolling)
-        key_config_map[CLDNNConfigParams::KEY_CLDNN_ENABLE_LOOP_UNROLLING] = PluginConfigParams::YES;
+        key_config_map[GPUConfigParams::KEY_GPU_ENABLE_LOOP_UNROLLING] = PluginConfigParams::YES;
     else
-        key_config_map[CLDNNConfigParams::KEY_CLDNN_ENABLE_LOOP_UNROLLING] = PluginConfigParams::NO;
+        key_config_map[GPUConfigParams::KEY_GPU_ENABLE_LOOP_UNROLLING] = PluginConfigParams::NO;
 }
+IE_SUPPRESS_DEPRECATED_END
+
 }  // namespace CLDNNPlugin
diff --git a/inference-engine/src/cldnn_engine/cldnn_engine.cpp b/inference-engine/src/cldnn_engine/cldnn_engine.cpp
index 86b9f2e4b9526b..171919a80775d6 100644
--- a/inference-engine/src/cldnn_engine/cldnn_engine.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_engine.cpp
@@ -79,7 +79,7 @@
 #include "cldnn_executable_network.h"
 #include "cldnn_custom_layer.h"
 #include "cldnn_itt.h"
-#include "cldnn/cldnn_config.hpp"
+#include "gpu/gpu_config.hpp"
 
 #ifdef __linux__
 # include <dlfcn.h>
diff --git a/inference-engine/src/cldnn_engine/cldnn_executable_network.cpp b/inference-engine/src/cldnn_engine/cldnn_executable_network.cpp
index c2289fa9fb0bde..5191da35c2e369 100644
--- a/inference-engine/src/cldnn_engine/cldnn_executable_network.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_executable_network.cpp
@@ -16,7 +16,6 @@
 #include "cldnn_itt.h"
 
 #include <description_buffer.hpp>
-#include <cldnn/cldnn_config.hpp>
 #include "cldnn_infer_request.h"
 #include <threading/ie_executor_manager.hpp>
 #include "cldnn_async_infer_request.h"
diff --git a/inference-engine/src/cldnn_engine/cldnn_graph.cpp b/inference-engine/src/cldnn_engine/cldnn_graph.cpp
index 04d40c9815d00d..1f835d8ac2c991 100644
--- a/inference-engine/src/cldnn_engine/cldnn_graph.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_graph.cpp
@@ -16,7 +16,6 @@
 #include "cldnn_graph.h"
 #include "simple_math.h"
 #include <description_buffer.hpp>
-#include <cldnn/cldnn_config.hpp>
 #include "cldnn_infer_request.h"
 #include <threading/ie_executor_manager.hpp>
 #include <fstream>
diff --git a/inference-engine/tests/functional/plugin/gpu/remote_blob_tests/cldnn_remote_blob_tests.cpp b/inference-engine/tests/functional/plugin/gpu/remote_blob_tests/cldnn_remote_blob_tests.cpp
index e6415688de2484..6837c0b84c33a8 100644
--- a/inference-engine/tests/functional/plugin/gpu/remote_blob_tests/cldnn_remote_blob_tests.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/remote_blob_tests/cldnn_remote_blob_tests.cpp
@@ -9,7 +9,7 @@
 
 #include <ie_compound_blob.h>
 
-#include <cldnn/cldnn_config.hpp>
+#include <gpu/gpu_config.hpp>
 #include <remote_blob_tests/remote_blob_helpers.hpp>
 #include <common_test_utils/test_common.hpp>
 #include <functional_test_utils/plugin_cache.hpp>
@@ -175,7 +175,7 @@ TEST_P(BatchedBlob_Test, canInputNV12) {
 
     /* XXX: is it correct to set KEY_CLDNN_NV12_TWO_INPUTS in case of remote blob? */
     auto exec_net_b = ie.LoadNetwork(net_remote, CommonTestUtils::DEVICE_GPU,
-                { { CLDNNConfigParams::KEY_CLDNN_NV12_TWO_INPUTS, PluginConfigParams::YES} });
+                { { GPUConfigParams::KEY_GPU_NV12_TWO_INPUTS, PluginConfigParams::YES} });
     auto inf_req_remote = exec_net_b.CreateInferRequest();
     auto cldnn_context = exec_net_b.GetContext();
     cl_context ctx = std::dynamic_pointer_cast<ClContext>(cldnn_context)->get();
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp
index e21d610db569cb..a8c039e43915a4 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp
@@ -4,6 +4,7 @@
 
 #include "behavior/config.hpp"
 #include "cldnn/cldnn_config.hpp"
+#include "gpu/gpu_config.hpp"
 
 using namespace BehaviorTestsDefinitions;
 namespace {
@@ -12,6 +13,7 @@ namespace {
             InferenceEngine::Precision::FP16
     };
 
+    IE_SUPPRESS_DEPRECATED_START
     const std::vector<std::map<std::string, std::string>> inconfigs = {
             {{InferenceEngine::PluginConfigParams::KEY_GPU_THROUGHPUT_STREAMS, "OFF"}},
             {{InferenceEngine::PluginConfigParams::KEY_PERF_COUNT, "ON"}},
@@ -46,6 +48,7 @@ namespace {
             {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_GPU},
                     {InferenceEngine::PluginConfigParams::KEY_DEVICE_ID, "DEVICE_UNKNOWN"}}
     };
+    IE_SUPPRESS_DEPRECATED_END
 
     INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigTests,
             ::testing::Combine(
@@ -73,6 +76,29 @@ namespace {
             {}
     };
 
+    IE_SUPPRESS_DEPRECATED_START
+    const std::vector<std::map<std::string, std::string>> conf_gpu = {
+            // Deprecated
+            {{InferenceEngine::CLDNNConfigParams::KEY_CLDNN_NV12_TWO_INPUTS, InferenceEngine::PluginConfigParams::YES}},
+            {{InferenceEngine::CLDNNConfigParams::KEY_CLDNN_NV12_TWO_INPUTS, InferenceEngine::PluginConfigParams::NO}},
+            {{InferenceEngine::CLDNNConfigParams::KEY_CLDNN_PLUGIN_THROTTLE, "0"}},
+            {{InferenceEngine::CLDNNConfigParams::KEY_CLDNN_PLUGIN_THROTTLE, "1"}},
+            {{InferenceEngine::CLDNNConfigParams::KEY_CLDNN_PLUGIN_PRIORITY, "0"}},
+            {{InferenceEngine::CLDNNConfigParams::KEY_CLDNN_PLUGIN_PRIORITY, "1"}},
+
+            {{InferenceEngine::GPUConfigParams::KEY_GPU_NV12_TWO_INPUTS, InferenceEngine::PluginConfigParams::YES}},
+            {{InferenceEngine::GPUConfigParams::KEY_GPU_NV12_TWO_INPUTS, InferenceEngine::PluginConfigParams::NO}},
+            {{InferenceEngine::GPUConfigParams::KEY_GPU_PLUGIN_THROTTLE, "0"}},
+            {{InferenceEngine::GPUConfigParams::KEY_GPU_PLUGIN_THROTTLE, "1"}},
+            {{InferenceEngine::GPUConfigParams::KEY_GPU_PLUGIN_PRIORITY, "0"}},
+            {{InferenceEngine::GPUConfigParams::KEY_GPU_PLUGIN_PRIORITY, "1"}},
+            {{InferenceEngine::GPUConfigParams::KEY_GPU_MAX_NUM_THREADS, "1"}},
+            {{InferenceEngine::GPUConfigParams::KEY_GPU_MAX_NUM_THREADS, "4"}},
+            {{InferenceEngine::GPUConfigParams::KEY_GPU_ENABLE_LOOP_UNROLLING, InferenceEngine::PluginConfigParams::YES}},
+            {{InferenceEngine::GPUConfigParams::KEY_GPU_ENABLE_LOOP_UNROLLING, InferenceEngine::PluginConfigParams::NO}},
+    };
+    IE_SUPPRESS_DEPRECATED_END
+
     const std::vector<std::map<std::string, std::string>> multiconf = {
             {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES , CommonTestUtils::DEVICE_GPU}}
     };
@@ -92,6 +118,13 @@ namespace {
                 ::testing::ValuesIn(conf)),
             CorrectConfigAPITests::getTestCaseName);
 
+    INSTANTIATE_TEST_CASE_P(smoke_GPU_BehaviorTests, CorrectConfigAPITests,
+            ::testing::Combine(
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(CommonTestUtils::DEVICE_GPU),
+                ::testing::ValuesIn(conf_gpu)),
+            CorrectConfigAPITests::getTestCaseName);
+
     INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigAPITests,
             ::testing::Combine(
                     ::testing::ValuesIn(netPrecisions),
@@ -142,4 +175,4 @@ namespace {
                             IncorrectConfigAPITests::getTestCaseName);
 
 
-} // namespace
\ No newline at end of file
+} // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_integration.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_integration.cpp
index 3765c75864fa5f..68b23831e47944 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_integration.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_integration.cpp
@@ -11,7 +11,7 @@
 #endif
 #include "gpu/gpu_context_api_ocl.hpp"
 
-#include "cldnn/cldnn_config.hpp"
+#include "gpu/gpu_config.hpp"
 
 using namespace BehaviorTestsDefinitions;
 
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp
index e15ea827caa814..59f4dd21677c24 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp
@@ -3,7 +3,7 @@
 //
 
 #include "behavior/infer_request_input.hpp"
-#include "cldnn/cldnn_config.hpp"
+#include "gpu/gpu_config.hpp"
 
 using namespace BehaviorTestsDefinitions;
 namespace {
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp
index 6c38f5c841c0b0..1135f6d9f7d8e4 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp
@@ -3,7 +3,7 @@
 //
 
 #include "behavior/infer_request_output.hpp"
-#include "cldnn/cldnn_config.hpp"
+#include "gpu/gpu_config.hpp"
 
 using namespace BehaviorTestsDefinitions;
 namespace {
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp
index 51979116646939..729bf57c64ac8a 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp
@@ -3,7 +3,7 @@
 //
 
 #include "behavior/test_plugin.hpp"
-#include "cldnn/cldnn_config.hpp"
+#include "gpu/gpu_config.hpp"
 
 using namespace BehaviorTestsDefinitions;
 namespace {
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/multi/gpu_remote_blob_tests.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/multi/gpu_remote_blob_tests.cpp
index da308c032e258b..4fffb2cad6e157 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/multi/gpu_remote_blob_tests.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/multi/gpu_remote_blob_tests.cpp
@@ -4,7 +4,7 @@
 
 #include <string>
 #include <vector>
-#include "cldnn/cldnn_config.hpp"
+#include "gpu/gpu_config.hpp"
 #include "multi/multi_remote_blob_tests.hpp"
 #include "common_test_utils/test_constants.hpp"
 
diff --git a/inference-engine/tests/functional/plugin/gpu/single_layer_tests/tensor_iterator.cpp b/inference-engine/tests/functional/plugin/gpu/single_layer_tests/tensor_iterator.cpp
index aae2e0db8fa3ac..bce1ef106917f2 100644
--- a/inference-engine/tests/functional/plugin/gpu/single_layer_tests/tensor_iterator.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/single_layer_tests/tensor_iterator.cpp
@@ -9,7 +9,7 @@
 #include <map>
 #include <vector>
 #include <ngraph/op/util/attr_types.hpp>
-#include <cldnn/cldnn_config.hpp>
+#include <gpu/gpu_config.hpp>
 #include <transformations/control_flow/unroll_tensor_iterator.hpp>
 #include "common_test_utils/test_constants.hpp"
 #include "ie_api.h"
@@ -289,8 +289,8 @@ namespace {
                 InferenceEngine::Precision::FP16,
             }), // precision
             ::testing::ValuesIn(std::vector<Config> {
-                {CommonTestUtils::DEVICE_GPU, {{CLDNNConfigParams::KEY_CLDNN_ENABLE_LOOP_UNROLLING, PluginConfigParams::YES}}},
-                {CommonTestUtils::DEVICE_GPU, {{CLDNNConfigParams::KEY_CLDNN_ENABLE_LOOP_UNROLLING, PluginConfigParams::NO}}}
+                {CommonTestUtils::DEVICE_GPU, {{GPUConfigParams::KEY_GPU_ENABLE_LOOP_UNROLLING, PluginConfigParams::YES}}},
+                {CommonTestUtils::DEVICE_GPU, {{GPUConfigParams::KEY_GPU_ENABLE_LOOP_UNROLLING, PluginConfigParams::NO}}}
             })), // configuration
         TensorIteratorWithConfigTest::getTestCaseName);
 }  // namespace
diff --git a/tools/benchmark/main.py b/tools/benchmark/main.py
index 29aff45742e8c1..26ef6246f0c9d5 100644
--- a/tools/benchmark/main.py
+++ b/tools/benchmark/main.py
@@ -152,7 +152,7 @@ def set_throughput_streams():
                 if MULTI_DEVICE_NAME in device_name and CPU_DEVICE_NAME in device_name:
                     logger.warning("Turn on GPU trottling. Multi-device execution with the CPU + GPU performs best with GPU trottling hint, " +
                                    "which releases another CPU thread (that is otherwise used by the GPU driver for active polling)")
-                    config[device]['CLDNN_PLUGIN_THROTTLE'] = '1'
+                    config[device]['GPU_PLUGIN_THROTTLE'] = '1'
             elif device == MYRIAD_DEVICE_NAME:
                 set_throughput_streams()
                 config[device]['LOG_LEVEL'] = 'LOG_INFO'

From 2c3abf8f42e9a47d9e701779d333a732bbb24f68 Mon Sep 17 00:00:00 2001
From: Roman Kazantsev <roman.kazantsev@intel.com>
Date: Wed, 9 Jun 2021 10:00:20 +0300
Subject: [PATCH 039/178] Use int64 for TopK indices output (#6085)

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
---
 model-optimizer/extensions/front/onnx/top_k_ext.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/model-optimizer/extensions/front/onnx/top_k_ext.py b/model-optimizer/extensions/front/onnx/top_k_ext.py
index bae2bec317686a..e073c593531eed 100644
--- a/model-optimizer/extensions/front/onnx/top_k_ext.py
+++ b/model-optimizer/extensions/front/onnx/top_k_ext.py
@@ -1,6 +1,8 @@
 # Copyright (C) 2018-2021 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 
+import numpy as np
+
 from extensions.ops.topk import TopK
 from mo.front.extractor import FrontExtractorOp
 from mo.front.onnx.extractors.utils import onnx_attr, onnx_node_has_attr
@@ -18,7 +20,8 @@ def extract(cls, node):
         TopK-11 (k as input, sorting manipulations through `sorted` and `largest` attrs)
         """
         attrs = {
-            'axis': onnx_attr(node, 'axis', 'i', default=-1)
+            'axis': onnx_attr(node, 'axis', 'i', default=-1),
+            'index_element_type': np.int64
         }
         if onnx_node_has_attr(node, 'k'):
             attrs['k'] = onnx_attr(node, 'k', 'i')

From 6e2d13937aea1b6a694767b49bb4ac8508f306e8 Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Wed, 9 Jun 2021 10:09:25 +0300
Subject: [PATCH 040/178] ImportNetwork with explicit device name only (#5689)

* Import with explicit name

* Fixed LoadHetero_MultiArchs tests

* Fixed MYRIAD tests on Windows

* Fixed compilation in tests

* Updated tesets

* Fixed test

* Removed useless lines

* Removed custom VPU tests, replaced with common ones

* Fixed Windows

* Reverted SKIP_IF_NOT_IMPLEMENTED macro
---
 docs/IE_PLUGIN_DG/ExecutableNetwork.md        |  11 +-
 docs/IE_PLUGIN_DG/Plugin.md                   |  10 +-
 .../src/template_executable_network.cpp       |   8 +-
 .../src/template_executable_network.hpp       |   2 +-
 docs/template_plugin/src/template_plugin.cpp  |   8 +-
 docs/template_plugin/src/template_plugin.hpp  |   2 +-
 inference-engine/include/ie_core.hpp          |  11 +-
 .../hetero_executable_network.cpp             |   2 +-
 .../hetero_executable_network.hpp             |   2 +-
 .../src/hetero_plugin/hetero_plugin.cpp       |   9 +-
 .../src/hetero_plugin/hetero_plugin.hpp       |   5 +-
 .../ie_iexecutable_network_internal.cpp       |  21 +-
 .../interface/ie_iplugin_internal.cpp         |  43 +-
 .../src/inference_engine/ie_core.cpp          |  52 +--
 .../ie_iexecutable_network_internal.hpp       |   8 -
 .../interface/ie_iplugin_internal.hpp         |  27 +-
 inference-engine/src/plugin_api/ie_icore.hpp  |  12 -
 .../myriad_plugin/myriad_executable_network.h |  10 -
 .../src/vpu/myriad_plugin/myriad_plugin.cpp   |  14 -
 .../src/vpu/myriad_plugin/myriad_plugin.h     |   4 -
 .../inference_engine/caching_test.cpp         | 377 +++++++++---------
 .../skip_tests_config.cpp                     |   2 +
 .../skip_tests_config.cpp                     |   2 +
 .../behavior/core_integration.cpp             |  35 +-
 .../include/behavior/core_integration.hpp     |  49 ++-
 .../impl/mock_inference_plugin_internal.hpp   |   7 +-
 .../mock_iexecutable_network_internal.hpp     |   4 -
 .../mocks/mock_engine/mock_plugin.cpp         |  10 +-
 .../mocks/mock_engine/mock_plugin.hpp         |   8 +-
 .../cpp_interfaces/ie_plugin_test.cpp         |  23 --
 30 files changed, 316 insertions(+), 462 deletions(-)

diff --git a/docs/IE_PLUGIN_DG/ExecutableNetwork.md b/docs/IE_PLUGIN_DG/ExecutableNetwork.md
index c5bfd889857793..ae82b05e4edce3 100644
--- a/docs/IE_PLUGIN_DG/ExecutableNetwork.md
+++ b/docs/IE_PLUGIN_DG/ExecutableNetwork.md
@@ -49,20 +49,15 @@ The function accepts a const shared pointer to `ngraph::Function` object and per
 
 This constructor creates a backend specific graph by importing from a stream object:
 
-> **NOTE**: The export of backend specific graph is done in the `ExportImpl` method, and data formats must be the same for both import and export.
+> **NOTE**: The export of backend specific graph is done in the `Export` method, and data formats must be the same for both import and export.
 
 @snippet src/template_executable_network.cpp executable_network:ctor_import_stream
 
-### `ExportImpl()`
-
-**Implementation details:**   
-Base InferenceEngine::ExecutableNetworkThreadSafeDefault class implements the public InferenceEngine::ExecutableNetworkThreadSafeDefault::Export method as following:
-- Writes `_plugin->GetName()` to the `model` stream.
-- Calls the `ExportImpl` method defined in a derived class to dump a backend specific graph.
+### `Export()`
 
 The implementation of the method should write all data to the `model` stream, which is required to import a backend specific graph later in the `Plugin::Import` method:
 
-@snippet src/template_executable_network.cpp executable_network:export_impl
+@snippet src/template_executable_network.cpp executable_network:export
 
 ### `CreateInferRequest()`
 
diff --git a/docs/IE_PLUGIN_DG/Plugin.md b/docs/IE_PLUGIN_DG/Plugin.md
index cadc8660fd3c30..6003eb691fc059 100644
--- a/docs/IE_PLUGIN_DG/Plugin.md
+++ b/docs/IE_PLUGIN_DG/Plugin.md
@@ -159,21 +159,13 @@ The snippet below provides an example of the implementation for `GetMetric`:
 
 > **NOTE**: If an unsupported metric key is passed to the function, it must throw an exception.
 
-### `ImportNetworkImpl()`
+### `ImportNetwork()`
 
 The importing network mechanism allows to import a previously exported backend specific graph and wrap it 
 using an [ExecutableNetwork](@ref executable_network) object. This functionality is useful if 
 backend specific graph compilation takes significant time and/or cannot be done on a target host 
 device due to other reasons.
 
-**Implementation details:** The base plugin class InferenceEngine::IInferencePlugin implements InferenceEngine::IInferencePlugin::ImportNetwork 
-as follows: exports a device type (InferenceEngine::IInferencePlugin::_pluginName) and then calls `ImportNetworkImpl`, 
-which is implemented in a derived class. 
-If a plugin cannot use the base implementation InferenceEngine::IInferencePlugin::ImportNetwork, it can override base 
-implementation and define an output blob structure up to its needs. This 
-can be useful if a plugin exports a blob in a special format for integration with other frameworks 
-where a common Inference Engine header from a base class implementation is not appropriate. 
-
 During export of backend specific graph using `ExecutableNetwork::Export`, a plugin may export any 
 type of information it needs to import a compiled graph properly and check its correctness. 
 For example, the export information may include:
diff --git a/docs/template_plugin/src/template_executable_network.cpp b/docs/template_plugin/src/template_executable_network.cpp
index e46bd63e5a0faa..4aba4622e50428 100644
--- a/docs/template_plugin/src/template_executable_network.cpp
+++ b/docs/template_plugin/src/template_executable_network.cpp
@@ -175,9 +175,9 @@ InferenceEngine::Parameter TemplatePlugin::ExecutableNetwork::GetMetric(const st
 }
 // ! [executable_network:get_metric]
 
-// ! [executable_network:export_impl]
-void TemplatePlugin::ExecutableNetwork::ExportImpl(std::ostream& modelStream) {
-    OV_ITT_SCOPED_TASK(itt::domains::TemplatePlugin, "ExecutableNetwork::ExportImpl");
+// ! [executable_network:export]
+void TemplatePlugin::ExecutableNetwork::Export(std::ostream& modelStream) {
+    OV_ITT_SCOPED_TASK(itt::domains::TemplatePlugin, "ExecutableNetwork::Export");
 
     // Note: custom ngraph extensions are not supported
     std::map<std::string, ngraph::OpSet> custom_opsets;
@@ -198,4 +198,4 @@ void TemplatePlugin::ExecutableNetwork::ExportImpl(std::ostream& modelStream) {
 
     // TODO: implement network precision, layout, preprocessing info serialization
 }
-// ! [executable_network:export_impl]
+// ! [executable_network:export]
diff --git a/docs/template_plugin/src/template_executable_network.hpp b/docs/template_plugin/src/template_executable_network.hpp
index ca3bca11ba847f..a68df02f958934 100644
--- a/docs/template_plugin/src/template_executable_network.hpp
+++ b/docs/template_plugin/src/template_executable_network.hpp
@@ -30,7 +30,7 @@ class ExecutableNetwork : public InferenceEngine::ExecutableNetworkThreadSafeDef
 
     // Methods from a base class ExecutableNetworkThreadSafeDefault
 
-    void ExportImpl(std::ostream& model) override;
+    void Export(std::ostream& model) override;
     InferenceEngine::IInferRequestInternal::Ptr CreateInferRequestImpl(InferenceEngine::InputsDataMap networkInputs,
                                                                        InferenceEngine::OutputsDataMap networkOutputs) override;
     InferenceEngine::IInferRequestInternal::Ptr CreateInferRequest() override;
diff --git a/docs/template_plugin/src/template_plugin.cpp b/docs/template_plugin/src/template_plugin.cpp
index 87a509c8a77bee..a0f7a30ee171cf 100644
--- a/docs/template_plugin/src/template_plugin.cpp
+++ b/docs/template_plugin/src/template_plugin.cpp
@@ -95,14 +95,14 @@ InferenceEngine::IExecutableNetworkInternal::Ptr Plugin::LoadExeNetworkImpl(cons
 }
 // ! [plugin:load_exe_network_impl]
 
-// ! [plugin:import_network_impl]
-InferenceEngine::IExecutableNetworkInternal::Ptr Plugin::ImportNetworkImpl(std::istream& modelStream, const std::map<std::string, std::string>& config) {
-    OV_ITT_SCOPED_TASK(itt::domains::TemplatePlugin, "Plugin::ImportNetworkImpl");
+// ! [plugin:import_network]
+InferenceEngine::IExecutableNetworkInternal::Ptr Plugin::ImportNetwork(std::istream& modelStream, const std::map<std::string, std::string>& config) {
+    OV_ITT_SCOPED_TASK(itt::domains::TemplatePlugin, "Plugin::ImportNetwork");
 
     auto fullConfig = Configuration {config, _cfg};
     return std::make_shared<ExecutableNetwork>(modelStream, fullConfig, std::static_pointer_cast<Plugin>(shared_from_this()));
 }
-// ! [plugin:import_network_impl]
+// ! [plugin:import_network]
 
 // ! [plugin:query_network]
 InferenceEngine::QueryNetworkResult Plugin::QueryNetwork(const InferenceEngine::CNNNetwork& network, const ConfigMap& config) const {
diff --git a/docs/template_plugin/src/template_plugin.hpp b/docs/template_plugin/src/template_plugin.hpp
index ef2b506d497500..71c37410ea717e 100644
--- a/docs/template_plugin/src/template_plugin.hpp
+++ b/docs/template_plugin/src/template_plugin.hpp
@@ -28,7 +28,7 @@ class Plugin : public InferenceEngine::IInferencePlugin {
     void AddExtension(const std::shared_ptr<InferenceEngine::IExtension>& extension) override;
     InferenceEngine::Parameter GetConfig(const std::string& name, const std::map<std::string, InferenceEngine::Parameter>& options) const override;
     InferenceEngine::Parameter GetMetric(const std::string& name, const std::map<std::string, InferenceEngine::Parameter>& options) const override;
-    InferenceEngine::IExecutableNetworkInternal::Ptr ImportNetworkImpl(std::istream& model, const std::map<std::string, std::string>& config) override;
+    InferenceEngine::IExecutableNetworkInternal::Ptr ImportNetwork(std::istream& model, const std::map<std::string, std::string>& config) override;
 
 private:
     friend class ExecutableNetwork;
diff --git a/inference-engine/include/ie_core.hpp b/inference-engine/include/ie_core.hpp
index e87f8c65719085..96f8d6b58af0e0 100644
--- a/inference-engine/include/ie_core.hpp
+++ b/inference-engine/include/ie_core.hpp
@@ -174,9 +174,18 @@ class INFERENCE_ENGINE_API_CLASS(Core) {
      * operation*
      * @return An executable network reference
      */
-    ExecutableNetwork ImportNetwork(std::istream& networkModel, const std::string& deviceName = {},
+    ExecutableNetwork ImportNetwork(std::istream& networkModel, const std::string& deviceName,
                                     const std::map<std::string, std::string>& config = {});
 
+    /**
+     * @deprecated Use Core::ImportNetwork with explicit device name
+     * @brief Creates an executable network from a previously exported network
+     * @param networkModel network model stream
+     * @return An executable network reference
+     */
+    INFERENCE_ENGINE_DEPRECATED("Use Core::ImportNetwork with explicit device name")
+    ExecutableNetwork ImportNetwork(std::istream& networkModel);
+
     /**
      * @brief Creates an executable network from a previously exported network within a specified
      * remote context.
diff --git a/inference-engine/src/hetero_plugin/hetero_executable_network.cpp b/inference-engine/src/hetero_plugin/hetero_executable_network.cpp
index 58fb35111affab..9f0135aa25e357 100644
--- a/inference-engine/src/hetero_plugin/hetero_executable_network.cpp
+++ b/inference-engine/src/hetero_plugin/hetero_executable_network.cpp
@@ -550,7 +550,7 @@ HeteroExecutableNetwork::HeteroExecutableNetwork(std::istream&
     this->SetPointerToPlugin(_heteroPlugin->shared_from_this());
 }
 
-void HeteroExecutableNetwork::ExportImpl(std::ostream& heteroModel) {
+void HeteroExecutableNetwork::Export(std::ostream& heteroModel) {
     pugi::xml_document doc;
     auto heteroNode = doc.append_child("hetero");
     heteroNode.append_attribute("name").set_value(_name.c_str());
diff --git a/inference-engine/src/hetero_plugin/hetero_executable_network.hpp b/inference-engine/src/hetero_plugin/hetero_executable_network.hpp
index 85fc8d9c19c9cd..59574ca2ce7a5f 100644
--- a/inference-engine/src/hetero_plugin/hetero_executable_network.hpp
+++ b/inference-engine/src/hetero_plugin/hetero_executable_network.hpp
@@ -56,7 +56,7 @@ class HeteroExecutableNetwork : public InferenceEngine::ExecutableNetworkThreadS
 
     InferenceEngine::Parameter GetMetric(const std::string &name) const override;
 
-    void ExportImpl(std::ostream& modelFile) override;
+    void Export(std::ostream& modelFile) override;
 
 private:
     void InitCNNImpl(const InferenceEngine::CNNNetwork&    network);
diff --git a/inference-engine/src/hetero_plugin/hetero_plugin.cpp b/inference-engine/src/hetero_plugin/hetero_plugin.cpp
index 1d8647716af8bf..09986b1e48eca8 100644
--- a/inference-engine/src/hetero_plugin/hetero_plugin.cpp
+++ b/inference-engine/src/hetero_plugin/hetero_plugin.cpp
@@ -57,13 +57,8 @@ InferenceEngine::IExecutableNetworkInternal::Ptr Engine::LoadExeNetworkImpl(cons
     return std::make_shared<HeteroExecutableNetwork>(network, mergeConfigs(_config, config), this);
 }
 
-InferenceEngine::IExecutableNetworkInternal::Ptr Engine::ImportNetworkImpl(std::istream& heteroModel, const Configs& config) {
-    if (GetCore() == nullptr) {
-        IE_THROW() << "Please, work with HETERO device via InferencEngine::Core object";
-    }
-
-    return std::make_shared<HeteroExecutableNetwork>(heteroModel,
-        mergeConfigs(_config, config), this);
+InferenceEngine::IExecutableNetworkInternal::Ptr Engine::ImportNetwork(std::istream& heteroModel, const std::map<std::string, std::string>& config) {
+    return std::make_shared<HeteroExecutableNetwork>(heteroModel, mergeConfigs(_config, config), this);
 }
 
 Engine::Configs Engine::GetSupportedConfig(const Engine::Configs& config, const std::string & deviceName) const {
diff --git a/inference-engine/src/hetero_plugin/hetero_plugin.hpp b/inference-engine/src/hetero_plugin/hetero_plugin.hpp
index 2b5a93b829b254..fbc602116d109e 100644
--- a/inference-engine/src/hetero_plugin/hetero_plugin.hpp
+++ b/inference-engine/src/hetero_plugin/hetero_plugin.hpp
@@ -37,10 +37,11 @@ class Engine : public InferenceEngine::IInferencePlugin {
     InferenceEngine::Parameter GetConfig(const std::string& name, const std::map<std::string,
                                          InferenceEngine::Parameter> & options) const override;
 
-    InferenceEngine::IExecutableNetworkInternal::Ptr ImportNetworkImpl(std::istream& heteroModel, const Configs& config) override;
+    InferenceEngine::IExecutableNetworkInternal::Ptr
+    ImportNetwork(std::istream& heteroModel, const std::map<std::string, std::string>& config) override;
 
     DeviceMetaInformationMap GetDevicePlugins(const std::string& targetFallback,
-        const Configs & localConfig) const;
+                                              const Configs & localConfig) const;
 
 private:
     Configs GetSupportedConfig(const Configs& config, const std::string & deviceName) const;
diff --git a/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_iexecutable_network_internal.cpp b/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_iexecutable_network_internal.cpp
index bf3086551c15fa..6b5bb34c97074e 100644
--- a/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_iexecutable_network_internal.cpp
+++ b/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_iexecutable_network_internal.cpp
@@ -49,19 +49,17 @@ std::shared_ptr<IInferRequestInternal> IExecutableNetworkInternal::CreateInferRe
 }
 
 void IExecutableNetworkInternal::Export(const std::string& modelFileName) {
-    // we need to write to stringstream first
-    // because in case of exception in ExportImpl the file is not created
-    std::stringstream strm;
-    ExportImpl(strm);
-    std::ofstream(modelFileName.c_str()) << strm.rdbuf();
+    std::ofstream modelFile(modelFileName, std::ios::out | std::ios::binary);
+
+    if (modelFile.is_open()) {
+        Export(modelFile);
+    } else {
+        IE_THROW() << "The " << modelFileName << " file can not be opened for Export";
+    }
 }
 
 void IExecutableNetworkInternal::Export(std::ostream& networkModel) {
-    std::stringstream strm;
-    strm.write(exportMagic.data(), exportMagic.size());
-    strm << _plugin->GetName() << std::endl;
-    ExportImpl(strm);
-    networkModel << strm.rdbuf();
+    IE_THROW(NotImplemented);
 }
 
 CNNNetwork IExecutableNetworkInternal::GetExecGraphInfo() {
@@ -97,7 +95,4 @@ std::shared_ptr<IInferRequestInternal> IExecutableNetworkInternal::CreateInferRe
     IE_THROW(NotImplemented);
 }
 
-void IExecutableNetworkInternal::ExportImpl(std::ostream&) {
-    IE_THROW(NotImplemented);
-}
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_iplugin_internal.cpp b/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_iplugin_internal.cpp
index 5637701754e3f5..88599aa78b37d2 100644
--- a/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_iplugin_internal.cpp
+++ b/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_iplugin_internal.cpp
@@ -16,24 +16,12 @@
 #include <blob_factory.hpp>
 
 #include <istream>
+#include <fstream>
 #include <map>
 #include <memory>
 #include <string>
 
 namespace InferenceEngine {
-namespace {
-void parsePluginName(std::istream& networkModel) {
-    ExportMagic magic = {};
-    auto currentPos = networkModel.tellg();
-    networkModel.read(magic.data(), magic.size());
-    auto exportedWithName = (exportMagic == magic);
-    if (exportedWithName) {
-        networkModel.ignore(std::numeric_limits<std::streamsize>::max(), '\n');
-    } else {
-        networkModel.seekg(currentPos, networkModel.beg);
-    }
-}
-}  // namespace
 
 PreProcessInfo copyPreProcess(const PreProcessInfo& from) {
     PreProcessInfo to = from;
@@ -170,22 +158,26 @@ RemoteContext::Ptr IInferencePlugin::GetDefaultContext(const ParamMap&) {
     IE_THROW(NotImplemented);
 }
 
-std::shared_ptr<IExecutableNetworkInternal> IInferencePlugin::ImportNetwork(const std::string&,
-                                                                            const std::map<std::string, std::string>&) {
-    IE_THROW(NotImplemented);
+std::shared_ptr<IExecutableNetworkInternal> IInferencePlugin::ImportNetwork(const std::string& modelFileName,
+                                                                            const std::map<std::string, std::string>& config) {
+    std::ifstream blobFile(modelFileName, std::ios::binary);
+
+    if (!blobFile.is_open()) {
+        IE_THROW(NetworkNotRead);
+    }
+
+    return ImportNetwork(blobFile, config);
 }
 
 std::shared_ptr<IExecutableNetworkInternal> IInferencePlugin::ImportNetwork(std::istream& networkModel,
                                                                             const std::map<std::string, std::string>& config) {
-    parsePluginName(networkModel);
-    return ImportNetworkImpl(networkModel, config);
+    IE_THROW(NotImplemented);
 }
 
 std::shared_ptr<IExecutableNetworkInternal> IInferencePlugin::ImportNetwork(std::istream& networkModel,
                                                                             const std::shared_ptr<RemoteContext>& context,
                                                                             const std::map<std::string, std::string>& config) {
-    parsePluginName(networkModel);
-    return ImportNetworkImpl(networkModel, context, config);
+   IE_THROW(NotImplemented);
 }
 
 void IInferencePlugin::SetCore(ICore* core) {
@@ -213,17 +205,6 @@ std::shared_ptr<IExecutableNetworkInternal> IInferencePlugin::LoadExeNetworkImpl
     IE_THROW(NotImplemented);
 }
 
-std::shared_ptr<IExecutableNetworkInternal> IInferencePlugin::ImportNetworkImpl(std::istream&,
-                                                                                const std::map<std::string, std::string>&) {
-    IE_THROW(NotImplemented);
-}
-
-std::shared_ptr<IExecutableNetworkInternal> IInferencePlugin::ImportNetworkImpl(std::istream&,
-                                                                                const std::shared_ptr<RemoteContext>&,
-                                                                                const std::map<std::string, std::string>&) {
-    IE_THROW(NotImplemented);
-}
-
 void IInferencePlugin::SetExeNetworkInfo(const std::shared_ptr<IExecutableNetworkInternal>& exeNetwork,
                                          const ConstInputsDataMap& inputs,
                                          const ConstOutputsDataMap& outputs) {
diff --git a/inference-engine/src/inference_engine/ie_core.cpp b/inference-engine/src/inference_engine/ie_core.cpp
index 28563a29b62d30..63814215037f16 100644
--- a/inference-engine/src/inference_engine/ie_core.cpp
+++ b/inference-engine/src/inference_engine/ie_core.cpp
@@ -395,6 +395,7 @@ class Core::Impl : public ICore {
         opsetNames.insert("opset4");
         opsetNames.insert("opset5");
         opsetNames.insert("opset6");
+        opsetNames.insert("opset7");
     }
 
     ~Impl() override = default;
@@ -566,18 +567,6 @@ class Core::Impl : public ICore {
     SoExecutableNetworkInternal ImportNetwork(std::istream& networkModel, const std::string& deviceName,
                                               const std::map<std::string, std::string>& config) override {
         auto parsed = parseDeviceNameIntoConfig(deviceName, config);
-
-        if (parsed._deviceName.empty()) {
-            ExportMagic magic = {};
-            auto currentPos = networkModel.tellg();
-            networkModel.read(magic.data(), magic.size());
-            auto exportedWithName = (exportMagic == magic);
-            if (exportedWithName) {
-                std::getline(networkModel, parsed._deviceName);
-            }
-            networkModel.seekg(currentPos, networkModel.beg);
-        }
-
         return GetCPPPluginByName(parsed._deviceName).ImportNetwork(networkModel, parsed._config);
     }
 
@@ -1022,18 +1011,6 @@ void Core::AddExtension(const IExtensionPtr& extension) {
 ExecutableNetwork Core::ImportNetwork(const std::string& modelFileName, const std::string& deviceName,
                                       const std::map<std::string, std::string>& config) {
     OV_ITT_SCOPED_TASK(itt::domains::IE, "Core::ImportNetwork");
-
-    // TODO: remove once NotImplemented exception is deprecated and not used
-    if (deviceName.find("HETERO") == 0) {
-        IE_THROW() << "HETERO device does not support ImportNetwork";
-    }
-    if (deviceName.find("MULTI") == 0) {
-        IE_THROW() << "MULTI device does not support ImportNetwork";
-    }
-    if (deviceName.find("AUTO") == 0) {
-        IE_THROW() << "AUTO device does not support ImportNetwork";
-    }
-
     auto parsed = parseDeviceNameIntoConfig(deviceName, config);
     auto exec = _impl->GetCPPPluginByName(parsed._deviceName).ImportNetwork(modelFileName, parsed._config);
     return { exec, exec };
@@ -1041,10 +1018,33 @@ ExecutableNetwork Core::ImportNetwork(const std::string& modelFileName, const st
 
 ExecutableNetwork Core::ImportNetwork(std::istream& networkModel, const std::string& deviceName,
                                       const std::map<std::string, std::string>& config) {
+    OV_ITT_SCOPED_TASK(itt::domains::IE, "Core::ImportNetwork");
     auto exec = _impl->ImportNetwork(networkModel, deviceName, config);
     return { exec, exec };
 }
 
+ExecutableNetwork Core::ImportNetwork(std::istream& networkModel) {
+    OV_ITT_SCOPED_TASK(itt::domains::IE, "Core::ImportNetwork");
+
+    using ExportMagic = std::array<char, 4>;
+    constexpr static const ExportMagic exportMagic = {{0x1, 0xE, 0xE, 0x1}};
+
+    std::string deviceName;
+    ExportMagic magic = {};
+    auto currentPos = networkModel.tellg();
+    networkModel.read(magic.data(), magic.size());
+    if (exportMagic == magic) {
+        std::getline(networkModel, deviceName);
+    } else {
+        IE_THROW() << "Passed compiled stream does not contain device name. "
+            "Please, provide device name manually";
+    }
+    networkModel.seekg(currentPos, networkModel.beg);
+
+    auto exec = _impl->GetCPPPluginByName(deviceName).ImportNetwork(networkModel, {});
+    return { exec, exec };
+}
+
 ExecutableNetwork Core::ImportNetwork(std::istream& networkModel,
                                       const RemoteContext::Ptr& context,
                                       const std::map<std::string, std::string>& config) {
@@ -1124,8 +1124,8 @@ Parameter Core::GetConfig(const std::string& deviceName, const std::string& name
             IE_THROW()
                 << "You can only GetConfig of the AUTO itself (without devices). "
                    "GetConfig is also possible for the individual devices before creating the AUTO on top.";
-      }
-  }
+        }
+    }
 
     auto parsed = parseDeviceNameIntoConfig(deviceName);
 
diff --git a/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_iexecutable_network_internal.hpp b/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_iexecutable_network_internal.hpp
index 589514103835c8..1f3eb681e4b897 100644
--- a/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_iexecutable_network_internal.hpp
+++ b/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_iexecutable_network_internal.hpp
@@ -140,14 +140,6 @@ class INFERENCE_ENGINE_API_CLASS(IExecutableNetworkInternal) : public std::enabl
     virtual std::shared_ptr<IInferRequestInternal> CreateInferRequestImpl(InputsDataMap networkInputs,
                                                                           OutputsDataMap networkOutputs);
 
-    /**
-     * @brief Exports an internal hardware-dependent model to a stream.
-     * @note The function is called from IExecutableNetworkInternal::Export(std::ostream&),
-     * which performs common export first and calls this plugin-dependent implementation after.
-     * @param networkModel A stream to export network to.
-     */
-    virtual void ExportImpl(std::ostream& networkModel);
-
     InferenceEngine::InputsDataMap _networkInputs;  //!< Holds information about network inputs info
     InferenceEngine::OutputsDataMap _networkOutputs;  //!< Holds information about network outputs data
 
diff --git a/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_iplugin_internal.hpp b/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_iplugin_internal.hpp
index 22c2f7e1c08780..56e3e1cf5f7dc6 100644
--- a/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_iplugin_internal.hpp
+++ b/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_iplugin_internal.hpp
@@ -286,29 +286,12 @@ class INFERENCE_ENGINE_API_CLASS(IInferencePlugin) : public std::enable_shared_f
                                                                            const std::map<std::string, std::string>& config);
 
     /**
-     * @brief Creates an executable network from an previously exported network
-     * @note The function is called from
-     * IInferencePlugin::ImportNetwork(std::istream&, const RemoteContext::Ptr&, const std::map<std::string, std::string>&)
-     * performs common steps first and calls this plugin-dependent implementation after.
-     * @param networkModel Reference to network model output stream
-     * @param config A string -> string map of parameters
-     * @return An Executable network
-     */
-    virtual std::shared_ptr<IExecutableNetworkInternal> ImportNetworkImpl(std::istream& networkModel,
-                                                                          const std::map<std::string, std::string>& config);
-
-    /**
-     * @brief Imports network wit RemoteContext
-     * @param networkModel Reference to network model output stream
-     * @param context - a pointer to plugin context derived from RemoteContext class used to
-     *        execute the network
-     * @param config A string -> string map of parameters
-     * @return An Executable network
+     * @brief Set input and output information to executable network. This method is used to
+     * set addtional information to InferenceEngine::IExecutableNetworkInternal create by device plugin.
+     * @param exeNetwork An executable network object to set information to
+     * @param inputs An input information to set
+     * @param outputs An output information to set
      */
-    virtual std::shared_ptr<IExecutableNetworkInternal> ImportNetworkImpl(std::istream& networkModel,
-                                                                          const std::shared_ptr<RemoteContext>& context,
-                                                                          const std::map<std::string, std::string>& config);
-
     void SetExeNetworkInfo(const std::shared_ptr<IExecutableNetworkInternal>& exeNetwork,
                            const ConstInputsDataMap& inputs,
                            const ConstOutputsDataMap& outputs);
diff --git a/inference-engine/src/plugin_api/ie_icore.hpp b/inference-engine/src/plugin_api/ie_icore.hpp
index 70d7aaff3a5e25..fb4ac0b3423ce6 100644
--- a/inference-engine/src/plugin_api/ie_icore.hpp
+++ b/inference-engine/src/plugin_api/ie_icore.hpp
@@ -141,18 +141,6 @@ class ICore {
     virtual ~ICore() = default;
 };
 
-/**
- * @brief Type of magic value
- * @ingroup ie_dev_api_plugin_api
- */
-using ExportMagic = std::array<char, 4>;
-
-/**
- * @brief Magic number used by ie core to identify exported network with plugin name
- * @ingroup ie_dev_api_plugin_api
- */
-constexpr static const ExportMagic exportMagic = {{0x1, 0xE, 0xE, 0x1}};
-
 /**
  * @private
  */
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.h b/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.h
index 8a6c42c4e97012..22824ee5ec1c47 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.h
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.h
@@ -93,16 +93,6 @@ class ExecutableNetwork : public ie::ExecutableNetworkThreadSafeDefault {
         model.write(_graphBlob.data(), _graphBlob.size());
     }
 
-    void Export(const std::string &modelFileName) override {
-        std::ofstream modelFile(modelFileName, std::ios::out | std::ios::binary);
-
-        if (modelFile.is_open()) {
-            Export(modelFile);
-        } else {
-            IE_THROW() << "The " << modelFileName << " file can not be opened for export";
-        }
-    }
-
     ie::Parameter GetMetric(const std::string &name) const override;
 
     ie::CNNNetwork GetExecGraphInfo() override;
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_plugin.cpp b/inference-engine/src/vpu/myriad_plugin/myriad_plugin.cpp
index f61b9fbf7fb9cb..75e7ef395d9266 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_plugin.cpp
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_plugin.cpp
@@ -151,20 +151,6 @@ InferenceEngine::IExecutableNetworkInternal::Ptr Engine::ImportNetwork(
     return executableNetwork;
 }
 
-InferenceEngine::IExecutableNetworkInternal::Ptr Engine::ImportNetwork(
-        const std::string& modelFileName,
-        const std::map<std::string, std::string>& config) {
-    VPU_PROFILE(ImportNetwork);
-
-    std::ifstream blobFile(modelFileName, std::ios::binary);
-
-    if (!blobFile.is_open()) {
-        IE_THROW(NetworkNotRead);
-    }
-
-    return ImportNetwork(blobFile, config);
-}
-
 InferenceEngine::Parameter Engine::GetMetric(const std::string& name,
                                      const std::map<std::string, InferenceEngine::Parameter> & options) const {
     const auto mvnc = _mvnc;
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_plugin.h b/inference-engine/src/vpu/myriad_plugin/myriad_plugin.h
index 07349f637e27ad..9fb074b5ac174c 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_plugin.h
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_plugin.h
@@ -37,10 +37,6 @@ class Engine : public ie::IInferencePlugin {
 
     using ie::IInferencePlugin::ImportNetwork;
 
-    ie::IExecutableNetworkInternal::Ptr ImportNetwork(
-            const std::string& modelFileName,
-            const std::map<std::string, std::string>& config) override;
-
     ie::IExecutableNetworkInternal::Ptr ImportNetwork(
             std::istream& model,
             const std::map<std::string, std::string>& config) override;
diff --git a/inference-engine/tests/functional/inference_engine/caching_test.cpp b/inference-engine/tests/functional/inference_engine/caching_test.cpp
index dd19dd3815d70c..cad8bd4428e3ba 100644
--- a/inference-engine/tests/functional/inference_engine/caching_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/caching_test.cpp
@@ -43,17 +43,18 @@ enum class TestLoadType {
     EContext,
     EModelName
 };
+
 using TestParam = std::tuple<TestLoadType, std::string, bool>;
 
 //  GCC4.8 limitation: have to specify type of each element in list
 static const std::vector<TestParam> loadVariants = {
-        TestParam { TestLoadType::ECNN, std::string("ByCNNNetwork"), false },
-        TestParam { TestLoadType::EContext, std::string("ByRemoteContext"), true },
-        TestParam { TestLoadType::EModelName, std::string("ByModelName"), false },
+    TestParam { TestLoadType::ECNN, std::string("ByCNNNetwork"), false },
+    TestParam { TestLoadType::EContext, std::string("ByRemoteContext"), true },
+    TestParam { TestLoadType::EModelName, std::string("ByModelName"), false },
 };
 
 static const std::vector<std::string> cacheFolders {
-        std::string("testCache"),
+    std::string("testCache"),
 };
 
 std::string getTestCaseName(const testing::TestParamInfo<std::tuple<TestParam, std::string>> &obj) {
@@ -100,12 +101,12 @@ class MockCachingInferencePlugin : public MockCachingInferencePluginBase {
 
     MOCK_CONST_METHOD0(OnLoadNetworkFromFile, void(void));
 
-    MOCK_METHOD2(ImportNetworkImpl, std::shared_ptr<IExecutableNetworkInternal>(std::istream& networkModel,
-                                                                                const std::map<std::string, std::string>& config));
+    MOCK_METHOD2(ImportNetwork, IExecutableNetworkInternal::Ptr(std::istream& networkModel,
+                                                               const std::map<std::string, std::string>& config));
 
-    MOCK_METHOD3(ImportNetworkImpl, std::shared_ptr<IExecutableNetworkInternal>(std::istream& networkModel,
-                                                                                const RemoteContext::Ptr& context,
-                                                                                const std::map<std::string, std::string>& config));
+    MOCK_METHOD3(ImportNetwork, IExecutableNetworkInternal::Ptr(std::istream& networkModel,
+                                                                const RemoteContext::Ptr& context,
+                                                                const std::map<std::string, std::string>& config));
 
     MOCK_CONST_METHOD2(QueryNetwork, QueryNetworkResult(const CNNNetwork& network,
                                                         const std::map<std::string, std::string>& config));
@@ -120,7 +121,7 @@ class MockExecutableNetwork : public IExecutableNetworkInternal {
 
 public:
     MockExecutableNetwork() {}
-    MOCK_METHOD1(ExportImpl, void(std::ostream& networkModel));
+    MOCK_METHOD1(Export, void(std::ostream& networkModel));
     MOCK_METHOD0(CreateInferRequest, IInferRequestInternal::Ptr());
     MOCK_CONST_METHOD0(GetInputsInfo, ConstInputsDataMap());
     MOCK_CONST_METHOD0(GetOutputsInfo, ConstOutputsDataMap());
@@ -130,10 +131,10 @@ class MockExecutableNetwork : public IExecutableNetworkInternal {
     MOCK_METHOD1(setNetworkInputs, void(const InputsDataMap& networkInputs));
     MOCK_METHOD1(setNetworkOutputs, void(const OutputsDataMap& networkOutputs));
 
-    void Export(std::ostream& networkModel) override {
-        std::lock_guard<std::mutex> guard(m_pluginMutex);
-        IExecutableNetworkInternal::Export(networkModel);
-    }
+    // void Export(std::ostream& networkModel) override {
+    //     std::lock_guard<std::mutex> guard(m_pluginMutex);
+    //     IExecutableNetworkInternal::Export(networkModel);
+    // }
 
     void SetPointerToPlugin(const IInferencePlugin::Ptr& plugin) override {
         std::lock_guard<std::mutex> guard(m_pluginMutex);
@@ -323,13 +324,13 @@ class CachingTest : public ::testing::TestWithParam<std::tuple<TestParam, std::s
         ON_CALL(plugin, GetMetric(METRIC_KEY(DEVICE_ARCHITECTURE), _)).
                 WillByDefault(Return("mock"));
 
-        ON_CALL(plugin, ImportNetworkImpl(_, _, _)).
+        ON_CALL(plugin, ImportNetwork(_, _, _)).
                 WillByDefault(Invoke([&](std::istream &istr, RemoteContext::Ptr,
                                          const std::map<std::string, std::string> &) {
             return createMockIExecutableNet();
         }));
 
-        ON_CALL(plugin, ImportNetworkImpl(_, _)).
+        ON_CALL(plugin, ImportNetwork(_, _)).
                 WillByDefault(Invoke([&](std::istream &istr, const std::map<std::string, std::string> &) {
             return createMockIExecutableNet();
         }));
@@ -403,9 +404,9 @@ TEST_P(CachingTest, TestLoad) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -415,9 +416,9 @@ TEST_P(CachingTest, TestLoad) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(!m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -426,43 +427,43 @@ TEST_P(CachingTest, TestLoad) {
 }
 
 TEST_P(CachingTest, TestLoadCustomImportExport) {
-    const int customNumber = 1234;
+    const char customData[] = {1, 2, 3, 4, 5};
     EXPECT_CALL(*mockPlugin, GetMetric(METRIC_KEY(SUPPORTED_METRICS), _)).Times(AnyNumber());
     EXPECT_CALL(*mockPlugin, GetMetric(METRIC_KEY(IMPORT_EXPORT_SUPPORT), _)).Times(AnyNumber());
     EXPECT_CALL(*mockPlugin, GetMetric(METRIC_KEY(DEVICE_ARCHITECTURE), _)).Times(AnyNumber());
-    ON_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).
+    ON_CALL(*mockPlugin, ImportNetwork(_, _, _)).
             WillByDefault(Invoke([&](std::istream& s, RemoteContext::Ptr,
                                      const std::map<std::string, std::string> &) {
-        int a;
-        s >> a;
-        EXPECT_EQ(customNumber, a);
+        char a[sizeof(customData)];
+        s.read(a, sizeof(customData));
+        EXPECT_EQ(memcmp(a, customData, sizeof(customData)), 0);
         auto mock = std::make_shared<MockExecutableNetwork>();
         EXPECT_CALL(*mock, GetInputsInfo()).Times(AnyNumber()).WillRepeatedly(Return(ConstInputsDataMap{}));
         EXPECT_CALL(*mock, GetOutputsInfo()).Times(AnyNumber()).WillRepeatedly(Return(ConstOutputsDataMap{}));
         return mock;
     }));
 
-    ON_CALL(*mockPlugin, ImportNetworkImpl(_, _)).
+    ON_CALL(*mockPlugin, ImportNetwork(_, _)).
             WillByDefault(Invoke([&](std::istream &s, const std::map<std::string, std::string> &) {
-        int a;
-        s >> a;
-        EXPECT_EQ(customNumber, a);
+        char a[sizeof(customData)];
+        s.read(a, sizeof(customData));
+        EXPECT_EQ(memcmp(a, customData, sizeof(customData)), 0);
         auto mock = std::make_shared<MockExecutableNetwork>();
         EXPECT_CALL(*mock, GetInputsInfo()).Times(AnyNumber()).WillRepeatedly(Return(ConstInputsDataMap{}));
         EXPECT_CALL(*mock, GetOutputsInfo()).Times(AnyNumber()).WillRepeatedly(Return(ConstOutputsDataMap{}));
         return mock;
     }));
 
-    ON_CALL(*net, ExportImpl(_)).WillByDefault(Invoke([&] (std::ostream& s) {
-        s << customNumber;
+    ON_CALL(*net, Export(_)).WillByDefault(Invoke([&] (std::ostream& s) {
+        s.write(customData, sizeof(customData));
     }));
 
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -472,9 +473,9 @@ TEST_P(CachingTest, TestLoadCustomImportExport) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(!m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -497,9 +498,9 @@ TEST_P(CachingTest, TestChangeLoadConfig) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunctionWithCfg(ie, {{CUSTOM_KEY, "0"}});
@@ -509,9 +510,9 @@ TEST_P(CachingTest, TestChangeLoadConfig) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunctionWithCfg(ie, {{CUSTOM_KEY, "1"}});
@@ -526,9 +527,9 @@ TEST_P(CachingTest, TestNoCacheEnabled) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             m_testFunction(ie);
         });
@@ -544,9 +545,9 @@ TEST_P(CachingTest, TestNoCacheSupported) {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, OnLoadNetworkFromFile()).Times(m_type == TestLoadType::EModelName ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -563,9 +564,9 @@ TEST_P(CachingTest, TestNoCacheMetricSupported) {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, OnLoadNetworkFromFile()).Times(m_type == TestLoadType::EModelName ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -652,8 +653,8 @@ TEST_P(CachingTest, TestNoCacheEnabled_cacheDirConfig) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
         testLoad([&](Core &ie) {
             m_testFunction(ie);
         });
@@ -667,9 +668,9 @@ TEST_P(CachingTest, TestLoadChangeCacheDir) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -681,9 +682,9 @@ TEST_P(CachingTest, TestLoadChangeCacheDir) {
         MkDirGuard dir(newCacheDir);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), newCacheDir}});
             m_testFunction(ie);
@@ -698,9 +699,9 @@ TEST_P(CachingTest, TestClearCacheDir) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), ""}});
@@ -716,9 +717,9 @@ TEST_P(CachingTest, TestChangeOtherConfig) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             ie.SetConfig({{"someKey", "someValue"}});
@@ -735,9 +736,9 @@ TEST_P(CachingTest, TestChangeCacheDirFailure) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -747,9 +748,9 @@ TEST_P(CachingTest, TestChangeCacheDirFailure) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(!m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             EXPECT_ANY_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir + "/" + longName}}));
@@ -769,9 +770,9 @@ TEST_P(CachingTest, TestCacheDirCreateRecursive) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             EXPECT_NO_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), newCacheDir3}}));
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -797,9 +798,9 @@ TEST_P(CachingTest, TestDeviceArchitecture) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             deviceToLoad = "mock.0";
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
@@ -810,9 +811,9 @@ TEST_P(CachingTest, TestDeviceArchitecture) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(!m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             deviceToLoad = "mock.1";
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
@@ -822,9 +823,9 @@ TEST_P(CachingTest, TestDeviceArchitecture) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             deviceToLoad = "mock.50";
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
@@ -835,9 +836,9 @@ TEST_P(CachingTest, TestDeviceArchitecture) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(!m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             deviceToLoad = "mock.51";
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
@@ -856,9 +857,9 @@ TEST_P(CachingTest, TestNoDeviceArchitecture) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             deviceToLoad = "mock.0";
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
@@ -869,9 +870,9 @@ TEST_P(CachingTest, TestNoDeviceArchitecture) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(!m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             deviceToLoad = "mock.50";
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
@@ -887,9 +888,9 @@ TEST_P(CachingTest, TestThrowOnExport) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1).WillOnce(Throw(1));
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1).WillOnce(Throw(1));
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             EXPECT_ANY_THROW(m_testFunction(ie));
@@ -906,9 +907,9 @@ TEST_P(CachingTest, TestThrowOnImport) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -918,13 +919,13 @@ TEST_P(CachingTest, TestThrowOnImport) {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
         if (m_remoteContext) {
-            EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(1).WillOnce(Throw(1));
-            EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
+            EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(1).WillOnce(Throw(1));
+            EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
         } else {
-            EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-            EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(1).WillOnce(Throw(1));
+            EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+            EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(1).WillOnce(Throw(1));
         }
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -933,9 +934,9 @@ TEST_P(CachingTest, TestThrowOnImport) {
     { // Step 3: same load, cache is re-created on export on step 2 and shall be successfully imported now
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(!m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -951,9 +952,9 @@ TEST_P(CachingTest, TestNetworkModified) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             EXPECT_NO_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}}));
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -974,9 +975,9 @@ TEST_P(CachingTest, TestNetworkModified) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             EXPECT_NO_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}}));
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -985,9 +986,9 @@ TEST_P(CachingTest, TestNetworkModified) {
     { // Step 3: same load, should be ok now
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(!m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             EXPECT_NO_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}}));
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -1003,9 +1004,9 @@ TEST_P(CachingTest, TestCacheFileCorrupted) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             EXPECT_NO_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}}));
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -1021,9 +1022,9 @@ TEST_P(CachingTest, TestCacheFileCorrupted) {
     { // Step 2. Cache is corrupted, will be silently removed
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             EXPECT_NO_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}}));
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -1032,9 +1033,9 @@ TEST_P(CachingTest, TestCacheFileCorrupted) {
     { // Step 3: same load, should be ok now due to re-creation of cache
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(!m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             EXPECT_NO_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}}));
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -1050,9 +1051,9 @@ TEST_P(CachingTest, TestCacheFileOldVersion) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             EXPECT_NO_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}}));
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -1083,9 +1084,9 @@ TEST_P(CachingTest, TestCacheFileOldVersion) {
     { // Step 2. Build number mismatch, cache will be silently removed
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             EXPECT_NO_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}}));
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -1094,9 +1095,9 @@ TEST_P(CachingTest, TestCacheFileOldVersion) {
     { // Step 3: same load, should be ok now due to re-creation of cache
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(!m_remoteContext ? 1 : 0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(!m_remoteContext ? 1 : 0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             EXPECT_NO_THROW(ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}}));
             EXPECT_NO_THROW(m_testFunction(ie));
@@ -1118,9 +1119,9 @@ TEST_P(CachingTest, LoadHetero_NoCacheMetric) {
     for (int i = 0; i < 2; i++) {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(1);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -1138,9 +1139,9 @@ TEST_P(CachingTest, LoadHetero_OneDevice) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(1);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -1152,9 +1153,9 @@ TEST_P(CachingTest, LoadHetero_OneDevice) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(1);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(1);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -1172,9 +1173,9 @@ TEST_P(CachingTest, LoadHetero_TargetFallbackFromCore) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(1);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             ie.SetConfig({{"TARGET_FALLBACK", "mock"}}, CommonTestUtils::DEVICE_HETERO);
@@ -1187,9 +1188,9 @@ TEST_P(CachingTest, LoadHetero_TargetFallbackFromCore) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(1);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(1);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             ie.SetConfig({{"TARGET_FALLBACK", "mock"}}, CommonTestUtils::DEVICE_HETERO);
@@ -1200,20 +1201,20 @@ TEST_P(CachingTest, LoadHetero_TargetFallbackFromCore) {
 
 TEST_P(CachingTest, LoadHetero_MultiArchs) {
     EXPECT_CALL(*mockPlugin, GetMetric(_, _)).Times(AnyNumber());
-    int customNumber = 1234;
-    ON_CALL(*mockPlugin, ImportNetworkImpl(_, _)).
+    const char customData[] = {1, 2, 3, 4, 5};
+    ON_CALL(*mockPlugin, ImportNetwork(_, _)).
             WillByDefault(Invoke([&](std::istream &s, const std::map<std::string, std::string> &) {
-        int a;
-        s >> a;
-        EXPECT_EQ(customNumber, a);
+        char a[sizeof(customData)];
+        s.read(a, sizeof(customData));
+        EXPECT_EQ(memcmp(a, customData, sizeof(customData)), 0);
         auto mock = std::make_shared<MockExecutableNetwork>();
         EXPECT_CALL(*mock, GetInputsInfo()).Times(AnyNumber()).WillRepeatedly(Return(ConstInputsDataMap{}));
         EXPECT_CALL(*mock, GetOutputsInfo()).Times(AnyNumber()).WillRepeatedly(Return(ConstOutputsDataMap{}));
         return mock;
     }));
 
-    ON_CALL(*net, ExportImpl(_)).WillByDefault(Invoke([&] (std::ostream& s) {
-        s << customNumber;
+    ON_CALL(*net, Export(_)).WillByDefault(Invoke([&] (std::ostream& s) {
+        s.write(customData, sizeof(customData));
     }));
     EXPECT_CALL(*mockPlugin, QueryNetwork(_, _)).Times(AnyNumber()).WillRepeatedly(
             Invoke([&](const CNNNetwork &network, const std::map<std::string, std::string> &config) {
@@ -1249,9 +1250,9 @@ TEST_P(CachingTest, LoadHetero_MultiArchs) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(AtLeast(2)); // for .1 and for .51
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(AtLeast(2)); // for .1 and for .51
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(AtLeast(2)); // for .1 and for .51
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -1264,9 +1265,9 @@ TEST_P(CachingTest, LoadHetero_MultiArchs) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(AtLeast(2)); // for .2 and for .52
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(AtLeast(2)); // for .2 and for .52
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -1276,9 +1277,9 @@ TEST_P(CachingTest, LoadHetero_MultiArchs) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(AtLeast(1));
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(AtLeast(1));
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(AtLeast(1));
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             m_testFunction(ie);
@@ -1305,9 +1306,9 @@ TEST_P(CachingTest, LoadHetero_MultiArchs_TargetFallback_FromCore) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(1);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             ie.SetConfig({{"TARGET_FALLBACK", "mock.1"}}, CommonTestUtils::DEVICE_HETERO);
@@ -1318,9 +1319,9 @@ TEST_P(CachingTest, LoadHetero_MultiArchs_TargetFallback_FromCore) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(1);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(1);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{"TARGET_FALLBACK", "mock.1"}}, CommonTestUtils::DEVICE_HETERO);
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
@@ -1330,9 +1331,9 @@ TEST_P(CachingTest, LoadHetero_MultiArchs_TargetFallback_FromCore) {
     {
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(1);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{"TARGET_FALLBACK", "mock.51"}}, CommonTestUtils::DEVICE_HETERO);
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
@@ -1366,9 +1367,9 @@ TEST_P(CachingTest, LoadMulti_race) {
 
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(1);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(devCount - 1);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(devCount - 1);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), cacheDir}});
             ASSERT_NO_THROW(m_testFunction(ie));
@@ -1394,9 +1395,9 @@ TEST_P(CachingTest, Load_threads) {
         MkDirGuard guard(cacheDir);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _, _)).Times(0);
         EXPECT_CALL(*mockPlugin, LoadExeNetworkImpl(_, _)).Times(1);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(THREADS_COUNT - 1);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(1);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(THREADS_COUNT - 1);
+        EXPECT_CALL(*net, Export(_)).Times(1);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), cacheDir}});
             std::vector<std::thread> threads;
@@ -1443,12 +1444,12 @@ TEST_P(CachingTest, LoadMulti_Archs) {
         // Load network from file shall not be called for plugins with caching supported
         EXPECT_CALL(*mockPlugin, OnLoadNetworkFromFile()).Times(0);
 
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(TEST_DEVICE_MAX_COUNT - 2)
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(TEST_DEVICE_MAX_COUNT - 2)
                 .WillRepeatedly(Invoke([&](std::istream &, const std::map<std::string, std::string> &) {
             return createMockIExecutableNet();
         }));
-        EXPECT_CALL(*net, ExportImpl(_)).Times(2);
+        EXPECT_CALL(*net, Export(_)).Times(2);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             ASSERT_NO_THROW(m_testFunction(ie));
@@ -1490,9 +1491,9 @@ TEST_P(CachingTest, LoadMulti_NoCachingOnDevice) {
         // Load network from file shall not be called by Multi plugin for devices with caching supported
         EXPECT_CALL(*mockPlugin, OnLoadNetworkFromFile()).Times(0);
 
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _, _)).Times(0);
-        EXPECT_CALL(*mockPlugin, ImportNetworkImpl(_, _)).Times(0);
-        EXPECT_CALL(*net, ExportImpl(_)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _, _)).Times(0);
+        EXPECT_CALL(*mockPlugin, ImportNetwork(_, _)).Times(0);
+        EXPECT_CALL(*net, Export(_)).Times(0);
         testLoad([&](Core &ie) {
             ie.SetConfig({{CONFIG_KEY(CACHE_DIR), m_cacheDir}});
             ExecutableNetwork exeNet;
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
index f013e544074644..3d4678bf5892b4 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
@@ -71,6 +71,8 @@ std::vector<std::string> disabledTestPatterns() {
         R"(.*smoke_SetBlobOfKindAUTO.*SetBlobOfKindTest.CompareWithRefs.*)",
         // reference doesn't cover I8, U8 cases. Issue: 55842
         R"(.*Gather7LayerTest.*netPRC=I8.*)",
+        // need to implement Export / Import
+        R"(.*IEClassImportExportTestP.*)"
     };
 #ifdef __APPLE__
         // TODO: Issue 55717
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/skip_tests_config.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/skip_tests_config.cpp
index 02889dd8ddef85..07bd2a260984aa 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/skip_tests_config.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/skip_tests_config.cpp
@@ -57,5 +57,7 @@ std::vector<std::string> disabledTestPatterns() {
             R"(.*LSTMSequence.*CompareWithRefs.*mode=CONVERT_TO_TI_RAND_SEQ_LEN_PARAM_seq.*direction=bidirectional_clip=0.7_netPRC=FP32.*)",
             // TODO: Issue: 54194
             R"(.*ActivationLayerTest.*SoftPlus.*)",
+            // need to implement Export / Import
+            R"(.*IEClassImportExportTestP.*)"
     };
 }
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_integration.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_integration.cpp
index 8edb52954a8295..2eaa15ec86612f 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_integration.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_integration.cpp
@@ -38,36 +38,7 @@ INSTANTIATE_TEST_CASE_P(
 // IEClassNetworkTestP tests, customized to add SKIP_IF_CURRENT_TEST_IS_DISABLED()
 //
 
-using IEClassNetworkTestP_VPU = IEClassNetworkTestP;
-
-TEST_P(IEClassNetworkTestP_VPU, smoke_ImportNetworkNoThrowWithDeviceName) {
-    SKIP_IF_CURRENT_TEST_IS_DISABLED();
-    Core ie;
-    std::stringstream strm;
-    ExecutableNetwork executableNetwork;
-    ASSERT_NO_THROW(executableNetwork = ie.LoadNetwork(actualNetwork, deviceName));
-    ASSERT_NO_THROW(executableNetwork.Export(strm));
-    ASSERT_NO_THROW(executableNetwork = ie.ImportNetwork(strm, deviceName));
-    ASSERT_NO_THROW(executableNetwork.CreateInferRequest());
-}
-
-TEST_P(IEClassNetworkTestP_VPU, smoke_ExportUsingFileNameImportFromStreamNoThrowWithDeviceName) {
-    SKIP_IF_CURRENT_TEST_IS_DISABLED();
-    Core ie;
-    ExecutableNetwork executableNetwork;
-    std::string fileName{"ExportedNetwork"};
-    {
-        ASSERT_NO_THROW(executableNetwork = ie.LoadNetwork(actualNetwork, deviceName));
-        ASSERT_NO_THROW(executableNetwork.Export(fileName));
-    }
-    {
-        std::ifstream strm(fileName);
-        ASSERT_NO_THROW(executableNetwork = ie.ImportNetwork(strm, deviceName));
-    }
-    ASSERT_NO_THROW(executableNetwork.CreateInferRequest());
-}
-
-using IEClassNetworkTestP_VPU_GetMetric = IEClassNetworkTestP_VPU;
+using IEClassNetworkTestP_VPU_GetMetric = IEClassNetworkTestP;
 
 TEST_P(IEClassNetworkTestP_VPU_GetMetric, smoke_OptimizationCapabilitiesReturnsFP16) {
     Core ie;
@@ -86,13 +57,13 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::ValuesIn(devices));
 
 INSTANTIATE_TEST_CASE_P(
-        smoke_IEClassImportExportTestP, IEClassNetworkTestP_VPU,
+        smoke_IEClassImportExportTestP, IEClassImportExportTestP,
         ::testing::Values(std::string(CommonTestUtils::DEVICE_MYRIAD), "HETERO:" + std::string(CommonTestUtils::DEVICE_MYRIAD)));
 
 #if defined(ENABLE_MKL_DNN) && ENABLE_MKL_DNN
 
 INSTANTIATE_TEST_CASE_P(
-        smoke_IEClassImportExportTestP_HETERO_CPU, IEClassNetworkTestP_VPU,
+        smoke_IEClassImportExportTestP_HETERO_CPU, IEClassImportExportTestP,
         ::testing::Values("HETERO:" + std::string(CommonTestUtils::DEVICE_MYRIAD) + ",CPU"));
 #endif
 
diff --git a/inference-engine/tests/functional/plugin/shared/include/behavior/core_integration.hpp b/inference-engine/tests/functional/plugin/shared/include/behavior/core_integration.hpp
index 834db01006f008..adcd0e525b43bf 100644
--- a/inference-engine/tests/functional/plugin/shared/include/behavior/core_integration.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/behavior/core_integration.hpp
@@ -61,6 +61,7 @@ namespace BehaviorTestsDefinitions {
     }                                                                           \
 }
 
+
 class IEClassBasicTestP : public ::testing::Test, public WithParamInterface<std::pair<std::string, std::string> > {
 protected:
     std::string deviceName;
@@ -424,7 +425,16 @@ TEST_P(IEClassBasicTestP, ImportNetworkThrows) {
 
     if (deviceName == CommonTestUtils::DEVICE_CPU ||
         deviceName == CommonTestUtils::DEVICE_GPU) {
-        ASSERT_THROW(ie.ImportNetwork("model", deviceName), NotImplemented);
+        ASSERT_THROW(ie.ImportNetwork("model", deviceName), NetworkNotRead);
+
+        const std::string modelName = "compiled_blob.blob";
+        {
+            std::ofstream file(modelName);
+            file << "content";
+        }
+
+        EXPECT_THROW(ie.ImportNetwork(modelName, deviceName), NotImplemented);
+        ASSERT_EQ(0, std::remove(modelName.c_str()));
     }
 }
 
@@ -432,13 +442,13 @@ TEST(IEClassBasicTest, smoke_ImportNetworkHeteroThrows) {
     SKIP_IF_CURRENT_TEST_IS_DISABLED()
     Core ie;
 
-    ASSERT_THROW(ie.ImportNetwork("model", CommonTestUtils::DEVICE_HETERO), Exception);
+    ASSERT_THROW(ie.ImportNetwork("model", CommonTestUtils::DEVICE_HETERO), NetworkNotRead);
 }
 
 TEST(IEClassBasicTest, smoke_ImportNetworkMultiThrows) {
     SKIP_IF_CURRENT_TEST_IS_DISABLED()
     InferenceEngine::Core ie;
-    ASSERT_THROW(ie.ImportNetwork("model", CommonTestUtils::DEVICE_MULTI), Exception);
+    ASSERT_THROW(ie.ImportNetwork("model", CommonTestUtils::DEVICE_MULTI), NetworkNotRead);
 }
 
 TEST_P(IEClassBasicTestP, ImportNetworkWithNullContextThrows) {
@@ -474,19 +484,18 @@ TEST_P(IEClassNetworkTestP, LoadNetworkActualHeteroDevice2NoThrow) {
 //
 // ImportExportNetwork
 //
-TEST_P(IEClassImportExportTestP, smoke_ImportNetworkNoThrowIfNoDeviceName) {
+
+TEST_P(IEClassImportExportTestP, smoke_ImportNetworkThrowsIfNoDeviceName) {
     SKIP_IF_CURRENT_TEST_IS_DISABLED()
     Core ie;
     std::stringstream strm;
     ExecutableNetwork executableNetwork;
     ASSERT_NO_THROW(executableNetwork = ie.LoadNetwork(actualNetwork, deviceName));
-    SKIP_IF_NOT_IMPLEMENTED(executableNetwork.Export(strm));
-    if (!strm.str().empty()) {
-        SKIP_IF_NOT_IMPLEMENTED(executableNetwork = ie.ImportNetwork(strm));
-    }
-    if (executableNetwork) {
-        ASSERT_NO_THROW(executableNetwork.CreateInferRequest());
-    }
+    ASSERT_NO_THROW(executableNetwork.Export(strm));
+
+    IE_SUPPRESS_DEPRECATED_START
+    ASSERT_THROW(executableNetwork = ie.ImportNetwork(strm), Exception);
+    IE_SUPPRESS_DEPRECATED_END
 }
 
 TEST_P(IEClassImportExportTestP, smoke_ImportNetworkNoThrowWithDeviceName) {
@@ -495,11 +504,9 @@ TEST_P(IEClassImportExportTestP, smoke_ImportNetworkNoThrowWithDeviceName) {
     std::stringstream strm;
     ExecutableNetwork executableNetwork;
     ASSERT_NO_THROW(executableNetwork = ie.LoadNetwork(actualNetwork, deviceName));
-    SKIP_IF_NOT_IMPLEMENTED(executableNetwork.Export(strm));
-    SKIP_IF_NOT_IMPLEMENTED(executableNetwork = ie.ImportNetwork(strm, deviceName));
-    if (executableNetwork) {
-        ASSERT_NO_THROW(executableNetwork.CreateInferRequest());
-    }
+    ASSERT_NO_THROW(executableNetwork.Export(strm));
+    ASSERT_NO_THROW(executableNetwork = ie.ImportNetwork(strm, deviceName));
+    ASSERT_NO_THROW(executableNetwork.CreateInferRequest());
 }
 
 TEST_P(IEClassImportExportTestP, smoke_ExportUsingFileNameImportFromStreamNoThrowWithDeviceName) {
@@ -509,18 +516,16 @@ TEST_P(IEClassImportExportTestP, smoke_ExportUsingFileNameImportFromStreamNoThro
     std::string fileName{"ExportedNetwork"};
     {
         ASSERT_NO_THROW(executableNetwork = ie.LoadNetwork(simpleNetwork, deviceName));
-        SKIP_IF_NOT_IMPLEMENTED(executableNetwork.Export(fileName));
+        ASSERT_NO_THROW(executableNetwork.Export(fileName));
     }
-    if (CommonTestUtils::fileExists(fileName)) {
+    {
         {
             std::ifstream strm(fileName);
-            SKIP_IF_NOT_IMPLEMENTED(executableNetwork = ie.ImportNetwork(strm, deviceName));
+            ASSERT_NO_THROW(executableNetwork = ie.ImportNetwork(strm, deviceName));
         }
         ASSERT_EQ(0, remove(fileName.c_str()));
     }
-    if (executableNetwork) {
-        ASSERT_NO_THROW(executableNetwork.CreateInferRequest());
-    }
+    ASSERT_NO_THROW(executableNetwork.CreateInferRequest());
 }
 
 //
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/impl/mock_inference_plugin_internal.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/impl/mock_inference_plugin_internal.hpp
index 2a68e96c19e790..82e2ff61f335f1 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/impl/mock_inference_plugin_internal.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/impl/mock_inference_plugin_internal.hpp
@@ -32,11 +32,8 @@ class MockInferencePluginInternal : public InferenceEngine::IInferencePlugin {
     MOCK_METHOD1(AddExtension, void(InferenceEngine::IExtensionPtr ext_ptr));
     MOCK_METHOD1(SetConfig, void(const std::map <std::string, std::string> &));
 
-    using InferenceEngine::IInferencePlugin::ImportNetwork;
-
-    std::shared_ptr<InferenceEngine::IExecutableNetworkInternal> ImportNetworkImpl(std::istream& stream,
-                                                                                   const std::map <std::string, std::string>&) {
-        std::getline(stream, importedString);
+    std::shared_ptr<InferenceEngine::IExecutableNetworkInternal>
+    ImportNetwork(std::istream& stream, const std::map <std::string, std::string>&) {
         return {};
     }
 
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_iexecutable_network_internal.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_iexecutable_network_internal.hpp
index 4e221be90f7945..c1cc30a944b430 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_iexecutable_network_internal.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_iexecutable_network_internal.hpp
@@ -34,8 +34,4 @@ class MockIExecutableNetworkInternal : public IExecutableNetworkInternal {
     void WrapOstreamExport(std::ostream& networkModel) {
         IExecutableNetworkInternal::Export(networkModel);
     }
-    const std::string exportString = "MockExecutableNetworkInternal";
-    void ExportImpl(std::ostream& networkModel) override {
-        networkModel << exportString << std::endl;
-    }
 };
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_engine/mock_plugin.cpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_engine/mock_plugin.cpp
index cd2e7b95f469d8..4408614f61c907 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_engine/mock_plugin.cpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_engine/mock_plugin.cpp
@@ -70,8 +70,8 @@ MockPlugin::LoadExeNetworkImpl(const CNNNetwork& network,
 }
 
 std::shared_ptr<InferenceEngine::IExecutableNetworkInternal>
-MockPlugin::ImportNetworkImpl(std::istream& networkModel,
-                              const std::map<std::string, std::string>& config) {
+MockPlugin::ImportNetwork(std::istream& networkModel,
+                          const std::map<std::string, std::string>& config) {
     if (_target) {
         return _target->ImportNetwork(networkModel, config);
     } else {
@@ -80,9 +80,9 @@ MockPlugin::ImportNetworkImpl(std::istream& networkModel,
 }
 
 std::shared_ptr<InferenceEngine::IExecutableNetworkInternal>
-MockPlugin::ImportNetworkImpl(std::istream& networkModel,
-                              const std::shared_ptr<InferenceEngine::RemoteContext>& context,
-                              const std::map<std::string, std::string>& config) {
+MockPlugin::ImportNetwork(std::istream& networkModel,
+                         const std::shared_ptr<InferenceEngine::RemoteContext>& context,
+                         const std::map<std::string, std::string>& config) {
     if (_target) {
         return _target->ImportNetwork(networkModel, context, config);
     } else {
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_engine/mock_plugin.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_engine/mock_plugin.hpp
index c01a8a8d1750f0..c2654061abd1f8 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_engine/mock_plugin.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_engine/mock_plugin.hpp
@@ -35,13 +35,13 @@ class MockPlugin : public InferenceEngine::IInferencePlugin {
                 const std::map<std::string, std::string> &config) override;
 
     std::shared_ptr<InferenceEngine::IExecutableNetworkInternal>
-    ImportNetworkImpl(std::istream& networkModel,
+    ImportNetwork(std::istream& networkModel,
         const std::map<std::string, std::string>& config) override;
 
     std::shared_ptr<InferenceEngine::IExecutableNetworkInternal>
-    ImportNetworkImpl(std::istream& networkModel,
-                      const std::shared_ptr<InferenceEngine::RemoteContext>& context,
-                      const std::map<std::string, std::string>& config) override;
+    ImportNetwork(std::istream& networkModel,
+        const std::shared_ptr<InferenceEngine::RemoteContext>& context,
+        const std::map<std::string, std::string>& config) override;
 
     InferenceEngine::Parameter GetMetric(const std::string& name,
                         const std::map<std::string, InferenceEngine::Parameter>& options) const override;
diff --git a/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_plugin_test.cpp b/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_plugin_test.cpp
index 0945510d7a00d0..2d26c7bd0e2a7f 100644
--- a/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_plugin_test.cpp
+++ b/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_plugin_test.cpp
@@ -149,29 +149,6 @@ TEST_F(InferenceEnginePluginInternalTest, failToSetNotAllocatedBlob) {
     }
 }
 
-TEST_F(InferenceEnginePluginInternalTest, executableNetworkInternalExportsMagicAndName) {
-    std::stringstream strm;
-    ASSERT_NO_THROW(mockIExeNetworkInternal->WrapOstreamExport(strm));
-    ExportMagic actualMagic = {};
-    strm.read(actualMagic.data(), actualMagic.size());
-    ASSERT_EQ(exportMagic, actualMagic);
-    std::string pluginName;
-    std::getline(strm, pluginName);
-    ASSERT_EQ(pluginId, pluginName);
-    std::string exportedString;
-    std::getline(strm, exportedString);
-    ASSERT_EQ(mockIExeNetworkInternal->exportString, exportedString);
-}
-
-TEST_F(InferenceEnginePluginInternalTest, pluginInternalEraseMagicAndNameWhenImports) {
-    std::stringstream strm;
-    ASSERT_NO_THROW(mockIExeNetworkInternal->WrapOstreamExport(strm));
-    ASSERT_NO_THROW(mock_plugin_impl->ImportNetwork(strm, {}));
-    ASSERT_EQ(mockIExeNetworkInternal->exportString, mock_plugin_impl->importedString);
-    mock_plugin_impl->importedString = {};
-}
-
-
 TEST(InferencePluginTests, throwsOnUninitializedGetVersion) {
     InferencePlugin plg;
     ASSERT_THROW(plg.GetVersion(), Exception);

From cbd48cf15f7c5c0d1779a5e2c492007af318f888 Mon Sep 17 00:00:00 2001
From: Nikolay Shchegolev <nikolay.shchegolev@intel.com>
Date: Wed, 9 Jun 2021 10:12:58 +0300
Subject: [PATCH 041/178] [CPU] Statically found issues on Windows. (#6075)

---
 .../src/mkldnn_plugin/nodes/mkldnn_convert_node.cpp          | 2 ++
 .../src/mkldnn_plugin/nodes/mkldnn_depth_to_space_node.cpp   | 5 +++--
 2 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_convert_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_convert_node.cpp
index d226dd73890ec6..678922f3a4b5b1 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_convert_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_convert_node.cpp
@@ -42,6 +42,8 @@ MKLDNNConvertNode::MKLDNNConvertNode(const InferenceEngine::SizeVector &dims, co
     addOriginalInputPrecision(inPrc);
     outDims.emplace_back(dims);
     addOriginalOutputPrecision(outPrc);
+
+    errorPrefix = "Convert node with name '" + getName() + "'";
 }
 
 void MKLDNNConvertNode::getSupportedDescriptors() {
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_depth_to_space_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_depth_to_space_node.cpp
index 2f97bbd2f85bda..38bebcd5271072 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_depth_to_space_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_depth_to_space_node.cpp
@@ -58,6 +58,8 @@ MKLDNNDepthToSpaceNode::MKLDNNDepthToSpaceNode(const std::shared_ptr<ngraph::Nod
         if (blockSize == 0)
             THROW_ERROR << "has incorrect block_size parameter is zero!";
 
+        size_t nSpatialDims = inDims[0].ndims() - 2;
+        blockStep = static_cast<size_t>(std::pow(blockSize, nSpatialDims));
     } else {
         IE_THROW(NotImplemented) << errorMessage;
     }
@@ -74,14 +76,13 @@ void MKLDNNDepthToSpaceNode::getSupportedDescriptors() {
     if (srcDims.size() != dstDims.size())
         THROW_ERROR << "has incorrect number of input/output dimensions";
 
-    size_t nSpatialDims = srcDims.size() - 2;
-    blockStep = static_cast<size_t>(std::pow(blockSize, nSpatialDims));
     if (srcDims[1] % blockStep)
         THROW_ERROR << "has block_size parameter which is incompatible with input tensor channels dimension size";
 
     if (srcDims[1] / blockStep != dstDims[1])
         THROW_ERROR << "has incompatible input/output channels";
 
+    size_t nSpatialDims = srcDims.size() - 2;
     for (size_t i = 0; i < nSpatialDims; ++i) {
         if (srcDims[i + 2] * blockSize != dstDims[i + 2])
             THROW_ERROR << "has incompatible spatial dims";

From 421465adc15521f91f6ea9dbbbb9e0bbeb0ea933 Mon Sep 17 00:00:00 2001
From: iliya mironov <iliya.mironov@intel.com>
Date: Wed, 9 Jun 2021 12:14:39 +0300
Subject: [PATCH 042/178] Add ShapeOfConstFolding transform (#5858)

* Add ShapeOfConstFolding transform

* Add unit tests

* Update bom file

* Update transform file

* Hot fix

* Fix midle replaser

* Update unit tests

* Fix get value

* Refactoring Const Folding transformation. Move to back

* Update bom file

* Remove unuse code

* Add some unit tests

* Refactoring unit test

* Hot fix
---
 model-optimizer/automation/package_BOM.txt    |   1 +
 .../extensions/back/ShapeOfConstFolding.py    |  29 +++
 .../back/ShapeOfConstFolding_test.py          | 170 ++++++++++++++++++
 3 files changed, 200 insertions(+)
 create mode 100644 model-optimizer/extensions/back/ShapeOfConstFolding.py
 create mode 100644 model-optimizer/unit_tests/extensions/back/ShapeOfConstFolding_test.py

diff --git a/model-optimizer/automation/package_BOM.txt b/model-optimizer/automation/package_BOM.txt
index be0dc1c37e0dd5..e7c779543cdda8 100644
--- a/model-optimizer/automation/package_BOM.txt
+++ b/model-optimizer/automation/package_BOM.txt
@@ -57,6 +57,7 @@ extensions/back/ReverseInputChannels.py
 extensions/back/RNNSequenceTypeRename.py
 extensions/back/ScalarConstNormalize.py
 extensions/back/SelectBroadcast.py
+extensions/back/ShapeOfConstFolding.py
 extensions/back/ShuffleChannelPatternOptimization.py
 extensions/back/ShufflenetReLUReorder.py
 extensions/back/SpecialNodesFinalization.py
diff --git a/model-optimizer/extensions/back/ShapeOfConstFolding.py b/model-optimizer/extensions/back/ShapeOfConstFolding.py
new file mode 100644
index 00000000000000..b97b46fc35dd09
--- /dev/null
+++ b/model-optimizer/extensions/back/ShapeOfConstFolding.py
@@ -0,0 +1,29 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+from mo.back.replacement import BackReplacementPattern
+from mo.graph.graph import Graph, rename_nodes
+from mo.ops.const import Const
+
+
+class ShapeOfConstFolding(BackReplacementPattern):
+    """
+    The transformation folds ShapeOf(Const) -> Const
+    """
+    enabled = True
+
+    def run_after(self):
+        from extensions.back.MatMulNormalizer import SmartReshape_HC_Reshape_MatMul
+        return [SmartReshape_HC_Reshape_MatMul]
+
+    def find_and_replace_pattern(self, graph: Graph):
+        for shapeof_node in graph.get_op_nodes(op='ShapeOf'):
+            in_node = shapeof_node.in_port(0).get_source().node
+            if in_node.op == 'Const':
+                shapeof_node.in_port(0).disconnect()
+                shape_name = shapeof_node.soft_get('name', shapeof_node.id)
+                shape_value = shapeof_node.out_port(0).data.get_value()
+                shape_const_node = Const(graph, {'name': shape_name + '/ExecutionConstValue',
+                                                 'value': shape_value}).create_node()
+                shapeof_node.out_port(0).get_connection().set_source(shape_const_node.out_port(0))
+                rename_nodes([(shapeof_node, shape_name + '/TBD'), (shape_const_node, shape_name)])
diff --git a/model-optimizer/unit_tests/extensions/back/ShapeOfConstFolding_test.py b/model-optimizer/unit_tests/extensions/back/ShapeOfConstFolding_test.py
new file mode 100644
index 00000000000000..562c1416342b04
--- /dev/null
+++ b/model-optimizer/unit_tests/extensions/back/ShapeOfConstFolding_test.py
@@ -0,0 +1,170 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import unittest
+
+import numpy as np
+
+from extensions.back.ShapeOfConstFolding import ShapeOfConstFolding
+from mo.front.common.partial_infer.eltwise import eltwise_infer
+from mo.front.common.partial_infer.utils import int64_array
+from mo.utils.ir_engine.compare_graphs import compare_graphs
+from unit_tests.utils.graph import build_graph
+
+const_value = np.random.rand(1, 3, 30, 30)
+nodes_attributes = {'input': {'shape': int64_array([1, 3, 30, 30]), 'type': 'Parameter', 'kind': 'op',
+                              'op': 'Parameter'},
+                    'input_data': {'value': None, 'shape': int64_array([1, 3, 30, 30]), 'kind': 'data'},
+                    'const': {'type': 'Const', 'kind': 'op', 'op': 'Const', 'value': const_value},
+                    'const_data': {'kind': 'data', 'value': const_value},
+                    'shapeof_input': {'kind': 'op', 'op': 'ShapeOf', 'value': int64_array([1, 3, 30, 30])},
+                    'shapeof_input_data': {'value': None, 'shape': None, 'kind': 'data',
+                                           'value': int64_array([1, 3, 30, 30])},
+
+                    'shapeof_const': {'kind': 'op', 'op': 'ShapeOf', 'value': int64_array([1, 3, 30, 30])},
+                    'shapeof_const_data': {'value': None, 'shape': None, 'kind': 'data',
+                                           'value': int64_array([1, 3, 30, 30])},
+
+                    'mul': {'kind': 'op', 'op': 'Mul', 'infer': lambda node: eltwise_infer(node, lambda a, b: a * b)},
+                    'mul_data': {'kind': 'data', 'value': np.array([1, 9, 900, 900])},
+                    'last': {'kind': 'op', 'op': 'Result'},
+
+                    # new nodes
+                    'new_const_shapeof': {'type': 'Const', 'kind': 'op', 'op': 'Const',
+                                          'value': int64_array([1, 3, 30, 30])}
+                    }
+
+const_value2 = np.random.rand(30, 30)
+nodes_attributes2 = {'input': {'shape': int64_array([1, 3, 30, 30]), 'type': 'Parameter', 'kind': 'op',
+                               'op': 'Parameter'},
+                     'input_data': {'value': None, 'shape': int64_array([1, 3, 30, 30]), 'kind': 'data'},
+
+                     'const': {'type': 'Const', 'kind': 'op', 'op': 'Const', 'value': const_value2},
+                     'const_data': {'kind': 'data', 'value': const_value2},
+
+                     'shapeof_const': {'kind': 'op', 'op': 'ShapeOf', 'value': int64_array([2700, 30])},
+                     'shapeof_const_data': {'value': int64_array([2700, 30]), 'shape': None, 'kind': 'data'},
+
+                     'gather': {'kind': 'op', 'op': 'Gather', 'batch_dims': 0},
+                     'gather_data': {'kind': 'data'},
+
+                     'const_concat': {'type': 'Const', 'kind': 'op', 'op': 'Const', 'value': [1]},
+                     'const_concat_data': {'kind': 'data', 'value': [1]},
+                     'concat': {'kind': 'op', 'op': 'Concat'},
+                     'concat_data': {'kind': 'data'},
+
+                     'reshape': {'kind': 'op', 'op': 'Reshape'},
+                     'reshape_data': {'kind': 'data'},
+
+                     'matmul': {'kind': 'op', 'op': 'MatMul'},
+                     'matmul_data': {'kind': 'data'},
+                     'last': {'kind': 'op', 'op': 'Result'},
+
+                     # new nodes
+                     'new_const_shapeof': {'type': 'Const', 'kind': 'op', 'op': 'Const',
+                                          'value': int64_array([2700, 30])},
+                     }
+
+
+class ShapeOfConstFoldingTests(unittest.TestCase):
+    def test_const_with_one_output(self):
+        graph = build_graph(nodes_attributes,
+                            [('input', 'input_data'),
+                             ('input_data', 'shapeof_input'),
+                             ('shapeof_input', 'shapeof_input_data'),
+                             ('shapeof_input_data', 'mul'),
+                             ('const', 'const_data'),
+                             ('const_data', 'shapeof_const'),
+                             ('shapeof_const', 'shapeof_const_data'),
+                             ('shapeof_const_data', 'mul'),
+                             ('mul', 'mul_data'),
+                             ('mul_data', 'last')],
+                            {
+                                'input': {'shape': int64_array([1, 3, 30, 30])},
+                                'input_data': {'shape': int64_array([1, 3, 30, 30])},
+                                'shapeof_input': {'value': int64_array([1, 3, 30, 30])},
+                                'shapeof_input_data': {'value': int64_array([1, 3, 30, 30])},
+                                'const': {'value': const_value},
+                                'const_data': {'value': const_value},
+                                'shapeof_const': {'value': int64_array([1, 3, 30, 30])},
+                                'shapeof_const_data': {'value': int64_array([1, 3, 30, 30])},
+                                'mul_data': {'value': int64_array([1, 9, 900, 900])},
+                            },
+                            nodes_with_edges_only=True)
+
+        graph_ref = build_graph(nodes_attributes,
+                                [('input', 'input_data'),
+                                 ('input_data', 'shapeof_input'),
+                                 ('shapeof_input', 'shapeof_input_data'),
+                                 ('shapeof_input_data', 'mul'),
+                                 ('new_const_shapeof', 'shapeof_const_data'),
+                                 ('shapeof_const_data', 'mul'),
+                                 ('mul', 'mul_data'),
+                                 ('mul_data', 'last')],
+                                {
+                                    'input': {'shape': int64_array([1, 3, 30, 30])},
+                                    'input_data': {'shape': int64_array([1, 3, 30, 30])},
+                                    'shapeof_input': {'value': int64_array([1, 3, 30, 30])},
+                                    'shapeof_input_data': {'value': int64_array([1, 3, 30, 30])},
+                                    'new_const_shapeof': {'value': int64_array([1, 3, 30, 30])},
+                                    'shapeof_const_data': {'value': int64_array([1, 3, 30, 30])},
+                                    'mul_data': {'value': int64_array([1, 9, 900, 900])},
+                                },
+                                nodes_with_edges_only=True)
+        ShapeOfConstFolding().find_and_replace_pattern(graph)
+        (flag, resp) = compare_graphs(graph, graph_ref, 'last')
+        self.assertTrue(flag, resp)
+
+    def test_const_with_two_outputs(self):
+        graph = build_graph(nodes_attributes2,
+                            [('input', 'input_data'),
+                             ('input_data', 'reshape'),
+                             ('const', 'const_data'),
+                             ('const_data', 'shapeof_const'),
+                             ('shapeof_const', 'shapeof_const_data'),
+                             ('shapeof_const_data', 'gather'),
+                             ('gather', 'gather_data'),
+                             ('const_concat', 'const_concat_data'),
+                             ('const_concat_data', 'concat'),
+                             ('gather_data', 'concat'),
+                             ('concat', 'reshape'),
+                             ('reshape', 'reshape_data'),
+                             ('reshape_data', 'matmul'),
+                             ('const_data', 'matmul'),
+                             ('matmul', 'matmul_data'),
+                             ('matmul_data', 'last')
+                             ],
+                            {
+                                'input': {'shape': int64_array([1, 3, 30, 30])},
+                                'input_data': {'shape': int64_array([1, 3, 30, 30])},
+                                'shapeof_const': {'value': int64_array([2700, 30])},
+                                'shapeof_const_data': {'value': int64_array([2700, 30])},
+                            },
+                            nodes_with_edges_only=True)
+
+        graph_ref = build_graph(nodes_attributes2,
+                                [('input', 'input_data'),
+                                 ('input_data', 'reshape'),
+                                 ('new_const_shapeof', 'shapeof_const_data'),
+                                 ('shapeof_const_data', 'gather'),
+                                 ('gather', 'gather_data'),
+                                 ('const_concat', 'const_concat_data'),
+                                 ('const_concat_data', 'concat'),
+                                 ('gather_data', 'concat'),
+                                 ('concat', 'reshape'),
+                                 ('reshape', 'reshape_data'),
+                                 ('reshape_data', 'matmul'),
+                                 ('const', 'const_data'),
+                                 ('const_data', 'matmul'),
+                                 ('matmul', 'matmul_data'),
+                                 ('matmul_data', 'last')],
+                                {
+                                    'input': {'shape': int64_array([1, 3, 30, 30])},
+                                    'input_data': {'shape': int64_array([1, 3, 30, 30])},
+                                    'new_const_shapeof': {'value': int64_array([2700, 30])},
+                                    'shapeof_const_data': {'value': int64_array([2700, 30])},
+                                },
+                                nodes_with_edges_only=True)
+        ShapeOfConstFolding().find_and_replace_pattern(graph)
+        (flag, resp) = compare_graphs(graph, graph_ref, 'last')
+        self.assertTrue(flag, resp)

From aa0c93300cc3ddd00e0cdcf0d10e1c57b16247cf Mon Sep 17 00:00:00 2001
From: Ivan Tikhonov <ivan.tikhonov@intel.com>
Date: Wed, 9 Jun 2021 12:19:36 +0300
Subject: [PATCH 043/178] Update for Python API to create networks with Memory
 layers (#6079)

* Update python API for Memory

* Run unit tests on CPU only

* add missed import

* Update tests

* fix py tests
---
 .../src/openvino/inference_engine/ie_api.pxd  |  2 +-
 .../python/tests/test_InferRequest.py         | 80 ++++++++++++++-----
 ngraph/python/src/pyngraph/function.cpp       | 38 +++++++++
 ngraph/python/tests/test_ngraph/test_basic.py | 22 +++++
 4 files changed, 122 insertions(+), 20 deletions(-)

diff --git a/inference-engine/ie_bridges/python/src/openvino/inference_engine/ie_api.pxd b/inference-engine/ie_bridges/python/src/openvino/inference_engine/ie_api.pxd
index 5d942f93050246..efb389259d36f8 100644
--- a/inference-engine/ie_bridges/python/src/openvino/inference_engine/ie_api.pxd
+++ b/inference-engine/ie_bridges/python/src/openvino/inference_engine/ie_api.pxd
@@ -2,7 +2,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 from .cimport ie_api_impl_defs as C
-from .ie_api_impl_defs cimport CBlob, CTensorDesc, InputInfo, CPreProcessChannel, CPreProcessInfo, CExecutableNetwork
+from .ie_api_impl_defs cimport CBlob, CTensorDesc, InputInfo, CPreProcessChannel, CPreProcessInfo, CExecutableNetwork, CVariableState
 
 import os
 
diff --git a/inference-engine/ie_bridges/python/tests/test_InferRequest.py b/inference-engine/ie_bridges/python/tests/test_InferRequest.py
index 6928944139d722..af79c0ff155bf8 100644
--- a/inference-engine/ie_bridges/python/tests/test_InferRequest.py
+++ b/inference-engine/ie_bridges/python/tests/test_InferRequest.py
@@ -16,6 +16,20 @@
 path_to_img = image_path()
 
 
+def create_function_with_memory(input_shape, data_type):
+    import ngraph as ng
+    from ngraph.impl import Function, Type
+
+    input_data = ng.parameter(input_shape, name="input_data", dtype=data_type)
+    rv = ng.read_value(input_data, "var_id_667")
+    add = ng.add(rv, input_data, name="MemoryAdd")
+    node = ng.assign(add, "var_id_667")
+    res = ng.result(add, "res")
+    func = Function(results=[res], sinks=[node], parameters=[input_data], name="name")
+    caps = Function.to_capsule(func)
+    return caps
+
+
 def read_image():
     import cv2
     n, c, h, w = (1, 3, 32, 32)
@@ -525,28 +539,56 @@ def test_resize_algorithm_work(device):
     assert np.allclose(res_1, res_2, atol=1e-2, rtol=1e-2)
 
 
-# issue 56653
-@pytest.mark.skip(reason="Test will enable when nGraph Python API allows to create network with memory")
-def test_query_state(device):
-    import ngraph as ng
-    from ngraph.impl import Function
-    input_data = ng.parameter([5, 7], name="input_data", dtype=np.float32)
-    rv = ng.read_value(input_data, "var_id_667")
-    #a = ng.add(rv, input_data)
-    node = ng.assign(rv, "var_id_667")
-    res = ng.result(rv, "res")
-    func = Function([res], sinks=[node], parameters=[input_data], name='test')
-    caps = Function.to_capsule(func)
+@pytest.mark.parametrize("mode", ["set_init_memory_state", "reset_memory_state", "normal"])
+@pytest.mark.parametrize("data_type", ["FP32", "FP16", "I32"])
+@pytest.mark.parametrize("input_shape", [[10], [10, 10], [10, 10, 10], [2, 10, 10, 10]])
+@pytest.mark.skipif(os.environ.get("TEST_DEVICE", "CPU") != "CPU",
+                    reason=f"Can't run test on device {os.environ.get('TEST_DEVICE', 'CPU')}, "
+                    "Memory layers fully supported only on CPU")
+def test_query_state_write_buffer(device, input_shape, data_type, mode):
+    ie_core = ie.IECore()
+    if device == "CPU":
+        if ie_core.get_metric(device, "FULL_DEVICE_NAME") == "arm_compute::NEON":
+            pytest.skip("Can't run on ARM plugin")
 
-    net = ie.IENetwork(caps)
+    layout = ["C", "HW", "CHW", "NCHW"]
+    np_data_type = {"FP32": np.float32, "FP16": np.float16, "I32": np.int32}
+
+    from openvino.inference_engine import TensorDesc, Blob
+
+    net = ie.IENetwork(create_function_with_memory(input_shape, np_data_type[data_type]))
     ie_core = ie.IECore()
     exec_net = ie_core.load_network(network=net, device_name=device, num_requests=1)
     request = exec_net.requests[0]
     mem_states = request.query_state()
     mem_state = mem_states[0]
-    with pytest.raises(ValueError) as e:
-        ones_arr = np.ones(shape=(1, 800), dtype=np.float32)
-        mem_state.state.buffer[:] = ones_arr
-    assert "assignment destination is read-only" in str(e.value)
-    assert mem_state.name == 'id_1'
-    assert mem_state.state.tensor_desc.precision == 'FP32'
+
+    assert mem_state.name == 'var_id_667'
+    # todo: Uncomment after fix 45611,
+    #  CPU plugin returns outputs and memory state in FP32 in case of FP16 original precision
+    #assert mem_state.state.tensor_desc.precision == data_type
+
+    for i in range(1, 10):
+        if mode == "set_init_memory_state":
+            # create initial value
+            const_init = 5
+            init_array = np.full(input_shape, const_init, dtype=np_data_type[mem_state.state.tensor_desc.precision])
+            tensor_desc = TensorDesc(mem_state.state.tensor_desc.precision, input_shape, layout[len(input_shape) - 1])
+            blob = Blob(tensor_desc, init_array)
+            mem_state.state = blob
+
+            res = exec_net.infer({"input_data": np.full(input_shape, 1, dtype=np_data_type[data_type])})
+            expected_res = np.full(input_shape, 1 + const_init, dtype=np_data_type[data_type])
+        elif mode == "reset_memory_state":
+            # reset initial state of ReadValue to zero
+            mem_state.reset()
+            res = exec_net.infer({"input_data": np.full(input_shape, 1, dtype=np_data_type[data_type])})
+
+            # always ones
+            expected_res = np.full(input_shape, 1, dtype=np_data_type[data_type])
+        else:
+            res = exec_net.infer({"input_data": np.full(input_shape, 1, dtype=np_data_type[data_type])})
+            expected_res = np.full(input_shape, i, dtype=np_data_type[data_type])
+
+        assert np.allclose(res['MemoryAdd'], expected_res, atol=1e-6), \
+            "Expected values: {} \n Actual values: {} \n".format(expected_res, res)
diff --git a/ngraph/python/src/pyngraph/function.cpp b/ngraph/python/src/pyngraph/function.cpp
index 6b60372a85aff2..ed84d9c458c0a4 100644
--- a/ngraph/python/src/pyngraph/function.cpp
+++ b/ngraph/python/src/pyngraph/function.cpp
@@ -7,6 +7,7 @@
 
 #include "ngraph/function.hpp"     // ngraph::Function
 #include "ngraph/op/parameter.hpp" // ngraph::op::Parameter
+#include "ngraph/op/sink.hpp"
 #include "pyngraph/function.hpp"
 
 namespace py = pybind11;
@@ -17,6 +18,42 @@ void regclass_pyngraph_Function(py::module m)
 {
     py::class_<ngraph::Function, std::shared_ptr<ngraph::Function>> function(m, "Function");
     function.doc() = "ngraph.impl.Function wraps ngraph::Function";
+
+    function.def(py::init([](const ngraph::ResultVector& res,
+                             const std::vector<std::shared_ptr<ngraph::Node>>& nodes,
+                             const ngraph::ParameterVector& params,
+                             const std::string& name) {
+                     ngraph::SinkVector sinks;
+                     for (const auto& node : nodes)
+                     {
+                         auto sink = std::dynamic_pointer_cast<ngraph::op::Sink>(node);
+                         NGRAPH_CHECK(sink != nullptr, "Node {} is not instance of Sink");
+                         sinks.push_back(sink);
+                     }
+                     return std::make_shared<ngraph::Function>(res, sinks, params, name);
+                 }),
+                 py::arg("results"),
+                 py::arg("sinks"),
+                 py::arg("parameters"),
+                 py::arg("name"),
+                 R"(
+                    Create user-defined Function which is a representation of a model.
+
+                    Parameters
+                    ----------
+                    results : List[op.Result]
+                        List of results.
+
+                    sinks : List[Node]
+                        List of Nodes to be used as Sinks (e.g. Assign ops).
+
+                    parameters : List[op.Parameter]
+                        List of parameters.
+
+                    name : str
+                        String to set as function's friendly name.
+                 )");
+
     function.def(py::init<const std::vector<std::shared_ptr<ngraph::Node>>&,
                           const std::vector<std::shared_ptr<ngraph::op::Parameter>>&,
                           const std::string&>(),
@@ -37,6 +74,7 @@ void regclass_pyngraph_Function(py::module m)
                     name : str
                         String to set as function's friendly name.
                  )");
+
     function.def(py::init<const std::shared_ptr<ngraph::Node>&,
                           const std::vector<std::shared_ptr<ngraph::op::Parameter>>&,
                           const std::string&>(),
diff --git a/ngraph/python/tests/test_ngraph/test_basic.py b/ngraph/python/tests/test_ngraph/test_basic.py
index 210bcb99ae0921..da6cf993d3f4f1 100644
--- a/ngraph/python/tests/test_ngraph/test_basic.py
+++ b/ngraph/python/tests/test_ngraph/test_basic.py
@@ -403,3 +403,25 @@ def test_mutiple_outputs():
     output = computation(input_data)
 
     assert np.equal(output, expected_output).all()
+
+
+def test_sink_function_ctor():
+    input_data = ng.parameter([2, 2], name="input_data", dtype=np.float32)
+    rv = ng.read_value(input_data, "var_id_667")
+    add = ng.add(rv, input_data, name="MemoryAdd")
+    node = ng.assign(add, "var_id_667")
+    res = ng.result(add, "res")
+    function = Function(results=[res], sinks=[node], parameters=[input_data], name="TestFunction")
+
+    ordered_ops = function.get_ordered_ops()
+    op_types = [op.get_type_name() for op in ordered_ops]
+    assert op_types == ["Parameter", "ReadValue", "Add", "Assign", "Result"]
+    assert len(function.get_ops()) == 5
+    assert function.get_output_size() == 1
+    assert function.get_output_op(0).get_type_name() == "Result"
+    assert function.get_output_element_type(0) == input_data.get_element_type()
+    assert list(function.get_output_shape(0)) == [2, 2]
+    assert (function.get_parameters()[0].get_partial_shape()) == PartialShape([2, 2])
+    assert len(function.get_parameters()) == 1
+    assert len(function.get_results()) == 1
+    assert function.get_friendly_name() == "TestFunction"

From 8c74a0a52c1cfa4a74fa2539e0ba4d91d80f6145 Mon Sep 17 00:00:00 2001
From: Ivan Novoselov <ivan.novoselov@intel.com>
Date: Wed, 9 Jun 2021 14:27:43 +0300
Subject: [PATCH 044/178] [CPU] Fixed reorder for strided both inputs and
 outputs (#6082)

---
 inference-engine/thirdparty/mkl-dnn | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/inference-engine/thirdparty/mkl-dnn b/inference-engine/thirdparty/mkl-dnn
index 1cb9d0615aaf51..87516e47dae71f 160000
--- a/inference-engine/thirdparty/mkl-dnn
+++ b/inference-engine/thirdparty/mkl-dnn
@@ -1 +1 @@
-Subproject commit 1cb9d0615aaf511b51b8f8fc3c3ff8805ad9be6c
+Subproject commit 87516e47dae71fc9c326d0f3685c1572c740e127

From 8dd914477c9a377e914de0318ea698ed01edfa3b Mon Sep 17 00:00:00 2001
From: Evgeny Talanin <evgeny.talanin@intel.com>
Date: Wed, 9 Jun 2021 15:06:21 +0300
Subject: [PATCH 045/178] Pin mypy dependency (#6098)

---
 ngraph/python/tox.ini | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ngraph/python/tox.ini b/ngraph/python/tox.ini
index e0ccc85785ef60..de7bb8337b284b 100644
--- a/ngraph/python/tox.ini
+++ b/ngraph/python/tox.ini
@@ -7,7 +7,7 @@ skip_install=True
 deps =
   -rrequirements.txt
   -rrequirements_test.txt
-  mypy
+  mypy<0.900
   flake8-bugbear
   pytest-xdist
 setenv =

From 3bc2c46693de0e5d14666068b069d00a196e0786 Mon Sep 17 00:00:00 2001
From: Daria Mityagina <daria.mityagina@intel.com>
Date: Wed, 9 Jun 2021 16:27:52 +0300
Subject: [PATCH 046/178] [IE][VPU] Yolo_v5 support (#5976)

Changing the threshold for VPU scales
---
 .../src/middleend/passes/weights_analysis.cpp                | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/weights_analysis.cpp b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/weights_analysis.cpp
index 213e06ee1f57c7..27e703ec4f53cc 100644
--- a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/weights_analysis.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/weights_analysis.cpp
@@ -92,7 +92,7 @@ bool checkGrowingOutput(const Model& model) {
         return false;
     }
 
-    static const float SCALE_THRESHOLD = 0.125f;
+    static const float SCALE_THRESHOLD = 0.1f;
 
     for (const auto& stage : model->getStages()) {
         if (stage->type() != StageType::Power &&
@@ -248,14 +248,13 @@ void PassImpl::run(const Model& model) {
                 if (firstStage && shift < 4 && isGrowingOutput && weights->desc().dim(Dim::C) > 1) {
                     normalVal = 5;
                 }
-
                 shift = correctShift(shift, firstStage, stage->origLayer()->type);
                 shift -= normalVal;
             }
 
             firstStage = false;
             scale = 1;
-            if (shift > scaleThreshold) {
+            if (shift >= scaleThreshold) {
                 scale = static_cast<float>(1ULL << static_cast<std::uint32_t>(shift));
             }
 

From 6a5b6b8b30165c71cdb1eb160136ce429c0332f3 Mon Sep 17 00:00:00 2001
From: Anton Chetverikov <Anton.Chetverikov@intel.com>
Date: Wed, 9 Jun 2021 17:26:52 +0300
Subject: [PATCH 047/178] Add transformation to normalize negative indices in
 Gather (#6028)

* Add transformation to normalize negative indices in Gather

* Update transformation name and add constant indices input to the pattern

* Apply comments

* Fix copyright year

* Add more tests

* Update type logic

* Add test with different types and update old ones

* Update pattern and logic to check only axis dimension is static, add appropriate tests

* Fix cdestyle

* Add axis normalization

* Fix wrong value in gather input

* Add related tests

* Add axis_constant to check

* Remove ngraph_check
---
 .../gather_normalize_negative_indices.hpp     |  29 ++
 .../common_optimizations.cpp                  |   2 +
 .../gather_normalize_negative_indices.cpp     |  77 +++++
 ...gather_normalize_negative_indices_test.cpp | 306 ++++++++++++++++++
 4 files changed, 414 insertions(+)
 create mode 100644 inference-engine/src/transformations/include/transformations/op_conversions/gather_normalize_negative_indices.hpp
 create mode 100644 inference-engine/src/transformations/src/transformations/op_conversions/gather_normalize_negative_indices.cpp
 create mode 100644 inference-engine/tests/functional/inference_engine/transformations/gather_normalize_negative_indices_test.cpp

diff --git a/inference-engine/src/transformations/include/transformations/op_conversions/gather_normalize_negative_indices.hpp b/inference-engine/src/transformations/include/transformations/op_conversions/gather_normalize_negative_indices.hpp
new file mode 100644
index 00000000000000..1ec1ffe628ed1a
--- /dev/null
+++ b/inference-engine/src/transformations/include/transformations/op_conversions/gather_normalize_negative_indices.hpp
@@ -0,0 +1,29 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <transformations_visibility.hpp>
+#include <ngraph/pass/graph_rewrite.hpp>
+
+namespace ngraph {
+namespace pass {
+
+    class TRANSFORMATIONS_API GatherNegativeConstIndicesNormalize;
+
+}  // namespace pass
+}  // namespace ngraph
+
+/**
+ * @ingroup ie_transformation_common_api
+ * @brief GatherNegativeConstIndicesNormalize checks if indices value is negative scalar and
+ * normalizes it using ShapeOf->Add->Cast subgraph.
+ * We need to remove this transformation after adding support of negative indices in
+ * future version of Gather operation.
+ */
+class ngraph::pass::GatherNegativeConstIndicesNormalize : public ngraph::pass::MatcherPass {
+public:
+    NGRAPH_RTTI_DECLARATION;
+    GatherNegativeConstIndicesNormalize();
+};
diff --git a/inference-engine/src/transformations/src/transformations/common_optimizations/common_optimizations.cpp b/inference-engine/src/transformations/src/transformations/common_optimizations/common_optimizations.cpp
index 79f1dee888256f..4ab5cf1e80dc53 100644
--- a/inference-engine/src/transformations/src/transformations/common_optimizations/common_optimizations.cpp
+++ b/inference-engine/src/transformations/src/transformations/common_optimizations/common_optimizations.cpp
@@ -70,6 +70,7 @@
 #include "transformations/op_conversions/log_softmax_decomposition.hpp"
 #include "transformations/op_conversions/mvn6_decomposition.hpp"
 #include "transformations/op_conversions/simplify_ctc_greedy_decoder_seq_len.hpp"
+#include "transformations/op_conversions/gather_normalize_negative_indices.hpp"
 
 #include <ngraph/pass/manager.hpp>
 #include <ngraph/pass/constant_folding.hpp>
@@ -157,6 +158,7 @@ bool ngraph::pass::CommonOptimizations::run_on_function(std::shared_ptr<ngraph::
     decomp->add_matcher<ngraph::pass::MVN6Decomposition>();
     decomp->add_matcher<ngraph::pass::SimplifyCTCGreedyDecoderSeqLen>();
     decomp->add_matcher<ngraph::pass::EinsumDecomposition>();
+    decomp->add_matcher<ngraph::pass::GatherNegativeConstIndicesNormalize>();
     decomp->set_name("ngraph::pass::CommonDecompositions");
 
     // CF is required after all decompositions
diff --git a/inference-engine/src/transformations/src/transformations/op_conversions/gather_normalize_negative_indices.cpp b/inference-engine/src/transformations/src/transformations/op_conversions/gather_normalize_negative_indices.cpp
new file mode 100644
index 00000000000000..86713451869345
--- /dev/null
+++ b/inference-engine/src/transformations/src/transformations/op_conversions/gather_normalize_negative_indices.cpp
@@ -0,0 +1,77 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "transformations/op_conversions/gather_normalize_negative_indices.hpp"
+
+#include <memory>
+
+#include <ngraph/opsets/opset7.hpp>
+#include <ngraph/rt_info.hpp>
+#include <ngraph/pattern/op/wrap_type.hpp>
+#include "itt.hpp"
+
+NGRAPH_RTTI_DEFINITION(ngraph::pass::GatherNegativeConstIndicesNormalize, "GatherNegativeConstIndicesNormalize", 0);
+
+ngraph::pass::GatherNegativeConstIndicesNormalize::GatherNegativeConstIndicesNormalize() {
+    MATCHER_SCOPE(GatherNegativeConstIndicesNormalize);
+    auto data_input = ngraph::pattern::any_input(pattern::has_static_rank());
+    auto axis_input = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
+    auto indices_input = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
+    auto gather_node = std::make_shared<ngraph::opset7::Gather>(data_input, indices_input, axis_input);
+
+    ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher& m) {
+        auto& pattern_to_output = m.get_pattern_value_map();
+        auto gather = std::dynamic_pointer_cast<ngraph::opset7::Gather>(pattern_to_output.at(gather_node).get_node_shared_ptr());
+        auto data = pattern_to_output.at(data_input);
+        auto axis_constant = std::dynamic_pointer_cast<ngraph::opset7::Constant>(pattern_to_output.at(axis_input).get_node_shared_ptr());
+        auto indices_constant = std::dynamic_pointer_cast<ngraph::opset7::Constant>(pattern_to_output.at(indices_input).get_node_shared_ptr());
+
+        if (!gather || !axis_constant || !indices_constant) {
+            return false;
+        }
+
+        auto indices = indices_constant->cast_vector<int64_t>();
+        if (indices.size() != 1 || indices[0] >= 0) {
+            return false;
+        }
+
+        auto axis = axis_constant->cast_vector<int64_t>();
+        if (axis.size() != 1) {
+            return false;
+        }
+
+        auto axis_value = axis[0];
+
+        // normalize `axis` value if it is negative
+        if (axis_value < 0) {
+            axis_value = axis_value + data.get_partial_shape().rank().get_length();
+        }
+
+        if (data.get_partial_shape().rank().get_length() < axis_value) {
+            return false;
+        }
+
+        // check `axis` dimension of data tensor is static
+        if (!data.get_partial_shape()[axis_value].is_static()) {
+            return false;
+        }
+
+        auto input_type = indices_constant->get_element_type();
+        auto shape_of = std::make_shared<ngraph::opset7::ShapeOf>(data, input_type);
+        auto input_gather = std::make_shared<ngraph::opset7::Gather>(shape_of,
+            ngraph::opset7::Constant::create(input_type, Shape{}, {axis_value}), ngraph::opset7::Constant::create(input_type, Shape{}, {0}));
+
+        auto add = std::make_shared<ngraph::opset7::Add>(input_gather, indices_constant);
+        auto gather_new = gather_node->copy_with_new_inputs({data, add, axis_constant});
+        gather_new->set_friendly_name(gather->get_friendly_name());
+
+        ngraph::copy_runtime_info(gather, {shape_of, input_gather, add, gather_new});
+        ngraph::replace_node(gather, gather_new);
+
+        return true;
+    };
+
+    auto m = std::make_shared<ngraph::pattern::Matcher>(gather_node, matcher_name);
+    register_matcher(m, callback);
+}
diff --git a/inference-engine/tests/functional/inference_engine/transformations/gather_normalize_negative_indices_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/gather_normalize_negative_indices_test.cpp
new file mode 100644
index 00000000000000..ec6c4204a9b8a0
--- /dev/null
+++ b/inference-engine/tests/functional/inference_engine/transformations/gather_normalize_negative_indices_test.cpp
@@ -0,0 +1,306 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <gtest/gtest.h>
+
+#include <string>
+#include <memory>
+
+#include <ngraph/function.hpp>
+#include <ngraph/opsets/opset7.hpp>
+#include <ngraph/pass/manager.hpp>
+#include <transformations/op_conversions/gather_normalize_negative_indices.hpp>
+#include <transformations/init_node_info.hpp>
+
+#include "common_test_utils/ngraph_test_utils.hpp"
+
+using namespace testing;
+
+TEST(TransformationTests, GatherNegativeIndicesNormalize) {
+    std::shared_ptr<ngraph::Function> f(nullptr), f_ref(nullptr);
+    {
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::Shape{1, 15, 128});
+        auto indices = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {1});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis, 0);
+
+        f = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+
+        ngraph::pass::Manager manager;
+        manager.register_pass<ngraph::pass::InitNodeInfo>();
+        manager.register_pass<ngraph::pass::GatherNegativeConstIndicesNormalize>();
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+
+    {
+        auto indices_type = ngraph::element::i32;
+
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::Shape{1, 15, 128});
+        auto indices = ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {1});
+
+        auto shape_of = std::make_shared<ngraph::opset7::ShapeOf>(data, indices_type);
+        auto input_gather = std::make_shared<ngraph::opset7::Gather>(shape_of,
+            ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {1}), ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {0}));
+        auto add = std::make_shared<ngraph::opset7::Add>(input_gather, indices);
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, add, axis);
+
+        f_ref = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+    }
+
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, GatherNegativeIndicesNormalize_neg_axis) {
+    std::shared_ptr<ngraph::Function> f(nullptr), f_ref(nullptr);
+    {
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::Shape{1, 15, 128});
+        auto indices = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-2});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis, 0);
+
+        f = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+
+        ngraph::pass::Manager manager;
+        manager.register_pass<ngraph::pass::InitNodeInfo>();
+        manager.register_pass<ngraph::pass::GatherNegativeConstIndicesNormalize>();
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+
+    {
+        auto indices_type = ngraph::element::i32;
+
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::Shape{1, 15, 128});
+        auto indices = ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-2});
+
+        auto shape_of = std::make_shared<ngraph::opset7::ShapeOf>(data, indices_type);
+        auto input_gather = std::make_shared<ngraph::opset7::Gather>(shape_of,
+             ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {1}), ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {0}));
+        auto add = std::make_shared<ngraph::opset7::Add>(input_gather, indices);
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, add, axis);
+
+        f_ref = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+    }
+
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, GatherNegativeIndicesNormalize_dif_input_types) {
+    std::shared_ptr<ngraph::Function> f(nullptr), f_ref(nullptr);
+    {
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::Shape{1, 15, 128});
+        auto indices = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i64, ngraph::Shape{}, {1});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis, 0);
+
+        f = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+
+        ngraph::pass::Manager manager;
+        manager.register_pass<ngraph::pass::InitNodeInfo>();
+        manager.register_pass<ngraph::pass::GatherNegativeConstIndicesNormalize>();
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+
+    {
+        auto indices_type = ngraph::element::i32;
+
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::Shape{1, 15, 128});
+        auto indices = ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i64, ngraph::Shape{}, {1});
+
+        auto shape_of = std::make_shared<ngraph::opset7::ShapeOf>(data, indices_type);
+        auto input_gather = std::make_shared<ngraph::opset7::Gather>(shape_of,
+            ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {1}), ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {0}));
+        auto add = std::make_shared<ngraph::opset7::Add>(input_gather, indices);
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, add, axis);
+
+        f_ref = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+    }
+
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, GatherNegativeIndicesNormalize_static_axis_dim) {
+    std::shared_ptr<ngraph::Function> f(nullptr), f_ref(nullptr);
+    {
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::PartialShape{DYN, 15, DYN});
+        auto indices = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {1});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis, 0);
+
+        f = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+
+        ngraph::pass::Manager manager;
+        manager.register_pass<ngraph::pass::InitNodeInfo>();
+        manager.register_pass<ngraph::pass::GatherNegativeConstIndicesNormalize>();
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+
+    {
+        auto indices_type = ngraph::element::i32;
+
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::PartialShape{DYN, 15, DYN});
+        auto indices = ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {1});
+
+        auto shape_of = std::make_shared<ngraph::opset7::ShapeOf>(data, indices_type);
+        auto input_gather = std::make_shared<ngraph::opset7::Gather>(shape_of,
+            ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {1}), ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {0}));
+        auto add = std::make_shared<ngraph::opset7::Add>(input_gather, indices);
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, add, axis);
+
+        f_ref = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+    }
+
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, GatherNegativeIndicesNormalize_static_axis_dim_neg_axis) {
+    std::shared_ptr<ngraph::Function> f(nullptr), f_ref(nullptr);
+    {
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::PartialShape{DYN, 15, DYN});
+        auto indices = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-2});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis, 0);
+
+        f = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+
+        ngraph::pass::Manager manager;
+        manager.register_pass<ngraph::pass::InitNodeInfo>();
+        manager.register_pass<ngraph::pass::GatherNegativeConstIndicesNormalize>();
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+
+    {
+        auto indices_type = ngraph::element::i32;
+
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::PartialShape{DYN, 15, DYN});
+        auto indices = ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-2});
+
+        auto shape_of = std::make_shared<ngraph::opset7::ShapeOf>(data, indices_type);
+        auto input_gather = std::make_shared<ngraph::opset7::Gather>(shape_of,
+            ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {1}), ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {0}));
+        auto add = std::make_shared<ngraph::opset7::Add>(input_gather, indices);
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, add, axis);
+
+        f_ref = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+    }
+
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, GatherNegativeIndicesNormalize_non_static_axis_dim) {
+    std::shared_ptr<ngraph::Function> f(nullptr), f_ref(nullptr);
+    {
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::PartialShape{DYN, DYN, DYN});
+        auto indices = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {1});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis, 0);
+
+        f = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+
+        ngraph::pass::Manager manager;
+        manager.register_pass<ngraph::pass::InitNodeInfo>();
+        manager.register_pass<ngraph::pass::GatherNegativeConstIndicesNormalize>();
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+
+    {
+        auto indices_type = ngraph::element::i32;
+
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::PartialShape{DYN, DYN, DYN});
+        auto indices = ngraph::opset7::Constant::create(indices_type, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {1});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis);
+
+        f_ref = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+    }
+
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, GatherNegativeIndicesNormalize_positive_ind) {
+    std::shared_ptr<ngraph::Function> f(nullptr), f_ref(nullptr);
+    {
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::Shape{2, 3});
+        auto indices = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {0});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis, 0);
+
+        f = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+
+        ngraph::pass::Manager manager;
+        manager.register_pass<ngraph::pass::InitNodeInfo>();
+        manager.register_pass<ngraph::pass::GatherNegativeConstIndicesNormalize>();
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+
+    {
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::Shape{2, 3});
+        auto indices = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {0});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis);
+
+        f_ref = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+    }
+
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}
+
+TEST(TransformationTests, GatherNegativeIndicesNormalize_non_static_rank) {
+    std::shared_ptr<ngraph::Function> f(nullptr), f_ref(nullptr);
+    {
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::PartialShape::dynamic(ngraph::Rank::dynamic()));
+        auto indices = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {0});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis, 0);
+
+        f = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+
+        ngraph::pass::Manager manager;
+        manager.register_pass<ngraph::pass::InitNodeInfo>();
+        manager.register_pass<ngraph::pass::GatherNegativeConstIndicesNormalize>();
+        manager.run_passes(f);
+        ASSERT_NO_THROW(check_rt_info(f));
+    }
+
+    {
+        auto data = std::make_shared<ngraph::opset7::Parameter>(ngraph::element::f32, ngraph::PartialShape::dynamic());
+        auto indices = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {-1});
+        auto axis = ngraph::opset7::Constant::create(ngraph::element::i32, ngraph::Shape{}, {0});
+
+        auto gather = std::make_shared<ngraph::opset7::Gather>(data, indices, axis);
+
+        f_ref = std::make_shared<ngraph::Function>(ngraph::NodeVector{gather}, ngraph::ParameterVector{data});
+    }
+
+    auto res = compare_functions(f, f_ref);
+    ASSERT_TRUE(res.first) << res.second;
+}

From f4ba0f28a94d4ab809b4ce9404fd0f5f7dc43c66 Mon Sep 17 00:00:00 2001
From: Elizaveta Lobanova <elizaveta.lobanova@intel.com>
Date: Wed, 9 Jun 2021 18:15:39 +0300
Subject: [PATCH 048/178] [GNA] Convert Matmul with batch size > 8 to pointwise
 convolution (#5991)

* [GNA] Convert Matmul with batch size > 8 to pointwise convolution.
Support Eltwise split to more than 2 parts.
Fake Quantize support fixes.

* Put convolution split into a separate transformation

* Add separate transformations for cases with bias and fake quantize

* Rollback restriction for diagonal layer reshape
---
 .../gna_plugin/backend/gna_limitations.hpp    |   5 +
 .../gna_plugin/frontend/scale_factor_calc.hpp |   7 +-
 .../src/gna_plugin/gna_graph_compiler.cpp     |  40 +--
 .../src/gna_plugin/gna_graph_tools.hpp        |  18 +-
 .../src/gna_plugin/gna_groups.hpp             |  11 +-
 .../src/gna_plugin/gna_plugin.cpp             |  28 ++-
 .../gna_plugin/optimizer/gna_pass_manager.cpp | 106 +++++---
 ...onvert_matmul_to_pointwise_convolution.cpp | 180 ++++++++++++++
 ...onvert_matmul_to_pointwise_convolution.hpp |  71 ++++++
 ...lit_convolution_with_large_buffer_size.cpp | 131 ++++++++++
 ...lit_convolution_with_large_buffer_size.hpp |  34 +++
 .../convert_matmul_to_pointwise_conv.cpp      | 230 ++++++++++++++++++
 .../eltwise_split_over_channels_pass.cpp      |  26 +-
 13 files changed, 798 insertions(+), 89 deletions(-)
 create mode 100644 inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.cpp
 create mode 100644 inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.hpp
 create mode 100644 inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.cpp
 create mode 100644 inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.hpp
 create mode 100644 inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp

diff --git a/inference-engine/src/gna_plugin/backend/gna_limitations.hpp b/inference-engine/src/gna_plugin/backend/gna_limitations.hpp
index 9b0eccaea59714..3a283cae895d15 100644
--- a/inference-engine/src/gna_plugin/backend/gna_limitations.hpp
+++ b/inference-engine/src/gna_plugin/backend/gna_limitations.hpp
@@ -10,13 +10,18 @@
 namespace GNAPluginNS {
 namespace GNALimitations {
 
+constexpr uint32_t bufferMaxSize = 65528;
+
 constexpr uint32_t convMinFiltersNum = 4;
 constexpr uint32_t convMaxFiltersNum = 65532;
 constexpr uint32_t convFiltersNumDivider = 4;
+constexpr uint32_t convFilterMaxSize = 768;
 constexpr uint32_t convEachKernelByteAlignment = 16;
 constexpr uint32_t noOfInputsDivisor = 8;
 constexpr uint32_t noOfInputsLowPrecDivisor = 16;
 
+constexpr uint32_t affineMaxBatchSize = 8;
+
 namespace Cnn2D {
 struct RangeLimit {
     uint32_t min;
diff --git a/inference-engine/src/gna_plugin/frontend/scale_factor_calc.hpp b/inference-engine/src/gna_plugin/frontend/scale_factor_calc.hpp
index 3c1fdaac0e729b..11f13a7a9acad7 100644
--- a/inference-engine/src/gna_plugin/frontend/scale_factor_calc.hpp
+++ b/inference-engine/src/gna_plugin/frontend/scale_factor_calc.hpp
@@ -370,14 +370,8 @@ class ScaleFactorPerLayer<InferenceEngine::CNNLayer *> {
             auto minOutValue = quantizedParams->_dst_quant.GetMinValues().front();
             auto maxOutValue = quantizedParams->_dst_quant.GetMaxValues().front();
             auto absMax = std::max(std::abs(minOutValue), std::abs(maxOutValue));
-            auto absMin = std::min(std::abs(minOutValue), std::abs(maxOutValue));
 
             result = (quantizedParams->_dst_quant.GetLevels() - 1) / (maxOutValue - minOutValue);
-            if (0 && fp32eq(absMin, 0.0f) && !fp32eq(absMax, 0.0f)) {
-                result = (quantizedParams->_dst_quant.GetLevels() - 1) / (2 * absMax);
-            }
-            //
-            //result = MAX_VAL_2B_FEAT / absMax;
             if (std::isinf(result) || fp32eq(absMax, 0.0f)) {
                 result = max_activation_scale_factor;
             }
@@ -401,6 +395,7 @@ class ScaleFactorPerLayer<InferenceEngine::CNNLayer *> {
                 (layer.isIdentity() || layer.isFakeQuantize()) && LayerInfo(prevLayer).isWeightableIdentity()) {
                 auto prevLayerQuant = InferenceEngine::getInjectedData<QuantizedLayerParams>(*prevLayer);
                 if (!fp32eq(prevLayerQuant->_src_quant.GetScale(), 1.0f) &&
+                    prevLayerQuant->_src_quant.IsStatsSet() &&
                     (prevLayer2 == nullptr || LayerInfo(prevLayer2).has8BOr16BOutput())) {
                     result = prevLayerQuant->_src_quant.GetScale();
                     usePrevScaleFactor = true;
diff --git a/inference-engine/src/gna_plugin/gna_graph_compiler.cpp b/inference-engine/src/gna_plugin/gna_graph_compiler.cpp
index 23685b4734faf9..bf44e437af0e6a 100644
--- a/inference-engine/src/gna_plugin/gna_graph_compiler.cpp
+++ b/inference-engine/src/gna_plugin/gna_graph_compiler.cpp
@@ -158,25 +158,27 @@ void GNAGraphCompiler::fillSplitConnections(InferenceEngine::CNNLayerPtr layer)
                 THROW_GNA_LAYER_EXCEPTION(layer) << " outData["<< i << "]" << " connected by " << j <<" connection doesnt connect to functional layer";
             }
 
-            auto dataOutput = outFunctionalLayer.first->insData[outFunctionalLayer.second].lock();
-
-            padding = std::max(padding, LayerInfo(outFunctionalLayer.first).paddingSize())
-                                                        * dataOutput->getPrecision().size();
-            output_layer_size =
-                    InferenceEngine::details::product(begin(dataOutput->getDims()),
-                                                     end(dataOutput->getDims())) * dataOutput->getPrecision().size();
-
-            if (LayerInfo(outFunctionalLayer.first).isAffineFilter()) {
-                size_t aligned64_offset = outFunctionalLayer.first->GetParamAsInt("offset");
-                layerInfoItem.splitOutputLayers.emplace_back(
-                    outFunctionalLayer.first,
-                    outFunctionalLayer.second,
-                    aligned64_offset * dataOutput->getPrecision().size(),
-                    output_layer_size);
-            } else {
-                layerInfoItem.splitOutputLayers.emplace_back(
-                    outFunctionalLayer.first, outFunctionalLayer.second, split_size, output_layer_size);
-            }
+            for (int idx : outFunctionalLayer.second) {
+                auto dataOutput = outFunctionalLayer.first->insData[idx].lock();
+
+                padding = std::max(padding, LayerInfo(outFunctionalLayer.first).paddingSize())
+                                                            * dataOutput->getPrecision().size();
+                output_layer_size =
+                        InferenceEngine::details::product(begin(dataOutput->getDims()),
+                                                        end(dataOutput->getDims())) * dataOutput->getPrecision().size();
+
+                if (LayerInfo(outFunctionalLayer.first).isAffineFilter()) {
+                    size_t aligned64_offset = outFunctionalLayer.first->GetParamAsInt("offset");
+                    layerInfoItem.splitOutputLayers.emplace_back(
+                        outFunctionalLayer.first,
+                        idx,
+                        aligned64_offset * dataOutput->getPrecision().size(),
+                        output_layer_size);
+                } else {
+                    layerInfoItem.splitOutputLayers.emplace_back(
+                        outFunctionalLayer.first, idx, split_size, output_layer_size);
+                }
+             }
         }
 
         // in case of unconnected split - we need properly increment size
diff --git a/inference-engine/src/gna_plugin/gna_graph_tools.hpp b/inference-engine/src/gna_plugin/gna_graph_tools.hpp
index e9cf70790acdb9..51701268209728 100644
--- a/inference-engine/src/gna_plugin/gna_graph_tools.hpp
+++ b/inference-engine/src/gna_plugin/gna_graph_tools.hpp
@@ -155,14 +155,14 @@ inline InferenceEngine::CNNLayerPtr  CNNNetPrevLayerSkipCertain(Layer layer, int
  */
 
 template <class Layer>
-inline std::pair<InferenceEngine::CNNLayerPtr, int>  CNNNetCheckNextLayerSkipCertain(Layer layer, int oidx, int iidx, bool bOnlyCheck,
+inline std::pair<InferenceEngine::CNNLayerPtr, std::vector<int>>  CNNNetCheckNextLayerSkipCertain(Layer layer, int oidx, int iidx, bool bOnlyCheck,
                                                                 const std::function<bool(CNNLayerPtr)> &shouldSkip) {
     if (oidx >= layer->outData.size()) {
-        if (bOnlyCheck) return {nullptr, 0};
+        if (bOnlyCheck) return {nullptr, {}};
         THROW_GNA_LAYER_EXCEPTION(layer) << " no next output layer for outdata: " << oidx;
     }
     if (getInputTo(layer->outData[oidx]).empty() || iidx >= getInputTo(layer->outData[oidx]).size()) {
-        if (bOnlyCheck) return {nullptr, 0};
+        if (bOnlyCheck) return {nullptr, {}};
         THROW_GNA_LAYER_EXCEPTION(layer) << " no next output layer for outdata: " << oidx << " and inputTo index: " << iidx;
     }
 
@@ -174,12 +174,12 @@ inline std::pair<InferenceEngine::CNNLayerPtr, int>  CNNNetCheckNextLayerSkipCer
 
     while (shouldSkip(outLayer->second)) {
         if (outLayer->second->outData.size() <= new_oidx) {
-            if (bOnlyCheck) return { nullptr, 0 };
+            if (bOnlyCheck) return { nullptr, {} };
             THROW_GNA_LAYER_EXCEPTION(outLayer->second) << " no next output layer for outdata: " << new_oidx;
         }
 
         if (getInputTo(outLayer->second->outData[new_oidx]).size() <= new_iidx) {
-            if (bOnlyCheck) return { nullptr, 0 };
+            if (bOnlyCheck) return { nullptr, {} };
             THROW_GNA_LAYER_EXCEPTION(outLayer->second) << " no next output layer for outdata: " << new_oidx << " and inputTo index: " << new_iidx;
         }
 
@@ -188,11 +188,7 @@ inline std::pair<InferenceEngine::CNNLayerPtr, int>  CNNNetCheckNextLayerSkipCer
     }
 
     auto insDataIdx = CNNLayerFindInsDataIdxes(layer->outData[new_oidx], outLayer->second);
-    if (insDataIdx.size() != 1) {
-        if (bOnlyCheck) return { nullptr, 0 };
-        THROW_GNA_LAYER_EXCEPTION(layer) << " has multiple connection to " << new_oidx << " outData";
-    }
-    return { outLayer->second, insDataIdx.front() };
+    return { outLayer->second, insDataIdx };
 }
 
 /**
@@ -256,7 +252,7 @@ inline std::pair<InferenceEngine::CNNLayerPtr, int>  CNNNetCheckNextLayerSkipCer
 
 /// @brief alias for strict checkNextLayer (false)
 template <class Layer>
-inline std::pair<InferenceEngine::CNNLayerPtr, int>  CNNNetGetNextLayerSkipCertain(Layer layer, int oidx, int iidx,
+inline std::pair<InferenceEngine::CNNLayerPtr, std::vector<int>>  CNNNetGetNextLayerSkipCertain(Layer layer, int oidx, int iidx,
                                                                                const std::function<bool(CNNLayerPtr)> &shouldSkip) {
     return CNNNetCheckNextLayerSkipCertain(layer, oidx, iidx, false, shouldSkip);
 }
diff --git a/inference-engine/src/gna_plugin/gna_groups.hpp b/inference-engine/src/gna_plugin/gna_groups.hpp
index 21abe5d01246f4..2449338821c109 100644
--- a/inference-engine/src/gna_plugin/gna_groups.hpp
+++ b/inference-engine/src/gna_plugin/gna_groups.hpp
@@ -46,14 +46,10 @@ inline InferenceEngine::DataPtr Get2DReshapedData(InferenceEngine::DataPtr input
  * @param layer
  */
 inline bool HasTo2DReshapeData(InferenceEngine::CNNLayerPtr layer) {
-    if (GNAPluginNS::LayerInfo(layer).isPower())
+    if (GNAPluginNS::LayerInfo(layer).isPower() || GNAPluginNS::LayerInfo(layer).isCopy())
         return true;
 
-    if (!GNAPluginNS::LayerInfo(layer).isScaleShift())
-        return false;
-
-    // Don't reshape user-defined ScaleShift layers
-    if (layer->name.rfind("SyntheticScaleShift", 0) == std::string::npos)
+    if (!GNAPluginNS::LayerInfo(layer).isSyntheticScaleShift())
         return false;
 
     // Don't reshape the first dnn layer since it breaks groups recognition
@@ -61,8 +57,7 @@ inline bool HasTo2DReshapeData(InferenceEngine::CNNLayerPtr layer) {
         return LayerInfo(ptr).isNonValuesChangable();
     });
     IE_ASSERT(prevLayer != nullptr);
-    if (LayerInfo(prevLayer).isInput())
-        return false;
+    if (LayerInfo(prevLayer).isInput()) return false;
 
     // Don't reshape diagonallayers with bias connection
     return !GNAPluginNS::LayerInfo(getCreatorLayer(layer->insData.front().lock()).lock()).has32BOutput();
diff --git a/inference-engine/src/gna_plugin/gna_plugin.cpp b/inference-engine/src/gna_plugin/gna_plugin.cpp
index e76eafa6d5372a..f49d543def1f37 100644
--- a/inference-engine/src/gna_plugin/gna_plugin.cpp
+++ b/inference-engine/src/gna_plugin/gna_plugin.cpp
@@ -54,12 +54,17 @@
 #include <transformations/common_optimizations/pull_transpose_through_fq.hpp>
 #include <transformations/common_optimizations/relu_fake_quantize_fusion.hpp>
 #include <transformations/common_optimizations/add_fake_quantize_fusion.hpp>
+#include <transformations/utils/utils.hpp>
 
 #include "transformations/remove_extra_reshapes.hpp"
 #include "transformations/insert_transpose_after_convolution_or_pooling.hpp"
 #include "transformations/insert_transpose_before_matmul.hpp"
 #include "transformations/reorder_activation_and_pooling.hpp"
 #include "transformations/swap_input_matmul_gna.hpp"
+#include "transformations/convert_matmul_to_pointwise_convolution.hpp"
+#include "transformations/split_convolution_with_large_buffer_size.hpp"
+
+#include <ngraph/opsets/opset7.hpp>
 
 #if GNA_LIB_VER == 2
 #include <gna2-model-api.h>
@@ -667,6 +672,15 @@ void GNAPlugin::LoadNetwork(CNNNetwork & _network) {
         // WA: ConvertPriorBox must be executed before the 1st ConstantFolding pass
         manager.register_pass<ngraph::pass::ConvertPriorBox>();
         manager.register_pass<ngraph::pass::CommonOptimizations>();
+        // TODO enable this transformation for networks with convolutions
+        if (!ngraph::op::util::has_op_with_type<ngraph::opset7::Convolution>(graph)) {
+            manager.register_pass<ConvertMatmulWithFqToPointWiseConvolution>();
+            manager.register_pass<ConvertMatmulWithBiasToPointWiseConvolution>();
+            manager.register_pass<ConvertMatmulToPointWiseConvolution>();
+        }
+        manager.register_pass<SplitConvolutionWithFq>();
+        manager.register_pass<SplitConvolutionWithBias>();
+        manager.register_pass<SplitConvolution>();
         manager.register_pass<InsertTransposeBeforeMatmul>();
         manager.register_pass<SwapInputMatMul>();
         manager.register_pass<InsertTransposeAfterConvOrPool>();
@@ -735,6 +749,7 @@ void GNAPlugin::LoadNetwork(CNNNetwork & _network) {
         passes->registerPass<SubstitutePReluPass>();
         passes->registerPass<SubstituteSoftSignPass>();
 
+        passes->registerPass<BroadcastConstPass>();
         passes->registerPass<ReorderMaxPoolPass>();
         passes->registerPass<EltwiseSplitOverChannelsPass>();
         passes->registerPass<InsertSplitAligningFilterPass>();
@@ -753,7 +768,6 @@ void GNAPlugin::LoadNetwork(CNNNetwork & _network) {
 
         passes->registerPass<InsertIdentityLayerPass>();
         passes->registerPass<BreakFusingOfOutputLayersPass>();
-        passes->registerPass<BroadcastConstPass>();
         passes->registerPass<InsertDiagonalLayerPass>();
         passes->registerPass<HandleMultipleActivationsForTheLayerPass>();
 #if GNA_LIB_VER == 2
@@ -1465,7 +1479,11 @@ static InferenceEngine::Layout GetLayoutForDims(const InferenceEngine::SizeVecto
 Blob::Ptr GNAPlugin::GetOutputBlob(const std::string& name, InferenceEngine::Precision precision) {
     // need to have intermediate blob for interleave conversion
     InferenceEngine::Blob::Ptr outputBlob;
-    auto outputDims = outputsDataMap[name]->getTensorDesc().getDims();
+    auto outputDataIt = outputsDataMap.find(name);
+    if (outputDataIt == std::end(outputsDataMap)) {
+        THROW_GNA_EXCEPTION << "Output " << name << " isn't found";
+    }
+    auto outputDims = outputDataIt->second->getTensorDesc().getDims();
     outputBlob = make_blob_with_precision(TensorDesc(precision, outputDims, GetLayoutForDims(outputDims)));
     outputBlob->allocate();
     return outputBlob;
@@ -1475,7 +1493,11 @@ Blob::Ptr GNAPlugin::GetInputBlob(const std::string& name, InferenceEngine::Prec
     InferenceEngine::Blob::Ptr inputBlob;
     // need to have intermediate blob for interleave conversion
     // TODO: NCHW format support is experimental = c++ MO did insert reshape, while TF mo - not
-    auto inputDims = inputsDataMap[name]->getTensorDesc().getDims();
+    auto inputDataIt = inputsDataMap.find(name);
+    if (inputDataIt == std::end(inputsDataMap)) {
+        THROW_GNA_EXCEPTION << "Input " << name << " isn't found";
+    }
+    auto inputDims = inputDataIt->second->getTensorDesc().getDims();
     inputBlob = make_blob_with_precision(TensorDesc(precision, inputDims, GetLayoutForDims(inputDims)));
     inputBlob->allocate();
     return inputBlob;
diff --git a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
index 4d3b71b9622237..b8962cebd36e23 100644
--- a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
+++ b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
@@ -86,7 +86,7 @@ static void insertDiagonalLayerBetween(InferenceEngine::CNNLayerPtr prevLayer,
     });
     IE_ASSERT(inputLayer != nullptr);
     size_t weightsSize = (LayerInfo(prevLayer).has32BOutput() || LayerInfo(inputLayer).isInput()) ?
-                         weightsSize = nextLayer->outData[0]->getDims().back() :
+                         nextLayer->outData[0]->getDims().back() :
                          Get2DReshapedData(nextLayer->outData[0], 8)->getDims()[1];
     std::vector<float> weightsValues(weightsSize, fillValue);
     IE_ASSERT(diagLayer != nullptr);
@@ -314,6 +314,7 @@ void HandleMultipleActivationsForTheLayerPass::run() {
                 LayerInfo info(inputTo.second);
 
                 if (info.isActivation()) {
+                    if (odata->getDims().empty()) continue;
                     if (!activations.empty() && odata->getDims()[0] != 1) {
                         THROW_GNA_EXCEPTION << "Unsupported batch size " << odata->getDims()[0]
                                             << " for diagonal layer insertion";
@@ -741,12 +742,17 @@ void RemovePermutationsNHWCToNCHWPass::run() {
         IE_ASSERT(!input_to.empty());
         auto current_layer = input_to.begin()->second;
         setNHWCOrder(current_layer->input());
-        while (current_layer != pattern_end) {
-            setNHWCOrder(current_layer->outData[0]);
-            input_to = getInputTo(current_layer->outData[0]);
-            IE_ASSERT(!input_to.empty());
-            current_layer = input_to.begin()->second;
-        }
+        std::function<void(CNNLayerPtr)> propogateNHWCOrderRecursive =
+            [pattern_end, &propogateNHWCOrderRecursive, &setNHWCOrder](CNNLayerPtr current_layer) {
+            if (current_layer == pattern_end) return;
+            for (size_t i = 0; i < current_layer->outData.size(); ++i) {
+                setNHWCOrder(current_layer->outData[i]);
+                auto input_to = getInputTo(current_layer->outData[i]);
+                IE_ASSERT(!input_to.empty());
+                propogateNHWCOrderRecursive(input_to.begin()->second);
+            }
+        };
+        propogateNHWCOrderRecursive(current_layer);
 
         if (LayerInfo(pattern_start).isPermute() && !getInputTo(pattern_start->outData.front()).empty()) {
             auto layer_before_permute = CNNNetPrevLayer(pattern_start);
@@ -1447,21 +1453,19 @@ void EltwiseSplitOverChannelsPass::run() {
             THROW_GNA_LAYER_EXCEPTION(l) << "number of outputs expected to be 1";
         }
         auto oData = l->outData.front();
+        auto out_width = GetDataDimSize(oData, DataDimName::W);
         auto totalElementsForOutput = details::product(oData->getDims().begin(), oData->getDims().end());
         auto maxAffineElements = getPassManager()->getPolicy().GNAAffineDiagonalPolicy.limitedTo;
         if (totalElementsForOutput <= maxAffineElements) {
             continue;
         }
 
-        // TODO: for now lets put split of 2 elements as restrictions
         auto totalSplits = 1 + totalElementsForOutput / maxAffineElements;
-        if (totalSplits > 2) {
-            THROW_GNA_LAYER_EXCEPTION(l) << "split layer over output channels on more than 2 layers unsupported";
-        }
 
         pass_trace() << "transforming " << LAYER_NAME(l) << " by splitting it to multiple eltwise operations\n";
         auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(l);
 
+        bool sameInputs = l->insData[0].lock() == l->insData[1].lock();
         std::vector<CNNLayerPtr> splitLayers(2);
         for (size_t kThEltwiseInput = 0; kThEltwiseInput != 2; kThEltwiseInput++) {
             // create split layer
@@ -1472,31 +1476,38 @@ void EltwiseSplitOverChannelsPass::run() {
 
             split->insData.push_back(l->insData[kThEltwiseInput]);
             auto inputDesc = l->insData[kThEltwiseInput].lock()->getTensorDesc();
-            // need to split this desc
-            if (inputDesc.getLayout() != Layout::NC) {
-                THROW_GNA_LAYER_EXCEPTION(l)
-                << "cannot split over channel: input " << std::to_string(kThEltwiseInput)
-                << " layout need to be NC";
-            }
 
             // create split layer outputs
-            for (size_t i = 0;; i++) {
-                auto elements_num = std::min(totalElementsForOutput - i * maxAffineElements,
+            size_t usedElements = 0;
+            for (size_t i = 0; i < totalSplits; i++) {
+                SizeVector newDims;
+                size_t elements_num = std::min(totalElementsForOutput - usedElements,
                         static_cast<size_t>(maxAffineElements));
+                if (inputDesc.getDims().size() == 2) {
+                    newDims = SizeVector{1, elements_num};
+                } else {
+                    elements_num = elements_num - elements_num % out_width;
+                    newDims = SizeVector{1, elements_num / out_width, out_width};
+                }
 
-                SizeVector newDims = {1, elements_num};
                 auto newDesc = TensorDesc(inputDesc.getPrecision(), newDims, inputDesc.getLayout());
                 auto data = std::make_shared<Data>(l->name + "/" + std::to_string(kThEltwiseInput) + "/1", newDesc);
                 getCreatorLayer(data) = split;
                 split->outData.push_back(data);
 
-                if (elements_num != maxAffineElements) {
+                usedElements += elements_num;
+                if (usedElements == totalElementsForOutput) {
                     break;
                 }
             }
             // replacing connection X->eltwise to X->split
             auto oData = CNNLayerFindOutData(l, kThEltwiseInput);
             oData.second->second = split;
+
+            if (sameInputs) {
+                splitLayers[1] = splitLayers[0];
+                break;
+            }
         }
 
         // create concatlayer
@@ -1507,8 +1518,6 @@ void EltwiseSplitOverChannelsPass::run() {
         concat->outData.push_back(masterEltwise->outData.front());
         getCreatorLayer(masterEltwise->outData.front()) = concat;
 
-
-        // create new eltwise layers - here 2 hardcode
         for (size_t k = 0; k != totalSplits; k++) {
             auto eltwiseRaw = std::make_shared<EltwiseLayer>(
                     LayerParams{l->name + "/eltwise/" + std::to_string(k), "Eltwise", Precision::FP32});
@@ -1517,7 +1526,6 @@ void EltwiseSplitOverChannelsPass::run() {
             eltwiseRaw->coeff = masterEltwise->coeff;
             auto eltwise = quantized ? InferenceEngine::injectData<QuantizedLayerParams>(eltwiseRaw) : eltwiseRaw;
 
-
             eltwise->insData.push_back(splitLayers[0]->outData[k]);
             eltwise->insData.push_back(splitLayers[1]->outData[k]);
             getInputTo(splitLayers[0]->outData[k])[eltwise->name] = eltwise;
@@ -1529,6 +1537,15 @@ void EltwiseSplitOverChannelsPass::run() {
             auto data = std::make_shared<Data>(l->name + "/elwise/out/" + std::to_string(k), newDesc);
             getCreatorLayer(data) = eltwise;
             eltwise->outData.push_back(data);
+            if (quantized) {
+                auto eltwiseQuant = InferenceEngine::getInjectedData<QuantizedLayerParams>(eltwise);
+                if (quantized->_src_quant.IsStatsSet()) {
+                    eltwiseQuant->_src_quant.CopyStats(quantized->_src_quant);
+                }
+                if (quantized->_dst_quant.IsStatsSet()) {
+                    eltwiseQuant->_dst_quant.CopyStats(quantized->_dst_quant);
+                }
+            }
             getInputTo(data)[concat->name] = concat;
             concat->insData.push_back(data);
         }
@@ -1919,13 +1936,20 @@ void FuseFQIntoWeightsPass::run() {
         }
 
         GNAFakeQuantizeLayer gnaFakeQuantizeLayer(fqLayer);
-        size_t layers_connected_to_fq_count = getInputTo(fqLayer->outData[0]).size();
+        auto inputTo = getInputTo(fqLayer->outData[0]);
+        size_t layers_connected_to_fq_count = inputTo.size();
+        auto layerBeforeWeightable = fqLayer;
+        while (layers_connected_to_fq_count == 1 && LayerInfo(inputTo.begin()->second).isNonFunctional()) {
+            layerBeforeWeightable = inputTo.begin()->second;
+            inputTo = getInputTo(layerBeforeWeightable->outData[0]);
+            layers_connected_to_fq_count = inputTo.size();
+        }
         for (int index = 0; index < layers_connected_to_fq_count; index++) {
-            auto weightableLayer = CNNNetGetNextLayerSkipCertain(fqLayer, 0, index, isNonFunctional).first;
+            auto weightableLayer = CNNNetGetNextLayerSkipCertain(layerBeforeWeightable, 0, index, isNonFunctional).first;
             if (!LayerInfo(weightableLayer).isWeightable()) {
                 continue;
             }
-            if (weightableLayer->insData.size() != 3) {
+            if (weightableLayer->insData.size() < 2) {
                 continue;
             }
 
@@ -1942,7 +1966,8 @@ void FuseFQIntoWeightsPass::run() {
             pass_trace() << "found " << LAYER_NAME(fqLayer) << " that will be converted to weights of "
                 << LAYER_NAME(weightableLayer) << "\n";
 
-            auto biases = LayerUtils::getParamFromInputAsBlob(weightableLayer, biasesIdx);
+            auto biases = weightableLayer->insData.size() == 3 ?
+                LayerUtils::getParamFromInputAsBlob(weightableLayer, biasesIdx) : nullptr;
             auto quantizedWeights = gnaFakeQuantizeLayer.getConstInputData();
 
             // 1. broke existing connections - by detaching fq subgraph from rest of graph
@@ -2149,8 +2174,11 @@ void MoveFakeQuantizeLayerIntoQuantParamsPass :: run() {
         }
         GNAFakeQuantizeLayer fqLayer(l);
         auto prevLayer = CNNNetPrevLayerSkipCertain(*fqLayer, 0, donotSkip);
-        if (prevLayer->outData.size() != 1) {
-            THROW_GNA_LAYER_EXCEPTION(prevLayer) << " fake quantize input that connected to something else not supported";
+        auto prevDataIt = std::find_if(std::begin(prevLayer->outData), std::end(prevLayer->outData), [l](DataPtr data) {
+            return getInputTo(data).find(l->name) != std::end(getInputTo(data));
+        });
+        if (prevDataIt == std::end(prevLayer->outData)) {
+            THROW_GNA_LAYER_EXCEPTION(fqLayer) << "Invalid connection between " << prevLayer->name << " and " << l->name;
         }
 
         auto inputRange = fqLayer.getInputRange();
@@ -2181,8 +2209,18 @@ void MoveFakeQuantizeLayerIntoQuantParamsPass :: run() {
         quantParamsPrevLayer->_dst_quant.SetMinValues({ outputRange.first[0] }, false);
         quantParamsPrevLayer->_dst_quant.SetMaxValues({ outputRange.second[0] }, false);
 
+        // Propogate destination statistics to multiply layer if it's set for the next sum/sub layer (is considered as bias)
+        if (LayerInfo(prevLayer).isEltwiseSum() || LayerInfo(prevLayer).isEltwiseSub()) {
+            auto eltwPrevLayer = CNNNetPrevLayerSkipCertain(prevLayer, 0, donotSkip);
+            auto constLayer = CNNNetPrevLayerSkipCertain(prevLayer, 1, donotSkip);
+            if (LayerInfo(eltwPrevLayer).isEltwise() && LayerInfo(constLayer).isConst()) {
+                auto quantParamsEltwLayer = InferenceEngine::getInjectedData<QuantizedLayerParams>(eltwPrevLayer);
+                quantParamsEltwLayer->_dst_quant.CopyStats(quantParamsPrevLayer->_dst_quant);
+            }
+        }
+
         auto fqQauntParams = InferenceEngine::getInjectedData<QuantizedLayerParams>(l);
-        fqQauntParams->_dst_quant.SetLevels(fqLevels);
+        fqQauntParams->_dst_quant.SetLevels(UINT16_MAX);
         fqQauntParams->_dst_quant.SetMinValues({ inputRange.first[0] }, true);
         fqQauntParams->_dst_quant.SetMaxValues({ inputRange.second[0] }, true);
         fqQauntParams->_dst_quant.SetMinValues({ outputRange.first[0] }, false);
@@ -2198,7 +2236,7 @@ void MoveFakeQuantizeLayerIntoQuantParamsPass :: run() {
         // FQ Layer is fused only when previous layer is const, memory or activation layer
         // or a next layer is activation layer.
         bool isFQFuseAllowed = allowFQFuse(l);
-        auto prevData = prevLayer->outData.front();
+        auto prevData = *prevDataIt;
 
         // Find all output layers connected to FQ
         auto nextLayers = CNNNetGetAllNextLayersSkipCertain(*fqLayer, -1, donotSkip);
@@ -2207,7 +2245,7 @@ void MoveFakeQuantizeLayerIntoQuantParamsPass :: run() {
         }
 
         if (isFQFuseAllowed) {
-            getInputTo(prevLayer->outData.front()).clear();
+            getInputTo(prevData).clear();
         }
 
         // Connect all next layers after FQ to the layer that is before FQ
@@ -2222,7 +2260,7 @@ void MoveFakeQuantizeLayerIntoQuantParamsPass :: run() {
                 for (int insDataIdx : insDatas) {
                     nextLayers[i]->insData[insDataIdx] = prevData;
                 }
-                getInputTo(prevLayer->outData.front())[nextLayers[i]->name] = nextLayers[i];
+                getInputTo(prevData)[nextLayers[i]->name] = nextLayers[i];
             }
 
             propagateStatistics(quantParamsPrevLayer, nextLayers[i]);
diff --git a/inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.cpp b/inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.cpp
new file mode 100644
index 00000000000000..da7e6279624c81
--- /dev/null
+++ b/inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.cpp
@@ -0,0 +1,180 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "transformations/convert_matmul_to_pointwise_convolution.hpp"
+
+#include <ngraph/opsets/opset7.hpp>
+#include <ngraph/pattern/op/or.hpp>
+#include <ngraph/pattern/op/wrap_type.hpp>
+
+#include "layers/gna_permute.hpp"
+#include "backend/gna_limitations.hpp"
+
+using namespace GNAPluginNS;
+
+NGRAPH_RTTI_DEFINITION(ConvertMatmulToPointWiseConvolution, "ConvertMatmulToPointWiseConvolution", 0);
+NGRAPH_RTTI_DEFINITION(ConvertMatmulWithBiasToPointWiseConvolution, "ConvertMatmulWithBiasToPointWiseConvolution", 0);
+NGRAPH_RTTI_DEFINITION(ConvertMatmulWithFqToPointWiseConvolution, "ConvertMatmulWithFqToPointWiseConvolution", 0);
+
+static std::tuple<bool, uint32_t, uint32_t, uint32_t> VerifyAndGetConvParams(std::shared_ptr<ngraph::Node> matmul_node) {
+    auto input1_shape = matmul_node->get_input_shape(0);
+    auto input2_shape = matmul_node->get_input_shape(1);
+    auto output_shape = matmul_node->get_output_shape(0);
+    if (input1_shape.size() == 3 && input1_shape.front() == 1) {
+        input1_shape.erase(std::begin(input1_shape));
+    }
+
+    if (input1_shape.size() != 2 || input2_shape.size() != 2 || output_shape.size() < 2) {
+        return std::make_tuple(false, 0, 0, 0);
+    }
+
+    // Check if MatMul or corresponding pointwise convolution are supported by GNA
+    const uint32_t width = input1_shape.front();
+    const uint32_t in_channels = input2_shape.back();
+    const uint32_t out_channels = input2_shape.front();
+    if (input1_shape.front() <= GNALimitations::affineMaxBatchSize ||
+        out_channels % GNALimitations::convFiltersNumDivider != 0 ||
+        out_channels > GNALimitations::convMaxFiltersNum ||
+        in_channels > GNALimitations::convFilterMaxSize) {
+        return std::make_tuple(false, 0, 0, 0);
+    }
+
+    return std::make_tuple(true, width, in_channels, out_channels);
+}
+
+static bool Convert(std::shared_ptr<ngraph::Node> matmul_node,
+                    std::shared_ptr<ngraph::Node> add,
+                    std::shared_ptr<ngraph::Node> bias,
+                    std::shared_ptr<ngraph::Node> fq) {
+    bool supported;
+    uint32_t width, in_channels, out_channels;
+    std::tie(supported, width, in_channels, out_channels) = VerifyAndGetConvParams(matmul_node);
+    if (!supported) return false;
+
+    auto input_node = matmul_node->input_value(0).get_node_shared_ptr();
+    auto weights_node = matmul_node->input_value(1).get_node_shared_ptr();
+    auto base_name = matmul_node->get_friendly_name();
+
+    auto reshape_const_before = std::make_shared<ngraph::opset7::Constant>(ngraph::element::Type_t::i64,
+                                                                            ngraph::Shape{4},
+                                                                            ngraph::Shape{1, 1, width, in_channels});
+    auto reshape_before =  std::make_shared<ngraph::opset7::Reshape>(input_node, reshape_const_before, false);
+    reshape_before->set_friendly_name(base_name + "/reshape_in");
+
+    auto transpose_before = std::make_shared<ngraph::opset7::Transpose>(reshape_before,
+        ngraph::opset7::Constant::create(ngraph::element::i64, ngraph::Shape{4},
+        GetPermuteOrder(InferenceEngine::Layout::NHWC, InferenceEngine::Layout::NCHW)));
+    transpose_before->set_friendly_name(base_name + "/transpose_in");
+
+    auto weights_reshape_const = std::make_shared<ngraph::opset7::Constant>(ngraph::element::Type_t::i64,
+        ngraph::Shape{4}, ngraph::Shape{out_channels, in_channels, 1, 1});
+    auto weights_reshaped =  std::make_shared<ngraph::opset7::Reshape>(weights_node, weights_reshape_const, false);
+
+    std::shared_ptr<ngraph::Node> conv_node = std::make_shared<ngraph::opset7::Convolution>(transpose_before, weights_reshaped,
+            ngraph::Strides{1, 1}, ngraph::CoordinateDiff{0, 0}, ngraph::CoordinateDiff{0, 0},
+            ngraph::Strides{1, 1}, ngraph::op::PadType::VALID);
+    conv_node->set_friendly_name(base_name + "/conv");
+
+    std::shared_ptr<ngraph::Node> root_node = matmul_node;
+    if (bias != nullptr) {
+         conv_node = std::make_shared<ngraph::opset7::Add>(conv_node, bias);
+         root_node = add;
+    }
+
+    if (fq != nullptr) {
+        conv_node = fq->clone_with_new_inputs({conv_node, fq->input_value(1), fq->input_value(2),
+            fq->input_value(3), fq->input_value(4)});
+        root_node = fq;
+    }
+
+    auto transpose_after = std::make_shared<ngraph::opset7::Transpose>(conv_node,
+        ngraph::opset7::Constant::create(ngraph::element::i64, ngraph::Shape{4},
+        GetPermuteOrder(InferenceEngine::Layout::NCHW, InferenceEngine::Layout::NHWC)));
+    transpose_after->set_friendly_name(base_name + "/transpose_out");
+
+    auto output_shape = matmul_node->get_output_shape(0);
+    output_shape[output_shape.size() - 1] = out_channels;
+    output_shape[output_shape.size() - 2] = width;
+    auto reshape_const_after = std::make_shared<ngraph::opset7::Constant>(ngraph::element::Type_t::i64,
+                                                                            ngraph::Shape{output_shape.size()},
+                                                                            output_shape);
+    auto reshape_after =  std::make_shared<ngraph::opset7::Reshape>(transpose_after, reshape_const_after, false);
+    reshape_after->set_friendly_name(base_name);
+
+    ngraph::replace_node(root_node, reshape_after);
+    return true;
+}
+
+ConvertMatmulToPointWiseConvolution::ConvertMatmulToPointWiseConvolution() {
+    auto const_input = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
+    auto const_fq = ngraph::pattern::wrap_type<ngraph::opset7::FakeQuantize>({const_input,
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>()});
+    auto second_input = std::make_shared<ngraph::pattern::op::Or>(ngraph::OutputVector{const_input, const_fq});
+    auto matmul = ngraph::pattern::wrap_type<ngraph::opset7::MatMul>({ngraph::pattern::any_input(), second_input});
+
+    ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher &m) {
+        const auto& pattern_map = m.get_pattern_value_map();
+        return Convert(pattern_map.at(matmul).get_node_shared_ptr(), nullptr, nullptr, nullptr);
+    };
+
+    auto m = std::make_shared<ngraph::pattern::Matcher>(matmul, "ConvertMatmulToPointWiseConvolution");
+    this->register_matcher(m, callback);
+}
+
+ConvertMatmulWithBiasToPointWiseConvolution::ConvertMatmulWithBiasToPointWiseConvolution() {
+    auto const_input = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
+    auto const_fq = ngraph::pattern::wrap_type<ngraph::opset7::FakeQuantize>({const_input,
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>()});
+    auto second_input = std::make_shared<ngraph::pattern::op::Or>(ngraph::OutputVector{const_input, const_fq});
+    auto matmul = ngraph::pattern::wrap_type<ngraph::opset7::MatMul>({ngraph::pattern::any_input(), second_input});
+    auto bias = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
+    auto add = ngraph::pattern::wrap_type<ngraph::opset7::Add>({matmul, bias});
+
+    ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher &m) {
+        const auto& pattern_map = m.get_pattern_value_map();
+        return Convert(pattern_map.at(matmul).get_node_shared_ptr(), pattern_map.at(add).get_node_shared_ptr(),
+            pattern_map.at(bias).get_node_shared_ptr(), nullptr);
+    };
+
+    auto m = std::make_shared<ngraph::pattern::Matcher>(add, "ConvertMatmulWithBiasToPointWiseConvolution");
+    this->register_matcher(m, callback);
+}
+
+ConvertMatmulWithFqToPointWiseConvolution::ConvertMatmulWithFqToPointWiseConvolution() {
+    auto const_input = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
+    auto const_fq = ngraph::pattern::wrap_type<ngraph::opset7::FakeQuantize>({const_input,
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>()});
+    auto second_input = std::make_shared<ngraph::pattern::op::Or>(ngraph::OutputVector{const_input, const_fq});
+    auto matmul = ngraph::pattern::wrap_type<ngraph::opset7::MatMul>({ngraph::pattern::any_input(), second_input});
+    auto bias = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
+    auto add = ngraph::pattern::wrap_type<ngraph::opset7::Add>({matmul, bias});
+    auto matmul_out = std::make_shared<ngraph::pattern::op::Or>(ngraph::OutputVector{add, matmul});
+    auto out_fq = ngraph::pattern::wrap_type<ngraph::opset7::FakeQuantize>({matmul_out,
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>()});
+
+    ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher &m) {
+        const auto& pattern_map = m.get_pattern_value_map();
+        auto add_it = pattern_map.find(add);
+        auto add_node = (add_it == std::end(pattern_map) ? nullptr : add_it->second.get_node_shared_ptr());
+        auto bias_it = pattern_map.find(bias);
+        auto bias_node = (bias_it == std::end(pattern_map) ? nullptr : bias_it->second.get_node_shared_ptr());
+        return Convert(pattern_map.at(matmul).get_node_shared_ptr(), add_node, bias_node,
+             pattern_map.at(out_fq).get_node_shared_ptr());
+    };
+
+    auto m = std::make_shared<ngraph::pattern::Matcher>(out_fq, "ConvertMatmulWithFqToPointWiseConvolution");
+    this->register_matcher(m, callback);
+}
\ No newline at end of file
diff --git a/inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.hpp b/inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.hpp
new file mode 100644
index 00000000000000..999b529194d860
--- /dev/null
+++ b/inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.hpp
@@ -0,0 +1,71 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ngraph/pass/graph_rewrite.hpp>
+
+namespace GNAPluginNS {
+
+/**
+ * @brief Convert a MatMul with batch size unsupported by GNA to a point-wise convolution with NHWC layout
+ * with transposes around it:
+ *                                      Transose (NHWC -> NCHW)
+ *                                                 |
+ * Matmul                               Convolution in NHWC layout
+ * Input1: [A, B] B > 8     ------->    Input: [1, 1, A, B]
+ * Input2: [B, C]                       Kernel: [C, B, 1, 1]
+ * Output: [A, C]                       Output: [1, 1, A, C]
+ *                                                  |
+ *                                      Transose (NCHW -> NHWC)
+ */
+class ConvertMatmulToPointWiseConvolution : public ngraph::pass::MatcherPass {
+public:
+  NGRAPH_RTTI_DECLARATION;
+  ConvertMatmulToPointWiseConvolution();
+};
+
+/**
+ * @brief Convert a MatMul with batch size unsupported by GNA to a point-wise convolution with NHWC layout
+ * with transposes around it, moved add with bias before the last transpose:
+ *                                      Transose (NHWC -> NCHW)
+ *                                                 |
+ * Matmul                               Convolution in NHWC layout
+ * Input1: [A, B] B > 8     ------->    Input: [1, 1, A, B]
+ * Input2: [B, C]                       Kernel: [C, B, 1, 1]
+ * Output: [A, C]                       Output: [1, 1, A, C]
+ *       |                                         |
+ *      Add (const)                            Add (const)
+ *                                                 |
+ *                                      Transose (NCHW -> NHWC)
+ */
+class ConvertMatmulWithBiasToPointWiseConvolution : public ngraph::pass::MatcherPass {
+public:
+  NGRAPH_RTTI_DECLARATION;
+  ConvertMatmulWithBiasToPointWiseConvolution();
+};
+
+/**
+ * @brief Convert a MatMul with batch size unsupported by GNA to a point-wise convolution with NHWC layout
+ * with transposes around it, moved add with bias and/or fake quantize before the last transpose:
+ *                                      Transose (NHWC -> NCHW)
+ *                                                 |
+ * Matmul                               Convolution in NHWC layout
+ * Input1: [A, B] B > 8     ------->    Input: [1, 1, A, B]
+ * Input2: [B, C]                       Kernel: [C, B, 1, 1]
+ * Output: [A, C]                       Output: [1, 1, A, C]
+ *       |                                         |
+ *      Add (const)                            Add (const)
+ *       |                                         |
+ *     FakeQuantize                            FakeQuantize
+ *                                                 |
+ *                                         Transose (NCHW -> NHWC)
+ */
+class ConvertMatmulWithFqToPointWiseConvolution : public ngraph::pass::MatcherPass {
+public:
+  NGRAPH_RTTI_DECLARATION;
+  ConvertMatmulWithFqToPointWiseConvolution();
+};
+
+} // namespace GNAPluginNS
\ No newline at end of file
diff --git a/inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.cpp b/inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.cpp
new file mode 100644
index 00000000000000..a9d79c831ab34d
--- /dev/null
+++ b/inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.cpp
@@ -0,0 +1,131 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "transformations/split_convolution_with_large_buffer_size.hpp"
+
+#include <numeric>
+
+#include <ngraph/opsets/opset7.hpp>
+#include <ngraph/pattern/op/or.hpp>
+#include <ngraph/pattern/op/wrap_type.hpp>
+
+#include "backend/gna_limitations.hpp"
+
+using namespace GNAPluginNS;
+
+NGRAPH_RTTI_DEFINITION(SplitConvolution, "SplitConvolution", 0);
+NGRAPH_RTTI_DEFINITION(SplitConvolutionWithBias, "SplitConvolutionWithBias", 0);
+NGRAPH_RTTI_DEFINITION(SplitConvolutionWithFq, "SplitConvolutionWithFq", 0);
+
+static std::vector<int64_t> GetConvSplitSizes(std::shared_ptr<ngraph::Node> conv) {
+    uint32_t width = conv->get_input_shape(0).back();
+    uint32_t in_channels = conv->get_input_shape(0).at(1);
+    uint32_t usedWidth = 0;
+    std::vector<int64_t> split_sizes;
+    uint32_t width_max_size = GNALimitations::bufferMaxSize / in_channels;
+    width_max_size = width_max_size - width_max_size % 64;
+    while (usedWidth < width) {
+        uint32_t width_part = std::min(width - usedWidth, width_max_size);
+        split_sizes.push_back(width_part);
+        usedWidth += width_part;
+    }
+    IE_ASSERT(usedWidth == width);
+    return split_sizes;
+}
+
+static bool Convert(std::shared_ptr<ngraph::Node> conv,
+                    std::shared_ptr<ngraph::Node> add,
+                    std::shared_ptr<ngraph::Node> bias,
+                    std::shared_ptr<ngraph::Node> fq) {
+    auto input_size = std::accumulate(std::begin(conv->get_input_shape(0)),
+        std::end(conv->get_input_shape(0)), 1, std::multiplies<size_t>());
+    if (input_size <= GNALimitations::bufferMaxSize) {
+        return false;
+    }
+
+    auto split_sizes = GetConvSplitSizes(conv);
+    IE_ASSERT(split_sizes.size() > 1);
+
+    /* TODO check if it's NHWC convolution wrapped with transposes or all input dimensions except of width == 1,
+        otherwise this split axis isn't supported */
+    const int64_t width_axis = conv->get_input_shape(0).size() - 1;
+    auto split_node = std::make_shared<ngraph::opset7::VariadicSplit>(conv->input_value(0),
+        ngraph::opset7::Constant::create(ngraph::element::i64, ngraph::Shape({1}), std::vector<int64_t>{width_axis}),
+        ngraph::opset7::Constant::create(ngraph::element::i64, ngraph::Shape({split_sizes.size()}), split_sizes));
+    split_node->set_friendly_name(conv->get_friendly_name() + "/split");
+    ngraph::OutputVector convOutputs;
+    std::shared_ptr<ngraph::Node> root_node = fq ? fq : (add ? add : conv);
+    for (int i = 0; i < split_sizes.size(); ++i) {
+        std::shared_ptr<ngraph::Node> output = conv->clone_with_new_inputs({split_node->output(i), conv->input_value(1)});
+        output->set_friendly_name(conv->get_friendly_name() + "_" + std::to_string(i));
+        if (bias) {
+            output = std::make_shared<ngraph::opset7::Add>(output, bias);
+        }
+
+        if (fq) {
+            output = fq->clone_with_new_inputs({output, fq->input_value(1), fq->input_value(2),
+                fq->input_value(3), fq->input_value(4)});
+        }
+        convOutputs.push_back(output);
+    }
+
+    auto concat = std::make_shared<ngraph::opset7::Concat>(convOutputs, width_axis);
+    concat->set_friendly_name(conv->get_friendly_name());
+    ngraph::replace_node(root_node, concat);
+    return true;
+}
+
+SplitConvolution::SplitConvolution() {
+    auto conv = ngraph::pattern::wrap_type<ngraph::opset7::Convolution>({ngraph::pattern::any_input(),
+        ngraph::pattern::any_input()});
+
+    ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher &m) {
+        const auto& pattern_map = m.get_pattern_value_map();
+        return Convert(pattern_map.at(conv).get_node_shared_ptr(), nullptr, nullptr, nullptr);
+    };
+
+    auto m = std::make_shared<ngraph::pattern::Matcher>(conv, "SplitConvolution");
+    this->register_matcher(m, callback);
+}
+
+SplitConvolutionWithBias::SplitConvolutionWithBias() {
+    auto conv = ngraph::pattern::wrap_type<ngraph::opset7::Convolution>({ngraph::pattern::any_input(),
+        ngraph::pattern::any_input()});
+    auto bias = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
+    auto add = ngraph::pattern::wrap_type<ngraph::opset7::Add>({conv, bias});
+
+    ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher &m) {
+        const auto& pattern_map = m.get_pattern_value_map();
+        return Convert(pattern_map.at(conv).get_node_shared_ptr(), pattern_map.at(add).get_node_shared_ptr(),
+            pattern_map.at(bias).get_node_shared_ptr(), nullptr);
+    };
+
+    auto m = std::make_shared<ngraph::pattern::Matcher>(add, "SplitConvolutionWithBias");
+    this->register_matcher(m, callback);
+}
+
+SplitConvolutionWithFq::SplitConvolutionWithFq() {
+    auto conv = ngraph::pattern::wrap_type<ngraph::opset7::Convolution>({ngraph::pattern::any_input(),
+        ngraph::pattern::any_input()});
+    auto bias = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
+    auto add = ngraph::pattern::wrap_type<ngraph::opset7::Add>({conv, bias});
+    auto conv_output = std::make_shared<ngraph::pattern::op::Or>(ngraph::OutputVector{conv, add});
+    auto out_fq = ngraph::pattern::wrap_type<ngraph::opset7::FakeQuantize>({conv_output,
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
+        ngraph::pattern::wrap_type<ngraph::opset7::Constant>()});
+
+    ngraph::matcher_pass_callback callback = [=](ngraph::pattern::Matcher &m) {
+        const auto& pattern_map = m.get_pattern_value_map();
+        auto add_it = pattern_map.find(add);
+        auto add_node = (add_it == std::end(pattern_map) ? nullptr : add_it->second.get_node_shared_ptr());
+        auto bias_it = pattern_map.find(bias);
+        auto bias_node = (bias_it == std::end(pattern_map) ? nullptr : bias_it->second.get_node_shared_ptr());
+        return Convert(pattern_map.at(conv).get_node_shared_ptr(), add_node, bias_node, pattern_map.at(out_fq).get_node_shared_ptr());
+    };
+
+    auto m = std::make_shared<ngraph::pattern::Matcher>(out_fq, "SplitConvolutionWithFq");
+    this->register_matcher(m, callback);
+}
\ No newline at end of file
diff --git a/inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.hpp b/inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.hpp
new file mode 100644
index 00000000000000..8667f4273bfaad
--- /dev/null
+++ b/inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.hpp
@@ -0,0 +1,34 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ngraph/pass/graph_rewrite.hpp>
+
+namespace GNAPluginNS {
+
+// @brief Splits convolution with large input buffer
+class SplitConvolution : public ngraph::pass::MatcherPass {
+public:
+  NGRAPH_RTTI_DECLARATION;
+  SplitConvolution();
+};
+
+// @brief Splits convolution with large input buffer, move add with bias to each convolution before concat
+class SplitConvolutionWithBias : public ngraph::pass::MatcherPass {
+public:
+  NGRAPH_RTTI_DECLARATION;
+  SplitConvolutionWithBias();
+};
+
+/* @brief Splits convolution with large input buffer,
+ * move add with bias and/or fake quantize to each convolution before concat
+ */
+class SplitConvolutionWithFq : public ngraph::pass::MatcherPass {
+public:
+  NGRAPH_RTTI_DECLARATION;
+  SplitConvolutionWithFq();
+};
+
+} // namespace GNAPluginNS
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp
new file mode 100644
index 00000000000000..7e3d15174f3e5f
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp
@@ -0,0 +1,230 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <vector>
+#include <memory>
+#include <tuple>
+#include <vector>
+#include <string>
+
+#include <ie_core.hpp>
+
+#include "common_test_utils/common_utils.hpp"
+#include "functional_test_utils/plugin_cache.hpp"
+#include "shared_test_classes/base/layer_test_utils.hpp"
+#include "functional_test_utils/blob_utils.hpp"
+#include "ngraph_functions/utils/ngraph_helpers.hpp"
+#include "ngraph_functions/builders.hpp"
+
+#include "ngraph_functions/pass/convert_prc.hpp"
+
+typedef std::tuple<
+    InferenceEngine::Precision,         // Network Precision
+    std::string,                        // Target Device
+    std::map<std::string, std::string>, // Configuration
+    std::vector<size_t>                 // Input Shape
+> convertMatmulToPointwiseConvParams;
+
+typedef std::tuple<
+    InferenceEngine::Precision,         // Network Precision
+    std::string,                        // Target Device
+    std::map<std::string, std::string>, // Configuration
+    std::vector<size_t>,                // Input Shape
+    std::pair<float, float>             // Input Min and Max
+> convertMatmulToPointwiseConvWithFqParams;
+
+namespace LayerTestsDefinitions {
+
+class ConvertMatmulToPointwiseConv : public testing::WithParamInterface<convertMatmulToPointwiseConvParams>,
+    public LayerTestsUtils::LayerTestsCommon {
+public:
+    static std::string getTestCaseName(testing::TestParamInfo<convertMatmulToPointwiseConvParams> obj) {
+        InferenceEngine::Precision netPrecision;
+        std::string targetDevice;
+        std::map<std::string, std::string> configuration;
+        std::vector<size_t> inputShape;
+        std::tie(netPrecision, targetDevice, configuration, inputShape) = obj.param;
+
+        std::ostringstream result;
+        result << "netPRC=" << netPrecision.name() << "_";
+        result << "targetDevice=" << targetDevice << "_";
+        for (auto const& configItem : configuration) {
+            result << "_configItem=" << configItem.first << "_" << configItem.second;
+        }
+        result << "_inputShape=" << CommonTestUtils::vec2str(inputShape);
+        return result.str();
+    }
+
+    InferenceEngine::Blob::Ptr GenerateInput(const InferenceEngine::InputInfo& info) const {
+        InferenceEngine::Blob::Ptr blob = make_blob_with_precision(info.getTensorDesc());
+        blob->allocate();
+
+        auto* rawBlobDataPtr = blob->buffer().as<float*>();
+        std::vector<float> values = CommonTestUtils::generate_float_numbers(blob->size(), -0.2f, 0.2f);
+        for (size_t i = 0; i < blob->size(); i++) {
+            rawBlobDataPtr[i] = values[i];
+        }
+        return blob;
+    }
+
+protected:
+    void SetUp() override {
+        InferenceEngine::Precision netPrecision;
+        std::vector<size_t> inputShape;
+        std::tie(netPrecision, targetDevice, configuration, inputShape) = this->GetParam();
+        auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
+
+        auto params = ngraph::builder::makeParams(ngPrc, {inputShape});
+
+        size_t batch = inputShape[inputShape.size() - 2];
+        size_t elemNum = inputShape[inputShape.size() - 1];
+        std::vector<float> weights = CommonTestUtils::generate_float_numbers(elemNum * elemNum, -0.1f, 0.1f);
+        auto weightsNode = std::make_shared<ngraph::opset7::Constant>(ngPrc, ngraph::Shape{elemNum, elemNum}, weights);
+        auto matmul = ngraph::builder::makeMatMul(params[0], weightsNode, false, true);
+
+        auto bias = ngraph::builder::makeConstant(ngPrc, std::vector<size_t>{1, batch, 1}, std::vector<float>{1.0f});
+        auto add = ngraph::builder::makeEltwise(matmul, bias, ngraph::helpers::EltwiseTypes::ADD);
+
+        auto pattern = std::make_shared<ngraph::opset7::Constant>(ngraph::element::Type_t::i64,
+            ngraph::Shape{ inputShape.size() }, inputShape);
+        auto reshape = std::make_shared<ngraph::opset7::Reshape>(matmul, pattern, false);
+        auto relu = std::make_shared<ngraph::opset7::Relu>(reshape);
+
+        ngraph::ResultVector results{ std::make_shared<ngraph::opset7::Result>(relu)};
+        function = std::make_shared<ngraph::Function>(results, params, "ConvertMatmulToPointwiseConv");
+    }
+};
+
+class ConvertMatmulToPointwiseConvWithFq : public testing::WithParamInterface<convertMatmulToPointwiseConvWithFqParams>,
+    public LayerTestsUtils::LayerTestsCommon {
+    float inputDataMin = -10.0f;
+    float inputDataMax = 10.0f;
+    float inputDataResolution = 1.0f;
+
+public:
+    static std::string getTestCaseName(testing::TestParamInfo<convertMatmulToPointwiseConvWithFqParams> obj) {
+        InferenceEngine::Precision netPrecision;
+        std::string targetDevice;
+        std::map<std::string, std::string> configuration;
+        std::vector<size_t> inputShape;
+        std::pair<float, float> inputMinMax;
+        std::tie(netPrecision, targetDevice, configuration, inputShape, inputMinMax) = obj.param;
+
+        std::ostringstream result;
+        result << "netPRC=" << netPrecision.name() << "_";
+        result << "targetDevice=" << targetDevice << "_";
+        for (auto const& configItem : configuration) {
+            result << "_configItem=" << configItem.first << "_" << configItem.second;
+        }
+        result << "_inputShape=" << CommonTestUtils::vec2str(inputShape);
+        result << "_inputMinMax=(" << inputMinMax.first << ".." << inputMinMax.second << ")";
+        return result.str();
+    }
+
+    InferenceEngine::Blob::Ptr GenerateInput(const InferenceEngine::InputInfo& info) const {
+        return FuncTestUtils::createAndFillBlob(info.getTensorDesc(), inputDataMax - inputDataMin, inputDataMin,
+            1 / inputDataResolution);
+    }
+
+protected:
+    void SetUp() override {
+        InferenceEngine::Precision netPrecision;
+        std::vector<size_t> inputShape;
+        std::pair<float, float> inputMinMax;
+        std::tie(netPrecision, targetDevice, configuration, inputShape, inputMinMax) = this->GetParam();
+        std::tie(inputDataMin, inputDataMax) = inputMinMax;
+        auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
+
+        auto params = ngraph::builder::makeParams(ngPrc, {inputShape});
+
+        auto inputLowNode = ngraph::builder::makeConstant(ngPrc, std::vector<size_t>{ 1 },
+            std::vector<float>{ inputDataMin });
+        auto inputHighNode = ngraph::builder::makeConstant(ngPrc, std::vector<size_t>{ 1 },
+            std::vector<float>{ inputDataMax });
+        auto inputFQ = std::make_shared<ngraph::opset7::FakeQuantize>(params[0],
+            inputLowNode, inputHighNode, inputLowNode, inputHighNode, UINT16_MAX);
+
+        size_t elemNum = inputShape[inputShape.size() - 1];
+
+        const float weightsMin = -0.2f;
+        const float weightsMax = 0.2f;
+        std::vector<float> weights = CommonTestUtils::generate_float_numbers(elemNum * elemNum, weightsMin, weightsMax);
+        auto weightsNode = std::make_shared<ngraph::opset7::Constant>(ngPrc, ngraph::Shape{elemNum, elemNum}, weights);
+        auto weightsLowNode = ngraph::builder::makeConstant(ngPrc, std::vector<size_t>{ 1 },
+            std::vector<float>{ weightsMin });
+        auto weightsHighNode = ngraph::builder::makeConstant(ngPrc, std::vector<size_t>{ 1 },
+              std::vector<float>{ weightsMax });
+        auto weightsFQNode = std::make_shared<ngraph::opset7::FakeQuantize>(weightsNode,
+            weightsLowNode, weightsHighNode, weightsLowNode, weightsHighNode, UINT16_MAX);
+        auto matmul = ngraph::builder::makeMatMul(inputFQ, weightsFQNode, false, true);
+
+        auto bias = ngraph::builder::makeConstant(ngPrc, std::vector<size_t>{1, 1, 1}, std::vector<float>{1.0f});
+        auto add = ngraph::builder::makeEltwise(matmul, bias, ngraph::helpers::EltwiseTypes::ADD);
+
+        auto outputLowNode = ngraph::builder::makeConstant(ngPrc, std::vector<size_t>{ 1 },
+            std::vector<float>{ -inputDataMax * weightsMax *  elemNum });
+        auto outputHighNode = ngraph::builder::makeConstant(ngPrc, std::vector<size_t>{ 1 },
+            std::vector<float>{ inputDataMax * weightsMax * elemNum });
+        auto outputFQ = std::make_shared<ngraph::opset7::FakeQuantize>(add,
+            outputLowNode, outputHighNode, outputLowNode, outputHighNode, UINT16_MAX);
+
+        auto pattern = std::make_shared<ngraph::opset7::Constant>(ngraph::element::Type_t::i64,
+            ngraph::Shape{ inputShape.size() }, inputShape);
+        auto reshape = std::make_shared<ngraph::opset7::Reshape>(outputFQ, pattern, false);
+
+        auto relu = std::make_shared<ngraph::opset7::Relu>(reshape);
+
+        ngraph::ResultVector results{ std::make_shared<ngraph::opset7::Result>(relu)};
+        function = std::make_shared<ngraph::Function>(results, params, "ConvertMatmulToPointwiseConv");
+    }
+};
+
+TEST_P(ConvertMatmulToPointwiseConv, CompareWithRefImpl) {
+    Run();
+};
+
+TEST_P(ConvertMatmulToPointwiseConvWithFq, CompareWithRefImpl) {
+    Run();
+};
+
+const std::vector<InferenceEngine::Precision> netPrecisions = {
+    InferenceEngine::Precision::FP32,
+    InferenceEngine::Precision::FP16
+};
+
+const std::vector<std::map<std::string, std::string>> configs = {
+    {
+        {"GNA_DEVICE_MODE", "GNA_SW_EXACT"},
+    }
+};
+
+const std::vector<std::vector<size_t>> inputShape = {
+    {1, 64, 64},
+    {1, 256, 128},
+    {1, 512, 128}
+};
+
+const std::vector<std::pair<float, float>> fqStats = {
+    {-0.5, 0.5}
+};
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvertMatmulToPointwiseConvTest, ConvertMatmulToPointwiseConv,
+    ::testing::Combine(
+        ::testing::ValuesIn(netPrecisions),
+        ::testing::Values(CommonTestUtils::DEVICE_GNA),
+        ::testing::ValuesIn(configs),
+        ::testing::ValuesIn(inputShape)),
+    ConvertMatmulToPointwiseConv::getTestCaseName);
+
+// Issue 55662
+INSTANTIATE_TEST_CASE_P(DISABLED_smoke_ConvertMatmulToPointwiseConvTest, ConvertMatmulToPointwiseConvWithFq,
+    ::testing::Combine(
+        ::testing::ValuesIn(netPrecisions),
+        ::testing::Values(CommonTestUtils::DEVICE_GNA),
+        ::testing::ValuesIn(configs),
+        ::testing::ValuesIn(inputShape),
+        ::testing::ValuesIn(fqStats)),
+    ConvertMatmulToPointwiseConvWithFq::getTestCaseName);
+
+} // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/eltwise_split_over_channels_pass.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/eltwise_split_over_channels_pass.cpp
index 17da73dfc99e18..f4c6cc98d34136 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/eltwise_split_over_channels_pass.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/eltwise_split_over_channels_pass.cpp
@@ -18,19 +18,21 @@
 typedef std::tuple<
         InferenceEngine::Precision,         // Network Precision
         std::string,                        // Target Device
-        std::map<std::string, std::string>  //Configuration
+        std::map<std::string, std::string>, // Configuration
+        std::vector<size_t>                 // Input Shape
 > EltwiseSplitOverChannelsPassParams;
 
 namespace LayerTestsDefinitions {
 
 class EltwiseSplitOverChannelsPassTest : public testing::WithParamInterface<EltwiseSplitOverChannelsPassParams>,
-                                             public LayerTestsUtils::LayerTestsCommon {
+                                         public LayerTestsUtils::LayerTestsCommon {
 public:
     static std::string getTestCaseName(testing::TestParamInfo<EltwiseSplitOverChannelsPassParams> obj) {
         InferenceEngine::Precision netPrecision;
         std::string targetDevice;
         std::map<std::string, std::string> configuration;
-        std::tie(netPrecision, targetDevice, configuration) = obj.param;
+        std::vector<size_t> inputShape;
+        std::tie(netPrecision, targetDevice, configuration, inputShape) = obj.param;
 
         std::ostringstream result;
         result << "netPRC=" << netPrecision.name() << "_";
@@ -38,20 +40,22 @@ class EltwiseSplitOverChannelsPassTest : public testing::WithParamInterface<Eltw
         for (auto const& configItem : configuration) {
             result << "_configItem=" << configItem.first << "_" << configItem.second;
         }
+        result << "_inputShape=" << CommonTestUtils::vec2str(inputShape);
         return result.str();
     }
 
 protected:
     void SetUp() override {
         InferenceEngine::Precision netPrecision;
-        std::tie(netPrecision, targetDevice, configuration) = this->GetParam();
+        std::vector<size_t> inputShape;
+        std::tie(netPrecision, targetDevice, configuration, inputShape) = this->GetParam();
         auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
 
-        auto params = ngraph::builder::makeParams(ngPrc, { {1, 67000} });
-        auto const_mult2 = ngraph::builder::makeConstant<float>(ngPrc, {1, 67000}, {-1.0f});
+        auto params = ngraph::builder::makeParams(ngPrc, { inputShape });
+        auto const_mult2 = ngraph::builder::makeConstant<float>(ngPrc, inputShape, {-1.0f});
 
         auto sum = ngraph::builder::makeEltwise(params[0], const_mult2, ngraph::helpers::EltwiseTypes::MULTIPLY);
-        function = std::make_shared<ngraph::Function>(sum, params, "RemovePermutationPass");
+        function = std::make_shared<ngraph::Function>(sum, params, "EltwiseSplitOverChannelsPassTest");
     }
 };
 
@@ -71,11 +75,17 @@ const std::vector<std::map<std::string, std::string>> configs = {
         }
 };
 
+const std::vector<std::vector<size_t>> inputShape = {
+    {1, 67000},
+    {1, 500000}
+};
+
 INSTANTIATE_TEST_CASE_P(smoke_EltwiseSplitOverChennels, EltwiseSplitOverChannelsPassTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA),
-                                ::testing::ValuesIn(configs)),
+                                ::testing::ValuesIn(configs),
+                                ::testing::ValuesIn(inputShape)),
                         EltwiseSplitOverChannelsPassTest::getTestCaseName);
 
 } // namespace LayerTestsDefinitions

From ec1134532a230020966d8e36a470d286859de503 Mon Sep 17 00:00:00 2001
From: Konstantin Satunin <konstantin.satunin@intel.com>
Date: Wed, 9 Jun 2021 18:52:44 +0300
Subject: [PATCH 049/178] Align logic and feature flag name (#6107)

---
 .ci/azure/linux.yml        | 1 +
 .ci/azure/mac.yml          | 2 +-
 .ci/azure/windows.yml      | 2 +-
 cmake/features.cmake       | 2 +-
 ngraph/test/CMakeLists.txt | 4 ++--
 5 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/.ci/azure/linux.yml b/.ci/azure/linux.yml
index f45f4e410c67ba..146775f6189f02 100644
--- a/.ci/azure/linux.yml
+++ b/.ci/azure/linux.yml
@@ -112,6 +112,7 @@ jobs:
         -DNGRAPH_ONNX_IMPORT_ENABLE=ON
         -DNGRAPH_ONNX_EDITOR_ENABLE=ON
         -DENABLE_FASTER_BUILD=ON
+        -DENABLE_STRICT_DEPENDENCIES=OFF
         -DIE_EXTRA_MODULES=$(OPENVINO_CONTRIB_REPO_DIR)/modules
         $(REPO_DIR)
       workingDirectory: $(BUILD_DIR)
diff --git a/.ci/azure/mac.yml b/.ci/azure/mac.yml
index 680ef281ac21fe..04d4c16ea23344 100644
--- a/.ci/azure/mac.yml
+++ b/.ci/azure/mac.yml
@@ -90,7 +90,7 @@ jobs:
       # Disable errors with Ninja
       export CXXFLAGS="-Wno-error=unused-command-line-argument"
       export CFLAGS="-Wno-error=unused-command-line-argument"
-      cmake -GNinja -DVERBOSE_BUILD=ON -DCMAKE_BUILD_TYPE=$(BUILD_TYPE) -DENABLE_PYTHON=ON -DENABLE_TESTS=ON -DIE_EXTRA_MODULES=$(OPENVINO_CONTRIB_REPO_DIR)/modules $(REPO_DIR)
+      cmake -GNinja -DVERBOSE_BUILD=ON -DCMAKE_BUILD_TYPE=$(BUILD_TYPE) -DENABLE_PYTHON=ON -DENABLE_TESTS=ON -DENABLE_STRICT_DEPENDENCIES=OFF -DIE_EXTRA_MODULES=$(OPENVINO_CONTRIB_REPO_DIR)/modules $(REPO_DIR)
     workingDirectory: $(BUILD_DIR)
     displayName: 'CMake'
 
diff --git a/.ci/azure/windows.yml b/.ci/azure/windows.yml
index 6b4e5203dd08f4..21a36392e33812 100644
--- a/.ci/azure/windows.yml
+++ b/.ci/azure/windows.yml
@@ -92,7 +92,7 @@ jobs:
 
   - script: |
       set PATH=$(WORK_DIR)\ninja-win;%PATH%
-      call "$(MSVS_VARS_PATH)" && cmake -GNinja -DENABLE_FASTER_BUILD=ON -DENABLE_TEMPLATE_PLUGIN=ON -DCMAKE_BUILD_TYPE=$(BUILD_TYPE) -DENABLE_TESTS=ON -DIE_EXTRA_MODULES=$(OPENVINO_CONTRIB_REPO_DIR)\modules -DCMAKE_C_COMPILER:PATH="$(MSVC_COMPILER_PATH)" -DCMAKE_CXX_COMPILER:PATH="$(MSVC_COMPILER_PATH)" $(REPO_DIR)
+      call "$(MSVS_VARS_PATH)" && cmake -GNinja -DENABLE_FASTER_BUILD=ON -DENABLE_TEMPLATE_PLUGIN=ON -DCMAKE_BUILD_TYPE=$(BUILD_TYPE) -DENABLE_TESTS=ON -DENABLE_STRICT_DEPENDENCIES=OFF -DIE_EXTRA_MODULES=$(OPENVINO_CONTRIB_REPO_DIR)\modules -DCMAKE_C_COMPILER:PATH="$(MSVC_COMPILER_PATH)" -DCMAKE_CXX_COMPILER:PATH="$(MSVC_COMPILER_PATH)" $(REPO_DIR)
     workingDirectory: $(BUILD_DIR)
     displayName: 'CMake'
 
diff --git a/cmake/features.cmake b/cmake/features.cmake
index 7518c99c86838c..aff805adb15a64 100644
--- a/cmake/features.cmake
+++ b/cmake/features.cmake
@@ -6,7 +6,7 @@ ie_dependent_option (ENABLE_MKL_DNN "MKL-DNN plugin for inference engine" ON "X8
 
 ie_option (ENABLE_TESTS "unit, behavior and functional tests" OFF)
 
-ie_option (ENABLE_STRICT_DEPENDENCIES "Skip configuring \"convinient\" dependencies for efficient parallel builds" OFF)
+ie_option (ENABLE_STRICT_DEPENDENCIES "Skip configuring \"convinient\" dependencies for efficient parallel builds" ON)
 
 ie_dependent_option (ENABLE_CLDNN "clDnn based plugin for inference engine" ON "X86_64;NOT APPLE;NOT MINGW;NOT WINDOWS_STORE;NOT WINDOWS_PHONE" OFF)
 
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 97f85a3128f926..7ee6b89580d0f6 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -286,7 +286,7 @@ set_source_files_properties(includes.cpp PROPERTIES COMPILE_DEFINITIONS
 if (ENABLE_MKL_DNN)
     message(STATUS "NGRAPH_TESTS: IE:CPU enabled")
     set(ACTIVE_BACKEND_LIST ${ACTIVE_BACKEND_LIST} "IE:CPU")
-    if (NOT ENABLE_STRICT_DEPENDENCIES)
+    if (ENABLE_STRICT_DEPENDENCIES)
         # For convinience add a runtime dependency to build along with this target.
         # Warning: Parallel build with -GNinja may not be efficient.
         list(APPEND UNIT_TESTS_DEPENDENCIES MKLDNNPlugin)
@@ -296,7 +296,7 @@ endif()
 if (ENABLE_CLDNN)
     message(STATUS "NGRAPH_TESTS: IE:GPU enabled")
     set(ACTIVE_BACKEND_LIST ${ACTIVE_BACKEND_LIST} "IE:GPU")
-    if (NOT ENABLE_STRICT_DEPENDENCIES)
+    if (ENABLE_STRICT_DEPENDENCIES)
         # For convinience add a runtime dependency to build along with this target.
         # Warning: Parallel build with -GNinja may not be efficient.
         list(APPEND UNIT_TESTS_DEPENDENCIES clDNNPlugin)

From 3bedd051dcf91389c957fdbca1ec4e8624a2f615 Mon Sep 17 00:00:00 2001
From: Ivan Tikhonov <ivan.tikhonov@intel.com>
Date: Wed, 9 Jun 2021 19:03:38 +0300
Subject: [PATCH 050/178] processing sinks in hetero plugin, update single
 layer tests (#6090)

---
 .../src/hetero_plugin/hetero_executable_network.cpp      | 6 +++++-
 .../shared_tests_instances/single_layer_tests/memory.cpp | 2 +-
 ngraph/core/include/ngraph/op/util/op_types.hpp          | 4 ++++
 ngraph/core/src/op/util/op_types.cpp                     | 9 +++++++++
 4 files changed, 19 insertions(+), 2 deletions(-)

diff --git a/inference-engine/src/hetero_plugin/hetero_executable_network.cpp b/inference-engine/src/hetero_plugin/hetero_executable_network.cpp
index 9f0135aa25e357..994ba866f7a2e9 100644
--- a/inference-engine/src/hetero_plugin/hetero_executable_network.cpp
+++ b/inference-engine/src/hetero_plugin/hetero_executable_network.cpp
@@ -312,6 +312,7 @@ HeteroExecutableNetwork::HeteroExecutableNetwork(const InferenceEngine::CNNNetwo
     struct Subgraph {
         ngraph::ResultVector    _results;
         ngraph::ParameterVector _parameters;
+        ngraph::SinkVector      _sinks;
         std::string             _affinity;
     };
     std::unordered_map<int, Subgraph> subgraphs;
@@ -325,6 +326,9 @@ HeteroExecutableNetwork::HeteroExecutableNetwork(const InferenceEngine::CNNNetwo
         } else if (ngraph::op::is_parameter(node)) {
             subgraph._parameters.emplace_back(
                 std::dynamic_pointer_cast<ngraph::op::v0::Parameter>(node->shared_from_this()));
+        } else if (ngraph::op::is_sink(node)) {
+            subgraph._sinks.emplace_back(
+                    std::dynamic_pointer_cast<ngraph::op::Sink>(node->shared_from_this()));
         }
         auto itAffinity = affinities.find(node);
         if (itAffinity != affinities.end()) {
@@ -373,7 +377,7 @@ HeteroExecutableNetwork::HeteroExecutableNetwork(const InferenceEngine::CNNNetwo
     for (auto&& subgraph : orderedSubgraphs) {
         _networks[id]._device = subgraph._affinity;
         subFunctions[id] =
-            std::make_shared<ngraph::Function>(subgraph._results, subgraph._parameters,
+            std::make_shared<ngraph::Function>(subgraph._results, subgraph._sinks, subgraph._parameters,
                                                      _name + '_' + std::to_string(id));
         _networks[id]._clonedNetwork = CNNNetwork{subFunctions[id]};
         // update of pre-processing info
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp
index 9ab20c3eda4cf7..062ea0cad9194e 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp
@@ -38,7 +38,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MemoryTest, MemoryTest,
                 ::testing::ValuesIn(iterationCount),
                 ::testing::ValuesIn(inShapes),
                 ::testing::ValuesIn(inputPrecisions),
-                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                ::testing::Values(CommonTestUtils::DEVICE_CPU, "HETERO:CPU")),
         MemoryTest::getTestCaseName);
 
 } // namespace
diff --git a/ngraph/core/include/ngraph/op/util/op_types.hpp b/ngraph/core/include/ngraph/op/util/op_types.hpp
index b672f5518c4086..6d162157ab581b 100644
--- a/ngraph/core/include/ngraph/op/util/op_types.hpp
+++ b/ngraph/core/include/ngraph/op/util/op_types.hpp
@@ -34,6 +34,8 @@ namespace ngraph
         NGRAPH_API
         bool is_output(const ngraph::Node* node);
         NGRAPH_API
+        bool is_sink(const ngraph::Node* node);
+        NGRAPH_API
         bool is_constant(const ngraph::Node* node);
         NGRAPH_API
         bool is_commutative(const ngraph::Node* node);
@@ -60,6 +62,8 @@ namespace ngraph
         NGRAPH_API
         bool is_output(const std::shared_ptr<ngraph::Node>& node);
         NGRAPH_API
+        bool is_sink(const std::shared_ptr<ngraph::Node>& node);
+        NGRAPH_API
         bool is_constant(const std::shared_ptr<ngraph::Node>& node);
         NGRAPH_API
         bool is_commutative(const std::shared_ptr<ngraph::Node>& node);
diff --git a/ngraph/core/src/op/util/op_types.cpp b/ngraph/core/src/op/util/op_types.cpp
index f0852233ec754f..354c605ced17be 100644
--- a/ngraph/core/src/op/util/op_types.cpp
+++ b/ngraph/core/src/op/util/op_types.cpp
@@ -76,6 +76,11 @@ bool ngraph::op::is_output(const ngraph::Node* node)
     return dynamic_cast<const ngraph::op::Result*>(node) != nullptr;
 }
 
+bool ngraph::op::is_sink(const ngraph::Node* node)
+{
+    return dynamic_cast<const ngraph::op::Sink*>(node) != nullptr;
+}
+
 bool ngraph::op::is_constant(const ngraph::Node* node)
 {
     return dynamic_cast<const ngraph::op::Constant*>(node) != nullptr;
@@ -134,6 +139,10 @@ bool ngraph::op::is_output(const std::shared_ptr<ngraph::Node>& node)
 {
     return is_output(node.get());
 }
+bool ngraph::op::is_sink(const std::shared_ptr<ngraph::Node>& node)
+{
+    return is_sink(node.get());
+}
 bool ngraph::op::is_constant(const std::shared_ptr<ngraph::Node>& node)
 {
     return is_constant(node.get());

From 1a6392eb5335f64a053d6301099e30838a3a2aa4 Mon Sep 17 00:00:00 2001
From: Mikhail Ryzhov <mikhail.ryzhov@intel.com>
Date: Wed, 9 Jun 2021 20:39:05 +0300
Subject: [PATCH 051/178] [GNA] Fixed export/import functionality (#5963)

* Rebase master

* [GNA] Fixed export/import functionality

* Extended import log

* Added logs

* Fixed importing issue for the old models

* Revert "Added logs"

This reverts commit 39a3882d560b241018b1ad20aa5318b991951379.

* Revert "Extended import log"

This reverts commit 59eb9d6fba91ba2ca86da159f2d55fe37fba0320.

* Reverted precision import

* Extended tests

* Enabled skipped tests

* Included gna2-common-api header

* Replaced included header

* Centos7 build fix
---
 .../src/gna_plugin/gna_model_serial.cpp       | 113 +++++++---
 .../src/gna_plugin/gna_model_serial.hpp       |   2 +
 .../serial/headers/2dot7/gna_model_header.hpp | 197 ++++++++++++++++++
 .../headers/latest/gna_model_header.hpp       |   6 +-
 .../skip_tests_config.cpp                     |   2 -
 .../import_export_base/import_export_base.cpp |   3 +
 .../base/layer_test_utils.hpp                 |   2 +
 .../src/base/layer_test_utils.cpp             |  11 +
 8 files changed, 301 insertions(+), 35 deletions(-)
 create mode 100644 inference-engine/src/gna_plugin/serial/headers/2dot7/gna_model_header.hpp

diff --git a/inference-engine/src/gna_plugin/gna_model_serial.cpp b/inference-engine/src/gna_plugin/gna_model_serial.cpp
index fdb99d7f273bc1..e32ded8a9e37ed 100644
--- a/inference-engine/src/gna_plugin/gna_model_serial.cpp
+++ b/inference-engine/src/gna_plugin/gna_model_serial.cpp
@@ -17,6 +17,7 @@
 #include <mm_malloc.h>
 #include <serial/headers/2dot2/gna_model_header.hpp>
 #include <serial/headers/2dot5/gna_model_header.hpp>
+#include <serial/headers/2dot6/gna_model_header.hpp>
 
 #endif
 
@@ -133,10 +134,11 @@ GNAPluginNS::HeaderLatest::ModelHeader GNAModelSerial::ReadHeader(std::istream &
                 }
                 case 5:
                 case 6:
+                case 7:
                     readNBytes(&header, sizeof(HeaderLatest::ModelHeader), is);
                     break;
                 default:
-                    THROW_GNA_EXCEPTION << "Imported file unsupported. minor version should have values in range 1 to 4 and is: " << header.version.minor;
+                    THROW_GNA_EXCEPTION << "Imported file unsupported. minor version should have values in range 1 to 7 and is: " << header.version.minor;
             }
             break;
         default:
@@ -154,6 +156,40 @@ GNAPluginNS::HeaderLatest::ModelHeader GNAModelSerial::ReadHeader(std::istream &
     return header;
 }
 
+GNAPluginNS::HeaderLatest::RuntimeEndPoint GNAModelSerial::ReadEndPoint(std::istream &is) {
+    is.exceptions(std::istream::failbit);
+
+    HeaderLatest::RuntimeEndPoint endPoint;
+    switch (modelHeader.version.major) {
+        case 2:
+            switch (modelHeader.version.minor) {
+                case 1:
+                case 2:
+                case 3:
+                case 4:
+                case 5:
+                case 6:
+                {
+                    Header2dot6::RuntimeEndPoint tempEndPoint2dot6;
+                    readBits(tempEndPoint2dot6, is);
+                    endPoint = HeaderLatest::RuntimeEndPoint(tempEndPoint2dot6, modelHeader.nGroup);
+                    break;
+                }
+                case 7:
+                    readNBytes(&endPoint, sizeof(HeaderLatest::RuntimeEndPoint), is);
+                    break;
+                default:
+                    THROW_GNA_EXCEPTION << "Imported file unsupported. minor version should have values in range 1 to 7 and is: " << modelHeader.version.minor;
+            }
+            break;
+        default:
+            THROW_GNA_EXCEPTION << "Imported file unsupported. Import for files with major version equal to: "
+            << modelHeader.version.major << " is not implemented";
+    }
+
+    return endPoint;
+}
+
 #define offsetFromBase(field)\
 getOffsetFromBase(field, #field)
 
@@ -324,18 +360,6 @@ void GNAModelSerial::Import(void *basePointer,
     is.read(reinterpret_cast<char*>(basePointer), gnaGraphSize);
 }
 
-
-uint32_t guessGrouping(Gna2Model const& model) {
-    if (model.NumberOfOperations == 0 ||
-        model.Operations == nullptr ||
-        model.Operations[0].Operands == nullptr ||
-        model.Operations[0].NumberOfOperands == 0 ||
-        model.Operations[0].Operands[0]->Shape.NumberOfDimensions < 2) {
-        THROW_GNA_EXCEPTION << "Can not guess grouping";
-    }
-    return (std::min)(model.Operations[0].Operands[0]->Shape.Dimensions[0], model.Operations[0].Operands[0]->Shape.Dimensions[1]);
-}
-
 void GNAModelSerial::Export(void * basePointer, size_t gnaGraphSize, std::ostream & os) const {
     os.exceptions(std::ostream::failbit);
 
@@ -366,6 +390,9 @@ void GNAModelSerial::Export(void * basePointer, size_t gnaGraphSize, std::ostrea
         out.descriptor_offset = offsetFromBase(ep.descriptor_ptr);
         out.scaleFactor = ep.scaleFactor;
         out.element_size = ep.element_size;
+        out.shape = ep.shape;
+        out.layout = ep.layout;
+        out.precision = ep.precision;
         out.orientation = ep.orientation;
         return out;
     };
@@ -381,7 +408,7 @@ void GNAModelSerial::Export(void * basePointer, size_t gnaGraphSize, std::ostrea
     header.headerSize = sizeof(HeaderLatest::ModelHeader);
     header.gnaMemSize = gnaGraphSize;
     header.layersCount = layers.size();
-    header.nGroup = guessGrouping(*gna2Model);
+    header.nGroup = 1; // just to support the old models
     header.nInputs = inputs.size();
     header.nOutputs = outputs.size();
     header.nTransposeInputs = transposeInputsInfo.size();
@@ -796,13 +823,22 @@ std::vector<HeaderLatest::RuntimeEndPoint> GNAModelSerial::serializeOutputs(cons
     std::size_t outputIndex = 0;
     for (auto const &output : outputsDataMap) {
         auto outputName = output.first;
-        auto inputDims = output.second->getTensorDesc().getDims();
-        uint32_t elementsCount = static_cast<uint32_t>(InferenceEngine::details::product(inputDims.begin(), inputDims.end()));
-
+        auto outputDims = output.second->getTensorDesc().getDims();
+        HeaderLatest::RuntimeEndPoint::Shape outputShape;
+        outputShape.NumberOfDimensions = outputDims.size();
+        for (size_t i=0; i < outputShape.NumberOfDimensions; ++i) {
+            outputShape.Dimensions[i] = static_cast<uint32_t>(outputDims[i]);
+        }
+        uint32_t elementsCount = static_cast<uint32_t>(InferenceEngine::details::product(outputDims.begin(), outputDims.end()));
+        InferenceEngine::Layout outputLayout = output.second->getLayout();
+        InferenceEngine::Precision::ePrecision outputPrecision = InferenceEngine::Precision::FP32;
         HeaderLatest::RuntimeEndPoint endPoint(outputsDesc[outputIndex].scale_factor,
                                                  outputsDesc[outputIndex].ptrs[0],
                                                  outputsDesc[outputIndex].num_bytes_per_element,
                                                  elementsCount,
+                                                 outputShape,
+                                                 outputLayout,
+                                                 outputPrecision,
                                                  outputsDesc[outputIndex].orientation);
         endPoints.push_back(endPoint);
         outputIndex++;
@@ -818,18 +854,26 @@ std::vector<HeaderLatest::RuntimeEndPoint> GNAModelSerial::serializeInputs(const
     for (auto const& input : inputsDataMap) {
         auto inputName = input.first;
         auto inputDims = input.second->getTensorDesc().getDims();
-
+        HeaderLatest::RuntimeEndPoint::Shape inputShape;
+        inputShape.NumberOfDimensions = inputDims.size();
+        for (size_t i=0; i < inputShape.NumberOfDimensions; ++i) {
+            inputShape.Dimensions[i] = static_cast<uint32_t>(inputDims[i]);
+        }
         double scaleFactor = inputDesc->getScaleFactor(inputIndex);
         std::vector<void *> descriptor_ptr = inputDesc->getPtrInputsGlobal(inputName);
         IE_ASSERT(descriptor_ptr.size() > 0);
         uint32_t element_size = 2u;
         uint32_t elementsCount = static_cast<uint32_t>(InferenceEngine::details::product(inputDims.begin(), inputDims.end()));
         intel_dnn_orientation_t orientation = inputDesc->getOrientation(inputName);
-
+        InferenceEngine::Layout inputLayout = input.second->getLayout();
+        InferenceEngine::Precision::ePrecision inputPrecision = InferenceEngine::Precision::FP32;
         HeaderLatest::RuntimeEndPoint endPoint(scaleFactor,
                                                  descriptor_ptr[0],
                                                  element_size,
                                                  elementsCount,
+                                                 inputShape,
+                                                 inputLayout,
+                                                 inputPrecision,
                                                  orientation);
         endPoints.push_back(endPoint);
         inputIndex++;
@@ -846,20 +890,24 @@ void GNAModelSerial::ImportInputs(std::istream &is,
     for (uint32_t inputIndex = 0; inputIndex < modelHeader.nInputs; inputIndex++) {
         const std::string& name = (modelHeader.version.major == 2 && modelHeader.version.minor >= 3)
                 ? inputNames.at(inputIndex) : std::string("input" + std::to_string(inputIndex));
-        HeaderLatest::RuntimeEndPoint input;
-        is.read(reinterpret_cast<char *>(&input), sizeof(input));
+
+        HeaderLatest::RuntimeEndPoint input = ReadEndPoint(is);
         inputsDesc->getPtrInputsGlobal(name).push_back(reinterpret_cast<float*>(reinterpret_cast<uint8_t *> (basePtr) + input.descriptor_offset));
         inputsDesc->orientation_in[name] = input.orientation;
         inputsDesc->bytes_allocated_for_input[name] = input.element_size * input.elements_count;
 
-        auto inputDims = InferenceEngine::SizeVector({modelHeader.nGroup, input.elements_count / modelHeader.nGroup});
-
+        auto inputDims = InferenceEngine::SizeVector();
+        for (auto i = 0; i < input.shape.NumberOfDimensions; ++i) {
+            inputDims.push_back(input.shape.Dimensions[i]);
+        }
+        InferenceEngine::Layout inputLayout = static_cast<InferenceEngine::Layout>(input.layout);
+        InferenceEngine::Precision inputPresicion = InferenceEngine::Precision(static_cast<InferenceEngine::Precision::ePrecision>(input.precision));
         dataMap[name] = std::make_shared<InferenceEngine::InputInfo>();
         dataMap[name]->setInputData(std::make_shared<InferenceEngine::Data>(name,
                                                             InferenceEngine::TensorDesc(
-                                                                    InferenceEngine::Precision::FP32,
+                                                                    inputPresicion,
                                                                     inputDims,
-                                                                    InferenceEngine::Layout::NC)));
+                                                                    inputLayout)));
         inputsDesc->inputScaleFactors.push_back(input.scaleFactor);
     }
 }
@@ -875,8 +923,8 @@ void GNAModelSerial::ImportOutputs(std::istream &is,
     for (uint32_t outputIndex = 0; outputIndex < modelHeader.nOutputs; outputIndex++) {
         const std::string& name = (modelHeader.version.major == 2 && modelHeader.version.minor >= 3)
                                   ? outputNames.at(outputIndex) : std::string("output" + std::to_string(outputIndex));
-        HeaderLatest::RuntimeEndPoint output;
-        is.read(reinterpret_cast<char *>(&output), sizeof(output));
+
+        HeaderLatest::RuntimeEndPoint output = ReadEndPoint(is);
         OutputDesc description;
         description.ptrs.push_back(reinterpret_cast<float*>(reinterpret_cast<uint8_t *> (basePtr) + output.descriptor_offset));
         description.orientation = kDnnInterleavedOrientation;
@@ -884,12 +932,17 @@ void GNAModelSerial::ImportOutputs(std::istream &is,
         description.num_bytes_per_element = output.element_size;
         description.scale_factor = output.scaleFactor;
 
-        auto outputDims = InferenceEngine::SizeVector({modelHeader.nGroup, output.elements_count / modelHeader.nGroup});
+        auto outputDims = InferenceEngine::SizeVector();
+        for (auto i = 0; i < output.shape.NumberOfDimensions; ++i) {
+            outputDims.push_back(output.shape.Dimensions[i]);
+        }
+        InferenceEngine::Layout outputLayout = static_cast<InferenceEngine::Layout>(output.layout);
+        InferenceEngine::Precision outputPresicion =  InferenceEngine::Precision(static_cast<InferenceEngine::Precision::ePrecision>(output.precision));
         dataMap[name] = std::make_shared<InferenceEngine::Data>(name,
                                                  InferenceEngine::TensorDesc(
-                                                         InferenceEngine::Precision::FP32,
+                                                         outputPresicion,
                                                          outputDims,
-                                                         InferenceEngine::Layout::NC));
+                                                         outputLayout));
         desc.at(outputIndex) = description;
     }
 }
diff --git a/inference-engine/src/gna_plugin/gna_model_serial.hpp b/inference-engine/src/gna_plugin/gna_model_serial.hpp
index d756a23f9fc0ab..f5310d826c4c5f 100644
--- a/inference-engine/src/gna_plugin/gna_model_serial.hpp
+++ b/inference-engine/src/gna_plugin/gna_model_serial.hpp
@@ -138,6 +138,8 @@ class GNAModelSerial {
      */
     static GNAPluginNS::HeaderLatest::ModelHeader ReadHeader(std::istream &is);
 
+    GNAPluginNS::HeaderLatest::RuntimeEndPoint ReadEndPoint(std::istream &is);
+
     /**
      * @brief Import model from FS into preallocated buffer,
      * buffers for pLayers, and pStructs are allocated here and required manual deallocation using mm_free
diff --git a/inference-engine/src/gna_plugin/serial/headers/2dot7/gna_model_header.hpp b/inference-engine/src/gna_plugin/serial/headers/2dot7/gna_model_header.hpp
new file mode 100644
index 00000000000000..14badf3adcf032
--- /dev/null
+++ b/inference-engine/src/gna_plugin/serial/headers/2dot7/gna_model_header.hpp
@@ -0,0 +1,197 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <cstdint>
+#include <map>
+#include "backend/dnn_types.h"
+#include "serial/headers/2dot4/gna_model_header.hpp"
+#include "serial/headers/2dot6/gna_model_header.hpp"
+#include "serial/headers/latest/gna_model_header.hpp"
+#include "gna_data_types.hpp"
+
+#pragma pack(push, 1)
+
+namespace GNAPluginNS {
+namespace Header2dot7 {
+
+/**
+ Maximal number of supported shape dimensions.
+ */
+#define GNA_SHAPE_MAXIMUM_NUMBER_OF_DIMENSIONS 8
+
+/**
+ * @brief Header version 2.7
+ */
+struct ModelHeader {
+    /**
+     *@brief MagicNumber – GNAM in ascii table, equals to hex 0x474e414d
+     */
+    char gnam[4] = {};
+    /**
+     * @brief if header size is not equal to sizeof ModelHeader - some reserved data append in the end of header
+     * usually it is an indicator of working with version of model different that is current export function produce
+     */
+    uint32_t headerSize = 0u;
+    struct Version {
+        /**
+         * @details Version of format Major – unsigned int, ex: 0x0001
+         * every change in the header or in the layers definition should be reflected in version change
+         * for backward compatibility new parsers can read old versions of model with certain restrictions
+         */
+        uint16_t major = 2u;
+        /**
+         * @details Version of Format Minor – unsigned int,  corresponding to build revision for example
+         * changes in minor version are not affected layout of model
+         */
+        uint32_t minor = 7u;
+    } version;
+    /**
+     * @brief Memory required to be allocated using GNAAlloc()
+     */
+    uint64_t gnaMemSize = 0ull;
+    /**
+     * @brief Number of GNA Layers
+     */
+    uint64_t layersCount = 0ull;
+    /**
+     * @brief Grouping level
+     * This is depricted field and used for old models only (<=2.6)
+     */
+    uint32_t nGroup = 0u;
+
+    /**
+     * Convolution related setting - they are affecting input transformation
+     */
+    uint32_t nRotateRows = 0u;
+    uint32_t nRotateColumns = 0u;
+    bool doRotateInput = false;
+
+    uint32_t nInputs = 0u;
+    uint32_t nOutputs = 0u;
+
+    /**
+     * Convolution related setting - they are affecting output transformation
+     */
+    uint32_t nRotateOutputRows = 0u;
+    uint32_t nRotateOutputColumns = 0u;
+    bool doRotateOutput = false;
+
+    uint32_t nTransposeInputs = 0u;
+    uint32_t nTransposeOutputs = 0u;
+
+    /**
+     * Reserved Data might be here
+     */
+    ModelHeader() = default;
+    ModelHeader(GNAPluginNS::Header2dot1::ModelHeader const &old) {
+        gnaMemSize = old.gnaMemSize;
+        layersCount = old.layersCount;
+        nGroup = old.nGroup;
+        nRotateRows = old.nRotateRows;
+        nRotateColumns = old.nRotateColumns;
+        nInputs = old.nInputs;
+        nOutputs = old.nOutputs;
+        version.minor = old.version.minor;
+    }
+    ModelHeader(GNAPluginNS::Header2dot4::ModelHeader const &old) {
+        gnaMemSize = old.gnaMemSize;
+        layersCount = old.layersCount;
+        nGroup = old.nGroup;
+        nRotateRows = old.nRotateRows;
+        nRotateColumns = old.nRotateColumns;
+        nInputs = old.nInputs;
+        nOutputs = old.nOutputs;
+        nRotateOutputRows = old.nRotateOutputRows;
+        nRotateOutputColumns = old.nRotateOutputColumns;
+        doRotateOutput = old.doRotateOutput;
+        version.minor = old.version.minor;
+    }
+};
+#pragma pack(pop)
+
+/*
+ * In runtime endpoint mostly same as in serial version, except of descriptor field
+ */
+struct RuntimeEndPoint {
+    /**
+     * if scale factor is different then pased into infer , network might need to be requantized
+     */
+    float scaleFactor = 0;
+    /**
+     * Pointer descriptor
+     */
+    void* descriptor_ptr = nullptr;
+    /**
+     * Endpoint resolution in bytes.
+     */
+    uint32_t element_size = 0;
+    /**
+     * Number of elements
+     */
+    uint32_t elements_count = 0;
+    /**
+     * Offset in bytes of pointer descriptor
+    */
+    uint64_t descriptor_offset = 0ull;
+    /**
+     Shape specifying dimension values.
+    */
+    struct Shape {
+        /**
+         Number of dimensions or rank or order.
+        */
+        uint32_t NumberOfDimensions = 0;
+        /**
+         array specifying value of each dimension.
+        Set all zeros for scalars.
+        */
+        uint32_t Dimensions[GNA_SHAPE_MAXIMUM_NUMBER_OF_DIMENSIONS] = {0};
+    } shape;
+    /**
+     * Blob layout
+     */
+    uint8_t layout = InferenceEngine::Layout::NC;
+    /**
+     * Blob precision
+     */
+    uint8_t precision = InferenceEngine::Precision::FP32;
+
+    intel_dnn_orientation_t orientation = kDnnUnknownOrientation;
+
+    RuntimeEndPoint() = default;
+    RuntimeEndPoint(const GNAPluginNS::Header2dot6::RuntimeEndPoint &old, uint32_t ngroup) {
+        scaleFactor = old.scaleFactor;
+        descriptor_ptr = old.descriptor_ptr;
+        element_size = old.element_size;
+        elements_count = old.elements_count;
+        orientation = old.orientation;
+        layout = InferenceEngine::Layout::NC;
+        precision = InferenceEngine::Precision::FP32;
+        descriptor_offset = old.descriptor_offset;
+        InferenceEngine::SizeVector dims = {ngroup, elements_count / ngroup};
+        shape.NumberOfDimensions = static_cast<uint32_t>(dims.size());
+        for (auto i = 0; i < dims.size(); i++) {
+            shape.Dimensions[i] = dims[i];
+        }
+    }
+    RuntimeEndPoint(double scaleFactor,
+                    void* descriptor_ptr,
+                    uint32_t element_size,
+                    uint32_t elements_count,
+                    Shape shape,
+                    uint8_t layout,
+                    uint8_t precision,
+                    intel_dnn_orientation_t orientation) : scaleFactor(scaleFactor),
+                                                           descriptor_ptr(descriptor_ptr),
+                                                           element_size(element_size),
+                                                           elements_count(elements_count),
+                                                           shape(shape),
+                                                           layout(layout),
+                                                           precision(precision),
+                                                           orientation(orientation) { }
+};
+} // namespace Header2dot7
+} // namespace GNAPluginNS
diff --git a/inference-engine/src/gna_plugin/serial/headers/latest/gna_model_header.hpp b/inference-engine/src/gna_plugin/serial/headers/latest/gna_model_header.hpp
index 89292ab88afb17..7ec27b2caed386 100644
--- a/inference-engine/src/gna_plugin/serial/headers/latest/gna_model_header.hpp
+++ b/inference-engine/src/gna_plugin/serial/headers/latest/gna_model_header.hpp
@@ -4,11 +4,11 @@
 
 #pragma once
 
-#include "serial/headers/2dot6/gna_model_header.hpp"
+#include "serial/headers/2dot7/gna_model_header.hpp"
 
 namespace GNAPluginNS {
 namespace HeaderLatest {
-using ModelHeader = GNAPluginNS::Header2dot6::ModelHeader;
-using RuntimeEndPoint = GNAPluginNS::Header2dot6::RuntimeEndPoint;
+using ModelHeader = GNAPluginNS::Header2dot7::ModelHeader;
+using RuntimeEndPoint = GNAPluginNS::Header2dot7::RuntimeEndPoint;
 }
 }
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/skip_tests_config.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/skip_tests_config.cpp
index cb4cc459a95e28..a59ad83eaed76e 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/skip_tests_config.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/skip_tests_config.cpp
@@ -60,8 +60,6 @@ std::vector<std::string> disabledTestPatterns() {
         R"(.*ConstantResultSubgraphTest.*inPrc=(U8|I8|I32|U64|I64|BOOL).*)",
         // TODO: Issue 51528
         R"(.*CachingSupport.*_(u8|i16)_.*)",
-        // TODO: Issue 51527
-        R"(.*CachingSupport.*_batch2_.*)",
         // TODO: Issue 51525
         R"(.*CachingSupport.*KSOFunction.*)",
         // TODO: Issue 57363 (Param -> Result subgraphs)
diff --git a/inference-engine/tests/functional/plugin/shared/src/base/import_export_base/import_export_base.cpp b/inference-engine/tests/functional/plugin/shared/src/base/import_export_base/import_export_base.cpp
index 0db7264cb74410..c30945dc914d83 100644
--- a/inference-engine/tests/functional/plugin/shared/src/base/import_export_base/import_export_base.cpp
+++ b/inference-engine/tests/functional/plugin/shared/src/base/import_export_base/import_export_base.cpp
@@ -69,13 +69,16 @@ void ImportNetworkTestBase::Run() {
 
     for (const auto& next_input : importedExecNetwork.GetInputsInfo()) {
         ASSERT_NO_THROW(compiledExecNetwork.GetInputsInfo()[next_input.first]);
+        Compare(next_input.second->getTensorDesc(), compiledExecNetwork.GetInputsInfo()[next_input.first]->getTensorDesc());
     }
     for (const auto& next_output : importedExecNetwork.GetOutputsInfo()) {
         ASSERT_NO_THROW(compiledExecNetwork.GetOutputsInfo()[next_output.first]);
     }
     auto importedOutputs = GetOutputs();
     ASSERT_EQ(actualOutputs.size(), importedOutputs.size());
+
     for (size_t i = 0; i < actualOutputs.size(); i++) {
+        Compare(actualOutputs[i]->getTensorDesc(), importedOutputs[i]->getTensorDesc());
         Compare(actualOutputs[i], importedOutputs[i]);
     }
 }
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/base/layer_test_utils.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/base/layer_test_utils.hpp
index 9b8b78b6ef0912..9d13251574331f 100644
--- a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/base/layer_test_utils.hpp
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/base/layer_test_utils.hpp
@@ -72,6 +72,8 @@ class LayerTestsCommon : public CommonTestUtils::TestsCommon {
 
     virtual void Compare(const InferenceEngine::Blob::Ptr &expected, const InferenceEngine::Blob::Ptr &actual);
 
+    virtual void Compare(const InferenceEngine::TensorDesc &actualDesc, const InferenceEngine::TensorDesc &expectedDesc);
+
     virtual void SetRefMode(RefMode mode);
 
     std::shared_ptr<ngraph::Function> GetFunction();
diff --git a/inference-engine/tests/functional/shared_test_classes/src/base/layer_test_utils.cpp b/inference-engine/tests/functional/shared_test_classes/src/base/layer_test_utils.cpp
index cc3927b25c53b3..056826aff86a5f 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/base/layer_test_utils.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/base/layer_test_utils.cpp
@@ -274,6 +274,17 @@ void LayerTestsCommon::Compare(const InferenceEngine::Blob::Ptr &expected, const
     }
 }
 
+void LayerTestsCommon::Compare(const InferenceEngine::TensorDesc &actualDesc, const InferenceEngine::TensorDesc &expectedDesc) {
+    auto expectedDims = actualDesc.getDims();
+    auto actualDims = expectedDesc.getDims();
+    ASSERT_EQ(actualDims.size(), expectedDims.size());
+    for (size_t j = 0; j < actualDims.size(); ++j) {
+        ASSERT_EQ(actualDims.at(j), expectedDims.at(j));
+    }
+    ASSERT_EQ(actualDesc.getLayout(), expectedDesc.getLayout());
+    ASSERT_EQ(actualDesc.getPrecision(), expectedDesc.getPrecision());
+}
+
 void LayerTestsCommon::ConfigureNetwork() {
     for (const auto &in : cnnNetwork.getInputsInfo()) {
         if (inLayout != InferenceEngine::Layout::ANY) {

From d4f323f3ddcafc6fbcb777c372a9222c437dd383 Mon Sep 17 00:00:00 2001
From: Maxim Shevtsov <maxim.y.shevtsov@intel.com>
Date: Wed, 9 Jun 2021 22:22:47 +0300
Subject: [PATCH 052/178] fix wording (#6108)

---
 docs/IE_DG/Intro_to_Performance.md | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/docs/IE_DG/Intro_to_Performance.md b/docs/IE_DG/Intro_to_Performance.md
index 94d0173dbbe2b4..0c9457ed4bfd79 100644
--- a/docs/IE_DG/Intro_to_Performance.md
+++ b/docs/IE_DG/Intro_to_Performance.md
@@ -22,7 +22,8 @@ $ benchmark_app -m <model.xml> -enforcebf16=false
 Notice that for quantized (e.g. INT8) models the bfloat16 calculations (of the layers that remain in FP32) is disabled by default.
 Refer to the [CPU Plugin documentation](supported_plugins/CPU.md) for more details.
 
-Similarly, the GPU device has a dedicated config key to enable FP16 execution of the layers that remain in FP32 in the quantized models (as the quantization is typically performed on the FP32 models), refer to the ENABLE_FP16_FOR_QUANTIZED_MODELS key in the [GPU Plugin documentation](supported_plugins/GPU.md)
+Similarly, the GPU device automatically executes FP16 for the layers that remain in FP16 in the quantized models (assuming that the FP16 model was quantized).
+Refer to the ENABLE_FP16_FOR_QUANTIZED_MODELS key in the [GPU Plugin documentation](supported_plugins/GPU.md).
 
 ## Latency vs. Throughput
 One way to increase computational efficiency is batching, which combines many (potentially tens) of

From 6ec725baac0113570e34c955ca3dfff2ac14b13b Mon Sep 17 00:00:00 2001
From: Ilya Znamenskiy <ilya.znamenskiy@intel.com>
Date: Wed, 9 Jun 2021 23:39:51 +0300
Subject: [PATCH 053/178] Revert "[IE CLDNN] Use fsv4 when feature-depth is
 shallow (#4398)" (#6103)

---
 .../thirdparty/clDNN/src/program.cpp          | 34 -------------------
 1 file changed, 34 deletions(-)

diff --git a/inference-engine/thirdparty/clDNN/src/program.cpp b/inference-engine/thirdparty/clDNN/src/program.cpp
index ef988cb12dac55..97eacdf9ae1122 100644
--- a/inference-engine/thirdparty/clDNN/src/program.cpp
+++ b/inference-engine/thirdparty/clDNN/src/program.cpp
@@ -1176,9 +1176,6 @@ void program_impl::set_layout_optimizer_attributes(layout_optimizer& lo) {
     size_t opt_deconv_layers_b_fs_zyx_fsv16 = 0;
     size_t total_crop_layers = 0;
 
-    size_t weighted_sum_feature_size = 0;
-    size_t weight_sum = 0;
-
     for (auto& node : get_processing_order()) {
         auto &prim = *node;
         if (prim.type() == cldnn::convolution::type_id()) {
@@ -1324,35 +1321,4 @@ void program_impl::set_layout_optimizer_attributes(layout_optimizer& lo) {
 
     if (should_use_bs_fs_yx_bsv16_fsv16)
         lo.set_optimization_attribute(layout_optimizer::optimization_attributes_type::bs_fs_yx_bsv16_fsv16_network, 1);
-
-
-    // This is to avoid using fsv16 for shallow-feature networks.
-    // This may not be exactly same as real execution graph as layer fusing is not done yet,
-    // but it is a reasonable approximation.
-    // Check the expected network efficiency after setting layer optimization attributes.
-    // If network depth is shallow, it is faster with fsv4.
-    for (auto& node : get_processing_order()) {
-        auto &prim = *node;
-
-        if (prim.is_in_data_flow() && prim.type() == cldnn::convolution::type_id()) {
-            size_t num_feature = prim.get_output_layout().size.feature.vector()[0];
-            size_t num_spatial = 1;
-            for (auto s : prim.get_output_layout().size.spatial.vector())
-                num_spatial *= s;
-
-            if (lo.get_preferred_format(prim) != format::b_fs_yx_fsv4) {
-                weight_sum += num_spatial;
-                weighted_sum_feature_size += num_spatial * num_feature;
-            }
-        }
-    }
-
-    size_t weighted_average_feature_depth = weighted_sum_feature_size / std::max(weight_sum, static_cast<size_t>(1));
-
-    // Need to confirm that weighted_average_feature_depth > 1 to keep unittest behavior.
-    if (is_quantized_int8_model && weighted_average_feature_depth < 8 && weighted_average_feature_depth > 1) {
-        lo.set_optimization_attribute(layout_optimizer::optimization_attributes_type::fs_b_yx_fsv32_network, 0);
-        lo.set_optimization_attribute(layout_optimizer::optimization_attributes_type::b_fs_yx_fsv16_network, 0);
-        lo.set_optimization_attribute(layout_optimizer::optimization_attributes_type::bs_fs_yx_bsv16_fsv16_network, 0);
-    }
 }

From 17a899ed0c2eff8d6b289203dcb532e3f775a784 Mon Sep 17 00:00:00 2001
From: Alexander Zhogov <alexander.zhogov@intel.com>
Date: Thu, 10 Jun 2021 10:33:47 +0300
Subject: [PATCH 054/178] Change version to 2022.1

---
 inference-engine/include/ie_version.hpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/inference-engine/include/ie_version.hpp b/inference-engine/include/ie_version.hpp
index 13215d0b68d253..10e649a09d32f8 100644
--- a/inference-engine/include/ie_version.hpp
+++ b/inference-engine/include/ie_version.hpp
@@ -20,8 +20,8 @@
  * @brief Defines Inference Engine patch version
  */
 
-#define IE_VERSION_MAJOR 2021
-#define IE_VERSION_MINOR 4
+#define IE_VERSION_MAJOR 2022
+#define IE_VERSION_MINOR 1
 #define IE_VERSION_PATCH 0
 
 #include "ie_api.h"

From 9f1720b0e4c73437818ed5c8ca0a827c42875eeb Mon Sep 17 00:00:00 2001
From: Tomasz Socha <tomasz.socha@intel.com>
Date: Thu, 10 Jun 2021 10:54:42 +0200
Subject: [PATCH 055/178] [ONNX Importer] Remove quantized matmuls (#5043)

---
 ngraph/frontend/onnx_import/CMakeLists.txt    |  4 --
 .../onnx_import/src/op/matmul_integer.cpp     | 41 -------------------
 .../onnx_import/src/op/matmul_integer.hpp     | 32 ---------------
 .../onnx_import/src/op/qlinear_matmul.cpp     | 41 -------------------
 .../onnx_import/src/op/qlinear_matmul.hpp     | 25 -----------
 .../frontend/onnx_import/src/ops_bridge.cpp   |  4 --
 6 files changed, 147 deletions(-)
 delete mode 100644 ngraph/frontend/onnx_import/src/op/matmul_integer.cpp
 delete mode 100644 ngraph/frontend/onnx_import/src/op/matmul_integer.hpp
 delete mode 100644 ngraph/frontend/onnx_import/src/op/qlinear_matmul.cpp
 delete mode 100644 ngraph/frontend/onnx_import/src/op/qlinear_matmul.hpp

diff --git a/ngraph/frontend/onnx_import/CMakeLists.txt b/ngraph/frontend/onnx_import/CMakeLists.txt
index 9f728a3f7e808c..f3ef7b10fe799f 100644
--- a/ngraph/frontend/onnx_import/CMakeLists.txt
+++ b/ngraph/frontend/onnx_import/CMakeLists.txt
@@ -11,14 +11,10 @@ file(GLOB_RECURSE LIBRARY_PUBLIC_HEADERS ${CMAKE_CURRENT_SOURCE_DIR}/include/*.h
 # Remove disabled ops
 list(REMOVE_ITEM LIBRARY_SRC
     ${CMAKE_CURRENT_SOURCE_DIR}/src/op/conv_integer.cpp
-    ${CMAKE_CURRENT_SOURCE_DIR}/src/op/matmul_integer.cpp
-    ${CMAKE_CURRENT_SOURCE_DIR}/src/op/qlinear_matmul.cpp
     ${CMAKE_CURRENT_SOURCE_DIR}/src/op/quant_conv.cpp
     )
 list(REMOVE_ITEM LIBRARY_HEADERS
     ${CMAKE_CURRENT_SOURCE_DIR}/src/op/conv_integer.hpp
-    ${CMAKE_CURRENT_SOURCE_DIR}/src/op/matmul_integer.hpp
-    ${CMAKE_CURRENT_SOURCE_DIR}/src/op/qlinear_matmul.hpp
     ${CMAKE_CURRENT_SOURCE_DIR}/src/op/quant_conv.hpp
     )
 
diff --git a/ngraph/frontend/onnx_import/src/op/matmul_integer.cpp b/ngraph/frontend/onnx_import/src/op/matmul_integer.cpp
deleted file mode 100644
index 231256950f9aba..00000000000000
--- a/ngraph/frontend/onnx_import/src/op/matmul_integer.cpp
+++ /dev/null
@@ -1,41 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "op/matmul_integer.hpp"
-#include "ngraph/builder/matmul_factory.hpp"
-#include "ngraph/log.hpp"
-
-namespace ngraph
-{
-    namespace onnx_import
-    {
-        namespace op
-        {
-            namespace set_1
-            {
-                OutputVector matmul_integer(const Node& node)
-                {
-                    auto ng_inputs = node.get_ng_inputs();
-                    auto factory = builder::MatmulIntegerFactory(
-                        OutputVector(std::begin(ng_inputs), std::end(ng_inputs)));
-                    std::size_t left_rank{ng_inputs.at(0).get_shape().size()};
-                    std::size_t right_rank{ng_inputs.at(1).get_shape().size()};
-
-                    if (left_rank == 0 || right_rank == 0)
-                    {
-                        NGRAPH_WARN
-                            << (node) << " "
-                            << "ONNX standard doesn't allow scalar operands, however nGraph "
-                               "accepts them. Consider use of element-wise multiplication instead "
-                               "to conform with ONNX standard.";
-                    }
-                    return factory.make_matmul_op();
-                }
-            } // namespace set_1
-
-        } // namespace op
-
-    } // namespace onnx_import
-
-} // namespace ngraph
diff --git a/ngraph/frontend/onnx_import/src/op/matmul_integer.hpp b/ngraph/frontend/onnx_import/src/op/matmul_integer.hpp
deleted file mode 100644
index 8e40e3b608b83b..00000000000000
--- a/ngraph/frontend/onnx_import/src/op/matmul_integer.hpp
+++ /dev/null
@@ -1,32 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-
-#include "ngraph/node.hpp"
-#include "onnx_import/core/node.hpp"
-
-namespace ngraph
-{
-    namespace onnx_import
-    {
-        namespace op
-        {
-            namespace set_1
-            {
-                /// \brief Performs ONNX MatMulInteger operation.
-                ///
-                /// \param node The ONNX node object representing this operation.
-                ///
-                /// \return The vector containing Ngraph nodes producing output of quantized ONNX
-                ///         matrix multiplication operation.
-                OutputVector matmul_integer(const Node& node);
-
-            } // namespace set_1
-
-        } // namespace op
-
-    } // namespace onnx_import
-
-} // namespace ngraph
diff --git a/ngraph/frontend/onnx_import/src/op/qlinear_matmul.cpp b/ngraph/frontend/onnx_import/src/op/qlinear_matmul.cpp
deleted file mode 100644
index 1baafa827ca1b2..00000000000000
--- a/ngraph/frontend/onnx_import/src/op/qlinear_matmul.cpp
+++ /dev/null
@@ -1,41 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "op/qlinear_matmul.hpp"
-#include "ngraph/builder/matmul_factory.hpp"
-#include "ngraph/log.hpp"
-
-namespace ngraph
-{
-    namespace onnx_import
-    {
-        namespace op
-        {
-            namespace set_1
-            {
-                OutputVector qlinear_matmul(const Node& node)
-                {
-                    auto ng_inputs = node.get_ng_inputs();
-                    auto factory = builder::QLinearMatmulFactory(
-                        (OutputVector(std::begin(ng_inputs), std::end(ng_inputs))));
-                    std::size_t left_rank{ng_inputs.at(0).get_shape().size()};
-                    std::size_t right_rank{ng_inputs.at(1).get_shape().size()};
-
-                    if (left_rank == 0 || right_rank == 0)
-                    {
-                        NGRAPH_WARN
-                            << (node) << " "
-                            << "ONNX standard doesn't allow scalar operands, however nGraph "
-                               "accepts them. Consider use of element-wise multiplication instead "
-                               "to conform with ONNX standard.";
-                    }
-                    return factory.make_matmul_op();
-                }
-            } // namespace set_1
-
-        } // namespace op
-
-    } // namespace onnx_import
-
-} // namespace ngraph
diff --git a/ngraph/frontend/onnx_import/src/op/qlinear_matmul.hpp b/ngraph/frontend/onnx_import/src/op/qlinear_matmul.hpp
deleted file mode 100644
index 8e38f33683654b..00000000000000
--- a/ngraph/frontend/onnx_import/src/op/qlinear_matmul.hpp
+++ /dev/null
@@ -1,25 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-
-#include "ngraph/node.hpp"
-#include "onnx_import/core/node.hpp"
-
-namespace ngraph
-{
-    namespace onnx_import
-    {
-        namespace op
-        {
-            namespace set_1
-            {
-                OutputVector qlinear_matmul(const Node& node);
-            } // namespace set_1
-
-        } // namespace op
-
-    } // namespace onnx_import
-
-} // namespace ngraph
diff --git a/ngraph/frontend/onnx_import/src/ops_bridge.cpp b/ngraph/frontend/onnx_import/src/ops_bridge.cpp
index b807a31212db64..f7864360ba7b43 100644
--- a/ngraph/frontend/onnx_import/src/ops_bridge.cpp
+++ b/ngraph/frontend/onnx_import/src/ops_bridge.cpp
@@ -73,7 +73,6 @@
 #include "op/lrn.hpp"
 #include "op/lstm.hpp"
 #include "op/matmul.hpp"
-//#include "op/matmul_integer.hpp"
 #include "op/max.hpp"
 #include "op/max_pool.hpp"
 #include "op/mean.hpp"
@@ -90,7 +89,6 @@
 #include "op/pad.hpp"
 #include "op/pow.hpp"
 #include "op/prelu.hpp"
-//#include "op/qlinear_matmul.hpp"
 // #include "op/quant_conv.hpp"
 #include "op/quantize_linear.hpp"
 #include "op/range.hpp"
@@ -376,7 +374,6 @@ namespace ngraph
             REGISTER_OPERATOR("LRN", 1, lrn);
             REGISTER_OPERATOR("LSTM", 1, lstm);
             REGISTER_OPERATOR("MatMul", 1, matmul);
-            // REGISTER_OPERATOR("MatMulInteger", 1, matmul_integer);
             REGISTER_OPERATOR("MaxPool", 1, max_pool);
             REGISTER_OPERATOR("Max", 1, max);
             REGISTER_OPERATOR("Max", 8, max);
@@ -399,7 +396,6 @@ namespace ngraph
             REGISTER_OPERATOR("Pow", 1, pow);
             REGISTER_OPERATOR("PRelu", 1, prelu);
             // REGISTER_OPERATOR("QLinearConv", 1, quant_conv);
-            // REGISTER_OPERATOR("QLinearMatMul", 1, qlinear_matmul);
             REGISTER_OPERATOR("QuantizeLinear", 1, quantize_linear);
             REGISTER_OPERATOR("QuantizeLinear", 13, quantize_linear);
             REGISTER_OPERATOR("Range", 1, range);

From 6040d433d395ea3df5242fcef3e62f31136cbec4 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Tomasz=20Do=C5=82bniak?= <tomasz.dolbniak@intel.com>
Date: Thu, 10 Jun 2021 11:06:30 +0200
Subject: [PATCH 056/178] Fix for the paddings attribute handling for ONNX
 Pad@1 (#6101)

---
 ngraph/frontend/onnx_import/src/utils/convpool.cpp | 5 +++++
 ngraph/python/tests/__init__.py                    | 1 -
 ngraph/python/tests/test_onnx/test_ops_convpool.py | 5 ++---
 3 files changed, 7 insertions(+), 4 deletions(-)

diff --git a/ngraph/frontend/onnx_import/src/utils/convpool.cpp b/ngraph/frontend/onnx_import/src/utils/convpool.cpp
index 64644c5e11e96b..52924d78116334 100644
--- a/ngraph/frontend/onnx_import/src/utils/convpool.cpp
+++ b/ngraph/frontend/onnx_import/src/utils/convpool.cpp
@@ -130,6 +130,11 @@ namespace ngraph
                     auto pads_int64 = node.get_attribute_value<std::vector<int64_t>>("pads");
                     pads = CoordinateDiff{std::begin(pads_int64), std::end(pads_int64)};
                 }
+                else if (node.has_attribute("paddings"))
+                {
+                    auto pads_int64 = node.get_attribute_value<std::vector<int64_t>>("paddings");
+                    pads = CoordinateDiff{std::begin(pads_int64), std::end(pads_int64)};
+                }
 
                 if (pads.size() == kernel_rank * 2)
                 {
diff --git a/ngraph/python/tests/__init__.py b/ngraph/python/tests/__init__.py
index c179f312f1e7e8..f54930f40d92e5 100644
--- a/ngraph/python/tests/__init__.py
+++ b/ngraph/python/tests/__init__.py
@@ -47,7 +47,6 @@ def xfail_test(reason="Mark the test as expected to fail", strict=True):
 xfail_issue_33581 = xfail_test(reason="RuntimeError: nGraph does not support the following ONNX operations:"
                                       "GatherElements")
 xfail_issue_33633 = xfail_test(reason="MaxPool: dilations unsupported")
-xfail_issue_35911 = xfail_test(reason="Assertion error: Pad model mismatch error")
 xfail_issue_35923 = xfail_test(reason="RuntimeError: PReLU without weights is not supported")
 xfail_issue_35927 = xfail_test(reason="RuntimeError: B has zero dimension that is not allowable")
 xfail_issue_36486 = xfail_test(reason="RuntimeError: HardSigmoid operation should be converted "
diff --git a/ngraph/python/tests/test_onnx/test_ops_convpool.py b/ngraph/python/tests/test_onnx/test_ops_convpool.py
index 7a507a2b4e000e..6637a06463b19c 100644
--- a/ngraph/python/tests/test_onnx/test_ops_convpool.py
+++ b/ngraph/python/tests/test_onnx/test_ops_convpool.py
@@ -5,10 +5,10 @@
 import onnx
 import pytest
 from onnx.helper import make_graph, make_model, make_node, make_tensor_value_info
+from onnx.onnx_cpp2py_export.checker import ValidationError
 
 from tests.runtime import get_runtime
 from tests.test_onnx.utils import get_node_model, import_onnx_model, run_model, run_node
-from tests import xfail_issue_35911
 
 
 @pytest.fixture
@@ -257,7 +257,6 @@ def test_2d_conv_transpose():
     )
 
 
-@xfail_issue_35911
 def test_pad_opset_1():
     x = np.ones((2, 2), dtype=np.float32)
     y = np.pad(x, pad_width=1, mode="constant")
@@ -281,7 +280,7 @@ def test_pad_opset_1():
 
     # no paddings arttribute
     model = get_node_model("Pad", x)
-    with pytest.raises(RuntimeError):
+    with pytest.raises(ValidationError):
         import_onnx_model(model)
 
 

From c7141973ece493f932c6b8f305372328d9be1aae Mon Sep 17 00:00:00 2001
From: Alexander Zhogov <alexander.zhogov@intel.com>
Date: Thu, 10 Jun 2021 12:46:24 +0300
Subject: [PATCH 057/178] GitHub CI: Fix checking commits in case of no email

---
 .github/org_control/check_pr.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.github/org_control/check_pr.py b/.github/org_control/check_pr.py
index 85e5b4e34c00f4..7bb8f89dd602f4 100644
--- a/.github/org_control/check_pr.py
+++ b/.github/org_control/check_pr.py
@@ -139,7 +139,7 @@ def update_labels(gh_api, pull, non_org_intel_pr_users, non_org_pr_users):
 
 def get_wrong_commits(pull):
     """Returns commits with incorrect user and email"""
-    pr_author_email = pull.user.email.lower()
+    pr_author_email = (pull.user.email or "").lower()
     print("GitHub PR author email:", pr_author_email)
     print("Check commits:")
     wrong_commits = set()
@@ -147,7 +147,7 @@ def get_wrong_commits(pull):
         # import pprint; pprint.pprint(commit.raw_data)
         print("Commit SHA:", commit.sha)
         # Use raw data because commit author can be non GitHub user
-        commit_email = commit.raw_data["commit"]["author"]["email"].lower()
+        commit_email = (commit.raw_data["commit"]["author"]["email"] or "").lower()
         print("    Commit email:", commit_email)
         if not github_api.is_valid_user(commit.author):
             print(

From c3ad7dc2a44b571a4c75c3e8809a2b5b8a1c16b2 Mon Sep 17 00:00:00 2001
From: Maxim Vafin <maxim.vafin@intel.com>
Date: Thu, 10 Jun 2021 18:09:35 +0300
Subject: [PATCH 058/178] Specify AdaptiveAvgPool-8 and AdaptiveMaxPool-8
 (#5773)

* Specify AdaptiveAvgPool-8 and AdaptiveMaxPool-8

* Apply review feedback

* Describe indices output

* Apply review feedback

* Apply review feedback

* Update AdaptiveMaxPool_8.md

* Apply suggestions from code review

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Apply review feedback

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
---
 docs/doxygen/ie_docs.xml              |   3 +
 docs/ops/opset8.md                    | 169 ++++++++++++++++++++++++++
 docs/ops/pooling/AdaptiveAvgPool_8.md |  70 +++++++++++
 docs/ops/pooling/AdaptiveMaxPool_8.md |  87 +++++++++++++
 4 files changed, 329 insertions(+)
 create mode 100644 docs/ops/opset8.md
 create mode 100644 docs/ops/pooling/AdaptiveAvgPool_8.md
 create mode 100644 docs/ops/pooling/AdaptiveMaxPool_8.md

diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index bb006c9f01c630..503f9f3898601b 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -88,6 +88,7 @@ limitations under the License.
         <!-- Intermediate Representation and Operations Sets -->
         <tab id="intermediate_representaton_and_operations_sets" type="usergroup" title="Intermediate Representation and Operations Sets" url="@ref openvino_docs_MO_DG_IR_and_opsets">
             <tab type="usergroup" title="Available Operations Sets" url="@ref openvino_docs_ops_opset">
+                <tab type="user" title="opset8 Specification" url="@ref openvino_docs_ops_opset8"/>
                 <tab type="user" title="opset7 Specification" url="@ref openvino_docs_ops_opset7"/>
                 <tab type="user" title="opset6 Specification" url="@ref openvino_docs_ops_opset6"/>
                 <tab type="user" title="opset5 Specification" url="@ref openvino_docs_ops_opset5"/>
@@ -100,6 +101,8 @@ limitations under the License.
                 <tab type="user" title="Abs-1" url="@ref openvino_docs_ops_arithmetic_Abs_1"/>
                 <tab type="user" title="Acos-1" url="@ref openvino_docs_ops_arithmetic_Acos_1"/>
                 <tab type="user" title="Acosh-3" url="@ref openvino_docs_ops_arithmetic_Acosh_3"/>
+                <tab type="user" title="AdaptiveAvgPool-8" url="@ref openvino_docs_ops_pooling_AdaptiveAvgPool_8"/>
+                <tab type="user" title="AdaptiveMaxPool-8" url="@ref openvino_docs_ops_pooling_AdaptiveMaxPool_8"/>
                 <tab type="user" title="Add-1" url="@ref openvino_docs_ops_arithmetic_Add_1"/>
                 <tab type="user" title="Asin-1" url="@ref openvino_docs_ops_arithmetic_Asin_1"/>
                 <tab type="user" title="Asinh-3" url="@ref openvino_docs_ops_arithmetic_Asinh_3"/>
diff --git a/docs/ops/opset8.md b/docs/ops/opset8.md
new file mode 100644
index 00000000000000..8f43927b5ec360
--- /dev/null
+++ b/docs/ops/opset8.md
@@ -0,0 +1,169 @@
+# Operation Set `opset8` Specification {#openvino_docs_ops_opset8}
+
+This specification document describes the `opset8` operation set supported in OpenVINO™.
+Support for each particular operation from the list below depends on the capabilities of an inference plugin
+and may vary among different hardware platforms and devices. Examples of operation instances are provided as IR V10 xml
+snippets. Such IR is generated by the Model Optimizer. The semantics match corresponding nGraph operation classes
+declared in `namespace opset8`.
+
+
+## Table of Contents <a name="toc"></a>
+
+* [Abs](arithmetic/Abs_1.md)
+* [Acos](arithmetic/Acos_1.md)
+* [Acosh](arithmetic/Acosh_3.md)
+* [AdaptiveAvgPool](pooling/AdaptiveAvgPool_8.md)
+* [AdaptiveMaxPool](pooling/AdaptiveMaxPool_8.md)
+* [Add](arithmetic/Add_1.md)
+* [Asin](arithmetic/Asin_1.md)
+* [Asinh](arithmetic/Asinh_3.md)
+* [Assign](infrastructure/Assign_3.md)
+* [Atan](arithmetic/Atan_1.md)
+* [Atanh](arithmetic/Atanh_3.md)
+* [AvgPool](pooling/AvgPool_1.md)
+* [BatchNormInference](normalization/BatchNormInference_5.md)
+* [BatchToSpace](movement/BatchToSpace_2.md)
+* [BinaryConvolution](convolution/BinaryConvolution_1.md)
+* [Broadcast](movement/Broadcast_3.md)
+* [Bucketize](condition/Bucketize_3.md)
+* [CTCGreedyDecoder](sequence/CTCGreedyDecoder_1.md)
+* [CTCGreedyDecoderSeqLen](sequence/CTCGreedyDecoderSeqLen_6.md)
+* [CTCLoss](sequence/CTCLoss_4.md)
+* [Ceiling](arithmetic/Ceiling_1.md)
+* [Clamp](activation/Clamp_1.md)
+* [Concat](movement/Concat_1.md)
+* [Constant](infrastructure/Constant_1.md)
+* [Convert](type/Convert_1.md)
+* [ConvertLike](type/ConvertLike_1.md)
+* [Convolution](convolution/Convolution_1.md)
+* [ConvolutionBackpropData](convolution/ConvolutionBackpropData_1.md)
+* [Cos](arithmetic/Cos_1.md)
+* [Cosh](arithmetic/Cosh_1.md)
+* [CumSum](arithmetic/CumSum_3.md)
+* [DeformableConvolution](convolution/DeformableConvolution_1.md)
+* [DeformablePSROIPooling](detection/DeformablePSROIPooling_1.md)
+* [DepthToSpace](movement/DepthToSpace_1.md)
+* [DetectionOutput](detection/DetectionOutput_1.md)
+* [DFT](signals/DFT_7.md)
+* [Divide](arithmetic/Divide_1.md)
+* [Einsum](matrix/Einsum_7.md)
+* [Elu](activation/Elu_1.md)
+* [EmbeddingBagOffsetsSum](sparse/EmbeddingBagOffsetsSum_3.md)
+* [EmbeddingBagPackedSum](sparse/EmbeddingBagPackedSum_3.md)
+* [EmbeddingSegmentsSum](sparse/EmbeddingSegmentsSum_3.md)
+* [Equal](comparison/Equal_1.md)
+* [Erf](arithmetic/Erf_1.md)
+* [Exp](activation/Exp_1.md)
+* [ExperimentalDetectronDetectionOutput_6](detection/ExperimentalDetectronDetectionOutput_6.md)
+* [ExperimentalDetectronGenerateProposalsSingleImage_6](detection/ExperimentalDetectronGenerateProposalsSingleImage_6.md)
+* [ExperimentalDetectronPriorGridGenerator_6](detection/ExperimentalDetectronPriorGridGenerator_6.md)
+* [ExperimentalDetectronROIFeatureExtractor_6](detection/ExperimentalDetectronROIFeatureExtractor_6.md)
+* [ExperimentalDetectronTopKROIs_6](sort/ExperimentalDetectronTopKROIs_6.md)
+* [ExtractImagePatches](movement/ExtractImagePatches_3.md)
+* [FakeQuantize](quantization/FakeQuantize_1.md)
+* [Floor](arithmetic/Floor_1.md)
+* [FloorMod](arithmetic/FloorMod_1.md)
+* [Gather](movement/Gather_7.md)
+* [GatherElements](movement/GatherElements_6.md)
+* [GatherND_5](movement/GatherND_5.md)
+* [GatherTree](movement/GatherTree_1.md)
+* [Gelu](activation/GELU_7.md)
+* [Greater](comparison/Greater_1.md)
+* [GreaterEqual](comparison/GreaterEqual_1.md)
+* [GRN](normalization/GRN_1.md)
+* [GroupConvolution](convolution/GroupConvolution_1.md)
+* [GroupConvolutionBackpropData](convolution/GroupConvolutionBackpropData_1.md)
+* [GRUCell](sequence/GRUCell_3.md)
+* [GRUSequence](sequence/GRUSequence_5.md)
+* [HardSigmoid](activation/HardSigmoid_1.md)
+* [HSigmoid](activation/HSigmoid_5.md)
+* [HSwish](activation/HSwish_4.md)
+* [IDFT](signals/IDFT_7.md)
+* [Interpolate](image/Interpolate_4.md)
+* [Less](comparison/Less_1.md)
+* [LessEqual](comparison/LessEqual_1.md)
+* [Log](arithmetic/Log_1.md)
+* [LogicalAnd](logical/LogicalAnd_1.md)
+* [LogicalNot](logical/LogicalNot_1.md)
+* [LogicalOr](logical/LogicalOr_1.md)
+* [LogicalXor](logical/LogicalXor_1.md)
+* [LogSoftmax](activation/LogSoftmax_5.md)
+* [Loop](infrastructure/Loop_5.md)
+* [LRN](normalization/LRN_1.md)
+* [LSTMCell](sequence/LSTMCell_1.md)
+* [LSTMSequence](sequence/LSTMSequence_1.md)
+* [MatMul](matrix/MatMul_1.md)
+* [MaxPool](pooling/MaxPool_1.md)
+* [Maximum](arithmetic/Maximum_1.md)
+* [Minimum](arithmetic/Minimum_1.md)
+* [Mish](activation/Mish_4.md)
+* [Mod](arithmetic/Mod_1.md)
+* [MVN](normalization/MVN_6.md)
+* [Multiply](arithmetic/Multiply_1.md)
+* [Negative](arithmetic/Negative_1.md)
+* [NonMaxSuppression](sort/NonMaxSuppression_5.md)
+* [NonZero](condition/NonZero_3.md)
+* [NormalizeL2](normalization/NormalizeL2_1.md)
+* [NotEqual](comparison/NotEqual_1.md)
+* [OneHot](sequence/OneHot_1.md)
+* [Pad](movement/Pad_1.md)
+* [Parameter](infrastructure/Parameter_1.md)
+* [Power](arithmetic/Power_1.md)
+* [PReLU](activation/PReLU_1.md)
+* [PriorBoxClustered](detection/PriorBoxClustered_1.md)
+* [PriorBox](detection/PriorBox_1.md)
+* [Proposal](detection/Proposal_4.md)
+* [PSROIPooling](detection/PSROIPooling_1.md)
+* [Range](generation/Range_4.md)
+* [ReLU](activation/ReLU_1.md)
+* [ReadValue](infrastructure/ReadValue_3.md)
+* [ReduceL1](reduction/ReduceL1_4.md)
+* [ReduceL2](reduction/ReduceL2_4.md)
+* [ReduceLogicalAnd](reduction/ReduceLogicalAnd_1.md)
+* [ReduceLogicalOr](reduction/ReduceLogicalOr_1.md)
+* [ReduceMax](reduction/ReduceMax_1.md)
+* [ReduceMean](reduction/ReduceMean_1.md)
+* [ReduceMin](reduction/ReduceMin_1.md)
+* [ReduceProd](reduction/ReduceProd_1.md)
+* [ReduceSum](reduction/ReduceSum_1.md)
+* [RegionYolo](detection/RegionYolo_1.md)
+* [ReorgYolo](detection/ReorgYolo_1.md)
+* [Reshape](shape/Reshape_1.md)
+* [Result](infrastructure/Result_1.md)
+* [ReverseSequence](movement/ReverseSequence_1.md)
+* [RNNCell](sequence/RNNCell_3.md)
+* [RNNSequence](sequence/RNNSequence_5.md)
+* [ROIAlign](detection/ROIAlign_3.md)
+* [ROIPooling](detection/ROIPooling_1.md)
+* [Roll](movement/Roll_7.md)
+* [Round](arithmetic/Round_5.md)
+* [ScatterElementsUpdate](movement/ScatterElementsUpdate_3.md)
+* [ScatterNDUpdate](movement/ScatterNDUpdate_3.md)
+* [ScatterUpdate](movement/ScatterUpdate_3.md)
+* [Select](condition/Select_1.md)
+* [Selu](activation/Selu_1.md)
+* [ShapeOf](shape/ShapeOf_3.md)
+* [ShuffleChannels](movement/ShuffleChannels_1.md)
+* [Sigmoid](activation/Sigmoid_1.md)
+* [Sign](arithmetic/Sign_1.md)
+* [Sin](arithmetic/Sin_1.md)
+* [Sinh](arithmetic/Sinh_1.md)
+* [SoftMax](activation/SoftMax_1.md)
+* [SoftPlus](activation/SoftPlus_4.md)
+* [SpaceToBatch](movement/SpaceToBatch_2.md)
+* [SpaceToDepth](movement/SpaceToDepth_1.md)
+* [Split](movement/Split_1.md)
+* [Sqrt](arithmetic/Sqrt_1.md)
+* [SquaredDifference](arithmetic/SquaredDifference_1.md)
+* [Squeeze](shape/Squeeze_1.md)
+* [StridedSlice](movement/StridedSlice_1.md)
+* [Subtract](arithmetic/Subtract_1.md)
+* [Swish](activation/Swish_4.md)
+* [Tan](arithmetic/Tan_1.md)
+* [Tanh](arithmetic/Tanh_1.md)
+* [TensorIterator](infrastructure/TensorIterator_1.md)
+* [Tile](movement/Tile_1.md)
+* [TopK](sort/TopK_3.md)
+* [Transpose](movement/Transpose_1.md)
+* [Unsqueeze](shape/Unsqueeze_1.md)
+* [VariadicSplit](movement/VariadicSplit_1.md)
diff --git a/docs/ops/pooling/AdaptiveAvgPool_8.md b/docs/ops/pooling/AdaptiveAvgPool_8.md
new file mode 100644
index 00000000000000..beb2ec30492d3e
--- /dev/null
+++ b/docs/ops/pooling/AdaptiveAvgPool_8.md
@@ -0,0 +1,70 @@
+## AdaptiveAvgPool<a name="AdaptiveAvgPool"></a> {#openvino_docs_ops_pooling_AdaptiveAvgPool_8}
+
+**Versioned name**: *AdaptiveAvgPool-8*
+
+**Category**: *Pooling*
+
+**Short description**: Applies average pooling with adaptive kernel size over the input.
+
+**Detailed description**: This operation calculates the output based on the first input and `output_size` determined by the second input.
+The kernel dimensions are calculated using the following formulae for the `NCDHW` input case:
+
+\f[
+\begin{array}{lcl}
+d_{start} &=& floor(i*D_{in}/D_{out})\\
+d_{end}   &=& ceil((i+1)*D_{in}/D_{out})\\
+h_{start} &=& floor(j*H_{in}/H_{out})\\
+h_{end}   &=& ceil((j+1)*H_{in}/H_{out})\\
+w_{start} &=& floor(k*W_{in}/W_{out})\\
+w_{end}   &=& ceil((k+1)*W_{in}/W_{out})
+\end{array}
+\f]
+
+The output is calculated with the following formula:
+
+\f[
+Output(i,j,k) = \frac{Input[d_{start}:d_{end}, h_{start}:h_{end}, w_{start}:w_{end}]}{(d_{end}-d_{start})*(h_{end}-h_{start})*(w_{end}-w_{start})}
+\f]
+
+**Inputs**:
+
+*   **1**: 3D, 4D, or 5D input tensor of shape `[N, C, H]`, `[N, C, H, W]` or `[N, C, D, H, W]` and type *T*. Required.
+*   **2**: 1D tensor describing output shape for spatial dimensions. Can be `[H_out]` for 3D input, `[H_out, W_out]` for 4D input, `[D_out, H_out, W_out]` for 5D input and of type *T_SHAPE*. Required.
+
+**Outputs**:
+
+*   **1**: Output of type *T* and shape `[N, C, H_out]`, `[N, C, H_out, W_out]` or `[N, C, D_out, H_out, W_out]`.
+
+**Types**
+
+*   *T*: floating-point type.
+*   *T_SHAPE*: `int32` or `int64`.
+
+**Examples**
+
+```xml
+<layer ... type="AdaptiveAvgPool" ... >
+    <data output_type="i64"/>
+    <input> 
+        <port id="0">
+            <dim>1</dim>
+            <dim>3</dim>
+            <dim>32</dim>
+            <dim>32</dim>
+        </port>
+    </input>
+    <input> 
+        <port id="1">
+            <dim>2</dim>
+        </port>
+    </input>
+    <output>
+        <port id="2">
+            <dim>1</dim>
+            <dim>3</dim>
+            <dim>16</dim>
+            <dim>16</dim>
+        </port>
+    </output>
+</layer>
+```
diff --git a/docs/ops/pooling/AdaptiveMaxPool_8.md b/docs/ops/pooling/AdaptiveMaxPool_8.md
new file mode 100644
index 00000000000000..d7ad9a42412bd9
--- /dev/null
+++ b/docs/ops/pooling/AdaptiveMaxPool_8.md
@@ -0,0 +1,87 @@
+## AdaptiveMaxPool<a name="AdaptiveMaxPool"></a> {#openvino_docs_ops_pooling_AdaptiveMaxPool_8}
+
+**Versioned name**: *AdaptiveMaxPool-8*
+
+**Category**: *Pooling*
+
+**Short description**: Applies max pooling with adaptive kernel size over the input.
+
+**Detailed description**: This operation calculates the output based on the first input and `output_size` determined by the second input.
+The kernel dimensions are calculated using the following formulae for the `NCDHW` input case:
+
+\f[
+\begin{array}{lcl}
+d_{start} &=& floor(i*D_{in}/D_{out})\\
+d_{end}   &=& ceil((i+1)*D_{in}/D_{out})\\
+h_{start} &=& floor(j*H_{in}/H_{out})\\
+h_{end}   &=& ceil((j+1)*H_{in}/H_{out})\\
+w_{start} &=& floor(k*W_{in}/W_{out})\\
+w_{end}   &=& ceil((k+1)*W_{in}/W_{out})
+\end{array}
+\f]
+
+The output is calculated following this formula:
+
+\f[
+Output(i,j,k) = max(Input[d_{start}:d_{end}, h_{start}:h_{end}, w_{start}:w_{end}])
+\f]
+
+**Attributes**:
+
+*   *index_element_type*
+
+  * **Description**: the type of the second output containing indices
+  * **Range of values**: "i64" or "i32"
+  * **Type**: string
+  * **Default value**: "i64"
+  * **Required**: *No*
+
+**Inputs**:
+
+*   **1**: 3D, 4D, or 5D input tensor of shape `[N, C, H]`, `[N, C, H, W]` or `[N, C, D, H, W]` and type *T*. Required.
+*   **2**: 1D tensor describing output shape for spatial dimensions. Can be `[H_out]` for 3D input, `[H_out, W_out]` for 4D input, `[D_out, H_out, W_out]` for 5D input and of type *T_SHAPE*. Required.
+
+**Outputs**:
+
+*   **1**: Output of type *T* and shape `[N, C, H_out]`, `[N, C, H_out, W_out]` or `[N, C, D_out, H_out, W_out]`.
+*   **2**: Output of type specified by *index_element_type* and same shape as the first output containing indices of elements in the first output. The values of indices are computed as if input was flatten 1-D tensor, so the values are in the range `[0, N * C * H * W * D)`.
+
+**Types**
+
+*   *T*: floating-point type.
+*   *T_SHAPE*: `int32` or `int64`.
+
+**Examples**
+
+```xml
+<layer ... type="AdaptiveMaxPool" ... >
+    <data output_type="i64"/>
+    <input> 
+        <port id="0">
+            <dim>1</dim>
+            <dim>3</dim>
+            <dim>32</dim>
+            <dim>32</dim>
+        </port>
+    </input>
+    <input> 
+        <port id="1">
+            <dim>2</dim>
+        </port>
+    </input>
+    <output>
+        <port id="1">
+            <dim>1</dim>
+            <dim>3</dim>
+            <dim>16</dim>
+            <dim>16</dim>
+        </port>
+        <port id="2">
+            <dim>1</dim>
+            <dim>3</dim>
+            <dim>16</dim>
+            <dim>16</dim>
+        </port>
+    </output>
+</layer>
+```

From 9e9e33e772602d76f2ad23ad0c1b07292f116990 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Tomasz=20Do=C5=82bniak?= <tomasz.dolbniak@intel.com>
Date: Thu, 10 Jun 2021 17:13:56 +0200
Subject: [PATCH 059/178] Axes input handling in ReduceSum evaluator (#5614)

---
 ngraph/core/src/op/reduce_sum.cpp             | 12 ++++-
 ngraph/core/src/op/util/evaluate_helpers.hpp  | 27 +++++++++++
 .../onnx/reduce_sum_13_axes_as_input.prototxt | 15 +++---
 ngraph/test/onnx/onnx_import.in.cpp           | 47 +++++++------------
 .../test/util/engine/interpreter_engine.cpp   | 10 ++--
 5 files changed, 67 insertions(+), 44 deletions(-)
 create mode 100644 ngraph/core/src/op/util/evaluate_helpers.hpp

diff --git a/ngraph/core/src/op/reduce_sum.cpp b/ngraph/core/src/op/reduce_sum.cpp
index 1eab2dcefb86b0..70c2b39ddc064b 100644
--- a/ngraph/core/src/op/reduce_sum.cpp
+++ b/ngraph/core/src/op/reduce_sum.cpp
@@ -6,9 +6,11 @@
 #include "itt.hpp"
 #include "ngraph/graph_util.hpp"
 #include "ngraph/op/broadcast.hpp"
+#include "ngraph/op/util/op_types.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
 #include "ngraph/runtime/reference/sum.hpp"
 #include "ngraph/shape_util.hpp"
+#include "util/evaluate_helpers.hpp"
 
 using namespace std;
 using namespace ngraph;
@@ -73,7 +75,15 @@ bool op::v1::ReduceSum::evaluate(const HostTensorVector& outputs,
                                  const HostTensorVector& inputs) const
 {
     NGRAPH_OP_SCOPE(v1_ReduceSum_evaluate);
-    return reduce_sum::evaluate_sum(inputs[0], outputs[0], get_reduction_axes(), get_keep_dims());
+
+    NGRAPH_CHECK(inputs.size() == 2,
+                 "The ReduceSum operation expects 2 input tensors. Got: ",
+                 inputs.size());
+
+    const auto reduction_axes = get_normalized_axes_from_tensor(
+        inputs[1], get_input_partial_shape(0).rank(), get_friendly_name());
+
+    return reduce_sum::evaluate_sum(inputs[0], outputs[0], reduction_axes, get_keep_dims());
 }
 
 bool op::v1::ReduceSum::has_evaluate() const
diff --git a/ngraph/core/src/op/util/evaluate_helpers.hpp b/ngraph/core/src/op/util/evaluate_helpers.hpp
new file mode 100644
index 00000000000000..d67a212f1574b6
--- /dev/null
+++ b/ngraph/core/src/op/util/evaluate_helpers.hpp
@@ -0,0 +1,27 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "ngraph/axis_set.hpp"
+#include "ngraph/descriptor/tensor.hpp"
+#include "ngraph/util.hpp"
+#include "ngraph/validation_util.hpp"
+
+namespace ngraph
+{
+    /// \brief Extracts the tensor data and returns a set of normalized axes created out of it.
+    ///
+    /// \param tensor A pointer to a HostTensor object containing the raw axes data
+    /// \param rank Rank of an operator's input data tensor (used to normalize the axes)
+    /// \param node_description An identifier of the operator's node (used to report errors)
+    ///
+    /// \return Normalized (positive only) axes as an AxisSet object.
+    AxisSet get_normalized_axes_from_tensor(const HostTensorPtr tensor,
+                                            const ngraph::Rank& rank,
+                                            const std::string& node_description)
+    {
+        const auto axes_vector = host_tensor_2_vector<int64_t>(tensor);
+        const auto normalized_axes = ngraph::normalize_axes(node_description, axes_vector, rank);
+        return AxisSet{normalized_axes};
+    }
+} // namespace ngraph
diff --git a/ngraph/test/models/onnx/reduce_sum_13_axes_as_input.prototxt b/ngraph/test/models/onnx/reduce_sum_13_axes_as_input.prototxt
index 6a458e7439356d..273e436803a42a 100644
--- a/ngraph/test/models/onnx/reduce_sum_13_axes_as_input.prototxt
+++ b/ngraph/test/models/onnx/reduce_sum_13_axes_as_input.prototxt
@@ -15,16 +15,10 @@ graph {
         elem_type: 1
         shape {
           dim {
-            dim_value: 1
+            dim_value: 2
           }
           dim {
-            dim_value: 1
-          }
-          dim {
-            dim_value: 4
-          }
-          dim {
-            dim_value: 4
+            dim_value: 2
           }
         }
       }
@@ -37,7 +31,7 @@ graph {
         elem_type: 7
         shape {
           dim {
-            dim_value: 4
+            dim_value: 1
           }
         }
       }
@@ -50,6 +44,9 @@ graph {
       tensor_type {
         elem_type: 1
         shape {
+          dim {
+            dim_value: 2
+          }
           dim {
             dim_value: 1
           }
diff --git a/ngraph/test/onnx/onnx_import.in.cpp b/ngraph/test/onnx/onnx_import.in.cpp
index a8522c08fa1729..849a015c4c3807 100644
--- a/ngraph/test/onnx/onnx_import.in.cpp
+++ b/ngraph/test/onnx/onnx_import.in.cpp
@@ -1238,25 +1238,10 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_model_reduce_sum_13_axes_as_input)
         file_util::path_join(SERIALIZED_ZOO, "onnx/reduce_sum_13_axes_as_input.prototxt"));
 
     auto test_case = test::TestCase<TestEngine, test::TestCaseType::DYNAMIC>(function);
-    test_case.add_input<float>({1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f,
-                                1.0f});
-    test_case.add_input<int64_t>({0, 1, 2, 3});
+    test_case.add_input<float>({1.0f, 2.0f, 3.0f, 4.0f});
+    test_case.add_input<int64_t>({1});
 
-    test_case.add_expected_output<float>(Shape{1, 1, 1, 1}, {16.0f});
+    test_case.add_expected_output<float>(Shape{2, 1}, {3.0f, 7.0f});
     test_case.run();
 }
 
@@ -3123,7 +3108,7 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_model_mod_sign)
 
     test_case.add_input<int32_t>({-4, 7, 5, 4, -7, 8});
     test_case.add_input<int32_t>({2, -3, 8, -2, 3, 5});
-    test_case.add_expected_output<int32_t>(Shape{6}, {0, -2,  5,  0,  2,  3});
+    test_case.add_expected_output<int32_t>(Shape{6}, {0, -2, 5, 0, 2, 3});
 
     test_case.run();
 }
@@ -3136,7 +3121,7 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_model_mod_sign_i64)
 
     test_case.add_input<int64_t>({-4, 7, 5, 4, -7, 8});
     test_case.add_input<int64_t>({2, -3, 8, -2, 3, 5});
-    test_case.add_expected_output<int64_t>(Shape{6}, {0, -2,  5,  0,  2,  3});
+    test_case.add_expected_output<int64_t>(Shape{6}, {0, -2, 5, 0, 2, 3});
 
     test_case.run();
 }
@@ -3159,13 +3144,15 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_model_mod_sign_f32)
     try
     {
         const auto function = onnx_import::import_onnx_model(
-        file_util::path_join(SERIALIZED_ZOO, "onnx/mod_sign_f32.prototxt"));
+            file_util::path_join(SERIALIZED_ZOO, "onnx/mod_sign_f32.prototxt"));
         FAIL() << "Expected exception was not thrown";
     }
     catch (const ngraph::ngraph_error& e)
     {
-        EXPECT_HAS_SUBSTRING(e.what(),
-                             std::string("If the input type is floating point, then `fmod` attribute must be set to 1."));
+        EXPECT_HAS_SUBSTRING(
+            e.what(),
+            std::string(
+                "If the input type is floating point, then `fmod` attribute must be set to 1."));
     }
     catch (...)
     {
@@ -3207,7 +3194,8 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_model_mod_sign_fmod_f32)
 
     test_case.add_input<float>({-4.3, 7.2, 5.0, 4.3, -7.2, 8.0});
     test_case.add_input<float>({2.1, -3.4, 8.0, -2.1, 3.4, 5.0});
-    test_case.add_expected_output<float>(Shape{6}, {-0.10000038, 0.39999962, 5. , 0.10000038, -0.39999962, 3.});
+    test_case.add_expected_output<float>(
+        Shape{6}, {-0.10000038, 0.39999962, 5., 0.10000038, -0.39999962, 3.});
 
     test_case.run();
 }
@@ -3217,13 +3205,13 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_model_mod_incorrect_fmod)
     try
     {
         const auto function = onnx_import::import_onnx_model(
-        file_util::path_join(SERIALIZED_ZOO, "onnx/mod_incorrect_fmod.prototxt"));
+            file_util::path_join(SERIALIZED_ZOO, "onnx/mod_incorrect_fmod.prototxt"));
         FAIL() << "Expected exception was not thrown";
     }
     catch (const ngraph::ngraph_error& e)
     {
-        EXPECT_HAS_SUBSTRING(e.what(),
-                             std::string("Unsupported value of 'fmod' attribute (should be: 0 or 1)"));
+        EXPECT_HAS_SUBSTRING(
+            e.what(), std::string("Unsupported value of 'fmod' attribute (should be: 0 or 1)"));
     }
     catch (...)
     {
@@ -4364,8 +4352,8 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_negativelog_likelihood_loss)
 
 NGRAPH_TEST(${BACKEND_NAME}, onnx_constant_fill_input_as_shape_default_value)
 {
-    auto function = onnx_import::import_onnx_model(
-        file_util::path_join(SERIALIZED_ZOO, "onnx/constant_fill_input_as_shape_default_value.prototxt"));
+    auto function = onnx_import::import_onnx_model(file_util::path_join(
+        SERIALIZED_ZOO, "onnx/constant_fill_input_as_shape_default_value.prototxt"));
 
     auto test_case = test::TestCase<TestEngine>(function);
     test_case.add_expected_output<float>(Shape{1, 2, 3}, {0.f, 0.f, 0.f, 0.f, 0.f, 0.f});
@@ -4422,7 +4410,6 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_constant_bfloat_tensor)
     test_case.run();
 }
 
-
 NGRAPH_TEST(${BACKEND_NAME}, onnx_constant_float_scalar)
 {
     auto function = onnx_import::import_onnx_model(
diff --git a/ngraph/test/util/engine/interpreter_engine.cpp b/ngraph/test/util/engine/interpreter_engine.cpp
index 3254f395f08242..876bd63d146d02 100644
--- a/ngraph/test/util/engine/interpreter_engine.cpp
+++ b/ngraph/test/util/engine/interpreter_engine.cpp
@@ -133,8 +133,9 @@ testing::AssertionResult
         if (expected_shape != result_shape)
         {
             comparison_result = testing::AssertionFailure();
-            comparison_result << "Computed data shape does not match the expected shape for output "
-                              << i << std::endl;
+            comparison_result << "Computed data shape(" << result_shape
+                              << ") does not match the expected shape(" << expected_shape
+                              << ") for output " << i << std::endl;
             break;
         }
 
@@ -175,8 +176,9 @@ testing::AssertionResult test::INTERPRETER_Engine::compare_results(const size_t
         if (expected_shape != result_shape)
         {
             comparison_result = testing::AssertionFailure();
-            comparison_result << "Computed data shape does not match the expected shape for output "
-                              << i << std::endl;
+            comparison_result << "Computed data shape(" << result_shape
+                              << ") does not match the expected shape(" << expected_shape
+                              << ") for output " << i << std::endl;
             break;
         }
 

From 084aa4e5916fa2ed3e353dcd45d081ab11d9c75a Mon Sep 17 00:00:00 2001
From: Bartek Szmelczynski <bartosz.szmelczynski@intel.com>
Date: Thu, 10 Jun 2021 18:35:52 +0200
Subject: [PATCH 060/178] Revise log (#6061)

* refactor ngraph log class, add type_prop tests

* udpate inputs, outputs decscription in spec

Co-authored-by: jdanieck <jozef.daniecki@intel.com>
---
 docs/ops/arithmetic/Log_1.md          | 16 +++++++--------
 ngraph/core/include/ngraph/op/log.hpp |  3 +--
 ngraph/core/src/op/log.cpp            |  2 +-
 ngraph/test/CMakeLists.txt            |  1 +
 ngraph/test/type_prop/unary_ops.cpp   |  2 +-
 ngraph/test/visitors/op/log.cpp       | 29 +++++++++++++++++++++++++++
 6 files changed, 41 insertions(+), 12 deletions(-)
 create mode 100644 ngraph/test/visitors/op/log.cpp

diff --git a/docs/ops/arithmetic/Log_1.md b/docs/ops/arithmetic/Log_1.md
index 6f33b002b693b7..f13149198210a4 100644
--- a/docs/ops/arithmetic/Log_1.md
+++ b/docs/ops/arithmetic/Log_1.md
@@ -6,28 +6,28 @@
 
 **Short description**: *Log* performs element-wise natural logarithm operation with given tensor.
 
+**Detailed description**: *Log* does the following with the input tensor *a*:
+
+\f[
+a_{i} = log(a_{i})
+\f]
+
 **Attributes**:
 
     No attributes available.
 
 **Inputs**
 
-* **1**: An tensor of type T. **Required.**
+* **1**: An tensor of type T and arbitrary shape. **Required.**
 
 **Outputs**
 
-* **1**: The result of element-wise log operation. A tensor of type T.
+* **1**: The result of element-wise log operation. A tensor of type T and the same shape as input.
 
 **Types**
 
 * *T*: any numeric type.
 
-*Log* does the following with the input tensor *a*:
-
-\f[
-a_{i} = log(a_{i})
-\f]
-
 **Examples**
 
 *Example 1*
diff --git a/ngraph/core/include/ngraph/op/log.hpp b/ngraph/core/include/ngraph/op/log.hpp
index 6c147d43314b03..997732b61161a3 100644
--- a/ngraph/core/include/ngraph/op/log.hpp
+++ b/ngraph/core/include/ngraph/op/log.hpp
@@ -16,8 +16,7 @@ namespace ngraph
             class NGRAPH_API Log : public util::UnaryElementwiseArithmetic
             {
             public:
-                static constexpr NodeTypeInfo type_info{"Log", 0};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
+                NGRAPH_RTTI_DECLARATION;
                 /// \brief Constructs a natural log operation.
                 Log() = default;
                 /// \brief Constructs a natural log operation.
diff --git a/ngraph/core/src/op/log.cpp b/ngraph/core/src/op/log.cpp
index 49a41c0d93108f..7743dbed6bdf2c 100644
--- a/ngraph/core/src/op/log.cpp
+++ b/ngraph/core/src/op/log.cpp
@@ -13,7 +13,7 @@
 using namespace std;
 using namespace ngraph;
 
-constexpr NodeTypeInfo op::Log::type_info;
+NGRAPH_RTTI_DEFINITION(op::v0::Log, "Log", 0);
 
 op::Log::Log(const Output<Node>& arg)
     : UnaryElementwiseArithmetic(arg)
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 7ee6b89580d0f6..e25a40de184b6e 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -232,6 +232,7 @@ set(SRC
     visitors/op/grn.cpp
     visitors/op/group_conv.cpp
     visitors/op/interpolate.cpp
+    visitors/op/log.cpp
     visitors/op/logical_xor.cpp
     visitors/op/lrn.cpp
     visitors/op/lstm_cell.cpp
diff --git a/ngraph/test/type_prop/unary_ops.cpp b/ngraph/test/type_prop/unary_ops.cpp
index 3859dd7369d490..788b50917fa6a2 100644
--- a/ngraph/test/type_prop/unary_ops.cpp
+++ b/ngraph/test/type_prop/unary_ops.cpp
@@ -96,6 +96,6 @@ REGISTER_TYPED_TEST_CASE_P(UnaryOperator,
                            dynamic_rank_input_shape_3D,
                            dynamic_rank_input_shape_full);
 
-using Types = ::testing::Types<op::Acos, op::Asin, op::Abs, op::Sqrt, op::Sin, op::Exp, op::Floor>;
+using Types = ::testing::Types<op::Acos, op::Asin, op::Abs, op::Sqrt, op::Sin, op::Exp, op::Floor, op::Log>;
 
 INSTANTIATE_TYPED_TEST_CASE_P(type_prop, UnaryOperator, Types);
diff --git a/ngraph/test/visitors/op/log.cpp b/ngraph/test/visitors/op/log.cpp
new file mode 100644
index 00000000000000..02d6293871c028
--- /dev/null
+++ b/ngraph/test/visitors/op/log.cpp
@@ -0,0 +1,29 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+
+#include "ngraph/ngraph.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+#include "ngraph/opsets/opset1.hpp"
+
+#include "util/visitor.hpp"
+
+using namespace std;
+using namespace ngraph;
+using ngraph::test::NodeBuilder;
+
+TEST(attributes, log_op)
+{
+    using namespace opset1;
+
+    NodeBuilder::get_ops().register_factory<Log>();
+    const auto data_input = make_shared<Parameter>(element::f32, Shape{1, 2, 3});
+    const auto op = make_shared<Sqrt>(data_input);
+
+    NodeBuilder builder(op);
+    const auto expected_attr_count = 0;
+
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
+}
\ No newline at end of file

From d2adea950ff99b7d13cb6b6c33f1ac5e045af9fc Mon Sep 17 00:00:00 2001
From: Irina Burgun <irinax.burgun@intel.com>
Date: Thu, 10 Jun 2021 22:23:01 +0300
Subject: [PATCH 061/178] Freeze summary header while scrolling (#6121)

---
 .../layer_tests_summary/template/report_template.html       | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/report_template.html b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/report_template.html
index 26f0923e144fb2..7d6f751f91701a 100644
--- a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/report_template.html
+++ b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/report_template.html
@@ -43,11 +43,11 @@ <h2>Operations coverage summary: {{report_tag}} {{ timestamp }}</h2>
     </thead>
 </table>
 <table class="table table-hover" id="report">
-    <thead style="position: sticky; top: 0">
+    <thead>
     <tr>
-        <th class="table-primary" scope="col">Operation</th>
+        <th class="table-primary" scope="col" style="position: sticky; top: 0">Operation</th>
         {% for d in devices -%}
-        <th class="table-primary">{{ d }}</th>
+        <th class="table-primary" style="position: sticky; top: 0">{{ d }}</th>
         {% endfor %}
     </tr>
     </thead>

From 14e486b767dac559c6ade05544bc48ac0958ba3f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Tomasz=20Do=C5=82bniak?= <tomasz.dolbniak@intel.com>
Date: Thu, 10 Jun 2021 22:34:36 +0200
Subject: [PATCH 062/178] MVN test with opset6 version of the op (#6111)

---
 ngraph/python/tests/__init__.py               |  1 -
 .../tests/test_ngraph/test_ops_fused.py       | 29 +++++++++----------
 ngraph/test/runtime/ie/unit_test.manifest     |  8 -----
 3 files changed, 14 insertions(+), 24 deletions(-)

diff --git a/ngraph/python/tests/__init__.py b/ngraph/python/tests/__init__.py
index f54930f40d92e5..817d4584534934 100644
--- a/ngraph/python/tests/__init__.py
+++ b/ngraph/python/tests/__init__.py
@@ -51,7 +51,6 @@ def xfail_test(reason="Mark the test as expected to fail", strict=True):
 xfail_issue_35927 = xfail_test(reason="RuntimeError: B has zero dimension that is not allowable")
 xfail_issue_36486 = xfail_test(reason="RuntimeError: HardSigmoid operation should be converted "
                                       "to HardSigmoid_IE")
-xfail_issue_36487 = xfail_test(reason="Assertion error - mvn operator computation mismatch")
 xfail_issue_38084 = xfail_test(reason="RuntimeError: AssertionFailed: layer->get_output_partial_shape(i)"
                                       "is_static() nGraph <value> operation with name: <value> cannot be"
                                       "converted to <value> layer with name: <value> because output"
diff --git a/ngraph/python/tests/test_ngraph/test_ops_fused.py b/ngraph/python/tests/test_ngraph/test_ops_fused.py
index 48fff1623c5f61..d96f870f604bc0 100644
--- a/ngraph/python/tests/test_ngraph/test_ops_fused.py
+++ b/ngraph/python/tests/test_ngraph/test_ops_fused.py
@@ -6,9 +6,7 @@
 
 import ngraph as ng
 from tests.runtime import get_runtime
-from tests import (xfail_issue_36486,
-                   xfail_issue_36487,
-                   xfail_issue_44976)
+from tests import (xfail_issue_36486, xfail_issue_44976)
 
 
 def test_elu_operator_with_scalar_and_array():
@@ -424,14 +422,14 @@ def test_hard_sigmoid_operator():
     assert np.allclose(result, expected)
 
 
-@xfail_issue_36487
 def test_mvn_operator():
     runtime = get_runtime()
 
     data_shape = [3, 3, 3, 1]
-    across_channels = True
+    axes = [0, 2, 3]
     normalize_variance = True
     eps = np.float32(1e-9)
+    eps_mode = "outside_sqrt"
 
     data_value = np.array(
         [
@@ -456,7 +454,7 @@ def test_mvn_operator():
 
     parameter_data = ng.parameter(data_shape, name="Data", dtype=np.float32)
 
-    model = ng.mvn(parameter_data, across_channels, normalize_variance, eps)
+    model = ng.mvn(parameter_data, axes, normalize_variance, eps, eps_mode)
     computation = runtime.computation(model, parameter_data)
 
     result = computation(data_value)
@@ -464,21 +462,22 @@ def test_mvn_operator():
     expected = np.array(
         [
             [
-                [[0.9951074], [0.14548765], [-1.410561]],
-                [[-1.4999886], [-1.1923014], [-0.03975919]],
-                [[0.8463296], [1.2926502], [1.3340596]],
+                [[1.3546423], [0.33053496], [-1.5450814]],
+                [[-1.2106764], [-0.8925952], [0.29888135]],
+                [[0.38083088], [0.81808794], [0.85865635]],
             ],
             [
-                [[-1.0463363], [-0.1747985], [-0.7784088]],
-                [[0.47672555], [-1.5383], [0.32375798]],
-                [[1.2404392], [1.3878832], [-1.2228798]],
+                [[-1.1060555], [-0.05552877], [-0.78310335]],
+                [[0.83281356], [-1.250282], [0.67467856]],
+                [[0.7669372], [0.9113869], [-1.6463585]],
             ],
             [
-                [[-0.3228847], [1.2063044], [0.22751297]],
-                [[0.91956615], [0.81839436], [-1.2279599]],
-                [[0.5312334], [0.067952], [-1.3592235]],
+                [[-0.23402764], [1.6092131], [0.42940593]],
+                [[1.2906139], [1.1860244], [-0.92945826]],
+                [[0.0721334], [-0.38174], [-1.7799333]],
             ],
         ],
+        dtype=np.float32,
     )
 
     assert np.allclose(result, expected)
diff --git a/ngraph/test/runtime/ie/unit_test.manifest b/ngraph/test/runtime/ie/unit_test.manifest
index c476f2dfac3137..f95ce0d951e162 100644
--- a/ngraph/test/runtime/ie/unit_test.manifest
+++ b/ngraph/test/runtime/ie/unit_test.manifest
@@ -1667,14 +1667,6 @@ IE_CPU.evaluate_ctc_greedy_decoder_seq_len_f16
 minimum_u16
 IE_CPU/ElemTypesTests/1.onnx_test_add_abc_set_precission
 
-# not implemented in CPU and GPU
-evaluate_mvn_6_no_variance
-evaluate_mvn_6
-evaluate_mvn_6_inside_sqrt
-evaluate_mvn_6_across_chanells
-evaluate_mvn_6_across_batch
-IE_CPU.onnx_mvn_v6
-
 # not yet implemented on CPU/GPU Gather 7
 gather_v7_data_int32_2d_indices_axis_1_negative_batch_dims
 gather_v7_1d_int32

From 05469e95e99576f9fda84ae6f3ba8dc9d506d9a6 Mon Sep 17 00:00:00 2001
From: Mikhail Nosov <mikhail.nosov@intel.com>
Date: Fri, 11 Jun 2021 07:14:36 +0300
Subject: [PATCH 063/178] [NG Frontend] Frontend exceptions (#5993)

* [NG Frontend] Introduce classes and macros for Frontend-specific exceptions

* Add FrontEnd exception unit tests

* resolve review comments

* Import FE exception classes

* code style

Co-authored-by: Ivan Tikhonov <ivan.tikhonov@intel.com>
---
 .../frontend_manager/frontend_exceptions.hpp  | 123 +++++++++++++++
 .../frontend_manager/src/frontend_manager.cpp |  12 +-
 ngraph/test/frontend/frontend_manager.cpp     | 148 ++++++++++++++++++
 3 files changed, 274 insertions(+), 9 deletions(-)
 create mode 100644 ngraph/frontend/frontend_manager/include/frontend_manager/frontend_exceptions.hpp

diff --git a/ngraph/frontend/frontend_manager/include/frontend_manager/frontend_exceptions.hpp b/ngraph/frontend/frontend_manager/include/frontend_manager/frontend_exceptions.hpp
new file mode 100644
index 00000000000000..46fa2be0888f70
--- /dev/null
+++ b/ngraph/frontend/frontend_manager/include/frontend_manager/frontend_exceptions.hpp
@@ -0,0 +1,123 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <memory>
+#include <string>
+#include "frontend_manager_defs.hpp"
+#include "ngraph/check.hpp"
+
+namespace ngraph
+{
+    namespace frontend
+    {
+        class FRONTEND_API GeneralFailure : public CheckFailure
+        {
+        public:
+            GeneralFailure(const CheckLocInfo& check_loc_info,
+                           const std::string& context,
+                           const std::string& explanation)
+                : CheckFailure(check_loc_info,
+                               "FrontEnd API failed with GeneralFailure: " + context,
+                               explanation)
+            {
+            }
+        };
+
+        class FRONTEND_API InitializationFailure : public CheckFailure
+        {
+        public:
+            InitializationFailure(const CheckLocInfo& check_loc_info,
+                                  const std::string& context,
+                                  const std::string& explanation)
+                : CheckFailure(check_loc_info,
+                               "FrontEnd API failed with InitializationFailure: " + context,
+                               explanation)
+            {
+            }
+        };
+
+        class FRONTEND_API OpValidationFailure : public CheckFailure
+        {
+        public:
+            OpValidationFailure(const CheckLocInfo& check_loc_info,
+                                const std::string& context,
+                                const std::string& explanation)
+                : CheckFailure(check_loc_info,
+                               "FrontEnd API failed with OpValidationFailure: " + context,
+                               explanation)
+            {
+            }
+        };
+
+        class FRONTEND_API OpConversionFailure : public CheckFailure
+        {
+        public:
+            OpConversionFailure(const CheckLocInfo& check_loc_info,
+                                const std::string& context,
+                                const std::string& explanation)
+                : CheckFailure(check_loc_info,
+                               "FrontEnd API failed with OpConversionFailure: " + context,
+                               explanation)
+            {
+            }
+        };
+
+        class FRONTEND_API NotImplementedFailure : public CheckFailure
+        {
+        public:
+            NotImplementedFailure(const CheckLocInfo& check_loc_info,
+                                  const std::string& context,
+                                  const std::string& explanation)
+                : CheckFailure(check_loc_info,
+                               "FrontEnd API failed with NotImplementedFailure: " + context,
+                               explanation)
+            {
+            }
+        };
+
+/// \brief Macro to check whether a boolean condition holds.
+/// \param cond Condition to check
+/// \param ... Additional error message info to be added to the error message via the `<<`
+///            stream-insertion operator. Note that the expressions here will be evaluated lazily,
+///            i.e., only if the `cond` evalutes to `false`.
+/// \throws ::ngraph::frontend::GeneralFailure if `cond` is false.
+#define FRONT_END_GENERAL_CHECK(...)                                                               \
+    NGRAPH_CHECK_HELPER(::ngraph::frontend::GeneralFailure, "", __VA_ARGS__)
+
+/// \brief Macro to check whether a boolean condition holds.
+/// \param cond Condition to check
+/// \param ... Additional error message info to be added to the error message via the `<<`
+///            stream-insertion operator. Note that the expressions here will be evaluated lazily,
+///            i.e., only if the `cond` evalutes to `false`.
+/// \throws ::ngraph::frontend::InitializationFailure if `cond` is false.
+#define FRONT_END_INITIALIZATION_CHECK(...)                                                        \
+    NGRAPH_CHECK_HELPER(::ngraph::frontend::InitializationFailure, "", __VA_ARGS__)
+
+/// \brief Macro to check whether a boolean condition holds.
+/// \param cond Condition to check
+/// \param ... Additional error message info to be added to the error message via the `<<`
+///            stream-insertion operator. Note that the expressions here will be evaluated lazily,
+///            i.e., only if the `cond` evalutes to `false`.
+/// \throws ::ngraph::frontend::OpConversionFailure if `cond` is false.
+#define FRONT_END_OP_CONVERSION_CHECK(...)                                                         \
+    NGRAPH_CHECK_HELPER(::ngraph::frontend::OpConversionFailure, "", __VA_ARGS__)
+
+/// \brief Assert macro.
+/// \param NAME Name of the function that is not implemented
+/// \throws ::ngraph::frontend::NotImplementedFailure
+#define FRONT_END_NOT_IMPLEMENTED(NAME)                                                            \
+    NGRAPH_CHECK_HELPER(::ngraph::frontend::NotImplementedFailure,                                 \
+                        "",                                                                        \
+                        false,                                                                     \
+                        #NAME " is not implemented for this FrontEnd class")
+
+/// \brief Assert macro.
+/// \param MSG Error message
+/// \throws ::ngraph::frontend::GeneralFailure
+#define FRONT_END_THROW(MSG) FRONT_END_GENERAL_CHECK(false, MSG)
+
+    } // namespace frontend
+} // namespace ngraph
\ No newline at end of file
diff --git a/ngraph/frontend/frontend_manager/src/frontend_manager.cpp b/ngraph/frontend/frontend_manager/src/frontend_manager.cpp
index c1a864ffd6e27d..037a2522523dc4 100644
--- a/ngraph/frontend/frontend_manager/src/frontend_manager.cpp
+++ b/ngraph/frontend/frontend_manager/src/frontend_manager.cpp
@@ -5,20 +5,13 @@
 #include <ngraph/env_util.hpp>
 #include <ngraph/except.hpp>
 
+#include "frontend_manager/frontend_exceptions.hpp"
 #include "frontend_manager/frontend_manager.hpp"
 #include "plugin_loader.hpp"
 
 using namespace ngraph;
 using namespace ngraph::frontend;
 
-#define FRONT_END_NOT_IMPLEMENTED(NAME)                                                            \
-    throw std::runtime_error(#NAME " is not implemented for this FrontEnd class")
-#define FRONT_END_ASSERT(EXPRESSION)                                                               \
-    {                                                                                              \
-        if (!(EXPRESSION))                                                                         \
-            throw "AssertionFailed";                                                               \
-    }
-
 //----------- FrontEndManager ---------------------------
 class FrontEndManager::Impl
 {
@@ -32,7 +25,8 @@ class FrontEndManager::Impl
 
     FrontEnd::Ptr loadByFramework(const std::string& framework, FrontEndCapFlags fec)
     {
-        FRONT_END_ASSERT(m_factories.count(framework))
+        FRONT_END_INITIALIZATION_CHECK(
+            m_factories.count(framework), "FrontEnd for Framework ", framework, " is not found");
         return m_factories[framework](fec);
     }
 
diff --git a/ngraph/test/frontend/frontend_manager.cpp b/ngraph/test/frontend/frontend_manager.cpp
index 10e8eac8e60001..af70885d237901 100644
--- a/ngraph/test/frontend/frontend_manager.cpp
+++ b/ngraph/test/frontend/frontend_manager.cpp
@@ -2,6 +2,7 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+#include <frontend_manager/frontend_exceptions.hpp>
 #include <frontend_manager/frontend_manager.hpp>
 #include <memory>
 
@@ -161,3 +162,150 @@ TEST(FrontEndManagerTest, testDefaultPlace)
     ASSERT_ANY_THROW(place->is_equal(nullptr));
     ASSERT_ANY_THROW(place->is_equal_data(nullptr));
 }
+
+TEST(FrontEndExceptionTest, frontend_general_error_no_throw)
+{
+    EXPECT_NO_THROW(FRONT_END_GENERAL_CHECK(true));
+}
+
+TEST(FrontEndExceptionTest, frontend_general_error_no_throw_info)
+{
+    EXPECT_NO_THROW(FRONT_END_GENERAL_CHECK(true, "msg example"));
+}
+
+TEST(FrontEndExceptionTest, frontend_general_error_throw_no_info)
+{
+    EXPECT_THROW(FRONT_END_GENERAL_CHECK(false), ngraph::frontend::GeneralFailure);
+}
+
+TEST(FrontEndExceptionTest, frontend_initialization_error_no_throw)
+{
+    EXPECT_NO_THROW(FRONT_END_INITIALIZATION_CHECK(true));
+}
+
+TEST(FrontEndExceptionTest, frontend_initialization_error_no_throw_info)
+{
+    EXPECT_NO_THROW(FRONT_END_INITIALIZATION_CHECK(true, "msg example"));
+}
+
+TEST(FrontEndExceptionTest, frontend_initialization_error_throw_no_info)
+{
+    EXPECT_THROW(FRONT_END_INITIALIZATION_CHECK(false), ngraph::frontend::InitializationFailure);
+}
+
+TEST(FrontEndExceptionTest, frontend_op_conversion_error_no_throw)
+{
+    EXPECT_NO_THROW(FRONT_END_OP_CONVERSION_CHECK(true));
+}
+
+TEST(FrontEndExceptionTest, frontend_op_conversion_error_no_throw_info)
+{
+    EXPECT_NO_THROW(FRONT_END_OP_CONVERSION_CHECK(true, "msg example"));
+}
+
+TEST(FrontEndExceptionTest, frontend_op_conversion_error_throw_no_info)
+{
+    EXPECT_THROW(FRONT_END_OP_CONVERSION_CHECK(false), ngraph::frontend::OpConversionFailure);
+}
+
+TEST(FrontEndExceptionTest, frontend_assert_throw_check_info)
+{
+    std::string msg("msg example");
+    try
+    {
+        FRONT_END_THROW(msg);
+    }
+    catch (const ngraph::frontend::GeneralFailure& ex)
+    {
+        std::string caught_msg(ex.what());
+        EXPECT_NE(caught_msg.find(msg), std::string::npos);
+        return;
+    }
+    catch (...)
+    {
+        FAIL() << "Not expected exception type.";
+    }
+    FAIL() << "Test is expected to throw an exception.";
+}
+
+TEST(FrontEndExceptionTest, frontend_not_implemented_throw_check_info)
+{
+    struct TestClass
+    {
+    };
+    try
+    {
+        FRONT_END_NOT_IMPLEMENTED(TestClass);
+    }
+    catch (const ngraph::frontend::NotImplementedFailure& ex)
+    {
+        std::string caught_msg(ex.what());
+        EXPECT_NE(caught_msg.find("TestClass"), std::string::npos);
+        return;
+    }
+    catch (...)
+    {
+        FAIL() << "Not expected exception type.";
+    }
+    FAIL() << "Test is expected to throw an exception.";
+}
+
+TEST(FrontEndExceptionTest, frontend_general_error_throw_info)
+{
+    std::string msg("msg example");
+    try
+    {
+        FRONT_END_GENERAL_CHECK(false, msg);
+    }
+    catch (const ngraph::frontend::GeneralFailure& ex)
+    {
+        std::string caught_msg(ex.what());
+        EXPECT_NE(caught_msg.find(msg), std::string::npos);
+        return;
+    }
+    catch (...)
+    {
+        FAIL() << "Not expected exception type.";
+    }
+    FAIL() << "Test is expected to throw an exception.";
+}
+
+TEST(FrontEndExceptionTest, frontend_op_conversion_error_throw_info)
+{
+    std::string msg("msg example");
+    try
+    {
+        FRONT_END_OP_CONVERSION_CHECK(false, msg);
+    }
+    catch (const ngraph::frontend::OpConversionFailure& ex)
+    {
+        std::string caught_msg(ex.what());
+        EXPECT_NE(caught_msg.find(msg), std::string::npos);
+        return;
+    }
+    catch (...)
+    {
+        FAIL() << "Not expected exception type.";
+    }
+    FAIL() << "Test is expected to throw an exception.";
+}
+
+TEST(FrontEndExceptionTest, frontend_initialization_error_throw_info)
+{
+    std::string msg("msg example");
+    try
+    {
+        FRONT_END_INITIALIZATION_CHECK(false, msg);
+    }
+    catch (const ngraph::frontend::InitializationFailure& ex)
+    {
+        std::string caught_msg(ex.what());
+        EXPECT_NE(caught_msg.find(msg), std::string::npos);
+        return;
+    }
+    catch (...)
+    {
+        FAIL() << "Not expected exception type.";
+    }
+    FAIL() << "Test is expected to throw an exception.";
+}

From ddc27d3513dc021bd833d600ca1323aa24b3cb04 Mon Sep 17 00:00:00 2001
From: Gabriele Galiero Casay <gabriele.galiero.casay@intel.com>
Date: Fri, 11 Jun 2021 06:41:44 +0200
Subject: [PATCH 064/178] Revise reference implementation for ReduceSum
 operation (#5728)

* Revise reference implementation for ReduceSum operation

* Refactor backend unit tests

 * Move tests with zero dims and dynamic shapes to op_eval
 * Removed tests for f64 precision not supported

* Add support for precisions bf16, i8 and u8

* Clean up interpreter manifest unit tests

* Create ticket for result mismatch and accuracy failing tests

* Disable back interpreter tests failing for ubuntu18 i386 for large tensors

* Revert "Add support for precisions bf16, i8 and u8"

This reverts commit 8f1bba0dff37e8131adc59351c65e65ba1103b9a.

* Add to manifest tests with disable types of reference implementation

* Address review comments

 * Use constexpr variable to set keep dims as false in reduction output shape
 * Replace CoordinateTransform with CoordinateTransformBasic

* Fix typo in variable name

* Remove keep_dims parameter from sum function call
---
 .../ngraph/runtime/reference/log_softmax.hpp  |   2 +-
 .../ngraph/runtime/reference/normalize_l2.hpp |   2 +-
 .../ngraph/runtime/reference/softmax.hpp      |   2 +-
 .../include/ngraph/runtime/reference/sum.hpp  |  37 +-
 .../src/runtime/reference/einsum.cpp          |   6 +-
 ngraph/core/src/op/reduce_sum.cpp             |   9 +-
 ngraph/test/CMakeLists.txt                    |   1 +
 ngraph/test/backend/reduce_sum.in.cpp         | 581 ------------------
 ngraph/test/op_eval/reduce_sum.cpp            | 395 ++++++++++++
 ngraph/test/runtime/ie/unit_test.manifest     |  28 +-
 .../runtime/interpreter/unit_test.manifest    |   4 +-
 11 files changed, 429 insertions(+), 638 deletions(-)
 create mode 100644 ngraph/test/op_eval/reduce_sum.cpp

diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/log_softmax.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/log_softmax.hpp
index 11b9092e6e6489..27198f465b30ff 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/log_softmax.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/log_softmax.hpp
@@ -35,7 +35,7 @@ namespace ngraph
                         arg[transform.index(coord)] - temp_max[temp_transform.index(temp_coord)]);
                 }
 
-                sum(out, temp_sum.data(), shape, axes, true);
+                sum(out, temp_sum.data(), shape, axes);
 
                 for (const Coordinate& coord : transform)
                 {
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/normalize_l2.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/normalize_l2.hpp
index 5da76763465b9d..c32ec70488ede3 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/normalize_l2.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/normalize_l2.hpp
@@ -42,7 +42,7 @@ namespace ngraph
                 }
 
                 std::vector<T> sum_data(shape_size(reduce_shape));
-                sum(sqr_data.data(), sum_data.data(), data_shape, axes, true);
+                sum(sqr_data.data(), sum_data.data(), data_shape, axes);
                 autobroadcast_binop(data,
                                     sum_data.data(),
                                     out,
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/softmax.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/softmax.hpp
index c31c901012ce93..94cb0549b7310d 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/softmax.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/softmax.hpp
@@ -34,7 +34,7 @@ namespace ngraph
                         arg[transform.index(coord)] - temp_ptr[temp_transform.index(temp_coord)]);
                 }
 
-                sum(out, temp_ptr, shape, axes, true);
+                sum(out, temp_ptr, shape, axes);
 
                 for (const Coordinate& coord : transform)
                 {
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/sum.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/sum.hpp
index 0c20eaf431ab9f..2408e6f476d728 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/sum.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/sum.hpp
@@ -5,6 +5,7 @@
 #pragma once
 
 #include <cmath>
+#include <numeric>
 
 #include "ngraph/coordinate_transform.hpp"
 #include "ngraph/shape_util.hpp"
@@ -41,34 +42,34 @@ namespace ngraph
             }
 
             template <typename T>
-            void sum(const T* arg,
-                     T* out,
-                     const Shape& in_shape,
-                     const AxisSet& reduction_axes,
-                     bool keep_dims)
+            void sum(const T* arg, T* out, const Shape& in_shape, const AxisSet& reduction_axes)
             {
-                auto out_shape = reduce(in_shape, reduction_axes, keep_dims);
-                CoordinateTransform output_transform(out_shape);
-                std::vector<T> cs(shape_size(out_shape));
+                constexpr bool dont_keep_dims_in_output = false;
+                auto out_shape = reduce(in_shape, reduction_axes, dont_keep_dims_in_output);
 
-                for (const Coordinate& output_coord : output_transform)
-                {
-                    out[output_transform.index(output_coord)] = 0;
-                    cs[output_transform.index(output_coord)] = 0;
-                }
+                std::vector<T> cs(shape_size(out_shape), 0);
+                std::fill(out, out + shape_size(out_shape), 0);
 
-                CoordinateTransform input_transform(in_shape);
+                const auto in_strides = row_major_strides(in_shape);
+                const auto out_strides = row_major_strides(out_shape);
 
+                CoordinateTransformBasic input_transform(in_shape);
                 for (const Coordinate& input_coord : input_transform)
                 {
-                    Coordinate output_coord = reduce(input_coord, reduction_axes, keep_dims);
+                    Coordinate output_coord =
+                        reduce(input_coord, reduction_axes, dont_keep_dims_in_output);
+
+                    size_t in_idx = std::inner_product(
+                        input_coord.begin(), input_coord.end(), in_strides.begin(), 0);
+                    size_t out_idx = std::inner_product(
+                        output_coord.begin(), output_coord.end(), out_strides.begin(), 0);
 
-                    T x = arg[input_transform.index(input_coord)];
-                    T& z = out[output_transform.index(output_coord)];
+                    T x = arg[in_idx];
+                    T& z = out[out_idx];
 
                     if (is_finite(x) && is_finite(z))
                     {
-                        T& c = cs[output_transform.index(output_coord)];
+                        T& c = cs[out_idx];
                         T t = z + (x - c);
                         c = (t - z) - (x - c);
                         z = t;
diff --git a/ngraph/core/reference/src/runtime/reference/einsum.cpp b/ngraph/core/reference/src/runtime/reference/einsum.cpp
index e1356788c5ea8d..51352ddadac9db 100644
--- a/ngraph/core/reference/src/runtime/reference/einsum.cpp
+++ b/ngraph/core/reference/src/runtime/reference/einsum.cpp
@@ -436,8 +436,7 @@ namespace ngraph
                     ngraph::runtime::reference::sum<T>(input_ptr->get_data_ptr<T>(),
                                                        output_ptr->get_data_ptr<T>(),
                                                        input_shape,
-                                                       reduced_axes,
-                                                       false);
+                                                       reduced_axes);
 
                     // update a vector of inputs and input subscripts
                     inputs[input_ind] = output_ptr;
@@ -743,8 +742,7 @@ namespace ngraph
                     ngraph::runtime::reference::sum<T>(mul_output->get_data_ptr<T>(),
                                                        result->get_data_ptr<T>(),
                                                        mul_output->get_shape(),
-                                                       reduced_axes,
-                                                       false);
+                                                       reduced_axes);
                     inputs[input_ind] = result;
                     input_subscripts[input_ind] = resultant_subscript;
                 }
diff --git a/ngraph/core/src/op/reduce_sum.cpp b/ngraph/core/src/op/reduce_sum.cpp
index 70c2b39ddc064b..34262ec9efb32d 100644
--- a/ngraph/core/src/op/reduce_sum.cpp
+++ b/ngraph/core/src/op/reduce_sum.cpp
@@ -3,6 +3,7 @@
 //
 
 #include "ngraph/op/reduce_sum.hpp"
+#include <ngraph/validation_util.hpp>
 #include "itt.hpp"
 #include "ngraph/graph_util.hpp"
 #include "ngraph/op/broadcast.hpp"
@@ -47,7 +48,7 @@ namespace reduce_sum
     {
         out->set_shape(reduce(arg->get_shape(), axes, keep_dims));
         runtime::reference::sum(
-            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes, keep_dims);
+            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes);
         return true;
     }
 
@@ -75,10 +76,8 @@ bool op::v1::ReduceSum::evaluate(const HostTensorVector& outputs,
                                  const HostTensorVector& inputs) const
 {
     NGRAPH_OP_SCOPE(v1_ReduceSum_evaluate);
-
-    NGRAPH_CHECK(inputs.size() == 2,
-                 "The ReduceSum operation expects 2 input tensors. Got: ",
-                 inputs.size());
+    NGRAPH_CHECK(validate_host_tensor_vector(inputs, 2));
+    NGRAPH_CHECK(validate_host_tensor_vector(outputs, 1));
 
     const auto reduction_axes = get_normalized_axes_from_tensor(
         inputs[1], get_input_partial_shape(0).rank(), get_friendly_name());
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index e25a40de184b6e..c629111ca147c1 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -73,6 +73,7 @@ set(SRC
     op_eval/non_zero.cpp
     op_eval/reduce_l1.cpp
     op_eval/reduce_l2.cpp
+    op_eval/reduce_sum.cpp
     op_eval/roi_align.cpp
     op_eval/roi_pooling.cpp
     op_eval/round.cpp
diff --git a/ngraph/test/backend/reduce_sum.in.cpp b/ngraph/test/backend/reduce_sum.in.cpp
index 50560a07c6ee97..97bc81d98eb370 100644
--- a/ngraph/test/backend/reduce_sum.in.cpp
+++ b/ngraph/test/backend/reduce_sum.in.cpp
@@ -151,95 +151,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_matrix_rows)
     EXPECT_TRUE(test::all_close_f((vector<float>{3, 7, 11}), read_vector<float>(result)));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_matrix_rows_zero)
-{
-    Shape shape_a{3, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0, 0}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_matrix_cols_zero)
-{
-    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
-    Shape shape_a{0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_vector_zero)
-{
-    Shape shape_a{0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{0}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_matrix_to_scalar_zero_by_zero)
-{
-    Shape shape_a{0, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{0}), read_vector<float>(result)));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_3d_to_matrix_most_sig)
 {
     Shape shape_a{3, 3, 3};
@@ -376,56 +287,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_3d_to_scalar_int32)
               read_vector<int32_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_3d_eliminate_zero_dim)
-{
-    Shape shape_a{3, 0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-
-    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
-    // right value.
-    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0, 0, 0, 0, 0}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_3d_eliminate_zero_dim_int32)
-{
-    Shape shape_a{3, 0, 2};
-    auto A = make_shared<op::Parameter>(element::i32, shape_a);
-    Shape shape_rt{3, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::i32, shape_a);
-    copy_data(a, vector<int32_t>{});
-    auto result = backend->create_tensor(element::i32, shape_rt);
-
-    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
-    // right value.
-    copy_data(result, vector<int32_t>{2112, 2112, 2112, 2112, 2112, 2112});
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<int32_t>{0, 0, 0, 0, 0, 0}), read_vector<int32_t>(result));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_5d_to_scalar)
 {
     Shape shape_a{3, 3, 3, 3, 3};
@@ -489,27 +350,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_2d_to_scalar_int8)
     EXPECT_EQ(std::vector<int8_t>{45}, read_vector<int8_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_trivial_in_double)
-{
-    Shape shape{4, 3};
-    Shape rshape{3};
-    auto A = make_shared<op::Parameter>(element::f64, shape);
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f64, shape);
-    copy_data(a, vector<double>{12, 2, 10, 9, 8, 4, 6, 1, 5, 3, 11, 7});
-    auto result = backend->create_tensor(element::f64, rshape);
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<double>{30, 22, 26}), read_vector<double>(result)));
-}
-
 #if NGRAPH_INTERPRETER_ENABLE
 
 #ifndef _WIN32
@@ -548,39 +388,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_stable_acc)
 }
 #endif
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_stable_acc_double)
-{
-    std::string backend_name = "${BACKEND_NAME}";
-    if (backend_name == "INTERPRETER")
-    {
-        return;
-    }
-    Shape shape_a{10, 10, 20, 300};
-    auto A = make_shared<op::Parameter>(element::f64, shape_a);
-
-    Shape shape_rt{10};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{3}, vector<int32_t>{1, 2, 3});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
-
-    test::Uniform<double> rng(1000000000.0L, 1000000000.001L, 2112);
-    vector<vector<double>> args;
-    for (shared_ptr<op::Parameter> param : f->get_parameters())
-    {
-        vector<double> tensor_val(shape_size(param->get_shape()));
-        rng.initialize(tensor_val);
-        args.push_back(tensor_val);
-    }
-
-    auto ref_func = clone_function(*f);
-    auto bk_func = clone_function(*f);
-
-    auto ref_results = execute(ref_func, args, "INTERPRETER");
-    auto bk_results = execute(bk_func, args, "${BACKEND_NAME}");
-
-    EXPECT_TRUE(test::all_close(ref_results.at(0), bk_results.at(0), 0.0, 1e-5));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_stable_simple_float)
 {
     std::string backend_name = "${BACKEND_NAME}";
@@ -611,106 +418,8 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_stable_simple_float)
         test::all_close_f(ref_results.at(0), bk_results.at(0), DEFAULT_FLOAT_TOLERANCE_BITS - 1));
 }
 
-#ifndef _WIN32
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_stable_simple_double)
-{
-    std::string backend_name = "${BACKEND_NAME}";
-    if (backend_name == "INTERPRETER")
-    {
-        return;
-    }
-    Shape shape_a{20};
-    auto A = make_shared<op::Parameter>(element::f64, shape_a);
-
-    Shape shape_rt{};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
-
-    vector<vector<double>> args;
-    args.push_back(vector<double>{10000000000000000.0L,
-                                  0.2L,
-                                  0.3L,
-                                  0.4L,
-                                  0.5L,
-                                  0.6L,
-                                  0.7L,
-                                  0.8L,
-                                  0.9L,
-                                  0.7L,
-                                  0.9L,
-                                  0.7L,
-                                  0.3L,
-                                  0.6L,
-                                  0.8L,
-                                  0.4L,
-                                  0.6L,
-                                  0.5L,
-                                  0.8L,
-                                  0.7L});
-
-    auto ref_func = clone_function(*f);
-    auto bk_func = clone_function(*f);
-
-    auto ref_results = execute(ref_func, args, "INTERPRETER");
-    auto bk_results = execute(bk_func, args, "${BACKEND_NAME}");
-
-    EXPECT_TRUE(test::all_close(ref_results.at(0), bk_results.at(0), 0.0, 2.0));
-}
 #endif
 
-#endif
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_dynamic)
-{
-    // Create a graph for f(x,axes:int32) = Sum(x,Convert<int64>(axes)).
-    auto x = make_shared<op::Parameter>(element::f32, PartialShape::dynamic());
-    auto axes = make_shared<op::Parameter>(element::i32, PartialShape{Dimension::dynamic()});
-    auto axes_i64 = make_shared<op::Convert>(axes, element::i64);
-
-    auto sum = make_shared<op::v1::ReduceSum>(x, axes_i64, false);
-    ASSERT_TRUE(sum->get_output_partial_shape(0).rank().is_dynamic());
-
-    auto f = make_shared<Function>(NodeVector{sum}, ParameterVector{x, axes});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}", true);
-
-    auto ex = backend->compile(f);
-
-    auto t_r = backend->create_dynamic_tensor(element::f32, PartialShape::dynamic());
-
-    std::vector<Shape> x_shapes{
-        Shape{2, 3}, Shape{2, 3}, Shape{2, 3}, Shape{2, 3}, Shape{5}, Shape{5}};
-    std::vector<std::vector<int32_t>> axeses{{}, {0}, {1}, {0, 1}, {}, {0}};
-    std::vector<std::vector<float>> inputs{{1, 2, 3, 4, 5, 6},
-                                           {1, 2, 3, 4, 5, 6},
-                                           {1, 2, 3, 4, 5, 6},
-                                           {1, 2, 3, 4, 5, 6},
-                                           {1, 2, 3, 4, 5},
-                                           {1, 2, 3, 4, 5}};
-    std::vector<Shape> expected_result_shapes{
-        Shape{2, 3}, Shape{3}, Shape{2}, Shape{}, Shape{5}, Shape{}};
-    std::vector<std::vector<float>> expected_results{
-        {1, 2, 3, 4, 5, 6}, {5, 7, 9}, {6, 15}, {21}, {1, 2, 3, 4, 5}, {15}};
-
-    for (size_t i = 0; i < x_shapes.size(); i++)
-    {
-        auto t_x = backend->create_tensor(element::f32, x_shapes[i]);
-        auto t_axes = backend->create_tensor(element::i32, Shape{axeses[i].size()});
-
-        copy_data(t_x, inputs[i]);
-        copy_data(t_axes, axeses[i]);
-
-        ex->call_with_validate({t_r}, {t_x, t_axes});
-
-        ASSERT_EQ(t_r->get_shape(), expected_result_shapes[i]);
-
-        auto results = read_vector<float>(t_r);
-
-        ASSERT_TRUE(test::all_close_f(results, expected_results[i], MIN_FLOAT_TOLERANCE_BITS));
-    }
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_inf)
 {
     Shape shape{7, 4};
@@ -874,95 +583,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_matrix_rows)
     EXPECT_TRUE(test::all_close_f((vector<float>{3, 7, 11}), read_vector<float>(result)));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_matrix_rows_zero)
-{
-    Shape shape_a{3, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0, 0}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_matrix_cols_zero)
-{
-    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
-    Shape shape_a{0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_vector_zero)
-{
-    Shape shape_a{0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{0}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_matrix_to_scalar_zero_by_zero)
-{
-    Shape shape_a{0, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{0}), read_vector<float>(result)));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_3d_to_matrix_most_sig)
 {
     Shape shape_a{3, 3, 3};
@@ -1099,56 +719,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_3d_to_scalar_int32)
               read_vector<int32_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_3d_eliminate_zero_dim)
-{
-    Shape shape_a{3, 0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 1, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-
-    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
-    // right value.
-    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0, 0, 0, 0, 0}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_3d_eliminate_zero_dim_int32)
-{
-    Shape shape_a{3, 0, 2};
-    auto A = make_shared<op::Parameter>(element::i32, shape_a);
-    Shape shape_rt{3, 1, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::i32, shape_a);
-    copy_data(a, vector<int32_t>{});
-    auto result = backend->create_tensor(element::i32, shape_rt);
-
-    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
-    // right value.
-    copy_data(result, vector<int32_t>{2112, 2112, 2112, 2112, 2112, 2112});
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<int32_t>{0, 0, 0, 0, 0, 0}), read_vector<int32_t>(result));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_5d_to_scalar)
 {
     Shape shape_a{3, 3, 3, 3, 3};
@@ -1212,27 +782,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_2d_to_scalar_int8)
     EXPECT_EQ(std::vector<int8_t>{45}, read_vector<int8_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_trivial_in_double)
-{
-    Shape shape{4, 3};
-    Shape rshape{1, 3};
-    auto A = make_shared<op::Parameter>(element::f64, shape);
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f64, shape);
-    copy_data(a, vector<double>{12, 2, 10, 9, 8, 4, 6, 1, 5, 3, 11, 7});
-    auto result = backend->create_tensor(element::f64, rshape);
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<double>{30, 22, 26}), read_vector<double>(result)));
-}
-
 #if NGRAPH_INTERPRETER_ENABLE
 
 #ifndef _WIN32
@@ -1271,38 +820,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_stable_acc)
 }
 #endif
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_stable_acc_double)
-{
-    std::string backend_name = "${BACKEND_NAME}";
-    if (backend_name == "INTERPRETER")
-    {
-        return;
-    }
-    Shape shape_a{10, 10, 20, 300};
-    auto A = make_shared<op::Parameter>(element::f64, shape_a);
-
-    Shape shape_rt{10, 1, 1, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{3}, vector<int32_t>{1, 2, 3});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
-
-    test::Uniform<double> rng(1000000000.0L, 1000000000.001L, 2112);
-    vector<vector<double>> args;
-    for (shared_ptr<op::Parameter> param : f->get_parameters())
-    {
-        vector<double> tensor_val(shape_size(param->get_shape()));
-        rng.initialize(tensor_val);
-        args.push_back(tensor_val);
-    }
-
-    auto ref_func = clone_function(*f);
-    auto bk_func = clone_function(*f);
-
-    auto ref_results = execute(ref_func, args, "INTERPRETER");
-    auto bk_results = execute(bk_func, args, "${BACKEND_NAME}");
-
-    EXPECT_TRUE(test::all_close(ref_results.at(0), bk_results.at(0), 0.0, 1e-5));
-}
 
 NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_stable_simple_float)
 {
@@ -1334,106 +851,8 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_stable_simple_float)
         test::all_close_f(ref_results.at(0), bk_results.at(0), DEFAULT_FLOAT_TOLERANCE_BITS - 1));
 }
 
-#ifndef _WIN32
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_stable_simple_double)
-{
-    std::string backend_name = "${BACKEND_NAME}";
-    if (backend_name == "INTERPRETER")
-    {
-        return;
-    }
-    Shape shape_a{20};
-    auto A = make_shared<op::Parameter>(element::f64, shape_a);
-
-    Shape shape_rt{1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
-
-    vector<vector<double>> args;
-    args.push_back(vector<double>{10000000000000000.0L,
-                                  0.2L,
-                                  0.3L,
-                                  0.4L,
-                                  0.5L,
-                                  0.6L,
-                                  0.7L,
-                                  0.8L,
-                                  0.9L,
-                                  0.7L,
-                                  0.9L,
-                                  0.7L,
-                                  0.3L,
-                                  0.6L,
-                                  0.8L,
-                                  0.4L,
-                                  0.6L,
-                                  0.5L,
-                                  0.8L,
-                                  0.7L});
-
-    auto ref_func = clone_function(*f);
-    auto bk_func = clone_function(*f);
-
-    auto ref_results = execute(ref_func, args, "INTERPRETER");
-    auto bk_results = execute(bk_func, args, "${BACKEND_NAME}");
-
-    EXPECT_TRUE(test::all_close(ref_results.at(0), bk_results.at(0), 0.0, 2.0));
-}
-#endif
-
 #endif
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_dynamic)
-{
-    // Create a graph for f(x,axes:int32) = Sum(x,Convert<int64>(axes)).
-    auto x = make_shared<op::Parameter>(element::f32, PartialShape::dynamic());
-    auto axes = make_shared<op::Parameter>(element::i32, PartialShape{Dimension::dynamic()});
-    auto axes_i64 = make_shared<op::Convert>(axes, element::i64);
-
-    auto sum = make_shared<op::v1::ReduceSum>(x, axes_i64, true);
-    ASSERT_TRUE(sum->get_output_partial_shape(0).rank().is_dynamic());
-
-    auto f = make_shared<Function>(NodeVector{sum}, ParameterVector{x, axes});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}", true);
-
-    auto ex = backend->compile(f);
-
-    auto t_r = backend->create_dynamic_tensor(element::f32, PartialShape::dynamic());
-
-    std::vector<Shape> x_shapes{
-        Shape{2, 3}, Shape{2, 3}, Shape{2, 3}, Shape{2, 3}, Shape{5}, Shape{5}};
-    std::vector<std::vector<int32_t>> axeses{{}, {0}, {1}, {0, 1}, {}, {0}};
-    std::vector<std::vector<float>> inputs{{1, 2, 3, 4, 5, 6},
-                                           {1, 2, 3, 4, 5, 6},
-                                           {1, 2, 3, 4, 5, 6},
-                                           {1, 2, 3, 4, 5, 6},
-                                           {1, 2, 3, 4, 5},
-                                           {1, 2, 3, 4, 5}};
-    std::vector<Shape> expected_result_shapes{
-        Shape{2, 3}, Shape{1, 3}, Shape{2, 1}, Shape{1, 1}, Shape{5}, Shape{1}};
-    std::vector<std::vector<float>> expected_results{
-        {1, 2, 3, 4, 5, 6}, {5, 7, 9}, {6, 15}, {21}, {1, 2, 3, 4, 5}, {15}};
-
-    for (size_t i = 0; i < x_shapes.size(); i++)
-    {
-        auto t_x = backend->create_tensor(element::f32, x_shapes[i]);
-        auto t_axes = backend->create_tensor(element::i32, Shape{axeses[i].size()});
-
-        copy_data(t_x, inputs[i]);
-        copy_data(t_axes, axeses[i]);
-
-        ex->call_with_validate({t_r}, {t_x, t_axes});
-
-        ASSERT_EQ(t_r->get_shape(), expected_result_shapes[i]);
-
-        auto results = read_vector<float>(t_r);
-
-        ASSERT_TRUE(test::all_close_f(results, expected_results[i], MIN_FLOAT_TOLERANCE_BITS));
-    }
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_sum_keep_inf)
 {
     Shape shape{7, 4};
diff --git a/ngraph/test/op_eval/reduce_sum.cpp b/ngraph/test/op_eval/reduce_sum.cpp
new file mode 100644
index 00000000000000..bad7aad21edf90
--- /dev/null
+++ b/ngraph/test/op_eval/reduce_sum.cpp
@@ -0,0 +1,395 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/test_control.hpp"
+#include "util/all_close.hpp"
+#include "util/all_close_f.hpp"
+#include "util/ndarray.hpp"
+
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+
+TEST(op_eval, reduce_sum_matrix_rows_zero)
+{
+    Shape shape_a{3, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0, 0}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_sum_vector_zero)
+{
+    Shape shape_a{0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{0}), read_vector<float>(result)));
+}
+
+
+TEST(op_eval, reduce_sum_matrix_cols_zero)
+{
+    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
+    Shape shape_a{0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_sum_matrix_to_scalar_zero_by_zero)
+{
+    Shape shape_a{0, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{0}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_sum_3d_eliminate_zero_dim)
+{
+    Shape shape_a{3, 0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+
+    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
+    // right value.
+    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0, 0, 0, 0, 0}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_sum_3d_eliminate_zero_dim_int32)
+{
+    Shape shape_a{3, 0, 2};
+    auto A = make_shared<op::Parameter>(element::i32, shape_a);
+    Shape shape_rt{3, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::i32, shape_a);
+    copy_data(a, vector<int32_t>{});
+    auto result = backend->create_tensor(element::i32, shape_rt);
+
+    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
+    // right value.
+    copy_data(result, vector<int32_t>{2112, 2112, 2112, 2112, 2112, 2112});
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<int32_t>{0, 0, 0, 0, 0, 0}), read_vector<int32_t>(result));
+}
+
+TEST(op_eval, reduce_sum_dynamic)
+{
+    // Create a graph for f(x,axes:int32) = Sum(x,Convert<int64>(axes)).
+    auto x = make_shared<op::Parameter>(element::f32, PartialShape::dynamic());
+    auto axes = make_shared<op::Parameter>(element::i32, PartialShape{Dimension::dynamic()});
+    auto axes_i64 = make_shared<op::Convert>(axes, element::i64);
+
+    auto sum = make_shared<op::v1::ReduceSum>(x, axes_i64, false);
+    ASSERT_TRUE(sum->get_output_partial_shape(0).rank().is_dynamic());
+
+    auto f = make_shared<Function>(NodeVector{sum}, ParameterVector{x, axes});
+
+    auto backend = runtime::Backend::create("INTERPRETER", true);
+
+    auto ex = backend->compile(f);
+
+    auto t_r = backend->create_dynamic_tensor(element::f32, PartialShape::dynamic());
+
+    std::vector<Shape> x_shapes{
+        Shape{2, 3}, Shape{2, 3}, Shape{2, 3}, Shape{2, 3}, Shape{5}, Shape{5}};
+    std::vector<std::vector<int32_t>> axeses{{}, {0}, {1}, {0, 1}, {}, {0}};
+    std::vector<std::vector<float>> inputs{{1, 2, 3, 4, 5, 6},
+                                           {1, 2, 3, 4, 5, 6},
+                                           {1, 2, 3, 4, 5, 6},
+                                           {1, 2, 3, 4, 5, 6},
+                                           {1, 2, 3, 4, 5},
+                                           {1, 2, 3, 4, 5}};
+    std::vector<Shape> expected_result_shapes{
+        Shape{2, 3}, Shape{3}, Shape{2}, Shape{}, Shape{5}, Shape{}};
+    std::vector<std::vector<float>> expected_results{
+        {1, 2, 3, 4, 5, 6}, {5, 7, 9}, {6, 15}, {21}, {1, 2, 3, 4, 5}, {15}};
+
+    for (size_t i = 0; i < x_shapes.size(); i++)
+    {
+        auto t_x = backend->create_tensor(element::f32, x_shapes[i]);
+        auto t_axes = backend->create_tensor(element::i32, Shape{axeses[i].size()});
+
+        copy_data(t_x, inputs[i]);
+        copy_data(t_axes, axeses[i]);
+
+        ex->call_with_validate({t_r}, {t_x, t_axes});
+
+        ASSERT_EQ(t_r->get_shape(), expected_result_shapes[i]);
+
+        auto results = read_vector<float>(t_r);
+
+        ASSERT_TRUE(test::all_close_f(results, expected_results[i], MIN_FLOAT_TOLERANCE_BITS));
+    }
+}
+
+TEST(op_eval, reduce_sum_keep_matrix_rows_zero)
+{
+    Shape shape_a{3, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0, 0}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_sum_keep_matrix_cols_zero)
+{
+    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
+    Shape shape_a{0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_sum_keep_vector_zero)
+{
+    Shape shape_a{0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{0}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_sum_keep_matrix_to_scalar_zero_by_zero)
+{
+    Shape shape_a{0, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1, 1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{0}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_sum_keep_3d_eliminate_zero_dim)
+{
+    Shape shape_a{3, 0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 1, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+
+    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
+    // right value.
+    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0, 0, 0, 0, 0}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_sum_keep_3d_eliminate_zero_dim_int32)
+{
+    Shape shape_a{3, 0, 2};
+    auto A = make_shared<op::Parameter>(element::i32, shape_a);
+    Shape shape_rt{3, 1, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceSum>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::i32, shape_a);
+    copy_data(a, vector<int32_t>{});
+    auto result = backend->create_tensor(element::i32, shape_rt);
+
+    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
+    // right value.
+    copy_data(result, vector<int32_t>{2112, 2112, 2112, 2112, 2112, 2112});
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<int32_t>{0, 0, 0, 0, 0, 0}), read_vector<int32_t>(result));
+}
+
+TEST(op_eval, reduce_sum_keep_dynamic)
+{
+    // Create a graph for f(x,axes:int32) = Sum(x,Convert<int64>(axes)).
+    auto x = make_shared<op::Parameter>(element::f32, PartialShape::dynamic());
+    auto axes = make_shared<op::Parameter>(element::i32, PartialShape{Dimension::dynamic()});
+    auto axes_i64 = make_shared<op::Convert>(axes, element::i64);
+
+    auto sum = make_shared<op::v1::ReduceSum>(x, axes_i64, true);
+    ASSERT_TRUE(sum->get_output_partial_shape(0).rank().is_dynamic());
+
+    auto f = make_shared<Function>(NodeVector{sum}, ParameterVector{x, axes});
+
+    auto backend = runtime::Backend::create("INTERPRETER", true);
+
+    auto ex = backend->compile(f);
+
+    auto t_r = backend->create_dynamic_tensor(element::f32, PartialShape::dynamic());
+
+    std::vector<Shape> x_shapes{
+        Shape{2, 3}, Shape{2, 3}, Shape{2, 3}, Shape{2, 3}, Shape{5}, Shape{5}};
+    std::vector<std::vector<int32_t>> axeses{{}, {0}, {1}, {0, 1}, {}, {0}};
+    std::vector<std::vector<float>> inputs{{1, 2, 3, 4, 5, 6},
+                                           {1, 2, 3, 4, 5, 6},
+                                           {1, 2, 3, 4, 5, 6},
+                                           {1, 2, 3, 4, 5, 6},
+                                           {1, 2, 3, 4, 5},
+                                           {1, 2, 3, 4, 5}};
+    std::vector<Shape> expected_result_shapes{
+        Shape{2, 3}, Shape{1, 3}, Shape{2, 1}, Shape{1, 1}, Shape{5}, Shape{1}};
+    std::vector<std::vector<float>> expected_results{
+        {1, 2, 3, 4, 5, 6}, {5, 7, 9}, {6, 15}, {21}, {1, 2, 3, 4, 5}, {15}};
+
+    for (size_t i = 0; i < x_shapes.size(); i++)
+    {
+        auto t_x = backend->create_tensor(element::f32, x_shapes[i]);
+        auto t_axes = backend->create_tensor(element::i32, Shape{axeses[i].size()});
+
+        copy_data(t_x, inputs[i]);
+        copy_data(t_axes, axeses[i]);
+
+        ex->call_with_validate({t_r}, {t_x, t_axes});
+
+        ASSERT_EQ(t_r->get_shape(), expected_result_shapes[i]);
+
+        auto results = read_vector<float>(t_r);
+
+        ASSERT_TRUE(test::all_close_f(results, expected_results[i], MIN_FLOAT_TOLERANCE_BITS));
+    }
+}
diff --git a/ngraph/test/runtime/ie/unit_test.manifest b/ngraph/test/runtime/ie/unit_test.manifest
index f95ce0d951e162..425ec0cdf48e30 100644
--- a/ngraph/test/runtime/ie/unit_test.manifest
+++ b/ngraph/test/runtime/ie/unit_test.manifest
@@ -368,36 +368,16 @@ any_2x0_to_scalar
 all_trivial
 all_2x0_to_scalar
 all_dynamic
-reduce_sum_keep_dynamic
-reduce_sum_keep_stable_simple_double
-reduce_sum_keep_stable_acc_double
-reduce_sum_keep_stable_acc
-reduce_sum_keep_3d_eliminate_zero_dim
-reduce_sum_keep_vector_zero
-reduce_sum_keep_matrix_rows_zero
-reduce_sum_dynamic
-reduce_sum_3d_eliminate_zero_dim_int32
-reduce_sum_keep_3d_eliminate_zero_dim_int32
-reduce_sum_keep_trivial_in_double
-reduce_sum_3d_eliminate_zero_dim
-reduce_sum_matrix_to_scalar_zero_by_zero
-reduce_sum_vector_zero
-reduce_sum_matrix_cols_zero
-reduce_sum_matrix_rows_zero
-reduce_sum_keep_matrix_to_scalar_zero_by_zero
-reduce_sum_keep_matrix_cols_zero
-reduce_product_matrix_columns
 
+# Reduce ops disabled/accuracy: 56520
+# disabled reference implementation
+reduce_sum_keep_2d_to_scalar_int8
+reduce_sum_2d_to_scalar_int8
 # accuracy
 reduce_sum_keep_stable_acc
-reduce_sum_keep_2d_to_scalar_int8
 reduce_sum_keep_3d_to_scalar_int32
 reduce_sum_keep_large_1d_to_scalar
-reduce_sum_stable_simple_double
-reduce_sum_stable_acc_double
 reduce_sum_stable_acc
-reduce_sum_trivial_in_double
-reduce_sum_2d_to_scalar_int8
 reduce_sum_3d_to_scalar_int32
 reduce_sum_large_1d_to_scalar
 
diff --git a/ngraph/test/runtime/interpreter/unit_test.manifest b/ngraph/test/runtime/interpreter/unit_test.manifest
index b5003831265c9b..9e1231491521e3 100644
--- a/ngraph/test/runtime/interpreter/unit_test.manifest
+++ b/ngraph/test/runtime/interpreter/unit_test.manifest
@@ -3,6 +3,7 @@ fake_quantize_pdpd
 INTERPRETER.onnx_model_quant_conv_linear
 INTERPRETER.onnx_top_k_opset_10
 
+# Failed in ubuntu18 i386
 reduce_sum_large_1d_to_scalar
 reduce_sum_keep_large_1d_to_scalar
 
@@ -31,11 +32,8 @@ INTERPRETER.onnx_model_matmul_integer_4d_no_zero_point
 # Disabled tests for disabled reference implementations
 INTERPRETER.onnx_dyn_shapes_expand_uint16_dyn_shape
 INTERPRETER.sum_2d_to_scalar_int8
-INTERPRETER.sum_trivial_in_double
 INTERPRETER.reduce_sum_2d_to_scalar_int8
-INTERPRETER.reduce_sum_trivial_in_double
 INTERPRETER.reduce_sum_keep_2d_to_scalar_int8
-INTERPRETER.reduce_sum_keep_trivial_in_double
 INTERPRETER.reduce_product_to_scalar_int8
 INTERPRETER.reduce_product_keep_to_scalar_int8
 INTERPRETER.reduce_min_to_scalar_int8

From c8af311774385b22949a208aa1713d50b23c2224 Mon Sep 17 00:00:00 2001
From: Gabriele Galiero Casay <gabriele.galiero.casay@intel.com>
Date: Fri, 11 Jun 2021 09:23:08 +0200
Subject: [PATCH 065/178] Revise reduction operations base classes (#5609)

* Revise reduction operations base class

* Fix visitor tests for logical reduction ops

* Remove node validation checks that break backward compatibility with tests
  * Check for integer element type of axes input
  * Check for unique elements provided in axes input

* Set mvn fusion test to have axes input with element type integral for ReduceMean nodes

* [LPT] Revise reduction operations base classes support in LPT

* Add node validation checks for integer element type of axes input in reduction operations

* Add end of line to files

* Implement output shape inference of reduction ops as member function

* Add ReduceBase class to have a common method to infer output shape

Co-authored-by: Edward Shogulin <edward.shogulin@intel.com>
---
 .../src/normalize_l2.cpp                      |   4 +-
 .../transformations/mvn_fusion_test.cpp       |  20 +-
 .../src/normalize_l2_function.cpp             |   6 +-
 .../transformations_after_split_function.cpp  |   2 +-
 .../ngraph/op/util/arithmetic_reduction.hpp   |   3 +-
 .../ngraph/op/util/logical_reduction.hpp      |   3 +-
 .../include/ngraph/op/util/reduction_base.hpp |  39 ++
 .../core/src/op/util/arithmetic_reduction.cpp |  52 +--
 .../util/arithmetic_reductions_keep_dims.cpp  |  79 ++--
 ngraph/core/src/op/util/logical_reduction.cpp |  71 +---
 .../op/util/logical_reduction_keep_dims.cpp   |  71 +---
 ngraph/core/src/op/util/reduction_base.cpp    |  74 ++++
 ngraph/test/CMakeLists.txt                    |  10 +-
 ngraph/test/type_prop/reduce_l1.cpp           |  47 +--
 ngraph/test/type_prop/reduce_l2.cpp           |  47 +--
 ngraph/test/type_prop/reduce_logical_and.cpp  |   9 +
 ngraph/test/type_prop/reduce_logical_or.cpp   |   9 +
 ngraph/test/type_prop/reduce_max.cpp          |   9 +
 ngraph/test/type_prop/reduce_mean.cpp         |  71 +---
 ngraph/test/type_prop/reduce_min.cpp          |   9 +
 ngraph/test/type_prop/reduce_ops.hpp          | 388 ++++++++++++++++++
 ngraph/test/type_prop/reduce_prod.cpp         |  48 +--
 ngraph/test/type_prop/reduce_sum.cpp          |  48 +--
 ngraph/test/visitors/op/reduce_ops.hpp        |   2 +-
 24 files changed, 649 insertions(+), 472 deletions(-)
 create mode 100644 ngraph/core/include/ngraph/op/util/reduction_base.hpp
 create mode 100644 ngraph/core/src/op/util/reduction_base.cpp
 create mode 100644 ngraph/test/type_prop/reduce_logical_and.cpp
 create mode 100644 ngraph/test/type_prop/reduce_logical_or.cpp
 create mode 100644 ngraph/test/type_prop/reduce_max.cpp
 create mode 100644 ngraph/test/type_prop/reduce_min.cpp
 create mode 100644 ngraph/test/type_prop/reduce_ops.hpp

diff --git a/inference-engine/src/low_precision_transformations/src/normalize_l2.cpp b/inference-engine/src/low_precision_transformations/src/normalize_l2.cpp
index 4368a48075f324..93dcdf8afbe551 100644
--- a/inference-engine/src/low_precision_transformations/src/normalize_l2.cpp
+++ b/inference-engine/src/low_precision_transformations/src/normalize_l2.cpp
@@ -120,10 +120,10 @@ bool NormalizeL2Transformation::transform(TransformationContext &context, ngraph
     }
 
     auto newNormalize = std::make_shared<op::TypeRelaxed<opset1::NormalizeL2>>(
-        std::vector<ngraph::element::Type>{ element::f32, element::f32 },
+        std::vector<ngraph::element::Type>{ element::f32, axes->output(0).get_element_type() },
         std::vector<ngraph::element::Type>{deqPrecision},
         ngraph::op::TemporaryReplaceOutputType(dequantization.subtract == nullptr ? dequantization.data : dequantization.subtract, element::f32).get(),
-        ngraph::op::TemporaryReplaceOutputType(axes->clone_with_new_inputs({}), element::f32).get(),
+        axes,
         normalize->get_eps(),
         normalize->get_eps_mode());
     NetworkHelper::copyInfo(normalize, newNormalize);
diff --git a/inference-engine/tests/functional/inference_engine/transformations/mvn_fusion_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/mvn_fusion_test.cpp
index e45258a63e67b3..ecad9adff3df5f 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/mvn_fusion_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/mvn_fusion_test.cpp
@@ -30,7 +30,7 @@ TEST(TransformationTests, MVNFusionTestOutside) {
         auto sub2 = std::make_shared<ngraph::opset6::Subtract>(input, mean2);
         auto const_2 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 2 });
         auto power_sqr = std::make_shared<ngraph::opset6::Power>(sub2, const_2);
-        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::i32, ngraph::Shape{ 3 }, { 1, 2, 3 });
         auto mean3 = std::make_shared<ngraph::opset6::ReduceMean>(power_sqr, mean3_axes);
         auto const_0_5 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 0.5 });
         auto power_sqrt = std::make_shared<ngraph::opset6::Power>(mean3, const_0_5);
@@ -70,7 +70,7 @@ TEST(TransformationTests, MVNFusionTestReuseSub) {
         auto sub1 = std::make_shared<ngraph::opset6::Subtract>(input, mean1);
         auto const_2 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 2 });
         auto power_sqr = std::make_shared<ngraph::opset6::Power>(sub1, const_2);
-        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::i32, ngraph::Shape{ 3 }, { 1, 2, 3 });
         auto mean3 = std::make_shared<ngraph::opset6::ReduceMean>(power_sqr, mean3_axes);
         auto const_0_5 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 0.5 });
         auto power_sqrt = std::make_shared<ngraph::opset6::Power>(mean3, const_0_5);
@@ -111,7 +111,7 @@ TEST(TransformationTests, MVNFusionTestWithConvert) {
         auto cast = std::make_shared<ngraph::opset6::Convert>(sub1, ngraph::element::f32);
         auto const_2 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 2 });
         auto power_sqr = std::make_shared<ngraph::opset6::Power>(cast, const_2);
-        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::i32, ngraph::Shape{ 3 }, { 1, 2, 3 });
         auto mean3 = std::make_shared<ngraph::opset6::ReduceMean>(power_sqr, mean3_axes);
         auto const_0_5 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 0.5 });
         auto power_sqrt = std::make_shared<ngraph::opset6::Power>(mean3, const_0_5);
@@ -151,7 +151,7 @@ TEST(TransformationTests, MVNFusionTestSqrt) {
         auto sub1 = std::make_shared<ngraph::opset6::Subtract>(input, mean1);
         auto const_2 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 2 });
         auto power_sqr = std::make_shared<ngraph::opset6::Power>(sub1, const_2);
-        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::i32, ngraph::Shape{ 3 }, { 1, 2, 3 });
         auto mean3 = std::make_shared<ngraph::opset6::ReduceMean>(power_sqr, mean3_axes);
         auto power_sqrt = std::make_shared<ngraph::opset6::Sqrt>(mean3);
         auto eps = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 1e-9 });
@@ -190,7 +190,7 @@ TEST(TransformationTests, MVNFusionTestAltDiv) {
         auto sub1 = std::make_shared<ngraph::opset6::Subtract>(input, mean1);
         auto const_2 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 2 });
         auto power_sqr = std::make_shared<ngraph::opset6::Power>(sub1, const_2);
-        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::i32, ngraph::Shape{ 3 }, { 1, 2, 3 });
         auto mean3 = std::make_shared<ngraph::opset6::ReduceMean>(power_sqr, mean3_axes);
         auto const_0_5 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 0.5 });
         auto power_sqrt = std::make_shared<ngraph::opset6::Power>(mean3, const_0_5);
@@ -231,7 +231,7 @@ TEST(TransformationTests, MVNFusionTestInsideSqrt) {
         auto sub2 = std::make_shared<ngraph::opset6::Subtract>(input, mean2);
         auto const_2 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 2 });
         auto power_sqr = std::make_shared<ngraph::opset6::Power>(sub2, const_2);
-        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::i32, ngraph::Shape{ 3 }, { 1, 2, 3 });
         auto mean3 = std::make_shared<ngraph::opset6::ReduceMean>(power_sqr, mean3_axes);
         auto eps = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 1e-9 });
         auto add_eps = std::make_shared<ngraph::opset6::Add>(mean3, eps);
@@ -271,7 +271,7 @@ TEST(TransformationTests, MVNFusionTestReuseSubInsideSqrt) {
         auto sub1 = std::make_shared<ngraph::opset6::Subtract>(input, mean1);
         auto const_2 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 2 });
         auto power_sqr = std::make_shared<ngraph::opset6::Power>(sub1, const_2);
-        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::i32, ngraph::Shape{ 3 }, { 1, 2, 3 });
         auto mean3 = std::make_shared<ngraph::opset6::ReduceMean>(power_sqr, mean3_axes);
         auto eps = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 1e-9 });
         auto add_eps = std::make_shared<ngraph::opset6::Add>(mean3, eps);
@@ -312,7 +312,7 @@ TEST(TransformationTests, MVNFusionTestWithConvertInsideSqrt) {
         auto cast = std::make_shared<ngraph::opset6::Convert>(sub1, ngraph::element::f32);
         auto const_2 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 2 });
         auto power_sqr = std::make_shared<ngraph::opset6::Power>(cast, const_2);
-        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::i32, ngraph::Shape{ 3 }, { 1, 2, 3 });
         auto mean3 = std::make_shared<ngraph::opset6::ReduceMean>(power_sqr, mean3_axes);
         auto eps = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 1e-9 });
         auto add_eps = std::make_shared<ngraph::opset6::Add>(mean3, eps);
@@ -352,7 +352,7 @@ TEST(TransformationTests, MVNFusionTestSqrtInsideSqrt) {
         auto sub1 = std::make_shared<ngraph::opset6::Subtract>(input, mean1);
         auto const_2 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 2 });
         auto power_sqr = std::make_shared<ngraph::opset6::Power>(sub1, const_2);
-        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::i32, ngraph::Shape{ 3 }, { 1, 2, 3 });
         auto mean3 = std::make_shared<ngraph::opset6::ReduceMean>(power_sqr, mean3_axes);
         auto eps = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 1e-9 });
         auto add_eps = std::make_shared<ngraph::opset6::Add>(mean3, eps);
@@ -391,7 +391,7 @@ TEST(TransformationTests, MVNFusionTestAltDivInsideSqrt) {
         auto sub1 = std::make_shared<ngraph::opset6::Subtract>(input, mean1);
         auto const_2 = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 2 });
         auto power_sqr = std::make_shared<ngraph::opset6::Power>(sub1, const_2);
-        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        auto mean3_axes = ngraph::opset6::Constant::create(ngraph::element::i32, ngraph::Shape{ 3 }, { 1, 2, 3 });
         auto mean3 = std::make_shared<ngraph::opset6::ReduceMean>(power_sqr, mean3_axes);
         auto eps = ngraph::opset6::Constant::create(ngraph::element::f32, ngraph::Shape{}, { 1e-9 });
         auto add_eps = std::make_shared<ngraph::opset6::Add>(mean3, eps);
diff --git a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/normalize_l2_function.cpp b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/normalize_l2_function.cpp
index 553ec6575ec5cd..4d1ef00a598fea 100644
--- a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/normalize_l2_function.cpp
+++ b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/normalize_l2_function.cpp
@@ -36,7 +36,7 @@ std::shared_ptr<ngraph::Function> NormalizeL2Function::getOriginal(
 
     fakeQuantize->set_friendly_name("fakeQuantize");
 
-    const auto axesNode = std::make_shared<ngraph::op::Constant>(ngraph::element::u64, ngraph::Shape{ axes.size() }, axes);
+    const auto axesNode = std::make_shared<ngraph::op::Constant>(ngraph::element::i64, ngraph::Shape{ axes.size() }, axes);
     axesNode->set_friendly_name("axes");
     const auto normalizeL2 = std::make_shared<ngraph::opset1::NormalizeL2>(fakeQuantize->output(0), axesNode, 1e-6, ngraph::op::EpsMode::ADD);
     normalizeL2->set_friendly_name("normalizeL2");
@@ -104,10 +104,10 @@ std::shared_ptr<ngraph::Function> NormalizeL2Function::getReference(
 
     const auto axesNode = std::make_shared<ngraph::op::Constant>(ngraph::element::i64, ngraph::Shape{ axes.size() }, axes);
     const auto normalizeL2 = std::make_shared<ngraph::op::TypeRelaxed<ngraph::opset1::NormalizeL2>>(
-        std::vector<ngraph::element::Type>{ element::f32, element::f32 },
+        std::vector<ngraph::element::Type>{ element::f32, axesNode->output(0).get_element_type() },
         std::vector<ngraph::element::Type>{dequantizationAfter.empty() ? precision : element::f32},
         ngraph::op::TemporaryReplaceOutputType(deqBefore, element::f32).get(),
-        ngraph::op::TemporaryReplaceOutputType(axesNode, element::f32).get(),
+        axesNode,
         1e-6,
         epsMode);
     auto& rtInfo = normalizeL2->get_rt_info();
diff --git a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/transformations_after_split_function.cpp b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/transformations_after_split_function.cpp
index 8b2010abb76a12..ad8fd6715925ee 100644
--- a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/transformations_after_split_function.cpp
+++ b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/transformations_after_split_function.cpp
@@ -137,7 +137,7 @@ std::shared_ptr<Node> TransformationsAfterSplitFunction::getLayerByTransformatio
     }
     if (transformationName == "NormalizeL2Transformation") {
         const auto dequantization = makeDequantization(parent, { {element::f32}, {}, { 0.1f } });
-        const auto axesNode = opset1::Constant::create(element::u64, ngraph::Shape{ 3 }, { 1, 2, 3 });
+        const auto axesNode = opset1::Constant::create(element::i64, ngraph::Shape{ 3 }, { 1, 2, 3 });
         return std::make_shared<ngraph::opset1::NormalizeL2>(dequantization, axesNode, 1e-6, ngraph::op::EpsMode::ADD);
     }
     if (transformationName == "PReluTransformation") {
diff --git a/ngraph/core/include/ngraph/op/util/arithmetic_reduction.hpp b/ngraph/core/include/ngraph/op/util/arithmetic_reduction.hpp
index 2d433e340d21de..6f9484ab340c05 100644
--- a/ngraph/core/include/ngraph/op/util/arithmetic_reduction.hpp
+++ b/ngraph/core/include/ngraph/op/util/arithmetic_reduction.hpp
@@ -5,6 +5,7 @@
 #pragma once
 
 #include "ngraph/op/op.hpp"
+#include "ngraph/op/util/reduction_base.hpp"
 
 namespace ngraph
 {
@@ -15,7 +16,7 @@ namespace ngraph
             /// \brief Abstract base class for arithmetic reduction operations, i.e., operations
             ///        where chosen axes of the input tensors are eliminated (reduced out) by
             ///        repeated application of a particular binary arithmetic operation.
-            class NGRAPH_API ArithmeticReduction : public Op
+            class NGRAPH_API ArithmeticReduction : public ReductionBase
             {
             protected:
                 /// \brief Constructs an arithmetic reduction operation.
diff --git a/ngraph/core/include/ngraph/op/util/logical_reduction.hpp b/ngraph/core/include/ngraph/op/util/logical_reduction.hpp
index 0fa85369b3187c..1c8953ba1fd9c1 100644
--- a/ngraph/core/include/ngraph/op/util/logical_reduction.hpp
+++ b/ngraph/core/include/ngraph/op/util/logical_reduction.hpp
@@ -5,6 +5,7 @@
 #pragma once
 
 #include "ngraph/op/op.hpp"
+#include "ngraph/op/util/reduction_base.hpp"
 
 namespace ngraph
 {
@@ -15,7 +16,7 @@ namespace ngraph
             /// \brief Abstract base class for logical reduction operations, i.e., operations where
             ///        chosen axes of the input tensors are eliminated (reduced out) by repeated
             ///        application of a particular binary logical operation.
-            class NGRAPH_API LogicalReduction : public Op
+            class NGRAPH_API LogicalReduction : public ReductionBase
             {
             protected:
                 /// \brief Constructs a logical reduction operation.
diff --git a/ngraph/core/include/ngraph/op/util/reduction_base.hpp b/ngraph/core/include/ngraph/op/util/reduction_base.hpp
new file mode 100644
index 00000000000000..5cc10b0be7e217
--- /dev/null
+++ b/ngraph/core/include/ngraph/op/util/reduction_base.hpp
@@ -0,0 +1,39 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "ngraph/op/op.hpp"
+
+namespace ngraph
+{
+    namespace op
+    {
+        namespace util
+        {
+            class NGRAPH_API ReductionBase : public Op
+            {
+            protected:
+                /// \brief Constructs a reduction operation.
+                ReductionBase();
+
+                /// \brief Constructs a reduction operation.
+                ///
+                /// \param arg Output that produces the first input tensor.
+                /// \param reduction_axes The axis positions (0-based) to be eliminated.
+                ReductionBase(const Output<Node>& arg, const Output<Node>& reduction_axes);
+
+                /// \brief      Infers reduction operations output shape.
+                ///
+                /// \param[in] keep_dims    Reduction operation keeps dimensions.
+                ///
+                /// \return Partial shape of the output.
+                PartialShape infer_reduction_output_shape(const bool keep_dims);
+
+            public:
+                NGRAPH_RTTI_DECLARATION;
+            };
+        } // namespace util
+    }     // namespace op
+} // namespace ngraph
\ No newline at end of file
diff --git a/ngraph/core/src/op/util/arithmetic_reduction.cpp b/ngraph/core/src/op/util/arithmetic_reduction.cpp
index beb40801ea9636..67b7abeaf20bbd 100644
--- a/ngraph/core/src/op/util/arithmetic_reduction.cpp
+++ b/ngraph/core/src/op/util/arithmetic_reduction.cpp
@@ -16,7 +16,7 @@ op::util::ArithmeticReduction::ArithmeticReduction() {}
 
 op::util::ArithmeticReduction::ArithmeticReduction(const Output<Node>& arg,
                                                    const Output<Node>& reduction_axes)
-    : Op({arg, reduction_axes})
+    : ReductionBase(arg, reduction_axes)
 {
 }
 
@@ -49,51 +49,15 @@ void op::util::ArithmeticReduction::set_reduction_axes(const AxisSet& reduction_
 void op::util::ArithmeticReduction::validate_and_infer_types()
 {
     NGRAPH_OP_SCOPE(util_ArithmeticReduction_validate_and_infer_types);
-    auto input_shape = get_input_partial_shape(0);
-    const auto input_rank = input_shape.rank();
 
-    PartialShape result_shape{PartialShape::dynamic()};
-
-    auto axes = get_constant_from_source(input_value(1));
-    if (input_rank.is_static() && axes)
-    {
-        AxisSet reduction_axes;
-        const auto reduction_axes_val = axes->cast_vector<int64_t>();
-        for (auto axis : reduction_axes_val)
-        {
-            try
-            {
-                axis = normalize_axis(this, axis, input_rank);
-            }
-            catch (const ngraph_error&)
-            {
-                NODE_VALIDATION_CHECK(this,
-                                      false,
-                                      "Reduction axis (",
-                                      axis,
-                                      ") is out of bounds ",
-                                      "(argument shape: ",
-                                      input_shape,
-                                      ", reduction axes: ",
-                                      reduction_axes,
-                                      ")");
-            }
-            reduction_axes.insert(axis);
-        }
-
-        std::vector<Dimension> dims;
-        for (int64_t i = 0; i < input_rank.get_length(); i++)
-        {
-            if (reduction_axes.count(i) == 0)
-            {
-                dims.push_back(input_shape[i]);
-            }
-        }
-
-        result_shape = PartialShape(dims);
-    }
+    const PartialShape& axes_shape = get_input_partial_shape(1);
+    const Rank axes_rank = axes_shape.rank();
+    NODE_VALIDATION_CHECK(this,
+                          axes_rank.compatible(0) || axes_rank.compatible(1),
+                          "Axes input must be a scalar or 1D input. Got: ",
+                          axes_shape);
 
+    PartialShape result_shape = infer_reduction_output_shape(false);
     set_input_is_relevant_to_shape(1);
-
     set_output_type(0, get_input_element_type(0), result_shape);
 }
diff --git a/ngraph/core/src/op/util/arithmetic_reductions_keep_dims.cpp b/ngraph/core/src/op/util/arithmetic_reductions_keep_dims.cpp
index dcb420c083893b..0a7a5eeb792b8a 100644
--- a/ngraph/core/src/op/util/arithmetic_reductions_keep_dims.cpp
+++ b/ngraph/core/src/op/util/arithmetic_reductions_keep_dims.cpp
@@ -32,61 +32,28 @@ bool ngraph::op::util::ArithmeticReductionKeepDims::visit_attributes(AttributeVi
 void op::util::ArithmeticReductionKeepDims::validate_and_infer_types()
 {
     NGRAPH_OP_SCOPE(v0_util_ArithmeticReductionKeepDims_validate_and_infer_types);
-    if (m_keep_dims)
-    {
-        auto input_shape = get_input_partial_shape(0);
-        auto input_rank = input_shape.rank();
-        PartialShape result_shape{PartialShape::dynamic()};
 
-        if (input_rank.is_static())
-            result_shape = PartialShape::dynamic(input_rank);
-
-        const auto& axes = get_constant_from_source(input_value(1));
-        if (input_rank.is_static() && axes)
-        {
-            AxisSet reduction_axes;
-            auto reduction_axes_val = axes->cast_vector<int64_t>();
-            for (auto axis : reduction_axes_val)
-            {
-                try
-                {
-                    axis = normalize_axis(this, axis, input_rank);
-                }
-                catch (const ngraph_error&)
-                {
-                    NODE_VALIDATION_CHECK(this,
-                                          false,
-                                          "Reduction axis (",
-                                          axis,
-                                          ") is out of bounds ",
-                                          "(argument shape: ",
-                                          input_shape,
-                                          ", reduction axes: ",
-                                          reduction_axes,
-                                          ")");
-                }
-                reduction_axes.insert(axis);
-            }
-
-            std::vector<Dimension> dims;
-            for (int64_t i = 0; i < input_rank.get_length(); i++)
-            {
-                if (reduction_axes.count(i) == 0)
-                {
-                    dims.push_back(input_shape[i]);
-                }
-                else
-                {
-                    dims.emplace_back(Dimension{1});
-                }
-            }
-            result_shape = PartialShape(dims);
-        }
-        set_input_is_relevant_to_shape(1);
-        set_output_type(0, get_input_element_type(0), result_shape);
-    }
-    else
-    {
-        ArithmeticReduction::validate_and_infer_types();
-    }
+    const element::Type& data_et = get_input_element_type(0);
+    const PartialShape& axes_shape = get_input_partial_shape(1);
+    const element::Type& axes_et = get_input_element_type(1);
+
+    NODE_VALIDATION_CHECK(this,
+                          data_et.is_real() || data_et.is_integral_number(),
+                          "Element type of data input must be numeric. Got: ",
+                          data_et);
+
+    NODE_VALIDATION_CHECK(this,
+                          axes_et.is_integral_number(),
+                          "Element type of axes input must be integer. Got: ",
+                          axes_et);
+
+    const Rank axes_rank = axes_shape.rank();
+    NODE_VALIDATION_CHECK(this,
+                          axes_rank.compatible(0) || axes_rank.compatible(1),
+                          "Axes input must be a scalar or 1D input. Got: ",
+                          axes_shape);
+
+    PartialShape result_shape = infer_reduction_output_shape(m_keep_dims);
+    set_input_is_relevant_to_shape(1);
+    set_output_type(0, data_et, result_shape);
 }
diff --git a/ngraph/core/src/op/util/logical_reduction.cpp b/ngraph/core/src/op/util/logical_reduction.cpp
index bd3895834d8be4..5d5ad8222ea3b6 100644
--- a/ngraph/core/src/op/util/logical_reduction.cpp
+++ b/ngraph/core/src/op/util/logical_reduction.cpp
@@ -15,17 +15,17 @@ NGRAPH_RTTI_DEFINITION(op::util::LogicalReduction, "LogicalReduction", 1);
 op::util::LogicalReduction::LogicalReduction() {}
 
 op::util::LogicalReduction::LogicalReduction(const Output<Node>& arg, const AxisSet& reduction_axes)
-    : Op({arg,
-          op::Constant::create(
-              element::i64, Shape{reduction_axes.size()}, reduction_axes.to_vector())
-              ->output(0)})
+    : ReductionBase(arg,
+                    op::Constant::create(
+                        element::i64, Shape{reduction_axes.size()}, reduction_axes.to_vector())
+                        ->output(0))
 {
     add_provenance_group_member(input_value(1).get_node_shared_ptr());
 }
 
 op::util::LogicalReduction::LogicalReduction(const Output<Node>& arg,
                                              const Output<Node>& reduction_axes)
-    : Op({arg, reduction_axes})
+    : ReductionBase(arg, reduction_axes)
 {
 }
 
@@ -54,59 +54,20 @@ void op::util::LogicalReduction::set_reduction_axes(const AxisSet& reduction_axe
 void op::util::LogicalReduction::validate_and_infer_types()
 {
     NGRAPH_OP_SCOPE(util_LogicalReduction_validate_and_infer_types);
-    auto input_shape = get_input_partial_shape(0);
-    auto input_rank = input_shape.rank();
 
-    PartialShape result_shape{PartialShape::dynamic()};
+    const element::Type& data_et = get_input_element_type(0);
+    const PartialShape& axes_shape = get_input_partial_shape(1);
 
-    set_input_is_relevant_to_shape(1);
+    NODE_VALIDATION_CHECK(
+        this, data_et.compatible(element::boolean), "Element type of data input must be boolean.");
 
+    const Rank axes_rank = axes_shape.rank();
     NODE_VALIDATION_CHECK(this,
-                          get_input_element_type(0).compatible(element::boolean),
-                          "Input element type must be boolean.");
-
-    set_output_type(0, element::boolean, result_shape);
-
-    if (input_rank.is_dynamic())
-        return;
-
-    if (const auto axes_const = get_constant_from_source(input_value(1)))
-    {
-        AxisSet reduction_axes;
-        auto reduction_axes_val = axes_const->cast_vector<int64_t>();
-        for (auto axis : reduction_axes_val)
-        {
-            try
-            {
-                axis = normalize_axis(this, axis, input_rank);
-            }
-            catch (const ngraph_error&)
-            {
-                NODE_VALIDATION_CHECK(this,
-                                      false,
-                                      "Reduction axis (",
-                                      axis,
-                                      ") is out of bounds ",
-                                      "(argument shape: ",
-                                      input_shape,
-                                      ", reduction axes: ",
-                                      reduction_axes,
-                                      ")");
-            }
-            reduction_axes.insert(axis);
-        }
+                          axes_rank.compatible(0) || axes_rank.compatible(1),
+                          "Axes input must be a scalar or 1D input. Got: ",
+                          axes_shape);
 
-        std::vector<Dimension> dims;
-        for (int64_t i = 0; i < input_rank.get_length(); i++)
-        {
-            if (reduction_axes.count(i) == 0)
-            {
-                dims.push_back(input_shape[i]);
-            }
-        }
-
-        result_shape = PartialShape(dims);
-    }
-
-    set_output_type(0, element::boolean, result_shape);
+    PartialShape result_shape = infer_reduction_output_shape(false);
+    set_input_is_relevant_to_shape(1);
+    set_output_type(0, data_et, result_shape);
 }
diff --git a/ngraph/core/src/op/util/logical_reduction_keep_dims.cpp b/ngraph/core/src/op/util/logical_reduction_keep_dims.cpp
index ea7188eaa844af..f2d366e0a97ce8 100644
--- a/ngraph/core/src/op/util/logical_reduction_keep_dims.cpp
+++ b/ngraph/core/src/op/util/logical_reduction_keep_dims.cpp
@@ -32,63 +32,26 @@ bool ngraph::op::util::LogicalReductionKeepDims::visit_attributes(AttributeVisit
 void op::util::LogicalReductionKeepDims::validate_and_infer_types()
 {
     NGRAPH_OP_SCOPE(v0_util_LogicalReductionKeepDims_validate_and_infer_types);
-    if (m_keep_dims)
-    {
-        const auto input_shape = get_input_partial_shape(0);
-        const auto input_rank = input_shape.rank();
-        PartialShape result_shape{PartialShape::dynamic(input_rank)};
 
-        set_input_is_relevant_to_shape(1);
-        set_output_type(0, get_input_element_type(0), result_shape);
+    const element::Type& data_et = get_input_element_type(0);
+    const PartialShape& axes_shape = get_input_partial_shape(1);
+    const element::Type& axes_et = get_input_element_type(1);
 
-        if (input_shape.is_dynamic())
-            return;
+    NODE_VALIDATION_CHECK(
+        this, data_et.compatible(element::boolean), "Element type of data input must be boolean.");
 
-        if (auto axes_const = get_constant_from_source(input_value(1)))
-        {
-            AxisSet reduction_axes;
-            auto reduction_axes_val = axes_const->cast_vector<int64_t>();
-            for (auto axis : reduction_axes_val)
-            {
-                try
-                {
-                    axis = normalize_axis(this, axis, input_rank);
-                }
-                catch (const ngraph_error&)
-                {
-                    NODE_VALIDATION_CHECK(this,
-                                          false,
-                                          "Reduction axis (",
-                                          axis,
-                                          ") is out of bounds ",
-                                          "(argument shape: ",
-                                          input_shape,
-                                          ", reduction axes: ",
-                                          reduction_axes,
-                                          ")");
-                }
-                reduction_axes.insert(axis);
-            }
+    NODE_VALIDATION_CHECK(this,
+                          axes_et.is_integral_number(),
+                          "Element type of axes input must be integer. Got: ",
+                          axes_et);
 
-            std::vector<Dimension> dims;
-            for (int64_t i = 0; i < input_rank.get_length(); i++)
-            {
-                if (reduction_axes.count(i) == 0)
-                {
-                    dims.push_back(input_shape[i]);
-                }
-                else
-                {
-                    dims.emplace_back(Dimension{1});
-                }
-            }
-            result_shape = PartialShape(dims);
-        }
+    const Rank axes_rank = axes_shape.rank();
+    NODE_VALIDATION_CHECK(this,
+                          axes_rank.compatible(0) || axes_rank.compatible(1),
+                          "Axes input must be a scalar or 1D input. Got: ",
+                          axes_shape);
 
-        set_output_type(0, get_input_element_type(0), result_shape);
-    }
-    else
-    {
-        LogicalReduction::validate_and_infer_types();
-    }
+    PartialShape result_shape = infer_reduction_output_shape(m_keep_dims);
+    set_input_is_relevant_to_shape(1);
+    set_output_type(0, data_et, result_shape);
 }
diff --git a/ngraph/core/src/op/util/reduction_base.cpp b/ngraph/core/src/op/util/reduction_base.cpp
new file mode 100644
index 00000000000000..bdfead647fb888
--- /dev/null
+++ b/ngraph/core/src/op/util/reduction_base.cpp
@@ -0,0 +1,74 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "ngraph/op/util/reduction_base.hpp"
+#include "itt.hpp"
+#include "ngraph/op/constant.hpp"
+#include "ngraph/validation_util.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+NGRAPH_RTTI_DEFINITION(op::util::ReductionBase, "ReductionBase", 0);
+
+op::util::ReductionBase::ReductionBase() {}
+
+op::util::ReductionBase::ReductionBase(const Output<Node>& arg, const Output<Node>& reduction_axes)
+    : Op({arg, reduction_axes})
+{
+}
+
+PartialShape op::util::ReductionBase::infer_reduction_output_shape(const bool keep_dims)
+{
+    const PartialShape& data_ps = get_input_partial_shape(0);
+    PartialShape result_ps{PartialShape::dynamic()};
+    Rank data_rank = data_ps.rank();
+
+    if (data_rank.is_static() && keep_dims)
+    {
+        result_ps = PartialShape::dynamic(data_rank);
+    }
+
+    const auto& axes = get_constant_from_source(input_value(1));
+    if (data_rank.is_static() && axes)
+    {
+        AxisSet reduction_axes;
+        auto reduction_axes_val = axes->cast_vector<int64_t>();
+        for (auto axis : reduction_axes_val)
+        {
+            try
+            {
+                axis = normalize_axis(this, axis, data_rank);
+            }
+            catch (const ngraph_error&)
+            {
+                NODE_VALIDATION_CHECK(this,
+                                      false,
+                                      "Reduction axis (",
+                                      axis,
+                                      ") is out of bounds ",
+                                      "(argument shape: ",
+                                      data_ps,
+                                      ", reduction axes: ",
+                                      reduction_axes,
+                                      ")");
+            }
+            reduction_axes.insert(axis);
+        }
+        std::vector<Dimension> dims;
+        for (int64_t i = 0; i < data_rank.get_length(); i++)
+        {
+            if (reduction_axes.count(i) == 0)
+            {
+                dims.push_back(data_ps[i]);
+            }
+            else if (keep_dims)
+            {
+                dims.emplace_back(Dimension{1});
+            }
+        }
+        result_ps = PartialShape(dims);
+    }
+    return result_ps;
+}
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index c629111ca147c1..a2a619007b619f 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -173,6 +173,13 @@ set(SRC
     type_prop/read_value.cpp
     type_prop/reduce_l1.cpp
     type_prop/reduce_l2.cpp
+    type_prop/reduce_logical_and.cpp
+    type_prop/reduce_logical_or.cpp
+    type_prop/reduce_max.cpp
+    type_prop/reduce_mean.cpp
+    type_prop/reduce_min.cpp
+    type_prop/reduce_prod.cpp
+    type_prop/reduce_sum.cpp
     type_prop/reorg_yolo.cpp
     type_prop/reshape.cpp
     type_prop/result.cpp
@@ -199,9 +206,6 @@ set(SRC
     type_prop/squared_difference.cpp
     type_prop/squeeze.cpp
     type_prop/swish.cpp
-    type_prop/reduce_mean.cpp
-    type_prop/reduce_prod.cpp
-    type_prop/reduce_sum.cpp
     type_prop/ti.cpp
     type_prop/tile.cpp
     type_prop/top_k.cpp
diff --git a/ngraph/test/type_prop/reduce_l1.cpp b/ngraph/test/type_prop/reduce_l1.cpp
index 07947615747da8..03f22e20f82bde 100644
--- a/ngraph/test/type_prop/reduce_l1.cpp
+++ b/ngraph/test/type_prop/reduce_l1.cpp
@@ -2,47 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gtest/gtest.h"
-#include "ngraph/ngraph.hpp"
-#include "util/type_prop.hpp"
+#include "reduce_ops.hpp"
 
-using namespace std;
-using namespace ngraph;
-
-TEST(type_prop, reduce_l1_v4_axis_out_of_range)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{1, 2, 3});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{2, 3});
-    try
-    {
-        auto reduce_sum = make_shared<op::v4::ReduceL1>(arg, axes);
-        // Should have thrown, so fail if it didn't
-        FAIL() << "Incorrect axes values exception not thrown";
-    }
-    catch (const NodeValidationFailure& error)
-    {
-        EXPECT_HAS_SUBSTRING(error.what(), std::string("Reduction axis ("));
-    }
-    catch (...)
-    {
-        FAIL() << "Deduced type check failed for unexpected reason";
-    }
-}
-
-TEST(type_prop, reduce_l1_v4_shape_if_keep_dims)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{3, 4, 5});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = true;
-    auto reduce_prod = make_shared<op::v4::ReduceL1>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(PartialShape{3, 1, 1}));
-}
-
-TEST(type_prop, reduce_l1_v4_shape_if_not_keep_dims)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{3, 4, 5});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = false;
-    auto reduce_prod = make_shared<op::v4::ReduceL1>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(PartialShape{3}));
-}
+using Type = ::testing::Types<op::v4::ReduceL1>;
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_l1, ReduceTest, Type);
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_l1_et, ReduceArithmeticTest, Type);
diff --git a/ngraph/test/type_prop/reduce_l2.cpp b/ngraph/test/type_prop/reduce_l2.cpp
index f2c2541b3852a4..64a9c02fb8751c 100644
--- a/ngraph/test/type_prop/reduce_l2.cpp
+++ b/ngraph/test/type_prop/reduce_l2.cpp
@@ -2,47 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gtest/gtest.h"
-#include "ngraph/ngraph.hpp"
-#include "util/type_prop.hpp"
+#include "reduce_ops.hpp"
 
-using namespace std;
-using namespace ngraph;
-
-TEST(type_prop, reduce_l2_v4_axis_out_of_range)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{1, 2, 3});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{2, 3});
-    try
-    {
-        auto reduce_sum = make_shared<op::v4::ReduceL2>(arg, axes);
-        // Should have thrown, so fail if it didn't
-        FAIL() << "Incorrect axes values exception not thrown";
-    }
-    catch (const NodeValidationFailure& error)
-    {
-        EXPECT_HAS_SUBSTRING(error.what(), std::string("Reduction axis ("));
-    }
-    catch (...)
-    {
-        FAIL() << "Deduced type check failed for unexpected reason";
-    }
-}
-
-TEST(type_prop, reduce_l2_v4_shape_if_keep_dims)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{3, 4, 5});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = true;
-    auto reduce_prod = make_shared<op::v4::ReduceL2>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(PartialShape{3, 1, 1}));
-}
-
-TEST(type_prop, reduce_l2_v4_shape_if_not_keep_dims)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{3, 4, 5});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = false;
-    auto reduce_prod = make_shared<op::v4::ReduceL2>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(PartialShape{3}));
-}
+using Type = ::testing::Types<op::v4::ReduceL2>;
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_l2, ReduceTest, Type);
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_l2_et, ReduceArithmeticTest, Type);
diff --git a/ngraph/test/type_prop/reduce_logical_and.cpp b/ngraph/test/type_prop/reduce_logical_and.cpp
new file mode 100644
index 00000000000000..f6e051bb50f12d
--- /dev/null
+++ b/ngraph/test/type_prop/reduce_logical_and.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "reduce_ops.hpp"
+
+using Type = ::testing::Types<op::v1::ReduceLogicalAnd>;
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_logical_and, ReduceTest, Type);
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_logical_and_et, ReduceLogicalTest, Type);
diff --git a/ngraph/test/type_prop/reduce_logical_or.cpp b/ngraph/test/type_prop/reduce_logical_or.cpp
new file mode 100644
index 00000000000000..b73d1789a20ebc
--- /dev/null
+++ b/ngraph/test/type_prop/reduce_logical_or.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "reduce_ops.hpp"
+
+using Type = ::testing::Types<op::v1::ReduceLogicalOr>;
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_logical_or, ReduceTest, Type);
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_logical_or_et, ReduceLogicalTest, Type);
diff --git a/ngraph/test/type_prop/reduce_max.cpp b/ngraph/test/type_prop/reduce_max.cpp
new file mode 100644
index 00000000000000..d907d1066af086
--- /dev/null
+++ b/ngraph/test/type_prop/reduce_max.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "reduce_ops.hpp"
+
+using Type = ::testing::Types<op::v1::ReduceMax>;
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_max, ReduceTest, Type);
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_max_et, ReduceArithmeticTest, Type);
diff --git a/ngraph/test/type_prop/reduce_mean.cpp b/ngraph/test/type_prop/reduce_mean.cpp
index 5f184fab97dd4c..607e8392f1e37b 100644
--- a/ngraph/test/type_prop/reduce_mean.cpp
+++ b/ngraph/test/type_prop/reduce_mean.cpp
@@ -2,71 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gtest/gtest.h"
-#include "ngraph/ngraph.hpp"
-#include "util/type_prop.hpp"
+#include "reduce_ops.hpp"
 
-using namespace std;
-using namespace ngraph;
-
-TEST(type_prop, reduce_mean_v1_axis_out_of_range)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{1, 2, 3});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{2, 3});
-    try
-    {
-        auto reduce_sum = make_shared<op::v1::ReduceMean>(arg, axes);
-
-        // Should have thrown, so fail if it didn't
-        FAIL() << "Incorrect axes values exception not thrown";
-    }
-    catch (const NodeValidationFailure& error)
-    {
-        EXPECT_HAS_SUBSTRING(error.what(), std::string("Reduction axis ("));
-    }
-    catch (...)
-    {
-        FAIL() << "Deduced type check failed for unexpected reason";
-    }
-}
-
-TEST(type_prop, reduce_mean_v1_shape_if_keep_dims)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{3, 4, 5});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = true;
-    auto reduce_prod = make_shared<op::v1::ReduceMean>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(PartialShape{3, 1, 1}));
-}
-
-TEST(type_prop, reduce_mean_v1_shape_if_not_keep_dims)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{3, 4, 5});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = false;
-    auto reduce_prod = make_shared<op::v1::ReduceMean>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(PartialShape{3}));
-}
-
-TEST(type_prop, reduce_mean_dynamic_shape)
-{
-    auto arg =
-        make_shared<op::Parameter>(element::f32, PartialShape{3, 4, 5, Dimension::dynamic()});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = true;
-    auto reduce_prod = make_shared<op::v1::ReduceMean>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(
-        PartialShape{3, 1, 1, Dimension::dynamic()}));
-}
-
-TEST(type_prop, reduce_mean_reduce_dynamic_shape)
-{
-    auto arg =
-        make_shared<op::Parameter>(element::f32, PartialShape{3, 4, 5, Dimension::dynamic()});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 3});
-    auto keep_dims = true;
-    auto reduce_prod = make_shared<op::v1::ReduceMean>(arg, axes, keep_dims);
-
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(
-        PartialShape{3, 1, 5, Dimension::dynamic()}));
-}
+using Type = ::testing::Types<op::v1::ReduceMean>;
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_mean, ReduceTest, Type);
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_mean_et, ReduceArithmeticTest, Type);
diff --git a/ngraph/test/type_prop/reduce_min.cpp b/ngraph/test/type_prop/reduce_min.cpp
new file mode 100644
index 00000000000000..d02f12f8a0a8f7
--- /dev/null
+++ b/ngraph/test/type_prop/reduce_min.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "reduce_ops.hpp"
+
+using Type = ::testing::Types<op::v1::ReduceMin>;
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_min, ReduceTest, Type);
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_min_et, ReduceArithmeticTest, Type);
diff --git a/ngraph/test/type_prop/reduce_ops.hpp b/ngraph/test/type_prop/reduce_ops.hpp
new file mode 100644
index 00000000000000..9aa7503e448bcd
--- /dev/null
+++ b/ngraph/test/type_prop/reduce_ops.hpp
@@ -0,0 +1,388 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/type_prop.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+struct ReduceParams
+{
+    PartialShape data_ps;
+    element::Type data_et;
+    Shape axes_ps;
+    std::vector<int64_t> axes;
+    element::Type axes_et;
+    bool keep_dims;
+};
+
+template<class T>
+static std::shared_ptr<Node> makeReduceOp(const ReduceParams& p, bool axes_as_param = false)
+{
+    auto in_data = make_shared<op::Parameter>(p.data_et, p.data_ps);
+    shared_ptr<Node> in_axes;
+    if (axes_as_param)
+    {
+        in_axes = make_shared<op::Parameter>(p.axes_et, p.axes_ps);
+    }
+    else
+    {
+        if (shape_size(p.axes_ps) != p.axes.size())
+        {
+            throw ngraph_error("Axes shape does not match with axes elements");
+        }
+        in_axes = make_shared<op::Constant>(p.axes_et, p.axes_ps, p.axes);
+    }
+    return make_shared<T>(in_data, in_axes, p.keep_dims);
+}
+
+template<class T>
+class ReduceTest : public testing::Test
+{
+};
+
+TYPED_TEST_CASE_P(ReduceTest);
+
+TYPED_TEST_P(ReduceTest, reduce_basic_shape_infer)
+{
+    PartialShape data_ps{3, 4, 5};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{2};
+    element::Type axes_et = element::i64;
+    std::vector<int64_t> axes{1, 2};
+
+    bool keep_dims = false;
+
+    PartialShape out_ps{3};
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    auto reduce_op = makeReduceOp<TypeParam>(params);
+    ASSERT_TRUE(reduce_op->get_output_partial_shape(0).same_scheme(out_ps));
+}
+
+TYPED_TEST_P(ReduceTest, reduce_basic_shape_infer_keep_dims)
+{
+    PartialShape data_ps{3, 4, 5};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{2};
+    element::Type axes_et = element::i64;
+    std::vector<int64_t> axes{1, 2};
+
+    bool keep_dims = true;
+
+    PartialShape out_ps{3, 1, 1};
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    auto reduce_op = makeReduceOp<TypeParam>(params);
+    ASSERT_TRUE(reduce_op->get_output_partial_shape(0).same_scheme(out_ps));
+}
+
+TYPED_TEST_P(ReduceTest, reduce_basic_shape_infer_scalar_axis)
+{
+    PartialShape data_ps{3, 4, 5};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{};
+    element::Type axes_et = element::i64;
+    std::vector<int64_t> axes{1};
+
+    bool keep_dims = false;
+
+    PartialShape out_ps{3, 5};
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    auto reduce_op = makeReduceOp<TypeParam>(params);
+    ASSERT_TRUE(reduce_op->get_output_partial_shape(0).same_scheme(out_ps));
+}
+
+TYPED_TEST_P(ReduceTest, reduce_basic_shape_infer_axes_as_param)
+{
+    PartialShape data_ps{3, 4, 5};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{2};
+    element::Type axes_et = element::i32;
+    std::vector<int64_t> axes;
+
+    bool keep_dims = false;
+
+    PartialShape out_ps{PartialShape::dynamic()};
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    bool axes_as_param = true;
+    auto reduce_op = makeReduceOp<TypeParam>(params, axes_as_param);
+    ASSERT_TRUE(reduce_op->get_output_partial_shape(0).same_scheme(out_ps));
+}
+
+TYPED_TEST_P(ReduceTest, reduce_dynamic_shape_reduced_axes_static)
+{
+    PartialShape data_ps{3, 4, 5, Dimension::dynamic()};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{2};
+    element::Type axes_et = element::i64;
+    std::vector<int64_t> axes{1, 2};
+
+    bool keep_dims = false;
+
+    PartialShape out_ps{3, Dimension::dynamic()};
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    auto reduce_op = makeReduceOp<TypeParam>(params);
+    ASSERT_TRUE(reduce_op->get_output_partial_shape(0).same_scheme(out_ps));
+}
+
+TYPED_TEST_P(ReduceTest, reduce_dynamic_shape_reduced_axes_static_keep_dims)
+{
+    PartialShape data_ps{3, 4, 5, Dimension::dynamic()};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{2};
+    element::Type axes_et = element::i64;
+    std::vector<int64_t> axes{1, 2};
+
+    bool keep_dims = true;
+
+    PartialShape out_ps{3, 1, 1, Dimension::dynamic()};
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    auto reduce_op = makeReduceOp<TypeParam>(params);
+    ASSERT_TRUE(reduce_op->get_output_partial_shape(0).same_scheme(out_ps));
+}
+
+TYPED_TEST_P(ReduceTest, reduce_dynamic_shape_reduced_axes_not_static)
+{
+    PartialShape data_ps{Dimension::dynamic(), 4, 5, Dimension::dynamic()};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{2};
+    element::Type axes_et = element::i64;
+    std::vector<int64_t> axes{2, 3};
+
+    bool keep_dims = false;
+
+    PartialShape out_ps{Dimension::dynamic(), 4};
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    auto reduce_op = makeReduceOp<TypeParam>(params);
+    ASSERT_TRUE(reduce_op->get_output_partial_shape(0).same_scheme(out_ps));
+}
+
+TYPED_TEST_P(ReduceTest, reduce_dynamic_shape_reduced_axes_not_static_keep_dims)
+{
+    PartialShape data_ps{Dimension::dynamic(), 4, 5, Dimension::dynamic()};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{2};
+    element::Type axes_et = element::i64;
+    std::vector<int64_t> axes{2, 3};
+
+    bool keep_dims = true;
+
+    PartialShape out_ps{Dimension::dynamic(), 4, 1, 1};
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    auto reduce_op = makeReduceOp<TypeParam>(params);
+    ASSERT_TRUE(reduce_op->get_output_partial_shape(0).same_scheme(out_ps));
+}
+
+TYPED_TEST_P(ReduceTest, reduce_dynamic_shape_data)
+{
+    PartialShape data_ps{PartialShape::dynamic()};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{2};
+    element::Type axes_et = element::i64;
+    std::vector<int64_t> axes{1, 2};
+
+    bool keep_dims = false;
+
+    PartialShape out_ps{PartialShape::dynamic()};
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    auto reduce_op = makeReduceOp<TypeParam>(params);
+    ASSERT_TRUE(reduce_op->get_output_partial_shape(0).same_scheme(out_ps));
+}
+
+TYPED_TEST_P(ReduceTest, reduce_invalid_axis_out_of_range)
+{
+    PartialShape data_ps{1, 2, 3};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{2};
+    element::Type axes_et = element::i64;
+    std::vector<int64_t> axes{2, 3};
+
+    bool keep_dims = false;
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    try
+    {
+        auto reduce_op = makeReduceOp<TypeParam>(params);
+        FAIL() << "Invalid axes values not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), "Reduction axis (");
+    }
+    catch (...)
+    {
+        FAIL() << "Axes input values validation check failed for unexpected reason";
+    }
+}
+
+TYPED_TEST_P(ReduceTest, reduce_invalid_axes_shape)
+{
+    PartialShape data_ps{1, 2, 3};
+    element::Type data_et = element::dynamic;
+
+    Shape axes_ps{2, 1};
+    element::Type axes_et = element::i64;
+    std::vector<int64_t> axes{0, 1};
+
+    bool keep_dims = true;
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    try
+    {
+        auto reduce_op = makeReduceOp<TypeParam>(params);
+        FAIL() << "Invalid shape of axes input not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), "Axes input must be a scalar or 1D input.");
+    }
+    catch (...)
+    {
+        FAIL() << "Axes input shape validation check failed for unexpected reason";
+    }
+}
+
+TYPED_TEST_P(ReduceTest, reduce_invalid_axes_et)
+{
+    element::Type data_et = element::dynamic;
+    PartialShape data_ps{1, 2, 3};
+
+    element::Type axes_et = element::f32;
+    Shape axes_ps{2};
+    std::vector<int64_t> axes{0, 1};
+
+    bool keep_dims = true;
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    try
+    {
+        auto reduce_op = makeReduceOp<TypeParam>(params);
+        FAIL() << "Invalid element type of axes input not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), "Element type of axes input must be integer.");
+    }
+    catch (...)
+    {
+        FAIL() << "Axes input element type validation check failed for unexpected reason";
+    }
+}
+
+REGISTER_TYPED_TEST_CASE_P(
+    ReduceTest,
+    reduce_basic_shape_infer,
+    reduce_basic_shape_infer_keep_dims,
+    reduce_basic_shape_infer_scalar_axis,
+    reduce_basic_shape_infer_axes_as_param,
+    reduce_dynamic_shape_data,
+    reduce_dynamic_shape_reduced_axes_static,
+    reduce_dynamic_shape_reduced_axes_static_keep_dims,
+    reduce_dynamic_shape_reduced_axes_not_static,
+    reduce_dynamic_shape_reduced_axes_not_static_keep_dims,
+    reduce_invalid_axis_out_of_range,
+    reduce_invalid_axes_shape,
+    reduce_invalid_axes_et);
+
+template<class T>
+class ReduceArithmeticTest : public testing::Test
+{
+};
+
+TYPED_TEST_CASE_P(ReduceArithmeticTest);
+
+TYPED_TEST_P(ReduceArithmeticTest, reduce_arithmetic_invalid_data_et)
+{
+    element::Type data_et = element::boolean;
+    PartialShape data_ps{1, 2, 3};
+
+    element::Type axes_et = element::i32;
+    Shape axes_ps{2};
+    std::vector<int64_t> axes{0, 1};
+
+    bool keep_dims = true;
+
+    const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+    try
+    {
+        auto reduce_op = makeReduceOp<TypeParam>(params);
+        FAIL() << "Invalid element type of data input not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), "Element type of data input must be numeric.");
+    }
+    catch (...)
+    {
+        FAIL() << "Data input element type validation check failed for unexpected reason";
+    }
+}
+
+REGISTER_TYPED_TEST_CASE_P(
+    ReduceArithmeticTest,
+    reduce_arithmetic_invalid_data_et);
+
+template<class T>
+class ReduceLogicalTest : public testing::Test
+{
+};
+
+TYPED_TEST_CASE_P(ReduceLogicalTest);
+
+TYPED_TEST_P(ReduceLogicalTest, reduce_logical_invalid_data_et)
+{
+    std::vector<element::Type> element_types{
+        element::f32,
+        element::i32,
+        element::u32};
+    PartialShape data_ps{1, 2, 3};
+
+    element::Type axes_et = element::i32;
+    Shape axes_ps{2};
+    std::vector<int64_t> axes{0, 1};
+
+    bool keep_dims = true;
+
+    for (const auto& data_et : element_types)
+    {
+        const ReduceParams params{data_ps, data_et, axes_ps, axes, axes_et, keep_dims};
+        try
+        {
+            auto reduce_op = makeReduceOp<TypeParam>(params);
+            FAIL() << "Invalid element type of data input not detected";
+        }
+        catch (const NodeValidationFailure& error)
+        {
+            EXPECT_HAS_SUBSTRING(error.what(), "Element type of data input must be boolean.");
+        }
+        catch (...)
+        {
+            FAIL() << "Data input element type validation check failed for unexpected reason";
+        }
+    }  
+}
+
+REGISTER_TYPED_TEST_CASE_P(
+    ReduceLogicalTest,
+    reduce_logical_invalid_data_et);
diff --git a/ngraph/test/type_prop/reduce_prod.cpp b/ngraph/test/type_prop/reduce_prod.cpp
index 243f158d7e25a2..3a6ea0a9e89af2 100644
--- a/ngraph/test/type_prop/reduce_prod.cpp
+++ b/ngraph/test/type_prop/reduce_prod.cpp
@@ -2,48 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gtest/gtest.h"
-#include "ngraph/ngraph.hpp"
-#include "util/type_prop.hpp"
+#include "reduce_ops.hpp"
 
-using namespace std;
-using namespace ngraph;
-
-TEST(type_prop, reduce_prod_v1_axis_out_of_range)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{1, 2, 3});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{2, 3});
-    try
-    {
-        auto reduce_prod = make_shared<op::v1::ReduceProd>(arg, axes);
-
-        // Should have thrown, so fail if it didn't
-        FAIL() << "Incorrect axes values exception not thrown";
-    }
-    catch (const NodeValidationFailure& error)
-    {
-        EXPECT_HAS_SUBSTRING(error.what(), std::string("Reduction axis ("));
-    }
-    catch (...)
-    {
-        FAIL() << "Deduced type check failed for unexpected reason";
-    }
-}
-
-TEST(type_prop, reduce_prod_v1_shape_if_keep_dims)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{3, 4, 5});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = true;
-    auto reduce_prod = make_shared<op::v1::ReduceProd>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(PartialShape{3, 1, 1}));
-}
-
-TEST(type_prop, reduce_prod_v1_shape_if_not_keep_dims)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{3, 4, 5});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = false;
-    auto reduce_prod = make_shared<op::v1::ReduceProd>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(PartialShape{3}));
-}
+using Type = ::testing::Types<op::v1::ReduceProd>;
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_prod, ReduceTest, Type);
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_prod_et, ReduceArithmeticTest, Type);
diff --git a/ngraph/test/type_prop/reduce_sum.cpp b/ngraph/test/type_prop/reduce_sum.cpp
index 0a1f0422a3ad64..7bfce1dfc708f0 100644
--- a/ngraph/test/type_prop/reduce_sum.cpp
+++ b/ngraph/test/type_prop/reduce_sum.cpp
@@ -2,48 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gtest/gtest.h"
-#include "ngraph/ngraph.hpp"
-#include "util/type_prop.hpp"
+#include "reduce_ops.hpp"
 
-using namespace std;
-using namespace ngraph;
-
-TEST(type_prop, reduce_sum_v1_axis_out_of_range)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{1, 2, 3});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{2, 3});
-    try
-    {
-        auto reduce_sum = make_shared<op::v1::ReduceSum>(arg, axes);
-
-        // Should have thrown, so fail if it didn't
-        FAIL() << "Incorrect axes values exception not thrown";
-    }
-    catch (const NodeValidationFailure& error)
-    {
-        EXPECT_HAS_SUBSTRING(error.what(), std::string("Reduction axis ("));
-    }
-    catch (...)
-    {
-        FAIL() << "Deduced type check failed for unexpected reason";
-    }
-}
-
-TEST(type_prop, reduce_sum_v1_shape_if_keep_dims)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{3, 4, 5});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = true;
-    auto reduce_prod = make_shared<op::v1::ReduceSum>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(PartialShape{3, 1, 1}));
-}
-
-TEST(type_prop, reduce_sum_v1_shape_if_not_keep_dims)
-{
-    auto arg = make_shared<op::Parameter>(element::f32, Shape{3, 4, 5});
-    auto axes = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 2});
-    auto keep_dims = false;
-    auto reduce_prod = make_shared<op::v1::ReduceSum>(arg, axes, keep_dims);
-    ASSERT_TRUE(reduce_prod->get_output_partial_shape(0).compatible(PartialShape{3}));
-}
+using Type = ::testing::Types<op::v1::ReduceSum>;
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_sum, ReduceTest, Type);
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_reduce_sum_et, ReduceArithmeticTest, Type);
diff --git a/ngraph/test/visitors/op/reduce_ops.hpp b/ngraph/test/visitors/op/reduce_ops.hpp
index dfbf1b6eba8757..4c5a165b2bcac5 100644
--- a/ngraph/test/visitors/op/reduce_ops.hpp
+++ b/ngraph/test/visitors/op/reduce_ops.hpp
@@ -24,7 +24,7 @@ TYPED_TEST_CASE_P(ReduceOpsAttrTest);
 TYPED_TEST_P(ReduceOpsAttrTest, reduce_ops)
 {
     Shape in_shape{3, 4, 5};
-    element::Type in_et = element::f32;
+    element::Type in_et = element::dynamic;
 
     Shape axes_shape{2};
     element::Type axes_et = element::i64;

From 8830dcabf3af456fa213594151abce6b118359b5 Mon Sep 17 00:00:00 2001
From: Alexander Zhogov <alexander.zhogov@intel.com>
Date: Fri, 11 Jun 2021 11:18:25 +0300
Subject: [PATCH 066/178] GitHub CI: Check committer email

---
 .github/org_control/check_pr.py | 18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)

diff --git a/.github/org_control/check_pr.py b/.github/org_control/check_pr.py
index 7bb8f89dd602f4..0e23251e1571a4 100644
--- a/.github/org_control/check_pr.py
+++ b/.github/org_control/check_pr.py
@@ -147,21 +147,29 @@ def get_wrong_commits(pull):
         # import pprint; pprint.pprint(commit.raw_data)
         print("Commit SHA:", commit.sha)
         # Use raw data because commit author can be non GitHub user
-        commit_email = (commit.raw_data["commit"]["author"]["email"] or "").lower()
-        print("    Commit email:", commit_email)
+        commit_author_email = (commit.raw_data["commit"]["author"]["email"] or "").lower()
+        commit_committer_email = (commit.raw_data["commit"]["committer"]["email"] or "").lower()
+        print("    Commit author email:", commit_author_email)
+        print("    Commit committer email:", commit_committer_email)
         if not github_api.is_valid_user(commit.author):
             print(
-                "    ERROR: User with the commit email is absent in GitHub:",
+                "    ERROR: User with the commit author email is absent in GitHub:",
                 commit.raw_data["commit"]["author"]["name"],
             )
             wrong_commits.add(commit.sha)
+        if not github_api.is_valid_user(commit.committer):
+            print(
+                "    ERROR: User with the commit committer email is absent in GitHub:",
+                commit.raw_data["commit"]["committer"]["name"],
+            )
+            wrong_commits.add(commit.sha)
         if not commit.raw_data["commit"]["verification"]["verified"]:
             print(
                 "    WARNING: The commit is not verified. Reason:",
                 commit.raw_data["commit"]["verification"]["reason"],
             )
-            if pr_author_email != commit_email:
-                print("    WARNING: Commit email and GitHub PR author public email are differnt")
+            if pr_author_email != commit_author_email or pr_author_email != commit_committer_email:
+                print("    WARNING: Commit emails and GitHub PR author public email are differnt")
     return wrong_commits
 
 

From 167a5519b170cfdb98f2d1e4ab7b897c4c72aeaf Mon Sep 17 00:00:00 2001
From: Tomasz Socha <tomasz.socha@intel.com>
Date: Fri, 11 Jun 2021 11:27:58 +0200
Subject: [PATCH 067/178] =?UTF-8?q?[ONNX=20Importer]=20=F0=9F=95=B8?=
 =?UTF-8?q?=EF=B8=8F=20Make=20global=20pools=20more=20dynamic=20(#6122)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .../src/op/global_average_pool.cpp            | 34 +++++++++----------
 .../onnx_import/src/op/global_max_pool.cpp    | 34 +++++++++----------
 2 files changed, 32 insertions(+), 36 deletions(-)

diff --git a/ngraph/frontend/onnx_import/src/op/global_average_pool.cpp b/ngraph/frontend/onnx_import/src/op/global_average_pool.cpp
index 2d99e69da5dad5..d0a7ea4020f939 100644
--- a/ngraph/frontend/onnx_import/src/op/global_average_pool.cpp
+++ b/ngraph/frontend/onnx_import/src/op/global_average_pool.cpp
@@ -19,19 +19,6 @@ namespace ngraph
             {
                 OutputVector global_average_pool(const Node& node)
                 {
-                    auto data = node.get_ng_inputs()[0];
-                    auto data_rank = data.get_partial_shape().rank();
-
-                    NGRAPH_CHECK(data_rank.is_static(),
-                                 "The input data tensor's rank has to be known (static)");
-
-                    auto data_rank_value = data_rank.get_length();
-
-                    NGRAPH_CHECK(data_rank_value > 2,
-                                 "The input data tensor's rank has to be greater than 2."
-                                 "Provided data rank is: ",
-                                 data_rank_value);
-
                     // Generate axes for reduce operation which contain all spatial dims indexes.
                     // Examples:
                     // Input shape: [N, C, H, W]
@@ -41,11 +28,22 @@ namespace ngraph
                     // Input shape: [N, C, H, W, D]
                     // Input spatial dimensions are H, W and D
                     // Expected spatial dims indexes: [2, 3, 4]
-                    size_t data_spatial_rank = data_rank_value - 2;
-                    auto reduce_axes_vector = std::vector<std::int64_t>(data_spatial_rank);
-                    std::iota(reduce_axes_vector.begin(), reduce_axes_vector.end(), 2);
-                    auto reduce_axes = default_opset::Constant::create(
-                        element::i64, Shape{data_spatial_rank}, reduce_axes_vector);
+                    auto data = node.get_ng_inputs()[0];
+
+                    const auto zero_node =
+                        default_opset::Constant::create(element::i64, Shape{}, {0});
+                    const auto one_node =
+                        default_opset::Constant::create(element::i64, Shape{}, {1});
+                    const auto two_node =
+                        default_opset::Constant::create(element::i64, Shape{}, {2});
+
+                    const auto data_shape = std::make_shared<default_opset::ShapeOf>(data);
+                    const auto data_rank = std::make_shared<default_opset::ShapeOf>(data_shape);
+                    const auto data_rank_as_scalar =
+                        std::make_shared<default_opset::Squeeze>(data_rank);
+
+                    const auto reduce_axes = std::make_shared<default_opset::Range>(
+                        two_node, data_rank_as_scalar, one_node, element::i64);
 
                     return {std::make_shared<default_opset::ReduceMean>(data, reduce_axes, true)};
                 }
diff --git a/ngraph/frontend/onnx_import/src/op/global_max_pool.cpp b/ngraph/frontend/onnx_import/src/op/global_max_pool.cpp
index 866b7ac0201bd3..49fec836f21e53 100644
--- a/ngraph/frontend/onnx_import/src/op/global_max_pool.cpp
+++ b/ngraph/frontend/onnx_import/src/op/global_max_pool.cpp
@@ -19,19 +19,6 @@ namespace ngraph
             {
                 OutputVector global_max_pool(const Node& node)
                 {
-                    auto data = node.get_ng_inputs()[0];
-                    auto data_rank = data.get_partial_shape().rank();
-
-                    NGRAPH_CHECK(data_rank.is_static(),
-                                 "The input data tensor's rank has to be known (static)");
-
-                    auto data_rank_value = data_rank.get_length();
-
-                    NGRAPH_CHECK(data_rank_value > 2,
-                                 "The input data tensor's rank has to be greater than 2."
-                                 "Provided data rank is: ",
-                                 data_rank_value);
-
                     // Generate axes for reduce operation which contain all spatial dims indexes.
                     // Examples:
                     // Input shape: [N, C, H, W]
@@ -41,11 +28,22 @@ namespace ngraph
                     // Input shape: [N, C, H, W, D]
                     // Input spatial dimensions are H, W and D
                     // Expected spatial dims indexes: [2, 3, 4]
-                    size_t data_spatial_rank = data_rank_value - 2;
-                    auto reduce_axes_vector = std::vector<std::int64_t>(data_spatial_rank);
-                    std::iota(reduce_axes_vector.begin(), reduce_axes_vector.end(), 2);
-                    auto reduce_axes = default_opset::Constant::create(
-                        element::i64, Shape{data_spatial_rank}, reduce_axes_vector);
+                    auto data = node.get_ng_inputs()[0];
+
+                    const auto zero_node =
+                        default_opset::Constant::create(element::i64, Shape{}, {0});
+                    const auto one_node =
+                        default_opset::Constant::create(element::i64, Shape{}, {1});
+                    const auto two_node =
+                        default_opset::Constant::create(element::i64, Shape{}, {2});
+
+                    const auto data_shape = std::make_shared<default_opset::ShapeOf>(data);
+                    const auto data_rank = std::make_shared<default_opset::ShapeOf>(data_shape);
+                    const auto data_rank_as_scalar =
+                        std::make_shared<default_opset::Squeeze>(data_rank);
+
+                    const auto reduce_axes = std::make_shared<default_opset::Range>(
+                        two_node, data_rank_as_scalar, one_node, element::i64);
 
                     return {std::make_shared<default_opset::ReduceMax>(data, reduce_axes, true)};
                 }

From 089f76cc7f599a719adef25bd6651fee25ea255a Mon Sep 17 00:00:00 2001
From: Victor Kuznetsov <32412802+just-sparta@users.noreply.github.com>
Date: Fri, 11 Jun 2021 12:57:01 +0300
Subject: [PATCH 068/178] Add `branch` field to submit to DB in
 MemCheck/TimeTests (#6073)

---
 tests/stress_tests/scripts/memcheck_upload.py | 1 +
 tests/time_tests/test_runner/conftest.py      | 3 ++-
 tests/time_tests/test_runner/utils.py         | 3 ++-
 3 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/tests/stress_tests/scripts/memcheck_upload.py b/tests/stress_tests/scripts/memcheck_upload.py
index 663a513c8566bb..a3eeb3906b4968 100644
--- a/tests/stress_tests/scripts/memcheck_upload.py
+++ b/tests/stress_tests/scripts/memcheck_upload.py
@@ -67,6 +67,7 @@ def metadata_from_manifest(manifest):
         'commit_sha': repo_trigger['revision'],
         'commit_date': repo_trigger['commit_time'],
         'repo_url': repo_trigger['url'],
+        'branch': repo_trigger['branch'],
         'target_branch': repo_trigger['target_branch'],
         'event_type': manifest['components'][PRODUCT_NAME]['build_event'].lower(),
         f'{PRODUCT_NAME}_version': manifest['components'][PRODUCT_NAME]['version'],
diff --git a/tests/time_tests/test_runner/conftest.py b/tests/time_tests/test_runner/conftest.py
index d0cb928ce07411..f60be60d963238 100644
--- a/tests/time_tests/test_runner/conftest.py
+++ b/tests/time_tests/test_runner/conftest.py
@@ -325,10 +325,11 @@ def manifest_metadata(request):
                 "repo_url": {"type": "string"},
                 "commit_sha": {"type": "string"},
                 "commit_date": {"type": "string"},
+                "branch": {"type": "string"},
                 "target_branch": {"type": "string"},
                 "version": {"type": "string"}
             },
-            "required": ["product_type", "repo_url", "commit_sha", "commit_date", "target_branch", "version"],
+            "required": ["product_type", "repo_url", "commit_sha", "commit_date", "branch", "target_branch", "version"],
             "additionalProperties": false
         }
         """
diff --git a/tests/time_tests/test_runner/utils.py b/tests/time_tests/test_runner/utils.py
index 990af4c9b6b70e..c68cc4fd7c2295 100644
--- a/tests/time_tests/test_runner/utils.py
+++ b/tests/time_tests/test_runner/utils.py
@@ -57,7 +57,8 @@ def metadata_from_manifest(manifest: Path):
         'commit_sha': repo_trigger['revision'],
         'commit_date': repo_trigger['commit_time'],
         'repo_url': repo_trigger['url'],
-        'target_branch': repo_trigger['branch'],
+        'branch': repo_trigger['branch'],
+        'target_branch': repo_trigger['target_branch'],
         'version': manifest['components'][PRODUCT_NAME]['version']
     }
 

From 458435ad753123ccaf33563aae977c3b29378814 Mon Sep 17 00:00:00 2001
From: Bartek Szmelczynski <bartosz.szmelczynski@intel.com>
Date: Fri, 11 Jun 2021 17:01:52 +0200
Subject: [PATCH 069/178] Revise selu (#5513)

* remove FusedOp from selu operator

* add type_prop tests for selu op

* add vistors test for selu op

* update CMakeLists and fix wrong path to visitors test for selu

* fix style

* add backend tests for selu op

* refactor validate_and_infer_types function for selu op

* refactor type_prop tests for selu to catch particular errors

Co-authored-by: jdanieck <jozef.daniecki@intel.com>
---
 ngraph/core/include/ngraph/op/selu.hpp |  19 +--
 ngraph/core/src/op/selu.cpp            |  68 ++++------
 ngraph/test/CMakeLists.txt             |   3 +
 ngraph/test/backend/selu.in.cpp        |  99 ++++++++++++++
 ngraph/test/type_prop/selu.cpp         | 171 +++++++++++++++++++++++++
 ngraph/test/visitors/op/selu.cpp       |  30 +++++
 6 files changed, 338 insertions(+), 52 deletions(-)
 create mode 100644 ngraph/test/backend/selu.in.cpp
 create mode 100644 ngraph/test/type_prop/selu.cpp
 create mode 100644 ngraph/test/visitors/op/selu.cpp

diff --git a/ngraph/core/include/ngraph/op/selu.hpp b/ngraph/core/include/ngraph/op/selu.hpp
index 9d71f71936b1fb..9bc1c71acd6937 100644
--- a/ngraph/core/include/ngraph/op/selu.hpp
+++ b/ngraph/core/include/ngraph/op/selu.hpp
@@ -6,10 +6,6 @@
 
 #include "ngraph/node.hpp"
 #include "ngraph/op/op.hpp"
-#include "ngraph/op/util/fused_op.hpp"
-
-NGRAPH_SUPPRESS_DEPRECATED_START
-
 namespace ngraph
 {
     namespace op
@@ -17,12 +13,12 @@ namespace ngraph
         namespace v0
         {
             /// \brief Performs a SELU activation function on all elements of the input node
-            class NGRAPH_API Selu : public ngraph::op::util::FusedOp
+            class NGRAPH_API Selu : public ngraph::op::Op
             {
             public:
-                static constexpr NodeTypeInfo type_info{"Selu", 0};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
-                Selu();
+                NGRAPH_RTTI_DECLARATION;
+
+                Selu() = default;
                 /// \brief Constructs a Selu node.
                 ///
                 /// \param data - Node producing the input tensor
@@ -31,9 +27,10 @@ namespace ngraph
                 Selu(const Output<Node>& data,
                      const Output<Node>& alpha,
                      const Output<Node>& lambda);
-                virtual void pre_validate_and_infer_types() override;
+
+                void validate_and_infer_types() override;
+
                 bool visit_attributes(AttributeVisitor& visitor) override;
-                virtual OutputVector decompose_op() const override;
 
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
@@ -42,5 +39,3 @@ namespace ngraph
         using v0::Selu;
     } // namespace op
 } // namespace ngraph
-
-NGRAPH_SUPPRESS_DEPRECATED_END
diff --git a/ngraph/core/src/op/selu.cpp b/ngraph/core/src/op/selu.cpp
index 680b67fc012462..fd9aae0182a06e 100644
--- a/ngraph/core/src/op/selu.cpp
+++ b/ngraph/core/src/op/selu.cpp
@@ -5,65 +5,53 @@
 #include "ngraph/op/selu.hpp"
 #include "itt.hpp"
 
-#include "ngraph/op/add.hpp"
-#include "ngraph/op/constant.hpp"
-#include "ngraph/op/exp.hpp"
-#include "ngraph/op/maximum.hpp"
-#include "ngraph/op/minimum.hpp"
-#include "ngraph/op/multiply.hpp"
-#include "ngraph/op/subtract.hpp"
-
 using namespace std;
 using namespace ngraph;
 
-NGRAPH_SUPPRESS_DEPRECATED_START
-
-constexpr NodeTypeInfo op::v0::Selu::type_info;
-
-op::v0::Selu::Selu()
-    : FusedOp()
-{
-}
+NGRAPH_RTTI_DEFINITION(op::v0::Selu, "Selu", 0);
 
 op::v0::Selu::Selu(const Output<Node>& data, const Output<Node>& alpha, const Output<Node>& lambda)
-    : FusedOp({data, alpha, lambda})
+    : Op({data, alpha, lambda})
 {
     constructor_validate_and_infer_types();
 }
 
-void ngraph::op::v0::Selu::pre_validate_and_infer_types()
+void op::v0::Selu::validate_and_infer_types()
 {
-    set_output_type(0, get_input_element_type(0), get_input_partial_shape(0));
+    NGRAPH_OP_SCOPE(v0_Selu_validate_and_infer_types);
+    auto data_et = get_input_element_type(0);
+    auto alpha_et = get_input_element_type(1);
+    auto lambda_et = get_input_element_type(2);
+    auto result_et = element::dynamic;
+
+    NODE_VALIDATION_CHECK(this,
+                          element::Type::merge(result_et, result_et, data_et) &&
+                              element::Type::merge(result_et, result_et, alpha_et) &&
+                              element::Type::merge(result_et, result_et, lambda_et),
+                          "Input element types do not match : ",
+                          data_et,
+                          " and ",
+                          alpha_et,
+                          " and ",
+                          lambda_et);
+
+    NODE_VALIDATION_CHECK(this,
+                          result_et.is_dynamic() || result_et.is_real(),
+                          "Input element types must be floating-point. Got: ",
+                          result_et);
+
+    set_output_type(0, result_et, get_input_partial_shape(0));
 }
 
-bool ngraph::op::v0::Selu::visit_attributes(AttributeVisitor& visitor)
+bool op::v0::Selu::visit_attributes(AttributeVisitor& visitor)
 {
     NGRAPH_OP_SCOPE(v0_Selu_visit_attributes);
     return true;
 }
 
-OutputVector op::v0::Selu::decompose_op() const
-{
-    const auto data = input_value(0);
-    const auto alpha = input_value(1);
-    const auto lambda = input_value(2);
-    const auto zero_node = op::Constant::create(data.get_element_type(), Shape{1}, {0});
-
-    // lambda * ((max(data, 0) + (alpha * exp(min(data, 0)) - alpha))
-    return {std::make_shared<op::v1::Multiply>(
-        lambda,
-        std::make_shared<op::v1::Add>(
-            std::make_shared<op::v1::Maximum>(data, zero_node),
-            std::make_shared<op::v1::Subtract>(
-                std::make_shared<op::v1::Multiply>(
-                    alpha,
-                    std::make_shared<op::Exp>(std::make_shared<op::v1::Minimum>(data, zero_node))),
-                alpha)))};
-}
-
 shared_ptr<Node> op::v0::Selu::clone_with_new_inputs(const OutputVector& new_args) const
 {
     NGRAPH_OP_SCOPE(v0_Selu_clone_with_new_inputs);
     check_new_args_count(this, new_args);
-    return make_shared<v0::Selu>(new_args.at(0), new_args.at(1), new_args.at(2));
+    return make_shared<op::v0::Selu>(new_args.at(0), new_args.at(1), new_args.at(2));
 }
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index a2a619007b619f..d04486563866ef 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -196,6 +196,7 @@ set(SRC
     type_prop/scatter_nd_update.cpp
     type_prop/scatter_update.cpp
     type_prop/select.cpp
+    type_prop/selu.cpp
     type_prop/shape_of.cpp
     type_prop/shuffle_channels.cpp
     type_prop/softmax.cpp
@@ -270,6 +271,7 @@ set(SRC
     visitors/op/reverse_sequence.cpp
     visitors/op/rnn_cell.cpp
     visitors/op/roi_pooling.cpp
+    visitors/op/selu.cpp
     visitors/op/shuffle_channels.cpp
     visitors/op/softmax.cpp
     visitors/op/space_to_depth.cpp
@@ -435,6 +437,7 @@ set(MULTI_TEST_SRC
     backend/round.in.cpp
     backend/scatter_nd_update.in.cpp
     backend/select.in.cpp
+    backend/selu.in.cpp
     backend/shape_of.in.cpp
     backend/sigmoid.in.cpp
     backend/sign.in.cpp
diff --git a/ngraph/test/backend/selu.in.cpp b/ngraph/test/backend/selu.in.cpp
new file mode 100644
index 00000000000000..2524ee045fda41
--- /dev/null
+++ b/ngraph/test/backend/selu.in.cpp
@@ -0,0 +1,99 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/engine/test_engines.hpp"
+#include "util/test_case.hpp"
+#include "util/test_control.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+using TestEngine = test::ENGINE_CLASS_NAME(${BACKEND_NAME});
+
+NGRAPH_TEST(${BACKEND_NAME}, selu_2Dfprop)
+{
+    Shape rt_shape{2};
+    Shape c_shape{1};
+    element::Type et = element::f32;
+
+    auto input = make_shared<op::Parameter>(et, rt_shape);
+    auto alpha = op::Constant::create(et, c_shape, {1.67326324});
+    auto lambda = op::Constant::create(et, c_shape, {1.05070098});
+    auto selu = make_shared<op::v0::Selu>(input, alpha, lambda);
+    auto f = make_shared<Function>(selu, ParameterVector{input});
+
+    vector<float> input_data{-1, 3};
+    vector<float> expected_out{-1.1113307, 3.152103};
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_input<float>(rt_shape, input_data);
+    test_case.add_expected_output(rt_shape, expected_out);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, selu_4Dfprop)
+{
+    Shape in_shape{4};
+    Shape c_shape{1};
+    element::Type et = element::f32;
+
+    auto input = make_shared<op::Parameter>(et, in_shape);
+    auto alpha = op::Constant::create(et, c_shape, {1.67326324});
+    auto lambda = op::Constant::create(et, c_shape, {1.05070098});
+    auto selu = make_shared<op::v0::Selu>(input, alpha, lambda);
+    auto f = make_shared<Function>(selu, ParameterVector{input});
+
+    vector<float> in_vec{-1.0, 0.0, 1.0, 2.0};
+    vector<float> out_vec{-1.1113307, 0., 1.050701, 2.101402};
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_input<float>(in_shape, in_vec);
+    test_case.add_expected_output<float>(in_shape, out_vec);
+    test_case.run_with_tolerance_as_fp(1e-4f);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, selu_1Dfprop)
+{
+    Shape in_shape{1};
+    Shape c_shape{1};
+    element::Type et = element::f32;
+
+    auto input = make_shared<op::Parameter>(et, in_shape);
+    auto alpha = op::Constant::create(et, c_shape, {1.67326324});
+    auto lambda = op::Constant::create(et, c_shape, {1.05070098});
+    auto selu = make_shared<op::v0::Selu>(input, alpha, lambda);
+    auto f = make_shared<Function>(selu, ParameterVector{input});
+
+    vector<float> in_vec{112.0};
+    vector<float> out_vec{117.67851};
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_input<float>(in_shape, in_vec);
+    test_case.add_expected_output<float>(in_shape, out_vec);
+    test_case.run_with_tolerance_as_fp(1e-4f);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, selu_3Dfprop_negative)
+{
+    Shape in_shape{3};
+    Shape c_shape{1};
+    element::Type et = element::f32;
+
+    auto input = make_shared<op::Parameter>(et, in_shape);
+    auto alpha = op::Constant::create(et, c_shape, {1.67326324});
+    auto lambda = op::Constant::create(et, c_shape, {1.05070098});
+    auto selu = make_shared<op::v0::Selu>(input, alpha, lambda);
+    auto f = make_shared<Function>(selu, ParameterVector{input});
+
+    vector<float> in_vec{-3.0, -12.5, -7.0};
+    vector<float> out_vec{-1.6705687, -1.7580928, -1.7564961};
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_input<float>(in_shape, in_vec);
+    test_case.add_expected_output<float>(in_shape, out_vec);
+    test_case.run_with_tolerance_as_fp(1e-4f);
+}
diff --git a/ngraph/test/type_prop/selu.cpp b/ngraph/test/type_prop/selu.cpp
new file mode 100644
index 00000000000000..66f89a84f11b15
--- /dev/null
+++ b/ngraph/test/type_prop/selu.cpp
@@ -0,0 +1,171 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/type_prop.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+TEST(type_prop, selu_basic_inference_f32_3D)
+{
+    const auto param = make_shared<op::Parameter>(element::f32, Shape{1, 32, 32});
+    const auto alpha = make_shared<op::Parameter>(element::f32, Shape{1});
+    const auto lambda = make_shared<op::Parameter>(element::f32, Shape{1});
+    const auto selu = make_shared<op::Selu>(param, alpha, lambda);
+
+    ASSERT_EQ(selu->get_element_type(), element::f32);
+    ASSERT_EQ(selu->get_shape(), (Shape{1, 32, 32}));
+}
+
+TEST(type_prop, selu_basic_inference_f16_3D)
+{
+    const auto param = make_shared<op::Parameter>(element::f16, Shape{1, 32, 32});
+    const auto alpha = make_shared<op::Parameter>(element::f16, Shape{1});
+    const auto lambda = make_shared<op::Parameter>(element::f16, Shape{1});
+    const auto selu = make_shared<op::Selu>(param, alpha, lambda);
+
+    ASSERT_EQ(selu->get_element_type(), element::f16);
+    ASSERT_EQ(selu->get_shape(), (Shape{1, 32, 32}));
+}
+
+TEST(type_prop, selu_basic_inference_f32_5D)
+{
+    const auto param = make_shared<op::Parameter>(element::f32, Shape{12, 135, 221, 31, 15});
+    const auto alpha = make_shared<op::Parameter>(element::f32, Shape{1});
+    const auto lambda = make_shared<op::Parameter>(element::f32, Shape{1});
+    const auto selu = make_shared<op::Selu>(param, alpha, lambda);
+
+    ASSERT_EQ(selu->get_element_type(), element::f32);
+    ASSERT_EQ(selu->get_shape(), (Shape{12, 135, 221, 31, 15}));
+}
+
+TEST(type_prop, selu_basic_inference_f16_5D)
+{
+    const auto param = make_shared<op::Parameter>(element::f16, Shape{12, 135, 221, 31, 15});
+    const auto alpha = make_shared<op::Parameter>(element::f16, Shape{1});
+    const auto lambda = make_shared<op::Parameter>(element::f16, Shape{1});
+    const auto selu = make_shared<op::Selu>(param, alpha, lambda);
+
+    ASSERT_EQ(selu->get_element_type(), element::f16);
+    ASSERT_EQ(selu->get_shape(), (Shape{12, 135, 221, 31, 15}));
+}
+
+TEST(type_prop, selu_incompatible_input_type_boolean)
+{
+    // Invalid data input element type
+    try
+    {
+        auto data = make_shared<op::Parameter>(element::boolean, Shape{1, 2, 3, 4});
+        const auto alpha = make_shared<op::Parameter>(element::boolean, Shape{1});
+        const auto lambda = make_shared<op::Parameter>(element::boolean, Shape{1});
+        auto selu = make_shared<op::Selu>(data, alpha, lambda);
+        // Data input expected to be of numeric type
+        FAIL() << "Invalid input type not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), std::string("Input element types must be floating-point"));
+    }
+    catch (...)
+    {
+        FAIL() << "Input type check failed for unexpected reason";
+    }
+}
+
+TEST(type_prop, selu_incompatible_input_type_i32)
+{
+    // Invalid data input element type
+    try
+    {
+        auto data = make_shared<op::Parameter>(element::i32, Shape{1, 2, 3, 4});
+        const auto alpha = make_shared<op::Parameter>(element::i32, Shape{1});
+        const auto lambda = make_shared<op::Parameter>(element::i32, Shape{1});
+        auto selu = make_shared<op::Selu>(data, alpha, lambda);
+        // Data input expected to be of numeric type
+        FAIL() << "Invalid input type not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), std::string("Input element types must be floating-point"));
+    }
+    catch (...)
+    {
+        FAIL() << "Input type check failed for unexpected reason";
+    }
+}
+
+TEST(type_prop, selu_incompatible_input_type_u16)
+{
+    // Invalid data input element type
+    try
+    {
+        auto data = make_shared<op::Parameter>(element::u16, Shape{1, 2, 3, 4});
+        const auto alpha = make_shared<op::Parameter>(element::u16, Shape{1});
+        const auto lambda = make_shared<op::Parameter>(element::u16, Shape{1});
+        auto selu = make_shared<op::Selu>(data, alpha, lambda);
+        // Data input expected to be of numeric type
+        FAIL() << "Invalid input type not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), std::string("Input element types must be floating-point"));
+    }
+    catch (...)
+    {
+        FAIL() << "Input type check failed for unexpected reason";
+    }
+}
+
+TEST(type_prop, selu_incompatible_input_types)
+{
+    // Invalid data input element type
+    try
+    {
+        auto data = make_shared<op::Parameter>(element::f32, Shape{1, 2, 3, 4});
+        const auto alpha = make_shared<op::Parameter>(element::f32, Shape{1});
+        const auto lambda = make_shared<op::Parameter>(element::u16, Shape{1});
+        auto selu = make_shared<op::Selu>(data, alpha, lambda);
+        // Data input expected to be of numeric type
+        FAIL() << "Inavlid input types not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), std::string("Input element types do not match"));
+    }
+    catch (...)
+    {
+        FAIL() << "Input type check failed for unexpected reason";
+    }
+}
+
+TEST(type_prop, selu_dynamic_rank_input_shape_2D)
+{
+    const PartialShape param_shape{Dimension::dynamic(), 10};
+    const auto param = std::make_shared<op::Parameter>(element::f32, param_shape);
+    const auto alpha = make_shared<op::Parameter>(element::f32, Shape{2, 1});
+    const auto lambda = make_shared<op::Parameter>(element::f32, Shape{1});
+    const auto op = std::make_shared<op::Selu>(param, alpha, lambda);
+    ASSERT_TRUE(op->get_output_partial_shape(0).same_scheme(PartialShape{Dimension(), 10}));
+}
+
+TEST(type_prop, selu_dynamic_rank_input_shape_3D)
+{
+    const PartialShape param_shape{100, Dimension::dynamic(), 58};
+    const auto param = std::make_shared<op::Parameter>(element::f32, param_shape);
+    const auto alpha = make_shared<op::Parameter>(element::f32, Shape{1});
+    const auto lambda = make_shared<op::Parameter>(element::f32, Shape{1});
+    const auto op = std::make_shared<op::Selu>(param, alpha, lambda);
+    ASSERT_TRUE(op->get_output_partial_shape(0).same_scheme(PartialShape{100, Dimension(), 58}));
+}
+
+TEST(type_prop, selu_dynamic_rank_input_shape_full)
+{
+    const auto param = std::make_shared<op::Parameter>(element::f32, PartialShape::dynamic());
+    const auto alpha = make_shared<op::Parameter>(element::f32, Shape{1});
+    const auto lambda = make_shared<op::Parameter>(element::f32, Shape{1});
+    const auto op = std::make_shared<op::Selu>(param, alpha, lambda);
+    ASSERT_TRUE(op->get_output_partial_shape(0).same_scheme(PartialShape::dynamic()));
+}
diff --git a/ngraph/test/visitors/op/selu.cpp b/ngraph/test/visitors/op/selu.cpp
new file mode 100644
index 00000000000000..3a83deaea53086
--- /dev/null
+++ b/ngraph/test/visitors/op/selu.cpp
@@ -0,0 +1,30 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+
+#include "ngraph/ngraph.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+#include "ngraph/opsets/opset1.hpp"
+
+#include "util/visitor.hpp"
+
+using namespace std;
+using namespace ngraph;
+using ngraph::test::NodeBuilder;
+
+TEST(attributes, selu_op)
+{
+    NodeBuilder::get_ops().register_factory<opset1::Selu>();
+    const auto data_input = make_shared<op::Parameter>(element::f32, Shape{1, 2, 3});
+    const auto alpha = make_shared<op::Parameter>(element::f32, Shape{1});
+    const auto lambda = make_shared<op::Parameter>(element::f32, Shape{1});
+
+    const auto op = make_shared<opset1::Selu>(data_input, alpha, lambda);
+
+    NodeBuilder builder(op);
+    const auto expected_attr_count = 0;
+
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
+}

From dc415573d4b99e8510f38ba987ebbff19bd8b34e Mon Sep 17 00:00:00 2001
From: Gabriele Galiero Casay <gabriele.galiero.casay@intel.com>
Date: Mon, 14 Jun 2021 11:29:33 +0200
Subject: [PATCH 070/178] Revise reference implementation for ReduceProd
 operation (#5774)

---
 .../ngraph}/op/util/evaluate_helpers.hpp      |   9 +-
 .../ngraph/runtime/reference/product.hpp      |  31 ++-
 .../include/ngraph/runtime/reference/sum.hpp  |   8 +-
 ngraph/core/src/op/reduce_prod.cpp            |  10 +-
 ngraph/core/src/op/reduce_sum.cpp             |   4 +-
 ngraph/core/src/op/util/evaluate_helpers.cpp  |  17 ++
 ngraph/test/CMakeLists.txt                    |   1 +
 ngraph/test/backend/reduce_prod.in.cpp        | 228 ----------------
 ngraph/test/op_eval/reduce_prod.cpp           | 244 ++++++++++++++++++
 ngraph/test/runtime/ie/unit_test.manifest     |  25 +-
 10 files changed, 295 insertions(+), 282 deletions(-)
 rename ngraph/core/{src => include/ngraph}/op/util/evaluate_helpers.hpp (79%)
 create mode 100644 ngraph/core/src/op/util/evaluate_helpers.cpp
 create mode 100644 ngraph/test/op_eval/reduce_prod.cpp

diff --git a/ngraph/core/src/op/util/evaluate_helpers.hpp b/ngraph/core/include/ngraph/op/util/evaluate_helpers.hpp
similarity index 79%
rename from ngraph/core/src/op/util/evaluate_helpers.hpp
rename to ngraph/core/include/ngraph/op/util/evaluate_helpers.hpp
index d67a212f1574b6..abcd8729a98d86 100644
--- a/ngraph/core/src/op/util/evaluate_helpers.hpp
+++ b/ngraph/core/include/ngraph/op/util/evaluate_helpers.hpp
@@ -2,6 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+#pragma once
+
 #include "ngraph/axis_set.hpp"
 #include "ngraph/descriptor/tensor.hpp"
 #include "ngraph/util.hpp"
@@ -18,10 +20,5 @@ namespace ngraph
     /// \return Normalized (positive only) axes as an AxisSet object.
     AxisSet get_normalized_axes_from_tensor(const HostTensorPtr tensor,
                                             const ngraph::Rank& rank,
-                                            const std::string& node_description)
-    {
-        const auto axes_vector = host_tensor_2_vector<int64_t>(tensor);
-        const auto normalized_axes = ngraph::normalize_axes(node_description, axes_vector, rank);
-        return AxisSet{normalized_axes};
-    }
+                                            const std::string& node_description);
 } // namespace ngraph
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/product.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/product.hpp
index 14615dc8b7cf9d..addd17eb82a647 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/product.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/product.hpp
@@ -5,6 +5,7 @@
 #pragma once
 
 #include <cmath>
+#include <numeric>
 
 #include "ngraph/coordinate_transform.hpp"
 #include "ngraph/shape_util.hpp"
@@ -16,29 +17,27 @@ namespace ngraph
         namespace reference
         {
             template <typename T>
-            void product(const T* arg,
-                         T* out,
-                         const Shape& in_shape,
-                         const AxisSet& reduction_axes,
-                         bool keep_dims)
+            void product(const T* arg, T* out, const Shape& in_shape, const AxisSet& reduction_axes)
             {
-                auto out_shape = reduce(in_shape, reduction_axes, keep_dims);
-                CoordinateTransform output_transform(out_shape);
+                constexpr bool dont_keep_dims_in_output = false;
+                const auto out_shape = reduce(in_shape, reduction_axes, dont_keep_dims_in_output);
+                std::fill(out, out + shape_size(out_shape), 1);
 
-                for (const Coordinate& output_coord : output_transform)
-                {
-                    out[output_transform.index(output_coord)] = 1;
-                }
-
-                CoordinateTransform input_transform(in_shape);
+                const auto in_strides = row_major_strides(in_shape);
+                const auto out_strides = row_major_strides(out_shape);
 
+                CoordinateTransformBasic input_transform(in_shape);
                 for (const Coordinate& input_coord : input_transform)
                 {
-                    Coordinate output_coord = reduce(input_coord, reduction_axes, keep_dims);
+                    const Coordinate output_coord =
+                        reduce(input_coord, reduction_axes, dont_keep_dims_in_output);
 
-                    size_t output_index = output_transform.index(output_coord);
+                    const size_t in_idx = std::inner_product(
+                        input_coord.begin(), input_coord.end(), in_strides.begin(), 0);
+                    const size_t out_idx = std::inner_product(
+                        output_coord.begin(), output_coord.end(), out_strides.begin(), 0);
 
-                    out[output_index] = out[output_index] * arg[input_transform.index(input_coord)];
+                    out[out_idx] = out[out_idx] * arg[in_idx];
                 }
             }
         } // namespace reference
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/sum.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/sum.hpp
index 2408e6f476d728..d73ec441f7b3a5 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/sum.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/sum.hpp
@@ -45,7 +45,7 @@ namespace ngraph
             void sum(const T* arg, T* out, const Shape& in_shape, const AxisSet& reduction_axes)
             {
                 constexpr bool dont_keep_dims_in_output = false;
-                auto out_shape = reduce(in_shape, reduction_axes, dont_keep_dims_in_output);
+                const auto out_shape = reduce(in_shape, reduction_axes, dont_keep_dims_in_output);
 
                 std::vector<T> cs(shape_size(out_shape), 0);
                 std::fill(out, out + shape_size(out_shape), 0);
@@ -56,12 +56,12 @@ namespace ngraph
                 CoordinateTransformBasic input_transform(in_shape);
                 for (const Coordinate& input_coord : input_transform)
                 {
-                    Coordinate output_coord =
+                    const Coordinate output_coord =
                         reduce(input_coord, reduction_axes, dont_keep_dims_in_output);
 
-                    size_t in_idx = std::inner_product(
+                    const size_t in_idx = std::inner_product(
                         input_coord.begin(), input_coord.end(), in_strides.begin(), 0);
-                    size_t out_idx = std::inner_product(
+                    const size_t out_idx = std::inner_product(
                         output_coord.begin(), output_coord.end(), out_strides.begin(), 0);
 
                     T x = arg[in_idx];
diff --git a/ngraph/core/src/op/reduce_prod.cpp b/ngraph/core/src/op/reduce_prod.cpp
index bbacf23cc7e4d4..c7a78beb51f30b 100644
--- a/ngraph/core/src/op/reduce_prod.cpp
+++ b/ngraph/core/src/op/reduce_prod.cpp
@@ -6,6 +6,7 @@
 #include <ngraph/validation_util.hpp>
 #include "itt.hpp"
 #include "ngraph/graph_util.hpp"
+#include "ngraph/op/util/evaluate_helpers.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
 #include "ngraph/runtime/reference/product.hpp"
 #include "ngraph/shape_util.hpp"
@@ -45,7 +46,7 @@ namespace reduce_prod
     {
         out->set_shape(reduce(arg->get_shape(), axes, keep_dims));
         runtime::reference::product(
-            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes, keep_dims);
+            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes);
         return true;
     }
 
@@ -75,8 +76,11 @@ bool op::v1::ReduceProd::evaluate(const HostTensorVector& outputs,
     NGRAPH_OP_SCOPE(v1_ReduceProd_evaluate);
     NGRAPH_CHECK(validate_host_tensor_vector(inputs, 2));
     NGRAPH_CHECK(validate_host_tensor_vector(outputs, 1));
-    return reduce_prod::evaluate_product(
-        inputs[0], outputs[0], get_reduction_axes(), get_keep_dims());
+
+    const auto reduction_axes = get_normalized_axes_from_tensor(
+        inputs[1], inputs[0]->get_partial_shape().rank(), get_friendly_name());
+
+    return reduce_prod::evaluate_product(inputs[0], outputs[0], reduction_axes, get_keep_dims());
 }
 
 bool op::v1::ReduceProd::has_evaluate() const
diff --git a/ngraph/core/src/op/reduce_sum.cpp b/ngraph/core/src/op/reduce_sum.cpp
index 34262ec9efb32d..50b87763e84c4d 100644
--- a/ngraph/core/src/op/reduce_sum.cpp
+++ b/ngraph/core/src/op/reduce_sum.cpp
@@ -7,11 +7,11 @@
 #include "itt.hpp"
 #include "ngraph/graph_util.hpp"
 #include "ngraph/op/broadcast.hpp"
+#include "ngraph/op/util/evaluate_helpers.hpp"
 #include "ngraph/op/util/op_types.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
 #include "ngraph/runtime/reference/sum.hpp"
 #include "ngraph/shape_util.hpp"
-#include "util/evaluate_helpers.hpp"
 
 using namespace std;
 using namespace ngraph;
@@ -80,7 +80,7 @@ bool op::v1::ReduceSum::evaluate(const HostTensorVector& outputs,
     NGRAPH_CHECK(validate_host_tensor_vector(outputs, 1));
 
     const auto reduction_axes = get_normalized_axes_from_tensor(
-        inputs[1], get_input_partial_shape(0).rank(), get_friendly_name());
+        inputs[1], inputs[0]->get_partial_shape().rank(), get_friendly_name());
 
     return reduce_sum::evaluate_sum(inputs[0], outputs[0], reduction_axes, get_keep_dims());
 }
diff --git a/ngraph/core/src/op/util/evaluate_helpers.cpp b/ngraph/core/src/op/util/evaluate_helpers.cpp
new file mode 100644
index 00000000000000..ac55bb5fe21b3f
--- /dev/null
+++ b/ngraph/core/src/op/util/evaluate_helpers.cpp
@@ -0,0 +1,17 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "ngraph/op/util/evaluate_helpers.hpp"
+
+namespace ngraph
+{
+    AxisSet get_normalized_axes_from_tensor(const HostTensorPtr tensor,
+                                            const ngraph::Rank& rank,
+                                            const std::string& node_description)
+    {
+        const auto axes_vector = host_tensor_2_vector<int64_t>(tensor);
+        const auto normalized_axes = ngraph::normalize_axes(node_description, axes_vector, rank);
+        return AxisSet{normalized_axes};
+    }
+} // namespace ngraph
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index d04486563866ef..97065dc966eb6a 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -73,6 +73,7 @@ set(SRC
     op_eval/non_zero.cpp
     op_eval/reduce_l1.cpp
     op_eval/reduce_l2.cpp
+    op_eval/reduce_prod.cpp
     op_eval/reduce_sum.cpp
     op_eval/roi_align.cpp
     op_eval/roi_pooling.cpp
diff --git a/ngraph/test/backend/reduce_prod.in.cpp b/ngraph/test/backend/reduce_prod.in.cpp
index 6f63089acaa6a9..dbd7d3c743bb78 100644
--- a/ngraph/test/backend/reduce_prod.in.cpp
+++ b/ngraph/test/backend/reduce_prod.in.cpp
@@ -80,95 +80,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_product_matrix_rows)
     EXPECT_TRUE(test::all_close_f((vector<float>{2, 12, 30}), read_vector<float>(result)));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_product_matrix_rows_zero)
-{
-    Shape shape_a{3, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1, 1}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_product_matrix_cols_zero)
-{
-    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
-    Shape shape_a{0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_product_vector_zero)
-{
-    Shape shape_a{0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{1}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_product_matrix_to_scalar_zero_by_zero)
-{
-    Shape shape_a{0, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{1}), read_vector<float>(result)));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_product_3d_to_matrix_most_sig)
 {
     Shape shape_a{3, 3, 3};
@@ -283,31 +194,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_product_3d_to_scalar)
                                   read_vector<float>(result)));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_product_3d_eliminate_zero_dim)
-{
-    Shape shape_a{3, 0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-
-    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
-    // right value.
-    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1, 1, 1, 1, 1}), read_vector<float>(result)));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_product_2d_to_scalar_int32)
 {
     Shape shape_a{3, 3};
@@ -433,95 +319,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_product_keep_matrix_rows)
     EXPECT_TRUE(test::all_close_f((vector<float>{2, 12, 30}), read_vector<float>(result)));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_product_keep_matrix_rows_zero)
-{
-    Shape shape_a{3, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1, 1}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_product_keep_matrix_cols_zero)
-{
-    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
-    Shape shape_a{0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_product_keep_vector_zero)
-{
-    Shape shape_a{0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{1}), read_vector<float>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_product_keep_matrix_to_scalar_zero_by_zero)
-{
-    Shape shape_a{0, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{1}), read_vector<float>(result)));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_product_keep_3d_to_matrix_most_sig)
 {
     Shape shape_a{3, 3, 3};
@@ -636,31 +433,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_product_keep_3d_to_scalar)
                                   read_vector<float>(result)));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_product_keep_3d_eliminate_zero_dim)
-{
-    Shape shape_a{3, 0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 1, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-
-    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
-    // right value.
-    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1, 1, 1, 1, 1}), read_vector<float>(result)));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_product_keep_2d_to_scalar_int32)
 {
     Shape shape_a{3, 3};
diff --git a/ngraph/test/op_eval/reduce_prod.cpp b/ngraph/test/op_eval/reduce_prod.cpp
new file mode 100644
index 00000000000000..d2cd89888ea31e
--- /dev/null
+++ b/ngraph/test/op_eval/reduce_prod.cpp
@@ -0,0 +1,244 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/test_control.hpp"
+#include "util/all_close.hpp"
+#include "util/all_close_f.hpp"
+#include "util/ndarray.hpp"
+
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+
+TEST(op_eval, reduce_product_matrix_rows_zero)
+{
+    Shape shape_a{3, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1, 1}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_product_matrix_cols_zero)
+{
+    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
+    Shape shape_a{0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_product_vector_zero)
+{
+    Shape shape_a{0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{1}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_product_matrix_to_scalar_zero_by_zero)
+{
+    Shape shape_a{0, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{1}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_product_3d_eliminate_zero_dim)
+{
+    Shape shape_a{3, 0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+
+    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
+    // right value.
+    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1, 1, 1, 1, 1}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_product_keep_matrix_rows_zero)
+{
+    Shape shape_a{3, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1, 1}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_product_keep_matrix_cols_zero)
+{
+    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
+    Shape shape_a{0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_product_keep_vector_zero)
+{
+    Shape shape_a{0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{1}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_product_keep_matrix_to_scalar_zero_by_zero)
+{
+    Shape shape_a{0, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1, 1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{1}), read_vector<float>(result)));
+}
+
+TEST(op_eval, reduce_product_keep_3d_eliminate_zero_dim)
+{
+    Shape shape_a{3, 0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 1, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceProd>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+
+    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
+    // right value.
+    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1, 1, 1, 1, 1}), read_vector<float>(result)));
+}
diff --git a/ngraph/test/runtime/ie/unit_test.manifest b/ngraph/test/runtime/ie/unit_test.manifest
index 425ec0cdf48e30..80f2421fb41d8f 100644
--- a/ngraph/test/runtime/ie/unit_test.manifest
+++ b/ngraph/test/runtime/ie/unit_test.manifest
@@ -373,6 +373,8 @@ all_dynamic
 # disabled reference implementation
 reduce_sum_keep_2d_to_scalar_int8
 reduce_sum_2d_to_scalar_int8
+reduce_product_to_scalar_int8
+reduce_product_keep_to_scalar_int8
 # accuracy
 reduce_sum_keep_stable_acc
 reduce_sum_keep_3d_to_scalar_int32
@@ -458,17 +460,6 @@ onnx_dyn_shapes_model_tile_static
 gather_4d_indices_axis_0_uint8
 tensor_constant_with_op
 constant_equality_bool
-reduce_product_matrix_rows
-reduce_product_3d_to_matrix_most_sig
-reduce_product_3d_to_matrix_least_sig
-reduce_product_keep_matrix_columns
-reduce_product_keep_matrix_rows
-reduce_product_keep_3d_to_matrix_most_sig
-reduce_product_keep_3d_to_matrix_least_sig
-reduce_product_matrix_columns_dynamic
-reduce_product_matrix_rows_dynamic
-reduce_product_keep_matrix_columns_dynamic
-reduce_product_keep_matrix_rows_dynamic
 reduce_min_matrix_columns
 reduce_min_matrix_rows
 reduce_min_matrix_rows_int32
@@ -485,18 +476,6 @@ reduce_min_keep_matrix_columns_dynamic
 reduce_min_keep_matrix_rows_dynamic
 
 # zero dimension / result mismatch
-reduce_product_matrix_rows_zero
-reduce_product_matrix_cols_zero
-reduce_product_vector_zero
-reduce_product_matrix_to_scalar_zero_by_zero
-reduce_product_3d_eliminate_zero_dim
-reduce_product_to_scalar_int8
-reduce_product_keep_matrix_rows_zero
-reduce_product_keep_matrix_cols_zero
-reduce_product_keep_vector_zero
-reduce_product_keep_matrix_to_scalar_zero_by_zero
-reduce_product_keep_3d_eliminate_zero_dim
-reduce_product_keep_to_scalar_int8
 reduce_min_to_scalar_int8
 reduce_min_matrix_rows_zero
 reduce_min_matrix_cols_zero

From 826638e52349992cc5c85ea41e95d5bd83a9bd83 Mon Sep 17 00:00:00 2001
From: Bartek Szmelczynski <bartosz.szmelczynski@intel.com>
Date: Mon, 14 Jun 2021 12:18:22 +0200
Subject: [PATCH 071/178] Revise space_to_batch (#6034)

---
 .../single_layer/space_to_batch.cpp           |  38 ++++++
 .../single_layer_tests/space_to_batch.cpp     |  91 ++++++++------
 .../core/include/ngraph/op/space_to_batch.hpp |   7 +-
 ngraph/core/src/op/space_to_batch.cpp         |   6 +-
 ngraph/test/CMakeLists.txt                    |   2 +
 ngraph/test/backend/space_to_batch.in.cpp     | 115 ++++++++++++++++++
 ngraph/test/type_prop/space_to_batch.cpp      | 101 +++++++++++++++
 ngraph/test/visitors/op/space_to_batch.cpp    |  33 +++++
 8 files changed, 349 insertions(+), 44 deletions(-)
 create mode 100644 inference-engine/tests/functional/inference_engine/serialization/single_layer/space_to_batch.cpp
 create mode 100644 ngraph/test/backend/space_to_batch.in.cpp
 create mode 100644 ngraph/test/visitors/op/space_to_batch.cpp

diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/space_to_batch.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/space_to_batch.cpp
new file mode 100644
index 00000000000000..1b55e2fbac71f1
--- /dev/null
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/space_to_batch.cpp
@@ -0,0 +1,38 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "shared_test_classes/single_layer/space_to_batch.hpp"
+
+#include <vector>
+
+using namespace LayerTestsDefinitions;
+
+namespace {
+
+TEST_P(SpaceToBatchLayerTest, Serialize) {
+    Serialize();
+}
+
+const std::vector<std::vector<int64_t>> blockShapes4D{{1, 1, 2, 2}};
+const std::vector<std::vector<int64_t>> padsBegins4D{{0, 0, 0, 0},
+                                                     {0, 0, 0, 2}};
+const std::vector<std::vector<int64_t>> padsEnds4D{{0, 0, 0, 0}, {0, 0, 0, 2}};
+const std::vector<std::vector<size_t>> dataShapes4D{
+    {1, 1, 2, 2}, {1, 3, 2, 2}, {1, 1, 4, 4}, {2, 1, 2, 4}};
+
+const auto SpaceToBatch4D = ::testing::Combine(
+    ::testing::ValuesIn(blockShapes4D), ::testing::ValuesIn(padsBegins4D),
+    ::testing::ValuesIn(padsEnds4D), ::testing::ValuesIn(dataShapes4D),
+    ::testing::Values(InferenceEngine::Precision::FP32),
+    ::testing::Values(InferenceEngine::Precision::FP32),
+    ::testing::Values(InferenceEngine::Precision::FP32),
+    ::testing::Values(InferenceEngine::Layout::ANY),
+    ::testing::Values(InferenceEngine::Layout::ANY),
+    ::testing::Values(CommonTestUtils::DEVICE_CPU));
+
+INSTANTIATE_TEST_CASE_P(smoke_spacetobatch4D_Serialization,
+                        SpaceToBatchLayerTest, SpaceToBatch4D,
+                        SpaceToBatchLayerTest::getTestCaseName);
+
+}  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp
index ac4801df822a54..884440c7bdedbc 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp
@@ -11,47 +11,64 @@ using namespace LayerTestsDefinitions;
 
 namespace {
 
-spaceToBatchParamsTuple stb_only_test_cases[] = {
-        spaceToBatchParamsTuple({1, 1, 2, 2}, {0, 0, 0, 0}, {0, 0, 0, 0}, {1, 1, 2, 2},
-                                InferenceEngine::Precision::FP32,
-                                InferenceEngine::Precision::UNSPECIFIED,
-                                InferenceEngine::Precision::UNSPECIFIED,
-                                InferenceEngine::Layout::ANY,
-                                InferenceEngine::Layout::ANY,
-                                CommonTestUtils::DEVICE_CPU),
-        spaceToBatchParamsTuple({1, 1, 2, 2}, {0, 0, 0, 0}, {0, 0, 0, 0}, {1, 3, 2, 2},
-                                InferenceEngine::Precision::FP32,
-                                InferenceEngine::Precision::UNSPECIFIED,
-                                InferenceEngine::Precision::UNSPECIFIED,
-                                InferenceEngine::Layout::ANY,
-                                InferenceEngine::Layout::ANY,
-                                CommonTestUtils::DEVICE_CPU),
-        spaceToBatchParamsTuple({1, 1, 2, 2}, {0, 0, 0, 0}, {0, 0, 0, 0}, {1, 1, 4, 4},
-                                InferenceEngine::Precision::FP32,
-                                InferenceEngine::Precision::UNSPECIFIED,
-                                InferenceEngine::Precision::UNSPECIFIED,
-                                InferenceEngine::Layout::ANY,
-                                InferenceEngine::Layout::ANY,
-                                CommonTestUtils::DEVICE_CPU),
-        spaceToBatchParamsTuple({1, 1, 2, 2}, {0, 0, 0, 2}, {0, 0, 0, 0}, {2, 1, 2, 4},
-                                InferenceEngine::Precision::FP32,
-                                InferenceEngine::Precision::UNSPECIFIED,
-                                InferenceEngine::Precision::UNSPECIFIED,
-                                InferenceEngine::Layout::ANY,
-                                InferenceEngine::Layout::ANY,
-                                CommonTestUtils::DEVICE_CPU),
-        spaceToBatchParamsTuple({1, 1, 3, 2, 2}, {0, 0, 1, 0, 3}, {0, 0, 2, 0, 0}, {1, 1, 3, 2, 1},
-                                InferenceEngine::Precision::FP32,
-                                InferenceEngine::Precision::UNSPECIFIED,
-                                InferenceEngine::Precision::UNSPECIFIED,
-                                InferenceEngine::Layout::ANY,
-                                InferenceEngine::Layout::ANY,
-                                CommonTestUtils::DEVICE_CPU),
+const std::vector<std::vector<int64_t >> blockShapes4D {
+        {1, 1, 2, 2}
+};
+const std::vector<std::vector<int64_t >> padsBegins4D {
+        {0, 0, 0, 0}, {0, 0, 0, 2}
+};
+const std::vector<std::vector<int64_t >> padsEnds4D {
+        {0, 0, 0, 0}, {0, 0, 0, 2}
+};
+const std::vector<std::vector<size_t >> dataShapes4D {
+        {1, 1, 2, 2}, {1, 3, 2, 2}, {1, 1, 4, 4}, {2, 1, 2, 4}
 };
 
+const auto SpaceToBatch4D = ::testing::Combine(
+        ::testing::ValuesIn(blockShapes4D),
+        ::testing::ValuesIn(padsBegins4D),
+        ::testing::ValuesIn(padsEnds4D),
+        ::testing::ValuesIn(dataShapes4D),
+        ::testing::Values(InferenceEngine::Precision::FP32),
+        ::testing::Values(InferenceEngine::Precision::FP32),
+        ::testing::Values(InferenceEngine::Precision::FP32),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(CommonTestUtils::DEVICE_CPU)
+);
+
 INSTANTIATE_TEST_CASE_P(
-        smoke_MKLDNN, SpaceToBatchLayerTest, ::testing::ValuesIn(stb_only_test_cases),
+        smoke_spacetobatch4D, SpaceToBatchLayerTest, SpaceToBatch4D,
         SpaceToBatchLayerTest::getTestCaseName);
 
+const std::vector<std::vector<int64_t >> blockShapes5D {
+        {1, 1, 3, 2, 2}
+};
+const std::vector<std::vector<int64_t >> padsBegins5D {
+        {0, 0, 1, 0, 3}
+};
+const std::vector<std::vector<int64_t >> padsEnds5D {
+        {0, 0, 2, 0, 0}
+};
+const std::vector<std::vector<size_t >> dataShapes5D {
+        {1, 1, 3, 2, 1}
+};
+
+const auto SpaceToBatch5D = ::testing::Combine(
+        ::testing::ValuesIn(blockShapes5D),
+        ::testing::ValuesIn(padsBegins5D),
+        ::testing::ValuesIn(padsEnds5D),
+        ::testing::ValuesIn(dataShapes5D),
+        ::testing::Values(InferenceEngine::Precision::FP32),
+        ::testing::Values(InferenceEngine::Precision::FP32),
+        ::testing::Values(InferenceEngine::Precision::FP32),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(CommonTestUtils::DEVICE_CPU)
+);
+
+INSTANTIATE_TEST_CASE_P(
+        smoke_spacetobatch5D, SpaceToBatchLayerTest, SpaceToBatch5D,
+        SpaceToBatchLayerTest::getTestCaseName);
 
 }  // namespace
\ No newline at end of file
diff --git a/ngraph/core/include/ngraph/op/space_to_batch.hpp b/ngraph/core/include/ngraph/op/space_to_batch.hpp
index 35587652384100..9f6220b3293cbd 100644
--- a/ngraph/core/include/ngraph/op/space_to_batch.hpp
+++ b/ngraph/core/include/ngraph/op/space_to_batch.hpp
@@ -4,8 +4,7 @@
 
 #pragma once
 
-#include "ngraph/node.hpp"
-#include "ngraph/op/util/fused_op.hpp"
+#include "ngraph/op/op.hpp"
 
 namespace ngraph
 {
@@ -27,8 +26,8 @@ namespace ngraph
             class NGRAPH_API SpaceToBatch : public Op
             {
             public:
-                static constexpr NodeTypeInfo type_info{"SpaceToBatch", 1};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
+                NGRAPH_RTTI_DECLARATION;
+
                 SpaceToBatch() = default;
 
                 /// \brief Constructs a SpaceToBatch operation.
diff --git a/ngraph/core/src/op/space_to_batch.cpp b/ngraph/core/src/op/space_to_batch.cpp
index 75c6ad435b050e..885228536529a6 100644
--- a/ngraph/core/src/op/space_to_batch.cpp
+++ b/ngraph/core/src/op/space_to_batch.cpp
@@ -21,7 +21,7 @@
 using namespace std;
 using namespace ngraph;
 
-constexpr NodeTypeInfo op::v1::SpaceToBatch::type_info;
+NGRAPH_RTTI_DEFINITION(op::v1::SpaceToBatch, "SpaceToBatch", 1);
 
 ngraph::op::v1::SpaceToBatch::SpaceToBatch(const ngraph::Output<ngraph::Node>& data,
                                            const ngraph::Output<ngraph::Node>& block_shape,
@@ -49,13 +49,13 @@ void op::v1::SpaceToBatch::validate_and_infer_types()
 
     NODE_VALIDATION_CHECK(this,
                           pads_begin_type.is_integral_number(),
-                          "crops_begin must be an integral number but got (",
+                          "pads_begin must be an integral number but got (",
                           pads_begin_type,
                           ").");
 
     NODE_VALIDATION_CHECK(this,
                           pads_end_type.is_integral_number(),
-                          "crops_end must be an integral number but got (",
+                          "pads_end must be an integral number but got (",
                           pads_end_type,
                           ").");
 
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 97065dc966eb6a..5ca3049b6e2095 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -275,6 +275,7 @@ set(SRC
     visitors/op/selu.cpp
     visitors/op/shuffle_channels.cpp
     visitors/op/softmax.cpp
+    visitors/op/space_to_batch.cpp
     visitors/op/space_to_depth.cpp
     visitors/op/split.cpp
     visitors/op/squared_difference.cpp
@@ -445,6 +446,7 @@ set(MULTI_TEST_SRC
     backend/sin.in.cpp
     backend/sinh.in.cpp
     backend/softmax.in.cpp
+    backend/space_to_batch.in.cpp
     backend/split.in.cpp
     backend/sqrt.in.cpp
     backend/squared_difference.in.cpp
diff --git a/ngraph/test/backend/space_to_batch.in.cpp b/ngraph/test/backend/space_to_batch.in.cpp
new file mode 100644
index 00000000000000..6e9f20804345bb
--- /dev/null
+++ b/ngraph/test/backend/space_to_batch.in.cpp
@@ -0,0 +1,115 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/engine/test_engines.hpp"
+#include "util/test_case.hpp"
+#include "util/test_control.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+using TestEngine = test::ENGINE_CLASS_NAME(${BACKEND_NAME});
+
+static void SpaceToBatchTest(const std::vector<float>& inputs,
+                            const Shape inputs_shape,
+                            const std::vector<int64_t>& block_shapes,
+                            const Shape blocks_shape,
+                            const std::vector<int64_t>& pads_begins,
+                            const std::vector<int64_t>& pads_ends,
+                            const Shape pads_shape,
+                            const std::vector<float>& outputs,
+                            const Shape outputs_shape)
+{
+    auto inputs_param = make_shared<op::Parameter>(element::f32, inputs_shape);
+    auto block_shapes_param = make_shared<op::Constant>(element::i64, blocks_shape, block_shapes);
+    auto pads_begins_param = make_shared<op::Constant>(element::i64, pads_shape, pads_begins);
+    auto pads_ends_param = make_shared<op::Constant>(element::i64, pads_shape, pads_ends);
+
+    auto space_to_batch = make_shared<op::v1::SpaceToBatch>(
+        inputs_param, block_shapes_param, pads_begins_param, pads_ends_param);
+    auto f = make_shared<Function>(space_to_batch, ParameterVector{inputs_param});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_input<float>(inputs);
+    test_case.add_expected_output<float>(outputs_shape, outputs);
+    test_case.run();
+}
+
+
+NGRAPH_TEST(${BACKEND_NAME}, space_to_batch_4D)
+{   
+    const Shape inputs_shape{1, 1, 2, 2};
+    const std::vector<float> inputs{1.0f, 1.0f, 
+                                    1.0f, 1.0f};
+
+    const Shape blocks_shape{4};
+    const std::vector<int64_t> block_shapes{1, 1, 1, 1};
+
+    const Shape pads_shape{4};
+    const std::vector<int64_t> pads_begins{0, 0 ,0, 0};
+    const std::vector<int64_t> pads_ends{0, 0, 0, 0};
+
+    const Shape outputs_shape{1, 1, 2, 2};
+    const std::vector<float> outputs{1.0f, 1.0f,
+                                     1.0f, 1.0f};
+
+
+    SpaceToBatchTest(inputs, inputs_shape, block_shapes, blocks_shape, pads_begins, 
+                     pads_ends, pads_shape, outputs, outputs_shape);    
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, space_to_batch_5D)
+{   
+    const Shape inputs_shape{1, 1, 3, 2, 1};
+    const std::vector<float> inputs{1.0f, 1.0f, 1.0f,
+                                    1.0f, 1.0f, 1.0f};
+
+    const Shape blocks_shape{5};
+    const std::vector<int64_t> block_shapes{1, 1, 3, 2, 2};
+
+    const Shape pads_shape{5};
+    const std::vector<int64_t> pads_begins{0, 0 ,1, 0, 3};
+    const std::vector<int64_t> pads_ends{0, 0, 2, 0, 0};
+
+    const Shape outputs_shape{12, 1, 2, 1, 2};
+    const std::vector<float> outputs{0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f,
+                                     0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f,
+                                     0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f,
+                                     0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f,
+                                     0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f,
+                                     0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f};
+
+
+    SpaceToBatchTest(inputs, inputs_shape, block_shapes, blocks_shape, pads_begins, 
+                     pads_ends, pads_shape, outputs, outputs_shape);    
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, space_to_batch_4x4)
+{   
+    const Shape inputs_shape{1, 1, 4, 4};
+    const std::vector<float> inputs{1.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 1.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 1.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 1.0f};
+
+    const Shape blocks_shape{4};
+    const std::vector<int64_t> block_shapes{1, 1, 1, 1};
+
+    const Shape pads_shape{4};
+    const std::vector<int64_t> pads_begins{0, 0, 1, 0};
+    const std::vector<int64_t> pads_ends{0, 0, 0, 0};
+
+    const Shape outputs_shape{1, 1, 5, 4};
+    const std::vector<float> outputs{0.0f, 0.0f, 0.0f, 0.0f,
+                                     1.0f, 0.0f, 0.0f, 0.0f,
+                                     0.0f, 1.0f, 0.0f, 0.0f,
+                                     0.0f, 0.0f, 1.0f, 0.0f,
+                                     0.0f, 0.0f, 0.0f, 1.0f};
+
+    SpaceToBatchTest(inputs, inputs_shape, block_shapes, blocks_shape, pads_begins, 
+                     pads_ends, pads_shape, outputs, outputs_shape);    
+}
diff --git a/ngraph/test/type_prop/space_to_batch.cpp b/ngraph/test/type_prop/space_to_batch.cpp
index cf4b3ffdf7f9cd..18cce1d1648b49 100644
--- a/ngraph/test/type_prop/space_to_batch.cpp
+++ b/ngraph/test/type_prop/space_to_batch.cpp
@@ -110,3 +110,104 @@ TEST(type_prop, space_to_batch_dynamic_shape_dynamic_rank)
     ASSERT_EQ(space_to_batch->get_element_type(), element::f32);
     ASSERT_EQ(space_to_batch->get_output_partial_shape(0), PartialShape::dynamic());
 }
+
+TEST(type_prop, space_to_batch_invalid_element_type_block_shape)
+{
+    auto data = make_shared<op::Parameter>(element::f32, Shape{2, 128});
+    auto block_shape = make_shared<op::Constant>(element::f32, Shape{2}, vector<int64_t>{1, 5});
+    auto pads_begin = make_shared<op::Constant>(element::i64, Shape{2}, vector<float>{0, 2});
+    auto pads_end = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{0, 0});
+
+    try
+    {
+        auto space_to_batch =
+            make_shared<op::v1::SpaceToBatch>(data, block_shape, pads_begin, pads_end);
+        // Input element type is float32
+        FAIL() << "Invalid f32 element type for block_shape not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), "block_shape must be an integral number");
+    }
+    catch(...)
+    {
+        FAIL() << "Integral element type node validation check failed for unexpected reason";
+    }
+
+}
+
+TEST(type_prop, space_to_batch_invalid_element_type_pads_begin)
+{
+    auto data = make_shared<op::Parameter>(element::f32, Shape{2, 128});
+    auto block_shape = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 5});
+    auto pads_begin = make_shared<op::Constant>(element::f32, Shape{2}, vector<float>{0, 2});
+    auto pads_end = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{0, 0});
+
+    try
+    {
+        auto space_to_batch =
+            make_shared<op::v1::SpaceToBatch>(data, block_shape, pads_begin, pads_end);
+        // Input element type is float32
+        FAIL() << "Invalid f32 element type for pads_begin not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), "pads_begin must be an integral number but got");
+    }
+    catch(...)
+    {
+        FAIL() << "Integral element type node validation check failed for unexpected reason";
+    }
+
+}
+
+TEST(type_prop, space_to_batch_invalid_element_type_pads_end)
+{
+    auto data = make_shared<op::Parameter>(element::f32, Shape{2, 128});
+    auto block_shape = make_shared<op::Constant>(element::i16, Shape{2}, vector<int64_t>{1, 5});
+    auto pads_begin = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{0, 2});
+    auto pads_end = make_shared<op::Constant>(element::f32, Shape{2}, vector<float>{0, 0});
+
+    try
+    {
+        auto space_to_batch =
+            make_shared<op::v1::SpaceToBatch>(data, block_shape, pads_begin, pads_end);
+        // Input element type is float32
+        FAIL() << "Invalid f32 element type for pads_end not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), "pads_end must be an integral number but got");
+    }
+    catch(...)
+    {
+        FAIL() << "Integral element type node validation check failed for unexpected reason";
+    }
+
+}
+
+TEST(type_prop, space_to_batch_invalid_value_block_shape)
+{
+    auto data = make_shared<op::Parameter>(element::f32, Shape{2, 128});
+    auto block_shape = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{-1, -5});
+    auto pads_begin = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{0, 2});
+    auto pads_end = make_shared<op::Constant>(element::i64, Shape{2}, vector<float>{0, 0});
+
+    try
+    {
+        auto space_to_batch =
+            make_shared<op::v1::SpaceToBatch>(data, block_shape, pads_begin, pads_end);
+        // Input element type is float32
+        FAIL() << "Invalid block_shape value not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), "block_shape values must be greater than 0");
+    }
+    catch(...)
+    {
+        FAIL() << "block_shape value node validation check failed for unexpected reason";
+    }
+
+}
+
diff --git a/ngraph/test/visitors/op/space_to_batch.cpp b/ngraph/test/visitors/op/space_to_batch.cpp
new file mode 100644
index 00000000000000..c105cd2a454ac1
--- /dev/null
+++ b/ngraph/test/visitors/op/space_to_batch.cpp
@@ -0,0 +1,33 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+
+#include "ngraph/ngraph.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+#include "ngraph/opsets/opset2.hpp"
+
+#include "util/visitor.hpp"
+
+using namespace std;
+using namespace ngraph;
+using ngraph::test::NodeBuilder;
+
+
+TEST(attributes, space_to_batch_op)
+{
+    using namespace opset2;
+    
+    NodeBuilder::get_ops().register_factory<SpaceToBatch>();
+    auto data = make_shared<op::Parameter>(element::f32, Shape{2, 128});
+    auto block_shape = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{1, 5});
+    auto pads_begin = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{0, 2});
+    auto pads_end = make_shared<op::Constant>(element::i64, Shape{2}, vector<int64_t>{0, 0});
+    auto op = make_shared<SpaceToBatch>(data, block_shape, pads_begin, pads_end);
+
+    NodeBuilder builder(op);
+    const auto expected_attr_count = 0;
+
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
+}

From 978121e91ed1a974b4576e78b7fea89186f39438 Mon Sep 17 00:00:00 2001
From: Szymon Durawa <szymon.durawa@intel.com>
Date: Mon, 14 Jun 2021 13:28:32 +0200
Subject: [PATCH 072/178] Ref implementation transposed convolution revise
 (#4999)

---
 .../ngraph/runtime/reference/convolution.hpp  |   23 +-
 .../reference/convolution_backprop_data.hpp   |  589 +++++---
 .../group_convolution_backprop_data.hpp       |   25 +-
 ngraph/test/CMakeLists.txt                    |    1 +
 .../test/backend/convolution_backprop.in.cpp  | 1221 +++++++++++++++++
 .../runtime/interpreter/evaluates_map.cpp     |    3 +-
 6 files changed, 1615 insertions(+), 247 deletions(-)
 create mode 100644 ngraph/test/backend/convolution_backprop.in.cpp

diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/convolution.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/convolution.hpp
index e5e12bac9acbc9..adee512d975c2b 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/convolution.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/convolution.hpp
@@ -18,8 +18,6 @@
 #include "ngraph/runtime/reference/split.hpp"
 #include "ngraph/util.hpp"
 
-// can't be removed currently due to arm-plugin dependency
-#include "ngraph/runtime/reference/convolution_backprop_data.hpp"
 namespace ngraph
 {
     namespace runtime
@@ -42,15 +40,18 @@ namespace ngraph
                     std::vector<int> dilation;
                     std::vector<int> pads_begin;
                     std::vector<int> pads_end;
+                    std::vector<int> output_padding;
 
                     ConvolutionParams(const Strides& strides_,
                                       const Strides& dilation_,
                                       const CoordinateDiff& pads_begin_,
-                                      const CoordinateDiff& pads_end_)
+                                      const CoordinateDiff& pads_end_,
+                                      const CoordinateDiff& output_padding_ = {0, 0, 0})
                         : strides{strides_.begin(), strides_.end()}
                         , dilation{dilation_.begin(), dilation_.end()}
                         , pads_begin{pads_begin_.begin(), pads_begin_.end()}
-                        , pads_end{pads_end_.begin(), pads_end_.end()} {};
+                        , pads_end{pads_end_.begin(), pads_end_.end()}
+                        , output_padding{output_padding_.begin(), output_padding_.end()} {};
                 };
 
                 template <typename Int>
@@ -86,15 +87,18 @@ namespace ngraph
                     const size_t filter_channel_size = shape_size(filter_channel_shape);
 
                     for (int i_z = -p.pads_begin[0];
-                         i_z <= (p.pads_end[0] + input_size_z - dilated_filter_size_z);
+                         i_z <= (p.pads_end[0] + input_size_z - dilated_filter_size_z +
+                                 p.output_padding[0]);
                          i_z += p.strides[0])
                     {
                         for (int i_y = -p.pads_begin[1];
-                             i_y <= (p.pads_end[1] + input_size_y - dilated_filter_size_y);
+                             i_y <= (p.pads_end[1] + input_size_y - dilated_filter_size_y +
+                                     p.output_padding[1]);
                              i_y += p.strides[1])
                         {
                             for (int i_x = -p.pads_begin[2];
-                                 i_x <= (p.pads_end[2] + input_size_x - dilated_filter_size_x);
+                                 i_x <= (p.pads_end[2] + input_size_x - dilated_filter_size_x +
+                                         p.output_padding[2]);
                                  i_x += p.strides[2])
                             {
                                 auto input_channel = batch;
@@ -154,6 +158,8 @@ namespace ngraph
                             std::prev(p.pads_begin.end(), spatial_rank), missing_dims, 0);
                         p.pads_end.insert(
                             std::prev(p.pads_end.end(), spatial_rank), missing_dims, 0);
+                        p.output_padding.insert(
+                            std::prev(p.output_padding.end(), spatial_rank), missing_dims, 0);
                         in_shape.insert(std::next(in_shape.end(), -spatial_rank), missing_dims, 1);
                         filter_shape.insert(
                             std::prev(filter_shape.end(), spatial_rank), missing_dims, 1);
@@ -324,3 +330,6 @@ namespace ngraph
         } // namespace reference
     }     // namespace runtime
 } // namespace ngraph
+
+// can't be removed currently due to arm-plugin dependency
+#include "ngraph/runtime/reference/convolution_backprop_data.hpp"
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/convolution_backprop_data.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/convolution_backprop_data.hpp
index 3fa325f7726e05..1c755198163a7a 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/convolution_backprop_data.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/convolution_backprop_data.hpp
@@ -10,11 +10,7 @@
 #include <numeric>
 
 #include "ngraph/axis_vector.hpp"
-#include "ngraph/coordinate_transform.hpp"
-#include "ngraph/runtime/reference/concat.hpp"
-#include "ngraph/runtime/reference/helpers.hpp"
-#include "ngraph/runtime/reference/reverse.hpp"
-#include "ngraph/runtime/reference/split.hpp"
+#include "ngraph/runtime/reference/convolution.hpp"
 #include "ngraph/util.hpp"
 
 namespace ngraph
@@ -23,217 +19,302 @@ namespace ngraph
     {
         namespace reference
         {
-            // in: NC_I...
-            // filter: C_OC_I...
-            // out: NC_O...
-            template <typename INPUT,
-                      typename FILTER,
-                      typename OUTPUT,
-                      typename ACCUMULATION = typename widen<OUTPUT>::type>
-            void convolution_backprop_impl(const INPUT* in,
-                                           const FILTER* filter,
-                                           OUTPUT* out,
-                                           const Shape& in_shape,
-                                           const Shape& filter_shape,
-                                           const Shape& out_shape,
-                                           const Strides& stride,
-                                           const Strides& filter_dilation,
-                                           const CoordinateDiff& in_pad_below,
-                                           const CoordinateDiff& in_pad_above,
-                                           const Strides& in_dilation,
-                                           size_t in_batch_axis,
-                                           size_t in_channel_axis,
-                                           size_t filter_out_channel_axis,
-                                           size_t filter_in_channel_axis,
-                                           size_t out_batch_axis,
-                                           size_t out_channel_axis)
+            namespace
             {
-                auto old_mode = std::fegetround();
-                std::fesetround(FE_TONEAREST);
-                // Comments throughout assume without loss of generality that:
-                //
-                // * batch axes for both in and out are 0
-                // * in channel axes for both in and filter are 1
-                // * out channel axes for filter is 0
-                // * out channel axis for out is 1
-
-                // At the outermost level we will walk over every out coordinate O.
-                CoordinateTransform out_transform(out_shape);
-
-                for (const Coordinate& out_coord : out_transform)
+                constexpr size_t filter_input_ch_axis = 0;
+
+                template <typename T>
+                void extend_with_zeros(const Strides& strides,
+                                       const Shape& input_shape,
+                                       const T* in,
+                                       Shape& output_shape,
+                                       std::vector<T>& input_zeros)
                 {
-                    // Our out coordinate O will have the form:
-                    //
-                    //   (N,chan_out,i_1,...,i_n)
-
-                    size_t batch_index = out_coord[out_batch_axis];
-                    size_t out_channel = out_coord[out_channel_axis];
-
-                    // For the in we need to iterate the coordinate:
-                    //
-                    //   I:
-                    //
-                    // over the range (noninclusive on the right):
-                    //
-                    //   (N,0,s_1*i_1,s_2*i_2,...,s_n*i_n) ->
-                    //
-                    //     (N+1,
-                    //      chans_in_count,
-                    //      s_1*i_1+ l_1*filter_dims_1,
-                    ///       ...,
-                    ///     s_n*i_n +l_n*filter_dims_n)
-                    //
-                    // with strides:
-                    //
-                    //   (1,l_1,...,l_n).
-                    //
-                    // Note that we are iterating within the *padded* and *dilated* in batch, so
-                    // further down we must check the current coordinate is in the pad or dilation
-                    // gap.
-
-                    size_t n_spatial_dimensions = in_shape.size() - 2;
-                    size_t n_in_channels = in_shape[in_channel_axis];
-
-                    Coordinate in_transform_start(2 + n_spatial_dimensions);
-                    Coordinate in_transform_end(2 + n_spatial_dimensions);
-                    Strides in_transform_movement_strides(2 + n_spatial_dimensions, 1);
-                    CoordinateDiff in_transform_pad_below(2 + n_spatial_dimensions, 0);
-                    CoordinateDiff in_transform_pad_above(2 + n_spatial_dimensions, 0);
-                    Strides in_transform_dilation_strides(2 + n_spatial_dimensions, 1);
-
-                    in_transform_start[in_batch_axis] = batch_index;
-                    in_transform_end[in_batch_axis] = batch_index + 1;
-                    in_transform_start[in_channel_axis] = 0;
-                    in_transform_end[in_channel_axis] = 1;
-
-                    for (size_t i = 2; i < n_spatial_dimensions + 2; i++)
+                    std::vector<int> input_3d(3, 1);
+                    std::vector<int> strides_3d(3, 1);
+                    std::vector<int> output_3d(3, 1);
+
+                    for (size_t i = 0; i < strides.size(); ++i)
                     {
-                        size_t filter_dilation_stride = filter_dilation[i - 2];
-                        size_t filter_movement_stride = stride[i - 2];
-                        std::ptrdiff_t below_pad = in_pad_below[i - 2];
-                        std::ptrdiff_t above_pad = in_pad_above[i - 2];
-                        size_t in_dilation_stride = in_dilation[i - 2];
-
-                        in_transform_start[i] = filter_movement_stride * out_coord[i];
-                        in_transform_end[i] = in_transform_start[i] +
-                                              (filter_shape[i] - 1) * filter_dilation_stride + 1;
-                        in_transform_movement_strides[i] = filter_dilation_stride;
-                        in_transform_pad_below[i] = below_pad;
-                        in_transform_pad_above[i] = above_pad;
-                        in_transform_dilation_strides[i] = in_dilation_stride;
+                        output_shape[i + 2] =
+                            input_shape[i + 2] + (strides[i] - 1) * (input_shape[i + 2] - 1);
+                        input_3d[input_3d.size() - strides.size() + i] = input_shape[i + 2];
+                        strides_3d[strides_3d.size() - strides.size() + i] = strides[i];
+                        output_3d[output_3d.size() - strides.size() + i] = output_shape[i + 2];
                     }
 
-                    AxisVector in_transform_axis_order(2 + n_spatial_dimensions);
-                    for (size_t i = 0; i < in_transform_axis_order.size(); i++)
+                    const size_t input_size = shape_size(input_3d);
+                    if (input_size == 1)
                     {
-                        in_transform_axis_order[i] = i;
+                        for (size_t i = 0; i < shape_size(input_shape); ++i)
+                        {
+                            input_zeros.push_back(in[i]);
+                        }
                     }
-                    CoordinateTransform in_transform(in_shape,
-                                                     in_transform_start,
-                                                     in_transform_end,
-                                                     in_transform_movement_strides,
-                                                     in_transform_axis_order,
-                                                     in_transform_pad_below,
-                                                     in_transform_pad_above,
-                                                     in_transform_dilation_strides);
-
-                    // Simultaneously with iterating I, for the filter we need to iterate the
-                    // coordinate:
-                    //
-                    //   F
-                    //
-                    // over the range (noninclusive on the right):
-                    //
-                    //   (chan_out,0,0,...,0) ->
-                    //     (chan_out+1,
-                    //      chans_in_count,
-                    //      filter_dims_1,
-                    //        ...,
-                    //      filter_dims_n)
-                    //
-                    // with unit stride.
-
-                    Shape filter_transform_start(2 + n_spatial_dimensions);
-                    Shape filter_transform_end(2 + n_spatial_dimensions);
-
-                    filter_transform_start[filter_out_channel_axis] = out_channel;
-                    filter_transform_end[filter_out_channel_axis] = out_channel + 1;
-                    filter_transform_start[filter_in_channel_axis] = 0;
-                    filter_transform_end[filter_in_channel_axis] = 1;
-
-                    for (size_t i = 2; i < n_spatial_dimensions + 2; i++)
+                    else
                     {
-                        filter_transform_start[i] = 0;
-                        filter_transform_end[i] = filter_shape[i];
+                        for (size_t batch = 0; batch < input_shape[0]; ++batch)
+                        {
+                            const auto offset_batch = batch * input_size * input_shape[1];
+                            for (size_t channel = 0; channel < input_shape[1]; ++channel)
+                            {
+                                const auto offset_channel = offset_batch + channel * input_size;
+                                for (int i_z = 0; i_z < input_3d[0]; ++i_z)
+                                {
+                                    const auto offset_i_z = i_z * input_3d[2] * input_3d[1];
+                                    for (int i_y = 0; i_y < input_3d[1]; ++i_y)
+                                    {
+                                        const auto offset_i_y = i_y * input_3d[2];
+                                        for (int i_x = 0; i_x < input_3d[2]; ++i_x)
+                                        {
+                                            input_zeros.push_back(
+                                                in[offset_channel + i_x + offset_i_y + offset_i_z]);
+
+                                            if (i_x < input_3d[2] - 1)
+                                            {
+                                                for (int k = 0; k < strides_3d[2] - 1; k++)
+                                                {
+                                                    input_zeros.push_back(0);
+                                                }
+                                            }
+                                        }
+
+                                        if (i_y < input_3d[1] - 1)
+                                        {
+                                            const auto new_size =
+                                                output_3d[2] * (strides_3d[1] - 1);
+                                            input_zeros.insert(input_zeros.begin() +
+                                                                   input_zeros.size(),
+                                                               new_size,
+                                                               0);
+                                        }
+                                    }
+
+                                    if (i_z < input_3d[0] - 1)
+                                    {
+                                        const auto new_size =
+                                            output_3d[1] * output_3d[2] * (strides_3d[0] - 1);
+                                        input_zeros.insert(
+                                            input_zeros.begin() + input_zeros.size(), new_size, 0);
+                                    }
+                                }
+                            }
+                        }
                     }
+                }
 
-                    CoordinateTransform filter_transform(
-                        filter_shape, filter_transform_start, filter_transform_end);
-
-                    // As we go, we sum up:
-                    //
-                    //   out[O] += in[I] * filter[F].
+                void infer_forward_convbackprop_output_shape(const Shape& in_spatial_shape,
+                                                             const Shape& f_spatial_shape,
+                                                             const Shape& out_spatial_shape,
+                                                             Shape& infer_spatial_shape,
+                                                             const Strides& strides,
+                                                             const Strides& dilations,
+                                                             const CoordinateDiff& output_padding)
+                {
+                    for (size_t idx = 0; idx < in_spatial_shape.size(); idx++)
+                    {
+                        int total_padding = strides[idx] * (in_spatial_shape[idx] - 1) +
+                                            dilations[idx] * (f_spatial_shape[idx] - 1) + 1 -
+                                            out_spatial_shape[idx] + output_padding[idx];
+                        size_t padded_dim = std::max<size_t>(total_padding, 0);
+                        size_t filter_dilated_dim = dilations[idx] * (f_spatial_shape[idx] - 1) + 1;
+                        size_t out_spatial_dim = (in_spatial_shape[idx] - 1) * strides[idx] +
+                                                 filter_dilated_dim - padded_dim +
+                                                 output_padding[idx];
+                        infer_spatial_shape.push_back(out_spatial_dim);
+                    }
+                }
 
-                    ACCUMULATION result = 0;
+                void validate_convolution_backprop_parameters(const Shape& in_shape,
+                                                              const Shape& f_shape,
+                                                              const Shape& out_shape,
+                                                              const Strides& strides,
+                                                              const Strides& dilations,
+                                                              const CoordinateDiff& pads_begin,
+                                                              const CoordinateDiff& pads_end,
+                                                              const CoordinateDiff& output_padding)
+                {
+                    // this implementation supports 1D, 2D and 3D convolutions
+                    NGRAPH_CHECK(in_shape.size() >= 3 && in_shape.size() <= 5,
+                                 "Unsupported input rank: ",
+                                 in_shape);
+
+                    NGRAPH_CHECK(in_shape.size() == f_shape.size(),
+                                 "Incompatible input ranks: ",
+                                 in_shape.size(),
+                                 " and ",
+                                 f_shape.size());
+
+                    NGRAPH_CHECK(in_shape[in_channel_axis] == f_shape[filter_input_ch_axis],
+                                 "Incompatible input channels in data batch and filters shapes: ",
+                                 in_shape[in_channel_axis],
+                                 " and ",
+                                 f_shape[filter_input_ch_axis]);
+
+                    NGRAPH_CHECK(in_shape.size() == out_shape.size(),
+                                 "Incompatible input and output ranks: ",
+                                 in_shape.size(),
+                                 " and ",
+                                 out_shape.size());
+
+                    const auto spatial_dims = in_shape.size() - 2;
+                    NGRAPH_CHECK(strides.size() == spatial_dims,
+                                 "Strides not definied for all and only spatial dimensions.");
+
+                    NGRAPH_CHECK(dilations.size() == spatial_dims,
+                                 "Dilations not defined for all and only spatial dimensions.");
+
+                    NGRAPH_CHECK((pads_begin.size() == pads_end.size()) &&
+                                     (pads_begin.size() == spatial_dims),
+                                 "Pads not defined for all and only spatial dimensions.");
+
+                    NGRAPH_CHECK(!output_padding.empty() && output_padding.size() == spatial_dims,
+                                 "Output padding not defined for all and only spatial dimensions.");
+
+                    Shape out_spatial_shape{std::next(out_shape.begin(), 2), std::end(out_shape)};
+                    Shape infered_out_spatial_shape{};
+                    infer_forward_convbackprop_output_shape(
+                        Shape{std::next(in_shape.begin(), 2), std::end(in_shape)},
+                        Shape{std::next(f_shape.begin(), 2), std::end(f_shape)},
+                        Shape{std::next(out_shape.begin(), 2), std::end(out_shape)},
+                        infered_out_spatial_shape,
+                        strides,
+                        dilations,
+                        output_padding);
+                    NGRAPH_CHECK(out_spatial_shape == infered_out_spatial_shape,
+                                 "Incorrect output shape provided");
+                }
+            } // namespace
 
-                    CoordinateTransform::Iterator in_it = in_transform.begin();
-                    CoordinateTransform::Iterator filter_it = filter_transform.begin();
-                    CoordinateTransform::Iterator in_it_end = in_transform.end();
-                    CoordinateTransform::Iterator filter_it_end = filter_transform.end();
+            template <typename T>
+            void convolution_backprop_impl(const T* in,
+                                           const T* f,
+                                           T* out,
+                                           const Shape& in_shape,
+                                           const Shape& f_shape,
+                                           const Shape& out_shape,
+                                           const Strides& strides,
+                                           const Strides& dilation,
+                                           const CoordinateDiff& pads_begin,
+                                           const CoordinateDiff& pads_end,
+                                           const CoordinateDiff& output_padding)
 
-                    size_t in_channel_stride = row_major_strides(in_shape).at(in_channel_axis);
-                    size_t filter_in_channel_stride =
-                        row_major_strides(filter_shape).at(filter_in_channel_axis);
+            {
+                // here we are converting all param types to int's to avoid arithmetic issues
+                // (e.g signed + unsigned) in indexes calculation later
+                ConvolutionParams params{strides, dilation, pads_begin, pads_end, output_padding};
+
+                // here we are extending spatial dimensions to 3D, because we are going to use 3D
+                // convolution implementation to convolve also in 1D & 2D case
+                Shape input_shape{in_shape};
+                Shape filters_shape{f_shape};
+                if (in_shape.size() < 5)
+                {
+                    extend_to_3D(params, input_shape, filters_shape);
+                }
 
-                    while (in_it != in_it_end && filter_it != filter_it_end)
+                for (size_t i = 0; i < input_shape.size() - 2; ++i)
+                {
+                    if (input_shape[i + 2] > 1 || filters_shape[i + 2] > 1)
                     {
-                        const Coordinate& in_coord = *in_it;
-                        if (in_transform.has_source_coordinate(in_coord))
-                        {
-                            size_t in_idx = in_transform.index(in_coord);
-                            const Coordinate& filter_coord = *filter_it;
-                            size_t filter_idx = filter_transform.index(filter_coord);
-                            for (size_t in_channel = 0; in_channel < n_in_channels; ++in_channel)
-                            {
-                                ACCUMULATION in_v = static_cast<ACCUMULATION>(in[in_idx]);
-                                ACCUMULATION f_v = static_cast<ACCUMULATION>(filter[filter_idx]);
+                        params.pads_begin[i] = filters_shape[i + 2] - params.pads_begin[i] - 1;
+                        params.pads_end[i] = filters_shape[i + 2] - params.pads_end[i] - 1;
+                    }
+                    else
+                    {
+                        params.pads_begin[i] = 0;
+                        params.pads_end[i] = 0;
+                    }
+                }
 
-                                result += in_v * f_v;
-                                in_idx += in_channel_stride;
-                                filter_idx += filter_in_channel_stride;
-                            }
-                        }
-                        ++in_it;
-                        ++filter_it;
+                // convert output shape to 3D, contains only dimensions
+                Shape out_shape_3d{out_shape.begin() + 2, out_shape.end()};
+
+                int out_shape_rank = out_shape.size() - 2;
+                if (out_shape_rank < 3)
+                {
+                    int missing_dims = 3 - out_shape_rank;
+                    out_shape_3d.insert(
+                        std::prev(out_shape_3d.end(), out_shape_rank), missing_dims, 1);
+                }
+
+                // modify params.pads_end when output_shape was provided in ctor in order to
+                // calculate expected number of output elements
+                for (size_t i = 0; i < out_shape_3d.size(); i++)
+                {
+                    if (out_shape_3d[i] > 1)
+                    {
+                        // expected_dim = (in - 1)* strides + filter - 2*padding + out_padding
+                        // strides is already applied (through 0's extension in input)
+                        // padding = pads_begin + pads_end, formula below is using
+                        // params.pad_begin/params.pads_end:
+                        const size_t expected_dim =
+                            out_shape_3d[i] - ((input_shape[i + 2] - 1) - filters_shape[i + 2] +
+                                               params.pads_begin[i] + params.pads_end[i] + 2 +
+                                               params.output_padding[i]);
+                        params.pads_end[i] += expected_dim;
                     }
+                }
+
+                const size_t filters_count = filters_shape[filter_out_ch_axis];
+                const Shape filter_shape(++filters_shape.begin(), filters_shape.end());
+                const size_t filter_size = shape_size(filter_shape);
+
+                const size_t batches_count = input_shape[in_batch_axis];
+                Shape batch_shape(++input_shape.begin(), input_shape.end());
+                const size_t batch_size = shape_size(batch_shape);
+
+                auto batch = in;
 
-                    out[out_transform.index(out_coord)] = result;
+                for (size_t batch_idx = 0; batch_idx < batches_count; ++batch_idx)
+                {
+                    auto filter = f;
+                    for (size_t f_idx = 0; f_idx < filters_count; ++f_idx)
+                    {
+                        convolve_3D_channels(params, batch, batch_shape, filter, filter_shape, out);
+                        filter += filter_size;
+                    }
+                    batch += batch_size;
                 }
-                std::fesetround(old_mode);
             }
 
-            template <typename OUTPUT,
-                      typename FILTER,
-                      typename INPUT,
-                      typename ACCUMULATION = typename widen<INPUT>::type>
-            void convolution_backprop_in(const OUTPUT* delta_out,
-                                         const FILTER* filter,
-                                         INPUT* delta_in,
-                                         const Shape& out_shape,
-                                         const Shape& filter_shape,
+            template <typename T>
+            void convolution_backprop_in(const T* delta_in,
+                                         const T* filter,
+                                         T* delta_out,
                                          const Shape& in_shape,
+                                         const Shape& filter_shape,
+                                         const Shape& out_shape,
                                          const Strides& in_dilation,
                                          const Strides& filter_dilation,
                                          const CoordinateDiff& forward_in_pad_bellow,
                                          const CoordinateDiff& forward_in_pad_above,
-                                         const Strides& stride)
+                                         const Strides& stride,
+                                         const CoordinateDiff& output_padding)
             {
+                std::vector<T> extended_input;
+                std::vector<T> extended_filter;
+                AxisSet reverse_axes;
+
+                Shape conv_input_shape = in_shape;
+                Shape conv_filter_shape = filter_shape;
+                Strides conv_stride = stride;
+                Strides conv_filter_dilation = filter_dilation;
+                auto conv_input_data = delta_in;
+
+                validate_convolution_backprop_parameters(in_shape,
+                                                         filter_shape,
+                                                         out_shape,
+                                                         stride,
+                                                         filter_dilation,
+                                                         forward_in_pad_bellow,
+                                                         forward_in_pad_above,
+                                                         output_padding);
+
                 // Note that we only reverse the spatial dimensions here (loop
                 // starts at 2)
-                std::vector<INPUT> reversed(shape_size(filter_shape));
-                AxisSet reverse_axes;
-                size_t reverse_axes_start = 2;
-                for (size_t i = reverse_axes_start; i < filter_shape.size(); ++i)
+                std::vector<T> reversed(shape_size(filter_shape));
+                for (size_t i = 2; i < filter_shape.size(); ++i)
                 {
                     reverse_axes.insert(i);
                 }
@@ -242,55 +323,109 @@ namespace ngraph
                         filter_shape,
                         filter_shape,
                         reverse_axes,
-                        sizeof(FILTER));
-                size_t filter_out_channel_axis = 1;
-                size_t filter_in_channel_axis = 0;
+                        sizeof(T));
 
-                // Compute backward pad out pad bellow
-                size_t spatial_dim_count = in_shape.size() - 2;
+                auto conv_filter_data = &reversed[0];
 
-                CoordinateDiff backward_delta_out_pad_below;
-                backward_delta_out_pad_below.resize(spatial_dim_count);
+                // if channel number for output is > 1 then reverse layout of filter coefficients as
+                // it is required by convolve_3D_channels() function.
+                // Current layout:
+                // batch0_ch0|batch0_ch1|...|batch0_chN|...|batch1_ch0|batch1_ch1|...|batch1_chN|...
+                // Expected layout:
+                // batch0_ch0|batch1_ch0|...|batchN_ch0|...|batch0_ch1|batch1_ch1|...|batch1_chN|...
+                if (filter_shape[1] > 1)
+                {
+                    std::vector<T> temp_reversed(reversed);
+                    const Shape filter_dim_shape(filter_shape.begin() + 2, filter_shape.end());
+                    const size_t filter_size = shape_size(filter_dim_shape);
 
-                for (size_t i = 0; i < spatial_dim_count; i++)
+                    for (size_t i = 0; i < filter_shape[1]; i++)
+                    {
+                        for (size_t j = 0; j < filter_shape[0]; j++)
+                        {
+                            const auto delta = temp_reversed.begin() +
+                                               j * filter_shape[1] * filter_size + i * filter_size;
+                            const auto out = reversed.begin() + i * filter_shape[0] * filter_size +
+                                             j * filter_size;
+                            std::copy(delta, delta + filter_size, out);
+                        }
+                    }
+                }
+
+                // swap filter batch and channels
+                std::iter_swap(conv_filter_shape.begin(), conv_filter_shape.begin() + 1);
+
+                // extend stride and filter inputs with zero padding for stride and filter_dilation
+                // > 1, after that set stride and filter params to 1.
+                const size_t stride_dim =
+                    std::accumulate(stride.begin(), stride.end(), 1, std::multiplies<size_t>());
+                if (stride_dim >= 2)
                 {
-                    backward_delta_out_pad_below[i] =
-                        (static_cast<ptrdiff_t>(filter_shape[i + 2]) - 1) * filter_dilation[i] -
-                        forward_in_pad_bellow[i];
+                    extend_with_zeros(stride, in_shape, delta_in, conv_input_shape, extended_input);
+                    std::fill(conv_stride.begin(), conv_stride.end(), 1);
+                    conv_input_data = &extended_input[0];
                 }
-                // Compute backward pad out pad above
-                CoordinateDiff backward_delta_out_pad_above;
-                backward_delta_out_pad_above.resize(spatial_dim_count);
 
-                for (size_t i = 0; i < spatial_dim_count; i++)
+                const size_t dilation_dim = std::accumulate(
+                    filter_dilation.begin(), filter_dilation.end(), 1, std::multiplies<size_t>());
+                if (dilation_dim >= 2)
                 {
-                    backward_delta_out_pad_above[i] =
-                        (static_cast<ptrdiff_t>(filter_shape[i + 2]) - 1) * filter_dilation[i] +
-                        ((forward_in_pad_bellow[i] + ((in_shape[i + 2]) - 1) * in_dilation[i] +
-                          forward_in_pad_above[i] -
-                          (static_cast<ptrdiff_t>(filter_shape[i + 2]) - 1) * filter_dilation[i]) %
-                         stride[i]) -
-                        forward_in_pad_above[i];
+                    extend_with_zeros<T>(filter_dilation,
+                                         filter_shape,
+                                         reinterpret_cast<const T*>(&reversed[0]),
+                                         conv_filter_shape,
+                                         extended_filter);
+                    std::fill(conv_filter_dilation.begin(), conv_filter_dilation.end(), 1);
+                    conv_filter_data = &extended_filter[0];
                 }
 
-                convolution_backprop_impl<OUTPUT, FILTER, INPUT, ACCUMULATION>(
-                    delta_out,
-                    &reversed[0],
-                    delta_in,
-                    out_shape,
-                    filter_shape,
-                    in_shape,
-                    in_dilation,
-                    filter_dilation,
-                    backward_delta_out_pad_below,
-                    backward_delta_out_pad_above,
-                    stride,
-                    0,
-                    1,
-                    filter_out_channel_axis,
-                    filter_in_channel_axis,
-                    0,
-                    1);
+                convolution_backprop_impl(conv_input_data,
+                                          conv_filter_data,
+                                          delta_out,
+                                          conv_input_shape,
+                                          conv_filter_shape,
+                                          out_shape,
+                                          conv_stride,
+                                          conv_filter_dilation,
+                                          forward_in_pad_bellow,
+                                          forward_in_pad_above,
+                                          output_padding);
+            }
+
+            // DEPRECATED, can't be removed currently due to arm-plugin dependency
+            template <typename OUTPUT,
+                      typename FILTER,
+                      typename INPUT,
+                      typename ACCUMULATION = typename widen<INPUT>::type>
+            NGRAPH_DEPRECATED(
+                "convolution_backprop_in function with 4 template types is deprecated, use "
+                "function with 1 template and output_padding parameter.")
+            void convolution_backprop_in(const INPUT* delta_in,
+                                         const FILTER* filter,
+                                         OUTPUT* delta_out,
+                                         const Shape& in_shape,
+                                         const Shape& filter_shape,
+                                         const Shape& out_shape,
+                                         const Strides& in_dilation,
+                                         const Strides& filter_dilation,
+                                         const CoordinateDiff& forward_in_pad_bellow,
+                                         const CoordinateDiff& forward_in_pad_above,
+                                         const Strides& stride)
+            {
+                const ngraph::CoordinateDiff output_padding(in_shape.size() - 2, 0);
+
+                convolution_backprop_in(delta_in,
+                                        filter,
+                                        delta_out,
+                                        in_shape,
+                                        filter_shape,
+                                        out_shape,
+                                        in_dilation,
+                                        filter_dilation,
+                                        forward_in_pad_bellow,
+                                        forward_in_pad_above,
+                                        stride,
+                                        output_padding);
             }
         } // namespace reference
     }     // namespace runtime
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/group_convolution_backprop_data.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/group_convolution_backprop_data.hpp
index b70c0d3ed9ab9d..306ddf047bfad6 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/group_convolution_backprop_data.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/group_convolution_backprop_data.hpp
@@ -178,23 +178,24 @@ namespace ngraph
                 const size_t group_out_size = shape_size(group_out_shape);
 
                 Strides in_dilation(in_shape.size(), 1);
+                const ngraph::CoordinateDiff output_padding(in_shape.size() - 2, 0);
                 for (size_t batch_idx = 0; batch_idx < in_shape[in_batch_axis]; ++batch_idx)
                 {
                     group_filter = f;
                     for (size_t group_idx = 0; group_idx < group_count; ++group_idx)
                     {
-                        runtime::reference::convolution_backprop_in<INPUT, FILTER, OUTPUT, ACCU>(
-                            group_batch,
-                            group_filter,
-                            group_out,
-                            group_batch_shape,
-                            group_filter_shape,
-                            group_out_shape,
-                            in_dilation,
-                            dilation,
-                            pads_begin,
-                            pads_end,
-                            strides);
+                        runtime::reference::convolution_backprop_in(group_batch,
+                                                                    group_filter,
+                                                                    group_out,
+                                                                    group_batch_shape,
+                                                                    group_filter_shape,
+                                                                    group_out_shape,
+                                                                    in_dilation,
+                                                                    dilation,
+                                                                    pads_begin,
+                                                                    pads_end,
+                                                                    strides,
+                                                                    output_padding);
                         group_batch += group_batch_size;
                         group_filter += group_filter_size;
                         group_out += group_out_size;
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 5ca3049b6e2095..c3c5ab80405e23 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -356,6 +356,7 @@ set(MULTI_TEST_SRC
     backend/constant.in.cpp
     backend/convert.in.cpp
     backend/convert_like.in.cpp
+    backend/convolution_backprop.in.cpp
     backend/convolution.in.cpp
     backend/binary_convolution.in.cpp
     backend/clamp.in.cpp
diff --git a/ngraph/test/backend/convolution_backprop.in.cpp b/ngraph/test/backend/convolution_backprop.in.cpp
new file mode 100644
index 00000000000000..0426999595656a
--- /dev/null
+++ b/ngraph/test/backend/convolution_backprop.in.cpp
@@ -0,0 +1,1221 @@
+//*****************************************************************************
+// Copyright 2017-2021 Intel Corporation
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//*****************************************************************************
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "ngraph/runtime/tensor.hpp"
+#include "runtime/backend.hpp"
+#include "util/all_close.hpp"
+#include "util/all_close_f.hpp"
+#include "util/engine/test_engines.hpp"
+#include "util/known_element_types.hpp"
+#include "util/ndarray.hpp"
+#include "util/test_case.hpp"
+#include "util/test_control.hpp"
+#include "util/test_tools.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+using TestEngine = test::ENGINE_CLASS_NAME(${BACKEND_NAME});
+
+static void ConvolutionBackpropTest(const std::vector<float>& inputs,
+                                    const Shape inputs_shape,
+                                    const std::vector<float>& filters,
+                                    const Shape filter_shape,
+                                    const std::vector<float>& outputs,
+                                    const Shape outputs_shape,
+                                    const Strides& strides,
+                                    const CoordinateDiff& padding,
+                                    const Strides& dilations,
+                                    const CoordinateDiff& output_padding)
+{
+    const CoordinateDiff pads_begin{padding};
+    const CoordinateDiff pads_end{padding};
+    const op::PadType auto_pad{op::PadType::EXPLICIT};
+    const CoordinateDiff out_padding{output_padding};
+
+    auto inputs_param = make_shared<op::Parameter>(element::f32, inputs_shape);
+    auto filters_param = make_shared<op::Parameter>(element::f32, filter_shape);
+    auto conv = make_shared<op::v1::ConvolutionBackpropData>(inputs_param,
+                                                             filters_param,
+                                                             strides,
+                                                             pads_begin,
+                                                             pads_end,
+                                                             dilations,
+                                                             auto_pad,
+                                                             out_padding);
+    auto f = make_shared<Function>(conv, ParameterVector{inputs_param, filters_param});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_input<float>(inputs);
+    test_case.add_input<float>(filters);
+    test_case.add_expected_output<float>(outputs_shape, outputs);
+    test_case.run();
+}
+
+// --------------------- 1D convolution ------------------------------------------
+// clang-format off
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_1D_1batch_1channel)
+{
+    const Strides strides{1};
+    const CoordinateDiff padding{0};
+    const Strides dilations{1};
+    const CoordinateDiff output_padding{0};
+
+    const Shape inputs_shape{1, 1, 4};
+    const std::vector<float> inputs{5.0f, 6.0f, 7.0f, 2.0f};
+
+    const Shape filter_shape{1, 1, 3};
+    const std::vector<float> filters{2.0f, 0.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 6};
+    const std::vector<float> outputs{10.0f, 12.0f, 19.0f, 10.0f, 7.0f, 2.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_1D_1batch_1channel_padding)
+{
+    const Strides strides{1};
+    const CoordinateDiff padding{1};
+    const Strides dilations{1};
+    const CoordinateDiff output_padding{0};
+
+    const Shape inputs_shape{1, 1, 4};
+    const std::vector<float> inputs{5.0f, 6.0f, 7.0f, 2.0f};
+
+    const Shape filter_shape{1, 1, 3};
+    const std::vector<float> filters{2.0f, 0.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 4};
+    const std::vector<float> outputs{12.0f, 19.0f, 10.0f, 7.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_1D_1batch_1channel_stride)
+{
+    const Strides strides{2};
+    const CoordinateDiff padding{0};
+    const Strides dilations{1};
+    const CoordinateDiff output_padding{0};
+
+    const Shape inputs_shape{1, 1, 2};
+    const std::vector<float> inputs{5.0f, 7.0f};
+
+    const Shape filter_shape{1, 1, 3};
+    const std::vector<float> filters{2.0f, 0.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 5};
+    const std::vector<float> outputs{10.0f, 0.0f, 19.0f, 0.0f, 7.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_1D_1batch_1channel_output_padding)
+{
+    const Strides strides{1};
+    const CoordinateDiff padding{1};
+    const Strides dilations{1};
+    const CoordinateDiff output_padding{1};
+
+    const Shape inputs_shape{1, 1, 4};
+    const std::vector<float> inputs{5.0f, 6.0f, 7.0f, 2.0f};
+
+    const Shape filter_shape{1, 1, 3};
+    const std::vector<float> filters{2.0f, 0.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 5};
+    const std::vector<float> outputs{12.0f, 19.0f, 10.0f, 7.0f, 2.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_1D_1batch_1channel_dilation)
+{
+    const Strides strides{1};
+    const CoordinateDiff padding{0};
+    const Strides dilations{2};
+    const CoordinateDiff output_padding{0};
+
+    const Shape inputs_shape{1, 1, 3};
+    const std::vector<float> inputs{8.0f, 5.0f, 1.0f};
+
+    const Shape filter_shape{1, 1, 3};
+    const std::vector<float> filters{2.0f, 0.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 7};
+    const std::vector<float> outputs{16.0f, 10.0f, 2.0f, 0.0f, 8.0f, 5.0f, 1.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_1D_1batch_1channel_padding_stride_dilation)
+{
+    const Strides strides{2};
+    const CoordinateDiff padding{2};
+    const Strides dilations{2};
+    const CoordinateDiff output_padding{0};
+
+    const Shape inputs_shape{1, 1, 4};
+    const std::vector<float> inputs{3.0f, 9.0f, 1.0f, 2.0f};
+
+    const Shape filter_shape{1, 1, 3};
+    const std::vector<float> filters{2.0f, 0.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 7};
+    const std::vector<float> outputs{18.0f, 0.0f, 5.0f, 0.0f, 13.0f, 0.0f, 1.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_1D_1batch_2channel)
+{
+    const Strides strides{1};
+    const CoordinateDiff padding{0};
+    const Strides dilations{1};
+    const CoordinateDiff output_padding{0};
+
+    const Shape inputs_shape{1, 1, 2};
+    const std::vector<float> inputs{10.0f, 3.0f};
+
+    const Shape filter_shape{1, 2, 3};
+    const std::vector<float> filters{
+                                    // channel 1
+                                    2.0f, 0.0f, 1.0f,
+                                    // channel 2
+                                    1.0f, 0.0f, 2.0f};
+
+    const Shape outputs_shape{1, 2, 4};
+    const std::vector<float> outputs{
+                                    // channel 1
+                                    20.0f, 6.0f, 10.0f, 3.0f,
+                                    // channel 2
+                                    10.0f, 3.0f, 20.0f, 6.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_1D_1batch_2filter)
+{
+    const Strides strides{1};
+    const CoordinateDiff padding{0};
+    const Strides dilations{1};
+    const CoordinateDiff output_padding{0};
+
+    const Shape inputs_shape{1, 2, 2};
+    const std::vector<float> inputs{
+                                    // channel 1
+                                    4.0f, 7.0f,
+                                    // channel 2
+                                    5.0f, 5.0f};
+
+    const Shape filter_shape{2, 1, 3};
+    const std::vector<float> filters{
+                                    // filter 1
+                                    2.0f, 0.0f, 1.0f,
+                                    // filter 2
+                                    1.0f, 0.0f, 2.0f};
+
+    const Shape outputs_shape{1, 1, 4};
+    const std::vector<float> outputs{13.0f, 19.0f, 14.0f, 17.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_1D_2batch_1channel)
+{
+    const Strides strides{1};
+    const CoordinateDiff padding{0};
+    const Strides dilations{1};
+    const CoordinateDiff output_padding{0};
+
+    const Shape inputs_shape{2, 1, 2};
+    const std::vector<float> inputs{
+                                    // batch 1
+                                    1.0f, 3.0f,
+                                    // batch 2
+                                    2.0f, 2.0f};
+
+    const Shape filter_shape{1, 1, 3};
+    const std::vector<float> filters{2.0f, 0.0f, 1.0f};
+
+    const Shape outputs_shape{2, 1, 4};
+    const std::vector<float> outputs{
+                                    // batch 1
+                                    2.0f, 6.0f, 1.0f, 3.0f,
+                                    // batch 2
+                                    4.0f, 4.0f, 2.0f, 2.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+// --------------------- 2D convolution ------------------------------------------
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_2D_1batch_1channel)
+{
+    const Strides strides{1, 1};
+    const CoordinateDiff padding{0, 0};
+    const Strides dilations{1, 1};
+    const CoordinateDiff output_padding{0, 0};
+
+    const Shape inputs_shape{1, 1, 2, 2};
+    const std::vector<float> inputs{1.0f, 3.0f,
+                                    7.0f, 5.0f};
+
+    const Shape filter_shape{1, 1, 3, 3};
+    const std::vector<float> filters{1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    3.0f, 2.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 4, 4};
+    const std::vector<float> outputs{1.0f, 5.0f, 9.0f, 9.0f,
+                                    7.0f, 20.0f, 34.0f, 15.0f,
+                                    3.0f, 18.0f, 12.0f, 3.0f,
+                                    21.0f, 29.0f, 17.0f, 5.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_2D_1batch_1channel_output_padding)
+{
+    const Strides strides{1, 1};
+    const CoordinateDiff padding{1, 1};
+    const Strides dilations{1, 1};
+    const CoordinateDiff output_padding{1, 1};
+
+    const Shape inputs_shape{1, 1, 2, 2};
+    const std::vector<float> inputs{1.0f, 3.0f,
+                                    7.0f, 5.0f};
+
+    const Shape filter_shape{1, 1, 3, 3};
+    const std::vector<float> filters{1.0f, 2.0f, 3.0f,
+                                    1.0f, 1.0f, 1.0f,
+                                    3.0f, 2.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 3, 3};
+    const std::vector<float> outputs{23.0f, 35.0f, 18.0f,
+                                    23.0f, 19.0f, 8.0f,
+                                    29.0f, 17.0f, 5.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_2D_1batch_1channel_padding)
+{
+    const Strides strides{1, 1};
+    const CoordinateDiff padding{1, 1};
+    const Strides dilations{1, 1};
+    const CoordinateDiff output_padding{0, 0};
+
+    const Shape inputs_shape{1, 1, 4, 4};
+    const std::vector<float> inputs{1.0f, 3.0f, 5.0f, 7.0f,
+                                    7.0f, 5.0f, 3.0f, 1.0f,
+                                    2.0f, 4.0f, 6.0f, 8.0f,
+                                    8.0f, 6.0f, 4.0f, 2.0f};
+
+    const Shape filter_shape{1, 1, 3, 3};
+    const std::vector<float> filters{1.0f, 2.0f, 3.0f,
+                                     0.0f, 1.0f, 0.0f,
+                                     2.0f, 1.0f, 2.0f};
+
+    const Shape outputs_shape{1, 1, 4, 4};
+    const std::vector<float> outputs{20.0f, 37.0f, 27.0f, 18.0f,
+                                     22.0f, 40.0f, 60.0f, 52.0f,
+                                     41.0f, 69.0f, 49.0f, 31.0f,
+                                     18.0f, 26.0f, 34.0f, 22.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_2D_1batch_1channel_stride)
+{
+    const Strides strides{2, 2};
+    const CoordinateDiff padding{0, 0};
+    const Strides dilations{1, 1};
+    const CoordinateDiff output_padding{0, 0};
+
+    const Shape inputs_shape{1, 1, 2, 2};
+    const std::vector<float> inputs{2.0f, 5.0f,
+                                    4.0f, 3.0f};
+
+    const Shape filter_shape{1, 1, 3, 3};
+    const std::vector<float> filters{1.0f, 2.0f, 3.0f,
+                                     1.0f, 1.0f, 1.0f,
+                                     3.0f, 2.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 5, 5};
+    const std::vector<float> outputs{2.0f, 4.0f, 11.0f, 10.0f, 15.0f,
+                                     2.0f, 2.0f, 7.0f, 5.0f, 5.0f,
+                                     10.0f, 12.0f, 32.0f, 16.0f, 14.0f,
+                                     4.0f, 4.0f, 7.0f, 3.0f, 3.0f,
+                                     12.0f, 8.0f, 13.0f, 6.0f, 3.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_2D_1batch_1channel_dilation)
+{
+    const Strides strides{1, 1};
+    const CoordinateDiff padding{0, 0};
+    const Strides dilations{2, 2};
+    const CoordinateDiff output_padding{0, 0};
+
+    const Shape inputs_shape{1, 1, 2, 2};
+    const std::vector<float> inputs{2.0f, 3.0f,
+                                    4.0f, 3.0f};
+
+    const Shape filter_shape{1, 1, 3, 3};
+    const std::vector<float> filters{1.0f, 2.0f, 3.0f,
+                                     1.0f, 1.0f, 1.0f,
+                                     3.0f, 2.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 6, 6};
+    const std::vector<float> outputs{2.f, 3.f, 4.f, 6.f, 6.f, 9.f,
+                                     4.f, 3.f, 8.f, 6.f, 12.f, 9.f,
+                                     2.f, 3.f, 2.f, 3.f, 2.f, 3.f,
+                                     4.f, 3.f, 4.f, 3.f, 4.f, 3.f,
+                                     6.f, 9.f, 4.f, 6.f, 2.f, 3.f,
+                                     12.f, 9.f, 8.f, 6.f, 4.f, 3.f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_2D_1batch_1channel_padding_strides_dilation)
+{
+    const Strides strides{2, 2};
+    const CoordinateDiff padding{2, 2};
+    const Strides dilations{2, 2};
+    const CoordinateDiff output_padding{0, 0};
+
+    const Shape inputs_shape{1, 1, 3, 3};
+    const std::vector<float> inputs{1.0f, 3.0f, 5.0f,
+                                    7.0f, 5.0f, 3.0f,
+                                    2.0f, 4.0f, 6.0f};
+
+    const Shape filter_shape{1, 1, 3, 3};
+    const std::vector<float> filters{1.0f, 2.0f, 3.0f,
+                                     1.0f, 1.0f, 1.0f,
+                                     3.0f, 2.0f, 1.0f};
+
+    const Shape outputs_shape{1, 1, 5, 5};
+    const std::vector<float> outputs{23.0f, 0.0f, 43.0f, 0.0f, 29.0f,
+                                     0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                     31.0f, 0.0f, 57.0f, 0.0f, 45.0f,
+                                     0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                     35.0f, 0.0f, 38.0f, 0.0f, 21.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_2D_1batch_2channel)
+{
+    const Strides strides{1, 1};
+    const CoordinateDiff padding{0, 0};
+    const Strides dilations{1, 1};
+    const CoordinateDiff output_padding{0, 0};
+
+    const Shape inputs_shape{1, 1, 2, 2};
+    const std::vector<float> inputs{1.0f, 3.0f,
+                                    7.0f, 5.0f};
+
+    const Shape filter_shape{1, 2, 3, 3};
+    const std::vector<float> filters{
+                                    // channel 1
+                                    5.0f, 3.0f, 5.0f,
+                                    1.0f, 3.0f, 1.0f,
+                                    4.0f, 2.0f, 4.0f,
+                                    // channel 2
+                                    -5.0f, 3.0f, 5.0f,
+                                    1.0f, -3.0f, 1.0f,
+                                    4.0f, 2.0f, -4.0f};
+
+    const Shape outputs_shape{1, 2, 4, 4};
+    const std::vector<float> outputs{
+                                    // channel 1
+                                    5.0f, 18.0f, 14.0f, 15.0f,
+                                    36.0f, 52.0f, 60.0f, 28.0f,
+                                    11.0f, 40.0f, 32.0f, 17.0f,
+                                    28.0f, 34.0f, 38.0f, 20.0f,
+                                    // channel 2
+                                    -5.0f, -12.0f, 14.0f, 15.0f,
+                                    -34.0f, -4.0f, 42.0f, 28.0f,
+                                    11.0f, -2.0f, -6.0f, -7.0f,
+                                    28.0f, 34.0f, -18.0f, -20.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_2D_1batch_2filter)
+{
+    const Strides strides{1, 1};
+    const CoordinateDiff padding{0, 0};
+    const Strides dilations{1, 1};
+    const CoordinateDiff output_padding{0, 0};
+
+    const Shape inputs_shape{1, 2, 2, 2};
+    const std::vector<float> inputs{
+                                    // channel 1
+                                    1.0f, 3.0f,
+                                    7.0f, 5.0f,
+                                    // channel 2
+                                    2.0f, 4.0f,
+                                    8.0f, 6.0f};
+
+    const Shape filter_shape{2, 1, 3, 3};
+    const std::vector<float> filters{
+                                    // channel 1
+                                    5.0f, 3.0f, 5.0f,
+                                    1.0f, 3.0f, 1.0f,
+                                    4.0f, 2.0f, 4.0f,
+                                    // channel 2
+                                   -5.0f, 3.0f, 5.0f,
+                                    1.0f, -3.0f, 1.0f,
+                                    4.0f, 2.0f, -4.0f};
+
+    const Shape outputs_shape{1, 1, 4, 4};
+    const std::vector<float> outputs{
+                                     -5.0f, 4.0f, 36.0f, 35.0f,
+                                     -2.0f, 44.0f, 108.0f, 62.0f,
+                                     27.0f, 42.0f, 22.0f, 7.0f,
+                                     60.0f, 74.0f, 18.0f, -4.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_2D_2batch_2filter)
+{
+    const Strides strides{1, 1};
+    const CoordinateDiff padding{0, 0};
+    const Strides dilations{1, 1};
+    const CoordinateDiff output_padding{0, 0};
+
+    const Shape inputs_shape{1, 2, 1, 1};
+    const std::vector<float> inputs{
+                                    // channel 1
+                                    2.0f,
+                                    // channel 2
+                                    3.0f};
+
+    const Shape filter_shape{2, 2, 2, 2};
+    const std::vector<float> filters{
+                                    // batch 0
+                                    // channel 1
+                                    5.0f, 3.0f,
+                                    1.0f, 3.0f,
+                                    // channel 2
+                                   -5.0f, 3.0f,
+                                    1.0f, -3.0f,
+                                    // batch 1
+                                    // channel 1
+                                    5.0f, 3.0f,
+                                    1.0f, 3.0f,
+                                    // channel 2
+                                   -5.0f, 3.0f,
+                                    1.0f, -3.0f};
+
+    const Shape outputs_shape{1, 2, 2, 2};
+    const std::vector<float> outputs{
+                                     25.0f, 15.0f, 5.0f, 15.0f, -25.0f, 15.0f, 5.0f, -15.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_2D_2batch_1channel)
+{
+    const Strides strides{1, 1};
+    const CoordinateDiff padding{0, 0};
+    const Strides dilations{1, 1};
+    const CoordinateDiff output_padding{0, 0};
+
+    const Shape inputs_shape{2, 1, 2, 2};
+    const std::vector<float> inputs{
+                                    // batch 1
+                                    1.0f, 3.0f, 
+                                    1.0f, 3.0f,
+                                    // batch 2
+                                    -1.0f, 3.0f,
+                                    1.0f, 3.0f};
+
+    const Shape filter_shape{1, 1, 3, 3};
+    const std::vector<float> filters{-5.0f, 3.0f, 5.0f,
+                                    1.0f, -3.0f, 1.0f,
+                                    4.0f, 2.0f, -4.0f};
+
+    const Shape outputs_shape{2, 1, 4, 4};
+    const std::vector<float> outputs{
+                                    // batch 1
+                                    -5.0f, -12.0f, 14.0f, 15.0f,
+                                    -4.0f, -12.0f, 6.0f, 18.0f,
+                                    5.0f, 14.0f, -6.0f, -9.0f,
+                                    4.0f, 14.0f, 2.0f, -12.0f,
+                                    // batch 2
+                                    5.0f, -18.0f, 4.0f, 15.0f,
+                                    -6.0f, -6.0f, 4.0f, 18.0f,
+                                    -3.0f, 10.0f, 2.0f, -9.0f,
+                                    4.0f, 14.0f, 2.0f, -12.0f};
+
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+// --------------------- 3D convolution ------------------------------------------
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_3D_1batch_1channel)
+{
+    const Strides strides{1, 1, 1};
+    const CoordinateDiff padding{0, 0, 0};
+    const Strides dilations{1, 1, 1};
+    const CoordinateDiff output_padding{0, 0, 0};
+
+    const Shape inputs_shape{1, 1, 2, 2, 2};
+    const std::vector<float> inputs{
+                                    // depth: 1
+                                    15.0f, 3.0f,
+                                    21.0f, 10.0f,
+                                    // depth: 2
+                                    10.0f, 13.0f,
+                                    11.0f, 17.0f};
+
+    const Shape filter_shape{1, 1, 3, 3, 3};
+    const std::vector<float> filters{
+                                    // depth: 1
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 2
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 3
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f};
+
+    const Shape outputs_shape{1, 1, 4, 4, 4};
+    const std::vector<float> outputs{
+                                    // depth: 1
+                                    15.0f, 33.0f, 51.0f, 9.0f,
+                                    21.0f, 67.0f, 86.0f, 30.0f,
+                                    30.0f, 42.0f, 43.0f, 6.0f,
+                                    42.0f, 41.0f, 52.0f, 20.0f,
+                                    // depth: 2
+                                    25.0f, 66.0f, 107.0f, 48.0f,
+                                    32.0f, 116.0f, 166.0f, 81.0f,
+                                    50.0f, 89.0f, 93.0f, 32.0f,
+                                    64.0f, 86.0f, 91.0f, 54.0f,
+                                    // depth: 3
+                                    25.0f, 66.0f, 107.0f, 48.0f,
+                                    32.0f, 116.0f, 166.0f, 81.0f,
+                                    50.0f, 89.0f, 93.0f, 32.0f,
+                                    64.0f, 86.0f, 91.0f, 54.0f,
+                                    // depth: 4
+                                    10.0f, 33.0f, 56.0f, 39.0f,
+                                    11.0f, 49.0f, 80.0f, 51.0f,
+                                    20.0f, 47.0f, 50.0f, 26.0f,
+                                    22.0f, 45.0f, 39.0f, 34.0f
+                                    };
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_3D_1batch_1channel_output_padding)
+{
+    const Strides strides{1, 1, 1};
+    const CoordinateDiff padding{1, 1, 1};
+    const Strides dilations{1, 1, 1};
+    const CoordinateDiff output_padding{1, 1, 1};
+
+    const Shape inputs_shape{1, 1, 2, 2, 2};
+    const std::vector<float> inputs{
+                                    // depth: 1
+                                    15.0f, 3.0f,
+                                    21.0f, 10.0f,
+                                    // depth: 2
+                                    10.0f, 13.0f,
+                                    11.0f, 17.0f};
+
+    const Shape filter_shape{1, 1, 3, 3, 3};
+    const std::vector<float> filters{
+                                    // depth: 1
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 2
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 3
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f};
+
+    const Shape outputs_shape{1, 1, 3, 3, 3};
+    const std::vector<float> outputs{
+                                    // depth: 1
+                                    116.0f, 166.0f, 81.0f,
+                                    89.0f, 93.0f, 32.0f,
+                                    86.0f, 91.0f, 54.0f,
+                                    // depth: 2
+                                    116.0f, 166.0f, 81.0f,
+                                    89.0f, 93.0f, 32.0f,
+                                    86.0f, 91.0f, 54.0f,
+                                    // depth: 3
+                                    49.0f, 80.0f, 51.0f,
+                                    47.0f, 50.0f, 26.0f,
+                                    45.0f, 39.0f, 34.0f
+                                    };
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_3D_1batch_1channel_padding)
+{
+    const Strides strides{1, 1, 1};
+    const CoordinateDiff padding{1, 1, 1};
+    const Strides dilations{1, 1, 1};
+    const CoordinateDiff output_padding{0, 0, 0};
+
+    const Shape inputs_shape{1, 1, 4, 4, 4};
+    const std::vector<float> inputs{
+                                    // depth: 1
+                                    1.0f, 3.0f, 2.0f, 1.0f,
+                                    1.0f, 3.0f, 3.0f, 1.0f,
+                                    2.0f, 1.0f, 1.0f, 3.0f,
+                                    3.0f, 2.0f, 3.0f, 3.0f,
+                                    // depth: 2
+                                    1.0f, 3.0f, 2.0f, 1.0f,
+                                    1.0f, 3.0f, 3.0f, 1.0f,
+                                    2.0f, 1.0f, 1.0f, 3.0f,
+                                    3.0f, 2.0f, 3.0f, 3.0f,
+                                    // depth: 3
+                                    1.0f, 3.0f, 2.0f, 1.0f,
+                                    1.0f, 3.0f, 3.0f, 1.0f,
+                                    2.0f, 1.0f, 1.0f, 3.0f,
+                                    3.0f, 2.0f, 3.0f, 3.0f,
+                                    // depth: 4
+                                    1.0f, 3.0f, 2.0f, 1.0f,
+                                    1.0f, 3.0f, 3.0f, 1.0f,
+                                    2.0f, 1.0f, 1.0f, 3.0f,
+                                    3.0f, 2.0f, 3.0f, 3.0f
+                                    };
+
+    const Shape filter_shape{1, 1, 3, 3, 3};
+    const std::vector<float> filters{
+                                    // depth: 1
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 2
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 3
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f};
+
+    const Shape outputs_shape{1, 1, 4, 4, 4};
+    const std::vector<float> outputs{
+                                     // depth: 1
+                                     12.0f, 30.0f, 36.0f, 24.0f,
+                                     26.0f, 42.0f, 42.0f, 30.0f,
+                                     34.0f, 56.0f, 54.0f, 50.0f,
+                                     14.0f, 18.0f, 24.0f, 16.0f,
+                                     // depth: 2
+                                     18.0f, 45.0f, 54.0f, 36.0f,
+                                     39.0f, 63.0f, 63.0f, 45.0f,
+                                     51.0f, 84.0f, 81.0f, 75.0f,
+                                     21.0f, 27.0f, 36.0f, 24.0f,
+                                     // depth: 3
+                                     18.0f, 45.0f, 54.0f, 36.0f,
+                                     39.0f, 63.0f, 63.0f, 45.0f,
+                                     51.0f, 84.0f, 81.0f, 75.0f,
+                                     21.0f, 27.0f, 36.0f, 24.0f,
+                                     // depth: 4
+                                     12.0f, 30.0f, 36.0f, 24.0f,
+                                     26.0f, 42.0f, 42.0f, 30.0f,
+                                     34.0f, 56.0f, 54.0f, 50.0f,
+                                     14.0f, 18.0f, 24.0f, 16.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_3D_1batch_1channel_stride)
+{
+    const Strides strides{2, 2, 2};
+    const CoordinateDiff padding{0, 0, 0};
+    const Strides dilations{1, 1, 1};
+    const CoordinateDiff output_padding{0, 0, 0};
+
+    const Shape inputs_shape{1, 1, 2, 2, 2};
+    const std::vector<float> inputs{
+                                    // depth: 1
+                                    15.0f, 3.0f,
+                                    21.0f, 10.0f,
+                                    // depth: 2
+                                    10.0f, 13.0f,
+                                    11.0f, 17.0f};
+
+    const Shape filter_shape{1, 1, 3, 3, 3};
+    const std::vector<float> filters{
+                                    // depth: 1
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 2
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 3
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f};
+
+    const Shape outputs_shape{1, 1, 5, 5, 5};
+    const std::vector<float> outputs{
+                                    // depth: 1
+                                    15.0f, 30.0f, 48.0f, 6.0f, 9.0f,
+                                    0.0f, 15.0f, 0.0f, 3.0f, 0.0f,
+                                    51.0f, 57.0f, 109.0f, 23.0f, 36.0f,
+                                    0.0f, 21.0f, 0.0f, 10.0f, 0.0f,
+                                    42.0f, 21.0f, 62.0f, 10.0f, 20.0f,
+                                    // depth: 2
+                                    15.0f, 30.0f, 48.0f, 6.0f, 9.0f,
+                                    0.0f, 15.0f, 0.0f, 3.0f, 0.0f,
+                                    51.0f, 57.0f, 109.0f, 23.0f, 36.0f,
+                                    0.0f, 21.0f, 0.0f, 10.0f, 0.0f,
+                                    42.0f, 21.0f, 62.0f, 10.0f, 20.0f,
+                                    // depth: 3
+                                    25.0f, 50.0f, 91.0f, 32.0f, 48.0f,
+                                    0.0f, 25.0f, 0.0f, 16.0f, 0.0f,
+                                    82.0f, 89.0f, 205.0f, 70.0f, 113.0f,
+                                    0.0f, 32.0f, 0.0f, 27.0f, 0.0f,
+                                    64.0f, 32.0f, 118.0f, 27.0f, 54.0f,
+                                    // depth: 4
+                                    10.0f, 20.0f, 43.0f, 26.0f, 39.0f,
+                                    0.0f, 10.0f, 0.0f, 13.0f, 0.0f,
+                                    31.0f, 32.0f, 96.0f, 47.0f, 77.0f,
+                                    0.0f, 11.0f, 0.0f, 17.0f, 0.0f,
+                                    22.0f, 11.0f, 56.0f, 17.0f, 34.0f,
+                                    // depth: 5
+                                    10.0f, 20.0f, 43.0f, 26.0f, 39.0f,
+                                    0.0f, 10.0f, 0.0f, 13.0f, 0.0f,
+                                    31.0f, 32.0f, 96.0f, 47.0f, 77.0f,
+                                    0.0f, 11.0f, 0.0f, 17.0f, 0.0f,
+                                    22.0f, 11.0f, 56.0f, 17.0f, 34.0f
+                                    };
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_3D_1batch_1channel_padding_strides_dilation)
+{
+    const Strides strides{2, 2, 2};
+    const CoordinateDiff padding{2, 2, 2};
+    const Strides dilations{2, 2, 2};
+    const CoordinateDiff output_padding{0, 0, 0};
+
+    const Shape inputs_shape{1, 1, 4, 4, 4};
+    const std::vector<float> inputs{
+                                    // depth: 1
+                                    1.0f, 3.0f, 2.0f, 1.0f,
+                                    1.0f, 3.0f, 3.0f, 1.0f,
+                                    2.0f, 1.0f, 1.0f, 3.0f,
+                                    3.0f, 2.0f, 3.0f, 3.0f,
+                                    // depth: 2
+                                    1.0f, 3.0f, 2.0f, 1.0f,
+                                    1.0f, 3.0f, 3.0f, 1.0f,
+                                    2.0f, 1.0f, 1.0f, 3.0f,
+                                    3.0f, 2.0f, 3.0f, 3.0f,
+                                    // depth: 3
+                                    1.0f, 3.0f, 2.0f, 1.0f,
+                                    1.0f, 3.0f, 3.0f, 1.0f,
+                                    2.0f, 1.0f, 1.0f, 3.0f,
+                                    3.0f, 2.0f, 3.0f, 3.0f,
+                                    // depth: 4
+                                    1.0f, 3.0f, 2.0f, 1.0f,
+                                    1.0f, 3.0f, 3.0f, 1.0f,
+                                    2.0f, 1.0f, 1.0f, 3.0f,
+                                    3.0f, 2.0f, 3.0f, 3.0f
+                                    };
+
+    const Shape filter_shape{1, 1, 3, 3, 3};
+    const std::vector<float> filters{
+                                    // depth: 1
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 2
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 3
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f};
+
+    const Shape outputs_shape{1, 1, 7, 7, 7};
+    const std::vector<float> outputs{
+                                    // depth: 1
+                                    12.0f, 0.0f, 30.0f, 0.0f, 36.0f, 0.0f, 24.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    26.0f, 0.0f, 42.0f, 0.0f, 42.0f, 0.0f, 30.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    34.0f, 0.0f, 56.0f, 0.0f, 54.0f, 0.0f, 50.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    14.0f, 0.0f, 18.0f, 0.0f, 24.0f, 0.0f, 16.0f,
+                                    // depth: 2
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    // depth: 3
+                                    18.0f, 0.0f, 45.0f, 0.0f, 54.0f, 0.0f, 36.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    39.0f, 0.0f, 63.0f, 0.0f, 63.0f, 0.0f, 45.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    51.0f, 0.0f, 84.0f, 0.0f, 81.0f, 0.0f, 75.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    21.0f, 0.0f, 27.0f, 0.0f, 36.0f, 0.0f, 24.0f,
+                                    // depth: 4
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    // depth: 5
+                                    18.0f, 0.0f, 45.0f, 0.0f, 54.0f, 0.0f, 36.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    39.0f, 0.0f, 63.0f, 0.0f, 63.0f, 0.0f, 45.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    51.0f, 0.0f, 84.0f, 0.0f, 81.0f, 0.0f, 75.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    21.0f, 0.0f, 27.0f, 0.0f, 36.0f, 0.0f, 24.0f,
+                                    // depth: 6
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    // depth: 7
+                                    12.0f, 0.0f, 30.0f, 0.0f, 36.0f, 0.0f, 24.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    26.0f, 0.0f, 42.0f, 0.0f, 42.0f, 0.0f, 30.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    34.0f, 0.0f, 56.0f, 0.0f, 54.0f, 0.0f, 50.0f,
+                                    0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+                                    14.0f, 0.0f, 18.0f, 0.0f, 24.0f, 0.0f, 16.0f
+                                    };
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_3D_1batch_2channel)
+{
+    const Strides strides{1, 1, 1};
+    const CoordinateDiff padding{0, 0, 0};
+    const Strides dilations{1, 1, 1};
+    const CoordinateDiff output_padding{0, 0, 0};
+
+    const Shape inputs_shape{1, 1, 2, 2, 2};
+    const std::vector<float> inputs{
+                                    // depth: 1
+                                    1.0f, 8.0f,
+                                    1.0f, 3.0f,
+                                    // depth: 2
+                                    1.0f, 7.0f,
+                                    3.0f, 8.0f};
+
+    const Shape filter_shape{1, 2, 3, 3, 3};
+    const std::vector<float> filters{
+                                    // -- channel 1 --
+                                    // depth: 1
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 2
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 3
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // -- channel 2 --
+                                    // depth: 1
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 2
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 3
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f
+                                    };
+
+    const Shape outputs_shape{1, 2, 4, 4, 4};
+    const std::vector<float> outputs{
+                                    // -- channel 1 --
+                                    // depth: 1
+                                    1.0f, 10.0f, 19.0f, 24.0f,
+                                    1.0f, 6.0f, 17.0f, 9.0f,
+                                    2.0f, 18.0f, 13.0f, 16.0f,
+                                    2.0f, 7.0f, 5.0f, 6.0f,
+                                    // depth: 2
+                                    2.0f, 19.0f, 36.0f, 45.0f,
+                                    4.0f, 21.0f, 49.0f, 33.0f,
+                                    4.0f, 36.0f, 30.0f, 30.0f,
+                                    8.0f, 26.0f, 19.0f, 22.0f,
+                                    // depth: 3
+                                    2.0f, 19.0f, 36.0f, 45.0f,
+                                    4.0f, 21.0f, 49.0f, 33.0f,
+                                    4.0f, 36.0f, 30.0f, 30.0f,
+                                    8.0f, 26.0f, 19.0f, 22.0f,
+                                    // depth: 4
+                                    1.0f, 9.0f, 17.0f, 21.0f,
+                                    3.0f, 15.0f, 32.0f, 24.0f,
+                                    2.0f, 18.0f, 17.0f, 14.0f,
+                                    6.0f, 19.0f, 14.0f, 16.0f,
+                                    // -- channel 2 --
+                                    // depth: 1
+                                    1.0f, 10.0f, 19.0f, 24.0f,
+                                    1.0f, 6.0f, 17.0f, 9.0f,
+                                    2.0f, 18.0f, 13.0f, 16.0f,
+                                    2.0f, 7.0f, 5.0f, 6.0f,
+                                    // depth: 2
+                                    2.0f, 19.0f, 36.0f, 45.0f,
+                                    4.0f, 21.0f, 49.0f, 33.0f,
+                                    4.0f, 36.0f, 30.0f, 30.0f,
+                                    8.0f, 26.0f, 19.0f, 22.0f,
+                                    // depth: 3
+                                    2.0f, 19.0f, 36.0f, 45.0f,
+                                    4.0f, 21.0f, 49.0f, 33.0f,
+                                    4.0f, 36.0f, 30.0f, 30.0f,
+                                    8.0f, 26.0f, 19.0f, 22.0f,
+                                    // depth: 4
+                                    1.0f, 9.0f, 17.0f, 21.0f,
+                                    3.0f, 15.0f, 32.0f, 24.0f,
+                                    2.0f, 18.0f, 17.0f, 14.0f,
+                                    6.0f, 19.0f, 14.0f, 16.0f
+                                    };
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_3D_1batch_2filter)
+{
+    const Strides strides{1, 1, 1};
+    const CoordinateDiff padding{0, 0, 0};
+    const Strides dilations{1, 1, 1};
+    const CoordinateDiff output_padding{0, 0, 0};
+
+    const Shape inputs_shape{1, 2, 2, 2, 2};
+    const std::vector<float> inputs{
+                                    // -- in 1 --
+                                    // depth: 1
+                                    1.0f, 3.0f,
+                                    2.0f, 5.0f,
+                                    // depth: 2
+                                    1.0f, 0.0f,
+                                    3.0f, 6.0f,
+                                    // -- in 2 --
+                                    // depth: 1
+                                    1.0f, 3.0f,
+                                    2.0f, 5.0f,
+                                    // depth: 2
+                                    3.0f, 0.0f,
+                                    1.0f, 8.0f};
+
+    const Shape filter_shape{2, 1, 3, 3, 3};
+    const std::vector<float> filters{
+                                    // -- filter 1 --
+                                    // depth: 1
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 2
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 3
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // -- filter 2 --
+                                    // depth: 1
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 2
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 3
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f
+                                    };
+
+    const Shape outputs_shape{1, 1, 4, 4, 4};
+    const std::vector<float> outputs{
+                                     // depth: 1
+                                     2.0f, 10.0f, 18.0f, 18.0f,
+                                     4.0f, 20.0f, 38.0f, 30.0f,
+                                     4.0f, 18.0f, 20.0f, 12.0f,
+                                     8.0f, 24.0f, 18.0f, 20.0f,
+                                     // depth: 2
+                                     6.0f, 18.0f, 30.0f, 18.0f,
+                                     8.0f, 46.0f, 78.0f, 72.0f,
+                                     12.0f, 26.0f, 42.0f, 12.0f,
+                                     16.0f, 56.0f, 40.0f, 48.0f,
+                                     // depth: 3
+                                     6.0f, 18.0f, 30.0f, 18.0f,
+                                     8.0f, 46.0f, 78.0f, 72.0f,
+                                     12.0f, 26.0f, 42.0f, 12.0f,
+                                     16.0f, 56.0f, 40.0f, 48.0f,
+                                     // depth: 4
+                                     4.0f, 8.0f, 12.0f, 0.0f,
+                                     4.0f, 26.0f, 40.0f, 42.0f,
+                                     8.0f, 8.0f, 22.0f, 0.0f,
+                                     8.0f, 32.0f, 22.0f, 28.0f
+                                     };
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, convolution_backprop_3D_2batch_1channel)
+{
+    const Strides strides{1, 1, 1};
+    const CoordinateDiff padding{0, 0, 0};
+    const Strides dilations{1, 1, 1};
+    const CoordinateDiff output_padding{0, 0, 0};
+
+    const Shape inputs_shape{2, 1, 2, 2, 2};
+    const std::vector<float> inputs{
+                                    // -- batch 1 --
+                                    // depth: 1
+                                    1.0f, 3.0f,
+                                    2.0f, 5.0f,
+                                    // depth: 2
+                                    1.0f, 0.0f,
+                                    6.0f, 4.0f,
+                                    // -- batch 2 --
+                                    // depth: 1
+                                    1.0f, 5.0f,
+                                    2.0f, 8.0f,
+                                    // depth: 2
+                                    2.0f, 1.0f,
+                                    0.0f, 5.0f};
+    const Shape filter_shape{1, 1, 3, 3, 3};
+    const std::vector<float> filters{
+                                    // depth: 1
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 2
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f,
+                                    // depth: 3
+                                    1.0f, 2.0f, 3.0f,
+                                    0.0f, 1.0f, 0.0f,
+                                    2.0f, 1.0f, 2.0f};
+
+    const Shape outputs_shape{2, 1, 4, 4, 4};
+    const std::vector<float> outputs{
+                                     // -- batch 1 --
+                                     // depth: 1
+                                     1.0f, 5.0f, 9.0f, 9.0f,
+                                     2.0f, 10.0f, 19.0f, 15.0f,
+                                     2.0f, 9.0f, 10.0f, 6.0f,
+                                     4.0f, 12.0f, 9.0f, 10.0f,
+                                     // depth: 2
+                                     2.0f, 7.0f, 12.0f, 9.0f,
+                                     8.0f, 27.0f, 45.0f, 27.0f,
+                                     4.0f, 16.0f, 16.0f, 6.0f,
+                                     16.0f, 26.0f, 25.0f, 18.0f,
+                                     // depth: 3
+                                     2.0f, 7.0f, 12.0f, 9.0f,
+                                     8.0f, 27.0f, 45.0f, 27.0f,
+                                     4.0f, 16.0f, 16.0f, 6.0f,
+                                     16.0f, 26.0f, 25.0f, 18.0f,
+                                     // depth: 4
+                                     1.0f, 2.0f, 3.0f, 0.0f,
+                                     6.0f, 17.0f, 26.0f, 12.0f,
+                                     2.0f, 7.0f, 6.0f, 0.0f,
+                                     12.0f, 14.0f, 16.0f, 8.0f,
+                                     // -- batch 2 --
+                                     // depth: 1
+                                     1.0f, 7.0f, 13.0f, 15.0f,
+                                     2.0f, 13.0f, 27.0f, 24.0f,
+                                     2.0f, 13.0f, 15.0f, 10.0f,
+                                     4.0f, 18.0f, 12.0f, 16.0f,
+                                     // depth: 2
+                                     3.0f, 12.0f, 21.0f, 18.0f,
+                                     2.0f, 20.0f, 38.0f, 39.0f,
+                                     6.0f, 17.0f, 25.0f, 12.0f,
+                                     4.0f, 28.0f, 17.0f, 26.0f,
+                                     // depth: 3
+                                     3.0f, 12.0f, 21.0f, 18.0f,
+                                     2.0f, 20.0f, 38.0f, 39.0f,
+                                     6.0f, 17.0f, 25.0f, 12.0f,
+                                     4.0f, 28.0f, 17.0f, 26.0f,
+                                     // depth: 4
+                                     2.0f, 5.0f, 8.0f, 3.0f,
+                                     0.0f, 7.0f, 11.0f, 15.0f,
+                                     4.0f, 4.0f, 10.0f, 2.0f,
+                                     0.0f, 10.0f, 5.0f, 10.0f};
+
+    ConvolutionBackpropTest(inputs, inputs_shape, filters, filter_shape, outputs, outputs_shape,
+                    strides, padding, dilations, output_padding);
+}
diff --git a/ngraph/test/runtime/interpreter/evaluates_map.cpp b/ngraph/test/runtime/interpreter/evaluates_map.cpp
index 610101177797fd..4446a4a24347de 100644
--- a/ngraph/test/runtime/interpreter/evaluates_map.cpp
+++ b/ngraph/test/runtime/interpreter/evaluates_map.cpp
@@ -283,7 +283,8 @@ namespace
             op->get_dilations(),
             op->get_pads_begin(),
             op->get_pads_end(),
-            op->get_strides());
+            op->get_strides(),
+            op->get_output_padding());
         return true;
     }
 

From 17b5240065304a37fb35c9e8485983d84537a03e Mon Sep 17 00:00:00 2001
From: Mingyu Kim <mingyu.kim@intel.com>
Date: Mon, 14 Jun 2021 22:47:59 +0900
Subject: [PATCH 073/178] Typo fix (#6144)

---
 inference-engine/samples/benchmark_app/main.cpp | 4 ++--
 tools/benchmark/main.py                         | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/inference-engine/samples/benchmark_app/main.cpp b/inference-engine/samples/benchmark_app/main.cpp
index cd7ddc641dc256..2a5252ba443a85 100644
--- a/inference-engine/samples/benchmark_app/main.cpp
+++ b/inference-engine/samples/benchmark_app/main.cpp
@@ -277,8 +277,8 @@ int main(int argc, char* argv[]) {
                 setThroughputStreams();
 
                 if ((device_name.find("MULTI") != std::string::npos) && (device_name.find("CPU") != std::string::npos)) {
-                    slog::warn << "Turn on GPU trottling. Multi-device execution with "
-                                  "the CPU + GPU performs best with GPU trottling hint,"
+                    slog::warn << "Turn on GPU throttling. Multi-device execution with "
+                                  "the CPU + GPU performs best with GPU throttling hint, "
                                << "which releases another CPU thread (that is otherwise "
                                   "used by the GPU driver for active polling)"
                                << slog::endl;
diff --git a/tools/benchmark/main.py b/tools/benchmark/main.py
index 26ef6246f0c9d5..0cc0a511848d65 100644
--- a/tools/benchmark/main.py
+++ b/tools/benchmark/main.py
@@ -150,7 +150,7 @@ def set_throughput_streams():
                 set_throughput_streams()
 
                 if MULTI_DEVICE_NAME in device_name and CPU_DEVICE_NAME in device_name:
-                    logger.warning("Turn on GPU trottling. Multi-device execution with the CPU + GPU performs best with GPU trottling hint, " +
+                    logger.warning("Turn on GPU throttling. Multi-device execution with the CPU + GPU performs best with GPU throttling hint, " +
                                    "which releases another CPU thread (that is otherwise used by the GPU driver for active polling)")
                     config[device]['GPU_PLUGIN_THROTTLE'] = '1'
             elif device == MYRIAD_DEVICE_NAME:

From fb7435a64758d2b1bce56e78e31d99cab9742ba8 Mon Sep 17 00:00:00 2001
From: Andrey Somsikov <andrey.somsikov@intel.com>
Date: Mon, 14 Jun 2021 17:00:49 +0300
Subject: [PATCH 074/178] Add cnpy fuzz test and fix issues (#6109)

---
 tests/fuzz/README.md                      | 70 +++++++++++++++++++++--
 tests/fuzz/fuzz-testhelper/CMakeLists.txt |  2 +-
 tests/fuzz/fuzz-testhelper/fuzz-utils.cc  | 40 +++++++++++++
 tests/fuzz/fuzz-testhelper/fuzz-utils.h   | 19 ++++++
 tests/fuzz/src/CMakeLists.txt             |  5 +-
 tests/fuzz/src/cnpy_npy_load-fuzzer.cc    | 21 +++++++
 thirdparty/cnpy/cnpy.cpp                  |  8 ++-
 thirdparty/cnpy/cnpy.h                    |  5 ++
 8 files changed, 160 insertions(+), 10 deletions(-)
 create mode 100644 tests/fuzz/fuzz-testhelper/fuzz-utils.cc
 create mode 100644 tests/fuzz/fuzz-testhelper/fuzz-utils.h
 create mode 100644 tests/fuzz/src/cnpy_npy_load-fuzzer.cc

diff --git a/tests/fuzz/README.md b/tests/fuzz/README.md
index 5f16f9f9998775..19b34e80c1764d 100644
--- a/tests/fuzz/README.md
+++ b/tests/fuzz/README.md
@@ -18,14 +18,72 @@ To run fuzzing you will need [LLVM](https://apt.llvm.org/) components:
 - lld (linker)
 - libc++
 
-## Reproducing Failure Found by Fuzzing
 
-1. Build `fuzz` test target:
+## Building fuzz tests
+
+1. Build openvino
+
+Build openvino with options `ENABLE_FUZZING` and `ENABLE_SANITIZER` enabled. It
+is recommended to use clang compiler.
+
+```bash
+(\
+mkdir -p build && cd build && \
+CC=clang CXX=clang++ cmake .. -DENABLE_FUZZING=ON -DENABLE_SANITIZER=ON -DTREAT_WARNING_AS_ERROR=OFF && \
+cmake --build . \
+)
+```
+
+2. Build fuzz tests
+
+Build fuzz tests with options `ENABLE_FUZZING` and `ENABLE_SANITIZER` enabled.
+You should use the same compiler as was used for the openvino build.
+
+```bash
+(\
+mkdir -p tests/fuzz/build && cd tests/fuzz/build && \
+CC=clang CXX=clang++ cmake .. -DENABLE_FUZZING=ON -DENABLE_SANITIZER=ON -DTREAT_WARNING_AS_ERROR=OFF -DInferenceEngine_DIR=$(pwd)/../../../build && \
+cmake --build . \
+)
+```
+
+## Running fuzz tests
+
+1. Prepare fuzzing corpus
+
+Fuzzing engine needs a set of valid inputs to start fuzzing from. Those files
+are called a fuzzing corpus. Place valid inputs for the fuzzing test into
+directory.
+
+Intel employees can get the corpus as described here
+https://wiki.ith.intel.com/x/2N42bg. 
+
+2. Run fuzzing
+
 ```bash
-cmake -DENABLE_TESTS=ON .. && ninja fuzz
+./read_network-fuzzer -max_total_time=600 ./read_network-corpus
 ```
+Consider adding those useful command line options:
+- `-jobs=$(nproc)` runs multiple fuzzing jobs in parallel.
+- `-rss_limit_mb=0` to ignore out-of-memory issues.
+
+## Analyzing fuzzing quality
 
-2. Run fuzzing test passing a failure reproducer as a command-line argument:
-``` bash
-./read_network-fuzzer crash-reproducer
+### Explore code coverage
+
+To build coverage report after fuzz test execution run:
+
+```
+llvm-profdata merge -sparse *.profraw -o default.profdata && \
+llvm-cov show ./read_network-fuzzer -instr-profile=default.profdata -format=html -output-dir=read_network-coverage
 ```
+
+## Reproducing findings
+
+Fuzzing run halts on the first issue identified, prints issue details to stdout and save data to reproduce the issue as a file in the current folder. To debug the issue pass reproducer as command line argument to fuzz test
+
+```bash
+./read_network-fuzzer crash-409b5eeed46a8445b7f7b7a2ce5b60a9ad895e3b
+```
+
+It is recommended but not required to use binaries built for fuzzing to debug the issues. A binaries built without `ENABLE_FUZZING` options can also be used to reproduce and debug the issues.
\ No newline at end of file
diff --git a/tests/fuzz/fuzz-testhelper/CMakeLists.txt b/tests/fuzz/fuzz-testhelper/CMakeLists.txt
index 2c47dbd2d3f86a..52e92798d0c01e 100644
--- a/tests/fuzz/fuzz-testhelper/CMakeLists.txt
+++ b/tests/fuzz/fuzz-testhelper/CMakeLists.txt
@@ -6,7 +6,7 @@ set(TARGET_NAME fuzz-testhelper)
 
 file(
     GLOB SRC_FILES
-    ${CMAKE_CURRENT_SOURCE_DIR}/*-testhelper.cc)
+    ${CMAKE_CURRENT_SOURCE_DIR}/*.cc)
 
 add_library(
     ${TARGET_NAME} STATIC
diff --git a/tests/fuzz/fuzz-testhelper/fuzz-utils.cc b/tests/fuzz/fuzz-testhelper/fuzz-utils.cc
new file mode 100644
index 00000000000000..16e63e6576dc12
--- /dev/null
+++ b/tests/fuzz/fuzz-testhelper/fuzz-utils.cc
@@ -0,0 +1,40 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "fuzz-utils.h"
+#include <stdexcept>
+#include <stdlib.h>
+#include <string.h>
+#include <string>
+#ifndef _WIN32
+#include <unistd.h>
+#endif  // _WIN32
+
+MemoryFile::MemoryFile(const void *data, size_t size) {
+#ifdef _WIN32
+    throw std::exception("MemoryFile is not implemented for Windows");
+#else  // _WIN32
+    m_name = strdup("/dev/shm/fuzz-XXXXXX");
+    if (!m_name)
+        throw std::bad_alloc();
+    int fd = mkstemp(m_name);
+    if (size) {
+        size_t nbytes = write(fd, data, size);
+        if (nbytes != size) {
+            free(m_name);
+            close(fd);
+            throw std::runtime_error("Failed to write " + std::to_string(size) +
+                                     " bytes to " + m_name);
+        }
+    }
+    close(fd);
+#endif  // _WIN32
+}
+
+MemoryFile::~MemoryFile() {
+#ifndef _WIN32
+    unlink(m_name);
+    free(m_name);
+#endif  // _WIN32
+}
diff --git a/tests/fuzz/fuzz-testhelper/fuzz-utils.h b/tests/fuzz/fuzz-testhelper/fuzz-utils.h
new file mode 100644
index 00000000000000..f167587eb65e9d
--- /dev/null
+++ b/tests/fuzz/fuzz-testhelper/fuzz-utils.h
@@ -0,0 +1,19 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <stddef.h>
+
+class MemoryFile {
+  public:
+    /// Create a memory backed file
+    MemoryFile(const void *data, size_t size);
+    /// Delete memory backed file
+    ~MemoryFile();
+
+    /// Get path to a file.
+    const char *name() { return m_name; }
+
+  private:
+    char *m_name;
+};
diff --git a/tests/fuzz/src/CMakeLists.txt b/tests/fuzz/src/CMakeLists.txt
index b58a2d018aae5a..b9400d9e7e1d7d 100644
--- a/tests/fuzz/src/CMakeLists.txt
+++ b/tests/fuzz/src/CMakeLists.txt
@@ -9,11 +9,14 @@ add_custom_target(fuzz)
 # Fuzz test target name is source file name without extension.
 FILE(GLOB tests "*-fuzzer.cc")
 
+add_subdirectory(../../../thirdparty/cnpy ${CMAKE_CURRENT_BINARY_DIR}/cnpy)
+add_subdirectory(../../../thirdparty/zlib ${CMAKE_CURRENT_BINARY_DIR}/zlib)
+
 foreach(test_source ${tests})
     get_filename_component(test_name ${test_source} NAME_WE)
     add_fuzzer(${test_name} ${test_source})
 
-    target_link_libraries(${test_name} PRIVATE IE::inference_engine)
+    target_link_libraries(${test_name} PRIVATE IE::inference_engine cnpy zlib)
 
     add_dependencies(fuzz ${test_name})
 endforeach()
diff --git a/tests/fuzz/src/cnpy_npy_load-fuzzer.cc b/tests/fuzz/src/cnpy_npy_load-fuzzer.cc
new file mode 100644
index 00000000000000..257dc22908b85d
--- /dev/null
+++ b/tests/fuzz/src/cnpy_npy_load-fuzzer.cc
@@ -0,0 +1,21 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <stdio.h>
+#include <cnpy.h>
+#include "fuzz-utils.h"
+
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t * inputData, size_t inputSize) {
+    MemoryFile file(inputData, inputSize);
+    
+    try {
+        cnpy::NpyArray array = cnpy::npy_load(file.name());
+    }
+    catch (const std::exception&) {
+        return 0;  // fail gracefully on expected exceptions
+    }
+
+    return 0;
+}
\ No newline at end of file
diff --git a/thirdparty/cnpy/cnpy.cpp b/thirdparty/cnpy/cnpy.cpp
index 26d0614bca1906..ed277deb5fec23 100644
--- a/thirdparty/cnpy/cnpy.cpp
+++ b/thirdparty/cnpy/cnpy.cpp
@@ -90,7 +90,9 @@ void cnpy::parse_npy_header(unsigned char* buffer,size_t& word_size, std::vector
     //byte order code | stands for not applicable. 
     //not sure when this applies except for byte array
     loc1 = header.find("descr")+9;
-    bool littleEndian = (header[loc1] == '<' || header[loc1] == '|' ? true : false);
+    bool littleEndian = false;
+    if (loc1 < header.size())
+        littleEndian = (header[loc1] == '<' || header[loc1] == '|' ? true : false);
     assert(littleEndian);
 
     //char type = header[loc1+1];
@@ -148,7 +150,9 @@ void cnpy::parse_npy_header(FILE* fp, size_t& word_size, std::vector<size_t>& sh
     if (loc1 == std::string::npos)
         throw std::runtime_error("parse_npy_header: failed to find header keyword: 'descr'");
     loc1 += 9;
-    bool littleEndian = (header[loc1] == '<' || header[loc1] == '|' ? true : false);
+    bool littleEndian = false;
+    if (loc1 < header.size())
+        littleEndian = (header[loc1] == '<' || header[loc1] == '|' ? true : false);
     assert(littleEndian);
 
     //char type = header[loc1+1];
diff --git a/thirdparty/cnpy/cnpy.h b/thirdparty/cnpy/cnpy.h
index 750251f480eaf4..e8935e8937e89c 100644
--- a/thirdparty/cnpy/cnpy.h
+++ b/thirdparty/cnpy/cnpy.h
@@ -27,6 +27,11 @@ namespace cnpy {
         {
             num_vals = 1;
             for(size_t i = 0;i < shape.size();i++) num_vals *= shape[i];
+            if (word_size &&
+                num_vals > std::vector<char>().max_size() / word_size)
+                throw std::length_error("NpyArray of " + std::to_string(num_vals) +
+                                        "*" + std::to_string(word_size) +
+                                        " elements is too big.");
             data_holder = std::shared_ptr<std::vector<char>>(
                 new std::vector<char>(num_vals * word_size));
         }

From f8759e19821cbb61cf905fbc48e4e511d94939e3 Mon Sep 17 00:00:00 2001
From: Bartosz Lesniewski <bartosz.lesniewski@intel.com>
Date: Tue, 15 Jun 2021 05:55:48 +0200
Subject: [PATCH 075/178] Revise ExtractImagePatches op (#6104)

* use ngraph rtti macros

* Add attribute count check to visitor tests

* Add sslt for ExtractImagePatches

* refactoring variable names

* remove whitespaces from empty line
---
 .../single_layer/extractimagepatches.cpp      | 43 +++++++++++++++++++
 .../include/ngraph/op/extractimagepatches.hpp |  4 +-
 ngraph/core/src/op/extractimagepatches.cpp    | 36 ++++++++--------
 .../test/visitors/op/extractimagepatches.cpp  |  2 +
 4 files changed, 65 insertions(+), 20 deletions(-)
 create mode 100644 inference-engine/tests/functional/inference_engine/serialization/single_layer/extractimagepatches.cpp

diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/extractimagepatches.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/extractimagepatches.cpp
new file mode 100644
index 00000000000000..7e48a0ad0e93a7
--- /dev/null
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/extractimagepatches.cpp
@@ -0,0 +1,43 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <vector>
+#include "shared_test_classes/single_layer/extract_image_patches.hpp"
+
+using namespace ngraph;
+using namespace LayerTestsDefinitions;
+
+namespace {
+TEST_P(ExtractImagePatchesTest, Serialize) {
+    Serialize();
+}
+
+const std::vector<std::vector<size_t>> inShapes = {{2, 3, 13, 37}};
+const std::vector<std::vector<size_t>> kSizes = {{1, 5}, {3, 4}, {3, 1}};
+const std::vector<std::vector<size_t>> strides = {{1, 2}, {2, 2}, {2, 1}};
+const std::vector<std::vector<size_t>> rates = {{1, 3}, {3, 3}, {3, 1}};
+
+const std::vector<ngraph::op::PadType> autoPads = {
+    ngraph::op::PadType::VALID, ngraph::op::PadType::SAME_UPPER,
+    ngraph::op::PadType::SAME_LOWER
+};
+const std::vector<InferenceEngine::Precision> netPrecision = {
+    InferenceEngine::Precision::I8, InferenceEngine::Precision::BF16,
+    InferenceEngine::Precision::FP32
+};
+
+INSTANTIATE_TEST_CASE_P(smoke_ExtractImagePatchesLayerTest, ExtractImagePatchesTest,
+        ::testing::Combine(::testing::ValuesIn(inShapes),
+                            ::testing::ValuesIn(kSizes),
+                            ::testing::ValuesIn(strides),
+                            ::testing::ValuesIn(rates),
+                            ::testing::ValuesIn(autoPads),
+                            ::testing::ValuesIn(netPrecision),
+                            ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                            ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                            ::testing::Values(InferenceEngine::Layout::ANY),
+                            ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+        ExtractImagePatchesTest::getTestCaseName);
+
+} // namespace
\ No newline at end of file
diff --git a/ngraph/core/include/ngraph/op/extractimagepatches.hpp b/ngraph/core/include/ngraph/op/extractimagepatches.hpp
index e9c89813d06497..2cd405618768ec 100644
--- a/ngraph/core/include/ngraph/op/extractimagepatches.hpp
+++ b/ngraph/core/include/ngraph/op/extractimagepatches.hpp
@@ -15,8 +15,8 @@ namespace ngraph
             class NGRAPH_API ExtractImagePatches : public Op
             {
             public:
-                static constexpr NodeTypeInfo type_info{"ExtractImagePatches", 3};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
+                NGRAPH_RTTI_DECLARATION;
+
                 ExtractImagePatches() = default;
                 /// \brief Constructs a ExtractImagePatches operation
                 ///
diff --git a/ngraph/core/src/op/extractimagepatches.cpp b/ngraph/core/src/op/extractimagepatches.cpp
index 0eba4645db0778..eb66bbc384808c 100644
--- a/ngraph/core/src/op/extractimagepatches.cpp
+++ b/ngraph/core/src/op/extractimagepatches.cpp
@@ -11,7 +11,7 @@ using namespace ngraph;
 
 // ExtractImagePatches v3
 
-constexpr NodeTypeInfo op::v3::ExtractImagePatches::type_info;
+NGRAPH_RTTI_DEFINITION(op::v3::ExtractImagePatches, "ExtractImagePatches", 3);
 
 op::v3::ExtractImagePatches::ExtractImagePatches(const Output<Node>& image,
                                                  const Shape& sizes,
@@ -30,9 +30,9 @@ op::v3::ExtractImagePatches::ExtractImagePatches(const Output<Node>& image,
 void op::v3::ExtractImagePatches::validate_and_infer_types()
 {
     NGRAPH_OP_SCOPE(v3_ExtractImagePatches_validate_and_infer_types);
-    const PartialShape input_Pshape = get_input_partial_shape(0);
+    const PartialShape input_pshape = get_input_partial_shape(0);
 
-    NODE_VALIDATION_CHECK(this, input_Pshape.rank() == 4, "input tensor must be 4D tensor.");
+    NODE_VALIDATION_CHECK(this, input_pshape.rank() == 4, "input tensor must be 4D tensor.");
 
     NODE_VALIDATION_CHECK(this,
                           m_patch_sizes.size() == 2,
@@ -60,18 +60,18 @@ void op::v3::ExtractImagePatches::validate_and_infer_types()
             m_padding == PadType::SAME_UPPER,
         "Attribute padding should be in either valid or same_lower or same_upper.");
 
-    if (input_Pshape[1].is_dynamic() || input_Pshape[2].is_dynamic() ||
-        input_Pshape[3].is_dynamic())
+    if (input_pshape[1].is_dynamic() || input_pshape[2].is_dynamic() ||
+        input_pshape[3].is_dynamic())
     {
         set_input_is_relevant_to_shape(0);
-        auto output_Pshape = PartialShape::dynamic(4);
-        set_output_type(0, get_input_element_type(0), output_Pshape);
+        auto output_pshape = PartialShape::dynamic(4);
+        set_output_type(0, get_input_element_type(0), output_pshape);
     }
     else
     {
-        int32_t input_depth = input_Pshape[1].get_length();
-        int32_t input_rows = input_Pshape[2].get_length();
-        int32_t input_cols = input_Pshape[3].get_length();
+        int32_t input_depth = input_pshape[1].get_length();
+        int32_t input_rows = input_pshape[2].get_length();
+        int32_t input_cols = input_pshape[3].get_length();
         int32_t out_rows(0);
         int32_t out_cols(0);
 
@@ -113,26 +113,26 @@ void op::v3::ExtractImagePatches::validate_and_infer_types()
         ngraph::Dimension::value_type out_cols_cast =
             static_cast<ngraph::Dimension::value_type>(out_cols);
 
-        PartialShape output_Pshape;
-        if (input_Pshape[0].is_dynamic())
+        PartialShape output_pshape;
+        if (input_pshape[0].is_dynamic())
         {
-            output_Pshape =
-                PartialShape{input_Pshape[0], out_depth_cast, out_rows_cast, out_cols_cast};
+            output_pshape =
+                PartialShape{input_pshape[0], out_depth_cast, out_rows_cast, out_cols_cast};
         }
         else
         {
             ngraph::Dimension::value_type input_batch_cast =
-                static_cast<ngraph::Dimension::value_type>(input_Pshape[0].get_length());
-            output_Pshape =
+                static_cast<ngraph::Dimension::value_type>(input_pshape[0].get_length());
+            output_pshape =
                 PartialShape{input_batch_cast, out_depth_cast, out_rows_cast, out_cols_cast};
         }
 
         if (input_rows == 0 || input_cols == 0)
         {
-            output_Pshape = input_Pshape;
+            output_pshape = input_pshape;
         }
 
-        set_output_type(0, get_input_element_type(0), output_Pshape);
+        set_output_type(0, get_input_element_type(0), output_pshape);
     }
 }
 
diff --git a/ngraph/test/visitors/op/extractimagepatches.cpp b/ngraph/test/visitors/op/extractimagepatches.cpp
index 6d5c99146eac10..a22b0a14b03433 100644
--- a/ngraph/test/visitors/op/extractimagepatches.cpp
+++ b/ngraph/test/visitors/op/extractimagepatches.cpp
@@ -33,6 +33,8 @@ TEST(attributes, extractimagepatches_op)
     NodeBuilder builder(extractimagepatches);
     auto g_extractimagepatches = as_type_ptr<opset3::ExtractImagePatches>(builder.create());
 
+    const auto expected_attr_count = 4;
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
     EXPECT_EQ(g_extractimagepatches->get_sizes(), sizes);
     EXPECT_EQ(g_extractimagepatches->get_strides(), strides);
     EXPECT_EQ(g_extractimagepatches->get_rates(), rates);

From 8dff04df2882cf04a1ab984a17ae0a59a87e3a83 Mon Sep 17 00:00:00 2001
From: Katarzyna Mitrus <katarzyna.mitrus@intel.com>
Date: Tue, 15 Jun 2021 06:04:06 +0200
Subject: [PATCH 076/178] ShuffleChannels ng op and reference implementation
 revision (#5764)

* Unblock shuffle channels tests from ie test manifest

* Add more backend tests

* ShiffleChannel reference impl update

* Update attr visitor test

* Remove unused get_pre_shuffle_shape helper function

* Update class descriprion

* Add type prop shape tests

* Remove NGRAPH_SUPPRESS_DEPRECATED macro

* Add single layer tests

* Update layer tests

* Remove unused header

* Move implementation to cpp file
---
 .../single_layer_tests/shuffle_channels.cpp   |  61 ++++++
 .../single_layer_tests/shuffle_channels.cpp   |  37 ++++
 .../include/ngraph/op/shuffle_channels.hpp    |  19 +-
 .../runtime/reference/shuffle_channels.hpp    |  27 +++
 .../runtime/reference/shuffle_channels.cpp    |  60 ++++++
 ngraph/core/src/op/shuffle_channels.cpp       |  73 +------
 ngraph/test/CMakeLists.txt                    |   1 +
 ngraph/test/backend/fused_op.in.cpp           |  60 ------
 ngraph/test/backend/shuffle_channels.in.cpp   | 192 ++++++++++++++++++
 ngraph/test/runtime/ie/unit_test.manifest     |   5 -
 ngraph/test/type_prop/shuffle_channels.cpp    | 112 +++++++++-
 ngraph/test/visitors/op/shuffle_channels.cpp  |  16 +-
 12 files changed, 508 insertions(+), 155 deletions(-)
 create mode 100644 ngraph/core/reference/include/ngraph/runtime/reference/shuffle_channels.hpp
 create mode 100644 ngraph/core/reference/src/runtime/reference/shuffle_channels.cpp
 create mode 100644 ngraph/test/backend/shuffle_channels.in.cpp

diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
index b36a9e0713b0b0..b338d02efc14bd 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
@@ -36,4 +36,65 @@ INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels4D, ShuffleChannelsLayerTest,
                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ShuffleChannelsLayerTest::getTestCaseName);
 
+// ND support tests
+INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels6D, ShuffleChannelsLayerTest,
+        ::testing::Combine(
+                ::testing::Values(std::tuple<int, int>(2, 3)),
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(std::vector<size_t >({24, 6, 12, 18, 30, 36})),
+                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+        ShuffleChannelsLayerTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels5D, ShuffleChannelsLayerTest,
+        ::testing::Combine(
+                ::testing::Values(std::tuple<int, int>(2, 3)),
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(std::vector<size_t >({6, 12, 18, 30, 36})),
+                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+        ShuffleChannelsLayerTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels3D, ShuffleChannelsLayerTest,
+        ::testing::Combine(
+                ::testing::Values(std::tuple<int, int>(1, 3)),
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(std::vector<size_t >({18, 30, 36})),
+                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+        ShuffleChannelsLayerTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels2D, ShuffleChannelsLayerTest,
+        ::testing::Combine(
+                ::testing::Values(std::tuple<int, int>(1, 3)),
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(std::vector<size_t >({18, 30})),
+                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+        ShuffleChannelsLayerTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels1D, ShuffleChannelsLayerTest,
+        ::testing::Combine(
+                ::testing::Values(std::tuple<int, int>(0, 3)),
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(std::vector<size_t >({30})),
+                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+        ShuffleChannelsLayerTest::getTestCaseName);
+
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
index 6f7bdc1fcf2225..0eb593e16f45e4 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
@@ -35,3 +35,40 @@ const auto testCases = ::testing::Combine(::testing::ValuesIn(shuffleParameters)
 
 
 INSTANTIATE_TEST_CASE_P(smoke_GPU_ShuffleChannels, ShuffleChannelsLayerTest, testCases, ShuffleChannelsLayerTest::getTestCaseName);
+
+// ND support tests
+INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels3D, ShuffleChannelsLayerTest,
+        ::testing::Combine(
+                ::testing::Values(std::tuple<int, int>(1, 3)),
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(std::vector<size_t >({18, 30, 36})),
+                ::testing::Values(CommonTestUtils::DEVICE_GPU)),
+        ShuffleChannelsLayerTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels2D, ShuffleChannelsLayerTest,
+        ::testing::Combine(
+                ::testing::Values(std::tuple<int, int>(1, 3)),
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(std::vector<size_t >({18, 30})),
+                ::testing::Values(CommonTestUtils::DEVICE_GPU)),
+        ShuffleChannelsLayerTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels1D, ShuffleChannelsLayerTest,
+        ::testing::Combine(
+                ::testing::Values(std::tuple<int, int>(0, 3)),
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(std::vector<size_t >({30})),
+                ::testing::Values(CommonTestUtils::DEVICE_GPU)),
+        ShuffleChannelsLayerTest::getTestCaseName);
diff --git a/ngraph/core/include/ngraph/op/shuffle_channels.hpp b/ngraph/core/include/ngraph/op/shuffle_channels.hpp
index fd4539b0b3ac35..a09878c574f349 100644
--- a/ngraph/core/include/ngraph/op/shuffle_channels.hpp
+++ b/ngraph/core/include/ngraph/op/shuffle_channels.hpp
@@ -24,14 +24,12 @@ namespace ngraph
                 ShuffleChannels() = default;
                 /// \brief Constructs a ShuffleChannels node.
                 ///
-                /// \param data - Node producing the input tensor
-                /// \param axis - channel dimension index in the data tensor. A negative value means
-                ///               that the index should be calculated from the back of the input
-                ///               data
-                ///               shape.
-                /// \param group - number of group the channel dimension specified by axis should
-                /// be
-                ///                 split into
+                /// \param data     Node producing the input tensor.
+                /// \param axis     Channel dimension index in the data tensor.
+                ///                 A negative value means that the index should be
+                ///                 calculated from the back of the input data shape.
+                /// \param group    Number of group the channel dimension should be split into.
+                ///
                 ShuffleChannels(const Output<Node>& data,
                                 const int64_t axis = 1,
                                 const int64_t group = 1);
@@ -51,11 +49,6 @@ namespace ngraph
                 bool has_evaluate() const override;
 
             private:
-                /// \brief Generates a shape required to permute the data
-                ///
-                /// \param data_shape - Shape of the original input data tensor
-                /// \return A 4D tensor to be used to reshape the input data before shuffling it
-                Shape get_pre_shuffle_shape(const Shape& data_shape) const;
                 bool evaluate_shuffle_channels(const HostTensorVector& outputs,
                                                const HostTensorVector& inputs) const;
 
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/shuffle_channels.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/shuffle_channels.hpp
new file mode 100644
index 00000000000000..0d6fe7bed51cd9
--- /dev/null
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/shuffle_channels.hpp
@@ -0,0 +1,27 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <cmath>
+#include <numeric>
+#include <vector>
+
+#include "ngraph/shape.hpp"
+
+namespace ngraph
+{
+    namespace runtime
+    {
+        namespace reference
+        {
+            void shuffle_channels(const char* arg,
+                                  char* out,
+                                  const Shape& data_shape,
+                                  size_t elem_size,
+                                  const int64_t axis,
+                                  const int64_t group);
+        } // namespace reference
+    }     // namespace runtime
+} // namespace ngraph
diff --git a/ngraph/core/reference/src/runtime/reference/shuffle_channels.cpp b/ngraph/core/reference/src/runtime/reference/shuffle_channels.cpp
new file mode 100644
index 00000000000000..5a12f3787bf0e5
--- /dev/null
+++ b/ngraph/core/reference/src/runtime/reference/shuffle_channels.cpp
@@ -0,0 +1,60 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "ngraph/runtime/reference/shuffle_channels.hpp"
+#include "ngraph/runtime/opt_kernel/reshape.hpp"
+
+namespace ngraph
+{
+    namespace runtime
+    {
+        namespace reference
+        {
+            void shuffle_channels(const char* arg,
+                                  char* out,
+                                  const Shape& data_shape,
+                                  size_t elem_size,
+                                  const int64_t axis,
+                                  const int64_t group)
+            {
+                // Input ND tensor of data_shape (ds) is always considered as 4D tensor with the
+                // following shape:
+                // dim 0: ds[0] * ds[1] * ... * ds[axis-1] (or 1 if axis == 0)
+                // dim 1: group
+                // dim 2: ds[axis] / group
+                // dim 3: ds[axis+1] * ds[axis+2] * ... * ds[ds.size()-1]
+                // (or 1 if axis points to last dimension)
+
+                // The representation of ND tensor as 4D tensor doesn't affect flat data order
+                Shape reshaped_input_shape(4, 1);
+                const size_t axis_zb =
+                    axis >= 0 ? axis : axis + data_shape.size(); // Allow negative indices
+                for (size_t i = 0; i < axis_zb; ++i)
+                {
+                    // All dimensions before input channels dim axis
+                    reshaped_input_shape[0] *= data_shape[i];
+                }
+                reshaped_input_shape[1] = group;
+                reshaped_input_shape[2] = data_shape[axis_zb] / group;
+                for (size_t i = axis_zb + 1; i < data_shape.size(); ++i)
+                {
+                    // All dimensions after input channels dim axis
+                    reshaped_input_shape[3] *= data_shape[i];
+                }
+
+                // The two dimensions in the middle are swapped
+                const Shape transposed_shape{reshaped_input_shape[0],
+                                             reshaped_input_shape[2],
+                                             reshaped_input_shape[1],
+                                             reshaped_input_shape[3]};
+                AxisVector axis_vector{0, 2, 1, 3};
+                runtime::opt_kernel::reshape(
+                    arg, out, reshaped_input_shape, axis_vector, transposed_shape, elem_size);
+
+                // Reshaped 4D tensor is interpreted as ND output tensor with original shape of data
+                // input
+            }
+        } // namespace reference
+    }     // namespace runtime
+} // namespace ngraph
diff --git a/ngraph/core/src/op/shuffle_channels.cpp b/ngraph/core/src/op/shuffle_channels.cpp
index 03859b3cb606bd..71683af4030d61 100644
--- a/ngraph/core/src/op/shuffle_channels.cpp
+++ b/ngraph/core/src/op/shuffle_channels.cpp
@@ -10,14 +10,13 @@
 #include "ngraph/op/shuffle_channels.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
 #include "ngraph/runtime/opt_kernel/reshape.hpp"
+#include "ngraph/runtime/reference/shuffle_channels.hpp"
 #include "ngraph/type/element_type.hpp"
 #include "ngraph/type/element_type_traits.hpp"
 
 using namespace std;
 using namespace ngraph;
 
-NGRAPH_SUPPRESS_DEPRECATED_START
-
 NGRAPH_RTTI_DEFINITION(op::v0::ShuffleChannels, "ShuffleChannels", 0);
 
 op::ShuffleChannels::ShuffleChannels(const Output<Node>& data,
@@ -87,7 +86,8 @@ void op::ShuffleChannels::validate_and_infer_types()
     }
     else
     {
-        set_output_type(0, data_type, PartialShape::dynamic());
+        const auto shape = get_input_partial_shape(0);
+        set_output_type(0, data_type, shape);
     }
 }
 
@@ -103,76 +103,19 @@ shared_ptr<Node> op::ShuffleChannels::clone_with_new_inputs(const OutputVector&
     return make_shared<ShuffleChannels>(new_args.at(0), m_axis, m_group);
 }
 
-Shape op::ShuffleChannels::get_pre_shuffle_shape(const Shape& data_shape) const
-{
-    const Shape& ds = data_shape;
-
-    // in general the resulting shape should contain the following values:
-    // [0]: ds[0] * ds[1] * ... * ds[m_axis-1] (or 1 if m_axis == 0)
-    // [1]: m_group
-    // [2]: ds[axis] / m_group
-    // [3]: ds[axis+1] * ds[axis+2] * ... * ds[ds.size()-1] (or 1 if m_axis points to the last elem
-    //                                                       of ds)
-    Shape res(4, 1);
-
-    size_t axis_zb = get_zero_based_axis();
-    for (size_t i = 0; i < axis_zb; ++i)
-    {
-        res[0] *= ds[i];
-    }
-
-    res[1] = m_group;
-    res[2] = ds[axis_zb] / m_group;
-
-    for (size_t i = axis_zb + 1; i < ds.size(); ++i)
-    {
-        res[3] *= ds[i];
-    }
-
-    return res;
-}
-
 bool op::ShuffleChannels::evaluate_shuffle_channels(const HostTensorVector& outputs,
                                                     const HostTensorVector& inputs) const
 {
     const auto arg = inputs[0]->get_data_ptr<const char>();
     auto out = outputs[0]->get_data_ptr<char>();
-    Shape data_shape = inputs[0]->get_shape();
-    const Shape& ds = data_shape;
-    size_t elem_size = inputs[0]->get_element_type().size();
-
-    Shape reshaped_out_shape(4, 1);
-    size_t axis_zb = m_axis >= 0 ? m_axis : m_axis + data_shape.size();
-    for (size_t i = 0; i < axis_zb; ++i)
-    {
-        reshaped_out_shape[0] *= ds[i];
-    }
+    const auto data_shape = inputs[0]->get_shape();
+    const size_t elem_size = inputs[0]->get_element_type().size();
 
-    reshaped_out_shape[1] = m_group;
-    reshaped_out_shape[2] = ds[axis_zb] / m_group;
+    outputs[0]->set_element_type(inputs[0]->get_element_type());
+    outputs[0]->set_shape(data_shape);
 
-    for (size_t i = axis_zb + 1; i < ds.size(); ++i)
-    {
-        reshaped_out_shape[3] *= ds[i];
-    }
-
-    // first reshape from data_shape to reshaped_out_shape is skipped since it doesn't affect
-    // out
-    // data
-
-    Shape transpose_axes_order = {0, 2, 1, 3};
-    Shape transposed_shape(transpose_axes_order.size());
-
-    for (size_t i = 0; i < transpose_axes_order.size(); ++i)
-    {
-        transposed_shape[i] = data_shape.at(transpose_axes_order.at(i));
-    }
-    auto axis_vector = AxisVector{begin(transpose_axes_order), end(transpose_axes_order)};
-    runtime::opt_kernel::reshape(
-        arg, out, reshaped_out_shape, axis_vector, transposed_shape, elem_size);
+    runtime::reference::shuffle_channels(arg, out, data_shape, elem_size, m_axis, m_group);
 
-    // last reshape from transposed_shape to data_shape is skipped since it doesn't affect out
-    // data
     return true;
 }
 bool op::ShuffleChannels::evaluate(const HostTensorVector& outputs,
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index c3c5ab80405e23..2f85b09fc606d3 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -442,6 +442,7 @@ set(MULTI_TEST_SRC
     backend/select.in.cpp
     backend/selu.in.cpp
     backend/shape_of.in.cpp
+    backend/shuffle_channels.in.cpp
     backend/sigmoid.in.cpp
     backend/sign.in.cpp
     backend/sin.in.cpp
diff --git a/ngraph/test/backend/fused_op.in.cpp b/ngraph/test/backend/fused_op.in.cpp
index 4b6bb15cb5171f..c5e89e84e40b0a 100644
--- a/ngraph/test/backend/fused_op.in.cpp
+++ b/ngraph/test/backend/fused_op.in.cpp
@@ -570,66 +570,6 @@ NGRAPH_TEST(${BACKEND_NAME}, DISABLED_grn_2d_with_bias)
     test_case.run();
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, shuffle_channels_simple)
-{
-    const auto data = make_shared<op::Parameter>(element::i32, Shape{1, 15, 2, 2});
-    auto tested_op = make_shared<op::ShuffleChannels>(data, 1, 5);
-    auto function = make_shared<Function>(tested_op, ParameterVector{data});
-
-    auto test_case = test::TestCase<TestEngine>(function);
-
-    std::vector<int32_t> input_data(60);
-    std::iota(std::begin(input_data), std::end(input_data), 0);
-    test_case.add_input(input_data);
-
-    test_case.add_expected_output<int32_t>(
-        Shape{1, 15, 2, 2},
-        {0, 1, 2,  3,  12, 13, 14, 15, 24, 25, 26, 27, 36, 37, 38, 39, 48, 49, 50, 51,
-         4, 5, 6,  7,  16, 17, 18, 19, 28, 29, 30, 31, 40, 41, 42, 43, 52, 53, 54, 55,
-         8, 9, 10, 11, 20, 21, 22, 23, 32, 33, 34, 35, 44, 45, 46, 47, 56, 57, 58, 59});
-
-    test_case.run();
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, shuffle_channels_negative_axis)
-{
-    // in this test the output is the same as in shuffle_channels_simple but
-    // the axis value is negative and the C(channels) value is in a different dimension(0) of the
-    // shape
-    const auto data = make_shared<op::Parameter>(element::i32, Shape{15, 2, 1, 2});
-    auto tested_op = make_shared<op::ShuffleChannels>(data, -4, 5);
-    auto function = make_shared<Function>(tested_op, ParameterVector{data});
-
-    auto test_case = test::TestCase<TestEngine>(function);
-
-    std::vector<int32_t> input_data(60);
-    std::iota(std::begin(input_data), std::end(input_data), 0);
-    test_case.add_input(input_data);
-
-    test_case.add_expected_output<int32_t>(
-        Shape{15, 2, 1, 2},
-        {0, 1, 2,  3,  12, 13, 14, 15, 24, 25, 26, 27, 36, 37, 38, 39, 48, 49, 50, 51,
-         4, 5, 6,  7,  16, 17, 18, 19, 28, 29, 30, 31, 40, 41, 42, 43, 52, 53, 54, 55,
-         8, 9, 10, 11, 20, 21, 22, 23, 32, 33, 34, 35, 44, 45, 46, 47, 56, 57, 58, 59});
-
-    test_case.run();
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, shuffle_channels_float)
-{
-    const auto data = make_shared<op::Parameter>(element::f32, Shape{6, 1, 1, 1});
-    auto tested_op = make_shared<op::ShuffleChannels>(data, 0, 2);
-    auto function = make_shared<Function>(tested_op, ParameterVector{data});
-
-    auto test_case = test::TestCase<TestEngine>(function);
-
-    test_case.add_input<float>({0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
-
-    test_case.add_expected_output<float>(Shape{6, 1, 1, 1}, {0.0f, 3.0f, 1.0f, 4.0f, 2.0f, 5.0f});
-
-    test_case.run();
-}
-
 // TODO: Issue: 37534
 NGRAPH_TEST(${BACKEND_NAME}, DISABLED_squared_difference)
 {
diff --git a/ngraph/test/backend/shuffle_channels.in.cpp b/ngraph/test/backend/shuffle_channels.in.cpp
new file mode 100644
index 00000000000000..16e3afb108c503
--- /dev/null
+++ b/ngraph/test/backend/shuffle_channels.in.cpp
@@ -0,0 +1,192 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "runtime/backend.hpp"
+#include "util/all_close.hpp"
+#include "util/all_close_f.hpp"
+#include "util/engine/test_engines.hpp"
+#include "util/test_case.hpp"
+#include "util/test_control.hpp"
+#include "util/test_tools.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+using TestEngine = test::ENGINE_CLASS_NAME(${BACKEND_NAME});
+
+
+NGRAPH_TEST(${BACKEND_NAME}, shuffle_channels_simple)
+{
+    const auto data = make_shared<op::Parameter>(element::i32, Shape{1, 15, 2, 2});
+    auto tested_op = make_shared<op::ShuffleChannels>(data, 1, 5);
+    auto function = make_shared<Function>(tested_op, ParameterVector{data});
+    auto test_case = test::TestCase<TestEngine>(function);
+
+    std::vector<int32_t> input_data(60);
+    std::iota(std::begin(input_data), std::end(input_data), 0);
+    test_case.add_input(input_data);
+
+    test_case.add_expected_output<int32_t>(
+        Shape{1, 15, 2, 2},
+        {0, 1, 2,  3,  12, 13, 14, 15, 24, 25, 26, 27, 36, 37, 38, 39, 48, 49, 50, 51,
+         4, 5, 6,  7,  16, 17, 18, 19, 28, 29, 30, 31, 40, 41, 42, 43, 52, 53, 54, 55,
+         8, 9, 10, 11, 20, 21, 22, 23, 32, 33, 34, 35, 44, 45, 46, 47, 56, 57, 58, 59});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, shuffle_channels_negative_axis)
+{
+    // In this test the output is the same as in shuffle_channels_simple but
+    // the axis value is negative and the C(channels) value is in a different dimension(0) of the
+    // shape
+    const auto data = make_shared<op::Parameter>(element::i32, Shape{15, 2, 1, 2});
+    auto tested_op = make_shared<op::ShuffleChannels>(data, -4, 5);
+    auto function = make_shared<Function>(tested_op, ParameterVector{data});
+
+    auto test_case = test::TestCase<TestEngine>(function);
+
+    std::vector<int32_t> input_data(60);
+    std::iota(std::begin(input_data), std::end(input_data), 0);
+    test_case.add_input(input_data);
+
+    test_case.add_expected_output<int32_t>(
+        Shape{15, 2, 1, 2},
+        {0, 1, 2,  3,  12, 13, 14, 15, 24, 25, 26, 27, 36, 37, 38, 39, 48, 49, 50, 51,
+         4, 5, 6,  7,  16, 17, 18, 19, 28, 29, 30, 31, 40, 41, 42, 43, 52, 53, 54, 55,
+         8, 9, 10, 11, 20, 21, 22, 23, 32, 33, 34, 35, 44, 45, 46, 47, 56, 57, 58, 59});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, shuffle_channels_float)
+{
+    const auto data = make_shared<op::Parameter>(element::f32, Shape{6, 1, 1, 1});
+    auto tested_op = make_shared<op::ShuffleChannels>(data, 0, 2);
+    auto function = make_shared<Function>(tested_op, ParameterVector{data});
+
+    auto test_case = test::TestCase<TestEngine>(function);
+
+    test_case.add_input<float>({0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
+    test_case.add_expected_output<float>(Shape{6, 1, 1, 1}, {0.0f, 3.0f, 1.0f, 4.0f, 2.0f, 5.0f});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, shuffle_channels_1d)
+{
+    Shape data_shape{15};
+    const auto data = make_shared<op::Parameter>(element::i32, data_shape);
+    auto tested_op = make_shared<op::ShuffleChannels>(data, 0, 5);
+    auto function = make_shared<Function>(tested_op, ParameterVector{data});
+    auto test_case = test::TestCase<TestEngine>(function);
+
+    std::vector<int32_t> input_data{0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14};
+
+    test_case.add_input(input_data);
+    test_case.add_expected_output<int32_t>(
+        data_shape,
+        {0, 3, 6, 9, 12,
+        1, 4, 7, 10, 13,
+        2, 5, 8, 11, 14});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, shuffle_channels_2d)
+{
+    Shape data_shape{15, 4};
+    const auto data = make_shared<op::Parameter>(element::i32, data_shape);
+    auto tested_op = make_shared<op::ShuffleChannels>(data, 0, 5);
+    auto function = make_shared<Function>(tested_op, ParameterVector{data});
+    auto test_case = test::TestCase<TestEngine>(function);
+
+    std::vector<int32_t> input_data{
+        0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
+        20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39,
+        40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59};
+
+    test_case.add_input(input_data);
+    test_case.add_expected_output<int32_t>(
+        data_shape,
+        {0, 1, 2,  3,  12, 13, 14, 15, 24, 25, 26, 27, 36, 37, 38, 39, 48, 49, 50, 51,
+         4, 5, 6,  7,  16, 17, 18, 19, 28, 29, 30, 31, 40, 41, 42, 43, 52, 53, 54, 55,
+         8, 9, 10, 11, 20, 21, 22, 23, 32, 33, 34, 35, 44, 45, 46, 47, 56, 57, 58, 59});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, shuffle_channels_3d)
+{
+    Shape data_shape{15, 2, 2};
+    const auto data = make_shared<op::Parameter>(element::i32, data_shape);
+    auto tested_op = make_shared<op::ShuffleChannels>(data, 0, 5);
+    auto function = make_shared<Function>(tested_op, ParameterVector{data});
+    auto test_case = test::TestCase<TestEngine>(function);
+
+
+    std::vector<int32_t> input_data{
+        0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
+        20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39,
+        40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59};
+
+    test_case.add_input(input_data);
+    test_case.add_expected_output<int32_t>(
+        data_shape,
+        {0, 1, 2,  3,  12, 13, 14, 15, 24, 25, 26, 27, 36, 37, 38, 39, 48, 49, 50, 51,
+         4, 5, 6,  7,  16, 17, 18, 19, 28, 29, 30, 31, 40, 41, 42, 43, 52, 53, 54, 55,
+         8, 9, 10, 11, 20, 21, 22, 23, 32, 33, 34, 35, 44, 45, 46, 47, 56, 57, 58, 59});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, shuffle_channels_5d)
+{
+    Shape data_shape{2, 2, 15, 2, 2};
+    const auto data = make_shared<op::Parameter>(element::i32, data_shape);
+    auto tested_op = make_shared<op::ShuffleChannels>(data, 2, 5);
+    auto function = make_shared<Function>(tested_op, ParameterVector{data});
+    auto test_case = test::TestCase<TestEngine>(function);
+
+    std::vector<int32_t> input_data{
+        0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
+        20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39,
+        40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59,
+
+        0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
+        20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39,
+        40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59,
+
+        0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
+        20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39,
+        40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59,
+
+        0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
+        20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39,
+        40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59};
+
+    test_case.add_input(input_data);
+    test_case.add_expected_output<int32_t>(
+        data_shape,
+        {0, 1, 2,  3,  12, 13, 14, 15, 24, 25, 26, 27, 36, 37, 38, 39, 48, 49, 50, 51,
+         4, 5, 6,  7,  16, 17, 18, 19, 28, 29, 30, 31, 40, 41, 42, 43, 52, 53, 54, 55,
+         8, 9, 10, 11, 20, 21, 22, 23, 32, 33, 34, 35, 44, 45, 46, 47, 56, 57, 58, 59,
+
+         0, 1, 2,  3,  12, 13, 14, 15, 24, 25, 26, 27, 36, 37, 38, 39, 48, 49, 50, 51,
+         4, 5, 6,  7,  16, 17, 18, 19, 28, 29, 30, 31, 40, 41, 42, 43, 52, 53, 54, 55,
+         8, 9, 10, 11, 20, 21, 22, 23, 32, 33, 34, 35, 44, 45, 46, 47, 56, 57, 58, 59,
+
+         0, 1, 2,  3,  12, 13, 14, 15, 24, 25, 26, 27, 36, 37, 38, 39, 48, 49, 50, 51,
+         4, 5, 6,  7,  16, 17, 18, 19, 28, 29, 30, 31, 40, 41, 42, 43, 52, 53, 54, 55,
+         8, 9, 10, 11, 20, 21, 22, 23, 32, 33, 34, 35, 44, 45, 46, 47, 56, 57, 58, 59,
+
+         0, 1, 2,  3,  12, 13, 14, 15, 24, 25, 26, 27, 36, 37, 38, 39, 48, 49, 50, 51,
+         4, 5, 6,  7,  16, 17, 18, 19, 28, 29, 30, 31, 40, 41, 42, 43, 52, 53, 54, 55,
+         8, 9, 10, 11, 20, 21, 22, 23, 32, 33, 34, 35, 44, 45, 46, 47, 56, 57, 58, 59});
+
+    test_case.run();
+}
diff --git a/ngraph/test/runtime/ie/unit_test.manifest b/ngraph/test/runtime/ie/unit_test.manifest
index 80f2421fb41d8f..e76ac6f2b97d91 100644
--- a/ngraph/test/runtime/ie/unit_test.manifest
+++ b/ngraph/test/runtime/ie/unit_test.manifest
@@ -752,11 +752,6 @@ gemm_broadcast_axes_1_input_C
 scale_shift_no_broadcast
 scale_shift
 
-# Cannot cast ngraph node ShuffleChannels to CNNLayer!
-shuffle_channels_simple
-shuffle_channels_negative_axis
-shuffle_channels_float
-
 # Detected op not belonging to opset1!
 onnx_model_quant_conv_linear
 onnx_model_quant_conv_linear_2d
diff --git a/ngraph/test/type_prop/shuffle_channels.cpp b/ngraph/test/type_prop/shuffle_channels.cpp
index 95cd2fd3cca804..3f18ba5023044f 100644
--- a/ngraph/test/type_prop/shuffle_channels.cpp
+++ b/ngraph/test/type_prop/shuffle_channels.cpp
@@ -9,12 +9,116 @@
 using namespace std;
 using namespace ngraph;
 
+TEST(type_prop, shuffle_channels_default_4D)
+{
+    const auto data_input_shape = Shape{3, 9, 4, 5};
+    const auto data = make_shared<op::Parameter>(element::f32, data_input_shape);
+    const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data);
+
+    EXPECT_EQ(shuffle_channels->get_element_type(), element::f32);
+    EXPECT_EQ(shuffle_channels->get_output_partial_shape(0), data_input_shape);
+}
+
+TEST(type_prop, shuffle_channels_basic_4D)
+{
+    const auto data_input_shape = Shape{3, 9, 4, 5};
+    const auto data = make_shared<op::Parameter>(element::f32, data_input_shape);
+    const auto axis = 1;
+    const auto group = 3;
+    const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, axis, group);
+
+    EXPECT_EQ(shuffle_channels->get_element_type(), element::f32);
+    EXPECT_EQ(shuffle_channels->get_output_partial_shape(0), data_input_shape);
+}
+
+TEST(type_prop, shuffle_channels_dynamic_4D)
+{
+    const auto data_input_shape = PartialShape{Dimension::dynamic(), Dimension(3, 9), 4, Dimension(4, 15)};
+    const auto data = make_shared<op::Parameter>(element::f32, data_input_shape);
+    const auto axis = 1;
+    const auto group = 3;
+    const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, axis, group);
+
+    EXPECT_EQ(shuffle_channels->get_element_type(), element::f32);
+    EXPECT_EQ(shuffle_channels->get_output_partial_shape(0), data_input_shape);
+}
+
+TEST(type_prop, shuffle_channels_dynamic_fully)
+{
+    const auto data_input_shape = PartialShape::dynamic();
+    const auto data = make_shared<op::Parameter>(element::f32, data_input_shape);
+    const auto axis = 1;
+    const auto group = 3;
+    const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, axis, group);
+
+    EXPECT_EQ(shuffle_channels->get_element_type(), element::f32);
+    EXPECT_EQ(shuffle_channels->get_output_partial_shape(0), data_input_shape);
+}
+
+TEST(type_prop, shuffle_channels_ND_bigger)
+{
+    {
+        // 5D
+        const auto data_input_shape = Shape{2, 3, 9, 4, 5};
+        const auto data = make_shared<op::Parameter>(element::f32, data_input_shape);
+        const auto axis = 2;
+        const auto group = 3;
+        const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, axis, group);
+
+        EXPECT_EQ(shuffle_channels->get_output_partial_shape(0), data_input_shape);
+    }
+    {
+        // 6D
+        const auto data_input_shape = Shape{6, 2, 3, 9, 4, 5};
+        const auto data = make_shared<op::Parameter>(element::f32, data_input_shape);
+        const auto axis = 3;
+        const auto group = 3;
+        const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, axis, group);
+
+        EXPECT_EQ(shuffle_channels->get_output_partial_shape(0), data_input_shape);
+    }
+}
+
+TEST(type_prop, shuffle_channels_ND_smaller)
+{
+    {
+        // 3D
+        const auto data_input_shape = Shape{5, 4, 9};
+        const auto data = make_shared<op::Parameter>(element::f32, data_input_shape);
+        const auto axis = 2;
+        const auto group = 3;
+        const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, axis, group);
+
+        EXPECT_EQ(shuffle_channels->get_output_partial_shape(0), data_input_shape);
+    }
+    {
+        // 2D
+        const auto data_input_shape = Shape{9, 20};
+        const auto data = make_shared<op::Parameter>(element::f32, data_input_shape);
+        const auto axis = 0;
+        const auto group = 3;
+        const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, axis, group);
+
+        EXPECT_EQ(shuffle_channels->get_output_partial_shape(0), data_input_shape);
+    }
+    {
+        // 1D
+        const auto data_input_shape = Shape{9};
+        const auto data = make_shared<op::Parameter>(element::f32, data_input_shape);
+        const auto axis = 0;
+        const auto group = 3;
+        const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, axis, group);
+
+        EXPECT_EQ(shuffle_channels->get_output_partial_shape(0), data_input_shape);
+    }
+}
+
 TEST(type_prop, shuffle_channels_axis_validation)
 {
     try
     {
         const auto data = make_shared<op::Parameter>(element::f64, Shape{1, 2, 3, 4});
-        const auto shuffle_channels = make_shared<op::ShuffleChannels>(data, -5, 5);
+        const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, -5, 5);
         FAIL() << "ShuffleChannels validation did not work. Op node was created with incorrect "
                   "params.";
     }
@@ -30,7 +134,7 @@ TEST(type_prop, shuffle_channels_negative_axis_calculation)
 {
     const auto data = make_shared<op::Parameter>(element::f64, Shape{1, 2, 3, 4});
 
-    const auto shuffle_channels = make_shared<op::ShuffleChannels>(data, -3, 2);
+    const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, -3, 2);
 
     EXPECT_EQ(shuffle_channels->get_zero_based_axis(), 1);
 }
@@ -40,7 +144,7 @@ TEST(type_prop, shuffle_channels_invalid_input_shape)
     try
     {
         const auto data = make_shared<op::Parameter>(element::f64, Shape{});
-        const auto shuffle_channels = make_shared<op::ShuffleChannels>(data, 0, 1);
+        const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, 0, 1);
         FAIL() << "ShuffleChannels validation did not work. Op node was created with incorrect "
                   "params.";
     }
@@ -56,7 +160,7 @@ TEST(type_prop, shuffle_channels_invalid_groups_value)
     try
     {
         const auto data = make_shared<op::Parameter>(element::f64, Shape{1, 2, 3, 15});
-        const auto shuffle_channels = make_shared<op::ShuffleChannels>(data, -1, 2);
+        const auto shuffle_channels = make_shared<op::v0::ShuffleChannels>(data, -1, 2);
         FAIL() << "ShuffleChannels validation did not work. Op node was created with incorrect "
                   "params.";
     }
diff --git a/ngraph/test/visitors/op/shuffle_channels.cpp b/ngraph/test/visitors/op/shuffle_channels.cpp
index ae1a5867779350..b9b36ba6d736f5 100644
--- a/ngraph/test/visitors/op/shuffle_channels.cpp
+++ b/ngraph/test/visitors/op/shuffle_channels.cpp
@@ -5,28 +5,28 @@
 #include "gtest/gtest.h"
 
 #include "ngraph/ngraph.hpp"
-#include "ngraph/op/util/attr_types.hpp"
 #include "ngraph/opsets/opset1.hpp"
-#include "ngraph/opsets/opset3.hpp"
-#include "ngraph/opsets/opset4.hpp"
-#include "ngraph/opsets/opset5.hpp"
 
 #include "util/visitor.hpp"
 
 using namespace std;
 using namespace ngraph;
 using ngraph::test::NodeBuilder;
-using ngraph::test::ValueMap;
 
 TEST(attributes, shuffle_channels_op)
 {
-    NodeBuilder::get_ops().register_factory<opset1::ShuffleChannels>();
+    using ShuffleChannels = opset1::ShuffleChannels;
+
+    NodeBuilder::get_ops().register_factory<ShuffleChannels>();
     auto data = make_shared<op::Parameter>(element::i32, Shape{200});
     auto axis = 0;
     auto groups = 2;
-    auto shuffle_channels = make_shared<opset1::ShuffleChannels>(data, axis, groups);
+    auto shuffle_channels = make_shared<ShuffleChannels>(data, axis, groups);
     NodeBuilder builder(shuffle_channels);
-    auto g_shuffle_channels = as_type_ptr<opset1::ShuffleChannels>(builder.create());
+    auto g_shuffle_channels = as_type_ptr<ShuffleChannels>(builder.create());
+
+    const auto expected_attr_count = 2;
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
 
     EXPECT_EQ(g_shuffle_channels->get_axis(), shuffle_channels->get_axis());
     EXPECT_EQ(g_shuffle_channels->get_group(), shuffle_channels->get_group());

From 6deec50b0b599568b93a8b725e2cef0fa292cd28 Mon Sep 17 00:00:00 2001
From: Patryk Elszkowski <patryk.elszkowski@intel.com>
Date: Tue, 15 Jun 2021 06:07:18 +0200
Subject: [PATCH 077/178] Reshape OP: add SLT for special `-1` value in new
 shape dimensions (#5648)

* add test for special `-1` value in new shape dimensions

* add ticket with next steps
---
 .../single_layer_tests/reshape.cpp            | 75 +++++++++++--------
 .../serialization/single_layer/reshape.cpp    |  8 +-
 .../single_layer_tests/reshape.cpp            | 28 +++++--
 .../single_layer_tests/reshape.cpp            | 26 +++++--
 .../single_layer/reshape.hpp                  | 64 +++++++++++-----
 .../src/single_layer/reshape.cpp              | 47 +++++++++++-
 6 files changed, 182 insertions(+), 66 deletions(-)

diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/reshape.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/reshape.cpp
index 85313f410d3606..e0c986ad8b52e5 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/reshape.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/reshape.cpp
@@ -2,43 +2,58 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+#include "single_layer_tests/reshape.hpp"
+
 #include <vector>
 
-#include "single_layer_tests/reshape.hpp"
 #include "common_test_utils/test_constants.hpp"
 
 using namespace LayerTestsDefinitions;
 
 namespace {
 const std::vector<InferenceEngine::Precision> netPrecisions = {
-        InferenceEngine::Precision::FP32,
+    InferenceEngine::Precision::FP32,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheckDynBatch, ReshapeLayerTest,
-        ::testing::Combine(
-                ::testing::Values(true),
-                ::testing::ValuesIn(netPrecisions),
-                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
-                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
-                ::testing::Values(InferenceEngine::Layout::ANY),
-                ::testing::Values(InferenceEngine::Layout::ANY),
-                ::testing::Values(std::vector<size_t>({30, 30, 30, 30})),
-                ::testing::Values(std::vector<size_t>({30, 30, 30, 30})),
-                ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
-                ::testing::Values(std::map<std::string, std::string>({}))),
-                ReshapeLayerTest::getTestCaseName);
-
-INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTest,
-        ::testing::Combine(
-                ::testing::Values(true),
-                ::testing::ValuesIn(netPrecisions),
-                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
-                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
-                ::testing::Values(InferenceEngine::Layout::ANY),
-                ::testing::Values(InferenceEngine::Layout::ANY),
-                ::testing::Values(std::vector<size_t>({10, 10, 10, 10})),
-                ::testing::Values(std::vector<size_t>({10, 0, 100})),
-                ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
-                ::testing::Values(std::map<std::string, std::string>({}))),
-                ReshapeLayerTest::getTestCaseName);
-}  // namespace
\ No newline at end of file
+INSTANTIATE_TEST_CASE_P(
+    smoke_ReshapeCheckDynBatch, ReshapeLayerTestRevise,
+    ::testing::Combine(
+        ::testing::Values(true), ::testing::ValuesIn(netPrecisions),
+        ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+        ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(std::vector<size_t>({30, 30, 30, 30})),
+        ::testing::Values(std::vector<int64_t>({30, 30, 30, 30})),
+        ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
+        ::testing::Values(std::map<std::string, std::string>({}))),
+    ReshapeLayerTestRevise::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(
+    smoke_ReshapeCheck, ReshapeLayerTestRevise,
+    ::testing::Combine(
+        ::testing::Values(true), ::testing::ValuesIn(netPrecisions),
+        ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+        ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(std::vector<size_t>({10, 10, 10, 10})),
+        ::testing::Values(std::vector<int64_t>({10, 0, 100})),
+        ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
+        ::testing::Values(std::map<std::string, std::string>({}))),
+    ReshapeLayerTestRevise::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(
+    smoke_ReshapeCheckNegative, ReshapeLayerTestRevise,
+    ::testing::Combine(
+        ::testing::Values(true), ::testing::ValuesIn(netPrecisions),
+        ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+        ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(std::vector<size_t>({10, 10, 10, 10})),
+        ::testing::Values(std::vector<int64_t>({10, -1, 100})),
+        ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
+        ::testing::Values(std::map<std::string, std::string>({}))),
+    ReshapeLayerTestRevise::getTestCaseName);
+}  // namespace
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/reshape.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/reshape.cpp
index c7b60da690fcd5..9ba9aec704b19d 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/reshape.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/reshape.cpp
@@ -10,7 +10,7 @@
 using namespace LayerTestsDefinitions;
 
 namespace {
-    TEST_P(ReshapeLayerTest, Serialize) {
+    TEST_P(ReshapeLayerTestRevise, Serialize) {
         Serialize();
     }
 
@@ -19,7 +19,7 @@ namespace {
             InferenceEngine::Precision::FP16
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_ReshapeSerialization, ReshapeLayerTest,
+    INSTANTIATE_TEST_CASE_P(smoke_ReshapeSerialization, ReshapeLayerTestRevise,
             ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
@@ -28,8 +28,8 @@ namespace {
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(std::vector<size_t>({30, 30, 30, 30})),
-                ::testing::Values(std::vector<size_t>({30, 30, 30, 30})),
+                ::testing::Values(std::vector<int64_t>({30, 30, 30, 30})),
                 ::testing::Values(CommonTestUtils::DEVICE_CPU),
                 ::testing::Values(std::map<std::string, std::string>({{CONFIG_KEY(DYN_BATCH_ENABLED), CONFIG_VALUE(YES)}}))),
-                ReshapeLayerTest::getTestCaseName);
+                ReshapeLayerTestRevise::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reshape.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reshape.cpp
index 0bdef6426580a1..28419ec971fca7 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reshape.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reshape.cpp
@@ -15,7 +15,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP16
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheckDynBatch, ReshapeLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheckDynBatch, ReshapeLayerTestRevise,
         ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
@@ -24,12 +24,26 @@ INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheckDynBatch, ReshapeLayerTest,
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(std::vector<size_t>({30, 30, 30, 30})),
-                ::testing::Values(std::vector<size_t>({30, 30, 30, 30})),
+                ::testing::Values(std::vector<int64_t>({30, 30, 30, 30})),
                 ::testing::Values(CommonTestUtils::DEVICE_CPU),
                 ::testing::Values(std::map<std::string, std::string>({{CONFIG_KEY(DYN_BATCH_ENABLED), CONFIG_VALUE(YES)}}))),
-                ReshapeLayerTest::getTestCaseName);
+                ReshapeLayerTestRevise::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTestRevise,
+        ::testing::Combine(
+                ::testing::Values(true),
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(std::vector<size_t>({10, 10, 10, 10})),
+                ::testing::Values(std::vector<int64_t>({10, 0, 100})),
+                ::testing::Values(CommonTestUtils::DEVICE_CPU),
+                ::testing::Values(std::map<std::string, std::string>({}))),
+                ReshapeLayerTestRevise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheckNegative, ReshapeLayerTestRevise,
         ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
@@ -38,8 +52,8 @@ INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTest,
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(std::vector<size_t>({10, 10, 10, 10})),
-                ::testing::Values(std::vector<size_t>({10, 0, 100})),
+                ::testing::Values(std::vector<int64_t>({10, -1, 100})),
                 ::testing::Values(CommonTestUtils::DEVICE_CPU),
                 ::testing::Values(std::map<std::string, std::string>({}))),
-                ReshapeLayerTest::getTestCaseName);
-}  // namespace
\ No newline at end of file
+                ReshapeLayerTestRevise::getTestCaseName);
+}  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reshape.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reshape.cpp
index a304c345776a06..84d30807341593 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reshape.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reshape.cpp
@@ -16,7 +16,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 };
 
 //TODO: Issue : - 28981
-INSTANTIATE_TEST_CASE_P(DISABLE_smoke_ReshapeCheckDynBatch, ReshapeLayerTest,
+INSTANTIATE_TEST_CASE_P(DISABLE_smoke_ReshapeCheckDynBatch, ReshapeLayerTestRevise,
         ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
@@ -25,12 +25,12 @@ INSTANTIATE_TEST_CASE_P(DISABLE_smoke_ReshapeCheckDynBatch, ReshapeLayerTest,
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(std::vector<size_t>({1, 16, 16, 16})),
-                ::testing::Values(std::vector<size_t>({1, 0, 256})),
+                ::testing::Values(std::vector<int64_t>({1, 0, 256})),
                  ::testing::Values(CommonTestUtils::DEVICE_GPU),
                 ::testing::Values(std::map<std::string, std::string>({{CONFIG_KEY(DYN_BATCH_ENABLED), CONFIG_VALUE(YES)}}))),
-                ReshapeLayerTest::getTestCaseName);
+                ReshapeLayerTestRevise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTestRevise,
         ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
@@ -39,8 +39,22 @@ INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTest,
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(std::vector<size_t>({10, 10, 10, 10})),
-                ::testing::Values(std::vector<size_t>({10, 0, 100})),
+                ::testing::Values(std::vector<int64_t>({10, 0, 100})),
                 ::testing::Values(CommonTestUtils::DEVICE_GPU),
                 ::testing::Values(std::map<std::string, std::string>({}))),
-                ReshapeLayerTest::getTestCaseName);
+                ReshapeLayerTestRevise::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheckNegative, ReshapeLayerTestRevise,
+        ::testing::Combine(
+                ::testing::Values(true),
+                ::testing::ValuesIn(netPrecisions),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(InferenceEngine::Layout::ANY),
+                ::testing::Values(std::vector<size_t>({10, 10, 10, 10})),
+                ::testing::Values(std::vector<int64_t>({10, -1, 100})),
+                ::testing::Values(CommonTestUtils::DEVICE_GPU),
+                ::testing::Values(std::map<std::string, std::string>({}))),
+                ReshapeLayerTestRevise::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/reshape.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/reshape.hpp
index d0acf57931dbfa..9ddeeed509aa48 100644
--- a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/reshape.hpp
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/reshape.hpp
@@ -4,36 +4,66 @@
 
 #pragma once
 
-#include <tuple>
+#include <memory>
 #include <string>
+#include <tuple>
 #include <vector>
-#include <memory>
+
 #include "ngraph_functions/builders.hpp"
 #include "ngraph_functions/utils/ngraph_helpers.hpp"
-
 #include "shared_test_classes/base/layer_test_utils.hpp"
 
 namespace LayerTestsDefinitions {
-typedef std::tuple<
-        bool,                               // SpecialZero
-        InferenceEngine::Precision,         // Network precision
-        InferenceEngine::Precision,         // Input precision
-        InferenceEngine::Precision,         // Output precision
-        InferenceEngine::Layout,            // Input layout
-        InferenceEngine::Layout,            // Output layout
-        std::vector<size_t>,                // Input shapes
-        std::vector<size_t>,                // OutForm Shapes
-        std::string,                        // Device name
-        std::map<std::string, std::string>  // Config
-> reshapeParams;
 
+//TODO: remove this alias when ticket 57975 is done - ticket: 57976
+typedef std::tuple<bool,                               // SpecialZero
+                   InferenceEngine::Precision,         // Network precision
+                   InferenceEngine::Precision,         // Input precision
+                   InferenceEngine::Precision,         // Output precision
+                   InferenceEngine::Layout,            // Input layout
+                   InferenceEngine::Layout,            // Output layout
+                   std::vector<size_t>,                // Input shapes
+                   std::vector<size_t>,                // OutForm Shapes
+                   std::string,                        // Device name
+                   std::map<std::string, std::string>  // Config
+                   >
+    reshapeParams;
+
+//TODO: remove this class when ticket 57975 is done - ticket: 57976
 class ReshapeLayerTest : public testing::WithParamInterface<reshapeParams>,
                          virtual public LayerTestsUtils::LayerTestsCommon {
 public:
-    static std::string getTestCaseName(testing::TestParamInfo<reshapeParams> obj);
+    static std::string getTestCaseName(
+        testing::TestParamInfo<reshapeParams> obj);
 
 protected:
     void SetUp() override;
 };
 
-}  // namespace LayerTestsDefinitions
\ No newline at end of file
+
+//TODO: use this alias in all dependencies - ticket: 57975
+typedef std::tuple<bool,                               // SpecialZero
+                   InferenceEngine::Precision,         // Network precision
+                   InferenceEngine::Precision,         // Input precision
+                   InferenceEngine::Precision,         // Output precision
+                   InferenceEngine::Layout,            // Input layout
+                   InferenceEngine::Layout,            // Output layout
+                   std::vector<size_t>,                // Input shapes
+                   std::vector<int64_t>,               // OutForm Shapes
+                   std::string,                        // Device name
+                   std::map<std::string, std::string>  // Config
+                   >
+    reshapeParamsRevise;
+
+//TODO: use this class in all dependencies - ticket: 57975
+class ReshapeLayerTestRevise
+    : public testing::WithParamInterface<reshapeParamsRevise>,
+      virtual public LayerTestsUtils::LayerTestsCommon {
+public:
+    static std::string getTestCaseName(
+        testing::TestParamInfo<reshapeParamsRevise> obj);
+
+protected:
+    void SetUp() override;
+};
+}  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/shared_test_classes/src/single_layer/reshape.cpp b/inference-engine/tests/functional/shared_test_classes/src/single_layer/reshape.cpp
index 3063b01b27b1ea..708c40ebbe058d 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/single_layer/reshape.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/single_layer/reshape.cpp
@@ -5,7 +5,7 @@
 #include "shared_test_classes/single_layer/reshape.hpp"
 
 namespace LayerTestsDefinitions {
-    std::string ReshapeLayerTest::getTestCaseName(testing::TestParamInfo<reshapeParams> obj) {
+std::string ReshapeLayerTest::getTestCaseName(testing::TestParamInfo<reshapeParams> obj) {
     InferenceEngine::Precision netPrecision;
     InferenceEngine::Precision inPrc, outPrc;
     InferenceEngine::Layout inLayout, outLayout;
@@ -44,4 +44,47 @@ void ReshapeLayerTest::SetUp() {
     ngraph::ResultVector results{std::make_shared<ngraph::opset1::Result>(reshape)};
     function = std::make_shared<ngraph::Function>(results, paramsIn, "Reshape");
 }
-}  // namespace LayerTestsDefinitions
\ No newline at end of file
+
+std::string ReshapeLayerTestRevise::getTestCaseName(testing::TestParamInfo<reshapeParamsRevise> obj) {
+    InferenceEngine::Precision netPrecision;
+    InferenceEngine::Precision inPrc, outPrc;
+    InferenceEngine::Layout inLayout, outLayout;
+    InferenceEngine::SizeVector inputShapes;
+    std::vector<int64_t> outFormShapes;
+    std::string targetDevice;
+    std::map<std::string, std::string> config;
+    bool specialZero;
+    std::tie(specialZero, netPrecision, inPrc, outPrc, inLayout, outLayout, inputShapes, outFormShapes, targetDevice, config) = obj.param;
+    std::ostringstream result;
+    result << "IS=" << CommonTestUtils::vec2str(inputShapes) << "_";
+    result << "OS=" << CommonTestUtils::vec2str(outFormShapes) << "_";
+    result << "specialZero=" << specialZero << "_";
+    result << "netPRC=" << netPrecision.name() << "_";
+    result << "inPRC=" << inPrc.name() << "_";
+    result << "outPRC=" << outPrc.name() << "_";
+    result << "inL=" << inLayout << "_";
+    result << "outL=" << outLayout << "_";
+    result << "trgDev=" << targetDevice;
+    return result.str();
+}
+
+void ReshapeLayerTestRevise::SetUp() {
+    InferenceEngine::SizeVector inputShapes;
+    std::vector<int64_t> outFormShapes;
+    bool specialZero;
+    InferenceEngine::Precision netPrecision;
+    std::tie(specialZero, netPrecision, inPrc, outPrc, inLayout, outLayout, inputShapes, outFormShapes, targetDevice, configuration) =
+        this->GetParam();
+    auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
+    auto paramsIn = ngraph::builder::makeParams(ngPrc, {inputShapes});
+    auto paramIn = ngraph::helpers::convert2OutputVector(
+            ngraph::helpers::castOps2Nodes<ngraph::op::Parameter>(paramsIn));
+    auto constNode = std::make_shared<ngraph::opset1::Constant>(
+            ngraph::element::Type_t::i64, ngraph::Shape{outFormShapes.size()}, outFormShapes);
+    auto reshape = std::dynamic_pointer_cast<ngraph::opset1::Reshape>(
+            std::make_shared<ngraph::opset1::Reshape>(paramIn[0], constNode, specialZero));
+    ngraph::ResultVector results{std::make_shared<ngraph::opset1::Result>(reshape)};
+    function = std::make_shared<ngraph::Function>(results, paramsIn, "Reshape");
+}
+
+}  // namespace LayerTestsDefinitions

From 772465da1ebd72f95298bf876e649e5057f4329d Mon Sep 17 00:00:00 2001
From: Szymon Durawa <szymon.durawa@intel.com>
Date: Tue, 15 Jun 2021 06:08:10 +0200
Subject: [PATCH 078/178] Add output shape and output padding for Convolution
 Backprop SLTs. (#5576)

* Create output shape for Convoution Backprop SLTs.

* Add output_padding attribute to SLT scope.

* Introduce SLT for Serializaton.

* Introduce new test layer class ConvolutionBackpropLayerTest which contains output_padding attribute and output_shape input. Old one is deprecated, but cannot be removed due to kmb plugin dependency.

* Add ConvolutionBackpropDataLayerTest into TEST_P.

* ConvolutionBackpropDataLayerTest left as legacy class used by kmb_plugin.

* Remove redundant variables.

* Switch to new API for gpu SLTs.

* Remove legacy API.

* Introduce legacy API to match dependency for KMB and ARM plugins.

* Create test cases for output_padding attribute.

* Fixing smoke_Deconv tests.
---
 .../single_layer/convolution_backprop.cpp     |  58 ++++++
 .../convolution_backprop_data.cpp             | 168 ++++++++++++++++--
 .../convolution_backprop_data.cpp             |  58 ++++--
 .../convolution_backprop_data.cpp             | 136 ++++++++++++--
 .../convolution_backprop.hpp                  |  15 ++
 .../convolution_backprop_data.hpp             |   1 +
 .../single_layer/convolution_backprop.hpp     |  49 +++++
 .../convolution_backprop_data.hpp             |   2 +
 .../src/single_layer/convolution_backprop.cpp |  71 ++++++++
 .../convolution_backprop_data.cpp             |   4 +-
 .../include/ngraph_functions/builders.hpp     |  17 ++
 .../src/convolution_backprop_data.cpp         |  45 ++++-
 12 files changed, 579 insertions(+), 45 deletions(-)
 create mode 100644 inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution_backprop.cpp
 create mode 100644 inference-engine/tests/functional/plugin/shared/include/single_layer_tests/convolution_backprop.hpp
 create mode 100644 inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/convolution_backprop.hpp
 create mode 100644 inference-engine/tests/functional/shared_test_classes/src/single_layer/convolution_backprop.cpp

diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution_backprop.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution_backprop.cpp
new file mode 100644
index 00000000000000..2f9383ad914475
--- /dev/null
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution_backprop.cpp
@@ -0,0 +1,58 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <vector>
+
+#include "shared_test_classes/single_layer/convolution_backprop.hpp"
+
+using namespace LayerTestsDefinitions;
+
+namespace {
+
+TEST_P(ConvolutionBackpropLayerTest, Serialize) {
+    Serialize();
+}
+
+const std::vector<InferenceEngine::Precision> precisions = {
+    InferenceEngine::Precision::FP64, InferenceEngine::Precision::FP32,
+    InferenceEngine::Precision::FP16, InferenceEngine::Precision::BF16,
+    InferenceEngine::Precision::I8,   InferenceEngine::Precision::I16,
+    InferenceEngine::Precision::I32,  InferenceEngine::Precision::I64,
+    InferenceEngine::Precision::U8,   InferenceEngine::Precision::U16,
+    InferenceEngine::Precision::U32,  InferenceEngine::Precision::U64,
+};
+const std::vector<std::vector<size_t>> kernels = {{3, 3}};
+const std::vector<std::vector<size_t>> strides = {{1, 1}};
+const std::vector<std::vector<ptrdiff_t>> padBegins = {{0, 0}};
+const std::vector<std::vector<ptrdiff_t>> padEnds = {{0, 0}};
+const std::vector<std::vector<size_t>> dilations = {{1, 1}};
+const std::vector<std::vector<ptrdiff_t>> outPadding = {{}, {1, 1}};
+const std::vector<size_t> numOutChannels = {8, 16};
+const std::vector<ngraph::op::PadType> pad_types = {
+    ngraph::op::PadType::EXPLICIT, ngraph::op::PadType::VALID,
+    ngraph::op::PadType::SAME_LOWER, ngraph::op::PadType::SAME_UPPER};
+const auto inputShapes = std::vector<size_t>({1, 16, 20, 20});
+const std::vector<std::vector<size_t >> emptyOutputShape = {{}};
+
+const auto convolutionBackpropData2DParams = ::testing::Combine(
+    ::testing::ValuesIn(kernels), ::testing::ValuesIn(strides),
+    ::testing::ValuesIn(padBegins), ::testing::ValuesIn(padEnds),
+    ::testing::ValuesIn(dilations), ::testing::ValuesIn(numOutChannels),
+    ::testing::ValuesIn(pad_types), ::testing::ValuesIn(outPadding));
+
+INSTANTIATE_TEST_CASE_P(
+    smoke_convolutionBackpropData2D_Serialization, ConvolutionBackpropLayerTest,
+    ::testing::Combine(
+        convolutionBackpropData2DParams,
+        ::testing::ValuesIn(precisions),
+        ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+        ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(InferenceEngine::Layout::ANY),
+        ::testing::Values(inputShapes),
+        ::testing::ValuesIn(emptyOutputShape),
+        ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+    ConvolutionBackpropLayerTest::getTestCaseName);
+
+}   // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
index a8b4c01497fb81..1a5f3885c9364b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
@@ -4,7 +4,7 @@
 
 #include <vector>
 
-#include "single_layer_tests/convolution_backprop_data.hpp"
+#include "single_layer_tests/convolution_backprop.hpp"
 #include "common_test_utils/test_constants.hpp"
 
 using namespace LayerTestsDefinitions;
@@ -17,6 +17,8 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 };
 
 const std::vector<size_t> numOutChannels = {1, 5, 16};
+const std::vector<std::vector<size_t >> emptyOutputShape = {{}};
+const std::vector<std::vector<ptrdiff_t >> emptyOutputPadding = {{}};
 
 /* ============= 2D ConvolutionBackpropData ============= */
 const std::vector<std::vector<size_t >> inputShapes2D = {{1, 3, 30, 30},
@@ -35,7 +37,8 @@ const auto conv2DParams_ExplicitPadding = ::testing::Combine(
         ::testing::ValuesIn(padEnds2D),
         ::testing::ValuesIn(dilations2D),
         ::testing::ValuesIn(numOutChannels),
-        ::testing::Values(ngraph::op::PadType::EXPLICIT)
+        ::testing::Values(ngraph::op::PadType::EXPLICIT),
+        ::testing::ValuesIn(emptyOutputPadding)
 );
 const auto conv2DParams_AutoPadValid = ::testing::Combine(
         ::testing::ValuesIn(kernels2D),
@@ -44,10 +47,11 @@ const auto conv2DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(std::vector<ptrdiff_t>({0, 0})),
         ::testing::ValuesIn(dilations2D),
         ::testing::ValuesIn(numOutChannels),
-        ::testing::Values(ngraph::op::PadType::VALID)
+        ::testing::Values(ngraph::op::PadType::VALID),
+        ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, ConvolutionBackpropDataLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -56,10 +60,11 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, Convolu
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::ValuesIn(inputShapes2D),
+                                ::testing::ValuesIn(emptyOutputShape),
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
-                        ConvolutionBackpropDataLayerTest::getTestCaseName);
+                        ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, ConvolutionBackpropDataLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -68,8 +73,75 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, Convolutio
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::ValuesIn(inputShapes2D),
+                                ::testing::ValuesIn(emptyOutputShape),
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
-                        ConvolutionBackpropDataLayerTest::getTestCaseName);
+                        ConvolutionBackpropLayerTest::getTestCaseName);
+
+const std::vector<std::vector<size_t >> inputShape2D = {{1, 3, 9, 12}};
+const std::vector<std::vector<size_t >> outputShapes2D = {{6, 6}, {4, 9}};
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_OutputShapeDefined, ConvolutionBackpropLayerTest,
+                        ::testing::Combine(
+                                conv2DParams_AutoPadValid,
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::ValuesIn(inputShape2D),
+                                ::testing::ValuesIn(outputShapes2D),
+                                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                        ConvolutionBackpropLayerTest::getTestCaseName);
+
+const std::vector<std::vector<ptrdiff_t>> outputPadding2D = {{1, 1}, {2, 2}};
+const std::vector<std::vector<size_t >> testStrides2D = {{3, 3}};
+
+const auto conv2DParams_ExplicitPadding_output_padding = ::testing::Combine(
+        ::testing::ValuesIn(kernels2D),
+        ::testing::ValuesIn(testStrides2D),
+        ::testing::ValuesIn(padBegins2D),
+        ::testing::ValuesIn(padEnds2D),
+        ::testing::ValuesIn(dilations2D),
+        ::testing::ValuesIn(numOutChannels),
+        ::testing::Values(ngraph::op::PadType::EXPLICIT),
+        ::testing::ValuesIn(outputPadding2D)
+);
+const auto conv2DParams_AutoPadValid_output_padding = ::testing::Combine(
+        ::testing::ValuesIn(kernels2D),
+        ::testing::ValuesIn(testStrides2D),
+        ::testing::Values(std::vector<ptrdiff_t>({0, 0})),
+        ::testing::Values(std::vector<ptrdiff_t>({0, 0})),
+        ::testing::ValuesIn(dilations2D),
+        ::testing::ValuesIn(numOutChannels),
+        ::testing::Values(ngraph::op::PadType::VALID),
+        ::testing::ValuesIn(outputPadding2D)
+);
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+                        ::testing::Combine(
+                                conv2DParams_AutoPadValid_output_padding,
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::ValuesIn(inputShapes2D),
+                                ::testing::ValuesIn(emptyOutputShape),
+                                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                        ConvolutionBackpropLayerTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+                        ::testing::Combine(
+                                conv2DParams_ExplicitPadding_output_padding,
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::ValuesIn(inputShapes2D),
+                                ::testing::ValuesIn(emptyOutputShape),
+                                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                        ConvolutionBackpropLayerTest::getTestCaseName);
 
 /* ============= 3D ConvolutionBackpropData ============= */
 const std::vector<std::vector<size_t >> inputShapes3D = {{1, 3, 10, 10, 10},
@@ -88,7 +160,8 @@ const auto conv3DParams_ExplicitPadding = ::testing::Combine(
         ::testing::ValuesIn(padEnds3D),
         ::testing::ValuesIn(dilations3D),
         ::testing::ValuesIn(numOutChannels),
-        ::testing::Values(ngraph::op::PadType::EXPLICIT)
+        ::testing::Values(ngraph::op::PadType::EXPLICIT),
+        ::testing::ValuesIn(emptyOutputPadding)
 );
 const auto conv3DParams_AutoPadValid = ::testing::Combine(
         ::testing::ValuesIn(kernels3D),
@@ -97,10 +170,11 @@ const auto conv3DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(std::vector<ptrdiff_t>({0, 0, 0})),
         ::testing::ValuesIn(dilations3D),
         ::testing::ValuesIn(numOutChannels),
-        ::testing::Values(ngraph::op::PadType::VALID)
+        ::testing::Values(ngraph::op::PadType::VALID),
+        ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, ConvolutionBackpropDataLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -109,10 +183,11 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, Convolu
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::ValuesIn(inputShapes3D),
+                                ::testing::ValuesIn(emptyOutputShape),
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
-                        ConvolutionBackpropDataLayerTest::getTestCaseName);
+                        ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, ConvolutionBackpropDataLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -121,7 +196,74 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, Convolutio
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::ValuesIn(inputShapes3D),
+                                ::testing::ValuesIn(emptyOutputShape),
+                                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                        ConvolutionBackpropLayerTest::getTestCaseName);
+
+const std::vector<std::vector<size_t >> inputShape3D = {{1, 3, 10, 10, 10}};
+const std::vector<std::vector<size_t >> outputShapes3D = {{8, 8, 8}, {10, 10, 10}};
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_OutputShapeDefined, ConvolutionBackpropLayerTest,
+                        ::testing::Combine(
+                                conv3DParams_AutoPadValid,
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::ValuesIn(inputShape3D),
+                                ::testing::ValuesIn(outputShapes3D),
+                                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                        ConvolutionBackpropLayerTest::getTestCaseName);
+
+const std::vector<std::vector<ptrdiff_t>> outputPadding3D = {{1, 1, 1}, {2, 2, 2}};
+const std::vector<std::vector<size_t >> testStrides3D = {{3, 3, 3}};
+
+const auto conv3DParams_ExplicitPadding_output_padding = ::testing::Combine(
+        ::testing::ValuesIn(kernels3D),
+        ::testing::ValuesIn(testStrides3D),
+        ::testing::ValuesIn(padBegins3D),
+        ::testing::ValuesIn(padEnds3D),
+        ::testing::ValuesIn(dilations3D),
+        ::testing::ValuesIn(numOutChannels),
+        ::testing::Values(ngraph::op::PadType::EXPLICIT),
+        ::testing::ValuesIn(outputPadding3D)
+);
+const auto conv3DParams_AutoPadValid_output_padding = ::testing::Combine(
+        ::testing::ValuesIn(kernels3D),
+        ::testing::ValuesIn(testStrides3D),
+        ::testing::Values(std::vector<ptrdiff_t>({0, 0, 0})),
+        ::testing::Values(std::vector<ptrdiff_t>({0, 0, 0})),
+        ::testing::ValuesIn(dilations3D),
+        ::testing::ValuesIn(numOutChannels),
+        ::testing::Values(ngraph::op::PadType::VALID),
+        ::testing::ValuesIn(outputPadding3D)
+);
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+                        ::testing::Combine(
+                                conv3DParams_AutoPadValid_output_padding,
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::ValuesIn(inputShapes3D),
+                                ::testing::ValuesIn(emptyOutputShape),
+                                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                        ConvolutionBackpropLayerTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+                        ::testing::Combine(
+                                conv3DParams_ExplicitPadding_output_padding,
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::ValuesIn(inputShapes3D),
+                                ::testing::ValuesIn(emptyOutputShape),
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
-                        ConvolutionBackpropDataLayerTest::getTestCaseName);
+                        ConvolutionBackpropLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution_backprop_data.cpp
index 5fdcbef5747dad..f1d144f666ebf9 100755
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution_backprop_data.cpp
@@ -8,18 +8,18 @@
 #include "shared_test_classes/base/layer_test_utils.hpp"
 #include "ngraph_functions/utils/ngraph_helpers.hpp"
 #include "ngraph_functions/builders.hpp"
-#include <shared_test_classes/single_layer/convolution_backprop_data.hpp>
+#include <shared_test_classes/single_layer/convolution_backprop.hpp>
 
 
 using namespace InferenceEngine;
 using namespace CPUTestUtils;
 
 namespace CPULayerTestsDefinitions {
-using LayerTestsDefinitions::convBackpropDataSpecificParams;
-using LayerTestsDefinitions::convBackpropDataLayerTestParamsSet;
+using LayerTestsDefinitions::convBackpropSpecificParams;
+using LayerTestsDefinitions::convBackpropLayerTestParamsSet;
 
 typedef std::tuple<
-    convBackpropDataLayerTestParamsSet,
+    convBackpropLayerTestParamsSet,
     CPUSpecificParams,
     fusingSpecificParams,
     std::map<std::string, std::string> > deconvLayerCPUTestParamsSet;
@@ -28,14 +28,14 @@ class DeconvolutionLayerCPUTest : public testing::WithParamInterface<deconvLayer
     virtual public LayerTestsUtils::LayerTestsCommon, public CpuTestWithFusing {
 public:
     static std::string getTestCaseName(testing::TestParamInfo<deconvLayerCPUTestParamsSet> obj) {
-        convBackpropDataLayerTestParamsSet basicParamsSet;
+        convBackpropLayerTestParamsSet basicParamsSet;
         CPUSpecificParams cpuParams;
         fusingSpecificParams fusingParams;
         std::map<std::string, std::string> additionalConfig;
         std::tie(basicParamsSet, cpuParams, fusingParams, additionalConfig) = obj.param;
 
         std::ostringstream result;
-        result << LayerTestsDefinitions::ConvolutionBackpropDataLayerTest::getTestCaseName(testing::TestParamInfo<convBackpropDataLayerTestParamsSet>(
+        result << LayerTestsDefinitions::ConvolutionBackpropLayerTest::getTestCaseName(testing::TestParamInfo<convBackpropLayerTestParamsSet>(
             basicParamsSet, 0));
 
         result << CPUTestsBase::getTestCaseName(cpuParams);
@@ -52,7 +52,7 @@ class DeconvolutionLayerCPUTest : public testing::WithParamInterface<deconvLayer
     }
 protected:
     void SetUp() override {
-        convBackpropDataLayerTestParamsSet basicParamsSet;
+        convBackpropLayerTestParamsSet basicParamsSet;
         CPUSpecificParams cpuParams;
         fusingSpecificParams fusingParams;
         std::map<std::string, std::string> additionalConfig;
@@ -63,10 +63,11 @@ class DeconvolutionLayerCPUTest : public testing::WithParamInterface<deconvLayer
         std::tie(inFmts, outFmts, priority, selectedType) = cpuParams;
         std::tie(postOpMgrPtr, fusedOps) = fusingParams;
 
-        convBackpropDataSpecificParams convParams;
+        convBackpropSpecificParams convParams;
         std::vector<size_t> inputShape;
+        std::vector<size_t> outputShape;
         auto netPrecision = InferenceEngine::Precision::UNSPECIFIED;
-        std::tie(convParams, netPrecision, inPrc, outPrc, inLayout, outLayout, inputShape, targetDevice) = basicParamsSet;
+        std::tie(convParams, netPrecision, inPrc, outPrc, inLayout, outLayout, inputShape, outputShape, targetDevice) = basicParamsSet;
 
         if (inPrc == Precision::UNSPECIFIED) {
             selectedType += std::string("_") + Precision(Precision::FP32).name();
@@ -76,16 +77,22 @@ class DeconvolutionLayerCPUTest : public testing::WithParamInterface<deconvLayer
 
         ngraph::op::PadType padType;
         InferenceEngine::SizeVector kernel, stride, dilation;
-        std::vector<ptrdiff_t> padBegin, padEnd;
+        std::vector<ptrdiff_t> padBegin, padEnd, outPadding;
         size_t convOutChannels;
-        std::tie(kernel, stride, padBegin, padEnd, dilation, convOutChannels, padType) = convParams;
+        std::tie(kernel, stride, padBegin, padEnd, dilation, convOutChannels, padType, outPadding) = convParams;
         auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
 
         auto inputParams = ngraph::builder::makeParams(ngraph::element::f32, { inputShape });
         auto paramOuts = ngraph::helpers::convert2OutputVector(ngraph::helpers::castOps2Nodes<ngraph::op::Parameter>(inputParams));
 
         auto deconvolutionNode = ngraph::builder::makeConvolutionBackpropData(paramOuts.front(), ngPrc, kernel, stride, padBegin,
-            padEnd, dilation, padType, convOutChannels);
+                padEnd, dilation, padType, convOutChannels, false, outPadding);
+
+        if (!outputShape.empty()) {
+            auto outShape = ngraph::opset3::Constant::create(ngraph::element::i64, {outputShape.size()}, outputShape);
+            deconvolutionNode = ngraph::builder::makeConvolutionBackpropData(paramOuts.front(), outShape, ngPrc, kernel, stride, padBegin,
+                padEnd, dilation, padType, convOutChannels);
+        }
 
         function = makeNgraphFunction(ngPrc, inputParams, deconvolutionNode, "convolutionBackpropData");
     }
@@ -108,6 +115,8 @@ const std::vector<fusingSpecificParams> fusingParamsSet{
 
 const std::map<std::string, std::string> cpuEmptyPluginConfig;
 const std::map<std::string, std::string> cpuBF16PluginConfig = { { PluginConfigParams::KEY_ENFORCE_BF16, PluginConfigParams::YES } };
+const std::vector<SizeVector> emptyOutputShape = { {} };
+const std::vector<std::vector<ptrdiff_t>> emptyOutputPadding = { {} };
 
 /* ============= Deconvolution params (planar layout) ============= */
 const SizeVector numOutChannels_Planar = { 6 };
@@ -139,7 +148,8 @@ const auto convParams_ExplicitPadding_Planar_2D = ::testing::Combine(
     ::testing::ValuesIn(padEnds2d),
     ::testing::ValuesIn(dilations2d),
     ::testing::ValuesIn(numOutChannels_Planar),
-    ::testing::Values(ngraph::op::PadType::EXPLICIT)
+    ::testing::Values(ngraph::op::PadType::EXPLICIT),
+    ::testing::ValuesIn(emptyOutputPadding)
 );
 
 INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Planar_FP32, DeconvolutionLayerCPUTest,
@@ -152,6 +162,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Planar_FP32, DeconvolutionLayerCPUTest,
             ::testing::Values(Layout::ANY),
             ::testing::Values(Layout::ANY),
             ::testing::Values(std::vector<size_t >({ 2, 12, 7, 7 })),
+            ::testing::ValuesIn(emptyOutputShape),
             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(filterCPUInfoForDevice({conv_gemm_2D})),
         ::testing::ValuesIn(fusingParamsSet),
@@ -168,6 +179,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Planar_BF16, DeconvolutionLayerCPUTest,
             ::testing::Values(Layout::ANY),
             ::testing::Values(Layout::ANY),
             ::testing::Values(std::vector<size_t >({ 2, 12, 7, 7 })),
+            ::testing::ValuesIn(emptyOutputShape),
             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(filterCPUInfoForDevice({conv_gemm_2D})),
         ::testing::ValuesIn(fusingParamsSet),
@@ -182,7 +194,8 @@ const auto convParams_ExplicitPadding_Planar_3D = ::testing::Combine(
     ::testing::ValuesIn(padEnds3d),
     ::testing::ValuesIn(dilations3d),
     ::testing::ValuesIn(numOutChannels_Planar),
-    ::testing::Values(ngraph::op::PadType::EXPLICIT)
+    ::testing::Values(ngraph::op::PadType::EXPLICIT),
+    ::testing::ValuesIn(emptyOutputPadding)
 );
 
 INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Planar_FP32, DeconvolutionLayerCPUTest,
@@ -195,6 +208,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Planar_FP32, DeconvolutionLayerCPUTest,
             ::testing::Values(Layout::ANY),
             ::testing::Values(Layout::ANY),
             ::testing::Values(std::vector<size_t >({ 2, 12, 7, 7, 7 })),
+            ::testing::ValuesIn(emptyOutputShape),
             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(filterCPUInfoForDevice({conv_gemm_3D})),
         ::testing::ValuesIn(fusingParamsSet),
@@ -211,6 +225,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Planar_BF16, DeconvolutionLayerCPUTest,
             ::testing::Values(Layout::ANY),
             ::testing::Values(Layout::ANY),
             ::testing::Values(std::vector<size_t >({ 2, 12, 7, 7, 7 })),
+            ::testing::ValuesIn(emptyOutputShape),
             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(filterCPUInfoForDevice({conv_gemm_3D})),
         ::testing::ValuesIn(fusingParamsSet),
@@ -225,7 +240,8 @@ const auto convParams_ExplicitPadding_Blocked_2D = ::testing::Combine(
     ::testing::ValuesIn(padEnds2d),
     ::testing::ValuesIn(dilations2d),
     ::testing::ValuesIn(numOutChannels_Blocked),
-    ::testing::Values(ngraph::op::PadType::EXPLICIT)
+    ::testing::Values(ngraph::op::PadType::EXPLICIT),
+    ::testing::ValuesIn(emptyOutputPadding)
 );
 
 INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Blocked_FP32, DeconvolutionLayerCPUTest,
@@ -238,6 +254,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Blocked_FP32, DeconvolutionLayerCPUTest,
             ::testing::Values(Layout::ANY),
             ::testing::Values(Layout::ANY),
             ::testing::Values(std::vector<size_t >({ 2, 67, 7, 7 })),
+            ::testing::ValuesIn(emptyOutputShape),
             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(filterCPUInfoForDevice({conv_avx512_2D})),
         ::testing::ValuesIn(fusingParamsSet),
@@ -254,6 +271,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Blocked_BF16, DeconvolutionLayerCPUTest,
             ::testing::Values(Layout::ANY),
             ::testing::Values(Layout::ANY),
             ::testing::Values(std::vector<size_t >({ 2, 67, 7, 7 })),
+            ::testing::ValuesIn(emptyOutputShape),
             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(filterCPUInfoForDevice({conv_avx512_2D})),
         ::testing::ValuesIn(fusingParamsSet),
@@ -268,7 +286,8 @@ const auto convParams_ExplicitPadding_Blocked_3D = ::testing::Combine(
     ::testing::ValuesIn(padEnds3d),
     ::testing::ValuesIn(dilations3d),
     ::testing::ValuesIn(numOutChannels_Blocked),
-    ::testing::Values(ngraph::op::PadType::EXPLICIT)
+    ::testing::Values(ngraph::op::PadType::EXPLICIT),
+    ::testing::ValuesIn(emptyOutputPadding)
 );
 
 INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Blocked_FP32, DeconvolutionLayerCPUTest,
@@ -281,6 +300,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Blocked_FP32, DeconvolutionLayerCPUTest,
             ::testing::Values(Layout::ANY),
             ::testing::Values(Layout::ANY),
             ::testing::Values(std::vector<size_t >({ 2, 67, 7, 7, 7 })),
+            ::testing::ValuesIn(emptyOutputShape),
             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(filterCPUInfoForDevice({conv_avx512_3D})),
         ::testing::ValuesIn(fusingParamsSet),
@@ -297,6 +317,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Blocked_BF16, DeconvolutionLayerCPUTest,
             ::testing::Values(Layout::ANY),
             ::testing::Values(Layout::ANY),
             ::testing::Values(std::vector<size_t >({ 2, 67, 7, 7, 7 })),
+            ::testing::ValuesIn(emptyOutputShape),
             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(filterCPUInfoForDevice({conv_avx512_3D})),
         ::testing::ValuesIn(fusingParamsSet),
@@ -312,7 +333,8 @@ const auto convParams_ExplicitPadding_1x1_2D = ::testing::Combine(
         ::testing::Values(std::vector<ptrdiff_t>({0, 0})),
         ::testing::Values(SizeVector({1, 1})),
         ::testing::ValuesIn(numOutChannels_Blocked),
-        ::testing::Values(ngraph::op::PadType::EXPLICIT)
+        ::testing::Values(ngraph::op::PadType::EXPLICIT),
+        ::testing::ValuesIn(emptyOutputPadding)
 );
 
 INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_1x1_FP32, DeconvolutionLayerCPUTest,
@@ -325,6 +347,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_1x1_FP32, DeconvolutionLayerCPUTest,
             ::testing::Values(Layout::ANY),
             ::testing::Values(Layout::ANY),
             ::testing::Values(std::vector<size_t >({ 2, 67, 7, 7 })),
+            ::testing::ValuesIn(emptyOutputShape),
             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(filterCPUInfoForDevice({conv_avx512_2D_1x1})),
         ::testing::ValuesIn(fusingParamsSet),
@@ -341,6 +364,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_1x1_BF16, DeconvolutionLayerCPUTest,
             ::testing::Values(Layout::ANY),
             ::testing::Values(Layout::ANY),
             ::testing::Values(std::vector<size_t >({ 2, 67, 7, 7 })),
+            ::testing::ValuesIn(emptyOutputShape),
             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(filterCPUInfoForDevice({conv_avx512_2D_1x1})),
         ::testing::ValuesIn(fusingParamsSet),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
index f18346cde9fb10..f404d932bc86a0 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
@@ -4,7 +4,7 @@
 
 #include <vector>
 
-#include "single_layer_tests/convolution_backprop_data.hpp"
+#include "single_layer_tests/convolution_backprop.hpp"
 #include "common_test_utils/test_constants.hpp"
 
 using namespace LayerTestsDefinitions;
@@ -17,6 +17,8 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 };
 
 const std::vector<size_t> numOutChannels = {1, 5, 16};
+const std::vector<std::vector<size_t >> emptyOutputShape = {{}};
+const std::vector<std::vector<ptrdiff_t>> emptyOutputPadding = {{}};
 
 /* ============= 2D ConvolutionBackpropData ============= */
 const std::vector<InferenceEngine::Precision> netPrecisions2D = {
@@ -40,7 +42,8 @@ const auto conv2DParams_ExplicitPadding = ::testing::Combine(
         ::testing::ValuesIn(padEnds2D),
         ::testing::ValuesIn(dilations2D),
         ::testing::ValuesIn(numOutChannels),
-        ::testing::Values(ngraph::op::PadType::EXPLICIT)
+        ::testing::Values(ngraph::op::PadType::EXPLICIT),
+        ::testing::ValuesIn(emptyOutputPadding)
 );
 const auto conv2DParams_AutoPadValid = ::testing::Combine(
         ::testing::ValuesIn(kernels2D),
@@ -49,10 +52,11 @@ const auto conv2DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(std::vector<ptrdiff_t>({0, 0})),
         ::testing::ValuesIn(dilations2D),
         ::testing::ValuesIn(numOutChannels),
-        ::testing::Values(ngraph::op::PadType::VALID)
+        ::testing::Values(ngraph::op::PadType::VALID),
+        ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, ConvolutionBackpropDataLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions2D),
@@ -61,10 +65,11 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, Convolu
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::ValuesIn(inputShapes2D),
+                                ::testing::ValuesIn(emptyOutputShape),
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
-                        ConvolutionBackpropDataLayerTest::getTestCaseName);
+                        ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, ConvolutionBackpropDataLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions2D),
@@ -73,8 +78,59 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, Convolutio
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::ValuesIn(inputShapes2D),
+                                ::testing::ValuesIn(emptyOutputShape),
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
-                        ConvolutionBackpropDataLayerTest::getTestCaseName);
+                        ConvolutionBackpropLayerTest::getTestCaseName);
+
+const std::vector<std::vector<ptrdiff_t>> outputPadding2D = {{1, 1}, {2, 2}};
+const std::vector<std::vector<size_t >> testStrides2D = {{3, 3}};
+
+const auto conv2DParams_ExplicitPadding_output_padding = ::testing::Combine(
+        ::testing::ValuesIn(kernels2D),
+        ::testing::ValuesIn(testStrides2D),
+        ::testing::ValuesIn(padBegins2D),
+        ::testing::ValuesIn(padEnds2D),
+        ::testing::ValuesIn(dilations2D),
+        ::testing::ValuesIn(numOutChannels),
+        ::testing::Values(ngraph::op::PadType::EXPLICIT),
+        ::testing::ValuesIn(outputPadding2D)
+);
+const auto conv2DParams_AutoPadValid_output_padding = ::testing::Combine(
+        ::testing::ValuesIn(kernels2D),
+        ::testing::ValuesIn(testStrides2D),
+        ::testing::Values(std::vector<ptrdiff_t>({0, 0})),
+        ::testing::Values(std::vector<ptrdiff_t>({0, 0})),
+        ::testing::ValuesIn(dilations2D),
+        ::testing::ValuesIn(numOutChannels),
+        ::testing::Values(ngraph::op::PadType::VALID),
+        ::testing::ValuesIn(outputPadding2D)
+);
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+                        ::testing::Combine(
+                                conv2DParams_AutoPadValid_output_padding,
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::ValuesIn(inputShapes2D),
+                                ::testing::ValuesIn(emptyOutputShape),
+                                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                        ConvolutionBackpropLayerTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+                        ::testing::Combine(
+                                conv2DParams_ExplicitPadding_output_padding,
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::ValuesIn(inputShapes2D),
+                                ::testing::ValuesIn(emptyOutputShape),
+                                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                        ConvolutionBackpropLayerTest::getTestCaseName);
 
 /* ============= 3D ConvolutionBackpropData ============= */
 const std::vector<InferenceEngine::Precision> netPrecisions3D = {
@@ -96,7 +152,8 @@ const auto conv3DParams_ExplicitPadding = ::testing::Combine(
         ::testing::ValuesIn(padEnds3D),
         ::testing::ValuesIn(dilations3D),
         ::testing::ValuesIn(numOutChannels),
-        ::testing::Values(ngraph::op::PadType::EXPLICIT)
+        ::testing::Values(ngraph::op::PadType::EXPLICIT),
+        ::testing::ValuesIn(emptyOutputPadding)
 );
 const auto conv3DParams_AutoPadValid = ::testing::Combine(
         ::testing::ValuesIn(kernels3D),
@@ -105,10 +162,11 @@ const auto conv3DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(std::vector<ptrdiff_t>({0, 0, 0})),
         ::testing::ValuesIn(dilations3D),
         ::testing::ValuesIn(numOutChannels),
-        ::testing::Values(ngraph::op::PadType::VALID)
+        ::testing::Values(ngraph::op::PadType::VALID),
+        ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, ConvolutionBackpropDataLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions3D),
@@ -117,10 +175,11 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, Convolu
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::ValuesIn(inputShapes3D),
+                                ::testing::ValuesIn(emptyOutputShape),
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
-                        ConvolutionBackpropDataLayerTest::getTestCaseName);
+                        ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, ConvolutionBackpropDataLayerTest,
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions3D),
@@ -129,7 +188,58 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, Convolutio
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::ValuesIn(inputShapes3D),
+                                ::testing::ValuesIn(emptyOutputShape),
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
-                        ConvolutionBackpropDataLayerTest::getTestCaseName);
+                        ConvolutionBackpropLayerTest::getTestCaseName);
+
+const std::vector<std::vector<ptrdiff_t>> outputPadding3D = {{1, 1, 1}, {2, 2, 2}};
+const std::vector<std::vector<size_t >> testStrides3D = {{3, 3, 3}};
+
+const auto conv3DParams_ExplicitPadding_output_padding = ::testing::Combine(
+        ::testing::ValuesIn(kernels3D),
+        ::testing::ValuesIn(testStrides3D),
+        ::testing::ValuesIn(padBegins3D),
+        ::testing::ValuesIn(padEnds3D),
+        ::testing::ValuesIn(dilations3D),
+        ::testing::ValuesIn(numOutChannels),
+        ::testing::Values(ngraph::op::PadType::EXPLICIT),
+        ::testing::ValuesIn(outputPadding3D)
+);
+const auto conv3DParams_AutoPadValid_output_padding = ::testing::Combine(
+        ::testing::ValuesIn(kernels3D),
+        ::testing::ValuesIn(testStrides3D),
+        ::testing::Values(std::vector<ptrdiff_t>({0, 0, 0})),
+        ::testing::Values(std::vector<ptrdiff_t>({0, 0, 0})),
+        ::testing::ValuesIn(dilations3D),
+        ::testing::ValuesIn(numOutChannels),
+        ::testing::Values(ngraph::op::PadType::VALID),
+        ::testing::ValuesIn(outputPadding3D)
+);
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+                        ::testing::Combine(
+                                conv3DParams_AutoPadValid_output_padding,
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::ValuesIn(inputShapes3D),
+                                ::testing::ValuesIn(emptyOutputShape),
+                                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                        ConvolutionBackpropLayerTest::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+                        ::testing::Combine(
+                                conv3DParams_ExplicitPadding_output_padding,
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::Values(InferenceEngine::Layout::ANY),
+                                ::testing::ValuesIn(inputShapes3D),
+                                ::testing::ValuesIn(emptyOutputShape),
+                                ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+                        ConvolutionBackpropLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/shared/include/single_layer_tests/convolution_backprop.hpp b/inference-engine/tests/functional/plugin/shared/include/single_layer_tests/convolution_backprop.hpp
new file mode 100644
index 00000000000000..45563d86a34f43
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/shared/include/single_layer_tests/convolution_backprop.hpp
@@ -0,0 +1,15 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "shared_test_classes/single_layer/convolution_backprop.hpp"
+
+namespace LayerTestsDefinitions {
+
+TEST_P(ConvolutionBackpropLayerTest, CompareWithRefs) {
+    Run();
+}
+
+}
diff --git a/inference-engine/tests/functional/plugin/shared/include/single_layer_tests/convolution_backprop_data.hpp b/inference-engine/tests/functional/plugin/shared/include/single_layer_tests/convolution_backprop_data.hpp
index 67fb0c56efdc1b..3b2947db121cc5 100644
--- a/inference-engine/tests/functional/plugin/shared/include/single_layer_tests/convolution_backprop_data.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/single_layer_tests/convolution_backprop_data.hpp
@@ -2,6 +2,7 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+// DEPRECATED, can't be removed currently due to arm and kmb-plugin dependency (#55568)
 #pragma once
 
 #include "shared_test_classes/single_layer/convolution_backprop_data.hpp"
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/convolution_backprop.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/convolution_backprop.hpp
new file mode 100644
index 00000000000000..794782396da858
--- /dev/null
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/convolution_backprop.hpp
@@ -0,0 +1,49 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <tuple>
+#include <vector>
+#include <string>
+#include <memory>
+
+#include "shared_test_classes/base/layer_test_utils.hpp"
+#include "ngraph_functions/builders.hpp"
+#include "ngraph_functions/utils/ngraph_helpers.hpp"
+
+namespace LayerTestsDefinitions {
+
+typedef std::tuple<
+        InferenceEngine::SizeVector,    // Kernel size
+        InferenceEngine::SizeVector,    // Strides
+        std::vector<ptrdiff_t>,         // Pad begin
+        std::vector<ptrdiff_t>,         // Pad end
+        InferenceEngine::SizeVector,    // Dilation
+        size_t,                         // Num out channels
+        ngraph::op::PadType,            // Padding type
+        std::vector<ptrdiff_t>          // Output padding
+> convBackpropSpecificParams;
+typedef std::tuple<
+        convBackpropSpecificParams,
+        InferenceEngine::Precision,     // Net precision
+        InferenceEngine::Precision,     // Input precision
+        InferenceEngine::Precision,     // Output precision
+        InferenceEngine::Layout,        // Input layout
+        InferenceEngine::Layout,        // Output layout
+        InferenceEngine::SizeVector,    // Input shapes
+        InferenceEngine::SizeVector,    // Output shapes
+        LayerTestsUtils::TargetDevice   // Device name
+> convBackpropLayerTestParamsSet;
+
+class ConvolutionBackpropLayerTest : public testing::WithParamInterface<convBackpropLayerTestParamsSet>,
+                                         virtual public LayerTestsUtils::LayerTestsCommon {
+public:
+    static std::string getTestCaseName(testing::TestParamInfo<convBackpropLayerTestParamsSet> obj);
+
+protected:
+    void SetUp() override;
+};
+
+}  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/convolution_backprop_data.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/convolution_backprop_data.hpp
index ecfd6e4f1f7769..9aeb9a1a2be72c 100644
--- a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/convolution_backprop_data.hpp
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/convolution_backprop_data.hpp
@@ -2,6 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+// DEPRECATED, can't be removed currently due to arm and kmb-plugin dependency (#55568)
+
 #pragma once
 
 #include <tuple>
diff --git a/inference-engine/tests/functional/shared_test_classes/src/single_layer/convolution_backprop.cpp b/inference-engine/tests/functional/shared_test_classes/src/single_layer/convolution_backprop.cpp
new file mode 100644
index 00000000000000..55aae5e0a215b3
--- /dev/null
+++ b/inference-engine/tests/functional/shared_test_classes/src/single_layer/convolution_backprop.cpp
@@ -0,0 +1,71 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "shared_test_classes/single_layer/convolution_backprop.hpp"
+
+namespace LayerTestsDefinitions {
+
+std::string ConvolutionBackpropLayerTest::getTestCaseName(testing::TestParamInfo<convBackpropLayerTestParamsSet> obj) {
+    convBackpropSpecificParams convBackpropDataParams;
+    InferenceEngine::Precision netPrecision;
+    InferenceEngine::Precision inPrc, outPrc;
+    InferenceEngine::Layout inLayout, outLayout;
+    InferenceEngine::SizeVector inputShapes;
+    InferenceEngine::SizeVector outputShapes;
+    std::string targetDevice;
+    std::tie(convBackpropDataParams, netPrecision, inPrc, outPrc, inLayout, outLayout, inputShapes, outputShapes, targetDevice) = obj.param;
+    ngraph::op::PadType padType;
+    InferenceEngine::SizeVector kernel, stride, dilation;
+    std::vector<ptrdiff_t> padBegin, padEnd, outPadding;
+    size_t convOutChannels;
+    std::tie(kernel, stride, padBegin, padEnd, dilation, convOutChannels, padType, outPadding) = convBackpropDataParams;
+
+    std::ostringstream result;
+    result << "IS=" << CommonTestUtils::vec2str(inputShapes) << "_";
+    result << "OS=" << CommonTestUtils::vec2str(outputShapes) << "_";
+    result << "K" << CommonTestUtils::vec2str(kernel) << "_";
+    result << "S" << CommonTestUtils::vec2str(stride) << "_";
+    result << "PB" << CommonTestUtils::vec2str(padBegin) << "_";
+    result << "PE" << CommonTestUtils::vec2str(padEnd) << "_";
+    result << "D=" << CommonTestUtils::vec2str(dilation) << "_";
+    result << "OP=" << CommonTestUtils::vec2str(outPadding) << "_";
+    result << "O=" << convOutChannels << "_";
+    result << "AP=" << padType << "_";
+    result << "netPRC=" << netPrecision.name() << "_";
+    result << "inPRC=" << inPrc.name() << "_";
+    result << "outPRC=" << outPrc.name() << "_";
+    result << "inL=" << inLayout << "_";
+    result << "outL=" << outLayout << "_";
+    result << "trgDev=" << targetDevice;
+    return result.str();
+}
+
+void ConvolutionBackpropLayerTest::SetUp() {
+    convBackpropSpecificParams convBackpropDataParams;
+    std::vector<size_t> inputShape;
+    std::vector<size_t> outputShape;
+    auto netPrecision = InferenceEngine::Precision::UNSPECIFIED;
+    std::tie(convBackpropDataParams, netPrecision, inPrc, outPrc, inLayout, outLayout, inputShape, outputShape, targetDevice) = this->GetParam();
+    ngraph::op::PadType padType;
+    InferenceEngine::SizeVector kernel, stride, dilation;
+    std::vector<ptrdiff_t> padBegin, padEnd, outPadding;
+    size_t convOutChannels;
+    std::tie(kernel, stride, padBegin, padEnd, dilation, convOutChannels, padType, outPadding) = convBackpropDataParams;
+    auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
+    auto params = ngraph::builder::makeParams(ngPrc, {inputShape});
+    auto paramOuts = ngraph::helpers::convert2OutputVector(
+            ngraph::helpers::castOps2Nodes<ngraph::op::Parameter>(params));
+    auto convBackpropData = std::dynamic_pointer_cast<ngraph::opset1::ConvolutionBackpropData>(
+            ngraph::builder::makeConvolutionBackpropData(paramOuts[0], ngPrc, kernel, stride, padBegin,
+                                                        padEnd, dilation, padType, convOutChannels, false, outPadding));
+    if (!outputShape.empty()) {
+        auto outShape = ngraph::opset3::Constant::create(ngraph::element::i64, {outputShape.size()}, outputShape);
+        convBackpropData = std::dynamic_pointer_cast<ngraph::opset1::ConvolutionBackpropData>(
+        ngraph::builder::makeConvolutionBackpropData(paramOuts[0], outShape, ngPrc, kernel, stride, padBegin,
+                                                        padEnd, dilation, padType, convOutChannels));
+    }
+    ngraph::ResultVector results{std::make_shared<ngraph::opset1::Result>(convBackpropData)};
+    function = std::make_shared<ngraph::Function>(results, params, "convolutionBackpropData");
+}
+}  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/shared_test_classes/src/single_layer/convolution_backprop_data.cpp b/inference-engine/tests/functional/shared_test_classes/src/single_layer/convolution_backprop_data.cpp
index c6730a3aaecf30..f2656a3c2ab581 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/single_layer/convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/single_layer/convolution_backprop_data.cpp
@@ -2,6 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+// DEPRECATED, can't be removed currently due to arm and kmb-plugin dependency (#55568)
+
 #include "shared_test_classes/single_layer/convolution_backprop_data.hpp"
 
 namespace LayerTestsDefinitions {
@@ -54,7 +56,7 @@ void ConvolutionBackpropDataLayerTest::SetUp() {
             ngraph::helpers::castOps2Nodes<ngraph::op::Parameter>(params));
     auto convBackpropData = std::dynamic_pointer_cast<ngraph::opset1::ConvolutionBackpropData>(
             ngraph::builder::makeConvolutionBackpropData(paramOuts[0], ngPrc, kernel, stride, padBegin,
-                                                         padEnd, dilation, padType, convOutChannels));
+                                                        padEnd, dilation, padType, convOutChannels));
     ngraph::ResultVector results{std::make_shared<ngraph::opset1::Result>(convBackpropData)};
     function = std::make_shared<ngraph::Function>(results, params, "convolutionBackpropData");
 }
diff --git a/inference-engine/tests/ngraph_helpers/ngraph_functions/include/ngraph_functions/builders.hpp b/inference-engine/tests/ngraph_helpers/ngraph_functions/include/ngraph_functions/builders.hpp
index 802535430cd134..1643f31f761f28 100644
--- a/inference-engine/tests/ngraph_helpers/ngraph_functions/include/ngraph_functions/builders.hpp
+++ b/inference-engine/tests/ngraph_helpers/ngraph_functions/include/ngraph_functions/builders.hpp
@@ -125,6 +125,7 @@ std::shared_ptr<ngraph::Node> makeConvolutionBackpropData(const ngraph::Output<N
                                                           const op::PadType &autoPad,
                                                           size_t numOutChannels,
                                                           bool addBiases = false,
+                                                          const std::vector<ptrdiff_t> &outputPadding = {},
                                                           const std::vector<float> &filterWeights = {},
                                                           const std::vector<float> &biasesWeights = {});
 
@@ -137,6 +138,22 @@ std::shared_ptr<ngraph::Node> makeConvolutionBackpropData(const ngraph::Output<N
                                                           const std::vector<size_t> &dilations,
                                                           const op::PadType &autoPad,
                                                           bool addBiases = false,
+                                                          const std::vector<ptrdiff_t> &outputPadding = {},
+                                                          const std::vector<float> &biasesWeights = {});
+
+std::shared_ptr<ngraph::Node> makeConvolutionBackpropData(const ngraph::Output<Node> &in,
+                                                          const ngraph::Output<Node> &outputShape,
+                                                          const element::Type &type,
+                                                          const std::vector<size_t> &filterSize,
+                                                          const std::vector<size_t> &strides,
+                                                          const std::vector<ptrdiff_t> &padsBegin,
+                                                          const std::vector<ptrdiff_t> &padsEnd,
+                                                          const std::vector<size_t> &dilations,
+                                                          const op::PadType &autoPad,
+                                                          size_t numOutChannels,
+                                                          bool addBiases = false,
+                                                          const std::vector<ptrdiff_t> &outputPadding = {},
+                                                          const std::vector<float> &filterWeights = {},
                                                           const std::vector<float> &biasesWeights = {});
 
 std::shared_ptr<ngraph::Node> makeCTCGreedyDecoder(
diff --git a/inference-engine/tests/ngraph_helpers/ngraph_functions/src/convolution_backprop_data.cpp b/inference-engine/tests/ngraph_helpers/ngraph_functions/src/convolution_backprop_data.cpp
index 0edf339ce7f3da..91d6c0fc0852e6 100644
--- a/inference-engine/tests/ngraph_helpers/ngraph_functions/src/convolution_backprop_data.cpp
+++ b/inference-engine/tests/ngraph_helpers/ngraph_functions/src/convolution_backprop_data.cpp
@@ -20,6 +20,7 @@ std::shared_ptr<Node> makeConvolutionBackpropData(const ngraph::Output<Node> &in
                                                   const op::PadType &autoPad,
                                                   size_t numOutChannels,
                                                   bool addBiases,
+                                                  const std::vector<ptrdiff_t> &outputPadding,
                                                   const std::vector<float> &filterWeights,
                                                   const std::vector<float> &biasesWeights) {
     bool randomFilterWeights = filterWeights.empty();
@@ -28,7 +29,7 @@ std::shared_ptr<Node> makeConvolutionBackpropData(const ngraph::Output<Node> &in
     filterWeightsShape.insert(filterWeightsShape.end(), filterSize.begin(), filterSize.end());
     auto filterWeightsNode = makeConstant(type, filterWeightsShape, filterWeights, randomFilterWeights);
 
-    return makeConvolutionBackpropData(in, filterWeightsNode, type, strides, padsBegin, padsEnd, dilations, autoPad, addBiases, biasesWeights);
+    return makeConvolutionBackpropData(in, filterWeightsNode, type, strides, padsBegin, padsEnd, dilations, autoPad, addBiases, outputPadding, biasesWeights);
 }
 
 std::shared_ptr<Node> makeConvolutionBackpropData(const ngraph::Output<Node> &in,
@@ -40,9 +41,51 @@ std::shared_ptr<Node> makeConvolutionBackpropData(const ngraph::Output<Node> &in
                                                   const std::vector<size_t> &dilations,
                                                   const op::PadType &autoPad,
                                                   bool addBiases,
+                                                  const std::vector<ptrdiff_t> &outputPadding,
                                                   const std::vector<float> &biasesWeights) {
     auto deconv = std::make_shared<opset1::ConvolutionBackpropData>(in, weights, strides, padsBegin, padsEnd, dilations, autoPad);
 
+    if (!outputPadding.empty()) {
+        deconv = std::make_shared<opset1::ConvolutionBackpropData>(in, weights, strides, padsBegin, padsEnd, dilations, autoPad, outputPadding);
+    }
+
+    if (addBiases) {
+        bool randomBiases = biasesWeights.empty();
+        auto biasesWeightsNode = makeConstant(type, {}, biasesWeights, randomBiases);
+        auto add = std::make_shared<ngraph::opset1::Add>(deconv, biasesWeightsNode);
+        return add;
+    } else {
+        return deconv;
+    }
+}
+
+std::shared_ptr<Node> makeConvolutionBackpropData(const ngraph::Output<Node> &in,
+                                                  const ngraph::Output<Node> &outputShape,
+                                                  const element::Type &type,
+                                                  const std::vector<size_t> &filterSize,
+                                                  const std::vector<size_t> &strides,
+                                                  const std::vector<ptrdiff_t> &padsBegin,
+                                                  const std::vector<ptrdiff_t> &padsEnd,
+                                                  const std::vector<size_t> &dilations,
+                                                  const op::PadType &autoPad,
+                                                  size_t numOutChannels,
+                                                  bool addBiases,
+                                                  const std::vector<ptrdiff_t> &outputPadding,
+                                                  const std::vector<float> &filterWeights,
+                                                  const std::vector<float> &biasesWeights) {
+    bool randomFilterWeights = filterWeights.empty();
+    auto shape = in.get_shape();
+    std::vector<size_t> filterWeightsShape = {shape[1], numOutChannels};
+    filterWeightsShape.insert(filterWeightsShape.end(), filterSize.begin(), filterSize.end());
+    auto filterWeightsNode = makeConstant(type, filterWeightsShape, filterWeights, randomFilterWeights);
+
+    auto deconv = std::make_shared<opset1::ConvolutionBackpropData>(in, filterWeightsNode, outputShape, strides, padsBegin, padsEnd, dilations, autoPad);
+
+    if (!outputPadding.empty()) {
+        deconv = std::make_shared<opset1::ConvolutionBackpropData>(in, filterWeightsNode, outputShape, strides, padsBegin,
+                                                                   padsEnd, dilations, autoPad, outputPadding);
+    }
+
     if (addBiases) {
         bool randomBiases = biasesWeights.empty();
         auto biasesWeightsNode = makeConstant(type, {}, biasesWeights, randomBiases);

From b4a4c9110c4c5e629277dae02d4ab8d41670edcc Mon Sep 17 00:00:00 2001
From: Gabriele Galiero Casay <gabriele.galiero.casay@intel.com>
Date: Tue, 15 Jun 2021 06:18:57 +0200
Subject: [PATCH 079/178] Revise Reference Implementations ReduceLp operations
 (#6065)

* Revise reference implementation for ReduceL1 operation

* Revise reference implementation for ReduceL2 operation

* Move op_eval tests to backend unit tests

* Added minor changes

 * Replace CoordinateTransform for CoordinateTransformBasic
 * Added constant expression to set keep_dims as false

* Add const qualifier to local variables

* Use rank from host tensor to normalize axes
---
 .../ngraph/runtime/reference/reduce_l1.hpp    | 29 ++++-----
 .../ngraph/runtime/reference/reduce_l2.hpp    | 37 ++++++-----
 ngraph/core/src/op/reduce_l1.cpp              | 12 +++-
 ngraph/core/src/op/reduce_l2.cpp              | 13 +++-
 ngraph/test/CMakeLists.txt                    |  4 +-
 ngraph/test/backend/reduce_l1.in.cpp          | 61 +++++++++++++++++++
 ngraph/test/backend/reduce_l2.in.cpp          | 61 +++++++++++++++++++
 ngraph/test/op_eval/reduce_l1.cpp             | 59 ------------------
 ngraph/test/op_eval/reduce_l2.cpp             | 61 -------------------
 9 files changed, 176 insertions(+), 161 deletions(-)
 create mode 100644 ngraph/test/backend/reduce_l1.in.cpp
 create mode 100644 ngraph/test/backend/reduce_l2.in.cpp
 delete mode 100644 ngraph/test/op_eval/reduce_l1.cpp
 delete mode 100644 ngraph/test/op_eval/reduce_l2.cpp

diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/reduce_l1.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/reduce_l1.hpp
index 86eac8ceb26183..83a3fbce98c721 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/reduce_l1.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/reduce_l1.hpp
@@ -5,6 +5,7 @@
 #pragma once
 
 #include <cmath>
+#include <numeric>
 
 #include "ngraph/coordinate_transform.hpp"
 #include "ngraph/shape_util.hpp"
@@ -19,27 +20,27 @@ namespace ngraph
             void reduce_l1(const T* arg,
                            T* out,
                            const Shape& in_shape,
-                           const AxisSet& reduction_axes,
-                           bool keep_dims)
+                           const AxisSet& reduction_axes)
             {
-                auto out_shape = reduce(in_shape, reduction_axes, keep_dims);
-                CoordinateTransform output_transform(out_shape);
+                constexpr bool dont_keep_dims_in_output = false;
+                const auto out_shape = reduce(in_shape, reduction_axes, dont_keep_dims_in_output);
+                std::fill(out, out + shape_size(out_shape), 0);
 
-                for (const Coordinate& output_coord : output_transform)
-                {
-                    out[output_transform.index(output_coord)] = 0;
-                }
-
-                CoordinateTransform input_transform(in_shape);
+                const auto in_strides = row_major_strides(in_shape);
+                const auto out_strides = row_major_strides(out_shape);
 
+                CoordinateTransformBasic input_transform(in_shape);
                 for (const Coordinate& input_coord : input_transform)
                 {
-                    Coordinate output_coord = reduce(input_coord, reduction_axes, keep_dims);
+                    const Coordinate output_coord =
+                        reduce(input_coord, reduction_axes, dont_keep_dims_in_output);
 
-                    size_t output_index = output_transform.index(output_coord);
+                    const size_t in_idx = std::inner_product(
+                        input_coord.begin(), input_coord.end(), in_strides.begin(), 0);
+                    const size_t out_idx = std::inner_product(
+                        output_coord.begin(), output_coord.end(), out_strides.begin(), 0);
 
-                    out[output_index] =
-                        out[output_index] + std::abs(arg[input_transform.index(input_coord)]);
+                    out[out_idx] = out[out_idx] + std::abs(arg[in_idx]);
                 }
             }
         } // namespace reference
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/reduce_l2.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/reduce_l2.hpp
index b7ae96586b2c08..aeb4eecbe0f97e 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/reduce_l2.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/reduce_l2.hpp
@@ -5,6 +5,7 @@
 #pragma once
 
 #include <cmath>
+#include <numeric>
 
 #include "ngraph/coordinate_transform.hpp"
 #include "ngraph/shape_util.hpp"
@@ -19,34 +20,30 @@ namespace ngraph
             void reduce_l2(const T* arg,
                            T* out,
                            const Shape& in_shape,
-                           const AxisSet& reduction_axes,
-                           bool keep_dims)
+                           const AxisSet& reduction_axes)
             {
-                auto out_shape = reduce(in_shape, reduction_axes, keep_dims);
-                CoordinateTransform output_transform(out_shape);
+                constexpr bool dont_keep_dims_in_output = false;
+                const auto out_shape = reduce(in_shape, reduction_axes, dont_keep_dims_in_output);
+                std::fill(out, out + shape_size(out_shape), 0);
 
-                for (const Coordinate& output_coord : output_transform)
-                {
-                    out[output_transform.index(output_coord)] = 0;
-                }
-
-                CoordinateTransform input_transform(in_shape);
+                const auto in_strides = row_major_strides(in_shape);
+                const auto out_strides = row_major_strides(out_shape);
 
+                CoordinateTransformBasic input_transform(in_shape);
                 for (const Coordinate& input_coord : input_transform)
                 {
-                    Coordinate output_coord = reduce(input_coord, reduction_axes, keep_dims);
+                    const Coordinate output_coord =
+                        reduce(input_coord, reduction_axes, dont_keep_dims_in_output);
 
-                    size_t output_index = output_transform.index(output_coord);
+                    const size_t in_idx = std::inner_product(
+                        input_coord.begin(), input_coord.end(), in_strides.begin(), 0);
+                    const size_t out_idx = std::inner_product(
+                        output_coord.begin(), output_coord.end(), out_strides.begin(), 0);
 
-                    out[output_index] =
-                        out[output_index] + arg[input_transform.index(input_coord)] *
-                                                arg[input_transform.index(input_coord)];
-                }
-                for (const Coordinate& output_coord : output_transform)
-                {
-                    out[output_transform.index(output_coord)] =
-                        sqrt(out[output_transform.index(output_coord)]);
+                    out[out_idx] = out[out_idx] + arg[in_idx] * arg[in_idx];
                 }
+                std::transform(
+                    out, out + shape_size(out_shape), out, [](T elem) { return sqrt(elem); });
             }
         } // namespace reference
     }     // namespace runtime
diff --git a/ngraph/core/src/op/reduce_l1.cpp b/ngraph/core/src/op/reduce_l1.cpp
index 29de7e4e03f09e..02a336ac60d661 100644
--- a/ngraph/core/src/op/reduce_l1.cpp
+++ b/ngraph/core/src/op/reduce_l1.cpp
@@ -3,8 +3,10 @@
 //
 
 #include "ngraph/op/reduce_l1.hpp"
+#include <ngraph/validation_util.hpp>
 #include "itt.hpp"
 #include "ngraph/graph_util.hpp"
+#include "ngraph/op/util/evaluate_helpers.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
 #include "ngraph/runtime/reference/reduce_l1.hpp"
 #include "ngraph/shape_util.hpp"
@@ -44,7 +46,7 @@ namespace reduce_l1
     {
         out->set_shape(reduce(arg->get_shape(), axes, keep_dims));
         runtime::reference::reduce_l1(
-            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes, keep_dims);
+            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes);
         return true;
     }
 
@@ -71,7 +73,13 @@ bool op::v4::ReduceL1::evaluate(const HostTensorVector& outputs,
                                 const HostTensorVector& inputs) const
 {
     NGRAPH_OP_SCOPE(v4_ReduceL1_evaluate);
-    return reduce_l1::evaluate_sum(inputs[0], outputs[0], get_reduction_axes(), get_keep_dims());
+    NGRAPH_CHECK(validate_host_tensor_vector(inputs, 2));
+    NGRAPH_CHECK(validate_host_tensor_vector(outputs, 1));
+
+    const auto reduction_axes = get_normalized_axes_from_tensor(
+        inputs[1], inputs[0]->get_partial_shape().rank(), get_friendly_name());
+
+    return reduce_l1::evaluate_sum(inputs[0], outputs[0], reduction_axes, get_keep_dims());
 }
 
 bool op::v4::ReduceL1::has_evaluate() const
diff --git a/ngraph/core/src/op/reduce_l2.cpp b/ngraph/core/src/op/reduce_l2.cpp
index e3ee81b1875474..7ffecb90377735 100644
--- a/ngraph/core/src/op/reduce_l2.cpp
+++ b/ngraph/core/src/op/reduce_l2.cpp
@@ -3,8 +3,10 @@
 //
 
 #include "ngraph/op/reduce_l2.hpp"
+#include <ngraph/validation_util.hpp>
 #include "itt.hpp"
 #include "ngraph/graph_util.hpp"
+#include "ngraph/op/util/evaluate_helpers.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
 #include "ngraph/runtime/reference/reduce_l2.hpp"
 #include "ngraph/shape_util.hpp"
@@ -44,7 +46,7 @@ namespace reduce_l2
     {
         out->set_shape(reduce(arg->get_shape(), axes, keep_dims));
         runtime::reference::reduce_l2(
-            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes, keep_dims);
+            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes);
         return true;
     }
 
@@ -69,8 +71,13 @@ bool op::v4::ReduceL2::evaluate(const HostTensorVector& outputs,
                                 const HostTensorVector& inputs) const
 {
     NGRAPH_OP_SCOPE(v4_ReduceL2_evaluate);
-    return reduce_l2::evaluate_reduce_l2(
-        inputs[0], outputs[0], get_reduction_axes(), get_keep_dims());
+    NGRAPH_CHECK(validate_host_tensor_vector(inputs, 2));
+    NGRAPH_CHECK(validate_host_tensor_vector(outputs, 1));
+
+    const auto reduction_axes = get_normalized_axes_from_tensor(
+        inputs[1], inputs[0]->get_partial_shape().rank(), get_friendly_name());
+
+    return reduce_l2::evaluate_reduce_l2(inputs[0], outputs[0], reduction_axes, get_keep_dims());
 }
 
 bool op::v4::ReduceL2::has_evaluate() const
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 2f85b09fc606d3..77903f5fc254c2 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -71,8 +71,6 @@ set(SRC
     op_eval/memory.cpp
     op_eval/mish.cpp
     op_eval/non_zero.cpp
-    op_eval/reduce_l1.cpp
-    op_eval/reduce_l2.cpp
     op_eval/reduce_prod.cpp
     op_eval/reduce_sum.cpp
     op_eval/roi_align.cpp
@@ -423,6 +421,8 @@ set(MULTI_TEST_SRC
     backend/psroi_pooling.in.cpp
     backend/range.in.cpp
     backend/recurrent_cells.in.cpp
+    backend/reduce_l1.in.cpp
+    backend/reduce_l2.in.cpp
     backend/reduce_max.in.cpp
     backend/reduce_mean.in.cpp
     backend/reduce_min.in.cpp
diff --git a/ngraph/test/backend/reduce_l1.in.cpp b/ngraph/test/backend/reduce_l1.in.cpp
new file mode 100644
index 00000000000000..d421ebc443dbc0
--- /dev/null
+++ b/ngraph/test/backend/reduce_l1.in.cpp
@@ -0,0 +1,61 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/all_close.hpp"
+#include "util/all_close_f.hpp"
+#include "util/test_control.hpp"
+#include "util/test_tools.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+
+NGRAPH_TEST(${BACKEND_NAME}, reduce_l1_one_axis_keep_dims)
+{
+    auto data = make_shared<op::Parameter>(element::f32, Shape{3, 2, 2});
+    auto axes = op::Constant::create(element::i32, Shape{1}, {2});
+    auto reduce_l1 = make_shared<op::v4::ReduceL1>(data, axes, true);
+    auto f = make_shared<Function>(OutputVector{reduce_l1}, ParameterVector{data});
+
+    auto backend = runtime::Backend::create("${BACKEND_NAME}");
+
+    // Create tensors for input/output
+    std::vector<float> input{1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0};
+    std::vector<float> expected_result{3.0, 7.0, 11.0, 15.0, 19.0, 23.0};
+
+    auto data_tensor = backend->create_tensor(element::f32, Shape{3, 2, 2});
+    copy_data(data_tensor, input);
+
+    auto result_tensor = backend->create_tensor(element::f32, Shape{3, 2, 1});
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result_tensor}, {data_tensor});
+    EXPECT_TRUE(test::all_close_f((expected_result), read_vector<float>(result_tensor)));
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, reduce_l1_one_axis_do_not_keep_dims)
+{
+    auto data = make_shared<op::Parameter>(element::f32, Shape{3, 2, 2});
+    auto axes = op::Constant::create(element::i32, Shape{1}, {2});
+    auto reduce_l1 = make_shared<op::v4::ReduceL1>(data, axes, false);
+    auto f = make_shared<Function>(OutputVector{reduce_l1}, ParameterVector{data});
+
+    auto backend = runtime::Backend::create("${BACKEND_NAME}");
+
+    // Create tensors for input/output
+    std::vector<float> input{1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0};
+    std::vector<float> expected_result{3.0, 7.0, 11.0, 15.0, 19.0, 23.0};
+
+    auto data_tensor = backend->create_tensor(element::f32, Shape{3, 2, 2});
+    copy_data(data_tensor, input);
+
+    auto result_tensor = backend->create_tensor(element::f32, Shape{3, 2});
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result_tensor}, {data_tensor});
+    EXPECT_TRUE(test::all_close_f((expected_result), read_vector<float>(result_tensor)));
+}
diff --git a/ngraph/test/backend/reduce_l2.in.cpp b/ngraph/test/backend/reduce_l2.in.cpp
new file mode 100644
index 00000000000000..a1cd3eebab6cdd
--- /dev/null
+++ b/ngraph/test/backend/reduce_l2.in.cpp
@@ -0,0 +1,61 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/all_close.hpp"
+#include "util/all_close_f.hpp"
+#include "util/test_control.hpp"
+#include "util/test_tools.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+
+NGRAPH_TEST(${BACKEND_NAME}, reduce_l2_one_axis_keep_dims)
+{
+    auto data = make_shared<op::Parameter>(element::f32, Shape{3, 2, 2});
+    auto axes = op::Constant::create(element::i32, Shape{1}, {2});
+    auto reduce_l2 = make_shared<op::v4::ReduceL2>(data, axes, true);
+    auto f = make_shared<Function>(OutputVector{reduce_l2}, ParameterVector{data});
+
+    auto backend = runtime::Backend::create("${BACKEND_NAME}");
+
+    std::vector<float> input{1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0};
+    std::vector<float> expected_result{
+        2.23606798, 5.0, 7.81024968, 10.63014581, 13.45362405, 16.2788206};
+
+    auto data_tensor = backend->create_tensor(element::f32, Shape{3, 2, 2});
+    copy_data(data_tensor, input);
+
+    auto result_tensor = backend->create_tensor(element::f32, Shape{3, 2, 1});
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result_tensor}, {data_tensor});
+    EXPECT_TRUE(test::all_close_f((expected_result), read_vector<float>(result_tensor)));
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, reduce_l2_one_axis_do_not_keep_dims)
+{
+    auto data = make_shared<op::Parameter>(element::f32, Shape{3, 2, 2});
+    auto axes = op::Constant::create(element::i32, Shape{1}, {2});
+    auto reduce_l2 = make_shared<op::v4::ReduceL2>(data, axes, false);
+    auto f = make_shared<Function>(OutputVector{reduce_l2}, ParameterVector{data});
+
+    auto backend = runtime::Backend::create("${BACKEND_NAME}");
+
+    std::vector<float> input{1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0};
+    std::vector<float> expected_result{
+        2.23606798, 5.0, 7.81024968, 10.63014581, 13.45362405, 16.2788206};
+
+    auto data_tensor = backend->create_tensor(element::f32, Shape{3, 2, 2});
+    copy_data(data_tensor, input);
+
+    auto result_tensor = backend->create_tensor(element::f32, Shape{3, 2});
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result_tensor}, {data_tensor});
+    EXPECT_TRUE(test::all_close_f((expected_result), read_vector<float>(result_tensor)));
+}
diff --git a/ngraph/test/op_eval/reduce_l1.cpp b/ngraph/test/op_eval/reduce_l1.cpp
deleted file mode 100644
index 0f64ee495bffab..00000000000000
--- a/ngraph/test/op_eval/reduce_l1.cpp
+++ /dev/null
@@ -1,59 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include <string>
-#include <vector>
-
-#include "gtest/gtest.h"
-
-#include "ngraph/opsets/opset4.hpp"
-#include "ngraph/runtime/host_tensor.hpp"
-#include "ngraph/validation_util.hpp"
-#include "runtime/backend.hpp"
-#include "util/test_tools.hpp"
-
-using namespace std;
-using namespace ngraph;
-
-TEST(op_eval, reduce_l1_one_axis_keep_dims)
-{
-    auto data = make_shared<opset4::Parameter>(element::f32, Shape{3, 2, 2});
-    auto axes = opset4::Constant::create(element::i32, Shape{1}, {2});
-    auto reduce = make_shared<opset4::ReduceL1>(data, axes, true);
-    auto fun = make_shared<Function>(OutputVector{reduce}, ParameterVector{data});
-
-    std::vector<float> inputs{1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0};
-    std::vector<float> expected_result{3.0, 7.0, 11.0, 15.0, 19.0, 23.0};
-
-    auto result = make_shared<HostTensor>();
-    ASSERT_TRUE(fun->evaluate({result},
-                              {make_host_tensor<element::Type_t::f32>(Shape{3, 2, 2}, inputs),
-                               make_host_tensor<element::Type_t::i32>(Shape{1}, {2})}));
-    EXPECT_EQ(result->get_element_type(), element::f32);
-    EXPECT_EQ(result->get_shape(), Shape{std::vector<size_t>({3, 2, 1})});
-    auto result_data = read_vector<float>(result);
-    for (size_t i = 0; i < expected_result.size(); i++)
-        EXPECT_NEAR(result_data[i], expected_result[i], 0.000001);
-}
-
-TEST(op_eval, reduce_l1_one_axis_do_not_keep_dims)
-{
-    auto data = make_shared<opset4::Parameter>(element::f32, Shape{3, 2, 2});
-    auto axes = opset4::Constant::create(element::i32, Shape{1}, {2});
-    auto reduce = make_shared<opset4::ReduceL1>(data, axes, false);
-    auto fun = make_shared<Function>(OutputVector{reduce}, ParameterVector{data});
-
-    std::vector<float> inputs{1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0};
-    std::vector<float> expected_result{3.0, 7.0, 11.0, 15.0, 19.0, 23.0};
-
-    auto result = make_shared<HostTensor>();
-    ASSERT_TRUE(fun->evaluate({result},
-                              {make_host_tensor<element::Type_t::f32>(Shape{3, 2, 2}, inputs),
-                               make_host_tensor<element::Type_t::i32>(Shape{1}, {2})}));
-    EXPECT_EQ(result->get_element_type(), element::f32);
-    EXPECT_EQ(result->get_shape(), Shape{std::vector<size_t>({3, 2})});
-    auto result_data = read_vector<float>(result);
-    for (size_t i = 0; i < expected_result.size(); i++)
-        EXPECT_NEAR(result_data[i], expected_result[i], 0.000001);
-}
diff --git a/ngraph/test/op_eval/reduce_l2.cpp b/ngraph/test/op_eval/reduce_l2.cpp
deleted file mode 100644
index 7aab1907c4aac0..00000000000000
--- a/ngraph/test/op_eval/reduce_l2.cpp
+++ /dev/null
@@ -1,61 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include <string>
-#include <vector>
-
-#include "gtest/gtest.h"
-
-#include "ngraph/opsets/opset4.hpp"
-#include "ngraph/runtime/host_tensor.hpp"
-#include "ngraph/validation_util.hpp"
-#include "runtime/backend.hpp"
-#include "util/test_tools.hpp"
-
-using namespace std;
-using namespace ngraph;
-
-TEST(op_eval, reduce_l2_one_axis_keep_dims)
-{
-    auto data = make_shared<opset4::Parameter>(element::f32, Shape{3, 2, 2});
-    auto axes = opset4::Constant::create(element::i32, Shape{1}, {2});
-    auto reduce = make_shared<op::v4::ReduceL2>(data, axes, true);
-    auto fun = make_shared<Function>(OutputVector{reduce}, ParameterVector{data});
-
-    std::vector<float> inputs{1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0};
-    std::vector<float> expected_result{
-        2.23606798, 5.0, 7.81024968, 10.63014581, 13.45362405, 16.2788206};
-
-    auto result = make_shared<HostTensor>();
-    ASSERT_TRUE(fun->evaluate({result},
-                              {make_host_tensor<element::Type_t::f32>(Shape{3, 2, 2}, inputs),
-                               make_host_tensor<element::Type_t::i32>(Shape{1}, {2})}));
-    EXPECT_EQ(result->get_element_type(), element::f32);
-    EXPECT_EQ(result->get_shape(), Shape{std::vector<size_t>({3, 2, 1})});
-    auto result_data = read_vector<float>(result);
-    for (size_t i = 0; i < expected_result.size(); i++)
-        EXPECT_NEAR(result_data[i], expected_result[i], 0.000001);
-}
-
-TEST(op_eval, reduce_l2_one_axis_do_not_keep_dims)
-{
-    auto data = make_shared<opset4::Parameter>(element::f32, Shape{3, 2, 2});
-    auto axes = opset4::Constant::create(element::i32, Shape{1}, {2});
-    auto reduce = make_shared<op::v4::ReduceL2>(data, axes, false);
-    auto fun = make_shared<Function>(OutputVector{reduce}, ParameterVector{data});
-
-    std::vector<float> inputs{1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0};
-    std::vector<float> expected_result{
-        2.23606798, 5.0, 7.81024968, 10.63014581, 13.45362405, 16.2788206};
-
-    auto result = make_shared<HostTensor>();
-    ASSERT_TRUE(fun->evaluate({result},
-                              {make_host_tensor<element::Type_t::f32>(Shape{3, 2, 2}, inputs),
-                               make_host_tensor<element::Type_t::i32>(Shape{1}, {2})}));
-    EXPECT_EQ(result->get_element_type(), element::f32);
-    EXPECT_EQ(result->get_shape(), Shape{std::vector<size_t>({3, 2})});
-    auto result_data = read_vector<float>(result);
-    for (size_t i = 0; i < expected_result.size(); i++)
-        EXPECT_NEAR(result_data[i], expected_result[i], 0.000001);
-}

From 766d011b06af935f5e2efcb5411bf848856e9029 Mon Sep 17 00:00:00 2001
From: Patryk Elszkowski <patryk.elszkowski@intel.com>
Date: Tue, 15 Jun 2021 06:25:19 +0200
Subject: [PATCH 080/178] reshape reference implementation - new implementation
 (#5559)

---
 .../ngraph/runtime/opt_kernel/reshape.hpp     |   1 -
 .../ngraph/runtime/reference/reshape.hpp      |   6 +-
 .../src/runtime/opt_kernel/reshape.cpp        |  18 +-
 .../src/runtime/reference/reshape.cpp         |  91 ++++---
 ngraph/test/CMakeLists.txt                    |   1 +
 ngraph/test/reshape_opt_kernel.cpp            | 232 ++++++++++++++++++
 ngraph/test/visitors/op/reshape.cpp           |   4 +
 7 files changed, 310 insertions(+), 43 deletions(-)
 create mode 100644 ngraph/test/reshape_opt_kernel.cpp

diff --git a/ngraph/core/reference/include/ngraph/runtime/opt_kernel/reshape.hpp b/ngraph/core/reference/include/ngraph/runtime/opt_kernel/reshape.hpp
index 30ba487ecc602d..27d960ee6db5fa 100644
--- a/ngraph/core/reference/include/ngraph/runtime/opt_kernel/reshape.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/opt_kernel/reshape.hpp
@@ -5,7 +5,6 @@
 #pragma once
 
 #include "ngraph/axis_vector.hpp"
-#include "ngraph/runtime/reference/reshape.hpp"
 #include "ngraph/shape.hpp"
 
 namespace ngraph
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/reshape.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/reshape.hpp
index d6696b8feeb024..ee1c2f998ed16f 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/reshape.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/reshape.hpp
@@ -4,12 +4,8 @@
 
 #pragma once
 
-#include <cmath>
-
 #include "ngraph/axis_vector.hpp"
-#include "ngraph/check.hpp"
-#include "ngraph/coordinate_transform.hpp"
-#include "ngraph/type/element_type.hpp"
+#include "ngraph/shape.hpp"
 
 namespace ngraph
 {
diff --git a/ngraph/core/reference/src/runtime/opt_kernel/reshape.cpp b/ngraph/core/reference/src/runtime/opt_kernel/reshape.cpp
index f0d559f1847114..38c29f9387c60a 100644
--- a/ngraph/core/reference/src/runtime/opt_kernel/reshape.cpp
+++ b/ngraph/core/reference/src/runtime/opt_kernel/reshape.cpp
@@ -2,11 +2,12 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <cmath>
-#include <stdio.h>
+#include <algorithm>
+#include <cstring>
 
 #include "ngraph/check.hpp"
 #include "ngraph/runtime/opt_kernel/reshape.hpp"
+#include "ngraph/runtime/reference/reshape.hpp"
 
 using namespace ngraph;
 
@@ -233,6 +234,13 @@ namespace
             }
         }
     }
+    bool no_axis_reordering(const AxisVector& axis_order)
+    {
+        auto tmp = axis_order;
+        std::sort(begin(tmp), end(tmp));
+        tmp.erase(std::unique(begin(tmp), end(tmp)), end(tmp));
+        return tmp == axis_order;
+    }
 } // namespace
 void runtime::opt_kernel::reshape(const char* in,
                                   char* out,
@@ -241,6 +249,12 @@ void runtime::opt_kernel::reshape(const char* in,
                                   const Shape& out_shape,
                                   size_t elem_size)
 {
+    if (no_axis_reordering(in_axis_order))
+    {
+        std::memcpy(out, in, shape_size(in_shape) * elem_size);
+        return;
+    }
+
     switch (in_shape.size())
     {
     case 0: reshape_in0(in, out, in_shape, in_axis_order, out_shape, elem_size); break;
diff --git a/ngraph/core/reference/src/runtime/reference/reshape.cpp b/ngraph/core/reference/src/runtime/reference/reshape.cpp
index f4c100b27fcabe..27ac18e0190e02 100644
--- a/ngraph/core/reference/src/runtime/reference/reshape.cpp
+++ b/ngraph/core/reference/src/runtime/reference/reshape.cpp
@@ -2,46 +2,67 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <cmath>
-#include <stdio.h>
+#include <cstring>
+#include <numeric>
 
 #include "ngraph/check.hpp"
+#include "ngraph/coordinate_range.hpp"
+#include "ngraph/coordinate_transform.hpp"
 #include "ngraph/runtime/reference/reshape.hpp"
 
-using namespace ngraph;
-
-void runtime::reference::reshape(const char* arg,
-                                 char* out,
-                                 const Shape& in_shape,
-                                 const AxisVector& in_axis_order,
-                                 const Shape& out_shape,
-                                 size_t elem_size)
+namespace ngraph
 {
-    // Unfortunately we don't yet have a constructor for CoordinateTransform that lets
-    // us pass only source_space_shape
-    // and source_axis_order so we have to construct the defaults here.
-    Shape in_start_corner(in_shape.size(), 0); // (0,...0)
-    Strides in_strides(in_shape.size(), 1);    // (1,...,1)
-
-    CoordinateTransform input_transform(
-        in_shape, in_start_corner, in_shape, in_strides, in_axis_order);
-    CoordinateTransform output_transform(out_shape);
-
-    NGRAPH_CHECK(shape_size(input_transform.get_target_shape()) ==
-                 shape_size(output_transform.get_target_shape()));
-
-    CoordinateTransform::Iterator output_it = output_transform.begin();
-
-    for (const Coordinate& input_coord : input_transform)
+    namespace runtime
     {
-        if (output_it == output_transform.end())
-            break;
-        const Coordinate& output_coord = *output_it;
+        namespace reference
+        {
+            namespace
+            {
+                std::vector<size_t> reorder(const std::vector<size_t>& origin,
+                                            const AxisVector& order)
+                {
+                    std::vector<size_t> reordered = origin;
+                    auto out = begin(reordered);
+                    NGRAPH_CHECK(origin.size() <= order.size());
+                    for (size_t i = 0; i < origin.size(); ++i)
+                    {
+                        *out = origin.at(order[i]);
+                        ++out;
+                    }
+                    return reordered;
+                }
+            } // namespace
 
-        memcpy(out + output_transform.index(output_coord) * elem_size,
-               arg + input_transform.index(input_coord) * elem_size,
-               elem_size);
+            void reshape(const char* arg,
+                         char* out,
+                         const Shape& in_shape,
+                         const AxisVector& in_axis_order,
+                         const Shape& out_shape,
+                         size_t elem_size)
+            {
+                if (shape_size(in_shape) == 1)
+                {
+                    std::memcpy(out, arg, elem_size);
+                    return;
+                }
 
-        ++output_it;
-    }
-}
+                char* output = out;
+                const char* const output_end = out + shape_size(out_shape) * elem_size;
+                const auto axis_strides = reorder(row_major_strides(in_shape), in_axis_order);
+                for (const auto& coordinate :
+                     CoordinateTransformBasic(reorder(in_shape, in_axis_order)))
+                {
+                    if (output >= output_end)
+                    {
+                        break;
+                    }
+                    const auto elem_offset = std::inner_product(
+                        begin(coordinate), end(coordinate), begin(axis_strides), 0ll);
+                    const auto input = arg + elem_offset * elem_size;
+                    std::memcpy(output, input, elem_size);
+                    output += elem_size;
+                }
+            }
+        } // namespace reference
+    }     // namespace runtime
+} // namespace ngraph
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 77903f5fc254c2..47dfbbd3e495b3 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -92,6 +92,7 @@ set(SRC
     pattern.cpp
     provenance.cpp
     replace_node.cpp
+    reshape_opt_kernel.cpp
     shape.cpp
     span.cpp
     specialize_function.cpp
diff --git a/ngraph/test/reshape_opt_kernel.cpp b/ngraph/test/reshape_opt_kernel.cpp
new file mode 100644
index 00000000000000..01a2358e29eac9
--- /dev/null
+++ b/ngraph/test/reshape_opt_kernel.cpp
@@ -0,0 +1,232 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <numeric>
+#include <vector>
+
+#include "gtest/gtest.h"
+
+#include "ngraph/axis_vector.hpp"
+#include "ngraph/runtime/opt_kernel/reshape.hpp"
+#include "ngraph/shape.hpp"
+
+#include "util/ndarray.hpp"
+
+using namespace ngraph;
+
+namespace
+{
+    using ElementValue = int32_t;
+    enum class AxisOrder
+    {
+        straight,
+        reverse,
+    };
+
+    AxisVector get_axis_order(AxisOrder order, size_t size)
+    {
+        AxisVector v(size);
+        std::iota(begin(v), end(v), 0);
+        if (order == AxisOrder::reverse)
+        {
+            std::reverse(begin(v), end(v));
+        }
+        return v;
+    }
+
+    struct TestParams
+    {
+        AxisOrder order;
+        test::NDArrayBase<ElementValue> input;
+        test::NDArrayBase<ElementValue> output;
+    };
+
+    struct ReshapeOptKernel : ::testing::TestWithParam<TestParams>
+    {
+    };
+
+} // namespace
+
+TEST_P(ReshapeOptKernel, reshape_opt_kernel)
+{
+    const TestParams& p = GetParam();
+
+    const AxisVector axis_order = get_axis_order(p.order, p.input.get_shape().size());
+    std::vector<ElementValue> output_buff(p.input.get_vector().size());
+
+    runtime::opt_kernel::reshape((const char*)p.input.data(),
+                                 (char*)output_buff.data(),
+                                 p.input.get_shape(),
+                                 axis_order,
+                                 p.output.get_shape(),
+                                 sizeof(ElementValue));
+    EXPECT_EQ(p.output.get_vector(), output_buff);
+}
+
+INSTANTIATE_TEST_CASE_P(reshape_opt_kernel,
+                        ReshapeOptKernel,
+                        ::testing::Values(TestParams{AxisOrder::straight,
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {1, 2},
+                                                         {3, 4},
+                                                         {5, 6},
+                                                     },
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {1, 2, 3},
+                                                         {4, 5, 6},
+                                                     }},
+                                          TestParams{AxisOrder::straight,
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {1, 2},
+                                                         {3, 4},
+                                                         {5, 6},
+                                                     },
+                                                     test::NDArray<ElementValue, 1>{
+                                                         {1, 2, 3, 4, 5, 6},
+                                                     }},
+                                          TestParams{AxisOrder::straight,
+                                                     test::NDArray<ElementValue, 3>{
+                                                         {
+                                                             {11, 12},
+                                                             {13, 14},
+                                                             {15, 16},
+                                                         },
+                                                         {
+                                                             {21, 22},
+                                                             {23, 24},
+                                                             {25, 26},
+                                                         },
+                                                     },
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {11, 12, 13, 14, 15, 16},
+                                                         {21, 22, 23, 24, 25, 26},
+                                                     }},
+                                          TestParams{AxisOrder::straight,
+                                                     test::NDArray<ElementValue, 4>{
+                                                         {
+                                                             {
+                                                                 {11, 12},
+                                                                 {13, 14},
+                                                                 {15, 16},
+                                                             },
+                                                             {
+                                                                 {21, 22},
+                                                                 {23, 24},
+                                                                 {25, 26},
+                                                             },
+                                                         },
+                                                     },
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {11, 12, 13, 14, 15, 16},
+                                                         {21, 22, 23, 24, 25, 26},
+                                                     }},
+                                          TestParams{AxisOrder::reverse,
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {1, 2},
+                                                         {3, 4},
+                                                         {5, 6},
+                                                     },
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {1, 3, 5},
+                                                         {2, 4, 6},
+                                                     }},
+                                          TestParams{AxisOrder::reverse,
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {1, 2},
+                                                         {3, 4},
+                                                         {5, 6},
+                                                     },
+                                                     test::NDArray<ElementValue, 1>{
+                                                         {1, 3, 5, 2, 4, 6},
+                                                     }},
+                                          TestParams{AxisOrder::reverse,
+                                                     test::NDArray<ElementValue, 3>{
+                                                         {
+                                                             {11, 12},
+                                                             {13, 14},
+                                                             {15, 16},
+                                                         },
+                                                         {
+                                                             {21, 22},
+                                                             {23, 24},
+                                                             {25, 26},
+                                                         },
+                                                     },
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {11, 21, 13, 23, 15, 25},
+                                                         {12, 22, 14, 24, 16, 26},
+                                                     }},
+                                          TestParams{AxisOrder::reverse,
+                                                     test::NDArray<ElementValue, 4>{
+                                                         {
+                                                             {
+                                                                 {11, 12},
+                                                                 {13, 14},
+                                                                 {15, 16},
+                                                             },
+                                                             {
+                                                                 {21, 22},
+                                                                 {23, 24},
+                                                                 {25, 26},
+                                                             },
+                                                         },
+                                                     },
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {11, 21, 13, 23, 15, 25},
+                                                         {12, 22, 14, 24, 16, 26},
+                                                     }}));
+
+// input shape with size > 6 should be covered by reference implementation:
+INSTANTIATE_TEST_CASE_P(reshape_opt_kernel_ref_impl_fallback,
+                        ReshapeOptKernel,
+                        ::testing::Values(TestParams{AxisOrder::straight,
+                                                     test::NDArray<ElementValue, 7>{
+                                                         {
+                                                             {
+                                                                 {
+                                                                     {
+                                                                         {
+                                                                             {11, 12},
+                                                                             {13, 14},
+                                                                             {15, 16},
+                                                                         },
+                                                                         {
+                                                                             {21, 22},
+                                                                             {23, 24},
+                                                                             {25, 26},
+                                                                         },
+                                                                     },
+                                                                 },
+                                                             },
+                                                         },
+                                                     },
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {11, 12, 13, 14, 15, 16},
+                                                         {21, 22, 23, 24, 25, 26},
+                                                     }},
+                                          TestParams{AxisOrder::reverse,
+                                                     test::NDArray<ElementValue, 7>{
+                                                         {
+                                                             {
+                                                                 {
+                                                                     {
+                                                                         {
+                                                                             {11, 12},
+                                                                             {13, 14},
+                                                                             {15, 16},
+                                                                         },
+                                                                         {
+                                                                             {21, 22},
+                                                                             {23, 24},
+                                                                             {25, 26},
+                                                                         },
+                                                                     },
+                                                                 },
+                                                             },
+                                                         },
+                                                     },
+                                                     test::NDArray<ElementValue, 2>{
+                                                         {11, 21, 13, 23, 15, 25},
+                                                         {12, 22, 14, 24, 16, 26},
+                                                     }}));
diff --git a/ngraph/test/visitors/op/reshape.cpp b/ngraph/test/visitors/op/reshape.cpp
index 8acad56678edb3..7d48e11442f8ad 100644
--- a/ngraph/test/visitors/op/reshape.cpp
+++ b/ngraph/test/visitors/op/reshape.cpp
@@ -30,5 +30,9 @@ TEST(attributes, reshape_op)
     NodeBuilder builder(reshape);
     auto g_reshape = as_type_ptr<opset1::Reshape>(builder.create());
 
+    const auto expected_attr_count = 1;
+
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
+
     EXPECT_EQ(g_reshape->get_special_zero(), reshape->get_special_zero());
 }

From 134c66a9339477f3ba1e271cac6a7d7de25c5689 Mon Sep 17 00:00:00 2001
From: Gabriele Galiero Casay <gabriele.galiero.casay@intel.com>
Date: Tue, 15 Jun 2021 11:06:30 +0200
Subject: [PATCH 081/178] Reference Implementation of Logical Reduce Operations
 (#6004)

* Remove CoordinateTransform call to index function to calculate tensor element indexes

* Allow negative axis values in axes host tensor

* Added constant expression to set keep_dims as false

* Use rank from host tensor to normalize axes

* Address minor comments

* Add const qualifier to local variables
* Add deprecated macro for arm plugin dependent function signatures

* Remove duplicate helper functions
---
 .../ngraph/runtime/reference/eval_helpers.hpp | 15 ----
 .../runtime/reference/logical_reduction.hpp   | 83 ++++++++++++-------
 .../src/runtime/reference/eval_helpers.cpp    | 33 --------
 ngraph/core/src/op/reduce_logical_and.cpp     | 32 +++----
 ngraph/core/src/op/reduce_logical_or.cpp      | 32 +++----
 ngraph/test/eval.cpp                          | 22 -----
 6 files changed, 87 insertions(+), 130 deletions(-)
 delete mode 100644 ngraph/core/reference/include/ngraph/runtime/reference/eval_helpers.hpp
 delete mode 100644 ngraph/core/reference/src/runtime/reference/eval_helpers.cpp

diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/eval_helpers.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/eval_helpers.hpp
deleted file mode 100644
index 935237778d15f6..00000000000000
--- a/ngraph/core/reference/include/ngraph/runtime/reference/eval_helpers.hpp
+++ /dev/null
@@ -1,15 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-
-#include "ngraph/runtime/host_tensor.hpp"
-
-namespace ngraph
-{
-    namespace eval
-    {
-        AxisSet extract_reduction_axes(const HostTensorPtr& axes, const char* op_name);
-    }
-} // namespace ngraph
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/logical_reduction.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/logical_reduction.hpp
index f9e24d89292992..f2789da27ded86 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/logical_reduction.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/logical_reduction.hpp
@@ -5,6 +5,7 @@
 #pragma once
 
 #include <cmath>
+#include <numeric>
 
 #include "ngraph/coordinate_transform.hpp"
 #include "ngraph/shape_util.hpp"
@@ -17,53 +18,75 @@ namespace ngraph
         {
             static inline void reduce_logical_and(const char* arg,
                                                   char* out,
-                                                  const Shape& input_shape,
-                                                  const AxisSet& reduction_axes,
-                                                  bool keep_dims)
+                                                  const Shape& in_shape,
+                                                  const AxisSet& reduction_axes)
             {
-                CoordinateTransform output_transform(
-                    reduce(input_shape, reduction_axes, keep_dims));
+                constexpr bool dont_keep_dims_in_output = false;
+                const auto out_shape = reduce(in_shape, reduction_axes, dont_keep_dims_in_output);
+                std::fill(out, out + shape_size(out_shape), 1);
 
-                for (const Coordinate& output_coord : output_transform)
-                {
-                    out[output_transform.index(output_coord)] = 1;
-                }
-
-                CoordinateTransform input_transform(input_shape);
+                const auto in_strides = row_major_strides(in_shape);
+                const auto out_strides = row_major_strides(out_shape);
 
+                CoordinateTransformBasic input_transform(in_shape);
                 for (const Coordinate& input_coord : input_transform)
                 {
-                    Coordinate output_coord = reduce(input_coord, reduction_axes, keep_dims);
-                    out[output_transform.index(output_coord)] =
-                        out[output_transform.index(output_coord)] &&
-                        arg[input_transform.index(input_coord)];
+                    const Coordinate output_coord =
+                        reduce(input_coord, reduction_axes, dont_keep_dims_in_output);
+
+                    const size_t in_idx = std::inner_product(
+                        input_coord.begin(), input_coord.end(), in_strides.begin(), 0);
+                    const size_t out_idx = std::inner_product(
+                        output_coord.begin(), output_coord.end(), out_strides.begin(), 0);
+
+                    out[out_idx] = out[out_idx] && arg[in_idx];
                 }
             }
 
+            NGRAPH_DEPRECATED("Remove when arm plugin supports the new signature")
+            static inline void reduce_logical_and(const char* arg,
+                                                  char* out,
+                                                  const Shape& input_shape,
+                                                  const AxisSet& reduction_axes,
+                                                  bool)
+            {
+                reduce_logical_and(arg, out, input_shape, reduction_axes);
+            }
+
             static inline void reduce_logical_or(const char* arg,
                                                  char* out,
-                                                 const Shape& input_shape,
-                                                 const AxisSet& reduction_axes,
-                                                 bool keep_dims)
+                                                 const Shape& in_shape,
+                                                 const AxisSet& reduction_axes)
             {
-                CoordinateTransform output_transform(
-                    reduce(input_shape, reduction_axes, keep_dims));
+                const auto out_shape = reduce(in_shape, reduction_axes, false);
+                std::fill(out, out + shape_size(out_shape), 0);
 
-                for (const Coordinate& output_coord : output_transform)
-                {
-                    out[output_transform.index(output_coord)] = 0;
-                }
-
-                CoordinateTransform input_transform(input_shape);
+                const auto in_strides = row_major_strides(in_shape);
+                const auto out_strides = row_major_strides(out_shape);
 
+                CoordinateTransformBasic input_transform(in_shape);
                 for (const Coordinate& input_coord : input_transform)
                 {
-                    Coordinate output_coord = reduce(input_coord, reduction_axes, keep_dims);
-                    out[output_transform.index(output_coord)] =
-                        out[output_transform.index(output_coord)] ||
-                        arg[input_transform.index(input_coord)];
+                    const Coordinate output_coord = reduce(input_coord, reduction_axes, false);
+
+                    const size_t in_idx = std::inner_product(
+                        input_coord.begin(), input_coord.end(), in_strides.begin(), 0);
+                    const size_t out_idx = std::inner_product(
+                        output_coord.begin(), output_coord.end(), out_strides.begin(), 0);
+
+                    out[out_idx] = out[out_idx] || arg[in_idx];
                 }
             }
+
+            NGRAPH_DEPRECATED("Remove when arm plugin supports the new signature")
+            static inline void reduce_logical_or(const char* arg,
+                                                 char* out,
+                                                 const Shape& input_shape,
+                                                 const AxisSet& reduction_axes,
+                                                 bool)
+            {
+                reduce_logical_or(arg, out, input_shape, reduction_axes);
+            }
         } // namespace reference
     }     // namespace runtime
 } // namespace ngraph
diff --git a/ngraph/core/reference/src/runtime/reference/eval_helpers.cpp b/ngraph/core/reference/src/runtime/reference/eval_helpers.cpp
deleted file mode 100644
index 9aae80310a347f..00000000000000
--- a/ngraph/core/reference/src/runtime/reference/eval_helpers.cpp
+++ /dev/null
@@ -1,33 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include <algorithm>
-
-#include "ngraph/check.hpp"
-#include "ngraph/runtime/reference/eval_helpers.hpp"
-#include "ngraph/util.hpp"
-
-namespace ngraph
-{
-    namespace eval
-    {
-        AxisSet extract_reduction_axes(const HostTensorPtr& axes, const char* op_name)
-        {
-            const auto axes_in_tensor = host_tensor_2_vector<int64_t>(axes);
-
-            const bool negative_axis_received =
-                std::any_of(axes_in_tensor.begin(), axes_in_tensor.end(), [](const int64_t axis) {
-                    return axis < 0;
-                });
-
-            NGRAPH_CHECK(!negative_axis_received,
-                         "Negative axis value received in the ",
-                         op_name,
-                         " evaluation. This case is not supported.");
-
-            return AxisSet(
-                std::vector<AxisSet::value_type>(axes_in_tensor.begin(), axes_in_tensor.end()));
-        }
-    } // namespace eval
-} // namespace ngraph
diff --git a/ngraph/core/src/op/reduce_logical_and.cpp b/ngraph/core/src/op/reduce_logical_and.cpp
index c75c244d59056b..6b578c894ac4ec 100644
--- a/ngraph/core/src/op/reduce_logical_and.cpp
+++ b/ngraph/core/src/op/reduce_logical_and.cpp
@@ -3,10 +3,11 @@
 //
 
 #include "ngraph/op/reduce_logical_and.hpp"
+#include <ngraph/validation_util.hpp>
 #include "itt.hpp"
 #include "ngraph/log.hpp"
+#include "ngraph/op/util/evaluate_helpers.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
-#include "ngraph/runtime/reference/eval_helpers.hpp"
 #include "ngraph/runtime/reference/logical_reduction.hpp"
 
 using namespace ngraph;
@@ -32,28 +33,20 @@ shared_ptr<Node> op::v1::ReduceLogicalAnd::clone_with_new_inputs(const OutputVec
     return make_shared<op::v1::ReduceLogicalAnd>(new_args.at(0), new_args.at(1), get_keep_dims());
 }
 
-namespace
+namespace reduce_and
 {
     bool evaluate_reduce_logical_and(const HostTensorPtr& data,
-                                     const HostTensorPtr& axes,
                                      const HostTensorPtr& out,
+                                     const AxisSet& reduction_axes,
                                      bool keep_dims)
     {
-        if (data->get_element_type() != element::boolean ||
-            !axes->get_element_type().is_integral_number())
-        {
-            return false;
-        }
+        out->set_shape(reduce(data->get_shape(), reduction_axes, keep_dims));
         try
         {
-            const AxisSet reduction_axes = eval::extract_reduction_axes(axes, "ReduceLogicalAnd");
-
             runtime::reference::reduce_logical_and(data->get_data_ptr<char>(),
                                                    out->get_data_ptr<char>(),
                                                    data->get_shape(),
-                                                   reduction_axes,
-                                                   keep_dims);
-
+                                                   reduction_axes);
             return true;
         }
         catch (const ngraph_error& e)
@@ -62,16 +55,25 @@ namespace
             return false;
         }
     }
-} // namespace
+} // namespace reduce_and
 
 bool op::v1::ReduceLogicalAnd::evaluate(const HostTensorVector& outputs,
                                         const HostTensorVector& inputs) const
 {
     NGRAPH_OP_SCOPE(v1_ReduceLogicalAnd_evaluate);
+    NGRAPH_CHECK(validate_host_tensor_vector(inputs, 2));
+    NGRAPH_CHECK(validate_host_tensor_vector(outputs, 1));
     const auto& data = inputs[0];
     const auto& axes = inputs[1];
     const auto& out = outputs[0];
-    return evaluate_reduce_logical_and(data, axes, out, get_keep_dims());
+    if (data->get_element_type() != element::boolean ||
+        !axes->get_element_type().is_integral_number())
+    {
+        return false;
+    }
+    const auto reduction_axes = get_normalized_axes_from_tensor(
+        axes, data->get_partial_shape().rank(), get_friendly_name());
+    return reduce_and::evaluate_reduce_logical_and(data, out, reduction_axes, get_keep_dims());
 }
 
 bool op::v1::ReduceLogicalAnd::has_evaluate() const
diff --git a/ngraph/core/src/op/reduce_logical_or.cpp b/ngraph/core/src/op/reduce_logical_or.cpp
index a6afa5f77c8503..427b576bf9138e 100644
--- a/ngraph/core/src/op/reduce_logical_or.cpp
+++ b/ngraph/core/src/op/reduce_logical_or.cpp
@@ -3,10 +3,11 @@
 //
 
 #include "ngraph/op/reduce_logical_or.hpp"
+#include <ngraph/validation_util.hpp>
 #include "itt.hpp"
 #include "ngraph/log.hpp"
+#include "ngraph/op/util/evaluate_helpers.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
-#include "ngraph/runtime/reference/eval_helpers.hpp"
 #include "ngraph/runtime/reference/logical_reduction.hpp"
 
 using namespace ngraph;
@@ -32,28 +33,20 @@ shared_ptr<Node> op::v1::ReduceLogicalOr::clone_with_new_inputs(const OutputVect
     return make_shared<op::v1::ReduceLogicalOr>(new_args.at(0), new_args.at(1), get_keep_dims());
 }
 
-namespace
+namespace reduce_or
 {
     bool evaluate_reduce_logical_or(const HostTensorPtr& data,
-                                    const HostTensorPtr& axes,
                                     const HostTensorPtr& out,
+                                    const AxisSet& reduction_axes,
                                     bool keep_dims)
     {
-        if (data->get_element_type() != element::boolean ||
-            !axes->get_element_type().is_integral_number())
-        {
-            return false;
-        }
+        out->set_shape(reduce(data->get_shape(), reduction_axes, keep_dims));
         try
         {
-            const AxisSet reduction_axes = eval::extract_reduction_axes(axes, "ReduceLogicalOr");
-
             runtime::reference::reduce_logical_or(data->get_data_ptr<char>(),
                                                   out->get_data_ptr<char>(),
                                                   data->get_shape(),
-                                                  reduction_axes,
-                                                  keep_dims);
-
+                                                  reduction_axes);
             return true;
         }
         catch (const ngraph_error& e)
@@ -62,16 +55,25 @@ namespace
             return false;
         }
     }
-} // namespace
+} // namespace reduce_or
 
 bool op::v1::ReduceLogicalOr::evaluate(const HostTensorVector& outputs,
                                        const HostTensorVector& inputs) const
 {
     NGRAPH_OP_SCOPE(v1_ReduceLogicalOr_evaluate);
+    NGRAPH_CHECK(validate_host_tensor_vector(inputs, 2));
+    NGRAPH_CHECK(validate_host_tensor_vector(outputs, 1));
     const auto& data = inputs[0];
     const auto& axes = inputs[1];
     const auto& out = outputs[0];
-    return evaluate_reduce_logical_or(data, axes, out, get_keep_dims());
+    if (data->get_element_type() != element::boolean ||
+        !axes->get_element_type().is_integral_number())
+    {
+        return false;
+    }
+    const auto reduction_axes = get_normalized_axes_from_tensor(
+        axes, data->get_partial_shape().rank(), get_friendly_name());
+    return reduce_or::evaluate_reduce_logical_or(data, out, reduction_axes, get_keep_dims());
 }
 
 bool op::v1::ReduceLogicalOr::has_evaluate() const
diff --git a/ngraph/test/eval.cpp b/ngraph/test/eval.cpp
index f5602023c5c675..1ba827190ae0fb 100644
--- a/ngraph/test/eval.cpp
+++ b/ngraph/test/eval.cpp
@@ -1810,28 +1810,6 @@ TEST(eval, topk_v1_param_dyn_k0)
     ASSERT_EQ(result1_val, expec1);
 }
 
-TEST(eval, reduce_logical_and__neg_axis)
-{
-    const auto data = make_shared<op::Parameter>(element::boolean, Shape{2, 2, 2});
-    const auto axes = make_shared<op::Parameter>(element::i64, Shape{});
-
-    const auto op = make_shared<op::v1::ReduceLogicalAnd>(data, axes);
-
-    auto fun = make_shared<Function>(op, ParameterVector{data, axes});
-
-    auto result = make_shared<HostTensor>();
-
-    // when ReduceLogicalAnd node evaluator returns false -> the Function object throws
-    EXPECT_THROW(
-        fun->evaluate({result},
-                      {
-                          make_host_tensor<element::Type_t::boolean>(
-                              Shape{2, 2, 2}, {true, false, true, false, true, false, true, false}),
-                          make_host_tensor<element::Type_t::i64>(Shape{}, {-1}),
-                      }),
-        ngraph::ngraph_error);
-}
-
 TEST(eval, evaluate_static_scatter_update_basic_axes_indices_i32)
 {
     const Shape data_shape{3, 3};

From d777679de94559f4f2626a5d27caecce18ec4dc3 Mon Sep 17 00:00:00 2001
From: Anton Romanov <anton.romanov@intel.com>
Date: Tue, 15 Jun 2021 13:19:58 +0300
Subject: [PATCH 082/178] Added copyrights note into CMakeLists (#6155)

---
 thirdparty/zlib/CMakeLists.txt | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/thirdparty/zlib/CMakeLists.txt b/thirdparty/zlib/CMakeLists.txt
index b24d4abf323cd0..33b9222ec53103 100644
--- a/thirdparty/zlib/CMakeLists.txt
+++ b/thirdparty/zlib/CMakeLists.txt
@@ -1,3 +1,7 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+#
+
 PROJECT(zlib)
 
 if(NOT WIN32)

From 67c93ec6fa0954a8d9977a93b7ab56edc73858b2 Mon Sep 17 00:00:00 2001
From: Alina Kladieva <alina.kladieva@intel.com>
Date: Tue, 15 Jun 2021 13:47:06 +0300
Subject: [PATCH 083/178] Return sporadic GPU test cases (cannot reproduce
 54436) (#6127)

---
 .../plugin/gpu/shared_tests_instances/skip_tests_config.cpp    | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/skip_tests_config.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/skip_tests_config.cpp
index 07bd2a260984aa..5863309efe4fcf 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/skip_tests_config.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/skip_tests_config.cpp
@@ -52,9 +52,6 @@ std::vector<std::string> disabledTestPatterns() {
             R"(.*(LPT/StridedSliceTransformation).*)",
             // TODO: Issue: 48106
             R"(.*ConstantResultSubgraphTest.*inPrc=I16.*)",
-            // TODO: Issue: 54436
-            R"(.*LSTMSequence.*CompareWithRefs.*mode=PURE_SEQ_RAND_SEQ_LEN_PARAM.*direction=bidirectional_clip=0.7_netPRC=FP32.*)",
-            R"(.*LSTMSequence.*CompareWithRefs.*mode=CONVERT_TO_TI_RAND_SEQ_LEN_PARAM_seq.*direction=bidirectional_clip=0.7_netPRC=FP32.*)",
             // TODO: Issue: 54194
             R"(.*ActivationLayerTest.*SoftPlus.*)",
             // need to implement Export / Import

From a248d84894ddf8681192c40dbfc2a9fa90b9d5f0 Mon Sep 17 00:00:00 2001
From: Katarzyna Mitrus <katarzyna.mitrus@intel.com>
Date: Tue, 15 Jun 2021 14:02:37 +0200
Subject: [PATCH 084/178] [Spec] ShuffleChannels spec update (#5813)

---
 docs/ops/movement/ShuffleChannels_1.md | 47 ++++++++++++++++++--------
 1 file changed, 32 insertions(+), 15 deletions(-)

diff --git a/docs/ops/movement/ShuffleChannels_1.md b/docs/ops/movement/ShuffleChannels_1.md
index ec7cfc75d9db6f..e97f3350d2a405 100644
--- a/docs/ops/movement/ShuffleChannels_1.md
+++ b/docs/ops/movement/ShuffleChannels_1.md
@@ -8,12 +8,37 @@
 
 **Short description**: *ShuffleChannels* permutes data in the channel dimension of the input tensor.
 
+**Detailed description**:
+
+Input tensor of `data_shape` is always interpreted as 4D tensor with the following shape:
+
+        dim 0: data_shape[0] * data_shape[1] * ... * data_shape[axis-1]
+               (or 1 if axis == 0)
+        dim 1: group
+        dim 2: data_shape[axis] / group
+        dim 3: data_shape[axis+1] * data_shape[axis+2] * ... * data_shape[data_shape.size()-1]
+               (or 1 if axis points to last dimension)
+
+
+Trailing and leading to `axis` dimensions are flattened and reshaped back to the original shape after channels shuffling.
+
+
+The operation is equivalent to the following transformation of the input tensor `x` of shape `[N, C, H, W]` and `axis = 1`:
+
+\f[
+x' = reshape(x, [N, group, C / group, H * W])\\
+x'' = transpose(x', [0, 2, 1, 3])\\
+y = reshape(x'', [N, C, H, W])\\
+\f]
+
+where `group` is the layer attribute described below.
+
 **Attributes**:
 
 * *axis*
 
   * **Description**: *axis* specifies the index of a channel dimension.
-  * **Range of values**: an integer number in the range [-4, 3]
+  * **Range of values**: an integer number in the range `[-rank(data_shape), rank(data_shape) - 1]`
   * **Type**: `int`
   * **Default value**: 1
   * **Required**: *No*
@@ -21,30 +46,22 @@
 * *group*
 
   * **Description**: *group* specifies the number of groups to split the channel dimension into. This number must evenly divide the channel dimension size.
-  * **Range of values**: a positive integer
+  * **Range of values**: a positive integer in the range `[1, data_shape[axis]]`
   * **Type**: `int`
   * **Default value**: 1
   * **Required**: *No*
 
 **Inputs**:
 
-*   **1**: 4D input tensor of any supported data type. Required.
+*   **1**: `data` input tensor of type *T* and rank greater or equal to 1. **Required.**
 
 **Outputs**:
 
-*   **1**: 4D input tensor with shape and element type as for the input tensor.
+*   **1**: Output tensor with element type *T* and same shape as the input tensor.
 
-**Mathematical Formulation**
+**Types**
 
-The operation is the equivalent with the following transformation of the input tensor *x* of shape *[N, C, H, W]*:
-
-```
-x' = reshape(x, [N, group, C / group, H * W])
-x'' = transpose(x', [0, 2, 1, 3])
-y = reshape(x'', [N, C, H, W])
-```
-
-where `group` is the layer parameter described above and the `axis = 1`.
+* *T*: any supported numeric type.
 
 **Example**
 
@@ -68,4 +85,4 @@ where `group` is the layer parameter described above and the `axis = 1`.
         </port>
     </output>
 </layer>
-```
\ No newline at end of file
+```

From d7b2e4e50ec4b7ee7a7b21aae36a7be853e3b349 Mon Sep 17 00:00:00 2001
From: Bartek Szmelczynski <bartosz.szmelczynski@intel.com>
Date: Tue, 15 Jun 2021 14:24:40 +0200
Subject: [PATCH 085/178] revise space_to_batch spec (#5767)

* refactor part of the docs file to use \dots

* refector docs

* add function enclosure for docs

* split function enclosurs across lines

* add latex operations to spec

* fix style

* fix missing index

* remove link to tensorflow operation

* Remove commas from formula.

Co-authored-by: jdanieck <jozef.daniecki@intel.com>
---
 docs/ops/movement/SpaceToBatch_2.md | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)

diff --git a/docs/ops/movement/SpaceToBatch_2.md b/docs/ops/movement/SpaceToBatch_2.md
index 66c064e27bee35..eea9df8ced1acf 100644
--- a/docs/ops/movement/SpaceToBatch_2.md
+++ b/docs/ops/movement/SpaceToBatch_2.md
@@ -8,20 +8,20 @@
 
 **Detailed description**:
 
-The *SpaceToBatch* operation is similar to the TensorFlow* operation [SpaceToBatchND](https://www.tensorflow.org/api_docs/python/tf/space_to_batch_nd)
-
 The operation is equivalent to the following transformation of the input tensor `data` of shape `[batch, D_1, D_2 ... D_{N - 1}]` and `block_shape`, `pads_begin`, `pads_end` of shapes `[N]` to *Y* output tensor.
 
-    Zero-pad the start and end of dimensions [D_0, ..., D_{N - 1}] of the input according to `pads_begin` and `pads_end`:
-    note: P_0 for batch dimension is expected to be 0 (no-padding).
-    x = [batch + P_0, D_1 + P_1, D_2 + P_2, ..., D_{N - 1} + P_{N - 1}], where P_i = pads_begin[i] + pads_end[i]
-
-    note: B_0 for batch is ignored.
-    x' = reshape(x, [batch, (D_1 + P_1) / B_1, B_1, (D_2 + P_2) / B_2, B_2, ..., (D_{N - 1} + P_{N - 1}) / B_{N - 1}, B_{N - 1}]), where B_i = block_shape[i]
+Zero-pad the start and end of dimensions  \f$[D_0, \dots, D_{N - 1}]\f$ of the input according to `pads_begin` and `pads_end`:
 
-    x'' = transpose(x',  [2, 4, ..., (N - 1) + (N - 1), 0, 1, 3, ..., N + (N - 1)])
+\f[x = [batch + P_0, D_1 + P_1, D_2 + P_2, \dots, D_{N - 1} + P_{N - 1}]\f]
+\f[x' = reshape(x, [batch, \frac{D_1 + P_1}{B_1}, B_1, \frac{D_2 + P_2}{B_2}, B_2, \dots, \frac{D_{N - 1} + P_{N - 1}}{B_{N - 1}}, B_{N - 1}])\f]
+\f[x'' = transpose(x',  [2, 4, \dots, (N - 1) + (N - 1), 0, 1, 3, \dots, N + (N - 1)])\f]
+\f[y = reshape(x'', [batch \times B_1 \times \dots \times B_{N - 1}, \frac{D_1 + P_1}{B_1}, \frac{D_2 + P_2}{B_2}, \dots, \frac{D_{N - 1} + P_{N - 1}}{B_{N - 1}}]\f]
 
-    y = reshape(x'', [batch * B_1 * ... * B_{N - 1}, (D_1 + P_1) / B_1, (D_2 + P_2) / B_2, ... , (D_{N - 1} + P_{N - 1}) / B_{N - 1}])
+where
+- \f$P_i\f$ = pads_begin[i] + pads_end[i]
+- \f$B_i\f$ = block_shape[i]
+- \f$P_0\f$ for batch dimension is expected to be 0 (no-padding)
+- \f$B_0\f$ for batch is ignored
 
 **Attributes**
 
@@ -36,7 +36,7 @@ The operation is equivalent to the following transformation of the input tensor
  
 **Outputs**
 
-*   **1**: N-D tensor with shape `[batch * block_shape[0] * block_shape[1] * ... * block_shape[N - 1], (pads_begin[1] + D_1 + pads_end[1]) / block_shape[1], (pads_begin[2] + D_2 + pads_end[2]) / block_shape[2], ..., (pads_begin[N - 1] + D_{N - 1} + pads_end[N - 1]) / block_shape[N - 1]` of the same type as `data` input. 
+*   **1**: N-D tensor with shape `[batch * block_shape[0] * block_shape[1] * ... * block_shape[N - 1], (D_1 + pads_begin[1] + pads_end[1]) / block_shape[1], (D_2 + pads_begin[2] + pads_end[2]) / block_shape[2], ..., (D_{N -1} + pads_begin[N - 1] + pads_end[N - 1]) / block_shape[N - 1]` of the same type as `data` input. 
 
 **Types**
 

From 5ebdcfe15b36d587e323da949f3deaaea21b78c4 Mon Sep 17 00:00:00 2001
From: Jozef Daniecki <jozef.daniecki@intel.com>
Date: Tue, 15 Jun 2021 14:35:47 +0200
Subject: [PATCH 086/178] StridedSlice operation specification refactoring
 (#5887)

* StridedSlice spec refactored against explicit type indication.

* Add name to data input.

* Add new examples.

* Changed T to 'any supported type'.

* Remove mention about 'generalized python indexing' from short
description.
---
 docs/ops/movement/StridedSlice_1.md | 97 +++++++++++++++++++++++------
 1 file changed, 79 insertions(+), 18 deletions(-)

diff --git a/docs/ops/movement/StridedSlice_1.md b/docs/ops/movement/StridedSlice_1.md
index 6c07665d8f930c..41742e20652b13 100644
--- a/docs/ops/movement/StridedSlice_1.md
+++ b/docs/ops/movement/StridedSlice_1.md
@@ -4,14 +4,13 @@
 
 **Category**: Data movement operation
 
-**Short description**: *StridedSlice* extracts a strided slice of a tensor.
- It is similar to generalized array indexing in Python\*.
+**Short description**: *StridedSlice* extracts a strided slice of a tensor. 
 
 **Attributes**
 
 *   *begin_mask*
 
-    * **Description**: *begin_mask* is a bit mask. *begin_mask[i]* equal to 1 means that the corresponding dimension of the `begin` input is ignored and the 'real' beginning of the tensor is used along corresponding dimension.
+    * **Description**: *begin_mask* is a bit mask. *begin_mask[i]* equal to `1` means that the corresponding dimension of the `begin` input is ignored and the 'real' beginning of the tensor is used along corresponding dimension.
     * **Range of values**: a list of `0`s and `1`s
     * **Type**: `int[]`
     * **Default value**: None
@@ -19,7 +18,7 @@
 
 *   *end_mask*
 
-    * **Description**: *end_mask* is a bit mask. If *end_mask[i]* is 1, the corresponding dimension of the `end` input is ignored and the real 'end' of the tensor is used along corresponding dimension.
+    * **Description**: *end_mask* is a bit mask. If *end_mask[i]* is `1`, the corresponding dimension of the `end` input is ignored and the real 'end' of the tensor is used along corresponding dimension.
     * **Range of values**: a list of `0`s and `1`s
     * **Type**: `int[]`
     * **Default value**: None
@@ -27,7 +26,7 @@
 
 *   *new_axis_mask*
 
-    * **Description**: *new_axis_mask* is a bit mask. If *new_axis_mask[i]* is 1, a length 1 dimension is inserted on the `i`-th position of input tensor.
+    * **Description**: *new_axis_mask* is a bit mask. If *new_axis_mask[i]* is `1`, a length 1 dimension is inserted on the `i`-th position of input tensor.
     * **Range of values**: a list of `0`s and `1`s
     * **Type**: `int[]`
     * **Default value**: `[0]`
@@ -35,7 +34,7 @@
 
 *   *shrink_axis_mask*
 
-    * **Description**: *shrink_axis_mask* is a bit mask. If *shrink_axis_mask[i]* is 1, the dimension on the `i`-th position is deleted.
+    * **Description**: *shrink_axis_mask* is a bit mask. If *shrink_axis_mask[i]* is `1`, the dimension on the `i`-th position is deleted.
     * **Range of values**: a list of `0`s and `1`s
     * **Type**: `int[]`
     * **Default value**: `[0]`
@@ -51,21 +50,83 @@
 
 **Inputs**:
 
-*   **1**: Multidimensional input tensor to be sliced. Required.
+*   **1**: `data` - input tensor to be sliced of type `T` and arbitrary shape. **Required.**
 
-*   **2**: `begin` input - 1D input tensor with begin indexes for input tensor slicing. Required.
-           Out-of-bounds values are silently clamped. If `begin_mask[i]` is 1, the value of `begin[i]` is ignored
-           and the range of the appropriate dimension starts from 0.
-           Negative values mean indexing starts from the end. For example, if `foo=[1,2,3]`, `begin[0]=-1` means `begin[0]=3`.
+*   **2**: `begin` - 1D tensor of type `T_IND` with begin indexes for input tensor slicing. **Required.**  
+    Out-of-bounds values are silently clamped. If `begin_mask[i]` is `1`, the value of `begin[i]` is ignored and the range of the appropriate dimension starts from `0`. Negative values mean indexing starts from the end. For example, if `data=[1,2,3]`, `begin[0]=-1` means `begin[0]=3`.
 
-*   **3**: `end` input - 1D input tensor with end indexes for input tensor slicing. Required.
-           Out-of-bounds values will be silently clamped. If `end_mask[i]` is 1, the value of `end[i]` is ignored
-           and the full range of the appropriate dimension is used instead.
-           Negative values mean indexing starts from the end. For example, if `foo=[1,2,3]`, `end[0]=-1` means `end[0]=3`.
+*   **3**: `end` - 1D tensor of type `T_IND` with end indexes for input tensor slicing. **Required.**  
+    Out-of-bounds values will be silently clamped. If `end_mask[i]` is `1`, the value of `end[i]` is ignored and the full range of the appropriate dimension is used instead. Negative values mean indexing starts from the end. For example, if `data=[1,2,3]`, `end[0]=-1` means `end[0]=3`.
 
-*   **4**: `stride` input - 1D input tensor with strides. Optional.
+*   **4**: `stride` - 1D tensor of type `T_IND` with strides. **Optional.**
 
-**Example**
+**Types**
+* *T*: any supported type.
+* *T_IND*: any supported integer type.
+
+**Example**  
+Example of `begin_mask` & `end_mask` usage.  
+```xml
+<layer ... type="StridedSlice" ...>
+    <data begin_mask="0,1,1" ellipsis_mask="0,0,0" end_mask="1,1,0" new_axis_mask="0,0,0" shrink_axis_mask="0,0,0"/>
+    <input>
+        <port id="0">
+            <dim>2</dim>
+            <dim>3</dim>
+            <dim>4</dim>
+        </port>
+        <port id="1">
+            <dim>2</dim> <!-- begin: [1, 0, 0] -->
+        </port>
+        <port id="2">
+            <dim>2</dim> <!-- end: [0, 0, 2] -->
+        </port>
+        <port id="3">
+            <dim>2</dim> <!-- stride: [1, 1, 1] -->
+        </port>
+    </input>
+    <output>
+        <port id="4">            
+            <dim>1</dim>
+            <dim>3</dim>
+            <dim>2</dim>
+        </port>
+    </output>
+</layer>
+```
+
+Example of `new_axis_mask` usage.
+```xml
+<layer ... type="StridedSlice" ...>
+    <data begin_mask="0,1,1" ellipsis_mask="0,0,0" end_mask="0,1,1" new_axis_mask="1,0,0" shrink_axis_mask="0,0,0"/>
+    <input>
+        <port id="0">
+            <dim>2</dim>
+            <dim>3</dim>
+            <dim>4</dim>
+        </port>
+        <port id="1">
+            <dim>2</dim>
+        </port>
+        <port id="2">
+            <dim>2</dim>
+        </port>
+        <port id="3">
+            <dim>2</dim>
+        </port>
+    </input>
+    <output>
+        <port id="4">
+            <dim>1</dim>
+            <dim>2</dim>
+            <dim>3</dim>
+            <dim>4</dim>
+        </port>
+    </output>
+</layer>
+```
+
+Example of `shrink_axis_mask` usage.
 ```xml
 <layer ... type="StridedSlice" ...>
     <data begin_mask="1,0,1,1,1" ellipsis_mask="0,0,0,0,0" end_mask="1,0,1,1,1" new_axis_mask="0,0,0,0,0" shrink_axis_mask="0,1,0,0,0"/>
@@ -96,4 +157,4 @@
         </port>
     </output>
 </layer>
-```
\ No newline at end of file
+```

From db74707835cd2ed08e11dd38cbfac9c038b33def Mon Sep 17 00:00:00 2001
From: Patryk Elszkowski <patryk.elszkowski@intel.com>
Date: Tue, 15 Jun 2021 15:08:36 +0200
Subject: [PATCH 087/178] Enable reshape test (#6161)

---
 ngraph/test/runtime/ie/unit_test.manifest | 19 -------------------
 1 file changed, 19 deletions(-)

diff --git a/ngraph/test/runtime/ie/unit_test.manifest b/ngraph/test/runtime/ie/unit_test.manifest
index e76ac6f2b97d91..1537b823c1795d 100644
--- a/ngraph/test/runtime/ie/unit_test.manifest
+++ b/ngraph/test/runtime/ie/unit_test.manifest
@@ -386,31 +386,12 @@ reduce_sum_large_1d_to_scalar
 # Doesn't throw expected exception type.
 unhandled_op
 
-# Const layer Constant_6325 has incorrect dimensions in the output data 0
-reshape_t2s
-
 # Expected equality of these values:
 #  (vector<char>{42})
 #    Which is: { '*' (42, 0x2A) }
 #  read_vector<char>(result)
 #    Which is: { '\0' }
 reshape_s2t1
-reshape_v2m_col
-reshape_v2m_row
-reshape_v2t_middle
-reshape_m2m_same
-reshape_m2m_transpose
-reshape_m2m_dim_change_transpose
-reshape_3d_transpose_021
-reshape_3d_transpose_210
-reshape_3d_transpose_201
-reshape_3d_transpose_102
-reshape_3d_transpose_120
-reshape_4d_transpose
-reshape_4d_no_transpose
-reshape_transposed_shape_change
-reshape_6d
-reshape_shufflenet_5d
 dyn_group_convolution_backprop_data
 group_conv
 group_conv_striding

From c3d1c2e420cd8f77b1e8a4946eedd1c1774c1287 Mon Sep 17 00:00:00 2001
From: Gabriele Galiero Casay <gabriele.galiero.casay@intel.com>
Date: Tue, 15 Jun 2021 15:12:12 +0200
Subject: [PATCH 088/178] Revise reference implementation for ReduceMax
 operation (#5792)

* Revise reference implementation for ReduceMax operation

* Refactor backend unit tests

 * Move tests with zero dims to op_eval
 * Remove test with double elem type

* Fix code style

* Added minor changes

 * Replace CoordinateTransform for CoordinateTransformBasic
 * Added constant expression to set keep_dims as false

* Add const qualifier to local variables

* Use host tensor to retrieve and normalize axes
---
 .../ngraph/runtime/reference/log_softmax.hpp  |   2 +-
 .../include/ngraph/runtime/reference/max.hpp  |  35 +-
 .../ngraph/runtime/reference/softmax.hpp      |   2 +-
 ngraph/core/src/op/max.cpp                    |  12 +-
 ngraph/test/CMakeLists.txt                    |   1 +
 ngraph/test/backend/reduce_max.in.cpp         | 336 ------------------
 ngraph/test/op_eval/reduce_max.cpp            | 312 ++++++++++++++++
 ngraph/test/runtime/ie/unit_test.manifest     |  14 -
 .../runtime/interpreter/unit_test.manifest    |   2 -
 9 files changed, 343 insertions(+), 373 deletions(-)
 create mode 100644 ngraph/test/op_eval/reduce_max.cpp

diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/log_softmax.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/log_softmax.hpp
index 27198f465b30ff..2cf445ee54f367 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/log_softmax.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/log_softmax.hpp
@@ -24,7 +24,7 @@ namespace ngraph
                 auto temp_max = std::vector<T>(temp_elements, 0);
                 auto temp_sum = std::vector<T>(temp_elements, 0);
 
-                max(arg, temp_max.data(), shape, axes, true);
+                max(arg, temp_max.data(), shape, axes);
 
                 CoordinateTransform transform(shape);
                 CoordinateTransform temp_transform(temp_shape);
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/max.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/max.hpp
index d8e39fefd65aee..8bed66cdefbfe0 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/max.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/max.hpp
@@ -6,6 +6,7 @@
 
 #include <cmath>
 #include <limits>
+#include <numeric>
 
 #include "ngraph/coordinate_transform.hpp"
 #include "ngraph/shape_util.hpp"
@@ -17,35 +18,35 @@ namespace ngraph
         namespace reference
         {
             template <typename T>
-            void max(const T* arg,
-                     T* out,
-                     const Shape& in_shape,
-                     const AxisSet& reduction_axes,
-                     bool keep_dims)
+            void max(const T* arg, T* out, const Shape& in_shape, const AxisSet& reduction_axes)
             {
                 T minval = std::numeric_limits<T>::has_infinity
                                ? T(-std::numeric_limits<T>::infinity())
                                : std::numeric_limits<T>::min();
 
-                auto out_shape = reduce(in_shape, reduction_axes, keep_dims);
-                CoordinateTransform output_transform(out_shape);
+                constexpr bool dont_keep_dims_in_output = false;
+                const auto out_shape = reduce(in_shape, reduction_axes, dont_keep_dims_in_output);
+                std::fill(out, out + shape_size(out_shape), minval);
 
-                for (const Coordinate& output_coord : output_transform)
-                {
-                    out[output_transform.index(output_coord)] = minval;
-                }
-
-                CoordinateTransform input_transform(in_shape);
+                const auto in_strides = row_major_strides(in_shape);
+                const auto out_strides = row_major_strides(out_shape);
 
+                CoordinateTransformBasic input_transform(in_shape);
                 for (const Coordinate& input_coord : input_transform)
                 {
-                    Coordinate output_coord = reduce(input_coord, reduction_axes, keep_dims);
+                    const Coordinate output_coord =
+                        reduce(input_coord, reduction_axes, dont_keep_dims_in_output);
+
+                    const size_t in_idx = std::inner_product(
+                        input_coord.begin(), input_coord.end(), in_strides.begin(), 0);
+                    const size_t out_idx = std::inner_product(
+                        output_coord.begin(), output_coord.end(), out_strides.begin(), 0);
 
-                    T x = arg[input_transform.index(input_coord)];
-                    T max = out[output_transform.index(output_coord)];
+                    const T x = arg[in_idx];
+                    const T max = out[out_idx];
                     if (x > max)
                     {
-                        out[output_transform.index(output_coord)] = x;
+                        out[out_idx] = x;
                     }
                 }
             }
diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/softmax.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/softmax.hpp
index 94cb0549b7310d..a8544177aec44e 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/softmax.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/softmax.hpp
@@ -23,7 +23,7 @@ namespace ngraph
                 auto temp_elements = shape_size(temp_shape);
                 auto temp_ptr = new T[temp_elements];
 
-                max(arg, temp_ptr, shape, axes, true);
+                max(arg, temp_ptr, shape, axes);
 
                 CoordinateTransform transform(shape);
                 CoordinateTransform temp_transform(temp_shape);
diff --git a/ngraph/core/src/op/max.cpp b/ngraph/core/src/op/max.cpp
index 318910aa70ce6c..f4332ef9317de7 100644
--- a/ngraph/core/src/op/max.cpp
+++ b/ngraph/core/src/op/max.cpp
@@ -3,8 +3,10 @@
 //
 
 #include "ngraph/op/max.hpp"
+#include <ngraph/validation_util.hpp>
 #include "itt.hpp"
 #include "ngraph/graph_util.hpp"
+#include "ngraph/op/util/evaluate_helpers.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
 #include "ngraph/runtime/reference/max.hpp"
 #include "ngraph/shape_util.hpp"
@@ -22,7 +24,7 @@ namespace maxop
     {
         out->set_shape(reduce(arg->get_shape(), axes, keep_dims));
         runtime::reference::max(
-            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes, keep_dims);
+            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes);
         return true;
     }
 
@@ -67,7 +69,13 @@ bool op::v1::ReduceMax::evaluate(const HostTensorVector& outputs,
                                  const HostTensorVector& inputs) const
 {
     NGRAPH_OP_SCOPE(v1_ReduceMax_evaluate);
-    return maxop::evaluate_max(inputs[0], outputs[0], get_reduction_axes(), get_keep_dims());
+    NGRAPH_CHECK(validate_host_tensor_vector(inputs, 2));
+    NGRAPH_CHECK(validate_host_tensor_vector(outputs, 1));
+
+    const auto reduction_axes = get_normalized_axes_from_tensor(
+        inputs[1], inputs[0]->get_partial_shape().rank(), get_friendly_name());
+
+    return maxop::evaluate_max(inputs[0], outputs[0], reduction_axes, get_keep_dims());
 }
 
 bool op::v1::ReduceMax::has_evaluate() const
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 47dfbbd3e495b3..b9e65209bb8367 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -71,6 +71,7 @@ set(SRC
     op_eval/memory.cpp
     op_eval/mish.cpp
     op_eval/non_zero.cpp
+    op_eval/reduce_max.cpp
     op_eval/reduce_prod.cpp
     op_eval/reduce_sum.cpp
     op_eval/roi_align.cpp
diff --git a/ngraph/test/backend/reduce_max.in.cpp b/ngraph/test/backend/reduce_max.in.cpp
index a1288f416a8409..28c1a3b97cd343 100644
--- a/ngraph/test/backend/reduce_max.in.cpp
+++ b/ngraph/test/backend/reduce_max.in.cpp
@@ -104,123 +104,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_max_matrix_rows_int32)
     test_case.run();
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_matrix_rows_zero)
-{
-    Shape shape_a{3, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
-
-    // Create some tensors for input/output
-    std::vector<float> a{};
-
-    auto test_case = test::TestCase<TestEngine>(f);
-    test_case.add_input<float>({a});
-    test_case.add_expected_output<float>(shape_rt,
-                                         {-std::numeric_limits<float>::infinity(),
-                                          -std::numeric_limits<float>::infinity(),
-                                          -std::numeric_limits<float>::infinity()});
-    test_case.run();
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_matrix_rows_zero_int32)
-{
-    Shape shape_a{3, 0};
-    auto A = make_shared<op::Parameter>(element::i32, shape_a);
-    Shape shape_rt{3};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::i32, shape_a);
-    copy_data(a, vector<int32_t>{});
-    auto result = backend->create_tensor(element::i32, shape_rt);
-    copy_data(result, vector<int32_t>({3, 3, 3}));
-
-    int32_t minval = std::numeric_limits<int32_t>::has_infinity
-                         ? -std::numeric_limits<int32_t>::infinity()
-                         : std::numeric_limits<int32_t>::min();
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<int32_t>{minval, minval, minval}), read_vector<int32_t>(result));
-    EXPECT_EQ((vector<int32_t>{}), read_vector<int32_t>(a));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_matrix_cols_zero)
-{
-    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
-    Shape shape_a{0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity(),
-                             -std::numeric_limits<float>::infinity()}),
-              read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_vector_zero)
-{
-    Shape shape_a{0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity()}), read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_matrix_to_scalar_zero_by_zero)
-{
-    Shape shape_a{0, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity()}), read_vector<float>(result));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_max_3d_to_matrix_most_sig)
 {
     Shape shape_a{3, 3, 3};
@@ -319,55 +202,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_max_3d_to_scalar_int32)
     EXPECT_EQ((vector<int32_t>{14}), read_vector<int32_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_3d_to_scalar_double)
-{
-    Shape shape_a{3, 3, 3};
-    auto A = make_shared<op::Parameter>(element::f64, shape_a);
-    Shape shape_rt{};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{3}, vector<int32_t>{0, 1, 2});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f64, shape_a);
-    copy_data(a, vector<double>{1,  2,  3,  4,  5, 6, 7, 8, 9, 10, 11, 12, 13, 14,
-                                13, 12, 11, 10, 9, 8, 7, 6, 5, 4,  3,  2,  1});
-    auto result = backend->create_tensor(element::f64, shape_rt);
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<double>{14}), read_vector<double>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_3d_eliminate_zero_dim)
-{
-    Shape shape_a{3, 0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-
-    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
-    // right value.
-    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
-
-    float mi = -std::numeric_limits<float>::infinity();
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{mi, mi, mi, mi, mi, mi}), read_vector<float>(result));
-}
-
 // ----------------------- keep dims = true ----------------------- //
 
 NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_to_scalar)
@@ -473,127 +307,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_matrix_rows_int32)
     EXPECT_EQ((vector<int32_t>{2, 4, 6}), read_vector<int32_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_matrix_rows_zero)
-{
-    Shape shape_a{3, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity(),
-                             -std::numeric_limits<float>::infinity(),
-                             -std::numeric_limits<float>::infinity()}),
-              read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_matrix_rows_zero_int32)
-{
-    Shape shape_a{3, 0};
-    auto A = make_shared<op::Parameter>(element::i32, shape_a);
-    Shape shape_rt{3, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::i32, shape_a);
-    copy_data(a, vector<int32_t>{});
-    auto result = backend->create_tensor(element::i32, shape_rt);
-    copy_data(result, vector<int32_t>({3, 3, 3}));
-
-    int32_t minval = std::numeric_limits<int32_t>::has_infinity
-                         ? -std::numeric_limits<int32_t>::infinity()
-                         : std::numeric_limits<int32_t>::min();
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<int32_t>{minval, minval, minval}), read_vector<int32_t>(result));
-    EXPECT_EQ((vector<int32_t>{}), read_vector<int32_t>(a));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_matrix_cols_zero)
-{
-    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
-    Shape shape_a{0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity(),
-                             -std::numeric_limits<float>::infinity()}),
-              read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_vector_zero)
-{
-    Shape shape_a{0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity()}), read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_matrix_to_scalar_zero_by_zero)
-{
-    Shape shape_a{0, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity()}), read_vector<float>(result));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_3d_to_matrix_most_sig)
 {
     Shape shape_a{3, 3, 3};
@@ -711,55 +424,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_3d_to_scalar_int32)
     EXPECT_EQ((vector<int32_t>{14}), read_vector<int32_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_3d_to_scalar_double)
-{
-    Shape shape_a{3, 3, 3};
-    auto A = make_shared<op::Parameter>(element::f64, shape_a);
-    Shape shape_rt{1, 1, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{3}, vector<int32_t>{0, 1, 2});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f64, shape_a);
-    copy_data(a, vector<double>{1,  2,  3,  4,  5, 6, 7, 8, 9, 10, 11, 12, 13, 14,
-                                13, 12, 11, 10, 9, 8, 7, 6, 5, 4,  3,  2,  1});
-    auto result = backend->create_tensor(element::f64, shape_rt);
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_TRUE(test::all_close_f((vector<double>{14}), read_vector<double>(result)));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_max_keep_3d_eliminate_zero_dim)
-{
-    Shape shape_a{3, 0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 1, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-
-    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
-    // right value.
-    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
-
-    float mi = -std::numeric_limits<float>::infinity();
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{mi, mi, mi, mi, mi, mi}), read_vector<float>(result));
-}
-
 // Dynamic
 
 NGRAPH_TEST(${BACKEND_NAME}, reduce_max_matrix_columns_dynamic)
diff --git a/ngraph/test/op_eval/reduce_max.cpp b/ngraph/test/op_eval/reduce_max.cpp
new file mode 100644
index 00000000000000..2ce244f9eb204b
--- /dev/null
+++ b/ngraph/test/op_eval/reduce_max.cpp
@@ -0,0 +1,312 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/test_control.hpp"
+#include "util/all_close.hpp"
+#include "util/all_close_f.hpp"
+#include "util/ndarray.hpp"
+
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+
+TEST(op_eval, reduce_max_matrix_rows_zero)
+{
+    Shape shape_a{3, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+
+    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity(),
+                             -std::numeric_limits<float>::infinity(),
+                             -std::numeric_limits<float>::infinity()}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_max_matrix_rows_zero_int32)
+{
+    Shape shape_a{3, 0};
+    auto A = make_shared<op::Parameter>(element::i32, shape_a);
+    Shape shape_rt{3};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::i32, shape_a);
+    copy_data(a, vector<int32_t>{});
+    auto result = backend->create_tensor(element::i32, shape_rt);
+    copy_data(result, vector<int32_t>({3, 3, 3}));
+
+    int32_t minval = std::numeric_limits<int32_t>::has_infinity
+                         ? -std::numeric_limits<int32_t>::infinity()
+                         : std::numeric_limits<int32_t>::min();
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<int32_t>{minval, minval, minval}), read_vector<int32_t>(result));
+    EXPECT_EQ((vector<int32_t>{}), read_vector<int32_t>(a));
+}
+
+TEST(op_eval, reduce_max_matrix_cols_zero)
+{
+    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
+    Shape shape_a{0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity(),
+                             -std::numeric_limits<float>::infinity()}),
+              read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_max_vector_zero)
+{
+    Shape shape_a{0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity()}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_max_matrix_to_scalar_zero_by_zero)
+{
+    Shape shape_a{0, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity()}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_max_3d_eliminate_zero_dim)
+{
+    Shape shape_a{3, 0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+
+    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
+    // right value.
+    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
+
+    float mi = -std::numeric_limits<float>::infinity();
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{mi, mi, mi, mi, mi, mi}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_max_keep_matrix_rows_zero)
+{
+    Shape shape_a{3, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity(),
+                             -std::numeric_limits<float>::infinity(),
+                             -std::numeric_limits<float>::infinity()}),
+              read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_max_keep_matrix_rows_zero_int32)
+{
+    Shape shape_a{3, 0};
+    auto A = make_shared<op::Parameter>(element::i32, shape_a);
+    Shape shape_rt{3, 1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::i32, shape_a);
+    copy_data(a, vector<int32_t>{});
+    auto result = backend->create_tensor(element::i32, shape_rt);
+    copy_data(result, vector<int32_t>({3, 3, 3}));
+
+    int32_t minval = std::numeric_limits<int32_t>::has_infinity
+                         ? -std::numeric_limits<int32_t>::infinity()
+                         : std::numeric_limits<int32_t>::min();
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<int32_t>{minval, minval, minval}), read_vector<int32_t>(result));
+    EXPECT_EQ((vector<int32_t>{}), read_vector<int32_t>(a));
+}
+
+TEST(op_eval, reduce_max_keep_matrix_cols_zero)
+{
+    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
+    Shape shape_a{0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity(),
+                             -std::numeric_limits<float>::infinity()}),
+              read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_max_keep_vector_zero)
+{
+    Shape shape_a{0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity()}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_max_keep_matrix_to_scalar_zero_by_zero)
+{
+    Shape shape_a{0, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1, 1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{-std::numeric_limits<float>::infinity()}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_max_keep_3d_eliminate_zero_dim)
+{
+    Shape shape_a{3, 0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 1, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMax>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+
+    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
+    // right value.
+    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
+
+    float mi = -std::numeric_limits<float>::infinity();
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{mi, mi, mi, mi, mi, mi}), read_vector<float>(result));
+}
\ No newline at end of file
diff --git a/ngraph/test/runtime/ie/unit_test.manifest b/ngraph/test/runtime/ie/unit_test.manifest
index 1537b823c1795d..a0a7c8735f58d3 100644
--- a/ngraph/test/runtime/ie/unit_test.manifest
+++ b/ngraph/test/runtime/ie/unit_test.manifest
@@ -474,21 +474,7 @@ reduce_mean_matrix_rows_int32
 reduce_mean_keep_to_scalar_int8
 reduce_mean_keep_matrix_rows_int32
 reduce_max_to_scalar_int8
-reduce_max_matrix_rows_zero
-reduce_max_matrix_rows_zero_int32
-reduce_max_matrix_cols_zero
-reduce_max_vector_zero
-reduce_max_matrix_to_scalar_zero_by_zero
-reduce_max_3d_to_scalar_double
-reduce_max_3d_eliminate_zero_dim
 reduce_max_keep_to_scalar_int8
-reduce_max_keep_matrix_rows_zero
-reduce_max_keep_matrix_rows_zero_int32
-reduce_max_keep_matrix_cols_zero
-reduce_max_keep_vector_zero
-reduce_max_keep_matrix_to_scalar_zero_by_zero
-reduce_max_keep_3d_to_scalar_double
-reduce_max_keep_3d_eliminate_zero_dim
 
 # Incorrect precision f64!
 sum_trivial_in_double
diff --git a/ngraph/test/runtime/interpreter/unit_test.manifest b/ngraph/test/runtime/interpreter/unit_test.manifest
index 9e1231491521e3..799025f3b92752 100644
--- a/ngraph/test/runtime/interpreter/unit_test.manifest
+++ b/ngraph/test/runtime/interpreter/unit_test.manifest
@@ -41,9 +41,7 @@ INTERPRETER.reduce_min_keep_to_scalar_int8
 INTERPRETER.reduce_mean_to_scalar_int8
 INTERPRETER.reduce_mean_keep_to_scalar_int8
 INTERPRETER.reduce_max_to_scalar_int8
-INTERPRETER.reduce_max_3d_to_scalar_double
 INTERPRETER.reduce_max_keep_to_scalar_int8
-INTERPRETER.reduce_max_keep_3d_to_scalar_double
 INTERPRETER.product_to_scalar_int8
 INTERPRETER.max_pool_uint8
 INTERPRETER.max_pool_int8

From b4e6028f8b4bdc8fca3bb646b31688bd0d894cf9 Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Tue, 15 Jun 2021 16:29:26 +0300
Subject: [PATCH 089/178] Fixed shellcheck issues (#6141)

---
 CMakeLists.txt                                |  9 +++--
 .../scripts/run_tests_myriad_multistick.sh    |  2 +-
 scripts/demo/demo_benchmark_app.sh            | 26 ++++++-------
 scripts/demo/demo_security_barrier_camera.sh  | 20 +++++-----
 .../demo_squeezenet_download_convert_run.sh   | 27 +++++++------
 scripts/demo/utils.sh                         |  2 +
 .../install_4_14_kernel.sh                    | 38 +++++++++----------
 .../install_NCS_udev_rules.sh                 |  2 -
 8 files changed, 63 insertions(+), 63 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index e0706a72e87e74..3602750435c550 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -169,10 +169,11 @@ ie_shellcheck_process(DIRECTORY "${OpenVINO_MAIN_SOURCE_DIR}"
                            "${IE_MAIN_SOURCE_DIR}/thirdparty"
                            "${IE_MAIN_SOURCE_DIR}/temp"
                            # TODO fix and enable back:
-                           "${OpenVINO_MAIN_SOURCE_DIR}/scripts/install_dependencies"
-                           "${OpenVINO_MAIN_SOURCE_DIR}/scripts/demo"
-                           "${OpenVINO_MAIN_SOURCE_DIR}/ngraph"
-                           "${IE_MAIN_SOURCE_DIR}/scripts")
+                           "${OpenVINO_MAIN_SOURCE_DIR}/inference-engine/scripts/dependencies.sh"
+                           "${OpenVINO_MAIN_SOURCE_DIR}/scripts/install_dependencies/install_NEO_OCL_driver.sh"
+                           "${OpenVINO_MAIN_SOURCE_DIR}/scripts/install_dependencies/install_openvino_dependencies.sh"
+                           "${OpenVINO_MAIN_SOURCE_DIR}/ngraph/python/tests/test_onnx/model_zoo_preprocess.sh"
+                           )
 
 #
 # cpack
diff --git a/inference-engine/scripts/run_tests_myriad_multistick.sh b/inference-engine/scripts/run_tests_myriad_multistick.sh
index 468817c6f75109..73befb9b366e2d 100755
--- a/inference-engine/scripts/run_tests_myriad_multistick.sh
+++ b/inference-engine/scripts/run_tests_myriad_multistick.sh
@@ -55,7 +55,7 @@ fi
 
 if [[ "${APPS_TO_RUN}" -ge 4 ]] ; then
     # For more then 4 multidevice testing
-    for (( VAR = 4; VAR <= ${APPS_TO_RUN}; ++VAR )); do
+    for (( VAR = 4; VAR <= APPS_TO_RUN; ++VAR )); do
         ./${APP_NAME} --gtest_filter=*VPURegTest*YOLO*myriad* &
         pids+=" $!"
     done
diff --git a/scripts/demo/demo_benchmark_app.sh b/scripts/demo/demo_benchmark_app.sh
index 8c84e8a724ef4c..25d1f94ae8eca7 100755
--- a/scripts/demo/demo_benchmark_app.sh
+++ b/scripts/demo/demo_benchmark_app.sh
@@ -50,7 +50,7 @@ fi
 
 target_precision="FP16"
 
-printf "target_precision = ${target_precision}\n"
+printf "target_precision = %s\n" ${target_precision}
 
 models_path="$HOME/openvino_models/models"
 models_cache="$HOME/openvino_models/cache"
@@ -71,12 +71,12 @@ else
 fi
 
 if ! . "$setupvars_path" ; then
-    printf "Unable to run ./setupvars.sh. Please check its presence. ${run_again}"
+    printf "Unable to run ./setupvars.sh. Please check its presence. %s" "${run_again}"
     exit 1
 fi
 
 # Step 1. Download the Caffe model and the prototxt of the model
-printf "${dashes}"
+printf "%s" "${dashes}"
 printf "\n\nDownloading the Caffe model and the prototxt"
 
 cur_path=$PWD
@@ -119,7 +119,7 @@ elif [[ $DISTRO == "ubuntu" ]]; then
     python_binary=python3
     pip_binary=pip3
 
-    system_ver=`cat /etc/lsb-release | grep -i "DISTRIB_RELEASE" | cut -d "=" -f2`
+    system_ver=$(grep -i "DISTRIB_RELEASE" -f /etc/lsb-release | cut -d "=" -f2)
     if [ "$system_ver" = "16.04" ]; then
         sudo -E apt-get install -y libpng12-dev
     else
@@ -143,7 +143,7 @@ elif [[ "$OSTYPE" == "darwin"* ]]; then
 fi
 
 if ! command -v $python_binary &>/dev/null; then
-    printf "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. ${run_again}"
+    printf "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. %s" "${run_again}"
     exit 1
 fi
 
@@ -166,14 +166,14 @@ ir_dir="${irs_path}/${model_dir}/${target_precision}"
 
 if [ ! -e "$ir_dir" ]; then
     # Step 2. Configure Model Optimizer
-    printf "${dashes}"
+    printf "%s" "${dashes}"
     printf "Install Model Optimizer dependencies\n\n"
     cd "${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/install_prerequisites"
     . ./install_prerequisites.sh caffe
     cd "$cur_path"
 
     # Step 3. Convert a model with Model Optimizer
-    printf "${dashes}"
+    printf "%s" "${dashes}"
     printf "Convert a model with Model Optimizer\n\n"
 
     mo_path="${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/mo.py"
@@ -181,12 +181,12 @@ if [ ! -e "$ir_dir" ]; then
     export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=cpp
     print_and_run "$python_binary" "$downloader_dir/converter.py" --mo "$mo_path" --name "$model_name" -d "$models_path" -o "$irs_path" --precisions "$target_precision"
 else
-    printf "\n\nTarget folder ${ir_dir} already exists. Skipping IR generation  with Model Optimizer."
-    printf "If you want to convert a model again, remove the entire ${ir_dir} folder. ${run_again}"
+    printf "\n\nTarget folder %s already exists. Skipping IR generation  with Model Optimizer." "${ir_dir}"
+    printf "If you want to convert a model again, remove the entire %s folder. %s" "${ir_dir}" "${run_again}"
 fi
 
 # Step 4. Build samples
-printf "${dashes}"
+printf "%s" "${dashes}"
 printf "Build Inference Engine samples\n\n"
 
 OS_PATH=$(uname -m)
@@ -211,15 +211,15 @@ cmake -DCMAKE_BUILD_TYPE=Release "$samples_path"
 make $NUM_THREADS benchmark_app
 
 # Step 5. Run samples
-printf "${dashes}"
+printf "%s" "${dashes}"
 printf "Run Inference Engine benchmark app\n\n"
 
 cd "$binaries_dir"
 
 cp -f "$ROOT_DIR/${model_name}.labels" "${ir_dir}/"
 
-print_and_run ./benchmark_app -d "$target" -i "$target_image_path" -m "${ir_dir}/${model_name}.xml" -pc ${sampleoptions}
+print_and_run ./benchmark_app -d "$target" -i "$target_image_path" -m "${ir_dir}/${model_name}.xml" -pc "${sampleoptions}"
 
-printf "${dashes}"
+printf "%s" "${dashes}"
 
 printf "Inference Engine benchmark app completed successfully.\n\n"
diff --git a/scripts/demo/demo_security_barrier_camera.sh b/scripts/demo/demo_security_barrier_camera.sh
index 4913aaf1d314ca..eaf6cd6784e094 100755
--- a/scripts/demo/demo_security_barrier_camera.sh
+++ b/scripts/demo/demo_security_barrier_camera.sh
@@ -88,7 +88,7 @@ elif [[ $DISTRO == "ubuntu" ]]; then
     python_binary=python3
     pip_binary=pip3
 
-    system_ver=`cat /etc/lsb-release | grep -i "DISTRIB_RELEASE" | cut -d "=" -f2`
+    system_ver=$(grep -i "DISTRIB_RELEASE" -f /etc/lsb-release | cut -d "=" -f2)
     if [ "$system_ver" = "16.04" ]; then
         sudo -E apt-get install -y libpng12-dev
     else
@@ -112,7 +112,7 @@ elif [[ "$OSTYPE" == "darwin"* ]]; then
 fi
 
 if ! command -v $python_binary &>/dev/null; then
-    printf "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. ${run_again}"
+    printf "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. %s" "${run_again}"
     exit 1
 fi
 
@@ -128,18 +128,18 @@ else
     printf "Error: setupvars.sh is not found\n"
 fi
 if ! . "$setupvars_path" ; then
-    printf "Unable to run ./setupvars.sh. Please check its presence. ${run_again}"
+    printf "Unable to run ./setupvars.sh. Please check its presence. %s" "${run_again}"
     exit 1
 fi
 
 # Step 1. Downloading Intel models
-printf "${dashes}"
+printf "%s" "${dashes}"
 printf "Downloading Intel models\n\n"
 
 
 target_precision="FP16"
 
-printf "target_precision = ${target_precision}\n"
+printf "target_precision = %s\n" "${target_precision}"
 
 downloader_dir="${INTEL_OPENVINO_DIR}/deployment_tools/open_model_zoo/tools/downloader"
 
@@ -161,13 +161,13 @@ while read -r model_opt model_name; do
 done < "$ROOT_DIR/demo_security_barrier_camera.conf"
 
 # Step 2. Build samples
-printf "${dashes}"
+printf "%s" "${dashes}"
 printf "Build Inference Engine demos\n\n"
 
 demos_path="${INTEL_OPENVINO_DIR}/deployment_tools/open_model_zoo/demos"
 
 if ! command -v cmake &>/dev/null; then
-    printf "\n\nCMAKE is not installed. It is required to build Inference Engine demos. Please install it. ${run_again}"
+    printf "\n\nCMAKE is not installed. It is required to build Inference Engine demos. Please install it. %s" "${run_again}"
     exit 1
 fi
 
@@ -189,13 +189,13 @@ cmake -DCMAKE_BUILD_TYPE=Release "$demos_path"
 make $NUM_THREADS security_barrier_camera_demo
 
 # Step 3. Run samples
-printf "${dashes}"
+printf "%s" "${dashes}"
 printf "Run Inference Engine security_barrier_camera demo\n\n"
 
 binaries_dir="${build_dir}/${OS_PATH}/Release"
 cd "$binaries_dir"
 
-print_and_run ./security_barrier_camera_demo -d "$target" -d_va "$target" -d_lpr "$target" -i "$target_image_path" "${model_args[@]}" ${sampleoptions}
+print_and_run ./security_barrier_camera_demo -d "$target" -d_va "$target" -d_lpr "$target" -i "$target_image_path" "${model_args[@]}" "${sampleoptions}"
 
-printf "${dashes}"
+printf "%s" "${dashes}"
 printf "Demo completed successfully.\n\n"
diff --git a/scripts/demo/demo_squeezenet_download_convert_run.sh b/scripts/demo/demo_squeezenet_download_convert_run.sh
index 00daead548e10c..91a9055375b7ef 100755
--- a/scripts/demo/demo_squeezenet_download_convert_run.sh
+++ b/scripts/demo/demo_squeezenet_download_convert_run.sh
@@ -46,7 +46,7 @@ done
 
 target_precision="FP16"
 
-printf "target_precision = ${target_precision}\n"
+printf "target_precision = %s\n" "${target_precision}"
 
 models_path="$HOME/openvino_models/models"
 models_cache="$HOME/openvino_models/cache"
@@ -67,12 +67,12 @@ else
 fi
 
 if ! . "$setupvars_path" ; then
-    printf "Unable to run ./setupvars.sh. Please check its presence. ${run_again}"
+    printf "Unable to run ./setupvars.sh. Please check its presence. %s" "${run_again}"
     exit 1
 fi
 
 # Step 1. Download the Caffe model and the prototxt of the model
-printf "${dashes}"
+printf "%s" "${dashes}"
 printf "\n\nDownloading the Caffe model and the prototxt"
 
 cur_path=$PWD
@@ -115,7 +115,7 @@ elif [[ $DISTRO == "ubuntu" ]]; then
     python_binary=python3
     pip_binary=pip3
 
-    system_ver=`cat /etc/lsb-release | grep -i "DISTRIB_RELEASE" | cut -d "=" -f2`
+    system_ver=$(grep -i "DISTRIB_RELEASE" -f /etc/lsb-release | cut -d "=" -f2)
     if [ "$system_ver" = "16.04" ]; then
         sudo -E apt-get install -y libpng12-dev
     else
@@ -139,7 +139,7 @@ elif [[ "$OSTYPE" == "darwin"* ]]; then
 fi
 
 if ! command -v $python_binary &>/dev/null; then
-    printf "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. ${run_again}"
+    printf "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. %s" "${run_again}"
     exit 1
 fi
 
@@ -162,14 +162,14 @@ ir_dir="${irs_path}/${model_dir}/${target_precision}"
 
 if [ ! -e "$ir_dir" ]; then
     # Step 2. Configure Model Optimizer
-    printf "${dashes}"
+    printf "%s" "${dashes}"
     printf "Install Model Optimizer dependencies\n\n"
     cd "${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/install_prerequisites"
     . ./install_prerequisites.sh caffe
     cd "$cur_path"
 
     # Step 3. Convert a model with Model Optimizer
-    printf "${dashes}"
+    printf "%s" "${dashes}"
     printf "Convert a model with Model Optimizer\n\n"
 
     mo_path="${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/mo.py"
@@ -177,12 +177,12 @@ if [ ! -e "$ir_dir" ]; then
     export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=cpp
     print_and_run "$python_binary" "$downloader_dir/converter.py" --mo "$mo_path" --name "$model_name" -d "$models_path" -o "$irs_path" --precisions "$target_precision"
 else
-    printf "\n\nTarget folder ${ir_dir} already exists. Skipping IR generation  with Model Optimizer."
-    printf "If you want to convert a model again, remove the entire ${ir_dir} folder. ${run_again}"
+    printf "\n\nTarget folder %s already exists. Skipping IR generation  with Model Optimizer." "${ir_dir}"
+    printf "If you want to convert a model again, remove the entire %s folder. %s" "${ir_dir}" "${run_again}"
 fi
 
 # Step 4. Build samples
-printf "${dashes}"
+printf "%s" "${dashes}"
 printf "Build Inference Engine samples\n\n"
 
 OS_PATH=$(uname -m)
@@ -207,15 +207,14 @@ cmake -DCMAKE_BUILD_TYPE=Release "$samples_path"
 make $NUM_THREADS classification_sample_async
 
 # Step 5. Run samples
-printf "${dashes}"
+printf "%s" "${dashes}"
 printf "Run Inference Engine classification sample\n\n"
 
 cd "$binaries_dir"
 
 cp -f "$ROOT_DIR/${model_name}.labels" "${ir_dir}/"
 
-print_and_run ./classification_sample_async -d "$target" -i "$target_image_path" -m "${ir_dir}/${model_name}.xml" ${sampleoptions}
-
-printf "${dashes}"
+print_and_run ./classification_sample_async -d "$target" -i "$target_image_path" -m "${ir_dir}/${model_name}.xml" "${sampleoptions}"
 
+printf "%s" "${dashes}"
 printf "Demo completed successfully.\n\n"
diff --git a/scripts/demo/utils.sh b/scripts/demo/utils.sh
index 552ba20aefa99f..b7349e0e3a8492 100755
--- a/scripts/demo/utils.sh
+++ b/scripts/demo/utils.sh
@@ -1,3 +1,5 @@
+#!/usr/bin/env bash
+
 # Copyright (C) 2018-2021 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 
diff --git a/scripts/install_dependencies/install_4_14_kernel.sh b/scripts/install_dependencies/install_4_14_kernel.sh
index a7e01309f19505..bcef687d7875b1 100755
--- a/scripts/install_dependencies/install_4_14_kernel.sh
+++ b/scripts/install_dependencies/install_4_14_kernel.sh
@@ -6,28 +6,28 @@
 # This script installs Linux kernel 4.14 required for Intel NEO OpenCL driver on Ubuntu and CentOS
 
 if grep -i "rhel" /etc/os-release &>/dev/null; then
-	# Cent OS
-	echo "install kernel build dependencies"
-	sudo -E yum install -y git gcc gcc-c++ ncurses-devel openssl-devel bc xz elfutils-libelf-devel xorg-x11-drv-nouveau rpm-build
+    # Cent OS
+    echo "install kernel build dependencies"
+    sudo -E yum install -y git gcc gcc-c++ ncurses-devel openssl-devel bc xz elfutils-libelf-devel xorg-x11-drv-nouveau rpm-build
 
-	echo "download 4.14.20 kernel"
-	if [ ! -f ./linux-4.14.20.tar.xz ]; then
-		wget https://www.kernel.org/pub/linux/kernel/v4.x/linux-4.14.20.tar.xz
-	fi
+    echo "download 4.14.20 kernel"
+    if [ ! -f ./linux-4.14.20.tar.xz ]; then
+        wget https://www.kernel.org/pub/linux/kernel/v4.x/linux-4.14.20.tar.xz
+    fi
 
-	tar -xJf linux-4.14.20.tar.xz
-	cd linux-4.14.20
-	echo "build 4.14.20 kernel"
-	make olddefconfig
+    tar -xJf linux-4.14.20.tar.xz
+    cd linux-4.14.20
+    echo "build 4.14.20 kernel"
+    make olddefconfig
 
-	make -j 8 binrpm-pkg
-	cd ~/rpmbuild/RPMS/x86_64
-	sudo -E yum -y localinstall *.rpm
-	sudo -E grub2-set-default 0
+    make -j 8 binrpm-pkg
+    cd ~/rpmbuild/RPMS/x86_64
+    sudo -E yum -y localinstall ./*.rpm
+    sudo -E grub2-set-default 0
 
 elif grep -i "ubuntu" /etc/os-release &>/dev/null; then
-	# Ubuntu
-	sudo -E add-apt-repository ppa:teejee2008/ppa
-	sudo -E apt-get update && sudo apt-get install -y ukuu
-	sudo -E ukuu --install v4.14.20
+    # Ubuntu
+    sudo -E add-apt-repository ppa:teejee2008/ppa
+    sudo -E apt-get update && sudo apt-get install -y ukuu
+    sudo -E ukuu --install v4.14.20
 fi
\ No newline at end of file
diff --git a/scripts/install_dependencies/install_NCS_udev_rules.sh b/scripts/install_dependencies/install_NCS_udev_rules.sh
index 3666b4502e9a44..7062e9d753db92 100755
--- a/scripts/install_dependencies/install_NCS_udev_rules.sh
+++ b/scripts/install_dependencies/install_NCS_udev_rules.sh
@@ -22,5 +22,3 @@ else
     echo "File '97-myriad-usbboot.rules' is missing. Please make sure you installed 'Inference Engine Runtime for Intel® Movidius™ VPU'."
     exit -1
 fi 
-
-

From 58eef532e95bfc41149cb3b47cdb2b2f4fa66f2a Mon Sep 17 00:00:00 2001
From: Bartosz Lesniewski <bartosz.lesniewski@intel.com>
Date: Tue, 15 Jun 2021 15:54:00 +0200
Subject: [PATCH 090/178] Revise ExtractImagePatches op - specification (#6105)

* Change example input matrices to tex

* Remove leftover matrix template

* Remove reference to TF operation
---
 docs/ops/movement/ExtractImagePatches_3.md | 165 +++++++++++----------
 1 file changed, 89 insertions(+), 76 deletions(-)

diff --git a/docs/ops/movement/ExtractImagePatches_3.md b/docs/ops/movement/ExtractImagePatches_3.md
index 3604d3b49ca19d..5046854ee22269 100644
--- a/docs/ops/movement/ExtractImagePatches_3.md
+++ b/docs/ops/movement/ExtractImagePatches_3.md
@@ -8,9 +8,7 @@
 
 **Detailed description**:
 
-The *ExtractImagePatches* operation is similar to the TensorFlow* operation [ExtractImagePatches](https://www.tensorflow.org/api_docs/python/tf/image/extract_patches).
-
-This op extracts patches of shape `sizes` which are `strides` apart in the input image. The output elements are taken from the input at intervals given by the `rate` argument, as in dilated convolutions.
+The *ExtractImagePatches* operation extracts patches of shape `sizes` which are `strides` apart in the input image. The output elements are taken from the input at intervals given by the `rate` argument, as in dilated convolutions.
 
 The result is a 4D tensor containing image patches with size `size[0] * size[1] * depth` vectorized in the "depth" dimension.
 
@@ -92,20 +90,23 @@ The "auto_pad" attribute has no effect on the size of each patch, it determines
 Image is a `1 x 1 x 10 x 10` array that contains the numbers 1 through 100. We use the symbol `x` to mark output patches.
 
 1. `sizes="3,3", strides="5,5", rates="1,1", auto_pad="valid"`
+\f[
+    \begin{bmatrix}
+        x & x & x & 4 & 5 & x & x & x & 9 & 10 \\
+        x & x & x & 14 & 15 & x & x & x & 19 & 20 \\
+        x & x & x & 24 & 25 & x & x & x & 29 & 30 \\
+        31 & 32 & 33 & 34 & 35 & 36 & 37 & 38 & 39 & 40 \\
+        41 & 42 & 43 & 44 & 45 & 46 & 47 & 48 & 49 & 50 \\
+        x & x & x & 54 & 55 & x & x & x & 59 & 60 \\
+        x & x & x & 64 & 65 & x & x & x & 69 & 70 \\
+        x & x & x & 74 & 75 & x & x & x & 79 & 80 \\
+        81 & 82 & 83 & 84 & 85 & 86 & 87 & 88 & 89 & 90 \\
+        91 & 92 & 93 & 94 & 95 & 96 & 79 & 98 & 99 & 100
+    \end{bmatrix}
+\f]
 
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp; 4 &nbsp; 5 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp; 9 10    
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;14 15 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x 19 20  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;24 25 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x 29 30  
-  31 32 33 34 35 36 37 38 39 40  
-  41 42 43 44 45 46 47 48 49 50  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;54 55 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x 59 60  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;64 65 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x 69 70  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;74 75 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x 79 80  
-  81 82 83 84 85 86 87 88 89 90  
-  91 92 93 94 95 96 97 98 99 100  
-  
   output:  
-
+```
     [[[[ 1  6]
        [51 56]]
 
@@ -132,25 +133,28 @@ Image is a `1 x 1 x 10 x 10` array that contains the numbers 1 through 100. We u
 
       [[23 28]
        [73 78]]]] 
-
+```
   output shape: `[1, 9, 2, 2]`
   
 2. `sizes="4,4", strides="8,8", rates="1,1", auto_pad="valid"`
-
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;&nbsp;5 &nbsp;&nbsp;6 &nbsp;&nbsp;7 &nbsp;&nbsp;8 &nbsp;&nbsp;9 10  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;15 16 17 18 19 20  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;25 26 27 28 29 30  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;35 36 37 38 39 40  
-  41 42 43 44 45 46 47 48 49 50  
-  51 52 53 54 55 56 57 58 59 60  
-  61 62 63 64 65 66 67 68 69 70  
-  71 72 73 74 75 76 77 78 79 80  
-  81 82 83 84 85 86 87 88 89 90  
-  91 92 93 94 95 96 97 98 99 100  
+\f[
+    \begin{bmatrix}
+        x & x & x & x & 5 & 6 & 7 & 8 & 9 & 10 \\
+        x & x & x & x & 15 & 16 & 17 & 18 & 19 & 20 \\
+        x & x & x & x & 25 & 26 & 27 & 28 & 29 & 30 \\
+        x & x & x & x & 35 & 36 & 37 & 38 & 39 & 40 \\
+        41 & 42 & 43 & 44 & 45 & 46 & 47 & 48 & 49 & 50 \\
+        51 & 52 & 53 & 54 & 55 & 56 & 57 & 58 & 59 & 60 \\
+        61 & 62 & 63 & 64 & 65 & 66 & 67 & 68 & 69 & 70 \\
+        71 & 72 & 73 & 74 & 75 & 76 & 77 & 78 & 79 & 80 \\
+        81 & 82 & 83 & 84 & 85 & 86 & 87 & 88 & 89 & 90 \\
+        91 & 92 & 93 & 94 & 95 & 96 & 79 & 98 & 99 & 100
+    \end{bmatrix}
+\f]
   
   output:  
-
-    [[[[ 1]]
+``` 
+     [[[[ 1]]
 
       [[ 2]]
 
@@ -180,28 +184,30 @@ Image is a `1 x 1 x 10 x 10` array that contains the numbers 1 through 100. We u
 
       [[33]]
 
-      [[34]]]] 
-      
+      [[34]]]]
+```
   output shape: `[1, 16, 1, 1]`
 
 3. `sizes="4,4", strides="9,9", rates="1,1", auto_pad="same_upper"`
-
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;&nbsp;4 &nbsp;&nbsp;5 &nbsp;&nbsp;6 &nbsp;&nbsp;7 &nbsp;&nbsp;8 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;14 15 16 17 18 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;24 25 26 27 28 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;0 31 32 33 34 35 36 37 38 39 40 &nbsp;&nbsp;0 &nbsp;&nbsp;0  
-   &nbsp;&nbsp;0 41 42 43 44 45 46 47 48 49 50 &nbsp;&nbsp;0 &nbsp;&nbsp;0  
-   &nbsp;&nbsp;0 51 52 53 54 55 56 57 58 59 60 &nbsp;&nbsp;0 &nbsp;&nbsp;0  
-   &nbsp;&nbsp;0 61 62 63 64 65 66 67 68 69 70 &nbsp;&nbsp;0 &nbsp;&nbsp;0  
-   &nbsp;&nbsp;0 71 72 73 74 75 76 77 78 79 80 &nbsp;&nbsp;0 &nbsp;&nbsp;0  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;84 85 86 87 88 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;94 95 96 97 98 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;0 &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;x    
-
+\f[
+    \begin{bmatrix}
+        x & x & x & x & 0 & 0 & 0 & 0 & 0 & x & x & x & x\\
+        x & x & x & x & 4 & 5 & 6 & 7 & 8 & x & x & x & x\\
+        x & x & x & x & 14 & 15 & 16 & 17 & 18 & x & x & x & x\\
+        x & x & x & x & 24 & 25 & 26 & 27 & 28 & x & x & x & x\\
+        0 & 31 & 32 & 33 & 34 & 35 & 36 & 37 & 38 & 39 & 40 & 0 & 0\\
+        0 & 41 & 42 & 43 & 44 & 45 & 46 & 47 & 48 & 49 & 50 & 0 & 0\\
+        0 & 51 & 52 & 53 & 54 & 55 & 56 & 57 & 58 & 59 & 60 & 0 & 0\\
+        0 & 61 & 62 & 63 & 64 & 65 & 66 & 67 & 68 & 69 & 70 & 0 & 0\\
+        0 & 71 & 72 & 73 & 74 & 75 & 76 & 77 & 78 & 79 & 80 & 0 & 0\\
+        x & x & x & x & 84 & 85 & 86 & 87 & 88 & x & x & x & x\\
+        x & x & x & x & 94 & 95 & 96 & 79 & 98 & x & x & x & x\\
+        x & x & x & x & 0 & 0 & 0 & 0 & 0 & x & x & x & x\\
+        x & x & x & x & 0 & 0 & 0 & 0 & 0 & x & x & x & x
+    \end{bmatrix}
+\f]
   output:  
-
+```
     [[[[  0   0]
        [  0  89]]
 
@@ -249,25 +255,28 @@ Image is a `1 x 1 x 10 x 10` array that contains the numbers 1 through 100. We u
 
       [[ 23   0]
        [  0   0]]]] 
-
+```
   output shape: `[1, 16, 2, 2]`
 
 4. `sizes="3,3", strides="5,5", rates="2,2", auto_pad="valid"`  
 This time we use the symbols `x`, `y`, `z` and `k` to distinguish the patches:
-  
-   &nbsp;&nbsp;x &nbsp;&nbsp;2 &nbsp;&nbsp;x &nbsp;&nbsp;4 &nbsp;&nbsp;x &nbsp;&nbsp;y &nbsp;&nbsp;7 &nbsp;&nbsp;y &nbsp;&nbsp;9 &nbsp;&nbsp;y  
-  11 12 13 14 15 16 17 18 19 20  
-   &nbsp;&nbsp;x &nbsp;22 &nbsp;&nbsp;x 24 &nbsp;&nbsp;x &nbsp;&nbsp;y 27 &nbsp;&nbsp;y 29 &nbsp;&nbsp;y  
-  31 32 33 34 35 36 37 38 39 40  
-   &nbsp;&nbsp;x &nbsp;42 &nbsp;&nbsp;x 44 &nbsp;&nbsp;x &nbsp;&nbsp;y 47 &nbsp;&nbsp;y 49 &nbsp;&nbsp;y  
-   &nbsp;&nbsp;z &nbsp;52 &nbsp;&nbsp;z 54 &nbsp;&nbsp;z &nbsp;&nbsp;k 57 &nbsp;&nbsp;k 59 &nbsp;&nbsp;k  
-  61 62 63 64 65 66 67 68 69 70  
-   &nbsp;&nbsp;z &nbsp;72 &nbsp;&nbsp;z 74 &nbsp;&nbsp;z &nbsp;&nbsp;k 77 &nbsp;&nbsp;k 79 &nbsp;&nbsp;k  
-  81 82 83 84 85 86 87 88 89 90  
-   &nbsp;&nbsp;z &nbsp;92 &nbsp;&nbsp;z 94 &nbsp;&nbsp;z &nbsp;&nbsp;k 97 &nbsp;&nbsp;k 99 &nbsp;&nbsp;k
-   
-  output:
+\f[
+    \begin{bmatrix}
+        x & 2 & x & 4 & x & y & 7 & y & 9 & y \\
+        11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 & 20 \\
+        x & 22 & x & 24 & x & y & 27 & y & 29 & y \\
+        31 & 32 & 33 & 34 & 35 & 36 & 37 & 38 & 39 & 40 \\
+        x & 42 & x & 44 & x & y & 47 & y & 49 & y \\
+        z & 52 & z & 54 & z & k & 57 & k & 59 & k \\
+        61 & 62 & 63 & 64 & 65 & 66 & 67 & 68 & 69 & 70 \\
+        z & 72 & z & 74 & z & k & 77 & k & 79 & k \\
+        81 & 82 & 83 & 84 & 85 & 86 & 87 & 88 & 89 & 90 \\
+        z & 92 & z & 94 & z & k & 79 & k & 99 & k
+    \end{bmatrix}
+\f]
 
+  output:
+```
     [[[[  1   6]
        [ 51  56]]
 
@@ -294,26 +303,30 @@ This time we use the symbols `x`, `y`, `z` and `k` to distinguish the patches:
 
       [[ 45  50]
        [ 95 100]]]] 
-
+```
   output_shape: `[1, 9, 2, 2]`
 
 5. `sizes="2,2", strides="3,3", rates="1,1", auto_pad="valid"`  
 Image is a `1 x 2 x 5 x 5` array that contains two feature maps where feature map with coordinate 0 contains numbers in a range `[1, 25]` and feature map with coordinate 1 contains numbers in a range `[26, 50]`  
 
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;&nbsp;3 &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;6 &nbsp;&nbsp;7 &nbsp;&nbsp;8 &nbsp;&nbsp;x &nbsp;&nbsp;x  
-  11 12 13 14 15  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;18 &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;23 &nbsp;&nbsp;x &nbsp;&nbsp;x  
-  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;28 &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;33 &nbsp;&nbsp;x &nbsp;&nbsp;x  
-  36 37 38 39 40  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;43 &nbsp;&nbsp;x &nbsp;&nbsp;x  
-   &nbsp;&nbsp;x &nbsp;&nbsp;x &nbsp;48 &nbsp;&nbsp;x &nbsp;&nbsp;x  
-
+\f[
+    \begin{bmatrix}
+        x & x & 3 & x & x\\
+        x & x & 8 & x & x\\
+        11 & 12 & 13 & 14 & 15\\
+        x & x & 18 & x & x\\
+        x & x & 23 & x & x
+    \end{bmatrix}\\
+    \begin{bmatrix}
+        x & x & 28 & x & x\\
+        x & x & 33 & x & x\\
+        36 & 37 & 38 & 39 & 40\\
+        x & x & 43 & x & x\\
+        x & x & 48 & x & x
+    \end{bmatrix}
+\f]
   output:
-
+```
     [[[[ 1  4]
        [16 19]]
 
@@ -337,5 +350,5 @@ Image is a `1 x 2 x 5 x 5` array that contains two feature maps where feature ma
 
       [[32 35]
        [47 50]]]] 
-   
+``` 
   output shape: `[1, 8, 2, 2]`

From a16af0d2ea550c955a34659510831577e43d0b43 Mon Sep 17 00:00:00 2001
From: Edward Shogulin <edward.shogulin@intel.com>
Date: Tue, 15 Jun 2021 16:59:10 +0300
Subject: [PATCH 091/178] [LPT] FakeQuantize folding fix to support
 ConvolutionBackpropData with FQ on weights (#6160)

---
 .../src/network_helper.cpp                    |  2 +-
 ...nvolution_backprop_data_transformation.cpp | 14 +++++++----
 ...nvolution_backprop_data_transformation.cpp | 23 +++++++++++++++----
 ...nvolution_backprop_data_transformation.hpp |  2 +-
 ...nvolution_backprop_data_transformation.cpp | 18 +++++++++------
 .../convolution_backprop_data_function.cpp    |  1 +
 6 files changed, 42 insertions(+), 18 deletions(-)

diff --git a/inference-engine/src/low_precision_transformations/src/network_helper.cpp b/inference-engine/src/low_precision_transformations/src/network_helper.cpp
index 4a1e942e5753ba..90995bc0ffb22d 100644
--- a/inference-engine/src/low_precision_transformations/src/network_helper.cpp
+++ b/inference-engine/src/low_precision_transformations/src/network_helper.cpp
@@ -683,7 +683,7 @@ std::shared_ptr<Node> NetworkHelper::foldFakeQuantize(
         auto levels_1 = fq->get_levels() - 1.f;
 
         const size_t DHW = D * H * W;
-        const size_t IDHW = IC * D * H * W;
+        const size_t IDHW = outChannelsShapeIndex == 0 ? IC * D * H * W : OC * D * H * W;
 
         const auto values = constant->cast_vector<float>();
         std::vector<float> quantizedValues(OC * IC * D * H * W);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
index 44d26d231b617a..a6e2e93a37ebb8 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
@@ -15,11 +15,17 @@ const std::vector<ngraph::element::Type> netPrecisions = {
 };
 
 const std::vector<ngraph::pass::low_precision::LayerTransformation::Params> trasformationParamValues = {
-    LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParams().setUpdatePrecisions(true),
-    LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParams().setUpdatePrecisions(false)
+    LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParams().setUpdatePrecisions(true)
 };
 
 const std::vector<LayerTestsDefinitions::ConvolutionBackpropDataTransformationParam> params = {
+    // FQ on weights
+    {
+        {256ul, ngraph::Shape{1, 1, 1, 1}, { 0.f }, { 25.5f }, { 0.f }, { 25.5f }},
+        {255ul, ngraph::Shape{1, 1, 1, 1}, { -12.7f }, { 12.7f }, { -12.7f }, { 12.7f }},
+        "convolutionBackpropData_original",
+        "U8"
+    },
     // FQ on weights
     // with zero point
     {
@@ -87,8 +93,8 @@ const std::vector<LayerTestsDefinitions::ConvolutionBackpropDataTransformationPa
     }
 };
 
-const std::vector<ngraph::Shape> inputShapes = {
-    { 1, 8, 16, 16 }
+const std::vector<std::pair<ngraph::Shape, bool>> inputShapes = {
+    {{ 1, 8, 16, 16 }, true}
 };
 
 const std::vector<ngraph::Shape> outputShapes = {
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
index d33e3c42f9e242..2cf625b58628a6 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
@@ -16,11 +16,24 @@ const std::vector<ngraph::element::Type> netPrecisions = {
 };
 
 const std::vector<ngraph::pass::low_precision::LayerTransformation::Params> trasformationParamValues = {
-        LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParams().setUpdatePrecisions(true),
-        LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParams().setUpdatePrecisions(false)
+        LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParams().setUpdatePrecisions(true)
 };
 
 const std::vector<LayerTestsDefinitions::ConvolutionBackpropDataTransformationParam> params = {
+    // FQ on weights
+    {
+        {256ul, ngraph::Shape{1, 1, 1, 1}, { 0.f }, { 25.5f }, { 0.f }, { 25.5f }},
+        {255ul, ngraph::Shape{1, 1, 1, 1}, { -12.7f }, { 12.7f }, { -12.7f }, { 12.7f }},
+        "convolutionBackpropData_original",
+        "U8"
+    },
+    // FQ on weights
+    {
+        {256ul, ngraph::Shape{1, 1, 1, 1}, { -12.8f }, { 12.7f }, { -12.8f }, { 12.7f }},
+        {255ul, ngraph::Shape{1, 1, 1, 1}, { -12.7f }, { 12.7f }, { -12.7f }, { 12.7f }},
+        "convolutionBackpropData_original",
+        "I8"
+    },
     // FQ on weights
     // with zero point
     {
@@ -82,9 +95,9 @@ const std::vector<LayerTestsDefinitions::ConvolutionBackpropDataTransformationPa
     }
 };
 
-const std::vector<ngraph::Shape> inputShapes = {
-        { 1, 8, 16, 16 },
-        { 1, 32, 16, 16 }
+const std::vector<std::pair<ngraph::Shape, bool>> inputShapes = {
+        {{ 1, 8, 16, 16 }, false},
+        {{ 1, 32, 16, 16 }, true}
 };
 
 const std::vector<ngraph::Shape> outputShapes = {
diff --git a/inference-engine/tests/functional/plugin/shared/include/low_precision_transformations/convolution_backprop_data_transformation.hpp b/inference-engine/tests/functional/plugin/shared/include/low_precision_transformations/convolution_backprop_data_transformation.hpp
index 39d5ea583916e5..8b690e2df2eee3 100644
--- a/inference-engine/tests/functional/plugin/shared/include/low_precision_transformations/convolution_backprop_data_transformation.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/low_precision_transformations/convolution_backprop_data_transformation.hpp
@@ -43,7 +43,7 @@ class ConvolutionBackpropDataTransformationParam {
 
 typedef std::tuple<
     ngraph::element::Type, // netPrecision
-    ngraph::Shape,         // inputShape
+    std::pair<ngraph::Shape, bool>, // input shape and shape support flag
     ngraph::Shape,         // outputShape
     std::string,           // targetDevice
     ngraph::pass::low_precision::LayerTransformation::Params,
diff --git a/inference-engine/tests/functional/plugin/shared/src/low_precision_transformations/convolution_backprop_data_transformation.cpp b/inference-engine/tests/functional/plugin/shared/src/low_precision_transformations/convolution_backprop_data_transformation.cpp
index 951af4fdd4e0e0..c8d15a5702e407 100644
--- a/inference-engine/tests/functional/plugin/shared/src/low_precision_transformations/convolution_backprop_data_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/shared/src/low_precision_transformations/convolution_backprop_data_transformation.cpp
@@ -14,7 +14,7 @@ namespace LayerTestsDefinitions {
 
 std::string ConvolutionBackpropDataTransformation::getTestCaseName(testing::TestParamInfo<ConvolutionBackpropDataTransformationParams> obj) {
     ngraph::element::Type netPrecision;
-    ngraph::Shape inputShape;
+    std::pair<ngraph::Shape, bool> inputShape;
     ngraph::Shape outputShape;
     std::string targetDevice;
     ngraph::pass::low_precision::LayerTransformation::Params params;
@@ -22,7 +22,7 @@ std::string ConvolutionBackpropDataTransformation::getTestCaseName(testing::Test
     std::tie(netPrecision, inputShape, outputShape, targetDevice, params, param) = obj.param;
 
     std::ostringstream result;
-    result << getTestCaseNameByParams(netPrecision, inputShape, targetDevice, params) << "_" <<
+    result << getTestCaseNameByParams(netPrecision, inputShape.first, targetDevice, params) << "_" <<
         outputShape << "_" <<
         param.fakeQuantizeOnData << "_" <<
         param.fakeQuantizeOnWeights << "_" <<
@@ -34,14 +34,15 @@ void ConvolutionBackpropDataTransformation::SetUp() {
     threshold = 0.1f;
 
     ngraph::element::Type netPrecision;
-    ngraph::Shape inputShape;
+    std::pair<ngraph::Shape, bool> inputShapeAndHandling;
     ngraph::Shape outputShape;
     ngraph::pass::low_precision::LayerTransformation::Params params;
     ConvolutionBackpropDataTransformationParam param;
-    std::tie(netPrecision, inputShape, outputShape, targetDevice, params, param) = this->GetParam();
+    std::tie(netPrecision, inputShapeAndHandling, outputShape, targetDevice, params, param) = this->GetParam();
 
     std::shared_ptr<ngraph::Node> weights;
 
+    const auto inputShape = inputShapeAndHandling.first;
     if (!param.fakeQuantizeOnWeights.empty()) {
         weights = ngraph::builder::subgraph::ConvolutionBackpropDataFunction::getWeights(
             ngraph::Shape{inputShape[1], inputShape[1] / 2, 1, 1},
@@ -65,9 +66,12 @@ void ConvolutionBackpropDataTransformation::SetUp() {
 void ConvolutionBackpropDataTransformation::Run() {
     LayerTestsCommon::Run();
 
-    const auto params = std::get<5>(GetParam());
-    const auto actualType = getRuntimePrecision(params.layerName);
-    EXPECT_EQ(actualType, params.expectedKernelType);
+    const auto inputShape = std::get<1>(GetParam());
+    if (inputShape.second) {
+        const auto params = std::get<5>(GetParam());
+        const auto actualType = getRuntimePrecision(params.layerName);
+        EXPECT_EQ(actualType, params.expectedKernelType);
+    }
 }
 
 TEST_P(ConvolutionBackpropDataTransformation, CompareWithRefImpl) {
diff --git a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/convolution_backprop_data_function.cpp b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/convolution_backprop_data_function.cpp
index ae7d3847f69866..969e1394a0fb02 100644
--- a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/convolution_backprop_data_function.cpp
+++ b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/convolution_backprop_data_function.cpp
@@ -38,6 +38,7 @@ std::shared_ptr<Function> ConvolutionBackpropDataFunction::get(
         CoordinateDiff{ 0, 0 },
         CoordinateDiff{ 0, 0 },
         Strides{ 1, 1 });
+    convolutionBackpropData->set_friendly_name("convolutionBackpropData");
 
     ngraph::ResultVector results{ std::make_shared<opset1::Result>(convolutionBackpropData) };
     return std::make_shared<ngraph::Function>(results, ParameterVector{ input }, "ConvolutionBackpropDataTransformation");

From 0981a1584677d447e6997a7d86694e599431ddff Mon Sep 17 00:00:00 2001
From: Mikhail Nosov <mikhail.nosov@intel.com>
Date: Tue, 15 Jun 2021 22:40:46 +0300
Subject: [PATCH 092/178] [Caching] Klocwork fixes (#6157)

---
 inference-engine/src/inference_engine/compilation_context.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/inference-engine/src/inference_engine/compilation_context.cpp b/inference-engine/src/inference_engine/compilation_context.cpp
index ae6aa698eafa48..1b656200996657 100644
--- a/inference-engine/src/inference_engine/compilation_context.cpp
+++ b/inference-engine/src/inference_engine/compilation_context.cpp
@@ -42,7 +42,7 @@ static int32_t as_int32_t(T v) {
 }
 
 class OstreamHashWrapper final: public std::streambuf {
-    std::size_t    m_res = {};
+    std::size_t m_res = 0;
 public:
     std::size_t getResult() const { return m_res; }
     std::streamsize xsputn(const char* s, std::streamsize n) override {
@@ -65,7 +65,7 @@ class OstreamHashWrapper final: public std::streambuf {
 //////////////////////////////////////////////////
 
 std::string NetworkCompilationContext::calculateFileInfo(const std::string& filePath) {
-    size_t seed {};
+    size_t seed = 0;
     auto absPath = filePath;
     try {
         absPath = FileUtils::absoluteFilePath(filePath);

From 613bb981ce67f8cbbdda816aa12500dd37243726 Mon Sep 17 00:00:00 2001
From: Szymon Durawa <szymon.durawa@intel.com>
Date: Tue, 15 Jun 2021 23:44:34 +0200
Subject: [PATCH 093/178] external_port_id is calculated based on number of op
 inputs. (#6037)

* External_port_id is calcultaed based on number of op inputs.

* Add test for external_port_id serialization.

* Restore data section appearance in xml file.
---
 .../src/transformations/serialize.cpp         | 38 +++++++++++--------
 .../ir_serialization/tensor_iterator.cpp      | 38 +++++++++++++++++++
 2 files changed, 60 insertions(+), 16 deletions(-)

diff --git a/inference-engine/src/transformations/src/transformations/serialize.cpp b/inference-engine/src/transformations/src/transformations/serialize.cpp
index 5f5c8144f2098a..73b72313e5f48a 100644
--- a/inference-engine/src/transformations/src/transformations/serialize.cpp
+++ b/inference-engine/src/transformations/src/transformations/serialize.cpp
@@ -235,7 +235,7 @@ class XmlSerializer : public ngraph::AttributeVisitor {
 
     void output_descriptions_on_adapter(const std::vector<std::shared_ptr<
                                         ngraph::op::util::SubGraphOp::OutputDescription>>& output_descriptions,
-                                        const std::vector<std::string>& parameter_mapping,
+                                        const uint32_t& input_count,
                                         const std::vector<std::string>& result_mapping,
                                         pugi::xml_node& port_map) {
         NGRAPH_CHECK(!result_mapping.empty(), "No results found in body Function.");
@@ -246,7 +246,7 @@ class XmlSerializer : public ngraph::AttributeVisitor {
 
         for (const auto& output_description : output_descriptions) {
             pugi::xml_node output = port_map.append_child("output");
-            output.append_attribute("external_port_id").set_value(parameter_mapping.size() + output_description->m_output_index);
+            output.append_attribute("external_port_id").set_value(input_count + output_description->m_output_index);
             output.append_attribute("internal_layer_id").set_value(result_mapping[output_description->m_body_value_index].c_str());
 
             if (auto concat_output = as_type_ptr<ngraph::op::util::SubGraphOp::ConcatOutputDescription>(output_description)) {
@@ -306,7 +306,11 @@ class XmlSerializer : public ngraph::AttributeVisitor {
                 input_descriptions_on_adapter(a->get(), parameter_mapping, result_mapping, port_map);
             } else if (const auto& a = ngraph::as_type<ngraph::AttributeAdapter<std::vector<std::shared_ptr
                                 <ngraph::op::util::SubGraphOp::OutputDescription>>>>(&adapter)) {
-                output_descriptions_on_adapter(a->get(), parameter_mapping, result_mapping, port_map);
+                uint32_t op_input_count = 0;
+                for (auto c = m_xml_node.parent().child("input").first_child(); !c.empty(); c = c.next_sibling()) {
+                    op_input_count++;
+                }
+                output_descriptions_on_adapter(a->get(), op_input_count, result_mapping, port_map);
             } else if (const auto& a = ngraph::as_type<ngraph::AttributeAdapter<ngraph::op::v5::Loop::SpecialBodyPorts>>(&adapter)) {
                 special_body_ports_on_adapter(a->get(), parameter_mapping, result_mapping, port_map);
             }
@@ -700,19 +704,6 @@ void ngfunction_2_irv10(pugi::xml_node& netXml,
 
         // <layers/data> general attributes
         pugi::xml_node data = layer.append_child("data");
-        XmlSerializer visitor(data, node_type_name, custom_opsets, constant_node_write_handler);
-        NGRAPH_CHECK(node->visit_attributes(visitor), "Visitor API is not supported in ", node);
-        rt_info::XmlSerializer{data}.serialize(node->get_rt_info());
-
-        if (exec_graph) {
-            visit_exec_graph_node(layer, node);
-        }
-
-        const bool data_attr_size =
-            data.attributes().begin() == data.attributes().end();
-        if (data_attr_size) {
-            layer.remove_child(data);
-        }
 
         int port_id = 0;
         // <layers/input>
@@ -780,6 +771,21 @@ void ngfunction_2_irv10(pugi::xml_node& netXml,
                 layer.insert_move_after(output, layer.first_child());
             }
         }
+
+        // fill <data> general attributes
+        XmlSerializer visitor(data, node_type_name, custom_opsets, constant_node_write_handler);
+        NGRAPH_CHECK(node->visit_attributes(visitor), "Visitor API is not supported in ", node);
+        rt_info::XmlSerializer{data}.serialize(node->get_rt_info());
+
+        if (exec_graph) {
+            visit_exec_graph_node(layer, node);
+        }
+
+        const bool data_attr_size =
+            data.attributes().begin() == data.attributes().end();
+        if (data_attr_size) {
+            layer.remove_child(data);
+        }
     }
     // <edges>
     const std::vector<Edge> edge_mapping = create_edge_mapping(layer_ids, f);
diff --git a/inference-engine/tests/functional/inference_engine/ir_serialization/tensor_iterator.cpp b/inference-engine/tests/functional/inference_engine/ir_serialization/tensor_iterator.cpp
index c63371baf89925..cccad0f70a5766 100644
--- a/inference-engine/tests/functional/inference_engine/ir_serialization/tensor_iterator.cpp
+++ b/inference-engine/tests/functional/inference_engine/ir_serialization/tensor_iterator.cpp
@@ -9,6 +9,7 @@
 #include "ie_core.hpp"
 #include "ie_blob.h"
 #include "common_test_utils/data_utils.hpp"
+#include "pugixml.hpp"
 
 #ifndef IR_SERIALIZATION_MODELS_PATH  // should be already defined by cmake
 #define IR_SERIALIZATION_MODELS_PATH ""
@@ -84,3 +85,40 @@ TEST_F(SerializationTensorIteratorTest, TiNegativeStride) {
 
     serialize_and_compare(model_path, weights);
 }
+
+TEST_F(SerializationTensorIteratorTest, SerializationExternalPortIdInXmlFile) {
+    const std::string model_path = IR_SERIALIZATION_MODELS_PATH "loop_2d_add.xml";
+    const std::string binary_path = IR_SERIALIZATION_MODELS_PATH "loop_2d_add.bin";
+
+    InferenceEngine::Core ie;
+    InferenceEngine::CNNNetwork expected;
+    pugi::xml_document loop_orig;
+    pugi::xml_document loop_serialized;
+
+    expected = ie.ReadNetwork(model_path, binary_path);
+    expected.serialize(m_out_xml_path, m_out_bin_path);
+
+    pugi::xml_parse_result result = loop_orig.load_file(model_path.c_str());
+    ASSERT_FALSE(result.status) << result.description();
+    result = loop_serialized.load_file(m_out_xml_path.c_str());
+    ASSERT_FALSE(result.status) << result.description();
+
+    auto node1 = loop_orig.child("net").child("layers").find_child_by_attribute("type", "Loop");
+    auto node2 = loop_serialized.child("net").child("layers").find_child_by_attribute("type", "Loop");
+    auto node2_port_map = node2.child("port_map").first_child();
+
+    for (auto ch = node1.child("port_map").first_child(); !ch.empty(); ch = ch.next_sibling()) {
+        auto node1_external_port_id = std::stoi(ch.attribute("external_port_id").value());
+        auto node2_external_port_id = std::stoi(node2_port_map.attribute("external_port_id").value());
+
+        if (node1_external_port_id == -1) {
+            continue;
+        }
+        if (node2_external_port_id == -1) {
+            node2_external_port_id = std::stoi(node2_port_map.next_sibling().attribute("external_port_id").value());
+        }
+        node2_port_map = node2_port_map.next_sibling();
+
+        EXPECT_EQ(node1_external_port_id, node2_external_port_id);
+    }
+}

From ad852f78b47e0580a7d6cefce43bcd1ff5019115 Mon Sep 17 00:00:00 2001
From: Marina Kolpakova <marina.kolpakova@intel.com>
Date: Wed, 16 Jun 2021 01:00:19 +0300
Subject: [PATCH 094/178] =?UTF-8?q?[=C2=A7]=20cleans=20snippets=20interfac?=
 =?UTF-8?q?e=20and=20adds=20subgraph=20tests=20(#6123)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .../src/snippets/include/snippets/emitter.hpp | 56 ++++++++++++++
 .../snippets/include/snippets/generator.hpp   | 77 ++++++++++---------
 .../snippets/include/snippets/op/kernel.hpp   | 37 +++++++++
 .../src/snippets/include/snippets/op/nop.hpp  |  4 +
 .../snippets/include/snippets/op/subgraph.hpp | 47 ++++++++++-
 .../src/snippets/include/snippets/op/tile.hpp | 36 +++++++++
 .../src/snippets/src/generator.cpp            | 59 +++++++++++++-
 .../src/snippets/src/op/kernel.cpp            | 14 ++++
 .../src/snippets/src/op/subgraph.cpp          | 20 ++++-
 inference-engine/src/snippets/src/op/tile.cpp | 14 ++++
 .../snippets/src/pass/collapse_subgraph.cpp   | 25 ++++--
 .../subgraph_tests/codegen_add.cpp            | 29 +++++++
 .../subgraph_tests/codegen_bert.cpp           | 26 +++++++
 .../subgraph_tests/codegen_gelu.cpp           | 26 +++++++
 .../include/subgraph_tests/codegen_add.hpp    | 34 ++++++++
 .../include/subgraph_tests/codegen_bert.hpp   | 34 ++++++++
 .../include/subgraph_tests/codegen_gelu.hpp   | 34 ++++++++
 .../shared/src/subgraph_tests/codegen_add.cpp | 60 +++++++++++++++
 .../src/subgraph_tests/codegen_bert.cpp       | 76 ++++++++++++++++++
 .../src/subgraph_tests/codegen_gelu.cpp       | 74 ++++++++++++++++++
 20 files changed, 734 insertions(+), 48 deletions(-)
 create mode 100644 inference-engine/src/snippets/include/snippets/emitter.hpp
 create mode 100644 inference-engine/src/snippets/include/snippets/op/kernel.hpp
 create mode 100644 inference-engine/src/snippets/include/snippets/op/tile.hpp
 create mode 100644 inference-engine/src/snippets/src/op/kernel.cpp
 create mode 100644 inference-engine/src/snippets/src/op/tile.cpp
 create mode 100644 inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_add.cpp
 create mode 100644 inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_bert.cpp
 create mode 100644 inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_gelu.cpp
 create mode 100644 inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_add.hpp
 create mode 100644 inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_bert.hpp
 create mode 100644 inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_gelu.hpp
 create mode 100644 inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_add.cpp
 create mode 100644 inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_bert.cpp
 create mode 100644 inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_gelu.cpp

diff --git a/inference-engine/src/snippets/include/snippets/emitter.hpp b/inference-engine/src/snippets/include/snippets/emitter.hpp
new file mode 100644
index 00000000000000..78d54098324010
--- /dev/null
+++ b/inference-engine/src/snippets/include/snippets/emitter.hpp
@@ -0,0 +1,56 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <transformations_visibility.hpp>
+
+#include <vector>
+#include <cstdint>
+
+namespace ngraph {
+namespace snippets {
+
+using code = const uint8_t *;
+using RegInfo = std::pair<std::vector<size_t>, std::vector<size_t>>;
+
+/**
+ * @interface Emitter
+ * @brief Base class for all target specific code emitters used by generator.
+ * @ingroup snippets
+ */
+class TRANSFORMATIONS_API Emitter {
+public:
+    /**
+     * @brief Default constructor
+     */
+    Emitter(const std::shared_ptr<ngraph::Node>& n) {
+    }
+
+    Emitter(std::vector<std::pair<std::shared_ptr<Emitter>, RegInfo>>& region) {
+    }
+
+    /**
+     * @brief called by generator to generate code to produce target code for a specific operation
+     * @param in vector of vector argument registers
+     * @param out vector of vector resulting registers
+     * @param pool optional vector of free vector registers which might be used inside method
+     * @param gpr vector of free generam puproce registers which might be used inside method
+     * @return void
+     */
+    virtual void emit_code(const std::vector<size_t>& in,
+                           const std::vector<size_t>& out,
+                           const std::vector<size_t>& pool = {},
+                           const std::vector<size_t>& gpr  = {}) const = 0;
+
+    /**
+     * @brief called by generator to generate data section, if needed for a specific operation
+     * @return void
+     */
+    virtual void emit_data() const {
+    }
+};
+
+} // namespace snippets
+} // namespace ngraph
\ No newline at end of file
diff --git a/inference-engine/src/snippets/include/snippets/generator.hpp b/inference-engine/src/snippets/include/snippets/generator.hpp
index a7408134c5e657..1a7ace69854f90 100644
--- a/inference-engine/src/snippets/include/snippets/generator.hpp
+++ b/inference-engine/src/snippets/include/snippets/generator.hpp
@@ -10,63 +10,60 @@
 
 #include <transformations_visibility.hpp>
 #include "snippets_isa.hpp"
+#include "emitter.hpp"
 
 namespace ngraph {
 namespace snippets {
 
-using code = const uint8_t *;
-using RegInfo = std::pair<std::vector<size_t>, std::vector<size_t>>;
-
 TRANSFORMATIONS_API auto getRegisters(std::shared_ptr<ngraph::Node>& n) -> ngraph::snippets::RegInfo;
 
 /**
- * @interface Emitter
- * @brief Base class for all target specific code emitters used by generator.
+ * @interface TargetMachine
+ * @brief Base class Target machine representation. Target derives from this class to provide generator information about supported emittors
  * @ingroup snippets
  */
-class TRANSFORMATIONS_API Emitter {
+class TRANSFORMATIONS_API TargetMachine {
 public:
     /**
-     * @brief Default constructor
+     * @brief checks if target is natively supported
+     * @return true, if supported
      */
-    Emitter(const std::shared_ptr<ngraph::Node>& n) {
-    }
+    virtual bool is_supported() const = 0;
 
     /**
-     * @brief called by generator to generate code to produce target code for a specific operation
-     * @param in vector of vector argument registers
-     * @param out vector of vector resulting registers
-     * @param pool optional vector of free vector registers which might be used inside method
-     * @param gpr vector of free generam puproce registers which might be used inside method
-     * @return void
+     * @brief finalizes code generation
+     * @return generated kernel binary
      */
-    virtual void emit_code(const std::vector<size_t>& in,
-                           const std::vector<size_t>& out,
-                           const std::vector<size_t>& pool = {},
-                           const std::vector<size_t>& gpr  = {}) const = 0;
+    virtual code get_snippet() const = 0;
 
     /**
-     * @brief called by generator to generate data section, if needed for a specific operation
-     * @return void
+     * @brief gets number of lanes supported by target's vector ISA
+     * @return number of lanes
      */
-    virtual void emit_data() const {
-    }
-};
+    virtual size_t get_lanes() const = 0;
 
-/**
- * @interface TargetMachine
- * @brief Base class Target machine representation. Target derives from this class to provide generator information about supported emittors
- * @ingroup snippets
- */
-class TRANSFORMATIONS_API TargetMachine {
-public:
     /**
-     * @brief called by generator to all the emittors available for a target machine
+     * @brief called by generator to all the emittor for a target machine
      * @return a map by node's type info with callbacks to create an instance of emmitter for corresponding operation type
      */
-    virtual auto getJitters() -> std::map<const ngraph::DiscreteTypeInfo, std::function<std::shared_ptr<Emitter>(std::shared_ptr<ngraph::Node>)>>{
-        return {};
+    std::function<std::shared_ptr<Emitter>(std::shared_ptr<ngraph::Node>)> get(const ngraph::DiscreteTypeInfo type) const {
+        auto jitter = jitters.find(type);
+        if (jitter == jitters.end()) {
+            throw ngraph_error(std::string("Target code emitter is not available for ") + type.name + " operation.");
+        }
+        return jitter->second;
+    }
+
+    /**
+     * @brief checks if emitter for a specific operation is supported
+     * @return true, if supported
+     */
+    bool has(const ngraph::DiscreteTypeInfo type) const {
+        return jitters.find(type) != jitters.end();
     }
+
+protected:
+    std::map<const ngraph::DiscreteTypeInfo, std::function<std::shared_ptr<Emitter>(std::shared_ptr<ngraph::Node>)>> jitters;
 };
 
 /**
@@ -87,6 +84,12 @@ class TRANSFORMATIONS_API Schedule {
      * @param p pointer to generated code
      */
     Schedule(const Shape& ws, bool f, code p) : work_size(ws), is_flat(f), ptr(p) {}
+    /**
+     * @brief Returns callable instanse of code pointer
+     */
+    template<typename K> K get_callable() const {
+        return reinterpret_cast<K>(const_cast<unsigned char*>(ptr));
+    }
 
     Shape work_size {};
     bool is_flat {false};
@@ -103,7 +106,7 @@ class TRANSFORMATIONS_API Generator {
     /**
      * @brief Default constructor
      */
-    Generator() = default;
+    Generator(const std::shared_ptr<TargetMachine>& t) : target(t) {}
     /**
      * @brief Default destructor
      */
@@ -113,10 +116,10 @@ class TRANSFORMATIONS_API Generator {
      * @param f runction in canonical for for table-based code generation
      * @return pointer to generated code
      */
-    virtual code generate(std::shared_ptr<Function>& f) const = 0;
+    code generate(std::shared_ptr<Function>& f) const;
 
 protected:
-    mutable std::map<const ngraph::DiscreteTypeInfo, std::function<std::shared_ptr<Emitter>(std::shared_ptr<ngraph::Node>)>> jitters;
+    std::shared_ptr<TargetMachine> target;
 };
 
 } // namespace snippets
diff --git a/inference-engine/src/snippets/include/snippets/op/kernel.hpp b/inference-engine/src/snippets/include/snippets/op/kernel.hpp
new file mode 100644
index 00000000000000..f93b71096166b4
--- /dev/null
+++ b/inference-engine/src/snippets/include/snippets/op/kernel.hpp
@@ -0,0 +1,37 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <transformations_visibility.hpp>
+
+#include "ngraph/op/op.hpp"
+#include "snippets/emitter.hpp"
+
+namespace ngraph {
+namespace snippets {
+namespace op {
+
+/**
+ * @interface Kernel
+ * @brief Generated by Canonicalization and represents compute kernel legal for sheduling
+ * @ingroup snippets
+ */
+class TRANSFORMATIONS_API Kernel : public ngraph::op::Op {
+public:
+    NGRAPH_RTTI_DECLARATION;
+
+    Kernel(const std::vector<std::pair<std::shared_ptr<ngraph::snippets::Emitter>, ngraph::snippets::RegInfo>>& region);
+    Kernel() = default;
+
+    std::vector<std::pair<std::shared_ptr<ngraph::snippets::Emitter>, ngraph::snippets::RegInfo>> region;
+
+    std::shared_ptr<Node> clone_with_new_inputs(const OutputVector& inputs) const override {
+        return std::make_shared<Kernel>(region);
+    }
+};
+
+} // namespace op
+} // namespace snippets
+} // namespace ngraph
\ No newline at end of file
diff --git a/inference-engine/src/snippets/include/snippets/op/nop.hpp b/inference-engine/src/snippets/include/snippets/op/nop.hpp
index 8bf6b21a2b6b8e..815776d59a0326 100644
--- a/inference-engine/src/snippets/include/snippets/op/nop.hpp
+++ b/inference-engine/src/snippets/include/snippets/op/nop.hpp
@@ -23,6 +23,10 @@ class TRANSFORMATIONS_API Nop : public ngraph::op::Op {
 
     Nop(const OutputVector& arguments, const OutputVector& results);
     Nop() = default;
+
+    std::shared_ptr<Node> clone_with_new_inputs(const OutputVector& inputs) const override {
+        return std::make_shared<Nop>();
+    }
 };
 
 } // namespace op
diff --git a/inference-engine/src/snippets/include/snippets/op/subgraph.hpp b/inference-engine/src/snippets/include/snippets/op/subgraph.hpp
index b204fdd22fc795..4b00b2bb9a07c8 100644
--- a/inference-engine/src/snippets/include/snippets/op/subgraph.hpp
+++ b/inference-engine/src/snippets/include/snippets/op/subgraph.hpp
@@ -10,6 +10,7 @@
 #include <ngraph/function.hpp>
 #include <ngraph/op/op.hpp>
 #include <ngraph/rt_info.hpp>
+#include <ngraph/pass/manager.hpp>
 
 #include "snippets/generator.hpp"
 
@@ -24,6 +25,47 @@ namespace op {
  */
 class TRANSFORMATIONS_API Subgraph : public ngraph::op::Op {
 public:
+    // < 1, 42, 17, 15, 16> < 0, 1, 2, 3, 1>
+    // should be:
+    // A = < 1, 42, 17, 15> -> < 1, 3, 17, 15, 16> < 0, 1, 2, 3, 1>
+    // B = < 1,  1, 17, 15> -> < 1, 1, 17, 15, 16> < 0, 1, 2, 3, 1>
+    // D = < 1, 42,  1, 1 > -> < 1, 3,  1,  1, 16> < 0, 1, 2, 3, 1> ???
+    // C = A + B
+    // C = < 1, 42, 17, 15> -> < 1, 3, 17, 15, 16> < 0, 1, 2, 3, 1>
+    //
+    // how it works now (multi-demention broadcast):
+    // [BroadcastLoad] doesn't perform post increment
+    // [Load] performs += vlan
+    // [ScalarLoad] performs += 1
+    // A = < 1, 42, 17, 15> -> < 1, 3, 17, 15, 16> < 0, 1, 2, 3, 1>
+    // B = < 1,  1, 17, 15> -> < 1, 1, 17, 15,  1> < 0, 1, 2, 3, 1>
+    // [A]     [B]
+    // [Load]  [ScalarLoad] <- should consider AxisVector to choose right type of load
+    //         [Broadcast]
+    //   [Add]
+    //  [Store]
+    //    [C]
+    // C = A + B
+    // C = < 1, 42, 17, 15> -> < 1, 3, 17, 15, 16> < 0, 1, 2, 3, 1>
+    //
+    // Multiple-dimension broadcasts support?
+    // A = < 1, 42, 17, 15> -> < 1, 3, 17, 15, 16> < 0, 1, 2, 3, 4>
+    // B = < 1,  1, 17, 15> -> < 1, 1, 17, 15,  1> < 0, 1, 2, 3, 4>
+    //
+    // A = < 1, 42, 17, 15> -> < 1, 3, 17, 15, 16> < 0, 1, 2, 3, 4>
+    // B = < 1,  1, 17, 15> -> < 1, 3, 17, 15,  1> < 0, 1, 2, 3, 4>
+    //
+    // Collapse moat varying dimensions with broadcast
+    // A = < 1, 42, 17, 15> -> < 1, 3, 17, 15, 16> < 0, 1, 2, 3, 1>
+    // B = < 1,  1, 17, 15> -> < 1, 3, 17, 15,  1> < 0, 1, 2, 3, 1>
+    //
+    // Collapse for mixed broadcast
+    // A = < 1, 3, 17, 15, 32> < 0, 1, 2, 3, 4>
+    // B = < 1, 3, 17,  1, 32> < 0, 1, 2, 3, 4>
+    // C = < 1, 3,  1, 15, 32> < 0, 1, 2, 3, 4>
+    //
+    // D = < 1, 3, 17, 15, 32> < 0, 1, 2, 3, 4>
+    // E = < 1, 3, 17,  1, 32> < 0, 1, 2, 3, 4>
     using BlockedShape = std::tuple<ngraph::Shape, ngraph::AxisVector, ngraph::element::Type>;
     using BlockedShapeVector = std::vector<BlockedShape>;
 
@@ -49,7 +91,8 @@ class TRANSFORMATIONS_API Subgraph : public ngraph::op::Op {
 
     std::shared_ptr<Subgraph> make_canonical_from_this();
 
-    snippets::Schedule generate(const BlockedShapeVector& output_shapes, const BlockedShapeVector& input_shapes);
+    snippets::Schedule generate(const BlockedShapeVector& output_shapes, const BlockedShapeVector& input_shapes,
+                                ngraph::pass::Manager opt = ngraph::pass::Manager());
     bool evaluate(const HostTensorVector& output_values, const HostTensorVector& input_values) const override;
 
     /// Set a new body for the op; body needs to satisfy requirements on inputs/outputs
@@ -62,6 +105,8 @@ class TRANSFORMATIONS_API Subgraph : public ngraph::op::Op {
     void print() const;
     void print_statistics(bool verbose);
 
+    void serialize() const;
+
     static auto wrap_node_as_subgraph(const std::shared_ptr<ngraph::Node>& node) -> std::shared_ptr<Subgraph>;
 
 private:
diff --git a/inference-engine/src/snippets/include/snippets/op/tile.hpp b/inference-engine/src/snippets/include/snippets/op/tile.hpp
new file mode 100644
index 00000000000000..874e962ef09536
--- /dev/null
+++ b/inference-engine/src/snippets/include/snippets/op/tile.hpp
@@ -0,0 +1,36 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <transformations_visibility.hpp>
+
+#include "ngraph/op/op.hpp"
+#include "snippets/emitter.hpp"
+
+namespace ngraph {
+namespace snippets {
+namespace op {
+
+/**
+ * @interface Tile
+ * @brief Generated by Canonicalization and represents Loop in affine notation
+ * @ingroup snippets
+ */
+class TRANSFORMATIONS_API Tile : public ngraph::op::Op {
+public:
+    NGRAPH_RTTI_DECLARATION;
+
+    Tile(const std::vector<std::pair<std::shared_ptr<ngraph::snippets::Emitter>, ngraph::snippets::RegInfo>>& region);
+    Tile() = default;
+    std::vector<std::pair<std::shared_ptr<ngraph::snippets::Emitter>, ngraph::snippets::RegInfo>> region;
+
+    std::shared_ptr<Node> clone_with_new_inputs(const OutputVector& inputs) const override {
+        return std::make_shared<Tile>(region);
+    }
+};
+
+} // namespace op
+} // namespace snippets
+} // namespace ngraph
\ No newline at end of file
diff --git a/inference-engine/src/snippets/src/generator.cpp b/inference-engine/src/snippets/src/generator.cpp
index 650c91992343d0..23fb777d4de2d8 100644
--- a/inference-engine/src/snippets/src/generator.cpp
+++ b/inference-engine/src/snippets/src/generator.cpp
@@ -4,10 +4,18 @@
 
 #include "snippets/generator.hpp"
 #include "snippets/register_info.hpp"
+#include "snippets/pass/assign_registers.hpp"
+#include "snippets/pass/vector_to_scalar.hpp"
+#include "snippets/pass/insert_load_store.hpp"
+#include "snippets/op/tile.hpp"
+#include "snippets/op/kernel.hpp"
+
+#include <ngraph/pass/manager.hpp>
 
 auto ngraph::snippets::getRegisters(std::shared_ptr<ngraph::Node>& n) -> ngraph::snippets::RegInfo {
     auto rt = n->get_rt_info();
 
+    // ToDo: change to reg_t
     std::vector<size_t> rout;
     if (auto rinfo = rt["reginfo"]) {
         auto reginfo = ngraph::as_type_ptr<ngraph::VariantWrapper<std::vector<size_t>>>(rinfo)->get();
@@ -27,4 +35,53 @@ auto ngraph::snippets::getRegisters(std::shared_ptr<ngraph::Node>& n) -> ngraph:
         }
     }
     return std::make_pair(rin, rout);
-}
\ No newline at end of file
+}
+
+ngraph::snippets::code ngraph::snippets::Generator::generate(std::shared_ptr<ngraph::Function>& f) const {
+    if (!target->is_supported())
+        throw ngraph_error("unsupported architecture for code genration");
+
+    auto params = f->get_parameters();
+    auto results = f->get_results();
+    auto nptrs = results.size() + params.size();
+
+    if (nptrs > 7) {
+        throw ngraph_error("snippet signature should not exceed 7 arguments. got " + std::to_string(nptrs));
+    }
+
+    // vector tile
+    std::vector<std::pair<std::shared_ptr<ngraph::snippets::Emitter>, ngraph::snippets::RegInfo>> lowered;
+    for (auto n : f->get_ordered_ops()) {
+        lowered.push_back(std::make_pair(target->get(n->get_type_info())(n), ngraph::snippets::getRegisters(n)));
+    }
+
+    // scalar tile
+    auto f_scalar = ngraph::clone_function(*f.get());
+    ngraph::pass::Manager m;
+    m.register_pass<ngraph::snippets::pass::ReplaceLoadsWithScalarLoads>();
+    m.register_pass<ngraph::snippets::pass::ReplaceStoresWithScalarStores>();
+    m.run_passes(f_scalar);
+
+    std::vector<std::pair<std::shared_ptr<Emitter>, RegInfo>> scalar_lowered;
+    for (auto n : f_scalar->get_ordered_ops()) {
+        scalar_lowered.push_back(std::make_pair(target->get(n->get_type_info())(n), ngraph::snippets::getRegisters(n)));
+    }
+
+    // wrapping into tiles
+    std::vector<std::pair<std::shared_ptr<Emitter>, RegInfo>> tiles;
+    tiles.push_back(std::make_pair(target->get(ngraph::snippets::op::Tile::type_info)(std::make_shared<ngraph::snippets::op::Tile>(lowered)),
+                                   std::make_pair(std::vector<size_t>({target->get_lanes(), nptrs}), std::vector<size_t>{})));
+    tiles.push_back(std::make_pair(target->get(ngraph::snippets::op::Tile::type_info)(std::make_shared<ngraph::snippets::op::Tile>(scalar_lowered)),
+                    std::make_pair(std::vector<size_t>{{1, nptrs}}, std::vector<size_t>{})));
+
+    // emission
+    std::shared_ptr<Emitter> kernel = target->get(ngraph::snippets::op::Kernel::type_info)(std::make_shared<ngraph::snippets::op::Kernel>(tiles));
+    kernel->emit_code({params.size(), results.size()}, {});
+
+    lowered.insert(lowered.end(), scalar_lowered.begin(), scalar_lowered.end());
+    for (auto& op : lowered) {
+        op.first->emit_data();
+    }
+
+    return target->get_snippet();
+}
diff --git a/inference-engine/src/snippets/src/op/kernel.cpp b/inference-engine/src/snippets/src/op/kernel.cpp
new file mode 100644
index 00000000000000..381db1a5076f6a
--- /dev/null
+++ b/inference-engine/src/snippets/src/op/kernel.cpp
@@ -0,0 +1,14 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "snippets/op/kernel.hpp"
+#include "snippets/generator.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+NGRAPH_RTTI_DEFINITION(snippets::op::Kernel, "Kernel", 0);
+
+snippets::op::Kernel::Kernel(const std::vector<std::pair<std::shared_ptr<snippets::Emitter>, snippets::RegInfo>>& nested) : Op(), region(nested) {
+}
diff --git a/inference-engine/src/snippets/src/op/subgraph.cpp b/inference-engine/src/snippets/src/op/subgraph.cpp
index 544f398c21ffc0..f58b1d383dceea 100644
--- a/inference-engine/src/snippets/src/op/subgraph.cpp
+++ b/inference-engine/src/snippets/src/op/subgraph.cpp
@@ -12,6 +12,7 @@
 #include "snippets/pass/assign_registers.hpp"
 
 #include <ngraph/pass/manager.hpp>
+#include <transformations/serialize.hpp>
 
 #include <algorithm>
 #include <memory>
@@ -176,9 +177,7 @@ void snippets::op::Subgraph::canonicalize(const BlockedShapeVector& output_shape
             if (param->get_element_type() != std::get<2>(input_shapes[i])) {
                 throw ngraph::ngraph_error("changes in presision. Is it legal??");
             }
-            if (param->get_shape().size() != std::get<0>(input_shapes[i]).size()) {
-                m_body->replace_parameter(i, std::make_shared<opset1::Parameter>(std::get<2>(input_shapes[i]), std::get<0>(input_shapes[i])));
-            }
+            m_body->replace_parameter(i, std::make_shared<opset1::Parameter>(std::get<2>(input_shapes[i]), std::get<0>(input_shapes[i])));
         }
     }
 
@@ -204,16 +203,20 @@ void snippets::op::Subgraph::convert_to_snippet_dialect() {
     manager.run_passes(m_body);
 }
 
-snippets::Schedule snippets::op::Subgraph::generate(const BlockedShapeVector& output_shapes, const BlockedShapeVector& input_shapes) {
+snippets::Schedule snippets::op::Subgraph::generate(const BlockedShapeVector& output_shapes, const BlockedShapeVector& input_shapes,
+                                                    ngraph::pass::Manager opt) {
     INTERNAL_OP_SCOPE(Subgraph);
     NGRAPH_CHECK(m_generator != nullptr, "generate is called while generator is not set");
 
     canonicalize(output_shapes, input_shapes);
     convert_to_snippet_dialect();
+    opt.run_passes(m_body);
 
     // generation flow
     snippets::pass::AssignRegisters().run_on_function(m_body);
 
+    // shedule generation should go here and be target agnostic
+
     // actual code emission
     ngraph::snippets::code ptr = m_generator->generate(m_body);
 
@@ -342,3 +345,12 @@ void snippets::op::Subgraph::print_statistics(bool verbose) {
         this->print();
     }
 }
+
+void snippets::op::Subgraph::serialize() const {
+    std::stringstream xmlFile, binFile;
+    ngraph::pass::Serialize serializer(xmlFile, xmlFile, ngraph::pass::Serialize::Version::IR_V10);
+    serializer.run_on_function(get_body());
+    auto m_constants = binFile.str();
+    auto m_model = xmlFile.str();
+    std::cout << m_model << std::endl;
+}
diff --git a/inference-engine/src/snippets/src/op/tile.cpp b/inference-engine/src/snippets/src/op/tile.cpp
new file mode 100644
index 00000000000000..6e630585e6554d
--- /dev/null
+++ b/inference-engine/src/snippets/src/op/tile.cpp
@@ -0,0 +1,14 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "snippets/op/tile.hpp"
+#include "snippets/generator.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+NGRAPH_RTTI_DEFINITION(snippets::op::Tile, "Tile", 0);
+
+snippets::op::Tile::Tile(const std::vector<std::pair<std::shared_ptr<snippets::Emitter>, snippets::RegInfo>>& nested) : Op(), region(nested) {
+}
diff --git a/inference-engine/src/snippets/src/pass/collapse_subgraph.cpp b/inference-engine/src/snippets/src/pass/collapse_subgraph.cpp
index 6c864ec5c4f03f..6f05719d9d5252 100644
--- a/inference-engine/src/snippets/src/pass/collapse_subgraph.cpp
+++ b/inference-engine/src/snippets/src/pass/collapse_subgraph.cpp
@@ -12,7 +12,6 @@
 #include <ngraph/rt_info.hpp>
 #include <ngraph/op/loop.hpp>
 
-
 #include <memory>
 #include <vector>
 #include <cassert>
@@ -66,10 +65,6 @@ auto has_cycles_of_dependencies(const std::vector<std::set<ngraph::Input<ngraph:
             ngraph::Node* curr = stack.front();
             visited.insert(curr);
 
-            if (ngraph::op::is_output(curr)) {
-                return false;
-            }
-
             stack.pop();
 
             if (curr != to) {
@@ -189,6 +184,14 @@ auto has_supported_in_out(std::shared_ptr<Node> n) -> bool {
         if (in.get_tensor().get_element_type() != ngraph::element::f32) {
             return false;
         }
+
+        if (in.get_partial_shape().is_dynamic()) {
+            return false;
+        }
+
+        if (in.get_partial_shape().is_static() && in.get_shape().size() > 6) {
+            return false;
+        }
     }
 
     for (auto out : n->outputs()) {
@@ -196,10 +199,22 @@ auto has_supported_in_out(std::shared_ptr<Node> n) -> bool {
             return false;
         }
 
+        if (out.get_partial_shape().is_dynamic()) {
+            return false;
+        }
+
+        if (out.get_partial_shape().is_static() && out.get_shape().size() > 6) {
+            return false;
+        }
+
         for (auto in_out : out.get_target_inputs()) {
             if (!!as_type_ptr<ngraph::op::v5::Loop>(in_out.get_node()->shared_from_this())) {
                 return false;
             }
+
+            if (!!as_type_ptr<ngraph::op::v0::Result>(in_out.get_node()->shared_from_this())) {
+                return false;
+            }
         }
     }
 
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_add.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_add.cpp
new file mode 100644
index 00000000000000..72c50bce2a4f7f
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_add.cpp
@@ -0,0 +1,29 @@
+
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <vector>
+
+#include "subgraph_tests/codegen_add.hpp"
+#include "common_test_utils/test_constants.hpp"
+
+using namespace LayerTestsDefinitions;
+
+namespace {
+
+    const std::vector<InferenceEngine::Precision> netPrecisions = {
+            InferenceEngine::Precision::FP32
+    };
+
+    INSTANTIATE_TEST_CASE_P(NoReshape, CodegenAdd,
+            ::testing::Combine(
+            ::testing::ValuesIn(netPrecisions),
+            ::testing::Values(InferenceEngine::SizeVector({1, 42, 16, 64})),
+            ::testing::Values(InferenceEngine::SizeVector({1, 42, 16, 64}),
+                              InferenceEngine::SizeVector({1, 42, 16,  1}),
+                              InferenceEngine::SizeVector({1, 42,  1, 64}),
+                              InferenceEngine::SizeVector({1,  1, 16, 64})),
+            ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+            CodegenAdd::getTestCaseName);
+}  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_bert.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_bert.cpp
new file mode 100644
index 00000000000000..97b3737d69366d
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_bert.cpp
@@ -0,0 +1,26 @@
+
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <vector>
+
+#include "subgraph_tests/codegen_bert.hpp"
+#include "common_test_utils/test_constants.hpp"
+
+using namespace LayerTestsDefinitions;
+
+namespace {
+
+    const std::vector<InferenceEngine::Precision> netPrecisions = {
+            InferenceEngine::Precision::FP32
+    };
+
+    INSTANTIATE_TEST_CASE_P(NoReshape, CodegenBert,
+            ::testing::Combine(
+            ::testing::ValuesIn(netPrecisions),
+            ::testing::Values(InferenceEngine::SizeVector({1, 42, 16, 64})),
+            ::testing::Values(InferenceEngine::SizeVector({1, 42, 64, 64})),
+            ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+            CodegenBert::getTestCaseName);
+}  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_gelu.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_gelu.cpp
new file mode 100644
index 00000000000000..81b03911dabe41
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_gelu.cpp
@@ -0,0 +1,26 @@
+
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <vector>
+
+#include "subgraph_tests/codegen_gelu.hpp"
+#include "common_test_utils/test_constants.hpp"
+
+using namespace LayerTestsDefinitions;
+
+namespace {
+
+    const std::vector<InferenceEngine::Precision> netPrecisions = {
+            InferenceEngine::Precision::FP32
+    };
+
+    INSTANTIATE_TEST_CASE_P(NoReshape, CodegenGelu,
+            ::testing::Combine(
+            ::testing::ValuesIn(netPrecisions),
+            ::testing::Values(InferenceEngine::SizeVector({1, 384, 4096})),
+            ::testing::Values(true, false),
+            ::testing::Values(CommonTestUtils::DEVICE_CPU)),
+            CodegenGelu::getTestCaseName);
+}  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_add.hpp b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_add.hpp
new file mode 100644
index 00000000000000..969f15c3c5103e
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_add.hpp
@@ -0,0 +1,34 @@
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <tuple>
+#include <vector>
+#include <string>
+#include <memory>
+
+#include "shared_test_classes/base/layer_test_utils.hpp"
+#include "ngraph_functions/utils/ngraph_helpers.hpp"
+#include "ngraph_functions/builders.hpp"
+
+namespace LayerTestsDefinitions {
+
+typedef std::tuple<
+        InferenceEngine::Precision,  // Network Precision
+        InferenceEngine::SizeVector, // Input 0 Shape
+        InferenceEngine::SizeVector, // Input 1 Shape
+        std::string                  // Target Device
+> multiInputParams;
+
+class CodegenAdd : public testing::WithParamInterface<LayerTestsDefinitions::multiInputParams>,
+public LayerTestsUtils::LayerTestsCommon {
+public:
+    static std::string getTestCaseName(testing::TestParamInfo<LayerTestsDefinitions::multiInputParams> obj);
+
+protected:
+    void SetUp() override;
+};
+
+}  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_bert.hpp b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_bert.hpp
new file mode 100644
index 00000000000000..449f259f34801e
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_bert.hpp
@@ -0,0 +1,34 @@
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <tuple>
+#include <vector>
+#include <string>
+#include <memory>
+
+#include "shared_test_classes/base/layer_test_utils.hpp"
+#include "ngraph_functions/utils/ngraph_helpers.hpp"
+#include "ngraph_functions/builders.hpp"
+
+namespace LayerTestsDefinitions {
+
+typedef std::tuple<
+        InferenceEngine::Precision,  // Network Precision
+        InferenceEngine::SizeVector, // Input 0 Shape
+        InferenceEngine::SizeVector, // Input 1 Shape
+        std::string                  // Target Device
+> multiInputParams;
+
+class CodegenBert : public testing::WithParamInterface<LayerTestsDefinitions::multiInputParams>,
+public LayerTestsUtils::LayerTestsCommon {
+public:
+    static std::string getTestCaseName(testing::TestParamInfo<LayerTestsDefinitions::multiInputParams> obj);
+
+protected:
+    void SetUp() override;
+};
+
+}  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_gelu.hpp b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_gelu.hpp
new file mode 100644
index 00000000000000..d4c90a036d4000
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/shared/include/subgraph_tests/codegen_gelu.hpp
@@ -0,0 +1,34 @@
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <tuple>
+#include <vector>
+#include <string>
+#include <memory>
+
+#include "shared_test_classes/base/layer_test_utils.hpp"
+#include "ngraph_functions/utils/ngraph_helpers.hpp"
+#include "ngraph_functions/builders.hpp"
+
+namespace LayerTestsDefinitions {
+
+typedef std::tuple<
+        InferenceEngine::Precision,  // Network Precision
+        InferenceEngine::SizeVector, // Input Shape,
+        bool,
+        std::string                  // Target Device
+> multiInputParams;
+
+class CodegenGelu : public testing::WithParamInterface<LayerTestsDefinitions::multiInputParams>,
+public LayerTestsUtils::LayerTestsCommon {
+public:
+    static std::string getTestCaseName(testing::TestParamInfo<LayerTestsDefinitions::multiInputParams> obj);
+
+protected:
+    void SetUp() override;
+};
+
+}  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_add.cpp b/inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_add.cpp
new file mode 100644
index 00000000000000..662ef3313b5cde
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_add.cpp
@@ -0,0 +1,60 @@
+
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <memory>
+#include <tuple>
+#include <vector>
+#include <string>
+
+#include <ie_core.hpp>
+
+#include "common_test_utils/common_utils.hpp"
+#include "functional_test_utils/plugin_cache.hpp"
+#include "shared_test_classes/base/layer_test_utils.hpp"
+#include "functional_test_utils/blob_utils.hpp"
+
+#include "ngraph_functions/pass/convert_prc.hpp"
+
+#include "subgraph_tests/codegen_add.hpp"
+
+namespace LayerTestsDefinitions {
+
+    std::string CodegenAdd::getTestCaseName(testing::TestParamInfo<LayerTestsDefinitions::multiInputParams> obj) {
+        InferenceEngine::Precision netPrecision;
+        InferenceEngine::SizeVector inputShapes0, inputShapes1, newInputShapes;
+        std::string targetDevice;
+        std::tie(netPrecision, inputShapes0, inputShapes1, targetDevice) = obj.param;
+
+        std::ostringstream result;
+        result << "IS[0]=" << CommonTestUtils::vec2str(inputShapes0) << "_";
+        result << "IS[1]=" << CommonTestUtils::vec2str(inputShapes1) << "_";
+        result << "netPRC=" << netPrecision.name() << "_";
+        result << "targetDevice=" << targetDevice;
+        return result.str();
+    }
+
+    void CodegenAdd::SetUp() {
+        std::vector<size_t> inputShape0, inputShape1;
+        InferenceEngine::Precision netPrecision;
+        std::tie(netPrecision, inputShape0, inputShape1, targetDevice) = this->GetParam();
+
+        auto input0 = std::make_shared<ngraph::opset1::Parameter>(ngraph::element::f32, ngraph::Shape{inputShape0});
+        auto input1 = std::make_shared<ngraph::opset1::Parameter>(ngraph::element::f32, ngraph::Shape{inputShape1});
+
+        auto add = std::make_shared<ngraph::opset1::Add>(input0, input1);
+        auto neg = std::make_shared<ngraph::opset1::Negative>(add);
+        auto result = std::make_shared<ngraph::opset1::Result>(neg);
+
+        function = std::make_shared<ngraph::Function>(
+            ngraph::ResultVector{result},
+            ngraph::ParameterVector{input0, input1},
+            "CodegenAdd");
+    }
+
+TEST_P(CodegenAdd, CompareWithRefImpl) {
+    Run();
+};
+
+}  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_bert.cpp b/inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_bert.cpp
new file mode 100644
index 00000000000000..d4e328e19b8a95
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_bert.cpp
@@ -0,0 +1,76 @@
+
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <memory>
+#include <tuple>
+#include <vector>
+#include <string>
+
+#include <ie_core.hpp>
+
+#include "common_test_utils/common_utils.hpp"
+#include "functional_test_utils/plugin_cache.hpp"
+#include "shared_test_classes/base/layer_test_utils.hpp"
+#include "functional_test_utils/blob_utils.hpp"
+
+#include "ngraph_functions/pass/convert_prc.hpp"
+
+#include "subgraph_tests/codegen_bert.hpp"
+
+namespace LayerTestsDefinitions {
+
+    std::string CodegenBert::getTestCaseName(testing::TestParamInfo<LayerTestsDefinitions::multiInputParams> obj) {
+        InferenceEngine::Precision netPrecision;
+        InferenceEngine::SizeVector inputShapes0, inputShapes1, newInputShapes;
+        std::string targetDevice;
+        std::tie(netPrecision, inputShapes0, inputShapes1, targetDevice) = obj.param;
+
+        std::ostringstream result;
+        result << "IS[0]=" << CommonTestUtils::vec2str(inputShapes0) << "_";
+        result << "IS[1]=" << CommonTestUtils::vec2str(inputShapes1) << "_";
+        result << "netPRC=" << netPrecision.name() << "_";
+        result << "targetDevice=" << targetDevice;
+        return result.str();
+    }
+
+    // the simplest possible eltwise operation with streaming access to the data
+    void CodegenBert::SetUp() {
+        std::vector<size_t> inputShape0, inputShape1;
+        InferenceEngine::Precision netPrecision;
+        std::tie(netPrecision, inputShape0, inputShape1, targetDevice) = this->GetParam();
+
+        auto shape = ngraph::Shape{inputShape0};
+        auto input1 = std::make_shared<ngraph::opset1::Parameter>(ngraph::element::f32, shape);
+        auto input2 = std::make_shared<ngraph::opset1::Parameter>(ngraph::element::f32, shape);
+
+        auto shapeMM = ngraph::Shape{inputShape1};
+        auto input3 = std::make_shared<ngraph::opset1::Parameter>(ngraph::element::f32, shapeMM);
+
+        auto add    = std::make_shared<ngraph::opset1::Add>(input1, input2);
+        auto mm     = std::make_shared<ngraph::opset1::MatMul>(add, input3);
+
+        std::vector<float> vals(ngraph::shape_size(shape));
+        for (int i = 0; i < vals.size(); i++) {
+            vals[i] = static_cast<float>(i)*vals.size();
+        }
+
+        auto c0 = std::make_shared<ngraph::opset1::Parameter>(ngraph::element::f32, shape);
+        auto add2    = std::make_shared<ngraph::opset1::Subtract>(mm, c0);
+
+        auto add3    = std::make_shared<ngraph::opset1::Multiply>(add, add2);
+        auto result = std::make_shared<ngraph::opset1::Result>(add3);
+
+        function = std::make_shared<ngraph::Function>(
+            ngraph::ResultVector{result},
+            // it should be some topological order to pass parameters for reference code to be executed correctly
+            ngraph::ParameterVector{input1, input2, c0, input3},
+            "CodegenBert");
+    }
+
+TEST_P(CodegenBert, CompareWithRefImpl) {
+    Run();
+};
+
+}  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_gelu.cpp b/inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_gelu.cpp
new file mode 100644
index 00000000000000..bbf7402d4fa6da
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/shared/src/subgraph_tests/codegen_gelu.cpp
@@ -0,0 +1,74 @@
+
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <memory>
+#include <tuple>
+#include <vector>
+#include <string>
+
+#include <ie_core.hpp>
+
+#include "common_test_utils/common_utils.hpp"
+#include "functional_test_utils/plugin_cache.hpp"
+#include "shared_test_classes/base/layer_test_utils.hpp"
+#include "functional_test_utils/blob_utils.hpp"
+
+#include "ngraph_functions/pass/convert_prc.hpp"
+
+#include "subgraph_tests/codegen_gelu.hpp"
+
+#include <ngraph/pass/constant_folding.hpp>
+#include <ngraph/pass/visualize_tree.hpp>
+
+#include <transformations/init_node_info.hpp>
+#include <transformations/utils/utils.hpp>
+
+namespace LayerTestsDefinitions {
+
+    std::string CodegenGelu::getTestCaseName(testing::TestParamInfo<LayerTestsDefinitions::multiInputParams> obj) {
+        InferenceEngine::Precision netPrecision;
+        InferenceEngine::SizeVector inputShapes0, newInputShapes;
+        bool useSubgraph;
+        std::string targetDevice;
+        std::tie(netPrecision, inputShapes0, useSubgraph, targetDevice) = obj.param;
+
+        std::ostringstream result;
+        result << "IS[0]=" << CommonTestUtils::vec2str(inputShapes0) << "_";
+        result << "netPRC=" << netPrecision.name() << "_";
+        result << "overSnippet=" << (useSubgraph ? "yes" : "no") << "_";
+        result << "targetDevice=" << targetDevice;
+        return result.str();
+    }
+
+    // Gelu from bert-large-uncased-whole-word-masking-squad-fp32-onnx-0001
+    void CodegenGelu::SetUp() {
+        std::vector<size_t> inputShape0;
+        InferenceEngine::Precision netPrecision;
+        bool useSubgraph;
+        std::tie(netPrecision, inputShape0, useSubgraph, targetDevice) = this->GetParam();
+
+        auto input0 = std::make_shared<ngraph::opset1::Parameter>(ngraph::element::f32, ngraph::Shape{inputShape0});
+        auto input1 = std::make_shared<ngraph::opset1::Parameter>(ngraph::element::f32, ngraph::Shape{inputShape0});
+        auto add = std::make_shared<ngraph::opset1::Add>(input0, input1);
+
+        auto gelu = std::make_shared<ngraph::opset2::Gelu>(add);
+        auto result = std::make_shared<ngraph::opset1::Result>(gelu);
+
+        function = std::make_shared<ngraph::Function>(
+            ngraph::ResultVector{result},
+            ngraph::ParameterVector{input0, input1},
+            "CodegenGelu");
+
+        if (useSubgraph) {
+            ngraph::pass::InitNodeInfo().run_on_function(function);
+            ngraph::pass::ConstantFolding().run_on_function(function);
+        }
+    }
+
+TEST_P(CodegenGelu, CompareWithRefImpl) {
+    Run();
+};
+
+}  // namespace LayerTestsDefinitions

From ea3ed8af21b78da54e8e0d89b38cd236d3b6c7ba Mon Sep 17 00:00:00 2001
From: Gleb Kazantaev <gleb.kazantaev@intel.com>
Date: Wed, 16 Jun 2021 03:12:24 +0300
Subject: [PATCH 095/178] Turn on IE and NG python APIs by default inside Model
 Optimizer (#5721)

* Turn on IE and NG python APIs by default inside Model Optimizer

* Remove fallback

* Fix mo_ut

* Remove MO wheel tests

* Add model_optimizer custom target to gather all MO deps inside single traget

* Fix PR comments
---
 .github/workflows/mo.yml                      | 38 ----------
 .../ie_bridges/python/CMakeLists.txt          |  4 ++
 model-optimizer/CMakeLists.txt                |  8 +++
 model-optimizer/mo/main.py                    | 71 ++++++-------------
 model-optimizer/mo/utils/check_ie_bindings.py |  8 +++
 model-optimizer/mo/utils/cli_parser.py        |  9 +--
 .../unit_tests/mo/utils/cli_parser_test.py    |  4 +-
 7 files changed, 47 insertions(+), 95 deletions(-)

diff --git a/.github/workflows/mo.yml b/.github/workflows/mo.yml
index eb2ea91484e7ca..ebc0827adb0622 100644
--- a/.github/workflows/mo.yml
+++ b/.github/workflows/mo.yml
@@ -62,42 +62,4 @@ jobs:
           mkdir ../mo-ut-logs
           python3 -m xmlrunner discover -p *_test.py --output=../mo-ut-logs
         working-directory: model-optimizer
-  
-  build_wheel:
-    name: Build Python wheel
-    runs-on: ubuntu-18.04
-    steps:
-      - uses: actions/checkout@v2
 
-      - name: Install dependencies
-        run: |
-          python3 -m pip install --upgrade pip
-          python3 -m pip install wheel setuptools
-          python3 -m pip install tensorflow==2.3.0
-      
-      - name: Build
-        run: |
-          python3 setup.py sdist bdist_wheel
-        working-directory: model-optimizer
-      
-      - name: Test package content
-        run: |
-          echo "src = open('openvino_mo.egg-info/SOURCES.txt', 'rt').read().split()" | tee -a test_wheel.py
-          echo "ref = open('automation/package_BOM.txt', 'rt').read().split()"       | tee -a test_wheel.py
-          echo "for name in ref:"                                                    | tee -a test_wheel.py
-          echo "  if name.endswith('.py'):"                                          | tee -a test_wheel.py
-          echo "    assert name in src or './' + name in src, name + ' file missed'" | tee -a test_wheel.py
-          python3 test_wheel.py
-        working-directory: model-optimizer
-
-      - name: Test conversion
-        run: |
-          wget -q http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_1.0_224.tgz
-          tar -xf mobilenet_v1_1.0_224.tgz
-          python3 -m pip install model-optimizer/dist/*.whl
-          python3 -m mo --input_model mobilenet_v1_1.0_224_frozen.pb --input_shape "[1,224,224,3]"
-
-      - uses: actions/upload-artifact@v2
-        with:
-          name: mo_wheel
-          path: "model-optimizer/dist/*.whl"
diff --git a/inference-engine/ie_bridges/python/CMakeLists.txt b/inference-engine/ie_bridges/python/CMakeLists.txt
index 5c80af7f574f72..b8216b0cb3435a 100644
--- a/inference-engine/ie_bridges/python/CMakeLists.txt
+++ b/inference-engine/ie_bridges/python/CMakeLists.txt
@@ -68,6 +68,10 @@ if(ENABLE_WHEEL)
     add_subdirectory(wheel)
 endif()
 
+if (NGRAPH_PYTHON_BUILD_ENABLE)
+    add_dependencies(ie_api _pyngraph)
+endif()
+
 # install
 
 ie_cpack_add_component(${PYTHON_VERSION})
diff --git a/model-optimizer/CMakeLists.txt b/model-optimizer/CMakeLists.txt
index e9d54d6081867d..b0e75505cc2cd3 100644
--- a/model-optimizer/CMakeLists.txt
+++ b/model-optimizer/CMakeLists.txt
@@ -1,6 +1,14 @@
 # Copyright (C) 2018-2021 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 
+if (NOT NGRAPH_PYTHON_BUILD_ENABLE)
+    message(WARNING "Please enable nGraph Python API (_pyngraph) target to enable Model Optimizer target")
+elseif(NOT ENABLE_PYTHON)
+    message(WARNING "Please enable IE Python API (ie_api and offline_transformations_api) targets to enable Model Optimizer target")
+else()
+    add_custom_target(model_optimizer DEPENDS ie_api offline_transformations_api inference_engine_ir_reader)
+endif()
+
 # install
 ie_cpack_add_component(model_optimizer)
 
diff --git a/model-optimizer/mo/main.py b/model-optimizer/mo/main.py
index aa0eb628ad8095..b4b32fa786ecbc 100644
--- a/model-optimizer/mo/main.py
+++ b/model-optimizer/mo/main.py
@@ -144,20 +144,18 @@ def prepare_ir(argv: argparse.Namespace):
 
     # This try-except is additional reinsurance that the IE
     # dependency search does not break the MO pipeline
-    try:
-        argv.ie_is_available = find_ie_version(silent=argv.silent)
-
-        if not argv.ie_is_available and not argv.silent:
-            print("[ WARNING ] Could not find the Inference Engine Python API. At this moment, the Inference Engine dependency is not required, but will be required in future releases.")
-            print("[ WARNING ] Consider building the Inference Engine Python API from sources or try to install OpenVINO (TM) Toolkit using \"install_prerequisites.{}\"".format(
+    def raise_ie_not_found():
+        raise Error("Could not find the Inference Engine or nGraph Python API.\n"
+                    "Consider building the Inference Engine and nGraph Python APIs from sources or try to install OpenVINO (TM) Toolkit using \"install_prerequisites.{}\"".format(
                     "bat" if sys.platform == "windows" else "sh"))
-            # If the IE was not found, it will not print the MO version, so we have to print it manually
-            print("{}: \t{}".format("Model Optimizer version", get_version()))
+    try:
+        if not find_ie_version(silent=argv.silent):
+            raise_ie_not_found()
     except Exception as e:
-        argv.ie_is_available = False
+        raise_ie_not_found()
 
     # This is just to check that transform key is valid and transformations are available
-    check_available_transforms(parse_transform(argv.transform), argv.ie_is_available)
+    check_available_transforms(parse_transform(argv.transform))
 
     if argv.legacy_ir_generation and len(argv.transform) != 0:
         raise Error("--legacy_ir_generation and --transform keys can not be used at the same time.")
@@ -261,10 +259,6 @@ def emit_ir(graph: Graph, argv: argparse.Namespace):
     mean_data = deepcopy(graph.graph['mf']) if 'mf' in graph.graph else None
     input_names = deepcopy(graph.graph['input_names']) if 'input_names' in graph.graph else []
 
-    # Remove temporary ie_is_available key from argv no to have it in IR
-    ie_is_available = argv.ie_is_available
-    del argv.ie_is_available
-
     prepare_emit_ir(graph=graph,
                     data_type=graph.graph['cmd_params'].data_type,
                     output_dir=argv.output_dir,
@@ -285,7 +279,7 @@ def emit_ir(graph: Graph, argv: argparse.Namespace):
         # This try-except is additional reinsurance that the IE
         # dependency search does not break the MO pipeline
         try:
-            if not argv.legacy_ir_generation and ie_is_available:
+            if not argv.legacy_ir_generation:
                 path_to_offline_transformations = os.path.join(os.path.realpath(os.path.dirname(__file__)), 'back',
                                                                'offline_transformations.py')
                 status = subprocess.run([sys.executable, path_to_offline_transformations,
@@ -295,7 +289,7 @@ def emit_ir(graph: Graph, argv: argparse.Namespace):
                 return_code = status.returncode
         except Exception as e:
             return_code = "failed"
-            log.error(e, extra={'is_warning': True})
+            log.error(e)
 
         message = str(dict({
             "platform": platform.system(),
@@ -307,39 +301,20 @@ def emit_ir(graph: Graph, argv: argparse.Namespace):
         t = tm.Telemetry()
         t.send_event('mo', 'offline_transformations_status', message)
 
-        # if IR wasn't produced by offline_transformations step we need to fallback to IR
-        # produced by prepare_ir. This IR needs to be renamed from XXX_tmp.xml to XXX.xml
-        suffixes = [".xml", ".bin", ".mapping"]
         if return_code != 0:
-            if len(argv.transform) != 0:
-                # Remove temporary IR before throwing exception
-                for suf in suffixes:
-                    path_to_file = orig_model_name + "_tmp" + suf
-                    if os.path.exists(path_to_file):
-                        os.remove(path_to_file)
-                raise Error("Failed to apply transformations: {}".format(argv.transform))
-
-            log.error("Using fallback to produce IR.", extra={'is_warning': True})
-            for suf in suffixes:
-                # remove existing files
-                path_to_file = orig_model_name + suf
-                if os.path.exists(path_to_file):
-                    os.remove(path_to_file)
-
-                # rename tmp IR to original name
-                os.rename(orig_model_name + "_tmp" + suf, orig_model_name + suf)
-        else:
-            for suf in suffixes:
-                # remove existing files
-                path_to_file = orig_model_name + "_tmp" + suf
-                if os.path.exists(path_to_file):
-                    os.remove(path_to_file)
-
-            # add meta information to IR
-            append_ir_info(file=orig_model_name,
-                           meta_info=get_meta_info(argv),
-                           mean_data=mean_data,
-                           input_names=input_names)
+            raise Error("offline transformations step has failed.")
+
+        for suf in [".xml", ".bin", ".mapping"]:
+            # remove existing files
+            path_to_file = orig_model_name + "_tmp" + suf
+            if os.path.exists(path_to_file):
+                os.remove(path_to_file)
+
+        # add meta information to IR
+        append_ir_info(file=orig_model_name,
+                       meta_info=get_meta_info(argv),
+                       mean_data=mean_data,
+                       input_names=input_names)
 
         print('[ SUCCESS ] Generated IR version {} model.'.format(get_ir_version(argv)))
         print('[ SUCCESS ] XML file: {}.xml'.format(orig_model_name))
diff --git a/model-optimizer/mo/utils/check_ie_bindings.py b/model-optimizer/mo/utils/check_ie_bindings.py
index fe03294bad149b..dae7aeb9958b12 100644
--- a/model-optimizer/mo/utils/check_ie_bindings.py
+++ b/model-optimizer/mo/utils/check_ie_bindings.py
@@ -51,7 +51,13 @@ def import_core_modules(silent: bool, path_to_module: str):
         from openvino.offline_transformations import ApplyMOCTransformations, ApplyLowLatencyTransformation, \
             GenerateMappingFile  # pylint: disable=import-error,no-name-in-module
 
+        # TODO: it is temporary import to check that nGraph python API is available. But in future
+        # we need to replace it with Frontend imports
+        from ngraph.impl.op import Parameter  # pylint: disable=import-error,no-name-in-module
+        from _pyngraph import PartialShape, Dimension  # pylint: disable=import-error,no-name-in-module
+
         import openvino  # pylint: disable=import-error,no-name-in-module
+        import ngraph  # pylint: disable=import-error,no-name-in-module
 
         if silent:
             return True
@@ -60,6 +66,8 @@ def import_core_modules(silent: bool, path_to_module: str):
         mo_version = str(v.get_version())  # pylint: disable=no-member,no-name-in-module
 
         print("\t- {}: \t{}".format("Inference Engine found in", os.path.dirname(openvino.__file__)))
+        # TODO: when nGraph version will be available we need to start compare it to IE and MO versions. Ticket: 58091
+        print("\t- {}: \t{}".format("nGraph found in", os.path.dirname(ngraph.__file__)))
         print("{}: \t{}".format("Inference Engine version", ie_version))
         print("{}: \t{}".format("Model Optimizer version", mo_version))
 
diff --git a/model-optimizer/mo/utils/cli_parser.py b/model-optimizer/mo/utils/cli_parser.py
index e6b2f2d2517f83..a7dd5df023b995 100644
--- a/model-optimizer/mo/utils/cli_parser.py
+++ b/model-optimizer/mo/utils/cli_parser.py
@@ -1217,17 +1217,12 @@ def parse_transform(transform: str) -> list:
     return transforms
 
 
-def check_available_transforms(transforms: list, ie_is_available: bool):
+def check_available_transforms(transforms: list):
     """
     This function check that transformations specified by user are available.
     :param transforms: list of user specified transformations
-    :param ie_is_available: True if IE Python API is available and False if it is not
-    :return: raises an Error if IE or transformation is not available
+    :return: raises an Error if transformation is not available
     """
-    if not ie_is_available and len(transforms) != 0:
-        raise Error('Can not apply {} transformations due to missing Inference Engine Python API'.format(
-            ','.join([name for name, _ in transforms])))
-
     from mo.back.offline_transformations import get_available_transformations
     available_transforms = get_available_transformations()
 
diff --git a/model-optimizer/unit_tests/mo/utils/cli_parser_test.py b/model-optimizer/unit_tests/mo/utils/cli_parser_test.py
index c6c1c96e46a66e..c4d36780a686ca 100644
--- a/model-optimizer/unit_tests/mo/utils/cli_parser_test.py
+++ b/model-optimizer/unit_tests/mo/utils/cli_parser_test.py
@@ -959,11 +959,11 @@ def test_single_pass_with_args_neg6(self):
     def test_check_low_latency_is_available(self, available_transformations):
         available_transformations.return_value = {"LowLatency2": None}
         try:
-            check_available_transforms([("LowLatency2", "")], True)
+            check_available_transforms([("LowLatency2", "")])
         except Error as e:
             self.assertTrue(False, "Exception \"{}\" is unexpected".format(e))
 
     @patch("mo.back.offline_transformations.get_available_transformations")
     def test_check_dummy_pass_is_available(self, available_transformations):
         available_transformations.return_value = {"LowLatency2": None}
-        self.assertRaises(Error, check_available_transforms, [("DummyPass", "")], True)
+        self.assertRaises(Error, check_available_transforms, [("DummyPass", "")])

From 3292543252c657d49de03f2d006920e217bfad0d Mon Sep 17 00:00:00 2001
From: Roman Kazantsev <roman.kazantsev@intel.com>
Date: Wed, 16 Jun 2021 06:43:12 +0300
Subject: [PATCH 096/178] Fix recovery of output subscript in Einsum implicit
 mode (#6131)

* Fix recovery of output subscript in Einsum implicit mode

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>

* Fix code style

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
---
 model-optimizer/extensions/ops/einsum.py      | 23 ++++++++++-
 .../unit_tests/extensions/ops/einsum_test.py  |  5 +++
 ngraph/core/src/op/einsum.cpp                 | 39 +++++++++++++++++--
 ngraph/test/type_prop/einsum.cpp              | 28 +++++++++++++
 4 files changed, 90 insertions(+), 5 deletions(-)

diff --git a/model-optimizer/extensions/ops/einsum.py b/model-optimizer/extensions/ops/einsum.py
index 907989216acf46..a30bb96bdfcb89 100644
--- a/model-optimizer/extensions/ops/einsum.py
+++ b/model-optimizer/extensions/ops/einsum.py
@@ -28,6 +28,22 @@ def __init__(self, graph: Graph, attrs: dict):
     def backend_attrs(self):
         return ['equation']
 
+    @staticmethod
+    def is_label_elsewhere(input_subscripts: list, label_to_check: str, excluded_subscript_inds: list) -> bool:
+        """
+        Check if the given label is met in input subscripts excluding ones specified by a list of indices
+        excluded_subscript_inds
+
+        :param input_subscripts: input subscripts among which to check if the label is met
+        :param label_to_check: a label to check
+        :param excluded_subscript_inds: indices of input subscripts to be excluded for this check
+        :return: True - met, False - otherwise
+        """
+        for ind, input_subscript in enumerate(input_subscripts):
+            if ind not in excluded_subscript_inds and label_to_check in input_subscript:
+                return True
+        return False
+
     @staticmethod
     def parse_equation(node_name: str, equation: str) -> (list, str):
         """
@@ -70,7 +86,12 @@ def parse_equation(node_name: str, equation: str) -> (list, str):
                     "The output subscript of Einsum node {} must contain ellipsis".format(node_name)
         elif len(splitted_equation) == 1:
             # recover output subscript in case implicit mode
-            output_subscript = ''.join(input_subscripts_list)
+            output_subscript = ""
+            for ind, input_subscript in enumerate(input_subscripts_list):
+                labels = Einsum.extract_subscript_labels(node_name, input_subscript)
+                for label in labels:
+                    if Einsum.is_label_elsewhere(input_subscripts_list, label, [ind]) is False:
+                        output_subscript += label
             output_subscript = ''.join(sorted(list(set(output_subscript) - {'.'})))
             if is_ellipsis_met:
                 output_subscript = "..." + output_subscript
diff --git a/model-optimizer/unit_tests/extensions/ops/einsum_test.py b/model-optimizer/unit_tests/extensions/ops/einsum_test.py
index 9407fe63d5f488..960db04310dea9 100644
--- a/model-optimizer/unit_tests/extensions/ops/einsum_test.py
+++ b/model-optimizer/unit_tests/extensions/ops/einsum_test.py
@@ -60,6 +60,11 @@ class TestEinsum(unittest.TestCase):
         ([int64_array([1, 3, 5])], "AbC", int64_array([1, 5, 3])),
         # mixed case letters and equation in implicit mode
         ([int64_array([3, 11, 1, 5]), int64_array([1, 3, 1, 7])], "a...b,B...", int64_array([3, 11, 7, 1, 3, 5])),
+        # inner product in implicit mode
+        ([int64_array([3]), int64_array([3])], "i,i", int64_array([])),
+        # equation with ellipsis and repeated labels in implicit mode
+        # "a...b,b..." is equivalent to "a...b,b...->...a"
+        ([int64_array([9, 1, 4, 3]), int64_array([3, 11, 7, 1])], "a...b,b...", int64_array([11, 7, 4, 9])),
     ])
     def test_einsum(self, input_shapes, equation, ref_output_shape):
         graph = create_einsum_graph(input_shapes, equation)
diff --git a/ngraph/core/src/op/einsum.cpp b/ngraph/core/src/op/einsum.cpp
index fbf52ef888b7b3..721a8e0211e9de 100644
--- a/ngraph/core/src/op/einsum.cpp
+++ b/ngraph/core/src/op/einsum.cpp
@@ -60,11 +60,40 @@ bool is_subscript_correct(const std::string& subscript, bool& is_ellipsis_met)
     return true;
 }
 
+/// \brief      Check if the given label is met in input subscripts excluding ones
+/// specified by a vector excluded_indices
+///
+/// \param      input_subscripts         The vector of the input subscripts
+/// \param      label_to_check           A label to check
+/// \param      excluded_indices         A vector of input subscript indices to be excluded
+///
+/// \return     true - met, false - otherwise
+///
+bool is_label_elsewhere(const std::vector<std::string>& input_subscripts,
+                        const std::string& label_to_check,
+                        const std::vector<size_t>& excluded_indices)
+{
+    for (size_t input_ind = 0; input_ind < input_subscripts.size(); ++input_ind)
+    {
+        const auto& input_subscript = input_subscripts[input_ind];
+        // the subscript is checked only if its index is not in excluded indices list
+        bool check_subscript =
+            (std::find(excluded_indices.begin(), excluded_indices.end(), input_ind) ==
+             excluded_indices.end());
+        if (check_subscript && input_subscript.find(label_to_check) != std::string::npos)
+        {
+            return true;
+        }
+    }
+    return false;
+}
+
 void op::v7::Einsum::parse_equation(const std::string& equation,
                                     std::vector<std::string>& input_subscripts,
                                     std::string& output_subscript)
 {
     NGRAPH_OP_SCOPE(v7_Einsum_parse_equation);
+    constexpr char ellipsis[] = "...";
 
     // split equation to input subscripts and an output subscript
     auto pos_output_delimeter = equation.find("->");
@@ -93,13 +122,15 @@ void op::v7::Einsum::parse_equation(const std::string& equation,
 
     if (pos_output_delimeter == std::string::npos)
     {
-        // recover output subscript
+        // equation is in implicit mode so recover output subscript
         output_subscript = "";
-        for (auto const& input_subscript : input_subscripts)
+        for (size_t ind = 0; ind < input_subscripts.size(); ++ind)
         {
-            for (auto const& label : input_subscript)
+            auto const& input_subscript = input_subscripts[ind];
+            for (auto const& label : extract_labels(input_subscript))
             {
-                if (std::isalpha(label) && output_subscript.find(label) == std::string::npos)
+                if (label != ellipsis &&
+                    (is_label_elsewhere(input_subscripts, label, {ind}) == false))
                 {
                     output_subscript += label;
                 }
diff --git a/ngraph/test/type_prop/einsum.cpp b/ngraph/test/type_prop/einsum.cpp
index a65fb0677f4ad9..cea89ea7076416 100644
--- a/ngraph/test/type_prop/einsum.cpp
+++ b/ngraph/test/type_prop/einsum.cpp
@@ -186,6 +186,34 @@ TEST(type_prop, einsum_implicitmode_mixedcaseletters2)
     ASSERT_TRUE(O->get_output_partial_shape(0).same_scheme(out_shape));
 }
 
+TEST(type_prop, einsum_implicitmode_repeatedlabels)
+{
+    // the following equation is equivalent to "a...b,b...->...a"
+    std::string equation = "a...b,b...";
+    const auto input1_shape = PartialShape{Dimension(3, 5), 11, 1, 3};
+    const auto input2_shape = PartialShape{Dimension(1, 3), 3, 1, 7};
+    const auto out_shape = PartialShape{3, 11, 7, Dimension(3, 5)};
+    auto I1 = make_shared<op::Parameter>(element::f32, input1_shape);
+    auto I2 = make_shared<op::Parameter>(element::f32, input2_shape);
+    auto O = make_shared<op::v7::Einsum>(OutputVector{I1, I2}, equation);
+    ASSERT_EQ(O->get_element_type(), element::f32);
+    ASSERT_TRUE(O->get_output_partial_shape(0).same_scheme(out_shape));
+}
+
+TEST(type_prop, einsum_implicitmode_innerprod)
+{
+    // the following equation is equivalent to "i,i->"
+    std::string equation = "i,i";
+    const auto input1_shape = PartialShape{11};
+    const auto input2_shape = PartialShape{Dimension(1, 20)};
+    const auto out_shape = PartialShape{};
+    auto I1 = make_shared<op::Parameter>(element::f32, input1_shape);
+    auto I2 = make_shared<op::Parameter>(element::f32, input2_shape);
+    auto O = make_shared<op::v7::Einsum>(OutputVector{I1, I2}, equation);
+    ASSERT_EQ(O->get_element_type(), element::f32);
+    ASSERT_TRUE(O->get_output_partial_shape(0).same_scheme(out_shape));
+}
+
 TEST(type_prop, einsum_dynamicrank_multimatmul)
 {
     std::string equation = "ab,bcd,bc->ca";

From 4fb2d83db9b4bd524498800e50de9cb9bc3e669e Mon Sep 17 00:00:00 2001
From: Patryk Elszkowski <patryk.elszkowski@intel.com>
Date: Wed, 16 Jun 2021 06:09:57 +0200
Subject: [PATCH 097/178] use RTTI declaration macro and add visitor API test
 (#6147)

---
 ngraph/core/include/ngraph/op/parameter.hpp |  5 ++--
 ngraph/core/src/op/parameter.cpp            |  2 +-
 ngraph/test/CMakeLists.txt                  |  3 +-
 ngraph/test/visitors/op/parameter.cpp       | 33 +++++++++++++++++++++
 4 files changed, 38 insertions(+), 5 deletions(-)
 create mode 100644 ngraph/test/visitors/op/parameter.cpp

diff --git a/ngraph/core/include/ngraph/op/parameter.hpp b/ngraph/core/include/ngraph/op/parameter.hpp
index 50de93b529b7a3..01b57f5b80f37c 100644
--- a/ngraph/core/include/ngraph/op/parameter.hpp
+++ b/ngraph/core/include/ngraph/op/parameter.hpp
@@ -21,8 +21,7 @@ namespace ngraph
             class NGRAPH_API Parameter : public op::Op
             {
             public:
-                static constexpr NodeTypeInfo type_info{"Parameter", 0};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
+                NGRAPH_RTTI_DECLARATION;
                 /// \brief Constructions a tensor-typed parameter node.
                 Parameter() = default;
                 /// \brief Constructions a tensor-typed parameter node.
@@ -56,7 +55,7 @@ namespace ngraph
             protected:
                 PartialShape m_partial_shape;
                 element::Type m_element_type;
-                bool m_is_relevant_to_shapes;
+                bool m_is_relevant_to_shapes{false};
             };
         } // namespace v0
         using v0::Parameter;
diff --git a/ngraph/core/src/op/parameter.cpp b/ngraph/core/src/op/parameter.cpp
index c8d33c49db966e..ef979410a864a7 100644
--- a/ngraph/core/src/op/parameter.cpp
+++ b/ngraph/core/src/op/parameter.cpp
@@ -11,7 +11,7 @@
 using namespace std;
 using namespace ngraph;
 
-constexpr NodeTypeInfo op::Parameter::type_info;
+NGRAPH_RTTI_DEFINITION(op::v0::Parameter, "Parameter", 0);
 
 op::Parameter::Parameter(const element::Type& element_type, const PartialShape& pshape)
     : m_partial_shape(pshape)
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index b9e65209bb8367..2850bd9d2e6508 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -252,6 +252,7 @@ set(SRC
     visitors/op/normalize_l2.cpp
     visitors/op/one_hot.cpp
     visitors/op/pad.cpp
+    visitors/op/parameter.cpp
     visitors/op/prior_box.cpp
     visitors/op/proposal.cpp
     visitors/op/psroi_pooling.cpp
@@ -595,4 +596,4 @@ target_include_directories(unit-test PRIVATE ${FRONTEND_INCLUDE_PATH})
 target_link_libraries(unit-test PRIVATE frontend_manager)
 
 add_subdirectory(frontend)
-### END FRONTEND ###
\ No newline at end of file
+### END FRONTEND ###
diff --git a/ngraph/test/visitors/op/parameter.cpp b/ngraph/test/visitors/op/parameter.cpp
new file mode 100644
index 00000000000000..991bf6a3e9e5d5
--- /dev/null
+++ b/ngraph/test/visitors/op/parameter.cpp
@@ -0,0 +1,33 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+
+#include "ngraph/ngraph.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+#include "ngraph/opsets/opset1.hpp"
+#include "ngraph/opsets/opset3.hpp"
+#include "ngraph/opsets/opset4.hpp"
+#include "ngraph/opsets/opset5.hpp"
+
+#include "util/visitor.hpp"
+
+using namespace ngraph;
+using ngraph::test::NodeBuilder;
+
+TEST(attributes, parameter_op)
+{
+    NodeBuilder::get_ops().register_factory<opset1::Parameter>();
+    auto parameter =
+        std::make_shared<op::Parameter>(element::f32, PartialShape{Dimension{1}, Dimension{4}});
+
+    NodeBuilder builder(parameter);
+    auto g_parameter = as_type_ptr<opset1::Parameter>(builder.create());
+
+    const auto expected_attr_count = 2;
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
+
+    EXPECT_EQ(g_parameter->get_partial_shape(), parameter->get_partial_shape());
+    EXPECT_EQ(g_parameter->get_element_type(), parameter->get_element_type());
+}

From c4274d4c32def52189ee8fd5f92d663c9fa00321 Mon Sep 17 00:00:00 2001
From: Jan Iwaszkiewicz <jan.iwaszkiewicz@intel.com>
Date: Wed, 16 Jun 2021 06:36:06 +0200
Subject: [PATCH 098/178] Change version of pybind (#6145)

---
 ngraph/python/CMakeLists.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ngraph/python/CMakeLists.txt b/ngraph/python/CMakeLists.txt
index 501bfffb4b2feb..58d30c300fdf3b 100644
--- a/ngraph/python/CMakeLists.txt
+++ b/ngraph/python/CMakeLists.txt
@@ -20,7 +20,7 @@ include(FetchContent)
 FetchContent_Declare(
     pybind11
     GIT_REPOSITORY "https://github.com/pybind/pybind11.git"
-    GIT_TAG        "v2.5.0"
+    GIT_TAG        "v2.6.2"
 )
 
 FetchContent_GetProperties(pybind11)

From 59fd456ba2b7dc9e2d5bc8bdf72e891ec811dc3f Mon Sep 17 00:00:00 2001
From: Patryk Elszkowski <patryk.elszkowski@intel.com>
Date: Wed, 16 Jun 2021 06:37:37 +0200
Subject: [PATCH 099/178] minor fixes in Parameter spec (#6146)

* minor fixes in Parameter spec

* expand supported type list and add output shape info
---
 docs/ops/infrastructure/Parameter_1.md | 23 ++++++++++++++++-------
 1 file changed, 16 insertions(+), 7 deletions(-)

diff --git a/docs/ops/infrastructure/Parameter_1.md b/docs/ops/infrastructure/Parameter_1.md
index 807a606a375036..879880002e6319 100644
--- a/docs/ops/infrastructure/Parameter_1.md
+++ b/docs/ops/infrastructure/Parameter_1.md
@@ -11,18 +11,27 @@
 * *element_type*
 
   * **Description**: the type of element of output tensor
-  * **Range of values**: u8, u16, u32, u64, i8, i16, i32, i64, f16, f32, boolean, bf16
-  * **Type**: string
+  * **Range of values**: u1, u4, u8, u16, u32, u64, i4, i8, i16, i32, i64, f16, f32, boolean, bf16
+  * **Type**: `string`
   * **Default value**: None
-  * **Required**: *Yes*
+  * **Required**: *yes*
 
 * *shape*
 
   * **Description**: the shape of the output tensor
-  * **Range of values**: list of non-negative integers, empty list is allowed that means 0D or scalar tensor
-  * **Type**: int[]
+  * **Range of values**: list of non-negative integers, empty list is allowed, which means 0D or scalar tensor
+  * **Type**: `int[]`
   * **Default value**: None
-  * **Required**: *Yes*
+  * **Required**: *yes*
+
+
+**Outputs**
+
+* **1**: Output tensor of type *T* and shape equal to *shape* attribute.
+
+**Types**
+
+* *T*: any type from *element type* values.
 
 **Example**
 
@@ -38,4 +47,4 @@
         </port>
     </output>
 </layer>
-```
\ No newline at end of file
+```

From 8c68f8b43c1cc0d56f767446550d032033edf79a Mon Sep 17 00:00:00 2001
From: Gabriele Galiero Casay <gabriele.galiero.casay@intel.com>
Date: Wed, 16 Jun 2021 08:01:57 +0200
Subject: [PATCH 100/178] Revise reference implementation for ReduceMin
 operation (#5797)

* Revise reference implementation for ReduceMin operation

* Refactor backend unit tests

 * Move tests with zero dims to op_eval

* Fix code style

* Added minor changes

 * Replace CoordinateTransform for CoordinateTransformBasic
 * Added constant expression to set keep_dims as false

* Add const qualifier to local variables

* Use host tensor to retrieve and normalize axes

* Re-arrange unit tests in manifest
---
 .../include/ngraph/runtime/reference/min.hpp  |  35 +--
 ngraph/core/src/op/min.cpp                    |   9 +-
 ngraph/test/CMakeLists.txt                    |   1 +
 ngraph/test/backend/reduce_min.in.cpp         | 241 ----------------
 ngraph/test/op_eval/reduce_min.cpp            | 258 ++++++++++++++++++
 ngraph/test/runtime/ie/unit_test.manifest     |  28 +-
 6 files changed, 286 insertions(+), 286 deletions(-)
 create mode 100644 ngraph/test/op_eval/reduce_min.cpp

diff --git a/ngraph/core/reference/include/ngraph/runtime/reference/min.hpp b/ngraph/core/reference/include/ngraph/runtime/reference/min.hpp
index 9cfa7c16c4d2c9..ada3198834503d 100644
--- a/ngraph/core/reference/include/ngraph/runtime/reference/min.hpp
+++ b/ngraph/core/reference/include/ngraph/runtime/reference/min.hpp
@@ -6,6 +6,7 @@
 
 #include <cmath>
 #include <limits>
+#include <numeric>
 
 #include "ngraph/coordinate_transform.hpp"
 #include "ngraph/shape_util.hpp"
@@ -21,34 +22,34 @@ namespace ngraph
         namespace reference
         {
             template <typename T>
-            void min(const T* arg,
-                     T* out,
-                     const Shape& in_shape,
-                     const AxisSet& reduction_axes,
-                     const bool keep_dims)
+            void min(const T* arg, T* out, const Shape& in_shape, const AxisSet& reduction_axes)
             {
                 T minval = std::numeric_limits<T>::has_infinity ? std::numeric_limits<T>::infinity()
                                                                 : std::numeric_limits<T>::max();
 
-                const auto out_shape = reduce(in_shape, reduction_axes, keep_dims);
-                CoordinateTransform output_transform(out_shape);
+                constexpr bool dont_keep_dims_in_output = false;
+                const auto out_shape = reduce(in_shape, reduction_axes, dont_keep_dims_in_output);
+                std::fill(out, out + shape_size(out_shape), minval);
 
-                for (const Coordinate& output_coord : output_transform)
-                {
-                    out[output_transform.index(output_coord)] = minval;
-                }
-
-                CoordinateTransform input_transform(in_shape);
+                const auto in_strides = row_major_strides(in_shape);
+                const auto out_strides = row_major_strides(out_shape);
 
+                CoordinateTransformBasic input_transform(in_shape);
                 for (const Coordinate& input_coord : input_transform)
                 {
-                    Coordinate output_coord = reduce(input_coord, reduction_axes, keep_dims);
+                    const Coordinate output_coord =
+                        reduce(input_coord, reduction_axes, dont_keep_dims_in_output);
+
+                    const size_t in_idx = std::inner_product(
+                        input_coord.begin(), input_coord.end(), in_strides.begin(), 0);
+                    const size_t out_idx = std::inner_product(
+                        output_coord.begin(), output_coord.end(), out_strides.begin(), 0);
 
-                    T x = arg[input_transform.index(input_coord)];
-                    T min = out[output_transform.index(output_coord)];
+                    const T x = arg[in_idx];
+                    const T min = out[out_idx];
                     if (x < min)
                     {
-                        out[output_transform.index(output_coord)] = x;
+                        out[out_idx] = x;
                     }
                 }
             }
diff --git a/ngraph/core/src/op/min.cpp b/ngraph/core/src/op/min.cpp
index 37f3512413b53e..36b248ed69b47e 100644
--- a/ngraph/core/src/op/min.cpp
+++ b/ngraph/core/src/op/min.cpp
@@ -5,6 +5,7 @@
 #include "ngraph/op/min.hpp"
 #include <ngraph/validation_util.hpp>
 #include "itt.hpp"
+#include "ngraph/op/util/evaluate_helpers.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
 #include "ngraph/runtime/reference/min.hpp"
 #include "ngraph/shape_util.hpp"
@@ -22,7 +23,7 @@ namespace minop
     {
         out->set_shape(reduce(arg->get_shape(), axes, keep_dims));
         runtime::reference::min(
-            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes, keep_dims);
+            arg->get_data_ptr<ET>(), out->get_data_ptr<ET>(), arg->get_shape(), axes);
         return true;
     }
 
@@ -69,7 +70,11 @@ bool op::v1::ReduceMin::evaluate(const HostTensorVector& outputs,
     NGRAPH_OP_SCOPE(v1_ReduceMin_evaluate);
     NGRAPH_CHECK(validate_host_tensor_vector(inputs, 2));
     NGRAPH_CHECK(validate_host_tensor_vector(outputs, 1));
-    return minop::evaluate_min(inputs[0], outputs[0], get_reduction_axes(), get_keep_dims());
+
+    const auto reduction_axes = get_normalized_axes_from_tensor(
+        inputs[1], inputs[0]->get_partial_shape().rank(), get_friendly_name());
+
+    return minop::evaluate_min(inputs[0], outputs[0], reduction_axes, get_keep_dims());
 }
 
 bool op::v1::ReduceMin::has_evaluate() const
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 2850bd9d2e6508..ca45c965c9ffa3 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -72,6 +72,7 @@ set(SRC
     op_eval/mish.cpp
     op_eval/non_zero.cpp
     op_eval/reduce_max.cpp
+    op_eval/reduce_min.cpp
     op_eval/reduce_prod.cpp
     op_eval/reduce_sum.cpp
     op_eval/roi_align.cpp
diff --git a/ngraph/test/backend/reduce_min.in.cpp b/ngraph/test/backend/reduce_min.in.cpp
index 21565248d08098..d43c96b17f7dc4 100644
--- a/ngraph/test/backend/reduce_min.in.cpp
+++ b/ngraph/test/backend/reduce_min.in.cpp
@@ -124,100 +124,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_min_matrix_rows_int32)
     EXPECT_EQ((vector<int32_t>{1, 3, 5}), read_vector<int32_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_min_matrix_rows_zero)
-{
-    Shape shape_a{3, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity(),
-                             std::numeric_limits<float>::infinity(),
-                             std::numeric_limits<float>::infinity()}),
-              read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_min_matrix_cols_zero)
-{
-    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
-    Shape shape_a{0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity(),
-                             std::numeric_limits<float>::infinity()}),
-              read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_min_vector_zero)
-{
-    Shape shape_a{0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity()}), read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_min_matrix_to_scalar_zero_by_zero)
-{
-    Shape shape_a{0, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity()}), read_vector<float>(result));
-}
-
 NGRAPH_TEST(${BACKEND_NAME}, reduce_min_3d_to_matrix_most_sig)
 {
     Shape shape_a{3, 3, 3};
@@ -334,33 +240,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_min_3d_to_scalar_int32)
     EXPECT_EQ((vector<int32_t>{1}), read_vector<int32_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_min_3d_eliminate_zero_dim)
-{
-    Shape shape_a{3, 0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, false), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-
-    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
-    // right value.
-    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
-
-    float inf = std::numeric_limits<float>::infinity();
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{inf, inf, inf, inf, inf, inf}), read_vector<float>(result));
-}
-
 // ----------------------- keep dims = true ----------------------- //
 
 NGRAPH_TEST(${BACKEND_NAME}, reduce_min_keep_to_scalar)
@@ -469,99 +348,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_min_keep_matrix_rows_int32)
     EXPECT_EQ((vector<int32_t>{1, 3, 5}), read_vector<int32_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_min_keep_matrix_rows_zero)
-{
-    Shape shape_a{3, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity(),
-                             std::numeric_limits<float>::infinity(),
-                             std::numeric_limits<float>::infinity()}),
-              read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_min_keep_matrix_cols_zero)
-{
-    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
-    Shape shape_a{0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3, 3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity(),
-                             std::numeric_limits<float>::infinity()}),
-              read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_min_keep_vector_zero)
-{
-    Shape shape_a{0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity()}), read_vector<float>(result));
-}
-
-NGRAPH_TEST(${BACKEND_NAME}, reduce_min_keep_matrix_to_scalar_zero_by_zero)
-{
-    Shape shape_a{0, 0};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{1, 1};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-    copy_data(result, vector<float>({3}));
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity()}), read_vector<float>(result));
-}
 
 NGRAPH_TEST(${BACKEND_NAME}, reduce_min_keep_3d_to_matrix_most_sig)
 {
@@ -679,33 +465,6 @@ NGRAPH_TEST(${BACKEND_NAME}, reduce_min_keep_3d_to_scalar_int32)
     EXPECT_EQ((vector<int32_t>{1}), read_vector<int32_t>(result));
 }
 
-NGRAPH_TEST(${BACKEND_NAME}, reduce_min_keep_3d_eliminate_zero_dim)
-{
-    Shape shape_a{3, 0, 2};
-    auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    Shape shape_rt{3, 1, 2};
-    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
-    auto f =
-        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, true), ParameterVector{A});
-
-    auto backend = runtime::Backend::create("${BACKEND_NAME}");
-
-    // Create some tensors for input/output
-    auto a = backend->create_tensor(element::f32, shape_a);
-    copy_data(a, vector<float>{});
-    auto result = backend->create_tensor(element::f32, shape_rt);
-
-    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
-    // right value.
-    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
-
-    float inf = std::numeric_limits<float>::infinity();
-
-    auto handle = backend->compile(f);
-    handle->call_with_validate({result}, {a});
-    EXPECT_EQ((vector<float>{inf, inf, inf, inf, inf, inf}), read_vector<float>(result));
-}
-
 // Dynamic
 
 NGRAPH_TEST(${BACKEND_NAME}, reduce_min_matrix_columns_dynamic)
diff --git a/ngraph/test/op_eval/reduce_min.cpp b/ngraph/test/op_eval/reduce_min.cpp
new file mode 100644
index 00000000000000..cc74b0b6f591e8
--- /dev/null
+++ b/ngraph/test/op_eval/reduce_min.cpp
@@ -0,0 +1,258 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/test_control.hpp"
+#include "util/all_close.hpp"
+#include "util/all_close_f.hpp"
+#include "util/ndarray.hpp"
+
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+
+TEST(op_eval, reduce_min_matrix_rows_zero)
+{
+    Shape shape_a{3, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity(),
+                             std::numeric_limits<float>::infinity(),
+                             std::numeric_limits<float>::infinity()}),
+              read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_min_matrix_cols_zero)
+{
+    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
+    Shape shape_a{0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity(),
+                             std::numeric_limits<float>::infinity()}),
+              read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_min_vector_zero)
+{
+    Shape shape_a{0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity()}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_min_matrix_to_scalar_zero_by_zero)
+{
+    Shape shape_a{0, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity()}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_min_3d_eliminate_zero_dim)
+{
+    Shape shape_a{3, 0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, false), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+
+    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
+    // right value.
+    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
+
+    float inf = std::numeric_limits<float>::infinity();
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{inf, inf, inf, inf, inf, inf}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_min_keep_matrix_rows_zero)
+{
+    Shape shape_a{3, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity(),
+                             std::numeric_limits<float>::infinity(),
+                             std::numeric_limits<float>::infinity()}),
+              read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_min_keep_matrix_cols_zero)
+{
+    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
+    Shape shape_a{0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3, 3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity(),
+                             std::numeric_limits<float>::infinity()}),
+              read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_min_keep_vector_zero)
+{
+    Shape shape_a{0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 0);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity()}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_min_keep_matrix_to_scalar_zero_by_zero)
+{
+    Shape shape_a{0, 0};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{1, 1};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{2}, vector<int32_t>{0, 1});
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+    copy_data(result, vector<float>({3}));
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{std::numeric_limits<float>::infinity()}), read_vector<float>(result));
+}
+
+TEST(op_eval, reduce_min_keep_3d_eliminate_zero_dim)
+{
+    Shape shape_a{3, 0, 2};
+    auto A = make_shared<op::Parameter>(element::f32, shape_a);
+    Shape shape_rt{3, 1, 2};
+    auto axes = make_shared<op::Constant>(element::i32, Shape{}, 1);
+    auto f =
+        make_shared<Function>(make_shared<op::v1::ReduceMin>(A, axes, true), ParameterVector{A});
+
+    auto backend = runtime::Backend::create("INTERPRETER");
+
+    // Create some tensors for input/output
+    auto a = backend->create_tensor(element::f32, shape_a);
+    copy_data(a, vector<float>{});
+    auto result = backend->create_tensor(element::f32, shape_rt);
+
+    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
+    // right value.
+    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});
+
+    float inf = std::numeric_limits<float>::infinity();
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {a});
+    EXPECT_EQ((vector<float>{inf, inf, inf, inf, inf, inf}), read_vector<float>(result));
+}
diff --git a/ngraph/test/runtime/ie/unit_test.manifest b/ngraph/test/runtime/ie/unit_test.manifest
index a0a7c8735f58d3..adfebe2b57aac4 100644
--- a/ngraph/test/runtime/ie/unit_test.manifest
+++ b/ngraph/test/runtime/ie/unit_test.manifest
@@ -375,6 +375,8 @@ reduce_sum_keep_2d_to_scalar_int8
 reduce_sum_2d_to_scalar_int8
 reduce_product_to_scalar_int8
 reduce_product_keep_to_scalar_int8
+reduce_min_to_scalar_int8
+reduce_min_keep_to_scalar_int8
 # accuracy
 reduce_sum_keep_stable_acc
 reduce_sum_keep_3d_to_scalar_int32
@@ -441,34 +443,8 @@ onnx_dyn_shapes_model_tile_static
 gather_4d_indices_axis_0_uint8
 tensor_constant_with_op
 constant_equality_bool
-reduce_min_matrix_columns
-reduce_min_matrix_rows
-reduce_min_matrix_rows_int32
-reduce_min_3d_to_matrix_most_sig
-reduce_min_3d_to_matrix_least_sig
-reduce_min_keep_matrix_columns
-reduce_min_keep_matrix_rows
-reduce_min_keep_matrix_rows_int32
-reduce_min_keep_3d_to_matrix_most_sig
-reduce_min_keep_3d_to_matrix_least_sig
-reduce_min_matrix_columns_dynamic
-reduce_min_matrix_rows_dynamic
-reduce_min_keep_matrix_columns_dynamic
-reduce_min_keep_matrix_rows_dynamic
 
 # zero dimension / result mismatch
-reduce_min_to_scalar_int8
-reduce_min_matrix_rows_zero
-reduce_min_matrix_cols_zero
-reduce_min_vector_zero
-reduce_min_matrix_to_scalar_zero_by_zero
-reduce_min_3d_eliminate_zero_dim
-reduce_min_keep_to_scalar_int8
-reduce_min_keep_matrix_rows_zero
-reduce_min_keep_matrix_cols_zero
-reduce_min_keep_vector_zero
-reduce_min_keep_matrix_to_scalar_zero_by_zero
-reduce_min_keep_3d_eliminate_zero_dim
 reduce_mean_to_scalar_int8
 reduce_mean_matrix_rows_int32
 reduce_mean_keep_to_scalar_int8

From b36c7736da8fd4f08cdc3f4d6d81af653622cf31 Mon Sep 17 00:00:00 2001
From: Evgenya Stepyreva <evgenya.stepyreva@intel.com>
Date: Wed, 16 Jun 2021 09:06:19 +0300
Subject: [PATCH 101/178] Deprecated legacy maximum_value method (#6178)

* Deprecated legacy maximum_value method since it is less reliable than the existing one

* Style and doxygen \deprecated comment
---
 .../common/src/ngraph/operations/static_shape_broadcast.cpp    | 1 -
 inference-engine/src/vpu/common/src/ngraph/utilities.cpp       | 1 -
 ngraph/core/include/ngraph/validation_util.hpp                 | 3 +++
 3 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/inference-engine/src/vpu/common/src/ngraph/operations/static_shape_broadcast.cpp b/inference-engine/src/vpu/common/src/ngraph/operations/static_shape_broadcast.cpp
index a52bf97e707dd4..1490bd408e116d 100644
--- a/inference-engine/src/vpu/common/src/ngraph/operations/static_shape_broadcast.cpp
+++ b/inference-engine/src/vpu/common/src/ngraph/operations/static_shape_broadcast.cpp
@@ -8,7 +8,6 @@
 #include "vpu/utils/error.hpp"
 
 #include "ngraph/opsets/opset3.hpp"
-#include "ngraph/evaluator.hpp"
 #include <ngraph/validation_util.hpp>
 
 namespace ngraph { namespace vpu { namespace op {
diff --git a/inference-engine/src/vpu/common/src/ngraph/utilities.cpp b/inference-engine/src/vpu/common/src/ngraph/utilities.cpp
index 47afb6a83a6a58..23576617f67994 100644
--- a/inference-engine/src/vpu/common/src/ngraph/utilities.cpp
+++ b/inference-engine/src/vpu/common/src/ngraph/utilities.cpp
@@ -6,7 +6,6 @@
 
 #include "ngraph/opsets/opset3.hpp"
 #include "ngraph/opsets/opset5.hpp"
-#include "ngraph/evaluator.hpp"
 
 #include <numeric>
 
diff --git a/ngraph/core/include/ngraph/validation_util.hpp b/ngraph/core/include/ngraph/validation_util.hpp
index d8922f132a9cc6..de81ebfd171bb8 100644
--- a/ngraph/core/include/ngraph/validation_util.hpp
+++ b/ngraph/core/include/ngraph/validation_util.hpp
@@ -234,6 +234,9 @@ namespace ngraph
     /// \brief Try to compute the maximum value of value
     /// \return (true, max_value) if can be determined, or (false, numeric_limits<uint64_t>::max())
     /// if not.
+    /// \deprecated Use evaluate_upper_bound instead
+    NGRAPH_DEPRECATED(
+        "Use evaluate_upper_bound: it would return HostTensorPtr to the value instead of a pair")
     NGRAPH_API std::pair<bool, uint64_t> maximum_value(const Output<Node>& value);
 
     /// \brief Evaluates outputs, treating values in value_map as already computed. value_map is

From 2f81968a3106dbd8f4adaa7bcf8bec778c84ce1c Mon Sep 17 00:00:00 2001
From: Vladimir Paramuzov <vladimir.paramuzov@intel.com>
Date: Wed, 16 Jun 2021 09:27:16 +0300
Subject: [PATCH 102/178] [IE CLDNN] Introduced new runtime API (#5417)

---
 .../src/cldnn_engine/CMakeLists.txt           |    2 +-
 .../src/cldnn_engine/cldnn_common_utils.h     |    2 +-
 .../src/cldnn_engine/cldnn_config.h           |    2 +-
 .../src/cldnn_engine/cldnn_custom_layer.h     |    2 +-
 .../src/cldnn_engine/cldnn_engine.cpp         |   19 +-
 .../src/cldnn_engine/cldnn_engine.h           |    4 +-
 .../cldnn_engine/cldnn_executable_network.cpp |   11 +-
 .../src/cldnn_engine/cldnn_graph.cpp          |   34 +-
 .../src/cldnn_engine/cldnn_graph.h            |    6 +-
 .../src/cldnn_engine/cldnn_infer_request.cpp  |  225 +-
 .../src/cldnn_engine/cldnn_infer_request.h    |    9 +-
 .../src/cldnn_engine/cldnn_program.cpp        |    4 +-
 .../src/cldnn_engine/cldnn_program.h          |   14 +-
 .../src/cldnn_engine/cldnn_remote_context.cpp |  109 +-
 .../src/cldnn_engine/cldnn_remote_context.h   |   39 +-
 .../src/cldnn_engine/ops/batch_to_space.cpp   |    2 +-
 .../src/cldnn_engine/ops/broadcast.cpp        |    6 +-
 .../src/cldnn_engine/ops/concat.cpp           |    2 +-
 .../src/cldnn_engine/ops/constant.cpp         |    9 +-
 .../src/cldnn_engine/ops/convert.cpp          |    2 +-
 .../src/cldnn_engine/ops/convolution.cpp      |   10 +-
 .../cldnn_engine/ops/ctc_greedy_decoder.cpp   |   10 +-
 .../src/cldnn_engine/ops/cum_sum.cpp          |    2 +-
 .../src/cldnn_engine/ops/custom.cpp           |    4 +-
 .../src/cldnn_engine/ops/depth_to_space.cpp   |    2 +-
 .../src/cldnn_engine/ops/detection_output.cpp |    2 +-
 .../src/cldnn_engine/ops/eltwise.cpp          |    8 +-
 .../src/cldnn_engine/ops/embedding_bag.cpp    |    4 +-
 .../ops/extract_image_patches.cpp             |    2 +-
 .../src/cldnn_engine/ops/fake_quantize.cpp    |    2 +-
 .../src/cldnn_engine/ops/gather tree.cpp      |    4 +-
 .../src/cldnn_engine/ops/gather.cpp           |    4 +-
 .../src/cldnn_engine/ops/gather_nd.cpp        |    2 +-
 inference-engine/src/cldnn_engine/ops/grn.cpp |    2 +-
 .../src/cldnn_engine/ops/interpolate.cpp      |    2 +-
 inference-engine/src/cldnn_engine/ops/lrn.cpp |    2 +-
 .../src/cldnn_engine/ops/matmul.cpp           |   10 +-
 inference-engine/src/cldnn_engine/ops/mvn.cpp |    3 +-
 .../cldnn_engine/ops/non_max_suppression.cpp  |   12 +-
 .../src/cldnn_engine/ops/normalize_l2.cpp     |    8 +-
 .../src/cldnn_engine/ops/one_hot.cpp          |    2 +-
 inference-engine/src/cldnn_engine/ops/pad.cpp |    2 +-
 .../src/cldnn_engine/ops/parameter.cpp        |   12 +-
 .../src/cldnn_engine/ops/pooling.cpp          |    2 +-
 .../src/cldnn_engine/ops/prior_box.cpp        |    2 +-
 .../src/cldnn_engine/ops/proposal.cpp         |    6 +-
 .../src/cldnn_engine/ops/reduce.cpp           |    6 +-
 .../src/cldnn_engine/ops/region_yolo.cpp      |    2 +-
 .../src/cldnn_engine/ops/reorg_yolo.cpp       |    2 +-
 .../src/cldnn_engine/ops/reshape.cpp          |    4 +-
 .../src/cldnn_engine/ops/result.cpp           |    2 +-
 .../src/cldnn_engine/ops/reverse_sequence.cpp |    2 +-
 inference-engine/src/cldnn_engine/ops/rnn.cpp |   12 +-
 .../src/cldnn_engine/ops/roi_pooling.cpp      |    2 +-
 .../ops/scatter_elements_update.cpp           |    2 +-
 .../cldnn_engine/ops/scatter_nd_update.cpp    |    2 +-
 .../src/cldnn_engine/ops/scatter_update.cpp   |    2 +-
 .../src/cldnn_engine/ops/select.cpp           |    6 +-
 .../src/cldnn_engine/ops/shuffle_channels.cpp |    2 +-
 .../src/cldnn_engine/ops/softmax.cpp          |    4 +-
 .../src/cldnn_engine/ops/space_to_batch.cpp   |    2 +-
 .../src/cldnn_engine/ops/space_to_depth.cpp   |    2 +-
 .../src/cldnn_engine/ops/split.cpp            |    2 +-
 .../src/cldnn_engine/ops/strided_slice.cpp    |    6 +-
 .../src/cldnn_engine/ops/tensor_iterator.cpp  |   17 +-
 .../src/cldnn_engine/ops/tile.cpp             |    2 +-
 .../src/cldnn_engine/ops/topk.cpp             |    6 +-
 .../src/cldnn_engine/ops/transpose.cpp        |    2 +-
 .../src/cldnn_engine/ops/unary.cpp            |    2 +-
 .../functional/plugin/gpu/CMakeLists.txt      |    2 +-
 .../thirdparty/clDNN/CMakeLists.txt           |   58 +-
 .../thirdparty/clDNN/api/cldnn.hpp            |  244 --
 .../clDNN/api/{ => cldnn/graph}/network.hpp   |   67 +-
 .../clDNN/api/{ => cldnn/graph}/program.hpp   |   25 +-
 .../clDNN/api/{ => cldnn/graph}/topology.hpp  |   25 +-
 .../api/{ => cldnn/primitives}/activation.hpp |    0
 .../{ => cldnn/primitives}/arg_max_min.hpp    |    0
 .../primitives}/average_unpooling.hpp         |    0
 .../{ => cldnn/primitives}/batch_to_space.hpp |    0
 .../primitives}/binary_convolution.hpp        |    0
 .../api/{ => cldnn/primitives}/border.hpp     |    0
 .../api/{ => cldnn/primitives}/broadcast.hpp  |    0
 .../{ => cldnn/primitives}/concatenation.hpp  |    0
 .../api/{ => cldnn/primitives}/condition.hpp  |    2 +-
 .../{ => cldnn/primitives}/convolution.hpp    |    0
 .../clDNN/api/{ => cldnn/primitives}/crop.hpp |    0
 .../primitives}/ctc_greedy_decoder.hpp        |    0
 .../api/{ => cldnn/primitives}/cum_sum.hpp    |    0
 .../primitives}/custom_gpu_primitive.hpp      |    2 +-
 .../clDNN/api/{ => cldnn/primitives}/data.hpp |    6 +-
 .../{ => cldnn/primitives}/deconvolution.hpp  |    0
 .../{ => cldnn/primitives}/depth_to_space.hpp |    0
 .../primitives}/detection_output.hpp          |    0
 .../api/{ => cldnn/primitives}/eltwise.hpp    |    0
 .../{ => cldnn/primitives}/embedding_bag.hpp  |    0
 .../primitives}/extract_image_patches.hpp     |    0
 .../primitives}/fully_connected.hpp           |    0
 .../cldnn/primitives}/fused_conv_eltwise.hpp  |    4 +-
 .../api/{ => cldnn/primitives}/gather.hpp     |    0
 .../api/{ => cldnn/primitives}/gather_nd.hpp  |   17 +-
 .../{ => cldnn/primitives}/gather_tree.hpp    |    0
 .../clDNN/api/{ => cldnn/primitives}/gemm.hpp |    0
 .../clDNN/api/{ => cldnn/primitives}/grn.hpp  |    0
 .../{ => cldnn/primitives}/input_layout.hpp   |    2 +-
 .../clDNN/api/{ => cldnn/primitives}/loop.hpp |    2 +-
 .../clDNN/api/{ => cldnn/primitives}/lrn.hpp  |    0
 .../clDNN/api/{ => cldnn/primitives}/lstm.hpp |    0
 .../{ => cldnn/primitives}/lstm_dynamic.hpp   |    0
 .../cldnn/primitives}/lstm_dynamic_input.hpp  |    2 +-
 .../primitives}/lstm_dynamic_timeloop.hpp     |    2 +-
 .../{ => cldnn/primitives}/max_unpooling.hpp  |    0
 .../{ => cldnn/primitives}/mutable_data.hpp   |    8 +-
 .../clDNN/api/{ => cldnn/primitives}/mvn.hpp  |    0
 .../primitives}/non_max_suppression.hpp       |    0
 .../api/{ => cldnn/primitives}/normalize.hpp  |    0
 .../api/{ => cldnn/primitives}/one_hot.hpp    |    0
 .../api/{ => cldnn/primitives}/permute.hpp    |    0
 .../api/{ => cldnn/primitives}/pooling.hpp    |    0
 .../api/{ => cldnn/primitives}/primitive.hpp  |    6 +-
 .../api/{ => cldnn/primitives}/prior_box.hpp  |    0
 .../api/{ => cldnn/primitives}/proposal.hpp   |    0
 .../primitives}/pyramid_roi_align.hpp         |    0
 .../api/{ => cldnn/primitives}/quantize.hpp   |    0
 .../api/{ => cldnn/primitives}/reduce.hpp     |    0
 .../{ => cldnn/primitives}/region_yolo.hpp    |    0
 .../api/{ => cldnn/primitives}/reorder.hpp    |    2 +-
 .../api/{ => cldnn/primitives}/reorg_yolo.hpp |    0
 .../api/{ => cldnn/primitives}/resample.hpp   |    0
 .../api/{ => cldnn/primitives}/reshape.hpp    |    0
 .../primitives}/reverse_sequence.hpp          |    0
 .../{ => cldnn/primitives}/roi_pooling.hpp    |    0
 .../api/{ => cldnn/primitives}/scale.hpp      |    0
 .../primitives}/scatter_elements_update.hpp   |    0
 .../primitives}/scatter_nd_update.hpp         |    0
 .../{ => cldnn/primitives}/scatter_update.hpp |    0
 .../api/{ => cldnn/primitives}/select.hpp     |    0
 .../primitives}/shuffle_channels.hpp          |    0
 .../api/{ => cldnn/primitives}/softmax.hpp    |    0
 .../{ => cldnn/primitives}/space_to_batch.hpp |    0
 .../{ => cldnn/primitives}/space_to_depth.hpp |    0
 .../api/{ => cldnn/primitives}/split.hpp      |    0
 .../{ => cldnn/primitives}/strided_slice.hpp  |    0
 .../clDNN/api/{ => cldnn/primitives}/tile.hpp |    0
 .../runtime/compounds.hpp}                    |    3 +-
 .../clDNN/api/cldnn/runtime/device.hpp        |   24 +
 .../clDNN/api/cldnn/runtime/device_info.hpp   |   76 +
 .../clDNN/api/cldnn/runtime/device_query.hpp  |   29 +
 .../clDNN/api/cldnn/runtime/engine.hpp        |  150 ++
 .../cldnn/runtime/engine_configuration.hpp    |  110 +
 .../cldnn/runtime/error_handler.hpp}          |    5 +-
 .../cldnn/runtime/event.hpp}                  |   20 +-
 .../clDNN/api/cldnn/runtime/half.hpp          |   55 +
 .../clDNN/api/cldnn/runtime/kernel.hpp        |   24 +
 .../clDNN/api/cldnn/runtime/kernel_args.hpp   |  133 ++
 .../clDNN/api/{ => cldnn/runtime}/layout.hpp  |    4 +-
 .../clDNN/api/cldnn/runtime/memory.hpp        |  150 ++
 .../clDNN/api/cldnn/runtime/memory_caps.hpp   |   84 +
 .../cldnn/runtime/memory_pool.hpp}            |   81 +-
 .../api/{ => cldnn/runtime}/profiling.hpp     |    1 +
 .../clDNN/api/cldnn/runtime/stream.hpp        |   46 +
 .../clDNN/api/{ => cldnn/runtime}/tensor.hpp  |    7 +-
 .../clDNN/api/cldnn/runtime/utils.hpp         |  144 ++
 .../thirdparty/clDNN/api/device.hpp           |  161 --
 .../thirdparty/clDNN/api/engine.hpp           |  189 --
 .../thirdparty/clDNN/api/event.hpp            |   83 -
 .../thirdparty/clDNN/api/memory.hpp           |  278 ---
 .../thirdparty/clDNN/api/meta_utils.hpp       |   43 -
 .../clDNN/api_test_builds/CMakeLists.txt      |   38 +-
 .../clDNN/kernel_selector/CMakeLists.txt      |    1 +
 .../kernel_selector/common/tensor_type.h      |    2 +
 .../activation/activation_kernel_base.cpp     |    2 +-
 .../arg_max_min/arg_max_min_kernel_axis.cpp   |    2 +-
 .../binary_convolution_kernel_base.cpp        |    4 +-
 .../concatenation_kernel_base.cpp             |   14 +-
 .../concatenation_kernel_fs_b_yx_fsv32.cpp    |   14 +-
 .../concatenation_kernel_ref.cpp              |    6 +-
 .../convolution/convolution_kernel_base.cpp   |   12 +-
 .../ctc_greedy_decoder_kernel_base.cpp        |    2 +-
 .../cum_sum/cum_sum_kernel_partial_sum.cpp    |   12 +-
 .../deconvolution_kernel_base.cpp             |    2 +-
 .../eltwise/eltwise_kernel_b_fs_yx_fsv16.cpp  |    8 +-
 .../eltwise/eltwise_kernel_b_fs_yx_fsv4.cpp   |    8 +-
 .../eltwise/eltwise_kernel_base.cpp           |    8 +-
 .../eltwise/eltwise_kernel_fs_b_yx_fsv32.cpp  |    9 +-
 ...se_kernel_mixed_byxf_and_fs_b_yx_fsv32.cpp |    8 +-
 .../eltwise/eltwise_kernel_vload8.cpp         |    8 +-
 .../fused_conv_eltwise_kernel_base.cpp        |    6 +-
 .../gather/gather_nd_kernel_ref.cpp           |   16 +-
 .../gather/gather_nd_kernel_ref.h             |   16 +-
 .../gather/gather_nd_kernel_selector.cpp      |   16 +-
 .../gather/gather_nd_kernel_selector.h        |   16 +-
 .../lstm/lstm_elt_kernel_base.cpp             |   10 +-
 .../lstm/lstm_gemm_kernel_base.cpp            |   16 +-
 ...m_gemv_gpu_subgroup1x64_bfyx_ff_simd16.cpp |    2 +-
 ...m_gemv_gpu_subgroup1x64_bfyx_hh_simd16.cpp |    2 +-
 .../lstm_dynamic_input_bfyx_opt.cpp           |    6 +-
 .../lstm_dynamic_input_kernel_base.cpp        |   14 +-
 .../lstm_dynamic_timeloop_kernel_base.cpp     |   22 +-
 .../max_unpooling_kernel_base.cpp             |    2 +-
 .../mvn/mvn_kernel_b_fs_yx_fsv16_imad.cpp     |   30 +-
 .../normalize/normalize_kernel_base.cpp       |    2 +-
 .../pooling/pooling_kernel_base.cpp           |    2 +-
 .../quantize/quantize_kernel_base.cpp         |    8 +-
 .../reorder/reorder_biplanar_nv12.cpp         |    2 +-
 .../reorder/reorder_kernel_base.cpp           |    6 +-
 .../reshape/reshape_kernel_ref.cpp            |    8 +-
 .../roi_pooling/roi_pooling_kernel_base.cpp   |    4 +-
 .../select/select_kernel_base.cpp             |    8 +-
 .../clDNN/kernel_selector/core/auto_tuner.cpp |    4 +-
 .../clDNN/kernel_selector/core/auto_tuner.h   |    2 +-
 .../core/cl_kernels/gather_nd_ref.cl          |   14 +-
 .../core/cl_kernels/max_unpooling_gpu_ref.cl  |    4 +-
 .../core/common/kernel_base_opencl.cpp        |    8 +-
 .../core}/device_cache_reader.cpp             |   27 +-
 .../core/device_cache_reader.h                |   14 +
 .../kernel_selector/core/kernel_selector.cpp  |    9 +-
 .../core/kernel_selector_common.h             |  115 +-
 .../thirdparty/clDNN/runtime/CMakeLists.txt   |   79 +
 .../cldnn_itt.h => runtime/cldnn_itt.hpp}     |    2 +-
 .../thirdparty/clDNN/runtime/device_query.cpp |   30 +
 .../thirdparty/clDNN/runtime/engine.cpp       |  160 ++
 .../thirdparty/clDNN/runtime/event.cpp        |   69 +
 .../{src/gpu => runtime}/kernels_cache.cpp    |  109 +-
 .../kernels_cache.hpp}                        |   55 +-
 .../clDNN/runtime/kernels_factory.cpp         |   22 +
 .../clDNN/runtime/kernels_factory.hpp         |   22 +
 .../thirdparty/clDNN/runtime/memory.cpp       |   36 +
 .../clDNN/{src => runtime}/memory_pool.cpp    |  221 +-
 .../gpu => runtime/ocl}/ocl_base_event.cpp    |   14 +-
 .../ocl/ocl_base_event.hpp}                   |   33 +-
 .../ocl/ocl_command_queues_builder.cpp}       |   35 +-
 .../ocl/ocl_command_queues_builder.hpp}       |   21 +-
 .../clDNN/runtime/ocl/ocl_common.hpp          |   30 +
 .../ocl/ocl_device.cpp}                       |  144 +-
 .../clDNN/runtime/ocl/ocl_device.hpp          |   42 +
 .../ocl/ocl_device_detector.cpp}              |  106 +-
 .../clDNN/runtime/ocl/ocl_device_detector.hpp |   30 +
 .../clDNN/runtime/ocl/ocl_engine.cpp          |  204 ++
 .../clDNN/runtime/ocl/ocl_engine.hpp          |   51 +
 .../clDNN/runtime/ocl/ocl_engine_factory.hpp  |   19 +
 .../ocl/ocl_events_pool.hpp}                  |   39 +-
 .../cl2_ext.hpp => runtime/ocl/ocl_ext.hpp}   |   41 +-
 .../clDNN/runtime/ocl/ocl_kernel.hpp          |   33 +
 .../clDNN/runtime/ocl/ocl_kernels_factory.cpp |   21 +
 .../clDNN/runtime/ocl/ocl_memory.cpp          |  373 +++
 .../clDNN/runtime/ocl/ocl_memory.hpp          |  128 ++
 .../clDNN/runtime/ocl/ocl_stream.cpp          |  428 ++++
 .../clDNN/runtime/ocl/ocl_stream.hpp          |   99 +
 .../gpu => runtime/ocl}/ocl_user_event.cpp    |    4 +-
 .../ocl/ocl_user_event.hpp}                   |   14 +-
 .../ocl/ocl_wrapper.hpp}                      |    7 +-
 .../thirdparty/clDNN/src/CMakeLists.txt       |   13 +-
 .../thirdparty/clDNN/src/activation.cpp       |    2 +-
 .../thirdparty/clDNN/src/arg_max_min.cpp      |    3 +-
 .../clDNN/src/average_unpooling.cpp           |    2 +-
 .../thirdparty/clDNN/src/batch_to_space.cpp   |    2 +-
 .../clDNN/src/binary_convolution.cpp          |    2 +-
 .../thirdparty/clDNN/src/border.cpp           |    2 +-
 .../thirdparty/clDNN/src/broadcast.cpp        |    2 +-
 .../thirdparty/clDNN/src/concatenation.cpp    |    2 +-
 .../thirdparty/clDNN/src/condition.cpp        |    8 +-
 .../thirdparty/clDNN/src/convolution.cpp      |    2 +-
 .../thirdparty/clDNN/src/crop.cpp             |    4 +-
 .../thirdparty/clDNN/src/cum_sum.cpp          |    2 +-
 .../thirdparty/clDNN/src/data.cpp             |   22 +-
 .../thirdparty/clDNN/src/deconvolution.cpp    |    2 +-
 .../clDNN/src/deformable_convolution.cpp      |    2 +-
 .../thirdparty/clDNN/src/depth_to_space.cpp   |    2 +-
 .../thirdparty/clDNN/src/detection_output.cpp |    2 +-
 .../thirdparty/clDNN/src/device.cpp           |   58 -
 .../thirdparty/clDNN/src/eltwise.cpp          |    2 +-
 .../thirdparty/clDNN/src/embedding_bag.cpp    |    2 +-
 .../thirdparty/clDNN/src/engine.cpp           |  298 ---
 .../thirdparty/clDNN/src/error_handler.cpp    |    2 +-
 .../thirdparty/clDNN/src/event.cpp            |  102 -
 .../clDNN/src/extract_image_patches.cpp       |    2 +-
 .../thirdparty/clDNN/src/fully_connected.cpp  |    2 +-
 .../clDNN/src/fused_conv_eltwise.cpp          |    2 +-
 .../thirdparty/clDNN/src/gather.cpp           |    2 +-
 .../thirdparty/clDNN/src/gather_nd.cpp        |   18 +-
 .../thirdparty/clDNN/src/gather_tree.cpp      |    2 +-
 .../thirdparty/clDNN/src/gemm.cpp             |    2 +-
 .../clDNN/src/gpu/activation_gpu.cpp          |   14 +-
 .../clDNN/src/gpu/arg_max_min_gpu.cpp         |   11 +-
 .../clDNN/src/gpu/average_unpooling_gpu.cpp   |   11 +-
 .../clDNN/src/gpu/batch_to_space_gpu.cpp      |    6 +-
 .../clDNN/src/gpu/binary_convolution_gpu.cpp  |   19 +-
 .../thirdparty/clDNN/src/gpu/border_gpu.cpp   |    6 +-
 .../clDNN/src/gpu/broadcast_gpu.cpp           |    6 +-
 .../clDNN/src/gpu/concatenation_gpu.cpp       |    7 +-
 .../clDNN/src/gpu/condition_gpu.cpp           |   32 +-
 .../clDNN/src/gpu/configuration.cpp           |   29 -
 .../thirdparty/clDNN/src/gpu/configuration.h  |   37 -
 .../clDNN/src/gpu/convolution_gpu.cpp         |   30 +-
 .../thirdparty/clDNN/src/gpu/crop_gpu.cpp     |    6 +-
 .../clDNN/src/gpu/ctc_greedy_decoder_gpu.cpp  |    6 +-
 .../thirdparty/clDNN/src/gpu/cum_sum_gpu.cpp  |    6 +-
 .../src/gpu/custom_gpu_primitive_gpu.cpp      |   78 +-
 .../clDNN/src/gpu/deconvolution_gpu.cpp       |   15 +-
 .../src/gpu/deformable_convolution_gpu.cpp    |   19 +-
 .../clDNN/src/gpu/depth_to_space_gpu.cpp      |    6 +-
 .../clDNN/src/gpu/detection_output_cpu.cpp    |   92 +-
 .../clDNN/src/gpu/device_cache_reader.h       |   27 -
 .../thirdparty/clDNN/src/gpu/device_info.h    |   58 -
 .../thirdparty/clDNN/src/gpu/eltwise_gpu.cpp  |   11 +-
 .../clDNN/src/gpu/embedding_bag_gpu.cpp       |    6 +-
 .../thirdparty/clDNN/src/gpu/events_waiter.h  |   26 -
 .../src/gpu/extract_image_patches_gpu.cpp     |    6 +-
 .../clDNN/src/gpu/fully_connected_gpu.cpp     |   19 +-
 .../clDNN/src/gpu/fused_conv_eltwise_gpu.cpp  |   15 +-
 .../thirdparty/clDNN/src/gpu/gather_gpu.cpp   |    6 +-
 .../clDNN/src/gpu/gather_nd_gpu.cpp           |   22 +-
 .../clDNN/src/gpu/gather_tree_gpu.cpp         |    7 +-
 .../thirdparty/clDNN/src/gpu/gemm_gpu.cpp     |    6 +-
 .../clDNN/src/gpu/generic_layer_gpu.cpp       |   88 +-
 .../thirdparty/clDNN/src/gpu/grn_gpu.cpp      |    6 +-
 .../thirdparty/clDNN/src/gpu/kernel.cpp       |  275 ---
 .../thirdparty/clDNN/src/gpu/kernel.h         |   96 -
 .../clDNN/src/gpu/kernel_runner.cpp           |   90 +-
 .../thirdparty/clDNN/src/gpu/kernel_runner.h  |   26 +-
 .../thirdparty/clDNN/src/gpu/loop_gpu.cpp     |   77 +-
 .../thirdparty/clDNN/src/gpu/lrn_gpu.cpp      |    6 +-
 .../clDNN/src/gpu/lstm_dynamic_input_gpu.cpp  |   19 +-
 .../src/gpu/lstm_dynamic_timeloop_gpu.cpp     |   25 +-
 .../thirdparty/clDNN/src/gpu/lstm_elt_gpu.cpp |   15 +-
 .../clDNN/src/gpu/lstm_gemm_gpu.cpp           |   23 +-
 .../clDNN/src/gpu/max_unpooling_gpu.cpp       |   21 +-
 .../thirdparty/clDNN/src/gpu/memory_gpu.cpp   |  310 ---
 .../thirdparty/clDNN/src/gpu/memory_gpu.h     |  176 --
 .../clDNN/src/gpu/mutable_data_gpu.cpp        |   12 +-
 .../thirdparty/clDNN/src/gpu/mvn_gpu.cpp      |    6 +-
 .../clDNN/src/gpu/non_max_suppression_cpu.cpp |  126 +-
 .../clDNN/src/gpu/normalize_gpu.cpp           |   13 +-
 .../thirdparty/clDNN/src/gpu/ocl_builder.h    |   38 -
 .../clDNN/src/gpu/ocl_queue_wrapper.cpp       |  158 --
 .../clDNN/src/gpu/ocl_queue_wrapper.h         |   96 -
 .../thirdparty/clDNN/src/gpu/ocl_toolkit.cpp  |  236 --
 .../thirdparty/clDNN/src/gpu/ocl_toolkit.h    |  147 --
 .../thirdparty/clDNN/src/gpu/one_hot_gpu.cpp  |    6 +-
 .../thirdparty/clDNN/src/gpu/permute_gpu.cpp  |    6 +-
 .../thirdparty/clDNN/src/gpu/pooling_gpu.cpp  |   13 +-
 .../clDNN/src/gpu/primitive_gpu_base.h        |  133 +-
 .../thirdparty/clDNN/src/gpu/proposal_gpu.cpp |   59 +-
 .../thirdparty/clDNN/src/gpu/push_pop_map.h   |   47 -
 .../clDNN/src/gpu/pyramid_roi_align_gpu.cpp   |    6 +-
 .../thirdparty/clDNN/src/gpu/quantize_gpu.cpp |   23 +-
 .../thirdparty/clDNN/src/gpu/reduce_gpu.cpp   |    6 +-
 .../clDNN/src/gpu/region_yolo_gpu.cpp         |    6 +-
 .../thirdparty/clDNN/src/gpu/register_gpu.hpp |  124 +-
 .../thirdparty/clDNN/src/gpu/reorder_gpu.cpp  |   14 +-
 .../clDNN/src/gpu/reorg_yolo_gpu.cpp          |    6 +-
 .../thirdparty/clDNN/src/gpu/resample_gpu.cpp |    6 +-
 .../thirdparty/clDNN/src/gpu/reshape_gpu.cpp  |    6 +-
 .../clDNN/src/gpu/reverse_sequence_gpu.cpp    |    6 +-
 .../clDNN/src/gpu/roi_pooling_gpu.cpp         |   21 +-
 .../thirdparty/clDNN/src/gpu/scale_gpu.cpp    |   16 +-
 .../src/gpu/scatter_elements_update_gpu.cpp   |    6 +-
 .../clDNN/src/gpu/scatter_nd_update_gpu.cpp   |    6 +-
 .../clDNN/src/gpu/scatter_update_gpu.cpp      |    6 +-
 .../thirdparty/clDNN/src/gpu/select_gpu.cpp   |    6 +-
 .../clDNN/src/gpu/shuffle_channels_gpu.cpp    |    6 +-
 .../thirdparty/clDNN/src/gpu/softmax_gpu.cpp  |    6 +-
 .../clDNN/src/gpu/space_to_batch_gpu.cpp      |    6 +-
 .../clDNN/src/gpu/space_to_depth_gpu.cpp      |    6 +-
 .../clDNN/src/gpu/strided_slice_gpu.cpp       |   15 +-
 .../thirdparty/clDNN/src/gpu/tile_gpu.cpp     |    6 +-
 .../clDNN/src/gpu/wait_for_events_gpu.cpp     |   14 +-
 .../graph_optimizer/add_required_reorders.cpp |    2 +-
 .../basic_memory_dependencies.cpp             |    2 +-
 .../graph_optimizer/calculate_prior_boxes.cpp |    3 +-
 .../src/graph_optimizer/compile_graph.cpp     |    7 +-
 .../graph_optimizer/concat_input_order.cpp    |   35 +-
 .../graph_optimizer/eltwise_remove_stride.cpp |    2 +-
 .../graph_optimizer/handle_input_padding.cpp  |    4 +-
 .../oooq_memory_dependencies.cpp              |    2 +-
 .../graph_optimizer/post_input_reorder.cpp    |    5 +-
 .../graph_optimizer/post_optimize_weights.cpp |    4 +-
 .../graph_optimizer/pre_replace_deconv.cpp    |   22 +-
 .../graph_optimizer/prepare_buffer_fusing.cpp |    3 +-
 .../prepare_primitive_fusing.cpp              |   23 +-
 .../graph_optimizer/prepare_quantization.cpp  |  243 +-
 .../graph_optimizer/propagate_constants.cpp   |   19 +-
 .../src/graph_optimizer/reorder_inputs.cpp    |    3 +-
 .../skipped_branch_memory_dependencies.cpp    |    2 +-
 .../strided_slice_optimize.cpp                |    2 +-
 .../clDNN/src/include/activation_inst.h       |    5 +-
 .../clDNN/src/include/arg_max_min_inst.h      |    2 +-
 .../src/include/average_unpooling_inst.h      |    3 +-
 .../clDNN/src/include/batch_to_space_inst.h   |    3 +-
 .../src/include/binary_convolution_inst.h     |    6 +-
 .../clDNN/src/include/border_inst.h           |    5 +-
 .../clDNN/src/include/broadcast_inst.h        |    2 +-
 .../clDNN/src/include/concatenation_inst.h    |    3 +-
 .../clDNN/src/include/condition_inst.h        |   17 +-
 .../clDNN/src/include/convolution_inst.h      |   30 +-
 .../thirdparty/clDNN/src/include/crop_inst.h  |    3 +-
 .../src/include/ctc_greedy_decoder_inst.h     |    3 +-
 .../clDNN/src/include/cum_sum_inst.h          |    3 +-
 .../src/include/custom_gpu_primitive_inst.h   |    3 +-
 .../thirdparty/clDNN/src/include/data_inst.h  |   10 +-
 .../clDNN/src/include/deconvolution_inst.h    |   16 +-
 .../src/include/deformable_convolution_inst.h |   16 +-
 .../clDNN/src/include/depth_to_space_inst.h   |    3 +-
 .../clDNN/src/include/detection_output_inst.h |    9 +-
 .../clDNN/src/include/device_impl.h           |  105 -
 .../clDNN/src/include/eltwise_inst.h          |    5 +-
 .../clDNN/src/include/embedding_bag_inst.h    |    4 +-
 .../clDNN/src/include/engine_impl.h           |  131 --
 .../src/include/extract_image_patches_inst.h  |    3 +-
 .../clDNN/src/include/fully_connected_inst.h  |    7 +-
 .../src/include/fused_conv_eltwise_inst.h     |   10 +-
 .../clDNN/src/include/gather_inst.h           |    3 +-
 .../clDNN/src/include/gather_nd_inst.h        |   21 +-
 .../clDNN/src/include/gather_tree_inst.h      |    5 +-
 .../thirdparty/clDNN/src/include/gemm_inst.h  |    3 +-
 .../clDNN/src/include/generic_layer.hpp       |    5 +-
 .../clDNN/src/include/generic_layer_inst.h    |    1 +
 .../thirdparty/clDNN/src/include/grn_inst.h   |    3 +-
 .../clDNN/src/include/implementation_map.h    |    2 +
 .../clDNN/src/include/input_layout_inst.h     |    7 +-
 .../clDNN/src/include/internal_primitive.h    |   36 -
 .../include/internal_primitive_type_base.h    |   69 -
 .../src/include/kernel_selector_helper.h      |   14 +-
 .../clDNN/src/include/layout_optimizer.h      |    6 +-
 .../thirdparty/clDNN/src/include/loop_inst.h  |   89 +-
 .../thirdparty/clDNN/src/include/lrn_inst.h   |    3 +-
 .../src/include/lstm_dynamic_input_inst.h     |   12 +-
 .../clDNN/src/include/lstm_dynamic_inst.h     |    4 +-
 .../src/include/lstm_dynamic_timeloop_inst.h  |   18 +-
 .../clDNN/src/include/lstm_elt_inst.h         |    5 +-
 .../clDNN/src/include/lstm_gemm_inst.h        |   11 +-
 .../thirdparty/clDNN/src/include/lstm_inst.h  |   15 +-
 .../clDNN/src/include/max_unpooling_inst.h    |    3 +-
 .../clDNN/src/include/memory_impl.h           |  115 -
 .../thirdparty/clDNN/src/include/meta_utils.h |   32 +-
 .../clDNN/src/include/mutable_data_inst.h     |   16 +-
 .../thirdparty/clDNN/src/include/mvn_inst.h   |    3 +-
 .../clDNN/src/include/network_impl.h          |   68 +-
 .../src/include/non_max_suppression_inst.h    |   34 +-
 .../clDNN/src/include/normalize_inst.h        |    5 +-
 .../clDNN/src/include/one_hot_inst.h          |    5 +-
 .../clDNN/src/include/pass_manager.h          |    2 +-
 .../clDNN/src/include/permute_inst.h          |    3 +-
 .../clDNN/src/include/pooling_inst.h          |    3 +-
 .../clDNN/src/include/primitive_inst.h        |  140 +-
 .../clDNN/src/include/primitive_type.h        |   17 +-
 .../clDNN/src/include/primitive_type_base.h   |   24 +-
 .../clDNN/src/include/prior_box_inst.h        |    9 +-
 .../clDNN/src/include/program_dump_graph.h    |    3 +-
 .../clDNN/src/include/program_helpers.h       |    6 +-
 .../clDNN/src/include/program_impl.h          |   55 +-
 .../clDNN/src/include/program_node.h          |   88 +-
 .../clDNN/src/include/proposal_inst.h         |    3 +-
 .../src/include/pyramid_roi_align_inst.h      |   12 +-
 .../clDNN/src/include/quantize_inst.h         |    2 +-
 .../clDNN/src/include/reduce_inst.h           |    3 +-
 .../clDNN/src/include/refcounted_obj.h        |  135 --
 .../clDNN/src/include/region_yolo_inst.h      |    3 +-
 .../clDNN/src/include/reorder_inst.h          |    8 +-
 .../clDNN/src/include/reorg_yolo_inst.h       |    3 +-
 .../clDNN/src/include/resample_inst.h         |    5 +-
 .../clDNN/src/include/reshape_inst.h          |    5 +-
 .../clDNN/src/include/reverse_sequence_inst.h |    3 +-
 .../clDNN/src/include/roi_pooling_inst.h      |    7 +-
 .../thirdparty/clDNN/src/include/scale_inst.h |    9 +-
 .../include/scatter_elements_update_inst.h    |    2 +-
 .../src/include/scatter_nd_update_inst.h      |    2 +-
 .../clDNN/src/include/scatter_update_inst.h   |    3 +-
 .../clDNN/src/include/select_inst.h           |    4 +-
 .../clDNN/src/include/shuffle_channels_inst.h |    3 +-
 .../clDNN/src/include/sliding_window_utils.h  |    8 +-
 .../clDNN/src/include/softmax_inst.h          |    3 +-
 .../clDNN/src/include/space_to_batch_inst.h   |    4 +-
 .../clDNN/src/include/space_to_depth_inst.h   |    4 +-
 .../thirdparty/clDNN/src/include/split_inst.h |    4 +-
 .../clDNN/src/include/strided_slice_inst.h    |    6 +-
 .../thirdparty/clDNN/src/include/tile_inst.h  |    4 +-
 .../clDNN/src/include/to_string_utils.h       |   11 +-
 .../clDNN/src/include/topology_impl.h         |    9 +-
 .../thirdparty/clDNN/src/input_layout.cpp     |   16 +-
 .../clDNN/src/kernel_selector_helper.cpp      |   26 +-
 .../thirdparty/clDNN/src/layout_optimizer.cpp |    2 +-
 .../thirdparty/clDNN/src/loop.cpp             |   57 +-
 inference-engine/thirdparty/clDNN/src/lrn.cpp |    2 +-
 .../thirdparty/clDNN/src/lstm.cpp             |    2 +-
 .../thirdparty/clDNN/src/lstm_dynamic.cpp     |    2 +-
 .../clDNN/src/lstm_dynamic_input.cpp          |    2 +-
 .../clDNN/src/lstm_dynamic_timeloop.cpp       |    2 +-
 .../thirdparty/clDNN/src/lstm_elt.cpp         |    2 +-
 .../thirdparty/clDNN/src/lstm_gemm.cpp        |    2 +-
 .../thirdparty/clDNN/src/max_unpooling.cpp    |    2 +-
 .../thirdparty/clDNN/src/memory.cpp           |  149 --
 .../thirdparty/clDNN/src/mutable_data.cpp     |   77 +-
 .../thirdparty/clDNN/src/network.cpp          |  298 +--
 .../thirdparty/clDNN/src/nodes_ordering.cpp   |    2 +-
 .../thirdparty/clDNN/src/normalize.cpp        |    2 +-
 .../thirdparty/clDNN/src/one_hot.cpp          |    2 +-
 .../thirdparty/clDNN/src/permute.cpp          |    2 +-
 .../thirdparty/clDNN/src/pooling.cpp          |    2 +-
 .../thirdparty/clDNN/src/primitive_inst.cpp   |   64 +-
 .../thirdparty/clDNN/src/prior_box.cpp        |   20 +-
 .../thirdparty/clDNN/src/program.cpp          |  171 +-
 .../clDNN/src/program_dump_graph.cpp          |    7 +-
 .../thirdparty/clDNN/src/program_helpers.cpp  |   11 +-
 .../thirdparty/clDNN/src/program_node.cpp     |   18 +-
 .../clDNN/src/pyramid_roi_align.cpp           |    2 +-
 .../thirdparty/clDNN/src/quantize.cpp         |    4 +-
 .../thirdparty/clDNN/src/reduce.cpp           |    2 +-
 .../thirdparty/clDNN/src/reorder.cpp          |    4 +-
 .../thirdparty/clDNN/src/resample.cpp         |    2 +-
 .../thirdparty/clDNN/src/reshape.cpp          |    4 +-
 .../thirdparty/clDNN/src/reverse_sequence.cpp |    2 +-
 .../thirdparty/clDNN/src/roi_pooling.cpp      |    2 +-
 .../thirdparty/clDNN/src/scale.cpp            |    2 +-
 .../clDNN/src/scatter_elements_update.cpp     |    2 +-
 .../clDNN/src/scatter_nd_update.cpp           |    2 +-
 .../thirdparty/clDNN/src/scatter_update.cpp   |    2 +-
 .../thirdparty/clDNN/src/select.cpp           |    2 +-
 .../thirdparty/clDNN/src/shuffle_channels.cpp |    2 +-
 .../thirdparty/clDNN/src/space_to_batch.cpp   |    2 +-
 .../thirdparty/clDNN/src/space_to_depth.cpp   |    2 +-
 .../thirdparty/clDNN/src/split.cpp            |    4 +-
 .../thirdparty/clDNN/src/strided_slice.cpp    |    2 +-
 .../thirdparty/clDNN/src/tile.cpp             |    4 +-
 .../thirdparty/clDNN/src/topology.cpp         |   11 +-
 .../thirdparty/clDNN/tests/CMakeLists.txt     |   21 +-
 .../tests/module_tests/events_pool_test.cpp   |   17 +-
 .../tests/module_tests/gpu_toolkit_test.cpp   |  225 --
 .../graph_manipulation_gpu_test.cpp           |   77 +-
 .../prepare_conv_eltw_fusing.cpp              |   33 +-
 .../module_tests}/program_impl_wrapper.h      |   16 +
 .../module_tests}/reorder_inputs_test.cpp     |   56 +-
 .../module_tests/test_uqr_distribution.cpp    |    8 +-
 .../module_tests}/usm_memory_test.cpp         |  104 +-
 .../test_cases/activation_simple_gpu_test.cpp |  304 ++-
 .../test_cases/add_reorders_gpu_test.cpp      |   85 +-
 .../tests/test_cases/arg_max_gpu_test.cpp     |  282 +--
 .../test_cases/average_unpooling_gpu_test.cpp |   68 +-
 .../clDNN/tests/test_cases/barriers_test.cpp  |   51 +-
 .../test_cases/batch_to_space_gpu_test.cpp    |  125 +-
 .../binary_convolution_gpu_test.cpp           |  128 +-
 .../tests/test_cases/border_gpu_test.cpp      |  150 +-
 .../tests/test_cases/broadcast_gpu_test.cpp   |   75 +-
 .../clDNN/tests/test_cases/cache_test.cpp     |   33 +-
 .../tests/test_cases/cl_mem_input_test.cpp    |   92 +-
 .../tests/test_cases/command_queue_test.cpp   |   62 +-
 .../test_cases/concatenation_gpu_test.cpp     |  180 +-
 .../tests/test_cases/condition_gpu_test.cpp   |  195 +-
 .../tests/test_cases/convolution_gpu_test.cpp | 1458 ++++++------
 .../clDNN/tests/test_cases/crop_gpu_test.cpp  |  204 +-
 .../tests/test_cases/cum_sum_gpu_test.cpp     |   91 +-
 .../test_cases/custom_gpu_primitive_test.cpp  |  197 +-
 .../test_cases/deconvolution_gpu_test.cpp     |  478 ++--
 .../test_cases/depth_concatenate_gpu_test.cpp |  300 ++-
 .../test_cases/depth_to_space_gpu_test.cpp    |   87 +-
 .../test_cases/detection_output_test.cpp      |  311 ++-
 .../tests/test_cases/eltwise_gpu_test.cpp     |  765 +++----
 .../test_cases/embedding_bag_gpu_test.cpp     |  287 ++-
 .../extract_image_patches_gpu_test.cpp        |   71 +-
 .../test_cases/fully_connected_gpu_test.cpp   |  246 +-
 .../fused_conv_eltwise_gpu_test.cpp           |   91 +-
 .../tests/test_cases/fusings_gpu_test.cpp     |  109 +-
 .../tests/test_cases/gather_gpu_test.cpp      |  367 ++-
 .../tests/test_cases/gather_nd_gpu_test.cpp   |  141 +-
 .../clDNN/tests/test_cases/gemm_gpu_test.cpp  |  216 +-
 .../clDNN/tests/test_cases/loop_gpu_test.cpp  |  117 +-
 .../clDNN/tests/test_cases/lrn_gpu_test.cpp   |   44 +-
 .../test_cases/lstm_dynamic_gpu_test.cpp      |  181 +-
 .../clDNN/tests/test_cases/lstm_gpu_test.cpp  |  301 ++-
 .../test_cases/max_unpooling_gpu_test.cpp     |   88 +-
 .../clDNN/tests/test_cases/memory_test.cpp    |  213 +-
 .../clDNN/tests/test_cases/mvn_gpu_test.cpp   |  176 +-
 .../test_cases/non_max_suppression_test.cpp   |   59 +-
 .../tests/test_cases/normalizel2_gpu_test.cpp |   53 +-
 .../tests/test_cases/one_hot_gpu_test.cpp     |  154 +-
 .../tests/test_cases/permute_gpu_test.cpp     |  303 ++-
 .../tests/test_cases/pooling_gpu_test.cpp     |  587 +++--
 .../propagate_constants_gpu_test.cpp          |   39 +-
 .../tests/test_cases/proposal_cpu_test.cpp    |   72 +-
 .../test_cases/pyramid_roi_align_gpu_test.cpp |   30 +-
 .../tests/test_cases/quantize_gpu_test.cpp    |  164 +-
 .../tests/test_cases/reduce_gpu_test.cpp      |  278 ++-
 .../tests/test_cases/region_yolo_gpu_test.cpp |   26 +-
 .../test_cases/removing_output_node_test.cpp  |   52 +-
 .../tests/test_cases/reorder_gpu_test.cpp     |  434 ++--
 .../tests/test_cases/resample_gpu_test.cpp    |  311 ++-
 .../tests/test_cases/reshape_gpu_test.cpp     |  106 +-
 .../test_cases/reverse_sequence_gpu_test.cpp  |  189 +-
 .../clDNN/tests/test_cases/scale_gpu_test.cpp | 1994 -----------------
 .../scatter_elements_update_gpu_test.cpp      |   30 +-
 .../test_cases/scatter_nd_update_gpu_test.cpp |  544 +++--
 .../test_cases/scatter_update_gpu_test.cpp    |  507 +++--
 .../tests/test_cases/select_gpu_test.cpp      |  641 +++---
 .../test_cases/shuffle_channels_test.cpp      |   83 +-
 .../tests/test_cases/softmax_gpu_test.cpp     |  187 +-
 .../test_cases/space_to_batch_gpu_test.cpp    |  125 +-
 .../test_cases/space_to_depth_gpu_test.cpp    |  155 +-
 .../spatial_concatenate_gpu_test.cpp          |  322 ++-
 .../clDNN/tests/test_cases/split_gpu_test.cpp |  125 +-
 .../clDNN/tests/test_cases/streams_test.cpp   |  145 +-
 .../test_cases/strided_slice_gpu_test.cpp     |  352 ++-
 .../clDNN/tests/test_cases/tensor_test.cpp    |    3 +-
 .../clDNN/tests/test_cases/tile_gpu_test.cpp  |  107 +-
 .../clDNN/tests/test_cases/topology_test.cpp  |   43 +-
 .../test_cases/trim_to_outputs_gpu_test.cpp   |   61 +-
 .../tests/test_utils/instrumentation.cpp      |  386 ----
 .../clDNN/tests/test_utils/instrumentation.h  |   39 -
 .../clDNN/tests/test_utils/network_test.h     |   67 +-
 .../clDNN/tests/test_utils/test_utils.cpp     |  611 +++--
 .../clDNN/tests/test_utils/test_utils.h       |  439 ++--
 .../uniform_quantized_real_distribution.hpp   |    7 +-
 .../clDNN/tests_core_internal/CMakeLists.txt  |  137 --
 .../clDNN/tests_core_internal/main.cpp        |   11 -
 613 files changed, 13711 insertions(+), 18482 deletions(-)
 delete mode 100644 inference-engine/thirdparty/clDNN/api/cldnn.hpp
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/graph}/network.hpp (79%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/graph}/program.hpp (97%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/graph}/topology.hpp (84%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/activation.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/arg_max_min.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/average_unpooling.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/batch_to_space.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/binary_convolution.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/border.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/broadcast.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/concatenation.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/condition.hpp (98%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/convolution.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/crop.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/ctc_greedy_decoder.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/cum_sum.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/custom_gpu_primitive.hpp (98%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/data.hpp (92%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/deconvolution.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/depth_to_space.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/detection_output.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/eltwise.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/embedding_bag.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/extract_image_patches.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/fully_connected.hpp (100%)
 rename inference-engine/thirdparty/clDNN/{api_extension => api/cldnn/primitives}/fused_conv_eltwise.hpp (98%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/gather.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/gather_nd.hpp (63%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/gather_tree.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/gemm.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/grn.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/input_layout.hpp (97%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/loop.hpp (99%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/lrn.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/lstm.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/lstm_dynamic.hpp (100%)
 rename inference-engine/thirdparty/clDNN/{api_extension => api/cldnn/primitives}/lstm_dynamic_input.hpp (98%)
 rename inference-engine/thirdparty/clDNN/{api_extension => api/cldnn/primitives}/lstm_dynamic_timeloop.hpp (99%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/max_unpooling.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/mutable_data.hpp (92%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/mvn.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/non_max_suppression.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/normalize.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/one_hot.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/permute.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/pooling.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/primitive.hpp (98%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/prior_box.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/proposal.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/pyramid_roi_align.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/quantize.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/reduce.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/region_yolo.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/reorder.hpp (99%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/reorg_yolo.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/resample.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/reshape.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/reverse_sequence.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/roi_pooling.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/scale.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/scatter_elements_update.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/scatter_nd_update.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/scatter_update.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/select.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/shuffle_channels.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/softmax.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/space_to_batch.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/space_to_depth.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/split.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/strided_slice.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/primitives}/tile.hpp (100%)
 rename inference-engine/thirdparty/clDNN/api/{compounds.h => cldnn/runtime/compounds.hpp} (98%)
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/device.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/device_info.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/device_query.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/engine.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/engine_configuration.hpp
 rename inference-engine/thirdparty/clDNN/{src/include/error_handler.h => api/cldnn/runtime/error_handler.hpp} (99%)
 rename inference-engine/thirdparty/clDNN/{src/include/event_impl.h => api/cldnn/runtime/event.hpp} (82%)
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/half.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/kernel.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/kernel_args.hpp
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/runtime}/layout.hpp (99%)
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory_caps.hpp
 rename inference-engine/thirdparty/clDNN/{src/include/memory_pool.h => api/cldnn/runtime/memory_pool.hpp} (52%)
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/runtime}/profiling.hpp (99%)
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/stream.hpp
 rename inference-engine/thirdparty/clDNN/api/{ => cldnn/runtime}/tensor.hpp (99%)
 create mode 100644 inference-engine/thirdparty/clDNN/api/cldnn/runtime/utils.hpp
 delete mode 100644 inference-engine/thirdparty/clDNN/api/device.hpp
 delete mode 100644 inference-engine/thirdparty/clDNN/api/engine.hpp
 delete mode 100644 inference-engine/thirdparty/clDNN/api/event.hpp
 delete mode 100644 inference-engine/thirdparty/clDNN/api/memory.hpp
 delete mode 100644 inference-engine/thirdparty/clDNN/api/meta_utils.hpp
 rename inference-engine/thirdparty/clDNN/{src/gpu => kernel_selector/core}/device_cache_reader.cpp (65%)
 create mode 100644 inference-engine/thirdparty/clDNN/kernel_selector/core/device_cache_reader.h
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/CMakeLists.txt
 rename inference-engine/thirdparty/clDNN/{src/include/cldnn_itt.h => runtime/cldnn_itt.hpp} (92%)
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/device_query.cpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/engine.cpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/event.cpp
 rename inference-engine/thirdparty/clDNN/{src/gpu => runtime}/kernels_cache.cpp (82%)
 rename inference-engine/thirdparty/clDNN/{src/gpu/kernels_cache.h => runtime/kernels_cache.hpp} (70%)
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/kernels_factory.cpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/kernels_factory.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/memory.cpp
 rename inference-engine/thirdparty/clDNN/{src => runtime}/memory_pool.cpp (55%)
 rename inference-engine/thirdparty/clDNN/{src/gpu => runtime/ocl}/ocl_base_event.cpp (93%)
 rename inference-engine/thirdparty/clDNN/{src/gpu/ocl_base_event.h => runtime/ocl/ocl_base_event.hpp} (76%)
 rename inference-engine/thirdparty/clDNN/{src/gpu/command_queues_builder.cpp => runtime/ocl/ocl_command_queues_builder.cpp} (80%)
 rename inference-engine/thirdparty/clDNN/{src/gpu/command_queues_builder.h => runtime/ocl/ocl_command_queues_builder.hpp} (57%)
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_common.hpp
 rename inference-engine/thirdparty/clDNN/{src/gpu/device_info.cpp => runtime/ocl/ocl_device.cpp} (54%)
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device.hpp
 rename inference-engine/thirdparty/clDNN/{src/gpu/ocl_builder.cpp => runtime/ocl/ocl_device_detector.cpp} (68%)
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device_detector.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine.cpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine_factory.hpp
 rename inference-engine/thirdparty/clDNN/{src/gpu/events_pool.h => runtime/ocl/ocl_events_pool.hpp} (57%)
 rename inference-engine/thirdparty/clDNN/{common/include/cl2_ext.hpp => runtime/ocl/ocl_ext.hpp} (97%)
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_kernel.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_kernels_factory.cpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_memory.cpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_memory.hpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_stream.cpp
 create mode 100644 inference-engine/thirdparty/clDNN/runtime/ocl/ocl_stream.hpp
 rename inference-engine/thirdparty/clDNN/{src/gpu => runtime/ocl}/ocl_user_event.cpp (94%)
 rename inference-engine/thirdparty/clDNN/{src/gpu/ocl_user_event.h => runtime/ocl/ocl_user_event.hpp} (71%)
 rename inference-engine/thirdparty/clDNN/{common/include/cl2_wrapper.h => runtime/ocl/ocl_wrapper.hpp} (95%)
 delete mode 100644 inference-engine/thirdparty/clDNN/src/device.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/src/engine.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/src/event.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/configuration.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/configuration.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/device_cache_reader.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/device_info.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/events_waiter.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/kernel.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/kernel.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/memory_gpu.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/memory_gpu.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/ocl_builder.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/ocl_queue_wrapper.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/ocl_queue_wrapper.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/ocl_toolkit.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/ocl_toolkit.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/gpu/push_pop_map.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/include/device_impl.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/include/engine_impl.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/include/internal_primitive.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/include/internal_primitive_type_base.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/include/memory_impl.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/include/refcounted_obj.h
 delete mode 100644 inference-engine/thirdparty/clDNN/src/memory.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/tests/module_tests/gpu_toolkit_test.cpp
 rename inference-engine/thirdparty/clDNN/{tests_core_internal/test_cases => tests/module_tests}/graph_manipulation_gpu_test.cpp (67%)
 rename inference-engine/thirdparty/clDNN/{tests_core_internal/test_cases => tests/module_tests}/prepare_conv_eltw_fusing.cpp (78%)
 rename inference-engine/thirdparty/clDNN/{tests_core_internal => tests/module_tests}/program_impl_wrapper.h (70%)
 rename inference-engine/thirdparty/clDNN/{tests_core_internal/test_cases => tests/module_tests}/reorder_inputs_test.cpp (79%)
 rename inference-engine/thirdparty/clDNN/{tests_core_internal/test_cases => tests/module_tests}/usm_memory_test.cpp (68%)
 delete mode 100644 inference-engine/thirdparty/clDNN/tests/test_cases/scale_gpu_test.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/tests/test_utils/instrumentation.cpp
 delete mode 100644 inference-engine/thirdparty/clDNN/tests/test_utils/instrumentation.h
 delete mode 100644 inference-engine/thirdparty/clDNN/tests_core_internal/CMakeLists.txt
 delete mode 100644 inference-engine/thirdparty/clDNN/tests_core_internal/main.cpp

diff --git a/inference-engine/src/cldnn_engine/CMakeLists.txt b/inference-engine/src/cldnn_engine/CMakeLists.txt
index ff138843dc45a6..161d6f16a8d2c8 100644
--- a/inference-engine/src/cldnn_engine/CMakeLists.txt
+++ b/inference-engine/src/cldnn_engine/CMakeLists.txt
@@ -33,7 +33,7 @@ target_include_directories(${TARGET_NAME} PRIVATE
         ${CMAKE_CURRENT_SOURCE_DIR}
         $<TARGET_PROPERTY:inference_engine_transformations,INTERFACE_INCLUDE_DIRECTORIES>
         ${CLDNN__OCL_ICD_INCDIRS}
-        ${CLDNN_TOP_FOLDER})
+        ${CLDNN_TOP_FOLDER}/api)
 
 set_target_properties(${TARGET_NAME} PROPERTIES INTERPROCEDURAL_OPTIMIZATION_RELEASE ${ENABLE_LTO})
 
diff --git a/inference-engine/src/cldnn_engine/cldnn_common_utils.h b/inference-engine/src/cldnn_engine/cldnn_common_utils.h
index c374a71a4653f3..f41f2d8e134b1c 100644
--- a/inference-engine/src/cldnn_engine/cldnn_common_utils.h
+++ b/inference-engine/src/cldnn_engine/cldnn_common_utils.h
@@ -5,7 +5,7 @@
 #pragma once
 
 #include <ie_layouts.h>
-#include <api/layout.hpp>
+#include <cldnn/runtime/layout.hpp>
 
 #include "ngraph/type/element_type.hpp"
 
diff --git a/inference-engine/src/cldnn_engine/cldnn_config.h b/inference-engine/src/cldnn_engine/cldnn_config.h
index 8c6d5d6c921da8..873c01e2188445 100644
--- a/inference-engine/src/cldnn_engine/cldnn_config.h
+++ b/inference-engine/src/cldnn_engine/cldnn_config.h
@@ -9,7 +9,7 @@
 
 #include "cldnn_custom_layer.h"
 
-#include <api/network.hpp>
+#include <cldnn/graph/network.hpp>
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/cldnn_custom_layer.h b/inference-engine/src/cldnn_engine/cldnn_custom_layer.h
index 95a6ff4c5c9f82..cbe41f8831b3f8 100644
--- a/inference-engine/src/cldnn_engine/cldnn_custom_layer.h
+++ b/inference-engine/src/cldnn_engine/cldnn_custom_layer.h
@@ -10,7 +10,7 @@
 #include <map>
 #include <ie_common.h>
 #include "pugixml.hpp"
-#include "api/tensor.hpp"
+#include "cldnn/runtime/tensor.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/cldnn_engine.cpp b/inference-engine/src/cldnn_engine/cldnn_engine.cpp
index 171919a80775d6..e3e67d95f152f1 100644
--- a/inference-engine/src/cldnn_engine/cldnn_engine.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_engine.cpp
@@ -81,6 +81,8 @@
 #include "cldnn_itt.h"
 #include "gpu/gpu_config.hpp"
 
+#include "cldnn/runtime/device_query.hpp"
+
 #ifdef __linux__
 # include <dlfcn.h>
 #endif
@@ -117,13 +119,13 @@ struct clDNNEngine::impl {
 };
 
 cldnn::device_info clDNNEngine::GetDeviceInfo(const std::map<std::string, std::string> &config) const {
-    auto device_info = device_map.begin()->second.get_info();
+    auto device_info = device_map.begin()->second->get_info();
     if (config.find(PluginConfigParams::KEY_DEVICE_ID) != config.end()) {
         auto val = config.at(PluginConfigParams::KEY_DEVICE_ID);
         if (device_map.find(val) == device_map.end()) {
             IE_THROW() << "Invalid device ID: " << val;
         }
-        device_info = device_map.at(val).get_info();
+        device_info = device_map.at(val)->get_info();
     }
 
     return device_info;
@@ -445,7 +447,8 @@ clDNNEngine::clDNNEngine() : m_defaultContext(nullptr) {
     RegisterPrimitives();
     // try loading clDNN engine and get info from it
     {
-        cldnn::device_query device_query;
+        // Set OCL runtime which should be always available
+        cldnn::device_query device_query(cldnn::engine_types::ocl, cldnn::runtime_types::ocl);
         device_map = device_query.get_available_devices();
     }
     // locate global custom kernel config
@@ -851,8 +854,8 @@ auto StringRightTrim = [](std::string string, std::string substring, bool case_s
 };
 
 static float GetGOPS(cldnn::device_info info, cldnn::data_types dt) {
-    auto freqGHz = info.core_frequency / 1000.f;
-    auto numEUs = info.cores_count;
+    auto freqGHz = info.gpu_frequency / 1000.f;
+    auto numEUs = info.execution_units_count;
     auto opsPerComputeBlock = 0;
     auto computeBlockIPC = 1.0f;
     switch (dt) {
@@ -894,8 +897,8 @@ Parameter clDNNEngine::GetMetric(const std::string& name, const std::map<std::st
 
     auto iter = device_map.find(device_id);
     auto device_info = iter != device_map.end() ?
-        iter->second.get_info() :
-        device_map.begin()->second.get_info();
+        iter->second->get_info() :
+        device_map.begin()->second->get_info();
 
     if (name == METRIC_KEY(SUPPORTED_METRICS)) {
         std::vector<std::string> metrics;
@@ -931,7 +934,7 @@ Parameter clDNNEngine::GetMetric(const std::string& name, const std::map<std::st
         gops[InferenceEngine::Precision::FP32] = GetGOPS(device_info, cldnn::data_types::f32);
         IE_SET_METRIC_RETURN(DEVICE_GOPS, gops);
     } else if (name == GPU_METRIC_KEY(EXECUTION_UNITS_COUNT)) {
-        IE_SET_METRIC_RETURN(GPU_EXECUTION_UNITS_COUNT, device_info.cores_count);
+        IE_SET_METRIC_RETURN(GPU_EXECUTION_UNITS_COUNT, device_info.execution_units_count);
     } else if (name == GPU_METRIC_KEY(UARCH_VERSION)) {
         std::stringstream s;
         if (device_info.gfx_ver.major == 0 && device_info.gfx_ver.minor == 0 && device_info.gfx_ver.revision == 0) {
diff --git a/inference-engine/src/cldnn_engine/cldnn_engine.h b/inference-engine/src/cldnn_engine/cldnn_engine.h
index f092bcd488e21e..10933cd7287729 100644
--- a/inference-engine/src/cldnn_engine/cldnn_engine.h
+++ b/inference-engine/src/cldnn_engine/cldnn_engine.h
@@ -7,7 +7,7 @@
 #include <map>
 #include <string>
 #include <memory>
-#include <api/engine.hpp>
+#include <cldnn/runtime/engine.hpp>
 #include <cpp_interfaces/interface/ie_iplugin_internal.hpp>
 #include <cpp_interfaces/interface/ie_iexecutable_network_internal.hpp>
 #include "cldnn_remote_context.h"
@@ -22,7 +22,7 @@ class clDNNEngine : public InferenceEngine::IInferencePlugin,
     std::shared_ptr<impl> _impl;
 
     // key: device_id, value: cldnn device
-    std::map<std::string, cldnn::device> device_map;
+    std::map<std::string, cldnn::device::ptr> device_map;
     std::mutex engine_mutex;
 
     mutable CLDNNRemoteCLContext::Ptr m_defaultContext;
diff --git a/inference-engine/src/cldnn_engine/cldnn_executable_network.cpp b/inference-engine/src/cldnn_engine/cldnn_executable_network.cpp
index 5191da35c2e369..94245c1d3b6bee 100644
--- a/inference-engine/src/cldnn_engine/cldnn_executable_network.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_executable_network.cpp
@@ -2,13 +2,13 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <list>
-#include <set>
-#include <unordered_set>
 
 #include "ie_metric_helpers.hpp"
-#include <api/cldnn.hpp>
-#include <api/data.hpp>
+#include <chrono>
+#include <cmath>
+#include <algorithm>
+
+#include "ie_metric_helpers.hpp"
 #include <chrono>
 #include <cmath>
 #include <algorithm>
@@ -27,7 +27,6 @@
 #include "threading/ie_cpu_streams_executor.hpp"
 #include "cpp_interfaces/interface/ie_iinfer_request_internal.hpp"
 
-
 using namespace InferenceEngine;
 using namespace InferenceEngine::details;
 
diff --git a/inference-engine/src/cldnn_engine/cldnn_graph.cpp b/inference-engine/src/cldnn_engine/cldnn_graph.cpp
index 1f835d8ac2c991..2b333a38ee9d20 100644
--- a/inference-engine/src/cldnn_engine/cldnn_graph.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_graph.cpp
@@ -2,22 +2,28 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+#include <cldnn/graph/network.hpp>
+#include <cldnn/runtime/profiling.hpp>
+
+#include "cldnn_graph.h"
+#include "simple_math.h"
+#include <cldnn/cldnn_config.hpp>
+#include "cldnn_infer_request.h"
+
+#include <description_buffer.hpp>
+#include <threading/ie_executor_manager.hpp>
+#include <exec_graph_info.hpp>
+
+#include <ie_ngraph_utils.hpp>
+#include <ngraph/variant.hpp>
+
 #include <list>
 #include <set>
 #include <unordered_set>
 #include <sstream>
-#include <api/cldnn.hpp>
-#include <api/network.hpp>
-#include <api/profiling.hpp>
-#include <api/custom_gpu_primitive.hpp>
 #include <chrono>
 #include <cmath>
 #include <algorithm>
-#include "cldnn_graph.h"
-#include "simple_math.h"
-#include <description_buffer.hpp>
-#include "cldnn_infer_request.h"
-#include <threading/ie_executor_manager.hpp>
 #include <fstream>
 #include <utility>
 #include <sys/types.h>
@@ -71,12 +77,10 @@ void CLDNNGraph::Build() {
         for (int b = m_bv_sz - 1; b >= 0; b--) {
             auto network = BuildNetwork(m_program->GetCompiledProgram(b));
             m_networks.insert(m_networks.begin(), network);
-            GetEngine()->release_pending_memory(network->get_id());
         }
     } else {
         auto network = BuildNetwork(m_program->GetCompiledProgram());
         m_networks.emplace_back(network);
-        GetEngine()->release_pending_memory(network->get_id());
     }
 
     UpdateImplementationsMap();
@@ -499,7 +503,7 @@ void CLDNNGraph::UpdatePerfStatistics() {
         }
     };
 
-    std::map<cldnn::primitive_id, cldnn::event> executedPrimitives = GetNetwork()->get_executed_primitives();
+    std::map<cldnn::primitive_id, cldnn::event::ptr> executedPrimitives = GetNetwork()->get_executed_primitives();
     auto allPrimitives = GetNetwork()->get_all_primitives();
 
     // Get profiling info for all layers
@@ -521,7 +525,7 @@ void CLDNNGraph::UpdatePerfStatistics() {
         auto event = execIter->second;
         executedPrimitives.erase(execIter);
 
-        cldnn::instrumentation::profiling_info cldnnInfo{profiledID, event.get_profiling_info()};
+        cldnn::instrumentation::profiling_info cldnnInfo{profiledID, event->get_profiling_info()};
 
         collectTimings(cldnnInfo, perfCount);
         perfCount.num++;
@@ -534,7 +538,7 @@ void CLDNNGraph::UpdatePerfStatistics() {
             pcIter = perfMap.find(executedID.first);
             auto& perfCount = pcIter->second.second;
 
-            cldnn::instrumentation::profiling_info cldnnInfo{executedID.first, executedID.second.get_profiling_info()};
+            cldnn::instrumentation::profiling_info cldnnInfo{executedID.first, executedID.second->get_profiling_info()};
 
             collectTimings(cldnnInfo, perfCount);
             perfCount.num++;
@@ -675,7 +679,7 @@ std::map<std::string, InferenceEngine::InferenceEngineProfileInfo> CLDNNGraph::G
             executedPrimitives.find(primId) != executedPrimitives.end()) {
             auto event = executedPrimitives.at(primId);
 
-            cldnn::instrumentation::profiling_info cldnnInfo{primId, event.get_profiling_info()};
+            cldnn::instrumentation::profiling_info cldnnInfo{primId, event->get_profiling_info()};
 
             // Collect timings
             long long cpuTime = 0;
diff --git a/inference-engine/src/cldnn_engine/cldnn_graph.h b/inference-engine/src/cldnn_engine/cldnn_graph.h
index 774b159a16c2b9..5ce64712fefa9b 100644
--- a/inference-engine/src/cldnn_engine/cldnn_graph.h
+++ b/inference-engine/src/cldnn_engine/cldnn_graph.h
@@ -17,8 +17,8 @@
 #include "ie_blob.h"
 #include "cpp/ie_cnn_network.h"
 
-#include <api/network.hpp>
-#include <api/topology.hpp>
+#include <cldnn/graph/network.hpp>
+#include <cldnn/graph/topology.hpp>
 
 #include <cpp_interfaces/impl/ie_executable_network_thread_safe_default.hpp>
 #include "cldnn_custom_layer.h"
@@ -43,7 +43,7 @@ class CLDNNGraph {
 
     const Config& getConfig() const { return m_config; }
     InferenceEngine::gpu::ClContext::Ptr GetContext() { return m_context; }
-    std::shared_ptr<const cldnn::engine> GetEngine() const { return getContextImpl(m_context)->GetEngine(); }
+    std::shared_ptr<cldnn::engine> GetEngine() const { return getContextImpl(m_context)->GetEngine(); }
     int GetMaxDynamicBatchSize() const { return getConfig().max_dynamic_batch; }
     const std::map<std::string, cldnn::layout>& GetInputLayouts() const { return m_program->GetInputLayouts(); }
     size_t GetNetworksCount() const { return m_networks.size(); }
diff --git a/inference-engine/src/cldnn_engine/cldnn_infer_request.cpp b/inference-engine/src/cldnn_engine/cldnn_infer_request.cpp
index 23f9895970ddf1..bb923f373b9e14 100644
--- a/inference-engine/src/cldnn_engine/cldnn_infer_request.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_infer_request.cpp
@@ -19,7 +19,7 @@ using namespace InferenceEngine;
 
 namespace CLDNNPlugin {
 
-const char CLDNNInferRequest::fp32_suffix[] = "_fp32";
+const char fp32_suffix[] = "_fp32";
 const char str_not_allocated[] = "Input data was not allocated.";
 const char cannot_set_compound[] = "cannot set compound blob: supported only for input pre-processing";
 const char wrong_nv12_blob[] = "NV12 input blob is expected for input with NV12 color format";
@@ -110,7 +110,7 @@ Blob::Ptr CLDNNInferRequest::createOutputBlob(const TensorDesc& desc, uint8_t* m
     }
 }
 
-void CLDNNInferRequest::input_attach(cldnn::primitive_id name, cldnn::memory& inputMem) {
+void CLDNNInferRequest::input_attach(cldnn::primitive_id name, cldnn::memory::ptr inputMem) {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNNPlugin, "CLDNNInferRequest::input_attach");
     auto impl = getContextImpl(m_graph->GetContext());
     impl->acquire_lock();
@@ -127,159 +127,66 @@ void CLDNNInferRequest::input_attach(cldnn::primitive_id name, cldnn::memory& in
 
 void CLDNNInferRequest::input_alloc(cldnn::primitive_id name, const cldnn::layout& layout) {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNNPlugin, "CLDNNInferRequest::input_alloc");
-    cldnn::memory input_mem = cldnn::memory::allocate(*(m_graph->GetEngine()), layout);
+    cldnn::memory::ptr input_mem = m_graph->GetEngine()->allocate_memory(layout);
     input_attach(name, input_mem);
 }
 
-void CLDNNInferRequest::copyOutputData(const cldnn::memory& outputMemory,
-                                        Blob::Ptr bptr,
-                                        buf_info* bi) {
-    OV_ITT_SCOPED_TASK(itt::domains::CLDNNPlugin, "CLDNNInferRequest::copyOutputData");
-    size_t n = (bi == nullptr) ? bptr->size() : bi->buf_size;
+template<typename T>
+void copyResultToOutputBlob(cldnn::memory::ptr src, Blob::Ptr dst, buf_info* bi, cldnn::stream& stream) {
+    size_t n = (bi == nullptr) ? dst->size() : bi->buf_size;
     size_t offset = (bi == nullptr) ? 0 : bi->buf_offset;
 
-    auto layout = outputMemory.get_layout();
+    auto layout = src->get_layout();
     auto size = layout.size;
-    auto l_padd = layout.data_padding.lower_size();
-    auto u_padd = layout.data_padding.upper_size();
-
-    auto h_padding = u_padd.spatial[0] + l_padd.spatial[0];
-    auto v_padding_l = (h_padding + size.spatial[0]) * u_padd.spatial[1];
-    auto v_padding_u = (h_padding + size.spatial[0]) * l_padd.spatial[1];
 
-    auto locked = bptr->buffer();
-    switch (bptr->getTensorDesc().getPrecision()) {
-    case Precision::FP32: {
-        auto out_f = locked.as<float*>();
-        if (out_f == nullptr) {
-            IE_THROW() << "Invalid output blob";
-        }
-        auto resPtr = outputMemory.pointer<float>();
-        float *resVec = out_f + offset;
-
-        if (h_padding || v_padding_l || v_padding_u) {
-            size_t i = 0;
-            for (size_t b = 0; b < size.batch[0]; b++) {
-                for (size_t f = 0; f < size.feature[0]; f++) {
-                    i += v_padding_l;
-                    for (size_t y = 0; y < size.spatial[1]; y++) {
-                        i += l_padd.spatial[0];
-                        for (size_t x = 0; x < size.spatial[0]; x++, i++) {
-                            *resVec++ = resPtr[i];
-                        }
-                        i += u_padd.spatial[0];
-                    }
-                    i += v_padding_u;
-                }
-            }
-        } else {
-            for (size_t i = 0; i < n; i++) {
-                resVec[i] = resPtr[i];
-            }
-        }
-    }
-    break;
-    case Precision::FP16: {
-        auto out_f = locked.as<uint16_t*>();
-        if (out_f == nullptr) {
-            IE_THROW() << "Invalid output blob";
-        }
-        auto resPtr = outputMemory.pointer<uint16_t>();
-        uint16_t* resVec = out_f + offset;
-
-        if (h_padding || v_padding_l || v_padding_u) {
-            size_t i = 0;
-            for (size_t b = 0; b < size.batch[0]; b++) {
-                for (size_t f = 0; f < size.feature[0]; f++) {
-                    i += v_padding_l;
-                    for (size_t y = 0; y < size.spatial[1]; y++) {
-                        i += l_padd.spatial[0];
-                        for (size_t x = 0; x < size.spatial[0]; x++, i++) {
-                            *resVec++ = resPtr[i];
-                        }
-                        i += u_padd.spatial[0];
-                    }
-                    i += v_padding_u;
-                }
-            }
-        } else {
-            for (size_t i = 0; i < n; i++) {
-                resVec[i] = resPtr[i];
-            }
-        }
-    }
-    break;
-    case Precision::I32: {
-        auto out_f = locked.as<int32_t*>();
-        if (out_f == nullptr) {
-            IE_THROW() << "Invalid output blob";
-        }
-        auto resPtr = outputMemory.pointer<int32_t>();
-        int32_t* resVec = out_f + offset;
-
-        if (h_padding || v_padding_l || v_padding_u) {
-            size_t i = 0;
-            for (size_t b = 0; b < size.batch[0]; b++) {
-                for (size_t f = 0; f < size.feature[0]; f++) {
-                    i += v_padding_l;
-                    for (size_t y = 0; y < size.spatial[1]; y++) {
-                        i += l_padd.spatial[0];
-                        for (size_t x = 0; x < size.spatial[0]; x++, i++) {
-                            *resVec++ = resPtr[i];
+    auto locked_dst = dst->buffer();
+    auto dst_ptr = locked_dst.as<T*>();
+    if (dst_ptr == nullptr) {
+        IE_THROW() << "Invalid output blob";
+    }
+    cldnn::mem_lock<T> src_lock{ src, stream };
+    T* src_ptr = src_lock.data();
+    dst_ptr += offset;
+
+    if (layout.data_padding) {
+        for (size_t b = 0; b < size.batch[0]; b++) {
+            for (size_t f = 0; f < size.feature[0]; f++) {
+                for (size_t w = 0; w < size.spatial[3]; w++) {
+                    for (size_t z = 0; z < size.spatial[2]; z++) {
+                        for (size_t y = 0; y < size.spatial[1]; y++) {
+                            for (size_t x = 0; x < size.spatial[0]; x++) {
+                                *dst_ptr++ = src_ptr[layout.get_linear_offset(cldnn::tensor(b, f, x, y, z, w))];
+                            }
                         }
-                        i += u_padd.spatial[0];
                     }
-                    i += v_padding_u;
                 }
             }
-        } else {
-            for (size_t i = 0; i < n; i++) {
-                resVec[i] = resPtr[i];
-            }
         }
-    }
-    break;
-    case Precision::I64: {
-        auto out_f = locked.as<int64_t*>();
-        if (out_f == nullptr) {
-            IE_THROW() << "Invalid output blob";
-        }
-        auto resPtr = outputMemory.pointer<int64_t>();
-        int64_t* resVec = out_f + offset;
-
-        if (h_padding || v_padding_l || v_padding_u) {
-            size_t i = 0;
-            for (size_t b = 0; b < size.batch[0]; b++) {
-                for (size_t f = 0; f < size.feature[0]; f++) {
-                    i += v_padding_l;
-                    for (size_t y = 0; y < size.spatial[1]; y++) {
-                        i += l_padd.spatial[0];
-                        for (size_t x = 0; x < size.spatial[0]; x++, i++) {
-                            *resVec++ = resPtr[i];
-                        }
-                        i += u_padd.spatial[0];
-                    }
-                    i += v_padding_u;
-                }
-            }
-        } else {
-            for (size_t i = 0; i < n; i++) {
-                resVec[i] = resPtr[i];
-            }
+    } else {
+        for (size_t i = 0; i < n; i++) {
+            dst_ptr[i] = src_ptr[i];
         }
     }
-    break;
-    default:
-        IE_THROW() << "The plugin does not support output " << bptr->getTensorDesc().getPrecision() << " precision";
+}
+
+void CLDNNInferRequest::copyOutputData(cldnn::memory::ptr src, Blob::Ptr dst, buf_info* bi) {
+    OV_ITT_SCOPED_TASK(itt::domains::CLDNNPlugin, "CLDNNInferRequest::copyOutputData");
+    auto& stream = m_graph->GetNetwork()->get_stream();
+    switch (dst->getTensorDesc().getPrecision()) {
+    case Precision::FP32: copyResultToOutputBlob<float>(src, dst, bi, stream);    break;
+    case Precision::FP16: copyResultToOutputBlob<uint16_t>(src, dst, bi, stream); break;
+    case Precision::I32:  copyResultToOutputBlob<int32_t>(src, dst, bi, stream);  break;
+    case Precision::I64:  copyResultToOutputBlob<int64_t>(src, dst, bi, stream);  break;
+    default: IE_THROW(NotImplemented) << "The plugin does not support output " << dst->getTensorDesc().getPrecision() << " precision";
     }
 }
 
 void CLDNNInferRequest::copyInputData(std::shared_ptr<cldnn::network> network,
-                                    const cldnn::primitive_id &inputName,
-                                    const cldnn::layout& inputLayout,
-                                    const Blob &inputBlob, buf_info* bi) {
+                                      const cldnn::primitive_id &inputName,
+                                      const cldnn::layout& inputLayout,
+                                      const Blob &inputBlob, buf_info* bi) {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNNPlugin, "CLDNNInferRequest::copyInputData");
-    size_t n = (bi == nullptr) ? inputBlob.size() : bi->buf_size;
+
     size_t offset = (bi == nullptr) ? 0 : bi->buf_offset;
 
     cldnn::primitive_id internalName = "parameter:" + inputName;
@@ -287,37 +194,37 @@ void CLDNNInferRequest::copyInputData(std::shared_ptr<cldnn::network> network,
     switch (inputBlob.getTensorDesc().getPrecision()) {
     case Precision::FP32: {
         float* blob_ptr = const_cast<float*>(locked.as<const float*>()) + offset;
-        network->set_input_data(internalName, cldnn::memory::attach(inputLayout, blob_ptr, n));
+        network->set_input_data(internalName, network->get_engine().attach_memory(inputLayout, blob_ptr));
         break;
     }
     case Precision::I32: {
         int32_t* blob_ptr = const_cast<int32_t*>(locked.as<const int32_t*>()) + offset;
-        network->set_input_data(internalName, cldnn::memory::attach(inputLayout, blob_ptr, n));
+        network->set_input_data(internalName, network->get_engine().attach_memory(inputLayout, blob_ptr));
         break;
     }
     case Precision::I64: {
         int64_t* blob_ptr = const_cast<int64_t*>(locked.as<const int64_t*>()) + offset;
-        network->set_input_data(internalName, cldnn::memory::attach(inputLayout, blob_ptr, n));
+        network->set_input_data(internalName, network->get_engine().attach_memory(inputLayout, blob_ptr));
         break;
     }
     case Precision::FP16: {
         uint16_t* blob_ptr = const_cast<uint16_t*>(locked.as<const uint16_t*>()) + offset;
-        network->set_input_data(internalName, cldnn::memory::attach(inputLayout, blob_ptr, n));
+        network->set_input_data(internalName, network->get_engine().attach_memory(inputLayout, blob_ptr));
         break;
     }
     case Precision::I8: {
         int8_t* blob_ptr = const_cast<int8_t*>(locked.as<const int8_t*>()) + offset;
-        network->set_input_data(internalName, cldnn::memory::attach(inputLayout, blob_ptr, n));
+        network->set_input_data(internalName, network->get_engine().attach_memory(inputLayout, blob_ptr));
         break;
     }
     case Precision::U8: {
         uint8_t* blob_ptr = const_cast<uint8_t*>(locked.as<const uint8_t*>()) + offset;
-        network->set_input_data(internalName, cldnn::memory::attach(inputLayout, blob_ptr, n));
+        network->set_input_data(internalName, network->get_engine().attach_memory(inputLayout, blob_ptr));
         break;
     }
     case Precision::BOOL: {
         uint8_t* blob_ptr = const_cast<uint8_t*>(locked.as<const uint8_t*>()) + offset;
-        network->set_input_data(internalName, cldnn::memory::attach(inputLayout, blob_ptr, n));
+        network->set_input_data(internalName, network->get_engine().attach_memory(inputLayout, blob_ptr));
         break;
     }
     default:
@@ -601,6 +508,7 @@ void CLDNNInferRequest::SetBlob(const std::string& name, const Blob::Ptr &data)
 void CLDNNInferRequest::AllocateInputs() {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNNPlugin, "CLDNNInferRequest::AllocateInputs");
     auto inputLayouts = m_graph->GetInputLayouts();
+    auto& stream = m_graph->GetNetwork()->get_stream();
     // allocate inputs
     for (auto& ni : _networkInputs) {
         std::string name = ni.first;
@@ -623,25 +531,24 @@ void CLDNNInferRequest::AllocateInputs() {
                 input_alloc(UVName, inputLayouts.at(UVName));
 
                 size_t height = desc.getDims()[2], width = desc.getDims()[3];
-                cldnn::pointer<uint8_t> input_mem_ptr_Y = inputsMemory.at(YName).pointer<uint8_t>();
+                cldnn::mem_lock<uint8_t> input_mem_ptr_Y{inputsMemory.at(YName), stream};
                 TensorDesc ydesc(Precision::U8, { 1, 1, height, width }, Layout::NHWC);
                 auto blobY = createInputBlob(ydesc, input_mem_ptr_Y.data());
 
-                cldnn::pointer<uint8_t> input_mem_ptr_UV = inputsMemory.at(UVName).pointer<uint8_t>();
+                cldnn::mem_lock<uint8_t> input_mem_ptr_UV{ inputsMemory.at(UVName), stream };
                 TensorDesc uvdesc(Precision::U8, { 1, 2, height / 2, width / 2 }, Layout::NHWC);
                 auto blobUV = createInputBlob(uvdesc, input_mem_ptr_UV.data());
 
                 blobs.push_back(make_shared_blob<NV12Blob>(blobY, blobUV));
             }
             _inputs[name] = desc.getDims()[0] == 1 ? blobs[0] : make_shared_blob<BatchedBlob>(blobs);
-
         } else {
             if (inputLayouts.find(name) == inputLayouts.end()) {
                 IE_THROW() << "Input layout for " << name << " is not found";
             }
             cldnn::layout layout = inputLayouts.at(name);
             input_alloc(name, layout);
-            cldnn::pointer<uint8_t> mem_ptr = inputsMemory.at(name).pointer<uint8_t>();
+            cldnn::mem_lock<uint8_t> mem_ptr{inputsMemory.at(name), stream};
             _inputs[name] = createInputBlob(desc, mem_ptr.data());
 
             if (desc.getPrecision() == Precision::I16 || desc.getPrecision() == Precision::U16) {
@@ -685,8 +592,8 @@ void CLDNNInferRequest::AllocateOutputs() {
     bool can_reuse_internal_mem = !m_useStreams;
     for (auto& no : _networkOutputs) {
         std::string outputID = m_graph->MapOutputName(no.first);
-        cldnn::memory output_mem = m_graph->GetNetwork()->get_output_memory(outputID);
-        cldnn::pointer<uint8_t> output_mem_ptr = output_mem.pointer<uint8_t>();
+        cldnn::memory::ptr output_mem = m_graph->GetNetwork()->get_output_memory(outputID);
+        cldnn::mem_lock<uint8_t> output_mem_ptr{output_mem, m_graph->GetNetwork()->get_stream()};
         if (output_mem_ptr.data() == nullptr) {
             IE_THROW() << "Empty output memory for primitive " << outputID;
         }
@@ -824,6 +731,7 @@ CLDNNInferRequest::CLDNNInferRequest(InputsDataMap networkInputs, OutputsDataMap
 void CLDNNInferRequest::execAndParse() {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNNPlugin, "CLDNNInferRequest::execAndParse");
     auto networkOutputs = m_graph->GetNetwork()->execute();
+    auto& stream = m_graph->GetNetwork()->get_stream();
 
     // Collect outputs as requested by the model
     for (auto& no : _networkOutputs) {
@@ -835,12 +743,12 @@ void CLDNNInferRequest::execAndParse() {
         // mapping remote blobs not needed -
         // let the user take care of them explicitly
         if (!bptr->is<gpu::ClBlob>()) {
-            auto out_ptr = outputMemory.pointer<uint8_t>();
+            cldnn::mem_lock<uint8_t> out_ptr{outputMemory, stream};
             auto blob_ptr = bptr->buffer().as<uint8_t*>();
 
             // If Async API is used, copy of output blobs is not needed, unless SetBlob function was called.
             // But in the case when old API is used we have to copy data to memory provided by user.
-            if (blob_ptr != &out_ptr[0]) {
+            if (blob_ptr != out_ptr.data()) {
                 copyOutputData(outputMemory, bptr);
             }
         }
@@ -965,19 +873,20 @@ void CLDNNInferRequest::PrepareInput(const cldnn::primitive_id &inputName, const
         IE_THROW() << "Input name mismatch.";
     }
     auto inputLayout = m_graph->GetInputLayouts().at(inputName);
-    auto is_same_buffer = [](const Blob& blob, const cldnn::memory& memory) -> bool {
+    auto is_same_buffer = [&](const Blob& blob, cldnn::memory::ptr memory) -> bool {
         const std::string str_not_allocated("Input data was not allocated.");
-        cldnn::pointer<const uint8_t> ptr = memory.pointer<const uint8_t>();
+        cldnn::mem_lock<uint8_t> ptr{memory, m_graph->GetNetwork()->get_stream()};
         const uint8_t* blob_ptr = blob.cbuffer().as<const uint8_t*>();
         const uint8_t* mem_ptr = ptr.data();
         if (blob_ptr == nullptr || mem_ptr == nullptr) {
             IE_THROW() << str_not_allocated;
         }
-        return (blob_ptr == mem_ptr) && (blob.byteSize() == memory.size());
+        return (blob_ptr == mem_ptr) && (blob.byteSize() == memory->size());
     };
 
     cldnn::primitive_id internalName = "parameter:" + inputName;
-    const cldnn::memory& memory = inputsMemory.at(inputName);
+    cldnn::memory::ptr memory = inputsMemory.at(inputName);
+    auto& stream = m_graph->GetNetwork()->get_stream();
     auto _nw_ptr = m_graph->GetNetwork();
     auto prec = inputBlob.getTensorDesc().getPrecision();
 
@@ -986,8 +895,8 @@ void CLDNNInferRequest::PrepareInput(const cldnn::primitive_id &inputName, const
         _nw_ptr->set_input_data(internalName, memory);
     } else if (prec == Precision::I16 || prec == Precision::U16) {
         // clDNN doesn't support I16 input precision, so we always have to convert input data to fp32 precision
-        const cldnn::memory& fp32_mem = inputsMemory.at(inputName+fp32_suffix);
-        cldnn::pointer<float> ptr = fp32_mem.pointer<float>();
+        cldnn::memory::ptr fp32_mem = inputsMemory.at(inputName+fp32_suffix);
+        cldnn::mem_lock<float> ptr {fp32_mem, stream};
         if (prec == Precision::I16) {
             copyToFloat<int16_t>(ptr.data(), &inputBlob);
         } else {
@@ -1031,4 +940,4 @@ void CLDNNInferRequest::PrepareInputDyn(const cldnn::primitive_id &inputName, co
     }
 }
 
-};  // namespace CLDNNPlugin
+}  // namespace CLDNNPlugin
diff --git a/inference-engine/src/cldnn_engine/cldnn_infer_request.h b/inference-engine/src/cldnn_engine/cldnn_infer_request.h
index f9ec4d94db54f5..a988438e8d6657 100644
--- a/inference-engine/src/cldnn_engine/cldnn_infer_request.h
+++ b/inference-engine/src/cldnn_engine/cldnn_infer_request.h
@@ -46,7 +46,7 @@ class CLDNNInferRequest : public InferenceEngine::IInferRequestInternal {
     void EnableStreams() { m_useStreams = true; }
 
 protected:
-    std::map<std::string, cldnn::memory> inputsMemory;
+    std::map<std::string, cldnn::memory::ptr> inputsMemory;
     std::map<std::string, cldnn::primitive_id> outputsMap;
 
     bool m_useProfiling;
@@ -60,12 +60,12 @@ class CLDNNInferRequest : public InferenceEngine::IInferRequestInternal {
 
     InferenceEngine::Blob::Ptr createInputBlob(const InferenceEngine::TensorDesc& desc, uint8_t* mem_ptr = nullptr);
     InferenceEngine::Blob::Ptr createOutputBlob(const InferenceEngine::TensorDesc& desc, uint8_t* mem_ptr = nullptr);
-    void copyOutputData(const cldnn::memory& outputMemory, InferenceEngine::Blob::Ptr bptr, buf_info* bi = nullptr);
+    void copyOutputData(cldnn::memory::ptr outputMemory, InferenceEngine::Blob::Ptr bptr, buf_info* bi = nullptr);
     void copyInputData(std::shared_ptr<cldnn::network> network, const cldnn::primitive_id &inputName,
                        const cldnn::layout& inputLayout, const InferenceEngine::Blob &inputBlob,
                        buf_info* bi = nullptr);
 
-    void input_attach(cldnn::primitive_id name, cldnn::memory& inputMem);
+    void input_attach(cldnn::primitive_id name, cldnn::memory::ptr inputMem);
     void input_alloc(cldnn::primitive_id name, const cldnn::layout& layout);
     void AllocateInputs();
     void AllocateOutputs();
@@ -76,9 +76,6 @@ class CLDNNInferRequest : public InferenceEngine::IInferRequestInternal {
 
     void PrepareInput(const cldnn::primitive_id &inputName, const InferenceEngine::Blob &inputBlob);
     void PrepareInputDyn(const cldnn::primitive_id &inputName, const InferenceEngine::Blob &inputBlob);
-
-private:
-    static const char fp32_suffix[];
 };
 
 };  // namespace CLDNNPlugin
diff --git a/inference-engine/src/cldnn_engine/cldnn_program.cpp b/inference-engine/src/cldnn_engine/cldnn_program.cpp
index d30434397fdedc..6ff0d4ecef3169 100644
--- a/inference-engine/src/cldnn_engine/cldnn_program.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_program.cpp
@@ -92,7 +92,7 @@ bool Program::CanProcessDynBatch(std::vector<std::shared_ptr<ngraph::Node>> ops,
     return true;
 }
 
-Program::Program(InferenceEngine::CNNNetwork& network, std::shared_ptr<const cldnn::engine> engine, const Config& config, bool createTopologyOnly)
+Program::Program(InferenceEngine::CNNNetwork& network, std::shared_ptr<cldnn::engine> engine, const Config& config, bool createTopologyOnly)
     : m_config(config)
     , m_engine(engine)
     , m_curBatch(-1)
@@ -128,11 +128,9 @@ Program::Program(InferenceEngine::CNNNetwork& network, std::shared_ptr<const cld
 
             ChangeInputBatch(1U << static_cast<unsigned>(b));
             m_programs.insert(m_programs.begin(), BuildProgram(ops, networkInputs, networkOutputs, createTopologyOnly));
-            m_engine->release_pending_memory(0);
         }
     } else {
         m_programs.emplace_back(BuildProgram(ops, networkInputs, networkOutputs, createTopologyOnly));
-        m_engine->release_pending_memory(0);
     }
 }
 
diff --git a/inference-engine/src/cldnn_engine/cldnn_program.h b/inference-engine/src/cldnn_engine/cldnn_program.h
index a5299d810f9a03..8f90b4fabb7712 100644
--- a/inference-engine/src/cldnn_engine/cldnn_program.h
+++ b/inference-engine/src/cldnn_engine/cldnn_program.h
@@ -15,8 +15,8 @@
 
 #include "cldnn_config.h"
 
-#include <api/engine.hpp>
-#include <api/topology.hpp>
+#include <cldnn/runtime/engine.hpp>
+#include <cldnn/graph/topology.hpp>
 
 // Forward declarations for cldnn part
 namespace cldnn {
@@ -69,8 +69,8 @@ struct PerfCounter {
 
 class Program {
 public:
-    Program(InferenceEngine::CNNNetwork& network, std::shared_ptr<const cldnn::engine> engine, const Config& config, bool createTopologyOnly = false);
-    Program(std::shared_ptr<const cldnn::engine> engine, const Config& config) : m_config(config), m_engine(engine),
+    Program(InferenceEngine::CNNNetwork& network, std::shared_ptr<cldnn::engine> engine, const Config& config, bool createTopologyOnly = false);
+    Program(std::shared_ptr<cldnn::engine> engine, const Config& config) : m_config(config), m_engine(engine),
             m_curBatch(-1), queryMode(false), m_max_batch(1) {}
     Program() : m_config({}), m_engine(nullptr), m_curBatch(-1), queryMode(false), m_max_batch(1) {}
 
@@ -100,8 +100,8 @@ class Program {
     const std::map<std::string, cldnn::layout>& GetInputLayouts() const { return inputLayouts; }
     InferenceEngine::InputsDataMap GetNetworkInputs() const { return m_networkInputs; }
     InferenceEngine::OutputsDataMap GetNetworkOutputs() const { return m_networkOutputs; }
-    const cldnn::engine& GetEngine() const { return *m_engine; }
-    std::shared_ptr<const cldnn::engine> GetEnginePtr() const { return m_engine; }
+    cldnn::engine& GetEngine() const { return *m_engine; }
+    std::shared_ptr<cldnn::engine> GetEnginePtr() const { return m_engine; }
     const Config& GetConfig() const { return m_config; }
     int GetMaxBatchSizeForSingleProgram();
 
@@ -150,7 +150,7 @@ class Program {
 private:
     static factories_map_t factories_map;
     std::vector<std::shared_ptr<cldnn::program>> m_programs;
-    std::shared_ptr<const cldnn::engine> m_engine;
+    std::shared_ptr<cldnn::engine> m_engine;
     Config m_config;
 
     std::shared_ptr<cldnn::topology> m_topology;
diff --git a/inference-engine/src/cldnn_engine/cldnn_remote_context.cpp b/inference-engine/src/cldnn_engine/cldnn_remote_context.cpp
index f03db1c4834e0d..ce52a5eea074bd 100644
--- a/inference-engine/src/cldnn_engine/cldnn_remote_context.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_remote_context.cpp
@@ -6,21 +6,23 @@
 #include "cldnn_remote_context.h"
 #include "cldnn_itt.h"
 
+#include "cldnn/runtime/device_query.hpp"
+
 using namespace InferenceEngine;
 using namespace InferenceEngine::gpu;
 using namespace InferenceEngine::details;
 
 namespace CLDNNPlugin {
-static const char unsupported_str[] = "Unsupported shared object type ";
 CLDNNRemoteAllocator CLDNNRemoteBlobImpl::m_allocator;
 
 CLDNNRemoteBlobImpl::CLDNNRemoteBlobImpl(ClContext::Ptr context,
+    cldnn::stream& stream,
     const cldnn::layout& layout,
     cldnn::shared_handle mem,
     cldnn::shared_surface surf,
     uint32_t plane,
     BlobType mem_type) :
-    m_context(context), m_layout(layout), m_mem_type(mem_type), m_mem(mem), m_surf(surf), m_plane(plane),
+    m_context(context), m_stream(stream), m_layout(layout), m_mem_type(mem_type), m_mem(mem), m_surf(surf), m_plane(plane),
     _handle(nullptr), _allocator(nullptr), m_memObject(nullptr), lockedHolder(nullptr) {
 }
 
@@ -67,8 +69,7 @@ ParamMap CLDNNRemoteBlobImpl::getParams() const {
 }
 
 bool CLDNNRemoteBlobImpl::deallocate() noexcept {
-    if (m_memObject != nullptr)
-        m_memObject.reset();
+    m_memObject.reset();
     return m_memObject == nullptr;
 }
 
@@ -86,32 +87,7 @@ void CLDNNRemoteBlobImpl::allocate_if_needed() {
     _impl->acquire_lock();
 
     if (m_memObject == nullptr) {
-        auto eng = _impl->GetEngine();
-        switch (m_mem_type) {
-        case BlobType::BT_BUF_INTERNAL:
-            m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::allocate(*eng, m_layout)));
-            break;
-        case BlobType::BT_BUF_SHARED:
-            m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::share_buffer(*eng, m_layout, m_mem)));
-            break;
-#ifdef _WIN32
-        case BlobType::BT_SURF_SHARED:
-            m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::share_surface(*eng, m_layout, m_mem, m_plane)));
-            break;
-        case BlobType::BT_DX_BUF_SHARED:
-            m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::share_dx_buffer(*eng, m_layout, m_mem)));
-            break;
-#else
-        case BlobType::BT_SURF_SHARED:
-            m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::share_surface(*eng, m_layout, m_surf, m_plane)));
-            break;
-#endif
-        case BlobType::BT_IMG_SHARED:
-            m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::share_image(*eng, m_layout, m_mem)));
-            break;
-        default:
-            IE_THROW() << unsupported_str << m_mem_type;
-        }
+        allocate();
     }
 
     _impl->release_lock();
@@ -120,32 +96,38 @@ void CLDNNRemoteBlobImpl::allocate_if_needed() {
 void CLDNNRemoteBlobImpl::allocate() noexcept {
     assert(m_memObject == nullptr);
 
-    std::shared_ptr<const cldnn::engine> eng = getContextImpl(m_context.lock())->GetEngine();
+    std::shared_ptr<cldnn::engine> eng = getContextImpl(m_context.lock())->GetEngine();
 
     switch (m_mem_type) {
-    case BlobType::BT_BUF_INTERNAL:
-        m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::allocate(*eng, m_layout)));
+    case BlobType::BT_BUF_INTERNAL: {
+        m_memObject = eng->allocate_memory(m_layout);
         break;
-    case BlobType::BT_BUF_SHARED:
-        m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::share_buffer(*eng, m_layout, m_mem)));
+    }
+    case BlobType::BT_BUF_SHARED: {
+        m_memObject = eng->share_buffer(m_layout, m_mem);
         break;
+    }
 #ifdef _WIN32
-    case BlobType::BT_SURF_SHARED:
-        m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::share_surface(*eng, m_layout, m_mem, m_plane)));
+    case BlobType::BT_SURF_SHARED: {
+        m_memObject = eng->share_surface(m_layout, m_mem, m_plane);
         break;
-    case BlobType::BT_DX_BUF_SHARED:
-        m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::share_dx_buffer(*eng, m_layout, m_mem)));
+    }
+    case BlobType::BT_DX_BUF_SHARED: {
+        m_memObject = eng->share_dx_buffer(m_layout, m_mem);
         break;
+    }
 #else
-    case BlobType::BT_SURF_SHARED:
-        m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::share_surface(*eng, m_layout, m_surf, m_plane)));
+    case BlobType::BT_SURF_SHARED: {
+        m_memObject = eng->share_surface(m_layout, m_surf, m_plane);
         break;
+    }
 #endif
-    case BlobType::BT_IMG_SHARED:
-        m_memObject = std::unique_ptr<cldnn::memory>(new cldnn::memory(cldnn::memory::share_image(*eng, m_layout, m_mem)));
+    case BlobType::BT_IMG_SHARED: {
+        m_memObject = eng->share_image(m_layout, m_mem);
         break;
+    }
     default:
-        m_memObject = nullptr;
+        m_memObject.reset();
     }
 }
 
@@ -165,7 +147,7 @@ std::shared_ptr<RemoteContext> CLDNNRemoteBlobImpl::getContext() const noexcept
 }
 
 void CLDNNRemoteBlobImpl::lock() const {
-    lockedHolder = std::unique_ptr<cldnn::pointer<uint8_t>>(new cldnn::pointer<uint8_t>(m_memObject->pointer<uint8_t>()));
+    lockedHolder = std::unique_ptr<cldnn::mem_lock<uint8_t>>(new cldnn::mem_lock<uint8_t>(m_memObject, m_stream));
     auto ptr = lockedHolder->data();
     _handle = reinterpret_cast<void*>(ptr);
     m_allocator.regLockedBlob(_handle, this);
@@ -244,7 +226,11 @@ CLDNNExecutionContextImpl::CLDNNExecutionContextImpl(const std::shared_ptr<IInfe
         }
     }
 
-    cldnn::device_query device_query(_context_id, _va_device);
+    // TODO: Parameterize this based on plugin config and compilation options
+    auto engine_type = cldnn::engine_types::ocl;
+    auto runtime_type = cldnn::runtime_types::ocl;
+    // Use actual runtime and engine types
+    cldnn::device_query device_query(engine_type, runtime_type, _context_id, _va_device);
     auto device_map = device_query.get_available_devices();
 
     auto iter = device_map.find(m_config.device_id);
@@ -252,28 +238,25 @@ CLDNNExecutionContextImpl::CLDNNExecutionContextImpl(const std::shared_ptr<IInfe
 
     {
         OV_ITT_SCOPED_TASK(itt::domains::CLDNNPlugin, "CLDNNExecutionContextImpl::Create");
-        m_engine = std::make_shared<cldnn::engine>(dev,
-            cldnn::engine_configuration((m_config.useProfiling ||
+        bool enable_profiling = (m_config.useProfiling ||
                 (m_config.tuningConfig.mode == cldnn::tuning_mode::tuning_tune_and_cache) ||
-                (m_config.tuningConfig.mode == cldnn::tuning_mode::tuning_retune_and_cache)),
-                false,
-                m_config.dumpCustomKernels,
-                std::string(),
-                std::string(),
-                true,
-                std::string(),
-                m_config.sources_dumps_dir,
-                m_config.queuePriority,
-                m_config.queueThrottle,
-                m_config.memory_pool_on,
-                m_config.throughput_streams,
-                m_config.kernels_cache_dir,
-                m_config.n_threads));
+                (m_config.tuningConfig.mode == cldnn::tuning_mode::tuning_retune_and_cache));
+        cldnn::queue_types queue_type = cldnn::queue_types::out_of_order;
+        bool use_unified_shared_memory = true;
+        m_engine = cldnn::engine::create(engine_type, runtime_type, dev, cldnn::engine_configuration(enable_profiling,
+                                                                                                     queue_type,
+                                                                                                     m_config.sources_dumps_dir,
+                                                                                                     m_config.queuePriority,
+                                                                                                     m_config.queueThrottle,
+                                                                                                     m_config.memory_pool_on,
+                                                                                                     use_unified_shared_memory,
+                                                                                                     m_config.kernels_cache_dir,
+                                                                                                     m_config.n_threads));
     }
 }
 
 ParamMap CLDNNExecutionContextImpl::getParams() const {
-    ParamMap ret = { { GPU_PARAM_KEY(OCL_CONTEXT), m_engine->get_context() } };
+    ParamMap ret = { { GPU_PARAM_KEY(OCL_CONTEXT), m_engine->get_user_context() } };
 
     switch (m_type) {
     case OCL:
diff --git a/inference-engine/src/cldnn_engine/cldnn_remote_context.h b/inference-engine/src/cldnn_engine/cldnn_remote_context.h
index a2ce1729bd10fc..f6a92e82c48764 100644
--- a/inference-engine/src/cldnn_engine/cldnn_remote_context.h
+++ b/inference-engine/src/cldnn_engine/cldnn_remote_context.h
@@ -4,15 +4,11 @@
 
 #pragma once
 
-#include <string>
-#include <map>
-#include <memory>
-#include <atomic>
+#include <cldnn/runtime/memory.hpp>
+#include <cldnn/runtime/engine.hpp>
 #include <ie_parameter.hpp>
 #include <cpp_interfaces/interface/ie_iplugin_internal.hpp>
 #include "cldnn_config.h"
-#include <api/memory.hpp>
-#include <api/engine.hpp>
 #include "cldnn_common_utils.h"
 
 #ifndef NOMINMAX
@@ -25,6 +21,11 @@
 # include <gpu/gpu_context_api_va.hpp>
 #endif
 
+#include <string>
+#include <map>
+#include <memory>
+#include <atomic>
+
 namespace CLDNNPlugin {
 class CLDNNRemoteAllocator;
 
@@ -41,6 +42,7 @@ class CLDNNRemoteBlobImpl : public InferenceEngine::gpu::details::param_map_obj_
     };
 
     explicit CLDNNRemoteBlobImpl(InferenceEngine::gpu::ClContext::Ptr context,
+                                 cldnn::stream& stream,
                                  const cldnn::layout& layout,
                                  cldnn::shared_handle mem,
                                  cldnn::shared_surface surf,
@@ -63,11 +65,12 @@ class CLDNNRemoteBlobImpl : public InferenceEngine::gpu::details::param_map_obj_
     bool is_allocated() const noexcept;
     bool is_locked() const noexcept;
     void allocate_if_needed();
-    cldnn::memory& getMemory() { return *m_memObject; }
+    cldnn::memory::ptr getMemory() { return m_memObject; }
 
 protected:
     static CLDNNRemoteAllocator m_allocator;
     std::weak_ptr<InferenceEngine::gpu::ClContext> m_context;
+    cldnn::stream& m_stream;
 
     // constructor stuff
     cldnn::shared_handle m_mem;
@@ -77,9 +80,9 @@ class CLDNNRemoteBlobImpl : public InferenceEngine::gpu::details::param_map_obj_
     cldnn::layout m_layout;
     BlobType m_mem_type;
 
-    std::unique_ptr<cldnn::memory> m_memObject;
+    cldnn::memory::ptr m_memObject;
 
-    mutable std::unique_ptr<cldnn::pointer<uint8_t>> lockedHolder;
+    mutable std::unique_ptr<cldnn::mem_lock<uint8_t>> lockedHolder;
     mutable void* _handle;
     mutable std::shared_ptr<InferenceEngine::IAllocator> _allocator;
 
@@ -93,13 +96,14 @@ class typedCLDNNRemoteBlob : public TpublicAPI {
     using Ptr = std::shared_ptr<typedCLDNNRemoteBlob>;
 
     explicit typedCLDNNRemoteBlob(InferenceEngine::gpu::ClContext::Ptr context,
+                                  cldnn::stream& stream,
                                   const InferenceEngine::TensorDesc& desc,
                                   const cldnn::layout& layout,
                                   cldnn::shared_handle mem,
                                   cldnn::shared_surface surf,
                                   uint32_t plane,
                                   CLDNNRemoteBlobImpl::BlobType mem_type)
-        : _impl(context, layout, mem, surf, plane, mem_type)
+        : _impl(context, stream, layout, mem, surf, plane, mem_type)
         , TpublicAPI(desc) {}
 
     void allocate() noexcept override { _impl.allocate(); }
@@ -231,6 +235,7 @@ class CLDNNExecutionContextImpl : public InferenceEngine::gpu::details::param_ma
     }
 
 protected:
+    // TODO: refactor to unique_ptr
     std::shared_ptr<cldnn::engine> m_engine;
     InferenceEngine::gpu_handle_param m_va_display;
     Config m_config;
@@ -267,6 +272,7 @@ class typedCLDNNExecutionContext : public TpublicContextAPI,
         using namespace InferenceEngine;
         using InferenceEngine::gpu::details::param_map_obj_getter;
         InferenceEngine::RemoteBlob::Ptr ret = nullptr;
+        auto& stream = _impl.GetEngine()->get_program_stream();
         uint32_t plane = param_map_obj_getter::_ObjFromParamSimple<uint32_t>(params, GPU_PARAM_KEY(VA_PLANE));
 #ifdef _WIN32
         cldnn::shared_handle mem = param_map_obj_getter::_ObjFromParamSimple<cldnn::shared_handle>(params, GPU_PARAM_KEY(DEV_OBJECT_HANDLE));
@@ -290,11 +296,11 @@ class typedCLDNNExecutionContext : public TpublicContextAPI,
                 std::dynamic_pointer_cast<InferenceEngine::gpu::ClContext>
                 (std::enable_shared_from_this<typedCLDNNExecutionContext<TpublicContextAPI>>::shared_from_this());
 #ifdef _WIN32
-            ret = std::make_shared<CLDNNRemoteD3DSurface>(smart_this,
+            ret = std::make_shared<CLDNNRemoteD3DSurface>(smart_this, stream,
                 tensorDesc, layout, mem, 0, plane,
                 CLDNNRemoteBlobImpl::BlobType::BT_SURF_SHARED);
 #else
-            ret = std::make_shared<CLDNNRemoteVASurface>(smart_this,
+            ret = std::make_shared<CLDNNRemoteVASurface>(smart_this, stream,
                 tensorDesc, layout, nullptr, surf, plane,
                 CLDNNRemoteBlobImpl::BlobType::BT_SURF_SHARED);
 #endif
@@ -311,6 +317,7 @@ class typedCLDNNExecutionContext : public TpublicContextAPI,
         InferenceEngine::RemoteBlob::Ptr ret = nullptr;
 
         _impl.acquire_lock();
+        auto& stream = _impl.GetEngine()->get_program_stream();
 
         // try to locate previously shared object
         auto itr = shared_obj_reg.find(mem);
@@ -327,15 +334,15 @@ class typedCLDNNExecutionContext : public TpublicContextAPI,
 
             switch (blob_type) {
             case CLDNNRemoteBlobImpl::BlobType::BT_BUF_SHARED:
-                ret = std::make_shared<CLDNNRemoteCLbuffer>(smart_this, tensorDesc, layout, mem, 0, 0, blob_type);
+                ret = std::make_shared<CLDNNRemoteCLbuffer>(smart_this, stream, tensorDesc, layout, mem, 0, 0, blob_type);
                 break;
             case CLDNNRemoteBlobImpl::BlobType::BT_IMG_SHARED:
                 layout.format = ImageFormatFromLayout(tensorDesc.getLayout());
-                ret = std::make_shared<CLDNNRemoteCLImage2D>(smart_this, tensorDesc, layout, mem, 0, 0, blob_type);
+                ret = std::make_shared<CLDNNRemoteCLImage2D>(smart_this, stream, tensorDesc, layout, mem, 0, 0, blob_type);
                 break;
 #ifdef _WIN32
             case CLDNNRemoteBlobImpl::BlobType::BT_DX_BUF_SHARED:
-                ret = std::make_shared<CLDNNRemoteD3DBuffer>(smart_this, tensorDesc, layout, mem, 0, 0, blob_type);
+                ret = std::make_shared<CLDNNRemoteD3DBuffer>(smart_this, stream, tensorDesc, layout, mem, 0, 0, blob_type);
                 break;
 #endif
             default:
@@ -354,7 +361,9 @@ class typedCLDNNExecutionContext : public TpublicContextAPI,
                              CldnnTensorFromIEDims(tensorDesc.getDims()));
         auto smart_this = std::dynamic_pointer_cast<InferenceEngine::gpu::ClContext>
             (std::enable_shared_from_this<typedCLDNNExecutionContext<TpublicContextAPI>>::shared_from_this());
+        auto& stream = _impl.GetEngine()->get_program_stream();
         return std::make_shared<CLDNNRemoteCLbuffer>(smart_this,
+                                                     stream,
                                                      tensorDesc,
                                                      layout,
                                                      nullptr, 0, 0,
diff --git a/inference-engine/src/cldnn_engine/ops/batch_to_space.cpp b/inference-engine/src/cldnn_engine/ops/batch_to_space.cpp
index 51d70e05f1a525..e46643f0617ca9 100644
--- a/inference-engine/src/cldnn_engine/ops/batch_to_space.cpp
+++ b/inference-engine/src/cldnn_engine/ops/batch_to_space.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/batch_to_space.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/batch_to_space.hpp"
+#include "cldnn/primitives/batch_to_space.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/broadcast.cpp b/inference-engine/src/cldnn_engine/ops/broadcast.cpp
index af0ed9b7b751da..6d6e63032005c3 100644
--- a/inference-engine/src/cldnn_engine/ops/broadcast.cpp
+++ b/inference-engine/src/cldnn_engine/ops/broadcast.cpp
@@ -8,9 +8,9 @@
 #include "ngraph/op/broadcast.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/broadcast.hpp"
-#include "api/reorder.hpp"
-#include "api/reshape.hpp"
+#include "cldnn/primitives/broadcast.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/reshape.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/concat.cpp b/inference-engine/src/cldnn_engine/ops/concat.cpp
index 5a300c3dc8f394..9d37f959f03fb7 100644
--- a/inference-engine/src/cldnn_engine/ops/concat.cpp
+++ b/inference-engine/src/cldnn_engine/ops/concat.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/concat.hpp"
 
-#include "api/concatenation.hpp"
+#include "cldnn/primitives/concatenation.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/constant.cpp b/inference-engine/src/cldnn_engine/ops/constant.cpp
index e8630f67a6bc31..fea42f31d98227 100644
--- a/inference-engine/src/cldnn_engine/ops/constant.cpp
+++ b/inference-engine/src/cldnn_engine/ops/constant.cpp
@@ -17,7 +17,7 @@
 #include "ngraph/op/variadic_split.hpp"
 #include "ngraph/op/util/op_types.hpp"
 
-#include "api/data.hpp"
+#include "cldnn/primitives/data.hpp"
 
 namespace CLDNNPlugin {
 
@@ -169,9 +169,10 @@ void CreateConstantOp(Program& p, const std::shared_ptr<ngraph::op::v0::Constant
     if (bufIter != p.blobMemCache.end()) {
         constPrimID = bufIter->second;
     } else {
-        auto mem = cldnn::memory::allocate(p.GetEngine(), constLayout, 0, false);
-        auto tmpPointer = mem.pointer<char>();  // implicitly maps buffer - unmap in destructor
-        auto buf = tmpPointer.data();
+        cldnn::memory::ptr mem = p.GetEngine().allocate_memory(constLayout, false);
+        auto& stream = p.GetEngine().get_program_stream();
+        cldnn::mem_lock<char> lock{mem, stream};
+        auto buf = lock.data();
         auto bufSize = constLayout.bytes_count();
 
         // Do actual weights reorder and change O and I channels order
diff --git a/inference-engine/src/cldnn_engine/ops/convert.cpp b/inference-engine/src/cldnn_engine/ops/convert.cpp
index 29fb037258f092..6af5bee759d683 100644
--- a/inference-engine/src/cldnn_engine/ops/convert.cpp
+++ b/inference-engine/src/cldnn_engine/ops/convert.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/convert.hpp"
 #include "ngraph/op/convert_like.hpp"
 
-#include "api/reorder.hpp"
+#include "cldnn/primitives/reorder.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/convolution.cpp b/inference-engine/src/cldnn_engine/ops/convolution.cpp
index 3207dab623aefa..83f536a68b72fb 100644
--- a/inference-engine/src/cldnn_engine/ops/convolution.cpp
+++ b/inference-engine/src/cldnn_engine/ops/convolution.cpp
@@ -13,11 +13,11 @@
 #include "ngraph/op/fake_quantize.hpp"
 #include "ngraph/op/util/op_types.hpp"
 
-#include "api/convolution.hpp"
-#include "api/deconvolution.hpp"
-#include "api/binary_convolution.hpp"
-#include "api/permute.hpp"
-#include "api/reorder.hpp"
+#include "cldnn/primitives/convolution.hpp"
+#include "cldnn/primitives/deconvolution.hpp"
+#include "cldnn/primitives/binary_convolution.hpp"
+#include "cldnn/primitives/permute.hpp"
+#include "cldnn/primitives/reorder.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/ctc_greedy_decoder.cpp b/inference-engine/src/cldnn_engine/ops/ctc_greedy_decoder.cpp
index 2887360724535e..c8bd8d54e078ea 100644
--- a/inference-engine/src/cldnn_engine/ops/ctc_greedy_decoder.cpp
+++ b/inference-engine/src/cldnn_engine/ops/ctc_greedy_decoder.cpp
@@ -8,9 +8,9 @@
 #include "ngraph/op/ctc_greedy_decoder.hpp"
 #include "ngraph/op/ctc_greedy_decoder_seq_len.hpp"
 
-#include "api/ctc_greedy_decoder.hpp"
-#include "api/reorder.hpp"
-#include "api/mutable_data.hpp"
+#include "cldnn/primitives/ctc_greedy_decoder.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/mutable_data.hpp"
 
 #include "transformations/utils/utils.hpp"
 
@@ -58,7 +58,7 @@ void CreateCommonCTCGreedyDecoderOp(Program& p, const std::shared_ptr<ngraph::No
 
     std::size_t num_output = op->get_output_size();
 
-    std::vector<cldnn::memory> shared_memory;
+    std::vector<cldnn::memory::ptr> shared_memory;
     if (num_output == 2) {
         auto mutable_precision = op->get_output_element_type(1);
          if (mutable_precision == ngraph::element::i64) {
@@ -70,7 +70,7 @@ void CreateCommonCTCGreedyDecoderOp(Program& p, const std::shared_ptr<ngraph::No
             DefaultFormatForDims(op->get_output_shape(1).size()),
             CldnnTensorFromIEDims(op->get_output_shape(1)));
 
-        shared_memory.emplace_back(cldnn::memory::allocate(p.GetEngine(), mutableLayout));
+        shared_memory.emplace_back(p.GetEngine().allocate_memory(mutableLayout));
 
         cldnn::primitive_id ctc_gd_mutable_id_w = layer_type_name_ID(op) + "_md_write";
         auto ctc_gd_mutable_prim = cldnn::mutable_data(ctc_gd_mutable_id_w, shared_memory[0]);
diff --git a/inference-engine/src/cldnn_engine/ops/cum_sum.cpp b/inference-engine/src/cldnn_engine/ops/cum_sum.cpp
index 6252a258269899..1bdcec2957eda5 100644
--- a/inference-engine/src/cldnn_engine/ops/cum_sum.cpp
+++ b/inference-engine/src/cldnn_engine/ops/cum_sum.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/cum_sum.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/cum_sum.hpp"
+#include "cldnn/primitives/cum_sum.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/custom.cpp b/inference-engine/src/cldnn_engine/ops/custom.cpp
index 5211fc94efd3aa..85945bfbdb9698 100644
--- a/inference-engine/src/cldnn_engine/ops/custom.cpp
+++ b/inference-engine/src/cldnn_engine/ops/custom.cpp
@@ -9,8 +9,8 @@
 #include "ngraph/attribute_visitor.hpp"
 #include "ngraph/node.hpp"
 
-#include "api/custom_gpu_primitive.hpp"
-#include "api/reorder.hpp"
+#include "cldnn/primitives/custom_gpu_primitive.hpp"
+#include "cldnn/primitives/reorder.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/depth_to_space.cpp b/inference-engine/src/cldnn_engine/ops/depth_to_space.cpp
index 4c407a126d50e0..b53262ab23dec7 100644
--- a/inference-engine/src/cldnn_engine/ops/depth_to_space.cpp
+++ b/inference-engine/src/cldnn_engine/ops/depth_to_space.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/depth_to_space.hpp"
 
-#include "api/depth_to_space.hpp"
+#include "cldnn/primitives/depth_to_space.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/detection_output.cpp b/inference-engine/src/cldnn_engine/ops/detection_output.cpp
index 8bbe102bcfd05d..aa2b505f0e76df 100644
--- a/inference-engine/src/cldnn_engine/ops/detection_output.cpp
+++ b/inference-engine/src/cldnn_engine/ops/detection_output.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/detection_output.hpp"
 
-#include "api/detection_output.hpp"
+#include "cldnn/primitives/detection_output.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/eltwise.cpp b/inference-engine/src/cldnn_engine/ops/eltwise.cpp
index 66dcca7b459c47..817512d5bf92ec 100644
--- a/inference-engine/src/cldnn_engine/ops/eltwise.cpp
+++ b/inference-engine/src/cldnn_engine/ops/eltwise.cpp
@@ -25,10 +25,10 @@
 #include "ngraph/op/power.hpp"
 #include "ngraph/op/floor_mod.hpp"
 
-#include "api/activation.hpp"
-#include "api/eltwise.hpp"
-#include "api/reorder.hpp"
-#include "api/reshape.hpp"
+#include "cldnn/primitives/activation.hpp"
+#include "cldnn/primitives/eltwise.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/reshape.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/embedding_bag.cpp b/inference-engine/src/cldnn_engine/ops/embedding_bag.cpp
index 35b9f00096d009..2e97a60aebf8be 100644
--- a/inference-engine/src/cldnn_engine/ops/embedding_bag.cpp
+++ b/inference-engine/src/cldnn_engine/ops/embedding_bag.cpp
@@ -9,8 +9,8 @@
 #include "ngraph/op/embeddingbag_offsets_sum.hpp"
 #include "ngraph/op/embeddingbag_packedsum.hpp"
 
-#include "api/embedding_bag.hpp"
-#include "api/reorder.hpp"
+#include "cldnn/primitives/embedding_bag.hpp"
+#include "cldnn/primitives/reorder.hpp"
 
 #include "transformations/utils/utils.hpp"
 
diff --git a/inference-engine/src/cldnn_engine/ops/extract_image_patches.cpp b/inference-engine/src/cldnn_engine/ops/extract_image_patches.cpp
index 42f7c4e8e99666..23b5f01432001a 100644
--- a/inference-engine/src/cldnn_engine/ops/extract_image_patches.cpp
+++ b/inference-engine/src/cldnn_engine/ops/extract_image_patches.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/extractimagepatches.hpp"
 
-#include "api/extract_image_patches.hpp"
+#include "cldnn/primitives/extract_image_patches.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/fake_quantize.cpp b/inference-engine/src/cldnn_engine/ops/fake_quantize.cpp
index ca3d950694c4f3..345a70f34bbce3 100644
--- a/inference-engine/src/cldnn_engine/ops/fake_quantize.cpp
+++ b/inference-engine/src/cldnn_engine/ops/fake_quantize.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/fake_quantize.hpp"
 
-#include "api/quantize.hpp"
+#include "cldnn/primitives/quantize.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/gather tree.cpp b/inference-engine/src/cldnn_engine/ops/gather tree.cpp
index a6c806bcbe8c1b..6b73131fd29d84 100644
--- a/inference-engine/src/cldnn_engine/ops/gather tree.cpp	
+++ b/inference-engine/src/cldnn_engine/ops/gather tree.cpp	
@@ -7,8 +7,8 @@
 
 #include "ngraph/op/gather_tree.hpp"
 
-#include "api/gather_tree.hpp"
-#include "api/reorder.hpp"
+#include "cldnn/primitives/gather_tree.hpp"
+#include "cldnn/primitives/reorder.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/gather.cpp b/inference-engine/src/cldnn_engine/ops/gather.cpp
index b80e26661e5353..362854cc32aaeb 100644
--- a/inference-engine/src/cldnn_engine/ops/gather.cpp
+++ b/inference-engine/src/cldnn_engine/ops/gather.cpp
@@ -7,8 +7,8 @@
 
 #include "ngraph/op/gather.hpp"
 
-#include "api/gather.hpp"
-#include "api/reorder.hpp"
+#include "cldnn/primitives/gather.hpp"
+#include "cldnn/primitives/reorder.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/gather_nd.cpp b/inference-engine/src/cldnn_engine/ops/gather_nd.cpp
index 6a1cd65132928e..cbdc5659bb3197 100644
--- a/inference-engine/src/cldnn_engine/ops/gather_nd.cpp
+++ b/inference-engine/src/cldnn_engine/ops/gather_nd.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/gather_nd.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/gather_nd.hpp"
+#include "cldnn/primitives/gather_nd.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/grn.cpp b/inference-engine/src/cldnn_engine/ops/grn.cpp
index 3eb750f85a920c..960dd03494735d 100644
--- a/inference-engine/src/cldnn_engine/ops/grn.cpp
+++ b/inference-engine/src/cldnn_engine/ops/grn.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/grn.hpp"
 
-#include "api/grn.hpp"
+#include "cldnn/primitives/grn.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/interpolate.cpp b/inference-engine/src/cldnn_engine/ops/interpolate.cpp
index f9241b8ca0fd75..b77999289c631b 100644
--- a/inference-engine/src/cldnn_engine/ops/interpolate.cpp
+++ b/inference-engine/src/cldnn_engine/ops/interpolate.cpp
@@ -9,7 +9,7 @@
 #include "ngraph/op/interpolate.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/resample.hpp"
+#include "cldnn/primitives/resample.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/lrn.cpp b/inference-engine/src/cldnn_engine/ops/lrn.cpp
index f8e7c601645cec..c13c17daaeb13d 100644
--- a/inference-engine/src/cldnn_engine/ops/lrn.cpp
+++ b/inference-engine/src/cldnn_engine/ops/lrn.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/lrn.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/lrn.hpp"
+#include "cldnn/primitives/lrn.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/matmul.cpp b/inference-engine/src/cldnn_engine/ops/matmul.cpp
index 5fe6f1432586c3..a8818c9e6f67ee 100644
--- a/inference-engine/src/cldnn_engine/ops/matmul.cpp
+++ b/inference-engine/src/cldnn_engine/ops/matmul.cpp
@@ -9,11 +9,11 @@
 #include "ngraph/op/constant.hpp"
 #include "ngraph/op/fake_quantize.hpp"
 
-#include "api/gemm.hpp"
-#include "api/fully_connected.hpp"
-#include "api/reshape.hpp"
-#include "api/reorder.hpp"
-#include "api/permute.hpp"
+#include "cldnn/primitives/gemm.hpp"
+#include "cldnn/primitives/fully_connected.hpp"
+#include "cldnn/primitives/reshape.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/permute.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/mvn.cpp b/inference-engine/src/cldnn_engine/ops/mvn.cpp
index 79001b812e1854..b9cb376a24e34e 100644
--- a/inference-engine/src/cldnn_engine/ops/mvn.cpp
+++ b/inference-engine/src/cldnn_engine/ops/mvn.cpp
@@ -8,7 +8,8 @@
 #include "ngraph/op/mvn.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/mvn.hpp"
+#include "cldnn/primitives/mvn.hpp"
+
 #include <algorithm>
 
 namespace CLDNNPlugin {
diff --git a/inference-engine/src/cldnn_engine/ops/non_max_suppression.cpp b/inference-engine/src/cldnn_engine/ops/non_max_suppression.cpp
index 8b124309467e10..a3d4834c51eed1 100644
--- a/inference-engine/src/cldnn_engine/ops/non_max_suppression.cpp
+++ b/inference-engine/src/cldnn_engine/ops/non_max_suppression.cpp
@@ -9,9 +9,9 @@
 #include <ngraph/opsets/opset3.hpp>
 #include <ngraph_ops/nms_ie_internal.hpp>
 
-#include "api/reorder.hpp"
-#include "api/mutable_data.hpp"
-#include "api/non_max_suppression.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/mutable_data.hpp"
+#include "cldnn/primitives/non_max_suppression.hpp"
 
 namespace CLDNNPlugin {
 
@@ -62,7 +62,7 @@ void CreateNonMaxSuppressionIEInternalOp(Program& p, const std::shared_ptr<ngrap
 
     std::size_t num_output = op->get_output_size();
 
-    std::vector<cldnn::memory> shared_memory;
+    std::vector<cldnn::memory::ptr> shared_memory;
     switch (num_output) {
         case 3: {
             auto mutable_precision_second = op->get_output_element_type(2);
@@ -74,7 +74,7 @@ void CreateNonMaxSuppressionIEInternalOp(Program& p, const std::shared_ptr<ngrap
                 DefaultFormatForDims(op->get_output_shape(2).size()),
                 CldnnTensorFromIEDims(op->get_output_shape(2)));
 
-            shared_memory.emplace_back(cldnn::memory::allocate(p.GetEngine(), mutableLayoutSecond));
+            shared_memory.emplace_back(p.GetEngine().allocate_memory(mutableLayoutSecond));
 
             cldnn::primitive_id non_max_supression_mutable_id_w_second = layer_type_name_ID(op) + "_md_write_second";
             auto nms_mutable_prim_second = cldnn::mutable_data(non_max_supression_mutable_id_w_second, shared_memory.back());
@@ -91,7 +91,7 @@ void CreateNonMaxSuppressionIEInternalOp(Program& p, const std::shared_ptr<ngrap
                 cldnn::format::bfyx,
                 cldnn::tensor(outputIndices, 3, 1, 1));
 
-            shared_memory.emplace_back(cldnn::memory::allocate(p.GetEngine(), mutableLayoutFirst));
+            shared_memory.emplace_back(p.GetEngine().allocate_memory(mutableLayoutFirst));
 
             cldnn::primitive_id non_max_supression_mutable_id_w_first = layer_type_name_ID(op) + "_md_write_first";
             auto nms_mutable_prim_first = cldnn::mutable_data(non_max_supression_mutable_id_w_first, shared_memory.back());
diff --git a/inference-engine/src/cldnn_engine/ops/normalize_l2.cpp b/inference-engine/src/cldnn_engine/ops/normalize_l2.cpp
index 1d679d1dc85f84..85f2eb95de80bc 100644
--- a/inference-engine/src/cldnn_engine/ops/normalize_l2.cpp
+++ b/inference-engine/src/cldnn_engine/ops/normalize_l2.cpp
@@ -8,8 +8,8 @@
 #include "ngraph/op/normalize_l2.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/normalize.hpp"
-#include "api/data.hpp"
+#include "cldnn/primitives/normalize.hpp"
+#include "cldnn/primitives/data.hpp"
 
 namespace CLDNNPlugin {
 
@@ -35,8 +35,8 @@ void CreateNormalizeL2Op(Program& p, const std::shared_ptr<ngraph::op::v0::Norma
     // We create fake scale constant and fill it with ones to keep the same behavior as current primitive
     auto scale = std::make_shared<ngraph::op::v0::Constant>(op->get_output_element_type(0), ngraph::Shape{1}, std::vector<float>{1.0});
     cldnn::layout constLayout = cldnn::layout(DataTypeFromPrecision(op->get_output_element_type(0)), cldnn::format::bfyx, cldnn::tensor{1});
-    auto mem = cldnn::memory::allocate(p.GetEngine(), constLayout, 0, false);
-    auto tmpPointer = mem.pointer<char>();  // implicitly maps buffer - unmap in destructor
+    auto mem = p.GetEngine().allocate_memory(constLayout, false);
+    cldnn::mem_lock<int8_t> tmpPointer{mem, p.GetEngine().get_program_stream()};
     auto buf = tmpPointer.data();
     auto bufSize = scale->get_output_tensor(0).size();
 
diff --git a/inference-engine/src/cldnn_engine/ops/one_hot.cpp b/inference-engine/src/cldnn_engine/ops/one_hot.cpp
index 1076bf595efb0f..3d792bda8aee0e 100644
--- a/inference-engine/src/cldnn_engine/ops/one_hot.cpp
+++ b/inference-engine/src/cldnn_engine/ops/one_hot.cpp
@@ -8,7 +8,7 @@
 
 #include "ngraph/op/one_hot.hpp"
 
-#include "api/one_hot.hpp"
+#include "cldnn/primitives/one_hot.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/pad.cpp b/inference-engine/src/cldnn_engine/ops/pad.cpp
index a3503318a77f79..0d409414b58776 100644
--- a/inference-engine/src/cldnn_engine/ops/pad.cpp
+++ b/inference-engine/src/cldnn_engine/ops/pad.cpp
@@ -8,7 +8,7 @@
 
 #include "ngraph/op/pad.hpp"
 
-#include "api/border.hpp"
+#include "cldnn/primitives/border.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/parameter.cpp b/inference-engine/src/cldnn_engine/ops/parameter.cpp
index 7c61d8261d7881..b68593dd0a5b0f 100644
--- a/inference-engine/src/cldnn_engine/ops/parameter.cpp
+++ b/inference-engine/src/cldnn_engine/ops/parameter.cpp
@@ -7,10 +7,10 @@
 
 #include "ngraph/op/parameter.hpp"
 
-#include "api/input_layout.hpp"
-#include "api/reorder.hpp"
-#include "api/data.hpp"
-#include "api/concatenation.hpp"
+#include "cldnn/primitives/input_layout.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/data.hpp"
+#include "cldnn/primitives/concatenation.hpp"
 
 using namespace InferenceEngine;
 
@@ -158,8 +158,8 @@ void CreateParameterOp(Program& p, const std::shared_ptr<ngraph::op::v0::Paramet
         if (bufIter != p.blobMemCache.end()) {
             meanBlobID = bufIter->second;
         } else {
-            auto mem = cldnn::memory::allocate(p.GetEngine(), meanBlobLayout, 0, false);
-            auto tmpPointer = mem.pointer<char>();  // implicitly maps buffer - unmap in destructor
+            auto mem = p.GetEngine().allocate_memory(meanBlobLayout, false);
+            cldnn::mem_lock<int8_t> tmpPointer{ mem, p.GetEngine().get_program_stream() };
             auto buf = tmpPointer.data();
             auto bufSize = meanBlobLayout.bytes_count();
 
diff --git a/inference-engine/src/cldnn_engine/ops/pooling.cpp b/inference-engine/src/cldnn_engine/ops/pooling.cpp
index 16ca93a6879d3c..f1bf6952292056 100644
--- a/inference-engine/src/cldnn_engine/ops/pooling.cpp
+++ b/inference-engine/src/cldnn_engine/ops/pooling.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/max_pool.hpp"
 #include "ngraph/op/avg_pool.hpp"
 
-#include "api/pooling.hpp"
+#include "cldnn/primitives/pooling.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/prior_box.cpp b/inference-engine/src/cldnn_engine/ops/prior_box.cpp
index 07c6a4ca3ee240..6cf0aaa65355d3 100644
--- a/inference-engine/src/cldnn_engine/ops/prior_box.cpp
+++ b/inference-engine/src/cldnn_engine/ops/prior_box.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/prior_box.hpp"
 #include "ngraph/op/prior_box_clustered.hpp"
 
-#include "api/prior_box.hpp"
+#include "cldnn/primitives/prior_box.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/proposal.cpp b/inference-engine/src/cldnn_engine/ops/proposal.cpp
index 70ff3d154f107d..d5b906e5e6e057 100644
--- a/inference-engine/src/cldnn_engine/ops/proposal.cpp
+++ b/inference-engine/src/cldnn_engine/ops/proposal.cpp
@@ -7,8 +7,8 @@
 
 #include "ngraph/op/proposal.hpp"
 
-#include "api/proposal.hpp"
-#include "api/mutable_data.hpp"
+#include "cldnn/primitives/proposal.hpp"
+#include "cldnn/primitives/mutable_data.hpp"
 
 namespace CLDNNPlugin {
 
@@ -62,7 +62,7 @@ void CreateProposalOp(Program& p, const std::shared_ptr<ngraph::op::v0::Proposal
                                                     DefaultFormatForDims(op->get_output_shape(1).size()),
                                                     CldnnTensorFromIEDims(op->get_output_shape(1)));
 
-        auto shared_memory = cldnn::memory::allocate(p.GetEngine(), mutableLayout);
+        auto shared_memory = p.GetEngine().allocate_memory(mutableLayout);
 
         cldnn::primitive_id proposal_mutable_id_w = layer_type_name_ID(op) + "_md_write";
         auto argmax_mutable_prim = cldnn::mutable_data(proposal_mutable_id_w, shared_memory);
diff --git a/inference-engine/src/cldnn_engine/ops/reduce.cpp b/inference-engine/src/cldnn_engine/ops/reduce.cpp
index 26343ffb813129..b336a2e78faaf5 100644
--- a/inference-engine/src/cldnn_engine/ops/reduce.cpp
+++ b/inference-engine/src/cldnn_engine/ops/reduce.cpp
@@ -16,9 +16,9 @@
 #include "ngraph/op/max.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/reduce.hpp"
-#include "api/reorder.hpp"
-#include "api/reshape.hpp"
+#include "cldnn/primitives/reduce.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/reshape.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/region_yolo.cpp b/inference-engine/src/cldnn_engine/ops/region_yolo.cpp
index 6e5fd660e3fa91..348dd0f7eeb581 100644
--- a/inference-engine/src/cldnn_engine/ops/region_yolo.cpp
+++ b/inference-engine/src/cldnn_engine/ops/region_yolo.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/region_yolo.hpp"
 
-#include "api/region_yolo.hpp"
+#include "cldnn/primitives/region_yolo.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/reorg_yolo.cpp b/inference-engine/src/cldnn_engine/ops/reorg_yolo.cpp
index 7ddc45221a6cfb..4a7f54cf810bc2 100644
--- a/inference-engine/src/cldnn_engine/ops/reorg_yolo.cpp
+++ b/inference-engine/src/cldnn_engine/ops/reorg_yolo.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/reorg_yolo.hpp"
 
-#include "api/reorg_yolo.hpp"
+#include "cldnn/primitives/reorg_yolo.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/reshape.cpp b/inference-engine/src/cldnn_engine/ops/reshape.cpp
index b2111e77f64aa7..f0084bb6a1cb21 100644
--- a/inference-engine/src/cldnn_engine/ops/reshape.cpp
+++ b/inference-engine/src/cldnn_engine/ops/reshape.cpp
@@ -9,8 +9,8 @@
 #include "ngraph/op/squeeze.hpp"
 #include "ngraph/op/unsqueeze.hpp"
 
-#include "api/reshape.hpp"
-#include "api/reorder.hpp"
+#include "cldnn/primitives/reshape.hpp"
+#include "cldnn/primitives/reorder.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/result.cpp b/inference-engine/src/cldnn_engine/ops/result.cpp
index 4974f1ec56cdc7..fe0d0f05658018 100644
--- a/inference-engine/src/cldnn_engine/ops/result.cpp
+++ b/inference-engine/src/cldnn_engine/ops/result.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/result.hpp"
 
-#include "api/reorder.hpp"
+#include "cldnn/primitives/reorder.hpp"
 
 using namespace InferenceEngine;
 
diff --git a/inference-engine/src/cldnn_engine/ops/reverse_sequence.cpp b/inference-engine/src/cldnn_engine/ops/reverse_sequence.cpp
index 4537d3d34b7f0a..766bbc89a31d57 100644
--- a/inference-engine/src/cldnn_engine/ops/reverse_sequence.cpp
+++ b/inference-engine/src/cldnn_engine/ops/reverse_sequence.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/reverse_sequence.hpp"
 
-#include "api/reverse_sequence.hpp"
+#include "cldnn/primitives/reverse_sequence.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/rnn.cpp b/inference-engine/src/cldnn_engine/ops/rnn.cpp
index a1bab946db42ec..2d4705f1a910fa 100644
--- a/inference-engine/src/cldnn_engine/ops/rnn.cpp
+++ b/inference-engine/src/cldnn_engine/ops/rnn.cpp
@@ -8,12 +8,12 @@
 #include "ngraph/op/lstm_cell.hpp"
 #include "ngraph/op/lstm_sequence.hpp"
 
-#include "api/reshape.hpp"
-#include "api/reorder.hpp"
-#include "api/fully_connected.hpp"
-#include "api/lstm.hpp"
-#include "api/crop.hpp"
-#include "api/concatenation.hpp"
+#include "cldnn/primitives/reshape.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/fully_connected.hpp"
+#include "cldnn/primitives/lstm.hpp"
+#include "cldnn/primitives/crop.hpp"
+#include "cldnn/primitives/concatenation.hpp"
 
 namespace CLDNNPlugin {
 cldnn::activation_func GetActivationFunc(std::string name) {
diff --git a/inference-engine/src/cldnn_engine/ops/roi_pooling.cpp b/inference-engine/src/cldnn_engine/ops/roi_pooling.cpp
index 8dc57e3a5c63ff..f2087756405404 100644
--- a/inference-engine/src/cldnn_engine/ops/roi_pooling.cpp
+++ b/inference-engine/src/cldnn_engine/ops/roi_pooling.cpp
@@ -9,7 +9,7 @@
 #include "ngraph/op/psroi_pooling.hpp"
 #include "ngraph/op/deformable_psroi_pooling.hpp"
 
-#include "api/roi_pooling.hpp"
+#include "cldnn/primitives/roi_pooling.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/scatter_elements_update.cpp b/inference-engine/src/cldnn_engine/ops/scatter_elements_update.cpp
index 98c22e2a19966e..19f63dbf3a3986 100644
--- a/inference-engine/src/cldnn_engine/ops/scatter_elements_update.cpp
+++ b/inference-engine/src/cldnn_engine/ops/scatter_elements_update.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/scatter_elements_update.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/scatter_elements_update.hpp"
+#include "cldnn/primitives/scatter_elements_update.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/scatter_nd_update.cpp b/inference-engine/src/cldnn_engine/ops/scatter_nd_update.cpp
index 1b422736575054..cf5d059772a245 100644
--- a/inference-engine/src/cldnn_engine/ops/scatter_nd_update.cpp
+++ b/inference-engine/src/cldnn_engine/ops/scatter_nd_update.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/scatter_nd_update.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/scatter_nd_update.hpp"
+#include "cldnn/primitives/scatter_nd_update.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/scatter_update.cpp b/inference-engine/src/cldnn_engine/ops/scatter_update.cpp
index 85388032de8690..c6fb4e9c0aa003 100644
--- a/inference-engine/src/cldnn_engine/ops/scatter_update.cpp
+++ b/inference-engine/src/cldnn_engine/ops/scatter_update.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/scatter_update.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/scatter_update.hpp"
+#include "cldnn/primitives/scatter_update.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/select.cpp b/inference-engine/src/cldnn_engine/ops/select.cpp
index 0d3ae8e182c901..6b3866ac5dbaa9 100644
--- a/inference-engine/src/cldnn_engine/ops/select.cpp
+++ b/inference-engine/src/cldnn_engine/ops/select.cpp
@@ -7,9 +7,9 @@
 
 #include "ngraph/op/select.hpp"
 
-#include "api/select.hpp"
-#include "api/reorder.hpp"
-#include "api/reshape.hpp"
+#include "cldnn/primitives/select.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/reshape.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/shuffle_channels.cpp b/inference-engine/src/cldnn_engine/ops/shuffle_channels.cpp
index 0c0cafd03ef633..f3066ace26acc4 100644
--- a/inference-engine/src/cldnn_engine/ops/shuffle_channels.cpp
+++ b/inference-engine/src/cldnn_engine/ops/shuffle_channels.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/shuffle_channels.hpp"
 
-#include "api/shuffle_channels.hpp"
+#include "cldnn/primitives/shuffle_channels.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/softmax.cpp b/inference-engine/src/cldnn_engine/ops/softmax.cpp
index 566aca800ea398..cbaffc04accd49 100644
--- a/inference-engine/src/cldnn_engine/ops/softmax.cpp
+++ b/inference-engine/src/cldnn_engine/ops/softmax.cpp
@@ -8,8 +8,8 @@
 #include "ngraph/op/softmax.hpp"
 #include "ngraph/op/log_softmax.hpp"
 
-#include "api/softmax.hpp"
-#include "api/activation.hpp"
+#include "cldnn/primitives/softmax.hpp"
+#include "cldnn/primitives/activation.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/space_to_batch.cpp b/inference-engine/src/cldnn_engine/ops/space_to_batch.cpp
index 7846375dff700c..fa57d178038882 100644
--- a/inference-engine/src/cldnn_engine/ops/space_to_batch.cpp
+++ b/inference-engine/src/cldnn_engine/ops/space_to_batch.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/space_to_batch.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/space_to_batch.hpp"
+#include "cldnn/primitives/space_to_batch.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/space_to_depth.cpp b/inference-engine/src/cldnn_engine/ops/space_to_depth.cpp
index a8d73eea79d545..df4a25b469811a 100644
--- a/inference-engine/src/cldnn_engine/ops/space_to_depth.cpp
+++ b/inference-engine/src/cldnn_engine/ops/space_to_depth.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/space_to_depth.hpp"
 
-#include "api/space_to_depth.hpp"
+#include "cldnn/primitives/space_to_depth.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/split.cpp b/inference-engine/src/cldnn_engine/ops/split.cpp
index 6706ccebd8b88c..33b48c6d17fb2a 100644
--- a/inference-engine/src/cldnn_engine/ops/split.cpp
+++ b/inference-engine/src/cldnn_engine/ops/split.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/split.hpp"
 #include "ngraph/op/variadic_split.hpp"
 
-#include "api/crop.hpp"
+#include "cldnn/primitives/crop.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/strided_slice.cpp b/inference-engine/src/cldnn_engine/ops/strided_slice.cpp
index 0bf5adb74e22e4..8c796290a747da 100644
--- a/inference-engine/src/cldnn_engine/ops/strided_slice.cpp
+++ b/inference-engine/src/cldnn_engine/ops/strided_slice.cpp
@@ -8,9 +8,9 @@
 #include "ngraph/op/strided_slice.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/strided_slice.hpp"
-#include "api/reshape.hpp"
-#include "api/crop.hpp"
+#include "cldnn/primitives/strided_slice.hpp"
+#include "cldnn/primitives/reshape.hpp"
+#include "cldnn/primitives/crop.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/tensor_iterator.cpp b/inference-engine/src/cldnn_engine/ops/tensor_iterator.cpp
index e4e78208c0297e..d9843481979f23 100644
--- a/inference-engine/src/cldnn_engine/ops/tensor_iterator.cpp
+++ b/inference-engine/src/cldnn_engine/ops/tensor_iterator.cpp
@@ -13,11 +13,11 @@
 #include "ngraph/op/constant.hpp"
 #include "ngraph/op/util/sub_graph_base.hpp"
 
-#include "api/loop.hpp"
-#include "api/mutable_data.hpp"
-#include "api/data.hpp"
-#include "api/reorder.hpp"
-#include "api/topology.hpp"
+#include "cldnn/primitives/loop.hpp"
+#include "cldnn/primitives/mutable_data.hpp"
+#include "cldnn/primitives/data.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/graph/topology.hpp"
 
 #include <vector>
 #include <algorithm>
@@ -28,9 +28,8 @@ namespace CLDNNPlugin {
 
 template<class DATA_TYPE>
 static DATA_TYPE CreateScalarData(Program &p, const cldnn::primitive_id& id, int64_t num) {
-    auto mem = cldnn::memory::allocate(p.GetEngine(),
-        { cldnn::data_types::i64, cldnn::format::bfyx, { 1, 1, 1, 1 } });
-    auto ptr = mem.pointer<int64_t>();
+    auto mem = p.GetEngine().allocate_memory({ cldnn::data_types::i64, cldnn::format::bfyx, { 1, 1, 1, 1 } });
+    cldnn::mem_lock<int64_t> ptr{mem, p.GetEngine().get_program_stream()};
     *ptr.begin() = num;
     return {id, mem};
 }
@@ -42,7 +41,7 @@ static cldnn::mutable_data CreateAdditionalOutputData(Program &p, const std::sha
     const auto format = DefaultFormatForDims(op->get_output_shape(output_idx).size());
     const auto tensor = CldnnTensorFromIEDims(op->get_output_shape(output_idx));
     cldnn::layout output_layout = cldnn::layout(precision, format, tensor);
-    auto mem = cldnn::memory::allocate(p.GetEngine(), output_layout);
+    auto mem = p.GetEngine().allocate_memory(output_layout);
     auto md = cldnn::mutable_data(id, {input}, mem); // cldnn::data cannot set dependency
     return md;
 }
diff --git a/inference-engine/src/cldnn_engine/ops/tile.cpp b/inference-engine/src/cldnn_engine/ops/tile.cpp
index a4856ad80dc210..aa91fbd3d5a410 100644
--- a/inference-engine/src/cldnn_engine/ops/tile.cpp
+++ b/inference-engine/src/cldnn_engine/ops/tile.cpp
@@ -7,7 +7,7 @@
 
 #include "ngraph/op/tile.hpp"
 
-#include "api/tile.hpp"
+#include "cldnn/primitives/tile.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/topk.cpp b/inference-engine/src/cldnn_engine/ops/topk.cpp
index d527aee1d61b6e..3d8f7e6521a112 100644
--- a/inference-engine/src/cldnn_engine/ops/topk.cpp
+++ b/inference-engine/src/cldnn_engine/ops/topk.cpp
@@ -7,8 +7,8 @@
 
 #include "ngraph/op/topk.hpp"
 
-#include "api/arg_max_min.hpp"
-#include "api/mutable_data.hpp"
+#include "cldnn/primitives/arg_max_min.hpp"
+#include "cldnn/primitives/mutable_data.hpp"
 
 namespace CLDNNPlugin {
 
@@ -71,7 +71,7 @@ void CreateTopKOp(Program& p, const std::shared_ptr<ngraph::op::v1::TopK>& op) {
                                                     DefaultFormatForDims(op->get_output_shape(1).size()),
                                                     CldnnTensorFromIEDims(op->get_output_shape(1)));
 
-        auto shared_memory = cldnn::memory::allocate(p.GetEngine(), mutableLayout);
+        auto shared_memory = p.GetEngine().allocate_memory(mutableLayout);
 
         cldnn::primitive_id argmax_mutable_id_w = layer_type_name_ID(op) + "_md_write";
         auto argmax_mutable_prim = cldnn::mutable_data(argmax_mutable_id_w, shared_memory);
diff --git a/inference-engine/src/cldnn_engine/ops/transpose.cpp b/inference-engine/src/cldnn_engine/ops/transpose.cpp
index a1c8ce63caef6f..f5de62923a2c3b 100644
--- a/inference-engine/src/cldnn_engine/ops/transpose.cpp
+++ b/inference-engine/src/cldnn_engine/ops/transpose.cpp
@@ -8,7 +8,7 @@
 #include "ngraph/op/transpose.hpp"
 #include "ngraph/op/constant.hpp"
 
-#include "api/permute.hpp"
+#include "cldnn/primitives/permute.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/src/cldnn_engine/ops/unary.cpp b/inference-engine/src/cldnn_engine/ops/unary.cpp
index 334d68c5f2379f..9a277a8be2aa04 100644
--- a/inference-engine/src/cldnn_engine/ops/unary.cpp
+++ b/inference-engine/src/cldnn_engine/ops/unary.cpp
@@ -41,7 +41,7 @@
 #include "ngraph/op/hsigmoid.hpp"
 #include "ngraph/op/round.hpp"
 
-#include "api/activation.hpp"
+#include "cldnn/primitives/activation.hpp"
 
 namespace CLDNNPlugin {
 
diff --git a/inference-engine/tests/functional/plugin/gpu/CMakeLists.txt b/inference-engine/tests/functional/plugin/gpu/CMakeLists.txt
index d96dd234c1fdf2..f175b4c1a1f5e3 100644
--- a/inference-engine/tests/functional/plugin/gpu/CMakeLists.txt
+++ b/inference-engine/tests/functional/plugin/gpu/CMakeLists.txt
@@ -17,7 +17,7 @@ addIeTargetTest(
             AutoPlugin
         LINK_LIBRARIES
             funcSharedTests
-            ${CLDNN__OCL_ICD_LIBPATH}
+            ${OpenCL_LIBRARY}
         ADD_CPPLINT
         LABELS
             GPU
diff --git a/inference-engine/thirdparty/clDNN/CMakeLists.txt b/inference-engine/thirdparty/clDNN/CMakeLists.txt
index d14da47860cbeb..41833b58dc58f2 100644
--- a/inference-engine/thirdparty/clDNN/CMakeLists.txt
+++ b/inference-engine/thirdparty/clDNN/CMakeLists.txt
@@ -83,6 +83,9 @@ set(CLDNN__API_EXTENSION_DIR       "${CMAKE_CURRENT_SOURCE_DIR}/api_extension")
 # Path which points to directory with interface for framework.
 set(CLDNN__KERNEL_SELECTOR_DIR       "${CMAKE_CURRENT_SOURCE_DIR}/kernel_selector")
 
+# Path which points to directory with runtime
+set(CLDNN__RUNTIME_DIR       "${CMAKE_CURRENT_SOURCE_DIR}/runtime")
+
 # Path which points to directory with binaries for Khronos OpenCL ICD Loader (Installable Client Driver).
 set(CLDNN__OCL_ICD_DIR  "${CLDNN__COMMON_DIR}/khronos_ocl_icd")
 
@@ -270,12 +273,6 @@ mark_as_advanced(CLDNN__INCLUDE_TESTS)
 
 # ======================================================================================================
 
-# Include and build: Core Internal Tests (unit tests and small acceptance tests) for core internal clDNN framework mechanisms.
-set(CLDNN__INCLUDE_CORE_INTERNAL_TESTS ON CACHE BOOL "Include and build: clDNN framework's core internal tests.")
-mark_as_advanced(CLDNN__INCLUDE_CORE_INTERNAL_TESTS)
-
-# ======================================================================================================
-
 
 # Include and build: clDNN tutorial.
 set(CLDNN__INCLUDE_TUTORIAL ON CACHE BOOL "Include and build: clDNN Tutorial.")
@@ -289,12 +286,6 @@ mark_as_advanced(CLDNN__RUN_TESTS)
 
 # ======================================================================================================
 
-# Run (requires CLDNN__INCLUDE_CORE_INTERNAL_TESTS to be true): Tests (unit tests and small acceptance core internal tests) for clDNN framework.
-set(CLDNN__RUN_CORE_INTERNAL_TESTS OFF CACHE BOOL "Run: clDNN framework's core internal tests.")
-mark_as_advanced(CLDNN__RUN_CORE_INTERNAL_TESTS)
-
-# ======================================================================================================
-
 # Compile / Link: Use static C++ Runtime library.
 set(CLDNN__COMPILE_LINK_USE_STATIC_RUNTIME OFF CACHE BOOL "Compile / Link: Use static version of C++ Runtime library instead of shared one.")
 mark_as_advanced(CLDNN__COMPILE_LINK_USE_STATIC_RUNTIME)
@@ -330,14 +321,6 @@ endif()
 
 # ======================================================================================================
 
-# Checking whether tests can be run.
-if((NOT CLDNN__INCLUDE_CORE_INTERNAL_TESTS) AND CLDNN__RUN_CORE_INTERNAL_TESTS)
-  message(WARNING "[clDNN] CLDNN__INCLUDE_CORE_INTERNAL_TESTS: Selected running of core internal tests, but test are not built. Option will be disabled.")
-  set(CLDNN__RUN_CORE_INTERNAL_TESTS OFF)
-endif()
-
-# ======================================================================================================
-
 # Check for python 3 interpreter (required tool).
 find_package(PythonInterp 3 QUIET)
 if(NOT PYTHONINTERP_FOUND)
@@ -375,19 +358,26 @@ set(CLDNN__OCL_ICD_ROOT "${CLDNN__OCL_ICD_DIR}" CACHE INTERNAL "Path to Khronos
 set(CLDNN__OCL_ICD_INCDIRS "${CLDNN__KHR_CLHPP_DIR}" CACHE INTERNAL "Paths to interface headers for OpenCL.")
 set(CLDNN__OCL_ICD_STLDIRS "${CLDNN__OCL_ICD_ROOT}/${__CLDNN_TargetOs}/${CLDNN__TARGET_CFG_VAR}/lib/${__CLDNN_TargetCpuDir}" CACHE INTERNAL "Paths to static libraries for OpenCL ICD Loader.")
 set(CLDNN__OCL_ICD_SHLDIRS "${CLDNN__OCL_ICD_ROOT}/${__CLDNN_TargetOs}/${CLDNN__TARGET_CFG_VAR}/bin/${__CLDNN_TargetCpuDir}" CACHE INTERNAL "Paths to shared libraries for OpenCL ICD Loader.")
+include_directories(SYSTEM "${CLDNN__KHR_CLHPP_DIR}")
 
 # Select link directory based on targeted OS.
 # - on Windows: static libraries directory.
 # - on others:  shared libraries directory.
 if(__CLDNN_TargetOs MATCHES "^windows$")
-  set(CLDNN__OCL_ICD_LIBDIRS ${CLDNN__OCL_ICD_STLDIRS} CACHE INTERNAL "Paths to libraries to link for OpenCL ICD Loader.")
-  set(CLDNN__OCL_ICD_LIBPATH ${CLDNN__OCL_ICD_LIBDIRS}/${CMAKE_STATIC_LIBRARY_PREFIX}OpenCL${CMAKE_STATIC_LIBRARY_SUFFIX} CACHE INTERNAL "")
+set(CLDNN__OCL_ICD_LIBDIRS ${CLDNN__OCL_ICD_STLDIRS} CACHE INTERNAL "Paths to libraries to link for OpenCL ICD Loader.")
+set(CLDNN__OCL_ICD_LIBPATH ${CLDNN__OCL_ICD_LIBDIRS}/${CMAKE_STATIC_LIBRARY_PREFIX}OpenCL${CMAKE_STATIC_LIBRARY_SUFFIX} CACHE INTERNAL "")
 else()
-  set(CLDNN__OCL_ICD_LIBDIRS ${CLDNN__OCL_ICD_SHLDIRS} CACHE INTERNAL "Paths to libraries to link for OpenCL ICD Loader.")
-  set(CLDNN__OCL_ICD_LIBPATH ${CLDNN__OCL_ICD_LIBDIRS}/${CMAKE_SHARED_LIBRARY_PREFIX}OpenCL${CMAKE_SHARED_LIBRARY_SUFFIX} CACHE INTERNAL "")
+set(CLDNN__OCL_ICD_LIBDIRS ${CLDNN__OCL_ICD_SHLDIRS} CACHE INTERNAL "Paths to libraries to link for OpenCL ICD Loader.")
+set(CLDNN__OCL_ICD_LIBPATH ${CLDNN__OCL_ICD_LIBDIRS}/${CMAKE_SHARED_LIBRARY_PREFIX}OpenCL${CMAKE_SHARED_LIBRARY_SUFFIX} CACHE INTERNAL "")
 endif()
 
-
+set(OpenCL_LIBRARY "${CLDNN__OCL_ICD_LIBPATH}" CACHE INTERNAL "Paths to shared libraries for OpenCL ICD Loader.")
+set(OpenCL_INCLUDE_DIR "${CLDNN__OCL_ICD_INCDIRS}" CACHE INTERNAL "Paths to interface headers for OpenCL.")
+add_library(clDNN_OpenCL UNKNOWN IMPORTED)
+set_target_properties(clDNN_OpenCL
+    PROPERTIES
+        IMPORTED_LOCATION ${CLDNN__OCL_ICD_LIBPATH}
+  )
 # ======================================================================================================
 set(CLDNN_UTILS__RAPIDJSON_INCDIRS "utils/rapidjson" CACHE INTERNAL "Paths to interface headers for rapidjson.")
 # ====================================== Version Calculation ===========================================
@@ -414,11 +404,9 @@ message(STATUS "[clDNN]")
 message(STATUS "[clDNN]  - Include/Build cldnn core:          ${CLDNN__INCLUDE_CORE}")
 message(STATUS "[clDNN]  - Include/Build kernel selector:     ${CLDNN__INCLUDE_KERNEL_SELECTOR}")
 message(STATUS "[clDNN]  - Include/Build tests:               ${CLDNN__INCLUDE_TESTS}")
-message(STATUS "[clDNN]  - Include/Build core internal tests: ${CLDNN__INCLUDE_CORE_INTERNAL_TESTS}")
 message(STATUS "[clDNN]  - Include/Build tutorial:            ${CLDNN__INCLUDE_TUTORIAL}")
 message(STATUS "[clDNN]")
 message(STATUS "[clDNN]  - Run tests:                     ${CLDNN__RUN_TESTS}")
-message(STATUS "[clDNN]  - Run core internal tests:       ${CLDNN__RUN_CORE_INTERNAL_TESTS}")
 message(STATUS "[clDNN]")
 message(STATUS "[clDNN]  - Use static C++ Runtime:        ${CLDNN__COMPILE_LINK_USE_STATIC_RUNTIME}")
 message(STATUS "[clDNN]  - Allow unsafe size opts:        ${CLDNN__COMPILE_LINK_ALLOW_UNSAFE_SIZE_OPT}")
@@ -781,30 +769,18 @@ set_property(DIRECTORY APPEND PROPERTY COMPILE_DEFINITIONS
 include_directories(
     ${CLDNN_UTILS__RAPIDJSON_INCDIRS}
     "${CLDNN__CODEGEN_INCDIR}"
+    "${CLDNN__API_DIR}"
   )
 
-include_directories(SYSTEM
-   "${CLDNN__KHR_CLHPP_DIR}"
-  )
-
-add_library(clDNN_OpenCL UNKNOWN IMPORTED)
-set_target_properties(clDNN_OpenCL
-    PROPERTIES
-        IMPORTED_LOCATION ${CLDNN__OCL_ICD_LIBPATH}
-  )
-
-
 # =================================== Link targets and dependencies ====================================
 if(CLDNN__INCLUDE_CORE)
   add_subdirectory(src)
+  add_subdirectory(runtime)
   add_subdirectory(api_test_builds)
 endif()
 if(CLDNN__INCLUDE_TESTS)
   add_subdirectory(tests)
 endif()
-if(CLDNN__INCLUDE_CORE_INTERNAL_TESTS)
-  add_subdirectory(tests_core_internal)
-endif()
 if(CLDNN__INCLUDE_KERNEL_SELECTOR)
   add_subdirectory(kernel_selector)
 endif()
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn.hpp b/inference-engine/thirdparty/clDNN/api/cldnn.hpp
deleted file mode 100644
index 4b531ab050bb45..00000000000000
--- a/inference-engine/thirdparty/clDNN/api/cldnn.hpp
+++ /dev/null
@@ -1,244 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-/*! @mainpage clDNN Documentation
-* @section intro Introduction
-* Compute Library for Deep Neural Networks (clDNN) is a middle-ware software
-*  for accelerating DNN inference on Intel&reg; HD and Iris&trade; Pro Graphics.
-*  This project includes CNN primitives implementations on Intel GPUs with C and C++ interfaces.
-*
-* clDNN Library implements set of primitives:
-*  - Convolution
-*  - Fully connected (inner product)
-*  - Pooling
-*   * average
-*   * maximum
-*  - Normalization
-*   * across channel
-*   * within channel
-*   * batch
-*  - Activation
-*   * logistic
-*   * tanh
-*   * rectified linear unit (ReLU)
-*   * softplus (softReLU)
-*   * abs
-*   * square
-*   * sqrt
-*   * linear
-*  - Softmax
-*  - Crop
-*  - Deconvolution
-*  - Depth concatenation
-*  - Eltwise
-*  - ROI pooling
-*  - Simpler NMS
-*  - Prior box
-*  - Detection output
-*
-*  With this primitive set, user can build and execute most common image recognition, semantic segmentation and object detection networks topologies like:
-*   - Alexnet
-*   - Googlenet(v1-v3)
-*   - ResNet
-*   - VGG
-*   - faster-rCNN
-* and other.
-*
-*
-* @section model Programming Model
-*  Intel&reg; clDNN is graph oriented library. To execute CNN you have to build, compile graph/topology and run to get results.
-*
-*  <B> Terminology: </B>
-*  - Primitive - dnn base functionality i.e. convolution, pooling, softmax.
-*  - Data - special primitive type representing primitive parameters (weights and biases), inputs and outputs
-*  - Engine - type of accelerator that is executing network. Currently ocl engine is the only available.
-*  - Topology - container of primitives, data, and relations between them. Topology represents graph.
-*  - Program - optional step between Topology and Network. It is compiled Topology without memory allocation.
-*  - Network - compiled Topology with memory allocation. Ready to be executed. During compilation, buidling parameters trigger special optimizations like fusing, data reordering.
-*
-*  <B> Execution Steps: </B>
-*
-* \image html workflow.jpg
-* -# Create Engine
-* -# Declare or define primitives parameters (weights and biases) if needed.
-* -# Create primitives. It is required to provide name for each primitive. This is a name of primitive which output will be input to current one. Name can be used before primitive definition.
-* -# Create topology
-* -# Add primitives to topology
-* -# Build Network from topology
-* -# Set Inputs data
-* -# Execute Network
-*
-*
-* @section graph_compilation Graph compilation
-*
-* If user choose build option optimize_data when program is being created - explicit or implicit over network creation, clDNN perform some graph optimizations as follows:
-* * <B> Stage 0: Graph initiation:</B>
-*  * build nodes from primitives
-*  * node replacement:
-*   * replace each split node with series of crop nodes. Name of crop primitive will be concatenation of split + port names.
-*   * replace upsampling node with deconvolution node if upsampling mode is bilinear.
-*  * set outputs - mark nodes that are defined by user as output (blocks fusing etc) or have no users (leafs).
-*  * calculate processing order - using dfs on graph to establish processing order
-* * <B> Stage 1: Priorboxes:</B>
-*  * priorbox is primitive that is executed during network compilation. Node is removed from a network execution.
-* * <B> Stage 2: Graph analysis:</B>
-*  * mark constatns
-*  * mark data flow
-* * <B> Stage 3: Trimming:</B>
-*  * apply backward bfs on each output to find unnecessary nodes/branches, then remove those.
-* * <B> Stage 4: Inputs and biases:</B>
-*  * reorder input - format of convolution's input/output is being selected.
-*  * reorder biases for conv,fc and deconv nodes
-* * <B> Stage 5: Redundant reorders:</B>
-*  * previous stages can provide additional reorders due to format changes per primitive. This stage removes redundant and fuses series of reorders into one.
-* * <B> Stage 6: Constant propagation:</B>
-*  * prepare padding - goes thrugh all primitves and checks if its user requires padding, if so, set output padding.
-*  * prepare depthwise separable opt - if split param is greater than 16 and number of IFM <= 8*split in conv or deconv, this stage changes execution from multi kernels into one.
-*  * constant propagation - replace constant nodes, that are not outputs with data type nodes. Constant primitive is the primitive that doesn't depend on any non-constant primitive and doesn't have to be executed: priorbox, data.
-* * <B> Stage 7: Fusing:</B>
-*  * buffer fusing
-*   * concat - if concatenation is the only user of its dependencies then remove concat node and setting proper output paddings in every dependencies.
-*   * crop - if crop has only one dependecy, and its users doesn't require padding, remove crop and set proper output padding in its dependecy.
-*   * reorder - if primitive before reorder supports different input vs output type reorder can be fused with previous node.
-*  * primitive fusing - right now this stage fuses activation node with previous node only, only if previous node supports activation fusing.
-* * <B> Stage 8: Compile graph:</B>
-*  * at this stage using kernel selector, graph chooses the best kernel implementation for each node.
-* * <B> Stage 9: reorder weights:</B>
-*  * at this stage weights are converted into format suitable for selected kernel implementation.
-* * <B> Stage 10 & 11: Redundant reorders and constant propagation:</B>
-*  * check again if whole graph compilation didn't provide any redundant reorders and constants.
-* * <B> Stage 12: Compile program:</B>
-*  * at this stage engine compiles cl_kernels.
-*
-* @section example C++ API Example MNIST network
-* @include example_cldnn.cpp
-*/
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-
-#include <stdint.h>
-#include <stddef.h>
-#include <memory>
-#include <string>
-#include <type_traits>
-#include <thread>
-
-namespace cldnn {
-
-/// @addtogroup cpp_api C++ API
-/// @{
-
-float half_to_float(uint16_t value);
-uint16_t float_to_half(float value);
-
-// There is no portable half precision floating point support.
-// Using wrapped integral type with the same size and alignment restrictions.
-class half_impl {
-public:
-    half_impl() = default;
-
-    template <typename T, typename = typename std::enable_if<!std::is_floating_point<T>::value>::type>
-    explicit half_impl(T data, int /*direct_creation_tag*/) : _data(data) {}
-
-    operator uint16_t() const { return _data; }
-    operator float() const {
-        return half_to_float(_data);
-    }
-
-    explicit half_impl(float value)
-        : _data(float_to_half(value))
-    {}
-
-    template <typename T, typename = typename std::enable_if<std::is_convertible<T, float>::value>::type>
-    explicit half_impl(T value)
-        : half_impl(static_cast<float>(value))
-    {}
-
-private:
-    uint16_t _data;
-};
-
-// Use complete implementation if necessary.
-#if defined HALF_HALF_HPP
-using half_t = half;
-#else
-using half_t = half_impl;
-#endif
-
-/// @cond CPP_HELPERS
-
-/// @defgroup cpp_helpers Helpers
-/// @{
-
-#define CLDNN_API_CLASS(the_class) static_assert(std::is_standard_layout<the_class>::value, #the_class " has to be 'standard layout' class");
-
-template <typename T>
-typename std::enable_if<std::is_integral<T>::value, T>::type align_to(T size, size_t align) {
-    return static_cast<T>((size % align == 0) ? size : size - size % align + align);
-}
-
-template <typename T>
-typename std::enable_if<std::is_integral<T>::value, T>::type pad_to(T size, size_t align) {
-    return static_cast<T>((size % align == 0) ? 0 : align - size % align);
-}
-
-template <typename T>
-typename std::enable_if<std::is_integral<T>::value, bool>::type is_aligned_to(T size, size_t align) {
-    return !(size % align);
-}
-
-/// Computes ceil(@p val / @p divider) on unsigned integral numbers.
-///
-/// Computes division of unsigned integral numbers and rounds result up to full number (ceiling).
-/// The function works for unsigned integrals only. Signed integrals are converted to corresponding
-/// unsigned ones.
-///
-/// @tparam T1   Type of @p val. Type must be integral (SFINAE).
-/// @tparam T2   Type of @p divider. Type must be integral (SFINAE).
-///
-/// @param val       Divided value. If value is signed, it will be converted to corresponding unsigned type.
-/// @param divider   Divider value. If value is signed, it will be converted to corresponding unsigned type.
-///
-/// @return   Result of ceil(@p val / @p divider). The type of result is determined as if in normal integral
-///           division, except each operand is converted to unsigned type if necessary.
-template <typename T1, typename T2>
-constexpr auto ceil_div(T1 val, T2 divider)
--> typename std::enable_if<std::is_integral<T1>::value && std::is_integral<T2>::value,
-    decltype(std::declval<typename std::make_unsigned<T1>::type>() / std::declval<typename std::make_unsigned<T2>::type>())>::type {
-    typedef typename std::make_unsigned<T1>::type UT1;
-    typedef typename std::make_unsigned<T2>::type UT2;
-    typedef decltype(std::declval<UT1>() / std::declval<UT2>()) RetT;
-
-    return static_cast<RetT>((static_cast<UT1>(val) + static_cast<UT2>(divider) - 1U) / static_cast<UT2>(divider));
-}
-
-/// Rounds @p val to nearest multiply of @p rounding that is greater or equal to @p val.
-///
-/// The function works for unsigned integrals only. Signed integrals are converted to corresponding
-/// unsigned ones.
-///
-/// @tparam T1       Type of @p val. Type must be integral (SFINAE).
-/// @tparam T2       Type of @p rounding. Type must be integral (SFINAE).
-///
-/// @param val        Value to round up. If value is signed, it will be converted to corresponding unsigned type.
-/// @param rounding   Rounding value. If value is signed, it will be converted to corresponding unsigned type.
-///
-/// @return   @p val rounded up to nearest multiply of @p rounding. The type of result is determined as if in normal integral
-///           division, except each operand is converted to unsigned type if necessary.
-template <typename T1, typename T2>
-constexpr auto round_up_to(T1 val, T2 rounding)
--> typename std::enable_if<std::is_integral<T1>::value && std::is_integral<T2>::value,
-    decltype(std::declval<typename std::make_unsigned<T1>::type>() / std::declval<typename std::make_unsigned<T2>::type>())>::type {
-    typedef typename std::make_unsigned<T1>::type UT1;
-    typedef typename std::make_unsigned<T2>::type UT2;
-    typedef decltype(std::declval<UT1>() / std::declval<UT2>()) RetT;
-
-    return static_cast<RetT>(ceil_div(val, rounding) * static_cast<UT2>(rounding));
-}
-
-/// @}
-/// @endcond
-/// @}
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/network.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/graph/network.hpp
similarity index 79%
rename from inference-engine/thirdparty/clDNN/api/network.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/graph/network.hpp
index ee089e7e42f4d1..a6829259d79aba 100644
--- a/inference-engine/thirdparty/clDNN/api/network.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/graph/network.hpp
@@ -3,12 +3,14 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
+
 #pragma once
-#include "cldnn.hpp"
-#include "compounds.h"
-#include "memory.hpp"
+
+#include "cldnn/runtime/compounds.hpp"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/event.hpp"
+#include "cldnn/runtime/stream.hpp"
 #include "program.hpp"
-#include "event.hpp"
 
 #include <cstdint>
 #include <algorithm>
@@ -28,18 +30,25 @@ namespace cldnn {
 /// @brief Represents network output returned by @ref network::get_output().
 struct network_output {
     /// @brief Returns @ref event associated with the output.
-    event get_event() const { return _event; }
+    event::ptr get_event() const { return _event; }
 
     /// @brief Returns @ref memory object of the output. Blocked until associated @ref event is not complete.
-    memory get_memory() const {
-        _event.wait();
+    memory::ptr get_memory() const {
+        // TODO: in_order queue doesn't create proper output event in some cases which leads to syncronization issues with user app
+        // So call finish for associated stream to enusre that the output data is ready.
+        if (_stream.get_queue_type() == queue_types::in_order) {
+            _stream.finish();
+        } else {
+            _event->wait();
+        }
         return _result;
     }
 
 private:
-    event _event;
-    memory _result;
-    network_output(event evt, memory mem) : _event(evt), _result(mem) {}
+    event::ptr _event;
+    memory::ptr _result;
+    stream& _stream;
+    network_output(event::ptr evt, memory::ptr mem, stream& stream) : _event(evt), _result(mem), _stream(stream) {}
     friend struct network;
 };
 
@@ -58,52 +67,49 @@ struct network {
     /// @param topology
     /// @param options
     /// @param options
-    network(const engine& engine,
+    network(engine& engine,
             const topology& topology,
             const build_options& options = build_options(),
             uint16_t stream_id = 0)
         : network(program(engine, topology, options), stream_id) {}
 
     /// @brief Constructs network object from C API @ref cldnn_network.
-    explicit network(network_impl* impl) : _impl(impl) {
+    explicit network(std::shared_ptr<network_impl> impl) : _impl(impl) {
         if (_impl == nullptr)
             throw std::invalid_argument("implementation pointer should not be null");
     }
 
     /// @brief Copy construction.
-    network(const network& other) : _impl(other._impl) { retain(); }
+    network(const network& other) : _impl(other._impl) { }
 
     /// @brief Copy assignment.
     network& operator=(const network& other) {
         if (_impl == other._impl)
             return *this;
-        release();
         _impl = other._impl;
-        retain();
         return *this;
     }
 
-    /// @brief Releases wrapped C API @ref cldnn_network.
-    ~network() { release(); }
-
     friend bool operator==(const network& lhs, const network& rhs) { return lhs._impl == rhs._impl; }
     friend bool operator!=(const network& lhs, const network& rhs) { return !(lhs == rhs); }
 
     /// @brief Returns @ref engine by which network was built.
-    engine get_engine() const;
+    engine& get_engine() const;
 
     /// @brief Returns network internal @ref program.
     program get_program() const;
 
     /// @brief Provides @ref memory for @ref input_layout primitives defined by user in source @ref topology.
-    void set_input_data(const primitive_id& id, const memory& mem) const;
+    void set_input_data(const primitive_id& id, memory::ptr mem) const;
 
     /// @brief Provides user-supplied @ref memory for output primitives defined by user in source @ref topology.
-    void set_output_memory(const primitive_id& id, const memory& mem) const;
+    void set_output_memory(const primitive_id& id, memory::ptr mem) const;
 
     /// @brief Return stream id.
     uint16_t get_stream_id();
 
+    stream& get_stream() const;
+
     /// @brief Return internal network id.
     uint32_t get_id();
 
@@ -131,18 +137,18 @@ struct network {
     std::vector<primitive_id> get_output_ids() const;
 
     /// @brief Returns @ref memory object for particular @p output. Can be called before network execution
-    memory get_output_memory(const primitive_id& output_id) const;
+    memory::ptr get_output_memory(const primitive_id& output_id) const;
 
     /// @brief Returns @ref event object for particular @p primitive. Can't be called before network execution
-    event get_primitive_event(const primitive_id& output_id) const;
+    event::ptr get_primitive_event(const primitive_id& output_id) const;
 
     /// @brief Returns @ref network_output object for particular @p output. Can't be called before network execution
     network_output get_output(const primitive_id& output_id) const {
-        return network_output(get_primitive_event(output_id), get_output_memory(output_id));
+        return network_output(get_primitive_event(output_id), get_output_memory(output_id), get_stream());
     }
 
     /// @brief Returns the list of @ref event for the primitives that were executed in network.
-    std::map<primitive_id, event> get_executed_primitives() const {
+    std::map<primitive_id, event::ptr> get_executed_primitives() const {
         auto primitive_ids = get_executed_primitive_ids();
         auto all_primitive_ids = get_all_primitive_ids();
         auto all_primitive_org_ids = get_all_primitive_org_ids();
@@ -152,7 +158,7 @@ struct network {
             if (all_primitive_ids[i] == "_optimized_")
                 optimized_primitives.push_back(all_primitive_org_ids[i]);
         }
-        std::map<primitive_id, event> result;
+        std::map<primitive_id, event::ptr> result;
         for (auto& id : primitive_ids) {
             if (std::find(optimized_primitives.begin(), optimized_primitives.end(), id) == optimized_primitives.end())
                 result.emplace(id, get_primitive_event(id));
@@ -177,16 +183,13 @@ struct network {
     /// @param dependencies List of @ref event objects to be waited before network execution.
     /// @note User should call set_input_data() for every @ref input_layout defined in source @ref topology
     /// before network execution.
-    std::map<primitive_id, network_output> execute(const std::vector<event>& dependencies = {}) const;
+    std::map<primitive_id, network_output> execute(const std::vector<event::ptr>& dependencies = {}) const;
 
     /// @brief Returns wrapped C API @ref cldnn_network handler.
-    network_impl* get() const { return _impl; }
+    network_impl* get() const { return _impl.get(); }
 
 private:
-    network_impl* _impl;
-
-    void retain();
-    void release();
+    std::shared_ptr<network_impl> _impl;
 };
 CLDNN_API_CLASS(network)
 /// @}
diff --git a/inference-engine/thirdparty/clDNN/api/program.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/graph/program.hpp
similarity index 97%
rename from inference-engine/thirdparty/clDNN/api/program.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/graph/program.hpp
index b50754e1fd07b9..95dfca9fc14bb9 100644
--- a/inference-engine/thirdparty/clDNN/api/program.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/graph/program.hpp
@@ -3,11 +3,12 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
+
 #pragma once
-#include "cldnn.hpp"
+
+#include "cldnn/runtime/engine.hpp"
+
 #include "topology.hpp"
-#include "engine.hpp"
-#include <iostream>
 
 #include <memory>
 #include <vector>
@@ -477,21 +478,19 @@ struct program {
     /// @param[in] engine The engine which will be used to build the program.
     /// @param[in] topology The user-defined topology on which the network will be based.
     /// @param[in] options Program build options. See @ref build_option and @ref build_options for details.
-    program(engine const& engine, topology const& topology, build_options const& options = build_options());
+    program(engine& engine, const topology& topology, const build_options& options = build_options());
 
     /// @brief Copy constructor.
-    program(program const& other) : _impl(other._impl) { retain(); }
+    program(const program& other) : _impl(other._impl) { }
 
     /// @brief Dereferences the counter of the underlying C API @ref cldnn_program handler.
-    ~program() { release(); }
+    ~program() { }
 
     /// @brief Assigns new value by releasing previously referenced C API @ref cldnn_program handler and retaining the one referenced by @p other.
     program& operator=(const program& other) {
         if (_impl == other._impl)
             return *this;
-        release();
         _impl = other._impl;
-        retain();
         return *this;
     }
 
@@ -500,19 +499,15 @@ struct program {
     /// @brief Checks whether @p lhs and @p rhs reference different C API @ref cldnn_program handlers
     friend bool operator!=(const program& lhs, const program& rhs) { return !(lhs == rhs); }
 
-    /// @brief Returns wrapped C API @ref cldnn_program handler.
-    program_impl* get() const { return _impl; }
+    std::shared_ptr<program_impl> get() const { return _impl; }
 
 private:
-    program_impl* _impl;
+    std::shared_ptr<program_impl> _impl;
 
-    explicit program(program_impl* impl) : _impl(impl) {
+    explicit program(std::shared_ptr<program_impl> impl) : _impl(impl) {
         if (_impl == nullptr)
             throw std::invalid_argument("implementation pointer should not be null");
     }
-
-    void retain();
-    void release();
 };
 /// @}
 /// @}
diff --git a/inference-engine/thirdparty/clDNN/api/topology.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/graph/topology.hpp
similarity index 84%
rename from inference-engine/thirdparty/clDNN/api/topology.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/graph/topology.hpp
index 78e4fc875cf757..4155c7ac529e8b 100644
--- a/inference-engine/thirdparty/clDNN/api/topology.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/graph/topology.hpp
@@ -3,11 +3,13 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
+
 #pragma once
+
+#include "cldnn/runtime/compounds.hpp"
+#include "cldnn/primitives/primitive.hpp"
+
 #include <cstdint>
-#include "cldnn.hpp"
-#include "compounds.h"
-#include "primitive.hpp"
 #include <vector>
 #include <memory>
 
@@ -33,26 +35,24 @@ struct topology {
     }
 
     /// @brief Copy construction.
-    topology(const topology& other) : _impl(other._impl) { retain(); }
+    topology(const topology& other) : _impl(other._impl) { }
 
     /// @brief Copy assignment.
     topology& operator=(const topology& other) {
         if (_impl == other._impl)
             return *this;
-        release();
         _impl = other._impl;
-        retain();
         return *this;
     }
 
     /// Construct C++ topology based on C API @p cldnn_topology
-    explicit topology(topology_impl* other) : _impl(other) {
+    explicit topology(std::shared_ptr<topology_impl> other) : _impl(other) {
         if (_impl == nullptr)
             throw std::invalid_argument("implementation pointer should not be null");
     }
 
     /// @brief Releases wrapped C API @ref cldnn_topology.
-    ~topology() { release(); }
+    ~topology() { }
 
     friend bool operator==(const topology& lhs, const topology& rhs) { return lhs._impl == rhs._impl; }
     friend bool operator!=(const topology& lhs, const topology& rhs) { return !(lhs == rhs); }
@@ -73,7 +73,7 @@ struct topology {
     }
 
     /// @brief Returns wrapped implementation pointer.
-    topology_impl* get() const { return _impl; }
+    std::shared_ptr<topology_impl> get() const { return _impl; }
 
     const std::vector<primitive_id> get_primitive_ids() const;
 
@@ -82,12 +82,9 @@ struct topology {
     const std::shared_ptr<primitive>& at(const primitive_id& id) const;
 
 private:
-    friend struct engine;
+    friend class engine;
     friend struct network;
-    topology_impl* _impl;
-
-    void retain();
-    void release();
+    std::shared_ptr<topology_impl> _impl;
 };
 
 CLDNN_API_CLASS(topology)
diff --git a/inference-engine/thirdparty/clDNN/api/activation.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/activation.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/activation.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/activation.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/arg_max_min.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/arg_max_min.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/arg_max_min.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/arg_max_min.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/average_unpooling.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/average_unpooling.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/average_unpooling.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/average_unpooling.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/batch_to_space.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/batch_to_space.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/batch_to_space.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/batch_to_space.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/binary_convolution.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/binary_convolution.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/binary_convolution.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/binary_convolution.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/border.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/border.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/border.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/border.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/broadcast.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/broadcast.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/broadcast.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/broadcast.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/concatenation.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/concatenation.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/concatenation.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/concatenation.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/condition.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/condition.hpp
similarity index 98%
rename from inference-engine/thirdparty/clDNN/api/condition.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/condition.hpp
index 1becbc1aff54f2..aa63586c4a3157 100644
--- a/inference-engine/thirdparty/clDNN/api/condition.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/condition.hpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 #include "primitive.hpp"
-#include "topology.hpp"
+#include "cldnn/graph/topology.hpp"
 #include <vector>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/api/convolution.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/convolution.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/convolution.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/convolution.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/crop.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/crop.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/crop.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/crop.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/ctc_greedy_decoder.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/ctc_greedy_decoder.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/ctc_greedy_decoder.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/ctc_greedy_decoder.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/cum_sum.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/cum_sum.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/cum_sum.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/cum_sum.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/custom_gpu_primitive.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/custom_gpu_primitive.hpp
similarity index 98%
rename from inference-engine/thirdparty/clDNN/api/custom_gpu_primitive.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/custom_gpu_primitive.hpp
index 22b43fedd3c01f..25da079a5de315 100644
--- a/inference-engine/thirdparty/clDNN/api/custom_gpu_primitive.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/custom_gpu_primitive.hpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 #include "primitive.hpp"
-#include "memory.hpp"
+#include "cldnn/runtime/memory.hpp"
 #include <vector>
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/api/data.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/data.hpp
similarity index 92%
rename from inference-engine/thirdparty/clDNN/api/data.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/data.hpp
index 68e43ee1abd773..e7fe1f889fc619 100644
--- a/inference-engine/thirdparty/clDNN/api/data.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/data.hpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 #include "primitive.hpp"
-#include "memory.hpp"
+#include "cldnn/runtime/memory.hpp"
 
 namespace cldnn {
 /// @addtogroup cpp_api C++ API
@@ -26,12 +26,12 @@ struct data : public primitive_base<data> {
     /// @param id This primitive id.
     /// @param mem @ref memory object which contains data.
     /// @note If memory is attached by memory::attach(), the attached buffer should be valid till network build.
-    data(const primitive_id& id, const memory& mem)
+    data(const primitive_id& id, memory::ptr mem)
         : primitive_base(id, {}, padding()), mem(mem) {}
 
     /// @brief @ref memory object which contains data.
     /// @note If memory is attached by memory::attach(), the attached buffer should be valid till network build.
-    memory mem;
+    memory::ptr mem;
 };
 /// @}
 /// @}
diff --git a/inference-engine/thirdparty/clDNN/api/deconvolution.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/deconvolution.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/deconvolution.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/deconvolution.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/depth_to_space.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/depth_to_space.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/depth_to_space.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/depth_to_space.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/detection_output.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/detection_output.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/detection_output.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/detection_output.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/eltwise.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/eltwise.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/eltwise.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/eltwise.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/embedding_bag.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/embedding_bag.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/embedding_bag.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/embedding_bag.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/extract_image_patches.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/extract_image_patches.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/extract_image_patches.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/extract_image_patches.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/fully_connected.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/fully_connected.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/fully_connected.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/fully_connected.hpp
diff --git a/inference-engine/thirdparty/clDNN/api_extension/fused_conv_eltwise.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/fused_conv_eltwise.hpp
similarity index 98%
rename from inference-engine/thirdparty/clDNN/api_extension/fused_conv_eltwise.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/fused_conv_eltwise.hpp
index ff1c274fb1c945..6aed6cf6a7568a 100644
--- a/inference-engine/thirdparty/clDNN/api_extension/fused_conv_eltwise.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/fused_conv_eltwise.hpp
@@ -4,8 +4,8 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/primitive.hpp"
-#include "api/eltwise.hpp"
+#include "cldnn/primitives/primitive.hpp"
+#include "cldnn/primitives/eltwise.hpp"
 #include <vector>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/api/gather.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/gather.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/gather.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/gather.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/gather_nd.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/gather_nd.hpp
similarity index 63%
rename from inference-engine/thirdparty/clDNN/api/gather_nd.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/gather_nd.hpp
index ebf3953cf7fd65..885f7297510f11 100644
--- a/inference-engine/thirdparty/clDNN/api/gather_nd.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/gather_nd.hpp
@@ -1,20 +1,7 @@
-/*
-// Copyright (c) 2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
 //
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-*/
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 #include "primitive.hpp"
 
diff --git a/inference-engine/thirdparty/clDNN/api/gather_tree.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/gather_tree.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/gather_tree.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/gather_tree.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/gemm.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/gemm.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/gemm.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/gemm.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/grn.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/grn.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/grn.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/grn.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/input_layout.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/input_layout.hpp
similarity index 97%
rename from inference-engine/thirdparty/clDNN/api/input_layout.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/input_layout.hpp
index 61cc5544e7acb6..c6be1bf5b39de6 100644
--- a/inference-engine/thirdparty/clDNN/api/input_layout.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/input_layout.hpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 #include "primitive.hpp"
-#include "memory.hpp"
+#include "cldnn/runtime/memory.hpp"
 
 namespace cldnn {
 /// @addtogroup cpp_api C++ API
diff --git a/inference-engine/thirdparty/clDNN/api/loop.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/loop.hpp
similarity index 99%
rename from inference-engine/thirdparty/clDNN/api/loop.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/loop.hpp
index 8268b728e12467..63215f3ad980b5 100644
--- a/inference-engine/thirdparty/clDNN/api/loop.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/loop.hpp
@@ -7,7 +7,7 @@
 #include <vector>
 #include <functional>
 #include "primitive.hpp"
-#include "topology.hpp"
+#include "cldnn/graph/topology.hpp"
 
 #define DEFAULT_MAX_NUM_ITERATION 256
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/api/lrn.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/lrn.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/lrn.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/lrn.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/lstm.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/lstm.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/lstm.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/lstm.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/lstm_dynamic.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/lstm_dynamic.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/lstm_dynamic.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/lstm_dynamic.hpp
diff --git a/inference-engine/thirdparty/clDNN/api_extension/lstm_dynamic_input.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/lstm_dynamic_input.hpp
similarity index 98%
rename from inference-engine/thirdparty/clDNN/api_extension/lstm_dynamic_input.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/lstm_dynamic_input.hpp
index 1a3d0fb7bff92d..3733a989fe68d4 100644
--- a/inference-engine/thirdparty/clDNN/api_extension/lstm_dynamic_input.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/lstm_dynamic_input.hpp
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/primitive.hpp"
+#include "cldnn/primitives/primitive.hpp"
 #include <vector>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/api_extension/lstm_dynamic_timeloop.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/lstm_dynamic_timeloop.hpp
similarity index 99%
rename from inference-engine/thirdparty/clDNN/api_extension/lstm_dynamic_timeloop.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/lstm_dynamic_timeloop.hpp
index 5449c95b01b8d9..c7b77fb1afe064 100644
--- a/inference-engine/thirdparty/clDNN/api_extension/lstm_dynamic_timeloop.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/lstm_dynamic_timeloop.hpp
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/primitive.hpp"
+#include "cldnn/primitives/primitive.hpp"
 #include <vector>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/api/max_unpooling.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/max_unpooling.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/max_unpooling.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/max_unpooling.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/mutable_data.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/mutable_data.hpp
similarity index 92%
rename from inference-engine/thirdparty/clDNN/api/mutable_data.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/mutable_data.hpp
index 6a8acc0770a1bc..a5aca1f4392063 100644
--- a/inference-engine/thirdparty/clDNN/api/mutable_data.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/mutable_data.hpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 #include "primitive.hpp"
-#include "memory.hpp"
+#include "cldnn/runtime/memory.hpp"
 #include <vector>
 
 namespace cldnn {
@@ -31,7 +31,7 @@ struct mutable_data : public primitive_base<mutable_data> {
     /// @param mem @ref memory object which contains data.
     /// @param filler_type @ref data filling function, default is zero
     /// @note If memory is attached by memory::attach(), the attached buffer should be valid till network build.
-    mutable_data(const primitive_id& id, const memory& mem, filler_type fill_type = filler_type::no_fill)
+    mutable_data(const primitive_id& id, memory::ptr mem, filler_type fill_type = filler_type::no_fill)
         : primitive_base(id, {}, padding()), mem(mem), fill_type(fill_type) {}
 
     /// @brief Constructs mutable_data primitive with inputs.
@@ -42,13 +42,13 @@ struct mutable_data : public primitive_base<mutable_data> {
     /// @param filler_type @ref data filling function, default is zero
     mutable_data(const primitive_id& id,
                  const std::vector<primitive_id>& input,
-                 const memory& mem,
+                 memory::ptr mem,
                  filler_type fill_type = filler_type::no_fill)
         : primitive_base(id, {input}, padding()), mem(mem), fill_type(fill_type) {}
 
     /// @brief @ref memory object which contains data.
     /// @note If memory is attached by memory::attach(), the attached buffer should be valid till network build.
-    memory mem;
+    memory::ptr mem;
 
     /// @brief Specifies function which will be used to fill weights.
     filler_type fill_type;
diff --git a/inference-engine/thirdparty/clDNN/api/mvn.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/mvn.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/mvn.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/mvn.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/non_max_suppression.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/non_max_suppression.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/non_max_suppression.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/non_max_suppression.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/normalize.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/normalize.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/normalize.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/normalize.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/one_hot.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/one_hot.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/one_hot.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/one_hot.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/permute.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/permute.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/permute.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/permute.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/pooling.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/pooling.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/pooling.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/pooling.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/primitive.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/primitive.hpp
similarity index 98%
rename from inference-engine/thirdparty/clDNN/api/primitive.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/primitive.hpp
index 1c8ac408493c0b..856df2e56496af 100644
--- a/inference-engine/thirdparty/clDNN/api/primitive.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/primitive.hpp
@@ -3,11 +3,11 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
+
 #pragma once
 
-#include "cldnn.hpp"
-#include "compounds.h"
-#include "layout.hpp"
+#include "cldnn/runtime/compounds.hpp"
+#include "cldnn/runtime/layout.hpp"
 
 #include <algorithm>
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/api/prior_box.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/prior_box.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/prior_box.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/prior_box.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/proposal.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/proposal.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/proposal.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/proposal.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/pyramid_roi_align.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/pyramid_roi_align.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/pyramid_roi_align.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/pyramid_roi_align.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/quantize.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/quantize.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/quantize.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/quantize.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/reduce.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/reduce.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/reduce.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/reduce.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/region_yolo.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/region_yolo.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/region_yolo.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/region_yolo.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/reorder.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/reorder.hpp
similarity index 99%
rename from inference-engine/thirdparty/clDNN/api/reorder.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/reorder.hpp
index ab597d7de189ca..c980c34f0d4216 100644
--- a/inference-engine/thirdparty/clDNN/api/reorder.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/reorder.hpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 #include "primitive.hpp"
-#include "memory.hpp"
+#include "cldnn/runtime/memory.hpp"
 #include <vector>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/api/reorg_yolo.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/reorg_yolo.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/reorg_yolo.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/reorg_yolo.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/resample.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/resample.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/resample.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/resample.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/reshape.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/reshape.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/reshape.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/reshape.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/reverse_sequence.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/reverse_sequence.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/reverse_sequence.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/reverse_sequence.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/roi_pooling.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/roi_pooling.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/roi_pooling.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/roi_pooling.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/scale.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/scale.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/scale.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/scale.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/scatter_elements_update.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/scatter_elements_update.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/scatter_elements_update.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/scatter_elements_update.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/scatter_nd_update.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/scatter_nd_update.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/scatter_nd_update.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/scatter_nd_update.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/scatter_update.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/scatter_update.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/scatter_update.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/scatter_update.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/select.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/select.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/select.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/select.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/shuffle_channels.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/shuffle_channels.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/shuffle_channels.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/shuffle_channels.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/softmax.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/softmax.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/softmax.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/softmax.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/space_to_batch.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/space_to_batch.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/space_to_batch.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/space_to_batch.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/space_to_depth.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/space_to_depth.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/space_to_depth.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/space_to_depth.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/split.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/split.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/split.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/split.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/strided_slice.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/strided_slice.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/strided_slice.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/strided_slice.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/tile.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/primitives/tile.hpp
similarity index 100%
rename from inference-engine/thirdparty/clDNN/api/tile.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/primitives/tile.hpp
diff --git a/inference-engine/thirdparty/clDNN/api/compounds.h b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/compounds.hpp
similarity index 98%
rename from inference-engine/thirdparty/clDNN/api/compounds.h
rename to inference-engine/thirdparty/clDNN/api/cldnn/runtime/compounds.hpp
index 45429e584f9b46..ade492705a5c19 100644
--- a/inference-engine/thirdparty/clDNN/api/compounds.h
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/compounds.hpp
@@ -4,6 +4,8 @@
 
 #pragma once
 
+#include "utils.hpp"
+
 #include <vector>
 #include <cassert>
 #include <iterator>
@@ -11,7 +13,6 @@
 #include <string>
 #include <stdexcept>
 
-#include "meta_utils.hpp"
 
 namespace cldnn {
 
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/device.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/device.hpp
new file mode 100644
index 00000000000000..7f5e1f13d7d1e7
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/device.hpp
@@ -0,0 +1,24 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "device_info.hpp"
+#include "memory_caps.hpp"
+
+#include <memory>
+
+namespace cldnn {
+
+/// @brief Represents detected GPU device object. Use device_query to get list of available objects.
+struct device {
+public:
+    using ptr = std::shared_ptr<device>;
+    virtual device_info get_info() const = 0;
+    virtual memory_capabilities get_mem_caps() const = 0;
+
+    virtual ~device() = default;
+};
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/device_info.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/device_info.hpp
new file mode 100644
index 00000000000000..60ac318340d241
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/device_info.hpp
@@ -0,0 +1,76 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <string>
+
+namespace cldnn {
+/// @addtogroup cpp_api C++ API
+/// @{
+
+/// @defgroup cpp_device GPU Device
+/// @{
+
+/// @brief Enumeration of supported device types
+enum class device_type {
+    integrated_gpu = 0,
+    discrete_gpu = 1
+};
+
+/// @brief Defines version of GFX IP
+struct gfx_version {
+    uint16_t major;
+    uint8_t minor;
+    uint8_t revision;
+};
+
+/// @brief Information about the device properties and capabilities.
+struct device_info {
+    uint32_t execution_units_count;             ///< Number of available execution units.
+    uint32_t gpu_frequency;                     ///< Clock frequency in MHz.
+    uint32_t max_threads_per_execution_unit;    ///< Number of available HW threads on EU.
+    uint32_t max_threads_per_device;            ///< Maximum number of HW threads on device.
+
+    uint64_t max_work_group_size;               ///< Maximum number of work-items in a work-group executing a kernel using the data parallel execution model.
+    uint64_t max_local_mem_size;                ///< Maximum size of local memory arena in bytes.
+    uint64_t max_global_mem_size;               ///< Maximum size of global device memory in bytes.
+    uint64_t max_alloc_mem_size;                ///< Maximum size of memory object allocation in bytes.
+
+    uint64_t max_image2d_width;                 ///< Maximum image 2d width supported by the device.
+    uint64_t max_image2d_height;                ///< Maximum image 2d height supported by the device.
+
+    bool supports_fp16;                         ///< Does engine support FP16.
+    bool supports_fp64;                         ///< Does engine support FP64.
+    bool supports_fp16_denorms;                 ///< Does engine support denormalized FP16.
+    bool supports_subgroups;                    ///< Does engine support cl_intel_subgroups extension.
+    bool supports_subgroups_short;              ///< Does engine support cl_intel_subgroups_short extension.
+    bool supports_subgroups_char;               ///< Does engine support cl_intel_subgroups_char extension.
+    bool supports_local_block_io;               ///< Does engine support cl_intel_subgroup_local_block_io extension.
+    bool supports_image;                        ///< Does engine support images (CL_DEVICE_IMAGE_SUPPORT cap).
+
+    bool supports_imad;                         ///< Does engine support int8 mad.
+    bool supports_immad;                        ///< Does engine support int8 multi mad.
+
+    bool supports_usm;                          ///< Does engine support unified shared memory.
+
+    uint32_t vendor_id;                         ///< Vendor ID
+    std::string dev_name;                       ///< Device ID string
+    std::string driver_version;                 ///< Version of OpenCL driver
+
+    device_type dev_type;                       ///< Defines type of current GPU device (integrated or discrete)
+
+    gfx_version gfx_ver;                        ///< Defines GFX IP version
+    uint32_t device_id;                         ///< ID of current GPU
+    uint32_t num_slices;                        ///< Number of slices
+    uint32_t num_sub_slices_per_slice;          ///< Number of subslices in a slice
+    uint32_t num_eus_per_sub_slice;             ///< Number of execution units per subslice
+    uint32_t num_threads_per_eu;                ///< Number of hardware threads per execution unit
+};
+
+/// @}
+
+/// @}
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/device_query.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/device_query.hpp
new file mode 100644
index 00000000000000..af458360288641
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/device_query.hpp
@@ -0,0 +1,29 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "device.hpp"
+#include "engine_configuration.hpp"
+
+#include <map>
+#include <string>
+#include <algorithm>
+
+namespace cldnn {
+
+// Fetches all available gpu devices with specific runtime and engine types and (optionally) user context/device handles
+struct device_query {
+public:
+    explicit device_query(engine_types engine_type, runtime_types runtime_type, void* user_context = nullptr, void* user_device = nullptr);
+
+    std::map<std::string, device::ptr> get_available_devices() const {
+        return _available_devices;
+    }
+
+    ~device_query() = default;
+private:
+    std::map<std::string, device::ptr> _available_devices;
+};
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/engine.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/engine.hpp
new file mode 100644
index 00000000000000..b81f626c5e64f4
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/engine.hpp
@@ -0,0 +1,150 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "device.hpp"
+#include "engine_configuration.hpp"
+#include "event.hpp"
+#include "memory_caps.hpp"
+#include "memory_pool.hpp"
+#include "layout.hpp"
+
+#include <memory>
+#include <set>
+#include <utility>
+#include <string>
+
+namespace cldnn {
+
+class stream;
+
+using memory_ptr = std::shared_ptr<memory>;
+using stream_ptr = std::shared_ptr<stream>;
+
+using primitive_id = std::string;
+
+class engine {
+public:
+    /// Default destructor
+    virtual ~engine() = default;
+
+    /// Returns type of the engine
+    virtual engine_types type() const = 0;
+    /// Returns runtime type used in the engine
+    virtual runtime_types runtime_type() const = 0;
+
+    /// Create memory object with specified @p layout and allocation @p type for primitive with @p id
+    /// Underlying memory handle can be reused with other primitives from memory pool based on @p dependencies
+    memory_ptr get_memory_from_pool(const layout& layout,
+                                    primitive_id id,
+                                    uint32_t network_id,
+                                    std::set<primitive_id> dependencies,
+                                    allocation_type type,
+                                    bool reusable = true);
+
+    /// Create memory object attached to the buffer allocated by user.
+    /// @param ptr  The pointer to user allocated buffer.
+    /// @note Size (in bytes) of the buffer should be equal to @p layout.bytes_count()
+    /// User is responsible for buffer deallocation. Buffer lifetime should be bigger than lifetime of the memory object.
+    memory_ptr attach_memory(const layout& layout, void* ptr);
+
+    /// Allocate gpu memory using specified @p layout and alloation @p type
+    virtual memory_ptr allocate_memory(const layout& layout, allocation_type type, bool reset = true) = 0;
+
+    /// Allocate gpu memory using specified @p layout. Allocation type is selected automatically based on engine/device configuration
+    memory_ptr allocate_memory(const layout& layout, bool reset = true);
+
+    /// Created memory object from memory @p params and reinterpred the data using specified @p layout
+    virtual memory_ptr reinterpret_handle(const layout& new_layout, shared_mem_params params) = 0;
+
+    /// Created memory object from the other @p memory and reinterpred the data using specified @p new_layout
+    virtual memory_ptr reinterpret_buffer(const memory& memory, const layout& new_layout) = 0;
+
+    /// Create shared memory object using user-supplied memory buffer @p buf using specified @p layout
+    memory_ptr share_buffer(const layout& layout, shared_handle buf);
+
+    /// Create shared memory object using user-supplied 2D image @p img using specified @p layout
+    memory_ptr share_image(const layout& layout, shared_handle img);
+
+    /// Create shared memory object over specified @p plane of video decoder surface @p surf using specified @p layout
+#ifdef _WIN32
+    memory_ptr share_surface(const layout& layout, shared_handle surf, uint32_t plane);
+    memory_ptr share_dx_buffer(const layout& layout, shared_handle res);
+#else
+    memory_ptr share_surface(const layout& layout, shared_surface surf, uint32_t plane);
+#endif
+
+    /// Checks whether two memory objects represents the same physical memory
+    virtual bool is_the_same_buffer(const memory& mem1, const memory& mem2) = 0;
+
+    /// Returns basic allocation type which will be used as a fallback when allocation type is not specified or device doesn't support some features.
+    virtual allocation_type get_default_allocation_type() const = 0;
+
+    /// Returns preferred allocation type which can be mapped to host ptr
+    allocation_type get_lockable_preffered_memory_allocation_type(bool is_image_layout = false) const;
+
+    /// Checks if the current engine supports speicied allocation @p type
+    bool supports_allocation(allocation_type type) const;
+
+    /// Returns configuration of current engine
+    const engine_configuration& configuration() const { return _configuration; }
+
+    /// Returns device structure which represents stores device capabilities
+    device_info get_device_info() const;
+
+    /// Returns device object associated with the engine
+    const device::ptr get_device() const;
+
+    /// Returns memory pool for the engine
+    memory_pool& get_memory_pool();
+
+    /// Returns user context handle which was used to create the engine
+    virtual void* get_user_context() const = 0;
+
+    /// Returns the maximum amount of GPU memory that engine allocated in current process
+    uint64_t get_max_used_device_memory() const;
+
+    /// Returns the amount of GPU memory currently used by the engine
+    uint64_t get_used_device_memory() const;
+
+    /// Returns true if USM is enabled in engine config and device/driver supports required features
+    bool use_unified_shared_memory() const;
+
+    /// Create stream object for current engine
+    virtual stream_ptr create_stream() const = 0;
+
+    /// Returns service stream which can be used during program build and optimizations
+    virtual stream& get_program_stream() const = 0;
+
+    /// Factory method which creates engine object with impl configured by @p engine_type
+    /// @param engine_type requested engine type
+    /// @param runtime_type requested execution runtime for the engine. @note some runtime/engine types configurations might be unsupported
+    /// @param device specifies the device which the engine is created for
+    /// @param configuration options for the engine
+    static std::shared_ptr<cldnn::engine> create(engine_types engine_type,
+                                                 runtime_types runtime_type,
+                                                 const device::ptr device,
+                                                 const engine_configuration& configuration = engine_configuration());
+
+    /// Factory method which creates engine object with impl configured by @p engine_type
+    /// @param engine_type requested engine type
+    /// @param runtime_type requested execution runtime for the engine. @note some runtime/engine types configurations might be unsupported
+    /// @param configuration options for the engine
+    /// @note engine is created for the first device returned by devices query
+    static std::shared_ptr<cldnn::engine> create(engine_types engine_type,
+                                                 runtime_types runtime_type,
+                                                 const engine_configuration& configuration = engine_configuration());
+
+protected:
+    /// Create engine for given @p device and @p configuration
+    engine(const device::ptr device, const engine_configuration& configuration);
+
+    // TODO: Consider moving memory pool to cldnn::network
+    std::unique_ptr<memory_pool> _memory_pool;
+    const device::ptr _device;
+    engine_configuration _configuration;
+};
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/engine_configuration.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/engine_configuration.hpp
new file mode 100644
index 00000000000000..2cbaca414ab594
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/engine_configuration.hpp
@@ -0,0 +1,110 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "utils.hpp"
+
+#include <string>
+#include <stdexcept>
+#include <thread>
+
+namespace cldnn {
+
+/// @addtogroup cpp_api C++ API
+/// @{
+
+/// @defgroup cpp_engine Execution Engine
+/// @{
+
+/// @brief Defines available engine types
+enum class engine_types : int32_t {
+    ocl,
+};
+
+/// @brief Defines available runtime types
+enum class runtime_types : int32_t {
+    ocl,
+};
+
+/// @brief Defines available priority mode types
+enum class priority_mode_types : int16_t {
+    disabled,
+    low,
+    med,
+    high
+};
+
+/// @brief Defines available throttle mode types
+enum class throttle_mode_types : int16_t {
+    disabled,
+    low,
+    med,
+    high
+};
+
+/// @brief Defines supported queue types
+enum class queue_types : int16_t {
+    in_order,
+    out_of_order
+};
+
+/// @brief Configuration parameters for created engine.
+struct engine_configuration {
+    const bool enable_profiling;              ///< Enable per-primitive profiling.
+    const queue_types queue_type;             ///< Specifies type of queue used by the runtime
+    const std::string sources_dumps_dir;      ///< Specifies a directory where sources of cldnn::program objects should be dumped.
+                                              ///< Empty by default (means no dumping).
+    const priority_mode_types priority_mode;  ///< Priority mode (support of priority hints in command queue). If cl_khr_priority_hints extension
+                                              ///< is not supported by current OpenCL implementation, the value must be set to cldnn_priority_disabled.
+
+    const throttle_mode_types throttle_mode;  ///< Throttle mode (support of throttle hints in command queue). If cl_khr_throttle_hints extension
+                                              ///< is not supported by current OpenCL implementation, the value must be set to cldnn_throttle_disabled.
+
+    bool use_memory_pool;                     ///< Enables memory usage optimization. memory objects will be reused when possible
+                                              ///< (switched off for older drivers then NEO).
+    bool use_unified_shared_memory;           ///< Enables USM usage
+    const std::string kernels_cache_path;     ///< Path to compiled kernels cache
+    uint16_t n_threads;                       ///< Max number of host threads used in gpu plugin
+    const std::string tuning_cache_path;      ///< Path to tuning kernel cache
+
+    /// @brief Constructs engine configuration with specified options.
+    /// @param enable_profiling Enable per-primitive profiling.
+    /// @param queue_type Specifies type of queue used by the runtime
+    /// @param sources_dumps_dir Specifies a directory where sources of cldnn::program objects should be dumped
+    /// @param priority_mode Priority mode for all streams created within the engine
+    /// @param throttle_mode Throttle mode for all streams created within the engine
+    /// @param use_memory_pool Controls whether engine is allowed to reuse intermediate memory buffers whithin a network
+    /// @param use_unified_shared_memory If this option it true and device supports USM, then engine will use USM for all memory allocations
+    /// @param kernels_cache_path Path to existing directory where plugin can cache compiled kernels
+    /// @param n_threads Max number of host threads used in gpu plugin
+    /// @param tuning_cache_path Path to tuning kernel cache
+    engine_configuration(
+        bool enable_profiling = false,
+        queue_types queue_type = queue_types::out_of_order,
+        const std::string& sources_dumps_dir = std::string(),
+        priority_mode_types priority_mode = priority_mode_types::disabled,
+        throttle_mode_types throttle_mode = throttle_mode_types::disabled,
+        bool use_memory_pool = true,
+        bool use_unified_shared_memory = true,
+        const std::string& kernels_cache_path = "",
+        uint16_t n_threads = std::max(static_cast<uint16_t>(std::thread::hardware_concurrency()), static_cast<uint16_t>(1)),
+        const std::string& tuning_cache_path = "cache.json")
+        : enable_profiling(enable_profiling)
+        , queue_type(queue_type)
+        , sources_dumps_dir(sources_dumps_dir)
+        , priority_mode(priority_mode)
+        , throttle_mode(throttle_mode)
+        , use_memory_pool(use_memory_pool)
+        , use_unified_shared_memory(use_unified_shared_memory)
+        , kernels_cache_path(kernels_cache_path)
+        , n_threads(n_threads)
+        , tuning_cache_path(tuning_cache_path) { }
+};
+
+/// @}
+
+/// @}
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/error_handler.h b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/error_handler.hpp
similarity index 99%
rename from inference-engine/thirdparty/clDNN/src/include/error_handler.h
rename to inference-engine/thirdparty/clDNN/api/cldnn/runtime/error_handler.hpp
index 1ce66e39c9f134..7f57d64861c413 100644
--- a/inference-engine/thirdparty/clDNN/src/include/error_handler.h
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/error_handler.hpp
@@ -2,14 +2,15 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
+
+#include "layout.hpp"
+
 #include <sstream>
 #include <vector>
 #include <array>
 #include <algorithm>
 #include <type_traits>
-#include "api/layout.hpp"
 #include <string>
 #include <utility>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/event_impl.h b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/event.hpp
similarity index 82%
rename from inference-engine/thirdparty/clDNN/src/include/event_impl.h
rename to inference-engine/thirdparty/clDNN/api/cldnn/runtime/event.hpp
index c87e34a3f84507..db2eda1a501bae 100644
--- a/inference-engine/thirdparty/clDNN/src/include/event_impl.h
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/event.hpp
@@ -2,21 +2,26 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/event.hpp"
-#include "refcounted_obj.h"
+
+#include "profiling.hpp"
 
 #include <list>
 #include <mutex>
 #include <utility>
+#include <utility>
+#include <functional>
 
 namespace cldnn {
 struct user_event;
 
-struct event_impl : public refcounted_obj<event_impl> {
+/// @brief user-defined event handler callback.
+using event_handler = std::function<void(void*)>;
+
+struct event {
 public:
-    event_impl() = default;
+    using ptr = std::shared_ptr<event>;
+    event() = default;
 
     void wait();
     bool is_set();
@@ -27,10 +32,11 @@ struct event_impl : public refcounted_obj<event_impl> {
         _profiling_captured = false;
         _profiling_info.clear();
     }
+
     // returns true if handler has been successfully added
     bool add_event_handler(event_handler handler, void* data);
 
-    const std::list<instrumentation::profiling_interval>& get_profiling_info();
+    std::vector<instrumentation::profiling_interval> get_profiling_info();
 
 private:
     std::mutex _handlers_mutex;
@@ -54,7 +60,7 @@ struct event_impl : public refcounted_obj<event_impl> {
     virtual bool get_profiling_info_impl(std::list<instrumentation::profiling_interval>&) { return true; }
 };
 
-struct user_event : virtual public event_impl {
+struct user_event : virtual public event {
 public:
     explicit user_event(bool set = false) { _set = set; }
 
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/half.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/half.hpp
new file mode 100644
index 00000000000000..785cc711d8f977
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/half.hpp
@@ -0,0 +1,55 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <stdint.h>
+#include <stddef.h>
+#include <memory>
+#include <string>
+#include <type_traits>
+
+namespace cldnn {
+
+/// @addtogroup cpp_api C++ API
+/// @{
+
+float half_to_float(uint16_t value);
+uint16_t float_to_half(float value);
+
+// There is no portable half precision floating point support.
+// Using wrapped integral type with the same size and alignment restrictions.
+class half_impl {
+public:
+    half_impl() = default;
+
+    template <typename T, typename = typename std::enable_if<!std::is_floating_point<T>::value>::type>
+    explicit half_impl(T data, int /*direct_creation_tag*/) : _data(data) {}
+
+    operator uint16_t() const { return _data; }
+    operator float() const {
+        return half_to_float(_data);
+    }
+
+    explicit half_impl(float value)
+        : _data(float_to_half(value))
+    {}
+
+    template <typename T, typename = typename std::enable_if<std::is_convertible<T, float>::value>::type>
+    explicit half_impl(T value)
+        : half_impl(static_cast<float>(value))
+    {}
+
+private:
+    uint16_t _data;
+};
+
+// Use complete implementation if necessary.
+#if defined HALF_HALF_HPP
+using half_t = half;
+#else
+using half_t = half_impl;
+#endif
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/kernel.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/kernel.hpp
new file mode 100644
index 00000000000000..0d75173e7723d9
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/kernel.hpp
@@ -0,0 +1,24 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "kernel_args.hpp"
+#include "event.hpp"
+
+#include <memory>
+#include <vector>
+
+namespace cldnn {
+
+using kernel_id = std::string;
+
+class kernel {
+public:
+    using ptr = std::shared_ptr<kernel>;
+    virtual std::shared_ptr<kernel> clone() const = 0;
+    virtual ~kernel() = default;
+};
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/kernel_args.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/kernel_args.hpp
new file mode 100644
index 00000000000000..f70d04ee33a771
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/kernel_args.hpp
@@ -0,0 +1,133 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "memory.hpp"
+
+#include <memory>
+#include <vector>
+
+namespace cldnn {
+
+struct work_group_sizes {
+    std::vector<size_t> global;
+    std::vector<size_t> local;
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Scalar
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+struct scalar_desc {
+    union ValueT {
+        uint8_t u8;
+        uint16_t u16;
+        uint32_t u32;
+        uint64_t u64;
+        int8_t s8;
+        int16_t s16;
+        int32_t s32;
+        int64_t s64;
+        float f32;
+        double f64;
+    };
+
+    enum class Types {
+        UINT8,
+        UINT16,
+        UINT32,
+        UINT64,
+        INT8,
+        INT16,
+        INT32,
+        INT64,
+        FLOAT32,
+        FLOAT64,
+    };
+
+    Types t;
+    ValueT v;
+};
+
+using scalars_desc = std::vector<scalar_desc>;
+
+
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// ArgumentDescpirtor
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+struct argument_desc {
+    enum class Types {
+        INPUT,
+        OUTPUT,
+        WEIGHTS,
+        BIAS,
+        SCALE_TABLE,
+        SLOPE,
+        SPLIT,
+        INTERNAL_BUFFER,
+        SCALAR,
+        RECURRENT,  // RNN/LSTM/GRU recurrent weights
+        HIDDEN,     // RNN/LSTM/GRU hidden input
+        CELL,       // LSTM cell input
+        LSTM_PACK,  // LSTM packed output
+        WEIGHTS_ZERO_POINTS,
+        ACTIVATIONS_ZERO_POINTS,
+        COMPENSATION,
+        INPUT_OF_FUSED_PRIMITIVE
+    };
+
+    Types t;
+    uint32_t index;
+};
+
+using arguments_desc = std::vector<argument_desc>;
+
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// KernelParams
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+struct kernel_arguments_desc {
+    work_group_sizes workGroups;
+    arguments_desc arguments;
+    scalars_desc scalars;
+    std::string layerID;
+};
+
+struct kernel_arguments_data {
+    std::vector<memory::cptr> inputs;
+    std::vector<memory::cptr> intermediates;
+    memory::cptr output;
+    memory::cptr weights;
+    memory::cptr recurrent;
+    memory::cptr hidden;
+    memory::cptr cell;
+    memory::cptr bias;
+    memory::cptr weights_zero_points;
+    memory::cptr activations_zero_points;
+    memory::cptr compensation;
+    memory::cptr lookup_table;
+    memory::cptr scale_table;
+    memory::cptr slope;
+
+    std::vector<memory::cptr> fused_op_inputs;
+    int32_t split = 0;
+    const scalars_desc* scalars = nullptr;
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// KernelString
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+struct kernel_string {
+    std::string str;
+    std::string jit;
+    std::string undefs;
+    std::string options;
+    std::string entry_point;
+    bool batch_compilation;
+
+    kernel_string() : str(""), jit(""), undefs(""), options(""), entry_point(""), batch_compilation(false) {}
+
+    std::string get_hash() { return str + jit + undefs + options + entry_point; }
+};
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/layout.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/layout.hpp
similarity index 99%
rename from inference-engine/thirdparty/clDNN/api/layout.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/runtime/layout.hpp
index ab4b17f03c9efb..4962cd009639cc 100644
--- a/inference-engine/thirdparty/clDNN/api/layout.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/layout.hpp
@@ -2,9 +2,11 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
+
 #include "tensor.hpp"
+#include "half.hpp"
+
 #include <cmath>
 #include <cstdlib>
 #include <vector>
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory.hpp
new file mode 100644
index 00000000000000..e054e90a477bd9
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory.hpp
@@ -0,0 +1,150 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "layout.hpp"
+#include "memory_caps.hpp"
+#include "event.hpp"
+#include "engine_configuration.hpp"
+
+namespace cldnn {
+
+class engine;
+class stream;
+
+struct memory {
+    using ptr = std::shared_ptr<memory>;
+    using cptr = std::shared_ptr<const memory>;
+    memory(engine* engine, const layout& layout,  allocation_type type, bool reused = false);
+
+    virtual ~memory();
+    virtual void* lock(const stream& stream) = 0;
+    virtual void unlock(const stream& stream) = 0;
+    virtual event::ptr fill(stream& stream, unsigned char pattern) = 0;
+    virtual event::ptr fill(stream& stream) = 0;
+
+    size_t size() const { return _bytes_count; }
+    size_t count() const { return _layout.count(); }
+    virtual shared_mem_params get_internal_params() const = 0;
+    virtual bool is_allocated_by(const engine& engine) const { return &engine == _engine; }
+    engine* get_engine() const { return _engine; }
+    const layout& get_layout() const { return _layout; }
+    allocation_type get_allocation_type() const { return _type; }
+    // TODO: must be moved outside memory class
+    virtual bool is_memory_reset_needed(layout l) {
+        // To avoid memory reset, output memory must meet the following requirements:
+        // - To be Weights format (Data memory can be reused by memory_pool, which can lead to errors)
+        // - To have zero paddings
+        // - To be completely filled with data
+        if ((!format::is_weights_format(l.format) && !format::is_simple_data_format(l.format)) ||
+             format::is_winograd(l.format) || format::is_image_2d(l.format)) {
+            return true;
+        }
+
+        if (l.data_padding.lower_size() != tensor(0) || l.data_padding.upper_size() != tensor(0)) {
+            return true;
+        }
+
+        if (_bytes_count == (l.data_type == data_types::bin ? ceil_div(l.count(), 32) : l.count()) * data_type_traits::size_of(l.data_type)) {
+            return false;
+        }
+
+        return true;
+    }
+
+    virtual event::ptr copy_from(stream& /* stream */, const memory& /* other */) = 0;
+    virtual event::ptr copy_from(stream& /* stream */, const void* /* host_ptr */) = 0;
+
+protected:
+    engine* _engine;
+    const layout _layout;
+    // layout bytes count, needed because of traits static map destruction
+    // before run of memory destructor, when engine is static
+    size_t _bytes_count;
+
+private:
+    allocation_type _type;
+    bool _reused;
+};
+
+struct simple_attached_memory : memory {
+    simple_attached_memory(const layout& layout, void* pointer)
+        : memory(nullptr, layout, allocation_type::unknown), _pointer(pointer) {}
+
+    void* lock(const stream& /* stream */) override { return _pointer; }
+    void unlock(const stream& /* stream */) override {}
+    event::ptr fill(stream& /* stream */, unsigned char) override { return nullptr; }
+    event::ptr fill(stream& /* stream */) override { return nullptr; }
+    shared_mem_params get_internal_params() const override { return { shared_mem_type::shared_mem_empty, nullptr, nullptr, nullptr,
+#ifdef _WIN32
+        nullptr,
+#else
+        0,
+#endif
+        0}; };
+
+    event::ptr copy_from(stream& /* stream */, const memory& /* other */) override { return nullptr; };
+    event::ptr copy_from(stream& /* stream */, const void* /* host_ptr */) override { return nullptr; }
+
+private:
+    void* _pointer;
+};
+
+template <class T>
+struct mem_lock {
+    explicit mem_lock(memory::ptr mem, const stream& stream) : _mem(mem), _stream(stream), _ptr(reinterpret_cast<T*>(_mem->lock(_stream))) {}
+
+    ~mem_lock() {
+        _ptr = nullptr;
+        _mem->unlock(_stream);
+    }
+
+    size_t size() const { return _mem->size() / sizeof(T); }
+
+    mem_lock(const mem_lock& other) = delete;
+    mem_lock& operator=(const mem_lock& other) = delete;
+
+#if defined(_SECURE_SCL) && (_SECURE_SCL > 0)
+    auto begin() & { return stdext::make_checked_array_iterator(_ptr, size()); }
+    auto end() & { return stdext::make_checked_array_iterator(_ptr, size(), size()); }
+#else
+    T* begin() & { return _ptr; }
+    T* end() & { return _ptr + size(); }
+#endif
+
+    /// @brief Provides indexed access to pointed memory.
+    T& operator[](size_t idx) const& {
+        assert(idx < size());
+        return _ptr[idx];
+    }
+
+    T* data() const & { return _ptr; }
+
+    /// Prevents to use mem_lock as temporary object
+    T* data() && = delete;
+    /// Prevents to use mem_lock as temporary object
+    T* begin() && = delete;
+    /// Prevents to use mem_lock as temporary object
+    T* end() && = delete;
+    /// Prevents to use mem_lock as temporary object
+    T& operator[](size_t idx) && = delete;
+
+private:
+    memory::ptr _mem;
+    const stream& _stream;
+    T* _ptr;
+};
+
+struct surfaces_lock {
+    surfaces_lock() = default;
+    virtual ~surfaces_lock() = default;
+
+    surfaces_lock(const surfaces_lock& other) = delete;
+    surfaces_lock& operator=(const surfaces_lock& other) = delete;
+
+    static std::unique_ptr<surfaces_lock> create(engine_types engine_type, std::vector<memory::ptr> mem, const stream& stream);
+};
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory_caps.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory_caps.hpp
new file mode 100644
index 00000000000000..98f9eee658103f
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory_caps.hpp
@@ -0,0 +1,84 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <map>
+#include <string>
+#include <vector>
+#include <algorithm>
+
+namespace cldnn {
+
+enum class allocation_type {
+    unknown,     // Not specified (i.e simple_attached_memory class).
+    cl_mem,      // Use standard OpenCL cl_mem allocations.
+    usm_host,    // Accessible by host and device. Not Migratable
+    usm_shared,  // Accessible by host and device. Migrtable.
+    usm_device,  // Accessible only by device. Not migratable.
+};
+
+class memory_capabilities {
+public:
+    memory_capabilities(std::vector<allocation_type> supported_allocation_types) : _caps(supported_allocation_types) {}
+    bool supports_usm() const {
+        return find_in_caps(allocation_type::usm_host) ||
+               find_in_caps(allocation_type::usm_shared) ||
+               find_in_caps(allocation_type::usm_device);
+    }
+    bool support_allocation_type(allocation_type type) const { return find_in_caps(type); }
+
+    static bool is_usm_type(allocation_type type) {
+        if (type == allocation_type::usm_host ||
+            type == allocation_type::usm_shared ||
+            type == allocation_type::usm_device)
+            return true;
+        return false;
+    }
+
+private:
+    std::vector<allocation_type> _caps;
+
+    bool find_in_caps(const allocation_type& type) const {
+        return std::find_if(_caps.begin(), _caps.end(), [&](const allocation_type& t) { return t == type; }) != _caps.end();
+    }
+};
+
+
+/// @brief Shared memory descriptor type.
+enum class shared_mem_type {
+    /// @brief Structure unitialized or contains no information.
+    shared_mem_empty,
+
+    /// @brief Structure describes shared CL buffer.
+    shared_mem_buffer,
+
+    /// @brief Structure describes shared CL image.
+    shared_mem_image,
+
+    /// @brief Structure describes shared VA/DXVA surface
+    shared_mem_vasurface,
+
+    /// @brief Structure describes shared D3D11 buffer
+    shared_mem_dxbuffer
+};
+
+using shared_handle = void*;
+using shared_surface = uint32_t;
+
+/// @brief Low-level API handles required for using cldnn memory objects in external API calls.
+struct shared_mem_params {
+    shared_mem_type mem_type;     ///< shared buffer type
+    shared_handle context;        ///< OpenCL context for external operations
+    shared_handle user_device;    ///< DX/VA device for external operations
+    shared_handle mem;            ///< memory object handle
+#ifdef _WIN32
+    shared_handle surface;        ///< VA/DXVA surface handle
+#else
+    shared_surface surface;
+#endif
+    uint32_t plane;               ///< shared surface plane
+};
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/memory_pool.h b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory_pool.hpp
similarity index 52%
rename from inference-engine/thirdparty/clDNN/src/include/memory_pool.h
rename to inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory_pool.hpp
index 739cf23515a9f0..6c3d5d4d7660ef 100644
--- a/inference-engine/thirdparty/clDNN/src/include/memory_pool.h
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/memory_pool.hpp
@@ -2,28 +2,29 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/layout.hpp"
-#include "api/primitive.hpp"
-#include "device_impl.h"
-#include "refcounted_obj.h"
+
+#include "layout.hpp"
+#include "memory_caps.hpp"
 
 #include <vector>
 #include <set>
 #include <map>
 #include <list>
 #include <string>
+#include <atomic>
 
 namespace cldnn {
 
-struct memory_impl;
+struct memory;
 struct shared_mem_params;
-struct engine_impl;
-struct program_impl;
+class engine;
 struct memory_user;
 struct memory_user_comparer;
 using memory_set = std::set<memory_user, memory_user_comparer>;
+using primitive_id = std::string;
+
+using memory_ptr = std::shared_ptr<memory>;
 
 struct memory_user {
     primitive_id _id;
@@ -48,10 +49,10 @@ struct memory_user_comparer {
 
 struct memory_record {
     memory_set _users;  // list of primitives that already use this memory object
-    refcounted_obj_ptr<memory_impl> _memory;
+    memory_ptr _memory;
     uint32_t _network_id;
     allocation_type _type;
-    memory_record(memory_set users, refcounted_obj_ptr<memory_impl>& memory, uint32_t net_id, allocation_type type);
+    memory_record(memory_set users, memory_ptr& memory, uint32_t net_id, allocation_type type);
 };
 
 struct padded_pool_comparer {
@@ -83,55 +84,49 @@ struct padded_pool_comparer {
 // - immutable - if user request for non reusable resource don't use pool, return
 
 // TODO list:
-// - resolve engine <--> memory_pool circular dependency
-// - add padded buffers pool
-// - add decreasing memory limit in gpu_buffer/image dctor
-// - add support for multi networks reuse
+// - Move from runtime to graph part
+// - Improve memory consumption
 
 class memory_pool {
     memory_pool();
 
-    refcounted_obj_ptr<memory_impl> alloc_memory(const layout& layout, allocation_type type, uint32_t network_id, bool reset = true);
+    memory_ptr alloc_memory(const layout& layout, allocation_type type);
     static bool has_conflict(const memory_set&, const std::set<primitive_id>&, uint32_t network_id);
 
     std::multimap<uint64_t, memory_record> _non_padded_pool;
     std::map<layout, std::list<memory_record>, padded_pool_comparer> _padded_pool;
     std::multimap<uint64_t, memory_record> _no_reusable_pool;
-    engine_impl* _engine;
+    engine* _engine;
     std::atomic<uint64_t> _temp_memory_used;
     std::atomic<uint64_t> _max_peak_memory_used;
 
 public:
-    explicit memory_pool(engine_impl& engine);
+    explicit memory_pool(engine& engine);
     ~memory_pool();
-    refcounted_obj_ptr<memory_impl> get_memory(const layout& layout,
-                                               const primitive_id& id,
-                                               uint32_t network_id,
-                                               const std::set<primitive_id>& restrictions,
-                                               allocation_type type,
-                                               bool reusable = true);  // get from pool or create memory allocation
-    refcounted_obj_ptr<memory_impl> get_memory(const layout& layout, allocation_type type, uint32_t network_id, bool reset = true);
-    refcounted_obj_ptr<memory_impl> get_memory(const layout& layout, const shared_mem_params* params, uint32_t network_id);
-    refcounted_obj_ptr<memory_impl> get_from_non_padded_pool(const layout& layout,
-                                                             const primitive_id& id,
-                                                             uint32_t network_id,
-                                                             const std::set<primitive_id>&,
-                                                             allocation_type type);
-    refcounted_obj_ptr<memory_impl> get_from_padded_pool(const layout& layout,
-                                                         const primitive_id& id,
-                                                         uint32_t network_id,
-                                                         const std::set<primitive_id>& restrictions,
-                                                         allocation_type type);
-    refcounted_obj_ptr<memory_impl> get_from_across_networks_pool(const layout& layout,
-                                                                  const primitive_id& id,
-                                                                  uint32_t network_id,
-                                                                  allocation_type type);
+    memory_ptr get_memory(const layout& layout,
+                          const primitive_id& id,
+                          uint32_t network_id,
+                          const std::set<primitive_id>& restrictions,
+                          allocation_type type,
+                          bool reusable = true);  // get from pool or create memory allocation
+    memory_ptr get_memory(const layout& layout, allocation_type type);
+    memory_ptr get_from_non_padded_pool(const layout& layout,
+                                        const primitive_id& id,
+                                        uint32_t network_id,
+                                        const std::set<primitive_id>&,
+                                        allocation_type type);
+    memory_ptr get_from_padded_pool(const layout& layout,
+                                    const primitive_id& id,
+                                    uint32_t network_id,
+                                    const std::set<primitive_id>& restrictions,
+                                    allocation_type type);
+    memory_ptr get_from_across_networks_pool(const layout& layout,
+                                             const primitive_id& id,
+                                             uint32_t network_id,
+                                             allocation_type type);
     void clear_pool();
     void clear_pool_for_network(uint32_t network_id);
-    void release_memory(memory_impl* memory,
-                        const primitive_id& id);
-    void color_graph(const program_impl&);
-    void dump_memory_pool(const program_impl&, std::string&, std::string&);
+    void release_memory(memory* memory, const primitive_id& id, uint32_t network_id);
 
     uint64_t get_temp_memory_used() const { return _temp_memory_used; }
     uint64_t get_max_peak_device_memory_used() const { return _max_peak_memory_used; }
diff --git a/inference-engine/thirdparty/clDNN/api/profiling.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/profiling.hpp
similarity index 99%
rename from inference-engine/thirdparty/clDNN/api/profiling.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/runtime/profiling.hpp
index 575fb06defa69c..98c18a483f9fe4 100644
--- a/inference-engine/thirdparty/clDNN/api/profiling.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/profiling.hpp
@@ -3,6 +3,7 @@
 //
 
 #pragma once
+
 #include <chrono>
 #include <memory>
 #include <vector>
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/stream.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/stream.hpp
new file mode 100644
index 00000000000000..10575a9d6c4c05
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/stream.hpp
@@ -0,0 +1,46 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "event.hpp"
+#include "kernel.hpp"
+#include "kernel_args.hpp"
+
+#include <memory>
+#include <vector>
+
+namespace cldnn {
+
+class stream {
+public:
+    using ptr = std::shared_ptr<stream>;
+    explicit stream(queue_types queue_type) : queue_type(queue_type) {}
+    virtual ~stream() = default;
+
+    virtual void flush() const = 0;
+    virtual void finish() const = 0;
+
+    virtual void set_arguments(kernel& kernel, const kernel_arguments_desc& args_desc, const kernel_arguments_data& args) = 0;
+    virtual event::ptr enqueue_kernel(kernel& kernel,
+                                      const kernel_arguments_desc& args_desc,
+                                      const kernel_arguments_data& args,
+                                      std::vector<event::ptr> const& deps,
+                                      bool is_output_event = false) = 0;
+    virtual event::ptr enqueue_marker(std::vector<event::ptr> const& deps, bool is_output_event = false) = 0;
+    virtual void enqueue_barrier() = 0;
+    virtual event::ptr group_events(std::vector<event::ptr> const& deps) = 0;
+    virtual void wait_for_events(const std::vector<event::ptr>& events) = 0;
+    virtual void reset_events() = 0;
+    virtual event::ptr create_user_event(bool set) = 0;
+    virtual event::ptr create_base_event() = 0;
+    virtual void release_events_pool() = 0;
+
+    queue_types get_queue_type() const { return queue_type; }
+
+protected:
+    queue_types queue_type;
+};
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/tensor.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/tensor.hpp
similarity index 99%
rename from inference-engine/thirdparty/clDNN/api/tensor.hpp
rename to inference-engine/thirdparty/clDNN/api/cldnn/runtime/tensor.hpp
index 4bed1bcdb3a5d6..9c6a02d9e2e044 100644
--- a/inference-engine/thirdparty/clDNN/api/tensor.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/tensor.hpp
@@ -2,11 +2,10 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "cldnn.hpp"
-#include "compounds.h"
-#include "meta_utils.hpp"
+
+#include "compounds.hpp"
+#include "utils.hpp"
 
 #include <map>
 #include <list>
diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/runtime/utils.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/utils.hpp
new file mode 100644
index 00000000000000..7b66ff43a7bc22
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/runtime/utils.hpp
@@ -0,0 +1,144 @@
+// Copyright (C) 2017-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <type_traits>
+#include <stdint.h>
+#include <stddef.h>
+#include <memory>
+
+namespace cldnn {
+
+struct primitive;
+
+namespace meta {
+
+// helper struct to tell wheter type T is any of given types U...
+// termination case when U... is empty -> return std::false_type
+template <class T, class... U>
+struct is_any_of : public std::false_type {};
+
+// helper struct to tell whether type is any of given types (U, Rest...)
+// recurrence case when at least one type U is present -> returns std::true_type if std::same<T, U>::value is true,
+// otherwise call is_any_of<T, Rest...> recurrently
+template <class T, class U, class... Rest>
+struct is_any_of<T, U, Rest...>
+    : public std::conditional<std::is_same<T, U>::value, std::true_type, is_any_of<T, Rest...>>::type {};
+
+template <class T>
+struct always_false : public std::false_type {};
+
+template <typename Ty, Ty Val>
+struct always_false_ty_val : public std::false_type {};
+
+template <typename Ty, Ty... Vals>
+struct val_tuple {};
+
+template <bool... Values>
+struct all : public std::true_type {};
+
+template <bool Val, bool... Values>
+struct all<Val, Values...> : public std::integral_constant<bool, Val && all<Values...>::value> {};
+
+}  // namespace meta
+
+/// @cond CPP_HELPERS
+
+/// @defgroup cpp_helpers Helpers
+/// @{
+
+#define CLDNN_API_CLASS(the_class) static_assert(std::is_standard_layout<the_class>::value, #the_class " has to be 'standard layout' class");
+
+template <typename T>
+typename std::enable_if<std::is_integral<T>::value, T>::type align_to(T size, size_t align) {
+    return static_cast<T>((size % align == 0) ? size : size - size % align + align);
+}
+
+template <typename T>
+typename std::enable_if<std::is_integral<T>::value, T>::type pad_to(T size, size_t align) {
+    return static_cast<T>((size % align == 0) ? 0 : align - size % align);
+}
+
+template <typename T>
+typename std::enable_if<std::is_integral<T>::value, bool>::type is_aligned_to(T size, size_t align) {
+    return !(size % align);
+}
+
+/// Computes ceil(@p val / @p divider) on unsigned integral numbers.
+///
+/// Computes division of unsigned integral numbers and rounds result up to full number (ceiling).
+/// The function works for unsigned integrals only. Signed integrals are converted to corresponding
+/// unsigned ones.
+///
+/// @tparam T1   Type of @p val. Type must be integral (SFINAE).
+/// @tparam T2   Type of @p divider. Type must be integral (SFINAE).
+///
+/// @param val       Divided value. If value is signed, it will be converted to corresponding unsigned type.
+/// @param divider   Divider value. If value is signed, it will be converted to corresponding unsigned type.
+///
+/// @return   Result of ceil(@p val / @p divider). The type of result is determined as if in normal integral
+///           division, except each operand is converted to unsigned type if necessary.
+template <typename T1, typename T2>
+constexpr auto ceil_div(T1 val, T2 divider)
+-> typename std::enable_if<std::is_integral<T1>::value && std::is_integral<T2>::value,
+    decltype(std::declval<typename std::make_unsigned<T1>::type>() / std::declval<typename std::make_unsigned<T2>::type>())>::type {
+    typedef typename std::make_unsigned<T1>::type UT1;
+    typedef typename std::make_unsigned<T2>::type UT2;
+    typedef decltype(std::declval<UT1>() / std::declval<UT2>()) RetT;
+
+    return static_cast<RetT>((static_cast<UT1>(val) + static_cast<UT2>(divider) - 1U) / static_cast<UT2>(divider));
+}
+
+/// Rounds @p val to nearest multiply of @p rounding that is greater or equal to @p val.
+///
+/// The function works for unsigned integrals only. Signed integrals are converted to corresponding
+/// unsigned ones.
+///
+/// @tparam T1       Type of @p val. Type must be integral (SFINAE).
+/// @tparam T2       Type of @p rounding. Type must be integral (SFINAE).
+///
+/// @param val        Value to round up. If value is signed, it will be converted to corresponding unsigned type.
+/// @param rounding   Rounding value. If value is signed, it will be converted to corresponding unsigned type.
+///
+/// @return   @p val rounded up to nearest multiply of @p rounding. The type of result is determined as if in normal integral
+///           division, except each operand is converted to unsigned type if necessary.
+template <typename T1, typename T2>
+constexpr auto round_up_to(T1 val, T2 rounding)
+-> typename std::enable_if<std::is_integral<T1>::value && std::is_integral<T2>::value,
+    decltype(std::declval<typename std::make_unsigned<T1>::type>() / std::declval<typename std::make_unsigned<T2>::type>())>::type {
+    typedef typename std::make_unsigned<T1>::type UT1;
+    typedef typename std::make_unsigned<T2>::type UT2;
+    typedef decltype(std::declval<UT1>() / std::declval<UT2>()) RetT;
+
+    return static_cast<RetT>(ceil_div(val, rounding) * static_cast<UT2>(rounding));
+}
+
+template <typename T, typename... Args>
+std::unique_ptr<T> make_unique(Args&& ... args) {
+    return std::unique_ptr<T>( new T(std::forward<Args>(args)...) );
+}
+
+template <typename derived_type, typename base_type, typename std::enable_if<std::is_base_of<base_type, derived_type>::value, int>::type = 0>
+inline derived_type* downcast(base_type* base) {
+    if (auto casted = dynamic_cast<derived_type*>(base))
+        return casted;
+
+    throw std::runtime_error("Unable to cast pointer from base to derived type");
+}
+
+template <typename derived_type, typename base_type, typename std::enable_if<std::is_base_of<base_type, derived_type>::value, int>::type = 0>
+inline derived_type& downcast(base_type& base) {
+    try {
+        return dynamic_cast<derived_type&>(base);
+    } catch (std::bad_cast& /* ex */) {
+        throw std::runtime_error("Unable to cast reference from base to derived type");
+    }
+    throw std::runtime_error("downcast failed with unhadnled exception");
+}
+
+/// @}
+/// @endcond
+/// @}
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/device.hpp b/inference-engine/thirdparty/clDNN/api/device.hpp
deleted file mode 100644
index e8889d5e86dfd8..00000000000000
--- a/inference-engine/thirdparty/clDNN/api/device.hpp
+++ /dev/null
@@ -1,161 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include "cldnn.hpp"
-#include <string>
-#include <stdexcept>
-#include <vector>
-#include <map>
-
-namespace cldnn {
-/// @addtogroup cpp_api C++ API
-/// @{
-
-/// @defgroup cpp_device GPU Device
-/// @{
-
-/// @brief Enumeration of supported device types
-enum class device_type {
-    integrated_gpu = 0,
-    discrete_gpu = 1
-};
-
-struct gfx_version {
-    uint16_t major;
-    uint8_t minor;
-    uint8_t revision;
-};
-
-/// @brief Information about the device properties and capabilities.
-struct device_info {
-    uint32_t cores_count;     ///< Number of available HW cores.
-    uint32_t core_frequency;  ///< Clock frequency in MHz.
-    uint32_t max_threads_per_execution_unit;    ///< Number of available HW threads on EU.
-    uint32_t max_threads_per_device;            ///< Maximum number of HW threads on device.
-
-    uint64_t max_work_group_size;  ///< Maximum number of work-items in a work-group executing a kernel using the data parallel execution model.
-    uint64_t max_local_mem_size;   ///< Maximum size of local memory arena in bytes.
-    uint64_t max_global_mem_size;  ///< Maximum size of global device memory in bytes.
-    uint64_t max_alloc_mem_size;   ///< Maximum size of memory object allocation in bytes.
-
-    uint64_t max_image2d_width;   ///< Maximum image 2d width supported by the device.
-    uint64_t max_image2d_height;  ///< Maximum image 2d height supported by the device.
-
-    // Flags (for layout compatibility fixed size types are used).
-    uint8_t supports_fp16;             ///< Does engine support FP16.
-    uint8_t supports_fp16_denorms;     ///< Does engine support denormalized FP16.
-    uint8_t supports_subgroups_short;  ///< Does engine support cl_intel_subgroups_short.
-    uint8_t supports_image;            ///< Does engine support images (CL_DEVICE_IMAGE_SUPPORT cap).
-
-    uint8_t supports_imad;   ///< Does engine support int8 mad.
-    uint8_t supports_immad;  ///< Does engine support int8 multi mad.
-
-    uint8_t supports_usm;    ///< Does engine support unified shared memory.
-
-    std::string dev_name;     ///< Device ID string
-    std::string driver_version;  ///< Version of OpenCL driver
-
-    device_type dev_type;  ///< Defines type of current GPU device (integrated or discrete)
-
-    gfx_version gfx_ver;
-    uint32_t device_id;
-    uint32_t num_slices;
-    uint32_t num_sub_slices_per_slice;
-    uint32_t num_eus_per_sub_slice;
-    uint32_t num_threads_per_eu;
-};
-
-struct device_impl;
-
-/// @brief Represents clDNN detected device object. Use device_query to get list of available objects.
-struct device {
-    static device create_default();
-
-    explicit device(device_impl* data)
-        : _impl(data) {
-        if (_impl == nullptr)
-            throw std::invalid_argument("implementation pointer should not be null");
-    }
-
-    /// @brief Returns information about properties and capabilities of the device.
-    device_info get_info() const;
-
-    // TODO add move construction/assignment
-    device(const device& other) : _impl(other._impl) {
-        retain();
-    }
-
-    device& operator=(const device& other) {
-        if (_impl == other._impl) return *this;
-        release();
-        _impl = other._impl;
-        retain();
-        return *this;
-    }
-
-    ~device() {
-        release();
-    }
-
-    friend bool operator==(const device& lhs, const device& rhs) { return lhs._impl == rhs._impl; }
-    friend bool operator!=(const device& lhs, const device& rhs) { return !(lhs == rhs); }
-
-    device_impl* get() const { return _impl; }
-
-private:
-    device_impl* _impl;
-
-    void retain();
-    void release();
-};
-
-struct device_query_impl;
-
-/// @brief Represents clDNN object, which allows to query for list of devices.
-struct device_query {
-    /// @brief Constructs engine configuration with specified options.
-    /// @param Query only for devices, which supports out of order execution (default in cldnn).
-    /// @param Query for devices in user provided opencl context.
-    explicit device_query(void* clcontext = nullptr, void* user_device = nullptr);
-    // TODO add move construction/assignment
-    device_query(const device_query& other) : _impl(other._impl) {
-        retain();
-    }
-
-    /// Returns map of {device_id, device object} of available devices on system.
-    /// Device_id is string. First device will have id: "0", second "1" etc.
-    std::map<std::string, device> get_available_devices() const;
-
-    device_query& operator=(const device_query& other) {
-        if (_impl == other._impl) return *this;
-        release();
-        _impl = other._impl;
-        retain();
-        return *this;
-    }
-
-    ~device_query() {
-        release();
-    }
-
-    friend bool operator==(const device_query& lhs, const device_query& rhs) { return lhs._impl == rhs._impl; }
-    friend bool operator!=(const device_query& lhs, const device_query& rhs) { return !(lhs == rhs); }
-
-    device_query_impl* get() const { return _impl; }
-
-private:
-    device_query_impl* _impl;
-
-    void retain();
-    void release();
-};
-CLDNN_API_CLASS(device_query)
-
-/// @}
-
-/// @}
-
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/engine.hpp b/inference-engine/thirdparty/clDNN/api/engine.hpp
deleted file mode 100644
index c65d89df60196d..00000000000000
--- a/inference-engine/thirdparty/clDNN/api/engine.hpp
+++ /dev/null
@@ -1,189 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include "cldnn.hpp"
-#include "device.hpp"
-#include <string>
-#include <stdexcept>
-#include <vector>
-#include <map>
-#include <algorithm>
-
-namespace cldnn {
-
-/// @addtogroup cpp_api C++ API
-/// @{
-
-/// @defgroup cpp_engine Execution Engine
-/// @{
-
-/// @brief Defines available engine types
-enum class engine_types : int32_t {
-    ocl
-};
-
-/// @brief Defines available priority mode types
-enum class priority_mode_types : int16_t {
-    disabled,
-    low,
-    med,
-    high
-};
-
-/// @brief Defines available priority mode types
-enum class throttle_mode_types : int16_t {
-    disabled,
-    low,
-    med,
-    high
-};
-
-/// @brief Configuration parameters for created engine.
-struct engine_configuration {
-    const bool enable_profiling;              ///< Enable per-primitive profiling.
-    const bool meaningful_kernels_names;      ///< Generate meaniful names fo OpenCL kernels.
-    const bool dump_custom_program;           ///< Dump the user OpenCL programs to files
-    const std::string compiler_options;       ///< OpenCL compiler options string.
-    const std::string single_kernel_name;     ///< If provided, runs specific layer.
-    const bool enable_parallelisation;        ///< Enables parallel execution of primitives which don't depend on each other. Disabled by default.
-    const std::string engine_log;             ///< Specifies a file to which engine log should be dumped. Empty by default (means no logging).
-    const std::string sources_dumps_dir;      ///< Specifies a directory where sources of cldnn::program objects should be dumped.
-                                              ///< Empty by default (means no dumping).
-    const priority_mode_types priority_mode;  ///< Priority mode (support of priority hints in command queue). If cl_khr_priority_hints extension
-                                              ///< is not supported by current OpenCL implementation, the value must be set to cldnn_priority_disabled.
-
-    const throttle_mode_types throttle_mode;  ///< Throttle mode (support of throttle hints in command queue). If cl_khr_throttle_hints extension
-                                              ///< is not supported by current OpenCL implementation, the value must be set to cldnn_throttle_disabled.
-
-    bool enable_memory_pool;                  ///< Enables memory usage optimization. memory objects will be reused when possible
-                                              ///< (switched off for older drivers then NEO).
-    uint16_t n_streams;                       ///< Number of queues executed in parallel
-    const std::string kernels_cache_path;     ///< Path to compiled kernels cache
-    uint16_t n_threads;                       ///< Number of threads
-    const std::string tuning_cache_path;      ///< Path to tuning kernel cache
-
-    /// @brief Constructs engine configuration with specified options.
-    /// @param profiling Enable per-primitive profiling.
-    /// @param decorate_kernel_names Generate meaniful names fo OpenCL kernels.
-    /// @param dump_custom_program Dump the custom OpenCL programs to files
-    /// @param options OpenCL compiler options string.
-    /// @param single_kernel If provided, runs specific layer.
-    engine_configuration(
-        bool profiling = false,
-        bool decorate_kernel_names = false,
-        bool dump_custom_program = false,
-        const std::string& options = std::string(),
-        const std::string& single_kernel = std::string(),
-        bool primitives_parallelisation = true,
-        const std::string& engine_log = std::string(),
-        const std::string& sources_dumps_dir = std::string(),
-        priority_mode_types priority_mode = priority_mode_types::disabled,
-        throttle_mode_types throttle_mode = throttle_mode_types::disabled,
-        bool memory_pool = true,
-        uint16_t n_streams = 1,
-        const std::string& kernels_cache_path = "",
-        uint16_t n_threads = std::max(static_cast<uint16_t>(std::thread::hardware_concurrency()), static_cast<uint16_t>(1)),
-        const std::string& tuning_cache_path = "cache.json")
-        : enable_profiling(profiling)
-        , meaningful_kernels_names(decorate_kernel_names)
-        , dump_custom_program(dump_custom_program)
-        , compiler_options(options)
-        , single_kernel_name(single_kernel)
-        , enable_parallelisation(primitives_parallelisation)
-        , engine_log(engine_log)
-        , sources_dumps_dir(sources_dumps_dir)
-        , priority_mode(priority_mode)
-        , throttle_mode(throttle_mode)
-        , enable_memory_pool(memory_pool)
-        , n_streams(n_streams)
-        , kernels_cache_path(kernels_cache_path)
-        , n_threads(n_threads)
-        , tuning_cache_path(tuning_cache_path) {
-        if (n_streams == 0) {
-            throw std::invalid_argument("Invalid streams count set in engine config");
-        }
-    }
-};
-
-struct engine_impl;
-
-/// @brief Represents clDNN engine object.
-struct engine {
-    /// @brief Constructs @p OpenCL engine
-    explicit engine(const engine_configuration& configuration = engine_configuration())
-        : engine(engine_types::ocl, device::create_default(), configuration) {}
-
-    /// @brief Constructs @p OpenCL engine
-    explicit engine(const device& device, const engine_configuration& configuration = engine_configuration())
-        : engine(engine_types::ocl, device, configuration) {}
-
-    /// @brief Construct engine of the specified @p type, @p engine_num, and @p configuration options.
-    /// @param[in] type Engine type @ref cldnn_engine_type. Only OCL engine is supported.
-    /// @param[in] engine_num Engine index. Should be 0.
-    /// @param[in] configuration Engine configuration options.
-    engine(engine_types type, const device& device, const engine_configuration& configuration = engine_configuration());
-
-    // TODO add move construction/assignment
-    engine(const engine& other) : _impl(other._impl) {
-        retain();
-    }
-
-    engine& operator=(const engine& other) {
-        if (_impl == other._impl) return *this;
-        release();
-        _impl = other._impl;
-        retain();
-        return *this;
-    }
-
-    ~engine() {
-        release();
-    }
-
-    friend bool operator==(const engine& lhs, const engine& rhs) { return lhs._impl == rhs._impl; }
-    friend bool operator!=(const engine& lhs, const engine& rhs) { return !(lhs == rhs); }
-
-    /// @brief Returns number of available engines of the particular @p type.
-    static uint32_t engine_count(engine_types type);
-
-    /// @brief Release pending memory allocated in OpenCL context.
-    void release_pending_memory(uint32_t net_id) const;
-
-    /// @brief Returns information about properties and capabilities of the device used for allocation of the engine.
-    device_info get_info() const;
-
-    /// @brief Returns OpenCL context handle of the engine.
-    void* get_context() const;
-
-    /// @brief Returns total size of all resources allocated using given engine
-    uint64_t get_max_used_device_memory_size() const;
-
-    /// @brief Returns total size of currently resources allocated using given engine
-    uint64_t get_temp_used_device_memory_size() const;
-
-    /// @brief Returns type of the engine.
-    engine_types get_type() const;
-
-    /// @brief get C API engine handler.
-    engine_impl* get() const { return _impl; }
-
-private:
-    friend struct network;
-    friend struct memory;
-    friend struct event;
-
-    engine_impl* _impl;
-
-    void retain();
-    void release();
-};
-CLDNN_API_CLASS(engine)
-
-/// @}
-
-/// @}
-
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/event.hpp b/inference-engine/thirdparty/clDNN/api/event.hpp
deleted file mode 100644
index 3b54171d74549d..00000000000000
--- a/inference-engine/thirdparty/clDNN/api/event.hpp
+++ /dev/null
@@ -1,83 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include "cldnn.hpp"
-#include "engine.hpp"
-#include "profiling.hpp"
-#include <algorithm>
-#include <cassert>
-#include <vector>
-#include <memory>
-#include <functional>
-#include <stdexcept>
-
-namespace cldnn {
-
-/// @addtogroup cpp_api C++ API
-/// @{
-
-/// @addtogroup cpp_event Events Support
-/// @{
-
-struct event_impl;
-
-/// @brief user-defined event handler callback.
-using event_handler = std::function<void(void*)>;
-
-/// @brief Represents an clDNN Event object
-struct event {
-    /// @brief Create an event which can be set to 'completed' by user.
-    static event create_user_event(const engine& engine, uint32_t net_id);
-
-    /// @brief Construct from C API handler @ref ::cldnn_event.
-    explicit event(event_impl* impl) : _impl(impl) {
-        if (_impl == nullptr) throw std::invalid_argument("implementation pointer should not be null");
-    }
-
-    event(const event& other) : _impl(other._impl) {
-        retain();
-    }
-
-    event& operator=(const event& other) {
-        if (_impl == other._impl) return *this;
-        release();
-        _impl = other._impl;
-        retain();
-        return *this;
-    }
-
-    ~event() {
-        release();
-    }
-
-    friend bool operator==(const event& lhs, const event& rhs) { return lhs._impl == rhs._impl; }
-    friend bool operator!=(const event& lhs, const event& rhs) { return !(lhs == rhs); }
-
-    /// @brief Wait for event completion.
-    void wait() const;
-
-    /// @brief Set event status to 'completed'.
-    void set() const;
-
-    /// @brief Register call back to be called on event completion.
-    void set_event_handler(event_handler handler, void* param) const;
-
-    /// @brief Get profiling info for the event associated with network output.
-    std::vector<instrumentation::profiling_interval> get_profiling_info() const;
-
-    /// @brief Returns C API event handler.
-    event_impl* get() const { return _impl; }
-
-private:
-    event_impl* _impl;
-    void retain();
-    void release();
-};
-CLDNN_API_CLASS(event)
-
-/// @}
-/// @}
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/memory.hpp b/inference-engine/thirdparty/clDNN/api/memory.hpp
deleted file mode 100644
index 30dfd5c7eef0e9..00000000000000
--- a/inference-engine/thirdparty/clDNN/api/memory.hpp
+++ /dev/null
@@ -1,278 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include <cstdint>
-#include "cldnn.hpp"
-#include "compounds.h"
-#include "layout.hpp"
-#include "engine.hpp"
-#include <memory>
-#include <iterator>
-#include <string>
-
-namespace cldnn {
-
-/// @addtogroup cpp_api C++ API
-/// @{
-
-/// @defgroup cpp_memory Memory description and management
-/// @{
-
-template <typename T>
-struct pointer;
-
-struct memory_impl;
-
-/// @brief Shared memory descriptor type.
-enum class shared_mem_type {
-    /// @brief Structure unitialized or contains no information.
-    shared_mem_empty,
-
-    /// @brief Structure describes shared CL buffer.
-    shared_mem_buffer,
-
-    /// @brief Structure describes shared CL image.
-    shared_mem_image,
-
-    /// @brief Structure describes shared VA/DXVA surface
-    shared_mem_vasurface,
-
-    /// @brief Structure describes shared D3D11 buffer
-    shared_mem_dxbuffer
-};
-
-using shared_handle = void*;
-using shared_surface = uint32_t;
-
-/// @brief Low-level API handles required for using cldnn memory objects in external API calls.
-struct shared_mem_params {
-    shared_mem_type mem_type;     ///< shared buffer type
-    shared_handle context;        ///< OpenCL context for external operations
-    shared_handle user_device;    ///< DX/VA device for external operations
-    shared_handle mem;            ///< memory object handle
-#ifdef _WIN32
-    shared_handle surface;        ///< VA/DXVA surface handle
-#else
-    shared_surface surface;
-#endif
-    uint32_t plane;               ///< shared surface plane
-};
-/// @brief Represents buffer with particular @ref layout.
-/// @details Usually allocated by @ref engine except cases when attached to user-allocated buffer.
-struct memory {
-    friend struct data;
-    friend struct mutable_data;
-    friend struct network;
-    friend struct network_output;
-
-    /// Allocate memory on @p engine using specified @p layout
-    static memory allocate(const engine& engine, const layout& layout, uint32_t net_id = 0, bool reset = true);
-
-    /// Create shared memory object on @p engine using user-supplied memory buffer @p buf using specified @p layout
-    static memory share_buffer(const engine& engine, const layout& layout, shared_handle buf, uint32_t net_id = 0);
-
-    /// Create shared memory object on @p engine using user-supplied 2D image @p img using specified @p layout
-    static memory share_image(const engine& engine, const layout& layout, shared_handle img, uint32_t net_id = 0);
-
-    /// Create shared memory object on @p engine over specified @p plane of video decoder surface @p surf using specified @p layout
-#ifdef _WIN32
-    static memory share_surface(const engine& engine, const layout& layout, shared_handle surf, uint32_t plane,
-        uint32_t net_id = 0);
-    static memory share_dx_buffer(const engine& engine, const layout& layout, shared_handle res, uint32_t net_id = 0);
-#else
-    static memory share_surface(const engine& engine, const layout& layout, shared_surface surf, uint32_t plane,
-        uint32_t net_id = 0);
-#endif
-
-    /// Create memory object attached to the buffer allocated by user.
-    /// @param ptr  The pointer to user allocated buffer.
-    /// @param size Size (in bytes) of the buffer. Should be equal to @p layout.data_size()
-    /// @note User is responsible for buffer deallocation. Buffer lifetime should be bigger than lifetime of the memory object.
-    template <typename T>
-    static memory attach(const cldnn::layout& layout, T* ptr, size_t size, uint32_t net_id = 0) {
-        if (!ptr)
-            throw std::invalid_argument("pointer should not be null");
-        size_t data_size = size * sizeof(T);
-        if (data_size != layout.bytes_count()) {
-            std::string err_str("buffer size mismatch - input size " + std::to_string(data_size) + " layout size " +
-                                std::to_string(layout.bytes_count()));
-            throw std::invalid_argument(err_str);
-        }
-
-        return attach_impl(layout, static_cast<void*>(ptr), net_id);
-    }
-
-    explicit memory(memory_impl* data)
-        : _impl(data) {
-        if (_impl == nullptr)
-            throw std::invalid_argument("implementation pointer should not be null");
-    }
-
-    memory(const memory& other) : _impl(other._impl) {
-        retain();
-    }
-
-    memory& operator=(const memory& other) {
-        if (_impl == other._impl)
-            return *this;
-        release();
-        _impl = other._impl;
-        retain();
-        return *this;
-    }
-
-    ~memory() { release(); }
-
-    friend bool operator==(const memory& lhs, const memory& rhs) { return lhs._impl == rhs._impl; }
-    friend bool operator!=(const memory& lhs, const memory& rhs) { return !(lhs == rhs); }
-
-    /// number of elements of _layout.data_type stored in memory
-    size_t count() const;
-
-    /// number of bytes used by memory
-    size_t size() const;
-
-    /// Associated @ref layout
-    const layout& get_layout() const;
-    int get_net_id() const;
-
-    /// Test if memory is allocated by @p engine
-    bool is_allocated_by(const engine& engine) const;
-
-    bool is_the_same_buffer(const memory& other) const;
-
-    shared_mem_params get_internal_params() const;
-
-    /// Creates the @ref pointer object to get an access memory data
-    template <typename T>
-    friend struct cldnn::pointer;
-    template <typename T>
-    cldnn::pointer<T> pointer() const;
-
-    /// C API memory handle
-    memory_impl* get() const { return _impl; }
-
-    void reset();
-
-private:
-    friend struct engine;
-    memory_impl* _impl;
-
-    template <typename T>
-    T* lock() const {
-        if (data_type_traits::align_of(get_layout().data_type) % alignof(T) != 0) {
-            throw std::logic_error("memory data type alignment do not match");
-        }
-        return static_cast<T*>(lock_impl());
-    }
-
-    void unlock() const;
-
-    void* lock_impl() const;
-    static memory attach_impl(const cldnn::layout& layout, void* ptr, uint32_t net_id);
-
-    void retain();
-    void release();
-};
-
-/// @brief Helper class to get an access @ref memory data
-/// @details
-/// This class provides an access to @ref memory data following RAII idiom and exposes basic C++ collection members.
-/// @ref memory object is locked on construction of pointer and "unlocked" on descruction.
-/// Objects of this class could be used in many STL utility functions like copy(), transform(), etc.
-/// As well as in range-for loops.
-template <typename T>
-struct pointer {
-    /// @brief Constructs pointer from @ref memory and locks @c (pin) ref@ memory object.
-    explicit pointer(const memory& mem) : _mem(mem), _size(_mem.size() / sizeof(T)), _ptr(_mem.lock<T>()) {}
-
-    /// @brief Unlocks @ref memory
-    ~pointer() { _mem.unlock(); }
-
-    /// @brief Copy construction.
-    pointer(const pointer& other) : pointer(other._mem) {}
-
-    /// @brief Copy assignment.
-    pointer& operator=(const pointer& other) {
-        if (this->_mem != other._mem)
-            do_copy(other._mem);
-        return *this;
-    }
-
-    /// @brief Returns the number of elements (of type T) stored in memory
-    size_t size() const { return _size; }
-
-#if defined(_SECURE_SCL) && (_SECURE_SCL > 0)
-    typedef stdext::checked_array_iterator<T*> iterator;
-    typedef stdext::checked_array_iterator<const T*> const_iterator;
-
-    iterator begin() & { return stdext::make_checked_array_iterator(_ptr, size()); }
-    iterator end() & { return stdext::make_checked_array_iterator(_ptr, size(), size()); }
-
-    const_iterator begin() const& { return stdext::make_checked_array_iterator(_ptr, size()); }
-    const_iterator end() const& { return stdext::make_checked_array_iterator(_ptr, size(), size()); }
-#else
-    typedef T* iterator;
-    typedef const T* const_iterator;
-    iterator begin() & { return _ptr; }
-    iterator end() & { return _ptr + size(); }
-    const_iterator begin() const& { return _ptr; }
-    const_iterator end() const& { return _ptr + size(); }
-#endif
-
-    /// @brief Provides indexed access to pointed memory.
-    T& operator[](size_t idx) const& {
-        assert(idx < _size);
-        return _ptr[idx];
-    }
-
-    /// @brief Returns the raw pointer to pointed memory.
-    T* data() & { return _ptr; }
-    /// @brief Returns the constant raw pointer to pointed memory
-    const T* data() const& { return _ptr; }
-
-    friend bool operator==(const pointer& lhs, const pointer& rhs) { return lhs._mem == rhs._mem; }
-    friend bool operator!=(const pointer& lhs, const pointer& rhs) { return !(lhs == rhs); }
-
-    // do not use this class as temporary object
-    // ReSharper disable CppMemberFunctionMayBeStatic, CppMemberFunctionMayBeConst
-    /// Prevents to use pointer as temporary object
-    void data() && {}
-    /// Prevents to use pointer as temporary object
-    void begin() && {}
-    /// Prevents to use pointer as temporary object
-    void end() && {}
-    /// Prevents to use pointer as temporary object
-    void operator[](size_t idx) && {}
-    // ReSharper restore CppMemberFunctionMayBeConst, CppMemberFunctionMayBeStatic
-
-private:
-    memory _mem;
-    size_t _size;
-    T* _ptr;
-
-    // TODO implement exception safe code.
-    void do_copy(const memory& mem) {
-        auto ptr = mem.lock<T>();
-        _mem.unlock();
-        _mem = mem;
-        _size = _mem.size() / sizeof(T);
-        _ptr = ptr;
-    }
-};
-
-#ifndef DOXYGEN_SHOULD_SKIP_THIS
-template <typename T>
-pointer<T> memory::pointer() const {
-    return cldnn::pointer<T>(*this);
-}
-#endif
-
-/// @}
-
-/// @}
-
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api/meta_utils.hpp b/inference-engine/thirdparty/clDNN/api/meta_utils.hpp
deleted file mode 100644
index c71073a8dfd0fa..00000000000000
--- a/inference-engine/thirdparty/clDNN/api/meta_utils.hpp
+++ /dev/null
@@ -1,43 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-
-#include <type_traits>
-
-namespace cldnn {
-
-struct primitive;
-
-namespace meta {
-
-// helper struct to tell wheter type T is any of given types U...
-// termination case when U... is empty -> return std::false_type
-template <class T, class... U>
-struct is_any_of : public std::false_type {};
-
-// helper struct to tell whether type is any of given types (U, Rest...)
-// recurrence case when at least one type U is present -> returns std::true_type if std::same<T, U>::value is true,
-// otherwise call is_any_of<T, Rest...> recurrently
-template <class T, class U, class... Rest>
-struct is_any_of<T, U, Rest...>
-    : public std::conditional<std::is_same<T, U>::value, std::true_type, is_any_of<T, Rest...>>::type {};
-
-template <class T>
-struct always_false : public std::false_type {};
-
-template <typename Ty, Ty Val>
-struct always_false_ty_val : public std::false_type {};
-
-template <typename Ty, Ty... Vals>
-struct val_tuple {};
-
-template <bool... Values>
-struct all : public std::true_type {};
-
-template <bool Val, bool... Values>
-struct all<Val, Values...> : public std::integral_constant<bool, Val && all<Values...>::value> {};
-
-}  // namespace meta
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/api_test_builds/CMakeLists.txt b/inference-engine/thirdparty/clDNN/api_test_builds/CMakeLists.txt
index 38772c1e8941c2..9e08c013ca1dce 100644
--- a/inference-engine/thirdparty/clDNN/api_test_builds/CMakeLists.txt
+++ b/inference-engine/thirdparty/clDNN/api_test_builds/CMakeLists.txt
@@ -30,28 +30,20 @@ intel_config_flag_apply_settings(CompilerOptions CMAKE_C_FLAGS ALL_PATTERN ""
 # ========================================= Source/Header files ========================================
 
 set(__CLDNN_Label__api                  "api")
-file(GLOB __CLDNN_Headers__api
+file(GLOB_RECURSE __CLDNN_Headers__api
     "${CLDNN__API_DIR}/*.hpp"
   )
 
-set(__CLDNN_Directory__api__cpp         "${CLDNN__API_DIR}/CPP")
-set(__CLDNN_Label__api__cpp             "${__CLDNN_Label__api}\\CPP")
-file(GLOB __CLDNN_Headers__api__cpp
-    "${__CLDNN_Directory__api__cpp}/*.hpp"
-  )
-
 set(__CLDNN_Directory__test_builds      "${CLDNN__CODEGEN_DIR}/test_builds")
 set(__CLDNN_CGDirectory__test_builds    "${CLDNN__CODEGEN_BASEDIR}/test_builds")
 set(__CLDNN_Label__test_builds          "test builds\\codegen")
-set(__CLDNN_File__test_builds__api__cpp "api_cpp_test.cpp")
+set(__CLDNN_File__test_builds__api      "api_cpp_test.cpp")
 set(__CLDNN_Sources__test_builds
-    "${__CLDNN_Directory__test_builds}/${__CLDNN_File__test_builds__api__cpp}"
+    "${__CLDNN_Directory__test_builds}/${__CLDNN_File__test_builds__api}"
   )
 
 set(__CLDNN_AllSources
     ${__CLDNN_Headers__api}
-    ${__CLDNN_Headers__api__cpp}
-    ${__CLDNN_Headers__api__c}
     ${__CLDNN_Sources__test_builds}
   )
 
@@ -63,9 +55,9 @@ set_property(SOURCE ${__CLDNN_Sources__test_builds} PROPERTY GENERATED TRUE)
 file(MAKE_DIRECTORY "${__CLDNN_CGDirectory__test_builds}")
 
 # C++ API testing (C++11 compatibility).
-set(__CLDNN_CGFile__api__cpp "${__CLDNN_CGDirectory__test_builds}/${__CLDNN_File__test_builds__api__cpp}")
+set(__CLDNN_CGFile__api__cpp "${__CLDNN_CGDirectory__test_builds}/${__CLDNN_File__test_builds__api}")
 file(WRITE "${__CLDNN_CGFile__api__cpp}" "// This file is auto-generated. Please, do not modify it directly.\n\n")
-foreach(__CLDNN_Header ${__CLDNN_Headers__api} ${__CLDNN_Headers__api__cpp})
+foreach(__CLDNN_Header ${__CLDNN_Headers__api})
   string(REPLACE ";" "\;" __CLDNN_Header "${__CLDNN_Header}") # [WA#1] Must escape ; again if occurred in item.
   file(APPEND "${__CLDNN_CGFile__api__cpp}" "#include \"${__CLDNN_Header}\"\n")
 endforeach()
@@ -74,21 +66,17 @@ file(APPEND "${__CLDNN_CGFile__api__cpp}" "\n")
 # =============================================== Filters ==============================================
 
 source_group("${__CLDNN_Label__api}"         FILES ${__CLDNN_Headers__api})
-source_group("${__CLDNN_Label__api__cpp}"    FILES ${__CLDNN_Headers__api__cpp})
 source_group("${__CLDNN_Label__test_builds}" FILES ${__CLDNN_Sources__test_builds})
 
 # ===================================== Include/Link directories =======================================
 
-include_directories(
-    "${CLDNN__MAIN_DIR}"
-  )
+include_directories("${CLDNN__MAIN_DIR}")
 
 # =================================== Link targets and dependencies ====================================
 
 # Test static library.
-add_library("${CLDNN_BUILD__PROJ}" STATIC
-    ${__CLDNN_AllSources}
-  )
+add_library("${CLDNN_BUILD__PROJ}" STATIC ${__CLDNN_AllSources})
+
 set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY PROJECT_LABEL            "${CLDNN_BUILD__PROJ_LABEL}")
 set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY ARCHIVE_OUTPUT_DIRECTORY "${CLDNN_BUILD__PROJ_OUTPUT_DIR}")
 set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY OUTPUT_NAME              "${CLDNN_BUILD__PROJ_OUTPUT_NAME}")
@@ -96,16 +84,14 @@ set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY OUTPUT_NAME              "${
 target_link_libraries("${CLDNN_BUILD__PROJ}" ${CLDNN__SYSTEM_LINK_LIBRARIES})
 
 
-add_dependencies("${CLDNN_BUILD__PROJ__clDNN}"
-    "${CLDNN_BUILD__PROJ}"
-  )
+add_dependencies("${CLDNN_BUILD__PROJ__clDNN}" "${CLDNN_BUILD__PROJ}")
 
 # =================================== Custom pre- and post-steps =======================================
 
-add_custom_command(OUTPUT "${__CLDNN_Directory__test_builds}/${__CLDNN_File__test_builds__api__cpp}"
-    COMMAND "${CMAKE_COMMAND}" -E copy_if_different "${__CLDNN_CGFile__api__cpp}" "${__CLDNN_Directory__test_builds}/${__CLDNN_File__test_builds__api__cpp}"
+add_custom_command(OUTPUT "${__CLDNN_Directory__test_builds}/${__CLDNN_File__test_builds__api}"
+    COMMAND "${CMAKE_COMMAND}" -E copy_if_different "${__CLDNN_CGFile__api__cpp}" "${__CLDNN_Directory__test_builds}/${__CLDNN_File__test_builds__api}"
     DEPENDS "${__CLDNN_CGFile__api__cpp}" ${__CLDNN_Headers__api} ${__CLDNN_Headers__api__cpp}
-    COMMENT "Updating file if the file changed (${__CLDNN_File__test_builds__api__cpp}) ..."
+    COMMENT "Updating file if the file changed (${__CLDNN_File__test_builds__api}) ..."
   )
 
 # ======================================================================================================
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/CMakeLists.txt b/inference-engine/thirdparty/clDNN/kernel_selector/CMakeLists.txt
index da5dd9b2333d8a..2f8271441bd28d 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/CMakeLists.txt
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/CMakeLists.txt
@@ -124,6 +124,7 @@ source_group("${__CLDNN_Label__cg_cache}"         FILES ${__CLDNN_Sources__cg_ca
 # ===================================== Include/Link directories =======================================
 
 include_directories(
+    "${CLDNN__MAIN_DIR}"
     "${__CLDNN_Directory__main}"
     "${__CLDNN_Directory__core}"
     "${__CLDNN_Directory__core}/common"
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/common/tensor_type.h b/inference-engine/thirdparty/clDNN/kernel_selector/common/tensor_type.h
index 19dbc972aba8b6..f9c9b622e48e69 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/common/tensor_type.h
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/common/tensor_type.h
@@ -385,6 +385,8 @@ struct TensorBase {
 
         return differ;
     }
+
+    virtual ~TensorBase() = default;
 };
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/activation/activation_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/activation/activation_kernel_base.cpp
index 0b31d6ce496cb3..2a24a9f18dfb53 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/activation/activation_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/activation/activation_kernel_base.cpp
@@ -81,7 +81,7 @@ KernelsData ActivationKernelBase::GetCommonKernelsData(const Params& params, con
                      DEFAULT, false, false, 1, GetFusedPrimitiveInputsCount(params));
 
     if (!newParams.inputActivationParams.empty()) {
-        kernel.arguments.push_back({ArgumentDescriptor::Types::SLOPE, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::SLOPE, 0});
     }
 
     return {kd};
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/arg_max_min/arg_max_min_kernel_axis.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/arg_max_min/arg_max_min_kernel_axis.cpp
index ab4e2068371679..734dd1c9ebb8cb 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/arg_max_min/arg_max_min_kernel_axis.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/arg_max_min/arg_max_min_kernel_axis.cpp
@@ -93,7 +93,7 @@ KernelsData ArgMaxMinKernelAxis::GetKernelsData(const Params& params, const opti
     FillCLKernelData(kernel, dispatchData, params.engineInfo, kernelName, jit, entry_point);
 
     if (orgParams.outputs_num == 2) {
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
     }
 
     return {kd};
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/binary_convolution/binary_convolution_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/binary_convolution/binary_convolution_kernel_base.cpp
index 6747a78d418955..d7a7ba3ec37800 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/binary_convolution/binary_convolution_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/binary_convolution/binary_convolution_kernel_base.cpp
@@ -137,7 +137,7 @@ KernelsData BinaryConvolutionKernelBase::GetCommonKernelsData(const Params& para
     uint32_t fused_deps_total = 0;
     for (auto& fused_dep : newParams.fused_ops) {
         for (int i = 0; i < static_cast<int>(fused_dep.dep_size); i++) {
-            kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT_OF_FUSED_PRIMITIVE, fused_deps_total});
+            kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT_OF_FUSED_PRIMITIVE, fused_deps_total});
             fused_deps_total++;
         }
     }
@@ -153,7 +153,7 @@ KernelsData BinaryConvolutionKernelBase::GetCommonKernelsData(const Params& para
                      !newParams.bias.empty(),
                      1,
                      fused_deps_total);
-    kernel.arguments.push_back({ArgumentDescriptor::Types::SPLIT, 0});
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::SPLIT, 0});
 
     kd.autoTuneIndex = autoTuneIndex;
 
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_base.cpp
index 649afada9992b3..f1c20f88767c33 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_base.cpp
@@ -111,17 +111,17 @@ KernelsData ConcatenationKernelBase::GetCommonKernelsData(const Params& params,
         auto entryPoint = GetEntryPoint(kernelName, newParams.layerID, options);
         auto jit = CreateJit(kernelName, cldnnJit, entryPoint);
 
-        kernel.workGroups.global = dispatchData.gws;
-        kernel.workGroups.local = dispatchData.lws;
-        kernel.kernelString = GetKernelString(kernelName, jit, entryPoint, params.engineInfo);
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, (uint32_t)i });
-        kernel.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
+        kernel.code.kernelString = GetKernelString(kernelName, jit, entryPoint, params.engineInfo);
+        kernel.params.workGroups.global = dispatchData.gws;
+        kernel.params.workGroups.local = dispatchData.lws;
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, (uint32_t)i });
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
 
         ScalarDescriptor s;
         s.t = ScalarDescriptor::Types::UINT32;
         s.v.u32 = lastOffset;
-        kernel.scalars.push_back(s);
-        kernel.arguments.push_back({ArgumentDescriptor::Types::SCALAR, 0});
+        kernel.params.scalars.push_back(s);
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::SCALAR, 0});
 
         lastOffset += (uint32_t)input.GetDims()[concatChannelIndex].v;
     }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_fs_b_yx_fsv32.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_fs_b_yx_fsv32.cpp
index 07912f47634e5e..26397613b41245 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_fs_b_yx_fsv32.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_fs_b_yx_fsv32.cpp
@@ -108,17 +108,17 @@ KernelsData ConcatenationKernel_fs_b_yx_fsv32::GetKernelsData(const Params& para
         auto entryPoint = GetEntryPoint(kernelName, newParams.layerID, optParams);
         auto jit = CreateJit(kernelName, cldnnJit, entryPoint);
 
-        kernel.workGroups.global = dispatchData.gws;
-        kernel.workGroups.local = dispatchData.lws;
-        kernel.kernelString = GetKernelString(kernelName, jit, entryPoint, params.engineInfo);
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, (uint32_t)i});
-        kernel.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
+        kernel.code.kernelString = GetKernelString(kernelName, jit, entryPoint, params.engineInfo);
+        kernel.params.workGroups.global = dispatchData.gws;
+        kernel.params.workGroups.local = dispatchData.lws;
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, (uint32_t)i});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
 
         ScalarDescriptor s;
         s.t = ScalarDescriptor::Types::UINT32;
         s.v.u32 = lastOffset;
-        kernel.scalars.push_back(s);
-        kernel.arguments.push_back({ArgumentDescriptor::Types::SCALAR, 0});
+        kernel.params.scalars.push_back(s);
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::SCALAR, 0});
 
         lastOffset += (uint32_t)input.GetDims()[concatChannelIndex].v;
     }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_ref.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_ref.cpp
index d9f630b9a57520..618043fb52e609 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_ref.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/concatenation/concatenation_kernel_ref.cpp
@@ -100,9 +100,9 @@ KernelsData ConcatenationKernelRef::GetKernelsData(const Params& params, const o
             auto& kernel = kd[0].kernels[i];
 
             // to avoid cases when we execute with local work sizes 1x1x1
-            if (kernel.workGroups.local[0] == 1 && kernel.workGroups.global[1] != 1) {
-                kernel.workGroups.global[1] = Align(kernel.workGroups.global[1], 32);
-                kernel.workGroups.local[1] = 32;
+            if (kernel.params.workGroups.local[0] == 1 && kernel.params.workGroups.global[1] != 1) {
+                kernel.params.workGroups.global[1] = Align(kernel.params.workGroups.global[1], 32);
+                kernel.params.workGroups.local[1] = 32;
             }
         }
     }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/convolution/convolution_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/convolution/convolution_kernel_base.cpp
index bcc7bbd29e7c9b..a5674f56f55167 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/convolution/convolution_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/convolution/convolution_kernel_base.cpp
@@ -230,24 +230,24 @@ KernelsData ConvolutionKernelBase::GetCommonKernelsData(const Params& params,
                      1);
 
     if (newParams.deformable_mode) {
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
     }
 
     if (!newParams.weights_zero_points.empty())
-        kernel.arguments.push_back({ArgumentDescriptor::Types::WEIGHTS_ZERO_POINTS, 1});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::WEIGHTS_ZERO_POINTS, 1});
     if (!newParams.activations_zero_points.empty())
-        kernel.arguments.push_back({ArgumentDescriptor::Types::ACTIVATIONS_ZERO_POINTS, 1});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::ACTIVATIONS_ZERO_POINTS, 1});
     if (!newParams.compensation.empty())
-        kernel.arguments.push_back({ArgumentDescriptor::Types::COMPENSATION, 1});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::COMPENSATION, 1});
 
     uint32_t fused_deps_total = 0;
     for (auto& fused_dep : newParams.fused_ops) {
         for (int i = 0; i < static_cast<int>(fused_dep.dep_size); i++) {
-            kernel.arguments.push_back({ ArgumentDescriptor::Types::INPUT_OF_FUSED_PRIMITIVE, fused_deps_total });
+            kernel.params.arguments.push_back({ ArgumentDescriptor::Types::INPUT_OF_FUSED_PRIMITIVE, fused_deps_total });
             fused_deps_total++;
         }
     }
-    kernel.arguments.push_back({ArgumentDescriptor::Types::SPLIT, 0});
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::SPLIT, 0});
 
     kd.autoTuneIndex = autoTuneIndex;
 
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/ctc_greedy_decoder/ctc_greedy_decoder_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/ctc_greedy_decoder/ctc_greedy_decoder_kernel_base.cpp
index db690ef39a4131..664ed70fabc64b 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/ctc_greedy_decoder/ctc_greedy_decoder_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/ctc_greedy_decoder/ctc_greedy_decoder_kernel_base.cpp
@@ -73,7 +73,7 @@ KernelsData CTCGreedyDecoderKernelBase::GetCommonKernelsData(const Params& param
                      GetFusedPrimitiveInputsCount(params));
 
     if (orgParams.outputs_num == 2) {
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 2});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 2});
     }
 
     return {kd};
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/cum_sum/cum_sum_kernel_partial_sum.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/cum_sum/cum_sum_kernel_partial_sum.cpp
index 085ce54e7a0944..87d2a3cedccc5c 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/cum_sum/cum_sum_kernel_partial_sum.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/cum_sum/cum_sum_kernel_partial_sum.cpp
@@ -44,9 +44,9 @@ KernelsData CumSumKernelPartialSum::GetMultiStageKernelsData(const Params& param
         auto jit = CreateJit(kernelName, cldnn_jit, entry_point);
         auto& kernel = kd.kernels[0];
         FillCLKernelData(kernel, dispatchData.stage_1, params.engineInfo, kernelName, jit, entry_point);
-        kernel.arguments.clear();  // Clear original output argument
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 0});
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
+        kernel.params.arguments.clear();  // Clear original output argument
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
         kd.internalBufferSizes.push_back(newParams.output.PhysicalSizeInBytes());
     }
     {
@@ -59,9 +59,9 @@ KernelsData CumSumKernelPartialSum::GetMultiStageKernelsData(const Params& param
 
         FillCLKernelData(kernel, dispatchData.stage_final, params.engineInfo, kernelName, jit, entry_point);
 
-        kernel.arguments.clear();  // Clear original output argument
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
-        kernel.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
+        kernel.params.arguments.clear();  // Clear original output argument
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
     }
     kd.internalBufferDataType = Datatype::F32;
 
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/deconvolution/deconvolution_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/deconvolution/deconvolution_kernel_base.cpp
index 5ec4e19da1e8a4..383427b6075ebd 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/deconvolution/deconvolution_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/deconvolution/deconvolution_kernel_base.cpp
@@ -137,7 +137,7 @@ KernelsData DeconvolutionKernelBase::GetKernelsData(const Params& params, const
                      !newParams.bias.empty(),
                      1,
                      GetFusedPrimitiveInputsCount(params));
-    kernel.arguments.push_back({ArgumentDescriptor::Types::SPLIT, 0});
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::SPLIT, 0});
 
     return {kd};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_b_fs_yx_fsv16.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_b_fs_yx_fsv16.cpp
index afaa202ffe002c..73b404f89e97e5 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_b_fs_yx_fsv16.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_b_fs_yx_fsv16.cpp
@@ -334,11 +334,11 @@ KernelsData EltwiseKernel_b_fs_yx_fsv16::GetKernelsData(const Params& params, co
 
     auto& kernel = kd.kernels[0];
 
-    kernel.workGroups.global = dispatchData.gws;
-    kernel.workGroups.local = dispatchData.lws;
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
 
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
-    kernel.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(),
+    kernel.params.workGroups.global = dispatchData.gws;
+    kernel.params.workGroups.local = dispatchData.lws;
+    kernel.params.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(),
                                    false,
                                    false,
                                    GetFusedPrimitiveInputsCount(params));
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_b_fs_yx_fsv4.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_b_fs_yx_fsv4.cpp
index 3f136eaf35b9ee..451145b63265ac 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_b_fs_yx_fsv4.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_b_fs_yx_fsv4.cpp
@@ -47,11 +47,11 @@ KernelsData EltwiseKernel_b_fs_yx_fsv4::GetKernelsData(const Params& params, con
 
     auto& kernel = kd.kernels[0];
 
-    kernel.workGroups.global = dispatchData.gws;
-    kernel.workGroups.local = dispatchData.lws;
+    kernel.params.workGroups.global = dispatchData.gws;
+    kernel.params.workGroups.local = dispatchData.lws;
 
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
-    kernel.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(),
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
+    kernel.params.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(),
                                    false,
                                    false,
                                    GetFusedPrimitiveInputsCount(params));
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_base.cpp
index 6a40df57a68691..370efc6f969bd5 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_base.cpp
@@ -621,11 +621,11 @@ KernelsData EltwiseKernelBase::GetCommonKernelsData(const Params& params, const
 
     auto& kernel = kd.kernels[0];
 
-    kernel.workGroups.global = dispatchData.gws;
-    kernel.workGroups.local = dispatchData.lws;
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
 
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
-    kernel.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(),
+    kernel.params.workGroups.global = dispatchData.gws;
+    kernel.params.workGroups.local = dispatchData.lws;
+    kernel.params.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(),
                                    false,
                                    false,
                                    GetFusedPrimitiveInputsCount(params));
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_fs_b_yx_fsv32.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_fs_b_yx_fsv32.cpp
index f8b43ea01a987f..91975fde6f1a09 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_fs_b_yx_fsv32.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_fs_b_yx_fsv32.cpp
@@ -97,11 +97,12 @@ KernelsData EltwiseKernel_fs_b_yx_fsv32::GetKernelsData(const Params& params, co
     size_t x = input.X().v;
     size_t global_size = featuresRoundedUp * batches * x * y;
 
-    kernel.workGroups.global = {std::max(global_size / 8, (size_t)1), 1, 1};
-    kernel.workGroups.local = GetOptimalLocalWorkGroupSizes(kernel.workGroups.global, params.engineInfo);
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
 
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
-    kernel.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(), false, false);
+    kernel.params.workGroups.global = {std::max(global_size / 8, (size_t)1), 1, 1};
+    kernel.params.workGroups.local = GetOptimalLocalWorkGroupSizes(kernel.params.workGroups.global, params.engineInfo);
+
+    kernel.params.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(), false, false);
 
     return {kd};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_mixed_byxf_and_fs_b_yx_fsv32.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_mixed_byxf_and_fs_b_yx_fsv32.cpp
index 288453139ccd8b..bf2ee1076120a3 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_mixed_byxf_and_fs_b_yx_fsv32.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_mixed_byxf_and_fs_b_yx_fsv32.cpp
@@ -131,12 +131,12 @@ KernelsData EltwiseKernel_mixed_byxf_and_fs_b_yx_fsv32::GetKernelsData(const Par
 
     // in fs_b_yx_fsv32 format we will process 2 features per work item, so reads/writes are done in full writes for
     // fp16
-    kernel.workGroups.global = {x, y, (featuresRoundedUp * batches) / 2};
+    kernel.params.workGroups.global = {x, y, (featuresRoundedUp * batches) / 2};
 
-    kernel.workGroups.local = {1, 1, 16};
+    kernel.params.workGroups.local = {1, 1, 16};
 
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
-    kernel.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(), false, false);
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
+    kernel.params.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(), false, false);
 
     return {kd};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_vload8.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_vload8.cpp
index 05910383aaee5c..a9470fbb0b7b42 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_vload8.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_vload8.cpp
@@ -105,10 +105,10 @@ KernelsData EltwiseKernel_vload8::GetKernelsData(const Params& params, const opt
     }
 
     auto& kernel = kd.kernels[0];
-    kernel.workGroups.global = {std::max(newParams.inputs[0].LogicalSize() / 8, (size_t)1), 1, 1};
-    kernel.workGroups.local = GetOptimalLocalWorkGroupSizes(kernel.workGroups.global, params.engineInfo);
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
-    kernel.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(), false, false);
+    kernel.params.workGroups.global = {std::max(newParams.inputs[0].LogicalSize() / 8, (size_t)1), 1, 1};
+    kernel.params.workGroups.local = GetOptimalLocalWorkGroupSizes(kernel.params.workGroups.global, params.engineInfo);
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
+    kernel.params.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(), false, false);
 
     return {kd};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/fused_conv_eltwise/fused_conv_eltwise_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/fused_conv_eltwise/fused_conv_eltwise_kernel_base.cpp
index ee6087e5417429..a1d86baf9a22c3 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/fused_conv_eltwise/fused_conv_eltwise_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/fused_conv_eltwise/fused_conv_eltwise_kernel_base.cpp
@@ -282,12 +282,12 @@ KernelsData fused_conv_eltwise_kernel_base::GetCommonKernelsData(const Params& p
                      true,
                      !newParams.bias.empty(),
                      1);
-    kernel.arguments.push_back({ArgumentDescriptor::Types::SPLIT, 0});
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::SPLIT, 0});
     // eltwise's second input
     if (newParams.second_input_in_output) {
-        kernel.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
     } else {
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
     }
 
     kd.autoTuneIndex = autoTuneIndex;
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_ref.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_ref.cpp
index 3da6f44337bd3c..fa523d35ef97e7 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_ref.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_ref.cpp
@@ -1,18 +1,6 @@
-/*
-// Copyright (c) 2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
 //
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-*/
 
 #include "gather_nd_kernel_ref.h"
 #include "kernel_selector_utils.h"
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_ref.h b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_ref.h
index 82fa0913413f34..ce449e229cbe0a 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_ref.h
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_ref.h
@@ -1,18 +1,6 @@
-/*
-// Copyright (c) 2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
 //
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-*/
 
 #pragma once
 
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_selector.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_selector.cpp
index 82c938d08c9871..306f2b1a730342 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_selector.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_selector.cpp
@@ -1,18 +1,6 @@
-/*
-// Copyright (c) 2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
 //
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-*/
 
 #include "gather_nd_kernel_selector.h"
 #include "gather_nd_kernel_ref.h"
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_selector.h b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_selector.h
index e4d2ca1edb4149..e97add6ee369e8 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_selector.h
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/gather/gather_nd_kernel_selector.h
@@ -1,18 +1,6 @@
-/*
-// Copyright (c) 2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
 //
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-*/
 
 #pragma once
 
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_elt_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_elt_kernel_base.cpp
index d91ff0926bb3fc..587832da828038 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_elt_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_elt_kernel_base.cpp
@@ -80,12 +80,12 @@ KernelsData LSTMEltKernelBase::GetCommonKernelsData(const Params& params, const
     auto entryPoint = GetEntryPoint(kernelName, newParams.layerID, options);
     auto jit = CreateJit(kernelName, cldnnJit, entryPoint);
 
-    kernel.workGroups.global = {out.X().v, out.Batch().v, 1};
-    kernel.kernelString = GetKernelString(kernelName, jit, entryPoint, params.engineInfo);
-    kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 0});
-    kernel.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
+    kernel.params.workGroups.global = {out.X().v, out.Batch().v, 1};
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entryPoint, params.engineInfo);
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 0});
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
     if (orgParams.has_cell) {
-        kernel.arguments.push_back({ArgumentDescriptor::Types::CELL, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::CELL, 0});
     }
 
     return {kd};
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemm_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemm_kernel_base.cpp
index b7b4d4b0113be3..e4b547deacbcc5 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemm_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemm_kernel_base.cpp
@@ -50,17 +50,17 @@ KernelsData LSTMGemmKernelBase::GetCommonKernelsData(const Params& params, const
     auto entryPoint = GetEntryPoint(kernelName, newParams.layerID, options);
     auto jit = CreateJit(kernelName, cldnnJit, entryPoint);
 
-    kernel.workGroups.global = {out.X().v, out.Batch().v, 1};
-    kernel.kernelString = GetKernelString(kernelName, jit, entryPoint, params.engineInfo);
-    kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 0});
-    kernel.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
-    kernel.arguments.push_back({ArgumentDescriptor::Types::WEIGHTS, 0});
+    kernel.params.workGroups.global = {out.X().v, out.Batch().v, 1};
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entryPoint, params.engineInfo);
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 0});
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::OUTPUT, 0});
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::WEIGHTS, 0});
     if (orgParams.hasHidden) {
-        kernel.arguments.push_back({ArgumentDescriptor::Types::HIDDEN, 0});
-        kernel.arguments.push_back({ArgumentDescriptor::Types::RECURRENT, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::HIDDEN, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::RECURRENT, 0});
     }
     if (orgParams.hasBias) {
-        kernel.arguments.push_back({ArgumentDescriptor::Types::BIAS, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::BIAS, 0});
     }
 
     return {kd};
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemv_gpu_subgroup1x64_bfyx_ff_simd16.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemv_gpu_subgroup1x64_bfyx_ff_simd16.cpp
index b5c2d1a482a787..f94f1c6bb418d0 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemv_gpu_subgroup1x64_bfyx_ff_simd16.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemv_gpu_subgroup1x64_bfyx_ff_simd16.cpp
@@ -36,7 +36,7 @@ KernelsData LSTMGemvKernel_subgroup1x64_bfyx_ff_SIMD16::GetKernelsData(const Par
     const auto& out = orgParams.output;
 
     if ((input.Batch().v == 1) && (input.X().v >= 64) && (input.Y().v == 1))
-        kernel.workGroups.global = {16, out.X().v, out.Batch().v};
+        kernel.params.workGroups.global = {16, out.X().v, out.Batch().v};
 
     return kernelsData;
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemv_gpu_subgroup1x64_bfyx_hh_simd16.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemv_gpu_subgroup1x64_bfyx_hh_simd16.cpp
index 437fc621fbc320..437dc458c265c7 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemv_gpu_subgroup1x64_bfyx_hh_simd16.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm/lstm_gemv_gpu_subgroup1x64_bfyx_hh_simd16.cpp
@@ -36,7 +36,7 @@ KernelsData LSTMGemvKernel_subgroup1x64_bfyx_hh_SIMD16::GetKernelsData(const Par
     const auto& out = orgParams.output;
 
     if ((input.Batch().v == 1) && (input.X().v >= 64) && (input.Y().v == 1))
-        kernel.workGroups.global = {16, out.X().v, out.Batch().v};
+        kernel.params.workGroups.global = {16, out.X().v, out.Batch().v};
 
     return kernelsData;
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_input_bfyx_opt.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_input_bfyx_opt.cpp
index 16271d23130e47..dfd13d8ce5bc83 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_input_bfyx_opt.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_input_bfyx_opt.cpp
@@ -89,9 +89,9 @@ KernelsData LSTM_DynamicInputKernelBfyxOpt::GetKernelsData(const Params& params,
     auto jit = CreateJit(kernelName, cldnn_jit, entry_point);
 
     auto& kernel = kd.kernels[0];
-    kernel.workGroups.global = dispatchData.gws;
-    kernel.workGroups.local = dispatchData.lws;
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo);
+    kernel.params.workGroups.global = dispatchData.gws;
+    kernel.params.workGroups.local = dispatchData.lws;
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo);
     SetKernelArguments(dlstm_params, kernel);
 
     return { kd };
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_input_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_input_kernel_base.cpp
index 8066bf61831f32..2a1491fb41b583 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_input_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_input_kernel_base.cpp
@@ -36,12 +36,12 @@ LSTM_DynamicInputKernelBase::DispatchData LSTM_DynamicInputKernelBase::SetDefaul
 }
 
 void kernel_selector::LSTM_DynamicInputKernelBase::SetKernelArguments(const lstm_dynamic_input_params& params, clKernelData& kernel) const {
-    kernel.arguments.push_back({ ArgumentDescriptor::Types::INPUT, 0 });
-    kernel.arguments.push_back({ ArgumentDescriptor::Types::INPUT, 1 });
-    kernel.arguments.push_back({ ArgumentDescriptor::Types::OUTPUT, 0 });
-    kernel.arguments.push_back({ ArgumentDescriptor::Types::WEIGHTS, 0 });
+    kernel.params.arguments.push_back({ ArgumentDescriptor::Types::INPUT, 0 });
+    kernel.params.arguments.push_back({ ArgumentDescriptor::Types::INPUT, 1 });
+    kernel.params.arguments.push_back({ ArgumentDescriptor::Types::OUTPUT, 0 });
+    kernel.params.arguments.push_back({ ArgumentDescriptor::Types::WEIGHTS, 0 });
     if (!params.bias.empty()) {
-        kernel.arguments.push_back({ ArgumentDescriptor::Types::BIAS, 0 });
+        kernel.params.arguments.push_back({ ArgumentDescriptor::Types::BIAS, 0 });
     }
 }
 
@@ -61,8 +61,8 @@ KernelsData LSTM_DynamicInputKernelBase::GetCommonKernelsData(const Params& para
     auto jit = CreateJit(kernelName, cldnn_jit, entry_point);
 
     auto& kernel = k_data.kernels[0];
-    kernel.workGroups.global = dispatchData.gws;
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo);
+    kernel.params.workGroups.global = dispatchData.gws;
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo);
     SetKernelArguments(orgParams, kernel);
 
     return {k_data};
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_timeloop_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_timeloop_kernel_base.cpp
index a39c724c92751f..8327a025c8e985 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_timeloop_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/lstm_dynamic/lstm_dynamic_timeloop_kernel_base.cpp
@@ -87,21 +87,21 @@ LSTM_DynamicTimeloopKernelBase::DispatchData LSTM_DynamicTimeloopKernelBase::Set
 
 void kernel_selector::LSTM_DynamicTimeloopKernelBase::SetKernelArguments(const lstm_dynamic_timeloop_params& params, clKernelData& kernel) const {
     uint32_t input_idx = 0;
-    kernel.arguments.push_back({ ArgumentDescriptor::Types::INPUT, input_idx++ });
-    kernel.arguments.push_back({ ArgumentDescriptor::Types::INPUT, input_idx++ });
-    kernel.arguments.push_back({ ArgumentDescriptor::Types::OUTPUT, 0 });
-    kernel.arguments.push_back({ ArgumentDescriptor::Types::RECURRENT, 0 });
+    kernel.params.arguments.push_back({ ArgumentDescriptor::Types::INPUT, input_idx++ });
+    kernel.params.arguments.push_back({ ArgumentDescriptor::Types::INPUT, input_idx++ });
+    kernel.params.arguments.push_back({ ArgumentDescriptor::Types::OUTPUT, 0 });
+    kernel.params.arguments.push_back({ ArgumentDescriptor::Types::RECURRENT, 0 });
     if (params.has_hidden) {
-        kernel.arguments.push_back({ ArgumentDescriptor::Types::HIDDEN, 0 });
+        kernel.params.arguments.push_back({ ArgumentDescriptor::Types::HIDDEN, 0 });
     }
     if (params.has_cell) {
-        kernel.arguments.push_back({ ArgumentDescriptor::Types::CELL, 0 });
+        kernel.params.arguments.push_back({ ArgumentDescriptor::Types::CELL, 0 });
     }
     if (params.has_last_hidden_output) {
-        kernel.arguments.push_back({ ArgumentDescriptor::Types::INPUT, input_idx++ });
+        kernel.params.arguments.push_back({ ArgumentDescriptor::Types::INPUT, input_idx++ });
     }
     if (params.has_last_cell_output) {
-        kernel.arguments.push_back({ ArgumentDescriptor::Types::INPUT, input_idx++ });
+        kernel.params.arguments.push_back({ ArgumentDescriptor::Types::INPUT, input_idx++ });
     }
 }
 
@@ -122,9 +122,9 @@ KernelsData LSTM_DynamicTimeloopKernelBase::GetCommonKernelsData(const Params& p
     auto jit = CreateJit(kernelName, cldnn_jit, entry_point);
 
     auto& kernel = k_data.kernels[0];
-    kernel.workGroups.global = dispatchData.gws;
-    kernel.workGroups.local  = dispatchData.lws;
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo);
+    kernel.params.workGroups.global = dispatchData.gws;
+    kernel.params.workGroups.local  = dispatchData.lws;
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo);
     SetKernelArguments(org_params, kernel);
     return {k_data};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/max_unpooling/max_unpooling_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/max_unpooling/max_unpooling_kernel_base.cpp
index b34c674a75500c..5534217a9ce900 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/max_unpooling/max_unpooling_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/max_unpooling/max_unpooling_kernel_base.cpp
@@ -70,7 +70,7 @@ KernelsData MaxUnpoolingKernelBase::GetCommonKernelsData(const Params& params,
 
     auto& kernel = kd.kernels[0];
     FillCLKernelData(kernel, dispatchData, params.engineInfo, kernelName, jit, entry_point);
-    kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
 
     return {kd};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/mvn/mvn_kernel_b_fs_yx_fsv16_imad.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/mvn/mvn_kernel_b_fs_yx_fsv16_imad.cpp
index 670f831594d51c..32fd8e5744cc4c 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/mvn/mvn_kernel_b_fs_yx_fsv16_imad.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/mvn/mvn_kernel_b_fs_yx_fsv16_imad.cpp
@@ -200,9 +200,9 @@ KernelsData MVNKernel_b_fs_yx_fsv16_imad::GetMultiStageKernelsData(const mvn_par
                          false,
                          0,
                          0);
-        kernel.arguments.clear();  // Clear original output argument
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 0});
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
+        kernel.params.arguments.clear();  // Clear original output argument
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
         kd.internalBufferSizes.push_back(params.output.Batch().v * Align(params.output.Feature().v, fsv) *
                                          dispatchData.item_groups * intermidiate_bytes);
     }
@@ -224,9 +224,9 @@ KernelsData MVNKernel_b_fs_yx_fsv16_imad::GetMultiStageKernelsData(const mvn_par
                          false,
                          0,
                          0);
-        kernel.arguments.clear();  // Clear original output argument
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 1});
+        kernel.params.arguments.clear();  // Clear original output argument
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 1});
         kd.internalBufferSizes.push_back(params.output.Batch().v * Align(params.output.Feature().v, fsv) *
                                          intermidiate_bytes);
     }
@@ -248,10 +248,10 @@ KernelsData MVNKernel_b_fs_yx_fsv16_imad::GetMultiStageKernelsData(const mvn_par
                          false,
                          0,
                          0);
-        kernel.arguments.clear();  // Clear original output argument
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 0});
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 1});
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
+        kernel.params.arguments.clear();  // Clear original output argument
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 1});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
     }
     if (params.mvnNormalizeVariance) {
         // Variance second stage
@@ -271,9 +271,9 @@ KernelsData MVNKernel_b_fs_yx_fsv16_imad::GetMultiStageKernelsData(const mvn_par
                          false,
                          0,
                          0);
-        kernel.arguments.clear();  // Clear original output argument
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 2});
+        kernel.params.arguments.clear();  // Clear original output argument
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 2});
         kd.internalBufferSizes.push_back(params.output.Batch().v * Align(params.output.Feature().v, fsv) *
                                          intermidiate_bytes);
     }
@@ -296,9 +296,9 @@ KernelsData MVNKernel_b_fs_yx_fsv16_imad::GetMultiStageKernelsData(const mvn_par
                          false,
                          1,
                          GetFusedPrimitiveInputsCount(params));
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 1});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 1});
         if (params.mvnNormalizeVariance) {
-            kernel.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 2});
+            kernel.params.arguments.push_back({ArgumentDescriptor::Types::INTERNAL_BUFFER, 2});
         }
     }
     kd.internalBufferDataType = Datatype::F32;
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/normalize/normalize_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/normalize/normalize_kernel_base.cpp
index 2e36b028e28b15..a5c6dde629d8c7 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/normalize/normalize_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/normalize/normalize_kernel_base.cpp
@@ -72,7 +72,7 @@ KernelsData NormalizeKernelBase::GetCommonKernelsData(const Params& params,
                      1,
                      GetFusedPrimitiveInputsCount(params));
 
-    kernel.arguments.push_back({ArgumentDescriptor::Types::SCALE_TABLE, 0});
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::SCALE_TABLE, 0});
 
     return {kd};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/pooling/pooling_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/pooling/pooling_kernel_base.cpp
index eb4313f39bc71d..eb42f283703225 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/pooling/pooling_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/pooling/pooling_kernel_base.cpp
@@ -182,7 +182,7 @@ KernelsData PoolingKernelBase::GetCommonKernelsData(const Params& params,
     FillCLKernelData(kernel, dispatchData, params.engineInfo, kernelName, jit, entry_point, DEFAULT, false, false, 1,
                      GetFusedPrimitiveInputsCount(params));
     if (orgParams.poolType == PoolType::MAX_WITH_ARGMAX)
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
 
 
     return {kd};
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/quantize/quantize_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/quantize/quantize_kernel_base.cpp
index 457e90ba97e0fa..b6a6a6a71dba8b 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/quantize/quantize_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/quantize/quantize_kernel_base.cpp
@@ -68,10 +68,10 @@ KernelsData QuantizeKernelBase::GetKernelsData(const Params& params, const optio
 
     auto& kernel = kd.kernels[0];
 
-    kernel.workGroups.global = dispatchData.gws;
-    kernel.workGroups.local = dispatchData.lws;
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
-    kernel.arguments = GetArgsDesc(static_cast<int>(newParams.inputs.size()), false, false);
+    kernel.params.workGroups.global = dispatchData.gws;
+    kernel.params.workGroups.local = dispatchData.lws;
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
+    kernel.params.arguments = GetArgsDesc(static_cast<int>(newParams.inputs.size()), false, false);
 
     return {kd};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reorder/reorder_biplanar_nv12.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reorder/reorder_biplanar_nv12.cpp
index e88e3bb10195f3..2e8c89238eb2cf 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reorder/reorder_biplanar_nv12.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reorder/reorder_biplanar_nv12.cpp
@@ -33,7 +33,7 @@ KernelsData reorder_biplanar_nv12::GetKernelsData(const Params& params, const op
         return {};
     }
     KernelsData kd = GetCommonKernelsData(orgParams, options);
-    kd[0].kernels[0].arguments = GetArgsDesc(2, false, false);
+    kd[0].kernels[0].params.arguments = GetArgsDesc(2, false, false);
     return kd;
 }
 
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reorder/reorder_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reorder/reorder_kernel_base.cpp
index cf227d83504e4d..9629aa2dbc1039 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reorder/reorder_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reorder/reorder_kernel_base.cpp
@@ -206,7 +206,7 @@ KernelsData ReorderKernelBase::GetCommonKernelsData(const reorder_weights_params
 
     FillCLKernelData(kernel, dispatchData, params.engineInfo, kernelName, jit, entry_point);
 
-    kernel.arguments = GetArgsDesc(1, false, false);
+    kernel.params.arguments = GetArgsDesc(1, false, false);
 
     return {kd};
 }
@@ -230,9 +230,9 @@ KernelsData ReorderKernelBase::GetCommonKernelsData(const reorder_params& params
 
     FillCLKernelData(kernel, dispatchData, params.engineInfo, kernelName, jit, entry_point);
 
-    kernel.arguments = GetArgsDesc(1, false, false);
+    kernel.params.arguments = GetArgsDesc(1, false, false);
     if (newParams.mode == MeanSubtractMode::IN_BUFFER) {
-        kernel.arguments.push_back({ArgumentDescriptor::Types::BIAS, 0});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::BIAS, 0});
     }
 
     return {kd};
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reshape/reshape_kernel_ref.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reshape/reshape_kernel_ref.cpp
index 0a76fed62c7a47..401ffc73d9cf89 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reshape/reshape_kernel_ref.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/reshape/reshape_kernel_ref.cpp
@@ -56,10 +56,10 @@ KernelsData ReshapeKernelRef::GetKernelsData(const Params& params, const optiona
         gws2 *= in_dims[i].v;
     }
 
-    kernel.workGroups.global = {gws0, gws1, gws2};
-    kernel.workGroups.local = GetOptimalLocalWorkGroupSizes(kernel.workGroups.global, params.engineInfo);
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
-    kernel.arguments = GetArgsDesc(1, false, false);
+    kernel.params.workGroups.global = {gws0, gws1, gws2};
+    kernel.params.workGroups.local = GetOptimalLocalWorkGroupSizes(kernel.params.workGroups.global, params.engineInfo);
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
+    kernel.params.arguments = GetArgsDesc(1, false, false);
 
     return {kd};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/roi_pooling/roi_pooling_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/roi_pooling/roi_pooling_kernel_base.cpp
index 147c2551b281be..271929559c42cc 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/roi_pooling/roi_pooling_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/roi_pooling/roi_pooling_kernel_base.cpp
@@ -55,9 +55,9 @@ KernelsData ROIPoolingKernelBase::GetCommonKernelsData(const Params& params,
 
     auto& kernel = kd.kernels[0];
     FillCLKernelData(kernel, dispatchData, params.engineInfo, kernelName, jit, entry_point);
-    kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
+    kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 1});
     if (orgParams.mode == PoolType::DEFORMABLE_BILINEAR && !orgParams.no_trans)
-        kernel.arguments.push_back({ArgumentDescriptor::Types::INPUT, 2});
+        kernel.params.arguments.push_back({ArgumentDescriptor::Types::INPUT, 2});
 
     return {kd};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/select/select_kernel_base.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/select/select_kernel_base.cpp
index b827527d22af58..4228d9fabfba36 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/select/select_kernel_base.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/select/select_kernel_base.cpp
@@ -131,11 +131,11 @@ KernelsData SelectKernelBase::GetCommonKernelsData(const Params& params, const o
 
     auto& kernel = kd.kernels[0];
 
-    kernel.workGroups.global = dispatchData.gws;
-    kernel.workGroups.local = dispatchData.lws;
+    kernel.params.workGroups.global = dispatchData.gws;
+    kernel.params.workGroups.local = dispatchData.lws;
 
-    kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
-    kernel.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(), false, false);
+    kernel.code.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
+    kernel.params.arguments = GetArgsDesc((uint32_t)newParams.inputs.size(), false, false);
 
     return {kd};
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/auto_tuner.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/auto_tuner.cpp
index bdef07b6f2964e..0c39d4e69231cd 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/auto_tuner.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/auto_tuner.cpp
@@ -315,9 +315,11 @@ void AutoTuner::RemoveKernel(const std::string& cacheFilePath,
     }
 }
 
-std::tuple<std::string, int> AutoTuner::LoadKernelOffline(std::shared_ptr<TuningCache> deviceCache,
+std::tuple<std::string, int> AutoTuner::LoadKernelOffline(TuningCache* deviceCache,
                                                           const Params& params) {
     static const uint32_t defaultComputeUnits = 24;
+    if (!deviceCache)
+        return {};
     auto result = deviceCache->LoadKernel(params, false);
     if (std::get<0>(result).empty() && params.engineInfo.computeUnitsCount != defaultComputeUnits) {
         result = deviceCache->LoadKernel(params, defaultComputeUnits);
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/auto_tuner.h b/inference-engine/thirdparty/clDNN/kernel_selector/core/auto_tuner.h
index 709f2955b6e657..22284622c3bf69 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/auto_tuner.h
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/auto_tuner.h
@@ -74,7 +74,7 @@ class AutoTuner {
                      const int tuneIndex);
     void RemoveKernel(const std::string& cacheFilePath,
                       const Params& params);
-    std::tuple<std::string, int> LoadKernelOffline(std::shared_ptr<TuningCache> cache,
+    std::tuple<std::string, int> LoadKernelOffline(TuningCache* cache,
                                                    const Params& params);
 
 private:
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/gather_nd_ref.cl b/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/gather_nd_ref.cl
index 91cb7d9be773e9..d59d93fec17326 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/gather_nd_ref.cl
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/gather_nd_ref.cl
@@ -1,16 +1,6 @@
-// Copyright (c) 2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
 //
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
 
 #include "include/fetch.cl"
 
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/max_unpooling_gpu_ref.cl b/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/max_unpooling_gpu_ref.cl
index 94e77fa113fb09..d8ad1abfd7a692 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/max_unpooling_gpu_ref.cl
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/max_unpooling_gpu_ref.cl
@@ -4,7 +4,7 @@
 
 #include "include/include_all.cl"
 
-KERNEL(pooling_gpu)(const __global UNIT_TYPE* input, __global UNIT_TYPE* output, const __global float* arg_max)
+KERNEL(pooling_gpu)(const __global INPUT0_TYPE* input, __global OUTPUT_TYPE* output, const __global float* arg_max)
 {
 #if OUTPUT_LAYOUT_BFYX  || OUTPUT_LAYOUT_BYXF
     const uint x    = (uint)get_global_id(0);
@@ -12,7 +12,7 @@ KERNEL(pooling_gpu)(const __global UNIT_TYPE* input, __global UNIT_TYPE* output,
     const uint bf   = (uint)get_global_id(2);
     const uint f    = bf % INPUT0_FEATURE_NUM;
     const uint b    = bf / INPUT0_FEATURE_NUM;
-    
+
     if (x >= INPUT0_SIZE_X)
     {
         return;
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/common/kernel_base_opencl.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/common/kernel_base_opencl.cpp
index 392d0ed2bc2bb2..aae0e195279f8f 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/common/kernel_base_opencl.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/common/kernel_base_opencl.cpp
@@ -182,9 +182,9 @@ void KernelBaseOpenCL::FillCLKernelData(clKernelData& kernel,
                                         int number_of_inputs,
                                         uint32_t number_of_inputs_for_fused_prims) const {
     KernelBase::CheckDispatchData(kernelMapName, dispatchData);
-    kernel.workGroups.global = dispatchData.gws;
-    kernel.workGroups.local = dispatchData.lws;
-    kernel.kernelString = GetKernelString(kernelMapName, jit, entryPoint, engine_info, exeMode);
-    kernel.arguments = GetArgsDesc(number_of_inputs, weights, bias, number_of_inputs_for_fused_prims);
+    kernel.code.kernelString = GetKernelString(kernelMapName, jit, entryPoint, engine_info, exeMode);
+    kernel.params.workGroups.global = dispatchData.gws;
+    kernel.params.workGroups.local = dispatchData.lws;
+    kernel.params.arguments = GetArgsDesc(number_of_inputs, weights, bias, number_of_inputs_for_fused_prims);
 }
 }  // namespace kernel_selector
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/device_cache_reader.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/device_cache_reader.cpp
similarity index 65%
rename from inference-engine/thirdparty/clDNN/src/gpu/device_cache_reader.cpp
rename to inference-engine/thirdparty/clDNN/kernel_selector/core/device_cache_reader.cpp
index 0d0b310a1793a3..5a61c3c3131820 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/device_cache_reader.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/device_cache_reader.cpp
@@ -3,13 +3,13 @@
 //
 
 #include "device_cache_reader.h"
-#include "include/to_string_utils.h"
 #include "auto_tuner.h"
 #include <limits>
 #include "istreamwrapper.h"
 
 #ifdef _WIN32
 #define WIN32_LEAN_AND_MEAN
+#define NOMINMAX
 #include <windows.h>
 #include <SetupAPI.h>
 #include <devguid.h>
@@ -25,17 +25,15 @@
 #include <iostream>
 #include <utility>
 
-namespace cldnn {
-namespace gpu {
-namespace {
+namespace kernel_selector {
 
-std::shared_ptr<kernel_selector::TuningCache> get_cache_from_file(std::string tuning_cache_path) {
+std::shared_ptr<kernel_selector::TuningCache> CreateTuningCacheFromFile(std::string tuning_cache_path) {
     if (tuning_cache_path.compare("cache.json") == 0) {
 #ifdef _WIN32
         char path[MAX_PATH];
         HMODULE hm = NULL;
         GetModuleHandleEx(GET_MODULE_HANDLE_EX_FLAG_FROM_ADDRESS | GET_MODULE_HANDLE_EX_FLAG_UNCHANGED_REFCOUNT,
-            (LPCSTR)&get_cache_from_file,
+            (LPCSTR)&CreateTuningCacheFromFile,
             &hm);
         GetModuleFileName(hm, path, sizeof(path));
         std::string bin_path(path);
@@ -51,19 +49,4 @@ std::shared_ptr<kernel_selector::TuningCache> get_cache_from_file(std::string tu
 
     return std::make_shared<kernel_selector::TuningCache>(tuning_cache_path, false);
 }
-}  // namespace
-
-device_cache_reader::device_cache_reader(const std::string tuning_file_path) {
-    {
-        try {
-            _dev_cache = get_cache_from_file(tuning_file_path);
-        }
-        catch (...) {
-            std::cout << "[WARNING] error during parsing cache file, tuning data won't be used" << std::endl;
-            _dev_cache = std::make_shared<kernel_selector::TuningCache>();
-        }
-    }
-}
-
-}  // namespace gpu
-}  // namespace cldnn
+}  // namespace kernel_selector
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/device_cache_reader.h b/inference-engine/thirdparty/clDNN/kernel_selector/core/device_cache_reader.h
new file mode 100644
index 00000000000000..9621296ddae07c
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/device_cache_reader.h
@@ -0,0 +1,14 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+#include <memory>
+#include <string>
+
+namespace kernel_selector {
+class TuningCache;
+
+std::shared_ptr<kernel_selector::TuningCache> CreateTuningCacheFromFile(std::string tuning_cache_path);
+
+}  // namespace kernel_selector
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/kernel_selector.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/kernel_selector.cpp
index 4d4a06b53fe8e9..2105b0436589b1 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/kernel_selector.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/kernel_selector.cpp
@@ -13,6 +13,7 @@
 #include <vector>
 #include <tuple>
 #include <set>
+#include <iostream>
 
 // #define ENABLE_ENV
 // #define ENABLE_ENV_PRINT
@@ -105,7 +106,7 @@ KernelsData kernel_selector_base::GetNaiveBestKernel(const Params& params,
     // TODO: find a better place to located this assignment
     if (kernelsData.size()) {
         kernelsData[0].kernelName = kernelName;
-        kernelsData[0].kernels[0].layerID = params.layerID;
+        kernelsData[0].kernels[0].params.layerID = params.layerID;
     }
 
     return kernelsData;
@@ -123,7 +124,7 @@ KernelsData kernel_selector_base::GetAutoTuneBestKernel(const Params& params,
     std::tuple<std::string, int> cachedKernelConfig;
     if (options.tuningParams.mode == TuningMode::TUNING_DISABLED && !int8_kernel) {  // Try to load kernel/config from offline cache
 #if ENABLE_OFFLINE_TUNING_CACHE
-        cachedKernelConfig = autoTuner.LoadKernelOffline(params.engineInfo.deviceCache, params);
+        cachedKernelConfig = autoTuner.LoadKernelOffline(params.engineInfo.deviceCache.get(), params);
 #else
         return GetNaiveBestKernel(params, options, kType);
 #endif
@@ -145,7 +146,7 @@ KernelsData kernel_selector_base::GetAutoTuneBestKernel(const Params& params,
                 if (kds.size() && kds[0].kernels.size()) {
                     kernelsData = kds;
                     kernelsData[0].kernelName = cachedkernelName;
-                    kernelsData[0].kernels[0].layerID = params.layerID;
+                    kernelsData[0].kernels[0].params.layerID = params.layerID;
                 }
                 break;
             }
@@ -218,7 +219,7 @@ KernelsData kernel_selector_base::GetAutoTuneBestKernel(const Params& params,
 
     if (kernelsData.size()) {
         kernelsData[0].kernelName = kernelName;
-        kernelsData[0].kernels[0].layerID = params.layerID;
+        kernelsData[0].kernels[0].params.layerID = params.layerID;
         autoTuner.StoreKernel(options.tuningParams.cacheFilePath,
                                 params,
                                 kernelName,
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/kernel_selector_common.h b/inference-engine/thirdparty/clDNN/kernel_selector/core/kernel_selector_common.h
index 5a0d62336cbd60..d0b0054f1b2dda 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/kernel_selector_common.h
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/kernel_selector_common.h
@@ -5,6 +5,7 @@
 #pragma once
 
 #include "kernel_selector_params.h"
+#include "cldnn/runtime/kernel_args.hpp"
 
 #include <cfloat>
 #include <cstdint>
@@ -41,118 +42,28 @@ namespace kernel_selector {
 
 std::string GetStringEnv(const char* varName);
 
-////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-// KernelString
-////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-struct KernelString {
-    std::string str;
-    std::string jit;
-    std::string undefs;
-    std::string options;
-    std::string entry_point;
-    bool batch_compilation;
-
-    KernelString() : str(""), jit(""), undefs(""), options(""), entry_point(""), batch_compilation(false) {}
-
-    std::string get_hash() { return str + jit + undefs + options + entry_point; }
-};
-
-////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-// WorkGroupSizes
-////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-struct WorkGroupSizes {
-    std::vector<size_t> global;
-    std::vector<size_t> local;
-};
-
-////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-// Scalar
-////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-struct ScalarDescriptor {
-    union ValueT {
-        uint8_t u8;
-        uint16_t u16;
-        uint32_t u32;
-        uint64_t u64;
-        int8_t s8;
-        int16_t s16;
-        int32_t s32;
-        int64_t s64;
-        float f32;
-        double f64;
-    };
-
-    enum class Types {
-        UINT8,
-        UINT16,
-        UINT32,
-        UINT64,
-        INT8,
-        INT16,
-        INT32,
-        INT64,
-        FLOAT32,
-        FLOAT64,
-    };
-
-    Types t;
-    ValueT v;
-};
+using KernelString = cldnn::kernel_string;
+using WorkGroupSizes = cldnn::work_group_sizes;
+using ScalarDescriptor = cldnn::scalar_desc;
+using Scalars = cldnn::scalars_desc;
+using ArgumentDescriptor = cldnn::argument_desc;
+using Arguments = cldnn::arguments_desc;
+using KernelParams = cldnn::kernel_arguments_desc;
 
-using Scalars = std::vector<ScalarDescriptor>;
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-// ArgumentDescpirtor
+// KernelCode
 ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-struct ArgumentDescriptor {
-    enum class Types {
-        INPUT,
-        OUTPUT,
-        WEIGHTS,
-        BIAS,
-        SCALE_TABLE,
-        SLOPE,
-        SPLIT,
-        INTERNAL_BUFFER,
-        SCALAR,
-        RECURRENT,  // RNN/LSTM/GRU recurrent weights
-        HIDDEN,     // RNN/LSTM/GRU hidden input
-        CELL,       // LSTM cell input
-        LSTM_PACK,  // LSTM packed output
-        WEIGHTS_ZERO_POINTS,
-        ACTIVATIONS_ZERO_POINTS,
-        COMPENSATION,
-        INPUT_OF_FUSED_PRIMITIVE
-    };
-
-    enum class ScalarTypes {
-        UINT8,
-        UINT16,
-        UINT32,
-        UINT64,
-        INT8,
-        INT16,
-        INT32,
-        INT64,
-        FLOAT32,
-        FLOAT64,
-    };
-
-    Types t;
-    uint32_t index;
+struct KernelCode {
+    std::shared_ptr<KernelString> kernelString;
 };
 
-using Arguments = std::vector<ArgumentDescriptor>;
-
 ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 // clKernelData
 ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 struct clKernelData {
-    std::shared_ptr<KernelString> kernelString;
-    WorkGroupSizes workGroups;
-    Arguments arguments;
-    Scalars scalars;
-    std::string layerID;  // TODO: in order to support run single layer. think about more appropriate place
+    KernelCode code;
+    KernelParams params;
 };
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
diff --git a/inference-engine/thirdparty/clDNN/runtime/CMakeLists.txt b/inference-engine/thirdparty/clDNN/runtime/CMakeLists.txt
new file mode 100644
index 00000000000000..1089d3f9baddf5
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/CMakeLists.txt
@@ -0,0 +1,79 @@
+# Copyright (C) 2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+#
+
+# =================================== Name / Output settings from parent ===============================
+
+set(CLDNN_BUILD__PROJ             "cldnn_runtime")
+set(CLDNN_BUILD__PROJ_LABEL       "${CLDNN_BUILD__PROJ}")
+set(CLDNN_BUILD__PROJ_OUTPUT_NAME "${CLDNN_BUILD__PROJ}${CLDNN__OUT_CPU_SUFFIX}")
+
+# ========================================== Compiler options ==========================================
+
+intel_config_flag_apply_settings(
+    CompilerOptions
+    CMAKE_CXX_FLAGS
+    ALL_PATTERN ""
+    SET
+      RttiEnabled
+  )
+
+# ========================================= Source/Header files ========================================
+
+set(__CLDNN_Directory__main           "${CMAKE_CURRENT_SOURCE_DIR}")
+set(__CLDNN_Label__main                "")
+file(GLOB __CLDNN_Sources__main
+    "${__CLDNN_Directory__main}/*.h"
+    "${__CLDNN_Directory__main}/*.hpp"
+    "${__CLDNN_Directory__main}/*.cpp"
+  )
+
+file(GLOB __CLDNN_Sources__ocl
+    "${__CLDNN_Directory__main}/ocl/*.h"
+    "${__CLDNN_Directory__main}/ocl/*.hpp"
+    "${__CLDNN_Directory__main}/ocl/*.cpp"
+)
+
+set(__CLDNN_AllSources
+    ${__CLDNN_Sources__main}
+    ${__CLDNN_Sources__ocl}
+  )
+
+# =============================================== Filters ==============================================
+
+source_group("${__CLDNN_Label__main}"             FILES ${__CLDNN_Sources__main})
+
+# ===================================== Include/Link directories =======================================
+
+include_directories(
+    "${CMAKE_CURRENT_SOURCE_DIR}/include"
+    "${__CLDNN_Directory__main}"
+  )
+
+# =================================== Link targets and dependencies ====================================
+
+# Main static library.
+add_library("${CLDNN_BUILD__PROJ}" STATIC
+    ${__CLDNN_AllSources}
+  )
+
+set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY PROJECT_LABEL "${CLDNN_BUILD__PROJ_LABEL}")
+set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY OUTPUT_NAME   "${CLDNN_BUILD__PROJ_OUTPUT_NAME}")
+
+if(COMMAND set_ie_threading_interface_for)
+  set_ie_threading_interface_for("${CLDNN_BUILD__PROJ}")
+endif()
+
+target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE
+    clDNN_OpenCL
+    openvino::itt
+  )
+
+if(WIN32)
+  target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE setupapi)
+elseif((NOT ANDROID) AND (UNIX))
+  target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE pthread)
+endif()
+target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE ${CLDNN__SYSTEM_LINK_LIBRARIES})
+
+# ======================================================================================================
diff --git a/inference-engine/thirdparty/clDNN/src/include/cldnn_itt.h b/inference-engine/thirdparty/clDNN/runtime/cldnn_itt.hpp
similarity index 92%
rename from inference-engine/thirdparty/clDNN/src/include/cldnn_itt.h
rename to inference-engine/thirdparty/clDNN/runtime/cldnn_itt.hpp
index b28db96f29567f..e3ab2d3ddb49fd 100644
--- a/inference-engine/thirdparty/clDNN/src/include/cldnn_itt.h
+++ b/inference-engine/thirdparty/clDNN/runtime/cldnn_itt.hpp
@@ -4,7 +4,7 @@
 
 /**
  * @brief Defines openvino domains for tracing
- * @file cldnn_itt.h
+ * @file cldnn_itt.hpp
  */
 
 #pragma once
diff --git a/inference-engine/thirdparty/clDNN/runtime/device_query.cpp b/inference-engine/thirdparty/clDNN/runtime/device_query.cpp
new file mode 100644
index 00000000000000..46f2de82c8c800
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/device_query.cpp
@@ -0,0 +1,30 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "cldnn/runtime/device_query.hpp"
+#include "ocl/ocl_device_detector.hpp"
+
+#include <map>
+#include <string>
+
+namespace cldnn {
+
+device_query::device_query(engine_types engine_type, runtime_types runtime_type, void* user_context, void* user_device) {
+    switch (engine_type) {
+    case engine_types::ocl: {
+        if (runtime_type != runtime_types::ocl)
+            throw std::runtime_error("Unsupported runtime type for ocl engine");
+
+        ocl::ocl_device_detector ocl_detector;
+        _available_devices = ocl_detector.get_available_devices(user_context, user_device);
+        break;
+    }
+    default: throw std::runtime_error("Unsupported engine type in device_query");
+    }
+
+    if (_available_devices.empty()) {
+        throw std::runtime_error("No suitable devices found for requested engine and runtime types");
+    }
+}
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/engine.cpp b/inference-engine/thirdparty/clDNN/runtime/engine.cpp
new file mode 100644
index 00000000000000..561f38f00f4231
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/engine.cpp
@@ -0,0 +1,160 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/event.hpp"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/stream.hpp"
+#include "cldnn/runtime/device_query.hpp"
+
+#include "ocl/ocl_engine_factory.hpp"
+
+#include <string>
+#include <vector>
+#include <memory>
+#include <set>
+#include <stdexcept>
+
+namespace cldnn {
+
+engine::engine(const device::ptr device, const engine_configuration& configuration)
+: _memory_pool(new memory_pool(*this))
+, _device(device)
+, _configuration(configuration) {}
+
+device_info engine::get_device_info() const {
+    return _device->get_info();
+}
+
+const device::ptr engine::get_device() const {
+    return _device;
+}
+
+bool engine::use_unified_shared_memory() const {
+    if (_device->get_mem_caps().supports_usm() && _configuration.use_unified_shared_memory) {
+        return true;
+    }
+    return false;
+}
+
+bool engine::supports_allocation(allocation_type type) const {
+    if (memory_capabilities::is_usm_type(type) && !use_unified_shared_memory())
+        return false;
+    if (allocation_type::usm_shared == type)
+        return false;
+    return _device->get_mem_caps().support_allocation_type(type);
+}
+
+allocation_type engine::get_lockable_preffered_memory_allocation_type(bool is_image_layout) const {
+    if (!use_unified_shared_memory() || is_image_layout)
+        return get_default_allocation_type();
+
+    /*
+        We do not check device allocation here.
+        Device allocation is reserved for buffers of hidden layers.
+        Const buffers are propagated to device if possible.
+    */
+
+    bool support_usm_host = supports_allocation(allocation_type::usm_host);
+    bool support_usm_shared = supports_allocation(allocation_type::usm_shared);
+
+    if (support_usm_shared)
+        return allocation_type::usm_shared;
+    if (support_usm_host)
+        return allocation_type::usm_host;
+
+    throw std::runtime_error("[clDNN internal error] Could not find proper allocation type!");
+}
+
+memory::ptr engine::get_memory_from_pool(const layout& layout,
+                                         primitive_id id,
+                                         uint32_t network_id,
+                                         std::set<primitive_id> dependencies,
+                                         allocation_type type,
+                                         bool reusable) {
+    if (_configuration.use_memory_pool)
+        return _memory_pool->get_memory(layout, id, network_id, dependencies, type, reusable);
+    return _memory_pool->get_memory(layout, type);
+}
+
+memory::ptr engine::attach_memory(const layout& layout, void* ptr) {
+    return std::make_shared<simple_attached_memory>(layout, ptr);
+}
+
+memory::ptr engine::allocate_memory(const layout& layout, bool reset) {
+    allocation_type type = get_lockable_preffered_memory_allocation_type(layout.format.is_image_2d());
+    return allocate_memory(layout, type, reset);
+}
+
+memory_ptr engine::share_buffer(const layout& layout, shared_handle buf) {
+    shared_mem_params params = { shared_mem_type::shared_mem_buffer, nullptr, nullptr, buf,
+#ifdef _WIN32
+        nullptr,
+#else
+        0,
+#endif
+        0 };
+    return reinterpret_handle(layout, params);
+}
+
+memory::ptr engine::share_image(const layout& layout, shared_handle img) {
+    shared_mem_params params = { shared_mem_type::shared_mem_image, nullptr, nullptr, img,
+#ifdef _WIN32
+        nullptr,
+#else
+        0,
+#endif
+        0 };
+    return reinterpret_handle(layout, params);
+}
+
+#ifdef _WIN32
+memory_ptr engine::share_surface(const layout& layout, shared_handle surf, uint32_t plane) {
+    shared_mem_params params = { shared_mem_type::shared_mem_vasurface, nullptr, nullptr, nullptr, surf, plane };
+    return reinterpret_handle(layout, params);
+}
+
+memory_ptr engine::share_dx_buffer(const layout& layout, shared_handle res) {
+    shared_mem_params params = { shared_mem_type::shared_mem_dxbuffer, nullptr, nullptr, res, nullptr, 0 };
+    return reinterpret_handle(layout, params);
+}
+#else
+memory_ptr engine::share_surface(const layout& layout, shared_surface surf, uint32_t plane) {
+    shared_mem_params params = { shared_mem_type::shared_mem_vasurface, nullptr, nullptr, nullptr, surf, plane };
+    return reinterpret_handle(layout, params);
+}
+#endif  // _WIN32
+
+memory_pool& engine::get_memory_pool() {
+    return *_memory_pool.get();
+}
+
+uint64_t engine::get_max_used_device_memory() const {
+    return _memory_pool->get_max_peak_device_memory_used();
+}
+
+uint64_t engine::get_used_device_memory() const {
+    return _memory_pool->get_temp_memory_used();
+}
+
+std::shared_ptr<cldnn::engine> engine::create(engine_types engine_type,
+                                              runtime_types runtime_type,
+                                              const device::ptr device,
+                                              const engine_configuration& configuration) {
+    switch (engine_type) {
+        case engine_types::ocl: return ocl::create_ocl_engine(device, runtime_type, configuration);
+        default: throw std::runtime_error("Invalid engine type");
+    }
+}
+
+std::shared_ptr<cldnn::engine> engine::create(engine_types engine_type,
+                                              runtime_types runtime_type,
+                                              const engine_configuration& configuration) {
+    device_query query(engine_type, runtime_type);
+    device::ptr default_device = query.get_available_devices().begin()->second;
+
+    return engine::create(engine_type, runtime_type, default_device, configuration);
+}
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/event.cpp b/inference-engine/thirdparty/clDNN/runtime/event.cpp
new file mode 100644
index 00000000000000..b9a61138cf12bf
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/event.cpp
@@ -0,0 +1,69 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "cldnn/runtime/event.hpp"
+#include "cldnn/runtime/engine.hpp"
+
+#include <list>
+#include <vector>
+#include <algorithm>
+
+namespace cldnn {
+
+void event::wait() {
+    if (_set)
+        return;
+
+    // TODO: refactor in context of multiple simultaneous calls (for generic engine)
+    wait_impl();
+    _set = true;
+    return;
+}
+
+bool event::is_set() {
+    if (_set)
+        return true;
+
+    // TODO: refactor in context of multiple simultaneous calls (for generic engine)
+    _set = is_set_impl();
+    return _set;
+}
+
+bool event::add_event_handler(event_handler handler, void* data) {
+    if (is_set()) {
+        handler(data);
+        return true;
+    }
+
+    std::lock_guard<std::mutex> lock(_handlers_mutex);
+    auto itr = _handlers.insert(_handlers.end(), {handler, data});
+    auto ret = add_event_handler_impl(handler, data);
+    if (!ret)
+        _handlers.erase(itr);
+
+    return ret;
+}
+
+std::vector<instrumentation::profiling_interval> event::get_profiling_info() {
+    if (!_profiling_captured) {
+        _profiling_captured = get_profiling_info_impl(_profiling_info);
+    }
+
+    std::vector<instrumentation::profiling_interval> result(_profiling_info.size());
+    std::copy(_profiling_info.begin(), _profiling_info.end(), result.begin());
+    return result;
+}
+
+void event::call_handlers() {
+    std::lock_guard<std::mutex> lock(_handlers_mutex);
+    for (auto& pair : _handlers) {
+        try {
+            pair.first(pair.second);
+        } catch (...) {
+        }
+    }
+    _handlers.clear();
+}
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/kernels_cache.cpp b/inference-engine/thirdparty/clDNN/runtime/kernels_cache.cpp
similarity index 82%
rename from inference-engine/thirdparty/clDNN/src/gpu/kernels_cache.cpp
rename to inference-engine/thirdparty/clDNN/runtime/kernels_cache.cpp
index 1451d68de5f402..014486841f04bc 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/kernels_cache.cpp
+++ b/inference-engine/thirdparty/clDNN/runtime/kernels_cache.cpp
@@ -2,9 +2,10 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "kernels_cache.h"
-#include "ocl_toolkit.h"
+#include "kernels_factory.hpp"
+#include "kernels_cache.hpp"
+#include "ocl/ocl_engine.hpp"
+
 #include <algorithm>
 #include <cassert>
 #include <sstream>
@@ -13,8 +14,8 @@
 #include <string>
 #include <memory>
 #include <utility>
-#include "kernel_selector_helper.h"
-#include "cldnn_itt.h"
+
+#include "cldnn_itt.hpp"
 #if (CLDNN_THREADING == CLDNN_THREADING_TBB)
 #include <tbb/parallel_for.h>
 #include <tbb/blocked_range.h>
@@ -137,10 +138,11 @@ inline bool does_options_support_batch_compilation(const std::string& options) {
 }  // namespace
 
 namespace cldnn {
-namespace gpu {
+
+std::mutex kernels_cache::_mutex;
 
 std::string kernels_cache::get_cache_path() const {
-    auto path = _context.get_configuration().kernels_cache_path;
+    auto path = _engine.configuration().kernels_cache_path;
     if (path.empty()) {
         return {};
     }
@@ -152,7 +154,7 @@ std::string kernels_cache::get_cache_path() const {
 }
 
 bool kernels_cache::is_cache_enabled() const {
-    return !_context.get_configuration().kernels_cache_path.empty();
+    return !_engine.configuration().kernels_cache_path.empty();
 }
 
 size_t kernels_cache::get_max_kernels_per_batch() const {
@@ -171,7 +173,6 @@ void kernels_cache::get_program_source(const kernels_code& kernels_source_code,
         std::string options = code.kernel_strings->options;
         bool batch_compilation = code.kernel_strings->batch_compilation;
         bool dump_custom_program = code.dump_custom_program;
-        bool one_time_kernel = code.one_time_kernel;
 
         batch_compilation &= does_options_support_batch_compilation(options);
 
@@ -189,9 +190,6 @@ void kernels_cache::get_program_source(const kernels_code& kernels_source_code,
             key += " __DUMP_CUSTOM_PROGRAM__";  // Adding label to key so it would be separated from other programs
         }
 
-        if (one_time_kernel) {
-            key += " __ONE_TIME__";
-        }
         auto& current_bucket = program_buckets[key];
         if (current_bucket.empty()) { // new bucket
             const auto& bucket_id = program_buckets.size() - 1;
@@ -212,7 +210,6 @@ void kernels_cache::get_program_source(const kernels_code& kernels_source_code,
 
         auto& current_batch = current_bucket.back();
         current_batch.dump_custom_program = dump_custom_program;
-        current_batch.one_time = one_time_kernel;
         current_batch.entry_point_to_id[entry_point] = code.id;
 
         assert(org_source_code.size() == 1);
@@ -230,7 +227,7 @@ void kernels_cache::get_program_source(const kernels_code& kernels_source_code,
         auto options = c.first;
         auto& batches = c.second;
         for (auto& b : batches) {
-            std::string full_code = options + " " + _context.get_device_info().driver_version;
+            std::string full_code = options + " " + _engine.get_device_info().driver_version;
             for (auto& ss : b.source)
                 full_code += ss;
             b.hash_value = std::hash<std::string>()(full_code);
@@ -239,20 +236,17 @@ void kernels_cache::get_program_source(const kernels_code& kernels_source_code,
     }
 }
 
-kernels_cache::kernels_cache(gpu_toolkit& context, uint32_t prog_id) : _context(context), _prog_id(prog_id) {
-}
+kernels_cache::kernels_cache(engine& engine) : _engine(engine) { }
 
-kernels_cache::kernel_id kernels_cache::set_kernel_source(
-    const std::shared_ptr<kernel_selector::kernel_string>& kernel_string,
-    bool dump_custom_program,
-    bool one_time_kernel) {
-    std::lock_guard<std::mutex> lock(_context.get_cache_mutex());
+kernel_id kernels_cache::set_kernel_source(
+    const std::shared_ptr<kernel_string>& kernel_string,
+    bool dump_custom_program) {
 
     // we need unique id in order to avoid conflict across topologies.
     const auto kernel_num = _kernels.size() + _kernels_code.size();
-    kernels_cache::kernel_id id = kernel_string->entry_point + "_" + std::to_string(kernel_num);
+    kernel_id id = kernel_string->entry_point + "_" + std::to_string(kernel_num);
 
-    auto res = _kernels_code.emplace(kernel_string, id, dump_custom_program, one_time_kernel);
+    auto res = _kernels_code.emplace(kernel_string, id, dump_custom_program);
 
     assert(_kernels.find(id) == _kernels.end());
     if (res.second) {
@@ -277,16 +271,19 @@ static std::vector<unsigned char> getProgramBinaries(cl::Program program) {
     return program.getInfo<CL_PROGRAM_BINARIES>().front();
 }
 
-void kernels_cache::build_batch(const batch_program& batch) {
-    OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "KernelsCache::BuildProgram");
+// TODO: This build_batch method should be backend specific
+void kernels_cache::build_batch(const engine& build_engine, const batch_program& batch) {
+    OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "KernelsCache::build_batch");
 
-    bool dump_sources = !_context.get_configuration().ocl_sources_dumps_dir.empty() || batch.dump_custom_program;
+    auto& cl_build_engine = dynamic_cast<const ocl::ocl_engine&>(build_engine);
+
+    bool dump_sources = !_engine.configuration().sources_dumps_dir.empty() || batch.dump_custom_program;
 
     std::string err_log;  // accumulated build log from all program's parts (only contains messages from parts which
 
     std::string current_dump_file_name = "";
     if (dump_sources) {
-        current_dump_file_name = _context.get_configuration().ocl_sources_dumps_dir;
+        current_dump_file_name = _engine.configuration().sources_dumps_dir;
         if (!current_dump_file_name.empty() && current_dump_file_name.back() != '/')
             current_dump_file_name += '/';
 
@@ -318,10 +315,10 @@ void kernels_cache::build_batch(const batch_program& batch) {
 
         // Run compilation
         if (precompiled_kernels.empty()) {
-            cl::Program program(_context.context(), batch.source);
+            cl::Program program(cl_build_engine.get_cl_context(), batch.source);
             {
                 OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "KernelsCache::BuildProgram::RunCompilation");
-                program.build(_context.device(), batch.options.c_str());
+                program.build(cl_build_engine.get_cl_device(), batch.options.c_str());
             }
 
             if (dump_sources && dump_file.good()) {
@@ -342,23 +339,21 @@ void kernels_cache::build_batch(const batch_program& batch) {
                 saveBinaryToFile(cached_bin_name, getProgramBinaries(program));
             }
         } else {
-            cl::Program program(_context.context(), {_context.device()}, precompiled_kernels);
-            program.build(_context.device(), batch.options.c_str());
+            cl::Program program(cl_build_engine.get_cl_context(), {cl_build_engine.get_cl_device()}, precompiled_kernels);
+            program.build(cl_build_engine.get_cl_device(), batch.options.c_str());
             program.createKernels(&kernels);
         }
         {
-            std::lock_guard<std::mutex> lock(_context.get_cache_mutex());
+            std::lock_guard<std::mutex> lock(_mutex);
             for (auto& k : kernels) {
                 const auto& entry_point = k.getInfo<CL_KERNEL_FUNCTION_NAME>();
                 const auto& k_id = batch.entry_point_to_id.find(entry_point);
-                const auto& k_type = kernel_type(k, _context.get_device_info().supports_usm);
                 if (k_id != batch.entry_point_to_id.end()) {
-                    const auto& kmap = std::make_pair(k_id->second, k_type);
-                    if (batch.one_time) {
-                        _one_time_kernels.insert(kmap);
-                    } else {
-                        _kernels.insert(kmap);
-                    }
+                    cl_kernel kern = k.get();
+                    cl_context context = cl_build_engine.get_cl_context().get();
+                    kernel::ptr kernel = kernels_factory::create(_engine, context, kern, entry_point);
+                    const auto& kmap = std::make_pair(k_id->second, kernel);
+                    _kernels.insert(kmap);
                 } else {
                     throw std::runtime_error("Could not find entry point");
                 }
@@ -381,13 +376,12 @@ void kernels_cache::build_batch(const batch_program& batch) {
     }
 }
 
-kernels_cache::kernel_type kernels_cache::get_kernel(kernel_id id, bool one_time_kernel) const {
+kernel::ptr kernels_cache::get_kernel(kernel_id id) const {
     if (_pending_compilation)
         throw std::runtime_error("Kernel cache is not compiled, call build_all() first!");
 
-    const auto& kernels = one_time_kernel ?  _one_time_kernels : _kernels;
-    auto res = kernels.find(id);
-    if (kernels.end() == res)
+    auto res = _kernels.find(id);
+    if (_kernels.end() == res)
         throw std::runtime_error("Kernel " + id + " not found in the kernel cache!");
     return res->second;
 }
@@ -396,46 +390,50 @@ void kernels_cache::build_all() {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "KernelsCache::BuildAll");
     if (!_pending_compilation)
         return;
+
+    std::unique_ptr<ocl::ocl_engine> _build_engine = nullptr;
+    if (_engine.type() == engine_types::ocl) {
+        _build_engine = std::unique_ptr<ocl::ocl_engine>(new ocl::ocl_engine(_engine.get_device(), runtime_types::ocl, _engine.configuration()));
+    }
     std::vector<batch_program> batches;
     {
-        std::lock_guard<std::mutex> lock(_context.get_cache_mutex());
+        std::lock_guard<std::mutex> lock(_mutex);
         get_program_source(_kernels_code, &batches);
-        _one_time_kernels.clear();
 #if (CLDNN_THREADING == CLDNN_THREADING_TBB)
-        int n_threads = _context.get_configuration().n_threads;
+        int n_threads = _engine.configuration().n_threads;
         arena = std::unique_ptr<tbb::task_arena>(new tbb::task_arena());
         arena->initialize(n_threads);
 #elif(CLDNN_THREADING == CLDNN_THREADING_THREADPOOL)
-        int n_threads = _context.get_configuration().n_threads;
+        int n_threads = _engine.configuration().n_threads;
         pool = std::unique_ptr<thread_pool>(new thread_pool(n_threads));
 #endif
     }
 
 #if (CLDNN_THREADING == CLDNN_THREADING_TBB)
-    arena->execute([this, &batches] {
-        tbb::parallel_for(tbb::blocked_range<size_t>(0, batches.size()), [this, &batches](const tbb::blocked_range<size_t>& r) {
+    arena->execute([this, &_build_engine, &batches] {
+        tbb::parallel_for(tbb::blocked_range<size_t>(0, batches.size()), [this, &_build_engine, &batches](const tbb::blocked_range<size_t>& r) {
             for (auto i = r.begin(); i != r.end(); ++i) {
-                build_batch(batches[i]);
+                build_batch(*_build_engine, batches[i]);
             }
         });
     });
 #elif(CLDNN_THREADING == CLDNN_THREADING_THREADPOOL)
     std::vector<std::future<void>> builds;
     for (size_t i = 0; i < batches.size(); ++i) {
-        builds.push_back(pool->enqueue([this, &batches, i] () {
-            build_batch(batches[i]);
+        builds.push_back(pool->enqueue([this, &_build_engine, &batches, i] () {
+            build_batch(*_build_engine, batches[i]);
         }));
     }
     std::for_each(builds.begin(), builds.end(), [] (std::future<void>& f) { f.wait(); });
 #else
     // no parallel build
     for (const auto& batch : batches) {
-        build_batch(batch);
+        build_batch(*_build_engine, batch);
     }
 #endif
 
     {
-        std::lock_guard<std::mutex> lock(_context.get_cache_mutex());
+        std::lock_guard<std::mutex> lock(_mutex);
         _kernels_code.clear();
         _pending_compilation = false;
 #if (CLDNN_THREADING == CLDNN_THREADING_TBB)
@@ -459,9 +457,8 @@ void kernels_cache::build_all() {
 
 void kernels_cache::reset() {
     _kernels.clear();
-    _one_time_kernels.clear();
     _kernels_code.clear();
     _pending_compilation = false;
 }
-}  // namespace gpu
+
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/kernels_cache.h b/inference-engine/thirdparty/clDNN/runtime/kernels_cache.hpp
similarity index 70%
rename from inference-engine/thirdparty/clDNN/src/gpu/kernels_cache.h
rename to inference-engine/thirdparty/clDNN/runtime/kernels_cache.hpp
index 7cf31598bbf01a..931d267e955513 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/kernels_cache.h
+++ b/inference-engine/thirdparty/clDNN/runtime/kernels_cache.hpp
@@ -2,8 +2,11 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
+
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/kernel.hpp"
+
 #include <map>
 #include <mutex>
 #include <vector>
@@ -11,7 +14,6 @@
 #include <atomic>
 #include <string>
 #include <unordered_set>
-#include <kernel_selector_common.h>
 
 #define CLDNN_THREADING_SEQ 0
 #define CLDNN_THREADING_TBB 1
@@ -26,19 +28,8 @@
 #include <condition_variable>
 #endif
 
-namespace cl {
-class Kernel;
-class KernelIntel;
-}
-
-namespace kernel_selector {
-using kernel_string = kernel_selector::KernelString;
-}
-
 namespace cldnn {
-namespace gpu {
 
-class gpu_toolkit;
 #if (CLDNN_THREADING == CLDNN_THREADING_THREADPOOL)
 class thread_pool {
 public:
@@ -101,6 +92,7 @@ class thread_pool {
     }
 };
 #endif
+
 class kernels_cache {
 public:
     using source_code = std::vector<std::string>;
@@ -112,24 +104,20 @@ class kernels_cache {
         uint32_t kernels_counter = 0;
         std::string options;
         bool dump_custom_program = false;
-        bool one_time = false;
         std::map<std::string, std::string> entry_point_to_id;
     };
 
     struct kernel_code {
-        std::shared_ptr<kernel_selector::kernel_string> kernel_strings;
+        std::shared_ptr<kernel_string> kernel_strings;
         std::string id;
         bool dump_custom_program;
-        bool one_time_kernel;
 
-        kernel_code(const std::shared_ptr<kernel_selector::kernel_string>& _kernel_strings,
+        kernel_code(const std::shared_ptr<kernel_string>& _kernel_strings,
                     const std::string& _id,
-                    bool _dump_custom_program,
-                    bool _one_time_kernel)
+                    bool _dump_custom_program)
             : kernel_strings(_kernel_strings),
               id(_id),
-              dump_custom_program(_dump_custom_program),
-              one_time_kernel(_one_time_kernel) {}
+              dump_custom_program(_dump_custom_program) {}
 
         bool operator == (const kernel_code& c2) const {
             return kernel_strings->get_hash() == c2.kernel_strings->get_hash();
@@ -142,44 +130,35 @@ class kernels_cache {
         }
     };
 
-    typedef std::string kernel_id;
-    typedef cl::KernelIntel kernel_type;
-    using kernels_map = std::map<std::string, kernel_type>;
     using kernels_code = std::unordered_set<kernel_code, hash_kernel_code>;
 
 private:
-    gpu_toolkit& _context;
+    static std::mutex _mutex;
+    engine& _engine;
     kernels_code _kernels_code;
     std::atomic<bool> _pending_compilation{false};
-    std::map<const std::string, const kernel_type> _kernels;
-    std::map<const std::string, const kernel_type> _one_time_kernels;  // These kernels are intended to be executed only once (can
-                                                           // be removed later from the cache).
-    uint32_t _prog_id;
+    std::map<const std::string, kernel::ptr> _kernels;
 #if (CLDNN_THREADING == CLDNN_THREADING_TBB)
     std::unique_ptr<tbb::task_arena> arena;
 #elif(CLDNN_THREADING == CLDNN_THREADING_THREADPOOL)
     std::unique_ptr<thread_pool> pool;
 #endif
 
-
     void get_program_source(const kernels_code& kernels_source_code, std::vector<batch_program>*) const;
-    void build_batch(const batch_program& batch);
+    void build_batch(const engine& build_engine, const batch_program& batch);
 
     std::string get_cache_path() const;
     bool is_cache_enabled() const;
     size_t get_max_kernels_per_batch() const;
 
 public:
-    explicit kernels_cache(gpu_toolkit& context, uint32_t prog_id);
-    kernel_id set_kernel_source(const std::shared_ptr<kernel_selector::kernel_string>& kernel_string,
-                                bool dump_custom_program,
-                                bool one_time_kernel);
-    kernel_type get_kernel(kernel_id id, bool one_time_kernel) const;
-    gpu_toolkit& get_context() { return _context; }
+    explicit kernels_cache(engine& engine);
+    kernel_id set_kernel_source(const std::shared_ptr<kernel_string>& kernel_string,
+                                bool dump_custom_program);
+    kernel::ptr get_kernel(kernel_id id) const;
     // forces compilation of all pending kernels/programs
     void build_all();
     void reset();
 };
 
-}  // namespace gpu
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/kernels_factory.cpp b/inference-engine/thirdparty/clDNN/runtime/kernels_factory.cpp
new file mode 100644
index 00000000000000..1cb366518a43aa
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/kernels_factory.cpp
@@ -0,0 +1,22 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "kernels_factory.hpp"
+
+namespace cldnn {
+namespace ocl {
+std::shared_ptr<kernel> create_ocl_kernel(engine& engine, cl_context context, cl_kernel kernel, std::string  entry_point);
+}  // namespace ocl
+
+namespace kernels_factory {
+
+std::shared_ptr<kernel> create(engine& engine, cl_context context, cl_kernel kernel, std::string  entry_point) {
+    switch (engine.type()) {
+        case engine_types::ocl: return ocl::create_ocl_kernel(engine, context, kernel, entry_point);
+        default: throw std::runtime_error("Unsupported engine type in kernels_factory::create");
+    }
+}
+
+}  // namespace kernels_factory
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/kernels_factory.hpp b/inference-engine/thirdparty/clDNN/runtime/kernels_factory.hpp
new file mode 100644
index 00000000000000..b3d03defcb285b
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/kernels_factory.hpp
@@ -0,0 +1,22 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "cldnn/runtime/kernel.hpp"
+#include "cldnn/runtime/engine.hpp"
+#include "ocl/ocl_common.hpp"
+
+#include <memory>
+
+namespace cldnn {
+
+namespace kernels_factory {
+
+// Creates instance of kernel for selected engine type.
+// For ocl engine it creates a copy of kernel object
+std::shared_ptr<kernel> create(engine& engine, cl_context context, cl_kernel kernel, kernel_id kernel_id);
+
+}  // namespace kernels_factory
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/memory.cpp b/inference-engine/thirdparty/clDNN/runtime/memory.cpp
new file mode 100644
index 00000000000000..1a6ff65795b461
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/memory.cpp
@@ -0,0 +1,36 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/stream.hpp"
+
+#include "ocl/ocl_memory.hpp"
+
+#include <string>
+#include <vector>
+#include <memory>
+#include <set>
+#include <stdexcept>
+
+namespace cldnn {
+
+memory::memory(engine* engine, const layout& layout, allocation_type type, bool reused)
+    : _engine(engine), _layout(layout), _bytes_count(_layout.bytes_count()), _type(type), _reused(reused) {}
+
+memory::~memory() {
+    if (!_reused && _engine) {
+        // TODO: Make memory usage tracker static in memory class
+        _engine->get_memory_pool().subtract_memory_used(_bytes_count);
+    }
+}
+
+std::unique_ptr<surfaces_lock> surfaces_lock::create(engine_types engine_type, std::vector<memory::ptr> mem, const stream& stream) {
+    switch (engine_type) {
+    case engine_types::ocl: return std::unique_ptr<ocl::ocl_surfaces_lock>(new ocl::ocl_surfaces_lock(mem, stream));
+    default: throw std::runtime_error("Unsupported engine type in surfaces_lock::create");
+    }
+}
+
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/memory_pool.cpp b/inference-engine/thirdparty/clDNN/runtime/memory_pool.cpp
similarity index 55%
rename from inference-engine/thirdparty/clDNN/src/memory_pool.cpp
rename to inference-engine/thirdparty/clDNN/runtime/memory_pool.cpp
index f80e8e9cd986bb..85091ccdd18e88 100644
--- a/inference-engine/thirdparty/clDNN/src/memory_pool.cpp
+++ b/inference-engine/thirdparty/clDNN/runtime/memory_pool.cpp
@@ -2,20 +2,14 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-
 #include <algorithm>
 #include <fstream>
 #include <vector>
 
-#include "memory_impl.h"
-#include "memory_pool.h"
-#include "engine_impl.h"
-#include "program_impl.h"
-
-#include "program_node.h"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/memory_pool.hpp"
 
-#include "gpu/memory_gpu.h"
 #include <list>
 #include <string>
 #include <utility>
@@ -24,88 +18,17 @@
 
 namespace cldnn {
 memory_record::memory_record(memory_set users,
-                             refcounted_obj_ptr<memory_impl>& memory,
+                             std::shared_ptr<memory>& memory,
                              uint32_t net_id,
                              allocation_type type)
     : _users(users), _memory(memory), _network_id(net_id), _type(type) {}
 
-memory_impl::ptr memory_pool::alloc_memory(const layout& layout, allocation_type type, uint32_t net_id, bool reset) {
-    auto context = _engine->get_context();
-    if (layout.bytes_count() > context->get_device_info().max_alloc_mem_size) {
-        throw std::runtime_error("exceeded max size of memory object allocation");
-    }
-
-    add_memory_used(layout.bytes_count());
-
-    if (_max_peak_memory_used > context->get_device_info().max_global_mem_size) {
+memory::ptr memory_pool::alloc_memory(const layout& layout, allocation_type type) {
+    if (_max_peak_memory_used > _engine->get_device_info().max_global_mem_size) {
         throw std::runtime_error("exceeded global device memory");
     }
 
-    try {
-        if (layout.format.is_image_2d()) {
-            memory_impl::ptr mem_impl {new gpu::gpu_image2d(engine_impl::ptr(_engine), layout, net_id, reset), false};
-            return mem_impl;
-        } else if (type == allocation_type::cl_mem) {
-            memory_impl::ptr mem_impl{ new gpu::gpu_buffer(engine_impl::ptr(_engine), layout, net_id, reset), false };
-            return mem_impl;
-        } else {
-            memory_impl::ptr mem_impl{ new gpu::gpu_usm(engine_impl::ptr(_engine), layout, net_id, type, reset), false };
-            return mem_impl;
-        }
-    } catch (const cl::Error& clErr) {
-        switch (clErr.err()) {
-            case CL_MEM_OBJECT_ALLOCATION_FAILURE:
-            case CL_OUT_OF_RESOURCES:
-            case CL_OUT_OF_HOST_MEMORY:
-            case CL_INVALID_BUFFER_SIZE:
-                throw std::runtime_error("out of GPU resources");
-            default:
-                throw std::runtime_error("GPU buffer allocation failed");
-        }
-    }
-}
-
-memory_impl::ptr memory_pool::get_memory(const layout& layout, const shared_mem_params* params, uint32_t net_id) {
-    try {
-        if (layout.format.is_image_2d() && params->mem_type == shared_mem_type::shared_mem_image) {
-            cl::Image2D img(static_cast<cl_mem>(params->mem), true);
-            memory_impl::ptr mem_impl{ new gpu::gpu_image2d(engine_impl::ptr(_engine), layout,
-                img,
-                net_id), false };
-            return mem_impl;
-        } else if (layout.format.is_image_2d() && params->mem_type == shared_mem_type::shared_mem_vasurface) {
-            memory_impl::ptr mem_impl{ new gpu::gpu_media_buffer(engine_impl::ptr(_engine), layout,
-                params,
-                net_id), false };
-            return mem_impl;
-#ifdef _WIN32
-        } else if (params->mem_type == shared_mem_type::shared_mem_dxbuffer) {
-            memory_impl::ptr mem_impl{ new gpu::gpu_dx_buffer(engine_impl::ptr(_engine), layout,
-                params,
-                net_id), false };
-            return mem_impl;
-#endif
-        } else if (params->mem_type == shared_mem_type::shared_mem_buffer) {
-            cl::Buffer buf(static_cast<cl_mem>(params->mem), true);
-            memory_impl::ptr mem_impl{ new gpu::gpu_buffer(engine_impl::ptr(_engine), layout,
-                buf,
-                net_id), false };
-            return mem_impl;
-        } else {
-            throw std::runtime_error("unknown shared object fromat or type");
-        }
-    }
-    catch (const cl::Error& clErr) {
-        switch (clErr.err()) {
-        case CL_MEM_OBJECT_ALLOCATION_FAILURE:
-        case CL_OUT_OF_RESOURCES:
-        case CL_OUT_OF_HOST_MEMORY:
-        case CL_INVALID_BUFFER_SIZE:
-            throw std::runtime_error("out of GPU resources");
-        default:
-            throw std::runtime_error("GPU buffer allocation failed");
-        }
-    }
+    return _engine->allocate_memory(layout, type);
 }
 
 memory_pool::~memory_pool() {}
@@ -129,12 +52,10 @@ bool memory_pool::has_conflict(const memory_set& a,
     return !intersection.empty();
 }
 
-void memory_pool::release_memory(memory_impl* mem,
-    const primitive_id& id) {
+void memory_pool::release_memory(memory* mem, const primitive_id& id, uint32_t network_id) {
     // check nonpadded pool first
     auto _layout = mem->get_layout();
     auto type = mem->get_allocation_type();
-    auto network_id = mem->get_net_id();
 
     {
         auto range = _non_padded_pool.equal_range(_layout.bytes_count());
@@ -198,11 +119,11 @@ void memory_pool::release_memory(memory_impl* mem,
     }
 }
 
-memory_impl::ptr memory_pool::get_from_non_padded_pool(const layout& layout,
-                                                       const primitive_id& id,
-                                                       uint32_t network_id,
-                                                       const std::set<primitive_id>& restrictions,
-                                                       allocation_type type) {
+memory::ptr memory_pool::get_from_non_padded_pool(const layout& layout,
+                                                  const primitive_id& id,
+                                                  uint32_t network_id,
+                                                  const std::set<primitive_id>& restrictions,
+                                                  allocation_type type) {
     auto it = _non_padded_pool.lower_bound(layout.bytes_count());
     while (it != _non_padded_pool.end()) {
         if (it->second._network_id == network_id &&
@@ -220,7 +141,7 @@ memory_impl::ptr memory_pool::get_from_non_padded_pool(const layout& layout,
         }
     }
     // didn't find anything for you? create new resource
-    auto mem = alloc_memory(layout, type, network_id);
+    auto mem = alloc_memory(layout, type);
     {
         _non_padded_pool.emplace(layout.bytes_count(),
                                  memory_record({{id, network_id}}, mem, network_id, type));
@@ -228,11 +149,11 @@ memory_impl::ptr memory_pool::get_from_non_padded_pool(const layout& layout,
     return mem;
 }
 
-memory_impl::ptr memory_pool::get_from_padded_pool(const layout& layout,
-                                                   const primitive_id& id,
-                                                   uint32_t network_id,
-                                                   const std::set<primitive_id>& restrictions,
-                                                   allocation_type type) {
+memory::ptr memory_pool::get_from_padded_pool(const layout& layout,
+                                              const primitive_id& id,
+                                              uint32_t network_id,
+                                              const std::set<primitive_id>& restrictions,
+                                              allocation_type type) {
     auto first_level_cache = _padded_pool.find(layout);
 
     if (first_level_cache != _padded_pool.end()) {
@@ -252,12 +173,12 @@ memory_impl::ptr memory_pool::get_from_padded_pool(const layout& layout,
                 return ret_mem;
             }
         }
-        auto mem = alloc_memory(layout, type, network_id);
+        auto mem = alloc_memory(layout, type);
         first_level_cache->second.emplace_back(
             memory_record({{id, network_id}}, mem, network_id, type));
         return mem;
     }
-    auto mem = alloc_memory(layout, type, network_id);
+    auto mem = alloc_memory(layout, type);
     std::list<memory_record> list = {memory_record({{id, network_id}}, mem, network_id, type)};
     _padded_pool.emplace(layout, std::move(list));
     return mem;
@@ -267,10 +188,10 @@ memory_impl::ptr memory_pool::get_from_padded_pool(const layout& layout,
         This is not reusable within one network or it's internal micronetworks. But we can use this memory records
    between networks.
     */
-memory_impl::ptr memory_pool::get_from_across_networks_pool(const layout& layout,
-                                                            const primitive_id& id,
-                                                            uint32_t network_id,
-                                                            allocation_type type) {
+memory::ptr memory_pool::get_from_across_networks_pool(const layout& layout,
+                                                       const primitive_id& id,
+                                                       uint32_t network_id,
+                                                       allocation_type type) {
     auto it = _no_reusable_pool.lower_bound(layout.bytes_count());
 
     while (it != _no_reusable_pool.end()) {
@@ -284,7 +205,7 @@ memory_impl::ptr memory_pool::get_from_across_networks_pool(const layout& layout
         }
         ++it;
     }
-    auto mem = alloc_memory(layout, type, network_id);
+    auto mem = alloc_memory(layout, type);
     {
         _no_reusable_pool.emplace(layout.bytes_count(),
                                   memory_record({{id, network_id}}, mem, network_id, type));
@@ -292,16 +213,16 @@ memory_impl::ptr memory_pool::get_from_across_networks_pool(const layout& layout
     return mem;
 }
 
-memory_impl::ptr memory_pool::get_memory(const layout& layout, allocation_type type, uint32_t net_id, bool reset) {
-    return alloc_memory(layout, type, net_id, reset);
+memory::ptr memory_pool::get_memory(const layout& layout, allocation_type type) {
+    return alloc_memory(layout, type);
 }
 
-memory_impl::ptr memory_pool::get_memory(const layout& layout,
-                                         const primitive_id& id,
-                                         uint32_t network_id,
-                                         const std::set<primitive_id>& restrictions,
-                                         allocation_type type,
-                                         bool reusable_across_network) {
+memory::ptr memory_pool::get_memory(const layout& layout,
+                                    const primitive_id& id,
+                                    uint32_t network_id,
+                                    const std::set<primitive_id>& restrictions,
+                                    allocation_type type,
+                                    bool reusable_across_network) {
     if (reusable_across_network) {
         // reusable within the same network
         if (!layout.format.is_image() && layout.data_padding == padding{{0, 0, 0, 0}, 0}) {
@@ -312,10 +233,10 @@ memory_impl::ptr memory_pool::get_memory(const layout& layout,
             return get_from_padded_pool(layout, id, network_id, restrictions, type);
         } else {
             // images (reuse not yet implemented)
-            return alloc_memory(layout, type, network_id);
+            return alloc_memory(layout, type);
         }
     } else {
-        return alloc_memory(layout, type, network_id);
+        return alloc_memory(layout, type);
     }
 }
 
@@ -329,8 +250,7 @@ void memory_pool::clear_pool_for_network(uint32_t network_id) {
         while (itr != _non_padded_pool.end()) {
             auto& record = itr->second;
 
-            if (record._memory->get_net_id() == network_id &&
-                record._network_id == network_id) {
+            if (record._network_id == network_id) {
                 itr = _non_padded_pool.erase(itr);
             } else {
                 itr++;
@@ -347,8 +267,7 @@ void memory_pool::clear_pool_for_network(uint32_t network_id) {
             auto list_itr = list.begin();
 
             while (list_itr != list.end()) {
-                if (list_itr->_memory->get_net_id() == network_id &&
-                    list_itr->_network_id == network_id) {
+                if (list_itr->_network_id == network_id) {
                     list_itr = list.erase(list_itr);
                 } else {
                     list_itr++;
@@ -370,8 +289,7 @@ void memory_pool::clear_pool_for_network(uint32_t network_id) {
         while (itr != _no_reusable_pool.end()) {
             auto& record = itr->second;
 
-            if (record._memory->get_net_id() == network_id &&
-                record._network_id == network_id) {
+            if (record._network_id == network_id) {
                 itr = _no_reusable_pool.erase(itr);
             } else {
                 itr++;
@@ -380,59 +298,14 @@ void memory_pool::clear_pool_for_network(uint32_t network_id) {
     }
 }
 
-memory_pool::memory_pool(engine_impl& engine) : _engine(&engine), _temp_memory_used(0), _max_peak_memory_used(0) {
-}
-
-void memory_pool::dump_memory_pool(const program_impl& program, std::string& path, std::string& dep) {
-    using namespace std;
-    ofstream log(path);
-
-    log << "\nNon-padded pool:" << endl;
-    log << "Size\tUsers:" << endl;
-    for (const auto& record : _non_padded_pool) {
-        log << record.first;
-        for (const auto& usr : record.second._users) log << ", " << usr;
-        log << endl;
-    }
-
-    log << "\n--- Padded pool: ---" << endl;
-    log << "Size\tUsers:" << endl;
-    for (const auto& record : _padded_pool) {
-        for (const auto& mem : record.second) {
-            log << mem._memory->size();
-            for (const auto& usr : mem._users) log << ", " << usr;
-            log << endl;
-        }
-    }
-    log << dep;
-    log.close();
-    color_graph(program);
-}
-
-void memory_pool::color_graph(const program_impl& program) {
-    uint32_t color = 0;
-    for (const auto& record : _non_padded_pool) {
-        for (const auto& usr : record.second._users) {
-            if (program.has_node(usr._id))
-                program.get_node(usr._id).set_reused_memory_color(color);
-        }
-        ++color;
-    }
-
-    for (const auto& list : _padded_pool) {
-        for (const auto& record : list.second) {
-            if (record._users.size() > 1) {  // one user doesn't mean reusing
-                for (const auto& usr : record._users) {
-                    if (program.has_node(usr._id))
-                        program.get_node(usr._id).set_reused_memory_color(color);
-                }
-            }
-            ++color;
-        }
-    }
-}
+memory_pool::memory_pool(engine& engine) : _engine(&engine), _temp_memory_used(0), _max_peak_memory_used(0) { }
 
 void memory_pool::add_memory_used(size_t value) {
+    // std::cerr << "ADD MEM: " << value
+    //           << " max: " << _engine->get_device_info().max_global_mem_size
+    //           << " peak: " << _max_peak_memory_used
+    //           << " tmp: " << _temp_memory_used << std::endl;
+
     _temp_memory_used += value;
     if (_temp_memory_used > _max_peak_memory_used) {
         _max_peak_memory_used = _temp_memory_used.load();
@@ -441,6 +314,10 @@ void memory_pool::add_memory_used(size_t value) {
 
 void memory_pool::subtract_memory_used(size_t value) {
     _temp_memory_used -= value;
+    // std::cerr << "FREE MEM: " << value
+    //           << " max: " << _engine->get_device_info().max_global_mem_size
+    //           << " peak: " << _max_peak_memory_used
+    //           << " tmp: " << _temp_memory_used << std::endl;
 }
 
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ocl_base_event.cpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_base_event.cpp
similarity index 93%
rename from inference-engine/thirdparty/clDNN/src/gpu/ocl_base_event.cpp
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_base_event.cpp
index 09e2cfda0b9254..e08cfd6fa553e0 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/ocl_base_event.cpp
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_base_event.cpp
@@ -2,14 +2,16 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "ocl_base_event.h"
+#include "ocl_base_event.hpp"
 
 #include <cassert>
 #include <iostream>
 #include <vector>
 #include <list>
+#include <map>
+
 using namespace cldnn;
-using namespace gpu;
+using namespace ocl;
 
 namespace {
 bool is_event_profiled(const cl::Event& event) {
@@ -48,9 +50,6 @@ void base_event::set_ocl_callback() {
 void base_event::wait_impl() {
     if (_event.get() != nullptr) {
         _event.wait();
-        if (get_context()->logging_enabled()) {
-            get_context()->log(0, "Wait for event: " + std::to_string(_queue_stamp));
-        }
     }
 }
 
@@ -92,9 +91,6 @@ bool base_event::get_profiling_info_impl(std::list<instrumentation::profiling_in
 void base_events::wait_impl() {
     if (_last_ocl_event.get() != nullptr) {
         _last_ocl_event.wait();
-        if (get_context()->logging_enabled()) {
-            get_context()->log(0, "Wait for event: " + std::to_string(_queue_stamp));
-        }
     }
 }
 
@@ -112,7 +108,7 @@ bool base_events::get_profiling_info_impl(std::list<instrumentation::profiling_i
     std::map<std::string, std::vector<std::pair<unsigned long long, unsigned long long>>> all_durations;
 
     for (size_t i = 0; i < _events.size(); i++) {
-        auto be = dynamic_cast<base_event*>(_events[i].get());
+        auto be = downcast<base_event>(_events[i].get());
         if (!is_event_profiled(be->_event))
             continue;
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ocl_base_event.h b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_base_event.hpp
similarity index 76%
rename from inference-engine/thirdparty/clDNN/src/gpu/ocl_base_event.h
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_base_event.hpp
index 988628497dad1a..85f32c0b0b6def 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/ocl_base_event.h
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_base_event.hpp
@@ -4,13 +4,16 @@
 
 #pragma once
 
-#include "ocl_toolkit.h"
+#include "ocl_common.hpp"
+#include "cldnn/runtime/event.hpp"
+#include "cldnn/runtime/utils.hpp"
+
 #include <vector>
 #include <memory>
 #include <list>
 
 namespace cldnn {
-namespace gpu {
+namespace ocl {
 
 struct profiling_period_ocl_start_stop {
     const char* name;
@@ -18,7 +21,7 @@ struct profiling_period_ocl_start_stop {
     cl_profiling_info stop;
 };
 
-struct ocl_base_event : virtual public event_impl {
+struct ocl_base_event : virtual public event {
 public:
     explicit ocl_base_event(uint64_t queue_stamp = 0, bool valid = false) : _queue_stamp(queue_stamp) { _attached = valid; }
     uint64_t get_queue_stamp() const { return _queue_stamp; }
@@ -30,10 +33,10 @@ struct ocl_base_event : virtual public event_impl {
 
 struct base_event : virtual public ocl_base_event {
 public:
-    base_event(std::shared_ptr<gpu_toolkit> ctx, cl::Event const& ev, uint64_t queue_stamp = 0)
-        : ocl_base_event(queue_stamp, true), _ctx(ctx), _event(ev) {}
+    base_event(const cl::Context& /* ctx */, cl::Event const& ev, uint64_t queue_stamp = 0)
+        : ocl_base_event(queue_stamp, true), _event(ev) {}
 
-    explicit base_event(std::shared_ptr<gpu_toolkit> ctx) : ocl_base_event(0, false), _ctx(ctx) {}
+    base_event(const cl::Context& /* ctx */) : ocl_base_event(0, false) {}
 
     void attach_ocl_event(const cl::Event& ev, const uint64_t q_stamp) {
         _event = ev;
@@ -42,11 +45,9 @@ struct base_event : virtual public ocl_base_event {
         _set = false;
     }
 
-    std::shared_ptr<gpu_toolkit> get_context() const { return _ctx; }
     cl::Event get() override { return _event; }
 
 private:
-    std::shared_ptr<gpu_toolkit> _ctx;
     bool _callback_set = false;
     void set_ocl_callback();
     static void CL_CALLBACK ocl_event_completion_callback(cl_event, cl_int, void* me);
@@ -65,14 +66,14 @@ struct base_event : virtual public ocl_base_event {
 
 struct base_events : virtual public ocl_base_event {
 public:
-    base_events(std::shared_ptr<gpu_toolkit> ctx, std::vector<event_impl::ptr> const& ev)
-        : ocl_base_event(0, true), _ctx(ctx) {
+    base_events(const cl::Context& /* ctx */, std::vector<event::ptr> const& ev)
+        : ocl_base_event(0, true) {
         process_events(ev);
     }
 
-    explicit base_events(std::shared_ptr<gpu_toolkit> ctx) : ocl_base_event(0, false), _ctx(ctx) {}
+    base_events(const cl::Context& /* ctx */) : ocl_base_event(0, false) {}
 
-    void attach_events(const std::vector<event_impl::ptr>& ev) {
+    void attach_events(const std::vector<event::ptr>& ev) {
         if (_attached)
             throw std::runtime_error("Trying to attach events to valid event object.");
         process_events(ev);
@@ -80,7 +81,6 @@ struct base_events : virtual public ocl_base_event {
     }
 
     cl::Event get() override { return _last_ocl_event; }
-    std::shared_ptr<gpu_toolkit> get_context() const { return _ctx; }
 
     void reset() override {
         ocl_base_event::reset();
@@ -91,7 +91,7 @@ struct base_events : virtual public ocl_base_event {
     void wait_impl() override;
     bool is_set_impl() override;
 
-    void process_events(const std::vector<event_impl::ptr>& ev) {
+    void process_events(const std::vector<event::ptr>& ev) {
         for (size_t i = 0; i < ev.size(); i++) {
             auto multiple_events = dynamic_cast<base_events*>(ev[i].get());
             if (multiple_events) {
@@ -121,9 +121,8 @@ struct base_events : virtual public ocl_base_event {
     bool get_profiling_info_impl(std::list<instrumentation::profiling_interval>& info) override;
 
     cl::Event _last_ocl_event;
-    std::shared_ptr<gpu_toolkit> _ctx;
-    std::vector<event_impl::ptr> _events;
+    std::vector<event::ptr> _events;
 };
 
-}  // namespace gpu
+}  // namespace ocl
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/command_queues_builder.cpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_command_queues_builder.cpp
similarity index 80%
rename from inference-engine/thirdparty/clDNN/src/gpu/command_queues_builder.cpp
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_command_queues_builder.cpp
index 25f842cd4c07d2..c2772434304ccb 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/command_queues_builder.cpp
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_command_queues_builder.cpp
@@ -2,21 +2,15 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "command_queues_builder.h"
-#include "error_handler.h"
+#include "ocl_command_queues_builder.hpp"
+#include "cldnn/runtime/error_handler.hpp"
 #include <string>
 
 namespace cldnn {
-namespace gpu {
-
-command_queues_builder::command_queues_builder(const cl::Context& context,
-                                               const cl::Device& device,
-                                               const cl_platform_id& platform_id)
-    : _context(context),
-      _device(device),
-      _platform_id(platform_id),
-      _profiling(false),
+namespace ocl {
+
+command_queues_builder::command_queues_builder()
+    : _profiling(false),
       _out_of_order(false),
       _priority_mode(priority_mode_types::disabled),
       _throttle_mode(throttle_mode_types::disabled) {}
@@ -27,12 +21,13 @@ cl_command_queue_properties command_queues_builder::get_properties() {
     return ret;
 }
 
-void command_queues_builder::build() {
+ocl_queue_type command_queues_builder::build(const cl::Context& context, const cl::Device& device) {
     auto properties = get_properties();
 
+    ocl_queue_type queue;
+
     if (_priority_mode == priority_mode_types::disabled && _throttle_mode == throttle_mode_types::disabled) {
-        _queue = queue_type(_context, _device, properties);
-        return;
+        queue = ocl_queue_type(context, device, properties);
     }
 
     unsigned cl_queue_priority_value = CL_QUEUE_PRIORITY_MED_KHR;
@@ -72,7 +67,7 @@ void command_queues_builder::build() {
                                                 properties,
                                                 0};
 
-        _queue = queue_type(clCreateCommandQueueWithProperties(_context.get(), _device.get(), properties_low, &error_code));
+        queue = ocl_queue_type(clCreateCommandQueueWithProperties(context.get(), device.get(), properties_low, &error_code));
     } else if (_priority_mode != priority_mode_types::disabled) {
         cl_queue_properties properties_low[] = {CL_QUEUE_PRIORITY_KHR,
                                                 cl_queue_priority_value,
@@ -80,7 +75,7 @@ void command_queues_builder::build() {
                                                 properties,
                                                 0};
 
-        _queue = queue_type(clCreateCommandQueueWithProperties(_context.get(), _device.get(), properties_low, &error_code));
+        queue = ocl_queue_type(clCreateCommandQueueWithProperties(context.get(), device.get(), properties_low, &error_code));
     } else if (_throttle_mode != throttle_mode_types::disabled) {
         cl_queue_properties properties_low[] = {CL_QUEUE_THROTTLE_KHR,
                                                 cl_queue_throttle_value,
@@ -88,13 +83,15 @@ void command_queues_builder::build() {
                                                 properties,
                                                 0};
 
-        _queue = queue_type(clCreateCommandQueueWithProperties(_context.get(), _device.get(), properties_low, &error_code));
+        queue = ocl_queue_type(clCreateCommandQueueWithProperties(context.get(), device.get(), properties_low, &error_code));
     }
 
     if (error_code != CL_SUCCESS) {
         CLDNN_ERROR_MESSAGE("Command queues builders",
                             "clCreateCommandQueueWithPropertiesINTEL error " + std::to_string(error_code));
     }
+
+    return queue;
 }
 
 void command_queues_builder::set_priority_mode(priority_mode_types priority, bool extension_support) {
@@ -115,5 +112,5 @@ void command_queues_builder::set_throttle_mode(throttle_mode_types throttle, boo
     }
     _throttle_mode = throttle;
 }
-}  // namespace gpu
+}  // namespace ocl
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/command_queues_builder.h b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_command_queues_builder.hpp
similarity index 57%
rename from inference-engine/thirdparty/clDNN/src/gpu/command_queues_builder.h
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_command_queues_builder.hpp
index 7d99bcccbd2ecd..cddd350760034b 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/command_queues_builder.h
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_command_queues_builder.hpp
@@ -2,28 +2,24 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "ocl_toolkit.h"
+
+#include "ocl_common.hpp"
+#include "cldnn/runtime/engine.hpp"
 
 namespace cldnn {
-namespace gpu {
+namespace ocl {
+
 class command_queues_builder {
 public:
-    command_queues_builder(const cl::Context& context, const cl::Device& device, const cl_platform_id& platform_id);
-    void build();
+    command_queues_builder();
+    ocl_queue_type build(const cl::Context& context, const cl::Device& device);
     void set_throttle_mode(throttle_mode_types throttle, bool extension_support);
     void set_priority_mode(priority_mode_types priority, bool extension_support);
     void set_profiling(bool flag) { _profiling = flag; }
     void set_out_of_order(bool flag) { _out_of_order = flag; }
-    queue_type& queue() { return _queue; }
-    queue_type queue() const { return _queue; }
 
 private:
-    queue_type _queue;
-    cl::Context _context;
-    cl::Device _device;
-    cl_platform_id _platform_id;
     bool _profiling;
     bool _out_of_order;
     priority_mode_types _priority_mode;
@@ -31,5 +27,6 @@ class command_queues_builder {
 
     cl_command_queue_properties get_properties();
 };
-}  // namespace gpu
+
+}  // namespace ocl
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_common.hpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_common.hpp
new file mode 100644
index 00000000000000..07be51ff38c7d1
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_common.hpp
@@ -0,0 +1,30 @@
+// Copyright (C) 2020-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#pragma once
+
+#include "ocl_wrapper.hpp"
+
+#include <vector>
+
+namespace cldnn {
+namespace ocl {
+
+typedef cl::vector<cl::vector<unsigned char>> kernels_binaries_vector;
+typedef cl::vector<kernels_binaries_vector> kernels_binaries_container;
+typedef CL_API_ENTRY cl_command_queue(CL_API_CALL* pfn_clCreateCommandQueueWithPropertiesINTEL)(
+    cl_context context,
+    cl_device_id device,
+    const cl_queue_properties* properties,
+    cl_int* errcodeRet);
+
+using ocl_queue_type = cl::CommandQueueIntel;
+using ocl_kernel_type = cl::KernelIntel;
+
+class ocl_error : public std::runtime_error {
+public:
+    explicit ocl_error(cl::Error const& err);
+};
+
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/device_info.cpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device.cpp
similarity index 54%
rename from inference-engine/thirdparty/clDNN/src/gpu/device_info.cpp
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device.cpp
index 09e069dda3f776..e1ba4bb5a5ceb7 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/device_info.cpp
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device.cpp
@@ -2,16 +2,19 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "device_info.h"
-#include "include/to_string_utils.h"
+#include "ocl_device.hpp"
+#include "ocl_common.hpp"
+
+#include <map>
+#include <string>
+#include <vector>
+#include <algorithm>
 #include <unordered_map>
 #include <string>
 #include <cassert>
 #include <time.h>
 #include <limits>
 #include <chrono>
-#include "ocl_builder.h"
-
 #include <fstream>
 #include <iostream>
 #include <utility>
@@ -30,7 +33,7 @@
 #endif
 
 namespace cldnn {
-namespace gpu {
+namespace ocl {
 
 namespace {
 int driver_dev_id() {
@@ -103,7 +106,7 @@ int driver_dev_id() {
         return result.back();
 }
 
-static device_type get_device_type(const cl::Device& device) {
+device_type get_device_type(const cl::Device& device) {
     auto unified_mem = device.getInfo<CL_DEVICE_HOST_UNIFIED_MEMORY>();
 
     return unified_mem ? device_type::integrated_gpu : device_type::discrete_gpu;
@@ -117,7 +120,7 @@ gfx_version parse_version(cl_uint ver) {
     return {major, minor, revision};
 }
 
-static bool get_imad_support(const cl::Device& device) {
+bool get_imad_support(const cl::Device& device) {
     std::string dev_name = device.getInfo<CL_DEVICE_NAME>();
 
     if (dev_name.find("Gen12") != std::string::npos ||
@@ -183,74 +186,111 @@ bool is_local_block_io_supported(const cl::Device& device) {
     }
 }
 
-}  // namespace
-
-device_info_internal::device_info_internal(const cl::Device& device) {
-    dev_name = device.getInfo<CL_DEVICE_NAME>();
-    driver_version = device.getInfo<CL_DRIVER_VERSION>();
-    dev_type = get_device_type(device);
+device_info init_device_info(const cl::Device& device) {
+    device_info info;
+    info.vendor_id = static_cast<uint32_t>(device.getInfo<CL_DEVICE_VENDOR_ID>());
+    info.dev_name = device.getInfo<CL_DEVICE_NAME>();
+    info.driver_version = device.getInfo<CL_DRIVER_VERSION>();
+    info.dev_type = get_device_type(device);
 
-    compute_units_count = device.getInfo<CL_DEVICE_MAX_COMPUTE_UNITS>();
+    info.execution_units_count = device.getInfo<CL_DEVICE_MAX_COMPUTE_UNITS>();
 
-    cores_count = static_cast<uint32_t>(device.getInfo<CL_DEVICE_MAX_COMPUTE_UNITS>());
-    core_frequency = static_cast<uint32_t>(device.getInfo<CL_DEVICE_MAX_CLOCK_FREQUENCY>());
+    info.gpu_frequency = static_cast<uint32_t>(device.getInfo<CL_DEVICE_MAX_CLOCK_FREQUENCY>());
 
-    max_work_group_size = static_cast<uint64_t>(device.getInfo<CL_DEVICE_MAX_WORK_GROUP_SIZE>());
+    info.max_work_group_size = static_cast<uint64_t>(device.getInfo<CL_DEVICE_MAX_WORK_GROUP_SIZE>());
 
-    if (max_work_group_size > 256)
-        max_work_group_size = 256;
+    // looks like WA. Do we still need it?
+    if (info.max_work_group_size > 256)
+        info.max_work_group_size = 256;
 
-    max_local_mem_size = static_cast<uint64_t>(device.getInfo<CL_DEVICE_LOCAL_MEM_SIZE>());
-    max_global_mem_size = static_cast<uint64_t>(device.getInfo<CL_DEVICE_GLOBAL_MEM_SIZE>());
-    max_alloc_mem_size = static_cast<uint64_t>(device.getInfo<CL_DEVICE_MAX_MEM_ALLOC_SIZE>());
+    info.max_local_mem_size = static_cast<uint64_t>(device.getInfo<CL_DEVICE_LOCAL_MEM_SIZE>());
+    info.max_global_mem_size = static_cast<uint64_t>(device.getInfo<CL_DEVICE_GLOBAL_MEM_SIZE>());
+    info.max_alloc_mem_size = static_cast<uint64_t>(device.getInfo<CL_DEVICE_MAX_MEM_ALLOC_SIZE>());
 
-    supports_image = static_cast<uint8_t>(device.getInfo<CL_DEVICE_IMAGE_SUPPORT>());
-    max_image2d_width = static_cast<uint64_t>(device.getInfo<CL_DEVICE_IMAGE2D_MAX_WIDTH>());
-    max_image2d_height = static_cast<uint64_t>(device.getInfo<CL_DEVICE_IMAGE2D_MAX_HEIGHT>());
+    info.supports_image = static_cast<uint8_t>(device.getInfo<CL_DEVICE_IMAGE_SUPPORT>());
+    info.max_image2d_width = static_cast<uint64_t>(device.getInfo<CL_DEVICE_IMAGE2D_MAX_WIDTH>());
+    info.max_image2d_height = static_cast<uint64_t>(device.getInfo<CL_DEVICE_IMAGE2D_MAX_HEIGHT>());
 
     // Check for supported features.
     auto extensions = device.getInfo<CL_DEVICE_EXTENSIONS>();
     extensions.push_back(' ');  // Add trailing space to ease searching (search with keyword with trailing space).
 
-    supports_fp16 = extensions.find("cl_khr_fp16 ") != std::string::npos;
-    supports_fp16_denorms = supports_fp16 && (device.getInfo<CL_DEVICE_HALF_FP_CONFIG>() & CL_FP_DENORM) != 0;
-
-    supports_subgroups_short = extensions.find("cl_intel_subgroups_short") != std::string::npos;
+    info.supports_fp16 = extensions.find("cl_khr_fp16 ") != std::string::npos;
+    info.supports_fp64 = extensions.find("cl_khr_fp64 ") != std::string::npos;
+    info.supports_fp16_denorms = info.supports_fp16 && (device.getInfo<CL_DEVICE_HALF_FP_CONFIG>() & CL_FP_DENORM) != 0;
 
-    supports_imad = get_imad_support(device);
-    supports_immad = false;
+    info.supports_subgroups = extensions.find("cl_intel_subgroups") != std::string::npos;
+    info.supports_subgroups_short = extensions.find("cl_intel_subgroups_short") != std::string::npos;
+    info.supports_subgroups_char = extensions.find("cl_intel_subgroups_char") != std::string::npos;
 
-    max_threads_per_execution_unit = 7;
-    max_threads_per_device = static_cast<uint32_t>(cores_count * max_threads_per_execution_unit);
+    info.supports_imad = get_imad_support(device);
+    info.supports_immad = false;
 
-    vendor_id = static_cast<uint32_t>(device.getInfo<CL_DEVICE_VENDOR_ID>());
+    info.max_threads_per_execution_unit = 7;
+    info.max_threads_per_device = static_cast<uint32_t>(info.execution_units_count * info.max_threads_per_execution_unit);
 
-    supports_usm = extensions.find("cl_intel_unified_shared_memory") != std::string::npos;
+    info.supports_usm = extensions.find("cl_intel_unified_shared_memory") != std::string::npos;
 
-    supports_optimization_hints = false;
-    supports_local_block_io = extensions.find("cl_intel_subgroup_local_block_io") != std::string::npos &&
-                              is_local_block_io_supported(device);
+    info.supports_local_block_io = extensions.find("cl_intel_subgroup_local_block_io") != std::string::npos &&
+                                   is_local_block_io_supported(device);
 
     bool device_attr_supported = extensions.find("cl_intel_device_attribute_query") != std::string::npos;
 
     if (device_attr_supported) {
-        gfx_ver = parse_version(device.getInfo<CL_DEVICE_IP_VERSION_INTEL>());
-        device_id = device.getInfo<CL_DEVICE_ID_INTEL>();
-        num_slices = device.getInfo<CL_DEVICE_NUM_SLICES_INTEL>();
-        num_sub_slices_per_slice = device.getInfo<CL_DEVICE_NUM_SUB_SLICES_PER_SLICE_INTEL>();
-        num_eus_per_sub_slice = device.getInfo<CL_DEVICE_NUM_EUS_PER_SUB_SLICE_INTEL>();
-        num_threads_per_eu = device.getInfo<CL_DEVICE_NUM_THREADS_PER_EU_INTEL>();
+        info.gfx_ver = parse_version(device.getInfo<CL_DEVICE_IP_VERSION_INTEL>());
+        info.device_id = device.getInfo<CL_DEVICE_ID_INTEL>();
+        info.num_slices = device.getInfo<CL_DEVICE_NUM_SLICES_INTEL>();
+        info.num_sub_slices_per_slice = device.getInfo<CL_DEVICE_NUM_SUB_SLICES_PER_SLICE_INTEL>();
+        info.num_eus_per_sub_slice = device.getInfo<CL_DEVICE_NUM_EUS_PER_SUB_SLICE_INTEL>();
+        info.num_threads_per_eu = device.getInfo<CL_DEVICE_NUM_THREADS_PER_EU_INTEL>();
         auto features = device.getInfo<CL_DEVICE_FEATURE_CAPABILITIES_INTEL>();
 
-        supports_imad = supports_imad || (features & CL_DEVICE_FEATURE_FLAG_DP4A_INTEL);
+        info.supports_imad = info.supports_imad || (features & CL_DEVICE_FEATURE_FLAG_DP4A_INTEL);
     } else {
-        gfx_ver = {0, 0, 0};
-        device_id = driver_dev_id();
-        num_slices = 0;
-        num_sub_slices_per_slice = 0;
-        num_eus_per_sub_slice = 0;
-        num_threads_per_eu = 0;
+        info.gfx_ver = {0, 0, 0};
+        info.device_id = driver_dev_id();
+        info.num_slices = 0;
+        info.num_sub_slices_per_slice = 0;
+        info.num_eus_per_sub_slice = 0;
+        info.num_threads_per_eu = 0;
     }
+
+    return info;
+}
+
+bool does_device_support(int32_t param, const cl::Device& device) {
+    cl_device_unified_shared_memory_capabilities_intel capabilities;
+    auto err = clGetDeviceInfo(device.get(), param, sizeof(cl_device_unified_shared_memory_capabilities_intel), &capabilities, NULL);
+    if (err) throw std::runtime_error("[CLDNN ERROR]. clGetDeviceInfo error " + std::to_string(err));
+    return !((capabilities & CL_UNIFIED_SHARED_MEMORY_ACCESS_INTEL) == 0u);
 }
-}  // namespace gpu
+
+memory_capabilities init_memory_caps(const cl::Device& device, const device_info& info) {
+    std::vector<allocation_type> memory_caps;
+    if (info.supports_usm) {
+        if (does_device_support(CL_DEVICE_HOST_MEM_CAPABILITIES_INTEL, device)) {
+            memory_caps.push_back(allocation_type::usm_host);
+        }
+        if (does_device_support(CL_DEVICE_SINGLE_DEVICE_SHARED_MEM_CAPABILITIES_INTEL, device)) {
+            memory_caps.push_back(allocation_type::usm_shared);
+        }
+        if (does_device_support(CL_DEVICE_DEVICE_MEM_CAPABILITIES_INTEL, device)) {
+            memory_caps.push_back(allocation_type::usm_device);
+        }
+    }
+
+    return memory_capabilities(memory_caps);
+}
+
+}  // namespace
+
+
+ocl_device::ocl_device(const cl::Device dev, const cl::Context& ctx, const cl_platform_id platform)
+: _context(ctx)
+, _device(dev)
+, _platform(platform)
+, _info(init_device_info(dev))
+, _mem_caps(init_memory_caps(dev, _info)) { }
+
+}  // namespace ocl
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device.hpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device.hpp
new file mode 100644
index 00000000000000..ecb3498e4886b1
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device.hpp
@@ -0,0 +1,42 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/device.hpp"
+#include "ocl_common.hpp"
+
+#include <map>
+#include <string>
+#include <vector>
+#include <algorithm>
+
+namespace cldnn {
+namespace ocl {
+
+struct ocl_device : public device {
+public:
+    ocl_device(const cl::Device dev, const cl::Context& ctx, const cl_platform_id platform);
+
+    device_info get_info() const override { return _info; }
+    memory_capabilities get_mem_caps() const override { return _mem_caps; }
+
+    const cl::Device& get_device() const { return _device; }
+    cl::Device& get_device() { return _device; }
+    const cl::Context& get_context() const { return _context; }
+    cl_platform_id get_platform() const { return _platform; }
+
+    ~ocl_device() = default;
+
+private:
+    cl::Context _context;
+    cl::Device _device;
+    cl_platform_id _platform;
+    device_info _info;
+    memory_capabilities _mem_caps;
+};
+
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ocl_builder.cpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device_detector.cpp
similarity index 68%
rename from inference-engine/thirdparty/clDNN/src/gpu/ocl_builder.cpp
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device_detector.cpp
index 60fcc9b043542e..1ba46d836d4579 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/ocl_builder.cpp
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device_detector.cpp
@@ -2,11 +2,10 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "ocl_builder.h"
-#include "configuration.h"
-#include "include/to_string_utils.h"
-#include "api/device.hpp"
+#include "ocl_device_detector.hpp"
+#include "ocl_device.hpp"
+#include "ocl_common.hpp"
+
 #include <string>
 #include <vector>
 #include <list>
@@ -21,8 +20,29 @@
 #pragma GCC diagnostic ignored "-Wignored-attributes"
 #endif
 
+namespace {
+bool does_device_match_config(bool out_of_order, const cl::Device& device) {
+// Is it intel gpu
+if (device.getInfo<CL_DEVICE_TYPE>() != CL_DEVICE_TYPE_GPU ||
+    device.getInfo<CL_DEVICE_VENDOR_ID>() != 0x8086) {
+    return false;
+}
+
+// Does device support OOOQ?
+if (out_of_order) {
+    auto queue_properties = device.getInfo<CL_DEVICE_QUEUE_PROPERTIES>();
+    using cmp_t = std::common_type<decltype(queue_properties),
+        typename std::underlying_type<cl::QueueProperties>::type>::type;
+    if (!(static_cast<cmp_t>(queue_properties) & static_cast<cmp_t>(cl::QueueProperties::OutOfOrder))) {
+        return false;
+    }
+}
+
+return true;
+}
+}  // namespace
 namespace cldnn {
-namespace gpu {
+namespace ocl {
 static constexpr auto INTEL_PLATFORM_VENDOR = "Intel(R) Corporation";
 
 static std::vector<cl::Device> getSubDevices(cl::Device& rootDevice) {
@@ -68,18 +88,18 @@ static std::vector<cl::Device> getSubDevices(cl::Device& rootDevice) {
     return subDevices;
 }
 
-std::map<std::string, device_impl::ptr> ocl_builder::get_available_devices(void* user_context, void* user_device) const {
+std::map<std::string, device::ptr> ocl_device_detector::get_available_devices(void* user_context, void* user_device) const {
     bool host_out_of_order = true;  // Change to false, if debug requires in-order queue.
-    std::vector<device_impl::ptr> dev_orig, dev_sorted;
+    std::vector<device::ptr> dev_orig, dev_sorted;
     if (user_context != nullptr) {
-        dev_orig = build_device_list_from_user_context(host_out_of_order, user_context);
+        dev_orig = create_device_list_from_user_context(host_out_of_order, user_context);
     } else if (user_device != nullptr) {
-        dev_orig = build_device_list_from_user_device(host_out_of_order, user_device);
+        dev_orig = create_device_list_from_user_device(host_out_of_order, user_device);
     } else {
-        dev_orig = build_device_list(host_out_of_order);
+        dev_orig = create_device_list(host_out_of_order);
     }
 
-    std::map<std::string, device_impl::ptr> ret;
+    std::map<std::string, device::ptr> ret;
     for (auto& dptr : dev_orig) {
         if (dptr->get_info().dev_type == cldnn::device_type::integrated_gpu)
             dev_sorted.insert(dev_sorted.begin(), dptr);
@@ -91,14 +111,16 @@ std::map<std::string, device_impl::ptr> ocl_builder::get_available_devices(void*
         auto map_id = std::to_string(idx++);
         ret[map_id] = dptr;
 
-        auto rootDevice = dptr->get_device();
-        auto subDevices = getSubDevices(rootDevice);
+        auto rootDevice = std::dynamic_pointer_cast<ocl_device>(dptr);
+        if (!rootDevice) {
+            throw std::runtime_error("Invalid device type created in ocl_device_detector");
+        }
+
+        auto subDevices = getSubDevices(rootDevice->get_device());
         if (!subDevices.empty()) {
             uint32_t sub_idx = 0;
             for (auto& subdevice : subDevices) {
-                auto subdPtr = device_impl::ptr(new device_impl(subdevice, cl::Context(subdevice),
-                                                                dptr->get_platform(),
-                                                                device_info_internal(subdevice)), false);
+                auto subdPtr = std::make_shared<ocl_device>(subdevice, cl::Context(subdevice), rootDevice->get_platform());
                 ret[map_id+"."+std::to_string(sub_idx++)] = subdPtr;
             }
         }
@@ -106,7 +128,7 @@ std::map<std::string, device_impl::ptr> ocl_builder::get_available_devices(void*
     return ret;
 }
 
-std::vector<device_impl::ptr> ocl_builder::build_device_list(bool out_out_order) const {
+std::vector<device::ptr> ocl_device_detector::create_device_list(bool out_out_order) const {
     cl_uint n = 0;
     // Get number of platforms availible
     cl_int err = clGetPlatformIDs(0, NULL, &n);
@@ -121,7 +143,7 @@ std::vector<device_impl::ptr> ocl_builder::build_device_list(bool out_out_order)
         throw std::runtime_error("[CLDNN ERROR]. clGetPlatformIDs error " + std::to_string(err));
     }
 
-    std::vector<device_impl::ptr> ret;
+    std::vector<device::ptr> ret;
     for (auto& id : platform_ids) {
         cl::Platform platform = cl::Platform(id);
 
@@ -131,9 +153,9 @@ std::vector<device_impl::ptr> ocl_builder::build_device_list(bool out_out_order)
         std::vector<cl::Device> devices;
         platform.getDevices(CL_DEVICE_TYPE_ALL, &devices);
         for (auto& device : devices) {
-            if (!does_device_match_config(out_out_order, device)) continue;
-            ret.emplace_back(device_impl::ptr{ new device_impl(device, cl::Context(device),
-                                                        id, device_info_internal(device)), false});
+            if (!does_device_match_config(out_out_order, device))
+                continue;
+            ret.emplace_back(std::make_shared<ocl_device>(device, cl::Context(device), id));
         }
     }
     if (ret.empty()) {
@@ -142,16 +164,15 @@ std::vector<device_impl::ptr> ocl_builder::build_device_list(bool out_out_order)
     return ret;
 }
 
-std::vector<device_impl::ptr>  ocl_builder::build_device_list_from_user_context(bool out_out_order, void* user_context) const {
+std::vector<device::ptr>  ocl_device_detector::create_device_list_from_user_context(bool out_out_order, void* user_context) const {
     cl::Context ctx = cl::Context(static_cast<cl_context>(user_context), true);
     auto all_devices = ctx.getInfo<CL_CONTEXT_DEVICES>();
 
-    std::vector<device_impl::ptr> ret;
+    std::vector<device::ptr> ret;
     for (auto& device : all_devices) {
-        if (!does_device_match_config(out_out_order, device)) continue;
-        ret.emplace_back(device_impl::ptr{ new device_impl(device, cl::Context(device),
-                                                        device.getInfo<CL_DEVICE_PLATFORM>(),
-                                                        device_info_internal(device)), false});
+        if (!does_device_match_config(out_out_order, device))
+            continue;
+        ret.emplace_back(std::make_shared<ocl_device>(device, cl::Context(device), device.getInfo<CL_DEVICE_PLATFORM>()));
     }
 
     if (ret.empty()) {
@@ -160,7 +181,7 @@ std::vector<device_impl::ptr>  ocl_builder::build_device_list_from_user_context(
     return ret;
 }
 
-std::vector<device_impl::ptr>  ocl_builder::build_device_list_from_user_device(bool out_out_order, void* user_device) const {
+std::vector<device::ptr>  ocl_device_detector::create_device_list_from_user_device(bool out_out_order, void* user_device) const {
     cl_uint n = 0;
     // Get number of platforms availible
     cl_int err = clGetPlatformIDs(0, NULL, &n);
@@ -175,7 +196,7 @@ std::vector<device_impl::ptr>  ocl_builder::build_device_list_from_user_device(b
         throw std::runtime_error("[CLDNN ERROR]. clGetPlatformIDs error " + std::to_string(err));
     }
 
-    std::vector<device_impl::ptr> ret;
+    std::vector<device::ptr> ret;
     for (auto& id : platform_ids) {
         cl::PlatformVA platform = cl::PlatformVA(id);
 
@@ -206,8 +227,7 @@ std::vector<device_impl::ptr>  ocl_builder::build_device_list_from_user_device(b
                 CL_CONTEXT_INTEROP_USER_SYNC, CL_FALSE,
                 CL_CONTEXT_PLATFORM, (cl_context_properties)id,
                 0 };
-            ret.emplace_back(device_impl::ptr{ new device_impl(device, cl::Context(device, props),
-                                                            id, device_info_internal(device)), false });
+            ret.emplace_back(std::make_shared<ocl_device>(device, cl::Context(device, props), id));
         }
     }
     if (ret.empty()) {
@@ -216,25 +236,5 @@ std::vector<device_impl::ptr>  ocl_builder::build_device_list_from_user_device(b
     return ret;
 }
 
-bool ocl_builder::does_device_match_config(bool out_of_order, const cl::Device& device) const {
-    // Is it intel gpu
-    if (device.getInfo<CL_DEVICE_TYPE>() != device_type ||
-        device.getInfo<CL_DEVICE_VENDOR_ID>() != device_vendor) {
-        return false;
-    }
-
-    // Does device support OOOQ?
-    if (out_of_order) {
-        auto queue_properties = device.getInfo<CL_DEVICE_QUEUE_PROPERTIES>();
-        using cmp_t = std::common_type<decltype(queue_properties),
-            typename std::underlying_type<cl::QueueProperties>::type>::type;
-        if (!(static_cast<cmp_t>(queue_properties) & static_cast<cmp_t>(cl::QueueProperties::OutOfOrder))) {
-            return false;
-        }
-    }
-
-    return true;
-}
-
-}  // namespace gpu
+}  // namespace ocl
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device_detector.hpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device_detector.hpp
new file mode 100644
index 00000000000000..d738285989c5ca
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device_detector.hpp
@@ -0,0 +1,30 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "cldnn/runtime/device.hpp"
+
+#include <list>
+#include <string>
+#include <utility>
+#include <vector>
+#include <map>
+
+namespace cldnn {
+namespace ocl {
+
+class ocl_device_detector {
+public:
+    ocl_device_detector() = default;
+
+    std::map<std::string, device::ptr> get_available_devices(void* user_context, void* user_device) const;
+private:
+    std::vector<device::ptr> create_device_list(bool out_out_order) const;
+    std::vector<device::ptr> create_device_list_from_user_context(bool out_out_order, void* user_context) const;
+    std::vector<device::ptr> create_device_list_from_user_device(bool out_out_order, void* user_device) const;
+};
+
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine.cpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine.cpp
new file mode 100644
index 00000000000000..95b2950ec61537
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine.cpp
@@ -0,0 +1,204 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "ocl_engine.hpp"
+#include "ocl_common.hpp"
+#include "ocl_memory.hpp"
+#include "ocl_stream.hpp"
+#include <string>
+#include <vector>
+#include <memory>
+#include <set>
+#include <stdexcept>
+
+// NOTE: Due to buggy scope transition of warnings we need to disable warning in place of use/instantation
+//       of some types (even though we already disabled them in scope of definition of these types).
+//       Moreover this warning is pretty much now only for annoyance: it is generated due to lack
+//       of proper support for mangling of custom GCC attributes into type name (usually when used
+//       with templates, even from standard library).
+#if defined __GNUC__ && __GNUC__ >= 6
+#pragma GCC diagnostic ignored "-Wignored-attributes"
+#endif
+
+// static class memebers - pointers to dynamically obtained OpenCL extension functions
+cl::PFN_clEnqueueAcquireMediaSurfacesINTEL cl::SharedSurfLock::pfn_acquire = NULL;
+cl::PFN_clEnqueueReleaseMediaSurfacesINTEL cl::SharedSurfLock::pfn_release = NULL;
+cl::PFN_clCreateFromMediaSurfaceINTEL cl::ImageVA::pfn_clCreateFromMediaSurfaceINTEL = NULL;
+#ifdef _WIN32
+cl::PFN_clCreateFromD3D11Buffer cl::BufferDX::pfn_clCreateFromD3D11Buffer = NULL;
+#endif
+
+namespace cldnn {
+namespace ocl {
+
+ocl_error::ocl_error(cl::Error const& err)
+    : std::runtime_error(err.what() + std::string(", error code: ") + std::to_string(err.err())) {}
+
+ocl_engine::ocl_engine(const device::ptr dev, runtime_types runtime_type, const engine_configuration& conf)
+    : engine(dev, conf) {
+    if (runtime_type != runtime_types::ocl) {
+        throw std::runtime_error("Invalid runtime type specified for OCL engine. Only OCL runtime is supported");
+    }
+
+    auto casted = dynamic_cast<ocl_device*>(dev.get());
+    if (!casted)
+        throw std::runtime_error("[CLDNN] Invalid device type passed to ocl engine");
+    casted->get_device().getInfo(CL_DEVICE_EXTENSIONS, &_extensions);
+
+    _program_stream.reset(new ocl_stream(*this));
+}
+
+const cl::Context& ocl_engine::get_cl_context() const {
+    auto cl_device = std::dynamic_pointer_cast<ocl_device>(_device);
+    if (!cl_device)
+        throw std::runtime_error("Invalid device type for ocl_engine");
+    return cl_device->get_context();
+}
+
+const cl::Device& ocl_engine::get_cl_device() const {
+    auto cl_device = std::dynamic_pointer_cast<ocl_device>(_device);
+    if (!cl_device)
+        throw std::runtime_error("Invalid device type for ocl_engine");
+    return cl_device->get_device();
+}
+
+memory::ptr ocl_engine::allocate_memory(const layout& layout, allocation_type type, bool reset) {
+    if (layout.bytes_count() > get_device_info().max_alloc_mem_size) {
+        throw std::runtime_error("exceeded max size of memory object allocation");
+    }
+
+    _memory_pool->add_memory_used(layout.bytes_count());
+
+    try {
+        memory::ptr res = nullptr;
+        if (layout.format.is_image_2d()) {
+            res = std::make_shared<ocl::gpu_image2d>(this, layout);
+        } else if (type == allocation_type::cl_mem) {
+            res = std::make_shared<ocl::gpu_buffer>(this, layout);
+        } else {
+            res = std::make_shared<ocl::gpu_usm>(this, layout, type);
+        }
+
+        if (reset || res->is_memory_reset_needed(layout)) {
+            res->fill(get_program_stream());
+        }
+
+        return res;
+    } catch (const cl::Error& clErr) {
+        switch (clErr.err()) {
+            case CL_MEM_OBJECT_ALLOCATION_FAILURE:
+            case CL_OUT_OF_RESOURCES:
+            case CL_OUT_OF_HOST_MEMORY:
+            case CL_INVALID_BUFFER_SIZE:
+                throw std::runtime_error("out of GPU resources");
+            default:
+                throw std::runtime_error("GPU buffer allocation failed");
+        }
+    }
+}
+
+memory::ptr ocl_engine::reinterpret_buffer(const memory& memory, const layout& new_layout) {
+    if (memory.get_engine() != this)
+        throw std::runtime_error("trying to reinterpret buffer allocated by a different engine");
+
+    if (new_layout.format.is_image() && !memory.get_layout().format.is_image())
+        throw std::runtime_error("trying to reinterpret non-image buffer as image");
+
+    if (!new_layout.format.is_image() && memory.get_layout().format.is_image())
+        throw std::runtime_error("trying to reinterpret image buffer as non-image buffer");
+
+    try {
+        if (new_layout.format.is_image_2d()) {
+           return std::make_shared<ocl::gpu_image2d>(this,
+                                     new_layout,
+                                     reinterpret_cast<const ocl::gpu_image2d&>(memory).get_buffer());
+        } else if (memory_capabilities::is_usm_type(memory.get_allocation_type())) {
+            return std::make_shared<ocl::gpu_usm>(this,
+                                     new_layout,
+                                     reinterpret_cast<const ocl::gpu_usm&>(memory).get_buffer(),
+                                     memory.get_allocation_type());
+        } else {
+           return std::make_shared<ocl::gpu_buffer>(this,
+                                    new_layout,
+                                    reinterpret_cast<const ocl::gpu_buffer&>(memory).get_buffer());
+        }
+    } catch (cl::Error const& err) {
+        throw ocl::ocl_error(err);
+    }
+}
+
+memory::ptr ocl_engine::reinterpret_handle(const layout& new_layout, shared_mem_params params) {
+   try {
+        if (new_layout.format.is_image_2d() && params.mem_type == shared_mem_type::shared_mem_image) {
+            cl::Image2D img(static_cast<cl_mem>(params.mem), true);
+            return std::make_shared<ocl::gpu_image2d>(this, new_layout, img);
+        } else if (new_layout.format.is_image_2d() && params.mem_type == shared_mem_type::shared_mem_vasurface) {
+            return std::make_shared<ocl::gpu_media_buffer>(this, new_layout, params);
+#ifdef _WIN32
+        } else if (params.mem_type == shared_mem_type::shared_mem_dxbuffer) {
+            return std::make_shared<ocl::gpu_dx_buffer>(this, new_layout, params);
+#endif
+        } else if (params.mem_type == shared_mem_type::shared_mem_buffer) {
+            cl::Buffer buf(static_cast<cl_mem>(params.mem), true);
+            return std::make_shared<ocl::gpu_buffer>(this, new_layout, buf);
+        } else {
+            throw std::runtime_error("unknown shared object fromat or type");
+        }
+    }
+    catch (const cl::Error& clErr) {
+        switch (clErr.err()) {
+        case CL_MEM_OBJECT_ALLOCATION_FAILURE:
+        case CL_OUT_OF_RESOURCES:
+        case CL_OUT_OF_HOST_MEMORY:
+        case CL_INVALID_BUFFER_SIZE:
+            throw std::runtime_error("out of GPU resources");
+        default:
+            throw std::runtime_error("GPU buffer allocation failed");
+        }
+    }
+}
+
+bool ocl_engine::is_the_same_buffer(const memory& mem1, const memory& mem2) {
+    if (mem1.get_engine() != this || mem2.get_engine() != this)
+        return false;
+    if (mem1.get_allocation_type() != mem2.get_allocation_type())
+        return false;
+    if (&mem1 == &mem2)
+        return true;
+
+    if (!memory_capabilities::is_usm_type(mem1.get_allocation_type()))
+        return (reinterpret_cast<const ocl::gpu_buffer&>(mem1).get_buffer() ==
+                reinterpret_cast<const ocl::gpu_buffer&>(mem2).get_buffer());
+    else
+        return (reinterpret_cast<const ocl::gpu_usm&>(mem1).get_buffer() ==
+                reinterpret_cast<const ocl::gpu_usm&>(mem2).get_buffer());
+}
+
+void* ocl_engine::get_user_context() const {
+    auto& cl_device = downcast<ocl_device>(*_device);
+    return static_cast<void*>(cl_device.get_context().get());
+}
+
+bool ocl_engine::extension_supported(std::string extension) const {
+    return _extensions.find(extension) != std::string::npos;
+}
+
+stream::ptr ocl_engine::create_stream() const {
+    return std::make_shared<ocl_stream>(*this);
+}
+
+stream& ocl_engine::get_program_stream() const {
+    return *_program_stream;
+}
+
+std::shared_ptr<cldnn::engine> ocl_engine::create(const device::ptr device, runtime_types runtime_type, const engine_configuration& configuration) {
+    return std::make_shared<ocl::ocl_engine>(device, runtime_type, configuration);
+}
+
+std::shared_ptr<cldnn::engine> create_ocl_engine(const device::ptr device, runtime_types runtime_type, const engine_configuration& configuration) {
+    return ocl_engine::create(device, runtime_type, configuration);
+}
+
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine.hpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine.hpp
new file mode 100644
index 00000000000000..d0916fcc6458f7
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine.hpp
@@ -0,0 +1,51 @@
+// Copyright (C) 2016-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/stream.hpp"
+#include "ocl_device.hpp"
+
+#include <memory>
+#include <set>
+#include <vector>
+#include <utility>
+#include <string>
+
+namespace cldnn {
+namespace ocl {
+
+class ocl_engine : public engine {
+public:
+    ocl_engine(const device::ptr dev, runtime_types runtime_type, const engine_configuration& conf);
+    engine_types type() const override { return engine_types::ocl; };
+    runtime_types runtime_type() const override { return runtime_types::ocl; };
+
+    memory_ptr allocate_memory(const layout& layout, allocation_type type, bool reset = true) override;
+    memory_ptr reinterpret_handle(const layout& new_layout, shared_mem_params params) override;
+    memory_ptr reinterpret_buffer(const memory& memory, const layout& new_layout) override;
+    bool is_the_same_buffer(const memory& mem1, const memory& mem2) override;
+
+    void* get_user_context() const override;
+
+    allocation_type get_default_allocation_type() const override { return allocation_type::cl_mem; }
+
+    const cl::Context& get_cl_context() const;
+    const cl::Device& get_cl_device() const;
+
+    bool extension_supported(std::string extension) const;
+
+    stream_ptr create_stream() const override;
+    stream& get_program_stream() const override;
+
+    static std::shared_ptr<cldnn::engine> create(const device::ptr device, runtime_types runtime_type, const engine_configuration& configuration);
+private:
+    std::string _extensions;
+    std::unique_ptr<stream> _program_stream;
+};
+
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine_factory.hpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine_factory.hpp
new file mode 100644
index 00000000000000..3d2530c234281f
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_engine_factory.hpp
@@ -0,0 +1,19 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "cldnn/runtime/device.hpp"
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/stream.hpp"
+
+namespace cldnn {
+namespace ocl {
+
+// Factory for ocl_engine creation. It's moved outside of ocl_engine class to avoid possible CL includes conflict
+// between different engines in engine.cpp file
+std::shared_ptr<cldnn::engine> create_ocl_engine(const device::ptr device, runtime_types runtime_type, const engine_configuration& configuration);
+
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/events_pool.h b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_events_pool.hpp
similarity index 57%
rename from inference-engine/thirdparty/clDNN/src/gpu/events_pool.h
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_events_pool.hpp
index 9fdf0a23fabe44..d8fde6242b5bcc 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/events_pool.h
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_events_pool.hpp
@@ -2,18 +2,15 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "refcounted_obj.h"
-#include "event_impl.h"
-#include "meta_utils.h"
+#include "cldnn/runtime/utils.hpp"
+#include "cldnn/runtime/event.hpp"
+
 #include <iostream>
 #include <vector>
 #include <memory>
 
 namespace cldnn {
-namespace gpu {
-
-class gpu_toolkit;
+namespace ocl {
 
 template <typename Type,
           typename U = typename std::enable_if<meta::is_any_of<Type, base_event, user_event, base_events>::value>::type>
@@ -23,14 +20,14 @@ class event_pool_impl {
 
     using type = Type;
 
-    event_impl::ptr get_from_pool(std::shared_ptr<gpu_toolkit>& ctx) {
+    event::ptr get_from_pool(const cl::Context& ctx) {
         for (auto& ev : _events) {
             if (!ev->is_valid()) {
                 ev->reset();
                 return ev;
             }
         }
-        const event_impl::ptr ev_impl { new Type(ctx), false };
+        auto ev_impl = std::make_shared<Type>(ctx);
         return allocate(ev_impl);
     }
 
@@ -39,36 +36,36 @@ class event_pool_impl {
     }
 
 private:
-    std::vector<event_impl::ptr> _events;
+    std::vector<event::ptr> _events;
 
-    event_impl::ptr allocate(const event_impl::ptr& obj) {
+    event::ptr allocate(const event::ptr& obj) {
         _events.emplace_back(obj);
         return _events.back();
     }
 };
 
 struct base_event_pool : event_pool_impl<base_event> {
-    event_impl::ptr get(std::shared_ptr<gpu_toolkit>& ctx, const cl::Event& ev, const uint64_t q_stamp) {
+    event::ptr get(const cl::Context& ctx, const cl::Event& ev, const uint64_t q_stamp) {
         auto ret = get_from_pool(ctx);
-        dynamic_cast<type*>(ret.get())->attach_ocl_event(ev, q_stamp);
+        std::dynamic_pointer_cast<type>(ret)->attach_ocl_event(ev, q_stamp);
         return ret;
     }
     void reset() { reset_events(); }
 };
 
 struct user_event_pool : event_pool_impl<user_event> {
-    event_impl::ptr get(std::shared_ptr<gpu_toolkit>& ctx, bool set = false) {
+    event::ptr get(const cl::Context& ctx, bool set = false) {
         auto ret = get_from_pool(ctx);
-        dynamic_cast<type*>(ret.get())->attach_event(set);
+        downcast<type>(ret.get())->attach_event(set);
         return ret;
     }
     void reset() { reset_events(); }
 };
 
 struct group_event_pool : event_pool_impl<base_events> {
-    event_impl::ptr get(std::shared_ptr<gpu_toolkit>& ctx, const std::vector<event_impl::ptr>& deps) {
+    event::ptr get(const cl::Context& ctx, const std::vector<event::ptr>& deps) {
         auto ret_ev = get_from_pool(ctx);
-        dynamic_cast<type*>(ret_ev.get())->attach_events(deps);
+        downcast<type>(ret_ev.get())->attach_events(deps);
         return ret_ev;
     }
     void reset() { reset_events(); }
@@ -78,15 +75,15 @@ class events_pool {
 public:
     events_pool() = default;
 
-    event_impl::ptr get_from_base_pool(std::shared_ptr<gpu_toolkit> ctx, const cl::Event& ev, const uint64_t q_stamp) {
+    event::ptr get_from_base_pool(const cl::Context& ctx, const cl::Event& ev, const uint64_t q_stamp) {
         return _base_pool.get(ctx, ev, q_stamp);
     }
 
-    event_impl::ptr get_from_user_pool(std::shared_ptr<gpu_toolkit> ctx, bool set = false) {
+    event::ptr get_from_user_pool(const cl::Context& ctx, bool set = false) {
         return _user_pool.get(ctx, set);
     }
 
-    event_impl::ptr get_from_group_pool(std::shared_ptr<gpu_toolkit> ctx, const std::vector<event_impl::ptr>& deps) {
+    event::ptr get_from_group_pool(const cl::Context& ctx, const std::vector<event::ptr>& deps) {
         return _group_pool.get(ctx, deps);
     }
 
@@ -101,5 +98,5 @@ class events_pool {
     user_event_pool _user_pool;
     group_event_pool _group_pool;
 };
-}  // namespace gpu
+}  // namespace ocl
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/common/include/cl2_ext.hpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_ext.hpp
similarity index 97%
rename from inference-engine/thirdparty/clDNN/common/include/cl2_ext.hpp
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_ext.hpp
index 8af34068af8d99..bacaeb03b008df 100644
--- a/inference-engine/thirdparty/clDNN/common/include/cl2_ext.hpp
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_ext.hpp
@@ -52,11 +52,9 @@ CL_HPP_DECLARE_PARAM_TRAITS_(cl_device_info, CL_DEVICE_FEATURE_CAPABILITIES_INTE
 namespace {
 template <typename T>
 T load_entrypoint(const cl_platform_id platform, const std::string name) {
-    T p = reinterpret_cast<T>(
-        clGetExtensionFunctionAddressForPlatform(platform, name.c_str()));
+    T p = reinterpret_cast<T>(clGetExtensionFunctionAddressForPlatform(platform, name.c_str()));
     if (!p) {
-        throw std::runtime_error("clGetExtensionFunctionAddressForPlatform(" +
-            name + ") returned NULL.");
+        throw std::runtime_error("clGetExtensionFunctionAddressForPlatform(" + name + ") returned NULL.");
     }
     return p;
 }
@@ -64,11 +62,9 @@ T load_entrypoint(const cl_platform_id platform, const std::string name) {
 template <typename T>
 T load_entrypoint(const cl_device_id device, const std::string name) {
     cl_platform_id platform;
-    cl_int error = clGetDeviceInfo(device, CL_DEVICE_PLATFORM, sizeof(platform),
-        &platform, nullptr);
+    cl_int error = clGetDeviceInfo(device, CL_DEVICE_PLATFORM, sizeof(platform), &platform, nullptr);
     if (error) {
-        throw std::runtime_error("Failed to retrieve CL_DEVICE_PLATFORM: " +
-            std::to_string(error));
+        throw std::runtime_error("Failed to retrieve CL_DEVICE_PLATFORM: " + std::to_string(error));
     }
     return load_entrypoint<T>(platform, name);
 }
@@ -78,20 +74,16 @@ T load_entrypoint(const cl_device_id device, const std::string name) {
 template <typename T>
 T load_entrypoint(const cl_context context, const std::string name) {
     size_t size = 0;
-    cl_int error =
-        clGetContextInfo(context, CL_CONTEXT_DEVICES, 0, nullptr, &size);
+    cl_int error = clGetContextInfo(context, CL_CONTEXT_DEVICES, 0, nullptr, &size);
     if (error) {
-        throw std::runtime_error("Failed to retrieve CL_CONTEXT_DEVICES size: " +
-            std::to_string(error));
+        throw std::runtime_error("Failed to retrieve CL_CONTEXT_DEVICES size: " + std::to_string(error));
     }
 
     std::vector<cl_device_id> devices(size / sizeof(cl_device_id));
 
-    error = clGetContextInfo(context, CL_CONTEXT_DEVICES, size, devices.data(),
-        nullptr);
+    error = clGetContextInfo(context, CL_CONTEXT_DEVICES, size, devices.data(), nullptr);
     if (error) {
-        throw std::runtime_error("Failed to retrieve CL_CONTEXT_DEVICES: " +
-            std::to_string(error));
+        throw std::runtime_error("Failed to retrieve CL_CONTEXT_DEVICES: " + std::to_string(error));
     }
 
     return load_entrypoint<T>(devices.front(), name);
@@ -148,11 +140,6 @@ inline void* deviceMemAlloc(const cl::Device& cpp_device, const cl::Context& cpp
     return fn(cpp_context.get(), cpp_device.get(), properties, size, alignment, err_code_ret);
 }
 
-inline cl_int memFree(const cl::Context& cpp_context, void* ptr) {
-    clMemFreeINTEL_fn fn = load_entrypoint<clMemFreeINTEL_fn>(cpp_context.get(), "clMemFreeINTEL");
-    return fn(cpp_context.get(), ptr);
-}
-
 inline cl_int set_kernel_arg_mem_pointer(const cl::Kernel& kernel, uint32_t index, const void* ptr, clSetKernelArgMemPointerINTEL_fn fn) {
     return fn(kernel.get(), index, ptr);
 }
@@ -605,12 +592,20 @@ typedef CL_API_ENTRY cl_mem(CL_API_CALL * PFN_clCreateFromMediaSurfaceINTEL)(
     */
     class UsmHolder {
     public:
-        explicit UsmHolder(Context& ctx, void* ptr) : _ctx(ctx), _ptr(ptr) {}
+        explicit UsmHolder(Context& ctx, void* ptr) : _ctx(ctx), _ptr(ptr), deleter(nullptr) {
+            deleter = load_entrypoint<clMemFreeINTEL_fn>(_ctx.get(), "clMemFreeINTEL");
+            if (!deleter) {
+                throw std::runtime_error("clMemFreeINTEL is nullptr in UsmHolder");
+            }
+        }
         void* ptr() { return _ptr; }
-        ~UsmHolder() { usm::memFree(_ctx, _ptr); }
+        ~UsmHolder() {
+            deleter(_ctx.get(), _ptr);
+        }
     private:
         Context _ctx;
         void* _ptr;
+        clMemFreeINTEL_fn deleter;
     };
 
     /*
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_kernel.hpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_kernel.hpp
new file mode 100644
index 00000000000000..4bda3ef8f31608
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_kernel.hpp
@@ -0,0 +1,33 @@
+// Copyright (C) 2016-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "ocl_common.hpp"
+#include "ocl_memory.hpp"
+#include "cldnn/runtime/kernel_args.hpp"
+#include "cldnn/runtime/kernel.hpp"
+
+#include <memory>
+#include <vector>
+
+namespace cldnn {
+namespace ocl {
+
+class ocl_kernel : public kernel {
+    ocl_kernel_type _compiled_kernel;
+    std::string _kernel_id;
+
+public:
+    ocl_kernel(ocl_kernel_type compiled_kernel, const std::string& kernel_id)
+        : _compiled_kernel(compiled_kernel)
+        , _kernel_id(kernel_id) { }
+
+    const ocl_kernel_type& get_handle() const { return _compiled_kernel; }
+    ocl_kernel_type& get_handle() { return _compiled_kernel; }
+    std::shared_ptr<kernel> clone() const override { return std::make_shared<ocl_kernel>(get_handle().clone(), _kernel_id); }
+};
+
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_kernels_factory.cpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_kernels_factory.cpp
new file mode 100644
index 00000000000000..29a431442616ab
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_kernels_factory.cpp
@@ -0,0 +1,21 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "ocl_kernel.hpp"
+#include "kernels_factory.hpp"
+
+#include <memory>
+#include <vector>
+
+namespace cldnn {
+namespace ocl {
+
+std::shared_ptr<kernel> create_ocl_kernel(engine& engine, cl_context /* context */, cl_kernel kernel, std::string entry_point) {
+    // Retain kernel to keep it valid
+    cl::Kernel k(kernel, true);
+    return std::make_shared<ocl::ocl_kernel>(ocl::ocl_kernel_type(k, engine.use_unified_shared_memory()), entry_point);
+}
+
+}  // namespace kernels_factory
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_memory.cpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_memory.cpp
new file mode 100644
index 00000000000000..9a5c2b3aa52ab0
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_memory.cpp
@@ -0,0 +1,373 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "cldnn/runtime/error_handler.hpp"
+#include "cldnn/runtime/utils.hpp"
+#include "ocl_memory.hpp"
+#include "ocl_engine.hpp"
+#include "ocl_stream.hpp"
+#include "ocl_base_event.hpp"
+#include <stdexcept>
+#include <vector>
+
+namespace cldnn {
+namespace ocl {
+
+gpu_buffer::gpu_buffer(ocl_engine* engine,
+                       const layout& layout)
+    : lockable_gpu_mem(), memory(engine, layout, allocation_type::cl_mem, false)
+    , _buffer(engine->get_cl_context(), CL_MEM_READ_WRITE, size()) { }
+
+gpu_buffer::gpu_buffer(ocl_engine* engine,
+                       const layout& new_layout,
+                       const cl::Buffer& buffer)
+    : lockable_gpu_mem(), memory(engine, new_layout, allocation_type::cl_mem, true)
+    , _buffer(buffer) {}
+
+void* gpu_buffer::lock(const stream& stream) {
+    auto& cl_stream = downcast<const ocl_stream>(stream);
+    std::lock_guard<std::mutex> locker(_mutex);
+    if (0 == _lock_count) {
+        _mapped_ptr = cl_stream.get_cl_queue().enqueueMapBuffer(_buffer, CL_TRUE, CL_MAP_WRITE, 0, size());
+    }
+    _lock_count++;
+    return _mapped_ptr;
+}
+
+void gpu_buffer::unlock(const stream& stream) {
+    auto& cl_stream = downcast<const ocl_stream>(stream);
+    std::lock_guard<std::mutex> locker(_mutex);
+    _lock_count--;
+    if (0 == _lock_count) {
+        cl_stream.get_cl_queue().enqueueUnmapMemObject(_buffer, _mapped_ptr);
+        _mapped_ptr = nullptr;
+    }
+}
+
+event::ptr gpu_buffer::fill(stream& stream) {
+    return fill(stream, 0);
+}
+
+event::ptr gpu_buffer::fill(stream& stream, unsigned char pattern) {
+    auto& cl_stream = downcast<ocl_stream>(stream);
+    auto ev = stream.create_base_event();
+    cl::Event ev_ocl = std::dynamic_pointer_cast<base_event>(ev)->get();
+    cl_stream.get_cl_queue().enqueueFillBuffer<unsigned char>(_buffer, pattern, 0, size(), nullptr, &ev_ocl);
+
+    // TODO: do we need sync here?
+    cl_stream.finish();
+
+    return ev;
+}
+
+shared_mem_params gpu_buffer::get_internal_params() const {
+    auto cl_engine = downcast<const ocl_engine>(_engine);
+    return {shared_mem_type::shared_mem_buffer, static_cast<shared_handle>(cl_engine->get_cl_context().get()), nullptr,
+            static_cast<shared_handle>(_buffer.get()),
+#ifdef _WIN32
+        nullptr,
+#else
+        0,
+#endif
+        0};
+}
+
+event::ptr gpu_buffer::copy_from(stream& /* stream */, const memory& /* other */) {
+    throw std::runtime_error("[clDNN] copy_from is not implemented for gpu_buffer");
+}
+
+event::ptr gpu_buffer::copy_from(stream& stream, const void* host_ptr) {
+    auto& cl_stream = downcast<ocl_stream>(stream);
+    auto ev = stream.create_base_event();
+    cl::Event ev_ocl = std::dynamic_pointer_cast<base_event>(ev)->get();
+    cl_stream.get_cl_queue().enqueueWriteBuffer(_buffer, false, 0, size(), host_ptr, nullptr, &ev_ocl);
+
+    return ev;
+}
+
+gpu_image2d::gpu_image2d(ocl_engine* engine, const layout& layout)
+    : lockable_gpu_mem(), memory(engine, layout, allocation_type::cl_mem, false), _row_pitch(0), _slice_pitch(0) {
+    cl_channel_type type = layout.data_type == data_types::f16 ? CL_HALF_FLOAT : CL_FLOAT;
+    cl_channel_order order = CL_R;
+    switch (layout.format) {
+        case format::image_2d_weights_c1_b_fyx:
+            _width = layout.size.batch[0];
+            _height = layout.size.spatial[0] * layout.size.feature[0] * layout.size.spatial[1];
+            break;
+        case format::image_2d_weights_winograd_6x3_s1_fbxyb:
+            _height = layout.size.feature[0];
+            _width = layout.size.spatial[0] * layout.size.batch[0] * layout.size.spatial[1] * 8 / 3;
+            break;
+        case format::image_2d_weights_winograd_6x3_s1_xfbyb:
+            _height = layout.size.feature[0] * layout.size.spatial[0] * 8 / 3;
+            _width = layout.size.batch[0] * layout.size.spatial[1];
+            break;
+        case format::image_2d_weights_c4_fyx_b:
+            _width = layout.size.batch[0];
+            _height = layout.size.spatial[0] * layout.size.feature[0] * layout.size.spatial[1];
+            order = CL_RGBA;
+            break;
+        case format::image_2d_rgba:
+            _width = layout.size.spatial[0];
+            _height = layout.size.spatial[1];
+            order = CL_RGBA;
+            if (layout.size.feature[0] != 3 && layout.size.feature[0] != 4) {
+                CLDNN_ERROR_MESSAGE("2D image allocation", "invalid number of channels in image_2d_rgba input image (should be 3 or 4)!");
+            }
+            type = CL_UNORM_INT8;
+            break;
+        case format::nv12:
+            _width = layout.size.spatial[1];
+            _height = layout.size.spatial[0];
+            if (layout.size.feature[0] == 2) {
+                order = CL_RG;
+            } else if (layout.size.feature[0] > 2) {
+                CLDNN_ERROR_MESSAGE("2D image allocation", "invalid number of channels in NV12 input image!");
+            }
+            type = CL_UNORM_INT8;
+            break;
+        default:
+            CLDNN_ERROR_MESSAGE("2D image allocation", "unsupported image type!");
+    }
+
+    cl::ImageFormat imageFormat(order, type);
+    _buffer = cl::Image2D(engine->get_cl_context(), CL_MEM_READ_WRITE, imageFormat, _width, _height, 0);
+}
+
+gpu_image2d::gpu_image2d(ocl_engine* engine,
+                         const layout& new_layout,
+                         const cl::Image2D& buffer)
+    : lockable_gpu_mem(), memory(engine, new_layout, allocation_type::cl_mem, true),
+      _buffer(buffer) {
+    _width = _buffer.getImageInfo<CL_IMAGE_WIDTH>();
+    _height = _buffer.getImageInfo<CL_IMAGE_HEIGHT>();
+    _row_pitch = _buffer.getImageInfo<CL_IMAGE_ROW_PITCH>();
+    _slice_pitch = _buffer.getImageInfo<CL_IMAGE_SLICE_PITCH>();
+}
+
+event::ptr gpu_image2d::fill(stream& stream) {
+    return fill(stream, 0);
+}
+
+event::ptr gpu_image2d::fill(stream& stream, unsigned char pattern) {
+    auto& cl_stream = downcast<ocl_stream>(stream);
+    auto ev = stream.create_base_event();
+    cl::Event ev_ocl = downcast<base_event>(ev.get())->get();
+    cl_uint4 pattern_uint4 = {pattern, pattern, pattern, pattern};
+    cl_stream.get_cl_queue().enqueueFillImage(_buffer, pattern_uint4, {0, 0, 0}, {_width, _height, 1}, 0, &ev_ocl);
+
+    // TODO: do we need sync here?
+    cl_stream.finish();
+
+    return ev;
+}
+
+void* gpu_image2d::lock(const stream& stream) {
+    auto& cl_stream = downcast<const ocl_stream>(stream);
+    std::lock_guard<std::mutex> locker(_mutex);
+    if (0 == _lock_count) {
+        _mapped_ptr = cl_stream.get_cl_queue()
+                          .enqueueMapImage(_buffer,
+                                           CL_TRUE,
+                                           CL_MAP_WRITE,
+                                           {0, 0, 0},
+                                           {_width, _height, 1},
+                                           &_row_pitch,
+                                           &_slice_pitch);
+    }
+    _lock_count++;
+    return _mapped_ptr;
+}
+
+void gpu_image2d::unlock(const stream& stream) {
+    auto& cl_stream = downcast<const ocl_stream>(stream);
+    std::lock_guard<std::mutex> locker(_mutex);
+    _lock_count--;
+    if (0 == _lock_count) {
+        cl_stream.get_cl_queue().enqueueUnmapMemObject(_buffer, _mapped_ptr);
+        _mapped_ptr = nullptr;
+    }
+}
+
+
+shared_mem_params gpu_image2d::get_internal_params() const {
+    auto cl_engine = downcast<const ocl_engine>(_engine);
+    return {shared_mem_type::shared_mem_image, static_cast<shared_handle>(cl_engine->get_cl_context().get()), nullptr,
+            static_cast<shared_handle>(_buffer.get()),
+#ifdef _WIN32
+        nullptr,
+#else
+        0,
+#endif
+        0};
+}
+
+event::ptr gpu_image2d::copy_from(stream& /* stream */, const memory& /* other */) {
+    throw std::runtime_error("[clDNN] copy_from is not implemented for gpu_image2d");
+}
+
+event::ptr gpu_image2d::copy_from(stream& /* stream */, const void* /* host_ptr */) {
+    throw std::runtime_error("[clDNN] copy_from is not implemented for gpu_image2d");
+}
+
+gpu_media_buffer::gpu_media_buffer(ocl_engine* engine,
+                                   const layout& new_layout,
+                                   shared_mem_params params)
+    : gpu_image2d(engine, new_layout, cl::ImageVA(engine->get_cl_context(), CL_MEM_READ_WRITE, params.surface, params.plane)),
+    device(params.user_device),
+    surface(params.surface),
+    plane(params.plane) { }
+
+shared_mem_params gpu_media_buffer::get_internal_params() const {
+    auto cl_engine = downcast<const ocl_engine>(_engine);
+    return {shared_mem_type::shared_mem_vasurface, static_cast<shared_handle>(cl_engine->get_cl_context().get()), device,
+            static_cast<shared_handle>(_buffer.get()), surface, plane };
+}
+
+#ifdef _WIN32
+gpu_dx_buffer::gpu_dx_buffer(ocl_engine* engine,
+                             const layout& new_layout,
+                             shared_mem_params params)
+    : gpu_buffer(engine, new_layout,
+                cl::BufferDX(engine->get_cl_context(), CL_MEM_READ_WRITE, params.mem)),
+    device(params.user_device),
+    resource(params.mem) { }
+
+shared_mem_params gpu_dx_buffer::get_internal_params() const {
+    auto cl_engine = downcast<const ocl_engine>(_engine);
+    return {shared_mem_type::shared_mem_dxbuffer, static_cast<shared_handle>(cl_engine->get_cl_context().get()), device,
+            static_cast<shared_handle>(_buffer.get()), resource, 0 };
+}
+#endif
+
+gpu_usm::gpu_usm(ocl_engine* engine,
+                 const layout& new_layout, const cl::UsmMemory& buffer,
+                 allocation_type type)
+    : lockable_gpu_mem()
+    , memory(engine, new_layout, type, true)
+    , _buffer(buffer) {
+}
+
+gpu_usm::gpu_usm(ocl_engine* engine, const layout& layout, allocation_type type)
+    : lockable_gpu_mem()
+    , memory(engine, layout, type, false)
+    , _buffer(engine->get_cl_context()) {
+    auto device = engine->get_cl_device();
+    switch (get_allocation_type()) {
+    case allocation_type::usm_host:
+        _buffer.allocateHost(_bytes_count);
+        break;
+    case allocation_type::usm_shared:
+        _buffer.allocateShared(device, _bytes_count);
+        break;
+    case allocation_type::usm_device:
+        _buffer.allocateDevice(device, _bytes_count);
+        break;
+    default:
+        CLDNN_ERROR_MESSAGE("gpu_usm allocation type",
+            "Unknown unified shared memory type!");
+    }
+}
+
+void* gpu_usm::lock(const stream& stream) {
+    assert(get_allocation_type() != allocation_type::usm_device && "Can't lock usm device memory!");
+    std::lock_guard<std::mutex> locker(_mutex);
+    if (0 == _lock_count) {
+        stream.finish();  // Synchronization needed for OOOQ.
+        _mapped_ptr = _buffer.get();
+    }
+    _lock_count++;
+    return _mapped_ptr;
+}
+
+void gpu_usm::unlock(const stream& /* stream */) {
+    std::lock_guard<std::mutex> locker(_mutex);
+    _lock_count--;
+    if (0 == _lock_count) {
+        _mapped_ptr = nullptr;
+    }
+}
+
+event::ptr gpu_usm::fill(stream& stream, unsigned char pattern) {
+    auto& cl_stream = downcast<ocl_stream>(stream);
+    auto ev = stream.create_base_event();
+    cl::Event ev_ocl = downcast<base_event>(ev.get())->get();
+    // enqueueFillUsm call will never finish. Driver bug? Uncomment when fixed. Some older drivers doesn't support enqueueFillUsm call at all.
+    // cl_stream.get_cl_queue().enqueueFillUsm<unsigned char>(_buffer, pattern, _bytes_count, nullptr, &ev_ocl)
+    // Workarounded with enqeue_memcopy. ToDo: Remove below code. Uncomment above.
+    std::vector<unsigned char> temp_buffer(_bytes_count, pattern);
+    // TODO: Do we really need blocking call here? Non-blocking one causes accuracy issues right now, but hopefully it can be fixed in more performant way.
+    const bool blocking = true;
+    cl::usm::enqueue_memcpy(cl_stream.get_cl_queue(), _buffer.get(), temp_buffer.data(), _bytes_count, blocking, nullptr, &ev_ocl);
+
+    return ev;
+}
+
+event::ptr gpu_usm::fill(stream& stream) {
+    // event::ptr ev{ new base_event(_context), false };
+    // cl::Event ev_ocl = downcast<base_event>(ev.get())->get();
+    // cl::usm::enqueue_set_mem(cl_stream.get_cl_queue(), _buffer.get(), 0, _bytes_count, nullptr, &ev_ocl);
+    // ev->wait();
+
+    // [WA]
+    return fill(stream, 0);
+}
+
+event::ptr gpu_usm::copy_from(stream& stream, const memory& other) {
+    auto& cl_stream = downcast<const ocl_stream>(stream);
+    auto& casted = downcast<const gpu_usm>(other);
+    cl_stream.get_cl_queue().enqueueCopyUsm(casted.get_buffer(), get_buffer(), _bytes_count, true);
+    return stream.create_user_event(true);
+}
+
+event::ptr gpu_usm::copy_from(stream& /* stream */, const void* /* host_ptr */) {
+    throw std::runtime_error("[clDNN] copy_from is not implemented for gpu_usm");
+}
+
+shared_mem_params gpu_usm::get_internal_params() const {
+    auto cl_engine = downcast<const ocl_engine>(_engine);
+    return {
+        shared_mem_type::shared_mem_empty,  // shared_mem_type
+        static_cast<shared_handle>(cl_engine->get_cl_context().get()),  // context handle
+        nullptr,  // user_device handle
+        nullptr,  // mem handle
+#ifdef _WIN32
+        nullptr,  // surface handle
+#else
+        0,  // surface handle
+#endif
+        0  // plane
+    };
+}
+
+std::vector<cl_mem> ocl_surfaces_lock::get_handles(std::vector<memory::ptr> mem) const {
+    std::vector<cl_mem> res;
+    for (auto& m : mem) {
+        auto mem_type = m->get_internal_params().mem_type;
+        if (mem_type == shared_mem_type::shared_mem_vasurface || mem_type == shared_mem_type::shared_mem_dxbuffer) {
+            res.push_back(static_cast<cl_mem>(m->get_internal_params().mem));
+        }
+    }
+
+    return res;
+}
+
+ocl_surfaces_lock::ocl_surfaces_lock(std::vector<memory::ptr> mem, const stream& stream)
+    : surfaces_lock()
+    , _stream(stream)
+    , _handles(get_handles(mem))
+    , _lock(nullptr) {
+    cl_int err = CL_SUCCESS;
+
+    auto& cl_stream = downcast<const ocl_stream>(stream);
+    auto queue = cl_stream.get_cl_queue();
+    _lock.reset(new cl::SharedSurfLock(queue.get(), _handles, &err));
+    // TODO: err code for some reason is 32766
+    if (/* err != CL_SUCCESS || */ !_lock) {
+        throw std::runtime_error("Unable to lock shared surface (" + std::to_string(err) + ")");
+    }
+}
+
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_memory.hpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_memory.hpp
new file mode 100644
index 00000000000000..f7590e366b880e
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_memory.hpp
@@ -0,0 +1,128 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "ocl_common.hpp"
+#include "ocl_engine.hpp"
+#include "ocl_stream.hpp"
+#include "cldnn/runtime/memory.hpp"
+
+#include <cassert>
+#include <iterator>
+#include <mutex>
+#include <memory>
+
+namespace cldnn {
+namespace ocl {
+struct lockable_gpu_mem {
+    lockable_gpu_mem() :
+        _lock_count(0),
+        _mapped_ptr(nullptr) {}
+
+    std::mutex _mutex;
+    unsigned _lock_count;
+    void* _mapped_ptr;
+};
+
+struct gpu_buffer : public lockable_gpu_mem, public memory {
+    gpu_buffer(ocl_engine* engine, const layout& new_layout, const cl::Buffer& buffer);
+    gpu_buffer(ocl_engine* engine, const layout& layout);
+
+    void* lock(const stream& stream) override;
+    void unlock(const stream& stream) override;
+    event::ptr fill(stream& stream, unsigned char pattern) override;
+    event::ptr fill(stream& stream) override;
+    shared_mem_params get_internal_params() const override;
+    const cl::Buffer& get_buffer() const {
+        assert(0 == _lock_count);
+        return _buffer;
+    }
+
+    event::ptr copy_from(stream& stream, const memory& other) override;
+    event::ptr copy_from(stream& stream, const void* host_ptr) override;
+
+protected:
+    cl::Buffer _buffer;
+};
+
+struct gpu_image2d : public lockable_gpu_mem, public memory {
+    gpu_image2d(ocl_engine* engine, const layout& new_layout, const cl::Image2D& buffer);
+    gpu_image2d(ocl_engine* engine, const layout& layout);
+
+    void* lock(const stream& stream) override;
+    void unlock(const stream& stream) override;
+    event::ptr fill(stream& stream, unsigned char pattern) override;
+    event::ptr fill(stream& stream) override;
+    shared_mem_params get_internal_params() const override;
+    const cl::Image2D& get_buffer() const {
+        assert(0 == _lock_count);
+        return _buffer;
+    }
+
+    event::ptr copy_from(stream& /* stream */, const memory& /* other */) override;
+    event::ptr copy_from(stream& /* stream */, const void* /* other */) override;
+
+protected:
+    cl::Image2D _buffer;
+    size_t _width;
+    size_t _height;
+    size_t _row_pitch;
+    size_t _slice_pitch;
+};
+
+struct gpu_media_buffer : public gpu_image2d {
+    gpu_media_buffer(ocl_engine* engine, const layout& new_layout, shared_mem_params params);
+    shared_mem_params get_internal_params() const override;
+private:
+    void* device;
+#ifdef _WIN32
+    void* surface;
+#else
+    uint32_t surface;
+#endif
+    uint32_t plane;
+};
+
+#ifdef _WIN32
+struct gpu_dx_buffer : public gpu_buffer {
+    gpu_dx_buffer(ocl_engine* engine, const layout& new_layout, shared_mem_params VAEncMiscParameterTypeSubMbPartPel);
+    shared_mem_params get_internal_params() const override;
+private:
+    void* device;
+    void* resource;
+};
+#endif
+
+struct gpu_usm : public lockable_gpu_mem, public memory {
+    gpu_usm(ocl_engine* engine, const layout& new_layout, const cl::UsmMemory& usm_buffer, allocation_type type);
+    gpu_usm(ocl_engine* engine, const layout& layout, allocation_type type);
+
+    void* lock(const stream& stream) override;
+    void unlock(const stream& stream) override;
+    const cl::UsmMemory& get_buffer() const { return _buffer; }
+    cl::UsmMemory& get_buffer() { return _buffer; }
+
+    event::ptr fill(stream& stream, unsigned char pattern) override;
+    event::ptr fill(stream& stream) override;
+    shared_mem_params get_internal_params() const override;
+
+    event::ptr copy_from(stream& stream, const memory& other) override;
+    event::ptr copy_from(stream& stream, const void* host_ptr) override;
+protected:
+    cl::UsmMemory _buffer;
+};
+
+struct ocl_surfaces_lock : public surfaces_lock {
+    ocl_surfaces_lock(std::vector<memory::ptr> mem, const stream& stream);
+
+    ~ocl_surfaces_lock() = default;
+private:
+    std::vector<cl_mem> get_handles(std::vector<memory::ptr> mem) const;
+    const stream& _stream;
+    std::vector<cl_mem> _handles;
+    std::unique_ptr<cl::SharedSurfLock> _lock;
+};
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_stream.cpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_stream.cpp
new file mode 100644
index 00000000000000..508f2114214a2f
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_stream.cpp
@@ -0,0 +1,428 @@
+// Copyright (C) 2019-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "ocl_stream.hpp"
+#include "ocl_base_event.hpp"
+#include "ocl_user_event.hpp"
+#include "ocl_command_queues_builder.hpp"
+#include "ocl_events_pool.hpp"
+#include "ocl_kernel.hpp"
+#include "ocl_common.hpp"
+
+#include <cassert>
+#include <iomanip>
+#include <ios>
+
+#include <fstream>
+#include <thread>
+#include <string>
+#include <vector>
+#include <memory>
+
+// NOTE: Due to buggy scope transition of warnings we need to disable warning in place of use/instantation
+//       of some types (even though we already disabled them in scope of definition of these types).
+//       Moreover this warning is pretty much now only for annoyance: it is generated due to lack
+//       of proper support for mangling of custom GCC attributes into type name (usually when used
+//       with templates, even from standard library).
+#if defined __GNUC__ && __GNUC__ >= 6
+#pragma GCC diagnostic ignored "-Wignored-attributes"
+#endif
+
+namespace cldnn {
+namespace ocl {
+
+namespace {
+inline cl::NDRange toNDRange(const std::vector<size_t>& v) {
+    switch (v.size()) {
+        case 1:
+            return cl::NDRange(v[0]);
+        case 2:
+            return cl::NDRange(v[0], v[1]);
+        case 3:
+            return cl::NDRange(v[0], v[1], v[2]);
+        default:
+            return cl::NullRange;
+    }
+}
+
+void set_arguments_impl(ocl_kernel_type& kernel,
+                        const arguments_desc& args,
+                        const kernel_arguments_data& data) {
+    using args_t = argument_desc::Types;
+    using scalar_t = scalar_desc::Types;
+    for (uint32_t i = 0; i < static_cast<uint32_t>(args.size()); i++) {
+        cl_int status = CL_INVALID_ARG_VALUE;
+        switch (args[i].t) {
+            case args_t::INPUT:
+                if (args[i].index < data.inputs.size() && data.inputs[args[i].index]) {
+                    const auto& input_mem = data.inputs[args[i].index];
+                    if (input_mem) {
+                        if (input_mem->get_layout().format.is_image_2d())
+                            status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_image2d>(input_mem)->get_buffer());
+                        else if (memory_capabilities::is_usm_type(input_mem->get_allocation_type()))
+                            status = kernel.setArgUsm(i, std::dynamic_pointer_cast<const ocl::gpu_usm>(input_mem)->get_buffer());
+                        else
+                            status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_buffer>(input_mem)->get_buffer());
+                    }
+                }
+                break;
+            case args_t::INPUT_OF_FUSED_PRIMITIVE:
+                if (args[i].index < data.fused_op_inputs.size() && data.fused_op_inputs[args[i].index]) {
+                    const auto& input_mem = data.fused_op_inputs[args[i].index];
+                    if (input_mem) {
+                        if (memory_capabilities::is_usm_type(input_mem->get_allocation_type()))
+                            status = kernel.setArgUsm(i, std::dynamic_pointer_cast<const ocl::gpu_usm>(input_mem)->get_buffer());
+                        else
+                            status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_buffer>(input_mem)->get_buffer());
+                    }
+                }
+                break;
+            case args_t::INTERNAL_BUFFER:
+                if (args[i].index < data.intermediates.size() && data.intermediates[args[i].index]) {
+                    const auto& input_mem = data.intermediates[args[i].index];
+                    if (input_mem) {
+                        if (memory_capabilities::is_usm_type(input_mem->get_allocation_type()))
+                            status = kernel.setArgUsm(i, std::dynamic_pointer_cast<const ocl::gpu_usm>(input_mem)->get_buffer());
+                        else
+                            status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_buffer>(input_mem)->get_buffer());
+                    }
+                }
+                break;
+            case args_t::OUTPUT:
+                if (data.output) {
+                     if (data.output->get_layout().format.is_image_2d())
+                        status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_image2d>(data.output)->get_buffer());
+                     else if (memory_capabilities::is_usm_type(data.output->get_allocation_type()))
+                         status = kernel.setArgUsm(i, std::dynamic_pointer_cast<const ocl::gpu_usm>(data.output)->get_buffer());
+                     else
+                        status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_buffer>(data.output)->get_buffer());
+                }
+                break;
+            case args_t::WEIGHTS:
+                if (data.weights) {
+                    if (data.weights->get_layout().format.is_image_2d())
+                        status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_image2d>(data.weights)->get_buffer());
+                    else if (memory_capabilities::is_usm_type(data.weights->get_allocation_type()))
+                        status = kernel.setArgUsm(i, std::dynamic_pointer_cast<const ocl::gpu_usm>(data.weights)->get_buffer());
+                    else
+                        status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_buffer>(data.weights)->get_buffer());
+                }
+                break;
+            case args_t::BIAS:
+                if (data.bias) {
+                    if (memory_capabilities::is_usm_type(data.bias->get_allocation_type()))
+                        status = kernel.setArgUsm(i, std::dynamic_pointer_cast<const ocl::gpu_usm>(data.bias)->get_buffer());
+                    else
+                        status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_buffer>(data.bias)->get_buffer());
+                }
+                break;
+            case args_t::WEIGHTS_ZERO_POINTS:
+                if (data.weights_zero_points) {
+                    if (memory_capabilities::is_usm_type(data.weights_zero_points->get_allocation_type()))
+                        status = kernel.setArgUsm(
+                            i,
+                            std::dynamic_pointer_cast<const ocl::gpu_usm>(data.weights_zero_points)->get_buffer());
+                    else
+                        status = kernel.setArg(
+                            i,
+                            std::dynamic_pointer_cast<const ocl::gpu_buffer>(data.weights_zero_points)->get_buffer());
+                }
+                break;
+            case args_t::ACTIVATIONS_ZERO_POINTS:
+                if (data.activations_zero_points) {
+                    if (memory_capabilities::is_usm_type(data.activations_zero_points->get_allocation_type()))
+                        status = kernel.setArgUsm(
+                            i,
+                            std::dynamic_pointer_cast<const ocl::gpu_usm>(data.activations_zero_points)->get_buffer());
+                    else
+                        status = kernel.setArg(
+                            i,
+                            std::dynamic_pointer_cast<const ocl::gpu_buffer>(data.activations_zero_points)->get_buffer());
+                }
+                break;
+            case args_t::COMPENSATION:
+                if (data.compensation) {
+                    if (memory_capabilities::is_usm_type(data.compensation->get_allocation_type()))
+                        status = kernel.setArgUsm(
+                                i,
+                                std::dynamic_pointer_cast<const ocl::gpu_usm>(data.compensation)->get_buffer());
+                    else
+                        status = kernel.setArg(
+                                 i,
+                                 std::dynamic_pointer_cast<const ocl::gpu_buffer>(data.compensation)->get_buffer());
+                }
+                break;
+            case args_t::SCALE_TABLE:
+                if (data.scale_table) {
+                    if (memory_capabilities::is_usm_type(data.scale_table->get_allocation_type()))
+                        status = kernel.setArgUsm(i, std::dynamic_pointer_cast<const ocl::gpu_usm>(data.scale_table)->get_buffer());
+                    else
+                        status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_buffer>(data.scale_table)->get_buffer());
+                }
+                break;
+            case args_t::SLOPE:
+                if (data.slope) {
+                    if (memory_capabilities::is_usm_type(data.slope->get_allocation_type()))
+                        status = kernel.setArgUsm(i, std::dynamic_pointer_cast<const ocl::gpu_usm>(data.slope)->get_buffer());
+                    else
+                        status = kernel.setArg(i, std::dynamic_pointer_cast<const ocl::gpu_buffer>(data.slope)->get_buffer());
+                }
+                break;
+            case args_t::SPLIT:
+                status = kernel.setArg(i, data.split);
+                break;
+            case args_t::SCALAR:
+                if (data.scalars && args[i].index < data.scalars->size()) {
+                    const auto& scalar = (*data.scalars)[args[i].index];
+                    switch (scalar.t) {
+                        case scalar_t::UINT8:
+                            status = kernel.setArg(i, scalar.v.u8);
+                            break;
+                        case scalar_t::UINT16:
+                            status = kernel.setArg(i, scalar.v.u16);
+                            break;
+                        case scalar_t::UINT32:
+                            status = kernel.setArg(i, scalar.v.u32);
+                            break;
+                        case scalar_t::UINT64:
+                            status = kernel.setArg(i, scalar.v.u64);
+                            break;
+                        case scalar_t::INT8:
+                            status = kernel.setArg(i, scalar.v.s8);
+                            break;
+                        case scalar_t::INT16:
+                            status = kernel.setArg(i, scalar.v.s16);
+                            break;
+                        case scalar_t::INT32:
+                            status = kernel.setArg(i, scalar.v.s32);
+                            break;
+                        case scalar_t::INT64:
+                            status = kernel.setArg(i, scalar.v.s64);
+                            break;
+                        case scalar_t::FLOAT32:
+                            status = kernel.setArg(i, scalar.v.f32);
+                            break;
+                        case scalar_t::FLOAT64:
+                            status = kernel.setArg(i, scalar.v.f64);
+                            break;
+                        default:
+                            break;
+                    }
+                }
+                break;
+            case args_t::RECURRENT:  // RNN/LSTM/GRU layers
+                if (data.recurrent) {
+                    if (data.recurrent->get_layout().format.is_image_2d())
+                        status = kernel.setArg(i, dynamic_cast<const ocl::gpu_image2d&>(*data.recurrent).get_buffer());
+                    else if (memory_capabilities::is_usm_type(data.recurrent->get_allocation_type()))
+                        status = kernel.setArgUsm(i, dynamic_cast<const ocl::gpu_usm&>(*data.recurrent).get_buffer());
+                    else
+                        status = kernel.setArg(i, dynamic_cast<const ocl::gpu_buffer&>(*data.recurrent).get_buffer());
+                }
+                break;
+            case args_t::HIDDEN:  // RNN/LSTM/GRU layers
+                if (data.hidden) {
+                    if (data.hidden->get_layout().format.is_image_2d())
+                        status = kernel.setArg(i, dynamic_cast<const ocl::gpu_image2d&>(*data.hidden).get_buffer());
+                    else if (memory_capabilities::is_usm_type(data.hidden->get_allocation_type()))
+                        status = kernel.setArgUsm(i, dynamic_cast<const ocl::gpu_usm&>(*data.hidden).get_buffer());
+                    else
+                        status = kernel.setArg(i, dynamic_cast<const ocl::gpu_buffer&>(*data.hidden).get_buffer());
+                }
+                break;
+            case args_t::CELL:  // LSTMlayers
+                if (data.cell) {
+                    if (data.cell->get_layout().format.is_image_2d())
+                        status = kernel.setArg(i, dynamic_cast<const ocl::gpu_image2d&>(*data.cell).get_buffer());
+                    else if (memory_capabilities::is_usm_type(data.cell->get_allocation_type()))
+                        status = kernel.setArgUsm(i, dynamic_cast<const ocl::gpu_usm&>(*data.cell).get_buffer());
+                    else
+                        status = kernel.setArg(i, dynamic_cast<const ocl::gpu_buffer&>(*data.cell).get_buffer());
+                }
+                break;
+            default:
+                break;
+        }
+
+        if (status != CL_SUCCESS) {
+            throw std::runtime_error("Error set arg " + std::to_string(i) + ", error code: " + std::to_string(status) + "\n");
+        }
+    }
+}
+}  // namespace
+
+ocl_stream::ocl_stream(const ocl_engine& engine) : stream(engine.configuration().queue_type), _engine(engine) {
+    auto context = engine.get_cl_context();
+    auto device = engine.get_cl_device();
+    auto config = engine.configuration();
+    ocl::command_queues_builder queue_builder;
+    queue_builder.set_profiling(config.enable_profiling);
+    queue_builder.set_out_of_order((config.queue_type == queue_types::out_of_order));
+
+    sync_method = _engine.configuration().enable_profiling ? sync_methods::events :
+                  config.queue_type == queue_types::out_of_order ? sync_methods::barriers : sync_methods::none;
+
+    if (sync_method == sync_methods::none && config.queue_type == queue_types::out_of_order) {
+        throw std::runtime_error("[CLDNN] Unexpected sync method (none) is specified for out_of_order queue");
+    }
+
+    bool priorty_extensions = engine.extension_supported("cl_khr_priority_hints") && engine.extension_supported("cl_khr_create_command_queue");
+    queue_builder.set_priority_mode(config.priority_mode, priorty_extensions);
+
+    bool throttle_extensions = engine.extension_supported("cl_khr_throttle_hints") && engine.extension_supported("cl_khr_create_command_queue");
+    queue_builder.set_throttle_mode(config.throttle_mode, throttle_extensions);
+
+    _command_queue = queue_builder.build(context, device);
+    _events_pool.reset(new events_pool());
+}
+
+void ocl_stream::set_arguments(kernel& kernel, const kernel_arguments_desc& args_desc, const kernel_arguments_data& args) {
+    static std::mutex m;
+    std::lock_guard<std::mutex> guard(m);
+
+    auto& ocl_kernel = downcast<ocl::ocl_kernel>(kernel);
+
+    auto& kern = ocl_kernel.get_handle();
+
+    try {
+        set_arguments_impl(kern, args_desc.arguments, args);
+    } catch (cl::Error const& err) {
+        throw ocl_error(err);
+    }
+}
+
+event::ptr ocl_stream::enqueue_kernel(kernel& kernel,
+                                      const kernel_arguments_desc& args_desc,
+                                      const kernel_arguments_data& /* args */,
+                                      std::vector<event::ptr> const& deps,
+                                      bool is_output) {
+    auto& ocl_kernel = downcast<ocl::ocl_kernel>(kernel);
+
+    auto& kern = ocl_kernel.get_handle();
+    auto global = toNDRange(args_desc.workGroups.global);
+    auto local = toNDRange(args_desc.workGroups.local);
+    std::vector<cl::Event> dep_events;
+    std::vector<cl::Event>* dep_events_ptr = nullptr;
+    if (sync_method == sync_methods::events) {
+        for (auto& dep : deps) {
+            if (auto ocl_base_ev = std::dynamic_pointer_cast<ocl_base_event>(dep)) {
+                if (ocl_base_ev->get().get() != nullptr)
+                    dep_events.push_back(ocl_base_ev->get());
+            }
+        }
+        dep_events_ptr = &dep_events;
+    } else if (sync_method == sync_methods::barriers) {
+        sync_events(deps, is_output);
+    }
+
+    cl::Event ret_ev;
+
+    bool set_output_event = sync_method == sync_methods::events || is_output;
+
+    try {
+        _command_queue.enqueueNDRangeKernel(kern, cl::NullRange, global, local, dep_events_ptr, set_output_event ? &ret_ev : nullptr);
+    } catch (cl::Error const& err) {
+        throw ocl_error(err);
+    }
+
+    return _events_pool->get_from_base_pool(_engine.get_cl_context(), ret_ev, ++_queue_counter);
+}
+
+void ocl_stream::enqueue_barrier() {
+    _command_queue.enqueueBarrierWithWaitList(nullptr, nullptr);
+}
+
+event::ptr ocl_stream::enqueue_marker(std::vector<event::ptr> const& deps, bool is_output) {
+    if (deps.empty())
+        return _events_pool->get_from_user_pool(_engine.get_cl_context(), true);
+
+    if (sync_method == sync_methods::events) {
+        cl::Event ret_ev;
+        std::vector<cl::Event> dep_events;
+        for (auto& dep : deps) {
+            if (auto ocl_base_ev = dynamic_cast<ocl_base_event*>(dep.get()))
+                if (ocl_base_ev->get().get() != nullptr)
+                    dep_events.push_back(ocl_base_ev->get());
+        }
+
+        try {
+            if (dep_events.empty()) {
+                return create_user_event(true);
+            }
+            _command_queue.enqueueMarkerWithWaitList(&dep_events, &ret_ev);
+        } catch (cl::Error const& err) {
+            throw ocl_error(err);
+        }
+
+        return _events_pool->get_from_base_pool(_engine.get_cl_context(), ret_ev, ++_queue_counter);
+    } else if (sync_method == sync_methods::barriers) {
+        sync_events(deps, is_output);
+        return _events_pool->get_from_base_pool(_engine.get_cl_context(), _last_barrier_ev, _last_barrier);
+    } else {
+        return _events_pool->get_from_user_pool(_engine.get_cl_context(), true);
+    }
+}
+
+event::ptr ocl_stream::group_events(std::vector<event::ptr> const& deps) {
+    return _events_pool->get_from_group_pool(_engine.get_cl_context(), deps);
+}
+
+event::ptr ocl_stream::create_user_event(bool set) {
+    return _events_pool->get_from_user_pool(_engine.get_cl_context(), set);
+}
+
+event::ptr ocl_stream::create_base_event() {
+    cl::Event ret_ev;
+    return _events_pool->get_from_base_pool(_engine.get_cl_context(), ret_ev, ++_queue_counter);
+}
+
+void ocl_stream::reset_events() { _events_pool->reset_events(); }
+
+void ocl_stream::release_events_pool() { _events_pool.reset(); }
+
+void ocl_stream::flush() const { get_cl_queue().flush(); }
+void ocl_stream::finish() const { get_cl_queue().finish(); }
+
+void ocl_stream::wait_for_events(const std::vector<event::ptr>& events) {
+    if (events.empty())
+        return;
+
+    std::vector<cl::Event> clevents;
+    for (auto& ev : events) {
+        if (auto ocl_base_ev = dynamic_cast<ocl_base_event*>(ev.get()))
+            clevents.push_back(ocl_base_ev->get());
+    }
+
+    try {
+        cl::WaitForEvents(clevents);
+    } catch (cl::Error const& err) {
+        throw ocl_error(err);
+    }
+}
+
+void ocl_stream::sync_events(std::vector<event::ptr> const& deps, bool is_output) {
+    bool needs_barrier = false;
+    for (auto& dep : deps) {
+        auto* ocl_base_ev = dynamic_cast<ocl_base_event*>(dep.get());
+        if (ocl_base_ev->get_queue_stamp() > _last_barrier) {
+            needs_barrier = true;
+        }
+    }
+
+    if (needs_barrier) {
+        try {
+            if (is_output)
+                _command_queue.enqueueBarrierWithWaitList(nullptr, &_last_barrier_ev);
+            else
+                _command_queue.enqueueBarrierWithWaitList(nullptr, nullptr);
+        } catch (cl::Error const& err) {
+            throw ocl_error(err);
+        }
+
+        _last_barrier = ++_queue_counter;
+    }
+}
+
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_stream.hpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_stream.hpp
new file mode 100644
index 00000000000000..3545d6d17e5b55
--- /dev/null
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_stream.hpp
@@ -0,0 +1,99 @@
+// Copyright (C) 2019-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "cldnn/runtime/event.hpp"
+#include "cldnn/runtime/stream.hpp"
+#include "ocl_common.hpp"
+#include "ocl_engine.hpp"
+
+#include <memory>
+#include <chrono>
+#include <thread>
+#include <iostream>
+#include <sstream>
+#include <utility>
+#include <vector>
+
+namespace cldnn {
+namespace ocl {
+
+class events_pool;
+
+// Possible sync methods for kernels in stream
+enum class sync_methods {
+    /* Build dependency graph using events. Each kernel creates proper cl_event which is set as dependency of users
+       At this moment it requires multiple retain/release calls for cl_event after each enqueueNDRange
+       which is less performant comparing to the barriers version
+    */
+    events = 0,
+    /* Enqueue barriers between dependent kernels. For example consider the following dimond dependency graph:
+             kernel_0
+             /      \
+         kernel_1   kernel_2
+             \      /
+             kernel_3
+       In that case we do the following:
+       1. Enqueue kernel_0
+       2. Enqueue barrier (ensures kernel_0 is completed)
+       3. Enqueue kernel_1
+       4. Enqueue kernel_2 (doesn't depend on kernel_1)
+       5. Enqueue barrier (ensures kernel_1 and kernel_2 are completed)
+       6. Enqueue kernel_3
+    */
+    barriers = 1,
+    /* No explicit syncronization is needed. Applicable for in-order queue only */
+    none = 2
+};
+class ocl_stream : public stream {
+public:
+    const ocl_queue_type& get_cl_queue() const { return _command_queue; }
+
+    explicit ocl_stream(const ocl_engine& engine);
+    ocl_stream(ocl_stream&& other)
+        : stream(other._engine.configuration().queue_type)
+        , _engine(other._engine)
+        , _command_queue(other._command_queue)
+        , _queue_counter(other._queue_counter.load())
+        , _last_barrier(other._last_barrier.load())
+        , _events_pool(std::move(other._events_pool))
+        , _last_barrier_ev(other._last_barrier_ev)
+        , sync_method(other.sync_method) {}
+
+    ~ocl_stream() = default;
+
+    void flush() const override;
+    void finish() const override;
+
+    void set_arguments(kernel& kernel, const kernel_arguments_desc& args_desc, const kernel_arguments_data& args) override;
+    event::ptr enqueue_kernel(kernel& kernel,
+                              const kernel_arguments_desc& args_desc,
+                              const kernel_arguments_data& args,
+                              std::vector<event::ptr> const& deps,
+                              bool is_output = false) override;
+    event::ptr enqueue_marker(std::vector<event::ptr> const& deps, bool is_output) override;
+    event::ptr group_events(std::vector<event::ptr> const& deps) override;
+    void wait_for_events(const std::vector<event::ptr>& events) override;
+    void enqueue_barrier() override;
+    void reset_events() override;
+    event::ptr create_user_event(bool set) override;
+    event::ptr create_base_event() override;
+    void release_events_pool() override;
+
+private:
+    void sync_events(std::vector<event::ptr> const& deps, bool is_output = false);
+
+    const ocl_engine& _engine;
+    ocl_queue_type _command_queue;
+    std::atomic<uint64_t> _queue_counter{0};
+    std::atomic<uint64_t> _last_barrier{0};
+    std::shared_ptr<events_pool> _events_pool;
+    cl::Event _last_barrier_ev;
+
+    sync_methods sync_method;
+};
+
+}  // namespace ocl
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ocl_user_event.cpp b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_user_event.cpp
similarity index 94%
rename from inference-engine/thirdparty/clDNN/src/gpu/ocl_user_event.cpp
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_user_event.cpp
index c5663fb549d696..86da8d90bbe3ea 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/ocl_user_event.cpp
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_user_event.cpp
@@ -2,10 +2,10 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "ocl_user_event.h"
+#include "ocl_user_event.hpp"
 #include <list>
 
-using namespace cldnn::gpu;
+using namespace cldnn::ocl;
 
 void user_event::set_impl() {
     // we simulate "wrapper_cast" here to cast from cl::Event to cl::UserEvent which both wrap the same cl_event
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ocl_user_event.h b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_user_event.hpp
similarity index 71%
rename from inference-engine/thirdparty/clDNN/src/gpu/ocl_user_event.h
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_user_event.hpp
index 45e44509face3a..199a8a8a26010f 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/ocl_user_event.h
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_user_event.hpp
@@ -4,8 +4,8 @@
 
 #pragma once
 
-#include "ocl_base_event.h"
-#include "api/profiling.hpp"
+#include "cldnn/runtime/profiling.hpp"
+#include "ocl_base_event.hpp"
 #include <memory>
 #include <list>
 
@@ -16,14 +16,15 @@
 #endif
 
 namespace cldnn {
-namespace gpu {
+namespace ocl {
 
 struct user_event : public base_event, public cldnn::user_event {
-    explicit user_event(std::shared_ptr<gpu_toolkit> ctx) : base_event(ctx), cldnn::user_event(false) {}
+    explicit user_event(const cl::Context& ctx) : base_event(ctx, cl::UserEvent(ctx)), cldnn::user_event(false), _ctx(ctx) {}
 
     void set_impl() override;
     void attach_event(bool set) {
-        _event = cl::UserEvent(get_context()->context());
+        // Event handle must be created again as clSetUserEventStatus can't be called twice for the same object
+        _event = cl::UserEvent(_ctx);
         // we need to reset the timer(since attach_ocl_event is called only when this object is being reused)
         _timer = cldnn::instrumentation::timer<>();
         if (set) {
@@ -36,11 +37,12 @@ struct user_event : public base_event, public cldnn::user_event {
 protected:
     cldnn::instrumentation::timer<> _timer;
     std::unique_ptr<cldnn::instrumentation::profiling_period_basic> _duration;
+    const cl::Context& _ctx;
 };
 
 #ifdef _WIN32
 #pragma warning(pop)
 #endif
 
-}  // namespace gpu
+}  // namespace ocl
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/common/include/cl2_wrapper.h b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_wrapper.hpp
similarity index 95%
rename from inference-engine/thirdparty/clDNN/common/include/cl2_wrapper.h
rename to inference-engine/thirdparty/clDNN/runtime/ocl/ocl_wrapper.hpp
index 3beb802304cafc..f75779075ff5f4 100644
--- a/inference-engine/thirdparty/clDNN/common/include/cl2_wrapper.h
+++ b/inference-engine/thirdparty/clDNN/runtime/ocl/ocl_wrapper.hpp
@@ -6,8 +6,7 @@
 /// \file This file wraps cl2.hpp and disables temporary some warnings that this header can emit.
 ///
 
-#ifndef CL2_WRAPPER_H_
-#define CL2_WRAPPER_H_
+#pragma once
 
 // Check for C++.
 #ifndef __cplusplus
@@ -58,7 +57,7 @@
 #endif
 
 
-#include "cl2_ext.hpp"
+#include "ocl_ext.hpp"
 
 namespace cl {
 namespace detail {
@@ -76,5 +75,3 @@ CL_HPP_PARAM_NAME_INFO_1_1_DEPRECATED_IN_2_0_(CL_HPP_DECLARE_PARAM_TRAITS_)
 #elif defined __GNUC__
     #pragma GCC diagnostic pop
 #endif
-
-#endif // CL2_WRAPPER_H_
diff --git a/inference-engine/thirdparty/clDNN/src/CMakeLists.txt b/inference-engine/thirdparty/clDNN/src/CMakeLists.txt
index 0e4bcb0c5d9022..de841fe48bebb1 100644
--- a/inference-engine/thirdparty/clDNN/src/CMakeLists.txt
+++ b/inference-engine/thirdparty/clDNN/src/CMakeLists.txt
@@ -38,16 +38,11 @@ set_property(DIRECTORY APPEND PROPERTY COMPILE_DEFINITIONS
 # ========================================= Source/Header files ========================================
 
 set(__CLDNN_Label__api                 "api")
-file(GLOB __CLDNN_Headers__api
+file(GLOB_RECURSE __CLDNN_Headers__api
     "${CLDNN__API_DIR}/*.h"
     "${CLDNN__API_DIR}/*.hpp"
   )
 
-set(__CLDNN_Label__api_extension       "api_extension")
-file(GLOB __CLDNN_Headers__api_extension
-    "${CLDNN__API_EXTENSION_DIR}/*.hpp"
-  )
-
 set(__CLDNN_Label__main                "")
 file(GLOB __CLDNN_Sources__main
     "${CMAKE_CURRENT_SOURCE_DIR}/*.h"
@@ -108,7 +103,6 @@ set(__CLDNN_AllSources
     ${__CLDNN_Headers__api}
     ${__CLDNN_Sources__graph_opt}
     ${__CLDNN_Headers__include}
-    ${__CLDNN_Headers__api_extension}
     ${__CLDNN_Sources__main}
     ${__CLDNN_Sources__gpu}
     ${__CLDNN_Sources__cache}
@@ -121,7 +115,6 @@ set_property(SOURCE ${__CLDNN_Sources__cg_cache} PROPERTY GENERATED TRUE)
 # =============================================== Filters ==============================================
 
 source_group("${__CLDNN_Label__api}"                  FILES ${__CLDNN_Headers__api})
-source_group("${__CLDNN_Label__api_extension}"        FILES ${__CLDNN_Headers__api_extension})
 source_group("${__CLDNN_Label__include}"              FILES ${__CLDNN_Headers__include})
 source_group("${__CLDNN_Label__graph_opt}"            FILES ${__CLDNN_Sources__graph_opt})
 source_group("${__CLDNN_Label__main}"                 FILES ${__CLDNN_Sources__main})
@@ -135,6 +128,7 @@ source_group("${__CLDNN_Label__cg_cache}"             FILES ${__CLDNN_Sources__c
 include_directories(
     "${CLDNN__MAIN_DIR}"
     "${CMAKE_CURRENT_SOURCE_DIR}"
+    "${CLDNN__RUNTIME_DIR}"
     "${__CLDNN_Directory__include}"
     "${__CLDNN_Directory__ks_core}"
     "${__CLDNN_Directory__ks_core}/common"
@@ -155,6 +149,7 @@ target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE
     clDNN_OpenCL
     cldnn_kernel_selector
     openvino::itt
+    cldnn_runtime
   )
 
 if(COMMAND add_cpplint_target)
@@ -171,3 +166,5 @@ elseif((NOT ANDROID) AND (UNIX))
   target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE pthread)
 endif()
 target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE ${CLDNN__SYSTEM_LINK_LIBRARIES})
+
+# ======================================================================================================
diff --git a/inference-engine/thirdparty/clDNN/src/activation.cpp b/inference-engine/thirdparty/clDNN/src/activation.cpp
index dba83df3583043..0987286ab4e51b 100644
--- a/inference-engine/thirdparty/clDNN/src/activation.cpp
+++ b/inference-engine/thirdparty/clDNN/src/activation.cpp
@@ -4,7 +4,7 @@
 
 #include "activation_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 #include <vector>
diff --git a/inference-engine/thirdparty/clDNN/src/arg_max_min.cpp b/inference-engine/thirdparty/clDNN/src/arg_max_min.cpp
index f4a5a188e61e58..0d2f1f4beae215 100644
--- a/inference-engine/thirdparty/clDNN/src/arg_max_min.cpp
+++ b/inference-engine/thirdparty/clDNN/src/arg_max_min.cpp
@@ -2,11 +2,10 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "arg_max_min_inst.h"
 #include "primitive_type_base.h"
 #include "sliding_window_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 #include <limits>
diff --git a/inference-engine/thirdparty/clDNN/src/average_unpooling.cpp b/inference-engine/thirdparty/clDNN/src/average_unpooling.cpp
index c4ba3ac8418e19..3082862f302e74 100644
--- a/inference-engine/thirdparty/clDNN/src/average_unpooling.cpp
+++ b/inference-engine/thirdparty/clDNN/src/average_unpooling.cpp
@@ -5,7 +5,7 @@
 #include "average_unpooling_inst.h"
 #include "primitive_type_base.h"
 #include "sliding_window_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/batch_to_space.cpp b/inference-engine/thirdparty/clDNN/src/batch_to_space.cpp
index 7540b49e754652..aa34ab8da536ff 100644
--- a/inference-engine/thirdparty/clDNN/src/batch_to_space.cpp
+++ b/inference-engine/thirdparty/clDNN/src/batch_to_space.cpp
@@ -5,7 +5,7 @@
 #include "batch_to_space_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include "data_inst.h"
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/src/binary_convolution.cpp b/inference-engine/thirdparty/clDNN/src/binary_convolution.cpp
index ece928a51d0d87..a2ffcdc02ca996 100644
--- a/inference-engine/thirdparty/clDNN/src/binary_convolution.cpp
+++ b/inference-engine/thirdparty/clDNN/src/binary_convolution.cpp
@@ -8,7 +8,7 @@
 #include "reorder_inst.h"
 #include "primitive_type_base.h"
 #include "sliding_window_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/border.cpp b/inference-engine/thirdparty/clDNN/src/border.cpp
index 810027192e0494..cdeca46d8b1e35 100644
--- a/inference-engine/thirdparty/clDNN/src/border.cpp
+++ b/inference-engine/thirdparty/clDNN/src/border.cpp
@@ -4,7 +4,7 @@
 
 #include "border_inst.h"
 
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include "primitive_type_base.h"
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/src/broadcast.cpp b/inference-engine/thirdparty/clDNN/src/broadcast.cpp
index 790a1222fba116..d6cc15067075e1 100644
--- a/inference-engine/thirdparty/clDNN/src/broadcast.cpp
+++ b/inference-engine/thirdparty/clDNN/src/broadcast.cpp
@@ -4,7 +4,7 @@
 
 #include "broadcast_inst.h"
 
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include "primitive_type_base.h"
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/src/concatenation.cpp b/inference-engine/thirdparty/clDNN/src/concatenation.cpp
index e6e622a93e9f6b..d341566679ee22 100644
--- a/inference-engine/thirdparty/clDNN/src/concatenation.cpp
+++ b/inference-engine/thirdparty/clDNN/src/concatenation.cpp
@@ -4,7 +4,7 @@
 
 #include "concatenation_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 #include <vector>
diff --git a/inference-engine/thirdparty/clDNN/src/condition.cpp b/inference-engine/thirdparty/clDNN/src/condition.cpp
index f6b2d237392309..c31e6cb80e780a 100644
--- a/inference-engine/thirdparty/clDNN/src/condition.cpp
+++ b/inference-engine/thirdparty/clDNN/src/condition.cpp
@@ -4,7 +4,7 @@
 
 #include "condition_inst.h"
 
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include "primitive_type_base.h"
 #include <string>
@@ -71,10 +71,8 @@ Condition primitive is resuing memory with the input.
 */
 condition_inst::typed_primitive_inst(network_impl& network, condition_node const& node)
     : parent(network, node),
-      _net_true(
-          node.get_program().get_engine().allocate_network(*node.get_branch_true(), true)),
-      _net_false(
-          node.get_program().get_engine().allocate_network(*node.get_branch_false(), true)) {
+      _net_true(network_impl::allocate_network(node.get_program().get_engine(), node.get_branch_true(), true)),
+      _net_false(network_impl::allocate_network(node.get_program().get_engine(), node.get_branch_false(), true)) {
     auto compare_tensor = node.compare().get_output_layout().size;
     auto input_tensor = node.input().get_output_layout().size;
     CLDNN_ERROR_TENSOR_SIZES_GREATER_THAN(node.id(),
diff --git a/inference-engine/thirdparty/clDNN/src/convolution.cpp b/inference-engine/thirdparty/clDNN/src/convolution.cpp
index 5b49b614f67117..c145dfc339cec9 100644
--- a/inference-engine/thirdparty/clDNN/src/convolution.cpp
+++ b/inference-engine/thirdparty/clDNN/src/convolution.cpp
@@ -7,7 +7,7 @@
 #include "convolution_inst.h"
 #include "primitive_type_base.h"
 #include "sliding_window_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/crop.cpp b/inference-engine/thirdparty/clDNN/src/crop.cpp
index 4dbcc9ce24727b..7d35ad1e04f3a0 100644
--- a/inference-engine/thirdparty/clDNN/src/crop.cpp
+++ b/inference-engine/thirdparty/clDNN/src/crop.cpp
@@ -4,8 +4,8 @@
 
 #include "crop_inst.h"
 #include "primitive_type_base.h"
-#include "memory_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/cum_sum.cpp b/inference-engine/thirdparty/clDNN/src/cum_sum.cpp
index ebb176ec74591c..b7e0b144b90e44 100644
--- a/inference-engine/thirdparty/clDNN/src/cum_sum.cpp
+++ b/inference-engine/thirdparty/clDNN/src/cum_sum.cpp
@@ -5,7 +5,7 @@
 #include "cum_sum_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/data.cpp b/inference-engine/thirdparty/clDNN/src/data.cpp
index 7d0d89ea3d9603..c08b6a28ce394c 100644
--- a/inference-engine/thirdparty/clDNN/src/data.cpp
+++ b/inference-engine/thirdparty/clDNN/src/data.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "data_inst.h"
 #include "primitive_type_base.h"
-#include "memory_impl.h"
+#include "cldnn/runtime/memory.hpp"
 
 #include "json_object.h"
 #include <string>
@@ -19,28 +19,28 @@ primitive_type_id data::type_id() {
 }
 
 namespace {
-memory_impl::ptr attach_or_copy_data(network_impl& network, memory_impl& mem) {
+memory::ptr attach_or_copy_data(network_impl& network, memory::ptr mem) {
     auto& engine = network.get_engine();
-    if (mem.is_allocated_by(engine))
-        return (memory_impl::ptr) &mem;
+    if (mem->is_allocated_by(engine))
+        return mem;
 
-    memory_impl::ptr result = engine.allocate_memory(mem.get_layout(), network.get_id(), false);
-    mem_lock<char> src(mem);
-    mem_lock<char> dst(result);
+    memory::ptr result = engine.allocate_memory(mem->get_layout(), false);
+    mem_lock<char> src(mem, network.get_stream());
+    mem_lock<char> dst(result, network.get_stream());
     std::copy(src.begin(), src.end(), dst.begin());
     return result;
 }
 }  // namespace
 
 data_node::typed_program_node(const std::shared_ptr<data> dprim, program_impl& prog)
-    : parent(dprim, prog), mem(dprim->mem.get()) {
+    : parent(dprim, prog), mem(dprim->mem) {
     constant = true;
     can_share_buffer(false);
     recalc_output_layout(false);
 }
 
-void data_node::attach_memory(memory_impl& new_mem, bool invalidate_users_if_changed) {
-    mem = (memory_impl::ptr) &new_mem;
+void data_node::attach_memory(memory::ptr new_mem, bool invalidate_users_if_changed) {
+    mem = new_mem;
     recalc_output_layout(invalidate_users_if_changed);
 }
 
@@ -54,6 +54,6 @@ std::string data_inst::to_string(data_node const& node) {
 }
 
 data_inst::typed_primitive_inst(network_impl& network, data_node const& node)
-    : parent(network, node, *attach_or_copy_data(network, node.get_attached_memory())) {}
+    : parent(network, node, attach_or_copy_data(network, node.get_attached_memory_ptr())) {}
 
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/deconvolution.cpp b/inference-engine/thirdparty/clDNN/src/deconvolution.cpp
index 9642e3e3289939..6ddc0373f5f376 100644
--- a/inference-engine/thirdparty/clDNN/src/deconvolution.cpp
+++ b/inference-engine/thirdparty/clDNN/src/deconvolution.cpp
@@ -6,7 +6,7 @@
 #include "deconvolution_inst.h"
 #include "primitive_type_base.h"
 #include "sliding_window_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/deformable_convolution.cpp b/inference-engine/thirdparty/clDNN/src/deformable_convolution.cpp
index d3e04b2e1c8833..237ed6929f67da 100644
--- a/inference-engine/thirdparty/clDNN/src/deformable_convolution.cpp
+++ b/inference-engine/thirdparty/clDNN/src/deformable_convolution.cpp
@@ -6,7 +6,7 @@
 #include "deformable_convolution_inst.h"
 #include "primitive_type_base.h"
 #include "sliding_window_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/depth_to_space.cpp b/inference-engine/thirdparty/clDNN/src/depth_to_space.cpp
index 3c8622f6b8ae42..09f16b6f7c6877 100644
--- a/inference-engine/thirdparty/clDNN/src/depth_to_space.cpp
+++ b/inference-engine/thirdparty/clDNN/src/depth_to_space.cpp
@@ -5,7 +5,7 @@
 #include "depth_to_space_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/detection_output.cpp b/inference-engine/thirdparty/clDNN/src/detection_output.cpp
index 78a770caa54b51..c4e0fa86e2b810 100644
--- a/inference-engine/thirdparty/clDNN/src/detection_output.cpp
+++ b/inference-engine/thirdparty/clDNN/src/detection_output.cpp
@@ -5,7 +5,7 @@
 #include "detection_output_inst.h"
 #include "primitive_type_base.h"
 #include "network_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/device.cpp b/inference-engine/thirdparty/clDNN/src/device.cpp
deleted file mode 100644
index b3470db4ff93bd..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/device.cpp
+++ /dev/null
@@ -1,58 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "device_impl.h"
-#include "gpu/ocl_builder.h"
-
-#include <map>
-#include <string>
-
-namespace cldnn {
-
-device device::create_default() {
-    device_query query;
-    auto devices = query.get_available_devices();
-    // ToDo Maybe some heuristic should be added to decide what device is the default? (i.e number of EUs)
-    return devices.begin()->second;
-}
-
-device_info device::get_info() const {
-    return _impl->get_info().convert_to_api();
-}
-
-void device::retain() {
-    _impl->add_ref();
-}
-void device::release() {
-    _impl->release();
-}
-
-// --- device query ---
-device_query::device_query(void* clcontext, void* user_device)
-    : _impl(new device_query_impl(clcontext, user_device)) {
-}
-
-std::map<std::string, device> device_query::get_available_devices() const {
-    std::map<std::string, device> ret;
-    auto device_list = _impl->get_available_devices();
-    for (auto dev : device_list) {
-        ret.insert({ dev.first, device(dev.second.detach())});
-    }
-    return ret;
-}
-
-void device_query::retain() {
-    _impl->add_ref();
-}
-void device_query::release() {
-    _impl->release();
-}
-
-// --- device query impl ---
-device_query_impl::device_query_impl(void* user_context, void* user_device) {
-    gpu::ocl_builder builder;
-    _available_devices = builder.get_available_devices(user_context, user_device);
-}
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/eltwise.cpp b/inference-engine/thirdparty/clDNN/src/eltwise.cpp
index 760b504a013e89..a7f544a3a75af8 100644
--- a/inference-engine/thirdparty/clDNN/src/eltwise.cpp
+++ b/inference-engine/thirdparty/clDNN/src/eltwise.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "eltwise_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 #include <vector>
diff --git a/inference-engine/thirdparty/clDNN/src/embedding_bag.cpp b/inference-engine/thirdparty/clDNN/src/embedding_bag.cpp
index 951ff190049b68..84e5483dff0d6c 100644
--- a/inference-engine/thirdparty/clDNN/src/embedding_bag.cpp
+++ b/inference-engine/thirdparty/clDNN/src/embedding_bag.cpp
@@ -5,7 +5,7 @@
 #include "embedding_bag_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/engine.cpp b/inference-engine/thirdparty/clDNN/src/engine.cpp
deleted file mode 100644
index e64c0051827714..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/engine.cpp
+++ /dev/null
@@ -1,298 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "engine_impl.h"
-#include "event_impl.h"
-#include "program_impl.h"
-#include "network_impl.h"
-#include "gpu/ocl_toolkit.h"
-#include "gpu/memory_gpu.h"
-#include "gpu/ocl_user_event.h"
-#include "gpu/register_gpu.hpp"
-#include <string>
-#include <vector>
-#include <memory>
-#include <set>
-#include <stdexcept>
-
-namespace cldnn {
-
-engine::engine(engine_types type, const device& dev, const engine_configuration& configuration)
-    : _impl(new engine_impl(*dev.get(), configuration)) {
-    if (type != engine_types::ocl)
-        throw std::invalid_argument("Invalid engine type, should be ocl.");
-}
-
-uint32_t engine::engine_count(engine_types type) {
-    if (type == engine_types::ocl) {
-        return 1;
-    } else {
-        return 0;
-    }
-}
-
-void engine::release_pending_memory(uint32_t net_id) const {
-    _impl->release_pending_memory(net_id);
-}
-
-device_info engine::get_info() const {
-    auto info = _impl->get_device_info();
-    return info.convert_to_api();
-}
-
-void* engine::get_context() const {
-    return _impl->get_user_context();
-}
-
-uint64_t engine::get_max_used_device_memory_size() const {
-    return _impl->get_max_used_device_memory();
-}
-
-uint64_t engine::get_temp_used_device_memory_size() const {
-    return _impl->get_used_device_memory();
-}
-
-engine_types engine::get_type() const {
-    return _impl->type();
-}
-
-void engine::retain() {
-    _impl->add_ref();
-}
-void engine::release() {
-    _impl->release();
-}
-
-using gpu_toolkit_config = gpu::configuration;
-
-gpu_toolkit_config convert_configuration(const engine_configuration conf) {
-    gpu_toolkit_config result;
-    result.compiler_options = conf.compiler_options;
-    result.enable_profiling = conf.enable_profiling != 0;
-    result.meaningful_kernels_names = conf.meaningful_kernels_names != 0;
-    result.dump_custom_program = conf.dump_custom_program != 0;
-    result.single_kernel_name = conf.single_kernel_name;
-    result.host_out_of_order = true;
-    result.use_unifed_shared_memory = true;  // Switch on/off USM.
-    result.log = conf.engine_log;
-    result.ocl_sources_dumps_dir = conf.sources_dumps_dir;
-    result.priority_mode = conf.priority_mode;
-    result.throttle_mode = conf.throttle_mode;
-    result.queues_num = conf.n_streams;
-    result.kernels_cache_path = conf.kernels_cache_path;
-    result.tuning_cache_path = conf.tuning_cache_path;
-    result.n_threads = conf.n_threads;
-    return result;
-}
-
-engine_impl::engine_impl(const device_impl& dev, const engine_configuration& conf)
-    : _configuration(conf), _context(gpu_toolkit::create(dev, convert_configuration(conf))), _memory_pool(*this) {
-    gpu::register_implementations_gpu();
-}
-
-engine_impl::~engine_impl() {
-    /*
-        Engine, which is main owner of context deallocate events pool manually, because
-        of the event_impl <-> gpu_toolkit dependencies.
-    */
-    _context->release_all_events_pools();
-}
-
-memory_impl::ptr engine_impl::allocate_memory(const layout& layout, uint32_t net_id, bool reset) {
-    allocation_type type = get_lockable_preffered_memory_allocation_type(layout.format.is_image_2d());
-    return _memory_pool.get_memory(layout, type, net_id, reset);
-}
-
-memory_impl::ptr engine_impl::allocate_memory(const layout& layout, allocation_type type, uint32_t net_id, bool reset) {
-    return _memory_pool.get_memory(layout, type, net_id, reset);
-}
-
-memory_impl::ptr engine_impl::allocate_memory(const layout& layout,
-                                              primitive_id id,
-                                              uint32_t network_id,
-                                              std::set<primitive_id> dependencies,
-                                              allocation_type type,
-                                              bool reusable) {
-    if (use_memory_pool())
-        return _memory_pool.get_memory(layout, id, network_id, dependencies, type, reusable);
-    return _memory_pool.get_memory(layout, type, network_id);
-}
-
-memory_impl::ptr engine_impl::reinterpret_buffer(const memory_impl& memory, const layout& new_layout) {
-    if (memory.get_engine() != (const refcounted_obj_ptr<engine_impl>) this)
-        throw std::runtime_error("trying to reinterpret buffer allocated by a different engine");
-
-    if (new_layout.format.is_image() && !memory.get_layout().format.is_image())
-        throw std::runtime_error("trying to reinterpret non-image buffer as image");
-
-    if (!new_layout.format.is_image() && memory.get_layout().format.is_image())
-        throw std::runtime_error("trying to reinterpret image buffer as non-image buffer");
-
-    try {
-        if (new_layout.format.is_image_2d()) {
-           memory_impl::ptr mem_impl {
-                new gpu::gpu_image2d((refcounted_obj_ptr<engine_impl>) this,
-                                     new_layout,
-                                     reinterpret_cast<const gpu::gpu_image2d&>(memory).get_buffer(),
-                                     memory.get_net_id()),
-                                     false };
-            return mem_impl;
-        } else if (memory_capabilities::is_usm_type(memory.get_allocation_type())) {
-            memory_impl::ptr mem_impl{
-                    new gpu::gpu_usm((refcounted_obj_ptr<engine_impl>) this,
-                                        new_layout,
-                                        reinterpret_cast<const gpu::gpu_usm&>(memory).get_buffer(),
-                                        memory.get_allocation_type(),
-                                        memory.get_net_id()),
-                                        false };
-            return mem_impl;
-        } else {
-           memory_impl::ptr mem_impl {
-                new gpu::gpu_buffer((refcounted_obj_ptr<engine_impl>) this,
-                                    new_layout,
-                                    reinterpret_cast<const gpu::gpu_buffer&>(memory).get_buffer(),
-                                    memory.get_net_id()),
-                                    false};
-            return mem_impl;
-        }
-    } catch (cl::Error const& err) {
-        throw gpu::ocl_error(err);
-    }
-}
-
-memory_impl::ptr engine_impl::reinterpret_handle(const layout& new_layout,
-                                    const shared_mem_params* params,
-                                    uint32_t net_id) {
-    return _memory_pool.get_memory(new_layout, params, net_id);
-}
-
-
-bool engine_impl::is_the_same_buffer(const memory_impl& mem1, const memory_impl& mem2) {
-    if (mem1.get_engine() != (refcounted_obj_ptr<engine_impl>)this || mem2.get_engine() != (refcounted_obj_ptr<engine_impl>) this)
-        return false;
-    if (mem1.get_net_id() != mem2.get_net_id())
-        return false;
-    if (mem1.get_allocation_type() != mem2.get_allocation_type())
-        return false;
-    if (&mem1 == &mem2)
-        return true;
-
-    if (!memory_capabilities::is_usm_type(mem1.get_allocation_type()))
-        return (reinterpret_cast<const gpu::gpu_buffer&>(mem1).get_buffer() ==
-            reinterpret_cast<const gpu::gpu_buffer&>(mem2).get_buffer());
-    else
-        return (reinterpret_cast<const gpu::gpu_usm&>(mem1).get_buffer() ==
-            reinterpret_cast<const gpu::gpu_usm&>(mem2).get_buffer());
-}
-
-event_impl::ptr engine_impl::create_user_event(uint32_t net_id, bool set) {
-    try {
-        return _context->create_user_event(net_id, set);
-    } catch (cl::Error const& err) {
-        throw gpu::ocl_error(err);
-    }
-}
-
-void engine_impl::flush_network(uint32_t net_id) { get_context()->flush(net_id); }
-
-void engine_impl::release_pending_memory(uint32_t net_id) { get_context()->release_pending_memory(net_id); }
-
-program_impl::ptr engine_impl::build_program(const topology_impl& topology,
-                                            const build_options& options,
-                                            bool is_internal,
-                                            bool no_optimizations) {
-    program_impl::ptr progr_impl{ new program_impl(*this, topology, options, is_internal, no_optimizations), false };
-    return progr_impl;
-}
-
-program_impl::ptr engine_impl::build_program(const std::set<std::shared_ptr<program_node>>& nodes,
-                                            const build_options& options,
-                                            bool is_internal) {
-    program_impl::ptr progr_impl{ new program_impl(*this, nodes, options, is_internal), false };
-    return progr_impl;
-}
-
-network_impl::ptr engine_impl::build_network(const topology_impl& topology,
-                                            const build_options& options,
-                                            uint16_t stream_id,
-                                            bool is_internal) {
-    network_impl::ptr netw_impl{ new network_impl(*this, topology, options, stream_id, is_internal), false };
-    return netw_impl;
-}
-
-network_impl::ptr engine_impl::build_network(const std::set<std::shared_ptr<program_node>>& nodes,
-                                            const build_options& options,
-                                            bool is_internal) {
-    network_impl::ptr netw_impl{ new network_impl(*this, nodes, options, is_internal), false };
-    return netw_impl;
-}
-
-network_impl::ptr engine_impl::allocate_network(const program_impl& program, uint16_t stream_id, bool is_internal) {
-    if (stream_id >= _configuration.n_streams)
-        throw std::invalid_argument("Unable to create network with stream_id=" + std::to_string(stream_id));
-    network_impl::ptr netw_impl{ new network_impl(program, stream_id, is_internal), false };
-    return netw_impl;
-}
-
-void engine_impl::wait_for_events(std::vector<event_impl::ptr> const& events) {
-    if (!events.empty())
-        _context->wait_for_events(events);
-}
-
-gpu::device_info_internal engine_impl::get_device_info() const { return _context->get_device_info(); }
-
-void* engine_impl::get_user_context() const { return static_cast<void*>(_context->context().get()); }
-
-void engine_impl::compile_program(program_impl& program) {
-    auto& cache = _context->get_kernels_cache(program.get_id());
-    if (!program.get_options().get<build_option_type::serialize_network>()->serialization_network_name.empty())
-        cache.get_context().set_serialization_flag(true);
-    // TODO: better compilation logic instead of a simple 'compile all'?
-    cache.build_all();
-}
-
-bool engine_impl::use_memory_pool() const {
-    if (configuration().enable_memory_pool && get_context()->is_neo_driver()) {
-        return true;
-    }
-    return false;
-}
-
-bool engine_impl::use_unified_shared_memory() const {
-    if (get_context()->memory_caps().supports_usm() && get_context()->get_configuration().use_unifed_shared_memory) {
-        return true;
-    }
-    return false;
-}
-
-bool engine_impl::supports_allocation(allocation_type type) const {
-    if (memory_capabilities::is_usm_type(type) && !use_unified_shared_memory())
-        return false;
-    if (allocation_type::usm_shared == type)
-        return false;
-    return get_context()->memory_caps().support_allocation_type(type);
-}
-
-allocation_type engine_impl::get_lockable_preffered_memory_allocation_type(bool is_image_layout) const {
-    if (!use_unified_shared_memory() || is_image_layout)
-        return allocation_type::cl_mem;
-
-    /*
-        We do not check device allocation here.
-        Device allocation is reserved for buffers of hidden layers.
-        Const buffers are propagated to device if possible.
-    */
-
-    bool support_usm_host = supports_allocation(allocation_type::usm_host);
-    bool support_usm_shared = supports_allocation(allocation_type::usm_shared);
-
-    if (support_usm_shared)
-        return allocation_type::usm_shared;
-    if (support_usm_host)
-        return allocation_type::usm_host;
-
-    throw std::runtime_error("[clDNN internal error] Could not find proper allocation type!");
-}
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/error_handler.cpp b/inference-engine/thirdparty/clDNN/src/error_handler.cpp
index 1e41407b27d82b..bb7ce5224128f3 100644
--- a/inference-engine/thirdparty/clDNN/src/error_handler.cpp
+++ b/inference-engine/thirdparty/clDNN/src/error_handler.cpp
@@ -3,7 +3,7 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include <string>
 #include <vector>
 
diff --git a/inference-engine/thirdparty/clDNN/src/event.cpp b/inference-engine/thirdparty/clDNN/src/event.cpp
deleted file mode 100644
index 48b6dfaff31175..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/event.cpp
+++ /dev/null
@@ -1,102 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "api/event.hpp"
-#include "event_impl.h"
-#include "engine_impl.h"
-#include <list>
-#include <vector>
-#include <algorithm>
-
-namespace cldnn {
-
-event event::create_user_event(const engine& engine, uint32_t net_id) {
-    return event(engine.get()->create_user_event(net_id).detach());
-}
-
-void event::wait() const {
-    _impl->wait();
-}
-
-void event::set() const {
-    if (auto user_ev = dynamic_cast<user_event*>(_impl))
-        user_ev->set();
-    else
-        throw std::invalid_argument("Event passed to cldnn_set_event should be an user event");
-}
-
-void event::set_event_handler(event_handler handler, void* param) const {
-    _impl->add_event_handler(handler, param);
-}
-
-std::vector<instrumentation::profiling_interval> event::get_profiling_info() const {
-    auto interval_list = _impl->get_profiling_info();
-    std::vector<instrumentation::profiling_interval> result(interval_list.size());
-    std::copy(interval_list.begin(), interval_list.end(), result.begin());
-    return result;
-}
-
-void event::retain() {
-    _impl->add_ref();
-}
-
-void event::release() {
-    _impl->release();
-}
-
-void event_impl::wait() {
-    if (_set)
-        return;
-
-    // TODO: refactor in context of multiple simultaneous calls (for generic engine)
-    wait_impl();
-    _set = true;
-    return;
-}
-
-bool event_impl::is_set() {
-    if (_set)
-        return true;
-
-    // TODO: refactor in context of multiple simultaneous calls (for generic engine)
-    _set = is_set_impl();
-    return _set;
-}
-
-bool event_impl::add_event_handler(event_handler handler, void* data) {
-    if (is_set()) {
-        handler(data);
-        return true;
-    }
-
-    std::lock_guard<std::mutex> lock(_handlers_mutex);
-    auto itr = _handlers.insert(_handlers.end(), {handler, data});
-    auto ret = add_event_handler_impl(handler, data);
-    if (!ret)
-        _handlers.erase(itr);
-
-    return ret;
-}
-
-const std::list<instrumentation::profiling_interval>& event_impl::get_profiling_info() {
-    if (_profiling_captured)
-        return _profiling_info;
-
-    _profiling_captured = get_profiling_info_impl(_profiling_info);
-    return _profiling_info;
-}
-
-void event_impl::call_handlers() {
-    std::lock_guard<std::mutex> lock(_handlers_mutex);
-    for (auto& pair : _handlers) {
-        try {
-            pair.first(pair.second);
-        } catch (...) {
-        }
-    }
-    _handlers.clear();
-}
-
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/extract_image_patches.cpp b/inference-engine/thirdparty/clDNN/src/extract_image_patches.cpp
index a4cc4ff1fe65a0..fffa064d13b236 100644
--- a/inference-engine/thirdparty/clDNN/src/extract_image_patches.cpp
+++ b/inference-engine/thirdparty/clDNN/src/extract_image_patches.cpp
@@ -5,7 +5,7 @@
 #include "extract_image_patches_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/fully_connected.cpp b/inference-engine/thirdparty/clDNN/src/fully_connected.cpp
index 4c6aeabe80eaae..71130b19a188cd 100644
--- a/inference-engine/thirdparty/clDNN/src/fully_connected.cpp
+++ b/inference-engine/thirdparty/clDNN/src/fully_connected.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "fully_connected_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/fused_conv_eltwise.cpp b/inference-engine/thirdparty/clDNN/src/fused_conv_eltwise.cpp
index dc4db14393b499..9d11b1ad4c5056 100644
--- a/inference-engine/thirdparty/clDNN/src/fused_conv_eltwise.cpp
+++ b/inference-engine/thirdparty/clDNN/src/fused_conv_eltwise.cpp
@@ -6,7 +6,7 @@
 #include "fused_conv_eltwise_inst.h"
 #include "primitive_type_base.h"
 #include "sliding_window_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/gather.cpp b/inference-engine/thirdparty/clDNN/src/gather.cpp
index c688ea19823615..5a264d2505b448 100644
--- a/inference-engine/thirdparty/clDNN/src/gather.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gather.cpp
@@ -5,7 +5,7 @@
 #include "gather_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/gather_nd.cpp b/inference-engine/thirdparty/clDNN/src/gather_nd.cpp
index f01d82dffa3ea8..998df5a76b2198 100644
--- a/inference-engine/thirdparty/clDNN/src/gather_nd.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gather_nd.cpp
@@ -1,23 +1,11 @@
-/*
-// Copyright (c) 2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
 //
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-*/
 
 #include "gather_nd_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/gather_tree.cpp b/inference-engine/thirdparty/clDNN/src/gather_tree.cpp
index 98b687be5b7013..6610df3523c777 100644
--- a/inference-engine/thirdparty/clDNN/src/gather_tree.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gather_tree.cpp
@@ -4,7 +4,7 @@
 
 #include "gather_tree_inst.h"
 
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include "primitive_type_base.h"
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/src/gemm.cpp b/inference-engine/thirdparty/clDNN/src/gemm.cpp
index 1c3191b87d9184..2652116b90a7c3 100644
--- a/inference-engine/thirdparty/clDNN/src/gemm.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gemm.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "gemm_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 #include <utility>
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/activation_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/activation_gpu.cpp
index 7ea703cab3bc87..9a90f972752dec 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/activation_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/activation_gpu.cpp
@@ -5,11 +5,10 @@
 #include "activation_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "activation/activation_kernel_selector.h"
 #include "activation/activation_kernel_base.h"
-#include "api/activation.hpp"
 #include "register_gpu.hpp"
 
 namespace cldnn {
@@ -19,12 +18,15 @@ struct activation_gpu : typed_primitive_gpu_impl<activation> {
     using parent = typed_primitive_gpu_impl<activation>;
     using parent::parent;
 
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<activation>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<activation_gpu>(*this);
+    }
+
+    kernel_arguments_data get_arguments(typed_primitive_inst<activation>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
 
         if (_outer.is_parameterized()) {
-            args.slope = (memory_impl::cptr) &instance.slope_memory();
+            args.slope = instance.slope_memory();
         }
 
         return args;
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/arg_max_min_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/arg_max_min_gpu.cpp
index 3b898761653687..7717ea901ded89 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/arg_max_min_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/arg_max_min_gpu.cpp
@@ -5,7 +5,7 @@
 #include "arg_max_min_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "arg_max_min/arg_max_min_kernel_selector.h"
 #include "arg_max_min/arg_max_min_kernel_base.h"
@@ -18,10 +18,13 @@ struct arg_max_min_gpu : typed_primitive_gpu_impl<arg_max_min> {
     using parent = typed_primitive_gpu_impl<arg_max_min>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<arg_max_min_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<arg_max_min>& instance,
-                                                        int32_t) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, 0);
+    kernel_arguments_data get_arguments(typed_primitive_inst<arg_max_min>& instance, int32_t) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, 0);
 
         if (args.inputs.size() == 3) {
             args.inputs.erase(args.inputs.begin() + 1);  // erase constant input in case of TOP_K
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/average_unpooling_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/average_unpooling_gpu.cpp
index 4ed6597d63ab30..55ad712d4d8a7e 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/average_unpooling_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/average_unpooling_gpu.cpp
@@ -5,7 +5,7 @@
 #include "average_unpooling_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "average_unpooling/average_unpooling_kernel_selector.h"
 #include "average_unpooling/average_unpooling_kernel_base.h"
@@ -17,10 +17,13 @@ struct average_unpooling_gpu : typed_primitive_gpu_impl<average_unpooling> {
     using parent = typed_primitive_gpu_impl<average_unpooling>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<average_unpooling_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<average_unpooling>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
+    kernel_arguments_data get_arguments(typed_primitive_inst<average_unpooling>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
         return args;
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/batch_to_space_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/batch_to_space_gpu.cpp
index 820e62561670ea..d2ef8e70df9f5c 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/batch_to_space_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/batch_to_space_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "batch_to_space/batch_to_space_kernel_selector.h"
 #include "batch_to_space/batch_to_space_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "data_inst.h"
 #include <vector>
 
@@ -20,6 +20,10 @@ struct batch_to_space_gpu : typed_primitive_gpu_impl<batch_to_space> {
     using parent = typed_primitive_gpu_impl<batch_to_space>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<batch_to_space_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const batch_to_space_node& arg) {
         auto batch_to_space_params = get_default_params<kernel_selector::batch_to_space_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/binary_convolution_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/binary_convolution_gpu.cpp
index 8800cc2c5ad111..c9b4c78eee552f 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/binary_convolution_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/binary_convolution_gpu.cpp
@@ -2,12 +2,12 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <api/scale.hpp>
-#include <api/quantize.hpp>
+#include "cldnn/primitives/scale.hpp"
+#include "cldnn/primitives/quantize.hpp"
 #include "binary_convolution_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "kernel_runner.h"
 #include "kernel_selector/core/actual_kernels/binary_convolution/binary_convolution_kernel_selector.h"
@@ -22,6 +22,10 @@ struct binary_convolution_gpu : typed_primitive_gpu_impl<binary_convolution> {
     using parent = typed_primitive_gpu_impl<binary_convolution>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<binary_convolution_gpu>(*this);
+    }
+
 protected:
     bool validate_impl(const typed_primitive_inst<binary_convolution>& instance) const override {
         bool res = true;
@@ -40,17 +44,16 @@ struct binary_convolution_gpu : typed_primitive_gpu_impl<binary_convolution> {
                                                     "Input memory",
                                                     data_type,
                                                     "filter memory",
-                                                    instance.weights_memory(0).get_layout().data_type,
+                                                    instance.weights_memory(0)->get_layout().data_type,
                                                     "");
 
         return res;
     }
 
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<binary_convolution>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
+    kernel_arguments_data get_arguments(typed_primitive_inst<binary_convolution>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
 
-        args.weights = (memory_impl::cptr) &instance.weights_memory(split);
+        args.weights = instance.weights_memory(split);
         return args;
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/border_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/border_gpu.cpp
index 33985b48c9736b..01c1d54ad1099a 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/border_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/border_gpu.cpp
@@ -9,7 +9,7 @@
 #include "kernel_selector_helper.h"
 #include "border/border_kernel_selector.h"
 #include "border/border_kernel_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -18,6 +18,10 @@ struct border_gpu : typed_primitive_gpu_impl<border> {
     using parent = typed_primitive_gpu_impl<border>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<border_gpu>(*this);
+    }
+
     static primitive_impl* create(const border_node& arg) {
         auto b_params = get_default_params<kernel_selector::border_params>(arg, 1);
         auto b_optional_params =
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/broadcast_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/broadcast_gpu.cpp
index a3fa9715eb90e8..1c07420d7a9365 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/broadcast_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/broadcast_gpu.cpp
@@ -9,7 +9,7 @@
 #include "kernel_selector_helper.h"
 #include "broadcast/broadcast_kernel_selector.h"
 #include "broadcast/broadcast_kernel_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -18,6 +18,10 @@ struct broadcast_gpu : typed_primitive_gpu_impl<broadcast> {
     using parent = typed_primitive_gpu_impl<broadcast>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<broadcast_gpu>(*this);
+    }
+
     static primitive_impl* create(const broadcast_node& arg) {
         auto bc_params = get_default_params<kernel_selector::broadcast_params>(arg, 1);
         auto bc_optional_params =
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/concatenation_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/concatenation_gpu.cpp
index b2abee8ab58ccf..590e97a2a8d274 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/concatenation_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/concatenation_gpu.cpp
@@ -5,8 +5,7 @@
 #include "concatenation_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "events_waiter.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "concatenation/concatenation_kernel_selector.h"
 #include "concatenation/concatenation_kernel_base.h"
@@ -40,6 +39,10 @@ kernel_selector::concat_axis convert_axis(concatenation::concatenation_axis axis
 struct concatenation_gpu : typed_primitive_gpu_impl<concatenation> {
     using parent = typed_primitive_gpu_impl<concatenation>;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<concatenation_gpu>(*this);
+    }
+
     concatenation_gpu(const concatenation_node& arg, const kernel_selector::kernel_data& kd) : parent(arg, kd) {
         if (!_outer.can_be_optimized()) {
             CLDNN_ERROR_NOT_EQUAL(_outer.id(),
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/condition_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/condition_gpu.cpp
index e3df38f722980f..2f84605cc19c24 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/condition_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/condition_gpu.cpp
@@ -17,23 +17,27 @@ namespace gpu {
 struct condition_gpu : typed_primitive_impl<condition> {
     const condition_node& outer;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<condition_gpu>(*this);
+    }
+
     explicit condition_gpu(const condition_node& outer) : outer(outer) {}
 
-    event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& events, condition_inst& instance) override {
+    event::ptr execute_impl(const std::vector<event::ptr>& events, condition_inst& instance) override {
         for (auto& a : events) {
             a->wait();
         }
-        auto ev = instance.get_network().get_engine().create_user_event(instance.get_network().get_id(), false);
+        auto ev = instance.get_network().get_stream().create_user_event(false);
 
         bool exec_branch = choose_branch_to_exec(instance);
-        memory_impl::ptr memory_to_copy;
+        memory::ptr memory_to_copy;
         if (exec_branch)
-            memory_to_copy = (memory_impl::ptr) &execute_branch(instance.get_net_true(), instance.result_id(), instance.input_memory());
+            memory_to_copy = execute_branch(instance.get_net_true(), instance.result_id(), instance.input_memory_ptr());
         else
-            memory_to_copy = (memory_impl::ptr) &execute_branch(instance.get_net_false(), instance.result_id(), instance.input_memory());
+            memory_to_copy = execute_branch(instance.get_net_false(), instance.result_id(), instance.input_memory_ptr());
         // just copy memory
-        mem_lock<float> inp_ptr{memory_to_copy};
-        mem_lock<float> out_ptr{instance.output_memory()};
+        mem_lock<float> inp_ptr{memory_to_copy, instance.get_network().get_stream()};
+        mem_lock<float> out_ptr{instance.output_memory_ptr(), instance.get_network().get_stream()};
         std::copy(inp_ptr.begin(), inp_ptr.end(), out_ptr.begin());
         dynamic_cast<cldnn::user_event*>(ev.get())->set();  // set as complete
         return ev;
@@ -41,6 +45,8 @@ struct condition_gpu : typed_primitive_impl<condition> {
 
     static primitive_impl* create(const condition_node& arg) { return new condition_gpu(arg); }
 
+    void init_kernels() override {}
+
 private:
     /*
     Add functions here.
@@ -67,11 +73,11 @@ struct condition_gpu : typed_primitive_impl<condition> {
     Returns boolean flag, which says what branch should be executed.
     */
     bool choose_branch_to_exec(condition_inst& instance) const {
-        mem_lock<float> lock_compare_data{instance.compare_memory()};
+        mem_lock<float> lock_compare_data{instance.compare_memory_ptr(), instance.get_network().get_stream()};
         auto compare_layout = instance.compare_memory().get_layout();
         auto compare_ptr = lock_compare_data.begin();
 
-        mem_lock<float> lock_input{instance.input_memory()};
+        mem_lock<float> lock_input{instance.input_memory_ptr(), instance.get_network().get_stream()};
         auto input_layout = instance.input_memory().get_layout();
         auto input_ptr = lock_input.begin();
 
@@ -101,12 +107,12 @@ struct condition_gpu : typed_primitive_impl<condition> {
         return true;
     }
 
-    memory_impl& execute_branch(network_impl::ptr branch,
-                                const primitive_id& input_id,
-                                memory_impl& input_memory) const {
+    memory::ptr execute_branch(network_impl::ptr branch,
+                           const primitive_id& input_id,
+                           memory::ptr input_memory) const {
         branch->set_input_data(input_id, input_memory);
         branch->execute({});
-        return branch->get_outputs().at(0)->output_memory();
+        return branch->get_outputs().at(0)->output_memory_ptr();
     }
 };
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/configuration.cpp b/inference-engine/thirdparty/clDNN/src/gpu/configuration.cpp
deleted file mode 100644
index 2536bf0b0c1d10..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/configuration.cpp
+++ /dev/null
@@ -1,29 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "configuration.h"
-#include <algorithm>
-
-namespace cldnn {
-namespace gpu {
-
-configuration::configuration()
-    : enable_profiling(false),
-      meaningful_kernels_names(false),
-      dump_custom_program(false),
-      host_out_of_order(true),
-      use_unifed_shared_memory(false),
-      compiler_options(""),
-      single_kernel_name(""),
-      log(""),
-      ocl_sources_dumps_dir(""),
-      priority_mode(priority_mode_types::disabled),
-      throttle_mode(throttle_mode_types::disabled),
-      queues_num(0),
-      tuning_cache_path("cache.json"),
-      kernels_cache_path(""),
-      n_threads(std::max(static_cast<uint16_t>(std::thread::hardware_concurrency()), static_cast<uint16_t>(1))) {}
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/configuration.h b/inference-engine/thirdparty/clDNN/src/gpu/configuration.h
deleted file mode 100644
index f0792ad8dac0d2..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/configuration.h
+++ /dev/null
@@ -1,37 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include <string>
-#include "api/cldnn.hpp"
-#include "api/engine.hpp"
-
-namespace cl {
-class Context;
-}
-namespace cldnn {
-namespace gpu {
-
-struct configuration {
-    configuration();
-
-    bool enable_profiling;
-    bool meaningful_kernels_names;
-    bool dump_custom_program;
-    bool host_out_of_order;
-    bool use_unifed_shared_memory;
-    std::string compiler_options;
-    std::string single_kernel_name;
-    std::string log;
-    std::string ocl_sources_dumps_dir;
-    priority_mode_types priority_mode;
-    throttle_mode_types throttle_mode;
-    uint16_t queues_num;
-    std::string tuning_cache_path;
-    std::string kernels_cache_path;
-    uint16_t n_threads;
-};
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/convolution_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/convolution_gpu.cpp
index 1c37c2073032c5..63065d8be870a6 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/convolution_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/convolution_gpu.cpp
@@ -6,7 +6,7 @@
 #include "eltwise_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "kernel_runner.h"
 #include "convolution/convolution_kernel_selector.h"
@@ -21,6 +21,10 @@ struct convolution_gpu : typed_primitive_gpu_impl<convolution> {
     using parent = typed_primitive_gpu_impl<convolution>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<convolution_gpu>(*this);
+    }
+
 protected:
     bool validate_impl(const typed_primitive_inst<convolution>& instance) const override {
         bool res = true;
@@ -33,26 +37,20 @@ struct convolution_gpu : typed_primitive_gpu_impl<convolution> {
                                                     "Input memory",
                                                     data_type,
                                                     "filter memory",
-                                                    instance.weights_memory(0).get_layout().data_type,
+                                                    instance.weights_memory(0)->get_layout().data_type,
                                                     "");
 
         return res;
     }
 
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<convolution>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
-
-        args.weights = (memory_impl::cptr) &instance.weights_memory(split);
-        args.bias = (memory_impl::cptr) (instance.bias_term() ? &instance.bias_memory(split) : nullptr);
-        args.weights_zero_points = (memory_impl::cptr) (instance.weights_zero_points_term() ? &instance.weights_zero_points_memory(split)
-                                                                                            : nullptr);
-        args.activations_zero_points = (memory_impl::cptr) (instance.activations_zero_points_term()
-                                       ? &instance.activations_zero_points_memory(split)
-                                       : nullptr);
-        args.compensation = (memory_impl::cptr) (instance.compensation_term()
-                                       ? &instance.compensation_memory(split)
-                                       : nullptr);
+    kernel_arguments_data get_arguments(typed_primitive_inst<convolution>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
+
+        args.weights = instance.weights_memory(split);
+        args.bias = instance.bias_term() ? instance.bias_memory(split) : nullptr;
+        args.weights_zero_points = instance.weights_zero_points_term() ? instance.weights_zero_points_memory(split) : nullptr;
+        args.activations_zero_points = instance.activations_zero_points_term() ? instance.activations_zero_points_memory(split) : nullptr;
+        args.compensation = instance.compensation_term() ? instance.compensation_memory(split) : nullptr;
 
         return args;
     }
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/crop_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/crop_gpu.cpp
index 0a4498973d52a5..07244dde5aeb0c 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/crop_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/crop_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "eltwise/eltwise_kernel_selector.h"
 #include "eltwise/eltwise_kernel_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -17,6 +17,10 @@ struct crop_gpu : typed_primitive_gpu_impl<crop> {
     using parent = typed_primitive_gpu_impl<crop>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<crop_gpu>(*this);
+    }
+
 protected:
     bool optimized_out(crop_inst& instance) const override {
         return parent::optimized_out(instance) || _outer.can_be_optimized();
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ctc_greedy_decoder_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/ctc_greedy_decoder_gpu.cpp
index 1161e4aee482ae..4c6c7f39c3c0dc 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/ctc_greedy_decoder_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/ctc_greedy_decoder_gpu.cpp
@@ -5,7 +5,7 @@
 #include "ctc_greedy_decoder_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "ctc_greedy_decoder/ctc_greedy_decoder_kernel_selector.h"
 #include "ctc_greedy_decoder/ctc_greedy_decoder_kernel_base.h"
@@ -21,6 +21,10 @@ struct ctc_greedy_decoder_gpu : typed_primitive_gpu_impl<ctc_greedy_decoder> {
     using parent = typed_primitive_gpu_impl<ctc_greedy_decoder>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<ctc_greedy_decoder_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const ctc_greedy_decoder_node& arg) {
         auto ctc_gd_params = get_default_params<kernel_selector::ctc_greedy_decoder_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/cum_sum_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/cum_sum_gpu.cpp
index d6d82ae2055dec..82e44cf07a3395 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/cum_sum_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/cum_sum_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "cum_sum/cum_sum_kernel_selector.h"
 #include "cum_sum/cum_sum_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -40,6 +40,10 @@ struct cum_sum_gpu : typed_primitive_gpu_impl<cum_sum> {
     using parent = typed_primitive_gpu_impl<cum_sum>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<cum_sum_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const cum_sum_node& arg) {
         auto cum_sum_params = get_default_params<kernel_selector::cum_sum_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/custom_gpu_primitive_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/custom_gpu_primitive_gpu.cpp
index 493f40203cdc52..eaecee357b685e 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/custom_gpu_primitive_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/custom_gpu_primitive_gpu.cpp
@@ -3,13 +3,12 @@
 //
 
 #include "custom_gpu_primitive_inst.h"
-#include "kernel.h"
+#include "cldnn/runtime/engine.hpp"
 #include "implementation_map.h"
 #include "kernel_selector_helper.h"
 #include "network_impl.h"
-#include "engine_impl.h"
 #include "jitter.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "register_gpu.hpp"
 
 #include <map>
@@ -28,37 +27,52 @@ namespace neural {
 struct custom_gpu_primitive_gpu : typed_primitive_impl<custom_gpu_primitive> {
     const custom_gpu_primitive_node& outer;
     std::shared_ptr<kernel_selector::cl_kernel_data> cl_kernel;
-    gpu::kernel _kernel;
+    std::vector<kernel::ptr> _kernels;
+    kernel_id _kernel_id;
+
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<custom_gpu_primitive_gpu>(*this);
+    }
+
+    custom_gpu_primitive_gpu(const custom_gpu_primitive_gpu& other)
+    : outer(other.outer)
+    , cl_kernel(other.cl_kernel)
+    , _kernels({})
+    , _kernel_id(other._kernel_id) {
+        _kernels.emplace_back(std::move(outer.get_program().get_kernel(_kernel_id)->clone()));
+    }
 
     custom_gpu_primitive_gpu(const custom_gpu_primitive_node& arg,
                              std::shared_ptr<kernel_selector::cl_kernel_data>& cl_kernel)
-        : outer(arg),
-          cl_kernel(cl_kernel),
-          _kernel(arg.get_program().get_engine().get_context(),
-                  cl_kernel->kernelString,
-                  arg.get_program().get_id(),
-                  arg.get_program().get_engine().get_context()->get_configuration().dump_custom_program) {}
+        : outer(arg)
+        , cl_kernel(cl_kernel)
+        , _kernels() {
+        _kernel_id = outer.get_program().add_kernel(cl_kernel->code.kernelString);
+    }
+
+    void init_kernels() override {
+        _kernels.emplace_back(std::move(outer.get_program().get_kernel(_kernel_id)));
+    }
 
     void set_arguments_impl(custom_gpu_primitive_inst& instance) override {
-        auto net_id = instance.get_network().get_id();
-        gpu::kernel::kernel_arguments_data args;
+        auto& stream = instance.get_network().get_stream();
+        kernel_arguments_data args;
         for (auto& dep : instance.dependencies()) {
-            args.inputs.push_back((memory_impl::cptr) &(dep->output_memory()));
+            args.inputs.push_back(dep->output_memory_ptr());
         }
-        args.output = (memory_impl::cptr) &instance.output_memory();
-        _kernel.set_arguments(net_id, *cl_kernel.get(), args);
-    }
-
-    void cleanup_impl(custom_gpu_primitive_inst& instance) override {
-        auto net_id = instance.get_network().get_id();
-        _kernel.cleanup(net_id);
+        args.output = instance.output_memory_ptr();
+        stream.set_arguments(*_kernels.front(), cl_kernel.get()->params, args);
     }
 
-    event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& events,
+    event::ptr execute_impl(const std::vector<event::ptr>& events,
                                  custom_gpu_primitive_inst& instance) override {
-        auto net_id = instance.get_network().get_id();
-        _kernel.set_output_event(net_id, instance.node.is_output());
-        return _kernel.run(net_id, *cl_kernel.get(), events);
+        auto& stream = instance.get_network().get_stream();
+        kernel_arguments_data args;
+        for (auto& dep : instance.dependencies()) {
+            args.inputs.push_back(dep->output_memory_ptr());
+        }
+        args.output = instance.output_memory_ptr();
+        return stream.enqueue_kernel(*_kernels.front(), cl_kernel.get()->params, args, events, instance.node.is_output());
     }
 };
 
@@ -195,19 +209,19 @@ static primitive_impl* create(const custom_gpu_primitive_node& arg) {
     const auto primitive = arg.get_primitive().get();
 
     auto cl_kernel = std::make_shared<kernel_selector::cl_kernel_data>();
-    cl_kernel->kernelString = std::make_shared<kernel_selector::kernel_string>();
-    cl_kernel->kernelString->entry_point = primitive->kernel_entry_point;
-    cl_kernel->kernelString->options = primitive->build_options;
-    cl_kernel->kernelString->jit = get_jit_constant(arg);
+    cl_kernel->code.kernelString = std::make_shared<kernel_selector::kernel_string>();
+    cl_kernel->code.kernelString->entry_point = primitive->kernel_entry_point;
+    cl_kernel->code.kernelString->options = primitive->build_options;
+    cl_kernel->code.kernelString->jit = get_jit_constant(arg);
     for (const auto& s : primitive->kernels_code) {
-        cl_kernel->kernelString->str += s + "\n";
+        cl_kernel->code.kernelString->str += s + "\n";
     }
 
-    cl_kernel->workGroups.global = primitive->gws;
-    cl_kernel->workGroups.local = primitive->lws;
+    cl_kernel->params.workGroups.global = primitive->gws;
+    cl_kernel->params.workGroups.local = primitive->lws;
 
     for (const auto& p : primitive->kernel_arguments) {
-        cl_kernel->arguments.push_back(get_arg(p));
+        cl_kernel->params.arguments.push_back(get_arg(p));
     }
 
     return new custom_gpu_primitive_gpu(arg, cl_kernel);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/deconvolution_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/deconvolution_gpu.cpp
index 352bdf00d5d0de..9c1532ecfd45a7 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/deconvolution_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/deconvolution_gpu.cpp
@@ -5,7 +5,7 @@
 #include "deconvolution_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "deconvolution/deconvolution_kernel_selector.h"
 #include "deconvolution/deconvolution_kernel_base.h"
@@ -18,6 +18,10 @@ struct deconvolution_gpu : typed_primitive_gpu_impl<deconvolution> {
     using parent = typed_primitive_gpu_impl<deconvolution>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<deconvolution_gpu>(*this);
+    }
+
 protected:
     // TODO: share it with convolution and fully connected
     bool validate_impl(const typed_primitive_inst<deconvolution>&) const override {
@@ -33,12 +37,11 @@ struct deconvolution_gpu : typed_primitive_gpu_impl<deconvolution> {
         return res;
     }
 
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<deconvolution>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
+    kernel_arguments_data get_arguments(typed_primitive_inst<deconvolution>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
 
-        args.weights = (memory_impl::cptr) &instance.weights_memory(split);
-        args.bias = (memory_impl::cptr) (instance.bias_term() ? &instance.bias_memory(split) : nullptr);
+        args.weights = instance.weights_memory(split);
+        args.bias = instance.bias_term() ? instance.bias_memory(split) : nullptr;
 
         return args;
     }
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/deformable_convolution_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/deformable_convolution_gpu.cpp
index 897f1a07a763f7..f2b61a505e27b1 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/deformable_convolution_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/deformable_convolution_gpu.cpp
@@ -5,7 +5,7 @@
 #include "deformable_convolution_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "kernel_runner.h"
 #include "convolution/convolution_kernel_selector.h"
@@ -19,13 +19,16 @@ struct deformable_conv_gpu : typed_primitive_gpu_impl<deformable_conv> {
     using parent = typed_primitive_gpu_impl<deformable_conv>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<deformable_conv_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<deformable_conv>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
+    kernel_arguments_data get_arguments(typed_primitive_inst<deformable_conv>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
 
-        args.weights = (memory_impl::cptr) &instance.weights_memory(split);
-        args.bias = memory_impl::cptr(instance.bias_term() ? &instance.bias_memory(split) : nullptr);
+        args.weights = instance.weights_memory(split);
+        args.bias = instance.bias_term() ? instance.bias_memory(split) : nullptr;
         return args;
     }
 
@@ -78,6 +81,10 @@ struct deformable_interp_gpu : typed_primitive_gpu_impl<deformable_interp> {
     using parent = typed_primitive_gpu_impl<deformable_interp>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<deformable_interp_gpu>(*this);
+    }
+
 protected:
     int32_t get_split() const override { return 1; }
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/depth_to_space_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/depth_to_space_gpu.cpp
index fc3c5dd9f56279..3a62c9e0a9c810 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/depth_to_space_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/depth_to_space_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "depth_to_space/depth_to_space_kernel_selector.h"
 #include "depth_to_space/depth_to_space_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "common_types.h"
 
 using namespace cldnn;
@@ -19,6 +19,10 @@ struct depth_to_space_gpu : typed_primitive_gpu_impl<depth_to_space> {
     using parent = typed_primitive_gpu_impl<depth_to_space>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<depth_to_space_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const depth_to_space_node& arg) {
         auto depth_to_space_params = get_default_params<kernel_selector::depth_to_space_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/detection_output_cpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/detection_output_cpu.cpp
index dfe22f193d6a1b..9673270ab264dc 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/detection_output_cpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/detection_output_cpu.cpp
@@ -3,7 +3,6 @@
 //
 
 #include "detection_output_inst.h"
-#include "kernel.h"
 #include "network_impl.h"
 #include "implementation_map.h"
 #include "math_utils.h"
@@ -37,6 +36,9 @@ struct detection_output_cpu : typed_primitive_impl<detection_output> {
     const detection_output_node& outer;
     NMSType nms_type;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<detection_output_cpu>(*this);
+    }
     explicit detection_output_cpu(const detection_output_node& outer)
         : outer(outer)
         , nms_type(outer.get_primitive()->decrease_label_id ? MXNET : CAFFE) {}
@@ -249,12 +251,12 @@ struct detection_output_cpu : typed_primitive_impl<detection_output> {
     }
 
     template <typename dtype>
-    void generate_detections(const detection_output_inst& instance,
+    void generate_detections(stream& stream, const detection_output_inst& instance,
                              const int num_of_images,
                              const std::vector<std::vector<std::vector<bounding_box>>>& all_bboxes,
                              std::vector<std::vector<std::vector<std::pair<float, int>>>>& confidences,
                              std::vector<std::vector<std::pair<float, std::pair<int, int>>>>& scoreIndexPairs) {
-        mem_lock<dtype> lock{instance.output_memory()};
+        mem_lock<dtype> lock{instance.output_memory_ptr(), stream};
         auto out_ptr = lock.begin();
 
         const auto& args = instance.argument;
@@ -405,22 +407,23 @@ struct detection_output_cpu : typed_primitive_impl<detection_output> {
     }
 
     template <typename dtype>
-    void extract_locations_per_image(const detection_output_inst& instance,
+    void extract_locations_per_image(stream& stream, const detection_output_inst& instance,
                                      std::vector<std::vector<std::vector<bounding_box>>>& locations,
                                      const int num_of_priors,
                                      const int num_loc_classes) {
         const bool share_location = instance.argument.share_location;
-        auto& input_location = instance.location_memory();
+        auto input_location = instance.location_memory();
+        auto location_layout = input_location->get_layout();
         const int num_of_images = static_cast<int>(locations.size());
-        mem_lock<dtype> lock{input_location};
+        mem_lock<dtype> lock{input_location, stream};
         auto location_data = lock.begin();
-        assert(num_of_priors * num_loc_classes * PRIOR_BOX_SIZE == input_location.get_layout().size.feature[0]);
+        assert(num_of_priors * num_loc_classes * PRIOR_BOX_SIZE == input_location->get_layout().size.feature[0]);
 
-        const auto& input_buffer_size = input_location.get_layout().get_buffer_size();
+        const auto& input_buffer_size = location_layout.get_buffer_size();
         const int input_buffer_size_x = input_buffer_size.spatial[0];
         const int input_buffer_size_y = input_buffer_size.spatial[1];
         const int input_buffer_size_f = input_buffer_size.feature[0];
-        const auto& input_padding = input_location.get_layout().data_padding;
+        const auto& input_padding = location_layout.data_padding;
         const int input_padding_lower_x = input_padding.lower_size().spatial[0];
         const int input_padding_lower_y = input_padding.lower_size().spatial[1];
 
@@ -467,16 +470,16 @@ struct detection_output_cpu : typed_primitive_impl<detection_output> {
     }
 
     template <typename dtype>
-    void extract_prior_boxes_and_variances(const detection_output_inst& instance,
+    void extract_prior_boxes_and_variances(stream& stream, const detection_output_inst& instance,
                                            const bool variance_encoded_in_target,
                                            const int32_t prior_info_size,
                                            const int32_t prior_coordinates_offset,
                                            const int32_t images_count,
                                            std::vector<bounding_box>& prior_bboxes,
                                            std::vector<std::array<float, PRIOR_BOX_SIZE>>& prior_variances) {
-        auto& input_prior_box = instance.prior_box_memory();
+        auto input_prior_box = instance.prior_box_memory();
         const int num_of_priors = static_cast<int>(prior_bboxes.size()) / images_count;
-        mem_lock<dtype> lock{input_prior_box};
+        mem_lock<dtype> lock{input_prior_box, stream};
         for (int i = 0; i < images_count; i++) {
             auto prior_box_data =
                 lock.begin() + i * num_of_priors * prior_info_size * (variance_encoded_in_target ? 1 : 2);
@@ -503,25 +506,25 @@ struct detection_output_cpu : typed_primitive_impl<detection_output> {
     }
 
     template <typename dtype>
-    void extract_confidences_per_image_caffe(const detection_output_inst& instance,
+    void extract_confidences_per_image_caffe(stream& stream, const detection_output_inst& instance,
                                              std::vector<std::vector<std::vector<std::pair<float, int>>>>& confidences,
                                              const int num_of_priors) {
         const int num_classes = instance.argument.num_classes;
 
         const int num_of_images = static_cast<int>(confidences.size());
-        auto& input_confidence = instance.confidence_memory();
+        auto input_confidence = instance.confidence_memory();
         const float confidence_threshold = instance.argument.confidence_threshold;
 
-        mem_lock<dtype> lock{(memory_impl::ptr) &input_confidence};
+        mem_lock<dtype> lock{input_confidence, stream};
         auto confidence_data = lock.begin();
 
-        assert(num_of_priors * num_classes == input_confidence.get_layout().size.feature[0]);
+        assert(num_of_priors * num_classes == input_confidence->get_layout().size.feature[0]);
 
-        const auto& input_buffer_size = input_confidence.get_layout().get_buffer_size();
+        const auto& input_buffer_size = input_confidence->get_layout().get_buffer_size();
         const int input_buffer_size_x = input_buffer_size.spatial[0];
         const int input_buffer_size_y = input_buffer_size.spatial[1];
         const int input_buffer_size_f = input_buffer_size.feature[0];
-        const auto& input_padding = input_confidence.get_layout().data_padding;
+        const auto& input_padding = input_confidence->get_layout().data_padding;
         const int input_padding_lower_x = input_padding.lower_size().spatial[0];
         const int input_padding_lower_y = input_padding.lower_size().spatial[1];
         const int stride = input_buffer_size_y * input_buffer_size_x;
@@ -593,26 +596,27 @@ struct detection_output_cpu : typed_primitive_impl<detection_output> {
     }
 
     template <typename dtype>
-    void extract_confidences_per_image_mxnet(const detection_output_inst& instance,
+    void extract_confidences_per_image_mxnet(stream& stream, const detection_output_inst& instance,
                                              std::vector<std::vector<std::vector<std::pair<float, int>>>>& confidences,
                                              const int num_of_priors,
                                              std::vector<std::vector<std::pair<float, std::pair<int, int>>>>& scoreIndexPairs) {
         const int num_classes = instance.argument.num_classes;
 
         const int num_of_images = static_cast<int>(confidences.size());
-        auto& input_confidence = instance.confidence_memory();
+        auto input_confidence = instance.confidence_memory();
         const float confidence_threshold = instance.argument.confidence_threshold;
+        auto confidence_layout = input_confidence->get_layout();
 
-        mem_lock<dtype> lock{(memory_impl::ptr) &input_confidence};
+        mem_lock<dtype> lock{input_confidence, stream};
         auto confidence_data = lock.begin();
 
-        assert(num_of_priors * num_classes == input_confidence.get_layout().size.feature[0]);
+        assert(num_of_priors * num_classes == confidence_layout.size.feature[0]);
 
-        const auto& input_buffer_size = input_confidence.get_layout().get_buffer_size();
+        const auto& input_buffer_size = confidence_layout.get_buffer_size();
         const int input_buffer_size_x = input_buffer_size.spatial[0];
         const int input_buffer_size_y = input_buffer_size.spatial[1];
         const int input_buffer_size_f = input_buffer_size.feature[0];
-        const auto& input_padding = input_confidence.get_layout().data_padding;
+        const auto& input_padding = confidence_layout.data_padding;
         const int input_padding_lower_x = input_padding.lower_size().spatial[0];
         const int input_padding_lower_y = input_padding.lower_size().spatial[1];
         const int stride = input_buffer_size_y * input_buffer_size_x;
@@ -703,7 +707,7 @@ struct detection_output_cpu : typed_primitive_impl<detection_output> {
     }
 
     template <typename dtype>
-    void prepare_data(const detection_output_inst& instance,
+    void prepare_data(stream& stream, const detection_output_inst& instance,
                       std::vector<std::vector<std::vector<bounding_box>>>& bboxes,
                       std::vector<std::vector<std::vector<std::pair<float, int>>>>& confidences,
                       std::vector<std::vector<std::pair<float, std::pair<int, int>>>>& scoreIndexPairs) {
@@ -711,23 +715,26 @@ struct detection_output_cpu : typed_primitive_impl<detection_output> {
 
         const auto& args = instance.argument;
 
+        auto priors_layout = instance.prior_box_memory()->get_layout();
+
         const int num_of_images = static_cast<int>(bboxes.size());
-        const int num_of_priors = instance.prior_box_memory().get_layout().size.spatial[1] / args.prior_info_size;
+        const int num_of_priors = priors_layout.size.spatial[1] / args.prior_info_size;
         const int num_loc_classes = args.share_location ? 1 : args.num_classes;
 
         // Extract locations per image.
         std::vector<std::vector<std::vector<bounding_box>>> locations(
             num_of_images);  // Per image : label -> bounding boxes.
-        extract_locations_per_image<dtype>(instance, locations, num_of_priors, num_loc_classes);
+        extract_locations_per_image<dtype>(stream, instance, locations, num_of_priors, num_loc_classes);
 
-        int32_t batches_in_prior_boxes = instance.prior_box_memory().get_layout().size.batch[0];
+        int32_t batches_in_prior_boxes = priors_layout.size.batch[0];
         std::vector<bounding_box> prior_bboxes(batches_in_prior_boxes *
                                                num_of_priors);  // Prior-Boxes (identical for all images since we assume
                                                                 // all images in a batch are of same dimension).
         std::vector<std::array<float, PRIOR_BOX_SIZE>> prior_variances(
             batches_in_prior_boxes * num_of_priors);  // Variances per prior-box (identical for all images since we
                                                       // assume all images in a batch are of same dimension).
-        extract_prior_boxes_and_variances<dtype>(instance,
+        extract_prior_boxes_and_variances<dtype>(stream,
+                                                 instance,
                                                  args.variance_encoded_in_target,
                                                  args.prior_info_size,
                                                  args.prior_coordinates_offset,
@@ -770,38 +777,39 @@ struct detection_output_cpu : typed_primitive_impl<detection_output> {
         }
         // Extract confidences per image.
         if (nms_type == CAFFE) {
-            extract_confidences_per_image_caffe<dtype>(instance, confidences, num_of_priors);
+            extract_confidences_per_image_caffe<dtype>(stream, instance, confidences, num_of_priors);
         } else {
-            extract_confidences_per_image_mxnet<dtype>(instance, confidences, num_of_priors, scoreIndexPairs);
+            extract_confidences_per_image_mxnet<dtype>(stream, instance, confidences, num_of_priors, scoreIndexPairs);
         }
     }
 
-    event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& events, detection_output_inst& instance) override {
+    event::ptr execute_impl(const std::vector<event::ptr>& events, detection_output_inst& instance) override {
         for (auto& a : events) {
             a->wait();
         }
 
-        auto ev = instance.get_network().get_engine().create_user_event(instance.get_network().get_id(), false);
+        auto& stream = instance.get_network().get_stream();
 
-        const int num_of_images = instance.location_memory().get_layout().size.batch[0];  // batch size
+        const int num_of_images = instance.location_memory()->get_layout().size.batch[0];  // batch size
         // Per image : label -> decoded bounding boxes.
         std::vector<std::vector<std::vector<bounding_box>>> bboxes(num_of_images);
         // Per image : class -> confidences per bounding box.
         std::vector<std::vector<std::vector<std::pair<float, int>>>> confidences(num_of_images);
+
         std::vector<std::vector<std::pair<float, std::pair<int, int>>>> scoreIndexPairs;
-        if (instance.location_memory().get_layout().data_type == data_types::f32) {
-            prepare_data<data_type_to_type<data_types::f32>::type>(instance, bboxes, confidences, scoreIndexPairs);
-            generate_detections<data_type_to_type<data_types::f32>::type>(instance, num_of_images, bboxes, confidences, scoreIndexPairs);
+        if (instance.location_memory()->get_layout().data_type == data_types::f32) {
+            prepare_data<data_type_to_type<data_types::f32>::type>(stream, instance, bboxes, confidences, scoreIndexPairs);
+            generate_detections<data_type_to_type<data_types::f32>::type>(stream, instance, num_of_images, bboxes, confidences, scoreIndexPairs);
         } else {
-            prepare_data<data_type_to_type<data_types::f16>::type>(instance, bboxes, confidences, scoreIndexPairs);
-            generate_detections<data_type_to_type<data_types::f16>::type>(instance, num_of_images, bboxes, confidences, scoreIndexPairs);
+            prepare_data<data_type_to_type<data_types::f16>::type>(stream, instance, bboxes, confidences, scoreIndexPairs);
+            generate_detections<data_type_to_type<data_types::f16>::type>(stream, instance, num_of_images, bboxes, confidences, scoreIndexPairs);
         }
 
-        dynamic_cast<cldnn::user_event*>(ev.get())->set();  // set as complete
-        // TODO: consider refactoring create_user_event() to return cldnn::user_event*
-        return ev;
+        return stream.create_user_event(true);
     }
 
+    void init_kernels() override {}
+
     static primitive_impl* create(const detection_output_node& arg) { return new detection_output_cpu(arg); }
 };
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/device_cache_reader.h b/inference-engine/thirdparty/clDNN/src/gpu/device_cache_reader.h
deleted file mode 100644
index 51c50b0df774fb..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/device_cache_reader.h
+++ /dev/null
@@ -1,27 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-#include <memory>
-#include "document.h"
-#include <string>
-
-namespace kernel_selector {
-class TuningCache;
-}
-
-namespace cldnn {
-namespace gpu {
-
-class device_cache_reader {
-public:
-    explicit device_cache_reader(const std::string tuning_file_path);
-    std::shared_ptr<kernel_selector::TuningCache> get() { return _dev_cache; }
-
-private:
-    std::shared_ptr<kernel_selector::TuningCache> _dev_cache;
-};
-
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/device_info.h b/inference-engine/thirdparty/clDNN/src/gpu/device_info.h
deleted file mode 100644
index 225ed453c2b890..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/device_info.h
+++ /dev/null
@@ -1,58 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-#include <cstdint>
-#include <memory>
-#include "api/device.hpp"
-#include <string>
-
-namespace cl {
-class Device;
-}
-namespace cldnn {
-namespace gpu {
-
-struct device_info_internal : cldnn::device_info {
-    std::uint32_t compute_units_count;
-    uint32_t vendor_id;
-    uint8_t supports_usm;
-    bool supports_optimization_hints;
-    bool supports_local_block_io;
-
-    explicit device_info_internal(const cl::Device& device);
-
-    device_info convert_to_api() {
-        return { cores_count,
-         core_frequency,
-         max_threads_per_execution_unit,
-         max_threads_per_device,
-         max_work_group_size,
-         max_local_mem_size,
-         max_global_mem_size,
-         max_alloc_mem_size,
-         max_image2d_width,
-         max_image2d_height,
-         supports_fp16,
-         supports_fp16_denorms,
-         supports_subgroups_short,
-         supports_image,
-         supports_imad,
-         supports_immad,
-         supports_usm,
-         dev_name,
-         driver_version,
-         dev_type,
-         gfx_ver,
-         device_id,
-         num_slices,
-         num_sub_slices_per_slice,
-         num_eus_per_sub_slice,
-         num_threads_per_eu,
-        };
-    }
-};
-
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/eltwise_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/eltwise_gpu.cpp
index 5ff1be35b761da..8ffc149e160609 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/eltwise_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/eltwise_gpu.cpp
@@ -5,7 +5,7 @@
 #include "eltwise_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "eltwise/eltwise_kernel_selector.h"
 #include "eltwise/eltwise_kernel_base.h"
@@ -18,10 +18,13 @@ struct eltwise_gpu : typed_primitive_gpu_impl<eltwise> {
     using parent = typed_primitive_gpu_impl<eltwise>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<eltwise_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<eltwise>& instance,
-                                                int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
+    kernel_arguments_data get_arguments(typed_primitive_inst<eltwise>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
         return args;
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/embedding_bag_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/embedding_bag_gpu.cpp
index 4de2c9cda6db33..9e705b8bcb5433 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/embedding_bag_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/embedding_bag_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "embedding_bag/embedding_bag_kernel_selector.h"
 #include "embedding_bag/embedding_bag_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "data_inst.h"
 
 using namespace cldnn;
@@ -19,6 +19,10 @@ struct embedding_bag_gpu : typed_primitive_gpu_impl<embedding_bag> {
     using parent = typed_primitive_gpu_impl<embedding_bag>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<embedding_bag_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const embedding_bag_node& arg) {
         auto embedding_bag_params = get_default_params<kernel_selector::embedding_bag_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/events_waiter.h b/inference-engine/thirdparty/clDNN/src/gpu/events_waiter.h
deleted file mode 100644
index 55be7d2e092e73..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/events_waiter.h
+++ /dev/null
@@ -1,26 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include "ocl_toolkit.h"
-#include "event_impl.h"
-#include <vector>
-#include <memory>
-
-namespace cldnn {
-namespace gpu {
-class events_waiter : public context_holder {
-public:
-    explicit events_waiter(std::shared_ptr<gpu_toolkit> context) : context_holder(context) {}
-
-    event_impl::ptr run(uint32_t queue_id, const std::vector<event_impl::ptr>& dependencies) {
-        if (dependencies.size() == 1)
-            return dependencies[0];
-
-        return context()->enqueue_marker(queue_id, dependencies);
-    }
-};
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/extract_image_patches_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/extract_image_patches_gpu.cpp
index bff7dafc638d6f..c696821b727583 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/extract_image_patches_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/extract_image_patches_gpu.cpp
@@ -5,7 +5,7 @@
 #include "extract_image_patches_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 
 #include "extract_image_patches/extract_image_patches_kernel_selector.h"
@@ -18,6 +18,10 @@ struct extract_image_patches_gpu : typed_primitive_gpu_impl<extract_image_patche
     using parent = typed_primitive_gpu_impl<extract_image_patches>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<extract_image_patches_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const extract_image_patches_node& arg) {
         auto params = get_default_params<kernel_selector::extract_image_patches_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/fully_connected_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/fully_connected_gpu.cpp
index 05cdc2eb3e1d82..2f4a6b48f7ee97 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/fully_connected_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/fully_connected_gpu.cpp
@@ -12,11 +12,11 @@
 #include "fully_connected/fully_connected_params.h"
 
 #include "network_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_runner.h"
 
-#include "api/reorder.hpp"
-#include "api/input_layout.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/input_layout.hpp"
 #include <memory>
 
 namespace cldnn {
@@ -26,13 +26,16 @@ struct fully_connected_gpu : typed_primitive_gpu_impl<fully_connected> {
     using parent = typed_primitive_gpu_impl<fully_connected>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<fully_connected_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<fully_connected>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
+    kernel_arguments_data get_arguments(typed_primitive_inst<fully_connected>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
 
-        args.weights = (memory_impl::cptr) &instance.weights_memory();
-        args.bias = (memory_impl::cptr) (instance.bias_term() ? &instance.bias_memory() : nullptr);
+        args.weights = instance.weights_memory();
+        args.bias = instance.bias_term() ? instance.bias_memory() : nullptr;
 
         return args;
     }
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/fused_conv_eltwise_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/fused_conv_eltwise_gpu.cpp
index 28d8990c9eaa23..9e608af17b13c7 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/fused_conv_eltwise_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/fused_conv_eltwise_gpu.cpp
@@ -5,7 +5,7 @@
 #include "fused_conv_eltwise_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "kernel_runner.h"
 #include "fused_conv_eltwise/fused_conv_eltwise_kernel_selector.h"
@@ -20,6 +20,10 @@ struct fused_conv_eltwise_gpu : typed_primitive_gpu_impl<fused_conv_eltwise> {
     using parent = typed_primitive_gpu_impl<fused_conv_eltwise>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<fused_conv_eltwise_gpu>(*this);
+    }
+
 protected:
     bool validate_impl(const typed_primitive_inst<fused_conv_eltwise>& instance) const override {
         (void)instance;
@@ -36,12 +40,11 @@ struct fused_conv_eltwise_gpu : typed_primitive_gpu_impl<fused_conv_eltwise> {
         return res;
     }
 
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<fused_conv_eltwise>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
+    kernel_arguments_data get_arguments(typed_primitive_inst<fused_conv_eltwise>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
 
-        args.weights = (memory_impl::cptr) &instance.weights_memory(split);
-        args.bias = (memory_impl::cptr) (instance.bias_term() ? &instance.bias_memory(split) : nullptr);
+        args.weights = instance.weights_memory(split);
+        args.bias = instance.bias_term() ? instance.bias_memory(split) : nullptr;
         return args;
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/gather_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/gather_gpu.cpp
index c596bbef83fde1..e0d51b3f0009ba 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/gather_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/gather_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "gather/gather_kernel_selector.h"
 #include "gather/gather_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -37,6 +37,10 @@ struct gather_gpu : typed_primitive_gpu_impl<gather> {
     using parent = typed_primitive_gpu_impl<gather>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<gather_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const gather_node& arg) {
         auto gather_params = get_default_params<kernel_selector::gather_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/gather_nd_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/gather_nd_gpu.cpp
index dc05203f37981e..cf7d692651a413 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/gather_nd_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/gather_nd_gpu.cpp
@@ -1,18 +1,6 @@
-/*
-// Copyright (c) 2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
 //
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-*/
 
 #include "gather_nd_inst.h"
 #include "primitive_gpu_base.h"
@@ -20,7 +8,6 @@
 #include "kernel_selector_helper.h"
 #include "gather/gather_nd_kernel_selector.h"
 #include "gather/gather_nd_kernel_ref.h"
-#include "error_handler.h"
 
 using namespace cldnn;
 
@@ -31,7 +18,10 @@ struct gather_nd_gpu : typed_primitive_gpu_impl<gather_nd> {
     using parent = typed_primitive_gpu_impl<gather_nd>;
     using parent::parent;
 
-public:
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<gather_nd_gpu>(*this);
+    }
+
     static primitive_impl* create(const gather_nd_node& arg) {
         auto gather_nd_params = get_default_params<kernel_selector::gather_nd_params>(arg);
         auto gather_nd_optional_params =
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/gather_tree_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/gather_tree_gpu.cpp
index db98e0b543c67d..f0ae5fcdbc6617 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/gather_tree_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/gather_tree_gpu.cpp
@@ -9,7 +9,8 @@
 #include "kernel_selector_helper.h"
 #include "gather_tree/gather_tree_kernel_selector.h"
 #include "gather_tree/gather_tree_kernel_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
+
 namespace cldnn {
 namespace gpu {
 
@@ -17,6 +18,10 @@ struct gather_tree_gpu : typed_primitive_gpu_impl<gather_tree> {
     using parent = typed_primitive_gpu_impl<gather_tree>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<gather_tree_gpu>(*this);
+    }
+
     static primitive_impl* create(const gather_tree_node& arg) {
         auto b_params = get_default_params<kernel_selector::gather_tree_params>(arg, 1);
         auto b_optional_params = get_default_optional_params<kernel_selector::gather_tree_optional_params>(arg.get_program());
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/gemm_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/gemm_gpu.cpp
index bc475596c7ccc8..9a6a76802dcf90 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/gemm_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/gemm_gpu.cpp
@@ -9,7 +9,7 @@
 #include "kernel_selector_helper.h"
 #include "gemm/gemm_kernel_selector.h"
 #include "gemm/gemm_kernel_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -18,6 +18,10 @@ struct gemm_gpu : typed_primitive_gpu_impl<gemm> {
     using parent = typed_primitive_gpu_impl<gemm>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<gemm_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const gemm_node& arg) {
         auto gemm_params = get_default_params<kernel_selector::gemm_params>(arg, 1);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/generic_layer_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/generic_layer_gpu.cpp
index c6643ed65beb95..a7d3d610b31081 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/generic_layer_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/generic_layer_gpu.cpp
@@ -3,11 +3,10 @@
 //
 
 #include "generic_layer_inst.h"
-#include "kernel.h"
+#include "cldnn/runtime/engine.hpp"
 #include "implementation_map.h"
 #include "kernel_selector_helper.h"
 #include "network_impl.h"
-#include "engine_impl.h"
 #include "register_gpu.hpp"
 #include <vector>
 
@@ -18,36 +17,57 @@ namespace neural {
 struct generic_layer_gpu : typed_primitive_impl<generic_layer> {
     const generic_layer_node& outer;
     const kernel_selector::cl_kernel_data& _cl_kernel_data;
-    gpu::kernel _kernel;
+    std::vector<kernel::ptr> _kernels;
+    kernel_id _kernel_id;
 
-    explicit generic_layer_gpu(const generic_layer_node& arg)
-        : outer(arg),
-          _cl_kernel_data(*outer.get_primitive()->generic_params.clKernel.get()),
-          _kernel(arg.get_program().get_engine().get_context(),
-                  outer.get_primitive()->generic_params.clKernel->kernelString,
-                  arg.get_program().get_id()) {}
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<generic_layer_gpu>(*this);
+    }
+
+    generic_layer_gpu(const generic_layer_gpu& other)
+    : outer(other.outer)
+    , _cl_kernel_data(other._cl_kernel_data)
+    , _kernels({})
+    , _kernel_id(other._kernel_id) {
+        if (other._kernels.empty()) {
+            throw std::runtime_error("Can't copy generic_layer_gpu node: kernels vector is empty");
+        }
+        _kernels.push_back(other._kernels.front()->clone());
+    }
+
+    generic_layer_gpu(const generic_layer_node& arg)
+        : outer(arg)
+        , _cl_kernel_data(*outer.get_primitive()->generic_params.clKernel.get())
+        , _kernels() {
+        _kernel_id = outer.get_program().add_kernel(outer.get_primitive()->generic_params.clKernel->code.kernelString);
+    }
+
+    void init_kernels() override {
+        _kernels.push_back(outer.get_program().get_kernel(_kernel_id));
+    }
 
     void set_arguments_impl(generic_layer_inst& instance) override {
-        auto net_id = instance.get_network().get_id();
-        gpu::kernel::kernel_arguments_data args;
-        args.scalars = &_cl_kernel_data.scalars;
+        stream& stream = instance.get_network().get_stream();
+        kernel_arguments_data args;
+        args.scalars = &_cl_kernel_data.params.scalars;
 
         for (size_t i = 0; i < instance.inputs_memory_count(); i++) {
-            args.inputs.push_back((memory_impl::cptr) &instance.input_memory(i));
+            args.inputs.push_back(instance.input_memory_ptr(i));
         }
-        args.output = (memory_impl::cptr) &instance.output_memory();
-        _kernel.set_arguments(net_id, _cl_kernel_data, args);
+        args.output = instance.output_memory_ptr();
+        stream.set_arguments(*_kernels.front(), _cl_kernel_data.params, args);
     }
 
-    void cleanup_impl(generic_layer_inst& instance) override {
-        auto net_id = instance.get_network().get_id();
-        _kernel.cleanup(net_id);
-    }
+    event::ptr execute_impl(const std::vector<event::ptr>& events, generic_layer_inst& instance) override {
+        stream& stream = instance.get_network().get_stream();
+        kernel_arguments_data args;
+        args.scalars = &_cl_kernel_data.params.scalars;
 
-    event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& events, generic_layer_inst& instance) override {
-        uint32_t net_id = instance.get_network().get_id();
-        _kernel.set_output_event(net_id, instance.node.is_output());
-        return _kernel.run(net_id, _cl_kernel_data, events);
+        for (size_t i = 0; i < instance.inputs_memory_count(); i++) {
+            args.inputs.push_back(instance.input_memory_ptr(i));
+        }
+        args.output = instance.output_memory_ptr();
+        return stream.enqueue_kernel(*_kernels.front(), _cl_kernel_data.params, args, events, true);
     }
 };
 
@@ -55,28 +75,34 @@ struct generic_layer_gpu : typed_primitive_impl<generic_layer> {
 struct generic_layer_cpu : typed_primitive_impl<generic_layer> {
     const generic_layer_node& outer;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<generic_layer_cpu>(*this);
+    }
+
     explicit generic_layer_cpu(const generic_layer_node& arg) : outer(arg) {}
 
-    event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& events, generic_layer_inst& instance) override {
-        uint32_t net_id = instance.get_network().get_id();
-        auto& input_mem = instance.input_memory();
-        auto& output_mem = instance.output_memory();
+    event::ptr execute_impl(const std::vector<event::ptr>& events, generic_layer_inst& instance) override {
+        stream& stream = instance.get_network().get_stream();
+        auto input_mem = instance.input_memory_ptr();
+        auto output_mem = instance.output_memory_ptr();
 
-        std::vector<event_impl::ptr> tmp_events(events);
+        std::vector<event::ptr> tmp_events(events);
 
         for (auto& a : events) {
             a->wait();
         }
 
-        mem_lock<uint8_t> old_pointer(input_mem);
-        mem_lock<uint8_t> new_pointer(output_mem);
+        mem_lock<uint8_t> old_pointer(input_mem, stream);
+        mem_lock<uint8_t> new_pointer(output_mem, stream);
 
         const auto& cpu_kernel = *outer.get_primitive()->generic_params.cpuKernel.get();
 
         cpu_kernel.Execute(old_pointer.data(), old_pointer.size(), new_pointer.data(), new_pointer.size());
 
-        return instance.get_network().get_engine().create_user_event(net_id, true);
+        return stream.create_user_event(true);
     }
+
+    void init_kernels() override {}
 };
 
 static primitive_impl* create(const generic_layer_node& arg) {
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/grn_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/grn_gpu.cpp
index 7612be5ccffc37..5b9d8c9145ca55 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/grn_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/grn_gpu.cpp
@@ -5,7 +5,7 @@
 #include "grn_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "grn/grn_kernel_selector.h"
 #include "grn/grn_kernel_base.h"
@@ -21,6 +21,10 @@ struct grn_gpu : typed_primitive_gpu_impl<grn> {
     using parent = typed_primitive_gpu_impl<grn>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<grn_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const grn_node& arg) {
         auto grn_params = get_default_params<kernel_selector::grn_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/kernel.cpp b/inference-engine/thirdparty/clDNN/src/gpu/kernel.cpp
deleted file mode 100644
index 0ba0ece0443359..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/kernel.cpp
+++ /dev/null
@@ -1,275 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-
-#include <iterator>
-#include "kernel.h"
-#include "memory_gpu.h"
-#include "memory_impl.h"
-#include "refcounted_obj.h"
-#include <vector>
-
-namespace cldnn {
-namespace gpu {
-
-namespace {
-inline cl::NDRange toNDRange(const std::vector<size_t>& v) {
-    switch (v.size()) {
-        case 1:
-            return cl::NDRange(v[0]);
-        case 2:
-            return cl::NDRange(v[0], v[1]);
-        case 3:
-            return cl::NDRange(v[0], v[1], v[2]);
-        default:
-            return cl::NullRange;
-    }
-}
-
-void set_arguments_impl(kernels_cache::kernel_type& kernel,
-                        const kernel_selector::kernel_arguments& args,
-                        const kernel::kernel_arguments_data& data) {
-    for (uint32_t i = 0; i < static_cast<uint32_t>(args.size()); i++) {
-        cl_int status = CL_INVALID_ARG_VALUE;
-        switch (args[i].t) {
-            case kernel_selector::kernel_argument_types::INPUT:
-                if (args[i].index < data.inputs.size() && data.inputs[args[i].index]) {
-                    const auto& input_mem = data.inputs[args[i].index];
-                    if (input_mem) {
-                        if (input_mem->get_layout().format.is_image_2d())
-                            status = kernel.setArg(i, dynamic_cast<const gpu::gpu_image2d&>(*input_mem).get_buffer());
-                        else if (memory_capabilities::is_usm_type(input_mem->get_allocation_type()))
-                            status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*input_mem).get_buffer());
-                        else
-                            status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*input_mem).get_buffer());
-                    }
-                }
-                break;
-            case kernel_selector::kernel_argument_types::INPUT_OF_FUSED_PRIMITIVE:
-                if (args[i].index < data.fused_op_inputs.size() && data.fused_op_inputs[args[i].index]) {
-                    const auto& input_mem = data.fused_op_inputs[args[i].index];
-                    if (input_mem) {
-                        if (memory_capabilities::is_usm_type(input_mem->get_allocation_type()))
-                            status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*input_mem).get_buffer());
-                        else
-                            status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*input_mem).get_buffer());
-                    }
-                }
-                break;
-            case kernel_selector::kernel_argument_types::INTERNAL_BUFFER:
-                if (args[i].index < data.intermediates.size() && data.intermediates[args[i].index]) {
-                    const auto& input_mem = data.intermediates[args[i].index];
-                    if (input_mem) {
-                        if (memory_capabilities::is_usm_type(input_mem->get_allocation_type()))
-                            status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*input_mem).get_buffer());
-                        else
-                            status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*input_mem).get_buffer());
-                    }
-                }
-                break;
-            case kernel_selector::kernel_argument_types::OUTPUT:
-                if (data.output) {
-                     if (data.output->get_layout().format.is_image_2d())
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_image2d&>(*data.output).get_buffer());
-                     else if (memory_capabilities::is_usm_type(data.output->get_allocation_type()))
-                         status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*data.output).get_buffer());
-                     else
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*data.output).get_buffer());
-                }
-                break;
-            case kernel_selector::kernel_argument_types::WEIGHTS:
-                if (data.weights) {
-                    if (data.weights->get_layout().format.is_image_2d())
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_image2d&>(*data.weights).get_buffer());
-                    else if (memory_capabilities::is_usm_type(data.weights->get_allocation_type()))
-                        status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*data.weights).get_buffer());
-                    else
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*data.weights).get_buffer());
-                }
-                break;
-            case kernel_selector::kernel_argument_types::BIAS:
-                if (data.bias) {
-                    if (memory_capabilities::is_usm_type(data.bias->get_allocation_type()))
-                        status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*data.bias).get_buffer());
-                    else
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*data.bias).get_buffer());
-                }
-                break;
-            case kernel_selector::kernel_argument_types::WEIGHTS_ZERO_POINTS:
-                if (data.weights_zero_points) {
-                    if (memory_capabilities::is_usm_type(data.weights_zero_points->get_allocation_type()))
-                        status = kernel.setArgUsm(
-                            i,
-                            dynamic_cast<const gpu::gpu_usm&>(*data.weights_zero_points).get_buffer());
-                    else
-                        status = kernel.setArg(
-                            i,
-                            dynamic_cast<const gpu::gpu_buffer&>(*data.weights_zero_points).get_buffer());
-                }
-                break;
-            case kernel_selector::kernel_argument_types::ACTIVATIONS_ZERO_POINTS:
-                if (data.activations_zero_points) {
-                    if (memory_capabilities::is_usm_type(data.activations_zero_points->get_allocation_type()))
-                        status = kernel.setArgUsm(
-                            i,
-                            dynamic_cast<const gpu::gpu_usm&>(*data.activations_zero_points).get_buffer());
-                    else
-                        status = kernel.setArg(
-                            i,
-                            dynamic_cast<const gpu::gpu_buffer&>(*data.activations_zero_points).get_buffer());
-                }
-                break;
-            case kernel_selector::kernel_argument_types::COMPENSATION:
-                if (data.compensation) {
-                    if (memory_capabilities::is_usm_type(data.compensation->get_allocation_type()))
-                        status = kernel.setArgUsm(
-                                i,
-                                dynamic_cast<const gpu::gpu_usm&>(*data.compensation).get_buffer());
-                    else
-                        status = kernel.setArg(
-                                 i,
-                                 dynamic_cast<const gpu::gpu_buffer&>(*data.compensation).get_buffer());
-                }
-                break;
-            case kernel_selector::kernel_argument_types::SCALE_TABLE:
-                if (data.scale_table) {
-                    if (memory_capabilities::is_usm_type(data.scale_table->get_allocation_type()))
-                        status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*data.scale_table).get_buffer());
-                    else
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*data.scale_table).get_buffer());
-                }
-                break;
-            case kernel_selector::kernel_argument_types::SLOPE:
-                if (data.slope) {
-                    if (memory_capabilities::is_usm_type(data.slope->get_allocation_type()))
-                        status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*data.slope).get_buffer());
-                    else
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*data.slope).get_buffer());
-                }
-                break;
-            case kernel_selector::kernel_argument_types::SPLIT:
-                status = kernel.setArg(i, data.split);
-                break;
-            case kernel_selector::kernel_argument_types::SCALAR:
-                if (data.scalars && args[i].index < data.scalars->size()) {
-                    const auto& scalar = (*data.scalars)[args[i].index];
-                    switch (scalar.t) {
-                        case kernel_selector::kernel_scalar_argument_types::UINT8:
-                            status = kernel.setArg(i, scalar.v.u8);
-                            break;
-                        case kernel_selector::kernel_scalar_argument_types::UINT16:
-                            status = kernel.setArg(i, scalar.v.u16);
-                            break;
-                        case kernel_selector::kernel_scalar_argument_types::UINT32:
-                            status = kernel.setArg(i, scalar.v.u32);
-                            break;
-                        case kernel_selector::kernel_scalar_argument_types::UINT64:
-                            status = kernel.setArg(i, scalar.v.u64);
-                            break;
-                        case kernel_selector::kernel_scalar_argument_types::INT8:
-                            status = kernel.setArg(i, scalar.v.s8);
-                            break;
-                        case kernel_selector::kernel_scalar_argument_types::INT16:
-                            status = kernel.setArg(i, scalar.v.s16);
-                            break;
-                        case kernel_selector::kernel_scalar_argument_types::INT32:
-                            status = kernel.setArg(i, scalar.v.s32);
-                            break;
-                        case kernel_selector::kernel_scalar_argument_types::INT64:
-                            status = kernel.setArg(i, scalar.v.s64);
-                            break;
-                        case kernel_selector::kernel_scalar_argument_types::FLOAT32:
-                            status = kernel.setArg(i, scalar.v.f32);
-                            break;
-                        case kernel_selector::kernel_scalar_argument_types::FLOAT64:
-                            status = kernel.setArg(i, scalar.v.f64);
-                            break;
-                        default:
-                            break;
-                    }
-                }
-                break;
-            case kernel_selector::kernel_argument_types::RECURRENT:  // RNN/LSTM/GRU layers
-                if (data.recurrent) {
-                    if (data.recurrent->get_layout().format.is_image_2d())
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_image2d&>(*data.recurrent).get_buffer());
-                    else if (memory_capabilities::is_usm_type(data.recurrent->get_allocation_type()))
-                        status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*data.recurrent).get_buffer());
-                    else
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*data.recurrent).get_buffer());
-                }
-                break;
-            case kernel_selector::kernel_argument_types::HIDDEN:  // RNN/LSTM/GRU layers
-                if (data.hidden) {
-                    if (data.hidden->get_layout().format.is_image_2d())
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_image2d&>(*data.hidden).get_buffer());
-                    else if (memory_capabilities::is_usm_type(data.hidden->get_allocation_type()))
-                        status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*data.hidden).get_buffer());
-                    else
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*data.hidden).get_buffer());
-                }
-                break;
-            case kernel_selector::kernel_argument_types::CELL:  // LSTMlayers
-                if (data.cell) {
-                    if (data.cell->get_layout().format.is_image_2d())
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_image2d&>(*data.cell).get_buffer());
-                    else if (memory_capabilities::is_usm_type(data.cell->get_allocation_type()))
-                        status = kernel.setArgUsm(i, dynamic_cast<const gpu::gpu_usm&>(*data.cell).get_buffer());
-                    else
-                        status = kernel.setArg(i, dynamic_cast<const gpu::gpu_buffer&>(*data.cell).get_buffer());
-                }
-                break;
-            default:
-                break;
-        }
-
-        if (status != CL_SUCCESS) {
-            throw std::runtime_error("Error set arg " + std::to_string(i) + ", error code: " + std::to_string(status) + "\n");
-        }
-    }
-}
-}  // namespace
-
-void kernel::set_arguments(uint32_t queue_id,
-                           const kernel_selector::cl_kernel_data& kernel_data,
-                           const kernel_arguments_data& args) {
-    static std::mutex m;
-    std::lock_guard<std::mutex> guard(m);
-    auto compiled_kernel = context()->get_kernels_cache(_prog_id).get_kernel(_kernel_id, _one_time_kernel);
-
-    // Create a copy of cl kernel for each stream if it doesn't exist
-    // Copy is needed to avoid data races between streams, but we create it only once for each stream
-    // because the cloning is quite expensive.
-    // Mutex is still needed to ensure that insert operation into the map is thread safe
-    if (_cl_kernels.find(queue_id) == _cl_kernels.end())
-        _cl_kernels[queue_id] = compiled_kernel.clone();
-
-    try {
-        set_arguments_impl(_cl_kernels.at(queue_id), kernel_data.arguments, args);
-    } catch (cl::Error const& err) {
-        throw ocl_error(err);
-    }
-}
-
-void kernel::cleanup(uint32_t queue_id) {
-    _cl_kernels.erase(queue_id);
-}
-
-event_impl::ptr kernel::run(uint32_t queue_id,
-                            const kernel_selector::cl_kernel_data& kernel_data,
-                            const std::vector<event_impl::ptr>& dependencies) const {
-    if (_cl_kernels.find(queue_id) == _cl_kernels.end() || _cl_kernels.at(queue_id).get() == NULL) {
-        throw std::runtime_error("[clDNN] Kernel for layer " + kernel_data.layerID + " is not found for stream " + std::to_string(queue_id));
-    }
-
-    return context()->enqueue_kernel(queue_id,
-                                     _cl_kernels.at(queue_id),
-                                     toNDRange(kernel_data.workGroups.global),
-                                     toNDRange(kernel_data.workGroups.local),
-                                     dependencies);
-}
-
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/kernel.h b/inference-engine/thirdparty/clDNN/src/gpu/kernel.h
deleted file mode 100644
index c55f2ca615b3ef..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/kernel.h
+++ /dev/null
@@ -1,96 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-
-#include "ocl_toolkit.h"
-#include "memory_impl.h"
-#include "kernels_cache.h"
-#include "event_impl.h"
-
-#include "kernel_selector_helper.h"
-#include <memory>
-#include <vector>
-
-namespace cldnn {
-namespace gpu {
-
-class kernel : public context_holder {
-    uint32_t _prog_id;
-    kernels_cache::kernel_id _kernel_id;
-    bool _one_time_kernel;  // If this flag is true, the kernel is intended to be executed only once (can be removed
-                            // later from the cache).
-
-    std::map<uint32_t, kernels_cache::kernel_type> _cl_kernels;
-
-public:
-    explicit kernel(std::shared_ptr<gpu_toolkit> context,
-                    const std::shared_ptr<kernel_selector::kernel_string>& kernel_string,
-                    uint32_t prog_id,
-                    bool dump_custom_program = false,
-                    bool one_time_kernel = false)
-        : context_holder(context)
-        , _prog_id(prog_id)
-        , _kernel_id(context->get_kernels_cache(prog_id).set_kernel_source(kernel_string, dump_custom_program, one_time_kernel))
-        , _one_time_kernel(one_time_kernel)
-        , _cl_kernels({}) {}
-
-    kernel(const kernel& other)
-        : context_holder(other.context())
-        , _prog_id(other._prog_id)
-        , _kernel_id(other._kernel_id)
-        , _one_time_kernel(other._one_time_kernel)
-        , _cl_kernels(other._cl_kernels) {}
-
-    kernel& operator=(const kernel& other) {
-        if (this == &other) {
-            return *this;
-        }
-
-        _kernel_id = other._kernel_id;
-        _prog_id = other._prog_id;
-        _one_time_kernel = other._one_time_kernel;
-        _cl_kernels = other._cl_kernels;
-
-        return *this;
-    }
-
-    struct kernel_arguments_data {
-        std::vector<memory_impl::cptr> inputs;
-        std::vector<memory_impl::cptr> intermediates;
-        memory_impl::cptr output;
-        memory_impl::cptr weights;
-        memory_impl::cptr recurrent;
-        memory_impl::cptr hidden;
-        memory_impl::cptr cell;
-        memory_impl::cptr bias;
-        memory_impl::cptr weights_zero_points;
-        memory_impl::cptr activations_zero_points;
-        memory_impl::cptr compensation;
-        memory_impl::cptr lookup_table;
-        memory_impl::cptr scale_table;
-        memory_impl::cptr slope;
-        // used for fused primitives
-        std::vector<memory_impl::cptr> fused_op_inputs;
-        int32_t split = 0;
-        float lr;
-        const kernel_selector::kernel_scalar_arguments* scalars = nullptr;
-    };
-
-    void set_output_event(uint32_t net_id, bool is_out_event) {
-        context()->set_output_event(net_id, is_out_event);
-    }
-
-    void cleanup(uint32_t queue_id);
-    void set_arguments(uint32_t queue_id,
-                       const kernel_selector::cl_kernel_data& kernel_data,
-                       const kernel_arguments_data& args);
-    event_impl::ptr run(uint32_t queue_id,
-                        const kernel_selector::cl_kernel_data& kernel_data,
-                        const std::vector<event_impl::ptr>& dependencies) const;
-};
-
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/kernel_runner.cpp b/inference-engine/thirdparty/clDNN/src/gpu/kernel_runner.cpp
index 2faddb64a6ac5d..8f937863b52685 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/kernel_runner.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/kernel_runner.cpp
@@ -5,8 +5,10 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
 #include "kernel_runner.h"
-#include "kernel.h"
+#include "runtime/kernels_cache.hpp"
+#include "cldnn/runtime/stream.hpp"
 #include "weight_bias_params.h"
+#include "kernel_selector_helper.h"
 #include <chrono>
 #include <vector>
 #include <limits>
@@ -15,19 +17,18 @@
 namespace cldnn {
 namespace gpu {
 
-kernel_runner::kernel_runner(engine_impl& engine_ref, uint32_t program_id, bool weights_and_bias_exist, bool zero_points_exist)
-    : engine(&engine_ref), program_id(program_id), weights_and_bias_exist(weights_and_bias_exist), zero_points_exist(zero_points_exist) {}
+kernel_runner::kernel_runner(engine& engine_ref, uint32_t program_id, bool weights_and_bias_exist, bool zero_points_exist)
+    : _engine(engine_ref), program_id(program_id), weights_and_bias_exist(weights_and_bias_exist), zero_points_exist(zero_points_exist) {}
 
 void kernel_runner::prepare_kernel_args(const kernel_selector::KernelsData& kernels_data,
-                                        gpu::kernel::kernel_arguments_data& args) {
+                                        kernel_arguments_data& args) {
     const auto& base_params = *static_cast<kernel_selector::base_params*>(kernels_data[0].params.get());
     // Prepare input buffers
     if (input_buffers.empty()) {
         for (const auto& input : base_params.inputs) {
             int num_of_input_elements = static_cast<int>(input.PhysicalSize());
-            input_buffers.push_back(engine->allocate_memory(
-                {from_data_type(input.GetDType()), format::bfyx, tensor(1, 1, num_of_input_elements, 1)},
-                0));
+            input_buffers.push_back(_engine.allocate_memory(
+                {from_data_type(input.GetDType()), format::bfyx, tensor(1, 1, num_of_input_elements, 1)}));
         }
     }
     for (const auto& input : input_buffers) {
@@ -38,9 +39,8 @@ void kernel_runner::prepare_kernel_args(const kernel_selector::KernelsData& kern
         for (auto& fused_op : base_params.fused_ops) {
             for (auto& fused_ops_input : fused_op.tensors) {
                 auto num_of_elements = static_cast<int>(fused_ops_input.PhysicalSize());
-                fused_ops_buffers.push_back(engine->allocate_memory(
-                    { from_data_type(fused_ops_input.GetDType()), format::bfyx, tensor(1, 1, num_of_elements, 1) },
-                    0));
+                fused_ops_buffers.push_back(_engine.allocate_memory(
+                    { from_data_type(fused_ops_input.GetDType()), format::bfyx, tensor(1, 1, num_of_elements, 1) }));
             }
         }
     }
@@ -50,9 +50,8 @@ void kernel_runner::prepare_kernel_args(const kernel_selector::KernelsData& kern
     // Prepare output buffer
     if (output_buffers.empty()) {
         int num_of_output_elements = static_cast<int>(base_params.output.PhysicalSize());
-        output_buffers.push_back(engine->allocate_memory(
-            {from_data_type(base_params.output.GetDType()), format::bfyx, tensor(1, 1, num_of_output_elements, 1)},
-            0));
+        output_buffers.push_back(_engine.allocate_memory(
+            {from_data_type(base_params.output.GetDType()), format::bfyx, tensor(1, 1, num_of_output_elements, 1)}));
     }
 
     args.output = output_buffers[0];
@@ -72,17 +71,15 @@ void kernel_runner::prepare_kernel_args(const kernel_selector::KernelsData& kern
         if (!cldnn::format::is_image_2d(from_weights_layout(weights_bias_params.weights.GetLayout()))) {
             if (weight_buffers.empty())
                 weight_buffers.push_back(
-                    engine->allocate_memory({from_weights_type(weights_bias_params.weights.GetDType()),
+                    _engine.allocate_memory({from_weights_type(weights_bias_params.weights.GetDType()),
                                              fmt,
-                                             tensor(num_of_weight_elements_ofm, 1, num_of_weight_elements_spatial, 1)},
-                                            0));
+                                             tensor(num_of_weight_elements_ofm, 1, num_of_weight_elements_spatial, 1)}));
 
             if (weight_buffers[0]->get_layout().format != fmt)
                 weight_buffers[0] =
-                    engine->allocate_memory({from_weights_type(weights_bias_params.weights.GetDType()),
+                    _engine.allocate_memory({from_weights_type(weights_bias_params.weights.GetDType()),
                                              fmt,
-                                             tensor(num_of_weight_elements_ofm, 1, num_of_weight_elements_spatial, 1)},
-                                            0);
+                                             tensor(num_of_weight_elements_ofm, 1, num_of_weight_elements_spatial, 1)});
 
             while (weight_buffers[0]->get_layout().bytes_count() < weights_bias_params.weights.PhysicalSizeInBytes()) {
                 // Weights layout depends on the kernel. Multiply the buffer size by 2 until it is big enough
@@ -90,22 +87,20 @@ void kernel_runner::prepare_kernel_args(const kernel_selector::KernelsData& kern
                 weight_buffers.clear();
                 num_of_weight_elements_spatial *= 2;
                 weight_buffers.push_back(
-                    engine->allocate_memory({from_weights_type(weights_bias_params.weights.GetDType()),
+                    _engine.allocate_memory({from_weights_type(weights_bias_params.weights.GetDType()),
                                              fmt,
-                                             tensor(num_of_weight_elements_ofm, 1, num_of_weight_elements_spatial, 1)},
-                                            0));
+                                             tensor(num_of_weight_elements_ofm, 1, num_of_weight_elements_spatial, 1)}));
             }
         } else {
             weight_buffers.clear();
             fmt = from_weights_layout(weights_bias_params.weights.GetLayout());
             num_of_weight_elements_ofm = static_cast<int>(weights_bias_params.weights.OFM().v);
-            weight_buffers.push_back(engine->allocate_memory({from_weights_type(weights_bias_params.weights.GetDType()),
+            weight_buffers.push_back(_engine.allocate_memory({from_weights_type(weights_bias_params.weights.GetDType()),
                                                               fmt,
                                                               tensor(num_of_weight_elements_ofm,
                                                                      num_of_weight_elements_ifm,
                                                                      num_of_weight_elements_spatial_x,
-                                                                     num_of_weight_elements_spatial_y)},
-                                                             0));
+                                                                     num_of_weight_elements_spatial_y)}));
         }
         args.weights = weight_buffers[0];
 
@@ -113,10 +108,9 @@ void kernel_runner::prepare_kernel_args(const kernel_selector::KernelsData& kern
         if (!weights_bias_params.bias.empty()) {
             if (bias_buffers.empty()) {
                 int num_of_bias_elements = static_cast<int>(weights_bias_params.bias[0].PhysicalSize());
-                bias_buffers.push_back(engine->allocate_memory({from_data_type(weights_bias_params.bias[0].GetDType()),
+                bias_buffers.push_back(_engine.allocate_memory({from_data_type(weights_bias_params.bias[0].GetDType()),
                                                                 format::bfyx,
-                                                                tensor(1, num_of_bias_elements, 1, 1)},
-                                                               0));
+                                                                tensor(1, num_of_bias_elements, 1, 1)}));
             }
             args.bias = bias_buffers[0];
         }
@@ -128,11 +122,10 @@ void kernel_runner::prepare_kernel_args(const kernel_selector::KernelsData& kern
                     auto& weight_zero_point = zero_point_params.weights_zero_points[0];
                     auto num_of_elements = static_cast<int>(weight_zero_point.PhysicalSize());
                     weight_zero_point_buffers.push_back(
-                        engine->allocate_memory({
+                        _engine.allocate_memory({
                             from_data_type(weight_zero_point.GetDType()),
                             format::bfyx,
-                            tensor(1, num_of_elements, 1, 1) },
-                            0));
+                            tensor(1, num_of_elements, 1, 1) }));
                 }
                 args.weights_zero_points = weight_zero_point_buffers[0];
             }
@@ -141,11 +134,10 @@ void kernel_runner::prepare_kernel_args(const kernel_selector::KernelsData& kern
                     auto& activation_zero_point = zero_point_params.activations_zero_points[0];
                     auto num_of_elements = static_cast<int>(activation_zero_point.PhysicalSize());
                     activation_zero_point_buffers.push_back(
-                        engine->allocate_memory({
+                        _engine.allocate_memory({
                             from_data_type(activation_zero_point.GetDType()),
                             format::bfyx,
-                            tensor(1, num_of_elements, 1, 1) },
-                            0));
+                            tensor(1, num_of_elements, 1, 1) }));
                 }
                 args.activations_zero_points = activation_zero_point_buffers[0];
             }
@@ -154,11 +146,10 @@ void kernel_runner::prepare_kernel_args(const kernel_selector::KernelsData& kern
                     auto& compensation = zero_point_params.compensation[0];
                     auto num_of_elements = static_cast<int>(compensation.PhysicalSize());
                     compensation_buffers.push_back(
-                        engine->allocate_memory({
+                        _engine.allocate_memory({
                             from_data_type(compensation.GetDType()),
                             format::bfyx,
-                            tensor(1, num_of_elements, 1, 1) },
-                            0));
+                            tensor(1, num_of_elements, 1, 1) }));
                 }
                 args.compensation = compensation_buffers[0];
             }
@@ -169,10 +160,10 @@ void kernel_runner::prepare_kernel_args(const kernel_selector::KernelsData& kern
 }
 
 std::vector<std::chrono::nanoseconds> kernel_runner::run_kernels(const kernel_selector::KernelsData& kernels_data) {
-    auto context = engine->get_context();
-
     std::vector<std::chrono::nanoseconds> run_times;
 
+    stream::ptr stream = _engine.create_stream();
+
     int num_of_kernels_to_run = static_cast<int>(kernels_data.size());
     int num_of_kernels_run = 0;
 
@@ -182,28 +173,31 @@ std::vector<std::chrono::nanoseconds> kernel_runner::run_kernels(const kernel_se
         int current_compilation_batch = std::min(num_of_kernels_to_run, compilation_batch_size);
         batch_end = batch_start + current_compilation_batch;
 
-        std::vector<gpu::kernel> kernels;
+        std::vector<kernel::ptr> kernels;
+        kernels_cache cache(_engine);
 
         for (auto it = batch_start; it < batch_end; it++) {
-            kernels.push_back(kernel(context, it->kernels[0].kernelString, program_id, false, true));
+            auto kernel_id = cache.set_kernel_source(it->kernels[0].code.kernelString, false);
+
+            kernels.push_back(cache.get_kernel(kernel_id));
         }
 
-        gpu::kernel::kernel_arguments_data args;
+        kernel_arguments_data args;
 
         prepare_kernel_args(kernels_data, args);
-        context->queue(0).finish();
+        stream->finish();
 
         int i = 0;
         for (auto it = batch_start; it < batch_end; it++) {
-            std::vector<event_impl::ptr> events;
+            std::vector<event::ptr> events;
             auto kernel_run_time = std::chrono::nanoseconds::max();
             int num_of_runs = 0;
 
             for (int iteration = 0; iteration < runs_per_kernel; iteration++) {
-                event_impl::ptr event;
+                event::ptr event;
                 try {
-                    kernels[i].set_arguments(0, it->kernels[0], args);
-                    event = kernels[i].run(0, it->kernels[0], {});
+                    stream->set_arguments(*kernels[i], it->kernels[0].params, args);
+                    event = stream->enqueue_kernel(*kernels[i], it->kernels[0].params, args, {});
                 } catch (std::exception& e) {
                     std::cout << "[clDNN] Could not run kernel for auto-tune: " << it->kernelName
                               << " with auto-tune index " << it->autoTuneIndex << std::endl
@@ -215,7 +209,7 @@ std::vector<std::chrono::nanoseconds> kernel_runner::run_kernels(const kernel_se
                 }
                 events.push_back(event);
             }
-            context->queue(0).finish();
+            stream->finish();
 
             for (auto& event : events) {
                 if (event.get() != NULL) {
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/kernel_runner.h b/inference-engine/thirdparty/clDNN/src/gpu/kernel_runner.h
index 6e257fef15f8d3..53eadc9095696f 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/kernel_runner.h
+++ b/inference-engine/thirdparty/clDNN/src/gpu/kernel_runner.h
@@ -5,10 +5,10 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 
-#include "engine_impl.h"
+#include "cldnn/runtime/engine.hpp"
 #include "kernel_selector_common.h"
+#include "kernel_selector_helper.h"
 #include "kernel_runner_interface.h"
-#include "kernel.h"
 #include <vector>
 
 namespace cldnn {
@@ -16,7 +16,7 @@ namespace gpu {
 
 class kernel_runner : public kernel_selector::KernelRunnerInterface {
 public:
-    kernel_runner(engine_impl& engine_ref, uint32_t program_id, bool weights_and_bias_exist = false, bool zero_points_exist = false);
+    kernel_runner(engine& engine_ref, uint32_t program_id, bool weights_and_bias_exist = false, bool zero_points_exist = false);
 
     std::vector<std::chrono::nanoseconds> run_kernels(const kernel_selector::KernelsData& kernelsData) override;
 
@@ -25,20 +25,20 @@ class kernel_runner : public kernel_selector::KernelRunnerInterface {
     const int runs_per_kernel = 15;
 
     void prepare_kernel_args(const kernel_selector::KernelsData& kernels_data,
-                             gpu::kernel::kernel_arguments_data& args);
+                             kernel_arguments_data& args);
 
-    engine_impl::ptr engine;
+    engine& _engine;
     uint32_t program_id;
     bool weights_and_bias_exist;
     bool zero_points_exist;
-    std::vector<memory_impl::cptr> input_buffers;
-    std::vector<memory_impl::cptr> fused_ops_buffers;
-    std::vector<memory_impl::ptr> output_buffers;
-    std::vector<memory_impl::cptr> weight_buffers;
-    std::vector<memory_impl::cptr> bias_buffers;
-    std::vector<memory_impl::cptr> weight_zero_point_buffers;
-    std::vector<memory_impl::cptr> activation_zero_point_buffers;
-    std::vector<memory_impl::cptr> compensation_buffers;
+    std::vector<memory::cptr> input_buffers;
+    std::vector<memory::cptr> fused_ops_buffers;
+    std::vector<memory::ptr> output_buffers;
+    std::vector<memory::cptr> weight_buffers;
+    std::vector<memory::cptr> bias_buffers;
+    std::vector<memory::cptr> weight_zero_point_buffers;
+    std::vector<memory::cptr> activation_zero_point_buffers;
+    std::vector<memory::cptr> compensation_buffers;
 };
 
 //////////////////////////////////////////////////////////////////////////////////////////////////////////
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/loop_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/loop_gpu.cpp
index ea8965ec427c41..828fb6ca3d1b31 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/loop_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/loop_gpu.cpp
@@ -10,7 +10,6 @@
 #include "register_gpu.hpp"
 #include "mutable_data_inst.h"
 #include "input_layout_inst.h"
-#include "memory_impl.h"
 #include <vector>
 #include <algorithm>
 
@@ -18,32 +17,38 @@ namespace cldnn {
 namespace gpu {
 struct loop_gpu : typed_primitive_impl<loop> {
     const loop_node& node;
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<loop_gpu>(*this);
+    }
+
+    void init_kernels() override {}
 
+    loop_gpu(const loop_gpu& other) : typed_primitive_impl<loop>(other), node(other.node) {}
     explicit loop_gpu(const loop_node& node) : node(node) {}
 
     // read scala value from data primitive
-    static int64_t read_scalar_value(memory_impl& mem) {
+    static int64_t read_scalar_value(memory::ptr mem, stream& stream) {
         int64_t trip_count = 0;
-        const layout& prim_layout = mem.get_layout();
+        const layout& prim_layout = mem->get_layout();
 
         switch (prim_layout.data_type) {
         case data_types::u8: {
-            mem_lock<uint8_t> lock_prim_output{mem};
+            mem_lock<uint8_t> lock_prim_output{mem, stream};
             trip_count = *lock_prim_output.data();
             break;
         }
         case data_types::i8: {
-            mem_lock<int8_t> lock_prim_output{mem};
+            mem_lock<int8_t> lock_prim_output{mem, stream};
             trip_count = *lock_prim_output.data();
             break;
         }
         case data_types::i32: {
-            mem_lock<int32_t> lock_prim_output{mem};
+            mem_lock<int32_t> lock_prim_output{mem, stream};
             trip_count = *lock_prim_output.data();
             break;
         }
         case data_types::i64: {
-            mem_lock<int64_t> lock_prim_output{mem};
+            mem_lock<int64_t> lock_prim_output{mem, stream};
             trip_count = *lock_prim_output.data();
             break;
         }
@@ -53,33 +58,33 @@ struct loop_gpu : typed_primitive_impl<loop> {
         return trip_count;
     }
 
-    static void write_scalar_value(memory_impl& mem, int64_t input) {
-        const layout& prim_layout = mem.get_layout();
+    static void write_scalar_value(memory::ptr mem, stream& stream, int64_t input) {
+        const layout& prim_layout = mem->get_layout();
 
         switch (prim_layout.data_type) {
         case data_types::u8: {
             assert(input >= std::numeric_limits<uint8_t>::min() &&
                    input <= std::numeric_limits<uint8_t>::max());
-            mem_lock<uint8_t> lock_prim_output{mem};
+            mem_lock<uint8_t> lock_prim_output{mem, stream};
             *lock_prim_output.data() = static_cast<uint8_t>(input);
             break;
         }
         case data_types::i8: {
             assert(input >= std::numeric_limits<int8_t>::min() &&
                    input <= std::numeric_limits<int8_t>::max());
-            mem_lock<int8_t> lock_prim_output{mem};
+            mem_lock<int8_t> lock_prim_output{mem, stream};
             *lock_prim_output.data() = static_cast<int8_t>(input);
             break;
         }
         case data_types::i32: {
             assert(input >= std::numeric_limits<int32_t>::min() &&
                    input <= std::numeric_limits<int32_t>::max());
-            mem_lock<int32_t> lock_prim_output{mem};
+            mem_lock<int32_t> lock_prim_output{mem, stream};
             *lock_prim_output.data() = static_cast<int32_t>(input);
             break;
         }
         case data_types::i64: {
-            mem_lock<int64_t> lock_prim_output{mem};
+            mem_lock<int64_t> lock_prim_output{mem, stream};
             *lock_prim_output.data() = input;
             break;
         }
@@ -88,10 +93,9 @@ struct loop_gpu : typed_primitive_impl<loop> {
         }
     }
 
-    event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& events, loop_inst& instance) override {
+    event::ptr execute_impl(const std::vector<event::ptr>& events, loop_inst& instance) override {
         auto& outer_network = instance.get_network();
-        const uint32_t& net_id = instance.get_network().get_id();
-        auto ev = outer_network.get_engine().create_user_event(net_id, false);
+        auto& stream = outer_network.get_stream();
 
         auto body_network = instance.get_body_network();
 
@@ -104,8 +108,8 @@ struct loop_gpu : typed_primitive_impl<loop> {
 
         // read trip_count from outer network
         const primitive_id& trip_count_id = node.get_trip_count_id();
-        memory_impl& trip_count_mem = outer_network.get_primitive(trip_count_id)->output_memory();
-        int64_t trip_count = read_scalar_value(trip_count_mem);
+        memory::ptr trip_count_mem = outer_network.get_primitive(trip_count_id)->output_memory_ptr();
+        int64_t trip_count = read_scalar_value(trip_count_mem, stream);
         if (trip_count < 0) {
             const int64_t max_iteration = node.get_max_iteration();
             trip_count = max_iteration;
@@ -113,26 +117,26 @@ struct loop_gpu : typed_primitive_impl<loop> {
 
         // read initial execution condition from outer network
         const primitive_id& initial_execution_id = node.get_initial_execution_id();
-        memory_impl& initial_execution_mem = outer_network.get_primitive(initial_execution_id)->output_memory();
-        int64_t execution_condition = read_scalar_value(initial_execution_mem);
+        memory::ptr initial_execution_mem = outer_network.get_primitive(initial_execution_id)->output_memory_ptr();
+        int64_t execution_condition = read_scalar_value(initial_execution_mem, stream);
 
         // shortcut of current_iteration memory in body network (slice of input)
-        memory_impl* current_iteration_mem = nullptr;
+        memory::ptr current_iteration_mem = nullptr;
         if (node.is_current_iteration_used()) {
             const primitive_id& current_iteration_id = node.get_current_iteration_id();
-            current_iteration_mem = &body_network->get_primitive(current_iteration_id)->output_memory();
+            current_iteration_mem = body_network->get_primitive(current_iteration_id)->output_memory_ptr();
         }
 
         // shortcut of execution_condition memory in body network
-        memory_impl* execution_condition_mem = nullptr;
+        memory::ptr execution_condition_mem = nullptr;
         if (node.is_execution_condition_used()) {
             const primitive_id& condition_id = node.get_condition_id();
-            execution_condition_mem = &body_network->get_primitive(condition_id)->output_memory();
+            execution_condition_mem = body_network->get_primitive(condition_id)->output_memory_ptr();
         }
 
         int64_t current_iteration = 0;
         if (node.is_current_iteration_used()) {
-            write_scalar_value(*current_iteration_mem, current_iteration);
+            write_scalar_value(current_iteration_mem, stream, current_iteration);
         }
 
         const auto& concatenated_input_mem_mappings = instance.concatenated_input_mem_mappings;
@@ -141,23 +145,23 @@ struct loop_gpu : typed_primitive_impl<loop> {
         // Set sliced input data
         for (size_t i = 0; i < concatenated_input_mem_mappings.size(); ++i) {
             const auto& concatenated_input = concatenated_input_mem_mappings.at(i);
-            memory_impl::ptr mem = concatenated_input.get_sliced_mem(0);
+            memory::ptr mem = concatenated_input.get_sliced_mem(0);
             if (mem) {
-                body_network->set_input_data(concatenated_input.sliced_data_prim->id(), *mem);
+                body_network->set_input_data(concatenated_input.sliced_data_prim->id(), mem);
             } else {
                 CLDNN_ERROR_MESSAGE(node.id(), "sliced input memory of loop is not allocated properly");
             }
         }
 
-        std::vector<event_impl::ptr> loop_carried_dep(events.begin(), events.end());
+        std::vector<event::ptr> loop_carried_dep(events.begin(), events.end());
 
         while (current_iteration < trip_count && execution_condition) {
             // Copy & Set sliced input memory
             for (size_t i = 0; i < concatenated_input_mem_mappings.size(); ++i) {
                 const auto& concatenated_input = concatenated_input_mem_mappings.at(i);
-                memory_impl::ptr mem = concatenated_input.get_sliced_mem(current_iteration);
+                memory::ptr mem = concatenated_input.get_sliced_mem(current_iteration);
                 if (mem) {
-                    concatenated_input.sliced_data_prim->set_output_memory(*mem);
+                    concatenated_input.sliced_data_prim->set_output_memory(mem);
                 } else {
                     CLDNN_ERROR_MESSAGE(node.id(), "sliced input memory of loop is not allocated properly");
                 }
@@ -178,7 +182,7 @@ struct loop_gpu : typed_primitive_impl<loop> {
 
             loop_carried_dep.clear();
             for (const auto& backedge : node.get_back_edges()) {
-                event_impl::ptr body_event = body_network->get_primitive_event(backedge.from);
+                event::ptr body_event = body_network->get_primitive_event(backedge.from);
                 loop_carried_dep.emplace_back(body_event);
             }
 
@@ -186,10 +190,10 @@ struct loop_gpu : typed_primitive_impl<loop> {
             //as they are presented in the ngraph opset document for loop operation.
             //However they are not being used yet and only TensorIterator which has fixed sequence length is being validated.
             if (node.is_current_iteration_used()) {
-                write_scalar_value(*current_iteration_mem, current_iteration);
+                write_scalar_value(current_iteration_mem, stream, current_iteration);
             }
             if (node.is_execution_condition_used()) {
-                execution_condition = read_scalar_value(*execution_condition_mem);
+                execution_condition = read_scalar_value(execution_condition_mem, stream);
             }
             // update index & execution condition for the next iteration
             ++current_iteration;
@@ -204,11 +208,10 @@ struct loop_gpu : typed_primitive_impl<loop> {
         }
 
         const primitive_id& num_iteration_id = node.get_num_iteration_id();
-        memory_impl& num_actual_iterations_mem = outer_network.get_primitive(num_iteration_id)->output_memory();
-        write_scalar_value(num_actual_iterations_mem, current_iteration);
+        memory::ptr num_actual_iterations_mem = outer_network.get_primitive(num_iteration_id)->output_memory_ptr();
+        write_scalar_value(num_actual_iterations_mem, stream, current_iteration);
 
-        dynamic_cast<cldnn::user_event*>(ev.get())->set();
-        return ev;
+        return stream.create_user_event(true);
     }
 
     static primitive_impl* create(const loop_node& arg) { return new loop_gpu(arg); }
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/lrn_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/lrn_gpu.cpp
index c12c07aad6753d..b829688a736e8d 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/lrn_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/lrn_gpu.cpp
@@ -5,7 +5,7 @@
 #include "lrn_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "lrn/lrn_kernel_selector.h"
 #include "lrn/lrn_kernel_base.h"
@@ -17,6 +17,10 @@ struct lrn_gpu : typed_primitive_gpu_impl<lrn> {
     using parent = typed_primitive_gpu_impl<lrn>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<lrn_gpu>(*this);
+    }
+
     static primitive_impl* create(const lrn_node& arg) {
         auto lrn_params = get_default_params<kernel_selector::lrn_params>(arg);
         auto lrn_optional_params = get_default_optional_params<kernel_selector::lrn_optional_params>(arg.get_program());
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/lstm_dynamic_input_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/lstm_dynamic_input_gpu.cpp
index 30640edd24e06a..4a8a4ee5d75b94 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/lstm_dynamic_input_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/lstm_dynamic_input_gpu.cpp
@@ -11,7 +11,7 @@
 #include "lstm_dynamic/lstm_dynamic_input_kernel_selector.h"
 #include "lstm_dynamic/lstm_dynamic_input_kernel_base.h"
 #include "network_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -20,14 +20,17 @@ struct lstm_dynamic_input_gpu : typed_primitive_gpu_impl<lstm_dynamic_input> {
     using parent = typed_primitive_gpu_impl<lstm_dynamic_input>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<lstm_dynamic_input_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<lstm_dynamic_input>& instance,
-                                                        int32_t) const override {
-        kernel::kernel_arguments_data args;
-        args.inputs = { (memory_impl::cptr) &instance.input_memory(), (memory_impl::cptr) &instance.dyn_length_memory()};
-        args.output = (memory_impl::cptr) &instance.output_memory();
-        args.weights = (memory_impl::cptr) &instance.weights_memory();
-        args.bias = (memory_impl::cptr) (instance.bias_term() ? &instance.bias_memory() : nullptr);
+    kernel_arguments_data get_arguments(typed_primitive_inst<lstm_dynamic_input>& instance, int32_t) const override {
+        kernel_arguments_data args;
+        args.inputs = { instance.input_memory_ptr(), instance.dyn_length_memory()};
+        args.output = instance.output_memory_ptr();
+        args.weights = instance.weights_memory();
+        args.bias = instance.bias_term() ? instance.bias_memory() : nullptr;
         return args;
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/lstm_dynamic_timeloop_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/lstm_dynamic_timeloop_gpu.cpp
index bbacc878735fa9..9c367d5d994a5d 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/lstm_dynamic_timeloop_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/lstm_dynamic_timeloop_gpu.cpp
@@ -11,7 +11,7 @@
 #include "lstm_dynamic/lstm_dynamic_timeloop_kernel_selector.h"
 #include "lstm_dynamic/lstm_dynamic_timeloop_kernel_base.h"
 #include "network_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -20,19 +20,22 @@ struct lstm_dynamic_timeloop_gpu : typed_primitive_gpu_impl<lstm_dynamic_timeloo
     using parent = typed_primitive_gpu_impl<lstm_dynamic_timeloop>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<lstm_dynamic_timeloop_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<lstm_dynamic_timeloop>& instance,
-                                                        int32_t) const override {
-        kernel::kernel_arguments_data args;
-        args.inputs = {(memory_impl::cptr) &instance.input_memory(), (memory_impl::cptr) &instance.dyn_length_memory()};
+    kernel_arguments_data get_arguments(typed_primitive_inst<lstm_dynamic_timeloop>& instance, int32_t) const override {
+        kernel_arguments_data args;
+        args.inputs = {instance.input_memory_ptr(), instance.dyn_length_memory()};
         if (instance.last_hidden_output_term())
-            args.inputs.push_back((memory_impl::cptr) &instance.last_hidden_output_memory());
+            args.inputs.push_back(instance.last_hidden_output_memory());
         if (instance.last_cell_output_term())
-            args.inputs.push_back((memory_impl::cptr) &instance.last_cell_output_memory());
-        args.output = (memory_impl::cptr) &instance.output_memory();
-        args.recurrent = (memory_impl::cptr) &instance.recurrent_memory();
-        args.hidden = (memory_impl::cptr) (instance.initial_hidden_term() ? &instance.initial_hidden_memory() : nullptr);
-        args.cell = (memory_impl::cptr) (instance.initial_cell_term() ? &instance.initial_cell_memory() : nullptr);
+            args.inputs.push_back(instance.last_cell_output_memory());
+        args.output = instance.output_memory_ptr();
+        args.recurrent = instance.recurrent_memory();
+        args.hidden = instance.initial_hidden_term() ? instance.initial_hidden_memory() : nullptr;
+        args.cell = instance.initial_cell_term() ? instance.initial_cell_memory() : nullptr;
         return args;
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/lstm_elt_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/lstm_elt_gpu.cpp
index b4a45bfe41500c..ed558133306753 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/lstm_elt_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/lstm_elt_gpu.cpp
@@ -11,7 +11,7 @@
 #include "lstm/lstm_elt_kernel_selector.h"
 #include "lstm/lstm_elt_kernel_base.h"
 #include "network_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -20,13 +20,16 @@ struct lstm_elt_gpu : typed_primitive_gpu_impl<lstm_elt> {
     using parent = typed_primitive_gpu_impl<lstm_elt>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<lstm_elt_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<lstm_elt>& instance,
-                                                        int32_t) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, 0);
+    kernel_arguments_data get_arguments(typed_primitive_inst<lstm_elt>& instance, int32_t) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, 0);
 
-        args.cell = (memory_impl::cptr) (instance.cell_term() ? &instance.cell_memory() : nullptr);
-        args.output = (memory_impl::cptr) &instance.output_memory();
+        args.cell = instance.cell_term() ? instance.cell_memory() : nullptr;
+        args.output = instance.output_memory_ptr();
 
         return args;
     }
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/lstm_gemm_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/lstm_gemm_gpu.cpp
index 2b27aee688ee99..f6c532decd08c6 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/lstm_gemm_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/lstm_gemm_gpu.cpp
@@ -11,7 +11,7 @@
 #include "lstm/lstm_gemm_kernel_selector.h"
 #include "lstm/lstm_gemm_kernel_base.h"
 #include "network_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -20,16 +20,19 @@ struct lstm_gemm_gpu : typed_primitive_gpu_impl<lstm_gemm> {
     using parent = typed_primitive_gpu_impl<lstm_gemm>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<lstm_gemm_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<lstm_gemm>& instance,
-                                                        int32_t) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, 0);
-
-        args.output = (memory_impl::cptr) &instance.output_memory();
-        args.weights = (memory_impl::cptr) &instance.weights_memory();
-        args.recurrent = (memory_impl::cptr) &instance.recurrent_memory();
-        args.bias = (memory_impl::cptr) (instance.bias_term() ? &instance.bias_memory() : nullptr);
-        args.hidden = (memory_impl::cptr) (instance.hidden_term() ? &instance.hidden_memory() : nullptr);
+    kernel_arguments_data get_arguments(typed_primitive_inst<lstm_gemm>& instance, int32_t) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, 0);
+
+        args.output = instance.output_memory_ptr();
+        args.weights = instance.weights_memory();
+        args.recurrent = instance.recurrent_memory();
+        args.bias = instance.bias_term() ? instance.bias_memory() : nullptr;
+        args.hidden = instance.hidden_term() ? instance.hidden_memory() : nullptr;
 
         return args;
     }
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/max_unpooling_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/max_unpooling_gpu.cpp
index 7e903fc65b0fdd..2daaaa0ddc2579 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/max_unpooling_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/max_unpooling_gpu.cpp
@@ -5,7 +5,7 @@
 #include "max_unpooling_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "network_impl.h"
 #include "kernel_selector_helper.h"
 #include "max_unpooling/max_unpooling_kernel_selector.h"
@@ -19,20 +19,23 @@ struct max_unpooling_gpu : typed_primitive_gpu_impl<max_unpooling> {
     using parent = typed_primitive_gpu_impl<max_unpooling>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<max_unpooling_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<max_unpooling>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
-        args.inputs.push_back((memory_impl::cptr) &instance.dep_memory(1));
+    kernel_arguments_data get_arguments(typed_primitive_inst<max_unpooling>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
+        args.inputs.push_back(instance.dep_memory_ptr(1));
         return args;
     }
 
 public:
-    event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& events, max_unpooling_inst& instance) override {
+    event::ptr execute_impl(const std::vector<event::ptr>& events, max_unpooling_inst& instance) override {
         // clear output buffer
-        std::vector<event_impl::ptr> tmp_events(events);
-        auto ev = instance.get_network().get_engine().create_user_event(instance.get_network().get_id(), false);
-        instance.output_memory().fill(0, ev);
+        std::vector<event::ptr> tmp_events(events);
+        auto& stream = instance.get_network().get_stream();
+        auto ev = instance.output_memory().fill(stream);
         tmp_events.push_back(ev);
         return parent::execute_impl(tmp_events, instance);
     }
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/memory_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/memory_gpu.cpp
deleted file mode 100644
index 5a839905ac38c3..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/memory_gpu.cpp
+++ /dev/null
@@ -1,310 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "error_handler.h"
-#include "memory_gpu.h"
-#include "engine_impl.h"
-#include "ocl_base_event.h"
-#include <stdexcept>
-#include <vector>
-
-namespace cldnn {
-namespace gpu {
-
-gpu_buffer::gpu_buffer(const refcounted_obj_ptr<engine_impl>& engine,
-                       const layout& layout,
-                       uint32_t net_id,
-                       bool reset)
-    : lockable_gpu_mem(engine), memory_impl(engine, layout, net_id, allocation_type::cl_mem, false),
-      _buffer(_context->context(), CL_MEM_READ_WRITE, size()) {
-    if (reset || is_memory_reset_needed(_layout)) zero_buffer();
-}
-
-gpu_buffer::gpu_buffer(const refcounted_obj_ptr<engine_impl>& engine,
-                       const layout& new_layout,
-                       const cl::Buffer& buffer,
-                       uint32_t net_id)
-    : lockable_gpu_mem(engine), memory_impl(engine, new_layout, net_id, allocation_type::cl_mem, true),
-      _buffer(buffer) {}
-
-void* gpu_buffer::lock() {
-    std::lock_guard<std::mutex> locker(_mutex);
-    if (0 == _lock_count) {
-        _mapped_ptr = _context->queue(_net_id).enqueueMapBuffer(_buffer, CL_TRUE, CL_MAP_WRITE, 0, size());
-    }
-    _lock_count++;
-    return _mapped_ptr;
-}
-
-void gpu_buffer::unlock() {
-    std::lock_guard<std::mutex> locker(_mutex);
-    _lock_count--;
-    if (0 == _lock_count) {
-        _context->queue(_net_id).enqueueUnmapMemObject(_buffer, _mapped_ptr);
-        _mapped_ptr = nullptr;
-    }
-}
-
-void gpu_buffer::zero_buffer() {
-    _context->queue(_net_id).enqueueFillBuffer<unsigned char>(_buffer, 0, 0, size());
-    _context->queue(_net_id).flush();
-}
-
-void gpu_buffer::fill(unsigned char pattern, event_impl::ptr ev) {
-    cl::Event ev_ocl = dynamic_cast<base_event*>(ev.get())->get();
-    _context->queue(_net_id).enqueueFillBuffer<unsigned char>(_buffer, pattern, 0, size(), 0, &ev_ocl);
-}
-
-shared_mem_params gpu_buffer::get_internal_params() const {
-    return {shared_mem_type::shared_mem_buffer, static_cast<shared_handle>(_context->context().get()), nullptr,
-            static_cast<shared_handle>(_buffer.get()),
-#ifdef _WIN32
-        nullptr,
-#else
-        0,
-#endif
-        0};
-}
-
-gpu_image2d::gpu_image2d(const refcounted_obj_ptr<engine_impl>& engine, const layout& layout, uint32_t net_id,
-                         bool reset)
-    : lockable_gpu_mem(engine), memory_impl(engine, layout, net_id, allocation_type::cl_mem, false),
-    _row_pitch(0), _slice_pitch(0) {
-    cl_channel_type type = layout.data_type == data_types::f16 ? CL_HALF_FLOAT : CL_FLOAT;
-    cl_channel_order order = CL_R;
-    switch (layout.format) {
-        case format::image_2d_weights_c1_b_fyx:
-            _width = layout.size.batch[0];
-            _height = layout.size.spatial[0] * layout.size.feature[0] * layout.size.spatial[1];
-            break;
-        case format::image_2d_weights_winograd_6x3_s1_fbxyb:
-            _height = layout.size.feature[0];
-            _width = layout.size.spatial[0] * layout.size.batch[0] * layout.size.spatial[1] * 8 / 3;
-            break;
-        case format::image_2d_weights_winograd_6x3_s1_xfbyb:
-            _height = layout.size.feature[0] * layout.size.spatial[0] * 8 / 3;
-            _width = layout.size.batch[0] * layout.size.spatial[1];
-            break;
-        case format::image_2d_weights_c4_fyx_b:
-            _width = layout.size.batch[0];
-            _height = layout.size.spatial[0] * layout.size.feature[0] * layout.size.spatial[1];
-            order = CL_RGBA;
-            break;
-        case format::image_2d_rgba:
-            _width = layout.size.spatial[0];
-            _height = layout.size.spatial[1];
-            order = CL_RGBA;
-            if (layout.size.feature[0] != 3 && layout.size.feature[0] != 4) {
-                CLDNN_ERROR_MESSAGE("2D image allocation", "invalid number of channels in image_2d_rgba input image (should be 3 or 4)!");
-            }
-            type = CL_UNORM_INT8;
-            break;
-        case format::nv12:
-            _width = layout.size.spatial[1];
-            _height = layout.size.spatial[0];
-            if (layout.size.feature[0] == 2) {
-                order = CL_RG;
-            } else if (layout.size.feature[0] > 2) {
-                CLDNN_ERROR_MESSAGE("2D image allocation", "invalid number of channels in NV12 input image!");
-            }
-            type = CL_UNORM_INT8;
-            break;
-        default:
-            CLDNN_ERROR_MESSAGE("2D image allocation", "unsupported image type!");
-    }
-
-    cl::ImageFormat imageFormat(order, type);
-    _buffer = cl::Image2D(_context->context(), CL_MEM_READ_WRITE, imageFormat, _width, _height, 0);
-
-    if (reset) zero_image();
-}
-
-gpu_image2d::gpu_image2d(const refcounted_obj_ptr<engine_impl>& engine,
-                         const layout& new_layout,
-                         const cl::Image2D& buffer,
-                         uint32_t net_id)
-    : lockable_gpu_mem(engine), memory_impl(engine, new_layout, net_id, allocation_type::cl_mem, true),
-      _buffer(buffer) {
-    _width = _buffer.getImageInfo<CL_IMAGE_WIDTH>();
-    _height = _buffer.getImageInfo<CL_IMAGE_HEIGHT>();
-    _row_pitch = _buffer.getImageInfo<CL_IMAGE_ROW_PITCH>();
-    _slice_pitch = _buffer.getImageInfo<CL_IMAGE_SLICE_PITCH>();
-}
-
-void gpu_image2d::zero_image() {
-    cl_uint4 pattern_uint4 = { 0, 0, 0, 0 };
-    _context->queue(_net_id).enqueueFillImage(_buffer, pattern_uint4, { 0, 0, 0 }, { _width, _height, 1 });
-    _context->queue(_net_id).flush();
-}
-
-void* gpu_image2d::lock() {
-    std::lock_guard<std::mutex> locker(_mutex);
-    if (0 == _lock_count) {
-        _mapped_ptr = _context->queue(_net_id)
-                          .enqueueMapImage(_buffer,
-                                           CL_TRUE,
-                                           CL_MAP_WRITE,
-                                           {0, 0, 0},
-                                           {_width, _height, 1},
-                                           &_row_pitch,
-                                           &_slice_pitch);
-    }
-    _lock_count++;
-    return _mapped_ptr;
-}
-
-void gpu_image2d::unlock() {
-    std::lock_guard<std::mutex> locker(_mutex);
-    _lock_count--;
-    if (0 == _lock_count) {
-        _context->queue(_net_id).enqueueUnmapMemObject(_buffer, _mapped_ptr);
-        _mapped_ptr = nullptr;
-    }
-}
-
-void gpu_image2d::fill(unsigned char pattern, event_impl::ptr ev) {
-    cl::Event ev_ocl = dynamic_cast<base_event*>(ev.get())->get();
-    cl_uint4 pattern_uint4 = {pattern, pattern, pattern, pattern};
-    _context->queue(_net_id).enqueueFillImage(_buffer, pattern_uint4, {0, 0, 0}, {_width, _height, 1}, 0, &ev_ocl);
-}
-
-shared_mem_params gpu_image2d::get_internal_params() const {
-    return {shared_mem_type::shared_mem_image, static_cast<shared_handle>(_context->context().get()), nullptr,
-            static_cast<shared_handle>(_buffer.get()),
-#ifdef _WIN32
-        nullptr,
-#else
-        0,
-#endif
-        0};
-}
-
-gpu_media_buffer::gpu_media_buffer(const refcounted_obj_ptr<engine_impl>& engine,
-    const layout& new_layout,
-    const shared_mem_params* params,
-    uint32_t net_id)
-    : gpu_image2d(engine, new_layout,
-        cl::ImageVA(engine->get_context()->context(), CL_MEM_READ_WRITE,
-                    params->surface, params->plane),
-        net_id),
-    device(params->user_device),
-    surface(params->surface),
-    plane(params->plane) {
-}
-
-shared_mem_params gpu_media_buffer::get_internal_params() const {
-    return {shared_mem_type::shared_mem_vasurface, static_cast<shared_handle>(_context->context().get()), device,
-            static_cast<shared_handle>(_buffer.get()), surface, plane };
-}
-
-#ifdef _WIN32
-gpu_dx_buffer::gpu_dx_buffer(const refcounted_obj_ptr<engine_impl>& engine,
-    const layout& new_layout,
-    const shared_mem_params* params,
-    uint32_t net_id)
-    : gpu_buffer(engine, new_layout,
-                cl::BufferDX(engine->get_context()->context(), CL_MEM_READ_WRITE, params->mem),
-                net_id),
-    device(params->user_device),
-    resource(params->mem) { }
-
-shared_mem_params gpu_dx_buffer::get_internal_params() const {
-    return {shared_mem_type::shared_mem_dxbuffer, static_cast<shared_handle>(_context->context().get()), device,
-            static_cast<shared_handle>(_buffer.get()), resource, 0 };
-}
-#endif
-
-gpu_usm::gpu_usm(const refcounted_obj_ptr<engine_impl>& engine,
-    const layout& new_layout, const cl::UsmMemory& buffer,
-    allocation_type type, uint32_t net_id)
-    : lockable_gpu_mem(engine)
-    , memory_impl(engine, new_layout, net_id, type, true)
-    , _buffer(buffer) {
-}
-
-gpu_usm::gpu_usm(const refcounted_obj_ptr<engine_impl>& engine, const layout& layout, uint32_t net_id, allocation_type type, bool reset)
-    : lockable_gpu_mem(engine)
-    , memory_impl(engine, layout, net_id, type, false)
-    , _buffer(_engine->get_context()->context()) {
-    auto device = _engine->get_context()->device();
-    switch (get_allocation_type()) {
-    case allocation_type::usm_host:
-        _buffer.allocateHost(_bytes_count);
-        break;
-    case allocation_type::usm_shared:
-        _buffer.allocateShared(device, _bytes_count);
-        break;
-    case allocation_type::usm_device:
-        _buffer.allocateDevice(device, _bytes_count);
-        break;
-    default:
-        CLDNN_ERROR_MESSAGE("gpu_usm allocation type",
-            "Unknown unified shared memory type!");
-    }
-
-    if (reset || is_memory_reset_needed(_layout)) zero_buffer();
-}
-
-void* gpu_usm::lock() {
-    assert(get_allocation_type() != allocation_type::usm_device && "Can't lock usm device memory!");
-    std::lock_guard<std::mutex> locker(_mutex);
-    if (0 == _lock_count) {
-        _engine->get_context()->queue(_net_id).finish();  // Synchronization needed for OOOQ.
-        _mapped_ptr = _buffer.get();
-    }
-    _lock_count++;
-    return _mapped_ptr;
-}
-
-void gpu_usm::unlock() {
-    std::lock_guard<std::mutex> locker(_mutex);
-    _lock_count--;
-    if (0 == _lock_count) {
-        _mapped_ptr = nullptr;
-    }
-}
-
-void gpu_usm::fill(unsigned char pattern, event_impl::ptr ev) {
-    cl::Event ev_ocl = dynamic_cast<base_event*>(ev.get())->get();
-    // enqueueFillUsm call will never finish. Driver bug? Uncomment when fixed. Some older drivers doesn't support enqueueFillUsm call at all.
-    // _engine->get_context()->queue(_net_id).enqueueFillUsm<unsigned char>(_buffer, pattern, _bytes_count, nullptr, &ev_ocl)
-    // Workarounded with enqeue_memcopy. ToDo: Remove below code. Uncomment above.
-    std::vector<unsigned char> temp_buffer(_bytes_count, pattern);
-    cl::usm::enqueue_memcpy(_engine->get_context()->queue(_net_id), _buffer.get(), temp_buffer.data(), _bytes_count, true, nullptr, &ev_ocl);
-}
-
-void gpu_usm::zero_buffer() {
-    // event_impl::ptr ev{ new base_event(_engine->get_context()), false };
-    // cl::Event ev_ocl = dynamic_cast<base_event*>(ev.get())->get();
-    // cl::usm::enqueue_set_mem(_engine->get_context()->queue(_net_id), _buffer.get(), 0, _bytes_count, nullptr, &ev_ocl);
-    // ev->wait();
-
-    // [WA]
-    event_impl::ptr ev{ new base_event(_engine->get_context()), false };
-    fill(0, ev);
-    ev->wait();
-}
-
-void gpu_usm::copy_from_other(const gpu_usm& other) {
-    _engine->get_context()->queue(_net_id).enqueueCopyUsm(other.get_buffer(), get_buffer(), _bytes_count, true);
-}
-
-shared_mem_params gpu_usm::get_internal_params() const {
-    return {
-        shared_mem_type::shared_mem_empty,  // shared_mem_type
-        static_cast<shared_handle>(_engine->get_context()->context().get()),  // context handle
-        nullptr,  // user_device handle
-        nullptr,  // mem handle
-#ifdef _WIN32
-        nullptr,  // surface handle
-#else
-        0,  // surface handle
-#endif
-        0  // plane
-    };
-}
-
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/memory_gpu.h b/inference-engine/thirdparty/clDNN/src/gpu/memory_gpu.h
deleted file mode 100644
index e1ac301f997c55..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/memory_gpu.h
+++ /dev/null
@@ -1,176 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include "ocl_toolkit.h"
-#include "memory_impl.h"
-#include <cassert>
-#include <iterator>
-#include <mutex>
-#include <memory>
-
-#define BUFFER_ALIGNMENT 4096
-#define CACHE_ALIGNMENT 64
-
-namespace cldnn {
-namespace gpu {
-
-template <typename T>
-T* allocate_aligned(size_t size, size_t align) {
-    assert(sizeof(T) <= size);
-    assert(alignof(T) <= align);
-    return reinterpret_cast<T*>(_mm_malloc(align_to(size, align), align));
-}
-
-template <typename T>
-void deallocate_aligned(T* ptr) {
-    _mm_free(ptr);
-}
-
-#if defined(_SECURE_SCL) && (_SECURE_SCL > 0)
-template <typename T>
-stdext::checked_array_iterator<T*> arr_begin(T* buf, size_t count) {
-    return stdext::make_checked_array_iterator(buf, count);
-}
-
-template <typename T>
-stdext::checked_array_iterator<T*> arr_end(T* buf, size_t count) {
-    return stdext::make_checked_array_iterator(buf, count, count);
-}
-
-#else
-template <typename T>
-T* arr_begin(T* buf, size_t) {
-    return buf;
-}
-
-template <typename T>
-T* arr_end(T* buf, size_t count) {
-    return buf + count;
-}
-#endif
-
-struct lockable_gpu_mem {
-    explicit lockable_gpu_mem(const refcounted_obj_ptr<engine_impl>& engine) : _context(engine->get_context()),
-        _lock_count(0),
-        _mapped_ptr(nullptr) {}
-
-    std::shared_ptr<gpu_toolkit> _context;
-    std::mutex _mutex;
-    unsigned _lock_count;
-    void* _mapped_ptr;
-};
-
-struct gpu_buffer : public lockable_gpu_mem, public memory_impl {
-    friend cldnn::memory_pool;
-
-    gpu_buffer(const refcounted_obj_ptr<engine_impl>& engine,
-               const layout& new_layout,
-               const cl::Buffer& buffer,
-               uint32_t net_id);
-
-    void* lock() override;
-    void unlock() override;
-    void fill(unsigned char pattern, event_impl::ptr ev) override;
-    shared_mem_params get_internal_params() const override;
-    const cl::Buffer& get_buffer() const {
-        assert(0 == _lock_count);
-        return _buffer;
-    }
-
-    void zero_buffer();
-
-protected:
-    gpu_buffer(const refcounted_obj_ptr<engine_impl>& engine, const layout& layout, uint32_t net_id,
-               bool reset = true);
-    cl::Buffer _buffer;
-};
-
-struct gpu_image2d : public lockable_gpu_mem, public memory_impl {
-    friend cldnn::memory_pool;
-
-    gpu_image2d(const refcounted_obj_ptr<engine_impl>& engine,
-                const layout& new_layout,
-                const cl::Image2D& buffer,
-                uint32_t net_id);
-    void* lock() override;
-    void unlock() override;
-    void fill(unsigned char pattern, event_impl::ptr ev) override;
-    shared_mem_params get_internal_params() const override;
-    const cl::Image2D& get_buffer() const {
-        assert(0 == _lock_count);
-        return _buffer;
-    }
-
-    void zero_image();
-
-protected:
-    gpu_image2d(const refcounted_obj_ptr<engine_impl>& engine, const layout& layout, uint32_t net_id,
-                bool reset = true);
-
-    cl::Image2D _buffer;
-    size_t _width;
-    size_t _height;
-    size_t _row_pitch;
-    size_t _slice_pitch;
-};
-
-struct gpu_media_buffer : public gpu_image2d {
-    friend cldnn::memory_pool;
-
-    gpu_media_buffer(const refcounted_obj_ptr<engine_impl>& engine,
-        const layout& new_layout,
-        const shared_mem_params* params,
-        uint32_t net_id);
-    shared_mem_params get_internal_params() const override;
-private:
-    void* device;
-#ifdef _WIN32
-    void* surface;
-#else
-    uint32_t surface;
-#endif
-    uint32_t plane;
-};
-
-#ifdef _WIN32
-struct gpu_dx_buffer : public gpu_buffer {
-    friend cldnn::memory_pool;
-
-    gpu_dx_buffer(const refcounted_obj_ptr<engine_impl>& engine,
-        const layout& new_layout,
-        const shared_mem_params* params,
-        uint32_t net_id);
-    shared_mem_params get_internal_params() const override;
-private:
-    void* device;
-    void* resource;
-};
-#endif
-
-struct gpu_usm : public lockable_gpu_mem, public memory_impl {
-    friend cldnn::memory_pool;
-
-    gpu_usm(const refcounted_obj_ptr<engine_impl>& engine,
-        const layout& new_layout,
-        const cl::UsmMemory& usm_buffer,
-        allocation_type type,
-        uint32_t net_id);
-
-    void* lock() override;
-    void unlock() override;
-    const cl::UsmMemory& get_buffer() const { return _buffer; }
-    cl::UsmMemory& get_buffer() { return _buffer; }
-
-    void fill(unsigned char pattern, event_impl::ptr ev) override;
-    void zero_buffer();
-    void copy_from_other(const gpu_usm& other);
-    shared_mem_params get_internal_params() const override;
-protected:
-    gpu_usm(const refcounted_obj_ptr<engine_impl>& engine, const layout& layout, uint32_t net_id, allocation_type type, bool reset = true);
-    cl::UsmMemory _buffer;
-};
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/mutable_data_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/mutable_data_gpu.cpp
index 0b2285af85915d..24a716b5dbe0b2 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/mutable_data_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/mutable_data_gpu.cpp
@@ -13,17 +13,11 @@ struct mutable_data_gpu : public typed_primitive_gpu_impl<mutable_data> {
     using parent = typed_primitive_gpu_impl<mutable_data>;
     using parent::parent;
 
-public:
-    bool validate_impl(const typed_primitive_inst<mutable_data>& instance) const override {
-        bool is_primary = instance.get_network().is_primary_stream();
-
-        auto net_id = instance.get_network().get_id();
-        auto mem_net_id = instance.output_memory().get_net_id();
-
-        bool res = is_primary || net_id == mem_net_id;
-        return res;
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<mutable_data_gpu>(*this);
     }
 
+public:
     static primitive_impl* create(mutable_data_node const& arg) { return new mutable_data_gpu(arg, {}); }
 };
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/mvn_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/mvn_gpu.cpp
index d2b839e7a9de6a..fc06edd6312db5 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/mvn_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/mvn_gpu.cpp
@@ -5,7 +5,7 @@
 #include "mvn_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "mvn/mvn_kernel_selector.h"
 #include "mvn/mvn_kernel_base.h"
@@ -21,6 +21,10 @@ struct mvn_gpu : typed_primitive_gpu_impl<mvn> {
     using parent = typed_primitive_gpu_impl<mvn>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<mvn_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const mvn_node& arg) {
         auto mvn_params = get_default_params<kernel_selector::mvn_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/non_max_suppression_cpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/non_max_suppression_cpu.cpp
index bcb8f006c3a565..55004113150bd6 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/non_max_suppression_cpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/non_max_suppression_cpu.cpp
@@ -110,14 +110,14 @@ std::vector<result_indices> run_nms(
 }
 
 template <typename T>
-vector2D<bounding_box> load_boxes_impl(memory_impl& mem, bool center_point) {
+vector2D<bounding_box> load_boxes_impl(stream& stream, memory::ptr mem, bool center_point) {
     vector2D<bounding_box> result;
-    auto lay = mem.get_layout();
+    auto lay = mem->get_layout();
     auto batch_size = lay.size.batch[0];
     auto boxes_num = lay.size.feature[0];
     result.resize(batch_size);
 
-    mem_lock<T> boxes_lock(mem);
+    mem_lock<T> boxes_lock(mem, stream);
     auto ptr = boxes_lock.data();
 
     for (int bi = 0; bi < batch_size; ++bi) {
@@ -145,28 +145,28 @@ vector2D<bounding_box> load_boxes_impl(memory_impl& mem, bool center_point) {
     return result;
 }
 
-vector2D<bounding_box> load_boxes(memory_impl& mem, bool center_point) {
-    auto data_type = mem.get_layout().data_type;
+vector2D<bounding_box> load_boxes(stream& stream, memory::ptr mem, bool center_point) {
+    auto data_type = mem->get_layout().data_type;
     switch (data_type) {
     case cldnn::data_types::f16:
-        return load_boxes_impl<data_type_to_type<data_types::f16>::type>(mem, center_point);
+        return load_boxes_impl<data_type_to_type<data_types::f16>::type>(stream, mem, center_point);
     case cldnn::data_types::f32:
-        return load_boxes_impl<data_type_to_type<data_types::f32>::type>(mem, center_point);
+        return load_boxes_impl<data_type_to_type<data_types::f32>::type>(stream, mem, center_point);
     default:
         throw std::runtime_error("Non max supression - unsupported boxes data type");
     }
 }
 
 template <typename T>
-vector3D<float> load_scores_impl(memory_impl& mem) {
-    auto lay = mem.get_layout();
+vector3D<float> load_scores_impl(stream& stream, memory::ptr mem) {
+    auto lay = mem->get_layout();
     auto batch_size = lay.size.batch[0];
     auto classes_num = lay.size.feature[0];
     auto boxes_num = lay.size.spatial[1];
 
     vector3D<float> result(batch_size, vector2D<float>(classes_num));
 
-    mem_lock<T> lock(mem);
+    mem_lock<T> lock(mem, stream);
     auto ptr = lock.data();
 
     for (int bi = 0; bi < batch_size; ++bi) {
@@ -182,47 +182,47 @@ vector3D<float> load_scores_impl(memory_impl& mem) {
     return result;
 }
 
-vector3D<float> load_scores(memory_impl& mem) {
-    auto data_type = mem.get_layout().data_type;
+vector3D<float> load_scores(stream& stream, memory::ptr mem) {
+    auto data_type = mem->get_layout().data_type;
     switch (data_type) {
     case cldnn::data_types::f16:
-        return load_scores_impl<data_type_to_type<data_types::f16>::type>(mem);
+        return load_scores_impl<data_type_to_type<data_types::f16>::type>(stream, mem);
     case cldnn::data_types::f32:
-        return load_scores_impl<data_type_to_type<data_types::f32>::type>(mem);
+        return load_scores_impl<data_type_to_type<data_types::f32>::type>(stream, mem);
     default:
         throw std::runtime_error("Non max supression - unsupported scores data type");
     }
 }
 
 template <typename T, typename MemT>
-T load_scalar_impl(memory_impl& mem) {
-    mem_lock<MemT> lock(mem);
+T load_scalar_impl(stream& stream, memory::ptr mem) {
+    mem_lock<MemT> lock(mem, stream);
     auto ptr = lock.data();
 
     return static_cast<T>(ptr[0]);
 }
 
 template <typename T>
-T load_scalar(memory_impl& mem) {
-    auto data_type = mem.get_layout().data_type;
+T load_scalar(stream& stream, memory::ptr mem) {
+    auto data_type = mem->get_layout().data_type;
     switch (data_type) {
     case cldnn::data_types::i32:
-        return load_scalar_impl<T, data_type_to_type<data_types::i32>::type>(mem);
+        return load_scalar_impl<T, data_type_to_type<data_types::i32>::type>(stream, mem);
     case cldnn::data_types::f16:
-        return load_scalar_impl<T, data_type_to_type<data_types::f16>::type>(mem);
+        return load_scalar_impl<T, data_type_to_type<data_types::f16>::type>(stream, mem);
     case cldnn::data_types::f32:
-        return load_scalar_impl<T, data_type_to_type<data_types::f32>::type>(mem);
+        return load_scalar_impl<T, data_type_to_type<data_types::f32>::type>(stream, mem);
     default:
         throw std::runtime_error("Non max supression - unsupported data type");
     }
 }
 
 template <typename T>
-void store_result_impl(memory_impl& mem, const std::vector<result_indices>& result) {
-    mem_lock<T> lock(mem);
+void store_result_impl(stream& stream, memory::ptr mem, const std::vector<result_indices>& result) {
+    mem_lock<T> lock(mem, stream);
     auto ptr = lock.data();
 
-    auto output_size = static_cast<size_t>(mem.get_layout().size.batch[0]);
+    auto output_size = static_cast<size_t>(mem->get_layout().size.batch[0]);
     auto results_size = result.size();
 
     size_t si = 0;
@@ -240,31 +240,31 @@ void store_result_impl(memory_impl& mem, const std::vector<result_indices>& resu
     }
 }
 
-void store_result(memory_impl& mem, const std::vector<result_indices>& result) {
-    auto data_type = mem.get_layout().data_type;
+void store_result(stream& stream, memory::ptr mem, const std::vector<result_indices>& result) {
+    auto data_type = mem->get_layout().data_type;
     switch (data_type) {
     case cldnn::data_types::i32:
-        store_result_impl<data_type_to_type<data_types::i32>::type>(mem, result);
+        store_result_impl<data_type_to_type<data_types::i32>::type>(stream, mem, result);
         break;
     case cldnn::data_types::f16:
-        store_result_impl<data_type_to_type<data_types::f16>::type>(mem, result);
+        store_result_impl<data_type_to_type<data_types::f16>::type>(stream, mem, result);
         break;
     case cldnn::data_types::f32:
-        store_result_impl<data_type_to_type<data_types::f32>::type>(mem, result);
+        store_result_impl<data_type_to_type<data_types::f32>::type>(stream, mem, result);
         break;
     default:
         throw std::runtime_error("Non max supression - unsupported output data type");
     }
 }
 
-void store_first_output(memory_impl& mem, const std::vector<result_indices>& result) {
-    auto data_type = mem.get_layout().data_type;
+void store_first_output(stream& stream, memory::ptr mem, const std::vector<result_indices>& result) {
+    auto data_type = mem->get_layout().data_type;
     switch (data_type) {
     case cldnn::data_types::i32:
-        store_result_impl<data_type_to_type<data_types::i32>::type>(mem, result);
+        store_result_impl<data_type_to_type<data_types::i32>::type>(stream, mem, result);
         break;
     case cldnn::data_types::i64:
-        store_result_impl<data_type_to_type<data_types::i32>::type>(mem, result);
+        store_result_impl<data_type_to_type<data_types::i32>::type>(stream, mem, result);
         break;
     default:
         throw std::runtime_error("Non max supression - unsupported output data type");
@@ -272,11 +272,11 @@ void store_first_output(memory_impl& mem, const std::vector<result_indices>& res
 }
 
 template <typename T>
-void store_second_output_impl(memory_impl& mem, const std::vector<result_indices>& result) {
-    mem_lock<T> lock(mem);
+void store_second_output_impl(stream& stream, memory::ptr mem, const std::vector<result_indices>& result) {
+    mem_lock<T> lock(mem, stream);
     auto ptr = lock.data();
 
-    auto output_size = static_cast<size_t>(mem.get_layout().size.batch[0]);
+    auto output_size = static_cast<size_t>(mem->get_layout().size.batch[0]);
     auto results_size = result.size();
 
     size_t si = 0;
@@ -294,14 +294,14 @@ void store_second_output_impl(memory_impl& mem, const std::vector<result_indices
     }
 }
 
-void store_second_output(memory_impl& mem, const std::vector<result_indices>& result) {
-    auto data_type = mem.get_layout().data_type;
+void store_second_output(stream& stream, memory::ptr mem, const std::vector<result_indices>& result) {
+    auto data_type = mem->get_layout().data_type;
     switch (data_type) {
     case cldnn::data_types::f16:
-        store_second_output_impl<data_type_to_type<data_types::f16>::type>(mem, result);
+        store_second_output_impl<data_type_to_type<data_types::f16>::type>(stream, mem, result);
         break;
     case cldnn::data_types::f32:
-        store_second_output_impl<data_type_to_type<data_types::f32>::type>(mem, result);
+        store_second_output_impl<data_type_to_type<data_types::f32>::type>(stream, mem, result);
         break;
     default:
         throw std::runtime_error("Non max supression - unsupported second output data type");
@@ -309,20 +309,20 @@ void store_second_output(memory_impl& mem, const std::vector<result_indices>& re
 }
 
 template <typename T>
-void store_third_output_impl(memory_impl& mem, const std::vector<result_indices>& result) {
-    mem_lock<T> lock(mem);
+void store_third_output_impl(stream& stream, memory::ptr mem, const std::vector<result_indices>& result) {
+    mem_lock<T> lock(mem, stream);
     auto ptr = lock.data();
     ptr[0] = static_cast<T>(result.size());
 }
 
-void store_third_output(memory_impl& mem, const std::vector<result_indices>& result) {
-    auto data_type = mem.get_layout().data_type;
+void store_third_output(stream& stream, memory::ptr mem, const std::vector<result_indices>& result) {
+    auto data_type = mem->get_layout().data_type;
     switch (data_type) {
     case cldnn::data_types::i32:
-        store_third_output_impl<data_type_to_type<data_types::i32>::type>(mem, result);
+        store_third_output_impl<data_type_to_type<data_types::i32>::type>(stream, mem, result);
         break;
     case cldnn::data_types::i64:
-        store_third_output_impl<data_type_to_type<data_types::i32>::type>(mem, result);
+        store_third_output_impl<data_type_to_type<data_types::i32>::type>(stream, mem, result);
         break;
     default:
         throw std::runtime_error("Non max supression - unsupported third output data type");
@@ -331,9 +331,10 @@ void store_third_output(memory_impl& mem, const std::vector<result_indices>& res
 
 void run(non_max_suppression_inst& instance) {
     auto prim = instance.node.get_primitive();
+    auto& stream = instance.get_network().get_stream();
 
-    auto boxes = load_boxes(instance.input_boxes_mem(), prim->center_point_box);
-    auto scores = load_scores(instance.input_scores_mem());
+    auto boxes = load_boxes(stream, instance.input_boxes_mem(), prim->center_point_box);
+    auto scores = load_scores(stream, instance.input_scores_mem());
 
     int num_select_per_class = 0;
     float iou_threshold = 1.f;
@@ -341,58 +342,61 @@ void run(non_max_suppression_inst& instance) {
     float soft_nms_sigma = 0.f;
 
     if (instance.has_num_select_per_class()) {
-        num_select_per_class = load_scalar<int>(instance.num_select_per_class_mem());
+        num_select_per_class = load_scalar<int>(stream, instance.num_select_per_class_mem());
     }
 
     if (instance.has_iou_threshold()) {
-        iou_threshold = load_scalar<float>(instance.iou_threshold_mem());
+        iou_threshold = load_scalar<float>(stream, instance.iou_threshold_mem());
     }
 
     if (instance.has_score_threshold()) {
-        score_threshold = load_scalar<float>(instance.score_threshold_mem());
+        score_threshold = load_scalar<float>(stream, instance.score_threshold_mem());
     }
 
     if (instance.has_soft_nms_sigma()) {
-        soft_nms_sigma = load_scalar<float>(instance.soft_nms_sigma_mem());
+        soft_nms_sigma = load_scalar<float>(stream, instance.soft_nms_sigma_mem());
     }
 
     auto result = run_nms(boxes, scores, num_select_per_class, score_threshold, iou_threshold, soft_nms_sigma, prim->sort_result_descending);
 
     if (instance.has_third_output()) {
-        store_third_output(instance.third_output_mem(), result);
+        store_third_output(stream, instance.third_output_mem(), result);
     }
 
     if (instance.has_second_output()) {
-        store_second_output(instance.second_output_mem(), result);
-        store_first_output(instance.output_memory(), result);
+        store_second_output(stream, instance.second_output_mem(), result);
+        store_first_output(stream, instance.output_memory_ptr(), result);
         return;
     }
 
-    store_result(instance.output_memory(), result);
+    store_result(stream, instance.output_memory_ptr(), result);
 }
 
 struct non_max_suppression_cpu : typed_primitive_impl<non_max_suppression> {
     using parent = typed_primitive_impl<non_max_suppression>;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<non_max_suppression_cpu>(*this);
+    }
+
     non_max_suppression_cpu() : parent(kernel_selector::weights_reorder_params(), "non_max_suppression_cpu") {}
 
-    virtual event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& event,
-                                         typed_primitive_inst<non_max_suppression>& instance) {
+    virtual event::ptr execute_impl(const std::vector<event::ptr>& event, typed_primitive_inst<non_max_suppression>& instance) {
         for (auto e : event) {
             e->wait();
         }
 
-        auto ev = instance.get_network().get_engine().create_user_event(instance.get_network().get_id(), false);
+        auto& stream = instance.get_network().get_stream();
 
         run(instance);
 
-        dynamic_cast<cldnn::user_event*>(ev.get())->set();  // set as complete
-        return ev;
+        return stream.create_user_event(true);
     }
 
     static primitive_impl* create(const non_max_suppression_node&) {
         return new non_max_suppression_cpu();
     }
+    void init_kernels() override {}
 };
 }  // namespace
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/normalize_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/normalize_gpu.cpp
index 9a8a025ffb78bb..7ed0653035fc53 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/normalize_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/normalize_gpu.cpp
@@ -5,7 +5,7 @@
 #include "normalize_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "normalize/normalize_kernel_selector.h"
 #include "normalize/normalize_kernel_base.h"
@@ -21,11 +21,14 @@ struct normalize_gpu : typed_primitive_gpu_impl<normalize> {
     using parent = typed_primitive_gpu_impl<normalize>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<normalize_gpu>(*this);
+    }
+
 protected:
-     kernel::kernel_arguments_data get_arguments(typed_primitive_inst<normalize>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
-        args.scale_table = (memory_impl::cptr) &instance.scale_memory();
+     kernel_arguments_data get_arguments(typed_primitive_inst<normalize>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
+        args.scale_table = instance.scale_memory();
         return args;
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ocl_builder.h b/inference-engine/thirdparty/clDNN/src/gpu/ocl_builder.h
deleted file mode 100644
index 5f426c9042f27d..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/ocl_builder.h
+++ /dev/null
@@ -1,38 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-
-#include <cl2_wrapper.h>
-#include <list>
-#include <string>
-#include <utility>
-#include <vector>
-#include <map>
-#include "device_impl.h"
-
-namespace cldnn {
-namespace gpu {
-struct configuration;
-
-class ocl_builder {
-private:
-    const uint32_t device_type = CL_DEVICE_TYPE_GPU;  // cldnn supports only gpu devices
-    const uint32_t device_vendor = 0x8086;  // Intel vendor
-public:
-    ocl_builder() = default;
-
-    std::map<std::string, device_impl::ptr> get_available_devices(void* user_context, void* user_device) const;
-    uint32_t get_device_type() const { return device_type; }
-    uint32_t get_device_vendor() const { return device_vendor; }
-private:
-    bool does_device_match_config(bool out_of_order, const cl::Device& device) const;
-    std::vector<device_impl::ptr> build_device_list(bool out_out_order) const;
-    std::vector<device_impl::ptr> build_device_list_from_user_context(bool out_out_order, void* user_context) const;
-    std::vector<device_impl::ptr> build_device_list_from_user_device(bool out_out_order, void* user_device) const;
-};
-
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ocl_queue_wrapper.cpp b/inference-engine/thirdparty/clDNN/src/gpu/ocl_queue_wrapper.cpp
deleted file mode 100644
index 60298e65a8547f..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/ocl_queue_wrapper.cpp
+++ /dev/null
@@ -1,158 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "ocl_queue_wrapper.h"
-#include "ocl_base_event.h"
-#include "ocl_user_event.h"
-#include "command_queues_builder.h"
-#include "events_pool.h"
-
-#include <cassert>
-#include <iomanip>
-#include <ios>
-
-#include <fstream>
-#include <thread>
-#include <string>
-#include <vector>
-#include <memory>
-
-// NOTE: Due to buggy scope transition of warnings we need to disable warning in place of use/instantation
-//       of some types (even though we already disabled them in scope of definition of these types).
-//       Moreover this warning is pretty much now only for annoyance: it is generated due to lack
-//       of proper support for mangling of custom GCC attributes into type name (usually when used
-//       with templates, even from standard library).
-#if defined __GNUC__ && __GNUC__ >= 6
-#pragma GCC diagnostic ignored "-Wignored-attributes"
-#endif
-
-namespace cldnn {
-namespace gpu {
-
-gpu_queue::gpu_queue(uint32_t id, queue_type queue, std::shared_ptr<gpu_toolkit> context)
-    : id(id), _context(context), _command_queue(queue), _events_pool(new events_pool()) {}
-
-event_impl::ptr gpu_queue::enqueue_kernel(kernels_cache::kernel_type const& kern,
-                                          cl::NDRange const& global,
-                                          cl::NDRange const& local,
-                                          std::vector<event_impl::ptr> const& deps) {
-    std::vector<cl::Event> dep_events;
-    auto dep_events_ptr = &dep_events;
-    if (!context()->get_configuration().host_out_of_order) {
-        for (auto& dep : deps) {
-            if (auto ocl_base_ev = dynamic_cast<ocl_base_event*>(dep.get())) {
-                dep_events.push_back(ocl_base_ev->get());
-            }
-        }
-    } else {
-        dep_events_ptr = nullptr;
-
-        sync_events(deps);
-    }
-
-    cl::Event ret_ev;
-
-    try {
-        if (!context()->get_configuration().host_out_of_order || _output_event ||
-            context()->get_configuration().enable_profiling) {
-            _command_queue.enqueueNDRangeKernel(kern, cl::NullRange, global, local, dep_events_ptr, &ret_ev);
-        } else {
-            _command_queue.enqueueNDRangeKernel(kern, cl::NullRange, global, local, dep_events_ptr, nullptr);
-        }
-    } catch (cl::Error const& err) {
-        throw ocl_error(err);
-    }
-
-    return _events_pool->get_from_base_pool(context(), ret_ev, ++_queue_counter);
-}
-
-event_impl::ptr gpu_queue::enqueue_marker(std::vector<event_impl::ptr> const& deps) {
-    if (deps.empty())
-        return _events_pool->get_from_user_pool(context(), true);
-
-    bool enabled_single_kernel = context()->get_configuration().single_kernel_name == "" ? false : true;
-    if (!context()->get_configuration().host_out_of_order) {
-        cl::Event ret_ev;
-        if (!enabled_single_kernel) {
-            std::vector<cl::Event> dep_events;
-            for (auto& dep : deps) {
-                if (auto ocl_base_ev = dynamic_cast<ocl_base_event*>(dep.get()))
-                    dep_events.push_back(ocl_base_ev->get());
-            }
-
-            try {
-                _command_queue.enqueueMarkerWithWaitList(&dep_events, &ret_ev);
-            } catch (cl::Error const& err) {
-                throw ocl_error(err);
-            }
-        } else {
-            try {
-                _command_queue.enqueueMarkerWithWaitList(nullptr, &ret_ev);
-            } catch (cl::Error const& err) {
-                throw ocl_error(err);
-            }
-        }
-
-        return _events_pool->get_from_base_pool(context(), ret_ev, ++_queue_counter);
-    } else {
-        sync_events(deps);
-        return _events_pool->get_from_base_pool(context(), _last_barrier_ev, _last_barrier);
-    }
-}
-
-event_impl::ptr gpu_queue::group_events(std::vector<event_impl::ptr> const& deps) {
-    return _events_pool->get_from_group_pool(context(), deps);
-}
-
-event_impl::ptr gpu_queue::create_user_event(bool set) { return _events_pool->get_from_user_pool(context(), set); }
-
-void gpu_queue::reset_events() { _events_pool->reset_events(); }
-
-void gpu_queue::release_events_pool() { _events_pool.reset(); }
-
-void gpu_queue::flush() { queue().flush(); }
-
-void gpu_queue::release_pending_memory() {
-    /*
-    TODO: Temp. solution, untill proper API calls from OpenCL are released.
-    */
-    void* ptr = nullptr;
-    ptr = _mm_malloc(4096, 4096);
-    queue().finish();
-    try {
-        cl::Buffer flusher(context()->context(), CL_MEM_USE_HOST_PTR, (size_t)4096, ptr);
-        flusher = (cl_mem) nullptr;  // clear buffer
-    } catch (...) {
-        _mm_free(ptr);
-        throw;
-    }
-    _mm_free(ptr);
-}
-
-void gpu_queue::sync_events(std::vector<event_impl::ptr> const& deps) {
-    bool needs_barrier = false;
-    for (auto& dep : deps) {
-        auto* ocl_base_ev = dynamic_cast<ocl_base_event*>(dep.get());
-        if (ocl_base_ev->get_queue_stamp() > _last_barrier) {
-            needs_barrier = true;
-        }
-    }
-
-    if (needs_barrier) {
-        try {
-            if (_output_event)
-                _command_queue.enqueueBarrierWithWaitList(nullptr, &_last_barrier_ev);
-            else
-                _command_queue.enqueueBarrierWithWaitList(nullptr, nullptr);
-        } catch (cl::Error const& err) {
-            throw ocl_error(err);
-        }
-
-        _last_barrier = ++_queue_counter;
-    }
-}
-
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ocl_queue_wrapper.h b/inference-engine/thirdparty/clDNN/src/gpu/ocl_queue_wrapper.h
deleted file mode 100644
index 55dabcdcfd04db..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/ocl_queue_wrapper.h
+++ /dev/null
@@ -1,96 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-
-#include "ocl_builder.h"
-
-#include "kernels_cache.h"
-#include "device_info.h"
-#include "event_impl.h"
-#include "configuration.h"
-
-#include <memory>
-#include <chrono>
-#include <thread>
-#include <iostream>
-#include <sstream>
-#include <utility>
-#include <vector>
-
-namespace cldnn {
-typedef cl::vector<cl::vector<unsigned char>> kernels_binaries_vector;
-typedef cl::vector<kernels_binaries_vector> kernels_binaries_container;
-using queue_type = cl::CommandQueueIntel;
-namespace gpu {
-typedef CL_API_ENTRY cl_command_queue(CL_API_CALL* pfn_clCreateCommandQueueWithPropertiesINTEL)(
-    cl_context context,
-    cl_device_id device,
-    const cl_queue_properties_khr* properties,
-    cl_int* errcodeRet);
-
-class gpu_toolkit;
-class events_pool;
-
-class gpu_queue {
-public:
-    const queue_type& queue() const { return _command_queue; }
-    gpu_queue(uint32_t id, queue_type queue, std::shared_ptr<gpu_toolkit> context);
-    gpu_queue(gpu_queue&& other)
-        : id(other.id),
-          _context(other._context),
-          _command_queue(other._command_queue),
-          _queue_counter(other._queue_counter.load()),
-          _last_barrier(other._last_barrier.load()),
-          _events_pool(std::move(other._events_pool)),
-          _last_barrier_ev(other._last_barrier_ev),
-          _output_event(other._output_event) {}
-
-    gpu_queue& operator=(gpu_queue&& other) {
-        if (this != &other) {
-            id = other.id;
-            _context = std::move(other._context);
-            _command_queue = std::move(other._command_queue);
-            _queue_counter = std::move(other._queue_counter.load());
-            _last_barrier = std::move(other._last_barrier.load());
-            _events_pool = std::move(std::move(other._events_pool));
-            _last_barrier_ev = std::move(other._last_barrier_ev);
-            _output_event = std::move(other._output_event);
-        }
-        return *this;
-    }
-
-    ~gpu_queue() = default;
-
-    void sync_events(std::vector<event_impl::ptr> const& deps);
-    void release_pending_memory();
-    void flush();
-
-    void set_output_event(bool out_event) { _output_event = out_event; }
-
-    event_impl::ptr enqueue_kernel(kernels_cache::kernel_type const& kern,
-                                   cl::NDRange const& global,
-                                   cl::NDRange const& local,
-                                   std::vector<event_impl::ptr> const& deps);
-    event_impl::ptr enqueue_marker(std::vector<event_impl::ptr> const& deps);
-    event_impl::ptr group_events(std::vector<event_impl::ptr> const& deps);
-    void reset_events();
-    event_impl::ptr create_user_event(bool set);
-    void release_events_pool();
-    std::shared_ptr<gpu_toolkit> context() { return _context.lock(); }
-
-private:
-    uint32_t id;
-    std::weak_ptr<gpu_toolkit> _context;
-    queue_type _command_queue;
-    std::atomic<uint64_t> _queue_counter{0};
-    std::atomic<uint64_t> _last_barrier{0};
-    std::shared_ptr<events_pool> _events_pool;
-    cl::Event _last_barrier_ev;
-    bool _output_event = false;
-};
-
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ocl_toolkit.cpp b/inference-engine/thirdparty/clDNN/src/gpu/ocl_toolkit.cpp
deleted file mode 100644
index 3ee0842eb26f95..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/ocl_toolkit.cpp
+++ /dev/null
@@ -1,236 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "ocl_toolkit.h"
-#include "ocl_base_event.h"
-#include "ocl_user_event.h"
-#include "command_queues_builder.h"
-#include "events_pool.h"
-
-#include <cassert>
-#include <iomanip>
-#include <ios>
-
-#include <fstream>
-#include <thread>
-#include <string>
-#include <vector>
-#include <memory>
-#include <utility>
-
-// NOTE: Due to buggy scope transition of warnings we need to disable warning in place of use/instantation
-//       of some types (even though we already disabled them in scope of definition of these types).
-//       Moreover this warning is pretty much now only for annoyance: it is generated due to lack
-//       of proper support for mangling of custom GCC attributes into type name (usually when used
-//       with templates, even from standard library).
-#if defined __GNUC__ && __GNUC__ >= 6
-#pragma GCC diagnostic ignored "-Wignored-attributes"
-#endif
-
-// static class memebers - pointers to dynamically obtained OpenCL extension functions
-cl::PFN_clEnqueueAcquireMediaSurfacesINTEL cl::SharedSurfLock::pfn_acquire = NULL;
-cl::PFN_clEnqueueReleaseMediaSurfacesINTEL cl::SharedSurfLock::pfn_release = NULL;
-cl::PFN_clCreateFromMediaSurfaceINTEL cl::ImageVA::pfn_clCreateFromMediaSurfaceINTEL = NULL;
-#ifdef _WIN32
-cl::PFN_clCreateFromD3D11Buffer cl::BufferDX::pfn_clCreateFromD3D11Buffer = NULL;
-#endif
-
-namespace cldnn {
-namespace gpu {
-
-ocl_error::ocl_error(cl::Error const& err)
-    : std::runtime_error(err.what() + std::string(", error code: ") + std::to_string(err.err())) {}
-
-std::mutex gpu_toolkit::cache_mutex;
-
-std::shared_ptr<gpu_toolkit> gpu_toolkit::create(const device_impl& device, const configuration& cfg) {
-    struct make_shared_wa : public gpu_toolkit {
-        explicit make_shared_wa(const device_impl& device, const configuration& cfg)
-            : gpu_toolkit(device, cfg) {}
-    };
-    try {
-        auto ctx = std::make_shared<make_shared_wa>(device, cfg);
-        ctx->add_network(0);
-        return ctx;
-    } catch (cl::Error const& err) {
-        throw ocl_error(err);
-    }
-}
-
-struct gpu_toolkit::ocl_logger {
-    std::ofstream _log_file;
-};
-
-gpu_toolkit::gpu_toolkit(const device_impl& device_impl, const configuration& config)
-    : _configuration(config),
-      _device(&device_impl),
-      _neo_driver(strstr(get_device_version().c_str(), "NEO") ? true : false) {
-    device().getInfo(CL_DEVICE_EXTENSIONS, &_extensions);
-
-    device_cache_reader dc_reader(_configuration.tuning_cache_path);
-    _device_cache = dc_reader.get();
-
-    _logger = std::unique_ptr<ocl_logger>(new ocl_logger());
-    if (logging_enabled()) {
-        auto device_info = get_device_info();
-        open_log() << "Engine configuration:\n"
-                   << "    profiling: " << std::boolalpha << _configuration.enable_profiling << "\n"
-                   << "    meaningful names: " << std::boolalpha << _configuration.meaningful_kernels_names << "\n"
-                   << "    dump custom program: " << std::boolalpha << _configuration.dump_custom_program << "\n"
-                   << "    vendor type: " << std::hex << std::setfill('0') << std::setw(4) << std::right
-                   << std::to_string(device_info.vendor_id) << "\n"
-                   << std::dec << std::setfill(' ') << std::right
-                   << "    compiler options: " << _configuration.compiler_options << "\n"
-                   << "    single kernel name: " << _configuration.single_kernel_name << "\n"
-                   << "    out-of-order: " << std::boolalpha << config.host_out_of_order << "\n"
-                   << "    engine log: " << _configuration.log << "\n"
-                   << "    sources dumps: " << _configuration.ocl_sources_dumps_dir << "\n"
-                   << "\nEngine info:\n"
-                   << "    cores count: " << device_info.cores_count << "\n"
-                   << "    core frequencey: " << device_info.core_frequency << "\n"
-                   << "    max work group size: " << device_info.max_work_group_size << "\n"
-                   << "    local memory size: " << device_info.max_local_mem_size << "\n"
-                   << "    fp16: " << std::boolalpha << (device_info.supports_fp16 != 0) << "\n"
-                   << "    fp16 denorms: " << std::boolalpha << (device_info.supports_fp16_denorms != 0) << "\n"
-                   << "    subgroups short: " << std::boolalpha << (device_info.supports_subgroups_short != 0) << "\n"
-                   << "    local block io: " << std::boolalpha << device_info.supports_local_block_io << "\n"
-                   << "    optimization hints: " << std::boolalpha << device_info.supports_optimization_hints << std::endl;
-    }
-}
-
-gpu_queue& gpu_toolkit::get_command_queue(uint32_t id) {
-    return _command_queues_w.at(id);
-}
-
-gpu_program_state& gpu_toolkit::get_program_state(uint32_t id) {
-    std::lock_guard<std::mutex> lock(toolkit_mutex);
-    return *_program_states.at(id);
-}
-
-void gpu_toolkit::add_program(uint32_t prog_id) {
-    std::lock_guard<std::mutex> lock(toolkit_mutex);
-    _program_states.emplace(std::make_pair(prog_id, std::make_shared<gpu_program_state>(*this, prog_id)));
-}
-
-void gpu_toolkit::remove_program(uint32_t prog_id) {
-    std::lock_guard<std::mutex> lock(toolkit_mutex);
-    auto state_iter = _program_states.find(prog_id);
-
-    if (state_iter != _program_states.end()) {
-        _program_states.erase(state_iter);
-    }
-}
-
-kernels_cache& gpu_toolkit::get_kernels_cache(uint32_t prog_id) {
-    return get_program_state(prog_id)._kernels_cache;
-}
-
-void gpu_toolkit::add_network(uint32_t net_id) {
-    std::lock_guard<std::mutex> lock(toolkit_mutex);
-    command_queues_builder queue_builder(context(), device(), _device->get_platform());
-    queue_builder.set_profiling(_configuration.enable_profiling);
-    queue_builder.set_out_of_order((_configuration.host_out_of_order && _neo_driver));
-
-    bool priorty_extensions =
-        extension_supported("cl_khr_priority_hints") && extension_supported("cl_khr_create_command_queue");
-    queue_builder.set_priority_mode(_configuration.priority_mode, priorty_extensions);
-
-    bool throttle_extensions =
-        extension_supported("cl_khr_throttle_hints") && extension_supported("cl_khr_create_command_queue");
-    queue_builder.set_throttle_mode(_configuration.throttle_mode, throttle_extensions);
-
-    queue_builder.build();
-    _command_queues_w.emplace(std::make_pair(net_id,
-        gpu_queue(net_id, queue_builder.queue(), shared_from_this())));
-}
-
-void gpu_toolkit::remove_network(uint32_t net_id) {
-    std::lock_guard<std::mutex> lock(toolkit_mutex);
-    auto net_iter = _command_queues_w.find(net_id);
-    if (net_iter != _command_queues_w.end()) {
-        // net_iter->second.release_pending_memory();
-        _command_queues_w.erase(net_iter);
-    }
-}
-
-event_impl::ptr gpu_toolkit::enqueue_kernel(uint32_t queue_id,
-                                            kernels_cache::kernel_type const& kern,
-                                            cl::NDRange const& global,
-                                            cl::NDRange const& local,
-                                            std::vector<event_impl::ptr> const& deps) {
-    return get_command_queue(queue_id).enqueue_kernel(kern, global, local, deps);
-}
-
-event_impl::ptr gpu_toolkit::enqueue_marker(uint32_t queue_id, std::vector<event_impl::ptr> const& deps) {
-    return get_command_queue(queue_id).enqueue_marker(deps);
-}
-
-event_impl::ptr gpu_toolkit::group_events(uint32_t queue_id, std::vector<event_impl::ptr> const& deps) {
-    return get_command_queue(queue_id).group_events(deps);
-}
-
-event_impl::ptr gpu_toolkit::create_user_event(uint32_t queue_id, bool set) {
-    return get_command_queue(queue_id).create_user_event(set);
-}
-
-void gpu_toolkit::reset_events(uint32_t queue_id) { get_command_queue(queue_id).reset_events(); }
-
-void gpu_toolkit::release_events_pool(uint32_t queue_id) { get_command_queue(queue_id).release_events_pool(); }
-
-void gpu_toolkit::release_all_events_pools() {
-    for (auto& queue : _command_queues_w) {
-        queue.second.release_events_pool();
-    }
-}
-
-void gpu_toolkit::flush(uint32_t queue_id) { get_command_queue(queue_id).flush(); }
-
-void gpu_toolkit::release_pending_memory(uint32_t queue_id) { get_command_queue(queue_id).release_pending_memory(); }
-
-void gpu_toolkit::wait_for_events(std::vector<event_impl::ptr> const& events) {
-    std::vector<cl::Event> clevents;
-    for (auto& ev : events) {
-        if (auto ocl_base_ev = dynamic_cast<ocl_base_event*>(ev.get()))
-            clevents.push_back(ocl_base_ev->get());
-    }
-
-    try {
-        cl::WaitForEvents(clevents);
-    } catch (cl::Error const& err) {
-        throw ocl_error(err);
-    }
-}
-
-void gpu_toolkit::log(uint64_t id, std::string const& msg) {
-    if (_configuration.log.empty())
-        return;
-
-    open_log() << "[" << id << "] " << msg << std::endl;
-}
-
-void gpu_toolkit::set_output_event(uint32_t queue_id, bool out_event) {
-    get_command_queue(queue_id).set_output_event(out_event);
-}
-
-std::ofstream& gpu_toolkit::open_log() {
-    if (!_logger->_log_file.is_open()) {
-        _logger->_log_file.open(_configuration.log, std::ios::out | std::ios::trunc);
-        if (!_logger->_log_file.good()) {
-            _logger->_log_file.close();
-            throw std::runtime_error("Could not initialize ocl_toolkit log file");
-        }
-
-        if (!_logger->_log_file.is_open()) {
-            _logger->_log_file.close();
-            throw std::runtime_error("Could not open ocl_toolkit log file '" + _configuration.log + "' for writing");
-        }
-    }
-
-    return _logger->_log_file;
-}
-
-}  // namespace gpu
-
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/ocl_toolkit.h b/inference-engine/thirdparty/clDNN/src/gpu/ocl_toolkit.h
deleted file mode 100644
index ba8a3cb8d26e1e..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/ocl_toolkit.h
+++ /dev/null
@@ -1,147 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include "device_info.h"
-#include "device_impl.h"
-#include "kernels_cache.h"
-#include "event_impl.h"
-#include "configuration.h"
-#include "ocl_queue_wrapper.h"
-#include "device_cache_reader.h"
-
-#include <memory>
-#include <chrono>
-#include <string>
-#include <vector>
-#include <map>
-#include <stdexcept>
-
-namespace cldnn {
-typedef cl::vector<cl::vector<unsigned char>> kernels_binaries_vector;
-typedef cl::vector<kernels_binaries_vector> kernels_binaries_container;
-using queue_type = cl::CommandQueueIntel;
-namespace gpu {
-typedef CL_API_ENTRY cl_command_queue(CL_API_CALL* pfn_clCreateCommandQueueWithPropertiesINTEL)(
-    cl_context context,
-    cl_device_id device,
-    const cl_queue_properties_khr* properties,
-    cl_int* errcodeRet);
-
-class ocl_error : public std::runtime_error {
-public:
-    explicit ocl_error(cl::Error const& err);
-};
-
-class events_pool;
-class gpu_toolkit;
-
-class context_holder {
-protected:
-    explicit context_holder(std::shared_ptr<gpu_toolkit> context) : _context(context) {}
-    virtual ~context_holder() = default;
-
-    const std::shared_ptr<gpu_toolkit>& context() const { return _context; }
-
-    std::shared_ptr<gpu_toolkit> _context;
-};
-
-struct gpu_program_state {
-    kernels_cache _kernels_cache;
-
-    gpu_program_state(gpu_toolkit& context, uint32_t prog_id) :
-        _kernels_cache(context, prog_id) {}
-};
-
-class gpu_toolkit : public std::enable_shared_from_this<gpu_toolkit> {
-    friend class context_holder;
-
-protected:
-    explicit gpu_toolkit(const device_impl& device_impl,
-        const configuration& aconfiguration = configuration());
-
-public:
-    static std::shared_ptr<gpu_toolkit> create(const device_impl& device_impl,
-        const configuration& cfg = configuration());
-    const cl::Context context() const { return _device->get_context(); }
-    const cl::Device device() const { return _device->get_device(); }
-    const memory_capabilities memory_caps() const { return _device->mem_caps(); }
-    const queue_type& queue(uint32_t id) { return get_command_queue(id).queue(); }
-
-    const configuration& get_configuration() const { return _configuration; }
-    device_info_internal get_device_info() const { return _device->get_info(); }
-    std::shared_ptr<kernel_selector::TuningCache> get_device_cache() const { return _device_cache; }
-    kernels_cache& get_kernels_cache(uint32_t prog_id);
-    bool get_serialization_flag() { return _serialize; }
-    void set_serialization_flag(bool serialization_flag) { _serialize = serialization_flag; }
-
-    inline bool extension_supported(const std::string ext) { return _extensions.find(ext) != std::string::npos; }
-
-    gpu_toolkit(const gpu_toolkit& other) = delete;
-    gpu_toolkit(gpu_toolkit&& other) = delete;
-    gpu_toolkit& operator=(const gpu_toolkit& other) = delete;
-    gpu_toolkit& operator=(gpu_toolkit&& other) = delete;
-    std::string single_kernel_name() const { return _configuration.single_kernel_name; }
-    bool enabled_single_kernel() const { return single_kernel_name() == "" ? false : true; }
-
-    void set_output_event(uint32_t queue_id, bool out_event);
-
-    event_impl::ptr enqueue_kernel(uint32_t queue_id,
-                                   kernels_cache::kernel_type const& kern,
-                                   cl::NDRange const& global,
-                                   cl::NDRange const& local,
-                                   std::vector<event_impl::ptr> const& deps);
-    event_impl::ptr enqueue_marker(uint32_t queue_id, std::vector<event_impl::ptr> const& deps);
-    event_impl::ptr group_events(uint32_t queue_id, std::vector<event_impl::ptr> const& deps);
-    void reset_events(uint32_t queue_id);
-    event_impl::ptr create_user_event(uint32_t queue_id, bool set);
-    void release_events_pool(uint32_t queue_id);
-    void release_all_events_pools();
-
-    void flush(uint32_t queue_id);
-    void release_pending_memory(uint32_t queue_id);
-    void wait_for_events(std::vector<event_impl::ptr> const& events);
-
-    void log(uint64_t id, std::string const& msg);
-    bool logging_enabled() const { return !_configuration.log.empty(); }
-    bool is_neo_driver() { return _neo_driver; }
-    void add_network(uint32_t net_id);
-    void remove_network(uint32_t net_id);
-
-    void add_program(uint32_t prog_id);
-    void remove_program(uint32_t prog_id);
-
-    std::mutex& get_cache_mutex() { return cache_mutex; }
-
-private:
-    configuration _configuration;
-    device_impl::cptr _device;
-    bool _neo_driver = false;
-    std::map<uint32_t, std::shared_ptr<gpu_program_state>> _program_states;
-    std::map<uint32_t, gpu_queue> _command_queues_w;
-    std::shared_ptr<kernel_selector::TuningCache> _device_cache;
-    bool _serialize = false;
-
-    std::string _extensions;
-
-    struct ocl_logger;
-    std::unique_ptr<ocl_logger> _logger;
-
-    // returns whether a barrier has been added
-    std::ofstream& open_log();
-
-    std::string get_device_version() { return device().getInfo<CL_DEVICE_VERSION>(); }
-
-    gpu_queue& get_command_queue(uint32_t id);
-    gpu_program_state& get_program_state(uint32_t id);
-
-    std::mutex toolkit_mutex;
-    // mutex for kernels_cache must be static to ensure that all threads run program build in a thread-safe fashion
-    // including the case when multiple IE cores are created.
-    static std::mutex cache_mutex;
-};
-
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/one_hot_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/one_hot_gpu.cpp
index 1c2704a5b26c15..218e50f018d571 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/one_hot_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/one_hot_gpu.cpp
@@ -9,7 +9,7 @@
 #include "kernel_selector_helper.h"
 #include "one_hot/one_hot_kernel_selector.h"
 #include "one_hot/one_hot_kernel_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include <vector>
 
 namespace cldnn {
@@ -19,6 +19,10 @@ struct one_hot_gpu : typed_primitive_gpu_impl<one_hot> {
     using parent = typed_primitive_gpu_impl<one_hot>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<one_hot_gpu>(*this);
+    }
+
     static primitive_impl* create(const one_hot_node& arg) {
         auto oh_params = get_default_params<kernel_selector::one_hot_params>(arg, 1);
         auto oh_optional_params =
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/permute_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/permute_gpu.cpp
index 31bc335b90470d..4c8e1f549752c4 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/permute_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/permute_gpu.cpp
@@ -5,7 +5,7 @@
 #include "permute_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "permute/permute_kernel_selector.h"
 #include "permute/permute_kernel_ref.h"
@@ -19,6 +19,10 @@ struct permute_gpu : typed_primitive_gpu_impl<permute> {
     using parent = typed_primitive_gpu_impl<permute>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<permute_gpu>(*this);
+    }
+
     static primitive_impl* create(const permute_node& arg) {
         auto permute_params = get_default_params<kernel_selector::permute_params>(arg);
         auto permute_optional_params =
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/pooling_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/pooling_gpu.cpp
index 138a235d6ce76a..a10439bb1c92cf 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/pooling_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/pooling_gpu.cpp
@@ -5,7 +5,7 @@
 #include "pooling_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "pooling/pooling_kernel_selector.h"
 #include "pooling/pooling_kernel_base.h"
@@ -68,12 +68,15 @@ struct pooling_gpu : typed_primitive_gpu_impl<pooling> {
     using parent = typed_primitive_gpu_impl<pooling>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<pooling_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<pooling>& instance,
-                                                        int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
+    kernel_arguments_data get_arguments(typed_primitive_inst<pooling>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
         if (!instance.argument.argmax.empty())
-            args.inputs.push_back((memory_impl::cptr) &instance.dep_memory(1));
+            args.inputs.push_back(instance.dep_memory_ptr(1));
         return args;
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/primitive_gpu_base.h b/inference-engine/thirdparty/clDNN/src/gpu/primitive_gpu_base.h
index c08058d9c67950..9cd55a0f9f46e6 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/primitive_gpu_base.h
+++ b/inference-engine/thirdparty/clDNN/src/gpu/primitive_gpu_base.h
@@ -8,9 +8,7 @@
 #include <thread>
 #include "primitive_inst.h"
 #include "program_impl.h"
-#include "kernel.h"
-#include "events_waiter.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "network_impl.h"
 #include "register_gpu.hpp"
@@ -31,27 +29,42 @@ For example, all gpu convolution implementations should derive from typed_primit
 template <class PType>
 struct typed_primitive_gpu_impl : public typed_primitive_impl<PType> {
     const typed_program_node<PType>& _outer;
-    device_info_internal _device_info;
     kernel_selector::kernel_data _kernel_data;
-    std::vector<gpu::kernel> _kernels;
-    std::vector<memory_impl::cptr> _intermediates_memory;
+    std::vector<kernel_id> _kernel_ids;
+    std::vector<kernel::ptr> _kernels;
+    std::vector<memory::cptr> _intermediates_memory;
+
+    typed_primitive_gpu_impl(const typed_primitive_gpu_impl<PType>& other)
+    : typed_primitive_impl<PType>(other._weights_reorder_params, other._kernel_name)
+    , _outer(other._outer)
+    , _kernel_data(other._kernel_data)
+    , _kernel_ids(other._kernel_ids)
+    , _kernels({})
+    , _intermediates_memory({}) {
+        _kernels.reserve(other._kernels.size());
+        for (size_t k = 0; k < other._kernels.size(); ++k) {
+            _kernels.emplace_back(other._kernels[k]->clone());
+        }
+        for (auto& mem : other._intermediates_memory) {
+            auto& engine = _outer.get_program().get_engine();
+            auto new_mem = engine.allocate_memory(mem->get_layout(), mem->get_allocation_type());
+            _intermediates_memory.push_back(new_mem);
+        }
+    }
 
     typed_primitive_gpu_impl(const typed_program_node<PType>& arg, const kernel_selector::kernel_data& kd)
         : typed_primitive_impl<PType>(kd.weightsReorderParams, kd.kernelName),
           _outer(arg),
-          _device_info(arg.get_program().get_engine().get_context()->get_device_info()),
           _kernel_data(kd) {
         // weights reorder params got copied to parent, clear in _kernel_data to release shared ptr
         _kernel_data.weightsReorderParams.engine = kernel_selector::generic_kernel_params::Engine::NONE;
         _kernel_data.weightsReorderParams.cpuKernel = nullptr;
         _kernel_data.weightsReorderParams.clKernel = nullptr;
 
-        _kernels.reserve(kd.kernels.size());
+        _kernel_ids.reserve(kd.kernels.size());
+        // Add selected kernels to kernels_cache for the following compilation and save output ids
         for (size_t i = 0; i < kd.kernels.size(); ++i) {
-            gpu::kernel kernel(_outer.get_program().get_engine().get_context(),
-                               kd.kernels[i].kernelString,
-                               _outer.get_program().get_id());
-            _kernels.emplace_back(std::move(kernel));
+            _kernel_ids.emplace_back(_outer.get_program().add_kernel(kd.kernels[i].code.kernelString));
         }
 
         for (auto size : kd.internalBufferSizes) {
@@ -62,7 +75,7 @@ struct typed_primitive_gpu_impl : public typed_primitive_impl<PType> {
                                       {1, 1, 1, (tensor::value_type)(size / bpp)}};
 
             auto& eimpl = arg.get_program().get_engine();
-            _intermediates_memory.push_back(eimpl.allocate_memory(expected_layout, 0));
+            _intermediates_memory.push_back(eimpl.allocate_memory(expected_layout));
         }
     }
     bool is_cpu() const override { return false; }
@@ -70,22 +83,21 @@ struct typed_primitive_gpu_impl : public typed_primitive_impl<PType> {
 protected:
     virtual bool optimized_out(typed_primitive_inst<PType>&) const { return false; }
 
-    virtual kernel::kernel_arguments_data get_arguments(typed_primitive_inst<PType>& instance,
-                                                        int32_t /*split*/) const {
-        kernel::kernel_arguments_data args;
+    virtual kernel_arguments_data get_arguments(typed_primitive_inst<PType>& instance, int32_t /*split*/) const {
+        kernel_arguments_data args;
 
         for (size_t i = 0; i < instance.inputs_memory_count(); i++) {
-            args.inputs.push_back((memory_impl::cptr)&instance.input_memory(i));
+            args.inputs.push_back(instance.input_memory_ptr(i));
         }
 
         if (instance.has_fused_primitives()) {
             size_t count = instance.get_fused_mem_count();
             for (size_t i = 0; i < count; i++) {
-                args.fused_op_inputs.push_back((memory_impl::cptr) &instance.fused_memory(i));
+                args.fused_op_inputs.push_back(instance.fused_memory(i));
             }
         }
 
-        args.output = (memory_impl::cptr) &instance.output_memory();
+        args.output = instance.output_memory_ptr();
 
         return args;
     }
@@ -94,93 +106,96 @@ struct typed_primitive_gpu_impl : public typed_primitive_impl<PType> {
     virtual uint32_t get_groups() const { return 1; }
     virtual bool get_depthwise_sep_opt() const { return false; }
 
-    event_impl::ptr aggregate_events(const std::vector<event_impl::ptr>& events,
-                                     uint32_t net_id,
-                                     bool group = false) const {
-        if (events.size() == 1)
+    event::ptr aggregate_events(const std::vector<event::ptr>& events, stream& stream, bool group = false, bool is_output = false) const {
+        if (events.size() == 1 && !is_output)
             return events[0];
 
-        if (group)
-            return _outer.get_program().get_engine().get_context()->group_events(net_id, events);
+        if (group && !is_output)
+            return stream.group_events(events);
 
-        return events_waiter(_outer.get_program().get_engine().get_context()).run(net_id, events);
+        return stream.enqueue_marker(events, is_output);
+    }
+
+    void init_kernels() override {
+        if (is_cpu()) {
+            return;
+        }
+        _kernels.clear();
+
+        _kernels.reserve(_kernel_ids.size());
+        for (size_t k = 0; k < _kernel_ids.size(); ++k) {
+            _kernels.emplace_back(std::move(_outer.get_program().get_kernel(_kernel_ids[k])));
+        }
     }
 
     void set_arguments_impl(typed_primitive_inst<PType>& instance) override {
-        uint32_t net_id = instance.get_network().get_id();
         if (optimized_out(instance) || is_cpu()) {
             return;
         }
 
         auto split = get_split();
 
+        stream& stream = instance.get_network().get_stream();
+
         // we iterate over split first in order to be able parallelism with OOOQ mechanism.
         for (size_t k = 0; k < _kernels.size(); ++k) {
             for (decltype(split) i = 0; i < split; i++) {
                 auto args = get_arguments(instance, i);
-                args.scalars = &_kernel_data.kernels[k].scalars;
+                args.scalars = &_kernel_data.kernels[k].params.scalars;
                 args.split = i;
 
                 for (const auto& m : _intermediates_memory) {
                     args.intermediates.push_back(m);
                 }
 
-                _kernels[k].set_arguments(net_id, _kernel_data.kernels[k], args);
-            }
-        }
-    }
 
-    void cleanup_impl(typed_primitive_inst<PType>& instance) override {
-        uint32_t net_id = instance.get_network().get_id();
-        if (optimized_out(instance) || is_cpu()) {
-            return;
-        }
-
-        for (size_t k = 0; k < _kernels.size(); ++k) {
-            _kernels[k].cleanup(net_id);
+                stream.set_arguments(*_kernels[k], _kernel_data.kernels[k].params, args);
+            }
         }
     }
 
-    event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& events,
-                                 typed_primitive_inst<PType>& instance) override {
-        uint32_t net_id = instance.get_network().get_id();
+    event::ptr execute_impl(const std::vector<event::ptr>& events,
+                            typed_primitive_inst<PType>& instance) override {
+        stream& stream = instance.get_network().get_stream();
         if (optimized_out(instance)) {
-            return aggregate_events(events, net_id);
+            return aggregate_events(events, stream, false, instance.is_output());
         }
 
-        std::vector<event_impl::ptr> tmp_events(events);
-        std::vector<event_impl::ptr> all_events;
+        std::vector<event::ptr> tmp_events(events);
+        std::vector<event::ptr> all_events;
 
         // TODO - split should be handle in kernel selector by providing multiple kernels.
         auto split = get_split();
 
         // we iterate over split first in order to be able parallelism with OOOQ mechanism.
         for (size_t k = 0; k < _kernels.size(); ++k) {
-            std::vector<event_impl::ptr> new_events;
+            std::vector<event::ptr> new_events;
             for (decltype(split) i = 0; i < split; i++) {
-                // is any user of the prim's users is an detecion output, set prim as a output event (event won't be
-                // nullptr)
+                // is any user of the prim's users is an detecion output, set prim as a output event (event won't be nullptr)
                 auto users = instance.node.get_users();
-                bool next_prim_is_cpu = is_any_user_cpu(users);
-                if (next_prim_is_cpu) {
-                    _kernels[k].set_output_event(net_id, true);
-                } else {
-                    _kernels[k].set_output_event(net_id, instance.node.is_output());
+                bool is_output_event = is_any_user_cpu(users) || instance.node.is_output();
+
+                auto args = get_arguments(instance, i);
+                args.scalars = &_kernel_data.kernels[k].params.scalars;
+                args.split = i;
+
+                for (const auto& m : _intermediates_memory) {
+                    args.intermediates.push_back(m);
                 }
 
-                auto event = _kernels[k].run(net_id, _kernel_data.kernels[k], tmp_events);
-                new_events.push_back(event);
-                all_events.push_back(event);
+                auto ev = stream.enqueue_kernel(*_kernels[k], _kernel_data.kernels[k].params, args, tmp_events, is_output_event);
+                new_events.push_back(ev);
+                all_events.push_back(ev);
             }
 
             tmp_events = new_events;
         }
 
         if ((all_events.size() == 0) && (tmp_events.size() > 0))
-            return aggregate_events(tmp_events, net_id);
+            return aggregate_events(tmp_events, stream);
 
         bool group_events = (all_events.size() > 1);
-        return aggregate_events(all_events, net_id, group_events);
+        return aggregate_events(all_events, stream, group_events);
     }
 };
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/proposal_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/proposal_gpu.cpp
index 889703234673a1..3bcc61c5ad06ff 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/proposal_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/proposal_gpu.cpp
@@ -3,12 +3,11 @@
 //
 
 #include "proposal_inst.h"
-#include "kernel.h"
+#include "cldnn/runtime/engine.hpp"
 #include "implementation_map.h"
 #include "network_impl.h"
-#include "engine_impl.h"
 #include "math_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "register_gpu.hpp"
 
 #include <algorithm>
@@ -197,10 +196,14 @@ struct proposal_gpu : typed_primitive_impl<proposal> {
 
     explicit proposal_gpu(const proposal_node& arg) : outer(arg) {}
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<proposal_gpu>(*this);
+    }
+
     template <typename dtype>
-    void read_image_info(proposal_inst& instance, im_info_t& im_info) {
-        auto& image_info = instance.dep_memory(proposal_inst::image_info_index);
-        mem_lock<dtype> image_info_ptr{image_info};
+    void read_image_info(stream& stream, proposal_inst& instance, im_info_t& im_info) {
+        auto image_info = instance.dep_memory_ptr(proposal_inst::image_info_index);
+        mem_lock<dtype> image_info_ptr{image_info, stream};
         const dtype* image_info_mem = image_info_ptr.data();
 
         bool swap_xy = instance.argument.swap_xy;
@@ -212,7 +215,7 @@ struct proposal_gpu : typed_primitive_impl<proposal> {
         int min_bbox_x = 1;
         int min_bbox_y = 1;
 
-        auto image_info_size = image_info.get_layout().size;
+        auto image_info_size = image_info->get_layout().size;
         auto image_info_count = image_info_size.feature[0] == 1 ? image_info_size.batch[0] : image_info_size.feature[0];
 
         int scaled_min_bbox_size = instance.argument.min_bbox_size;
@@ -259,13 +262,13 @@ struct proposal_gpu : typed_primitive_impl<proposal> {
     }
 
     template <typename dtype>
-    void execute(proposal_inst& instance, im_info_t im_info, dtype* proposal_prob_ptr = nullptr) {
+    void execute(stream& stream, proposal_inst& instance, im_info_t im_info, dtype* proposal_prob_ptr = nullptr) {
         const std::vector<proposal_inst::anchor>& anchors = instance.get_anchors();
 
         size_t anchors_num = anchors.size();
 
-        auto& cls_scores = instance.dep_memory(proposal_inst::cls_scores_index);
-        auto& bbox_pred = instance.dep_memory(proposal_inst::bbox_pred_index);
+        auto cls_scores = instance.dep_memory_ptr(proposal_inst::cls_scores_index);
+        auto bbox_pred = instance.dep_memory_ptr(proposal_inst::bbox_pred_index);
 
         bool swap_xy = instance.argument.swap_xy;
         bool initial_clip = instance.argument.initial_clip;
@@ -277,14 +280,14 @@ struct proposal_gpu : typed_primitive_impl<proposal> {
         bool for_deformable = instance.argument.for_deformable;
 
         // feat map sizes
-        const auto& score_size = cls_scores.get_layout().size;
+        const auto& score_size = cls_scores->get_layout().size;
         int fm_h = score_size.spatial[1];
         int fm_w = score_size.spatial[0];
 
         int fm_sz = fm_w * fm_h;
 
-        mem_lock<dtype> cls_scores_ptr{cls_scores};
-        mem_lock<dtype> bbox_pred_ptr{bbox_pred};
+        mem_lock<dtype> cls_scores_ptr{cls_scores, stream};
+        mem_lock<dtype> bbox_pred_ptr{bbox_pred, stream};
         const dtype* cls_scores_mem = cls_scores_ptr.data();
         const dtype* bbox_pred_mem = bbox_pred_ptr.data();
 
@@ -347,9 +350,9 @@ struct proposal_gpu : typed_primitive_impl<proposal> {
                                                  instance.argument.post_nms_topn,
                                                  coordinates_offset);
 
-            auto& output = instance.output_memory();
+            auto output = instance.output_memory_ptr();
 
-            mem_lock<dtype> output_ptr{output};
+            mem_lock<dtype> output_ptr{output, stream};
             dtype* top_data = output_ptr.data() + n * instance.argument.post_nms_topn * 5;
 
             dtype* top_data_prob = proposal_prob_ptr == nullptr ? nullptr : proposal_prob_ptr + n * instance.argument.post_nms_topn;
@@ -386,17 +389,19 @@ struct proposal_gpu : typed_primitive_impl<proposal> {
         }
     }
 
-    event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& events, proposal_inst& instance) override {
+    event::ptr execute_impl(const std::vector<event::ptr>& events, proposal_inst& instance) override {
         for (auto& a : events) {
             a->wait();
         }
 
-        auto ev = instance.get_network().get_engine().create_user_event(instance.get_network().get_id(), false);
+        auto& stream = instance.get_network().get_stream();
+
+        auto ev = instance.get_network().get_stream().create_user_event(false);
         im_info_t im_info;
         if (instance.dep_memory(proposal_inst::image_info_index).get_layout().data_type == data_types::f16) {
-            read_image_info<data_type_to_type<data_types::f16>::type>(instance, im_info);
+            read_image_info<data_type_to_type<data_types::f16>::type>(stream, instance, im_info);
         } else {
-            read_image_info<data_type_to_type<data_types::f32>::type>(instance, im_info);
+            read_image_info<data_type_to_type<data_types::f32>::type>(stream, instance, im_info);
         }
 
         if (instance.dep_memory(proposal_inst::cls_scores_index).get_layout().data_type !=
@@ -404,19 +409,19 @@ struct proposal_gpu : typed_primitive_impl<proposal> {
             throw std::runtime_error("clDNN: proposal primitive doesn't support mixed bbox and scores types");
 
         if (instance.dependencies().size() == 4) {
-            auto &proposal_probabilities = instance.dep_memory(proposal_inst::proposal_probabilities_out);
+            auto proposal_probabilities = instance.dep_memory_ptr(proposal_inst::proposal_probabilities_out);
             if (instance.dep_memory(proposal_inst::cls_scores_index).get_layout().data_type == data_types::f16) {
-                mem_lock<data_type_to_type<data_types::f16>::type> proposal_prob_ptr{proposal_probabilities};
-                execute<data_type_to_type<data_types::f16>::type>(instance, im_info, proposal_prob_ptr.data());
+                mem_lock<data_type_to_type<data_types::f16>::type> proposal_prob_ptr{proposal_probabilities, stream};
+                execute<data_type_to_type<data_types::f16>::type>(stream, instance, im_info, proposal_prob_ptr.data());
             } else {
-                mem_lock<data_type_to_type<data_types::f32>::type> proposal_prob_ptr{proposal_probabilities};
-                execute<data_type_to_type<data_types::f32>::type>(instance, im_info, proposal_prob_ptr.data());
+                mem_lock<data_type_to_type<data_types::f32>::type> proposal_prob_ptr{proposal_probabilities, stream};
+                execute<data_type_to_type<data_types::f32>::type>(stream, instance, im_info, proposal_prob_ptr.data());
             }
         } else {
             if (instance.dep_memory(proposal_inst::cls_scores_index).get_layout().data_type == data_types::f16) {
-                execute<data_type_to_type<data_types::f16>::type>(instance, im_info);
+                execute<data_type_to_type<data_types::f16>::type>(stream, instance, im_info);
             } else {
-                execute<data_type_to_type<data_types::f32>::type>(instance, im_info);
+                execute<data_type_to_type<data_types::f32>::type>(stream, instance, im_info);
             }
         }
 
@@ -424,6 +429,8 @@ struct proposal_gpu : typed_primitive_impl<proposal> {
         return ev;
     }
 
+    void init_kernels() override {}
+
     static primitive_impl* create(const proposal_node& arg) {
         const layout& l = arg.image_info().get_output_layout();
         const size_t count = l.size.feature[0] == 1 ? static_cast<size_t>(l.size.batch[0]) : static_cast<size_t>(l.size.feature[0]);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/push_pop_map.h b/inference-engine/thirdparty/clDNN/src/gpu/push_pop_map.h
deleted file mode 100644
index c4fc57ed14c1c9..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/gpu/push_pop_map.h
+++ /dev/null
@@ -1,47 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include <map>
-#include <mutex>
-#include <memory>
-#include <utility>
-#include <functional>
-
-namespace cldnn {
-namespace gpu {
-
-template <typename Key,
-          typename Type,
-          class Traits = std::less<Key>,
-          class Allocator = std::allocator<std::pair<const Key, Type>>>
-class push_pop_map {
-    std::mutex _mutex;
-    std::map<Key, Type, Traits, Allocator> _map;
-
-public:
-    void push(const Key& key, Type value) {
-        std::lock_guard<std::mutex> lock{_mutex};
-        _map.insert({key, std::move(value)});
-    }
-
-    Type pop(const Key& key) {
-        std::lock_guard<std::mutex> lock{_mutex};
-        auto it = _map.find(key);
-        if (it == _map.end())
-            throw std::out_of_range("Invalud push_pop_map<K, T> key");
-        auto x = std::move(it->second);
-        _map.erase(it);
-        return std::move(x);
-    }
-
-    bool empty() {
-        std::lock_guard<std::mutex> lock{_mutex};
-        return _map.empty();
-    }
-};
-
-}  // namespace gpu
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/pyramid_roi_align_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/pyramid_roi_align_gpu.cpp
index 3b16e678c2ceb0..8259257d92c9aa 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/pyramid_roi_align_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/pyramid_roi_align_gpu.cpp
@@ -7,7 +7,7 @@
 #include "kernel_selector_helper.h"
 #include "pyramid_roi_align/pyramid_roi_align_kernel_selector.h"
 #include "pyramid_roi_align/pyramid_roi_align_kernel_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "pyramid_roi_align_inst.h"
 #include "network_impl.h"
 
@@ -20,6 +20,10 @@ struct pyramid_roi_align_gpu : typed_primitive_gpu_impl<pyramid_roi_align> {
     using parent = typed_primitive_gpu_impl<pyramid_roi_align>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<pyramid_roi_align_gpu>(*this);
+    }
+
     static primitive_impl* create(const pyramid_roi_align_node& arg) {
         auto prim = arg.get_primitive();
         auto params = get_default_params<kernel_selector::PyramidROIAlign_params>(arg, 1);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/quantize_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/quantize_gpu.cpp
index aefaa20eb8cc06..627a2e1799ac0d 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/quantize_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/quantize_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "quantize/quantize_kernel_selector.h"
 #include "quantize/quantize_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -19,23 +19,26 @@ struct quantize_gpu : typed_primitive_gpu_impl<quantize> {
     using parent = typed_primitive_gpu_impl<quantize>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<quantize_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<quantize>& instance,
-                                                int32_t) const override {
-        kernel::kernel_arguments_data args;
+    kernel_arguments_data get_arguments(typed_primitive_inst<quantize>& instance, int32_t) const override {
+        kernel_arguments_data args;
 
         for (size_t i = 0; i < instance.inputs_memory_count(); i++) {
-            args.inputs.push_back((memory_impl::cptr) &instance.input_memory(i));
+            args.inputs.push_back(instance.input_memory_ptr(i));
         }
         if (instance.node.get_scale_shift_opt()) {
             if (instance.node.get_dependencies().size() == 9) {
-                args.inputs.push_back((memory_impl::cptr) &instance.dep_memory(5));
-                args.inputs.push_back((memory_impl::cptr) &instance.dep_memory(6));
-                args.inputs.push_back((memory_impl::cptr) &instance.dep_memory(7));
-                args.inputs.push_back((memory_impl::cptr) &instance.dep_memory(8));
+                args.inputs.push_back(instance.dep_memory_ptr(5));
+                args.inputs.push_back(instance.dep_memory_ptr(6));
+                args.inputs.push_back(instance.dep_memory_ptr(7));
+                args.inputs.push_back(instance.dep_memory_ptr(8));
             }
         }
-        args.output = (memory_impl::cptr) &instance.output_memory();
+        args.output = instance.output_memory_ptr();
         return args;
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/reduce_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/reduce_gpu.cpp
index 2bf2f7535b2e1e..5ad63259500d3f 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/reduce_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/reduce_gpu.cpp
@@ -9,7 +9,7 @@
 #include "reduce/reduce_kernel_selector.h"
 #include "reduce/reduce_kernel_ref.h"
 #include "reduce/reduce_kernel_b_fs_yx_fsv16.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "data_inst.h"
 
 using namespace cldnn;
@@ -53,6 +53,10 @@ struct reduce_gpu : typed_primitive_gpu_impl<reduce> {
     using parent = typed_primitive_gpu_impl<reduce>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<reduce_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const reduce_node& arg) {
         auto reduce_params = get_default_params<kernel_selector::reduce_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/region_yolo_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/region_yolo_gpu.cpp
index 5bbe70ce8ed00e..e5518f66cb8911 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/region_yolo_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/region_yolo_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "region_yolo/region_yolo_kernel_selector.h"
 #include "region_yolo/region_yolo_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -17,6 +17,10 @@ struct region_yolo_gpu : typed_primitive_gpu_impl<region_yolo> {
     using parent = typed_primitive_gpu_impl<region_yolo>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<region_yolo_gpu>(*this);
+    }
+
     static primitive_impl* create(const region_yolo_node& arg) {
         auto ry_params = get_default_params<kernel_selector::region_yolo_params>(arg);
         auto ry_optional_params =
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/register_gpu.hpp b/inference-engine/thirdparty/clDNN/src/gpu/register_gpu.hpp
index 6bba0170822970..d6d2db15d6b617 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/register_gpu.hpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/register_gpu.hpp
@@ -5,69 +5,69 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 
-#include "api/activation.hpp"
-#include "api/arg_max_min.hpp"
-#include "api/average_unpooling.hpp"
-#include "api/batch_to_space.hpp"
-#include "api/binary_convolution.hpp"
-#include "api/border.hpp"
-#include "api/broadcast.hpp"
-#include "api/concatenation.hpp"
-#include "api/condition.hpp"
-#include "api/convolution.hpp"
-#include "api/crop.hpp"
-#include "api/custom_gpu_primitive.hpp"
-#include "api/data.hpp"
-#include "api/deconvolution.hpp"
-#include "api/depth_to_space.hpp"
-#include "api/detection_output.hpp"
-#include "api/eltwise.hpp"
-#include "api/fully_connected.hpp"
-#include "api/gather.hpp"
-#include "api/gather_nd.hpp"
-#include "api/gemm.hpp"
-#include "api/input_layout.hpp"
-#include "api/lrn.hpp"
-#include "api/lstm.hpp"
-#include "api/lstm_dynamic.hpp"
-#include "api/max_unpooling.hpp"
-#include "api/mutable_data.hpp"
-#include "api/mvn.hpp"
-#include "api/normalize.hpp"
-#include "api/one_hot.hpp"
-#include "api/permute.hpp"
-#include "api/pooling.hpp"
-#include "api/prior_box.hpp"
-#include "api/proposal.hpp"
-#include "api/pyramid_roi_align.hpp"
-#include "api/quantize.hpp"
-#include "api/reduce.hpp"
-#include "api/region_yolo.hpp"
-#include "api/reorder.hpp"
-#include "api/reorg_yolo.hpp"
-#include "api/reshape.hpp"
-#include "api/reverse_sequence.hpp"
-#include "api/roi_pooling.hpp"
-#include "api/scale.hpp"
-#include "api/scatter_update.hpp"
-#include "api/scatter_elements_update.hpp"
-#include "api/scatter_nd_update.hpp"
-#include "api/select.hpp"
-#include "api/shuffle_channels.hpp"
-#include "api/softmax.hpp"
-#include "api/space_to_batch.hpp"
-#include "api/strided_slice.hpp"
-#include "api/tile.hpp"
-#include "api/resample.hpp"
-#include "api/gather_tree.hpp"
-#include "api_extension/fused_conv_eltwise.hpp"
-#include "api_extension/lstm_dynamic_input.hpp"
-#include "api_extension/lstm_dynamic_timeloop.hpp"
+#include "cldnn/primitives/activation.hpp"
+#include "cldnn/primitives/arg_max_min.hpp"
+#include "cldnn/primitives/average_unpooling.hpp"
+#include "cldnn/primitives/batch_to_space.hpp"
+#include "cldnn/primitives/binary_convolution.hpp"
+#include "cldnn/primitives/border.hpp"
+#include "cldnn/primitives/broadcast.hpp"
+#include "cldnn/primitives/concatenation.hpp"
+#include "cldnn/primitives/condition.hpp"
+#include "cldnn/primitives/convolution.hpp"
+#include "cldnn/primitives/crop.hpp"
+#include "cldnn/primitives/custom_gpu_primitive.hpp"
+#include "cldnn/primitives/data.hpp"
+#include "cldnn/primitives/deconvolution.hpp"
+#include "cldnn/primitives/depth_to_space.hpp"
+#include "cldnn/primitives/detection_output.hpp"
+#include "cldnn/primitives/eltwise.hpp"
+#include "cldnn/primitives/fully_connected.hpp"
+#include "cldnn/primitives/gather.hpp"
+#include "cldnn/primitives/gather_nd.hpp"
+#include "cldnn/primitives/gemm.hpp"
+#include "cldnn/primitives/input_layout.hpp"
+#include "cldnn/primitives/lrn.hpp"
+#include "cldnn/primitives/lstm.hpp"
+#include "cldnn/primitives/lstm_dynamic.hpp"
+#include "cldnn/primitives/max_unpooling.hpp"
+#include "cldnn/primitives/mutable_data.hpp"
+#include "cldnn/primitives/mvn.hpp"
+#include "cldnn/primitives/normalize.hpp"
+#include "cldnn/primitives/one_hot.hpp"
+#include "cldnn/primitives/permute.hpp"
+#include "cldnn/primitives/pooling.hpp"
+#include "cldnn/primitives/prior_box.hpp"
+#include "cldnn/primitives/proposal.hpp"
+#include "cldnn/primitives/pyramid_roi_align.hpp"
+#include "cldnn/primitives/quantize.hpp"
+#include "cldnn/primitives/reduce.hpp"
+#include "cldnn/primitives/region_yolo.hpp"
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/reorg_yolo.hpp"
+#include "cldnn/primitives/reshape.hpp"
+#include "cldnn/primitives/reverse_sequence.hpp"
+#include "cldnn/primitives/roi_pooling.hpp"
+#include "cldnn/primitives/scale.hpp"
+#include "cldnn/primitives/scatter_update.hpp"
+#include "cldnn/primitives/scatter_elements_update.hpp"
+#include "cldnn/primitives/scatter_nd_update.hpp"
+#include "cldnn/primitives/select.hpp"
+#include "cldnn/primitives/shuffle_channels.hpp"
+#include "cldnn/primitives/softmax.hpp"
+#include "cldnn/primitives/space_to_batch.hpp"
+#include "cldnn/primitives/strided_slice.hpp"
+#include "cldnn/primitives/tile.hpp"
+#include "cldnn/primitives/resample.hpp"
+#include "cldnn/primitives/gather_tree.hpp"
+#include "cldnn/primitives/fused_conv_eltwise.hpp"
+#include "cldnn/primitives/lstm_dynamic_input.hpp"
+#include "cldnn/primitives/lstm_dynamic_timeloop.hpp"
+#include "cldnn/primitives/non_max_suppression.hpp"
+#include "cldnn/primitives/grn.hpp"
+#include "cldnn/primitives/ctc_greedy_decoder.hpp"
+#include "cldnn/primitives/loop.hpp"
 #include "generic_layer.hpp"
-#include "api/non_max_suppression.hpp"
-#include "api/grn.hpp"
-#include "api/ctc_greedy_decoder.hpp"
-#include "api/loop.hpp"
 
 
 namespace cldnn { namespace gpu {
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/reorder_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/reorder_gpu.cpp
index e3961faaa000dc..93fafb1d14cb1d 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/reorder_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/reorder_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "reorder/reorder_kernel_selector.h"
 #include "reorder/reorder_kernel_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -17,20 +17,24 @@ struct reorder_gpu : typed_primitive_gpu_impl<reorder> {
     using parent = typed_primitive_gpu_impl<reorder>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<reorder_gpu>(*this);
+    }
+
 protected:
     bool optimized_out(reorder_inst& instance) const override {
         return parent::optimized_out(instance) || _outer.can_be_optimized();
     }
 
-    kernel::kernel_arguments_data get_arguments(reorder_inst& instance, int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
+    kernel_arguments_data get_arguments(reorder_inst& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
         auto input = &instance.input_memory();
         auto input_layout = input->get_layout();
         if (_outer.has_mean()) {
             if (input_layout.format == cldnn::format::nv12) {
-                args.bias = (memory_impl::cptr) &instance.mean_nv12_memory();
+                args.bias = instance.mean_nv12_memory();
             } else {
-                args.bias = (memory_impl::cptr) &instance.mean_memory();
+                args.bias = instance.mean_memory();
             }
         }
         return args;
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/reorg_yolo_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/reorg_yolo_gpu.cpp
index b8badf3584e524..5cc78810ad4c8a 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/reorg_yolo_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/reorg_yolo_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "reorg_yolo/reorg_yolo_kernel_selector.h"
 #include "reorg_yolo/reorg_yolo_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -17,6 +17,10 @@ struct reorg_yolo_gpu : typed_primitive_gpu_impl<reorg_yolo> {
     using parent = typed_primitive_gpu_impl<reorg_yolo>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<reorg_yolo_gpu>(*this);
+    }
+
     static primitive_impl* create(const reorg_yolo_node& arg) {
         auto ry_params = get_default_params<kernel_selector::reorg_yolo_params>(arg);
         auto ry_optional_params =
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/resample_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/resample_gpu.cpp
index 18697841ac478d..1ec432673ee4b0 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/resample_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/resample_gpu.cpp
@@ -5,7 +5,7 @@
 #include "resample_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "kernel_selector/core/actual_kernels/resample/resample_kernel_selector.h"
 #include "kernel_selector/core/actual_kernels/resample/resample_kernel_base.h"
@@ -100,6 +100,10 @@ struct resample_gpu : typed_primitive_gpu_impl<resample> {
     using parent = typed_primitive_gpu_impl<resample>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<resample_gpu>(*this);
+    }
+
     static primitive_impl* create(const resample_node& arg) {
         auto us_params = get_default_params<kernel_selector::resample_params>(arg);
         auto us_optional_params =
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/reshape_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/reshape_gpu.cpp
index 04a89b42002b03..5a78f82cca122e 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/reshape_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/reshape_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "reshape/reshape_kernel_ref.h"
 #include "reshape/reshape_kernel_selector.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -17,6 +17,10 @@ struct reshape_gpu : public typed_primitive_gpu_impl<reshape> {
     using parent = typed_primitive_gpu_impl<reshape>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<reshape_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(reshape_node const& arg) {
         if (arg.can_be_optimized()) {
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/reverse_sequence_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/reverse_sequence_gpu.cpp
index 01a03b9a251460..01d6aaf89334dd 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/reverse_sequence_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/reverse_sequence_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "reverse_sequence/reverse_sequence_kernel_selector.h"
 #include "reverse_sequence/reverse_sequence_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -18,6 +18,10 @@ struct reverse_sequence_gpu : typed_primitive_gpu_impl<reverse_sequence> {
     using parent = typed_primitive_gpu_impl<reverse_sequence>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<reverse_sequence_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const reverse_sequence_node& arg) {
         auto reverse_sequence_params = get_default_params<kernel_selector::reverse_sequence_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/roi_pooling_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/roi_pooling_gpu.cpp
index c8f1fde2290976..fef441b15b8ce1 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/roi_pooling_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/roi_pooling_gpu.cpp
@@ -5,7 +5,7 @@
 #include "roi_pooling_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "roi_pooling/roi_pooling_kernel_selector.h"
 #include "roi_pooling/roi_pooling_kernel_ref.h"
@@ -37,20 +37,23 @@ struct roi_pooling_gpu : typed_primitive_gpu_impl<roi_pooling> {
     using parent = typed_primitive_gpu_impl<roi_pooling>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<roi_pooling_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<roi_pooling>& instance,
-                                                        int32_t) const override {
-        kernel::kernel_arguments_data args;
+    kernel_arguments_data get_arguments(typed_primitive_inst<roi_pooling>& instance, int32_t) const override {
+        kernel_arguments_data args;
 
         if (instance.argument.mode == pooling_mode::deformable_bilinear && !instance.argument.no_trans)
             args.inputs = {
-                (memory_impl::cptr) &instance.input_memory(),
-                (memory_impl::cptr) &instance.rois_memory(),
-                (memory_impl::cptr) &instance.trans_memory()};
+                instance.input_memory_ptr(),
+                instance.rois_memory(),
+                instance.trans_memory()};
         else
-            args.inputs = {(memory_impl::cptr) &instance.input_memory(), (memory_impl::cptr) &instance.rois_memory()};
+            args.inputs = {instance.input_memory_ptr(), instance.rois_memory()};
 
-        args.output = (memory_impl::cptr) &instance.output_memory();
+        args.output = instance.output_memory_ptr();
 
         return args;
     }
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/scale_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/scale_gpu.cpp
index 0f6a91b3f546b8..cfdf1a3085cc41 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/scale_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/scale_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "eltwise/eltwise_kernel_selector.h"
 #include "eltwise/eltwise_kernel_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -19,14 +19,18 @@ struct scale_gpu : typed_primitive_gpu_impl<scale> {
     using parent = typed_primitive_gpu_impl<scale>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<scale_gpu>(*this);
+    }
+
 protected:
-    kernel::kernel_arguments_data get_arguments(typed_primitive_inst<scale>& instance, int32_t split) const override {
-        kernel::kernel_arguments_data args = parent::get_arguments(instance, split);
-        args.inputs = {(memory_impl::cptr) &instance.input_memory(), (memory_impl::cptr) &instance.scale_memory()};
-        args.output = (memory_impl::cptr) &instance.output_memory();
+    kernel_arguments_data get_arguments(typed_primitive_inst<scale>& instance, int32_t split) const override {
+        kernel_arguments_data args = parent::get_arguments(instance, split);
+        args.inputs = {instance.input_memory_ptr(), instance.scale_memory()};
+        args.output = instance.output_memory_ptr();
 
         if (_outer.bias_term()) {
-            args.inputs.push_back((memory_impl::cptr) &instance.bias_memory());
+            args.inputs.push_back(instance.bias_memory());
         }
         return args;
     }
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/scatter_elements_update_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/scatter_elements_update_gpu.cpp
index 00d756142fddf5..85e482e02e0dff 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/scatter_elements_update_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/scatter_elements_update_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "scatter_update/scatter_elements_update_kernel_selector.h"
 #include "scatter_update/scatter_elements_update_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -38,6 +38,10 @@ struct scatter_elements_update_gpu : typed_primitive_gpu_impl<scatter_elements_u
     using parent = typed_primitive_gpu_impl<scatter_elements_update>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<scatter_elements_update_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const scatter_elements_update_node& arg) {
         auto scatter_elements_update_params = get_default_params<kernel_selector::scatter_elements_update_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/scatter_nd_update_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/scatter_nd_update_gpu.cpp
index c9014120b51d65..445361bf5b85b7 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/scatter_nd_update_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/scatter_nd_update_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "scatter_update/scatter_nd_update_kernel_selector.h"
 #include "scatter_update/scatter_nd_update_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -19,6 +19,10 @@ struct scatter_nd_update_gpu : typed_primitive_gpu_impl<scatter_nd_update> {
     using parent = typed_primitive_gpu_impl<scatter_nd_update>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<scatter_nd_update_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const scatter_nd_update_node& arg) {
         auto scatter_nd_update_params = get_default_params<kernel_selector::scatter_nd_update_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/scatter_update_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/scatter_update_gpu.cpp
index 8bf3e7122a270a..10629a080909eb 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/scatter_update_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/scatter_update_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "scatter_update/scatter_update_kernel_selector.h"
 #include "scatter_update/scatter_update_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -38,6 +38,10 @@ struct scatter_update_gpu : typed_primitive_gpu_impl<scatter_update> {
     using parent = typed_primitive_gpu_impl<scatter_update>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<scatter_update_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const scatter_update_node& arg) {
         auto scatter_update_params = get_default_params<kernel_selector::scatter_update_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/select_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/select_gpu.cpp
index 88d088113054ac..09572da455a535 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/select_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/select_gpu.cpp
@@ -5,7 +5,7 @@
 #include "select_inst.h"
 #include "primitive_gpu_base.h"
 #include "implementation_map.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
 #include "select/select_kernel_selector.h"
 #include "select/select_kernel_base.h"
@@ -17,6 +17,10 @@ struct select_gpu : typed_primitive_gpu_impl<select> {
     using parent = typed_primitive_gpu_impl<select>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<select_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const select_node& arg) {
         auto select_params = get_default_params<kernel_selector::select_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/shuffle_channels_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/shuffle_channels_gpu.cpp
index 95a7705aa3a61f..b7a29684772e9b 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/shuffle_channels_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/shuffle_channels_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "shuffle_channels/shuffle_channels_kernel_selector.h"
 #include "shuffle_channels/shuffle_channels_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -19,6 +19,10 @@ struct shuffle_channels_gpu : typed_primitive_gpu_impl<shuffle_channels> {
     using parent = typed_primitive_gpu_impl<shuffle_channels>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<shuffle_channels_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const shuffle_channels_node& arg) {
         auto shuffle_channels_params = get_default_params<kernel_selector::shuffle_channels_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/softmax_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/softmax_gpu.cpp
index 70cdf63d3cff4d..868ca8913dde97 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/softmax_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/softmax_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "softmax/softmax_kernel_selector.h"
 #include "softmax/softmax_kernel_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 namespace cldnn {
 namespace gpu {
@@ -17,6 +17,10 @@ struct softmax_gpu : typed_primitive_gpu_impl<softmax> {
     using parent = typed_primitive_gpu_impl<softmax>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<softmax_gpu>(*this);
+    }
+
     static primitive_impl* create(const softmax_node& arg) {
         auto sm_params = get_default_params<kernel_selector::softmax_params>(arg);
         auto sm_optional_params =
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/space_to_batch_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/space_to_batch_gpu.cpp
index 5961ee1858cb5f..ebc97b591f28ea 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/space_to_batch_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/space_to_batch_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "space_to_batch/space_to_batch_kernel_selector.h"
 #include "space_to_batch/space_to_batch_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "data_inst.h"
 #include <vector>
 
@@ -20,6 +20,10 @@ struct space_to_batch_gpu : typed_primitive_gpu_impl<space_to_batch> {
     using parent = typed_primitive_gpu_impl<space_to_batch>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<space_to_batch_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const space_to_batch_node& arg) {
         auto space_to_batch_params = get_default_params<kernel_selector::space_to_batch_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/space_to_depth_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/space_to_depth_gpu.cpp
index 0866cebd143b0f..bf92acea8b4211 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/space_to_depth_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/space_to_depth_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "space_to_depth/space_to_depth_kernel_selector.h"
 #include "space_to_depth/space_to_depth_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -18,6 +18,10 @@ struct space_to_depth_gpu : typed_primitive_gpu_impl<space_to_depth> {
     using parent = typed_primitive_gpu_impl<space_to_depth>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<space_to_depth_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const space_to_depth_node& arg) {
         auto space_to_depth_params = get_default_params<kernel_selector::space_to_depth_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/strided_slice_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/strided_slice_gpu.cpp
index d78a08eef69fd5..d1ddf2d4d69db2 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/strided_slice_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/strided_slice_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "strided_slice/strided_slice_kernel_ref.h"
 #include "strided_slice/strided_slice_kernel_selector.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "data_inst.h"
 #include <vector>
 
@@ -21,6 +21,10 @@ struct strided_slice_gpu : typed_primitive_gpu_impl<strided_slice> {
     using parent = typed_primitive_gpu_impl<strided_slice>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<strided_slice_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const strided_slice_node& arg) {
         auto params = get_default_params<kernel_selector::strided_slice_params>(arg);
@@ -30,21 +34,22 @@ struct strided_slice_gpu : typed_primitive_gpu_impl<strided_slice> {
         // Getting data from constant inputs. There are 3 args: Begin, End, Stride
         for (size_t i = 1; i < arg.get_dependencies().size(); ++i) {
             auto& input = arg.get_dependency(i).as<data>();
-            auto& mem = input.get_attached_memory();
+            auto mem = input.get_attached_memory_ptr();
             std::vector<int32_t> sizes;
             if (input.get_output_layout().data_type == cldnn::data_types::i64) {
-                int64_t* data = static_cast<int64_t*>(mem.lock());
+                mem_lock<int64_t> lock{mem, arg.get_program().get_stream()};
+                int64_t* data = lock.data();
                 std::vector<int64_t> sizes_i64 = std::vector<int64_t>(data, data + input.get_output_layout().count());
                 sizes.resize(sizes_i64.size());
                 for (size_t j = 0; j < sizes.size(); j++)
                     sizes[j] = static_cast<int32_t>(sizes_i64[j]);
             } else {
-                int32_t* data = static_cast<int32_t*>(mem.lock());
+                mem_lock<int32_t> lock{mem, arg.get_program().get_stream()};
+                int32_t* data = lock.data();
                 sizes = std::vector<int32_t>(data, data + input.get_output_layout().count());
             }
             pad_vector_to_size(sizes, dims_num, i != 1);  // for "begin" completion used 0 value, for other - 1
             params.striding_params.push_back(sizes);
-            mem.unlock();
         }
 
         params.end_mask = arg.get_primitive()->end_mask;
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/tile_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/tile_gpu.cpp
index 02f95954f58e79..d508e2bea312dd 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/tile_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/tile_gpu.cpp
@@ -8,7 +8,7 @@
 #include "kernel_selector_helper.h"
 #include "tile/tile_kernel_selector.h"
 #include "tile/tile_kernel_ref.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 using namespace cldnn;
 
@@ -19,6 +19,10 @@ struct tile_gpu : typed_primitive_gpu_impl<tile> {
     using parent = typed_primitive_gpu_impl<tile>;
     using parent::parent;
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<tile_gpu>(*this);
+    }
+
 public:
     static primitive_impl* create(const tile_node& arg) {
         auto tile_params = get_default_params<kernel_selector::tile_params>(arg);
diff --git a/inference-engine/thirdparty/clDNN/src/gpu/wait_for_events_gpu.cpp b/inference-engine/thirdparty/clDNN/src/gpu/wait_for_events_gpu.cpp
index 38a822dca525d8..00c74080a38400 100644
--- a/inference-engine/thirdparty/clDNN/src/gpu/wait_for_events_gpu.cpp
+++ b/inference-engine/thirdparty/clDNN/src/gpu/wait_for_events_gpu.cpp
@@ -10,7 +10,6 @@
 #include "register_gpu.hpp"
 
 #include "network_impl.h"
-#include "events_waiter.h"
 #include <vector>
 
 namespace cldnn {
@@ -20,13 +19,16 @@ class wait_for_events_gpu : public primitive_impl {
 public:
     explicit wait_for_events_gpu(const program_node& /*node*/) {}
 
+    std::unique_ptr<primitive_impl> clone() const override {
+        return make_unique<wait_for_events_gpu>(*this);
+    }
+
+    void init_kernels() override {}
     void set_arguments(primitive_inst& /*instance*/) override {}
-    void cleanup(primitive_inst& /*instance*/) override {}
 
-    event_impl::ptr execute(const std::vector<event_impl::ptr>& events, primitive_inst& instance) override {
-        uint32_t net_id = instance.get_network().get_id();
-        events_waiter events_waiter(instance.get_network().get_engine().get_context());
-        return events_waiter.run(net_id, events);
+    event::ptr execute(const std::vector<event::ptr>& events, primitive_inst& instance) override {
+        auto& stream = instance.get_network().get_stream();
+        return stream.enqueue_marker(events);
     }
 
     bool validate(const primitive_inst&) const override { return true; }
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/add_required_reorders.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/add_required_reorders.cpp
index 7d61c77158d9d0..249325eaaccf75 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/add_required_reorders.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/add_required_reorders.cpp
@@ -49,7 +49,7 @@ void add_required_reorders::run(program_impl& p) {
         auto& usr = *usr_itr++;
         if (usr->get_dependencies().size() == 0)
             continue;  // only nodes with dependencies
-        if (usr->is_type<internal_primitive>() || usr->is_type<data>())
+        if (usr->is_type<data>())
             continue;
         if (usr->type()->does_an_implementation_exist(p.get_engine(), *usr))
             continue;
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/basic_memory_dependencies.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/basic_memory_dependencies.cpp
index 540a81707aae69..1dd8b0af12adc1 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/basic_memory_dependencies.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/basic_memory_dependencies.cpp
@@ -9,7 +9,7 @@
 #include "layout_optimizer.h"
 #include "program_impl.h"
 #include "program_helpers.h"
-#include "cldnn_itt.h"
+#include "runtime/cldnn_itt.hpp"
 #include <vector>
 #include <memory>
 #include <list>
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/calculate_prior_boxes.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/calculate_prior_boxes.cpp
index 387512cae04c2f..00f74be859345a 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/calculate_prior_boxes.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/calculate_prior_boxes.cpp
@@ -25,9 +25,8 @@ void calculate_prior_boxes::run(program_impl& p) {
         p.remove_connection(pb_node.input(), pb_node);
 
         auto result = pb_node.get_result_buffer();
-        auto cpp_mem = memory(result.detach());
 
-        auto& data_node = p.get_or_create(std::make_shared<data>("_cldnn_tmp_" + pb_node.id() + "_result", cpp_mem));
+        auto& data_node = p.get_or_create(std::make_shared<data>("_cldnn_tmp_" + pb_node.id() + "_result", result));
         p.replace(pb_node, data_node);
     }
 }
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/compile_graph.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/compile_graph.cpp
index fe236c180d8c41..8c3335aecbb5bb 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/compile_graph.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/compile_graph.cpp
@@ -5,19 +5,18 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
 #include "pass_manager.h"
-#include "internal_primitive.h"
 #include "data_inst.h"
 #include "mutable_data_inst.h"
 #include "program_node.h"
-#include "engine_impl.h"
-#include "cldnn_itt.h"
+#include "cldnn/runtime/engine.hpp"
+#include "runtime/cldnn_itt.hpp"
 
 using namespace cldnn;
 
 void compile_graph::run(program_impl& p) {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "CLDNN::pass::CompileGraph");
     for (auto& node : p.get_processing_order()) {
-        if (!node->is_type<internal_primitive>() && !node->is_type<data>()) {
+        if (!node->is_type<data>()) {
             node->get_output_layout();
             if (!node->is_type<data>() && !(node->is_type<mutable_data>() && node->get_dependencies().empty())) {
                 node->selected_impl = node->type()->choose_impl(p.get_engine(), *node);
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/concat_input_order.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/concat_input_order.cpp
index 2e949169e3d35b..9acd92fc7266bb 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/concat_input_order.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/concat_input_order.cpp
@@ -9,7 +9,7 @@
 #include "convolution_inst.h"
 #include "fully_connected_inst.h"
 #include "data_inst.h"
-#include "memory_impl.h"
+#include "cldnn/runtime/memory.hpp"
 #include "program_impl.h"
 
 #include <vector>
@@ -21,7 +21,7 @@ namespace {
 
 using shuffle_range = std::pair<int32_t, int32_t>;
 
-bool can_shuffle_features(program_node& node) {
+bool can_shuffle_features(program_node& node, stream& stream) {
     if (node.is_type<convolution>()) {
         auto& conv_node = node.as<convolution>();
         auto& wei_node = conv_node.weights();
@@ -46,7 +46,7 @@ bool can_shuffle_features(program_node& node) {
     if (pass_through) {
         // Primitives that are feature order invariant, pass-through shuffled features to users
         for (auto& user : node.get_users()) {
-            if (!can_shuffle_features(*user))
+            if (!can_shuffle_features(*user, stream))
                 return false;
         }
         return true;
@@ -55,17 +55,19 @@ bool can_shuffle_features(program_node& node) {
     return false;
 }
 
-void shuffle_weights(data_node& node, const std::vector<shuffle_range>& ranges) {
+void shuffle_weights(data_node& node, const std::vector<shuffle_range>& ranges, stream& stream) {
     // Correct for shuffled features by shuffling input feature dimension in weights.
     // This allows to restore correct feature order on output and only changes calculation order.
     auto wei_layout = node.get_output_layout();
-    auto& old_weights_memory = node.get_attached_memory();
+    auto old_weights_memory = node.get_attached_memory_ptr();
     bool need_reset = static_cast<bool>(wei_layout.data_padding) || wei_layout.format.is_blocked();
-    auto new_weights_memory = old_weights_memory.get_engine()->allocate_memory(wei_layout, old_weights_memory.get_net_id(), need_reset);
+    auto new_weights_memory = old_weights_memory->get_engine()->allocate_memory(wei_layout, old_weights_memory->get_allocation_type(), need_reset);
 
     auto bytes_per_elem = data_type_traits::size_of(wei_layout.data_type);
-    auto old_ptr = static_cast<char*>(old_weights_memory.lock());
-    auto new_ptr = static_cast<char*>(new_weights_memory->lock());
+    mem_lock<uint8_t> old_weights_memory_lock{old_weights_memory, stream};
+    mem_lock<uint8_t> new_weights_memory_lock{new_weights_memory, stream};
+    auto old_ptr = old_weights_memory_lock.data();
+    auto new_ptr = new_weights_memory_lock.data();
     for (int32_t ofi = 0; ofi < wei_layout.size.batch[0]; ++ofi) {
         int32_t new_ifi = 0;
         for (auto& range : ranges) {
@@ -88,23 +90,21 @@ void shuffle_weights(data_node& node, const std::vector<shuffle_range>& ranges)
             }
         }
     }
-    old_weights_memory.unlock();
-    new_weights_memory->unlock();
 
-    node.attach_memory(*new_weights_memory, false);
+    node.attach_memory(new_weights_memory, false);
 }
 
-void shuffle_features(program_node& node, const std::vector<shuffle_range>& ranges) {
+void shuffle_features(program_node& node, const std::vector<shuffle_range>& ranges, stream& stream) {
     if (node.is_type<convolution>()) {
         auto& conv = node.as<convolution>();
-        shuffle_weights(conv.weights().as<data>(), ranges);
+        shuffle_weights(conv.weights().as<data>(), ranges, stream);
     } else if (node.is_type<fully_connected>()) {
         auto& fc = node.as<fully_connected>();
-        shuffle_weights(fc.weights().as<data>(), ranges);
+        shuffle_weights(fc.weights().as<data>(), ranges, stream);
     } else {
         // General case for pass-through layers
         for (auto& user : node.get_users()) {
-            shuffle_features(*user, ranges);
+            shuffle_features(*user, ranges, stream);
         }
     }
 }
@@ -155,7 +155,7 @@ void concat_input_order::run(program_impl& p) {
         // Check that we can fuse shuffling to users
         bool can_shuffle_users = true;
         for (auto user : concat_node.get_users()) {
-            can_shuffle_users &= can_shuffle_features(*user);
+            can_shuffle_users &= can_shuffle_features(*user, p.get_stream());
         }
 
         if (!along_f || !no_fusing || !correct_format || !single_format || already_aligned || !can_shuffle_users)
@@ -207,8 +207,7 @@ void concat_input_order::run(program_impl& p) {
         mutable_prim->input = new_input_ids;
         // Correct users for shuffled features
         for (auto& user : concat_node.get_users()) {
-            shuffle_features(*user, shuffled_ranges);
+            shuffle_features(*user, shuffled_ranges, p.get_stream());
         }
     }
 }
-
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/eltwise_remove_stride.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/eltwise_remove_stride.cpp
index c30a3348548f5c..b0c28bd93a60d0 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/eltwise_remove_stride.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/eltwise_remove_stride.cpp
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include "api/tensor.hpp"
+#include "cldnn/runtime/tensor.hpp"
 
 #include "pass_manager.h"
 
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/handle_input_padding.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/handle_input_padding.cpp
index 137ced81fc25aa..84496690536dca 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/handle_input_padding.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/handle_input_padding.cpp
@@ -7,7 +7,7 @@
 #include "pass_manager.h"
 #include "border_inst.h"
 #include "convolution_inst.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include <memory>
 
 using namespace cldnn;
@@ -113,4 +113,4 @@ void handle_input_padding::run(program_impl& p) {
             }
         }
     }
-}
\ No newline at end of file
+}
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/oooq_memory_dependencies.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/oooq_memory_dependencies.cpp
index 380b18b951f66a..ccf9903701a897 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/oooq_memory_dependencies.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/oooq_memory_dependencies.cpp
@@ -9,7 +9,7 @@
 #include "layout_optimizer.h"
 #include "program_impl.h"
 #include "program_helpers.h"
-#include "cldnn_itt.h"
+#include "runtime/cldnn_itt.hpp"
 #include <vector>
 #include <memory>
 #include <list>
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/post_input_reorder.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/post_input_reorder.cpp
index f98eabd2cf6514..b4371112d11767 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/post_input_reorder.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/post_input_reorder.cpp
@@ -45,9 +45,8 @@ void post_input_reorder::run(program_impl& p) {
         const auto impl = node->get_selected_impl();
         // add a reorder if primitive's input format doesn't match implementation's input format
         if (node->is_type<fully_connected>()) {
-            const auto& fc_impl = dynamic_cast<gpu::typed_primitive_gpu_impl<fully_connected>&>(*impl);
-            const auto& fc_params =
-                *static_cast<kernel_selector::fully_connected_params*>(fc_impl._kernel_data.params.get());
+            const auto& fc_impl = dynamic_cast<const gpu::typed_primitive_gpu_impl<fully_connected>&>(*impl);
+            const auto& fc_params = *static_cast<kernel_selector::fully_connected_params*>(fc_impl._kernel_data.params.get());
 
             auto layout_format = from_data_layout(fc_params.inputs[0].GetLayout());
             auto& input = node->get_dependencies()[0];
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/post_optimize_weights.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/post_optimize_weights.cpp
index 271293e4dc43ea..85ea94cd476c8b 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/post_optimize_weights.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/post_optimize_weights.cpp
@@ -6,7 +6,7 @@
 
 #include "pass_manager.h"
 #include "program_helpers.h"
-#include "api_extension/fused_conv_eltwise.hpp"
+#include "cldnn/primitives/fused_conv_eltwise.hpp"
 #include "include/fused_conv_eltwise_inst.h"
 #include "include/binary_convolution_inst.h"
 #include "include/deformable_convolution_inst.h"
@@ -36,7 +36,7 @@ post_optimize_weights::weights_bias_offset post_optimize_weights::get_weights_bi
 template<typename T>
 void post_optimize_weights::optimize_weights(T& node, program_impl& p) {
     auto offsets = get_weights_bias_offset(node);
-    auto* impl = node.get_selected_impl().get();
+    auto impl = node.get_selected_impl();
     auto output_layout = node.get_output_layout();
     auto& weights_reorder_params = impl->_weights_reorder_params;
 
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/pre_replace_deconv.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/pre_replace_deconv.cpp
index 874eedb91ce759..ccb00d3726e5b6 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/pre_replace_deconv.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/pre_replace_deconv.cpp
@@ -16,10 +16,13 @@
 #include <memory>
 #include <string>
 #include <utility>
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 void pre_replace_deconv::run(program_impl& p) {
     bool update_processing_order = false;
+
+    auto& stream = p.get_stream();
+
     auto itr = p.nodes_map.begin();
     while (itr != p.nodes_map.end()) {
         auto node_itr = itr++;
@@ -242,16 +245,16 @@ void pre_replace_deconv::run(program_impl& p) {
                 auto target_weights_layout = layout{ weights_layout.data_type, weights_layout.format, target_weights_size };
 
                 {
-                     memory_impl::ptr data_to_allocate = p.get_engine().allocate_memory(target_weights_layout, 0);
+                     memory::ptr data_to_allocate = p.get_engine().allocate_memory(target_weights_layout);
 
                      std::vector<float> weights_vec_float;
 
                      if (weights_data_type == data_types::f16) {
-                         mem_lock<half_t> src{ cur_weights_node_ptr->as<data>().get_attached_memory() };
+                         mem_lock<half_t> src{ cur_weights_node_ptr->as<data>().get_attached_memory_ptr(), stream };
                          for (uint32_t i = 0; i < weights_layout.size.count(); i++)
                              weights_vec_float.push_back(static_cast<float>(src.data()[i]));
                      } else {
-                         mem_lock<float> src{ cur_weights_node_ptr->as<data>().get_attached_memory() };
+                         mem_lock<float> src{ cur_weights_node_ptr->as<data>().get_attached_memory_ptr(), stream };
                          for (uint32_t i = 0; i < weights_layout.size.count(); i++)
                              weights_vec_float.push_back(src.data()[i]);
                      }
@@ -266,17 +269,16 @@ void pre_replace_deconv::run(program_impl& p) {
                          subpixel_weights);
 
                      if (weights_data_type == data_types::f16) {
-                         mem_lock<half_t> dst{ data_to_allocate };
+                         mem_lock<half_t> dst{ data_to_allocate, stream};
                          program_helpers::set_weights_values<half_t>(dst.data(), subpixel_weights);
                      } else if (weights_data_type == data_types::f32) {
-                         mem_lock<float> dst{ data_to_allocate };
+                         mem_lock<float> dst{ data_to_allocate, stream };
                          program_helpers::set_weights_values<float>(dst.data(), subpixel_weights);
                      } else {
                          throw std::logic_error("Not supported data type.");
                      }
 
-                     memory api_memory = memory(data_to_allocate.detach());
-                     auto data_node_weights_replace = std::make_shared<data>(weights_vec[0] + "_conv_rpl", api_memory);
+                     auto data_node_weights_replace = std::make_shared<data>(weights_vec[0] + "_conv_rpl", data_to_allocate);
                      p.get_or_create(data_node_weights_replace);
                      auto data_node_weights_replace_node_ptr = p.nodes_map.find(weights_vec[0] + "_conv_rpl")->second;
                      auto& data_node = data_node_weights_replace_node_ptr->as<data>();
@@ -285,10 +287,10 @@ void pre_replace_deconv::run(program_impl& p) {
                 float bias = 0;
 
                 if (bias_data_type == data_types::f16) {
-                    mem_lock<half_t> src{ bias_id_node_ptr->as<data>().get_attached_memory() };
+                    mem_lock<half_t> src{ bias_id_node_ptr->as<data>().get_attached_memory_ptr(), stream };
                     bias = static_cast<float>(src.data()[0]);
                 } else {
-                    mem_lock<float> src{ bias_id_node_ptr->as<data>().get_attached_memory() };
+                    mem_lock<float> src{ bias_id_node_ptr->as<data>().get_attached_memory_ptr(), stream };
                     bias = src.data()[0];
                 }
 
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_buffer_fusing.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_buffer_fusing.cpp
index abd9985cb72292..eebd65149ad27a 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_buffer_fusing.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_buffer_fusing.cpp
@@ -4,8 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include "api/eltwise.hpp"
-#include "api/pooling.hpp"
+#include "pooling_inst.h"
 #include "fused_conv_eltwise_inst.h"
 #include "primitive_inst.h"
 #include "activation_inst.h"
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_primitive_fusing.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_primitive_fusing.cpp
index 8dc449053ce22a..ffabb96f2e40bf 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_primitive_fusing.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_primitive_fusing.cpp
@@ -4,13 +4,12 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include "api/pooling.hpp"
-#include "api/proposal.hpp"
-#include "api/roi_pooling.hpp"
-
 #include "program_helpers.h"
 #include "pass_manager.h"
 
+#include "pooling_inst.h"
+#include "proposal_inst.h"
+#include "roi_pooling_inst.h"
 #include "quantize_inst.h"
 #include "binary_convolution_inst.h"
 #include "activation_inst.h"
@@ -51,7 +50,7 @@
 #include <string>
 #include <utility>
 #include <deque>
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 void prepare_primitive_fusing::run(program_impl& p) {
     fuse_reorders(p);
@@ -1334,24 +1333,16 @@ void prepare_conv_eltw_read_write_opt::conv_eltwise_read_write_opt(program_impl&
     // buffer shared between primitives, if second input is mutable data, then we can reuse this memory
     auto shared_buffer_mem = second_input_node->is_type<mutable_data>()
                                  ? second_input_node->as<mutable_data>().get_attached_memory_ptr()
-                                 : p.get_engine().allocate_memory(node->get_output_layout(), 0);
-
-    float zero = 0.0f;
-    layout dummy_layout(data_types::f32, format::bfyx, tensor(1, 1, 1, 1));
+                                 : p.get_engine().allocate_memory(node->get_output_layout());
 
     // this one is the first one to write data to
-    auto rw_output_prim0 = std::make_shared<mutable_data>(fused_conv_eltw_node->id() + "_RW_OPT_use",
-                                                          memory::attach(dummy_layout, &zero, 1));
+    auto rw_output_prim0 = std::make_shared<mutable_data>(fused_conv_eltw_node->id() + "_RW_OPT_use", shared_buffer_mem);
     // this one already expects data to be inside
-    auto rw_output_prim1 = std::make_shared<mutable_data>(fused_conv_eltw_node->id() + "_RW_OPT_reuse",
-                                                          memory::attach(dummy_layout, &zero, 1));
+    auto rw_output_prim1 = std::make_shared<mutable_data>(fused_conv_eltw_node->id() + "_RW_OPT_reuse", shared_buffer_mem);
 
     auto& rw_output_node0 = p.get_or_create(rw_output_prim0);
     auto& rw_output_node1 = p.get_or_create(rw_output_prim1);
 
-    rw_output_node0.as<mutable_data>().attach_memory(*shared_buffer_mem, false);
-    rw_output_node1.as<mutable_data>().attach_memory(*shared_buffer_mem, false);
-
     // add connection between second input node -> rw_output_node0 -> node
     p.add_intermediate(rw_output_node0, *node, 1, true);
     // replace other connections with rw_output_node0
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_quantization.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_quantization.cpp
index 134494d2ba7f2e..35d31d3a71d211 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_quantization.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_quantization.cpp
@@ -4,11 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include "api/quantize.hpp"
-#include "api/binary_convolution.hpp"
-#include "api/scale.hpp"
-#include "api/pooling.hpp"
-
+#include "pooling_inst.h"
 #include "quantize_inst.h"
 #include "binary_convolution_inst.h"
 #include "scale_inst.h"
@@ -16,10 +12,10 @@
 #include "data_inst.h"
 #include "pass_manager.h"
 #include "program_helpers.h"
-#include <algorithm>
 #include "to_string_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
+#include <algorithm>
 #include <string>
 #include <memory>
 #include <vector>
@@ -46,19 +42,23 @@ void prepare_quantization::prepare_packed_quantize(program_impl& p) {
             auto &input_low = input_low_node.as<data>();
             auto &input_high = input_high_node.as<data>();
 
-            auto &mem_input_low = input_low.get_attached_memory();
-            auto &mem_input_high = input_high.get_attached_memory();
+            auto mem_input_low = input_low.get_attached_memory_ptr();
+            auto mem_input_high = input_high.get_attached_memory_ptr();
 
             auto output_dt = quantize_node.get_output_layout().data_type;
 
+            auto& stream = p.get_stream();
+
             if (levels == 2) {
                 bool is_binarization = true;
-                switch (mem_input_high.get_layout().data_type) {
+                switch (mem_input_high->get_layout().data_type) {
                     case data_types::f32: {
-                        auto data_input_low = static_cast<float*>(mem_input_low.lock());
-                        auto data_input_high = static_cast<float*>(mem_input_high.lock());
+                        mem_lock<float> data_input_low_lock{mem_input_low, stream};
+                        mem_lock<float> data_input_high_lock{mem_input_high, stream};
+                        auto data_input_low = data_input_low_lock.data();
+                        auto data_input_high = data_input_high_lock.data();
 
-                        for (size_t i = 0; i < mem_input_high.get_layout().count(); i++) {
+                        for (size_t i = 0; i < mem_input_high->get_layout().count(); i++) {
                             if (data_input_high[i] != data_input_low[i]) {
                                 is_binarization = false;
                                 break;
@@ -67,10 +67,12 @@ void prepare_quantization::prepare_packed_quantize(program_impl& p) {
                         break;
                     }
                     case data_types::f16: {
-                        auto data_input_low = static_cast<uint16_t*>(mem_input_low.lock());
-                        auto data_input_high = static_cast<uint16_t*>(mem_input_high.lock());
+                        mem_lock<uint16_t> data_input_low_lock{mem_input_low, stream};
+                        mem_lock<uint16_t> data_input_high_lock{mem_input_high, stream};
+                        auto data_input_low = data_input_low_lock.data();
+                        auto data_input_high = data_input_high_lock.data();
 
-                        for (size_t i = 0; i < mem_input_high.get_layout().count(); i++) {
+                        for (size_t i = 0; i < mem_input_high->get_layout().count(); i++) {
                             if (data_input_high[i] != data_input_low[i]) {
                                 is_binarization = false;
                                 break;
@@ -81,8 +83,6 @@ void prepare_quantization::prepare_packed_quantize(program_impl& p) {
                     default:
                         CLDNN_ERROR_MESSAGE(node->id(), "prepare_quantization: Unsupported precision of quantize inputs");
                 }
-                mem_input_low.unlock();
-                mem_input_high.unlock();
 
                 if (is_binarization) {
                     output_dt = data_types::bin;
@@ -106,6 +106,8 @@ void prepare_quantization::prepare_scale_shift_opt(program_impl &p) {
             if (levels == 2 || levels > 256 || quantize_node.get_scale_shift_opt() || quantize_node.is_constant())
                 return;
 
+            auto& stream = p.get_stream();
+
             program_node &input_low_node = quantize_node.get_dependency(1);
             program_node &input_high_node = quantize_node.get_dependency(2);
             program_node &output_low_node = quantize_node.get_dependency(3);
@@ -121,20 +123,20 @@ void prepare_quantization::prepare_scale_shift_opt(program_impl &p) {
             auto &output_low = output_low_node.as<data>();
             auto &output_high = output_high_node.as<data>();
 
-            auto &mem_input_low = input_low.get_attached_memory();
-            auto &mem_input_high = input_high.get_attached_memory();
-            auto &mem_output_low = output_low.get_attached_memory();
-            auto &mem_output_high = output_high.get_attached_memory();
+            auto mem_input_low = input_low.get_attached_memory_ptr();
+            auto mem_input_high = input_high.get_attached_memory_ptr();
+            auto mem_output_low = output_low.get_attached_memory_ptr();
+            auto mem_output_high = output_high.get_attached_memory_ptr();
 
-            auto scales_layout = mem_input_low.get_layout();
-            scales_layout.size = tensor::max(scales_layout.size, mem_input_high.get_layout().size);
-            scales_layout.size = tensor::max(scales_layout.size, mem_output_low.get_layout().size);
-            scales_layout.size = tensor::max(scales_layout.size, mem_output_high.get_layout().size);
+            auto scales_layout = mem_input_low->get_layout();
+            scales_layout.size = tensor::max(scales_layout.size, mem_input_high->get_layout().size);
+            scales_layout.size = tensor::max(scales_layout.size, mem_output_low->get_layout().size);
+            scales_layout.size = tensor::max(scales_layout.size, mem_output_high->get_layout().size);
 
-            auto mem_input_scale  = p.get_engine().allocate_memory(scales_layout, mem_input_low.get_net_id(), false);
-            auto mem_input_shift  = p.get_engine().allocate_memory(scales_layout, mem_input_high.get_net_id(), false);
-            auto mem_output_scale = p.get_engine().allocate_memory(scales_layout, mem_output_low.get_net_id(), false);
-            auto mem_output_shift = p.get_engine().allocate_memory(scales_layout, mem_output_high.get_net_id(), false);
+            auto mem_input_scale  = p.get_engine().allocate_memory(scales_layout, false);
+            auto mem_input_shift  = p.get_engine().allocate_memory(scales_layout, false);
+            auto mem_output_scale = p.get_engine().allocate_memory(scales_layout, false);
+            auto mem_output_shift = p.get_engine().allocate_memory(scales_layout, false);
 
             auto get_offset_safe = [](layout l, tensor idx) -> int {
                 auto sizes = l.size;
@@ -164,17 +166,26 @@ void prepare_quantization::prepare_scale_shift_opt(program_impl &p) {
             float out_shift_val = 0.0f;
             float in_lo_val = 0.0f;
             float in_hi_val = 0.0f;
-            switch (mem_output_high.get_layout().data_type) {
+            switch (mem_output_high->get_layout().data_type) {
                 case data_types::f32: {
                     // TODO [LOW PRECISION]: Output low/high values can be removed.
-                    auto data_input_low = static_cast<float*>(mem_input_low.lock());
-                    auto data_input_high = static_cast<float*>(mem_input_high.lock());
-                    auto data_output_low = static_cast<float*>(mem_output_low.lock());
-                    auto data_output_high = static_cast<float*>(mem_output_high.lock());
-                    auto data_input_scale = static_cast<float*>(mem_input_scale->lock());
-                    auto data_input_shift = static_cast<float*>(mem_input_shift->lock());
-                    auto data_output_scale = static_cast<float*>(mem_output_scale->lock());
-                    auto data_output_shift = static_cast<float*>(mem_output_shift->lock());
+                    mem_lock<float> data_input_low_lock{mem_input_low, stream};
+                    mem_lock<float> data_input_high_lock{mem_input_high, stream};
+                    mem_lock<float> data_output_low_lock{mem_output_low, stream};
+                    mem_lock<float> data_output_high_lock{mem_output_high, stream};
+                    mem_lock<float> data_input_scale_lock{mem_input_scale, stream};
+                    mem_lock<float> data_input_shift_lock{mem_input_shift, stream};
+                    mem_lock<float> data_output_scale_lock{mem_output_scale, stream};
+                    mem_lock<float> data_output_shift_lock{mem_output_shift, stream};
+
+                    auto data_input_low = data_input_low_lock.data();
+                    auto data_input_high = data_input_high_lock.data();
+                    auto data_output_low = data_output_low_lock.data();
+                    auto data_output_high = data_output_high_lock.data();
+                    auto data_input_scale = data_input_scale_lock.data();
+                    auto data_input_shift = data_input_shift_lock.data();
+                    auto data_output_scale = data_output_scale_lock.data();
+                    auto data_output_shift = data_output_shift_lock.data();
 
                     for (int b = 0; b < scales_layout.size.batch[0]; b++) {
                         for (int f = 0; f < scales_layout.size.feature[0]; f++) {
@@ -182,11 +193,11 @@ void prepare_quantization::prepare_scale_shift_opt(program_impl &p) {
                                 for (int x = 0; x < scales_layout.size.spatial[0]; x++) {
                                     auto idx = cldnn::tensor(format::bfyx, {b, f, y, x}, 0);
                                     auto s_offset = scales_layout.get_linear_offset(idx);
-                                    auto in_lo = data_input_low[get_offset_safe(mem_input_low.get_layout(), idx)];
-                                    auto in_hi = data_input_high[get_offset_safe(mem_input_high.get_layout(), idx)];
+                                    auto in_lo = data_input_low[get_offset_safe(mem_input_low->get_layout(), idx)];
+                                    auto in_hi = data_input_high[get_offset_safe(mem_input_high->get_layout(), idx)];
 
-                                    auto out_lo = data_output_low[get_offset_safe(mem_output_low.get_layout(), idx)];
-                                    auto out_hi = data_output_high[get_offset_safe(mem_output_high.get_layout(), idx)];
+                                    auto out_lo = data_output_low[get_offset_safe(mem_output_low->get_layout(), idx)];
+                                    auto out_hi = data_output_high[get_offset_safe(mem_output_high->get_layout(), idx)];
                                     data_input_scale[s_offset] = (static_cast<float>(levels) - 1) / (in_hi - in_lo);
                                     data_input_shift[s_offset] = - in_lo * (static_cast<float>(levels) - 1) / (in_hi - in_lo);
                                     data_output_scale[s_offset] = (out_hi - out_lo) / (static_cast<float>(levels) - 1);
@@ -225,21 +236,30 @@ void prepare_quantization::prepare_scale_shift_opt(program_impl &p) {
                         if (data_input_shift[i] != 0.0f)
                             need_pre_shift = true;
 
-                        if (in_lo_val != data_input_low[i % mem_input_low.get_layout().count()] ||
-                            in_hi_val != data_input_high[i % mem_input_high.get_layout().count()])
+                        if (in_lo_val != data_input_low[i % mem_input_low->get_layout().count()] ||
+                            in_hi_val != data_input_high[i % mem_input_high->get_layout().count()])
                             per_tensor_in_range = false;
                     }
                     break;
                 }
                 case data_types::f16: {
-                    auto data_input_low = static_cast<uint16_t*>(mem_input_low.lock());
-                    auto data_input_high = static_cast<uint16_t*>(mem_input_high.lock());
-                    auto data_output_low = static_cast<uint16_t*>(mem_output_low.lock());
-                    auto data_output_high = static_cast<uint16_t*>(mem_output_high.lock());
-                    auto data_input_scale = static_cast<uint16_t*>(mem_input_scale->lock());
-                    auto data_input_shift = static_cast<uint16_t*>(mem_input_shift->lock());
-                    auto data_output_scale = static_cast<uint16_t*>(mem_output_scale->lock());
-                    auto data_output_shift = static_cast<uint16_t*>(mem_output_shift->lock());
+                    mem_lock<uint16_t> data_input_low_lock{mem_input_low, stream};
+                    mem_lock<uint16_t> data_input_high_lock{mem_input_high, stream};
+                    mem_lock<uint16_t> data_output_low_lock{mem_output_low, stream};
+                    mem_lock<uint16_t> data_output_high_lock{mem_output_high, stream};
+                    mem_lock<uint16_t> data_input_scale_lock{mem_input_scale, stream};
+                    mem_lock<uint16_t> data_input_shift_lock{mem_input_shift, stream};
+                    mem_lock<uint16_t> data_output_scale_lock{mem_output_scale, stream};
+                    mem_lock<uint16_t> data_output_shift_lock{mem_output_shift, stream};
+
+                    auto data_input_low = data_input_low_lock.data();
+                    auto data_input_high = data_input_high_lock.data();
+                    auto data_output_low = data_output_low_lock.data();
+                    auto data_output_high = data_output_high_lock.data();
+                    auto data_input_scale = data_input_scale_lock.data();
+                    auto data_input_shift = data_input_shift_lock.data();
+                    auto data_output_scale = data_output_scale_lock.data();
+                    auto data_output_shift = data_output_shift_lock.data();
 
                     for (int b = 0; b < scales_layout.size.batch[0]; b++) {
                         for (int f = 0; f < scales_layout.size.feature[0]; f++) {
@@ -247,11 +267,11 @@ void prepare_quantization::prepare_scale_shift_opt(program_impl &p) {
                                 for (int x = 0; x < scales_layout.size.spatial[0]; x++) {
                                     auto idx = cldnn::tensor(format::bfyx, {b, f, y, x}, 0);
                                     auto s_offset = scales_layout.get_linear_offset(idx);
-                                    auto in_lo = half_to_float(data_input_low[get_offset_safe(mem_input_low.get_layout(), idx)]);
-                                    auto in_hi = half_to_float(data_input_high[get_offset_safe(mem_input_high.get_layout(), idx)]);
+                                    auto in_lo = half_to_float(data_input_low[get_offset_safe(mem_input_low->get_layout(), idx)]);
+                                    auto in_hi = half_to_float(data_input_high[get_offset_safe(mem_input_high->get_layout(), idx)]);
 
-                                    auto out_lo = half_to_float(data_output_low[get_offset_safe(mem_output_low.get_layout(), idx)]);
-                                    auto out_hi = half_to_float(data_output_high[get_offset_safe(mem_output_high.get_layout(), idx)]);
+                                    auto out_lo = half_to_float(data_output_low[get_offset_safe(mem_output_low->get_layout(), idx)]);
+                                    auto out_hi = half_to_float(data_output_high[get_offset_safe(mem_output_high->get_layout(), idx)]);
                                     data_input_scale[s_offset] = float_to_half((static_cast<float>(levels) - 1) / (in_hi - in_lo));
                                     data_input_shift[s_offset] = float_to_half(- in_lo * (static_cast<float>(levels) - 1) / (in_hi - in_lo));
                                     data_output_scale[s_offset] = float_to_half((out_hi - out_lo) / (static_cast<float>(levels) - 1));
@@ -289,8 +309,8 @@ void prepare_quantization::prepare_scale_shift_opt(program_impl &p) {
                         if (half_to_float(data_input_shift[i]) != 0.0f)
                             need_pre_shift = true;
 
-                        if (in_lo_val != half_to_float(data_input_low[i % mem_input_low.get_layout().count()]) ||
-                            in_hi_val != half_to_float(data_input_high[i % mem_input_high.get_layout().count()]))
+                        if (in_lo_val != half_to_float(data_input_low[i % mem_input_low->get_layout().count()]) ||
+                            in_hi_val != half_to_float(data_input_high[i % mem_input_high->get_layout().count()]))
                             per_tensor_in_range = false;
                     }
                     break;
@@ -303,22 +323,15 @@ void prepare_quantization::prepare_scale_shift_opt(program_impl &p) {
                 return;
             }
 
-            layout dummy_layout(data_types::f32, format::bfyx, tensor(1, 1, 1, 1));
-            float zero = 0.f;
-            auto in_scale_prim = std::make_shared<data>(quantize_node.id() + "_in_scale", memory::attach(dummy_layout, &zero, 1));
-            auto in_shift_prim = std::make_shared<data>(quantize_node.id() + "_in_shift", memory::attach(dummy_layout, &zero, 1));
-            auto out_scale_prim = std::make_shared<data>(quantize_node.id() + "_output_scale", memory::attach(dummy_layout, &zero, 1));
-            auto out_shift_prim = std::make_shared<data>(quantize_node.id() + "_output_shift", memory::attach(dummy_layout, &zero, 1));
+            auto in_scale_prim = std::make_shared<data>(quantize_node.id() + "_in_scale", mem_input_scale);
+            auto in_shift_prim = std::make_shared<data>(quantize_node.id() + "_in_shift", mem_input_shift);
+            auto out_scale_prim = std::make_shared<data>(quantize_node.id() + "_output_scale", mem_output_scale);
+            auto out_shift_prim = std::make_shared<data>(quantize_node.id() + "_output_shift", mem_output_shift);
             auto& in_scale_node = p.get_or_create(in_scale_prim);
             auto& in_shift_node = p.get_or_create(in_shift_prim);
             auto& out_scale_node = p.get_or_create(out_scale_prim);
             auto& out_shift_node = p.get_or_create(out_shift_prim);
 
-            in_scale_node.as<data>().attach_memory(*mem_input_scale);
-            in_shift_node.as<data>().attach_memory(*mem_input_shift);
-            out_scale_node.as<data>().attach_memory(*mem_output_scale);
-            out_shift_node.as<data>().attach_memory(*mem_output_shift);
-
             auto& inputs = p.get_inputs();
 
             inputs.push_back(&in_scale_node);
@@ -381,15 +394,6 @@ void prepare_quantization::prepare_scale_shift_opt(program_impl &p) {
                 quantize_node.set_per_tensor_output_shift();
                 quantize_node.set_output_shift_val(out_shift_val);
             }
-
-            mem_input_low.unlock();
-            mem_input_high.unlock();
-            mem_output_low.unlock();
-            mem_output_high.unlock();
-            mem_input_scale->unlock();
-            mem_input_shift->unlock();
-            mem_output_scale->unlock();
-            mem_output_shift->unlock();
         });
     }
 }
@@ -453,6 +457,8 @@ void prepare_quantization::prepare_asymmetric_quantization(program_impl &p) {
         auto node_itr = itr++;
         auto& node = (*node_itr);
 
+        auto& stream = p.get_stream();
+
         // Detects if given eltwise node performs zero point subtraction
         auto is_zero_point_node = [](eltwise_node& node) -> bool {
             auto prim = node.get_primitive();
@@ -480,8 +486,7 @@ void prepare_quantization::prepare_asymmetric_quantization(program_impl &p) {
             return true;
         };
 
-        auto fill_compensation = [&](int groups, memory_impl* w, memory_impl* azp, memory_impl* wzp,
-                                     memory_impl::ptr compensation) {
+        auto fill_compensation = [&](int groups, const memory::ptr w, const  memory::ptr azp, const memory::ptr wzp, memory::ptr compensation) {
             auto wl = w->get_layout();
 
             int GS = groups;
@@ -492,40 +497,40 @@ void prepare_quantization::prepare_asymmetric_quantization(program_impl &p) {
             auto w_dt = wl.data_type;
             auto azp_dt = azp->get_layout().data_type;
 
-            mem_lock<float> comp_lock{compensation};
+            mem_lock<float> comp_lock{compensation, stream};
 
             if (w_dt == data_types::u8 && azp_dt == data_types::u8) {
-                mem_lock<uint8_t> w_lock(*w);
-                mem_lock<uint8_t> azp_lock(*azp);
+                mem_lock<uint8_t> w_lock(w, stream);
+                mem_lock<uint8_t> azp_lock(azp, stream);
                 if (wzp) {
-                    mem_lock<uint8_t> wzp_lock(*wzp);
+                    mem_lock<uint8_t> wzp_lock(wzp, stream);
                     fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
                 } else {
                     fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<uint8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
                 }
             } else if (w_dt == data_types::i8 && azp_dt == data_types::u8) {
-                mem_lock<int8_t> w_lock(*w);
-                mem_lock<uint8_t> azp_lock(*azp);
+                mem_lock<int8_t> w_lock(w, stream);
+                mem_lock<uint8_t> azp_lock(azp, stream);
                 if (wzp) {
-                    mem_lock<int8_t> wzp_lock(*wzp);
+                    mem_lock<int8_t> wzp_lock(wzp, stream);
                     fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
                 } else {
                     fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<int8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
                 }
             } else if (w_dt == data_types::i8 && azp_dt == data_types::i8) {
-                mem_lock<int8_t> w_lock(*w);
-                mem_lock<int8_t> azp_lock(*azp);
+                mem_lock<int8_t> w_lock(w, stream);
+                mem_lock<int8_t> azp_lock(azp, stream);
                 if (wzp) {
-                    mem_lock<int8_t> wzp_lock(*wzp);
+                    mem_lock<int8_t> wzp_lock(wzp, stream);
                     fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
                 } else {
                     fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<int8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
                 }
             } else if (w_dt == data_types::u8 && azp_dt == data_types::i8) {
-                mem_lock<uint8_t> w_lock(*w);
-                mem_lock<int8_t> azp_lock(*azp);
+                mem_lock<uint8_t> w_lock(w, stream);
+                mem_lock<int8_t> azp_lock(azp, stream);
                 if (wzp) {
-                    mem_lock<uint8_t> wzp_lock(*wzp);
+                    mem_lock<uint8_t> wzp_lock(wzp, stream);
                     fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
                 } else {
                     fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<uint8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
@@ -590,13 +595,14 @@ void prepare_quantization::prepare_asymmetric_quantization(program_impl &p) {
 
                 auto l = layout{new_a_zp->get_output_layout().data_type, format::bfyx, tensor{1, ifm_aligned, 1, 1}};
                 int s = new_a_zp->get_output_layout().size.feature[0];
-                auto azp_aligned = p.get_engine().allocate_memory(l, 0, false);
-                mem_lock<int8_t> new_data{azp_aligned};
-                mem_lock<int8_t> old_data{new_a_zp->as<data>().get_attached_memory()};
+                auto azp_aligned = p.get_engine().allocate_memory(l);
+                auto old_ptr = new_a_zp->as<data>().get_attached_memory_ptr();
+                mem_lock<int8_t> new_data{azp_aligned, stream};
+                mem_lock<int8_t> old_data{old_ptr, stream};
                 for (int i = 0; i < ifm_aligned; i++) {
                     new_data.data()[i] = old_data.data()[i % s];
                 }
-                new_a_zp->as<data>().attach_memory(*azp_aligned);
+                new_a_zp->as<data>().attach_memory(azp_aligned);
 
                 input = new_input->id();
                 a_zero_points.push_back(new_a_zp->id());
@@ -609,13 +615,14 @@ void prepare_quantization::prepare_asymmetric_quantization(program_impl &p) {
 
                 auto l = layout{new_w_zp->get_output_layout().data_type, format::bfyx, tensor{ofm_aligned, 1, 1, 1}};
                 int s = new_w_zp->get_output_layout().size.batch[0];
-                auto wzp_aligned = p.get_engine().allocate_memory(l, 0, false);
-                mem_lock<int8_t> new_data{wzp_aligned};
-                mem_lock<int8_t> old_data{new_w_zp->as<data>().get_attached_memory()};
+                auto wzp_aligned = p.get_engine().allocate_memory(l);
+                auto old_ptr = new_w_zp->as<data>().get_attached_memory_ptr();
+                mem_lock<int8_t> new_data{wzp_aligned, stream};
+                mem_lock<int8_t> old_data{old_ptr, stream};
                 for (int i = 0; i < ofm_aligned; i++) {
                     new_data.data()[i] = old_data.data()[i % s];
                 }
-                new_w_zp->as<data>().attach_memory(*wzp_aligned);
+                new_w_zp->as<data>().attach_memory(wzp_aligned);
 
                 weights = { new_weights->id() };
                 w_zero_points.push_back(new_w_zp->id());
@@ -623,19 +630,16 @@ void prepare_quantization::prepare_asymmetric_quantization(program_impl &p) {
 
             if (need_compensation) {
                 auto l = layout{data_types::f32, format::bfyx, tensor{1, ofm_aligned, 1, 1}};
-                auto data_to_allocate = p.get_engine().allocate_memory(l, 0, false);
-                auto w = &new_weights->as<data>().get_attached_memory();
-                auto azp = asymmetric_data ? &new_a_zp->as<data>().get_attached_memory() : nullptr;
-                auto wzp = asymmetric_weights ? &new_w_zp->as<data>().get_attached_memory() : nullptr;
+                auto data_to_allocate = p.get_engine().allocate_memory(l);
+                auto w = new_weights->as<data>().get_attached_memory_ptr();
+                auto azp = asymmetric_data ? new_a_zp->as<data>().get_attached_memory_ptr() : nullptr;
+                auto wzp = asymmetric_weights ? new_w_zp->as<data>().get_attached_memory_ptr() : nullptr;
                 fill_compensation(groups, w, azp, wzp, data_to_allocate);
-                layout dummy_layout(data_types::f32, format::bfyx, tensor(1, 1, 1, 1));
-                float zero = 0.f;
 
-                auto compensation_prim = std::make_shared<data>(convolution_node.id() + "_compensation", memory::attach(dummy_layout, &zero, 1));
+                auto compensation_prim = std::make_shared<data>(convolution_node.id() + "_compensation", data_to_allocate);
                 new_compenstation = &p.get_or_create(compensation_prim);
                 p.get_inputs().push_back(new_compenstation);
                 compensation.push_back(new_compenstation->id());
-                new_compenstation->as<data>().attach_memory(*data_to_allocate);
             }
 
             // Collect dependencies of a new convolution node
@@ -745,11 +749,11 @@ void prepare_quantization::prepare_dequantize_merge(program_impl &p) {
                 }
             }
 
-            auto get_scale_shift_mem = [](const eltwise_node& eltw, size_t dep_id) -> memory_impl& {
+            auto get_scale_shift_mem = [](const eltwise_node& eltw, size_t dep_id) -> memory::ptr {
                 if (dep_id >= eltw.get_dependencies().size())
                     CLDNN_ERROR_MESSAGE(eltw.id(), "Invalid dependency id in dequantize optimization");
 
-                return eltw.get_dependency(dep_id).as<data>().get_attached_memory();
+                return eltw.get_dependency(dep_id).as<data>().get_attached_memory_ptr();
             };
 
             auto eltw_mode = node.get_primitive()->mode;
@@ -757,6 +761,7 @@ void prepare_quantization::prepare_dequantize_merge(program_impl &p) {
                 return;
 
             auto& input = node.input();
+            auto& stream = p.get_stream();
 
             for (auto& user : input.get_users()) {
                 if (user == &node)
@@ -781,20 +786,20 @@ void prepare_quantization::prepare_dequantize_merge(program_impl &p) {
 
                 bool same_params = true;
                 for (size_t i = 1; i < node.get_dependencies().size(); i++) {
-                    auto& mem0 = get_scale_shift_mem(eltwise_dep, i);
-                    auto& mem1 = get_scale_shift_mem(node, i);
+                    auto mem0 = get_scale_shift_mem(eltwise_dep, i);
+                    auto mem1 = get_scale_shift_mem(node, i);
 
-                    auto ptr0 = static_cast<uint8_t*>(mem0.lock());
-                    auto ptr1 = static_cast<uint8_t*>(mem1.lock());
+                    mem_lock<uint8_t> mem0_lock{mem0, stream};
+                    mem_lock<uint8_t> mem1_lock{mem1, stream};
+                    auto ptr0 = mem0_lock.data();
+                    auto ptr1 = mem1_lock.data();
 
-                    for (size_t j = 0; j < mem0.get_layout().bytes_count(); j++) {
+                    for (size_t j = 0; j < mem0->get_layout().bytes_count(); j++) {
                         if (ptr0[j] != ptr1[j]) {
                             same_params = false;
                             break;
                         }
                     }
-                    mem0.unlock();
-                    mem1.unlock();
                 }
 
                 if (same_params) {
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/propagate_constants.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/propagate_constants.cpp
index 0229027776ca20..83e2742063f713 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/propagate_constants.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/propagate_constants.cpp
@@ -6,11 +6,11 @@
 
 #include "pass_manager.h"
 #include "program_node.h"
-#include "engine_impl.h"
+#include "cldnn/runtime/engine.hpp"
 #include "program_impl.h"
 #include "network_impl.h"
 #include "data_inst.h"
-#include "cldnn_itt.h"
+#include "runtime/cldnn_itt.hpp"
 #include <vector>
 #include <list>
 #include <memory>
@@ -71,10 +71,8 @@ void propagate_constants::run(program_impl& p) {
         auto& id_to_replace = cout.first;
         auto mem_impl = cout.second;
 
-        memory api_memory = memory(mem_impl.detach());
-
         auto const_data =
-            std::make_shared<data>("_cldnn_const_prop_" + id_to_replace, api_memory /* <<< REMOVE ME WHEN POSSIBLE */);
+            std::make_shared<data>("_cldnn_const_prop_" + id_to_replace, mem_impl /* <<< REMOVE ME WHEN POSSIBLE */);
         auto& new_node = p.get_or_create(const_data);
         auto& curr_node = p.get_node(id_to_replace);
 
@@ -109,21 +107,22 @@ bool propagate_constants::has_non_const_user(program_node& node) const {
     return false;
 }
 
-std::list<std::pair<primitive_id, memory_impl::ptr>> propagate_constants::calculate(engine_impl& engine, build_options bo) {
+std::list<std::pair<primitive_id, memory::ptr>> propagate_constants::calculate(engine& engine, build_options bo) {
     if (!has_non_trivial_constants)
         return {};
 
     bo.set_option(build_option::optimize_data(false));
     bo.set_option(build_option::outputs(const_outputs));
-    network_impl::ptr net = engine.build_network(nodes, bo, true);
-    for (auto& cin : const_inputs) net->set_input_data(cin->id(), cin->get_attached_memory());
+    network_impl::ptr net = network_impl::build_network(engine, nodes, bo, true);
+    for (auto& cin : const_inputs)
+        net->set_input_data(cin->id(), cin->get_attached_memory_ptr());
 
     net->execute({});
     net->reset_execution(true);  // wait for computations to complete
     auto outputs = net->get_outputs();
 
-    std::list<std::pair<primitive_id, memory_impl::ptr>> ret;
-    for (auto& out : outputs) ret.push_back({out->id(), (memory_impl::ptr) &out->output_memory()});
+    std::list<std::pair<primitive_id, memory::ptr>> ret;
+    for (auto& out : outputs) ret.push_back({out->id(), out->output_memory_ptr()});
 
     return ret;
 }
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/reorder_inputs.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/reorder_inputs.cpp
index 689157262ccc08..15257fe0e95c90 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/reorder_inputs.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/reorder_inputs.cpp
@@ -4,13 +4,14 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include "api/binary_convolution.hpp"
 #include "pass_manager.h"
 #include "program_node.h"
 #include "layout_optimizer.h"
 #include "program_impl.h"
 #include "program_helpers.h"
+#include "binary_convolution_inst.h"
 #include "mvn_inst.h"
+
 #include <vector>
 #include <memory>
 #include <list>
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/skipped_branch_memory_dependencies.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/skipped_branch_memory_dependencies.cpp
index bb9578142b203a..a3ac6478673d13 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/skipped_branch_memory_dependencies.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/skipped_branch_memory_dependencies.cpp
@@ -9,7 +9,7 @@
 #include "layout_optimizer.h"
 #include "program_impl.h"
 #include "program_helpers.h"
-#include "cldnn_itt.h"
+#include "runtime/cldnn_itt.hpp"
 #include <vector>
 #include <memory>
 #include <list>
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/strided_slice_optimize.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/strided_slice_optimize.cpp
index 50ddea1771a183..3546a7d427cab4 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/strided_slice_optimize.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/strided_slice_optimize.cpp
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include <src/include/error_handler.h>
+#include "cldnn/runtime/error_handler.hpp"
 #include "pass_manager.h"
 #include "program_helpers.h"
 #include "strided_slice_inst.h"
diff --git a/inference-engine/thirdparty/clDNN/src/include/activation_inst.h b/inference-engine/thirdparty/clDNN/src/include/activation_inst.h
index f5c66dac45dbc7..989b2b557242a2 100644
--- a/inference-engine/thirdparty/clDNN/src/include/activation_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/activation_inst.h
@@ -4,9 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/activation.hpp"
+#include "cldnn/primitives/activation.hpp"
 #include "primitive_inst.h"
 #include "kernel_selector/core/actual_kernels/activation/activation_kernel_base.h"
+
 #include <memory>
 #include <string>
 
@@ -49,7 +50,7 @@ class typed_primitive_inst<activation> : public typed_primitive_inst_base<activa
 public:
     typed_primitive_inst(network_impl& network, activation_node const& node);
 
-    memory_impl& slope_memory() const { return dep_memory(1); }
+    memory::ptr slope_memory() const { return dep_memory_ptr(1); }
 
     bool is_parameterized() const { return !argument.additional_params_input.empty(); }
 };
diff --git a/inference-engine/thirdparty/clDNN/src/include/arg_max_min_inst.h b/inference-engine/thirdparty/clDNN/src/include/arg_max_min_inst.h
index 48a899dbc1ccf5..605a6bb7e95d4f 100644
--- a/inference-engine/thirdparty/clDNN/src/include/arg_max_min_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/arg_max_min_inst.h
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/arg_max_min.hpp"
+#include "cldnn/primitives/arg_max_min.hpp"
 #include "primitive_inst.h"
 
 #include <memory>
diff --git a/inference-engine/thirdparty/clDNN/src/include/average_unpooling_inst.h b/inference-engine/thirdparty/clDNN/src/include/average_unpooling_inst.h
index e20df8e759750a..0eef5fe75e221d 100644
--- a/inference-engine/thirdparty/clDNN/src/include/average_unpooling_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/average_unpooling_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/average_unpooling.hpp"
+#include "cldnn/primitives/average_unpooling.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/batch_to_space_inst.h b/inference-engine/thirdparty/clDNN/src/include/batch_to_space_inst.h
index 930db1ae93f0f2..1b0feedd2f3dcd 100644
--- a/inference-engine/thirdparty/clDNN/src/include/batch_to_space_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/batch_to_space_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/batch_to_space.hpp"
+#include "cldnn/primitives/batch_to_space.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/binary_convolution_inst.h b/inference-engine/thirdparty/clDNN/src/include/binary_convolution_inst.h
index f4ce140d4f804a..ef7f7afb2a854c 100644
--- a/inference-engine/thirdparty/clDNN/src/include/binary_convolution_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/binary_convolution_inst.h
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/binary_convolution.hpp"
+#include "cldnn/primitives/binary_convolution.hpp"
 #include "primitive_inst.h"
 
 #include <memory>
@@ -62,11 +62,11 @@ class typed_primitive_inst<binary_convolution> : public typed_primitive_inst_bas
 public:
     typed_primitive_inst(network_impl& network, binary_convolution_node const& node);
 
-    memory_impl& weights_memory(size_t index) const {
+    memory::ptr weights_memory(size_t index) const {
         if (static_cast<int32_t>(index) >= node.get_split())
             throw std::range_error("weights offset too big");
 
-        return dep_memory(1 + index);
+        return dep_memory_ptr(1 + index);
     }
 };
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/border_inst.h b/inference-engine/thirdparty/clDNN/src/include/border_inst.h
index d724b426891787..ed181ed0b16c3f 100644
--- a/inference-engine/thirdparty/clDNN/src/include/border_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/border_inst.h
@@ -4,10 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-
-#include <api/border.hpp>
-
+#include "cldnn/primitives/border.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/broadcast_inst.h b/inference-engine/thirdparty/clDNN/src/include/broadcast_inst.h
index 9b8471d44db2ee..75be67373c787e 100644
--- a/inference-engine/thirdparty/clDNN/src/include/broadcast_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/broadcast_inst.h
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 
-#include <api/broadcast.hpp>
+#include "cldnn/primitives/broadcast.hpp"
 
 #include "primitive_inst.h"
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/src/include/concatenation_inst.h b/inference-engine/thirdparty/clDNN/src/include/concatenation_inst.h
index ffdb51a63de9e3..82d3d39676f9ff 100644
--- a/inference-engine/thirdparty/clDNN/src/include/concatenation_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/concatenation_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/concatenation.hpp"
+#include "cldnn/primitives/concatenation.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/condition_inst.h b/inference-engine/thirdparty/clDNN/src/include/condition_inst.h
index 8a8859b55b6f88..4bff69b6e4953b 100644
--- a/inference-engine/thirdparty/clDNN/src/include/condition_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/condition_inst.h
@@ -5,10 +5,10 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 
-#include <api/condition.hpp>
-
+#include <cldnn/primitives/condition.hpp>
 #include "network_impl.h"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
@@ -26,9 +26,10 @@ struct typed_program_node<condition> : public typed_program_node_base<condition>
 
         void set(const program_node& node) {
             add_or_change_input_layout(node);
-            _program = node.get_program().get_engine().build_program(_topology,
-                                                                     node.get_program().get_options(),
-                                                                     true);  // rebuild program
+            _program = program_impl::build_program(node.get_program().get_engine(),
+                                                   _topology,
+                                                   node.get_program().get_options(),
+                                                   true);  // rebuild program
         }
         program_impl::ptr get() const { return _program; }
 
@@ -89,8 +90,10 @@ class typed_primitive_inst<condition> : public typed_primitive_inst_base<conditi
     static std::string to_string(condition_node const& node);
     typed_primitive_inst(network_impl& network, condition_node const& node);
 
-    memory_impl& input_memory() const { return dep_memory(0); }
-    memory_impl& compare_memory() const { return dep_memory(1); }
+    memory::ptr input_memory_ptr() const { return dep_memory_ptr(0); }
+    memory::ptr compare_memory_ptr() const { return dep_memory_ptr(1); }
+    memory& input_memory() const { return dep_memory(0); }
+    memory& compare_memory() const { return dep_memory(1); }
     network_impl::ptr get_net_true() const { return _net_true; }
     network_impl::ptr get_net_false() const { return _net_false; }
     primitive_id result_id() const { return node.result_id(); }
diff --git a/inference-engine/thirdparty/clDNN/src/include/convolution_inst.h b/inference-engine/thirdparty/clDNN/src/include/convolution_inst.h
index b90d667134ce90..899ad44dc9bc2c 100644
--- a/inference-engine/thirdparty/clDNN/src/include/convolution_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/convolution_inst.h
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/convolution.hpp"
+#include "cldnn/primitives/convolution.hpp"
 #include "primitive_inst.h"
 
 #include <memory>
@@ -120,48 +120,48 @@ class typed_primitive_inst<convolution> : public typed_primitive_inst_base<convo
 public:
     typed_primitive_inst(network_impl& network, convolution_node const& node);
 
-    memory_impl& weights_memory(size_t index) const {
+    memory::ptr weights_memory(size_t index) const {
         if (node.get_groups() == 1) {
             if (static_cast<int32_t>(index) >= node.get_split())
                 throw std::range_error("weights offset too big");
-            return dep_memory(1 + index + node.get_trans_dep_offset());
+            return dep_memory_ptr(1 + index + node.get_trans_dep_offset());
         } else {  // all weights are in one buffer
-            return dep_memory(1 + node.get_trans_dep_offset());
+            return dep_memory_ptr(1 + node.get_trans_dep_offset());
         }
     }
 
-    memory_impl& bias_memory(size_t index) const {
+    memory::ptr bias_memory(size_t index) const {
         if (node.get_groups() == 1) {
             if (static_cast<int32_t>(index) >= node.get_split())
                 throw std::range_error("bias offset too big");
-            return dep_memory(1 + node.get_split() + index + node.get_trans_dep_offset());
+            return dep_memory_ptr(1 + node.get_split() + index + node.get_trans_dep_offset());
         } else {  // all bias are in one buffer
-            return dep_memory(2 + node.get_trans_dep_offset());
+            return dep_memory_ptr(2 + node.get_trans_dep_offset());
         }
     }
 
-    memory_impl& weights_zero_points_memory(size_t) const {
+    memory::ptr weights_zero_points_memory(size_t) const {
         if (node.get_split() > 1)
             throw std::range_error("Split is unsupported for quantized convolutions");
-        return dep_memory(2 + 1 * bias_term() + node.get_trans_dep_offset());
+        return dep_memory_ptr(2 + 1 * bias_term() + node.get_trans_dep_offset());
     }
 
-    memory_impl& trans_memory() const {
+    memory::ptr trans_memory() const {
         if (!node.get_trans_dep_offset())
             throw std::range_error("trans input exists only in deformable mode");
-        return dep_memory(1);
+        return dep_memory_ptr(1);
     }
 
-    memory_impl& activations_zero_points_memory(size_t) const {
+    memory::ptr activations_zero_points_memory(size_t) const {
         if (node.get_split() > 1)
             throw std::range_error("Split is unsupported for quantized convolutions");
-        return dep_memory(2 + 1 * bias_term() + 1 * weights_zero_points_term() + node.get_trans_dep_offset());
+        return dep_memory_ptr(2 + 1 * bias_term() + 1 * weights_zero_points_term() + node.get_trans_dep_offset());
     }
 
-    memory_impl& compensation_memory(size_t) const {
+    memory::ptr compensation_memory(size_t) const {
         if (node.get_split() > 1)
             throw std::range_error("Split is unsupported for quantized convolutions");
-        return dep_memory(2 + 1 * bias_term() + 1 * weights_zero_points_term() + 1*activations_zero_points_term() + node.get_trans_dep_offset());
+        return dep_memory_ptr(2 + 1 * bias_term() + 1 * weights_zero_points_term() + 1*activations_zero_points_term() + node.get_trans_dep_offset());
     }
 
     bool bias_term() const { return node.bias_term(); }
diff --git a/inference-engine/thirdparty/clDNN/src/include/crop_inst.h b/inference-engine/thirdparty/clDNN/src/include/crop_inst.h
index 4ae7a095f0b495..e8842e5bed9ebf 100644
--- a/inference-engine/thirdparty/clDNN/src/include/crop_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/crop_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/crop.hpp"
+#include "cldnn/primitives/crop.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/ctc_greedy_decoder_inst.h b/inference-engine/thirdparty/clDNN/src/include/ctc_greedy_decoder_inst.h
index 7e856ff2e0cd9f..ea5d2bcda58359 100644
--- a/inference-engine/thirdparty/clDNN/src/include/ctc_greedy_decoder_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/ctc_greedy_decoder_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/ctc_greedy_decoder.hpp"
+#include "cldnn/primitives/ctc_greedy_decoder.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/cum_sum_inst.h b/inference-engine/thirdparty/clDNN/src/include/cum_sum_inst.h
index 452076299e8b7e..fb9cecf7fc5caa 100644
--- a/inference-engine/thirdparty/clDNN/src/include/cum_sum_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/cum_sum_inst.h
@@ -4,8 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/cum_sum.hpp"
-
+#include "cldnn/primitives/cum_sum.hpp"
 #include "primitive_inst.h"
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/custom_gpu_primitive_inst.h b/inference-engine/thirdparty/clDNN/src/include/custom_gpu_primitive_inst.h
index 125dbcfc2b65b3..20c40ecdf741c6 100644
--- a/inference-engine/thirdparty/clDNN/src/include/custom_gpu_primitive_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/custom_gpu_primitive_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/custom_gpu_primitive.hpp"
+#include "cldnn/primitives/custom_gpu_primitive.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/data_inst.h b/inference-engine/thirdparty/clDNN/src/include/data_inst.h
index 311257992c9f51..e1cafad9c726ad 100644
--- a/inference-engine/thirdparty/clDNN/src/include/data_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/data_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/data.hpp"
+#include "cldnn/primitives/data.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
@@ -17,11 +18,12 @@ struct typed_program_node<data> : public typed_program_node_base<data> {
 
     typed_program_node(const std::shared_ptr<data> prim, program_impl& prog);
 
-    memory_impl& get_attached_memory() const { return *mem; }
-    void attach_memory(memory_impl& new_mem, bool invalidate_users_if_changed = true);
+    memory& get_attached_memory() const { return *mem; }
+    memory::ptr get_attached_memory_ptr() const { return mem; }
+    void attach_memory(memory::ptr new_mem, bool invalidate_users_if_changed = true);
 
 private:
-    memory_impl::ptr mem;
+    memory::ptr mem;
 };
 
 using data_node = typed_program_node<data>;
diff --git a/inference-engine/thirdparty/clDNN/src/include/deconvolution_inst.h b/inference-engine/thirdparty/clDNN/src/include/deconvolution_inst.h
index c96d179aa9800f..f8d933002440a4 100644
--- a/inference-engine/thirdparty/clDNN/src/include/deconvolution_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/deconvolution_inst.h
@@ -4,8 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/deconvolution.hpp"
+
+#include "cldnn/primitives/deconvolution.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
@@ -90,25 +92,25 @@ class typed_primitive_inst<deconvolution> : public typed_primitive_inst_base<dec
 public:
     typed_primitive_inst(network_impl& network, deconvolution_node const& node);
 
-    memory_impl& weights_memory(size_t index) const {
+    memory::ptr weights_memory(size_t index) const {
         if (node.get_groups() == 1) {
             if (static_cast<int32_t>(index) >= node.get_split())
                 throw std::range_error("weights offset too big");
-            return dep_memory(1 + index);
+            return dep_memory_ptr(1 + index);
         } else {  // all weights are in one buffer
-            return dep_memory(1);
+            return dep_memory_ptr(1);
         }
     }
 
-    memory_impl& bias_memory(size_t index) const {
+    memory::ptr bias_memory(size_t index) const {
         if (node.get_groups() == 1) {
             if (argument.bias.size() == 0 && static_cast<int32_t>(index) >= node.get_split())
                 throw std::range_error("no bias data");
             if (static_cast<int32_t>(index) > node.get_split())
                 throw std::range_error("bias offset too big");
-            return dep_memory(1 + node.get_split() + index);
+            return dep_memory_ptr(1 + node.get_split() + index);
         } else {  // all bias are in one buffer
-            return dep_memory(2);
+            return dep_memory_ptr(2);
         }
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/deformable_convolution_inst.h b/inference-engine/thirdparty/clDNN/src/include/deformable_convolution_inst.h
index 63c4556b65fa32..c7889f7860016e 100644
--- a/inference-engine/thirdparty/clDNN/src/include/deformable_convolution_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/deformable_convolution_inst.h
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/convolution.hpp"
+#include "cldnn/primitives/convolution.hpp"
 #include "primitive_inst.h"
 
 #include <memory>
@@ -75,23 +75,23 @@ class typed_primitive_inst<deformable_conv> : public typed_primitive_inst_base<d
 public:
     typed_primitive_inst(network_impl& network, deformable_conv_node const& node);
 
-    memory_impl& weights_memory(size_t index) const {
+    memory::ptr weights_memory(size_t index) const {
         if (node.get_groups() == 1) {
             if (static_cast<int32_t>(index) >= node.get_split())
                 throw std::range_error("weights offset too big");
-            return dep_memory(1 + index);
+            return dep_memory_ptr(1 + index);
         } else {  // all weights are in one buffer
-            return dep_memory(1);
+            return dep_memory_ptr(1);
         }
     }
 
-    memory_impl& bias_memory(size_t index) const {
+    memory::ptr bias_memory(size_t index) const {
         if (node.get_groups() == 1) {
             if (static_cast<int32_t>(index) >= node.get_split())
                 throw std::range_error("bias offset too big");
-            return dep_memory(1 + node.get_split());
+            return dep_memory_ptr(1 + node.get_split());
         } else {  // all bias are in one buffer
-            return dep_memory(2);
+            return dep_memory_ptr(2);
         }
     }
 
@@ -154,7 +154,7 @@ class typed_primitive_inst<deformable_interp> : public typed_primitive_inst_base
 public:
     typed_primitive_inst(network_impl& network, deformable_interp_node const& node);
 
-    memory_impl& trans_memory() const { return dep_memory(1); }
+    memory& trans_memory() const { return dep_memory(1); }
 };
 
 using deformable_interp_inst = typed_primitive_inst<deformable_interp>;
diff --git a/inference-engine/thirdparty/clDNN/src/include/depth_to_space_inst.h b/inference-engine/thirdparty/clDNN/src/include/depth_to_space_inst.h
index 0d85e72663b30f..f551ee7ae341c5 100644
--- a/inference-engine/thirdparty/clDNN/src/include/depth_to_space_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/depth_to_space_inst.h
@@ -4,9 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/depth_to_space.hpp"
+#include "cldnn/primitives/depth_to_space.hpp"
 #include "primitive_inst.h"
 #include "kernel_selector/core/actual_kernels/depth_to_space/depth_to_space_kernel_base.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/detection_output_inst.h b/inference-engine/thirdparty/clDNN/src/include/detection_output_inst.h
index 398767ed581b1d..6ce0a82eb4b9b7 100644
--- a/inference-engine/thirdparty/clDNN/src/include/detection_output_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/detection_output_inst.h
@@ -4,9 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/detection_output.hpp"
+#include "cldnn/primitives/detection_output.hpp"
 #include "primitive_inst.h"
 #include "topology_impl.h"
+
 #include <string>
 
 #define PRIOR_BOX_SIZE 4  // Each prior-box consists of [xmin, ymin, xmax, ymax].
@@ -41,9 +42,9 @@ class typed_primitive_inst<detection_output> : public typed_primitive_inst_base<
 public:
     typed_primitive_inst(network_impl& network, detection_output_node const& node);
 
-    memory_impl& location_memory() const { return dep_memory(0); }
-    memory_impl& confidence_memory() const { return dep_memory(1); }
-    memory_impl& prior_box_memory() const { return dep_memory(2); }
+    memory::ptr location_memory() const { return dep_memory_ptr(0); }
+    memory::ptr confidence_memory() const { return dep_memory_ptr(1); }
+    memory::ptr prior_box_memory() const { return dep_memory_ptr(2); }
 };
 
 using detection_output_inst = typed_primitive_inst<detection_output>;
diff --git a/inference-engine/thirdparty/clDNN/src/include/device_impl.h b/inference-engine/thirdparty/clDNN/src/include/device_impl.h
deleted file mode 100644
index 9c692ae49bc110..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/include/device_impl.h
+++ /dev/null
@@ -1,105 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include <cl2_wrapper.h>
-#include "gpu/device_info.h"
-#include "api/device.hpp"
-#include "refcounted_obj.h"
-#include "gpu/configuration.h"
-
-#include <map>
-#include <string>
-#include <vector>
-#include <algorithm>
-
-namespace cldnn {
-enum class allocation_type {
-    unknown,     // Not specified (i.e simple_attached_memory class).
-    cl_mem,      // Use standard OpenCL cl_mem allocations.
-    usm_host,    // Accessible only by host. Not Migratable
-    usm_shared,  // Accessible by host and device. Migrtable.
-    usm_device,  // Accessible only by device. Not migratable.
-};
-
-struct device_impl;
-
-class memory_capabilities {
-public:
-    memory_capabilities(bool support_usm, const cl::Device& cl_dev) : _caps({ allocation_type::cl_mem }) {
-        if (support_usm) {
-            if (does_device_support(CL_DEVICE_HOST_MEM_CAPABILITIES_INTEL, cl_dev)) {
-                _caps.push_back(allocation_type::usm_host);
-            }
-            if (does_device_support(CL_DEVICE_SINGLE_DEVICE_SHARED_MEM_CAPABILITIES_INTEL, cl_dev)) {
-                _caps.push_back(allocation_type::usm_shared);
-            }
-            if (does_device_support(CL_DEVICE_DEVICE_MEM_CAPABILITIES_INTEL, cl_dev)) {
-                 _caps.push_back(allocation_type::usm_device);
-            }
-        }
-    }
-
-    bool supports_usm() const { return find_in_caps(allocation_type::cl_mem) && _caps.size() > 1; }
-    bool support_allocation_type(allocation_type type) const { return find_in_caps(type); }
-
-    static bool is_usm_type(allocation_type type) {
-        if (type == allocation_type::usm_host ||
-            type == allocation_type::usm_shared ||
-            type == allocation_type::usm_device)
-            return true;
-        return false;
-    }
-
-private:
-    std::vector<allocation_type> _caps;
-
-    bool does_device_support(int32_t param, const cl::Device& device) {
-        cl_device_unified_shared_memory_capabilities_intel capabilities;
-        auto err = clGetDeviceInfo(device.get(), param, sizeof(cl_device_unified_shared_memory_capabilities_intel), &capabilities, NULL);
-        if (err) throw std::runtime_error("[CLDNN ERROR]. clGetDeviceInfo error " + std::to_string(err));
-        return !((capabilities & CL_UNIFIED_SHARED_MEMORY_ACCESS_INTEL) == 0u);
-    }
-
-    bool find_in_caps(const allocation_type& type) const {
-        return std::find_if(_caps.begin(), _caps.end(), [&](const allocation_type& t) { return t == type; }) != _caps.end();
-    }
-};
-
-
-struct device_impl : public refcounted_obj<device_impl> {
-public:
-    explicit device_impl(const cl::Device dev, const cl::Context& ctx, const cl_platform_id platform, const gpu::device_info_internal& info)
-        : _context(ctx), _device(dev), _platform(platform), _info(info), _mem_caps(_info.supports_usm, _device) { }
-
-    gpu::device_info_internal get_info() const { return _info; }
-    cl::Device get_device() const { return _device; }
-    cl::Context get_context() const { return _context; }
-    cl_platform_id get_platform() const { return _platform; }
-    memory_capabilities mem_caps() const { return _mem_caps; }
-
-    ~device_impl() = default;
-
-private:
-    cl::Context _context;
-    cl::Device _device;
-    cl_platform_id _platform;
-    gpu::device_info_internal _info;
-    memory_capabilities _mem_caps;
-};
-
-struct device_query_impl : public refcounted_obj<device_query_impl> {
-public:
-    explicit device_query_impl(void* user_context = nullptr, void* user_device = nullptr);
-
-    std::map<std::string, device_impl::ptr> get_available_devices() const {
-        return _available_devices;
-    }
-
-    ~device_query_impl() = default;
-private:
-    std::map<std::string, device_impl::ptr> _available_devices;
-};
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/eltwise_inst.h b/inference-engine/thirdparty/clDNN/src/include/eltwise_inst.h
index f1c72138fad089..9b62e0b73b2f64 100644
--- a/inference-engine/thirdparty/clDNN/src/include/eltwise_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/eltwise_inst.h
@@ -4,11 +4,12 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/eltwise.hpp"
+#include "cldnn/primitives/eltwise.hpp"
 #include "primitive_inst.h"
-#include <memory>
 #include "topology_impl.h"
 #include "kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_base.h"
+
+#include <memory>
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/embedding_bag_inst.h b/inference-engine/thirdparty/clDNN/src/include/embedding_bag_inst.h
index 1b5982083ffa6a..79151251ce97cb 100644
--- a/inference-engine/thirdparty/clDNN/src/include/embedding_bag_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/embedding_bag_inst.h
@@ -4,9 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/embedding_bag.hpp"
-
+#include "cldnn/primitives/embedding_bag.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/engine_impl.h b/inference-engine/thirdparty/clDNN/src/include/engine_impl.h
deleted file mode 100644
index addc1e1daf0542..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/include/engine_impl.h
+++ /dev/null
@@ -1,131 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include "api/memory.hpp"
-#include "event_impl.h"
-#include "refcounted_obj.h"
-#include "implementation_map.h"
-#include "memory_pool.h"
-#include "device_impl.h"
-
-#include <memory>
-#include <set>
-#include <vector>
-#include <utility>
-#include <string>
-
-namespace cldnn {
-namespace gpu {
-class gpu_toolkit;
-}
-
-class build_options;
-using gpu_toolkit = gpu::gpu_toolkit;
-
-struct memory_impl;
-struct event_impl;
-struct topology_impl;
-struct program_impl;
-struct network_impl;
-struct program_node;
-
-template <class>
-struct typed_program_node;
-
-struct engine_impl : public refcounted_obj<engine_impl> {
-public:
-    explicit engine_impl(const device_impl& dev, const engine_configuration& conf);
-    ~engine_impl();
-    engine_types type() const { return engine_types::ocl; }
-    refcounted_obj_ptr<memory_impl> allocate_memory(const layout& layout, uint32_t net_id, bool reset = true);
-    refcounted_obj_ptr<memory_impl> allocate_memory(const layout& layout, allocation_type type, uint32_t net_id = 0, bool reset = true);
-    refcounted_obj_ptr<memory_impl> allocate_memory(const layout& layout,
-                                                    primitive_id,
-                                                    uint32_t network_id,
-                                                    std::set<primitive_id>,
-                                                    allocation_type type,
-                                                    bool reusable = true);
-    refcounted_obj_ptr<memory_impl> reinterpret_buffer(const memory_impl& memory, const layout& new_layout);
-    refcounted_obj_ptr<memory_impl> reinterpret_handle(const layout& new_layout,
-                                                       const shared_mem_params* params,
-                                                       uint32_t net_id);
-    bool is_the_same_buffer(const memory_impl& mem1, const memory_impl& mem2);
-
-    refcounted_obj_ptr<event_impl> create_user_event(uint32_t net_id, bool set = false);
-    void wait_for_events(std::vector<event_impl::ptr> const& events);
-
-    refcounted_obj_ptr<program_impl> build_program(const topology_impl& topology,
-                                                   const build_options& options,
-                                                   bool is_internal = false,
-                                                   bool no_optimizations = false);
-    refcounted_obj_ptr<program_impl> build_program(const std::set<std::shared_ptr<program_node>>& nodes,
-                                                   const build_options& options,
-                                                   bool is_internal);
-    void compile_program(program_impl& prog);
-
-    refcounted_obj_ptr<network_impl> allocate_network(const program_impl& program,
-                                                      uint16_t stream_id,
-                                                      bool is_internal = false);
-    refcounted_obj_ptr<network_impl> build_network(const topology_impl& topology,
-                                                   const build_options& options,
-                                                   uint16_t stream_id,
-                                                   bool is_internal = false);
-    refcounted_obj_ptr<network_impl> build_network(const std::set<std::shared_ptr<program_node>>& nodes,
-                                                   const build_options& options,
-                                                   bool is_internal);
-    void flush_network(uint32_t net_id);
-    void release_pending_memory(uint32_t net_id);
-
-    template <class T>
-    std::unique_ptr<primitive_impl> create_primitive_impl(typed_program_node<T> const& node) {
-        if (&node.get_program().get_engine() != this)
-            throw std::invalid_argument(
-                "engine_impl::create_primitive_impl: program's engine does not match called engine");
-
-        auto factory = implementation_map<T>::get(type(), node);
-        return std::move(std::unique_ptr<primitive_impl>(factory(node)));
-    }
-
-    template <class T>
-    bool does_an_implementation_exist(typed_program_node<T> const& node) {
-        if (&node.get_program().get_engine() != this)
-            throw std::invalid_argument(
-                "engine_impl::create_primitive_impl: program's engine does not match called engine");
-        return implementation_map<T>::check(type(), node);
-    }
-
-    template <class T>
-    bool does_possible_implementation_exist(typed_program_node<T> const& node) {
-        if (&node.get_program().get_engine() != this)
-            throw std::invalid_argument(
-                "engine_impl::create_primitive_impl: program's engine does not match called engine");
-        return implementation_map<T>::check_io_eq(type(), node);
-    }
-
-    const engine_configuration& configuration() const { return _configuration; }
-    void set_mem_pool(bool flag) { _configuration.enable_memory_pool = flag; }
-    std::shared_ptr<gpu_toolkit> get_context() const { return _context; }
-    gpu::device_info_internal get_device_info() const;
-    void* get_user_context() const;
-    memory_pool& get_memory_pool() { return _memory_pool; }
-
-    uint64_t get_max_used_device_memory() const { return _memory_pool.get_max_peak_device_memory_used(); }
-    uint64_t get_used_device_memory() const { return _memory_pool.get_temp_memory_used(); }
-
-    void dump_memory_pool(const program_impl& program, std::string& path, std::string& dependencies) {
-        _memory_pool.dump_memory_pool(program, path, dependencies);
-    }
-    bool use_memory_pool() const;
-    bool use_unified_shared_memory() const;
-    bool supports_allocation(allocation_type type) const;
-    allocation_type get_lockable_preffered_memory_allocation_type(bool is_image_layout = false) const;
-
-private:
-    engine_configuration _configuration;
-    std::shared_ptr<gpu_toolkit> _context;
-    memory_pool _memory_pool;
-};
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/extract_image_patches_inst.h b/inference-engine/thirdparty/clDNN/src/include/extract_image_patches_inst.h
index 00ec93d57dbe6d..2eb4dd7f7926bd 100644
--- a/inference-engine/thirdparty/clDNN/src/include/extract_image_patches_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/extract_image_patches_inst.h
@@ -4,8 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-
-#include "api/extract_image_patches.hpp"
+#include "cldnn/primitives/extract_image_patches.hpp"
 #include "primitive_inst.h"
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/fully_connected_inst.h b/inference-engine/thirdparty/clDNN/src/include/fully_connected_inst.h
index 75e00f7dd92015..db3a156dec5aae 100644
--- a/inference-engine/thirdparty/clDNN/src/include/fully_connected_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/fully_connected_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/fully_connected.hpp"
+#include "cldnn/primitives/fully_connected.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
@@ -37,8 +38,8 @@ class typed_primitive_inst<fully_connected> : public typed_primitive_inst_base<f
 public:
     typed_primitive_inst(network_impl& network, fully_connected_node const& node);
 
-    memory_impl& weights_memory() const { return dep_memory(1); }
-    memory_impl& bias_memory() const { return dep_memory(2); }
+    memory::ptr weights_memory() const { return dep_memory_ptr(1); }
+    memory::ptr bias_memory() const { return dep_memory_ptr(2); }
 
     bool bias_term() const { return !argument.bias.empty(); }
 };
diff --git a/inference-engine/thirdparty/clDNN/src/include/fused_conv_eltwise_inst.h b/inference-engine/thirdparty/clDNN/src/include/fused_conv_eltwise_inst.h
index 6f4aa5a47a0c30..65b23e3d8ecd6d 100644
--- a/inference-engine/thirdparty/clDNN/src/include/fused_conv_eltwise_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/fused_conv_eltwise_inst.h
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api_extension/fused_conv_eltwise.hpp"
+#include "cldnn/primitives/fused_conv_eltwise.hpp"
 #include "primitive_inst.h"
 
 #include <memory>
@@ -83,18 +83,18 @@ class typed_primitive_inst<fused_conv_eltwise> : public typed_primitive_inst_bas
 public:
     typed_primitive_inst(network_impl& network, fused_conv_eltwise_node const& node);
 
-    memory_impl& weights_memory(size_t index) const {
+    memory::ptr weights_memory(size_t index) const {
         if (static_cast<int32_t>(index) >= node.get_split())
             throw std::range_error("weights offset too big");
 
-        return dep_memory(2 + index);
+        return dep_memory_ptr(2 + index);
     }
 
-    memory_impl& bias_memory(size_t index) const {
+    memory::ptr bias_memory(size_t index) const {
         if (static_cast<int32_t>(index) >= node.get_split())
             throw std::range_error("bias offset too big");
 
-        return dep_memory(2 + node.get_split() + index);
+        return dep_memory_ptr(2 + node.get_split() + index);
     }
 
     bool bias_term() const { return node.bias_term(); }
diff --git a/inference-engine/thirdparty/clDNN/src/include/gather_inst.h b/inference-engine/thirdparty/clDNN/src/include/gather_inst.h
index f8e2c924cea462..9acd82a499c303 100644
--- a/inference-engine/thirdparty/clDNN/src/include/gather_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/gather_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/gather.hpp"
+#include "cldnn/primitives/gather.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/gather_nd_inst.h b/inference-engine/thirdparty/clDNN/src/include/gather_nd_inst.h
index b8732f7171c635..86a3b0b4cec956 100644
--- a/inference-engine/thirdparty/clDNN/src/include/gather_nd_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/gather_nd_inst.h
@@ -1,22 +1,9 @@
-/*
-// Copyright (c) 2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
 //
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-*/
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
+
 #pragma once
-#include "api/gather_nd.hpp"
+#include "cldnn/primitives/gather_nd.hpp"
 #include "primitive_inst.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/gather_tree_inst.h b/inference-engine/thirdparty/clDNN/src/include/gather_tree_inst.h
index cb18b6e713391e..589c004af1a095 100644
--- a/inference-engine/thirdparty/clDNN/src/include/gather_tree_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/gather_tree_inst.h
@@ -3,10 +3,9 @@
 //
 
 #pragma once
-
-#include <api/gather_tree.hpp>
-
+#include "cldnn/primitives/gather_tree.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/gemm_inst.h b/inference-engine/thirdparty/clDNN/src/include/gemm_inst.h
index 501f3029c566c6..670edf130639cf 100644
--- a/inference-engine/thirdparty/clDNN/src/include/gemm_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/gemm_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/gemm.hpp"
+#include "cldnn/primitives/gemm.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/generic_layer.hpp b/inference-engine/thirdparty/clDNN/src/include/generic_layer.hpp
index 7809c2d58f3722..79f0616b9161e1 100644
--- a/inference-engine/thirdparty/clDNN/src/include/generic_layer.hpp
+++ b/inference-engine/thirdparty/clDNN/src/include/generic_layer.hpp
@@ -4,9 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/primitive.hpp"
-#include "api/memory.hpp"
+#include "cldnn/primitives/primitive.hpp"
+#include "cldnn/runtime/memory.hpp"
 #include "kernel_selector_helper.h"
+
 #include <vector>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/generic_layer_inst.h b/inference-engine/thirdparty/clDNN/src/include/generic_layer_inst.h
index e04fa801f3ed47..32c90371abb0d4 100644
--- a/inference-engine/thirdparty/clDNN/src/include/generic_layer_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/generic_layer_inst.h
@@ -6,6 +6,7 @@
 #pragma once
 #include "generic_layer.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/grn_inst.h b/inference-engine/thirdparty/clDNN/src/include/grn_inst.h
index 1023d3ba4fa7bd..c470a44d9dd9a9 100644
--- a/inference-engine/thirdparty/clDNN/src/include/grn_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/grn_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/grn.hpp"
+#include "cldnn/primitives/grn.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/implementation_map.h b/inference-engine/thirdparty/clDNN/src/include/implementation_map.h
index 44b36bc94781ba..1a6bb51600682f 100644
--- a/inference-engine/thirdparty/clDNN/src/include/implementation_map.h
+++ b/inference-engine/thirdparty/clDNN/src/include/implementation_map.h
@@ -132,6 +132,8 @@ class implementation_map {
     using factory_type = std::function<primitive_impl*(const typed_program_node<primitive_kind>&)>;
     using map_type = singleton_map<key_type, factory_type>;
 
+    // TODO: Replace enigne_type here with impl_type
+    // And add a check that engine do support specific impl_type
     static factory_type get(engine_types engine_type, const typed_program_node<primitive_kind>& primitive) {
         // lookup in database; throw if not found
         auto key = key_builder()(engine_type, primitive);
diff --git a/inference-engine/thirdparty/clDNN/src/include/input_layout_inst.h b/inference-engine/thirdparty/clDNN/src/include/input_layout_inst.h
index 3db3e7db12a19a..c6199634634a01 100644
--- a/inference-engine/thirdparty/clDNN/src/include/input_layout_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/input_layout_inst.h
@@ -4,13 +4,14 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/input_layout.hpp"
+#include "cldnn/primitives/input_layout.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
 namespace cldnn {
-struct memory_impl;
+struct memory;
 
 template <>
 struct typed_program_node<input_layout> : public typed_program_node_base<input_layout> {
@@ -33,7 +34,7 @@ class typed_primitive_inst<input_layout> : public typed_primitive_inst_base<inpu
 public:
     typed_primitive_inst(network_impl& network, input_layout_node const& node);
 
-    void set_data(memory_impl& mem);
+    void set_data(memory::ptr mem);
 };
 
 using input_layout_inst = typed_primitive_inst<input_layout>;
diff --git a/inference-engine/thirdparty/clDNN/src/include/internal_primitive.h b/inference-engine/thirdparty/clDNN/src/include/internal_primitive.h
deleted file mode 100644
index 927871b63d6b72..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/include/internal_primitive.h
+++ /dev/null
@@ -1,36 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-
-#include "api/primitive.hpp"
-
-#include "primitive_type.h"
-
-namespace cldnn {
-
-struct internal_primitive : public primitive {
-public:
-    // a helper structure which returns true when compared with any primitive_type which is internal
-    struct internal_primitive_generic_type {
-        friend bool operator==(internal_primitive_generic_type, primitive_type_id type) {
-            return type->is_internal_type();
-        }
-
-        friend bool operator==(primitive_type_id type, internal_primitive_generic_type) {
-            return type->is_internal_type();
-        }
-
-        friend bool operator==(internal_primitive_generic_type, internal_primitive_generic_type) { return true; }
-    };
-
-    static internal_primitive_generic_type type_id() { return {}; }
-
-private:
-    internal_primitive() = delete;
-    internal_primitive(internal_primitive const&) = delete;
-    internal_primitive(internal_primitive&&) = delete;
-};
-
-}  // namespace cldnn
\ No newline at end of file
diff --git a/inference-engine/thirdparty/clDNN/src/include/internal_primitive_type_base.h b/inference-engine/thirdparty/clDNN/src/include/internal_primitive_type_base.h
deleted file mode 100644
index 64aa1342a1fa93..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/include/internal_primitive_type_base.h
+++ /dev/null
@@ -1,69 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-
-#include "meta_utils.h"
-#include "primitive_type.h"
-#include "internal_primitive.h"
-#include "program_node.h"
-#include "primitive_inst.h"
-#include <string>
-#include <memory>
-
-namespace cldnn {
-
-template <class PType>
-struct internal_primitive_type_base : public primitive_type {
-    static_assert(meta::is_internal_primitive<PType>::value,
-                  "Primitive type passed to internal_primitive_type_base should derive from internal_primitive");
-
-    [[noreturn]] std::shared_ptr<cldnn::program_node> create_node(program_impl&,
-                                                                  const std::shared_ptr<primitive>) const override {
-        throw std::runtime_error(
-            "Trying to create generic program_node for an internal primitive - internal primitives' nodes should be "
-            "created manually");
-    }
-
-    std::shared_ptr<cldnn::primitive_inst> create_instance(network_impl& network,
-                                                           const cldnn::program_node& node) const override {
-        if (node.type() != this)
-            throw std::invalid_argument("internal_primitive_type_base::create_instance: primitive type mismatch");
-
-        return std::make_shared<typed_primitive_inst<PType>>(network, node);
-    }
-
-    [[noreturn]] std::unique_ptr<primitive_impl> choose_impl(cldnn::engine_impl&,
-                                                             const cldnn::program_node&) const override {
-        throw std::runtime_error(
-            "primitive_type_id::choose_impl called for internal primitive - internal primitives should have manually "
-            "attached executable");
-    }
-
-    [[noreturn]] cldnn::layout calc_output_layout(const cldnn::program_node&) const override {
-        throw std::runtime_error(
-            "primitive_type_id::calc_output_layout called for internal primitive - internal primitives should have "
-            "output layouts precalculated");
-    }
-
-    std::string to_string(const cldnn::program_node& node) const override {
-        if (node.type() != this)
-            throw std::invalid_argument("primitive_type_base::to_string: primitive type mismatch");
-
-        return typed_primitive_inst<PType>::to_string(node);
-    }
-
-    bool is_internal_type() const override { return true; }
-};
-
-#define CLDNN_DEFINE_INTERNAL_PRIM(PType)                        \
-    struct PType : public internal_primitive {                   \
-        static primitive_type_id type_id() {                     \
-            static internal_primitive_type_base<PType> instance; \
-            return &instance;                                    \
-        }                                                        \
-    };                                                           \
-    using PType##_node = typed_program_node<PType>;
-
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/kernel_selector_helper.h b/inference-engine/thirdparty/clDNN/src/include/kernel_selector_helper.h
index 46926eecbca919..f97f74ebbbcda8 100644
--- a/inference-engine/thirdparty/clDNN/src/include/kernel_selector_helper.h
+++ b/inference-engine/thirdparty/clDNN/src/include/kernel_selector_helper.h
@@ -4,17 +4,17 @@
 
 #pragma once
 
-#include "api/cldnn.hpp"
-#include "api/tensor.hpp"
-#include "api/eltwise.hpp"
-#include "api/scale.hpp"
-#include "api/quantize.hpp"
-#include "api/activation.hpp"
+#include "cldnn/runtime/utils.hpp"
+#include "cldnn/runtime/tensor.hpp"
+#include "cldnn/runtime/error_handler.hpp"
+#include "cldnn/primitives/eltwise.hpp"
+#include "cldnn/primitives/scale.hpp"
+#include "cldnn/primitives/quantize.hpp"
+#include "cldnn/primitives/activation.hpp"
 
 #include "kernel_selector_params.h"
 #include "kernel_selector_common.h"
 #include "tensor_type.h"
-#include "error_handler.h"
 
 #include <cstdint>
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/src/include/layout_optimizer.h b/inference-engine/thirdparty/clDNN/src/include/layout_optimizer.h
index 8bd0da81f09763..ab15eed29beafa 100644
--- a/inference-engine/thirdparty/clDNN/src/include/layout_optimizer.h
+++ b/inference-engine/thirdparty/clDNN/src/include/layout_optimizer.h
@@ -4,9 +4,9 @@
 
 #pragma once
 
-#include "memory_impl.h"
-#include "engine_impl.h"
-#include "meta_utils.h"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/utils.hpp"
 
 #include "data_inst.h"
 #include "reorder_inst.h"
diff --git a/inference-engine/thirdparty/clDNN/src/include/loop_inst.h b/inference-engine/thirdparty/clDNN/src/include/loop_inst.h
index b555f37e716175..e39bb90f43185a 100644
--- a/inference-engine/thirdparty/clDNN/src/include/loop_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/loop_inst.h
@@ -5,17 +5,17 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 
-#include "api/loop.hpp"
-#include "api/mutable_data.hpp"
-#include "api/input_layout.hpp"
-#include "api/memory.hpp"
+#include "cldnn/primitives/loop.hpp"
+#include "cldnn/primitives/mutable_data.hpp"
+#include "cldnn/primitives/input_layout.hpp"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/error_handler.hpp"
 
 #include "network_impl.h"
 #include "primitive_inst.h"
 #include <string>
 #include <memory>
 #include <vector>
-#include "error_handler.h"
 
 namespace cldnn {
 template<>
@@ -31,7 +31,7 @@ struct typed_program_node<loop> : public typed_program_node_base<loop> {
     bool use_current_iteration;
     bool use_execution_condition;
     mutable program_impl::ptr body_program;
-    mutable std::map<primitive_id, memory_impl::ptr> backedge_mem_impls;
+    mutable std::map<primitive_id, memory::ptr> backedge_mem_impls;
     mutable std::map<primitive_id, std::shared_ptr<mutable_data>> backedge_layers;
     mutable std::map<primitive_id, std::shared_ptr<memory>> backedge_mem;
 
@@ -40,8 +40,8 @@ struct typed_program_node<loop> : public typed_program_node_base<loop> {
     void setup_internal_mutabledata_node(primitive_id md_id, layout md_layout, std::vector<primitive_id> md_inputs_id = {}, uint32_t net_id = 0) const {
         if (body.get_primitives().count(md_id) == 0) {
             backedge_mem_impls[md_id] = get_program().get_engine().allocate_memory(md_layout, net_id);
-            backedge_mem[md_id] = std::make_shared<memory>(backedge_mem_impls[md_id].get());
-            backedge_layers[md_id] = std::make_shared<mutable_data>(md_id, md_inputs_id, *backedge_mem[md_id]);
+            backedge_mem[md_id] = backedge_mem_impls[md_id];
+            backedge_layers[md_id] = std::make_shared<mutable_data>(md_id, md_inputs_id, backedge_mem[md_id]);
             body.add(backedge_layers[md_id]);
         }
     }
@@ -266,7 +266,7 @@ struct typed_program_node<loop> : public typed_program_node_base<loop> {
         auto opts = get_program().get_options();
         std::vector<primitive_id> output_names_vec(output_names.begin(), output_names.end());
         opts.set_option(build_option::outputs(output_names_vec));
-        body_program = get_program().get_engine().build_program(body, opts, false);
+        body_program = program_impl::build_program(get_program().get_engine(), body, opts, false);
     }
 
     const primitive_id& get_trip_count_id() const { return get_primitive()->trip_count_id; }
@@ -298,17 +298,19 @@ class typed_primitive_inst<loop> : public typed_primitive_inst_base<loop> {
         };
         std::shared_ptr<primitive_inst> from_primitive;
         std::shared_ptr<primitive_inst> to_primitive;
-        std::vector<memory_impl::ptr> from_mems;
-        memory_impl::ptr initial_mem;
+        std::vector<memory::ptr> from_mems;
+        memory::ptr initial_mem;
+        cldnn::stream& stream;
         backedge_type type;
         size_t total_bytes;
 
         backedge_memory_mapping(
             std::shared_ptr<primitive_inst> from_primitive, std::shared_ptr<primitive_inst> to_primitive,
-            std::vector<memory_impl::ptr> from_mems, memory_impl::ptr initial_mem, backedge_type type = CONCAT_OUTPUT):
+            std::vector<memory::ptr> from_mems, memory::ptr initial_mem, cldnn::stream& stream, backedge_type type = CONCAT_OUTPUT):
             from_primitive(from_primitive),
             to_primitive(to_primitive),
             from_mems(from_mems),
+            stream(stream),
             type(type),
             total_bytes(initial_mem->get_layout().bytes_count()) {
                 validate_backedge_memory();
@@ -316,11 +318,12 @@ class typed_primitive_inst<loop> : public typed_primitive_inst_base<loop> {
 
         backedge_memory_mapping(
             std::shared_ptr<primitive_inst> from_primitive, std::shared_ptr<primitive_inst> to_primitive,
-            memory_impl::ptr from_mem, memory_impl::ptr initial_mem, backedge_type type = SINGLE_SHARED):
+            memory::ptr from_mem, memory::ptr initial_mem, cldnn::stream& stream, backedge_type type = SINGLE_SHARED):
             from_primitive(from_primitive),
             to_primitive(to_primitive),
             from_mems{from_mem},
             initial_mem(initial_mem),
+            stream(stream),
             type(type),
             total_bytes(initial_mem->get_layout().bytes_count()) {
                 validate_backedge_memory();
@@ -328,10 +331,11 @@ class typed_primitive_inst<loop> : public typed_primitive_inst_base<loop> {
 
         backedge_memory_mapping(
             std::shared_ptr<primitive_inst> from_primitive, std::shared_ptr<primitive_inst> to_primitive,
-            memory_impl::ptr initial_mem, backedge_type type = SINGLE):
+            memory::ptr initial_mem, cldnn::stream& stream, backedge_type type = SINGLE):
             from_primitive(from_primitive),
             to_primitive(to_primitive),
             initial_mem(initial_mem),
+            stream(stream),
             type(type),
             total_bytes(initial_mem->get_layout().bytes_count()) {
                 validate_backedge_memory();
@@ -340,22 +344,22 @@ class typed_primitive_inst<loop> : public typed_primitive_inst_base<loop> {
         void setup_iteration(int64_t iter) const {
             if (type == CONCAT_OUTPUT) {
                 if (iter == 0) {
-                    to_primitive->set_output_memory(*initial_mem);
+                    to_primitive->set_output_memory(initial_mem);
                 } else if (iter > 0) {
-                    to_primitive->set_output_memory(*from_mems.at(iter - 1));
+                    to_primitive->set_output_memory(from_mems.at(iter - 1));
                 } else {
                     throw std::runtime_error("Invalid iteraton count" + std::to_string(iter));
                 }
             } else if (type == SINGLE_SHARED && iter == 0) {
-                copy_data(initial_mem, from_mems.front());
+                from_mems.front()->copy_from(stream, *initial_mem);
             } else if (type == SINGLE) {
-                memory_impl::ptr mem1 = (memory_impl::ptr)&to_primitive->output_memory();
+                memory::ptr mem1 = to_primitive->output_memory_ptr();
                 if (iter == 0) {
-                    copy_data(initial_mem, mem1);
+                    mem1->copy_from(stream, *initial_mem);
                 } else {
-                    memory_impl::ptr mem2 = (memory_impl::ptr)&from_primitive->output_memory();
-                    to_primitive->set_output_memory(*mem2);
-                    from_primitive->set_output_memory(*mem1);
+                    memory::ptr mem2 = from_primitive->output_memory_ptr();
+                    to_primitive->set_output_memory(mem2);
+                    from_primitive->set_output_memory(mem1);
                 }
             }
         }
@@ -370,26 +374,20 @@ class typed_primitive_inst<loop> : public typed_primitive_inst_base<loop> {
                 }
             }
         }
-
-        void copy_data(cldnn::memory_impl::ptr src_mem, cldnn::memory_impl::ptr dst_mem) const {
-            mem_lock<uint8_t> from_lock {src_mem};
-            mem_lock<uint8_t> to_lock {dst_mem};
-            const auto src = from_lock.begin();
-            const auto dst = to_lock.begin();
-            std::copy(src, src + total_bytes, dst);
-        }
     };
 
     struct concatenated_memory_mapping {
         concatenated_memory_mapping(int64_t axis,
-                            memory_impl::ptr concatenated_mem,
-                            std::vector<memory_impl::ptr> sliced_mems,
-                            int64_t iteration_elements = 0,
-                            int64_t stride = 0,
-                            int64_t initial_offset = 0) :
+                                    memory::ptr concatenated_mem,
+                                    std::vector<memory::ptr> sliced_mems,
+                                    stream& stream,
+                                    int64_t iteration_elements = 0,
+                                    int64_t stride = 0,
+                                    int64_t initial_offset = 0) :
             axis(axis),
             concatenated_mem(concatenated_mem),
             sliced_mems(sliced_mems),
+            stream(stream),
             bytes_per_element(data_type_traits::size_of(concatenated_mem->get_layout().data_type)),
             batch_size(get_batch_size(concatenated_mem->get_layout(), axis)),
             bytes_batch_stride((static_cast<int64_t>(concatenated_mem->get_layout().count()) / batch_size) * bytes_per_element),
@@ -410,13 +408,13 @@ class typed_primitive_inst<loop> : public typed_primitive_inst_base<loop> {
         }
 
         void restore_concatenated_mem() const {
-            mem_lock<uint8_t> concat_mem_lock{ concatenated_mem };
+            mem_lock<uint8_t> concat_mem_lock{ concatenated_mem, stream };
             int64_t iteration_offset = bytes_iteration_initial_offset;
             for (const auto& sliced_mem : sliced_mems) {
                 for (int64_t batch = 0; batch < batch_size; ++batch) {
                     const int64_t src_offset = batch * bytes_iteration;
                     const int64_t dst_offset = batch * bytes_batch_stride + iteration_offset;
-                    mem_lock<uint8_t> sliced_mem_lock{ sliced_mem };
+                    mem_lock<uint8_t> sliced_mem_lock{ sliced_mem, stream };
                     uint8_t* src = sliced_mem_lock.data() + src_offset;
                     uint8_t* dst = concat_mem_lock.data() + dst_offset;
                     std::copy(src, src + bytes_iteration, dst);
@@ -427,18 +425,18 @@ class typed_primitive_inst<loop> : public typed_primitive_inst_base<loop> {
 
         void setup_concatenated_output_memory(uint64_t iteration) const {
             const auto& sliced_output_mem = sliced_mems.at(iteration);
-            concat_data_prim->set_output_memory(*sliced_output_mem);
+            concat_data_prim->set_output_memory(sliced_output_mem);
         }
 
-        memory_impl::ptr get_sliced_mem(int64_t iteration) const {
-            mem_lock<uint8_t> from_lock{ concatenated_mem };
+        memory::ptr get_sliced_mem(int64_t iteration) const {
+            mem_lock<uint8_t> from_lock{ concatenated_mem, stream };
             int64_t batch_offset = 0;
             const int64_t iteration_offset = bytes_iteration_initial_offset +
                 bytes_iteration_stride * iteration;
             for (int64_t batch = 0; batch < batch_size; ++batch) {
                 const int64_t src_offset = batch_offset + iteration_offset;
                 const int64_t dst_offset = batch * bytes_iteration;
-                mem_lock<uint8_t> to_lock{ sliced_mems.at(iteration) };
+                mem_lock<uint8_t> to_lock{ sliced_mems.at(iteration), stream };
                 const auto src = from_lock.begin() + src_offset;
                 const auto dst = to_lock.begin() + dst_offset;
                 std::copy(src, src + bytes_iteration, dst);
@@ -450,8 +448,9 @@ class typed_primitive_inst<loop> : public typed_primitive_inst_base<loop> {
         const int64_t axis;
         std::shared_ptr<primitive_inst> concat_data_prim;
         std::shared_ptr<primitive_inst> sliced_data_prim;
-        memory_impl::ptr concatenated_mem;
-        std::vector<memory_impl::ptr> sliced_mems;
+        memory::ptr concatenated_mem;
+        std::vector<memory::ptr> sliced_mems;
+        cldnn::stream& stream;
         // element size
         const int64_t bytes_per_element;
         // number of higher level of dimension of slicing axis
@@ -483,8 +482,8 @@ class typed_primitive_inst<loop> : public typed_primitive_inst_base<loop> {
 
 private:
     network_impl::ptr body_network;
-    memory_impl::ptr get_external_memory(const primitive_id& external_id) const;
-    std::vector<memory_impl::ptr> get_sliced_mem(const primitive_id& internal_id) const;
+    memory::ptr get_external_memory(const primitive_id& external_id) const;
+    std::vector<memory::ptr> get_sliced_mem(const primitive_id& internal_id) const;
 };
 
 using loop_inst = typed_primitive_inst<loop>;
diff --git a/inference-engine/thirdparty/clDNN/src/include/lrn_inst.h b/inference-engine/thirdparty/clDNN/src/include/lrn_inst.h
index 47df087be8f6e5..a7bcd3f33c0744 100644
--- a/inference-engine/thirdparty/clDNN/src/include/lrn_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/lrn_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/lrn.hpp"
+#include "cldnn/primitives/lrn.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_input_inst.h b/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_input_inst.h
index 5676fe49914ba8..9ad70dffef0681 100644
--- a/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_input_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_input_inst.h
@@ -4,9 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api_extension/lstm_dynamic_input.hpp"
+#include "cldnn/primitives/lstm_dynamic_input.hpp"
 #include "primitive_inst.h"
-#include "error_handler.h"
+
 #include <memory>
 #include <string>
 
@@ -47,11 +47,11 @@ class typed_primitive_inst<lstm_dynamic_input> : public typed_primitive_inst_bas
 public:
     typed_primitive_inst(network_impl& network, lstm_dynamic_input_node const& node);
 
-    memory_impl& dyn_length_memory() const { return dep_memory(1); }
-    memory_impl& weights_memory() const { return dep_memory(2); }
-    memory_impl& bias_memory() const {
+    memory::ptr dyn_length_memory() const { return dep_memory_ptr(1); }
+    memory::ptr weights_memory() const { return dep_memory_ptr(2); }
+    memory::ptr bias_memory() const {
         CLDNN_ERROR_BOOL(id(), "Bias term", !bias_term(), "Trying to get non existing bias memory.");
-        return dep_memory(3);
+        return dep_memory_ptr(3);
     }
     int32_t direction() const { return node.direction(); }
     bool bias_term() const { return node.bias_term(); }
diff --git a/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_inst.h b/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_inst.h
index c64530ffae9840..fe281d35ff4ba7 100644
--- a/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_inst.h
@@ -4,9 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/lstm_dynamic.hpp"
+#include "cldnn/primitives/lstm_dynamic.hpp"
 #include "primitive_inst.h"
-#include "error_handler.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_timeloop_inst.h b/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_timeloop_inst.h
index 6330abb5efcfec..b7ded70f51bf5e 100644
--- a/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_timeloop_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/lstm_dynamic_timeloop_inst.h
@@ -4,9 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api_extension/lstm_dynamic_timeloop.hpp"
+#include "cldnn/primitives/lstm_dynamic_timeloop.hpp"
 #include "primitive_inst.h"
-#include "error_handler.h"
+
 #include <memory>
 #include <string>
 #include <vector>
@@ -68,12 +68,12 @@ class typed_primitive_inst<lstm_dynamic_timeloop> : public typed_primitive_inst_
 public:
     typed_primitive_inst(network_impl& network, lstm_dynamic_timeloop_node const& node);
 
-    memory_impl& dyn_length_memory() const { return get_dependency_memory("dyn_length"); }
-    memory_impl& recurrent_memory() const { return get_dependency_memory("recurrent"); }
-    memory_impl& last_hidden_output_memory() const { return get_dependency_memory("last_hidden_output"); }
-    memory_impl& last_cell_output_memory() const { return get_dependency_memory("last_cell_output"); }
-    memory_impl& initial_hidden_memory() const { return get_dependency_memory("initial_hidden"); }
-    memory_impl& initial_cell_memory() const { return get_dependency_memory("initial_cell"); }
+    memory::ptr dyn_length_memory() const { return get_dependency_memory("dyn_length"); }
+    memory::ptr recurrent_memory() const { return get_dependency_memory("recurrent"); }
+    memory::ptr last_hidden_output_memory() const { return get_dependency_memory("last_hidden_output"); }
+    memory::ptr last_cell_output_memory() const { return get_dependency_memory("last_cell_output"); }
+    memory::ptr initial_hidden_memory() const { return get_dependency_memory("initial_hidden"); }
+    memory::ptr initial_cell_memory() const { return get_dependency_memory("initial_cell"); }
 
     bool dyn_length_term() const { return node.dyn_length_term(); }
     bool initial_hidden_term() const { return node.initial_hidden_term(); }
@@ -82,7 +82,7 @@ class typed_primitive_inst<lstm_dynamic_timeloop> : public typed_primitive_inst_
     bool last_cell_output_term() const { return node.last_cell_output_term(); }
 
 private:
-    memory_impl& get_dependency_memory(std::string val) const { return dep_memory(node.get_dependency_idx(val)); }
+    memory::ptr get_dependency_memory(std::string val) const { return dep_memory_ptr(node.get_dependency_idx(val)); }
 };
 
 using lstm_dynamic_timeloop_inst = typed_primitive_inst<lstm_dynamic_timeloop>;
diff --git a/inference-engine/thirdparty/clDNN/src/include/lstm_elt_inst.h b/inference-engine/thirdparty/clDNN/src/include/lstm_elt_inst.h
index 7cdb1483f4d300..e87d75dba08ecc 100644
--- a/inference-engine/thirdparty/clDNN/src/include/lstm_elt_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/lstm_elt_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/lstm.hpp"
+#include "cldnn/primitives/lstm.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
@@ -43,7 +44,7 @@ class typed_primitive_inst<lstm_elt> : public typed_primitive_inst_base<lstm_elt
 public:
     typed_primitive_inst(network_impl& network, lstm_elt_node const& node);
 
-    memory_impl& cell_memory() const { return dep_memory(1); }
+    memory::ptr cell_memory() const { return dep_memory_ptr(1); }
     bool cell_term() const { return !argument.cell.empty(); }
     lstm_weights_order offset_order() const { return argument.offset_order; }
     float clip() const {
diff --git a/inference-engine/thirdparty/clDNN/src/include/lstm_gemm_inst.h b/inference-engine/thirdparty/clDNN/src/include/lstm_gemm_inst.h
index 7d88493fccb15c..bc2ae664afed45 100644
--- a/inference-engine/thirdparty/clDNN/src/include/lstm_gemm_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/lstm_gemm_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/lstm.hpp"
+#include "cldnn/primitives/lstm.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
@@ -39,10 +40,10 @@ class typed_primitive_inst<lstm_gemm> : public typed_primitive_inst_base<lstm_ge
 public:
     typed_primitive_inst(network_impl& network, lstm_gemm_node const& node);
 
-    memory_impl& weights_memory() const { return dep_memory(1); }
-    memory_impl& recurrent_memory() const { return dep_memory(2); }
-    memory_impl& bias_memory() const { return dep_memory(3); }
-    memory_impl& hidden_memory() const { return bias_term() ? dep_memory(4) : dep_memory(3); }
+    memory::ptr weights_memory() const { return dep_memory_ptr(1); }
+    memory::ptr recurrent_memory() const { return dep_memory_ptr(2); }
+    memory::ptr bias_memory() const { return dep_memory_ptr(3); }
+    memory::ptr hidden_memory() const { return bias_term() ? dep_memory_ptr(4) : dep_memory_ptr(3); }
     bool bias_term() const { return !argument.bias.empty(); }
     bool hidden_term() const { return !argument.hidden.empty(); }
     uint32_t direction() const { return argument.direction; }
diff --git a/inference-engine/thirdparty/clDNN/src/include/lstm_inst.h b/inference-engine/thirdparty/clDNN/src/include/lstm_inst.h
index 6f42da675f0833..2f94e236283af3 100644
--- a/inference-engine/thirdparty/clDNN/src/include/lstm_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/lstm_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/lstm.hpp"
+#include "cldnn/primitives/lstm.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <vector>
 
@@ -51,14 +52,14 @@ class typed_primitive_inst<lstm> : public typed_primitive_inst_base<lstm> {
 public:
     typed_primitive_inst(network_impl& network, lstm_node const& node);
 
-    memory_impl& weights_memory() const { return dep_memory(1); }
-    memory_impl& recurrent_memory() const { return dep_memory(2); }
-    memory_impl& bias_memory() const { return dep_memory(3); }
-    memory_impl& initial_hidden_memory() const { return dep_memory(bias_term() ? 4 : 3); }
-    memory_impl& initial_cell_memory() const {
+    memory& weights_memory() const { return dep_memory(1); }
+    memory& recurrent_memory() const { return dep_memory(2); }
+    memory& bias_memory() const { return dep_memory(3); }
+    memory& initial_hidden_memory() const { return dep_memory(bias_term() ? 4 : 3); }
+    memory& initial_cell_memory() const {
         return dep_memory(bias_term() ? (initial_hidden_term() ? 5 : 4) : (initial_hidden_term() ? 4 : 2));
     }
-    memory_impl& peepholes_memory() const { return dep_memory(6); }
+    memory& peepholes_memory() const { return dep_memory(6); }
     bool bias_term() const { return !argument.bias.empty(); }
     bool peepholes_term() const { return !argument.peepholes.empty(); }
     bool initial_hidden_term() const { return !argument.initial_hidden.empty(); }
diff --git a/inference-engine/thirdparty/clDNN/src/include/max_unpooling_inst.h b/inference-engine/thirdparty/clDNN/src/include/max_unpooling_inst.h
index 85079daf5b087b..543344011de040 100644
--- a/inference-engine/thirdparty/clDNN/src/include/max_unpooling_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/max_unpooling_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/max_unpooling.hpp"
+#include "cldnn/primitives/max_unpooling.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/memory_impl.h b/inference-engine/thirdparty/clDNN/src/include/memory_impl.h
deleted file mode 100644
index a0353637b3306d..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/include/memory_impl.h
+++ /dev/null
@@ -1,115 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include "api/memory.hpp"
-
-#include "engine_impl.h"
-#include "refcounted_obj.h"
-
-namespace cldnn {
-
-struct memory_impl : refcounted_obj<memory_impl> {
-    memory_impl(const engine_impl::ptr& engine, const layout& layout, uint32_t net_id,  allocation_type type, bool reused = false)
-        : _engine(engine.get()), _layout(layout), _net_id(net_id), _bytes_count(_layout.bytes_count()), _type(type), _reused(reused) {}
-
-    virtual ~memory_impl() {
-        if (_engine != nullptr && !_reused) {
-            _engine->get_memory_pool().subtract_memory_used(_bytes_count);
-        }
-    }
-    virtual void* lock() = 0;
-    virtual void unlock() = 0;
-    virtual void fill(unsigned char pattern, event_impl::ptr ev) = 0;
-    size_t size() const { return _bytes_count; }
-    virtual shared_mem_params get_internal_params() const = 0;
-    virtual bool is_allocated_by(const engine_impl& engine) const { return &engine == _engine; }
-    refcounted_obj_ptr<engine_impl> get_engine() const { return engine_impl::ptr(_engine); }
-    const layout& get_layout() const { return _layout; }
-    uint32_t get_net_id() const { return _net_id; }
-    void set_net(uint32_t id) { _net_id = id; }
-    allocation_type get_allocation_type() const { return _type; }
-    virtual bool is_memory_reset_needed(layout l) {
-        // To avoid memory reset, output memory must meet the following requirements:
-        // - To be Weights format (Data memory can be reused by memory_pool, which can lead to errors)
-        // - To have zero paddings
-        // - To be completely filled with data
-        if ((!format::is_weights_format(l.format) && !format::is_simple_data_format(l.format)) ||
-             format::is_winograd(l.format) || format::is_image_2d(l.format)) {
-            return true;
-        }
-
-        if (l.data_padding.lower_size() != tensor(0) || l.data_padding.upper_size() != tensor(0)) {
-            return true;
-        }
-
-        if (_bytes_count == (l.data_type == data_types::bin ? ceil_div(l.count(), 32) : l.count()) * data_type_traits::size_of(l.data_type)) {
-            return false;
-        }
-
-        return true;
-    }
-
-protected:
-    engine_impl *const _engine;
-    const layout _layout;
-    uint32_t _net_id;
-    size_t _bytes_count;
-
-private:
-    // layout bytes count, needed because of traits static map destruction
-    // before run of memory_impl destructor, when engine is static
-    allocation_type _type;
-    bool _reused;
-};
-
-struct simple_attached_memory : memory_impl {
-    simple_attached_memory(const layout& layout, void* pointer, uint32_t net_id)
-        : memory_impl((engine_impl::ptr) nullptr, layout, net_id, allocation_type::unknown), _pointer(pointer) {}
-
-    void* lock() override { return _pointer; }
-    void unlock() override {}
-    void fill(unsigned char, event_impl::ptr) override {}
-    shared_mem_params get_internal_params() const override { return { shared_mem_type::shared_mem_empty, nullptr, nullptr, nullptr,
-#ifdef _WIN32
-        nullptr,
-#else
-        0,
-#endif
-        0}; };
-
-private:
-    void* _pointer;
-};
-
-template <class T>
-struct mem_lock {
-    explicit mem_lock(memory_impl::ptr mem) : mem(mem), ptr(reinterpret_cast<T*>(mem->lock())) {}
-
-    explicit mem_lock(memory_impl& mem) : mem_lock((memory_impl::ptr) &mem) {}
-
-    ~mem_lock() {
-        ptr = nullptr;
-        mem->unlock();
-    }
-
-    size_t size() const { return mem->size() / sizeof(T); }
-
-#if defined(_SECURE_SCL) && (_SECURE_SCL > 0)
-    auto begin() & { return stdext::make_checked_array_iterator(ptr, size()); }
-    auto end() & { return stdext::make_checked_array_iterator(ptr, size(), size()); }
-#else
-    T* begin() & { return ptr; }
-    T* end() & { return ptr + size(); }
-#endif
-
-    T* data() const { return ptr; }
-
-private:
-    memory_impl::ptr mem;
-    T* ptr;
-};
-
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/meta_utils.h b/inference-engine/thirdparty/clDNN/src/include/meta_utils.h
index ac412a5ac0e058..7c523ef23599e3 100644
--- a/inference-engine/thirdparty/clDNN/src/include/meta_utils.h
+++ b/inference-engine/thirdparty/clDNN/src/include/meta_utils.h
@@ -4,9 +4,9 @@
 
 #pragma once
 
+#include "cldnn/runtime/utils.hpp"
+
 #include <type_traits>
-#include "api/meta_utils.hpp"
-#include "internal_primitive.h"
 
 namespace cldnn {
 
@@ -14,22 +14,6 @@ struct primitive;
 
 namespace meta {
 
-template <class... T>
-struct pack {};
-
-// helper type for deducing return type from member function pointer
-// doesn't require passing arguments like std::result_of
-template <class T>
-struct deduce_ret_type;
-
-template <class Ret, class C, class... Args>
-struct deduce_ret_type<Ret (C::*)(Args...)> {
-    using type = Ret;
-};
-
-template <class T>
-using deduce_ret_type_t = typename deduce_ret_type<T>::type;
-
 template <class T>
 struct is_primitive
     : public std::integral_constant<bool,
@@ -37,16 +21,6 @@ struct is_primitive
                                         !std::is_same<primitive, typename std::remove_cv<T>::type>::value &&
                                         std::is_same<T, typename std::remove_cv<T>::type>::value> {};
 
-template <class T>
-struct is_api_primitive
-    : public std::integral_constant<bool, is_primitive<T>::value && !std::is_base_of<internal_primitive, T>::value> {};
-
-template <class T>
-struct is_internal_primitive
-    : public std::integral_constant<bool,
-                                    std::is_base_of<internal_primitive, T>::value &&
-                                        !std::is_same<internal_primitive, typename std::remove_cv<T>::type>::value &&
-                                        std::is_same<T, typename std::remove_cv<T>::type>::value> {};
 
 }  // namespace meta
-}  // namespace cldnn
\ No newline at end of file
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/mutable_data_inst.h b/inference-engine/thirdparty/clDNN/src/include/mutable_data_inst.h
index 02ed1f1ff3b160..311873fb01556d 100644
--- a/inference-engine/thirdparty/clDNN/src/include/mutable_data_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/mutable_data_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/mutable_data.hpp"
+#include "cldnn/primitives/mutable_data.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
@@ -17,18 +18,14 @@ struct typed_program_node<mutable_data> : public typed_program_node_base<mutable
 
     typed_program_node(const std::shared_ptr<mutable_data> prim, program_impl& prog);
 
-    memory_impl& get_attached_memory() const { return *mem; }
-    memory_impl::ptr get_attached_memory_ptr() const { return mem; }
-    void attach_memory(memory_impl& new_mem, bool invalidate_users_if_changed = true);
+    memory& get_attached_memory() const { return *mem; }
+    memory::ptr get_attached_memory_ptr() const { return mem; }
+    void attach_memory(memory::ptr new_mem, bool invalidate_users_if_changed = true);
 
     program_node& input(size_t idx = 0) const { return get_dependency(idx); }
 
 private:
-    memory_impl::ptr mem;
-
-    void fill_memory();
-    void fill_memory_xavier();
-    void fill_memory_constant(float value);
+    memory::ptr mem;
 };
 
 using mutable_data_node = typed_program_node<mutable_data>;
@@ -41,7 +38,6 @@ class typed_primitive_inst<mutable_data> : public typed_primitive_inst_base<muta
     static layout calc_output_layout(mutable_data_node const& node) { return node.get_attached_memory().get_layout(); }
     static std::string to_string(mutable_data_node const& node);
 
-public:
     typed_primitive_inst(network_impl& network, mutable_data_node const& node);
 };
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/mvn_inst.h b/inference-engine/thirdparty/clDNN/src/include/mvn_inst.h
index 31b2537746e540..d4787d67608959 100644
--- a/inference-engine/thirdparty/clDNN/src/include/mvn_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/mvn_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/mvn.hpp"
+#include "cldnn/primitives/mvn.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/network_impl.h b/inference-engine/thirdparty/clDNN/src/include/network_impl.h
index a4bfcf1b26d91e..8c5c570cbd4942 100644
--- a/inference-engine/thirdparty/clDNN/src/include/network_impl.h
+++ b/inference-engine/thirdparty/clDNN/src/include/network_impl.h
@@ -5,12 +5,12 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
 
-#include "api/network.hpp"
-
-#include "engine_impl.h"
-#include "event_impl.h"
+#include "cldnn/graph/network.hpp"
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/event.hpp"
+#include "cldnn/runtime/stream.hpp"
 #include "program_impl.h"
-#include "refcounted_obj.h"
+#include "implementation_map.h"
 
 #include <map>
 #include <vector>
@@ -24,30 +24,49 @@ namespace cldnn {
 
 class primitive_inst;
 
-struct network_impl : public refcounted_obj<network_impl> {
+struct network_impl {
 public:
-    explicit network_impl(const program_impl& program, uint16_t stream_id, bool is_internal = false);
-    network_impl(engine_impl& engine,
+    using ptr = std::shared_ptr<network_impl>;
+    explicit network_impl(program_impl::ptr program, stream::ptr stream, bool is_internal = false, bool is_primary_stream = false);
+    network_impl(engine& engine,
                  const topology_impl& topo,
                  const build_options& options = build_options(),
-                 uint16_t stream_id = 0,
                  bool is_internal = false);
-    network_impl(engine_impl& engine,
+    network_impl(engine& engine,
                  const std::set<std::shared_ptr<program_node>>& nodes,
                  const build_options& options,
                  bool is_internal);
     ~network_impl();
 
-    const program_impl& get_program() const { return *_program; }
-    engine_impl& get_engine() const { return _program->get_engine(); }
+
+    static ptr build_network(engine& engine,
+                             const topology_impl& topology,
+                             const build_options& options,
+                             bool is_internal = false);
+    static ptr build_network(engine& engine,
+                             const std::set<std::shared_ptr<program_node>>& nodes,
+                             const build_options& options,
+                             bool is_internal);
+
+    static ptr allocate_network(stream::ptr stream,
+                                program_impl::ptr program,
+                                bool is_internal = false,
+                                bool is_primary_stream = false);
+
+    static ptr allocate_network(engine& engine,
+                                program_impl::ptr program,
+                                bool is_internal = false,
+                                bool is_primary_stream = false);
+    program_impl::cptr get_program() const { return _program; }
+    program_impl::ptr get_program() { return _program; }
+    engine& get_engine() const { return _program->get_engine(); }
 
     void reset_execution(bool wait = true);
-    void set_input_data(const primitive_id& id, memory_impl& data);
-    void set_output_memory(const primitive_id& id, memory_impl& mem);
+    void set_input_data(const primitive_id& id, memory::ptr data);
+    void set_output_memory(const primitive_id& id, memory::ptr mem);
 
     void set_learning_rate(const float lr);
     float get_learning_rate();
-    uint16_t get_stream_id() const { return _stream_id; }
 
     std::vector<std::shared_ptr<primitive_inst>> const& get_outputs() { return _outputs; }
 
@@ -62,31 +81,33 @@ struct network_impl : public refcounted_obj<network_impl> {
     std::vector<primitive_id> get_all_primitive_org_ids() const;
     const program_impl::primitives_info& get_primitives_info() const;
     const program_impl::graph_optimizer_info& get_optimizer_passes_info() const;
-    void execute(const std::vector<event_impl::ptr>& events);
+    void execute(const std::vector<event::ptr>& events);
     void validate_primitives();
     void set_arguments();
     // Implementation specific calls
     std::shared_ptr<primitive_inst> get_primitive(const primitive_id& id);
     std::string get_primitive_info(const primitive_id& id) const;
-    const event_impl::ptr& get_primitive_event(const primitive_id& id) const { return _events.at(id); }
+    const event::ptr& get_primitive_event(const primitive_id& id) const { return _events.at(id); }
     bool has_event(const primitive_id& id) const { return _events.count(id); }
     std::vector<std::shared_ptr<primitive_inst>> get_primitives(const std::vector<primitive_id>& ids);
     std::vector<std::shared_ptr<primitive_inst>> get_primitives(const std::vector<program_node*>& nodes);
     void execute_primitive(const std::shared_ptr<primitive_inst>& primitive,
-                           const std::vector<event_impl::ptr>& events);
+                           const std::vector<event::ptr>& events);
     void allocate_primitives();
     void build_insts_deps();
     uint32_t get_id() const { return net_id; }
+    stream& get_stream() const { return *_stream; }
+    stream::ptr get_stream_ptr() const { return _stream; }
     void build_exec_order();
     bool is_internal() const { return _internal; }
-    bool is_primary_stream();
-    bool is_secondary_stream();
+    bool is_primary_stream() { return _is_primary_stream; }
 
 private:
     uint32_t net_id = 0;
-    const program_impl::cptr _program;
-    uint16_t _stream_id;
+    program_impl::ptr _program;
+    stream::ptr _stream;
     bool _internal;
+    bool _is_primary_stream;
     bool _reset_arguments;
     float _learning_rate = static_cast<float>(0.00001);
 
@@ -96,11 +117,10 @@ struct network_impl : public refcounted_obj<network_impl> {
     std::list<std::shared_ptr<primitive_inst>> _exec_order;
     std::list<std::shared_ptr<primitive_inst>> _data_outputs;
 
-    std::unordered_map<primitive_id, event_impl::ptr> _events;
+    std::unordered_map<primitive_id, event::ptr> _events;
 
     void allocate_primitive_instance(program_node const& node);
     void transfer_memory_to_device(std::shared_ptr<primitive_inst> instance, program_node const& node);
-    void allocate_mutable_data_for_streams(std::vector<std::shared_ptr<program_node>>& mutable_data_nodes);
     void add_to_exec_order(const primitive_id& id);
     std::shared_ptr<primitive_inst> find_in_internal_networks(const primitive_id& id);
     std::shared_ptr<primitive_inst> find_primitive(const primitive_id& id);
diff --git a/inference-engine/thirdparty/clDNN/src/include/non_max_suppression_inst.h b/inference-engine/thirdparty/clDNN/src/include/non_max_suppression_inst.h
index b4ff4c3e4812ac..11e3151ec6bc09 100644
--- a/inference-engine/thirdparty/clDNN/src/include/non_max_suppression_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/non_max_suppression_inst.h
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/non_max_suppression.hpp"
+#include "cldnn/primitives/non_max_suppression.hpp"
 #include "primitive_inst.h"
 
 #include <memory>
@@ -94,62 +94,62 @@ class typed_primitive_inst<non_max_suppression> : public typed_primitive_inst_ba
     static layout calc_output_layout(non_max_suppression_node const& node);
     static std::string to_string(non_max_suppression_node const& node);
 
-    memory_impl& input_boxes_mem() const {
-        return dep_memory(0);
+    memory::ptr input_boxes_mem() const {
+        return dep_memory_ptr(0);
     }
 
-    memory_impl& input_scores_mem() const {
-        return dep_memory(1);
+    memory::ptr input_scores_mem() const {
+        return dep_memory_ptr(1);
     }
 
     bool has_num_select_per_class() const { return node.has_num_select_per_class(); }
-    memory_impl& num_select_per_class_mem() const {
-        return dep_memory(2);
+    memory::ptr num_select_per_class_mem() const {
+        return dep_memory_ptr(2);
     }
 
     bool has_iou_threshold() const { return node.has_iou_threshold(); }
-    memory_impl& iou_threshold_mem() const {
+    memory::ptr iou_threshold_mem() const {
         size_t offset = 2;
         offset += has_num_select_per_class();
-        return dep_memory(offset);
+        return dep_memory_ptr(offset);
     }
 
     bool has_score_threshold() const { return node.has_score_threshold(); }
-    memory_impl& score_threshold_mem() const {
+    memory::ptr score_threshold_mem() const {
         size_t offset = 2;
         offset += has_num_select_per_class();
         offset += has_iou_threshold();
-        return dep_memory(offset);
+        return dep_memory_ptr(offset);
     }
 
     bool has_soft_nms_sigma() const { return node.has_soft_nms_sigma(); }
-    memory_impl& soft_nms_sigma_mem() const {
+    memory::ptr soft_nms_sigma_mem() const {
         size_t offset = 2;
         offset += has_num_select_per_class();
         offset += has_iou_threshold();
         offset += has_score_threshold();
-        return dep_memory(offset);
+        return dep_memory_ptr(offset);
     }
 
     bool has_second_output() const { return node.has_second_output(); }
-    memory_impl& second_output_mem() const {
+    memory::ptr second_output_mem() const {
         size_t offset = 2;
         offset += has_num_select_per_class();
         offset += has_iou_threshold();
         offset += has_score_threshold();
         offset += has_soft_nms_sigma();
-        return dep_memory(offset);
+        return dep_memory_ptr(offset);
     }
 
     bool has_third_output() const { return node.has_third_output(); }
-    memory_impl& third_output_mem() const {
+    memory::ptr third_output_mem() const {
         size_t offset = 2;
         offset += has_num_select_per_class();
         offset += has_iou_threshold();
         offset += has_score_threshold();
         offset += has_soft_nms_sigma();
         offset += has_second_output();
-        return dep_memory(offset);
+        return dep_memory_ptr(offset);
     }
 };
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/normalize_inst.h b/inference-engine/thirdparty/clDNN/src/include/normalize_inst.h
index 7d93b24cd78fec..4f14fe2f3e88a2 100644
--- a/inference-engine/thirdparty/clDNN/src/include/normalize_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/normalize_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/normalize.hpp"
+#include "cldnn/primitives/normalize.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
@@ -34,7 +35,7 @@ class typed_primitive_inst<normalize> : public typed_primitive_inst_base<normali
 public:
     typed_primitive_inst(network_impl& network, normalize_node const& node);
 
-    memory_impl& scale_memory() const { return dep_memory(1); }
+    memory::ptr scale_memory() const { return dep_memory_ptr(1); }
 };
 
 using normalize_inst = typed_primitive_inst<normalize>;
diff --git a/inference-engine/thirdparty/clDNN/src/include/one_hot_inst.h b/inference-engine/thirdparty/clDNN/src/include/one_hot_inst.h
index 13477f9f3969f4..0db8759d4762f8 100644
--- a/inference-engine/thirdparty/clDNN/src/include/one_hot_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/one_hot_inst.h
@@ -4,10 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-
-#include <api/one_hot.hpp>
-
+#include "cldnn/primitives/one_hot.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/pass_manager.h b/inference-engine/thirdparty/clDNN/src/include/pass_manager.h
index 3faf9d996ce7e7..0047c7bec52347 100644
--- a/inference-engine/thirdparty/clDNN/src/include/pass_manager.h
+++ b/inference-engine/thirdparty/clDNN/src/include/pass_manager.h
@@ -265,7 +265,7 @@ class propagate_constants : public base_pass {
 
 private:
     void run(program_impl& p) override;
-    std::list<std::pair<primitive_id, memory_impl::ptr>> calculate(engine_impl& engine, build_options bo);
+    std::list<std::pair<primitive_id, memory::ptr>> calculate(engine& engine, build_options bo);
     bool has_non_const_user(program_node& node) const;
     void handle_constant(program_impl& prog, program_node& node);
     void add_constant(program_impl& prog, program_node& node);
diff --git a/inference-engine/thirdparty/clDNN/src/include/permute_inst.h b/inference-engine/thirdparty/clDNN/src/include/permute_inst.h
index be8b8d8049ccd6..b6250a2b54dc58 100644
--- a/inference-engine/thirdparty/clDNN/src/include/permute_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/permute_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/permute.hpp"
+#include "cldnn/primitives/permute.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/pooling_inst.h b/inference-engine/thirdparty/clDNN/src/include/pooling_inst.h
index 5fc2f05b4bd21a..6bcc208cefbd13 100644
--- a/inference-engine/thirdparty/clDNN/src/include/pooling_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/pooling_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/pooling.hpp"
+#include "cldnn/primitives/pooling.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/primitive_inst.h b/inference-engine/thirdparty/clDNN/src/include/primitive_inst.h
index 5796620722489c..80987966343c6f 100644
--- a/inference-engine/thirdparty/clDNN/src/include/primitive_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/primitive_inst.h
@@ -4,15 +4,15 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-
-#include "api/primitive.hpp"
-#include "api/concatenation.hpp"
-
-#include "event_impl.h"
-#include "memory_impl.h"
-#include "meta_utils.h"
+#include "cldnn/primitives/primitive.hpp"
+#include "cldnn/primitives/concatenation.hpp"
+#include "cldnn/runtime/error_handler.hpp"
+#include "cldnn/runtime/event.hpp"
+#include "cldnn/runtime/memory.hpp"
 #include "kernel_selector_helper.h"
+#include "meta_utils.h"
 #include "program_node.h"
+#include "primitive_type.h"
 
 #include <memory>
 #include <vector>
@@ -44,16 +44,17 @@ struct primitive_impl {
     virtual ~primitive_impl() = default;
 
     virtual void set_arguments(primitive_inst& instance) = 0;
-    virtual void cleanup(primitive_inst& instance) = 0;
-    virtual event_impl::ptr execute(const std::vector<event_impl::ptr>& events, primitive_inst& instance) = 0;
+    virtual event::ptr execute(const std::vector<event::ptr>& events, primitive_inst& instance) = 0;
     virtual bool validate(const primitive_inst& instance) const = 0;
     std::string get_kernel_name() const { return _kernel_name; }
     // TODO: added a derived class for weights reordering (maybe for all static data reordering)
     kernel_selector::weights_reorder_params _weights_reorder_params;
     // class typed_primitive_gpu_impl override this with return false;
     virtual bool is_cpu() const { return true; }
+    virtual void init_kernels() = 0;
+    virtual std::unique_ptr<primitive_impl> clone() const = 0;
 
-private:
+protected:
     std::string _kernel_name;
 };
 
@@ -74,8 +75,10 @@ class primitive_inst {
         return reinterpret_cast<std::vector<std::shared_ptr<const primitive_inst>> const&>(_deps);
     }
 
-    memory_impl& dep_memory(size_t index) const { return dependencies().at(index)->output_memory(); }
-    memory_impl& output_memory() const { return *_output; }
+    memory& dep_memory(size_t index) const { return dependencies().at(index)->output_memory(); }
+    memory::ptr dep_memory_ptr(size_t index) const { return dependencies().at(index)->output_memory_ptr(); }
+    memory& output_memory() const { return *_output; }
+    memory::ptr output_memory_ptr() const { return _output; }
     size_t inputs_memory_count() const { return _node.get_primitive()->input_size(); }
     primitive_type_id type() const { return _node.type(); }
     primitive_id id() const { return _node.id(); }
@@ -84,22 +87,28 @@ class primitive_inst {
     std::shared_ptr<const primitive> desc() const { return _node.get_primitive(); }
     network_impl& get_network() const { return _network; }
     uint32_t get_network_id() const;
-    void set_output_memory(memory_impl& mem);
-    void check_memory_to_set(const memory_impl& mem, const layout& layout) const;
+    void set_output_memory(memory::ptr mem);
+    void check_memory_to_set(const memory& mem, const layout& layout) const;
     const std::list<const cldnn::program_node *>& get_users() const { return _node.get_users(); }
 
     // return pointer to const to prevent arbitrary 'execute' call -> use primitive_inst.execute() instead
-    primitive_impl* get_impl() const { return _impl.get(); }
+    const primitive_impl* get_impl() const { return _impl.get(); }
 
-    memory_impl& input_memory(size_t index = 0) const {
+    memory& input_memory(size_t index = 0) const {
         if (index >= inputs_memory_count())
             throw std::range_error("input offset too big");
         return dep_memory(index);
     }
 
-    event_impl::ptr execute(const std::vector<event_impl::ptr>& events);
+    memory::ptr input_memory_ptr(size_t index = 0) const {
+        if (index >= inputs_memory_count())
+            throw std::range_error("input offset too big");
+        return dep_memory_ptr(index);
+    }
+
+    event::ptr execute(const std::vector<event::ptr>& events);
+    void init_kernels();
     void set_arguments();
-    void cleanup();
     bool validate() const {
         if (_impl == nullptr)
             throw std::invalid_argument("[Internal cldnn error].  Validation method for nullptr impl is not allowed.");
@@ -110,8 +119,8 @@ class primitive_inst {
 
     void build_deps();
 
-    memory_impl& fused_memory(size_t dep_id) const {
-        return dep_memory(get_fused_mem_offset() + dep_id);
+    memory::ptr fused_memory(size_t dep_id) const {
+        return dep_memory_ptr(get_fused_mem_offset() + dep_id);
     }
 
     bool has_fused_primitives() const { return !_node.get_fused_primitives().empty(); }
@@ -136,7 +145,7 @@ class primitive_inst {
     network_impl& _network;
     program_node const& _node;
 
-    std::shared_ptr<primitive_impl> _impl;
+    std::unique_ptr<primitive_impl> _impl;
 
     // this is a set of dependencies in terms of memory, if execution of this primitive requires data from another one,
     // it should be added to this set
@@ -154,14 +163,14 @@ class primitive_inst {
     // _output is optional because its initialization might be postponed (reshape_inst may either allocate it's own
     // buffer or attach input as output
     // depending on reshape_node.is_in_place())
-    memory_impl::ptr _output;
+    memory::ptr _output;
 
     bool _output_changed;  // todo: implement output reuse if neither of inputs has changed
     bool _has_valid_input =
         true;  // by default all primitives has valid inputs, exception is input_layout (see input_layout_inst)
     bool _has_mutable_input = false;
 
-    memory_impl::ptr allocate_output();
+    memory::ptr allocate_output();
     static std::vector<std::shared_ptr<primitive_inst>> build_exec_deps(
         std::vector<std::shared_ptr<primitive_inst>> const& mem_deps);
 
@@ -185,7 +194,7 @@ struct typed_primitive_impl : public primitive_impl {
     using primitive_impl::primitive_impl;
 
 private:
-    event_impl::ptr execute(const std::vector<refcounted_obj_ptr<event_impl>>& event,
+    event::ptr execute(const std::vector<event::ptr>& event,
                             primitive_inst& instance) override {
         if (instance.type() != PType::type_id())
             throw std::invalid_argument("Implementation type does not match primitive type");
@@ -206,19 +215,9 @@ struct typed_primitive_impl : public primitive_impl {
         return set_arguments_impl(reinterpret_cast<typed_primitive_inst<PType>&>(instance));
     }
 
-    void cleanup(primitive_inst& instance) override {
-        if (instance.type() != PType::type_id())
-            throw std::invalid_argument("Implementation type does not match primitive type");
-        if (instance.get_impl() != this)
-            throw std::invalid_argument(
-                "Trying to cleanup primitive implementation with mismatching primitive instance");
-
-        return cleanup_impl(reinterpret_cast<typed_primitive_inst<PType>&>(instance));
-    }
 
     virtual void set_arguments_impl(typed_primitive_inst<PType>& /*instance*/) {}
-    virtual void cleanup_impl(typed_primitive_inst<PType>& /*instance*/) {}
-    virtual event_impl::ptr execute_impl(const std::vector<event_impl::ptr>& event,
+    virtual event::ptr execute_impl(const std::vector<event::ptr>& event,
                                          typed_primitive_inst<PType>& instance) = 0;
 
     bool validate(const primitive_inst& instance) const override {
@@ -233,13 +232,8 @@ struct typed_primitive_impl : public primitive_impl {
     virtual bool validate_impl(const typed_primitive_inst<PType>&) const { return true; }
 };
 
-namespace details {
 template <class PType>
-class api_typed_primitive_inst_base : public primitive_inst {
-    static_assert(meta::is_api_primitive<PType>::value,
-                  "PType should name a non-const, non-volatile type derived from cldnn::primitive but not from "
-                  "cldnn::internal_primitive");
-
+class typed_primitive_inst_base : public primitive_inst {
 public:
     using typed_node = typed_program_node<PType>;
     using typed_impl = typed_primitive_impl<PType>;
@@ -247,16 +241,16 @@ class api_typed_primitive_inst_base : public primitive_inst {
     const typed_node& node;
     const PType& argument;
 
-    api_typed_primitive_inst_base(network_impl& network, typed_node const& node)
-        : api_typed_primitive_inst_base(network, node, do_allocate_memory(node)) {}
+    typed_primitive_inst_base(network_impl& network, typed_node const& node)
+        : typed_primitive_inst_base(network, node, do_allocate_memory(node)) {}
 
 protected:
-    api_typed_primitive_inst_base(network_impl& network, typed_node const& node, bool allocate_memory)
+    typed_primitive_inst_base(network_impl& network, typed_node const& node, bool allocate_memory)
         : primitive_inst(network, node, allocate_memory), node(_node), argument(*node.get_primitive()) {}
 
-    api_typed_primitive_inst_base(network_impl& network, typed_node const& node, memory_impl& buffer)
-        : api_typed_primitive_inst_base(network, node, false) {
-        _output = (memory_impl::ptr) &buffer;
+    typed_primitive_inst_base(network_impl& network, typed_node const& node, memory::ptr buffer)
+        : typed_primitive_inst_base(network, node, false) {
+        _output = buffer;
     }
 
 private:
@@ -269,48 +263,6 @@ class api_typed_primitive_inst_base : public primitive_inst {
     }
 };
 
-template <class PType>
-class internal_typed_primitive_inst_base : public primitive_inst {
-    static_assert(meta::is_internal_primitive<PType>::value,
-                  "PType should name a non-const, non-volatile type derived from cldnn::internal_primitive");
-
-public:
-    using typed_node = typed_program_node<PType>;
-    using typed_impl = typed_primitive_impl<PType>;
-
-    const typed_node& node;
-
-    internal_typed_primitive_inst_base(network_impl& network, typed_node const& node)
-        : internal_typed_primitive_inst_base(
-              network,
-              node,
-              false)  // by default, do not allocate output buffer automatically for internal primitives
-    {}
-
-    template <class... Guard>
-    [[noreturn]] void desc(Guard&&...) const {
-        static_assert(meta::always_false<meta::pack<Guard...>>::value, "Trying to get primitive from internal node");
-    }
-
-protected:
-    internal_typed_primitive_inst_base(network_impl& network, typed_node const& node, bool allocate_memory)
-        : primitive_inst(network, node, allocate_memory), node(_node) {}
-
-    internal_typed_primitive_inst_base(network_impl& network, typed_node const& node, memory_impl::ptr buffer)
-        : internal_typed_primitive_inst_base(network, node, false) {
-        _output = buffer;
-    }
-};
-}  // namespace details
-
-/*
-    Base class for all concrete primitive instances.
-*/
-template <class PType>
-using typed_primitive_inst_base = typename std::conditional<meta::is_api_primitive<PType>::value,
-                                                            details::api_typed_primitive_inst_base<PType>,
-                                                            details::internal_typed_primitive_inst_base<PType>>::type;
-
 /*
     Template class which represents instance of primitive 'PType'.
     Each new primitive should explicitly specialize this class.
@@ -330,14 +282,4 @@ class typed_primitive_inst : public typed_primitive_inst_base<PType> {
     static_assert(meta::always_false<PType>::value, "Missing typed_primitive_inst specialization");
 };
 
-#define CLDNN_DEFINE_SIMPLE_PRIM_INST(PType)                                       \
-    template <>                                                                    \
-    struct typed_primitive_inst<PType> : public typed_primitive_inst_base<PType> { \
-        using typed_primitive_inst_base<PType>::typed_primitive_inst_base;         \
-        static std::string to_string(PType##_node const& arg) {                    \
-            return primitive_inst::generic_to_string(arg, #PType);                 \
-        }                                                                          \
-    };                                                                             \
-    using PType##_inst = typed_primitive_inst<PType>;
-
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/primitive_type.h b/inference-engine/thirdparty/clDNN/src/include/primitive_type.h
index b608dff853615e..8ccd776183c436 100644
--- a/inference-engine/thirdparty/clDNN/src/include/primitive_type.h
+++ b/inference-engine/thirdparty/clDNN/src/include/primitive_type.h
@@ -4,10 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/memory.hpp"
-#include "api/primitive.hpp"
-#include "api/program.hpp"
-
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/primitives/primitive.hpp"
+#include "cldnn/graph/program.hpp"
 #include "topology_impl.h"
 
 #include <memory>
@@ -15,7 +14,7 @@
 
 namespace cldnn {
 struct network_impl;
-struct engine_impl;
+class engine;
 struct program_node;
 struct primitive_impl;
 class primitive_inst;
@@ -28,14 +27,12 @@ struct primitive_type {
                                                       const std::shared_ptr<primitive> prim) const = 0;
     virtual std::shared_ptr<primitive_inst> create_instance(network_impl& network,
                                                             const program_node& node) const = 0;
-    virtual std::unique_ptr<primitive_impl> choose_impl(engine_impl& engine,
+    virtual std::unique_ptr<primitive_impl> choose_impl(const engine& engine,
                                                         const program_node& node) const = 0;
-    virtual bool does_an_implementation_exist(engine_impl& engine, const program_node& node) const = 0;
-    virtual bool does_possible_implementation_exist(engine_impl& engine,
+    virtual bool does_an_implementation_exist(const engine& engine, const program_node& node) const = 0;
+    virtual bool does_possible_implementation_exist(const engine& engine,
                                                     const program_node& node) const = 0;
     virtual layout calc_output_layout(const program_node& node) const = 0;
     virtual std::string to_string(const program_node& node) const = 0;
-
-    virtual bool is_internal_type() const { return false; }
 };
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/primitive_type_base.h b/inference-engine/thirdparty/clDNN/src/include/primitive_type_base.h
index faffe78c4ea420..3dc14aa77d286d 100644
--- a/inference-engine/thirdparty/clDNN/src/include/primitive_type_base.h
+++ b/inference-engine/thirdparty/clDNN/src/include/primitive_type_base.h
@@ -4,21 +4,22 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
+
+#include "cldnn/runtime/engine.hpp"
+
 #include "meta_utils.h"
 #include "primitive_type.h"
 #include "program_node.h"
 #include "primitive_inst.h"
 #include "network_impl.h"
-#include "engine_impl.h"
+#include "implementation_map.h"
+
 #include <memory>
 #include <string>
 
 namespace cldnn {
 template <class PType>
 struct primitive_type_base : primitive_type {
-    static_assert(meta::is_api_primitive<PType>::value,
-                  "Primitive type passed to primitive_type_base should derive from cldnn::primitive");
-
     std::shared_ptr<cldnn::program_node> create_node(program_impl& program,
                                                      const std::shared_ptr<primitive> prim) const override {
         if (prim->type != this)
@@ -35,23 +36,26 @@ struct primitive_type_base : primitive_type {
         return std::make_shared<typed_primitive_inst<PType>>(network, node);
     }
 
-    std::unique_ptr<primitive_impl> choose_impl(engine_impl& engine, const cldnn::program_node& node) const override {
+    // TODO: Should we get rid of engine type in impl map? Or we must pass internal build engine to get real ocl type?
+    std::unique_ptr<primitive_impl> choose_impl(const engine& /* engine */, const cldnn::program_node& node) const override {
         if (node.type() != this)
             throw std::invalid_argument("primitive_type_base::choose_impl: primitive type mismatch");
 
-        return engine.create_primitive_impl(node.as<PType>());
+        auto factory = implementation_map<PType>::get(engine_types::ocl, node);
+        return std::move(std::unique_ptr<primitive_impl>(factory(node)));
     }
 
-    bool does_an_implementation_exist(engine_impl& engine, const cldnn::program_node& node) const override {
+    bool does_an_implementation_exist(const engine& /* engine */, const cldnn::program_node& node) const override {
         if (node.type() != this)
             throw std::invalid_argument("primitive_type_base::choose_impl: primitive type mismatch");
-        return engine.does_an_implementation_exist(node.as<PType>());
+
+        return implementation_map<PType>::check(engine_types::ocl, node);
     }
 
-    bool does_possible_implementation_exist(engine_impl& engine, const cldnn::program_node& node) const override {
+    bool does_possible_implementation_exist(const engine& /* engine */, const cldnn::program_node& node) const override {
         if (node.type() != this)
             throw std::invalid_argument("primitive_type_base::choose_impl: primitive type mismatch");
-        return engine.does_possible_implementation_exist(node.as<PType>());
+        return implementation_map<PType>::check_io_eq(engine_types::ocl, node);
     }
 
     cldnn::layout calc_output_layout(const cldnn::program_node& node) const override {
diff --git a/inference-engine/thirdparty/clDNN/src/include/prior_box_inst.h b/inference-engine/thirdparty/clDNN/src/include/prior_box_inst.h
index 8aba188276dda2..16c6681509fa1a 100644
--- a/inference-engine/thirdparty/clDNN/src/include/prior_box_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/prior_box_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/prior_box.hpp"
+#include "cldnn/primitives/prior_box.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
@@ -21,10 +22,10 @@ struct typed_program_node<prior_box> : typed_program_node_base<prior_box> {
 
     bool is_clustered() const { return get_primitive()->is_clustered(); }
     void calc_result();
-    memory_impl::ptr get_result_buffer() const { return result; }
+    memory::ptr get_result_buffer() const { return result; }
 
 private:
-    memory_impl::ptr result;
+    memory::ptr result;
 };
 
 using prior_box_node = typed_program_node<prior_box>;
@@ -40,7 +41,7 @@ class typed_primitive_inst<prior_box> : public typed_primitive_inst_base<prior_b
 public:
     typed_primitive_inst(network_impl& network, prior_box_node const& node);
 
-    memory_impl& input_memory() const { return dep_memory(0); }
+    memory& input_memory() const { return dep_memory(0); }
 };
 
 using prior_box_inst = typed_primitive_inst<prior_box>;
diff --git a/inference-engine/thirdparty/clDNN/src/include/program_dump_graph.h b/inference-engine/thirdparty/clDNN/src/include/program_dump_graph.h
index 5f8e687cd9089f..381b1ea9cdf1bd 100644
--- a/inference-engine/thirdparty/clDNN/src/include/program_dump_graph.h
+++ b/inference-engine/thirdparty/clDNN/src/include/program_dump_graph.h
@@ -6,7 +6,6 @@
 
 #include "program_impl.h"
 #include "program_node.h"
-#include "gpu/ocl_toolkit.h"
 #include <fstream>
 #include <string>
 
@@ -18,4 +17,4 @@ void dump_graph_optimized(std::ofstream&, const program_impl&);
 void dump_graph_processing_order(std::ofstream&, const program_impl&);
 void dump_graph_init(std::ofstream&, const program_impl&, std::function<bool(program_node const&)> const&);
 void dump_graph_info(std::ofstream&, const program_impl&, std::function<bool(program_node const&)> const&);
-}  // namespace cldnn
\ No newline at end of file
+}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/program_helpers.h b/inference-engine/thirdparty/clDNN/src/include/program_helpers.h
index cdef2a3b18d663..60291deb1e7557 100644
--- a/inference-engine/thirdparty/clDNN/src/include/program_helpers.h
+++ b/inference-engine/thirdparty/clDNN/src/include/program_helpers.h
@@ -7,8 +7,10 @@
 #pragma once
 
 #include "program_node.h"
-#include "engine_impl.h"
+#include "cldnn/runtime/engine.hpp"
 #include "program_impl.h"
+#include "data_inst.h"
+
 #include <string>
 #include <vector>
 #include <utility>
@@ -95,7 +97,7 @@ struct program_helpers {
         else
             do_for_types<RestOfT...>(node, rest...);
     }
-    static void merge_buffers(engine_impl& engine,
+    static void merge_buffers(engine& engine,
                               program_node& node,
                               const layout& target_layout,
                               size_t begin_offset,
diff --git a/inference-engine/thirdparty/clDNN/src/include/program_impl.h b/inference-engine/thirdparty/clDNN/src/include/program_impl.h
index 4588ec4ec801fc..fe8c8cb55f2795 100644
--- a/inference-engine/thirdparty/clDNN/src/include/program_impl.h
+++ b/inference-engine/thirdparty/clDNN/src/include/program_impl.h
@@ -6,10 +6,10 @@
 
 #pragma once
 
-#include "api/program.hpp"
-
-#include "refcounted_obj.h"
-#include "engine_impl.h"
+#include "cldnn/graph/program.hpp"
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/stream.hpp"
+#include "runtime/kernels_cache.hpp"
 
 #include <list>
 #include <string>
@@ -19,6 +19,10 @@
 #include <utility>
 #include <set>
 
+namespace kernel_selector {
+class TuningCache;
+}  // namespace kernel_selector
+
 namespace cldnn {
 
 struct topology_impl;
@@ -28,10 +32,19 @@ class layout_optimizer;
 class pass_manager;
 class base_pass;
 class program_impl_wrapper;
+
+struct gpu_program_state {
+    kernels_cache _kernels_cache;
+
+    gpu_program_state(engine& engine) : _kernels_cache(engine) {}
+};
+
 /*
     cldnn_program implementation
 */
-struct program_impl : public refcounted_obj<program_impl> {
+struct program_impl {
+    using ptr = std::shared_ptr<program_impl>;
+    using cptr = std::shared_ptr<const program_impl>;
     friend class calculate_prior_boxes;      // to be removed when possible
     friend class graph_initializations;      // to be removed when possible
     friend class prepare_padding;            // to be removed when possible
@@ -121,18 +134,18 @@ struct program_impl : public refcounted_obj<program_impl> {
     typedef std::vector<std::pair<std::string, primitives_info>> graph_optimizer_info;
     typedef std::pair<primitive_id, std::vector<primitive_id>> optimized_info;
 
-    program_impl(engine_impl& engine_ref,
+    program_impl(engine& engine_ref,
                  topology_impl const& topology,
                  build_options const& options,
                  bool is_internal,
                  bool no_optimizations = false);
     /* constructor used to build a program from subset of nodes of other program (used in propagate_constants) */
-    program_impl(engine_impl& engine_ref,
+    program_impl(engine& engine_ref,
                  std::set<std::shared_ptr<program_node>> const& nodes,
                  build_options const& options,
                  bool is_internal);
     ~program_impl();
-    engine_impl& get_engine() const { return *engine; }
+    engine& get_engine() const { return _engine; }
     const build_options& get_options() const { return options; }
     std::list<program_node*>& get_inputs() {
         return inputs;
@@ -144,6 +157,7 @@ struct program_impl : public refcounted_obj<program_impl> {
     const nodes_ordering& get_processing_order() const;
     nodes_ordering& get_processing_order();
     uint32_t get_prog_id() { return prog_id; }
+    stream& get_stream() { return *_stream; }
     const std::list<primitive_id>& get_optimized_out() const { return optimized_out; }
     const std::list<optimized_info>& get_optimized() const { return optimized; }
     bool has_node(const primitive_id& prim) const { return nodes_map.count(prim) > 0; }
@@ -151,7 +165,6 @@ struct program_impl : public refcounted_obj<program_impl> {
     program_node const& get_node(primitive_id const& id) const;
     std::shared_ptr<program_node> get_node_ptr(const primitive_id& prim) { return nodes_map.at(prim); }
     std::shared_ptr<program_node> get_node_ptr(const primitive_id& prim) const { return nodes_map.at(prim); }
-    void dump_memory_pool() const;
 
     // returns already existing program_node for given primitive 'prim' (lookup in 'nodes_map')
     // if it was previously created, otherwise creates and then returns program_node
@@ -211,14 +224,36 @@ struct program_impl : public refcounted_obj<program_impl> {
     void reset_program();
     uint32_t get_id() const { return prog_id; }
 
+    static ptr build_program(engine& engine,
+                             const topology_impl& topology,
+                             const build_options& options,
+                             bool is_internal = false,
+                             bool no_optimizations = false);
+    static ptr build_program(engine& engine,
+                             const std::set<std::shared_ptr<program_node>>& nodes,
+                             const build_options& options,
+                             bool is_internal);
+    static void init_primitives();
+    void compile();
+    void init_kernels();
+    kernel_id add_kernel(const std::shared_ptr<kernel_string> kernel_sring);
+    kernel::ptr get_kernel(kernel_id id);
+
+    void load_tuning_cache();
+    std::shared_ptr<kernel_selector::TuningCache> get_tuning_cache() const { return tuning_cache; }
+
 private:
     uint32_t prog_id = 0;
-    engine_impl::ptr engine;
+    engine& _engine;
+    stream::ptr _stream;
+    gpu_program_state program_state;
     build_options options;
     std::list<program_node*> inputs;
     std::vector<program_node*> outputs;
     nodes_ordering processing_order;
     std::unique_ptr<pass_manager> pm;
+    std::shared_ptr<kernel_selector::TuningCache> tuning_cache;
+
 
     std::map<primitive_id, std::shared_ptr<program_node>> nodes_map;
     std::list<primitive_id> optimized_out;
diff --git a/inference-engine/thirdparty/clDNN/src/include/program_node.h b/inference-engine/thirdparty/clDNN/src/include/program_node.h
index 3e8711d8625034..d0e904d3c56379 100644
--- a/inference-engine/thirdparty/clDNN/src/include/program_node.h
+++ b/inference-engine/thirdparty/clDNN/src/include/program_node.h
@@ -4,15 +4,14 @@
 
 #pragma once
 
-#include <set>
-#include <array>
+#include "cldnn/primitives/primitive.hpp"
+#include "cldnn/primitives/activation.hpp"
 
-#include "api/primitive.hpp"
-#include "api/activation.hpp"
-#include "internal_primitive.h"
 #include "kernel_selector_helper.h"
-
 #include "meta_utils.h"
+
+#include <set>
+#include <array>
 #include <vector>
 #include <memory>
 #include <list>
@@ -21,6 +20,7 @@
 namespace cldnn {
 
 struct program_impl;
+struct primitive_impl;
 class reorder_inputs;
 class graph_initializations;
 class prepare_quantization;
@@ -29,9 +29,6 @@ class pre_replace_deconv;
 template <class T>
 struct typed_program_node;
 
-template <class PType>
-struct internal_primitive_type_base;
-
 class json_composite;
 class xml_composite;
 
@@ -92,10 +89,10 @@ struct program_node {
     }
 
     program_impl& get_program() { return myprog; }
-    program_impl const& get_program() const { return myprog; }
+    program_impl& get_program() const { return myprog; }
 
-    std::shared_ptr<primitive_impl> get_selected_impl() const { return selected_impl; }
-    void set_selected_impl(std::shared_ptr<primitive_impl> impl) { selected_impl = impl; }
+    primitive_impl* get_selected_impl() const { return selected_impl.get(); }
+    void set_selected_impl(std::unique_ptr<primitive_impl> impl);
 
     std::vector<program_node*> const& get_dependencies() const { return dependencies; }
     program_node& get_dependency(size_t idx) const { return *dependencies.at(idx); }
@@ -312,7 +309,7 @@ struct program_node {
     std::shared_ptr<primitive> desc;
     program_impl& myprog;
 
-    std::shared_ptr<primitive_impl> selected_impl;
+    std::unique_ptr<primitive_impl> selected_impl;
 
     bool valid_output_layout = false;
     layout output_layout = layout(data_types::f32, format::bfyx, tensor());
@@ -353,12 +350,16 @@ struct program_node {
     void invalidate_users() const;
 };
 
-namespace details {
+/*
+Template class used to indicate that usage context requires 'program_node' to wrap primitive
+of type 'PType'. Successful conversion from 'program_node' to 'typed_program_node<PType>' means
+that this restriction in fact holds and functions/method/etc. may saftly use uderlaying primitive.
+
+This class shadows 'get_primitive' method from base class which now returns pointer to more specific
+type.
+*/
 template <class PType>
-struct api_typed_program_node_base : public program_node {
-    static_assert(meta::is_api_primitive<PType>::value,
-                  "PType should name a non-const, non-volatile type derived from cldnn::primitive but not from "
-                  "cldnn::internal_primitive");
+struct typed_program_node_base : public program_node {
     friend class cldnn::graph_initializations;
     friend class cldnn::pre_replace_deconv;
     friend class cldnn::prepare_quantization;
@@ -376,57 +377,6 @@ struct api_typed_program_node_base : public program_node {
     std::shared_ptr<PType> typed_desc() const { return std::static_pointer_cast<PType>(desc); }
 };
 
-struct internal_program_node_base : public program_node {
-    friend struct cldnn::program_impl;
-
-    explicit internal_program_node_base(program_impl& prog);
-
-    const primitive_id& id() const override { return internal_id; }
-
-    void set_implementation(std::unique_ptr<primitive_impl>&& impl);
-
-private:
-    primitive_id internal_id;
-
-    static primitive_id get_next_internal_id();
-};
-
-template <class PType>
-struct internal_typed_program_node_base : public internal_program_node_base {
-    static_assert(meta::is_internal_primitive<PType>::value,
-                  "PType should name a non-const, non-volatile type derived from cldnn::internal_primitive");
-
-public:
-    using internal_program_node_base::internal_program_node_base;
-
-    primitive_type_id type() const override { return PType::type_id(); }
-
-    template <class... Guard>
-    [[noreturn]] void get_primitive(Guard&&...) {
-        static_assert(meta::always_false<meta::pack<Guard...>>::value, "Trying to get primitive from internal node");
-    }
-
-protected:
-    template <class... Guard>
-    [[noreturn]] void typed_desc(Guard&&...) {
-        static_assert(meta::always_false<meta::pack<Guard...>>::value, "Trying to get primitive from internal node");
-    }
-};
-}  // namespace details
-
-/*
-Template class used to indicate that usage context requires 'program_node' to wrap primitive
-of type 'PType'. Successful conversion from 'program_node' to 'typed_program_node<PType>' means
-that this restriction in fact holds and functions/method/etc. may saftly use uderlaying primitive.
-
-This class shadows 'get_primitive' method from base class which now returns pointer to more specific
-type.
-*/
-template <class PType>
-using typed_program_node_base = typename std::conditional<meta::is_api_primitive<PType>::value,
-                                                          details::api_typed_program_node_base<PType>,
-                                                          details::internal_typed_program_node_base<PType>>::type;
-
 /*
     Actual template class used in context which requires 'program_node' to wrap
     primitive of type 'PType'. This class is introduced to provide possibility of explicit specialization.
diff --git a/inference-engine/thirdparty/clDNN/src/include/proposal_inst.h b/inference-engine/thirdparty/clDNN/src/include/proposal_inst.h
index d1fa03f8d9e58e..3c67afe4003cfa 100644
--- a/inference-engine/thirdparty/clDNN/src/include/proposal_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/proposal_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/proposal.hpp"
+#include "cldnn/primitives/proposal.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <vector>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/pyramid_roi_align_inst.h b/inference-engine/thirdparty/clDNN/src/include/pyramid_roi_align_inst.h
index 45cd6638f741ab..829f5228f61897 100644
--- a/inference-engine/thirdparty/clDNN/src/include/pyramid_roi_align_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/pyramid_roi_align_inst.h
@@ -3,7 +3,7 @@
 //
 
 #pragma once
-#include "api/pyramid_roi_align.hpp"
+#include "cldnn/primitives/pyramid_roi_align.hpp"
 #include "primitive_inst.h"
 
 #include <memory>
@@ -36,11 +36,11 @@ class typed_primitive_inst<pyramid_roi_align> : public typed_primitive_inst_base
     static std::string to_string(pyramid_roi_align_node const& node);
     typed_primitive_inst(network_impl& network, pyramid_roi_align_node const& node);
 
-    memory_impl& input() const { return dep_memory(0); }
-    memory_impl& P2() const { return dep_memory(1); }
-    memory_impl& P3() const { return dep_memory(2); }
-    memory_impl& P4() const { return dep_memory(3); }
-    memory_impl& P5() const { return dep_memory(4); }
+    memory& input() const { return dep_memory(0); }
+    memory& P2() const { return dep_memory(1); }
+    memory& P3() const { return dep_memory(2); }
+    memory& P4() const { return dep_memory(3); }
+    memory& P5() const { return dep_memory(4); }
 };
 
 using pyramid_roi_align_inst = typed_primitive_inst<pyramid_roi_align>;
diff --git a/inference-engine/thirdparty/clDNN/src/include/quantize_inst.h b/inference-engine/thirdparty/clDNN/src/include/quantize_inst.h
index 0cf0b30e81c0cc..1d680ecb3cce20 100644
--- a/inference-engine/thirdparty/clDNN/src/include/quantize_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/quantize_inst.h
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/quantize.hpp"
+#include "cldnn/primitives/quantize.hpp"
 #include "primitive_inst.h"
 #include "data_inst.h"
 #include "kernel_selector/core/actual_kernels/quantize/quantize_kernel_params.h"
diff --git a/inference-engine/thirdparty/clDNN/src/include/reduce_inst.h b/inference-engine/thirdparty/clDNN/src/include/reduce_inst.h
index 6574ce8de21a97..678f861a2a8694 100644
--- a/inference-engine/thirdparty/clDNN/src/include/reduce_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/reduce_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/reduce.hpp"
+#include "cldnn/primitives/reduce.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/refcounted_obj.h b/inference-engine/thirdparty/clDNN/src/include/refcounted_obj.h
deleted file mode 100644
index 82b931f65da3c2..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/include/refcounted_obj.h
+++ /dev/null
@@ -1,135 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma once
-#include <atomic>
-#include <type_traits>
-
-namespace cldnn {
-
-template <class T>
-struct refcounted_obj_ptr;
-
-/**
- * \brief Base class for all reference counted pointers aka PIMPL implementations
- */
-// TODO refine this code for multithreading support
-template <class T>
-class refcounted_obj {
-public:
-    using ptr = refcounted_obj_ptr<typename std::remove_const<T>::type>;
-    using cptr = refcounted_obj_ptr<typename std::add_const<T>::type>;
-
-    refcounted_obj() : _ref_count(1) {}
-
-    virtual ~refcounted_obj() = default;
-
-    void add_ref() const { ++_ref_count; }
-
-    void release() const {
-        if ((--_ref_count) == 0)
-            delete static_cast<const T*>(this);
-    }
-
-    int get_ref_count() const { return _ref_count; }
-
-private:
-    mutable std::atomic_int _ref_count;
-};
-
-template <class T>
-struct refcounted_obj_ptr {
-    template <class U = T>
-    explicit refcounted_obj_ptr(T* ptr, bool add_ref = true) : _ptr(ptr) {
-        static_assert(std::is_base_of<refcounted_obj<typename std::remove_const<U>::type>, U>::value,
-                      "Object handled with refcounted_obj_ptr should derive from refcounted_obj");
-        if (add_ref)
-            ptr_add_ref();
-    }
-
-    // for refcounted_obj_ptr<const T>, allow contruction from T*
-    template <class U = T, class = typename std::enable_if<std::is_const<U>::value>::type>
-    explicit refcounted_obj_ptr(typename std::remove_const<T>::type* ptr, bool add_ref = true) : _ptr(ptr) {
-        static_assert(std::is_base_of<refcounted_obj<typename std::remove_const<U>::type>, U>::value,
-                      "Object handled with refcounted_obj_ptr should derive from refcounted_obj");
-        if (add_ref)
-            ptr_add_ref();
-    }
-
-    constexpr refcounted_obj_ptr() : _ptr(nullptr) {}
-
-    refcounted_obj_ptr(const refcounted_obj_ptr& other) : _ptr(other._ptr) { ptr_add_ref(); }
-
-    refcounted_obj_ptr& operator=(const refcounted_obj_ptr& other) {
-        if (this == &other)
-            return *this;
-        ptr_release();
-        _ptr = other._ptr;
-        ptr_add_ref();
-        return *this;
-    }
-
-    refcounted_obj_ptr(refcounted_obj_ptr&& other) noexcept {
-        _ptr = other._ptr;
-        other._ptr = nullptr;
-    }
-
-    refcounted_obj_ptr& operator=(refcounted_obj_ptr&& other) {
-        if (this == &other)
-            return *this;
-        ptr_release();
-        _ptr = other._ptr;
-        other._ptr = nullptr;
-        return *this;
-    }
-
-    ~refcounted_obj_ptr() {
-        ptr_release();
-        _ptr = nullptr;
-    }
-
-    T* detach() {
-        T* result = _ptr;
-        _ptr = nullptr;
-        return result;
-    }
-
-    void reset(T* ptr, bool add_ref = true) {
-        ptr_release();
-        _ptr = ptr;
-        if (add_ref)
-            ptr_add_ref();
-    }
-
-    operator bool() const { return _ptr != nullptr; }
-    T* get() const { return _ptr; }
-    T& operator*() const { return *get(); }
-    T* operator->() const { return get(); }
-
-    friend bool operator==(const refcounted_obj_ptr& lhs, const refcounted_obj_ptr& rhs) {
-        return lhs._ptr == rhs._ptr;
-    }
-
-    friend bool operator!=(const refcounted_obj_ptr& lhs, const refcounted_obj_ptr& rhs) { return !(lhs == rhs); }
-
-    // for refcounted_obj_ptr<T>, allow conversion to refcounted_obj_ptr<const T>
-    template <class R>
-    operator refcounted_obj_ptr<const R>() const {
-        return refcounted_obj_ptr<const R>(_ptr);
-    }
-
-private:
-    T* _ptr;
-    void ptr_add_ref() {
-        if (_ptr)
-            _ptr->add_ref();
-    }
-    void ptr_release() {
-        if (_ptr)
-            _ptr->release();
-    }
-};
-
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/include/region_yolo_inst.h b/inference-engine/thirdparty/clDNN/src/include/region_yolo_inst.h
index 26553e215cecc9..0b6a24068cccc8 100644
--- a/inference-engine/thirdparty/clDNN/src/include/region_yolo_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/region_yolo_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/region_yolo.hpp"
+#include "cldnn/primitives/region_yolo.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/reorder_inst.h b/inference-engine/thirdparty/clDNN/src/include/reorder_inst.h
index c9fccb0d7c7106..ce851e482aebb5 100644
--- a/inference-engine/thirdparty/clDNN/src/include/reorder_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/reorder_inst.h
@@ -4,8 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/reorder.hpp"
+
+#include "cldnn/primitives/reorder.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 #include <memory>
 
@@ -50,8 +52,8 @@ class typed_primitive_inst<reorder> : public typed_primitive_inst_base<reorder>
 
 public:
     typed_primitive_inst(network_impl& network, reorder_node const& node);
-    memory_impl& mean_nv12_memory() const { return dep_memory(2); }
-    memory_impl& mean_memory() const { return dep_memory(1); }
+    memory::ptr mean_nv12_memory() const { return dep_memory_ptr(2); }
+    memory::ptr mean_memory() const { return dep_memory_ptr(1); }
 
     bool has_mean() const { return !argument.mean.empty(); }
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/reorg_yolo_inst.h b/inference-engine/thirdparty/clDNN/src/include/reorg_yolo_inst.h
index 43361d31760ef9..43efd38a9b5180 100644
--- a/inference-engine/thirdparty/clDNN/src/include/reorg_yolo_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/reorg_yolo_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/reorg_yolo.hpp"
+#include "cldnn/primitives/reorg_yolo.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/resample_inst.h b/inference-engine/thirdparty/clDNN/src/include/resample_inst.h
index 33bf98f2ddaf63..6bb8de03b27351 100644
--- a/inference-engine/thirdparty/clDNN/src/include/resample_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/resample_inst.h
@@ -4,10 +4,11 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/resample.hpp"
+#include "cldnn/primitives/resample.hpp"
 #include "primitive_inst.h"
-#include <memory>
 #include "topology_impl.h"
+
+#include <memory>
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/reshape_inst.h b/inference-engine/thirdparty/clDNN/src/include/reshape_inst.h
index 3c640441bad3f3..32adce5518613a 100644
--- a/inference-engine/thirdparty/clDNN/src/include/reshape_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/reshape_inst.h
@@ -4,9 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/reshape.hpp"
+#include "cldnn/primitives/reshape.hpp"
 #include "primitive_inst.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
+
 #include <string>
 #include <memory>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/reverse_sequence_inst.h b/inference-engine/thirdparty/clDNN/src/include/reverse_sequence_inst.h
index 51f4a24a8cbb6c..60ba5b3dd4e297 100644
--- a/inference-engine/thirdparty/clDNN/src/include/reverse_sequence_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/reverse_sequence_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/reverse_sequence.hpp"
+#include "cldnn/primitives/reverse_sequence.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/roi_pooling_inst.h b/inference-engine/thirdparty/clDNN/src/include/roi_pooling_inst.h
index f9c3dc3b069f46..6d3cfcf9f38355 100644
--- a/inference-engine/thirdparty/clDNN/src/include/roi_pooling_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/roi_pooling_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/roi_pooling.hpp"
+#include "cldnn/primitives/roi_pooling.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
@@ -34,8 +35,8 @@ class typed_primitive_inst<roi_pooling> : public typed_primitive_inst_base<roi_p
 public:
     using parent::parent;
 
-    memory_impl& rois_memory() const { return dep_memory(1); }
-    memory_impl& trans_memory() const { return dep_memory(2); }
+    memory::ptr rois_memory() const { return dep_memory_ptr(1); }
+    memory::ptr trans_memory() const { return dep_memory_ptr(2); }
 };
 
 using roi_pooling_inst = typed_primitive_inst<roi_pooling>;
diff --git a/inference-engine/thirdparty/clDNN/src/include/scale_inst.h b/inference-engine/thirdparty/clDNN/src/include/scale_inst.h
index 018daad2a8cfd3..a25fb49be994e7 100644
--- a/inference-engine/thirdparty/clDNN/src/include/scale_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/scale_inst.h
@@ -4,11 +4,12 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/scale.hpp"
+#include "cldnn/primitives/scale.hpp"
 #include "primitive_inst.h"
+#include "kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_base.h"
+
 #include <string>
 #include <memory>
-#include "kernel_selector/core/actual_kernels/eltwise/eltwise_kernel_base.h"
 
 namespace cldnn {
 
@@ -47,8 +48,8 @@ class typed_primitive_inst<scale> : public typed_primitive_inst_base<scale> {
 public:
     typed_primitive_inst(network_impl& network, scale_node const& desc);
 
-    memory_impl& scale_memory() const { return dep_memory(1); }
-    memory_impl& bias_memory() const { return dep_memory(2); }
+    memory::ptr scale_memory() const { return dep_memory_ptr(1); }
+    memory::ptr bias_memory() const { return dep_memory_ptr(2); }
 
     bool bias_term() const { return _node.as<scale>().bias_term(); }
 };
diff --git a/inference-engine/thirdparty/clDNN/src/include/scatter_elements_update_inst.h b/inference-engine/thirdparty/clDNN/src/include/scatter_elements_update_inst.h
index dd87a7facce92d..f70bed68bee509 100644
--- a/inference-engine/thirdparty/clDNN/src/include/scatter_elements_update_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/scatter_elements_update_inst.h
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/scatter_elements_update.hpp"
+#include "cldnn/primitives/scatter_elements_update.hpp"
 #include "primitive_inst.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/scatter_nd_update_inst.h b/inference-engine/thirdparty/clDNN/src/include/scatter_nd_update_inst.h
index 1da6e21445ccf9..acf3ee9c6c010d 100644
--- a/inference-engine/thirdparty/clDNN/src/include/scatter_nd_update_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/scatter_nd_update_inst.h
@@ -4,7 +4,7 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/scatter_nd_update.hpp"
+#include "cldnn/primitives/scatter_nd_update.hpp"
 #include "primitive_inst.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/scatter_update_inst.h b/inference-engine/thirdparty/clDNN/src/include/scatter_update_inst.h
index 54fa0fae09dc52..1c9b7934f25aa1 100644
--- a/inference-engine/thirdparty/clDNN/src/include/scatter_update_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/scatter_update_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/scatter_update.hpp"
+#include "cldnn/primitives/scatter_update.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/select_inst.h b/inference-engine/thirdparty/clDNN/src/include/select_inst.h
index bd59f29062db5f..4adde4e8c604e5 100644
--- a/inference-engine/thirdparty/clDNN/src/include/select_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/select_inst.h
@@ -4,9 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include <api/select.hpp>
-
+#include "cldnn/primitives/select.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/shuffle_channels_inst.h b/inference-engine/thirdparty/clDNN/src/include/shuffle_channels_inst.h
index 585d7440ba7a78..278f9c56fab065 100644
--- a/inference-engine/thirdparty/clDNN/src/include/shuffle_channels_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/shuffle_channels_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/shuffle_channels.hpp"
+#include "cldnn/primitives/shuffle_channels.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/sliding_window_utils.h b/inference-engine/thirdparty/clDNN/src/include/sliding_window_utils.h
index b832c49979fe06..369b78f4a8bd44 100644
--- a/inference-engine/thirdparty/clDNN/src/include/sliding_window_utils.h
+++ b/inference-engine/thirdparty/clDNN/src/include/sliding_window_utils.h
@@ -4,15 +4,15 @@
 
 #pragma once
 
-#include <api/layout.hpp>
-#include <api/tensor.hpp>
+#include "cldnn/runtime/layout.hpp"
+#include "cldnn/runtime/tensor.hpp"
+
+#include "meta_utils.h"
 
 #include <algorithm>
 #include <cassert>
 #include <stdexcept>
 
-#include "meta_utils.h"
-
 namespace cldnn {
 
 /// @brief Sliding window output range computation mode.
diff --git a/inference-engine/thirdparty/clDNN/src/include/softmax_inst.h b/inference-engine/thirdparty/clDNN/src/include/softmax_inst.h
index 1fc97c3288c8ce..8bf45e7392446f 100644
--- a/inference-engine/thirdparty/clDNN/src/include/softmax_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/softmax_inst.h
@@ -4,8 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/softmax.hpp"
+#include "cldnn/primitives/softmax.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/space_to_batch_inst.h b/inference-engine/thirdparty/clDNN/src/include/space_to_batch_inst.h
index 3fed298e57fbdb..db2f69719d9499 100644
--- a/inference-engine/thirdparty/clDNN/src/include/space_to_batch_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/space_to_batch_inst.h
@@ -4,8 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/space_to_batch.hpp"
+
+#include "cldnn/primitives/space_to_batch.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/space_to_depth_inst.h b/inference-engine/thirdparty/clDNN/src/include/space_to_depth_inst.h
index 527c17e9b85427..2cebeb7cdf8127 100644
--- a/inference-engine/thirdparty/clDNN/src/include/space_to_depth_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/space_to_depth_inst.h
@@ -4,8 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/space_to_depth.hpp"
+
+#include "cldnn/primitives/space_to_depth.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/split_inst.h b/inference-engine/thirdparty/clDNN/src/include/split_inst.h
index 16f74d64e4a0e7..af9a8e402eb08b 100644
--- a/inference-engine/thirdparty/clDNN/src/include/split_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/split_inst.h
@@ -4,8 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/split.hpp"
+
+#include "cldnn/primitives/split.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/strided_slice_inst.h b/inference-engine/thirdparty/clDNN/src/include/strided_slice_inst.h
index 812cd2137612d3..406d46b44effc7 100644
--- a/inference-engine/thirdparty/clDNN/src/include/strided_slice_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/strided_slice_inst.h
@@ -4,9 +4,11 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/strided_slice.hpp"
+
+#include "cldnn/primitives/strided_slice.hpp"
 #include "primitive_inst.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
+
 #include <string>
 #include <vector>
 
diff --git a/inference-engine/thirdparty/clDNN/src/include/tile_inst.h b/inference-engine/thirdparty/clDNN/src/include/tile_inst.h
index ea3966c5ee5a9d..4569201f79caad 100644
--- a/inference-engine/thirdparty/clDNN/src/include/tile_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/tile_inst.h
@@ -4,8 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/tile.hpp"
+
+#include "cldnn/primitives/tile.hpp"
 #include "primitive_inst.h"
+
 #include <string>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/to_string_utils.h b/inference-engine/thirdparty/clDNN/src/include/to_string_utils.h
index ae48ecd853a29f..299675318879e5 100644
--- a/inference-engine/thirdparty/clDNN/src/include/to_string_utils.h
+++ b/inference-engine/thirdparty/clDNN/src/include/to_string_utils.h
@@ -3,12 +3,15 @@
 //
 
 #pragma once
+
+#include "cldnn/runtime/tensor.hpp"
+#include "cldnn/runtime/layout.hpp"
+#include "cldnn/runtime/device.hpp"
+#include "cldnn/primitives/primitive.hpp"
+
+
 #include <string>
 #include <sstream>
-#include "api/tensor.hpp"
-#include "api/layout.hpp"
-#include "api/primitive.hpp"
-#include "device_impl.h"
 #include <memory>
 
 namespace cldnn {
diff --git a/inference-engine/thirdparty/clDNN/src/include/topology_impl.h b/inference-engine/thirdparty/clDNN/src/include/topology_impl.h
index 2e9b943ca146f8..bb1df4fc4b7ddb 100644
--- a/inference-engine/thirdparty/clDNN/src/include/topology_impl.h
+++ b/inference-engine/thirdparty/clDNN/src/include/topology_impl.h
@@ -4,9 +4,9 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #pragma once
-#include "api/primitive.hpp"
-#include "api/input_layout.hpp"
-#include "refcounted_obj.h"
+
+#include "cldnn/primitives/primitive.hpp"
+#include "cldnn/primitives/input_layout.hpp"
 
 #include <map>
 #include <memory>
@@ -16,8 +16,9 @@ namespace cldnn {
 
 typedef std::map<primitive_id, std::shared_ptr<primitive>> topology_map;
 
-struct topology_impl : public refcounted_obj<topology_impl> {
+struct topology_impl {
 public:
+    using ptr = std::shared_ptr<topology_impl>;
     explicit topology_impl(const topology_map& map = topology_map()) : _primitives(map) {}
 
     void add(std::shared_ptr<primitive> desc) {
diff --git a/inference-engine/thirdparty/clDNN/src/input_layout.cpp b/inference-engine/thirdparty/clDNN/src/input_layout.cpp
index b4df10dd23a66f..6aaefe16e65296 100644
--- a/inference-engine/thirdparty/clDNN/src/input_layout.cpp
+++ b/inference-engine/thirdparty/clDNN/src/input_layout.cpp
@@ -5,8 +5,8 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "input_layout_inst.h"
 #include "primitive_type_base.h"
-#include "memory_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 #include <memory>
@@ -27,16 +27,16 @@ input_layout_inst::typed_primitive_inst(network_impl& network, input_layout_node
     _has_valid_input = false;  // by default input for 'input_layout' is invalid as long as user doesn't call set_data
 }
 
-void input_layout_inst::set_data(memory_impl& mem) {
+void input_layout_inst::set_data(memory::ptr mem) {
     auto ol = node.get_output_layout();
 
-    check_memory_to_set(mem, ol);
+    check_memory_to_set(*mem, ol);
 
-    if (mem.is_allocated_by(get_network().get_engine())) {
-        _output = (memory_impl::ptr) &mem;
+    if (mem->is_allocated_by(get_network().get_engine())) {
+        _output = mem;
     } else {
-        mem_lock<char> src((memory_impl::ptr) &mem);
-        mem_lock<char> dst(_output);
+        mem_lock<char> src(mem, get_network().get_stream());
+        mem_lock<char> dst(_output, get_network().get_stream());
         std::copy(src.begin(), src.end(), dst.begin());
     }
 
diff --git a/inference-engine/thirdparty/clDNN/src/kernel_selector_helper.cpp b/inference-engine/thirdparty/clDNN/src/kernel_selector_helper.cpp
index 7ba2a515120ae4..716b5fa49eb494 100644
--- a/inference-engine/thirdparty/clDNN/src/kernel_selector_helper.cpp
+++ b/inference-engine/thirdparty/clDNN/src/kernel_selector_helper.cpp
@@ -6,7 +6,6 @@
 #include "kernel_selector_params.h"
 #include "to_string_utils.h"
 
-#include "gpu/ocl_toolkit.h"
 
 #include "program_node.h"
 #include "program_impl.h"
@@ -733,28 +732,27 @@ kernel_selector::activation_function get_kernel_selector_activation_param(activa
 
 void set_params(const program_node& node, kernel_selector::params& params) {
     const auto& program = node.get_program();
-    const auto& context = program.get_engine().get_context();
-    const auto& device_info = context->get_device_info();
-
-    params.engineInfo.bSubGroupSupport = context->extension_supported("cl_intel_subgroups");
-    params.engineInfo.bSubGroupShortSupport = context->extension_supported("cl_intel_subgroups_short");
-    params.engineInfo.bSubGroupCharSupport = context->extension_supported("cl_intel_subgroups_char");
-    params.engineInfo.bFP16Support = context->extension_supported("cl_khr_fp16");
-    params.engineInfo.bFP64Support = context->extension_supported("cl_khr_fp64");
+    const auto& device_info = program.get_engine().get_device_info();
+
+    params.engineInfo.bSubGroupSupport = device_info.supports_subgroups;
+    params.engineInfo.bSubGroupShortSupport = device_info.supports_subgroups_short;
+    params.engineInfo.bSubGroupCharSupport = device_info.supports_subgroups_char;
+    params.engineInfo.bFP16Support = device_info.supports_fp16;
+    params.engineInfo.bFP64Support = device_info.supports_fp64;
     params.engineInfo.bIMADSupport = device_info.supports_imad != 0;
     params.engineInfo.bIMMADSupport = device_info.supports_immad != 0;
     params.engineInfo.bImageSupport = device_info.supports_image != 0;
-    params.engineInfo.bOptHintsSupport = device_info.supports_optimization_hints;
+    params.engineInfo.bOptHintsSupport = false;
     params.engineInfo.bLocalBlockIOSupport = device_info.supports_local_block_io;
     params.engineInfo.deviceType = get_device_type(device_info.dev_type);
     params.engineInfo.maxWorkGroupSize = device_info.max_work_group_size;
     params.engineInfo.maxLocalMemSize = device_info.max_local_mem_size;
     params.engineInfo.maxImage2dWidth = device_info.max_image2d_width;
     params.engineInfo.maxImage2dHeight = device_info.max_image2d_height;
-    params.engineInfo.computeUnitsCount = device_info.compute_units_count;
+    params.engineInfo.computeUnitsCount = device_info.execution_units_count;
     params.engineInfo.maxThreadsPerExecutionUnit = device_info.max_threads_per_execution_unit;
     params.engineInfo.maxThreadsPerDevice = device_info.max_threads_per_device;
-    params.engineInfo.deviceCache = context->get_device_cache();
+    params.engineInfo.deviceCache = program.get_tuning_cache();
     params.engineInfo.driverVersion = device_info.driver_version;
 
     auto impl_forcing_bo = program.get_options().get<build_option_type::force_implementations>();
@@ -766,9 +764,7 @@ void set_params(const program_node& node, kernel_selector::params& params) {
 }
 
 void set_optional_params(const program_impl& program, kernel_selector::optional_params& params) {
-    const auto& context = program.get_engine().get_context();
-
-    params.meaningfulKernelsNames = context->get_configuration().meaningful_kernels_names;
+    params.meaningfulKernelsNames = false;
     params.allowStaticInputReordering = program.get_options().get<build_option_type::optimize_data>()->enabled() ||
                                         program.get_options().get<build_option_type::allow_static_input_reorder>()->enabled();
     params.allowInputReordering = false;
diff --git a/inference-engine/thirdparty/clDNN/src/layout_optimizer.cpp b/inference-engine/thirdparty/clDNN/src/layout_optimizer.cpp
index 6540c4f79625ee..9c26b09cf27dc3 100644
--- a/inference-engine/thirdparty/clDNN/src/layout_optimizer.cpp
+++ b/inference-engine/thirdparty/clDNN/src/layout_optimizer.cpp
@@ -6,7 +6,7 @@
 #include "topology_impl.h"
 #include "network_impl.h"
 #include "primitive_inst.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 
 #include "data_inst.h"
 #include "reorder_inst.h"
diff --git a/inference-engine/thirdparty/clDNN/src/loop.cpp b/inference-engine/thirdparty/clDNN/src/loop.cpp
index da79292fdaaf58..9399dfdfa6a144 100644
--- a/inference-engine/thirdparty/clDNN/src/loop.cpp
+++ b/inference-engine/thirdparty/clDNN/src/loop.cpp
@@ -5,11 +5,10 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "loop_inst.h"
 
-#include "error_handler.h"
 #include "json_object.h"
 #include "primitive_type_base.h"
-#include "api/data.hpp"
-#include "api/mutable_data.hpp"
+#include "cldnn/primitives/data.hpp"
+#include "cldnn/primitives/mutable_data.hpp"
 #include <string>
 #include <exception>
 #include <algorithm>
@@ -199,18 +198,18 @@ void loop_inst::preprocess_output_memory() {
         const primitive_id& external_id = output_mapping.external_id;
         const primitive_id& internal_id = output_mapping.internal_id;
         if (output_mapping.axis < 0) {
-            memory_impl::ptr memory = get_external_memory(external_id);
-            body_network->get_primitive(internal_id)->set_output_memory(*memory);
+            memory::ptr memory = get_external_memory(external_id);
+            body_network->get_primitive(internal_id)->set_output_memory(memory);
         } else {
-            memory_impl::ptr to_mem = get_external_memory(external_id);
+            memory::ptr to_mem = get_external_memory(external_id);
             auto output_prim = body_network->get_primitive(internal_id);
             layout sliced_layout = output_prim->output_memory().get_layout();
 
             const int64_t max_iteration = node.get_max_iteration();
-            std::vector<memory_impl::ptr> sliced_mems;
+            std::vector<memory::ptr> sliced_mems;
             sliced_mems.reserve(max_iteration);
             for (int j=0; j < max_iteration; ++j) {
-                memory_impl::ptr sliced_mem = engine.allocate_memory(sliced_layout, 0);
+                memory::ptr sliced_mem = engine.allocate_memory(sliced_layout, 0);
                 sliced_mems.push_back(sliced_mem);
             }
 
@@ -219,7 +218,7 @@ void loop_inst::preprocess_output_memory() {
             const int64_t num_elements_iteration = sliced_layout.count() / num_elements_batch;
             const int64_t start = output_mapping.start < 0? node.get_max_iteration() - 1: output_mapping.start;
             concatenated_memory_mapping memory_mapping_info(
-                output_mapping.axis, to_mem, sliced_mems,
+                output_mapping.axis, to_mem, sliced_mems, _network.get_stream(),
                 num_elements_iteration, output_mapping.stride, start);
             memory_mapping_info.concat_data_prim = body_network->get_primitive(internal_id);
             concatenated_output_mem_mappings.push_back(memory_mapping_info);
@@ -241,7 +240,7 @@ void loop_inst::preprocess_input_memory() {
             CLDNN_ERROR_MESSAGE(id(), "loop primitive_map is incomplete");
         }
 
-        memory_impl& memory = input_memory(memory_num);
+        auto memory = input_memory_ptr(memory_num);
         for (size_t i = 0; i < input_map_ptrs.size(); ++i) {
             const auto input_map = input_map_ptrs.at(i);
             bool is_concatenated_input = (input_map->axis >= 0);
@@ -249,10 +248,10 @@ void loop_inst::preprocess_input_memory() {
                 layout sliced_layout
                     = body_network->get_primitive(input_map->internal_id)->output_memory().get_layout();
                 const int64_t max_iteration = node.get_max_iteration();
-                std::vector<memory_impl::ptr> sliced_mems;
+                std::vector<memory::ptr> sliced_mems;
                 sliced_mems.reserve(max_iteration);
                 for (int j=0; j < max_iteration; ++j) {
-                    memory_impl::ptr sliced_mem = engine.allocate_memory(sliced_layout, 0);
+                    memory::ptr sliced_mem = engine.allocate_memory(sliced_layout, 0);
                     sliced_mems.push_back(sliced_mem);
                 }
                 const int64_t num_elements_batch = concatenated_memory_mapping::get_batch_size(
@@ -260,12 +259,12 @@ void loop_inst::preprocess_input_memory() {
                 const int64_t num_elements_iteration = sliced_layout.count() / num_elements_batch;
                 const int64_t start = input_map->start < 0? node.get_max_iteration() - 1: input_map->start;
                 concatenated_memory_mapping concatenated_input_mem_mapping_info(
-                    input_map->axis, (memory_impl::ptr)&memory, sliced_mems,
+                    input_map->axis, memory, sliced_mems, _network.get_stream(),
                     num_elements_iteration, input_map->stride, start);
                 concatenated_input_mem_mapping_info.sliced_data_prim = body_network->get_primitive(input_map->internal_id);
                 iteration_mem.push_back(concatenated_input_mem_mapping_info);
             } else {
-                if (memory.get_layout().data_type != body_network->get_primitive(input_map->internal_id)->output_memory().get_layout().data_type) {
+                if (memory->get_layout().data_type != body_network->get_primitive(input_map->internal_id)->output_memory().get_layout().data_type) {
                     CLDNN_ERROR_MESSAGE(id(), "incompatible datatypes");
                 }
                 body_network->set_input_data(input_map->internal_id, memory);
@@ -285,17 +284,17 @@ void loop_inst::preprocess_backedge_memory() {
         auto backedged_sliced_output_mems = get_sliced_mem(back_edge.from);
         const auto backedge_to_prim = body_network->get_primitive(back_edge.to);
         const auto backedge_from_prim = body_network->get_primitive(back_edge.from);
-        memory_impl::ptr initial_mem = get_external_memory(input_map->external_id);
+        memory::ptr initial_mem = get_external_memory(input_map->external_id);
         if (backedged_sliced_output_mems.empty()) {
             // backedge output which does not need concatenation
             // input memory = output memory = loop output memory
             const auto output_mapping = node.find_io_primitive_maps(back_edge.from, false);
-            memory_impl::ptr backedge_mem;
+            memory::ptr backedge_mem;
             if (output_mapping.empty()) {
                 // from and to primitives in backedge are connected directly
                 if (backedge_to_prim == backedge_from_prim->dependencies().front()) {
                     backedge_memory_mappings.emplace_back(
-                        backedge_from_prim, backedge_to_prim, initial_mem);
+                        backedge_from_prim, backedge_to_prim, initial_mem, body_network->get_stream());
                     continue;
                 } else {
                     auto output_prim = body_network->get_primitive(back_edge.from);
@@ -305,19 +304,19 @@ void loop_inst::preprocess_backedge_memory() {
             } else {
                 backedge_mem = get_external_memory(output_mapping.front()->external_id);
             }
-            body_network->set_input_data(back_edge.to, *backedge_mem);
-            body_network->set_output_memory(back_edge.from, *backedge_mem);
+            body_network->set_input_data(back_edge.to, backedge_mem);
+            body_network->set_output_memory(back_edge.from, backedge_mem);
             backedge_memory_mappings.emplace_back(
-                backedge_from_prim, backedge_to_prim, backedge_mem, initial_mem);
+                backedge_from_prim, backedge_to_prim, backedge_mem, initial_mem, body_network->get_stream());
         } else {
             // backedge output which needs concatenation
             backedge_memory_mappings.emplace_back(
-                backedge_from_prim, backedge_to_prim, backedged_sliced_output_mems, initial_mem);
+                backedge_from_prim, backedge_to_prim, backedged_sliced_output_mems, initial_mem, body_network->get_stream());
         }
     }
 }
 
-std::vector<memory_impl::ptr> loop_inst::get_sliced_mem(const primitive_id& internal_id) const {
+std::vector<memory::ptr> loop_inst::get_sliced_mem(const primitive_id& internal_id) const {
         for (const auto& mem_mapping : concatenated_input_mem_mappings) {
             if (mem_mapping.sliced_data_prim->id() == internal_id) {
                 return mem_mapping.sliced_mems;
@@ -331,20 +330,18 @@ std::vector<memory_impl::ptr> loop_inst::get_sliced_mem(const primitive_id& inte
         return {}; // not found
     }
 
-memory_impl::ptr loop_inst::get_external_memory(const primitive_id& external_id) const {
+memory::ptr loop_inst::get_external_memory(const primitive_id& external_id) const {
     const auto outputPrim = _network.get_primitive(external_id);
-    memory_impl& memory = outputPrim->output_memory();
-    return (memory_impl::ptr) &memory;
+    return outputPrim->output_memory_ptr();
 }
 
 loop_inst::typed_primitive_inst(network_impl & network, loop_node const & node)
     : parent(network, node),
       preproc_memories_done(false),
-      body_network(node.get_program()
-        .get_engine()
-        .allocate_network(*node.get_body_program(),
-                          network.get_stream_id(),
-                          false)) {
+      body_network(network_impl::allocate_network(network.get_stream_ptr(),
+                                                  node.get_body_program(),
+                                                  false,
+                                                  network.is_primary_stream())) {
     if (!check_if_axis_is_set_properly(node))
         CLDNN_ERROR_MESSAGE(node.id(), "axis is not set properly");
 
diff --git a/inference-engine/thirdparty/clDNN/src/lrn.cpp b/inference-engine/thirdparty/clDNN/src/lrn.cpp
index c1f0c1b3e1b4e7..2f84b0fad131ee 100644
--- a/inference-engine/thirdparty/clDNN/src/lrn.cpp
+++ b/inference-engine/thirdparty/clDNN/src/lrn.cpp
@@ -4,7 +4,7 @@
 
 #include "lrn_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/lstm.cpp b/inference-engine/thirdparty/clDNN/src/lstm.cpp
index 5c0d4c25ee38fa..c04b686e297577 100644
--- a/inference-engine/thirdparty/clDNN/src/lstm.cpp
+++ b/inference-engine/thirdparty/clDNN/src/lstm.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "lstm_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/lstm_dynamic.cpp b/inference-engine/thirdparty/clDNN/src/lstm_dynamic.cpp
index 2e60fff1285998..4d26546e99556f 100644
--- a/inference-engine/thirdparty/clDNN/src/lstm_dynamic.cpp
+++ b/inference-engine/thirdparty/clDNN/src/lstm_dynamic.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "lstm_dynamic_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/lstm_dynamic_input.cpp b/inference-engine/thirdparty/clDNN/src/lstm_dynamic_input.cpp
index 72e0967007249e..36b346c2f19851 100644
--- a/inference-engine/thirdparty/clDNN/src/lstm_dynamic_input.cpp
+++ b/inference-engine/thirdparty/clDNN/src/lstm_dynamic_input.cpp
@@ -6,7 +6,7 @@
 #include "lstm_dynamic_input_inst.h"
 #include "lstm_dynamic_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/lstm_dynamic_timeloop.cpp b/inference-engine/thirdparty/clDNN/src/lstm_dynamic_timeloop.cpp
index 6db06934be112b..44db5f6ff28a5d 100644
--- a/inference-engine/thirdparty/clDNN/src/lstm_dynamic_timeloop.cpp
+++ b/inference-engine/thirdparty/clDNN/src/lstm_dynamic_timeloop.cpp
@@ -6,7 +6,7 @@
 #include "lstm_dynamic_timeloop_inst.h"
 #include "lstm_dynamic_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/lstm_elt.cpp b/inference-engine/thirdparty/clDNN/src/lstm_elt.cpp
index 2c8346accbc4f8..3e66d985f6dd65 100644
--- a/inference-engine/thirdparty/clDNN/src/lstm_elt.cpp
+++ b/inference-engine/thirdparty/clDNN/src/lstm_elt.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "lstm_elt_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/lstm_gemm.cpp b/inference-engine/thirdparty/clDNN/src/lstm_gemm.cpp
index 4456e29675a042..77f2b62452905d 100644
--- a/inference-engine/thirdparty/clDNN/src/lstm_gemm.cpp
+++ b/inference-engine/thirdparty/clDNN/src/lstm_gemm.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "lstm_gemm_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/max_unpooling.cpp b/inference-engine/thirdparty/clDNN/src/max_unpooling.cpp
index 3e95bd95eb4e3d..b43b72ca084b1b 100644
--- a/inference-engine/thirdparty/clDNN/src/max_unpooling.cpp
+++ b/inference-engine/thirdparty/clDNN/src/max_unpooling.cpp
@@ -5,7 +5,7 @@
 #include "max_unpooling_inst.h"
 #include "primitive_type_base.h"
 #include "sliding_window_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 #include <memory>
diff --git a/inference-engine/thirdparty/clDNN/src/memory.cpp b/inference-engine/thirdparty/clDNN/src/memory.cpp
deleted file mode 100644
index 9cf892500f0bc8..00000000000000
--- a/inference-engine/thirdparty/clDNN/src/memory.cpp
+++ /dev/null
@@ -1,149 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "api/memory.hpp"
-#include "memory_impl.h"
-#include "engine_impl.h"
-
-namespace cldnn {
-
-memory memory::allocate(const engine& engine, const layout& layout, uint32_t net_id, bool reset) {
-    size_t size = layout.bytes_count();
-    if (size == 0)
-        throw std::invalid_argument("size should be more than 0");
-
-    allocation_type type = engine.get()->get_lockable_preffered_memory_allocation_type(layout.format.is_image_2d());
-    return memory(engine.get()->allocate_memory(layout, type, net_id, reset).detach());
-}
-
-memory memory::share_buffer(const engine& engine, const layout& layout, shared_handle buf, uint32_t net_id) {
-    shared_mem_params params = { shared_mem_type::shared_mem_buffer, nullptr, nullptr, buf,
-#ifdef _WIN32
-        nullptr,
-#else
-        0,
-#endif
-        0 };
-    return memory(engine.get()->reinterpret_handle(layout, &params, net_id).detach());
-}
-
-memory memory::share_image(const engine& engine, const layout& layout, shared_handle img, uint32_t net_id) {
-    shared_mem_params params = { shared_mem_type::shared_mem_image, nullptr, nullptr, img,
-#ifdef _WIN32
-        nullptr,
-#else
-        0,
-#endif
-        0 };
-    return memory(engine.get()->reinterpret_handle(layout, &params, net_id).detach());
-}
-
-#ifdef _WIN32
-memory memory::share_surface(const engine& engine, const layout& layout, shared_handle surf, uint32_t plane,
-    uint32_t net_id) {
-    shared_mem_params params = { shared_mem_type::shared_mem_vasurface, nullptr, nullptr, nullptr, surf, plane };
-    return memory(engine.get()->reinterpret_handle(layout, &params, net_id).detach());
-}
-
-memory memory::share_dx_buffer(const engine& engine, const layout& layout, shared_handle res, uint32_t net_id) {
-    shared_mem_params params = { shared_mem_type::shared_mem_dxbuffer, nullptr, nullptr, res, nullptr, 0 };
-    return memory(engine.get()->reinterpret_handle(layout, &params, net_id).detach());
-}
-#else
-memory memory::share_surface(const engine& engine, const layout& layout, shared_surface surf, uint32_t plane,
-    uint32_t net_id) {
-    shared_mem_params params = { shared_mem_type::shared_mem_vasurface, nullptr, nullptr, nullptr, surf, plane };
-    return memory(engine.get()->reinterpret_handle(layout, &params, net_id).detach());
-}
-#endif
-
-size_t memory::count() const {
-    if (_impl)
-        return get_layout().count();
-    else
-        return 0;
-}
-
-size_t memory::size() const {
-    if (_impl)
-        return _impl->size();
-    else
-        return 0;
-}
-
-const layout& memory::get_layout() const {
-    if (_impl)
-        return _impl->get_layout();
-    else
-        throw std::runtime_error("empty memory object");
-}
-
-int memory::get_net_id() const {
-    if (_impl)
-        return _impl->get_net_id();
-    else
-        throw std::runtime_error("empty memory object");
-}
-
-bool memory::is_allocated_by(const engine& engine) const {
-    if (_impl)
-        return _impl->is_allocated_by(*engine.get());
-    else
-        return false;
-}
-
-bool memory::is_the_same_buffer(const memory& other) const {
-    if (_impl == nullptr)
-        return false;
-
-    if (_impl == other.get())
-        return true;
-
-    if (_impl->get_engine() != other.get()->get_engine())
-        return false;
-
-    // User memory, check te pointers
-    if (!_impl->get_engine())
-        return lock_impl() == other.lock_impl();
-
-    // Engine memory, let it decide
-    return _impl->get_engine()->is_the_same_buffer(*_impl, *other.get());
-}
-
-shared_mem_params memory::get_internal_params() const {
-    if (_impl)
-        return _impl->get_internal_params();
-    else
-        throw std::runtime_error("empty memory object");
-}
-
-memory memory::attach_impl(const cldnn::layout& layout, void* ptr, uint32_t net_id) {
-    return memory(new simple_attached_memory(layout, ptr, net_id));
-}
-
-void* memory::lock_impl() const {
-    if (_impl)
-        return _impl->lock();
-    else
-        return nullptr;
-}
-
-void memory::unlock() const {
-    if (_impl)  _impl->unlock();
-}
-
-void memory::retain() {
-    if (_impl)  _impl->add_ref();
-}
-void memory::release() {
-    if (_impl)  _impl->release();
-}
-
-void memory::reset() {
-    release();
-    _impl = nullptr;
-}
-
-}  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/mutable_data.cpp b/inference-engine/thirdparty/clDNN/src/mutable_data.cpp
index 85f73d89c8ade1..7ba0640d9c192c 100644
--- a/inference-engine/thirdparty/clDNN/src/mutable_data.cpp
+++ b/inference-engine/thirdparty/clDNN/src/mutable_data.cpp
@@ -5,9 +5,9 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "mutable_data_inst.h"
 #include "primitive_type_base.h"
-#include "memory_impl.h"
+#include "cldnn/runtime/memory.hpp"
 #include <random>
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 #include <memory>
@@ -20,19 +20,17 @@ primitive_type_id mutable_data::type_id() {
 }
 
 namespace {
-memory_impl::ptr attach_or_copy_data(network_impl& network, memory_impl& mem) {
+memory::ptr attach_or_copy_data(network_impl& network, memory::ptr mem, bool reuse) {
     auto& engine = network.get_engine();
-    auto own_id = network.get_id();
+    auto& stream = network.get_stream();
 
-    if (mem.is_allocated_by(engine) &&
-        (own_id == mem.get_net_id() || network.is_primary_stream())) {
-        mem.set_net(own_id);
-        return (memory_impl::ptr) & mem;
+    if (mem->is_allocated_by(engine) && reuse) {
+        return mem;
     }
 
-    memory_impl::ptr result = engine.allocate_memory(mem.get_layout(), network.get_id(), false);
-    mem_lock<char> src(mem);
-    mem_lock<char> dst(result);
+    memory::ptr result = engine.allocate_memory(mem->get_layout(), false);
+    mem_lock<char> src(mem, stream);
+    mem_lock<char> dst(result, stream);
     std::copy(src.begin(), src.end(), dst.begin());
 
     return result;
@@ -40,65 +38,16 @@ memory_impl::ptr attach_or_copy_data(network_impl& network, memory_impl& mem) {
 }  // namespace
 
 mutable_data_node::typed_program_node(const std::shared_ptr<mutable_data> dprim, program_impl& prog)
-    : parent(dprim, prog), mem(dprim->mem.get()) {
+    : parent(dprim, prog), mem(dprim->mem) {
     recalc_output_layout(false);
     can_share_buffer(false);
-    fill_memory();
 }
 
-void mutable_data_node::attach_memory(memory_impl& new_mem, bool invalidate_users_if_changed) {
-    mem = (memory_impl::ptr) &new_mem;
+void mutable_data_node::attach_memory(memory::ptr new_mem, bool invalidate_users_if_changed) {
+    mem = new_mem;
     recalc_output_layout(invalidate_users_if_changed);
 }
 
-void mutable_data_node::fill_memory() {
-    auto prim = get_primitive();
-
-    if (prim->fill_type == mutable_data::filler_type::no_fill)
-        return;
-
-    auto memory = mem.get();
-    auto layout = memory->get_layout();
-    if (layout.data_type != data_types::f32)
-        CLDNN_ERROR_MESSAGE(id(), "only f32 data types can be filled");
-
-    switch (prim->fill_type) {
-        case mutable_data::filler_type::zero:
-            fill_memory_constant(0.f);
-            break;
-        case mutable_data::filler_type::one:
-            fill_memory_constant(1.f);
-            break;
-        case mutable_data::filler_type::xavier:
-            fill_memory_xavier();
-            break;
-        default:
-            break;
-    }
-}
-
-void mutable_data_node::fill_memory_xavier() {
-    auto memory = mem.get();
-    auto layout = memory->get_layout();
-    auto n = layout.count() / layout.size.batch[0];
-    float scale = static_cast<float>(sqrt(3.0f / static_cast<float>(n)));
-    std::default_random_engine generator(0);
-
-    mem_lock<float> lock(mem);
-    auto out_ptr = lock.begin();
-    std::uniform_real_distribution<float> distribution(-scale, scale);
-    for (uint32_t i = 0; i < (uint32_t)layout.count(); i++) out_ptr[i] = distribution(generator);
-}
-
-void mutable_data_node::fill_memory_constant(float value) {
-    auto memory = mem.get();
-    auto layout = memory->get_layout();
-    mem_lock<float> lock(mem);
-    auto out_ptr = lock.begin();
-
-    for (uint32_t i = 0; i < (uint32_t)layout.count(); i++) out_ptr[i] = value;
-}
-
 std::string mutable_data_inst::to_string(mutable_data_node const& node) {
     auto node_info = node.desc_to_json();
 
@@ -109,6 +58,6 @@ std::string mutable_data_inst::to_string(mutable_data_node const& node) {
 }
 
 mutable_data_inst::typed_primitive_inst(network_impl& network, mutable_data_node const& node)
-    : parent(network, node, *attach_or_copy_data(network, node.get_attached_memory())) {}
+    : parent(network, node, attach_or_copy_data(network, node.get_attached_memory_ptr(), network.is_primary_stream())) {}
 
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/network.cpp b/inference-engine/thirdparty/clDNN/src/network.cpp
index 2ce1c58a1ffcce..14cfd2791a50de 100644
--- a/inference-engine/thirdparty/clDNN/src/network.cpp
+++ b/inference-engine/thirdparty/clDNN/src/network.cpp
@@ -3,26 +3,28 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
+
+#include "cldnn/primitives/data.hpp"
+#include "cldnn/primitives/mutable_data.hpp"
+#include "cldnn/primitives/input_layout.hpp"
+
+#include "cldnn/runtime/error_handler.hpp"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/event.hpp"
+#include "cldnn/runtime/stream.hpp"
+
 #include "network_impl.h"
-#include "engine_impl.h"
-#include "event_impl.h"
 #include "program_impl.h"
-#include "api/data.hpp"
-#include "api/mutable_data.hpp"
-#include "api/input_layout.hpp"
-#include <src/include/to_string_utils.h>
-#include "cldnn_itt.h"
-
-#include "error_handler.h"
+#include "to_string_utils.h"
 #include "primitive_inst.h"
 #include "input_layout_inst.h"
 #include "mutable_data_inst.h"
 #include "condition_inst.h"
 #include "kernel_selector_helper.h"
-#include "gpu/memory_gpu.h"
-#include <algorithm>
+#include "runtime/cldnn_itt.hpp"
 
-#include "gpu/ocl_toolkit.h"
+#include <algorithm>
 #include <string>
 #include <vector>
 #include <memory>
@@ -44,24 +46,27 @@
 namespace cldnn {
 
 network::network(program const& program, uint16_t stream_id)
-    : _impl(program.get()->get_engine().allocate_network(*program.get(), stream_id).detach()) {}
+    : _impl(network_impl::allocate_network(program.get()->get_engine(), program.get(), false, stream_id == 0)) {}
 
-engine network::get_engine() const {
-    auto impl = engine_impl::ptr(&_impl->get_engine());
-    return engine(impl.detach());
+engine& network::get_engine() const {
+    return _impl->get_engine();
 }
 
 program network::get_program() const {
-    auto impl = program_impl::cptr(&_impl->get_program());
-    return program(const_cast<program_impl*>(impl.detach()));
+    auto impl = std::const_pointer_cast<program_impl>(_impl->get_program());
+    return program(impl);
 }
 
-void network::set_input_data(const primitive_id& id, const memory& mem) const {
-    _impl->set_input_data(id, *mem.get());
+void network::set_input_data(const primitive_id& id, memory::ptr mem) const {
+    _impl->set_input_data(id, mem);
 }
 
-void network::set_output_memory(const primitive_id& id, const memory& mem) const {
-    _impl->set_output_memory(id, *mem.get());
+void network::set_output_memory(const primitive_id& id, memory::ptr mem) const {
+    _impl->set_output_memory(id, mem);
+}
+
+stream& network::get_stream() const {
+    return _impl->get_stream();
 }
 
 uint32_t network::get_id() {
@@ -100,28 +105,18 @@ std::vector<primitive_id> network::get_output_ids() const {
     return _impl->get_output_ids();
 }
 
-memory network::get_output_memory(const primitive_id& output_id) const {
-    auto out_mem = memory_impl::ptr(&_impl->get_primitive(output_id)->output_memory());
-    return memory(out_mem.detach());
+memory::ptr network::get_output_memory(const primitive_id& output_id) const {
+    return _impl->get_primitive(output_id)->output_memory_ptr();
 }
 
-event network::get_primitive_event(const primitive_id& output_id) const {
-    auto out_event = _impl->get_primitive_event(output_id);
-    return event(out_event.detach());
+event::ptr network::get_primitive_event(const primitive_id& output_id) const {
+    return _impl->get_primitive_event(output_id);
 }
 
-std::map<primitive_id, network_output> network::execute(const std::vector<event>& dependencies) const {
-    std::vector<refcounted_obj_ptr<event_impl>> dep_impls(dependencies.size());
-
-    std::transform(
-        dependencies.begin(),
-        dependencies.end(),
-        dep_impls.begin(),
-        [](const event& ev) {
-            return event_impl::ptr(ev.get());
-    });
+std::map<primitive_id, network_output> network::execute(const std::vector<event::ptr>& dependencies) const {
+    std::vector<event::ptr> dep_impls(dependencies.size());
 
-    _impl->execute(dep_impls);
+    _impl->execute(dependencies);
 
     auto output_ids = get_output_ids();
     std::map<primitive_id, network_output> result;
@@ -131,14 +126,6 @@ std::map<primitive_id, network_output> network::execute(const std::vector<event>
     return result;
 }
 
-void network::retain() {
-    _impl->add_ref();
-}
-
-void network::release() {
-    _impl->release();
-}
-
 #ifdef DEBUG_DUMP_PATH
 static float convert_half_to_float(half_t val, bool flush_denorm_to_zero = false) {
 #if defined HALF_HALF_HPP
@@ -189,13 +176,14 @@ float convert_element(float f) { return f; }
 float convert_element(half_t h) { return convert_half_to_float(h); }
 
 template <class T>
-static void dump(memory_impl& mem, std::ofstream& file_stream) {
-    auto&& size = mem.get_layout().size;
+static void dump(memory::ptr mem, stream& stream, std::ofstream& file_stream) {
+    auto&& size = mem->get_layout().size;
 
     file_stream << "shape: " << size.to_string() << " ";
-    file_stream << "(count: " << size.count() << ", original format: " << cldnn::fmt_to_str(mem.get_layout().format) << ")" << std::endl;
+    file_stream << "(count: " << size.count() << ", original format: " << cldnn::fmt_to_str(mem->get_layout().format) << ")" << std::endl;
 
-    auto mem_ptr = static_cast<T*>(mem.lock());
+    mem_lock<T> lock(mem, stream);
+    auto mem_ptr = lock.data();
 
     for (cldnn::tensor::value_type g = 0; g < size.group[0]; ++g) {
         for (cldnn::tensor::value_type b = 0; b < size.batch[0]; ++b) {
@@ -205,7 +193,7 @@ static void dump(memory_impl& mem, std::ofstream& file_stream) {
                         for (cldnn::tensor::value_type y = 0; y < size.spatial[1]; ++y) {
                             for (cldnn::tensor::value_type x = 0; x < size.spatial[0]; ++x) {
                                 cldnn::tensor t(cldnn::group(g), cldnn::batch(b), cldnn::feature(f), cldnn::spatial(x, y, z, w));
-                                size_t input_it = mem.get_layout().get_linear_offset(t);
+                                size_t input_it = mem->get_layout().get_linear_offset(t);
                                 file_stream << std::fixed << std::setprecision(6) << convert_element(mem_ptr[input_it]) << std::endl;
                             }
                         }
@@ -214,12 +202,10 @@ static void dump(memory_impl& mem, std::ofstream& file_stream) {
             }
         }
     }
-
-    mem.unlock();
 }
 template <>
-void dump<uint32_t>(memory_impl& mem, std::ofstream& file_stream) {
-    auto&& size = mem.get_layout().size;
+void dump<uint32_t>(memory::ptr mem, stream& stream, std::ofstream& file_stream) {
+    auto&& size = mem->get_layout().size;
 
     file_stream << "shape: ";
     file_stream << size.batch[0] << " ";
@@ -228,7 +214,8 @@ void dump<uint32_t>(memory_impl& mem, std::ofstream& file_stream) {
     file_stream << size.spatial[0] << " ";
     file_stream << "(" << size.batch[0] * size.feature[0] * size.spatial[1] * size.spatial[0] << ")" << std::endl;
 
-    auto mem_ptr = static_cast<uint32_t*>(mem.lock());
+    mem_lock<uint32_t> lock(mem, stream);
+    auto mem_ptr = lock.data();
 
     for (cldnn::tensor::value_type b = 0; b < size.batch[0]; ++b) {
         for (cldnn::tensor::value_type f = 0; f < (cldnn::tensor::value_type)ceil_div(size.feature[0], 32); ++f) {
@@ -236,87 +223,95 @@ void dump<uint32_t>(memory_impl& mem, std::ofstream& file_stream) {
                 for (cldnn::tensor::value_type y = 0; y < size.spatial[1]; ++y) {
                     for (cldnn::tensor::value_type x = 0; x < size.spatial[0]; ++x) {
                         cldnn::tensor t(cldnn::batch(b), cldnn::feature(f), cldnn::spatial(x, y, z, 0));
-                        size_t input_it = mem.get_layout().get_linear_offset(t);
+                        size_t input_it = mem->get_layout().get_linear_offset(t);
                         file_stream << mem_ptr[input_it] << std::endl;
                     }
                 }
             }
         }
     }
-
-    mem.unlock();
 }
 
-static void log_memory_to_file(memory_impl& mem, std::string layerName) {
+static void log_memory_to_file(memory::ptr mem, stream& stream, std::string layerName) {
     std::string filename = layerName;
     std::replace(filename.begin(), filename.end(), '\\', '_');
     std::replace(filename.begin(), filename.end(), '/', '_');
     std::replace(filename.begin(), filename.end(), ' ', '_');
     std::replace(filename.begin(), filename.end(), ':', '_');
-    filename = DEBUG_DUMP_PATH + filename + ".txt";
+        filename = DEBUG_DUMP_PATH + filename + ".txt";
 
     std::ofstream file_stream(filename);
-    if (mem.get_layout().data_type == cldnn::data_types::f32)
-        dump<float>(mem, file_stream);
-    else if (mem.get_layout().data_type == cldnn::data_types::f16)
-        dump<half_t>(mem, file_stream);
-    else if (mem.get_layout().data_type == cldnn::data_types::bin)
-        dump<uint32_t>(mem, file_stream);
-    else if (mem.get_layout().data_type == cldnn::data_types::i32)
-        dump<int32_t>(mem, file_stream);
-    else if (mem.get_layout().data_type == cldnn::data_types::i8)
-        dump<int8_t>(mem, file_stream);
-    else if (mem.get_layout().data_type == cldnn::data_types::u8)
-        dump<uint8_t>(mem, file_stream);
+    auto mem_dt = mem->get_layout().data_type;
+    if (mem_dt == cldnn::data_types::f32)
+        dump<float>(mem, stream, file_stream);
+    else if (mem_dt == cldnn::data_types::f16)
+        dump<half_t>(mem, stream, file_stream);
+    else if (mem_dt == cldnn::data_types::bin)
+        dump<uint32_t>(mem, stream, file_stream);
+    else if (mem_dt == cldnn::data_types::i32)
+        dump<int32_t>(mem, stream, file_stream);
+    else if (mem_dt == cldnn::data_types::i8)
+        dump<int8_t>(mem, stream, file_stream);
+    else if (mem_dt == cldnn::data_types::u8)
+        dump<uint8_t>(mem, stream, file_stream);
 }
 #endif
 /*
 Network_impl will always have net_id = 0 when it will be cldnn internal micronetwork (created i.e by propagate_constants
 opt pass).
 */
-network_impl::network_impl(const program_impl& program, uint16_t stream_id, bool is_internal)
-    : _program(&program), _stream_id(stream_id), _internal(is_internal), _reset_arguments(true) {
+network_impl::network_impl(program_impl::ptr program, stream::ptr stream, bool is_internal, bool is_primary_stream)
+    : _program(program), _stream(stream), _internal(is_internal), _is_primary_stream(is_primary_stream), _reset_arguments(true) {
     static std::atomic<uint32_t> id_gen{0};
     if (!_internal) {
         net_id = ++id_gen;
     }
-    if (net_id) {
-        get_engine().get_context()->add_network(net_id);
-    }
 
     allocate_primitives();
     check_names();
     build_insts_deps();
     build_exec_order();
     validate_primitives();
-    _program->dump_memory_pool();
 }
 
 network_impl::~network_impl() {
-    for (auto const& prim : _exec_order) {
-        prim->cleanup();
-    }
-
-    auto toolkit = get_engine().get_context();
     get_engine().get_memory_pool().clear_pool_for_network(net_id);
-    toolkit->release_pending_memory(net_id);
-    if (net_id) {
-        toolkit->remove_network(net_id);
-    }
 }
 
-network_impl::network_impl(engine_impl& engine,
+network_impl::ptr network_impl::allocate_network(stream::ptr stream, program_impl::ptr program, bool is_internal, bool is_primary_stream) {
+    return std::make_shared<network_impl>(program, stream, is_internal, is_primary_stream);
+}
+
+network_impl::ptr network_impl::allocate_network(engine& engine, program_impl::ptr program, bool is_internal, bool is_primary_stream) {
+    auto stream = engine.create_stream();
+    return std::make_shared<network_impl>(program, stream, is_internal, is_primary_stream);
+}
+
+network_impl::ptr network_impl::build_network(engine& engine,
+                                              const topology_impl& topology,
+                                              const build_options& options,
+                                              bool is_internal) {
+    return std::make_shared<network_impl>(engine, topology, options, is_internal);
+}
+
+network_impl::ptr network_impl::build_network(engine& engine,
+                                              const std::set<std::shared_ptr<program_node>>& nodes,
+                                              const build_options& options,
+                                              bool is_internal) {
+    return std::make_shared<network_impl>(engine, nodes, options, is_internal);
+}
+
+network_impl::network_impl(engine& engine,
                            const topology_impl& topo,
                            const build_options& options,
-                           uint16_t stream_id,
                            bool is_internal)
-    : network_impl(*engine.build_program(topo, options, is_internal), stream_id, is_internal) {}
+    : network_impl(program_impl::build_program(engine, topo, options, is_internal), engine.create_stream(), is_internal) {}
 
-network_impl::network_impl(engine_impl& engine,
+network_impl::network_impl(engine& engine,
                            const std::set<std::shared_ptr<program_node>>& nodes,
                            const build_options& options,
                            bool is_internal)
-    : network_impl(*engine.build_program(nodes, options, is_internal), 0, is_internal) {}
+    : network_impl(program_impl::build_program(engine, nodes, options, is_internal), engine.create_stream(), is_internal) {}
 
 void network_impl::validate_primitives() {
     for (auto const& prim : _exec_order) {
@@ -337,7 +332,7 @@ void network_impl::set_arguments() {
 
 void network_impl::reset_execution(bool wait) {
     if (wait && _events.size() > 0) {
-        std::vector<event_impl::ptr> events;
+        std::vector<event::ptr> events;
         for (auto& pair : _events) {
             auto& ev = pair.second;
             if (ev->is_set())
@@ -346,12 +341,12 @@ void network_impl::reset_execution(bool wait) {
             events.push_back(ev);
         }
 
-        get_engine().wait_for_events(events);
+        get_stream().wait_for_events(events);
     }
     _events.clear();
 }
 
-void network_impl::set_input_data(const primitive_id& id, memory_impl& data) {
+void network_impl::set_input_data(const primitive_id& id, memory::ptr data) {
     std::shared_ptr<primitive_inst> primitive_inst;
 
     primitive_inst = find_primitive(id);
@@ -370,7 +365,7 @@ void network_impl::set_input_data(const primitive_id& id, memory_impl& data) {
     input->set_data(data);
 }
 
-void network_impl::set_output_memory(const primitive_id& id, memory_impl& mem) {
+void network_impl::set_output_memory(const primitive_id& id, memory::ptr mem) {
     std::shared_ptr<primitive_inst> primitive_inst;
 
     primitive_inst = find_primitive(id);
@@ -426,16 +421,6 @@ void network_impl::set_learning_rate(const float lr) { _learning_rate = lr; }
 
 float network_impl::get_learning_rate() { return _learning_rate; }
 
-bool network_impl::is_primary_stream() {
-    auto _nstreams = get_engine().configuration().n_streams;
-    return _nstreams == 1 || (_nstreams > 1 && _stream_id > 0);
-}
-
-bool network_impl::is_secondary_stream() {
-    auto _nstreams = get_engine().configuration().n_streams;
-    return _nstreams > 1 && _stream_id > 0;
-}
-
 std::string network_impl::get_primitive_info(const primitive_id& id) const {
     const auto& node = _program->get_node(id);
     return node.type()->to_string(node);
@@ -452,14 +437,6 @@ void network_impl::allocate_primitives() {
                   return (lhs->get_output_layout().bytes_count() > rhs->get_output_layout().bytes_count());
               });
 
-    std::vector<std::shared_ptr<program_node>> mutable_data_nodes;
-    for (auto const& node : nodes_to_allocate) {
-        if (node->is_type<mutable_data>())
-            mutable_data_nodes.push_back(node);
-    }
-
-    allocate_mutable_data_for_streams(mutable_data_nodes);
-
     for (auto const& node : nodes_to_allocate) {
         allocate_primitive_instance(*node);
     }
@@ -483,30 +460,21 @@ void network_impl::add_to_exec_order(const primitive_id& id) {
     _exec_order.push_back(inst);
 }
 
-void network_impl::execute(const std::vector<refcounted_obj_ptr<event_impl>>& events) {
+void network_impl::execute(const std::vector<event::ptr>& events) {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "NetworkImpl::Execute");
     // Wait for previous execution completion
     reset_execution(false);
 
-    // collect all shared media surfaces and enqueue acquire/relese
-    auto check_and_add_to_return_vec = [](std::shared_ptr<primitive_inst> prim, std::vector<cl_mem>& return_vec) {
-        const auto& mem = prim->output_memory().get_internal_params();
-        if (mem.mem_type == shared_mem_type::shared_mem_vasurface ||
-            mem.mem_type == shared_mem_type::shared_mem_dxbuffer) {
-            return_vec.push_back(static_cast<cl_mem>(mem.mem));
-        }
-    };
-    std::vector<cl_mem> surfaces;
-
+    std::vector<memory::ptr> in_out_mem;
     for (auto& inst : _inputs) {
-        check_and_add_to_return_vec(inst, surfaces);
+        in_out_mem.push_back(inst->output_memory_ptr());
     }
 
     for (auto& inst : _outputs) {
-        check_and_add_to_return_vec(inst, surfaces);
+        in_out_mem.push_back(inst->output_memory_ptr());
     }
-    cl_int err;
-    cl::SharedSurfLock lock(get_engine().get_context()->queue(get_id()).get(), surfaces, &err);
+
+    auto surf_lock = surfaces_lock::create(get_engine().type(), in_out_mem, get_stream());
 
     set_arguments();
 
@@ -523,7 +491,7 @@ void network_impl::execute(const std::vector<refcounted_obj_ptr<event_impl>>& ev
 #endif
             std::cerr << "Dump " << layer_name << " layer" << std::endl;
             for (size_t i = 0; i < get_primitive(inst->id())->dependencies().size(); i++) {
-                log_memory_to_file(get_primitive(inst->id())->dep_memory(i),
+                log_memory_to_file(get_primitive(inst->id())->dep_memory_ptr(i), get_stream(),
                                    layer_name + "_src_" + std::to_string(i));
             }
 #if DUMP_SINGLE_LAYER
@@ -537,15 +505,16 @@ void network_impl::execute(const std::vector<refcounted_obj_ptr<event_impl>>& ev
             inst->set_arguments();
         }
         execute_primitive(inst, events);
+
 #ifdef DEBUG_DUMP_PATH
+        get_stream().finish();
 #if DUMP_SINGLE_LAYER
         if (layer_name == DUMP_LAYER_NAME)
 #endif
         {
-            log_memory_to_file(get_primitive(inst->id())->output_memory(), layer_name + "_dst_0");
+            log_memory_to_file(get_primitive(inst->id())->output_memory_ptr(), get_stream(), layer_name + "_dst_0");
         }
 
-        get_engine().flush_network(get_id());
 #endif
     }
 
@@ -576,19 +545,19 @@ void network_impl::execute(const std::vector<refcounted_obj_ptr<event_impl>>& ev
 
     for (auto& dout : _data_outputs) {  // data primitives are not executed so if they are marked as output we need to add
                                         // them valid events manually
-        _events[dout->id()] = get_engine().create_user_event(get_id(), true);
+        _events[dout->id()] = get_stream().create_user_event(true);
     }
 
     for (auto& prim : _primitives) {
         prim.second->reset_output_change();
     }
 
-    get_engine().get_context()->reset_events(get_id());
+    get_stream().reset_events();
 
-    // Using output of previouse network as input to another one may cause hazard (in OOOQ mode) if user would not
+    // Using output of previous network as input to another one may cause hazard (in OOOQ mode) if user would not
     // provide proper event to execution. Flushing pipeline should prevent this kind of issues.
     // In scenarios with a big number of very small networks it can provide performance drop.
-    get_engine().flush_network(get_id());
+    get_stream().flush();
 }
 
 std::vector<primitive_id> network_impl::get_input_ids() const {
@@ -664,7 +633,7 @@ std::vector<std::shared_ptr<primitive_inst>> network_impl::get_primitives(const
 }
 
 void network_impl::execute_primitive(const std::shared_ptr<primitive_inst>& primitive,
-                                     const std::vector<refcounted_obj_ptr<event_impl>>& events) {
+                                     const std::vector<event::ptr>& events) {
     auto id = primitive->id();
     auto it = _events.find(id);
     bool found = (it != _events.end());
@@ -673,45 +642,10 @@ void network_impl::execute_primitive(const std::shared_ptr<primitive_inst>& prim
                      found,
                      "Primitive " + id + " is tried to be executed for the second time");
 
-    event_impl::ptr ev;
-    if (!get_engine().get_context()->enabled_single_kernel() || get_engine().get_context()->single_kernel_name() == id)
-        ev = primitive->execute(events);
-    else
-        ev = get_engine().create_user_event(get_id(), true);
+    event::ptr ev = primitive->execute(events);
     _events.insert({id, ev});
 }
 
-void network_impl::allocate_mutable_data_for_streams(std::vector<std::shared_ptr<program_node>>& mutable_data_nodes) {
-    // When multiple streams are used, mutable_data should be duplicated for each stream.
-    while (!mutable_data_nodes.empty()) {
-        auto it = mutable_data_nodes.begin();
-        mutable_data_node& node = (*it)->as<mutable_data>();
-        auto mem = node.get_attached_memory_ptr();
-
-        if (is_secondary_stream()) {
-            // Alloc new buffer for this stream and copy data to have valid initial state
-            memory_impl::ptr result = get_engine().allocate_memory(mem->get_layout(), get_id(), false);
-            {
-                mem_lock<char> src(mem);
-                mem_lock<char> dst(result);
-                std::copy(src.begin(), src.end(), dst.begin());
-            }
-
-            // It's possible that several mutable_data nodes use the same memory buffer, so replace all usages
-            for (auto it1 = it; it1 != mutable_data_nodes.end();) {
-                if (get_engine().is_the_same_buffer((*it1)->as<mutable_data>().get_attached_memory(), *mem)) {
-                    (*it1)->as<mutable_data>().attach_memory(*result, false);
-                    it1 = mutable_data_nodes.erase(it1);
-                } else {
-                    ++it1;
-                }
-            }
-        } else {
-            mutable_data_nodes.erase(it);
-        }
-    }
-}
-
 void network_impl::allocate_primitive_instance(program_node const& node) {
     if (_primitives.count(node.id()))
         return;
@@ -746,16 +680,16 @@ void network_impl::transfer_memory_to_device(std::shared_ptr<primitive_inst> ins
     if (node.need_lockable_memory())
         return;
 
+    if (!get_engine().supports_allocation(allocation_type::usm_device))
+        return;
+
     if (alloc_type == allocation_type::usm_host || alloc_type == allocation_type::usm_shared) {
         // Allocate and transfer memory
         auto& mem_pool = inst_mem.get_engine()->get_memory_pool();
-        auto device_mem = inst_mem.get_engine()->allocate_memory(
-            inst_mem.get_layout(),
-            allocation_type::usm_device,
-            inst_mem.get_net_id());
-        dynamic_cast<gpu::gpu_usm&>(*device_mem).copy_from_other(dynamic_cast<gpu::gpu_usm&>(inst_mem));
-        mem_pool.release_memory(&inst_mem, node.id());
-        instance->set_output_memory(*device_mem);
+        auto device_mem = inst_mem.get_engine()->allocate_memory(inst_mem.get_layout(), allocation_type::usm_device, false);
+        device_mem->copy_from(get_stream(), inst_mem);
+        mem_pool.release_memory(&inst_mem, node.id(), get_id());
+        instance->set_output_memory(device_mem);
     }
 }
 }  // namespace cldnn
diff --git a/inference-engine/thirdparty/clDNN/src/nodes_ordering.cpp b/inference-engine/thirdparty/clDNN/src/nodes_ordering.cpp
index a776234eae66cc..54fb3046ddbe4e 100644
--- a/inference-engine/thirdparty/clDNN/src/nodes_ordering.cpp
+++ b/inference-engine/thirdparty/clDNN/src/nodes_ordering.cpp
@@ -6,7 +6,7 @@
 
 #include "program_impl.h"
 #include "program_node.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include <vector>
 #include <map>
 #include <algorithm>
diff --git a/inference-engine/thirdparty/clDNN/src/normalize.cpp b/inference-engine/thirdparty/clDNN/src/normalize.cpp
index 4bafd2263eed20..d342e958b1eea1 100644
--- a/inference-engine/thirdparty/clDNN/src/normalize.cpp
+++ b/inference-engine/thirdparty/clDNN/src/normalize.cpp
@@ -4,7 +4,7 @@
 
 #include "normalize_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/one_hot.cpp b/inference-engine/thirdparty/clDNN/src/one_hot.cpp
index 3b985254018826..d3a02ec58c4f17 100644
--- a/inference-engine/thirdparty/clDNN/src/one_hot.cpp
+++ b/inference-engine/thirdparty/clDNN/src/one_hot.cpp
@@ -4,7 +4,7 @@
 
 #include "one_hot_inst.h"
 
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include "primitive_type_base.h"
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/src/permute.cpp b/inference-engine/thirdparty/clDNN/src/permute.cpp
index c820de4e4255b5..3d6671f923ab93 100644
--- a/inference-engine/thirdparty/clDNN/src/permute.cpp
+++ b/inference-engine/thirdparty/clDNN/src/permute.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "permute_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 
 #include <algorithm>
diff --git a/inference-engine/thirdparty/clDNN/src/pooling.cpp b/inference-engine/thirdparty/clDNN/src/pooling.cpp
index 8376b7404f1a61..73801ffa2b91d1 100644
--- a/inference-engine/thirdparty/clDNN/src/pooling.cpp
+++ b/inference-engine/thirdparty/clDNN/src/pooling.cpp
@@ -5,7 +5,7 @@
 #include "pooling_inst.h"
 #include "primitive_type_base.h"
 #include "sliding_window_utils.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/primitive_inst.cpp b/inference-engine/thirdparty/clDNN/src/primitive_inst.cpp
index 43ec27e3e1e200..9e1ad0902ce7f8 100644
--- a/inference-engine/thirdparty/clDNN/src/primitive_inst.cpp
+++ b/inference-engine/thirdparty/clDNN/src/primitive_inst.cpp
@@ -2,21 +2,19 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "primitive_inst.h"
 #include "data_inst.h"
 #include "mutable_data_inst.h"
 #include "generic_layer_inst.h"
 #include "input_layout_inst.h"
-#include "max_unpooling_inst.h"
 #include "arg_max_min_inst.h"
 #include "fused_conv_eltwise_inst.h"
 
 #include "network_impl.h"
-#include "engine_impl.h"
-#include "memory_impl.h"
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/memory.hpp"
 
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 #include <vector>
@@ -27,7 +25,7 @@ namespace cldnn {
 
 uint32_t primitive_inst::get_network_id() const { return _network.get_id(); }
 
-void primitive_inst::check_memory_to_set(const memory_impl& mem, const layout& layout) const {
+void primitive_inst::check_memory_to_set(const memory& mem, const layout& layout) const {
     CLDNN_ERROR_LAYOUT_MISMATCH("network layout",
         "set memory layout",
         mem.get_layout(),
@@ -60,15 +58,15 @@ void primitive_inst::check_memory_to_set(const memory_impl& mem, const layout& l
     }
 }
 
-void primitive_inst::set_output_memory(memory_impl& mem) {
+void primitive_inst::set_output_memory(memory::ptr mem) {
     auto ol = _node.get_output_layout();
 
-    check_memory_to_set(mem, ol);
+    check_memory_to_set(*mem, ol);
 
-    _output = (memory_impl::ptr) &mem;
+    _output = mem;
 }
 
-event_impl::ptr primitive_inst::execute(const std::vector<event_impl::ptr>& events) {
+event::ptr primitive_inst::execute(const std::vector<event::ptr>& events) {
     const auto primitive_id = id();
     CLDNN_ERROR_BOOL(primitive_id,
                      "Invalid/unset input",
@@ -79,12 +77,12 @@ event_impl::ptr primitive_inst::execute(const std::vector<event_impl::ptr>& even
     if (_exec_deps.empty())
         return _impl->execute(events, *this);
 
-    std::vector<event_impl::ptr> dependencies;
+    std::vector<event::ptr> dependencies;
     dependencies.reserve(_exec_deps.size());
     for (auto& input : _exec_deps) {
         auto id = input->id();
         try {
-            // if the requested event deos not exits it means that it has not been executed, so the processing_order is
+            // if the requested event does not exits it means that it has not been executed, so the processing_order is
             // wrong or synchronization failed.
             auto ev = get_network().get_primitive_event(id);
             dependencies.emplace_back(ev);
@@ -107,10 +105,6 @@ void primitive_inst::set_arguments() {
     _impl->set_arguments(*this);
 }
 
-void primitive_inst::cleanup() {
-    _impl->cleanup(*this);
-}
-
 void primitive_inst::build_deps() {
     if (_deps.empty() && !_node.get_dependencies().empty()) {
         _deps = _network.get_primitives(_node.get_dependencies());
@@ -119,7 +113,7 @@ void primitive_inst::build_deps() {
 }
 
 primitive_inst::primitive_inst(network_impl& network, program_node const& node, bool allocate_memory)
-    : _network(network), _node(node), _impl(node.get_selected_impl()), _output(), _output_changed(false) {
+    : _network(network), _node(node), _impl(node.get_selected_impl() ? node.get_selected_impl()->clone() : nullptr), _output(), _output_changed(false) {
     if (allocate_memory) {
         // In case when output is mutable_data primitive, and other users dependencies are only used for
         // suychronization, The output memory of such primitive will be fused with mutable_data
@@ -153,7 +147,7 @@ primitive_inst::primitive_inst(network_impl& network, program_node const& node,
     }
 }
 
-memory_impl::ptr primitive_inst::allocate_output() {
+memory::ptr primitive_inst::allocate_output() {
     auto layout = _node.get_output_layout();
     auto net_id = get_network_id();
     auto& engine = get_network().get_engine();
@@ -163,33 +157,35 @@ memory_impl::ptr primitive_inst::allocate_output() {
     auto use_lockable_memory = _node.is_output() || _node.get_selected_impl()->is_cpu()
                                || std::any_of(_node.get_users().begin(), _node.get_users().end(),
                                               [](const program_node* n) {return n->get_selected_impl()->is_cpu() || n->can_be_optimized(); })
-                               || engine.supports_allocation(allocation_type::usm_device) == false;
+                               || !engine.supports_allocation(allocation_type::usm_device);
     allocation_type alloc_type = use_lockable_memory ?
                                  engine.get_lockable_preffered_memory_allocation_type(layout.format.is_image_2d())
                                                      : allocation_type::usm_device;
+
     if (!_network.is_internal() && (_node.can_be_optimized() || _node.is_type<generic_layer>())) {
-        return engine.allocate_memory(layout,
-                                      _node.id(),
-                                      net_id,
-                                      _node.get_memory_dependencies(),
-                                      alloc_type,
-                                      false);
+        return engine.get_memory_from_pool(layout,
+                                           _node.id(),
+                                           net_id,
+                                           _node.get_memory_dependencies(),
+                                           alloc_type,
+                                           false);
     } else if (_network.is_internal() && _node.is_output() && _node.is_type<generic_layer>() &&
                engine.supports_allocation(allocation_type::usm_device)) {
-        return engine.allocate_memory(layout, allocation_type::usm_device, net_id, false);
+        return engine.allocate_memory(layout, allocation_type::usm_device, false);
     } else if (_network.is_internal() && !_node.is_output() && _node.is_type<input_layout>()) {
         // Skip memory reset for input_layout primitives, since data will be copied from cldnn::data primitive
         // or just reuse primitive's memory
-        return engine.allocate_memory(layout, alloc_type, net_id, false);
+        return engine.allocate_memory(layout, alloc_type, false);
     } else if (_network.is_internal() || (!_node.can_share_buffer()) || _node.can_be_optimized() || _node.is_output()) {
-        return engine.allocate_memory(layout, alloc_type, net_id);
+        return engine.allocate_memory(layout, alloc_type);
+    } else {
+        return engine.get_memory_from_pool(layout,
+                                           _node.id(),
+                                           net_id,
+                                           _node.get_memory_dependencies(),
+                                           alloc_type,
+                                           true);
     }
-    return engine.allocate_memory(layout,
-                                  _node.id(),
-                                  net_id,
-                                  _node.get_memory_dependencies(),
-                                  alloc_type,
-                                  true);
 }
 
 std::vector<std::shared_ptr<primitive_inst>> primitive_inst::build_exec_deps(
diff --git a/inference-engine/thirdparty/clDNN/src/prior_box.cpp b/inference-engine/thirdparty/clDNN/src/prior_box.cpp
index a85691bb9c6fc1..8d536ac6561d18 100644
--- a/inference-engine/thirdparty/clDNN/src/prior_box.cpp
+++ b/inference-engine/thirdparty/clDNN/src/prior_box.cpp
@@ -4,7 +4,7 @@
 
 #include "prior_box_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 
 #include <cmath>
@@ -21,7 +21,7 @@ primitive_type_id prior_box::type_id() {
 
 namespace {
 template <typename dtype>
-void calculate_prior_box_output(memory_impl& output_mem, layout const& input_layout, prior_box& argument) {
+void calculate_prior_box_output(memory::ptr output_mem, stream& stream, layout const& input_layout, prior_box& argument) {
     // Calculate output.
     // All the inputs for this layer are known at this point,
     // so the output buffer is written here and not in execute().
@@ -39,10 +39,10 @@ void calculate_prior_box_output(memory_impl& output_mem, layout const& input_lay
     const float offset = argument.offset;
     int num_priors = argument.is_clustered() ?
         static_cast<int>(argument.widths.size()) :
-        output_mem.get_layout().size.spatial[1] / 4 / layer_width / layer_height;
+        output_mem->get_layout().size.spatial[1] / 4 / layer_width / layer_height;
     int var_size = static_cast<int>(argument.variance.size());
 
-    mem_lock<dtype> lock{output_mem};
+    mem_lock<dtype> lock{output_mem, stream};
     auto out_ptr = lock.begin();
     int dim = layer_height * layer_width * num_priors * 4;
 
@@ -204,7 +204,7 @@ void calculate_prior_box_output(memory_impl& output_mem, layout const& input_lay
     }
 
     // set the variance.
-    int count = output_mem.get_layout().size.spatial[0] * output_mem.get_layout().size.spatial[1];
+    int count = output_mem->get_layout().size.spatial[0] * output_mem->get_layout().size.spatial[1];
     int var_loop_count = argument.is_clustered() ? var_size : 4;
     for (int h = 0; h < layer_height; ++h) {
         for (int w = 0; w < layer_width; ++w) {
@@ -224,7 +224,7 @@ prior_box_node::typed_program_node(std::shared_ptr<prior_box> prim, program_impl
 }
 
 void prior_box_node::calc_result() {
-    if (result != (memory_impl::ptr) nullptr)
+    if (result != nullptr)
         return;
 
     auto& argument = *typed_desc();
@@ -340,15 +340,17 @@ void prior_box_node::calc_result() {
     CLDNN_ERROR_BOOL(id(), "Prior box padding", is_padded(), "Prior-box layer doesn't support output padding.");
 
     // allocate storage
-    result = get_program().get_engine().allocate_memory(get_output_layout(), 0, false);
+    result = get_program().get_engine().allocate_memory(get_output_layout());
 
     // perform calculations
     if (get_output_layout().data_type == data_types::f16)
-        calculate_prior_box_output<data_type_to_type<data_types::f16>::type>(*result,
+        calculate_prior_box_output<data_type_to_type<data_types::f16>::type>(result,
+                                                                             get_program().get_stream(),
                                                                              input().get_output_layout(),
                                                                              *typed_desc());
     else
-        calculate_prior_box_output<data_type_to_type<data_types::f32>::type>(*result,
+        calculate_prior_box_output<data_type_to_type<data_types::f32>::type>(result,
+                                                                             get_program().get_stream(),
                                                                              input().get_output_layout(),
                                                                              *typed_desc());
 }
diff --git a/inference-engine/thirdparty/clDNN/src/program.cpp b/inference-engine/thirdparty/clDNN/src/program.cpp
index 97eacdf9ae1122..ec59a526be0ac8 100644
--- a/inference-engine/thirdparty/clDNN/src/program.cpp
+++ b/inference-engine/thirdparty/clDNN/src/program.cpp
@@ -4,10 +4,10 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "kernel_selector_helper.h"
-#include "internal_primitive.h"
-#include "internal_primitive_type_base.h"
+#include "device_cache_reader.h"
+#include "auto_tuner.h"
 #include "layout_optimizer.h"
 #include "pass_manager.h"
 #include "primitive_type.h"
@@ -53,12 +53,13 @@
 #include "reduce_inst.h"
 #include "region_yolo_inst.h"
 #include "strided_slice_inst.h"
-#include "to_string_utils.h"
-#include "gpu/memory_gpu.h"
-#include "cldnn_itt.h"
 #include "loop_inst.h"
+#include "to_string_utils.h"
+#include "gpu/register_gpu.hpp"
+#include "runtime/cldnn_itt.hpp"
 
-#include "gpu/ocl_toolkit.h"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/engine.hpp"
 
 #include "kernel_base.h"
 
@@ -76,25 +77,21 @@
 #include <vector>
 #include <stdexcept>
 
-program::program(engine const& engine, topology const& topology, build_options const& options)
-    : _impl(engine.get()->build_program(*topology.get(), options).detach()) {}
-
-void program::retain() {
-    _impl->add_ref();
-}
-
-void program::release() {
-    _impl->release();
-}
+program::program(engine& engine, const topology& topology, const build_options& options)
+    : _impl(program_impl::build_program(engine, *topology.get(), options)) {}
 
-program_impl::program_impl(engine_impl& engine_ref,
+program_impl::program_impl(engine& engine_ref,
                            topology_impl const& topology,
                            build_options const& options,
                            bool is_internal,
                            bool no_optimizations)
-    : engine(&engine_ref),
+    : _engine(engine_ref),
+      _stream(_engine.create_stream()),
+      program_state(_engine),
       options(options),
-      processing_order() {
+      processing_order(),
+      tuning_cache(nullptr) {
+    init_primitives();
     kernel_selector::KernelBase::ResetCounter();
     set_options();
     pm = std::unique_ptr<pass_manager>(new pass_manager(*this));
@@ -106,13 +103,16 @@ program_impl::program_impl(engine_impl& engine_ref,
     }
 }
 
-program_impl::program_impl(engine_impl& engine_ref,
+program_impl::program_impl(engine& engine_ref,
                            std::set<std::shared_ptr<program_node>> const& nodes,
                            build_options const& options,
                            bool is_internal)
-    : engine(&engine_ref),
+    : _engine(engine_ref),
+      program_state(_engine),
       options(options),
-      processing_order() {
+      processing_order(),
+      tuning_cache(nullptr) {
+    init_primitives();
     set_options();
     pm = std::unique_ptr<pass_manager>(new pass_manager(*this));
     prepare_nodes(nodes);
@@ -120,7 +120,58 @@ program_impl::program_impl(engine_impl& engine_ref,
 }
 
 program_impl::~program_impl() {
-    engine->get_context()->remove_program(prog_id);
+}
+
+void program_impl::init_primitives() {
+    static bool is_initialized = false;
+    if (!is_initialized) {
+        gpu::register_implementations_gpu();
+        is_initialized = true;
+    }
+}
+
+void program_impl::compile() {
+    auto& cache = program_state._kernels_cache;
+    cache.build_all();
+}
+
+void program_impl::init_kernels() {
+    for (auto& n : get_processing_order()) {
+        if (n->get_selected_impl())
+            n->get_selected_impl()->init_kernels();
+    }
+}
+
+void program_impl::load_tuning_cache() {
+    OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "ProgramImpl::LoadTuningCache");
+    try {
+        tuning_cache = kernel_selector::CreateTuningCacheFromFile(get_engine().configuration().tuning_cache_path);
+    } catch (...) {
+        tuning_cache = std::make_shared<kernel_selector::TuningCache>();
+    }
+}
+
+kernel_id program_impl::add_kernel(const std::shared_ptr<kernel_string> kernelSring) {
+    return program_state._kernels_cache.set_kernel_source(kernelSring, false);
+}
+
+kernel::ptr program_impl::get_kernel(kernel_id id) {
+    return program_state._kernels_cache.get_kernel(id);
+}
+
+program_impl::ptr program_impl::build_program(engine& engine,
+                                              const topology_impl& topology,
+                                              const build_options& options,
+                                              bool is_internal,
+                                              bool no_optimizations) {
+    return std::make_shared<program_impl>(engine, topology, options, is_internal, no_optimizations);
+}
+
+program_impl::ptr program_impl::build_program(engine& engine,
+                                              const std::set<std::shared_ptr<program_node>>& nodes,
+                                              const build_options& options,
+                                              bool is_internal) {
+    return std::make_shared<program_impl>(engine, nodes, options, is_internal);
 }
 
 program_node& program_impl::get_node(primitive_id const& id) {
@@ -248,7 +299,7 @@ void program_impl::prepare_nodes(std::set<std::shared_ptr<program_node>> const&
     for (const auto& itr : nodes) {
         if (itr.get()->is_type<data>()) {
             get_or_create(std::make_shared<input_layout>(itr.get()->id(),
-                                                         itr.get()->as<data>().get_primitive()->mem.get_layout()));
+                                                         itr.get()->as<data>().get_primitive()->mem->get_layout()));
         } else {
             get_or_create(itr->desc);
         }
@@ -341,11 +392,9 @@ void program_impl::set_options() {
     prog_id = ++id_gen;
     assert(prog_id != 0);
 
-    get_engine().get_context()->add_program(prog_id);
-
     if ((options.get<build_option_type::tuning_config>()->config.mode == tuning_mode::tuning_tune_and_cache ||
          options.get<build_option_type::tuning_config>()->config.mode == tuning_mode::tuning_retune_and_cache) &&
-        !engine->configuration().enable_profiling) {
+        !_engine.configuration().enable_profiling) {
         throw std::invalid_argument("Engine must be created with profiling enabled in tune_and_cache mode!");
     }
 
@@ -360,13 +409,13 @@ void program_impl::build_program(bool is_internal) {
     run_graph_compilation();
     { post_optimize_graph(is_internal); }
     prepare_memory_dependencies();
-    engine->compile_program(*this);
+    compile();
+    init_kernels();
 
-    if (!is_internal)
+    if (!is_internal) {
         prim_info = get_current_stage_info();
-
-    if (!is_internal)
         transfer_memory_to_device();
+    }
 
     cleanup();
 }
@@ -376,7 +425,7 @@ void program_impl::init_graph() {
     apply_opt_pass<graph_initializations>();
 
     for (auto& node : processing_order) {
-        if (!node->is_type<internal_primitive>() && !node->is_type<data>())
+        if (!node->is_type<data>())
             node->get_output_layout();
     }
 
@@ -389,6 +438,10 @@ void program_impl::run_graph_compilation() { apply_opt_pass<compile_graph>(); }
 
 void program_impl::pre_optimize_graph(bool is_internal) {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "ProgramImpl::PreOptimizeGraph");
+
+    if (!is_internal)
+        load_tuning_cache();
+
     // trim to outputs
     apply_opt_pass<trim_to_outputs>();  // ToDo remove hidden dependencies from trimm pass
 
@@ -401,7 +454,7 @@ void program_impl::pre_optimize_graph(bool is_internal) {
 
     bool output_size_handling_enabled = analyze_output_size_handling_need();
     for (auto& node : processing_order) {
-        if (!node->is_type<internal_primitive>() && !node->is_type<data>())
+        if (!node->is_type<data>())
             node->get_output_layout();
     }
 
@@ -522,6 +575,9 @@ void program_impl::mark_if_data_flow(program_node& node) {
 
 void program_impl::transfer_memory_to_device() {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "ProgramImpl::TransferMemory");
+    if (!get_engine().supports_allocation(allocation_type::usm_device))
+        return;
+
     for (auto& node : processing_order) {
         if (node->is_type<data>() && !node->need_lockable_memory()) {
             auto& data_node = node->as<data>();
@@ -535,15 +591,15 @@ void program_impl::transfer_memory_to_device() {
                 throw std::invalid_argument(err_str);
             }
 
+
             if (alloc_type == allocation_type::usm_host || alloc_type == allocation_type::usm_shared) {
                 // Allocate and transfer memory
-                auto device_mem = mem.get_engine()->allocate_memory(data_node_layout,
-                                                                    allocation_type::usm_device,
-                                                                    mem.get_net_id(),
-                                                                    false);
-                dynamic_cast<gpu::gpu_usm&>(*device_mem).copy_from_other(dynamic_cast<gpu::gpu_usm&>(mem));
-                data_node.attach_memory(*device_mem);
-                const_cast<memory&>(data_node.get_primitive()->mem).reset();
+                auto device_mem = mem.get_engine()->allocate_memory(data_node_layout, allocation_type::usm_device, false);
+                device_mem->copy_from(get_stream(), mem);
+                data_node.attach_memory(device_mem);
+                const_cast<memory::ptr&>(data_node.get_primitive()->mem).reset();
+                // TODO: Do we need finish call here? Maybe call it in network::execute() ?
+                get_stream().finish();
             }
         }
     }
@@ -551,8 +607,7 @@ void program_impl::transfer_memory_to_device() {
 
 void program_impl::cleanup() {
     for (auto& node : processing_order)
-        if (!node->is_type<internal_primitive>())
-            node->get_output_layout();
+        node->get_output_layout();
 
     // in debug build, at the end, mark all nodes as outputs so user can query for buffers of all not-optimized nodes,
     // including internal ones etc.
@@ -595,7 +650,7 @@ program_impl::nodes_ordering& program_impl::get_processing_order() { return proc
 const program_impl::nodes_ordering& program_impl::get_processing_order() const { return processing_order; }
 
 void program_impl::prepare_memory_dependencies() {
-    if (!get_engine().configuration().enable_memory_pool)
+    if (!get_engine().configuration().use_memory_pool)
         return;
 
     apply_opt_pass<basic_memory_dependencies>();
@@ -760,18 +815,12 @@ void program_impl::rename(program_node& node, primitive_id const& new_id) {
     nodes_map.emplace(new_id, node_ptr);
     nodes_map.erase(node.id());
 
-    if (!node.is_type<internal_primitive>())
-        const_cast<primitive_id&>(node.desc->id) = new_id;
-    else
-        reinterpret_cast<details::internal_program_node_base&>(node).internal_id = new_id;
+    const_cast<primitive_id&>(node.desc->id) = new_id;
 }
 
 void program_impl::swap_names(program_node& node1, program_node& node2) {
     const auto _extract_id = [](program_node& node) -> primitive_id& {
-        if (!node.is_type<internal_primitive>())
-            return const_cast<primitive_id&>(node.desc->id);
-        else
-            return reinterpret_cast<details::internal_program_node_base&>(node).internal_id;
+        return const_cast<primitive_id&>(node.desc->id);
     };
 
     nodes_map.at(node1.id()).swap(nodes_map.at(node2.id()));
@@ -1011,24 +1060,6 @@ void program_impl::remove_nodes(std::vector<program_node*>& to_remove) {
     }
 }
 
-void program_impl::dump_memory_pool() const {
-    if (!get_engine().configuration().enable_memory_pool)
-        return;
-    auto path = get_dir_path(options);
-    if (path.empty()) {
-        return;
-    }
-    path += "cldnn_memory_pool.log";
-    auto dep = get_memory_dependencies_string();
-    get_engine().dump_memory_pool(*this, path, dep);
-    std::string dump_file_name;
-    if (pm->get_pass_count() < 10)
-        dump_file_name += "0";
-    dump_file_name += std::to_string(pm->get_pass_count()) + "_memory_pool";
-    pm->inc_pass_count();
-    dump_program(dump_file_name.c_str(), true);
-}
-
 // TODO: break this function into number of smaller ones + add per-primitive fields (possibly use
 // primitive_inst::to_string?)
 void program_impl::dump_program(const char* stage,
diff --git a/inference-engine/thirdparty/clDNN/src/program_dump_graph.cpp b/inference-engine/thirdparty/clDNN/src/program_dump_graph.cpp
index 938c888d08939d..12e1abb4f34928 100644
--- a/inference-engine/thirdparty/clDNN/src/program_dump_graph.cpp
+++ b/inference-engine/thirdparty/clDNN/src/program_dump_graph.cpp
@@ -9,8 +9,6 @@
 #include "data_inst.h"
 #include "condition_inst.h"
 
-#include "gpu/ocl_toolkit.h"
-
 #include <algorithm>
 #include <vector>
 #include <string>
@@ -213,7 +211,7 @@ void dump_graph_init(std::ofstream& graph,
             !node->can_be_optimized()) {
             graph << "\\n Selected kernel: "
                   << (node->get_selected_impl() == nullptr ? "none"
-                                                           : node->get_selected_impl().get()->get_kernel_name()) +
+                                                           : node->get_selected_impl()->get_kernel_name()) +
                          "\n" + dump_mem_info(node);
         }
         graph << "\"";
@@ -227,9 +225,6 @@ void dump_graph_init(std::ofstream& graph,
         if (node->is_type<data>() || node->is_constant()) {
             graph << ", shape=box";
         }
-        if (node->is_type<internal_primitive>()) {
-            graph << ", color=blue";
-        }
 
         if (node->is_reusing_memory()) {
             graph << ", fillcolor=\"" << colors[node->get_reused_memory_color() % colors.size()] << "\" ";
diff --git a/inference-engine/thirdparty/clDNN/src/program_helpers.cpp b/inference-engine/thirdparty/clDNN/src/program_helpers.cpp
index 27c67c032116ea..ccd099c2d4a6a7 100644
--- a/inference-engine/thirdparty/clDNN/src/program_helpers.cpp
+++ b/inference-engine/thirdparty/clDNN/src/program_helpers.cpp
@@ -13,24 +13,25 @@
 
 namespace cldnn {
 // helper function for merging the weights/biases buffers on cpu side for depthwise separable convolution optimization
-void program_helpers::merge_buffers(engine_impl& engine,
+void program_helpers::merge_buffers(engine& engine,
                                     program_node& node,
                                     const layout& target_layout,
                                     size_t begin_offset,
                                     size_t end_offset) {
-    memory_impl::ptr data_to_allocate = engine.allocate_memory(target_layout, 0, false);
+    memory::ptr data_to_allocate = engine.allocate_memory(target_layout, false);
+    auto& stream = node.get_program().get_stream();
 
     for (size_t i = begin_offset; i < end_offset; i++) {
         auto& weights = node.get_dependency(i).as<data>();
-        mem_lock<char> src{weights.get_attached_memory()};
-        mem_lock<char> dst{data_to_allocate};
+        mem_lock<char> src{weights.get_attached_memory_ptr(), stream};
+        mem_lock<char> dst{data_to_allocate, stream};
         std::copy(src.begin(), src.end(), dst.begin() + (i - begin_offset) * src.size());
     }
 
     for (size_t i = 0; i < end_offset - begin_offset - 1; i++) node.remove_dependency(begin_offset + 1);
 
     auto& data_node = node.get_dependency(begin_offset).as<data>();
-    data_node.attach_memory(*data_to_allocate, false);
+    data_node.attach_memory(data_to_allocate, false);
 }
 
 void program_helpers::reshape_deconvolution_weights(const std::vector<float> &deconv_weights,
diff --git a/inference-engine/thirdparty/clDNN/src/program_node.cpp b/inference-engine/thirdparty/clDNN/src/program_node.cpp
index 63bdc8f58eb135..4e24eacf25a23b 100644
--- a/inference-engine/thirdparty/clDNN/src/program_node.cpp
+++ b/inference-engine/thirdparty/clDNN/src/program_node.cpp
@@ -71,7 +71,6 @@ std::unique_ptr<json_composite> program_node::desc_to_json() const {
     node_info->add("ptr", "node_" + std::to_string(reinterpret_cast<uintptr_t>(this)));
     node_info->add("id", id());
     node_info->add("type", desc->type_string());
-    node_info->add("internal", bool_to_str(this->is_type<internal_primitive>()));
     node_info->add("valid output layout", bool_to_str(valid_output_layout));
 
     json_composite output_layout_info;
@@ -267,6 +266,10 @@ bool program_node::is_padding_supported(int axis, int padding) const {
     return true;
 }
 
+ void program_node::set_selected_impl(std::unique_ptr<primitive_impl> impl) {
+    selected_impl = std::move(impl);
+}
+
 bool program_node::need_lockable_memory() const {
     bool need_lockable_mem = get_users().empty() || std::any_of(get_users().begin(), get_users().end(), [](const program_node* n) {
         return n->get_selected_impl()->is_cpu();
@@ -274,16 +277,3 @@ bool program_node::need_lockable_memory() const {
 
     return need_lockable_mem;
 }
-
-primitive_id details::internal_program_node_base::get_next_internal_id() {
-    static std::atomic<uint64_t> counter{0};
-    auto idx = counter++;
-    return primitive_id("_cldnn_internal_") + std::to_string(idx);
-}
-
-details::internal_program_node_base::internal_program_node_base(program_impl& prog)
-    : program_node(nullptr, prog), internal_id(get_next_internal_id()) {}
-
-void details::internal_program_node_base::set_implementation(std::unique_ptr<primitive_impl>&& impl) {
-    selected_impl = std::move(impl);
-}
diff --git a/inference-engine/thirdparty/clDNN/src/pyramid_roi_align.cpp b/inference-engine/thirdparty/clDNN/src/pyramid_roi_align.cpp
index bc2bbbc095d96c..37eb2bc16501a8 100644
--- a/inference-engine/thirdparty/clDNN/src/pyramid_roi_align.cpp
+++ b/inference-engine/thirdparty/clDNN/src/pyramid_roi_align.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "pyramid_roi_align_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/quantize.cpp b/inference-engine/thirdparty/clDNN/src/quantize.cpp
index 95f4e5c6be321d..6d26b7fdbdba5b 100644
--- a/inference-engine/thirdparty/clDNN/src/quantize.cpp
+++ b/inference-engine/thirdparty/clDNN/src/quantize.cpp
@@ -5,8 +5,8 @@
 #include "quantize_inst.h"
 #include "binary_convolution_inst.h"
 #include "primitive_type_base.h"
-#include "memory_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include "data_inst.h"
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/src/reduce.cpp b/inference-engine/thirdparty/clDNN/src/reduce.cpp
index d1d66553e9c6f0..e4f4634f062b71 100644
--- a/inference-engine/thirdparty/clDNN/src/reduce.cpp
+++ b/inference-engine/thirdparty/clDNN/src/reduce.cpp
@@ -5,7 +5,7 @@
 #include "reduce_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include "data_inst.h"
 #include <vector>
diff --git a/inference-engine/thirdparty/clDNN/src/reorder.cpp b/inference-engine/thirdparty/clDNN/src/reorder.cpp
index bbe1f897c4e8ce..7300a83e3a4e6a 100644
--- a/inference-engine/thirdparty/clDNN/src/reorder.cpp
+++ b/inference-engine/thirdparty/clDNN/src/reorder.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "reorder_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 
 #include <algorithm>
@@ -235,7 +235,7 @@ void reorder_inst::reuse_input() {
     if (node.requires_reinterpret()) {
         _output = _network.get_engine().reinterpret_buffer(input_memory(), node.get_output_layout());
     } else {
-        _output = (memory_impl::ptr) &input_memory();
+        _output = input_memory_ptr();
     }
 }
 
diff --git a/inference-engine/thirdparty/clDNN/src/resample.cpp b/inference-engine/thirdparty/clDNN/src/resample.cpp
index 882a2361d3e83d..187dbdfc8e46bb 100644
--- a/inference-engine/thirdparty/clDNN/src/resample.cpp
+++ b/inference-engine/thirdparty/clDNN/src/resample.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "resample_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include <string>
 #include <src/include/json_object.h>
 
diff --git a/inference-engine/thirdparty/clDNN/src/reshape.cpp b/inference-engine/thirdparty/clDNN/src/reshape.cpp
index ecdbeb289a9f48..e6fc316cc539f7 100644
--- a/inference-engine/thirdparty/clDNN/src/reshape.cpp
+++ b/inference-engine/thirdparty/clDNN/src/reshape.cpp
@@ -5,8 +5,8 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "reshape_inst.h"
 #include "primitive_type_base.h"
-#include "memory_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/reverse_sequence.cpp b/inference-engine/thirdparty/clDNN/src/reverse_sequence.cpp
index 77f650a8e19ad1..c0c2f1fb51f744 100644
--- a/inference-engine/thirdparty/clDNN/src/reverse_sequence.cpp
+++ b/inference-engine/thirdparty/clDNN/src/reverse_sequence.cpp
@@ -5,7 +5,7 @@
 #include "reverse_sequence_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/roi_pooling.cpp b/inference-engine/thirdparty/clDNN/src/roi_pooling.cpp
index 4c291ec1329fe7..caa6dca083d2dd 100644
--- a/inference-engine/thirdparty/clDNN/src/roi_pooling.cpp
+++ b/inference-engine/thirdparty/clDNN/src/roi_pooling.cpp
@@ -4,7 +4,7 @@
 
 #include "roi_pooling_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/scale.cpp b/inference-engine/thirdparty/clDNN/src/scale.cpp
index 1e881122bdb9c8..52916cf1699bbb 100644
--- a/inference-engine/thirdparty/clDNN/src/scale.cpp
+++ b/inference-engine/thirdparty/clDNN/src/scale.cpp
@@ -4,7 +4,7 @@
 
 #include "scale_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/scatter_elements_update.cpp b/inference-engine/thirdparty/clDNN/src/scatter_elements_update.cpp
index 11165e1b547c27..ccfb299221ea69 100644
--- a/inference-engine/thirdparty/clDNN/src/scatter_elements_update.cpp
+++ b/inference-engine/thirdparty/clDNN/src/scatter_elements_update.cpp
@@ -5,7 +5,7 @@
 #include "scatter_elements_update_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/scatter_nd_update.cpp b/inference-engine/thirdparty/clDNN/src/scatter_nd_update.cpp
index e2935e530f2ef8..435650c28f6d68 100644
--- a/inference-engine/thirdparty/clDNN/src/scatter_nd_update.cpp
+++ b/inference-engine/thirdparty/clDNN/src/scatter_nd_update.cpp
@@ -5,7 +5,7 @@
 #include "scatter_nd_update_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/scatter_update.cpp b/inference-engine/thirdparty/clDNN/src/scatter_update.cpp
index 2359494f4d9f24..cba37e0a4cbdff 100644
--- a/inference-engine/thirdparty/clDNN/src/scatter_update.cpp
+++ b/inference-engine/thirdparty/clDNN/src/scatter_update.cpp
@@ -5,7 +5,7 @@
 #include "scatter_update_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/select.cpp b/inference-engine/thirdparty/clDNN/src/select.cpp
index 5494449d3849df..433e8ba04d2565 100644
--- a/inference-engine/thirdparty/clDNN/src/select.cpp
+++ b/inference-engine/thirdparty/clDNN/src/select.cpp
@@ -5,7 +5,7 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 #include "select_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/shuffle_channels.cpp b/inference-engine/thirdparty/clDNN/src/shuffle_channels.cpp
index 530202d3fc1901..61e5578a8e97b8 100644
--- a/inference-engine/thirdparty/clDNN/src/shuffle_channels.cpp
+++ b/inference-engine/thirdparty/clDNN/src/shuffle_channels.cpp
@@ -5,7 +5,7 @@
 #include "shuffle_channels_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/space_to_batch.cpp b/inference-engine/thirdparty/clDNN/src/space_to_batch.cpp
index 98712851ae3ffb..9de115c3650477 100644
--- a/inference-engine/thirdparty/clDNN/src/space_to_batch.cpp
+++ b/inference-engine/thirdparty/clDNN/src/space_to_batch.cpp
@@ -5,7 +5,7 @@
 #include "space_to_batch_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include "data_inst.h"
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/src/space_to_depth.cpp b/inference-engine/thirdparty/clDNN/src/space_to_depth.cpp
index c3283b32265af6..3fc1f16c3ef116 100644
--- a/inference-engine/thirdparty/clDNN/src/space_to_depth.cpp
+++ b/inference-engine/thirdparty/clDNN/src/space_to_depth.cpp
@@ -5,7 +5,7 @@
 #include "space_to_depth_inst.h"
 
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/split.cpp b/inference-engine/thirdparty/clDNN/src/split.cpp
index 80ee4aac11a753..182fe49f0b083b 100644
--- a/inference-engine/thirdparty/clDNN/src/split.cpp
+++ b/inference-engine/thirdparty/clDNN/src/split.cpp
@@ -4,8 +4,8 @@
 
 #include "split_inst.h"
 #include "primitive_type_base.h"
-#include "memory_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/strided_slice.cpp b/inference-engine/thirdparty/clDNN/src/strided_slice.cpp
index eb409b3572c454..8faf64a3626a64 100644
--- a/inference-engine/thirdparty/clDNN/src/strided_slice.cpp
+++ b/inference-engine/thirdparty/clDNN/src/strided_slice.cpp
@@ -4,7 +4,7 @@
 
 #include "strided_slice_inst.h"
 #include "primitive_type_base.h"
-#include "error_handler.h"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include "data_inst.h"
 #include <string>
diff --git a/inference-engine/thirdparty/clDNN/src/tile.cpp b/inference-engine/thirdparty/clDNN/src/tile.cpp
index c6661655d926f5..798a835af45fe4 100644
--- a/inference-engine/thirdparty/clDNN/src/tile.cpp
+++ b/inference-engine/thirdparty/clDNN/src/tile.cpp
@@ -4,8 +4,8 @@
 
 #include "tile_inst.h"
 #include "primitive_type_base.h"
-#include "memory_impl.h"
-#include "error_handler.h"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/error_handler.hpp"
 #include "json_object.h"
 #include <string>
 
diff --git a/inference-engine/thirdparty/clDNN/src/topology.cpp b/inference-engine/thirdparty/clDNN/src/topology.cpp
index 7d0708669fd0db..4ba2c86a151197 100644
--- a/inference-engine/thirdparty/clDNN/src/topology.cpp
+++ b/inference-engine/thirdparty/clDNN/src/topology.cpp
@@ -3,7 +3,8 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include "api/topology.hpp"
+
+#include "cldnn/graph/topology.hpp"
 #include "topology_impl.h"
 #include <vector>
 #include <memory>
@@ -33,14 +34,6 @@ void topology::add_primitive(std::shared_ptr<primitive> desc) {
     _impl->add(desc);
 }
 
-void topology::retain() {
-    _impl->add_ref();
-}
-
-void topology::release() {
-    _impl->release();
-}
-
 const std::shared_ptr<primitive>& topology::at(const primitive_id& id) const {
     return _impl->at(id);
 }
diff --git a/inference-engine/thirdparty/clDNN/tests/CMakeLists.txt b/inference-engine/thirdparty/clDNN/tests/CMakeLists.txt
index f1107dba4317df..00011e54075fa6 100644
--- a/inference-engine/thirdparty/clDNN/tests/CMakeLists.txt
+++ b/inference-engine/thirdparty/clDNN/tests/CMakeLists.txt
@@ -65,7 +65,7 @@ if(UNIX)
     set_source_files_properties(${__CLDNN_Sources__gtest} PROPERTIES
         COMPILE_FLAGS "-Wno-undef")
 endif()
-  
+
 if (MSVC)
   file(GLOB __CLDNN_Sources__natvis
     "${CMAKE_CURRENT_SOURCE_DIR}/float16.natvis"
@@ -76,7 +76,7 @@ set(__CLDNN_Directory__ks_main           "${CLDNN__KERNEL_SELECTOR_DIR}")
 set(__CLDNN_Directory__ks_core           "${CLDNN__KERNEL_SELECTOR_DIR}/core")
 set(__CLDNN_Directory__ks_common         "${CLDNN__KERNEL_SELECTOR_DIR}/common")
 set(__CLDNN_Directory__ks_core_common    "${__CLDNN_Directory__ks_core}/common")
-set(__CLDNN_Directory__ks_actual_kernels "${__CLDNN_Directory__ks_core}/actual_kernels")  
+set(__CLDNN_Directory__ks_actual_kernels "${__CLDNN_Directory__ks_core}/actual_kernels")
 set(__CLDNN_Directory__ks_cache          "${__CLDNN_Directory__ks_core}/cache")
 
 set(__CLDNN_AllSources
@@ -101,7 +101,9 @@ source_group("${__CLDNN_Label__gtest}"        FILES ${__CLDNN_Sources__gtest})
 include_directories(
     "${CLDNN__MAIN_DIR}"
     "${CLDNN__MAIN_DIR}/src"
+    "${CLDNN__MAIN_DIR}/src/include/"
     "${CMAKE_CURRENT_SOURCE_DIR}"
+    "${CLDNN__RUNTIME_DIR}"
     "${__CLDNN_Directory__test_utils}"
     "${__CLDNN_Directory__ks_core}"
     "${__CLDNN_Directory__ks_core}/common"
@@ -118,20 +120,23 @@ include_directories(SYSTEM
 add_executable("${CLDNN_BUILD__PROJ}"
     ${__CLDNN_AllSources}
   )
+
+if(COMMAND set_ie_threading_interface_for)
+  set_ie_threading_interface_for("${CLDNN_BUILD__PROJ}")
+endif()
+
 set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY PROJECT_LABEL "${CLDNN_BUILD__PROJ_LABEL}")
 set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY OUTPUT_NAME   "${CLDNN_BUILD__PROJ_OUTPUT_NAME}")
 
 # Set library dependencies
-target_link_libraries("${CLDNN_BUILD__PROJ}"
-    "${CLDNN_BUILD__PROJ__clDNN}"
-  )
+target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE "${CLDNN_BUILD__PROJ__clDNN}")
 
 if(WIN32)
-  target_link_libraries("${CLDNN_BUILD__PROJ}" setupapi)
+  target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE setupapi)
 elseif((NOT ANDROID) AND (UNIX))
-  target_link_libraries("${CLDNN_BUILD__PROJ}" pthread)
+  target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE pthread)
 endif()
-target_link_libraries("${CLDNN_BUILD__PROJ}" ${CLDNN__SYSTEM_LINK_LIBRARIES})
+target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE ${CLDNN__SYSTEM_LINK_LIBRARIES})
 
 # =================================== Custom pre- and post-steps =======================================
 
diff --git a/inference-engine/thirdparty/clDNN/tests/module_tests/events_pool_test.cpp b/inference-engine/thirdparty/clDNN/tests/module_tests/events_pool_test.cpp
index 8dc4d57f426be7..b318e806d08f4a 100644
--- a/inference-engine/thirdparty/clDNN/tests/module_tests/events_pool_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/module_tests/events_pool_test.cpp
@@ -2,11 +2,7 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <gtest/gtest.h>
-#include "api/engine.hpp"
-#include "test_utils/test_utils.h"
-#include "api/input_layout.hpp"
-#include "api/network.hpp"
+#include "test_utils.h"
 
 using namespace tests;
 using namespace cldnn;
@@ -35,17 +31,16 @@ TEST(events_pool, DISABLED_basic_test)
 
     for (int i = 0; i < 20; i++)
     {
-        engine eng;// here we build new engine i times
-        auto input = memory::allocate(eng, { data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+        auto eng = engine::create(engine_types::ocl, runtime_types::ocl);// here we build new engine i times
+        auto input = eng->allocate_memory({ data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
         std::vector<float> input_vec = { -1.f, 2.f, -3.f, 4.f };
         for (int j = 0; j < 20; j++) //then we build network j times
         {
-            network network(eng, topology, bo);
+            network network(*eng, topology, bo);
             network.set_input_data("input", input);
             for(int k = 0; k < 20; k++) //and execute that network k times
-                network.execute();  
+                network.execute();
         }
-        EXPECT_EQ(eng.get_max_used_device_memory_size(), (uint64_t)80);
-        eng.~engine();
+        EXPECT_EQ(eng->get_max_used_device_memory(), (uint64_t)80);
     }
 }
diff --git a/inference-engine/thirdparty/clDNN/tests/module_tests/gpu_toolkit_test.cpp b/inference-engine/thirdparty/clDNN/tests/module_tests/gpu_toolkit_test.cpp
deleted file mode 100644
index d7efdc350450c7..00000000000000
--- a/inference-engine/thirdparty/clDNN/tests/module_tests/gpu_toolkit_test.cpp
+++ /dev/null
@@ -1,225 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include <gtest/gtest.h>
-#include "api/engine.hpp"
-#include "test_utils/test_utils.h"
-#include "api/network.hpp"
-#include "api/topology.hpp"
-#include "api/input_layout.hpp"
-#include "api/activation.hpp"
-#include "api/cldnn.hpp"
-
-#include "test_utils.h"
-
-#if defined __clang__
-#pragma clang diagnostic push
-#pragma clang diagnostic ignored "-Wmissing-braces"
-#elif defined __GNUC__ && __GNUC__ >= 6
-#pragma GCC diagnostic ignored "-Wignored-attributes"
-#endif
-
-#include <cl2_wrapper.h>
-
-using namespace cldnn;
-
-class user_gpu_toolkit
-{
-public:
-    user_gpu_toolkit()
-    {
-        get_platform_and_device(get_plaftorm());
-        create_context_from_one_device();
-    }
-
-    cl_context get_gpu_context() const { return _gpu_context; }
-
-private:
-    cl_platform_id _platform_id;
-    cl_device_id _gpu_device;
-    cl_context _gpu_context;
-
-    void create_context_from_one_device()
-    {
-        cl_int error = 0;
-        _gpu_context = clCreateContext(0, 1, &_gpu_device, 0, 0, &error);
-        if (error != CL_SUCCESS)
-        {
-            throw std::runtime_error("error creating context");
-        }
-    }
-
-    cl_platform_id get_plaftorm()
-    {
-        static constexpr auto INTEL_PLATFORM_VENDOR = "Intel(R) Corporation";
-        cl_uint n = 0;
-        cl_int err = clGetPlatformIDs(0, NULL, &n);
-        if (err != CL_SUCCESS) {
-            throw std::runtime_error("clGetPlatformIDs error " + std::to_string(err));
-        }
-
-        // Get platform list
-        std::vector<cl_platform_id> platform_ids(n);
-        err = clGetPlatformIDs(n, platform_ids.data(), NULL);
-        if (err != CL_SUCCESS) {
-            throw std::runtime_error("clGetPlatformIDs error " + std::to_string(err));
-        }
-
-        // find Intel platform
-        for (auto& id : platform_ids) {
-            size_t infoSize;
-            err = clGetPlatformInfo(id, CL_PLATFORM_VENDOR, 0, NULL, &infoSize);
-            if (err != CL_SUCCESS) {
-                throw std::runtime_error("clGetPlatformInfo error " + std::to_string(err));
-            }
-
-            std::vector<char> tmp(infoSize);
-
-            err = clGetPlatformInfo(id, CL_PLATFORM_VENDOR, infoSize, tmp.data(), NULL);
-            if (err != CL_SUCCESS) {
-                throw std::runtime_error("clGetPlatformInfo error " + std::to_string(err));
-            }
-
-            std::string vendor_id(tmp.data());
-            if (vendor_id == std::string(INTEL_PLATFORM_VENDOR))
-                return id;
-        }
-        return static_cast<cl_platform_id>(nullptr);
-    }
-
-    void get_platform_and_device(cl_platform_id platform_id)
-    {
-        _platform_id = platform_id;
-        cl_int err = clGetDeviceIDs(platform_id, CL_DEVICE_TYPE_GPU, 1, &_gpu_device, 0);
-        if (err != CL_SUCCESS) {
-            throw std::runtime_error("clGetDeviceIDs error " + std::to_string(err));
-        }
-    }
-};
-
-TEST(gpu_engine, DISABLED_user_context)
-{
-    user_gpu_toolkit gpu_toolkit;
-    cl_context user_context = gpu_toolkit.get_gpu_context();
-
-    device_query query(static_cast<void*>(user_context));
-    auto devices = query.get_available_devices();
-
-    //[0] Check if the user engine config works.
-    auto engine_config = cldnn::engine_configuration(false, false, false, "", "", true, "", "", cldnn::priority_mode_types::disabled, cldnn::throttle_mode_types::disabled, true, 1);
-
-    //[1]Check if the engine creation works.
-    engine engine(devices.begin()->second, engine_config);
-    auto info = engine.get_info();
-    EXPECT_GT(info.cores_count, 0u);
-    EXPECT_GT(info.core_frequency, 0u);
-
-    //[2]Now check if the queues works (run simple network).
-    topology topo;
-    auto inp_lay = cldnn::layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1,1,2,2 });
-    auto input_mem = cldnn::memory::allocate(engine, inp_lay);
-    tests::set_values<float>(input_mem, { 1.0f, 2.0f, 3.0f, 4.0f });
-    auto inp = input_layout("input", inp_lay);
-    auto activ = activation("this_needs_queue", "input", activation_func::abs);
-    topo.add(inp, activ);
-    network net(engine, topo);
-
-    net.set_input_data("input", input_mem);
-    auto out = net.execute();
-    auto out_ptr = out.at("this_needs_queue").get_memory().pointer<float>();
-    EXPECT_EQ(out.size(), size_t(1));
-    for (uint32_t i = 0; i < 4; i++)
-        EXPECT_EQ(out_ptr[i], float(i + 1));
-}
-
-void execute_simple_topology(cldnn::engine& engine) {
-    auto batch_num = 1;
-    auto feature_num = 4;
-    auto x_size = 1;
-    auto y_size = 1;
-    auto input_tensor = cldnn::tensor(cldnn::spatial(x_size, y_size), cldnn::feature(feature_num), cldnn::batch(batch_num));
-    auto topo = cldnn::topology(
-        cldnn::input_layout("input", { cldnn::data_types::f32, cldnn::format::bfyx, input_tensor  }),
-        cldnn::activation("relu", "input", cldnn::activation_func::relu));
-
-    cldnn::network net(engine, topo);
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfyx, input_tensor });
-    tests::set_values(input_mem, { -1.f, 2.f, -3.f, 4.f });
-    net.set_input_data("input", input_mem);
-    auto outs = net.execute();
-    auto output = outs.at("relu");
-    auto out_ptr = output.get_memory().pointer<float>();
-    ASSERT_EQ(out_ptr[0], 0.0f);
-    ASSERT_EQ(out_ptr[1], 2.0f);
-    ASSERT_EQ(out_ptr[2], 0.0f);
-    ASSERT_EQ(out_ptr[3], 4.0f);
-}
-
-
-TEST(gpu_device_query, get_device_info)
-{
-    cldnn::device_query query;
-    auto devices = query.get_available_devices();
-    auto device_id = devices.begin()->first;
-    auto device = devices.begin()->second;
-    auto device_info = device.get_info();
-
-    //check key and few members, so we know that device info was returned properly
-    ASSERT_EQ(device_id, "0");
-    ASSERT_GT(device_info.cores_count, 0u);
-    ASSERT_GT(device_info.core_frequency, 0u);
-    ASSERT_NE(device_info.dev_name, "");
-    ASSERT_NE(device_info.driver_version, "");
-}
-
-
-TEST(gpu_device_query, get_engine_info)
-{
-    const auto& engine = tests::get_test_engine();
-    auto info = engine.get_info();
-    EXPECT_GT(info.cores_count, 0u);
-    EXPECT_GT(info.core_frequency, 0u);
-}
-
-
-TEST(gpu_device_query, simple)
-{
-    cldnn::device_query query;
-    auto devices = query.get_available_devices();
-    auto device = devices.begin()->second;
-
-    cldnn::engine eng(device);
-    //check if simple execution was finished correctly
-    execute_simple_topology(eng);
-}
-
-TEST(gpu_device_query, DISABLED_release_query)
-{
-    cldnn::device_query query;
-    auto devices = query.get_available_devices();
-    auto device = devices.begin()->second;
-
-    //destroy query
-    query.~device_query();
-    //create engine
-    cldnn::engine eng(device);
-    //check if simple execution was finished correctly
-    execute_simple_topology(eng);
-}
-
-TEST(gpu_device_query, DISABLED_release_device)
-{
-    cldnn::device_query query;
-    auto devices = query.get_available_devices();
-    auto device = devices.begin()->second;
-
-    //destroy query
-    query.~device_query();
-    //create engine
-    cldnn::engine eng(device);
-    //destroy device
-    device.~device();
-    //check if simple execution was finished correctly
-    execute_simple_topology(eng);
-}
diff --git a/inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/graph_manipulation_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/module_tests/graph_manipulation_gpu_test.cpp
similarity index 67%
rename from inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/graph_manipulation_gpu_test.cpp
rename to inference-engine/thirdparty/clDNN/tests/module_tests/graph_manipulation_gpu_test.cpp
index 81e4665287533c..25e6260ea1236a 100644
--- a/inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/graph_manipulation_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/module_tests/graph_manipulation_gpu_test.cpp
@@ -2,14 +2,13 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <memory>
+#include "test_utils.h"
 
-#include <gtest/gtest.h>
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/memory.hpp"
 
 #include "program_impl.h"
 #include "topology_impl.h"
-#include "engine_impl.h"
-#include "memory_impl.h"
 #include "data_inst.h"
 #include "activation_inst.h"
 #include "convolution_inst.h"
@@ -18,29 +17,30 @@
 #include "reshape_inst.h"
 #include "pass_manager.h"
 
-#include "test_utils.h"
 #include "program_impl_wrapper.h"
 
+#include <memory>
+
 using namespace cldnn;
 using namespace ::tests;
 
 /* Basic test to show how the program can be build and run within internal tests
    in similar way as it is done in tests utilizing clDNN API */
 TEST(basic, test1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
     build_opt.set_option(build_option::optimize_data(true));
 
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
-    auto weights1 = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 1 } });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::byxf,{ 1, 1, 1, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto weights1 = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 1 } });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::byxf,{ 1, 1, 1, 2 } });
 
     set_values(input, { FLOAT16(1.1f), FLOAT16(1.2f), FLOAT16(1.3f), FLOAT16(1.4f) });
     set_values(weights1, { FLOAT16(2.1f), FLOAT16(3.1f) });
     set_values(weights2, { 1.1f, 0.1f });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights1", weights1));
     topology.add(data("weights2", weights2));
     topology.add(reshape("reshape1", "weights1", tensor(spatial(1, 2))));
@@ -49,9 +49,9 @@ TEST(basic, test1) {
     topology.add(concatenation("concat", { "reorder1", "weights2" }, concatenation::along_x));
     topology.add(convolution("conv2", { "reorder2" }, { "concat" }));
 
-    program_impl::ptr prog = engine.get()->build_program(*topology.get(), build_opt, false);
-    cldnn::refcounted_obj_ptr<cldnn::network_impl> net = engine.get()->allocate_network(*prog, 0);
-    network network = (cldnn::network) net.get();
+    program_impl::ptr prog = program_impl::build_program(engine, *topology.get(), build_opt, false);
+    std::shared_ptr<cldnn::network_impl> net = network_impl::allocate_network(engine, prog);
+    network network(net);
 
     network.set_input_data("input", input);
 
@@ -60,24 +60,22 @@ TEST(basic, test1) {
     float epsilon = 1e-2f;
     for (auto& it : outputs)
     {
-        auto output = it.second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output(it.second.get_memory(), get_test_stream());
         EXPECT_NEAR(7.8f, output[0], epsilon);
     }
 }
 
-/*
-    This test creates a program without optimization passes, even the compilation is being run manualy.
-    Thus, a single method from program_impl like add_intermediate might be tested separately.
-*/
+// This test creates a program without optimization passes, even the compilation is being run manualy.
+// Thus, a single method from program_impl like add_intermediate might be tested separately.
 TEST(add_intermediate_gpu, test1)
 {
     build_options build_opt;
     topology topology;
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, {2, 2, 2, 2} });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, {2, 2, 2, 2} });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, {2, 2, 2, 2} });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, {2, 2, 2, 2} });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 1, 1 } });
 
     set_values(input, { (1.1f), (1.2f), (1.3f), (1.4f),
                         (2.1f), (2.2f), (2.3f), (2.4f),
@@ -89,21 +87,21 @@ TEST(add_intermediate_gpu, test1)
                           (4.5f), (4.6f), (4.7f), (4.8f) });
 
     set_values(weights2, { (5.5f), (5.6f), (5.7f), (5.8f) });
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights", weights));
     topology.add(data("weights2", weights2));
     topology.add(cldnn::convolution("conv1a", { "input" }, { "weights" }));
     topology.add(cldnn::convolution("conv1b", { "input" }, { "weights" }));
     topology.add(cldnn::convolution("conv2a", { "conv1a" }, { "weights2" }));
-    auto new_reorder = std::make_shared<reorder>("reorder","nothing", input.get_layout());
-    program_impl::ptr prog = engine.get()->build_program(*topology.get(), build_opt, false, true);
+    auto new_reorder = std::make_shared<reorder>("reorder","nothing", input->get_layout());
+    program_impl::ptr prog = program_impl::build_program(engine, *topology.get(), build_opt, false, true);
     prog->add_intermediate(new_reorder, prog->get_node("conv1a"), 0);
     prog->dump_program("custom_dump", true);
 
-    program_impl_wrapper::run_graph_compilation(*prog);
+    program_impl_wrapper::build(*prog);
 
-    cldnn::refcounted_obj_ptr<cldnn::network_impl> net = engine.get()->allocate_network(*prog, 0);
-    network network = (cldnn::network) net.get();
+    std::shared_ptr<cldnn::network_impl> net = network_impl::allocate_network(engine, prog);
+    network network(net);
     network.set_input_data("input", input);
     auto outputs = network.execute();
 
@@ -116,7 +114,7 @@ TEST(add_intermediate_gpu, test1)
     uint32_t output_index = 0;
     for (auto& it : outputs)
     {
-        auto output = it.second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output(it.second.get_memory(), get_test_stream());
         for (uint32_t x = 0; x < output_size; x++)
         {
             EXPECT_FLOAT_EQ(expected_output_vec[x+output_size*output_index], output[x]);
@@ -126,15 +124,16 @@ TEST(add_intermediate_gpu, test1)
 }
 
 /* This test shows how to use private members (here: add_connection) of program_impl using program_impl_wraper */
+// Disabled for now as it produces wrong results
 TEST(add_intermediate_gpu, test2)
 {
     build_options build_opt;
     topology topology;
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 1, 1 } });
 
     set_values(input, { (1.1f), (1.2f), (1.3f), (1.4f),
         (2.1f), (2.2f), (2.3f), (2.4f),
@@ -147,7 +146,7 @@ TEST(add_intermediate_gpu, test2)
 
     set_values(weights2, { (5.5f), (5.6f), (5.7f), (5.8f) });
 
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights2", weights2));
 
     topology.add(cldnn::convolution("conv2a", { "input" }, { "weights2" }));
@@ -157,16 +156,16 @@ TEST(add_intermediate_gpu, test2)
     w_vec.push_back("weights");
     auto new_conv = std::make_shared<convolution>("conv1a", "input", w_vec);
     auto weights_node = std::make_shared<data>("weights", weights);
-    program_impl::ptr prog = engine.get()->build_program(*topology.get(), build_opt, false, true);
+    program_impl::ptr prog = program_impl::build_program(engine, *topology.get(), build_opt, false, true);
 
     prog->add_intermediate(new_conv, prog->get_node("conv2a"), 0, true, true);
     program_impl_wrapper::add_connection(*prog, prog->get_or_create(weights_node), prog->get_or_create(new_conv));
     prog->dump_program("custom_dump", true);
 
-    program_impl_wrapper::run_graph_compilation(*prog);
+    program_impl_wrapper::build(*prog);
 
-    cldnn::refcounted_obj_ptr<cldnn::network_impl> net = engine.get()->allocate_network(*prog, 0);
-    network network = (cldnn::network) net.get();
+    std::shared_ptr<cldnn::network_impl> net = network_impl::allocate_network(engine, prog);
+    network network(net);
     network.set_input_data("input", input);
     auto outputs = network.execute();
 
@@ -177,7 +176,7 @@ TEST(add_intermediate_gpu, test2)
     uint32_t output_size = 4;
     for (auto& it : outputs)
     {
-        auto output = it.second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output(it.second.get_memory(), get_test_stream());
         for (uint32_t x = 0; x < output_size; x++)
         {
             EXPECT_FLOAT_EQ(expected_output_vec[x], output[x]);
diff --git a/inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/prepare_conv_eltw_fusing.cpp b/inference-engine/thirdparty/clDNN/tests/module_tests/prepare_conv_eltw_fusing.cpp
similarity index 78%
rename from inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/prepare_conv_eltw_fusing.cpp
rename to inference-engine/thirdparty/clDNN/tests/module_tests/prepare_conv_eltw_fusing.cpp
index 1e6037fd98a773..1ca951b244d6ce 100644
--- a/inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/prepare_conv_eltw_fusing.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/module_tests/prepare_conv_eltw_fusing.cpp
@@ -2,20 +2,20 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <memory>
+#include "test_utils.h"
 
-#include <gtest/gtest.h>
+#include "cldnn/runtime/engine.hpp"
 
 #include "program_impl.h"
 #include "data_inst.h"
-#include "engine_impl.h"
 #include "eltwise_inst.h"
 #include "network_impl.h"
 #include "pass_manager.h"
 
-#include "test_utils.h"
 #include "program_impl_wrapper.h"
 
+#include <memory>
+
 using namespace cldnn;
 using namespace ::tests;
 
@@ -23,20 +23,19 @@ using namespace ::tests;
 std::map<primitive_id, network_output> test_prepare_conv_eltw_fusing(bool eltw1, bool eltw2)
 {
     build_options build_opt;
-    build_opt.set_option(build_option::graph_dumps_dir("dumps"));
     build_opt.set_option(build_option::optimize_data(true));
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { 1.1f, 1.2f, 1.3f, 1.4f });
     set_values(weights1, { 2.1f});
     set_values(weights2, { -1.5f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights1", weights1));
     topology.add(data("weights2", weights2));
     topology.add(convolution("conv1", { "input" }, { "weights1" }));
@@ -67,15 +66,17 @@ std::map<primitive_id, network_output> test_prepare_conv_eltw_fusing(bool eltw1,
     {
         topology.add(eltwise("eltw3", "conv1", "conv2", cldnn::eltwise_mode::sum));
     }
-    program_impl::ptr prog = engine.get()->build_program(*topology.get(), build_opt, false, true);
+    program_impl::ptr prog = program_impl::build_program(engine, *topology.get(), build_opt, false, true);
 
     layout_optimizer lo;
     program_impl_wrapper::apply_opt_pass<prepare_conv_eltw_fusing>(*prog, lo);
 
     program_impl_wrapper::run_graph_compilation(*prog);
     program_impl_wrapper::prepare_memory_dependencies(*prog);
-    cldnn::refcounted_obj_ptr<cldnn::network_impl> net = engine.get()->allocate_network(*prog, 0);
-    network network = (cldnn::network) net.get();
+    program_impl_wrapper::compile(*prog);
+    program_impl_wrapper::init_kernels(*prog);
+    std::shared_ptr<cldnn::network_impl> net = network_impl::allocate_network(engine, prog);
+    network network(net);
     network.set_input_data("input", input);
 
     return network.execute();
@@ -94,7 +95,7 @@ TEST(prepare_conv_eltw_fusing, testlp)
     float epsilon = 1e-3f;
     for (auto& it : outputs)
     {
-        auto output = it.second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output(it.second.get_memory(), get_test_stream());
         for (int i = 0; i < 4; i++)
             EXPECT_NEAR(ref_out[i], output[i], epsilon);
     }
@@ -114,7 +115,7 @@ TEST(prepare_conv_eltw_fusing, testl)
     float epsilon = 1e-3f;
     for (auto& it : outputs)
     {
-        auto output = it.second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output(it.second.get_memory(), get_test_stream());
         for (int i = 0; i < 4; i++)
             EXPECT_NEAR(ref_out[i], output[i], epsilon);
     }
@@ -135,7 +136,7 @@ TEST(prepare_conv_eltw_fusing, testp)
     float epsilon = 1e-3f;
     for (auto& it : outputs)
     {
-        auto output = it.second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output(it.second.get_memory(), get_test_stream());
         for (int i = 0; i < 4; i++)
             EXPECT_NEAR(ref_out[i], output[i], epsilon);
     }
diff --git a/inference-engine/thirdparty/clDNN/tests_core_internal/program_impl_wrapper.h b/inference-engine/thirdparty/clDNN/tests/module_tests/program_impl_wrapper.h
similarity index 70%
rename from inference-engine/thirdparty/clDNN/tests_core_internal/program_impl_wrapper.h
rename to inference-engine/thirdparty/clDNN/tests/module_tests/program_impl_wrapper.h
index 855f3f050db4d9..03b7bd279703e6 100644
--- a/inference-engine/thirdparty/clDNN/tests_core_internal/program_impl_wrapper.h
+++ b/inference-engine/thirdparty/clDNN/tests/module_tests/program_impl_wrapper.h
@@ -2,6 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+#pragma once
+
 namespace cldnn
 {
     struct program_node;
@@ -24,6 +26,20 @@ namespace cldnn
         {
             p.run_graph_compilation();
         }
+        static void compile(program_impl& p)
+        {
+            p.compile();
+        }
+        static void build(program_impl& p)
+        {
+            program_impl_wrapper::run_graph_compilation(p);
+            program_impl_wrapper::compile(p);
+            program_impl_wrapper::init_kernels(p);
+        }
+        static void init_kernels(program_impl& p)
+        {
+            p.init_kernels();
+        }
         static void prepare_memory_dependencies(program_impl& p)
         {
             p.prepare_memory_dependencies();
diff --git a/inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/reorder_inputs_test.cpp b/inference-engine/thirdparty/clDNN/tests/module_tests/reorder_inputs_test.cpp
similarity index 79%
rename from inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/reorder_inputs_test.cpp
rename to inference-engine/thirdparty/clDNN/tests/module_tests/reorder_inputs_test.cpp
index fb21acf2396ef4..f67512fc3a4ecf 100644
--- a/inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/reorder_inputs_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/module_tests/reorder_inputs_test.cpp
@@ -2,21 +2,21 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <memory>
+#include "test_utils.h"
 
-#include <gtest/gtest.h>
+#include "cldnn/runtime/engine.hpp"
 
 #include "program_impl.h"
 #include "data_inst.h"
-#include "engine_impl.h"
 #include "eltwise_inst.h"
 #include "network_impl.h"
 #include "pass_manager.h"
 #include "to_string_utils.h"
 
-#include "test_utils.h"
 #include "program_impl_wrapper.h"
 
+#include <memory>
+
 using namespace cldnn;
 using namespace ::tests;
 
@@ -31,13 +31,13 @@ TEST(reorder_inputs, propagation) {
     // Format of convolutions should be propagated through pooling.
     // At most single reorder should be inserted before first convolution.
 
-    auto engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb, { 2, 32, 1, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 32, 32, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb, { 2, 32, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 32, 32, 1, 1 } });
 
     topology topology;
     topology.add(data("weights", weights));
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(convolution("conv1", "input", { "weights" }));
     topology.add(pooling("pool", "conv1", pooling_mode::max, { 1, 1, 1, 1 }, { 1, 1, 1, 1 }));
     topology.add(convolution("conv2", "pool", { "weights" }));
@@ -70,11 +70,11 @@ TEST(reorder_inputs, propagation) {
 }
 
 TEST(reorder_inputs, impl_forcing_basic_format) {
-    auto engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 4, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 4, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(pooling("pool", "input", pooling_mode::max, { 1, 1, 2, 1 }, { 1, 1, 2, 1 }));
 
     implementation_desc pool_impl = { format::yxfb, "" };
@@ -91,14 +91,14 @@ TEST(reorder_inputs, impl_forcing_basic_format) {
     network.execute();
 
     auto network_impl = network.get();
-    auto& prog = network_impl->get_program();
-    auto& pool_node = prog.get_node("pool");
+    const auto& prog = network_impl->get_program();
+    auto& pool_node = prog->get_node("pool");
     auto pool_layout = pool_node.get_output_layout();
 
     EXPECT_EQ(pool_layout.format.value, format::yxfb);
 
     auto out_mem = network.get_output("pool").get_memory();
-    auto out_mem_ptr = out_mem.pointer<float>();
+    cldnn::mem_lock<float> out_mem_ptr(out_mem, get_test_stream());
 
     ASSERT_EQ(out_mem_ptr.size(), 4u);
 
@@ -109,11 +109,11 @@ TEST(reorder_inputs, impl_forcing_basic_format) {
 }
 
 TEST(reorder_inputs, impl_forcing_not_existing) {
-    auto engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 4, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 4, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(pooling("pool", "input", pooling_mode::max, { 1, 1, 2, 1 }, { 1, 1, 2, 1 }));
 
     implementation_desc pool_impl = { format::any, "NOT_EXISTING" };
@@ -125,11 +125,11 @@ TEST(reorder_inputs, impl_forcing_not_existing) {
 }
 
 TEST(reorder_inputs, impl_forcing_basic_format_kernel) {
-    auto engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 4, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 4, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(activation("actv", "input", activation_func::relu));
 
     implementation_desc actv_impl = { format::yxfb, "activation_ref" };
@@ -146,8 +146,8 @@ TEST(reorder_inputs, impl_forcing_basic_format_kernel) {
     network.execute();
 
     auto network_impl = network.get();
-    auto& prog = network_impl->get_program();
-    auto& node = prog.get_node("actv");
+    auto prog = network_impl->get_program();
+    auto& node = prog->get_node("actv");
     auto actv_layout = node.get_output_layout();
     auto kernel_name = node.get_selected_impl()->get_kernel_name();
 
@@ -155,7 +155,7 @@ TEST(reorder_inputs, impl_forcing_basic_format_kernel) {
     EXPECT_EQ(kernel_name, actv_impl.kernel_name);
 
     auto out_mem = network.get_output("actv").get_memory();
-    auto out_mem_ptr = out_mem.pointer<float>();
+    cldnn::mem_lock<float> out_mem_ptr(out_mem, get_test_stream());
 
     ASSERT_EQ(out_mem_ptr.size(), 8u);
 
@@ -171,13 +171,13 @@ TEST(reorder_inputs, impl_forcing_basic_format_kernel) {
 
 // TODO Not yet implemented
 //TEST(reorder_inputs, impl_forcing_conv_format_kernel) {
-//    auto engine = get_test_engine();
-//    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, {1, 2, 2, 2} });
-//    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, {2, 2, 1, 1} });
+//    auto& engine = get_test_engine();
+//    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, {1, 2, 2, 2} });
+//    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, {2, 2, 1, 1} });
 //
 //    topology topology;
 //    topology.add(data("weights", weights));
-//    topology.add(input_layout("input", input.get_layout()));
+//    topology.add(input_layout("input", input->get_layout()));
 //    topology.add(convolution("conv", "input", { "weights" }));
 //    topology.add(reorder("output", "conv", format::bfyx, data_types::f32));
 //
@@ -209,7 +209,7 @@ TEST(reorder_inputs, impl_forcing_basic_format_kernel) {
 //        EXPECT_EQ(conv_sel_impl->get_kernel_name(), impl.kernel);
 //
 //        auto out_mem = network.get_output("output").get_memory();
-//        auto out_mem_ptr = out_mem.pointer<float>();
+//        cldnn::mem_lock<float> out_mem_ptr(out_mem, get_test_stream());
 //
 //        EXPECT_EQ(out_mem_ptr.size(), 8);
 //
diff --git a/inference-engine/thirdparty/clDNN/tests/module_tests/test_uqr_distribution.cpp b/inference-engine/thirdparty/clDNN/tests/module_tests/test_uqr_distribution.cpp
index 88ecd67818bb34..0b85f0b1c1c973 100644
--- a/inference-engine/thirdparty/clDNN/tests/module_tests/test_uqr_distribution.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/module_tests/test_uqr_distribution.cpp
@@ -2,9 +2,7 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <gtest/gtest.h>
-
-#include "test_utils/uniform_quantized_real_distribution.hpp"
+#include "test_utils.h"
 
 #include <cmath>
 #include <iomanip>
@@ -17,7 +15,7 @@
 #include <type_traits>
 #include <utility>
 
-namespace cldnn { namespace tests {
+namespace tests {
 
 template <typename RealType>
 struct uniform_quantized_real_distribution_test : ::testing::Test
@@ -1137,4 +1135,4 @@ TYPED_TEST(uniform_quantized_real_distribution_test, DISABLED_generate_random_eq
     }
 }
 
-}} // namespace cldnn { namespace tests {
+}  // namespace tests
diff --git a/inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/usm_memory_test.cpp b/inference-engine/thirdparty/clDNN/tests/module_tests/usm_memory_test.cpp
similarity index 68%
rename from inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/usm_memory_test.cpp
rename to inference-engine/thirdparty/clDNN/tests/module_tests/usm_memory_test.cpp
index d1b35dc774c91b..c67b2516fc78bf 100644
--- a/inference-engine/thirdparty/clDNN/tests_core_internal/test_cases/usm_memory_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/module_tests/usm_memory_test.cpp
@@ -2,14 +2,18 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <memory>
+#include "test_utils.h"
 
-#include <gtest/gtest.h>
+#include "cldnn/runtime/engine.hpp"
+#include "cldnn/runtime/memory.hpp"
+#include "cldnn/runtime/device_query.hpp"
+#include "runtime/ocl/ocl_stream.hpp"
+#include "runtime/ocl/ocl_memory.hpp"
+#include "runtime/ocl/ocl_common.hpp"
+#include "runtime/ocl/ocl_base_event.hpp"
 
 #include "program_impl.h"
 #include "topology_impl.h"
-#include "engine_impl.h"
-#include "memory_impl.h"
 #include "data_inst.h"
 #include "activation_inst.h"
 #include "convolution_inst.h"
@@ -17,14 +21,9 @@
 #include "network_impl.h"
 #include "reshape_inst.h"
 #include "pass_manager.h"
-#include "api/engine.hpp"
-#include "test_utils.h"
 #include "program_impl_wrapper.h"
-#include "gpu/ocl_queue_wrapper.h"
-#include "gpu/memory_gpu.h"
-#include "gpu/ocl_toolkit.h"
-#include "gpu/command_queues_builder.h"
-#include "gpu/ocl_base_event.h"
+
+#include <memory>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -36,7 +35,7 @@ using namespace ::tests;
 #pragma GCC diagnostic ignored "-Wignored-attributes"
 #endif
 
-#include <cl2_wrapper.h>
+#include <ocl/ocl_wrapper.hpp>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -47,21 +46,24 @@ struct usm_test_params{
 
 class BaseUSMTest : public ::testing::TestWithParam<usm_test_params> {
 protected:
-    std::shared_ptr<device> _device = nullptr;
-    std::shared_ptr<engine> _engine = nullptr;
+    std::shared_ptr<ocl::ocl_device> _device = nullptr;
+    std::shared_ptr<ocl::ocl_engine> _engine = nullptr;
     bool _supports_usm = false;
 public:
     void SetUp() override {
         // Find device, which supports USMs.
-        device_query query;
+        device_query query(engine_types::ocl, runtime_types::ocl);
         auto devices = query.get_available_devices();
         for (const auto& d : devices) {
-            if (d.second.get()->mem_caps().supports_usm()) {
-                _device = std::make_shared<device>(d.second);
+            if (d.second->get_mem_caps().supports_usm()) {
+                _device = std::dynamic_pointer_cast<ocl::ocl_device>(d.second);
                 break;
             }
         }
-        _engine = std::make_shared<engine>(_device->get());
+        if (!_device) {
+            GTEST_SUCCEED();
+        }
+        _engine = std::dynamic_pointer_cast<ocl::ocl_engine>(engine::create(engine_types::ocl, runtime_types::ocl, _device));
         _supports_usm = true;
     }
 
@@ -76,8 +78,8 @@ TEST_P(ctor_test, basic) {
         return;
     }
     try {
-        cl::UsmMemory mem(_device->get()->get_context());
-        auto cl_dev = _device->get()->get_device();
+        cl::UsmMemory mem(_device->get_context());
+        auto cl_dev = _device->get_device();
         switch (p.type) {
         case allocation_type::usm_host: {
             mem.allocateHost(1);
@@ -115,35 +117,31 @@ TEST_P(copy_and_read_buffer, basic) {
         return;
     }
     try {
-        gpu::command_queues_builder q_builder(_device->get()->get_context(), _device->get()->get_device(), _device->get()->get_platform());
-        q_builder.build();
-        auto queue = cl::CommandQueueIntel(q_builder.queue());
+        ocl::ocl_stream stream(*_engine);
 
         size_t values_count = 100;
         size_t values_bytes_count = values_count * sizeof(float);
         std::vector<float> src_buffer(values_count);
         std::iota(src_buffer.begin(), src_buffer.end(), 0.0f);
         cldnn::layout linear_layout = cldnn::layout(cldnn::data_types::f32, cldnn::format::bfyx, cldnn::tensor(1, 1, int32_t(values_count), 1));
-        auto cldnn_mem_src = _engine->get()->allocate_memory(linear_layout, p.type);
-        auto ptr_to_fill = cldnn_mem_src->lock();
+        auto cldnn_mem_src = _engine->allocate_memory(linear_layout, p.type);
+
         // Fill src buffer
         switch (p.type) {
         case allocation_type::usm_host:
         case allocation_type::usm_shared: {
-            std::copy(src_buffer.begin(), src_buffer.end(), static_cast<float*>(ptr_to_fill));
-            cldnn_mem_src->unlock();
+            cldnn::mem_lock<float> lock(cldnn_mem_src, stream);
+            std::copy(src_buffer.begin(), src_buffer.end(), lock.data());
             break;
         }
         case allocation_type::usm_device: {
-            auto host_buf = _engine->get()->allocate_memory(linear_layout, allocation_type::usm_host);
-            std::copy(src_buffer.begin(), src_buffer.end(), static_cast<float*>(host_buf->lock()));
-            host_buf->unlock();
-            queue.enqueueCopyUsm(
-                dynamic_cast<gpu::gpu_usm&>(*host_buf).get_buffer(),
-                dynamic_cast<gpu::gpu_usm&>(*cldnn_mem_src).get_buffer(),
-                values_bytes_count,
-                true
-            );
+            auto casted = std::dynamic_pointer_cast<ocl::gpu_usm>(cldnn_mem_src);
+            auto host_buf = _engine->allocate_memory(linear_layout, allocation_type::usm_host);
+            {
+                cldnn::mem_lock<float> lock(host_buf, stream);
+                std::copy(src_buffer.begin(), src_buffer.end(), lock.data());
+            }
+            casted->copy_from(stream, *host_buf);
             break;
         }
         default:
@@ -155,22 +153,17 @@ TEST_P(copy_and_read_buffer, basic) {
         switch (p.type) {
         case allocation_type::usm_host:
         case allocation_type::usm_shared: {
-            auto values_ptr = cldnn_mem_src->lock();
-            std::memcpy(dst_buffer.data(), values_ptr, values_bytes_count);
-            cldnn_mem_src->unlock();
+            cldnn::mem_lock<float> lock(cldnn_mem_src, stream);
+            std::memcpy(dst_buffer.data(), lock.data(), values_bytes_count);
             break;
         }
         case allocation_type::usm_device: {
-            auto host_buf = _engine->get()->allocate_memory(linear_layout, allocation_type::usm_host);
-            queue.enqueueCopyUsm(
-                dynamic_cast<gpu::gpu_usm&>(*cldnn_mem_src).get_buffer(),
-                dynamic_cast<gpu::gpu_usm&>(*host_buf).get_buffer(),
-                values_bytes_count,
-                true
-                );
-            auto values_ptr = host_buf->lock();
-            std::memcpy(dst_buffer.data(), values_ptr, values_bytes_count);
-            host_buf->unlock();
+            auto host_buf = _engine->allocate_memory(linear_layout, allocation_type::usm_host);
+            host_buf->copy_from(stream, *cldnn_mem_src);
+            {
+                cldnn::mem_lock<float> lock(host_buf, stream);
+                std::memcpy(dst_buffer.data(), lock.data(), values_bytes_count);
+            }
             break;
         }
         default:
@@ -197,22 +190,21 @@ TEST_P(fill_buffer, DISABLED_basic) {
         return;
     }
     try {
-        gpu::command_queues_builder q_builder(_device->get()->get_context(), _device->get()->get_device(), _device->get()->get_platform());
-        q_builder.build();
-        auto queue = cl::CommandQueueIntel(q_builder.queue());
+        ocl::ocl_stream stream(*_engine);
+        auto queue = stream.get_cl_queue();
 
         size_t values_count = 100;
         size_t values_bytes_count = values_count * sizeof(float);
-        cl::UsmMemory mem(_device->get()->get_context());
+        cl::UsmMemory mem(_device->get_context());
         switch (p.type) {
         case allocation_type::usm_host:
             mem.allocateHost(values_bytes_count);
             break;
         case allocation_type::usm_shared:
-            mem.allocateShared(_device->get()->get_device(), values_bytes_count);
+            mem.allocateShared(_device->get_device(), values_bytes_count);
             break;
         case allocation_type::usm_device:
-            mem.allocateDevice(_device->get()->get_device(), values_bytes_count);
+            mem.allocateDevice(_device->get_device(), values_bytes_count);
             break;
         default:
             FAIL() << "Not supported allocation type!";
@@ -240,7 +232,7 @@ TEST_P(fill_buffer, DISABLED_basic) {
             break;
         }
         case allocation_type::usm_device: {
-            cl::UsmMemory host_mem(_device->get()->get_context());
+            cl::UsmMemory host_mem(_device->get_context());
             host_mem.allocateHost(values_bytes_count);
             queue.enqueueCopyUsm(
                 mem,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/activation_simple_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/activation_simple_gpu_test.cpp
index 77546a65f4d635..1760a85fdf641b 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/activation_simple_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/activation_simple_gpu_test.cpp
@@ -2,24 +2,18 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/activation.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/reorder.hpp>
 
 #include <cmath>
-#include <gtest/gtest.h>
 #include <algorithm>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/activation.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include <api/data.hpp>
-#include "test_utils/test_utils.h"
-#include "test_utils/float16.h"
-#include "api/reorder.hpp"
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 TEST(activation_f32_fw_gpu, not_basic_yxfb) {
     //  Input:
@@ -34,9 +28,9 @@ TEST(activation_f32_fw_gpu, not_basic_yxfb) {
     //  0, 0, 0, 1, 0,
     //  0, 0, 0, 0, 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     set_values(input,
     { 1.0f, 0.0f, -3.0f, 4.0f, 5.0f,
       0.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -49,7 +43,7 @@ TEST(activation_f32_fw_gpu, not_basic_yxfb) {
         0.0f, 0.0f, 0.0f, 0.0f, 1.0f };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         activation("not", "input", activation_func::negation));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -58,8 +52,8 @@ TEST(activation_f32_fw_gpu, not_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "not");
 
     auto output_memory = outputs.at("not").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -83,9 +77,9 @@ TEST(activation_f32_fw_gpu, erf_basic_yxfb) {
     //  3 -3  3  0  1
     //  1  1  1 -1  0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     set_values(input,
                { 1.0f, 0.0f, -3.0f, 4.0f, 5.0f,
                  0.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -93,7 +87,7 @@ TEST(activation_f32_fw_gpu, erf_basic_yxfb) {
                  1.0f, 1.0f, 1.0f, -1.0f, 0.0f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             activation("not", "input", activation_func::erf));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -102,9 +96,9 @@ TEST(activation_f32_fw_gpu, erf_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "not");
 
     auto output_memory = outputs.at("not").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
-    auto input_ptr = input.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -128,9 +122,9 @@ TEST(activation_f32_fw_gpu, hard_sigmoid_basic_yxfb) {
     //  3 -3  3  0  1
     //  1  1  1 -1  0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     activation_additional_params params = { 1.0f, 0.5f };
     set_values(input,
                { 1.0f, 0.0f, -3.0f, 4.0f, 5.0f,
@@ -139,7 +133,7 @@ TEST(activation_f32_fw_gpu, hard_sigmoid_basic_yxfb) {
                  1.0f, 1.0f, 1.0f, -1.0f, 0.0f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             activation("not", "input", activation_func::hard_sigmoid, params));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -148,9 +142,9 @@ TEST(activation_f32_fw_gpu, hard_sigmoid_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "not");
 
     auto output_memory = outputs.at("not").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
-    auto input_ptr = input.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -175,9 +169,9 @@ TEST(activation_f32_fw_gpu, reciprocal_basic_yxfb) {
     //  3 -3  3  0  1
     //  1  1  1 -1  0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     set_values(input,
                { 1.0f, 0.3f, -3.0f, 4.0f, 5.0f,
                  21.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -185,7 +179,7 @@ TEST(activation_f32_fw_gpu, reciprocal_basic_yxfb) {
                  1.0f, 1.0f, 1.0f, -1.0f, 0.1f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             activation("not", "input", activation_func::reciprocal));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -194,9 +188,9 @@ TEST(activation_f32_fw_gpu, reciprocal_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "not");
 
     auto output_memory = outputs.at("not").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
-    auto input_ptr = input.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -221,9 +215,9 @@ TEST(activation_f32_fw_gpu, selu_basic_yxfb) {
     //  3 -3  3  0  1
     //  1  1  1 -1  0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     activation_additional_params params = { 1.0f, 0.5f };
     set_values(input,
                { 1.0f, 0.3f, -3.0f, 4.0f, 5.0f,
@@ -232,7 +226,7 @@ TEST(activation_f32_fw_gpu, selu_basic_yxfb) {
                  1.0f, 1.0f, 1.0f, -1.0f, 0.1f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             activation("not", "input", activation_func::selu, params));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -241,9 +235,9 @@ TEST(activation_f32_fw_gpu, selu_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "not");
 
     auto output_memory = outputs.at("not").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
-    auto input_ptr = input.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -269,9 +263,9 @@ TEST(activation_f32_fw_gpu, softplus_basic_yxfb) {
     //  3 -3  3  0  1
     //  1  1  1 -1  0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     set_values(input,
                { 1.0f, 0.3f, -3.0f, 4.0f, 5.0f,
                  21.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -279,7 +273,7 @@ TEST(activation_f32_fw_gpu, softplus_basic_yxfb) {
                  1.0f, 1.0f, 1.0f, -1.0f, 0.1f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             activation("not", "input", activation_func::softplus));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -288,9 +282,9 @@ TEST(activation_f32_fw_gpu, softplus_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "not");
 
     auto output_memory = outputs.at("not").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
-    auto input_ptr = input.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -315,9 +309,9 @@ TEST(activation_f32_fw_gpu, softsign_basic_yxfb) {
     //  3 -3  3  0  1
     //  1  1  1 -1  0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     set_values(input,
                { 1.0f, 0.3f, -3.0f, 4.0f, 5.0f,
                  21.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -325,7 +319,7 @@ TEST(activation_f32_fw_gpu, softsign_basic_yxfb) {
                  1.0f, 1.0f, 1.0f, -1.0f, 0.1f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             activation("not", "input", activation_func::softsign));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -334,9 +328,9 @@ TEST(activation_f32_fw_gpu, softsign_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "not");
 
     auto output_memory = outputs.at("not").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
-    auto input_ptr = input.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -361,9 +355,9 @@ TEST(activation_f32_fw_gpu, sign_basic_yxfb) {
     //  3 -3  3  0  1
     //  1  1  1 -1  0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     set_values(input,
                { 1.0f, 0.0f, -3.0f, 4.0f, 5.0f,
                  21.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -371,7 +365,7 @@ TEST(activation_f32_fw_gpu, sign_basic_yxfb) {
                  1.0f, 1.0f, 1.0f, -1.0f, 0.1f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             activation("not", "input", activation_func::sign));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -380,9 +374,9 @@ TEST(activation_f32_fw_gpu, sign_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "not");
 
     auto output_memory = outputs.at("not").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
-    auto input_ptr = input.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -401,15 +395,15 @@ TEST(activation_f32_fw_gpu, sign_basic_yxfb) {
 }
 
 TEST(activation_f32_fw_gpu, pow_basic_yxfb) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
     set_values(input,
     { 1.0f, 2.0f, 3.0f, 4.0f });
     VF<float> output_vec = { 1.0f, 4.0f, 9.0f, 16.0f };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         activation("pow", "input", activation_func::pow, { 2.0f, 0.0f }));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -418,8 +412,8 @@ TEST(activation_f32_fw_gpu, pow_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "pow");
 
     auto output_memory = outputs.at("pow").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -437,15 +431,15 @@ TEST(activation_f32_fw_gpu, pow_basic_yxfb) {
 }
 
 TEST(activation_f16_fw_gpu, pow_basic_yxfb) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb, { 1, 1, 2, 2 } });
     set_values(input,
         { FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f), FLOAT16(4.5f) });
     VF<FLOAT16> output_vec = { FLOAT16(1.0f), FLOAT16(8.0f), FLOAT16(27.0f), FLOAT16(91.125f) };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         activation("pow", "input", activation_func::pow, { FLOAT16(3.0f), FLOAT16(0.0f) }));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -454,8 +448,8 @@ TEST(activation_f16_fw_gpu, pow_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "pow");
 
     auto output_memory = outputs.at("pow").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<FLOAT16>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<FLOAT16> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -487,9 +481,9 @@ TEST(activation_f32_fw_gpu, relu_basic_yxfb) {
     //  3   -1.5  3    5    1
     //  1    1    1   -0.5  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     set_values(input,
     { 1.0f, -2.0f, -3.0f, 4.0f, 5.0f,
       2.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -502,7 +496,7 @@ TEST(activation_f32_fw_gpu, relu_basic_yxfb) {
         1.0f, 1.0f, 1.0f, -0.5f, 1.0f };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         activation("relu", "input", activation_func::relu_negative_slope, { 0.5f, 0.f }, padding{ { 0, 0, 0, 0 }, 0 }));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -511,8 +505,8 @@ TEST(activation_f32_fw_gpu, relu_basic_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "relu");
 
     auto output_memory = outputs.at("relu").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -555,9 +549,9 @@ TEST(activation_f32_fw_gpu, relu_basic_bfzyx) {
     //  1    2    1   -1    2
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 1, 5, 4, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 1, 5, 4, 2 } });
     set_values(input,
     { 1.0f, -2.0f, -3.0f, 4.0f, 5.0f,
         2.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -578,7 +572,7 @@ TEST(activation_f32_fw_gpu, relu_basic_bfzyx) {
         1.0f, 2.0f, 1.0f, -1.0f, 2.0f };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         activation("relu", "input", activation_func::relu_negative_slope, { 0.5f, 0.f }, padding{ { 0, 0, 0, 0, 0 }, 0 }));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -587,8 +581,8 @@ TEST(activation_f32_fw_gpu, relu_basic_bfzyx) {
     EXPECT_EQ(outputs.begin()->first, "relu");
 
     auto output_memory = outputs.at("relu").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int z_size = output_layout.size.spatial[2];
     int y_size = output_layout.size.spatial[1];
@@ -618,10 +612,10 @@ TEST(activation_f32_fw_gpu, basic_yxfb_all_functions)
     //  a: 0.5, b: 2.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 5, 4 } });
-    auto input_params = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 5, 4 } });
+    auto input_params = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 2, 1, 1 } });
     set_values(input,
     { 0.0f, -2.0f, -3.0f, 4.0f, 5.0f,
         2.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -665,7 +659,7 @@ TEST(activation_f32_fw_gpu, basic_yxfb_all_functions)
     {
         for (auto func : funcs)
         {
-            topology topology(input_layout("input", input.get_layout()));
+            topology topology(input_layout("input", input->get_layout()));
 
             if (i == 0)
             {
@@ -684,9 +678,9 @@ TEST(activation_f32_fw_gpu, basic_yxfb_all_functions)
             EXPECT_EQ(outputs.begin()->first, "activation");
 
             auto output_memory = outputs.at("activation").get_memory();
-            auto output_layout = output_memory.get_layout();
-            auto output_ptr = output_memory.pointer<float>();
-            auto input_ptr = input.pointer<float>();
+            auto output_layout = output_memory->get_layout();
+            cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+            cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
             int y_size = output_layout.size.spatial[1];
             int x_size = output_layout.size.spatial[0];
@@ -796,10 +790,10 @@ TEST(activation_f32_fw_gpu, basic_yxfb_all_functions)
 
 TEST(activation_f16_fw_gpu, basic_bfyx_all_functions)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 1, 2, 4 } });
-    auto input_params = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 1, 2, 4 } });
+    auto input_params = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, { FLOAT16(-4.5f), FLOAT16(-2.5f), FLOAT16(-1.5f), FLOAT16(0.5f),
                         FLOAT16(0.9f),  FLOAT16(1.5f),  FLOAT16(2.0f),  FLOAT16(2.5f) });
@@ -818,7 +812,7 @@ TEST(activation_f16_fw_gpu, basic_bfyx_all_functions)
 
     for (uint8_t i = 0 ; i < 2 ; i++) {
         for (auto func : funcs) {
-            topology topology(input_layout("input", input.get_layout()));
+            topology topology(input_layout("input", input->get_layout()));
 
             if (i == 0) {
                 topology.add(activation("activation", "input", func, params));
@@ -834,9 +828,9 @@ TEST(activation_f16_fw_gpu, basic_bfyx_all_functions)
             EXPECT_EQ(outputs.begin()->first, "activation");
 
             auto output_memory = outputs.at("activation").get_memory();
-            auto output_layout = output_memory.get_layout();
-            auto output_ptr = output_memory.pointer<FLOAT16>();
-            auto input_ptr = input.pointer<FLOAT16>();
+            auto output_layout = output_memory->get_layout();
+            cldnn::mem_lock<FLOAT16> output_ptr(output_memory, get_test_stream());
+            cldnn::mem_lock<FLOAT16> input_ptr(input, get_test_stream());
 
             int y_size = output_layout.size.spatial[1];
             int x_size = output_layout.size.spatial[0];
@@ -884,9 +878,9 @@ TEST(activation_f16_fw_gpu, basic_bfyx_all_functions)
 
 TEST(activation_f32_fw_gpu, basic_yxfb_asin_acos_log_atan)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 4 } });
     set_values(input, { 0.12f, 0.56f, 0.45f, 0.789f, 0.546f, 0.999f, 0.7899f, 0.6677f});
 
     std::vector<activation_func> funcs = {
@@ -902,7 +896,7 @@ TEST(activation_f32_fw_gpu, basic_yxfb_asin_acos_log_atan)
 
     for (auto func : funcs)
     {
-        topology topology(input_layout("input", input.get_layout()));
+        topology topology(input_layout("input", input->get_layout()));
         topology.add(activation("activation", "input", func));
 
         network network(engine, topology);
@@ -912,9 +906,9 @@ TEST(activation_f32_fw_gpu, basic_yxfb_asin_acos_log_atan)
         EXPECT_EQ(outputs.begin()->first, "activation");
 
         auto output_memory = outputs.at("activation").get_memory();
-        auto output_layout = output_memory.get_layout();
-        auto output_ptr = output_memory.pointer<float>();
-        auto input_ptr = input.pointer<float>();
+        auto output_layout = output_memory->get_layout();
+        cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+        cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
         int y_size = output_layout.size.spatial[1];
         int x_size = output_layout.size.spatial[0];
@@ -977,9 +971,9 @@ TEST(activation_f32_fw_gpu, relu_basic_acosh_yxfb) {
     //
     //  Slope: 0.5
 
-    const auto &engine = get_test_engine();
+    auto &engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::yxfb, {1, 1, 5, 4}});
+    auto input = engine.allocate_memory({data_types::f32, format::yxfb, {1, 1, 5, 4}});
 
     set_values(input,
                {1.0f, 2.0f, 3.0f, 4.0f, 5.0f,
@@ -988,8 +982,8 @@ TEST(activation_f32_fw_gpu, relu_basic_acosh_yxfb) {
                 1.0f, 1.0f, 1.0f, 1.0f, 1.0f});
 
     topology topology(
-            input_layout("input", input.get_layout()),
-            reorder("reorder", "input", input.get_layout().with_padding(padding{{0, 0, 2, 1}, 0})),
+            input_layout("input", input->get_layout()),
+            reorder("reorder", "input", input->get_layout().with_padding(padding{{0, 0, 2, 1}, 0})),
             activation("relu", "reorder", activation_func::acosh, {0.5f, 0.f}, padding{{0, 0, 0, 0}, 0}));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -997,9 +991,9 @@ TEST(activation_f32_fw_gpu, relu_basic_acosh_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "relu");
 
     auto output_memory = outputs.at("relu").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
-    auto input_ptr = input.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -1038,9 +1032,9 @@ TEST(activation_f32_fw_gpu, relu_basic_input_padding_yxfb) {
     //  3   -1.5  3    5    1
     //  1    1    1   -0.5  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
 
     set_values(input,
     { 1.0f, -2.0f, -3.0f, 4.0f, 5.0f,
@@ -1054,8 +1048,8 @@ TEST(activation_f32_fw_gpu, relu_basic_input_padding_yxfb) {
          1.0f, 1.0f, 1.0f, -0.5f, 1.0f};
 
     topology topology(
-        input_layout("input", input.get_layout()),
-        reorder("reorder", "input", input.get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })),
+        input_layout("input", input->get_layout()),
+        reorder("reorder", "input", input->get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })),
         activation("relu", "reorder", activation_func::relu_negative_slope, { 0.5f, 0.f }, padding{ { 0, 0, 0, 0 }, 0 }));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -1063,8 +1057,8 @@ TEST(activation_f32_fw_gpu, relu_basic_input_padding_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "relu");
 
     auto output_memory = outputs.at("relu").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -1116,9 +1110,9 @@ TEST(activation_f32_fw_gpu, relu_basic_input_padding_bfzyx) {
     //  3   -1.5  3    5    1
     //  1    1    1   -0.5  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 1, 5, 4, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 1, 5, 4, 2 } });
     set_values(input,
     {
         1.0f, -2.0f, -3.0f, 4.0f, 5.0f,
@@ -1141,8 +1135,8 @@ TEST(activation_f32_fw_gpu, relu_basic_input_padding_bfzyx) {
         1.0f, 1.0f, 1.0f, -0.5f, 1.0f };
 
     topology topology(
-        input_layout("input", input.get_layout()),
-        reorder("reorder", "input", input.get_layout().with_padding(padding{ { 0, 0, 2, 1, 0 }, 0 })),
+        input_layout("input", input->get_layout()),
+        reorder("reorder", "input", input->get_layout().with_padding(padding{ { 0, 0, 2, 1, 0 }, 0 })),
         activation("relu", "reorder", activation_func::relu_negative_slope, { 0.5f, 0.f }, padding{ { 0, 0, 0, 0, 0 }, 0 }));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -1151,8 +1145,8 @@ TEST(activation_f32_fw_gpu, relu_basic_input_padding_bfzyx) {
 
     auto output_memory = outputs.at("relu").get_memory();
 
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int z_size = output_layout.size.spatial[2];
     int y_size = output_layout.size.spatial[1];
@@ -1194,9 +1188,9 @@ TEST(activation_f32_fw_gpu, relu_basic_output_padding_yxfb) {
     //  0    0    0    0    0    0    0    0    0    0    0
     //  0    0    0    0    0    0    0    0    0    0    0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     set_values(input,
     { 1.0f, -2.0f, -3.0f, 4.0f, 5.0f,
         2.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -1215,7 +1209,7 @@ TEST(activation_f32_fw_gpu, relu_basic_output_padding_yxfb) {
         0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         activation("relu", "input", activation_func::relu_negative_slope, { 0.5f, 0.f }, padding{ { 0, 0, 3, 3 }, 0 }));
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -1224,9 +1218,9 @@ TEST(activation_f32_fw_gpu, relu_basic_output_padding_yxfb) {
     EXPECT_EQ(outputs.begin()->first, "relu");
 
     auto output_memory = outputs.at("relu").get_memory();
-    auto output_layout = output_memory.get_layout();
+    auto output_layout = output_memory->get_layout();
     auto output_size = output_layout.get_buffer_size();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_size.spatial[1];
     int x_size = output_size.spatial[0];
@@ -1245,9 +1239,9 @@ TEST(activation_f32_fw_gpu, relu_basic_output_padding_yxfb) {
 
 TEST(activation_f32_fw_gpu, basic_yxfb_floor_ceil)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 4 } });
     set_values(input, { 0.01f, 0.99f, -0.01f, -0.99f, 1.1f, 1.0f, 0.0f, -1.1f });
 
     std::vector<activation_func> funcs = {
@@ -1257,7 +1251,7 @@ TEST(activation_f32_fw_gpu, basic_yxfb_floor_ceil)
 
     for (auto func : funcs)
     {
-        topology topology(input_layout("input", input.get_layout()));
+        topology topology(input_layout("input", input->get_layout()));
         topology.add(activation("activation", "input", func));
 
         network network(engine, topology);
@@ -1267,9 +1261,9 @@ TEST(activation_f32_fw_gpu, basic_yxfb_floor_ceil)
         EXPECT_EQ(outputs.begin()->first, "activation");
 
         auto output_memory = outputs.at("activation").get_memory();
-        auto output_layout = output_memory.get_layout();
-        auto output_ptr = output_memory.pointer<float>();
-        auto input_ptr = input.pointer<float>();
+        auto output_layout = output_memory->get_layout();
+        cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
+        cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
         int y_size = output_layout.size.spatial[1];
         int x_size = output_layout.size.spatial[0];
@@ -1300,8 +1294,8 @@ TEST(activation_f32_fw_gpu, basic_yxfb_floor_ceil)
 
 TEST(activation_i8_fw_gpu, basic_yxfb_all_funcs)
 {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::i8, format::yxfb,{ 2, 2, 2, 2 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::i8, format::yxfb,{ 2, 2, 2, 2 } });
 
     std::vector<int8_t> input_vec = {
         1,   0,  5,   1,
@@ -1321,7 +1315,7 @@ TEST(activation_i8_fw_gpu, basic_yxfb_all_funcs)
     for (auto func : funcs)
     {
         topology topology;
-        topology.add(input_layout("input", input.get_layout()));
+        topology.add(input_layout("input", input->get_layout()));
         topology.add(activation("activation", "input", func));
 
         network network(engine, topology);
@@ -1332,14 +1326,12 @@ TEST(activation_i8_fw_gpu, basic_yxfb_all_funcs)
         EXPECT_EQ(outputs.begin()->first, "activation");
 
         auto output_memory = outputs.at("activation").get_memory();
-        auto output_layout = output_memory.get_layout();
-        auto output_ptr = output_memory.pointer<int8_t>();
-        auto input_ptr = input.pointer<int8_t>();
+        auto output_layout = output_memory->get_layout();
+        cldnn::mem_lock<int8_t> output_ptr(output_memory, get_test_stream());
+        cldnn::mem_lock<int8_t> input_ptr(input, get_test_stream());
 
-        for (size_t i = 0; i < output_layout.get_linear_size(); ++i)
-        {
-            switch (func)
-            {
+        for (size_t i = 0; i < output_layout.get_linear_size(); ++i) {
+            switch (func) {
             case activation_func::none:
                 EXPECT_EQ((int8_t)input_ptr[i], output_ptr[i]);
                 break;
@@ -1356,10 +1348,9 @@ TEST(activation_i8_fw_gpu, basic_yxfb_all_funcs)
     }
 }
 
-TEST(activation_i32_fw_gpu, basic_yxfb_i32_funcs)
-{
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::i32, format::yxfb,{ 2, 2, 2, 2 } });
+TEST(activation_i32_fw_gpu, basic_yxfb_i32_funcs) {
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::i32, format::yxfb,{ 2, 2, 2, 2 } });
 
     std::vector<int32_t> input_vec = {
         1,   0,  5,   1,
@@ -1378,11 +1369,10 @@ TEST(activation_i32_fw_gpu, basic_yxfb_i32_funcs)
         activation_func::clamp
     };
 
-    for (auto func : funcs)
-    {
+    for (auto func : funcs) {
         topology topology;
         activation_additional_params params = {0.0, 1.0};
-        topology.add(input_layout("input", input.get_layout()));
+        topology.add(input_layout("input", input->get_layout()));
         topology.add(activation("activation", "input", func, params));
 
         network network(engine, topology);
@@ -1393,14 +1383,12 @@ TEST(activation_i32_fw_gpu, basic_yxfb_i32_funcs)
         EXPECT_EQ(outputs.begin()->first, "activation");
 
         auto output_memory = outputs.at("activation").get_memory();
-        auto output_layout = output_memory.get_layout();
-        auto output_ptr = output_memory.pointer<int32_t>();
-        auto input_ptr = input.pointer<int32_t>();
+        auto output_layout = output_memory->get_layout();
+        cldnn::mem_lock<int32_t> output_ptr(output_memory, get_test_stream());
+        cldnn::mem_lock<int32_t> input_ptr(input, get_test_stream());
 
-        for (size_t i = 0; i < output_layout.get_linear_size(); ++i)
-        {
-            switch (func)
-            {
+        for (size_t i = 0; i < output_layout.get_linear_size(); ++i) {
+            switch (func) {
             case activation_func::none:
                 EXPECT_EQ((int32_t)input_ptr[i], output_ptr[i]);
                 break;
@@ -1429,13 +1417,13 @@ TEST(activation_f32_fw_gpu, b_fs_yx_fsv16_prelu) {
     constexpr int x = 2;
     constexpr int y = 2;
 
-    auto eng = get_test_engine();
+    auto& eng = get_test_engine();
 
     auto in_lay = cldnn::layout(cldnn::data_types::f32, cldnn::format::bfyx, cldnn::tensor(b, f, x, y));
     auto params_lay = cldnn::layout(cldnn::data_types::f32, cldnn::format::bfyx, cldnn::tensor(1, f, 1, 1));
 
-    auto in_mem = cldnn::memory::allocate(eng, in_lay);
-    auto params_mem = cldnn::memory::allocate(eng, params_lay);
+    auto in_mem = eng.allocate_memory(in_lay);
+    auto params_mem = eng.allocate_memory(params_lay);
 
     auto in_data = generate_random_4d<float>(b, f, y, x, -1, 1);
     auto params_data = generate_random_1d<float>(f, -1, 1);
@@ -1464,7 +1452,7 @@ TEST(activation_f32_fw_gpu, b_fs_yx_fsv16_prelu) {
         }
     }
 
-    auto out_ptr = out_mem.pointer<float>();
+    cldnn::mem_lock<float> out_ptr(out_mem, get_test_stream());
     ASSERT_EQ(expected.size(), out_ptr.size());
 
     for (size_t i = 0; i < expected.size(); ++i) {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/add_reorders_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/add_reorders_gpu_test.cpp
index b9e26b77b94eb1..e8c18919b8428e 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/add_reorders_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/add_reorders_gpu_test.cpp
@@ -2,26 +2,19 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/reorder.hpp>
-#include <api/data.hpp>
-#include <api/activation.hpp>
-#include <api/mutable_data.hpp>
-#include <api/layout.hpp>
-#include <api/tile.hpp>
-#include <api/reshape.hpp>
-
-#include <api/concatenation.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/activation.hpp>
+#include <cldnn/primitives/mutable_data.hpp>
+#include <cldnn/primitives/tile.hpp>
+#include <cldnn/primitives/reshape.hpp>
+#include <cldnn/primitives/concatenation.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 /*
 These tests are inteded to check if additional reorders are being added  properly during
@@ -30,20 +23,20 @@ add_reorders optimization pass.
 
 //concatenation of incompatible convolutions
 TEST(add_reorders_gpu, two_convolutions_and_concatenation) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
     build_opt.set_option(build_option::optimize_data(false));
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::yxio,{ 1, 1, 1, 2 } });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::oiyx,{ 1, 1, 1, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::yxio,{ 1, 1, 1, 2 } });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::oiyx,{ 1, 1, 1, 2 } });
 
     set_values(input, { 1.1f, 1.2f, 1.3f, 1.4f });
     set_values(weights1, { 2.1f, 3.1f});
     set_values(weights2, { 1.1f, 0.1f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights1", weights1));
     topology.add(data("weights2", weights2));
 
@@ -63,23 +56,18 @@ TEST(add_reorders_gpu, two_convolutions_and_concatenation) {
     float expected_out[] = { 6.34f, 1.34f, 6.86f, 1.46f };
     float epsilon = 1e-3f;
 
-    for (auto& it : outputs)
-    {
-        auto output = it.second.get_memory().pointer<float>();
-        for (size_t cntr = 0; cntr < 2 * 2; cntr++)
-        {
+    for (auto& it : outputs) {
+        cldnn::mem_lock<float> output(it.second.get_memory(), get_test_stream());
+        for (size_t cntr = 0; cntr < 2 * 2; cntr++) {
             EXPECT_NEAR(expected_out[cntr], output[cntr], epsilon);
         }
     }
 }
 
 template<typename data_t>
-void tile_ref(const memory& input, memory& output, tile::tile_axis axis, int num_tiles)
-{
-    auto get_sizes = [](const tensor& size, tile::tile_axis axis) -> std::pair<int, int>
-    {
-        switch (axis)
-        {
+void tile_ref(const memory::ptr input, memory::ptr output, tile::tile_axis axis, int num_tiles) {
+    auto get_sizes = [](const tensor& size, tile::tile_axis axis) -> std::pair<int, int> {
+        switch (axis) {
         case tile::along_b: return std::make_pair(1, size.batch[0] * size.feature[0] * size.spatial[2] * size.spatial[1] * size.spatial[0]);
         case tile::along_f: return std::make_pair(size.batch[0], size.feature[0] * size.spatial[2] * size.spatial[1] * size.spatial[0]);
         case tile::along_z: return std::make_pair(size.batch[0] * size.feature[0], size.spatial[2] * size.spatial[1] * size.spatial[0]);
@@ -89,22 +77,19 @@ void tile_ref(const memory& input, memory& output, tile::tile_axis axis, int num
         }
     };
 
-    const pointer<data_t> src = input.pointer<data_t>();
-    pointer<data_t> dst = output.pointer<data_t>();
+    cldnn::mem_lock<data_t> src(input, get_test_stream());
+    cldnn::mem_lock<data_t> dst(output, get_test_stream());
 
     const data_t* psrc = src.data();
     data_t* pdst = dst.data();
 
-    auto sizes = get_sizes(input.get_layout().size, axis);
+    auto sizes = get_sizes(input->get_layout().size, axis);
     int outer_dim = sizes.first;
     int inner_dim = sizes.second;
 
-    for (int i = 0; i < outer_dim; i++)
-    {
-        for (int t = 0; t < num_tiles; t++)
-        {
-            for (int j = 0; j < inner_dim; j++)
-            {
+    for (int i = 0; i < outer_dim; i++) {
+        for (int t = 0; t < num_tiles; t++) {
+            for (int j = 0; j < inner_dim; j++) {
                 pdst[j] = psrc[j];
             }
             pdst += inner_dim;
@@ -114,13 +99,13 @@ void tile_ref(const memory& input, memory& output, tile::tile_axis axis, int num
 }
 
 TEST(add_reorders_gpu, basic_reshape_and_tile) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::byxf,{ 1, 2, 2, 1 } });
-    auto output_ref = memory::allocate(engine, { data_types::f32, format::byxf,{ 2, 1, 4, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::byxf,{ 1, 2, 2, 1 } });
+    auto output_ref = engine.allocate_memory({ data_types::f32, format::byxf,{ 2, 1, 4, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reshape("reshape", "input", tensor(2, 1, 2, 1)));
     topology.add(tile("tile", "reshape", tensor(2, 1, 2, 4)));
 
@@ -136,10 +121,10 @@ TEST(add_reorders_gpu, basic_reshape_and_tile) {
     auto outputs = network.execute();
 
     auto output = outputs.at("tile").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_ref_ptr = output_ref.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> output_ref_ptr(output_ref, get_test_stream());
 
-    for (unsigned int i = 0; i < output_ref.count(); ++i) {
+    for (unsigned int i = 0; i < output_ref->count(); ++i) {
         EXPECT_EQ(output_ptr[i], output_ref_ptr[i]);
     }
 }
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/arg_max_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/arg_max_gpu_test.cpp
index ac1dacfe4e325a..d7b4b630201304 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/arg_max_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/arg_max_gpu_test.cpp
@@ -2,32 +2,27 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/arg_max_min.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include <api/mutable_data.hpp>
-#include <api/data.hpp>
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/arg_max_min.hpp>
+#include <cldnn/primitives/mutable_data.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/input_layout.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 template <typename Tin, typename Tout>
-void generic_arg_max_test_xyf(int input_b, int input_f, int input_y, int input_x, arg_max_min::out_type mode, bool expect_throw = false)
-{
+void generic_arg_max_test_xyf(int input_b, int input_f, int input_y, int input_x, arg_max_min::out_type mode, bool expect_throw = false) {
     auto axis = arg_max_min::axis_name::xyf;
     auto sort_type = arg_max_min::sort_type::sort_by_values;
     auto test_input_fmt = format::bfyx;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     tensor input_tensor(input_b, input_f, input_x, input_y);
-    auto input = memory::allocate(engine, { type_to_data_type<Tin>::value, test_input_fmt, input_tensor });
+    auto input = engine.allocate_memory({ type_to_data_type<Tin>::value, test_input_fmt, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, mode, 1U, axis, sort_type, false, padding(), type_to_data_type<Tout>::value));
 
     int min_random = -2, max_random = 2;
@@ -51,12 +46,12 @@ void generic_arg_max_test_xyf(int input_b, int input_f, int input_y, int input_x
     int out_size = input_x * input_y * input_f;
 
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<Tout>();
+    cldnn::mem_lock<Tout> output_ptr(output, get_test_stream());
 
     Tout index;
     Tin value;
     for (auto i = 0; i < input_b; i++) {
-        index = get_value<Tout>(output_ptr, i);
+        index = get_value<Tout>(output_ptr.data(), i);
         EXPECT_GE(index, (Tout)0);
         EXPECT_LT(index, (Tout)out_size);
         value = input_rnd_vec[i*out_size + (int)index];
@@ -74,11 +69,11 @@ void generic_arg_max_test_xyf(int input_b, int input_f, int input_y, int input_x
 TEST(arg_max_gpu_batch_one, base) {
     //  Input  : 2x3x2x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 5, batch_num = 1, top_k = 8;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::max, top_k));
 
     std::vector<float> input_vec = {
@@ -100,11 +95,10 @@ TEST(arg_max_gpu_batch_one, base) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
 
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[batch_num * top_k];
-    for (uint32_t i = 0; i < batch_num * top_k; i++)
-    {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+    for (uint32_t i = 0; i < batch_num * top_k; i++) {
+        out_buffer[i] = get_value<float>(output_ptr.data(), i);
     }
      int size = x_size * y_size * feature_num;
      int index;
@@ -123,8 +117,7 @@ TEST(arg_max_gpu_batch_one, base) {
                  }
                  else
                      amount += same_values * (j - same_values + 1);
-             }
-             else if (input_vec[i*size + (int)out_buffer[i*top_k + j]] != input_vec[i*size + (int)out_buffer[i*top_k + j + 1]]) {
+             } else if (input_vec[i*size + (int)out_buffer[i*top_k + j]] != input_vec[i*size + (int)out_buffer[i*top_k + j + 1]]) {
                  if (same_values != j + 1) {
                      amount += same_values * (j - same_values + 1);
                      same_values = 1;
@@ -135,12 +128,10 @@ TEST(arg_max_gpu_batch_one, base) {
          }
          EXPECT_GE(out_buffer[i*top_k + top_k - 1], 0);
          EXPECT_LT(out_buffer[i*top_k + top_k - 1], size);
-         for (int j = 0; j < top_k; j++)
-         {
+         for (int j = 0; j < top_k; j++) {
              index = (int)out_buffer[i*top_k + j];
              value = input_vec[i*size + index];
-             for (int k = 0; k < size; k++)
-             {
+             for (int k = 0; k < size; k++) {
                  if (input_vec[i*size + k] > value)
                      count++;
              }
@@ -152,11 +143,11 @@ TEST(arg_max_gpu_batch_one, base) {
 TEST(arg_max_gpu_top_k, base) {
 	//  Input  : 2x3x2x2
 	static const int32_t x_size = 2, y_size = 2, feature_num = 5, batch_num = 2;
-	const auto& engine = get_test_engine();
+	auto& engine = get_test_engine();
 	const int top_k = 8;
-	auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
+	auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
 	topology topology;
-	topology.add(input_layout("input", input.get_layout()));
+	topology.add(input_layout("input", input->get_layout()));
 	topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::max, top_k));
 
 	std::vector<float> input_vec = {
@@ -184,11 +175,10 @@ TEST(arg_max_gpu_top_k, base) {
 	EXPECT_EQ(outputs.begin()->first, "arg_max");
 
 	auto output = outputs.at("arg_max").get_memory();
-	auto output_ptr = output.pointer<float>();
+	cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 	float out_buffer[batch_num * top_k];
-	for (uint32_t i = 0; i < batch_num * top_k; i++)
-	{
-		out_buffer[i] = get_value<float>(output_ptr, i);
+	for (uint32_t i = 0; i < batch_num * top_k; i++) {
+		out_buffer[i] = get_value<float>(output_ptr.data(), i);
 	}
 	int size = x_size * y_size * feature_num;
 	int index;
@@ -207,24 +197,21 @@ TEST(arg_max_gpu_top_k, base) {
 				}
 				else
 					amount += same_values * (j - same_values + 1);
-			}
-			else if (input_vec[i*size + (int)(int)out_buffer[i*top_k + j]] != input_vec[i*size + (int)(int)out_buffer[i*top_k + j + 1]]) {
+			} else if (input_vec[i*size + (int)(int)out_buffer[i*top_k + j]] != input_vec[i*size + (int)(int)out_buffer[i*top_k + j + 1]]) {
 				if (same_values != j+1) {
 					amount += same_values * (j - same_values + 1);
 					same_values = 1;
 				}
-			}
-			else
+			} else {
 				same_values++;
+            }
 		}
 		EXPECT_GE(out_buffer[i*top_k + top_k - 1], 0);
 		EXPECT_LT(out_buffer[i*top_k + top_k - 1], size);
-		for (int j = 0; j < top_k; j++)
-		{
+		for (int j = 0; j < top_k; j++) {
 			index = (int)out_buffer[i*top_k + j];
 			value = input_vec[i*size + index];
-			for (int k = 0; k < size; k++)
-			{
+			for (int k = 0; k < size; k++) {
 				if (input_vec[i*size + k] > value)
 					count++;
 			}
@@ -236,11 +223,11 @@ TEST(arg_max_gpu_top_k, base) {
 TEST(arg_max_gpu_min_top_k, base) {
 	//  Input  : 2x3x2x2
 	static const int32_t x_size = 2, y_size = 2, feature_num = 4, batch_num = 2;
-	const auto& engine = get_test_engine();
+	auto& engine = get_test_engine();
 	const int top_k = 3;
-	auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
+	auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
 	topology topology;
-	topology.add(input_layout("input", input.get_layout()));
+	topology.add(input_layout("input", input->get_layout()));
 	topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::min, top_k));
 
 	std::vector<float> input_vec = {
@@ -266,11 +253,10 @@ TEST(arg_max_gpu_min_top_k, base) {
 	EXPECT_EQ(outputs.begin()->first, "arg_max");
 
 	auto output = outputs.at("arg_max").get_memory();
-	auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 	float out_buffer[batch_num * top_k];
-	for (uint32_t i = 0; i < batch_num * top_k; i++)
-	{
-		out_buffer[i] = get_value<float>(output_ptr, i);
+	for (uint32_t i = 0; i < batch_num * top_k; i++) {
+		out_buffer[i] = get_value<float>(output_ptr.data(), i);
 	}
 	int size = x_size * y_size * feature_num;
 	int index;
@@ -289,24 +275,21 @@ TEST(arg_max_gpu_min_top_k, base) {
 				}
 				else
 					amount += same_values * (j - same_values + 1);
-			}
-			else if (input_vec[i*size + (int)out_buffer[i*top_k + j]] != input_vec[i*size + (int)out_buffer[i*top_k + j + 1]]) {
+			} else if (input_vec[i*size + (int)out_buffer[i*top_k + j]] != input_vec[i*size + (int)out_buffer[i*top_k + j + 1]]) {
 				if (same_values != j + 1) {
 					amount += same_values * (j - same_values + 1);
 					same_values = 1;
 				}
-			}
-			else
+			} else {
 				same_values++;
+            }
 		}
 		EXPECT_GE(out_buffer[i*top_k + top_k - 1], 0);
 		EXPECT_LT(out_buffer[i*top_k + top_k - 1], size);
-		for (int j = 0; j < top_k; j++)
-		{
+		for (int j = 0; j < top_k; j++) {
 			index = (int)out_buffer[i*top_k + j];
 			value = input_vec[i*size + index];
-			for (int k = 0; k < size; k++)
-			{
+			for (int k = 0; k < size; k++) {
 				if (input_vec[i*size + k] < value)
 					count++;
 			}
@@ -318,11 +301,11 @@ TEST(arg_max_gpu_min_top_k, base) {
 TEST(arg_max_gpu_min_axis_batch, base) {
     //  Input  : 2x3x2x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 4, batch_num = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int top_k = 2;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::min, top_k, arg_max_min::batch));
 
     std::vector<float> input_vec = {
@@ -348,14 +331,12 @@ TEST(arg_max_gpu_min_axis_batch, base) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
     const int out_size = y_size * feature_num * x_size * top_k;
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[out_size];
-    for (uint32_t i = 0; i < out_size; i++)
-    {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+    for (uint32_t i = 0; i < out_size; i++) {
+        out_buffer[i] = get_value<float>(output_ptr.data(), i);
     }
-    for (int i = 0; i < out_size; i++)
-    {
+    for (int i = 0; i < out_size; i++) {
         EXPECT_EQ(out_buffer[i], i < (out_size / 2) ? 0 : 1);
     }
 }
@@ -407,11 +388,11 @@ TEST(arg_max_gpu_min, i64) {
 TEST(arg_max_gpu_min_axis_batch, i32) {
     //  Input  : 2x3x2x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 4, batch_num = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int top_k = 2;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::min, top_k, arg_max_min::batch, arg_max_min::sort_by_values, false, padding(), data_types::i32));
 
     std::vector<float> input_vec = {
@@ -437,14 +418,12 @@ TEST(arg_max_gpu_min_axis_batch, i32) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
     const int out_size = y_size * feature_num * x_size * top_k;
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<int32_t>();
+    cldnn::mem_lock<int32_t> output_ptr(output, get_test_stream());
     int32_t out_buffer[out_size];
-    for (uint32_t i = 0; i < out_size; i++)
-    {
-        out_buffer[i] = get_value<int32_t>(output_ptr, i);
+    for (uint32_t i = 0; i < out_size; i++) {
+        out_buffer[i] = get_value<int32_t>(output_ptr.data(), i);
     }
-    for (int i = 0; i < out_size; i++)
-    {
+    for (int i = 0; i < out_size; i++) {
         EXPECT_EQ(out_buffer[i], i < (out_size / 2) ? 0 : 1);
     }
 }
@@ -452,11 +431,11 @@ TEST(arg_max_gpu_min_axis_batch, i32) {
 TEST(arg_max_gpu_min_axis_batch_bfzyx, i32) {
     //  Input  : 2x3x2x2
     static const int32_t x_size = 2, y_size = 2, z_size = 1, feature_num = 4, batch_num = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int top_k = 2;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ batch_num, feature_num, x_size , y_size, z_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ batch_num, feature_num, x_size , y_size, z_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::min, top_k, arg_max_min::batch, arg_max_min::sort_by_values, false, padding(), data_types::i32));
 
     std::vector<float> input_vec = {
@@ -482,25 +461,23 @@ TEST(arg_max_gpu_min_axis_batch_bfzyx, i32) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
     const int out_size = y_size * feature_num * x_size * top_k;
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<int32_t>();
+    cldnn::mem_lock<int32_t> output_ptr(output, get_test_stream());
     int32_t out_buffer[out_size];
-    for (uint32_t i = 0; i < out_size; i++)
-    {
-        out_buffer[i] = get_value<int32_t>(output_ptr, i);
+    for (uint32_t i = 0; i < out_size; i++) {
+        out_buffer[i] = get_value<int32_t>(output_ptr.data(), i);
     }
-    for (int i = 0; i < out_size; i++)
-    {
+    for (int i = 0; i < out_size; i++) {
         EXPECT_EQ(out_buffer[i], i < (out_size / 2) ? 0 : 1);
     }
 }
 
 TEST(arg_max_gpu_min_axis_y_yxfb, f32) {
     static const int32_t x_size = 2, y_size = 2, feature_num = 4, batch_num = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int top_k = 1;
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::max, top_k, arg_max_min::y, arg_max_min::sort_by_values, false, padding(), data_types::f32));
 
     std::vector<float> input_vec = {
@@ -548,25 +525,23 @@ TEST(arg_max_gpu_min_axis_y_yxfb, f32) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
     const int out_size = y_size * feature_num * x_size * top_k;
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[out_size];
-    for (uint32_t i = 0; i < out_size; i++)
-    {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+    for (uint32_t i = 0; i < out_size; i++) {
+        out_buffer[i] = get_value<float>(output_ptr.data(), i);
     }
-    for (int i = 0; i < out_size; i++)
-    {
+    for (int i = 0; i < out_size; i++) {
         EXPECT_EQ(out_buffer[i], ref_vec[i]);
     }
 }
 
 TEST(arg_max_gpu_min_axis_batch_yxfb, f32) {
     static const int32_t x_size = 2, y_size = 2, feature_num = 4, batch_num = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int top_k = 1;
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::max, top_k, arg_max_min::batch, arg_max_min::sort_by_values, false, padding(), data_types::f32));
 
     std::vector<float> input_vec = {
@@ -614,25 +589,23 @@ TEST(arg_max_gpu_min_axis_batch_yxfb, f32) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
     const int out_size = y_size * feature_num * x_size * top_k;
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[out_size];
-    for (uint32_t i = 0; i < out_size; i++)
-    {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+    for (uint32_t i = 0; i < out_size; i++) {
+        out_buffer[i] = get_value<float>(output_ptr.data(), i);
     }
-    for (int i = 0; i < out_size; i++)
-    {
+    for (int i = 0; i < out_size; i++) {
         EXPECT_EQ(out_buffer[i], ref_vec[i]);
     }
 }
 
 TEST(arg_max_gpu_min_axis_y_yxfb_topk_2, f32) {
     static const int32_t x_size = 2, y_size = 2, feature_num = 4, batch_num = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int top_k = 2;
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::max, top_k, arg_max_min::y, arg_max_min::sort_by_values, false, padding(), data_types::f32));
 
     std::vector<float> input_vec = {
@@ -690,27 +663,25 @@ TEST(arg_max_gpu_min_axis_y_yxfb_topk_2, f32) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
     const int out_size = y_size * feature_num * x_size * top_k;
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[out_size];
-    for (uint32_t i = 0; i < out_size; i++)
-    {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+    for (uint32_t i = 0; i < out_size; i++) {
+        out_buffer[i] = get_value<float>(output_ptr.data(), i);
     }
-    for (int i = 0; i < out_size; i++)
-    {
+    for (int i = 0; i < out_size; i++) {
         EXPECT_EQ(out_buffer[i], ref_vec[i]);
     }
 }
 
 TEST(top_k_layer_tests, second_output) {
     static const int32_t x_size = 2, y_size = 2, feature_num = 4, batch_num = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int top_k = 2;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
-    auto top_k_input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1 , 1 } });
-    auto second_output = memory::allocate(engine, { data_types::f32, format::bfyx, { top_k, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
+    auto top_k_input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1 , 1 } });
+    auto second_output = engine.allocate_memory({ data_types::f32, format::bfyx, { top_k, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(cldnn::data("const", top_k_input));
     topology.add(mutable_data("second_output", second_output));
     topology.add(arg_max_min("arg_max", { "input", "const", "second_output" }, arg_max_min::min, top_k, arg_max_min::batch));
@@ -738,17 +709,16 @@ TEST(top_k_layer_tests, second_output) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
     const int out_size = y_size * feature_num * x_size * top_k;
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto second_output_ptr = second_output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> second_output_ptr(second_output, get_test_stream());
+
     float out_buffer[out_size];
     float second_out_buffer[out_size];
-    for (uint32_t i = 0; i < out_size; i++)
-    {
-        out_buffer[i] = get_value<float>(output_ptr, i);
-        second_out_buffer[i] = get_value<float>(second_output_ptr, i);
+    for (uint32_t i = 0; i < out_size; i++) {
+        out_buffer[i] = get_value<float>(output_ptr.data(), i);
+        second_out_buffer[i] = get_value<float>(second_output_ptr.data(), i);
     }
-    for (int i = 0; i < out_size; i++)
-    {
+    for (int i = 0; i < out_size; i++) {
         EXPECT_EQ(out_buffer[i], i < (out_size / 2) ? 0 : 1);
         EXPECT_EQ(second_out_buffer[i], input_vec[i]);
     }
@@ -756,13 +726,13 @@ TEST(top_k_layer_tests, second_output) {
 
 TEST(top_k_layer_tests, second_output2) {
     static const int32_t x_size = 2, y_size = 2, feature_num = 4, batch_num = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int top_k = 1;
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
-    auto top_k_input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1 , 1 } });
-    auto second_output = memory::allocate(engine, { data_types::f32, format::yxfb, { top_k, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
+    auto top_k_input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1 , 1 } });
+    auto second_output = engine.allocate_memory({ data_types::f32, format::yxfb, { top_k, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(cldnn::data("const", top_k_input));
     topology.add(mutable_data("second_output", second_output));
     topology.add(arg_max_min("arg_max", { "input", "const", "second_output" }, arg_max_min::max, top_k, arg_max_min::batch, arg_max_min::sort_by_values, false, padding(), data_types::f32));
@@ -834,17 +804,15 @@ TEST(top_k_layer_tests, second_output2) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
     const int out_size = y_size * feature_num * x_size * top_k;
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto second_output_ptr = second_output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> second_output_ptr(second_output, get_test_stream());
     float out_buffer[out_size];
     float second_out_buffer[out_size];
-    for (uint32_t i = 0; i < out_size; i++)
-    {
-        out_buffer[i] = get_value<float>(output_ptr, i);
-        second_out_buffer[i] = get_value<float>(second_output_ptr, i);
+    for (uint32_t i = 0; i < out_size; i++) {
+        out_buffer[i] = get_value<float>(output_ptr.data(), i);
+        second_out_buffer[i] = get_value<float>(second_output_ptr.data(), i);
     }
-    for (int i = 0; i < out_size; i++)
-    {
+    for (int i = 0; i < out_size; i++) {
         EXPECT_EQ(out_buffer[i], ref_vec[i]);
         EXPECT_EQ(second_out_buffer[i], second_ref_vec[i]);
     }
@@ -852,11 +820,11 @@ TEST(top_k_layer_tests, second_output2) {
 
 TEST(arg_max_gpu_min_axis_y_yxfb_topk_2, sort_by_values) {
     static const int32_t x_size = 2, y_size = 2, feature_num = 4, batch_num = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int top_k = 2;
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::max, top_k, arg_max_min::y, arg_max_min::sort_by_values, false, padding(), data_types::f32));
 
     std::vector<float> input_vec = {
@@ -914,25 +882,23 @@ TEST(arg_max_gpu_min_axis_y_yxfb_topk_2, sort_by_values) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
     const int out_size = y_size * feature_num * x_size * top_k;
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[out_size];
-    for (uint32_t i = 0; i < out_size; i++)
-    {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+    for (uint32_t i = 0; i < out_size; i++) {
+        out_buffer[i] = get_value<float>(output_ptr.data(), i);
     }
-    for (int i = 0; i < out_size; i++)
-    {
+    for (int i = 0; i < out_size; i++) {
         EXPECT_EQ(out_buffer[i], ref_vec[i]);
     }
 }
 
 TEST(arg_max_gpu_min_axis_y_yxfb_topk_2, sort_by_indices) {
     static const int32_t x_size = 2, y_size = 2, feature_num = 4, batch_num = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int top_k = 2;
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::max, top_k, arg_max_min::y, arg_max_min::sort_by_indices, false, padding(), data_types::f32));
 
     std::vector<float> input_vec = {
@@ -990,14 +956,12 @@ TEST(arg_max_gpu_min_axis_y_yxfb_topk_2, sort_by_indices) {
     EXPECT_EQ(outputs.begin()->first, "arg_max");
     const int out_size = y_size * feature_num * x_size * top_k;
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[out_size];
-    for (uint32_t i = 0; i < out_size; i++)
-    {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+    for (uint32_t i = 0; i < out_size; i++) {
+        out_buffer[i] = get_value<float>(output_ptr.data(), i);
     }
-    for (int i = 0; i < out_size; i++)
-    {
+    for (int i = 0; i < out_size; i++) {
         EXPECT_EQ(out_buffer[i], ref_vec[i]);
     }
 }
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/average_unpooling_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/average_unpooling_gpu_test.cpp
index 01eafbde391ca0..7674100106094d 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/average_unpooling_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/average_unpooling_gpu_test.cpp
@@ -2,23 +2,17 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/average_unpooling.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/reorder.hpp>
-#include <api/data.hpp>
-#include <api/mutable_data.hpp>
-#include <api/pooling.hpp>
-#include "test_utils/float16.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/average_unpooling.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/mutable_data.hpp>
+#include <cldnn/primitives/pooling.hpp>
+#include <cldnn/primitives/input_layout.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 TEST(average_unpooling_gpu, basic_in2x2x2x1) {
     //  Input  : 2x2x2x1
@@ -40,9 +34,9 @@ TEST(average_unpooling_gpu, basic_in2x2x2x1) {
     //  f1: b0:  1.5     2.5    1       b1:   1.75   2.9375   1.1875
     //  f1: b0:  1.5     2.5    1       b1:   1.75   2.9375   1.1875
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
 
     set_values(input, {
         2.5f, -4.5f,
@@ -52,7 +46,7 @@ TEST(average_unpooling_gpu, basic_in2x2x2x1) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(average_unpooling("average_unpooling", "input", { 2, 2, 3, 2 }, { 1, 1, 2, 2 }, { 1, 1, 1, 1 }));
 
     network network(engine, topology);
@@ -62,8 +56,8 @@ TEST(average_unpooling_gpu, basic_in2x2x2x1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("average_unpooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 2);
@@ -107,9 +101,9 @@ TEST(average_unpooling_gpu, basic_in2x2x3x2_with_average_pooling_unpooling) {
     //  f1: b0:  1.5     1.5    0.5  b1:   1.75   1.75   1
     //  f1: b0:  1.5     1.5    0.5  b1:   1.75   1.75   1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
 
     set_values(input, {
         1.f, 2.f, -10.f,
@@ -123,9 +117,9 @@ TEST(average_unpooling_gpu, basic_in2x2x3x2_with_average_pooling_unpooling) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(pooling("pooling", "input", pooling_mode::average_no_padding, { 1, 1, 2, 2 }, { 1, 1, 2, 2 }));
-    topology.add(average_unpooling("average_unpooling", "pooling", input.get_layout().size, { 1, 1, 2, 2 }, { 1, 1, 2, 2 }));
+    topology.add(average_unpooling("average_unpooling", "pooling", input->get_layout().size, { 1, 1, 2, 2 }, { 1, 1, 2, 2 }));
 
     network network(engine, topology);
 
@@ -134,8 +128,8 @@ TEST(average_unpooling_gpu, basic_in2x2x3x2_with_average_pooling_unpooling) {
     auto outputs = network.execute();
 
     auto output = outputs.at("average_unpooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 2);
@@ -179,9 +173,9 @@ TEST(average_unpooling_gpu, basic_in2x2x2x1_output_padding) {
     //  f0: b0:  0.625   -0.5  -1.125   b1:   0  -1.6875  -1.6875
     //  f1: b0:  1.5     2.5    1       b1:   1.75   2.9375   1.1875
     //  f1: b0:  1.5     2.5    1       b1:   1.75   2.9375   1.1875
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         2.5f, -4.5f,
@@ -191,7 +185,7 @@ TEST(average_unpooling_gpu, basic_in2x2x2x1_output_padding) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(average_unpooling("average_unpooling", "input", { 2, 2, 3, 2 }, { 1, 1, 2, 2 }, { 1, 1, 1, 1 }, padding({ 0, 0, 1, 1 }, 0)));
 
     network network(engine, topology);
@@ -201,8 +195,8 @@ TEST(average_unpooling_gpu, basic_in2x2x2x1_output_padding) {
     auto outputs = network.execute();
 
     auto output = outputs.at("average_unpooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 2);
@@ -260,9 +254,9 @@ TEST(average_unpooling_gpu, basic_in2x2x2x1_fp16) {
     //  f1: b0:  1.5     2.5    1       b1:   1.75   2.9375   1.1875
     //  f1: b0:  1.5     2.5    1       b1:   1.75   2.9375   1.1875
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         FLOAT16(2.5f), FLOAT16(-4.5f),
@@ -272,7 +266,7 @@ TEST(average_unpooling_gpu, basic_in2x2x2x1_fp16) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(average_unpooling("average_unpooling", "input", { 2, 2, 3, 2 }, { 1, 1, 2, 2 }, { 1, 1, 1, 1 }));
 
     network network(engine, topology);
@@ -282,8 +276,8 @@ TEST(average_unpooling_gpu, basic_in2x2x2x1_fp16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("average_unpooling").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
-    auto output_layout = output.get_layout();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 2);
@@ -304,4 +298,4 @@ TEST(average_unpooling_gpu, basic_in2x2x2x1_fp16) {
     for (size_t i = 0; i < expected_output_vec.size(); ++i) {
         EXPECT_EQ(expected_output_vec[i], float16_to_float32(output_ptr[i]));
     }
-}
\ No newline at end of file
+}
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/barriers_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/barriers_test.cpp
index 2b15da70ed271c..2b72881c6180d3 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/barriers_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/barriers_test.cpp
@@ -2,28 +2,21 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
+#include "test_utils.h"
 
-#include <gtest/gtest.h>
-#include <api/engine.hpp>
-#include <api/memory.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/input_layout.hpp>
-#include <api/concatenation.hpp>
-#include <api/reorder.hpp>
-
-#include "test_utils/test_utils.h"
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/concatenation.hpp>
+#include <cldnn/primitives/reorder.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 TEST(DISABLED_oooq_test, simple)
 {
-    engine_configuration cfg{ false, false, false, std::string(), std::string(), true };
-    engine eng{ cfg };
+    engine_configuration cfg{ false, queue_types::out_of_order };
+    auto eng = engine::create(engine_types::ocl, runtime_types::ocl, cfg);
 
-    memory input_mem = memory::allocate(eng, layout{ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input_mem = eng->allocate_memory(layout{ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
     set_values(input_mem, { 50 });
 
     /*                 ---- r1 ---- r3 ----            -- r7 --
@@ -34,16 +27,16 @@ TEST(DISABLED_oooq_test, simple)
     */
 
     topology tpl;
-    tpl.add(input_layout("in", input_mem.get_layout()));
-    tpl.add(reorder("r0", "in", input_mem.get_layout(), std::vector<float>{ 0 }));
-    tpl.add(reorder("r1", "r0", input_mem.get_layout(), std::vector<float>{ 1 }));
-    tpl.add(reorder("r2", "r0", input_mem.get_layout(), std::vector<float>{ 2 }));
-    tpl.add(reorder("r3", "r1", input_mem.get_layout(), std::vector<float>{ 3 }));
-    tpl.add(reorder("r4", "r2", input_mem.get_layout(), std::vector<float>{ 4 }));
-    tpl.add(reorder("r5", "r4", input_mem.get_layout(), std::vector<float>{ 5 }));
+    tpl.add(input_layout("in", input_mem->get_layout()));
+    tpl.add(reorder("r0", "in", input_mem->get_layout(), std::vector<float>{ 0 }));
+    tpl.add(reorder("r1", "r0", input_mem->get_layout(), std::vector<float>{ 1 }));
+    tpl.add(reorder("r2", "r0", input_mem->get_layout(), std::vector<float>{ 2 }));
+    tpl.add(reorder("r3", "r1", input_mem->get_layout(), std::vector<float>{ 3 }));
+    tpl.add(reorder("r4", "r2", input_mem->get_layout(), std::vector<float>{ 4 }));
+    tpl.add(reorder("r5", "r4", input_mem->get_layout(), std::vector<float>{ 5 }));
 
     tpl.add(concatenation("c6", { "r3", "r5" }, concatenation::along_x));
-    layout concat_lay = input_mem.get_layout();
+    layout concat_lay = input_mem->get_layout();
     concat_lay.size.spatial[0] *= 2;
 
     tpl.add(reorder("r7", "c6", concat_lay, std::vector<float>{ 7 }));
@@ -52,13 +45,13 @@ TEST(DISABLED_oooq_test, simple)
     concat_lay.size.spatial[1] *= 2;
 
     build_options options;
-    network net{ eng, tpl, options };
+    network net{ *eng, tpl, options };
 
     net.set_input_data("in", input_mem);
     auto output = net.execute().at("c9").get_memory();
 
-    EXPECT_TRUE(output.get_layout().size.spatial[0] == 2);
-    EXPECT_TRUE(output.get_layout().size.spatial[1] == 2);
-    EXPECT_TRUE(output.get_layout().size.feature[0] == 1);
-    EXPECT_TRUE(output.get_layout().size.batch[0] == 1);
-}
\ No newline at end of file
+    EXPECT_TRUE(output->get_layout().size.spatial[0] == 2);
+    EXPECT_TRUE(output->get_layout().size.spatial[1] == 2);
+    EXPECT_TRUE(output->get_layout().size.feature[0] == 1);
+    EXPECT_TRUE(output->get_layout().size.batch[0] == 1);
+}
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/batch_to_space_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/batch_to_space_gpu_test.cpp
index 9299820f8ce448..e3d33ad250765f 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/batch_to_space_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/batch_to_space_gpu_test.cpp
@@ -2,18 +2,13 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
+#include "test_utils.h"
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/batch_to_space.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/data.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/batch_to_space.hpp>
+#include <cldnn/primitives/data.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -26,9 +21,9 @@ TEST(batch_to_space_fp16_gpu, i8111_bs1222_cb0000_ce0000) {
     //  Output :      1x2x2x2
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(8), feature(1), spatial(1, 1)};
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, input_shape });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f),
@@ -38,7 +33,7 @@ TEST(batch_to_space_fp16_gpu, i8111_bs1222_cb0000_ce0000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(batch_to_space("batch_to_space", "Input", tensor(format::bfyx, {1,2,2,2}, 1),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
@@ -50,7 +45,7 @@ TEST(batch_to_space_fp16_gpu, i8111_bs1222_cb0000_ce0000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("batch_to_space").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f
@@ -71,9 +66,9 @@ TEST(batch_to_space_fp16_gpu, i4321_bs1212_cb0000_ce0000) {
     //  Output :      1x6x2x2
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(4), feature(3), spatial(1, 2)};
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, input_shape });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f),
@@ -85,7 +80,7 @@ TEST(batch_to_space_fp16_gpu, i4321_bs1212_cb0000_ce0000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(batch_to_space("batch_to_space", "Input", tensor(format::bfyx, {1,2,1,2}, 1),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
@@ -97,7 +92,7 @@ TEST(batch_to_space_fp16_gpu, i4321_bs1212_cb0000_ce0000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("batch_to_space").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 6.f, 1.f, 7.f, 12.f, 18.f,
@@ -121,9 +116,9 @@ TEST(batch_to_space_fp16_gpu, i4321_bs1212_cb0010_ce0101) {
     //  Output :      1x5x1x1
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(4), feature(3), spatial(1, 2)};
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, input_shape });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f),
@@ -135,7 +130,7 @@ TEST(batch_to_space_fp16_gpu, i4321_bs1212_cb0010_ce0101) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(batch_to_space("batch_to_space", "Input", tensor(format::bfyx, {1,2,1,2}, 1),
                                                            tensor(format::bfyx, {0,0,1,0}, 0),
                                                            tensor(format::bfyx, {0,1,0,1}, 0),
@@ -147,7 +142,7 @@ TEST(batch_to_space_fp16_gpu, i4321_bs1212_cb0010_ce0101) {
     auto outputs = network.execute();
 
     auto output = outputs.at("batch_to_space").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f, 13.f, 3.f, 15.f, 5.f
@@ -168,9 +163,9 @@ TEST(batch_to_space_fp16_gpu, i62121_bs12311_cb02000_ce00110) {
     //  Output :      1x2x2x1x1
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(6), feature(2), spatial(1, 2, 1)};
-    auto input = memory::allocate(engine, { data_types::f16, format::bfzyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfzyx, input_shape });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f),
@@ -182,7 +177,7 @@ TEST(batch_to_space_fp16_gpu, i62121_bs12311_cb02000_ce00110) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(batch_to_space("batch_to_space", "Input", tensor(format::bfzyx, {1,2,3,1,1}, 1),
                                                            tensor(format::bfzyx, {0,2,0,0,0}, 0),
                                                            tensor(format::bfzyx, {0,0,1,1,0}, 0),
@@ -194,7 +189,7 @@ TEST(batch_to_space_fp16_gpu, i62121_bs12311_cb02000_ce00110) {
     auto outputs = network.execute();
 
     auto output = outputs.at("batch_to_space").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         2.f, 6.f, 14.f, 18.f
@@ -215,9 +210,9 @@ TEST(batch_to_space_fp16_gpu, i1212112_bs112321_cb02000_ce00110) {
     //  Output :      1x1x3x1x2x2
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(12), feature(1), spatial(2, 1, 1, 2)};
-    auto input = memory::allocate(engine, { data_types::f16, format::bfwzyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfwzyx, input_shape });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f),
@@ -231,7 +226,7 @@ TEST(batch_to_space_fp16_gpu, i1212112_bs112321_cb02000_ce00110) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(batch_to_space("batch_to_space", "Input", tensor(format::bfwzyx, {1,1,2,3,2,1}, 1),
                                                            tensor(format::bfwzyx, {0,0,1,0,0,0}, 0),
                                                            tensor(format::bfwzyx, {0,0,0,2,0,0}, 0),
@@ -243,7 +238,7 @@ TEST(batch_to_space_fp16_gpu, i1212112_bs112321_cb02000_ce00110) {
     auto outputs = network.execute();
 
     auto output = outputs.at("batch_to_space").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         24.f, 25.f, 28.f, 29.f,
@@ -266,9 +261,9 @@ TEST(batch_to_space_fp16_gpu, i21611_bs1112_cb0000_ce0000_b_fs_yx_fsv16) {
     //  Output :      1x16x1x2
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(2), feature(16), spatial(1, 1)};
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, input_shape });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f), FLOAT16(4.0f), FLOAT16(5.0f), FLOAT16(6.0f), FLOAT16(7.0f),
@@ -278,7 +273,7 @@ TEST(batch_to_space_fp16_gpu, i21611_bs1112_cb0000_ce0000_b_fs_yx_fsv16) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(reorder("input_fsv", "Input", format::b_fs_yx_fsv16, data_types::f16));
     topology.add(batch_to_space("batch_to_space", "input_fsv", tensor(format::bfyx, {1,1,1,2}, 1),
                                                                tensor(format::bfyx, {0,0,0,0}, 0),
@@ -293,7 +288,7 @@ TEST(batch_to_space_fp16_gpu, i21611_bs1112_cb0000_ce0000_b_fs_yx_fsv16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("bts_to_bfyx").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 16.f, 1.f, 17.f, 2.f, 18.f, 3.f, 19.f,
@@ -317,9 +312,9 @@ TEST(batch_to_space_fp16_gpu, i2812_bs1112_cb0000_ce0000_b_fs_yx_fsv16) {
     //  Output :      1x6x1x4
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(2), feature(8), spatial(2, 1)};
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, input_shape });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f), FLOAT16(4.0f), FLOAT16(5.0f), FLOAT16(6.0f), FLOAT16(7.0f),
@@ -329,7 +324,7 @@ TEST(batch_to_space_fp16_gpu, i2812_bs1112_cb0000_ce0000_b_fs_yx_fsv16) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(reorder("input_fsv", "Input", format::b_fs_yx_fsv16, data_types::f16));
     topology.add(batch_to_space("batch_to_space", "input_fsv", tensor(format::bfyx, {1,1,1,2}, 1),
                                                                tensor(format::bfyx, {0,2,0,0}, 0),
@@ -344,7 +339,7 @@ TEST(batch_to_space_fp16_gpu, i2812_bs1112_cb0000_ce0000_b_fs_yx_fsv16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("bts_to_bfyx").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         4.f, 20.f, 5.f, 21.f, 6.f, 22.f, 7.f, 23.f,
@@ -367,9 +362,9 @@ TEST(batch_to_space_fp32_gpu, i8111_bs1222_cb0000_ce0000) {
     //  Output :      1x2x2x2
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(8), feature(1), spatial(1, 1)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, input_shape });
 
     set_values(input, {
         0.0f, 1.0f, 2.0f, 3.0f,
@@ -377,7 +372,7 @@ TEST(batch_to_space_fp32_gpu, i8111_bs1222_cb0000_ce0000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(batch_to_space("batch_to_space", "Input", tensor(format::bfyx, {1,2,2,2}, 1),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
@@ -389,7 +384,7 @@ TEST(batch_to_space_fp32_gpu, i8111_bs1222_cb0000_ce0000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("batch_to_space").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f
@@ -410,9 +405,9 @@ TEST(batch_to_space_fp32_gpu, i4321_bs1212_cb0000_ce0000) {
     //  Output :      1x6x2x2
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(4), feature(3), spatial(1, 2)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, input_shape });
 
     set_values(input, {
         0.0f, 1.0f, 2.0f, 3.0f,
@@ -424,7 +419,7 @@ TEST(batch_to_space_fp32_gpu, i4321_bs1212_cb0000_ce0000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(batch_to_space("batch_to_space", "Input", tensor(format::bfyx, {1,2,1,2}, 1),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
@@ -436,7 +431,7 @@ TEST(batch_to_space_fp32_gpu, i4321_bs1212_cb0000_ce0000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("batch_to_space").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 6.f, 1.f, 7.f, 12.f, 18.f,
@@ -460,9 +455,9 @@ TEST(batch_to_space_fp32_gpu, i4321_bs1212_cb0010_ce0101) {
     //  Output :      1x5x1x1
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(4), feature(3), spatial(1, 2)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, input_shape });
 
     set_values(input, {
         0.0f, 1.0f, 2.0f, 3.0f,
@@ -474,7 +469,7 @@ TEST(batch_to_space_fp32_gpu, i4321_bs1212_cb0010_ce0101) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(batch_to_space("batch_to_space", "Input", tensor(format::bfyx, {1,2,1,2}, 1),
                                                            tensor(format::bfyx, {0,0,1,0}, 0),
                                                            tensor(format::bfyx, {0,1,0,1}, 0),
@@ -486,7 +481,7 @@ TEST(batch_to_space_fp32_gpu, i4321_bs1212_cb0010_ce0101) {
     auto outputs = network.execute();
 
     auto output = outputs.at("batch_to_space").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f, 13.f, 3.f, 15.f, 5.f
@@ -507,9 +502,9 @@ TEST(batch_to_space_fp32_gpu, i62121_bs12311_cb02000_ce00110) {
     //  Output :      1x2x2x1x1
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(6), feature(2), spatial(1, 2, 1)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, input_shape });
 
     set_values(input, {
         0.0f, 1.0f, 2.0f, 3.0f,
@@ -521,7 +516,7 @@ TEST(batch_to_space_fp32_gpu, i62121_bs12311_cb02000_ce00110) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(batch_to_space("batch_to_space", "Input", tensor(format::bfzyx, {1,2,3,1,1}, 1),
                                                            tensor(format::bfzyx, {0,2,0,0,0}, 0),
                                                            tensor(format::bfzyx, {0,0,1,1,0}, 0),
@@ -533,7 +528,7 @@ TEST(batch_to_space_fp32_gpu, i62121_bs12311_cb02000_ce00110) {
     auto outputs = network.execute();
 
     auto output = outputs.at("batch_to_space").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         2.f, 6.f, 14.f, 18.f
@@ -554,9 +549,9 @@ TEST(batch_to_space_fp32_gpu, i1212112_bs112321_cb02000_ce00110) {
     //  Output :      1x1x3x1x2x2
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(12), feature(1), spatial(2, 1, 1, 2)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfwzyx, input_shape });
 
     set_values(input, {
         0.0f, 1.0f, 2.0f, 3.0f,
@@ -570,7 +565,7 @@ TEST(batch_to_space_fp32_gpu, i1212112_bs112321_cb02000_ce00110) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(batch_to_space("batch_to_space", "Input", tensor(format::bfwzyx, {1,1,2,3,2,1}, 1),
                                                            tensor(format::bfwzyx, {0,0,1,0,0,0}, 0),
                                                            tensor(format::bfwzyx, {0,0,0,2,0,0}, 0),
@@ -582,7 +577,7 @@ TEST(batch_to_space_fp32_gpu, i1212112_bs112321_cb02000_ce00110) {
     auto outputs = network.execute();
 
     auto output = outputs.at("batch_to_space").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         24.f, 25.f, 28.f, 29.f,
@@ -605,9 +600,9 @@ TEST(batch_to_space_fp32_gpu, i21621_bs1112_cb0201_ce0810_b_fs_yx_fsv16) {
     //  Output :      1x6x1x1
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(2), feature(16), spatial(1, 2)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, input_shape });
 
     set_values(input, {
         0.0f,  1.0f,  2.0f,  3.0f,  4.0f,  5.0f,  6.0f,  7.0f,
@@ -621,7 +616,7 @@ TEST(batch_to_space_fp32_gpu, i21621_bs1112_cb0201_ce0810_b_fs_yx_fsv16) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(reorder("input_fsv", "Input", format::b_fs_yx_fsv16, data_types::f32));
     topology.add(batch_to_space("batch_to_space", "input_fsv", tensor(format::bfyx, {1,1,1,2}, 1),
                                                                tensor(format::bfyx, {0,2,0,1}, 0),
@@ -636,7 +631,7 @@ TEST(batch_to_space_fp32_gpu, i21621_bs1112_cb0201_ce0810_b_fs_yx_fsv16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("bts_to_bfyx").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         36.0f, 38.0f, 40.0f, 42.0f, 44.0f, 46.0f
@@ -657,9 +652,9 @@ TEST(batch_to_space_fp32_gpu, i41021_bs1221_cb0201_ce0810_b_fs_yx_fsv16) {
     //  Output :      1x8x3x1
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{batch(4), feature(10), spatial(1, 2)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, input_shape });
 
     set_values(input, {
         0.0f,  1.0f,  2.0f,  3.0f,  4.0f,  5.0f,  6.0f,  7.0f,  8.0f,  9.0f,
@@ -673,7 +668,7 @@ TEST(batch_to_space_fp32_gpu, i41021_bs1221_cb0201_ce0810_b_fs_yx_fsv16) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(reorder("input_fsv", "Input", format::b_fs_yx_fsv16, data_types::f32));
     topology.add(batch_to_space("batch_to_space", "input_fsv", tensor(format::bfyx, {1,2,2,1}, 1),
                                                                tensor(format::bfyx, {0,8,1,0}, 0),
@@ -688,7 +683,7 @@ TEST(batch_to_space_fp32_gpu, i41021_bs1221_cb0201_ce0810_b_fs_yx_fsv16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("bts_to_bfyx").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         28.0f, 9.0f,  29.0f, 68.0f, 49.0f, 69.0f,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/binary_convolution_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/binary_convolution_gpu_test.cpp
index a5535f3239fc77..e0ac110a817a8b 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/binary_convolution_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/binary_convolution_gpu_test.cpp
@@ -1,26 +1,18 @@
-// Copyright (C) 2018-2021 Intel Corporation
+// Copyright (C) 2018-2021 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-
-#include <gtest/gtest.h>
-#include <api/memory.hpp>
-#include <api/input_layout.hpp>
-#include "api/binary_convolution.hpp"
-#include "api/reorder.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <iostream>
-#include <api/data.hpp>
-#include <src/include/to_string_utils.h>
-#include "float16.h"
 #include "test_utils.h"
 
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/binary_convolution.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
+
+#include <iostream>
+
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 // Batch, groups, IC, IW, IH, OC, OW, OH, KH, KW, SH, SW, PH, PW
 struct TestParams {
@@ -74,23 +66,23 @@ struct TestParams {
     }
 };
 
-static void fill(cldnn::memory& mem) {
-    auto ptr = mem.pointer<uint32_t>();
-    for (size_t i = 0; i < div_up(mem.get_layout().count(), 32); i++) {
+static void fill(cldnn::memory::ptr mem) {
+    cldnn::mem_lock<uint32_t> ptr(mem, get_test_stream());
+    for (size_t i = 0; i < div_up(mem->get_layout().count(), 32); i++) {
         ptr[i] = (uint32_t)rand() % (1 << 31);
     }
 }
 
 template <typename data_t_src, typename data_t_wei,
           typename data_t_acc, typename data_t_dst>
-void compute_ref_conv_bin(const cldnn::memory &src,
-                          const cldnn::memory &weights,
-                          cldnn::memory &dst,
-                          TestParams &p)
-{
-    auto src_data     = src.pointer<data_t_src>();
-    auto weights_data = weights.pointer<data_t_wei>();
-    auto dst_data     = dst.pointer<data_t_dst>();
+void compute_ref_conv_bin(const cldnn::memory::ptr src,
+                          const cldnn::memory::ptr weights,
+                          cldnn::memory::ptr dst,
+                          TestParams &p) {
+
+    cldnn::mem_lock<data_t_src> src_data(src, get_test_stream());
+    cldnn::mem_lock<data_t_wei> weights_data(weights, get_test_stream());
+    cldnn::mem_lock<data_t_dst> dst_data(dst, get_test_stream());
 
     int pack_size = sizeof(data_t_src) * 8;
 
@@ -117,7 +109,7 @@ void compute_ref_conv_bin(const cldnn::memory &src,
         return (data_t_src)((val >> bit) & 0x1);
     };
 
-    auto ker = [=](data_t_acc &d, int g, int mb, int oc,int oh, int ow, int& ks) {
+    auto ker = [&](data_t_acc &d, int g, int mb, int oc,int oh, int ow, int& ks) {
         for (int ic = 0; ic < IC / NG; ++ic) {
             for (int kh = 0; kh < KH; ++kh)
                 for (int kw = 0; kw < KW; ++kw) {
@@ -133,15 +125,12 @@ void compute_ref_conv_bin(const cldnn::memory &src,
                     uint8_t w = extract_bit(weights_data[widx / pack_size], widx % pack_size);
                     uint8_t s = 0;
 
-                    if ((ih < 0 || ih >= IH || iw < 0 || iw >= IW))
-                    {
+                    if ((ih < 0 || ih >= IH || iw < 0 || iw >= IW)) {
                         if (p.pad_value == 0.0f)
                             continue;
                         else
                             s = (p.pad_value == -1.0f) ? 0 : 1;
-                    }
-                    else
-                    {
+                    } else {
                         if (ic == 0) ks++;
                         iidx = mb * div_up(IC, pack_size) * IH * IW
                                + g * div_up(IC, pack_size) / NG * IH * IW
@@ -180,18 +169,15 @@ void compute_ref_conv_bin(const cldnn::memory &src,
     }
 }
 
-class binary_convolution_test : public ::testing::TestWithParam<TestParams>
-{
-    void SetUp()
-    {
+class binary_convolution_test : public ::testing::TestWithParam<TestParams> {
+    void SetUp() {
         std::cout << GetParam() << std::endl;
         ASSERT_TRUE(GetParam().isConsistent());
     }
 };
 
-TEST_P(binary_convolution_test, conv)
-{
-    const auto& engine = get_test_engine();
+TEST_P(binary_convolution_test, conv) {
+    auto& engine = get_test_engine();
     cldnn::build_options options;
     options.set_option(cldnn::build_option::optimize_data(true));
     topology topology_bin;
@@ -217,9 +203,9 @@ TEST_P(binary_convolution_test, conv)
                             cldnn::feature(p.oc),
                             cldnn::spatial(p.ow, p.oh)};
 
-    auto input       = memory::allocate(engine, { cldnn::data_types::bin, cldnn::format::b_fs_yx_32fp, is_size });
-    auto weights     = memory::allocate(engine, { cldnn::data_types::bin, cldnn::format::bfyx, wei_size });
-    auto output_ref  = memory::allocate(engine, { cldnn::data_types::f32, cldnn::format::bfyx, os_size });
+    auto input       = engine.allocate_memory({ cldnn::data_types::bin, cldnn::format::b_fs_yx_32fp, is_size });
+    auto weights     = engine.allocate_memory({ cldnn::data_types::bin, cldnn::format::bfyx, wei_size });
+    auto output_ref  = engine.allocate_memory({ cldnn::data_types::f32, cldnn::format::bfyx, os_size });
 
     fill(input);
     fill(weights);
@@ -231,7 +217,7 @@ TEST_P(binary_convolution_test, conv)
 //    print_bin_blob(weights, "weights");
 //    print_blob(output_ref, "ref_out");
 
-    topology_bin.add(input_layout(input_name, input.get_layout()));
+    topology_bin.add(input_layout(input_name, input->get_layout()));
     topology_bin.add(data(output_name + weights_suffix, weights));
 
     topology_bin.add(binary_convolution(output_name, input_name, {output_name + weights_suffix},
@@ -243,18 +229,15 @@ TEST_P(binary_convolution_test, conv)
     std::map<primitive_id, network_output> outputs = network_bin.execute();
     auto outputMemory = outputs.at(output_name).get_memory();
 
-    for (size_t i = 0; i < output_ref.count(); i++) {
-        if (p.dt == data_types::f32)
-        {
-            auto ref = output_ref.pointer<float>();
-            auto opt = outputMemory.pointer<float>();
+    for (size_t i = 0; i < output_ref->count(); i++) {
+        if (p.dt == data_types::f32) {
+            cldnn::mem_lock<float> ref(output_ref, get_test_stream());
+            cldnn::mem_lock<float> opt(outputMemory, get_test_stream());
 
             ASSERT_EQ(ref[i], opt[i]) << i;
-        }
-        else if (p.dt == data_types::f16)
-        {
-            auto ref = output_ref.pointer<float>();
-            auto opt = outputMemory.pointer<uint16_t>();
+        } else if (p.dt == data_types::f16) {
+            cldnn::mem_lock<float> ref(output_ref, get_test_stream());
+            cldnn::mem_lock<uint16_t> opt(outputMemory, get_test_stream());
 
             ASSERT_EQ(ref[i], float16_to_float32(opt[i])) << i;
         }
@@ -320,20 +303,19 @@ INSTANTIATE_TEST_CASE_P(BinaryConvTest, binary_convolution_test, ::testing::Valu
 ),);
 
 template <typename T>
-static void set_binary_values(const cldnn::memory& mem, std::vector<T> args) {
-    auto ptr = mem.pointer<T>();
+static void set_binary_values(cldnn::memory::ptr mem, std::vector<T> args) {
+    cldnn::mem_lock<T> ptr(mem, get_test_stream());
 
     auto it = ptr.begin();
     for (auto x : args)
         *it++ = x;
 }
 
-TEST(binary_convolution, basic_convolution_1x1_single_packed_channel)
-{
-    const auto& engine = get_test_engine();
+TEST(binary_convolution, basic_convolution_1x1_single_packed_channel) {
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::bin, format::b_fs_yx_32fp, { 1, 16, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::bin, format::bfyx, { 4, 16, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::bin, format::b_fs_yx_32fp, { 1, 16, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::bin, format::bfyx, { 4, 16, 1, 1 } });
 
     // 0 0 1 0  0 1 0 0  1 0 1 0  1 0 1 0
     // 1 0 0 0  0 1 1 0  0 1 1 0  1 0 1 0
@@ -373,7 +355,7 @@ TEST(binary_convolution, basic_convolution_1x1_single_packed_channel)
              4.0f,  2.0f, -6.0f,  14.0f };
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             binary_convolution("binary_conv", "input", { "weights" },
                                { 1,1,1,1 },
@@ -396,8 +378,8 @@ TEST(binary_convolution, basic_convolution_1x1_single_packed_channel)
     EXPECT_EQ(outputs.begin()->first, "binary_conv");
 
     auto output_memory = outputs.at("binary_conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.data_type, data_types::f32);
@@ -413,10 +395,10 @@ TEST(binary_convolution, basic_convolution_1x1_single_packed_channel)
 }
 
 TEST(binary_convolution, basic_convolution_1x1_single_packed_channel_fp16) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::bin, format::b_fs_yx_32fp, { 1, 16, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::bin, format::bfyx, { 4, 16, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::bin, format::b_fs_yx_32fp, { 1, 16, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::bin, format::bfyx, { 4, 16, 1, 1 } });
 
     // 0 0 1 0  0 1 0 0  1 0 1 0  1 0 1 0
     // 1 0 0 0  0 1 1 0  0 1 1 0  1 0 1 0
@@ -456,7 +438,7 @@ TEST(binary_convolution, basic_convolution_1x1_single_packed_channel_fp16) {
              4.0f,  2.0f, -6.0f,  14.0f };
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             binary_convolution("binary_conv", "input", { "weights" },
                                { 1,1,1,1 },
@@ -479,8 +461,8 @@ TEST(binary_convolution, basic_convolution_1x1_single_packed_channel_fp16) {
     EXPECT_EQ(outputs.begin()->first, "binary_conv");
 
     auto output_memory = outputs.at("binary_conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<uint16_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<uint16_t> output_ptr(output_memory, get_test_stream());
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.data_type, data_types::f16);
@@ -489,9 +471,7 @@ TEST(binary_convolution, basic_convolution_1x1_single_packed_channel_fp16) {
     EXPECT_EQ(output_layout.size.spatial[1], 2);
     EXPECT_EQ(output_layout.size.spatial[0], 2);
 
-    for (size_t i = 0; i < output_layout.count(); i++)
-    {
+    for (size_t i = 0; i < output_layout.count(); i++) {
         EXPECT_EQ(float16_to_float32(output_ptr[i]), output_vec[i]) << "index="<< i;
     }
 }
-
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/border_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/border_gpu_test.cpp
index 1e206a2876cb39..df8a8d47c203a3 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/border_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/border_gpu_test.cpp
@@ -2,18 +2,10 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
+#include "test_utils.h"
 
-#include <api/engine.hpp>
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/border.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-
-#include "test_utils/test_utils.h"
-#include "test_utils/uniform_quantized_real_distribution.hpp"
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/border.hpp>
 
 #include <cstddef>
 
@@ -26,7 +18,7 @@ static std::vector<T> generate_rnd_real_input(
     const T min = static_cast<T>(0), const T max = static_cast<T>(1), const unsigned rnd_bits = 9)
 {
     static std::default_random_engine rnd_gen(random_seed);
-    cldnn::tests::distributions::uniform_quantized_real_distribution<T> rnd_dist(min, max, rnd_bits);
+    tests::distributions::uniform_quantized_real_distribution<T> rnd_dist(min, max, rnd_bits);
 
     auto acum = std::accumulate(sizes.begin(), sizes.end(), static_cast<std::size_t>(1), std::multiplies<std::size_t>());
 
@@ -62,12 +54,12 @@ TEST(border_gpu, basic_yxfb_0x0x1x2_0x0x3x4_border_constant) {
     constexpr auto out_size_y = in_size_y + blt_size_y + brb_size_y;
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::yxfb, {in_size_b, in_size_f, in_size_x, in_size_y}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::yxfb, {in_size_b, in_size_f, in_size_x, in_size_y}});
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -97,7 +89,7 @@ TEST(border_gpu, basic_yxfb_0x0x1x2_0x0x3x4_border_constant) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(out_data.size(), static_cast<std::size_t>(out_size_b * out_size_f * out_size_y * out_size_x));
 
@@ -140,12 +132,12 @@ TEST(border_gpu, basic_bfzyx_0x0x1x01_0x0x0x0x3_border_constant) {
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
     constexpr auto out_size_z = in_size_z + blt_size_z + brb_size_z;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ in_size_b, in_size_f, in_size_x, in_size_y, in_size_z } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ in_size_b, in_size_f, in_size_x, in_size_y, in_size_z } });
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -200,7 +192,7 @@ TEST(border_gpu, basic_bfzyx_0x0x1x01_0x0x0x0x3_border_constant) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(out_data.size(), static_cast<std::size_t>(out_size_b * out_size_f * out_size_y * out_size_x * out_size_z));
 
@@ -249,12 +241,12 @@ TEST(border_gpu, basic_bfwzyx_0x0x0x1x0x1_0x0x0x1x0x1_border_constant) {
     constexpr auto out_size_z = in_size_z + blt_size_z + brb_size_z;
     constexpr auto out_size_w = in_size_w + blt_size_w + brb_size_w;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx, tensor{ batch(in_size_b), feature(in_size_f), spatial(in_size_x, in_size_y, in_size_z, in_size_w) } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfwzyx, tensor{ batch(in_size_b), feature(in_size_f), spatial(in_size_x, in_size_y, in_size_z, in_size_w) } });
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -308,7 +300,7 @@ TEST(border_gpu, basic_bfwzyx_0x0x0x1x0x1_0x0x0x1x0x1_border_constant) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(out_data.size(), static_cast<std::size_t>(out_size_b * out_size_f * out_size_y * out_size_x * out_size_z * out_size_w));
 
@@ -353,12 +345,12 @@ TEST(border_gpu, basic_yxfb_0x0x1x2_0x0x3x4_border_constant_non_constant) {
     constexpr auto out_size_y = in_size_y + blt_size_y + brb_size_y;
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::yxfb, {in_size_b, in_size_f, in_size_x, in_size_y}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::yxfb, {in_size_b, in_size_f, in_size_x, in_size_y}});
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -388,7 +380,7 @@ TEST(border_gpu, basic_yxfb_0x0x1x2_0x0x3x4_border_constant_non_constant) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(out_data.size(), static_cast<std::size_t>(out_size_b * out_size_f * out_size_y * out_size_x));
 
@@ -429,12 +421,12 @@ TEST(border_gpu, basic_yxfb_0x0x1x2_0x0x3x4_border_mirror) {
     constexpr auto out_size_y = in_size_y + blt_size_y + brb_size_y;
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::yxfb, {in_size_b, in_size_f, in_size_x, in_size_y}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::yxfb, {in_size_b, in_size_f, in_size_x, in_size_y}});
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -464,7 +456,7 @@ TEST(border_gpu, basic_yxfb_0x0x1x2_0x0x3x4_border_mirror) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(out_data.size(), static_cast<std::size_t>(out_size_b * out_size_f * out_size_y * out_size_x));
 
@@ -507,12 +499,12 @@ TEST(border_gpu, basic_bfzyx_0x0x0x0x1_0x0x0x0x1_border_mirror) {
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
     constexpr auto out_size_z = in_size_z + blt_size_z + brb_size_z;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ in_size_b, in_size_f, in_size_x, in_size_y, in_size_z } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ in_size_b, in_size_f, in_size_x, in_size_y, in_size_z } });
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -532,7 +524,7 @@ TEST(border_gpu, basic_bfzyx_0x0x0x0x1_0x0x0x0x1_border_mirror) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (auto b = 0; b < out_size_b; ++b) {             // B
         for (auto f = 0; f < out_size_f; ++f) {         // F
@@ -587,12 +579,12 @@ TEST(border_gpu, basic_bfzyxw_0x0x0x0x1_0x0x0x0x1_border_mirror) {
     constexpr auto out_size_z = in_size_z + blt_size_z + brb_size_z;
     constexpr auto out_size_w = in_size_w + blt_size_w + brb_size_w;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx, tensor{ batch(in_size_b), feature(in_size_f), spatial(in_size_x, in_size_y, in_size_z, in_size_w) } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfwzyx, tensor{ batch(in_size_b), feature(in_size_f), spatial(in_size_x, in_size_y, in_size_z, in_size_w) } });
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -612,7 +604,7 @@ TEST(border_gpu, basic_bfzyxw_0x0x0x0x1_0x0x0x0x1_border_mirror) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (auto b = 0; b < out_size_b; ++b) {             // B
         for (auto f = 0; f < out_size_f; ++f) {         // F
@@ -664,12 +656,12 @@ TEST(border_gpu, basic_yxfb_0x0x1x2_0x0x3x4_border_mirror_101) {
     constexpr auto out_size_y = in_size_y + blt_size_y + brb_size_y;
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::yxfb, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::yxfb, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -701,7 +693,7 @@ TEST(border_gpu, basic_yxfb_0x0x1x2_0x0x3x4_border_mirror_101) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(out_data.size(), static_cast<std::size_t>(out_size_b * out_size_f * out_size_y * out_size_x));
 
@@ -743,12 +735,12 @@ TEST(border_gpu, basic_bfzyx_0x0x0x0x1_0x0x0x0x1_border_mirror_101) {
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
     constexpr auto out_size_z = in_size_z + blt_size_z + brb_size_z;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, tensor{ in_size_b, in_size_f, in_size_x, in_size_y, in_size_z } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, tensor{ in_size_b, in_size_f, in_size_x, in_size_y, in_size_z } });
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -781,7 +773,7 @@ TEST(border_gpu, basic_bfzyx_0x0x0x0x1_0x0x0x0x1_border_mirror_101) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(out_data.size(), static_cast<std::size_t>(out_size_b * out_size_f * out_size_y * out_size_x * out_size_z));
 
@@ -829,12 +821,12 @@ TEST(border_gpu, basic_bfwzyx_0x0x0x0x1x1_0x0x0x0x1x1_border_mirror_101) {
     constexpr auto out_size_z = in_size_z + blt_size_z + brb_size_z;
     constexpr auto out_size_w = in_size_w + blt_size_w + brb_size_w;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx, tensor{ batch(in_size_b), feature(in_size_f), spatial(in_size_x, in_size_y, in_size_z, in_size_w) } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfwzyx, tensor{ batch(in_size_b), feature(in_size_f), spatial(in_size_x, in_size_y, in_size_z, in_size_w) } });
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -847,26 +839,26 @@ TEST(border_gpu, basic_bfwzyx_0x0x0x0x1x1_0x0x0x0x1x1_border_mirror_101) {
         1, -2,  3,  -4,
         5,  6,  7,   8,
 
-        2, -3,  4,  -5, 
+        2, -3,  4,  -5,
         15,  4,  4,   4,
 
         2, -6,  13,  -14,
-        3,  7,  7,   7, 
+        3,  7,  7,   7,
     };
     std::vector<float> out_data = {
-        2, -3,  4,  -5, 
+        2, -3,  4,  -5,
         15,  4,  4,   4,
 
         1, -2,  3,  -4,
         5,  6,  7,   8,
 
-        2, -3,  4,  -5, 
-        15,  4,  4,   4, 
+        2, -3,  4,  -5,
+        15,  4,  4,   4,
 
-        2, -6,  13,  -14, 
+        2, -6,  13,  -14,
         3,  7,  7,   7,
 
-        2, -3,  4,  -5, 
+        2, -3,  4,  -5,
         15,  4,  4,   4,
     };
     set_values(input, input_data);
@@ -876,7 +868,7 @@ TEST(border_gpu, basic_bfwzyx_0x0x0x0x1x1_0x0x0x0x1x1_border_mirror_101) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(out_data.size(), static_cast<std::size_t>(out_size_b * out_size_f * out_size_y * out_size_x * out_size_z * out_size_w));
 
@@ -921,12 +913,12 @@ TEST(border_gpu, basic_yxfb_0x0x1x2_0x0x3x4_border_edge) {
     constexpr auto out_size_y = in_size_y + blt_size_y + brb_size_y;
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::yxfb, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::yxfb, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -958,7 +950,7 @@ TEST(border_gpu, basic_yxfb_0x0x1x2_0x0x3x4_border_edge) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(out_data.size(), static_cast<std::size_t>(out_size_b * out_size_f * out_size_y * out_size_x));
 
@@ -996,12 +988,12 @@ TEST(border_gpu, basic_bfyx_2x1x2x3_1x2x3x4_border_constant) {
     constexpr auto out_size_y = in_size_y + blt_size_y + brb_size_y;
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -1021,7 +1013,7 @@ TEST(border_gpu, basic_bfyx_2x1x2x3_1x2x3x4_border_constant) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (auto b = 0; b < out_size_b; ++b) {             // B
         for (auto f = 0; f < out_size_f; ++f) {         // F
@@ -1068,12 +1060,12 @@ TEST(border_gpu, basic_bfyx_2x1x2x3_1x2x3x4_border_mirror) {
     constexpr auto out_size_y = in_size_y + blt_size_y + brb_size_y;
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -1092,7 +1084,7 @@ TEST(border_gpu, basic_bfyx_2x1x2x3_1x2x3x4_border_mirror) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (auto b = 0; b < out_size_b; ++b) {             // B
         for (auto f = 0; f < out_size_f; ++f) {         // F
@@ -1135,12 +1127,12 @@ TEST(border_gpu, basic_bfyx_2x1x2x3_1x2x3x4_border_mirror_101) {
     constexpr auto out_size_y = in_size_y + blt_size_y + brb_size_y;
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -1158,7 +1150,7 @@ TEST(border_gpu, basic_bfyx_2x1x2x3_1x2x3x4_border_mirror_101) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (auto b = 0; b < out_size_b; ++b) {             // B
         for (auto f = 0; f < out_size_f; ++f) {         // F
@@ -1201,12 +1193,12 @@ TEST(border_gpu, basic_bfyx_2x1x2x3_1x2x3x4_border_edge) {
     constexpr auto out_size_y = in_size_y + blt_size_y + brb_size_y;
     constexpr auto out_size_x = in_size_x + blt_size_x + brb_size_x;
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, tensor{in_size_b, in_size_f, in_size_x, in_size_y}});
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
         border("output", "input",
@@ -1224,7 +1216,7 @@ TEST(border_gpu, basic_bfyx_2x1x2x3_1x2x3x4_border_edge) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (auto b = 0; b < out_size_b; ++b) {             // B
         for (auto f = 0; f < out_size_f; ++f) {         // F
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/broadcast_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/broadcast_gpu_test.cpp
index 9784a70a916b8d..8ae447221ed9c1 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/broadcast_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/broadcast_gpu_test.cpp
@@ -2,18 +2,10 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
+#include "test_utils.h"
 
-#include <api/engine.hpp>
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/broadcast.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-
-#include "test_utils/test_utils.h"
-#include "test_utils/uniform_quantized_real_distribution.hpp"
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/broadcast.hpp>
 
 #include <cstddef>
 
@@ -23,8 +15,7 @@ using namespace ::tests;
 template<typename T>
 void start_broadcast_test(data_types cldnn_data_type, std::vector<size_t> output_shape,
                           std::vector<size_t> input_shape, std::vector<size_t> broadcast_axes,
-                          std::vector<T> golden_data)
-{
+                          std::vector<T> golden_data) {
     size_t input_data_size = accumulate(input_shape.rbegin(), input_shape.rend(), (size_t)1, std::multiplies<size_t>());
     EXPECT_GE(input_data_size, (size_t)1);
     std::vector<T> input_data = {};
@@ -51,11 +42,11 @@ void start_broadcast_test(data_types cldnn_data_type, std::vector<size_t> output
         fixed_b_axes.push_back((uint16_t) (broadcast_axes.at(i) + shift));
     }
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {cldnn_data_type, format::bfyx, {input_4d.at(0), input_4d.at(1), input_4d.at(3), input_4d.at(2)}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({cldnn_data_type, format::bfyx, {input_4d.at(0), input_4d.at(1), input_4d.at(3), input_4d.at(2)}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(broadcast("output", "input", {output_4d.at(0), output_4d.at(1), output_4d.at(3), output_4d.at(2)}, fixed_b_axes));
 
     set_values(input, input_data);
@@ -65,7 +56,7 @@ void start_broadcast_test(data_types cldnn_data_type, std::vector<size_t> output
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<T>();
+    cldnn::mem_lock<T> output_ptr(output, get_test_stream());
 
     for (tensor::value_type b = 0; b < output_4d.at(0); ++b) {
         for (tensor::value_type f = 0; f < output_4d.at(1); ++f) {
@@ -110,11 +101,11 @@ void start_broadcast_test_5d(data_types cldnn_data_type, std::vector<size_t> out
         fixed_b_axes.push_back((uint16_t)(broadcast_axes.at(i) + shift));
     }
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { cldnn_data_type, format::bfzyx,{ input_5d.at(0), input_5d.at(1), input_5d.at(4), input_5d.at(3), input_5d.at(2) } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ cldnn_data_type, format::bfzyx,{ input_5d.at(0), input_5d.at(1), input_5d.at(4), input_5d.at(3), input_5d.at(2) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(broadcast("output", "input", { output_5d.at(0), output_5d.at(1), output_5d.at(4), output_5d.at(3), output_5d.at(2) }, fixed_b_axes));
 
     set_values(input, input_data);
@@ -124,7 +115,7 @@ void start_broadcast_test_5d(data_types cldnn_data_type, std::vector<size_t> out
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<T>();
+    cldnn::mem_lock<T> output_ptr(output, get_test_stream());
 
     for (tensor::value_type b = 0; b < output_5d.at(0); ++b) {
         for (tensor::value_type f = 0; f < output_5d.at(1); ++f) {
@@ -1438,11 +1429,11 @@ TEST(broadcast_gpu_int64_t, bfyx_2_to_2x3x4x5_w_b_axes_1_2_3) {
 
 TEST(broadcast_gpu, basic_error_wrong_b_axes_size) {
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 1, 1, 1}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(broadcast("output", "input", {2, 3, 4, 5}, {0, 1, 2, 3, 4}));
 
     std::string msg_to_find = "Incorrect parameters configuration: broadcast_axes size should be less or equal 4.";
@@ -1451,11 +1442,11 @@ TEST(broadcast_gpu, basic_error_wrong_b_axes_size) {
 
 TEST(broadcast_gpu, basic_error_wrong_b_axis_value) {
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 1, 1, 1}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(broadcast("output", "input", {2, 3, 4, 5}, {0, 4}));
 
     std::string msg_to_find = "Incorrect parameters configuration: broadcast_axes index should be within broadcast_sizes range.";
@@ -1464,11 +1455,11 @@ TEST(broadcast_gpu, basic_error_wrong_b_axis_value) {
 
 TEST(broadcast_gpu, basic_error_duplicate_b_axis_values) {
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 1, 1, 1}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(broadcast("output", "input", {2, 3, 4, 5}, {0, 1, 1}));
 
     std::string msg_to_find = "Incorrect parameters configuration: Duplicate axes numbers was found in broadcast_axes.";
@@ -1477,11 +1468,11 @@ TEST(broadcast_gpu, basic_error_duplicate_b_axis_values) {
 
 TEST(broadcast_gpu, basic_error_wrong_input_dimension_0) {
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {2, 3, 4, 5}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {2, 3, 4, 5}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(broadcast("output", "input", {2, 3, 4, 5}, {1}));
 
     std::string msg_to_find = "Input size on dimension number 0(=2) is not equal to: (=1)";
@@ -1490,11 +1481,11 @@ TEST(broadcast_gpu, basic_error_wrong_input_dimension_0) {
 
 TEST(broadcast_gpu, basic_error_not_dividable_2x3x4x5_to_3x3x4x5) {
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {2, 3, 4, 5}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {2, 3, 4, 5}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(broadcast("output", "input", {3, 3, 4, 5}, {}));
 
     std::string msg_to_find = "Invalid broadcast size: not dividable by input size";
@@ -1503,11 +1494,11 @@ TEST(broadcast_gpu, basic_error_not_dividable_2x3x4x5_to_3x3x4x5) {
 
 TEST(broadcast_gpu, basic_error_not_dividable_3_to_2x3x4x5_w_b_axes_0x1x3) {
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 1, 3, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 1, 3, 1}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(broadcast("output", "input", {2, 3, 4, 5}, {0, 1, 3}));
 
     std::string msg_to_find = "Invalid broadcast size: not dividable by input size";
@@ -1516,11 +1507,11 @@ TEST(broadcast_gpu, basic_error_not_dividable_3_to_2x3x4x5_w_b_axes_0x1x3) {
 
 TEST(broadcast_gpu, basic_error_not_dividable_4x5_to_3x4x5_w_b_axes_1) {
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 3, 5, 4}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 3, 5, 4}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(broadcast("output", "input", {2, 3, 4, 5}, {1}));
 
     std::string msg_to_find = "Invalid broadcast size: not dividable by input size";
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/cache_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/cache_test.cpp
index 89935b79549406..469d0b3103f1f8 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/cache_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/cache_test.cpp
@@ -2,18 +2,11 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-
 #include "test_utils.h"
 
-#include <api/engine.hpp>
-#include <api/program.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/input_layout.hpp>
-#include <api/convolution.hpp>
-#include <api/data.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/convolution.hpp>
+#include <cldnn/primitives/data.hpp>
 
 #include <iostream>
 #include <fstream>
@@ -172,13 +165,13 @@ void remove(const std::string& filename) {
 
 class cache_test_helper {
 public:
-    cache_test_helper(cldnn::engine engine, cache_version v)
+    cache_test_helper(cldnn::engine& engine, cache_version v)
         : _engine(engine)
         , _mode(cldnn::tuning_mode::tuning_disabled)
         , cache_filename(get_temporary_cache_file())
     {
         auto cache = get_cache_version(v);
-        auto eus = engine.get_info().cores_count;
+        auto eus = engine.get_device_info().execution_units_count;
         replace(cache, eus_marker, eus);
 
         write(cache_filename, cache);
@@ -210,7 +203,7 @@ class cache_test_helper {
     }
 
     void test() {
-        auto w_mem = cldnn::memory::allocate(_engine, cldnn::layout(cldnn::data_types::f32, cldnn::format::bfyx, { 16, 16, 1, 1 }));
+        auto w_mem = _engine.allocate_memory(cldnn::layout(cldnn::data_types::f32, cldnn::format::bfyx, { 16, 16, 1, 1 }));
         auto topology = cldnn::topology(
             cldnn::input_layout("input", cldnn::layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1, 16, 3, 3 })),
             cldnn::data("weights", w_mem),
@@ -225,7 +218,7 @@ class cache_test_helper {
             cldnn::build_option::optimize_data(true)
         );
         auto network = cldnn::network(_engine, topology, build_opts);
-        auto in_mem = cldnn::memory::allocate(_engine, cldnn::layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1, 16, 3, 3 }));
+        auto in_mem = _engine.allocate_memory(cldnn::layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1, 16, 3, 3 }));
         network.set_input_data("input", in_mem);
         network.execute();
 
@@ -247,7 +240,7 @@ class cache_test_helper {
         if (compare_cache.compare) {
             auto cache = read(cache_filename);
             auto expected_cache = get_cache_version(compare_cache.value);
-            auto eus = _engine.get_info().cores_count;
+            auto eus = _engine.get_device_info().execution_units_count;
             replace(expected_cache, eus_marker, eus);
 
             EXPECT_EQ(cache, expected_cache);
@@ -266,7 +259,7 @@ class cache_test_helper {
         optional_compare(T v, bool neq) : compare(true), not_equal(neq), value(v) {}
     };
 
-    cldnn::engine _engine;
+    cldnn::engine& _engine;
 
     cldnn::tuning_mode _mode;
 
@@ -311,7 +304,7 @@ class cache_version_test : public testing::TestWithParam<cache_version> {
 
 TEST(cache_test, no_cache_baseline) {
     SCOPED_TRACE("default implementation same as reference, cache tests may provide invalid pass");
-    auto engine = tests::get_test_engine();
+    auto& engine = tests::get_test_engine();
     auto helper = cache_test_helper(engine, cache_version::version_2);
 
     helper.with_mode(cldnn::tuning_mode::tuning_disabled)
@@ -321,7 +314,7 @@ TEST(cache_test, no_cache_baseline) {
 
 TEST_P(cache_version_test, use_only) {
     auto version = GetParam();
-    auto engine = tests::get_test_engine();
+    auto& engine = tests::get_test_engine();
 
     cache_test_helper helper(engine, version);
     helper.with_mode(cldnn::tuning_mode::tuning_use_cache)
@@ -337,7 +330,7 @@ TEST_P(cache_version_test, update) {
         ex_version = cache_version::version_2_from_1;
     }
 
-    auto engine = tests::get_test_engine();
+    auto& engine = tests::get_test_engine();
 
     cache_test_helper helper(engine, version);
     helper.with_mode(cldnn::tuning_mode::tuning_use_and_update)
@@ -353,7 +346,7 @@ INSTANTIATE_TEST_CASE_P(
     cache_version_test::to_string);
 
 TEST(cache_test, remove_invalid) {
-    auto engine = tests::get_test_engine();
+    auto& engine = tests::get_test_engine();
 
     cache_test_helper helper(engine, cache_version::version_2_invalid);
     helper.with_mode(cldnn::tuning_mode::tuning_use_and_update)
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/cl_mem_input_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/cl_mem_input_test.cpp
index 2ff579ca8ffcd6..bf5d79b74fa86c 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/cl_mem_input_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/cl_mem_input_test.cpp
@@ -2,23 +2,18 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/activation.hpp"
-#include <api/topology.hpp>
-#include <api/tensor.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include <api/data.hpp>
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
 
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/activation.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/runtime/device_query.hpp>
 
-#include <cl2_wrapper.h>
+
+#include <ocl/ocl_wrapper.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 typedef std::chrono::high_resolution_clock Time;
 typedef std::chrono::nanoseconds ns;
@@ -26,10 +21,8 @@ typedef std::chrono::duration<double, std::ratio<1, 1000>> ms;
 typedef std::chrono::duration<float> fsec;
 
 
-void checkStatus(int status, const char *message)
-{
-    if (status != 0)
-    {
+void checkStatus(int status, const char *message) {
+    if (status != 0) {
         std::string str_message(message + std::string(": "));
         std::string str_number(std::to_string(status));
 
@@ -37,8 +30,7 @@ void checkStatus(int status, const char *message)
     }
 }
 
-std::vector<unsigned char> createSampleData(int width, int height)
-{
+std::vector<unsigned char> createSampleData(int width, int height) {
     int data_size = width * (height + height / 2);
     auto data = std::vector<unsigned char>(data_size);
     srand((unsigned)time(0));
@@ -57,8 +49,7 @@ std::vector<unsigned char> createSampleData(int width, int height)
     return data;
 }
 
-std::vector<float> createReferenceData(std::vector<unsigned char> data, int width, int height, cldnn::format format)
-{
+std::vector<float> createReferenceData(std::vector<unsigned char> data, int width, int height, cldnn::format format) {
     auto img = std::vector<float>(width * height * 3);
     for (int i = 0; i < height; i++) {
         for (int j = 0; j < width; j++) {
@@ -78,8 +69,7 @@ std::vector<float> createReferenceData(std::vector<unsigned char> data, int widt
                 img[j + width * i] = R;
                 img[j + width * i + width * height] = G;
                 img[j + width * i + width * height * 2] = B;
-            }
-            else { //byxf
+            } else { //byxf
                 img[3* width*i + 3 * j] = R;
                 img[3 * width * i + 3 * j + 1] = G;
                 img[3 * width*i + 3 * j + 2] = B;
@@ -90,14 +80,12 @@ std::vector<float> createReferenceData(std::vector<unsigned char> data, int widt
     return img;
 }
 
-struct OpenCL
-{
+struct OpenCL {
     cl::Context _context;
     cl::Device _device;
     cl::CommandQueue _queue;
 
-    OpenCL()
-    {
+    OpenCL() {
         // get Intel iGPU OCL device, create context and queue
         {
             static constexpr auto INTEL_PLATFORM_VENDOR = "Intel(R) Corporation";
@@ -136,8 +124,7 @@ struct OpenCL
             _queue = cl::CommandQueue(_context, _device, props);
         }
     }
-    void releaseOclImage(std::shared_ptr<cl_mem> image)
-    {
+    void releaseOclImage(std::shared_ptr<cl_mem> image) {
         checkStatus(clReleaseMemObject(*image), "clReleaseMemObject");
     }
 };
@@ -153,9 +140,9 @@ TEST(cl_mem_check, check_2_inputs) {
     image_format.image_channel_order = CL_R;
     image_format.image_channel_data_type = CL_UNORM_INT8;
     cl_image_desc image_desc = { CL_MEM_OBJECT_IMAGE2D, (size_t)width, (size_t)height, 0,
-                                 0, 0, 0, 0, 0, NULL };
+                                 0, 0, 0, 0, 0, { nullptr } };
 
-    cl_mem nv12_image_plane_y = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, NULL, &err);
+    cl_mem nv12_image_plane_y = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, nullptr, &err);
     checkStatus(err, "Creating nv12 image plane_y failed");
 
     image_format.image_channel_order = CL_RG;
@@ -163,38 +150,38 @@ TEST(cl_mem_check, check_2_inputs) {
     image_desc.image_height = height / 2;
     image_desc.image_depth = 1;
 
-    cl_mem nv12_image_plane_uv = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, NULL, &err);
+    cl_mem nv12_image_plane_uv = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, nullptr, &err);
     checkStatus(err, "Creating nv12 image plane_uv failed");
 
     size_t origin[3] = { 0, 0, 0 };
     size_t y_region[3] = { (size_t)width, (size_t)height, 1 };
     size_t uv_region[3] = { (size_t)width / 2, (size_t)height / 2, 1 };
 
-    err = clEnqueueWriteImage(ocl_instance->_queue.get(), nv12_image_plane_y, true, origin, y_region, 0, 0, &data[0], 0, NULL, NULL);
+    err = clEnqueueWriteImage(ocl_instance->_queue.get(), nv12_image_plane_y, true, origin, y_region, 0, 0, &data[0], 0, nullptr, nullptr);
     checkStatus(err, "Writing nv12 image plane_y failed");
 
-    err = clEnqueueWriteImage(ocl_instance->_queue.get(), nv12_image_plane_uv, true, origin, uv_region, 0, 0, &data[width * height], 0, NULL, NULL);
+    err = clEnqueueWriteImage(ocl_instance->_queue.get(), nv12_image_plane_uv, true, origin, uv_region, 0, 0, &data[width * height], 0, nullptr, nullptr);
     checkStatus(err, "Writing nv12 image plane_uv failed");
 
-    device_query query(static_cast<void*>(ocl_instance->_context.get()));
+    device_query query(engine_types::ocl, runtime_types::ocl, static_cast<void*>(ocl_instance->_context.get()));
     auto devices = query.get_available_devices();
 
     auto engine_config = cldnn::engine_configuration();
-    engine engine(devices.begin()->second, engine_config);
+    auto engine = engine::create(engine_types::ocl, runtime_types::ocl, devices.begin()->second, engine_config);
 
     auto input = input_layout("input", { data_types::i8, format::nv12, {1,1,height,width} });
     auto input2 = input_layout("input2", { data_types::i8, format::nv12, {1,1,height / 2,width / 2} });
     auto output_format = cldnn::format::byxf;
     layout output_layout(data_types::f32, output_format, { 1,3,height,width });
-    auto input_memory = cldnn::memory::share_image(engine, input.layout, nv12_image_plane_y,  0);
-    auto input_memory2 = cldnn::memory::share_image(engine,  input2.layout, nv12_image_plane_uv, 0);
+    auto input_memory = engine->share_image(input.layout, nv12_image_plane_y);
+    auto input_memory2 = engine->share_image(input2.layout, nv12_image_plane_uv);
 
     topology topology;
     topology.add(input);
     topology.add(input2);
     topology.add(reorder("reorder", "input", "input2", output_layout));
 
-    network network(engine, topology);
+    network network(*engine, topology);
     network.set_input_data("input", input_memory);
     network.set_input_data("input2", input_memory2);
 
@@ -202,7 +189,7 @@ TEST(cl_mem_check, check_2_inputs) {
 
     std::vector<float> reference_results = createReferenceData(data, width, height, output_format);
     auto output_prim = outputs.begin()->second.get_memory();
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
     int size = width * height * 3;
     for (auto i = 0; i < size; i++) {
         EXPECT_NEAR(reference_results[i], output_ptr[i], 1.001f);
@@ -222,26 +209,26 @@ TEST(cl_mem_check, check_input) {
     image_format.image_channel_order = CL_R;
     image_format.image_channel_data_type = CL_UNORM_INT8;
     cl_image_desc image_desc = { CL_MEM_OBJECT_IMAGE2D, (size_t)width, (size_t)height, 0,
-        0, 0, 0, 0, 0, NULL };
+        0, 0, 0, 0, 0, { nullptr } };
 
-    cl_mem nv12_image_plane_y = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, NULL, &err);
+    cl_mem nv12_image_plane_y = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, nullptr, &err);
     checkStatus(err, "Creating nv12 image plane_y failed");
 
     image_format.image_channel_order = CL_RG;
     image_desc.image_width = width / 2;
     image_desc.image_height = height / 2;
 
-    cl_mem nv12_image_plane_uv = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, NULL, &err);
+    cl_mem nv12_image_plane_uv = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, nullptr, &err);
     checkStatus(err, "Creating nv12 image plane_uv failed");
 
     size_t origin[3] = { 0, 0, 0 };
     size_t y_region[3] = { (size_t)width, (size_t)height, 1 };
     size_t uv_region[3] = { (size_t)width / 2, (size_t)height / 2, 1 };
 
-    err = clEnqueueWriteImage(ocl_instance->_queue.get(), nv12_image_plane_y, true, origin, y_region, 0, 0, &data[0], 0, NULL, NULL);
+    err = clEnqueueWriteImage(ocl_instance->_queue.get(), nv12_image_plane_y, true, origin, y_region, 0, 0, &data[0], 0, nullptr, nullptr);
     checkStatus(err, "Writing nv12 image plane_y failed");
 
-    err = clEnqueueWriteImage(ocl_instance->_queue.get(), nv12_image_plane_uv, true, origin, uv_region, 0, 0, &data[width * height], 0, NULL, NULL);
+    err = clEnqueueWriteImage(ocl_instance->_queue.get(), nv12_image_plane_uv, true, origin, uv_region, 0, 0, &data[width * height], 0, nullptr, nullptr);
     checkStatus(err, "Writing nv12 image plane_uv failed");
 
     image_format.image_channel_order = CL_NV12_INTEL;
@@ -267,12 +254,12 @@ TEST(cl_mem_check, check_input) {
     image_desc.image_depth = 0;
     image_format.image_channel_order = CL_R;
 
-    cl_mem img_y = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, NULL, &err);
+    cl_mem img_y = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, nullptr, &err);
     checkStatus(err, "Creating nv12 image plane_y failed");
 
     image_desc.image_depth = 1;
     image_format.image_channel_order = CL_RG;
-    cl_mem img_uv = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, NULL, &err);
+    cl_mem img_uv = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, nullptr, &err);
     checkStatus(err, "Creating nv12 image plane_uv failed");
 
     size_t regionY[] = { (size_t)width, (size_t)height, 1 };
@@ -288,30 +275,29 @@ TEST(cl_mem_check, check_input) {
     checkStatus(clReleaseMemObject(nv12_image_plane_uv), "clReleaseMemObject");
     checkStatus(clReleaseMemObject(nv12_image_plane_y), "clReleaseMemObject");
 
-    device_query query(static_cast<void*>(ocl_instance->_context.get()));
+    device_query query(engine_types::ocl, runtime_types::ocl, static_cast<void*>(ocl_instance->_context.get()));
     auto devices = query.get_available_devices();
 
-    auto engine_config = cldnn::engine_configuration();
-    engine engine(devices.begin()->second, engine_config);
+    auto engine = engine::create(engine_types::ocl, runtime_types::ocl, devices.begin()->second);
 
     auto input = input_layout("input", { data_types::i8, format::nv12, {1,1,height,width} });
     auto output_format = cldnn::format::byxf;
     layout output_layout(data_types::f32, output_format, { 1,3,height,width });
-    auto input_memory = cldnn::memory::share_image(engine, input.layout, img,  0);
+    auto input_memory = engine->share_image(input.layout, img);
 
     topology topology;
 
     topology.add(input);
     topology.add(reorder("reorder", "input", output_layout));
 
-    network network(engine, topology);
+    network network(*engine, topology);
     network.set_input_data("input", input_memory);
 
     auto outputs = network.execute();
 
     std::vector<float> reference_results = createReferenceData(data, width, height, output_format);
     auto output_prim = outputs.begin()->second.get_memory();
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
     int size = width * height * 3;
     for (auto i = 0; i < size; i++) {
         EXPECT_NEAR(reference_results[i], output_ptr[i], 1.001f);
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/command_queue_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/command_queue_test.cpp
index a07688ee671a6e..0fed8ea95e4c08 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/command_queue_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/command_queue_test.cpp
@@ -3,27 +3,24 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include <api/input_layout.hpp>
 #include "test_utils/test_utils.h"
-#include "api/arg_max_min.hpp"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/arg_max_min.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 using namespace std;
 
 // Run some topology too see if command queue does work correctly
 // Coppied from arg_max_gpu.base test.
-void exexute_network(cldnn::engine engine)
-{
+void exexute_network(cldnn::engine& engine) {
     //  Input  : 2x3x2x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 3, batch_num = 2;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(arg_max_min("arg_max", { "input" }, arg_max_min::max));
 
     vector<float> input_vec = {
@@ -48,11 +45,10 @@ void exexute_network(cldnn::engine engine)
     EXPECT_EQ(outputs.begin()->first, "arg_max");
 
     auto output = outputs.at("arg_max").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[batch_num];
-    for (uint32_t i = 0; i < batch_num; i++)
-    {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+    for (uint32_t i = 0; i < batch_num; i++) {
+        out_buffer[i] = get_value<float>(output_ptr.data(), i);
     }
     int size = x_size * y_size * feature_num;
     int index;
@@ -62,8 +58,7 @@ void exexute_network(cldnn::engine engine)
         EXPECT_LT(out_buffer[i], size);
         index = (int)out_buffer[i];
         value = input_vec[i*size + (int)index];
-        for (int j = 0; j < size; j++)
-        {
+        for (int j = 0; j < size; j++) {
             EXPECT_LE(input_vec[i*size + j], value);
         }
     }
@@ -73,49 +68,34 @@ TEST(command_queue_test, test_priority_hints) {
     engine_configuration configuration =
         engine_configuration(
             false,          // profiling
-            false,          // decorate_kernel_names
-            false,          // dump_custom_program
-            "",             // options
-            "",             // single_kernel
-            true,           // primitives_parallelisation
-            "",             // engine_log
+            queue_types::out_of_order,
             "",             // sources_dumps_dir
             priority_mode_types::low,
             throttle_mode_types::disabled);
-    cldnn::engine engine(configuration);
-    exexute_network(engine);
+    auto engine = engine::create(engine_types::ocl, runtime_types::ocl, configuration);
+    exexute_network(*engine);
 }
 
 TEST(command_queue_test, test_throttle_hints) {
     engine_configuration configuration =
         engine_configuration(
             false,          // profiling
-            false,          // decorate_kernel_names
-            false,          // dump_custom_program
-            "",             // options
-            "",             // single_kernel
-            true,           // primitives_parallelisation
-            "",             // engine_log
+            queue_types::out_of_order,
             "",             // sources_dumps_dir
             priority_mode_types::disabled,
             throttle_mode_types::high);
-    cldnn::engine engine(configuration);
-    exexute_network(engine);
+    auto engine = engine::create(engine_types::ocl, runtime_types::ocl, configuration);
+    exexute_network(*engine);
 }
 
 TEST(command_queue_test, test_priority_and_throttle_hints) {
     engine_configuration configuration =
         engine_configuration(
             false,          // profiling
-            false,          // decorate_kernel_names
-            false,          // dump_custom_program
-            "",             // options
-            "",             // single_kernel
-            true,           // primitives_parallelisation
-            "",             // engine_log
+            queue_types::out_of_order,
             "",             // sources_dumps_dir
             priority_mode_types::high,
             throttle_mode_types::low);
-    cldnn::engine engine(configuration);
-    exexute_network(engine);
-}
\ No newline at end of file
+    auto engine = engine::create(engine_types::ocl, runtime_types::ocl, configuration);
+    exexute_network(*engine);
+}
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/concatenation_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/concatenation_gpu_test.cpp
index 9294d5e5d7cbbe..10f051b9ff435c 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/concatenation_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/concatenation_gpu_test.cpp
@@ -4,17 +4,13 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include <gtest/gtest.h>
-#include <gmock/gmock.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/convolution.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include "test_utils/float16.h"
-#include <api/data.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/convolution.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/reorder.hpp>
+
 #include <algorithm>
 #include <cmath>
 #include <iostream>
@@ -22,10 +18,9 @@
 #include <thread>
 #include <type_traits>
 #include <fstream>
-#include <api/reorder.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 namespace cldnn
 {
@@ -34,13 +29,13 @@ namespace cldnn
 
 
 TEST(concat_gpu, mixed_input_types) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 4, 3 } });
-    auto input1 = memory::allocate(engine, { data_types::i32, format::bfyx, { 1, 1, 4, 3 } });
-    auto input2 = memory::allocate(engine, { data_types::i8, format::bfyx, { 1, 1, 4, 3 } });
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 1, 4, 3 } });
-    auto input4 = memory::allocate(engine, { data_types::i64, format::bfyx, { 1, 1, 4, 3 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 4, 3 } });
+    auto input1 = engine.allocate_memory({ data_types::i32, format::bfyx, { 1, 1, 4, 3 } });
+    auto input2 = engine.allocate_memory({ data_types::i8, format::bfyx, { 1, 1, 4, 3 } });
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 1, 4, 3 } });
+    auto input4 = engine.allocate_memory({ data_types::i64, format::bfyx, { 1, 1, 4, 3 } });
 
     set_values<float>(input0, { 1.0f, 2.0f, 3.0f, 4.0f, 2.0f, 2.0f, 3.0f, 4.0f, 3.0f, 3.0f, 3.0f, 5.0f });
     set_values<int32_t>(input1, { 11, 12, 13, 14, 12, 12, 13, 14, 13, 13, 13, 15 });
@@ -59,11 +54,11 @@ TEST(concat_gpu, mixed_input_types) {
             41.0f, 42.0f, 43.0f, 44.0f, 42.0f, 42.0f, 43.0f, 44.0f, 43.0f, 43.0f, 43.0f, 45.0f };
 
     topology topology(
-            input_layout("input0", input0.get_layout()),
-            input_layout("input1", input1.get_layout()),
-            input_layout("input2", input2.get_layout()),
-            input_layout("input3", input3.get_layout()),
-            input_layout("input4", input4.get_layout()),
+            input_layout("input0", input0->get_layout()),
+            input_layout("input1", input1->get_layout()),
+            input_layout("input2", input2->get_layout()),
+            input_layout("input3", input3->get_layout()),
+            input_layout("input4", input4->get_layout()),
             concatenation("concat",
                           { "input0", "input1", "input2", "input3", "input4" },
                           concatenation::concatenation_axis::along_f,
@@ -83,8 +78,8 @@ TEST(concat_gpu, mixed_input_types) {
     EXPECT_EQ(outputs.begin()->first, "concat");
 
     auto output_memory = outputs.at("concat").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -102,12 +97,12 @@ TEST(concat_gpu, mixed_input_types) {
 }
 
 TEST(concat_gpu, mixed_input_types_5d) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 1, 1, 1, 4, 3 } });
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 1, 1, 1, 4, 3 } });
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 1, 1, 1, 4, 3 } });
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 1, 1, 1, 4, 3 } });
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 1, 1, 1, 4, 3 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 1, 1, 1, 4, 3 } });
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 1, 1, 1, 4, 3 } });
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 1, 1, 1, 4, 3 } });
 
     set_values(input0, { half_t(1.0f), half_t(2.0f), half_t(3.0f),
                          half_t(4.0f), half_t(2.0f), half_t(2.0f),
@@ -133,10 +128,10 @@ TEST(concat_gpu, mixed_input_types_5d) {
             31.0f, 32.0f, 33.0f, 34.0f, 32.0f, 32.0f, 33.0f, 34.0f, 33.0f, 33.0f, 33.0f, 35.0f };
 
     topology topology(
-            input_layout("input0", input0.get_layout()),
-            input_layout("input1", input1.get_layout()),
-            input_layout("input2", input2.get_layout()),
-            input_layout("input3", input3.get_layout()),
+            input_layout("input0", input0->get_layout()),
+            input_layout("input1", input1->get_layout()),
+            input_layout("input2", input2->get_layout()),
+            input_layout("input3", input3->get_layout()),
             concatenation("concat",
                           { "input0", "input1", "input2", "input3" },
                           concatenation::concatenation_axis::along_f,
@@ -155,8 +150,8 @@ TEST(concat_gpu, mixed_input_types_5d) {
     EXPECT_EQ(outputs.begin()->first, "concat");
 
     auto output_memory = outputs.at("concat").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int z_size = output_layout.size.spatial[2];
     int y_size = output_layout.size.spatial[1];
@@ -176,10 +171,10 @@ TEST(concat_gpu, mixed_input_types_5d) {
 }
 
 TEST(concat_gpu, i8_optimization_with_pool) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, {data_types::i8, format::bfyx, {1, 1, 8, 3}});
-    auto input1 = memory::allocate(engine, {data_types::i8, format::bfyx, {1, 1, 8, 3}});
+    auto input0 = engine.allocate_memory({data_types::i8, format::bfyx, {1, 1, 8, 3}});
+    auto input1 = engine.allocate_memory({data_types::i8, format::bfyx, {1, 1, 8, 3}});
 
 
     set_values<int8_t>(input0, { 11, 12, 13,
@@ -207,8 +202,8 @@ TEST(concat_gpu, i8_optimization_with_pool) {
                         18, 14, -13, 15};
 
     layout reorder_layout(data_types::i8, format::yxfb, {7, 2, 2, 1});
-    topology topology(input_layout("input0", input0.get_layout()),
-                      input_layout("input1", input1.get_layout()),
+    topology topology(input_layout("input0", input0->get_layout()),
+                      input_layout("input1", input1->get_layout()),
                       pooling("pool0", "input0", pooling_mode::max, {1, 1, 2, 2}, {1, 1, 1, 1}),
                       pooling("pool1", "input1", pooling_mode::max, {1, 1, 2, 2}, {1, 1, 1, 1}),
                       concatenation("concat",
@@ -228,8 +223,8 @@ TEST(concat_gpu, i8_optimization_with_pool) {
     EXPECT_EQ(outputs.begin()->first, "reorder");
 
     auto output_memory = outputs.at("reorder").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int8_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int8_t> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[0];
     int x_size = output_layout.size.spatial[1];
@@ -279,12 +274,12 @@ TEST(concat_gpu, i8_optimization_with_conv) {
     //  Output:
     // 53  54  30
     // 52  47  37
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, {data_types::i8, format::bfyx, {1, 1, 5, 4}});
-    auto input1 = memory::allocate(engine, {data_types::i8, format::bfyx, {1, 1, 5, 4}});
-    auto input2 = memory::allocate(engine, {data_types::i8, format::bfyx, {1, 1, 5, 4}});
-    auto weights = memory::allocate(engine, { data_types::i8, format::bfyx, { 1, 3, 3, 2 } });
+    auto input0 = engine.allocate_memory({data_types::i8, format::bfyx, {1, 1, 5, 4}});
+    auto input1 = engine.allocate_memory({data_types::i8, format::bfyx, {1, 1, 5, 4}});
+    auto input2 = engine.allocate_memory({data_types::i8, format::bfyx, {1, 1, 5, 4}});
+    auto weights = engine.allocate_memory({ data_types::i8, format::bfyx, { 1, 3, 3, 2 } });
 
     set_values<int8_t>(weights, { 1, 2, 1,
                           2, 1, 2, 1, 2, 1,
@@ -303,14 +298,14 @@ TEST(concat_gpu, i8_optimization_with_conv) {
                           1, 2, -2, 4, 2,
                           3, 5, 3, -3, 1,
                           5, 4, 3, 2, 1 });
-    
+
     VF<int8_t> output_vec = { 53, 54, 30, 52, 47, 37 };
 
-    
+
     layout reorder_layout(data_types::i8, format::bfyx, {1, 1, 2, 3});
-    topology topology(input_layout("input0", input0.get_layout()),
-                      input_layout("input1", input1.get_layout()),
-                      input_layout("input2", input2.get_layout()),
+    topology topology(input_layout("input0", input0->get_layout()),
+                      input_layout("input1", input1->get_layout()),
+                      input_layout("input2", input2->get_layout()),
                       concatenation("concat",
                                     {"input0", "input1", "input2"},
                                     concatenation::concatenation_axis::along_f,
@@ -331,8 +326,8 @@ TEST(concat_gpu, i8_optimization_with_conv) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int8_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int8_t> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -358,20 +353,20 @@ TEST(concat_gpu, i8_optimization_with_pool_conv) {
     //  Output : 1x1x3
     //
     //  Input0:
-    // -3 6 0 2 -1 -1 6 0 5 4 1 6 2 4 0 5 
-    // -2 -1 1 0 2 3 3 3 6 2 4 7 3 6 7 -1 
-    // 7 7 5 -3 1 -1 5 4 0 3 -2 6 2 5 2 4 
-    // 5 -1 3 6 2 0 -3 -1 0 3 0 -1 1 6 1 6 
-    // 5 -2 2 -1 5 6 3 4 1 0 6 6 7 2 6 3 
-    // 6 7 -1 5 5 6 -1 0 -1 5 5 2 3 -1 -3 4 
+    // -3 6 0 2 -1 -1 6 0 5 4 1 6 2 4 0 5
+    // -2 -1 1 0 2 3 3 3 6 2 4 7 3 6 7 -1
+    // 7 7 5 -3 1 -1 5 4 0 3 -2 6 2 5 2 4
+    // 5 -1 3 6 2 0 -3 -1 0 3 0 -1 1 6 1 6
+    // 5 -2 2 -1 5 6 3 4 1 0 6 6 7 2 6 3
+    // 6 7 -1 5 5 6 -1 0 -1 5 5 2 3 -1 -3 4
     //
     //  Input1:
-    //  4 -2 0 0 6 2 0 4 6 4 4 4 -3 -1 4 -3 
-    //  1 0 -1 5 -1 1 4 2 7 7 0 2 3 4 -1 3 
-    //  7 7 2 -3 -1 5 -2 2 6 -3 0 7 0 3 3 3 
-    //  -1 0 -2 -2 7 -3 -3 -1 5 0 3 4 0 -1 2 5 
-    //  2 -1 2 -3 0 -3 -3 2 4 3 3 5 5 7 5 1 
-    //  2 2 -3 6 6 7 1 -1 -2 5 1 -1 4 5 -3 -2 
+    //  4 -2 0 0 6 2 0 4 6 4 4 4 -3 -1 4 -3
+    //  1 0 -1 5 -1 1 4 2 7 7 0 2 3 4 -1 3
+    //  7 7 2 -3 -1 5 -2 2 6 -3 0 7 0 3 3 3
+    //  -1 0 -2 -2 7 -3 -3 -1 5 0 3 4 0 -1 2 5
+    //  2 -1 2 -3 0 -3 -3 2 4 3 3 5 5 7 5 1
+    //  2 2 -3 6 6 7 1 -1 -2 5 1 -1 4 5 -3 -2
     //
     // Filters:
     // -1, 2, -2, 2, -2, 1, 1, 0, -1, 1, 2, -2, 2, 1, -2, 0,
@@ -382,11 +377,11 @@ TEST(concat_gpu, i8_optimization_with_pool_conv) {
     //  Output:
     //  -14, -35, -10
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, {data_types::i8, format::bfyx, {1, 16, 3, 2}});
-    auto input1 = memory::allocate(engine, {data_types::i8, format::bfyx, {1, 16, 3, 2}});
-    auto weights = memory::allocate(engine, {data_types::i8, format::bfyx, {1, 32, 2, 1}});
+    auto input0 = engine.allocate_memory({data_types::i8, format::bfyx, {1, 16, 3, 2}});
+    auto input1 = engine.allocate_memory({data_types::i8, format::bfyx, {1, 16, 3, 2}});
+    auto weights = engine.allocate_memory({data_types::i8, format::bfyx, {1, 32, 2, 1}});
 
     set_values<int8_t>(weights, {-1, 2, -2, 2, -2, 1, 1, 0, -1, 1, 2, -2, 2, 1, -2, 0, 0, -2, -2, -2, -2, -1, 2, 1, 2, -1, -1, 0, 2, -2, -2, 1,
                                 0, -2, 0, 1, -2, -1, -2, 0, -1, -1, -2, 1, -2, 0, 1, 2, 2, 2, 2, -2, 0, 2, 1, -2, -1, -1, 0, -2, 2, -1, 2, -1});
@@ -394,7 +389,7 @@ TEST(concat_gpu, i8_optimization_with_pool_conv) {
     set_values<int8_t>(input0, {-3, 6, 0, 2, -1, -1, 6, 0, 5, 4, 1, 6, 2, 4, 0, 5,
                                 -2, -1, 1, 0, 2, 3, 3, 3, 6, 2, 4, 7, 3, 6, 7, -1,
                                 7, 7, 5, -3, 1, -1, 5, 4, 0, 3, -2, 6, 2, 5, 2, 4,
-                                5, -1, 3, 6, 2, 0, -3, -1, 0, 3, 0, -1, 1, 6, 1, 6, 
+                                5, -1, 3, 6, 2, 0, -3, -1, 0, 3, 0, -1, 1, 6, 1, 6,
                                 5, -2, 2, -1, 5, 6, 3, 4, 1, 0, 6, 6, 7, 2, 6, 3,
                                 6, 7, -1, 5, 5, 6, -1, 0, -1, 5, 5, 2, 3, -1, -3, 4 });
 
@@ -404,12 +399,12 @@ TEST(concat_gpu, i8_optimization_with_pool_conv) {
                                  -1, 0, -2, -2, 7, -3, -3, -1, 5, 0, 3, 4, 0, -1, 2, 5,
                                  2, -1, 2, -3, 0, -3, -3, 2, 4, 3, 3, 5, 5, 7, 5, 1,
                                  2, 2, -3, 6, 6, 7, 1, -1, -2, 5, 1, -1, 4, 5, -3, -2});
-    
+
     VF<int8_t> output_vec = { -14, -35, -10 };
 
     layout reorder_layout(data_types::i8, format::bfyx, {1, 1, 3, 1});
-    topology topology(input_layout("input0", input0.get_layout()),
-                      input_layout("input1", input1.get_layout()),
+    topology topology(input_layout("input0", input0->get_layout()),
+                      input_layout("input1", input1->get_layout()),
                       pooling("pool0", "input0", pooling_mode::max, {1, 1, 2, 2}, {1, 1, 1, 1}),
                       pooling("pool1", "input1", pooling_mode::max, {1, 1, 2, 2}, {1, 1, 1, 1}),
                       concatenation("concat",
@@ -431,8 +426,8 @@ TEST(concat_gpu, i8_optimization_with_pool_conv) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int8_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int8_t> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[0];
     int x_size = output_layout.size.spatial[1];
@@ -527,7 +522,7 @@ struct concat_gpu_4d : public concat_gpu {
     void test(format::type fmt) {
         auto data_type = type_to_data_type<Type>::value;
 
-        const auto& engine = get_test_engine();
+        auto& engine = get_test_engine();
         const size_t batch_num = testing::get<0>(GetParam());
         const std::vector<size_t> in_features = testing::get<1>(GetParam());
         const size_t input_y = testing::get<2>(GetParam());
@@ -539,7 +534,7 @@ struct concat_gpu_4d : public concat_gpu {
         topology topology;
 
         std::vector<VVVVF<Type>> in_data;
-        std::vector<memory> in_memory;
+        std::vector<memory::ptr> in_memory;
         std::vector<primitive_id> input_ids;
         for (size_t i = 0; i < in_features.size(); i++) {
             auto size = tensor(static_cast<int32_t>(batch_num),
@@ -563,7 +558,7 @@ struct concat_gpu_4d : public concat_gpu {
                 }
             }
 
-            auto in_mem = memory::allocate(engine, in_lay);
+            auto in_mem = engine.allocate_memory(in_lay);
             set_values(in_mem, data_flat);
             in_memory.push_back(in_mem);
 
@@ -585,7 +580,7 @@ struct concat_gpu_4d : public concat_gpu {
         network.execute();
 
         auto out_mem = network.get_output("concat").get_memory();
-        auto out_ptr = out_mem.pointer<Type>();
+        cldnn::mem_lock<Type> out_ptr(out_mem, get_test_stream());
 
         for (size_t bi = 0; bi < batch_num; bi++) {
             size_t f_sum = 0;
@@ -594,7 +589,7 @@ struct concat_gpu_4d : public concat_gpu {
                     for (size_t yi = 0; yi < input_y; yi++) {
                         for (size_t xi = 0; xi < input_x; xi++) {
                             auto output_coords = tensor(batch(bi), feature(f_sum + fi), spatial(xi, yi, 0, 0));
-                            auto output_offset = out_mem.get_layout().get_linear_offset(output_coords);
+                            auto output_offset = out_mem->get_layout().get_linear_offset(output_coords);
 
                             auto ref_val = in_data[in_i][bi][fi][yi][xi];
                             auto actual_val = out_ptr[output_offset];
@@ -651,7 +646,7 @@ struct concat_id_conv_gpu_4d : public concat_gpu {
     void test(format::type fmt) {
         auto data_type = type_to_data_type<Type>::value;
 
-        const auto& engine = get_test_engine();
+        auto& engine = get_test_engine();
         const size_t batch_num = testing::get<0>(GetParam());
         const std::vector<size_t> in_features = testing::get<1>(GetParam());
         const size_t input_y = testing::get<2>(GetParam());
@@ -663,7 +658,7 @@ struct concat_id_conv_gpu_4d : public concat_gpu {
         topology topology;
 
         std::vector<VVVVF<Type>> in_data;
-        std::vector<memory> in_memory;
+        std::vector<memory::ptr> in_memory;
         std::vector<primitive_id> input_ids;
         for (size_t i = 0; i < in_features.size(); i++) {
             auto size = tensor(static_cast<int32_t>(batch_num),
@@ -687,7 +682,7 @@ struct concat_id_conv_gpu_4d : public concat_gpu {
                 }
             }
 
-            auto in_mem = memory::allocate(engine, in_lay);
+            auto in_mem = engine.allocate_memory(in_lay);
             set_values(in_mem, data_flat);
             in_memory.push_back(in_mem);
 
@@ -699,9 +694,11 @@ struct concat_id_conv_gpu_4d : public concat_gpu {
         topology.add(concatenation("concat", input_ids, concatenation::concatenation_axis::along_f));
         // Add identity convolution
         auto weights_lay = cldnn::layout(data_type, cldnn::format::bfyx, tensor(batch(output_f), feature(output_f)));
-        auto weights_mem = cldnn::memory::allocate(engine, weights_lay);
+        auto weights_mem = engine.allocate_memory(weights_lay);
+        weights_mem->fill(get_test_stream());
+        get_test_stream().finish();
         {
-            auto weights_ptr = weights_mem.pointer<Type>();
+            cldnn::mem_lock<Type> weights_ptr(weights_mem, get_test_stream());
             for (size_t fi = 0; fi < output_f; ++fi) {
                 auto coords = tensor(batch(fi), feature(fi), spatial(0, 0, 0, 0));
                 auto offset = weights_lay.get_linear_offset(coords);
@@ -724,7 +721,8 @@ struct concat_id_conv_gpu_4d : public concat_gpu {
         network.execute();
 
         auto out_mem = network.get_output("conv").get_memory();
-        auto out_ptr = out_mem.pointer<OutputT>();
+        cldnn::mem_lock<OutputT> out_ptr(out_mem, get_test_stream());
+        ASSERT_EQ(out_mem->get_layout().format, fmt);
 
         for (size_t bi = 0; bi < batch_num; bi++) {
             size_t f_sum = 0;
@@ -733,11 +731,11 @@ struct concat_id_conv_gpu_4d : public concat_gpu {
                     for (size_t yi = 0; yi < input_y; yi++) {
                         for (size_t xi = 0; xi < input_x; xi++) {
                             auto output_coords = tensor(batch(bi), feature(f_sum + fi), spatial(xi, yi, 0, 0));
-                            auto output_offset = out_mem.get_layout().get_linear_offset(output_coords);
+                            auto output_offset = out_mem->get_layout().get_linear_offset(output_coords);
 
                             auto ref_val = in_data[in_i][bi][fi][yi][xi];
                             auto actual_val = static_cast<Type>(out_ptr[output_offset]);
-                            EXPECT_EQ(ref_val, actual_val)
+                            ASSERT_EQ(ref_val, actual_val)
                                 << " b=" << bi << ", f=" << f_sum + fi << "(input " << in_i << "), y=" << yi << ", x=" << xi;
                         }
                     }
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/condition_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/condition_gpu_test.cpp
index 34a7c99dfb349c..c8c066e4948754 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/condition_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/condition_gpu_test.cpp
@@ -3,31 +3,26 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-
-#include <api/engine.hpp>
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/concatenation.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/pooling.hpp>
-#include <api/condition.hpp>
-#include <api/softmax.hpp>
-#include <api/scale.hpp>
-#include <api/data.hpp>
-#include "test_utils/test_utils.h"
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/concatenation.hpp>
+#include <cldnn/primitives/pooling.hpp>
+#include <cldnn/primitives/condition.hpp>
+#include <cldnn/primitives/softmax.hpp>
+#include <cldnn/primitives/scale.hpp>
+#include <cldnn/primitives/data.hpp>
 
 #include <cstddef>
 
 using namespace cldnn;
 using namespace ::tests;
 
-bool is_output_equal(const cldnn::memory& mem, const std::vector<float>& ref)
+bool is_output_equal(const cldnn::memory::ptr mem, const std::vector<float>& ref)
 {
-    auto ptr = mem.pointer<float>();
-    for (size_t i = 0; i < mem.get_layout().count(); i++)
-    {
+    cldnn::mem_lock<float> ptr(mem, get_test_stream());
+    for (size_t i = 0; i < mem->get_layout().count(); i++) {
         if (!are_equal(ptr[i], ref[i])) return false;
     }
     return true;
@@ -36,14 +31,11 @@ bool is_output_equal(const cldnn::memory& mem, const std::vector<float>& ref)
 topology generate_simple_branch (bool branch_true_false, const primitive_id& input_id)
 {
     topology branch;
-    if (branch_true_false)
-    {
+    if (branch_true_false) {
         branch.add(
             pooling(input_id + "_when_true", input_id, cldnn::pooling_mode::max, { 0, 0, 2, 1 }, { 0, 0, 2, 1 })
         );
-    }
-    else
-    {
+    } else {
         branch.add(
             pooling(input_id + "_when_false", input_id, cldnn::pooling_mode::average, { 0, 0, 2, 1 }, { 0, 0, 2, 1 })
         );
@@ -52,29 +44,29 @@ topology generate_simple_branch (bool branch_true_false, const primitive_id& inp
 }
 
 TEST(DISABLED_condition_gpu, basic_equal_comp) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options bs;
     bs.set_option(build_option::optimize_data(true));
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
-    auto compare = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto scale_mem = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
+    auto compare = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto scale_mem = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+
     topology branch_true = generate_simple_branch(true, "condi");
     topology branch_false = generate_simple_branch(false, "condi");
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("compare", compare.get_layout())
-    );    
+        input_layout("compare", compare->get_layout())
+    );
     topology.add(
-        input_layout("scale_data", scale_mem.get_layout())
+        input_layout("scale_data", scale_mem->get_layout())
     );
     topology.add(
         condition("condi", "input", branch_true, branch_false, "compare", cond_functions::EQUAL)
-    );  
+    );
     topology.add(
         scale("output", "condi", "scale_data")
     );
@@ -104,32 +96,31 @@ TEST(DISABLED_condition_gpu, basic_equal_comp) {
 }
 
 TEST(DISABLED_condition_gpu, basic_range_equal_comp) {
-
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options bs;
     bs.set_option(build_option::optimize_data(true));
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
 
-    auto compare = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 3, 1 } });
+    auto compare = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 3, 1 } });
 
     topology branch_true = generate_simple_branch(true, "condi");
     topology branch_false = generate_simple_branch(false, "condi");
 
     topology topology;
     topology.add(
-        input_layout("input0", input0.get_layout())
+        input_layout("input0", input0->get_layout())
     );
     topology.add(
-        input_layout("input1", input1.get_layout())
+        input_layout("input1", input1->get_layout())
     );
     topology.add(
-        input_layout("compare", compare.get_layout())
+        input_layout("compare", compare->get_layout())
     );
     topology.add(
         concatenation("concat", { "input0", "input1" }, concatenation::along_x)
     );
-    topology.add( 
+    topology.add(
         condition("condi", "concat", branch_true, branch_false, "compare", cond_functions::EQUAL)
     );
 
@@ -182,14 +173,10 @@ std::pair<std::vector<float>, std::vector<float>> get_values_to_compare(const cl
     std::vector<float> ret_true;
     std::vector<float> ret_false;
     auto mem_desc = generic_test::get_linear_memory_desc(input_lay);
-    for (int32_t b = 0; b < range.batch[0]; b++)
-    {
-        for (int32_t f = 0; f < range.feature[0]; f++)
-        {
-            for (int32_t y = 0; y < range.spatial[1]; y++)
-            {
-                for (int32_t x = 0; x < range.spatial[0]; x++)
-                {
+    for (int32_t b = 0; b < range.batch[0]; b++) {
+        for (int32_t f = 0; f < range.feature[0]; f++) {
+            for (int32_t y = 0; y < range.spatial[1]; y++) {
+                for (int32_t x = 0; x < range.spatial[0]; x++) {
                     auto linear_idx = generic_test::get_linear_index(
                         input_lay,
                         offset.batch[0] + b,
@@ -198,17 +185,16 @@ std::pair<std::vector<float>, std::vector<float>> get_values_to_compare(const cl
                         offset.spatial[0] + x,
                         mem_desc);
 
-                    switch (func)
-                    {
+                    switch (func) {
                     case cond_functions::EQUAL:
                         ret_true.push_back(values.at(linear_idx));
                         ret_false.push_back(-1.0f);
                         break;
-                    case cond_functions::GREATER: 
+                    case cond_functions::GREATER:
                         ret_true.push_back(values.at(linear_idx) - 1.0f);
                         ret_false.push_back(99.0f);
                         break;
-                    case cond_functions::LESS: 
+                    case cond_functions::LESS:
                         ret_true.push_back(values.at(linear_idx) + 1.0f);
                         ret_false.push_back(-1.0f);
                         break;
@@ -222,10 +208,10 @@ std::pair<std::vector<float>, std::vector<float>> get_values_to_compare(const cl
 
 TEST(DISABLED_condition_gpu, generic_test_true_false) {
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options bs;
     bs.set_option(build_option::optimize_data(true));
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 5, 2, 5, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 5, 2, 5, 1 } });
     std::vector<float> input_data(50);
     std::iota(input_data.begin(), input_data.end(), 0.0f);
 
@@ -263,17 +249,14 @@ TEST(DISABLED_condition_gpu, generic_test_true_false) {
         33, 36, 38, 41, 43, 46, 48
     };
 
-    for (auto const& func : functions)
-    {
-        for (auto const& range : ranges)
-        {
-            for (auto const& offset : offsets)
-            {
-                auto comp_values = get_values_to_compare(offset, range, input_data, input.get_layout(), func);
+    for (auto const& func : functions) {
+        for (auto const& range : ranges) {
+            for (auto const& offset : offsets) {
+                auto comp_values = get_values_to_compare(offset, range, input_data, input->get_layout(), func);
                 auto comp_values_true = comp_values.first;
                 auto comp_values_false = comp_values.second;
 
-                auto compare = memory::allocate(engine, { data_types::f32, format::bfyx, range });
+                auto compare = engine.allocate_memory({ data_types::f32, format::bfyx, range });
 
                 topology branch_true;
                 topology branch_false;
@@ -286,10 +269,10 @@ TEST(DISABLED_condition_gpu, generic_test_true_false) {
 
                 topology topology;
                 topology.add(
-                    input_layout("input", input.get_layout())
+                    input_layout("input", input->get_layout())
                 );
                 topology.add(
-                    input_layout("compare", compare.get_layout())
+                    input_layout("compare", compare->get_layout())
                 );
                 topology.add(
                     condition("condi", "input", branch_true, branch_false, "compare", func, offset)
@@ -324,7 +307,7 @@ TEST(DISABLED_condition_gpu, generic_test_true_false) {
 
 TEST(DISABLED_condition_gpu, basic_stacked_ifs) {
 
-    /*   
+    /*
         <prims...>
         <if>
         <...>
@@ -333,14 +316,14 @@ TEST(DISABLED_condition_gpu, basic_stacked_ifs) {
         <if>
         <...>
         <end_if>
-        <prims...>    
+        <prims...>
     */
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options bs;
     bs.set_option(build_option::optimize_data(true));
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
-    auto compare = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto compare2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
+    auto compare = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto compare2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 1 } });
 
     topology condi_1_true = generate_simple_branch(true, "condi");
     topology condi_1_false = generate_simple_branch(false, "condi");
@@ -355,16 +338,16 @@ TEST(DISABLED_condition_gpu, basic_stacked_ifs) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("compare", compare.get_layout())
+        input_layout("compare", compare->get_layout())
     );
     topology.add(
         condition("condi", "input", condi_1_true, condi_1_false, "compare", cond_functions::EQUAL)
     );
     topology.add(
-        input_layout("compare2", compare2.get_layout())
+        input_layout("compare2", compare2->get_layout())
     );
     topology.add(
         condition("condi2", "condi", condi_2_true, condi_2_false, "compare2", cond_functions::GREATER)
@@ -406,15 +389,15 @@ TEST(DISABLED_condition_gpu, basic_nested_ifs) {
     <end_if 0>
     <prims...>
     */
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options bs;
     bs.set_option(build_option::optimize_data(true));
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
-    auto compare = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto compare2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 1 } });
-    auto scale_5_mem = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
+    auto compare = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto compare2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 1 } });
+    auto scale_5_mem = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
     set_values(scale_5_mem, { 5.0f });
-    auto scale_10_mem = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto scale_10_mem = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
     set_values(scale_10_mem, { 10.0f });
 
     topology nested_true;
@@ -433,7 +416,7 @@ TEST(DISABLED_condition_gpu, basic_nested_ifs) {
         pooling("pooling_when_true", "condi", cldnn::pooling_mode::max, { 0, 0, 2, 1 }, { 0, 0, 2, 1 })
     );
     branch_true.add(
-        input_layout("compare2", compare2.get_layout())
+        input_layout("compare2", compare2->get_layout())
     );
 
     branch_true.add(
@@ -453,11 +436,11 @@ TEST(DISABLED_condition_gpu, basic_nested_ifs) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
 
     topology.add(
-        input_layout("compare", compare.get_layout())
+        input_layout("compare", compare->get_layout())
     );
 
     topology.add(
@@ -488,21 +471,21 @@ TEST(DISABLED_condition_gpu, basic_nested_ifs) {
 }
 
 TEST(DISABLED_condition_gpu, negative_compare_wrong_layout) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options bs;
     bs.set_option(build_option::optimize_data(true));
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
-    auto compare = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 5, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
+    auto compare = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 5, 1 } });
 
     topology branch_true = generate_simple_branch(true, "condi");
     topology branch_false = generate_simple_branch(false, "condi");
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("compare", compare.get_layout())
+        input_layout("compare", compare->get_layout())
     );
     topology.add(
         condition("condi", "input", branch_true, branch_false, "compare", cond_functions::EQUAL)
@@ -512,21 +495,21 @@ TEST(DISABLED_condition_gpu, negative_compare_wrong_layout) {
 }
 
 TEST(DISABLED_condition_gpu, negative_too_big_offset) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options bs;
     bs.set_option(build_option::optimize_data(true));
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
-    auto compare = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 3, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
+    auto compare = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 3, 1 } });
 
     topology branch_true = generate_simple_branch(true, "condi");
     topology branch_false = generate_simple_branch(false, "condi");
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("compare", compare.get_layout())
+        input_layout("compare", compare->get_layout())
     );
     topology.add(
         condition("condi", "input", branch_true, branch_false, "compare", cond_functions::EQUAL, {1, 1, 2, 1})
@@ -536,12 +519,12 @@ TEST(DISABLED_condition_gpu, negative_too_big_offset) {
 }
 
 TEST(DISABLED_condition_gpu, negative_not_same_layouts) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options bs;
     bs.set_option(build_option::optimize_data(true));
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
-    auto compare = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
+    auto compare = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+
     topology branch_true;
     branch_true.add(
         pooling("pooling_when_true", "condi", cldnn::pooling_mode::max, { 0, 0, 2, 1 }, { 0, 0, 2, 1 })
@@ -554,10 +537,10 @@ TEST(DISABLED_condition_gpu, negative_not_same_layouts) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("compare", compare.get_layout())
+        input_layout("compare", compare->get_layout())
     );
     topology.add(
         condition("condi", "input", branch_true, branch_false, "compare", cond_functions::EQUAL)
@@ -567,11 +550,11 @@ TEST(DISABLED_condition_gpu, negative_not_same_layouts) {
 }
 
 TEST(DISABLED_condition_gpu, negative_same_names_within_different_networks) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options bs;
     bs.set_option(build_option::optimize_data(true));
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
-    auto compare = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
+    auto compare = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     topology branch_true;
     branch_true.add(
@@ -585,10 +568,10 @@ TEST(DISABLED_condition_gpu, negative_same_names_within_different_networks) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("compare", compare.get_layout())
+        input_layout("compare", compare->get_layout())
     );
     topology.add(
         condition("condi", "input", branch_true, branch_false, "compare", cond_functions::EQUAL)
@@ -596,6 +579,6 @@ TEST(DISABLED_condition_gpu, negative_same_names_within_different_networks) {
     topology.add(
         pooling("pooling_check_name", "condi", cldnn::pooling_mode::max, { 0, 0, 2, 1 }, { 0, 0, 2, 1 })
     );
-    
+
     EXPECT_ANY_THROW(network net(engine, topology, bs););
 }
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/convolution_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/convolution_gpu_test.cpp
index 66855dba938b0e..658bd6625f3403 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/convolution_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/convolution_gpu_test.cpp
@@ -4,18 +4,15 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include <gtest/gtest.h>
-#include <gmock/gmock.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/convolution.hpp"
-#include "api/eltwise.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include "test_utils/float16.h"
-#include <api/data.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/convolution.hpp>
+#include <cldnn/primitives/eltwise.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/crop.hpp>
+#include <cldnn/primitives/reorder.hpp>
+
 #include <algorithm>
 #include <array>
 #include <cmath>
@@ -25,12 +22,9 @@
 #include <type_traits>
 #include <fstream>
 #include <tuple>
-#include <api/crop.hpp>
-#include <api/reorder.hpp>
-#include <src/include/to_string_utils.h>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 namespace cldnn
 {
@@ -226,29 +220,23 @@ VVF<T> reference_scale_post_op(const VVF<T>& input, const T& scale, const T& shi
     return output_shrinked;
 }
 
-void dump_buffer(memory const& mem, std::string const& name)
-{
+void dump_buffer(memory::ptr mem, std::string const& name) {
     std::ofstream out(name);
-    auto size = mem.get_layout().get_buffer_size();
-    auto ptr = mem.pointer<const float>();
-    auto pitches = mem.get_layout().get_pitches();
-    out << "Data size: " << mem.get_layout().size << "\n";
-    out << "Lower padding: " << mem.get_layout().data_padding.lower_size() << "\n";
-    out << "Upper padding: " << mem.get_layout().data_padding.upper_size() << "\n";
+    auto size = mem->get_layout().get_buffer_size();
+    cldnn::mem_lock<const float> ptr(mem, get_test_stream());
+    auto pitches = mem->get_layout().get_pitches();
+    out << "Data size: " << mem->get_layout().size << "\n";
+    out << "Lower padding: " << mem->get_layout().data_padding.lower_size() << "\n";
+    out << "Upper padding: " << mem->get_layout().data_padding.upper_size() << "\n";
     out << "\n";
 
-    for (int b = 0; b < size.batch[0]; ++b)
-    {
+    for (int b = 0; b < size.batch[0]; ++b) {
         out << " ================ BATCH " << b << " =================\n\n";
-        for (int f = 0; f < size.feature[0]; ++f)
-        {
+        for (int f = 0; f < size.feature[0]; ++f) {
             out << "feature " << f << ":\n";
-            for (int z = 0; z < size.spatial[2]; ++z)
-            {
-                for (int y = 0; y < size.spatial[1]; ++y)
-                {
-                    for (int x = 0; x < size.spatial[0]; ++x)
-                    {
+            for (int z = 0; z < size.spatial[2]; ++z) {
+                for (int y = 0; y < size.spatial[1]; ++y) {
+                    for (int x = 0; x < size.spatial[0]; ++x) {
                         size_t idx = b * pitches.batch[0] + f * pitches.feature[0] + z * pitches.spatial[2] + y * pitches.spatial[1] + x * pitches.spatial[0];
                         out << ptr[idx] << " ";
                     }
@@ -274,12 +262,12 @@ TEST(deformable_convolution_f32_fw_gpu, basic_deformable_convolution_def_group1_
     //  Group    : 1
     //  Def_group: 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 4, 4 } });
-    auto trans = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 18, 4, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 4, 4, 3, 3 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 4, 4 } });
+    auto trans = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 18, 4, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 4, 4, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 1, 1 } });
 
     set_values(input, { 0.680375f, -0.211234f, 0.566198f, 0.59688f, 0.823295f, -0.604897f, -0.329554f, 0.536459f,
                         -0.444451f, 0.10794f, -0.0452059f, 0.257742f, -0.270431f, 0.0268018f, 0.904459f, 0.83239f,
@@ -350,8 +338,8 @@ TEST(deformable_convolution_f32_fw_gpu, basic_deformable_convolution_def_group1_
             -2.6067f, 0.562893f, 0.671884f, 0.404735f, 1.45044f, 0.950113f };
 
     topology topology(
-            input_layout("input", input.get_layout()),
-            input_layout("trans", trans.get_layout()),
+            input_layout("input", input->get_layout()),
+            input_layout("trans", trans->get_layout()),
             data("weights", weights),
             data("biases", biases),
             convolution(
@@ -377,8 +365,8 @@ TEST(deformable_convolution_f32_fw_gpu, basic_deformable_convolution_def_group1_
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -406,12 +394,12 @@ TEST(deformable_convolution_f32_fw_gpu, basic_deformable_convolution_def_group1)
     //  Group    : 1
     //  Def_group: 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 4, 4 } });
-    auto trans = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 18, 4, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 4, 4, 3, 3 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 4, 4 } });
+    auto trans = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 18, 4, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 4, 4, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 1, 1 } });
 
     set_values(input, { 0.680375f, -0.211234f, 0.566198f, 0.59688f, 0.823295f, -0.604897f, -0.329554f, 0.536459f,
                         -0.444451f, 0.10794f, -0.0452059f, 0.257742f, -0.270431f, 0.0268018f, 0.904459f, 0.83239f,
@@ -482,8 +470,8 @@ TEST(deformable_convolution_f32_fw_gpu, basic_deformable_convolution_def_group1)
             -0.210789f, -0.973089f, -0.407542f, 1.11818f, 0.843776f, 0.628229f, 1.29095f, 1.18637f, 0.808982f, 1.43841f };
 
     topology topology(
-            input_layout("input", input.get_layout()),
-            input_layout("trans", trans.get_layout()),
+            input_layout("input", input->get_layout()),
+            input_layout("trans", trans->get_layout()),
             data("weights", weights),
             data("biases", biases),
             convolution(
@@ -509,8 +497,8 @@ TEST(deformable_convolution_f32_fw_gpu, basic_deformable_convolution_def_group1)
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -538,12 +526,12 @@ TEST(deformable_convolution_f32_fw_gpu, basic_deformable_convolution) {
     //  Group    : 1
     //  Def_group: 2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 4, 4 } });
-    auto trans = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 36, 4, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 4, 4, 3, 3 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 4, 4 } });
+    auto trans = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 36, 4, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 4, 4, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 1, 1 } });
 
     set_values(input, { 0.680375f, -0.211234f, 0.566198f, 0.59688f, 0.823295f, -0.604897f, -0.329554f, 0.536459f,
                         -0.444451f, 0.10794f, -0.0452059f, 0.257742f, -0.270431f, 0.0268018f, 0.904459f, 0.83239f,
@@ -646,8 +634,8 @@ TEST(deformable_convolution_f32_fw_gpu, basic_deformable_convolution) {
             1.18929f, 0.382556f, 0.499048f, 1.16212f, 1.62688f, 1.31246f, 1.82684f };
 
     topology topology(
-            input_layout("input", input.get_layout()),
-            input_layout("trans", trans.get_layout()),
+            input_layout("input", input->get_layout()),
+            input_layout("trans", trans->get_layout()),
             data("weights", weights),
             data("biases", biases),
             convolution(
@@ -673,8 +661,8 @@ TEST(deformable_convolution_f32_fw_gpu, basic_deformable_convolution) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -711,10 +699,10 @@ TEST(convolution_f32_fw_gpu, basic_convolution_no_bias) {
     // 21  28  39
     // 18  20  20
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32,format::yxfb,{ 1, 1, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 3, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32,format::yxfb,{ 1, 1, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 3, 2 } });
 
     set_values(input, { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 2.0f, 2.0f, 3.0f, 4.0f, 6.0f, 3.0f, 3.0f, 3.0f, 5.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f });
     set_values(weights, { 1.0f, 2.0f, 1.0f, 2.0f, 1.0f, 2.0f });
@@ -723,7 +711,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution_no_bias) {
         { 17.0f, 19.0f, 19.0f } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         convolution("conv", "input", { "weights" }, { 1,1,1,2 }));
 
@@ -735,8 +723,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution_no_bias) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -782,10 +770,10 @@ TEST(convolution_f32_fw_gpu, basic_convolution_int8_no_bias) {
     // 21  28  39
     // 18  20  20
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::i8,format::bfyx,{ 1, 1, 3, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::i8,format::bfyx,{ 1, 1, 3, 2 } });
 
     set_values(input, { 1.1f, 2.4f, 3.5f, 4.5f, 5.8f,
                         2.9f, 2.3f, 3.5f, 4.4f, 6.6f,
@@ -798,7 +786,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution_int8_no_bias) {
         { 22.0f, 20.0f, 21.0f } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("to_int","input", { data_types::i8,format::bfyx,{ 1, 1, 5, 4 } }),
         data("weights", weights),
         convolution("conv", "to_int", { "weights" }, { 1,1,1,2 }),
@@ -812,8 +800,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution_int8_no_bias) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -839,10 +827,10 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D_no_bias) {
     //  Input  : 4x5x1
     //  Output : 2x3x1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 3, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 3, 2 } });
 
     set_values(input, { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 2.0f, 2.0f, 3.0f, 4.0f, 6.0f, 3.0f, 3.0f, 3.0f, 5.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f });
     set_values(weights, { 1.0f, 2.0f, 1.0f, 2.0f, 1.0f, 2.0f });
@@ -852,7 +840,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D_no_bias) {
         { 17.0f, 19.0f, 19.0f } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         convolution("conv", "input", { "weights" }, { 1,1,1,2 }));
 
@@ -864,8 +852,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D_no_bias) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int z_size = output_layout.size.spatial[2];
     int y_size = output_layout.size.spatial[1];
@@ -934,11 +922,11 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D) {
     //  Bias:
     //  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 1, 4, 4, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 1, 2, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 1, 4, 4, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 1, 2, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1, 1 } });
 
     set_values(input, {
         1.0f,  0.0f,  1.0f,  0.0f,
@@ -987,7 +975,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D) {
     };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", { "weights" }, { "biases" }));
@@ -1000,8 +988,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int z_size = output_layout.size.spatial[2];
     int y_size = output_layout.size.spatial[1];
@@ -1026,10 +1014,10 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D) {
 
 TEST(convolution_f32_fw_gpu, basic_convolution3D_split2) {
     //  data is similar as in basic_convolution3D
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 2, 4, 4, 4 } });
-    auto weights_1 = memory::allocate(engine, { data_types::f32, format::goizyx, tensor(cldnn::group(2), cldnn::batch(1), cldnn::feature(1), cldnn::spatial(2, 2, 2))});
-    auto biases_1 = memory::allocate(engine, { data_types::f32, format::bfyx, tensor(feature(2)) });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 2, 4, 4, 4 } });
+    auto weights_1 = engine.allocate_memory({ data_types::f32, format::goizyx, tensor(cldnn::group(2), cldnn::batch(1), cldnn::feature(1), cldnn::spatial(2, 2, 2))});
+    auto biases_1 = engine.allocate_memory({ data_types::f32, format::bfyx, tensor(feature(2)) });
 
     set_values(input, {
         1.0f,  0.0f,  1.0f,  0.0f,
@@ -1117,7 +1105,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D_split2) {
     };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights_1", weights_1),
         data("biases_1", biases_1),
         convolution("conv", "input", { "weights_1" }, { "biases_1" }, 2, tensor(1), tensor(0), tensor(1), tensor{1, 2, 3, 3, 3}, data_types::f32, true));
@@ -1130,8 +1118,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D_split2) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int z_size = output_layout.size.spatial[2];
     int y_size = output_layout.size.spatial[1];
@@ -1158,10 +1146,10 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D_split2) {
 
 TEST(convolution_f32_fw_gpu, basic_convolution3D_group2) {
     //  data is similar as in basic_convolution3D_split2
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 2, 4, 4, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 1, 2, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 2, 4, 4, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 1, 2, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1, 1 } });
 
     set_values(input, {
         1.0f,  0.0f,  1.0f,  0.0f,
@@ -1249,7 +1237,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D_group2) {
     };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", { "weights" }, { "biases" }));
@@ -1262,8 +1250,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D_group2) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int z_size = output_layout.size.spatial[2];
     int y_size = output_layout.size.spatial[1];
@@ -1289,15 +1277,14 @@ TEST(convolution_f32_fw_gpu, basic_convolution3D_group2) {
 }
 
 TEST(convolution_f32_fw_gpu, with_output_size_same_input) {
+    auto& engine = get_test_engine();
 
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 320, 320 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 64, 4, 7, 7 } });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 64, 4, 7, 7 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 320, 320 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 64, 4, 7, 7 } });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 64, 4, 7, 7 } });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("weights2", weights2),
         convolution::create_with_output_size("conv1", "input", { "weights" }, {1, 64, 160, 160}, {1, 1, 2, 2}, {0, 0, -3, -3}),
@@ -1329,16 +1316,16 @@ TEST(convolution_f32_fw_gpu, three_convolutions_same_weights) {
     //  8  8   8  8
     //  8  8   8  8
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, {1,2,2,2} });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 2,2,1,1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, {1,2,2,2} });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 2,2,1,1 } });
 
     set_values(input, { 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f });
     set_values(weights, { 1.0f, 1.0f, 1.0f, 1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         convolution("conv1", "input", { "weights" }),
         convolution("conv2", "conv1", { "weights" }),
@@ -1353,8 +1340,8 @@ TEST(convolution_f32_fw_gpu, three_convolutions_same_weights) {
     auto outputs = network.execute();
 
     auto output_memory = outputs.at("conv3").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -1397,11 +1384,11 @@ TEST(convolution_f32_fw_gpu, basic_convolution) {
     //  Bias:
     //  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 3, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 3, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 2.0f, 2.0f, 3.0f, 4.0f, 6.0f, 3.0f, 3.0f, 3.0f, 5.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f });
     set_values(weights, { 1.0f, 2.0f, 1.0f, 2.0f, 1.0f, 2.0f });
@@ -1411,7 +1398,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution) {
         { 18.0f, 20.0f, 20.0f } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution( "conv", "input", { "weights" }, { "biases" }, { 0,0,1,2 }));
@@ -1424,8 +1411,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -1445,14 +1432,14 @@ TEST(convolution_f32_fw_gpu, basic_convolution) {
 
 TEST(convolution_f32_fw_gpu, basic_convolution_bfyx_weights_as_input_layout) {
     //Same params as convolution_f32_fw_gpu, basic_convolution but with bfyx optimized data and weights set as input_layout
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,
     { 1, 1, 5, 4 }
     });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx,
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx,
     { 1, 1, 3, 2 }
     });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,
     { 1, 1, 1, 1 }
     });
     set_values(input,
@@ -1470,9 +1457,9 @@ TEST(convolution_f32_fw_gpu, basic_convolution_bfyx_weights_as_input_layout) {
         { 18.0f, 20.0f, 20.0f }
     };
     topology topology(
-        input_layout("input", input.get_layout()),
-        input_layout("weights", weights.get_layout()),
-        input_layout("biases", biases.get_layout()),
+        input_layout("input", input->get_layout()),
+        input_layout("weights", weights->get_layout()),
+        input_layout("biases", biases->get_layout()),
         convolution("conv", "input",
         { "weights" }
             ,
@@ -1491,8 +1478,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution_bfyx_weights_as_input_layout) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -1542,11 +1529,11 @@ TEST(convolution_f32_fw_gpu, basic_convolution_input_padding) {
     //  Bias:
     //  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 4, 3 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 4, 3 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 1.0f, 2.0f, 3.0f, 4.0f, 2.0f, 2.0f, 3.0f, 4.0f, 3.0f, 3.0f, 3.0f, 5.0f });
     set_values(weights, { 1.0f, 1.0f, 1.0f, 1.0f });
@@ -1560,7 +1547,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution_input_padding) {
         { 1.0f, 1.0f, 1.0f, 1.0f, 1.0f } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -1582,8 +1569,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution_input_padding) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -1642,11 +1629,11 @@ TEST(convolution_f32_fw_gpu, basic_convolution_sym_input_padding) {
     //  Bias:
     //  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 4, 3 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 4, 3 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { 1.0f, 2.0f, 3.0f, 4.0f, 2.0f, 2.0f, 3.0f, 4.0f, 3.0f, 3.0f, 3.0f, 5.0f });
     set_values(weights, { 1.0f, 1.0f, 1.0f, 1.0f });
@@ -1660,7 +1647,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution_sym_input_padding) {
         { 1.0f, 1.0f, 1.0f, 1.0f, 1.0f } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -1684,8 +1671,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution_sym_input_padding) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -1738,11 +1725,11 @@ TEST(convolution_f32_fw_gpu, basic_convolution_asym_input_padding) {
     //  Bias:
     //  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 4, 3 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 4, 3 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { 1.0f, 2.0f, 3.0f, 4.0f, 2.0f, 2.0f, 3.0f, 4.0f, 3.0f, 3.0f, 3.0f, 5.0f });
     set_values(weights, { 1.0f, 1.0f, 1.0f, 1.0f });
@@ -1757,7 +1744,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution_asym_input_padding) {
         { 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -1781,8 +1768,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution_asym_input_padding) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -1842,11 +1829,11 @@ TEST(convolution_f32_fw_gpu, basic_convolution_sym_input_padding_with_input_offs
     //  Bias:
     //  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 4, 3 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 4, 3 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { 1.0f, 2.0f, 3.0f, 4.0f, 2.0f, 2.0f, 3.0f, 4.0f, 3.0f, 3.0f, 3.0f, 5.0f });
     set_values(weights, { 1.0f, 1.0f, 1.0f, 1.0f });
@@ -1864,7 +1851,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution_sym_input_padding_with_input_offs
         { 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -1888,8 +1875,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution_sym_input_padding_with_input_offs
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -1951,11 +1938,11 @@ TEST(convolution_f32_fw_gpu, basic_convolution_asym_input_padding_with_input_off
     //  Bias:
     //  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 4, 3 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 4, 3 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { 1.0f, 2.0f, 3.0f, 4.0f, 2.0f, 2.0f, 3.0f, 4.0f, 3.0f, 3.0f, 3.0f, 5.0f });
     set_values(weights, { 1.0f, 1.0f, 1.0f, 1.0f });
@@ -1974,7 +1961,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution_asym_input_padding_with_input_off
         { 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -1998,8 +1985,8 @@ TEST(convolution_f32_fw_gpu, basic_convolution_asym_input_padding_with_input_off
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -2052,11 +2039,11 @@ TEST(convolution_f32_fw_gpu, basic_convolution_input_and_output_padding) {
     //  Bias:
     //  1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 4, 3 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 4, 3 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 1.0f, 2.0f, 3.0f, 4.0f, 2.0f, 2.0f, 3.0f, 4.0f, 3.0f, 3.0f, 3.0f, 5.0f });
     set_values(weights, { 1.0f, 1.0f, 1.0f, 1.0f });
@@ -2074,7 +2061,7 @@ TEST(convolution_f32_fw_gpu, basic_convolution_input_and_output_padding) {
     const int x_pad = 2;
     const int y_pad = 1;
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -2096,9 +2083,9 @@ TEST(convolution_f32_fw_gpu, basic_convolution_input_and_output_padding) {
     EXPECT_EQ(outputs.begin()->first, "conv");
 
     auto output_memory = outputs.at("conv").get_memory();
-    auto output_layout = output_memory.get_layout();
+    auto output_layout = output_memory->get_layout();
     auto output_size = output_layout.get_buffer_size();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_size.spatial[1];
     int x_size = output_size.spatial[0];
@@ -2165,19 +2152,19 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x1x1_nopad_random) {
     }
     VF<float> output_rnd_vec = flatten_4d<float>(format::yxfb, output_rnd);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32,  format::yxfb, { 1, 1, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32,  format::yxfb, { 1, 1, 4, 4 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1,{ 2, 2 }, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32,  format::bfyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32,  format::bfyx, { 1, 1, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32,  format::bfyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32,  format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, input_rnd_vec);
     set_values(weights, filter_rnd_vec);
     set_values(biases, bias_rnd);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", {"weights"}, {"biases"}, {1,1,2,2})
@@ -2192,7 +2179,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x1x1_nopad_random) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     for (size_t i = 0; i < output_rnd.size(); ++i) {
         float x = float_round(output_rnd_vec[i]), y = float_round(output_ptr[i]);
@@ -2235,19 +2222,19 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in2x2x1x2_nopad_random) {
     }
     VF<float> output_rnd_vec = flatten_4d<float>(format::yxfb, output_rnd);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 1, 2, 2 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 2,{ 1, 1 }, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, input_rnd_vec);
     set_values(weights, filter_rnd_vec);
     set_values(biases, bias_rnd);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", { "weights" }, { "biases" }, { 1,1,2,2 })
@@ -2262,7 +2249,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in2x2x1x2_nopad_random) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     for (size_t i = 0; i < output_rnd.size(); ++i) {
         float x = float_round(output_rnd_vec[i]), y = float_round(output_ptr[i]);
@@ -2293,19 +2280,19 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x1x1_nopad) {
     //  8  0.5
     //  6  9
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 4, 4 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1,{ 2, 2 }, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { -0.5f, 1.0f, 0.5f, 2.0f, 1.5f, -0.5f, 0.0f, -1.0f, 0.5f, 0.5f, -1.0f, 1.0f, 0.5f, 2.0f, 1.5f, -0.5f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f });
     set_values(biases, { 2.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", { "weights" }, { "biases" }, { 1,1,2,2 })
@@ -2320,7 +2307,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     EXPECT_FLOAT_EQ(8.0f, output_ptr[0]);
     EXPECT_FLOAT_EQ(0.5f, output_ptr[1]);
@@ -2347,19 +2334,19 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in2x2x1x2_nopad) {
     //
     //  Output:
     //  3.65 -5.36
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 1, 2, 2 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 2,{ 1, 1 }, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 0.5f, 2.3f, 1.5f, -0.4f, 2.0f, 1.0f, -4.0f, 3.0f });
     set_values(weights, { -1.2f, 1.5f, 0.5f, -0.5f });
     set_values(biases, { -1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", { "weights" }, { "biases" }, { 1,1,2,2 } )
@@ -2374,7 +2361,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in2x2x1x2_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     EXPECT_FLOAT_EQ(3.65f, output_ptr[0]);
     EXPECT_FLOAT_EQ(-5.36f, output_ptr[1]);
@@ -2399,19 +2386,19 @@ TEST(convolution_f32_fw_gpu, basic_ofm_wsiz2x1x2x1_in1x2x1_nopad) {
     //   5.1  f=0
     //  -5.2  f=1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 1, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 1, 2 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1 ,{ 1, 1 }, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, { 1.0f, 2.0f });
     set_values(weights, { 1.0f, 2.0f, -1.0f, -2.0f });
     set_values(biases, { 0.1f, -0.2f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", { "weights" }, { "biases" }, { 1,1,5,5 })
@@ -2426,7 +2413,7 @@ TEST(convolution_f32_fw_gpu, basic_ofm_wsiz2x1x2x1_in1x2x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     EXPECT_FLOAT_EQ(5.1f, output_ptr[0]);
     EXPECT_FLOAT_EQ(-5.2f, output_ptr[1]);
@@ -2458,19 +2445,19 @@ TEST(convolution_f32_fw_gpu, basic_ofm_wsiz3x2x2x1_in2x2x1_nopad) {
     //   64,0  f=1
     //  103.0  f=2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 2, 1, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 2, 1, 2 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1 ,{ 1, 1 }, 3 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 2, 1, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 3, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 2, 1, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 3, 1, 1 } });
 
     set_values(input, { 1.0f, 3.0f, 2.0f, 4.0f });
     set_values(weights, { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f, 9.0f, 10.0f, 11.0f, 12.0f });
     set_values(biases, { -5.0f, -6.0f, -7.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", { "weights" }, { "biases" }, { 1,1,5,5 })
@@ -2485,7 +2472,7 @@ TEST(convolution_f32_fw_gpu, basic_ofm_wsiz3x2x2x1_in2x2x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     EXPECT_FLOAT_EQ(25.0f, output_ptr[0]);
     EXPECT_FLOAT_EQ(64.0f, output_ptr[1]);
@@ -2514,19 +2501,19 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2x1x3_wstr2x2_in2x2x1x1_nopad) {
     //   2.12
     //   3.08
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1 ,{ 1, 1 }, 3 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 3, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 3, 1, 1 } });
 
     set_values(input, { -2.3f, -0.1f, 3.1f, 1.9f });
     set_values(weights, { -1.1f, 1.5f, 0.5f, -0.5f, 0.1f, 0.2f, 0.4f, 0.7f, 2.0f, -1.0f, 2.5f, -1.5f });
     set_values(biases, { 0.1f, -0.2f, 0.3f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", { "weights" }, { "biases" }, { 1,1,2,2 })
@@ -2541,7 +2528,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2x1x3_wstr2x2_in2x2x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     EXPECT_TRUE(are_equal(3.08f, output_ptr[0]));
     EXPECT_TRUE(are_equal(2.12f, output_ptr[1]));
@@ -2570,19 +2557,19 @@ TEST(convolution_f32_fw_gpu, wsiz3x3_wstr2x2_in2x2x1x1_zeropad) {
     //
     //  Output:
     //  12.25
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1,{ 1, 1 }, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 3, 3 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { -0.5f, 1.0f, 0.5f, 2.0f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f, 4.0f, -5.0f, 0.5f, 1.5f, -1.5f });
     set_values(biases, { 2.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", { "weights" }, { "biases" }, { 1,1,2,2 })
@@ -2597,7 +2584,7 @@ TEST(convolution_f32_fw_gpu, wsiz3x3_wstr2x2_in2x2x1x1_zeropad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     EXPECT_FLOAT_EQ(12.25f, output_ptr[0]);
 }
@@ -2627,19 +2614,19 @@ TEST(convolution_f32_fw_gpu, offsets_wsiz3x3_wstr2x2_in2x2x1x1_zeropad) {
     //   Output:
     //   rnd   rnd
     //   rnd   2.0
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1 ,{ 2, 2 }, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 3, 3 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { -0.5f, 1.0f, 0.5f, 2.0f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f, 4.0f, -5.0f, 0.5f, 1.5f, -1.5f });
     set_values(biases, { 2.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -2662,7 +2649,7 @@ TEST(convolution_f32_fw_gpu, offsets_wsiz3x3_wstr2x2_in2x2x1x1_zeropad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     EXPECT_FLOAT_EQ(-7.25f, output_ptr[4]);
 }
@@ -2702,12 +2689,12 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_split2) {
     //   8  3.65 0.5 -5.36
     //   6  3.65 9   -5.36
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 2, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 2, 4, 4 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1,{ 2, 2 }, 2 } });
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2,2))});
-    auto biases1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2,2))});
+    auto biases1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, {
         -0.5f,  0.5f,  1.0f,  1.5f,  0.5f,  2.3f,  2.0f, -0.4f,
@@ -2719,7 +2706,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_split2) {
     set_values(biases1, { 2.0f, -1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights1", weights1),
         data("biases1", biases1),
         convolution(
@@ -2742,16 +2729,16 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_split2) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
-    EXPECT_FLOAT_EQ(8.0f,   get_value<float>(output_ptr, 0));
-    EXPECT_FLOAT_EQ(3.65f,  get_value<float>(output_ptr, 1));
-    EXPECT_FLOAT_EQ(0.5f,   get_value<float>(output_ptr, 2));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 3));
-    EXPECT_FLOAT_EQ(6.0f,   get_value<float>(output_ptr, 4));
-    EXPECT_FLOAT_EQ(3.65f,  get_value<float>(output_ptr, 5));
-    EXPECT_FLOAT_EQ(9.0f,   get_value<float>(output_ptr, 6));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 7));
+    EXPECT_FLOAT_EQ(8.0f,   output_ptr[0]);
+    EXPECT_FLOAT_EQ(3.65f,  output_ptr[1]);
+    EXPECT_FLOAT_EQ(0.5f,   output_ptr[2]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[3]);
+    EXPECT_FLOAT_EQ(6.0f,   output_ptr[4]);
+    EXPECT_FLOAT_EQ(3.65f,  output_ptr[5]);
+    EXPECT_FLOAT_EQ(9.0f,   output_ptr[6]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[7]);
 }
 
 TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2) {
@@ -2800,12 +2787,12 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2) {
     //   8  8 3.65 3.65 0.5  0.5 -5.36 -5.36
     //   6  6 3.65 3.65 9    9   -5.36 -5.36
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 4, 4 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 2,{ 2, 2 }, 2 } });
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2,2)) });
-    auto biases1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2,2)) });
+    auto biases1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, {
        -0.5f, -0.5f,  0.5f,  0.5f,  1.0f,  1.0f,  1.5f,  1.5f,  0.5f,  0.5f,  2.3f,  2.3f,  2.0f,  2.0f, -0.4f, -0.4f,
@@ -2817,7 +2804,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2) {
     set_values(biases1, { 2.0f, -1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights1", weights1),
         data("biases1", biases1),
         convolution(
@@ -2840,33 +2827,33 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
-
-    EXPECT_FLOAT_EQ(8.0f,   get_value<float>(output_ptr, 0));
-    EXPECT_FLOAT_EQ(8.0f,   get_value<float>(output_ptr, 1));
-    EXPECT_FLOAT_EQ(3.65f,  get_value<float>(output_ptr, 2));
-    EXPECT_FLOAT_EQ(3.65f,  get_value<float>(output_ptr, 3));
-    EXPECT_FLOAT_EQ(0.5f,   get_value<float>(output_ptr, 4));
-    EXPECT_FLOAT_EQ(0.5f,   get_value<float>(output_ptr, 5));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 6));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 7));
-    EXPECT_FLOAT_EQ(6.0f,   get_value<float>(output_ptr, 8));
-    EXPECT_FLOAT_EQ(6.0f,   get_value<float>(output_ptr, 9));
-    EXPECT_FLOAT_EQ(3.65f,  get_value<float>(output_ptr, 10));
-    EXPECT_FLOAT_EQ(3.65f,  get_value<float>(output_ptr, 11));
-    EXPECT_FLOAT_EQ(9.0f,   get_value<float>(output_ptr, 12));
-    EXPECT_FLOAT_EQ(9.0f,   get_value<float>(output_ptr, 13));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 14));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 15));
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
+
+    EXPECT_FLOAT_EQ(8.0f,   output_ptr[0]);
+    EXPECT_FLOAT_EQ(8.0f,   output_ptr[1]);
+    EXPECT_FLOAT_EQ(3.65f,  output_ptr[2]);
+    EXPECT_FLOAT_EQ(3.65f,  output_ptr[3]);
+    EXPECT_FLOAT_EQ(0.5f,   output_ptr[4]);
+    EXPECT_FLOAT_EQ(0.5f,   output_ptr[5]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[6]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[7]);
+    EXPECT_FLOAT_EQ(6.0f,   output_ptr[8]);
+    EXPECT_FLOAT_EQ(6.0f,   output_ptr[9]);
+    EXPECT_FLOAT_EQ(3.65f,  output_ptr[10]);
+    EXPECT_FLOAT_EQ(3.65f,  output_ptr[11]);
+    EXPECT_FLOAT_EQ(9.0f,   output_ptr[12]);
+    EXPECT_FLOAT_EQ(9.0f,   output_ptr[13]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[14]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[15]);
 }
 
 TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_group2) {
     //  data is similar as in basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_split2
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 2, 4, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx ,tensor(group(2), batch(1), feature(1), spatial(2,2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 2, 4, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx ,tensor(group(2), batch(1), feature(1), spatial(2,2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values(input, {
         -0.5f,  0.5f,  1.0f,  1.5f,  0.5f,  2.3f,  2.0f, -0.4f,
@@ -2881,7 +2868,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_group2) {
     set_values(biases, { 2.0f, -1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -2904,26 +2891,26 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_group2) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
-    EXPECT_FLOAT_EQ(8.0f, get_value<float>(output_ptr, 0));
-    EXPECT_FLOAT_EQ(3.65f, get_value<float>(output_ptr, 1));
-    EXPECT_FLOAT_EQ(0.5f, get_value<float>(output_ptr, 2));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 3));
-    EXPECT_FLOAT_EQ(6.0f, get_value<float>(output_ptr, 4));
-    EXPECT_FLOAT_EQ(3.65f, get_value<float>(output_ptr, 5));
-    EXPECT_FLOAT_EQ(9.0f, get_value<float>(output_ptr, 6));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 7));
+    EXPECT_FLOAT_EQ(8.0f, output_ptr[0]);
+    EXPECT_FLOAT_EQ(3.65f, output_ptr[1]);
+    EXPECT_FLOAT_EQ(0.5f, output_ptr[2]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[3]);
+    EXPECT_FLOAT_EQ(6.0f, output_ptr[4]);
+    EXPECT_FLOAT_EQ(3.65f, output_ptr[5]);
+    EXPECT_FLOAT_EQ(9.0f, output_ptr[6]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[7]);
 }
 
 TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_group2_bfyx) {
     //  data is similar as in basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_split2
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 2, 4, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx ,tensor(group(2), batch(1), feature(1), spatial(2,2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 2, 4, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx ,tensor(group(2), batch(1), feature(1), spatial(2,2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values(input, {
         -0.5f,  0.5f,  1.0f,  1.5f,  0.5f,  2.3f,  2.0f, -0.4f,
@@ -2938,7 +2925,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_group2_bfyx)
     set_values(biases, { 2.0f, -1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("input_1", "input", { data_types::f32,format::bfyx,{ 1, 2, 4, 4 } }),
         data("weights", weights),
         data("biases", biases),
@@ -2962,26 +2949,26 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x1_nopad_group2_bfyx)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
-    EXPECT_FLOAT_EQ(8.0f, get_value<float>(output_ptr, 0));
-    EXPECT_FLOAT_EQ(0.5f, get_value<float>(output_ptr, 1));
-    EXPECT_FLOAT_EQ(6.0f, get_value<float>(output_ptr, 2));
-    EXPECT_FLOAT_EQ(9.0f, get_value<float>(output_ptr, 3));
-    EXPECT_FLOAT_EQ(3.65f, get_value<float>(output_ptr, 4));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 5));
-    EXPECT_FLOAT_EQ(3.65f, get_value<float>(output_ptr, 6));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 7));
+    EXPECT_FLOAT_EQ(8.0f, output_ptr[0]);
+    EXPECT_FLOAT_EQ(0.5f, output_ptr[1]);
+    EXPECT_FLOAT_EQ(6.0f, output_ptr[2]);
+    EXPECT_FLOAT_EQ(9.0f, output_ptr[3]);
+    EXPECT_FLOAT_EQ(3.65f, output_ptr[4]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[5]);
+    EXPECT_FLOAT_EQ(3.65f, output_ptr[6]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[7]);
 }
 
 TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group2) {
     //  data is similar as in basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 4, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx ,tensor(group(2), batch(1), feature(1), spatial(2,2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 4, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx ,tensor(group(2), batch(1), feature(1), spatial(2,2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values(input, {
         -0.5f, -0.5f,  0.5f,  0.5f,  1.0f,  1.0f,  1.5f,  1.5f,  0.5f,  0.5f,  2.3f,  2.3f,  2.0f,  2.0f, -0.4f, -0.4f,
@@ -2996,7 +2983,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group2) {
     set_values(biases, { 2.0f, -1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -3019,33 +3006,33 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group2) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
-
-    EXPECT_FLOAT_EQ(8.0f, get_value<float>(output_ptr, 0));
-    EXPECT_FLOAT_EQ(8.0f, get_value<float>(output_ptr, 1));
-    EXPECT_FLOAT_EQ(3.65f, get_value<float>(output_ptr, 2));
-    EXPECT_FLOAT_EQ(3.65f, get_value<float>(output_ptr, 3));
-    EXPECT_FLOAT_EQ(0.5f, get_value<float>(output_ptr, 4));
-    EXPECT_FLOAT_EQ(0.5f, get_value<float>(output_ptr, 5));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 6));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 7));
-    EXPECT_FLOAT_EQ(6.0f, get_value<float>(output_ptr, 8));
-    EXPECT_FLOAT_EQ(6.0f, get_value<float>(output_ptr, 9));
-    EXPECT_FLOAT_EQ(3.65f, get_value<float>(output_ptr, 10));
-    EXPECT_FLOAT_EQ(3.65f, get_value<float>(output_ptr, 11));
-    EXPECT_FLOAT_EQ(9.0f, get_value<float>(output_ptr, 12));
-    EXPECT_FLOAT_EQ(9.0f, get_value<float>(output_ptr, 13));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 14));
-    EXPECT_FLOAT_EQ(-5.36f, get_value<float>(output_ptr, 15));
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
+
+    EXPECT_FLOAT_EQ(8.0f, output_ptr[0]);
+    EXPECT_FLOAT_EQ(8.0f, output_ptr[1]);
+    EXPECT_FLOAT_EQ(3.65f, output_ptr[2]);
+    EXPECT_FLOAT_EQ(3.65f, output_ptr[3]);
+    EXPECT_FLOAT_EQ(0.5f, output_ptr[4]);
+    EXPECT_FLOAT_EQ(0.5f, output_ptr[5]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[6]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[7]);
+    EXPECT_FLOAT_EQ(6.0f, output_ptr[8]);
+    EXPECT_FLOAT_EQ(6.0f, output_ptr[9]);
+    EXPECT_FLOAT_EQ(3.65f, output_ptr[10]);
+    EXPECT_FLOAT_EQ(3.65f, output_ptr[11]);
+    EXPECT_FLOAT_EQ(9.0f, output_ptr[12]);
+    EXPECT_FLOAT_EQ(9.0f, output_ptr[13]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[14]);
+    EXPECT_FLOAT_EQ(-5.36f, output_ptr[15]);
 }
 
 TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthwise_sep_opt) {
     //  Test for depthwise separable optimization, there are 16 weights and biases (split 16)
     //  data is similar as in basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2 but with batch 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 16, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 16, 4, 4 } });
 
     set_values(input, {
         -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f,
@@ -3066,8 +3053,8 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthw
         -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f,
     });
 
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::goiyx ,tensor(group(16), batch(1), feature(1), spatial(2,2)) });
-    auto biases1 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 16, 1, 1 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::goiyx ,tensor(group(16), batch(1), feature(1), spatial(2,2)) });
+    auto biases1 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 16, 1, 1 } });
 
     set_values(weights1, { -2.0f, 0.5f, 3.5f, 1.5f,  -1.2f, 1.5f, 0.5f, -0.5f,
                            -2.0f, 0.5f, 3.5f, 1.5f,  -1.2f, 1.5f, 0.5f, -0.5f,
@@ -3091,7 +3078,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthw
     primitive_id bias_id = "biases1";
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data(weights_id, weights1),
         data(bias_id, biases1),
         convolution(
@@ -3114,7 +3101,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthw
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f,
@@ -3132,9 +3119,9 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthw
 TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthwise_sep_opt_bfyx) {
     //  Test for depthwise separable optimization, there are 16 weights and biases (split 16)
     //  data is similar as in basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2 but with batch 1
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 16, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 16, 4, 4 } });
 
     set_values(input, {
         -0.5f, 1.0f, 0.5f, 2.0f, 1.5f, -0.5f, 0.0f, -1.0f, 0.5f, 0.5f, -1.0f, 1.0f, 0.5f, 2.0f, 1.5f, -0.5f,
@@ -3155,8 +3142,8 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthw
         0.5f, 1.5f, 2.3f, -0.4f, 2.0f, -4.0f, 1.0f, 3.0f, 0.5f, 1.5f, 2.3f, -0.4f, 2.0f, -4.0f, 1.0f, 3.0f,
     });
 
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::goiyx ,tensor(group(16), batch(1), feature(1), spatial(2,2)) });
-    auto biases1 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 16, 1, 1 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::goiyx ,tensor(group(16), batch(1), feature(1), spatial(2,2)) });
+    auto biases1 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 16, 1, 1 } });
 
     set_values(weights1, { -2.0f, 0.5f, 3.5f, 1.5f,  -1.2f, 1.5f, 0.5f, -0.5f,
                            -2.0f, 0.5f, 3.5f, 1.5f,  -1.2f, 1.5f, 0.5f, -0.5f,
@@ -3180,7 +3167,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthw
     primitive_id bias_id = "biases1";
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data(weights_id, weights1),
             data(bias_id, biases1),
             convolution(
@@ -3203,7 +3190,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthw
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         8.0f, 0.5f,  6.0f,  9.0f, 3.65f,-5.36f, 3.65f, -5.36f,
@@ -3226,9 +3213,9 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group16) {
     //  Test for grouped convolution, there are 16 joined weights and biases (group 16)
     //  data is similar as in basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthwise_sep_opt
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 16, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 16, 4, 4 } });
 
     set_values(input, {
         -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f, -0.5f, -0.5f,  0.5f,  0.5f,
@@ -3249,10 +3236,10 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group16) {
         -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f, -0.5f, -0.5f,  3.0f,  3.0f,
     });
 
-    topology topology(input_layout("input", input.get_layout()));
+    topology topology(input_layout("input", input->get_layout()));
 
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx ,tensor(group(16), batch(1), feature(1), spatial(2,2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 16, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx ,tensor(group(16), batch(1), feature(1), spatial(2,2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 16, 1, 1 } });
 
     set_values(weights,
         {
@@ -3302,7 +3289,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group16) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f, 8.0f, 8.0f, 3.65f, 3.65f,
@@ -3320,9 +3307,9 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group16) {
 TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group16_bfyx) {
     //  Test for grouped convolution, there are 16 joined weights and biases (group 16)
     //  data is similar as in basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_split2_depthwise_sep_opt_bfyx
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 16, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 16, 4, 4 } });
 
     set_values(input, {
         -0.5f, 1.0f, 0.5f, 2.0f, 1.5f, -0.5f, 0.0f, -1.0f, 0.5f, 0.5f, -1.0f, 1.0f, 0.5f, 2.0f, 1.5f, -0.5f,
@@ -3343,10 +3330,10 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group16_bfyx)
         0.5f, 1.5f, 2.3f, -0.4f, 2.0f, -4.0f, 1.0f, 3.0f, 0.5f, 1.5f, 2.3f, -0.4f, 2.0f, -4.0f, 1.0f, 3.0f,
     });
 
-    topology topology(input_layout("input", input.get_layout()));
+    topology topology(input_layout("input", input->get_layout()));
 
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx ,tensor(group(16), batch(1), feature(1), spatial(2,2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 16, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx ,tensor(group(16), batch(1), feature(1), spatial(2,2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 16, 1, 1 } });
 
     set_values(weights,
         {
@@ -3397,7 +3384,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group16_bfyx)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         8.0f, 0.5f,  6.0f,  9.0f, 3.65f,-5.36f, 3.65f, -5.36f,
@@ -3450,14 +3437,14 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group16_bfyx)
     //  -1.75
     //   2.25
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 4, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 4, 1, 1 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1,{ 1, 1 }, 4 } });
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
-    auto biases1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
-    auto biases2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
+    auto biases1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
+    auto biases2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, {
        1.5f, 0.5f, 0.0f, -0.5f
@@ -3468,7 +3455,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group16_bfyx)
     set_values(biases2, { -1.0f, 2.5f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights1", weights1),
         data("biases1", biases1),
         data("weights2", weights2),
@@ -3492,12 +3479,12 @@ TEST(convolution_f32_fw_gpu, basic_wsiz2x2_wstr2x2_in4x4x2x2_nopad_group16_bfyx)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
-    EXPECT_FLOAT_EQ(-2.25f, get_value<float>(output_ptr, 0));
-    EXPECT_FLOAT_EQ(7.5f, get_value<float>(output_ptr, 1));
-    EXPECT_FLOAT_EQ(-1.75f, get_value<float>(output_ptr, 2));
-    EXPECT_FLOAT_EQ(2.25f, get_value<float>(output_ptr, 3));
+    EXPECT_FLOAT_EQ(-2.25f, output_ptr[0]);
+    EXPECT_FLOAT_EQ(7.5f, output_ptr[1]);
+    EXPECT_FLOAT_EQ(-1.75f, output_ptr[2]);
+    EXPECT_FLOAT_EQ(2.25f, output_ptr[3]);
 }
 
 TEST(convolution_f32_fw_gpu, basic_wsiz1x1_wstr2x2_in1x1x2x1_nopad_split2) {
@@ -3531,14 +3518,14 @@ TEST(convolution_f32_fw_gpu, basic_wsiz1x1_wstr2x2_in1x1x2x1_nopad_split2) {
     //   1
     //   3.5
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 2, 1, 1 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1,{ 1, 1 }, 4 } });
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
-    auto biases1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
-    auto biases2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
+    auto biases1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
+    auto biases2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, {
         1.5f, 0.5f
@@ -3549,7 +3536,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz1x1_wstr2x2_in1x1x2x1_nopad_split2) {
     set_values(biases2, { -1.0f, 2.5f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights1", weights1),
         data("biases1", biases1),
         data("weights2", weights2),
@@ -3573,12 +3560,12 @@ TEST(convolution_f32_fw_gpu, basic_wsiz1x1_wstr2x2_in1x1x2x1_nopad_split2) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
-    EXPECT_FLOAT_EQ(-2.0f, get_value<float>(output_ptr, 0));
-    EXPECT_FLOAT_EQ(6.5f, get_value<float>(output_ptr, 1));
-    EXPECT_FLOAT_EQ(1.0f, get_value<float>(output_ptr, 2));
-    EXPECT_FLOAT_EQ(3.5f, get_value<float>(output_ptr, 3));
+    EXPECT_FLOAT_EQ(-2.0f, output_ptr[0]);
+    EXPECT_FLOAT_EQ(6.5f, output_ptr[1]);
+    EXPECT_FLOAT_EQ(1.0f, output_ptr[2]);
+    EXPECT_FLOAT_EQ(3.5f, output_ptr[3]);
 }
 
 TEST(convolution_f32_fw_gpu, basic_wsiz1x1_wstr2x2_in1x1x4x1_filter_1x3x2x1x1_nopad_split2) {
@@ -3618,14 +3605,14 @@ TEST(convolution_f32_fw_gpu, basic_wsiz1x1_wstr2x2_in1x1x4x1_filter_1x3x2x1x1_no
     //   6
     //  -2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 4, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 4, 1, 1 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1,{ 1, 1 }, 6 } });
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 2, 1, 1 } });
-    auto biases1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 3, 1, 1 } });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 2, 1, 1 } });
-    auto biases2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 3, 1, 1 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 2, 1, 1 } });
+    auto biases1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 3, 1, 1 } });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 2, 1, 1 } });
+    auto biases2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 3, 1, 1 } });
 
     set_values(input, {
         1.5f, 0.5f, 2.0f, -1.0f
@@ -3636,7 +3623,7 @@ TEST(convolution_f32_fw_gpu, basic_wsiz1x1_wstr2x2_in1x1x4x1_filter_1x3x2x1x1_no
     set_values(biases2, { -1.0f, 2.5f, 2.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights1", weights1),
         data("biases1", biases1),
         data("weights2", weights2),
@@ -3660,14 +3647,14 @@ TEST(convolution_f32_fw_gpu, basic_wsiz1x1_wstr2x2_in1x1x4x1_filter_1x3x2x1x1_no
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
-    EXPECT_FLOAT_EQ(-1.5f, get_value<float>(output_ptr, 0));
-    EXPECT_FLOAT_EQ(8.0f, get_value<float>(output_ptr, 1));
-    EXPECT_FLOAT_EQ(7.75f, get_value<float>(output_ptr, 2));
-    EXPECT_FLOAT_EQ(11.0f, get_value<float>(output_ptr, 3));
-    EXPECT_FLOAT_EQ(6.0f, get_value<float>(output_ptr, 4));
-    EXPECT_FLOAT_EQ(-2.0f, get_value<float>(output_ptr, 5));
+    EXPECT_FLOAT_EQ(-1.5f, output_ptr[0]);
+    EXPECT_FLOAT_EQ(8.0f, output_ptr[1]);
+    EXPECT_FLOAT_EQ(7.75f, output_ptr[2]);
+    EXPECT_FLOAT_EQ(11.0f, output_ptr[3]);
+    EXPECT_FLOAT_EQ(6.0f, output_ptr[4]);
+    EXPECT_FLOAT_EQ(-2.0f, output_ptr[5]);
 
 }*/
 
@@ -3695,12 +3682,12 @@ TEST(convolution_gpu, trivial_convolution_relu) {
     //  4  0.0
     //  2  5
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 4, 4 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1 ,{ 2, 2 }, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, {
         -0.5f,  1.0f,  0.5f,  2.0f,
@@ -3712,7 +3699,7 @@ TEST(convolution_gpu, trivial_convolution_relu) {
     set_values(biases, { -2.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -3739,12 +3726,12 @@ TEST(convolution_gpu, trivial_convolution_relu) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
-    EXPECT_FLOAT_EQ(4.0f, get_value<float>(output_ptr, 0));
-    EXPECT_FLOAT_EQ(0.0f, get_value<float>(output_ptr, 1));
-    EXPECT_FLOAT_EQ(2.0f, get_value<float>(output_ptr, 2));
-    EXPECT_FLOAT_EQ(5.0f, get_value<float>(output_ptr, 3));
+    EXPECT_FLOAT_EQ(4.0f, output_ptr[0]);
+    EXPECT_FLOAT_EQ(0.0f, output_ptr[1]);
+    EXPECT_FLOAT_EQ(2.0f, output_ptr[2]);
+    EXPECT_FLOAT_EQ(5.0f, output_ptr[3]);
 }
 
 TEST(convolution_gpu, relu_with_negative_slope) {
@@ -3772,12 +3759,12 @@ TEST(convolution_gpu, relu_with_negative_slope) {
     //  4  -0.35
     //  2  5
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 4, 4 } });
     //auto output = memory::allocate({ memory::format::yxfb_f32,{ 1 ,{ 2, 2 }, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, {
         -0.5f,  1.0f,  0.5f,  2.0f,
@@ -3789,7 +3776,7 @@ TEST(convolution_gpu, relu_with_negative_slope) {
     set_values(biases, { -2.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -3817,29 +3804,29 @@ TEST(convolution_gpu, relu_with_negative_slope) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
-    EXPECT_FLOAT_EQ(4.0f, get_value<float>(output_ptr, 0));
-    EXPECT_FLOAT_EQ(-0.35f, get_value<float>(output_ptr, 1));
-    EXPECT_FLOAT_EQ(2.0f, get_value<float>(output_ptr, 2));
-    EXPECT_FLOAT_EQ(5.0f, get_value<float>(output_ptr, 3));
+    EXPECT_FLOAT_EQ(4.0f, output_ptr[0]);
+    EXPECT_FLOAT_EQ(-0.35f, output_ptr[1]);
+    EXPECT_FLOAT_EQ(2.0f, output_ptr[2]);
+    EXPECT_FLOAT_EQ(5.0f, output_ptr[3]);
 }
 
 TEST(convolution_gpu, DISABLED_two_1x1_kernels_after_each_other) {
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     extern const std::vector<float> conv_1x1_output;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 16, 8, 16, 16 } });
-    auto weights_conv_1 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 8, 8, 1, 1 } });
-    auto weights_conv_2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 8, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 16, 8, 16, 16 } });
+    auto weights_conv_1 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 8, 8, 1, 1 } });
+    auto weights_conv_2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 8, 1, 1 } });
 
     set_random_values<float>(input);
     set_random_values<float>(weights_conv_1);
     set_random_values<float>(weights_conv_2);
 
-    auto inp_lay = input_layout("input", input.get_layout());
+    auto inp_lay = input_layout("input", input->get_layout());
     auto conv_1 = convolution(
         "conv_1",
         "input",
@@ -3867,8 +3854,8 @@ TEST(convolution_gpu, DISABLED_two_1x1_kernels_after_each_other) {
 
     auto output_prim = outputs.at("conv_2").get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
-    auto output_layout = output_prim.get_layout();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
+    auto output_layout = output_prim->get_layout();
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -3876,16 +3863,12 @@ TEST(convolution_gpu, DISABLED_two_1x1_kernels_after_each_other) {
     int b_size = output_layout.size.batch[0];
     int f_offset = y_size * x_size;
     int b_offset = f_size * f_offset;
-    for (int b = 0; b < b_size; ++b)
-    {
-        for (int f = 0; f < f_size; ++f)
-        {
-            for (int y = 0; y < y_size; ++y)
-            {
-                for (int x = 0; x < x_size; ++x)
-                {
+    for (int b = 0; b < b_size; ++b) {
+        for (int f = 0; f < f_size; ++f) {
+            for (int y = 0; y < y_size; ++y) {
+                for (int x = 0; x < x_size; ++x) {
                     int idx = b * b_offset + f * f_offset + y * x_size + x;
-                    EXPECT_TRUE(are_equal(conv_1x1_output[idx], get_value<float>(output_ptr, idx)));
+                    EXPECT_TRUE(are_equal(conv_1x1_output[idx], output_ptr[idx]));
                 }
             }
         }
@@ -3918,13 +3901,13 @@ TEST(convolution_gpu, basic_yxfb_4_4_yxfb_2_2_b16_if2_of16_st2_2_p0_sp1_fp32)
     const int32_t output_x = (input_x - weights_x) / stride_x + 1;
     const int32_t output_y = (input_y - weights_y) / stride_y + 1;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto input_size = tensor( batch_size, input_feature_count, input_x, input_y );
-    auto input = memory::allocate(engine, { data_types::f32, input_format, input_size });
+    auto input = engine.allocate_memory({ data_types::f32, input_format, input_size });
     auto weights_size = tensor( output_feature_count, input_feature_count, weights_x, weights_y );
-    auto weights = memory::allocate(engine, { data_types::f32, weights_format, weights_size });
-    auto biases = memory::allocate(engine, { data_types::f32, biases_format, {1,output_feature_count,1,1}});
+    auto weights = engine.allocate_memory({ data_types::f32, weights_format, weights_size });
+    auto biases = engine.allocate_memory({ data_types::f32, biases_format, {1,output_feature_count,1,1}});
 
     //auto output = memory::allocate({output_format, {batch_size, {output_x, output_y}, output_feature_count}});
 
@@ -4021,7 +4004,7 @@ TEST(convolution_gpu, basic_yxfb_4_4_yxfb_2_2_b16_if2_of16_st2_2_p0_sp1_fp32)
 
     // Computing convolution.
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution(
@@ -4049,7 +4032,7 @@ TEST(convolution_gpu, basic_yxfb_4_4_yxfb_2_2_b16_if2_of16_st2_2_p0_sp1_fp32)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     // Checking result.
     uint32_t i = 0;
@@ -4059,12 +4042,12 @@ TEST(convolution_gpu, basic_yxfb_4_4_yxfb_2_2_b16_if2_of16_st2_2_p0_sp1_fp32)
         {
             for (uint32_t bi = 0; bi < batch_size; ++bi, ++i)
             {
-                auto equal = are_equal(output_vals[i], get_value<float>(output_ptr, i));
+                auto equal = are_equal(output_vals[i], output_ptr[i]);
                 EXPECT_TRUE(equal);
                 if (!equal)
                 {
                     std::cout << "Failed at position (" << yxi << ", output feature = " << ofi << ", batch = " << bi << "): "
-                        << output_vals[i] << " != " << get_value<float>(output_ptr, i) << std::endl;
+                        << output_vals[i] << " != " << output_ptr[i] << std::endl;
                     return;
                 }
             }
@@ -4074,9 +4057,8 @@ TEST(convolution_gpu, basic_yxfb_4_4_yxfb_2_2_b16_if2_of16_st2_2_p0_sp1_fp32)
 #undef USE_OLD_WEIGHTS_FORMAT
 }
 
-void add_primitives(const engine& engine, topology& topology)
-{
-    auto weights = memory::allocate(engine, { data_types::i8, format::bfyx,{ 2, 1, 3, 2 } });
+void add_primitives(engine& engine, topology& topology) {
+    auto weights = engine.allocate_memory({ data_types::i8, format::bfyx,{ 2, 1, 3, 2 } });
 
     std::vector<char> weights_values = { 1, 2, 1,
                                          2, 1, 2,
@@ -4084,7 +4066,7 @@ void add_primitives(const engine& engine, topology& topology)
                                          19, 17, -1,
                                          -10, 32, 23 };
     set_values<char>(weights, weights_values);
-    cldnn::memory biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
     set_values(biases, { 1.0f, -8.0f });
 
     topology.add(
@@ -4123,9 +4105,8 @@ TEST(convolution_f32_fw_gpu, byte_activation) {
     //
     //  Bias:
     //  1 -8
-    auto eng_conf = get_test_engine();
-    engine engine{ eng_conf };
-    auto input = memory::allocate(engine, { data_types::i8, format::bfyx,{ 1, 1, 5, 4 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::i8, format::bfyx,{ 1, 1, 5, 4 } });
 
     VVVF<char> output_vec = {
         {
@@ -4146,7 +4127,7 @@ TEST(convolution_f32_fw_gpu, byte_activation) {
                               -1, -1, -1, -1, -1 });
 
     topology topology(
-        input_layout("input", input.get_layout()));
+        input_layout("input", input->get_layout()));
     add_primitives(engine, topology);
     network network(engine, topology, opts);
     network.set_input_data("input", input);
@@ -4155,8 +4136,8 @@ TEST(convolution_f32_fw_gpu, byte_activation) {
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output_memory = outputs.at("out").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -4176,11 +4157,11 @@ TEST(convolution_f32_fw_gpu, byte_activation) {
 }
 
 TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_symmetric) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::bfyx,{ 1, 1, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::i8, format::bfyx,{ 2, 1, 3, 3 } });
-    cldnn::memory biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::bfyx,{ 1, 1, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::i8, format::bfyx,{ 2, 1, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values<uint8_t>(input, { 1, 2, 3, 4, 5,
                                  2, 2, 3, 4, 6,
@@ -4206,7 +4187,7 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_symmetric) {
         } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         convolution("conv", "input", { "weights" }, { "biases" }, tensor{ 0, 0, 2, 2 }, tensor(0), tensor{1, 1, 1, 1}, tensor{1, 2, 3, 2}),
@@ -4221,9 +4202,9 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_symmetric) {
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output_memory = outputs.at("out").get_memory();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
-    auto output_layout = output_memory.get_layout();
+    auto output_layout = output_memory->get_layout();
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
     int f_size = output_layout.size.feature[0];
@@ -4243,13 +4224,13 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_symmetric) {
 }
 
 TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_weight_and_activations) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::bfyx,{ 1, 1, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::i8, format::bfyx,{ 2, 1, 3, 3 } });
-    cldnn::memory biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
-    auto w_zp = memory::allocate(engine, { data_types::i8, format::bfyx,{ 2, 1, 1, 1 } });
-    auto a_zp = memory::allocate(engine, { data_types::u8, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::bfyx,{ 1, 1, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::i8, format::bfyx,{ 2, 1, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto w_zp = engine.allocate_memory({ data_types::i8, format::bfyx,{ 2, 1, 1, 1 } });
+    auto a_zp = engine.allocate_memory({ data_types::u8, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values<uint8_t>(input, { 1, 2, 3, 4, 5,
                                  2, 2, 3, 4, 6,
@@ -4277,7 +4258,7 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_weight_an
         } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         data("a_zp", a_zp),
@@ -4295,9 +4276,9 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_weight_an
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output_memory = outputs.at("out").get_memory();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
-    auto output_layout = output_memory.get_layout();
+    auto output_layout = output_memory->get_layout();
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
     int f_size = output_layout.size.feature[0];
@@ -4317,12 +4298,12 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_weight_an
 }
 
 TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activations_per_tensor) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::bfyx,{ 1, 1, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::i8, format::bfyx,{ 2, 1, 3, 3 } });
-    cldnn::memory biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
-    auto a_zp = memory::allocate(engine, { data_types::u8, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::bfyx,{ 1, 1, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::i8, format::bfyx,{ 2, 1, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto a_zp = engine.allocate_memory({ data_types::u8, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values<uint8_t>(input, { 1, 2, 3, 4, 5,
                                  2, 2, 3, 4, 6,
@@ -4349,7 +4330,7 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activatio
         } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         data("a_zp", a_zp),
@@ -4366,9 +4347,9 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activatio
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output_memory = outputs.at("out").get_memory();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
-    auto output_layout = output_memory.get_layout();
+    auto output_layout = output_memory->get_layout();
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
     int f_size = output_layout.size.feature[0];
@@ -4388,12 +4369,12 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activatio
 }
 
 TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activations_per_channel) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::bfyx,{ 1, 2, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::i8, format::bfyx,{ 2, 2, 3, 3 } });
-    cldnn::memory biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
-    auto a_zp = memory::allocate(engine, { data_types::u8, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::bfyx,{ 1, 2, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::i8, format::bfyx,{ 2, 2, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto a_zp = engine.allocate_memory({ data_types::u8, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values<uint8_t>(input, { 1, 2, 3, 4, 5,
                                  2, 2, 3, 4, 6,
@@ -4434,7 +4415,7 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activatio
         } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         data("a_zp", a_zp),
@@ -4451,9 +4432,9 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activatio
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output_memory = outputs.at("out").get_memory();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
-    auto output_layout = output_memory.get_layout();
+    auto output_layout = output_memory->get_layout();
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
     int f_size = output_layout.size.feature[0];
@@ -4473,12 +4454,12 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activatio
 }
 
 TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activations_per_channel_3ic_with_sub) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::bfyx,{ 1, 3, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::i8, format::bfyx,{ 2, 3, 3, 3 } });
-    cldnn::memory biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
-    auto a_zp = memory::allocate(engine, { data_types::u8, format::bfyx,{ 1, 3, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::bfyx,{ 1, 3, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::i8, format::bfyx,{ 2, 3, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto a_zp = engine.allocate_memory({ data_types::u8, format::bfyx,{ 1, 3, 1, 1 } });
 
     set_values<uint8_t>(input, { 1, 2, 3, 4, 5,
                                  2, 2, 3, 4, 6,
@@ -4533,7 +4514,7 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activatio
         } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         data("a_zp", a_zp),
@@ -4552,9 +4533,9 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activatio
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output_memory = outputs.at("out").get_memory();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
-    auto output_layout = output_memory.get_layout();
+    auto output_layout = output_memory->get_layout();
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
     int f_size = output_layout.size.feature[0];
@@ -4574,12 +4555,12 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_activatio
 }
 
 TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_weights_per_channel) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::bfyx,{ 1, 1, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::i8, format::bfyx,{ 2, 1, 3, 3 } });
-    cldnn::memory biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
-    auto w_zp = memory::allocate(engine, { data_types::i8, format::bfyx,{ 2, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::bfyx,{ 1, 1, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::i8, format::bfyx,{ 2, 1, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto w_zp = engine.allocate_memory({ data_types::i8, format::bfyx,{ 2, 1, 1, 1 } });
 
     set_values<uint8_t>(input, { 1, 2, 3, 4, 5,
                                  2, 2, 3, 4, 6,
@@ -4606,7 +4587,7 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_weights_p
         } };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         data("w_zp", w_zp),
@@ -4623,9 +4604,9 @@ TEST(convolution_int8_fw_gpu, quantized_convolution_u8s8f32_asymmetric_weights_p
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output_memory = outputs.at("out").get_memory();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
-    auto output_layout = output_memory.get_layout();
+    auto output_layout = output_memory->get_layout();
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
     int f_size = output_layout.size.feature[0];
@@ -4648,10 +4629,9 @@ TEST(convolution_gpu, basic_yxfb_4_4_yxfb_2_2_b16_if2_of16_st2_2_p0_sp1_fp16)
 {
 #define USE_OLD_WEIGHTS_FORMAT 0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -4681,17 +4661,17 @@ TEST(convolution_gpu, basic_yxfb_4_4_yxfb_2_2_b16_if2_of16_st2_2_p0_sp1_fp16)
     const int32_t output_y = (input_y - weights_y) / stride_y + 1;
 
     auto input_size = tensor( batch_size, input_feature_count, input_x, input_y );
-    auto input = memory::allocate(engine, { data_types::f32, input_format, input_size });
+    auto input = engine.allocate_memory({ data_types::f32, input_format, input_size });
     auto weights_size = tensor( output_feature_count, input_feature_count, weights_x, weights_y );
-    auto weights = memory::allocate(engine, { data_types::f32, weights_format, weights_size });
+    auto weights = engine.allocate_memory({ data_types::f32, weights_format, weights_size });
     auto biases_size = tensor( 1,output_feature_count,1,1 );
-    auto biases = memory::allocate(engine, { data_types::f32, biases_format, biases_size });
+    auto biases = engine.allocate_memory({ data_types::f32, biases_format, biases_size });
     auto output_size = tensor( batch_size, output_feature_count, output_x, output_y );
     //auto output = memory::allocate({output_format, {batch_size, {output_x, output_y}, output_feature_count}});
 
-    //auto input_cvtd = memory::allocate(engine, { data_types::f16, input_size });
-    //auto weights_cvtd = memory::allocate(engine, { data_types::f16, weights_size });
-    //auto biases_cvtd = memory::allocate(engine, { data_types::f16, biases_size });
+    //auto input_cvtd = engine.allocate_memory({ data_types::f16, input_size });
+    //auto weights_cvtd = engine.allocate_memory({ data_types::f16, weights_size });
+    //auto biases_cvtd = engine.allocate_memory({ data_types::f16, biases_size });
     //auto output_cvtd  = memory::allocate({output_cvt_format, {batch_size, {output_x, output_y}, output_feature_count}});
 
     // input:
@@ -4785,9 +4765,9 @@ TEST(convolution_gpu, basic_yxfb_4_4_yxfb_2_2_b16_if2_of16_st2_2_p0_sp1_fp16)
         }
     }
 
-    //auto expected_float = memory::allocate(engine, { data_types::f32,{ format::x,{ static_cast<int32_t>(output_vals.size()) } } });
-    //auto expected_half  = memory::allocate(engine, { data_types::f16,{ format::x,{ static_cast<int32_t>(output_vals.size()) } } });
-    //auto expected       = memory::allocate(engine, { data_types::f32,{ format::x,{ static_cast<int32_t>(output_vals.size()) } } });
+    //auto expected_float = engine.allocate_memory({ data_types::f32,{ format::x,{ static_cast<int32_t>(output_vals.size()) } } });
+    //auto expected_half  = engine.allocate_memory({ data_types::f16,{ format::x,{ static_cast<int32_t>(output_vals.size()) } } });
+    //auto expected       = engine.allocate_memory({ data_types::f32,{ format::x,{ static_cast<int32_t>(output_vals.size()) } } });
 
 //    set_values(expected_float, output_vals);
 //    auto cvt_expected_f32_f16 = reorder::create({expected_float, expected_half});
@@ -4798,7 +4778,7 @@ TEST(convolution_gpu, basic_yxfb_4_4_yxfb_2_2_b16_if2_of16_st2_2_p0_sp1_fp16)
 
     // Computing convolution.
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("cvt_input", "input", {data_types::f16, input_format, input_size}),
         data("weights", weights),
         reorder("cvt_weights", "weights", {data_types::f16, weights_format, weights_size}),
@@ -4822,7 +4802,7 @@ TEST(convolution_gpu, basic_yxfb_4_4_yxfb_2_2_b16_if2_of16_st2_2_p0_sp1_fp16)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
 
     // Checking result.
     uint32_t i = 0;
@@ -5042,10 +5022,9 @@ INSTANTIATE_TEST_CASE_P(convolution_gpu_test,
 
 TEST_P(convolution_gpu_fs_byx_fsv32, fs_byx_fsv32)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -5066,20 +5045,20 @@ TEST_P(convolution_gpu_fs_byx_fsv32, fs_byx_fsv32)
     auto input_size = tensor(batch_num, input_f, input_xy, input_xy);
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_xy, input_xy, -1, 1);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f16, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f16, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy);
     auto weights_data = generate_random_4d<FLOAT16>(output_f, input_f, filter_xy, filter_xy, -1, 1);
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
-    auto weights_mem = memory::allocate(engine, { data_types::f16, format::bfyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f16, format::bfyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<FLOAT16>(batch_num, VVVF<FLOAT16>(output_f));
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
         data("weights_fsv", weights_mem));
 
     // Reorder input to fs_byx_fsv32
@@ -5090,7 +5069,7 @@ TEST_P(convolution_gpu_fs_byx_fsv32, fs_byx_fsv32)
         // Generate bias data
         auto biases_size = tensor(1, output_f, 1, 1);
         auto biases_data = generate_random_1d<FLOAT16>(output_f, -1, 1);
-        auto biases_mem = memory::allocate(engine, { data_types::f16, format::bfyx, biases_size });
+        auto biases_mem = engine.allocate_memory({ data_types::f16, format::bfyx, biases_size });
         set_values(biases_mem, biases_data);
 
         // Calculate reference values with bias
@@ -5152,9 +5131,9 @@ TEST_P(convolution_gpu_fs_byx_fsv32, fs_byx_fsv32)
     network.execute();
 
     auto out_mem = network.get_output("conv_fsv").get_memory();
-    auto out_ptr = out_mem.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr(out_mem, get_test_stream());
 
-    ASSERT_EQ(out_mem.get_layout().format, format::fs_b_yx_fsv32);
+    ASSERT_EQ(out_mem->get_layout().format, format::fs_b_yx_fsv32);
 
     for (int bi = 0; bi < batch_num; ++bi)
         for (int fi = 0; fi < output_f; ++fi)
@@ -5177,10 +5156,9 @@ TEST_P(convolution_gpu_fs_byx_fsv32, fs_byx_fsv32)
 }
 
 TEST(convolution_f16_fsv_gpu, convolution_f16_fsv_gpu_padding) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -5197,20 +5175,20 @@ TEST(convolution_f16_fsv_gpu, convolution_f16_fsv_gpu_padding) {
     auto input_size = tensor(batch_num, input_f, input_xy, input_xy);
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_xy, input_xy, -1, 1);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f16, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f16, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy);
     auto weights_data = generate_random_4d<FLOAT16>(output_f, input_f, filter_xy, filter_xy, -1, 1);
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
-    auto weights_mem = memory::allocate(engine, { data_types::f16, format::bfyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f16, format::bfyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<FLOAT16>(batch_num, VVVF<FLOAT16>(output_f));
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
         data("weights_fsv", weights_mem));
 
     // add input padding by X and Y
@@ -5220,7 +5198,7 @@ TEST(convolution_f16_fsv_gpu, convolution_f16_fsv_gpu_padding) {
     // Generate bias data
     auto biases_size = tensor(1, output_f, 1, 1);
     auto biases_data = generate_random_1d<FLOAT16>(output_f, -1, 1);
-    auto biases_mem = memory::allocate(engine, { data_types::f16, format::bfyx, biases_size });
+    auto biases_mem = engine.allocate_memory({ data_types::f16, format::bfyx, biases_size });
     set_values(biases_mem, biases_data);
 
     // Calculate reference values
@@ -5254,9 +5232,9 @@ TEST(convolution_f16_fsv_gpu, convolution_f16_fsv_gpu_padding) {
     network.execute();
 
     auto out_mem = network.get_output("conv_fsv").get_memory();
-    auto out_ptr = out_mem.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr(out_mem, get_test_stream());
 
-    ASSERT_EQ(out_mem.get_layout().format, format::fs_b_yx_fsv32);
+    ASSERT_EQ(out_mem->get_layout().format, format::fs_b_yx_fsv32);
 
     for (int bi = 0; bi < batch_num; ++bi)
         for (int fi = 0; fi < output_f; ++fi)
@@ -5309,10 +5287,9 @@ INSTANTIATE_TEST_CASE_P(convolution_gpu_with_crop,
 
 TEST_P(convolution_gpu_fs_byx_fsv32_crop, fs_byx_fsv32_crop)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -5333,7 +5310,7 @@ TEST_P(convolution_gpu_fs_byx_fsv32_crop, fs_byx_fsv32_crop)
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy);
     auto weights_data = generate_random_4d<FLOAT16>(output_f, input_f, filter_xy, filter_xy, -1, 1);
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
-    auto weights_mem = memory::allocate(engine, { data_types::f16, format::bfyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f16, format::bfyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // ref input
@@ -5357,11 +5334,11 @@ TEST_P(convolution_gpu_fs_byx_fsv32_crop, fs_byx_fsv32_crop)
 
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
     auto input_size = tensor(batch_num, input_f * 2, input_xy, input_xy);
-    auto input_mem = memory::allocate(engine, { data_types::f16, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f16, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
         data("weights_fsv", weights_mem));
 
     auto crop_batch_num = batch_num;
@@ -5383,7 +5360,7 @@ TEST_P(convolution_gpu_fs_byx_fsv32_crop, fs_byx_fsv32_crop)
         // Generate bias data
         auto biases_size = tensor(1, output_f, 1, 1);
         auto biases_data = generate_random_1d<FLOAT16>(output_f, -1, 1);
-        auto biases_mem = memory::allocate(engine, { data_types::f16, format::bfyx, biases_size });
+        auto biases_mem = engine.allocate_memory({ data_types::f16, format::bfyx, biases_size });
         set_values(biases_mem, biases_data);
 
         // Calculate reference values with bias
@@ -5462,9 +5439,9 @@ TEST_P(convolution_gpu_fs_byx_fsv32_crop, fs_byx_fsv32_crop)
     network.execute();
 
     auto out_mem = network.get_output("concat").get_memory();
-    auto out_ptr = out_mem.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr(out_mem, get_test_stream());
 
-    ASSERT_EQ(out_mem.get_layout().format, format::bfyx);
+    ASSERT_EQ(out_mem->get_layout().format, format::bfyx);
 
     for (int bi = 0; bi < batch_num; ++bi)
         for (int fi = 0; fi < output_f * 2; ++fi)
@@ -5495,29 +5472,29 @@ TEST(convolution_f32_fw_gpu, convolution_int8_b_fs_yx_fsv4_to_bfyx) {
     const int input_size_x = 1280;
     const int input_size_y = 720;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto input_size = tensor(batch_num, input_f, input_size_x, input_size_y);
     auto input_data = generate_random_4d<float>(batch_num, input_f, input_size_y, input_size_x, -10, 10);
 
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, input_size });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, input_size });
     set_values(input, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy);
     auto weights_data = generate_random_4d<int8_t>(output_f, input_f, filter_xy, filter_xy, -10, 10);
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
-    auto weights = memory::allocate(engine, { data_types::i8, format::bfyx, weights_size });
+    auto weights = engine.allocate_memory({ data_types::i8, format::bfyx, weights_size });
     set_values(weights, weights_data_bfyx);
 
     auto biases_size = tensor(1, output_f, 1, 1);
     auto biases_data = generate_random_4d<int8_t>(1, output_f, 1, 1, -10, 10);
     auto biases_data_bfyx = flatten_4d(format::bfyx, biases_data);
-    auto biases = memory::allocate(engine, { data_types::i8, format::bfyx, biases_size });
+    auto biases = engine.allocate_memory({ data_types::i8, format::bfyx, biases_size });
     set_values(biases, biases_data_bfyx);
 
     topology topology_ref(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("to_int", "input", { data_types::i8,format::bfyx,{ batch_num, input_f, input_size_x, input_size_y } }),
         data("weights", weights),
         data("biases", biases),
@@ -5535,11 +5512,11 @@ TEST(convolution_f32_fw_gpu, convolution_int8_b_fs_yx_fsv4_to_bfyx) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     topology topology_act(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("to_int", "input", { data_types::i8,format::b_fs_yx_fsv4,{ batch_num, input_f, input_size_x, input_size_y } }),
         data("weights", weights),
         data("biases", biases),
@@ -5559,7 +5536,7 @@ TEST(convolution_f32_fw_gpu, convolution_int8_b_fs_yx_fsv4_to_bfyx) {
     EXPECT_EQ(outputs_act.begin()->first, "output");
 
     auto output_memory_act = outputs_act.at("output").get_memory();
-    auto output_act_ptr = output_memory_act.pointer<float>();
+    cldnn::mem_lock<float> output_act_ptr(output_memory_act, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -5582,10 +5559,9 @@ TEST(convolution_f32_fw_gpu, convolution_int8_b_fs_yx_fsv4_to_bfyx) {
 TEST(convolution_gpu, bfyx_iyxo_5x5_fp16)
 {
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -5613,13 +5589,13 @@ TEST(convolution_gpu, bfyx_iyxo_5x5_fp16)
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_size_y, input_size_x, -1, 1);
 
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f16, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f16, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy);
     auto weights_data = generate_random_4d<FLOAT16>(output_f, input_f, filter_xy, filter_xy, -1, 1);
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
-    auto weights_mem = memory::allocate(engine, { data_types::f16, format::bfyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f16, format::bfyx, weights_size });
 
     set_values(weights_mem, weights_data_bfyx);
 
@@ -5627,7 +5603,7 @@ TEST(convolution_gpu, bfyx_iyxo_5x5_fp16)
     auto reference_result = VVVVF<FLOAT16>(batch_num, VVVF<FLOAT16>(output_f));
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
         data("weights_fsv", weights_mem)
     );
 
@@ -5636,7 +5612,7 @@ TEST(convolution_gpu, bfyx_iyxo_5x5_fp16)
         // Generate bias data
         auto biases_size = tensor(1, output_f, 1, 1);
         auto biases_data = generate_random_1d<FLOAT16>(output_f, -1, 1);
-        auto biases_mem = memory::allocate(engine, { data_types::f16, format::bfyx, biases_size });
+        auto biases_mem = engine.allocate_memory({ data_types::f16, format::bfyx, biases_size });
         set_values(biases_mem, biases_data);
 
         // Calculate reference values with bias
@@ -5698,7 +5674,7 @@ TEST(convolution_gpu, bfyx_iyxo_5x5_fp16)
     network.execute();
 
     auto out_mem = network.get_output("conv_fsv").get_memory();
-    auto out_ptr = out_mem.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr(out_mem, get_test_stream());
 
     for (int bi = 0; bi < batch_num; ++bi)
         for (int fi = 0; fi < output_f; ++fi)
@@ -5721,16 +5697,16 @@ TEST(convolution_gpu, bfyx_iyxo_5x5_fp16)
 }
 
 template<typename T>
-void blockedFormatZeroCheck(cldnn::memory out_mem) {
-    auto out_ptr = out_mem.pointer<T>();
+void blockedFormatZeroCheck(cldnn::memory::ptr out_mem) {
+    cldnn::mem_lock<T> out_ptr(out_mem, get_test_stream());
 
     bool batch_blocked = false;
-    if (out_mem.get_layout().format == format::bs_fs_zyx_bsv16_fsv16 ||
-        out_mem.get_layout().format == format::bs_fs_yx_bsv16_fsv16)
+    if (out_mem->get_layout().format == format::bs_fs_zyx_bsv16_fsv16 ||
+        out_mem->get_layout().format == format::bs_fs_yx_bsv16_fsv16)
         batch_blocked = true;
     const int block_size = 16;
 
-    auto output_tensor = out_mem.get_layout().get_buffer_size();
+    auto output_tensor = out_mem->get_layout().get_buffer_size();
     const int b = output_tensor.batch[0];
     const int f = output_tensor.feature[0];
     const int spatials = std::accumulate(output_tensor.spatial.begin(), output_tensor.spatial.end(), 1, std::multiplies<int>());
@@ -5820,7 +5796,7 @@ INSTANTIATE_TEST_CASE_P(convolution_gpu_block3D,
 
 TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int batch_num = testing::get<0>(GetParam());
     const int input_f = testing::get<1>(GetParam());
@@ -5838,7 +5814,7 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32)
     auto input_size = tensor(batch_num, input_f, input_xy, input_xy, 1);
     auto input_data = generate_random_4d<float>(batch_num, input_f, input_xy, input_xy, 1, 10);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfzyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::bfzyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy, 1);
@@ -5846,14 +5822,14 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32)
 
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
 
-    auto weights_mem = memory::allocate(engine, { data_types::f32, format::bfzyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f32, format::bfzyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<float>(batch_num, VVVF<float>(output_f));
 
     topology topology(
-            input_layout("input", input_mem.get_layout()),
+            input_layout("input", input_mem->get_layout()),
             data("weights", weights_mem));
 
     // Reorder input to correct format
@@ -5864,7 +5840,7 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32)
         // Generate bias data
         auto biases_size = tensor(1, output_f, 1, 1, 1);
         auto biases_data = generate_random_1d<float>(output_f, -1, 1);
-        auto biases_mem = memory::allocate(engine, { data_types::f32, format::bfzyx, biases_size });
+        auto biases_mem = engine.allocate_memory({ data_types::f32, format::bfzyx, biases_size });
         set_values(biases_mem, biases_data);
 
         // Calculate reference values with bias
@@ -5925,14 +5901,14 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32)
     network.execute();
 
     auto out_mem = network.get_output("conv_bsv16_fsv16").get_memory();
-    auto out_ptr = out_mem.pointer<float>();
+    cldnn::mem_lock<float> out_ptr(out_mem, get_test_stream());
 
     auto out_mem_bfyx = network.get_output("reorder_bfzyx").get_memory();
-    auto out_ptr_bfyx = out_mem_bfyx.pointer<float>();
+    cldnn::mem_lock<float> out_ptr_bfyx(out_mem_bfyx, get_test_stream());
 
     blockedFormatZeroCheck<float>(out_mem);
 
-    ASSERT_EQ(out_mem.get_layout().format, input_format);
+    ASSERT_EQ(out_mem->get_layout().format, input_format);
 
     auto flatten_ref = flatten_4d(format::bfyx, reference_result);
 
@@ -5950,10 +5926,9 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32)
 
 TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp16)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -5976,7 +5951,7 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp16)
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_xy, input_xy, 0, 1);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
 
-    auto input_mem = memory::allocate(engine, { data_types::f16, format::bfzyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f16, format::bfzyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy, 1);
@@ -5984,14 +5959,14 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp16)
 
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
 
-    auto weights_mem = memory::allocate(engine, { data_types::f16, format::bfzyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f16, format::bfzyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<FLOAT16>(batch_num, VVVF<FLOAT16>(output_f));
 
     topology topology(
-            input_layout("input", input_mem.get_layout()),
+            input_layout("input", input_mem->get_layout()),
             data("weights", weights_mem));
 
     // Reorder input to correct format
@@ -6002,7 +5977,7 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp16)
         // Generate bias data
         auto biases_size = tensor(1, output_f, 1, 1, 1);
         auto biases_data = generate_random_1d<FLOAT16>(output_f, -1, 1);
-        auto biases_mem = memory::allocate(engine, { data_types::f16, format::bfzyx, biases_size });
+        auto biases_mem = engine.allocate_memory({ data_types::f16, format::bfzyx, biases_size });
         set_values(biases_mem, biases_data);
 
         // Calculate reference values with bias
@@ -6063,14 +6038,14 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp16)
     network.execute();
 
     auto out_mem = network.get_output("conv_bsv16_fsv16").get_memory();
-    auto out_ptr = out_mem.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr(out_mem, get_test_stream());
 
     auto out_mem_bfyx = network.get_output("reorder_bfzyx").get_memory();
-    auto out_ptr_bfyx = out_mem_bfyx.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr_bfyx(out_mem_bfyx, get_test_stream());
 
     blockedFormatZeroCheck<FLOAT16>(out_mem);
 
-    ASSERT_EQ(out_mem.get_layout().format, input_format);
+    ASSERT_EQ(out_mem->get_layout().format, input_format);
 
     auto flatten_ref = flatten_4d(format::bfyx, reference_result);
 
@@ -6087,7 +6062,7 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp16)
 
 TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32_fused_ops)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int batch_num = testing::get<0>(GetParam());
     const int input_f = testing::get<1>(GetParam());
@@ -6105,7 +6080,7 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32_fused_ops)
     auto input_size = tensor(batch_num, input_f, input_xy, input_xy, 1);
     auto input_data = generate_random_4d<float>(batch_num, input_f, input_xy, input_xy, 1, 10);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfzyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::bfzyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy, 1);
@@ -6113,14 +6088,14 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32_fused_ops)
 
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
 
-    auto weights_mem = memory::allocate(engine, { data_types::f32, format::bfzyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f32, format::bfzyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<float>(batch_num, VVVF<float>(output_f));
 
     topology topology(
-            input_layout("input", input_mem.get_layout()),
+            input_layout("input", input_mem->get_layout()),
             data("weights", weights_mem));
 
     // Reorder input to correct format
@@ -6131,7 +6106,7 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32_fused_ops)
         // Generate bias data
         auto biases_size = tensor(1, output_f, 1, 1, 1);
         auto biases_data = generate_random_1d<float>(output_f, -1, 1);
-        auto biases_mem = memory::allocate(engine, { data_types::f32, format::bfzyx, biases_size });
+        auto biases_mem = engine.allocate_memory({ data_types::f32, format::bfzyx, biases_size });
         set_values(biases_mem, biases_data);
 
         // Calculate reference values with bias
@@ -6181,7 +6156,7 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32_fused_ops)
     }
 
     const float scalar = 5.5f;
-    auto scale_mem = memory::allocate(engine, { data_types::f32, format::bfzyx, {1, 1, 1, 1, 1} });
+    auto scale_mem = engine.allocate_memory({ data_types::f32, format::bfzyx, {1, 1, 1, 1, 1} });
     set_values(scale_mem, {scalar});
 
     topology.add(data("scalar", scale_mem));
@@ -6199,14 +6174,14 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32_fused_ops)
     network.execute();
 
     auto out_mem = network.get_output("conv_bsv16_fsv16").get_memory();
-    auto out_ptr = out_mem.pointer<float>();
+    cldnn::mem_lock<float> out_ptr(out_mem, get_test_stream());
 
     auto out_mem_bfyx = network.get_output("reorder_bfzyx").get_memory();
-    auto out_ptr_bfyx = out_mem_bfyx.pointer<float>();
+    cldnn::mem_lock<float> out_ptr_bfyx(out_mem_bfyx, get_test_stream());
 
     blockedFormatZeroCheck<float>(out_mem);
 
-    ASSERT_EQ(out_mem.get_layout().format, input_format);
+    ASSERT_EQ(out_mem->get_layout().format, input_format);
 
     auto flatten_ref = flatten_4d(format::bfyx, reference_result);
 
@@ -6251,7 +6226,7 @@ INSTANTIATE_TEST_CASE_P(convolution_gpu_block,
 
 TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int batch_num = testing::get<0>(GetParam());
     const int input_xy = 5;
@@ -6273,7 +6248,7 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32)
     auto input_size = tensor(batch_num, input_f, input_xy, input_xy);
     auto input_data = generate_random_4d<float>(batch_num, input_f, input_xy, input_xy, 1, 10);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy);
@@ -6281,14 +6256,14 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32)
 
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
 
-    auto weights_mem = memory::allocate(engine, { data_types::f32, format::bfyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f32, format::bfyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<float>(batch_num, VVVF<float>(output_f));
 
     topology topology(
-            input_layout("input", input_mem.get_layout()),
+            input_layout("input", input_mem->get_layout()),
             data("weights", weights_mem));
 
     // Reorder input to bs_fs_yx_bsv16_fsv16
@@ -6299,7 +6274,7 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32)
         // Generate bias data
         auto biases_size = tensor(1, output_f, 1, 1);
         auto biases_data = generate_random_1d<float>(output_f, -1, 1);
-        auto biases_mem = memory::allocate(engine, { data_types::f32, format::bfyx, biases_size });
+        auto biases_mem = engine.allocate_memory({ data_types::f32, format::bfyx, biases_size });
         set_values(biases_mem, biases_data);
 
         // Calculate reference values with bias
@@ -6362,12 +6337,12 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32)
     network.execute();
 
     auto out_mem = network.get_output("conv_bsv16_fsv16").get_memory();
-    auto out_ptr = out_mem.pointer<float>();
+    cldnn::mem_lock<float> out_ptr(out_mem, get_test_stream());
 
     auto out_mem_bfyx = network.get_output("reorder_bfyx").get_memory();
-    auto out_ptr_bfyx = out_mem_bfyx.pointer<float>();
+    cldnn::mem_lock<float> out_ptr_bfyx(out_mem_bfyx, get_test_stream());
 
-    ASSERT_EQ(out_mem.get_layout().format, format::bs_fs_yx_bsv16_fsv16);
+    ASSERT_EQ(out_mem->get_layout().format, format::bs_fs_yx_bsv16_fsv16);
 
     auto flatten_ref = flatten_4d(format::bfyx, reference_result);
 
@@ -6384,10 +6359,9 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32)
 
 TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp16)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -6414,7 +6388,7 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp16)
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_xy, input_xy, 0, 1);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
 
-    auto input_mem = memory::allocate(engine, { data_types::f16, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f16, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy);
@@ -6422,14 +6396,14 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp16)
 
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
 
-    auto weights_mem = memory::allocate(engine, { data_types::f16, format::bfyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f16, format::bfyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<FLOAT16>(batch_num, VVVF<FLOAT16>(output_f));
 
     topology topology(
-            input_layout("input", input_mem.get_layout()),
+            input_layout("input", input_mem->get_layout()),
             data("weights", weights_mem));
 
     // Reorder input to bs_fs_yx_bsv16_fsv16
@@ -6440,7 +6414,7 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp16)
         // Generate bias data
         auto biases_size = tensor(1, output_f, 1, 1);
         auto biases_data = generate_random_1d<FLOAT16>(output_f, -1, 1);
-        auto biases_mem = memory::allocate(engine, { data_types::f16, format::bfyx, biases_size });
+        auto biases_mem = engine.allocate_memory({ data_types::f16, format::bfyx, biases_size });
         set_values(biases_mem, biases_data);
 
         // Calculate reference values with bias
@@ -6503,20 +6477,19 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp16)
     network.execute();
 
     auto out_mem = network.get_output("conv_bsv16_fsv16").get_memory();
-    auto out_ptr = out_mem.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr(out_mem, get_test_stream());
 
     auto out_mem_bfyx = network.get_output("reorder_bfyx").get_memory();
-    auto out_ptr_bfyx = out_mem_bfyx.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr_bfyx(out_mem_bfyx, get_test_stream());
 
-    ASSERT_EQ(out_mem.get_layout().format, format::bs_fs_yx_bsv16_fsv16);
+    ASSERT_EQ(out_mem->get_layout().format, format::bs_fs_yx_bsv16_fsv16);
 
     auto flatten_ref = flatten_4d(format::bfyx, reference_result);
 
     for (size_t i = 0; i < out_ptr_bfyx.size(); i++) {
         auto equal = are_equal(flatten_ref[i], out_ptr_bfyx[i], 1);
         EXPECT_TRUE(equal);
-        if (!equal)
-        {
+        if (!equal) {
             std::cout << "Difference at idx = " << i << std::endl;
             return;
         }
@@ -6525,10 +6498,9 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp16)
 
 TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32_fused_ops)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -6547,7 +6519,7 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32_fused_ops)
     auto input_size = tensor(batch_num, input_f, input_xy, input_xy);
     auto input_data = generate_random_4d<float>(batch_num, input_f, input_xy, input_xy, 1, 10);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_xy, filter_xy, 1);
@@ -6555,14 +6527,14 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32_fused_ops)
 
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
 
-    auto weights_mem = memory::allocate(engine, { data_types::f32, format::bfyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f32, format::bfyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<float>(batch_num, VVVF<float>(output_f));
 
     topology topology(
-            input_layout("input", input_mem.get_layout()),
+            input_layout("input", input_mem->get_layout()),
             data("weights", weights_mem));
 
     // Reorder input to bs_fs_yx_bsv16_fsv16
@@ -6573,7 +6545,7 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32_fused_ops)
         // Generate bias data
         auto biases_size = tensor(1, output_f, 1, 1, 1);
         auto biases_data = generate_random_1d<float>(output_f, -1, 1);
-        auto biases_mem = memory::allocate(engine, { data_types::f32, format::bfyx, biases_size });
+        auto biases_mem = engine.allocate_memory({ data_types::f32, format::bfyx, biases_size });
         set_values(biases_mem, biases_data);
 
         // Calculate reference values with bias
@@ -6623,7 +6595,7 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32_fused_ops)
     }
 
     const float scalar = 5.5f;
-    auto scale_mem = memory::allocate(engine, { data_types::f32, format::bfyx, {1, 1, 1, 1} });
+    auto scale_mem = engine.allocate_memory({ data_types::f32, format::bfyx, {1, 1, 1, 1} });
     set_values(scale_mem, {scalar});
 
     topology.add(data("scalar", scale_mem));
@@ -6643,12 +6615,12 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32_fused_ops)
     network.execute();
 
     auto out_mem = network.get_output("conv_bsv16_fsv16").get_memory();
-    auto out_ptr = out_mem.pointer<float>();
+    cldnn::mem_lock<float> out_ptr(out_mem, get_test_stream());
 
     auto out_mem_bfyx = network.get_output("reorder_bfyx").get_memory();
-    auto out_ptr_bfyx = out_mem_bfyx.pointer<float>();
+    cldnn::mem_lock<float> out_ptr_bfyx(out_mem_bfyx, get_test_stream());
 
-    ASSERT_EQ(out_mem.get_layout().format, format::bs_fs_yx_bsv16_fsv16);
+    ASSERT_EQ(out_mem->get_layout().format, format::bs_fs_yx_bsv16_fsv16);
 
     auto flatten_ref = flatten_4d(format::bfyx, reference_result);
 
@@ -6701,10 +6673,9 @@ INSTANTIATE_TEST_CASE_P(convolution_depthwise_gpu_fs_b_yx_fsv32,
 
 TEST_P(convolution_depthwise_gpu, depthwise_conv_fs_b_yx_fsv32)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -6729,20 +6700,20 @@ TEST_P(convolution_depthwise_gpu, depthwise_conv_fs_b_yx_fsv32)
     auto input_size = tensor(batch_num, input_f, input_xy, input_xy);
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_xy, input_xy, -1, 1);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f16, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f16, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(group(groups), batch(1), feature(1), spatial(filter_x, filter_y));
     auto weights_data = generate_random_4d<FLOAT16>(output_f, 1, filter_y, filter_x, -1, 1);
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
-    auto weights_mem = memory::allocate(engine, { data_types::f16, format::goiyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f16, format::goiyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<FLOAT16>(batch_num, VVVF<FLOAT16>(output_f));
 
     topology topology(
-            input_layout("input", input_mem.get_layout()),
+            input_layout("input", input_mem->get_layout()),
             data("weights_fsv", weights_mem));
 
     // Reorder input to fs_byx_fsv32
@@ -6782,9 +6753,9 @@ TEST_P(convolution_depthwise_gpu, depthwise_conv_fs_b_yx_fsv32)
     network.execute();
 
     auto out_mem = network.get_output("conv_fsv").get_memory();
-    auto out_ptr = out_mem.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr(out_mem, get_test_stream());
 
-    ASSERT_EQ(out_mem.get_layout().format, format::fs_b_yx_fsv32);
+    ASSERT_EQ(out_mem->get_layout().format, format::fs_b_yx_fsv32);
 
     for (int bi = 0; bi < batch_num; ++bi)
         for (int fi = 0; fi < output_f; ++fi)
@@ -6844,10 +6815,9 @@ INSTANTIATE_TEST_CASE_P(convolution_depthwise_gpu_b_fs_yx_fsv16,
 
 TEST_P(convolution_depthwise_gpu_fsv16, depthwise_conv_b_fs_yx_fsv16)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -6873,20 +6843,20 @@ TEST_P(convolution_depthwise_gpu_fsv16, depthwise_conv_b_fs_yx_fsv16)
     auto input_size = tensor(batch_num, input_f, input_xy, input_xy);
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_xy, input_xy, -1, 1);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f16, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f16, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(group(output_f), batch(1), feature(1), spatial(filter_x, filter_y));
     auto weights_data = generate_random_4d<FLOAT16>(output_f, 1, filter_y, filter_x, -1, 1);
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
-    auto weights_mem = memory::allocate(engine, { data_types::f16, format::goiyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f16, format::goiyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<FLOAT16>(batch_num, VVVF<FLOAT16>(output_f));
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
         data("weights_fsv", weights_mem));
 
     // Reorder input to b_fs_yx_fsv16
@@ -6926,9 +6896,9 @@ TEST_P(convolution_depthwise_gpu_fsv16, depthwise_conv_b_fs_yx_fsv16)
     network.execute();
 
     auto out_mem = network.get_output("conv_fsv").get_memory();
-    auto out_ptr = out_mem.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr(out_mem, get_test_stream());
 
-    ASSERT_EQ(out_mem.get_layout().format, format::b_fs_yx_fsv16);
+    ASSERT_EQ(out_mem->get_layout().format, format::b_fs_yx_fsv16);
 
     for (int bi = 0; bi < batch_num; ++bi)
         for (int fi = 0; fi < output_f; ++fi)
@@ -6970,11 +6940,11 @@ TEST(convolution_depthwise_gpu_fsv16, depthwise_conv_b_fs_yx_fsv16_in_feature_pa
     auto input_lower_sizes = { 0, 16, 0, 0 };
     auto input_upper_sizes = { 0, 64, 0, 0 };
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, input_size });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, weights_size});
-    auto bias = memory::allocate(engine, { data_types::f32, format::bfyx, bias_size});
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, input_size });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, weights_size});
+    auto bias = engine.allocate_memory({ data_types::f32, format::bfyx, bias_size});
 
     set_values<float>(input, {
          3, -1, -1, -1,  2, -2,  2,  2,  0,  1, -5,  4, -1,  4,  1,  0,
@@ -7002,7 +6972,7 @@ TEST(convolution_depthwise_gpu_fsv16, depthwise_conv_b_fs_yx_fsv16_in_feature_pa
     layout reordered_input_layout = layout(data_types::f32, format::b_fs_yx_fsv16, input_size, input_padding);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("input_reordered", "input", reordered_input_layout),
         data("weights", weights),
         data("bias", bias),
@@ -7022,8 +6992,8 @@ TEST(convolution_depthwise_gpu_fsv16, depthwise_conv_b_fs_yx_fsv16_in_feature_pa
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output_memory = outputs.at("out").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -7055,10 +7025,9 @@ struct convolution_depthwise_gpu_bfyx : public convolution_depthwise_gpu {};
 
 TEST_P(convolution_depthwise_gpu_bfyx, depthwise_conv_bfyx)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -7084,20 +7053,20 @@ TEST_P(convolution_depthwise_gpu_bfyx, depthwise_conv_bfyx)
     auto input_size = tensor(batch_num, input_f, input_xy, input_xy);
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_xy, input_xy, -1, 1);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f16, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f16, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(group(output_f), batch(1), feature(1), spatial(filter_x, filter_y));
     auto weights_data = generate_random_4d<FLOAT16>(output_f, 1, filter_y, filter_x, -1, 1);
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
-    auto weights_mem = memory::allocate(engine, { data_types::f16, format::goiyx, weights_size });
+    auto weights_mem = engine.allocate_memory({ data_types::f16, format::goiyx, weights_size });
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
     auto reference_result = VVVVF<FLOAT16>(batch_num, VVVF<FLOAT16>(output_f));
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
         data("weights", weights_mem));
 
     // Calculate reference values without bias
@@ -7134,9 +7103,9 @@ TEST_P(convolution_depthwise_gpu_bfyx, depthwise_conv_bfyx)
     network.execute();
 
     auto out_mem = network.get_output("conv").get_memory();
-    auto out_ptr = out_mem.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> out_ptr(out_mem, get_test_stream());
 
-    ASSERT_EQ(out_mem.get_layout().format, format::bfyx);
+    ASSERT_EQ(out_mem->get_layout().format, format::bfyx);
 
     for (int bi = 0; bi < batch_num; ++bi)
         for (int fi = 0; fi < output_f; ++fi)
@@ -7260,7 +7229,7 @@ INSTANTIATE_TEST_CASE_P(convolution_grouped_fsv4_fsv16,
                         convolution_grouped_gpu::PrintToStringParamName);
 
 TEST_P(convolution_grouped_gpu, base) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int input_x = testing::get<0>(GetParam()),
               input_y = testing::get<1>(GetParam()),
@@ -7304,7 +7273,7 @@ TEST_P(convolution_grouped_gpu, base) {
                         size_t offset = input_lay.get_linear_offset(coords);
                         input_flat[offset] = input_rnd[b][f][z][y][x];
                     }
-    auto input = memory::allocate(engine, input_lay);
+    auto input = engine.allocate_memory(input_lay);
     set_values(input, input_flat);
 
     auto input_zp_rnd = std::vector<int8_t>(input_f);
@@ -7314,7 +7283,7 @@ TEST_P(convolution_grouped_gpu, base) {
         input_zp_prim_name = { "input_zp" };
     }
     auto input_zp_lay = layout(data_types::i8, format::bfyx, tensor(feature(input_f)));
-    auto input_zp = memory::allocate(engine, input_zp_lay);
+    auto input_zp = engine.allocate_memory(input_zp_lay);
     set_values(input_zp, input_zp_rnd);
 
     auto weights_size = tensor(group(groups), batch(output_f / groups), feature(input_f / groups), spatial(filter_x, filter_y, filter_z));
@@ -7335,7 +7304,7 @@ TEST_P(convolution_grouped_gpu, base) {
                             size_t offset = weights_lay.get_linear_offset(coords);
                             weights_flat[offset] = weights_rnd[gi][ofi][ifi][kzi][kyi][kxi];
                         }
-    auto weights = memory::allocate(engine, weights_lay);
+    auto weights = engine.allocate_memory(weights_lay);
     set_values(weights, weights_flat);
 
     auto weights_zp_rnd = std::vector<int8_t>(output_f);
@@ -7345,7 +7314,7 @@ TEST_P(convolution_grouped_gpu, base) {
         weights_zp_prim_name = { "weights_zp" };
     }
     auto weights_zp_lay = layout(data_types::i8, format::bfyx, tensor(batch(output_f)));
-    auto weights_zp = memory::allocate(engine, weights_zp_lay);
+    auto weights_zp = engine.allocate_memory(weights_zp_lay);
     set_values(weights_zp, weights_zp_rnd);
 
     VVVVVF<float> expected_result(batch_num, VVVVF<float>(output_f));
@@ -7405,7 +7374,7 @@ TEST_P(convolution_grouped_gpu, base) {
         comp_prim_name = { "compensation" };
     }
     auto comp_lay = layout(data_types::f32, format::bfyx, tensor(batch(output_f)));
-    auto comp = memory::allocate(engine, comp_lay);
+    auto comp = engine.allocate_memory(comp_lay);
     set_values(comp, comp_val);
 
     auto stride_tensor = tensor(batch(1), feature(1), spatial(stride, stride, stride, 1));
@@ -7413,7 +7382,7 @@ TEST_P(convolution_grouped_gpu, base) {
         stride_tensor = tensor(batch(1), feature(1), spatial(stride, stride, 1, 1));
     }
 
-    topology topology(input_layout("input", input.get_layout()),
+    topology topology(input_layout("input", input->get_layout()),
                       data("weights", weights),
                       reorder("input_fsv", "input", {data_types::i8, input_data_format, input_size}),
                       convolution("conv",
@@ -7443,6 +7412,7 @@ TEST_P(convolution_grouped_gpu, base) {
 
     build_options options;
     options.set_option(build_option::optimize_data(true));
+    options.set_option(build_option::outputs({"conv", "out"}));
     implementation_desc conv_impl = {input_data_format, impl_name};
     options.set_option(build_option::force_implementations({{"conv", conv_impl}}));
 
@@ -7451,10 +7421,10 @@ TEST_P(convolution_grouped_gpu, base) {
     network.execute();
 
     auto out_mem = network.get_output("conv").get_memory();
-    auto out_ptr = out_mem.pointer<float>();
-    auto out_lay = out_mem.get_layout();
+    cldnn::mem_lock<float> out_ptr(out_mem, get_test_stream());
+    auto out_lay = out_mem->get_layout();
 
-    ASSERT_EQ(out_mem.get_layout().format, input_data_format);
+    ASSERT_EQ(out_mem->get_layout().format, input_data_format);
     ASSERT_EQ(out_lay.size.batch[0], expected_result.size());
     ASSERT_EQ(out_lay.size.feature[0], expected_result[0].size());
     ASSERT_EQ(out_lay.size.spatial[2], expected_result[0][0].size());
@@ -7501,10 +7471,9 @@ INSTANTIATE_TEST_CASE_P(conv_fp16_cases,
                         convolution_general_gpu::PrintToStringParamName);
 
 TEST_P(convolution_general_gpu, conv_fp16_cases) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
-    {
+    if (!engine.get_device_info().supports_fp16) {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
@@ -7529,13 +7498,13 @@ TEST_P(convolution_general_gpu, conv_fp16_cases) {
     auto input_size = tensor(batch_num, input_f, input_x, input_y);
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_y, input_x, -1, 1);
     auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
-    auto input_mem = memory::allocate(engine, { data_types::f16, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f16, format::bfyx, input_size });
     set_values(input_mem, input_data_bfyx);
 
     auto weights_size = tensor(output_f, input_f, filter_y, filter_x, 1);
     auto weights_data = generate_random_4d<FLOAT16>(output_f, input_f, filter_y, filter_x, -1, 1);
     auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
-    auto weights_mem = memory::allocate(engine, {data_types::f16, format::bfyx, weights_size});
+    auto weights_mem = engine.allocate_memory({data_types::f16, format::bfyx, weights_size});
     set_values(weights_mem, weights_data_bfyx);
 
     // Will be used to store reference values calculated in branches depending on bias
@@ -7546,7 +7515,7 @@ TEST_P(convolution_general_gpu, conv_fp16_cases) {
     if (with_bias) {
         auto biases_size = tensor(1, output_f, 1, 1);
         auto biases_data = generate_random_1d<FLOAT16>(output_f, -1, 1);
-        auto biases_mem = memory::allocate(engine, {data_types::f16, format::bfyx, biases_size});
+        auto biases_mem = engine.allocate_memory({data_types::f16, format::bfyx, biases_size});
         set_values(biases_mem, biases_data);
 
         for (auto bi = 0; bi < batch_num; ++bi) {
@@ -7561,7 +7530,7 @@ TEST_P(convolution_general_gpu, conv_fp16_cases) {
             }
         }
 
-        topology.add(input_layout("input", input_mem.get_layout()),
+        topology.add(input_layout("input", input_mem->get_layout()),
                      data("weights_fsv", weights_mem),
                      data("bias", biases_mem),
                      reorder("input_fsv", "input", {data_types::f16, input_data_format, input_size}));
@@ -7589,7 +7558,7 @@ TEST_P(convolution_general_gpu, conv_fp16_cases) {
             }
         }
 
-        topology.add(input_layout("input", input_mem.get_layout()),
+        topology.add(input_layout("input", input_mem->get_layout()),
                      data("weights_fsv", weights_mem),
                      reorder("input_fsv", "input", {data_types::f16, input_data_format, input_size}));
 
@@ -7612,10 +7581,10 @@ TEST_P(convolution_general_gpu, conv_fp16_cases) {
     network.execute();
 
     auto out_mem = network.get_output("conv_fsv").get_memory();
-    auto out_ptr = out_mem.pointer<FLOAT16>();
-    auto out_lay = out_mem.get_layout();
+    cldnn::mem_lock<FLOAT16> out_ptr(out_mem, get_test_stream());
+    auto out_lay = out_mem->get_layout();
 
-    ASSERT_EQ(out_mem.get_layout().format, input_data_format);
+    ASSERT_EQ(out_mem->get_layout().format, input_data_format);
     ASSERT_EQ(out_lay.size.batch[0], expected_result.size());
     ASSERT_EQ(out_lay.size.feature[0], expected_result[0].size());
     ASSERT_EQ(out_lay.size.spatial[1], expected_result[0][0].size());
@@ -7643,11 +7612,11 @@ TEST_P(convolution_general_gpu, conv_fp16_cases) {
 template <typename InputT, typename WeightsT, typename OutputT>
 class convolution_test_base {
 public:
-    virtual topology build_topology(const cldnn::engine& engine) {
+    virtual topology build_topology(cldnn::engine& engine) {
         auto input_lay = layout(input_type(), format::bfyx, input_size(), padding_size());
         auto wei_lay = layout(weights_type(), format::bfyx, weights_size());
 
-        auto wei_mem = memory::allocate(engine, wei_lay);
+        auto wei_mem = engine.allocate_memory(wei_lay);
         auto weights_flat = flatten_4d(format::bfyx, _weights);
         set_values(wei_mem, weights_flat);
         layout reordered_layout = layout{input_type(), input_format(), input_size(), padding_size()};
@@ -7657,7 +7626,7 @@ class convolution_test_base {
         std::string input_id = "input_reorder";
         if (has_input_zp()) {
             auto input_zp_lay = layout(input_type(), format::bfyx, tensor(feature(input_features())));
-            auto input_zp_mem = memory::allocate(engine, input_zp_lay);
+            auto input_zp_mem = engine.allocate_memory(input_zp_lay);
             set_values(input_zp_mem, _input_zp);
             topo.add(data("input_zp", input_zp_mem));
             topo.add(eltwise("input_asymm", { "input_reorder", "input_zp" }, eltwise_mode::sub));
@@ -7667,7 +7636,7 @@ class convolution_test_base {
         std::string weights_id = "weights";
         if (has_weights_zp()) {
             auto weights_zp_lay = layout(weights_type(), format::bfyx, tensor(batch(output_features())));
-            auto weights_zp_mem = memory::allocate(engine, weights_zp_lay);
+            auto weights_zp_mem = engine.allocate_memory(weights_zp_lay);
             set_values(weights_zp_mem, _weights_zp);
             topo.add(data("weights_zp", weights_zp_mem));
             topo.add(eltwise("weights_asymm", { "weights", "weights_zp" }, eltwise_mode::sub));
@@ -7686,7 +7655,7 @@ class convolution_test_base {
             topo.add(conv_prim);
         } else {
             auto bias_lay = layout(output_type(), format::bfyx, tensor(feature(output_features())));
-            auto bias_mem = memory::allocate(engine, bias_lay);
+            auto bias_mem = engine.allocate_memory(bias_lay);
             set_values(bias_mem, _bias);
             topo.add(data("bias", bias_mem));
             auto conv_prim = convolution(
@@ -7710,7 +7679,7 @@ class convolution_test_base {
     }
 
     virtual void run_expect(const VVVVF<OutputT>& expected) {
-        auto engine = get_test_engine();
+        auto& engine = get_test_engine();
 
         auto topo = build_topology(engine);
 
@@ -7723,7 +7692,7 @@ class convolution_test_base {
         auto net = network(prog, 0);
 
         auto input_lay = layout(input_type(), format::bfyx, input_size(), padding_size());
-        auto input_mem = memory::allocate(engine, input_lay);
+        auto input_mem = engine.allocate_memory(input_lay);
         std::vector<InputT> input_flat(input_lay.get_linear_size(), static_cast<InputT>(0));
         for (size_t bi = 0; bi < batch_num(); ++bi)
             for (size_t fi = 0; fi < input_features(); ++fi)
@@ -7738,8 +7707,8 @@ class convolution_test_base {
         net.set_input_data("input", input_mem);
         auto result = net.execute();
         auto out_mem = result.at(output_primitive_id()).get_memory();
-        auto out_lay = out_mem.get_layout();
-        auto out_ptr = out_mem.cldnn::memory::template pointer<OutputT>();
+        auto out_lay = out_mem->get_layout();
+        cldnn::mem_lock<OutputT> out_ptr(out_mem, get_test_stream());
 
         std::stringstream description;
         for (auto i : net.get_primitives_info()) {
@@ -8008,11 +7977,11 @@ template <typename InputT, typename WeightsT, typename OutputT>
 class convolution_random_test_fsv4_input : public convolution_random_test_base<InputT, WeightsT, OutputT> {
 public:
     using parent = convolution_random_test_base<InputT, WeightsT, OutputT>;
-    topology build_topology(const cldnn::engine& engine) override {
+    topology build_topology(cldnn::engine& engine) override {
         auto input_lay = layout(this->input_type(), format::b_fs_yx_fsv4, this->input_size(), this->padding_size());
         auto wei_lay = layout(this->weights_type(), format::bfyx, this->weights_size());
 
-        auto wei_mem = memory::allocate(engine, wei_lay);
+        auto wei_mem = engine.allocate_memory(wei_lay);
         auto wei_flat = flatten_4d(format::bfyx, this->_weights);
         set_values(wei_mem, wei_flat);
         layout reordered_layout = layout{this->input_type(), this->input_format(), this->input_size(), this->padding_size()};
@@ -8022,7 +7991,7 @@ class convolution_random_test_fsv4_input : public convolution_random_test_base<I
         std::string input_id = "input_reorder";
         if (this->has_input_zp()) {
             auto input_zp_lay = layout(this->input_type(), format::bfyx, tensor(feature(this->input_features())));
-            auto input_zp_mem = memory::allocate(engine, input_zp_lay);
+            auto input_zp_mem = engine.allocate_memory(input_zp_lay);
             set_values(input_zp_mem, this->_input_zp);
             topo.add(data("input_zp", input_zp_mem));
             topo.add(eltwise("input_asymm", { "input_reorder", "input_zp" }, eltwise_mode::sub));
@@ -8032,7 +8001,7 @@ class convolution_random_test_fsv4_input : public convolution_random_test_base<I
         std::string weights_id = "weights";
         if (this->has_weights_zp()) {
             auto weights_zp_lay = layout(this->weights_type(), format::bfyx, tensor(batch(this->output_features())));
-            auto weights_zp_mem = memory::allocate(engine, weights_zp_lay);
+            auto weights_zp_mem = engine.allocate_memory(weights_zp_lay);
             set_values(weights_zp_mem, this->_weights_zp);
             topo.add(data("weights_zp", weights_zp_mem));
             topo.add(eltwise("weights_asymm", { "weights", "weights_zp" }, eltwise_mode::sub));
@@ -8051,7 +8020,7 @@ class convolution_random_test_fsv4_input : public convolution_random_test_base<I
             topo.add(conv_prim);
         } else {
             auto bias_lay = layout(this->output_type(), format::bfyx, tensor(feature(this->output_features())));
-            auto bias_mem = memory::allocate(engine, bias_lay);
+            auto bias_mem = engine.allocate_memory(bias_lay);
             set_values(bias_mem, this->_bias);
             topo.add(data("bias", bias_mem));
             auto conv_prim = convolution(
@@ -8070,7 +8039,7 @@ class convolution_random_test_fsv4_input : public convolution_random_test_base<I
         return topo;
     }
     void run_expect(const VVVVF<OutputT>& expected) override {
-        auto engine = get_test_engine();
+        auto& engine = get_test_engine();
 
         auto topo = this->build_topology(engine);
 
@@ -8083,7 +8052,7 @@ class convolution_random_test_fsv4_input : public convolution_random_test_base<I
         auto net = network(prog, 0);
 
         auto input_lay = layout(this->input_type(), format::b_fs_yx_fsv4,  this->input_size(), this->padding_size());
-        auto input_mem = memory::allocate(engine, input_lay);
+        auto input_mem = engine.allocate_memory(input_lay);
         std::vector<InputT> input_flat(input_lay.get_linear_size(), static_cast<InputT>(0));
         for (size_t bi = 0; bi < this->batch_num(); ++bi)
             for (size_t fi = 0; fi < this->input_features(); ++fi)
@@ -8098,8 +8067,8 @@ class convolution_random_test_fsv4_input : public convolution_random_test_base<I
         net.set_input_data("input", input_mem);
         auto result = net.execute();
         auto out_mem = result.at(this->output_primitive_id()).get_memory();
-        auto out_lay = out_mem.get_layout();
-        auto out_ptr = out_mem.cldnn::memory::template pointer<OutputT>();
+        auto out_lay = out_mem->get_layout();
+        cldnn::mem_lock<OutputT> out_ptr(out_mem, get_test_stream());
 
         std::stringstream description;
         for (auto i : net.get_primitives_info()) {
@@ -8142,14 +8111,14 @@ class convolution_scale_random_test : public convolution_random_test_base<InputT
         return "scale_wa_reorder";
     }
 
-    topology build_topology(const cldnn::engine& engine) override {
+    topology build_topology(cldnn::engine& engine) override {
         topology topo = parent::build_topology(engine);
 
         auto scale_lay = layout(this->output_type(), format::bfyx, tensor(batch(1), feature(this->output_features())));
         auto shift_lay = layout(this->output_type(), format::bfyx, tensor(batch(1), feature(this->output_features())));
 
-        auto scale_mem = memory::allocate(engine, scale_lay);
-        auto shift_mem = memory::allocate(engine, shift_lay);
+        auto scale_mem = engine.allocate_memory(scale_lay);
+        auto shift_mem = engine.allocate_memory(shift_lay);
 
         set_values(scale_mem, _scale);
         set_values(shift_mem, _shift);
@@ -8397,20 +8366,17 @@ INSTANTIATE_TEST_CASE_P(
     to_string_convolution_all_params
 );
 
-class convolution_test : public tests::generic_test
-{
+class convolution_test : public tests::generic_test {
 
 public:
 
-    static void TearDownTestCase()
-    {
+    static void TearDownTestCase() {
         all_generic_params.clear();
         all_layer_params.clear();
         all_test_params.clear();
     }
 
-    static std::vector<std::shared_ptr<cldnn::primitive>> generate_specific_test_params()
-    {
+    static std::vector<std::shared_ptr<cldnn::primitive>> generate_specific_test_params() {
         // TODO: check split
 
         // TODO: check convolution without bias
@@ -8443,8 +8409,7 @@ class convolution_test : public tests::generic_test
         return all_layer_params;
     }
 
-    static std::vector<std::tuple<std::shared_ptr<tests::test_params>, std::shared_ptr<cldnn::primitive>>> generate_all_test_params()
-    {
+    static std::vector<std::tuple<std::shared_ptr<tests::test_params>, std::shared_ptr<cldnn::primitive>>> generate_all_test_params() {
         generate_specific_test_params();
 
         std::vector<cldnn::format> input_formats = { cldnn::format::bfyx, cldnn::format::yxfb };
@@ -8457,23 +8422,16 @@ class convolution_test : public tests::generic_test
 
         auto data_types = test_data_types();
 
-        for (cldnn::data_types data_type : data_types)
-        {
-            for (cldnn::format input_format : input_formats)
-            {
-                for (cldnn::format weights_format : weights_formats)
-                {
+        for (cldnn::data_types data_type : data_types) {
+            for (cldnn::format input_format : input_formats) {
+                for (cldnn::format weights_format : weights_formats) {
                     cldnn::build_options network_build_options;
-                    if (input_format == cldnn::format::bfyx)
-                    {
+                    if (input_format == cldnn::format::bfyx) {
                         network_build_options.set_option(cldnn::build_option::optimize_data(true));
                     }
-                    for (cldnn::tensor input_size : input_tensor_size)
-                    {
-                        for (cldnn::tensor kernel_size : kernel_sizes)
-                        {
-                            for (auto output_features : output_features_sizes)
-                            {
+                    for (cldnn::tensor input_size : input_tensor_size) {
+                        for (cldnn::tensor kernel_size : kernel_sizes) {
+                            for (auto output_features : output_features_sizes) {
                                 std::shared_ptr<tests::test_params> params = std::make_shared<test_params>(data_type, input_format, input_size.batch[0], input_size.feature[0], tensor(1, 1, input_size.spatial[0], input_size.spatial[1]), network_build_options);
                                 int input_features = params->input_layouts[0].size.feature[0];
                                 params->input_layouts.push_back(cldnn::layout(params->data_type, weights_format, cldnn::tensor(output_features, input_features, kernel_size.spatial[0], kernel_size.spatial[1]))); // weights
@@ -8487,10 +8445,8 @@ class convolution_test : public tests::generic_test
         }
 
         // Create all the combinations for the test.
-        for (const auto& layer_param : all_layer_params)
-        {
-            for (auto test_param : all_generic_params)
-            {
+        for (const auto& layer_param : all_layer_params) {
+            for (auto test_param : all_generic_params) {
                 all_test_params.push_back(std::make_tuple(test_param, layer_param));
             }
         }
@@ -8498,13 +8454,11 @@ class convolution_test : public tests::generic_test
         return all_test_params;
     }
 
-    virtual bool is_format_supported(cldnn::format format)
-    {
+    virtual bool is_format_supported(cldnn::format format) {
         return ((format == cldnn::format::bfyx) || (format == cldnn::format::yxfb));
     }
 
-    virtual cldnn::tensor get_expected_output_tensor()
-    {
+    virtual cldnn::tensor get_expected_output_tensor() {
         auto convolution = std::static_pointer_cast<const cldnn::convolution>(layer_params);
         tensor input_size = generic_params->input_layouts[0].size;
         tensor dilation = convolution->dilation;
@@ -8523,58 +8477,52 @@ class convolution_test : public tests::generic_test
         return cldnn::tensor(input_size.batch[0], output_features, output_size_x, output_size_y);
     }
 
-    virtual void prepare_input_for_test(std::vector<cldnn::memory>& inputs)
-    {
-        if (generic_params->data_type == data_types::f32)
-        {
+    virtual void prepare_input_for_test(std::vector<cldnn::memory::ptr>& inputs) {
+        if (generic_params->data_type == data_types::f32) {
             prepare_input_for_test_typed<float>(inputs);
-        }
-        else
-        {
+        } else {
             prepare_input_for_test_typed<FLOAT16>(inputs);
         }
     }
 
     template<typename Type>
-    void prepare_input_for_test_typed(std::vector<cldnn::memory>& inputs)
-    {
+    void prepare_input_for_test_typed(std::vector<cldnn::memory::ptr>& inputs) {
         int k = (generic_params->data_type == data_types::f32) ? 8 : 4;
 
         // Update inputs.
         auto input = inputs[0];
-        auto input_size = inputs[0].get_layout().size;
+        auto input_size = inputs[0]->get_layout().size;
         VVVVF<Type> input_rnd = generate_random_4d<Type>(input_size.batch[0], input_size.feature[0], input_size.spatial[1], input_size.spatial[0], -2, 2, k);
-        VF<Type> input_rnd_vec = flatten_4d<Type>(input.get_layout().format, input_rnd);
+        VF<Type> input_rnd_vec = flatten_4d<Type>(input->get_layout().format, input_rnd);
         set_values(input, input_rnd_vec);
 
         // Update weights.
         auto weight_input = inputs[1];
-        auto weight_size = inputs[1].get_layout().size;
+        auto weight_size = inputs[1]->get_layout().size;
         VVVVF<Type> weight_rnd = generate_random_4d<Type>(weight_size.batch[0], weight_size.feature[0], weight_size.spatial[1], weight_size.spatial[0], -2, 2, k);
-        VF<Type> weight_rnd_vec = flatten_4d<Type>(weight_input.get_layout().format, weight_rnd);
+        VF<Type> weight_rnd_vec = flatten_4d<Type>(weight_input->get_layout().format, weight_rnd);
         set_values(weight_input, weight_rnd_vec);
 
         // Update biases.
         auto bias_input = inputs[2];
-        auto bias_size = inputs[2].get_layout().size;
+        auto bias_size = inputs[2]->get_layout().size;
         VF<Type> bias_rnd = generate_random_1d<Type>(bias_size.spatial[0], -2, 2, k);
         set_values(bias_input, bias_rnd);
     }
 
     template<typename Type>
-    memory generate_reference_typed(const std::vector<cldnn::memory>& inputs)
-    {
+    memory::ptr generate_reference_typed(const std::vector<cldnn::memory::ptr>& inputs) {
         // Output reference is always bfyx.
 
         auto convolution = std::static_pointer_cast<const cldnn::convolution>(layer_params);
 
-        data_types dt = inputs[0].get_layout().data_type;
+        data_types dt = inputs[0]->get_layout().data_type;
 
-        tensor input_size = inputs[0].get_layout().size;
+        tensor input_size = inputs[0]->get_layout().size;
         tensor dilation = convolution->dilation;
         tensor stride = convolution->stride;
         tensor input_offset = convolution->input_offset;
-        tensor weights_size = inputs[1].get_layout().size;
+        tensor weights_size = inputs[1]->get_layout().size;
         padding output_padding = convolution->output_padding;
 
         tensor output_size = get_expected_output_tensor();
@@ -8585,27 +8533,23 @@ class convolution_test : public tests::generic_test
         int output_features = weights_size.batch[0];
         int input_features = weights_size.feature[0];
 
-        auto output = memory::allocate( engine, cldnn::layout(dt, cldnn::format::bfyx, output_size, output_padding) );
+        auto output = engine.allocate_memory(cldnn::layout(dt, cldnn::format::bfyx, output_size, output_padding));
 
-        auto input_mem = inputs[0].pointer<Type>();
-        auto weights_mem = inputs[1].pointer<Type>();
-        auto bias_mem = inputs[2].pointer<Type>();
-        auto output_mem = output.pointer<Type>();
+        cldnn::mem_lock<Type> input_mem(inputs[0], get_test_stream());
+        cldnn::mem_lock<Type> weights_mem(inputs[1], get_test_stream());
+        cldnn::mem_lock<Type> bias_mem(inputs[2], get_test_stream());
+        cldnn::mem_lock<Type> output_mem(output, get_test_stream());
 
-        tensor output_buffer_size = output.get_layout().get_buffer_size();
+        tensor output_buffer_size = output->get_layout().get_buffer_size();
 
         // Initialized output with zeros.
         std::fill(output_mem.begin(), output_mem.end(), static_cast<Type>(0));
 
         // Add the bias
-        for (int b = 0; b < input_size.batch[0]; b++)
-        {
-            for (int out_f = 0; out_f < output_features; out_f++)
-            {
-                for (int y = 0; y < output_size_y; y++)
-                {
-                    for (int x = 0; x < output_size_x; x++)
-                    {
+        for (int b = 0; b < input_size.batch[0]; b++) {
+            for (int out_f = 0; out_f < output_features; out_f++) {
+                for (int y = 0; y < output_size_y; y++) {
+                    for (int x = 0; x < output_size_x; x++) {
                         int output_index = (b * output_buffer_size.feature[0] + out_f) * output_buffer_size.spatial[1] * output_buffer_size.spatial[0];
                         tensor lower_output_padding = convolution->output_padding.lower_size();
                         output_index += (lower_output_padding.spatial[1] + y) * output_buffer_size.spatial[0] + lower_output_padding.spatial[0] + x;
@@ -8616,22 +8560,17 @@ class convolution_test : public tests::generic_test
             }
         }
 
-        const auto input0_desc = get_linear_memory_desc(inputs[0].get_layout());
-        const auto input1_desc = get_linear_memory_desc(inputs[1].get_layout());
+        const auto input0_desc = get_linear_memory_desc(inputs[0]->get_layout());
+        const auto input1_desc = get_linear_memory_desc(inputs[1]->get_layout());
 
         // Convolve with weights
-        for (int b = 0; b < input_size.batch[0]; b++)
-        {
+        for (int b = 0; b < input_size.batch[0]; b++) {
             int input_bi = b;
-            for (int out_f = 0; out_f < output_features; out_f++)
-            {
-                for (int in_f = 0; in_f < input_features; in_f++)
-                {
+            for (int out_f = 0; out_f < output_features; out_f++) {
+                for (int in_f = 0; in_f < input_features; in_f++) {
                     int input_fi = in_f;
-                    for (int y = 0; y < output_size_y; y++)
-                    {
-                        for (int x = 0; x < output_size_x; x++)
-                        {
+                    for (int y = 0; y < output_size_y; y++) {
+                        for (int x = 0; x < output_size_x; x++) {
                             int output_bi = b;
                             int output_fi = out_f;
                             int output_yi = y;
@@ -8640,29 +8579,25 @@ class convolution_test : public tests::generic_test
                             tensor lower_output_padding = convolution->output_padding.lower_size();
                             output_index += (lower_output_padding.spatial[1] + output_yi) * output_buffer_size.spatial[0] + lower_output_padding.spatial[0] + output_xi;
 
-                            for (int kernel_y = 0; kernel_y < weights_size.spatial[1]; kernel_y++)
-                            {
+                            for (int kernel_y = 0; kernel_y < weights_size.spatial[1]; kernel_y++) {
                                 int input_yi = y * stride.spatial[1] + input_offset.spatial[1] + kernel_y * dilation.spatial[1];
-                                if ((input_yi < 0) || (input_yi >= input_size.spatial[1]))
-                                {
+                                if ((input_yi < 0) || (input_yi >= input_size.spatial[1])) {
                                     continue;
                                 }
 
-                                for (int kernel_x = 0; kernel_x < weights_size.spatial[0]; kernel_x++)
-                                {
+                                for (int kernel_x = 0; kernel_x < weights_size.spatial[0]; kernel_x++) {
                                     int input_xi = x * stride.spatial[0] + input_offset.spatial[0] + kernel_x * dilation.spatial[0];
-                                    if ((input_xi < 0) || (input_xi >= input_size.spatial[0]))
-                                    {
+                                    if ((input_xi < 0) || (input_xi >= input_size.spatial[0])) {
                                         continue;
                                     }
 
-                                    size_t input_index = get_linear_index(inputs[0].get_layout(), input_bi, input_fi, input_yi, input_xi, input0_desc);
+                                    size_t input_index = get_linear_index(inputs[0]->get_layout(), input_bi, input_fi, input_yi, input_xi, input0_desc);
 
                                     int weight_bi = out_f;
                                     int weight_fi = in_f;
                                     int weight_yi = kernel_y;
                                     int weight_xi = kernel_x;
-                                    size_t weight_index = get_linear_index(inputs[1].get_layout(), weight_bi, weight_fi, weight_yi, weight_xi, input1_desc);
+                                    size_t weight_index = get_linear_index(inputs[1]->get_layout(), weight_bi, weight_fi, weight_yi, weight_xi, input1_desc);
                                     output_mem[output_index] += input_mem[input_index] * weights_mem[weight_index];
                                 }
                             }
@@ -8675,14 +8610,10 @@ class convolution_test : public tests::generic_test
         return output;
     }
 
-    virtual memory generate_reference(const std::vector<cldnn::memory>& inputs)
-    {
-        if (generic_params->data_type == data_types::f32)
-        {
+    virtual memory::ptr generate_reference(const std::vector<cldnn::memory::ptr>& inputs) {
+        if (generic_params->data_type == data_types::f32) {
             return generate_reference_typed<float>(inputs);
-        }
-        else
-        {
+        } else {
             return generate_reference_typed<FLOAT16>(inputs);
         }
     }
@@ -8698,8 +8629,7 @@ std::vector<std::shared_ptr<tests::test_params>> convolution_test::all_generic_p
 std::vector<std::shared_ptr<cldnn::primitive>> convolution_test::all_layer_params = {};
 std::vector<std::tuple<std::shared_ptr<tests::test_params>, std::shared_ptr<cldnn::primitive>>> convolution_test::all_test_params = {};
 
-TEST_P(convolution_test, CONVOLUTION)
-{
+TEST_P(convolution_test, CONVOLUTION) {
     run_single_test();
 }
 
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/crop_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/crop_gpu_test.cpp
index da96bcab6f490c..e5b6ed490a9e3c 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/crop_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/crop_gpu_test.cpp
@@ -3,19 +3,15 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/crop.hpp"
-#include <api/eltwise.hpp>
-#include <api/reorder.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/crop.hpp>
+#include <cldnn/primitives/eltwise.hpp>
+#include <cldnn/primitives/reorder.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 template<typename T>
 std::vector<T> generate_random_input(size_t b, size_t f, size_t z, size_t y, size_t x, int min, int max) {
@@ -40,7 +36,7 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_all) {
     //  Input      : 2x3x4x5
     //  Output     : 1x2x2x3
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 2;
     auto feature_num = 3;
@@ -52,10 +48,10 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_all) {
     auto crop_x_size = x_size - 2;
     auto crop_y_size = y_size - 2;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size }, { 0, 0, 0, 0 }));
 
     std::vector<float> input_vec = generate_random_input<float>(batch_num, feature_num, y_size, x_size, -10, 10);
@@ -68,7 +64,7 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_all) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
@@ -84,7 +80,7 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_all) {
 }
 
 TEST(crop_gpu, basic_in2x2x2x3_crop_all) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 2;
     auto feature_num = 2;
@@ -96,10 +92,10 @@ TEST(crop_gpu, basic_in2x2x2x3_crop_all) {
     auto crop_x_size = x_size - 1;
     auto crop_y_size = y_size - 1;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size }, { 0, 0, 0, 0 }));
 
     std::vector<float> input_vec;
@@ -114,7 +110,7 @@ TEST(crop_gpu, basic_in2x2x2x3_crop_all) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     printf("Results:\n");
     for (int b = 0; b < crop_batch_num; ++b) { //B
@@ -135,7 +131,7 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_all) {
     //  Input      : 2x3x4x5
     //  Output     : 1x2x2x3
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 2;
     auto feature_num = 3;
@@ -147,10 +143,10 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_all) {
     auto crop_x_size = x_size - 2;
     auto crop_y_size = y_size - 2;
 
-    auto input = memory::allocate(engine, { data_types::i32, format::yxfb,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::i32, format::yxfb,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size }, { 0, 0, 0, 0 }));
 
     std::vector<int32_t> input_vec = generate_random_input<int32_t>(batch_num, feature_num, y_size, x_size, -10, 10);
@@ -163,7 +159,7 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_all) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<int32_t>();
+    cldnn::mem_lock<int32_t> output_ptr(output, get_test_stream());
 
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
@@ -183,7 +179,7 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_all) {
     //  Input      : 2x3x4x5
     //  Output     : 1x2x2x3
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 2;
     auto feature_num = 3;
@@ -195,10 +191,10 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_all) {
     auto crop_x_size = x_size - 2;
     auto crop_y_size = y_size - 2;
 
-    auto input = memory::allocate(engine, { data_types::i64, format::yxfb,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::i64, format::yxfb,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size }, { 0, 0, 0, 0 }));
 
     std::vector<int64_t> input_vec = generate_random_input<int64_t>(batch_num, feature_num, y_size, x_size, -10, 10);
@@ -211,7 +207,7 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_all) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<int64_t>();
+    cldnn::mem_lock<int64_t> output_ptr(output, get_test_stream());
 
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
@@ -231,7 +227,7 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_all_bfyx) {
     //  Input      : 6x2x4x3
     //  Output     : 3x1x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 2;
@@ -243,10 +239,10 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_all_bfyx) {
     auto crop_x_size = x_size - 2;
     auto crop_y_size = y_size - 1;
 
-    auto input = memory::allocate(engine, { data_types::f32,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::f32,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size }, {0, 0, 0, 0} ));
 
     std::vector<float> input_vec = generate_random_input<float>(batch_num, feature_num, y_size, x_size, -10, 10);
@@ -259,7 +255,7 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_all_bfyx) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     std::vector<float> a;
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
@@ -280,7 +276,7 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_all_bfyx) {
     //  Input      : 6x2x4x3
     //  Output     : 3x1x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 2;
@@ -292,10 +288,10 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_all_bfyx) {
     auto crop_x_size = x_size - 2;
     auto crop_y_size = y_size - 1;
 
-    auto input = memory::allocate(engine, { data_types::i32,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::i32,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size }, { 0, 0, 0, 0 }));
 
     std::vector<int32_t> input_vec = generate_random_input<int32_t>(batch_num, feature_num, y_size, x_size, -10, 10);
@@ -308,7 +304,7 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_all_bfyx) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<int32_t>();
+    cldnn::mem_lock<int32_t> output_ptr(output, get_test_stream());
     std::vector<int32_t> a;
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
@@ -329,7 +325,7 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_all_bfyx) {
     //  Input      : 6x2x4x3
     //  Output     : 3x1x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 2;
@@ -341,10 +337,10 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_all_bfyx) {
     auto crop_x_size = x_size - 2;
     auto crop_y_size = y_size - 1;
 
-    auto input = memory::allocate(engine, { data_types::i64,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::i64,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size }, { 0, 0, 0, 0 }));
 
     std::vector<int64_t> input_vec = generate_random_input<int64_t>(batch_num, feature_num, y_size, x_size, -10, 10);
@@ -357,7 +353,7 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_all_bfyx) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<int64_t>();
+    cldnn::mem_lock<int64_t> output_ptr(output, get_test_stream());
     std::vector<int64_t> a;
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
@@ -378,7 +374,7 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_all_fyxb) {
     //  Input      : 6x2x4x3
     //  Output     : 3x1x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 2;
@@ -390,10 +386,10 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_all_fyxb) {
     auto crop_x_size = x_size - 2;
     auto crop_y_size = y_size - 1;
 
-    auto input = memory::allocate(engine, { data_types::f32,format::fyxb,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::f32,format::fyxb,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size }, {0, 0, 0, 0} ));
 
     std::vector<float> input_vec = generate_random_input<float>(batch_num, feature_num, y_size, x_size, -10, 10);
@@ -406,7 +402,7 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_all_fyxb) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
             for (int y = 0; y < crop_y_size; ++y) { //Y
@@ -425,7 +421,7 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_all_fyxb) {
     //  Input      : 6x2x4x3
     //  Output     : 3x1x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 2;
@@ -437,10 +433,10 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_all_fyxb) {
     auto crop_x_size = x_size - 2;
     auto crop_y_size = y_size - 1;
 
-    auto input = memory::allocate(engine, { data_types::i32,format::fyxb,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::i32,format::fyxb,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size }, { 0, 0, 0, 0 }));
 
     std::vector<int32_t> input_vec = generate_random_input<int32_t>(batch_num, feature_num, y_size, x_size, -10, 10);
@@ -453,7 +449,7 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_all_fyxb) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<int32_t>();
+    cldnn::mem_lock<int32_t> output_ptr(output, get_test_stream());
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
             for (int y = 0; y < crop_y_size; ++y) { //Y
@@ -472,7 +468,7 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_all_fyxb) {
     //  Input      : 6x2x4x3
     //  Output     : 3x1x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 2;
@@ -484,10 +480,10 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_all_fyxb) {
     auto crop_x_size = x_size - 2;
     auto crop_y_size = y_size - 1;
 
-    auto input = memory::allocate(engine, { data_types::i64,format::fyxb,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::i64,format::fyxb,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size }, { 0, 0, 0, 0 }));
 
     std::vector<int64_t> input_vec = generate_random_input<int64_t>(batch_num, feature_num, y_size, x_size, -10, 10);
@@ -500,7 +496,7 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_all_fyxb) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<int64_t>();
+    cldnn::mem_lock<int64_t> output_ptr(output, get_test_stream());
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
             for (int y = 0; y < crop_y_size; ++y) { //Y
@@ -526,7 +522,7 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_offsets) {
     //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
     //  f1: b0:  7    8  -16   b1:   12   8    -17
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 2;
     auto feature_num = 2;
@@ -543,10 +539,10 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_offsets) {
     auto x_offset = 1;
     auto y_offset = 1;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num)), { tensor(feature(0)) }));
 
     std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f,
@@ -564,7 +560,7 @@ TEST(crop_gpu, basic_in2x3x2x2_crop_offsets) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
@@ -591,7 +587,7 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_offsets) {
     //  f1: b0:  5    6  -12   b1:   15   52   -13
     //  f1: b0:  7    8  -16   b1:   12   8    -17
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 2;
     auto feature_num = 2;
@@ -608,10 +604,10 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_offsets) {
     auto x_offset = 1;
     auto y_offset = 1;
 
-    auto input = memory::allocate(engine, { data_types::i32, format::yxfb,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine.allocate_memory({ data_types::i32, format::yxfb,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num)), { tensor(feature(0)) }));
 
     std::vector<int32_t> input_vec = { 1, 0, 5, 15,
@@ -629,7 +625,7 @@ TEST(crop_gpu, basic_i32_in2x3x2x2_crop_offsets) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<int32_t>();
+    cldnn::mem_lock<int32_t> output_ptr(output, get_test_stream());
 
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
@@ -656,7 +652,7 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_offsets) {
     //  f1: b0:  5    6  -12   b1:   15   52   -13
     //  f1: b0:  7    8  -16   b1:   12   8    -17
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 2;
     auto feature_num = 2;
@@ -673,10 +669,10 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_offsets) {
     auto x_offset = 1;
     auto y_offset = 1;
 
-    auto input = memory::allocate(engine, { data_types::i64, format::yxfb,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine.allocate_memory({ data_types::i64, format::yxfb,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num)), { tensor(feature(0)) }));
 
     std::vector<int64_t> input_vec = { 1, 0, 5, 15,
@@ -694,7 +690,7 @@ TEST(crop_gpu, basic_i64_in2x3x2x2_crop_offsets) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<int64_t>();
+    cldnn::mem_lock<int64_t> output_ptr(output, get_test_stream());
 
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
@@ -738,7 +734,7 @@ TEST(crop_gpu, basic_in1x4x1x1_split) {
 
     //  Out2:
     //  f0: 4.0
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 1;
     auto feature_num = 4;
@@ -752,10 +748,10 @@ TEST(crop_gpu, basic_in1x4x1x1_split) {
     auto crop_y_size = 1;
     auto feature_offset_1 = 0;
     auto feature_offset_2 = 3;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop1", "input", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num_1)), { tensor(feature(feature_offset_1), spatial(0,0),batch(0)) }));
     topology.add(crop("crop2", "input", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num_2)), { tensor(feature(feature_offset_2), spatial(0,0),batch(0)) }));
 
@@ -772,21 +768,21 @@ TEST(crop_gpu, basic_in1x4x1x1_split) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop1").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < out1.size();i++)
         EXPECT_EQ(output_ptr[i], out1[i]);
 
     std::cout << std::endl;
     auto output_2 = outputs.at("crop2").get_memory();
-    auto output_ptr_2 = output_2.pointer<float>();
+    cldnn::mem_lock<float> output_ptr_2(output_2, get_test_stream());
 
     for (size_t i = 0; i < out2.size();i++)
         EXPECT_EQ(output_ptr_2[i], out2[i]);
 }
 
 TEST(crop_gpu, basic_in1x4x1x1_crop_pad) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 1;
     auto feature_num = 4;
@@ -798,12 +794,12 @@ TEST(crop_gpu, basic_in1x4x1x1_crop_pad) {
     auto crop_x_size = 1;
     auto crop_y_size = 1;
     auto feature_offset_1 = 0;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     padding in_pad({0, 0, 1, 1}, {0, 0, 1, 1});
-    auto padded_layout = input.get_layout().with_padding(in_pad);
+    auto padded_layout = input->get_layout().with_padding(in_pad);
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reorder("input_reorder", "input", padded_layout));
     topology.add(crop("crop1", "input_reorder", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num_1)), { tensor(feature(feature_offset_1), spatial(0,0),batch(0)) }));
     topology.add(reorder("out_reorder", "crop1", format::bfyx, data_types::f32));
@@ -819,7 +815,7 @@ TEST(crop_gpu, basic_in1x4x1x1_crop_pad) {
     auto outputs = network.execute();
 
     auto output = outputs.at("out_reorder").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < out1.size();i++)
         EXPECT_EQ(output_ptr[i], out1[i]);
@@ -854,7 +850,7 @@ TEST(crop_gpu, basic_i32_in1x4x1x1_split) {
 
     //  Out2:
     //  f0: 4
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 1;
     auto feature_num = 4;
@@ -868,10 +864,10 @@ TEST(crop_gpu, basic_i32_in1x4x1x1_split) {
     auto crop_y_size = 1;
     auto feature_offset_1 = 0;
     auto feature_offset_2 = 3;
-    auto input = memory::allocate(engine, { data_types::i32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine.allocate_memory({ data_types::i32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop1", "input", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num_1)), { tensor(feature(feature_offset_1), spatial(0,0),batch(0)) }));
     topology.add(crop("crop2", "input", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num_2)), { tensor(feature(feature_offset_2), spatial(0,0),batch(0)) }));
 
@@ -888,14 +884,13 @@ TEST(crop_gpu, basic_i32_in1x4x1x1_split) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop1").get_memory();
-    auto output_ptr = output.pointer<int32_t>();
+    cldnn::mem_lock<int32_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < out1.size(); i++)
         EXPECT_EQ(output_ptr[i], out1[i]);
 
-    std::cout << std::endl;
     auto output_2 = outputs.at("crop2").get_memory();
-    auto output_ptr_2 = output_2.pointer<int32_t>();
+    cldnn::mem_lock<int32_t> output_ptr_2(output_2, get_test_stream());
 
     for (size_t i = 0; i < out2.size(); i++)
         EXPECT_EQ(output_ptr_2[i], out2[i]);
@@ -930,7 +925,7 @@ TEST(crop_gpu, basic_i64_in1x4x1x1_split) {
 
     //  Out2:
     //  f0: 4
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 1;
     auto feature_num = 4;
@@ -944,10 +939,10 @@ TEST(crop_gpu, basic_i64_in1x4x1x1_split) {
     auto crop_y_size = 1;
     auto feature_offset_1 = 0;
     auto feature_offset_2 = 3;
-    auto input = memory::allocate(engine, { data_types::i64, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine.allocate_memory({ data_types::i64, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop1", "input", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num_1)), { tensor(feature(feature_offset_1), spatial(0,0),batch(0)) }));
     topology.add(crop("crop2", "input", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num_2)), { tensor(feature(feature_offset_2), spatial(0,0),batch(0)) }));
 
@@ -964,14 +959,13 @@ TEST(crop_gpu, basic_i64_in1x4x1x1_split) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop1").get_memory();
-    auto output_ptr = output.pointer<int64_t>();
+    cldnn::mem_lock<int64_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < out1.size(); i++)
         EXPECT_EQ(output_ptr[i], out1[i]);
 
-    std::cout << std::endl;
     auto output_2 = outputs.at("crop2").get_memory();
-    auto output_ptr_2 = output_2.pointer<int64_t>();
+    cldnn::mem_lock<int64_t> output_ptr_2(output_2, get_test_stream());
 
     for (size_t i = 0; i < out2.size(); i++)
         EXPECT_EQ(output_ptr_2[i], out2[i]);
@@ -1007,8 +1001,8 @@ TEST(crop_gpu, basic_in1x4x1x1_split_w_relu) {
     //  Out2:
     //  f0: 4.0
     // disable memory pool when we want to check optimized out internal results
-    engine_configuration cfg{ false, false, false, std::string(), std::string(), true /*oooq*/, std::string(),std::string(), priority_mode_types::disabled,  throttle_mode_types::disabled, false /*mem_pool*/ };
-    engine engine{ cfg };
+    engine_configuration cfg{ false, queue_types::out_of_order, std::string(), priority_mode_types::disabled,  throttle_mode_types::disabled, false /*mem_pool*/ };
+    auto engine = engine::create(engine_types::ocl, runtime_types::ocl, cfg);
     auto batch_num = 1;
     auto feature_num = 4;
     auto x_size = 1;
@@ -1020,10 +1014,10 @@ TEST(crop_gpu, basic_in1x4x1x1_split_w_relu) {
     auto crop_y_size = 1;
     auto feature_offset_1 = 0;
     auto feature_offset_2 = 3;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine->allocate_memory({ data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(activation("relu", "input", activation_func::relu));
     topology.add(crop("crop1", "relu", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num_1)), { tensor(feature(feature_offset_1), spatial(0,0),batch(0)) }));
     topology.add(crop("crop2", "relu", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size), feature(crop_feature_num_2)), { tensor(feature(feature_offset_2), spatial(0,0),batch(0)) }));
@@ -1038,22 +1032,22 @@ TEST(crop_gpu, basic_in1x4x1x1_split_w_relu) {
     bo.set_option(build_option::optimize_data(true));
     bo.set_option(build_option::debug(true)); //required to have optimized crop despite the fact that it's specified as an output
 
-    network network(engine, topology, bo);
+    network network(*engine, topology, bo);
     network.set_input_data("input", input);
     auto outputs = network.execute();
 
     auto output = outputs.at("relu1").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     // check if crop has been executed in place
-    auto in_place = outputs.at("crop1").get_memory().is_the_same_buffer(outputs.at("relu").get_memory());
+    auto in_place = engine->is_the_same_buffer(*outputs.at("crop1").get_memory(), *outputs.at("relu").get_memory());
     EXPECT_TRUE(in_place);
 
     for (size_t i = 0; i < out1.size();i++)
         EXPECT_EQ(output_ptr[i], out1[i]);
 
     auto output_2 = outputs.at("relu2").get_memory();
-    auto output_ptr_2 = output_2.pointer<float>();
+    cldnn::mem_lock<float> output_ptr_2(output_2, get_test_stream());
 
     for (size_t i = 0; i < out2.size();i++)
         EXPECT_EQ(output_ptr_2[i], out2[i]);
@@ -1064,7 +1058,7 @@ TEST(crop_gpu, basic_in3x1x2x2x1_crop_all_bfzyx) {
     //  Input      : 6x2x4x3x2
     //  Output     : 3x1x2x2x1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 2;
@@ -1078,10 +1072,10 @@ TEST(crop_gpu, basic_in3x1x2x2x1_crop_all_bfzyx) {
     auto crop_y_size = y_size - 1;
     auto crop_z_size = z_size - 1;
 
-    auto input = memory::allocate(engine, { data_types::f32,format::bfzyx,{ batch_num, feature_num, x_size, y_size, z_size } });
+    auto input = engine.allocate_memory({ data_types::f32,format::bfzyx,{ batch_num, feature_num, x_size, y_size, z_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", { crop_batch_num, crop_feature_num, crop_x_size, crop_y_size, crop_z_size }, { 0, 0, 0, 0, 0 }));
 
     std::vector<float> input_vec = generate_random_input<float>(batch_num, feature_num, y_size, x_size, -10, 10);
@@ -1094,7 +1088,7 @@ TEST(crop_gpu, basic_in3x1x2x2x1_crop_all_bfzyx) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
             for (int z = 0; z < crop_z_size; ++z) { //Z
@@ -1115,7 +1109,7 @@ TEST(crop_gpu, basic_in3x1x3x2x2x1_crop_all_bfwzyx) {
     //  Input      : 6x2x6x4x3x2
     //  Output     : 3x1x3x2x2x1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 2;
@@ -1133,10 +1127,10 @@ TEST(crop_gpu, basic_in3x1x3x2x2x1_crop_all_bfwzyx) {
 
     tensor in_size = tensor(format::bfwzyx, { batch_num, feature_num, w_size, z_size, y_size, x_size });
     tensor crop_size = tensor(format::bfwzyx, { crop_batch_num, crop_feature_num, crop_w_size, crop_z_size, crop_y_size, crop_x_size });
-    auto input = memory::allocate(engine, { data_types::f32,format::bfwzyx, in_size });
+    auto input = engine.allocate_memory({ data_types::f32,format::bfwzyx, in_size });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(crop("crop", "input", crop_size, tensor{ 0 }));
 
     VVVVVVF<float> input_rnd = generate_random_6d<float>(batch_num, feature_num, w_size, z_size, y_size, x_size, -10, 10);
@@ -1150,7 +1144,7 @@ TEST(crop_gpu, basic_in3x1x3x2x2x1_crop_all_bfwzyx) {
     auto outputs = network.execute();
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int b = 0; b < crop_batch_num; ++b) { //B
         for (int f = 0; f < crop_feature_num; ++f) { //F
             for (int w = 0; w < crop_w_size; ++w) { //W
@@ -1176,7 +1170,7 @@ class crop_gpu : public ::testing::TestWithParam<crop_test_params> {};
 TEST_P(crop_gpu, pad_test) {
     auto p = GetParam();
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = std::get<0>(p);
     auto feature_num = std::get<1>(p);
@@ -1194,10 +1188,10 @@ TEST_P(crop_gpu, pad_test) {
     auto in_out_format = std::get<3>(p).first;
     auto crop_format = std::get<3>(p).second;
 
-    auto input = memory::allocate(engine, { data_types::f32, in_out_format, { tensor(spatial(x_size, y_size, z_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine.allocate_memory({ data_types::f32, in_out_format, { tensor(spatial(x_size, y_size, z_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reorder("reorder", "input", crop_format, data_types::f32));
     topology.add(crop("crop1", "reorder", tensor(batch(crop_batch_num), spatial(crop_x_size, crop_y_size, crop_z_size), feature(crop_feature_num_1)), { tensor(feature(feature_offset_1), spatial(0,0,0), batch(0)) }));
     topology.add(reorder("out", "crop1", in_out_format, data_types::f32));
@@ -1225,7 +1219,7 @@ TEST_P(crop_gpu, pad_test) {
     auto outputs = network.execute();
 
     auto output = outputs.at("out").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < res.size(); i++)
         EXPECT_EQ(output_ptr[i], res[i]);
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/cum_sum_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/cum_sum_gpu_test.cpp
index bb40415d16aed1..0020c53076a293 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/cum_sum_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/cum_sum_gpu_test.cpp
@@ -3,21 +3,18 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/input_layout.hpp>
-#include "api/cum_sum.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/data.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/cum_sum.hpp>
+#include <cldnn/primitives/data.hpp>
 
 #include <algorithm>
 #include <fstream>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 template<typename T = float>
 static std::vector<T> cumsum(const std::vector<T>& input,
@@ -150,7 +147,7 @@ class cum_sum_gpu : public ::testing::TestWithParam<cum_sum_test_params> {};
 
 TEST_P(cum_sum_gpu, basic_test) {
     auto p = GetParam();
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto b = std::get<0>(p);
     auto f = std::get<1>(p);
@@ -169,14 +166,14 @@ TEST_P(cum_sum_gpu, basic_test) {
     else if (in_out_format == format::bfzyx)
         size = 5;
 
-    auto input = memory::allocate(engine, { data_types::f32, in_out_format, shape });
+    auto input = engine.allocate_memory({ data_types::f32, in_out_format, shape });
     const int inputSize = b * f * w * z * y * x;
     auto inputVals = generateVector(inputSize);
 
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(cum_sum("cum_sum", "Input0", getCumSumAxis(axis, size), exclusive, reverse));
 
     network network(engine, topology);
@@ -189,7 +186,7 @@ TEST_P(cum_sum_gpu, basic_test) {
     EXPECT_EQ(outputs.begin()->first, "cum_sum");
 
     auto output = outputs.at("cum_sum").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     auto answers = cumsum(inputVals, in_out_format, { b, f, w, z, y, x }, axis, exclusive, reverse);
     ASSERT_EQ(output_ptr.size(), answers.size());
@@ -306,21 +303,22 @@ INSTANTIATE_TEST_CASE_P(
             ::testing::ValuesIn(variants)
             ), );
 
-TEST(cum_sum_gpu_f16, basic_1d) {
+// FIXME: This test fails on some driver versions. Looks like UB in impl or driver issue
+TEST(cum_sum_gpu_f16, DISABLED_basic_1d) {
     // Input : 5x1x1x1
     // Output : 5x1x1x1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor shape = { 5, 1, 1, 1 };
     std::vector<float> inputVals = {
         1.0f, 2.0f, 3.0f, 4.0f, 5.0f
     };
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, shape });
 
     set_values(input, vectorCast<FLOAT16>(inputVals));
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(cum_sum("cum_sum", "Input0"));
 
     network network(engine, topology);
@@ -333,65 +331,12 @@ TEST(cum_sum_gpu_f16, basic_1d) {
     EXPECT_EQ(outputs.begin()->first, "cum_sum");
 
     auto output = outputs.at("cum_sum").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     auto answers = cumsum(inputVals, format::bfyx, { 5, 1, 1, 1, 1, 1 });
 
-    ASSERT_EQ(output_ptr.size(), answers.size());
-    for (size_t i = 0; i < answers.size(); ++i)
-    {
+    ASSERT_EQ(output->count(), answers.size());
+    for (size_t i = 0; i < answers.size(); ++i) {
         EXPECT_TRUE(are_equal(answers[i], float16_to_float32(output_ptr[i]))) << i;
     }
 }
-
-TEST(cum_sum_gpu_f32, perf) {
-    // Input : 384x160x160x1
-    // Output : 384x160x160x1
-
-    constexpr int batch = 384;
-    constexpr int features = 160;
-    constexpr int y = 160;
-    constexpr int x = 1;
-    engine_configuration configuration(true);
-    engine engine(configuration);
-    tensor shape = { batch, features, y, x };
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
-    constexpr int inputSize = batch * features * y * x;
-    auto inputVals = generateVector(inputSize);
-
-    set_values(input, inputVals);
-
-    topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
-    topology.add(cum_sum("cum_sum", "Input0"));
-
-    network network(engine, topology);
-
-    network.set_input_data("Input0", input);
-
-    auto outputs = network.execute();
-
-    EXPECT_EQ(outputs.size(), size_t(1));
-    EXPECT_EQ(outputs.begin()->first, "cum_sum");
-
-    auto output = outputs.at("cum_sum").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    auto profilingTime = [](const primitive_id& id, const event& ev) {
-            cldnn::instrumentation::profiling_info cldnnInfo{id, ev.get_profiling_info()};
-            long long time = 0;
-            for (auto &interval : cldnnInfo.intervals) {
-                using duration_t = std::chrono::duration<long long, std::chrono::microseconds::period>;
-                time += std::chrono::duration_cast<duration_t>(interval.value->value()).count();
-            }
-            return time;
-        };
-
-
-    auto ep = network.get_executed_primitives();
-    auto cumSumEP = ep.find("cum_sum");
-    ASSERT_NE(cumSumEP, ep.end()) << "Cannot find 'cum_sum' id in executed primitives";
-
-    auto time = profilingTime(cumSumEP->first, cumSumEP->second);
-    std::cout << "Time, id: " << cumSumEP->first << ", time: " << time << std::endl;
-}
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/custom_gpu_primitive_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/custom_gpu_primitive_test.cpp
index a484f20291a126..f93009ee6397be 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/custom_gpu_primitive_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/custom_gpu_primitive_test.cpp
@@ -2,18 +2,12 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/eltwise.hpp"
-#include "api/reorder.hpp"
-#include "api/custom_gpu_primitive.hpp"
-#include <api/engine.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/eltwise.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/custom_gpu_primitive.hpp>
 
 namespace cldnn
 {
@@ -21,7 +15,7 @@ namespace cldnn
 }
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 TEST(custom_gpu_primitive_f32, add_basic_in2x2x2x2) {
     //  Input2   : 2x2x2
@@ -29,28 +23,28 @@ TEST(custom_gpu_primitive_f32, add_basic_in2x2x2x2) {
     //  Output : 2x2x2x2
 
     //  Input:
-    //  f0: b0:  1    2  b1:   0    0       
-    //  f0: b0:  3    4  b1:   0.5 -0.5     
-    //  f1: b0:  5    6  b1:   1.5  5.2     
-    //  f1: b0:  7    8  b1:   12   8       
+    //  f0: b0:  1    2  b1:   0    0
+    //  f0: b0:  3    4  b1:   0.5 -0.5
+    //  f1: b0:  5    6  b1:   1.5  5.2
+    //  f1: b0:  7    8  b1:   12   8
     //
     //  Input2
-    //  f0: b0: 0.5  5   b1: 2.5  7 
+    //  f0: b0: 0.5  5   b1: 2.5  7
     //  f0: b0: 15  -2   b1: 17   6.5
     //  f1: b0: 0.5  2   b1: 2.5  4
     //  f1: b0: 8   -0.5 b1: 10   -2.5
     //
     //  Output:
-    //  f0: b0:   1.5  7    b1:  2.5   7      
-    //  f0: b0:   18   2    b1:  17.5  6     
-    //  f1: b0:   5.5  8    b1:   4    9.2     
-    //  f1: b0:   15  16.5  b1:  22    16.5     
+    //  f0: b0:   1.5  7    b1:  2.5   7
+    //  f0: b0:   18   2    b1:  17.5  6
+    //  f1: b0:   5.5  8    b1:   4    9.2
+    //  f1: b0:   15  16.5  b1:  22    16.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
 
     std::string kernel_code =
         R"__krnl(
@@ -68,10 +62,10 @@ TEST(custom_gpu_primitive_f32, add_basic_in2x2x2x2) {
     layout output_layout = { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } };
     std::vector<size_t> gws = { output_layout.count() };
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(custom_gpu_primitive(
-        "user_kernel", 
+        "user_kernel",
         { "input", "input2" },
         { kernel_code },
         entry_point,
@@ -109,10 +103,9 @@ TEST(custom_gpu_primitive_f32, add_basic_in2x2x2x2) {
                           18.f,17.5f,   15.f,   22.f,
                           2.f,   6.f,   7.5f,  5.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
-    for (int i = 0; i < 16; i++)
-    {
+    for (int i = 0; i < 16; i++) {
         EXPECT_TRUE(are_equal(answers[i], output_ptr[i]));
     }
 }
@@ -125,28 +118,28 @@ void add_basic_in2x2x2x2_with_reorder()
     //  Output : 2x2x2x2
 
     //  Input:
-    //  f0: b0:  1    2  b1:   0    0       
-    //  f0: b0:  3    4  b1:   0.5 -0.5     
-    //  f1: b0:  5    6  b1:   1.5  5.2     
-    //  f1: b0:  7    8  b1:   12   8       
+    //  f0: b0:  1    2  b1:   0    0
+    //  f0: b0:  3    4  b1:   0.5 -0.5
+    //  f1: b0:  5    6  b1:   1.5  5.2
+    //  f1: b0:  7    8  b1:   12   8
     //
     //  Input2
-    //  f0: b0: 0.5  5   b1: 2.5  7 
+    //  f0: b0: 0.5  5   b1: 2.5  7
     //  f0: b0: 15  -2   b1: 17   6.5
     //  f1: b0: 0.5  2   b1: 2.5  4
     //  f1: b0: 8   -0.5 b1: 10   -2.5
     //
     //  Output:
-    //  f0: b0:   1.5  7    b1:  2.5   7      
-    //  f0: b0:   18   2    b1:  17.5  6     
-    //  f1: b0:   5.5  8    b1:   4    9.2     
-    //  f1: b0:   15  16.5  b1:  22    16.5     
+    //  f0: b0:   1.5  7    b1:  2.5   7
+    //  f0: b0:   18   2    b1:  17.5  6
+    //  f1: b0:   5.5  8    b1:   4    9.2
+    //  f1: b0:   15  16.5  b1:  22    16.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
     std::string data_type_string = "float";
     switch (DType)
@@ -172,8 +165,8 @@ void add_basic_in2x2x2x2_with_reorder()
     layout output_layout = { DType, format::yxfb,{ 2, 2, 2, 2 } };
     std::vector<size_t> gws = { output_layout.count() };
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(reorder("to_int1", "input", { DType, format::yxfb,{ 2,2,2,2 } }));
     topology.add(reorder("to_int2", "input2", { DType, format::yxfb,{ 2,2,2,2 } }));
     topology.add(custom_gpu_primitive(
@@ -216,7 +209,7 @@ void add_basic_in2x2x2x2_with_reorder()
         18.f,17.f,   15.f,   22.f,
         2.f,   6.f,   8.f,  6.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -238,28 +231,28 @@ TEST(custom_gpu_primitive_f32, eltwise_add_basic_in2x2x2x2) {
     //  Output : 2x2x2x2
 
     //  Input:
-    //  f0: b0:  1    2  b1:   0    0       
-    //  f0: b0:  3    4  b1:   0.5 -0.5     
-    //  f1: b0:  5    6  b1:   1.5  5.2     
-    //  f1: b0:  7    8  b1:   12   8       
+    //  f0: b0:  1    2  b1:   0    0
+    //  f0: b0:  3    4  b1:   0.5 -0.5
+    //  f1: b0:  5    6  b1:   1.5  5.2
+    //  f1: b0:  7    8  b1:   12   8
     //
     //  Input2
-    //  f0: b0: 0.5  5   b1: 2.5  7 
+    //  f0: b0: 0.5  5   b1: 2.5  7
     //  f0: b0: 15  -2   b1: 17   6.5
     //  f1: b0: 0.5  2   b1: 2.5  4
     //  f1: b0: 8   -0.5 b1: 10   -2.5
     //
     //  Output:
-    //  f0: b0:   1.5  7    b1:  2.5   7      
-    //  f0: b0:   18   2    b1:  17.5  6     
-    //  f1: b0:   5.5  8    b1:   4    9.2     
-    //  f1: b0:   15  16.5  b1:  22    16.5     
+    //  f0: b0:   1.5  7    b1:  2.5   7
+    //  f0: b0:   18   2    b1:  17.5  6
+    //  f1: b0:   5.5  8    b1:   4    9.2
+    //  f1: b0:   15  16.5  b1:  22    16.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
     std::string kernel_code =
         R"__krnl(
@@ -274,8 +267,8 @@ TEST(custom_gpu_primitive_f32, eltwise_add_basic_in2x2x2x2) {
     layout output_layout = { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } };
     std::vector<size_t> gws = { output_layout.count() };
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::sum));
     topology.add(custom_gpu_primitive(
         "user_kernel",
@@ -311,13 +304,13 @@ TEST(custom_gpu_primitive_f32, eltwise_add_basic_in2x2x2x2) {
 
     auto output = outputs.at("user_kernel").get_memory();
 
-    float answers[16] = 
+    float answers[16] =
       { 2.5f,  3.5f,  6.5f,   5.f,
          8.f,   8.f,   9.f, 10.2f,
         19.f, 18.5f,  16.f,  23.f,
          3.f,   7.f,  8.5f,  6.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -331,28 +324,28 @@ TEST(custom_gpu_primitive_f32, add_eltwise_basic_in2x2x2x2) {
     //  Output : 2x2x2x2
 
     //  Input:
-    //  f0: b0:  1    2  b1:   0    0       
-    //  f0: b0:  3    4  b1:   0.5 -0.5     
-    //  f1: b0:  5    6  b1:   1.5  5.2     
-    //  f1: b0:  7    8  b1:   12   8       
+    //  f0: b0:  1    2  b1:   0    0
+    //  f0: b0:  3    4  b1:   0.5 -0.5
+    //  f1: b0:  5    6  b1:   1.5  5.2
+    //  f1: b0:  7    8  b1:   12   8
     //
     //  Input2
-    //  f0: b0: 0.5  5   b1: 2.5  7 
+    //  f0: b0: 0.5  5   b1: 2.5  7
     //  f0: b0: 15  -2   b1: 17   6.5
     //  f1: b0: 0.5  2   b1: 2.5  4
     //  f1: b0: 8   -0.5 b1: 10   -2.5
     //
     //  Output:
-    //  f0: b0:   1.5  7    b1:  2.5   7      
-    //  f0: b0:   18   2    b1:  17.5  6     
-    //  f1: b0:   5.5  8    b1:   4    9.2     
-    //  f1: b0:   15  16.5  b1:  22    16.5     
+    //  f0: b0:   1.5  7    b1:  2.5   7
+    //  f0: b0:   18   2    b1:  17.5  6
+    //  f1: b0:   5.5  8    b1:   4    9.2
+    //  f1: b0:   15  16.5  b1:  22    16.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
     std::string kernel_code =
         R"__krnl(
@@ -367,8 +360,8 @@ TEST(custom_gpu_primitive_f32, add_eltwise_basic_in2x2x2x2) {
     layout output_layout = { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } };
     std::vector<size_t> gws = { output_layout.count() };
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(custom_gpu_primitive(
         "user_kernel",
         { "input" },
@@ -410,7 +403,7 @@ TEST(custom_gpu_primitive_f32, add_eltwise_basic_in2x2x2x2) {
         19.f, 18.5f,  16.f,  23.f,
         3.f,   7.f,  8.5f,  6.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -424,27 +417,27 @@ TEST(custom_gpu_primitive_f32, two_kernels_with_same_entry_point_basic_in2x2x2x2
     //  Output : 2x2x2x2
 
     //  Input:
-    //  f0: b0:  1    2  b1:   0    0       
-    //  f0: b0:  3    4  b1:   0.5 -0.5     
-    //  f1: b0:  5    6  b1:   1.5  5.2     
-    //  f1: b0:  7    8  b1:   12   8       
+    //  f0: b0:  1    2  b1:   0    0
+    //  f0: b0:  3    4  b1:   0.5 -0.5
+    //  f1: b0:  5    6  b1:   1.5  5.2
+    //  f1: b0:  7    8  b1:   12   8
     //
     //  Input2
-    //  f0: b0: 0.5  5   b1: 2.5  7 
+    //  f0: b0: 0.5  5   b1: 2.5  7
     //  f0: b0: 15  -2   b1: 17   6.5
     //  f1: b0: 0.5  2   b1: 2.5  4
     //  f1: b0: 8   -0.5 b1: 10   -2.5
     //
     //  Output:
-    //  f0: b0:   1.5  7    b1:  2.5   7      
-    //  f0: b0:   18   2    b1:  17.5  6     
-    //  f1: b0:   5.5  8    b1:   4    9.2     
-    //  f1: b0:   15  16.5  b1:  22    16.5     
+    //  f0: b0:   1.5  7    b1:  2.5   7
+    //  f0: b0:   18   2    b1:  17.5  6
+    //  f1: b0:   5.5  8    b1:   4    9.2
+    //  f1: b0:   15  16.5  b1:  22    16.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
     std::string kernel_code1 =
         R"__krnl(
@@ -454,7 +447,7 @@ TEST(custom_gpu_primitive_f32, two_kernels_with_same_entry_point_basic_in2x2x2x2
                 output[idx] = input0[idx] + SCALAR;
             }
         )__krnl";
-    
+
     std::string kernel_code2 =
         R"__krnl(
             __kernel void add_kernel(const __global float* input0, __global float* output)
@@ -468,7 +461,7 @@ TEST(custom_gpu_primitive_f32, two_kernels_with_same_entry_point_basic_in2x2x2x2
     layout output_layout = { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } };
     std::vector<size_t> gws = { output_layout.count() };
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(custom_gpu_primitive(
         "user_kernel1",
         { "input" },
@@ -505,20 +498,19 @@ TEST(custom_gpu_primitive_f32, two_kernels_with_same_entry_point_basic_in2x2x2x2
 
     auto output = outputs.at("user_kernel2").get_memory();
 
-    auto output_ptr = output.pointer<float>();
-    auto input_ptr = input.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
-    for (int i = 0; i < 16; i++)
-    {
+    for (int i = 0; i < 16; i++) {
         EXPECT_TRUE(are_equal(input_ptr[i] + 7, output_ptr[i]));
     }
 }
 
 TEST(custom_gpu_primitive_u8, add_basic_in2x2x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::u8, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::u8, format::yxfb,{ 2, 2, 2, 2 } });
 
     std::string kernel_code =
         R"__krnl(
@@ -533,8 +525,8 @@ TEST(custom_gpu_primitive_u8, add_basic_in2x2x2x2) {
     layout output_layout = { data_types::u8, format::yxfb,{ 2, 2, 2, 2 } };
     std::vector<size_t> gws = { output_layout.count() };
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(custom_gpu_primitive(
         "user_kernel",
         { "input", "input2" },
@@ -556,7 +548,7 @@ TEST(custom_gpu_primitive_u8, add_basic_in2x2x2x2) {
          0,  2,  0,  2,
         55, 75, 20,  4,
         15, 17, 80, 10,
-         2, 60,  0, 20 
+         2, 60,  0, 20
     });
 
     network network(engine, topology);
@@ -577,10 +569,9 @@ TEST(custom_gpu_primitive_u8, add_basic_in2x2x2x2) {
           6, 160,   8, 200
     };
 
-    auto output_ptr = output.pointer<unsigned char>();
+    cldnn::mem_lock<unsigned char> output_ptr(output, get_test_stream());
 
-    for (int i = 0; i < 16; i++)
-    {
+    for (int i = 0; i < 16; i++) {
         EXPECT_TRUE(are_equal(answers[i], output_ptr[i]));
     }
 }
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/deconvolution_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/deconvolution_gpu_test.cpp
index a5789188fbf506..eb82022c8b389e 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/deconvolution_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/deconvolution_gpu_test.cpp
@@ -4,26 +4,20 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/deconvolution.hpp"
-#include "api/crop.hpp"
-#include <api/data.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
 #include "test_utils/test_utils.h"
-#include "test_utils/float16.h"
-#include "api/reorder.hpp"
-#include "src/include/to_string_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/deconvolution.hpp>
+#include <cldnn/primitives/crop.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
 
 namespace cldnn {
 template<> struct type_to_data_type<FLOAT16> { static const data_types value = data_types::f16; };
 }
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 template <typename InputT>
 struct deconvolution_traits {
@@ -147,18 +141,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_nopad) {
     //   18    0.75  7.25
     //   23    42.5  15.5
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f });
     set_values(biases, { 2.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1,1,1,1 })
@@ -173,7 +167,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -14.f, 5.f, 2.25f,
@@ -208,16 +202,16 @@ TEST(deconvolution_f32_fw_gpu, no_bias_basic_wsiz2x2_in2x2x1x1_nopad) {
     //   18    0.75  7.25
     //   23    42.5  15.5
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx,{ 1, 1, 2, 2 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         deconvolution("deconv", "input", { "weights" })
     );
@@ -231,7 +225,7 @@ TEST(deconvolution_f32_fw_gpu, no_bias_basic_wsiz2x2_in2x2x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -16.f, 3.f, 0.25f,
@@ -265,18 +259,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_nopad_bfyx) {    //  Filt
     //   18    0.75  7.25
     //   23    42.5  15.5
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f });
     set_values(biases, { 2.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1,1,1,1 })
@@ -291,7 +285,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_nopad_bfyx) {    //  Filt
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -14.f, 5.f, 2.25f,
@@ -325,18 +319,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_pad1) {
     //  Output:
     //  0.75
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f });
     set_values(biases, { 2.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1, 1, 1, 1 }, { 0, 0, -1, -1 })
@@ -351,7 +345,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_pad1) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_FLOAT_EQ(0.75f, output_ptr[0]);
 }
@@ -376,18 +370,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride2_nopad) {
     //  Output:
     //  0.75
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f });
     set_values(biases, { 1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1,1,2,2 })
@@ -402,7 +396,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride2_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -15.f, 5.f, 0.f, 1.25f,
@@ -441,18 +435,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride4_pad2) {
     //  0    0    0
     //  6    0   -18
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx, { 1, 1, 3, 3 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx, { 1, 1, 3, 3 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.0f, 0.5f, 1.f, 3.5f, 1.5f, 2.f, 3.f, 4.f, 5.f });
     set_values(biases, { 0.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1, 1, 4, 4 }, { 0, 0, -2, -2 })
@@ -467,7 +461,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride4_pad2) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         40.f, 0.f, 1.5f,
@@ -503,18 +497,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x2_stride2_pad1) {
     //  -3    4.5    0.5   22
     //   13  -17     5    -7
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 1.f, 0.5f, 3.f, 6.f, 2.f, 9.f, 4.f });
     set_values(weights, { -2.f, 2.f, 7.f, -0.5f });
     set_values(biases, { 1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1, 1, 2, 2 }, { 0, 0, -1, -1 })
@@ -529,7 +523,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x2_stride2_pad1) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 0.5f, 4.5f, 22.f,
@@ -568,20 +562,20 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2x2_in2x2x1x1_stride2_pad1) {
     //  f1: 1    8.5
     //  f1: 17 - 13
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     cldnn::build_options options;
     options.set_option(cldnn::build_option::optimize_data(true));
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::yxio, { 2, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::yxio, { 2, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.f, -2.f, 2.f, 2.f, 7.f, 7.f, -0.5f, -0.5f });
     set_values(biases, { 1.0f, 5.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1, 1, 2, 2 }, { 0, 0, -1, -1 })
@@ -596,7 +590,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2x2_in2x2x1x1_stride2_pad1) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 1.f, 4.5f, 8.5f,
@@ -631,18 +625,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x2_bfyx_stride2_pad1) {
     //  -3    4.5    0.5   22
     //   13  -17     5    -7
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f });
     set_values(weights, { -2.f, 2.f, 7.f, -0.5f });
     set_values(biases, { 1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1, 1, 2, 2 }, { 0, 0, -1, -1 })
@@ -657,7 +651,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x2_bfyx_stride2_pad1) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 4.5f, 13.f, -17.f,
@@ -693,19 +687,19 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x2_bfyx_stride2_pad1_input_p
     //  -3    4.5    0.5   22
     //   13  -17     5    -7
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx,{ 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx,{ 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f });
     set_values(weights, { -2.f, 2.f, 7.f, -0.5f });
     set_values(biases, { 1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
-        reorder("reorder", "input", input.get_layout().with_padding(padding{ { 0, 0, 1, 2 }, 0 })),
+        input_layout("input", input->get_layout()),
+        reorder("reorder", "input", input->get_layout().with_padding(padding{ { 0, 0, 1, 2 }, 0 })),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "reorder", { "weights" }, { "biases" }, { 1, 1, 2, 2 }, { 0, 0, -1, -1 })
@@ -720,7 +714,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x2_bfyx_stride2_pad1_input_p
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 4.5f, 13.f, -17.f,
@@ -760,21 +754,21 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2x2_in2x2x1x1_stride2_pad1_input_padd
     //  f1: 1    8.5
     //  f1: 17 - 13
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     cldnn::build_options options;
     options.set_option(cldnn::build_option::optimize_data(true));
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::yxio,{ 2, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::yxio,{ 2, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.f, -2.f, 2.f, 2.f, 7.f, 7.f, -0.5f, -0.5f });
     set_values(biases, { 1.0f, 5.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
-        reorder("reorder", "input", input.get_layout().with_padding(padding{ { 0, 0, 1, 2 }, 0 })),
+        input_layout("input", input->get_layout()),
+        reorder("reorder", "input", input->get_layout().with_padding(padding{ { 0, 0, 1, 2 }, 0 })),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "reorder", { "weights" }, { "biases" }, { 1, 1, 2, 2 }, { 0, 0, -1, -1 })
@@ -789,7 +783,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2x2_in2x2x1x1_stride2_pad1_input_padd
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 1.f, 4.5f, 8.5f,
@@ -824,18 +818,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x2_bfyx_yxfb_stride2_pad1) {
     //  -3    4.5    0.5   22
     //   13  -17     5    -7
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f });
     set_values(weights, { -2.f, 2.f, 7.f, -0.5f });
     set_values(biases, { 1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1, 1, 2, 2 }, { 0, 0, -1, -1 })
@@ -850,7 +844,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x2_bfyx_yxfb_stride2_pad1) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 4.5f, 13.f, -17.f,
@@ -885,11 +879,11 @@ TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in2x2x1x2_bfyx_yxfb_stride2_pad1) {
     //  -3    4.5    0.5   22
     //   13  -17     5    -7
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx,{ 2, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx,{ 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx,{ 2, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx,{ 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     cldnn::build_options options;
     options.set_option(cldnn::build_option::optimize_data(true));
@@ -904,7 +898,7 @@ TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in2x2x1x2_bfyx_yxfb_stride2_pad1) {
     set_values(biases, { 1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1, 1, 2, 2 }, { 0, 0, -1, -1 })
@@ -919,7 +913,7 @@ TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in2x2x1x2_bfyx_yxfb_stride2_pad1) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 4.5f, 13.f, -17.f,
@@ -961,18 +955,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_split2)
     //  -3    4.5    -8   -28
     //   13  -17     1    -17
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f });
     set_values(weights, { -2.f, 2.f, 7.f, -0.5f, -4.f, 1.f, -9.f, -7.f });
     set_values(biases, { 1.0f, -1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, 2, { 1, 1, 2, 2 }, { 0, 0, -1, -1 })
@@ -987,7 +981,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_split2)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 4.5f, 13.f, -17.f,
@@ -1003,11 +997,11 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_split2)
 TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_group2) {
     //  data is similar as in basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_split2
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f });
     set_values(weights, {
@@ -1017,7 +1011,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_group2)
     set_values(biases, { 1.0f, -1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, 2, { 1, 1, 2, 2 }, { 0, 0, -1, -1 })
@@ -1032,7 +1026,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_group2)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 4.5f, 13.f, -17.f,
@@ -1049,9 +1043,9 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_group16
     //  Test for depthwise separable optimization, there are 16 joined weights and biases (group 16)
     //  data is similar as in basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_split2_depthwise_sep_opt
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 16, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 16, 2, 2 } });
     set_values(input,
     { 8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f,
         8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f,
@@ -1063,13 +1057,13 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_group16
         8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f
     });
 
-    topology topology(input_layout("input", input.get_layout()));
+    topology topology(input_layout("input", input->get_layout()));
 
     std::vector<primitive_id> weights_vec;
     std::vector<primitive_id> bias_vec;
 
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(16), batch(1), feature(1), spatial(2, 2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 16, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(16), batch(1), feature(1), spatial(2, 2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 16, 1, 1 } });
 
     set_values(weights,
         {
@@ -1108,7 +1102,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_group16
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 4.5f, 13.f, -17.f, -8.f, -28.f, 1.f, -17.f,
@@ -1131,9 +1125,9 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_group16
     //  Test for depthwise separable optimization, there are 16 joined weights and biases (group 16)
     //  data is similar as in basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_split2_depthwise_sep_opt_ofm2
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 16, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 16, 2, 2 } });
     set_values(input,
     { 8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f,
         8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f,
@@ -1145,13 +1139,13 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_group16
         8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f
     });
 
-    topology topology(input_layout("input", input.get_layout()));
+    topology topology(input_layout("input", input->get_layout()));
 
     std::vector<primitive_id> weights_vec;
     std::vector<primitive_id> bias_vec;
 
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(16), batch(2), feature(1), spatial(2, 2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 32, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(16), batch(2), feature(1), spatial(2, 2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 32, 1, 1 } });
 
     set_values(weights,
         {
@@ -1197,7 +1191,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_group16
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -3.f, 4.5f, 13.f, -17.f,-3.f, 4.5f, 13.f, -17.f, -8.f, -28.f, 1.f, -17.f, -8.f, -28.f, 1.f, -17.f,
@@ -1219,11 +1213,11 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_group16
 TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x6x1x1_bfyx_stride2_pad1_group2_ofm3) {
     //  data is similar as in basic_wsiz2x2_in1x6x1x1_bfyx_stride2_pad1_split2_ofm3
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 4, 1, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(3), feature(2), spatial(1, 1)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 6, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 4, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(3), feature(2), spatial(1, 1)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 6, 1, 1 } });
 
     set_values(input, {
         1.5f, 0.5f, 2.0f, -1.0f
@@ -1238,7 +1232,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x6x1x1_bfyx_stride2_pad1_group2_
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, 2, { 1, 1, 1, 1 }, { 0, 0, 0, 0 })
@@ -1253,7 +1247,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x6x1x1_bfyx_stride2_pad1_group2_
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -1.5f, 8.0f, 7.75f, 11.0f, 6.0f, -2.0f
@@ -1284,18 +1278,18 @@ TEST(deconvolution_f32_fw_gpu, basic3D_wsiz2x2x1_in1x1x2x2x1_nopad) {
     //   18    0.75  7.25
     //   23    42.5  15.5
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 1, 2, 2, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oizyx,{ 1, 1, 2, 2, 1 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 1, 2, 2, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oizyx,{ 1, 1, 2, 2, 1 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f });
     set_values(biases, { 2.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1,1,1,1,1 })
@@ -1310,7 +1304,7 @@ TEST(deconvolution_f32_fw_gpu, basic3D_wsiz2x2x1_in1x1x2x2x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -14.f, 5.f, 2.25f,
@@ -1410,10 +1404,10 @@ TEST(deconvolution_f32_fw_gpu, basic3D_wsiz3x3x3_in1x1x4x4x4_nopad) {
     //  1  2  3  3  2  1
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 1, 4, 4, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oizyx,{ 1, 1, 3, 3, 3 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 1, 4, 4, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oizyx,{ 1, 1, 3, 3, 3 } });
 
     set_values(input,
     {
@@ -1447,7 +1441,7 @@ TEST(deconvolution_f32_fw_gpu, basic3D_wsiz3x3x3_in1x1x4x4x4_nopad) {
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         deconvolution("deconv", "input", { "weights" })
     );
@@ -1461,7 +1455,7 @@ TEST(deconvolution_f32_fw_gpu, basic3D_wsiz3x3x3_in1x1x4x4x4_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         1.0f, 2.0f, 3.0f, 3.0f, 2.0f, 1.0f,
@@ -1532,10 +1526,10 @@ TEST(deconvolution_f32_fw_gpu, basic3D_wsiz2x2x2_in1x1x2x2x2_stride2_nopad) {
     //   3.5 1.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oizyx,{ 1, 1, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oizyx,{ 1, 1, 2, 2, 2 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f, 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f, -2.0f, 0.5f, 3.5f, 1.5f });
@@ -1543,7 +1537,7 @@ TEST(deconvolution_f32_fw_gpu, basic3D_wsiz2x2x2_in1x1x2x2x2_stride2_nopad) {
     //set_values(weights, { 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         deconvolution("deconv", "input", { "weights" }, { 1,1,2,2,2 })
     );
@@ -1556,7 +1550,7 @@ TEST(deconvolution_f32_fw_gpu, basic3D_wsiz2x2x2_in1x1x2x2x2_stride2_nopad) {
     EXPECT_EQ(outputs.begin()->first, "deconv");
 
     auto output_prim = outputs.begin()->second.get_memory();
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -16.f, 4.f, -1.f, 0.25f,
@@ -1607,16 +1601,16 @@ TEST(deconvolution_f32_fw_gpu, basic3D_wsiz2x2x2_in1x1x2x2x2_stride2_pad1) {
     //  12 1.75
     //   3 -18
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oizyx,{ 1, 1, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oizyx,{ 1, 1, 2, 2, 2 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f, 8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f, -2.0f, 0.5f, 3.5f, 1.5f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         deconvolution("deconv", "input", { "weights" }, { 1,1,2,2,2 }, { 0, 0, -1, -1, -1 })
     );
@@ -1629,7 +1623,7 @@ TEST(deconvolution_f32_fw_gpu, basic3D_wsiz2x2x2_in1x1x2x2x2_stride2_pad1) {
     EXPECT_EQ(outputs.begin()->first, "deconv");
 
     auto output_prim = outputs.begin()->second.get_memory();
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         12.f, 1.75f, 3.f, -18.f,
@@ -1649,8 +1643,8 @@ TEST(deconvolution_f16_gpu, basic_k9x9_s2x2_pad4x4) {
     //  Stride : 2x2
     //  Pad    : 4x4
 
-    //const auto& engine = get_test_engine();
-    engine engine;
+    //auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     VVVVF<FLOAT16> input_rnd = generate_random_4d<FLOAT16>(1, 32, 16, 16, -2, 2);
     VF<FLOAT16> input_rnd_vec = flatten_4d<FLOAT16>(format::bfyx, input_rnd);
@@ -1665,11 +1659,11 @@ TEST(deconvolution_f16_gpu, basic_k9x9_s2x2_pad4x4) {
     for (unsigned int i = 0; i < bias_rnd.size(); i++)
         bias_f32_rnd.push_back(float(bias_rnd[i]));
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 32, 16, 16 } });
-    auto weights = memory::allocate(engine, { data_types::f16, format::oiyx, { 1, 32, 9, 9 } });
-    auto biases = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 1, 1, 1 } });
-    auto weights_f32 = memory::allocate(engine, { data_types::f32, format::oiyx, { 1, 32, 9, 9 } });
-    auto biases_f32 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 32, 16, 16 } });
+    auto weights = engine.allocate_memory({ data_types::f16, format::oiyx, { 1, 32, 9, 9 } });
+    auto biases = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 1, 1, 1 } });
+    auto weights_f32 = engine.allocate_memory({ data_types::f32, format::oiyx, { 1, 32, 9, 9 } });
+    auto biases_f32 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, input_rnd_vec);
     set_values(weights, filter_rnd_vec);
@@ -1678,7 +1672,7 @@ TEST(deconvolution_f16_gpu, basic_k9x9_s2x2_pad4x4) {
     set_values(biases_f32, bias_f32_rnd);
 
     topology topology_ref(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1, 1, 2, 2 }, { 0, 0, -4, -4 }, tensor{ 1, 1, 32, 32 })
@@ -1691,16 +1685,15 @@ TEST(deconvolution_f16_gpu, basic_k9x9_s2x2_pad4x4) {
     EXPECT_EQ(outputs_ref.size(), size_t(1));
     EXPECT_EQ(outputs_ref.begin()->first, "deconv");
     auto output_ref_prim = outputs_ref.begin()->second.get_memory();
-    auto output_ref_ptr = output_ref_prim.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_ref_ptr(output_ref_prim, get_test_stream());
 
     std::vector<FLOAT16> output_vec_ref;
-    for (unsigned int i = 0; i < output_ref_prim.get_layout().count(); i++)
-    {
+    for (unsigned int i = 0; i < output_ref_prim->get_layout().count(); i++) {
         output_vec_ref.push_back(output_ref_ptr[i]);
     }
 
     topology topology_act(
-        input_layout("input_act", input.get_layout()),
+        input_layout("input_act", input->get_layout()),
         data("weights_f32", weights_f32),
         data("biases_f32", biases_f32),
         deconvolution("deconv_act", "input_act", { "weights_f32" }, { "biases_f32" }, { 1, 1, 2, 2 }, { 0, 0, -4, -4 }),
@@ -1716,11 +1709,10 @@ TEST(deconvolution_f16_gpu, basic_k9x9_s2x2_pad4x4) {
     EXPECT_EQ(outputs_act.size(), size_t(1));
     EXPECT_EQ(outputs_act.begin()->first, "out");
     auto output_act_prim = outputs_act.begin()->second.get_memory();
-    auto output_act_ptr = output_act_prim.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_act_ptr(output_act_prim, get_test_stream());
 
     std::vector<float> output_vec;
-    for (unsigned int i = 0; i < output_act_prim.get_layout().count(); i++)
-    {
+    for (unsigned int i = 0; i < output_act_prim->get_layout().count(); i++) {
         float x = float_round(output_act_ptr[i]), y = float_round(output_vec_ref[i]);
         EXPECT_NEAR(x, y, 1e-0f);
     }
@@ -1748,18 +1740,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x2_b_fs_yx_fsv16_stride2_pad
     //  -3    4.5    0.5   22
     //   13  -17     5    -7
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::yxio, { 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::yxio, { 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f });
     set_values(weights, { -2.f, 2.f, 7.f, -0.5f });
     set_values(biases, { 1.0f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             data("biases", biases),
             deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1, 1, 2, 2 }, { 0, 0, -1, -1 }),
@@ -1780,7 +1772,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x2_b_fs_yx_fsv16_stride2_pad
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
             -3.f, 4.5f, 13.f, -17.f,
@@ -1814,11 +1806,11 @@ TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in2x2x1x2_b_fs_yx_fsv16_stride2_pad
     //  -3    4.5    0.5   22
     //   13  -17     5    -7
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx,{ 2, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f16, format::oiyx,{ 1, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f16, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx,{ 2, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f16, format::oiyx,{ 1, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f16, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { FLOAT16(8.f), FLOAT16(0.5f),
                         FLOAT16(6.f), FLOAT16(9.f),
@@ -1830,7 +1822,7 @@ TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in2x2x1x2_b_fs_yx_fsv16_stride2_pad
     set_values(biases, { FLOAT16(1.0f) });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             data("biases", biases),
             deconvolution("deconv", "input", { "weights" }, { "biases" }, { 1, 1, 2, 2 }, { 0, 0, -1, -1 }),
@@ -1851,7 +1843,7 @@ TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in2x2x1x2_b_fs_yx_fsv16_stride2_pad
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
             -3.f, 4.5f, 13.f, -17.f,
@@ -1865,11 +1857,11 @@ TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in2x2x1x2_b_fs_yx_fsv16_stride2_pad
 
 TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_b_fs_yx_fsv16_stride2_pad1_group2) {
     //  data is similar as in basic_wsiz2x2_in1x2x2x2_bfyx_stride2_pad1_split2
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f });
     set_values(weights, {
@@ -1879,7 +1871,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_b_fs_yx_fsv16_stride2_pad
     set_values(biases, { 1.0f, -1.0f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             data("biases", biases),
             deconvolution("deconv", "input", { "weights" }, { "biases" }, 2, { 1, 1, 2, 2 }, { 0, 0, -1, -1 }),
@@ -1900,7 +1892,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_b_fs_yx_fsv16_stride2_pad
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
             -3.f, 4.5f, 13.f, -17.f,
@@ -1913,11 +1905,11 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_b_fs_yx_fsv16_stride2_pad
 }
 
 TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_b_fs_yx_fsv16_stride2_pad1_b_fs_yx_fsv16_dw) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f, 1.f, 3.f, 2.f, 4.f });
     set_values(weights, {
@@ -1927,7 +1919,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_b_fs_yx_fsv16_stride2_pad
     set_values(biases, { 0.0f, 0.0f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             data("biases", biases),
             deconvolution("deconv", "input", { "weights" }, { "biases" }, 2, { 1, 1, 2, 2 }, { 0, 0, -1, -1 }),
@@ -1948,7 +1940,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_b_fs_yx_fsv16_stride2_pad
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
             -4.f, 3.5f, 12.f, -18.f,
@@ -1961,18 +1953,18 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in1x2x2x2_b_fs_yx_fsv16_stride2_pad
 }
 
 TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_nopad_b_fs_yx_fsv16_dw) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f,  8.f, 0.5f, 6.f, 9.f });
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f,  -2.0f, 0.5f, 3.5f, 1.5f });
     set_values(biases, { 2.0f, 2.0f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             data("biases", biases),
             reorder("input_fsv16", "input", format::b_fs_yx_fsv16, data_types::f32),
@@ -1994,7 +1986,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_nopad_b_fs_yx_fsv16_dw) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
             -14.f, 5.f, 2.25f,
@@ -2013,11 +2005,11 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_nopad_b_fs_yx_fsv16_dw) {
 }
 
 TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_pad1_b_fs_yx_fsv16_dw) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f,
                         8.f, 0.5f, 6.f, 9.f});
@@ -2026,7 +2018,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_pad1_b_fs_yx_fsv16_dw) {
     set_values(biases, { 2.0f, 2.0f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             data("biases", biases),
             reorder("input_fsv16", "input", format::b_fs_yx_fsv16, data_types::f32),
@@ -2048,7 +2040,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_pad1_b_fs_yx_fsv16_dw) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_FLOAT_EQ(0.75f, output_ptr[0]);
     EXPECT_FLOAT_EQ(0.75f, output_ptr[1]);
@@ -2056,11 +2048,11 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_pad1_b_fs_yx_fsv16_dw) {
 
 
 TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride2_nopad_b_fs_yx_fsv16_dw) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(2, 2)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f,
                         8.f, 0.5f, 6.f, 9.f});
@@ -2069,7 +2061,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride2_nopad_b_fs_yx_fsv
     set_values(biases, { 1.0f, 1.0f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("biases", biases),
         deconvolution("deconv", "input", { "weights" }, { "biases" }, 2, { 1,1,2,2 }),
@@ -2090,7 +2082,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride2_nopad_b_fs_yx_fsv
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
         -15.f, 5.f, 0.f, 1.25f,
@@ -2112,11 +2104,11 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride2_nopad_b_fs_yx_fsv
 }
 
 TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride4_pad2_b_fs_yx_fsv16_dw) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(3, 3)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(3, 3)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f,
                         8.f, 0.5f, 6.f, 9.f});
@@ -2125,7 +2117,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride4_pad2_b_fs_yx_fsv1
     set_values(biases, { 0.0f, 0.0f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             data("biases", biases),
             deconvolution("deconv", "input", { "weights" }, { "biases" }, 2, { 1, 1, 4, 4 }, { 0, 0, -2, -2 }),
@@ -2146,7 +2138,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride4_pad2_b_fs_yx_fsv1
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
             40.f, 0.f, 1.5f,
@@ -2165,11 +2157,11 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride4_pad2_b_fs_yx_fsv1
 }
 
 TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride4_pad2_b_fs_yx_fsv16_dw_batch2) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(3, 3)) });
-    auto biases = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::goiyx, tensor(group(2), batch(1), feature(1), spatial(3, 3)) });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     set_values(input, { 8.f, 0.5f, 6.f, 9.f,
                         8.f, 0.5f, 6.f, 9.f,
@@ -2181,7 +2173,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride4_pad2_b_fs_yx_fsv1
     set_values(biases, { 0.0f, 0.0f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             data("biases", biases),
             deconvolution("deconv", "input", { "weights" }, { "biases" }, 2, { 1, 1, 4, 4 }, { 0, 0, -2, -2 }),
@@ -2202,7 +2194,7 @@ TEST(deconvolution_f32_fw_gpu, basic_wsiz2x2_in2x2x1x1_stride4_pad2_b_fs_yx_fsv1
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
             40.f, 0.f, 1.5f,
@@ -2250,10 +2242,10 @@ TEST(deconvolution_f32_fw_gpu, bs_fs_zyx_bsv16_fsv16_wsiz2x2x2_in1x1x2x2x2_strid
     //   3.5 1.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 32, 1, 2, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oizyx,{ 1, 1, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 32, 1, 2, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oizyx,{ 1, 1, 2, 2, 2 } });
 
     std::vector<float> input_single_batch = { 8.f, 0.5f, 6.f, 9.f, 8.f, 0.5f, 6.f, 9.f };
     std::vector<float> input_batched;
@@ -2267,7 +2259,7 @@ TEST(deconvolution_f32_fw_gpu, bs_fs_zyx_bsv16_fsv16_wsiz2x2x2_in1x1x2x2x2_strid
     set_values(weights, { -2.0f, 0.5f, 3.5f, 1.5f, -2.0f, 0.5f, 3.5f, 1.5f });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             deconvolution("deconv", "input", { "weights" }, { 1,1,2,2,2 }, { 0, 0, -1, -1, -1 }),
             reorder("out", "deconv", format::bfzyx, data_types::f32)
@@ -2286,7 +2278,7 @@ TEST(deconvolution_f32_fw_gpu, bs_fs_zyx_bsv16_fsv16_wsiz2x2x2_in1x1x2x2x2_strid
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output_prim = outputs.begin()->second.get_memory();
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
             12.f, 1.75f, 3.f, -18.f,
@@ -2301,11 +2293,11 @@ TEST(deconvolution_f32_fw_gpu, bs_fs_zyx_bsv16_fsv16_wsiz2x2x2_in1x1x2x2x2_strid
 }
 
 TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in1x2x2x2_fs_b_yx_fsv32_stride1_pad1_replace_to_conv) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx,{ 2, 1, 2, 2 } });
-    auto weights = memory::allocate(engine, { data_types::f16, format::bfyx,{ 2, 1, 2, 2 } });
-    auto biases = memory::allocate(engine, { data_types::f16, format::bfyx,{ 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx,{ 2, 1, 2, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f16, format::bfyx,{ 2, 1, 2, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f16, format::bfyx,{ 1, 2, 1, 1 } });
 
     set_values(input, { FLOAT16(8.f), FLOAT16(0.5f), FLOAT16(6.f), FLOAT16(9.f),
                         FLOAT16(1.f), FLOAT16(3.f), FLOAT16(2.f), FLOAT16(4.f)
@@ -2317,7 +2309,7 @@ TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in1x2x2x2_fs_b_yx_fsv32_stride1_pad
     set_values(biases, { FLOAT16(1.0f), FLOAT16(-1.0f) });
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             reorder("reorder", "input", format::fs_b_yx_fsv32, data_types::f16),
             data("weights", weights),
             data("biases", biases),
@@ -2337,7 +2329,7 @@ TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in1x2x2x2_fs_b_yx_fsv32_stride1_pad
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     std::vector<float> expected_output_vec = {
             -15.f, 16.f, 2.f, 45.f, -5.5f, 18.75f, 43.f, 61.f, -3.5f,
@@ -2345,7 +2337,7 @@ TEST(deconvolution_f16_fw_gpu, basic_wsiz2x2_in1x2x2x2_fs_b_yx_fsv32_stride1_pad
             -1.f, -3.f, 7.f, 4.f, 17.5f, 7.5f, 15.f, 28.f, -1.f,
             -5.f, -12.f, 2.f, -18.f, -49.f, -18.f, -19.f, -51.f, -29.f,
     };
-    ASSERT_EQ(expected_output_vec.size(), output_prim.count());
+    ASSERT_EQ(expected_output_vec.size(), output_prim->count());
 
     for (size_t i = 0; i < expected_output_vec.size(); i++) {
         EXPECT_FLOAT_EQ(expected_output_vec[i], output_ptr[i]) << " index=" << i;
@@ -2449,8 +2441,8 @@ template <typename InputT, typename WeightsT, typename OutputT>
 class deconvolution_random_test_base {
 public:
     template <typename T>
-    void set_memory(cldnn::memory& mem, const VVVVVF<T>& data) {
-        auto ptr = mem.pointer<T>();
+    void set_memory(cldnn::memory::ptr mem, const VVVVVF<T>& data) {
+        cldnn::mem_lock<T> ptr(mem, get_test_stream());
 
         auto b = data.size();
         auto f = data[0].size();
@@ -2464,7 +2456,7 @@ class deconvolution_random_test_base {
                     for (size_t yi = 0; yi < y; ++yi) {
                         for (size_t xi = 0; xi < x; ++xi) {
                             auto coords = cldnn::tensor(batch(bi), feature(fi), spatial(xi, yi, zi, 0));
-                            auto offset = mem.get_layout().get_linear_offset(coords);
+                            auto offset = mem->get_layout().get_linear_offset(coords);
                             ptr[offset] = data[bi][fi][zi][yi][xi];
                         }
                     }
@@ -2474,8 +2466,8 @@ class deconvolution_random_test_base {
     }
 
     template <typename T>
-    void set_memory_weights(cldnn::memory& mem, const VVVVVVF<T>& data) {
-        auto ptr = mem.pointer<T>();
+    void set_memory_weights(cldnn::memory::ptr mem, const VVVVVVF<T>& data) {
+        cldnn::mem_lock<T> ptr(mem, get_test_stream());
 
         auto g = data.size();
         auto b = data[0].size();
@@ -2491,7 +2483,7 @@ class deconvolution_random_test_base {
                         for (size_t yi = 0; yi < y; ++yi) {
                             for (size_t xi = 0; xi < x; ++xi) {
                                 auto coords = cldnn::tensor(group(gi), batch(bi), feature(fi), spatial(xi, yi, zi, 0));
-                                auto offset = mem.get_layout().get_linear_offset(coords);
+                                auto offset = mem->get_layout().get_linear_offset(coords);
                                 ptr[offset] = data[gi][bi][fi][zi][yi][xi];
                             }
                         }
@@ -2526,7 +2518,7 @@ class deconvolution_random_test_base {
             type_test_ranges<T>::max);
     }
 
-    void run(cldnn::engine eng, const deconvolution_random_test_params& params, cldnn::build_options build_opts) {
+    void run(cldnn::engine& eng, const deconvolution_random_test_params& params, cldnn::build_options build_opts) {
         uint32_t groups = params.weights_size.group[0];
         size_t ifm = params.weights_size.feature[0];
         size_t ofm = params.weights_size.batch[0];
@@ -2537,8 +2529,8 @@ class deconvolution_random_test_base {
         auto in_layout = cldnn::layout(cldnn::type_to_data_type<InputT>::value, params.input_format, params.input_size);
         auto wei_layout = cldnn::layout(cldnn::type_to_data_type<WeightsT>::value, params.weights_format, params.weights_size);
 
-        auto wei_mem = cldnn::memory::allocate(eng, wei_layout);
-        auto in_mem = cldnn::memory::allocate(eng, in_layout);
+        auto wei_mem = eng.allocate_memory(wei_layout);
+        auto in_mem = eng.allocate_memory(in_layout);
 
         this->set_memory_weights(wei_mem, weights_data);
         this->set_memory(in_mem, input_data);
@@ -2553,7 +2545,7 @@ class deconvolution_random_test_base {
         if (params.with_bias) {
             auto bias_size = cldnn::tensor(feature(params.weights_size.batch[0] * params.weights_size.group[0]));
             auto bias_lay = cldnn::layout(cldnn::type_to_data_type<OutputT>::value, cldnn::format::bfyx, bias_size);
-            auto bias_mem = cldnn::memory::allocate(eng, bias_lay);
+            auto bias_mem = eng.allocate_memory(bias_lay);
             bias_data = generate_random_1d<OutputT>(bias_lay.size.feature[0], -1, 1);
             set_values(bias_mem, bias_data);
             topo.add(cldnn::data("bias", bias_mem));
@@ -2581,10 +2573,10 @@ class deconvolution_random_test_base {
 
         // Compare results
         {
-            auto ptr = out_mem.pointer<OutputT>();
+            cldnn::mem_lock<OutputT> ptr(out_mem, get_test_stream());
 
-            auto b = static_cast<size_t>(out_mem.get_layout().size.batch[0]);
-            auto of = static_cast<size_t>(out_mem.get_layout().size.feature[0]);
+            auto b = static_cast<size_t>(out_mem->get_layout().size.batch[0]);
+            auto of = static_cast<size_t>(out_mem->get_layout().size.feature[0]);
 
             for (size_t bi = 0; bi < b; ++bi) {
                 for (size_t fi = 0; fi < of; ++fi) {
@@ -2597,16 +2589,16 @@ class deconvolution_random_test_base {
                         params.input_offset,
                         group * ifm);
 
-                    ASSERT_EQ(reference.size(), out_mem.get_layout().size.spatial[2]);
-                    ASSERT_EQ(reference[0].size(), out_mem.get_layout().size.spatial[1]);
-                    ASSERT_EQ(reference[0][0].size(), out_mem.get_layout().size.spatial[0]);
+                    ASSERT_EQ(reference.size(), out_mem->get_layout().size.spatial[2]);
+                    ASSERT_EQ(reference[0].size(), out_mem->get_layout().size.spatial[1]);
+                    ASSERT_EQ(reference[0][0].size(), out_mem->get_layout().size.spatial[0]);
 
                     for (size_t zi = 0; zi < reference.size(); zi++) {
                         for (size_t yi = 0; yi < reference[0].size(); yi++) {
                             for (size_t xi = 0; xi < reference[0][0].size(); xi++) {
                                 auto ref_val = reference[zi][yi][xi];
                                 auto out_coords = cldnn::tensor(batch(bi), feature(fi), spatial(xi, yi, zi, 0));
-                                auto out_offset = out_mem.get_layout().get_linear_offset(out_coords);
+                                auto out_offset = out_mem->get_layout().get_linear_offset(out_coords);
                                 auto out_val = ptr[out_offset];
                                 TYPED_ASSERT_EQ(ref_val, out_val)
                                     << "at b=" << bi << ", f=" << fi << ", z=" << zi << ", y=" << yi << ", x=" << xi << std::endl
@@ -2626,7 +2618,6 @@ class deconvolution_random_test_base {
 class deconvolution_random_test : public testing::TestWithParam<deconvolution_random_test_params> {
 protected:
     void SetUp() override {
-        eng = get_test_engine();
         build_opts.set_option(cldnn::build_option::optimize_data(true));
     }
 
@@ -2650,7 +2641,6 @@ class deconvolution_random_test : public testing::TestWithParam<deconvolution_ra
         }
     }
 
-    cldnn::engine eng;
     cldnn::build_options build_opts;
 
 private:
@@ -2658,7 +2648,7 @@ class deconvolution_random_test : public testing::TestWithParam<deconvolution_ra
     void run_typed() {
         auto& params = GetParam();
         deconvolution_random_test_base<InputT, WeightsT, OutputT> test;
-        test.run(eng, params, build_opts);
+        test.run(get_test_engine(), params, build_opts);
     }
 
     template <typename InputT, typename WeightsT>
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/depth_concatenate_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/depth_concatenate_gpu_test.cpp
index 6af2651aa2fc1b..116d6ba21f4d2b 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/depth_concatenate_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/depth_concatenate_gpu_test.cpp
@@ -4,23 +4,19 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/concatenation.hpp"
-#include "api/convolution.hpp"
-#include "api/data.hpp"
-#include "api/eltwise.hpp"
-#include "api/fully_connected.hpp"
-#include "api/pooling.hpp"
-#include "api/crop.hpp"
-#include "api/resample.hpp"
-#include "api/reshape.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
 #include "test_utils/test_utils.h"
 
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/concatenation.hpp>
+#include <cldnn/primitives/convolution.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/eltwise.hpp>
+#include <cldnn/primitives/fully_connected.hpp>
+#include <cldnn/primitives/pooling.hpp>
+#include <cldnn/primitives/crop.hpp>
+#include <cldnn/primitives/resample.hpp>
+#include <cldnn/primitives/reshape.hpp>
+
 using namespace cldnn;
 using namespace tests;
 
@@ -59,16 +55,16 @@ TEST(depth_concatenate_f32_gpu, test01) {
     //  0   -0.2  :f4
     //
 
-    const auto& engine = get_test_engine();
-    auto input1 = memory::allocate(engine, {data_types::f32, format::yxfb, {2, 2, 1, 1}});
-    auto input2 = memory::allocate(engine, {data_types::f32, format::yxfb, {2, 3, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({data_types::f32, format::yxfb, {2, 2, 1, 1}});
+    auto input2 = engine.allocate_memory({data_types::f32, format::yxfb, {2, 3, 1, 1}});
 
     set_values(input1, {0.5f, 0.7f, 0.2f, 0.4f});
     set_values(input2, {1.0f, 0.1f, 0.3f, -0.5f, 0.0f, -0.2f});
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(concatenation("depth1", {"input1", "input2"}, concatenation::along_f));
 
     network network(engine, topology);
@@ -82,7 +78,7 @@ TEST(depth_concatenate_f32_gpu, test01) {
 
     auto output = outputs.at("depth1").get_memory();
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     EXPECT_FLOAT_EQ(0.5f, output_ptr[0]);
     EXPECT_FLOAT_EQ(0.7f, output_ptr[1]);
     EXPECT_FLOAT_EQ(0.2f, output_ptr[2]);
@@ -118,16 +114,16 @@ void concat_basic_with_reorder() {
     //  0    0  :f4
     //
 
-    const auto& engine = get_test_engine();
-    auto input1 = memory::allocate(engine, {data_types::f32, format::yxfb, {2, 2, 1, 1}});
-    auto input2 = memory::allocate(engine, {data_types::f32, format::yxfb, {2, 3, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({data_types::f32, format::yxfb, {2, 2, 1, 1}});
+    auto input2 = engine.allocate_memory({data_types::f32, format::yxfb, {2, 3, 1, 1}});
     auto outs = {3.0f, 4.0f, 0.0f, 1.0f, 1.0f, 4.0f, -4.0f, -8.0f, 0.0f, 0.0f};
     set_values(input1, {2.5f, 3.7f, 0.2f, 1.4f});
     set_values(input2, {1.0f, 4.1f, -4.3f, -7.5f, 0.0f, -0.2f});
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(reorder("to_int1", "input1", {DType, format::yxfb, {2, 2, 1, 1}}));
     topology.add(reorder("to_int2", "input2", {DType, format::yxfb, {2, 3, 1, 1}}));
     topology.add(concatenation("depth1", {"to_int1", "to_int2"}, concatenation::along_f));
@@ -144,7 +140,7 @@ void concat_basic_with_reorder() {
 
     auto output = outputs.at("to_float").get_memory();
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     int ptr_cntr = 0;
     for (const auto& ref : outs) {
         EXPECT_FLOAT_EQ(ref, output_ptr[ptr_cntr++]);
@@ -194,19 +190,19 @@ TEST(depth_concatenate_f32_gpu, test02) {
     //  0   -0.2  :f7
     //
 
-    const auto& engine = get_test_engine();
-    auto input1 = memory::allocate(engine, {data_types::f32, format::yxfb, {2, 2, 1, 1}});
-    auto input2 = memory::allocate(engine, {data_types::f32, format::yxfb, {2, 3, 1, 1}});
-    auto input3 = memory::allocate(engine, {data_types::f32, format::bfyx, {2, 3, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({data_types::f32, format::yxfb, {2, 2, 1, 1}});
+    auto input2 = engine.allocate_memory({data_types::f32, format::yxfb, {2, 3, 1, 1}});
+    auto input3 = engine.allocate_memory({data_types::f32, format::bfyx, {2, 3, 1, 1}});
 
     set_values(input1, {0.5f, 0.7f, 0.2f, 0.4f});
     set_values(input2, {1.0f, 0.1f, 0.3f, -0.5f, 0.0f, -0.2f});
     set_values(input3, {1.0f, 0.3f, 0.0f, 0.1f, -0.5f, -0.2f});
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("input3", input3.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("input3", input3->get_layout()));
     topology.add(concatenation("depth1", {"input1", "input2", "input3"}, concatenation::along_f));
 
     network network(engine, topology);
@@ -221,7 +217,7 @@ TEST(depth_concatenate_f32_gpu, test02) {
 
     auto output = outputs.at("depth1").get_memory();
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     EXPECT_FLOAT_EQ(0.5f, output_ptr[0]);
     EXPECT_FLOAT_EQ(0.7f, output_ptr[1]);
     EXPECT_FLOAT_EQ(0.2f, output_ptr[2]);
@@ -241,14 +237,14 @@ TEST(depth_concatenate_f32_gpu, test02) {
 }
 
 TEST(concatenate_f32_gpu, test_concatenation_of_pool_and_unpool) {
-    engine engine;
-    auto input1 = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 1, 2, 2}});
-    auto weights = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 1, 2, 1}});
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({data_types::f32, format::bfyx, {1, 1, 2, 2}});
+    auto weights = engine.allocate_memory({data_types::f32, format::bfyx, {1, 1, 2, 1}});
 
     set_values(input1, {16.0f, 32.0f, 128.0f, 256.0f});
     set_values(weights, {.1f, .2f});
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
     topology.add(pooling("pool1", "input1",
                          cldnn::pooling_mode::max,
                          {1, 1, 2, 1}, /*kernel*/
@@ -267,7 +263,7 @@ TEST(concatenate_f32_gpu, test_concatenation_of_pool_and_unpool) {
     auto outputs = network.execute({});
     auto output = outputs.at("conv").get_memory();
     std::vector<float> out_ref = {6.4f, 8.f, 51.2f, 64.f};
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 4; i++) {
         EXPECT_NEAR(output_ptr[i], out_ref[i], 1e-3);
     }
@@ -278,13 +274,13 @@ TEST(depth_concatenate_f32_gpu, test03_cascade_concat_opt) {
     //  Despite having concatenations one after another and connected to different non padded activation primitives,
     //  graph should remove all concatenations from execution.
 
-    const auto& engine = get_test_engine();
-    auto input1 = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 2, 2, 1}});
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({data_types::f32, format::bfyx, {1, 2, 2, 1}});
 
     set_values(input1, {16.0f, 32.0f, 128.0f, 256.0f});
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
     topology.add(activation("relu1", "input1", activation_func::relu));
     topology.add(activation("relu2", "relu1", activation_func::sqrt));
     topology.add(concatenation("depth1", {"relu2", "relu1"}, concatenation::along_f));
@@ -303,7 +299,7 @@ TEST(depth_concatenate_f32_gpu, test03_cascade_concat_opt) {
     auto outputs = network.execute({});
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
     auto executed_primitives = network.get_executed_primitives();
 
     EXPECT_TRUE(executed_primitives.count("depth1") == 0);
@@ -331,9 +327,9 @@ TEST(depth_concatenate_f32_gpu, test03_cascade_concat_opt) {
 TEST(depth_concatenate_f32_gpu, test04_fused_relu) {
     // 2 inputs of size 3x10x10 concatenated on f axis with fused relu
 
-    const auto& engine = get_test_engine();
-    auto input1 = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 3, 10, 10}});
-    auto input2 = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 3, 10, 10}});
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({data_types::f32, format::bfyx, {1, 3, 10, 10}});
+    auto input2 = engine.allocate_memory({data_types::f32, format::bfyx, {1, 3, 10, 10}});
 
     std::vector<float> input1_vec = generate_random_input<float>(1, 3, 10, 10, -10, 10);
     set_values(input1, input1_vec);
@@ -341,8 +337,8 @@ TEST(depth_concatenate_f32_gpu, test04_fused_relu) {
     set_values(input2, input2_vec);
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(concatenation("depth1", {"input1", "input2"}, concatenation::along_f));
     topology.add(activation("relu1", "depth1", activation_func::relu));
 
@@ -359,7 +355,7 @@ TEST(depth_concatenate_f32_gpu, test04_fused_relu) {
 
     auto output = outputs.at("relu1").get_memory();
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     unsigned int input_element_count = 300;
     for (unsigned int i = 0; i < 600; i++) {
         if (i < input_element_count)
@@ -372,9 +368,9 @@ TEST(depth_concatenate_f32_gpu, test04_fused_relu) {
 TEST(depth_concatenate_f32_gpu, test05_different_formats) {
     // 2 inputs of size 3x2x2 concatenated on f axis
 
-    const auto& engine = get_test_engine();
-    auto input1 = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 3, 2, 2}});
-    auto input2 = memory::allocate(engine, {data_types::f32, format::yxfb, {1, 3, 2, 2}});
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({data_types::f32, format::bfyx, {1, 3, 2, 2}});
+    auto input2 = engine.allocate_memory({data_types::f32, format::yxfb, {1, 3, 2, 2}});
 
     set_values(input1, {1.0f, 1.0f, 1.0f, 1.0f,
                         2.0f, 2.0f, 2.0f, 2.0f,
@@ -393,8 +389,8 @@ TEST(depth_concatenate_f32_gpu, test05_different_formats) {
         -3.0f, -3.0f, -3.0f, -3.0f};
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(reshape("reshape1", "input1", {1, 3, 2, 2}));
     topology.add(reshape("reshape2", "input2", {1, 3, 2, 2}));
     topology.add(concatenation("depth1", {"reshape1", "reshape2"}, concatenation::along_f));
@@ -412,7 +408,7 @@ TEST(depth_concatenate_f32_gpu, test05_different_formats) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     int cntr = 0;
     for (float val : output_ptr) {
         EXPECT_EQ(val, out_ref[cntr++]);
@@ -429,16 +425,16 @@ TEST(depth_concatenate_f32_gpu, test06_padded_input) {
     const int32_t input_f = 32;
     const int32_t output_f = 3 * input_f;
 
-    const auto& engine = get_test_engine();
-    auto input1 = memory::allocate(engine, { data_types::f16, format::fs_b_yx_fsv32, {1, input_f, 1, 1} });
-    auto input2 = memory::allocate(engine, { data_types::f16, format::fs_b_yx_fsv32, {1, input_f, 1, 1} });
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({ data_types::f16, format::fs_b_yx_fsv32, {1, input_f, 1, 1} });
+    auto input2 = engine.allocate_memory({ data_types::f16, format::fs_b_yx_fsv32, {1, input_f, 1, 1} });
 
     auto input1_data = generate_random_4d<FLOAT16>(1, input_f, 1, 1, -1, 1);
     auto input2_data = generate_random_4d<FLOAT16>(1, input_f, 1, 1, -1, 1);
     set_values(input1, flatten_4d(format::bfyx, input1_data));
     set_values(input2, flatten_4d(format::bfyx, input2_data));
 
-    auto weights = memory::allocate(engine, { data_types::f16, format::oiyx, {input_f, input_f, 3, 3} });
+    auto weights = engine.allocate_memory({ data_types::f16, format::oiyx, {input_f, input_f, 3, 3} });
     // Construct weights for convolution that just double input values.
     VVVVF<FLOAT16> weights_data;
     weights_data.resize(input_f);
@@ -449,8 +445,8 @@ TEST(depth_concatenate_f32_gpu, test06_padded_input) {
     set_values(weights, flatten_4d(format::bfyx, weights_data));
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(activation("actv1", "input1", activation_func::linear, { 0.75f, 0.0f }));
     topology.add(activation("actv2", "input2", activation_func::linear, { 0.5f, 0.0f }));
     topology.add(data("weights", weights));
@@ -482,8 +478,8 @@ TEST(depth_concatenate_f32_gpu, test06_padded_input) {
     }
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
-    ASSERT_EQ(output.count(), output_f);
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    ASSERT_EQ(output->count(), output_f);
     for (size_t i = 0; i < output_f; ++i) {
         auto& val = output_ptr[i];
         float ref;
@@ -506,16 +502,16 @@ TEST(depth_concatenate_f32_gpu, test07_padded_output) {
     const int32_t input_f = 32;
     const int32_t output_f = 2 * input_f;
 
-    const auto& engine = get_test_engine();
-    auto input1 = memory::allocate(engine, { data_types::f16, format::fs_b_yx_fsv32, {1, input_f, 1, 1} });
-    auto input2 = memory::allocate(engine, { data_types::f16, format::fs_b_yx_fsv32, {1, input_f, 1, 1} });
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({ data_types::f16, format::fs_b_yx_fsv32, {1, input_f, 1, 1} });
+    auto input2 = engine.allocate_memory({ data_types::f16, format::fs_b_yx_fsv32, {1, input_f, 1, 1} });
 
     auto input1_data = generate_random_4d<FLOAT16>(1, input_f, 1, 1, -1, 1);
     auto input2_data = generate_random_4d<FLOAT16>(1, input_f, 1, 1, -1, 1);
     set_values(input1, flatten_4d(format::bfyx, input1_data));
     set_values(input2, flatten_4d(format::bfyx, input2_data));
 
-    auto weights = memory::allocate(engine, { data_types::f16, format::oiyx, {output_f, output_f, 3, 3} });
+    auto weights = engine.allocate_memory({ data_types::f16, format::oiyx, {output_f, output_f, 3, 3} });
     // Construct weights for convolution that just double input values.
     VVVVF<FLOAT16> weights_data;
     weights_data.resize(output_f);
@@ -526,8 +522,8 @@ TEST(depth_concatenate_f32_gpu, test07_padded_output) {
     set_values(weights, flatten_4d(format::bfyx, weights_data));
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(activation("actv1", "input1", activation_func::linear, { 0.75f, 0.0f }));
     topology.add(activation("actv2", "input2", activation_func::linear, { 0.5f, 0.0f }));
     topology.add(concatenation("depth1", { "actv1", "actv2" }, concatenation::along_f));
@@ -557,8 +553,8 @@ TEST(depth_concatenate_f32_gpu, test07_padded_output) {
     }
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
-    ASSERT_EQ(output.count(), output_f);
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    ASSERT_EQ(output->count(), output_f);
     for (size_t i = 0; i < output_f; ++i) {
         auto& val = output_ptr[i];
         float ref;
@@ -579,9 +575,9 @@ TEST(depth_concatenate_f32_gpu, test07_concat_is_output) {
     const int32_t input_f = 16;
     const int32_t output_f = 2 * input_f;
 
-    const auto& engine = get_test_engine();
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, {1, input_f, 1, 1} });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, {1, input_f, 1, 1} });
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, {1, input_f, 1, 1} });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, {1, input_f, 1, 1} });
 
     auto input1_data = generate_random_4d<float>(1, input_f, 1, 1, -1, 1);
     auto input2_data = generate_random_4d<float>(1, input_f, 1, 1, -1, 1);
@@ -589,8 +585,8 @@ TEST(depth_concatenate_f32_gpu, test07_concat_is_output) {
     set_values(input2, flatten_4d(format::bfyx, input2_data));
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(activation("actv1", "input1", activation_func::linear, { 0.75f, 0.0f }));
     topology.add(activation("actv2", "input2", activation_func::linear, { 0.5f, 0.0f }));
     topology.add(concatenation("depth1", { "actv1", "actv2" }, concatenation::along_f));
@@ -610,8 +606,8 @@ TEST(depth_concatenate_f32_gpu, test07_concat_is_output) {
     EXPECT_TRUE(executed_primitives.count("depth1") == 1);
 
     auto output = outputs.at("depth1").get_memory();
-    auto output_ptr = output.pointer<float>();
-    ASSERT_EQ(output.count(), output_f);
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    ASSERT_EQ(output->count(), output_f);
     for (size_t i = 0; i < output_f; ++i) {
         auto& val = output_ptr[i];
         float ref;
@@ -625,23 +621,23 @@ TEST(depth_concatenate_f32_gpu, test07_concat_is_output) {
 }
 
 TEST(depth_concatenate_f32_gpu, concat_with_different_format_inputs) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
     const int in1_f = 2, in2_f = 1;
     const int b = 2, x = 2, y = 4;
-    auto input1 = memory::allocate(engine, { data_types::f32, format::yxfb,{ b, in1_f, y, x } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ b, in2_f, y, x } });
-    unsigned input2_start_value = (unsigned)input1.count() + 1;
+    auto input1 = engine.allocate_memory({ data_types::f32, format::yxfb,{ b, in1_f, y, x } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ b, in2_f, y, x } });
+    unsigned input2_start_value = (unsigned)input1->count() + 1;
 
-    std::vector<float> in1(input1.count());
-    std::vector<float> in2(input2.count());
+    std::vector<float> in1(input1->count());
+    std::vector<float> in2(input2->count());
 
-    for (unsigned i = 0; i < input1.count(); i++)
+    for (unsigned i = 0; i < input1->count(); i++)
     {
         in1[i] = (float)(i + 1);
     }
 
-    for (unsigned i = 0; i < input2.count(); i++)
+    for (unsigned i = 0; i < input2->count(); i++)
     {
         in2[i] = (float)(i + input2_start_value);
     }
@@ -652,8 +648,8 @@ TEST(depth_concatenate_f32_gpu, concat_with_different_format_inputs) {
     // Special constrution of topology to run buffer fusing optimization
     // for concatenation with different format inputs
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(concatenation("depth1", { "input1" }, concatenation::along_f));
     topology.add(concatenation("depth2", { "input2" }, concatenation::along_f));
     // In the step below there will be run of buffer fusing optimization for concatenation with
@@ -672,7 +668,7 @@ TEST(depth_concatenate_f32_gpu, concat_with_different_format_inputs) {
     EXPECT_EQ(outputs.begin()->first, "depth4");
 
     auto output = outputs.at("depth4").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     int input1_values_count = in1_f * x;
     int input2_values_count = in2_f * x;
@@ -680,7 +676,7 @@ TEST(depth_concatenate_f32_gpu, concat_with_different_format_inputs) {
     int input2_batch_offset = x * y;
     int out_offset = 0;
 
-    for (unsigned i = 0; i < input1.count(); i++)
+    for (unsigned i = 0; i < input1->count(); i++)
     {
         int value = i + 1;
         EXPECT_FLOAT_EQ(float(value), output_ptr[out_offset++]);
@@ -692,7 +688,7 @@ TEST(depth_concatenate_f32_gpu, concat_with_different_format_inputs) {
     }
 
     out_offset = input1_values_count;
-    for (unsigned i = 0; i < input2.count() / b; i++)
+    for (unsigned i = 0; i < input2->count() / b; i++)
     {
         for (unsigned j = 0; j < b; j++)
         {
@@ -709,9 +705,9 @@ TEST(depth_concatenate_f32_gpu, concat_with_different_format_inputs) {
 
 TEST(depth_concatenate_f32_gpu, concat_with_reshape_input) {
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2,4,1,2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2,4,1,2 } });
 
     std::vector<float> values = {
         0.1f, 0.2f, 0.3f, 0.4f,
@@ -722,7 +718,7 @@ TEST(depth_concatenate_f32_gpu, concat_with_reshape_input) {
     set_values(input1, values);
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
     topology.add(reshape("reshape", "input1", tensor(2, 1, 4, 2)));
     topology.add(concatenation("depth1", { "reshape" }, concatenation::along_f));
     topology.add(concatenation("depth2", { "depth1" }, concatenation::along_f));
@@ -738,7 +734,7 @@ TEST(depth_concatenate_f32_gpu, concat_with_reshape_input) {
 
     auto output = outputs.at("depth2").get_memory();
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -747,13 +743,13 @@ TEST(depth_concatenate_f32_gpu, concat_with_reshape_input) {
 }
 
 TEST(depth_concatenate_i32_gpu, optimize_data01) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
-    auto input = memory::allocate(engine, {data_types::i32, format::bfyx, {1, 1, 1, 1}});
+    auto input = engine.allocate_memory({data_types::i32, format::bfyx, {1, 1, 1, 1}});
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout()));
+        input_layout("input", input->get_layout()));
     topology.add(cldnn::concatenation("int1", {"input"}, cldnn::concatenation::along_f));
     topology.add(cldnn::concatenation("result1", {"int1"}, cldnn::concatenation::along_f));
     topology.add(cldnn::concatenation("result2", {"int1"}, cldnn::concatenation::along_f));
@@ -768,28 +764,28 @@ TEST(depth_concatenate_i32_gpu, optimize_data01) {
     auto outputs = network.execute();
 
     for (auto& it : outputs) {
-        auto output_ptr = it.second.get_memory().pointer<int>();
+        cldnn::mem_lock<int> output_ptr(it.second.get_memory(), get_test_stream());
         EXPECT_EQ(output_ptr[0], out_data[0]);
     }
 }
 
 TEST(depth_concatenate_i32_gpu, optimize_data02) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
-    auto input1 = memory::allocate(engine, {data_types::i32, format::bfyx, {1, 1, 2, 2}});
-    auto input2 = memory::allocate(engine, {data_types::i32, format::bfyx, {1, 1, 2, 2}});
-    auto input3 = memory::allocate(engine, {data_types::i32, format::bfyx, {1, 1, 2, 2}});
-    auto input4 = memory::allocate(engine, {data_types::i32, format::bfyx, {1, 1, 2, 2}});
+    auto input1 = engine.allocate_memory({data_types::i32, format::bfyx, {1, 1, 2, 2}});
+    auto input2 = engine.allocate_memory({data_types::i32, format::bfyx, {1, 1, 2, 2}});
+    auto input3 = engine.allocate_memory({data_types::i32, format::bfyx, {1, 1, 2, 2}});
+    auto input4 = engine.allocate_memory({data_types::i32, format::bfyx, {1, 1, 2, 2}});
 
     topology topology;
     topology.add(
-        input_layout("input1", input1.get_layout()));
+        input_layout("input1", input1->get_layout()));
     topology.add(
-        input_layout("input2", input2.get_layout()));
+        input_layout("input2", input2->get_layout()));
     topology.add(
-        input_layout("input3", input3.get_layout()));
+        input_layout("input3", input3->get_layout()));
     topology.add(
-        input_layout("input4", input4.get_layout()));
+        input_layout("input4", input4->get_layout()));
 
     topology.add(cldnn::concatenation("concat1", {"input1", "input2"}, cldnn::concatenation::along_x));
     topology.add(cldnn::concatenation("concat2", {"input3", "input4"}, cldnn::concatenation::along_x));
@@ -833,7 +829,7 @@ TEST(depth_concatenate_i32_gpu, optimize_data02) {
     network.set_input_data("input4", input4);
     auto outputs = network.execute();
 
-    auto output_concat6 = outputs.at("concat6").get_memory().pointer<int>();
+    cldnn::mem_lock<int> output_concat6(outputs.at("concat6").get_memory(), get_test_stream());
 
     for (size_t i = 0; i < output_concat6.size(); i++) {
         EXPECT_EQ(output_concat6[i], c6_data[i]);
@@ -841,13 +837,13 @@ TEST(depth_concatenate_i32_gpu, optimize_data02) {
 }
 
 TEST(depth_concatenate_i32_gpu, optimize_data03) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
-    auto input1 = memory::allocate(engine, {data_types::i32, format::bfyx, {1, 1, 2, 2}});
+    auto input1 = engine.allocate_memory({data_types::i32, format::bfyx, {1, 1, 2, 2}});
 
     topology topology;
     topology.add(
-        input_layout("input1", input1.get_layout()));
+        input_layout("input1", input1->get_layout()));
 
     topology.add(cldnn::concatenation("concat1", {"input1"}, cldnn::concatenation::along_x));
 
@@ -873,7 +869,7 @@ TEST(depth_concatenate_i32_gpu, optimize_data03) {
     auto outputs = network.execute();
 
     for (auto& it : outputs) {
-        auto output_ptr = it.second.get_memory().pointer<int>();
+        cldnn::mem_lock<int> output_ptr(it.second.get_memory(), get_test_stream());
         for (size_t i = 0; i < output_ptr.size(); i++) {
             EXPECT_EQ(output_ptr[i], output_data[i]);
         }
@@ -881,13 +877,13 @@ TEST(depth_concatenate_i32_gpu, optimize_data03) {
 }
 
 TEST(depth_concatenate_i32_gpu, optimize_data04) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
-    auto input1 = memory::allocate(engine, {data_types::i32, format::bfyx, {1, 1, 2, 2}});
+    auto input1 = engine.allocate_memory({data_types::i32, format::bfyx, {1, 1, 2, 2}});
 
     topology topology;
     topology.add(
-        input_layout("input1", input1.get_layout()));
+        input_layout("input1", input1->get_layout()));
 
     topology.add(cldnn::concatenation("concat1", {"input1"}, cldnn::concatenation::along_x));
 
@@ -913,7 +909,7 @@ TEST(depth_concatenate_i32_gpu, optimize_data04) {
     auto outputs = network.execute();
 
     for (auto& it : outputs) {
-        auto output_ptr = it.second.get_memory().pointer<int>();
+        cldnn::mem_lock<int> output_ptr(it.second.get_memory(), get_test_stream());
         for (size_t i = 0; i < output_ptr.size(); i++) {
             EXPECT_EQ(output_ptr[i], output_data[i]);
         }
@@ -921,13 +917,13 @@ TEST(depth_concatenate_i32_gpu, optimize_data04) {
 }
 
 TEST(depth_concatenate_i32_gpu, optimize_data05) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
-    auto input1 = memory::allocate(engine, {data_types::i32, format::bfyx, {1, 1, 2, 2}});
+    auto input1 = engine.allocate_memory({data_types::i32, format::bfyx, {1, 1, 2, 2}});
 
     topology topology;
     topology.add(
-        input_layout("input1", input1.get_layout()));
+        input_layout("input1", input1->get_layout()));
 
     topology.add(cldnn::concatenation("concat1", {"input1"}, cldnn::concatenation::along_x));
 
@@ -953,7 +949,7 @@ TEST(depth_concatenate_i32_gpu, optimize_data05) {
 
     auto outputs = network.execute();
 
-    auto output_concat5 = outputs.at("concat5").get_memory().pointer<int>();
+    cldnn::mem_lock<int> output_concat5(outputs.at("concat5").get_memory(), get_test_stream());
 
     for (size_t i = 0; i < output_concat5.size(); i++) {
         EXPECT_EQ(output_concat5[i], c5_data[i]);
@@ -961,7 +957,7 @@ TEST(depth_concatenate_i32_gpu, optimize_data05) {
 }
 
 TEST(depth_concatenate_f32_gpu, basic_bfwzyx_along_w) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int b = 2;
     const int f = 3;
     const int x = 2;
@@ -970,16 +966,16 @@ TEST(depth_concatenate_f32_gpu, basic_bfwzyx_along_w) {
     const int w = 9;
 
     auto input1_layout = layout(data_types::f32, format::bfwzyx, tensor{batch(b), feature(f), spatial(x, y, z, w)});
-    auto input1 = memory::allocate(engine, input1_layout);
+    auto input1 = engine.allocate_memory(input1_layout);
     auto output_layout = layout(data_types::f32, format::bfwzyx, tensor{batch(b), feature(f), spatial(x, y, z, w * 2)});
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
     topology.add(concatenation("concat", {"input1", "input1"}, concatenation::along_w));
 
-    auto input_data = generate_random_1d<float>(input1.count(), -1, 1);
+    auto input_data = generate_random_1d<float>(input1->count(), -1, 1);
 
-    auto expected_output = std::vector<float>(input1.count() * 2);
+    auto expected_output = std::vector<float>(input1->count() * 2);
 
     for (int bi = 0; bi < b; bi++)
         for (int fi = 0; fi < f; fi++)
@@ -1002,7 +998,7 @@ TEST(depth_concatenate_f32_gpu, basic_bfwzyx_along_w) {
 
     auto outputs = network.execute();
 
-    auto output_concat = outputs.at("concat").get_memory().pointer<float>();
+    cldnn::mem_lock<float> output_concat(outputs.at("concat").get_memory(), get_test_stream());
 
     ASSERT_EQ(output_concat.size(), expected_output.size());
     for (size_t i = 0; i < output_concat.size(); i++) {
@@ -1021,19 +1017,19 @@ static network setup_depth_concatatenate_network(const std::vector<data_types> d
     assert(dts.size() == ts.size());
     const size_t sz = ts.size();
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     topology topology;
 
     std::vector<std::string> input_names;
     input_names.resize(sz);
 
     for (size_t i = 0; i < sz; ++i) {
-        auto input = memory::allocate(engine, {dts[i], fmt[i], ts[i]});
+        auto input = engine.allocate_memory({dts[i], fmt[i], ts[i]});
 
         input_names[i] = "input";
         input_names[i] += std::to_string(i);
 
-        topology.add(input_layout(input_names[i], input.get_layout()));
+        topology.add(input_layout(input_names[i], input->get_layout()));
     }
     //TODO: ask Uzi if something tests cases where there's missing input_names (nodes not present in the topology, etc.)
     topology.add(concatenation("depth_concat_node", input_names, concatenation::along_f));
@@ -1194,44 +1190,44 @@ class depth_concatenate_test : public tests::generic_test {
     }
 
     template <typename Type>
-    memory generate_reference_typed(const std::vector<memory>& inputs) {
+    memory::ptr generate_reference_typed(const std::vector<memory::ptr>& inputs) {
         assert(!inputs.empty());
 
-        const int in_b = inputs[0].get_layout().size.batch[0];
-        const int in_h = inputs[0].get_layout().size.spatial[1];
-        const int in_w = inputs[0].get_layout().size.spatial[0];
+        const int in_b = inputs[0]->get_layout().size.batch[0];
+        const int in_h = inputs[0]->get_layout().size.spatial[1];
+        const int in_w = inputs[0]->get_layout().size.spatial[0];
 
         int out_f = 0;
 
-        for (const memory& input : inputs) {
-            assert(input.get_layout().size.batch[0] == in_b);
-            assert(input.get_layout().size.spatial[1] == in_h);
-            assert(input.get_layout().size.spatial[0] == in_w);
+        for (const memory::ptr& input : inputs) {
+            assert(input->get_layout().size.batch[0] == in_b);
+            assert(input->get_layout().size.spatial[1] == in_h);
+            assert(input->get_layout().size.spatial[0] == in_w);
 
-            out_f += input.get_layout().size.feature[0];
+            out_f += input->get_layout().size.feature[0];
 
-            assert(input.get_layout().data_type == inputs[0].get_layout().data_type);
-            assert(input.get_layout().format.value == inputs[0].get_layout().format.value);
+            assert(input->get_layout().data_type == inputs[0]->get_layout().data_type);
+            assert(input->get_layout().format.value == inputs[0]->get_layout().format.value);
         }
 
         //Output is bfyx
-        auto output = memory::allocate(engine, cldnn::layout(inputs[0].get_layout().data_type, cldnn::format::bfyx, tensor(in_b, out_f, in_w, in_h)));
-        auto out_mem = output.pointer<Type>();
+        auto output = engine.allocate_memory(cldnn::layout(inputs[0]->get_layout().data_type, cldnn::format::bfyx, tensor(in_b, out_f, in_w, in_h)));
+        cldnn::mem_lock<Type> out_mem(output, get_test_stream());
 
         int out_f_off = 0;
-        for (const memory& input : inputs) {
-            const auto input_desc = get_linear_memory_desc(input.get_layout());
-            const auto output_desc = get_linear_memory_desc(output.get_layout());
+        for (const memory::ptr& input : inputs) {
+            const auto input_desc = get_linear_memory_desc(input->get_layout());
+            const auto output_desc = get_linear_memory_desc(output->get_layout());
 
-            const int in_f = input.get_layout().size.feature[0];
-            const auto in_mem = input.pointer<Type>();
+            const int in_f = input->get_layout().size.feature[0];
+            cldnn::mem_lock<Type> in_mem(input, get_test_stream());
 
             for (int n = 0; n < in_b; ++n)
                 for (int f = 0; f < in_f; ++f)
                     for (int y = 0; y < in_h; ++y)
                         for (int x = 0; x < in_w; ++x) {
-                            const size_t in_idx = get_linear_index(input.get_layout(), n, f, y, x, input_desc);
-                            const size_t out_idx = get_linear_index(output.get_layout(), n, out_f_off + f, y, x, output_desc);
+                            const size_t in_idx = get_linear_index(input->get_layout(), n, f, y, x, input_desc);
+                            const size_t out_idx = get_linear_index(output->get_layout(), n, out_f_off + f, y, x, output_desc);
 
                             out_mem[out_idx] = in_mem[in_idx];
                         }
@@ -1242,7 +1238,7 @@ class depth_concatenate_test : public tests::generic_test {
         return output;
     }
 
-    virtual memory generate_reference(const std::vector<memory>& inputs) override {
+    virtual memory::ptr generate_reference(const std::vector<memory::ptr>& inputs) override {
         if (generic_params->data_type == data_types::f32) {
             return generate_reference_typed<float>(inputs);
         } else {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/depth_to_space_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/depth_to_space_gpu_test.cpp
index 6f608e9efe5bf8..8e67dfbb4eb8cd 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/depth_to_space_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/depth_to_space_gpu_test.cpp
@@ -3,19 +3,16 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/depth_to_space.hpp>
-#include <api/topology.hpp>
-#include <api/reshape.hpp>
-#include <api/network.hpp>
-#include "api/permute.hpp"
-#include "api/reorder.hpp"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/depth_to_space.hpp>
+#include <cldnn/primitives/reshape.hpp>
+#include <cldnn/primitives/permute.hpp>
+#include <cldnn/primitives/reorder.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -26,9 +23,9 @@ TEST(depth_to_space_fp16_gpu, d1411_bs2) {
     //  Output : 1x1x2x2
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 4, 1, 1 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 4, 1, 1 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -37,7 +34,7 @@ TEST(depth_to_space_fp16_gpu, d1411_bs2) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
         depth_to_space("depth_to_space", "Input0", block_size, depth_to_space_mode::blocks_first)
     );
@@ -49,7 +46,7 @@ TEST(depth_to_space_fp16_gpu, d1411_bs2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("depth_to_space").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.f, 2.f, 3.f
@@ -66,9 +63,9 @@ TEST(depth_to_space_fp16_gpu, d1421_bs2) {
     //  Output : 1x1x4x2
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 4, 1, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 4, 1, 2 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -79,7 +76,7 @@ TEST(depth_to_space_fp16_gpu, d1421_bs2) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
         depth_to_space("depth_to_space", "Input0", block_size, depth_to_space_mode::blocks_first)
     );
@@ -91,7 +88,7 @@ TEST(depth_to_space_fp16_gpu, d1421_bs2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("depth_to_space").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.0f, 2.0f, 4.0f, 6.0f, 1.0f, 3.0f, 5.0f, 7.0f
@@ -108,9 +105,9 @@ TEST(depth_to_space_fp16_gpu, d1933_bs3) {
     //  Output : 1x1x9x9
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 9, 3, 3 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 9, 3, 3 } });
     size_t block_size = 3;
 
     set_values(input1, {
@@ -134,7 +131,7 @@ TEST(depth_to_space_fp16_gpu, d1933_bs3) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             depth_to_space("depth_to_space", "Input0", block_size, depth_to_space_mode::blocks_first)
     );
@@ -146,7 +143,7 @@ TEST(depth_to_space_fp16_gpu, d1933_bs3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("depth_to_space").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.0f, 9.0f, 18.0f, 1.0f, 10.0f, 19.0f, 2.0f, 11.0f, 20.0f, 27.0f,
@@ -171,9 +168,9 @@ TEST(depth_to_space_fp32_gpu, d1411_bs2) {
     //  Output : 1x1x2x2
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 1, 1 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 1, 1 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -181,7 +178,7 @@ TEST(depth_to_space_fp32_gpu, d1411_bs2) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
         depth_to_space("depth_to_space", "Input0", block_size, depth_to_space_mode::blocks_first)
     );
@@ -193,7 +190,7 @@ TEST(depth_to_space_fp32_gpu, d1411_bs2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("depth_to_space").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.f, 2.f, 3.f
@@ -210,9 +207,9 @@ TEST(depth_to_space_fp32_gpu, d112960540_bs2) {
     //  Output : 1x3x1920x1080
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 12, 960, 540 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 12, 960, 540 } });
     size_t block_size = 2;
 
     auto random_input = generate_random_4d<FLOAT16>(1, 12, 540, 960, -1, 1);
@@ -220,7 +217,7 @@ TEST(depth_to_space_fp32_gpu, d112960540_bs2) {
     set_values(input1, input_rnd_vec);
 
     topology topology_act;
-    topology_act.add(input_layout("Input0", input1.get_layout()));
+    topology_act.add(input_layout("Input0", input1->get_layout()));
     topology_act.add(
         depth_to_space("depth_to_space", "Input0", block_size, depth_to_space_mode::blocks_first)
     );
@@ -232,12 +229,12 @@ TEST(depth_to_space_fp32_gpu, d112960540_bs2) {
     auto outputs = network_act.execute();
 
     auto output = outputs.at("depth_to_space").get_memory();
-    auto output_ptr = output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_ptr (output, get_test_stream());
 
     std::vector<uint16_t> perm = { 0,4,5,2,1,3 };
 
     topology topology_ref;
-    topology_ref.add(input_layout("Input0", input1.get_layout()));
+    topology_ref.add(input_layout("Input0", input1->get_layout()));
     topology_ref.add(reorder("reorder1", "Input0", { data_types::f16, format::bfwzyx, tensor{ batch(1), feature(12), spatial(1, 1, 960, 540) }
         }));
     topology_ref.add(
@@ -259,9 +256,9 @@ TEST(depth_to_space_fp32_gpu, d112960540_bs2) {
     auto outputs_ref = network_ref.execute();
 
     auto output_ref = outputs_ref.at("reshape2").get_memory();
-    auto output_ptr_ref = output_ref.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_ptr_ref(output_ref, get_test_stream());
 
-    for (size_t i = 0; i < output.get_layout().count(); ++i) {
+    for (size_t i = 0; i < output->get_layout().count(); ++i) {
         EXPECT_EQ(output_ptr_ref[i], output_ptr[i]);
     }
 }
@@ -272,9 +269,9 @@ TEST(depth_to_space_fp32_gpu, d1933_bs3) {
     //  Output : 1x1x9x9
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 9, 3, 3 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 9, 3, 3 } });
     size_t block_size = 3;
 
     set_values(input1, {
@@ -290,7 +287,7 @@ TEST(depth_to_space_fp32_gpu, d1933_bs3) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
         depth_to_space("depth_to_space", "Input0", block_size, depth_to_space_mode::blocks_first)
     );
@@ -302,7 +299,7 @@ TEST(depth_to_space_fp32_gpu, d1933_bs3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("depth_to_space").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.0f, 9.0f, 18.0f, 1.0f, 10.0f, 19.0f, 2.0f, 11.0f, 20.0f, 27.0f,
@@ -328,9 +325,9 @@ TEST(depth_to_space_fp32_gpu, d1822_bs2_blocks_first) {
     //  Output : 1x2x4x4
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 8, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 8, 2, 2 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -345,7 +342,7 @@ TEST(depth_to_space_fp32_gpu, d1822_bs2_blocks_first) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
         depth_to_space("depth_to_space", "Input0", block_size, depth_to_space_mode::blocks_first)
     );
@@ -357,7 +354,7 @@ TEST(depth_to_space_fp32_gpu, d1822_bs2_blocks_first) {
     auto outputs = network.execute();
 
     auto output = outputs.at("depth_to_space").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.0f, 8.0f, 1.0f, 9.0f, 16.0f, 24.0f, 17.0f, 25.0f,
@@ -378,9 +375,9 @@ TEST(depth_to_space_fp32_gpu, d1822_bs2_depth_first) {
     //  Output : 1x2x4x4
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 8, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 8, 2, 2 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -395,7 +392,7 @@ TEST(depth_to_space_fp32_gpu, d1822_bs2_depth_first) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
         depth_to_space("depth_to_space", "Input0", block_size, depth_to_space_mode::depth_first)
     );
@@ -407,7 +404,7 @@ TEST(depth_to_space_fp32_gpu, d1822_bs2_depth_first) {
     auto outputs = network.execute();
 
     auto output = outputs.at("depth_to_space").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.0f, 4.0f, 1.0f, 5.0f, 8.0f, 12.0f, 9.0f, 13.0f,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/detection_output_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/detection_output_test.cpp
index 2112f227ca29fb..5d85613c1c5025 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/detection_output_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/detection_output_test.cpp
@@ -3,21 +3,18 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/detection_output.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/detection_output.hpp>
 
 namespace cldnn {
     template<> struct type_to_data_type<FLOAT16> { static const data_types value = data_types::f16; };
 }
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 template <typename T>
 class detection_output_test : public ::testing::Test {
@@ -26,12 +23,12 @@ class detection_output_test : public ::testing::Test {
     detection_output_test() :
         nms_threshold(0.1f) {}
 
-    void init_buffers(cldnn::memory prior_memory, cldnn::memory confidence_memory, cldnn::memory location_memory,
+    void init_buffers(cldnn::memory::ptr prior_memory, cldnn::memory::ptr confidence_memory, cldnn::memory::ptr location_memory,
                       bool share_location, bool variance_encoded_in_target = false,
                       int prior_info_size = 4, int prior_coordinates_offset = 0, bool prior_is_normalized = true) {
-        auto location_ptr = location_memory.pointer<T>();
-        auto confidence_ptr = confidence_memory.pointer<T>();
-        auto prior_box_ptr = prior_memory.pointer<T>();
+        cldnn::mem_lock<T> location_ptr(location_memory, get_test_stream());
+        cldnn::mem_lock<T> confidence_ptr(confidence_memory, get_test_stream());
+        cldnn::mem_lock<T> prior_box_ptr(prior_memory, get_test_stream());
 
         T* prior_data = prior_box_ptr.data();
         T* confidence_data = confidence_ptr.data();
@@ -93,10 +90,10 @@ class detection_output_test : public ::testing::Test {
         }
     }
 
-    void init_buffer_sort(cldnn::memory input_buff) {
-        auto input_data_ptr = input_buff.pointer<T>();
+    void init_buffer_sort(cldnn::memory::ptr input_buff) {
+        cldnn::mem_lock<T> input_data_ptr(input_buff, get_test_stream());
 
-        EXPECT_EQ((int)input_buff.count(), 128);
+        EXPECT_EQ((int)input_buff->count(), 128);
 
         T* input_data = input_data_ptr.data();
         input_data[0] = 8;
@@ -119,8 +116,8 @@ class detection_output_test : public ::testing::Test {
         input_data[121] = -1; input_data[122] = 0; input_data[123] = 0; input_data[124] = 0; input_data[125] = 0; input_data[126] = 0; input_data[127] = 0;
     }
 
-    void check_results(const memory& output, const int num, const std::string values) {
-        assert(num < output.get_layout().size.spatial[1]);
+    void check_results(const memory::ptr output, const int num, const std::string values) {
+        assert(num < output->get_layout().size.spatial[1]);
 
         // Split values to vector of items.
         std::vector<std::string> items;
@@ -129,13 +126,13 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ((int)items.size(), 7);
 
         // Check data.
-        auto out_ptr = output.pointer<T>();
+        cldnn::mem_lock<T> out_ptr(output, get_test_stream());
         const T* data = out_ptr.data();
         for (int i = 0; i < 2; ++i) {
-            EXPECT_EQ(static_cast<int>((float)data[num * output.get_layout().size.spatial[0] + i]), atoi(items[i].c_str()));
+            EXPECT_EQ(static_cast<int>((float)data[num * output->get_layout().size.spatial[0] + i]), atoi(items[i].c_str()));
         }
         for (int i = 2; i < 7; ++i) {
-            EXPECT_TRUE(floating_point_equal(data[num * output.get_layout().size.spatial[0] + i], (T)(float)atof(items[i].c_str())));
+            EXPECT_TRUE(floating_point_equal(data[num * output->get_layout().size.spatial[0] + i], (T)(float)atof(items[i].c_str())));
         }
     }
 
@@ -144,15 +141,15 @@ class detection_output_test : public ::testing::Test {
         const int num_loc_classes = share_location ? 1 : this->num_classes;
         const int keep_top_k = 150;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
 
         topology.add(detection_output("detection_output", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k));
 
@@ -167,10 +164,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
     }
 
     void setup_two_layers() {
@@ -178,15 +175,15 @@ class detection_output_test : public ::testing::Test {
         const int num_loc_classes = share_location ? 1 : this->num_classes;
         const int keep_top_k = 150;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
 
         topology.add(detection_output("detection_output_1", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k));
         topology.add(detection_output("detection_output_2", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k));
@@ -205,10 +202,10 @@ class detection_output_test : public ::testing::Test {
 
             EXPECT_EQ(it->first, "detection_output_" + std::to_string(i));
 
-            EXPECT_EQ(it->second.get_memory().get_layout().size.batch[0], 1);
-            EXPECT_EQ(it->second.get_memory().get_layout().size.feature[0], 1);
-            EXPECT_EQ(it->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-            EXPECT_EQ(it->second.get_memory().get_layout().size.spatial[0], 7);
+            EXPECT_EQ(it->second.get_memory()->get_layout().size.batch[0], 1);
+            EXPECT_EQ(it->second.get_memory()->get_layout().size.feature[0], 1);
+            EXPECT_EQ(it->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+            EXPECT_EQ(it->second.get_memory()->get_layout().size.spatial[0], 7);
             i++;
         }
     }
@@ -219,17 +216,17 @@ class detection_output_test : public ::testing::Test {
         const int keep_top_k = 4;
         const int background_label_id = 0;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         this->init_buffers(input_prior_box, input_confidence, input_location, share_location);
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
 
         topology.add(detection_output("detection_output", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k, share_location, background_label_id, this->nms_threshold));
 
@@ -244,10 +241,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
 
         auto output_prim = outputs.begin()->second.get_memory();
 
@@ -267,17 +264,17 @@ class detection_output_test : public ::testing::Test {
         const int keep_top_k = 1;
         const int background_label_id = 0;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         this->init_buffers(input_prior_box, input_confidence, input_location, share_location);
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
 
         topology.add(detection_output("detection_output", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k, share_location, background_label_id, this->nms_threshold));
 
@@ -292,10 +289,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
 
         auto output_prim = outputs.begin()->second.get_memory();
 
@@ -309,17 +306,17 @@ class detection_output_test : public ::testing::Test {
         const int keep_top_k = 6;
         const int background_label_id = 0;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         this->init_buffers(input_prior_box, input_confidence, input_location, share_location);
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
 
         topology.add(detection_output("detection_output", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k, share_location, background_label_id, this->nms_threshold));
 
@@ -334,10 +331,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
 
         auto output_prim = outputs.begin()->second.get_memory();
 
@@ -362,17 +359,17 @@ class detection_output_test : public ::testing::Test {
         const int top_k = 2;
         const int background_label_id = 0;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         this->init_buffers(input_prior_box, input_confidence, input_location, share_location);
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
 
         topology.add(detection_output("detection_output", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k, share_location, background_label_id, this->nms_threshold, top_k));
 
@@ -387,10 +384,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
 
         auto output_prim = outputs.begin()->second.get_memory();
 
@@ -406,17 +403,17 @@ class detection_output_test : public ::testing::Test {
         const int keep_top_k = 10;
         const int background_label_id = -1;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         this->init_buffers(input_prior_box, input_confidence, input_location, share_location);
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
 
         topology.add(detection_output("detection_output", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k, share_location, background_label_id, this->nms_threshold));
 
@@ -431,10 +428,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
 
         auto output_prim = outputs.begin()->second.get_memory();
 
@@ -467,17 +464,17 @@ class detection_output_test : public ::testing::Test {
         const int background_label_id = -1;
         const int top_k = 2;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         this->init_buffers(input_prior_box, input_confidence, input_location, share_location);
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
 
         topology.add(detection_output("detection_output", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k, share_location, background_label_id, this->nms_threshold, top_k));
 
@@ -492,10 +489,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
 
         auto output_prim = outputs.begin()->second.get_memory();
 
@@ -515,17 +512,17 @@ class detection_output_test : public ::testing::Test {
         const int keep_top_k = 5;
         const int background_label_id = 0;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         this->init_buffers(input_prior_box, input_confidence, input_location, share_location);
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
 
         topology.add(detection_output("detection_output", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k, share_location, background_label_id, this->nms_threshold));
 
@@ -540,10 +537,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
 
         auto output_prim = outputs.begin()->second.get_memory();
 
@@ -566,17 +563,17 @@ class detection_output_test : public ::testing::Test {
         const int background_label_id = 0;
         const int top_k = 2;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         this->init_buffers(input_prior_box, input_confidence, input_location, share_location);
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
 
         topology.add(detection_output("detection_output", "input_location", "input_confidence", "input_prior_box", this->num_classes, keep_top_k, share_location, background_label_id, this->nms_threshold, top_k));
 
@@ -591,10 +588,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
 
         auto output_prim = outputs.begin()->second.get_memory();
 
@@ -611,18 +608,18 @@ class detection_output_test : public ::testing::Test {
         const int background_label_id = -1;
         const int top_k = 2;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 2, 1, this->num_priors * 4 } });
 
         this->init_buffers(input_prior_box, input_confidence, input_location, share_location);
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
-        topology.add(reorder("input_location_padded", "input_location", input_location.get_layout().with_padding(padding{ { 0, 0, 12, 3 },{ 0, 0, 5, 11 } })));
-        topology.add(reorder("input_confidence_padded", "input_confidence", input_location.get_layout().with_padding(padding{ { 0, 0, 2, 7 },{ 0, 0, 13, 1 } })));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
+        topology.add(reorder("input_location_padded", "input_location", input_location->get_layout().with_padding(padding{ { 0, 0, 12, 3 },{ 0, 0, 5, 11 } })));
+        topology.add(reorder("input_confidence_padded", "input_confidence", input_location->get_layout().with_padding(padding{ { 0, 0, 2, 7 },{ 0, 0, 13, 1 } })));
 
         topology.add(detection_output("detection_output", "input_location_padded", "input_confidence_padded", "input_prior_box", this->num_classes, keep_top_k, share_location, background_label_id, this->nms_threshold, top_k));
 
@@ -637,10 +634,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
 
         auto output_prim = outputs.begin()->second.get_memory();
 
@@ -668,20 +665,20 @@ class detection_output_test : public ::testing::Test {
         const int32_t prior_coordinates_offset = 1;
         const bool prior_is_normalized = true;
 
-        const auto& engine = get_test_engine();
-        cldnn::memory input_location = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
-        cldnn::memory input_confidence = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
-        cldnn::memory input_prior_box = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1, 1, 1, this->num_priors * prior_info_size } });
+        auto& engine = get_test_engine();
+        cldnn::memory::ptr input_location = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * num_loc_classes * 4, 1, 1 } });
+        cldnn::memory::ptr input_confidence = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ this->num_of_images, this->num_priors * this->num_classes, 1, 1 } });
+        cldnn::memory::ptr input_prior_box = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1, 1, 1, this->num_priors * prior_info_size } });
 
         this->init_buffers(input_prior_box, input_confidence, input_location, share_location, variance_encoded_in_target,
             prior_info_size, prior_coordinates_offset, prior_is_normalized);
 
         topology topology;
-        topology.add(input_layout("input_location", input_location.get_layout()));
-        topology.add(input_layout("input_confidence", input_confidence.get_layout()));
-        topology.add(input_layout("input_prior_box", input_prior_box.get_layout()));
-        topology.add(reorder("input_location_padded", "input_location", input_location.get_layout().with_padding(padding{ { 0, 0, 12, 3 },{ 0, 0, 5, 11 } })));
-        topology.add(reorder("input_confidence_padded", "input_confidence", input_location.get_layout().with_padding(padding{ { 0, 0, 2, 7 },{ 0, 0, 13, 1 } })));
+        topology.add(input_layout("input_location", input_location->get_layout()));
+        topology.add(input_layout("input_confidence", input_confidence->get_layout()));
+        topology.add(input_layout("input_prior_box", input_prior_box->get_layout()));
+        topology.add(reorder("input_location_padded", "input_location", input_location->get_layout().with_padding(padding{ { 0, 0, 12, 3 },{ 0, 0, 5, 11 } })));
+        topology.add(reorder("input_confidence_padded", "input_confidence", input_location->get_layout().with_padding(padding{ { 0, 0, 2, 7 },{ 0, 0, 13, 1 } })));
 
         topology.add(detection_output("detection_output", "input_location_padded", "input_confidence_padded", "input_prior_box",
             this->num_classes, keep_top_k, share_location, background_label_id, this->nms_threshold, top_k,
@@ -700,10 +697,10 @@ class detection_output_test : public ::testing::Test {
         EXPECT_EQ(outputs.size(), size_t(1));
         EXPECT_EQ(outputs.begin()->first, "detection_output");
 
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.batch[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.feature[0], 1);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[1], keep_top_k * this->num_of_images);
-        EXPECT_EQ(outputs.begin()->second.get_memory().get_layout().size.spatial[0], 7);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.batch[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.feature[0], 1);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[1], keep_top_k * this->num_of_images);
+        EXPECT_EQ(outputs.begin()->second.get_memory()->get_layout().size.spatial[0], 7);
 
         auto output_prim = outputs.begin()->second.get_memory();
 
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/eltwise_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/eltwise_gpu_test.cpp
index 32456c52da362b..016f4b98c69bbc 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/eltwise_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/eltwise_gpu_test.cpp
@@ -3,16 +3,13 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/eltwise.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include <api/reorder.hpp>
-#include <api/data.hpp>
-#include "test_utils/test_utils.h"
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/eltwise.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
 
 namespace cldnn
 {
@@ -20,7 +17,7 @@ namespace cldnn
 }
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 template <typename T>
 T eltwise_execute(cldnn::eltwise_mode mode, T x, T y) {
@@ -88,17 +85,17 @@ void generic_eltwise_test(cldnn::format test_input_fmt, int input_b, int input_f
     VF<T> input1_rnd_vec = flatten_4d<T>(test_input_fmt, input1_rnd);
     VF<T> input2_rnd_vec = flatten_4d<T>(test_input_fmt, input2_rnd);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor( input_b, input_f, input_x, input_y );
-    auto input1 = memory::allocate(engine, { type_to_data_type<T>::value, test_input_fmt, input_tensor });
-    auto input2 = memory::allocate(engine, { type_to_data_type<T>::value, test_input_fmt, input_tensor });
+    auto input1 = engine.allocate_memory({ type_to_data_type<T>::value, test_input_fmt, input_tensor });
+    auto input2 = engine.allocate_memory({ type_to_data_type<T>::value, test_input_fmt, input_tensor });
     set_values(input1, input1_rnd_vec);
     set_values(input2, input2_rnd_vec);
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(reorder("reorder1", "input1", input1.get_layout().with_padding(padding{{ 0, 0, input_padding_x, input_padding_y }, 0 })));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(reorder("reorder1", "input1", input1->get_layout().with_padding(padding{{ 0, 0, input_padding_x, input_padding_y }, 0 })));
     topology.add(eltwise("eltwise", {"reorder1", "input2"}, mode, padding{ { 0, 0, output_padding_x, output_padding_y }, 0 }));
     primitive_id out_id = "eltwise";
     if (relu)
@@ -114,8 +111,8 @@ void generic_eltwise_test(cldnn::format test_input_fmt, int input_b, int input_f
     EXPECT_EQ(outputs.begin()->first, out_id);
 
     auto output_memory = outputs.at(out_id).get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<T>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<T> output_ptr(output_memory, get_test_stream());
 
     VVVVF<T> output_cpu = eltwise_reference<T>(input1_rnd, input2_rnd, mode, relu, slope, input_padding_y, input_padding_x, output_padding_y, output_padding_x);
     EXPECT_EQ(output_layout.format.value, test_input_fmt.value);
@@ -176,10 +173,10 @@ TEST(eltwise_gpu_f32, equal_in2_float_out1_int) {
     //  0, 0, 0, 0,
     //  0, 1, 0, 0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -196,8 +193,8 @@ TEST(eltwise_gpu_f32, equal_in2_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::eq));
 
     network network(engine, topology);
@@ -211,7 +208,7 @@ TEST(eltwise_gpu_f32, equal_in2_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 0, 1, 0, 1,
                                     0, 0, 1, 0,
@@ -246,10 +243,10 @@ TEST(eltwise_gpu_f32, not_equal_in2_float_out1_int) {
     //  1, 1, 1, 1,
     //  1, 0, 1, 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -266,8 +263,8 @@ TEST(eltwise_gpu_f32, not_equal_in2_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::ne));
 
     network network(engine, topology);
@@ -281,7 +278,7 @@ TEST(eltwise_gpu_f32, not_equal_in2_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 1, 0, 1, 0,
                                     1, 1, 0, 1,
@@ -316,10 +313,10 @@ TEST(eltwise_gpu_f32, less_in2_float_out1_int) {
     //  1, 1, 1, 0,
     //  0, 0, 0, 0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -336,8 +333,8 @@ TEST(eltwise_gpu_f32, less_in2_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::lt));
 
     network network(engine, topology);
@@ -351,7 +348,7 @@ TEST(eltwise_gpu_f32, less_in2_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 0, 0, 0, 0,
                                     1, 1, 0, 0,
@@ -386,10 +383,10 @@ TEST(eltwise_gpu_f32, less_equal_in2_float_out1_int) {
     //  1, 1, 1, 0,
     //  0, 1, 0, 0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -406,8 +403,8 @@ TEST(eltwise_gpu_f32, less_equal_in2_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::le));
 
     network network(engine, topology);
@@ -421,7 +418,7 @@ TEST(eltwise_gpu_f32, less_equal_in2_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 0, 1, 0, 1,
                                     1, 1, 1, 0,
@@ -456,10 +453,10 @@ TEST(eltwise_gpu_f32, greater_in2_float_out1_int) {
     //  0, 0, 0, 1,
     //  1, 0, 1, 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -476,8 +473,8 @@ TEST(eltwise_gpu_f32, greater_in2_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::gt));
 
     network network(engine, topology);
@@ -491,7 +488,7 @@ TEST(eltwise_gpu_f32, greater_in2_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 1, 0, 1, 0,
                                     0, 0, 0, 1,
@@ -526,10 +523,10 @@ TEST(eltwise_gpu_f32, greater_equal_in2_float_out1_int) {
     //  0, 0, 0, 1,
     //  1, 1, 1, 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -546,8 +543,8 @@ TEST(eltwise_gpu_f32, greater_equal_in2_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::ge));
 
     network network(engine, topology);
@@ -561,7 +558,7 @@ TEST(eltwise_gpu_f32, greater_equal_in2_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 1, 1, 1, 1,
                                     0, 0, 1, 1,
@@ -596,10 +593,10 @@ TEST(eltwise_gpu_f32, logicalAND_in2_float_out1_int) {
     //  1, 1, 1, 1,
     //  1, 0, 1, 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -616,8 +613,8 @@ TEST(eltwise_gpu_f32, logicalAND_in2_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::logic_and));
 
     network network(engine, topology);
@@ -631,7 +628,7 @@ TEST(eltwise_gpu_f32, logicalAND_in2_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 1, 1, 1, 1,
                                     1, 0, 1, 1,
@@ -673,11 +670,11 @@ TEST(eltwise_gpu_f32, logicalAND_in3_float_out1_int) {
     //  1, 1, 1, 1,
     //  1, 0, 1, 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -701,9 +698,9 @@ TEST(eltwise_gpu_f32, logicalAND_in3_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("input3", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("input3", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2", "input3"}, eltwise_mode::logic_and));
 
     network network(engine, topology);
@@ -718,7 +715,7 @@ TEST(eltwise_gpu_f32, logicalAND_in3_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 0, 0, 0, 0,
                                     0, 0, 0, 0,
@@ -753,10 +750,10 @@ TEST(eltwise_gpu_f32, logicalOR_in2_float_out1_int) {
     //  1, 1, 1, 1,
     //  1, 0, 1, 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -773,8 +770,8 @@ TEST(eltwise_gpu_f32, logicalOR_in2_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::logic_or));
 
     network network(engine, topology);
@@ -788,7 +785,7 @@ TEST(eltwise_gpu_f32, logicalOR_in2_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 1, 1, 1, 1,
                                     1, 1, 1, 1,
@@ -830,11 +827,11 @@ TEST(eltwise_gpu_f32, logicalOR_in3_float_out1_int) {
     //  1, 1, 1, 1,
     //  1, 1, 1, 1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -858,9 +855,9 @@ TEST(eltwise_gpu_f32, logicalOR_in3_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("input3", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("input3", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2", "input3"}, eltwise_mode::logic_or));
 
     network network(engine, topology);
@@ -875,7 +872,7 @@ TEST(eltwise_gpu_f32, logicalOR_in3_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 1, 1, 1, 1,
                                     1, 1, 1, 1,
@@ -910,10 +907,10 @@ TEST(eltwise_gpu_f32, logicalXOR_in2_float_out1_int) {
     //  0, 0, 0, 0,
     //  0, 0, 0, 0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     set_values(input1, {
             1.f,   2.5f, 5.f, 1.5f,
@@ -930,8 +927,8 @@ TEST(eltwise_gpu_f32, logicalXOR_in2_float_out1_int) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::logic_xor));
 
     network network(engine, topology);
@@ -945,7 +942,7 @@ TEST(eltwise_gpu_f32, logicalXOR_in2_float_out1_int) {
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output = outputs.at("eltwise").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
     std::vector<int8_t> answers = { 0, 0, 0, 0,
                                     0, 1, 0, 0,
@@ -981,14 +978,14 @@ TEST(eltwise_gpu_f32, add_basic_in4x4x2x2) {
     //  f1: b0:   15  16.5  b1:  22    16.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::sum));
 
     set_values(input, {
@@ -1020,7 +1017,7 @@ TEST(eltwise_gpu_f32, add_basic_in4x4x2x2) {
                           18.f,17.5f,   15.f,   22.f,
                           2.f,   6.f,   7.5f,  5.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1029,14 +1026,14 @@ TEST(eltwise_gpu_f32, add_basic_in4x4x2x2) {
 }
 
 TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_channel) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::sum));
 
     set_values(input, {
@@ -1083,7 +1080,7 @@ TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_channel) {
                           1.f,   7.f,
                           3.5f, -3.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1092,14 +1089,14 @@ TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_channel) {
 }
 
 TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_x) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::sum));
 
     set_values(input, {
@@ -1152,7 +1149,7 @@ TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_x) {
                           2.f,   10.5f,
                           -1.f, -3.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1161,14 +1158,14 @@ TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_x) {
 }
 
 TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_y) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::sum));
 
     set_values(input, {
@@ -1215,7 +1212,7 @@ TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_y) {
                           1.f,   7.f,
                           3.5f, -3.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1224,14 +1221,14 @@ TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_y) {
 }
 
 TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_batch) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::sum));
 
     set_values(input, {
@@ -1280,7 +1277,7 @@ TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_batch) {
                           2.f,   6.f,
                           3.5f, -3.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1289,14 +1286,14 @@ TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_batch) {
 }
 
 TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_multiple_dims) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::sum));
 
     set_values(input, {
@@ -1339,7 +1336,7 @@ TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_multiple_dims) {
                           0.f,   8.5f,
                           1.5f, -0.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1348,14 +1345,14 @@ TEST(eltwise_gpu_f32, add_in2x2x2x2_broadcast_multiple_dims) {
 }
 
 TEST(eltwise_gpu_f32, pow_in2x2x2x2_broadcast_all) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::pow));
 
     set_values(input, {
@@ -1396,7 +1393,7 @@ TEST(eltwise_gpu_f32, pow_in2x2x2x2_broadcast_all) {
                           169.f, 196.f,
                           225.f, 256.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1405,16 +1402,16 @@ TEST(eltwise_gpu_f32, pow_in2x2x2x2_broadcast_all) {
 }
 
 TEST(eltwise_gpu_f32, add_basic_in2x2x2x2_broadcast_2_inputs_same_dim) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("input3", input3.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("input3", input3->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2", "input3"}, eltwise_mode::sum));
 
     set_values(input, {
@@ -1474,7 +1471,7 @@ TEST(eltwise_gpu_f32, add_basic_in2x2x2x2_broadcast_2_inputs_same_dim) {
                           -2.f,  6.5f,
                           -0.5f, -2.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1483,16 +1480,16 @@ TEST(eltwise_gpu_f32, add_basic_in2x2x2x2_broadcast_2_inputs_same_dim) {
 }
 
 TEST(eltwise_gpu_f32, add_basic_in2x2x2x2_broadcast_2_inputs_diff_dim) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("input3", input3.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("input3", input3->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2", "input3"}, eltwise_mode::sum));
 
     set_values(input, {
@@ -1550,7 +1547,7 @@ TEST(eltwise_gpu_f32, add_basic_in2x2x2x2_broadcast_2_inputs_diff_dim) {
                            0.f,   7.5f,
                           -0.5f, -2.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1581,14 +1578,14 @@ TEST(eltwise_gpu_f32, max_basic_in4x4x4x4) {
     //  f1: b0:    5   6    b1:  2.5   5.2
     //  f1: b0:    8   8    b1:  12    8
     //
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::max));
 
     set_values(input, {
@@ -1621,7 +1618,7 @@ TEST(eltwise_gpu_f32, max_basic_in4x4x4x4) {
         15.f,  17.f,   8.f,  12.f,
          6.f,   8.f,   8.f,   8.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1653,13 +1650,13 @@ TEST(eltwise_gpu_f32, sub_basic_in4x4x4x4) {
     //  f1: b0:   -1    8.5  b1:  3.5   -2.5
     //
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::sub));
 
     set_values(input, {
@@ -1692,7 +1689,7 @@ TEST(eltwise_gpu_f32, sub_basic_in4x4x4x4) {
        -12.f,  -16.5f,  -1.f,    3.5f,
         -2.f,   -8.5f,   8.5f,  -2.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1710,13 +1707,13 @@ TEST(eltwise_gpu_int, basic_in4x4x4x4) {
     {
         for (auto& mode : eltwise_ops_to_test)
         {
-            const auto& engine = get_test_engine();
-            auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-            auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+            auto& engine = get_test_engine();
+            auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+            auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
             topology topology;
-            topology.add(input_layout("input", input.get_layout()));
-            topology.add(input_layout("input2", input2.get_layout()));
+            topology.add(input_layout("input", input->get_layout()));
+            topology.add(input_layout("input2", input2->get_layout()));
             topology.add(reorder("input_reorder", "input", { data_type, format::yxfb,{ 2, 2, 2, 2 } }));
             topology.add(reorder("input2_reorder", "input2", { data_type, format::yxfb,{ 2, 2, 2, 2 } }));
             topology.add(eltwise("eltwise", { "input_reorder", "input2_reorder" }, mode));
@@ -1747,7 +1744,7 @@ TEST(eltwise_gpu_int, basic_in4x4x4x4) {
 
             auto output = outputs.at("eltwise_reorder").get_memory();
 
-            auto output_ptr = output.pointer<float>();
+            cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
             for (int i = 0; i < 16; i++)
             {
@@ -1786,13 +1783,13 @@ TEST(eltwise_gpu_f32_int, basic_in4x4x4x4) {
     {
         for (auto& mode : eltwise_ops_to_test)
         {
-            const auto& engine = get_test_engine();
-            auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-            auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+            auto& engine = get_test_engine();
+            auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+            auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
             topology topology;
-            topology.add(input_layout("input", input.get_layout()));
-            topology.add(input_layout("input2", input2.get_layout()));
+            topology.add(input_layout("input", input->get_layout()));
+            topology.add(input_layout("input2", input2->get_layout()));
             topology.add(reorder("input_reorder", "input", { data_type, format::yxfb,{ 2, 2, 2, 2 } }));
             topology.add(eltwise("eltwise", { "input_reorder", "input2" }, mode));
             topology.add(reorder("eltwise_reorder", "eltwise", { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } }));
@@ -1822,7 +1819,7 @@ TEST(eltwise_gpu_f32_int, basic_in4x4x4x4) {
 
             auto output = outputs.at("eltwise_reorder").get_memory();
 
-            auto output_ptr = output.pointer<float>();
+            cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
             for (int i = 0; i < 16; i++)
             {
@@ -1872,12 +1869,12 @@ TEST(eltwise_gpu_f32, prod_basic_in4x4x4x4) {
     //  f1: b0:   119  80    b1:  96   -18.75
     //
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::prod));
 
     set_values(input, {
@@ -1910,7 +1907,7 @@ TEST(eltwise_gpu_f32, prod_basic_in4x4x4x4) {
         7.5f,   3.5f,   119.f,   96.0f,
        10.0f,  -2.0f,    80.f, -18.75f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1942,16 +1939,16 @@ TEST(eltwise_gpu_f32, max_basic_in4x4x4x4_input_padding) {
     //  f1: b0:    5   6    b1:  2.5   5.2
     //  f1: b0:    8   8    b1:  12    8
     //
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(reorder("reorder", "input", input.get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
-    topology.add(reorder("reorder2", "input2", input.get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(reorder("reorder", "input", input->get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
+    topology.add(reorder("reorder2", "input2", input->get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
     topology.add(eltwise("eltwise", {"reorder", "reorder2"}, eltwise_mode::max));
 
     set_values(input, {
@@ -1984,7 +1981,7 @@ TEST(eltwise_gpu_f32, max_basic_in4x4x4x4_input_padding) {
         15.f,  17.f,   8.f,  12.f,
         6.f,   8.f,   8.f,   8.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -2016,14 +2013,14 @@ TEST(eltwise_gpu_f32, add_basic_in4x4x2x2_with_coefficients) {
     //  f1: b0:   7.5  8.25  b1:  11    8.25
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2"}, eltwise_mode::sum, {0.5f, 0.5f}, data_types::f32));
 
     set_values(input, {
@@ -2055,7 +2052,7 @@ TEST(eltwise_gpu_f32, add_basic_in4x4x2x2_with_coefficients) {
                           9.f, 8.75f,  7.5f, 11.f,
                           1.f,   3.f, 3.75f, 2.75f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -2064,16 +2061,16 @@ TEST(eltwise_gpu_f32, add_basic_in4x4x2x2_with_coefficients) {
 }
 
 TEST(eltwise_gpu_f32, coefficients_count_check) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("input3", input3.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("input3", input3->get_layout()));
 
     std::vector<float> coeffs0 = {};
     std::vector<float> coeffs1 = {0.5f};
@@ -2124,16 +2121,16 @@ TEST(eltwise_gpu_f32, add_basic_in4x4x2x2_with_coefficients_3inputs) {
     //  f1: b0:   8.5   8.75   b1:  11    8.75
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("input3", input3.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("input3", input3->get_layout()));
     topology.add(eltwise("eltwise", {"input", "input2", "input3"}, eltwise_mode::sum, {0.5f, 0.5f, 0.5f}, data_types::f32));
 
     set_values(input, {
@@ -2172,7 +2169,7 @@ TEST(eltwise_gpu_f32, add_basic_in4x4x2x2_with_coefficients_3inputs) {
                           12.f, 8.75f,  8.5f, 11.f,
                           3.5f,  3.5f, 4.25f, 3.25f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -2211,19 +2208,19 @@ TEST(eltwise_gpu_f32, max_3inputs_in4x4x4x4_input_padding) {
     //  f1: b0:    5   6    b1:  2.5   7
     //  f1: b0:    9   8    b1:  12    8
     //
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("input3", input3.get_layout()));
-    topology.add(reorder("reorder", "input", input.get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
-    topology.add(reorder("reorder2", "input2", input.get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
-    topology.add(reorder("reorder3", "input3", input.get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("input3", input3->get_layout()));
+    topology.add(reorder("reorder", "input", input->get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
+    topology.add(reorder("reorder2", "input2", input->get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
+    topology.add(reorder("reorder3", "input3", input->get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
     topology.add(eltwise("eltwise", {"reorder", "reorder2", "reorder3"}, eltwise_mode::max));
 
     set_values(input, {
@@ -2263,7 +2260,7 @@ TEST(eltwise_gpu_f32, max_3inputs_in4x4x4x4_input_padding) {
         15.f,  17.f,   9.f,  12.f,
         6.f,   8.f,   8.f,   8.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -2300,14 +2297,14 @@ TEST(eltwise_gpu_f32, stride_test_2x2) {
     //  f1: b0: 33 35   b1:  49  51
     //  f1: b0: 41 43   b1:  57  59
     //
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 4, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 4, 4 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", "input", "input2", { {0,0,1,1}, {0,0,2,2} }, eltwise_mode::max));
 
     set_values(input, {
@@ -2352,7 +2349,7 @@ TEST(eltwise_gpu_f32, stride_test_2x2) {
         9,  25,  41,  57,
         11, 27,  43,  59 };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -2382,14 +2379,14 @@ TEST(eltwise_gpu_f32, broadcast_test_in4x4x2x2) {
     //  f1: b0:   15  16.5  b1:  22    16.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 1, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", { "input", "input2" }, eltwise_mode::sum));
 
     set_values(input, {
@@ -2420,7 +2417,7 @@ TEST(eltwise_gpu_f32, broadcast_test_in4x4x2x2) {
         3.5f,    3.f,   7.5f,  14.5f,
         4.5f,    2.f,   8.5f,  10.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -2452,23 +2449,23 @@ TEST(eltwise_gpu_f16, fs_b_yx_fsv32_basic)
         FLOAT16(2211),FLOAT16(2212),FLOAT16(2221),FLOAT16(2222)
     };
 
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
         return;
     }
 
-    auto input1 = memory::allocate(engine, { data_types::f16,format::bfyx, input_tensor });
-    auto input2 = memory::allocate(engine, { data_types::f16,format::bfyx, input_tensor });
+    auto input1 = engine.allocate_memory({ data_types::f16,format::bfyx, input_tensor });
+    auto input2 = engine.allocate_memory({ data_types::f16,format::bfyx, input_tensor });
 
     set_values(input1, fp16_bfyx_2x2x2x2_input);
     set_values(input2, fp16_bfyx_2x2x2x2_input);
 
     // GOLDEN BFYX ELTWISE
     topology golden_topology;
-    golden_topology.add(input_layout("input1", input1.get_layout()));
-    golden_topology.add(input_layout("input2", input2.get_layout()));
+    golden_topology.add(input_layout("input1", input1->get_layout()));
+    golden_topology.add(input_layout("input2", input2->get_layout()));
     golden_topology.add(eltwise("eltwise", "input1", "input2", eltwise_mode::sum));
 
     network golden_network(engine, golden_topology);
@@ -2477,12 +2474,12 @@ TEST(eltwise_gpu_f16, fs_b_yx_fsv32_basic)
 
     auto golden_outputs = golden_network.execute();
     auto golden_output = golden_outputs.at("eltwise").get_memory();
-    auto golden_ptr = golden_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> golden_ptr(golden_output, get_test_stream());
     // GOLDEN BFYX ELTWISE - END
     // FS_B_YX_FSV32 ELTWISE
     topology FSV32_topology;
-    FSV32_topology.add(input_layout("input1", input1.get_layout()));
-    FSV32_topology.add(input_layout("input2", input2.get_layout()));
+    FSV32_topology.add(input_layout("input1", input1->get_layout()));
+    FSV32_topology.add(input_layout("input2", input2->get_layout()));
     FSV32_topology.add(reorder("reorder1", "input1", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor)));
     FSV32_topology.add(reorder("reorder2", "input2", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor)));
     FSV32_topology.add(eltwise("eltwise", "reorder1", "reorder2", eltwise_mode::sum));
@@ -2494,13 +2491,12 @@ TEST(eltwise_gpu_f16, fs_b_yx_fsv32_basic)
 
     auto FSV32_outputs = FSV32_network.execute();
     auto FSV32_output = FSV32_outputs.at("reorderOutput").get_memory();
-    auto FSV32_ptr = FSV32_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> FSV32_ptr(FSV32_output, get_test_stream());
     // FS_B_YX_FSV32 ELTWISE - END
 
     ASSERT_EQ(golden_ptr.size(), FSV32_ptr.size());
 
-    for (size_t i = 0; i < golden_ptr.size(); i++)
-    {
+    for (size_t i = 0; i < golden_ptr.size(); i++) {
         EXPECT_EQ(float(golden_ptr[i]), float(FSV32_ptr[i]));
     }
 }
@@ -2510,14 +2506,14 @@ TEST(eltwise_gpu_f32, broadcast_test_in4x4x2x2x2) {
     //  Input2 : 2x2x1x1x2
     //  Output : 2x2x2x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 2, 2, 2, 1 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 2, 1, 1, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 2, 2, 2, 1 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 2, 1, 1, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(eltwise("eltwise", { "input", "input2" }, eltwise_mode::sum));
 
     set_values(input, {
@@ -2551,7 +2547,7 @@ TEST(eltwise_gpu_f32, broadcast_test_in4x4x2x2x2) {
         8.f,    3.5f,   12.f,   12.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 32; i++)
     {
@@ -2567,23 +2563,23 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_basic)
     VVVVF<FLOAT16> input_rnd = generate_random_4d<FLOAT16>(32, 96, 20, 20, 1, 3);
     VF<FLOAT16> fp16_bfyx_32x96x2x2_input = flatten_4d<FLOAT16>(format::bfyx, input_rnd);
 
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
         return;
     }
 
-    auto input1 = memory::allocate(engine, { data_types::f16,format::bfyx, input_tensor });
-    auto input2 = memory::allocate(engine, { data_types::f16,format::bfyx, input_tensor });
+    auto input1 = engine.allocate_memory({ data_types::f16,format::bfyx, input_tensor });
+    auto input2 = engine.allocate_memory({ data_types::f16,format::bfyx, input_tensor });
 
     set_values(input1, fp16_bfyx_32x96x2x2_input);
     set_values(input2, fp16_bfyx_32x96x2x2_input);
 
     // GOLDEN BFYX ELTWISE
     topology golden_topology;
-    golden_topology.add(input_layout("input1", input1.get_layout()));
-    golden_topology.add(input_layout("input2", input2.get_layout()));
+    golden_topology.add(input_layout("input1", input1->get_layout()));
+    golden_topology.add(input_layout("input2", input2->get_layout()));
     golden_topology.add(eltwise("eltwise", "input1", "input2", eltwise_mode::sum));
 
     network golden_network(engine, golden_topology);
@@ -2592,12 +2588,12 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_basic)
 
     auto golden_outputs = golden_network.execute();
     auto golden_output = golden_outputs.at("eltwise").get_memory();
-    auto golden_ptr = golden_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> golden_ptr(golden_output, get_test_stream());
     // GOLDEN BFYX ELTWISE - END
     // MIXED INPUT, FS_B_YX_FSV32 OUTPUT
     topology FS_B_YX_FSV32_OUTPUT_topology;
-    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input1", input1.get_layout()));
-    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input2", input2.get_layout()));
+    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input1", input1->get_layout()));
+    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input2", input2->get_layout()));
     FS_B_YX_FSV32_OUTPUT_topology.add(reorder("reorder1", "input1", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor)));
     FS_B_YX_FSV32_OUTPUT_topology.add(reorder("reorder2", "input2", layout(data_types::f16, format::byxf, input_tensor)));
     FS_B_YX_FSV32_OUTPUT_topology.add(eltwise("eltwise", "reorder1", "reorder2", eltwise_mode::sum));
@@ -2609,12 +2605,12 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_basic)
 
     auto FS_B_YX_FSV32_OUTPUT_outputs = FS_B_YX_FSV32_OUTPUT_network.execute();
     auto FS_B_YX_FSV32_OUTPUT_output = FS_B_YX_FSV32_OUTPUT_outputs.at("reorderOutput").get_memory();
-    auto FS_B_YX_FSV32_OUTPUT_ptr = FS_B_YX_FSV32_OUTPUT_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> FS_B_YX_FSV32_OUTPUT_ptr(FS_B_YX_FSV32_OUTPUT_output, get_test_stream());
     // MIXED INPUT, FS_B_YX_FSV32 OUTPUT - END
     // MIXED INPUT, BYXF OUTPUT
     topology BYXF_OUTPUT_topology;
-    BYXF_OUTPUT_topology.add(input_layout("input1", input1.get_layout()));
-    BYXF_OUTPUT_topology.add(input_layout("input2", input2.get_layout()));
+    BYXF_OUTPUT_topology.add(input_layout("input1", input1->get_layout()));
+    BYXF_OUTPUT_topology.add(input_layout("input2", input2->get_layout()));
     BYXF_OUTPUT_topology.add(reorder("reorder1", "input1", layout(data_types::f16, format::byxf, input_tensor)));
     BYXF_OUTPUT_topology.add(reorder("reorder2", "input2", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor)));
     BYXF_OUTPUT_topology.add(eltwise("eltwise", "reorder1", "reorder2", eltwise_mode::sum));
@@ -2626,47 +2622,44 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_basic)
 
     auto BYXF_OUTPUT_outputs = BYXF_OUTPUT_network.execute();
     auto BYXF_OUTPUT_output = BYXF_OUTPUT_outputs.at("reorderOutput").get_memory();
-    auto BYXF_OUTPUT_ptr = BYXF_OUTPUT_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> BYXF_OUTPUT_ptr(BYXF_OUTPUT_output, get_test_stream());
     // MIXED INPUT, BYXF OUTPUT - END
 
     ASSERT_EQ(golden_ptr.size(), FS_B_YX_FSV32_OUTPUT_ptr.size());
     ASSERT_EQ(golden_ptr.size(), BYXF_OUTPUT_ptr.size());
 
-    for (size_t i = 0; i < golden_ptr.size(); i++)
-    {
+    for (size_t i = 0; i < golden_ptr.size(); i++) {
         EXPECT_EQ(float(golden_ptr[i]), float(FS_B_YX_FSV32_OUTPUT_ptr[i]));
     }
-    for (size_t i = 0; i < golden_ptr.size(); i++)
-    {
+    for (size_t i = 0; i < golden_ptr.size(); i++) {
         EXPECT_EQ(float(golden_ptr[i]), float(BYXF_OUTPUT_ptr[i]));
     }
 }
 
-TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_output_padding)
-{
+TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_output_padding) {
     // Inputs are 32x96x2x2
 
     tensor input_tensor(32, 96, 20, 20);
     VVVVF<FLOAT16> input_rnd = generate_random_4d<FLOAT16>(32, 96, 20, 20, 1, 3);
     VF<FLOAT16> fp16_bfyx_32x96x2x2_input = flatten_4d<FLOAT16>(format::bfyx, input_rnd);
 
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
         return;
     }
 
-    auto input1 = memory::allocate(engine, { data_types::f16,format::bfyx, input_tensor });
-    auto input2 = memory::allocate(engine, { data_types::f16,format::bfyx, input_tensor });
+    auto input1 = engine.allocate_memory({ data_types::f16,format::bfyx, input_tensor });
+    auto input2 = engine.allocate_memory({ data_types::f16,format::bfyx, input_tensor });
 
     set_values(input1, fp16_bfyx_32x96x2x2_input);
     set_values(input2, fp16_bfyx_32x96x2x2_input);
 
     // GOLDEN BFYX ELTWISE
     topology golden_topology;
-    golden_topology.add(input_layout("input1", input1.get_layout()));
-    golden_topology.add(input_layout("input2", input2.get_layout()));
+    golden_topology.add(input_layout("input1", input1->get_layout()));
+    golden_topology.add(input_layout("input2", input2->get_layout()));
     golden_topology.add(eltwise("eltwise", "input1", "input2", eltwise_mode::sum, padding{ {0,0,5,10} , 0 }));
 
     network golden_network(engine, golden_topology);
@@ -2675,12 +2668,12 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_output_padding)
 
     auto golden_outputs = golden_network.execute();
     auto golden_output = golden_outputs.at("eltwise").get_memory();
-    auto golden_ptr = golden_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> golden_ptr(golden_output, get_test_stream());
     // GOLDEN BFYX ELTWISE - END
     // MIXED INPUT, FS_B_YX_FSV32 OUTPUT
     topology FS_B_YX_FSV32_OUTPUT_topology;
-    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input1", input1.get_layout()));
-    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input2", input2.get_layout()));
+    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input1", input1->get_layout()));
+    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input2", input2->get_layout()));
     FS_B_YX_FSV32_OUTPUT_topology.add(reorder("reorder1", "input1", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor)));
     FS_B_YX_FSV32_OUTPUT_topology.add(reorder("reorder2", "input2", layout(data_types::f16, format::byxf, input_tensor)));
     FS_B_YX_FSV32_OUTPUT_topology.add(eltwise("eltwise", "reorder1", "reorder2", eltwise_mode::sum, padding{ {0,0,5,10} , 0 }));
@@ -2693,12 +2686,12 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_output_padding)
 
     auto FS_B_YX_FSV32_OUTPUT_outputs = FS_B_YX_FSV32_OUTPUT_network.execute();
     auto FS_B_YX_FSV32_OUTPUT_output = FS_B_YX_FSV32_OUTPUT_outputs.at("reorderOutput").get_memory();
-    auto FS_B_YX_FSV32_OUTPUT_ptr = FS_B_YX_FSV32_OUTPUT_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> FS_B_YX_FSV32_OUTPUT_ptr(FS_B_YX_FSV32_OUTPUT_output, get_test_stream());
     // MIXED INPUT, FS_B_YX_FSV32 OUTPUT - END
     // MIXED INPUT, BYXF OUTPUT
     topology BYXF_OUTPUT_topology;
-    BYXF_OUTPUT_topology.add(input_layout("input1", input1.get_layout()));
-    BYXF_OUTPUT_topology.add(input_layout("input2", input2.get_layout()));
+    BYXF_OUTPUT_topology.add(input_layout("input1", input1->get_layout()));
+    BYXF_OUTPUT_topology.add(input_layout("input2", input2->get_layout()));
     BYXF_OUTPUT_topology.add(reorder("reorder1", "input1", layout(data_types::f16, format::byxf, input_tensor)));
     BYXF_OUTPUT_topology.add(reorder("reorder2", "input2", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor)));
     BYXF_OUTPUT_topology.add(eltwise("eltwise", "reorder1", "reorder2", eltwise_mode::sum, padding{ {0,0,5,10} , 0 }));
@@ -2711,18 +2704,16 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_output_padding)
 
     auto BYXF_OUTPUT_outputs = BYXF_OUTPUT_network.execute();
     auto BYXF_OUTPUT_output = BYXF_OUTPUT_outputs.at("reorderOutput").get_memory();
-    auto BYXF_OUTPUT_ptr = BYXF_OUTPUT_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> BYXF_OUTPUT_ptr(BYXF_OUTPUT_output, get_test_stream());
     // MIXED INPUT, BYXF OUTPUT - END
 
     ASSERT_EQ(golden_ptr.size(), FS_B_YX_FSV32_OUTPUT_ptr.size());
     ASSERT_EQ(golden_ptr.size(), BYXF_OUTPUT_ptr.size());
 
-    for (size_t i = 0; i < golden_ptr.size(); i++)
-    {
+    for (size_t i = 0; i < golden_ptr.size(); i++) {
         EXPECT_EQ(float(golden_ptr[i]), float(FS_B_YX_FSV32_OUTPUT_ptr[i]));
     }
-    for (size_t i = 0; i < golden_ptr.size(); i++)
-    {
+    for (size_t i = 0; i < golden_ptr.size(); i++) {
         EXPECT_EQ(float(golden_ptr[i]), float(BYXF_OUTPUT_ptr[i]));
     }
 }
@@ -2735,23 +2726,23 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_input_padding)
     VVVVF<FLOAT16> input_rnd = generate_random_4d<FLOAT16>(32, 96, 20, 20, 1, 3);
     VF<FLOAT16> fp16_bfyx_32x96x2x2_input = flatten_4d<FLOAT16>(format::bfyx, input_rnd);
 
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
         return;
     }
 
-    auto input1 = memory::allocate(engine, { data_types::f16,format::bfyx, input_tensor });
-    auto input2 = memory::allocate(engine, { data_types::f16,format::bfyx, input_tensor });
+    auto input1 = engine.allocate_memory({ data_types::f16,format::bfyx, input_tensor });
+    auto input2 = engine.allocate_memory({ data_types::f16,format::bfyx, input_tensor });
 
     set_values(input1, fp16_bfyx_32x96x2x2_input);
     set_values(input2, fp16_bfyx_32x96x2x2_input);
 
     // GOLDEN BFYX ELTWISE
     topology golden_topology;
-    golden_topology.add(input_layout("input1", input1.get_layout()));
-    golden_topology.add(input_layout("input2", input2.get_layout()));
+    golden_topology.add(input_layout("input1", input1->get_layout()));
+    golden_topology.add(input_layout("input2", input2->get_layout()));
     golden_topology.add(reorder("reorder1", "input1", layout(data_types::f16, format::bfyx, input_tensor, padding{ {0,0,10,15},0.0f })));
     golden_topology.add(reorder("reorder2", "input2", layout(data_types::f16, format::bfyx, input_tensor, padding{ {0,0,5,7},0.0f })));
     golden_topology.add(eltwise("eltwise", "input1", "input2", eltwise_mode::sum));
@@ -2762,12 +2753,12 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_input_padding)
 
     auto golden_outputs = golden_network.execute();
     auto golden_output = golden_outputs.at("eltwise").get_memory();
-    auto golden_ptr = golden_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> golden_ptr(golden_output, get_test_stream());
     // GOLDEN BFYX ELTWISE - END
     // MIXED INPUT, FS_B_YX_FSV32 OUTPUT
     topology FS_B_YX_FSV32_OUTPUT_topology;
-    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input1", input1.get_layout()));
-    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input2", input2.get_layout()));
+    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input1", input1->get_layout()));
+    FS_B_YX_FSV32_OUTPUT_topology.add(input_layout("input2", input2->get_layout()));
     FS_B_YX_FSV32_OUTPUT_topology.add(reorder("reorder1", "input1", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor, padding{ {0,0,10,15},0.0f })));
     FS_B_YX_FSV32_OUTPUT_topology.add(reorder("reorder2", "input2", layout(data_types::f16, format::byxf, input_tensor, padding{ {0,0,5,7},0.0f })));
     FS_B_YX_FSV32_OUTPUT_topology.add(eltwise("eltwise", "reorder1", "reorder2", eltwise_mode::sum));
@@ -2779,12 +2770,12 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_input_padding)
 
     auto FS_B_YX_FSV32_OUTPUT_outputs = FS_B_YX_FSV32_OUTPUT_network.execute();
     auto FS_B_YX_FSV32_OUTPUT_output = FS_B_YX_FSV32_OUTPUT_outputs.at("reorderOutput").get_memory();
-    auto FS_B_YX_FSV32_OUTPUT_ptr = FS_B_YX_FSV32_OUTPUT_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> FS_B_YX_FSV32_OUTPUT_ptr(FS_B_YX_FSV32_OUTPUT_output, get_test_stream());
     // MIXED INPUT, FS_B_YX_FSV32 OUTPUT - END
     // MIXED INPUT, BYXF OUTPUT
     topology BYXF_OUTPUT_topology;
-    BYXF_OUTPUT_topology.add(input_layout("input1", input1.get_layout()));
-    BYXF_OUTPUT_topology.add(input_layout("input2", input2.get_layout()));
+    BYXF_OUTPUT_topology.add(input_layout("input1", input1->get_layout()));
+    BYXF_OUTPUT_topology.add(input_layout("input2", input2->get_layout()));
     BYXF_OUTPUT_topology.add(reorder("reorder1", "input1", layout(data_types::f16, format::byxf, input_tensor, padding{ {0,0,10,15},0.0f })));
     BYXF_OUTPUT_topology.add(reorder("reorder2", "input2", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor, padding{ {0,0,5,7},0.0f })));
     BYXF_OUTPUT_topology.add(eltwise("eltwise", "reorder1", "reorder2", eltwise_mode::sum));
@@ -2796,7 +2787,7 @@ TEST(eltwise_gpu_f16, bfyx_and_fs_b_yx_fsv32_input_padding)
 
     auto BYXF_OUTPUT_outputs = BYXF_OUTPUT_network.execute();
     auto BYXF_OUTPUT_output = BYXF_OUTPUT_outputs.at("reorderOutput").get_memory();
-    auto BYXF_OUTPUT_ptr = BYXF_OUTPUT_output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> BYXF_OUTPUT_ptr(BYXF_OUTPUT_output, get_test_stream());
     // MIXED INPUT, BYXF OUTPUT - END
 
     ASSERT_EQ(golden_ptr.size(), FS_B_YX_FSV32_OUTPUT_ptr.size());
@@ -2874,17 +2865,17 @@ void generic_eltwise_bool_test(cldnn::format test_input_fmt, int input_b, int in
     VF<T> input1_rnd_vec = flatten_4d<T>(test_input_fmt, input1_rnd);
     VF<T> input2_rnd_vec = flatten_4d<T>(test_input_fmt, input2_rnd);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor( input_b, input_f, input_x, input_y );
-    auto input1 = memory::allocate(engine, { type_to_data_type<T>::value, test_input_fmt, input_tensor });
-    auto input2 = memory::allocate(engine, { type_to_data_type<T>::value, test_input_fmt, input_tensor });
+    auto input1 = engine.allocate_memory({ type_to_data_type<T>::value, test_input_fmt, input_tensor });
+    auto input2 = engine.allocate_memory({ type_to_data_type<T>::value, test_input_fmt, input_tensor });
     set_values(input1, input1_rnd_vec);
     set_values(input2, input2_rnd_vec);
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(reorder("reorder1", "input1", input1.get_layout().with_padding(padding{{ 0, 0, input_padding_x, input_padding_y }, 0 })));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(reorder("reorder1", "input1", input1->get_layout().with_padding(padding{{ 0, 0, input_padding_x, input_padding_y }, 0 })));
     topology.add(eltwise("eltwise", {"reorder1", "input2"}, mode, padding{ { 0, 0, output_padding_x, output_padding_y }, 0 }));
 
     network network(engine, topology);
@@ -2895,8 +2886,8 @@ void generic_eltwise_bool_test(cldnn::format test_input_fmt, int input_b, int in
     EXPECT_EQ(outputs.begin()->first, "eltwise");
 
     auto output_memory = outputs.at("eltwise").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int8_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int8_t> output_ptr(output_memory, get_test_stream());
 
     VVVVF<int8_t> output_cpu = eltwise_bool_reference<T>(input1_rnd, input2_rnd, mode, input_padding_y, input_padding_x, output_padding_y, output_padding_x);
     EXPECT_EQ(output_layout.format.value, test_input_fmt.value);
@@ -2978,8 +2969,8 @@ void run_eltwise_generic_test(cldnn::eltwise_mode mode)
     cldnn::format test_inputs_fmt = cldnn::format::bfyx;
     std::pair<int, int> input_size = { 227, 227 };
 
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
     }
@@ -3012,7 +3003,7 @@ TEST(eltwise_gpu, b_fs_yx_fsv4_wo_callib) {
     const int BATCH = 1;
     const int in_B = BATCH;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int in_F = 256;
 
@@ -3027,26 +3018,17 @@ TEST(eltwise_gpu, b_fs_yx_fsv4_wo_callib) {
 
     // Mem initialization
     // This is user data, no kernels here
-    auto input1 = memory::allocate(engine,
-                                   { data_types::i8,
-                                       format::bfyx,
-                                       { in_B, in_F, in_X, in_Y } });
+    auto input1 = engine.allocate_memory({ data_types::i8, format::bfyx, { in_B, in_F, in_X, in_Y } });
     std::vector<char> data_i1(DataGold);
     for (size_t i = 0; i < data_i1.size(); i++) data_i1[i] = data_i1[i] + 1;
     set_values(input1, std::move(data_i1));
 
-    auto input2 = memory::allocate(engine,
-                                  { data_types::i8,
-                                      format::bfyx,
-                                      { in_B, in_F, in_X, in_Y } });
+    auto input2 = engine.allocate_memory({ data_types::i8, format::bfyx, { in_B, in_F, in_X, in_Y } });
     std::vector<char> data_i2(DataGold);
     for (size_t i = 0; i < data_i2.size(); i++) data_i2[i] = data_i2[i] + 2;
     set_values(input2, std::move(data_i2));
 
-    auto input3 = memory::allocate(engine,
-                                  { data_types::i8,
-                                      format::bfyx,
-                                      { in_B, in_F, in_X, in_Y } });
+    auto input3 = engine.allocate_memory({ data_types::i8, format::bfyx, { in_B, in_F, in_X, in_Y } });
     std::vector<char> data_i3(DataGold);
     for (size_t i = 0; i < data_i3.size(); i++) data_i3[i] = data_i3[i] + 3;
     set_values(input3, std::move(data_i3));
@@ -3070,9 +3052,9 @@ TEST(eltwise_gpu, b_fs_yx_fsv4_wo_callib) {
             auto actv = activation("eltw_GOLD", eltw, activation_func::relu);
 
             // Create a topology
-            topology.add(input_layout("input1", input1.get_layout()),
-                         input_layout("input2", input2.get_layout()),
-                         input_layout("input3", input3.get_layout()),
+            topology.add(input_layout("input1", input1->get_layout()),
+                         input_layout("input2", input2->get_layout()),
+                         input_layout("input3", input3->get_layout()),
                          eltw, actv);
 
             // Network processing
@@ -3086,7 +3068,7 @@ TEST(eltwise_gpu, b_fs_yx_fsv4_wo_callib) {
             auto searchC = outputs.find("eltw_GOLD");
             EXPECT_NE(searchC, outputs.end());
             auto output = outputs.begin()->second.get_memory();
-            auto output_ptr = output.pointer<char>();
+            cldnn::mem_lock<char> output_ptr(output, get_test_stream());
             vGoldOutput.reserve(output_ptr.size());
             for (size_t i = 0; i < output_ptr.size(); i++)
                 vGoldOutput.push_back(output_ptr[i]);
@@ -3119,9 +3101,9 @@ TEST(eltwise_gpu, b_fs_yx_fsv4_wo_callib) {
                                   "reorder3_Swizzelled" },
                                 mode[i]);
             auto actv = activation("eltw_IMAD", eltw, activation_func::relu);
-            topology.add(input_layout("input1", input1.get_layout()),
-                         input_layout("input2", input2.get_layout()),
-                         input_layout("input3", input3.get_layout()),
+            topology.add(input_layout("input1", input1->get_layout()),
+                         input_layout("input2", input2->get_layout()),
+                         input_layout("input3", input3->get_layout()),
                          eltw, actv);
 
             // Back reordering (a-ka unswizzelling) output from MMAD/IMAD pooling
@@ -3142,7 +3124,7 @@ TEST(eltwise_gpu, b_fs_yx_fsv4_wo_callib) {
             auto searchC = outputs.find("reorder_UnSwizzelled");
             EXPECT_NE(searchC, outputs.end());
             auto output = outputs.begin()->second.get_memory();
-            auto output_ptr = output.pointer<char>();
+            cldnn::mem_lock<char> output_ptr(output, get_test_stream());
             vTestOutput.reserve(output_ptr.size());
             for (size_t i = 0; i < output_ptr.size(); i++)
                 vTestOutput.push_back(output_ptr[i]);
@@ -3162,8 +3144,8 @@ TEST(DISABLED_eltwise_gpu, generic_random) {
     VF<float> slopes = { 0.0f, -0.0f, -17.19f, 1028.8f, std::numeric_limits<float>::infinity(), -std::numeric_limits<float>::infinity() };
     std::vector<std::pair<int, int>> input_sizes = { { 100, 100 },{ 227, 227 },{ 400, 600 } };
 
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
     }
@@ -3207,21 +3189,21 @@ struct eltwise_same_input_test_params {
 struct eltwise_same_input_test : testing::TestWithParam<eltwise_same_input_test_params>
 {
     template <typename T>
-    void fill_random_typed(memory& mem, int min, int max, int k) {
-        auto size = mem.get_layout().size;
+    void fill_random_typed(memory::ptr mem, int min, int max, int k) {
+        auto size = mem->get_layout().size;
         size_t b = size.batch[0];
         size_t f = size.feature[0];
         size_t x = size.spatial[0];
         size_t y = size.spatial[1];
 
         auto data = generate_random_4d<T>(b, f, y, x, min, max, k);
-        auto ptr = mem.pointer<T>();
+        mem_lock<T> ptr{mem, get_test_stream()};
         for (size_t bi = 0; bi < b; ++bi) {
             for (size_t fi = 0; fi < f; ++fi) {
                 for (size_t yi = 0; yi < y; ++yi) {
                     for (size_t xi = 0; xi < x; ++xi) {
                         auto coords = tensor(batch(bi), feature(fi), spatial(xi, yi, 0, 0));
-                        auto offset = mem.get_layout().get_linear_offset(coords);
+                        auto offset = mem->get_layout().get_linear_offset(coords);
                         ptr[offset] = data[bi][fi][yi][xi];
                     }
                 }
@@ -3229,8 +3211,8 @@ struct eltwise_same_input_test : testing::TestWithParam<eltwise_same_input_test_
         }
     }
 
-    void fill_random(memory& mem) {
-        auto dt = mem.get_layout().data_type;
+    void fill_random(memory::ptr mem) {
+        auto dt = mem->get_layout().data_type;
         switch (dt) {
         case data_types::f32:
             fill_random_typed<float>(mem, -127, 127, 2);
@@ -3250,16 +3232,16 @@ struct eltwise_same_input_test : testing::TestWithParam<eltwise_same_input_test_
     }
 
     template <typename T>
-    bool compare_outputs(const memory& out_ref, const memory& input_ref) {
-        auto output_lay = out_ref.get_layout();
-        auto opt_output_lay = input_ref.get_layout();
+    bool compare_outputs(const memory::ptr out_ref, const memory::ptr input_ref) {
+        auto output_lay = out_ref->get_layout();
+        auto opt_output_lay = input_ref->get_layout();
 
         size_t b = output_lay.size.batch[0];
         size_t f = output_lay.size.feature[0];
         size_t x = output_lay.size.spatial[0];
         size_t y = output_lay.size.spatial[1];
-        auto ref_ptr = out_ref.pointer<T>();
-        auto input_ptr = input_ref.pointer<T>();
+        mem_lock<T> ref_ptr{out_ref, get_test_stream()};
+        mem_lock<T> input_ptr{input_ref, get_test_stream()};
         for (size_t bi = 0; bi < b; ++bi) {
             for (size_t fi = 0; fi < f; ++fi) {
                 for (size_t yi = 0; yi < y; ++yi) {
@@ -3282,22 +3264,22 @@ struct eltwise_same_input_test : testing::TestWithParam<eltwise_same_input_test_
     }
 
     void execute_same_input(const eltwise_same_input_test_params& params, bool check_result) {
-        auto eng = cldnn::engine();
+        auto& engine = get_test_engine();
 
         auto in_layout = layout(params.input_type, params.in_format, params.input_size);
-        auto input = memory::allocate(eng, in_layout);
+        auto input = engine.allocate_memory(in_layout);
         fill_random(input);
 
         cldnn::topology topo;
-        topo.add(input_layout("input1", input.get_layout()));
-        topo.add(input_layout("input2", input.get_layout()));
+        topo.add(input_layout("input1", input->get_layout()));
+        topo.add(input_layout("input2", input->get_layout()));
         auto prim = eltwise("eltwise", {"input1", "input2"}, eltwise_mode::sum);
         topo.add(prim);
 
         auto build_ops = build_options();
         build_ops.set_option(build_option::outputs({"eltwise"}));
 
-        auto net = network(eng, topo, build_ops);
+        auto net = network(engine, topo, build_ops);
         net.set_input_data("input1", input);
         net.set_input_data("input2", input);
 
@@ -3437,21 +3419,21 @@ TEST_P(eltwise_test, fsv16) {
     VF<float> input1_rnd_vec = flatten_6d<float>(format::bfwzyx, input1_rnd);
     VF<float> input2_rnd_vec = flatten_6d<float>(format::bfwzyx, input2_rnd);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto fmt_pln = input0_size.size() == 4 ? format::bfyx : format::bfzyx;
     auto fmt_fsv16 = input0_size.size() == 4 ? format::b_fs_yx_fsv16 : format::b_fs_zyx_fsv16;
 
     auto in0_size = tensor(fmt_pln, input0_size);
     auto in1_size = tensor(fmt_pln, input1_size);
 
-    auto input1 = memory::allocate(engine, { data_types::f32, fmt_pln, in0_size });
-    auto input2 = memory::allocate(engine, { data_types::f32, fmt_pln, in1_size });
+    auto input1 = engine.allocate_memory({ data_types::f32, fmt_pln, in0_size });
+    auto input2 = engine.allocate_memory({ data_types::f32, fmt_pln, in1_size });
     set_values(input1, input1_rnd_vec);
     set_values(input2, input2_rnd_vec);
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(reorder("reorder1", "input1", fmt_fsv16, dt));
     topology.add(reorder("reorder2", "input2", fmt_fsv16, dt));
     topology.add(eltwise("eltwise", {"reorder1", "reorder2"}, mode));
@@ -3469,7 +3451,7 @@ TEST_P(eltwise_test, fsv16) {
     EXPECT_EQ(outputs.begin()->first, out_id);
 
     auto output_memory = outputs.at(out_id).get_memory();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     VF<float> output_cpu_vec = eltwise_ref(input1_rnd, input2_rnd, in0_size, in1_size, mode);
     for (size_t i = 0; i < output_cpu_vec.size(); ++i) {
@@ -3549,15 +3531,15 @@ TEST_P(eltwise_test_6d, bfwzyx) {
     auto in0_size = tensor(format::bfwzyx, input0_size);
     auto in1_size = tensor(format::bfwzyx, input1_size);
 
-    const auto& engine = get_test_engine();
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfwzyx, in0_size });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfwzyx, in1_size });
+    auto& engine = get_test_engine();
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfwzyx, in0_size });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfwzyx, in1_size });
     set_values(input1, input1_rnd_vec);
     set_values(input2, input2_rnd_vec);
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(reorder("reorder1", "input1", format::bfwzyx, dt));
     topology.add(reorder("reorder2", "input2", format::bfwzyx, dt));
     topology.add(eltwise("eltwise", {"reorder1", "reorder2"}, mode));
@@ -3575,7 +3557,7 @@ TEST_P(eltwise_test_6d, bfwzyx) {
     EXPECT_EQ(outputs.begin()->first, out_id);
 
     auto output_memory = outputs.at(out_id).get_memory();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     VF<float> output_cpu_vec = eltwise_ref(input1_rnd, input2_rnd, in0_size, in1_size, mode);
     for (size_t i = 0; i < output_cpu_vec.size(); ++i) {
@@ -3628,21 +3610,21 @@ TEST_P(eltwise_test_mixed_precision, fsv16) {
     VF<float> input1_rnd_vec = flatten_6d<float>(format::bfwzyx, input1_rnd);
     VF<int> input2_rnd_vec = flatten_6d<int>(format::bfwzyx, input2_rnd);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto fmt_pln = input0_size.size() == 4 ? format::bfyx : format::bfzyx;
     auto fmt_fsv16 = input0_size.size() == 4 ? format::b_fs_yx_fsv16 : format::b_fs_zyx_fsv16;
 
     auto in0_size = tensor(fmt_pln, input0_size);
     auto in1_size = tensor(fmt_pln, input1_size);
 
-    auto input1 = memory::allocate(engine, { data_types::f32, fmt_pln, in0_size });
-    auto input2 = memory::allocate(engine, { data_types::i32, fmt_pln, in1_size });
+    auto input1 = engine.allocate_memory({ data_types::f32, fmt_pln, in0_size });
+    auto input2 = engine.allocate_memory({ data_types::i32, fmt_pln, in1_size });
     set_values(input1, input1_rnd_vec);
     set_values(input2, input2_rnd_vec);
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(reorder("reorder1", "input1", fmt_fsv16, input0_dt));
     topology.add(reorder("reorder2", "input2", fmt_fsv16, input1_dt));
     topology.add(eltwise("eltwise", {"reorder1", "reorder2"}, mode));
@@ -3660,7 +3642,7 @@ TEST_P(eltwise_test_mixed_precision, fsv16) {
     EXPECT_EQ(outputs.begin()->first, out_id);
 
     auto output_memory = outputs.at(out_id).get_memory();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     VF<float> output_cpu_vec = eltwise_ref(input1_rnd, input2_rnd, in0_size, in1_size, mode);
     for (size_t i = 0; i < output_cpu_vec.size(); ++i) {
@@ -3728,18 +3710,18 @@ TEST_P(eltwise_test_mixed_layout, mixed_layout) {
     VF<float> input1_rnd_vec = flatten_6d<float>(format::bfwzyx, input1_rnd);
     VF<float> input2_rnd_vec = flatten_6d<float>(format::bfwzyx, input2_rnd);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto in0_size = tensor(format::bfyx, input0_size);
     auto in1_size = tensor(format::bfyx, input1_size);
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, in0_size });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, in1_size });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, in0_size });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, in1_size });
     set_values(input1, input1_rnd_vec);
     set_values(input2, input2_rnd_vec);
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
     topology.add(reorder("reorder1", "input1", format0, data_types::f32));
     topology.add(reorder("reorder2", "input2", format1, data_types::f32));
     topology.add(eltwise("eltwise", {"reorder1", "reorder2"}, mode));
@@ -3757,7 +3739,7 @@ TEST_P(eltwise_test_mixed_layout, mixed_layout) {
     EXPECT_TRUE(network.get_primitive_info("eltwise").find(selected_kernel) != std::string::npos);
 
     auto output_memory = outputs.at(out_id).get_memory();
-    auto output_ptr = output_memory.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     VF<float> output_cpu_vec = eltwise_ref(input1_rnd, input2_rnd, in0_size, in1_size, mode);
     for (size_t i = 0; i < output_cpu_vec.size(); ++i) {
@@ -3794,21 +3776,21 @@ struct eltwise_random_test_params {
 struct eltwise_random_test : testing::TestWithParam<eltwise_random_test_params>
 {
     template <typename T>
-    void fill_random_typed(memory& mem, int min, int max, int k) {
-        auto size = mem.get_layout().size;
+    void fill_random_typed(memory::ptr mem, int min, int max, int k) {
+        auto size = mem->get_layout().size;
         size_t b = size.batch[0];
         size_t f = size.feature[0];
         size_t x = size.spatial[0];
         size_t y = size.spatial[1];
 
         auto data = generate_random_4d<T>(b, f, y, x, min, max, k);
-        auto ptr = mem.pointer<T>();
+        mem_lock<T> ptr{mem, get_test_stream()};
         for (size_t bi = 0; bi < b; ++bi) {
             for (size_t fi = 0; fi < f; ++fi) {
                 for (size_t yi = 0; yi < y; ++yi) {
                     for (size_t xi = 0; xi < x; ++xi) {
                         auto coords = tensor(batch(bi), feature(fi), spatial(xi, yi, 0, 0));
-                        auto offset = mem.get_layout().get_linear_offset(coords);
+                        auto offset = mem->get_layout().get_linear_offset(coords);
                         ptr[offset] = data[bi][fi][yi][xi];
                     }
                 }
@@ -3816,8 +3798,8 @@ struct eltwise_random_test : testing::TestWithParam<eltwise_random_test_params>
         }
     }
 
-    void fill_random(memory& mem) {
-        auto dt = mem.get_layout().data_type;
+    void fill_random(memory::ptr mem) {
+        auto dt = mem->get_layout().data_type;
         switch (dt) {
         case data_types::f32:
             fill_random_typed<float>(mem, -127, 127, 2);
@@ -3837,16 +3819,16 @@ struct eltwise_random_test : testing::TestWithParam<eltwise_random_test_params>
     }
 
     template <typename T>
-    bool compare_outputs(const memory& out_ref, const memory& out_opt) {
-        auto output_lay = out_ref.get_layout();
-        auto opt_output_lay = out_opt.get_layout();
+    bool compare_outputs(const memory::ptr out_ref, const memory::ptr out_opt) {
+        auto output_lay = out_ref->get_layout();
+        auto opt_output_lay = out_opt->get_layout();
 
         size_t b = output_lay.size.batch[0];
         size_t f = output_lay.size.feature[0];
         size_t x = output_lay.size.spatial[0];
         size_t y = output_lay.size.spatial[1];
-        auto ref_ptr = out_ref.pointer<T>();
-        auto opt_ptr = out_opt.pointer<T>();
+        mem_lock<T> ref_ptr{out_ref, get_test_stream()};
+        mem_lock<T> opt_ptr{out_opt, get_test_stream()};
         for (size_t bi = 0; bi < b; ++bi) {
             for (size_t fi = 0; fi < f; ++fi) {
                 for (size_t yi = 0; yi < y; ++yi) {
@@ -3869,18 +3851,18 @@ struct eltwise_random_test : testing::TestWithParam<eltwise_random_test_params>
     }
 
     void execute_compare(const eltwise_random_test_params& params, bool check_result) {
-        auto eng = cldnn::engine();
+        auto& engine = get_test_engine();
 
         auto in_layout1 = layout(params.input_type, params.in_format, params.first_input_size);
         auto in_layout2 = layout(params.input_type, params.in_format_second, params.second_input_size);
-        auto input1 = memory::allocate(eng, in_layout1);
-        auto input2 = memory::allocate(eng, in_layout2);
+        auto input1 = engine.allocate_memory(in_layout1);
+        auto input2 = engine.allocate_memory(in_layout2);
         fill_random(input1);
         fill_random(input2);
 
         cldnn::topology topo;
-        topo.add(input_layout("input1", input1.get_layout()));
-        topo.add(input_layout("input2", input2.get_layout()));
+        topo.add(input_layout("input1", input1->get_layout()));
+        topo.add(input_layout("input2", input2->get_layout()));
         auto prim = eltwise("eltwise", {"input1", "input2"}, params.mode);
         topo.add(prim);
 
@@ -3888,26 +3870,23 @@ struct eltwise_random_test : testing::TestWithParam<eltwise_random_test_params>
         build_ops.set_option(build_option::outputs({"eltwise"}));
         build_ops.set_option(build_option::force_implementations({ {"eltwise", {params.in_format, "generic_eltwise_ref"}} }));
 
-        auto net = network(eng, topo, build_ops);
+        auto net = network(engine, topo, build_ops);
         net.set_input_data("input1", input1);
         net.set_input_data("input2", input2);
 
         auto result = net.execute();
         auto output = result.at("eltwise").get_memory();
 
-        // Execute optimized eltwise 'eltwise_opt'
-        auto eng_opt = cldnn::engine();
-
         cldnn::topology topo_opt;
-        topo_opt.add(input_layout("input1", input1.get_layout()));
-        topo_opt.add(input_layout("input2", input2.get_layout()));
+        topo_opt.add(input_layout("input1", input1->get_layout()));
+        topo_opt.add(input_layout("input2", input2->get_layout()));
         auto prim_opt = eltwise("eltwise_opt", {"input1", "input2"}, params.mode);
         topo_opt.add(prim_opt);
 
         auto buildops_opt = build_options();
         buildops_opt.set_option(build_option::outputs({"eltwise_opt"}));
 
-        auto net_opt = network(eng_opt, topo_opt, buildops_opt);
+        auto net_opt = network(engine, topo_opt, buildops_opt);
         net_opt.set_input_data("input1", input1);
         net_opt.set_input_data("input2", input2);
 
@@ -3970,4 +3949,4 @@ INSTANTIATE_TEST_CASE_P(eltwise_smoke_fsv4,
                             .simple_params(data_types::f16, format::b_fs_yx_fsv4, format::b_fs_yx_fsv4)
                             .simple_params(data_types::i8, format::b_fs_yx_fsv4, format::b_fs_yx_fsv4)
                             .simple_params(data_types::u8, format::b_fs_yx_fsv4, format::b_fs_yx_fsv4)
-                        ), );
\ No newline at end of file
+                        ), );
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/embedding_bag_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/embedding_bag_gpu_test.cpp
index 531e290461dce5..b5be9527e6010d 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/embedding_bag_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/embedding_bag_gpu_test.cpp
@@ -3,17 +3,14 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/data.hpp>
-#include <api/embedding_bag.hpp>
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/network.hpp>
-#include <api/topology.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/embedding_bag.hpp>
+#include <cldnn/primitives/input_layout.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -24,11 +21,11 @@ TEST(embedding_bag_fp16_gpu, packed_sum_basic) {
     //  per_sample_weights : 3x2
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -51,8 +48,8 @@ TEST(embedding_bag_fp16_gpu, packed_sum_basic) {
 
     auto type = embedding_bag::packed_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
     topology.add(data("Input2", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2"}, type, output_shape)
@@ -66,7 +63,7 @@ TEST(embedding_bag_fp16_gpu, packed_sum_basic) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -1.05f, -1.2f,
@@ -84,10 +81,10 @@ TEST(embedding_bag_fp16_gpu, packed_sum_basic_without_weights) {
     //  indices : 3x2
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -105,8 +102,8 @@ TEST(embedding_bag_fp16_gpu, packed_sum_basic_without_weights) {
 
     auto type = embedding_bag::packed_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1"}, type, output_shape)
     );
@@ -119,7 +116,7 @@ TEST(embedding_bag_fp16_gpu, packed_sum_basic_without_weights) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -2.1f, -2.4f,
@@ -139,11 +136,11 @@ TEST(embedding_bag_fp16_gpu, packed_sum_dim2) {
     //  Output : 3x2x2
     //  Input values in fp16
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 2, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 2, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 1 } });
     tensor output_shape = {3, 2, 2, 1};
 
     /*
@@ -185,8 +182,8 @@ TEST(embedding_bag_fp16_gpu, packed_sum_dim2) {
 
     auto type = embedding_bag::packed_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
     topology.add(data("Input2", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2"}, type, output_shape)
@@ -200,7 +197,7 @@ TEST(embedding_bag_fp16_gpu, packed_sum_dim2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     /*
      * [ 3
@@ -232,11 +229,11 @@ TEST(embedding_bag_fp16_gpu, packed_sum_dim3) {
     //  per_sample_weights : 3x2
     //  Output : 3x2x3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 3, 2 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 3, 2 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 1 } });
     tensor output_shape = {3, 2, 3, 2};
 
     /*
@@ -308,8 +305,8 @@ TEST(embedding_bag_fp16_gpu, packed_sum_dim3) {
 
     auto type = embedding_bag::packed_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
     topology.add(data("Input2", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2"}, type, output_shape)
@@ -323,7 +320,7 @@ TEST(embedding_bag_fp16_gpu, packed_sum_dim3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     /*
      * [ 3
@@ -375,12 +372,12 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_basic) {
     //  default_index : 1x1
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto offsets = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto offsets = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -402,9 +399,9 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_basic) {
 
     auto type = embedding_bag::offsets_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", offsets.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", offsets->get_layout()));
     topology.add(data("Input3", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2", "Input3"}, type, output_shape, 0)
@@ -418,7 +415,7 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_basic) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -1.05f, -1.2f,
@@ -439,12 +436,12 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_basic_first_empty) {
     //  default_index : 1x1
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto offsets = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto offsets = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -466,9 +463,9 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_basic_first_empty) {
 
     auto type = embedding_bag::offsets_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", offsets.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", offsets->get_layout()));
     topology.add(data("Input3", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2", "Input3"}, type, output_shape, 2)
@@ -483,7 +480,7 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_basic_first_empty) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -1.9f, -1.8f,
@@ -504,12 +501,12 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_basic_last_empty) {
     //  default_index : 1x1
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto offsets = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto offsets = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -531,9 +528,9 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_basic_last_empty) {
 
     auto type = embedding_bag::offsets_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", offsets.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", offsets->get_layout()));
     topology.add(data("Input3", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2", "Input3"}, type, output_shape, 2)
@@ -548,7 +545,7 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_basic_last_empty) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -1.05f, -1.2f,
@@ -567,11 +564,11 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_without_weights_and_def_index) {
     //  offsets : 3x1
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto offsets = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto offsets = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -590,9 +587,9 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_without_weights_and_def_index) {
 
     auto type = embedding_bag::offsets_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", offsets.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", offsets->get_layout()));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2"}, type, output_shape)
     );
@@ -606,7 +603,7 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_without_weights_and_def_index) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -2.1f, -2.4f,
@@ -627,12 +624,12 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_dim3) {
     //  default_index : 1x1
     //  Output : 3x2x3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 3, 2 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto offsets = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 3, 2 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto offsets = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
     tensor output_shape = {3, 2, 3, 2};
 
     /*
@@ -704,9 +701,9 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_dim3) {
 
     auto type = embedding_bag::offsets_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", offsets.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", offsets->get_layout()));
     topology.add(data("Input3", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2", "Input3"}, type, output_shape, 0)
@@ -721,7 +718,7 @@ TEST(embedding_bag_fp16_gpu, offsets_sum_dim3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     /*
      * [ 3
@@ -773,12 +770,12 @@ TEST(embedding_bag_fp16_gpu, segments_sum_basic) {
     //  default_index : 1x1
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto segment_ids = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto segment_ids = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -800,9 +797,9 @@ TEST(embedding_bag_fp16_gpu, segments_sum_basic) {
 
     auto type = embedding_bag::segments_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", segment_ids.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", segment_ids->get_layout()));
     topology.add(data("Input3", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2", "Input3"}, type, output_shape, 0)
@@ -817,7 +814,7 @@ TEST(embedding_bag_fp16_gpu, segments_sum_basic) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -1.05f, -1.2f,
@@ -838,12 +835,12 @@ TEST(embedding_bag_fp16_gpu, segments_sum_basic_first_empty) {
     //  default_index : 1x1
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto segment_ids = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto segment_ids = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -865,9 +862,9 @@ TEST(embedding_bag_fp16_gpu, segments_sum_basic_first_empty) {
 
     auto type = embedding_bag::segments_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", segment_ids.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", segment_ids->get_layout()));
     topology.add(data("Input3", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2", "Input3"}, type, output_shape, 2)
@@ -882,7 +879,7 @@ TEST(embedding_bag_fp16_gpu, segments_sum_basic_first_empty) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -1.9f, -1.8f,
@@ -903,12 +900,12 @@ TEST(embedding_bag_fp16_gpu, segments_sum_basic_last_empty) {
     //  default_index : 1x1
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto segment_ids = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto segment_ids = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -930,9 +927,9 @@ TEST(embedding_bag_fp16_gpu, segments_sum_basic_last_empty) {
 
     auto type = embedding_bag::segments_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", segment_ids.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", segment_ids->get_layout()));
     topology.add(data("Input3", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2", "Input3"}, type, output_shape, 2)
@@ -947,7 +944,7 @@ TEST(embedding_bag_fp16_gpu, segments_sum_basic_last_empty) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -1.05f, -1.2f,
@@ -966,11 +963,11 @@ TEST(embedding_bag_fp16_gpu, segments_sum_without_weights_and_def_index) {
     //  segment_ids : 4x1
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto segment_ids = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto segment_ids = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -989,9 +986,9 @@ TEST(embedding_bag_fp16_gpu, segments_sum_without_weights_and_def_index) {
 
     auto type = embedding_bag::segments_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", segment_ids.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", segment_ids->get_layout()));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2"}, type, output_shape)
     );
@@ -1005,7 +1002,7 @@ TEST(embedding_bag_fp16_gpu, segments_sum_without_weights_and_def_index) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -2.1f, -2.4f,
@@ -1026,12 +1023,12 @@ TEST(embedding_bag_fp16_gpu, segments_sum_dim3) {
     //  default_index : 1x1
     //  Output : 3x2x3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f16, format::bfyx, { 5, 2, 3, 2 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto segment_ids = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f16, format::bfyx, { 5, 2, 3, 2 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto segment_ids = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } });
     tensor output_shape = {3, 2, 3, 2};
 
     /*
@@ -1103,9 +1100,9 @@ TEST(embedding_bag_fp16_gpu, segments_sum_dim3) {
 
     auto type = embedding_bag::segments_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", segment_ids.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", segment_ids->get_layout()));
     topology.add(data("Input3", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2", "Input3"}, type, output_shape, 0)
@@ -1120,7 +1117,7 @@ TEST(embedding_bag_fp16_gpu, segments_sum_dim3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     /*
      * [ 3
@@ -1170,11 +1167,11 @@ TEST(embedding_bag_fp32_gpu, packed_sum_basic) {
     //  per_sample_weights : 3x2
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f32, format::bfyx, { 5, 2, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 2, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f32, format::bfyx, { 5, 2, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 2, 1, 1 } });
     tensor output_shape = {3, 2, 1, 1};
 
     set_values(emb_table, {
@@ -1197,8 +1194,8 @@ TEST(embedding_bag_fp32_gpu, packed_sum_basic) {
 
     auto type = embedding_bag::packed_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
     topology.add(data("Input2", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2"}, type, output_shape)
@@ -1212,7 +1209,7 @@ TEST(embedding_bag_fp32_gpu, packed_sum_basic) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             -1.05f, -1.2f,
@@ -1231,11 +1228,11 @@ TEST(embedding_bag_fp32_gpu, packed_sum_dim3) {
     //  per_sample_weights : 3x2
     //  Output : 3x2x3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f32, format::bfyx, { 5, 2, 3, 2 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
-    auto per_sample_weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 2, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f32, format::bfyx, { 5, 2, 3, 2 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 2, 1, 1 } });
+    auto per_sample_weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 2, 1, 1 } });
     tensor output_shape = {3, 2, 3, 2};
 
     /*
@@ -1307,8 +1304,8 @@ TEST(embedding_bag_fp32_gpu, packed_sum_dim3) {
 
     auto type = embedding_bag::packed_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
     topology.add(data("Input2", per_sample_weights));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2"}, type, output_shape)
@@ -1322,7 +1319,7 @@ TEST(embedding_bag_fp32_gpu, packed_sum_dim3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     /*
      * [ 3
@@ -1372,11 +1369,11 @@ TEST(embedding_bag_fp32_gpu, extended5_6) {
     //  per_sample_weights : 3x2
     //  Output : 3x2
     //  Input values in fp16
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto emb_table = memory::allocate(engine, { data_types::f32, format::bfyx, { 5, 6, 1, 1 } });
-    auto indices = memory::allocate(engine, { data_types::i32, format::bfyx, { 5, 1, 1, 1 } });
-    auto segment_ids = memory::allocate(engine, { data_types::i32, format::bfyx, { 5, 1, 1, 1 } });
+    auto emb_table = engine.allocate_memory({ data_types::f32, format::bfyx, { 5, 6, 1, 1 } });
+    auto indices = engine.allocate_memory({ data_types::i32, format::bfyx, { 5, 1, 1, 1 } });
+    auto segment_ids = engine.allocate_memory({ data_types::i32, format::bfyx, { 5, 1, 1, 1 } });
     tensor output_shape = {5, 6, 1, 1};
 
     set_values(emb_table, {
@@ -1391,9 +1388,9 @@ TEST(embedding_bag_fp32_gpu, extended5_6) {
 
     auto type = embedding_bag::segments_sum;
     topology topology;
-    topology.add(input_layout("Input0", emb_table.get_layout()));
-    topology.add(input_layout("Input1", indices.get_layout()));
-    topology.add(input_layout("Input2", segment_ids.get_layout()));
+    topology.add(input_layout("Input0", emb_table->get_layout()));
+    topology.add(input_layout("Input1", indices->get_layout()));
+    topology.add(input_layout("Input2", segment_ids->get_layout()));
     topology.add(
             embedding_bag("embedding_bag", {"Input0", "Input1", "Input2"}, type, output_shape)
     );
@@ -1407,7 +1404,7 @@ TEST(embedding_bag_fp32_gpu, extended5_6) {
     auto outputs = network.execute();
 
     auto output = outputs.at("embedding_bag").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0, 8, 15,  15, 9,  7,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/extract_image_patches_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/extract_image_patches_gpu_test.cpp
index ccbe6f5a769390..c627b088d1011a 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/extract_image_patches_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/extract_image_patches_gpu_test.cpp
@@ -3,26 +3,23 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/memory.hpp>
-#include <api/input_layout.hpp>
-#include <api/extract_image_patches.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/data.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/extract_image_patches.hpp>
+#include <cldnn/primitives/data.hpp>
 
-#include <test_utils/test_utils.h>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 TEST(extract_image_patches_gpu, basic) {
     //  Input  : 1x1x10x10
     //  Output : 1x9x2x2
 
     tensor output_shape = {1, 9, 2, 2};
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto batch = 1;
     auto depth = 1;
     auto in_rows = 10;
@@ -32,7 +29,7 @@ TEST(extract_image_patches_gpu, basic) {
     std::vector<unsigned int> rates = {1, 1};
     std::string auto_pad = "valid";
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
 
     std::vector<float> inputVals(batch * depth * in_rows * in_cols);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -43,7 +40,7 @@ TEST(extract_image_patches_gpu, basic) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(extract_image_patches("extract_image_patches", "Input0", sizes, strides, rates, auto_pad, output_shape));
 
     network network(engine, topology);
@@ -54,7 +51,7 @@ TEST(extract_image_patches_gpu, basic) {
     EXPECT_EQ(outputs.begin()->first, "extract_image_patches");
 
     auto output = outputs.at("extract_image_patches").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          1,  6,
@@ -95,7 +92,7 @@ TEST(extract_image_patches_gpu, basic2) {
     //  Input  : 1x1x10x10
     //  Output : 1x16x1x1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto batch = 1;
     auto depth = 1;
     auto in_rows = 10;
@@ -106,7 +103,7 @@ TEST(extract_image_patches_gpu, basic2) {
     std::string auto_pad = "valid";
     tensor output_shape = {1, 16, 1, 1};
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
 
     std::vector<float> inputVals(batch * depth * in_rows * in_cols);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -117,7 +114,7 @@ TEST(extract_image_patches_gpu, basic2) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(extract_image_patches("extract_image_patches", "Input0", sizes, strides, rates, auto_pad, output_shape));
 
     network network(engine, topology);
@@ -128,7 +125,7 @@ TEST(extract_image_patches_gpu, basic2) {
     EXPECT_EQ(outputs.begin()->first, "extract_image_patches");
 
     auto output = outputs.at("extract_image_patches").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          1,
@@ -159,7 +156,7 @@ TEST(extract_image_patches_gpu, basic3) {
     //  Input  : 1x1x10x10
     //  Output : 1x16x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto batch = 1;
     auto depth = 1;
     auto in_rows = 10;
@@ -170,7 +167,7 @@ TEST(extract_image_patches_gpu, basic3) {
     std::string auto_pad = "same_upper";
     tensor output_shape = {1, 16, 2, 2};
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
 
     std::vector<float> inputVals(batch * depth * in_rows * in_cols);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -181,7 +178,7 @@ TEST(extract_image_patches_gpu, basic3) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(extract_image_patches("extract_image_patches", "Input0", sizes, strides, rates, auto_pad, output_shape));
 
     network network(engine, topology);
@@ -192,7 +189,7 @@ TEST(extract_image_patches_gpu, basic3) {
     EXPECT_EQ(outputs.begin()->first, "extract_image_patches");
 
     auto output = outputs.at("extract_image_patches").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          0,   0,
@@ -254,7 +251,7 @@ TEST(extract_image_patches_gpu, basic3_same_lower) {
     //  Input  : 1x1x10x10
     //  Output : 1x16x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto batch = 1;
     auto depth = 1;
     auto in_rows = 10;
@@ -265,7 +262,7 @@ TEST(extract_image_patches_gpu, basic3_same_lower) {
     std::string auto_pad = "same_lower";
     tensor output_shape = {1, 16, 2, 2};
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
 
     std::vector<float> inputVals(batch * depth * in_rows * in_cols);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -276,7 +273,7 @@ TEST(extract_image_patches_gpu, basic3_same_lower) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(extract_image_patches("extract_image_patches", "Input0", sizes, strides, rates, auto_pad, output_shape));
 
     network network(engine, topology);
@@ -287,7 +284,7 @@ TEST(extract_image_patches_gpu, basic3_same_lower) {
     EXPECT_EQ(outputs.begin()->first, "extract_image_patches");
 
     auto output = outputs.at("extract_image_patches").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          0,   0,
@@ -349,7 +346,7 @@ TEST(extract_image_patches_gpu, basic3_enough_space) {
     //  Input  : 1x1x10x10
     //  Output : 1x9x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto batch = 1;
     auto depth = 1;
     auto in_rows = 10;
@@ -360,7 +357,7 @@ TEST(extract_image_patches_gpu, basic3_enough_space) {
     std::string auto_pad = "same_upper";
     tensor output_shape = {1, 9, 2, 2};
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
 
     std::vector<float> inputVals(batch * depth * in_rows * in_cols);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -371,7 +368,7 @@ TEST(extract_image_patches_gpu, basic3_enough_space) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(extract_image_patches("extract_image_patches", "Input0", sizes, strides, rates, auto_pad, output_shape));
 
     network network(engine, topology);
@@ -382,7 +379,7 @@ TEST(extract_image_patches_gpu, basic3_enough_space) {
     EXPECT_EQ(outputs.begin()->first, "extract_image_patches");
 
     auto output = outputs.at("extract_image_patches").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          1,   8,
@@ -423,7 +420,7 @@ TEST(extract_image_patches_gpu, basic4) {
     //  Input  : 1x1x10x10
     //  Output : 1x9x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto batch = 1;
     auto depth = 1;
     auto in_rows = 10;
@@ -434,7 +431,7 @@ TEST(extract_image_patches_gpu, basic4) {
     std::string auto_pad = "valid";
     tensor output_shape = {1, 9, 2, 2};
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
 
     std::vector<float> inputVals(batch * depth * in_rows * in_cols);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -445,7 +442,7 @@ TEST(extract_image_patches_gpu, basic4) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(extract_image_patches("extract_image_patches", "Input0", sizes, strides, rates, auto_pad, output_shape));
 
     network network(engine, topology);
@@ -456,7 +453,7 @@ TEST(extract_image_patches_gpu, basic4) {
     EXPECT_EQ(outputs.begin()->first, "extract_image_patches");
 
     auto output = outputs.at("extract_image_patches").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          1,   6,
@@ -497,7 +494,7 @@ TEST(extract_image_patches_gpu, basic5) {
     //  Input  : 1x2x5x5
     //  Output : 1x8x2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto batch = 1;
     auto depth = 2;
     auto in_rows = 5;
@@ -508,7 +505,7 @@ TEST(extract_image_patches_gpu, basic5) {
     std::string auto_pad = "valid";
     tensor output_shape = {1, 8, 2, 2};
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { batch, depth, in_cols, in_rows } });
 
     std::vector<float> inputVals(batch * depth * in_rows * in_cols);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -519,7 +516,7 @@ TEST(extract_image_patches_gpu, basic5) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(extract_image_patches("extract_image_patches", "Input0", sizes, strides, rates, auto_pad, output_shape));
 
     network network(engine, topology);
@@ -530,7 +527,7 @@ TEST(extract_image_patches_gpu, basic5) {
     EXPECT_EQ(outputs.begin()->first, "extract_image_patches");
 
     auto output = outputs.at("extract_image_patches").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          1,  4,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/fully_connected_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/fully_connected_gpu_test.cpp
index 21f51050aa8185..b96ca63bb2e0e9 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/fully_connected_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/fully_connected_gpu_test.cpp
@@ -3,20 +3,14 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/fully_connected.hpp"
-#include <api/quantize.hpp>
-#include <api/topology.hpp>
-#include <api/tensor.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <src/include/to_string_utils.h>
-#include <api/data.hpp>
-#include "instrumentation.h"
-#include "test_utils/network_test.h"
+
+#include "test_utils.h"
+#include "network_test.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include "cldnn/primitives/fully_connected.hpp"
+#include <cldnn/primitives/quantize.hpp>
+#include <cldnn/primitives/data.hpp>
 
 #include <cmath>
 
@@ -26,7 +20,7 @@ namespace cldnn
 }
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 cldnn::format::type layout_4d(cldnn::format f) {
     switch (f.value) {
@@ -75,19 +69,19 @@ void generic_fully_connected_test(cldnn::format test_input_fmt, cldnn::format te
     VF<T> input_rnd_vec = flatten_4d<T>(test_input_fmt, input_rnd);
     VF<T> weights_rnd_vec = flatten_4d<T>(test_weights_fmt, weights_rnd);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(input_b, f, x, y);
     tensor weights_tensor(output_f, f, x, y);
-    auto input = memory::allocate(engine, { type_to_data_type<T>::value, test_input_fmt, input_tensor });
-    auto weights = memory::allocate(engine, { type_to_data_type<T>::value, test_weights_fmt, weights_tensor });
-    auto bias = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1,1,output_f,1 } });
+    auto input = engine.allocate_memory({ type_to_data_type<T>::value, test_input_fmt, input_tensor });
+    auto weights = engine.allocate_memory({ type_to_data_type<T>::value, test_weights_fmt, weights_tensor });
+    auto bias = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1,1,output_f,1 } });
     set_values(input, input_rnd_vec);
     set_values(weights, weights_rnd_vec);
     set_values(bias, bias_rnd_vec);
 
     primitive_id out_id = "fully_connected";
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         data("bias", bias),
         fully_connected(out_id, "input", "weights", "bias")
@@ -105,8 +99,8 @@ void generic_fully_connected_test(cldnn::format test_input_fmt, cldnn::format te
     EXPECT_EQ(outputs.begin()->first, out_id);
 
     auto output_memory = outputs.at(out_id).get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<T>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<T> output_ptr(output_memory, get_test_stream());
 
     //EXPECT_EQ(output_layout.format.value, test_input_fmt);
     tensor output_tensor = output_layout.size;
@@ -150,8 +144,8 @@ TEST(DISABLED_fully_connected_gpu, generic_random_short) {
     std::vector<std::pair<int, int>> input_sizes = { {28, 28}, {64, 64}, {100, 100}, {227, 227}, {1000, 1}, {1, 4096} };
     VF<int> outputs_x = { 5, 16 };
 
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = !!engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
     }
@@ -199,15 +193,15 @@ TEST(fully_connected_gpu, no_biases) {
     const int32_t input_x = 3, input_b = 1,  // size of whole input buffer
         weight_b = 4, weight_x = 3;  // size of whole weights buffer
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::yxfb,{ input_b, 1, input_x, 1} });
-    auto weights_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::yxfb,{ input_b, 1, input_x, 1} });
+    auto weights_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
 
     set_values(input_prim, { -0.5f, 2.0f, 0.5f });
     set_values(weights_prim, { 1.5f, 1.0f, 0.5f, -1.0f, 0.0f, 0.5f, 0.5f, -0.5f, -2.0f, -0.5f, 1.0f, 1.5f });
 
-    auto input = input_layout("input", input_prim.get_layout());
+    auto input = input_layout("input", input_prim->get_layout());
     auto w_data = data("weights", weights_prim);
     auto fc = fully_connected("full_con_prim", "input", "weights");
     topology topology;
@@ -224,7 +218,7 @@ TEST(fully_connected_gpu, no_biases) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(1.5f, output_ptr[0]);
     EXPECT_EQ(0.75f, output_ptr[1]);
@@ -256,15 +250,15 @@ TEST(fully_connected_gpu, no_biases_int8) {
     const int32_t input_x = 3, input_b = 1,  // size of whole input buffer
         weight_b = 4, weight_x = 3;  // size of whole weights buffer
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ input_b, 1, input_x, 1 } });
-    auto weights_prim = memory::allocate(engine, { data_types::i8,format::bfyx,{ weight_b, 1, weight_x, 1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ input_b, 1, input_x, 1 } });
+    auto weights_prim = engine.allocate_memory({ data_types::i8,format::bfyx,{ weight_b, 1, weight_x, 1 } });
 
     set_values(input_prim, { 8.4f, 2.3f, -4.49f });
     set_values<char>(weights_prim, { 2, 1, 0, -3, -2, 1, 0, -2, -4, -5, 10, 8 });
 
-    auto input = input_layout("input", input_prim.get_layout());
+    auto input = input_layout("input", input_prim->get_layout());
     auto w_data = data("weights", weights_prim);
     auto ri = reorder("reorder_to_int", "input", { data_types::i8,format::bfyx,{ input_b, 1, input_x, 1 } });
     auto fc = fully_connected("full_con_prim", "reorder_to_int", "weights");
@@ -284,7 +278,7 @@ TEST(fully_connected_gpu, no_biases_int8) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(18.0f, output_ptr[0]);
     EXPECT_EQ(-32.0f, output_ptr[1]);
@@ -317,18 +311,18 @@ TEST(fully_connected_gpu, xb_f32_batch_1) {
         input_x = 3, input_b = 1,  // size of whole input buffer
         weight_b = 4, weight_x = 3;  // size of whole weights buffer
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate( engine, { data_types::f32, format::yxfb, { input_b, 1, input_x, 1 } });
-    auto weights_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
-    auto bias_prim = memory::allocate(engine, { data_types::f32,format::bfyx, { 1,1,output_f, 1} });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::yxfb, { input_b, 1, input_x, 1 } });
+    auto weights_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
+    auto bias_prim = engine.allocate_memory({ data_types::f32,format::bfyx, { 1,1,output_f, 1} });
 
     set_values(input_prim, { -0.5f, 2.0f, 0.5f });
     set_values(weights_prim, { 1.5f, 1.0f, 0.5f, -1.0f, 0.0f, 0.5f, 0.5f, -0.5f, -2.0f, -0.5f, 1.0f, 1.5f });
     set_values(bias_prim, { 1.0f, 2.0f, 3.0f, 4.0f });
 
     topology topology(
-        input_layout("input", input_prim.get_layout()),
+        input_layout("input", input_prim->get_layout()),
         data("weights", weights_prim),
         data("bias", bias_prim),
         fully_connected("full_con_prim", "input", "weights", "bias")
@@ -343,7 +337,7 @@ TEST(fully_connected_gpu, xb_f32_batch_1) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.5f, output_ptr[0]);
     EXPECT_EQ(2.75f, output_ptr[1]);
@@ -377,18 +371,18 @@ TEST(fully_connected_gpu, xb_f32_batch_2) {
         input_x = 3, input_b = 2,  // size of whole input buffer
         weight_b = 4, weight_x = 3;  // size of whole weights buffer
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::yxfb,{ input_b,1,input_x, 1 } });
-    auto weights_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
-    auto bias_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1,1,output_f,1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::yxfb,{ input_b,1,input_x, 1 } });
+    auto weights_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
+    auto bias_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1,1,output_f,1 } });
 
     set_values(input_prim, { -0.5f, 1.0f, 2.0f, 1.5f, 0.5f, 0.0f });
     set_values(weights_prim, { 1.5f, 1.0f, 0.5f, -1.0f, 0.0f, 0.5f, 0.5f, -0.5f, -2.0f, -0.5f, 1.0f, 1.5f });
     set_values(bias_prim, { 1.0f, 2.0f, 3.0f, 4.0f });
 
     topology topology(
-        input_layout("input", input_prim.get_layout()),
+        input_layout("input", input_prim->get_layout()),
         data("weights", weights_prim),
         data("bias", bias_prim),
         fully_connected("full_con_prim", "input", "weights", "bias")
@@ -403,7 +397,7 @@ TEST(fully_connected_gpu, xb_f32_batch_2) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.50f, output_ptr[0]);
     EXPECT_EQ(4.00f, output_ptr[1]);
@@ -438,19 +432,19 @@ TEST(fully_connected_gpu, x_f32) {
         input_x = 3,                 // size of whole input buffer
         weight_b = 4, weight_x = 3;  // size of whole weights buffer
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::bfyx, { 1,1,input_x,1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::bfyx, { 1,1,input_x,1 } });
     //auto output_prim = memory::allocate({ memory::format::xb_f32,{ output_b,{ { output_f } },{ 1 } } });
-    auto weights_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
-    auto bias_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1,1,output_f,1 } });
+    auto weights_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
+    auto bias_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1,1,output_f,1 } });
 
     set_values(input_prim, { -0.5f, 2.0f, 0.5f });
     set_values(weights_prim, { 1.5f, 1.0f, 0.5f, -1.0f, 0.0f, 0.5f, 0.5f, -0.5f, -2.0f, -0.5f, 1.0f, 1.5f });
     set_values(bias_prim, { 1.0f, 2.0f, 3.0f, 4.0f });
 
     topology topology(
-        input_layout("input", input_prim.get_layout()),
+        input_layout("input", input_prim->get_layout()),
         data("weights", weights_prim),
         data("bias", bias_prim),
         fully_connected("full_con_prim", "input", "weights", "bias")
@@ -465,7 +459,7 @@ TEST(fully_connected_gpu, x_f32) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.50f, output_ptr[0]);
     EXPECT_EQ(2.75f, output_ptr[1]);
@@ -494,19 +488,19 @@ TEST(fully_connected_gpu, yxfn_f32) {
     //  Output:
     //   10  -28.5
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 2, 2, 1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 2, 2, 1 } });
     //auto output_prim = memory::allocate({ memory::format::xb_f32,{ 2 ,{ { 1 } }, 1 } });
-    auto weights_prim = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
-    auto bias_prim = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 1 } });
+    auto weights_prim = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
+    auto bias_prim = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 1 } });
 
     set_values(input_prim, { 1.f, 3.f, -2.f, -4.f });
     set_values(weights_prim, { 1.f, -1.f, 2.0f, 0.f, 3.0f, 4.0f, 0.5f, 5.0f });
     set_values(bias_prim, { 1.0f, -5.0f });
 
     topology topology(
-        input_layout("input", input_prim.get_layout()),
+        input_layout("input", input_prim->get_layout()),
         data("weights", weights_prim),
         data("bias", bias_prim),
         fully_connected("full_con_prim", "input", "weights", "bias")
@@ -521,7 +515,7 @@ TEST(fully_connected_gpu, yxfn_f32) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(10, output_ptr[0]);
     EXPECT_EQ(-28.5, output_ptr[1]);
@@ -552,19 +546,19 @@ TEST(fully_connected_gpu, xb_f32_batch_1_relu) {
         input_x = 3, input_b = 1,  // size of whole input buffer
         weight_b = 4, weight_x = 3;  // size of whole weights buffer
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::yxfb,{ input_b, 1, input_x, 1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::yxfb,{ input_b, 1, input_x, 1 } });
     //auto output_prim = memory::allocate({ memory::format::xb_f32,{ output_b,{ { output_f } },{ 1 } } });
-    auto weights_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
-    auto bias_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1,1,output_f, 1 } });
+    auto weights_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
+    auto bias_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1,1,output_f, 1 } });
 
     set_values(input_prim, { -0.5f, 2.0f, 0.5f });
     set_values(weights_prim, { 1.5f, 1.0f, 0.5f, -1.0f, 0.0f, 0.5f, 0.5f, -0.5f, -2.0f, -0.5f, 1.0f, 1.5f });
     set_values(bias_prim, { 1.0f, -2.0f, 3.0f, -4.0f });
 
     topology topology(
-        input_layout("input", input_prim.get_layout()),
+        input_layout("input", input_prim->get_layout()),
         data("weights", weights_prim),
         data("bias", bias_prim),
         fully_connected("full_con_prim", "input", "weights", "bias"),
@@ -580,7 +574,7 @@ TEST(fully_connected_gpu, xb_f32_batch_1_relu) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.50f, output_ptr[0]);
     EXPECT_EQ(0.00f, output_ptr[1]);
@@ -614,19 +608,19 @@ TEST(fully_connected_gpu, xb_f32_batch_2_relu) {
         input_x = 3, input_b = 2,  // size of whole input buffer
         weight_b = 4, weight_x = 3;  // size of whole weights buffer
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::yxfb,{ input_b, 1, input_x, 1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::yxfb,{ input_b, 1, input_x, 1 } });
     //auto output_prim = memory::allocate({ memory::format::xb_f32,{ output_b,{ { output_f } },{ 1 } } });
-    auto weights_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
-    auto bias_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1,1,output_f,1 } });
+    auto weights_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
+    auto bias_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1,1,output_f,1 } });
 
     set_values(input_prim, { -0.5f, 1.0f, 2.0f, 1.5f, 0.5f, 0.0f });
     set_values(weights_prim, { 1.5f, 1.0f, 0.5f, -1.0f, 0.0f, 0.5f, 0.5f, -0.5f, -2.0f, -0.5f, 1.0f, 1.5f });
     set_values(bias_prim, { 1.0f, -2.0f, 3.0f, -4.0f });
 
     topology topology(
-        input_layout("input", input_prim.get_layout()),
+        input_layout("input", input_prim->get_layout()),
         data("weights", weights_prim),
         data("bias", bias_prim),
         fully_connected("full_con_prim", "input", "weights", "bias"),
@@ -642,7 +636,7 @@ TEST(fully_connected_gpu, xb_f32_batch_2_relu) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.50f, output_ptr[0]);
     EXPECT_EQ(4.00f, output_ptr[1]);
@@ -677,19 +671,19 @@ TEST(fully_connected_gpu, x_f32_relu) {
         input_x = 3,                 // size of whole input buffer
         weight_b = 4, weight_x = 3;  // size of whole weights buffer
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1,1,input_x,1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1,1,input_x,1 } });
     //auto output_prim = memory::allocate({ memory::format::x_f32,{ 1       ,{ { output_f } }, 1 } });
-    auto weights_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
-    auto bias_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1,1,output_f,1 } });
+    auto weights_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
+    auto bias_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1,1,output_f,1 } });
 
     set_values(input_prim, { -0.5f, 2.0f, 0.5f });
     set_values(weights_prim, { 1.5f, 1.0f, 0.5f, -1.0f, 0.0f, 0.5f, 0.5f, -0.5f, -2.0f, -0.5f, 1.0f, 1.5f });
     set_values(bias_prim, { 1.0f, -2.0f, 3.0f, -4.0f });
 
     topology topology(
-        input_layout("input", input_prim.get_layout()),
+        input_layout("input", input_prim->get_layout()),
         data("weights", weights_prim),
         data("bias", bias_prim),
         fully_connected("full_con_prim", "input", "weights", "bias"),
@@ -705,7 +699,7 @@ TEST(fully_connected_gpu, x_f32_relu) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.50f, output_ptr[0]);
     EXPECT_EQ(0.00f, output_ptr[1]);
@@ -737,19 +731,19 @@ TEST(fully_connected_gpu, x_f32_relu_with_negative_slope) {
         input_x = 3,                 // size of whole input buffer
         weight_b = 4, weight_x = 3;  // size of whole weights buffer
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1,1,input_x,1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1,1,input_x,1 } });
     //auto output_prim = memory::allocate({ memory::format::x_f32,{ 1       ,{ { output_f } }, 1 } });
-    auto weights_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
-    auto bias_prim = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1,1,output_f,1 } });
+    auto weights_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ weight_b, 1, weight_x, 1 } });
+    auto bias_prim = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1,1,output_f,1 } });
 
     set_values(input_prim, { -0.5f, 2.0f, 0.5f });
     set_values(weights_prim, { 1.5f, 1.0f, 0.5f, -1.0f, 0.0f, 0.5f, 0.5f, -0.5f, -2.0f, -0.5f, 1.0f, 1.5f });
     set_values(bias_prim, { 1.0f, -2.0f, 3.0f, -4.0f });
 
     topology topology(
-        input_layout("input", input_prim.get_layout()),
+        input_layout("input", input_prim->get_layout()),
         data("weights", weights_prim),
         data("bias", bias_prim),
         fully_connected("full_con_prim", "input", "weights", "bias"),
@@ -765,7 +759,7 @@ TEST(fully_connected_gpu, x_f32_relu_with_negative_slope) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.50f, output_ptr[0]);
     EXPECT_EQ(-0.125f, output_ptr[1]);
@@ -775,7 +769,7 @@ TEST(fully_connected_gpu, x_f32_relu_with_negative_slope) {
 
 TEST(fully_connected_gpu, b_fs_yx_fsv4)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int in_B = 2;
     const int in_F = 2048;
@@ -791,11 +785,11 @@ TEST(fully_connected_gpu, b_fs_yx_fsv4)
     std::vector<char> Data(in_F * in_B); // in_X=in_Y=1
     int i = 0;
     std::generate(Data.begin(), Data.end(), [i]() mutable { return i++ % 9; });
-    auto input = memory::allocate(engine, {data_types::i8, format::bfyx, {in_B, in_F, in_X, in_Y}});
+    auto input = engine.allocate_memory({data_types::i8, format::bfyx, {in_B, in_F, in_X, in_Y}});
     set_values(input, std::move(Data));
 
     // Create a topology
-    topology topology(input_layout("input", input.get_layout()));
+    topology topology(input_layout("input", input->get_layout()));
 
     // Reorder
     topology.add(reorder("reorder_in",
@@ -809,15 +803,19 @@ TEST(fully_connected_gpu, b_fs_yx_fsv4)
         return i % 2 ? -(i++) / W_F - 1 : (i++) / W_F + 1;
     });
     auto weights_gold =
-        memory::allocate(engine, {data_types::i8, format::bfyx, {W_B, W_F, W_X, W_Y}});
+        engine.allocate_memory({data_types::i8, format::bfyx, {W_B, W_F, W_X, W_Y}});
     auto weights_imad =
-        memory::allocate(engine, {data_types::i8, format::bfyx, {W_B, W_F, W_X, W_Y}});
+        engine.allocate_memory({data_types::i8, format::bfyx, {W_B, W_F, W_X, W_Y}});
     set_values(weights_gold, Weights);
     set_values(weights_imad, std::move(Weights));
     topology.add(data("weights_gold", weights_gold), data("weights_imad", weights_imad));
 
-    auto bias_gold = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 1, in_F, 1}});
-    auto bias_imad = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 1, in_F, 1}});
+    auto bias_gold = engine.allocate_memory({data_types::f32, format::bfyx, {1, in_F, 1, 1}});
+    auto bias_imad = engine.allocate_memory({data_types::f32, format::bfyx, {1, in_F, 1, 1}});
+
+    std::vector<float> bias_data(in_F, 0);
+    set_values(bias_gold, bias_data);
+    set_values(bias_imad, bias_data);
 
     topology.add(data("bias_gold", bias_gold));
     topology.add(data("bias_imad", bias_imad));
@@ -830,10 +828,10 @@ TEST(fully_connected_gpu, b_fs_yx_fsv4)
     topology.add(fullc_gold, fullc_imad);
 
 
-    auto input_low_mem = memory::allocate(engine, { data_types::f32, format::bfyx, {1, W_B, 1, 1} });
-    auto input_high_mem = memory::allocate(engine, { data_types::f32, format::bfyx, {1, W_B, 1, 1} });
-    auto output_low_mem = memory::allocate(engine, { data_types::f32, format::bfyx, {1, 1, 1, 1} });
-    auto output_high_mem = memory::allocate(engine, { data_types::f32, format::bfyx, {1, 1, 1, 1} });
+    auto input_low_mem = engine.allocate_memory({ data_types::f32, format::bfyx, {1, W_B, 1, 1} });
+    auto input_high_mem = engine.allocate_memory({ data_types::f32, format::bfyx, {1, W_B, 1, 1} });
+    auto output_low_mem = engine.allocate_memory({ data_types::f32, format::bfyx, {1, 1, 1, 1} });
+    auto output_high_mem = engine.allocate_memory({ data_types::f32, format::bfyx, {1, 1, 1, 1} });
     set_values(input_low_mem,  generate_random_1d<float>(W_B, -200, 0));
     set_values(input_high_mem, generate_random_1d<float>(W_B, 1, 200));
     set_values(output_low_mem,  {-127.0f});
@@ -868,21 +866,19 @@ TEST(fully_connected_gpu, b_fs_yx_fsv4)
 
     ASSERT_NE(out_gold, outputs.end());
     ASSERT_NE(out_test, outputs.end());
-    auto gold_ptr = out_gold->second.get_memory().pointer<char>();
-    auto test_ptr = out_test->second.get_memory().pointer<char>();
+    cldnn::mem_lock<char> gold_ptr(out_gold->second.get_memory(), get_test_stream());
+    cldnn::mem_lock<char> test_ptr(out_test->second.get_memory(), get_test_stream());
 
     ASSERT_EQ(gold_ptr.size(), test_ptr.size());
-    for (size_t i = 0; i < gold_ptr.size(); i++)
-    {
+    for (size_t i = 0; i < gold_ptr.size(); i++) {
         ASSERT_EQ(gold_ptr[i], test_ptr[i]);
     }
 }
 
-TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b12)
-{
-    const auto& engine = get_test_engine();
+TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b12) {
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
+    if (!engine.get_device_info().supports_fp16)
     {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
@@ -896,9 +892,9 @@ TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b12)
     const int input_f = 64;
 
     // Allocate memory
-    auto input_prim = memory::allocate(engine, { data_types::f16, format::bfyx, { batch_num, input_f, input_y, input_x } });
-    auto weights_prim = memory::allocate(engine, { data_types::f16,format::bfyx,{ output_f, input_f, input_y, input_x } });
-    auto bias_prim = memory::allocate(engine, { data_types::f16,format::bfyx,{ 1, 1, output_f, 1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f16, format::bfyx, { batch_num, input_f, input_y, input_x } });
+    auto weights_prim = engine.allocate_memory({ data_types::f16,format::bfyx,{ output_f, input_f, input_y, input_x } });
+    auto bias_prim = engine.allocate_memory({ data_types::f16,format::bfyx,{ 1, 1, output_f, 1 } });
 
     // Generate random input data and set values
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_y, input_x, -1, 1);
@@ -917,7 +913,7 @@ TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b12)
 
     // Create topology to test
     topology topology(
-        input_layout("input", input_prim.get_layout()),
+        input_layout("input", input_prim->get_layout()),
         data("weights", weights_prim),
         data("bias", bias_prim),
         reorder("input_fsv", "input", {data_types::f16, format::fs_b_yx_fsv32, { batch_num, input_f, input_y, input_x } }),
@@ -935,7 +931,7 @@ TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b12)
     auto outputs = network.execute();
 
     auto output_prim = outputs.at("out").get_memory();
-    auto output_ptr = output_prim.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_ptr(output_prim, get_test_stream());
 
     for (size_t bi = 0; bi < batch_num; ++bi)
     {
@@ -956,9 +952,9 @@ TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b12)
 
 TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b34)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
+    if (!engine.get_device_info().supports_fp16)
     {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
@@ -972,9 +968,9 @@ TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b34)
     const int input_f = 64;
 
     // Allocate memory
-    auto input_prim = memory::allocate(engine, { data_types::f16, format::bfyx, { batch_num, input_f, input_y, input_x } });
-    auto weights_prim = memory::allocate(engine, { data_types::f16,format::bfyx,{ output_f, input_f, input_y, input_x } });
-    auto bias_prim = memory::allocate(engine, { data_types::f16,format::bfyx,{ 1, 1, output_f, 1 } });
+    auto input_prim = engine.allocate_memory({ data_types::f16, format::bfyx, { batch_num, input_f, input_y, input_x } });
+    auto weights_prim = engine.allocate_memory({ data_types::f16,format::bfyx,{ output_f, input_f, input_y, input_x } });
+    auto bias_prim = engine.allocate_memory({ data_types::f16,format::bfyx,{ 1, 1, output_f, 1 } });
 
     // Generate random input data and set values
     auto input_data = generate_random_4d<FLOAT16>(batch_num, input_f, input_y, input_x, -1, 1);
@@ -993,7 +989,7 @@ TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b34)
 
     // Create topology to test
     topology topology(
-        input_layout("input", input_prim.get_layout()),
+        input_layout("input", input_prim->get_layout()),
         data("weights", weights_prim),
         data("bias", bias_prim),
         reorder("input_fsv", "input", { data_types::f16, format::fs_b_yx_fsv32, { batch_num, input_f, input_y, input_x } }),
@@ -1011,7 +1007,7 @@ TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b34)
     auto outputs = network.execute();
 
     auto output_prim = outputs.at("out").get_memory();
-    auto output_ptr = output_prim.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_ptr(output_prim, get_test_stream());
 
     for (size_t bi = 0; bi < batch_num; ++bi)
     {
@@ -1054,7 +1050,7 @@ struct fully_connected_random_test : ::testing::TestWithParam<fully_connected_te
         auto weights_data = generate_smart_random_4d<WeightsT>(output_f, input_f, input_y, input_x);
         auto bias_data = generate_smart_random_2d<BiasT>(1, output_f);
 
-        auto eng = get_test_engine();
+        auto& eng = get_test_engine();
         auto net = network_test(eng);
         auto input = net.add_input_layout<InputT, 4>("input", input_format, std::move(input_data));
         auto weights = net.add_data<WeightsT, 4>("weights", format::oiyx, std::move(weights_data));
@@ -1152,7 +1148,7 @@ struct fully_connected_random_test_3d : ::testing::TestWithParam<fully_connected
         auto weights_data = generate_smart_random_4d<WeightsT>(output_y, input_y, 1, 1);
         auto bias_data = generate_smart_random_2d<BiasT>(1, output_y);
 
-        auto eng = get_test_engine();
+        auto& eng = get_test_engine();
         auto net = network_test(eng);
         auto input = net.add_input_layout<InputT, 4>("input", input_format, std::move(input_data));
         auto weights = net.add_data<WeightsT, 4>("weights", format::oiyx, std::move(weights_data));
@@ -1314,14 +1310,14 @@ class fully_connected_quantized_test : public ::testing::Test {
         auto input_size = tensor(TensorValue(batch_num()), TensorValue(input_f()), TensorValue(input_x()), TensorValue(input_y()));
         auto weights_size = tensor(TensorValue(output_f()), TensorValue(input_f()), TensorValue(input_x()), TensorValue(input_y()));
 
-        auto input_prim = memory::allocate(engine, { input_data_type(), _fmt, input_size });
-        auto weights_prim = memory::allocate(engine, { data_types::i8, format::bfyx, weights_size });
-        auto quantization_input_low = memory::allocate(engine, { data_types::f32, format::bfyx, tensor(feature(output_f())) });
-        auto quantization_input_high = memory::allocate(engine, { data_types::f32, format::bfyx, tensor(feature(output_f())) });
-        auto quantization_output_low = memory::allocate(engine, { data_types::f32, format::bfyx, tensor(feature(1)) });
-        auto quantization_output_high = memory::allocate(engine, { data_types::f32, format::bfyx, tensor(feature(1)) });
+        auto input_prim = engine.allocate_memory({ input_data_type(), _fmt, input_size });
+        auto weights_prim = engine.allocate_memory({ data_types::i8, format::bfyx, weights_size });
+        auto quantization_input_low = engine.allocate_memory({ data_types::f32, format::bfyx, tensor(feature(output_f())) });
+        auto quantization_input_high = engine.allocate_memory({ data_types::f32, format::bfyx, tensor(feature(output_f())) });
+        auto quantization_output_low = engine.allocate_memory({ data_types::f32, format::bfyx, tensor(feature(1)) });
+        auto quantization_output_high = engine.allocate_memory({ data_types::f32, format::bfyx, tensor(feature(1)) });
 
-        VF<InputT> input_flattened(input_prim.get_layout().get_linear_size());
+        VF<InputT> input_flattened(input_prim->get_layout().get_linear_size());
         for (size_t bi = 0; bi < batch_num(); ++bi)
             for (size_t fi = 0; fi < input_f(); ++fi)
                 for (size_t yi = 0; yi < input_y(); ++yi)
@@ -1338,14 +1334,14 @@ class fully_connected_quantized_test : public ::testing::Test {
         set_values(quantization_output_low, { _quantization.output_low });
         set_values(quantization_output_high, { _quantization.output_high });
 
-        auto bias_prim = memory::allocate(engine, { data_types::i32, format::bfyx, tensor(feature(output_f())) });
+        auto bias_prim = engine.allocate_memory({ data_types::i32, format::bfyx, tensor(feature(output_f())) });
         set_values(bias_prim, _bias);
 
         topology topo;
         topo.add(data("weights", weights_prim));
         topo.add(data("bias", bias_prim));
 
-        topo.add(input_layout("input", input_prim.get_layout()));
+        topo.add(input_layout("input", input_prim->get_layout()));
         auto fc_prim = fully_connected("fc", "input", "weights", "bias");
         fc_prim.output_data_type = type_to_data_type<OutputT>::value;
         topo.add(fc_prim);
@@ -1374,7 +1370,7 @@ class fully_connected_quantized_test : public ::testing::Test {
 
         auto output = net.execute();
         auto out_mem = output.at("output").get_memory();
-        auto out_ptr = out_mem.pointer<OutputT>();
+        cldnn::mem_lock<OutputT> out_ptr(out_mem, get_test_stream());
 
         for (size_t bi = 0; bi < batch_num(); ++bi) {
             for (size_t fi = 0; fi < output_f(); ++fi) {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/fused_conv_eltwise_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/fused_conv_eltwise_gpu_test.cpp
index 237c7aa12ebbe0..bcca8e4dd4fd7d 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/fused_conv_eltwise_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/fused_conv_eltwise_gpu_test.cpp
@@ -3,36 +3,31 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/convolution.hpp"
-#include "api/eltwise.hpp"
-#include "api/reorder.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/data.hpp>
-#include <api/depth_to_space.hpp>
-
-#include <api_extension/fused_conv_eltwise.hpp>
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/convolution.hpp>
+#include <cldnn/primitives/eltwise.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/depth_to_space.hpp>
+#include <cldnn/primitives/fused_conv_eltwise.hpp>
 
 #include <cassert>
 #include <cmath>
-#include <gmock/gmock.h>
 #include <limits>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 using namespace testing;
 
 TEST(fused_conv_eltwise, basic_0)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 4, 5 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 4, 5 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, {
         1.0f,  2.0f, -15.f,  3.0f, 4.0f, -15.f, 5.0f,  6.0f, -15.f, 7.0f,
@@ -40,7 +35,7 @@ TEST(fused_conv_eltwise, basic_0)
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         convolution("conv", "input", { "weights" }),
         eltwise("eltwise", "input", "conv", eltwise_mode::sum),
@@ -56,7 +51,7 @@ TEST(fused_conv_eltwise, basic_0)
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output = outputs.begin()->second.get_memory();
-    auto&& out_layout = output.get_layout();
+    auto&& out_layout = output->get_layout();
 
     EXPECT_EQ(out_layout.format, format::bfyx);
     EXPECT_EQ(out_layout.size.batch[0], 1);
@@ -67,11 +62,11 @@ TEST(fused_conv_eltwise, basic_0)
 
 TEST(fused_conv_eltwise, basic_image2d)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 4, 128, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 3, 256, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f16, format::bfyx, { 12, 4, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 4, 128, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 3, 256, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f16, format::bfyx, { 12, 4, 1, 1 } });
 
     auto input_data1 = generate_random_4d<FLOAT16>(1, 4, 2, 128, -1, 1);
     auto input_data1_bfyx = flatten_4d(format::bfyx, input_data1);
@@ -86,8 +81,8 @@ TEST(fused_conv_eltwise, basic_image2d)
     set_values(weights, weights_data_bfyx);
 
     topology topology_act(
-        input_layout("input", input.get_layout()),
-        input_layout("input2", input2.get_layout()),
+        input_layout("input", input->get_layout()),
+        input_layout("input2", input2->get_layout()),
         data("weights", weights),
         convolution("conv", "input", { "weights" }),
         depth_to_space("depth_to_space", "conv", 2, depth_to_space_mode::blocks_first),
@@ -105,11 +100,11 @@ TEST(fused_conv_eltwise, basic_image2d)
     EXPECT_EQ(outputs_act.begin()->first, "eltwise");
 
     auto output_act = outputs_act.begin()->second.get_memory();
-    auto out_act_ptr = output_act.pointer<uint8_t>();
+    cldnn::mem_lock<uint8_t> out_act_ptr(output_act, get_test_stream());
 
     topology topology_ref(
-        input_layout("input", input.get_layout()),
-        input_layout("input2", input2.get_layout()),
+        input_layout("input", input->get_layout()),
+        input_layout("input2", input2->get_layout()),
         data("weights", weights),
         convolution("conv", "input", { "weights" }),
         depth_to_space("depth_to_space", "conv", 2, depth_to_space_mode::blocks_first),
@@ -127,7 +122,7 @@ TEST(fused_conv_eltwise, basic_image2d)
     EXPECT_EQ(outputs_ref.begin()->first, "out");
 
     auto output_ref = outputs_ref.begin()->second.get_memory();
-    auto out_ref_ptr = output_ref.pointer<uint8_t>();
+    cldnn::mem_lock<uint8_t> out_ref_ptr(output_ref, get_test_stream());
 
     for (int i = 0;i < 3 * 256 * 4;i++) {
         EXPECT_EQ(out_act_ptr[i], out_ref_ptr[i]);
@@ -136,10 +131,10 @@ TEST(fused_conv_eltwise, basic_image2d)
 
 TEST(fused_conv_eltwise, dont_fuse_if_conv_elt_are_outputs)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 5 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 5 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, {
         1.0f,  2.0f, -15.f,  3.0f, 4.0f, -15.f, 5.0f,  6.0f, -15.f, 7.0f,
@@ -147,7 +142,7 @@ TEST(fused_conv_eltwise, dont_fuse_if_conv_elt_are_outputs)
         });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         convolution("conv", "input", { "weights" }),
         eltwise("out", "input", "conv", eltwise_mode::sum));
@@ -162,7 +157,7 @@ TEST(fused_conv_eltwise, dont_fuse_if_conv_elt_are_outputs)
     EXPECT_EQ(outputs.begin()->first, "out");
 
     auto output = outputs.begin()->second.get_memory();
-    auto&& out_layout = output.get_layout();
+    auto&& out_layout = output->get_layout();
 
     EXPECT_EQ(out_layout.format, format::bfyx);
     EXPECT_EQ(out_layout.size.batch[0], 1);
@@ -211,7 +206,7 @@ class FusedConvTest : public testing::Test
 
     void do_test(const fused_conv_eltwise& fused_prim)
     {
-        const auto& engine = get_test_engine();
+        auto& engine = get_test_engine();
 
         int n_features = static_cast<int>(biases_values.size());
 
@@ -220,19 +215,11 @@ class FusedConvTest : public testing::Test
         auto biases_shape = tensor(1, n_features, 1, 1);
         auto sum_input_shape = tensor(1, n_features, 2, 1);
 
-        auto input = memory::allocate(
-            engine,
-            {type_to_data_type<InputTy>::value, format::bfyx, input_shape});
-        auto weights = memory::allocate(
-            engine,
-            {type_to_data_type<WeightsTy>::value, format::bfyx, weights_shape});
+        auto input = engine.allocate_memory({type_to_data_type<InputTy>::value, format::bfyx, input_shape});
+        auto weights = engine.allocate_memory({type_to_data_type<WeightsTy>::value, format::bfyx, weights_shape});
 
-        auto biases = memory::allocate(
-            engine,
-            {type_to_data_type<BiasesTy>::value, format::bfyx, biases_shape});
-        auto sum_input = memory::allocate(
-            engine,
-            {type_to_data_type<InputTy>::value, format::bfyx, sum_input_shape});
+        auto biases = engine.allocate_memory({type_to_data_type<BiasesTy>::value, format::bfyx, biases_shape});
+        auto sum_input = engine.allocate_memory({type_to_data_type<InputTy>::value, format::bfyx, sum_input_shape});
 
         set_values(input, input_values);
         std::vector<WeightsTy> post_processed_weights_values(n_features
@@ -254,7 +241,7 @@ class FusedConvTest : public testing::Test
         set_values(biases, biases_values);
         set_values(sum_input, non_conv_input_values);
 
-        the_topology.add(input_layout("input", input.get_layout()));
+        the_topology.add(input_layout("input", input->get_layout()));
         the_topology.add(data("weights", weights));
         the_topology.add(data("biases", biases));
         the_topology.add(data("sum_input", sum_input));
@@ -269,8 +256,8 @@ class FusedConvTest : public testing::Test
         auto outputs = network.execute();
 
         auto output_memory = outputs.at("fused_conv").get_memory();
-        auto output_layout = output_memory.get_layout();
-        auto output_ptr = output_memory.pointer<OutputTy>();
+        auto output_layout = output_memory->get_layout();
+        cldnn::mem_lock<OutputTy> output_ptr(output_memory, get_test_stream());
         int y_size = output_layout.size.spatial[1];
         int x_size = output_layout.size.spatial[0];
         int f_size = output_layout.size.feature[0];
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/fusings_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/fusings_gpu_test.cpp
index 2b9ddcbd3f4417..a3a88b7b2e8c69 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/fusings_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/fusings_gpu_test.cpp
@@ -2,43 +2,35 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include "api/input_layout.hpp"
-#include "api/convolution.hpp"
-#include "api/quantize.hpp"
-#include "api/topology.hpp"
-#include "api/tensor.hpp"
-#include "api/network.hpp"
-#include "api/eltwise.hpp"
-#include "api/fully_connected.hpp"
-#include "api/gemm.hpp"
-#include "api/binary_convolution.hpp"
-#include "api/engine.hpp"
-#include "api/data.hpp"
-#include "api/resample.hpp"
-#include "api/mvn.hpp"
-#include "api/deconvolution.hpp"
-#include "api/permute.hpp"
-#include "api/gather.hpp"
-#include "api/gather_nd.hpp"
-#include "api/scatter_update.hpp"
-#include "api/scatter_nd_update.hpp"
-#include "api/scatter_elements_update.hpp"
-#include "api/depth_to_space.hpp"
-#include "api/space_to_depth.hpp"
-#include "api/batch_to_space.hpp"
-#include "api/space_to_batch.hpp"
-#include "api/reduce.hpp"
-
-
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/convolution.hpp>
+#include <cldnn/primitives/quantize.hpp>
+#include <cldnn/primitives/eltwise.hpp>
+#include <cldnn/primitives/fully_connected.hpp>
+#include <cldnn/primitives/gemm.hpp>
+#include <cldnn/primitives/binary_convolution.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/resample.hpp>
+#include <cldnn/primitives/mvn.hpp>
+#include <cldnn/primitives/deconvolution.hpp>
+#include <cldnn/primitives/permute.hpp>
+#include <cldnn/primitives/gather.hpp>
+#include <cldnn/primitives/gather_nd.hpp>
+#include <cldnn/primitives/scatter_update.hpp>
+#include <cldnn/primitives/scatter_nd_update.hpp>
+#include <cldnn/primitives/scatter_elements_update.hpp>
+#include <cldnn/primitives/depth_to_space.hpp>
+#include <cldnn/primitives/space_to_depth.hpp>
+#include <cldnn/primitives/batch_to_space.hpp>
+#include <cldnn/primitives/space_to_batch.hpp>
+#include <cldnn/primitives/reduce.hpp>
 
 #include <cmath>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 struct resample_test_params {
     tensor in_shape;
@@ -137,7 +129,7 @@ struct normalize_test_params {
 template<typename T>
 class BaseFusingTest : public ::testing::TestWithParam<T> {
 public:
-    cldnn::engine engine;
+    cldnn::engine& engine = get_test_engine();
     cldnn::topology topology_fused;
     cldnn::topology topology_non_fused;
     cldnn::build_options bo_fused;
@@ -162,7 +154,7 @@ class BaseFusingTest : public ::testing::TestWithParam<T> {
             for (auto& pi : net.get_primitives_info()) {
                 if (pi.type_id == "reorder") {
                     auto exec_prims = net.get_executed_primitives();
-                    auto it = std::find_if(exec_prims.begin(), exec_prims.end(), [&](const std::pair<primitive_id, event>& e) -> bool {
+                    auto it = std::find_if(exec_prims.begin(), exec_prims.end(), [&](const std::pair<primitive_id, event::ptr>& e) -> bool {
                         return e.first == pi.original_id;
                     });
                     // We count executed reorders only
@@ -194,23 +186,23 @@ class BaseFusingTest : public ::testing::TestWithParam<T> {
 
         auto output_not_fused_prim = outputs_ref.begin()->second.get_memory();
         auto output_fused_prim = outputs_fused.begin()->second.get_memory();
-        if (output_not_fused_prim.get_layout().data_type == data_types::f32) {
-            auto ref = output_not_fused_prim.pointer<float>();
-            auto output_ptr = output_fused_prim.pointer<float>();
-            for (size_t i = 0; i < output_fused_prim.get_layout().count(); i++) {
+        if (output_not_fused_prim->get_layout().data_type == data_types::f32) {
+            cldnn::mem_lock<float> ref(output_not_fused_prim, get_test_stream());
+            cldnn::mem_lock<float> output_ptr(output_fused_prim, get_test_stream());
+            for (size_t i = 0; i < output_fused_prim->get_layout().count(); i++) {
                 ASSERT_NEAR(ref[i], output_ptr[i], tolerance) << "i = " << i;
             }
         } else {
-            auto ref = output_not_fused_prim.pointer<int16_t>();
-            auto output_ptr = output_fused_prim.pointer<int16_t>();
-            for (size_t i = 0; i < output_fused_prim.get_layout().count(); i++) {
+            cldnn::mem_lock<int16_t> ref(output_not_fused_prim, get_test_stream());
+            cldnn::mem_lock<int16_t> output_ptr(output_fused_prim, get_test_stream());
+            for (size_t i = 0; i < output_fused_prim->get_layout().count(); i++) {
                 ASSERT_NEAR(float16_to_float32(ref[i]), float16_to_float32(output_ptr[i]), tolerance) << "i = " << i;
             }
         }
     }
 
-    cldnn::memory get_mem(cldnn::layout l) {
-        auto prim = memory::allocate(engine, l);
+    cldnn::memory::ptr get_mem(cldnn::layout l) {
+        auto prim = engine.allocate_memory(l);
         tensor s = l.size;
         if (l.data_type == data_types::bin) {
             VF<int32_t> rnd_vec = generate_random_1d<int32_t>(s.count() / 32, min_random, max_random);
@@ -229,8 +221,8 @@ class BaseFusingTest : public ::testing::TestWithParam<T> {
         return prim;
     }
 
-    cldnn::memory get_mem(cldnn::layout l, float fill_value) {
-        auto prim = memory::allocate(engine, l);
+    cldnn::memory::ptr get_mem(cldnn::layout l, float fill_value) {
+        auto prim = engine.allocate_memory(l);
         tensor s = l.size;
         if (l.data_type == data_types::bin) {
             VF<int32_t> rnd_vec(s.count() / 32, static_cast<int32_t>(fill_value));
@@ -238,16 +230,18 @@ class BaseFusingTest : public ::testing::TestWithParam<T> {
         } else if (l.data_type == data_types::f16) {
             VF<uint16_t> rnd_vec(s.count(), float32_to_float16(fill_value));
             set_values(prim, rnd_vec);
-        } else {
+        } else if (l.data_type == data_types::f32) {
             VF<float> rnd_vec(s.count(), fill_value);
             set_values(prim, rnd_vec);
+        } else {
+            throw std::runtime_error("get_mem: Unsupported precision");
         }
 
         return prim;
     }
 
-    cldnn::memory get_repeatless_mem(cldnn::layout l, int min, int max) {
-        auto prim = memory::allocate(engine, l);
+    cldnn::memory::ptr get_repeatless_mem(cldnn::layout l, int min, int max) {
+        auto prim = engine.allocate_memory(l);
         tensor s = l.size;
         if (l.data_type == data_types::f32) {
             VF<float> rnd_vec = generate_random_norepetitions_1d<float>(s.count(), min, max);
@@ -267,8 +261,8 @@ class BaseFusingTest : public ::testing::TestWithParam<T> {
         return prim;
     }
 
-    cldnn::memory get_mem(cldnn::layout l, int min, int max) {
-        auto prim = memory::allocate(engine, l);
+    cldnn::memory::ptr get_mem(cldnn::layout l, int min, int max) {
+        auto prim = engine.allocate_memory(l);
         tensor s = l.size;
         if (l.data_type == data_types::f32) {
             VF<float> rnd_vec = generate_random_1d<float>(s.count(), min, max);
@@ -292,7 +286,7 @@ class BaseFusingTest : public ::testing::TestWithParam<T> {
         return layout{ p.data_type, p.input_format, p.out_shape };
     }
 
-    layout get_weights_layout(T& p, const int32_t split = 1) {
+    layout get_weights_layout(T& p, const int32_t /* split */ = 1) {
         cldnn::tensor weights_tensor;
         if (p.groups == 1) {
             weights_tensor = cldnn::tensor(batch(p.out_shape.feature[0]), feature(p.in_shape.feature[0]),
@@ -304,7 +298,7 @@ class BaseFusingTest : public ::testing::TestWithParam<T> {
         return layout{p.weights_type, p.weights_format, weights_tensor};
     }
 
-    layout get_weights_layout(T& p, const int32_t split, cldnn::format f) {
+    layout get_weights_layout(T& p, const int32_t /* split */, cldnn::format f) {
         cldnn::tensor weights_tensor;
         weights_tensor = cldnn::tensor(batch(p.out_shape.feature[0]), feature(static_cast<int32_t>(p.in_shape.feature[0] / p.groups)),
                                        spatial(p.kernel.spatial[0], p.kernel.spatial[1], p.kernel.spatial[2]));
@@ -836,7 +830,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_prelu_eltwise,
 
 class conv_fp32_multi_eltwise_2 : public ConvFusingTest {};
 TEST_P(conv_fp32_multi_eltwise_2, basic) {
-    if (engine.get_info().supports_immad) {
+    if (engine.get_device_info().supports_immad) {
         return;
     }
 
@@ -871,7 +865,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_2,
 
 class conv_fp32_multi_eltwise_2_clamp : public ConvFusingTest {};
 TEST_P(conv_fp32_multi_eltwise_2_clamp, basic) {
-    if (engine.get_info().supports_immad) {
+    if (engine.get_device_info().supports_immad) {
         return;
     }
 
@@ -907,7 +901,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_2_clamp,
 
 class conv_fp32_multi_eltwise_4_clamp : public ConvFusingTest {};
 TEST_P(conv_fp32_multi_eltwise_4_clamp, basic) {
-    if (engine.get_info().supports_immad) {
+    if (engine.get_device_info().supports_immad) {
         return;
     }
 
@@ -947,7 +941,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_4_clamp,
 
 class conv_fp32_multi_eltwise_3_fusing : public ConvFusingTest {};
 TEST_P(conv_fp32_multi_eltwise_3_fusing, basic) {
-    if (engine.get_info().supports_immad) {
+    if (engine.get_device_info().supports_immad) {
         return;
     }
 
@@ -8419,4 +8413,3 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, gather_nd_activation_scale_eltwise,
         gather_nd_test_params{ CASE_GATHER_ND_FP32_6D_3, 2, 5 },
         gather_nd_test_params{ CASE_GATHER_ND_FP32_6D_4, 2, 5 },
 }), );
-
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/gather_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/gather_gpu_test.cpp
index fbfe62957f6ffd..c0387763873953 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/gather_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/gather_gpu_test.cpp
@@ -2,23 +2,18 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
+#include "test_utils.h"
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/gather.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/gather.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
 
 TEST(gather7_gpu_fp16, d222_axisX_bdim_m1) {
-    //  Dictionary : 2x2x2x2x2x2  
+    //  Dictionary : 2x2x2x2x2x2
     //  Indexes : 2x2x2x1
     //  Axis : 5
     //  batch_dim : -1
@@ -35,19 +30,19 @@ TEST(gather7_gpu_fp16, d222_axisX_bdim_m1) {
     //  55.f  56.f  57.f  58.f  59.f  60.f  61.f  62.f  63.f  64.f
     //
     //  Output:
-    //  1.f,  2.f,  3.f,  4.f,  5.f,  6.f,  7.f,  8.f, 
-    //  9.f,  10.f, 11.f, 12.f, 13.f, 14.f, 15.f, 16.f, 
-    //  17.f, 17.f, 19.f, 19.f, 21.f, 21.f, 23.f, 23.f, 
-    //  25.f, 25.f, 27.f, 27.f, 29.f, 29.f, 31.f, 31.f, 
-    //  33.f, 33.f, 35.f, 35.f, 37.f, 37.f, 39.f, 39.f, 
-    //  41.f, 41.f, 43.f, 43.f, 45.f, 45.f, 47.f, 47.f, 
-    //  50.f, 49.f, 52.f, 51.f, 54.f, 53.f, 56.f, 55.f, 
+    //  1.f,  2.f,  3.f,  4.f,  5.f,  6.f,  7.f,  8.f,
+    //  9.f,  10.f, 11.f, 12.f, 13.f, 14.f, 15.f, 16.f,
+    //  17.f, 17.f, 19.f, 19.f, 21.f, 21.f, 23.f, 23.f,
+    //  25.f, 25.f, 27.f, 27.f, 29.f, 29.f, 31.f, 31.f,
+    //  33.f, 33.f, 35.f, 35.f, 37.f, 37.f, 39.f, 39.f,
+    //  41.f, 41.f, 43.f, 43.f, 45.f, 45.f, 47.f, 47.f,
+    //  50.f, 49.f, 52.f, 51.f, 54.f, 53.f, 56.f, 55.f,
     //  58.f, 57.f, 60.f, 59.f, 62.f, 61.f, 64.f, 63.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2} }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 2 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2} }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 2 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_x;
     int64_t batch_dim = -1;
 
@@ -68,14 +63,14 @@ TEST(gather7_gpu_fp16, d222_axisX_bdim_m1) {
     set_values(input2, {
         0.f, 1.f,
         0.f, 0.f,
-        
+
         0.f, 0.f,
         1.f, 0.f
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfwzyx, tensor(2, 2, 2, 2, 2, 2), batch_dim)
     );
@@ -88,16 +83,16 @@ TEST(gather7_gpu_fp16, d222_axisX_bdim_m1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-        1.f,  2.f,  3.f,  4.f,  5.f,  6.f,  7.f,  8.f, 
-        9.f,  10.f, 11.f, 12.f, 13.f, 14.f, 15.f, 16.f, 
-        17.f, 17.f, 19.f, 19.f, 21.f, 21.f, 23.f, 23.f, 
-        25.f, 25.f, 27.f, 27.f, 29.f, 29.f, 31.f, 31.f, 
-        33.f, 33.f, 35.f, 35.f, 37.f, 37.f, 39.f, 39.f, 
-        41.f, 41.f, 43.f, 43.f, 45.f, 45.f, 47.f, 47.f, 
-        50.f, 49.f, 52.f, 51.f, 54.f, 53.f, 56.f, 55.f, 
+        1.f,  2.f,  3.f,  4.f,  5.f,  6.f,  7.f,  8.f,
+        9.f,  10.f, 11.f, 12.f, 13.f, 14.f, 15.f, 16.f,
+        17.f, 17.f, 19.f, 19.f, 21.f, 21.f, 23.f, 23.f,
+        25.f, 25.f, 27.f, 27.f, 29.f, 29.f, 31.f, 31.f,
+        33.f, 33.f, 35.f, 35.f, 37.f, 37.f, 39.f, 39.f,
+        41.f, 41.f, 43.f, 43.f, 45.f, 45.f, 47.f, 47.f,
+        50.f, 49.f, 52.f, 51.f, 54.f, 53.f, 56.f, 55.f,
         58.f, 57.f, 60.f, 59.f, 62.f, 61.f, 64.f, 63.f
     };
 
@@ -107,7 +102,7 @@ TEST(gather7_gpu_fp16, d222_axisX_bdim_m1) {
 }
 
 TEST(gather7_gpu_fp16, d323_axisY_bdim_m1) {
-    //  Dictionary : 3x2x3x4x2  
+    //  Dictionary : 3x2x3x4x2
     //  Indexes : 3x2x3x1
     //  Axis : 3
     //  batch_dim : -1
@@ -138,10 +133,10 @@ TEST(gather7_gpu_fp16, d323_axisY_bdim_m1) {
     //  119.f 120.f 115.f 116.f 113.f 114.f 125.f 126.f 121.f 122.f 123.f 124.f
     //  133.f 134.f 129.f 130.f 131.f 132.f 141.f 142.f 137.f 138.f 139.f 140.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 3, 2, 2, 4, 3} }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 2, 1, 3 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 3, 2, 2, 4, 3} }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 2, 1, 3 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_y;
     int64_t batch_dim = -1;
 
@@ -176,17 +171,17 @@ TEST(gather7_gpu_fp16, d323_axisY_bdim_m1) {
     set_values(input2, {
         0.f, 0.f, 0.f,
         3.f, 1.f, 0.f,
-        
+
         1.f, 1.f, 1.f,
         2.f, 0.f, 3.f,
-        
+
         3.f, 1.f, 0.f,
         2.f, 0.f, 1.f
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfzyx, tensor(3, 2, 2, 3, 3), batch_dim)
     );
@@ -199,7 +194,7 @@ TEST(gather7_gpu_fp16, d323_axisY_bdim_m1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f,   2.f,   1.f,   2.f,   1.f,   2.f,
@@ -218,7 +213,7 @@ TEST(gather7_gpu_fp16, d323_axisY_bdim_m1) {
         77.f,  78.f,  73.f,  74.f,  79.f,  80.f,
         85.f,  86.f,  81.f,  82.f,  87.f,  88.f,
         93.f,  94.f,  89.f,  90.f,  95.f,  96.f,
-        
+
 
         103.f, 104.f,  99.f,  100.f, 97.f,  98.f,
         111.f, 112.f, 107.f, 108.f, 105.f, 106.f,
@@ -256,28 +251,28 @@ TEST(gather7_gpu_fp16, d44_axisY_bdim1) {
     //  29.f 16.f 75.f 74.f 74.f  8.f 29.f 29.f  7.f 18.f 54.f 54.f 38.f 16.f 40.f 40.f 74.f 24.f
     //  74.f 25.f 82.f 74.f 71.f  9.f 92.f 71.f 80.f 64.f 27.f 80.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 3, 1, 5 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 4, 4, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 3, 1, 5 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 4, 4, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_y;
     int64_t batch_dim = 1;
 
     set_values(input1, {
         FLOAT16(84.f), FLOAT16( 7.f), FLOAT16(10.f), FLOAT16(69.f), FLOAT16(13.f),
-        FLOAT16(47.f), FLOAT16(75.f), FLOAT16( 8.f), FLOAT16(65.f), FLOAT16(28.f), 
-        FLOAT16( 5.f), FLOAT16(12.f), FLOAT16(56.f), FLOAT16(54.f), FLOAT16( 9.f), 
-        
-        FLOAT16(31.f), FLOAT16(12.f), FLOAT16(71.f), FLOAT16(55.f), FLOAT16( 8.f), 
-        FLOAT16(73.f), FLOAT16(16.f), FLOAT16(29.f), FLOAT16(81.f), FLOAT16(81.f), 
-        FLOAT16(75.f), FLOAT16( 8.f), FLOAT16(74.f), FLOAT16(75.f), FLOAT16(51.f), 
-
-        FLOAT16( 7.f), FLOAT16(29.f), FLOAT16( 6.f), FLOAT16(72.f), FLOAT16(18.f), 
-        FLOAT16(38.f), FLOAT16(54.f), FLOAT16(19.f), FLOAT16(70.f), FLOAT16(16.f), 
-        FLOAT16(74.f), FLOAT16(40.f), FLOAT16(72.f), FLOAT16(88.f), FLOAT16(24.f), 
-        
-        FLOAT16(14.f), FLOAT16(75.f), FLOAT16(74.f), FLOAT16(82.f), FLOAT16(25.f), 
-        FLOAT16(48.f), FLOAT16(13.f), FLOAT16(71.f), FLOAT16(92.f), FLOAT16( 9.f), 
+        FLOAT16(47.f), FLOAT16(75.f), FLOAT16( 8.f), FLOAT16(65.f), FLOAT16(28.f),
+        FLOAT16( 5.f), FLOAT16(12.f), FLOAT16(56.f), FLOAT16(54.f), FLOAT16( 9.f),
+
+        FLOAT16(31.f), FLOAT16(12.f), FLOAT16(71.f), FLOAT16(55.f), FLOAT16( 8.f),
+        FLOAT16(73.f), FLOAT16(16.f), FLOAT16(29.f), FLOAT16(81.f), FLOAT16(81.f),
+        FLOAT16(75.f), FLOAT16( 8.f), FLOAT16(74.f), FLOAT16(75.f), FLOAT16(51.f),
+
+        FLOAT16( 7.f), FLOAT16(29.f), FLOAT16( 6.f), FLOAT16(72.f), FLOAT16(18.f),
+        FLOAT16(38.f), FLOAT16(54.f), FLOAT16(19.f), FLOAT16(70.f), FLOAT16(16.f),
+        FLOAT16(74.f), FLOAT16(40.f), FLOAT16(72.f), FLOAT16(88.f), FLOAT16(24.f),
+
+        FLOAT16(14.f), FLOAT16(75.f), FLOAT16(74.f), FLOAT16(82.f), FLOAT16(25.f),
+        FLOAT16(48.f), FLOAT16(13.f), FLOAT16(71.f), FLOAT16(92.f), FLOAT16( 9.f),
         FLOAT16(73.f), FLOAT16( 8.f), FLOAT16(80.f), FLOAT16(27.f), FLOAT16(64.f)
     });
 
@@ -289,8 +284,8 @@ TEST(gather7_gpu_fp16, d44_axisY_bdim1) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(4, 3, 1, 4), batch_dim)
     );
@@ -303,23 +298,23 @@ TEST(gather7_gpu_fp16, d44_axisY_bdim1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-        69.f, 10.f, 69.f, 13.f, 
-        65.f,  8.f, 65.f, 28.f, 
-        54.f, 56.f, 54.f,  9.f, 
+        69.f, 10.f, 69.f, 13.f,
+        65.f,  8.f, 65.f, 28.f,
+        54.f, 56.f, 54.f,  9.f,
 
-        55.f, 71.f, 71.f, 12.f, 
-        81.f, 29.f, 29.f, 16.f, 
+        55.f, 71.f, 71.f, 12.f,
+        81.f, 29.f, 29.f, 16.f,
         75.f, 74.f, 74.f,  8.f,
 
-        29.f, 29.f,  7.f, 18.f, 
-        54.f, 54.f, 38.f, 16.f, 
+        29.f, 29.f,  7.f, 18.f,
+        54.f, 54.f, 38.f, 16.f,
         40.f, 40.f, 74.f, 24.f,
 
-        74.f, 25.f, 82.f, 74.f, 
-        71.f,  9.f, 92.f, 71.f, 
+        74.f, 25.f, 82.f, 74.f,
+        71.f,  9.f, 92.f, 71.f,
         80.f, 64.f, 27.f, 80.f
     };
 
@@ -345,10 +340,10 @@ TEST(gather7_gpu_fp16, d32_axisF_bdim_m1) {
     //  Output:
     //  1.f 1.f 4.f 3.f 5.f 5.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_f;
     size_t batch_dim = -1;
 
@@ -364,8 +359,8 @@ TEST(gather7_gpu_fp16, d32_axisF_bdim_m1) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(3, 2, 1, 1), batch_dim)
     );
@@ -378,10 +373,10 @@ TEST(gather7_gpu_fp16, d32_axisF_bdim_m1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-        1.f, 1.f, 
+        1.f, 1.f,
         4.f, 3.f,
         5.f, 5.f,
     };
@@ -408,10 +403,10 @@ TEST(gather7_gpu_fp16, d32_axisF_bdim1) {
     //  Output:
     //  1.f 1.f 4.f 3.f 5.f 5.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_f;
     int64_t batch_dim = 1;
 
@@ -427,8 +422,8 @@ TEST(gather7_gpu_fp16, d32_axisF_bdim1) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(3, 2, 1, 1), batch_dim)
     );
@@ -441,7 +436,7 @@ TEST(gather7_gpu_fp16, d32_axisF_bdim1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f, 1.f, 4.f,
@@ -470,10 +465,10 @@ TEST(gather7_gpu_fp16, d32_axisF_bdim0) {
     //  Output:
     //  1.f 1.f 4.f 3.f 5.f 5.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_f;
     size_t batch_dim = 0;
 
@@ -489,8 +484,8 @@ TEST(gather7_gpu_fp16, d32_axisF_bdim0) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(3, 3, 1, 2), batch_dim)
     );
@@ -503,18 +498,18 @@ TEST(gather7_gpu_fp16, d32_axisF_bdim0) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-        1.f, 1.f, 
+        1.f, 1.f,
         2.f, 1.f,
         1.f, 1.f,
 
-        3.f, 3.f, 
+        3.f, 3.f,
         4.f, 3.f,
         3.f, 3.f,
 
-        5.f, 5.f, 
+        5.f, 5.f,
         6.f, 5.f,
         5.f, 5.f
     };
@@ -540,10 +535,10 @@ TEST(gather_gpu_fp16, d14_axisB) {
     //  Output:
     //  1.f, 2.f, 3.f, 4.f, 3.f, 4.f, 1.f, 2.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_b;
 
     set_values(input1, {
@@ -557,8 +552,8 @@ TEST(gather_gpu_fp16, d14_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(1, 4, 1, 2))
     );
@@ -571,7 +566,7 @@ TEST(gather_gpu_fp16, d14_axisB) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f, 2.f, 3.f, 4.f, 3.f, 4.f, 1.f, 2.f
@@ -599,10 +594,10 @@ TEST(gather_gpu_fp16, d222_axisB) {
     //  Output:
     //  1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 5.f, 6.f, 7.f, 8.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_b;
 
     set_values(input1, {
@@ -619,8 +614,8 @@ TEST(gather_gpu_fp16, d222_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 2, 2, 2))
     );
@@ -633,7 +628,7 @@ TEST(gather_gpu_fp16, d222_axisB) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 5.f, 6.f, 7.f, 8.f
@@ -661,10 +656,10 @@ TEST(gather_gpu_fp16, d22_axisY) {
     //  Output:
     //  1.f, 2.f, 3.f, 2.f, 4.f, 5.f, 6.f, 5.f, 7.f, 8.f, 9.f, 8.f, 10.f, 11.f, 12.f, 11.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_y;
 
     set_values(input1, {
@@ -680,8 +675,8 @@ TEST(gather_gpu_fp16, d22_axisY) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 2, 2, 2))
     );
@@ -694,7 +689,7 @@ TEST(gather_gpu_fp16, d22_axisY) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f, 2.f, 3.f, 2.f, 4.f, 5.f, 6.f, 5.f, 7.f, 8.f, 9.f, 8.f, 10.f, 11.f, 12.f, 11.f
@@ -722,10 +717,10 @@ TEST(gather_gpu_fp16, d22_axisF) {
     //  Output:
     //  1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 3.f, 4.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 9.f, 10.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 1, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 1, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_f;
 
     set_values(input1, {
@@ -741,8 +736,8 @@ TEST(gather_gpu_fp16, d22_axisF) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
             gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 2, 2, 2))
     );
@@ -755,7 +750,7 @@ TEST(gather_gpu_fp16, d22_axisF) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 3.f, 4.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 9.f, 10.f
@@ -782,10 +777,10 @@ TEST(gather_gpu_fp32, d14_axisB) {
     //  Output:
     //  1.f, 2.f, 3.f, 4.f, 3.f, 4.f, 1.f, 2.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_b;
 
     set_values(input1, {
@@ -799,8 +794,8 @@ TEST(gather_gpu_fp32, d14_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(1, 4, 1, 2))
     );
@@ -813,7 +808,7 @@ TEST(gather_gpu_fp32, d14_axisB) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f, 2.f, 3.f, 4.f, 3.f, 4.f, 1.f, 2.f
@@ -841,10 +836,10 @@ TEST(gather_gpu_fp32, d222_axisB) {
     //  Output:
     //  1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 5.f, 6.f, 7.f, 8.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 2, 1, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 2, 1, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_b;
 
     set_values(input1, {
@@ -860,8 +855,8 @@ TEST(gather_gpu_fp32, d222_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 2, 2, 2))
     );
@@ -874,7 +869,7 @@ TEST(gather_gpu_fp32, d222_axisB) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 5.f, 6.f, 7.f, 8.f
@@ -902,10 +897,10 @@ TEST(gather_gpu_fp32, d22_axisY) {
     //  Output:
     //  1.f, 2.f, 3.f, 2.f, 4.f, 5.f, 6.f, 5.f, 7.f, 8.f, 9.f, 8.f, 10.f, 11.f, 12.f, 11.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_y;
 
     set_values(input1, {
@@ -921,8 +916,8 @@ TEST(gather_gpu_fp32, d22_axisY) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 2, 2, 2))
     );
@@ -935,7 +930,7 @@ TEST(gather_gpu_fp32, d22_axisY) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f, 2.f, 3.f, 2.f, 4.f, 5.f, 6.f, 5.f, 7.f, 8.f, 9.f, 8.f, 10.f, 11.f, 12.f, 11.f
@@ -963,10 +958,10 @@ TEST(gather_gpu_fp32, d22_axisF) {
     //  Output:
     //  1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 3.f, 4.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 9.f, 10.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 3, 1, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 3, 1, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_f;
 
     set_values(input1, {
@@ -982,8 +977,8 @@ TEST(gather_gpu_fp32, d22_axisF) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
             gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 2, 2, 2))
     );
@@ -996,7 +991,7 @@ TEST(gather_gpu_fp32, d22_axisF) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 3.f, 4.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 9.f, 10.f
@@ -1024,10 +1019,10 @@ TEST(gather_gpu_int32, d22_axisF) {
     //  Output:
     //  1, 2, 3, 4, 5, 6, 3, 4, 7, 8, 9, 10, 11, 12, 9, 10
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 3, 1, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 3, 1, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_f;
 
     set_values(input1, {
@@ -1043,8 +1038,8 @@ TEST(gather_gpu_int32, d22_axisF) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
             gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 2, 2, 2))
     );
@@ -1057,7 +1052,7 @@ TEST(gather_gpu_int32, d22_axisF) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<int>();
+    cldnn::mem_lock<int> output_ptr(output, get_test_stream());
 
     std::vector<int> expected_results = {
             1, 2, 3, 4, 5, 6, 3, 4, 7, 8, 9, 10, 11, 12, 9, 10
@@ -1084,10 +1079,10 @@ TEST(gather_gpu_int32, d14_axisB) {
     //  Output:
     //  1, 2, 3, 4, 3, 4, 1, 2
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 1, 4, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 1, 4, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_b;
 
     set_values(input1, {
@@ -1101,8 +1096,8 @@ TEST(gather_gpu_int32, d14_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
             gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(1, 4, 1, 2))
     );
@@ -1115,7 +1110,7 @@ TEST(gather_gpu_int32, d14_axisB) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<int>();
+    cldnn::mem_lock<int> output_ptr(output, get_test_stream());
 
     std::vector<int> expected_results = {
             1, 2, 3, 4, 3, 4, 1, 2
@@ -1143,10 +1138,10 @@ TEST(gather_gpu_int32, d222_axisB) {
     //  Output:
     //  1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 5, 6, 7, 8
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 2, 1, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 2, 1, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_b;
 
     set_values(input1, {
@@ -1162,8 +1157,8 @@ TEST(gather_gpu_int32, d222_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
             gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 2, 2, 2))
     );
@@ -1176,7 +1171,7 @@ TEST(gather_gpu_int32, d222_axisB) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<int>();
+    cldnn::mem_lock<int> output_ptr(output, get_test_stream());
 
     std::vector<int> expected_results = {
             1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 5, 6, 7, 8
@@ -1204,10 +1199,10 @@ TEST(gather_gpu_int32, d22_axisY) {
     //  Output:
     //  1, 2, 3, 2, 4, 5, 6, 5, 7, 8, 9, 8, 10, 11, 12, 11
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 2, 1, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 2, 1, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_y;
 
     set_values(input1, {
@@ -1223,8 +1218,8 @@ TEST(gather_gpu_int32, d22_axisY) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
             gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 2, 2, 2))
     );
@@ -1237,7 +1232,7 @@ TEST(gather_gpu_int32, d22_axisY) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<int>();
+    cldnn::mem_lock<int> output_ptr(output, get_test_stream());
 
     std::vector<int> expected_results = {
             1, 2, 3, 2, 4, 5, 6, 5, 7, 8, 9, 8, 10, 11, 12, 11
@@ -1268,10 +1263,10 @@ TEST(gather_gpu_fp32, d41_axisB) {
     //  7, 8, 9, 10, 11, 12
     //  1, 2, 3, 4, 5, 6,
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_b;
 
     set_values(input1, {
@@ -1287,8 +1282,8 @@ TEST(gather_gpu_fp32, d41_axisB) {
                });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(4, 1, 3, 2))
     );
@@ -1301,7 +1296,7 @@ TEST(gather_gpu_fp32, d41_axisB) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             1.f, 2.f, 3.f, 4.f, 5.f, 6.f,
@@ -1334,10 +1329,10 @@ TEST(gather_gpu_fp32, d41_axisF) {
     //  3, 4,   1, 2,   3, 4,   5, 6,
     //  9, 10,  7, 8,   9, 10,  11, 12
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 3, 1, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 3, 1, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_f;
 
     set_values(input1, {
@@ -1350,8 +1345,8 @@ TEST(gather_gpu_fp32, d41_axisF) {
                });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 4, 2, 1))
     );
@@ -1364,7 +1359,7 @@ TEST(gather_gpu_fp32, d41_axisF) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             3.f, 4.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f,
@@ -1393,10 +1388,10 @@ TEST(gather_gpu_fp32, d2_axisX) {
     //  Output:
     //  1, 1, 2, 2, 3, 3, 4, 4
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 1, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 1, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_x;
 
     set_values(input1, {
@@ -1409,8 +1404,8 @@ TEST(gather_gpu_fp32, d2_axisX) {
                });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(2, 2, 2, 1))
     );
@@ -1423,7 +1418,7 @@ TEST(gather_gpu_fp32, d2_axisX) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             1.f, 1.f, 2.f, 2.f,
@@ -1443,10 +1438,10 @@ TEST(gather_gpu_fp32, 322_axisF) {
     //  Output : 3x2x2x1
     //  Input values in i32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 3, 1, 1 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input1 = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 3, 1, 1 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
     auto axis = cldnn::gather::gather_axis::along_f;
 
     set_values(input1, {
@@ -1459,8 +1454,8 @@ TEST(gather_gpu_fp32, 322_axisF) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
     topology.add(
         gather("gather", "InputDictionary", "InputText", axis, format::bfyx, tensor(3, 2, 1, 2))
     );
@@ -1473,7 +1468,7 @@ TEST(gather_gpu_fp32, 322_axisF) {
     auto outputs = network.execute();
 
     auto output = outputs.at("gather").get_memory();
-    auto output_ptr = output.pointer<int>();
+    cldnn::mem_lock<int> output_ptr(output, get_test_stream());
 
     std::vector<int> expected_results = {
         1, 0, 2, 1,   11, 10, 12, 11,   21, 20, 22, 21
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/gather_nd_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/gather_nd_gpu_test.cpp
index 2999969928772b..9be5392902a836 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/gather_nd_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/gather_nd_gpu_test.cpp
@@ -1,41 +1,25 @@
-// Copyright (c) 2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
 //
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
+#include "test_utils.h"
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/gather_nd.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/gather_nd.hpp>
 
-#include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
 
-inline void DoTest(const engine& engine,
-    const cldnn::memory& input0,
-    const cldnn::memory& input1,
+inline void DoTest(engine& engine,
+    const cldnn::memory::ptr input0,
+    const cldnn::memory::ptr input1,
     const std::vector<float>& expected_results,
     const int indices_rank,
     const int batch_dims) {
     topology topology;
-    topology.add(input_layout("InputData", input0.get_layout()));
-    topology.add(input_layout("InputIndices", input1.get_layout()));
+    topology.add(input_layout("InputData", input0->get_layout()));
+    topology.add(input_layout("InputIndices", input1->get_layout()));
     topology.add(
         gather_nd("gather_nd", "InputData", "InputIndices", indices_rank, batch_dims)
     );
@@ -46,7 +30,7 @@ inline void DoTest(const engine& engine,
     network.set_input_data("InputIndices", input1);
     auto outputs = network.execute();
     auto output = outputs.at("gather_nd").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -54,12 +38,12 @@ inline void DoTest(const engine& engine,
 }
 
 TEST(gather_nd_gpu_fp16, d23322_i231312_ir6_batch2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 6;
     const int batch_dims = 2;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 3, 2, 2, 3 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 3, 2, 1, 3, 1 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 3, 2, 2, 3 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 3, 2, 1, 3, 1 } }); // indices
     // expected output dim: {6,1,3,1,2}
 
     set_values(input0, {
@@ -96,12 +80,12 @@ TEST(gather_nd_gpu_fp16, d23322_i231312_ir6_batch2) {
 }
 
 TEST(gather_nd_gpu_fp16, d231322_i231321_ir6_batch5) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 6;
     const int batch_dims = 5;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 3, 2, 2, 3, 1 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 3, 1, 2, 3, 1 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 3, 2, 2, 3, 1 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 3, 1, 2, 3, 1 } }); // indices
     // expected output dim: {36}
 
     set_values(input0, {
@@ -138,12 +122,12 @@ TEST(gather_nd_gpu_fp16, d231322_i231321_ir6_batch5) {
 }
 
 TEST(gather_nd_gpu_fp16, d23322_i23321_ir5_batch4) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 5;
     const int batch_dims = 4;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 3, 2, 2, 3 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 3, 1, 2, 3 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 3, 2, 2, 3 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 3, 1, 2, 3 } }); // indices
     // expected output dim: {36}
 
     set_values(input0, {
@@ -180,12 +164,12 @@ TEST(gather_nd_gpu_fp16, d23322_i23321_ir5_batch4) {
 }
 
 TEST(gather_nd_gpu_fp16, d23223_i2321_ir4_batch3) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 4;
     const int batch_dims = 3;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 3, 3, 2, 2 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 1, 2 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 3, 3, 2, 2 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 1, 2 } }); // indices
     // expected output dim: {2*3*2,3}
 
     set_values(input0, {
@@ -222,12 +206,12 @@ TEST(gather_nd_gpu_fp16, d23223_i2321_ir4_batch3) {
 }
 
 TEST(gather_nd_gpu_fp16, d2342_i2312_ir4_batch2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 4;
     const int batch_dims = 2;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 2, 4 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 2, 1 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 2, 4 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 2, 1 } }); // indices
     // expected output dim: {6,1}
 
     set_values(input0, {
@@ -264,12 +248,12 @@ TEST(gather_nd_gpu_fp16, d2342_i2312_ir4_batch2) {
 }
 
 TEST(gather_nd_gpu_fp16, d234_i2311_ir4_batch2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 4;
     const int batch_dims = 2;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 1, 4 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 1, 1 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 1, 4 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 1, 1 } }); // indices
     // expected output dim: {6,1,1}
 
     set_values(input0, {
@@ -307,12 +291,12 @@ TEST(gather_nd_gpu_fp16, d234_i2311_ir4_batch2) {
 }
 
 TEST(gather_nd_gpu_fp16, d234_i21_ir2_batch1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 2;
     const int batch_dims = 1;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 1, 4 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 1, 4 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
     // expected output dim: {2,4}
 
     set_values(input0, {
@@ -340,12 +324,12 @@ TEST(gather_nd_gpu_fp16, d234_i21_ir2_batch1) {
 }
 
 TEST(gather_nd_gpu_fp16, d22_i21_ir2_batch1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 2;
     const int batch_dims = 1;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
     // expected output dim: 2
 
     set_values(input0, {
@@ -367,12 +351,12 @@ TEST(gather_nd_gpu_fp16, d22_i21_ir2_batch1) {
 }
 
 TEST(gather_nd_gpu_fp16, d3223_i321113_ir6_batch0) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 6;
     const int batch_dims = 0;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 3, 2 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 3, 2, 3, 1, 1, 1 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 3, 2 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 3, 2, 3, 1, 1, 1 } }); // indices
     // expected output dim: 321113
 
     set_values(input0, {
@@ -412,12 +396,12 @@ TEST(gather_nd_gpu_fp16, d3223_i321113_ir6_batch0) {
 }
 
 TEST(gather_nd_gpu_fp16, d3221_i32312_ir3_batch0) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 3;
     const int batch_dims = 0;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 3, 2, 2, 1, 3 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 2 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 3, 2, 2, 1, 3 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 2 } }); // indices
     // expected output dim: 32312
 
     set_values(input0, {
@@ -457,12 +441,12 @@ TEST(gather_nd_gpu_fp16, d3221_i32312_ir3_batch0) {
 }
 
 TEST(gather_nd_gpu_fp16, d3231_i32312_ir3_batch0) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 3;
     const int batch_dims = 0;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 3, 2, 2, 1, 3 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 3 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 3, 2, 2, 1, 3 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 3 } }); // indices
     // expected output dim: {3,2,1,2}
 
     set_values(input0, {
@@ -502,12 +486,12 @@ TEST(gather_nd_gpu_fp16, d3231_i32312_ir3_batch0) {
 }
 
 TEST(gather_nd_gpu_fp16, d3112_i3221_ir4_batch0) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 4;
     const int batch_dims = 0;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 1, 2, 1 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 2 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 1, 2, 1 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 2 } }); // indices
     // expected output dim: {3,2,2,1,1,2}
 
     set_values(input0, {
@@ -542,12 +526,12 @@ TEST(gather_nd_gpu_fp16, d3112_i3221_ir4_batch0) {
 }
 
 TEST(gather_nd_gpu_fp16, d311211_i322111_ir4_batch0) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 4;
     const int batch_dims = 0;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 3, 1, 1, 1, 2, 1 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 3, 2, 1, 1, 1, 2 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 3, 1, 1, 1, 2, 1 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 3, 2, 1, 1, 1, 2 } }); // indices
     // expected output dim: {3,2,2,1,1,2,1,1}
 
     set_values(input0, {
@@ -582,12 +566,12 @@ TEST(gather_nd_gpu_fp16, d311211_i322111_ir4_batch0) {
 }
 
 TEST(gather_nd_gpu_fp16, d3332_i3223_ir4_batch0) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 4;
     const int batch_dims = 0;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 3, 3, 2 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 3, 2 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 3, 3, 2 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 3, 2 } }); // indices
 
     set_values(input0, {
         FLOAT16(1), FLOAT16(2), FLOAT16(3),     FLOAT16(4), FLOAT16(5), FLOAT16(6),
@@ -629,12 +613,12 @@ TEST(gather_nd_gpu_fp16, d3332_i3223_ir4_batch0) {
 }
 
 TEST(gather_nd_gpu_fp16, d3323_i322_ir3_batch0) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 3;
     const int batch_dims = 0;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 3, 3, 2 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 2 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 3, 3, 2 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 2 } }); // indices
 
     set_values(input0, {
         FLOAT16(1), FLOAT16(2), FLOAT16(3),     FLOAT16(4), FLOAT16(5), FLOAT16(6),
@@ -676,12 +660,12 @@ TEST(gather_nd_gpu_fp16, d3323_i322_ir3_batch0) {
 }
 
 TEST(gather_nd_gpu_fp16, d22_i21_ir2_batch0) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 2;
     const int batch_dims = 0;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
 
     set_values(input0, {
         FLOAT16(1), FLOAT16(2),
@@ -701,12 +685,12 @@ TEST(gather_nd_gpu_fp16, d22_i21_ir2_batch0) {
 }
 
 TEST(gather_nd_gpu_fp16, d22_i32_ir2_batch0) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int indices_rank = 2;
     const int batch_dims = 0;
-    auto input0 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // data
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // indices
+    auto input0 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // data
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // indices
 
     set_values(input0, {
         FLOAT16(1), FLOAT16(2),
@@ -727,4 +711,3 @@ TEST(gather_nd_gpu_fp16, d22_i32_ir2_batch0) {
 
     DoTest(engine,input0, input1, expected_results, indices_rank, batch_dims);
 }
-
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/gemm_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/gemm_gpu_test.cpp
index 136c1a3c107f04..ae1ca356924dcf 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/gemm_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/gemm_gpu_test.cpp
@@ -3,17 +3,11 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/engine.hpp>
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/gemm.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include "test_utils.h"
 
-#include "test_utils/test_utils.h"
-#include "test_utils/uniform_quantized_real_distribution.hpp"
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/gemm.hpp>
 
 #include <cstddef>
 
@@ -21,14 +15,14 @@ using namespace cldnn;
 using namespace ::tests;
 
 TEST(gemm_gpu, basic_bfyx_t1) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 3, 4 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 4 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 3, 4 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 4 } });
 
     std::vector<float> input_data = {
-        1.f, -2.f,  3.f,  
-        -4.f, 5.f,  6.f, 
-        1.f, 2.f, 3.f, 
+        1.f, -2.f,  3.f,
+        -4.f, 5.f,  6.f,
+        1.f, 2.f, 3.f,
         3.f, 2.f, -1.f,
 
     };
@@ -48,10 +42,10 @@ TEST(gemm_gpu, basic_bfyx_t1) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2" }, data_types::f32, true, false)
@@ -63,7 +57,7 @@ TEST(gemm_gpu, basic_bfyx_t1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     EXPECT_EQ(output_ptr.size(), (uint32_t)3);
     for (uint32_t i = 0; i < out_data.size(); ++i) {
@@ -71,13 +65,13 @@ TEST(gemm_gpu, basic_bfyx_t1) {
     }
 }
 TEST(gemm_gpu, basic_bfyx_t2) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 3 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 3 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 1 } });
 
     std::vector<float> input_data = {
-        1.f, -2.f,  3.f, -4.f, 
-        5.f,  6.f, 1.f, 2.f, 
+        1.f, -2.f,  3.f, -4.f,
+        5.f,  6.f, 1.f, 2.f,
         3.f, 3.f, 2.f, -1.f,
 
     };
@@ -94,10 +88,10 @@ TEST(gemm_gpu, basic_bfyx_t2) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2" }, data_types::f32, false, true)
@@ -109,7 +103,7 @@ TEST(gemm_gpu, basic_bfyx_t2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     EXPECT_EQ(output_ptr.size(), (uint32_t)3);
     for (uint32_t i = 0; i < out_data.size(); ++i) {
@@ -118,14 +112,14 @@ TEST(gemm_gpu, basic_bfyx_t2) {
 }
 
 TEST(gemm_gpu, basic_bfyx_t1t2) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 1, 3, 4 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 1, 4, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 1, 3, 4 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 1, 4, 1 } });
 
     std::vector<float> input_data = {
-        1.f, -2.f,  3.f, 
-        -4.f, 5.f,  6.f, 
-        1.f, 2.f, 3.f, 
+        1.f, -2.f,  3.f,
+        -4.f, 5.f,  6.f,
+        1.f, 2.f, 3.f,
         3.f, 2.f, -1.f,
 
         1.f, -2.f,  3.f,
@@ -150,10 +144,10 @@ TEST(gemm_gpu, basic_bfyx_t1t2) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2" }, data_types::f32, true, true)
@@ -165,7 +159,7 @@ TEST(gemm_gpu, basic_bfyx_t1t2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     EXPECT_EQ(output_ptr.size(), (uint32_t)6);
     for (uint32_t i = 0; i < out_data.size(); ++i) {
@@ -174,19 +168,19 @@ TEST(gemm_gpu, basic_bfyx_t1t2) {
 }
 
 TEST(gemm_gpu, basic_input3) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 3, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 3 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 3, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 3 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
     float alpha = 2.f;
     float beta = 10.f;
 
-    std::vector<float> input_data = { 
+    std::vector<float> input_data = {
         1.0f, 2.0f, 3.0f,
         1.0f, 0.0f, 1.0f
     };
 
-    std::vector<float> input_data2 = { 
+    std::vector<float> input_data2 = {
         3.0f, 3.0f,
         1.0f, 2.0f,
         1.0f, 2.0f,
@@ -208,13 +202,13 @@ TEST(gemm_gpu, basic_input3) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
-        input_layout("input3", input3.get_layout())
+        input_layout("input3", input3->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2", "input3" }, data_types::f32,  false, false, alpha, beta)
@@ -227,20 +221,20 @@ TEST(gemm_gpu, basic_input3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     EXPECT_EQ(output_ptr.size(), (uint32_t)4);
 
-    for (uint32_t i = 0; i < out_data.size(); ++i) {         
+    for (uint32_t i = 0; i < out_data.size(); ++i) {
         EXPECT_FLOAT_EQ(output_ptr[i], out_data[i]);
     }
 }
 
 TEST(gemm_gpu, basic_input3_t1t2) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 3 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 3, 2 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 4 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 3 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 3, 2 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 4 } });
     float alpha = 2.f;
     float beta = 3.f;
 
@@ -251,7 +245,7 @@ TEST(gemm_gpu, basic_input3_t1t2) {
     };
 
     std::vector<float> input_data2 = {
-        3.0f, 3.0f, 1.0f, 
+        3.0f, 3.0f, 1.0f,
         2.0f, 1.0f, 2.0f,
     };
 
@@ -275,13 +269,13 @@ TEST(gemm_gpu, basic_input3_t1t2) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
-        input_layout("input3", input3.get_layout())
+        input_layout("input3", input3->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2", "input3" }, data_types::f32, true, true, alpha, beta)
@@ -294,7 +288,7 @@ TEST(gemm_gpu, basic_input3_t1t2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     EXPECT_EQ(output_ptr.size(), (uint32_t)8);
 
@@ -303,10 +297,10 @@ TEST(gemm_gpu, basic_input3_t1t2) {
     }
 }
 TEST(gemm_gpu, basic_input3_1) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 3, 4 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 3 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 4 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 3, 4 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 3 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 4 } });
     float alpha = 2.f;
     float beta = 3.f;
 
@@ -343,13 +337,13 @@ TEST(gemm_gpu, basic_input3_1) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
-        input_layout("input3", input3.get_layout())
+        input_layout("input3", input3->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2", "input3" }, data_types::f32, false, false, alpha, beta)
@@ -363,7 +357,7 @@ TEST(gemm_gpu, basic_input3_1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     EXPECT_EQ(output_ptr.size(), (uint32_t)8);
 
@@ -373,10 +367,10 @@ TEST(gemm_gpu, basic_input3_1) {
 }
 
 TEST(gemm_gpu, basic_input3_t2) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 3, 4 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 3, 2 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 4 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 3, 4 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 3, 2 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 4 } });
     float alpha = 2.f;
     float beta = 3.f;
 
@@ -412,13 +406,13 @@ TEST(gemm_gpu, basic_input3_t2) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
-        input_layout("input3", input3.get_layout())
+        input_layout("input3", input3->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2", "input3" }, data_types::f32, false, true, alpha, beta)
@@ -431,7 +425,7 @@ TEST(gemm_gpu, basic_input3_t2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     EXPECT_EQ(output_ptr.size(), (uint32_t)8);
 
@@ -441,10 +435,10 @@ TEST(gemm_gpu, basic_input3_t2) {
 }
 
 TEST(gemm_gpu, basic_input3_t1) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 4, 3 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 3 } });
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 4 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 4, 3 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 3 } });
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 4 } });
     float alpha = 2.f;
     float beta = 3.f;
 
@@ -480,13 +474,13 @@ TEST(gemm_gpu, basic_input3_t1) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
-        input_layout("input3", input3.get_layout())
+        input_layout("input3", input3->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2", "input3" }, data_types::f32, true, false, alpha, beta)
@@ -499,7 +493,7 @@ TEST(gemm_gpu, basic_input3_t1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     EXPECT_EQ(output_ptr.size(), (uint32_t)8);
 
@@ -509,13 +503,13 @@ TEST(gemm_gpu, basic_input3_t1) {
 }
 
 TEST(gemm_gpu, basic_bfyx) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 1, 4, 3 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 1, 1, 4 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 1, 4, 3 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 1, 1, 4 } });
 
     std::vector<float> input_data = {
-        1.f, -2.f,  3.f,  -4.f, 
-        5.f,  6.f,  7.f,   8.f, 
+        1.f, -2.f,  3.f,  -4.f,
+        5.f,  6.f,  7.f,   8.f,
         -10.f, 12.f, 13.f, -13.f,
 
         1.f, -2.f,  3.f,  -4.f,
@@ -524,9 +518,9 @@ TEST(gemm_gpu, basic_bfyx) {
     };
 
     std::vector<float> input_data2 = {
-        2.f, 
-        5.f, 
-        -4.f,  
+        2.f,
+        5.f,
+        -4.f,
         -7.f,
         2.f,
         5.f,
@@ -542,10 +536,10 @@ TEST(gemm_gpu, basic_bfyx) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2" }, data_types::f32)
@@ -557,9 +551,9 @@ TEST(gemm_gpu, basic_bfyx) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+
 
-    
     EXPECT_EQ(output_ptr.size(), (uint32_t)6);
     for (uint32_t i = 0; i < out_data.size(); ++i) {
             EXPECT_FLOAT_EQ(output_ptr[i], out_data[i]);
@@ -567,9 +561,9 @@ TEST(gemm_gpu, basic_bfyx) {
 }
 
 TEST(gemm_gpu, basic3_bfyx) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 5, 1, 500, 9 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 5, 1, 1, 500 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 5, 1, 500, 9 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 5, 1, 1, 500 } });
 
     std::vector<float> input_data = {
         -0.000449777f, -0.000137405f, -0.0762075f, 0.000949166f, 0.0346124f, -0.0111424f, 0.0108357f, 0.0121679f, 0.0242749f, 0.052692f, -0.0017713f, 0.0053728f, 0.0128862f, -0.0162366f, 0.0125041f, -0.00602398f, 0.0107778f, -0.00323086f, -0.00914208f, -0.013884f, 0.00755173f, -0.0175622f, 0.00473339f, -0.015003f, -0.0238219f, 0.004502f, 0.00187154f, 0.0041163f, -9.36184e-05f, 0.00873372f, 0.0121869f, -0.020973f, -0.006006f, -0.0038208f, 0.00210471f, 0.00255549f, -0.0251856f, -0.0626372f, -0.0059258f, -0.0058662f, -0.0946306f, 0.00197436f, 0.00105865f, -0.0033595f, 0.0158977f, -0.0036025f, -0.00568902f, -0.0202577f, -0.000251319f, -0.0117895f, -0.0144239f, -0.0144024f, -0.0150431f, -0.0354826f, -0.0135123f, -0.000422157f, 0.0286438f, -0.000884989f, -0.00675718f, 0.013241f, -0.0118388f, 0.0321394f, -0.000803071f, 0.11408f, -0.00806301f, -0.00831608f, 0.0165189f, 0.016094f, -0.000449332f, -0.00695901f, 0.0437514f, -0.00172117f, 0.00180391f, -0.000859933f, -0.0144826f, 0.0262613f, -0.00194352f, -1.98829e-05f, -0.00902827f, -0.00400867f, -0.00600827f, 0.0120846f, -0.0162493f, 0.0418596f, 0.00131911f, -0.00631566f, 0.00270484f, -0.0950513f, 0.00726431f, -0.0169798f, -0.000554365f, -0.00256903f, -0.00885843f, 0.0104025f, 0.00590779f, -0.00175832f, 0.0168603f, 0.00964353f, -0.0180614f, 0.0213157f, 0.0209548f, -0.0231143f, -0.00121617f, -0.0129815f, -0.0199287f, 0.00863336f, -0.00464991f, 0.0162288f, -0.340115f, -0.011018f, -0.0593997f, 0.00644821f, 0.0416332f, 0.0394596f, 0.0172296f, 0.00494231f, 0.0143805f, -0.00819845f, 0.00196982f, 0.00393258f, 0.0246168f, -0.0235927f, 0.0131416f, -0.0190432f, -0.0237865f, -0.0155627f, 0.0265165f, 0.0162884f, 0.00321098f, 0.0136674f, -0.000966112f, -0.0100813f, -0.00604589f, 0.00889466f, 0.0113945f, 0.0264707f, 0.00371883f, -0.00843358f, 0.0145675f, 0.0048638f, 0.00110399f, -0.00130233f, 0.00740726f, -0.00393368f, -0.0242178f, 0.00341681f, 0.00115369f, -0.00297881f, -0.0844071f, 0.0537151f, -0.00209399f, 0.0310295f, 0.0383914f, 0.00456459f, 0.0188114f, -0.0177144f, 0.0133258f, 0.0584683f, -0.00640495f, 0.0175946f, 0.0186782f, 0.00213311f, 0.00393403f, 0.00382759f, 0.00267507f, 0.00493673f, -0.00856695f, -0.00627955f, -0.0103436f, -0.000671664f, -0.110419f, 0.0307264f, 0.0042176f, 0.0031638f, 0.0154172f, 0.00265482f, 0.0410853f, 0.00833895f, -0.0183989f, -0.000717906f, -0.0090387f, -0.00404523f, -0.00976238f, -0.0137555f, 0.000157289f, -0.00341186f, -0.0214878f, 0.0142639f, 0.00624623f, 0.000537292f, -0.0520912f, -0.0432221f, -0.00330415f, 0.0263942f, -0.00150974f, 0.00172088f, -0.0815726f, -0.0201155f, -0.00986346f, 0.0121252f, 0.00198959f, -0.0349936f, -0.00608366f, -0.00399543f, 0.0192487f, -0.0123156f, 0.0072797f, 0.000507143f, 0.0334805f, 0.000609379f, 0.00961966f, -0.00697663f, 0.00201967f, -0.0207349f, -0.0103385f, -0.00343849f, -0.00330492f, 0.035106f, -0.00456996f, 0.00197528f, 0.016148f, 0.0142903f, 0.0616483f, 0.0093118f, -0.0596028f, 0.00945764f, -0.00659242f, 0.118389f, -0.00259384f, -0.00285344f, 0.00567036f, 0.0195813f, -0.00461807f, -0.0608699f, 0.00380259f, 0.00143385f, -0.00466997f, 0.0194046f, -0.0198423f, -0.00334569f, -0.014399f, 0.0130021f, -0.0141619f, -0.00859914f, 0.00997122f, -0.0198446f, -0.0094162f, -0.0116609f, -0.0111888f, -0.00903524f, 0.00937981f, 0.01772f, -0.00236374f, -0.00870162f, 0.000141193f, -0.0343695f, -0.00997931f, 0.0073531f, -0.100394f, -0.00367661f, -0.00124499f, 0.00318026f, 0.0554203f, -0.00342582f, -0.0104147f, -0.0577869f, -0.0126485f, -0.0332496f, 0.0346141f, 0.0307962f, -0.0174745f, -0.0387339f, 0.0167707f, -0.0363424f, 0.0154902f, -0.0118644f, -4.63543e-06f, -0.0683506f, -0.0344076f, -0.00104884f, -0.00883997f, -0.00305185f, -0.0150299f, -0.0186403f, 0.0110238f, 0.00779224f, -0.0102231f, 0.0087488f, -0.0138988f, -0.0229105f, -0.0244903f, -0.0202919f, 0.00135903f, -0.00574432f, 0.00254918f, 0.0340209f, -0.046428f, -0.00670622f, 0.000925543f, -0.0249251f, -0.00275456f, 0.0199177f, 0.000210993f, 0.027762f, -0.0228046f, 0.0484813f, 0.00538959f, 0.0136714f, -0.00690097f, -0.0448533f, -0.00815204f, 0.00734891f, 0.0173959f, -0.0379109f, 0.0594617f, -0.00722084f, 0.0415935f, 0.014792f, -0.0170252f, -0.0139396f, 0.00146415f, 0.00117702f, 0.0685559f, 0.00727832f, -0.107566f, -0.0112505f, 0.033853f, 0.0046957f, -0.0242369f, 0.0148181f, -0.0723487f, -0.00961667f, 0.0304085f, -0.00520772f, -0.0316467f, 0.0327801f, -0.00755137f, 0.0166041f, -0.0557288f, -0.0227759f, -0.00314548f, 0.0152585f, 0.020071f, -0.0377076f, 0.00687613f, -0.0273935f, -0.00647955f, 0.0105047f, -0.0137238f, 0.023264f, -0.0455722f, -0.00221414f, -0.0258535f, -0.0236395f, 0.0593407f, 0.00448763f, 0.0150777f, 0.00437925f, 0.0295782f, -0.0344752f, 0.00365267f, 0.140464f, -0.0479012f, 0.025726f, 0.119063f, 0.000301925f, -0.00810565f, -0.354073f, -0.0723185f, -0.0046123f, 0.033882f, -0.044552f, -0.0138361f, 0.00384129f, 0.0139111f, -0.01667f, -0.0821503f, 0.0029974f, -0.0306725f, 0.0160366f, 0.0334754f, 0.0192693f, -0.00616713f, -0.00232275f, 0.0107987f, 0.00437057f, 0.0017298f, 0.0196916f, -0.0417255f, -0.00911193f, 0.00876709f, -0.00172422f, -0.00105248f, -0.0191631f, -0.00387423f, -0.0102766f, -0.025317f, -0.0416204f, -0.0319611f, -0.00359193f, 0.00424064f, -0.00575092f, -0.0282402f, 0.0745899f, -0.0126492f, -0.0162564f, -0.261967f, -0.705265f, -0.0403731f, -0.00209634f, -0.694297f, 0.00956909f, 0.0158826f, 0.0130207f, 0.003825f, -0.000300812f, -0.0121346f, 0.00642053f, -0.012902f, 0.0309272f, 0.0609192f, -0.00654145f, -0.0937578f, -0.00432024f, -0.00767539f, 0.0461248f, 0.00701077f, -0.0174477f, 0.00563833f, -0.0107107f, -0.0255275f, 0.00892488f, -0.00166062f, 0.039829f, -0.00150394f, 0.00742194f, -0.00885529f, -0.0103532f, 0.0777858f, 0.0885367f, -0.00425715f, 0.0423651f, -0.0446651f, -0.635069f, -0.00919329f, -0.00356176f, 0.00988705f, 0.0116529f, -0.0401253f, 0.00260105f, 0.00573955f, -0.0667439f, 0.101175f, 0.0765288f, -0.0120077f, 0.00322599f, -0.0192768f, 0.0382749f, -0.222119f, -0.0452036f, 0.0424303f, 0.0890699f, 0.0117557f, 0.0315167f, 0.0284256f, 0.00541845f, -0.250147f, 0.00420668f, -0.0189724f, -0.00416381f, -0.00162803f, -0.0108763f, -0.00970892f, 0.0134476f, -0.0254931f, 0.0307225f, 0.00128596f, 0.0171106f, 0.00467854f, -0.0124376f, 0.0183396f, 0.0021754f, 0.00170886f, -0.0226898f, 0.0250111f, -0.0533301f, -0.0163268f, 0.00618995f, 0.0416378f, 0.0475397f, 0.0105684f, -0.00440933f, 0.0496722f, -0.0215733f, -0.0256361f, -0.0285091f, -0.0276881f, -0.00102202f, -0.0720219f, -0.0296656f,
@@ -3145,10 +3139,10 @@ TEST(gemm_gpu, basic3_bfyx) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2" }, data_types::f32)
@@ -3160,7 +3154,7 @@ TEST(gemm_gpu, basic3_bfyx) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     EXPECT_EQ(output_ptr.size(), (uint32_t)45);
     for (uint32_t  i = 0; i < out_data.size(); ++i) {
@@ -3170,9 +3164,9 @@ TEST(gemm_gpu, basic3_bfyx) {
 }
 
 TEST(gemm_gpu, basic_smarcink2) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 1, 3, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 1, 2, 3 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 1, 3, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 1, 2, 3 } });
 
     std::vector<float> input_data = {
         1.0f, 2.0f, 3.0f,
@@ -3204,10 +3198,10 @@ TEST(gemm_gpu, basic_smarcink2) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout())
+        input_layout("input", input->get_layout())
     );
     topology.add(
-        input_layout("input2", input2.get_layout())
+        input_layout("input2", input2->get_layout())
     );
     topology.add(
         gemm("output", { "input", "input2" }, data_types::f32)
@@ -3219,7 +3213,7 @@ TEST(gemm_gpu, basic_smarcink2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     EXPECT_EQ(output_ptr.size(), (uint32_t)8);
     for (uint32_t i = 0; i < out_data.size(); ++i) {
@@ -3414,7 +3408,7 @@ class GemmBaseTest : public ::testing::TestWithParam<gemm_params> {
     }
 
     void execute(gemm_params& p) {
-        const auto& engine = get_test_engine();
+        auto& engine = get_test_engine();
 
         auto y0_size = p.m_size;
         auto y0_pitch = p.k_size;
@@ -3461,19 +3455,19 @@ class GemmBaseTest : public ::testing::TestWithParam<gemm_params> {
         auto input0_size = tensor((int)p.b0_num, (int)p.f0_num, (int)x0_size, (int)y0_size);
         VVVVF<input0_type> input0_data = generate_random_4d<input0_type>(p.b0_num, p.f0_num, x0_size, y0_size, p.range0[0], p.range0[1], p.range0[2]);
         auto input0_data_bfyx = flatten_4d(format::bfyx, input0_data);
-        auto input0_mem = memory::allocate(engine, { p.allocate0_type, format::bfyx, input0_size });
+        auto input0_mem = engine.allocate_memory({ p.allocate0_type, format::bfyx, input0_size });
         set_values(input0_mem, input0_data_bfyx);
 
         auto input1_size = tensor((int)p.b1_num, (int)p.f1_num, (int)x1_size, (int)y1_size);
         VVVVF<input1_type> input1_data = generate_random_4d<input1_type>(p.b1_num, p.f1_num, x1_size, y1_size, p.range1[0], p.range1[1], p.range1[2]);
         auto input1_data_bfyx = flatten_4d(format::bfyx, input1_data);
-        auto input1_mem = memory::allocate(engine, { p.allocate1_type, format::bfyx, input1_size });
+        auto input1_mem = engine.allocate_memory({ p.allocate1_type, format::bfyx, input1_size });
         set_values(input1_mem, input1_data_bfyx);
 
         auto input2_size = tensor((int)p.b2_num, (int)p.f2_num, (int)x2_size, (int)y2_size);
         VVVVF<input2_type> input2_data = generate_random_4d<input2_type>(p.b2_num, p.f2_num, x2_size, y2_size, p.range2[0], p.range2[1], p.range2[2]);
         auto input2_data_bfyx = flatten_4d(format::bfyx, input2_data);
-        auto input2_mem = memory::allocate(engine, { p.allocate2_type, format::bfyx, input2_size });
+        auto input2_mem = engine.allocate_memory({ p.allocate2_type, format::bfyx, input2_size });
         set_values(input2_mem, input2_data_bfyx);
 
         std::vector<output_type> out_data(p.b_out_num * p.f_out_num * p.m_size * p.n_size);
@@ -3505,9 +3499,9 @@ class GemmBaseTest : public ::testing::TestWithParam<gemm_params> {
         }
 
         topology topology;
-        topology.add(input_layout("input0", input0_mem.get_layout()));
-        topology.add(input_layout("input1", input1_mem.get_layout()));
-        topology.add(input_layout("input2", input2_mem.get_layout()));
+        topology.add(input_layout("input0", input0_mem->get_layout()));
+        topology.add(input_layout("input1", input1_mem->get_layout()));
+        topology.add(input_layout("input2", input2_mem->get_layout()));
         topology.add(gemm("gemm_bfyx", { "input0", "input1", "input2" }, p.output_type, p.transpose_input0, p.transpose_input1, p.alpha, p.beta));
         topology.add(reorder("reorder_bfyx", "gemm_bfyx", format::bfyx, data_types::f32));
 
@@ -3521,7 +3515,7 @@ class GemmBaseTest : public ::testing::TestWithParam<gemm_params> {
         network.set_input_data("input2", input2_mem);
         auto outputs = network.execute();
         auto output = outputs.at("reorder_bfyx").get_memory();
-        auto output_ptr = output.pointer<float>();
+        cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
         EXPECT_EQ(output_ptr.size(), (size_t)(p.b_out_num * p.f_out_num * p.m_size * p.n_size));
         if (sizeof(input0_type) == 1) {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/loop_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/loop_gpu_test.cpp
index 9a61d29f8e2652..37c0a0309497ac 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/loop_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/loop_gpu_test.cpp
@@ -3,17 +3,18 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/eltwise.hpp"
-#include <api/network.hpp>
-#include <api/engine.hpp>
+
 #include "test_utils/test_utils.h"
-#include <api/data.hpp>
-#include <api/loop.hpp>
-#include <api/mutable_data.hpp>
-#include <api/data.hpp>
+
+#include <cldnn/runtime/memory.hpp>
+#include <cldnn/runtime/engine.hpp>
+#include <cldnn/graph/network.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include "cldnn/primitives/eltwise.hpp"
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/loop.hpp>
+#include <cldnn/primitives/mutable_data.hpp>
+#include <cldnn/primitives/data.hpp>
 
 #include <cassert>
 #include <cmath>
@@ -26,13 +27,13 @@ using namespace testing;
 
 TEST(loop_gpu, basic_no_concat)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 4, 5 } });
-    auto operand_mem = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 4, 5 } });
-    auto trip_count_mem = memory::allocate(engine, { data_types::i32, format::bfyx, { 1, 1, 1, 1 } });
-    auto initial_condition_mem = memory::allocate(engine, { data_types::i32, format::bfyx, { 1, 1, 1, 1 } });
-    auto num_iteration_mem = memory::allocate(engine, { data_types::i32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 4, 5 } });
+    auto operand_mem = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 4, 5 } });
+    auto trip_count_mem = engine.allocate_memory({ data_types::i32, format::bfyx, { 1, 1, 1, 1 } });
+    auto initial_condition_mem = engine.allocate_memory({ data_types::i32, format::bfyx, { 1, 1, 1, 1 } });
+    auto num_iteration_mem = engine.allocate_memory({ data_types::i32, format::bfyx, { 1, 1, 1, 1 } });
 
     std::vector<float> input_data{
         1.0f,  2.0f, -15.f,  3.0f, 4.0f, -15.f, 5.0f,  6.0f, -15.f, 7.0f,
@@ -65,9 +66,9 @@ TEST(loop_gpu, basic_no_concat)
     };
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
-        input_layout("trip_count", trip_count_mem.get_layout()),
-        input_layout("initial_condition", initial_condition_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
+        input_layout("trip_count", trip_count_mem->get_layout()),
+        input_layout("initial_condition", initial_condition_mem->get_layout()),
         mutable_data("num_iteration", num_iteration_mem),
         loop("loop", {"input"}, body,
              "trip_count", "initial_condition", "num_iteration",
@@ -82,33 +83,33 @@ TEST(loop_gpu, basic_no_concat)
     auto outputs = network.execute();
     EXPECT_EQ(outputs.size(), 1);
     auto output = outputs.begin()->second.get_memory();
-    auto output_layout = output.get_layout();
+    auto output_layout = output->get_layout();
 
     EXPECT_EQ(output_layout.size.batch[0], 1);
     EXPECT_EQ(output_layout.size.feature[0], 1);
     EXPECT_EQ(output_layout.size.spatial[0], 4);
     EXPECT_EQ(output_layout.size.spatial[1], 5);
 
-    auto ptr = num_iteration_mem.pointer<int32_t>();
+    mem_lock<int32_t> ptr{num_iteration_mem, get_test_stream()};
     EXPECT_EQ(ptr[0], trip_count);
 
     // value check
-    auto output_ptr = output.pointer<float>();
+    mem_lock<float> output_ptr{output, get_test_stream()};
     EXPECT_EQ(output_ptr.size(), input_data.size());
-    for (size_t i=0, iend = input_data.size(); i<iend; ++i) {
+    for (size_t i = 0, iend = input_data.size(); i<iend; ++i) {
         EXPECT_FLOAT_EQ(output_ptr[i], input_data[i] + eltwise_operand[i] * trip_count);
     }
 }
 
 TEST(loop_gpu, basic_concat)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 4, 5 } }); // b,f,x,y
-    auto operand_mem = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 4, 1 } }); // b,f,x,y
-    auto trip_count_mem = memory::allocate(engine, { data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
-    auto initial_condition_mem = memory::allocate(engine, { data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
-    auto num_iteration_mem = memory::allocate(engine, { data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 4, 5 } }); // b,f,x,y
+    auto operand_mem = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 4, 1 } }); // b,f,x,y
+    auto trip_count_mem = engine.allocate_memory({ data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
+    auto initial_condition_mem = engine.allocate_memory({ data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
+    auto num_iteration_mem = engine.allocate_memory({ data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
 
     std::vector<float> input_data{
         1.0f,  2.0f, -15.f,  3.0f, 4.0f, -15.f, 5.0f,  6.0f, -15.f, 7.0f,
@@ -128,7 +129,7 @@ TEST(loop_gpu, basic_concat)
     set_values(initial_condition_mem, {initial_condition});
 
     topology body(
-        input_layout("input", operand_mem.get_layout()),
+        input_layout("input", operand_mem->get_layout()),
         data("eltwise_operand", operand_mem),
         eltwise("eltwise", "input", "eltwise_operand", eltwise_mode::sum)
     );
@@ -139,9 +140,9 @@ TEST(loop_gpu, basic_concat)
     std::vector<loop::backedge_mapping> back_edges {};
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
-        input_layout("trip_count", trip_count_mem.get_layout()),
-        input_layout("initial_condition", initial_condition_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
+        input_layout("trip_count", trip_count_mem->get_layout()),
+        input_layout("initial_condition", initial_condition_mem->get_layout()),
         mutable_data("num_iteration", num_iteration_mem),
         loop("loop", {"input"}, body,
              "trip_count", "initial_condition", "num_iteration",
@@ -156,19 +157,19 @@ TEST(loop_gpu, basic_concat)
     auto outputs = network.execute();
     EXPECT_EQ(outputs.size(), 1);
     auto output = outputs.begin()->second.get_memory();
-    auto output_layout = output.get_layout();
+    auto output_layout = output->get_layout();
 
     EXPECT_EQ(output_layout.size.batch[0], 1);
     EXPECT_EQ(output_layout.size.feature[0], 1);
     EXPECT_EQ(output_layout.size.spatial[0], 4);
     EXPECT_EQ(output_layout.size.spatial[1], 5);
 
-    auto ptr = num_iteration_mem.pointer<int32_t>();
+    mem_lock<int32_t> ptr{num_iteration_mem, get_test_stream()};
     const int32_t actual_iterations = ptr[0];
     EXPECT_EQ(actual_iterations, trip_count);
 
     // value check
-    auto output_ptr = output.pointer<float>();
+    mem_lock<float> output_ptr{output, get_test_stream()};
     for (size_t i=0, iend = input_data.size(); i<iend; ++i) {
         const size_t j = i % eltwise_operand.size();
         float expected = input_data[i] + eltwise_operand[j];
@@ -178,16 +179,16 @@ TEST(loop_gpu, basic_concat)
 
 TEST(loop_gpu, basic_concat_nested)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 4, 5 } }); // b,f,x,y
-    auto trip_count_mem = memory::allocate(engine, { data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
-    auto initial_condition_mem = memory::allocate(engine, { data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
-    auto num_iteration_mem = memory::allocate(engine, { data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
-    auto inner_trip_count_mem = memory::allocate(engine, { data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
-    auto inner_initial_condition_mem = memory::allocate(engine, { data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
-    auto inner_num_iteration_mem = memory::allocate(engine, { data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
-    auto inner_operand_mem = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 4, 1 } }); // b,f,x,y
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 4, 5 } }); // b,f,x,y
+    auto trip_count_mem = engine.allocate_memory({ data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
+    auto initial_condition_mem = engine.allocate_memory({ data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
+    auto num_iteration_mem = engine.allocate_memory({ data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
+    auto inner_trip_count_mem = engine.allocate_memory({ data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
+    auto inner_initial_condition_mem = engine.allocate_memory({ data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
+    auto inner_num_iteration_mem = engine.allocate_memory({ data_types::i64, format::bfyx, { 1, 1, 1, 1 } });
+    auto inner_operand_mem = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 4, 1 } }); // b,f,x,y
 
     /////////////////////////////////
     // set data
@@ -220,7 +221,7 @@ TEST(loop_gpu, basic_concat_nested)
     // set inner loop body
     /////////////////////////////////
     topology inner_loop_body(
-        input_layout("inner_input", input_mem.get_layout()),
+        input_layout("inner_input", input_mem->get_layout()),
         data("inner_eltwise_operand", inner_operand_mem),
         eltwise("inner_eltwise", "inner_input", "inner_eltwise_operand", eltwise_mode::sum)
     );
@@ -232,9 +233,9 @@ TEST(loop_gpu, basic_concat_nested)
     // set outer loop body
     /////////////////////////////////
     topology outer_loop_body(
-        input_layout("inner_input", input_mem.get_layout()),
-        input_layout("trip_count", inner_trip_count_mem.get_layout()),
-        input_layout("initial_condition", inner_initial_condition_mem.get_layout()),
+        input_layout("inner_input", input_mem->get_layout()),
+        input_layout("trip_count", inner_trip_count_mem->get_layout()),
+        input_layout("initial_condition", inner_initial_condition_mem->get_layout()),
         mutable_data("inner_num_iteration", inner_num_iteration_mem),
         loop("inner_loop", {"inner_input", "trip_count", "initial_condition"},
             inner_loop_body, "trip_count", "initial_condition", "inner_num_iteration",
@@ -254,12 +255,12 @@ TEST(loop_gpu, basic_concat_nested)
     // set main topology
     /////////////////////////////////
     topology main_topology(
-        input_layout("input", input_mem.get_layout()),
-        input_layout("trip_count", trip_count_mem.get_layout()),
-        input_layout("initial_condition", initial_condition_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
+        input_layout("trip_count", trip_count_mem->get_layout()),
+        input_layout("initial_condition", initial_condition_mem->get_layout()),
         mutable_data("num_iteration", num_iteration_mem),
-        input_layout("inner_trip_count", inner_trip_count_mem.get_layout()),
-        input_layout("inner_initial_condition", inner_initial_condition_mem.get_layout()),
+        input_layout("inner_trip_count", inner_trip_count_mem->get_layout()),
+        input_layout("inner_initial_condition", inner_initial_condition_mem->get_layout()),
         loop("loop", {"input", "inner_trip_count", "inner_initial_condition"},
             outer_loop_body, "trip_count", "initial_condition", "num_iteration",
             outer_input_primitive_maps, outer_output_primitive_maps, outer_back_edges, outer_trip_count)
@@ -278,7 +279,7 @@ TEST(loop_gpu, basic_concat_nested)
     auto outputs = network.execute();
     EXPECT_EQ(outputs.size(), 1);
     auto output = outputs.begin()->second.get_memory();
-    auto output_layout = output.get_layout();
+    auto output_layout = output->get_layout();
 
     /////////////////////////////////
     // calculate expected output
@@ -303,16 +304,16 @@ TEST(loop_gpu, basic_concat_nested)
     EXPECT_EQ(output_layout.size.spatial[1], 5);
 
     // check trip count = actual iteration
-    auto inner_num_iteration_ptr = inner_num_iteration_mem.pointer<int64_t>();
+    mem_lock<int64_t> inner_num_iteration_ptr{inner_num_iteration_mem, get_test_stream()};
     int64_t inner_actual_iterations = inner_num_iteration_ptr[0];
     EXPECT_EQ(inner_actual_iterations, inner_trip_count);
-    auto num_iteration_ptr = num_iteration_mem.pointer<int64_t>();
+    mem_lock<int64_t> num_iteration_ptr{num_iteration_mem, get_test_stream()};
     int64_t actual_iterations = num_iteration_ptr[0];
     EXPECT_EQ(actual_iterations, outer_trip_count);
 
     // check output values
     EXPECT_EQ(output_layout.count(), expected.size());
-    auto output_ptr = output.pointer<float>();
+    mem_lock<float> output_ptr{output, get_test_stream()};
     for (size_t i=0 ;i<output_layout.count(); ++i) {
         EXPECT_FLOAT_EQ(output_ptr[i], expected.at(i));
     }
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/lrn_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/lrn_gpu_test.cpp
index 8d0c39a9bad475..119a6fb1c52385 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/lrn_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/lrn_gpu_test.cpp
@@ -3,16 +3,12 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/data.hpp>
-#include <api/lrn.hpp>
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/network.hpp>
-#include <api/topology.hpp>
+#include "test_utils.h"
 
-#include <tests/test_utils/test_utils.h>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/lrn.hpp>
+#include <cldnn/primitives/input_layout.hpp>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -20,14 +16,14 @@ using namespace ::tests;
 TEST(lrn_fp32_gpu, basic) {
     //  input : 1x16x1x1
     //  Output : 1x16x1x1
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const size_t b = 1;
     const size_t f = 16;
     const size_t y = 1;
     const size_t x = 1;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::b_fs_yx_fsv16, { b, f, x, y } });
+    auto input = engine.allocate_memory({ data_types::f32, format::b_fs_yx_fsv16, { b, f, x, y } });
     std::vector<float> inputVals(b * f * y * x);
     std::generate(inputVals.begin(), inputVals.end(), []() {
         static float n = 0;
@@ -37,7 +33,7 @@ TEST(lrn_fp32_gpu, basic) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     uint32_t size = 2;
     float k = 0.5f;
     float alpha = 9.9e-05f;
@@ -51,7 +47,7 @@ TEST(lrn_fp32_gpu, basic) {
     auto outputs = network.execute();
 
     auto output = outputs.at("lrn").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.99901f, 3.99486f, 5.98519f,
@@ -69,14 +65,14 @@ TEST(lrn_fp32_gpu, basic) {
 TEST(lrn_fp32_gpu, basic2) {
     //  input : 1x16x1x1
     //  Output : 1x16x1x1
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const size_t b = 1;
     const size_t f = 16;
     const size_t y = 1;
     const size_t x = 1;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::b_fs_yx_fsv16, { b, f, x, y } });
+    auto input = engine.allocate_memory({ data_types::f32, format::b_fs_yx_fsv16, { b, f, x, y } });
     std::vector<float> inputVals(b * f * y * x);
     std::generate(inputVals.begin(), inputVals.end(), []() {
         static float n = 0;
@@ -86,7 +82,7 @@ TEST(lrn_fp32_gpu, basic2) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     uint32_t size = 5;
     float k = 0.5f;
     float alpha = 9.9e-05f;
@@ -100,7 +96,7 @@ TEST(lrn_fp32_gpu, basic2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("lrn").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.99889f, 3.99525f, 5.98696f,
@@ -118,14 +114,14 @@ TEST(lrn_fp32_gpu, basic2) {
 TEST(lrn_fp16_gpu, basic1) {
     //  input : 1x16x1x1
     //  Output : 1x16x1x1
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const size_t b = 1;
     const size_t f = 16;
     const size_t y = 1;
     const size_t x = 1;
 
-    auto input = memory::allocate(engine, { data_types::f16, format::b_fs_yx_fsv16, { b, f, x, y } });
+    auto input = engine.allocate_memory({ data_types::f16, format::b_fs_yx_fsv16, { b, f, x, y } });
     std::vector<half_t> inputVals(b * f * y * x);
     std::generate(inputVals.begin(), inputVals.end(), []() {
         static float n = 0;
@@ -135,7 +131,7 @@ TEST(lrn_fp16_gpu, basic1) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     uint32_t size = 5;
     float k = 0.5f;
     float alpha = 9.9e-05f;
@@ -149,7 +145,7 @@ TEST(lrn_fp16_gpu, basic1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("lrn").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.99889f, 3.99525f, 5.98696f,
@@ -167,14 +163,14 @@ TEST(lrn_fp16_gpu, basic1) {
 TEST(lrn_fp32_gpu, basic3) {
     //  input : 2x16x4x4
     //  Output : 2x16x4x4
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const size_t b = 2;
     const size_t f = 16;
     const size_t y = 4;
     const size_t x = 4;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::b_fs_yx_fsv16, { b, f, x, y } });
+    auto input = engine.allocate_memory({ data_types::f32, format::b_fs_yx_fsv16, { b, f, x, y } });
     std::vector<float> inputVals(b * f * y * x);
     std::generate(inputVals.begin(), inputVals.end(), []() {
         static float n = 0;
@@ -184,7 +180,7 @@ TEST(lrn_fp32_gpu, basic3) {
     set_values(input, inputVals);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     uint32_t size = 5;
     float k = 1.f;
     float alpha = 9.89999971e-05f;
@@ -198,7 +194,7 @@ TEST(lrn_fp32_gpu, basic3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("lrn").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f,      0.999792f, 1.99911f, 2.99755f, 3.99466f, 4.99f,    5.98313f, 6.97361f, 7.96102f, 8.94493f, 9.92493f,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_dynamic_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_dynamic_gpu_test.cpp
index 04100ead3c7b9d..c8a037bfbf157e 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_dynamic_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_dynamic_gpu_test.cpp
@@ -1,34 +1,30 @@
-// Copyright (C) 2018-2021 Intel Corporation
+// Copyright (C) 2018-2021 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include "api/mutable_data.hpp"
-#include "api/input_layout.hpp"
-#include "api/lstm.hpp"
-#include "api/lstm_dynamic.hpp"
-#include "api/reorder.hpp"
-#include "api_extension/lstm_dynamic_input.hpp"
-#include "api_extension/lstm_dynamic_timeloop.hpp"
-#include "api/topology.hpp"
-#include "api/tensor.hpp"
-#include "api/network.hpp"
-#include "api/engine.hpp"
-#include "test_utils/test_utils.h"
-#include "api/data.hpp"
-#include "instrumentation.h"
-#include <test_utils/float16.h>
+#include "test_utils.h"
+
+#include <cldnn/primitives/mutable_data.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/lstm.hpp>
+#include <cldnn/primitives/lstm_dynamic.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/lstm_dynamic_input.hpp>
+#include <cldnn/primitives/lstm_dynamic_timeloop.hpp>
+
 #include <chrono>
 #include <sstream>
 #include <iomanip>
 
+#ifndef __clang__
 #pragma warning( disable : 4503 )
+#endif
 
 #define MEASURE_PERF false
 #define MEASURE_LOOP 50
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 namespace {
     float sigmoid(float x) {
@@ -217,29 +213,28 @@ struct lstm_dynamic_input_layer_test : public ::testing::Test
         VF<T> ref_weights_vec = flatten_4d<T>(cldnn::format::bfyx, ref_weights);
         VF<T> ref_bias_vec = flatten_4d<T>(cldnn::format::bfyx, ref_bias);
 
-        const auto& engine = get_test_engine();
+        auto& engine = get_test_engine();
         VF<T> ref_dynamic_length;
         for (auto& v : dynamic_lengths)
             ref_dynamic_length.push_back((T)v);
         constexpr auto dt = std::is_same<T, float>::value ? data_types::f32 : data_types::f16;
 
-        memory input_mem = memory::allocate(engine, { dt, format::bfyx,{ batch_size, max_sequence_len, input_size, direction } });
+        auto input_mem = engine.allocate_memory({ dt, format::bfyx,{ batch_size, max_sequence_len, input_size, direction } });
         set_values<T>(input_mem, ref_input_vec);
-        memory weights_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
+        auto weights_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
         set_values<T>(weights_mem, ref_weights_vec);
-        memory dynamic_length_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, batch_size, 1 } });
+        auto dynamic_length_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, batch_size, 1 } });
         set_values<T>(dynamic_length_mem, ref_dynamic_length);
-        memory bias_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, 4 * hidden_size, direction } });
+        auto bias_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, 4 * hidden_size, direction } });
         set_values(bias_mem, ref_bias_vec);
 
         topology topology;
-        topology.add(input_layout("input", input_mem.get_layout()));
-        topology.add(input_layout("dyn_len", dynamic_length_mem.get_layout()));
+        topology.add(input_layout("input", input_mem->get_layout()));
+        topology.add(input_layout("dyn_len", dynamic_length_mem->get_layout()));
         topology.add(data("weights", weights_mem));
 
         std::string bias_id = "";
-        if (has_bias)
-        {
+        if (has_bias) {
             bias_id = "bias";
             topology.add(data(bias_id, bias_mem));
         }
@@ -277,8 +272,9 @@ struct lstm_dynamic_input_layer_test : public ::testing::Test
 
         auto outputs = network.execute();
         auto out = outputs.at("dynamic_lstm_input");
-        auto out_tensor = out.get_memory().get_layout().size;
-        auto out_ptr = out.get_memory().pointer<T>();
+        auto out_tensor = out.get_memory()->get_layout().size;
+        cldnn::mem_lock<T> out_ptr(out.get_memory(), get_test_stream());
+
 
         auto output_ref =  dynamic_lstm::lstm_dynamic_input_ref(ref_input, ref_weights, ref_bias, dynamic_lengths, max_sequence_len, has_bias, direction);
 
@@ -331,31 +327,31 @@ struct lstm_dynamic_single_layer_test : public ::testing::Test
         VF<T> ref_hidden_vec = flatten_4d<T>(cldnn::format::bfyx, ref_hidden);
         VF<T> ref_cell_vec = flatten_4d<T>(cldnn::format::bfyx, ref_cell);
 
-        const auto& engine = get_test_engine();
+        auto& engine = get_test_engine();
         constexpr auto dt = std::is_same<T, float>::value ? data_types::f32 : data_types::f16;
         VF<T> ref_dynamic_length;
         for (auto& v : dynamic_lengths)
             ref_dynamic_length.push_back((T)v);
 
-        memory input_mem = memory::allocate(engine, { dt, format::bfyx,{ batch_size, max_sequence_len, input_size, direction } });
+        auto input_mem = engine.allocate_memory({ dt, format::bfyx,{ batch_size, max_sequence_len, input_size, direction } });
         set_values<T>(input_mem, ref_input_vec);
 
-        memory weights_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
+        auto weights_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
         set_values<T>(weights_mem, ref_weights_vec);
-        memory recurrent_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, hidden_size, 4 * hidden_size } });
+        auto recurrent_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, hidden_size, 4 * hidden_size } });
         set_values<T>(recurrent_mem, ref_recurrent_vec);
-        memory dynamic_length_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, batch_size, 1 } });
+        auto dynamic_length_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, batch_size, 1 } });
         set_values<T>(dynamic_length_mem, ref_dynamic_length);
-        memory bias_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, 4 * hidden_size, direction } });
+        auto bias_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, 4 * hidden_size, direction } });
         set_values(bias_mem, ref_bias_vec);
-        memory initial_hidden_mem = memory::allocate(engine, { dt, format::bfyx,{ batch_size, 1, hidden_size, direction } });
+        auto initial_hidden_mem = engine.allocate_memory({ dt, format::bfyx,{ batch_size, 1, hidden_size, direction } });
         set_values<T>(initial_hidden_mem, ref_hidden_vec);
-        memory initial_cell_mem = memory::allocate(engine, { dt, format::bfyx,{ batch_size, 1, hidden_size, direction } });
+        auto initial_cell_mem = engine.allocate_memory({ dt, format::bfyx,{ batch_size, 1, hidden_size, direction } });
         set_values<T>(initial_cell_mem, ref_cell_vec);
 
         topology topology;
-        topology.add(input_layout("input", input_mem.get_layout()));
-        topology.add(input_layout("dyn_len", dynamic_length_mem.get_layout()));
+        topology.add(input_layout("input", input_mem->get_layout()));
+        topology.add(input_layout("dyn_len", dynamic_length_mem->get_layout()));
         topology.add(data("weights", weights_mem));
         topology.add(data("recurrent", recurrent_mem));
 
@@ -381,7 +377,9 @@ struct lstm_dynamic_single_layer_test : public ::testing::Test
         }
 
         std::string last_hidden_state = "";
-        memory last_hidden_mem = memory::allocate(engine, { dt, format::bfyx,{ batch_size, 1, hidden_size, direction } });
+        auto last_hidden_mem = engine.allocate_memory({ dt, format::bfyx,{ batch_size, 1, hidden_size, direction } });
+        last_hidden_mem->fill(get_test_stream());
+        get_test_stream().finish();
         if (has_last_hidden_state)
         {
             last_hidden_state = "last_hidden_state";
@@ -389,7 +387,9 @@ struct lstm_dynamic_single_layer_test : public ::testing::Test
         }
 
         std::string last_cell_state = "";
-        memory last_cell_mem = memory::allocate(engine, { dt, format::bfyx,{ batch_size, 1, hidden_size, direction } });
+        auto last_cell_mem = engine.allocate_memory({ dt, format::bfyx,{ batch_size, 1, hidden_size, direction } });
+        last_cell_mem->fill(get_test_stream());
+        get_test_stream().finish();
         if (has_last_cell_state)
         {
             last_cell_state = "last_cell_state";
@@ -436,10 +436,11 @@ struct lstm_dynamic_single_layer_test : public ::testing::Test
             clip_threshold, input_forget);
         auto real_outs = network.execute();
         auto out = real_outs.at("dynamic_lstm");
-        auto out_tensor = out.get_memory().get_layout().size;
-        auto out_ptr = out.get_memory().pointer<T>();
-        auto last_hidden_ptr = last_hidden_mem.pointer<T>();
-        auto last_cell_ptr = last_cell_mem.pointer<T>();
+        auto out_tensor = out.get_memory()->get_layout().size;
+
+        cldnn::mem_lock<T> out_ptr(out.get_memory(), get_test_stream());
+        cldnn::mem_lock<T> last_hidden_ptr(last_hidden_mem, get_test_stream());
+        cldnn::mem_lock<T> last_cell_ptr(last_cell_mem, get_test_stream());
         size_t i = 0, i_lh = 0, i_lc = 0;
         for (auto b = 0; b < out_tensor.batch[0]; b++)
         {
@@ -474,8 +475,8 @@ struct lstm_dynamic_single_layer_test : public ::testing::Test
                         //check optional last hidden state output
                         if(has_last_hidden_state && len == dynamic_lengths[b] - 1)
                         {
-                            auto ratio = (float)ref_output_hidden[b][len][dir][x] / (float)last_hidden_ptr[i_lh++];                 
-                            EXPECT_TRUE(std::abs((1.0f - ratio) < 0.01f))
+                            auto ratio = (float)ref_output_hidden[b][len][dir][x] / (float)last_hidden_ptr[i_lh++];
+                            EXPECT_TRUE(std::abs(1.0f - ratio) < 0.01f)
                             << "check has_last_hidden_state with ratio: " << ratio << ", "
                                 << "b:" << b << ", "
                                 << "len:" << len << ", "
@@ -499,7 +500,7 @@ struct lstm_dynamic_single_layer_test : public ::testing::Test
                         if(has_last_cell_state && len == dynamic_lengths[b] - 1)
                         {
                             auto ratio = (float)ref_output_cell[b][len][dir][x] / (float)last_cell_ptr[i_lc++];
-                            EXPECT_TRUE(std::abs((1.0f - ratio) < 0.01f))
+                            EXPECT_TRUE(std::abs(1.0f - ratio) < 0.01f)
                                 << "check has_last_cell_state with ratio: " << ratio << ", "
                                 << "b:" << b << ", "
                                 << "len:" << len << ", "
@@ -869,17 +870,17 @@ TEST(lstm_dynamic_negative, wrong_weights_size) {
 
     auto batch_size = 1, max_sequence_len = 10, input_size = 16, hidden_size = 32, direction = 1;
     auto wrong_value = 50;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     cldnn::data_types dt = cldnn::data_types::f32;
-    memory input_mem = memory::allocate(engine, { dt, format::bfyx, { batch_size, max_sequence_len, input_size, 1 } });
-    memory weights_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, input_size, wrong_value } });
-    memory recurrent_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, hidden_size, 4 * hidden_size } });
-    memory dynamic_length_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, batch_size, 1 } });
-    memory bias_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, 4 * hidden_size, 1 } });
+    auto input_mem = engine.allocate_memory({ dt, format::bfyx, { batch_size, max_sequence_len, input_size, 1 } });
+    auto weights_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, input_size, wrong_value } });
+    auto recurrent_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, hidden_size, 4 * hidden_size } });
+    auto dynamic_length_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, batch_size, 1 } });
+    auto bias_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, 4 * hidden_size, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input_mem.get_layout()));
-    topology.add(input_layout("dyn_len", dynamic_length_mem.get_layout()));
+    topology.add(input_layout("input", input_mem->get_layout()));
+    topology.add(input_layout("dyn_len", dynamic_length_mem->get_layout()));
     topology.add(data("weights", weights_mem));
     topology.add(data("recurrent", recurrent_mem));
     topology.add(lstm_dynamic("dynamic_lstm",
@@ -894,17 +895,17 @@ TEST(lstm_dynamic_negative, wrong_recurrent_size_0) {
 
     auto batch_size = 1, max_sequence_len = 10, input_size = 16, hidden_size = 32, direction = 1;
     auto wrong_value = 50;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     cldnn::data_types dt = cldnn::data_types::f32;
-    memory input_mem = memory::allocate(engine, { dt, format::bfyx,{ batch_size, max_sequence_len, input_size, 1 } });
-    memory weights_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
-    memory recurrent_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, wrong_value, 4 * hidden_size } });
-    memory dynamic_length_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, batch_size, 1 } });
-    memory bias_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, 4 * hidden_size, 1 } });
+    auto input_mem = engine.allocate_memory({ dt, format::bfyx,{ batch_size, max_sequence_len, input_size, 1 } });
+    auto weights_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
+    auto recurrent_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, wrong_value, 4 * hidden_size } });
+    auto dynamic_length_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, batch_size, 1 } });
+    auto bias_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, 4 * hidden_size, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input_mem.get_layout()));
-    topology.add(input_layout("dyn_len", dynamic_length_mem.get_layout()));
+    topology.add(input_layout("input", input_mem->get_layout()));
+    topology.add(input_layout("dyn_len", dynamic_length_mem->get_layout()));
     topology.add(data("weights", weights_mem));
     topology.add(data("recurrent", recurrent_mem));
     topology.add(lstm_dynamic("dynamic_lstm",
@@ -919,17 +920,17 @@ TEST(lstm_dynamic_negative, wrong_recurrent_size_1) {
 
     auto batch_size = 1, max_sequence_len = 10, input_size = 16, hidden_size = 32, direction = 1;
     auto wrong_value = 50;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     cldnn::data_types dt = cldnn::data_types::f32;
-    memory input_mem = memory::allocate(engine, { dt, format::bfyx,{ batch_size, max_sequence_len, input_size, 1 } });
-    memory weights_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
-    memory recurrent_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, wrong_value, 4 * hidden_size } });
-    memory dynamic_length_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, batch_size, 1 } });
-    memory bias_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, 4 * hidden_size, 1 } });
+    auto input_mem = engine.allocate_memory({ dt, format::bfyx,{ batch_size, max_sequence_len, input_size, 1 } });
+    auto weights_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
+    auto recurrent_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, wrong_value, 4 * hidden_size } });
+    auto dynamic_length_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, batch_size, 1 } });
+    auto bias_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, 4 * hidden_size, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input_mem.get_layout()));
-    topology.add(input_layout("dyn_len", dynamic_length_mem.get_layout()));
+    topology.add(input_layout("input", input_mem->get_layout()));
+    topology.add(input_layout("dyn_len", dynamic_length_mem->get_layout()));
     topology.add(data("weights", weights_mem));
     topology.add(data("recurrent", recurrent_mem));
     topology.add(lstm_dynamic("dynamic_lstm",
@@ -944,17 +945,17 @@ TEST(lstm_dynamic_negative, wrong_dynamic_length_size_0) {
 
     auto batch_size = 1, max_sequence_len = 10, input_size = 16, hidden_size = 32, direction = 1;
     auto wrong_value = 50;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     cldnn::data_types dt = cldnn::data_types::f32;
-    memory input_mem = memory::allocate(engine, { dt, format::bfyx,{ batch_size, max_sequence_len, input_size, 1 } });
-    memory weights_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
-    memory recurrent_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, hidden_size, 4 * hidden_size } });
-    memory dynamic_length_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, wrong_value, 1 } });
-    memory bias_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, 4 * hidden_size, 1 } });
+    auto input_mem = engine.allocate_memory({ dt, format::bfyx,{ batch_size, max_sequence_len, input_size, 1 } });
+    auto weights_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
+    auto recurrent_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, hidden_size, 4 * hidden_size } });
+    auto dynamic_length_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, wrong_value, 1 } });
+    auto bias_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, 4 * hidden_size, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input_mem.get_layout()));
-    topology.add(input_layout("dyn_len", dynamic_length_mem.get_layout()));
+    topology.add(input_layout("input", input_mem->get_layout()));
+    topology.add(input_layout("dyn_len", dynamic_length_mem->get_layout()));
     topology.add(data("weights", weights_mem));
     topology.add(data("recurrent", recurrent_mem));
     topology.add(lstm_dynamic("dynamic_lstm",
@@ -969,17 +970,17 @@ TEST(lstm_dynamic_negative, wrong_dynamic_length_size_1) {
 
     auto batch_size = 50, max_sequence_len = 10, input_size = 16, hidden_size = 32, direction = 1;
     auto wrong_value = 2;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     cldnn::data_types dt = cldnn::data_types::f32;
-    memory input_mem = memory::allocate(engine, { dt, format::bfyx,{ batch_size, max_sequence_len, input_size, 1 } });
-    memory weights_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
-    memory recurrent_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, direction, hidden_size, 4 * hidden_size } });
-    memory dynamic_length_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, wrong_value, 1 } });
-    memory bias_mem = memory::allocate(engine, { dt, format::bfyx,{ 1, 1, 4 * hidden_size, 1 } });
+    auto input_mem = engine.allocate_memory({ dt, format::bfyx,{ batch_size, max_sequence_len, input_size, 1 } });
+    auto weights_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, input_size, 4 * hidden_size } });
+    auto recurrent_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, direction, hidden_size, 4 * hidden_size } });
+    auto dynamic_length_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, wrong_value, 1 } });
+    auto bias_mem = engine.allocate_memory({ dt, format::bfyx,{ 1, 1, 4 * hidden_size, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input_mem.get_layout()));
-    topology.add(input_layout("dyn_len", dynamic_length_mem.get_layout()));
+    topology.add(input_layout("input", input_mem->get_layout()));
+    topology.add(input_layout("dyn_len", dynamic_length_mem->get_layout()));
     topology.add(data("weights", weights_mem));
     topology.add(data("recurrent", recurrent_mem));
     topology.add(lstm_dynamic("dynamic_lstm",
@@ -989,5 +990,3 @@ TEST(lstm_dynamic_negative, wrong_dynamic_length_size_1) {
         "recurrent"));
     ASSERT_ANY_THROW(network network(engine, topology));
 }
-
-
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_gpu_test.cpp
index 866343d92974f2..d9b968d205f503 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_gpu_test.cpp
@@ -3,22 +3,16 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/lstm.hpp"
-#include <api/split.hpp>
-#include <api/crop.hpp>
-#include <api/reshape.hpp>
-#include <api/concatenation.hpp>
-#include <api/topology.hpp>
-#include <api/tensor.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/data.hpp>
-#include "instrumentation.h"
-#include <test_utils/float16.h>
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/lstm.hpp>
+#include <cldnn/primitives/split.hpp>
+#include <cldnn/primitives/crop.hpp>
+#include <cldnn/primitives/reshape.hpp>
+#include <cldnn/primitives/concatenation.hpp>
+#include <cldnn/primitives/data.hpp>
 
 #include <sstream>
 #include <iomanip>
@@ -28,7 +22,7 @@
 #endif
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 #define FERROR 1E-4
 
@@ -219,10 +213,10 @@ void generic_lstm_gemm_gpu_test(int sequence_len, int direction, int batch_size,
     VVVVF<T> ref_output = lstm_gemm_reference(ref_input, ref_weights, ref_recurrent, ref_bias, ref_hidden, 0, hasBias, hasHidden);
 
     constexpr auto dt = std::is_same<T, float>::value ? data_types::f32 : data_types::f16;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    // If the input is of fp16 type then, the memory will be allocated as such
-    if (!engine.get_info().supports_fp16)
+    // If the input is of fp16 type then, the memory::ptr will be allocated as such
+    if (!engine.get_device_info().supports_fp16)
     {
         if (dt == data_types::f16)
         {
@@ -230,11 +224,11 @@ void generic_lstm_gemm_gpu_test(int sequence_len, int direction, int batch_size,
         }
     }
 
-    memory input = memory::allocate(engine, { dt, format::bfyx,     { batch_size,   sequence_len,  input_size,      1 } });
-    memory weights = memory::allocate(engine, { dt, format::bfyx,   { 1,            direction,     input_size,      4 * hidden_size } });
-    memory recurrent = memory::allocate(engine, { dt, format::bfyx, { 1,            direction,     hidden_size,     4 * hidden_size } });
-    memory biases = memory::allocate(engine, { dt, format::bfyx,    { 1,            1,             4 * hidden_size, direction } });
-    memory hidden = memory::allocate(engine, { dt, format::bfyx,    { batch_size,   direction,     hidden_size,     1 } });
+    memory::ptr input = engine.allocate_memory({ dt, format::bfyx,     { batch_size,   sequence_len,  input_size,      1 } });
+    memory::ptr weights = engine.allocate_memory({ dt, format::bfyx,   { 1,            direction,     input_size,      4 * hidden_size } });
+    memory::ptr recurrent = engine.allocate_memory({ dt, format::bfyx, { 1,            direction,     hidden_size,     4 * hidden_size } });
+    memory::ptr biases = engine.allocate_memory({ dt, format::bfyx,    { 1,            1,             4 * hidden_size, direction } });
+    memory::ptr hidden = engine.allocate_memory({ dt, format::bfyx,    { batch_size,   direction,     hidden_size,     1 } });
 
     set_values(input, ref_input_vec);
     set_values(weights, ref_weights_vec);
@@ -243,14 +237,14 @@ void generic_lstm_gemm_gpu_test(int sequence_len, int direction, int batch_size,
     set_values(hidden, ref_hidden_vec);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights", weights));
     topology.add(data("recurrent", recurrent));
     if (hasBias) {
         topology.add(data("biases", biases));
     }
     if (hasHidden) {
-        topology.add(input_layout("hidden", hidden.get_layout()));
+        topology.add(input_layout("hidden", hidden->get_layout()));
     }
 
     topology.add(lstm_gemm("lstm_gemm", "input", "weights", "recurrent", hasBias ? "biases" : "", hasHidden ? "hidden" : ""));
@@ -265,7 +259,7 @@ void generic_lstm_gemm_gpu_test(int sequence_len, int direction, int batch_size,
     EXPECT_EQ(outputs.size(), size_t(1));
 
     auto output = outputs.begin()->second.get_memory();
-    auto output_ptr = output.pointer<T>();
+    cldnn::mem_lock<T> output_ptr(output, get_test_stream());
     int i = 0;
     for (int b = 0; b < batch_size; ++b) {
         for (int x = 0; x < 4 * hidden_size; ++x)
@@ -295,10 +289,10 @@ void generic_lstm_elt_gpu_test(int /* sequence_len */, int direction, int batch_
     // to error from 1E-4 to 1E-2
     constexpr float ferror = std::is_same<T, float>::value ? (float)1E-4 : (float)1E-2;
     constexpr auto dt = std::is_same<T, float>::value ? data_types::f32 : data_types::f16;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    // If the input is of fp16 type then, the memory will be allocated as such
-    if (!engine.get_info().supports_fp16)
+    // If the input is of fp16 type then, the memory::ptr will be allocated as such
+    if (!engine.get_device_info().supports_fp16)
     {
         if (dt == data_types::f16)
         {
@@ -306,15 +300,15 @@ void generic_lstm_elt_gpu_test(int /* sequence_len */, int direction, int batch_
         }
     }
 
-    memory tempGEMM = memory::allocate(engine, { dt, format::bfyx,{ batch_size,    direction, 4 * hidden_size, 1 } });
-    memory cell = memory::allocate(engine, { dt, format::bfyx,{ batch_size,    direction,     hidden_size, 1 } });
+    memory::ptr tempGEMM = engine.allocate_memory({ dt, format::bfyx,{ batch_size,    direction, 4 * hidden_size, 1 } });
+    memory::ptr cell = engine.allocate_memory({ dt, format::bfyx,{ batch_size,    direction,     hidden_size, 1 } });
     set_values(tempGEMM, ref_tempGEMM_vec);
     set_values(cell, ref_cell_vec);
 
     topology topology;
-    topology.add(input_layout("tempGEMM", tempGEMM.get_layout()));
+    topology.add(input_layout("tempGEMM", tempGEMM->get_layout()));
     if (hasCell) {
-        topology.add(input_layout("cell", cell.get_layout()));
+        topology.add(input_layout("cell", cell->get_layout()));
     }
     topology.add(lstm_elt("lstm_elt", "tempGEMM", hasCell ? "cell" : "", clip_threshold, input_forget));
 
@@ -328,7 +322,7 @@ void generic_lstm_elt_gpu_test(int /* sequence_len */, int direction, int batch_
     EXPECT_EQ(outputs.size(), size_t(1));
 
     auto output = outputs.begin()->second.get_memory();
-    auto output_ptr = output.pointer<T>();
+    cldnn::mem_lock<T> output_ptr(output, get_test_stream());
     for (int b = 0; b < batch_size; ++b) {
         for (int j = 0; j < 2; ++j) {
             for (int x = 0; x < hidden_size; ++x)
@@ -349,11 +343,11 @@ std::string get_string_id(size_t i) {
 // --------------- Manually constructed LSTM ----------------------------------------
 // This function manually generates an lstm node sequence by conbining lstm_gemm and lstm_elt nodes
 // it requires that the output of the lstm_elt node is croped to obtain the corresponding hidden and cell outputs
-void generate_lstm_topology(topology& t, memory& input, memory& hidden, memory& cell,
-    memory& weights, memory& recurrent, memory& biases, int sequence_len,
+void generate_lstm_topology(topology& t, memory::ptr input, memory::ptr hidden, memory::ptr cell,
+    memory::ptr weights, memory::ptr recurrent, memory::ptr biases, int sequence_len,
     bool hasBias = true, bool hasInitialHidden = true, bool hasInitialCell = true) {
-    auto hidden_size = hidden.get_layout().size;
-    t.add(input_layout("input", input.get_layout()));
+    auto hidden_size = hidden->get_layout().size;
+    t.add(input_layout("input", input->get_layout()));
     std::vector<std::pair<primitive_id, tensor>> input_ids_offsets;
     std::vector<primitive_id> output_ids_offsets;
     for (int i = 0; i < sequence_len; ++i)
@@ -372,12 +366,12 @@ void generate_lstm_topology(topology& t, memory& input, memory& hidden, memory&
     }
     if (hasInitialHidden)
     {
-        t.add(input_layout("hidden", hidden.get_layout()));
+        t.add(input_layout("hidden", hidden->get_layout()));
         hiddenStr = "hidden";
     }
     if (hasInitialCell)
     {
-        t.add(input_layout("cell", cell.get_layout()));
+        t.add(input_layout("cell", cell->get_layout()));
         cellStr = "cell";
     }
     for (int i = 0; i < sequence_len; ++i) {
@@ -423,13 +417,13 @@ void generic_lstm_custom_gpu_test(int sequence_len, int direction, int batch_siz
     lstm_reference(ref_input, ref_hidden, ref_cell, ref_weights, ref_recurrent, ref_bias, ref_output, last_hidden, last_cell,
         hasBias, hasInitialHidden, hasInitialCell);
 
-    const auto& engine = get_test_engine();
-    memory input = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ batch_size, sequence_len,  input_size,       1 } });
-    memory weights = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1,          direction,     input_size,       4 * hidden_size } });
-    memory recurrent = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1,          direction,     hidden_size,      4 * hidden_size } });
-    memory biases = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ 1,          1,             4 * hidden_size,  direction } });
-    memory hidden = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ batch_size, direction,     hidden_size,      1 } });
-    memory cell = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx,{ batch_size, direction,     hidden_size,      1 } });
+    auto& engine = get_test_engine();
+    memory::ptr input = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ batch_size, sequence_len,  input_size,       1 } });
+    memory::ptr weights = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1,          direction,     input_size,       4 * hidden_size } });
+    memory::ptr recurrent = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1,          direction,     hidden_size,      4 * hidden_size } });
+    memory::ptr biases = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ 1,          1,             4 * hidden_size,  direction } });
+    memory::ptr hidden = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ batch_size, direction,     hidden_size,      1 } });
+    memory::ptr cell = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx,{ batch_size, direction,     hidden_size,      1 } });
     set_values(input, ref_input_vec);
     set_values(weights, ref_weights_vec);
     set_values(recurrent, ref_recurrent_vec);
@@ -448,11 +442,11 @@ void generic_lstm_custom_gpu_test(int sequence_len, int direction, int batch_siz
     auto outputs = network.execute();
 
     ASSERT_EQ(outputs.size(), size_t(1));
-    size_t output_size = outputs.begin()->second.get_memory().size() / sizeof(T);
+    size_t output_size = outputs.begin()->second.get_memory()->size() / sizeof(T);
     ASSERT_EQ(output_size, size_t(hidden_size * sequence_len * batch_size * direction));
 
     auto output = outputs.begin()->second.get_memory();
-    auto output_ptr = output.pointer<T>();
+    cldnn::mem_lock<T> output_ptr(output, get_test_stream());
     int i = 0;
     for (int b = 0; b < batch_size; ++b) {
         for (int s = 0; s < sequence_len; ++s) {
@@ -526,10 +520,10 @@ void generic_lstm_gpu_test(int layers, int sequence_len, int direction, int batc
     // to error from 1E-4 to 1E-2
     constexpr float ferror = std::is_same<T, float>::value ? (float)1E-4 : (float)1E-2;
     constexpr auto dt = std::is_same<T, float>::value ? data_types::f32 : data_types::f16;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    // If the input is of fp16 type then, the memory will be allocated as such
-    if (!engine.get_info().supports_fp16)
+    // If the input is of fp16 type then, the memory::ptr will be allocated as such
+    if (!engine.get_device_info().supports_fp16)
     {
         if (dt == data_types::f16)
         {
@@ -537,29 +531,29 @@ void generic_lstm_gpu_test(int layers, int sequence_len, int direction, int batc
         }
     }
 
-    memory input = memory::allocate(engine, { dt, format::bfyx, {batch_size, sequence_len, input_size, 1} });
+    memory::ptr input = engine.allocate_memory({ dt, format::bfyx, {batch_size, sequence_len, input_size, 1} });
     set_values(input, ref_input_vec);
 
-    std::vector<memory> weights;
-    std::vector<memory> recurrent;
-    std::vector<memory> biases;
-    std::vector<memory> hidden;
-    std::vector<memory> cell;
+    std::vector<memory::ptr> weights;
+    std::vector<memory::ptr> recurrent;
+    std::vector<memory::ptr> biases;
+    std::vector<memory::ptr> hidden;
+    std::vector<memory::ptr> cell;
     for(int i = 0; i < layers; ++i) {
-        weights.push_back(memory::allocate(engine, { dt, format::bfyx, { 1, direction, i==0 ? input_size : hidden_size, 4 * hidden_size } }));
+        weights.push_back(engine.allocate_memory({ dt, format::bfyx, { 1, direction, i==0 ? input_size : hidden_size, 4 * hidden_size } }));
         set_values(weights[i], ref_weights_vec[i]);
-        recurrent.push_back(memory::allocate(engine, { dt, format::bfyx, { 1, direction, hidden_size, 4 * hidden_size } }));
+        recurrent.push_back(engine.allocate_memory({ dt, format::bfyx, { 1, direction, hidden_size, 4 * hidden_size } }));
         set_values(recurrent[i], ref_recurrent_vec[i]);
         if (hasBias) {
-            biases.push_back(memory::allocate(engine, { dt, format::bfyx, { 1, 1, 4 * hidden_size, direction } }));
+            biases.push_back(engine.allocate_memory({ dt, format::bfyx, { 1, 1, 4 * hidden_size, direction } }));
             set_values(biases[i], ref_bias_vec[i]);
         }
         if (hasInitialHidden) {
-            hidden.push_back(memory::allocate(engine, { dt, format::bfyx, { batch_size, 1, hidden_size, direction } }));
+            hidden.push_back(engine.allocate_memory({ dt, format::bfyx, { batch_size, 1, hidden_size, direction } }));
             set_values(hidden[i], ref_hidden_vec[i]);
         }
         if (hasInitialCell) {
-            cell.push_back(memory::allocate(engine, { dt, format::bfyx, { batch_size, 1, hidden_size, direction} }));
+            cell.push_back(engine.allocate_memory({ dt, format::bfyx, { batch_size, 1, hidden_size, direction} }));
             set_values(cell[i], ref_cell_vec[i]);
         }
     }
@@ -569,7 +563,7 @@ void generic_lstm_gpu_test(int layers, int sequence_len, int direction, int batc
     std::vector<primitive_id> lstm_inputs;
     std::vector<primitive_id> output_ids_offsets;
 
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     for (int i = 0; i < sequence_len; ++i) {
         input_ids_offsets.push_back({get_string_id(i), {0, i, 0, 0}});
         lstm_inputs.push_back("inputSplit:"+get_string_id(i));
@@ -588,8 +582,8 @@ void generic_lstm_gpu_test(int layers, int sequence_len, int direction, int batc
         topology.add(data(weights_id, weights[i]));
         topology.add(data(recurrent_id, recurrent[i]));
         if (hasBias) topology.add(data(biases_id, biases[i]));
-        if (hasInitialHidden) topology.add(input_layout(hidden_id, hidden[i].get_layout()));
-        if (hasInitialCell) topology.add(input_layout(cell_id, cell[i].get_layout()));
+        if (hasInitialHidden) topology.add(input_layout(hidden_id, hidden[i]->get_layout()));
+        if (hasInitialCell) topology.add(input_layout(cell_id, cell[i]->get_layout()));
         if (i == 0) {
             topology.add(lstm(lstm_id, lstm_inputs, weights_id, recurrent_id,
                             hasBias ? biases_id : "", hasInitialHidden ? hidden_id : "", hasInitialCell ? cell_id : "", "",
@@ -617,13 +611,13 @@ void generic_lstm_gpu_test(int layers, int sequence_len, int direction, int batc
     auto outputs = network.execute();
     {
         ASSERT_EQ(outputs.size(), size_t(1));
-        size_t output_size = outputs.begin()->second.get_memory().size() / sizeof(T);
+        size_t output_size = outputs.begin()->second.get_memory()->size() / sizeof(T);
         ASSERT_EQ(output_size, size_t(hidden_size * sequence_len * batch_size * direction));
 
         auto output = outputs.begin()->second.get_memory();
 
         // Get the output tensor
-        cldnn::layout output_layout = output.get_layout();
+        cldnn::layout output_layout = output->get_layout();
         cldnn::tensor output_tensor = output_layout.size;
 
         // Compare the output tensor configuration against the reference value
@@ -633,7 +627,7 @@ void generic_lstm_gpu_test(int layers, int sequence_len, int direction, int batc
         ASSERT_EQ(direction, output_tensor.spatial[1]);
         ASSERT_EQ(hidden_size, output_tensor.spatial[0]);
 
-        auto output_ptr = output.pointer<T>();
+        cldnn::mem_lock<T> output_ptr(output, get_test_stream());
         int32_t i = 0;
         for (int32_t b = 0; b < batch_size; ++b) {
             for (int32_t s = 0; s < sequence_len; ++s) {
@@ -683,14 +677,14 @@ void lstm_gpu_output_test(const lstm_output_selection& output_selection, int dir
                    last_hidden, last_cell, true, true, true,
                    (T)0, false, true);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    memory input = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, {batch_size, sequence_len, input_size, 1} });
-    memory weights = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, directions, input_size , 4 * hidden_size } });
-    memory recurrent = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, directions, hidden_size, 4 * hidden_size } });
-    memory biases = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, 1, 4 * hidden_size, directions } });
-    memory hidden = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, directions } });
-    memory cell = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, directions } });
+    memory::ptr input = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, {batch_size, sequence_len, input_size, 1} });
+    memory::ptr weights = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, directions, input_size , 4 * hidden_size } });
+    memory::ptr recurrent = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, directions, hidden_size, 4 * hidden_size } });
+    memory::ptr biases = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, 1, 4 * hidden_size, directions } });
+    memory::ptr hidden = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, directions } });
+    memory::ptr cell = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, directions } });
 
     set_values(input, ref_input_vec);
     set_values(weights, ref_weights_vec);
@@ -709,7 +703,7 @@ void lstm_gpu_output_test(const lstm_output_selection& output_selection, int dir
     std::vector<primitive_id> lstm_inputs;
     std::vector<primitive_id> output_ids_offsets;
 
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     for (int i = 0; i < sequence_len; ++i)
     {
         input_ids_offsets.push_back({get_string_id(i), {0, i, 0, 0}});
@@ -719,8 +713,8 @@ void lstm_gpu_output_test(const lstm_output_selection& output_selection, int dir
     topology.add(data("weights", weights));
     topology.add(data("recurrent", recurrent));
     topology.add(data("biases", biases));
-    topology.add(input_layout("hidden", hidden.get_layout()));
-    topology.add(input_layout("cell", cell.get_layout()));
+    topology.add(input_layout("hidden", hidden->get_layout()));
+    topology.add(input_layout("cell", cell->get_layout()));
     topology.add(lstm("lstm", lstm_inputs, "weights", "recurrent",
                       "biases", "hidden", "cell", "", 0, false,
                       { activation_func::logistic, activation_func::hyperbolic_tan, activation_func::hyperbolic_tan }, {},
@@ -752,11 +746,11 @@ void lstm_gpu_output_test(const lstm_output_selection& output_selection, int dir
 
 	for (auto itr = outputs.begin(); itr != outputs.end(); itr++)
 	{
-        auto output_tensor = itr->second.get_memory().get_layout().size;
+        auto output_tensor = itr->second.get_memory()->get_layout().size;
         primitive_id primitive_name = itr->first;
 
-		cldnn::memory output_memory = itr->second.get_memory();
-        int32_t output_size = (int32_t)(itr->second.get_memory().size() / sizeof(T));
+		cldnn::memory::ptr output_memory = itr->second.get_memory();
+        int32_t output_size = (int32_t)(itr->second.get_memory()->size() / sizeof(T));
 		cldnn::tensor ref_output_tensor;
 		VVVVF<T> ref_primitive_output;
 
@@ -793,7 +787,7 @@ void lstm_gpu_output_test(const lstm_output_selection& output_selection, int dir
 		ASSERT_EQ(ref_directions, output_tensor.spatial[1]);	// directions should match
         ASSERT_EQ(ref_hidden_size, output_tensor.spatial[0]);	// input size should match
 
-        auto output_ptr = output_memory.pointer<T>();
+        cldnn::mem_lock<T> output_ptr(output_memory, get_test_stream());
 
 		int32_t i = 0;
 		for (int32_t b = 0; b < ref_batch_size; ++b) {
@@ -846,14 +840,14 @@ void lstm_gpu_format_test(const cldnn::format& format, int directions) {
                    last_hidden, last_cell, true, true, true,
                    (T)0, false, true);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    memory input = memory::allocate(engine, { type_to_data_type<T>::value,format, {batch_size, sequence_len, input_size, 1} });
-    memory weights = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, directions, input_size , 4 * hidden_size } });
-    memory recurrent = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, directions, hidden_size, 4 * hidden_size } });
-    memory biases = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, 1, 4 * hidden_size, directions } });
-    memory hidden = memory::allocate(engine, { type_to_data_type<T>::value, format, { batch_size, 1, hidden_size, directions } });
-    memory cell = memory::allocate(engine, { type_to_data_type<T>::value, format, { batch_size, 1, hidden_size, directions } });
+    memory::ptr input = engine.allocate_memory({ type_to_data_type<T>::value,format, {batch_size, sequence_len, input_size, 1} });
+    memory::ptr weights = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, directions, input_size , 4 * hidden_size } });
+    memory::ptr recurrent = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, directions, hidden_size, 4 * hidden_size } });
+    memory::ptr biases = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, 1, 4 * hidden_size, directions } });
+    memory::ptr hidden = engine.allocate_memory({ type_to_data_type<T>::value, format, { batch_size, 1, hidden_size, directions } });
+    memory::ptr cell = engine.allocate_memory({ type_to_data_type<T>::value, format, { batch_size, 1, hidden_size, directions } });
 
     set_values(input, ref_input_vec);
     set_values(weights, ref_weights_vec);
@@ -872,7 +866,7 @@ void lstm_gpu_format_test(const cldnn::format& format, int directions) {
     std::vector<primitive_id> lstm_inputs;
     std::vector<primitive_id> output_ids_offsets;
 
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     for (int i = 0; i < sequence_len; ++i)
     {
         input_ids_offsets.push_back({get_string_id(i), {0, i, 0, 0}});
@@ -882,8 +876,8 @@ void lstm_gpu_format_test(const cldnn::format& format, int directions) {
     topology.add(data("weights", weights));
     topology.add(data("recurrent", recurrent));
     topology.add(data("biases", biases));
-    topology.add(input_layout("hidden", hidden.get_layout()));
-    topology.add(input_layout("cell", cell.get_layout()));
+    topology.add(input_layout("hidden", hidden->get_layout()));
+    topology.add(input_layout("cell", cell->get_layout()));
     topology.add(lstm("lstm"+get_string_id(0), lstm_inputs, "weights", "recurrent",
                       "biases", "hidden", "cell", "", 0, false,
                       { activation_func::logistic, activation_func::hyperbolic_tan, activation_func::hyperbolic_tan }, {},
@@ -918,11 +912,11 @@ void lstm_gpu_format_test(const cldnn::format& format, int directions) {
 
     for (auto itr = outputs.begin(); itr != outputs.end(); itr++)
     {
-        auto output_tensor = itr->second.get_memory().get_layout().size;
+        auto output_tensor = itr->second.get_memory()->get_layout().size;
         primitive_id primitive_name = itr->first;
 
-        cldnn::memory output_memory = itr->second.get_memory();
-        int32_t output_size = (int32_t)(itr->second.get_memory().size() / sizeof(T));
+        cldnn::memory::ptr output_memory = itr->second.get_memory();
+        int32_t output_size = (int32_t)(itr->second.get_memory()->size() / sizeof(T));
         cldnn::tensor ref_output_tensor;
         VVVVF<T> ref_primitive_output;
 
@@ -959,7 +953,7 @@ void lstm_gpu_format_test(const cldnn::format& format, int directions) {
         ASSERT_EQ(ref_directions, output_tensor.spatial[1]);    // directions should match
         ASSERT_EQ(ref_hidden_size, output_tensor.spatial[0]);   // input size should match
 
-        auto output_ptr = output_memory.pointer<T>();
+        cldnn::mem_lock<T> output_ptr(output_memory, get_test_stream());
 
         int32_t i = 0;
         if (format == cldnn::format::bfyx) {
@@ -1025,14 +1019,14 @@ void lstm_gpu_users_test() {
     VVVVF<T> last_hidden(batch_size, VVVF<T>(1, VVF<T>(directions, VF<T>(hidden_size))));
     VVVVF<T> last_cell(batch_size, VVVF<T>(1, VVF<T>(directions, VF<T>(hidden_size))));
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    memory input = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, {batch_size, sequence_len, input_size, 1} });
-    memory weights = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, directions, input_size , 4 * hidden_size } });
-    memory recurrent = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, directions, hidden_size, 4 * hidden_size } });
-    memory biases = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, 1, 4 * hidden_size, directions } });
-    memory hidden = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, directions } });
-    memory cell = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, directions } });
+    memory::ptr input = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, {batch_size, sequence_len, input_size, 1} });
+    memory::ptr weights = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, directions, input_size , 4 * hidden_size } });
+    memory::ptr recurrent = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, directions, hidden_size, 4 * hidden_size } });
+    memory::ptr biases = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, 1, 4 * hidden_size, directions } });
+    memory::ptr hidden = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, directions } });
+    memory::ptr cell = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, directions } });
 
     set_values(input, ref_input_vec);
     set_values(weights, ref_weights_vec);
@@ -1045,7 +1039,7 @@ void lstm_gpu_users_test() {
     std::vector<std::pair<primitive_id, tensor>> input_ids_offsets;
     std::vector<primitive_id> lstm_inputs;
 
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     for (int i = 0; i < sequence_len; ++i)
     {
         input_ids_offsets.push_back({get_string_id(i), {0, i, 0, 0}});
@@ -1055,8 +1049,8 @@ void lstm_gpu_users_test() {
     topology.add(data("weights", weights));
     topology.add(data("recurrent", recurrent));
     topology.add(data("biases", biases));
-    topology.add(input_layout("hidden", hidden.get_layout()));
-    topology.add(input_layout("cell", cell.get_layout()));
+    topology.add(input_layout("hidden", hidden->get_layout()));
+    topology.add(input_layout("cell", cell->get_layout()));
     topology.add(lstm("lstm", lstm_inputs, "weights", "recurrent",
                       "biases", "hidden", "cell", "", 0, false,
                       { activation_func::logistic, activation_func::hyperbolic_tan, activation_func::hyperbolic_tan }, {},
@@ -1074,8 +1068,8 @@ void lstm_gpu_users_test() {
 
     // check if the number of returned primitives match the expected number of output primitives
     ASSERT_EQ(size_t(1), outputs.size());
-    cldnn::memory output_memory = outputs.begin()->second.get_memory();
-    auto output_ptr = output_memory.pointer<T>();
+    cldnn::memory::ptr output_memory = outputs.begin()->second.get_memory();
+    cldnn::mem_lock<T> output_ptr(output_memory, get_test_stream());
 
     for (int32_t b = 0; b < batch_size; ++b) {
         for (int32_t s = 0; s < 1; ++s) {
@@ -1148,31 +1142,31 @@ void lstm_gpu_concatenated_input_test(int layers, int sequence_len, int directio
 			clip_threshold, input_forget, false);
 	}
 
-	const auto& engine = get_test_engine();
+	auto& engine = get_test_engine();
 
-	memory input = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, {batch_size, sequence_len, input_size, 1} });
+	memory::ptr input = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, {batch_size, sequence_len, input_size, 1} });
 	set_values(input, ref_input_vec);
 
-	std::vector<memory> weights;
-	std::vector<memory> recurrent;
-	std::vector<memory> biases;
-	std::vector<memory> hidden;
-	std::vector<memory> cell;
+	std::vector<memory::ptr> weights;
+	std::vector<memory::ptr> recurrent;
+	std::vector<memory::ptr> biases;
+	std::vector<memory::ptr> hidden;
+	std::vector<memory::ptr> cell;
 	for (int i = 0; i < layers; ++i) {
-		weights.push_back(memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, direction, i == 0 ? input_size : hidden_size, 4 * hidden_size } }));
+		weights.push_back(engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, direction, i == 0 ? input_size : hidden_size, 4 * hidden_size } }));
 		set_values(weights[i], ref_weights_vec[i]);
-		recurrent.push_back(memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, direction, hidden_size, 4 * hidden_size } }));
+		recurrent.push_back(engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, direction, hidden_size, 4 * hidden_size } }));
 		set_values(recurrent[i], ref_recurrent_vec[i]);
 		if (has_bias) {
-			biases.push_back(memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { 1, 1, 4 * hidden_size, direction } }));
+			biases.push_back(engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { 1, 1, 4 * hidden_size, direction } }));
 			set_values(biases[i], ref_bias_vec[i]);
 		}
 		if (has_initial_hidden) {
-			hidden.push_back(memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, direction } }));
+			hidden.push_back(engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, direction } }));
 			set_values(hidden[i], ref_hidden_vec[i]);
 		}
 		if (has_initial_cell) {
-			cell.push_back(memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, direction} }));
+			cell.push_back(engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, { batch_size, 1, hidden_size, direction} }));
 			set_values(cell[i], ref_cell_vec[i]);
 		}
 	}
@@ -1182,7 +1176,7 @@ void lstm_gpu_concatenated_input_test(int layers, int sequence_len, int directio
 	std::vector<primitive_id> lstm_inputs;
 	std::vector<primitive_id> output_ids_offsets;
 
-	topology.add(input_layout("input", input.get_layout()));
+	topology.add(input_layout("input", input->get_layout()));
 	cldnn::primitive_id prev_node_id;
 
     for (int i = 0; i < layers; ++i) {
@@ -1198,8 +1192,8 @@ void lstm_gpu_concatenated_input_test(int layers, int sequence_len, int directio
 		topology.add(data(weights_id, weights[i]));
 		topology.add(data(recurrent_id, recurrent[i]));
 		if (has_bias) topology.add(data(biases_id, biases[i]));
-		if (has_initial_hidden) topology.add(input_layout(hidden_id, hidden[i].get_layout()));
-		if (has_initial_cell) topology.add(input_layout(cell_id, cell[i].get_layout()));
+		if (has_initial_hidden) topology.add(input_layout(hidden_id, hidden[i]->get_layout()));
+		if (has_initial_cell) topology.add(input_layout(cell_id, cell[i]->get_layout()));
 		if (i == 0) {
             topology.add(lstm(lstm_id, { "input" }, weights_id, recurrent_id,
 				has_bias ? biases_id : "", has_initial_hidden ? hidden_id : "", has_initial_cell ? cell_id : "", "",
@@ -1232,13 +1226,13 @@ void lstm_gpu_concatenated_input_test(int layers, int sequence_len, int directio
 	auto outputs = network.execute();
 	{
 		ASSERT_EQ(outputs.size(), size_t(1));
-		size_t output_size = outputs.begin()->second.get_memory().size() / sizeof(T);
+		size_t output_size = outputs.begin()->second.get_memory()->size() / sizeof(T);
 		ASSERT_EQ(output_size, size_t(hidden_size * sequence_len * batch_size * direction));
 
 		auto output = outputs.begin()->second.get_memory();
 
 		// Get the output tensor
-		cldnn::layout output_layout = output.get_layout();
+		cldnn::layout output_layout = output->get_layout();
 		cldnn::tensor output_tensor = output_layout.size;
 
 		// Compare the output tensor configuration against the reference value
@@ -1248,7 +1242,7 @@ void lstm_gpu_concatenated_input_test(int layers, int sequence_len, int directio
 		ASSERT_EQ(direction, output_tensor.spatial[1]);
 		ASSERT_EQ(hidden_size, output_tensor.spatial[0]);
 
-		auto output_ptr = output.pointer<T>();
+        cldnn::mem_lock<T> output_ptr(output, get_test_stream());
 		int32_t i = 0;
 		for (int32_t b = 0; b < batch_size; ++b) {
 			for (int32_t s = 0; s < sequence_len; ++s) {
@@ -1388,49 +1382,49 @@ void lstm_gpu_chain_test(int batch_size, int input_size, int hidden_size,
         }
     }
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor = { batch_size, sequence_len, input_size, 1 };
     layout layout = { type_to_data_type<T>::value, cldnn::format::bfyx, input_tensor };
 
-    memory input = memory::allocate(engine, layout);
+    memory::ptr input = engine.allocate_memory(layout);
     set_values(input, ref_input_vec);
 
     // 2-dim vectors to support chain and layers
-    std::vector<std::vector<memory>> weights;
-    std::vector<std::vector<memory>> recurrent;
-    std::vector<std::vector<memory>> biases;
-    std::vector<std::vector<memory>> hidden;
-    std::vector<std::vector<memory>> cell;
+    std::vector<std::vector<memory::ptr>> weights;
+    std::vector<std::vector<memory::ptr>> recurrent;
+    std::vector<std::vector<memory::ptr>> biases;
+    std::vector<std::vector<memory::ptr>> hidden;
+    std::vector<std::vector<memory::ptr>> cell;
 
     for (size_t chain = 0; chain < chains; chain++) {
-        std::vector<memory> per_chain_weights;
-        std::vector<memory> per_chain_recurrent;
-        std::vector<memory> per_chain_biases;
-        std::vector<memory> per_chain_hidden;
-        std::vector<memory> per_chain_cell;
+        std::vector<memory::ptr> per_chain_weights;
+        std::vector<memory::ptr> per_chain_recurrent;
+        std::vector<memory::ptr> per_chain_biases;
+        std::vector<memory::ptr> per_chain_hidden;
+        std::vector<memory::ptr> per_chain_cell;
 
         for (size_t layer = 0; layer < layers; layer++) {
-            per_chain_weights.push_back(memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, {1, directions, layer == 0 ? input_size : hidden_size, 4 * hidden_size} }));
+            per_chain_weights.push_back(engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, {1, directions, layer == 0 ? input_size : hidden_size, 4 * hidden_size} }));
             set_values(per_chain_weights[layer], ref_weights_vec[chain][layer]);
 
-            per_chain_recurrent.push_back(memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, {1, directions, hidden_size, 4 * hidden_size} }));
+            per_chain_recurrent.push_back(engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, {1, directions, hidden_size, 4 * hidden_size} }));
             set_values(per_chain_recurrent[layer], ref_recurrent_vec[chain][layer]);
 
             if (has_bias)
             {
-                per_chain_biases.push_back(memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, {1, 1, 4 * hidden_size, directions} }));
+                per_chain_biases.push_back(engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, {1, 1, 4 * hidden_size, directions} }));
                 set_values(per_chain_biases[layer], ref_bias_vec[chain][layer]);
             }
 
             if (has_initial_hidden)
             {
-                per_chain_hidden.push_back(memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, {1, 1, hidden_size, directions} }));
+                per_chain_hidden.push_back(engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, {1, 1, hidden_size, directions} }));
                 set_values(per_chain_hidden[layer], ref_hidden_vec[chain][layer]);
             }
 
             if (has_initial_cell)
             {
-                per_chain_cell.push_back(memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, {1, 1, hidden_size, directions} }));
+                per_chain_cell.push_back(engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, {1, 1, hidden_size, directions} }));
                 set_values(per_chain_cell[layer], ref_cell_vec[chain][layer]);
             }
         }
@@ -1448,7 +1442,7 @@ void lstm_gpu_chain_test(int batch_size, int input_size, int hidden_size,
     std::vector<primitive_id> lstm_inputs;
     std::vector<primitive_id> output_ids_offsets;
 
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
 
     for (int feature = 0; feature < sequence_len; feature++) {
         input_ids_offsets.push_back({ get_string_id(feature), {0, feature, 0, 0} });
@@ -1498,8 +1492,8 @@ void lstm_gpu_chain_test(int batch_size, int input_size, int hidden_size,
 
             if (chain == 0 && layer == 0)
             {
-                if (has_initial_hidden) topology.add(input_layout(hidden_id, hidden[chain][layer].get_layout()));
-                if (has_initial_cell) topology.add(input_layout(cell_id, cell[chain][layer].get_layout()));
+                if (has_initial_hidden) topology.add(input_layout(hidden_id, hidden[chain][layer]->get_layout()));
+                if (has_initial_cell) topology.add(input_layout(cell_id, cell[chain][layer]->get_layout()));
             }
 
             // Get the initial hidden and initial cell for each layer for each chain link
@@ -1577,7 +1571,7 @@ void lstm_gpu_chain_test(int batch_size, int input_size, int hidden_size,
     auto outputs = network.execute();
     for (auto itr = outputs.begin(); itr != outputs.end(); itr++)
     {
-        auto output_tensor = itr->second.get_memory().get_layout().size;
+        auto output_tensor = itr->second.get_memory()->get_layout().size;
         primitive_id primitive_name = itr->first;
 
         // Split the primitive id to get the chain id
@@ -1589,8 +1583,8 @@ void lstm_gpu_chain_test(int batch_size, int input_size, int hidden_size,
         size_t chain_id = stoi(chain_str);
         size_t layer_id = stoi(layer_str);
 
-        cldnn::memory output_memory = itr->second.get_memory();
-        int32_t output_size = (int32_t)(itr->second.get_memory().size() / sizeof(T));
+        cldnn::memory::ptr output_memory = itr->second.get_memory();
+        int32_t output_size = (int32_t)(itr->second.get_memory()->size() / sizeof(T));
         cldnn::tensor ref_output_tensor;
         VVVVF<T> ref_primitive_output;
 
@@ -1628,7 +1622,7 @@ void lstm_gpu_chain_test(int batch_size, int input_size, int hidden_size,
         ASSERT_EQ(ref_directions, output_tensor.spatial[1]);	// directions should match
         ASSERT_EQ(ref_hidden_size, output_tensor.spatial[0]);	// input size should match
 
-        auto output_ptr = output_memory.pointer<T>();
+        cldnn::mem_lock<T> output_ptr(output_memory, get_test_stream());
 
         int32_t i = 0;
         for (int32_t b = 0; b < ref_batch_size; ++b) {
@@ -2049,4 +2043,3 @@ TEST(lstm_gpu, generic_lstm_stacked_bi_f16) {
 // integration testing using multi-layer and chained LSTMs
 // LSTMs single input
 // optional activation list
-
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/max_unpooling_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/max_unpooling_gpu_test.cpp
index 1fbdba583d7b01..0cb3c2cb828f1a 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/max_unpooling_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/max_unpooling_gpu_test.cpp
@@ -3,22 +3,18 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/max_unpooling.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/reorder.hpp>
-#include <api/data.hpp>
-#include <api/mutable_data.hpp>
-#include <api/pooling.hpp>
-#include "test_utils/float16.h"
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/max_unpooling.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/mutable_data.hpp>
+#include <cldnn/primitives/pooling.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 TEST(max_unpooling_gpu, basic_in2x3x2x2) {
     //  Input  : 2x2x2x1
@@ -45,10 +41,10 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2) {
     //  f1: b0:  0    0  0   b1:   0    0    0
     //  f1: b0:  0    8  16  b1:   12   0    17
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
-    auto arg_max = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
+    auto arg_max = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         4.0f, 4.0f,
@@ -65,7 +61,7 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("arg_max", arg_max));
     topology.add(max_unpooling("max_unpooling", "input", "arg_max", { 1, 1, 2, 2 }, { 1, 1, 1, 1 }));
 
@@ -76,8 +72,8 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("max_unpooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 2);
@@ -127,10 +123,10 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2_output_padding) {
     //  f1: b0:  0    0  0   b1:   0    0    0
     //  f1: b0:  0    8  16  b1:   12   0    17
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
-    auto arg_max = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto arg_max = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         4.0f, 4.0f,
@@ -147,7 +143,7 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2_output_padding) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("arg_max", arg_max));
     topology.add(max_unpooling("max_unpooling", "input", "arg_max", { 1, 1, 2, 2 }, { 1, 1, 1, 1 }, { 0, 0, 0, 0 }, padding({ 0, 0, 1, 1 }, 0)));
 
@@ -158,8 +154,8 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2_output_padding) {
     auto outputs = network.execute();
 
     auto output = outputs.at("max_unpooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 2);
@@ -218,10 +214,10 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2_output_size) {
     //  f1: b0:  0    0  0   b1:   0    0    0
     //  f1: b0:  0    8  16  b1:   12   0    17
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
-    auto arg_max = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto arg_max = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         4.0f, 4.0f,
@@ -238,7 +234,7 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2_output_size) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("arg_max", arg_max));
     topology.add(max_unpooling("max_unpooling", "input", "arg_max", {2, 2, 3, 2}));
 
@@ -249,8 +245,8 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2_output_size) {
     auto outputs = network.execute();
 
     auto output = outputs.at("max_unpooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 2);
@@ -299,10 +295,10 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2_fp16) {
     //  f1: b0:  0    0  0   b1:   0    0    0
     //  f1: b0:  0    8  16  b1:   12   0    17
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx,{ 2, 2, 2, 1 } });
-    auto arg_max = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx,{ 2, 2, 2, 1 } });
+    auto arg_max = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         FLOAT16(4.0f), FLOAT16(4.0f),
@@ -319,7 +315,7 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2_fp16) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("arg_max", arg_max));
     topology.add(max_unpooling("max_unpooling", "input", "arg_max", { 1, 1, 2, 2 }, { 1, 1, 1, 1 }));
 
@@ -330,8 +326,8 @@ TEST(max_unpooling_gpu, basic_in2x3x2x2_fp16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("max_unpooling").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
-    auto output_layout = output.get_layout();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 2);
@@ -379,10 +375,10 @@ TEST(max_unpooling_gpu, basic_in2x2x3x2_max_with_argmax_pooling_unpooling) {
     //  f1: b0:  0    0  0   b1:   0    0    0
     //  f1: b0:  0    8  16  b1:   12   0    17
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
-    auto arg_max = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
+    auto arg_max = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         1.0f, 2.0f, -10.f,
@@ -396,7 +392,7 @@ TEST(max_unpooling_gpu, basic_in2x2x3x2_max_with_argmax_pooling_unpooling) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mutable_data("arg_max", arg_max));
     topology.add(pooling("pooling_max_with_argmax", "input", "arg_max", pooling_mode::max_with_argmax, { 1, 1, 2, 2 }, { 1, 1, 1, 1 }));
     topology.add(max_unpooling("max_unpooling", "pooling_max_with_argmax", "arg_max", { 1, 1, 2, 2 }, { 1, 1, 1, 1 }));
@@ -408,9 +404,9 @@ TEST(max_unpooling_gpu, basic_in2x2x3x2_max_with_argmax_pooling_unpooling) {
     auto outputs = network.execute();
 
     auto output = outputs.at("max_unpooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
-    auto argmax_ptr = arg_max.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
+    cldnn::mem_lock<float> argmax_ptr(arg_max, get_test_stream());
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 2);
@@ -443,4 +439,4 @@ TEST(max_unpooling_gpu, basic_in2x2x3x2_max_with_argmax_pooling_unpooling) {
     for (size_t i = 0; i < expected_argmax_vec.size(); ++i) {
         EXPECT_EQ(expected_argmax_vec[i], argmax_ptr[i]);
     }
-}
\ No newline at end of file
+}
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/memory_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/memory_test.cpp
index d303700be2ffd2..861cd2672cd4e9 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/memory_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/memory_test.cpp
@@ -4,24 +4,19 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include <gtest/gtest.h>
-#include <api/engine.hpp>
-#include <api/memory.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/input_layout.hpp>
-#include <api/activation.hpp>
-#include <api/pooling.hpp>
-#include <api/concatenation.hpp>
-#include <api/data.hpp>
-#include <api/reshape.hpp>
-#include <api/crop.hpp>
-#include <api/scale.hpp>
-
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/activation.hpp>
+#include <cldnn/primitives/pooling.hpp>
+#include <cldnn/primitives/concatenation.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/reshape.hpp>
+#include <cldnn/primitives/crop.hpp>
+#include <cldnn/primitives/scale.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 #if 0
 TEST(memory_tests, DISABLED_execution_loop)
@@ -31,12 +26,12 @@ TEST(memory_tests, DISABLED_execution_loop)
     memory in = memory::allocate(eng, layout{ data_types::f32, format::bfyx, { 1, 1, 1000, 1000 } });
 
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         activation("out", "in", activation_func::linear)
     };
 
     network net(eng, tpl);
-    
+
     while (true)
     {
         net.set_input_data("in", in);
@@ -51,7 +46,7 @@ TEST(memory_tests, DISABLED_network_creation_loop)
     memory in = memory::allocate(eng, layout{ data_types::f32, format::bfyx,{ 1, 1, 1000, 1000 } });
 
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         activation("out", "in", activation_func::linear)
     };
 
@@ -62,17 +57,19 @@ TEST(memory_tests, DISABLED_network_creation_loop)
 }
 #endif
 TEST(memory_pool, basic_non_padded_relu_pipe) {
-    // 5 relu's of size 1x4x1x1
-    const cldnn::engine engine;// here we need new engine
+    // We need a new engine here to get correct get_max_used_device_memory() result
+    // If we reuse common engine, then max memory value will be taken from some previously executed tests
+    // as it's tracked within engine instance
+    auto engine = create_test_engine();
     auto batch_num = 1;
     auto feature_num = 4;
     auto x_size = 1;
     auto y_size = 1;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine->allocate_memory({ data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(activation("relu", "input", activation_func::relu));
     topology.add(activation("relu1", "relu", activation_func::relu));
     topology.add(activation("relu2", "relu1", activation_func::relu));
@@ -85,27 +82,27 @@ TEST(memory_pool, basic_non_padded_relu_pipe) {
     build_options bo;
     bo.set_option(build_option::optimize_data(true));
 
-    network network(engine, topology, bo);
+    network network(*engine, topology, bo);
     network.set_input_data("input", input);
     auto outputs = network.execute();
 
-    EXPECT_EQ(engine.get_max_used_device_memory_size(), (uint64_t) 64);
+    EXPECT_EQ(engine->get_max_used_device_memory(), (uint64_t) 64);
  }
 
 TEST(memory_pool, basic_non_padded_relu_and_pooling_pipe) {
-    // uncomment this line to disable memory pool
-    /*engine_configuration cfg{ false, false, false, std::string(), std::string(), true, std::string(),std::string(), 0, false };
-    engine engine{ cfg };*/
-    const cldnn::engine engine;// here we need new engine
+    // We need a new engine here to get correct get_max_used_device_memory() result
+    // If we reuse common engine, then max memory value will be taken from some previously executed tests
+    // as it's tracked within engine instance
+    auto engine = create_test_engine();
     auto batch_num = 1;
     auto feature_num = 4;
     auto x_size = 4;
     auto y_size = 4;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine->allocate_memory({ data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(activation("relu", "input", activation_func::relu));
     topology.add(activation("relu1", "relu", activation_func::relu));
     topology.add(pooling("pool1", "relu1",pooling_mode::max, { 1,1,3,3 }, { 1,1,2,2 }));
@@ -117,11 +114,11 @@ TEST(memory_pool, basic_non_padded_relu_and_pooling_pipe) {
     build_options bo;
     bo.set_option(build_option::optimize_data(true));
 
-    network network(engine, topology, bo);
+    network network(*engine, topology, bo);
     network.set_input_data("input", input);
     auto outputs = network.execute();
 
-    EXPECT_EQ(engine.get_max_used_device_memory_size(), (uint64_t)896);
+    EXPECT_EQ(engine->get_max_used_device_memory(), (uint64_t)896);
 }
 
 TEST(memory_pool, multi_outputs_network) {
@@ -130,19 +127,16 @@ TEST(memory_pool, multi_outputs_network) {
     //            -- relu2 --  relu3 -- relu5--relu6--relu7
     // neither of relu5, relu6 nor relu7 can share resource with relu4.
 
-    // uncomment this line to disable memory pool
-    /*engine_configuration cfg{ false, false, false, std::string(), std::string(), true, std::string(),std::string(), 0, false };
-    engine engine{ cfg };*/
-    const cldnn::engine engine;// here we need new engine
+    auto engine = create_test_engine();
     auto batch_num = 1;
     auto feature_num = 4;
     auto x_size = 4;
     auto y_size = 4;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine->allocate_memory({ data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(activation("relu", "input", activation_func::relu));
     topology.add(activation("relu1", "relu", activation_func::relu));
     topology.add(activation("relu2", "input", activation_func::relu));
@@ -155,11 +149,11 @@ TEST(memory_pool, multi_outputs_network) {
     build_options bo;
     bo.set_option(build_option::optimize_data(true));
 
-    network network(engine, topology, bo);
+    network network(*engine, topology, bo);
     network.set_input_data("input", input);
     auto outputs = network.execute();
 
-    EXPECT_EQ(engine.get_max_used_device_memory_size(), (uint64_t)1536);
+    EXPECT_EQ(engine->get_max_used_device_memory(), (uint64_t)1536);
 }
 
 TEST(memory_pool, oooq) {
@@ -168,17 +162,19 @@ TEST(memory_pool, oooq) {
                 -- relu3 --  relu5 ---------
        neither of relu5, relu6 nor relu7 can share resource with relu4. */
 
-    engine_configuration cfg{ false, false, false, std::string(), std::string(), true /*oooq*/, std::string(),std::string(), priority_mode_types::disabled, throttle_mode_types::disabled, true /*mem_pool*/ };
-    engine engine{ cfg };
+    // We need a new engine here to get correct get_max_used_device_memory() result
+    // If we reuse common engine, then max memory value will be taken from some previously executed tests
+    // as it's tracked within engine instance
+    auto engine = create_test_engine();
     auto batch_num = 1;
     auto feature_num = 4;
     auto x_size = 4;
     auto y_size = 4;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine->allocate_memory({ data_types::f32, format::bfyx,{ tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num)) } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(activation("relu1", "input", activation_func::relu));
     topology.add(activation("relu2", "input", activation_func::relu));
     topology.add(activation("relu3", "input", activation_func::relu));
@@ -191,11 +187,11 @@ TEST(memory_pool, oooq) {
     build_options bo;
     bo.set_option(build_option::optimize_data(true));
 
-    network network(engine, topology, bo);
+    network network(*engine, topology, bo);
     network.set_input_data("input", input);
     auto outputs = network.execute();
 
-    EXPECT_EQ(engine.get_max_used_device_memory_size(), (uint64_t) 2560);
+    EXPECT_EQ(engine->get_max_used_device_memory(), (uint64_t) 2560);
 }
 
 TEST(memory_pool, DISABLED_shared_mem_pool_same_topology_twice) {
@@ -204,14 +200,16 @@ TEST(memory_pool, DISABLED_shared_mem_pool_same_topology_twice) {
                       -- relu3 --  relu5 ---------
     neither of relu5, relu6 nor relu7 can share resource with relu4. */
 
-    engine_configuration cfg{ false, false, false, std::string(), std::string(), true /*oooq*/, std::string(),std::string(), priority_mode_types::disabled, throttle_mode_types::disabled, true /*mem_pool*/ };
-    engine engine{ cfg };
+    // We need a new engine here to get correct get_max_used_device_memory() result
+    // If we reuse common engine, then max memory value will be taken from some previously executed tests
+    // as it's tracked within engine instance
+    auto engine = create_test_engine();
     auto batch_num = 1;
     auto feature_num = 4;
     auto inp_x_size = 4;
     auto inp_y_size = 4;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ tensor(spatial(inp_x_size, inp_y_size), feature(feature_num), batch(batch_num)) } });
+    auto input = engine->allocate_memory({ data_types::f32, format::bfyx,{ tensor(spatial(inp_x_size, inp_y_size), feature(feature_num), batch(batch_num)) } });
 
     set_values(input,
     {   1.0f, 2.5f, 3.0f, 4.0f, 5.0f, 2.0f, 2.0f, 3.0f, 6.1f, 4.7f, 1.0f, 1.0f, 8.2f, 1.0f, 2.0f, 1.0f,
@@ -221,7 +219,7 @@ TEST(memory_pool, DISABLED_shared_mem_pool_same_topology_twice) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(activation("relu1", "input", activation_func::relu));
     topology.add(activation("relu2", "input", activation_func::sqrt));
     topology.add(activation("relu3", "input", activation_func::square));
@@ -234,25 +232,25 @@ TEST(memory_pool, DISABLED_shared_mem_pool_same_topology_twice) {
     build_options bo;
     bo.set_option(build_option::optimize_data(true));
 
-    network network_first(engine, topology, bo);
+    network network_first(*engine, topology, bo);
     network_first.set_input_data("input", input);
     auto outputs = network_first.execute();
 
     auto output_memory_first = outputs.at("relu6").get_memory();
-    auto output_layout_first = output_memory_first.get_layout();
-    auto output_ptr_first = output_memory_first.pointer<float>();
+    auto output_layout_first = output_memory_first->get_layout();
+    cldnn::mem_lock<float> output_ptr_first(output_memory_first, get_test_stream());
 
-    EXPECT_EQ(engine.get_max_used_device_memory_size(), (uint64_t) 2560);
+    EXPECT_EQ(engine->get_max_used_device_memory(), (uint64_t) 2560);
 
-    network network_second(engine, topology, bo);
+    network network_second(*engine, topology, bo);
     network_second.set_input_data("input", input);
     auto outputs_second = network_second.execute();
 
     auto output_memory_second = outputs_second.at("relu6").get_memory();
-    auto output_layout_second = output_memory_second.get_layout();
-    auto output_ptr_second = output_memory_second.pointer<float>();
+    auto output_layout_second = output_memory_second->get_layout();
+    cldnn::mem_lock<float> output_ptr_second(output_memory_second, get_test_stream());
 
-    EXPECT_EQ(engine.get_max_used_device_memory_size(), (uint64_t) 3328);
+    EXPECT_EQ(engine->get_max_used_device_memory(), (uint64_t) 3328);
 
     EXPECT_EQ(output_layout_first, output_layout_second);
 
@@ -279,16 +277,17 @@ TEST(memory_pool, DISABLED_shared_mem_pool_same_topology_twice) {
 }
 
 TEST(memory_pool, DISABLED_shared_mem_pool_same_topology_twice_weights) {
-
-    engine_configuration cfg{ false, false, false, std::string(), std::string(), true /*oooq*/, std::string(),std::string(), priority_mode_types::disabled, throttle_mode_types::disabled, true /*mem_pool*/ };
-    engine engine{ cfg };
+    // We need a new engine here to get correct get_max_used_device_memory() result
+    // If we reuse common engine, then max memory value will be taken from some previously executed tests
+    // as it's tracked within engine instance
+    auto engine = create_test_engine();
     auto batch_num = 1;
     auto feature_num = 3;
     auto inp_x_size = 4;
     auto inp_y_size = 4;
 
-    auto input= memory::allocate(engine, { data_types::f32, format::bfyx,{ tensor(spatial(inp_x_size, inp_y_size), feature(feature_num), batch(batch_num)) } });
-    auto weights = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 3, 2 } });
+    auto input= engine->allocate_memory({ data_types::f32, format::bfyx,{ tensor(spatial(inp_x_size, inp_y_size), feature(feature_num), batch(batch_num)) } });
+    auto weights = engine->allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 3, 2 } });
 
     std::vector<float> dummy_input_data_1 = {
        /*f0 xy*/ 0.8f, 0.65f, 0.1f, 1.0f, 1.0f, 0.5f, 0.11f, 0.33f, 0.66f, 0.11f, 0.22f, 0.33f, 0.99f, 0.8f, 0.7f, 0.5f,
@@ -300,7 +299,7 @@ TEST(memory_pool, DISABLED_shared_mem_pool_same_topology_twice_weights) {
     set_values(weights, { 0.10f, 0.2f, 0.1f, 0.2f, 0.1f, 0.2f });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("weights", weights),
         convolution("conv", "input", { "weights" }, { 1, 1, 1, 2 }),
         softmax("softmax", "conv"));
@@ -308,31 +307,31 @@ TEST(memory_pool, DISABLED_shared_mem_pool_same_topology_twice_weights) {
     build_options bo;
     bo.set_option(build_option::optimize_data(true));
 
-    network network_first(engine, topology, bo);
+    network network_first(*engine, topology, bo);
     network_first.set_input_data("input", input);
     auto outputs = network_first.execute();
     uint64_t cl_mem_result = 824;
     uint64_t usm_result = 1208; // USM have a higher peak, since transfering memory to device adds temporay memory bytes allocated. Old memory is deallocated quickly, but max peak is higher.
-    auto is_correct = engine.get_max_used_device_memory_size() == cl_mem_result
-        || engine.get_max_used_device_memory_size() == usm_result;
+    auto is_correct = engine->get_max_used_device_memory() == cl_mem_result
+        || engine->get_max_used_device_memory() == usm_result;
     EXPECT_TRUE(is_correct) << "Memory max peak is not correct";
 
     auto output_memory_first = outputs.at("softmax").get_memory();
-    auto output_layout_first = output_memory_first.get_layout();
-    auto output_ptr_first = output_memory_first.pointer<float>();
+    auto output_layout_first = output_memory_first->get_layout();
+    cldnn::mem_lock<float> output_ptr_first(output_memory_first, get_test_stream());
 
-    network network_second(engine, topology, bo);
+    network network_second(*engine, topology, bo);
     network_second.set_input_data("input", input);
     auto outputs_second = network_second.execute();
 
     auto output_memory_second = outputs_second.at("softmax").get_memory();
-    auto output_layout_second = output_memory_second.get_layout();
-    auto output_ptr_second = output_memory_second.pointer<float>();
+    auto output_layout_second = output_memory_second->get_layout();
+    cldnn::mem_lock<float> output_ptr_second(output_memory_second, get_test_stream());
 
     cl_mem_result = 1224;
     usm_result = 1992; // USM have a higher peak, since transfering memory to device adds temporay memory bytes allocated. Old memory is deallocated quickly, but max peak is higher.
-    is_correct = engine.get_max_used_device_memory_size() == cl_mem_result
-        || engine.get_max_used_device_memory_size() == usm_result;
+    is_correct = engine->get_max_used_device_memory() == cl_mem_result
+        || engine->get_max_used_device_memory() == usm_result;
     EXPECT_TRUE(is_correct) << "Memory max peak is not correct";
     EXPECT_EQ(output_layout_first, output_layout_second);
 
@@ -359,9 +358,10 @@ TEST(memory_pool, DISABLED_shared_mem_pool_same_topology_twice_weights) {
 }
 
 TEST(memory_pool, shared_mem_pool_diff_batches) {
-
-    engine_configuration cfg{ false, false, false, std::string(), std::string(), true /*oooq*/, std::string(),std::string(), priority_mode_types::disabled, throttle_mode_types::disabled, true /*mem_pool*/ };
-    engine engine{ cfg };
+    // We need a new engine here to get correct get_max_used_device_memory() result
+    // If we reuse common engine, then max memory value will be taken from some previously executed tests
+    // as it's tracked within engine instance
+    auto engine = create_test_engine();
     auto batch_8 = 8;
     auto batch_1 = 1;
     auto feature_num = 3;
@@ -371,9 +371,9 @@ TEST(memory_pool, shared_mem_pool_diff_batches) {
     auto fmt = format::bfyx;
     layout lay_batch_1 = { dt, fmt, { tensor(spatial(inp_x_size, inp_y_size), feature(feature_num), batch(batch_1)) }};
     layout lay_batch_8 = { dt, fmt, { tensor(spatial(inp_x_size, inp_y_size), feature(feature_num), batch(batch_8)) }};
-    auto input_1 = memory::allocate(engine, lay_batch_1);
-    auto input_8 = memory::allocate(engine, lay_batch_8);
-    auto weights = memory::allocate(engine, { dt, fmt, { 1, 1, 3, 2 } });
+    auto input_1 = engine->allocate_memory(lay_batch_1);
+    auto input_8 = engine->allocate_memory(lay_batch_8);
+    auto weights = engine->allocate_memory({ dt, fmt, { 1, 1, 3, 2 } });
 
     std::vector<float> dummy_input_data_1 = generate_random_1d<float>(batch_1*feature_num*inp_x_size*inp_y_size, 0, 1);
     std::vector<float> dummy_input_data_8 = generate_random_1d<float>(batch_8*feature_num*inp_x_size*inp_y_size, 0, 1);
@@ -383,7 +383,7 @@ TEST(memory_pool, shared_mem_pool_diff_batches) {
     set_values(weights, { 0.10f, 0.2f, 0.1f, 0.2f, 0.1f, 0.2f });
 
     topology topo(
-        input_layout("input", input_8.get_layout()),
+        input_layout("input", input_8->get_layout()),
         data("weights", weights),
         convolution("conv", "input", { "weights" }, { 1, 1, 1, 2 }),
         softmax("softmax", "conv"));
@@ -391,25 +391,27 @@ TEST(memory_pool, shared_mem_pool_diff_batches) {
     build_options bo;
     bo.set_option(build_option::optimize_data(true));
 
-    network network_first(engine, topo, bo);
+    network network_first(*engine, topo, bo);
     network_first.set_input_data("input", input_8);
     auto outputs = network_first.execute();
 
-    auto dev_info = engine.get_info();
-    EXPECT_EQ(engine.get_max_used_device_memory_size(), (uint64_t)3928);
+    auto dev_info = engine->get_device_info();
+    EXPECT_EQ(engine->get_max_used_device_memory(), (uint64_t)3928);
 
-    topo.change_input_layout("input", input_1.get_layout());//change input layout to batch=1
+    topo.change_input_layout("input", input_1->get_layout());//change input layout to batch=1
 
-    network network_second(engine, topo, bo);
+    network network_second(*engine, topo, bo);
     network_second.set_input_data("input", input_1);
     auto outputs_second = network_second.execute();
-    EXPECT_EQ(engine.get_max_used_device_memory_size(), (uint64_t)3928);
+    EXPECT_EQ(engine->get_max_used_device_memory(), (uint64_t)3928);
 }
 
 TEST(memory_pool, shared_dep_two_output) {
+    // We need a new engine here to get correct get_max_used_device_memory() result
+    // If we reuse common engine, then max memory value will be taken from some previously executed tests
+    // as it's tracked within engine instance
+    auto engine = create_test_engine();
 
-    engine_configuration cfg{ false, false, false, std::string(), std::string(), true /*oooq*/, std::string(),std::string(), priority_mode_types::disabled, throttle_mode_types::disabled, true /*mem_pool*/ };
-    engine engine{ cfg };
     auto batch_1 = 1;
     auto feature_num = 1;
     auto inp_x_size = 4;
@@ -417,7 +419,7 @@ TEST(memory_pool, shared_dep_two_output) {
     auto dt = data_types::f32;
     auto fmt = format::bfyx;
     layout lay_batch_1 = { dt, fmt,{ tensor(spatial(inp_x_size, inp_y_size), feature(feature_num), batch(batch_1)) } };
-    auto input_1 = memory::allocate(engine, lay_batch_1);
+    auto input_1 = engine->allocate_memory(lay_batch_1);
     set_random_values<float>(input_1);
 
     //build primitives
@@ -445,21 +447,19 @@ TEST(memory_pool, shared_dep_two_output) {
     build_options bo;
     bo.set_option(build_option::optimize_data(true));
 
-    network network(engine, topo, bo);
+    network network(*engine, topo, bo);
     auto outputs = network.execute();
-    EXPECT_EQ(engine.get_max_used_device_memory_size(), (uint64_t)256);
+    EXPECT_EQ(engine->get_max_used_device_memory(), (uint64_t)256);
 }
 
 TEST(memory_pool, non_opt_intermidate_opt_after) {
-
-    engine_configuration cfg{ false, false, false, std::string(), std::string(), true /*oooq*/, std::string(),std::string(), priority_mode_types::disabled, throttle_mode_types::disabled, true /*mem_pool*/ };
-    engine engine{ cfg };
+    auto& engine = get_test_engine();
     auto input_layout1 = layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1, 1, 2, 2 });
     auto input_layout2 = layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1, 1, 2, 2 });
 
-    auto input_memory1 = cldnn::memory::allocate(engine, input_layout1);
-    auto input_memory2 = cldnn::memory::allocate(engine, input_layout2);
-    auto scale_memory = cldnn::memory::allocate(engine, layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1,1,1,1 }));
+    auto input_memory1 = engine.allocate_memory(input_layout1);
+    auto input_memory2 = engine.allocate_memory(input_layout2);
+    auto scale_memory = engine.allocate_memory(layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1,1,1,1 }));
     auto data_memory = cldnn::data("scale_mem", scale_memory);
 
     set_values(input_memory1, { 1.0f, 2.0f, 3.0f, 4.0f });
@@ -496,20 +496,19 @@ TEST(memory_pool, non_opt_intermidate_opt_after) {
     auto out1 = outputs.at("elt1");
     auto out2 = outputs.at("elt2");
 
-    auto out1_ptr = out1.get_memory().pointer<float>();
-    auto out2_ptr = out2.get_memory().pointer<float>();
+    cldnn::mem_lock<float> out1_ptr(out1.get_memory(), get_test_stream());
+    cldnn::mem_lock<float> out2_ptr(out2.get_memory(), get_test_stream());
     EXPECT_EQ(out1_ptr[0], 1.0f);
     EXPECT_EQ(out2_ptr[0], 2.0f);
 }
 
 TEST(memory_pool, add_mem_dep_test) {
+    auto& engine = get_test_engine();
 
-    engine_configuration cfg{ false, false, false, std::string(), std::string(), true /*oooq*/, std::string(),std::string(), priority_mode_types::disabled, throttle_mode_types::disabled, true /*mem_pool*/ };
-    engine engine{ cfg };
     auto input_layout1 = layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1, 2, 2, 2 });
 
-    auto input_memory1 = cldnn::memory::allocate(engine, input_layout1);
-    auto scale_memory = cldnn::memory::allocate(engine, layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1,1,1,1 }));
+    auto input_memory1 = engine.allocate_memory(input_layout1);
+    auto scale_memory = engine.allocate_memory(layout(cldnn::data_types::f32, cldnn::format::bfyx, { 1,1,1,1 }));
     auto data_memory = cldnn::data("scale_mem", scale_memory);
 
     set_values(input_memory1, { 1.0f, 2.0f, 3.0f, 4.0f,
@@ -545,8 +544,8 @@ TEST(memory_pool, add_mem_dep_test) {
     auto out1 = outputs.at("out3");
     auto out2 = outputs.at("out4");
 
-    auto out1_ptr = out1.get_memory().pointer<float>();
-    auto out2_ptr = out2.get_memory().pointer<float>();
+    cldnn::mem_lock<float> out1_ptr(out1.get_memory(), get_test_stream());
+    cldnn::mem_lock<float> out2_ptr(out2.get_memory(), get_test_stream());
     EXPECT_EQ(out1_ptr[0], 1.0f);
     EXPECT_EQ(out1_ptr[1], 2.0f);
     EXPECT_EQ(out1_ptr[2], 3.0f);
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/mvn_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/mvn_gpu_test.cpp
index 410fc6cad02bd3..5632a19cfd4ebe 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/mvn_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/mvn_gpu_test.cpp
@@ -4,26 +4,22 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include <gtest/gtest.h>
-#include <api/memory.hpp>
-#include <api/input_layout.hpp>
-#include "api/mvn.hpp"
-#include "api/reorder.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <iostream>
-#include "float16.h"
 #include "test_utils.h"
 
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/mvn.hpp>
+#include <cldnn/primitives/reorder.hpp>
+
+#include <iostream>
+
 using namespace cldnn;
+using namespace ::tests;
 
 class mvn_gpu_test : public ::testing::TestWithParam<cldnn::format> {};
 
 template <typename T>
-void mvn_compute_mean_across_channels(cldnn::memory& output, bool normalize_variance) {
-    auto output_size = output.get_layout().size;
+void mvn_compute_mean_across_channels(cldnn::memory::ptr output, bool normalize_variance) {
+    auto output_size = output->get_layout().size;
 
     uint32_t batch_size = output_size.batch[0];
     uint32_t feature_size = output_size.feature[0];
@@ -31,9 +27,9 @@ void mvn_compute_mean_across_channels(cldnn::memory& output, bool normalize_vari
     uint32_t y_size = output_size.spatial[1];
     uint32_t x_size = output_size.spatial[0];
 
-    auto buff = output.pointer<T>();
+    cldnn::mem_lock<T> buff(output, get_test_stream());
 
-    float err_margin = output.get_layout().data_type == data_types::f32 ? 1e-03F : 1e-02F;
+    float err_margin = output->get_layout().data_type == data_types::f32 ? 1e-03F : 1e-02F;
 
     for (uint32_t b = 0; b < batch_size; ++b) {
         float sum = 0.f;
@@ -43,7 +39,7 @@ void mvn_compute_mean_across_channels(cldnn::memory& output, bool normalize_vari
                 for (uint32_t y = 0; y < y_size; ++y) {
                     for (uint32_t x = 0; x < x_size; ++x) {
                         auto index_tensor = tensor(batch(b), feature(f), spatial(x, y, z, 0));
-                        size_t data_index = output.get_layout().get_linear_offset(index_tensor);
+                        size_t data_index = output->get_layout().get_linear_offset(index_tensor);
                         float data = static_cast<float>(buff[data_index]);
                         sum += data;
                         if (normalize_variance)
@@ -65,8 +61,8 @@ void mvn_compute_mean_across_channels(cldnn::memory& output, bool normalize_vari
 }
 
 template <typename T>
-void mvn_compute_mean_within_channels(cldnn::memory& output, bool normalize_variance) {
-    auto output_size = output.get_layout().size;
+void mvn_compute_mean_within_channels(cldnn::memory::ptr output, bool normalize_variance) {
+    auto output_size = output->get_layout().size;
 
     uint32_t batch_size = output_size.batch[0];
     uint32_t feature_size = output_size.feature[0];
@@ -74,9 +70,9 @@ void mvn_compute_mean_within_channels(cldnn::memory& output, bool normalize_vari
     uint32_t y_size = output_size.spatial[1];
     uint32_t x_size = output_size.spatial[0];
 
-    auto buff = output.pointer<T>();
+    cldnn::mem_lock<T> buff(output, get_test_stream());
 
-    float err_margin = output.get_layout().data_type == data_types::f32 ? 1e-03F : 1e-02F;
+    float err_margin = output->get_layout().data_type == data_types::f32 ? 1e-03F : 1e-02F;
 
     for (uint32_t b = 0; b < batch_size; ++b) {
         for (uint32_t f = 0; f < feature_size; ++f) {
@@ -86,7 +82,7 @@ void mvn_compute_mean_within_channels(cldnn::memory& output, bool normalize_vari
                 for (uint32_t y = 0; y < y_size; ++y) {
                     for (uint32_t x = 0; x < x_size; ++x) {
                         auto index_tensor = tensor(batch(b), feature(f), spatial(x, y, z, 0));
-                        size_t data_index = output.get_layout().get_linear_offset(index_tensor);
+                        size_t data_index = output->get_layout().get_linear_offset(index_tensor);
                         float data = static_cast<float>(buff[data_index]);
                         sum += data;
                         if (normalize_variance)
@@ -110,16 +106,16 @@ void mvn_compute_mean_within_channels(cldnn::memory& output, bool normalize_vari
 TEST(mvn_gpu_test, mvn_test_across_channels_outside_sqrt_bfyx) {
     // mvn across channels fp32 test with normalize_variance set to false
     using namespace cldnn;
-    using namespace tests;
+    using namespace ::tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<float>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", false, 1e-10f, false, true));
 
     network network(engine, topology);
@@ -139,14 +135,14 @@ TEST(mvn_gpu_test, mvn_test_across_channels_inside_sqrt_bfyx) {
     using namespace cldnn;
     using namespace tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<float>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", false, 1e-10f, true, true));
 
     network network(engine, topology);
@@ -164,16 +160,16 @@ TEST(mvn_gpu_test, mvn_test_across_channels_inside_sqrt_bfyx) {
 TEST(mvn_gpu_test, mvn_test_across_channels_bfyx_outside_sqrt_fp16) {
     // mvn across channels fp16 test with normalize_variance set to false
     using namespace cldnn;
-    using namespace tests;
+    using namespace ::tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f16, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<FLOAT16>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", false, 1e-10f, false, true));
 
     network network(engine, topology);
@@ -193,14 +189,14 @@ TEST(mvn_gpu_test, mvn_test_across_channels_inside_sqrt_bfyx_fp16) {
     using namespace cldnn;
     using namespace tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f16, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<FLOAT16>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", false, 1e-10f, true, true));
 
     network network(engine, topology);
@@ -218,16 +214,16 @@ TEST(mvn_gpu_test, mvn_test_across_channels_inside_sqrt_bfyx_fp16) {
 TEST(mvn_gpu_test, mvn_test_across_channels_outside_sqrt_bfyx_normalize_variance) {
     // mvn across channels fp32 test with normalize_variance set to true
     using namespace cldnn;
-    using namespace tests;
+    using namespace ::tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<float>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", true, 1e-10f, false, true));
 
     network network(engine, topology);
@@ -247,14 +243,14 @@ TEST(mvn_gpu_test, mvn_test_across_channels_inside_sqrt_bfyx_normalize_variance)
     using namespace cldnn;
     using namespace tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<float>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", true, 1e-10f, true, true));
 
     network network(engine, topology);
@@ -274,14 +270,14 @@ TEST(mvn_gpu_test, mvn_test_across_channels_outside_sqrt_bfyx_normalize_variance
     using namespace cldnn;
     using namespace tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f16, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<FLOAT16>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", true, 1e-10f, false, true));
 
     network network(engine, topology);
@@ -299,16 +295,16 @@ TEST(mvn_gpu_test, mvn_test_across_channels_outside_sqrt_bfyx_normalize_variance
 TEST(mvn_gpu_test, mvn_test_across_channels_inside_sqrt_bfyx_normalize_variance_fp16) {
     // mvn across channels fp16 test with normalize_variance set to true
     using namespace cldnn;
-    using namespace tests;
+    using namespace ::tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f16, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<FLOAT16>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", true, 1e-10f, true, true));
 
     network network(engine, topology);
@@ -328,14 +324,14 @@ TEST(mvn_gpu_test, mvn_test_within_channels_outside_sqrt_bfyx) {
     using namespace cldnn;
     using namespace tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<float>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", false, 1e-10f, false, false));
 
     network network(engine, topology);
@@ -353,16 +349,16 @@ TEST(mvn_gpu_test, mvn_test_within_channels_outside_sqrt_bfyx) {
 TEST(mvn_gpu_test, mvn_test_within_channels_inside_sqrt__bfyx) {
     // mvn within channels fp32 test with normalize_variance set to false
     using namespace cldnn;
-    using namespace tests;
+    using namespace ::tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<float>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", false, 1e-10f, true, false));
 
     network network(engine, topology);
@@ -380,16 +376,16 @@ TEST(mvn_gpu_test, mvn_test_within_channels_inside_sqrt__bfyx) {
 TEST(mvn_gpu_test, mvn_test_within_channels_outside_sqrt_bfyx_fp16) {
     // mvn within channels fp16 test with normalize_variance set to false
     using namespace cldnn;
-    using namespace tests;
+    using namespace ::tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f16, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<FLOAT16>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", false, 1e-10f, false, false));
 
     network network(engine, topology);
@@ -409,14 +405,14 @@ TEST(mvn_gpu_test, mvn_test_within_channels_inside_sqrt_bfyx_fp16) {
     using namespace cldnn;
     using namespace tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f16, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<FLOAT16>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", false, 1e-10f, true, false));
 
     network network(engine, topology);
@@ -434,16 +430,16 @@ TEST(mvn_gpu_test, mvn_test_within_channels_inside_sqrt_bfyx_fp16) {
 TEST(mvn_gpu_test, mvn_test_within_channels_outside_sqrt_bfyx_normalize_variance) {
     // mvn within channels fp32 test with normalize_variance set to true
     using namespace cldnn;
-    using namespace tests;
+    using namespace ::tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<float>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", true, 1e-10f, false, false));
 
     network network(engine, topology);
@@ -463,14 +459,14 @@ TEST(mvn_gpu_test, mvn_test_within_channels_inside_sqrt_bfyx_normalize_variance)
     using namespace cldnn;
     using namespace tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<float>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", true, 1e-10f, true, false));
 
     network network(engine, topology);
@@ -490,14 +486,14 @@ TEST(mvn_gpu_test, mvn_test_within_channels_outside_sqrt_bfyx_normalize_variance
     using namespace cldnn;
     using namespace tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f16, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<FLOAT16>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", true, 1e-10f, false, false));
 
     network network(engine, topology);
@@ -515,16 +511,16 @@ TEST(mvn_gpu_test, mvn_test_within_channels_outside_sqrt_bfyx_normalize_variance
 TEST(mvn_gpu_test, mvn_test_within_channels_inside_sqrt_bfyx_normalize_variance_fp16) {
     // mvn within channels fp16 test with normalize_variance set to true
     using namespace cldnn;
-    using namespace tests;
+    using namespace ::tests;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::bfyx, {7, 10, 17, 13}});
+    auto input = engine.allocate_memory({data_types::f16, format::bfyx, {7, 10, 17, 13}});
 
     tests::set_random_values<FLOAT16>(input, true, 8, 100);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mvn("mvn", "input", true, 1e-10f, true, false));
 
     network network(engine, topology);
@@ -551,16 +547,16 @@ struct mvn_basic_test_params {
 
 struct mvn_random_test : ::testing::TestWithParam<mvn_basic_test_params> {
     template <typename T>
-    void fill_data(memory& mem, const tests::VVVVVF<T>& data) {
-        auto size = mem.get_layout().size;
-        auto ptr = mem.pointer<T>();
+    void fill_data(memory::ptr mem, const tests::VVVVVF<T>& data) {
+        auto size = mem->get_layout().size;
+        cldnn::mem_lock<T> ptr(mem, get_test_stream());
         for (size_t bi = 0; bi < static_cast<size_t>(size.batch[0]); ++bi) {
             for (size_t fi = 0; fi < static_cast<size_t>(size.feature[0]); ++fi) {
                 for (size_t zi = 0; zi < static_cast<size_t>(size.spatial[2]); ++zi) {
                     for (size_t yi = 0; yi < static_cast<size_t>(size.spatial[1]); ++yi) {
                         for (size_t xi = 0; xi < static_cast<size_t>(size.spatial[0]); ++xi) {
                             auto tensor_addr = tensor(batch(bi), feature(fi), spatial(xi, yi, zi, 0));
-                            auto offset = mem.get_layout().get_linear_offset(tensor_addr);
+                            auto offset = mem->get_layout().get_linear_offset(tensor_addr);
                             ptr[offset] = data[bi][fi][xi][yi][zi];
                         }
                     }
@@ -570,8 +566,8 @@ struct mvn_random_test : ::testing::TestWithParam<mvn_basic_test_params> {
     }
 
     template <typename T>
-    void fill_random_data(memory& mem, int min, int max, int k = 8) {
-        auto size = mem.get_layout().size;
+    void fill_random_data(memory::ptr mem, int min, int max, int k = 8) {
+        auto size = mem->get_layout().size;
         auto input_data = tests::generate_random_5d<T>(size.batch[0],
                                                        size.feature[0],
                                                        size.spatial[0],
@@ -583,14 +579,14 @@ struct mvn_random_test : ::testing::TestWithParam<mvn_basic_test_params> {
         fill_data(mem, input_data);
     }
 
-    void check_result(memory& output, bool across_channels, bool normalize_variance) {
-        if (output.get_layout().data_type == data_types::f32) {
+    void check_result(memory::ptr output, bool across_channels, bool normalize_variance) {
+        if (output->get_layout().data_type == data_types::f32) {
             if (across_channels) {
                 mvn_compute_mean_across_channels<float>(output, normalize_variance);
             } else {
                 mvn_compute_mean_within_channels<float>(output, normalize_variance);
             }
-        } else if (output.get_layout().data_type == data_types::f16) {
+        } else if (output->get_layout().data_type == data_types::f16) {
             if (across_channels) {
                 mvn_compute_mean_across_channels<FLOAT16>(output, normalize_variance);
             } else {
@@ -599,11 +595,11 @@ struct mvn_random_test : ::testing::TestWithParam<mvn_basic_test_params> {
         }
     }
 
-    void execute(const mvn_basic_test_params& params, const engine& eng) {
+    void execute(const mvn_basic_test_params& params, engine& eng) {
         auto& size = params.input_size;
         auto& output_pad = params.output_pad;
 
-        auto input = memory::allocate(eng, {params.input_type, params.input_format, size});
+        auto input = eng.allocate_memory({params.input_type, params.input_format, size});
 
         switch (params.input_type) {
             case data_types::f32:
@@ -623,7 +619,7 @@ struct mvn_random_test : ::testing::TestWithParam<mvn_basic_test_params> {
         }
 
         topology topo;
-        topo.add(input_layout("input", input.get_layout()));
+        topo.add(input_layout("input", input->get_layout()));
         auto prim = mvn("mvn", "input", params.normalize_variance, 1e-10f, false, params.across_channels);
         prim.output_padding = output_pad;
         topo.add(prim);
@@ -642,7 +638,7 @@ struct mvn_random_test : ::testing::TestWithParam<mvn_basic_test_params> {
 };
 
 TEST_P(mvn_random_test, random) {
-    auto eng = tests::get_test_engine();
+    auto& eng = tests::get_test_engine();
     this->execute(GetParam(), eng);
 }
 
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/non_max_suppression_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/non_max_suppression_test.cpp
index 554e648608607a..7ca881a7b2747b 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/non_max_suppression_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/non_max_suppression_test.cpp
@@ -4,17 +4,14 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include <gtest/gtest.h>
-
 #include "test_utils.h"
 
-#include "api/topology.hpp"
-#include "api/network.hpp"
-#include "api/input_layout.hpp"
-#include "api/non_max_suppression.hpp"
-#include "api/data.hpp"
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/non_max_suppression.hpp>
+#include <cldnn/primitives/data.hpp>
 
 using namespace cldnn;
+using namespace ::tests;
 
 template <typename T>
 struct non_max_suppression_basic : public testing::Test {
@@ -73,14 +70,14 @@ struct non_max_suppression_basic : public testing::Test {
     const layout boxes_layout = layout(type_to_data_type<T>::value, format::bfyx, tensor(batch(batch_size), feature(boxes_num), spatial(1, 4)));
     const layout scores_layout = layout(type_to_data_type<T>::value, format::bfyx, tensor(batch(batch_size), feature(classes_num), spatial(1, boxes_num)));
 
-    memory get_boxes_memory(engine& engine) {
-        auto mem = memory::allocate(engine, boxes_layout);
+    memory::ptr get_boxes_memory(engine& engine) {
+        auto mem = engine.allocate_memory(boxes_layout);
         tests::set_values(mem, boxes_data);
         return mem;
     }
 
-    memory get_scores_memory(engine& engine) {
-        auto mem = memory::allocate(engine, scores_layout);
+    memory::ptr get_scores_memory(engine& engine) {
+        auto mem = engine.allocate_memory(scores_layout);
         tests::set_values(mem, scores_data);
         return mem;
     }
@@ -92,7 +89,7 @@ using nms_types = testing::Types<float, half_t>;
 TYPED_TEST_CASE(non_max_suppression_basic, nms_types);
 
 TYPED_TEST(non_max_suppression_basic, basic) {
-    auto engine = tests::get_test_engine();
+    auto& engine = tests::get_test_engine();
 
     topology topo;
     topo.add(input_layout("boxes", this->boxes_layout));
@@ -122,7 +119,7 @@ TYPED_TEST(non_max_suppression_basic, basic) {
     };
 
     auto out_mem = result.at("nms").get_memory();
-    auto out_ptr = out_mem.pointer<int>();
+    cldnn::mem_lock<int> out_ptr(out_mem, get_test_stream());
 
     ASSERT_EQ(expected_out.size(), out_ptr.size());
     for (size_t i = 0; i < expected_out.size(); ++i) {
@@ -131,9 +128,9 @@ TYPED_TEST(non_max_suppression_basic, basic) {
 }
 
 TYPED_TEST(non_max_suppression_basic, num_per_class) {
-    auto engine = tests::get_test_engine();
+    auto& engine = tests::get_test_engine();
 
-    auto num_per_class_mem = memory::allocate(engine, layout(data_types::f32, format::bfyx, tensor(batch(1))));
+    auto num_per_class_mem = engine.allocate_memory(layout(data_types::f32, format::bfyx, tensor(batch(1))));
     tests::set_values(num_per_class_mem, { 1.f });
 
     topology topo;
@@ -165,7 +162,7 @@ TYPED_TEST(non_max_suppression_basic, num_per_class) {
     };
 
     auto out_mem = result.at("nms").get_memory();
-    auto out_ptr = out_mem.pointer<int>();
+    cldnn::mem_lock<int> out_ptr(out_mem, get_test_stream());
 
     ASSERT_EQ(expected_out.size(), out_ptr.size());
     for (size_t i = 0; i < expected_out.size(); ++i) {
@@ -174,11 +171,11 @@ TYPED_TEST(non_max_suppression_basic, num_per_class) {
 }
 
 TYPED_TEST(non_max_suppression_basic, iou_threshold) {
-    auto engine = tests::get_test_engine();
+    auto& engine = tests::get_test_engine();
 
-    auto num_per_class_mem = memory::allocate(engine, layout(data_types::f32, format::bfyx, tensor(batch(1))));
+    auto num_per_class_mem = engine.allocate_memory(layout(data_types::f32, format::bfyx, tensor(batch(1))));
     tests::set_values(num_per_class_mem, { 3.f });
-    auto iou_threshold_mem = memory::allocate(engine, layout(data_types::f32, format::bfyx, tensor(batch(1))));
+    auto iou_threshold_mem = engine.allocate_memory(layout(data_types::f32, format::bfyx, tensor(batch(1))));
     tests::set_values(iou_threshold_mem, { 0.4f });
 
     topology topo;
@@ -211,7 +208,7 @@ TYPED_TEST(non_max_suppression_basic, iou_threshold) {
     };
 
     auto out_mem = result.at("nms").get_memory();
-    auto out_ptr = out_mem.pointer<int>();
+    cldnn::mem_lock<int> out_ptr(out_mem, get_test_stream());
 
     ASSERT_EQ(expected_out.size(), out_ptr.size());
     for (size_t i = 0; i < expected_out.size(); ++i) {
@@ -220,13 +217,13 @@ TYPED_TEST(non_max_suppression_basic, iou_threshold) {
 }
 
 TYPED_TEST(non_max_suppression_basic, score_threshold) {
-    auto engine = tests::get_test_engine();
+    auto& engine = tests::get_test_engine();
 
-    auto num_per_class_mem = memory::allocate(engine, layout(data_types::f32, format::bfyx, tensor(batch(1))));
+    auto num_per_class_mem = engine.allocate_memory(layout(data_types::f32, format::bfyx, tensor(batch(1))));
     tests::set_values(num_per_class_mem, { 3.f });
-    auto iou_threshold_mem = memory::allocate(engine, layout(data_types::f32, format::bfyx, tensor(batch(1))));
+    auto iou_threshold_mem = engine.allocate_memory(layout(data_types::f32, format::bfyx, tensor(batch(1))));
     tests::set_values(iou_threshold_mem, { 0.4f });
-    auto score_threshold_mem = memory::allocate(engine, layout(data_types::f32, format::bfyx, tensor(batch(1))));
+    auto score_threshold_mem = engine.allocate_memory(layout(data_types::f32, format::bfyx, tensor(batch(1))));
     tests::set_values(score_threshold_mem, { 0.4f });
 
     topology topo;
@@ -260,7 +257,7 @@ TYPED_TEST(non_max_suppression_basic, score_threshold) {
     };
 
     auto out_mem = result.at("nms").get_memory();
-    auto out_ptr = out_mem.pointer<int>();
+    cldnn::mem_lock<int> out_ptr(out_mem, get_test_stream());
 
     ASSERT_EQ(expected_out.size(), out_ptr.size());
     for (size_t i = 0; i < expected_out.size(); ++i) {
@@ -269,15 +266,15 @@ TYPED_TEST(non_max_suppression_basic, score_threshold) {
 }
 
 TYPED_TEST(non_max_suppression_basic, soft_nms_sigma) {
-    auto engine = tests::get_test_engine();
+    auto& engine = tests::get_test_engine();
 
-    auto num_per_class_mem = memory::allocate(engine, layout(data_types::f32, format::bfyx, tensor(batch(1))));
+    auto num_per_class_mem = engine.allocate_memory(layout(data_types::f32, format::bfyx, tensor(batch(1))));
     tests::set_values(num_per_class_mem, { 3.f });
-    auto iou_threshold_mem = memory::allocate(engine, layout(data_types::f32, format::bfyx, tensor(batch(1))));
+    auto iou_threshold_mem = engine.allocate_memory(layout(data_types::f32, format::bfyx, tensor(batch(1))));
     tests::set_values(iou_threshold_mem, { 0.4f });
-    auto score_threshold_mem = memory::allocate(engine, layout(data_types::f32, format::bfyx, tensor(batch(1))));
+    auto score_threshold_mem = engine.allocate_memory(layout(data_types::f32, format::bfyx, tensor(batch(1))));
     tests::set_values(score_threshold_mem, { 0.4f });
-    auto soft_nms_sigma_mem = memory::allocate(engine, layout(data_types::f32, format::bfyx, tensor(batch(1))));
+    auto soft_nms_sigma_mem = engine.allocate_memory(layout(data_types::f32, format::bfyx, tensor(batch(1))));
     tests::set_values(soft_nms_sigma_mem, { 0.5f });
 
     topology topo;
@@ -312,7 +309,7 @@ TYPED_TEST(non_max_suppression_basic, soft_nms_sigma) {
     };
 
     auto out_mem = result.at("nms").get_memory();
-    auto out_ptr = out_mem.pointer<int>();
+    cldnn::mem_lock<int> out_ptr(out_mem, get_test_stream());
 
     ASSERT_EQ(expected_out.size(), out_ptr.size());
     for (size_t i = 0; i < expected_out.size(); ++i) {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/normalizel2_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/normalizel2_gpu_test.cpp
index 00a400cce57453..a96185c7d8f9c3 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/normalizel2_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/normalizel2_gpu_test.cpp
@@ -3,15 +3,12 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/input_layout.hpp>
-#include <api/normalize.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/data.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/normalize.hpp>
+#include <cldnn/primitives/data.hpp>
 
 #include <vector>
 #include <iostream>
@@ -23,15 +20,15 @@ TEST(normalizel2_f32_gpu, basic) {
     //  Input  : 1x2x3x3
     //  Output : 1x2x3x3
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const unsigned b = 1;
     const unsigned f = 2;
     const unsigned y = 3;
     const unsigned x = 3;
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {b, f, y, x}});
-    auto weights = memory::allocate(engine, {data_types::f32, format::bfyx, {1, f, 1, 1}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {b, f, y, x}});
+    auto weights = engine.allocate_memory({data_types::f32, format::bfyx, {1, f, 1, 1}});
 
     std::vector<float> inputVals(b * f * y * x);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -47,7 +44,7 @@ TEST(normalizel2_f32_gpu, basic) {
     set_values(weights, weightVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(data("Input1", weights));
     topology.add(normalize("normalizel2", "Input0", "Input1", false));
 
@@ -58,7 +55,7 @@ TEST(normalizel2_f32_gpu, basic) {
     auto outputs = network.execute();
 
     auto output = outputs.at("normalizel2").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {0.f,
                                            0.0995037f,
@@ -88,15 +85,15 @@ TEST(normalizel2_f32_gpu, basic2) {
     //  Input  : 1x2x3x3
     //  Output : 1x2x3x3
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const unsigned b = 1;
     const unsigned f = 2;
     const unsigned y = 3;
     const unsigned x = 3;
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {b, f, y, x}});
-    auto weights = memory::allocate(engine, {data_types::f32, format::bfyx, {1, f, 1, 1}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {b, f, y, x}});
+    auto weights = engine.allocate_memory({data_types::f32, format::bfyx, {1, f, 1, 1}});
 
     std::vector<float> inputVals(b * f * y * x);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -112,7 +109,7 @@ TEST(normalizel2_f32_gpu, basic2) {
     set_values(weights, weightVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(data("Input1", weights));
     topology.add(normalize("normalizel2", "Input0", "Input1", true));
 
@@ -123,7 +120,7 @@ TEST(normalizel2_f32_gpu, basic2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("normalizel2").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {0.f,
                                            0.0236691f,
@@ -153,15 +150,15 @@ TEST(normalizel2_int8_gpu, basic) {
     //  Input  : 1x2x3x3
     //  Output : 1x2x3x3
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const unsigned b = 1;
     const unsigned f = 2;
     const unsigned y = 3;
     const unsigned x = 3;
 
-    auto input = memory::allocate(engine, {data_types::i8, format::bfyx, {b, f, y, x}});
-    auto weights = memory::allocate(engine, {data_types::f32, format::bfyx, {1, f, 1, 1}});
+    auto input = engine.allocate_memory({data_types::i8, format::bfyx, {b, f, y, x}});
+    auto weights = engine.allocate_memory({data_types::f32, format::bfyx, {1, f, 1, 1}});
 
     std::vector<int8_t> inputVals(b * f * y * x);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -177,7 +174,7 @@ TEST(normalizel2_int8_gpu, basic) {
     set_values(weights, weightVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(data("Input1", weights));
     topology.add(normalize("normalizel2", "Input0", "Input1", false));
 
@@ -188,7 +185,7 @@ TEST(normalizel2_int8_gpu, basic) {
     auto outputs = network.execute();
 
     auto output = outputs.at("normalizel2").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {0.f,
                                            0.0995037f,
@@ -218,15 +215,15 @@ TEST(normalizel2_int8_gpu, basic2) {
     //  Input  : 1x2x3x3
     //  Output : 1x2x3x3
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const unsigned b = 1;
     const unsigned f = 2;
     const unsigned y = 3;
     const unsigned x = 3;
 
-    auto input = memory::allocate(engine, {data_types::i8, format::bfyx, {b, f, y, x}});
-    auto weights = memory::allocate(engine, {data_types::f32, format::bfyx, {1, f, 1, 1}});
+    auto input = engine.allocate_memory({data_types::i8, format::bfyx, {b, f, y, x}});
+    auto weights = engine.allocate_memory({data_types::f32, format::bfyx, {1, f, 1, 1}});
 
     std::vector<int8_t> inputVals(b * f * y * x);
     std::generate(inputVals.begin(), inputVals.end(), []() {
@@ -242,7 +239,7 @@ TEST(normalizel2_int8_gpu, basic2) {
     set_values(weights, weightVals);
 
     topology topology;
-    topology.add(input_layout("Input0", input.get_layout()));
+    topology.add(input_layout("Input0", input->get_layout()));
     topology.add(data("Input1", weights));
     topology.add(normalize("normalizel2", "Input0", "Input1", true));
 
@@ -253,7 +250,7 @@ TEST(normalizel2_int8_gpu, basic2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("normalizel2").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {0.f,
                                            0.0236691f,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/one_hot_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/one_hot_gpu_test.cpp
index cb510e4a685119..08bb8690be0976 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/one_hot_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/one_hot_gpu_test.cpp
@@ -3,17 +3,11 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/engine.hpp>
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/one_hot.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include "test_utils.h"
 
-#include "test_utils/test_utils.h"
-#include "test_utils/uniform_quantized_real_distribution.hpp"
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/one_hot.hpp>
 
 #include <cstddef>
 
@@ -83,13 +77,13 @@ void generic_one_hot_test_int(cldnn::format test_input_fmt, int input_b, int inp
     VVVVF<T> input_rnd = generate_random_4d<T>(input_b, input_f, input_y, input_x, min_random, max_random);
     VF<T> input_rnd_vec = flatten_4d<T>(test_input_fmt, input_rnd);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(input_b, input_f, input_x, input_y);
-    auto input = memory::allocate(engine, { type_to_data_type<T>::value, test_input_fmt, input_tensor });
+    auto input = engine.allocate_memory({ type_to_data_type<T>::value, test_input_fmt, input_tensor });
     set_values(input, input_rnd_vec);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output", "input", shape, one_hot_axis));
 
     network network(engine, topology);
@@ -99,8 +93,8 @@ void generic_one_hot_test_int(cldnn::format test_input_fmt, int input_b, int inp
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<T>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<T> output_ptr(output_memory, get_test_stream());
 
     VVVVF<T> output_cpu = one_hot_cpu<T>(input_rnd, one_hot_axis, one_hot_limit, input_padding_y, input_padding_x, output_padding_y, output_padding_x);
     EXPECT_EQ(output_layout.format.value, test_input_fmt.value);
@@ -166,11 +160,11 @@ TEST(one_hot_gpu_i32, bfzyx_ax4) {
 
     VF<int32_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i32, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i32, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -182,8 +176,8 @@ TEST(one_hot_gpu_i32, bfzyx_ax4) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int32_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int32_t> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -225,11 +219,11 @@ TEST(one_hot_gpu_i64, bfzyx_ax4) {
 
     VF<int64_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i64, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i64, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -241,8 +235,8 @@ TEST(one_hot_gpu_i64, bfzyx_ax4) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int64_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int64_t> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -284,11 +278,11 @@ TEST(one_hot_gpu_i32_to_f32, bfyx_ax4) {
 
     VF<int32_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i32, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i32, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, data_types::f32, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -300,8 +294,8 @@ TEST(one_hot_gpu_i32_to_f32, bfyx_ax4) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -338,11 +332,11 @@ TEST(one_hot_gpu_i64_to_f32, bfyx_ax4) {
 
     VF<int64_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i64, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i64, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, data_types::f32, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -354,8 +348,8 @@ TEST(one_hot_gpu_i64_to_f32, bfyx_ax4) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -389,11 +383,11 @@ TEST(one_hot_gpu_i32, bfzyx_ax0) {
 
     VF<int32_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i32, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i32, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -405,8 +399,8 @@ TEST(one_hot_gpu_i32, bfzyx_ax0) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int32_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int32_t> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -444,11 +438,11 @@ TEST(one_hot_gpu_i64, bfzyx_ax0) {
 
     VF<int64_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i64, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i64, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -460,8 +454,8 @@ TEST(one_hot_gpu_i64, bfzyx_ax0) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int64_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int64_t> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -499,11 +493,11 @@ TEST(one_hot_gpu_i32, bfzyx_ax1) {
 
     VF<int32_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i32, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i32, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -515,8 +509,8 @@ TEST(one_hot_gpu_i32, bfzyx_ax1) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int32_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int32_t> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -554,11 +548,11 @@ TEST(one_hot_gpu_i64, bfzyx_ax1) {
 
     VF<int64_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i64, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i64, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -570,8 +564,8 @@ TEST(one_hot_gpu_i64, bfzyx_ax1) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int64_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int64_t> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -609,11 +603,11 @@ TEST(one_hot_gpu_i32, bfzyx_ax2) {
 
     VF<int32_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i32, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i32, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -625,8 +619,8 @@ TEST(one_hot_gpu_i32, bfzyx_ax2) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int32_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int32_t> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -664,11 +658,11 @@ TEST(one_hot_gpu_i64, bfzyx_ax2) {
 
     VF<int64_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i64, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i64, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -680,8 +674,8 @@ TEST(one_hot_gpu_i64, bfzyx_ax2) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int64_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int64_t> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -719,11 +713,11 @@ TEST(one_hot_gpu_i32, bfzyx_ax3) {
 
     VF<int32_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i32, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i32, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -735,8 +729,8 @@ TEST(one_hot_gpu_i32, bfzyx_ax3) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int32_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int32_t> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -774,11 +768,11 @@ TEST(one_hot_gpu_i64, bfzyx_ax3) {
 
     VF<int64_t> input_rnd_vec = {0, 1};
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     tensor input_tensor(in_b, in_f, in_x, in_y);
-    auto input = memory::allocate(engine, { data_types::i64, format::bfyx, input_tensor });
+    auto input = engine.allocate_memory({ data_types::i64, format::bfyx, input_tensor });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output","input", shape, one_hot_axis));
 
     set_values(input, input_rnd_vec);
@@ -790,8 +784,8 @@ TEST(one_hot_gpu_i64, bfzyx_ax3) {
     EXPECT_EQ(outputs.begin()->first, "output");
 
     auto output_memory = outputs.at("output").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<int64_t>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<int64_t> output_ptr(output_memory, get_test_stream());
 
     tensor output_tensor = output_layout.get_buffer_size();
     int z_size = output_tensor.spatial[2];
@@ -818,11 +812,11 @@ TEST(one_hot_gpu_i64, bfzyx_ax3) {
 }
 
 TEST(one_hot_error, basic_error_wrong_axis) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::i32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::i32, format::bfyx,{ 1, 1, 1, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output", "input", tensor(1, 1, 1, 50), 5));
 
     std::string msg_to_find = "Incorrect parameters configuration: one_hot_axis should be less or equal to 4.";
@@ -830,11 +824,11 @@ TEST(one_hot_error, basic_error_wrong_axis) {
 }
 
 TEST(one_hot_error, basic_error_bad_shape) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::i32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::i32, format::bfyx,{ 1, 1, 1, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(one_hot("output", "input", tensor(1, 5, 1, 50), 2));
 
     std::string msg_to_find = "Incorrect parameters configuration: shape does not fit input size.";
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/permute_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/permute_gpu_test.cpp
index d22a5bf8077e9d..7fb709d9845984 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/permute_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/permute_gpu_test.cpp
@@ -3,31 +3,28 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/permute.hpp"
-#include "api/reorder.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/data.hpp>
-#include <api/fully_connected.hpp>
-#include <api/reshape.hpp>
-#include <api/crop.hpp>
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/permute.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/fully_connected.hpp>
+#include <cldnn/primitives/reshape.hpp>
+#include <cldnn/primitives/crop.hpp>
+
 #include <cmath>
-#include <gmock/gmock.h>
 #include <limits>
 #include <type_traits>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 using namespace testing;
 
 TEST(permute_gpu_f32, output_ordering_test)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     std::vector<std::vector<int32_t>> input_tensors =
     {
@@ -60,9 +57,9 @@ TEST(permute_gpu_f32, output_ordering_test)
             for (auto const& perm : permutations)
             {
 
-                auto input = memory::allocate(engine, { data_types::f32, fr, tensor(inp_t) });
+                auto input = engine.allocate_memory({ data_types::f32, fr, tensor(inp_t) });
                 topology topology(
-                    input_layout("input", input.get_layout()),
+                    input_layout("input", input->get_layout()),
                     permute("permute", "input", perm));
 
                 network network(engine, topology);
@@ -72,7 +69,7 @@ TEST(permute_gpu_f32, output_ordering_test)
                 auto output_mem = output.get_memory();
                 EXPECT_EQ(outputs.size(), size_t(1));
                 auto ref_tensor = get_permutation(inp_t, perm);
-                auto out_tensor = output_mem.get_layout().size;
+                auto out_tensor = output_mem->get_layout().size;
                 EXPECT_EQ(out_tensor.batch[0], ref_tensor[0]);
                 EXPECT_EQ(out_tensor.feature[0], ref_tensor[1]);
                 EXPECT_EQ(out_tensor.spatial[0], ref_tensor[2]);
@@ -95,9 +92,9 @@ TEST(permute_gpu_f32, basic_bfyx_permute_0_1_2_3)
     //
     //  Output = input
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
 
     std::vector<float> values =
     {
@@ -117,7 +114,7 @@ TEST(permute_gpu_f32, basic_bfyx_permute_0_1_2_3)
     set_values(input, values);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 1, 2, 3 }));
 
     network network(engine, topology);
@@ -129,7 +126,7 @@ TEST(permute_gpu_f32, basic_bfyx_permute_0_1_2_3)
 
     auto output = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 24; i++)
     {
         EXPECT_FLOAT_EQ(values[i], output_ptr[i]);
@@ -157,9 +154,9 @@ TEST(permute_gpu_f32, basic_bfyx_permute_0_1_3_2)
     //  f1: b0:  -15 -15 b1:   -15   -15
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
 
     set_values(input, {
         1.0f,  2.0f, -15.f,
@@ -176,7 +173,7 @@ TEST(permute_gpu_f32, basic_bfyx_permute_0_1_3_2)
         });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 1, 3, 2 }));
 
     network network(engine, topology);
@@ -206,7 +203,7 @@ TEST(permute_gpu_f32, basic_bfyx_permute_0_1_3_2)
         -15.0f,  -15.0f,
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 24; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -216,14 +213,14 @@ TEST(permute_gpu_f32, basic_bfyx_permute_0_1_3_2)
 
 TEST(permute_gpu_f32, basic_yxfb_permute_1_0_2_3)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 100, 64, 1 } });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 100, 64, 1 } });
 
     tests::set_random_values<float>(input_mem);
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
         permute("permute", "input", { 1, 0, 2, 3 }));
 
     network network(engine, topology);
@@ -235,8 +232,8 @@ TEST(permute_gpu_f32, basic_yxfb_permute_1_0_2_3)
 
     auto output = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output.pointer<float>();
-    auto input_ptr = input_mem.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input_mem, get_test_stream());
     for (int i = 0; i < 6400; i++)
     {
         EXPECT_FLOAT_EQ(input_ptr[i], output_ptr[i]);
@@ -265,9 +262,9 @@ TEST(permute_gpu_f32, basic_bfyx_permute_0_1_3_2_input_padding)
     //  f1: b0:  -15 -15 b1:   -15   -15
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
 
     set_values(input, {
         1.0f,  2.0f, -15.f,
@@ -284,8 +281,8 @@ TEST(permute_gpu_f32, basic_bfyx_permute_0_1_3_2_input_padding)
         });
 
     topology topology(
-        input_layout("input", input.get_layout()),
-        reorder("reorder", "input", input.get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })),
+        input_layout("input", input->get_layout()),
+        reorder("reorder", "input", input->get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })),
         permute("permute", "reorder", { 0, 1, 3, 2 }));
 
     network network(engine, topology);
@@ -315,7 +312,7 @@ TEST(permute_gpu_f32, basic_bfyx_permute_0_1_3_2_input_padding)
         -15.0f,  -15.0f,
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 24; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -329,9 +326,9 @@ TEST(permute_gpu_f32, basic_yxfb_permute_batch_with_feature)
     //  Permute order       : { 1, 0, 2, 3 }
     //  Output              : yxfb:2x8x1x1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 8, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 8, 2, 1, 1 } });
 
     set_values(input, {
         //b0 - b7 for f=0
@@ -342,7 +339,7 @@ TEST(permute_gpu_f32, basic_yxfb_permute_batch_with_feature)
         });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 1, 0, 2, 3 }));
 
     network network(engine, topology);
@@ -353,7 +350,7 @@ TEST(permute_gpu_f32, basic_yxfb_permute_batch_with_feature)
     EXPECT_EQ(outputs.begin()->first, "permute");
 
     auto output = outputs.begin()->second.get_memory();
-    auto out_tensor = output.get_layout().size;
+    auto out_tensor = output->get_layout().size;
     EXPECT_EQ(out_tensor.batch[0], 2);
     EXPECT_EQ(out_tensor.feature[0], 8);
     EXPECT_EQ(out_tensor.spatial[0], 1);
@@ -370,7 +367,7 @@ TEST(permute_gpu_f32, basic_yxfb_permute_batch_with_feature)
         5.2f, 8.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -384,9 +381,9 @@ TEST(permute_gpu_f32, basic_bfyx_permute_batch_with_feature)
     //  Permute order       : { 1, 0, 2, 3 }
     //  Output              : yxfb:2x8x1x1
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 8, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 8, 1, 1 } });
 
     set_values(input, {
         //f0 - f7 for b=0
@@ -397,7 +394,7 @@ TEST(permute_gpu_f32, basic_bfyx_permute_batch_with_feature)
         });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 1, 0, 2, 3 }));
 
     network network(engine, topology);
@@ -408,7 +405,7 @@ TEST(permute_gpu_f32, basic_bfyx_permute_batch_with_feature)
     EXPECT_EQ(outputs.begin()->first, "permute");
 
     auto output = outputs.begin()->second.get_memory();
-    auto out_tensor = output.get_layout().size;
+    auto out_tensor = output->get_layout().size;
     EXPECT_EQ(out_tensor.batch[0], 8);
     EXPECT_EQ(out_tensor.feature[0], 2);
     EXPECT_EQ(out_tensor.spatial[0], 1);
@@ -425,7 +422,7 @@ TEST(permute_gpu_f32, basic_bfyx_permute_batch_with_feature)
         5.2f, 8.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -436,9 +433,9 @@ TEST(permute_gpu_f32, basic_bfyx_permute_batch_with_feature)
 template<data_types DType>
 void permute_test_with_reorder()
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
 
     set_values(input, {
         1.0f,  2.0f, -15.f,
@@ -455,7 +452,7 @@ void permute_test_with_reorder()
         });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", { DType, format::bfyx,{ 2, 2, 3, 2 } }),
         permute("permute", "reorder", { 0, 1, 3, 2 }),
         reorder("reorder_out", "permute", { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } }));
@@ -487,7 +484,7 @@ void permute_test_with_reorder()
         -15.0f,  -15.0f,
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 24; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -512,8 +509,8 @@ TEST(permute_fuse_reorder_gpu_f32, basic_b_fs_yx_fsv4_permute_1_8_16_1)
     //  Permute1 order       : {0, 3, 1, 2}
     //  Permute2 order       : {0, 2, 3, 1}
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, {1, 8, 1, 16}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, {1, 8, 1, 16}});
 
     std::vector<float> values = {
             0.0f, 1.0f, 2.0f, 3.0f,
@@ -553,7 +550,7 @@ TEST(permute_fuse_reorder_gpu_f32, basic_b_fs_yx_fsv4_permute_1_8_16_1)
     set_values(input, values);
     // unfused
     topology topology_unfused(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder1", "input", format::b_fs_yx_fsv4, data_types::f32),
         permute("permute", "reorder1", { 0, 3, 1, 2}),
         reorder("reorder2", "permute", format::bfyx, data_types::f32),
@@ -568,7 +565,7 @@ TEST(permute_fuse_reorder_gpu_f32, basic_b_fs_yx_fsv4_permute_1_8_16_1)
 
     // fused network
     topology topology_fused(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder1", "input", format::b_fs_yx_fsv4, data_types::f32),
         permute("permute", "reorder1", { 0, 3, 1, 2}),
         reorder("reorder2", "permute", format::bfyx, data_types::f32), // to be fused to previous permute
@@ -582,11 +579,11 @@ TEST(permute_fuse_reorder_gpu_f32, basic_b_fs_yx_fsv4_permute_1_8_16_1)
     auto outputs_fused = fused.execute();
     auto outputs_unfused = unfused.execute();
     auto output_fused = outputs_fused.begin()->second.get_memory();
-    auto output_fused_ptr = output_fused.pointer<float>();
+    cldnn::mem_lock<float> output_fused_ptr(output_fused, get_test_stream());
     auto output_unfused = outputs_unfused.begin()->second.get_memory();
-    auto output_unfused_ptr = output_unfused.pointer<float>();
-    EXPECT_EQ(output_fused.get_layout().format, cldnn::format::bfyx);
-    EXPECT_EQ(output_unfused.get_layout().format, cldnn::format::bfyx);
+    cldnn::mem_lock<float> output_unfused_ptr(output_unfused, get_test_stream());
+    EXPECT_EQ(output_fused->get_layout().format, cldnn::format::bfyx);
+    EXPECT_EQ(output_unfused->get_layout().format, cldnn::format::bfyx);
     EXPECT_EQ(fused.get_executed_primitives().size(), 4);
     EXPECT_EQ(unfused.get_executed_primitives().size(), 5);
 
@@ -599,13 +596,13 @@ TEST(permute_fuse_reorder_gpu_f32, basic_b_fs_yx_fsv4_permute_1_8_16_1)
 
 TEST(fc_permute_crop_gpu, basic_permute_yxfb)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 5, 1, 512 } });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 5, 1, 512 } });
 
     //Topolgy creates permute which "repalces" the batch with the feature.
     topology topology(
-        input_layout("input", input_mem.get_layout()),  // yxfb {1, 5, 1, 512 }}
+        input_layout("input", input_mem->get_layout()),  // yxfb {1, 5, 1, 512 }}
         permute("permute", "input", { 1, 0, 2, 3 })  // yxfb {5, 1, 1, 512}  --- without permute fix yxfb {1, 5, 512, 1}
     );
 
@@ -617,25 +614,25 @@ TEST(fc_permute_crop_gpu, basic_permute_yxfb)
     EXPECT_EQ(outputs.begin()->first, "permute");
 
     auto output = outputs.begin()->second.get_memory();
-    auto out_tensor = output.get_layout().size;
+    auto out_tensor = output->get_layout().size;
     EXPECT_EQ(out_tensor.batch[0], 5);
     EXPECT_EQ(out_tensor.feature[0], 1);
     EXPECT_EQ(out_tensor.spatial[0], 1);
     EXPECT_EQ(out_tensor.spatial[1], 512);
-    EXPECT_EQ(output.get_layout().format, cldnn::format::yxfb);
+    EXPECT_EQ(output->get_layout().format, cldnn::format::yxfb);
 }
 
 TEST(fc_permute_crop_gpu, basic_0)
 {
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::yxfb,{ 5, 11264, 1, 1 } });
-    auto weights_mem = memory::allocate(engine, { data_types::f32, format::yxio,{ 512, 11264, 1, 1 } });
-    auto bias_mem = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 512, 1 } });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::yxfb,{ 5, 11264, 1, 1 } });
+    auto weights_mem = engine.allocate_memory({ data_types::f32, format::yxio,{ 512, 11264, 1, 1 } });
+    auto bias_mem = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 512, 1 } });
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),                   // bfyx {5, 11264, 1, 1}}
+        input_layout("input", input_mem->get_layout()),                   // bfyx {5, 11264, 1, 1}}
         data("weights", weights_mem),
         data("bias", bias_mem),
         fully_connected("fully_connected", "input", "weights", "bias"),  // yxfb {5, 512, 1, 1}
@@ -652,25 +649,25 @@ TEST(fc_permute_crop_gpu, basic_0)
     EXPECT_EQ(outputs.begin()->first, "crop");
 
     auto output = outputs.begin()->second.get_memory();
-    auto out_tensor = output.get_layout().size;
+    auto out_tensor = output->get_layout().size;
     EXPECT_EQ(out_tensor.batch[0], 1);
     EXPECT_EQ(out_tensor.feature[0], 1);
     EXPECT_EQ(out_tensor.spatial[0], 1);
     EXPECT_EQ(out_tensor.spatial[1], 512);
-    EXPECT_EQ(output.get_layout().format, cldnn::format::yxfb);
+    EXPECT_EQ(output->get_layout().format, cldnn::format::yxfb);
 }
 
 TEST(fc_permute_gpu, basic_permute_bfyx)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 5, 1, 256 } });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 5, 1, 256 } });
 
     tests::set_random_values<float>(input_mem);
 
     //Topolgy creates permute which "repalces" the batch with the feature.
     topology topology(
-        input_layout("input", input_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
         permute("permute", "input", { 1, 0, 2, 3 })
     );
 
@@ -682,15 +679,15 @@ TEST(fc_permute_gpu, basic_permute_bfyx)
     EXPECT_EQ(outputs.begin()->first, "permute");
 
     auto output = outputs.begin()->second.get_memory();
-    auto out_tensor = output.get_layout().size;
+    auto out_tensor = output->get_layout().size;
     EXPECT_EQ(out_tensor.batch[0], 5);
     EXPECT_EQ(out_tensor.feature[0], 1);
     EXPECT_EQ(out_tensor.spatial[0], 1);
     EXPECT_EQ(out_tensor.spatial[1], 256);
-    EXPECT_EQ(output.get_layout().format, cldnn::format::bfyx);
+    EXPECT_EQ(output->get_layout().format, cldnn::format::bfyx);
 
-    auto input_ptr = input_mem.pointer<float>();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> input_ptr(input_mem, get_test_stream());
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 5 * 256; i++)
         EXPECT_NEAR(input_ptr[i], output_ptr[i], 1e-3f);
 
@@ -698,7 +695,7 @@ TEST(fc_permute_gpu, basic_permute_bfyx)
 
 TEST(permute_gpu_f32, permute_bfwzyx)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int b = 1;
     const int f = 2;
     const int x = 3;
@@ -708,8 +705,8 @@ TEST(permute_gpu_f32, permute_bfwzyx)
     std::vector<uint16_t> permute_order = { 1, 0, 5, 4, 3, 2 };
 
     auto input_size = cldnn::tensor(batch(b), feature(f), spatial(x, y, z, w));
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfwzyx, input_size });
-    auto input_data = generate_random_1d<float>(input_mem.get_layout().count(), -1, 1);
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::bfwzyx, input_size });
+    auto input_data = generate_random_1d<float>(input_mem->get_layout().count(), -1, 1);
 
     set_values(input_mem, input_data);
 
@@ -726,11 +723,11 @@ TEST(permute_gpu_f32, permute_bfwzyx)
         auto in_index = cldnn::tensor(batch(bi), feature(fi), spatial(xi, yi, zi, wi));
         auto out_index = cldnn::tensor(batch(fi), feature(bi), spatial(wi, zi, yi, xi));
         expected_output[expected_layout.get_linear_offset(out_index)] =
-            input_data[input_mem.get_layout().get_linear_offset(in_index)];
+            input_data[input_mem->get_layout().get_linear_offset(in_index)];
     }
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
         permute("permute", "input", permute_order)
     );
 
@@ -742,16 +739,16 @@ TEST(permute_gpu_f32, permute_bfwzyx)
     EXPECT_EQ(outputs.begin()->first, "permute");
 
     auto output = outputs.begin()->second.get_memory();
-    auto out_tensor = output.get_layout().size;
+    auto out_tensor = output->get_layout().size;
     EXPECT_EQ(out_tensor.batch[0], 2);
     EXPECT_EQ(out_tensor.feature[0], 1);
     EXPECT_EQ(out_tensor.spatial[0], 6);
     EXPECT_EQ(out_tensor.spatial[1], 5);
     EXPECT_EQ(out_tensor.spatial[2], 4);
     EXPECT_EQ(out_tensor.spatial[3], 3);
-    EXPECT_EQ(output.get_layout().format, cldnn::format::bfwzyx);
+    EXPECT_EQ(output->get_layout().format, cldnn::format::bfwzyx);
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < output_ptr.size(); ++i)
     {
@@ -787,7 +784,7 @@ TEST(permute_gpu_f32, 6D_reshape_permute_reshape)
     // 0 2   0 2   0 2   0 2
     // 1 3   1 3   1 3   1 3
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int b = 1;
     const int f = 4;
     const int x = 2;
@@ -800,7 +797,7 @@ TEST(permute_gpu_f32, 6D_reshape_permute_reshape)
     std::vector<uint16_t> permute_order = { 0, 1, 5, 4, 2, 3 };
 
     auto input_size = cldnn::tensor(batch(b), feature(f), spatial(x, y));
-    auto input_mem = memory::allocate(engine, { data_types::f32, format::bfyx, input_size });
+    auto input_mem = engine.allocate_memory({ data_types::f32, format::bfyx, input_size });
     std::vector<float> input_data = {
         0.f, 0.f, 0.f, 0.f,
         1.f, 1.f, 1.f, 1.f,
@@ -818,7 +815,7 @@ TEST(permute_gpu_f32, 6D_reshape_permute_reshape)
     set_values(input_mem, input_data);
 
     topology topology(
-        input_layout("input", input_mem.get_layout()),
+        input_layout("input", input_mem->get_layout()),
         reorder("input_6d", "input", { data_types::f32, format::bfwzyx, cldnn::tensor(batch(b), feature(f), spatial(x, y)) }),
         reshape("reshape_4_to_6", "input_6d", cldnn::tensor(batch(b), feature(f_reshape), spatial(x, y, z_reshape, w_reshape))),
         permute("permute", "reshape_4_to_6", permute_order),
@@ -835,7 +832,7 @@ TEST(permute_gpu_f32, 6D_reshape_permute_reshape)
 
     auto output = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < output_ptr.size(); ++i)
     {
@@ -847,9 +844,9 @@ TEST(permute_gpu_f32, basic_bfzyx_permute_0_2_3_4_1)
     //  Input               : bfzyx:2x2x2x2x3
     //  Permute order       : { 0,2,3,4,1 }
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 2, 3, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 2, 3, 2, 2 } });
 
     set_values(input, {
         1.0f,  2.0f, -15.f, //B0, F0,   // z0 y0 x-3
@@ -874,7 +871,7 @@ TEST(permute_gpu_f32, basic_bfzyx_permute_0_2_3_4_1)
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 2, 3, 4, 1 }));
 
     network network(engine, topology);
@@ -885,14 +882,14 @@ TEST(permute_gpu_f32, basic_bfzyx_permute_0_2_3_4_1)
     EXPECT_EQ(outputs.begin()->first, "permute");
 
     auto output = outputs.begin()->second.get_memory();
-    auto out_tensor = output.get_layout().size;
+    auto out_tensor = output->get_layout().size;
     EXPECT_EQ(out_tensor.batch[0], 2);
     EXPECT_EQ(out_tensor.feature[0], 3);
     EXPECT_EQ(out_tensor.spatial[0], 2);
     EXPECT_EQ(out_tensor.spatial[1], 2);
     EXPECT_EQ(out_tensor.spatial[2], 2);
 
-    EXPECT_EQ(output.get_layout().format, cldnn::format::bfzyx);
+    EXPECT_EQ(output->get_layout().format, cldnn::format::bfzyx);
 
     float answers[48] = {
         1.0f, 3.0f, 2.0f, 4.0f,
@@ -909,7 +906,7 @@ TEST(permute_gpu_f32, basic_bfzyx_permute_0_2_3_4_1)
         -15.0f, -15.0f, -15.0f, -15.0f,
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 48; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -929,9 +926,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, normal_bfyx_0_3_1_2) {
 
     constexpr size_t array_size = 256;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 8, 8, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 8, 8, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -941,7 +938,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, normal_bfyx_0_3_1_2) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 3, 1, 2 }));
 
     network network(engine, topology);
@@ -972,7 +969,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, normal_bfyx_0_3_1_2) {
         142.f, 158.f, 174.f, 190.f, 206.f, 222.f, 238.f, 254.f, 143.f, 159.f, 175.f, 191.f, 207.f, 223.f, 239.f, 255.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -985,9 +982,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, f_remainder_bfyx_0_3_1_2) {
 
     constexpr size_t array_size = 160;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 5, 8, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 5, 8, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -997,7 +994,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, f_remainder_bfyx_0_3_1_2) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 3, 1, 2 }));
 
     network network(engine, topology);
@@ -1022,7 +1019,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, f_remainder_bfyx_0_3_1_2) {
         156.f,  93.f, 109.f, 125.f, 141.f, 157.f,  94.f, 110.f, 126.f, 142.f, 158.f,  95.f, 111.f, 127.f, 143.f, 159.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1035,9 +1032,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, x_remainder_bfyx_0_3_1_2) {
 
     constexpr size_t array_size = 160;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 8, 5, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 8, 5, 2 } });
 
     set_values(input, {
         0.f,   1.f,   2.f,   3.f,   4.f,   5.f,   6.f,   7.f,   8.f,   9.f,  10.f,  11.f,  12.f,  13.f,  14.f,  15.f,
@@ -1053,7 +1050,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, x_remainder_bfyx_0_3_1_2) {
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 3, 1, 2 }));
 
     network network(engine, topology);
@@ -1078,7 +1075,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, x_remainder_bfyx_0_3_1_2) {
         88.f,  98.f, 108.f, 118.f, 128.f, 138.f, 148.f, 158.f,  89.f,  99.f, 109.f, 119.f, 129.f, 139.f, 149.f, 159.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1091,9 +1088,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, xf_remainder_bfyx_0_3_1_2) {
 
     constexpr size_t array_size = 100;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 5, 5, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 5, 5, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -1103,7 +1100,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, xf_remainder_bfyx_0_3_1_2) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 3, 1, 2 }));
 
     network network(engine, topology);
@@ -1128,7 +1125,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, xf_remainder_bfyx_0_3_1_2) {
         58.f,  68.f,  78.f,  88.f,  98.f,  59.f,  69.f,  79.f,  89.f,  99.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1141,9 +1138,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, normal_bfzyx_0_4_1_2_3) {
 
     constexpr size_t array_size = 512;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 8, 8, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 8, 8, 2, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -1153,7 +1150,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, normal_bfzyx_0_4_1_2_3) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 4, 1, 2, 3 }));
 
     network network(engine, topology);
@@ -1190,7 +1187,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, normal_bfzyx_0_4_1_2_3) {
         287.f, 319.f, 351.f, 383.f, 415.f, 447.f, 479.f, 511.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1203,9 +1200,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, f_remainder_bfzyx_0_4_1_2_3) {
 
     constexpr size_t array_size = 320;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 5, 8, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 5, 8, 2, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -1215,7 +1212,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, f_remainder_bfzyx_0_4_1_2_3) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 4, 1, 2, 3 }));
 
     network network(engine, topology);
@@ -1244,7 +1241,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, f_remainder_bfzyx_0_4_1_2_3) {
         254.f, 286.f, 318.f, 191.f, 223.f, 255.f, 287.f, 319.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1257,9 +1254,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, x_remainder_bfzyx_0_4_1_2_3) {
 
     constexpr size_t array_size = 320;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 8, 5, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 8, 5, 2, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -1269,7 +1266,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, x_remainder_bfzyx_0_4_1_2_3) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 4, 1, 2, 3 }));
 
     network network(engine, topology);
@@ -1298,7 +1295,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, x_remainder_bfzyx_0_4_1_2_3) {
         179.f, 199.f, 219.f, 239.f, 259.f, 279.f, 299.f, 319.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1311,9 +1308,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, xf_remainder_bfzyx_0_4_1_2_3) {
 
     constexpr size_t array_size = 200;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 5, 5, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 5, 5, 2, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -1323,7 +1320,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, xf_remainder_bfzyx_0_4_1_2_3) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 4, 1, 2, 3 }));
 
     network network(engine, topology);
@@ -1352,7 +1349,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, xf_remainder_bfzyx_0_4_1_2_3) {
         119.f, 139.f, 159.f, 179.f, 199.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1365,9 +1362,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, normal_bfwzyx_0_5_4_1_2_3) {
 
     constexpr size_t array_size = 1024;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx,{ 2, 8, 8, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfwzyx,{ 2, 8, 8, 2, 2, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -1377,7 +1374,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, normal_bfwzyx_0_5_4_1_2_3) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 5, 1, 2, 3, 4 }));
 
     network network(engine, topology);
@@ -1424,7 +1421,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, normal_bfwzyx_0_5_4_1_2_3) {
         572.f, 636.f, 700.f, 764.f, 828.f, 892.f, 956.f, 1020.f, 573.f, 637.f, 701.f, 765.f, 829.f, 893.f, 957.f, 1021.f, 574.f, 638.f, 702.f, 766.f, 830.f, 894.f, 958.f, 1022.f, 575.f, 639.f, 703.f, 767.f, 831.f, 895.f, 959.f, 1023.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1437,9 +1434,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, f_remainder_bfwzyx_0_5_4_1_2_3) {
 
     constexpr size_t array_size = 640;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx,{ 2, 5, 8, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfwzyx,{ 2, 5, 8, 2, 2, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -1449,7 +1446,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, f_remainder_bfwzyx_0_5_4_1_2_3) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 5, 1, 2, 3, 4 }));
 
     network network(engine, topology);
@@ -1484,7 +1481,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, f_remainder_bfwzyx_0_5_4_1_2_3) {
         569.f, 633.f, 378.f, 442.f, 506.f, 570.f, 634.f, 379.f, 443.f, 507.f, 571.f, 635.f, 380.f, 444.f, 508.f, 572.f, 636.f, 381.f, 445.f, 509.f, 573.f, 637.f, 382.f, 446.f, 510.f, 574.f, 638.f, 383.f, 447.f, 511.f, 575.f, 639.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1497,9 +1494,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, x_remainder_bfwzyx_0_5_4_1_2_3) {
 
     constexpr size_t array_size = 640;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx,{ 2, 8, 5, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfwzyx,{ 2, 8, 5, 2, 2, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -1509,7 +1506,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, x_remainder_bfwzyx_0_5_4_1_2_3) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 5, 1, 2, 3, 4 }));
 
     network network(engine, topology);
@@ -1544,7 +1541,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, x_remainder_bfwzyx_0_5_4_1_2_3) {
         356.f, 396.f, 436.f, 476.f, 516.f, 556.f, 596.f, 636.f, 357.f, 397.f, 437.f, 477.f, 517.f, 557.f, 597.f, 637.f, 358.f, 398.f, 438.f, 478.f, 518.f, 558.f, 598.f, 638.f, 359.f, 399.f, 439.f, 479.f, 519.f, 559.f, 599.f, 639.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1557,9 +1554,9 @@ TEST(permute_gpu_f32_tile_8x8_4x4, xf_remainder_bfwzyx_0_5_4_1_2_3) {
 
     constexpr size_t array_size = 400;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx,{ 2, 5, 5, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfwzyx,{ 2, 5, 5, 2, 2, 2 } });
 
     std::vector<float> input_data;
     input_data.reserve(array_size);
@@ -1569,7 +1566,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, xf_remainder_bfwzyx_0_5_4_1_2_3) {
     set_values(input, input_data);
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         permute("permute", "input", { 0, 5, 1, 2, 3, 4 }));
 
     network network(engine, topology);
@@ -1604,7 +1601,7 @@ TEST(permute_gpu_f32_tile_8x8_4x4, xf_remainder_bfwzyx_0_5_4_1_2_3) {
         236.f, 276.f, 316.f, 356.f, 396.f, 237.f, 277.f, 317.f, 357.f, 397.f, 238.f, 278.f, 318.f, 358.f, 398.f, 239.f, 279.f, 319.f, 359.f, 399.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (size_t i = 0; i < array_size; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1618,7 +1615,7 @@ struct TiledPermuteParam {
 
 class TiledPermuteTest : public ::testing::TestWithParam<TiledPermuteParam> {
 public:
-    const cldnn::engine engine;
+    cldnn::engine& engine = get_test_engine();
     TiledPermuteTest(): engine(get_test_engine()) { }
 
     template<typename T>
@@ -1627,7 +1624,7 @@ class TiledPermuteTest : public ::testing::TestWithParam<TiledPermuteParam> {
     }
 
     template<typename T>
-    void set_random_values(const cldnn::memory& mem) const {
+    void set_random_values(const cldnn::memory::ptr mem) const {
         tests::set_random_values<T>(mem);
     }
 
@@ -1647,11 +1644,11 @@ void TiledPermuteTest::compare_value(FLOAT16 a, FLOAT16 b) const {
 }
 
 template<>
-void TiledPermuteTest::set_random_values<int8_t>(const cldnn::memory& mem) const {
+void TiledPermuteTest::set_random_values<int8_t>(const cldnn::memory::ptr mem) const {
     // tests::set_random_values<int8_t>() is not supported
     std::mt19937 gen;
     static std::uniform_int_distribution<int32_t> uid(std::numeric_limits<int8_t>::min(), std::numeric_limits<int8_t>::max());
-    auto ptr = mem.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> ptr(mem, get_test_stream());
     for (auto it = ptr.begin(); it != ptr.end(); ++it) {
         *it = static_cast<int8_t>(uid(gen));
     }
@@ -1680,11 +1677,11 @@ void TiledPermuteTest::run_test(const std::vector<cldnn::tensor::value_type>& si
         order.push_back(i);
     }
 
-    auto input = memory::allocate(engine, {Data_Type, format, tensor});
+    auto input = engine.allocate_memory({Data_Type, format, tensor});
     set_random_values<type>(input);
 
     topology topology_ref = topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", {Data_Type, format_fsv, tensor}),
         permute("output", "reorder", order )
     );
@@ -1698,7 +1695,7 @@ void TiledPermuteTest::run_test(const std::vector<cldnn::tensor::value_type>& si
     network_ref.set_input_data("input", input);
     auto outputs_ref = network_ref.execute();
     auto output_ref = outputs_ref.begin()->second.get_memory();
-    auto output_ref_ptr = output_ref.pointer<type>();
+    cldnn::mem_lock<type> output_ref_ptr(output_ref, get_test_stream());
 
     // run with permute_tile_8x8_4x4_fsv16
     cldnn::build_options options_tile;
@@ -1709,10 +1706,10 @@ void TiledPermuteTest::run_test(const std::vector<cldnn::tensor::value_type>& si
     network_tile.set_input_data("input", input);
     auto outputs_tile = network_tile.execute();
     auto output_tile = outputs_tile.begin()->second.get_memory();
-    auto output_tile_ptr = output_tile.pointer<type>();
+    cldnn::mem_lock<type> output_tile_ptr(output_tile, get_test_stream());
 
     // compare results
-    const size_t output_size= output_ref.get_layout().get_linear_size();
+    const size_t output_size= output_ref->get_layout().get_linear_size();
     for (size_t i = 0; i < output_size; i++)
     {
         compare_value<type>(output_ref_ptr[i], output_tile_ptr[i]);
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/pooling_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/pooling_gpu_test.cpp
index a1dc98c8b0351e..009490ac681d44 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/pooling_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/pooling_gpu_test.cpp
@@ -3,21 +3,17 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/pooling.hpp"
-#include "api/mutable_data.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include "api/reorder.hpp"
-#include <api/data.hpp>
-#include "test_utils/float16.h"
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/pooling.hpp>
+#include <cldnn/primitives/mutable_data.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 namespace cldnn {
 template <>
@@ -223,12 +219,12 @@ TEST(pooling_forward_gpu, basic_max_byxf_f32_wsiz3x3_wstr1x1_i1x3x3x8_nopad) {
     //  Expected output:
     //  [ 8.0, 0.0, 0.0, 4,0, 0,5, -0.5, -0.5, -0.5 ]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,  format::byxf,{ 1, 8, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,  format::byxf,{ 1, 8, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::max, { 1,1,3,3 }, { 1,1,1,1 }));
     network network(engine, topology);
     set_values(input_prim, { 0.5f, -0.5f, -0.5f, -0.5f, 0.5f, -0.5f, -0.5f, -0.5f,
@@ -248,7 +244,7 @@ TEST(pooling_forward_gpu, basic_max_byxf_f32_wsiz3x3_wstr1x1_i1x3x3x8_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
     EXPECT_EQ(4.0f, output_ptr[3]);
 }
 
@@ -268,12 +264,12 @@ TEST(pooling_forward_gpu, basic_max_yxfb_f32_wsiz3x3_wstr1x1_i3x3x1x1_nopad) {
     //  Expected output:
     //  [ 2.0]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,  format::yxfb, { 1, 1, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,  format::yxfb, { 1, 1, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::max, { 1,1,3,3 }, { 1,1,1,1 }));
 
     network network(engine, topology);
@@ -286,7 +282,7 @@ TEST(pooling_forward_gpu, basic_max_yxfb_f32_wsiz3x3_wstr1x1_i3x3x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.0f, output_ptr[0]);
 }
@@ -306,12 +302,12 @@ TEST(pooling_forward_gpu, basic_max_yxfb_f32_global_i3x3x1x1_nopad) {
     //  Expected output:
     //  [ 2.0]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,  format::yxfb,{ 1, 1, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32,  format::yxfb,{ 1, 1, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::max));
 
     network network(engine, topology);
@@ -324,7 +320,7 @@ TEST(pooling_forward_gpu, basic_max_yxfb_f32_global_i3x3x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.0f, output_ptr[0]);
 }
@@ -336,12 +332,12 @@ TEST(pooling_forward_gpu, basic_max_b_fs_yx_fsv16_i8_global_i3x3x1x1_nopad) {
     //  Global pooling: true
     //  Padding: none
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::i8, format::b_fs_yx_fsv16, { 1, 16, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::i8, format::b_fs_yx_fsv16, { 1, 16, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::max));
 
     network network(engine, topology);
@@ -365,7 +361,7 @@ TEST(pooling_forward_gpu, basic_max_b_fs_yx_fsv16_i8_global_i3x3x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<char>();
+    cldnn::mem_lock<char> output_ptr(output_prim, get_test_stream());
 
     std::vector<char> answers = { 8, 44, 8, 81, 64, 8, 12, 66, 14, 8, 99, 64, 8, 11, 18, 21 };
 
@@ -382,12 +378,12 @@ TEST(pooling_forward_gpu, basic_avg_b_fs_yx_fsv16_i8_global_i3x3x1x1_nopad) {
     //  Global pooling: true
     //  Padding: none
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::i8, format::b_fs_yx_fsv16, { 1, 16, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::i8, format::b_fs_yx_fsv16, { 1, 16, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::average));
 
     network network(engine, topology);
@@ -411,7 +407,7 @@ TEST(pooling_forward_gpu, basic_avg_b_fs_yx_fsv16_i8_global_i3x3x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<uint8_t>();
+    cldnn::mem_lock<uint8_t> output_ptr(output_prim, get_test_stream());
 
     std::vector<uint8_t> answers = {
          29, 199, 241, 63,  85,  85, 213, 64,  85,  85,  21, 64, 142, 227,   8, 65,
@@ -428,7 +424,7 @@ TEST(pooling_forward_gpu, basic_avg_b_fs_yx_fsv16_i8_global_i3x3x1x1_nopad) {
 
 TEST(pooling_forward_gpu, basic_max_pooling_int8) {
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     layout in_layout = { type_to_data_type<float>::value,format::byxf,{ 1,1,3,3 } };
     layout out_layout = { type_to_data_type<float>::value,format::byxf,{ 1,1,1,1 } };
     layout byte_layout = { type_to_data_type<int8_t>::value, format::bfyx,{ 1,1,3,3 } };
@@ -436,7 +432,7 @@ TEST(pooling_forward_gpu, basic_max_pooling_int8) {
     std::list<float> final_results = { 10.0f };
 
     // Allocate memory for input image.
-    auto input_memory = memory::allocate(engine, in_layout);
+    auto input_memory = engine.allocate_memory(in_layout);
     set_values(input_memory, input_f);
 
     // Create input_layout description
@@ -464,7 +460,7 @@ TEST(pooling_forward_gpu, basic_max_pooling_int8) {
     auto outputs = network.execute();
 
     auto interm = outputs.at("reorder2").get_memory();
-    auto interm_ptr = interm.pointer<float>();
+    cldnn::mem_lock<float> interm_ptr(interm, get_test_stream());
     unsigned int cntr = 0;
     for (const auto& exp : final_results)
     {
@@ -474,7 +470,7 @@ TEST(pooling_forward_gpu, basic_max_pooling_int8) {
 
 TEST(pooling_forward_gpu, basic_avg_pooling_int8) {
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     layout in_layout = { type_to_data_type<float>::value,format::byxf,{ 1,1,3,3 } };
     layout out_layout = { type_to_data_type<float>::value,format::byxf,{ 1,1,1,1 } };
     layout byte_layout = { type_to_data_type<int8_t>::value, format::bfyx,{ 1,1,3,3 } };
@@ -488,7 +484,7 @@ TEST(pooling_forward_gpu, basic_avg_pooling_int8) {
     }
     final_result /= input_f.size();
     // Allocate memory for input image.
-    auto input_memory = memory::allocate(engine, in_layout);
+    auto input_memory = engine.allocate_memory(in_layout);
     set_values(input_memory, input_f);
 
     // Create input_layout description
@@ -516,7 +512,7 @@ TEST(pooling_forward_gpu, basic_avg_pooling_int8) {
     auto outputs = network.execute();
 
     auto interm = outputs.at("reorder2").get_memory();
-    auto interm_ptr = interm.pointer<float>();
+    cldnn::mem_lock<float> interm_ptr(interm, get_test_stream());
     EXPECT_EQ(final_result, interm_ptr[0]);
 }
 
@@ -537,12 +533,12 @@ TEST(pooling_forward_gpu, basic_max_yxfb_f32_wsiz2x2_wstr1x1_i3x3x1x1_nopad) {
     //  [ 2.0,  1.5]
     //  [ 2.0,  1.5]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::max, { 1,1,2,2 }, { 1,1,1,1 }));
 
     network network(engine, topology);
@@ -555,7 +551,7 @@ TEST(pooling_forward_gpu, basic_max_yxfb_f32_wsiz2x2_wstr1x1_i3x3x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.0f, output_ptr[0]);
     EXPECT_EQ(1.5f, output_ptr[1]);
@@ -581,12 +577,12 @@ TEST(pooling_forward_gpu, basic_max_yxfb_f32_wsiz2x2_wstr2x2_i4x4x1x1_nopad) {
     //  [ 2.0,  0.5]
     //  [ 0.5,  0.5]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 4, 4 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 4, 4 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::max, { 1,1,2,2 }, { 1,1,2,2 }));
 
     network network(engine, topology);
@@ -599,7 +595,7 @@ TEST(pooling_forward_gpu, basic_max_yxfb_f32_wsiz2x2_wstr2x2_i4x4x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(2.0f, output_ptr[0]);
     EXPECT_EQ(0.5f, output_ptr[1]);
@@ -635,12 +631,12 @@ TEST(pooling_forward_gpu, basic_max_yxfb_f32_wsiz2x2_wstr1x1_i3x3x2x2_nopad) {
     //  [ 0.5,  1.0]         [ 1.0,  0.5]
     //  [-0.5,  1.5]         [ 1.0,  0.0]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::max, { 1,1,2,2 }, { 1,1,1,1 }));
 
     network network(engine, topology);
@@ -653,7 +649,7 @@ TEST(pooling_forward_gpu, basic_max_yxfb_f32_wsiz2x2_wstr1x1_i3x3x2x2_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(1.0f, output_ptr[0]); EXPECT_EQ(0.0f, output_ptr[2]);
     EXPECT_EQ(0.5f, output_ptr[4]); EXPECT_EQ(1.5f, output_ptr[6]);
@@ -685,12 +681,12 @@ TEST(pooling_forward_gpu, offsets_max_yxfb_f32_wsiz2x2_wstr2x2_i2x2x1x1_zeropad)
     //  [ 1.5, -0.5]
     //  [   -1, 0.5]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::max, { 1,1,2,2 }, { 1,1,2,2 }, { 0, 0, -1,-1 }));
 
     network network(engine, topology);
@@ -703,7 +699,7 @@ TEST(pooling_forward_gpu, offsets_max_yxfb_f32_wsiz2x2_wstr2x2_i2x2x1x1_zeropad)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
     EXPECT_EQ( 1.5f, output_ptr[0]);
     EXPECT_EQ(-0.5f, output_ptr[1]);
     EXPECT_EQ(-1.0f, output_ptr[2]);
@@ -730,12 +726,12 @@ TEST(pooling_forward_gpu, offsets_max_yxfb_f32_wsiz2x2_wstr2x2_i3x3x1x1_zeropad)
     //  [ 1.5,  -0.5]
     //  [   1,  -0.5]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::max, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,-1,-1 }));
 
     network network(engine, topology);
@@ -753,13 +749,13 @@ TEST(pooling_forward_gpu, offsets_max_yxfb_f32_wsiz2x2_wstr2x2_i3x3x1x1_zeropad)
     EXPECT_EQ(outputs.begin()->first, "pool_prim");
 
     auto output_prim = outputs.begin()->second.get_memory();
-    EXPECT_EQ((int)output_prim.get_layout().size.count(), 4);
+    EXPECT_EQ((int)output_prim->get_layout().size.count(), 4);
 
-    auto output_ptr = output_prim.pointer<float>();
-    EXPECT_EQ(1.5f, get_value<float>(output_ptr, 0));
-    EXPECT_EQ(-0.5f, get_value<float>(output_ptr, 1));
-    EXPECT_EQ(1.0f, get_value<float>(output_ptr, 2));
-    EXPECT_EQ(-0.5f, get_value<float>(output_ptr, 3));
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
+    EXPECT_EQ(1.5f, output_ptr[0]);
+    EXPECT_EQ(-0.5f, output_ptr[1]);
+    EXPECT_EQ(1.0f, output_ptr[2]);
+    EXPECT_EQ(-0.5f, output_ptr[3]);
 }
 
 TEST(pooling_forward_gpu, basic_avg_yxfb_f32_wsiz2x2_wstr1x1_i3x3x1x1_nopad) {
@@ -779,12 +775,12 @@ TEST(pooling_forward_gpu, basic_avg_yxfb_f32_wsiz2x2_wstr1x1_i3x3x1x1_nopad) {
     //  [ 1.0,   0.625]
     //  [ 1.625, 0.875]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::average,{ 1,1,2,2 },{ 1,1,1,1 }));
 
     network network(engine, topology);
@@ -797,7 +793,7 @@ TEST(pooling_forward_gpu, basic_avg_yxfb_f32_wsiz2x2_wstr1x1_i3x3x1x1_nopad) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_EQ(1.0f,   output_ptr[0]);
     EXPECT_EQ(0.625f, output_ptr[1]);
@@ -824,12 +820,12 @@ TEST(pooling_forward_gpu, offsets_avg_yxfb_f32_wsiz2x2_wstr2x2_i2x2x1x1_zeropad)
     //  [ 0.375, -0.125]
     //  [ -0.25,  0.125]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::average, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,-1,-1 }));
 
     network network(engine, topology);
@@ -842,7 +838,7 @@ TEST(pooling_forward_gpu, offsets_avg_yxfb_f32_wsiz2x2_wstr2x2_i2x2x1x1_zeropad)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
     EXPECT_EQ(0.375f,  output_ptr[0]);
     EXPECT_EQ(-0.125f, output_ptr[1]);
     EXPECT_EQ(-0.25f,  output_ptr[2]);
@@ -869,12 +865,12 @@ TEST(pooling_forward_gpu, offsets_avg_bfyx_f32_wsiz3x3_wstr3x3_i1x1x3x3_zeropad)
     //  [ 0.177777, -0.133333]
     //  [ 0.333333,  0.55]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::average, { 1,1,3,3 }, { 1,1,3,3 }, { 0,0,-1,-1 }));
 
     network network(engine, topology);
@@ -890,7 +886,7 @@ TEST(pooling_forward_gpu, offsets_avg_bfyx_f32_wsiz3x3_wstr3x3_i1x1x3x3_zeropad)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
 
     EXPECT_NEAR(output_ptr[0], 0.177777f, 1e-05F);
     EXPECT_NEAR(output_ptr[1], -0.133333f, 1e-05F);
@@ -917,12 +913,12 @@ TEST(pooling_forward_gpu, offsets_avg_yxfb_f32_wsiz2x2_wstr2x2_i3x3x1x1_zeropad)
     //  [  0.375,    0.5]
     //  [ -0.125, -1.125]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input_prim = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(pooling("pool_prim", "input_prim", pooling_mode::average, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,-1,-1 }));
 
     network network(engine, topology);
@@ -934,9 +930,9 @@ TEST(pooling_forward_gpu, offsets_avg_yxfb_f32_wsiz2x2_wstr2x2_i3x3x1x1_zeropad)
     EXPECT_EQ(outputs.begin()->first, "pool_prim");
 
     auto output_prim = outputs.begin()->second.get_memory();
-    EXPECT_EQ((int)output_prim.get_layout().size.count(), 4);
+    EXPECT_EQ((int)output_prim->get_layout().size.count(), 4);
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
     EXPECT_EQ(0.375f,  output_ptr[0]);
     EXPECT_EQ(0.5f,    output_ptr[1]);
     EXPECT_EQ(-0.125f, output_ptr[2]);
@@ -966,7 +962,7 @@ TEST(pooling_forward_gpu, offsets_avg_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i2x2x1x1_out
     //  [0, 0, 0, 0, 0, 0]
     //  [0, 0, 0, 0, 0, 0]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     std::vector<format> formats_to_test = { format::yxfb , format::bfyx };
 
     for (std::vector<format>::iterator it = formats_to_test.begin(); it != formats_to_test.end(); ++it)
@@ -974,10 +970,10 @@ TEST(pooling_forward_gpu, offsets_avg_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i2x2x1x1_out
         std::cout << "Testing format: " << format::order(*it) << std::endl;
 
         tensor input_tensor( 1, 1, 2, 2 );
-        auto input_prim = memory::allocate(engine, { data_types::f32, *it, input_tensor });
+        auto input_prim = engine.allocate_memory({ data_types::f32, *it, input_tensor });
 
         topology topology;
-        topology.add(input_layout("input_prim", input_prim.get_layout()));
+        topology.add(input_layout("input_prim", input_prim->get_layout()));
         topology.add(pooling("pool_prim", "input_prim", pooling_mode::average, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,-1,-1 }, padding{ { 0,0,2,2 }, 0 }));
 
         network network(engine, topology);
@@ -998,7 +994,7 @@ TEST(pooling_forward_gpu, offsets_avg_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i2x2x1x1_out
         EXPECT_EQ(outputs.begin()->first, "pool_prim");
 
         auto output_prim = outputs.begin()->second.get_memory();
-        auto output_ptr = output_prim.pointer<float>();
+        cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
         for (size_t i = 0; i < expected.size(); ++i) {
             EXPECT_EQ(expected[i], output_ptr[i]);
         }
@@ -1027,7 +1023,7 @@ TEST(pooling_forward_gpu, offsets_max_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i3x3x1x1_out
     //  [0, 1, -0.5, 0, 0]
     //  [0, 0, 0, 0, 0]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     std::vector<format> formats_to_test = { format::yxfb , format::bfyx };
 
     for (std::vector<format>::iterator it = formats_to_test.begin(); it != formats_to_test.end(); ++it)
@@ -1035,10 +1031,10 @@ TEST(pooling_forward_gpu, offsets_max_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i3x3x1x1_out
         std::cout << "Testing format: " << format::order(*it) << std::endl;
 
         tensor input_tensor( 1, 1, 3, 3 );
-        auto input_prim = memory::allocate(engine, { data_types::f32, *it, input_tensor });
+        auto input_prim = engine.allocate_memory({ data_types::f32, *it, input_tensor });
 
         topology topology;
-        topology.add(input_layout("input_prim", input_prim.get_layout()));
+        topology.add(input_layout("input_prim", input_prim->get_layout()));
         topology.add(pooling("pool_prim", "input_prim", pooling_mode::max, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,-1,-1 }, padding{ { 0,0,1,1 }, 0 }));
 
         network network(engine, topology);
@@ -1063,10 +1059,10 @@ TEST(pooling_forward_gpu, offsets_max_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i3x3x1x1_out
         EXPECT_EQ(outputs.begin()->first, "pool_prim");
 
         auto output_prim = outputs.begin()->second.get_memory();
-        EXPECT_EQ((int)output_prim.get_layout().size.count(), 4);
-        EXPECT_EQ((int)output_prim.get_layout().get_buffer_size().count(), 16);
+        EXPECT_EQ((int)output_prim->get_layout().size.count(), 4);
+        EXPECT_EQ((int)output_prim->get_layout().get_buffer_size().count(), 16);
 
-        auto output_ptr = output_prim.pointer<float>();
+        cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
         for (size_t i = 0; i < expected.size(); ++i) {
             EXPECT_EQ(expected[i], output_ptr[i]);
         }
@@ -1097,7 +1093,7 @@ TEST(pooling_forward_gpu, offsets_avg_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i2x2x1x1_inp
     //  [0, 0, 0, 0, 0, 0]
     //  [0, 0, 0, 0, 0, 0]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     std::vector<format> formats_to_test = { format::yxfb , format::bfyx };
 
     for (std::vector<format>::iterator it = formats_to_test.begin(); it != formats_to_test.end(); ++it)
@@ -1105,11 +1101,11 @@ TEST(pooling_forward_gpu, offsets_avg_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i2x2x1x1_inp
         std::cout << "Testing format: " << format::order(*it) << std::endl;
 
         tensor input_tensor( 1, 1, 2, 2 );
-        auto input_prim = memory::allocate(engine, { data_types::f32, *it, input_tensor });
+        auto input_prim = engine.allocate_memory({ data_types::f32, *it, input_tensor });
 
         topology topology;
-        topology.add(input_layout("input_prim", input_prim.get_layout()));
-        topology.add(reorder("reorder", "input_prim", input_prim.get_layout().with_padding(padding{ {0,0,1,2}, 0 })));
+        topology.add(input_layout("input_prim", input_prim->get_layout()));
+        topology.add(reorder("reorder", "input_prim", input_prim->get_layout().with_padding(padding{ {0,0,1,2}, 0 })));
         topology.add(pooling("pool_prim", "reorder", pooling_mode::average, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,-1,-1 }, padding{ { 0,0,2,2 }, 0 }));
 
         network network(engine, topology);
@@ -1130,7 +1126,7 @@ TEST(pooling_forward_gpu, offsets_avg_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i2x2x1x1_inp
         EXPECT_EQ(outputs.begin()->first, "pool_prim");
 
         auto output_prim = outputs.begin()->second.get_memory();
-        auto output_ptr = output_prim.pointer<float>();
+        cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
         for (size_t i = 0; i < expected.size(); ++i) {
             EXPECT_EQ(expected[i], output_ptr[i]);
         }
@@ -1160,7 +1156,7 @@ TEST(pooling_forward_gpu, offsets_max_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i3x3x1x1_inp
     //  [0, 1, -0.5, 0]
     //  [0, 0, 0, 0, 0]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     std::vector<format> formats_to_test = { format::yxfb , format::bfyx };
 
     for (std::vector<format>::iterator it = formats_to_test.begin(); it != formats_to_test.end(); ++it)
@@ -1168,11 +1164,11 @@ TEST(pooling_forward_gpu, offsets_max_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i3x3x1x1_inp
         std::cout << "Testing format: " << format::order(*it) << std::endl;
 
         tensor input_tensor( 1, 1, 3, 3 );
-        auto input_prim = memory::allocate(engine, { data_types::f32, *it, input_tensor });
+        auto input_prim = engine.allocate_memory({ data_types::f32, *it, input_tensor });
 
         topology topology;
-        topology.add(input_layout("input_prim", input_prim.get_layout()));
-        topology.add(reorder("reorder", "input_prim", input_prim.get_layout().with_padding(padding{ { 0, 0, 1, 2 }, 0 })));
+        topology.add(input_layout("input_prim", input_prim->get_layout()));
+        topology.add(reorder("reorder", "input_prim", input_prim->get_layout().with_padding(padding{ { 0, 0, 1, 2 }, 0 })));
         topology.add(pooling("pool_prim", "reorder", pooling_mode::max, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,-1,-1 }, padding{ { 0,0,1,1 }, 0 }));
 
         network network(engine, topology);
@@ -1197,10 +1193,10 @@ TEST(pooling_forward_gpu, offsets_max_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i3x3x1x1_inp
         EXPECT_EQ(outputs.begin()->first, "pool_prim");
 
         auto output_prim = outputs.begin()->second.get_memory();
-        EXPECT_EQ((int)output_prim.get_layout().size.count(), 4);
-        EXPECT_EQ((int)output_prim.get_layout().get_buffer_size().count(), 16);
+        EXPECT_EQ((int)output_prim->get_layout().size.count(), 4);
+        EXPECT_EQ((int)output_prim->get_layout().get_buffer_size().count(), 16);
 
-        auto output_ptr = output_prim.pointer<float>();
+        cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
         for (size_t i = 0; i < expected.size(); ++i) {
             EXPECT_EQ(expected[i], output_ptr[i]);
         }
@@ -1231,7 +1227,7 @@ TEST(pooling_forward_gpu, avg_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i2x2x1x1_inpad2x1_ou
     //  [0, 0, 0, 0, 0, 0]
     //  [0, 0, 0, 0, 0, 0]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     std::vector<format> formats_to_test = { format::yxfb , format::bfyx };
 
     for (std::vector<format>::iterator it = formats_to_test.begin(); it != formats_to_test.end(); ++it)
@@ -1239,11 +1235,11 @@ TEST(pooling_forward_gpu, avg_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i2x2x1x1_inpad2x1_ou
         std::cout << "Testing format: " << format::order(*it) << std::endl;
 
         tensor input_tensor( 1, 1, 4, 4 );
-        auto input_prim = memory::allocate(engine, { data_types::f32, *it, input_tensor });
+        auto input_prim = engine.allocate_memory({ data_types::f32, *it, input_tensor });
 
         topology topology;
-        topology.add(input_layout("input_prim", input_prim.get_layout()));
-        topology.add(reorder("reorder", "input_prim", input_prim.get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
+        topology.add(input_layout("input_prim", input_prim->get_layout()));
+        topology.add(reorder("reorder", "input_prim", input_prim->get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
         topology.add(pooling("pool_prim", "reorder", pooling_mode::average, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,0,0 }, padding{ { 0,0,2,2 }, 0 }));
 
         network network(engine, topology);
@@ -1268,7 +1264,7 @@ TEST(pooling_forward_gpu, avg_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i2x2x1x1_inpad2x1_ou
         EXPECT_EQ(outputs.begin()->first, "pool_prim");
 
         auto output_prim = outputs.begin()->second.get_memory();
-        auto output_ptr = output_prim.pointer<float>();
+        cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
         for (size_t i = 0; i < expected.size(); ++i) {
             EXPECT_EQ(expected[i], output_ptr[i]);
         }
@@ -1299,7 +1295,7 @@ TEST(pooling_forward_gpu, max_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i3x3x1x1_inpad2x1_ou
     //  [0, 12, 14, 16, 0]
     //  [0, 0, 0, 0, 0]
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     std::vector<format> formats_to_test = { format::yxfb , format::bfyx };
 
     for (std::vector<format>::iterator it = formats_to_test.begin(); it != formats_to_test.end(); ++it)
@@ -1307,11 +1303,11 @@ TEST(pooling_forward_gpu, max_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i3x3x1x1_inpad2x1_ou
         std::cout << "Testing format: " << format::order(*it) << std::endl;
 
         tensor input_tensor( 1, 1, 5, 5 );
-        auto input_prim = memory::allocate(engine, { data_types::f32, *it, input_tensor });
+        auto input_prim = engine.allocate_memory({ data_types::f32, *it, input_tensor });
 
         topology topology;
-        topology.add(input_layout("input_prim", input_prim.get_layout()));
-        topology.add(reorder("reorder", "input_prim", input_prim.get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
+        topology.add(input_layout("input_prim", input_prim->get_layout()));
+        topology.add(reorder("reorder", "input_prim", input_prim->get_layout().with_padding(padding{ { 0, 0, 2, 1 }, 0 })));
         topology.add(pooling("pool_prim", "reorder", pooling_mode::max, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,-1,-1 }, padding{ { 0,0,1,1 }, 0 }));
 
         network network(engine, topology);
@@ -1339,10 +1335,10 @@ TEST(pooling_forward_gpu, max_yxfb_bfyx_f32_wsiz2x2_wstr2x2_i3x3x1x1_inpad2x1_ou
         EXPECT_EQ(outputs.begin()->first, "pool_prim");
 
         auto output_prim = outputs.begin()->second.get_memory();
-        EXPECT_EQ((int)output_prim.get_layout().size.count(), 9);
-        EXPECT_EQ((int)output_prim.get_layout().get_buffer_size().count(), 25);
+        EXPECT_EQ((int)output_prim->get_layout().size.count(), 9);
+        EXPECT_EQ((int)output_prim->get_layout().get_buffer_size().count(), 25);
 
-        auto output_ptr = output_prim.pointer<float>();
+        cldnn::mem_lock<float> output_ptr (output_prim, get_test_stream());
         for (size_t i = 0; i < expected.size(); ++i) {
             EXPECT_EQ(expected[i], output_ptr[i]);
         }
@@ -1368,10 +1364,10 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax) {
     //  f0: b0:  4    4   b1:   15    13
     //  f1: b0:  10  11   b1:   21    23
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
-    auto arg_max = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
+    auto arg_max = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         1.0f, 2.0f, -10.f,
@@ -1385,7 +1381,7 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mutable_data("arg_max", arg_max));
     topology.add(pooling("pooling", "input", "arg_max", pooling_mode::max_with_argmax, { 1, 1, 2, 2 }, { 1, 1, 1, 1 }));
 
@@ -1396,9 +1392,9 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax) {
     auto outputs = network.execute();
 
     auto output = outputs.at("pooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
-    auto argmax_ptr = arg_max.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
+    cldnn::mem_lock<float> argmax_ptr(arg_max, get_test_stream());
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 1);
@@ -1445,10 +1441,10 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2x1_max_with_argmax) {
     //  f0: b0:  4    4   b1:   15    13
     //  f1: b0:  10  11   b1:   21    23
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 2, 3, 2, 1 } });
-    auto arg_max = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 2, 2, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 2, 3, 2, 1 } });
+    auto arg_max = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 2, 2, 1, 1 } });
 
     set_values(input, {
         1.0f, 2.0f, -10.f,
@@ -1462,7 +1458,7 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2x1_max_with_argmax) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mutable_data("arg_max", arg_max));
     topology.add(pooling("pooling", "input", "arg_max", pooling_mode::max_with_argmax, { 1, 1, 2, 2, 1 }, { 1, 1, 1, 1, 1 }));
 
@@ -1473,9 +1469,9 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2x1_max_with_argmax) {
     auto outputs = network.execute();
 
     auto output = outputs.at("pooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
-    auto argmax_ptr = arg_max.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
+    cldnn::mem_lock<float> argmax_ptr(arg_max, get_test_stream());
 
     EXPECT_EQ(output_layout.format, format::bfzyx);
     EXPECT_EQ(output_layout.size.spatial[2], 1);
@@ -1524,10 +1520,10 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax_input_padding) {
     //  f0: b0:  4    4   b1:   15    13
     //  f1: b0:  10  11   b1:   21    23
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
-    auto arg_max = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
+    auto arg_max = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         1.0f, 2.0f, -10.f,
@@ -1541,8 +1537,8 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax_input_padding) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(reorder("reorder", "input", input.get_layout().with_padding(padding{ { 0, 0, 2, 2 }, 0 })));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(reorder("reorder", "input", input->get_layout().with_padding(padding{ { 0, 0, 2, 2 }, 0 })));
     topology.add(mutable_data("arg_max", arg_max));
     topology.add(pooling("pooling", "reorder", "arg_max", pooling_mode::max_with_argmax, { 1, 1, 2, 2 }, { 1, 1, 1, 1 }));
 
@@ -1553,9 +1549,9 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax_input_padding) {
     auto outputs = network.execute();
 
     auto output = outputs.at("pooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
-    auto argmax_ptr = arg_max.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
+    cldnn::mem_lock<float> argmax_ptr(arg_max, get_test_stream());
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 1);
@@ -1603,10 +1599,10 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax_output_padding) {
     //  f0: b0:  4    4   b1:   15    13
     //  f1: b0:  10  11   b1:   21    23
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
-    auto arg_max = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
+    auto arg_max = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         1.0f, 2.0f, -10.f,
@@ -1620,8 +1616,8 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax_output_padding) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(reorder("reorder", "input", input.get_layout().with_padding(padding{ { 0, 0, 2, 2 }, 0 })));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(reorder("reorder", "input", input->get_layout().with_padding(padding{ { 0, 0, 2, 2 }, 0 })));
     topology.add(mutable_data("arg_max", arg_max));
     topology.add(pooling("pooling", "reorder", "arg_max", pooling_mode::max_with_argmax, { 1, 1, 2, 2 }, { 1, 1, 1, 1 }, { 0, 0, 0, 0 }, padding({ 0, 0, 1, 1 }, 0)));
 
@@ -1632,9 +1628,9 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax_output_padding) {
     auto outputs = network.execute();
 
     auto output = outputs.at("pooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
-    auto argmax_ptr = arg_max.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
+    cldnn::mem_lock<float> argmax_ptr(arg_max, get_test_stream());
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 1);
@@ -1692,10 +1688,10 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax_with_output_size) {
     //  f0: b0:  4    4   b1:   15    13
     //  f1: b0:  10  11   b1:   21    23
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
-    auto arg_max = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 3, 2 } });
+    auto arg_max = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 1 } });
 
     set_values(input, {
         1.0f, 2.0f, -10.f,
@@ -1709,7 +1705,7 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax_with_output_size) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(mutable_data("arg_max", arg_max));
     topology.add(pooling("pooling", "input", "arg_max", pooling_mode::max_with_argmax, { 1, 1, 2, 2 }, { 1, 1, 1, 1 }, { 0, 0, 0, 0 }, { 2, 2, 2, 1 }));
 
@@ -1720,9 +1716,9 @@ TEST(pooling_forward_gpu, basic_in2x2x3x2_max_with_argmax_with_output_size) {
     auto outputs = network.execute();
 
     auto output = outputs.at("pooling").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_layout = output.get_layout();
-    auto argmax_ptr = arg_max.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_layout = output->get_layout();
+    cldnn::mem_lock<float> argmax_ptr(arg_max, get_test_stream());
 
     EXPECT_EQ(output_layout.format, format::bfyx);
     EXPECT_EQ(output_layout.size.spatial[1], 1);
@@ -1755,38 +1751,36 @@ static void generic_average_wo_padding_test(format fmt, tensor output, tensor in
 {
     constexpr auto dt = std::is_same<DataType, float>::value ? data_types::f32 : data_types::f16;
 
-    engine eng;
+    auto& engine = get_test_engine();
 
-    if (!eng.get_info().supports_fp16)
-    {
-        if(dt == data_types::f16)
-        {
+    if (!engine.get_device_info().supports_fp16) {
+        if (dt == data_types::f16) {
             return;
         }
     }
 
-    auto input_mem = memory::allocate(eng, layout{ dt, fmt, input });
+    auto input_mem = engine.allocate_memory(layout{ dt, fmt, input });
     set_values(input_mem, std::vector<DataType>(input.count(), DataType(1)));
     std::vector<DataType> expected_output(output.count(), DataType(1));
 
     topology tpl;
-    tpl.add(input_layout("in", input_mem.get_layout()));
+    tpl.add(input_layout("in", input_mem->get_layout()));
 
     auto pool_in = "in";
     if (offset != tensor())
     {
-        tpl.add(reorder("reorder", "in", input_mem.get_layout().with_padding((padding) offset.negate().sizes())));
+        tpl.add(reorder("reorder", "in", input_mem->get_layout().with_padding((padding) offset.negate().sizes())));
         pool_in = "reorder";
     }
     tpl.add(pooling("pool", pool_in, pooling_mode::average_no_padding, window, stride, offset));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in", input_mem);
     auto output_mem = net.execute().at("pool").get_memory();
 
-    ASSERT_EQ(output_mem.count(), expected_output.size());
-    EXPECT_EQ(output_mem.get_layout().size, output);
-    auto out_ptr = output_mem.pointer<DataType>();
+    ASSERT_EQ(output_mem->count(), expected_output.size());
+    EXPECT_EQ(output_mem->get_layout().size, output);
+    cldnn::mem_lock<DataType> out_ptr(output_mem, get_test_stream());
 
     for (size_t i = 0; i < expected_output.size(); ++i)
         EXPECT_FLOAT_EQ(out_ptr[i], expected_output[i]);
@@ -1959,16 +1953,13 @@ TEST(pooling_forward_gpu, b_fs_yx_fsv4)
         std::vector<char>  vGoldOutput;
         std::vector<char>  vTestOutput;
 
-        engine   engine;
+        auto& engine = get_test_engine();
 
         // "Golden" Pooling
         {
             // Mem initialization
             // This is user data, no kernels here
-            auto input = memory::allocate(engine,
-                                          { data_types::i8,
-                                              format::bfyx,
-                                              { in_B, in_F, in_X, in_Y } });
+            auto input = engine.allocate_memory({ data_types::i8, format::bfyx, { in_B, in_F, in_X, in_Y } });
             set_values(input, std::move(DataGold));
 
             auto pool = pooling("pool_GOLD",
@@ -1978,7 +1969,7 @@ TEST(pooling_forward_gpu, b_fs_yx_fsv4)
                                  { 1, 1, S_X, S_Y }); // stride
 
             // Create a topology with a simple Convolution layer
-            topology topology(input_layout("input", input.get_layout()),
+            topology topology(input_layout("input", input->get_layout()),
                               pool);
 
             // Network processing
@@ -1989,7 +1980,7 @@ TEST(pooling_forward_gpu, b_fs_yx_fsv4)
             auto searchC = outputs.find("pool_GOLD");
             ASSERT_FALSE(searchC == outputs.end());
             auto output = outputs.begin()->second.get_memory();
-            auto output_ptr = output.pointer<char>();
+            cldnn::mem_lock<char> output_ptr(output, get_test_stream());
             vGoldOutput.reserve(output_ptr.size());
             for (size_t i = 0; i < output_ptr.size(); i++)
                 vGoldOutput.push_back(output_ptr[i]);
@@ -2003,15 +1994,12 @@ TEST(pooling_forward_gpu, b_fs_yx_fsv4)
 
             // Mem initialization
             // This is user data, no kernels here
-            auto input = memory::allocate(engine,
-                                          { data_types::i8,
-                                              format::bfyx,
-                                              { in_B, in_F, in_X, in_Y } });
+            auto input = engine.allocate_memory({ data_types::i8, format::bfyx, { in_B, in_F, in_X, in_Y } });
             set_values(input, std::move(Data));
 
             // Add input to topology
             topology.add(
-                input_layout("input", input.get_layout()));
+                input_layout("input", input->get_layout()));
 
             // Reorder (a-ka swizzelling) input to MMAD/IMAD Pooling format
             topology.add(reorder("reorder_Swizzelled",
@@ -2041,7 +2029,7 @@ TEST(pooling_forward_gpu, b_fs_yx_fsv4)
             auto searchC = outputs.find("reorder_UnSwizzelled");
             ASSERT_FALSE(searchC == outputs.end());
             auto output = outputs.begin()->second.get_memory();
-            auto output_ptr = output.pointer<char>();
+            cldnn::mem_lock<char> output_ptr(output, get_test_stream());
             vTestOutput.reserve(output_ptr.size());
             for (size_t i = 0; i < output_ptr.size(); i++)
                 vTestOutput.push_back(output_ptr[i]);
@@ -2057,8 +2045,8 @@ TEST(pooling_forward_gpu, b_fs_yx_fsv4)
 
 TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_3x3_input_2x2_pool_1x1_stride_2x2_output)
 {
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = !!engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
         return;
@@ -2080,10 +2068,10 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_3x3_input_2x2_pool_1x1_stride_2x2_ou
     //  [ 1.0,   0.625]
     //  [ 1.625, 0.875]
 
-    auto input_prim = memory::allocate(engine, { data_types::f16, format::yxfb, { 1, 1, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f16, format::yxfb, { 1, 1, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input", input_prim.get_layout()));
+    topology.add(input_layout("input", input_prim->get_layout()));
     topology.add(reorder("reorder_input", "input", layout(data_types::f16, format::fs_b_yx_fsv32, { 1, 1, 3, 3 })));
     topology.add(pooling("avg_pooling", "reorder_input", pooling_mode::average, { 1,1,2,2 }, { 1,1,1,1 }));
     topology.add(reorder("reorder_after_pooling", "avg_pooling", layout(data_types::f16, format::bfyx, { 1,1,2,2 })));
@@ -2098,7 +2086,7 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_3x3_input_2x2_pool_1x1_stride_2x2_ou
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_ptr(output_prim, get_test_stream());
 
     EXPECT_EQ(1.0f, float(output_ptr[0]));
     EXPECT_EQ(0.625f, float(output_ptr[1]));
@@ -2109,8 +2097,8 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_3x3_input_2x2_pool_1x1_stride_2x2_ou
 
 TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_3x3_input_2x2_pool_2x2_stride)
 {
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = !!engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
         return;
@@ -2132,10 +2120,10 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_3x3_input_2x2_pool_2x2_stride)
     //  [ 1.0, 0  ]
     //  [ 1.5, 3.5]
 
-    auto input_prim = memory::allocate(engine, { data_types::f16, format::yxfb, { 1, 1, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f16, format::yxfb, { 1, 1, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input", input_prim.get_layout()));
+    topology.add(input_layout("input", input_prim->get_layout()));
     topology.add(reorder("reorder_input", "input", layout(data_types::f16, format::fs_b_yx_fsv32, { 1, 1, 3, 3 })));
     topology.add(pooling("avg_pooling", "reorder_input", pooling_mode::average, { 1,1,2,2 }, { 1,1,2,2 }));
     topology.add(reorder("reorder_after_pooling", "avg_pooling", layout(data_types::f16, format::bfyx, { 1,1,3,3 })));
@@ -2149,7 +2137,7 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_3x3_input_2x2_pool_2x2_stride)
     EXPECT_EQ(outputs.begin()->first, "reorder_after_pooling");
 
     auto output_prim = outputs.begin()->second.get_memory();
-    auto output_ptr = output_prim.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_ptr(output_prim, get_test_stream());
 
     EXPECT_EQ(1.0f, float(output_ptr[0]));
     EXPECT_EQ(0.f, float(output_ptr[1]));
@@ -2161,8 +2149,8 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_3x3_input_2x2_pool_2x2_stride)
 
 TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_2x2x3x3_input_2x2_pool_2x2_stride)
 {
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = !!engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
         return;
@@ -2198,10 +2186,10 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_2x2x3x3_input_2x2_pool_2x2_stride)
     const int out_x = 2;
     const int out_y = 2;
 
-    auto input_prim = memory::allocate(engine, { data_types::f16, format::bfyx, { batch_count, features_count, 3, 3 } });
+    auto input_prim = engine.allocate_memory({ data_types::f16, format::bfyx, { batch_count, features_count, 3, 3 } });
 
     topology topology;
-    topology.add(input_layout("input", input_prim.get_layout()));
+    topology.add(input_layout("input", input_prim->get_layout()));
     topology.add(reorder("reorder_input", "input", layout(data_types::f16, format::fs_b_yx_fsv32, { batch_count, features_count, 3, 3 })));
     topology.add(pooling("avg_pooling", "reorder_input", pooling_mode::average, { 1,1,2,2 }, { 1,1,2,2 }));
     topology.add(reorder("reorder_after_pooling", "avg_pooling", layout(data_types::f16, format::bfyx, { batch_count, features_count, out_y, out_x })));
@@ -2219,7 +2207,7 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_2x2x3x3_input_2x2_pool_2x2_stride)
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_ptr(output_prim, get_test_stream());
 
     ASSERT_EQ((int)output_ptr.size(), batch_count * features_count*out_x*out_y);
 
@@ -2240,8 +2228,8 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_2x2x3x3_input_2x2_pool_2x2_stride)
 }
 
 TEST(pooling_forward_gpu, fs_b_yx_fsv32_max_1x1x3x3_input_2x2_pool_2x2_stride_2x2_outpad) {
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = !!engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
         return;
@@ -2269,10 +2257,10 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_max_1x1x3x3_input_2x2_pool_2x2_stride_2x
     //  [0,    0,    0,  0,  0]
 
         tensor input_tensor(1, 1, 3, 3);
-        auto input_prim = memory::allocate(engine, { data_types::f16, format::bfyx, input_tensor });
+        auto input_prim = engine.allocate_memory({ data_types::f16, format::bfyx, input_tensor });
 
         topology topology;
-        topology.add(input_layout("input_prim", input_prim.get_layout()));
+        topology.add(input_layout("input_prim", input_prim->get_layout()));
         topology.add(reorder("reorder_input", "input_prim", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor)));
         topology.add(pooling("pool_prim", "reorder_input", pooling_mode::max, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,-1,-1 }, padding{ { 0,0,1,1 }, 0 }));
         topology.add(reorder("reorder_pooling", "pool_prim", layout(data_types::f16, format::bfyx, { 1,1,4,4 }, padding{ {0,0,1,1},0 })));
@@ -2299,10 +2287,10 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_max_1x1x3x3_input_2x2_pool_2x2_stride_2x
         EXPECT_EQ(outputs.begin()->first, "reorder_pooling");
 
         auto output_prim = outputs.begin()->second.get_memory();
-        EXPECT_EQ((int)output_prim.get_layout().size.count(), 4);
-        EXPECT_EQ((int)output_prim.get_layout().get_buffer_size().count(), 16);
+        EXPECT_EQ((int)output_prim->get_layout().size.count(), 4);
+        EXPECT_EQ((int)output_prim->get_layout().get_buffer_size().count(), 16);
 
-        auto output_ptr = output_prim.pointer<FLOAT16>();
+        cldnn::mem_lock<FLOAT16> output_ptr(output_prim, get_test_stream());
 
         for (size_t i = 0; i < expected.size(); ++i) {
             EXPECT_EQ(expected[i], float(output_ptr[i]));
@@ -2311,8 +2299,8 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_max_1x1x3x3_input_2x2_pool_2x2_stride_2x
 }
 
 TEST(pooling_forward_gpu, fs_b_yx_fsv32_max_1x1x5x5_input_2x2_pool_2x2_stride_2x2_outpad_2x1_inpad) {
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = !!engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
         return;
@@ -2342,10 +2330,10 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_max_1x1x5x5_input_2x2_pool_2x2_stride_2x
     //  [ 0,  0,  0,    0,  0]
 
     tensor input_tensor(1, 1, 5, 5);
-    auto input_prim = memory::allocate(engine, { data_types::f16, format::bfyx, input_tensor });
+    auto input_prim = engine.allocate_memory({ data_types::f16, format::bfyx, input_tensor });
 
     topology topology;
-    topology.add(input_layout("input_prim", input_prim.get_layout()));
+    topology.add(input_layout("input_prim", input_prim->get_layout()));
     topology.add(reorder("reorder_input", "input_prim", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor, padding{ { 0,0,2,1 } , 0 })));
     topology.add(pooling("pool_prim", "reorder_input", pooling_mode::max, { 1,1,2,2 }, { 1,1,2,2 }, { 0,0,-1,-1 }, padding{ { 0,0,1,1 }, 0 }));
     topology.add(reorder("reorder_pooling", "pool_prim", layout(data_types::f16, format::bfyx, input_tensor, padding{{0,0,1,1},0})));
@@ -2375,10 +2363,10 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_max_1x1x5x5_input_2x2_pool_2x2_stride_2x
     EXPECT_EQ(outputs.begin()->first, "reorder_pooling");
 
     auto output_prim = outputs.begin()->second.get_memory();
-    EXPECT_EQ((int)output_prim.get_layout().size.count(), 9);
-    EXPECT_EQ((int)output_prim.get_layout().get_buffer_size().count(), 25);
+    EXPECT_EQ((int)output_prim->get_layout().size.count(), 9);
+    EXPECT_EQ((int)output_prim->get_layout().get_buffer_size().count(), 25);
 
-    auto output_ptr = output_prim.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_ptr(output_prim, get_test_stream());
 
     for (size_t i = 0; i < expected.size(); ++i) {
         EXPECT_EQ(expected[i], float(output_ptr[i]));
@@ -2387,8 +2375,8 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_max_1x1x5x5_input_2x2_pool_2x2_stride_2x
 
 TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_65x5x6x7_input_3x3_pool_4x4_stride_3x2_outpad_2x3_inpad)
 {
-    const auto& engine = get_test_engine();
-    bool f16_supported = !!engine.get_info().supports_fp16;
+    auto& engine = get_test_engine();
+    bool f16_supported = !!engine.get_device_info().supports_fp16;
     if (!f16_supported) {
         std::cout << "[ SKIPPED  ] float16 combinations are skipped (cl_khr_fp16 is not supported)." << std::endl;
         return;
@@ -2414,7 +2402,7 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_65x5x6x7_input_3x3_pool_4x4_stride_3
         input_data[i] = FLOAT16((float)i/float(input_data.size()));
     }
 
-    auto input_prim = memory::allocate(engine, { data_types::f16,format::bfyx,input_tensor });
+    auto input_prim = engine.allocate_memory({ data_types::f16,format::bfyx,input_tensor });
     set_values(input_prim, input_data);
 
     std::vector<float> golden_results;
@@ -2422,15 +2410,15 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_65x5x6x7_input_3x3_pool_4x4_stride_3
 
     { //GOLDEN TOPOLOGY
         topology golden_topology;
-        golden_topology.add(input_layout("input", input_prim.get_layout()));
-        golden_topology.add(reorder("reorder_input", "input", input_prim.get_layout().with_padding(padding{ {0,0,x_in_pad,y_in_pad},0 })));
+        golden_topology.add(input_layout("input", input_prim->get_layout()));
+        golden_topology.add(reorder("reorder_input", "input", input_prim->get_layout().with_padding(padding{ {0,0,x_in_pad,y_in_pad},0 })));
         golden_topology.add(pooling("golden_pooling", "reorder_input", pooling_mode::average, { 1,1,pool_size,pool_size }, { 1,1,stride_size,stride_size }, { 0,0,0,0 }, padding{ { 0,0,x_out_pad,y_out_pad },0 }));
 
         network golden_network(engine, golden_topology);
         golden_network.set_input_data("input", input_prim);
 
         auto outputs = golden_network.execute();
-        auto output_ptr = outputs.begin()->second.get_memory().pointer<FLOAT16>();
+        cldnn::mem_lock<FLOAT16> output_ptr(outputs.begin()->second.get_memory(), get_test_stream());
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
             golden_results.push_back(float(output_ptr[i]));
@@ -2439,7 +2427,7 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_65x5x6x7_input_3x3_pool_4x4_stride_3
 
     { //FSV32 TOPOLOGY
         topology golden_topology;
-        golden_topology.add(input_layout("input", input_prim.get_layout()));
+        golden_topology.add(input_layout("input", input_prim->get_layout()));
         golden_topology.add(reorder("reorder_input", "input", layout(data_types::f16, format::fs_b_yx_fsv32, input_tensor, padding{ {0,0,x_in_pad, y_in_pad}, 0 })));
         golden_topology.add(pooling("fsv32_pooling", "reorder_input", pooling_mode::average, { 1,1,pool_size,pool_size }, { 1,1,stride_size,stride_size }, { 0,0,0,0 }, padding{ { 0,0,x_out_pad,y_out_pad },0 }));
         golden_topology.add(reorder("reorder_pooling", "fsv32_pooling", layout(data_types::f16, format::bfyx, input_tensor, padding{ { 0,0,x_out_pad,y_out_pad },0 })));
@@ -2448,7 +2436,7 @@ TEST(pooling_forward_gpu, fs_b_yx_fsv32_avg_65x5x6x7_input_3x3_pool_4x4_stride_3
         fsv32_network.set_input_data("input", input_prim);
 
         auto outputs = fsv32_network.execute();
-        auto output_ptr = outputs.begin()->second.get_memory().pointer<FLOAT16>();
+        cldnn::mem_lock<FLOAT16> output_ptr(outputs.begin()->second.get_memory(), get_test_stream());
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
             fsv32_results.push_back(float(output_ptr[i]));
@@ -2493,7 +2481,7 @@ class pooling_test_base {
 
     virtual void run_expect(const VVVVVF<output_t>& expected) {
 
-        auto eng = get_test_engine();
+        auto& eng = get_test_engine();
         auto topo = build_topology(eng);
         auto opts = build_options(
             build_option::optimize_data(true)
@@ -2504,7 +2492,7 @@ class pooling_test_base {
         auto input_lay = layout(input_type(),
                                 input_format(),
                                 input_size);
-        auto input_mem = memory::allocate(eng, input_lay);
+        auto input_mem = eng.allocate_memory(input_lay);
         std::vector<InputT> input_flat(input_lay.get_linear_size(), static_cast<InputT>(0));
         for (size_t bi = 0; bi < batch_num(); ++bi)
             for (size_t fi = 0; fi < input_features(); ++fi)
@@ -2520,8 +2508,8 @@ class pooling_test_base {
         net.set_input_data("input", input_mem);
         auto result = net.execute();
         auto out_mem = result.at(output_id()).get_memory();
-        auto out_lay = out_mem.get_layout();
-        auto out_ptr = out_mem.cldnn::memory::template pointer<output_t>();
+        auto out_lay = out_mem->get_layout();
+        cldnn::mem_lock<output_t> out_ptr(out_mem, get_test_stream());
 
         std::string kernel;
         for (auto i : net.get_primitives_info()) {
@@ -2779,12 +2767,12 @@ class pooling_scale_random_test_base : public pooling_random_test_base<InputT, M
     using parent = pooling_random_test_base<InputT, Mode>;
     using output_t = typename parent::output_t;
 
-    topology build_topology(const engine& eng) override {
+    topology build_topology(engine& eng) override {
         topology topo = parent::build_topology(eng);
 
         auto scale_lay = layout(this->output_type(), format::bfyx, tensor(batch(1), feature(this->input_features()), spatial(1, 1, 1, 1)));
-        auto scale_mem = memory::allocate(eng, scale_lay);
-        auto shift_mem = memory::allocate(eng, scale_lay);
+        auto scale_mem = eng.allocate_memory(scale_lay);
+        auto shift_mem = eng.allocate_memory(scale_lay);
         set_values(scale_mem, _scale);
         set_values(shift_mem, _shift);
 
@@ -2864,7 +2852,7 @@ INSTANTIATE_TEST_CASE_P(
 
 TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x8x8_input_2x2_pool_2x2_stride)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int features = 16;
     const int batches = 16;
@@ -2880,7 +2868,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x8x8_input_2x2_pool_2x2_stride)
 
     auto input_data = generate_random_1d<float>(batches * features * x_input * y_input, -10, 10);
 
-    auto input_prim = memory::allocate(engine, {data_types::f32, format::bfyx, input_tensor});
+    auto input_prim = engine.allocate_memory({data_types::f32, format::bfyx, input_tensor});
     set_values(input_prim, input_data);
 
     std::vector<float> golden_results;
@@ -2889,8 +2877,8 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x8x8_input_2x2_pool_2x2_stride)
     {
         //  golden topology
         topology golden_topology;
-        golden_topology.add(input_layout("input", input_prim.get_layout()));
-        golden_topology.add(reorder("reorder_input", "input", input_prim.get_layout()));
+        golden_topology.add(input_layout("input", input_prim->get_layout()));
+        golden_topology.add(reorder("reorder_input", "input", input_prim->get_layout()));
         golden_topology.add(pooling("golden_pooling", "reorder_input", pooling_mode::max, {1, 1, pool_size, pool_size},
                                     {1, 1, stride_size, stride_size}, {0, 0, -x_in_pad, -y_in_pad}));
 
@@ -2898,7 +2886,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x8x8_input_2x2_pool_2x2_stride)
         golden_network.set_input_data("input", input_prim);
 
         auto outputs = golden_network.execute();
-        auto output_ptr = outputs.begin()->second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.begin()->second.get_memory(), get_test_stream());
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
             golden_results.push_back(float(output_ptr[i]));
@@ -2908,7 +2896,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x8x8_input_2x2_pool_2x2_stride)
     {
         //  bfzyx_bsv16_fsv16 topology
         topology tested_topology;
-        tested_topology.add(input_layout("input", input_prim.get_layout()));
+        tested_topology.add(input_layout("input", input_prim->get_layout()));
         tested_topology.add(reorder("reorder_input", "input",
                                     layout(data_types::f32, format::bs_fs_yx_bsv16_fsv16, input_tensor)));
         tested_topology.add(pooling("bsv16_fsv16_pooling", "reorder_input", pooling_mode::max, {1, 1, pool_size, pool_size},
@@ -2922,9 +2910,9 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x8x8_input_2x2_pool_2x2_stride)
         bsv16_fsv16_network.set_input_data("input", input_prim);
 
         auto outputs = bsv16_fsv16_network.execute();
-        auto output_ptr = outputs.at("reorder_pooling").get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.at("reorder_pooling").get_memory(), get_test_stream());
 
-        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory().get_layout().format, format::bs_fs_yx_bsv16_fsv16);
+        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory()->get_layout().format, format::bs_fs_yx_bsv16_fsv16);
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -2948,7 +2936,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x8x8_input_2x2_pool_2x2_stride)
 
 TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x2x2_input_4x4_pool_1x1_stride_1x1_inpad)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int features = 16;
     const int batches = 16;
@@ -2964,7 +2952,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x2x2_input_4x4_pool_1x1_stride_1x
 
     auto input_data = generate_random_1d<float>(batches * features * x_input * y_input, -10, 10);
 
-    auto input_prim = memory::allocate(engine, {data_types::f32, format::bfyx, input_tensor});
+    auto input_prim = engine.allocate_memory({data_types::f32, format::bfyx, input_tensor});
     set_values(input_prim, input_data);
 
     std::vector<float> golden_results;
@@ -2973,8 +2961,8 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x2x2_input_4x4_pool_1x1_stride_1x
     {
         //  golden topology
         topology golden_topology;
-        golden_topology.add(input_layout("input", input_prim.get_layout()));
-        golden_topology.add(reorder("reorder_input", "input", input_prim.get_layout()));
+        golden_topology.add(input_layout("input", input_prim->get_layout()));
+        golden_topology.add(reorder("reorder_input", "input", input_prim->get_layout()));
         golden_topology.add(
                 pooling("golden_pooling", "reorder_input", pooling_mode::max, {1, 1, pool_size, pool_size},
                         {1, 1, stride_size, stride_size}, {0, 0, -x_in_pad, -y_in_pad}));
@@ -2983,7 +2971,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x2x2_input_4x4_pool_1x1_stride_1x
         golden_network.set_input_data("input", input_prim);
 
         auto outputs = golden_network.execute();
-        auto output_ptr = outputs.begin()->second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.begin()->second.get_memory(), get_test_stream());
         for (size_t i = 0; i < output_ptr.size(); i++) {
             golden_results.push_back(float(output_ptr[i]));
         }
@@ -2992,7 +2980,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x2x2_input_4x4_pool_1x1_stride_1x
     {
         //  bs_fs_yx_bsv16_fsv16 topology
         topology tested_topology;
-        tested_topology.add(input_layout("input", input_prim.get_layout()));
+        tested_topology.add(input_layout("input", input_prim->get_layout()));
         tested_topology.add(reorder("reorder_input", "input",
                                     layout(data_types::f32, format::bs_fs_yx_bsv16_fsv16, input_tensor)));
         tested_topology.add(
@@ -3006,9 +2994,9 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x2x2_input_4x4_pool_1x1_stride_1x
         bsv16_fsv16_network.set_input_data("input", input_prim);
 
         auto outputs = bsv16_fsv16_network.execute();
-        auto output_ptr = outputs.at("reorder_pooling").get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.at("reorder_pooling").get_memory(), get_test_stream());
 
-        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory().get_layout().format, format::bs_fs_yx_bsv16_fsv16);
+        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory()->get_layout().format, format::bs_fs_yx_bsv16_fsv16);
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3031,7 +3019,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x2x2_input_4x4_pool_1x1_stride_1x
 
 TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x3_stride)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int features = 16;
     const int batches = 16;
@@ -3047,7 +3035,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x3_stride)
 
     auto input_data = generate_random_1d<float>(batches * features * x_input * y_input, -10, 10);
 
-    auto input_prim = memory::allocate(engine, {data_types::f32, format::bfyx, input_tensor});
+    auto input_prim = engine.allocate_memory({data_types::f32, format::bfyx, input_tensor});
     set_values(input_prim, input_data);
 
     std::vector<float> golden_results;
@@ -3056,8 +3044,8 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x3_stride)
     {
         //  golden topology
         topology golden_topology;
-        golden_topology.add(input_layout("input", input_prim.get_layout()));
-        golden_topology.add(reorder("reorder_input", "input", input_prim.get_layout()));
+        golden_topology.add(input_layout("input", input_prim->get_layout()));
+        golden_topology.add(reorder("reorder_input", "input", input_prim->get_layout()));
         golden_topology.add(pooling("golden_pooling", "reorder_input", pooling_mode::average, {1, 1, pool_size, pool_size},
                                     {1, 1, stride_size, stride_size}, {0, 0, -x_in_pad, -y_in_pad}));
 
@@ -3065,7 +3053,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x3_stride)
         golden_network.set_input_data("input", input_prim);
 
         auto outputs = golden_network.execute();
-        auto output_ptr = outputs.begin()->second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.begin()->second.get_memory(), get_test_stream());
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3076,7 +3064,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x3_stride)
     {
         //  bs_fs_yx_bsv16_fsv16 topology
         topology tested_topology;
-        tested_topology.add(input_layout("input", input_prim.get_layout()));
+        tested_topology.add(input_layout("input", input_prim->get_layout()));
         tested_topology.add(reorder("reorder_input", "input",
                                     layout(data_types::f32, format::bs_fs_yx_bsv16_fsv16, input_tensor)));
         tested_topology.add(pooling("bsv16_fsv16_pooling", "reorder_input", pooling_mode::average, {1, 1, pool_size, pool_size},
@@ -3090,9 +3078,9 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x3_stride)
         bsv16_fsv16_network.set_input_data("input", input_prim);
 
         auto outputs = bsv16_fsv16_network.execute();
-        auto output_ptr = outputs.at("reorder_pooling").get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.at("reorder_pooling").get_memory(), get_test_stream());
 
-        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory().get_layout().format, format::bs_fs_yx_bsv16_fsv16);
+        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory()->get_layout().format, format::bs_fs_yx_bsv16_fsv16);
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3115,7 +3103,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x3_stride)
 
 TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x1_stride)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int features = 16;
     const int batches = 16;
@@ -3132,7 +3120,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x1_stride)
 
     auto input_data = generate_random_1d<float>(batches * features * x_input * y_input, -10, 10);
 
-    auto input_prim = memory::allocate(engine, {data_types::f32, format::bfyx, input_tensor});
+    auto input_prim = engine.allocate_memory({data_types::f32, format::bfyx, input_tensor});
     set_values(input_prim, input_data);
 
     std::vector<float> golden_results;
@@ -3141,8 +3129,8 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x1_stride)
     {
         //  golden topology
         topology golden_topology;
-        golden_topology.add(input_layout("input", input_prim.get_layout()));
-        golden_topology.add(reorder("reorder_input", "input", input_prim.get_layout()));
+        golden_topology.add(input_layout("input", input_prim->get_layout()));
+        golden_topology.add(reorder("reorder_input", "input", input_prim->get_layout()));
         golden_topology.add(pooling("golden_pooling", "reorder_input", pooling_mode::average, {1, 1, pool_size, pool_size},
                                     {1, 1, stride_size_x, stride_size_y}, {0, 0, -x_in_pad, -y_in_pad}));
 
@@ -3150,7 +3138,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x1_stride)
         golden_network.set_input_data("input", input_prim);
 
         auto outputs = golden_network.execute();
-        auto output_ptr = outputs.begin()->second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.begin()->second.get_memory(), get_test_stream());
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3161,7 +3149,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x1_stride)
     {
         //  bs_fs_yx_bsv16_fsv16 topology
         topology tested_topology;
-        tested_topology.add(input_layout("input", input_prim.get_layout()));
+        tested_topology.add(input_layout("input", input_prim->get_layout()));
         tested_topology.add(reorder("reorder_input", "input", layout(data_types::f32, format::bs_fs_yx_bsv16_fsv16, input_tensor)));
         tested_topology.add(pooling("bsv16_fsv16_pooling", "reorder_input", pooling_mode::average, {1, 1, pool_size, pool_size},
                                     {1, 1, stride_size_x, stride_size_y}, {0, 0, -x_in_pad, -y_in_pad}));
@@ -3173,9 +3161,9 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x1_stride)
         bsv16_fsv16_network.set_input_data("input", input_prim);
 
         auto outputs = bsv16_fsv16_network.execute();
-        auto output_ptr = outputs.at("reorder_pooling").get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.at("reorder_pooling").get_memory(), get_test_stream());
 
-        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory().get_layout().format, format::bs_fs_yx_bsv16_fsv16);
+        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory()->get_layout().format, format::bs_fs_yx_bsv16_fsv16);
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3198,7 +3186,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_avg_16x16x20x20_input_5x5_pool_3x1_stride)
 
 TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x20x20_input_5x5_pool_3x1_stride)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int features = 16;
     const int batches = 16;
@@ -3215,7 +3203,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x20x20_input_5x5_pool_3x1_stride)
 
     auto input_data = generate_random_1d<float>(batches * features * x_input * y_input, -10, 10);
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::bfyx,input_tensor });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::bfyx,input_tensor });
     set_values(input_prim, input_data);
 
     std::vector<float> golden_results;
@@ -3224,8 +3212,8 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x20x20_input_5x5_pool_3x1_stride)
     {
         //  golden topology
         topology golden_topology;
-        golden_topology.add(input_layout("input", input_prim.get_layout()));
-        golden_topology.add(reorder("reorder_input", "input", input_prim.get_layout()));
+        golden_topology.add(input_layout("input", input_prim->get_layout()));
+        golden_topology.add(reorder("reorder_input", "input", input_prim->get_layout()));
         golden_topology.add(pooling("golden_pooling", "reorder_input", pooling_mode::max, {1, 1, pool_size, pool_size},
                                     {1, 1, stride_size_x, stride_size_y}, {0, 0, -x_in_pad, -y_in_pad}));
 
@@ -3233,7 +3221,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x20x20_input_5x5_pool_3x1_stride)
         golden_network.set_input_data("input", input_prim);
 
         auto outputs = golden_network.execute();
-        auto output_ptr = outputs.begin()->second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.begin()->second.get_memory(), get_test_stream());
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3244,7 +3232,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x20x20_input_5x5_pool_3x1_stride)
     {
         //  bs_fs_yx_bsv16_fsv16 topology
         topology tested_topology;
-        tested_topology.add(input_layout("input", input_prim.get_layout()));
+        tested_topology.add(input_layout("input", input_prim->get_layout()));
         tested_topology.add(reorder("reorder_input", "input",
                                     layout(data_types::f32, format::bs_fs_yx_bsv16_fsv16, input_tensor)));
         tested_topology.add(
@@ -3258,9 +3246,9 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x20x20_input_5x5_pool_3x1_stride)
         bsv16_fsv16_network.set_input_data("input", input_prim);
 
         auto outputs = bsv16_fsv16_network.execute();
-        auto output_ptr = outputs.at("reorder_pooling").get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.at("reorder_pooling").get_memory(), get_test_stream());
 
-        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory().get_layout().format, format::bs_fs_yx_bsv16_fsv16);
+        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory()->get_layout().format, format::bs_fs_yx_bsv16_fsv16);
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3283,7 +3271,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_16x16x20x20_input_5x5_pool_3x1_stride)
 
 TEST(pooling_forward_gpu, bsv16_fsv16_max_32x32x20x20_input_5x5_pool_3x1_stride)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int features = 32;
     const int batches = 32;
@@ -3300,7 +3288,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x32x20x20_input_5x5_pool_3x1_stride)
 
     auto input_data = generate_random_1d<float>(batches * features * x_input * y_input, -10, 10);
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::bfyx,input_tensor });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::bfyx,input_tensor });
     set_values(input_prim, input_data);
 
     std::vector<float> golden_results;
@@ -3309,8 +3297,8 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x32x20x20_input_5x5_pool_3x1_stride)
     {
         //  golden topology
         topology golden_topology;
-        golden_topology.add(input_layout("input", input_prim.get_layout()));
-        golden_topology.add(reorder("reorder_input", "input", input_prim.get_layout()));
+        golden_topology.add(input_layout("input", input_prim->get_layout()));
+        golden_topology.add(reorder("reorder_input", "input", input_prim->get_layout()));
         golden_topology.add(pooling("golden_pooling", "reorder_input", pooling_mode::max, {1, 1, pool_size, pool_size},
                                     {1, 1, stride_size_x, stride_size_y}, {0, 0, -x_in_pad, -y_in_pad}));
 
@@ -3318,7 +3306,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x32x20x20_input_5x5_pool_3x1_stride)
         golden_network.set_input_data("input", input_prim);
 
         auto outputs = golden_network.execute();
-        auto output_ptr = outputs.begin()->second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.begin()->second.get_memory(), get_test_stream());
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3329,7 +3317,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x32x20x20_input_5x5_pool_3x1_stride)
     {
         //  bs_fs_yx_bsv16_fsv16 topology
         topology tested_topology;
-        tested_topology.add(input_layout("input", input_prim.get_layout()));
+        tested_topology.add(input_layout("input", input_prim->get_layout()));
         tested_topology.add(reorder("reorder_input", "input",
                                     layout(data_types::f32, format::bs_fs_yx_bsv16_fsv16, input_tensor)));
         tested_topology.add(
@@ -3343,9 +3331,9 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x32x20x20_input_5x5_pool_3x1_stride)
         bsv16_fsv16_network.set_input_data("input", input_prim);
 
         auto outputs = bsv16_fsv16_network.execute();
-        auto output_ptr = outputs.at("reorder_pooling").get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.at("reorder_pooling").get_memory(), get_test_stream());
 
-        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory().get_layout().format, format::bs_fs_yx_bsv16_fsv16);
+        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory()->get_layout().format, format::bs_fs_yx_bsv16_fsv16);
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3368,7 +3356,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x32x20x20_input_5x5_pool_3x1_stride)
 
 TEST(pooling_forward_gpu, bsv16_fsv16_max_32x16x20x20_input_5x5_pool_3x1_stride)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int features = 16;
     const int batches = 32;
@@ -3389,7 +3377,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x16x20x20_input_5x5_pool_3x1_stride)
         input_data[i] = static_cast<float>(i);
     }
 
-    auto input_prim = memory::allocate(engine, { data_types::f32,format::bfyx,input_tensor });
+    auto input_prim = engine.allocate_memory({ data_types::f32,format::bfyx,input_tensor });
     set_values(input_prim, input_data);
 
     std::vector<float> golden_results;
@@ -3398,8 +3386,8 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x16x20x20_input_5x5_pool_3x1_stride)
     {
         //  golden topology
         topology golden_topology;
-        golden_topology.add(input_layout("input", input_prim.get_layout()));
-        golden_topology.add(reorder("reorder_input", "input", input_prim.get_layout()));
+        golden_topology.add(input_layout("input", input_prim->get_layout()));
+        golden_topology.add(reorder("reorder_input", "input", input_prim->get_layout()));
         golden_topology.add(pooling("golden_pooling", "reorder_input", pooling_mode::max, {1, 1, pool_size, pool_size},
                                     {1, 1, stride_size_x, stride_size_y}, {0, 0, -x_in_pad, -y_in_pad}));
 
@@ -3407,7 +3395,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x16x20x20_input_5x5_pool_3x1_stride)
         golden_network.set_input_data("input", input_prim);
 
         auto outputs = golden_network.execute();
-        auto output_ptr = outputs.begin()->second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.begin()->second.get_memory(), get_test_stream());
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3418,7 +3406,7 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x16x20x20_input_5x5_pool_3x1_stride)
     {
         //  bs_fs_yx_bsv16_fsv16 topology
         topology tested_topology;
-        tested_topology.add(input_layout("input", input_prim.get_layout()));
+        tested_topology.add(input_layout("input", input_prim->get_layout()));
         tested_topology.add(reorder("reorder_input", "input",
                                     layout(data_types::f32, format::bs_fs_yx_bsv16_fsv16, input_tensor)));
         tested_topology.add(
@@ -3432,9 +3420,9 @@ TEST(pooling_forward_gpu, bsv16_fsv16_max_32x16x20x20_input_5x5_pool_3x1_stride)
         bsv16_fsv16_network.set_input_data("input", input_prim);
 
         auto outputs = bsv16_fsv16_network.execute();
-        auto output_ptr = outputs.at("reorder_pooling").get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(outputs.at("reorder_pooling").get_memory(), get_test_stream());
 
-        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory().get_layout().format, format::bs_fs_yx_bsv16_fsv16);
+        ASSERT_EQ(outputs.at("bsv16_fsv16_pooling").get_memory()->get_layout().format, format::bs_fs_yx_bsv16_fsv16);
 
         for (size_t i = 0; i < output_ptr.size(); i++)
         {
@@ -3521,7 +3509,7 @@ class pooling_test : public tests::generic_test
         return false;
     }
 
-    virtual void prepare_input_for_test(std::vector<cldnn::memory>& inputs)
+    virtual void prepare_input_for_test(std::vector<cldnn::memory::ptr>& inputs)
     {
         if (generic_params->data_type == data_types::f32)
         {
@@ -3534,13 +3522,13 @@ class pooling_test : public tests::generic_test
     }
 
     template<typename Type>
-    void prepare_input_for_test_typed(std::vector<cldnn::memory>& inputs)
+    void prepare_input_for_test_typed(std::vector<cldnn::memory::ptr>& inputs)
     {
         int k = (generic_params->data_type == data_types::f32) ? 8 : 4;
         auto input = inputs[0];
-        auto input_size = inputs[0].get_layout().size;
+        auto input_size = inputs[0]->get_layout().size;
         VVVVF<Type> input_rnd = generate_random_4d<Type>(input_size.batch[0], input_size.feature[0], input_size.spatial[1], input_size.spatial[0], -10, 10, k);
-        VF<Type> input_rnd_vec = flatten_4d<Type>(input.get_layout().format, input_rnd);
+        VF<Type> input_rnd_vec = flatten_4d<Type>(input->get_layout().format, input_rnd);
         set_values(input, input_rnd_vec);
     }
 
@@ -3579,14 +3567,13 @@ class pooling_test : public tests::generic_test
     }
 
     template<typename Type>
-    memory generate_reference_typed(const std::vector<cldnn::memory>& inputs)
-    {
+    memory::ptr generate_reference_typed(const std::vector<cldnn::memory::ptr>& inputs) {
         auto pooling = std::static_pointer_cast<cldnn::pooling>(layer_params);
 
-        int batch = inputs[0].get_layout().size.batch[0];
-        int feature = inputs[0].get_layout().size.feature[0];
-        int height = inputs[0].get_layout().size.spatial[1];
-        int width = inputs[0].get_layout().size.spatial[0];
+        int batch = inputs[0]->get_layout().size.batch[0];
+        int feature = inputs[0]->get_layout().size.feature[0];
+        int height = inputs[0]->get_layout().size.spatial[1];
+        int width = inputs[0]->get_layout().size.spatial[0];
 
         cldnn::pooling_mode pooling_mode = pooling->mode;
 
@@ -3605,22 +3592,22 @@ class pooling_test : public tests::generic_test
         int pooled_height = output_tensor.spatial[1];
 
         //Output is bfyx
-        auto output = memory::allocate(engine, cldnn::layout(inputs[0].get_layout().data_type, cldnn::format::bfyx, output_tensor, pooling->output_padding));
+        auto output = engine.allocate_memory(cldnn::layout(inputs[0]->get_layout().data_type, cldnn::format::bfyx, output_tensor, pooling->output_padding));
 
-        auto input_mem = inputs[0].pointer<Type>();
-        auto output_mem = output.pointer<Type>();
+        cldnn::mem_lock<Type> input_mem(inputs[0], get_test_stream());
+        cldnn::mem_lock<Type> output_mem(output, get_test_stream());
 
-        int output_width = output.get_layout().get_buffer_size().spatial[0];
-        int output_height = output.get_layout().get_buffer_size().spatial[1];
+        int output_width = output->get_layout().get_buffer_size().spatial[0];
+        int output_height = output->get_layout().get_buffer_size().spatial[1];
 
-        const auto input_desc = get_linear_memory_desc(inputs[0].get_layout());
-        const auto output_desc = get_linear_memory_desc(output.get_layout());
+        const auto input_desc = get_linear_memory_desc(inputs[0]->get_layout());
+        const auto output_desc = get_linear_memory_desc(output->get_layout());
 
         switch (pooling_mode)
         {
             case cldnn::pooling_mode::max:
             {
-                for (int i = 0; i < (int)output.get_layout().get_buffer_size().count(); i++)
+                for (int i = 0; i < (int)output->get_layout().get_buffer_size().count(); i++)
                 {
                     output_mem[i] = (generic_params->data_type == data_types::f32) ? -FLT_MAX : -65504;
                 }
@@ -3640,13 +3627,13 @@ class pooling_test : public tests::generic_test
                                 int input_offset_y_end = std::min(input_offset_y_start + kernel_height, height);
                                 input_offset_y_start = std::max(input_offset_y_start, 0);
 
-                                const size_t output_index = get_linear_index(output.get_layout(), b, f, h, w, output_desc);
+                                const size_t output_index = get_linear_index(output->get_layout(), b, f, h, w, output_desc);
 
                                 for (int y = input_offset_y_start; y < input_offset_y_end; y++)
                                 {
                                     for (int x = input_offset_x_start; x < input_offset_x_end; x++)
                                     {
-                                        const size_t input_index = get_linear_index(inputs[0].get_layout(), b, f, y, x, input_desc);
+                                        const size_t input_index = get_linear_index(inputs[0]->get_layout(), b, f, y, x, input_desc);
 
                                         if (input_mem[input_index] > output_mem[output_index])
                                         {
@@ -3683,7 +3670,7 @@ class pooling_test : public tests::generic_test
                     return y*x;
                 };
 
-                for (int i = 0; i < (int)output.get_layout().get_buffer_size().count(); i++)
+                for (int i = 0; i < (int)output->get_layout().get_buffer_size().count(); i++)
                 {
                     output_mem[i] = 0;
                 }
@@ -3712,7 +3699,7 @@ class pooling_test : public tests::generic_test
                                 {
                                     for (int x = input_offset_x_start; x < input_offset_x_end; x++)
                                     {
-                                        const size_t input_index = get_linear_index(inputs[0].get_layout(), b, f, y, x, input_desc);
+                                        const size_t input_index = get_linear_index(inputs[0]->get_layout(), b, f, y, x, input_desc);
                                         output_mem[output_index] += input_mem[input_index];
                                         if (!dynamic_mode || pooling_mode == cldnn::pooling_mode::average_no_padding)
                                         {
@@ -3746,7 +3733,7 @@ class pooling_test : public tests::generic_test
         return output;
     }
 
-    virtual memory generate_reference(const std::vector<cldnn::memory>& inputs)
+    virtual memory::ptr generate_reference(const std::vector<cldnn::memory::ptr>& inputs)
     {
         if (generic_params->data_type == data_types::f32)
         {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/propagate_constants_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/propagate_constants_gpu_test.cpp
index 264eb954d3936c..3425bfd427c773 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/propagate_constants_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/propagate_constants_gpu_test.cpp
@@ -2,40 +2,34 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/concatenation.hpp>
-#include <api/reorder.hpp>
-#include <api/data.hpp>
-#include <api/reshape.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/concatenation.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/reshape.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 //We expect additional reorder to be added in between "weights1" and "reshape1".
 //This situation should be handled properly by propagate constants optimization phase
 TEST(propagate_constants, copy_dependecies_from_nodes) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
     build_opt.set_option(build_option::optimize_data(true));
 
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
-    auto weights1 = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 1 } });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::byxf,{ 1, 1, 1, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto weights1 = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 1 } });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::byxf,{ 1, 1, 1, 2 } });
 
     set_values(input, { FLOAT16(1.1f), FLOAT16(1.2f), FLOAT16(1.3f), FLOAT16(1.4f) });
     set_values(weights1, { FLOAT16(2.1f), FLOAT16(3.1f) });
     set_values(weights2, { 1.1f, 0.1f });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights1", weights1));
     topology.add(data("weights2", weights2));
     topology.add(reshape("reshape1", "weights1", tensor(spatial(1, 2))));
@@ -49,9 +43,8 @@ TEST(propagate_constants, copy_dependecies_from_nodes) {
     auto outputs = network.execute();
 
     float epsilon = 1e-2f;
-    for (auto& it : outputs)
-    {
-        auto output = it.second.get_memory().pointer<float>();
+    for (auto& it : outputs) {
+        cldnn::mem_lock<float> output(it.second.get_memory(), get_test_stream());
         EXPECT_NEAR(7.8f, output[0], epsilon);
     }
-}
\ No newline at end of file
+}
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/proposal_cpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/proposal_cpu_test.cpp
index 7cbec32a861bfb..a7ea1040828e28 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/proposal_cpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/proposal_cpu_test.cpp
@@ -2,18 +2,12 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <fstream>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/proposal.hpp>
 
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include <api/proposal.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include "test_utils/float16.h"
+#include <fstream>
 
 namespace cldnn
 {
@@ -21,7 +15,7 @@ template<> struct type_to_data_type<FLOAT16> { static const data_types value = d
 }
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 using namespace std;
 
 extern float cls_scores_data[];
@@ -65,11 +59,10 @@ class TestRunnerProposal
     public:
         explicit TestRunnerProposal(cldnn::tensor image_info_size);
 
-        memory Run(std::vector<Dtype>& data,
-                   std::vector<Dtype>& rois);
+        memory::ptr Run(std::vector<Dtype>& data,
+                        std::vector<Dtype>& rois);
 
     private:
-        engine _engine;
         layout _cls_scores_layout;
         layout _bbox_pred_layout;
         layout _image_info_layout;
@@ -103,20 +96,21 @@ TestRunnerProposal<Dtype, ImInfoType>::TestRunnerProposal(cldnn::tensor image_in
 
     _topology.add(_test_layer);
 
-    _network.reset(new network(_engine, _topology));
+    _network.reset(new network(get_test_engine(), _topology));
 }
 
 template <typename Dtype, typename ImInfoType>
-memory TestRunnerProposal<Dtype, ImInfoType>::Run(std::vector<Dtype>& cls_scores_vals,
-                                                  std::vector<Dtype>& bbox_pred_vals)
+memory::ptr TestRunnerProposal<Dtype, ImInfoType>::Run(std::vector<Dtype>& cls_scores_vals,
+                                                       std::vector<Dtype>& bbox_pred_vals)
 {
-    memory cls_scores = memory::attach(_cls_scores_layout, cls_scores_vals.data(), cls_scores_vals.size());
-    memory bbox_pred  = memory::attach(_bbox_pred_layout, bbox_pred_vals.data(), bbox_pred_vals.size());
+    auto& engine = get_test_engine();
+    memory::ptr cls_scores = engine.attach_memory(_cls_scores_layout, cls_scores_vals.data());
+    memory::ptr bbox_pred  = engine.attach_memory(_bbox_pred_layout, bbox_pred_vals.data());
 
     std::vector<ImInfoType> image_info_vals = { (ImInfoType)((float)image_h - 0.0000001f), // check fp robustness of the layer
                                                 (ImInfoType)((float)image_w + 0.0000001f), // check fp robustness of the layer
                                                 (ImInfoType)((float)image_z) };
-    memory image_info = memory::allocate(_engine, _image_info_layout);
+    memory::ptr image_info = engine.allocate_memory(_image_info_layout);
     tests::set_values(image_info, image_info_vals);
 
     _network->set_input_data(cls_scores_name, cls_scores);
@@ -134,10 +128,10 @@ TEST(proposal, basic) {
 
     TestRunnerProposal<float> t({ 1, 3, 1, 1 });
 
-    const memory& output = t.Run(cls_scores, bbox_pred);
-    ASSERT_EQ(output.get_layout().count(), proposal_ref_size);
+    memory::ptr output = t.Run(cls_scores, bbox_pred);
+    ASSERT_EQ(output->get_layout().count(), proposal_ref_size);
 
-    auto f = output.pointer<float>();
+    cldnn::mem_lock<float> f(output, get_test_stream());
 
     for (size_t i = 0; i < proposal_ref_size; i++) {
         EXPECT_NEAR(f[i], proposal_ref[i], epsilon);
@@ -150,10 +144,10 @@ TEST(proposal, fp16) {
 
     TestRunnerProposal<FLOAT16> t({ 1, 3, 1, 1 });
 
-    const memory& output = t.Run(cls_scores, bbox_pred);
-    ASSERT_EQ(output.get_layout().count(), proposal_ref_size);
+    memory::ptr output = t.Run(cls_scores, bbox_pred);
+    ASSERT_EQ(output->get_layout().count(), proposal_ref_size);
 
-    auto d = output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> d(output, get_test_stream());
 
     for (size_t i = 0; i < proposal_ref_size; i++) {
         FLOAT16 ref(proposal_ref[i]);
@@ -167,10 +161,10 @@ TEST(proposal, scores_fp16_im_info_fp32) {
 
     TestRunnerProposal<FLOAT16, float> t({ 1, 3, 1, 1 });
 
-    const memory& output = t.Run(cls_scores, bbox_pred);
-    ASSERT_EQ(output.get_layout().count(), proposal_ref_size);
+    memory::ptr output = t.Run(cls_scores, bbox_pred);
+    ASSERT_EQ(output->get_layout().count(), proposal_ref_size);
 
-    auto d = output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> d(output, get_test_stream());
 
     for (size_t i = 0; i < proposal_ref_size; i++) {
         FLOAT16 ref(proposal_ref[i]);
@@ -184,10 +178,10 @@ TEST(proposal, scores_fp32_im_info_fp16) {
 
     TestRunnerProposal<float, FLOAT16> t({ 1, 3, 1, 1 });
 
-    const memory& output = t.Run(cls_scores, bbox_pred);
-    ASSERT_EQ(output.get_layout().count(), proposal_ref_size);
+    memory::ptr output = t.Run(cls_scores, bbox_pred);
+    ASSERT_EQ(output->get_layout().count(), proposal_ref_size);
 
-    auto d = output.pointer<float>();
+    cldnn::mem_lock<float> d(output, get_test_stream());
 
     for (size_t i = 0; i < proposal_ref_size; i++) {
         float ref(proposal_ref[i]);
@@ -201,10 +195,10 @@ TEST(proposal, img_info_batched) {
 
     TestRunnerProposal<float> t({ 2, 3, 1, 1 });
 
-    const memory& output = t.Run(cls_scores, bbox_pred);
-    ASSERT_EQ(output.get_layout().count(), proposal_ref_size);
+    memory::ptr output = t.Run(cls_scores, bbox_pred);
+    ASSERT_EQ(output->get_layout().count(), proposal_ref_size);
 
-    auto f = output.pointer<float>();
+    cldnn::mem_lock<float> f(output, get_test_stream());
 
     for (size_t i = 0; i < proposal_ref_size; i++) {
         EXPECT_NEAR(f[i], proposal_ref[i], epsilon);
@@ -217,10 +211,10 @@ TEST(proposal, img_info_batch_only) {
 
     TestRunnerProposal<float> t({ 3, 1, 1, 1 });
 
-    const memory& output = t.Run(cls_scores, bbox_pred);
-    ASSERT_EQ(output.get_layout().count(), proposal_ref_size);
+    memory::ptr output = t.Run(cls_scores, bbox_pred);
+    ASSERT_EQ(output->get_layout().count(), proposal_ref_size);
 
-    auto f = output.pointer<float>();
+    cldnn::mem_lock<float> f(output, get_test_stream());
 
     for (size_t i = 0; i < proposal_ref_size; i++) {
         EXPECT_NEAR(f[i], proposal_ref[i], epsilon);
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/pyramid_roi_align_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/pyramid_roi_align_gpu_test.cpp
index 58afd3089287c8..a4e168ee9a018c 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/pyramid_roi_align_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/pyramid_roi_align_gpu_test.cpp
@@ -3,21 +3,15 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/engine.hpp>
-#include <api/input_layout.hpp>
-#include <api/pyramid_roi_align.hpp>
-#include <api/memory.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/mutable_data.hpp>
-#include <api/data.hpp>
+#include "test_utils.h"
 
-#include "test_utils/test_utils.h"
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/pyramid_roi_align.hpp>
+#include <cldnn/primitives/data.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 template <typename T>
 struct pyramid_roi_align_typed_test : testing::Test {
@@ -31,7 +25,7 @@ TYPED_TEST_CASE(pyramid_roi_align_typed_test, pyramid_roi_align_types);
 TYPED_TEST(pyramid_roi_align_typed_test, smoke_4levels) {
     using Type = typename pyramid_roi_align_typed_test<TypeParam>::Type;
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     const int rois_num = 3;
     const int output_size = 2;
@@ -85,11 +79,11 @@ TYPED_TEST(pyramid_roi_align_typed_test, smoke_4levels) {
     auto P4_lay = layout(this->data_type, format::bfyx, tensor(1, 1, P4_size, P4_size));
     auto P5_lay = layout(this->data_type, format::bfyx, tensor(1, 1, P5_size, P5_size));
 
-    auto rois_mem = memory::allocate(engine, rois_lay);
-    auto P2_mem = memory::allocate(engine, P2_lay);
-    auto P3_mem = memory::allocate(engine, P3_lay);
-    auto P4_mem = memory::allocate(engine, P4_lay);
-    auto P5_mem = memory::allocate(engine, P5_lay);
+    auto rois_mem = engine.allocate_memory(rois_lay);
+    auto P2_mem = engine.allocate_memory(P2_lay);
+    auto P3_mem = engine.allocate_memory(P3_lay);
+    auto P4_mem = engine.allocate_memory(P4_lay);
+    auto P5_mem = engine.allocate_memory(P5_lay);
 
     tests::set_values(rois_mem, rois_data);
     tests::set_values(P2_mem, P2_data);
@@ -129,7 +123,7 @@ TYPED_TEST(pyramid_roi_align_typed_test, smoke_4levels) {
     auto result = net.execute();
 
     auto out_mem = result.at("pyramid").get_memory();
-    auto out_ptr = out_mem.pointer<Type>();
+    cldnn::mem_lock<Type> out_ptr(out_mem, get_test_stream());
 
     ASSERT_EQ(expected_out.size(), out_ptr.size());
     for (size_t i = 0; i < expected_out.size(); ++i) {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/quantize_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/quantize_gpu_test.cpp
index 1d0dfaa4beb219..f5090ab08ff1c0 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/quantize_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/quantize_gpu_test.cpp
@@ -3,31 +3,25 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/engine.hpp>
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/quantize.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include "test_utils.h"
 
-#include "test_utils/test_utils.h"
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/quantize.hpp>
+#include <cldnn/primitives/data.hpp>
 
 #include <cstddef>
-#include <api/data.hpp>
-#include <src/include/to_string_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
 
 TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 16, 2, 2}});
-    auto input_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
-    auto input_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
-    auto output_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
-    auto output_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 16, 2, 2}});
+    auto input_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
+    auto input_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
+    auto output_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto output_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { -1.0f, 2.0f, 3.0f, 4.0f,
                          5.0f, 2.0f, 2.0f, 3.0f,
@@ -83,7 +77,7 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("input_low", input_low),
         data("input_high", input_high),
         data("output_low", output_low),
@@ -96,13 +90,13 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("quantize").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     // Check that layout and memory contains logical size of tensor
-    ASSERT_EQ(output.count(), (size_t)64);
-    ASSERT_EQ(output.get_layout().count(), (size_t)64);
+    ASSERT_EQ(output->count(), (size_t)64);
+    ASSERT_EQ(output->get_layout().count(), (size_t)64);
 
-    ASSERT_EQ(output.size(), ref_data.size() * sizeof(uint32_t));
+    ASSERT_EQ(output->size(), ref_data.size() * sizeof(uint32_t));
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_EQ(output_ptr[i], ref_data[i]) << " index = " << i;
@@ -110,11 +104,11 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1) {
 }
 
 TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1_ch8) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 8, 2, 2}});
-    auto input_thresh = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 8, 1, 1 } });
-    auto output_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
-    auto output_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 8, 2, 2}});
+    auto input_thresh = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 8, 1, 1 } });
+    auto output_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto output_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { -1.0f, 2.0f, 3.0f, 4.0f,
                          5.0f, 2.0f, 2.0f, 3.0f,
@@ -147,7 +141,7 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1_ch8) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("input_low", input_thresh),
         data("input_high", input_thresh),
         data("output_low", output_low),
@@ -160,13 +154,13 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1_ch8) {
     auto outputs = network.execute();
 
     auto output = outputs.at("quantize").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     // Check that layout and memory contains logical size of tensor
-    ASSERT_EQ(output.count(), (size_t)32);
-    ASSERT_EQ(output.get_layout().count(), (size_t)32);
+    ASSERT_EQ(output->count(), (size_t)32);
+    ASSERT_EQ(output->get_layout().count(), (size_t)32);
 
-    ASSERT_EQ(output.size(), ref_data.size() * sizeof(uint32_t));
+    ASSERT_EQ(output->size(), ref_data.size() * sizeof(uint32_t));
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_EQ(output_ptr[i], ref_data[i]) << " index = " << i;
@@ -174,11 +168,11 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1_ch8) {
 }
 
 TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1_ch8_binary_pack) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 8, 2, 2}});
-    auto input_thresh = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 8, 1, 1 } });
-    auto output_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
-    auto output_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 8, 2, 2}});
+    auto input_thresh = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 8, 1, 1 } });
+    auto output_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto output_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { -1.0f, 2.0f, 3.0f, 4.0f,
                          5.0f, 2.0f, 2.0f, 3.0f,
@@ -210,7 +204,7 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1_ch8_binary_pack)
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("input_low", input_thresh),
         data("input_high", input_thresh),
         data("output_low", output_low),
@@ -226,13 +220,13 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1_ch8_binary_pack)
     auto outputs = network.execute();
 
     auto output = outputs.at("reorder").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     // Check that layout and memory contains logical size of tensor
-    ASSERT_EQ(output.count(), (size_t)32);
-    ASSERT_EQ(output.get_layout().count(), (size_t)32);
+    ASSERT_EQ(output->count(), (size_t)32);
+    ASSERT_EQ(output->get_layout().count(), (size_t)32);
 
-    ASSERT_EQ(output.size(), ref_data.size() * sizeof(uint32_t));
+    ASSERT_EQ(output->size(), ref_data.size() * sizeof(uint32_t));
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_EQ(output_ptr[i], ref_data[i]) << " index = " << i;
@@ -240,12 +234,12 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_1_ch8_binary_pack)
 }
 
 TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_2) {
-    const cldnn::engine& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 16, 2, 2}});
-    auto input_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
-    auto input_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
-    auto output_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
-    auto output_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    cldnn::engine& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 16, 2, 2}});
+    auto input_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto input_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto output_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto output_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { -1.0f, 2.0f, 3.0f, 4.0f,
                          5.0f, 2.0f, 2.0f, 3.0f,
@@ -291,7 +285,7 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_2) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("input_low", input_low),
         data("input_high", input_high),
         data("output_low", output_low),
@@ -304,13 +298,13 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("quantize").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     // Check that layout and memory contains logical size of tensor
-    ASSERT_EQ(output.count(), (size_t)64);
-    ASSERT_EQ(output.get_layout().count(), (size_t)64);
+    ASSERT_EQ(output->count(), (size_t)64);
+    ASSERT_EQ(output->get_layout().count(), (size_t)64);
 
-    ASSERT_EQ(output.size(), ref_data.size() * sizeof(float));
+    ASSERT_EQ(output->size(), ref_data.size() * sizeof(float));
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_EQ(output_ptr[i], ref_data[i]) << " index = " << i;
@@ -318,12 +312,12 @@ TEST(quantize_gpu, quantize_levels_2_output_broadcast_inputs_2) {
 }
 
 TEST(quantize_gpu, quantize_levels_3) {
-    const cldnn::engine& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 16, 2, 2}});
-    auto input_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
-    auto input_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
-    auto output_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
-    auto output_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    cldnn::engine& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 16, 2, 2}});
+    auto input_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
+    auto input_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
+    auto output_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto output_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { -1.0f, 2.0f, 3.0f, 4.0f,
                          5.0f, 2.0f, 2.0f, 3.0f,
@@ -380,7 +374,7 @@ TEST(quantize_gpu, quantize_levels_3) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("input_low", input_low),
         data("input_high", input_high),
         data("output_low", output_low),
@@ -393,14 +387,14 @@ TEST(quantize_gpu, quantize_levels_3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("quantize").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     // Check that layout and memory contains logical size of tensor
-    ASSERT_EQ(output.count(), ref_data.size());
-    ASSERT_EQ(output.get_layout().count(), ref_data.size());
+    ASSERT_EQ(output->count(), ref_data.size());
+    ASSERT_EQ(output->get_layout().count(), ref_data.size());
 
     // Check that memory physical size consider binary pack
-    ASSERT_EQ(output.size(), ref_data.size() * sizeof(float));
+    ASSERT_EQ(output->size(), ref_data.size() * sizeof(float));
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_EQ(output_ptr[i], ref_data[i]) << " i=" << i;
@@ -408,12 +402,12 @@ TEST(quantize_gpu, quantize_levels_3) {
 }
 
 TEST(quantize_gpu, quantize_levels_256_2d_unsigned) {
-    const cldnn::engine& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 16, 2, 2}});
-    auto input_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
-    auto input_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
-    auto output_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
-    auto output_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    cldnn::engine& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 16, 2, 2}});
+    auto input_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
+    auto input_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
+    auto output_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto output_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { -1.0f, 2.1f, 3.0f, 4.0f,
                          5.0f, 2.0f, 2.0f, 3.0f,
@@ -471,7 +465,7 @@ TEST(quantize_gpu, quantize_levels_256_2d_unsigned) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("input_low", input_low),
         data("input_high", input_high),
         data("output_low", output_low),
@@ -484,14 +478,14 @@ TEST(quantize_gpu, quantize_levels_256_2d_unsigned) {
     auto outputs = network.execute();
 
     auto output = outputs.at("quantize").get_memory();
-    auto output_ptr = output.pointer<uint8_t>();
+    cldnn::mem_lock<uint8_t> output_ptr(output, get_test_stream());
 
     // Check that layout and memory contains logical size of tensor
-    ASSERT_EQ(output.count(), ref_data.size());
-    ASSERT_EQ(output.get_layout().count(), ref_data.size());
+    ASSERT_EQ(output->count(), ref_data.size());
+    ASSERT_EQ(output->get_layout().count(), ref_data.size());
 
     // Check that memory physical size consider binary pack
-    ASSERT_EQ(output.size(), ref_data.size() * sizeof(uint8_t));
+    ASSERT_EQ(output->size(), ref_data.size() * sizeof(uint8_t));
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_EQ(output_ptr[i], ref_data[i]) << " i=" << i;
@@ -499,12 +493,12 @@ TEST(quantize_gpu, quantize_levels_256_2d_unsigned) {
 }
 
 TEST(quantize_gpu, quantize_levels_256_3d_unsigned) {
-    const cldnn::engine& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfzyx, {1, 16, 2, 1, 2}});
-    auto input_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
-    auto input_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
-    auto output_low = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
-    auto output_high = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    cldnn::engine& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfzyx, {1, 16, 2, 1, 2}});
+    auto input_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
+    auto input_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 16, 1, 1 } });
+    auto output_low = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
+    auto output_high = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { -1.0f, 2.1f, 3.0f, 4.0f,
                          5.0f, 2.0f, 2.0f, 3.0f,
@@ -562,7 +556,7 @@ TEST(quantize_gpu, quantize_levels_256_3d_unsigned) {
 
     topology topology;
     topology.add(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         data("input_low", input_low),
         data("input_high", input_high),
         data("output_low", output_low),
@@ -576,14 +570,14 @@ TEST(quantize_gpu, quantize_levels_256_3d_unsigned) {
     auto outputs = network.execute();
 
     auto output = outputs.at("out").get_memory();
-    auto output_ptr = output.pointer<uint8_t>();
+    cldnn::mem_lock<uint8_t> output_ptr(output, get_test_stream());
 
     // Check that layout and memory contains logical size of tensor
-    ASSERT_EQ(output.count(), ref_data.size());
-    ASSERT_EQ(output.get_layout().count(), ref_data.size());
+    ASSERT_EQ(output->count(), ref_data.size());
+    ASSERT_EQ(output->get_layout().count(), ref_data.size());
 
     // Check that memory physical size consider binary pack
-    ASSERT_EQ(output.size(), ref_data.size() * sizeof(uint8_t));
+    ASSERT_EQ(output->size(), ref_data.size() * sizeof(uint8_t));
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_EQ(output_ptr[i], ref_data[i]) << " i=" << i;
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/reduce_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/reduce_gpu_test.cpp
index f1b906058ab249..77a3e06d40c1dc 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/reduce_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/reduce_gpu_test.cpp
@@ -2,20 +2,17 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <gtest/gtest.h>
-#include <api/input_layout.hpp>
-#include "api/reduce.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/data.hpp>
-#include "test_utils/float16.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/reduce.hpp>
+#include <cldnn/primitives/data.hpp>
+
 #include <cmath>
 #include <algorithm>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 template <typename InputT>
 struct accumulator_type {
@@ -432,7 +429,7 @@ struct output_data_type<data_types::u8> {
 template <data_types InputT, data_types OutputT>
 class ReduceTestBase : public ::testing::TestWithParam<TestParamType_general_reduce_gpu> {
 protected:
-    cldnn::engine engine = get_test_engine();
+    cldnn::engine& engine = get_test_engine();
     int batch_num, input_f, input_w, input_z, input_y, input_x;
     cldnn::format input_format = format::any;
     cldnn::reduce_mode reduce_mode;
@@ -480,10 +477,10 @@ class ReduceTestBase : public ::testing::TestWithParam<TestParamType_general_red
         auto input_size = tensor(batch(batch_num), feature(input_f), spatial(input_x, input_y, input_z, input_w));
         auto input_data = generate_random_6d<input_t>(batch_num, input_f, input_x, input_y, input_z, input_w, 1, 10);
         auto input_lay = layout(input_dt, layout_format, input_size);
-        auto input_mem = memory::allocate(engine, input_lay);
+        auto input_mem = engine.allocate_memory(input_lay);
 
         {
-            auto input_ptr = input_mem.pointer<input_t>();
+            cldnn::mem_lock<input_t> input_ptr(input_mem, get_test_stream());
             for (int fi = 0; fi < input_f; fi++)
                 for (int wi = 0; wi < input_w; wi++)
                     for (int zi = 0; zi < input_z; zi++)
@@ -505,7 +502,7 @@ class ReduceTestBase : public ::testing::TestWithParam<TestParamType_general_red
         if (force_output_dt) {
             red.output_data_type = output_dt;
         }
-        topology.add(input_layout("input", input_mem.get_layout()));
+        topology.add(input_layout("input", input_mem->get_layout()));
         topology.add(red);
         build_options options;
         options.set_option(build_option::optimize_data(true));
@@ -517,8 +514,8 @@ class ReduceTestBase : public ::testing::TestWithParam<TestParamType_general_red
         network.execute();
 
         auto out_mem = network.get_output("reduce").get_memory();
-        auto out_ptr = out_mem.pointer<output_t>();
-        auto out_lay = out_mem.get_layout();
+        cldnn::mem_lock<output_t> out_ptr(out_mem, get_test_stream());
+        auto out_lay = out_mem->get_layout();
 
         ASSERT_EQ(out_lay.size.sizes()[0], reference_result.size());                 // b
         ASSERT_EQ(out_lay.size.sizes()[1], reference_result[0].size());              // f
@@ -543,11 +540,11 @@ class ReduceTestBase : public ::testing::TestWithParam<TestParamType_general_red
                                     std::cout << "Reference value at batch: " << bi << " output_f: " << fi
                                               << " y: " << yi << " x: " << xi << " = " << val_ref << " Val = " << val
                                               << std::endl;
-                                
+
                                 EXPECT_TRUE(equal);
 
                                 if (!equal)
-                                    break;    
+                                    break;
                             }
                         }
     }
@@ -752,13 +749,13 @@ INSTANTIATE_TEST_CASE_P(DISABLED_reduce_gpu_ref_f32_f32,
                         general_reduce_gpu::PrintToStringParamName);
 
 TEST(reduce_gpu, common_bfyx) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 1, 1, 1}});
 
     set_values(input, {1.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::sum, {cldnn::reduce::along_b}, 0));
 
     network network(engine, topology);
@@ -774,7 +771,7 @@ TEST(reduce_gpu, common_bfyx) {
 
     std::vector<float> ref_data = {1.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -782,13 +779,13 @@ TEST(reduce_gpu, common_bfyx) {
 }
 
 TEST(reduce_gpu, common_bfyx_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {1, 3, 4, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {1, 3, 4, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f, 9.0f, 10.0f, 11.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::sum, {cldnn::reduce::along_x, cldnn::reduce::along_y}, 1));
 
     network network(engine, topology);
@@ -804,7 +801,7 @@ TEST(reduce_gpu, common_bfyx_keepdims) {
 
     std::vector<float> ref_data = {6.0f, 22.0f, 38.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -812,13 +809,13 @@ TEST(reduce_gpu, common_bfyx_keepdims) {
 }
 
 TEST(reduce_gpu, regr_bfyx_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, {1, 3, 2, 2} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, {1, 3, 2, 2} });
 
     set_values(input, { 0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f, 9.0f, 10.0f, 11.0f });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::sum, { cldnn::reduce::along_b, cldnn::reduce::along_x }, 1));
 
     network network(engine, topology);
@@ -834,7 +831,7 @@ TEST(reduce_gpu, regr_bfyx_keepdims) {
 
     std::vector<float> ref_data = { 1.0f, 5.0f, 9.0f, 13.0f, 17.0f, 21.0f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -842,13 +839,13 @@ TEST(reduce_gpu, regr_bfyx_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfzyx) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfzyx, {1, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfzyx, {1, 1, 1, 1, 1}});
 
     set_values(input, {1.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::sum, {cldnn::reduce::along_b}, 0));
 
     network network(engine, topology);
@@ -864,7 +861,7 @@ TEST(reduce_gpu, common_bfzyx) {
 
     std::vector<float> ref_data = {1.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -872,13 +869,13 @@ TEST(reduce_gpu, common_bfzyx) {
 }
 
 TEST(reduce_gpu, common_bfzyx_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfzyx, {1, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfzyx, {1, 1, 1, 1, 1}});
 
     set_values(input, {1.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::sum, {cldnn::reduce::along_b}, 1));
 
     network network(engine, topology);
@@ -894,7 +891,7 @@ TEST(reduce_gpu, common_bfzyx_keepdims) {
 
     std::vector<float> ref_data = {1.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -902,13 +899,13 @@ TEST(reduce_gpu, common_bfzyx_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, tensor(format::bfwzyx, {1, 3, 4, 1, 1, 1})});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, tensor(format::bfwzyx, {1, 3, 4, 1, 1, 1})});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f, 9.0f, 10.0f, 11.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::sum, {cldnn::reduce::along_w, cldnn::reduce::along_z, cldnn::reduce::along_y, cldnn::reduce::along_x}, 0));
 
     network network(engine, topology);
@@ -924,7 +921,7 @@ TEST(reduce_gpu, common_bfwzyx) {
 
     std::vector<float> ref_data = {6.0f, 22.0f, 38.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -932,13 +929,13 @@ TEST(reduce_gpu, common_bfwzyx) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, tensor(format::bfwzyx, {1, 3, 4, 1, 1, 1})});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, tensor(format::bfwzyx, {1, 3, 4, 1, 1, 1})});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f, 9.0f, 10.0f, 11.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::sum, {cldnn::reduce::along_f, cldnn::reduce::along_w, cldnn::reduce::along_z}, 1));
 
     network network(engine, topology);
@@ -954,7 +951,7 @@ TEST(reduce_gpu, common_bfwzyx_keepdims) {
 
     std::vector<float> ref_data = {66.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -962,14 +959,14 @@ TEST(reduce_gpu, common_bfwzyx_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_max_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 4, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 4, 1, 1, 1}});
 
     set_values(input, {0.0f,  1.0f,  2.0f,  3.0f,  4.0f,  5.0f,  6.0f,  7.0f,  8.0f,  9.0f,  10.0f, 11.0f,
                        12.0f, 13.0f, 14.0f, 15.0f, 16.0f, 17.0f, 18.0f, 19.0f, 20.0f, 21.0f, 22.0f, 23.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::max, {cldnn::reduce::along_b, cldnn::reduce::along_f}, 1));
 
     network network(engine, topology);
@@ -985,7 +982,7 @@ TEST(reduce_gpu, common_bfwzyx_max_keepdims) {
 
     std::vector<float> ref_data = {20.0f, 21.0f, 22.0f, 23.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -993,13 +990,13 @@ TEST(reduce_gpu, common_bfwzyx_max_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_min) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::min, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 0));
 
     network network(engine, topology);
@@ -1015,7 +1012,7 @@ TEST(reduce_gpu, common_bfwzyx_min) {
 
     std::vector<float> ref_data = {0.0f, 3.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1023,13 +1020,13 @@ TEST(reduce_gpu, common_bfwzyx_min) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_min_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::min, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 1));
 
     network network(engine, topology);
@@ -1045,7 +1042,7 @@ TEST(reduce_gpu, common_bfwzyx_min_keepdims) {
 
     std::vector<float> ref_data = {0.0f, 3.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1053,13 +1050,13 @@ TEST(reduce_gpu, common_bfwzyx_min_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_mean) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::mean, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 0));
 
     network network(engine, topology);
@@ -1075,7 +1072,7 @@ TEST(reduce_gpu, common_bfwzyx_mean) {
 
     std::vector<float> ref_data = {1.0f, 4.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1083,13 +1080,13 @@ TEST(reduce_gpu, common_bfwzyx_mean) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_mean_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::mean, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 1));
 
     network network(engine, topology);
@@ -1105,7 +1102,7 @@ TEST(reduce_gpu, common_bfwzyx_mean_keepdims) {
 
     std::vector<float> ref_data = {1.0f, 4.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1113,13 +1110,13 @@ TEST(reduce_gpu, common_bfwzyx_mean_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_prod) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::prod, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 0));
 
     network network(engine, topology);
@@ -1135,7 +1132,7 @@ TEST(reduce_gpu, common_bfwzyx_prod) {
 
     std::vector<float> ref_data = {0.0f, 60.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1143,13 +1140,13 @@ TEST(reduce_gpu, common_bfwzyx_prod) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_prod_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::prod, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 1));
 
     network network(engine, topology);
@@ -1165,7 +1162,7 @@ TEST(reduce_gpu, common_bfwzyx_prod_keepdims) {
 
     std::vector<float> ref_data = {0.0f, 60.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1173,14 +1170,14 @@ TEST(reduce_gpu, common_bfwzyx_prod_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_sum_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 4, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 4, 1, 1, 1}});
 
     set_values(input, {0.0f,  1.0f,  2.0f,  3.0f,  4.0f,  5.0f,  6.0f,  7.0f,  8.0f,  9.0f,  10.0f, 11.0f,
                        12.0f, 13.0f, 14.0f, 15.0f, 16.0f, 17.0f, 18.0f, 19.0f, 20.0f, 21.0f, 22.0f, 23.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::sum, {cldnn::reduce::along_b, cldnn::reduce::along_f}, 1));
 
     network network(engine, topology);
@@ -1196,7 +1193,7 @@ TEST(reduce_gpu, common_bfwzyx_sum_keepdims) {
 
     std::vector<float> ref_data = {60.0f, 66.0f, 72.0f, 78.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1204,13 +1201,13 @@ TEST(reduce_gpu, common_bfwzyx_sum_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_logical_and) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::logical_and, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 0));
 
     network network(engine, topology);
@@ -1226,7 +1223,7 @@ TEST(reduce_gpu, common_bfwzyx_logical_and) {
 
     std::vector<char> ref_data = {0, 1};
 
-    auto output_ptr = output.pointer<char>();
+    cldnn::mem_lock<char> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1234,13 +1231,13 @@ TEST(reduce_gpu, common_bfwzyx_logical_and) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_logical_and_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::logical_and, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 1));
 
     network network(engine, topology);
@@ -1256,7 +1253,7 @@ TEST(reduce_gpu, common_bfwzyx_logical_and_keepdims) {
 
     std::vector<char> ref_data = {0, 1};
 
-    auto output_ptr = output.pointer<char>();
+    cldnn::mem_lock<char> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1264,13 +1261,13 @@ TEST(reduce_gpu, common_bfwzyx_logical_and_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_logical_or) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::logical_or, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 0));
 
     network network(engine, topology);
@@ -1286,7 +1283,7 @@ TEST(reduce_gpu, common_bfwzyx_logical_or) {
 
     std::vector<char> ref_data = {1, 1};
 
-    auto output_ptr = output.pointer<char>();
+    cldnn::mem_lock<char> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1294,13 +1291,13 @@ TEST(reduce_gpu, common_bfwzyx_logical_or) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_logical_or_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::logical_or, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 1));
 
     network network(engine, topology);
@@ -1316,7 +1313,7 @@ TEST(reduce_gpu, common_bfwzyx_logical_or_keepdims) {
 
     std::vector<char> ref_data = {1, 1};
 
-    auto output_ptr = output.pointer<char>();
+    cldnn::mem_lock<char> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1324,13 +1321,13 @@ TEST(reduce_gpu, common_bfwzyx_logical_or_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_sum_square) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::sum_square, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 0));
 
     network network(engine, topology);
@@ -1346,7 +1343,7 @@ TEST(reduce_gpu, common_bfwzyx_sum_square) {
 
     std::vector<float> ref_data = {5.0f, 50.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1354,13 +1351,13 @@ TEST(reduce_gpu, common_bfwzyx_sum_square) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_sum_square_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::sum_square, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 1));
 
     network network(engine, topology);
@@ -1376,7 +1373,7 @@ TEST(reduce_gpu, common_bfwzyx_sum_square_keepdims) {
 
     std::vector<float> ref_data = {5.0f, 50.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1384,13 +1381,13 @@ TEST(reduce_gpu, common_bfwzyx_sum_square_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_l1) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, -2.0f, 3.0f, 4.0f, -5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::l1, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 0));
 
     network network(engine, topology);
@@ -1406,7 +1403,7 @@ TEST(reduce_gpu, common_bfwzyx_l1) {
 
     std::vector<float> ref_data = {3.0f, 12.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1414,13 +1411,13 @@ TEST(reduce_gpu, common_bfwzyx_l1) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_l1_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, -2.0f, 3.0f, 4.0f, -5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::l1, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 1));
 
     network network(engine, topology);
@@ -1436,7 +1433,7 @@ TEST(reduce_gpu, common_bfwzyx_l1_keepdims) {
 
     std::vector<float> ref_data = {3.0f, 12.0f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1444,13 +1441,13 @@ TEST(reduce_gpu, common_bfwzyx_l1_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_l2) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, -2.0f, 3.0f, 4.0f, -5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::l2, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 0));
 
     network network(engine, topology);
@@ -1466,7 +1463,7 @@ TEST(reduce_gpu, common_bfwzyx_l2) {
 
     std::vector<float> ref_data = {2.236067977f, 7.071067812f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1474,13 +1471,13 @@ TEST(reduce_gpu, common_bfwzyx_l2) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_l2_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, -2.0f, 3.0f, 4.0f, -5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::l2, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 1));
 
     network network(engine, topology);
@@ -1496,7 +1493,7 @@ TEST(reduce_gpu, common_bfwzyx_l2_keepdims) {
 
     std::vector<float> ref_data = {2.236067977f, 7.071067812f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1504,13 +1501,13 @@ TEST(reduce_gpu, common_bfwzyx_l2_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_log_sum) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::log_sum, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 0));
 
     network network(engine, topology);
@@ -1526,7 +1523,7 @@ TEST(reduce_gpu, common_bfwzyx_log_sum) {
 
     std::vector<float> ref_data = {1.0986122887f, 2.4849066498f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1534,13 +1531,13 @@ TEST(reduce_gpu, common_bfwzyx_log_sum) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_log_sum_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::log_sum, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 1));
 
     network network(engine, topology);
@@ -1556,7 +1553,7 @@ TEST(reduce_gpu, common_bfwzyx_log_sum_keepdims) {
 
     std::vector<float> ref_data = {1.0986122887f, 2.4849066498f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1564,13 +1561,13 @@ TEST(reduce_gpu, common_bfwzyx_log_sum_keepdims) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_log_sum_exp) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::log_sum_exp, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 0));
 
     network network(engine, topology);
@@ -1586,7 +1583,7 @@ TEST(reduce_gpu, common_bfwzyx_log_sum_exp) {
 
     std::vector<float> ref_data = {2.407605964f, 5.407605964f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1594,13 +1591,13 @@ TEST(reduce_gpu, common_bfwzyx_log_sum_exp) {
 }
 
 TEST(reduce_gpu, common_bfwzyx_log_sum_exp_keepdims) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfwzyx, {2, 3, 1, 1, 1, 1}});
 
     set_values(input, {0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reduce("reduce", "input", reduce_mode::log_sum_exp, {cldnn::reduce::along_f, cldnn::reduce::along_w}, 1));
 
     network network(engine, topology);
@@ -1616,7 +1613,7 @@ TEST(reduce_gpu, common_bfwzyx_log_sum_exp_keepdims) {
 
     std::vector<float> ref_data = {2.407605964f, 5.407605964f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < ref_data.size(); ++i) {
         EXPECT_TRUE(are_equal(ref_data[i], output_ptr[i]));
@@ -1626,7 +1623,7 @@ TEST(reduce_gpu, common_bfwzyx_log_sum_exp_keepdims) {
 template <data_types InputT, data_types OutputT>
 class ReduceXYWithBigTensorTestBase : public ::testing::TestWithParam<TestParamType_general_reduce_gpu> {
 protected:
-    cldnn::engine engine = get_test_engine();
+    cldnn::engine& engine = get_test_engine();
     int batch_num, input_f, input_w, input_z, input_y, input_x;
     cldnn::format input_format = format::any;
     cldnn::reduce_mode reduce_mode;
@@ -1675,10 +1672,11 @@ class ReduceXYWithBigTensorTestBase : public ::testing::TestWithParam<TestParamT
         auto input_size = tensor(batch(batch_num), feature(input_f), spatial(input_x, input_y, input_z, input_w));
         auto input_data = generate_random_6d<input_t>(batch_num, input_f, input_x, input_y, input_z, input_w, 1, 5, 9);
         auto input_lay = layout(input_dt, layout_format, input_size);
-        auto input_mem = memory::allocate(engine, input_lay);
+        auto input_mem = engine.allocate_memory(input_lay);
 
         {
-            auto input_ptr = input_mem.pointer<input_t>();
+            cldnn::mem_lock<input_t> input_ptr(input_mem, get_test_stream());
+
             for (int fi = 0; fi < input_f; fi++)
                 for (int wi = 0; wi < input_w; wi++)
                     for (int zi = 0; zi < input_z; zi++)
@@ -1718,7 +1716,7 @@ class ReduceXYWithBigTensorTestBase : public ::testing::TestWithParam<TestParamT
             if (force_output_dt) {
                 red.output_data_type = output_dt;
             }
-            topology.add(input_layout("input", input_mem.get_layout()));
+            topology.add(input_layout("input", input_mem->get_layout()));
             topology.add(red);
             build_options options;
             options.set_option(build_option::optimize_data(true));
@@ -1730,8 +1728,8 @@ class ReduceXYWithBigTensorTestBase : public ::testing::TestWithParam<TestParamT
             network.execute();
 
             auto out_mem = network.get_output("reduce").get_memory();
-            auto out_ptr = out_mem.pointer<output_t>();
-            auto out_lay = out_mem.get_layout();
+            cldnn::mem_lock<output_t> out_ptr(out_mem, get_test_stream());
+            auto out_lay = out_mem->get_layout();
 
             ASSERT_EQ(out_lay.size.sizes()[0], reference_result.size());                 // b
             ASSERT_EQ(out_lay.size.sizes()[1], reference_result[0].size());              // f
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/region_yolo_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/region_yolo_gpu_test.cpp
index 62d02ef2abd54a..e1d5d9b2a71873 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/region_yolo_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/region_yolo_gpu_test.cpp
@@ -2,18 +2,16 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
+#include "test_utils.h"
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/region_yolo.hpp>
-#include <api/reorder.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/region_yolo.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/graph/topology.hpp>
+#include <cldnn/graph/network.hpp>
+#include <cldnn/runtime/memory.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -186,26 +184,26 @@ namespace internal
 template <typename T>
 static void runRegionTest(internal::region_yolo_test_params& params)
 {
-    engine eng;
+    auto& engine = get_test_engine();
     const tensor kInputTensor(params.tensor[0], params.tensor[1], params.tensor[2], params.tensor[3]);
     auto inputData = generate_random_1d<T>(params.tensor[0] * params.tensor[1] * params.tensor[2] * params.tensor[3], -1, 1);
 
-    auto inputPrim = memory::allocate(eng, { params.dataType, format::bfyx, kInputTensor });
+    auto inputPrim = engine.allocate_memory({ params.dataType, format::bfyx, kInputTensor });
     set_values(inputPrim, inputData);
 
     topology topology;
-    topology.add(input_layout("InputData", inputPrim.get_layout()));
+    topology.add(input_layout("InputData", inputPrim->get_layout()));
     topology.add(reorder("reorder_pre", "InputData", params.fmt, params.dataType));
     topology.add(region_yolo("region_yolo", "reorder_pre", params.coords, params.classes,
             params.regionNum, static_cast<uint32_t>(params.mask.size()), params.softMax));
     topology.add(reorder("reorder_post", "region_yolo", format::bfyx, params.dataType));
 
-    network network(eng, topology);
+    network network(engine, topology);
     network.set_input_data("InputData", inputPrim);
 
     auto outputs = network.execute();
     auto output = outputs.at("reorder_post").get_memory();
-    auto outputData = output.pointer<T>();
+    cldnn::mem_lock<T> outputData(output, get_test_stream());
 
     /// reference value
     std::vector<T> refOutputData(inputData.size());
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/removing_output_node_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/removing_output_node_test.cpp
index 65e1c9f0ec0410..f8151686b50dc3 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/removing_output_node_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/removing_output_node_test.cpp
@@ -2,22 +2,16 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
+#include "test_utils.h"
 
-#include <api/data.hpp>
-#include <api/reshape.hpp>
-#include <api/input_layout.hpp>
-#include <api/shuffle_channels.hpp>
-#include <api/strided_slice.hpp>
-
-#include "test_utils/test_utils.h"
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/reshape.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/shuffle_channels.hpp>
+#include <cldnn/primitives/strided_slice.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 using namespace testing;
 
 TEST(removing_output_node, multiple_outputs) {
@@ -28,7 +22,7 @@ TEST(removing_output_node, multiple_outputs) {
     //                                                  |_
     //                                                     reshape(bfyx);
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto batch_num = 6;
     auto feature_num = 1;
     auto x_size = 1;
@@ -40,10 +34,10 @@ TEST(removing_output_node, multiple_outputs) {
     tensor after_strided_slice = tensor(spatial(y_size, feature_num), feature(batch_num), batch(1));
     tensor after_reshape = tensor(feature(batch_num * feature_num * y_size * x_size));
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, initial_shape });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, initial_shape });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(begin, {
             1, 0, 1, 0
@@ -56,7 +50,7 @@ TEST(removing_output_node, multiple_outputs) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(shuffle_channels("shuffle_channels", "input", group, axis));
     topology.add(reshape("reshape", "shuffle_channels", after_reshape));
     topology.add(data("input2", begin));
@@ -76,9 +70,9 @@ TEST(removing_output_node, multiple_outputs) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reshape").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
-    ASSERT_TRUE(output.get_layout().size == after_reshape);
+    ASSERT_TRUE(output->get_layout().size == after_reshape);
 
     for (size_t i = 0; i < out_vec.size(); i++)
         EXPECT_EQ(output_ptr[i], out_vec[i]);
@@ -86,9 +80,9 @@ TEST(removing_output_node, multiple_outputs) {
     // checking the output node has the same name after output node deleting due to StridedSlice optimization
     ASSERT_TRUE(outputs.find("strided_slice") != outputs.end());
     auto output2 = outputs.at("strided_slice").get_memory();
-    auto output_ptr2 = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr2(output, get_test_stream());
 
-    ASSERT_TRUE(output2.get_layout().size == after_strided_slice);
+    ASSERT_TRUE(output2->get_layout().size == after_strided_slice);
 
     for (size_t i = 0; i < out_vec.size(); i++)
         EXPECT_EQ(output_ptr2[i], out_vec[i]);
@@ -114,10 +108,10 @@ TEST(removing_output_node, output_node_optimization) {
     //  21  28  39
     //  18  20  20
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32,format::yxfb,{ 1, 1, 5, 4 } });
-    auto weights = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 1, 3, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32,format::yxfb,{ 1, 1, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 3, 2 } });
 
     set_values(input, { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 2.0f, 2.0f, 3.0f, 4.0f, 6.0f, 3.0f, 3.0f, 3.0f, 5.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f });
     set_values(weights, { 1.0f, 2.0f, 1.0f, 2.0f, 1.0f, 2.0f });
@@ -126,7 +120,7 @@ TEST(removing_output_node, output_node_optimization) {
             { 17.0f, 19.0f, 19.0f } };
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights", weights));
     topology.add(convolution("conv", "input", { "weights" }, { 1,1,1,2 }));
     topology.add(activation("relu", "conv", activation_func::relu));
@@ -140,8 +134,8 @@ TEST(removing_output_node, output_node_optimization) {
     EXPECT_EQ(outputs.begin()->first, "relu");
 
     auto output_memory = outputs.at("relu").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/reorder_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/reorder_gpu_test.cpp
index f2c80c3b06130c..6f0b6a3dfc1eac 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/reorder_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/reorder_gpu_test.cpp
@@ -3,31 +3,27 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/reorder.hpp"
-#include "api/crop.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include <api/reshape.hpp>
-#include "test_utils/test_utils.h"
-#include <api/data.hpp>
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/reshape.hpp>
+#include "cldnn/primitives/reorder.hpp"
+#include "cldnn/primitives/crop.hpp"
+#include <cldnn/primitives/data.hpp>
 
 #include <cmath>
-#include <gmock/gmock.h>
 #include <limits>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 using namespace testing;
 
 static void compare_bfyx2blocked_with_ref(const std::string& kernel_name,
     const data_types input_data_type, const data_types output_data_type,
     cldnn::format input_format, cldnn::format output_format,
     int32_t b_in, int32_t f_in, int32_t x_in, int32_t y_in, int32_t z_in = 0, int32_t w_in = 0) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     tensor ts;
     if (input_format.dimension() == 4) {
@@ -40,11 +36,11 @@ static void compare_bfyx2blocked_with_ref(const std::string& kernel_name,
         ts = { b_in, f_in, x_in, y_in, z_in, w_in };
     }
 
-    auto input = memory::allocate(engine, { input_data_type, input_format, ts });
+    auto input = engine.allocate_memory({ input_data_type, input_format, ts });
     layout output_layout(output_data_type, output_format, ts);
 
     if (input_data_type == data_types::i8) {
-        auto input_ptr = input.pointer<unsigned char>();
+        mem_lock<uint8_t> input_ptr{input, get_test_stream()};
         unsigned char i = 1;
         for (auto it = input_ptr.begin(); it != input_ptr.end(); ++it)
         {
@@ -54,7 +50,7 @@ static void compare_bfyx2blocked_with_ref(const std::string& kernel_name,
             }
         }
     } else {
-        auto input_ptr = input.pointer<float>();
+        mem_lock<float> input_ptr{input, get_test_stream()};
         float i = 1.f;
         for (auto it = input_ptr.begin(); it != input_ptr.end(); ++it)
         {
@@ -64,7 +60,7 @@ static void compare_bfyx2blocked_with_ref(const std::string& kernel_name,
     }
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", output_layout));
 
     // run on reference(reorder_data) kernel
@@ -78,11 +74,11 @@ static void compare_bfyx2blocked_with_ref(const std::string& kernel_name,
     std::map<cldnn::primitive_id, cldnn::network_output> outputs_ref;
 
     outputs_ref = network_ref.execute();
-    cldnn::event e1 = outputs_ref.at("reorder").get_event();
-    e1.wait();
+    cldnn::event::ptr e1 = outputs_ref.at("reorder").get_event();
+    e1->wait();
 
     auto output_ref = outputs_ref.begin()->second.get_memory();
-    auto output_ref_ptr = output_ref.pointer<unsigned char>();
+    mem_lock<uint8_t> output_ref_ptr{output_ref, get_test_stream()};
 
     // run on optimized kernel
     cldnn::build_options options;
@@ -95,11 +91,11 @@ static void compare_bfyx2blocked_with_ref(const std::string& kernel_name,
     std::map<cldnn::primitive_id, cldnn::network_output> outputs;
 
     outputs = network.execute();
-    cldnn::event e2 = outputs.at("reorder").get_event();
-    e2.wait();
+    cldnn::event::ptr e2 = outputs.at("reorder").get_event();
+    e2->wait();
 
     auto output = outputs.begin()->second.get_memory();
-    auto output_ptr = output.pointer<unsigned char>();
+    mem_lock<uint8_t> output_ptr{output, get_test_stream()};
 
     // compare results
     const size_t output_size = output_ref_ptr.size();
@@ -206,17 +202,17 @@ TEST(reorder_gpu_optimization, compare_with_ref__bfyx_to_blocked_format_differen
 }
 
 TEST(reorder_gpu_optimization, bfyx_to_fsv16_without_f_remainder) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int32_t b_in = 1;
     const int32_t f_in = 8 * 4;
     const int32_t y_in = 4;
     const int32_t x_in = 8 * 2;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { b_in,f_in,x_in,y_in } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { b_in,f_in,x_in,y_in } });
     layout output_layout(data_types::f32, format::b_fs_yx_fsv16, { b_in,f_in,x_in,y_in });
 
     // Set incremental input value
-    auto input_ptr = input.pointer<float>();
+    mem_lock<float> input_ptr{input, get_test_stream()};
     float i = 0.f;
     for (auto it = input_ptr.begin(); it != input_ptr.end(); ++it)
     {
@@ -224,7 +220,7 @@ TEST(reorder_gpu_optimization, bfyx_to_fsv16_without_f_remainder) {
     }
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", output_layout));
 
     network network(engine, topology);
@@ -235,7 +231,7 @@ TEST(reorder_gpu_optimization, bfyx_to_fsv16_without_f_remainder) {
     EXPECT_EQ(outputs.begin()->first, "reorder");
 
     auto output = outputs.begin()->second.get_memory();
-    auto output_ptr = output.pointer<float>();
+    mem_lock<float> output_ptr{output, get_test_stream()};
 
     auto get_fsv16_index = [](int32_t /* b_size */, int32_t f_size, int32_t y_size, int32_t x_size,
         int32_t b, int32_t f, int32_t y, int32_t x) {
@@ -291,9 +287,9 @@ TEST(reorder_gpu_f32, basic) {
     //  b1 f1: 12    8
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
     layout output_layout(data_types::f32, format::bfyx,{ 2,2,2,2 });
 
     set_values(input, {
@@ -311,7 +307,7 @@ TEST(reorder_gpu_f32, basic) {
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", output_layout));
 
     network network(engine, topology);
@@ -337,7 +333,7 @@ TEST(reorder_gpu_f32, basic) {
         12.0f, 8.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -376,11 +372,11 @@ TEST(reorder_gpu_f32, basic_subtract) {
     //  b1 f1: 10    7
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32,  format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32,  format::yxfb, { 2, 2, 2, 2 } });
     layout output_layout( data_types::f32, format::bfyx, {2,2,2,2} );
-    auto subtract = memory::allocate(engine, { data_types::f32, format::byxf, { 1, 2, 2, 2 } });
+    auto subtract = engine.allocate_memory({ data_types::f32, format::byxf, { 1, 2, 2, 2 } });
 
     set_values(input, {
         1.f, 0.f,
@@ -402,8 +398,8 @@ TEST(reorder_gpu_f32, basic_subtract) {
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
-        input_layout("subtract", subtract.get_layout()),
+        input_layout("input", input->get_layout()),
+        input_layout("subtract", subtract->get_layout()),
         reorder("reorder", "input", output_layout, "subtract"));
 
     network network(engine, topology);
@@ -429,7 +425,7 @@ TEST(reorder_gpu_f32, basic_subtract) {
                          10.0f,  7.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -465,9 +461,9 @@ TEST(reorder_gpu_f32, basic_subtract_value) {
     //  b1 f1:  9.5  5.5
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
     layout output_layout(data_types::f32, format::bfyx,{ 2,2,2,2 });
     std::vector<float> subtract_val = { 0.5, 2.5 };
 
@@ -486,7 +482,7 @@ TEST(reorder_gpu_f32, basic_subtract_value) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()), reorder("reorder", "input", output_layout, subtract_val));
+    topology.add(input_layout("input", input->get_layout()), reorder("reorder", "input", output_layout, subtract_val));
 
     network network(engine, topology);
     network.set_input_data("input", input);
@@ -510,7 +506,7 @@ TEST(reorder_gpu_f32, basic_subtract_value) {
                           9.5f,  5.5f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_TRUE(are_equal(answers[i], output_ptr[i]));
@@ -549,18 +545,18 @@ TEST(reorder_gpu_f16, basic_subtract_f32_output_f32) {
     //  b1 f1: 10    7
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
+    if (!engine.get_device_info().supports_fp16)
     {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
     }
 
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb, { 2, 2, 2, 2 } });
     layout output_layout(data_types::f32, format::bfyx,{ 2,2,2,2 });
-    auto subtract = memory::allocate(engine, { data_types::f32, format::byxf, { 1, 2, 2, 2 } });
+    auto subtract = engine.allocate_memory({ data_types::f32, format::byxf, { 1, 2, 2, 2 } });
 
     set_values(input, {
         half_t(1.f), half_t(0.f),
@@ -582,7 +578,7 @@ TEST(reorder_gpu_f16, basic_subtract_f32_output_f32) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("subtract", subtract));
     topology.add(reorder("reorder", "input", output_layout, "subtract"));
 
@@ -608,7 +604,7 @@ TEST(reorder_gpu_f16, basic_subtract_f32_output_f32) {
                          10.0f,  7.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_TRUE(are_equal(answers[i], output_ptr[i]));
@@ -644,15 +640,15 @@ TEST(reorder_gpu_f16, basic_subtract_value) {
     //  b1 f1:  9.5  5.5
     //
 
-    const auto& engine = get_test_engine();
-    if (!engine.get_info().supports_fp16)
+    auto& engine = get_test_engine();
+    if (!engine.get_device_info().supports_fp16)
     {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
         return;
     }
 
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb, { 2, 2, 2, 2 } });
     layout output_layout(data_types::f16, format::bfyx,{ 2,2,2,2 });
     std::vector<float> subtract_val = { 0.5, 2.5 };
 
@@ -671,7 +667,7 @@ TEST(reorder_gpu_f16, basic_subtract_value) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reorder("reorder", "input", output_layout, subtract_val));
 
     network network(engine, topology);
@@ -696,7 +692,7 @@ TEST(reorder_gpu_f16, basic_subtract_value) {
                            half_t(9.5f), half_t(5.5f)
     };
 
-    auto output_ptr = output.pointer<half_t>();
+    cldnn::mem_lock<half_t> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_TRUE(are_equal(static_cast<uint16_t>(answers[i]), static_cast<uint16_t>(output_ptr[i])));
@@ -713,9 +709,9 @@ TEST(reorder_gpu, basic_convert_f16_f32_f16) {
     //  Output is expected to contain the same value as input in range of indices from 0x0000 to 0xF801.
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    if (!engine.get_info().supports_fp16)
+    if (!engine.get_device_info().supports_fp16)
     {
         std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
         EXPECT_EQ(1, 1);
@@ -736,14 +732,14 @@ TEST(reorder_gpu, basic_convert_f16_f32_f16) {
     expected_values[0xF802] = half_t(0x8000, 0);    // -0
     expected_values[0xF803] = half_t(0xFC12, 0);    // A NaN (sample: -NaN.0x12).
 
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb, { 1, static_cast<int32_t>(expected_values.size()) / 4, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb, { 1, static_cast<int32_t>(expected_values.size()) / 4, 2, 2 } });
     layout interm_layout( data_types::f32, format::byxf, { 1, static_cast<int32_t>(expected_values.size()) / 4, 2, 2 });
-    auto output_layout = input.get_layout();
+    auto output_layout = input->get_layout();
 
     set_values(input, expected_values);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reorder("reorder_f16_f32", "input", interm_layout));
     topology.add(reorder("reorder_f32_f16", "reorder_f16_f32", output_layout));
 
@@ -762,7 +758,7 @@ TEST(reorder_gpu, basic_convert_f16_f32_f16) {
     EXPECT_TRUE(outputs.find("reorder_f32_f16") != outputs.end());
 
     auto interm = outputs.at("reorder_f16_f32").get_memory();
-    auto interm_ptr = interm.pointer<float>();
+    cldnn::mem_lock<float> interm_ptr(interm, get_test_stream());
 
     // Sample positive.
     EXPECT_TRUE(are_equal(interm_ptr[0x3400], 0.25f));
@@ -783,7 +779,7 @@ TEST(reorder_gpu, basic_convert_f16_f32_f16) {
     EXPECT_TRUE(std::isnan(interm_ptr[0xF803]));
 
     auto output = outputs.at("reorder_f32_f16").get_memory();
-    auto output_ptr = output.pointer<half_t>();
+    cldnn::mem_lock<half_t> output_ptr(output, get_test_stream());
     for (int i = 0; i < 0xF802; ++i) // NOTE: do not test for possibly ambiguous values of floating point (-0, NaNs).
     {
         EXPECT_TRUE(are_equal(static_cast<uint16_t>(expected_values[i]), static_cast<uint16_t>(output_ptr[i])));
@@ -792,14 +788,14 @@ TEST(reorder_gpu, basic_convert_f16_f32_f16) {
 
 TEST(reorder_gpu, basic_convert_int8) {
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     layout in_layout = { type_to_data_type<float>::value,format::byxf,{ 1,1,3,3 } };
     layout byte_layout = { type_to_data_type<int8_t>::value, format::bfyx,{ 1,1,3,3 } };
     std::initializer_list<float> input_f = { 1.0f, -2.5f, 3.1f, -4.0f, 5.03f, -6.99f, 7.0f, -8.0f, 9.0f };
     std::list<float> final_results = { 1.0f, -3.0f, 3.0f, -4.0f, 5.0f, -7.0f, 7.0f, -8.0f, 9.0f };
 
     // Allocate memory for input image.
-    auto input_memory = memory::allocate(engine, in_layout);
+    auto input_memory = engine.allocate_memory(in_layout);
     set_values(input_memory, input_f);
 
     // Create input_layout description
@@ -830,7 +826,7 @@ TEST(reorder_gpu, basic_convert_int8) {
     auto outputs = network.execute();
 
     auto interm = outputs.at("reorder2").get_memory();
-    auto interm_ptr = interm.pointer<float>();
+    cldnn::mem_lock<float> interm_ptr(interm, get_test_stream());
     unsigned int cntr = 0;
     for (const auto& exp : final_results)
     {
@@ -849,9 +845,9 @@ TEST(reorder_gpu, basic_convert_uint8rgbabyxf_to_fp32_bfyx) {
 	//
 	const int kernel_size = 5;
 	const int feature_size = 4;
-	const auto& engine = get_test_engine();
+	auto& engine = get_test_engine();
 
-	if (!engine.get_info().supports_fp16)
+	if (!engine.get_device_info().supports_fp16)
 	{
 		std::cout << "[ SKIPPED ] The test is skipped (cl_khr_fp16 is not supported)." << std::endl;
 		EXPECT_EQ(1, 1);
@@ -869,7 +865,7 @@ TEST(reorder_gpu, basic_convert_uint8rgbabyxf_to_fp32_bfyx) {
 	layout output_layout = { type_to_data_type<float>::value, format::bfyx, {1,4,kernel_size,kernel_size } };
 
 	// Allocate memory for input image.
-	auto input_memory = memory::allocate(engine, in_layout);
+	auto input_memory = engine.allocate_memory(in_layout);
 	set_values(input_memory, input_i8);
 
     // Create input_layout description
@@ -916,8 +912,8 @@ TEST(reorder_gpu, basic_convert_uint8rgbabyxf_to_fp32_bfyx) {
     EXPECT_TRUE(outputs.find("crop") != outputs.end());
 
     auto interm = outputs.at("reorder_input").get_memory();
-    auto interm_ptr = interm.pointer<float>();
-    auto interm_size = outputs.at("reorder_input").get_memory().count();
+    cldnn::mem_lock<float> interm_ptr(interm, get_test_stream());
+    auto interm_size = outputs.at("reorder_input").get_memory()->count();
     EXPECT_EQ(interm_size,(size_t) (1*feature_size*kernel_size*kernel_size));
 
     // Sample positive.
@@ -939,8 +935,8 @@ TEST(reorder_gpu, basic_convert_uint8rgbabyxf_to_fp32_bfyx) {
     }
 
     auto output = outputs.at("crop").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_size = outputs.at("crop").get_memory().count();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    auto output_size = output->count();
     EXPECT_EQ(output_size,(size_t) (1 * (feature_size-1)*kernel_size*kernel_size));
 
     for (target_index = 0; target_index < output_size; target_index++)
@@ -980,9 +976,9 @@ TEST(reorder_gpu_f32, basic_yxfb_to_bfyx_input_padding)
     //  f1: b0:  5    6  b1:   1.5  5.2
     //  f1: b0:  7    8  b1:   12   8
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
     layout output_layout(data_types::f32, format::bfyx, { 2,2,2,2 });
 
     set_values(input, {
@@ -1000,8 +996,8 @@ TEST(reorder_gpu_f32, basic_yxfb_to_bfyx_input_padding)
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
-        reorder("reorder", "input", input.get_layout().format, input.get_layout().data_type, "", reorder_mean_mode::subtract, padding{ { 0, 0, 1, 2 }, 0 }),
+        input_layout("input", input->get_layout()),
+        reorder("reorder", "input", input->get_layout().format, input->get_layout().data_type, "", reorder_mean_mode::subtract, padding{ { 0, 0, 1, 2 }, 0 }),
         reorder("reorder2", "reorder", output_layout));
 
     network network(engine, topology);
@@ -1026,7 +1022,7 @@ TEST(reorder_gpu_f32, basic_yxfb_to_bfyx_input_padding)
         1.5f,  5.2f,
         12.0f, 8.0f
     };
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1059,9 +1055,9 @@ TEST(reorder_gpu_f32, basic_bfyx_to_yxfb_input_padding)
     //  b1 f1: 12    8
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
     layout output_layout(data_types::f32, format::yxfb, { 2,2,2,2 });
 
     set_values(input, {
@@ -1079,8 +1075,8 @@ TEST(reorder_gpu_f32, basic_bfyx_to_yxfb_input_padding)
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
-        reorder("reorder", "input", input.get_layout().format, input.get_layout().data_type, "", reorder_mean_mode::subtract, padding{ { 0, 0, 2, 1 }, 0 }),
+        input_layout("input", input->get_layout()),
+        reorder("reorder", "input", input->get_layout().format, input->get_layout().data_type, "", reorder_mean_mode::subtract, padding{ { 0, 0, 2, 1 }, 0 }),
         reorder("reorder2", "reorder", output_layout));
 
     network network(engine, topology);
@@ -1106,7 +1102,7 @@ TEST(reorder_gpu_f32, basic_bfyx_to_yxfb_input_padding)
         8.f, 8.f
     };
     std::vector<float> out;
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         out.push_back(output_ptr[i]);
@@ -1120,9 +1116,9 @@ TEST(reorder_gpu_f32, basic_bfyx_to_bfzyx)
     //  Input               : bfyx:2x2x2x2
     //  Output              : bfzyx:2x2x1X2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
 
     set_values(input, {
         1.f, 0.f,
@@ -1139,7 +1135,7 @@ TEST(reorder_gpu_f32, basic_bfyx_to_bfzyx)
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", format::bfzyx, data_types::f32));
 
     network network(engine, topology);
@@ -1150,8 +1146,8 @@ TEST(reorder_gpu_f32, basic_bfyx_to_bfzyx)
     EXPECT_EQ(outputs.begin()->first, "reorder");
 
     auto output = outputs.begin()->second.get_memory();
-    EXPECT_TRUE(output.get_layout().format == format::bfzyx);
-    auto sizes = output.get_layout().size;
+    EXPECT_TRUE(output->get_layout().format == format::bfzyx);
+    auto sizes = output->get_layout().size;
     EXPECT_TRUE(sizes.batch[0] == 2);
     EXPECT_TRUE(sizes.feature[0] == 2);
     EXPECT_TRUE(sizes.spatial[0] == 2);
@@ -1172,7 +1168,7 @@ TEST(reorder_gpu_f32, basic_bfyx_to_bfzyx)
         8.f, 8.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1184,9 +1180,9 @@ TEST(reorder_gpu_f32, basic_yxfb_to_bfzyx)
     //  Input               : yxfb:2x2x2x2
     //  Output              : bfzyx:2x2x1X2x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
     set_values(input, {
         1.f, 0.f,
@@ -1203,7 +1199,7 @@ TEST(reorder_gpu_f32, basic_yxfb_to_bfzyx)
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", format::bfzyx, data_types::f32));
 
     network network(engine, topology);
@@ -1214,8 +1210,8 @@ TEST(reorder_gpu_f32, basic_yxfb_to_bfzyx)
     EXPECT_EQ(outputs.begin()->first, "reorder");
 
     auto output = outputs.begin()->second.get_memory();
-    EXPECT_TRUE(output.get_layout().format == format::bfzyx);
-    auto sizes = output.get_layout().size;
+    EXPECT_TRUE(output->get_layout().format == format::bfzyx);
+    auto sizes = output->get_layout().size;
     EXPECT_TRUE(sizes.batch[0] == 2);
     EXPECT_TRUE(sizes.feature[0] == 2);
     EXPECT_TRUE(sizes.spatial[0] == 2);
@@ -1236,7 +1232,7 @@ TEST(reorder_gpu_f32, basic_yxfb_to_bfzyx)
         12.0f, 8.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1248,9 +1244,9 @@ TEST(reorder_gpu_f32, basic_bfzyx_to_bfyx)
     //  Input               : bfzyx:2x2x2x2x2
     //  Output              : bfyx:2x2x4x2
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 2, 2, 2, 2, 2 } });
 
     set_values(input, {
         1.f, 0.f,
@@ -1279,7 +1275,7 @@ TEST(reorder_gpu_f32, basic_bfzyx_to_bfyx)
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", format::bfyx, data_types::f32));
 
     network network(engine, topology);
@@ -1290,8 +1286,8 @@ TEST(reorder_gpu_f32, basic_bfzyx_to_bfyx)
     EXPECT_EQ(outputs.begin()->first, "reorder");
 
     auto output = outputs.begin()->second.get_memory();
-    EXPECT_TRUE(output.get_layout().format == format::bfyx);
-    auto sizes = output.get_layout().size;
+    EXPECT_TRUE(output->get_layout().format == format::bfyx);
+    auto sizes = output->get_layout().size;
     EXPECT_TRUE(sizes.batch[0] == 2);
     EXPECT_TRUE(sizes.feature[0] == 2);
     EXPECT_TRUE(sizes.spatial[0] == 2);
@@ -1324,7 +1320,7 @@ TEST(reorder_gpu_f32, basic_bfzyx_to_bfyx)
         8.f, 8.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]);
@@ -1333,11 +1329,11 @@ TEST(reorder_gpu_f32, basic_bfzyx_to_bfyx)
 
 TEST(reorder_gpu_opt, basic_remove_redundant)
 {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory in = memory::allocate(eng, { data_types::f32, format::bfyx, tensor{ 1, 2, 2, 1 } });
+    memory::ptr in = engine.allocate_memory({ data_types::f32, format::bfyx, tensor{ 1, 2, 2, 1 } });
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         reorder("r1", "in", format::bfyx, data_types::f32),
         reorder("r2", "r1", format::yxfb, data_types::f32)
     };
@@ -1345,26 +1341,26 @@ TEST(reorder_gpu_opt, basic_remove_redundant)
     build_options opts;
     opts.set_option(build_option::optimize_data(true));
 
-    network net(eng, tpl, opts);
+    network net(engine, tpl, opts);
     net.set_input_data("in", in);
     auto outputs = net.execute();
     auto executed_primitives = net.get_executed_primitives();
 
     EXPECT_TRUE(executed_primitives.count("r1") == 0);
     ASSERT_TRUE(outputs.count("r2") == 1);
-    EXPECT_TRUE(outputs.at("r2").get_memory().get_layout().format == format::yxfb);
+    EXPECT_TRUE(outputs.at("r2").get_memory()->get_layout().format == format::yxfb);
 }
 
 TEST(reorder_gpu_opt, remove_redundant_activation_fuse)
 {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory in = memory::allocate(eng, { data_types::f32, format::bfyx, tensor{ 1, 1, 2, 1 } });
+    memory::ptr in = engine.allocate_memory({ data_types::f32, format::bfyx, tensor{ 1, 1, 2, 1 } });
     set_values(in, { -1.0f, -1.0f });
-    memory scale_mem = memory::allocate(eng, { data_types::f32, format::bfyx, tensor{1, 1, 1, 1 } });
+    memory::ptr scale_mem = engine.allocate_memory({ data_types::f32, format::bfyx, tensor{1, 1, 1, 1 } });
     set_values(scale_mem, { 2.0f });
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         reorder("r1", "in", format::bfyx, data_types::f32),
         activation("relu", "r1", activation_func::relu_negative_slope, {0.01f, 0.0f}),
         data("scale_data", scale_mem),
@@ -1374,22 +1370,22 @@ TEST(reorder_gpu_opt, remove_redundant_activation_fuse)
     build_options opts;
     opts.set_option(build_option::optimize_data(true));
 
-    network net(eng, tpl, opts);
+    network net(engine, tpl, opts);
     net.set_input_data("in", in);
     auto outputs = net.execute();
-    auto out_ptr = outputs.begin()->second.get_memory().pointer<float>();
+    cldnn::mem_lock<float> out_ptr(outputs.begin()->second.get_memory(), get_test_stream());
     EXPECT_FLOAT_EQ(out_ptr[0], -0.02f);
     EXPECT_FLOAT_EQ(out_ptr[1], -0.02f);
 }
 
 TEST(reorder_gpu_opt, basic_remove_redundant_output_due_to_implicit_reorders)
 {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory in = memory::allocate(eng, { data_types::f32, format::yxfb, tensor{ 1, 2, 2, 1 } });
-    memory weights = memory::allocate(eng, { data_types::f32, format::bfyx, tensor{ 1, 2, 2, 1 } });
+    memory::ptr in = engine.allocate_memory({ data_types::f32, format::yxfb, tensor{ 1, 2, 2, 1 } });
+    memory::ptr weights = engine.allocate_memory({ data_types::f32, format::bfyx, tensor{ 1, 2, 2, 1 } });
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         convolution("conv", "in",{ "weights" }),
         data("weights", weights),
         reorder("r1", "conv", format::bfyx, data_types::f32) //optimize data should add conversion from yxfb to bfyx and 'conv' should output data in bfyx as well (IE case)
@@ -1401,23 +1397,23 @@ TEST(reorder_gpu_opt, basic_remove_redundant_output_due_to_implicit_reorders)
     opts.set_option(build_option::outputs({ "r1" }));
     opts.set_option(build_option::optimize_data(true));
 
-    network net(eng, tpl, opts);
+    network net(engine, tpl, opts);
     net.set_input_data("in", in);
     auto outputs = net.execute();
 
     EXPECT_TRUE(outputs.count("conv") == 0);
     ASSERT_TRUE(outputs.count("r1") == 1);
-    EXPECT_TRUE(outputs.at("r1").get_memory().get_layout().format == format::bfyx);
+    EXPECT_TRUE(outputs.at("r1").get_memory()->get_layout().format == format::bfyx);
 }
 
 TEST(reorder_gpu_opt, basic_remove_redundant_due_to_implicit_reorders)
 {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory in = memory::allocate(eng, { data_types::f32, format::yxfb, tensor{ 1, 2, 2, 1 } });
-    memory weights = memory::allocate(eng, { data_types::f32, format::bfyx, tensor{ 1, 2, 2, 1 } });
+    memory::ptr in = engine.allocate_memory({ data_types::f32, format::yxfb, tensor{ 1, 2, 2, 1 } });
+    memory::ptr weights = engine.allocate_memory({ data_types::f32, format::bfyx, tensor{ 1, 2, 2, 1 } });
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         convolution("conv", "in",{ "weights" }),
         data("weights", weights),
         reorder("r1", "conv", format::bfyx, data_types::f32), //optimize data should add conversion from yxfb to bfyx and 'conv' should output data in bfyx as well (IE case)
@@ -1427,7 +1423,7 @@ TEST(reorder_gpu_opt, basic_remove_redundant_due_to_implicit_reorders)
     build_options opts;
     opts.set_option(build_option::optimize_data(true));
 
-    network net(eng, tpl, opts);
+    network net(engine, tpl, opts);
     net.set_input_data("in", in);
     auto outputs = net.execute();
     auto executed_primitives = net.get_executed_primitives();
@@ -1436,16 +1432,16 @@ TEST(reorder_gpu_opt, basic_remove_redundant_due_to_implicit_reorders)
     EXPECT_TRUE(executed_primitives.count("r1") == 0);
     //all pirmitives in this test needs to be executed
     ASSERT_TRUE(outputs.count("output") == 1);
-    EXPECT_TRUE(outputs.at("output").get_memory().get_layout().format == format::bfyx);
+    EXPECT_TRUE(outputs.at("output").get_memory()->get_layout().format == format::bfyx);
 }
 
 TEST(reorder_gpu_opt, non_trivial_remove_redundant)
 {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory in = memory::allocate(eng, { data_types::f32, format::yxfb, tensor{ 1, 1, 5, 2 } });
+    memory::ptr in = engine.allocate_memory({ data_types::f32, format::yxfb, tensor{ 1, 1, 5, 2 } });
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         reorder("r1", "in", format::bfyx, data_types::f32)
     };
 
@@ -1453,7 +1449,7 @@ TEST(reorder_gpu_opt, non_trivial_remove_redundant)
 
     opts.set_option(build_option::optimize_data(true));
 
-    network net(eng, tpl, opts);
+    network net(engine, tpl, opts);
     net.set_input_data("in", in);
     auto outputs = net.execute();
     auto executed_primitives = net.get_executed_primitives();
@@ -1463,15 +1459,15 @@ TEST(reorder_gpu_opt, non_trivial_remove_redundant)
     //ASSERT_TRUE(all_primitives.at("r1") == "_optimized_");
     EXPECT_TRUE(executed_primitives.at("in") != outputs.at("r1").get_event());
     ASSERT_TRUE(outputs.count("r1") == 1);
-    EXPECT_TRUE(outputs.at("r1").get_memory().get_layout().format == format::bfyx);
+    EXPECT_TRUE(outputs.at("r1").get_memory()->get_layout().format == format::bfyx);
 }
 
 TEST(reorder_gpu_opt, mean_mul)
 {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory in  = memory::allocate(eng, { data_types::i8, format::bfyx, tensor{ 1, 3, 1, 2 } });
-    memory mul = memory::allocate(eng, { data_types::f32, format::bfyx, tensor{1, 3, 1, 2 } });
+    memory::ptr in  = engine.allocate_memory({ data_types::i8, format::bfyx, tensor{ 1, 3, 1, 2 } });
+    memory::ptr mul = engine.allocate_memory({ data_types::f32, format::bfyx, tensor{1, 3, 1, 2 } });
 
     set_values<char>(in,
     { 1, 2,
@@ -1481,7 +1477,7 @@ TEST(reorder_gpu_opt, mean_mul)
     { 0.5f, 2.5f, -5.0f, 4.3f, 1.2f, -3.5f });
 
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         data("mul",mul),
         reorder("r1", "in", format::bfyx, data_types::f32,"mul", reorder_mean_mode::mul)
     };
@@ -1489,12 +1485,12 @@ TEST(reorder_gpu_opt, mean_mul)
     float answers[] = { 0.5f, 5.0f, -15.0f, 17.2f, 6.0f, -21.0f };
     build_options opts;
     opts.set_option(build_option::optimize_data(true));
-    network net(eng, tpl, opts);
+    network net(engine, tpl, opts);
     net.set_input_data("in", in);
 
     auto outputs = net.execute();
     auto output = outputs.begin()->second.get_memory();
-    auto ptr = output.pointer<float>();
+    cldnn::mem_lock<float> ptr(output, get_test_stream());
     float* a_ptr = answers;
     for (auto& val : ptr)
         EXPECT_FLOAT_EQ(*(a_ptr++), val);;
@@ -1503,10 +1499,10 @@ TEST(reorder_gpu_opt, mean_mul)
 
 TEST(reorder_gpu_opt, mean_div)
 {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory in = memory::allocate(eng, { data_types::i8, format::bfyx, tensor{ 1, 3, 1, 2 } });
-    memory mul = memory::allocate(eng, { data_types::f32, format::bfyx, tensor{ 1, 3, 1, 2 } });
+    memory::ptr in = engine.allocate_memory({ data_types::i8, format::bfyx, tensor{ 1, 3, 1, 2 } });
+    memory::ptr mul = engine.allocate_memory({ data_types::f32, format::bfyx, tensor{ 1, 3, 1, 2 } });
 
     set_values<char>(in,
     { 1, 2,
@@ -1516,7 +1512,7 @@ TEST(reorder_gpu_opt, mean_div)
     { 0.5f, 2.0f, -3.0f, 8.0f, 1.25f, -3.0f });
 
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         data("mul",mul),
         reorder("r1", "in", format::bfyx, data_types::f32,"mul", reorder_mean_mode::div)
     };
@@ -1524,12 +1520,12 @@ TEST(reorder_gpu_opt, mean_div)
     float answers[] = { 2.0f, 1.0f, -1.0f, 0.5f, 4.0f, -2.0f };
     build_options opts;
     opts.set_option(build_option::optimize_data(true));
-    network net(eng, tpl, opts);
+    network net(engine, tpl, opts);
     net.set_input_data("in", in);
 
     auto outputs = net.execute();
     auto output = outputs.begin()->second.get_memory();
-    auto ptr = output.pointer<float>();
+    cldnn::mem_lock<float> ptr(output, get_test_stream());
     float* a_ptr = answers;
     for (auto& val : ptr)
         EXPECT_FLOAT_EQ(*(a_ptr++), val);;
@@ -1538,9 +1534,9 @@ TEST(reorder_gpu_opt, mean_div)
 
 TEST(reorder_gpu_opt, mean_mul_val)
 {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory in = memory::allocate(eng, { data_types::i8, format::bfyx, tensor{ 1, 3, 1, 2 } });
+    memory::ptr in = engine.allocate_memory({ data_types::i8, format::bfyx, tensor{ 1, 3, 1, 2 } });
 
     set_values<char>(in,
     { 1, 2,
@@ -1548,19 +1544,19 @@ TEST(reorder_gpu_opt, mean_mul_val)
       5, 60 });
     std::vector<float> mul_val = { 2.0f, 0.5f, 10.0f };
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         reorder("r1", "in", format::bfyx, data_types::f32, mul_val, reorder_mean_mode::mul)
     };
 
     float answers[] = { 2.0f, 4.0f, 1.5f, 2.0f, 50.0f, 600.0f };
     build_options opts;
     opts.set_option(build_option::optimize_data(true));
-    network net(eng, tpl, opts);
+    network net(engine, tpl, opts);
     net.set_input_data("in", in);
 
     auto outputs = net.execute();
     auto output = outputs.begin()->second.get_memory();
-    auto ptr = output.pointer<float>();
+    cldnn::mem_lock<float> ptr(output, get_test_stream());
     float* a_ptr = answers;
     for (auto& val : ptr)
         EXPECT_FLOAT_EQ(*(a_ptr++), val);;
@@ -1568,9 +1564,9 @@ TEST(reorder_gpu_opt, mean_mul_val)
 
 TEST(reorder_gpu_opt, mean_mul_val_float_to_int)
 {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory in = memory::allocate(eng, { data_types::f32, format::bfyx, tensor{ 1, 3, 1, 2 } });
+    memory::ptr in = engine.allocate_memory({ data_types::f32, format::bfyx, tensor{ 1, 3, 1, 2 } });
 
     set_values<float>(in,
     { 0.6f, 1.5f,
@@ -1578,19 +1574,19 @@ TEST(reorder_gpu_opt, mean_mul_val_float_to_int)
       5.0f, 60.0f });
     std::vector<float> mul_val = { 1.4f, 0.5f, 5.0f };
     topology tpl{
-        input_layout("in", in.get_layout()),
+        input_layout("in", in->get_layout()),
         reorder("r1", "in", format::bfyx, data_types::i8, mul_val, reorder_mean_mode::mul)
     };
 
     char answers[] = { 1, 2, 2, 2, 25, 127 };
     build_options opts;
     opts.set_option(build_option::optimize_data(true));
-    network net(eng, tpl, opts);
+    network net(engine, tpl, opts);
     net.set_input_data("in", in);
 
     auto outputs = net.execute();
     auto output = outputs.begin()->second.get_memory();
-    auto ptr = output.pointer<char>();
+    cldnn::mem_lock<char> ptr(output, get_test_stream());
     char* a_ptr = answers;
     for (auto& val : ptr)
         EXPECT_EQ(*(a_ptr++), val);
@@ -1599,9 +1595,9 @@ TEST(reorder_gpu_opt, mean_mul_val_float_to_int)
 TEST(reorder_gpu_i32, basic)
 {
     //  Test for converting data types f32->i32
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
     layout output_layout(data_types::i32, format::bfyx, { 2,2,2,2 });
 
     set_values(input, {
@@ -1612,7 +1608,7 @@ TEST(reorder_gpu_i32, basic)
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", output_layout));
 
     network network(engine, topology);
@@ -1632,7 +1628,7 @@ TEST(reorder_gpu_i32, basic)
     };
 
     int32_t* a_ptr = answers;
-    auto output_ptr = output.pointer<int32_t>();
+    cldnn::mem_lock<int32_t> output_ptr(output, get_test_stream());
     for (auto& val : output_ptr)
         EXPECT_EQ(*(a_ptr++), val);
 }
@@ -1640,9 +1636,9 @@ TEST(reorder_gpu_i32, basic)
 TEST(reorder_gpu_i64, basic)
 {
     //  Test for converting data types f32->i64
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
     layout output_layout(data_types::i64, format::bfyx, { 2,2,2,2 });
 
     set_values(input, {
@@ -1653,7 +1649,7 @@ TEST(reorder_gpu_i64, basic)
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", output_layout));
 
     network network(engine, topology);
@@ -1673,19 +1669,19 @@ TEST(reorder_gpu_i64, basic)
     };
 
     int64_t* a_ptr = answers;
-    auto output_ptr = output.pointer<int64_t>();
+    cldnn::mem_lock<int64_t> output_ptr(output, get_test_stream());
     for (auto& val : output_ptr)
         EXPECT_EQ(*(a_ptr++), val);
 }
 
 TEST(reorder_gpu_binary, binary_output)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     cldnn::build_options options;
     options.set_option(cldnn::build_option::optimize_data(true));
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
     layout output_layout(data_types::bin, format::b_fs_yx_32fp, { 2, 2, 2, 2 });
 
     // Data is supposed to be quantized to {0,1} values
@@ -1698,7 +1694,7 @@ TEST(reorder_gpu_binary, binary_output)
     });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", output_layout));
 
     network network(engine, topology);
@@ -1709,17 +1705,17 @@ TEST(reorder_gpu_binary, binary_output)
     EXPECT_EQ(outputs.begin()->first, "reorder");
 
     auto output = outputs.begin()->second.get_memory();
-    auto output_ptr = output.pointer<uint32_t>();
+    cldnn::mem_lock<uint32_t> output_ptr(output, get_test_stream());
 
     std::vector<uint32_t > answers = { 1, 2, 3, 1,
                                        1, 1, 0, 3 };
 
     // Check that layout and memory contains logical size of tensor
-    ASSERT_EQ(output.count(), input.get_layout().count());
-    ASSERT_EQ(output.get_layout().count(), input.get_layout().count());
+    ASSERT_EQ(output->count(), input->get_layout().count());
+    ASSERT_EQ(output->get_layout().count(), input->get_layout().count());
 
     // Check that memory physical size consider binary pack
-    ASSERT_EQ(output.size(), answers.size() * sizeof(uint32_t));
+    ASSERT_EQ(output->size(), answers.size() * sizeof(uint32_t));
 
     for (size_t i = 0; i < answers.size(); ++i) {
         EXPECT_EQ(answers[i], output_ptr[i]) << "index: " << i;
@@ -1728,12 +1724,12 @@ TEST(reorder_gpu_binary, binary_output)
 
 TEST(reorder_gpu_binary, binary_input)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     cldnn::build_options options;
     options.set_option(cldnn::build_option::optimize_data(true));
 
-    auto input = memory::allocate(engine, { data_types::bin, format::b_fs_yx_32fp,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::bin, format::b_fs_yx_32fp,{ 2, 2, 2, 2 } });
     layout output_layout(data_types::f32, format::bfyx, { 2, 2, 2, 2 });
 
     // Data is supposed to be quantized to {0,1} values
@@ -1749,7 +1745,7 @@ TEST(reorder_gpu_binary, binary_input)
                                  1, 1, 0, 3 });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", output_layout));
 
     network network(engine, topology);
@@ -1760,13 +1756,13 @@ TEST(reorder_gpu_binary, binary_input)
     EXPECT_EQ(outputs.begin()->first, "reorder");
 
     auto output = outputs.begin()->second.get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     // Check that layout and memory contains logical size of tensor
-    ASSERT_EQ(output.count(), input.get_layout().count());
-    ASSERT_EQ(output.get_layout().count(), input.get_layout().count());
+    ASSERT_EQ(output->count(), input->get_layout().count());
+    ASSERT_EQ(output->get_layout().count(), input->get_layout().count());
 
-    ASSERT_EQ(output.size(), answers.size() * sizeof(float));
+    ASSERT_EQ(output->size(), answers.size() * sizeof(float));
 
     for (size_t i = 0; i < answers.size(); ++i) {
         EXPECT_EQ(answers[i], output_ptr[i]) << "index: " << i;
@@ -1791,9 +1787,9 @@ TEST(reorder_gpu_f32, bfwzyx_bfyx_chain)
     // Expected output:
     // 0 1  -1 0    8  9   7  8
     // 2 3   1 2   10 11   9 10
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, layout{ data_types::f32, format::bfyx, tensor{ batch(1), feature(4), spatial(2, 2) } });
+    auto input = engine.allocate_memory(layout{ data_types::f32, format::bfyx, tensor{ batch(1), feature(4), spatial(2, 2) } });
 
     std::vector<float> data = {
         1.f, 2.f, 3.f, 4.f,
@@ -1813,7 +1809,7 @@ TEST(reorder_gpu_f32, bfwzyx_bfyx_chain)
     };
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder1", "input", format::bfwzyx, data_types::f32),
         reshape("reshape1", "reorder1", tensor(batch(2), feature(2), spatial(1, 1, 2, 2) )),
         reorder("reorder2", "reshape1", format::bfwzyx, data_types::f32, sub_bfwzyx),
@@ -1832,8 +1828,8 @@ TEST(reorder_gpu_f32, bfwzyx_bfyx_chain)
     EXPECT_EQ(outputs.begin()->first, "out_reorder");
 
     auto output = outputs.begin()->second.get_memory();
-    EXPECT_TRUE(output.get_layout().format == format::bfwzyx);
-    auto sizes = output.get_layout().size;
+    EXPECT_TRUE(output->get_layout().format == format::bfwzyx);
+    auto sizes = output->get_layout().size;
     EXPECT_EQ(sizes.batch[0], 1);
     EXPECT_EQ(sizes.feature[0], 4);
     EXPECT_EQ(sizes.spatial[0], 2);
@@ -1841,7 +1837,7 @@ TEST(reorder_gpu_f32, bfwzyx_bfyx_chain)
     EXPECT_EQ(sizes.spatial[2], 1);
     EXPECT_EQ(sizes.spatial[2], 1);
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     ASSERT_EQ(output_ptr.size(), expected.size());
 
     for (size_t i = 0; i < expected.size(); i++)
@@ -1852,20 +1848,20 @@ TEST(reorder_gpu_f32, bfwzyx_bfyx_chain)
 
 TEST(reorder_gpu_f32, bfzyx_to_bsv16_fsv16)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int32_t b_in = 2;
     const int32_t f_in = 2;
     const int32_t x_in = 2;
     const int32_t y_in = 2;
     const int32_t z_in = 2;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, { b_in,f_in,x_in,y_in,z_in } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, { b_in,f_in,x_in,y_in,z_in } });
     layout output_layout(data_types::f32, format::bs_fs_zyx_bsv16_fsv16,{ b_in,f_in,x_in,y_in,z_in });
 
     tests::set_random_values<float>(input);
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             reorder("reorder", "input", output_layout));
 
     network network(engine, topology);
@@ -1906,8 +1902,8 @@ TEST(reorder_gpu_f32, bfzyx_to_bsv16_fsv16)
         return output_offset;
     };
 
-    auto input_ptr = input.pointer<float>();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     int32_t linear_index = 0;
     for (int32_t b = 0; b < b_in; b++) {
         for (int32_t f = 0; f < f_in; f++) {
@@ -1930,7 +1926,7 @@ TEST(reorder_gpu_f32, bfzyx_to_bsv16_fsv16)
 
 TEST(reorder_gpu_f32, bfzyx_to_bsv16_fsv16_padded)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     const int32_t b_in = 2;
     const int32_t f_in = 2;
     const int32_t x_in = 2;
@@ -1941,13 +1937,13 @@ TEST(reorder_gpu_f32, bfzyx_to_bsv16_fsv16_padded)
     const int32_t y_pad= 2;
     const int32_t x_pad= 1;
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, { b_in,f_in,x_in,y_in,z_in } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, { b_in,f_in,x_in,y_in,z_in } });
     layout output_layout(data_types::f32, format::bs_fs_zyx_bsv16_fsv16,{ b_in,f_in,x_in,y_in,z_in });
 
     tests::set_random_values<float>(input);
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             reorder("reorder", "input", output_layout.with_padding(padding({0, 0, x_pad, y_pad, 0}, 0.f))));
 
     network network(engine, topology);
@@ -1988,8 +1984,8 @@ TEST(reorder_gpu_f32, bfzyx_to_bsv16_fsv16_padded)
         return output_offset;
     };
 
-    auto input_ptr = input.pointer<float>();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     int32_t linear_index = 0;
     for (int32_t b = 0; b < b_in; b++) {
         for (int32_t f = 0; f < f_in; f++) {
@@ -2011,16 +2007,16 @@ TEST(reorder_gpu_f32, bfzyx_to_bsv16_fsv16_padded)
 
 TEST(reorder_gpu_f32, b_fs_yx_fsv16_to_bfyx_opt_allowed)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::b_fs_yx_fsv16, { 2, 12, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::b_fs_yx_fsv16, { 2, 12, 1, 1 } });
 
     set_values(input, { 0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 13.f, 14.f, 15.f,
                         16.f, 17.f, 18.f, 19.f, 20.f, 21.f, 22.f, 23.f, 24.f, 25.f, 26.f, 27.f, 28.f, 29.f, 30.f, 31.f });
 
     const std::string reorder_name = "reorder_prim";
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             activation("first_activation", "input", activation_func::abs),
             reorder(reorder_name, "first_activation", format::bfyx, data_types::f32),
             activation("second_activation", reorder_name, activation_func::abs));
@@ -2045,7 +2041,7 @@ TEST(reorder_gpu_f32, b_fs_yx_fsv16_to_bfyx_opt_allowed)
             16.f, 17.f, 18.f, 19.f, 20.f, 21.f, 22.f, 23.f, 24.f, 25.f, 26.f, 27.f,
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     ASSERT_EQ(output_ptr.size(), 24);
     for (size_t i = 0; i < output_ptr.size(); i++)
     {
@@ -2055,19 +2051,19 @@ TEST(reorder_gpu_f32, b_fs_yx_fsv16_to_bfyx_opt_allowed)
 
 TEST(reorder_gpu_f32, b_fs_yx_fsv16_to_bfyx_opt_not_allowed)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::b_fs_yx_fsv16, { 1, 8, 1, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::oiyx, { 1, 8, 3, 3 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::b_fs_yx_fsv16, { 1, 8, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::oiyx, { 1, 8, 3, 3 } });
 
     set_values(input, { 0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f });
 
-    set_values(weights, std::vector<float>(weights.count(), 1));
+    set_values(weights, std::vector<float>(weights->count(), 1));
 
     const std::string reorder_name = "reorder";
     const std::string reorder_primitive_name = "reorder:" + reorder_name;
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             data("weights", weights),
             reorder(reorder_name, "input", format::bfyx, data_types::f32),
             convolution("convolution", reorder_name, {"weights"}, {1,1,1,1}, {0,0,-1,-1}, {1,1,1,1}));
@@ -2089,7 +2085,7 @@ TEST(reorder_gpu_f32, b_fs_yx_fsv16_to_bfyx_opt_not_allowed)
 
     float answers[1] = { 28.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 1; i++)
     {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]) << "i=" << i;
@@ -2098,9 +2094,9 @@ TEST(reorder_gpu_f32, b_fs_yx_fsv16_to_bfyx_opt_not_allowed)
 
 TEST(reorder_gpu_f32, b_fs_yx_fsv16_to_bfyx_opt_padded)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32,
+    auto input = engine.allocate_memory({ data_types::f32,
                                             format::b_fs_yx_fsv16,
                                             { 2, 4, 1, 1 },
                                             padding({1, 16, 0, 0}, {1, 0, 0, 0}) });
@@ -2124,7 +2120,7 @@ TEST(reorder_gpu_f32, b_fs_yx_fsv16_to_bfyx_opt_padded)
 
     const std::string reorder_name = "reorder_prim";
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder(reorder_name, "input", format::bfyx, data_types::f32),
         activation("activation", reorder_name, activation_func::abs));
 
@@ -2148,7 +2144,7 @@ TEST(reorder_gpu_f32, b_fs_yx_fsv16_to_bfyx_opt_padded)
             16.f, 17.f, 18.f, 19.f,
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     ASSERT_EQ(output_ptr.size(), 8);
     for (size_t i = 0; i < output_ptr.size(); i++) {
         EXPECT_FLOAT_EQ(answers[i], output_ptr[i]) << "i=" << i;
@@ -2158,21 +2154,21 @@ TEST(reorder_gpu_f32, b_fs_yx_fsv16_to_bfyx_opt_padded)
 TEST(reorder_gpu, any_format) {
     auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, layout(data_types::f32, format::yxfb, tensor(5, 7, 13, 9)));
+    auto input = engine.allocate_memory(layout(data_types::f32, format::yxfb, tensor(5, 7, 13, 9)));
 
     topology topo;
-    topo.add(input_layout("in", input.get_layout()));
+    topo.add(input_layout("in", input->get_layout()));
     topo.add(reorder("out", "in", format::any, data_types::f32));
 
     network net(engine, topo);
 
-    auto data = generate_random_1d<float>(input.count(), -1, 1);
+    auto data = generate_random_1d<float>(input->count(), -1, 1);
     set_values(input, data);
     net.set_input_data("in", input);
 
     auto outputs = net.execute();
     auto out_mem = outputs.at("out").get_memory();
-    auto output = out_mem.pointer<float>();
+    cldnn::mem_lock<float> output(out_mem, get_test_stream());
 
     for (size_t i = 0; i < data.size(); ++i) {
         EXPECT_EQ(output[i], data[i]) << "i = " << i;
@@ -2181,9 +2177,9 @@ TEST(reorder_gpu, any_format) {
 
 TEST(reorder_image2d_rgba_to_bfyx_gpu, basic)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::image_2d_rgba, { 1, 3, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::image_2d_rgba, { 1, 3, 2, 2 } });
     layout output_layout(data_types::f16, format::bfyx, { 1, 3, 2, 2 });
 
     set_values<unsigned char>(input, {
@@ -2194,7 +2190,7 @@ TEST(reorder_image2d_rgba_to_bfyx_gpu, basic)
         });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", output_layout));
 
     network network(engine, topology);
@@ -2217,7 +2213,7 @@ TEST(reorder_image2d_rgba_to_bfyx_gpu, basic)
         50.0f, 253.0f,
     };
 
-    auto output_ptr = output.pointer<FLOAT16>();
+    cldnn::mem_lock<FLOAT16> output_ptr (output, get_test_stream());
     for (int i = 0; i < 12; i++)
     {
         EXPECT_NEAR(FLOAT16(answers[i] / 255.f), output_ptr[i], 1e-3f);
@@ -2227,9 +2223,9 @@ TEST(reorder_image2d_rgba_to_bfyx_gpu, basic)
 
 TEST(reorder_bfyx_to_image2d_rgba_gpu, basic)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 3, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 3, 2, 2 } });
     layout output_layout(data_types::u8, format::image_2d_rgba, { 1, 3, 2, 2 });
 
     set_values<FLOAT16>(input, {
@@ -2244,7 +2240,7 @@ TEST(reorder_bfyx_to_image2d_rgba_gpu, basic)
         });
 
     topology topology(
-        input_layout("input", input.get_layout()),
+        input_layout("input", input->get_layout()),
         reorder("reorder", "input", output_layout));
 
     network network(engine, topology);
@@ -2263,7 +2259,7 @@ TEST(reorder_bfyx_to_image2d_rgba_gpu, basic)
         251, 252, 253, 0
     };
 
-    auto output_ptr = output.pointer<unsigned char>();
+    cldnn::mem_lock<unsigned char> output_ptr(output, get_test_stream());
     for (int i = 0; i < 16; i++)
     {
         EXPECT_EQ(answers[i], output_ptr[i]);
@@ -2329,7 +2325,7 @@ class reorder_test : public tests::generic_test
     }
 
     template<typename InputType, typename OutputType>
-    memory generate_reference_typed(const std::vector<cldnn::memory>& inputs)
+    memory::ptr generate_reference_typed(const std::vector<cldnn::memory::ptr>& inputs)
     {
         auto reorder = std::static_pointer_cast<cldnn::reorder>(layer_params);
         primitive_id mean = reorder->mean;
@@ -2337,12 +2333,12 @@ class reorder_test : public tests::generic_test
         assert(mean == "");
         assert(subtract_per_feature.size() == 0);
 
-        auto output = memory::allocate(engine, cldnn::layout(*reorder->output_data_type, inputs[0].get_layout().format, inputs[0].get_layout().size));
+        auto output = engine.allocate_memory(cldnn::layout(*reorder->output_data_type, inputs[0]->get_layout().format, inputs[0]->get_layout().size));
 
-        cldnn::pointer<InputType> input_mem = inputs[0].pointer<InputType>();
-        cldnn::pointer<OutputType> output_mem = output.pointer<OutputType>();
+        cldnn::mem_lock<InputType> input_mem(inputs[0], get_test_stream());
+        cldnn::mem_lock<OutputType> output_mem(output, get_test_stream());
 
-        for (size_t i = 0; i < inputs[0].get_layout().get_linear_size(); i++)
+        for (size_t i = 0; i < inputs[0]->get_layout().get_linear_size(); i++)
         {
             // Write the output in the same order as the input with type conversion as needed.
             // The correct order will be checked in generic_test::compare_buffers.
@@ -2352,7 +2348,7 @@ class reorder_test : public tests::generic_test
         return output;
     }
 
-    virtual memory generate_reference(const std::vector<cldnn::memory>& inputs)
+    virtual memory::ptr generate_reference(const std::vector<cldnn::memory::ptr>& inputs)
     {
         if (generic_params->data_type == data_types::f32)
         {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/resample_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/resample_gpu_test.cpp
index 1a02c57d79c0e7..ca7da7f9e83013 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/resample_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/resample_gpu_test.cpp
@@ -3,19 +3,16 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/resample.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/reorder.hpp>
-#include <api/data.hpp>
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/resample.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/data.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 TEST(resample_gpu, basic_in2x3x2x2_nearest) {
     //  Input  : 2x2x3x2
@@ -29,15 +26,15 @@ TEST(resample_gpu, basic_in2x3x2x2_nearest) {
     //  f1: b0:  7    8  -16   b1:   12   9     -17
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 3, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 3, 2 } });
 
     auto output_size = tensor(batch(2), feature(2), spatial(6, 4));
     uint32_t num_filter = 0u;
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(resample("upsampling", "input", output_size, num_filter, resample_type::nearest));
 
     set_values(input, {
@@ -58,7 +55,7 @@ TEST(resample_gpu, basic_in2x3x2x2_nearest) {
     auto outputs = net.execute();
 
     auto output = outputs.at("upsampling").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     float answers[96] = {
         1.f, 1.f, 2.f,   2.f,   -10.f,  -10.f,
@@ -101,15 +98,15 @@ TEST(resample_gpu, basic_in2x3x2x2_bilinear) {
     //  f0: b0:  3    4
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
 
     auto output_size = tensor(batch(1), feature(1), spatial(4, 4));
     uint32_t num_filter = 1u;
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(resample("upsampling", "input", output_size, num_filter, resample_type::caffe_bilinear));
 
     set_values(input, {
@@ -123,9 +120,9 @@ TEST(resample_gpu, basic_in2x3x2x2_bilinear) {
     auto outputs = net.execute();
 
     auto output = outputs.at("upsampling").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
-    EXPECT_EQ(output.get_layout().get_linear_size(), (size_t) 16);
+    EXPECT_EQ(output->get_layout().get_linear_size(), (size_t) 16);
 
     float answers[16] = {
         1.f, 1.25f, 1.75f, 2.f,
@@ -152,14 +149,14 @@ TEST(resample_gpu, basic_in1x1x2x2_interp) {
     //  f0: b0:  3    4
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
 
     auto output_size = tensor(batch(1), feature(1), spatial(4, 4));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(resample("upsampling", "input", output_size, {0, 0, 0, 0}, {0, 0, 0, 0}, 0, resample_type::bilinear));
 
     set_values(input, {
@@ -173,9 +170,9 @@ TEST(resample_gpu, basic_in1x1x2x2_interp) {
     auto outputs = net.execute();
 
     auto output = outputs.at("upsampling").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
-    EXPECT_EQ(output.get_layout().get_linear_size(), (size_t) 16);
+    EXPECT_EQ(output->get_layout().get_linear_size(), (size_t) 16);
 
     float answers[16] = {
         1.0f, 1.5f, 2.0f, 2.0f,
@@ -202,14 +199,14 @@ TEST(resample_gpu, basic_in1x1x2x2_interp_f16) {
     //  f0: b0:  3    4
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
 
     auto output_size = tensor(batch(1), feature(1), spatial(4, 4));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reorder("input_to_b_fs_yx_fsv16", "input", format::b_fs_yx_fsv16, data_types::f32));
     topology.add(resample("resample", "input_to_b_fs_yx_fsv16", output_size, {0, 0, 0, 0}, {0, 0, 0, 0}, 0, resample_type::bilinear));
     topology.add(reorder("res_to_bfyx", "resample", format::bfyx, data_types::f32));
@@ -228,12 +225,12 @@ TEST(resample_gpu, basic_in1x1x2x2_interp_f16) {
     auto outputs = net.execute();
 
     auto resample_out = outputs.at("resample").get_memory();
-    ASSERT_EQ(resample_out.get_layout().format, format::b_fs_yx_fsv16);
+    ASSERT_EQ(resample_out->get_layout().format, format::b_fs_yx_fsv16);
 
     auto output = outputs.at("res_to_bfyx").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
-    EXPECT_EQ(output.get_layout().get_linear_size(), (size_t) 16);
+    EXPECT_EQ(output->get_layout().get_linear_size(), (size_t) 16);
 
     float answers[16] = {
         1.0f, 1.5f, 2.0f, 2.0f,
@@ -260,14 +257,14 @@ TEST(resample_gpu, basic_in1x1x2x2_interp_fsv32) {
     //  f0: b0:  3    4
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
 
     auto output_size = tensor(batch(1), feature(1), spatial(4, 4));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reorder("input_to_fs_b_yx_fsv32", "input", format::fs_b_yx_fsv32, data_types::f16));
     topology.add(resample("resample", "input_to_fs_b_yx_fsv32", output_size, {0, 0, 0, 0}, {0, 0, 0, 0}, 0, resample_type::bilinear));
     topology.add(reorder("res_to_bfyx", "resample", format::bfyx, data_types::f32));
@@ -286,12 +283,12 @@ TEST(resample_gpu, basic_in1x1x2x2_interp_fsv32) {
     auto outputs = net.execute();
 
     auto resample_out = outputs.at("resample").get_memory();
-    ASSERT_EQ(resample_out.get_layout().format, format::fs_b_yx_fsv32);
+    ASSERT_EQ(resample_out->get_layout().format, format::fs_b_yx_fsv32);
 
     auto output = outputs.at("res_to_bfyx").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
-    EXPECT_EQ(output.get_layout().get_linear_size(), (size_t) 16);
+    EXPECT_EQ(output->get_layout().get_linear_size(), (size_t) 16);
 
     float answers[16] = {
         1.0f, 1.5f, 2.0f, 2.0f,
@@ -319,14 +316,14 @@ TEST(resample_gpu, basic_in1x1x2x2_interp_align_1) {
     //  f0: b0:  3    4
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
 
     auto output_size = tensor(batch(1), feature(1), spatial(4, 4));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(resample("upsampling", "input", output_size, {0, 0, 0, 0}, {0, 0, 0, 0}, 1, resample_type::bilinear));
 
     set_values(input, {
@@ -340,9 +337,9 @@ TEST(resample_gpu, basic_in1x1x2x2_interp_align_1) {
     auto outputs = net.execute();
 
     auto output = outputs.at("upsampling").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
-    EXPECT_EQ(output.get_layout().get_linear_size(), (size_t) 16);
+    EXPECT_EQ(output->get_layout().get_linear_size(), (size_t) 16);
 
     float answers[16] = {
             1.000000f, 1.333333f, 1.666667f, 2.000000f,
@@ -369,15 +366,15 @@ TEST(resample_gpu, nearest_asymmetric) {
     //  f0: b0:  3    4
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
 
     auto output_size = tensor(batch(1), feature(1), spatial(5, 4));
 
     topology topology;
     uint32_t num_filter = 1u;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(resample("upsampling", "input", output_size, num_filter, resample_type::nearest));
 
     set_values(input, {
@@ -391,9 +388,9 @@ TEST(resample_gpu, nearest_asymmetric) {
     auto outputs = net.execute();
 
     auto output = outputs.at("upsampling").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
-    EXPECT_EQ(output.get_layout().get_linear_size(), (size_t)20);
+    EXPECT_EQ(output->get_layout().get_linear_size(), (size_t)20);
 
     float answers[20] = {
         1.f, 1.f, 1.f, 2.f, 2.f,
@@ -420,15 +417,15 @@ TEST(resample_gpu, nearest_asymmetric_i8) {
     //  f0: b0:  3    4
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::i8, format::bfyx, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::i8, format::bfyx, { 1, 1, 2, 2 } });
 
     auto output_size = tensor(batch(1), feature(1), spatial(5, 4));
 
     topology topology;
     uint32_t num_filter = 1u;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(resample("upsampling", "input", output_size, num_filter, resample_type::nearest));
 
     set_values<int8_t>(input, {
@@ -442,9 +439,9 @@ TEST(resample_gpu, nearest_asymmetric_i8) {
     auto outputs = net.execute();
 
     auto output = outputs.at("upsampling").get_memory();
-    auto output_ptr = output.pointer<int8_t>();
+    cldnn::mem_lock<int8_t> output_ptr(output, get_test_stream());
 
-    EXPECT_EQ(output.get_layout().get_linear_size(), (size_t)20);
+    EXPECT_EQ(output->get_layout().get_linear_size(), (size_t)20);
 
     int8_t answers[20] = {
             1, 1, 1, 2, 2,
@@ -471,15 +468,15 @@ TEST(resample_gpu, bilinear_asymmetric) {
     //  f0: b0:  3    4
     //
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
 
     auto output_size = tensor(batch(1), feature(1), spatial(6, 4));
 
     topology topology;
     uint32_t num_filter = 1u;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(resample("upsampling", "input", output_size, num_filter, resample_type::caffe_bilinear));
 
     set_values(input, {
@@ -493,9 +490,9 @@ TEST(resample_gpu, bilinear_asymmetric) {
     auto outputs = net.execute();
 
     auto output = outputs.at("upsampling").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
-    EXPECT_EQ(output.get_layout().get_linear_size(), (size_t)24);
+    EXPECT_EQ(output->get_layout().get_linear_size(), (size_t)24);
 
     float answers[24] = {
         1.f, 1.f, 1.33333f, 1.66667f, 2.f, 2.f,
@@ -525,21 +522,21 @@ struct resample_random_test_params {
 
 struct resample_random_test : testing::TestWithParam<resample_random_test_params>{
     template <typename T>
-    void fill_random_typed(memory& mem, int min, int max, int k) {
-        auto size = mem.get_layout().size;
+    void fill_random_typed(memory::ptr mem, int min, int max, int k) {
+        auto size = mem->get_layout().size;
         size_t b = size.batch[0];
         size_t f = size.feature[0];
         size_t x = size.spatial[0];
         size_t y = size.spatial[1];
 
         auto data = generate_random_4d<T>(b, f, y, x, min, max, k);
-        auto ptr = mem.pointer<T>();
+        cldnn::mem_lock<T> ptr(mem, get_test_stream());
         for (size_t bi = 0; bi < b; ++bi) {
             for (size_t fi = 0; fi < f; ++fi) {
                 for (size_t yi = 0; yi < y; ++yi) {
                     for (size_t xi = 0; xi < x; ++xi) {
                         auto coords = tensor(batch(bi), feature(fi), spatial(xi, yi, 0, 0));
-                        auto offset = mem.get_layout().get_linear_offset(coords);
+                        auto offset = mem->get_layout().get_linear_offset(coords);
                         ptr[offset] = data[bi][fi][yi][xi];
                     }
                 }
@@ -547,8 +544,8 @@ struct resample_random_test : testing::TestWithParam<resample_random_test_params
         }
     }
 
-    void fill_random(memory& mem) {
-        auto dt = mem.get_layout().data_type;
+    void fill_random(memory::ptr mem) {
+        auto dt = mem->get_layout().data_type;
         switch (dt) {
         case data_types::f32:
             fill_random_typed<float>(mem, -127, 127, 2);
@@ -568,19 +565,19 @@ struct resample_random_test : testing::TestWithParam<resample_random_test_params
     }
 
     template <typename T>
-    void compare_nearest_typed(const memory& input, const memory& output, uint32_t align_corners) {
-        auto output_lay = output.get_layout();
+    void compare_nearest_typed(const memory::ptr input, const memory::ptr output, uint32_t align_corners) {
+        auto output_lay = output->get_layout();
         size_t b = output_lay.size.batch[0];
         size_t f = output_lay.size.feature[0];
         size_t x = output_lay.size.spatial[0];
         size_t y = output_lay.size.spatial[1];
-        size_t in_x = input.get_layout().size.spatial[0];
-        size_t in_y = input.get_layout().size.spatial[1];
+        size_t in_x = input->get_layout().size.spatial[0];
+        size_t in_y = input->get_layout().size.spatial[1];
         float x_ratio = x > align_corners ? static_cast<float>(in_x - align_corners) / static_cast<float>(x - align_corners) : 0.f;
         float y_ratio = y > align_corners ? static_cast<float>(in_y - align_corners) / static_cast<float>(y - align_corners) : 0.f;
 
-        auto in_ptr = input.pointer<T>();
-        auto out_ptr = output.pointer<T>();
+        cldnn::mem_lock<T> in_ptr(input, get_test_stream());
+        cldnn::mem_lock<T> out_ptr(output, get_test_stream());
         for (size_t bi = 0; bi < b; ++bi) {
             for (size_t fi = 0; fi < f; ++fi) {
                 for (size_t yi = 0; yi < y; ++yi) {
@@ -588,10 +585,10 @@ struct resample_random_test : testing::TestWithParam<resample_random_test_params
                         auto in_xi = static_cast<size_t>(floor(x_ratio * xi));
                         auto in_yi = static_cast<size_t>(floor(y_ratio * yi));
                         auto in_coords = tensor(batch(bi), feature(fi), spatial(in_xi, in_yi, 0, 0));
-                        auto in_offset = input.get_layout().get_linear_offset(in_coords);
+                        auto in_offset = input->get_layout().get_linear_offset(in_coords);
                         auto in_val = in_ptr[in_offset];
                         auto out_coords = tensor(batch(bi), feature(fi), spatial(xi, yi, 0, 0));
-                        auto out_offset = output.get_layout().get_linear_offset(out_coords);
+                        auto out_offset = output->get_layout().get_linear_offset(out_coords);
                         auto out_val = out_ptr[out_offset];
                         EXPECT_EQ(in_val, out_val) << " at bi=" << bi << ", fi=" << fi << ", xi=" << xi << ", yi=" << yi;
                     }
@@ -601,20 +598,20 @@ struct resample_random_test : testing::TestWithParam<resample_random_test_params
     }
 
     template <typename InT, typename OutT>
-    void compare_bilinear_typed(const memory& input, const memory& output, uint32_t align_corners) {
-        auto output_lay = output.get_layout();
+    void compare_bilinear_typed(const memory::ptr input, const memory::ptr output, uint32_t align_corners) {
+        auto output_lay = output->get_layout();
         size_t b = output_lay.size.batch[0];
         size_t f = output_lay.size.feature[0];
         size_t x = output_lay.size.spatial[0];
         size_t y = output_lay.size.spatial[1];
-        auto input_lay = input.get_layout();
+        auto input_lay = input->get_layout();
         size_t in_x = input_lay.size.spatial[0];
         size_t in_y = input_lay.size.spatial[1];
         float x_ratio = x > align_corners ? static_cast<float>(in_x - align_corners) / static_cast<float>(x - align_corners) : 0.f;
         float y_ratio = y > align_corners ? static_cast<float>(in_y - align_corners) / static_cast<float>(y - align_corners) : 0.f;
 
-        auto in_ptr = input.pointer<InT>();
-        auto out_ptr = output.pointer<OutT>();
+        cldnn::mem_lock<InT> in_ptr(input, get_test_stream());
+        cldnn::mem_lock<OutT> out_ptr(output, get_test_stream());
         for (size_t bi = 0; bi < b; ++bi) {
             for (size_t fi = 0; fi < f; ++fi) {
                 for (size_t yi = 0; yi < y; ++yi) {
@@ -658,8 +655,8 @@ struct resample_random_test : testing::TestWithParam<resample_random_test_params
         }
     }
 
-    void compare(const memory& input, const memory& output, resample_type operation, uint32_t align_corners) {
-        auto dt = input.get_layout().data_type;
+    void compare(const memory::ptr input, const memory::ptr output, resample_type operation, uint32_t align_corners) {
+        auto dt = input->get_layout().data_type;
         if (operation == resample_type::nearest) {
             // Nearest resampling implicitly ignores align_corners
             if (dt == data_types::f32) {
@@ -691,7 +688,7 @@ struct resample_random_test : testing::TestWithParam<resample_random_test_params
     }
 
     void execute(const resample_random_test_params& params) {
-        auto eng = get_test_engine();
+        auto& engine = get_test_engine();
 
         auto in_layout = layout(params.input_type, params.in_format, params.input_size);
 
@@ -704,9 +701,9 @@ struct resample_random_test : testing::TestWithParam<resample_random_test_params
         auto build_opts = build_options(
             build_option::force_implementations({ {"resample", {params.out_format, ""}} })
         );
-        auto net = network(eng, topo, build_opts);
+        auto net = network(engine, topo, build_opts);
 
-        auto in_mem = memory::allocate(eng, in_layout);
+        auto in_mem = engine.allocate_memory(in_layout);
         fill_random(in_mem);
         net.set_input_data("in", in_mem);
 
@@ -783,21 +780,21 @@ struct caffe_resample_random_test_params {
 struct caffe_resample_random_test : testing::TestWithParam<caffe_resample_random_test_params>
 {
     template <typename T>
-    void fill_random_typed(memory& mem, int min, int max, int k) {
-        auto size = mem.get_layout().size;
+    void fill_random_typed(memory::ptr mem, int min, int max, int k) {
+        auto size = mem->get_layout().size;
         size_t b = size.batch[0];
         size_t f = size.feature[0];
         size_t x = size.spatial[0];
         size_t y = size.spatial[1];
 
         auto data = generate_random_4d<T>(b, f, y, x, min, max, k);
-        auto ptr = mem.pointer<T>();
+        cldnn::mem_lock<T> ptr(mem, get_test_stream());
         for (size_t bi = 0; bi < b; ++bi) {
             for (size_t fi = 0; fi < f; ++fi) {
                 for (size_t yi = 0; yi < y; ++yi) {
                     for (size_t xi = 0; xi < x; ++xi) {
                         auto coords = tensor(batch(bi), feature(fi), spatial(xi, yi, 0, 0));
-                        auto offset = mem.get_layout().get_linear_offset(coords);
+                        auto offset = mem->get_layout().get_linear_offset(coords);
                         ptr[offset] = data[bi][fi][yi][xi];
                     }
                 }
@@ -805,8 +802,8 @@ struct caffe_resample_random_test : testing::TestWithParam<caffe_resample_random
         }
     }
 
-    void fill_random(memory& mem) {
-        auto dt = mem.get_layout().data_type;
+    void fill_random(memory::ptr mem) {
+        auto dt = mem->get_layout().data_type;
         switch (dt) {
         case data_types::f32:
             fill_random_typed<float>(mem, -127, 127, 2);
@@ -826,16 +823,16 @@ struct caffe_resample_random_test : testing::TestWithParam<caffe_resample_random
     }
 
     template <typename T>
-    bool compare_outputs(const memory& out_ref, const memory& out_opt) {
-        auto output_lay = out_ref.get_layout();
-        auto opt_output_lay = out_opt.get_layout();
+    bool compare_outputs(const memory::ptr out_ref, const memory::ptr out_opt) {
+        auto output_lay = out_ref->get_layout();
+        auto opt_output_lay = out_opt->get_layout();
 
         size_t b = output_lay.size.batch[0];
         size_t f = output_lay.size.feature[0];
         size_t x = output_lay.size.spatial[0];
         size_t y = output_lay.size.spatial[1];
-        auto ref_ptr = out_ref.pointer<T>();
-        auto opt_ptr = out_opt.pointer<T>();
+        cldnn::mem_lock<T> ref_ptr(out_ref, get_test_stream());
+        cldnn::mem_lock<T> opt_ptr(out_opt, get_test_stream());
         for (size_t bi = 0; bi < b; ++bi) {
             for (size_t fi = 0; fi < f; ++fi) {
                 for (size_t yi = 0; yi < y; ++yi) {
@@ -859,10 +856,10 @@ struct caffe_resample_random_test : testing::TestWithParam<caffe_resample_random
     }
 
     void execute_compare(const caffe_resample_random_test_params& params, bool check_result) {
-        auto eng = cldnn::engine();
+        auto& engine = get_test_engine();
 
         auto in_layout = layout(params.input_type, params.in_format, params.input_size);
-        auto in_mem = memory::allocate(eng, in_layout);
+        auto in_mem = engine.allocate_memory(in_layout);
         fill_random(in_mem);
 
         cldnn::topology topo;
@@ -877,15 +874,13 @@ struct caffe_resample_random_test : testing::TestWithParam<caffe_resample_random
         build_opts.set_option(build_option::outputs({"resample"}));
         build_opts.set_option(build_option::force_implementations({ {"resample", {params.in_format, "resample_ref"}} }));
 
-        auto net = network(eng, topo, build_opts);
+        auto net = network(engine, topo, build_opts);
         net.set_input_data("in", in_mem);
 
         auto result = net.execute();
         auto output = result.at("resample").get_memory();
 
         // Execute resample_opt
-        auto eng_opt = cldnn::engine();
-
         cldnn::topology topo_opt;
         topo_opt.add(input_layout("in", in_layout));
         auto prim_opt = resample("resample_opt", "in", params.output_size, params.num_filter, params.operation_type);
@@ -898,7 +893,7 @@ struct caffe_resample_random_test : testing::TestWithParam<caffe_resample_random
         build_opts_opt.set_option(build_option::outputs({"resample_opt"}));
         build_opts.set_option(build_option::force_implementations({ {"resample_opt", {params.in_format, "resample_opt"}} }));
 
-        auto net_opt = network(eng_opt, topo_opt, build_opts_opt);
+        auto net_opt = network(engine, topo_opt, build_opts_opt);
 
         // Use in_mem from ref network
         net_opt.set_input_data("in", in_mem);
@@ -969,19 +964,19 @@ TEST(resample_gpu, interpolate_in2x2x3x2_nearest1) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     auto output_size = tensor(batch(b), feature(f), spatial(x*2, y*2));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::nearest;
@@ -1009,7 +1004,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_nearest1) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     float answers[96] = {
          0.f,  1.f,  1.f,  1.f,
@@ -1058,19 +1053,19 @@ TEST(resample_gpu, interpolate_in2x2x3x2_nearest2) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     auto output_size = tensor(batch(b), feature(f), spatial(x*2, y*2));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::nearest;
@@ -1098,7 +1093,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_nearest2) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     float answers[96] = {
          0.f,  0.f,  1.f,  1.f,
@@ -1147,19 +1142,19 @@ TEST(resample_gpu, interpolate_in2x2x3x2_nearest3) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     auto output_size = tensor(batch(b), feature(f), spatial(x*2, y*2));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::nearest;
@@ -1187,7 +1182,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_nearest3) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     float answers[96] = {
          0.f,  0.f,  1.f,  1.f,
@@ -1236,19 +1231,19 @@ TEST(resample_gpu, interpolate_in2x2x3x2_nearest4) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     auto output_size = tensor(batch(b), feature(f), spatial(x*2, y*2));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::nearest;
@@ -1276,7 +1271,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_nearest4) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     float answers[96] = {
          0.f,  0.f,  0.f,  1.f,
@@ -1325,19 +1320,19 @@ TEST(resample_gpu, interpolate_in2x2x3x2_nearest5) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     auto output_size = tensor(batch(b), feature(f), spatial(x*2, y*2));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::nearest;
@@ -1365,7 +1360,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_nearest5) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     float answers[96] = {
          0.f,  0.f,  0.f,  1.f,
@@ -1414,21 +1409,21 @@ TEST(resample_gpu, interpolate_in2x2x3x2_coord_transform_mode1) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     y = 2;
     x = 3;
     auto output_size = tensor(batch(b), feature(f), spatial(x, y));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::nearest;
@@ -1456,7 +1451,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_coord_transform_mode1) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          0.f,  0.f,  1.f,
@@ -1483,21 +1478,21 @@ TEST(resample_gpu, interpolate_in2x2x3x2_coord_transform_mode2) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     y = 1;
     x = 3;
     auto output_size = tensor(batch(b), feature(f), spatial(x, y));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::nearest;
@@ -1525,7 +1520,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_coord_transform_mode2) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          0.f,  0.f,  1.f,
@@ -1546,21 +1541,21 @@ TEST(resample_gpu, interpolate_in2x2x3x2_coord_transform_mode3) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     y = 2;
     x = 3;
     auto output_size = tensor(batch(b), feature(f), spatial(x, y));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::nearest;
@@ -1588,7 +1583,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_coord_transform_mode3) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          0.f,  1.f,  1.f,
@@ -1615,21 +1610,21 @@ TEST(resample_gpu, interpolate_in2x2x3x2_coord_transform_mode4) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     y = 2;
     x = 3;
     auto output_size = tensor(batch(b), feature(f), spatial(x, y));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::nearest;
@@ -1657,7 +1652,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_coord_transform_mode4) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          2.f,  3.f,  3.f,
@@ -1684,21 +1679,21 @@ TEST(resample_gpu, interpolate_in2x2x3x2_coord_transform_mode5) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     y = 2;
     x = 3;
     auto output_size = tensor(batch(b), feature(f), spatial(x, y));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::nearest;
@@ -1726,7 +1721,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_coord_transform_mode5) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          0.f,  0.f,  1.f,
@@ -1753,21 +1748,21 @@ TEST(resample_gpu, interpolate_in2x2x3x2_cubic) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     y = 2;
     x = 3;
     auto output_size = tensor(batch(b), feature(f), spatial(x, y));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::cubic;
@@ -1793,7 +1788,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_cubic) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          0.29600694f,  0.8828125f,  1.46961806f,
@@ -1820,20 +1815,20 @@ TEST(resample_gpu, interpolate_in2x2x3x2_cubic2) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 1;
     int f = 1;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     x = 3;
     auto output_size = tensor(batch(b), feature(f), spatial(x, y));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::cubic;
@@ -1853,7 +1848,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_cubic2) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
           5.34722222f,  3.f, 0.65277778f,
@@ -1872,21 +1867,21 @@ TEST(resample_gpu, interpolate_in2x2x3x2_linear) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 2;
     int f = 2;
     int y = 3;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     y = 2;
     x = 3;
     auto output_size = tensor(batch(b), feature(f), spatial(x, y));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::caffe_bilinear;
@@ -1912,7 +1907,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_linear) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          0.5f,  1.f,  1.5f,
@@ -1939,21 +1934,21 @@ TEST(resample_gpu, interpolate_in2x2x3x2_linear_onnx) {
     //  Output : 2x2x6x4
     //  Sample Type: Nearest
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 1;
     int f = 1;
     int y = 2;
     int x = 2;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     y = 4;
     x = 4;
     auto output_size = tensor(batch(b), feature(f), spatial(x, y));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::bilinear;
@@ -1974,7 +1969,7 @@ TEST(resample_gpu, interpolate_in2x2x3x2_linear_onnx) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
              1.f, 1.33333f, 1.66667f,      2.f,
@@ -1994,21 +1989,21 @@ TEST(resample_gpu, interpolate_in1x1x2x4_linear_scale) {
     //  Output : 1x1x1x2
     //  Sample Type: Linear
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     int b = 1;
     int f = 1;
     int y = 2;
     int x = 4;
     tensor shape = tensor{batch(b), feature(f), spatial(x, y)};
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, shape });
 
     y = 1;
     x = 2;
     auto output_size = tensor(batch(b), feature(f), spatial(x, y));
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     int32_t antialias = 0;
     float cube_coeff = -0.75f;
     resample_type mode = resample_type::caffe_bilinear;
@@ -2031,7 +2026,7 @@ TEST(resample_gpu, interpolate_in1x1x2x4_linear_scale) {
     auto outputs = net.execute();
 
     auto output = outputs.at("interpolate").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> answers = {
          2.6666665f,  4.3333331f
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/reshape_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/reshape_gpu_test.cpp
index 69ad0f2be1f1af..ff35e58e75eb7f 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/reshape_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/reshape_gpu_test.cpp
@@ -3,19 +3,15 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
 
-#include <api/data.hpp>
-#include <api/reshape.hpp>
-#include <api/input_layout.hpp>
+#include "test_utils.h"
 
-#include "test_utils/test_utils.h"
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/reshape.hpp>
+#include <cldnn/primitives/input_layout.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 using namespace testing;
 
 void verify_float(const float& output_value, const float& value) {
@@ -30,7 +26,7 @@ template <class ElemType>
 void generic_reshape_test(format fmt, tensor const& input_size, tensor const& reshape_size,
     bool /* in_place */, padding const& input_padd = padding(),
     padding const& output_padd = padding()) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     //allocate input memory
     auto data_type = data_types::f32;
@@ -43,10 +39,10 @@ void generic_reshape_test(format fmt, tensor const& input_size, tensor const& re
     else if (std::is_same<ElemType, int64_t>::value)
         data_type = data_types::i64;
 
-    auto input = memory::allocate(engine, {data_type, fmt, input_size});
+    auto input = engine.allocate_memory({data_type, fmt, input_size});
 
     {
-        auto input_ptr = input.cldnn::memory::pointer<ElemType>();
+        cldnn::mem_lock<ElemType> input_ptr(input, get_test_stream());
         auto input_itr = input_ptr.begin();
 
         auto elements = input_size.count();
@@ -59,9 +55,9 @@ void generic_reshape_test(format fmt, tensor const& input_size, tensor const& re
     topology tpl;
     std::string reshape_input = "input";
 
-    tpl.add(input_layout("input", input.get_layout()));
+    tpl.add(input_layout("input", input->get_layout()));
     if (input_padd) {
-        auto padded_input_layout = input.get_layout();
+        auto padded_input_layout = input->get_layout();
         padded_input_layout.data_padding = input_padd;
         tpl.add(reorder("reorder", "input", padded_input_layout));
         reshape_input = "reorder";
@@ -79,15 +75,15 @@ void generic_reshape_test(format fmt, tensor const& input_size, tensor const& re
     auto net_input = outputs.at(reshape_input).get_memory();
     auto output = outputs.at("reshape").get_memory();
 
-    EXPECT_TRUE(output.get_layout().data_type == input.get_layout().data_type);        //reshape should not change data_type
-    EXPECT_TRUE(output.get_layout().format.value == input.get_layout().format.value);  //reshape should not change format
+    EXPECT_TRUE(output->get_layout().data_type == input->get_layout().data_type);        //reshape should not change data_type
+    EXPECT_TRUE(output->get_layout().format.value == input->get_layout().format.value);  //reshape should not change format
 
     //output size should be equal to requested plus output padding
-    ASSERT_TRUE(output.get_layout().size == reshape_size);
-    ASSERT_TRUE(output.get_layout().get_buffer_size() == reshape_size.add(output_padd.lower_size()).add(output_padd.upper_size()));
+    ASSERT_TRUE(output->get_layout().size == reshape_size);
+    ASSERT_TRUE(output->get_layout().get_buffer_size() == reshape_size.add(output_padd.lower_size()).add(output_padd.upper_size()));
 
     {
-        auto output_ptr = output.pointer<const ElemType>();
+        cldnn::mem_lock<const ElemType> output_ptr(output, get_test_stream());
         auto output_itr = output_ptr.begin();
 
         auto sizes = reshape_size.sizes(fmt);
@@ -439,15 +435,15 @@ TEST(reshape_gpu_f32, multiple_users_with_reorder) {
     //  b1f0:  0.0
     //  b1f1:  4.0
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto batch_num = 2;
     auto feature_num = 2;
     auto x_size = 1;
     auto y_size = 1;
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num))}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num))}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(activation("relu", "input", activation_func::relu));
     topology.add(reshape("reshape", "relu", tensor(batch(4))));
     topology.add(reorder("reorder1", "reshape", format::yxfb, data_types::f32));
@@ -464,13 +460,13 @@ TEST(reshape_gpu_f32, multiple_users_with_reorder) {
     auto outputs = network.execute();
 
     auto output = outputs.at("relu1").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < out1.size(); i++)
         EXPECT_EQ(output_ptr[i], out1[i]);
 
     auto output_2 = outputs.at("relu2").get_memory();
-    auto output_ptr_2 = output_2.pointer<float>();
+    cldnn::mem_lock<float> output_ptr_2(output_2, get_test_stream());
 
     for (size_t i = 0; i < out2.size(); i++)
         EXPECT_EQ(output_ptr_2[i], out2[i]);
@@ -487,12 +483,12 @@ TEST(reshape_gpu_f32, calc_output_shape) {
     //
     // output_shape (1, 1, 1, 4)
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {2, 2, 1, 1}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {2, 2, 1, 1}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reshape("reshape", "input", tensor(1, 1, 0, -1)));
 
     set_values(input, {-1.f, 2.f, -3.f, 4.f});
@@ -506,14 +502,14 @@ TEST(reshape_gpu_f32, calc_output_shape) {
 
     auto output = outputs.at("reshape").get_memory();
 
-    EXPECT_TRUE(output.get_layout().data_type == input.get_layout().data_type);
-    EXPECT_TRUE(output.get_layout().format == input.get_layout().format);
+    EXPECT_TRUE(output->get_layout().data_type == input->get_layout().data_type);
+    EXPECT_TRUE(output->get_layout().format == input->get_layout().format);
 
-    ASSERT_TRUE(output.get_layout().size == tensor(1, 1, 1, 4));
+    ASSERT_TRUE(output->get_layout().size == tensor(1, 1, 1, 4));
 
     float answers[4] = {-1.f, 2.f, -3.f, 4.f};
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     for (int i = 0; i < 4; i++) {
         EXPECT_TRUE(are_equal(answers[i], output_ptr[i]));
     }
@@ -523,12 +519,12 @@ TEST(reshape_gpu_f32, basic_bfwzyx) {
     // input:  bfwzyx, (3, 3, 2, 2, 1, 1)
     // reshape: (1, 1, 2, 2, 3, 3), pad (0, 0, 0, 0, 0, 1)
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, layout{data_types::f32, format::bfwzyx, tensor{batch(3), feature(3), spatial(1, 1, 2, 2)}});
+    auto input = engine.allocate_memory(layout{data_types::f32, format::bfwzyx, tensor{batch(3), feature(3), spatial(1, 1, 2, 2)}});
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(reshape("reshape", "input", tensor(batch(1), feature(1), spatial(2, 2, 3, 3)), padding({0, 0, 0, 0, 0, 1}, 0.f)));
 
     // clang-format off
@@ -573,10 +569,10 @@ TEST(reshape_gpu_f32, basic_bfwzyx) {
 
     auto output = outputs.at("reshape").get_memory();
 
-    EXPECT_TRUE(output.get_layout().data_type == input.get_layout().data_type);
-    EXPECT_TRUE(output.get_layout().format == input.get_layout().format);
+    EXPECT_TRUE(output->get_layout().data_type == input->get_layout().data_type);
+    EXPECT_TRUE(output->get_layout().format == input->get_layout().format);
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     ASSERT_EQ(output_ptr.size(), expected_out.size());
 
     for (size_t i = 0; i < expected_out.size(); i++) {
@@ -585,14 +581,14 @@ TEST(reshape_gpu_f32, basic_bfwzyx) {
 }
 
 TEST(reshape_gpu_f32, shrink_chain_partial) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     auto batch_num = 2;
     auto feature_num = 2;
     auto x_size = 1;
     auto y_size = 1;
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num))}});
-    auto scale_in = memory::allocate(engine, {data_types::f32, format::bfyx, { tensor(feature(4)) }});
-    auto shift_in = memory::allocate(engine, {data_types::f32, format::bfyx, { tensor(feature(4)) }});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {tensor(spatial(x_size, y_size), feature(feature_num), batch(batch_num))}});
+    auto scale_in = engine.allocate_memory({data_types::f32, format::bfyx, { tensor(feature(4)) }});
+    auto shift_in = engine.allocate_memory({data_types::f32, format::bfyx, { tensor(feature(4)) }});
 
     std::vector<float> scale_vals = {0.f, 1.f, 2.f, 3.f};
     std::vector<float> scale_shifts = {5.f, 10.f, 15.f, 20.0f};
@@ -600,7 +596,7 @@ TEST(reshape_gpu_f32, shrink_chain_partial) {
     set_values(shift_in, scale_shifts);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("scale_in", scale_in));
     topology.add(data("shift_in", shift_in));
     topology.add(activation("relu", "input", activation_func::relu));
@@ -621,17 +617,17 @@ TEST(reshape_gpu_f32, shrink_chain_partial) {
     auto outputs = network.execute();
 
     auto output = outputs.at("out_reorder").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < out.size(); i++)
         EXPECT_EQ(output_ptr[i], out[i]) << " i=" << i;
 }
 
 TEST(reshape_gpu_f32, shrink_chain_full) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, { tensor(feature(4)) }});
-    auto scale_in = memory::allocate(engine, {data_types::f32, format::bfyx, { tensor(feature(4)) }});
-    auto shift_in = memory::allocate(engine, {data_types::f32, format::bfyx, { tensor(feature(4)) }});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, { tensor(feature(4)) }});
+    auto scale_in = engine.allocate_memory({data_types::f32, format::bfyx, { tensor(feature(4)) }});
+    auto shift_in = engine.allocate_memory({data_types::f32, format::bfyx, { tensor(feature(4)) }});
 
     std::vector<float> scale_vals = {0.f, 1.f, 2.f, 3.f};
     std::vector<float> scale_shifts = {5.f, 10.f, 15.f, 20.0f};
@@ -639,7 +635,7 @@ TEST(reshape_gpu_f32, shrink_chain_full) {
     set_values(shift_in, scale_shifts);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("scale_in", scale_in));
     topology.add(data("shift_in", shift_in));
     topology.add(activation("relu", "input", activation_func::relu));
@@ -660,17 +656,17 @@ TEST(reshape_gpu_f32, shrink_chain_full) {
     auto outputs = network.execute();
 
     auto output = outputs.at("out_reorder").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < out.size(); i++)
         EXPECT_EQ(output_ptr[i], out[i]) << " i=" << i;
 }
 
 TEST(reshape_gpu_f32, shrink_chain_out) {
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, { tensor(feature(4)) }});
-    auto scale_in = memory::allocate(engine, {data_types::f32, format::bfyx, { tensor(feature(4)) }});
-    auto shift_in = memory::allocate(engine, {data_types::f32, format::bfyx, { tensor(feature(4)) }});
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, { tensor(feature(4)) }});
+    auto scale_in = engine.allocate_memory({data_types::f32, format::bfyx, { tensor(feature(4)) }});
+    auto shift_in = engine.allocate_memory({data_types::f32, format::bfyx, { tensor(feature(4)) }});
 
     std::vector<float> scale_vals = {0.f, 1.f, 2.f, 3.f};
     std::vector<float> scale_shifts = {5.f, 10.f, 15.f, 20.0f};
@@ -678,7 +674,7 @@ TEST(reshape_gpu_f32, shrink_chain_out) {
     set_values(shift_in, scale_shifts);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(activation("relu", "input", activation_func::relu));
     topology.add(reshape("reshape", "relu", tensor(spatial(2, 2))));
     topology.add(reorder("reorder", "reshape", format::bfyx, data_types::f32));
@@ -695,7 +691,7 @@ TEST(reshape_gpu_f32, shrink_chain_out) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reshape1").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < out.size(); i++)
         EXPECT_EQ(output_ptr[i], out[i]) << " i=" << i;
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/reverse_sequence_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/reverse_sequence_gpu_test.cpp
index b7de3b16fa6d52..cb674704b29f75 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/reverse_sequence_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/reverse_sequence_gpu_test.cpp
@@ -3,25 +3,22 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/reverse_sequence.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/reverse_sequence.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
 
 TEST(reverese_sequence_gpu_test, fp32_d2_2_ba1_sa0) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
     int32_t batch_axis = 1;
     int32_t seq_axis = 0;
 
@@ -34,8 +31,8 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_ba1_sa0) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -48,7 +45,7 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_ba1_sa0) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 3.0f, 2.0f, 1.0f
@@ -60,10 +57,10 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_ba1_sa0) {
 }
 
 TEST(reverese_sequence_gpu_test, fp32_d3_3_3_ba0_sa1) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 3, 1, 3 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 3, 1, 3 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
     int32_t batch_axis = 0;
     int32_t seq_axis = 1;
 
@@ -78,8 +75,8 @@ TEST(reverese_sequence_gpu_test, fp32_d3_3_3_ba0_sa1) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -92,7 +89,7 @@ TEST(reverese_sequence_gpu_test, fp32_d3_3_3_ba0_sa1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             3.0f, 4.0f, 5.0f, 0.0f, 1.0f, 2.0f, 6.0f, 7.0f, 8.0f,
@@ -106,10 +103,10 @@ TEST(reverese_sequence_gpu_test, fp32_d3_3_3_ba0_sa1) {
 }
 
 TEST(reverese_sequence_gpu_test, fp32_d3_3_3_ba2_sa0) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 3, 1, 3 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 3, 1, 3 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
     int32_t batch_axis = 2;
     int32_t seq_axis = 0;
 
@@ -124,8 +121,8 @@ TEST(reverese_sequence_gpu_test, fp32_d3_3_3_ba2_sa0) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -138,11 +135,11 @@ TEST(reverese_sequence_gpu_test, fp32_d3_3_3_ba2_sa0) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-            9.0f, 10.0f, 11.0f, 12.0f, 13.0f, 14.0f, 15.0f, 16.0f, 17.0f, 
-            0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f, 
+            9.0f, 10.0f, 11.0f, 12.0f, 13.0f, 14.0f, 15.0f, 16.0f, 17.0f,
+            0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f,
             18.0f, 19.0f, 20.0f, 21.0f, 22.0f, 23.0f, 24.0f, 25.0f, 26.0f
     };
 
@@ -152,10 +149,10 @@ TEST(reverese_sequence_gpu_test, fp32_d3_3_3_ba2_sa0) {
 }
 
 TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba0_sa3) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 3 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 3 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
     int32_t batch_axis = 0;
     int32_t seq_axis = 3;
 
@@ -170,8 +167,8 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba0_sa3) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -184,12 +181,12 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba0_sa3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-            0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 
-            6.0f, 7.0f, 8.0f, 9.0f, 10.0f, 11.0f, 
-            13.0f, 12.0f, 15.0f, 14.0f, 17.0f, 16.0f, 
+            0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f,
+            6.0f, 7.0f, 8.0f, 9.0f, 10.0f, 11.0f,
+            13.0f, 12.0f, 15.0f, 14.0f, 17.0f, 16.0f,
             19.0f, 18.0f, 21.0f, 20.0f, 23.0f, 22.0f
     };
 
@@ -199,10 +196,10 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba0_sa3) {
 }
 
 TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba0_sa2) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 3 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 3 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
     int32_t batch_axis = 0;
     int32_t seq_axis = 2;
 
@@ -217,8 +214,8 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba0_sa2) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -231,12 +228,12 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba0_sa2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-            2.0f, 3.0f, 0.0f, 1.0f, 4.0f, 5.0f, 
-            8.0f, 9.0f, 6.0f, 7.0f, 10.0f, 11.0f, 
-            14.0f, 15.0f, 12.0f, 13.0f, 16.0f, 17.0f, 
+            2.0f, 3.0f, 0.0f, 1.0f, 4.0f, 5.0f,
+            8.0f, 9.0f, 6.0f, 7.0f, 10.0f, 11.0f,
+            14.0f, 15.0f, 12.0f, 13.0f, 16.0f, 17.0f,
             20.0f, 21.0f, 18.0f, 19.0f, 22.0f, 23.0f
     };
 
@@ -246,10 +243,10 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba0_sa2) {
 }
 
 TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba2_sa0) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 3 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 3 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
     int32_t batch_axis = 2;
     int32_t seq_axis = 0;
 
@@ -264,8 +261,8 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba2_sa0) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -278,12 +275,12 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba2_sa0) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-            0.0f, 1.0f, 2.0f, 3.0f, 16.0f, 17.0f, 
-            6.0f, 7.0f, 8.0f, 9.0f, 22.0f, 23.0f, 
-            12.0f, 13.0f, 14.0f, 15.0f, 4.0f, 5.0f, 
+            0.0f, 1.0f, 2.0f, 3.0f, 16.0f, 17.0f,
+            6.0f, 7.0f, 8.0f, 9.0f, 22.0f, 23.0f,
+            12.0f, 13.0f, 14.0f, 15.0f, 4.0f, 5.0f,
             18.0f, 19.0f, 20.0f, 21.0f, 10.0f, 11.0f
     };
 
@@ -293,10 +290,10 @@ TEST(reverese_sequence_gpu_test, fp32_d2_2_3_2ba2_sa0) {
 }
 
 TEST(reverese_sequence_gpu_test, fp16_d2_2_ba1_sa0) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
     int32_t batch_axis = 1;
     int32_t seq_axis = 0;
 
@@ -309,8 +306,8 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_ba1_sa0) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -323,7 +320,7 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_ba1_sa0) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 3.0f, 2.0f, 1.0f
@@ -335,10 +332,10 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_ba1_sa0) {
 }
 
 TEST(reverese_sequence_gpu_test, fp16x2_d2_2_ba1_sa0) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } });
     int32_t batch_axis = 1;
     int32_t seq_axis = 0;
 
@@ -351,8 +348,8 @@ TEST(reverese_sequence_gpu_test, fp16x2_d2_2_ba1_sa0) {
         });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
         reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -365,7 +362,7 @@ TEST(reverese_sequence_gpu_test, fp16x2_d2_2_ba1_sa0) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 3.0f, 2.0f, 1.0f
@@ -377,10 +374,10 @@ TEST(reverese_sequence_gpu_test, fp16x2_d2_2_ba1_sa0) {
 }
 
 TEST(reverese_sequence_gpu_test, fp16_d3_3_3_ba0_sa1) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 3, 1, 3 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 3, 1, 3 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
     int32_t batch_axis = 0;
     int32_t seq_axis = 1;
 
@@ -395,8 +392,8 @@ TEST(reverese_sequence_gpu_test, fp16_d3_3_3_ba0_sa1) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -409,7 +406,7 @@ TEST(reverese_sequence_gpu_test, fp16_d3_3_3_ba0_sa1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<uint16_t >();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             3.0f, 4.0f, 5.0f, 0.0f, 1.0f, 2.0f, 6.0f, 7.0f, 8.0f,
@@ -423,10 +420,10 @@ TEST(reverese_sequence_gpu_test, fp16_d3_3_3_ba0_sa1) {
 }
 
 TEST(reverese_sequence_gpu_test, fp16_d3_3_3_ba2_sa0) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 3, 1, 3 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 3, 1, 3 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
     int32_t batch_axis = 2;
     int32_t seq_axis = 0;
 
@@ -441,8 +438,8 @@ TEST(reverese_sequence_gpu_test, fp16_d3_3_3_ba2_sa0) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -455,7 +452,7 @@ TEST(reverese_sequence_gpu_test, fp16_d3_3_3_ba2_sa0) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             9.0f, 10.0f, 11.0f, 12.0f, 13.0f, 14.0f, 15.0f, 16.0f, 17.0f,
@@ -469,10 +466,10 @@ TEST(reverese_sequence_gpu_test, fp16_d3_3_3_ba2_sa0) {
 }
 
 TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba0_sa3) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 2, 3 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 2, 3 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
     int32_t batch_axis = 0;
     int32_t seq_axis = 3;
 
@@ -487,8 +484,8 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba0_sa3) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -501,7 +498,7 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba0_sa3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f,
@@ -516,10 +513,10 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba0_sa3) {
 }
 
 TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba0_sa2) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 2, 3 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 2, 3 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 1 } });
     int32_t batch_axis = 0;
     int32_t seq_axis = 2;
 
@@ -534,8 +531,8 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba0_sa2) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -548,7 +545,7 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba0_sa2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             2.0f, 3.0f, 0.0f, 1.0f, 4.0f, 5.0f,
@@ -563,10 +560,10 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba0_sa2) {
 }
 
 TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba2_sa0) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 2, 3 } });
-    auto seq_lengths = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 2, 3 } });
+    auto seq_lengths = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 1, 1, 1 } });
     int32_t batch_axis = 2;
     int32_t seq_axis = 0;
 
@@ -581,8 +578,8 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba2_sa0) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("seq_lengths", seq_lengths.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("seq_lengths", seq_lengths->get_layout()));
     topology.add(
             reverse_sequence("reverse_sequence", "input", "seq_lengths", seq_axis, batch_axis)
     );
@@ -595,7 +592,7 @@ TEST(reverese_sequence_gpu_test, fp16_d2_2_3_2ba2_sa0) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reverse_sequence").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 1.0f, 2.0f, 3.0f, 16.0f, 17.0f,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/scale_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/scale_gpu_test.cpp
deleted file mode 100644
index 1e0359f1dada2e..00000000000000
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/scale_gpu_test.cpp
+++ /dev/null
@@ -1,1994 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/scale.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include "api/reorder.hpp"
-#include "api/data.hpp"
-
-#include <iostream>
-
-using namespace cldnn;
-using namespace tests;
-
-TEST(scale_gpu, basic_in2x3x2x2_mixed_types_in_fp32_out_fp16) {
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 3, 2 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
-    auto shift_input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 1 } });
-
-    std::vector<float> input_vec = { 1.0f, 0.0f,   5.0f,   1.5f,   2.0f,   0.0f,
-                                     6.0f, 5.0f, -10.0f, -11.0f, -12.0f, -13.0f,
-
-                                     3.0f, 0.5f,   7.0f,  12.0f,   4.0f,  -0.5f,
-                                     8.0f, 8.0f, -14.0f, -15.0f, -16.0f, -17.0f };
-    set_values(input, input_vec);
-    set_values(scale_input, { 2.0f, -1.0f });
-    set_values(shift_input, { -5.0f, 10.0f });
-
-    std::vector<float> result_vec = { -3.0f, -5.0f,  5.0f, -2.0f, -1.0f, -5.0f,
-                                       4.0f,  5.0f, 20.0f, 21.0f, 22.0f, 23.0f,
-
-                                       1.0f, -4.0f,  9.0f, 19.0f , 3.0f, -6.0f,
-                                       2.0f,  2.0f, 24.0f, 25.0f, 26.0f, 27.0f };
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(data("scale_input", scale_input));
-    topology.add(data("shift_input", shift_input));
-    topology.add(scale("scale", "input", "scale_input", "shift_input", optional_data_type{data_types::f16}));
-    topology.add(reorder("reorder", "scale", format::bfyx, data_types::f32));
-
-    build_options bo;
-    bo.set_option(build_option::optimize_data(true));
-    network network(engine, topology, bo);
-
-    network.set_input_data("input", input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("reorder").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    ASSERT_EQ(result_vec.size(), output.count());
-
-    for (unsigned int i = 0; i < result_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], result_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_mixed_types_in_fp16_out_fp32) {
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 3, 2 } });
-    auto scale_input = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 2, 1, 1 } });
-    auto shift_input = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 2, 1, 1 } });
-
-    std::vector<half_t> input_vec = { half_t(1.0f), half_t(0.0f), half_t(5.0f),   half_t(1.5f),   half_t(2.0f),   half_t(0.0f),
-                                      half_t(6.0f), half_t(5.0f), half_t(-10.0f), half_t(-11.0f), half_t(-12.0f), half_t(-13.0f),
-
-                                      half_t(3.0f), half_t(0.5f), half_t(  7.0f), half_t(12.0f),  half_t(4.0f),   half_t(-0.5f),
-                                      half_t(8.0f), half_t(8.0f), half_t(-14.0f), half_t(-15.0f), half_t(-16.0f), half_t(-17.0f) };
-    set_values(input, input_vec);
-    set_values(scale_input, { half_t(2.0f), half_t(-1.0f) });
-    set_values(shift_input, { half_t(-5.0f), half_t(10.0f) });
-
-    std::vector<float> result_vec = { -3.0f, -5.0f,  5.0f, -2.0f, -1.0f, -5.0f,
-                                       4.0f,  5.0f, 20.0f, 21.0f, 22.0f, 23.0f,
-
-                                       1.0f, -4.0f,  9.0f, 19.0f , 3.0f, -6.0f,
-                                       2.0f,  2.0f, 24.0f, 25.0f, 26.0f, 27.0f };
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(data("scale_input", scale_input));
-    topology.add(data("shift_input", shift_input));
-    topology.add(scale("scale", "input", "scale_input", "shift_input", optional_data_type{data_types::f32}));
-    topology.add(reorder("reorder", "scale", format::bfyx, data_types::f32));
-
-    build_options bo;
-    bo.set_option(build_option::optimize_data(true));
-    network network(engine, topology, bo);
-
-    network.set_input_data("input", input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("reorder").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    ASSERT_EQ(result_vec.size(), output.count());
-
-    for (unsigned int i = 0; i < result_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], result_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_mixed_types_in_fp32_scale_fp16_out_fp16) {
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 3, 2 } });
-    auto scale_input = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 2, 1, 1 } });
-    auto shift_input = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 2, 1, 1 } });
-
-    std::vector<float> input_vec = { 1.0f, 0.0f,   5.0f,   1.5f,   2.0f,   0.0f,
-                                     6.0f, 5.0f, -10.0f, -11.0f, -12.0f, -13.0f,
-
-                                     3.0f, 0.5f,   7.0f,  12.0f,   4.0f,  -0.5f,
-                                     8.0f, 8.0f, -14.0f, -15.0f, -16.0f, -17.0f };
-    set_values(input, input_vec);
-    set_values(scale_input, { half_t(2.0f), half_t(-1.0f) });
-    set_values(shift_input, { half_t(-5.0f), half_t(10.0f) });
-
-    std::vector<float> result_vec = { -3.0f, -5.0f,  5.0f, -2.0f, -1.0f, -5.0f,
-                                       4.0f,  5.0f, 20.0f, 21.0f, 22.0f, 23.0f,
-
-                                       1.0f, -4.0f,  9.0f, 19.0f , 3.0f, -6.0f,
-                                       2.0f,  2.0f, 24.0f, 25.0f, 26.0f, 27.0f };
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(data("scale_input", scale_input));
-    topology.add(data("shift_input", shift_input));
-    topology.add(scale("scale", "input", "scale_input", "shift_input", optional_data_type{data_types::f16}));
-    topology.add(reorder("reorder", "scale", format::bfyx, data_types::f32));
-
-    build_options bo;
-    bo.set_option(build_option::optimize_data(true));
-    network network(engine, topology, bo);
-
-    network.set_input_data("input", input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("reorder").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    ASSERT_EQ(result_vec.size(), output.count());
-
-    for (unsigned int i = 0; i < result_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], result_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_same_size) {
-    //  Scale  : 2x3x2x2
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale:
-    //  f0: b0:  0.1    0.2  0.25   b1:   0.3   0.4   0.5
-    //  f0: b0:  0.6    0.7  0.75   b1:   0.8   0.9   1
-    //  f1: b0:  1.1    1.2  1.25   b1:   1.3   1.4   1.5
-    //  f1: b0:  1.6    1.7  1.75   b1:   1.8   1.9   2
-
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 3, 2 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 3, 2 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f, 0.3f, 1.1f, 1.3f,
-        0.2f, 0.4f, 1.2f, 1.4f,
-        0.25f, 0.5f, 1.25f, 1.5f,
-        0.6f, 0.8f, 1.6f, 1.8f,
-        0.7f, 0.9f, 1.7f, 1.9f,
-        0.75f, 1.f, 1.75f, 2.f
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (unsigned int i = 0; i < input_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], input_vec[i] * scale_input_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_same_size_bfyx) {
-    //  Scale  : 2x3x2x2
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale:
-    //  f0: b0:  0.1    0.2  0.25   b1:   0.3   0.4   0.5
-    //  f0: b0:  0.6    0.7  0.75   b1:   0.8   0.9   1
-    //  f1: b0:  1.1    1.2  1.25   b1:   1.3   1.4   1.5
-    //  f1: b0:  1.6    1.7  1.75   b1:   1.8   1.9   2
-
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 3, 2 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 3, 2 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 2.f, -10.f, 0.f, 0.f, -11.f,
-        3.f, 4.f, -14.f, 0.5f, -0.5f, -15.f,
-        5.f, 6.f, -12.f, 1.5f, 5.2f, -13.f,
-        7.f, 8.f, -16.f, 12.f, 8.f, -17.f
-    };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f, 0.2f, 0.25f, 0.3f, 0.4f, 0.5f,
-        0.6f, 0.7f, 0.75f, 0.8f, 0.9f, 1.f,
-        1.1f, 1.2f, 1.25f, 1.3f, 1.4f, 1.5f,
-        1.6f, 1.7f, 1.75f, 1.8f, 1.9f, 2.f
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (unsigned int i = 0; i < input_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], input_vec[i] * scale_input_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_same_size_scale_bfyx) {
-    //  Scale  : 2x3x2x2
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale:
-    //  f0: b0:  0.1    0.2  0.25   b1:   0.3   0.4   0.5
-    //  f0: b0:  0.6    0.7  0.75   b1:   0.8   0.9   1
-    //  f1: b0:  1.1    1.2  1.25   b1:   1.3   1.4   1.5
-    //  f1: b0:  1.6    1.7  1.75   b1:   1.8   1.9   2
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto x_size = 3;
-    auto y_size = 2;
-
-    auto input = memory::allocate(engine, { data_types::f32,format::yxfb,{ batch_num, feature_num, x_size, y_size } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size, y_size } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f, 0.2f, 0.25f, 0.3f, 0.4f, 0.5f,
-        0.6f, 0.7f, 0.75f, 0.8f, 0.9f, 1.f,
-        1.1f, 1.2f, 1.25f, 1.3f, 1.4f, 1.5f,
-        1.6f, 1.7f, 1.75f, 1.8f, 1.9f, 2.f
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int j = 0; j < feature_num; ++j) { //F
-        for (int i = 0; i < batch_num; ++i) { //B
-            for (int k = 0; k < y_size; ++k) { //Y
-                for (int l = 0; l < x_size; ++l) { //X
-                    int linear_id = i + batch_num * (j + feature_num * (l + x_size * k));
-                    int linear_id_scale = l + x_size * (k + y_size * (j + i * feature_num));
-                    EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05F);
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_same_size_bias_term) {
-    //  Scale  : 2x3x2x2
-    //  Bias   : 2x3x2x2
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale:
-    //  f0: b0:  0.1    0.2  0.25   b1:   0.3   0.4   0.5
-    //  f0: b0:  0.6    0.7  0.75   b1:   0.8   0.9   1
-    //  f1: b0:  1.1    1.2  1.25   b1:   1.3   1.4   1.5
-    //  f1: b0:  1.6    1.7  1.75   b1:   1.8   1.9   2
-    //
-    //  Bias:
-    //  f0: b0:  1.1    1.2  1.25   b1:   1.3   1.4   1.5
-    //  f0: b0:  2.6    2.7  2.75   b1:   2.8   2.9   2
-    //  f1: b0:  3.1    3.2  3.25   b1:   3.3   3.4   3.5
-    //  f1: b0:  4.6    4.7  4.75   b1:   4.8   4.9   4
-
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 3, 2 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 3, 2 } });
-    auto bias = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 3, 2 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(input_layout("bias", bias.get_layout()));
-    topology.add(scale("scale", "input", "scale_input", "bias"));
-
-    std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f, 0.3f, 1.1f, 1.3f,
-        0.2f, 0.4f, 1.2f, 1.4f,
-        0.25f, 0.5f, 1.25f, 1.5f,
-        0.6f, 0.8f, 1.6f, 1.8f,
-        0.7f, 0.9f, 1.7f, 1.9f,
-        0.75f, 1.f, 1.75f, 2.f
-    };
-    set_values(scale_input, scale_input_vec);
-
-    std::vector<float> bias_vec = {
-        1.1f, 2.3f, 3.1f, 4.3f,
-        1.2f, 2.4f, 3.2f, 4.4f,
-        1.25f, 2.5f, 3.25f, 4.5f,
-        1.6f, 2.8f, 3.6f, 4.8f,
-        1.7f, 2.9f, 3.7f, 4.9f,
-        1.75f, 2.f, 3.75f, 4.f
-    };
-    set_values(bias, bias_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-    network.set_input_data("bias", bias);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (unsigned int i = 0; i < input_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], input_vec[i] * scale_input_vec[i] + bias_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_scalar) {
-    //  Scale  : 1
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale:
-    //  0.1    0.2
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto x_size = 3;
-    auto y_size = 2;
-
-    auto input = memory::allocate(engine, { data_types::f32,format::yxfb,{ batch_num, feature_num, x_size, y_size } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 1, 1 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f,
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int j = 0; j < feature_num; ++j) { //F
-        for (int i = 0; i < batch_num; ++i) { //B
-            for (int k = 0; k < y_size; ++k) { //Y
-                for (int l = 0; l < x_size; ++l) { //X
-                    int linear_id = i + batch_num * (j + feature_num * (l + x_size * k));
-                    int linear_id_scale = 0;
-                    EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05F);
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_y) {
-    //  Scale  : 2
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale:
-    //  0.1    0.2
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto x_size = 3;
-    auto y_size = 2;
-
-    auto input = memory::allocate(engine, { data_types::f32,format::yxfb,{ batch_num, feature_num, x_size, y_size } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1,1,1,y_size } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f,
-        0.2f,
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int j = 0; j < feature_num; ++j) { //F
-        for (int i = 0; i < batch_num; ++i) { //B
-            for (int k = 0; k < y_size; ++k) { //Y
-                for (int l = 0; l < x_size; ++l) { //X
-                    int linear_id = i + batch_num * (j + feature_num * (l + x_size * k));
-                    int linear_id_scale = k;
-                    EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05F);
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_fb) {
-    //  Scale  : 2x3x2x2
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale: per feature per batch
-    //  f0b0: 0.1   f0b1: 0.2
-    //  f1b0: 0.5   f1b1: 2.0
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto x_size = 3;
-    auto y_size = 2;
-
-    auto input = memory::allocate(engine, { data_types::f32,format::yxfb,{ batch_num, feature_num, x_size, y_size } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb,{ batch_num, feature_num, 1, 1 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f, 0.2f, 0.5f, 2.0f,
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int j = 0; j < feature_num; ++j) { //F
-        for (int i = 0; i < batch_num; ++i) { //B
-            for (int k = 0; k < y_size; ++k) { //Y
-                for (int l = 0; l < x_size; ++l) { //X
-                    int linear_id = i + batch_num * (j + feature_num * (l + x_size * k));
-                    int linear_id_scale = i + feature_num * j;
-                    EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05F);
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_f) {
-    //  Scale  : 2x3x2x2
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale: per feature
-    //  f0bx: 0.1   f1bx: 0.2
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto x_size = 3;
-    auto y_size = 2;
-
-    auto input = memory::allocate(engine, { data_types::f32,format::yxfb,{ batch_num, feature_num, x_size, y_size } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, feature_num, 1, 1 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        //f0bx  //f1bx
-        0.1f,   0.2f
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int j = 0; j < feature_num; ++j) { //F
-        for (int i = 0; i < batch_num; ++i) { //B
-            for (int k = 0; k < y_size; ++k) { //Y
-                for (int l = 0; l < x_size; ++l) { //X
-                    int linear_id = i + batch_num * (j + feature_num * (l + x_size * k));
-                    int linear_id_scale = j;
-                    EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05F);
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_x) {
-    //  Scale  : 3
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale:
-    //  0.1    0.2  0.25
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto x_size = 3;
-    auto y_size = 2;
-
-    auto input = memory::allocate(engine, { data_types::f32,format::yxfb,{ batch_num, feature_num, x_size, y_size } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, x_size, 1 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f,
-        0.2f,
-        0.25f
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int j = 0; j < feature_num; ++j) { //F
-        for (int i = 0; i < batch_num; ++i) { //B
-            for (int k = 0; k < y_size; ++k) { //Y
-                for (int l = 0; l < x_size; ++l) { //X
-                    int linear_id = i + batch_num * (j + feature_num * (l + x_size * k));
-                    int linear_id_scale = l;
-                    EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05F);
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_xy) {
-    //  Scale  : 2x3x1x1
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale:
-    //  f0:  0.1    0.2  0.25
-    //  f0:  0.6    0.7  0.75
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto x_size = 3;
-    auto y_size = 2;
-
-    auto input = memory::allocate(engine, { data_types::f32,format::yxfb,{ batch_num, feature_num, x_size, y_size } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, x_size, y_size } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f,
-        0.2f,
-        0.25f,
-        0.6f,
-        0.7f,
-        0.75f
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int j = 0; j < feature_num; ++j) { //F
-        for (int i = 0; i < batch_num; ++i) { //B
-            for (int k = 0; k < y_size; ++k) { //Y
-                for (int l = 0; l < x_size; ++l) { //X
-                    int linear_id = i + batch_num * (j + feature_num * (l + x_size * k));
-                    int linear_id_scale = l + x_size * k;
-                    EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05F);
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_batch1) {
-    //  Scale  : 2x3x2x1
-    //  Input  : 2x3x2x2
-    //  Output : 2x3x2x2
-
-    //  Input:
-    //  f0: b0:  1    2  -10   b1:   0    0    -11
-    //  f0: b0:  3    4  -14   b1:   0.5 -0.5  -15
-    //  f1: b0:  5    6  -12   b1:   1.5  5.2  -13
-    //  f1: b0:  7    8  -16   b1:   12   8    -17
-    //
-    //  Scale:
-    //  f0: b0:  0.1    0.2  0.25
-    //  f0: b0:  0.6    0.7  0.75
-    //  f1: b0:  1.1    1.2  1.25
-    //  f1: b0:  1.6    1.7  1.75
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto x_size = 3;
-    auto y_size = 2;
-
-    auto input = memory::allocate(engine, { data_types::f32,format::yxfb,{ batch_num, feature_num, x_size, y_size } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, feature_num, x_size, y_size } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f, 1.1f,
-        0.2f, 1.2f,
-        0.25f, 1.25f,
-        0.6f, 1.6f,
-        0.7f, 1.7f,
-        0.75f, 1.75f
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int j = 0; j < feature_num; ++j) { //F
-        for (int i = 0; i < batch_num; ++i) { //B
-            for (int k = 0; k < y_size; ++k) { //Y
-                for (int l = 0; l < x_size; ++l) { //X
-                    int linear_id = i + batch_num * (j + feature_num * (l + x_size * k));
-                    int linear_id_scale = j + feature_num * (l + x_size * k);
-                    EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05F);
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x3_scale_same_size_bx) {
-    //  Scale  : 2x3
-    //  Bias   : 2x3
-    //  Input  : 2x3
-    //  Output : 2x3
-
-    //  Input:
-    //  b0: 1  2  -0.75
-    //  b1: 0 -1.5  -3
-    //
-    //  Scale:
-    //  b0: 3.1  0.2  0.17
-    //  b1:  10   -3     1
-
-    //  Bias:
-    //  b0: -0.1 3.2  7
-    //  b1: 0    1   -1
-
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 3, 1 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 3, 1 } });
-    auto bias_input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 3, 1 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(input_layout("bias_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input", "bias_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 2.f, -0.75f,
-        0.f, -1.5f, -3.f,
-    };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_vec = {
-        3.1f, 0.2f, 0.17f,
-        10.f, -3.f, 1.f,
-    };
-    set_values(scale_input, scale_vec);
-
-    std::vector<float> bias_vec = {
-        -0.1f, 3.2f, 7.f,
-        0.f, 1.f, -1.f,
-    };
-    set_values(bias_input, bias_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-    network.set_input_data("bias_input", bias_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (unsigned int i = 0; i < input_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], input_vec[i] * scale_vec[i] + bias_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3_scale_same_size_xb) {
-    //  Scale  : 2x3
-    //  Bias   : 2x3
-    //  Input  : 2x3
-    //  Output : 2x3
-
-    //  Input:
-    //  x0: 1     2  -0.75
-    //  x1: 0  -1.5     -3
-    //
-    //  Scale:
-    //  x0: 3.1   0.2  0.17
-    //  x1: 10     -3     1
-
-    //  Bias:
-    //  x0: -0.1  3.2   7
-    //  x1: 0       1  -1
-
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 3, 1, 2, 1 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb, { 3, 1, 2, 1 } });
-    auto bias_input = memory::allocate(engine, { data_types::f32, format::yxfb, { 3, 1, 2, 1 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(input_layout("bias_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input", "bias_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 2.f, -0.75f,
-        0.f, -1.5f, -3.f,
-    };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_vec = {
-        3.1f, 0.2f, 0.17f,
-        10.f, -3.f, 1.f,
-    };
-    set_values(scale_input, scale_vec);
-
-    std::vector<float> bias_vec = {
-        -0.1f, 3.2f, 7.f,
-        0.f, 1.f, -1.f,
-    };
-    set_values(bias_input, bias_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-    network.set_input_data("bias_input", bias_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (unsigned int i = 0; i < input_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], input_vec[i] * scale_vec[i] + bias_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3_scale_single_value_bx) {
-    //  Scale  : 1x1
-    //  Bias   : 1x1
-    //  Input  : 2x3
-    //  Output : 2x3
-
-    //  Input:
-    //  b0: 1    2 -0.75
-    //  b1: 0 -1.5    -3
-    //
-    //  Scale:
-    //  3.1
-
-    //  Bias:
-    //  -0.1
-
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 3, 1 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
-    auto bias_input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(input_layout("bias_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input", "bias_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 2.f, -0.75f,
-        0.f, -1.5f, -3.f,
-    };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_vec = {
-        3.1f,
-    };
-    set_values(scale_input, scale_vec);
-
-    std::vector<float> bias_vec = {
-        -0.1f,
-    };
-    set_values(bias_input, bias_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-    network.set_input_data("bias_input", bias_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (unsigned int i = 0; i < input_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], input_vec[i] * scale_vec[0] + bias_vec[0], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3_scale_single_value_xb) {
-    //  Scale  : 1x1
-    //  Bias   : 1x1
-    //  Input  : 2x3
-    //  Output : 2x3
-
-    //  Input:
-    //  x0: 1     2 -0.75
-    //  x1: 0  -1.5    -3
-    //
-    //  Scale:
-    //  3.1
-
-    //  Bias:
-    //  -0.1
-
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 3, 1, 2, 1 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 1, 1 } });
-    auto bias_input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 1, 1 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(input_layout("bias_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input", "bias_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 2.f, -0.75f,
-        0.f, -1.5f, -3.f,
-    };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_vec = {
-        3.1f,
-    };
-    set_values(scale_input, scale_vec);
-
-    std::vector<float> bias_vec = {
-        -0.1f,
-    };
-    set_values(bias_input, bias_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-    network.set_input_data("bias_input", bias_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (unsigned int i = 0; i < input_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], input_vec[i] * scale_vec[0] + bias_vec[0], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3_scale_same_size_no_bias_bx) {
-    //  Scale  : 2x3
-    //  Input  : 2x3
-    //  Output : 2x3
-
-    //  Input:
-    //  b0: 1    2 -0.75
-    //  b1: 0 -1.5    -3
-    //
-    //  Scale:
-    //  b0: 3.1   0.2   0.17
-    //  b1: 10     -3      1
-
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 3, 1 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 3, 1 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 2.f, -0.75f,
-        0.f, -1.5f, -3.f,
-    };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_vec = {
-        3.1f, 0.2f, 0.17f,
-        10.f, -3.f, 1.f,
-    };
-    set_values(scale_input, scale_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (unsigned int i = 0; i < input_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], input_vec[i] * scale_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3_scale_same_size_no_bias_xb) {
-    //  Scale  : 2x3
-    //  Input  : 2x3
-    //  Output : 2x3
-
-    //  Input:
-    //  x0: 1     2  -0.75
-    //  x1: 0  -1.5     -3
-    //
-    //  Scale:
-    //  x0: 3.1    0.2   0.17
-    //  x1: 10      -3      1
-
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 3, 1, 2, 1 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::yxfb, { 3, 1, 2, 1 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 2.f, -0.75f,
-        0.f, -1.5f, -3.f,
-    };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_vec = {
-        3.1f, 0.2f, 0.17f,
-        10.f, -3.f, 1.f,
-    };
-    set_values(scale_input, scale_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (unsigned int i = 0; i < input_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], input_vec[i] * scale_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x3x2x2_scale_yxfb_bfyx_same_size_padding) {
-    //  Scale  : 2x2x1x1
-    //  Input  : 2x2x1x1
-    //  Output : 2x2x1x1
-    //  Output Padding: 2x2
-    //  Input Padding: 2x1 (with reorder)
-
-    //  Input:
-    //  1    2
-    //  3    4
-
-    //
-    //  Scale:
-    //  0.1    0.2
-    //  0.6    0.5
-
-    const auto& engine = get_test_engine();
-    std::vector<format> formats_to_test = { format::yxfb , format::bfyx };
-
-    for (std::vector<format>::iterator it = formats_to_test.begin(); it != formats_to_test.end(); ++it)
-    {
-        std::cout << "Testing format: " << format::order(*it) << std::endl;
-
-        tensor input_tensor(1, 1, 2, 2);
-
-        auto input = memory::allocate(engine, { data_types::f32, *it, input_tensor });
-        auto scale_input = memory::allocate(engine, { data_types::f32, *it, input_tensor });
-
-        topology topology;
-        topology.add(input_layout("input", input.get_layout()));
-        topology.add(reorder("reorder", "input", input.get_layout().with_padding(padding{ { 0, 0, 1, 2 }, 0 })));
-        topology.add(input_layout("scale_input", scale_input.get_layout()));
-        topology.add(scale("scale", "reorder", "scale_input", {}, padding( { 0, 0, 2, 2 }, 0 )));
-
-        std::vector<float> input_vec = { 1.f, 2.f, 3.f, 4.f };
-        set_values(input, input_vec);
-
-        std::vector<float> scale_input_vec = { 0.1f, 0.2f, 0.6f, 0.5f };
-        set_values(scale_input, scale_input_vec);
-
-        std::vector<float> expected = {
-            0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
-            0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
-            0.f, 0.f, 0.1f, 0.4f, 0.f, 0.f,
-            0.f, 0.f, 1.8f, 2.0f, 0.f, 0.f,
-            0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
-            0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
-        };
-
-        network network(engine, topology);
-
-        network.set_input_data("input", input);
-        network.set_input_data("scale_input", scale_input);
-
-        auto outputs = network.execute();
-
-        auto output = outputs.at("scale").get_memory();
-        auto output_ptr = output.pointer<float>();
-
-        for (unsigned int i = 0; i < expected.size(); ++i) {
-            EXPECT_NEAR(output_ptr[i], expected[i], 1e-05F);
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x2x2x3x2_scale_same_size_bfzyx) {
-    //  Scale  : 2x2x2x3x2
-    //  Input  : 2x2x2x3x2
-    //  Output : 2x2x2x3x2
-
-    const auto& engine = get_test_engine();
-
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 2, 2, 3, 2 } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 2, 2, 2, 3, 2 } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 2.f, -10.f, 0.f, 0.f, -11.f,
-        3.f, 4.f, -14.f, 0.5f, -0.5f, -15.f,
-        5.f, 6.f, -12.f, 1.5f, 5.2f, -13.f,
-        7.f, 8.f, -16.f, 12.f, 8.f, -17.f,
-        1.f, 2.f, -10.f, 0.f, 0.f, -11.f,
-        3.f, 4.f, -14.f, 0.5f, -0.5f, -15.f,
-        5.f, 6.f, -12.f, 1.5f, 5.2f, -13.f,
-        7.f, 8.f, -16.f, 12.f, 8.f, -17.f
-    };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f, 0.2f, 0.25f, 0.3f, 0.4f, 0.5f,
-        0.6f, 0.7f, 0.75f, 0.8f, 0.9f, 1.f,
-        1.1f, 1.2f, 1.25f, 1.3f, 1.4f, 1.5f,
-        1.6f, 1.7f, 1.75f, 1.8f, 1.9f, 2.f,
-        0.1f, 0.2f, 0.25f, 0.3f, 0.4f, 0.5f,
-        0.6f, 0.7f, 0.75f, 0.8f, 0.9f, 1.f,
-        1.1f, 1.2f, 1.25f, 1.3f, 1.4f, 1.5f,
-        1.6f, 1.7f, 1.75f, 1.8f, 1.9f, 2.f
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (unsigned int i = 0; i < input_vec.size(); ++i) {
-        EXPECT_NEAR(output_ptr[i], input_vec[i] * scale_input_vec[i], 1e-05F);
-    }
-}
-
-TEST(scale_gpu, basic_in2x2x2x2x3_scale_z) {
-    //  Scale  : 2
-    //  Input  : 2x2x2x2x3
-    //  Output : 2x2x2x2x3
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto z_size = 2;
-    auto y_size = 2;
-    auto x_size = 3;
-
-    auto input = memory::allocate(engine, { data_types::f32,format::bfzyx,{ batch_num, feature_num, x_size, y_size, z_size } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1,1,1,1,z_size } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f,
-        1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f
-    };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f,
-        0.2f,
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int i = 0; i < batch_num; ++i) { //B
-        for (int j = 0; j < feature_num; ++j) { //F
-            for (int m = 0; m < z_size; ++m) { //Z
-                for (int k = 0; k < y_size; ++k) { //Y
-                    for (int l = 0; l < x_size; ++l) { //X
-                        int linear_id = l + x_size * (k + y_size * (m + z_size * (j + feature_num*i)));
-                        int linear_id_scale = m;
-                        EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05F);
-                    }
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x2x2x2x3_scale_xyz) {
-    //  Scale  : 1x1x2x2x3
-    //  Input  : 2x2x2x2x3
-    //  Output : 2x2x2x2x3
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto z_size = 2;
-    auto y_size = 2;
-    auto x_size = 3;
-
-    auto input = memory::allocate(engine, { data_types::f32,format::bfzyx,{ batch_num, feature_num, x_size, y_size, z_size } });
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1,1, x_size, y_size, z_size } });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = {
-        1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f,
-        1.f, 0.f, 5.f, 1.5f,
-        2.f, 0.f, 6.f, 5.2f,
-        -10.f, -11.f, -12.f, -13.f,
-        3.f, 0.5f, 7.f, 12.f,
-        4.f, -0.5f, 8.f, 8.f,
-        -14.f, -15.f, -16.f, -17.f
-    };
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = {
-        0.1f,
-        0.2f,
-        0.25f,
-        0.6f,
-        0.7f,
-        0.75f,
-        0.1f,
-        0.2f,
-        0.25f,
-        0.6f,
-        0.7f,
-        0.75f
-    };
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int i = 0; i < batch_num; ++i) { //B
-        for (int j = 0; j < feature_num; ++j) { //F
-            for (int m = 0; m < z_size; ++m) { //Z
-                for (int k = 0; k < y_size; ++k) { //Y
-                    for (int l = 0; l < x_size; ++l) { //X
-                        int linear_id = l + x_size * (k + y_size * (m + z_size * (j + feature_num*i)));
-                        int linear_id_scale = l + x_size * (k + y_size * m);
-                        EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05F);
-                    }
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x2x2x2x2x3_scale_4d) {
-    //  Scale  : 1x2x1x1
-    //  Input  : 2x2x2x2x2x3
-    //  Output : 2x2x2x2x2x3
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto w_size = 2;
-    auto z_size = 2;
-    auto y_size = 2;
-    auto x_size = 3;
-
-    tensor in_size = tensor(format::bfwzyx, { batch_num, feature_num, x_size, y_size, z_size, w_size });
-    tensor scale_size = tensor(format::bfyx, { 1, feature_num, 1, 1 });
-
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx,  in_size});
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfyx, scale_size });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = generate_random_1d<float>(in_size.count(), -10, 10);
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = generate_random_1d<float>(scale_input.count(), -10, 10);
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int b = 0; b < batch_num; ++b) {
-        for (int f = 0; f < feature_num; ++f) {
-            for (int w = 0; w < w_size; ++w) {
-                for (int z = 0; z < z_size; ++z) {
-                    for (int y = 0; y < y_size; ++y) {
-                        for (int x = 0; x < x_size; ++x) {
-                            int linear_id = x + x_size * (y + y_size * (z + z_size * (w + w_size * (f + feature_num * b))));
-                            int linear_id_scale = f;
-                            EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05f);
-                        }
-                    }
-                }
-            }
-        }
-    }
-}
-
-TEST(scale_gpu, basic_in2x2x2x2x2x3_scale_6d) {
-    //  Scale  : 1x2x1x1x1x1
-    //  Input  : 2x2x2x2x2x3
-    //  Output : 2x2x2x2x2x3
-
-    const auto& engine = get_test_engine();
-
-    auto batch_num = 2;
-    auto feature_num = 2;
-    auto w_size = 2;
-    auto z_size = 2;
-    auto y_size = 2;
-    auto x_size = 3;
-
-    tensor in_size = tensor(format::bfwzyx, { batch_num, feature_num, x_size, y_size, z_size, w_size });
-    tensor scale_size = tensor(format::bfwzyx, { 1, feature_num, 1, 1, 1, 1 });
-
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx,  in_size});
-    auto scale_input = memory::allocate(engine, { data_types::f32, format::bfwzyx, scale_size });
-
-    topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input", scale_input.get_layout()));
-    topology.add(scale("scale", "input", "scale_input"));
-
-    std::vector<float> input_vec = generate_random_1d<float>(in_size.count(), -10, 10);
-    set_values(input, input_vec);
-
-    std::vector<float> scale_input_vec = generate_random_1d<float>(scale_input.count(), -10, 10);
-    set_values(scale_input, scale_input_vec);
-
-    network network(engine, topology);
-
-    network.set_input_data("input", input);
-    network.set_input_data("scale_input", scale_input);
-
-    auto outputs = network.execute();
-
-    auto output = outputs.at("scale").get_memory();
-    auto output_ptr = output.pointer<float>();
-
-    for (int b = 0; b < batch_num; ++b) {
-        for (int f = 0; f < feature_num; ++f) {
-            for (int w = 0; w < w_size; ++w) {
-                for (int z = 0; z < z_size; ++z) {
-                    for (int y = 0; y < y_size; ++y) {
-                        for (int x = 0; x < x_size; ++x) {
-                            int linear_id = x + x_size * (y + y_size * (z + z_size * (w + w_size * (f + feature_num * b))));
-                            int linear_id_scale = f;
-                            EXPECT_NEAR(output_ptr[linear_id], input_vec[linear_id] * scale_input_vec[linear_id_scale], 1e-05f);
-                        }
-                    }
-                }
-            }
-        }
-    }
-}
-
-//////////////////////////////////////////////////////////////////////////////
-//                                                                          //
-//                      Exhaustive Negative Matrix tests                    //
-//                                                                          //
-//////////////////////////////////////////////////////////////////////////////
-
-//TODO: this should be done using TEST_P or some equivallent construct
-static network setup_scale_network(
-    const data_types dt,
-    const tensor input_tensor,
-    const tensor scale_tensor,
-    const tensor bias_tensor,
-    const format f,
-    const format of,
-    bool pass_bias          //TODO: a WA for lack of std::optional<tensor> bias
-)
-{
-    const auto& engine = get_test_engine();
-    topology topology;
-
-    auto input_mem = memory::allocate(engine, { dt, f, input_tensor });
-    auto scale_mem = memory::allocate(engine, { dt, of, scale_tensor });
-    topology.add(input_layout("input", input_mem.get_layout()));
-    topology.add(input_layout("scale_input", scale_mem.get_layout()));
-
-    if (pass_bias)
-    {
-        auto bias_mem = memory::allocate(engine, { dt, f, bias_tensor });
-        topology.add(input_layout("bias_input", bias_mem.get_layout()));
-
-        topology.add(scale("scale", "input", "scale_input", "bias_input" ));
-    }
-    else
-    {
-        topology.add(scale("scale", "input", "scale_input"));
-    }
-//TODO: this will be supported after the API change
-//    else
-//    {
-//        assert(!pass_bias);
-//
-//        topology.add(scale("scale", "input", "scale_input"));
-//    }
-
-    return network(engine, topology);
-}
-
-TEST(NegativeScaleTest, TestAll) {
-    auto d = data_types::f32;
-    auto f = format::bfyx;
-    auto of = format::yxfb;
-
-    std::vector<int> t { 3, 4, 5, 6 };
-    std::vector<int> t2 { 5, 6, 4, 3 };
-
-    // broadcast rules mean that either the dim size is equal to input dim or is 1
-    std::vector<std::vector<int>> good_ts =
-    {
-        { 1, 4, 5, 6 }, { 3, 1, 5, 6 }, { 3, 4, 1, 6 }, { 3, 4, 5, 1 },
-        { 1, 1, 5, 6 }, { 1, 4, 1, 6 }, { 1, 4, 5, 1 }, { 3, 1, 1, 6 }, { 3, 1, 5, 1 }, { 3, 4, 1, 1 },
-        { 1, 1, 1, 6 }, { 1, 1, 5, 1 }, { 1, 4, 1, 1 }, { 3, 1, 1, 1 }
-    };
-    std::vector<std::vector<int>> bad_ts = { { 2, 4, 5, 6 }, { 3, 2, 5, 6 }, { 3, 4, 2, 6 }, { 3, 4, 5, 2 } };
-
-    //TODO: should be ASSERT_THROW(statement, exception_type) - but what exception type?
-    ASSERT_ANY_THROW(setup_scale_network(d, tensor{ }, tensor{ }, tensor{ }, f, of, false));
-    ASSERT_ANY_THROW(setup_scale_network(d, tensor{ }, tensor{ }, tensor{ }, f, of, true));
-
-    ASSERT_ANY_THROW(setup_scale_network(d, tensor(t), tensor(t2), tensor(t), f, of, true));
-    ASSERT_ANY_THROW(setup_scale_network(d, tensor(t), tensor(t2), tensor(t), f, of, false));
-
-    // make sure that it's the input that's masked in the scale/bias with a "1", not ther other way around
-    for (const auto & good : good_ts)
-    {
-        ASSERT_ANY_THROW(setup_scale_network(d, tensor(good), tensor(t), tensor(t), f, of, true));
-    }
-
-    // sizes must either be equal to input or at most have
-    for (const auto & bad : bad_ts)
-    {
-        ASSERT_ANY_THROW(setup_scale_network(d, tensor(t), tensor(bad), tensor(t), f, of, true));
-        ASSERT_ANY_THROW(setup_scale_network(d, tensor(t), tensor(t), tensor(bad), f, of, true));
-
-        for (const auto & good : good_ts)
-        {
-            ASSERT_ANY_THROW(setup_scale_network(d, tensor(t), tensor(bad), tensor(good), f, of, true));
-            ASSERT_ANY_THROW(setup_scale_network(d, tensor(t), tensor(good), tensor(bad), f, of, true));
-        }
-    }
-
-    // we expect the broadcast mask to be identical for scale and bias, when present
-    for (unsigned i = 0; i < good_ts.size(); ++i)
-    for (unsigned j = 0; j < good_ts.size(); ++j)
-        if (i != j)
-        {
-            ASSERT_ANY_THROW(setup_scale_network(d, tensor(t), tensor(good_ts[i]), tensor(good_ts[j]), f, of, true));
-        }
-
-}
-
-//////////////////////////////////////////////////////////////////////////////
-//                                                                          //
-//                      Exhaustive Positive Matrix tests                    //
-//                                                                          //
-//////////////////////////////////////////////////////////////////////////////
-
-using namespace cldnn;
-
-class scale_test : public tests::generic_test
-{
-public:
-    static void TearDownTestCase()
-    {
-        all_generic_params.clear();
-        all_layer_params.clear();
-    }
-
-    //TODO: use an enum instead of int i
-    static std::vector<std::shared_ptr<cldnn::primitive>> generate_specific_test_params(int variant)
-    {
-        std::vector<std::shared_ptr<cldnn::primitive>> all_layer_params;
-
-        switch(variant)
-        {
-            case 0: all_layer_params.emplace_back(new scale("scale", "input0", "input1")); break;
-            case 1: all_layer_params.emplace_back(new scale("scale", "input0", "input1", "input2")); break;
-                    //    case 3: all_layer_params.push_back(new scale("scale", "input0", "input1", true));    // This case should be checked by negative_scale_test
-                    //    case 4: all_layer_params.push_back(new scale("scale", "input0", "input1", false));    // This case should be checked by negative_scale_test
-            default: assert(0);
-        }
-
-        return all_layer_params;
-    }
-
-    static std::vector<std::shared_ptr<tests::test_params>> generate_generic_test_params(int variant)
-    {
-        assert(!variant || variant == 1);
-
-        std::vector<std::shared_ptr<tests::test_params>> all_generic_params;
-
-        auto data_types = test_data_types();
-
-        for (cldnn::data_types dt : data_types)
-        for (tensor & t : test_input_sizes)
-        {
-            std::vector<std::vector<int>> attempted_dims;
-
-            for (int32_t b : test_batch_sizes)
-            for (auto f : test_feature_sizes)
-            for (int mask = 0; mask < 16; ++mask)    //TODO: do we want to restrict it to some smaller subset like for (auto mask : { 0, 1, 3, 7, 15, 5, 10})? the problem is that because of the layout we might miss some interesting combinations since this is effectively hardcoded int he kernels
-            {
-                const int w = t.spatial[0];
-                const int h = t.spatial[1];
-
-                const auto mb = mask & 0x8 ? b : 1;
-                const auto mf = mask & 0x4 ? f : 1;
-                const auto mh = mask & 0x2 ? h : 1;
-                const auto mw = mask & 0x1 ? w : 1;
-
-                // avoid adding test cases with different masks leading to the same dimensions
-                if(attempted_dims.end() == std::find_if(attempted_dims.begin(), attempted_dims.end(), [=](const std::vector<int> & arr) { return arr[0] == mb && arr[1] == mf && arr[2] == mh && arr[3] == mw; }))
-                {
-                    std::vector<int> tmp { mb, mf, mh, mw };
-                    attempted_dims.push_back(tmp);
-
-                    std::shared_ptr<test_params> tp = std::make_shared<test_params>();
-                    tp->data_type = dt;
-
-                    tp->input_layouts.push_back(cldnn::layout(tp->data_type, tp->fmt, cldnn::tensor(  b, f, w, h  )));
-                    tp->input_layouts.push_back(cldnn::layout(tp->data_type, tp->fmt, cldnn::tensor(  mb, mf, mw, mh  )));
-                    if (variant)
-                            tp->input_layouts.push_back(cldnn::layout(tp->data_type, tp->fmt, cldnn::tensor(  mb, mf, mw, mh  )));
-
-                    all_generic_params.emplace_back(tp);
-                }
-            }
-        }
-
-        return all_generic_params;
-    }
-
-    static std::vector<std::tuple<std::shared_ptr<tests::test_params>, std::shared_ptr<cldnn::primitive>>> generate_all_test_params()
-    {
-        std::vector<std::tuple<std::shared_ptr<tests::test_params>, std::shared_ptr<cldnn::primitive>>> res;
-
-        for (int variant = 0; variant <= 1; ++variant)
-        {
-            auto tpv = generate_generic_test_params(variant);
-            auto pv = generate_specific_test_params(variant);
-
-            for (auto & tp : tpv)
-                all_generic_params.emplace_back(tp);
-
-            for (auto & p : pv)
-                all_layer_params.emplace_back(p);
-
-            for (auto & tp : tpv)
-            for (auto & p: pv)
-                res.emplace_back(tp, p);
-        }
-
-        return res;
-    }
-
-    virtual bool is_format_supported(cldnn::format format) override
-    {
-        return format == cldnn::format::bfyx;
-    }
-
-    template<typename Type>
-    memory generate_reference_typed(const std::vector<memory> & inputs)
-    {
-        assert(inputs.size() == 3 || inputs.size() == 2);
-        const bool bias_input_present = inputs.size() == 3;
-
-        const memory & input = inputs[0];
-        const memory & scale = inputs[1];
-        const memory * bias = bias_input_present ? &inputs[2] : nullptr;
-        assert(!bias_input_present || bias);
-
-        //Output is bfyx
-        auto output = memory::allocate(engine, cldnn::layout(input.get_layout().data_type, cldnn::format::bfyx, input.get_layout().size ));
-
-        const auto in0_mem = input.pointer<Type>();
-        const auto in1_mem = scale.pointer<Type>();
-        const auto in2_mem_ptr = bias ? std::make_shared<pointer<Type>>(*bias) : nullptr;
-        const Type * const in2_mem = in2_mem_ptr ? in2_mem_ptr->data() : nullptr; //TODO: is the condition needed or is it nullptr anyway?
-        auto out_mem = output.pointer<Type>();
-
-        const auto input_sizes = input.get_layout().size.sizes(cldnn::format::bfyx);
-
-        const int in0_b = input_sizes[0];
-        const int in0_f = input_sizes[1];
-        const int in0_h = input_sizes[2];
-        const int in0_w = input_sizes[3];
-
-        { // asserting dims
-            const auto output_sizes = output.get_layout().size.sizes(cldnn::format::bfyx);
-            const int out_b = output_sizes[0]; (void) out_b;
-            const int out_f = output_sizes[1]; (void) out_f;
-            const int out_h = output_sizes[2]; (void) out_h;
-            const int out_w = output_sizes[3]; (void) out_w;
-
-            const auto scale_sizes = scale.get_layout().size.sizes(cldnn::format::bfyx);
-            const int in1_b = scale_sizes[0]; (void) in1_b;
-            const int in1_f = scale_sizes[1]; (void) in1_f;
-            const int in1_h = scale_sizes[2]; (void) in1_h;
-            const int in1_w = scale_sizes[3]; (void) in1_w;
-            // input and output dims must match
-            assert(in0_b == out_b && in0_f == out_f && in0_h == out_h && in0_w == out_w);
-
-            // scale/bias dims must be equal to in/out or be 1 for broadcast
-            assert(in1_b == 1 || in1_b == in0_b);
-            assert(in1_f == 1 || in1_f == in0_f);
-            assert(in1_h == 1 || in1_h == in0_h);
-            assert(in1_w == 1 || in1_w == in0_w);
-
-            if (bias)
-            {
-                const auto bias_sizes = bias->get_layout().size.sizes(cldnn::format::bfyx);
-                const int in2_b = bias_sizes[0]; (void) in2_b;
-                const int in2_f = bias_sizes[1]; (void) in2_f;
-                const int in2_h = bias_sizes[2]; (void) in2_h;
-                const int in2_w = bias_sizes[3]; (void) in2_w;
-
-                // scale/bias dims must be equal to in/out or be 1 for broadcast
-                assert(in2_b == 1 || in2_b == in1_b);
-                assert(in2_b == 1 || in2_f == in1_f);
-                assert(in2_b == 1 || in2_h == in1_h);
-                assert(in2_b == 1 || in2_w == in1_w);
-            }
-        }
-
-        const auto input_desc = get_linear_memory_desc(input.get_layout());
-        const auto output_desc = get_linear_memory_desc(output.get_layout());
-        const auto scale_desc = get_linear_memory_desc(scale.get_layout());
-        const auto bias_desc =
-            bias ?
-            get_linear_memory_desc(bias->get_layout()) :
-            memory_desc();
-
-        for (int n = 0; n < in0_b; ++n)
-        for (int c = 0; c < in0_f; ++c)
-        for (int y = 0; y < in0_h; ++y)
-        for (int x = 0; x < in0_w; ++x)
-        {
-            const size_t in0_idx = get_linear_index(input.get_layout(), n, c, y, x, input_desc);
-            const size_t in1_idx = get_linear_index_with_broadcast(scale.get_layout(), n, c, y, x, scale_desc);
-            const size_t out_idx = get_linear_index(output.get_layout(), n, c, y, x, output_desc);
-
-            out_mem[out_idx] = in0_mem[in0_idx] * in1_mem[in1_idx];
-
-            if (bias)
-            {
-                const size_t in2_idx = get_linear_index_with_broadcast(bias->get_layout(), n, c, y, x, bias_desc);
-                out_mem[out_idx] += in2_mem[in2_idx];
-            }
-        }
-        return output;
-    }
-
-    virtual memory generate_reference(const std::vector<memory> & inputs) override
-    {
-        if (generic_params->data_type == data_types::f32)
-        {
-            return generate_reference_typed<float>(inputs);
-        }
-        else
-        {
-            return generate_reference_typed<FLOAT16>(inputs);
-        }
-    }
-
-    static std::string custom_param_name(const ::testing::TestParamInfo<std::tuple<std::shared_ptr<tests::test_params>, std::shared_ptr<cldnn::primitive>>>& info)
-    {
-        std::stringstream res;
-
-        const auto & p = std::get<0>(info.param);
-
-        assert (p->data_type == data_types::f32 ||
-                p->data_type == data_types::f16);
-
-        res << info.index
-            << "_" << (p->data_type == data_types::f32 ? "f32" : "f16");
-
-        for (unsigned i = 0; i < p->input_layouts.size(); ++i)
-        {
-            if (i == 0) res << "_Input";
-            if (i == 1) res << "_ScaleInput";
-            if (i == 2) res << "_BiasInput";
-
-            const auto chans = format::traits(p->fmt).order;
-
-            for (unsigned int j = 0; j < p->input_layouts[i].size.sizes(p->fmt).size(); ++j)
-            {
-                res << chans[j] << p->input_layouts[i].size.sizes(p->fmt)[j];
-            }
-        }
-        return res.str();
-    }
-
-private:
-    static std::vector<std::shared_ptr<tests::test_params>> all_generic_params;
-    static std::vector<std::shared_ptr<cldnn::primitive>> all_layer_params;
-};
-
-std::vector<std::shared_ptr<cldnn::primitive>> scale_test::all_layer_params = {};
-std::vector<std::shared_ptr<tests::test_params>> scale_test::all_generic_params = {};
-
-TEST_P(scale_test, SCALE)
-{
-    run_single_test();
-}
-
-INSTANTIATE_TEST_CASE_P(DISABLED_SCALE,
-    scale_test,
-    ::testing::ValuesIn(scale_test::generate_all_test_params()),
-    scale_test::custom_param_name);
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_elements_update_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_elements_update_gpu_test.cpp
index e190e590241b9f..02d5089f9c57d3 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_elements_update_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_elements_update_gpu_test.cpp
@@ -2,17 +2,15 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
+#include "test_utils.h"
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/scatter_elements_update.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/scatter_elements_update.hpp>
+#include <cldnn/runtime/memory.hpp>
+#include <cldnn/graph/topology.hpp>
+#include <cldnn/graph/network.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -42,11 +40,11 @@ TEST(scatter_elements_update_gpu_fp16, d2411_axisF) {
     //  10.f, 11.f, 5.f, 4.f,
     //  1.f, 7.f, 12.f, 13.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 4, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 4, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Updates
     auto axis = cldnn::scatter_elements_update::scatter_elements_update_axis::along_f;
 
     set_values(input1, {
@@ -65,9 +63,9 @@ TEST(scatter_elements_update_gpu_fp16, d2411_axisF) {
     });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_elements_update("scatter_elements_update", "InputData", "InputIndices", "InputUpdates", axis)
     );
@@ -81,7 +79,7 @@ TEST(scatter_elements_update_gpu_fp16, d2411_axisF) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_elements_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         10.f, 11.f, 5.f, 4.f,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_nd_update_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_nd_update_gpu_test.cpp
index 26dc684f5e52be..9977777fa0e4f9 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_nd_update_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_nd_update_gpu_test.cpp
@@ -2,29 +2,27 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
+#include "test_utils.h"
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/scatter_update.hpp>
-#include <api/scatter_nd_update.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/scatter_update.hpp>
+#include <cldnn/primitives/scatter_nd_update.hpp>
+#include <cldnn/runtime/memory.hpp>
+#include <cldnn/graph/topology.hpp>
+#include <cldnn/graph/network.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
 
 
 TEST(scatter_nd_update_gpu_fp16_test15, data5_indice3_update5) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 2, 2, 4, 3 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx,  { 1, 2, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 1, 2, 2, 4, 3, 2 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 2, 2, 4, 3 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx,  { 1, 2, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 1, 2, 2, 4, 3, 2 } }); // updates
 
     set_values(input1, {
         // 0
@@ -90,9 +88,9 @@ TEST(scatter_nd_update_gpu_fp16_test15, data5_indice3_update5) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 3)
     );
@@ -107,7 +105,7 @@ TEST(scatter_nd_update_gpu_fp16_test15, data5_indice3_update5) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -115,11 +113,11 @@ TEST(scatter_nd_update_gpu_fp16_test15, data5_indice3_update5) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test14, data5_indice2_update3) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 2, 2, 4, 3 } }); // data 2x2x3x4x2 (bfzyx)
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx,  { 3, 3, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 3, 4, 1, 1, 2 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 2, 2, 4, 3 } }); // data 2x2x3x4x2 (bfzyx)
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx,  { 3, 3, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 3, 4, 1, 1, 2 } }); // updates
 
     set_values(input1, {
         // 0
@@ -174,9 +172,9 @@ TEST(scatter_nd_update_gpu_fp16_test14, data5_indice2_update3) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -191,7 +189,7 @@ TEST(scatter_nd_update_gpu_fp16_test14, data5_indice2_update3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -199,11 +197,11 @@ TEST(scatter_nd_update_gpu_fp16_test14, data5_indice2_update3) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test13, data4_indice2_update2) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 4, 2 } }); // data 2x3x2x4 (bfyx)
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 3, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 4, 1, 1 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 4, 2 } }); // data 2x3x2x4 (bfyx)
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 3, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 4, 1, 1 } }); // updates
 
     set_values(input1, {
         FLOAT16(1.0f), FLOAT16(2.0f),  FLOAT16(3.0f), FLOAT16(4.0f),       FLOAT16(1.0f), FLOAT16(2.0f),  FLOAT16(3.0f), FLOAT16(4.0f),
@@ -238,9 +236,9 @@ TEST(scatter_nd_update_gpu_fp16_test13, data4_indice2_update2) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -255,7 +253,7 @@ TEST(scatter_nd_update_gpu_fp16_test13, data4_indice2_update2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -263,11 +261,11 @@ TEST(scatter_nd_update_gpu_fp16_test13, data4_indice2_update2) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test12, data3_indice3_update1) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 3, 1, 4 } }); // data 3x3x4 (bfy)
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 3, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 3, 1, 4 } }); // data 3x3x4 (bfy)
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 3, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } }); // updates
 
     set_values(input1, {
         FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f), FLOAT16(4.0f),
@@ -309,9 +307,9 @@ TEST(scatter_nd_update_gpu_fp16_test12, data3_indice3_update1) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -326,7 +324,7 @@ TEST(scatter_nd_update_gpu_fp16_test12, data3_indice3_update1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -334,11 +332,11 @@ TEST(scatter_nd_update_gpu_fp16_test12, data3_indice3_update1) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test11, data6_indice1_update6) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 2, 3, 4, 2 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 1, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 1, 2, 2, 3, 4, 2 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 2, 3, 4, 2 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 1, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 1, 2, 2, 3, 4, 2 } }); // updates
 
     set_values(input1, {
         // 0, 0, 0
@@ -440,9 +438,9 @@ TEST(scatter_nd_update_gpu_fp16_test11, data6_indice1_update6) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -457,7 +455,7 @@ TEST(scatter_nd_update_gpu_fp16_test11, data6_indice1_update6) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -465,11 +463,11 @@ TEST(scatter_nd_update_gpu_fp16_test11, data6_indice1_update6) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test10, data5_indice1_update5) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 2, 3, 4, 2 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 2, 3, 4, 2 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 2, 3, 4, 2 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 2, 3, 4, 2 } }); // updates
 
     set_values(input1, {
         // 0
@@ -537,9 +535,9 @@ TEST(scatter_nd_update_gpu_fp16_test10, data5_indice1_update5) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -554,7 +552,7 @@ TEST(scatter_nd_update_gpu_fp16_test10, data5_indice1_update5) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -562,11 +560,11 @@ TEST(scatter_nd_update_gpu_fp16_test10, data5_indice1_update5) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test9, data4_indice1_update4) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 3, 4, 2 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 4, 2 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 3, 4, 2 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 4, 2 } }); // updates
 
     set_values(input1, {
         // 0
@@ -616,9 +614,9 @@ TEST(scatter_nd_update_gpu_fp16_test9, data4_indice1_update4) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -633,7 +631,7 @@ TEST(scatter_nd_update_gpu_fp16_test9, data4_indice1_update4) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -641,11 +639,11 @@ TEST(scatter_nd_update_gpu_fp16_test9, data4_indice1_update4) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test8, data6_indice2_update5) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 1, 2, 2, 3, 4, 2 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx,   { 2, 2, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 1, 3, 4, 2 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 1, 2, 2, 3, 4, 2 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx,   { 2, 2, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 1, 3, 4, 2 } }); // updates
 
     set_values(input1, {
         //0,0
@@ -714,9 +712,9 @@ TEST(scatter_nd_update_gpu_fp16_test8, data6_indice2_update5) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -731,7 +729,7 @@ TEST(scatter_nd_update_gpu_fp16_test8, data6_indice2_update5) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -739,11 +737,11 @@ TEST(scatter_nd_update_gpu_fp16_test8, data6_indice2_update5) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test7, data5_indice2_update4) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 1, 2, 3, 4, 2 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx,  { 2, 2, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfzyx,  { 2, 2, 1, 3, 4 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 1, 2, 3, 4, 2 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx,  { 2, 2, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfzyx,  { 2, 2, 1, 3, 4 } }); // updates
 
 
     set_values(input1, {
@@ -782,9 +780,9 @@ TEST(scatter_nd_update_gpu_fp16_test7, data5_indice2_update4) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -799,7 +797,7 @@ TEST(scatter_nd_update_gpu_fp16_test7, data5_indice2_update4) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -808,11 +806,11 @@ TEST(scatter_nd_update_gpu_fp16_test7, data5_indice2_update4) {
 
 
 TEST(scatter_nd_update_gpu_fp16_test6, data4_indice2_update3) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 4, 2 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 4, 1, 2 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 4, 2 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 4, 1, 2 } }); // updates
 
 
     set_values(input1, {
@@ -848,9 +846,9 @@ TEST(scatter_nd_update_gpu_fp16_test6, data4_indice2_update3) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -865,7 +863,7 @@ TEST(scatter_nd_update_gpu_fp16_test6, data4_indice2_update3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -873,11 +871,11 @@ TEST(scatter_nd_update_gpu_fp16_test6, data4_indice2_update3) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test5, data3_indice2_update2) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 1, 4 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 4, 1, 1 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 1, 4 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 4, 1, 1 } }); // updates
 
 
     set_values(input1, {
@@ -913,9 +911,9 @@ TEST(scatter_nd_update_gpu_fp16_test5, data3_indice2_update2) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -930,7 +928,7 @@ TEST(scatter_nd_update_gpu_fp16_test5, data3_indice2_update2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -938,11 +936,11 @@ TEST(scatter_nd_update_gpu_fp16_test5, data3_indice2_update2) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test4, data2_indice2_update1) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 4, 1, 1 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 1, 1, 1 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 4, 1, 1 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 1, 1, 1 } }); // updates
 
 
     set_values(input1, {
@@ -968,9 +966,9 @@ TEST(scatter_nd_update_gpu_fp16_test4, data2_indice2_update1) {
         };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -985,7 +983,7 @@ TEST(scatter_nd_update_gpu_fp16_test4, data2_indice2_update1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -993,11 +991,11 @@ TEST(scatter_nd_update_gpu_fp16_test4, data2_indice2_update1) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test3, data3_indice1_update3) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 3, 4, 1 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 4, 1 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 3, 4, 1 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 4, 1 } }); // updates
 
 
     set_values(input1, {
@@ -1043,9 +1041,9 @@ TEST(scatter_nd_update_gpu_fp16_test3, data3_indice1_update3) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -1060,7 +1058,7 @@ TEST(scatter_nd_update_gpu_fp16_test3, data3_indice1_update3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -1069,11 +1067,11 @@ TEST(scatter_nd_update_gpu_fp16_test3, data3_indice1_update3) {
 
 
 TEST(scatter_nd_update_gpu_fp16_test2, data2_indice1_update2) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 4, 1, 1 } }); // data
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 4, 1, 1 } }); // updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 4, 1, 1 } }); // data
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // indices
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 4, 1, 1 } }); // updates
 
 
     set_values(input1, {
@@ -1098,9 +1096,9 @@ TEST(scatter_nd_update_gpu_fp16_test2, data2_indice1_update2) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -1115,7 +1113,7 @@ TEST(scatter_nd_update_gpu_fp16_test2, data2_indice1_update2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -1123,11 +1121,11 @@ TEST(scatter_nd_update_gpu_fp16_test2, data2_indice1_update2) {
 }
 
 TEST(scatter_nd_update_gpu_fp16_test1, data1_indice1_update1) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 8, 1, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 1, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 8, 1, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 1, 1, 1 } }); // Updates
 
 
     set_values(input1, {
@@ -1147,9 +1145,9 @@ TEST(scatter_nd_update_gpu_fp16_test1, data1_indice1_update1) {
     };
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -1164,7 +1162,7 @@ TEST(scatter_nd_update_gpu_fp16_test1, data1_indice1_update1) {
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
@@ -1181,11 +1179,11 @@ TEST(scatter_nd_update_gpu_fp16, d6661_i2311) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 6, 6, 1, 6 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 6, 6, 1, 6 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // Updates
 
     set_values(input1, {
         FLOAT16(100.f), FLOAT16(101.f), FLOAT16(102.f), FLOAT16(103.f), FLOAT16(104.f), FLOAT16(105.f),
@@ -1242,9 +1240,9 @@ TEST(scatter_nd_update_gpu_fp16, d6661_i2311) {
 
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -1260,7 +1258,7 @@ TEST(scatter_nd_update_gpu_fp16, d6661_i2311) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f, 102.f, 103.f, 104.f, 105.f,
@@ -1319,11 +1317,11 @@ TEST(scatter_nd_update_gpu_fp16, d6661_i2211) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 6, 6, 1, 6 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 6, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 6, 6, 1, 6 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 6, 1, 1 } }); // Updates
 
 
     set_values(input1, {
@@ -1381,9 +1379,9 @@ TEST(scatter_nd_update_gpu_fp16, d6661_i2211) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -1399,7 +1397,7 @@ TEST(scatter_nd_update_gpu_fp16, d6661_i2211) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f, 102.f, 103.f, 104.f, 105.f,
@@ -1458,11 +1456,11 @@ TEST(scatter_nd_update_gpu_fp16, d6661_i2111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 6, 6, 1, 6 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 6, 1, 6 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 6, 6, 1, 6 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 6, 1, 6 } }); // Updates
 
 
     set_values(input1, {
@@ -1531,9 +1529,9 @@ TEST(scatter_nd_update_gpu_fp16, d6661_i2111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -1549,7 +1547,7 @@ TEST(scatter_nd_update_gpu_fp16, d6661_i2111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         777.f, 999.f, 999.f, 999.f, 999.f, 999.f,
@@ -1609,11 +1607,11 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2411) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 2, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 4, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 2, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 4, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // Updates
 
 
     set_values(input1, {
@@ -1652,9 +1650,9 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2411) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -1670,7 +1668,7 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2411) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f,
@@ -1712,11 +1710,11 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2311) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 2, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 2, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Updates
 
 
     set_values(input1, {
@@ -1755,9 +1753,9 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2311) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -1773,7 +1771,7 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2311) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f,
@@ -1815,11 +1813,11 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2211) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 2, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 1, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 2, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 1, 2 } }); // Updates
 
 
     set_values(input1, {
@@ -1864,9 +1862,9 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2211) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -1882,7 +1880,7 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2211) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f,
@@ -1924,11 +1922,11 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 3, 2, 2, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 2, 3 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 3, 2, 2, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 2, 3 } }); // Updates
 
 
     set_values(input1, {
@@ -1981,9 +1979,9 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -1999,7 +1997,7 @@ TEST(scatter_nd_update_gpu_fp16, d3232_i2111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         666.f, 666.f,
@@ -2040,11 +2038,11 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i25111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 3, 2, 3, 2, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 5, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 1, 1, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 3, 2, 3, 2, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 5, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 1, 1, 1, 1 } }); // Updates
 
 
     set_values(input1, {
@@ -2115,9 +2113,9 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i25111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -2133,7 +2131,7 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i25111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f, 102.f,
@@ -2207,11 +2205,11 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i24111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 3, 2, 3, 2, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 4, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 3, 1, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 3, 2, 3, 2, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 4, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 3, 1, 1, 1 } }); // Updates
 
 
     set_values(input1, {
@@ -2283,9 +2281,9 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i24111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -2301,7 +2299,7 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i24111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f, 102.f,
@@ -2375,11 +2373,11 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i23111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 3, 2, 3, 2, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 3, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 2, 1, 1, 3 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 3, 2, 3, 2, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 3, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 2, 1, 1, 3 } }); // Updates
 
 
     set_values(input1, {
@@ -2454,9 +2452,9 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i23111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -2472,7 +2470,7 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i23111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f, 102.f,
@@ -2546,11 +2544,11 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i22111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 3, 2, 3, 2, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 2, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 3, 1, 3, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 3, 2, 3, 2, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 2, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 3, 1, 3, 2 } }); // Updates
 
 
     set_values(input1, {
@@ -2637,9 +2635,9 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i22111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -2655,7 +2653,7 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i22111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f, 102.f,
@@ -2729,11 +2727,11 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i21111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 3, 2, 3, 2, 3 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 1, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 2, 3, 2, 3 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 3, 2, 3, 2, 3 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 1, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 2, 3, 2, 3 } }); // Updates
 
 
     set_values(input1, {
@@ -2838,9 +2836,9 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i21111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -2856,7 +2854,7 @@ TEST(scatter_nd_update_gpu_fp16, d32323_i21111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         555.f, 555.f, 555.f,
@@ -2930,12 +2928,12 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i261111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
     // memory order is bfxyzw
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 6, 1, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 1, 1, 1, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 6, 1, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 1, 1, 1, 1, 1 } }); // Updates
 
 
     set_values(input1, {
@@ -2998,9 +2996,9 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i261111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -3016,7 +3014,7 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i261111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f,
@@ -3082,12 +3080,12 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i251111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
     // memory order is bfxyzw
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 5, 1, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 1, 1, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 5, 1, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 1, 1, 1, 1 } }); // Updates
 
 
     set_values(input1, {
@@ -3151,9 +3149,9 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i251111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -3169,7 +3167,7 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i251111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f,
@@ -3235,12 +3233,12 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i241111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
     // memory order is bfxyzw
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 4, 1, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 1, 1, 1, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 4, 1, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 1, 1, 1, 2 } }); // Updates
 
 
     set_values(input1, {
@@ -3307,9 +3305,9 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i241111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -3325,7 +3323,7 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i241111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f,
@@ -3392,12 +3390,12 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i231111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
     // memory order is bfxyzw
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 3, 1, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 1, 1, 2, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 3, 1, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 1, 1, 2, 2 } }); // Updates
 
 
     set_values(input1, {
@@ -3470,9 +3468,9 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i231111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -3488,7 +3486,7 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i231111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f,
@@ -3554,12 +3552,12 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i221111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
     // memory order is bfxyzw
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 1, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 1, 2, 2, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 1, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 1, 2, 2, 2 } }); // Updates
 
 
     set_values(input1, {
@@ -3644,9 +3642,9 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i221111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -3662,7 +3660,7 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i221111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         100.f, 101.f,
@@ -3727,12 +3725,12 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i211111) {
     //  Input values in fp16
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
     // memory order is bfxyzw
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 1, 1, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 1, 1, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfwzyx, { 2, 2, 2, 2, 2, 2 } }); // Updates
 
 
     set_values(input1, {
@@ -3841,9 +3839,9 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i211111) {
         });
 
     topology topology;
-    topology.add(input_layout("InputData", input1.get_layout()));
-    topology.add(input_layout("InputIndices", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputData", input1->get_layout()));
+    topology.add(input_layout("InputIndices", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_nd_update("scatter_nd_update", "InputData", "InputIndices", "InputUpdates", 2)
     );
@@ -3859,7 +3857,7 @@ TEST(scatter_nd_update_gpu_fp16, d222222_i211111) {
 
 
     auto output = outputs.at("scatter_nd_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         777.f, 777.f,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_update_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_update_gpu_test.cpp
index 01f2c7f02ef209..937695760e3a42 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_update_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/scatter_update_gpu_test.cpp
@@ -2,17 +2,12 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
+#include "test_utils.h"
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/scatter_update.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/scatter_update.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -38,14 +33,14 @@ TEST(scatter_update_gpu_fp16, d2411_axisB) {
     //  0.f, 0.f, 0.f, 0.f
     //
     //  Output:
-    //  3.f, 6.f, 5.f, 4.f, 
+    //  3.f, 6.f, 5.f, 4.f,
     //  1.f, 7.f, 2.f, 9.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 4, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 4, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 4, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 4, 1, 1 } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_b;
 
     set_values(input1, {
@@ -63,34 +58,34 @@ TEST(scatter_update_gpu_fp16, d2411_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-    
-    
+
+    network network(engine, topology);
+
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
-    
+
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-        3.f, 6.f, 5.f, 4.f, 
+        3.f, 6.f, 5.f, 4.f,
         1.f, 7.f, 2.f, 9.f
     };
 
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
-    }    
+    }
 }
 
 TEST(scatter_update_gpu_fp32, d8111_axisB) {
@@ -114,11 +109,11 @@ TEST(scatter_update_gpu_fp32, d8111_axisB) {
     //  1.f, 11.f, 3.f, 10.f, 9.f, 6.f, 7.f, 12.f
 
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 8, 1, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 4, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx, { 4, 1, 1, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 8, 1, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 4, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx, { 4, 1, 1, 1 } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_b;
 
     set_values(input1, {
@@ -134,32 +129,32 @@ TEST(scatter_update_gpu_fp32, d8111_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-    
-    
+
+    network network(engine, topology);
+
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
-    auto outputs = network.execute();    
+
+    auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         1.f, 11.f, 3.f, 10.f, 9.f, 6.f, 7.f, 12.f
     };
-    
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], output_ptr[i]);
-    }    
+    }
 }
 
 TEST(scatter_update_gpu_fp16, d4311_axisB) {
@@ -193,11 +188,11 @@ TEST(scatter_update_gpu_fp16, d4311_axisB) {
     //  6.f, 6.f, 6.f,
     //  7.f, 7.f, 7.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 4, 3, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 3 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 4, 3, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 3 } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_b;
 
     set_values(input1, {
@@ -221,34 +216,34 @@ TEST(scatter_update_gpu_fp16, d4311_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology);     
-    
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
-    auto outputs = network.execute();   
+
+    auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-        9.f, 10.f, 11.f, 
+        9.f, 10.f, 11.f,
         8.f, 8.f, 8.f,
-        6.f, 6.f, 6.f, 
+        6.f, 6.f, 6.f,
         7.f, 7.f, 7.f
     };
-    
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
-    } 
+    }
 }
 
 TEST(scatter_update_gpu_fp16, d2521_axisF) {
@@ -302,24 +297,24 @@ TEST(scatter_update_gpu_fp16, d2521_axisF) {
     //  16.f, 17.f,
     //  141.f, 151.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 5, 1, 2 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 2, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 5, 1, 2 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 2, 2 } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_f;
 
     set_values(input1, {
-        FLOAT16(0.0f), FLOAT16(1.0f), 
-        FLOAT16(2.0f), FLOAT16(3.0f), 
+        FLOAT16(0.0f), FLOAT16(1.0f),
+        FLOAT16(2.0f), FLOAT16(3.0f),
         FLOAT16(4.0f), FLOAT16(5.0f),
-        FLOAT16(6.0f), FLOAT16(7.0f), 
+        FLOAT16(6.0f), FLOAT16(7.0f),
         FLOAT16(8.0f), FLOAT16(9.0f),
 
         FLOAT16(10.0f), FLOAT16(11.0f),
         FLOAT16(12.0f), FLOAT16(13.0f),
-        FLOAT16(14.0f), FLOAT16(15.0f), 
-        FLOAT16(16.0f), FLOAT16(17.0f), 
+        FLOAT16(14.0f), FLOAT16(15.0f),
+        FLOAT16(16.0f), FLOAT16(17.0f),
         FLOAT16(18.0f), FLOAT16(19.0f)
     });
 
@@ -329,35 +324,35 @@ TEST(scatter_update_gpu_fp16, d2521_axisF) {
     });
 
     set_values(input3, {
-        FLOAT16(21.0f), FLOAT16(31.0f), 
+        FLOAT16(21.0f), FLOAT16(31.0f),
         FLOAT16(41.0f), FLOAT16(51.0f),
-        FLOAT16(61.0f), FLOAT16(71.0f), 
+        FLOAT16(61.0f), FLOAT16(71.0f),
         FLOAT16(81.0f), FLOAT16(91.0f),
 
         FLOAT16(101.0f), FLOAT16(111.0f),
         FLOAT16(121.0f), FLOAT16(131.0f),
-        FLOAT16(141.0f), FLOAT16(151.0f), 
+        FLOAT16(141.0f), FLOAT16(151.0f),
         FLOAT16(161.0f), FLOAT16(171.0f)
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-      
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         21.f, 31.f,
@@ -372,7 +367,7 @@ TEST(scatter_update_gpu_fp16, d2521_axisF) {
         16.f, 17.f,
         141.f, 151.f
     };
-  
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
     }
@@ -415,11 +410,11 @@ TEST(scatter_update_gpu_fp16, d2241_axisY) {
     //  90.f, 120.f, 100.f, 110.f,
     //  130.f, 160.f, 140.f, 150.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 1, 4 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 2, 2, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 1, 4 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 2, 2, 2 } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_y;
 
     set_values(input1, {
@@ -435,35 +430,35 @@ TEST(scatter_update_gpu_fp16, d2241_axisY) {
     });
 
     set_values(input3, {
-        FLOAT16(0.0f), FLOAT16(20.0f), 
+        FLOAT16(0.0f), FLOAT16(20.0f),
         FLOAT16(30.0f), FLOAT16(40.0f),
-        FLOAT16(50.0f), FLOAT16(60.0f), 
+        FLOAT16(50.0f), FLOAT16(60.0f),
         FLOAT16(70.0f), FLOAT16(80.0f),
 
         FLOAT16(90.0f), FLOAT16(100.0f),
         FLOAT16(110.0f), FLOAT16(120.0f),
-        FLOAT16(130.0f), FLOAT16(140.0f), 
+        FLOAT16(130.0f), FLOAT16(140.0f),
         FLOAT16(150.0f), FLOAT16(160.0f)
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-    
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 40.f, 20.f, 30.f,
@@ -472,7 +467,7 @@ TEST(scatter_update_gpu_fp16, d2241_axisY) {
         130.f, 160.f, 140.f, 150.f
     };
 
-    
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
     }
@@ -486,11 +481,11 @@ TEST(scatter_update_gpu_fp16, d8x2x20x1_axisB) {
     //  Output : 8x2x20x1
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 8, 2, 1, 20 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 3, 1, 1 } });  // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfyx, { 2, 3, 20, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 8, 2, 1, 20 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 3, 1, 1 } });  // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfyx, { 2, 3, 20, 2 } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_b;
 
     set_values(input1, {
@@ -510,8 +505,8 @@ TEST(scatter_update_gpu_fp16, d8x2x20x1_axisB) {
         FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),
 
         FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f),
-        FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f), 
-        
+        FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),
+
         FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f),FLOAT16(0.0f), FLOAT16(0.0f),
         FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),FLOAT16(1.0f), FLOAT16(1.0f),
 
@@ -525,43 +520,43 @@ TEST(scatter_update_gpu_fp16, d8x2x20x1_axisB) {
     });
 
     set_values(input3, {
-        FLOAT16(0), FLOAT16(1), FLOAT16(2), FLOAT16(3), FLOAT16(4), FLOAT16(5), FLOAT16(6), FLOAT16(7), FLOAT16(8), FLOAT16(9), FLOAT16(10), FLOAT16(11), FLOAT16(12), FLOAT16(13), FLOAT16(14), FLOAT16(15), FLOAT16(16), FLOAT16(17), FLOAT16(18), FLOAT16(19), 
+        FLOAT16(0), FLOAT16(1), FLOAT16(2), FLOAT16(3), FLOAT16(4), FLOAT16(5), FLOAT16(6), FLOAT16(7), FLOAT16(8), FLOAT16(9), FLOAT16(10), FLOAT16(11), FLOAT16(12), FLOAT16(13), FLOAT16(14), FLOAT16(15), FLOAT16(16), FLOAT16(17), FLOAT16(18), FLOAT16(19),
         FLOAT16(20), FLOAT16(21), FLOAT16(22), FLOAT16(23), FLOAT16(24), FLOAT16(25), FLOAT16(26), FLOAT16(27), FLOAT16(28), FLOAT16(29), FLOAT16(30), FLOAT16(31), FLOAT16(32), FLOAT16(33), FLOAT16(34), FLOAT16(35), FLOAT16(36), FLOAT16(37), FLOAT16(38), FLOAT16(39),
-        
+
         FLOAT16(40), FLOAT16(41), FLOAT16(42), FLOAT16(43), FLOAT16(44), FLOAT16(45), FLOAT16(46), FLOAT16(47), FLOAT16(48), FLOAT16(49), FLOAT16(50), FLOAT16(51), FLOAT16(52), FLOAT16(53), FLOAT16(54), FLOAT16(55), FLOAT16(56), FLOAT16(57), FLOAT16(58), FLOAT16(59),
         FLOAT16(60), FLOAT16(61), FLOAT16(62), FLOAT16(63), FLOAT16(64), FLOAT16(65), FLOAT16(66), FLOAT16(67), FLOAT16(68), FLOAT16(69), FLOAT16(70), FLOAT16(71), FLOAT16(72), FLOAT16(73), FLOAT16(74), FLOAT16(75), FLOAT16(76), FLOAT16(77), FLOAT16(78), FLOAT16(79),
-        
+
         FLOAT16(80), FLOAT16(81), FLOAT16(82), FLOAT16(83), FLOAT16(84), FLOAT16(85), FLOAT16(86), FLOAT16(87), FLOAT16(88), FLOAT16(89), FLOAT16(90), FLOAT16(91), FLOAT16(92), FLOAT16(93), FLOAT16(94), FLOAT16(95), FLOAT16(96), FLOAT16(97), FLOAT16(98), FLOAT16(99),
         FLOAT16(100), FLOAT16(101), FLOAT16(102), FLOAT16(103), FLOAT16(104), FLOAT16(105), FLOAT16(106), FLOAT16(107), FLOAT16(108), FLOAT16(109), FLOAT16(110), FLOAT16(111), FLOAT16(112), FLOAT16(113), FLOAT16(114), FLOAT16(115), FLOAT16(116), FLOAT16(117), FLOAT16(118), FLOAT16(119),
-        
+
         FLOAT16(120), FLOAT16(121), FLOAT16(122), FLOAT16(123), FLOAT16(124), FLOAT16(125), FLOAT16(126), FLOAT16(127), FLOAT16(128), FLOAT16(129), FLOAT16(130), FLOAT16(131), FLOAT16(132), FLOAT16(133), FLOAT16(134), FLOAT16(135), FLOAT16(136), FLOAT16(137), FLOAT16(138), FLOAT16(139),
         FLOAT16(140), FLOAT16(141), FLOAT16(142), FLOAT16(143), FLOAT16(144), FLOAT16(145), FLOAT16(146), FLOAT16(147), FLOAT16(148), FLOAT16(149), FLOAT16(150), FLOAT16(151), FLOAT16(152), FLOAT16(153), FLOAT16(154), FLOAT16(155), FLOAT16(156), FLOAT16(157), FLOAT16(158), FLOAT16(159),
- 
+
         FLOAT16(160), FLOAT16(161), FLOAT16(162), FLOAT16(163), FLOAT16(164), FLOAT16(165), FLOAT16(166), FLOAT16(167), FLOAT16(168), FLOAT16(169), FLOAT16(170), FLOAT16(171), FLOAT16(172), FLOAT16(173), FLOAT16(174), FLOAT16(175), FLOAT16(176), FLOAT16(177), FLOAT16(178), FLOAT16(179),
-        FLOAT16(180), FLOAT16(181), FLOAT16(182), FLOAT16(183), FLOAT16(184), FLOAT16(185), FLOAT16(186), FLOAT16(187), FLOAT16(188), FLOAT16(189), FLOAT16(190), FLOAT16(191), FLOAT16(192), FLOAT16(193), FLOAT16(194), FLOAT16(195), FLOAT16(196), FLOAT16(197), FLOAT16(198), FLOAT16(199), 
+        FLOAT16(180), FLOAT16(181), FLOAT16(182), FLOAT16(183), FLOAT16(184), FLOAT16(185), FLOAT16(186), FLOAT16(187), FLOAT16(188), FLOAT16(189), FLOAT16(190), FLOAT16(191), FLOAT16(192), FLOAT16(193), FLOAT16(194), FLOAT16(195), FLOAT16(196), FLOAT16(197), FLOAT16(198), FLOAT16(199),
 
         FLOAT16(200), FLOAT16(201), FLOAT16(202), FLOAT16(203), FLOAT16(204), FLOAT16(205), FLOAT16(206), FLOAT16(207), FLOAT16(208), FLOAT16(209), FLOAT16(210), FLOAT16(211), FLOAT16(212), FLOAT16(213), FLOAT16(214), FLOAT16(215), FLOAT16(216), FLOAT16(217), FLOAT16(218), FLOAT16(219),
         FLOAT16(220), FLOAT16(221), FLOAT16(222), FLOAT16(223), FLOAT16(224), FLOAT16(225), FLOAT16(226), FLOAT16(227), FLOAT16(228), FLOAT16(229), FLOAT16(230), FLOAT16(231), FLOAT16(232), FLOAT16(233), FLOAT16(234), FLOAT16(235), FLOAT16(236), FLOAT16(237), FLOAT16(238), FLOAT16(239)
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
+
     network network(engine, topology);
-    
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
-    
+
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
@@ -569,26 +564,26 @@ TEST(scatter_update_gpu_fp16, d8x2x20x1_axisB) {
 
         40.f, 41.f,  42.f,  43.f,  44.f,  45.f,  46.f,  47.f,  48.f,  49.f,  50.f,  51.f,  52.f,  53.f,  54.f,  55.f,  56.f, 57.f,  58.f,  59.f,
         60.f,  61.f,  62.f,  63.f,  64.f,  65.f,  66.f, 67.f,  68.f,  69.f, 70.f,  71.f,  72.f,  73.f,  74.f,  75.f,  76.f,  77.f,  78.f,  79.f,
-        
+
         120.f, 121.f, 122.f, 123.f, 124.f, 125.f, 126.f, 127.f, 128.f, 129.f,130.f, 131.f, 132.f, 133.f, 134.f, 135.f, 136.f, 137.f, 138.f, 139.f,
         140.f, 141.f, 142.f, 143.f, 144.f, 145.f, 146.f, 147.f, 148.f, 149.f,150.f, 151.f, 152.f, 153.f, 154.f, 155.f, 156.f, 157.f, 158.f, 159.f,
 
         0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 13.f, 14.f, 15.f, 16.f, 17.f, 18.f, 19.f,
         20.f, 21.f, 22.f, 23.f, 24.f, 25.f, 26.f, 27.f, 28.f, 29.f, 30.f, 31.f, 32.f, 33.f, 34.f , 35.f, 36.f, 37.f, 38.f, 39.f,
-        
-        200.f, 201.f, 202.f, 203.f, 204.f, 205.f, 206.f, 207.f, 208.f, 209.f,210.f, 211.f, 212.f, 213.f, 214.f, 215.f, 216.f, 217.f, 218.f, 219.f, 
+
+        200.f, 201.f, 202.f, 203.f, 204.f, 205.f, 206.f, 207.f, 208.f, 209.f,210.f, 211.f, 212.f, 213.f, 214.f, 215.f, 216.f, 217.f, 218.f, 219.f,
         220.f, 221.f, 222.f, 223.f, 224.f, 225.f, 226.f,227.f, 228.f, 229.f, 230.f, 231.f, 232.f, 233.f, 234.f, 235.f, 236.f, 237.f, 238.f, 239.f,
 
         0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
         1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f , 1.f, 1.f, 1.f, 1.f, 1.f,
 
-        80.f,  81.f,  82.f,  83.f,  84.f,  85.f,  86.f, 87.f,  88.f,  89.f, 90.f,  91.f,  92.f,  93.f,  94.f,  95.f,  96.f,  97.f,  98.f,  99.f, 
+        80.f,  81.f,  82.f,  83.f,  84.f,  85.f,  86.f, 87.f,  88.f,  89.f, 90.f,  91.f,  92.f,  93.f,  94.f,  95.f,  96.f,  97.f,  98.f,  99.f,
         100.f, 101.f, 102.f, 103.f, 104.f, 105.f, 106.f, 107.f, 108.f, 109.f,110.f, 111.f, 112.f, 113.f, 114.f, 115.f, 116.f, 117.f, 118.f, 119.f,
 
-        160.f, 161.f, 162.f, 163.f, 164.f, 165.f, 166.f, 167.f, 168.f, 169.f, 170.f, 171.f, 172.f, 173.f, 174.f, 175.f, 176.f, 177.f, 178.f, 179.f, 
+        160.f, 161.f, 162.f, 163.f, 164.f, 165.f, 166.f, 167.f, 168.f, 169.f, 170.f, 171.f, 172.f, 173.f, 174.f, 175.f, 176.f, 177.f, 178.f, 179.f,
         180.f, 181.f, 182.f, 183.f, 184.f, 185.f, 186.f, 187.f, 188.f, 189.f, 190.f, 191.f, 192.f, 193.f, 194.f, 195.f, 196.f, 197.f, 198.f, 199.f
     };
-    
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
     }
@@ -626,11 +621,11 @@ TEST(scatter_update_gpu_fp32, d2214_axisX) {
     //  90.f, 9.f, 80.f, 100.f,
     //  120.f, 13.f, 110.f, 130.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 4, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 3, 1, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 3, 1 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 4, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 3, 1, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 3, 1 } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_x;
 
     set_values(input1, {
@@ -652,23 +647,23 @@ TEST(scatter_update_gpu_fp32, d2214_axisX) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-    
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         30.f, 1.f, 20.f, 40.f,
@@ -677,7 +672,7 @@ TEST(scatter_update_gpu_fp32, d2214_axisX) {
         120.f, 13.f, 110.f, 130.f
     };
 
-    
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], output_ptr[i]);
     }
@@ -718,11 +713,11 @@ TEST(scatter_update_gpu_int32, d6211_axisB) {
     //   9,  10,
     //  60,  70
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::i32, format::bfyx, { 6, 2, 1, 1 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 1, 2, 1, 2 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::i32, format::bfyx, { 1, 2, 2, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::i32, format::bfyx, { 6, 2, 1, 1 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 1, 2, 1, 2 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::i32, format::bfyx, { 1, 2, 2, 2 } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_b;
 
     set_values(input1, {
@@ -747,23 +742,23 @@ TEST(scatter_update_gpu_int32, d6211_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology);  
-    
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<int>();
+    cldnn::mem_lock<int> output_ptr(output, get_test_stream());
 
     std::vector<int> expected_results = {
         1, 2,
@@ -811,11 +806,11 @@ TEST(scatter_update_gpu_int32, d3151_axisY) {
     //   70,  6,  60,  50, 80,
     //   110, 11, 100, 90, 120
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 5 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
-    auto input3 = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 2, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 5 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 2, 1, 1 } }); // Indexes
+    auto input3 = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 2, 2 } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_y;
 
     set_values(input1, {
@@ -839,30 +834,30 @@ TEST(scatter_update_gpu_int32, d3151_axisY) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-    
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<int>();
+    cldnn::mem_lock<int> output_ptr(output, get_test_stream());
 
     std::vector<int> expected_results = {
         30, 1, 20, 200, 40,
         70, 6, 60, 50, 80,
         110, 11, 100, 90, 120
     };
-    
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], output_ptr[i]);
     }
@@ -880,7 +875,7 @@ TEST(scatter_update_gpu_fp32, d24111_axisF_bfzyx) {
     //  2.f, 0.f
     //
     //  Updates:
-    //  1.f, 2.f, 
+    //  1.f, 2.f,
     //  3.f, 4.f
     //
     //  Dictionary:
@@ -888,14 +883,14 @@ TEST(scatter_update_gpu_fp32, d24111_axisF_bfzyx) {
     //  0.f, 0.f, 0.f, 0.f
     //
     //  Output:
-    //  2.f, 0.f, 1.f, 0.f, 
+    //  2.f, 0.f, 1.f, 0.f,
     //  4.f, 0.f, 3.f, 0.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 4, 1, 1 } });      // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 1 } });      // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfzyx, { 2, 1, 1, 2, 1 } });  // Updates
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 4, 1, 1 } });      // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 1 } });      // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfzyx, { 2, 1, 1, 2, 1 } });  // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_f;
 
     set_values(input1, {
@@ -913,29 +908,29 @@ TEST(scatter_update_gpu_fp32, d24111_axisF_bfzyx) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-    
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
-    auto outputs = network.execute(); 
+
+    auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
-        2.f, 0.f, 1.f, 0.f, 
+        2.f, 0.f, 1.f, 0.f,
         4.f, 0.f, 3.f, 0.f
     };
-    
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], output_ptr[i]);
     }
@@ -970,19 +965,19 @@ TEST(scatter_update_gpu_int32, d121251_bfwzyx_axisB) {
     //  0, 1, 2, 3, 4,
     //  5, 6, 7, 8, 9,
     //  10, 11, 12, 13, 14,
-    //  15, 16, 17, 18, 19 
+    //  15, 16, 17, 18, 19
     //
     //  Output:
     //  40,  30,   20,  3, 50,
     //  80,  70,   60,  8, 90,
     //  120, 110, 100, 13, 130,
-    //  160, 150, 140, 18, 170 
+    //  160, 150, 140, 18, 170
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::i32, format::bfwzyx, tensor{ batch(1), feature(2), spatial(1, 5, 2, 1) }}); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 2, 1, 1 } });                                       // Indexes
-    auto input3 = memory::allocate(engine, { data_types::i32, format::bfwzyx, tensor{ batch(1), feature(2), spatial(2, 2, 2, 1) }}); // Updates
+    auto input1 = engine.allocate_memory({ data_types::i32, format::bfwzyx, tensor{ batch(1), feature(2), spatial(1, 5, 2, 1) }}); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 2, 1, 1 } });                                       // Indexes
+    auto input3 = engine.allocate_memory({ data_types::i32, format::bfwzyx, tensor{ batch(1), feature(2), spatial(2, 2, 2, 1) }}); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_y;
 
     set_values(input1, {
@@ -1009,23 +1004,23 @@ TEST(scatter_update_gpu_int32, d121251_bfwzyx_axisB) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-    
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
-    
+
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<int>();
+    cldnn::mem_lock<int> output_ptr(output, get_test_stream());
 
     std::vector<int> expected_results = {
         40,  30,   20,  3, 50,
@@ -1052,7 +1047,7 @@ TEST(scatter_update_gpu_fp32, d21511_bfzyx_axisX) {
     //  0.f, 1.f
     //
     //  Updates:
-    //  10.f, 20.f, 
+    //  10.f, 20.f,
     //  30.f, 40.f,
     //  50.f, 60.f,
     //  70.f, 80.f
@@ -1066,11 +1061,11 @@ TEST(scatter_update_gpu_fp32, d21511_bfzyx_axisX) {
     //  70.f, 80.f, 7.f, 50.f, 60.f
     //
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfzyx, { 2, 1, 1, 1, 5 } }); // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } });     // Indices
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfzyx, { 2, 1, 1, 2, 2 } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfzyx, { 2, 1, 1, 1, 5 } }); // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } });     // Indices
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfzyx, { 2, 1, 1, 2, 2 } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_z;
 
     set_values(input1, {
@@ -1084,31 +1079,31 @@ TEST(scatter_update_gpu_fp32, d21511_bfzyx_axisX) {
     });
 
     set_values(input3, {
-        10.f, 20.f, 
+        10.f, 20.f,
         30.f, 40.f,
         50.f, 60.f,
         70.f, 80.f
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
+
     network network(engine, topology);
-    
-    
+
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         30.f, 40.f, 2.f, 10.f, 20.f,
@@ -1133,16 +1128,16 @@ TEST(scatter_update_gpu_fp32, d1252_axisY_bfwzyx) {
     //  3.f, 4.f
     //
     //  Updates:
-    //  20.f, 30.f, 
+    //  20.f, 30.f,
     //  40.f, 50.f
     //
-    //  60.f, 70.f, 
+    //  60.f, 70.f,
     //  80.f, 90.f
     //
-    //  100.f, 110.f, 
+    //  100.f, 110.f,
     //  120.f, 130.f
     //
-    //  140.f, 150.f, 
+    //  140.f, 150.f,
     //  160.f, 170.f
     //
     //  Dictionary:
@@ -1153,11 +1148,11 @@ TEST(scatter_update_gpu_fp32, d1252_axisY_bfwzyx) {
     //  40.f, 50.f,     2.f, 3.f,     20.f, 30.f,     60.f, 70.f,     80.f, 90.f,
     //  120.f, 130.f,   12.f, 13.f,   100.f, 110.f,   140.f, 150.f,   160.f, 170.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 5 } });                                         // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 1, 1, 2 } });                                         // Indices
-    auto input3 = memory::allocate(engine, { data_types::f32, format::bfwzyx, tensor{ batch(1), feature(2), spatial(2, 2, 1, 2) } });  // Updates
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 5 } });                                         // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 1, 1, 2 } });                                         // Indices
+    auto input3 = engine.allocate_memory({ data_types::f32, format::bfwzyx, tensor{ batch(1), feature(2), spatial(2, 2, 1, 2) } });  // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_y;
 
     set_values(input1, {
@@ -1171,43 +1166,43 @@ TEST(scatter_update_gpu_fp32, d1252_axisY_bfwzyx) {
     });
 
     set_values(input3, {
-        20.f, 30.f, 
+        20.f, 30.f,
         40.f, 50.f,
 
-        60.f, 70.f, 
+        60.f, 70.f,
         80.f, 90.f,
-      
-        100.f, 110.f, 
+
+        100.f, 110.f,
         120.f, 130.f,
-    
-        140.f, 150.f, 
+
+        140.f, 150.f,
         160.f, 170.f
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-    
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         40.f, 50.f,     2.f, 3.f,     20.f, 30.f,     60.f, 70.f,     80.f, 90.f,
         120.f, 130.f,   12.f, 13.f,   100.f, 110.f,   140.f, 150.f,   160.f, 170.f
     };
-    
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], output_ptr[i]);
     }
@@ -1240,11 +1235,11 @@ TEST(scatter_update_gpu_int32, d2115_axisX_bfwzyx) {
     //  0,  30,   20,  50, 40,
     //  5,  70,   60,  90, 80
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 1, 5, 1 }});                                        // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 2, 1, 1 } });                                       // Indexes
-    auto input3 = memory::allocate(engine, { data_types::i32, format::bfwzyx, tensor{ batch(2), feature(1), spatial(1, 2, 2, 1) }}); // Updates
+    auto input1 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 1, 5, 1 }});                                        // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 2, 1, 1 } });                                       // Indexes
+    auto input3 = engine.allocate_memory({ data_types::i32, format::bfwzyx, tensor{ batch(2), feature(1), spatial(1, 2, 2, 1) }}); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_x;
 
     set_values(input1, {
@@ -1265,29 +1260,29 @@ TEST(scatter_update_gpu_int32, d2115_axisX_bfwzyx) {
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-    
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<int>();
+    cldnn::mem_lock<int> output_ptr(output, get_test_stream());
 
     std::vector<int> expected_results = {
         0,  30,  20,  50, 40,
         5,  70,  60,  90, 80
     };
-    
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], output_ptr[i]);
     }
@@ -1322,11 +1317,11 @@ TEST(scatter_update_gpu_fp16, d21214_bfzyx_axisX_bfwzyx) {
     //  8.f, 100.f, 90.f, 80.f,
     //  12.f, 130.f, 120.f, 110.f
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfzyx, { 2, 1, 4, 1, 2 } });                                    // Dictionary
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 3, 1, 1 } });                                        // Indexes
-    auto input3 = memory::allocate(engine, { data_types::f16, format::bfwzyx, tensor{ batch(2), feature(1), spatial(3, 1, 1, 2) } }); // Updates
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfzyx, { 2, 1, 4, 1, 2 } });                                    // Dictionary
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 3, 1, 1 } });                                        // Indexes
+    auto input3 = engine.allocate_memory({ data_types::f16, format::bfwzyx, tensor{ batch(2), feature(1), spatial(3, 1, 1, 2) } }); // Updates
     auto axis = cldnn::scatter_update::scatter_update_axis::along_x;
 
     set_values(input1, {
@@ -1341,30 +1336,30 @@ TEST(scatter_update_gpu_fp16, d21214_bfzyx_axisX_bfwzyx) {
     });
 
     set_values(input3, {
-        FLOAT16(20.0f), FLOAT16(30.0f), FLOAT16(40.0f), 
+        FLOAT16(20.0f), FLOAT16(30.0f), FLOAT16(40.0f),
         FLOAT16(50.0f), FLOAT16(60.0f), FLOAT16(70.0f),
-        FLOAT16(80.0f), FLOAT16(90.0f), FLOAT16(100.0f), 
+        FLOAT16(80.0f), FLOAT16(90.0f), FLOAT16(100.0f),
         FLOAT16(110.0f), FLOAT16(120.0f), FLOAT16(130.0f)
     });
 
     topology topology;
-    topology.add(input_layout("InputDictionary", input1.get_layout()));
-    topology.add(input_layout("InputText", input2.get_layout()));
-    topology.add(input_layout("InputUpdates", input3.get_layout()));
+    topology.add(input_layout("InputDictionary", input1->get_layout()));
+    topology.add(input_layout("InputText", input2->get_layout()));
+    topology.add(input_layout("InputUpdates", input3->get_layout()));
     topology.add(
         scatter_update("scatter_update", "InputDictionary", "InputText", "InputUpdates", axis)
     );
-    
-    network network(engine, topology); 
-    
+
+    network network(engine, topology);
+
     network.set_input_data("InputDictionary", input1);
     network.set_input_data("InputText", input2);
     network.set_input_data("InputUpdates", input3);
-    
+
     auto outputs = network.execute();
 
     auto output = outputs.at("scatter_update").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 40.f, 30.f, 20.f,
@@ -1372,7 +1367,7 @@ TEST(scatter_update_gpu_fp16, d21214_bfzyx_axisX_bfwzyx) {
         8.f, 100.f, 90.f, 80.f,
         12.f, 130.f, 120.f, 110.f
     };
-    
+
     for (size_t i = 0; i < expected_results.size(); ++i) {
         EXPECT_EQ(expected_results[i], float16_to_float32(output_ptr[i]));
     }
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/select_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/select_gpu_test.cpp
index fff6997834b640..c4e698f82108b0 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/select_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/select_gpu_test.cpp
@@ -3,30 +3,27 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/select.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include "cldnn/primitives/select.hpp"
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 // select_gpu_f32
 TEST(select_gpu_f32, select_basic) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb, { 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values(input, {
@@ -62,7 +59,7 @@ TEST(select_gpu_f32, select_basic) {
                           15.f,   0.5f,   8.f,  12.f,
                            4.f,   6.5f,   8.f,  -2.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -71,16 +68,16 @@ TEST(select_gpu_f32, select_basic) {
 }
 
 TEST(select_gpu_f32, select_basic_negative) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values(input, {
@@ -116,7 +113,7 @@ TEST(select_gpu_f32, select_basic_negative) {
         15.f,   0.5f,   8.f,  12.f,
         4.f,   6.5f,   8.f,  -2.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -125,16 +122,16 @@ TEST(select_gpu_f32, select_basic_negative) {
 }
 
 TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_mask_2x2x1x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 1, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 1, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -202,7 +199,7 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_mask_2x2x1x2) {
         -0.5f, -2.5f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -211,16 +208,16 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_mask_2x2x1x2) {
 }
 
 TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_mask_1x1x1x1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -278,7 +275,7 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_mask_1x1x1x1) {
         -0.5f, -2.5f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -287,16 +284,16 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_mask_1x1x1x1) {
 }
 
 TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x2x2x1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::byxf, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::byxf ,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::byxf, { 2, 2, 2, 1 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::byxf, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::byxf ,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::byxf, { 2, 2, 2, 1 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -358,7 +355,7 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x2x2x1) {
         -0.5f, 8.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -367,16 +364,16 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x2x2x1) {
 }
 
 TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in2_2x2x1x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -444,7 +441,7 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in2_2x2x1x2) {
         8.f,   -0.5f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -453,16 +450,16 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in2_2x2x1x2) {
 }
 
 TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in1_2x2x2x1_bcast_in2_2x2x1x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 1 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -526,7 +523,7 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in1_2x2x2x1_bcast_in2_2x2x1
         4.f,   -0.5f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -535,16 +532,16 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in1_2x2x2x1_bcast_in2_2x2x1
 }
 
 TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_mask_2x1x2x2_in1_1x2x2x2_in2_2x2x1x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 1, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -600,7 +597,7 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_mask_2x1x2x2_in1_1x2x2x2_in
         -0.5f, 5.2f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -609,16 +606,16 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_mask_2x1x2x2_in1_1x2x2x2_in
 }
 
 TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x1x2x2_in1_2x2x2x1_in2_2x2x1x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::byxf, { 2, 2, 2, 1 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::byxf ,{ 2, 2, 1, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::byxf, { 2, 1, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::byxf, { 2, 2, 2, 1 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::byxf ,{ 2, 2, 1, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::byxf, { 2, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -644,7 +641,7 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x1x2x2_in1_2x2x
         0.f,
 
         0.1f,
-        0.5f,  
+        0.5f,
 
         -0.f,
         -0.5f,
@@ -676,7 +673,7 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x1x2x2_in1_2x2x
         -2.f,  6.5f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -685,16 +682,16 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x1x2x2_in1_2x2x
 }
 
 TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in2_1x1x1x1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -752,7 +749,7 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in2_1x1x1x1) {
         8.f,   1.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -761,16 +758,16 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in2_1x1x1x1) {
 }
 
 TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_in2_2x2x2x1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::byxf, { 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::byxf ,{ 2, 2, 2, 1 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::byxf, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::byxf, { 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::byxf ,{ 2, 2, 2, 1 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::byxf, { 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -800,7 +797,7 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_in2_2x2x2x1) {
         0.5f,  0.7f,
 
         0.f,   0.f,
-        0.f,   0.f,        
+        0.f,   0.f,
 
         -0.f,  -0.1f,
         -0.f,  -0.5f,
@@ -832,7 +829,7 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_in2_2x2x2x1) {
         8.f,  10.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -841,16 +838,16 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_in2_2x2x2x1) {
 }
 
 TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in1_2x2x1x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -918,7 +915,7 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in1_2x2x1x2) {
         8.f,   -2.5f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -927,16 +924,16 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in1_2x2x1x2) {
 }
 
 TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in1_1x1x1x1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -994,7 +991,7 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in1_1x1x1x1) {
         1.f,   -2.5f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1003,16 +1000,16 @@ TEST(select_gpu_f32, select_basic_bfyx_2x2x2x2_bcast_in1_1x1x1x1) {
 }
 
 TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_in1_2x2x2x1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::byxf, { 2, 2, 2, 1 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::byxf ,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::byxf, { 2, 2, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::byxf, { 2, 2, 2, 1 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::byxf ,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::byxf, { 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -1042,7 +1039,7 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_in1_2x2x2x1) {
         0.f,   0.f,
 
         0.1f,  0.3f,
-        0.5f,  0.7f,   
+        0.5f,  0.7f,
 
         -0.f,  -0.1f,
         -0.f,  -0.5f,
@@ -1074,7 +1071,7 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_in1_2x2x2x1) {
         7.f,  -2.5f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1083,16 +1080,16 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_in1_2x2x2x1) {
 }
 
 TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x1x2x2_in1_2x2x2x1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::byxf, { 2, 2, 2, 1 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::byxf ,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::byxf, { 2, 1, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::byxf, { 2, 2, 2, 1 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::byxf ,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::byxf, { 2, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input1", input1.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input1", input1->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input1", "input2"));
 
     set_values(input1, {
@@ -1122,7 +1119,7 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x1x2x2_in1_2x2x
         0.f,
 
         0.1f,
-        0.5f,  
+        0.5f,
 
         -0.f,
         -0.5f,
@@ -1154,7 +1151,7 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x1x2x2_in1_2x2x
         -0.5f, -2.5f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1163,16 +1160,16 @@ TEST(select_gpu_f32, select_basic_comma_byxf_2x2x2x2_bcast_mask_2x1x2x2_in1_2x2x
 }
 
 TEST(select_gpu_f32, select_basic_comma) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values(input, {
@@ -1208,7 +1205,7 @@ TEST(select_gpu_f32, select_basic_comma) {
         15.f,   0.5f,   8.f,  12.f,
         4.f,   6.5f,   8.f,  -2.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1217,79 +1214,79 @@ TEST(select_gpu_f32, select_basic_comma) {
 }
 
 TEST(select_gpu_f32, select_basic_error_input_sizes) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 3, 4, 5, 6 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 3, 4, 5, 6 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     EXPECT_ANY_THROW(network(engine, topology));
 }
 
 TEST(select_gpu_f32, select_basic_error_mask_sizes) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::yxfb,{ 3, 4, 5, 6 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::yxfb,{ 3, 4, 5, 6 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     EXPECT_ANY_THROW(network(engine, topology));
 }
 
 TEST(select_gpu_f32, select_basic_error_input_types) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::i8, format::yxfb,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::i8, format::yxfb,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
     EXPECT_ANY_THROW(network(engine, topology));
 }
 
 TEST(select_gpu_f32, select_basic_error_input_formats) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::bfyx,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     EXPECT_ANY_THROW(network(engine, topology));
 }
 
 TEST(select_gpu_f32, select_basic_byxf) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::byxf,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::byxf,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::byxf,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::byxf,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::byxf,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::byxf,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values(input, {
@@ -1325,7 +1322,7 @@ TEST(select_gpu_f32, select_basic_byxf) {
         15.f,   0.5f,   8.f,  12.f,
         4.f,   6.5f,   8.f,  -2.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1334,16 +1331,16 @@ TEST(select_gpu_f32, select_basic_byxf) {
 }
 
 TEST(select_gpu_f32, select_basic_mask_f16) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f16, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f16, format::yxfb,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values(input, {
@@ -1379,7 +1376,7 @@ TEST(select_gpu_f32, select_basic_mask_f16) {
         15.f,   0.5f,   8.f,  12.f,
         4.f,   6.5f,   8.f,  -2.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1388,16 +1385,16 @@ TEST(select_gpu_f32, select_basic_mask_f16) {
 }
 
 TEST(select_gpu_f32, select_basic_mask_i8) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::i8, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::i8, format::yxfb,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values(input, {
@@ -1433,7 +1430,7 @@ TEST(select_gpu_f32, select_basic_mask_i8) {
         15.f,   0.5f,   8.f,  12.f,
         4.f,   6.5f,   8.f,  -2.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1442,16 +1439,16 @@ TEST(select_gpu_f32, select_basic_mask_i8) {
 }
 
 TEST(select_gpu_f32, select_basic_mask_u8) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::u8, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 2, 2, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::u8, format::yxfb,{ 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values(input, {
@@ -1487,7 +1484,7 @@ TEST(select_gpu_f32, select_basic_mask_u8) {
         15.f,   0.5f,   8.f,  12.f,
         4.f,   6.5f,   8.f,  -2.5f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 16; i++)
     {
@@ -1496,16 +1493,16 @@ TEST(select_gpu_f32, select_basic_mask_u8) {
 }
 
 TEST(select_gpu_f32, select_basic_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values(input, {
@@ -1529,11 +1526,11 @@ TEST(select_gpu_f32, select_basic_1x1x2x2) {
 
     auto output = outputs.at("select").get_memory();
 
-    float answers[4] = { 
+    float answers[4] = {
         0.5f,    2.5f,    2.f,    0.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -1542,16 +1539,16 @@ TEST(select_gpu_f32, select_basic_1x1x2x2) {
 }
 
 TEST(select_gpu_f32, select_basic_bfyx_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values(input, {
@@ -1583,7 +1580,7 @@ TEST(select_gpu_f32, select_basic_bfyx_1x1x2x2) {
         2.f,   0.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -1592,16 +1589,16 @@ TEST(select_gpu_f32, select_basic_bfyx_1x1x2x2) {
 }
 
 TEST(select_gpu_f32, select_basic_byxf_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::byxf,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f32, format::byxf,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::byxf,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::byxf,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f32, format::byxf,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::byxf,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values(input, {
@@ -1633,7 +1630,7 @@ TEST(select_gpu_f32, select_basic_byxf_1x1x2x2) {
         2.f,   0.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -1643,16 +1640,16 @@ TEST(select_gpu_f32, select_basic_byxf_1x1x2x2) {
 
 // select_gpu_f16
 TEST(select_gpu_f16, select_basic_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<uint16_t>(input, {
@@ -1684,7 +1681,7 @@ TEST(select_gpu_f16, select_basic_1x1x2x2) {
         2,   0
     };
 
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -1693,16 +1690,16 @@ TEST(select_gpu_f16, select_basic_1x1x2x2) {
 }
 
 TEST(select_gpu_f16, select_basic_mask_f32_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<uint16_t>(input, {
@@ -1734,7 +1731,7 @@ TEST(select_gpu_f16, select_basic_mask_f32_1x1x2x2) {
         2,   0
     };
 
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -1743,16 +1740,16 @@ TEST(select_gpu_f16, select_basic_mask_f32_1x1x2x2) {
 }
 
 TEST(select_gpu_f16, select_basic_mask_i8_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<uint16_t>(input, {
@@ -1784,7 +1781,7 @@ TEST(select_gpu_f16, select_basic_mask_i8_1x1x2x2) {
         2,   0
     };
 
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -1793,16 +1790,16 @@ TEST(select_gpu_f16, select_basic_mask_i8_1x1x2x2) {
 }
 
 TEST(select_gpu_f16, select_basic_mask_u8_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<uint16_t>(input, {
@@ -1834,7 +1831,7 @@ TEST(select_gpu_f16, select_basic_mask_u8_1x1x2x2) {
         2,   0
     };
 
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -1844,16 +1841,16 @@ TEST(select_gpu_f16, select_basic_mask_u8_1x1x2x2) {
 
 // select_gpu_i8
 TEST(select_gpu_i8, select_basic_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<char>(input, {
@@ -1885,7 +1882,7 @@ TEST(select_gpu_i8, select_basic_1x1x2x2) {
         2,  0
     };
 
-    auto output_ptr = output.pointer<char>();
+    cldnn::mem_lock<char> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -1894,16 +1891,16 @@ TEST(select_gpu_i8, select_basic_1x1x2x2) {
 }
 
 TEST(select_gpu_i8, select_basic_mask_f32_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<char>(input, {
@@ -1935,7 +1932,7 @@ TEST(select_gpu_i8, select_basic_mask_f32_1x1x2x2) {
         2,  0
     };
 
-    auto output_ptr = output.pointer<char>();
+    cldnn::mem_lock<char> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -1944,16 +1941,16 @@ TEST(select_gpu_i8, select_basic_mask_f32_1x1x2x2) {
 }
 
 TEST(select_gpu_i8, select_basic_mask_f16_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<char>(input, {
@@ -1985,7 +1982,7 @@ TEST(select_gpu_i8, select_basic_mask_f16_1x1x2x2) {
         2,  0
     };
 
-    auto output_ptr = output.pointer<char>();
+    cldnn::mem_lock<char> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -1994,16 +1991,16 @@ TEST(select_gpu_i8, select_basic_mask_f16_1x1x2x2) {
 }
 
 TEST(select_gpu_i8, select_basic_mask_u8_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<char>(input, {
@@ -2035,7 +2032,7 @@ TEST(select_gpu_i8, select_basic_mask_u8_1x1x2x2) {
         2,  0
     };
 
-    auto output_ptr = output.pointer<char>();
+    cldnn::mem_lock<char> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -2045,16 +2042,16 @@ TEST(select_gpu_i8, select_basic_mask_u8_1x1x2x2) {
 
 // select_gpu_u8
 TEST(select_gpu_u8, select_basic_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<unsigned char>(input, {
@@ -2086,7 +2083,7 @@ TEST(select_gpu_u8, select_basic_1x1x2x2) {
         255,  0
     };
 
-    auto output_ptr = output.pointer<unsigned char>();
+    cldnn::mem_lock<unsigned char> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -2095,16 +2092,16 @@ TEST(select_gpu_u8, select_basic_1x1x2x2) {
 }
 
 TEST(select_gpu_u8, select_basic_mask_f32_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<unsigned char>(input, {
@@ -2136,7 +2133,7 @@ TEST(select_gpu_u8, select_basic_mask_f32_1x1x2x2) {
         255,  0
     };
 
-    auto output_ptr = output.pointer<unsigned char>();
+    cldnn::mem_lock<unsigned char> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -2145,16 +2142,16 @@ TEST(select_gpu_u8, select_basic_mask_f32_1x1x2x2) {
 }
 
 TEST(select_gpu_u8, select_basic_mask_f16_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::f16, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<unsigned char>(input, {
@@ -2186,7 +2183,7 @@ TEST(select_gpu_u8, select_basic_mask_f16_1x1x2x2) {
         255,  0
     };
 
-    auto output_ptr = output.pointer<unsigned char>();
+    cldnn::mem_lock<unsigned char> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
@@ -2195,16 +2192,16 @@ TEST(select_gpu_u8, select_basic_mask_f16_1x1x2x2) {
 }
 
 TEST(select_gpu_u8, select_basic_mask_i8_1x1x2x2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto input2 = memory::allocate(engine, { data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
-    auto mask = memory::allocate(engine, { data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto input2 = engine.allocate_memory({ data_types::u8, format::yxfb,{ 1, 1, 2, 2 } });
+    auto mask = engine.allocate_memory({ data_types::i8, format::yxfb,{ 1, 1, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("input2", input2.get_layout()));
-    topology.add(input_layout("mask", mask.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("input2", input2->get_layout()));
+    topology.add(input_layout("mask", mask->get_layout()));
     topology.add(cldnn::select("select", "mask", "input", "input2"));
 
     set_values<unsigned char>(input, {
@@ -2236,7 +2233,7 @@ TEST(select_gpu_u8, select_basic_mask_i8_1x1x2x2) {
         255,  0
     };
 
-    auto output_ptr = output.pointer<unsigned char>();
+    cldnn::mem_lock<unsigned char> output_ptr(output, get_test_stream());
 
     for (int i = 0; i < 4; i++)
     {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/shuffle_channels_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/shuffle_channels_test.cpp
index a1bbf1987715b5..55a5e8fa8979c0 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/shuffle_channels_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/shuffle_channels_test.cpp
@@ -3,24 +3,21 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/shuffle_channels.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/shuffle_channels.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
 
 TEST(shuffle_channels_fp32_gpu, d1_15_2_2_ax1_g5) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 15, 2, 2 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 15, 2, 2 } });
     int32_t axis = 1;
     int32_t group = 5;
 
@@ -34,7 +31,7 @@ TEST(shuffle_channels_fp32_gpu, d1_15_2_2_ax1_g5) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input0.get_layout()));
+    topology.add(input_layout("Input0", input0->get_layout()));
     topology.add(
             shuffle_channels("shuffle_channels", "Input0", group, axis)
     );
@@ -46,7 +43,7 @@ TEST(shuffle_channels_fp32_gpu, d1_15_2_2_ax1_g5) {
     auto outputs = network.execute();
 
     auto output = outputs.at("shuffle_channels").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 1.f, 2.f, 3.f, 12.f, 13.f, 14.f, 15.f, 24.f, 25.f, 26.f, 27.f, 36.f, 37.f, 38.f, 39.f, 48.f, 49.f, 50.f, 51.f,
@@ -60,9 +57,9 @@ TEST(shuffle_channels_fp32_gpu, d1_15_2_2_ax1_g5) {
 }
 
 TEST(shuffle_channels_fp32_gpu, d1_15_2_2_axm3_g5) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 15, 2, 2 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 15, 2, 2 } });
     int32_t axis = -3;
     int32_t group = 5;
 
@@ -76,7 +73,7 @@ TEST(shuffle_channels_fp32_gpu, d1_15_2_2_axm3_g5) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input0.get_layout()));
+    topology.add(input_layout("Input0", input0->get_layout()));
     topology.add(
             shuffle_channels("shuffle_channels", "Input0", group, axis)
     );
@@ -88,7 +85,7 @@ TEST(shuffle_channels_fp32_gpu, d1_15_2_2_axm3_g5) {
     auto outputs = network.execute();
 
     auto output = outputs.at("shuffle_channels").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 1.f, 2.f, 3.f, 12.f, 13.f, 14.f, 15.f, 24.f, 25.f, 26.f, 27.f, 36.f, 37.f, 38.f, 39.f, 48.f, 49.f, 50.f, 51.f,
@@ -102,9 +99,9 @@ TEST(shuffle_channels_fp32_gpu, d1_15_2_2_axm3_g5) {
 }
 
 TEST(shuffle_channels_fp32_gpu, d15_2_2_ax0_g5) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx, { 15, 2, 1, 2 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 15, 2, 1, 2 } });
     int32_t axis = 0;
     int32_t group = 5;
 
@@ -118,7 +115,7 @@ TEST(shuffle_channels_fp32_gpu, d15_2_2_ax0_g5) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input0.get_layout()));
+    topology.add(input_layout("Input0", input0->get_layout()));
     topology.add(
             shuffle_channels("shuffle_channels", "Input0", group, axis)
     );
@@ -130,7 +127,7 @@ TEST(shuffle_channels_fp32_gpu, d15_2_2_ax0_g5) {
     auto outputs = network.execute();
 
     auto output = outputs.at("shuffle_channels").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 1.f, 2.f, 3.f, 12.f, 13.f, 14.f, 15.f, 24.f, 25.f, 26.f, 27.f, 36.f, 37.f, 38.f, 39.f, 48.f, 49.f, 50.f, 51.f,
@@ -144,9 +141,9 @@ TEST(shuffle_channels_fp32_gpu, d15_2_2_ax0_g5) {
 }
 
 TEST(shuffle_channels_fp32_gpu, d15_2_2_axm4_g5) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx, { 15, 2, 1, 2 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 15, 2, 1, 2 } });
     int32_t axis = -4;
     int32_t group = 5;
 
@@ -160,7 +157,7 @@ TEST(shuffle_channels_fp32_gpu, d15_2_2_axm4_g5) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input0.get_layout()));
+    topology.add(input_layout("Input0", input0->get_layout()));
     topology.add(
             shuffle_channels("shuffle_channels", "Input0", group, axis)
     );
@@ -172,7 +169,7 @@ TEST(shuffle_channels_fp32_gpu, d15_2_2_axm4_g5) {
     auto outputs = network.execute();
 
     auto output = outputs.at("shuffle_channels").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 1.f, 2.f, 3.f, 12.f, 13.f, 14.f, 15.f, 24.f, 25.f, 26.f, 27.f, 36.f, 37.f, 38.f, 39.f, 48.f, 49.f, 50.f, 51.f,
@@ -186,9 +183,9 @@ TEST(shuffle_channels_fp32_gpu, d15_2_2_axm4_g5) {
 }
 
 TEST(shuffle_channels_fp32_gpu, d2_2_6_axm2_g3) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 6 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 6 } });
     int32_t axis = -2;
     int32_t group = 3;
 
@@ -199,7 +196,7 @@ TEST(shuffle_channels_fp32_gpu, d2_2_6_axm2_g3) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input0.get_layout()));
+    topology.add(input_layout("Input0", input0->get_layout()));
     topology.add(
             shuffle_channels("shuffle_channels", "Input0", group, axis)
     );
@@ -211,7 +208,7 @@ TEST(shuffle_channels_fp32_gpu, d2_2_6_axm2_g3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("shuffle_channels").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 2.f, 4.f, 1.f, 3.f, 5.f, 6.f, 8.f, 10.f, 7.f, 9.f, 11.f,
@@ -224,9 +221,9 @@ TEST(shuffle_channels_fp32_gpu, d2_2_6_axm2_g3) {
 }
 
 TEST(shuffle_channels_fp32_gpu, d2_6_2_axm3_g3) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 6, 1, 2 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 6, 1, 2 } });
     int32_t axis = -3;
     int32_t group = 3;
 
@@ -237,7 +234,7 @@ TEST(shuffle_channels_fp32_gpu, d2_6_2_axm3_g3) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input0.get_layout()));
+    topology.add(input_layout("Input0", input0->get_layout()));
     topology.add(
             shuffle_channels("shuffle_channels", "Input0", group, axis)
     );
@@ -249,7 +246,7 @@ TEST(shuffle_channels_fp32_gpu, d2_6_2_axm3_g3) {
     auto outputs = network.execute();
 
     auto output = outputs.at("shuffle_channels").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 1.f, 4.f, 5.f, 8.f, 9.f, 2.f, 3.f, 6.f, 7.f, 10.f, 11.f,
@@ -262,9 +259,9 @@ TEST(shuffle_channels_fp32_gpu, d2_6_2_axm3_g3) {
 }
 
 TEST(shuffle_channels_fp32_gpu, d2_2_6_axm2_g2) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 6 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 6 } });
     int32_t axis = -2;
     int32_t group = 2;
 
@@ -275,7 +272,7 @@ TEST(shuffle_channels_fp32_gpu, d2_2_6_axm2_g2) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input0.get_layout()));
+    topology.add(input_layout("Input0", input0->get_layout()));
     topology.add(
             shuffle_channels("shuffle_channels", "Input0", group, axis)
     );
@@ -287,7 +284,7 @@ TEST(shuffle_channels_fp32_gpu, d2_2_6_axm2_g2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("shuffle_channels").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 3.f, 1.f, 4.f, 2.f, 5.f, 6.f, 9.f, 7.f, 10.f, 8.f, 11.f,
@@ -300,9 +297,9 @@ TEST(shuffle_channels_fp32_gpu, d2_2_6_axm2_g2) {
 }
 
 TEST(shuffle_channels_fp32_gpu, d2_6_2_axm3_g2) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 6, 1, 2 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 6, 1, 2 } });
     int32_t axis = -3;
     int32_t group = 2;
 
@@ -313,7 +310,7 @@ TEST(shuffle_channels_fp32_gpu, d2_6_2_axm3_g2) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input0.get_layout()));
+    topology.add(input_layout("Input0", input0->get_layout()));
     topology.add(
             shuffle_channels("shuffle_channels", "Input0", group, axis)
     );
@@ -325,7 +322,7 @@ TEST(shuffle_channels_fp32_gpu, d2_6_2_axm3_g2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("shuffle_channels").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 1.f, 6.f, 7.f, 2.f, 3.f, 8.f, 9.f, 4.f, 5.f, 10.f, 11.f,
@@ -338,9 +335,9 @@ TEST(shuffle_channels_fp32_gpu, d2_6_2_axm3_g2) {
 }
 
 TEST(shuffle_channels_fp32_gpu, d6_axm0_g2) {
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input0 = memory::allocate(engine, { data_types::f32, format::bfyx, { 6, 1, 1, 1 } });
+    auto input0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 6, 1, 1, 1 } });
     int32_t axis = 0;
     int32_t group = 2;
 
@@ -349,7 +346,7 @@ TEST(shuffle_channels_fp32_gpu, d6_axm0_g2) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input0.get_layout()));
+    topology.add(input_layout("Input0", input0->get_layout()));
     topology.add(
             shuffle_channels("shuffle_channels", "Input0", group, axis)
     );
@@ -361,7 +358,7 @@ TEST(shuffle_channels_fp32_gpu, d6_axm0_g2) {
     auto outputs = network.execute();
 
     auto output = outputs.at("shuffle_channels").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 3.f, 1.f, 4.f, 2.f, 5.f
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/softmax_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/softmax_gpu_test.cpp
index 803ca0b3ae9c50..2ecc932b41e72d 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/softmax_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/softmax_gpu_test.cpp
@@ -2,18 +2,14 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/softmax.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/softmax.hpp>
 
 using namespace cldnn;
 using namespace std;
-using namespace tests;
+using namespace ::tests;
 
 class softmax_gpu_xb_f32_test_fixture: public ::testing::Test {
 public:
@@ -27,14 +23,14 @@ class softmax_gpu_xb_f32_test_fixture: public ::testing::Test {
     float out_buffer[out_size];
     float expected_buffer[out_size];
 
-    const cldnn::engine& engine;
-    cldnn::memory input;
+    cldnn::engine& engine;
+    cldnn::memory::ptr input;
 
     //neural::primitive output = memory::allocate({ memory::format::xb_f32, {output_b, {{output_x}}, 1}});
 
     softmax_gpu_xb_f32_test_fixture()
         : engine(get_test_engine())
-        ,input(memory::allocate(engine, { data_types::f32, format::yxfb, { input_b, 1, input_x, 1}}))
+        , input(engine.allocate_memory({ data_types::f32, format::yxfb, { input_b, 1, input_x, 1}}))
     {}
 
     void compare_out_buffer_with_expected() {
@@ -72,7 +68,7 @@ TEST_F(softmax_gpu_xb_f32_test_fixture, input_same_values) {
 
     set_values(input, in_b);
 
-    network network(engine, topology(input_layout("input", input.get_layout()), softmax("softmax", "input")));
+    network network(engine, topology(input_layout("input", input->get_layout()), softmax("softmax", "input")));
     network.set_input_data("input", input);
 
     auto outputs = network.execute();
@@ -81,10 +77,10 @@ TEST_F(softmax_gpu_xb_f32_test_fixture, input_same_values) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
     for (uint32_t i = 0; i < out_size; i++)
     {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+        out_buffer[i] = output_ptr[i];
     }
     compare_out_buffer_with_expected();
 }
@@ -102,7 +98,7 @@ TEST_F(softmax_gpu_xb_f32_test_fixture, input_same_values_batch_wise) {
     for(size_t i = 0; i < out_size; ++i)
         expected_buffer[i] = 0.1f;
 
-    network network(engine, topology(input_layout("input", input.get_layout()), softmax("softmax", "input")));
+    network network(engine, topology(input_layout("input", input->get_layout()), softmax("softmax", "input")));
     network.set_input_data("input", input);
 
     auto outputs = network.execute();
@@ -111,10 +107,10 @@ TEST_F(softmax_gpu_xb_f32_test_fixture, input_same_values_batch_wise) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
     for (uint32_t i = 0; i < out_size; i++)
     {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+        out_buffer[i] = output_ptr[i];
     }
     compare_out_buffer_with_expected_batch_wise();
 }
@@ -157,7 +153,7 @@ TEST_F(softmax_gpu_xb_f32_test_fixture, values_batch_wise) {
     for(size_t i = 0; i < out_size; ++i)
         out_buffer[i] = NAN;
 
-    network network(engine, topology(input_layout("input", input.get_layout()), softmax("softmax", "input")));
+    network network(engine, topology(input_layout("input", input->get_layout()), softmax("softmax", "input")));
     network.set_input_data("input", input);
 
     auto outputs = network.execute();
@@ -166,10 +162,10 @@ TEST_F(softmax_gpu_xb_f32_test_fixture, values_batch_wise) {
 
     auto output_prim = outputs.begin()->second.get_memory();
 
-    auto output_ptr = output_prim.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output_prim, get_test_stream());
     for (uint32_t i = 0; i < out_size; i++)
     {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+        out_buffer[i] = output_ptr[i];
     }
     compare_out_buffer_with_expected_batch_wise();
 }
@@ -178,11 +174,11 @@ TEST(softmax_gpu_bfyx_f32, normalize_fyx) {
     //  Input  : 2x3x2x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 3,
         batch_num = 2, buf_size = x_size*y_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input"));
 
     set_values(input, {  //bfyx
@@ -208,11 +204,11 @@ TEST(softmax_gpu_bfyx_f32, normalize_fyx) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[buf_size];
     for (uint32_t i = 0; i < buf_size; i++)
     {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+        out_buffer[i] = output_ptr[i];
     }
 
     float sum = 0;
@@ -251,11 +247,11 @@ TEST(softmax_gpu_bfyx_f32, normalize_y) {
     //  Input  : 2x3x2x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 3,
         batch_num = 2, buf_size = x_size*y_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input", softmax::normalize_y));
 
     vector<float> input_vec = {
@@ -299,11 +295,11 @@ TEST(softmax_gpu_bfyx_f32, normalize_y) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[buf_size];
     for (uint32_t i = 0; i < buf_size; i++)
     {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+        out_buffer[i] = output_ptr[i];
     }
 
     float temp_max = 0;
@@ -345,11 +341,11 @@ TEST(softmax_gpu_bfyx_f32, normalize_f) {
     //  Input  : 2x3x2x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 3,
         batch_num = 2, buf_size = x_size*y_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx,{ batch_num, feature_num, x_size , y_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input", softmax::normalize_f));
 
     vector<float> input_vec = {
@@ -387,11 +383,11 @@ TEST(softmax_gpu_bfyx_f32, normalize_f) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[buf_size];
     for (uint32_t i = 0; i < buf_size; i++)
     {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+        out_buffer[i] = output_ptr[i];
     }
 
     float temp_max = 0;
@@ -433,11 +429,11 @@ TEST(softmax_gpu_yxfb_f32, normalize_f) {
 
     static const int32_t x_size = 1, y_size = 2, feature_num = 1,
         batch_num = 12, buf_size = x_size*y_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ batch_num, feature_num, y_size , x_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ batch_num, feature_num, y_size , x_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input", softmax::normalize_fyx));
 
     set_values(input, {  //yxfb
@@ -470,11 +466,11 @@ TEST(softmax_gpu_yxfb_f32, normalize_f) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[buf_size];
     for (uint32_t i = 0; i < buf_size; i++)
     {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+        out_buffer[i] = output_ptr[i];
     }
 
     float expected_sum = 1.0f;
@@ -510,11 +506,11 @@ TEST(softmax_gpu_bfzyx_f32, normalize_z) {
     //  Input  : 2x3x2x2x2
     static const int32_t x_size = 2, y_size = 2, z_size = 2, feature_num = 3,
         batch_num = 2, buf_size = x_size  *y_size * z_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ batch_num, feature_num, x_size , y_size, z_size } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ batch_num, feature_num, x_size , y_size, z_size } });
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input", softmax::normalize_z));
 
     vector<float> input_vec = {
@@ -553,11 +549,11 @@ TEST(softmax_gpu_bfzyx_f32, normalize_z) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float out_buffer[buf_size];
     for (uint32_t i = 0; i < buf_size; i++)
     {
-        out_buffer[i] = get_value<float>(output_ptr, i);
+        out_buffer[i] = output_ptr[i];
     }
 
     float temp_max = 0;
@@ -602,11 +598,11 @@ TEST(softmax_gpu_bfyx_f32, normalize_all) {
     //  Input  : 2x3x2x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 3,
                          batch_num = 2, buf_size = x_size * y_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfyx, {batch_num, feature_num, x_size, y_size}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfyx, {batch_num, feature_num, x_size, y_size}});
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input", softmax::normalize_all));
 
     set_values(input, {//bfyx
@@ -627,11 +623,11 @@ TEST(softmax_gpu_bfyx_f32, normalize_all) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float sum = 0.0f;
     float expected_sum = 1.0f;
     for (uint32_t i = 0; i < buf_size; i++) {
-        sum += get_value<float>(output_ptr, i);
+        sum += output_ptr[i];
     }
     EXPECT_EQ(true, are_equal(sum, expected_sum));
 }
@@ -640,11 +636,11 @@ TEST(softmax_gpu_yxfb_f32, normalize_all) {
     //  Input  : 2x2x3x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 3,
                          batch_num = 2, buf_size = x_size * y_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::yxfb, {y_size, x_size, feature_num, batch_num}});
+    auto input = engine.allocate_memory({data_types::f32, format::yxfb, {y_size, x_size, feature_num, batch_num}});
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input", softmax::normalize_all));
 
     set_values(input, {//yxfb
@@ -665,11 +661,11 @@ TEST(softmax_gpu_yxfb_f32, normalize_all) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float sum = 0.0f;
     float expected_sum = 1.0f;
     for (uint32_t i = 0; i < buf_size; i++) {
-        sum += get_value<float>(output_ptr, i);
+        sum += output_ptr[i];
     }
     EXPECT_EQ(true, are_equal(sum, expected_sum));
 }
@@ -678,11 +674,11 @@ TEST(softmax_gpu_bfzyx_f32, normalize_all) {
     //  Input  : 2x3x2x2x2
     static const int32_t x_size = 2, y_size = 2, z_size = 2, feature_num = 3,
                          batch_num = 2, buf_size = x_size * y_size * z_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f32, format::bfzyx, {batch_num, feature_num, x_size, y_size, z_size}});
+    auto input = engine.allocate_memory({data_types::f32, format::bfzyx, {batch_num, feature_num, x_size, y_size, z_size}});
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input", softmax::normalize_all));
 
     set_values(input, {//    z0y0x0 z0y0x1 z0y1x0 z0y1x1 z1y0x0 z1y0x1 z1y1x0 z1y1x1
@@ -702,11 +698,11 @@ TEST(softmax_gpu_bfzyx_f32, normalize_all) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
     float sum = 0.0f;
     float expected_sum = 1.0f;
     for (uint32_t i = 0; i < buf_size; i++) {
-        sum += get_value<float>(output_ptr, i);
+        sum += output_ptr[i];
     }
     EXPECT_EQ(true, are_equal(sum, expected_sum));
 }
@@ -715,11 +711,11 @@ TEST(softmax_gpu_bfyx_f16, normalize_all) {
     //  Input  : 2x3x2x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 3,
                          batch_num = 2, buf_size = x_size * y_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::bfyx, {batch_num, feature_num, x_size, y_size}});
+    auto input = engine.allocate_memory({data_types::f16, format::bfyx, {batch_num, feature_num, x_size, y_size}});
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input", softmax::normalize_all));
 
     set_values(input, {//bfyx
@@ -740,11 +736,11 @@ TEST(softmax_gpu_bfyx_f16, normalize_all) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
     float sum = 0.0f;
     float expected_sum = 1.0f;
     for (uint32_t i = 0; i < buf_size; i++) {
-        sum += float16_to_float32(get_value<uint16_t>(output_ptr, i));
+        sum += float16_to_float32(output_ptr[i]);
     }
     ASSERT_NEAR(sum, expected_sum, 0.001);
 }
@@ -753,11 +749,11 @@ TEST(softmax_gpu_yxfb_f16, normalize_all) {
     //  Input  : 2x2x3x2
     static const int32_t x_size = 2, y_size = 2, feature_num = 3,
                          batch_num = 2, buf_size = x_size * y_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::yxfb, {y_size, x_size, feature_num, batch_num}});
+    auto input = engine.allocate_memory({data_types::f16, format::yxfb, {y_size, x_size, feature_num, batch_num}});
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input", softmax::normalize_all));
 
     set_values(input, {//yxfb
@@ -778,11 +774,11 @@ TEST(softmax_gpu_yxfb_f16, normalize_all) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
     float sum = 0.0f;
     float expected_sum = 1.0f;
     for (uint32_t i = 0; i < buf_size; i++) {
-        sum += float16_to_float32(get_value<uint16_t>(output_ptr, i));
+        sum += float16_to_float32(output_ptr[i]);
     }
     ASSERT_NEAR(sum, expected_sum, 0.001);
 }
@@ -791,11 +787,11 @@ TEST(softmax_gpu_bfzyx_f16, normalize_all) {
     //  Input  : 2x3x2x2x2
     static const int32_t x_size = 2, y_size = 2, z_size = 2, feature_num = 3,
                          batch_num = 2, buf_size = x_size * y_size * z_size * batch_num * feature_num;
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, {data_types::f16, format::bfzyx, {batch_num, feature_num, x_size, y_size, z_size}});
+    auto input = engine.allocate_memory({data_types::f16, format::bfzyx, {batch_num, feature_num, x_size, y_size, z_size}});
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(softmax("softmax", "input", softmax::normalize_all));
 
     set_values(input, {//           z0y0x0          z0y0x1          z0y1x0        z0y1x1        z1y0x0          z1y0x1          z1y1x0          z1y1x1
@@ -815,11 +811,11 @@ TEST(softmax_gpu_bfzyx_f16, normalize_all) {
     EXPECT_EQ(outputs.begin()->first, "softmax");
 
     auto output = outputs.at("softmax").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
     float sum = 0.0f;
     float expected_sum = 1.0f;
     for (uint32_t i = 0; i < buf_size; i++) {
-        sum += float16_to_float32(get_value<uint16_t>(output_ptr, i));
+        sum += float16_to_float32(output_ptr[i]);
     }
     ASSERT_NEAR(sum, expected_sum, 0.001);
 }
@@ -884,28 +880,26 @@ class softmax_test : public tests::generic_test
     }
 
     template<typename Type>
-    memory generate_reference_typed(const std::vector<memory> & inputs)
+    memory::ptr generate_reference_typed(const std::vector<memory::ptr> & inputs)
     {
         assert(inputs.size() == 1);
-        const memory & input = inputs[0];
+        const memory::ptr input = inputs[0];
 
         //Output is bfyx
-        auto output = memory::allocate(engine, cldnn::layout(input.get_layout().data_type, input.get_layout().format, input.get_layout().size));
-
-//        const auto params = static_cast<cldnn::softmax *>(layer_parmas);
+        auto output = engine.allocate_memory(cldnn::layout(input->get_layout().data_type, input->get_layout().format, input->get_layout().size));
 
-        const auto in0_mem = input.pointer<Type>();
-        auto out_mem = output.pointer<Type>();
+        cldnn::mem_lock<Type> in0_mem(input, get_test_stream());
+        cldnn::mem_lock<Type> out_mem(output, get_test_stream());
 
-        const int in0_b = input.get_layout().size.sizes()[0];
-        const int in0_f = input.get_layout().size.sizes()[1];
-        const int in0_h = input.get_layout().size.sizes()[3];
-        const int in0_w = input.get_layout().size.sizes()[2];
+        const int in0_b = input->get_layout().size.sizes()[0];
+        const int in0_f = input->get_layout().size.sizes()[1];
+        const int in0_h = input->get_layout().size.sizes()[3];
+        const int in0_w = input->get_layout().size.sizes()[2];
 
-//        const int out_b = output.get_layout().size.transform(cldnn::format::bfyx, 0).sizes()[0];
-//        const int out_f = output.get_layout().size.transform(cldnn::format::bfyx, 0).sizes()[1];
-//        const int out_h = output.get_layout().size.transform(cldnn::format::bfyx, 0).sizes()[2];
-//        const int out_w = output.get_layout().size.transform(cldnn::format::bfyx, 0).sizes()[3];
+//        const int out_b = output->get_layout().size.transform(cldnn::format::bfyx, 0).sizes()[0];
+//        const int out_f = output->get_layout().size.transform(cldnn::format::bfyx, 0).sizes()[1];
+//        const int out_h = output->get_layout().size.transform(cldnn::format::bfyx, 0).sizes()[2];
+//        const int out_w = output->get_layout().size.transform(cldnn::format::bfyx, 0).sizes()[3];
 
 //        assert(in0_b == out_b);
 //        assert(in0_f == out_f);
@@ -915,7 +909,7 @@ class softmax_test : public tests::generic_test
         std::vector<float> cached_exp_vals;
         cached_exp_vals.resize(in0_f);
 
-        const auto input_desc = get_linear_memory_desc(input.get_layout());
+        const auto input_desc = get_linear_memory_desc(input->get_layout());
 
         for (int n = 0; n < in0_b; ++n)
         for (int y = 0; y < in0_h; ++y)
@@ -925,7 +919,7 @@ class softmax_test : public tests::generic_test
 
             for (int c = 0; c < in0_f; ++c)
             {
-                const size_t in0_idx = get_linear_index(input.get_layout(), n, c, y, x, input_desc);
+                const size_t in0_idx = get_linear_index(input->get_layout(), n, c, y, x, input_desc);
 
                 max_val = std::max(max_val, static_cast<float>(in0_mem[in0_idx]));
             }
@@ -934,7 +928,7 @@ class softmax_test : public tests::generic_test
 
             for (int c = 0; c < in0_f; ++c)
             {
-                const size_t in0_idx = get_linear_index(input.get_layout(), n, c, y, x, input_desc);
+                const size_t in0_idx = get_linear_index(input->get_layout(), n, c, y, x, input_desc);
 
                 float tmp = static_cast<float>((Type)std::exp(static_cast<float>(in0_mem[in0_idx]) - max_val));
                 Z += tmp;
@@ -943,7 +937,7 @@ class softmax_test : public tests::generic_test
 
             for (int c = 0; c < in0_f; ++c)
             {
-                const size_t out_idx = get_linear_index(output.get_layout(), n, c, y, x, input_desc);
+                const size_t out_idx = get_linear_index(output->get_layout(), n, c, y, x, input_desc);
                 out_mem[out_idx] = (Type)(cached_exp_vals[c] / Z);
             }
         }
@@ -951,7 +945,7 @@ class softmax_test : public tests::generic_test
         return output;
     }
 
-    virtual memory generate_reference(const std::vector<memory> & inputs) override
+    virtual memory::ptr generate_reference(const std::vector<memory::ptr> & inputs) override
     {
         if (generic_params->data_type == data_types::f32)
         {
@@ -1008,4 +1002,3 @@ INSTANTIATE_TEST_CASE_P(DISABLED_SOFTMAX,
     softmax_test,
     ::testing::Combine(::testing::ValuesIn(softmax_test::generate_generic_test_params()), ::testing::ValuesIn(softmax_test::generate_specific_test_params())),
     softmax_test::custom_param_name);
-
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/space_to_batch_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/space_to_batch_gpu_test.cpp
index 9dec5cf8c7393a..cb46b0132079c1 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/space_to_batch_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/space_to_batch_gpu_test.cpp
@@ -3,17 +3,14 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/space_to_batch.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/data.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/space_to_batch.hpp>
+#include <cldnn/primitives/data.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -26,8 +23,8 @@ TEST(space_to_batch_fp16_gpu, i1222_bs1222_pb0000_pe0000) {
     // Output :      8x1x1x1
     // Input values in fp16
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, {1,2,2,2} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, {1,2,2,2} });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f),
@@ -37,7 +34,7 @@ TEST(space_to_batch_fp16_gpu, i1222_bs1222_pb0000_pe0000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(space_to_batch("space_to_batch", "Input", tensor(format::bfyx, {1,2,2,2}, 1),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
@@ -49,7 +46,7 @@ TEST(space_to_batch_fp16_gpu, i1222_bs1222_pb0000_pe0000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_batch").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f
@@ -70,8 +67,8 @@ TEST(space_to_batch_fp16_gpu, i1242_bs1221_pb0020_pe0000) {
     // Output :      4x1x3x2
     // Input values in fp16
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, {1,2,2,4} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, {1,2,2,4} });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f),
@@ -81,7 +78,7 @@ TEST(space_to_batch_fp16_gpu, i1242_bs1221_pb0020_pe0000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(space_to_batch("space_to_batch", "Input", tensor(format::bfyx, {1,2,2,1}, 1),
                                                            tensor(format::bfyx, {0,0,2,0}, 0),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
@@ -93,7 +90,7 @@ TEST(space_to_batch_fp16_gpu, i1242_bs1221_pb0020_pe0000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_batch").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 0.f, 0.f, 1.f, 4.f, 5.f,
@@ -117,8 +114,8 @@ TEST(space_to_batch_fp16_gpu, i2132_bs1222_pb0010_pe0100) {
     // Output :      16x1x2x1
     // Input values in fp16
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, {2,1,2,3} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, {2,1,2,3} });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f),
@@ -127,7 +124,7 @@ TEST(space_to_batch_fp16_gpu, i2132_bs1222_pb0010_pe0100) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(space_to_batch("space_to_batch", "Input", tensor(format::bfyx, {1,2,2,2}, 1),
                                                            tensor(format::bfyx, {0,0,1,0}, 0),
                                                            tensor(format::bfyx, {0,1,0,0}, 0),
@@ -139,7 +136,7 @@ TEST(space_to_batch_fp16_gpu, i2132_bs1222_pb0010_pe0100) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_batch").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 2.f, 0.f, 8.f, 0.f, 3.f, 0.f, 9.f,
@@ -163,8 +160,8 @@ TEST(space_to_batch_fp16_gpu, i12132_bs12122_pb00010_pe00000) {
     // Output :      8x1x1x2x1
     // Input values in fp16
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f16, format::bfzyx, {1,2,2,3,1} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f16, format::bfzyx, {1,2,2,3,1} });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f),
@@ -173,7 +170,7 @@ TEST(space_to_batch_fp16_gpu, i12132_bs12122_pb00010_pe00000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(space_to_batch("space_to_batch", "Input", tensor(format::bfzyx, {1,2,1,2,2}, 1),
                                                            tensor(format::bfzyx, {0,0,0,1,0}, 0),
                                                            tensor(format::bfzyx, {0,0,0,0,0}, 0),
@@ -185,7 +182,7 @@ TEST(space_to_batch_fp16_gpu, i12132_bs12122_pb00010_pe00000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_batch").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 2.f, 0.f, 3.f, 0.f, 4.f, 1.f, 5.f,
@@ -207,9 +204,9 @@ TEST(space_to_batch_fp16_gpu, i134121_bs142121_pb010100_pe000000) {
     // Output :      16x1x2x2x1x1
     // Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{ batch(1), feature(3), spatial(1, 2, 1, 4) };
-    auto input = memory::allocate(engine, { data_types::f16, format::bfwzyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f16, format::bfwzyx, input_shape });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f),
@@ -221,7 +218,7 @@ TEST(space_to_batch_fp16_gpu, i134121_bs142121_pb010100_pe000000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(space_to_batch("space_to_batch", "Input", tensor(format::bfwzyx, {1,4,2,1,2,1}, 1),
                                                            tensor(format::bfwzyx, {0,1,0,1,0,0}, 0),
                                                            tensor(format::bfwzyx, {0,0,0,0,0,0}, 0),
@@ -233,7 +230,7 @@ TEST(space_to_batch_fp16_gpu, i134121_bs142121_pb010100_pe000000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_batch").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
@@ -261,8 +258,8 @@ TEST(space_to_batch_fp16_gpu, i11611_bs1222_pb0010_pe0001_b_fs_yx_fsv16) {
     // Output :      8x8x1x1
     // Input values in fp16
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, {1,16,1,1} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, {1,16,1,1} });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f),
@@ -272,7 +269,7 @@ TEST(space_to_batch_fp16_gpu, i11611_bs1222_pb0010_pe0001_b_fs_yx_fsv16) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(reorder("input_fsv", "Input", format::b_fs_yx_fsv16, data_types::f16));
     topology.add(space_to_batch("space_to_batch", "input_fsv", tensor(format::bfyx, {1,2,2,2}, 1),
                                                                tensor(format::bfyx, {0,0,1,0}, 0),
@@ -286,7 +283,7 @@ TEST(space_to_batch_fp16_gpu, i11611_bs1222_pb0010_pe0001_b_fs_yx_fsv16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("stb_to_bfyx").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
@@ -314,8 +311,8 @@ TEST(space_to_batch_fp16_gpu, i1812_bs1221_pb0010_pe0200_b_fs_yx_fsv16) {
     // Output :      4x5x1x2
     // Input values in fp16
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f16, format::bfyx, {1,8,2,1} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f16, format::bfyx, {1,8,2,1} });
 
     set_values(input, {
         FLOAT16(0.0f), FLOAT16(1.0f), FLOAT16(2.0f), FLOAT16(3.0f),
@@ -325,7 +322,7 @@ TEST(space_to_batch_fp16_gpu, i1812_bs1221_pb0010_pe0200_b_fs_yx_fsv16) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(reorder("input_fsv", "Input", format::b_fs_yx_fsv16, data_types::f16));
     topology.add(space_to_batch("space_to_batch", "input_fsv", tensor(format::bfyx, {1,2,2,1}, 1),
                                                                tensor(format::bfyx, {0,0,1,0}, 0),
@@ -339,7 +336,7 @@ TEST(space_to_batch_fp16_gpu, i1812_bs1221_pb0010_pe0200_b_fs_yx_fsv16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("stb_to_bfyx").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
@@ -363,8 +360,8 @@ TEST(space_to_batch_fp32_gpu, i1222_bs1222_pb0000_pe0000) {
     // Output :      8x1x1x1
     // Input values in fp32
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, {1,2,2,2} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, {1,2,2,2} });
 
     set_values(input, {
         0.0f, 1.0f, 2.0f, 3.0f,
@@ -372,7 +369,7 @@ TEST(space_to_batch_fp32_gpu, i1222_bs1222_pb0000_pe0000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(space_to_batch("space_to_batch", "Input", tensor(format::bfyx, {1,2,2,2}, 1),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
@@ -384,7 +381,7 @@ TEST(space_to_batch_fp32_gpu, i1222_bs1222_pb0000_pe0000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_batch").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f
@@ -405,8 +402,8 @@ TEST(space_to_batch_fp32_gpu, i1242_bs1221_pb0020_pe0000) {
     // Output :      4x1x3x2
     // Input values in fp32
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, {1,2,2,4} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, {1,2,2,4} });
 
     set_values(input, {
         0.0f, 1.0f, 2.0f, 3.0f,
@@ -416,7 +413,7 @@ TEST(space_to_batch_fp32_gpu, i1242_bs1221_pb0020_pe0000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(space_to_batch("space_to_batch", "Input", tensor(format::bfyx, {1,2,2,1}, 1),
                                                            tensor(format::bfyx, {0,0,2,0}, 0),
                                                            tensor(format::bfyx, {0,0,0,0}, 0),
@@ -428,7 +425,7 @@ TEST(space_to_batch_fp32_gpu, i1242_bs1221_pb0020_pe0000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_batch").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 0.f, 0.f, 1.f, 4.f, 5.f,
@@ -452,8 +449,8 @@ TEST(space_to_batch_fp32_gpu, i2132_bs1222_pb0010_pe0100) {
     // Output :      16x1x2x1
     // Input values in fp32
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, {2,1,2,3} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, {2,1,2,3} });
 
     set_values(input, {
         0.0f, 1.0f, 2.0f, 3.0f,
@@ -462,7 +459,7 @@ TEST(space_to_batch_fp32_gpu, i2132_bs1222_pb0010_pe0100) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(space_to_batch("space_to_batch", "Input", tensor(format::bfyx, {1,2,2,2}, 1),
                                                            tensor(format::bfyx, {0,0,1,0}, 0),
                                                            tensor(format::bfyx, {0,1,0,0}, 0),
@@ -474,7 +471,7 @@ TEST(space_to_batch_fp32_gpu, i2132_bs1222_pb0010_pe0100) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_batch").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 2.f, 0.f, 8.f, 0.f, 3.f, 0.f, 9.f,
@@ -498,8 +495,8 @@ TEST(space_to_batch_fp32_gpu, i12132_bs12122_pb00010_pe00000) {
     // Output :      8x1x1x2x1
     // Input values in fp32
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, {1,2,2,3,1} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, {1,2,2,3,1} });
 
     set_values(input, {
         0.0f, 1.0f, 2.0f, 3.0f,
@@ -508,7 +505,7 @@ TEST(space_to_batch_fp32_gpu, i12132_bs12122_pb00010_pe00000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(space_to_batch("space_to_batch", "Input", tensor(format::bfzyx, {1,2,1,2,2}, 1),
                                                            tensor(format::bfzyx, {0,0,0,1,0}, 0),
                                                            tensor(format::bfzyx, {0,0,0,0,0}, 0),
@@ -520,7 +517,7 @@ TEST(space_to_batch_fp32_gpu, i12132_bs12122_pb00010_pe00000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_batch").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 2.f, 0.f, 3.f, 0.f, 4.f, 1.f, 5.f,
@@ -542,9 +539,9 @@ TEST(space_to_batch_fp32_gpu, i134121_bs142121_pb010100_pe000000) {
     // Output :      16x1x2x2x1x1
     // Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
     tensor input_shape = tensor{ batch(1), feature(3), spatial(1, 2, 1, 4) };
-    auto input = memory::allocate(engine, { data_types::f32, format::bfwzyx, input_shape });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfwzyx, input_shape });
 
     set_values(input, {
        0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f,
@@ -554,7 +551,7 @@ TEST(space_to_batch_fp32_gpu, i134121_bs142121_pb010100_pe000000) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(space_to_batch("space_to_batch", "Input", tensor(format::bfwzyx, {1,4,2,1,2,1}, 1),
                                                            tensor(format::bfwzyx, {0,1,0,1,0,0}, 0),
                                                            tensor(format::bfwzyx, {0,0,0,0,0,0}, 0),
@@ -566,7 +563,7 @@ TEST(space_to_batch_fp32_gpu, i134121_bs142121_pb010100_pe000000) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_batch").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
@@ -594,8 +591,8 @@ TEST(space_to_batch_fp32_gpu, i11622_bs1421_pb0000_pe0000_b_fs_yx_fsv16) {
     // Output :      8x4x1x2
     // Input values in fp32
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, {1,16,2,2} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, {1,16,2,2} });
 
     set_values(input, {
         0.0f,  1.0f,  2.0f,  3.0f,  4.0f,  5.0f,  6.0f,  7.0f,
@@ -609,7 +606,7 @@ TEST(space_to_batch_fp32_gpu, i11622_bs1421_pb0000_pe0000_b_fs_yx_fsv16) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(reorder("input_fsv", "Input", format::b_fs_yx_fsv16, data_types::f32));
     topology.add(space_to_batch("space_to_batch", "input_fsv", tensor(format::bfyx, {1,4,2,1}, 1),
                                                                tensor(format::bfyx, {0,0,0,0}, 0),
@@ -623,7 +620,7 @@ TEST(space_to_batch_fp32_gpu, i11622_bs1421_pb0000_pe0000_b_fs_yx_fsv16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("stb_to_bfyx").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.f, 16.f, 17.f, 32.f, 33.f, 48.f, 49.f,
@@ -651,8 +648,8 @@ TEST(space_to_batch_fp32_gpu, i1623_bs1312_pb0001_pe0000_b_fs_yx_fsv16) {
     // Output :      6x2x2x2
     // Input values in fp32
 
-    engine engine;
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, {1,6,3,2} });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, {1,6,3,2} });
 
     set_values(input, {
         0.0f,  1.0f,  2.0f,  3.0f,  4.0f,  5.0f,  6.0f,  7.0f,
@@ -663,7 +660,7 @@ TEST(space_to_batch_fp32_gpu, i1623_bs1312_pb0001_pe0000_b_fs_yx_fsv16) {
     });
 
     topology topology;
-    topology.add(input_layout("Input", input.get_layout()));
+    topology.add(input_layout("Input", input->get_layout()));
     topology.add(reorder("input_fsv", "Input", format::b_fs_yx_fsv16, data_types::f32));
     topology.add(space_to_batch("space_to_batch", "input_fsv", tensor(format::bfyx, {1,3,1,2}, 1),
                                                                tensor(format::bfyx, {0,0,0,1}, 0),
@@ -677,7 +674,7 @@ TEST(space_to_batch_fp32_gpu, i1623_bs1312_pb0001_pe0000_b_fs_yx_fsv16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("stb_to_bfyx").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
         0.f, 1.f, 0.f, 4.f, 0.f, 19.f, 0.f, 22.f,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/space_to_depth_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/space_to_depth_gpu_test.cpp
index 76611b0f60f1aa..6614d4d6b75e8a 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/space_to_depth_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/space_to_depth_gpu_test.cpp
@@ -3,16 +3,13 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
 
-#include <api/input_layout.hpp>
-#include <api/memory.hpp>
-#include <api/space_to_depth.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/space_to_depth.hpp>
 
 #include <cstddef>
-#include <tests/test_utils/test_utils.h>
 
 using namespace cldnn;
 using namespace ::tests;
@@ -27,9 +24,9 @@ TEST(space_to_depth_fp16_gpu, d1122_bs2_mbf) {
     //  Output : 1x4x1x1
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 1, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 1, 2, 2 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -38,7 +35,7 @@ TEST(space_to_depth_fp16_gpu, d1122_bs2_mbf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::blocks_first, block_size)
     );
@@ -50,7 +47,7 @@ TEST(space_to_depth_fp16_gpu, d1122_bs2_mbf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 1.f, 2.f, 3.f
@@ -67,9 +64,9 @@ TEST(space_to_depth_fp16_gpu, d1142_bs2_mbf) {
     //  Output : 1x4x2x1
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 1, 2, 4 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 1, 2, 4 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -80,7 +77,7 @@ TEST(space_to_depth_fp16_gpu, d1142_bs2_mbf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::blocks_first, block_size)
     );
@@ -92,7 +89,7 @@ TEST(space_to_depth_fp16_gpu, d1142_bs2_mbf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 4.0f, 1.0f, 5.0f, 2.0f, 6.0f, 3.0f, 7.0f
@@ -109,9 +106,9 @@ TEST(space_to_depth_fp16_gpu, d1264_bs2_mbf) {
     //  Output : 1x8x3x2
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 2, 4, 6 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 2, 4, 6 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -128,7 +125,7 @@ TEST(space_to_depth_fp16_gpu, d1264_bs2_mbf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
     space_to_depth("space_to_depth", "Input0", space_to_depth::blocks_first, block_size)
     );
@@ -140,7 +137,7 @@ TEST(space_to_depth_fp16_gpu, d1264_bs2_mbf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 2.0f, 8.0f, 10.0f, 16.0f, 18.0f,
@@ -164,9 +161,9 @@ TEST(space_to_depth_fp16_gpu, d1199_bs3_mbf) {
     //  Output : 1x9x3x3
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 1, 9, 9 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 1, 9, 9 } });
     size_t block_size = 3;
 
     set_values(input1, {
@@ -190,7 +187,7 @@ TEST(space_to_depth_fp16_gpu, d1199_bs3_mbf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::blocks_first, block_size)
     );
@@ -202,7 +199,7 @@ TEST(space_to_depth_fp16_gpu, d1199_bs3_mbf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 3.0f, 6.0f, 27.0f, 30.0f, 33.0f, 54.0f, 57.0f, 60.0f, 1.0f,
@@ -227,9 +224,9 @@ TEST(space_to_depth_fp32_gpu, d1122_bs2_mbf) {
     //  Output : 1x4x1x1
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -237,7 +234,7 @@ TEST(space_to_depth_fp32_gpu, d1122_bs2_mbf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::blocks_first, block_size)
     );
@@ -249,7 +246,7 @@ TEST(space_to_depth_fp32_gpu, d1122_bs2_mbf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 1.f, 2.f, 3.f
@@ -266,9 +263,9 @@ TEST(space_to_depth_fp32_gpu, d1142_bs2_mbf) {
     //  Output : 1x4x2x1
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 4 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 4 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -276,7 +273,7 @@ TEST(space_to_depth_fp32_gpu, d1142_bs2_mbf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::blocks_first, block_size)
     );
@@ -288,7 +285,7 @@ TEST(space_to_depth_fp32_gpu, d1142_bs2_mbf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 4.0f, 1.0f, 5.0f, 2.0f, 6.0f, 3.0f, 7.0f
@@ -305,9 +302,9 @@ TEST(space_to_depth_fp32_gpu, d1264_bs2_mbf) {
     //  Output : 1x8x3x2
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 4, 6 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 4, 6 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -324,7 +321,7 @@ TEST(space_to_depth_fp32_gpu, d1264_bs2_mbf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
         space_to_depth("space_to_depth", "Input0", space_to_depth::blocks_first, block_size)
     );
@@ -336,7 +333,7 @@ TEST(space_to_depth_fp32_gpu, d1264_bs2_mbf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 2.0f, 8.0f, 10.0f, 16.0f, 18.0f,
@@ -360,9 +357,9 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mbf) {
     //  Output : 1x9x3x3
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 9, 9 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 9, 9 } });
     size_t block_size = 3;
 
     set_values(input1, {
@@ -378,7 +375,7 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mbf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::blocks_first, block_size)
     );
@@ -390,7 +387,7 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mbf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 3.0f, 6.0f, 27.0f, 30.0f, 33.0f, 54.0f, 57.0f, 60.0f, 1.0f,
@@ -419,9 +416,9 @@ TEST(space_to_depth_fp16_gpu, d1122_bs2_mdf) {
     //  Output : 1x4x1x1
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 1, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 1, 2, 2 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -430,7 +427,7 @@ TEST(space_to_depth_fp16_gpu, d1122_bs2_mdf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::depth_first, block_size)
     );
@@ -442,7 +439,7 @@ TEST(space_to_depth_fp16_gpu, d1122_bs2_mdf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 1.f, 2.f, 3.f
@@ -459,9 +456,9 @@ TEST(space_to_depth_fp16_gpu, d1142_bs2_mdf) {
     //  Output : 1x4x2x1
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 1, 2, 4 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 1, 2, 4 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -472,7 +469,7 @@ TEST(space_to_depth_fp16_gpu, d1142_bs2_mdf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::depth_first, block_size)
     );
@@ -484,7 +481,7 @@ TEST(space_to_depth_fp16_gpu, d1142_bs2_mdf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 4.0f, 1.0f, 5.0f, 2.0f, 6.0f, 3.0f, 7.0f
@@ -501,9 +498,9 @@ TEST(space_to_depth_fp16_gpu, d1264_bs2_mdf) {
     //  Output : 1x8x3x2
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 2, 4, 6 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 2, 4, 6 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -520,7 +517,7 @@ TEST(space_to_depth_fp16_gpu, d1264_bs2_mdf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::depth_first, block_size)
     );
@@ -532,7 +529,7 @@ TEST(space_to_depth_fp16_gpu, d1264_bs2_mdf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 2.0f, 8.0f, 10.0f, 16.0f, 18.0f,
@@ -556,9 +553,9 @@ TEST(space_to_depth_fp16_gpu, d1199_bs3_mdf) {
     //  Output : 1x9x3x3
     //  Input values in fp16
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f16, format::bfyx, { 1, 1, 9, 9 } });
+    auto input1 = engine.allocate_memory({ data_types::f16, format::bfyx, { 1, 1, 9, 9 } });
     size_t block_size = 3;
 
     set_values(input1, {
@@ -582,7 +579,7 @@ TEST(space_to_depth_fp16_gpu, d1199_bs3_mdf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::depth_first, block_size)
     );
@@ -594,7 +591,7 @@ TEST(space_to_depth_fp16_gpu, d1199_bs3_mdf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<uint16_t>();
+    cldnn::mem_lock<uint16_t> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 3.0f, 6.0f, 27.0f, 30.0f, 33.0f, 54.0f, 57.0f, 60.0f, 1.0f,
@@ -619,9 +616,9 @@ TEST(space_to_depth_fp32_gpu, d1122_bs2_mdf) {
     //  Output : 1x4x1x1
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 2 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -629,7 +626,7 @@ TEST(space_to_depth_fp32_gpu, d1122_bs2_mdf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::depth_first, block_size)
     );
@@ -641,7 +638,7 @@ TEST(space_to_depth_fp32_gpu, d1122_bs2_mdf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.f, 1.f, 2.f, 3.f
@@ -658,9 +655,9 @@ TEST(space_to_depth_fp32_gpu, d1142_bs2_mdf) {
     //  Output : 1x4x2x1
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 2, 4 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 2, 4 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -668,7 +665,7 @@ TEST(space_to_depth_fp32_gpu, d1142_bs2_mdf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::depth_first, block_size)
     );
@@ -680,7 +677,7 @@ TEST(space_to_depth_fp32_gpu, d1142_bs2_mdf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 4.0f, 1.0f, 5.0f, 2.0f, 6.0f, 3.0f, 7.0f
@@ -697,9 +694,9 @@ TEST(space_to_depth_fp32_gpu, d1264_bs2_mdf) {
     //  Output : 1x8x3x2
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 4, 6 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 4, 6 } });
     size_t block_size = 2;
 
     set_values(input1, {
@@ -716,7 +713,7 @@ TEST(space_to_depth_fp32_gpu, d1264_bs2_mdf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::depth_first, block_size)
     );
@@ -728,7 +725,7 @@ TEST(space_to_depth_fp32_gpu, d1264_bs2_mdf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 2.0f, 8.0f, 10.0f, 16.0f, 18.0f,
@@ -752,9 +749,9 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mdf) {
     //  Output : 1x9x3x3
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 9, 9 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 9, 9 } });
     size_t block_size = 3;
 
     set_values(input1, {
@@ -770,7 +767,7 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mdf) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(
             space_to_depth("space_to_depth", "Input0", space_to_depth::depth_first, block_size)
     );
@@ -782,7 +779,7 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mdf) {
     auto outputs = network.execute();
 
     auto output = outputs.at("space_to_depth").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 3.0f, 6.0f, 27.0f, 30.0f, 33.0f, 54.0f, 57.0f, 60.0f, 1.0f,
@@ -807,9 +804,9 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mdf_fsv16) {
     //  Output : 1x9x3x3
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 9, 9 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 9, 9 } });
     size_t block_size = 3;
 
     set_values(input1, {
@@ -825,7 +822,7 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mdf_fsv16) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(reorder("reorder", "Input0", format::b_fs_yx_fsv16, data_types::f32));
     topology.add(space_to_depth("space_to_depth", "reorder", space_to_depth::depth_first, block_size));
     topology.add(reorder("reorder_out", "space_to_depth", format::bfyx, data_types::f32));
@@ -837,7 +834,7 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mdf_fsv16) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reorder_out").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 3.0f, 6.0f, 27.0f, 30.0f, 33.0f, 54.0f, 57.0f, 60.0f, 1.0f,
@@ -862,9 +859,9 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mdf_fsv4) {
     //  Output : 1x9x3x3
     //  Input values in fp32
 
-    engine engine;
+    auto& engine = get_test_engine();
 
-    auto input1 = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 9, 9 } });
+    auto input1 = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 9, 9 } });
     size_t block_size = 3;
 
     set_values(input1, {
@@ -880,7 +877,7 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mdf_fsv4) {
     });
 
     topology topology;
-    topology.add(input_layout("Input0", input1.get_layout()));
+    topology.add(input_layout("Input0", input1->get_layout()));
     topology.add(reorder("reorder", "Input0", format::b_fs_yx_fsv4, data_types::f32));
     topology.add(space_to_depth("space_to_depth", "reorder", space_to_depth::depth_first, block_size));
     topology.add(reorder("reorder_out", "space_to_depth", format::bfyx, data_types::f32));
@@ -892,7 +889,7 @@ TEST(space_to_depth_fp32_gpu, d1199_bs3_mdf_fsv4) {
     auto outputs = network.execute();
 
     auto output = outputs.at("reorder_out").get_memory();
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     std::vector<float> expected_results = {
             0.0f, 3.0f, 6.0f, 27.0f, 30.0f, 33.0f, 54.0f, 57.0f, 60.0f, 1.0f,
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/spatial_concatenate_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/spatial_concatenate_gpu_test.cpp
index 27dcab87f8bff5..286588efefa928 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/spatial_concatenate_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/spatial_concatenate_gpu_test.cpp
@@ -4,23 +4,19 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/concatenation.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include "cldnn/primitives/concatenation.hpp"
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 TEST(spatial_concatenate_f32_gpu, test01) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
 
     set_values(input1, {
         1.0f, 2.0f,
@@ -38,11 +34,11 @@ TEST(spatial_concatenate_f32_gpu, test01) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2" }, concatenation::along_x));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
 
@@ -50,16 +46,16 @@ TEST(spatial_concatenate_f32_gpu, test01) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0] + input2.get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0] + input2->get_layout().size.spatial[0]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
@@ -70,10 +66,10 @@ TEST(spatial_concatenate_f32_gpu, test01) {
 }
 
 TEST(spatial_concatenate_f32_gpu, test02) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfyx,{ 1,1,2,2 } });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfyx,{ 1,1,2,2 } });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfyx,{ 1,1,2,2 } });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfyx,{ 1,1,2,2 } });
 
     set_values(input1, {
         1.0f, 2.0f,
@@ -93,11 +89,11 @@ TEST(spatial_concatenate_f32_gpu, test02) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2" }, concatenation::along_y));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
 
@@ -105,16 +101,16 @@ TEST(spatial_concatenate_f32_gpu, test02) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1] + input2.get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1] + input2->get_layout().size.spatial[1]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
@@ -125,10 +121,10 @@ TEST(spatial_concatenate_f32_gpu, test02) {
 }
 
 TEST(spatial_concatenate_f32_gpu, test03) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfyx,{ 1,1,2,2 } });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfyx,{ 1,1,2,2 } });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfyx,{ 1,1,2,2 } });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfyx,{ 1,1,2,2 } });
 
     set_values(input1, {
         1.0f, 2.0f,
@@ -150,11 +146,11 @@ TEST(spatial_concatenate_f32_gpu, test03) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2" }, concatenation::along_y, padding({ 0, 0, 1, 1 }, 0.0f)));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
 
@@ -162,16 +158,16 @@ TEST(spatial_concatenate_f32_gpu, test03) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1] + input2.get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1] + input2->get_layout().size.spatial[1]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
@@ -182,10 +178,10 @@ TEST(spatial_concatenate_f32_gpu, test03) {
 }
 
 TEST(spatial_concatenate_f32_gpu, test04) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfyx,{ 1,1,2,2 }, padding({ 0,0,0,0 }, { 0,0,1,0 }) });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfyx,{ 1,1,2,2 }, padding({ 0,0,0,1 }, 0.0f) });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfyx,{ 1,1,2,2 }, padding({ 0,0,0,0 }, { 0,0,1,0 }) });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfyx,{ 1,1,2,2 }, padding({ 0,0,0,1 }, 0.0f) });
 
     set_values(input1, {
         1.0f, 2.0f, 0.0f,
@@ -205,11 +201,11 @@ TEST(spatial_concatenate_f32_gpu, test04) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2" }, concatenation::along_x, padding({ 0,0,2,0 }, { 0,0,0,0 })));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
 
@@ -217,16 +213,16 @@ TEST(spatial_concatenate_f32_gpu, test04) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0] + input2.get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0] + input2->get_layout().size.spatial[0]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
@@ -237,11 +233,11 @@ TEST(spatial_concatenate_f32_gpu, test04) {
 }
 
 TEST(spatial_concatenate_f32_gpu, inputs_3) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
-    memory input3 = memory::allocate(eng, layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
+    memory::ptr input3 = engine.allocate_memory(layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
 
     set_values(input1, {
         1.0f, 2.0f,
@@ -264,12 +260,12 @@ TEST(spatial_concatenate_f32_gpu, inputs_3) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
-    tpl.add(input_layout("in3", input3.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
+    tpl.add(input_layout("in3", input3->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2", "in3" }, concatenation::along_x));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
     net.set_input_data("in3", input3);
@@ -278,16 +274,16 @@ TEST(spatial_concatenate_f32_gpu, inputs_3) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0] + input2.get_layout().size.spatial[0] + input3.get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0] + input2->get_layout().size.spatial[0] + input3->get_layout().size.spatial[0]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
@@ -298,11 +294,11 @@ TEST(spatial_concatenate_f32_gpu, inputs_3) {
 }
 
 TEST(spatial_concatenate_f32_gpu, inputs_3_uneven_axis_b) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfyx, { 3,1,2,2 } });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
-    memory input3 = memory::allocate(eng, layout{ data_types::f32, format::bfyx, { 2,1,2,2 } });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfyx, { 3,1,2,2 } });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfyx, { 1,1,2,2 } });
+    memory::ptr input3 = engine.allocate_memory(layout{ data_types::f32, format::bfyx, { 2,1,2,2 } });
 
     set_values(input1, {
         1.0f, 2.0f,
@@ -352,12 +348,12 @@ TEST(spatial_concatenate_f32_gpu, inputs_3_uneven_axis_b) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
-    tpl.add(input_layout("in3", input3.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
+    tpl.add(input_layout("in3", input3->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2", "in3" }, concatenation::along_b));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
     net.set_input_data("in3", input3);
@@ -366,16 +362,16 @@ TEST(spatial_concatenate_f32_gpu, inputs_3_uneven_axis_b) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0] + input2.get_layout().size.batch[0] + input3.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0] + input2->get_layout().size.batch[0] + input3->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
@@ -386,10 +382,10 @@ TEST(spatial_concatenate_f32_gpu, inputs_3_uneven_axis_b) {
 }
 
 TEST(spatial_concatenate_f32_gpu, inputs3d_axis_x) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
 
     set_values(input1, {
         1.0f, 2.0f,
@@ -413,11 +409,11 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_x) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2" }, concatenation::along_x));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
 
@@ -425,17 +421,17 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_x) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0] + input2.get_layout().size.spatial[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1]);
-    ASSERT_EQ(output_layout.size.spatial[2], input1.get_layout().size.spatial[2]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0] + input2->get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.spatial[2], input1->get_layout().size.spatial[2]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
@@ -446,10 +442,10 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_x) {
 }
 
 TEST(spatial_concatenate_f32_gpu, inputs3d_axis_y) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
 
     set_values(input1, {
         1.0f, 2.0f,
@@ -477,11 +473,11 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_y) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2" }, concatenation::along_y));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
 
@@ -489,17 +485,17 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_y) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1] + input2.get_layout().size.spatial[1]);
-    ASSERT_EQ(output_layout.size.spatial[2], input1.get_layout().size.spatial[2]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1] + input2->get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.spatial[2], input1->get_layout().size.spatial[2]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
@@ -510,10 +506,10 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_y) {
 }
 
 TEST(spatial_concatenate_f32_gpu, inputs3d_axis_z) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
 
     set_values(input1, {
         1.0f, 2.0f,
@@ -541,11 +537,11 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_z) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2" }, concatenation::along_z));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
 
@@ -553,17 +549,17 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_z) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1]);
-    ASSERT_EQ(output_layout.size.spatial[2], input1.get_layout().size.spatial[2] + input2.get_layout().size.spatial[2]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.spatial[2], input1->get_layout().size.spatial[2] + input2->get_layout().size.spatial[2]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
@@ -574,10 +570,10 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_z) {
 }
 
 TEST(spatial_concatenate_f32_gpu, inputs3d_axis_b) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 2,1,2,2,2 } });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 2,1,2,2,2 } });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
 
     set_values(input1, {
         1.0f, 2.0f,
@@ -616,11 +612,11 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_b) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2" }, concatenation::along_b));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
 
@@ -628,17 +624,17 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_b) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0] + input2.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1]);
-    ASSERT_EQ(output_layout.size.spatial[2], input1.get_layout().size.spatial[2]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0] + input2->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.spatial[2], input1->get_layout().size.spatial[2]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
@@ -649,11 +645,11 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_axis_b) {
 }
 
 TEST(spatial_concatenate_f32_gpu, inputs3d_3_uneven_axis_b) {
-    engine eng;
+    auto& engine = get_test_engine();
 
-    memory input1 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 3,1,2,2,2 } });
-    memory input2 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
-    memory input3 = memory::allocate(eng, layout{ data_types::f32, format::bfzyx, { 2,1,2,2,2 } });
+    memory::ptr input1 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 3,1,2,2,2 } });
+    memory::ptr input2 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 1,1,2,2,2 } });
+    memory::ptr input3 = engine.allocate_memory(layout{ data_types::f32, format::bfzyx, { 2,1,2,2,2 } });
 
     set_values(input1, {
         //b0
@@ -745,12 +741,12 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_3_uneven_axis_b) {
     };
 
     topology tpl;
-    tpl.add(input_layout("in1", input1.get_layout()));
-    tpl.add(input_layout("in2", input2.get_layout()));
-    tpl.add(input_layout("in3", input3.get_layout()));
+    tpl.add(input_layout("in1", input1->get_layout()));
+    tpl.add(input_layout("in2", input2->get_layout()));
+    tpl.add(input_layout("in3", input3->get_layout()));
     tpl.add(concatenation("conc", { "in1", "in2", "in3" }, concatenation::along_b));
 
-    network net(eng, tpl);
+    network net(engine, tpl);
     net.set_input_data("in1", input1);
     net.set_input_data("in2", input2);
     net.set_input_data("in3", input3);
@@ -759,17 +755,17 @@ TEST(spatial_concatenate_f32_gpu, inputs3d_3_uneven_axis_b) {
     ASSERT_TRUE(outputs.size() == 1 && outputs.count("conc") == 1);
 
     auto output_mem = outputs.at("conc").get_memory();
-    auto output_layout = output_mem.get_layout();
+    auto output_layout = output_mem->get_layout();
 
-    ASSERT_EQ(output_layout.size.batch[0], input1.get_layout().size.batch[0] + input2.get_layout().size.batch[0] + input3.get_layout().size.batch[0]);
-    ASSERT_EQ(output_layout.size.feature[0], input1.get_layout().size.feature[0]);
-    ASSERT_EQ(output_layout.size.spatial[0], input1.get_layout().size.spatial[0]);
-    ASSERT_EQ(output_layout.size.spatial[1], input1.get_layout().size.spatial[1]);
-    ASSERT_EQ(output_layout.size.spatial[2], input1.get_layout().size.spatial[2]);
+    ASSERT_EQ(output_layout.size.batch[0], input1->get_layout().size.batch[0] + input2->get_layout().size.batch[0] + input3->get_layout().size.batch[0]);
+    ASSERT_EQ(output_layout.size.feature[0], input1->get_layout().size.feature[0]);
+    ASSERT_EQ(output_layout.size.spatial[0], input1->get_layout().size.spatial[0]);
+    ASSERT_EQ(output_layout.size.spatial[1], input1->get_layout().size.spatial[1]);
+    ASSERT_EQ(output_layout.size.spatial[2], input1->get_layout().size.spatial[2]);
 
-    ASSERT_EQ(output_mem.get_layout().get_linear_size(), expected_output.size());
+    ASSERT_EQ(output_mem->get_layout().get_linear_size(), expected_output.size());
     {
-        auto out_ptr = output_mem.pointer<const float>();
+        cldnn::mem_lock<const float> out_ptr(output_mem, get_test_stream());
 
         size_t idx = 0;
         for (auto const& value : out_ptr)
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/split_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/split_gpu_test.cpp
index 0d420abfefabcd..10cb13acc3137d 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/split_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/split_gpu_test.cpp
@@ -2,23 +2,18 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/split.hpp"
-#include "api/scale.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include <api/reorder.hpp>
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/split.hpp>
+#include <cldnn/primitives/scale.hpp>
+#include <cldnn/primitives/reorder.hpp>
 
 #include <sstream>
 #include <iomanip>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 template<typename T>
 std::vector<T> generate_random_input(size_t b, size_t f, size_t y, size_t x, int min, int max) {
@@ -34,7 +29,7 @@ std::vector<T> generate_random_input(size_t b, size_t f, size_t y, size_t x, int
 }
 
 template<typename T>
-void check_feature_map(cldnn::pointer<T> output_ptr, std::vector<T> &input_vec, size_t batch_num, size_t feature_num, size_t y_size, size_t x_size, size_t feature_id, size_t factor)
+void check_feature_map(T* output_ptr, std::vector<T> &input_vec, size_t batch_num, size_t feature_num, size_t y_size, size_t x_size, size_t feature_id, size_t factor)
 {
     for (size_t b = 0; b < batch_num; ++b) { //B
         for (size_t y = 0; y < y_size; ++y) { //Y
@@ -50,15 +45,15 @@ void check_feature_map(cldnn::pointer<T> output_ptr, std::vector<T> &input_vec,
 template<typename T>
 void split_test(int batch_num, int feature_num, int x_size, int y_size, std::vector<cldnn::tensor> split_offsets)
 {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     cldnn::tensor reference_input_size = { batch_num, feature_num, x_size, y_size };
 
-    cldnn::memory input = memory::allocate(engine, { type_to_data_type<T>::value, format::bfyx, reference_input_size });
+    cldnn::memory::ptr input = engine.allocate_memory({ type_to_data_type<T>::value, format::bfyx, reference_input_size });
     std::vector<std::pair<primitive_id, cldnn::tensor> > input_ids_offsets;
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    
+    topology.add(input_layout("input", input->get_layout()));
+
     // lambda exoression to create the primitive id for the splits
     auto create_split_id = [](size_t splitNum) {
         std::stringstream ss;
@@ -68,7 +63,7 @@ void split_test(int batch_num, int feature_num, int x_size, int y_size, std::vec
     };
 
     // Create the splits with the split ids for the topology
-    for (size_t splitNum = 0; splitNum < split_offsets.size(); splitNum++) 
+    for (size_t splitNum = 0; splitNum < split_offsets.size(); splitNum++)
     {
         input_ids_offsets.push_back({ create_split_id(splitNum), split_offsets[splitNum]});
     }
@@ -85,7 +80,7 @@ void split_test(int batch_num, int feature_num, int x_size, int y_size, std::vec
 
     // The number of splits should match the expected number of splits
     EXPECT_EQ(outputs.size(), size_t(split_offsets.size()));
-    
+
     std::vector<cldnn::tensor> expected_sizes;
     for (size_t splitNum = 0; splitNum < split_offsets.size(); splitNum++)  // Calculate the expected sizes
     {
@@ -110,15 +105,15 @@ void split_test(int batch_num, int feature_num, int x_size, int y_size, std::vec
         expected_sizes.push_back(size);
     }
 
-    pointer<T> input_ptr = input.pointer<T>();
+    cldnn::mem_lock<T> input_ptr(input, get_test_stream());
 
     for (size_t splitNum = 0; splitNum < split_offsets.size(); splitNum++)
     {
         primitive_id split_id = "split:" + create_split_id(splitNum);
-        cldnn::memory output = outputs.at(split_id).get_memory();
-        auto prim = output.get_layout();
+        cldnn::memory::ptr output = outputs.at(split_id).get_memory();
+        auto prim = output->get_layout();
         EXPECT_EQ(prim.size, expected_sizes[splitNum]);
-        auto output_ptr = output.pointer<T>();
+        cldnn::mem_lock<T> output_ptr(output, get_test_stream());
 
         // Output tensor size
         auto output_batch = prim.size.batch[0];
@@ -131,25 +126,25 @@ void split_test(int batch_num, int feature_num, int x_size, int y_size, std::vec
         auto input_feature_offset = split_offsets[splitNum].feature[0];
         auto input_y_offset = split_offsets[splitNum].spatial[1];
         auto input_x_offset = split_offsets[splitNum].spatial[0];
-        
+
         // iterator to iterate through input buffer
         auto input_batch_itr = input_batch_offset;
         auto input_feature_itr = input_feature_offset;
         auto input_y_itr = input_y_offset;
         auto input_x_itr = input_x_offset;
-        
+
         for (auto b = 0; b < output_batch; ++b) {  // B
-            
+
                 // reset the input feature iterator
-            input_feature_itr = input_feature_offset; 
+            input_feature_itr = input_feature_offset;
             for (auto f = 0; f < output_feature; f++) {  // F
-                
+
                 // reset the input y iterator
-                input_y_itr = input_y_offset;  
+                input_y_itr = input_y_offset;
                 for (auto y = 0; y < output_y; y++) {  // Y
-                    
+
                     // reset the input x iterator
-                    input_x_itr = input_x_offset;  
+                    input_x_itr = input_x_offset;
                     for (auto x = 0; x < output_x; x++) {  // X
                         auto linear_id = input_x_itr + x_size * (input_y_itr + y_size * (input_feature_itr + feature_num * input_batch_itr)); // index in input
                         auto output_linear_id = x + output_x * (y + output_y * (f + output_feature * b)); // index in output
@@ -180,7 +175,7 @@ TEST(split_gpu_f32, split_1d_uneven_2_splits) {
     auto y_size = 3;
     std::vector<cldnn::tensor> split_offsets = {
                                                 {0, 0, 0, 0},
-                                                {0, 1, 0, 0}                                                
+                                                {0, 1, 0, 0}
                                                };
 
     split_test<float>(batch_num, feature_num, x_size, y_size, split_offsets);
@@ -201,7 +196,7 @@ TEST(split_gpu_i64, split_1d_uneven_2_splits) {
     auto y_size = 3;
     std::vector<cldnn::tensor> split_offsets = {
                                                 {0, 0, 0, 0},
-                                                {0, 1, 0, 0}                                                
+                                                {0, 1, 0, 0}
                                                };
 
     split_test<int64_t>(batch_num, feature_num, x_size, y_size, split_offsets);
@@ -209,13 +204,13 @@ TEST(split_gpu_i64, split_1d_uneven_2_splits) {
 
 TEST(split_gpu_f32, basic_split_concat_optimization) {
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32,format::bfyx,{ 1, 25, 1, 256 } });
+    auto input = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 25, 1, 256 } });
     tests::set_random_values<float>(input);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     std::vector<std::pair<primitive_id, tensor>> offsets;
     std::vector<primitive_id> ids;
     for (int i = 0; i < 25; i++)
@@ -238,8 +233,8 @@ TEST(split_gpu_f32, basic_split_concat_optimization) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto input_ptr = input.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> input_ptr(input, get_test_stream());
 
     for (int i = 0; i < 25*256; ++i)
     {
@@ -249,13 +244,13 @@ TEST(split_gpu_f32, basic_split_concat_optimization) {
 
 TEST(split_gpu_i64, basic_split_concat_optimization) {
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::i64,format::bfyx,{ 1, 25, 1, 256 } });
+    auto input = engine.allocate_memory({ data_types::i64,format::bfyx,{ 1, 25, 1, 256 } });
     tests::set_random_values<int64_t>(input);
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     std::vector<std::pair<primitive_id, tensor>> offsets;
     std::vector<primitive_id> ids;
     for (int i = 0; i < 25; i++)
@@ -278,8 +273,8 @@ TEST(split_gpu_i64, basic_split_concat_optimization) {
     auto outputs = network.execute();
 
     auto output = outputs.at("output").get_memory();
-    auto output_ptr = output.pointer<int64_t>();
-    auto input_ptr = input.pointer<int64_t>();
+    cldnn::mem_lock<int64_t> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<int64_t> input_ptr(input, get_test_stream());
 
     for (int i = 0; i < 25*256; ++i)
     {
@@ -523,17 +518,17 @@ TEST(split_gpu_f32, basic_in2x3x2x2_split_feature_bfyx) {
     //  id: "out1", offsets: { 0, 1, 0, 0 }
     //  id: "out2", offsets: { 0, 2, 0, 0 }
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 3;
     auto x_size = 4;
     auto y_size = 3;
 
-    auto input = memory::allocate(engine, { data_types::f32,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::f32,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(split("split", "input",
     {
         { "out0", { 0, 0, 0, 0 } },
@@ -556,8 +551,8 @@ TEST(split_gpu_f32, basic_in2x3x2x2_split_feature_bfyx) {
     {
         auto split_id = "split:out" + std::to_string(i);
         auto output = outputs.at(split_id).get_memory();
-        auto output_ptr = output.pointer<float>();
-        check_feature_map<float>(output_ptr, input_vec, batch_num, feature_num, y_size, x_size, i, 1);
+        cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+        check_feature_map<float>(output_ptr.data(), input_vec, batch_num, feature_num, y_size, x_size, i, 1);
     }
 }
 
@@ -569,17 +564,17 @@ TEST(split_gpu_i64, basic_in2x3x2x2_split_feature_bfyx) {
     //  id: "out1", offsets: { 0, 1, 0, 0 }
     //  id: "out2", offsets: { 0, 2, 0, 0 }
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 3;
     auto x_size = 4;
     auto y_size = 3;
 
-    auto input = memory::allocate(engine, { data_types::i64,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
+    auto input = engine.allocate_memory({ data_types::i64,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(split("split", "input",
     {
         { "out0", { 0, 0, 0, 0 } },
@@ -602,8 +597,8 @@ TEST(split_gpu_i64, basic_in2x3x2x2_split_feature_bfyx) {
     {
         auto split_id = "split:out" + std::to_string(i);
         auto output = outputs.at(split_id).get_memory();
-        auto output_ptr = output.pointer<int64_t>();
-        check_feature_map<int64_t>(output_ptr, input_vec, batch_num, feature_num, y_size, x_size, i, 1);
+        cldnn::mem_lock<int64_t> output_ptr(output, get_test_stream());
+        check_feature_map<int64_t>(output_ptr.data(), input_vec, batch_num, feature_num, y_size, x_size, i, 1);
     }
 }
 
@@ -616,23 +611,23 @@ TEST(split_gpu_f32, basic_in2x3x2x2_split_scale_feature_bfyx) {
     //  id: "out2", offsets: { 0, 2, 0, 0 }
     //  Additional scale layer at the end
 
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
     auto batch_num = 6;
     auto feature_num = 3;
     auto x_size = 4;
     auto y_size = 3;
 
-    auto input = memory::allocate(engine, { data_types::f32,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
-    auto scale_input0 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto scale_input1 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto scale_input2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32,format::bfyx,{ batch_num, feature_num, x_size, y_size } });
+    auto scale_input0 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto scale_input1 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto scale_input2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
-    topology.add(input_layout("scale_input0", scale_input0.get_layout()));
-    topology.add(input_layout("scale_input1", scale_input1.get_layout()));
-    topology.add(input_layout("scale_input2", scale_input2.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
+    topology.add(input_layout("scale_input0", scale_input0->get_layout()));
+    topology.add(input_layout("scale_input1", scale_input1->get_layout()));
+    topology.add(input_layout("scale_input2", scale_input2->get_layout()));
     topology.add(split("split", "input",
     {
         { "out0",{ 0, 0, 0, 0 } },
@@ -649,7 +644,7 @@ TEST(split_gpu_f32, basic_in2x3x2x2_split_scale_feature_bfyx) {
     set_values(scale_input1, scale_input_vec1);
     std::vector<float> scale_input_vec2 = { 3.f };
     set_values(scale_input2, scale_input_vec2);
-   
+
     std::vector<float> input_vec = generate_random_input<float>(batch_num, feature_num, y_size, x_size, -10, 10);
     set_values(input, input_vec);
 
@@ -668,7 +663,7 @@ TEST(split_gpu_f32, basic_in2x3x2x2_split_scale_feature_bfyx) {
     {
         auto split_id = "scale" + std::to_string(i);
         auto output = outputs.at(split_id).get_memory();
-        auto output_ptr = output.pointer<float>();
-        check_feature_map<float>(output_ptr, input_vec, batch_num, feature_num, y_size, x_size, i, i + 1);
+        cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+        check_feature_map<float>(output_ptr.data(), input_vec, batch_num, feature_num, y_size, x_size, i, i + 1);
     }
 }
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/streams_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/streams_test.cpp
index b679ac71a6be9e..ed1dfa1110a1b0 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/streams_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/streams_test.cpp
@@ -2,50 +2,19 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <gtest/gtest.h>
-#include <api/tensor.hpp>
-#include <api/engine.hpp>
-#include <api/memory.hpp>
-#include <api/input_layout.hpp>
-#include <api/data.hpp>
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/mutable_data.hpp>
 
 using namespace cldnn;
-using namespace tests;
-
-static engine_configuration streams_config(false,
-                                                 false,
-                                                 false,
-                                                 "",
-                                                 "",
-                                                 true,
-                                                 "",
-                                                 "",
-                                                 priority_mode_types::disabled,
-                                                 throttle_mode_types::disabled,
-                                                 true,
-                                                 2);
-
-TEST(gpu_streams, DISABLED_can_allocate_memory_for_stream)
-{
-    engine _engine(streams_config);
-
-    ASSERT_NO_THROW(memory::allocate(_engine, layout(data_types::f32, format::bfyx, {1, 2, 3, 4})));
-    ASSERT_NO_THROW(memory::allocate(_engine, layout(data_types::f32, format::bfyx, {1, 2, 3, 4}), 0));
-    ASSERT_NO_THROW(memory::allocate(_engine, layout(data_types::f32, format::bfyx, {1, 2, 3, 4}), 1));
-    ASSERT_ANY_THROW(memory::allocate(_engine, layout(data_types::f32, format::bfyx, {1, 2, 3, 4}), 2));
-
-    auto mem0 = memory::allocate(_engine, layout(data_types::f32, format::bfyx, {1, 2, 3, 4}), 0);
-    ASSERT_EQ(mem0.get_net_id(), 0);
-    auto mem1 = memory::allocate(_engine, layout(data_types::f32, format::bfyx, {1, 2, 3, 4}), 1);
-    ASSERT_EQ(mem1.get_net_id(), 1);
-}
+using namespace ::tests;
 
-TEST(gpu_streams, can_create_networks_for_stream)
-{
-    engine _engine(streams_config);
+TEST(gpu_streams, can_create_networks_for_stream) {
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(_engine, { data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 5, 4 } });
     set_values(input,
                { 1.0f, -2.0f, -3.0f, 4.0f, 5.0f,
                  2.0f, 2.0f, 3.0f, 4.0f, -6.0f,
@@ -58,9 +27,9 @@ TEST(gpu_streams, can_create_networks_for_stream)
             1.0f, 1.0f, 1.0f, -0.5f, 1.0f };
 
     topology topology(
-            input_layout("input", input.get_layout()),
+            input_layout("input", input->get_layout()),
             activation("relu", "input", activation_func::relu_negative_slope, activation_additional_params{ 0.5f, 0.f }, padding{ { 0, 0, 0, 0 }, 0 }));
-    network network(_engine, topology, build_options());
+    network network(engine, topology, build_options());
 
     network.set_input_data("input", input);
     auto outputs = network.execute();
@@ -68,10 +37,8 @@ TEST(gpu_streams, can_create_networks_for_stream)
     EXPECT_EQ(outputs.begin()->first, "relu");
 
     auto output_memory = outputs.at("relu").get_memory();
-    auto output_layout = output_memory.get_layout();
-    auto output_ptr = output_memory.pointer<float>();
-
-    EXPECT_EQ(output_memory.get_net_id(), network.get_id());
+    auto output_layout = output_memory->get_layout();
+    cldnn::mem_lock<float> output_ptr(output_memory, get_test_stream());
 
     int y_size = output_layout.size.spatial[1];
     int x_size = output_layout.size.spatial[0];
@@ -89,9 +56,9 @@ TEST(gpu_streams, can_create_networks_for_stream)
 }
 
 TEST(gpu_streams, check_networks_can_use_the_same_weights) {
-    engine _engine(streams_config);
+    auto& engine = get_test_engine();
 
-    auto weights = memory::allocate(_engine, { data_types::f32,format::bfyx,{ 1, 1, 3, 2 } });
+    auto weights = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 3, 2 } });
 
     VVF<float> output_vec = {
             { 20.0f, 27.0f, 38.0f },
@@ -105,12 +72,12 @@ TEST(gpu_streams, check_networks_can_use_the_same_weights) {
             convolution("conv", "input", { "weights" }, { 1,1,1,2 }));
 
     set_values(weights, { 1.0f, 2.0f, 1.0f, 2.0f, 1.0f, 2.0f });
-    program prog(_engine, topology, build_options());
+    program prog(engine, topology, build_options());
     network network0(prog, 0);
     network network1(prog, 1);
 
-    auto input0 = memory::allocate(_engine, input0_layout, network0.get_id());
-    auto input1 = memory::allocate(_engine, input0_layout, network1.get_id());
+    auto input0 = engine.allocate_memory(input0_layout);
+    auto input1 = engine.allocate_memory(input0_layout);
     set_values(input0, { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 2.0f, 2.0f, 3.0f, 4.0f, 6.0f, 3.0f, 3.0f, 3.0f, 5.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f });
     set_values(input1, { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 2.0f, 2.0f, 3.0f, 4.0f, 6.0f, 3.0f, 3.0f, 3.0f, 5.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f });
 
@@ -126,12 +93,10 @@ TEST(gpu_streams, check_networks_can_use_the_same_weights) {
 
     auto output_memory0 = outputs0.at("conv").get_memory();
     auto output_memory1 = outputs1.at("conv").get_memory();
-    auto output_layout = output_memory0.get_layout();
-    auto output_ptr0 = output_memory0.pointer<float>();
-    auto output_ptr1 = output_memory1.pointer<float>();
+    auto output_layout = output_memory0->get_layout();
+    cldnn::mem_lock<float> output_ptr0(output_memory0, get_test_stream());
+    cldnn::mem_lock<float> output_ptr1(output_memory1, get_test_stream());
 
-    EXPECT_EQ(output_memory0.get_net_id(), network0.get_id());
-    EXPECT_EQ(output_memory1.get_net_id(), network1.get_id());
     auto wmem0 = network0.get_output_memory("weights");
     auto wmem1 = network1.get_output_memory("weights");
 
@@ -153,3 +118,71 @@ TEST(gpu_streams, check_networks_can_use_the_same_weights) {
         }
     }
 }
+
+TEST(gpu_streams, check_networks_use_unique_mutable_data_per_stream) {
+    auto& engine = get_test_engine();
+
+    auto weights = engine.allocate_memory({ data_types::f32,format::bfyx,{ 1, 1, 3, 2 } });
+
+    VVF<float> output_vec = {
+            { 20.0f, 27.0f, 38.0f },
+            { 17.0f, 19.0f, 19.0f } };
+
+    layout input0_layout(data_types::f32, format::bfyx, { 1, 1, 5, 4 });
+
+    topology topology(
+            input_layout("input", input0_layout),
+            mutable_data("weights", weights),
+            convolution("conv", "input", { "weights" }, { 1,1,1,2 }));
+
+    set_values(weights, { 1.0f, 2.0f, 1.0f, 2.0f, 1.0f, 2.0f });
+    program prog(engine, topology, build_options());
+    network network0(prog, 0);
+    network network1(prog, 1);
+
+    auto input0 = engine.allocate_memory(input0_layout);
+    auto input1 = engine.allocate_memory(input0_layout);
+    set_values(input0, { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 2.0f, 2.0f, 3.0f, 4.0f, 6.0f, 3.0f, 3.0f, 3.0f, 5.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f });
+    set_values(input1, { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 2.0f, 2.0f, 3.0f, 4.0f, 6.0f, 3.0f, 3.0f, 3.0f, 5.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f });
+
+    network0.set_input_data("input", input0);
+    network1.set_input_data("input", input1);
+
+    auto outputs0 = network0.execute();
+    auto outputs1 = network1.execute();
+    EXPECT_EQ(outputs0.size(), size_t(1));
+    EXPECT_EQ(outputs1.size(), size_t(1));
+    EXPECT_EQ(outputs0.begin()->first, "conv");
+    EXPECT_EQ(outputs1.begin()->first, "conv");
+
+    auto output_memory0 = outputs0.at("conv").get_memory();
+    auto output_memory1 = outputs1.at("conv").get_memory();
+    auto output_layout = output_memory0->get_layout();
+    cldnn::mem_lock<float> output_ptr0(output_memory0, get_test_stream());
+    cldnn::mem_lock<float> output_ptr1(output_memory1, get_test_stream());
+
+    auto wmem0 = network0.get_output_memory("weights");
+    auto wmem1 = network1.get_output_memory("weights");
+
+    // check that each stream has unique weights data
+    ASSERT_NE(wmem0, wmem1);
+
+    // check that initial memory is reused by the primary stream
+    ASSERT_EQ(wmem0, weights);
+
+    int y_size = output_layout.size.spatial[1];
+    int x_size = output_layout.size.spatial[0];
+    int f_size = output_layout.size.feature[0];
+    int b_size = output_layout.size.batch[0];
+    EXPECT_EQ(output_layout.format, format::bfyx);
+    EXPECT_EQ(y_size, 2);
+    EXPECT_EQ(x_size, 3);
+    EXPECT_EQ(f_size, 1);
+    EXPECT_EQ(b_size, 1);
+    for (int y = 0; y < y_size; ++y) {
+        for (int x = 0; x < x_size; ++x) {
+            EXPECT_EQ(output_vec[y][x], output_ptr0[y * x_size + x]);
+            EXPECT_EQ(output_vec[y][x], output_ptr1[y * x_size + x]);
+        }
+    }
+}
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/strided_slice_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/strided_slice_gpu_test.cpp
index 2634f03962a2c3..aa4ec3aa0d3c32 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/strided_slice_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/strided_slice_gpu_test.cpp
@@ -3,17 +3,15 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include <api/input_layout.hpp>
-#include "api/strided_slice.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
-#include <api/data.hpp>
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/strided_slice.hpp>
+#include <cldnn/primitives/data.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_full) {
     // Input (BFYX): 2x2x2x2
@@ -22,11 +20,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_full) {
     // Stride (BFYX): 1x1x1x1
     // Output (BFYX): 2x2x2x2
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f,
@@ -43,7 +41,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_full) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -63,7 +61,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_full) {
     std::vector<float> answers = {
             0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 13.f, 14.f, 15.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -79,11 +77,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_full) {
     // Stride (BFYX): 1x1x1x1
     // Output (BFYX): 2x2x2x2
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f,
@@ -100,7 +98,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_full) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -120,7 +118,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_full) {
     std::vector<float> answers = {
             0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f, 9.f, 10.f, 11.f, 12.f, 13.f, 14.f, 15.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -136,11 +134,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_ignore) {
     // Stride (BFYX): 1x1x1x1
     // Output (BFYX): 2x2x2x2
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f,
@@ -157,7 +155,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_ignore) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -179,7 +177,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_ignore) {
         9.0f, 10.0f, 11.0f, 12.0f, 13.0f, 14.0f, 15.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -195,11 +193,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_ignore) {
     // Stride (BFYX): 1x1x1x1
     // Output (BFYX): 2x2x2x2
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f,
@@ -216,7 +214,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_ignore) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -238,7 +236,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_ignore) {
         9.0f, 10.0f, 11.0f, 12.0f, 13.0f, 14.0f, 15.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -254,11 +252,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_single) {
     // Stride (BFYX): 1x1x1x1
     // Output (BFYX): 1x1x1x1
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f,
@@ -275,7 +273,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_single) {
                });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -294,7 +292,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_single) {
 
     std::vector<float> answers = { 15.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -310,11 +308,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_single) {
     // Stride (BFYX): 1x1x1x1
     // Output (BFYX): 1x1x1x1
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f,
@@ -331,7 +329,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_single) {
                });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -350,7 +348,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_single) {
 
     std::vector<float> answers = { 15.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -366,11 +364,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x4x3_stride) {
     // Stride (BFYX): 1x1x2x1
     // Output (BFYX): 2x2x2x3
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 3, 4 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 3, 4 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f,
@@ -391,7 +389,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x4x3_stride) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -413,7 +411,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x4x3_stride) {
             24.f, 25.f, 26.f, 30.f, 31.f, 32.f, 36.f, 37.f, 38.f, 42.f, 43.f, 44.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -429,11 +427,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x4x3_stride) {
     // Stride (BFYX): 1x1x2x1
     // Output (BFYX): 2x2x2x3
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 3, 4 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 3, 4 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.f, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f,
@@ -454,7 +452,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x4x3_stride) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -476,7 +474,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x4x3_stride) {
             24.f, 25.f, 26.f, 30.f, 31.f, 32.f, 36.f, 37.f, 38.f, 42.f, 43.f, 44.f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -492,11 +490,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x4x4_part_stride) {
     // Stride (BFYX): 1x1x1x2
     // Output (BFYX): 1x2x4x2
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 4, 4 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 4, 4 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f,
@@ -530,7 +528,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x4x4_part_stride) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -559,7 +557,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x4x4_part_stride) {
             61.0f, 63.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -575,11 +573,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x4x4_part_stride) {
     // Stride (BFYX): 1x1x1x2
     // Output (BFYX): 1x2x4x2
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 4, 4 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 4, 4 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f,
@@ -613,7 +611,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x4x4_part_stride) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -642,7 +640,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x4x4_part_stride) {
             61.0f, 63.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -656,11 +654,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x4x1_new_axis_mask) {
     // New_axis_mask: 1
     // Output (BFYX): 1x2x2x4
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 4 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 4 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f, 9.0f,
@@ -677,7 +675,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x4x1_new_axis_mask) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -699,7 +697,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x4x1_new_axis_mask) {
             10.0f, 11.0f, 12.0f, 13.0f, 14.0f, 15.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -712,11 +710,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x4x1_new_axis_mask) {
     // New_axis_mask: 1
     // Output (BFYX): 1x2x2x4
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 4 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 4 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f, 9.0f,
@@ -733,7 +731,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x4x1_new_axis_mask) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -755,7 +753,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x4x1_new_axis_mask) {
             10.0f, 11.0f, 12.0f, 13.0f, 14.0f, 15.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -768,11 +766,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x1x1_new_axis_mask_2) {
     // New_axis_mask: 101
     // Output (BFYX): 1x2x1x2
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f
@@ -788,7 +786,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x1x1_new_axis_mask_2) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -809,7 +807,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x1x1_new_axis_mask_2) {
             0.0f, 1.0f, 2.0f, 3.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -822,11 +820,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x1x1_new_axis_mask_2) {
     // New_axis_mask: 101
     // Output (BFYX): 1x2x1x2
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f
@@ -842,7 +840,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x1x1_new_axis_mask_2) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -863,7 +861,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x1x1_new_axis_mask_2) {
             0.0f, 1.0f, 2.0f, 3.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -875,11 +873,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x1x1) {
     // Input (BFYX): 2x2x1x1
     // Output (BFYX): 2x2x1x1
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 2, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 2, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f
@@ -895,7 +893,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x1x1) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -916,7 +914,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x1x1) {
             0.0f, 1.0f, 2.0f, 3.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -928,11 +926,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x1x1) {
     // Input (BFYX): 2x2x1x1
     // Output (BFYX): 2x2x1x1
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 2, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 2, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 2, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 1, 1 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 2, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 2, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 2, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f
@@ -948,7 +946,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x1x1) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -969,7 +967,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x1x1) {
             0.0f, 1.0f, 2.0f, 3.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -981,11 +979,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x1x1) {
     // Input (BFZYX): 2x2x2x1x1
     // Output (BFZYX): 1x2x2x1x1
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f
@@ -1001,7 +999,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x1x1) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -1022,7 +1020,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x1x1) {
             0.0f, 1.0f, 2.0f, 3.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -1034,11 +1032,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x1x1) {
     // Input (BFZYX): 2x2x2x1x1
     // Output (BFZYX): 1x2x2x1x1
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f
@@ -1054,7 +1052,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x1x1) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -1075,7 +1073,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x1x1) {
             0.0f, 1.0f, 2.0f, 3.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -1087,11 +1085,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x1x1_2) {
     // Input (BFZYX): 2x2x2x1x1
     // Output (BFZYX): 2x1x1x1x1
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f
@@ -1107,7 +1105,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x1x1_2) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -1128,7 +1126,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x1x1_2) {
             0.0f, 4.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -1140,11 +1138,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x1x1_2) {
     // Input (BFZYX): 2x2x2x1x1
     // Output (BFZYX): 2x1x1x1x1
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f
@@ -1160,7 +1158,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x1x1_2) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -1181,7 +1179,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x1x1_2) {
             0.0f, 4.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -1196,11 +1194,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_full_negative_stride) {
     // Stride (BFYX): -1x1x1x1
     // Output (BFYX): 2x2x2x2
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f,
@@ -1217,7 +1215,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_full_negative_stride) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -1237,7 +1235,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x2_full_negative_stride) {
     std::vector<float> answers = {
             12.f, 13.f, 14.f, 15.f, 8.f, 9.f, 10.f, 11.f, 4.f, 5.f, 6.f, 7.f, 0.f, 1.f, 2.f, 3.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -1253,11 +1251,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_full_negative_stride) {
     // Stride (BFYX): -1x1x1x1
     // Output (BFYX): 2x2x2x2
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 4, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f,
@@ -1274,7 +1272,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_full_negative_stride) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -1294,7 +1292,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x2_full_negative_stride) {
     std::vector<float> answers = {
             12.f, 13.f, 14.f, 15.f, 8.f, 9.f, 10.f, 11.f, 4.f, 5.f, 6.f, 7.f, 0.f, 1.f, 2.f, 3.f };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     ASSERT_EQ(output_ptr.size(), answers.size());
     for (size_t i = 0; i < answers.size(); ++i)
@@ -1307,11 +1305,11 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x1x1_2_negative_all) {
     // Input (BFZYX): 2x2x2x1x1
     // Output (BFZYX): 2x1x1x1x1
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i32, format::bfyx, { 3, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f
@@ -1327,7 +1325,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x1x1_2_negative_all) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -1348,7 +1346,7 @@ TEST(strided_slice_gpu_f32_i32, test_2x2x2x1x1_2_negative_all) {
             0.0f, 4.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
@@ -1360,11 +1358,11 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x1x1_2_negative_all) {
     // Input (BFZYX): 2x2x2x1x1
     // Output (BFZYX): 2x1x1x1x1
 
-    const auto& engine = get_test_engine();
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
-    auto begin = memory::allocate(engine, { data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
-    auto end = memory::allocate(engine, { data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
-    auto strides = memory::allocate(engine, { data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
+    auto& engine = get_test_engine();
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx, { 2, 2, 1, 1, 2 } });
+    auto begin = engine.allocate_memory({ data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
+    auto end = engine.allocate_memory({ data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
+    auto strides = engine.allocate_memory({ data_types::i64, format::bfyx, { 3, 1, 1, 1 } });
 
     set_values(input, {
             0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f
@@ -1380,7 +1378,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x1x1_2_negative_all) {
     });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("input2", begin));
     topology.add(data("input3", end));
     topology.add(data("input4", strides));
@@ -1401,7 +1399,7 @@ TEST(strided_slice_gpu_f32_i64, test_2x2x2x1x1_2_negative_all) {
             0.0f, 4.0f
     };
 
-    auto output_ptr = output.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
 
     for (size_t i = 0; i < answers.size(); ++i)
     {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/tensor_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/tensor_test.cpp
index 22abf3fbc818eb..c7563f4e861c5e 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/tensor_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/tensor_test.cpp
@@ -2,8 +2,7 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <gtest/gtest.h>
-#include <api/tensor.hpp>
+#include "test_utils.h"
 
 TEST(tensor_api, order_new_notation)
 {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/tile_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/tile_gpu_test.cpp
index b734a9e7b4f666..b80f4eedd6d581 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/tile_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/tile_gpu_test.cpp
@@ -3,22 +3,19 @@
 //
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/tile.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
+
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/tile.hpp>
 
 #include <iostream>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 template<typename data_t>
-void tile_ref(const memory& input, memory& output, tile::tile_axis axis, int num_tiles)
+void tile_ref(const memory::ptr input, memory::ptr output, tile::tile_axis axis, int num_tiles)
 {
     auto get_sizes = [](const tensor& size, tile::tile_axis axis) -> std::pair<int, int>
     {
@@ -33,13 +30,13 @@ void tile_ref(const memory& input, memory& output, tile::tile_axis axis, int num
         }
     };
 
-    const pointer<data_t> src = input.pointer<data_t>();
-    pointer<data_t> dst = output.pointer<data_t>();
+    cldnn::mem_lock<data_t> src(input, get_test_stream());
+    cldnn::mem_lock<data_t> dst(output, get_test_stream());
 
     const data_t* psrc = src.data();
     data_t* pdst = dst.data();
 
-    auto sizes = get_sizes(input.get_layout().size, axis);
+    auto sizes = get_sizes(input->get_layout().size, axis);
     int outer_dim = sizes.first;
     int inner_dim = sizes.second;
 
@@ -58,13 +55,13 @@ void tile_ref(const memory& input, memory& output, tile::tile_axis axis, int num
 }
 
 TEST(tile_gpu, basic_in1x2x2x2_axis_b) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
-    auto output_ref = memory::allocate(engine, { data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
+    auto output_ref = engine.allocate_memory({ data_types::f32, format::bfyx, { 2, 2, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(tile("tile", "input", tensor(2, 2, 2, 2)));
 
     std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f,
@@ -78,22 +75,22 @@ TEST(tile_gpu, basic_in1x2x2x2_axis_b) {
     auto outputs = network.execute();
 
     auto output = outputs.at("tile").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_ref_ptr = output_ref.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> output_ref_ptr(output_ref, get_test_stream());
 
-    for (unsigned int i = 0; i < output_ref.count(); ++i) {
+    for (unsigned int i = 0; i < output_ref->count(); ++i) {
         EXPECT_EQ(output_ptr[i], output_ref_ptr[i]) << "Index=" << i;
     }
 }
 
 TEST(tile_gpu, basic_in1x2x2x2_axis_f) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
-    auto output_ref = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 4, 2, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
+    auto output_ref = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 4, 2, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(tile("tile", "input", tensor(1, 4, 2, 2)));
 
     std::vector<float> input_vec = { 1.f, 0.f,
@@ -110,22 +107,22 @@ TEST(tile_gpu, basic_in1x2x2x2_axis_f) {
     auto outputs = network.execute();
 
     auto output = outputs.at("tile").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_ref_ptr = output_ref.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> output_ref_ptr(output_ref, get_test_stream());
 
-    for (unsigned int i = 0; i < output_ref.count(); ++i) {
+    for (unsigned int i = 0; i < output_ref->count(); ++i) {
         EXPECT_EQ(output_ptr[i], output_ref_ptr[i]) << "Index=" << i;
     }
 }
 
 TEST(tile_gpu, basic_in1x2x2x2_axis_y) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
-    auto output_ref = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
+    auto output_ref = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 4 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(tile("tile", "input", tensor(1, 2, 2, 4)));
 
     std::vector<float> input_vec = { 1.f, 0.f,
@@ -142,22 +139,22 @@ TEST(tile_gpu, basic_in1x2x2x2_axis_y) {
     auto outputs = network.execute();
 
     auto output = outputs.at("tile").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_ref_ptr = output_ref.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> output_ref_ptr(output_ref, get_test_stream());
 
-    for (unsigned int i = 0; i < output_ref.count(); ++i) {
+    for (unsigned int i = 0; i < output_ref->count(); ++i) {
         EXPECT_EQ(output_ptr[i], output_ref_ptr[i]) << "Index=" << i;
     }
 }
 
 TEST(tile_gpu, basic_in1x2x2x2_axis_x) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
-    auto output_ref = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 4, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 2, 2 } });
+    auto output_ref = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 4, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(tile("tile", "input", tensor(1, 2, 4, 2)));
 
     std::vector<float> input_vec = { 1.f, 0.f,
@@ -174,22 +171,22 @@ TEST(tile_gpu, basic_in1x2x2x2_axis_x) {
     auto outputs = network.execute();
 
     auto output = outputs.at("tile").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_ref_ptr = output_ref.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> output_ref_ptr(output_ref, get_test_stream());
 
-    for (unsigned int i = 0; i < output_ref.count(); ++i) {
+    for (unsigned int i = 0; i < output_ref->count(); ++i) {
         EXPECT_EQ(output_ptr[i], output_ref_ptr[i]) << "Index=" << i;
     }
 }
 
 TEST(tile_gpu, basic_in1x2x2x2_axis_x_dense) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 1, 2 } });
-    auto output_ref = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 2, 4, 2 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 1, 2 } });
+    auto output_ref = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 2, 4, 2 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(tile("tile", "input", tensor(1, 2, 4, 2)));
 
     std::vector<float> input_vec = { 1.f, 0.f, 5.f, 1.5f};
@@ -202,22 +199,22 @@ TEST(tile_gpu, basic_in1x2x2x2_axis_x_dense) {
     auto outputs = network.execute();
 
     auto output = outputs.at("tile").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_ref_ptr = output_ref.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> output_ref_ptr(output_ref, get_test_stream());
 
-    for (unsigned int i = 0; i < output_ref.count(); ++i) {
+    for (unsigned int i = 0; i < output_ref->count(); ++i) {
         EXPECT_EQ(output_ptr[i], output_ref_ptr[i]) << "Index=" << i;
     }
 }
 
 TEST(tile_gpu, basic_in1x2x2x2_axis_z) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
 
-    auto input = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 2, 2, 2, 2 } });
-    auto output_ref = memory::allocate(engine, { data_types::f32, format::bfzyx,{ 1, 2, 2, 2, 4 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 2, 2, 2, 2 } });
+    auto output_ref = engine.allocate_memory({ data_types::f32, format::bfzyx,{ 1, 2, 2, 2, 4 } });
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(tile("tile", "input", tensor(1, 2, 2, 2, 4)));
 
     std::vector<float> input_vec = {
@@ -239,10 +236,10 @@ TEST(tile_gpu, basic_in1x2x2x2_axis_z) {
     auto outputs = network.execute();
 
     auto output = outputs.at("tile").get_memory();
-    auto output_ptr = output.pointer<float>();
-    auto output_ref_ptr = output_ref.pointer<float>();
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    cldnn::mem_lock<float> output_ref_ptr(output_ref, get_test_stream());
 
-    for (unsigned int i = 0; i < output_ref.count(); ++i) {
+    for (unsigned int i = 0; i < output_ref->count(); ++i) {
         EXPECT_EQ(output_ptr[i], output_ref_ptr[i]) << "Index=" << i;
     }
 }
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/topology_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/topology_test.cpp
index 11e34e2f60158f..42eae68031cb9e 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/topology_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/topology_test.cpp
@@ -2,25 +2,23 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <gtest/gtest.h>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/lrn.hpp>
+#include <cldnn/primitives/convolution.hpp>
+#include <cldnn/primitives/fully_connected.hpp>
+#include <cldnn/primitives/pooling.hpp>
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/scale.hpp>
+#include <cldnn/primitives/eltwise.hpp>
+#include <cldnn/primitives/softmax.hpp>
+#include <cldnn/primitives/activation.hpp>
+#include <cldnn/primitives/concatenation.hpp>
+
 #include <include/topology_impl.h>
+
 #include <iostream>
-#include "api/memory.hpp"
-#include <api/lrn.hpp>
-#include <api/convolution.hpp>
-#include <api/fully_connected.hpp>
-#include <api/pooling.hpp>
-#include <api/data.hpp>
-#include <api/reorder.hpp>
-#include <api/scale.hpp>
-#include <api/eltwise.hpp>
-#include <api/softmax.hpp>
-#include <api/activation.hpp>
-#include <api/concatenation.hpp>
 #include <deque>
 #include <set>
 
@@ -55,6 +53,7 @@ class topology_test : public ::testing::TestWithParam<topology_params>
         public:
             // return false for invalid output_layout
             virtual bool AddPrimitive(cldnn::topology& topology, cldnn::primitive_id id, cldnn::layout output_layout, std::deque<named_layout>& input_layouts) = 0;
+            virtual ~topology_layer_type() = default;
         };
         static std::vector<std::shared_ptr<topology_layer_type>> layer_types;
         static cldnn::topology* CreateTopology(cldnn::layout output_layout, const std::vector<unsigned> generator_vec)
@@ -112,7 +111,7 @@ class topology_test : public ::testing::TestWithParam<topology_params>
         static void AddRandomMemory(cldnn::topology& topology, cldnn::primitive_id id, cldnn::layout layout)
         {
             //todo: allocate mem, randomize values by type, add to topology
-            auto mem_primitive = cldnn::memory::allocate(topology_test::engine, layout);
+            auto mem_primitive = topology_test::engine.allocate_memory(layout);
             switch (layout.data_type)
             {
             case cldnn::data_types::f32:
@@ -368,8 +367,8 @@ class topology_test : public ::testing::TestWithParam<topology_params>
         EXPECT_NE(topology, nullptr);
         cldnn::build_options options;
         options.set_option(cldnn::build_option::optimize_data(true));
-        cldnn::engine temp_engine;// using temp_engine since reusing the same one does not free all resources (network build becomes slower and slower)
-        cldnn::network network(temp_engine, *topology, options);
+        auto& engine = tests::get_test_engine();
+        cldnn::network network(engine, *topology, options);
         auto outputs = network.execute();
         EXPECT_NE(outputs.find(topology_generator::output_layer_id), outputs.end());
 
@@ -443,11 +442,11 @@ class topology_test : public ::testing::TestWithParam<topology_params>
     cldnn::layout output_layout;
     std::vector<unsigned> generator;
 
-    static const cldnn::engine& engine;
+    static cldnn::engine& engine;
     static std::vector<cldnn::layout> all_output_layouts;//just for tear-down
 };
 
-const cldnn::engine& topology_test::engine = tests::get_test_engine();
+cldnn::engine& topology_test::engine = tests::get_test_engine();
 std::vector<cldnn::layout> topology_test::all_output_layouts = {};
 
 std::vector<std::shared_ptr<topology_test::topology_generator::topology_layer_type>> topology_test::topology_generator::layer_types = {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/trim_to_outputs_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/trim_to_outputs_gpu_test.cpp
index bb613fccd13955..94689c2ce4b752 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/trim_to_outputs_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/trim_to_outputs_gpu_test.cpp
@@ -4,18 +4,14 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include <gtest/gtest.h>
-#include "api/memory.hpp"
-#include <api/input_layout.hpp>
-#include "api/concatenation.hpp"
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
-#include <api/data.hpp>
-#include "test_utils/test_utils.h"
+#include "test_utils.h"
+
+#include <cldnn/primitives/input_layout.hpp>
+#include "cldnn/primitives/concatenation.hpp"
+#include <cldnn/primitives/data.hpp>
 
 using namespace cldnn;
-using namespace tests;
+using namespace ::tests;
 
 /*
     This set of tests has been designed to check the correctness of trim_to_outputs optimization pass
@@ -29,14 +25,14 @@ using namespace tests;
                             ---> conv2 (to be eliminated)
 */
 TEST(trim_to_outputs, one_node_to_eliminate_case1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
     build_opt.set_option(cldnn::build_option::outputs({ "conv1" }));
     build_opt.set_option(build_option::optimize_data(false));             // to avoid adding reorders
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb, { 1, 1, 1, 1 } });
-    auto weights = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
-    auto bias = memory::allocate(engine, { data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb, { 1, 1, 1, 1 } });
+    auto weights = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto bias = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
 
     set_values(input, { 1.1f });
     set_values(weights, { 2.1f });
@@ -45,7 +41,7 @@ TEST(trim_to_outputs, one_node_to_eliminate_case1) {
     std::vector<float> out_data = { 3.91f };
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights", weights));
     topology.add(data("bias", bias));
     topology.add(cldnn::convolution("conv1", { "input" }, { "weights" }, { "bias" }));
@@ -61,7 +57,7 @@ TEST(trim_to_outputs, one_node_to_eliminate_case1) {
 
     for (auto& it : outputs)
     {
-        auto output_ptr = it.second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(it.second.get_memory(), get_test_stream());
         for (size_t cntr = 0; cntr < out_data.size(); cntr++)
         {
             EXPECT_NEAR(output_ptr[cntr], out_data[cntr], 1e-4);
@@ -78,16 +74,16 @@ Network structure:  input  -> conv1 (output)
                          ---> conv2 (to be eliminated along with its weights and bias)
 */
 TEST(trim_to_outputs, one_node_to_eliminate_case2) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
     build_opt.set_option(cldnn::build_option::outputs({ "conv1" }));
     build_opt.set_option(build_option::optimize_data(false));             // to avoid adding reorders
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 1, 1 } });
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto weights2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto bias1 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto bias2 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 1, 1 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto weights2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto bias1 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto bias2 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { 1.1f });
     set_values(weights1, { 2.1f });
@@ -98,7 +94,7 @@ TEST(trim_to_outputs, one_node_to_eliminate_case2) {
     std::vector<float> out_data = { 3.91f };
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights1", weights1));
     topology.add(data("bias1", bias1));
     topology.add(cldnn::convolution("conv1", { "input" }, { "weights1" }, { "bias1" }));
@@ -116,7 +112,7 @@ TEST(trim_to_outputs, one_node_to_eliminate_case2) {
 
     for (auto& it : outputs)
     {
-        auto output_ptr = it.second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(it.second.get_memory(), get_test_stream());
 
         for (size_t cntr = 0; cntr < out_data.size(); cntr++)
         {
@@ -135,16 +131,16 @@ Network structure:  input ---> conv1 --- ---> conv4 (output)
 Convolutions conv2, conv3 should be optimized out along with weights23 shered by conv2 and conv3.
 */
 TEST(trim_to_outputs, two_nodes_to_eliminate_case1) {
-    const auto& engine = get_test_engine();
+    auto& engine = get_test_engine();
     build_options build_opt;
     build_opt.set_option(cldnn::build_option::outputs({ "conv4" }));
     build_opt.set_option(build_option::optimize_data(false));             // to avoid adding reorders
 
-    auto input = memory::allocate(engine, { data_types::f32, format::yxfb,{ 1, 1, 1, 1 } });
-    auto weights1 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto weights23 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto weights4 = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
-    auto bias = memory::allocate(engine, { data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto input = engine.allocate_memory({ data_types::f32, format::yxfb,{ 1, 1, 1, 1 } });
+    auto weights1 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto weights23 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto weights4 = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
+    auto bias = engine.allocate_memory({ data_types::f32, format::bfyx,{ 1, 1, 1, 1 } });
 
     set_values(input, { 1.1f });
     set_values(weights1, { 2.1f });
@@ -155,7 +151,7 @@ TEST(trim_to_outputs, two_nodes_to_eliminate_case1) {
     std::vector<float> out_data = { 9.42f };
 
     topology topology;
-    topology.add(input_layout("input", input.get_layout()));
+    topology.add(input_layout("input", input->get_layout()));
     topology.add(data("weights1", weights1));
     topology.add(data("bias", bias));
     topology.add(cldnn::convolution("conv1", { "input" }, { "weights1" }, { "bias" }));
@@ -175,7 +171,7 @@ TEST(trim_to_outputs, two_nodes_to_eliminate_case1) {
 
     for (auto& it : outputs)
     {
-        auto output_ptr = it.second.get_memory().pointer<float>();
+        cldnn::mem_lock<float> output_ptr(it.second.get_memory(), get_test_stream());
 
         for (size_t cntr = 0; cntr < out_data.size(); cntr++)
         {
@@ -184,4 +180,3 @@ TEST(trim_to_outputs, two_nodes_to_eliminate_case1) {
         EXPECT_EQ(it.first, "conv4");
     }
 }
-
diff --git a/inference-engine/thirdparty/clDNN/tests/test_utils/instrumentation.cpp b/inference-engine/thirdparty/clDNN/tests/test_utils/instrumentation.cpp
deleted file mode 100644
index d8b2c73cc1145a..00000000000000
--- a/inference-engine/thirdparty/clDNN/tests/test_utils/instrumentation.cpp
+++ /dev/null
@@ -1,386 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "instrumentation.h"
-
-#include <fstream>
-#include <iomanip>
-#include <string>
-#include <vector>
-#include <iostream>
-
-namespace instrumentation {
-    // initalize dumping directory for whole run
-    const std::string logger::dump_dir = DUMP_DIRECTORY;
-
-    static float convert_half_to_float(cldnn::half_t val)
-    {
-        return static_cast<float>(val);
-    }
-
-    float convert_element(float f)
-    {
-        return f;
-    }
-
-    float convert_element(cldnn::half_t h)
-    {
-        return convert_half_to_float(h);
-    }
-
-    template<typename elemType>
-    void dump_byxf(const cldnn::memory& mem, bool single_batch, cldnn::tensor::value_type batch_id, bool single_feature, cldnn::tensor::value_type feature_id, std::vector<std::vector<std::stringstream>> & streams)
-    {
-        auto mem_arg = mem.get_layout();
-        auto mem_ptr = mem.pointer<elemType>();
-
-        unsigned int input_it = 0;
-        for (cldnn::tensor::value_type b = 0; b < mem_arg.size.batch[0]; b++)
-        {
-            for (cldnn::tensor::value_type y = 0; y < mem_arg.size.spatial[1]; y++)
-            {
-                for (cldnn::tensor::value_type x = 0; x < mem_arg.size.spatial[0]; x++)
-                {
-                    for (cldnn::tensor::value_type f = 0; f < mem_arg.size.feature[0]; f++)
-                    {
-                        if ((!single_batch || b == batch_id) && (!single_feature || f == feature_id))
-                        {
-                            streams[b][f] << convert_element(mem_ptr[input_it]) << " ";
-                            if (x == mem_arg.size.spatial[0] - 1)
-                            {
-                                streams[b][f] << std::endl;
-                            }
-                        }
-                        input_it++;
-                    }
-                }
-            }
-        }
-    }
-
-    template<typename elemType>
-    void dump_bfyx(const cldnn::memory& mem, bool single_batch, cldnn::tensor::value_type batch_id, bool single_feature, cldnn::tensor::value_type feature_id, std::vector<std::vector<std::stringstream>> & streams)
-    {
-        auto mem_arg = mem.get_layout();
-        auto mem_ptr = mem.pointer<elemType>();
-
-        unsigned int input_it = 0;
-        for (cldnn::tensor::value_type b = 0; b < mem_arg.size.batch[0]; b++)
-        {
-            for (cldnn::tensor::value_type f = 0; f < mem_arg.size.feature[0]; f++)
-            {
-                for (cldnn::tensor::value_type y = 0; y < mem_arg.size.spatial[1]; y++)
-                {
-                    for (cldnn::tensor::value_type x = 0; x < mem_arg.size.spatial[0]; x++)
-                    {
-                        if ((!single_batch || b == batch_id) && (!single_feature || f == feature_id))
-                        {
-                            streams[b][f] << convert_element(mem_ptr[input_it]) << " ";
-                            if (x == mem_arg.size.spatial[0] - 1)
-                            {
-                                streams[b][f] << std::endl;
-                            }
-                        }
-                        input_it++;
-                    }
-                }
-            }
-        }
-    }
-
-    template<typename elemType>
-    void dump_yxfb(const cldnn::memory& mem, bool single_batch, cldnn::tensor::value_type batch_id, bool single_feature, cldnn::tensor::value_type feature_id, std::vector<std::vector<std::stringstream>> & streams)
-    {
-        auto mem_arg = mem.get_layout();
-        auto mem_ptr = mem.pointer<elemType>();
-
-        unsigned int input_it = 0;
-        for (cldnn::tensor::value_type y = 0; y < mem_arg.size.spatial[1]; y++)
-        {
-            for (cldnn::tensor::value_type x = 0; x < mem_arg.size.spatial[0]; x++)
-            {
-                for (cldnn::tensor::value_type f = 0; f < mem_arg.size.feature[0]; f++)
-                {
-                    for (cldnn::tensor::value_type b = 0; b < mem_arg.size.batch[0]; b++)
-                    {
-                        if ((!single_batch || b == batch_id) && (!single_feature || f == feature_id))
-                        {
-                            streams[b][f] << convert_element(mem_ptr[input_it]) << " ";
-                            if (x == mem_arg.size.spatial[0] - 1)
-                            {
-                                streams[b][f] << std::endl;
-                            }
-                        }
-                        input_it++;
-                    }
-                }
-            }
-        }
-    }
-
-    template<typename elemType>
-    void dump_xb(const cldnn::memory& mem, bool single_batch, cldnn::tensor::value_type batch_id, std::vector<std::vector<std::stringstream>> & streams)
-    {
-        auto mem_arg = mem.get_layout();
-        auto mem_ptr = mem.pointer<elemType>();
-
-        unsigned int input_it = 0;
-        for (cldnn::tensor::value_type x = 0; x < mem_arg.size.spatial[0]; x++)
-        {
-            for (cldnn::tensor::value_type b = 0; b < mem_arg.size.batch[0]; b++)
-            {
-                if (!single_batch || b == batch_id)
-                {
-                    streams[b][0] << convert_element(mem_ptr[input_it]) << std::endl;
-                }
-                input_it++;
-            }
-        }
-    }
-
-    template<typename elemType>
-    void dump_bx(const cldnn::memory& mem, bool single_batch, cldnn::tensor::value_type batch_id, std::vector<std::vector<std::stringstream>> & streams)
-    {
-        auto mem_arg = mem.get_layout();
-        auto mem_ptr = mem.pointer<elemType>();
-
-        unsigned int input_it = 0;
-        for (cldnn::tensor::value_type b = 0; b < mem_arg.size.batch[0]; b++)
-        {
-            for (cldnn::tensor::value_type x = 0; x < mem_arg.size.spatial[0]; x++)
-            {
-                if (!single_batch || b == batch_id)
-                {
-                    streams[b][0] << convert_element(mem_ptr[input_it]) << std::endl;
-                }
-                input_it++;
-            }
-        }
-    }
-
-    template<typename elemType>
-    void dump_yxio(const cldnn::memory& mem, std::stringstream & stream)
-    {
-        auto mem_arg = mem.get_layout();
-        auto mem_ptr = mem.pointer<elemType>();
-
-        auto i_size = mem_arg.size.batch[0];
-        auto o_size = mem_arg.size.feature[0];
-        auto x_size = mem_arg.size.spatial[0];
-        auto y_size = mem_arg.size.spatial[1];
-        unsigned int input_it = 0;
-        for (cldnn::tensor::value_type o = 0; o < o_size; o++)
-        {
-            for (cldnn::tensor::value_type i = 0; i < i_size; i++)
-            {
-                for (cldnn::tensor::value_type x = 0; x < x_size; x++)
-                {
-                    for (cldnn::tensor::value_type y = 0; y < y_size; y++)
-                    {
-                        stream<< convert_element(mem_ptr[input_it]) << " ";
-                        input_it++;
-                    }
-                    stream<< std::endl;
-                }
-            }
-        }
-    }
-
-    template<typename elemType>
-    void dump_oiyx(const cldnn::memory& mem, std::stringstream & stream)
-    {
-        auto mem_arg = mem.get_layout();
-        auto mem_ptr = mem.pointer<elemType>();
-
-        auto i_size = mem_arg.size.batch[0];
-        auto o_size = mem_arg.size.feature[0];
-        auto x_size = mem_arg.size.spatial[0];
-        auto y_size = mem_arg.size.spatial[1];
-        unsigned int input_it = 0;
-        for (cldnn::tensor::value_type x = 0; x < x_size; x++)
-        {
-            for (cldnn::tensor::value_type y = 0; y < y_size; y++)
-            {
-                for (cldnn::tensor::value_type i = 0; i < i_size; i++)
-                {
-                    for (cldnn::tensor::value_type o = 0; o < o_size; o++)
-                    {
-                        stream << convert_element(mem_ptr[input_it]) << " ";
-                        input_it++;
-                    }
-                    stream << std::endl;
-                }
-            }
-        }
-    }
-
-    template<typename elemType>
-    void dump_os_iyx_osv16(const cldnn::memory& mem, std::stringstream & stream)
-    {
-        auto mem_arg = mem.get_layout();
-        auto mem_ptr = mem.pointer<elemType>();
-
-        auto i_size = mem_arg.size.batch[0];
-        auto o_size = mem_arg.size.feature[0];
-        auto x_size = mem_arg.size.spatial[0];
-        auto y_size = mem_arg.size.spatial[1];
-        auto weights_size = i_size * o_size * x_size * y_size; //count() also counts feature[1]
-        int slice_value = 16;
-        cldnn::tensor::value_type it = 0;
-        while (it < weights_size)
-        {
-            stream << convert_element(mem_ptr[it]) << " ";
-            it++;
-            if (it % slice_value == 0) //separate every bsv with a new line
-                stream << std::endl;
-        };
-    }
-
-    template<typename elemType>
-    void dump_bs_xs_xsv8_bsv8(const cldnn::memory& mem, std::stringstream & stream)
-    {
-        auto mem_arg = mem.get_layout();
-        auto mem_ptr = mem.pointer<elemType>();
-
-        auto i_size = mem_arg.size.batch[0]; //batch = input feature map
-        auto x_size = mem_arg.size.spatial[0]; // spatial_x = output feature map
-        auto weights_size = mem_arg.size.count();
-        int xsv = 8, bsv = 8;
-        unsigned int input_it = 0, input_i_it= 0 , input_o_it = 0;
-        for (cldnn::tensor::value_type it = 0; it < weights_size; it++)
-        {
-                stream << convert_element(mem_ptr[input_it]) << " ";
-                input_i_it++;
-                if (input_i_it % bsv == 0) //separete every input slice with a new line
-                {
-                    stream << std::endl;
-                    input_o_it++;
-                    input_i_it = 0;
-                }
-                input_it = input_o_it*bsv + input_i_it;
-
-                if (input_it % (xsv*bsv) == 0) // seperate every block (8x8) with a new line
-                    stream << std::endl;
-        }
-    }
-
-    template<typename elemType>
-    void dump_bs_x_bsv16(const cldnn::memory& mem, std::stringstream & stream)
-    {
-        auto mem_arg = mem.get_layout();
-        auto mem_ptr = mem.pointer<elemType>();
-
-        auto i_size = mem_arg.size.batch[0]; //batch = input feature map
-        auto x_size = mem_arg.size.spatial[0]; // spatial_x = output feature map
-        auto weights_size = mem_arg.size.count();
-        int bsv = 16;
-        cldnn::tensor::value_type it = 0;
-        while (it < weights_size)
-        {
-            stream << convert_element(mem_ptr[it]) << " ";
-            it++;
-            if (it % bsv == 0) //separate every bsv with a new line
-                stream << std::endl;
-        }
-    }
-
-    template <class T>
-    void dump(const cldnn::memory& mem, std::stringstream& stream)
-    {
-        auto mem_ptr = mem.pointer<T>();
-
-        auto&& pitches = mem.get_layout().get_pitches();
-        auto&& size = mem.get_layout().size;
-        for (cldnn::tensor::value_type b = 0; b < size.batch[0]; ++b)
-        {
-            stream << "============= BATCH " << b << " ============\n\n";
-            for (cldnn::tensor::value_type f = 0; f < size.feature[0]; ++f)
-            {
-                stream << "feature " << f << ":\n";
-                for (cldnn::tensor::value_type y = 0; y < size.spatial[1]; ++y)
-                {
-                    for (cldnn::tensor::value_type x = 0; x < size.spatial[0]; ++x)
-                    {
-                        unsigned int input_it = b*pitches.batch[0] + f*pitches.feature[0] + y*pitches.spatial[1] + x*pitches.spatial[0];
-                        stream << convert_element(mem_ptr[input_it]) << " ";
-                        input_it++;
-                    }
-                    stream << '\n';
-                }
-                stream << std::endl;
-            }
-        }
-    }
-
-    template <class T>
-    void dump(const cldnn::memory& mem, std::vector<std::vector<std::string>>& dump_strings)
-    {
-        auto mem_ptr = mem.pointer<T>();
-        std::stringstream stream;
-
-        auto&& pitches = mem.get_layout().get_pitches();
-        auto&& size = mem.get_layout().size;
-        for (cldnn::tensor::value_type b = 0; b < size.batch[0]; ++b)
-        {
-            for (cldnn::tensor::value_type f = 0; f < size.feature[0]; ++f)
-            {
-                for (cldnn::tensor::value_type y = 0; y < size.spatial[1]; ++y)
-                {
-                    for (cldnn::tensor::value_type x = 0; x < size.spatial[0]; ++x)
-                    {
-                        unsigned int input_it = b*pitches.batch[0] + f*pitches.feature[0] + y*pitches.spatial[1] + x*pitches.spatial[0];
-                        stream << convert_element(mem_ptr[input_it]) << " ";
-                        input_it++;
-                    }
-                    stream << std::endl;
-                    dump_strings[b][f] = stream.str();
-                }
-            }
-        }
-    }
-
-    void logger::log_memory_to_file(const cldnn::memory& mem, std::string prefix, bool single_batch, cldnn::tensor::value_type batch_id, cldnn::tensor::value_type feature_id)
-    {
-        auto batch = mem.get_layout().size.batch[0];
-        auto feature = mem.get_layout().size.feature[0];
-        auto eng_type =  "gpu" ;
-        std::vector<std::vector<std::string>> dump_strings(batch);
-        for(cldnn::tensor::value_type b = 0; b < batch; b++)
-        {
-            dump_strings[b].resize(feature);
-        }
-
-        if (mem.get_layout().data_type == cldnn::data_types::f32)
-            dump<float>(mem, dump_strings);
-        else
-            dump<cldnn::half_t>(mem, dump_strings);
-
-        for (cldnn::tensor::value_type b = 0; b < batch; b++)
-            for (cldnn::tensor::value_type f = 0; f < feature; f++)
-            {
-                if (!single_batch || (b == batch_id && f == feature_id))
-                {
-                    std::string filename((dump_dir + "/" + prefix + "_" + eng_type + "_b" + std::to_string(b) + "_f" + std::to_string(f) + ".txt"));
-                    std::ofstream file_stream(filename);
-                    file_stream << dump_strings[b][f];
-                    file_stream.close();
-                }
-            }
-    }
-
-    void logger::log_weights_to_file(const cldnn::memory& mem, std::string prefix)
-    {
-        std::stringstream stream;
-
-        if (mem.get_layout().data_type == cldnn::data_types::f32)
-            dump<float>(mem, stream);
-        else
-            dump<cldnn::half_t>(mem, stream);
-
-        std::string filename((dump_dir + "/" + prefix + ".txt"));
-        std::ofstream file_stream(filename);
-        file_stream << stream.str();
-        file_stream.close();
-    }
-}
diff --git a/inference-engine/thirdparty/clDNN/tests/test_utils/instrumentation.h b/inference-engine/thirdparty/clDNN/tests/test_utils/instrumentation.h
deleted file mode 100644
index 8bb105356320c9..00000000000000
--- a/inference-engine/thirdparty/clDNN/tests/test_utils/instrumentation.h
+++ /dev/null
@@ -1,39 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-#include <chrono>
-#include <sstream>
-#include <iomanip>
-#include "api/memory.hpp"
-
-#define DUMP_DIRECTORY "./"
-
-namespace instrumentation {
-
-    template<class Rep, class Period>
-    std::string to_string(const std::chrono::duration<Rep, Period> val) {
-        namespace  ch = std::chrono;
-        const ch::microseconds us(1);
-        const ch::milliseconds ms(1);
-        const ch::seconds s(1);
-        const std::chrono::duration<Rep, Period> abs_val(std::abs(val.count()));
-
-        std::ostringstream os;
-        os << std::setprecision(3) << std::fixed;
-        if (abs_val > s)       os << ch::duration_cast<ch::duration<double, ch::seconds::period>>(val).count() << " s";
-        else if (abs_val > ms) os << ch::duration_cast<ch::duration<double, ch::milliseconds::period>>(val).count() << " ms";
-        else if (abs_val > us) os << ch::duration_cast<ch::duration<double, ch::microseconds::period>>(val).count() << " us";
-        else               os << ch::duration_cast<ch::nanoseconds>(val).count() << " ns";
-        return os.str();
-    }
-
-    struct logger
-    {
-        static void log_memory_to_file(const cldnn::memory&, std::string prefix = "", bool single_batch = false, cldnn::tensor::value_type batch_id = 0, cldnn::tensor::value_type feature_id = 0);
-        static void log_weights_to_file(const cldnn::memory&, std::string prefix = "");
-    private:
-        static const std::string dump_dir;
-    };
-}
diff --git a/inference-engine/thirdparty/clDNN/tests/test_utils/network_test.h b/inference-engine/thirdparty/clDNN/tests/test_utils/network_test.h
index e83e3b50335753..efc9d7b6cbfd41 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_utils/network_test.h
+++ b/inference-engine/thirdparty/clDNN/tests/test_utils/network_test.h
@@ -7,14 +7,15 @@
 #include <gtest/gtest.h>
 #include "test_utils/test_utils.h"
 
-#include "api/data.hpp"
-#include "api/engine.hpp"
-#include "api/fully_connected.hpp"
-#include "api/input_layout.hpp"
-#include "api/layout.hpp"
-#include "api/memory.hpp"
-#include "api/primitive.hpp"
-#include "api/tensor.hpp"
+#include <cldnn/runtime/engine.hpp>
+#include <cldnn/runtime/layout.hpp>
+#include <cldnn/runtime/memory.hpp>
+#include <cldnn/runtime/tensor.hpp>
+
+#include <cldnn/primitives/data.hpp>
+#include <cldnn/primitives/fully_connected.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/primitive.hpp>
 
 #include <map>
 #include <set>
@@ -57,7 +58,7 @@ struct typed_comparator<FLOAT16> {
 // =====================================================================================================================
 // Reference tensor
 struct reference_tensor {
-    virtual void compare(cldnn::memory actual) = 0;
+    virtual void compare(cldnn::memory::ptr actual) = 0;
 };
 
 template <typename T, size_t N>
@@ -68,22 +69,23 @@ struct reference_tensor_typed<T, 1> : reference_tensor {
     using vector_type = VF<T>;
     reference_tensor_typed(vector_type data) : reference(std::move(data)) {}
 
-    void compare(cldnn::memory actual) override {
-        auto ptr = actual.pointer<T>();
+    void compare(cldnn::memory::ptr actual) override {
+        cldnn::mem_lock<T> ptr(actual, get_test_stream());
+
         for (size_t bi = 0; bi < reference.size(); ++bi) {
             auto coords = cldnn::tensor(cldnn::batch(bi), cldnn::feature(0), cldnn::spatial(0, 0, 0, 0));
-            size_t offset = actual.get_layout().get_linear_offset(coords);
+            size_t offset = actual->get_layout().get_linear_offset(coords);
             auto& ref = reference[bi];
             auto& val = ptr[offset];
             TYPED_EXPECT_EQ(ref, val) << " at bi=" << bi;
         }
     }
 
-    void fill_memory(cldnn::memory mem) {
-        auto ptr = mem.pointer<T>();
+    void fill_memory(cldnn::memory::ptr mem) {
+        cldnn::mem_lock<T> ptr(mem, get_test_stream());
         for (size_t bi = 0; bi < reference.size(); ++bi) {
             auto coords = cldnn::tensor(cldnn::batch(bi), cldnn::feature(0), cldnn::spatial(0, 0, 0, 0));
-            size_t offset = mem.get_layout().get_linear_offset(coords);
+            size_t offset = mem->get_layout().get_linear_offset(coords);
             ptr[offset] = reference[bi];
         }
     }
@@ -100,12 +102,12 @@ struct reference_tensor_typed<T, 2> : reference_tensor {
     using vector_type = VVF<T>;
     reference_tensor_typed(vector_type data) : reference(std::move(data)) {}
 
-    void compare(cldnn::memory actual) override {
-        auto ptr = actual.pointer<T>();
+    void compare(cldnn::memory::ptr actual) override {
+        cldnn::mem_lock<T> ptr(actual, get_test_stream());
         for (size_t bi = 0; bi < reference.size(); ++bi) {
             for (size_t fi = 0; fi < reference[0].size(); ++fi) {
                 auto coords = cldnn::tensor(cldnn::batch(bi), cldnn::feature(fi), cldnn::spatial(0, 0, 0, 0));
-                size_t offset = actual.get_layout().get_linear_offset(coords);
+                size_t offset = actual->get_layout().get_linear_offset(coords);
                 auto& ref = reference[bi][fi];
                 auto& val = ptr[offset];
                 TYPED_EXPECT_EQ(ref, val) << "at bi=" << bi << " fi=" << fi;
@@ -113,12 +115,12 @@ struct reference_tensor_typed<T, 2> : reference_tensor {
         }
     }
 
-    void fill_memory(cldnn::memory mem) {
-        auto ptr = mem.pointer<T>();
+    void fill_memory(cldnn::memory::ptr mem) {
+        cldnn::mem_lock<T> ptr(mem, get_test_stream());
         for (size_t bi = 0; bi < reference.size(); ++bi) {
             for (size_t fi = 0; fi < reference[0].size(); ++fi) {
                 auto coords = cldnn::tensor(cldnn::batch(bi), cldnn::feature(fi), cldnn::spatial(0, 0, 0, 0));
-                size_t offset = mem.get_layout().get_linear_offset(coords);
+                size_t offset = mem->get_layout().get_linear_offset(coords);
                 ptr[offset] = reference[bi][fi];
             }
         }
@@ -135,14 +137,14 @@ template <typename T>
 struct reference_tensor_typed<T, 4> : reference_tensor {
     using vector_type = VVVVF<T>;
     reference_tensor_typed(vector_type data) : reference(std::move(data)) {}
-    void compare(cldnn::memory actual) override {
-        auto ptr = actual.pointer<T>();
+    void compare(cldnn::memory::ptr actual) override {
+        cldnn::mem_lock<T> ptr(actual, get_test_stream());
         for (size_t bi = 0; bi < reference.size(); ++bi) {
             for (size_t fi = 0; fi < reference[0].size(); ++fi) {
                 for (size_t yi = 0; yi < reference[0][0].size(); ++yi) {
                     for (size_t xi = 0; xi < reference[0][0][0].size(); ++xi) {
                         auto coords = cldnn::tensor(cldnn::batch(bi), cldnn::feature(fi), cldnn::spatial(xi, yi, 0, 0));
-                        size_t offset = actual.get_layout().get_linear_offset(coords);
+                        size_t offset = actual->get_layout().get_linear_offset(coords);
                         auto& ref = reference[bi][fi][yi][xi];
                         auto& val = ptr[offset];
                         TYPED_EXPECT_EQ(ref, val) << "at bi=" << bi << " fi=" << fi << " yi=" << yi << " xi=" << xi;
@@ -152,14 +154,14 @@ struct reference_tensor_typed<T, 4> : reference_tensor {
         }
     }
 
-    void fill_memory(cldnn::memory mem) {
-        auto ptr = mem.pointer<T>();
+    void fill_memory(cldnn::memory::ptr mem) {
+        cldnn::mem_lock<T> ptr(mem, get_test_stream());
         for (size_t bi = 0; bi < reference.size(); ++bi) {
             for (size_t fi = 0; fi < reference[0].size(); ++fi) {
                 for (size_t yi = 0; yi < reference[0][0].size(); ++yi) {
                     for (size_t xi = 0; xi < reference[0][0][0].size(); ++xi) {
                         auto coords = cldnn::tensor(cldnn::batch(bi), cldnn::feature(fi), cldnn::spatial(xi, yi, 0, 0));
-                        size_t offset = mem.get_layout().get_linear_offset(coords);
+                        size_t offset = mem->get_layout().get_linear_offset(coords);
                         ptr[offset] = reference[bi][fi][yi][xi];
                     }
                 }
@@ -257,6 +259,7 @@ struct reference_node_interface {
 
     virtual reference_tensor& get_reference() = 0;
     virtual cldnn::primitive_id get_id() = 0;
+    virtual ~reference_node_interface() = default;
 };
 
 template <typename T, size_t N>
@@ -275,7 +278,7 @@ struct reference_node : reference_node_interface {
 
 class network_test {
 public:
-    explicit network_test(cldnn::engine eng) : eng(eng) {}
+    explicit network_test(cldnn::engine& eng) : eng(eng) {}
 
     template <typename T, size_t N>
     typename reference_node<T, N>::ptr add_input_layout(cldnn::primitive_id id,
@@ -285,7 +288,7 @@ class network_test {
         auto shape = output.get_shape();
         auto lt = cldnn::layout(cldnn::type_to_data_type<T>::value, fmt, shape);
         topo.add(cldnn::input_layout(id, lt));
-        auto mem = cldnn::memory::allocate(eng, lt);
+        auto mem = eng.allocate_memory(lt);
         output.fill_memory(mem);
         inputs.emplace(id, mem);
         return add_node(id, std::move(output), {});
@@ -298,7 +301,7 @@ class network_test {
         auto output = reference_tensor_typed<T, N>(std::move(data));
         auto shape = output.get_shape();
         auto lt = cldnn::layout(cldnn::type_to_data_type<T>::value, fmt, shape);
-        auto mem = cldnn::memory::allocate(eng, lt);
+        auto mem = eng.allocate_memory(lt);
         output.fill_memory(mem);
         topo.add(cldnn::data(id, mem));
         return add_node(id, std::move(output), {});
@@ -384,10 +387,10 @@ class network_test {
         return node;
     }
 
-    cldnn::engine eng;
+    cldnn::engine& eng;
     cldnn::topology topo;
     std::map<cldnn::primitive_id, cldnn::implementation_desc> forced_impls;
-    std::map<cldnn::primitive_id, cldnn::memory> inputs;
+    std::map<cldnn::primitive_id, cldnn::memory::ptr> inputs;
     std::set<reference_node_interface::ptr> outputs;
 };
 
diff --git a/inference-engine/thirdparty/clDNN/tests/test_utils/test_utils.cpp b/inference-engine/thirdparty/clDNN/tests/test_utils/test_utils.cpp
index 86f0cfd7805c21..bfa964350dd7ab 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_utils/test_utils.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_utils/test_utils.cpp
@@ -4,415 +4,366 @@
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
-#include "api/memory.hpp"
-#include <api/primitive.hpp>
-#include <api/input_layout.hpp>
-#include <api/data.hpp>
-#include <api/topology.hpp>
-#include <api/network.hpp>
-#include <api/engine.hpp>
 #include "test_utils.h"
 #include "float16.h"
-#include "instrumentation.h"
 #include <iostream>
 
 using namespace cldnn;
 
-namespace tests
-{
-    const std::string graph_dump_dir = DUMP_DIRECTORY;
+namespace tests {
 
-    generic_test::generic_test() : generic_params(std::get<0>(GetParam())), layer_params(std::get<1>(GetParam())), max_ulps_diff_allowed(4), random_values(true), dump_graphs(false), dump_memory(false)
-    {
-    }
-    void generic_test::run_single_test()
-    {
-        assert((generic_params->data_type == data_types::f32) || (generic_params->data_type == data_types::f16));
-        if (dump_graphs)
-        {
-            generic_params->network_build_options.set_option(cldnn::build_option::graph_dumps_dir(DUMP_DIRECTORY));
-        }
-        topology topology;
-        topology.add_primitive(layer_params);
+generic_test::generic_test()
+    : generic_params(std::get<0>(GetParam()))
+    , layer_params(std::get<1>(GetParam()))
+    , max_ulps_diff_allowed(4)
+    , random_values(true) { }
 
-        std::vector<memory> input_mems;
-        std::vector<std::string> input_layouts_names = {};
+void generic_test::run_single_test() {
+    assert((generic_params->data_type == data_types::f32) || (generic_params->data_type == data_types::f16));
+    topology topology;
+    topology.add_primitive(layer_params);
 
-        size_t multipler = 0;
-        for (size_t i = 0 ; i < generic_params->input_layouts.size() ; i++)
-        {
-            input_mems.push_back( memory::allocate(engine, generic_params->input_layouts[i]) );
+    std::vector<memory::ptr> input_mems;
+    std::vector<std::string> input_layouts_names = {};
 
-            if (random_values)
-            {
-                if (generic_params->data_type == data_types::f32)
-                {
-                    tests::set_random_values<float>(input_mems[i], true, 7, 10);
-                }
-                else
-                {
-                    tests::set_random_values<FLOAT16>(input_mems[i], true, 5, 10);
-                }
+    size_t multipler = 0;
+    for (size_t i = 0 ; i < generic_params->input_layouts.size() ; i++) {
+        input_mems.push_back( engine.allocate_memory(generic_params->input_layouts[i]) );
+
+        if (random_values) {
+            if (generic_params->data_type == data_types::f32) {
+                tests::set_random_values<float>(input_mems[i], true, 7, 10);
+            } else {
+                tests::set_random_values<FLOAT16>(input_mems[i], true, 5, 10);
             }
-            else
-            {
-                size_t size = generic_params->input_layouts[i].size.batch[0] * generic_params->input_layouts[i].size.feature[0];
+        } else {
+            size_t size = generic_params->input_layouts[i].size.batch[0] * generic_params->input_layouts[i].size.feature[0];
 
-                if (generic_params->data_type == data_types::f32)
-                {
-                    std::vector<float> values;
-                    for (size_t j = 1; j <= size; j++)
-                    {
-                        values.push_back(static_cast<float>(multipler + j));
-                    }
-                    tests::set_values_per_batch_and_feature<float>(input_mems[i], values);
-                    multipler = values.size();
+            if (generic_params->data_type == data_types::f32) {
+                std::vector<float> values;
+                for (size_t j = 1; j <= size; j++) {
+                    values.push_back(static_cast<float>(multipler + j));
                 }
-                else
-                {
-                    std::vector<FLOAT16> values;
-                    for (size_t j = 1; j <= size; j++)
-                    {
-                        values.push_back(FLOAT16(static_cast<float>(multipler + j)));
-                    }
-                    tests::set_values_per_batch_and_feature<FLOAT16>(input_mems[i], values);
-                    multipler = values.size();
+                tests::set_values_per_batch_and_feature<float>(input_mems[i], values);
+                multipler = values.size();
+            } else {
+                std::vector<FLOAT16> values;
+                for (size_t j = 1; j <= size; j++) {
+                    values.push_back(FLOAT16(static_cast<float>(multipler + j)));
                 }
-            }
-            std::string input_name = "input" + std::to_string(i);
-            if ( (i == 0) && generic_params->network_build_options.get<cldnn::build_option_type::optimize_data>()->enabled() )
-            {
-                // Add reorder after the first input in case of optimize data flag since it might change the input layout.
-                input_name = "input0_init";
-            }
-
-            // First input is provided to the network as input_layout.
-            // Other inputs are provided as input_layout if optimize data flag is off. Otherwise they are provided as data.
-            if ( (i == 0) || !generic_params->network_build_options.get<cldnn::build_option_type::optimize_data>()->enabled())
-            {
-                topology.add(input_layout(input_name, input_mems[i].get_layout()));
-                input_layouts_names.push_back(input_name);
-            }
-            else
-            {
-                topology.add(data(input_name, input_mems[i]));
-            }
-
-            if (!is_format_supported(generic_params->fmt))
-            {
-                ASSERT_THROW(network bad(engine, topology), std::exception);
-                return;
+                tests::set_values_per_batch_and_feature<FLOAT16>(input_mems[i], values);
+                multipler = values.size();
             }
         }
-
-        if (generic_params->network_build_options.get<cldnn::build_option_type::optimize_data>()->enabled())
-        {
+        std::string input_name = "input" + std::to_string(i);
+        if ((i == 0) && generic_params->network_build_options.get<cldnn::build_option_type::optimize_data>()->enabled()) {
             // Add reorder after the first input in case of optimize data flag since it might change the input layout.
-            topology.add(reorder("input0", "input0_init", input_mems[0].get_layout()));
+            input_name = "input0_init";
         }
 
-        if (layer_params->input[0] == "reorder0")
-        {
-            // Add reorder layer with output padding as input to the tested layer.
-            topology.add(reorder("reorder0", "input0", input_mems[0].get_layout().with_padding(padding{ { 0, 0, 1, 3 },{ 0, 0, 5, 2 } })));
+        // First input is provided to the network as input_layout.
+        // Other inputs are provided as input_layout if optimize data flag is off. Otherwise they are provided as data.
+        if ((i == 0) || !generic_params->network_build_options.get<cldnn::build_option_type::optimize_data>()->enabled()) {
+            topology.add(input_layout(input_name, input_mems[i]->get_layout()));
+            input_layouts_names.push_back(input_name);
+        } else {
+            topology.add(data(input_name, input_mems[i]));
+        }
+
+        if (!is_format_supported(generic_params->fmt)) {
+            ASSERT_THROW(network bad(engine, topology), std::exception);
+            return;
         }
+    }
 
-        prepare_input_for_test(input_mems);
+    if (generic_params->network_build_options.get<cldnn::build_option_type::optimize_data>()->enabled()) {
+        // Add reorder after the first input in case of optimize data flag since it might change the input layout.
+        topology.add(reorder("input0", "input0_init", input_mems[0]->get_layout()));
+    }
 
-        network network(engine, topology, generic_params->network_build_options);
+    if (layer_params->input[0] == "reorder0") {
+        // Add reorder layer with output padding as input to the tested layer.
+        topology.add(reorder("reorder0", "input0", input_mems[0]->get_layout().with_padding(padding{ { 0, 0, 1, 3 },{ 0, 0, 5, 2 } })));
+    }
 
-        for (size_t i = 0 ; i < input_layouts_names.size() ; i++)
-        {
-            network.set_input_data(input_layouts_names[i], input_mems[i]);
-        }
+    prepare_input_for_test(input_mems);
 
-        auto outputs = network.execute();
-        EXPECT_EQ(outputs.size(), size_t(1));
+    network network(engine, topology, generic_params->network_build_options);
 
-        auto output = outputs.begin()->second.get_memory();
+    for (size_t i = 0 ; i < input_layouts_names.size() ; i++) {
+        network.set_input_data(input_layouts_names[i], input_mems[i]);
+    }
 
-        auto output_ref = generate_reference(input_mems);
+    auto outputs = network.execute();
+    EXPECT_EQ(outputs.size(), size_t(1));
 
+    auto output = outputs.begin()->second.get_memory();
 
-        if (dump_memory)
-        {
-            std::string prefix = test_info.name();
-            for (size_t i = 0; i < generic_params->input_layouts.size(); i++)
-            {
-                ::instrumentation::logger::log_memory_to_file(input_mems[i], prefix + "input" + std::to_string(i));
-            }
-            for (size_t i = 0; i < outputs.size(); i++)
-            {
-                ::instrumentation::logger::log_memory_to_file(output, prefix + "output" + std::to_string(i));
-            }
-        }
+    auto output_ref = generate_reference(input_mems);
 
-        if (output.get_layout().data_type == data_types::f32)
-        {
-            compare_buffers<float>(output, output_ref);
-        }
-        else
-        {
-            compare_buffers<FLOAT16>(output, output_ref);
-        }
+    if (output->get_layout().data_type == data_types::f32) {
+        compare_buffers<float>(output, output_ref);
+    } else {
+        compare_buffers<FLOAT16>(output, output_ref);
     }
+}
 
-    template<typename Type>
-    void generic_test::compare_buffers(const memory& out, const memory& ref)
-    {
-        auto out_layout = out.get_layout();
-        auto ref_layout = ref.get_layout();
+template<typename Type>
+void generic_test::compare_buffers(const memory::ptr out, const memory::ptr ref) {
+    auto out_layout = out->get_layout();
+    auto ref_layout = ref->get_layout();
 
-        EXPECT_EQ(out_layout.size, ref_layout.size);
-        EXPECT_EQ(out_layout.data_type, ref_layout.data_type);
-        EXPECT_EQ(get_expected_output_tensor(), out_layout.size);
-        EXPECT_EQ(out_layout.get_linear_size(), ref_layout.get_linear_size());
-        EXPECT_EQ(out_layout.data_padding, ref_layout.data_padding);
+    EXPECT_EQ(out_layout.size, ref_layout.size);
+    EXPECT_EQ(out_layout.data_type, ref_layout.data_type);
+    EXPECT_EQ(get_expected_output_tensor(), out_layout.size);
+    EXPECT_EQ(out_layout.get_linear_size(), ref_layout.get_linear_size());
+    EXPECT_EQ(out_layout.data_padding, ref_layout.data_padding);
 
-        auto output_size = out_layout.size;
+    auto output_size = out_layout.size;
 
-        int batch_size = output_size.batch[0];
-        int feature_size = output_size.feature[0];
-        int y_size = output_size.spatial[1];
-        int x_size = output_size.spatial[0];
+    int batch_size = output_size.batch[0];
+    int feature_size = output_size.feature[0];
+    int y_size = output_size.spatial[1];
+    int x_size = output_size.spatial[0];
 
-        auto res_data = out.pointer<Type>();
-        auto ref_data = ref.pointer<Type>();
+    mem_lock<Type> res_data(out, get_test_stream());
+    mem_lock<Type> ref_data(ref, get_test_stream());
 
-        const auto out_desc = get_linear_memory_desc(out_layout);
-        const auto ref_desc = get_linear_memory_desc(ref_layout);
+    const auto out_desc = get_linear_memory_desc(out_layout);
+    const auto ref_desc = get_linear_memory_desc(ref_layout);
 
-        for (int b = 0; b < batch_size; b++)
-        {
-            for (int f = 0; f < feature_size; f++)
-            {
-                for (int y = 0; y < y_size; y++)
-                {
-                    for (int x = 0; x < x_size; x++)
-                    {
-                        size_t res_index = get_linear_index(out_layout, b, f, y, x, out_desc);
-                        size_t ref_index = get_linear_index(ref_layout, b, f, y, x, ref_desc);
+    for (int b = 0; b < batch_size; b++) {
+        for (int f = 0; f < feature_size; f++) {
+            for (int y = 0; y < y_size; y++) {
+                for (int x = 0; x < x_size; x++) {
+                    size_t res_index = get_linear_index(out_layout, b, f, y, x, out_desc);
+                    size_t ref_index = get_linear_index(ref_layout, b, f, y, x, ref_desc);
 
-                        EXPECT_TRUE(floating_point_equal(res_data[res_index], ref_data[ref_index], max_ulps_diff_allowed))
-                            << "Expected " << (float)res_data[res_index] << " to be almost equal (within " << max_ulps_diff_allowed << " ULP's) to " << (float)ref_data[ref_index]
-                            << " (ref index = " << ref_index << ", B " << b << ", F "<< f << ", Y " << y << ", X " << x << ")!";
+                    EXPECT_TRUE(floating_point_equal(res_data[res_index], ref_data[ref_index], max_ulps_diff_allowed))
+                        << "Expected " << (float)res_data[res_index] << " to be almost equal (within "
+                        << max_ulps_diff_allowed << " ULP's) to " << (float)ref_data[ref_index]
+                        << " (ref index = " << ref_index << ", B " << b << ", F "<< f << ", Y " << y << ", X " << x << ")!";
 
-                        if (HasFailure())
-                        {
-                            return;
-                        }
+                    if (HasFailure()) {
+                        return;
                     }
                 }
             }
         }
     }
+}
 
-    static size_t calc_offfset(const layout & layout, const pitches& p)
-    {
-        auto lower_padding = layout.data_padding.lower_size();
-        if (layout.format == format::bfzyx) {
-            return
-                p.b * lower_padding.batch[0] +
-                p.f * lower_padding.feature[0] +
-                p.z * lower_padding.spatial[2] +
-                p.y * lower_padding.spatial[1] +
-                p.x * lower_padding.spatial[0];
-        }
-        else {
-            return
-                p.b * lower_padding.batch[0] +
-                p.f * lower_padding.feature[0] +
-                p.y * lower_padding.spatial[1] +
-                p.x * lower_padding.spatial[0];
-        }
+static size_t calc_offfset(const layout & layout, const pitches& p) {
+    auto lower_padding = layout.data_padding.lower_size();
+    if (layout.format == format::bfzyx) {
+        return
+            p.b * lower_padding.batch[0] +
+            p.f * lower_padding.feature[0] +
+            p.z * lower_padding.spatial[2] +
+            p.y * lower_padding.spatial[1] +
+            p.x * lower_padding.spatial[0];
+    } else {
+        return
+            p.b * lower_padding.batch[0] +
+            p.f * lower_padding.feature[0] +
+            p.y * lower_padding.spatial[1] +
+            p.x * lower_padding.spatial[0];
     }
+}
 
-    memory_desc generic_test::get_linear_memory_desc(const layout & layout)
-    {
-        pitches p;
+memory_desc generic_test::get_linear_memory_desc(const layout & layout) {
+    pitches p;
 
-        switch (layout.format)
-        {
-            case format::bfyx:
-            {
-                p.x = 1;
-                p.y = layout.get_buffer_size().sizes(format::bfyx)[3] * p.x;
-                p.f = layout.get_buffer_size().sizes(format::bfyx)[2] * p.y;
-                p.b = layout.get_buffer_size().sizes(format::bfyx)[1] * p.f;
-                break;
-            }
-            case format::yxfb:
-            {
-                p.b = 1;
-                p.f = layout.get_buffer_size().sizes(format::yxfb)[3] * p.b;
-                p.x = layout.get_buffer_size().sizes(format::yxfb)[2] * p.f;
-                p.y = layout.get_buffer_size().sizes(format::yxfb)[1] * p.x;
-                break;
-            }
-            case format::fyxb:
-            {
-                p.b = 1;
-                p.x = layout.get_buffer_size().sizes(format::fyxb)[3] * p.b;
-                p.y = layout.get_buffer_size().sizes(format::fyxb)[2] * p.x;
-                p.f = layout.get_buffer_size().sizes(format::fyxb)[1] * p.y;
-                break;
-            }
-            case format::byxf:
-            {
-                p.f = 1;
-                p.x = layout.get_buffer_size().sizes(format::byxf)[3] * p.f;
-                p.y = layout.get_buffer_size().sizes(format::byxf)[2] * p.x;
-                p.b = layout.get_buffer_size().sizes(format::byxf)[1] * p.y;
-                break;
-            }
-            case format::bfzyx:
-            {
-                p.x = 1;
-                p.y = layout.get_buffer_size().sizes(format::bfzyx)[4] * p.x;
-                p.z = layout.get_buffer_size().sizes(format::bfzyx)[3] * p.y;
-                p.f = layout.get_buffer_size().sizes(format::bfzyx)[2] * p.z;
-                p.b = layout.get_buffer_size().sizes(format::bfzyx)[1] * p.f;
-                break;
-            }
-            default:
-            {
-                throw std::runtime_error("Format not supported yet.");
-            }
+    switch (layout.format) {
+        case format::bfyx: {
+            p.x = 1;
+            p.y = layout.get_buffer_size().sizes(format::bfyx)[3] * p.x;
+            p.f = layout.get_buffer_size().sizes(format::bfyx)[2] * p.y;
+            p.b = layout.get_buffer_size().sizes(format::bfyx)[1] * p.f;
+            break;
+        }
+        case format::yxfb: {
+            p.b = 1;
+            p.f = layout.get_buffer_size().sizes(format::yxfb)[3] * p.b;
+            p.x = layout.get_buffer_size().sizes(format::yxfb)[2] * p.f;
+            p.y = layout.get_buffer_size().sizes(format::yxfb)[1] * p.x;
+            break;
+        }
+        case format::fyxb: {
+            p.b = 1;
+            p.x = layout.get_buffer_size().sizes(format::fyxb)[3] * p.b;
+            p.y = layout.get_buffer_size().sizes(format::fyxb)[2] * p.x;
+            p.f = layout.get_buffer_size().sizes(format::fyxb)[1] * p.y;
+            break;
+        }
+        case format::byxf: {
+            p.f = 1;
+            p.x = layout.get_buffer_size().sizes(format::byxf)[3] * p.f;
+            p.y = layout.get_buffer_size().sizes(format::byxf)[2] * p.x;
+            p.b = layout.get_buffer_size().sizes(format::byxf)[1] * p.y;
+            break;
+        }
+        case format::bfzyx: {
+            p.x = 1;
+            p.y = layout.get_buffer_size().sizes(format::bfzyx)[4] * p.x;
+            p.z = layout.get_buffer_size().sizes(format::bfzyx)[3] * p.y;
+            p.f = layout.get_buffer_size().sizes(format::bfzyx)[2] * p.z;
+            p.b = layout.get_buffer_size().sizes(format::bfzyx)[1] * p.f;
+            break;
+        }
+        default: {
+            throw std::runtime_error("Format not supported yet.");
         }
-
-        return{ p, calc_offfset(layout, p) };
     }
 
-    size_t generic_test::get_linear_index(const layout&, size_t b, size_t f, size_t y, size_t x, const memory_desc& desc)
-    {
-        return
-            desc.offset +
-            b*desc.pitch.b +
-            f*desc.pitch.f +
-            y*desc.pitch.y +
-            x*desc.pitch.x;
-    }
+    return {p, calc_offfset(layout, p)};
+}
 
-    size_t generic_test::get_linear_index(const layout&, size_t b, size_t f, size_t z, size_t y, size_t x, const memory_desc& desc)
-    {
-        return
-            desc.offset +
-            b*desc.pitch.b +
-            f*desc.pitch.f +
-            z*desc.pitch.z +
-            y*desc.pitch.y +
-            x*desc.pitch.x;
-    }
+size_t generic_test::get_linear_index(const layout&, size_t b, size_t f, size_t y, size_t x, const memory_desc& desc)
+{
+    return
+        desc.offset +
+        b*desc.pitch.b +
+        f*desc.pitch.f +
+        y*desc.pitch.y +
+        x*desc.pitch.x;
+}
 
-    size_t generic_test::get_linear_index_with_broadcast(const layout& in_layout, size_t b, size_t f, size_t y, size_t x, const memory_desc& desc)
-    {
-        return
-            desc.offset +
-            (b % in_layout.size.batch[0]) * desc.pitch.b +
-            (f % in_layout.size.feature[0]) * desc.pitch.f +
-            (y % in_layout.size.spatial[1]) * desc.pitch.y +
-            (x % in_layout.size.spatial[0]) * desc.pitch.x;
-    }
+size_t generic_test::get_linear_index(const layout&, size_t b, size_t f, size_t z, size_t y, size_t x, const memory_desc& desc)
+{
+    return
+        desc.offset +
+        b*desc.pitch.b +
+        f*desc.pitch.f +
+        z*desc.pitch.z +
+        y*desc.pitch.y +
+        x*desc.pitch.x;
+}
 
-    //Default implementation. Should be overridden in derived class otherwise.
-    cldnn::tensor generic_test::get_expected_output_tensor()
-    {
-        return generic_params->input_layouts[0].size;
-    }
+size_t generic_test::get_linear_index_with_broadcast(const layout& in_layout, size_t b, size_t f, size_t y, size_t x, const memory_desc& desc)
+{
+    return
+        desc.offset +
+        (b % in_layout.size.batch[0]) * desc.pitch.b +
+        (f % in_layout.size.feature[0]) * desc.pitch.f +
+        (y % in_layout.size.spatial[1]) * desc.pitch.y +
+        (x % in_layout.size.spatial[0]) * desc.pitch.x;
+}
 
-    std::vector<std::shared_ptr<test_params>> generic_test::generate_generic_test_params(std::vector<std::shared_ptr<test_params>>& all_generic_params)
-    {
-        // , { format::yx,{ 531,777 } } , { format::yx,{ 4096,1980 } } ,
-        //{ format::bfyx,{ 1,1,1,1 } } , { format::bfyx,{ 1,1,2,2 } } , { format::yx,{ 3,3 } } , { format::yx,{ 4,4 } } , { format::bfyx,{ 1,1,5,5 } } , { format::yx,{ 6,6 } } , { format::yx,{ 7,7 } } ,
-        //{ format::yx,{ 8,8 } } , { format::yx,{ 9,9 } } , { format::yx,{ 10,10 } } , { format::yx,{ 11,11 } } , { format::yx,{ 12,12 } } , { format::yx,{ 13,13 } } ,
-        //{ format::yx,{ 14,14 } } , { format::yx,{ 15,15 } } , { format::yx,{ 16,16 } } };
+//Default implementation. Should be overridden in derived class otherwise.
+cldnn::tensor generic_test::get_expected_output_tensor()
+{
+    return generic_params->input_layouts[0].size;
+}
 
-        auto data_types = test_data_types();
+std::vector<std::shared_ptr<test_params>> generic_test::generate_generic_test_params(std::vector<std::shared_ptr<test_params>>& all_generic_params)
+{
+    // , { format::yx,{ 531,777 } } , { format::yx,{ 4096,1980 } } ,
+    //{ format::bfyx,{ 1,1,1,1 } } , { format::bfyx,{ 1,1,2,2 } } , { format::yx,{ 3,3 } } , { format::yx,{ 4,4 } } , { format::bfyx,{ 1,1,5,5 } } , { format::yx,{ 6,6 } } , { format::yx,{ 7,7 } } ,
+    //{ format::yx,{ 8,8 } } , { format::yx,{ 9,9 } } , { format::yx,{ 10,10 } } , { format::yx,{ 11,11 } } , { format::yx,{ 12,12 } } , { format::yx,{ 13,13 } } ,
+    //{ format::yx,{ 14,14 } } , { format::yx,{ 15,15 } } , { format::yx,{ 16,16 } } };
+
+    auto data_types = test_data_types();
 
-        for (cldnn::data_types data_type : data_types)
+    for (cldnn::data_types data_type : data_types)
+    {
+        for (cldnn::format fmt : test_input_formats)
         {
-            for (cldnn::format fmt : test_input_formats)
+            for (int batch_size : test_batch_sizes)
             {
-                for (int batch_size : test_batch_sizes)
+                for (int feature_size : test_feature_sizes)
                 {
-                    for (int feature_size : test_feature_sizes)
+                    for (tensor input_size : test_input_sizes)
                     {
-                        for (tensor input_size : test_input_sizes)
-                        {
-                            all_generic_params.emplace_back(new test_params(data_type, fmt, batch_size, feature_size, input_size));
-                        }
+                        all_generic_params.emplace_back(new test_params(data_type, fmt, batch_size, feature_size, input_size));
                     }
                 }
             }
         }
-
-        return all_generic_params;
     }
 
-    const cldnn::engine & get_test_engine()
-    {
-        static const cldnn::engine engine;
-        return engine;
-    }
+    return all_generic_params;
+}
 
-    const std::string test_dump::name() const
-    {
-        std::string temp = name_str;
-        std::replace(temp.begin(), temp.end(), '/', '_');
-        return temp;
+cldnn::engine_configuration get_test_engine_config() {
+    const bool enable_profiling = false;
+    const cldnn::queue_types queue_type = cldnn::queue_types::out_of_order;
+    std::string sources_dumps_dir = "";
+    priority_mode_types priority_mode = priority_mode_types::disabled;
+    throttle_mode_types throttle_mode = throttle_mode_types::disabled;
+    bool use_memory_pool = true;
+    bool use_unified_shared_memory = true;
+    return engine_configuration(enable_profiling, queue_type, sources_dumps_dir, priority_mode, throttle_mode, use_memory_pool, use_unified_shared_memory);
+}
+
+std::shared_ptr<cldnn::engine> create_test_engine() {
+    return cldnn::engine::create(engine_types::ocl, runtime_types::ocl, get_test_engine_config());
+}
+
+cldnn::engine& get_test_engine() {
+    static std::shared_ptr<cldnn::engine> test_engine = nullptr;
+    if (!test_engine) {
+        test_engine = create_test_engine();
     }
+    return *test_engine;
+}
 
-    const std::string test_dump::test_case_name() const
-    {
-        size_t pos = test_case_name_str.find("/");
-        if (pos > test_case_name_str.length())
-        {
-            pos = 0;
-        }
-        std::string temp = test_case_name_str.substr(pos);
-        return temp;
+cldnn::stream& get_test_stream() {
+    static std::shared_ptr<cldnn::stream> test_stream = nullptr;
+    if (!test_stream)
+        test_stream = get_test_engine().create_stream();
+    return *test_stream;
+}
+
+const std::string test_dump::name() const {
+    std::string temp = name_str;
+    std::replace(temp.begin(), temp.end(), '/', '_');
+    return temp;
+}
+
+const std::string test_dump::test_case_name() const {
+    size_t pos = test_case_name_str.find("/");
+    if (pos > test_case_name_str.length()) {
+        pos = 0;
     }
+    std::string temp = test_case_name_str.substr(pos);
+    return temp;
+}
 
-    std::string test_params::print_tensor(cldnn::tensor t)
-    {
-        std::stringstream str;
-        for (size_t i = 0; i < t.sizes(format::bfyx).size(); i++)
-        {
-            str << t.sizes(format::bfyx)[i] << " ";
-        }
-        str << "]";
-        return str.str();
+std::string test_params::print_tensor(cldnn::tensor t) {
+    std::stringstream str;
+    for (size_t i = 0; i < t.sizes(format::bfyx).size(); i++) {
+        str << t.sizes(format::bfyx)[i] << " ";
     }
+    str << "]";
+    return str.str();
+}
 
-    std::string test_params::print()
-    {
-        std::stringstream str;
-        str << "Data type: " << data_type_traits::name(data_type) << std::endl;
+std::string test_params::print() {
+    std::stringstream str;
+    str << "Data type: " << data_type_traits::name(data_type) << std::endl;
 
-        for (int j = 0 ; j < (int)input_layouts.size(); j++)
-        {
-            const cldnn::tensor& t = input_layouts[j].size;
+    for (int j = 0 ; j < (int)input_layouts.size(); j++) {
+        const cldnn::tensor& t = input_layouts[j].size;
 
-            str << "Input " << j << ": " << print_tensor(t) << std::endl;
-        }
-        return str.str();
+        str << "Input " << j << ": " << print_tensor(t) << std::endl;
     }
+    return str.str();
+}
 
-    std::vector<cldnn::data_types> generic_test::test_data_types()
-    {
-        std::vector<cldnn::data_types> result;
-        result.push_back(cldnn::data_types::f32);
+std::vector<cldnn::data_types> generic_test::test_data_types() {
+    std::vector<cldnn::data_types> result;
+    result.push_back(cldnn::data_types::f32);
 
-        if(get_test_engine().get_info().supports_fp16)
-        {
-            result.push_back(cldnn::data_types::f16);
-        }
-        return result;
+    if (get_test_engine().get_device_info().supports_fp16) {
+        result.push_back(cldnn::data_types::f16);
     }
+    return result;
+}
 
-    std::vector<cldnn::format> generic_test::test_input_formats = { cldnn::format::bfyx , cldnn::format::yxfb, cldnn::format::fyxb, cldnn::format::byxf };
-    std::vector<int32_t> generic_test::test_batch_sizes = { 1, 2 };// 4, 8, 16};
-    std::vector<int32_t> generic_test::test_feature_sizes = { 1, 2 };// , 3, 15};
-    std::vector<tensor> generic_test::test_input_sizes = { { 1, 1, 100, 100 } ,{ 1, 1, 277, 277 } ,{ 1, 1, 400, 600 } };
+std::vector<cldnn::format> generic_test::test_input_formats = { cldnn::format::bfyx , cldnn::format::yxfb, cldnn::format::fyxb, cldnn::format::byxf };
+std::vector<int32_t> generic_test::test_batch_sizes = { 1, 2 };// 4, 8, 16};
+std::vector<int32_t> generic_test::test_feature_sizes = { 1, 2 };// , 3, 15};
+std::vector<tensor> generic_test::test_input_sizes = { { 1, 1, 100, 100 } ,{ 1, 1, 277, 277 } ,{ 1, 1, 400, 600 } };
 
-}
+}  // namespace tests
diff --git a/inference-engine/thirdparty/clDNN/tests/test_utils/test_utils.h b/inference-engine/thirdparty/clDNN/tests/test_utils/test_utils.h
index 436df5abb1e34e..bfe6cfc17988cc 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_utils/test_utils.h
+++ b/inference-engine/thirdparty/clDNN/tests/test_utils/test_utils.h
@@ -6,35 +6,49 @@
 
 #pragma once
 
-#include "api/memory.hpp"
-#include "api/tensor.hpp"
-#include "api/program.hpp"
-#include "api/network.hpp"
+#include <cldnn/runtime/memory.hpp>
+#include <cldnn/runtime/tensor.hpp>
+#include <cldnn/runtime/engine.hpp>
+#include <cldnn/runtime/stream.hpp>
+#include <cldnn/graph/program.hpp>
+#include <cldnn/graph/network.hpp>
+#include <cldnn/graph/topology.hpp>
+#include <cldnn/primitives/primitive.hpp>
+#include <cldnn/primitives/concatenation.hpp>
+#include <cldnn/primitives/lrn.hpp>
+#include <cldnn/primitives/roi_pooling.hpp>
+#include <cldnn/primitives/scale.hpp>
+#include <cldnn/primitives/softmax.hpp>
+#include <cldnn/primitives/reorder.hpp>
+#include <cldnn/primitives/normalize.hpp>
+#include <cldnn/primitives/convolution.hpp>
+#include <cldnn/primitives/activation.hpp>
+#include <cldnn/primitives/pooling.hpp>
+#include <cldnn/primitives/input_layout.hpp>
+#include <cldnn/primitives/data.hpp>
+
+#include <gtest/gtest.h>
+#include <gmock/gmock.h>
+#include "float16.h"
+#include "random_gen.h"
+#include "uniform_quantized_real_distribution.hpp"
+#include <src/include/to_string_utils.h>
+
 #include <iostream>
 #include <limits>
 #include <random>
 #include <algorithm>
 #include <memory>
-#include <gtest/gtest.h>
-#include <api/primitive.hpp>
-#include "float16.h"
-#include "random_gen.h"
-#include "api/concatenation.hpp"
-#include "api/lrn.hpp"
-#include "api/roi_pooling.hpp"
-#include "api/scale.hpp"
-#include "api/softmax.hpp"
-#include "api/reorder.hpp"
-#include "api/normalize.hpp"
-#include "api/convolution.hpp"
-#include "api/activation.hpp"
-#include "api/pooling.hpp"
-
 #include <chrono>
 
 #define ARRAY_SIZE(a) (sizeof(a) / sizeof(a[0]))
 
 namespace tests {
+
+std::shared_ptr<cldnn::engine> create_test_engine();
+cldnn::engine& get_test_engine();
+cldnn::stream& get_test_stream();
+
 #define USE_RANDOM_SEED 0
 #if USE_RANDOM_SEED
     std::random_device rnd_device;
@@ -228,12 +242,12 @@ VVVVVVF<T> generate_random_6d(size_t a, size_t b, size_t c, size_t d, size_t e,
     return v;
 }
 
-template <class T> void set_value(const cldnn::pointer<T>& ptr, uint32_t index, T value) { ptr[index] = value; }
-template <class T> T    get_value(const cldnn::pointer<T>& ptr, uint32_t index) { return ptr[index]; }
+template <class T> void set_value(T* ptr, uint32_t index, T value) { ptr[index] = value; }
+template <class T> T    get_value(T* ptr, uint32_t index) { return ptr[index]; }
 
 template<typename T>
-void set_values(const cldnn::memory& mem, std::initializer_list<T> args ){
-    auto ptr = mem.pointer<T>();
+void set_values(cldnn::memory::ptr mem, std::initializer_list<T> args) {
+    cldnn::mem_lock<T> ptr(mem, get_test_stream());
 
     auto it = ptr.begin();
     for(auto x : args)
@@ -241,8 +255,8 @@ void set_values(const cldnn::memory& mem, std::initializer_list<T> args ){
 }
 
 template<typename T>
-void set_values(const cldnn::memory& mem, std::vector<T> args) {
-    auto ptr = mem.pointer<T>();
+void set_values(cldnn::memory::ptr mem, std::vector<T> args) {
+    cldnn::mem_lock<T> ptr(mem, get_test_stream());
 
     auto it = ptr.begin();
     for (auto x : args)
@@ -250,19 +264,14 @@ void set_values(const cldnn::memory& mem, std::vector<T> args) {
 }
 
 template<typename T>
-void set_values_per_batch_and_feature(const cldnn::memory& mem, std::vector<T> args)
-{
-    auto mem_ptr = mem.pointer<T>();
-    auto&& pitches = mem.get_layout().get_pitches();
-    auto&& size = mem.get_layout().size;
-    for (cldnn::tensor::value_type b = 0; b < size.batch[0]; ++b)
-    {
-        for (cldnn::tensor::value_type f = 0; f < size.feature[0]; ++f)
-        {
-            for (cldnn::tensor::value_type y = 0; y < size.spatial[1]; ++y)
-            {
-                for (cldnn::tensor::value_type x = 0; x < size.spatial[0]; ++x)
-                {
+void set_values_per_batch_and_feature(cldnn::memory::ptr mem, std::vector<T> args) {
+    cldnn::mem_lock<T> mem_ptr(mem, get_test_stream());
+    auto&& pitches = mem->get_layout().get_pitches();
+    auto&& size = mem->get_layout().size;
+    for (cldnn::tensor::value_type b = 0; b < size.batch[0]; ++b) {
+        for (cldnn::tensor::value_type f = 0; f < size.feature[0]; ++f) {
+            for (cldnn::tensor::value_type y = 0; y < size.spatial[1]; ++y) {
+                for (cldnn::tensor::value_type x = 0; x < size.spatial[0]; ++x) {
                     unsigned int input_it = b*pitches.batch[0] + f*pitches.feature[0] + y*pitches.spatial[1] + x*pitches.spatial[0];
                     mem_ptr[input_it] = args[b*size.feature[0] + f];
                 }
@@ -274,42 +283,37 @@ void set_values_per_batch_and_feature(const cldnn::memory& mem, std::vector<T> a
 
 template<typename T, typename std::enable_if<std::is_floating_point<T>::value ||
                                              std::is_same<T, FLOAT16>::value>::type* = nullptr>
-void set_random_values(const cldnn::memory& mem, bool sign = false, unsigned significand_bit = 8, unsigned scale = 1)
+void set_random_values(cldnn::memory::ptr mem, bool sign = false, unsigned significand_bit = 8, unsigned scale = 1)
 {
-    auto ptr = mem.pointer<T>();
+    cldnn::mem_lock<T> ptr(mem, get_test_stream());
 
     std::mt19937 gen;
-    for (auto it = ptr.begin(); it != ptr.end(); ++it)
-    {
+    for (auto it = ptr.begin(); it != ptr.end(); ++it) {
         *it = rnd_generators::gen_number<T>(gen, significand_bit, sign, false, scale);
     }
 }
 
 template<class T, typename std::enable_if<std::is_integral<T>::value>::type* = nullptr>
-void set_random_values(const cldnn::memory& mem)
+void set_random_values(cldnn::memory::ptr mem)
 {
-    auto ptr = mem.pointer<T>();
+    cldnn::mem_lock<T> ptr(mem, get_test_stream());
 
     std::mt19937 gen;
     static std::uniform_int_distribution<T> uid(std::numeric_limits<T>::min(), std::numeric_limits<T>::max());
-    for (auto it = ptr.begin(); it != ptr.end(); ++it)
-    {
+    for (auto it = ptr.begin(); it != ptr.end(); ++it) {
         *it = uid(gen);
     }
 }
 
 // Tries to construct a network, checking if an expected error appears
-inline void check_exception_massage(const cldnn::engine& engine, cldnn::topology& topology, std::string msg_to_find)
-{
+inline void check_exception_massage(cldnn::engine& engine, cldnn::topology& topology, std::string msg_to_find) {
     try {
         cldnn::network(engine, topology);
-    }
-    catch (std::exception & exc) {
+    } catch (std::exception & exc) {
         std::string msg(exc.what());
         if (msg.find(msg_to_find) != std::string::npos) {
             throw;
-        }
-        else {
+        } else {
             printf("%s\n", exc.what());
         }
     }
@@ -372,20 +376,15 @@ inline bool floating_point_equal(float x, float y, int max_ulps_diff = 4) {
     }
 }
 
-class test_params
-{
+class test_params {
 public:
 
-    test_params() :
-        fmt(cldnn::format::bfyx)
-    {
-    }
+    test_params() : fmt(cldnn::format::bfyx) { }
 
     test_params(cldnn::data_types dt, cldnn::format input_format, int32_t batch_size, int32_t feature_size, cldnn::tensor input_size, cldnn::build_options const& options = cldnn::build_options()) :
         data_type(dt),
         fmt(input_format),
-        network_build_options(options)
-    {
+        network_build_options(options) {
         cldnn::tensor t = cldnn::tensor(batch_size, feature_size, input_size.spatial[0],  input_size.spatial[1] );
         input_layouts.push_back( cldnn::layout(dt, fmt, t) );
     }
@@ -402,38 +401,32 @@ class test_params
     static std::string print_tensor(cldnn::tensor tensor);
 };
 
-struct pitches
-{
+struct pitches {
     size_t b, f, y, x, z;
 };
 
-struct memory_desc
-{
+struct memory_desc {
     pitches pitch;
     size_t offset;
 };
 
-const cldnn::engine & get_test_engine();
-
-struct test_dump
-{
+struct test_dump {
     const std::string name() const;
     const std::string test_case_name() const;
+
 private:
     const std::string test_case_name_str = ::testing::UnitTest::GetInstance()->current_test_info()->test_case_name();
     const std::string name_str = ::testing::UnitTest::GetInstance()->current_test_info()->name();
 };
 
-class generic_test : public ::testing::TestWithParam<std::tuple<std::shared_ptr<test_params>, std::shared_ptr<cldnn::primitive>>>
-{
-
+class generic_test : public ::testing::TestWithParam<std::tuple<std::shared_ptr<test_params>, std::shared_ptr<cldnn::primitive>>> {
 public:
     generic_test();
 
     void run_single_test();
 
     template<typename Type>
-    void compare_buffers(const cldnn::memory& out, const cldnn::memory& ref);
+    void compare_buffers(const cldnn::memory::ptr out, const cldnn::memory::ptr ref);
 
     static size_t get_linear_index(const cldnn::layout & layout, size_t b, size_t f, size_t y, size_t x, const memory_desc& desc);
     static size_t get_linear_index(const cldnn::layout & layout, size_t b, size_t f, size_t z, size_t y, size_t x, const memory_desc& desc);
@@ -456,7 +449,7 @@ class generic_test : public ::testing::TestWithParam<std::tuple<std::shared_ptr<
     };
 
 protected:
-    const cldnn::engine& engine = get_test_engine();
+    cldnn::engine& engine = get_test_engine();
     std::shared_ptr<test_params> generic_params;
     test_dump test_info;
     std::shared_ptr<cldnn::primitive> layer_params;
@@ -464,10 +457,10 @@ class generic_test : public ::testing::TestWithParam<std::tuple<std::shared_ptr<
     bool random_values; // if set memory buffers will be filled with random values
     bool dump_graphs; // if set tests will dump graphs to file
     bool dump_memory; // if set memory buffers will be dumped to file
-    virtual cldnn::memory generate_reference(const std::vector<cldnn::memory>& inputs) = 0;
+    virtual cldnn::memory::ptr generate_reference(const std::vector<cldnn::memory::ptr>& inputs) = 0;
     // Allows the test to override the random input data that the framework generates
 
-    virtual void prepare_input_for_test(std::vector<cldnn::memory>& /*inputs*/) { }
+    virtual void prepare_input_for_test(std::vector<cldnn::memory::ptr>& /*inputs*/) { }
 
     static std::vector<cldnn::data_types> test_data_types();
     static std::vector<cldnn::format> test_input_formats;
@@ -480,8 +473,7 @@ class generic_test : public ::testing::TestWithParam<std::tuple<std::shared_ptr<
 // When a test assertion such as EXPECT_EQ fails, Google-Test prints the argument values to help with debugging.
 // It does this using a user - extensible value printer.
 // This function will be used to print the test params in case of an error.
-inline void PrintTupleTo(const std::tuple<std::shared_ptr<test_params>, std::shared_ptr<cldnn::primitive>>& t, ::std::ostream* os)
-{
+inline void PrintTupleTo(const std::tuple<std::shared_ptr<test_params>, std::shared_ptr<cldnn::primitive>>& t, ::std::ostream* os) {
     std::stringstream str;
 
     auto test_param = std::get<0>(t);
@@ -495,13 +487,10 @@ inline void PrintTupleTo(const std::tuple<std::shared_ptr<test_params>, std::sha
 
     //TODO: do layers not have param dumping? we could consider adding it
 
-    if (primitive->type == cldnn::concatenation::type_id())
-    {
+    if (primitive->type == cldnn::concatenation::type_id()) {
         auto dc = std::static_pointer_cast<cldnn::concatenation>(primitive);
         (void)dc;
-    }
-    else if(primitive->type == cldnn::lrn::type_id())
-    {
+    } else if(primitive->type == cldnn::lrn::type_id()) {
         auto lrn = std::static_pointer_cast<cldnn::lrn >(primitive);
         std::string norm_region = (lrn->norm_region == cldnn::lrn_norm_region_across_channel) ? "across channel" : "within channel";
         str << "Norm region: " << norm_region
@@ -509,9 +498,7 @@ inline void PrintTupleTo(const std::tuple<std::shared_ptr<test_params>, std::sha
             << " Alpha: " << lrn->alpha
             << " Beta: " << lrn->beta
             << " K: " << lrn->k;
-    }
-    else if(primitive->type == cldnn::roi_pooling::type_id())
-    {
+    } else if(primitive->type == cldnn::roi_pooling::type_id()) {
         auto p = std::static_pointer_cast<cldnn::roi_pooling >(primitive);
         str << "Pooling mode: " << (p->mode == cldnn::pooling_mode::max ? "MAX" : "AVG")
             << " Pooled width: " << p->pooled_width
@@ -520,51 +507,35 @@ inline void PrintTupleTo(const std::tuple<std::shared_ptr<test_params>, std::sha
             << " Spatial bins x: " << p->spatial_bins_x
             << " Spatial bins y: " << p->spatial_bins_y
             << " Output dim: " << p->output_dim;
-    }
-    else if(primitive->type == cldnn::scale::type_id())
-    {
+    } else if(primitive->type == cldnn::scale::type_id()) {
         auto s = std::static_pointer_cast<cldnn::scale >(primitive);
         (void)s;
-    }
-    else if(primitive->type == cldnn::softmax::type_id())
-    {
+    } else if(primitive->type == cldnn::softmax::type_id()) {
         auto sm = std::static_pointer_cast<cldnn::softmax >(primitive);
         (void)sm;
-    }
-    else if (primitive->type == cldnn::reorder::type_id())
-    {
+    } else if (primitive->type == cldnn::reorder::type_id()) {
         auto reorder = std::static_pointer_cast<cldnn::reorder>(primitive);
         str << "Output data type: " << cldnn::data_type_traits::name(*reorder->output_data_type) << " Mean: " << reorder->mean << "Subtract per feature: " << "TODO" /*std::vector<float> subtract_per_feature*/;
-    }
-    else if (primitive->type == cldnn::normalize::type_id())
-    {
+    } else if (primitive->type == cldnn::normalize::type_id()) {
         auto normalize = std::static_pointer_cast<cldnn::normalize>(primitive);
         std::string norm_region = normalize->across_spatial ? "across_spatial" : "within_spatial";
         str << "Norm region: " << norm_region << " Epsilon: " << normalize->epsilon << " Scale input id: " << normalize->scale_input;
-    }
-    else if (primitive->type == cldnn::convolution::type_id())
-    {
+    } else if (primitive->type == cldnn::convolution::type_id()) {
         auto convolution = std::static_pointer_cast<cldnn::convolution>(primitive);
         str << "Stride x: " << convolution->stride.spatial[0] << " Stride y: " << convolution->stride.spatial[1]
             << " Dilation x: " << convolution->dilation.spatial[0] << " Dilation y: " << convolution->dilation.spatial[1]
             << " Input offset x: " << convolution->input_offset.spatial[0] << " Input offset y: " << convolution->input_offset.spatial[1];
-    }
-    else if (primitive->type == cldnn::activation::type_id())
-    {
+    } else if (primitive->type == cldnn::activation::type_id()) {
         auto activation = std::static_pointer_cast<cldnn::activation>(primitive);
         str << "Negative slope: " << activation->additional_params.a << " Negative slope input id: " << activation->additional_params_input;
-    }
-    else if (primitive->type == cldnn::pooling::type_id())
-    {
+    } else if (primitive->type == cldnn::pooling::type_id()) {
         auto pooling = std::static_pointer_cast<cldnn::pooling>(primitive);
         std::string pooling_mode = (pooling->mode == cldnn::pooling_mode::max) ? "max" : "average";
         str << "Pooling mode: " << pooling_mode
             << " Input offset x: " << pooling->input_offset.spatial[0] << " Input offset y: " << pooling->input_offset.spatial[1]
             << " Stride x: " << pooling->stride.spatial[0] << " Stride y: " << pooling->stride.spatial[1]
             << " Size x: " << pooling->size.spatial[0] << " Size y: " << pooling->size.spatial[1];
-    }
-    else
-    {
+    } else {
         throw std::runtime_error("Not implemented yet for this primitive.");
     }
 
@@ -577,130 +548,130 @@ T div_up(const T a, const U b) {
     return (a + b - 1) / b;
 }
 
-inline void print_bin_blob(cldnn::memory& mem, std::string name)
-{
-    auto&& size = mem.get_layout().size;
-
-    std::cerr << name;
-    std::cerr << " shape: ";
-    std::cerr << size.batch[0] << " ";
-    std::cerr << size.feature[0] << " ";
-    std::cerr << size.spatial[1] << " ";
-    std::cerr << size.spatial[0] << " ";
-    std::cerr << "(" << size.batch[0] * size.feature[0] * size.spatial[1] * size.spatial[0] << ")" << std::endl;
-
-    auto mem_ptr = mem.pointer<uint32_t>();
-
-    bool packed_ic = mem.get_layout().format == cldnn::format::b_fs_yx_32fp ? 1 : 0;
-    int B = size.batch[0];
-    int C = size.feature[0];
-    int H = size.spatial[1];
-    int W = size.spatial[0];
-
-    for (cldnn::tensor::value_type b = 0; b < B; ++b)
-    {
-        for (cldnn::tensor::value_type f = 0; f < C; ++f)
-        {
-            for (cldnn::tensor::value_type y = 0; y < H; ++y)
-            {
-                for (cldnn::tensor::value_type x = 0; x < W; ++x)
-                {
-                    if (!packed_ic)
-                    {
-                        size_t input_it = b * C*H*W + f * W*H + y * W + x;
-                        size_t elem = input_it / 32;
-                        size_t bit = input_it % 32;
-                        std::cerr << ((mem_ptr[elem] & (1 << bit)) >> bit) << " ";
-                    }
-                    else
-                    {
-                        size_t input_it = b * (C / 32)*W*H + (f / 32)*W*H + y * W + x;
-                        size_t bit = f % 32;
-                        std::cerr << ((mem_ptr[input_it] & (1 << bit)) >> bit) << " ";
-                    }
-                }
-                std::cerr << std::endl;
-            }
-            std::cerr << std::endl;
-        }
-        std::cerr << "==============" << std::endl;
-    }
-}
-
-inline void print_bin_blob_packed(cldnn::memory& mem, std::string name)
-{
-    auto&& size = mem.get_layout().size;
-
-    std::cerr << name;
-    std::cerr << " shape: ";
-    std::cerr << size.batch[0] << " ";
-    std::cerr << size.feature[0] << " ";
-    std::cerr << size.spatial[1] << " ";
-    std::cerr << size.spatial[0] << " ";
-    std::cerr << "(" << size.batch[0] * size.feature[0] * size.spatial[1] * size.spatial[0] << ")" << std::endl;
-
-    auto mem_ptr = mem.pointer<uint32_t>();
-
-    int B = size.batch[0];
-    int C = size.feature[0];
-    int H = size.spatial[1];
-    int W = size.spatial[0];
-
-    for (cldnn::tensor::value_type b = 0; b < B; ++b)
-    {
-        for (cldnn::tensor::value_type f = 0; f < div_up(C, 32); ++f)
-        {
-            for (cldnn::tensor::value_type y = 0; y < H; ++y)
-            {
-                for (cldnn::tensor::value_type x = 0; x < W; ++x)
-                {
-                    size_t input_it = b * div_up(C, 32)*W*H + f * W*H + y * W + x;
-                    std::cerr << mem_ptr[input_it] << " ";
-                }
-                std::cerr << std::endl;
-            }
-            std::cerr << std::endl;
-        }
-        std::cerr << "==============" << std::endl;
-    }
-}
+// inline void print_bin_blob(cldnn::memory& mem, std::string name)
+// {
+//     auto&& size = mem.get_layout().size;
+
+//     std::cerr << name;
+//     std::cerr << " shape: ";
+//     std::cerr << size.batch[0] << " ";
+//     std::cerr << size.feature[0] << " ";
+//     std::cerr << size.spatial[1] << " ";
+//     std::cerr << size.spatial[0] << " ";
+//     std::cerr << "(" << size.batch[0] * size.feature[0] * size.spatial[1] * size.spatial[0] << ")" << std::endl;
+
+//     auto mem_ptr = mem.pointer<uint32_t>();
+
+//     bool packed_ic = mem.get_layout().format == cldnn::format::b_fs_yx_32fp ? 1 : 0;
+//     int B = size.batch[0];
+//     int C = size.feature[0];
+//     int H = size.spatial[1];
+//     int W = size.spatial[0];
+
+//     for (cldnn::tensor::value_type b = 0; b < B; ++b)
+//     {
+//         for (cldnn::tensor::value_type f = 0; f < C; ++f)
+//         {
+//             for (cldnn::tensor::value_type y = 0; y < H; ++y)
+//             {
+//                 for (cldnn::tensor::value_type x = 0; x < W; ++x)
+//                 {
+//                     if (!packed_ic)
+//                     {
+//                         size_t input_it = b * C*H*W + f * W*H + y * W + x;
+//                         size_t elem = input_it / 32;
+//                         size_t bit = input_it % 32;
+//                         std::cerr << ((mem_ptr[elem] & (1 << bit)) >> bit) << " ";
+//                     }
+//                     else
+//                     {
+//                         size_t input_it = b * (C / 32)*W*H + (f / 32)*W*H + y * W + x;
+//                         size_t bit = f % 32;
+//                         std::cerr << ((mem_ptr[input_it] & (1 << bit)) >> bit) << " ";
+//                     }
+//                 }
+//                 std::cerr << std::endl;
+//             }
+//             std::cerr << std::endl;
+//         }
+//         std::cerr << "==============" << std::endl;
+//     }
+// }
+
+// inline void print_bin_blob_packed(cldnn::memory& mem, std::string name)
+// {
+//     auto&& size = mem.get_layout().size;
+
+//     std::cerr << name;
+//     std::cerr << " shape: ";
+//     std::cerr << size.batch[0] << " ";
+//     std::cerr << size.feature[0] << " ";
+//     std::cerr << size.spatial[1] << " ";
+//     std::cerr << size.spatial[0] << " ";
+//     std::cerr << "(" << size.batch[0] * size.feature[0] * size.spatial[1] * size.spatial[0] << ")" << std::endl;
+
+//     auto mem_ptr = mem.pointer<uint32_t>();
+
+//     int B = size.batch[0];
+//     int C = size.feature[0];
+//     int H = size.spatial[1];
+//     int W = size.spatial[0];
+
+//     for (cldnn::tensor::value_type b = 0; b < B; ++b)
+//     {
+//         for (cldnn::tensor::value_type f = 0; f < div_up(C, 32); ++f)
+//         {
+//             for (cldnn::tensor::value_type y = 0; y < H; ++y)
+//             {
+//                 for (cldnn::tensor::value_type x = 0; x < W; ++x)
+//                 {
+//                     size_t input_it = b * div_up(C, 32)*W*H + f * W*H + y * W + x;
+//                     std::cerr << mem_ptr[input_it] << " ";
+//                 }
+//                 std::cerr << std::endl;
+//             }
+//             std::cerr << std::endl;
+//         }
+//         std::cerr << "==============" << std::endl;
+//     }
+// }
+
+// inline void print_blob(cldnn::memory& mem, std::string name)
+// {
+//     auto&& size = mem.get_layout().size;
+
+//     std::cerr << name;
+//     std::cerr << " shape: ";
+//     std::cerr << size.batch[0] << " ";
+//     std::cerr << size.feature[0] << " ";
+//     std::cerr << size.spatial[1] << " ";
+//     std::cerr << size.spatial[0] << " ";
+//     std::cerr << "(" << size.batch[0] * size.feature[0] * size.spatial[1] * size.spatial[0] << ")" << std::endl;
+
+//     auto mem_ptr = mem.pointer<float>();
+
+//     int B = size.batch[0];
+//     int C = size.feature[0];
+//     int H = size.spatial[1];
+//     int W = size.spatial[0];
+
+//     for (cldnn::tensor::value_type b = 0; b < B; ++b)
+//     {
+//         for (cldnn::tensor::value_type f = 0; f < C; ++f)
+//         {
+//             for (cldnn::tensor::value_type y = 0; y < H; ++y)
+//             {
+//                 for (cldnn::tensor::value_type x = 0; x < W; ++x)
+//                 {
+//                     size_t input_it = b * C*W*H + f * W*H + y * W + x;
+//                     std::cerr << std::setw(4) << mem_ptr[input_it] << " ";
+//                 }
+//                 std::cerr << std::endl;
+//             }
+//             std::cerr << std::endl;
+//         }
+//         std::cerr << "==============" << std::endl;
+//     }
+// }
 
-inline void print_blob(cldnn::memory& mem, std::string name)
-{
-    auto&& size = mem.get_layout().size;
-
-    std::cerr << name;
-    std::cerr << " shape: ";
-    std::cerr << size.batch[0] << " ";
-    std::cerr << size.feature[0] << " ";
-    std::cerr << size.spatial[1] << " ";
-    std::cerr << size.spatial[0] << " ";
-    std::cerr << "(" << size.batch[0] * size.feature[0] * size.spatial[1] * size.spatial[0] << ")" << std::endl;
-
-    auto mem_ptr = mem.pointer<float>();
-
-    int B = size.batch[0];
-    int C = size.feature[0];
-    int H = size.spatial[1];
-    int W = size.spatial[0];
-
-    for (cldnn::tensor::value_type b = 0; b < B; ++b)
-    {
-        for (cldnn::tensor::value_type f = 0; f < C; ++f)
-        {
-            for (cldnn::tensor::value_type y = 0; y < H; ++y)
-            {
-                for (cldnn::tensor::value_type x = 0; x < W; ++x)
-                {
-                    size_t input_it = b * C*W*H + f * W*H + y * W + x;
-                    std::cerr << std::setw(4) << mem_ptr[input_it] << " ";
-                }
-                std::cerr << std::endl;
-            }
-            std::cerr << std::endl;
-        }
-        std::cerr << "==============" << std::endl;
-    }
-}
 } // namespace tests
-
diff --git a/inference-engine/thirdparty/clDNN/tests/test_utils/uniform_quantized_real_distribution.hpp b/inference-engine/thirdparty/clDNN/tests/test_utils/uniform_quantized_real_distribution.hpp
index ea817fae29e242..464e15db83fa65 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_utils/uniform_quantized_real_distribution.hpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_utils/uniform_quantized_real_distribution.hpp
@@ -20,7 +20,7 @@
 #include <string>
 #include <utility>
 
-namespace cldnn { namespace tests { namespace distributions {
+namespace tests { namespace distributions {
 
 /// @cond PRIVATE
 namespace detail
@@ -812,7 +812,7 @@ class data_type_mask
         return os;
     }
 
-    
+
 
     /// @brief Deserialize from stream.
     ///
@@ -1537,4 +1537,5 @@ class uniform_quantized_real_distribution
     underlying_dist_type _base_distribution; ///< Base distribution used to generate input numbers.
 };
 
-}}} // namespace cldnn { namespace tests { namespace distributions
+}  // namespace distributions
+}  // namespace tests
diff --git a/inference-engine/thirdparty/clDNN/tests_core_internal/CMakeLists.txt b/inference-engine/thirdparty/clDNN/tests_core_internal/CMakeLists.txt
deleted file mode 100644
index 663f50000cc599..00000000000000
--- a/inference-engine/thirdparty/clDNN/tests_core_internal/CMakeLists.txt
+++ /dev/null
@@ -1,137 +0,0 @@
-# Copyright (C) 2018-2021 Intel Corporation
-# SPDX-License-Identifier: Apache-2.0
-#
-
-# ========================================= Name / Output settings =====================================
-
-set(CLDNN_BUILD__PROJ             "clDNN_tests_core_internal")
-set(CLDNN_BUILD__PROJ_LABEL       "${CLDNN_BUILD__PROJ}")
-set(CLDNN_BUILD__PROJ_OUTPUT_NAME "${CLDNN_BUILD__PROJ}${CLDNN__OUT_CPU_SUFFIX}")
-
-# =========================================== Compiler options =========================================
-intel_config_flag_apply_settings(CompilerOptions CMAKE_CXX_FLAGS ALL_PATTERN ""
-    SET
-	  StandardCxx11
-      RttiEnabled
-  )
-
-if (NOT MSVC)
-  intel_config_flag_apply_settings(CompilerOptions CMAKE_CXX_FLAGS ALL_PATTERN ""
-      SET_RAW
-        "-Wno-error=conversion-null"
-        "-Wno-error=type-limits"
-    )
-endif ()
-
-find_package(OpenMP)
-if (OPENMP_FOUND)
-    add_definitions(-DOPENMP_FOUND)
-    set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}")
-    set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
-endif()
-
-# ================================== Compiler preprocessor definitions =================================
-
-# ========================================= Source/Header files ========================================
-
-set(__CLDNN_Label__main                "")
-file(GLOB __CLDNN_Sources__main
-    "${CMAKE_CURRENT_SOURCE_DIR}/*.h"
-    "${CMAKE_CURRENT_SOURCE_DIR}/*.hpp"
-    "${CMAKE_CURRENT_SOURCE_DIR}/*.cpp"
-  )
-
-set(__CLDNN_Directory__test_cases      "${CMAKE_CURRENT_SOURCE_DIR}/test_cases")
-set(__CLDNN_Label__test_cases          "test cases")
-file(GLOB __CLDNN_Sources__test_cases
-    "${__CLDNN_Directory__test_cases}/*.h"
-    "${__CLDNN_Directory__test_cases}/*.hpp"
-    "${__CLDNN_Directory__test_cases}/*.cpp"
-  )
-
-set(__CLDNN_Directory__test_utils      "${CMAKE_CURRENT_SOURCE_DIR}/../tests/test_utils")
-set(__CLDNN_Label__test_utils          "test utils")
-file(GLOB __CLDNN_Sources__test_utils
-    "${__CLDNN_Directory__test_utils}/*.h"
-    "${__CLDNN_Directory__test_utils}/*.hpp"
-    "${__CLDNN_Directory__test_utils}/*.cpp"
-  )
-
-set(__CLDNN_Directory__gtest           "${CLDNN__GTEST_DIR}")
-set(__CLDNN_Label__gtest               "google test framework")
-file(GLOB __CLDNN_Sources__gtest
-    "${__CLDNN_Directory__gtest}/*.cc"
-  )
-
-if(UNIX)
-    set_source_files_properties(${__CLDNN_Sources__gtest} PROPERTIES
-        COMPILE_FLAGS "-Wno-undef")
-endif()
-
-set(__CLDNN_AllSources
-    ${__CLDNN_Sources__main}
-    ${__CLDNN_Sources__test_cases}
-    ${__CLDNN_Sources__test_utils}
-    ${__CLDNN_Sources__gtest}
-  )
-# Helping with some generators.
-set_property(SOURCE ${__CLDNN_Sources__cg_cache} PROPERTY GENERATED TRUE)
-
-
-# =============================================== Filters ==============================================
-
-source_group("${__CLDNN_Label__main}"                 FILES ${__CLDNN_Sources__main})
-source_group("${__CLDNN_Label__test_cases}"           FILES ${__CLDNN_Sources__test_cases})
-source_group("${__CLDNN_Label__test_utils}"           FILES ${__CLDNN_Sources__test_utils})
-
-# ===================================== Include/Link directories =======================================
-
-include_directories(
-    "${CLDNN__MAIN_DIR}"
-    "${CLDNN__MAIN_DIR}/src"
-    "${CLDNN__MAIN_DIR}/src/include"
-    "${CLDNN__KERNEL_SELECTOR_DIR}/core"
-    "${CLDNN__KERNEL_SELECTOR_DIR}/core/common"
-    "${CLDNN__KERNEL_SELECTOR_DIR}/common"
-    "${__CLDNN_Directory__test_utils}"
-    "${CMAKE_CURRENT_SOURCE_DIR}"
-  )
-
-include_directories(SYSTEM
-    "${CLDNN__GTEST_DIR}")
-
-# =================================== Link targets and dependencies ====================================
-
-# Tests executable.
-add_executable("${CLDNN_BUILD__PROJ}"
-    ${__CLDNN_AllSources}
-  )
-
-if(COMMAND set_ie_threading_interface_for)
-  set_ie_threading_interface_for("${CLDNN_BUILD__PROJ}")
-endif()
-
-set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY PROJECT_LABEL "${CLDNN_BUILD__PROJ_LABEL}")
-set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY OUTPUT_NAME   "${CLDNN_BUILD__PROJ_OUTPUT_NAME}")
-
-# Set library dependencies
-target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE "${CLDNN_BUILD__PROJ__clDNN}")
-
-if(WIN32)
-  target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE setupapi)
-elseif((NOT ANDROID) AND (UNIX))
-  target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE pthread)
-endif()
-target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE ${CLDNN__SYSTEM_LINK_LIBRARIES})
-
-# =================================== Custom pre- and post-steps =======================================
-
-if(CLDNN__RUN_TESTS)
-  add_custom_command(TARGET "${CLDNN_BUILD__PROJ}" POST_BUILD
-      WORKING_DIRECTORY "${CMAKE_RUNTIME_OUTPUT_DIRECTORY}"
-      COMMAND "${CLDNN_BUILD__PROJ}"
-      COMMENT "Executing tests..."
-    )
-endif()
-
-# ======================================================================================================
diff --git a/inference-engine/thirdparty/clDNN/tests_core_internal/main.cpp b/inference-engine/thirdparty/clDNN/tests_core_internal/main.cpp
deleted file mode 100644
index 7e583e9b068903..00000000000000
--- a/inference-engine/thirdparty/clDNN/tests_core_internal/main.cpp
+++ /dev/null
@@ -1,11 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "gtest/gtest.h"
-
-int main(int argc, char* argv[])
-{
-    ::testing::InitGoogleTest(&argc, argv);
-    return RUN_ALL_TESTS();
-}
\ No newline at end of file

From b6d7af12ae84653c622b62a46756e5ddd030b523 Mon Sep 17 00:00:00 2001
From: Anna Khakimova <anna.khakimova@intel.com>
Date: Wed, 16 Jun 2021 10:07:46 +0300
Subject: [PATCH 103/178] PP: Color conversion kernels refactoring (#5985)

* [PP] POC of vectorised kernels refactoring

 - base/framework changes

* [PP] POC of vectorised kernels refactoring

 - ChanToPlane implementation

* [PP] POC of vectorised kernels refactoring

 - ChanToPlane, switched order of dispatching to first choose ISA type

* [PP] POC of vectorised kernels refactoring

 - ChanToPlane, moved choose ISA stage to kernel package generation

* Preparing

* NV12ToRGB kernel refactoring

* * I420ToRGB kernel refactoring

Co-authored-by: Anton Potapov <anton.potapov@intel.com>
---
 .../ie_preprocess_gapi_kernels_neon.cpp       |  31 +-
 .../ie_preprocess_gapi_kernels_neon.hpp       |  29 +-
 .../ie_preprocess_gapi_kernels_avx2.cpp       |  31 +-
 .../ie_preprocess_gapi_kernels_avx2.hpp       |  34 +-
 .../ie_preprocess_gapi_kernels_avx512.cpp     |  31 +-
 .../ie_preprocess_gapi_kernels_avx512.hpp     |  31 +-
 .../ie_preprocess_gapi_kernels_sse42.cpp      |  30 +-
 .../ie_preprocess_gapi_kernels_sse42.hpp      |  30 +-
 .../ie_preprocess_gapi_kernels.cpp            | 579 ++++++++++--------
 .../ie_preprocess_gapi_kernels_impl.hpp       |  31 +
 .../ie_preprocess_gapi_kernels_simd_impl.hpp  | 134 ++--
 11 files changed, 504 insertions(+), 487 deletions(-)

diff --git a/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.cpp b/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.cpp
index 779db927c32cb5..8290fda25ed8ea 100644
--- a/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.cpp
+++ b/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.cpp
@@ -90,21 +90,6 @@ void splitRow_32FC4(const float in[], float out0[], float out1[],
     splitRow_32FC4_Impl(in, out0, out1, out2, out3, length);
 }
 
-void calculate_nv12_to_rgb(const  uchar **srcY,
-                           const  uchar *srcUV,
-                                  uchar **dstRGBx,
-                                    int width) {
-    calculate_nv12_to_rgb_impl(srcY, srcUV, dstRGBx, width);
-}
-
-void calculate_i420_to_rgb(const  uchar **srcY,
-                           const  uchar *srcU,
-                           const  uchar *srcV,
-                                  uchar **dstRGBx,
-                                    int width) {
-    calculate_i420_to_rgb_impl(srcY, srcU, srcV, dstRGBx, width);
-}
-
 void calcRowArea_8U(uchar dst[], const uchar *src[], const Size& inSz,
                     const Size& outSz, Q0_16 yalpha, const MapperUnit8U &ymap,
                     int xmaxdf, const short xindex[], const Q0_16 xalpha[],
@@ -119,14 +104,6 @@ void calcRowArea_32F(float dst[], const float *src[], const Size& inSz,
     calcRowArea_impl(dst, src, inSz, outSz, yalpha, ymap, xmaxdf, xindex, xalpha, vbuf);
 }
 
-void copyRow_8U(const uint8_t in[], uint8_t out[], int length) {
-    copyRow_8U_impl(in, out, length);
-}
-
-void copyRow_32F(const float in[], float out[], int length) {
-    copyRow_32F_impl(in, out, length);
-}
-
 // Resize (bi-linear, 32F)
 void calcRowLinear_32F(float* dst[],
                        const float* src0[],
@@ -708,6 +685,14 @@ void calcRowLinear_8UC1(uint8_t* dst[],
     }
 }
 }  // namespace neon
+
+template void chanToPlaneRowImpl(neon_tag, const uint8_t* in, int chan, int chs, uint8_t* out, const int length);
+template void chanToPlaneRowImpl(neon_tag, const float*   in, int chan, int chs, float  * out, const int length);
+
+template void nv12ToRgbRowImpl(neon_tag, const uint8_t** y_rows, const uint8_t* uv_row, uint8_t** out_rows, const int buf_width);
+
+template void i420ToRgbRowImpl(neon_tag, const uint8_t** y_rows, const uint8_t* u_row,
+                               const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.hpp b/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.hpp
index f3b21f73a95b44..4e0b82a62596c7 100644
--- a/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.hpp
+++ b/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.hpp
@@ -167,26 +167,31 @@ void splitRow_32FC4(const float in[],
                           float out3[],
                             int length);
 
-void calculate_nv12_to_rgb(const  uchar **srcY,
-                           const  uchar *srcUV,
-                                  uchar **dstRGBx,
-                                    int width);
-
 void calculate_i420_to_rgb(const  uchar **srcY,
                            const  uchar *srcU,
                            const  uchar *srcV,
                                   uchar **dstRGBx,
                                     int width);
 
-void copyRow_8U(const uint8_t in[],
-                uint8_t out[],
-                int length);
+}  // namespace neon
 
-void copyRow_32F(const float in[],
-                 float out[],
-                 int length);
+template<typename isa_tag_t, typename T>
+void chanToPlaneRowImpl(isa_tag_t, const T* in, const int chan, const int chs, T* out, const int length);
 
-}  // namespace neon
+extern template void chanToPlaneRowImpl(neon_tag, const uint8_t* in, const int chan, const int chs, uint8_t* out, const int length);
+extern template void chanToPlaneRowImpl(neon_tag, const float*   in, const int chan, const int chs, float  * out, const int length);
+
+template<typename isa_tag_t>
+void nv12ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* uv_row, uint8_t** out_rows, const int buf_width);
+
+extern template void nv12ToRgbRowImpl(neon_tag, const uint8_t** y_rows, const uint8_t* uv_row, uint8_t** out_rows, const int buf_width);
+
+template<typename isa_tag_t>
+void i420ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* u_row,
+                             const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+extern template void i420ToRgbRowImpl(neon_tag, const uint8_t** y_rows, const uint8_t* u_row,
+                                      const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.cpp b/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.cpp
index 22d86e10538df4..aefd2e2dfc33ce 100644
--- a/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.cpp
+++ b/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.cpp
@@ -107,21 +107,6 @@ void splitRow_32FC4(const float in[], float out0[], float out1[],
     splitRow_32FC4_Impl(in, out0, out1, out2, out3, length);
 }
 
-void calculate_nv12_to_rgb(const  uchar **srcY,
-                           const  uchar *srcUV,
-                                  uchar **dstRGBx,
-                                    int width) {
-    calculate_nv12_to_rgb_impl(srcY, srcUV, dstRGBx, width);
-}
-
-void calculate_i420_to_rgb(const  uchar **srcY,
-                           const  uchar *srcU,
-                           const  uchar *srcV,
-                                  uchar **dstRGBx,
-                                    int width) {
-    calculate_i420_to_rgb_impl(srcY, srcU, srcV, dstRGBx, width);
-}
-
 void calcRowArea_8U(uchar dst[], const uchar *src[], const Size& inSz,
                     const Size& outSz, Q0_16 yalpha, const MapperUnit8U &ymap,
                     int xmaxdf, const short xindex[], const Q0_16 xalpha[],
@@ -555,13 +540,6 @@ void calcRowLinear_8U(C4, std::array<std::array<uint8_t*, 4>, 4> &dst,
     calcRowLinear_8UC_Impl<chanNum>(dst, src0, src1, alpha, clone, mapsx, beta, tmp, inSz, outSz, lpi);
 }
 
-void copyRow_8U(const uint8_t in[], uint8_t out[], int length) {
-    copyRow_8U_impl(in, out, length);
-}
-void copyRow_32F(const float in[], float out[], int length) {
-    copyRow_32F_impl(in, out, length);
-}
-
 void calcRowLinear_32F(float *dst[],
                        const float *src0[],
                        const float *src1[],
@@ -575,6 +553,15 @@ void calcRowLinear_32F(float *dst[],
 }
 
 }  // namespace avx
+
+template void chanToPlaneRowImpl(avx2_tag, const uint8_t* in, const int chan, const int chs, uint8_t* out, const int length);
+template void chanToPlaneRowImpl(avx2_tag, const float*   in, const int chan, const int chs, float*   out, const int length);
+
+template void nv12ToRgbRowImpl(avx2_tag, const uint8_t** y_rows, const uint8_t* uv_row,
+                               uint8_t** out_rows, const int buf_width);
+
+template void i420ToRgbRowImpl(avx2_tag, const uint8_t** y_rows, const uint8_t* u_row,
+                               const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.hpp b/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.hpp
index 7226a51ddca1e1..512121b0f3dc76 100644
--- a/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.hpp
+++ b/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.hpp
@@ -181,27 +181,29 @@ void splitRow_32FC4(const float in[],
                           float out2[],
                           float out3[],
                             int length);
+}  // namespace avx
 
-void calculate_nv12_to_rgb(const  uchar **srcY,
-                           const  uchar *srcUV,
-                                  uchar **dstRGBx,
-                                    int width);
 
-void calculate_i420_to_rgb(const  uchar **srcY,
-                           const  uchar *srcU,
-                           const  uchar *srcV,
-                                  uchar **dstRGBx,
-                                    int width);
+template<typename isa_tag_t, typename T>
+void chanToPlaneRowImpl(isa_tag_t, const T* in, const int chan, const int chs, T* out, const int length);
 
-void copyRow_8U(const uint8_t in[],
-                uint8_t out[],
-                int length);
+extern template void chanToPlaneRowImpl(avx2_tag, const uint8_t* in, const int chan, const int chs, uint8_t* out, const int length);
+extern template void chanToPlaneRowImpl(avx2_tag, const float*   in, const int chan, const int chs, float  * out, const int length);
 
-void copyRow_32F(const float in[],
-                 float out[],
-                 int length);
+template<typename isa_tag_t>
+void nv12ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* uv_row,
+                             uint8_t** out_rows, const int buf_width);
 
-}  // namespace avx
+extern template void nv12ToRgbRowImpl(avx2_tag, const uint8_t** y_rows,
+                                      const uint8_t* uv_row, uint8_t** out_rows,
+                                      const int buf_width);
+
+template<typename isa_tag_t>
+void i420ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* u_row,
+                             const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+extern template void i420ToRgbRowImpl(avx2_tag, const uint8_t** y_rows, const uint8_t* u_row,
+                                      const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.cpp b/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.cpp
index 05925d45c195c3..78b74f532d62a6 100644
--- a/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.cpp
+++ b/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.cpp
@@ -101,21 +101,6 @@ void splitRow_32FC4(const float in[], float out0[], float out1[],
     splitRow_32FC4_Impl(in, out0, out1, out2, out3, length);
 }
 
-void calculate_nv12_to_rgb(const  uchar **srcY,
-                           const  uchar *srcUV,
-                                  uchar **dstRGBx,
-                                    int width) {
-    calculate_nv12_to_rgb_impl(srcY, srcUV, dstRGBx, width);
-}
-
-void calculate_i420_to_rgb(const  uchar **srcY,
-                           const  uchar *srcU,
-                           const  uchar *srcV,
-                                  uchar **dstRGBx,
-                                    int width) {
-    calculate_i420_to_rgb_impl(srcY, srcU, srcV, dstRGBx, width);
-}
-
 void calcRowArea_8U(uchar dst[], const uchar *src[], const Size& inSz,
                     const Size& outSz, Q0_16 yalpha, const MapperUnit8U &ymap,
                     int xmaxdf, const short xindex[], const Q0_16 xalpha[],
@@ -636,14 +621,6 @@ void calcRowLinear_8U(C4, std::array<std::array<uint8_t*, 4>, 4> &dst,
     calcRowLinear_8UC_Impl<chanNum>(dst, src0, src1, alpha, clone, mapsx, beta, tmp, inSz, outSz, lpi);
 }
 
-void copyRow_8U(const uint8_t in[], uint8_t out[], int length) {
-    copyRow_8U_impl(in, out, length);
-}
-
-void copyRow_32F(const float in[], float out[], int length) {
-    copyRow_32F_impl(in, out, length);
-}
-
 void calcRowLinear_32F(float *dst[],
                        const float *src0[],
                        const float *src1[],
@@ -657,6 +634,14 @@ void calcRowLinear_32F(float *dst[],
 }
 
 }  // namespace avx512
+
+template void chanToPlaneRowImpl(avx512_tag, const uint8_t* in, const int chan, const int chs, uint8_t* out, const int length);
+template void chanToPlaneRowImpl(avx512_tag, const float*   in, const int chan, const int chs, float*   out, const int length);
+
+template void nv12ToRgbRowImpl(avx512_tag, const uint8_t** y_rows, const uint8_t* uv_row, uint8_t** out_rows, const int buf_width);
+
+template void i420ToRgbRowImpl(avx512_tag, const uint8_t** y_rows, const uint8_t* u_row,
+                               const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.hpp b/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.hpp
index 2323d9345d4321..8d2778781a35b4 100644
--- a/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.hpp
+++ b/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.hpp
@@ -180,27 +180,26 @@ void splitRow_32FC4(const float in[],
                           float out2[],
                           float out3[],
                             int length);
+}  // namespace avx512
 
-void calculate_nv12_to_rgb(const  uchar **srcY,
-                           const  uchar *srcUV,
-                                  uchar **dstRGBx,
-                                    int width);
 
-void calculate_i420_to_rgb(const  uchar **srcY,
-                           const  uchar *srcU,
-                           const  uchar *srcV,
-                                  uchar **dstRGBx,
-                                    int width);
+template<typename isa_tag_t, typename T>
+void chanToPlaneRowImpl(isa_tag_t, const T* in, const int chan, const int chs, T* out, const int length);
 
-void copyRow_8U(const uint8_t in[],
-                uint8_t out[],
-                int length);
+extern template void chanToPlaneRowImpl(avx512_tag, const uint8_t* in, const int chan, const int chs, uint8_t* out, const int length);
+extern template void chanToPlaneRowImpl(avx512_tag, const float*   in, const int chan, const int chs, float*   out, const int length);
 
-void copyRow_32F(const float in[],
-                 float out[],
-                 int length);
+template<typename isa_tag_t>
+void nv12ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* uv_row, uint8_t** out_rows, const int buf_width);
 
-}  // namespace avx512
+extern template void nv12ToRgbRowImpl(avx512_tag, const uint8_t** y_rows, const uint8_t* uv_row, uint8_t** out_rows, const int buf_width);
+
+template<typename isa_tag_t>
+void i420ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* u_row,
+                             const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+extern template void i420ToRgbRowImpl(avx512_tag, const uint8_t** y_rows, const uint8_t* u_row,
+                                      const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.cpp b/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.cpp
index 86fb156931ede4..4823e33537141a 100644
--- a/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.cpp
+++ b/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.cpp
@@ -1365,33 +1365,13 @@ void splitRow_32FC4(const float in[],
     splitRow_32FC4_Impl(in, out0, out1, out2, out3, length);
 }
 
-void calculate_nv12_to_rgb(const  uchar **srcY,
-                           const  uchar *srcUV,
-                                  uchar **dstRGBx,
-                                    int width) {
-    calculate_nv12_to_rgb_impl(srcY, srcUV, dstRGBx, width);
-}
-
-void calculate_i420_to_rgb(const  uchar **srcY,
-                           const  uchar *srcU,
-                           const  uchar *srcV,
-                                  uchar **dstRGBx,
-                                    int width) {
-    calculate_i420_to_rgb_impl(srcY, srcU, srcV, dstRGBx, width);
-}
+template void chanToPlaneRowImpl(sse42_tag, const uint8_t* in, const int chan, const int chs, uint8_t* out, const int length);
+template void chanToPlaneRowImpl(sse42_tag, const float* in, const int chan, const int chs, float* out, const int length);
 
-void copyRow_8U(const uint8_t in[],
-                 uint8_t out[],
-                 int length) {
-    copyRow_8U_impl(in, out, length);
-}
-
-void copyRow_32F(const float in[],
-                 float out[],
-                 int length) {
-    copyRow_32F_impl(in, out, length);
-}
+template void nv12ToRgbRowImpl(sse42_tag, const uint8_t** y_rows, const uint8_t* uv_row, uint8_t** out_rows, const int buf_width);
 
+template void i420ToRgbRowImpl(sse42_tag, const uint8_t** y_rows, const uint8_t* u_row,
+                               const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.hpp b/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.hpp
index 4e1eb471cd5aa9..8726013357e0fa 100644
--- a/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.hpp
+++ b/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.hpp
@@ -180,25 +180,25 @@ void splitRow_32FC4(const float in[],
                           float out3[],
                             int length);
 
-void calculate_nv12_to_rgb(const  uchar **srcY,
-                           const  uchar *srcUV,
-                                  uchar **dstRGBx,
-                                    int width);
+template<typename isa_tag_t, typename T>
+void chanToPlaneRowImpl(isa_tag_t, const T* in, const int chan, const int chs,
+                        T* out, const int length);
 
-void calculate_i420_to_rgb(const  uchar **srcY,
-                           const  uchar *srcU,
-                           const  uchar *srcV,
-                                  uchar **dstRGBx,
-                                    int width);
+extern template void chanToPlaneRowImpl(sse42_tag, const uint8_t* in, const int chan,
+                                        const int chs, uint8_t* out, const int length);
+extern template void chanToPlaneRowImpl(sse42_tag, const float*   in, const int chan,
+                                        const int chs, float*   out, const int length);
+template<typename isa_tag_t>
+void nv12ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* uv_row, uint8_t** out_rows, const int buf_width);
 
-void copyRow_8U(const uint8_t in[],
-                uint8_t out[],
-                int length);
+extern template void nv12ToRgbRowImpl(sse42_tag, const uint8_t** y_rows, const uint8_t* uv_row, uint8_t** out_rows, const int buf_width);
 
-void copyRow_32F(const float in[],
-                 float out[],
-                 int length);
+template<typename isa_tag_t>
+void i420ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* u_row,
+                             const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
 
+extern template void i420ToRgbRowImpl(sse42_tag, const uint8_t** y_rows, const uint8_t* u_row,
+                                      const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels.cpp b/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels.cpp
index 0dd374bd3ffd4b..4cf4fffd74be9f 100644
--- a/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels.cpp
+++ b/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels.cpp
@@ -468,15 +468,86 @@ struct type_to_type {};
 template <typename typelist>
 struct type_dispatch_impl;
 
+//FIXME: add test for type_dispatch
 template <template<typename ...> class typelist, typename... type>
 struct type_dispatch_impl<typelist<type...>> {
     template <typename result_t, typename default_t, typename type_id_t, typename type_to_id_t, typename type_to_value_t>
     static result_t dispatch(type_id_t type_id, type_to_id_t&& type_to_id, type_to_value_t&& type_to_value, default_t default_value) {
         result_t res = default_value;
 
-        std::initializer_list<int> ({(type_id == type_to_id(type_to_type<type>{}) ? (res = type_to_value(type_to_type<type>{})), 0 : 0)...});
+        bool matched = false;
+        std::initializer_list<int> ({
+            !matched && (type_id == type_to_id(type_to_type<type>{})) ?
+                    (matched = true, res = type_to_value(type_to_type<type>{})), 0
+                    : 0
+            ...
+        });
         return res;
     }
+
+    template <typename result_t, typename default_t, typename pred_t, typename type_to_value_t>
+    static result_t dispatch(pred_t&& pred, type_to_value_t&& type_to_value, default_t default_value) {
+        result_t res = default_value;
+
+        bool matched = false;
+        std::initializer_list<int> ({
+            !matched && pred(type_to_type<type>{}) ?
+                    (matched = true, res = type_to_value(type_to_type<type>{})), 0
+                    : 0
+            ...
+        });
+        return res;
+    }
+};
+
+template<typename left_typelsist, typename right_typelsist>
+struct concat;
+
+template<typename left_typelsist, typename right_typelsist>
+using concat_t = typename concat<left_typelsist, right_typelsist>::type;
+
+template<template<typename ...> class left_list, typename ... left_types, template<typename ...> class right_list, typename ... right_types>
+struct concat<left_list<left_types...>, right_list<right_types...>>{
+    using type = left_list<left_types... , right_types...>;
+};
+
+template< class T, class U >
+using is_same_t = typename std::is_same<T, U>::type;
+
+template<bool C, class T, class E> struct if_c_impl;
+
+template<class T, class E> struct if_c_impl<true, T, E> {
+    using type = T;
+};
+
+template<class T, class E> struct if_c_impl<false, T, E> {
+    using type = E;
+};
+
+template<bool C, class T, class E>
+using if_c = typename if_c_impl<C, T, E>::type;
+
+template<class C, class T, class E>
+using if_ = typename if_c_impl<C::value != 0, T, E>::type;
+
+template<typename typelist, typename type>
+struct remove;
+
+template<typename typelist, typename type>
+using remove_t = typename remove<typelist, type>::type;
+
+
+template<template<typename ...> class list, typename head_t, typename ... types, typename t>
+struct remove<list<head_t, types...>, t> {
+    using type = concat_t<
+            if_<is_same_t<head_t, t>, list<>, list<head_t>>,
+            remove_t<list<types...>, t>
+            >;
+};
+
+template<template<typename ...> class list, typename t>
+struct remove<list<>, t> {
+    using type = list<>;
 };
 
 }  // namespace
@@ -490,6 +561,13 @@ result_t type_dispatch(type_id_t type_id, type_to_id_t&& type_to_id, type_to_val
                                                                      std::forward<default_t>(default_value));
 }
 
+template <typename typelist, typename default_t, typename pred_t, typename type_to_value_t,
+          typename result_t = decltype(std::declval<type_to_value_t>()(type_to_type<head_t<typelist>> {}))>
+result_t type_dispatch(pred_t&& pred, type_to_value_t&& type_to_value, default_t default_value = {}) {
+    return type_dispatch_impl<typelist>::template dispatch<result_t>(std::forward<pred_t>(pred),
+                                                                     std::forward<type_to_value_t>(type_to_value),
+                                                                     std::forward<default_t>(default_value));
+}
 namespace {
 
 struct cv_type_id {
@@ -668,81 +746,47 @@ GAPI_FLUID_KERNEL(FSplit4, Split4, false) {
 };
 
 //----------------------------------------------------------------------
-
-template<typename T>
-static void chanToPlaneRow(const uint8_t* in, int chan, int chs, uint8_t* out, int length) {
-// AVX512 implementation of wide universal intrinsics is slower than AVX2.
-// It is turned off until the cause isn't found out.
-#if 0
-    #ifdef HAVE_AVX512
-    if (with_cpu_x86_avx512f()) {
-        if (std::is_same<T, uint8_t>::value && chs == 1) {
-            avx512::copyRow_8U(in, out, length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 1) {
-            avx512::copyRow_32F(reinterpret_cast<const float*>(in),
-                                reinterpret_cast<float*>(out),
-                                length);
-            return;
-        }
-    }
-    #endif  // HAVE_AVX512
+using isas_set = typelist<
+#ifdef HAVE_AVX512
+        avx512_tag,
 #endif
+#ifdef HAVE_AVX2
+        avx2_tag,
+#endif
+#ifdef HAVE_SSE
+        sse42_tag,
+#endif
+#ifdef HAVE_NEON
+        neon_tag,
+#endif
+        //scalar "ISA" have to be the last one in the list,
+        //as the search for supported ISA is performed until first match
+        scalar_tag>;
+#ifdef HAVE_AVX512
+bool is_present(avx512_tag) { return with_cpu_x86_avx512f(); }
+#endif  // HAVE_AVX512
 
-    #ifdef HAVE_AVX2
-    if (with_cpu_x86_avx2()) {
-        if (std::is_same<T, uint8_t>::value && chs == 1) {
-            avx::copyRow_8U(in, out, length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 1) {
-            avx::copyRow_32F(reinterpret_cast<const float*>(in),
-                             reinterpret_cast<float*>(out),
-                             length);
-            return;
-        }
-    }
-    #endif  // HAVE_AVX2
-    #ifdef HAVE_SSE
-    if (with_cpu_x86_sse42()) {
-        if (std::is_same<T, uint8_t>::value && chs == 1) {
-            copyRow_8U(in, out, length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 1) {
-            copyRow_32F(reinterpret_cast<const float*>(in),
-                        reinterpret_cast<float*>(out),
-                        length);
-            return;
-        }
-    }
-    #endif  // HAVE_SSE
+#ifdef HAVE_AVX2
+bool is_present(avx2_tag)   { return with_cpu_x86_avx2();    }
+#endif  // HAVE_AVX2
 
-    #ifdef HAVE_NEON
-    if (std::is_same<T, uint8_t>::value && chs == 1) {
-        neon::copyRow_8U(in, out, length);
-        return;
-    }
+#ifdef HAVE_SSE
+bool is_present(sse42_tag)  { return with_cpu_x86_sse42();   }
+#endif  // HAVE_SSE
 
-    if (std::is_same<T, float>::value && chs == 1) {
-        neon::copyRow_32F(reinterpret_cast<const float*>(in),
-                          reinterpret_cast<float*>(out),
-                          length);
-        return;
-    }
-    #endif  // HAVE_NEON
+#ifdef HAVE_NEON
+bool is_present(neon_tag)   { return true; }
+#endif  // HAVE_NEON
 
-    const auto inT  = reinterpret_cast<const T*>(in);
-          auto outT = reinterpret_cast<      T*>(out);
+//scalar version of kernels is always available
+bool is_present(scalar_tag) { return true; }
 
-    for (int x = 0; x < length; x++) {
-        outT[x] = inT[x*chs + chan];
+struct is_isa_present {
+    template< typename isa_tag_t>
+    bool operator()(type_to_type<isa_tag_t>) {
+        return is_present(isa_tag_t{});
     }
-}
+};
 
 //    GAPI_OCV_KERNEL(OCVChanToPlane, ChanToPlane) {
 //        static void run(const cv::Mat &in, int chan, cv::Mat &out) {
@@ -774,15 +818,225 @@ static void chanToPlaneRow(const uint8_t* in, int chan, int chs, uint8_t* out, i
 //        }
 //    };
 
+namespace {
+
+using chan_to_plane_supported_types = typelist<uint8_t, float>;
+
+template<typename T>
+void chanToPlaneRowImpl(scalar_tag, const T* in, int chan, int chs, T* out, int length) {
+    for (int x = 0; x < length; x++) {
+        out[x] = in[x*chs + chan];
+    }
+}
+
+template<typename isa_tag_t>
+struct typed_chan_to_plane_row {
+    using p_f = void (*)(const uint8_t* in, int chan, int chs, uint8_t* out, int length);
+
+    template <typename type>
+    p_f operator()(type_to_type<type> ) {
+        return [](const uint8_t* in, int chan, int chs, uint8_t* out, int length){
+            const auto inT  = reinterpret_cast<const type*>(in);
+                  auto outT = reinterpret_cast<      type*>(out);
+
+            chanToPlaneRowImpl(isa_tag_t{}, inT, chan, chs, outT, length);
+        };
+    }
+};
+} //namespace
+
+namespace {
+
+using nv12_to_rgb_supported_types = typelist<uint8_t>;
+
+void nv12ToRgbRowImpl(scalar_tag, const uint8_t** y_rows, const uint8_t* uv_row,
+    uint8_t** out_rows, const int buf_width) {
+    for (int i = 0; i < buf_width; i += 2) {
+        uint8_t u = uv_row[i];
+        uint8_t v = uv_row[i + 1];
+        int ruv, guv, buv;
+        uvToRGBuv(u, v, ruv, guv, buv);
+
+        for (int y = 0; y < 2; y++) {
+            for (int x = 0; x < 2; x++) {
+                uint8_t vy = y_rows[y][i + x];
+                uint8_t r, g, b;
+                yRGBuvToRGB(vy, ruv, guv, buv, r, g, b);
+
+                out_rows[y][3 * (i + x)] = r;
+                out_rows[y][3 * (i + x) + 1] = g;
+                out_rows[y][3 * (i + x) + 2] = b;
+            }
+        }
+    }
+}
+
+template<typename isa_tag_t>
+struct typed_nv12_to_rgb_row {
+    using p_f = void (*)(const uint8_t** y_rows, const uint8_t* uv_row,
+        uint8_t** out_rows, const int buf_width);
+
+    template <typename type>
+    p_f operator()(type_to_type<type>) {
+        return [](const uint8_t** y_rows, const uint8_t* uv_row,
+            uint8_t** out_rows, const int buf_width) {
+                const auto inT1 = reinterpret_cast<const type**>(y_rows);
+                const auto inT2 = reinterpret_cast<const type*>(uv_row);
+                auto outT = reinterpret_cast<type**>(out_rows);
+
+                nv12ToRgbRowImpl(isa_tag_t{}, inT1, inT2, outT, buf_width);
+        };
+    }
+};
+}  // namespace
+
+namespace {
+
+using i420_to_rgb_supported_types = typelist<uint8_t>;
+
+static void i420ToRgbRowImpl(scalar_tag, const  uint8_t** y_rows,
+    const  uint8_t* u_row,
+    const  uint8_t* v_row,
+    uint8_t** out_rows,
+    const int buf_width) {
+    for (int i = 0; i < buf_width; i += 2) {
+        uchar u = u_row[i / 2];
+        uchar v = v_row[i / 2];
+        int ruv, guv, buv;
+        uvToRGBuv(u, v, ruv, guv, buv);
+
+        for (int y = 0; y < 2; y++) {
+            for (int x = 0; x < 2; x++) {
+                uchar vy = y_rows[y][i + x];
+                uchar r, g, b;
+                yRGBuvToRGB(vy, ruv, guv, buv, r, g, b);
+
+                out_rows[y][3 * (i + x)] = r;
+                out_rows[y][3 * (i + x) + 1] = g;
+                out_rows[y][3 * (i + x) + 2] = b;
+            }
+        }
+    }
+}
+
+template<typename isa_tag_t>
+struct typed_i420_to_rgb_row {
+    using p_f = void (*)(const uint8_t** y_rows, const uint8_t* u_row, const uint8_t* v_row,
+                         uint8_t** out_rows, const int buf_width);
+
+    template <typename type>
+    p_f operator()(type_to_type<type>) {
+        return [](const uint8_t** y_rows, const uint8_t* u_row, const uint8_t* v_row,
+                  uint8_t** out_rows, const int buf_width) {
+                const auto inT1 = reinterpret_cast<const type**>(y_rows);
+                const auto inT2 = reinterpret_cast<const type*>(u_row);
+                const auto inT3 = reinterpret_cast<const type*>(v_row);
+                auto outT = reinterpret_cast<type**>(out_rows);
+
+                i420ToRgbRowImpl(isa_tag_t{}, inT1, inT2, inT3, outT, buf_width);
+        };
+    }
+};
+}  // namespace
+
+template <typename isa_tag_t>
+struct choose_impl {
 GAPI_FLUID_KERNEL(FChanToPlane, ChanToPlane, false) {
     static const int Window = 1;
     static void run(const cv::gapi::fluid::View& in, int chan,
                     cv::gapi::fluid::Buffer& out) {
-        const auto rowFunc = (in.meta().depth == CV_8U) ? &chanToPlaneRow<uint8_t> : &chanToPlaneRow<float>;
+        GAPI_DbgAssert(is_cv_type_in_list<chan_to_plane_supported_types>(out.meta().depth));
+
+        const auto rowFunc = type_dispatch<chan_to_plane_supported_types>(out.meta().depth, cv_type_id{}, typed_chan_to_plane_row<isa_tag_t>{}, nullptr);
+
+        GAPI_DbgAssert(rowFunc);
+
         rowFunc(in.InLineB(0), chan, in.meta().chan, out.OutLineB(), in.length());
     }
 };
 
+GAPI_FLUID_KERNEL(FNV12toRGB, NV12toRGB, false) {
+    static const int Window = 1;
+    static const int LPI = 2;
+    static const auto Kind = cv::GFluidKernel::Kind::YUV420toRGB;
+
+    static void run(const cv::gapi::fluid::View & in_y,
+                    const cv::gapi::fluid::View & in_uv,
+                    cv::gapi::fluid::Buffer & out) {
+        GAPI_DbgAssert(is_cv_type_in_list<nv12_to_rgb_supported_types>(out.meta().depth));
+
+        const uchar* uv_row = in_uv.InLineB(0);
+        const uchar* y_rows[2] = { in_y.InLineB(0), in_y.InLineB(1) };
+        uchar* out_rows[2] = { out.OutLineB(0), out.OutLineB(1) };
+
+        int buf_width = out.length();
+
+        const auto rowFunc = type_dispatch<nv12_to_rgb_supported_types>(out.meta().depth, cv_type_id{}, typed_nv12_to_rgb_row<isa_tag_t>{}, nullptr);
+
+        GAPI_DbgAssert(rowFunc);
+
+        rowFunc(y_rows, uv_row, out_rows, buf_width);
+    }
+};
+
+GAPI_FLUID_KERNEL(FI420toRGB, I420toRGB, false) {
+    static const int Window = 1;
+    static const int LPI = 2;
+    static const auto Kind = cv::GFluidKernel::Kind::YUV420toRGB;
+
+    static void run(const cv::gapi::fluid::View & in_y,
+        const cv::gapi::fluid::View & in_u,
+        const cv::gapi::fluid::View & in_v,
+        cv::gapi::fluid::Buffer & out) {
+        GAPI_DbgAssert(is_cv_type_in_list<i420_to_rgb_supported_types>(out.meta().depth));
+
+        const uchar* u_row = in_u.InLineB(0);
+        const uchar* v_row = in_v.InLineB(0);
+        const uchar* y_rows[2] = { in_y.InLineB(0), in_y.InLineB(1) };
+        uchar* out_rows[2] = { out.OutLineB(0), out.OutLineB(1) };
+
+        int buf_width = out.length();
+        GAPI_DbgAssert(in_u.length() == in_v.length());
+
+        const auto rowFunc = type_dispatch<i420_to_rgb_supported_types>(out.meta().depth, cv_type_id{}, typed_i420_to_rgb_row<isa_tag_t>{}, nullptr);
+
+        GAPI_DbgAssert(rowFunc);
+
+        rowFunc(y_rows, u_row, v_row, out_rows, buf_width);
+    }
+};
+};
+
+namespace {
+struct ColorConversionISA {
+    cv::gapi::GKernelPackage& pckg;
+
+    ColorConversionISA(cv::gapi::GKernelPackage& _pckg) : pckg(_pckg) {}
+
+    template<typename isa_tag_t>
+    bool operator()(type_to_type<isa_tag_t>) {
+        pckg.include<typename choose_impl<isa_tag_t>::FI420toRGB>();
+        pckg.include<typename choose_impl<isa_tag_t>::FNV12toRGB>();
+        pckg.include<typename choose_impl<isa_tag_t>::FChanToPlane>();
+        //at the moment type_dispatch requires something to be returned by the lambda
+        return true;
+    }
+};
+}  //namespace
+
+cv::gapi::GKernelPackage FColorConversionChooseISA() {
+    // At the moment AVX512 implementation of wide universal intrinsics is slower than AVX2.
+    // So, disable it for now.
+    using isas = remove_t<isas_set, avx512_tag>;
+
+    cv::gapi::GKernelPackage pckg;
+    ColorConversionISA ctpISA{pckg};
+
+    type_dispatch<isas>(is_isa_present{}, ctpISA, false);
+
+    return pckg;
+}
+
 //----------------------------------------------------------------------
 
 G_TYPED_KERNEL(ScalePlane8u, <cv::GMat(cv::GMat, Size, int)>, "com.intel.ie.scale_plane_8u") {
@@ -2234,180 +2488,6 @@ GAPI_FLUID_KERNEL(FScalePlaneArea8u, ScalePlaneArea8u, true) {
     }
 };
 
-static const int ITUR_BT_601_CY = 1220542;
-static const int ITUR_BT_601_CUB = 2116026;
-static const int ITUR_BT_601_CUG = -409993;
-static const int ITUR_BT_601_CVG = -852492;
-static const int ITUR_BT_601_CVR = 1673527;
-static const int ITUR_BT_601_SHIFT = 20;
-
-static inline void uvToRGBuv(const uchar u, const uchar v, int& ruv, int& guv, int& buv) {
-    int uu, vv;
-    uu = static_cast<int>(u) - 128;
-    vv = static_cast<int>(v) - 128;
-
-    ruv = (1 << (ITUR_BT_601_SHIFT - 1)) + ITUR_BT_601_CVR * vv;
-    guv = (1 << (ITUR_BT_601_SHIFT - 1)) + ITUR_BT_601_CVG * vv + ITUR_BT_601_CUG * uu;
-    buv = (1 << (ITUR_BT_601_SHIFT - 1)) + ITUR_BT_601_CUB * uu;
-}
-
-static inline void yRGBuvToRGB(const uchar vy, const int ruv, const int guv, const int buv,
-                                uchar& r, uchar& g, uchar& b) {
-    int yy = static_cast<int>(vy);
-    int y = std::max(0, yy - 16) * ITUR_BT_601_CY;
-    r = saturate_cast<uchar>((y + ruv) >> ITUR_BT_601_SHIFT);
-    g = saturate_cast<uchar>((y + guv) >> ITUR_BT_601_SHIFT);
-    b = saturate_cast<uchar>((y + buv) >> ITUR_BT_601_SHIFT);
-}
-
-static void calculate_nv12_to_rgb_fallback(const  uchar **y_rows,
-                                           const  uchar *uv_row,
-                                                  uchar **out_rows,
-                                           int buf_width) {
-    for (int i = 0; i < buf_width; i += 2) {
-        uchar u = uv_row[i];
-        uchar v = uv_row[i + 1];
-        int ruv, guv, buv;
-        uvToRGBuv(u, v, ruv, guv, buv);
-
-        for (int y = 0; y < 2; y++) {
-            for (int x = 0; x < 2; x++) {
-                uchar vy = y_rows[y][i + x];
-                uchar r, g, b;
-                yRGBuvToRGB(vy, ruv, guv, buv, r, g, b);
-
-                out_rows[y][3*(i + x)]     = r;
-                out_rows[y][3*(i + x) + 1] = g;
-                out_rows[y][3*(i + x) + 2] = b;
-            }
-        }
-    }
-}
-
-static void calculate_i420_to_rgb_fallback(const  uchar **y_rows,
-                                           const  uchar *u_row,
-                                           const  uchar *v_row,
-                                                  uchar **out_rows,
-                                           int buf_width) {
-    for (int i = 0; i < buf_width; i += 2) {
-        uchar u = u_row[i / 2];
-        uchar v = v_row[i / 2];
-        int ruv, guv, buv;
-        uvToRGBuv(u, v, ruv, guv, buv);
-
-        for (int y = 0; y < 2; y++) {
-            for (int x = 0; x < 2; x++) {
-                uchar vy = y_rows[y][i + x];
-                uchar r, g, b;
-                yRGBuvToRGB(vy, ruv, guv, buv, r, g, b);
-
-                out_rows[y][3*(i + x)]     = r;
-                out_rows[y][3*(i + x) + 1] = g;
-                out_rows[y][3*(i + x) + 2] = b;
-            }
-        }
-    }
-}
-
-GAPI_FLUID_KERNEL(FNV12toRGB, NV12toRGB, false) {
-    static const int Window = 1;
-    static const int LPI    = 2;
-    static const auto Kind = cv::GFluidKernel::Kind::YUV420toRGB;
-
-    static void run(const cv::gapi::fluid::View &in_y,
-                    const cv::gapi::fluid::View &in_uv,
-                          cv::gapi::fluid::Buffer &out) {
-        const uchar* uv_row = in_uv.InLineB(0);
-        const uchar* y_rows[2] = {in_y. InLineB(0), in_y. InLineB(1)};
-        uchar* out_rows[2] = {out.OutLineB(0), out.OutLineB(1)};
-
-        int buf_width = out.length();
-
-// AVX512 implementation of wide universal intrinsics is slower than AVX2.
-// It is turned off until the cause isn't found out.
-    #if 0
-    #ifdef HAVE_AVX512
-        if (with_cpu_x86_avx512_core()) {
-            #define CV_AVX_512DQ 1
-            avx512::calculate_nv12_to_rgb(y_rows, uv_row, out_rows, buf_width);
-            return;
-        }
-    #endif  // HAVE_AVX512
-    #endif
-
-    #ifdef HAVE_AVX2
-        if (with_cpu_x86_avx2()) {
-            avx::calculate_nv12_to_rgb(y_rows, uv_row, out_rows, buf_width);
-            return;
-        }
-    #endif  // HAVE_AVX2
-    #ifdef HAVE_SSE
-        if (with_cpu_x86_sse42()) {
-            calculate_nv12_to_rgb(y_rows, uv_row, out_rows, buf_width);
-            return;
-        }
-    #endif  // HAVE_SSE
-
-    #ifdef HAVE_NEON
-        neon::calculate_nv12_to_rgb(y_rows, uv_row, out_rows, buf_width);
-        return;
-    #endif  // HAVE_NEON
-
-        calculate_nv12_to_rgb_fallback(y_rows, uv_row, out_rows, buf_width);
-    }
-};
-
-GAPI_FLUID_KERNEL(FI420toRGB, I420toRGB, false) {
-    static const int Window = 1;
-    static const int LPI    = 2;
-    static const auto Kind = cv::GFluidKernel::Kind::YUV420toRGB;
-
-    static void run(const cv::gapi::fluid::View &in_y,
-                    const cv::gapi::fluid::View &in_u,
-                    const cv::gapi::fluid::View &in_v,
-                          cv::gapi::fluid::Buffer &out) {
-        const uchar* u_row = in_u.InLineB(0);
-        const uchar* v_row = in_v.InLineB(0);
-        const uchar* y_rows[2] = {in_y. InLineB(0), in_y. InLineB(1)};
-        uchar* out_rows[2] = {out.OutLineB(0), out.OutLineB(1)};
-
-        int buf_width = out.length();
-        GAPI_DbgAssert(in_u.length() ==  in_v.length());
-
-        // AVX512 implementation of wide universal intrinsics is slower than AVX2.
-        // It is turned off until the cause isn't found out.
-        #if 0
-        #ifdef HAVE_AVX512
-            if (with_cpu_x86_avx512_core()) {
-               #define CV_AVX_512DQ 1
-               avx512::calculate_i420_to_rgb(y_rows, u_row, v_row, out_rows, buf_width);
-               return;
-            }
-        #endif  // HAVE_AVX512
-        #endif
-
-        #ifdef HAVE_AVX2
-            if (with_cpu_x86_avx2()) {
-               avx::calculate_i420_to_rgb(y_rows, u_row, v_row, out_rows, buf_width);
-               return;
-            }
-        #endif  // HAVE_AVX2
-        #ifdef HAVE_SSE
-            if (with_cpu_x86_sse42()) {
-               calculate_i420_to_rgb(y_rows, u_row, v_row, out_rows, buf_width);
-               return;
-            }
-        #endif  // HAVE_SSE
-
-        #ifdef HAVE_NEON
-            neon::calculate_i420_to_rgb(y_rows, u_row, v_row, out_rows, buf_width);
-            return;
-        #endif  // HAVE_NEON
-
-        calculate_i420_to_rgb_fallback(y_rows, u_row, v_row, out_rows, buf_width);
-    }
-};
-
 namespace {
 
 template <typename src_t, typename dst_t>
@@ -2520,9 +2600,10 @@ GAPI_FLUID_KERNEL(FDivC, GDivC, false) {
 using namespace kernels;
 
 cv::gapi::GKernelPackage preprocKernels() {
-    return cv::gapi::kernels
-        < FChanToPlane
-        , FScalePlanes
+    return combine(
+        FColorConversionChooseISA(),
+        cv::gapi::kernels
+        <FScalePlanes
         , FScalePlanes4
         , FScalePlane
         , FScalePlane32f
@@ -2537,12 +2618,10 @@ cv::gapi::GKernelPackage preprocKernels() {
         , FSplit2
         , FSplit3
         , FSplit4
-        , FNV12toRGB
-        , FI420toRGB
         , FConvertDepth
         , FSubC
         , FDivC
-        >();
+        >());
 }
 
 }  // namespace gapi
diff --git a/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_impl.hpp b/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_impl.hpp
index 452a29b2e2daec..3acf5a56b26884 100644
--- a/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_impl.hpp
+++ b/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_impl.hpp
@@ -34,6 +34,12 @@ namespace InferenceEngine {
 namespace gapi {
 namespace kernels {
 
+struct avx512_tag {};
+struct avx2_tag {};
+struct sse42_tag {};
+struct neon_tag {};
+struct scalar_tag {};
+
 template<typename DST, typename SRC> static inline DST saturate_cast(SRC x);
 template<> inline short saturate_cast(int x) { return (std::min)(SHRT_MAX, (std::max)(SHRT_MIN, x)); }
 template<> inline short saturate_cast(float x) { return saturate_cast<short>(static_cast<int>(std::rint(x))); }
@@ -116,6 +122,31 @@ static inline Q8_8 mulaw(Q0_16 a, Q8_8 w) { return static_cast<Q8_8>((a * w) >>
 static inline float mulas(float a, float s) { return a * s; }
 static inline float mulaw(float a, float w) { return a * w; }
 
+static const int ITUR_BT_601_CY = 1220542;
+static const int ITUR_BT_601_CUB = 2116026;
+static const int ITUR_BT_601_CUG = -409993;
+static const int ITUR_BT_601_CVG = -852492;
+static const int ITUR_BT_601_CVR = 1673527;
+static const int ITUR_BT_601_SHIFT = 20;
+
+static inline void uvToRGBuv(const uchar u, const uchar v, int& ruv, int& guv, int& buv) {
+    int uu, vv;
+    uu = static_cast<int>(u) - 128;
+    vv = static_cast<int>(v) - 128;
+
+    ruv = (1 << (ITUR_BT_601_SHIFT - 1)) + ITUR_BT_601_CVR * vv;
+    guv = (1 << (ITUR_BT_601_SHIFT - 1)) + ITUR_BT_601_CVG * vv + ITUR_BT_601_CUG * uu;
+    buv = (1 << (ITUR_BT_601_SHIFT - 1)) + ITUR_BT_601_CUB * uu;
+}
+
+static inline void yRGBuvToRGB(const uchar vy, const int ruv, const int guv, const int buv,
+    uchar& r, uchar& g, uchar& b) {
+    int yy = static_cast<int>(vy);
+    int y = std::max(0, yy - 16) * ITUR_BT_601_CY;
+    r = saturate_cast<uchar>((y + ruv) >> ITUR_BT_601_SHIFT);
+    g = saturate_cast<uchar>((y + guv) >> ITUR_BT_601_SHIFT);
+    b = saturate_cast<uchar>((y + buv) >> ITUR_BT_601_SHIFT);
+}
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_simd_impl.hpp b/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_simd_impl.hpp
index a59111b86b64fe..e559b83b87e405 100644
--- a/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_simd_impl.hpp
+++ b/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_simd_impl.hpp
@@ -374,23 +374,6 @@ CV_ALWAYS_INLINE void splitRow_32FC4_Impl(const float in[], float out0[], float
 
 //------------------------------------------------------------------------------
 
-static const int ITUR_BT_601_CY = 1220542;
-static const int ITUR_BT_601_CUB = 2116026;
-static const int ITUR_BT_601_CUG = -409993;
-static const int ITUR_BT_601_CVG = -852492;
-static const int ITUR_BT_601_CVR = 1673527;
-static const int ITUR_BT_601_SHIFT = 20;
-
-CV_ALWAYS_INLINE void uvToRGBuv(const uchar u, const uchar v, int& ruv, int& guv, int& buv) {
-    int uu, vv;
-    uu = static_cast<int>(u) - 128;
-    vv = static_cast<int>(v) - 128;
-
-    ruv = (1 << (ITUR_BT_601_SHIFT - 1)) + ITUR_BT_601_CVR * vv;
-    guv = (1 << (ITUR_BT_601_SHIFT - 1)) + ITUR_BT_601_CVG * vv + ITUR_BT_601_CUG * uu;
-    buv = (1 << (ITUR_BT_601_SHIFT - 1)) + ITUR_BT_601_CUB * uu;
-}
-
 CV_ALWAYS_INLINE void uvToRGBuv(const v_uint8& u, const v_uint8& v,
                                 v_int32 (&ruv)[4], v_int32 (&guv)[4],
                                 v_int32 (&buv)[4]) {
@@ -418,15 +401,6 @@ CV_ALWAYS_INLINE void uvToRGBuv(const v_uint8& u, const v_uint8& v,
     }
 }
 
-CV_ALWAYS_INLINE void yRGBuvToRGB(const uchar vy, const int ruv, const int guv,
-                                  const int buv, uchar& r, uchar& g, uchar& b) {
-    int yy = static_cast<int>(vy);
-    int y = std::max(0, yy - 16) * ITUR_BT_601_CY;
-    r = saturate_cast<uchar>((y + ruv) >> ITUR_BT_601_SHIFT);
-    g = saturate_cast<uchar>((y + guv) >> ITUR_BT_601_SHIFT);
-    b = saturate_cast<uchar>((y + buv) >> ITUR_BT_601_SHIFT);
-}
-
 CV_ALWAYS_INLINE void yRGBuvToRGB(const v_uint8& vy,
                                   const v_int32 (&ruv)[4],
                                   const v_int32 (&guv)[4],
@@ -464,16 +438,15 @@ CV_ALWAYS_INLINE void yRGBuvToRGB(const v_uint8& vy,
     bb = v_pack_u(b0, b1);
 }
 
-CV_ALWAYS_INLINE void calculate_nv12_to_rgb_impl(const  uchar **srcY,
-                                                 const  uchar *srcUV,
-                                                 uchar **dstRGBx,
-                                                 int width) {
+template<typename isa_tag_t>
+CV_ALWAYS_INLINE void nv12ToRgbRowImpl(isa_tag_t, const uchar** srcY, const uchar* srcUV,
+                                       uchar** dstRGBx, const int width) {
     int i = 0;
 
 #if MANUAL_SIMD
     constexpr int nlanes = v_uint8::nlanes;
 
-    for ( ; i <= width - 2*nlanes; i += 2*nlanes) {
+    for (; i <= width - 2 * nlanes; i += 2 * nlanes) {
         v_uint8 u, v;
         v_load_deinterleave(srcUV + i, u, v);
 
@@ -510,9 +483,7 @@ CV_ALWAYS_INLINE void calculate_nv12_to_rgb_impl(const  uchar **srcY,
         v_store_interleave(dstRGBx[1] + i * 3, b1_0, g1_0, r1_0);
         v_store_interleave(dstRGBx[1] + i * 3 + 3 * nlanes, b1_1, g1_1, r1_1);
     }
-
-    vx_cleanup();
-
+    //vx_cleanup();
 #endif
 
     for (; i < width; i += 2) {
@@ -527,25 +498,25 @@ CV_ALWAYS_INLINE void calculate_nv12_to_rgb_impl(const  uchar **srcY,
                 uchar r, g, b;
                 yRGBuvToRGB(vy, ruv, guv, buv, r, g, b);
 
-                dstRGBx[y][3*(i + x)]     = r;
-                dstRGBx[y][3*(i + x) + 1] = g;
-                dstRGBx[y][3*(i + x) + 2] = b;
+                dstRGBx[y][3 * (i + x)] = r;
+                dstRGBx[y][3 * (i + x) + 1] = g;
+                dstRGBx[y][3 * (i + x) + 2] = b;
             }
         }
     }
 }
 
-CV_ALWAYS_INLINE void calculate_i420_to_rgb_impl(const  uchar **srcY, const  uchar *srcU,
-                                                 const  uchar *srcV, uchar **dstRGBx,
-                                                 int width) {
+template<typename isa_tag_t>
+CV_ALWAYS_INLINE void i420ToRgbRowImpl(isa_tag_t, const uint8_t** srcY, const uint8_t* srcU,
+                                       const uint8_t* srcV, uint8_t** dstRGBx, const int width) {
     int i = 0;
 
 #if MANUAL_SIMD
     constexpr int nlanes = v_uint8::nlanes;
 
-    for ( ; i <= width - 2*nlanes; i += 2*nlanes) {
-        v_uint8 u = vx_load(srcU + i/2);
-        v_uint8 v = vx_load(srcV + i/2);
+    for (; i <= width - 2 * nlanes; i += 2 * nlanes) {
+        v_uint8 u = vx_load(srcU + i / 2);
+        v_uint8 v = vx_load(srcV + i / 2);
 
         v_uint8 vy[4];
         v_load_deinterleave(srcY[0] + i, vy[0], vy[1]);
@@ -580,14 +551,11 @@ CV_ALWAYS_INLINE void calculate_i420_to_rgb_impl(const  uchar **srcY, const  uch
         v_store_interleave(dstRGBx[1] + i * 3, b1_0, g1_0, r1_0);
         v_store_interleave(dstRGBx[1] + i * 3 + 3 * nlanes, b1_1, g1_1, r1_1);
     }
-
-    vx_cleanup();
-
-    #endif
-
+    //vx_cleanup();
+#endif
     for (; i < width; i += 2) {
-        uchar u = srcU[i/2];
-        uchar v = srcV[i/2];
+        uchar u = srcU[i / 2];
+        uchar v = srcV[i / 2];
         int ruv, guv, buv;
         uvToRGBuv(u, v, ruv, guv, buv);
 
@@ -597,9 +565,9 @@ CV_ALWAYS_INLINE void calculate_i420_to_rgb_impl(const  uchar **srcY, const  uch
                 uchar r, g, b;
                 yRGBuvToRGB(vy, ruv, guv, buv, r, g, b);
 
-                dstRGBx[y][3*(i + x)]     = r;
-                dstRGBx[y][3*(i + x) + 1] = g;
-                dstRGBx[y][3*(i + x) + 2] = b;
+                dstRGBx[y][3 * (i + x)] = r;
+                dstRGBx[y][3 * (i + x) + 1] = g;
+                dstRGBx[y][3 * (i + x) + 2] = b;
             }
         }
     }
@@ -737,48 +705,24 @@ CV_ALWAYS_INLINE void calcRowArea_impl(T dst[], const T *src[], const Size& inSz
 
 //------------------------------------------------------------------------------
 
-#if MANUAL_SIMD
 template <typename VecT, typename T>
-CV_ALWAYS_INLINE void copyRow_impl(const T in[], T out[], int l) {
-    VecT r;
-    r = vx_load(&in[l]);
-    vx_store(&out[l], r);
-}
-#endif
-
-CV_ALWAYS_INLINE void copyRow_8U_impl(const uint8_t in[], uint8_t out[], int length) {
+CV_ALWAYS_INLINE void copyRow_Impl(const T in[], T out[], int length) {
     int l = 0;
 
 #if MANUAL_SIMD
-    constexpr int nlanes = v_uint8::nlanes;
+    const int nlanes = VecT::nlanes;
 
-    for (; l <= length - nlanes; l += nlanes) {
-        copyRow_impl<v_uint8>(in, out, l);
-    }
-
-    if (l < length && length >= nlanes) {
-        copyRow_impl<v_uint8>(in, out, length - nlanes);
-        l = length;
-    }
-#endif
-
-    for (; l < length; l++) {
-        out[l] = in[l];
-    }
-}
-
-CV_ALWAYS_INLINE void copyRow_32F_impl(const float in[], float out[], int length) {
-    int l = 0;
-
-#if MANUAL_SIMD
-    constexpr int nlanes = v_float32::nlanes;
+    auto copy_row = [](const T in[], T out[], int l) {
+        VecT r = vx_load(&in[l]);
+        vx_store(&out[l], r);
+    };
 
     for (; l <= length - nlanes; l += nlanes) {
-        copyRow_impl<v_float32>(in, out, l);
+        copy_row(in, out, l);
     }
 
     if (l < length && length >= nlanes) {
-        copyRow_impl<v_float32>(in, out, length - nlanes);
+        copy_row(in, out, length - nlanes);
         l = length;
     }
 #endif
@@ -916,6 +860,26 @@ CV_ALWAYS_INLINE void calcRowLinear_32FC1(float *dst[],
     }
 }
 
+template<typename isa_tag_t, typename scalar_t>
+struct vector_type_of;
+
+template<typename isa_tag_t, typename scalar_t>
+using vector_type_of_t = typename vector_type_of<isa_tag_t, scalar_t>::type;
+
+template<typename isa_tag_t> struct vector_type_of<isa_tag_t, uint8_t> { using type = v_uint8;  };
+template<typename isa_tag_t> struct vector_type_of<isa_tag_t, float>   { using type = v_float32;};
+
+template<typename isa_tag_t, typename T>
+CV_ALWAYS_INLINE void chanToPlaneRowImpl(isa_tag_t, const T* in, const int chan, const int chs, T* out, const int length) {
+    if (chs == 1) {
+        copyRow_Impl<vector_type_of_t<isa_tag_t, T>, T>(in, out, length);
+        return;
+    }
+
+    for (int x = 0; x < length; x++) {
+        out[x] = in[x*chs + chan];
+    }
+}
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine

From dbe1c1ab49b3e59ec3c41ca715a9e7f01714bf77 Mon Sep 17 00:00:00 2001
From: Chenhu Wang <chenhu.wang@intel.com>
Date: Wed, 16 Jun 2021 15:32:32 +0800
Subject: [PATCH 104/178] [CPU] Change rounding type in load/store emitters
 (#6134)

---
 .../src/mkldnn_plugin/emitters/jit_load_store_emitters.cpp      | 2 --
 1 file changed, 2 deletions(-)

diff --git a/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.cpp b/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.cpp
index 57689d6302d70a..83bc04c530db07 100644
--- a/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.cpp
+++ b/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.cpp
@@ -106,7 +106,6 @@ void jit_load_emitter::emit_isa(const Xbyak::Reg64 &reg_src, int offset_byte, In
                 break;
             case Precision::I32:
                 if ((src_prc == Precision::FP32) || (src_prc == Precision::BF16)) {
-                    h->uni_vroundps(Vmm(out_vec_idx), Vmm(out_vec_idx), 3); // rounding to zero
                     h->uni_vcvtps2dq(Vmm(out_vec_idx), Vmm(out_vec_idx));
                 }
                 break;
@@ -552,7 +551,6 @@ template <mkldnn::impl::cpu::x64::cpu_isa_t isa>
             switch (src_prc) {
                 case Precision::FP32:
                     if ((dst_prc != Precision::FP32) && (dst_prc != Precision::BF16)) {
-                        h->uni_vroundps(Vmm(in_vec_idx), Vmm(in_vec_idx), 3); // rounding to zero
                         h->uni_vcvtps2dq(Vmm(in_vec_idx), Vmm(in_vec_idx));
                     }
                     break;

From 72cb75ffc7df4bf4fc59791b81d9627df34e4e83 Mon Sep 17 00:00:00 2001
From: Edward Shogulin <edward.shogulin@intel.com>
Date: Wed, 16 Jun 2021 11:14:29 +0300
Subject: [PATCH 105/178] [LPT] Reshape folding extending: cherry-pick to
 master (#6151)

* [LPT] Reshape folding extending

* [LPT] tests addition

* typo quick fix
---
 .../include/low_precision/network_helper.hpp  |  6 ++
 .../group_convolution_transformation.cpp      | 75 +++++++++++++++++++
 .../group_convolution_transformation.cpp      | 28 +++++++
 .../group_convolution_transformation.cpp      | 28 +++++++
 .../group_convolution_transformation.hpp      |  5 ++
 .../group_convolution_transformation.cpp      | 16 ++++
 .../group_convolution_function.hpp            |  3 +
 .../src/group_convolution_function.cpp        | 14 +++-
 8 files changed, 174 insertions(+), 1 deletion(-)

diff --git a/inference-engine/src/low_precision_transformations/include/low_precision/network_helper.hpp b/inference-engine/src/low_precision_transformations/include/low_precision/network_helper.hpp
index 8cf52a13fe20ca..bf45cdeae94133 100644
--- a/inference-engine/src/low_precision_transformations/include/low_precision/network_helper.hpp
+++ b/inference-engine/src/low_precision_transformations/include/low_precision/network_helper.hpp
@@ -270,6 +270,12 @@ template <typename T, typename... Args>
 std::shared_ptr<Node> fold_reshape(Args&&... args) {
     std::shared_ptr<Node> node = std::make_shared<T>(std::forward<Args>(args)...);
     if (node->get_output_size() == 1) {
+        // issue #57985: remove fold_reshape & reuse nGraph implementation
+        const auto values = as_type_ptr<opset1::Constant>(node->input_value(1).get_node_shared_ptr())->template cast_vector<int64_t>();
+        if (std::any_of(values.begin(), values.end(), [](const int64_t value) { return (value == 0) || (value == -1); })) {
+            return fold<opset1::Reshape>(std::forward<Args>(args)...);
+        }
+
         OutputVector folded;
         if (is_type<opset1::Constant>(node->input_value(0).get_node_shared_ptr()) &&
             is_type<opset1::Constant>(node->input_value(1).get_node_shared_ptr())) {
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/group_convolution_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/group_convolution_transformation.cpp
index d90999bb8ccad4..22844037c98d75 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/group_convolution_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/group_convolution_transformation.cpp
@@ -53,6 +53,7 @@ class GroupConvolutionTestValues {
     ngraph::Shape inputShape;
     ngraph::Shape outputShape;
     size_t group;
+    int groupCalculationDimention;
     Actual actual;
     Expected expected;
 };
@@ -67,6 +68,7 @@ class GroupConvolutionTransformation : public LayerTransformation, public testin
             testValues.inputShape,
             testValues.outputShape,
             testValues.group,
+            testValues.groupCalculationDimention,
             testValues.actual.dequantization,
             testValues.actual.weights,
             testValues.actual.fakeQuantizeOnWeights,
@@ -84,6 +86,7 @@ class GroupConvolutionTransformation : public LayerTransformation, public testin
             testValues.inputShape,
             testValues.outputShape,
             testValues.group,
+            testValues.groupCalculationDimention,
             testValues.expected.dequantizationBefore,
             testValues.expected.weights,
             testValues.expected.fakeQuantizeOnWeights,
@@ -101,6 +104,7 @@ class GroupConvolutionTransformation : public LayerTransformation, public testin
             testValues.inputShape << "_" <<
             testValues.outputShape << "_" <<
             testValues.group << "_" <<
+            testValues.groupCalculationDimention << "_" <<
             testValues.actual.precisionBeforeDequantization << "_" <<
             testValues.actual.dequantization << "_" << "_weights_" <<
             testValues.actual.weights->get_element_type() << "_" << "{ " <<
@@ -123,6 +127,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 24, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
@@ -142,12 +147,68 @@ const std::vector<GroupConvolutionTestValues> testValues = {
             {{}, {}, {{ 0.0002f }, ngraph::element::f32, { 1, 24, 1, 1 }}} // 0.0002 = 0.02 (on data) * 0.01 (on weights)
         }
     },
+
+    // group convolution, tensor quantization, with zero point
+    {
+        LayerTransformation::createParamsU8I8().setSupportAsymmetricQuantization(true),
+        { 1, 6, 224, 224 },
+        { 1, 24, 218, 218 },
+        3ul,
+        0,
+        // ActualValues
+        {
+            ngraph::element::u8,
+            {{ngraph::element::f32}, { 128.f }, { 0.02f }},
+            op::Constant::create(ngraph::element::f32, ngraph::Shape{}, std::vector<float>{ 2.f }),
+            { 255ul, Shape({ 1, 1, 1, 1 }), { 0.f }, { 254.f }, { -1.27f }, { 1.27f } },
+            {}
+        },
+        // ExpectedValues
+        {
+            ngraph::element::u8,
+            {{}, { { 128.f }, ngraph::element::f32, { 1, 6, 1, 1 }, false }, {}},
+            op::Constant::create(ngraph::element::i8, ngraph::Shape{}, std::vector<float>{ -125.f }),
+            {},
+            {},
+            ngraph::element::f32,
+            {{}, {}, {{ 0.0002f }, ngraph::element::f32, { 1, 24, 1, 1 }}} // 0.0002 = 0.02 (on data) * 0.01 (on weights)
+        }
+    },
+
+    // group convolution, tensor quantization, with zero point
+    {
+        LayerTransformation::createParamsU8I8().setSupportAsymmetricQuantization(true),
+        { 1, 6, 224, 224 },
+        { 1, 24, 218, 218 },
+        3ul,
+        1,
+        // ActualValues
+        {
+            ngraph::element::u8,
+            {{ngraph::element::f32}, { 128.f }, { 0.02f }},
+            op::Constant::create(ngraph::element::f32, ngraph::Shape{}, std::vector<float>{ 2.f }),
+            { 255ul, Shape({ 1, 1, 1, 1 }), { 0.f }, { 254.f }, { -1.27f }, { 1.27f } },
+            {}
+        },
+        // ExpectedValues
+        {
+            ngraph::element::u8,
+            {{}, { { 128.f }, ngraph::element::f32, { 1, 6, 1, 1 }, false }, {}},
+            op::Constant::create(ngraph::element::i8, ngraph::Shape{}, std::vector<float>{ -125.f }),
+            {},
+            {},
+            ngraph::element::f32,
+            {{}, {}, {{ 0.0002f }, ngraph::element::f32, { 1, 24, 1, 1 }}} // 0.0002 = 0.02 (on data) * 0.01 (on weights)
+        }
+    },
+
     // group convolution, tensor quantization, with zero point
     {
         LayerTransformation::createParamsU8I8().setSupportAsymmetricQuantization(false),
         { 1, 6, 224, 224 },
         { 1, 24, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
@@ -173,6 +234,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 24, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::f32,
@@ -198,6 +260,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 24, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
@@ -241,6 +304,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 24, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
@@ -274,6 +338,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 24, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::f32,
@@ -299,6 +364,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 24, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
@@ -324,6 +390,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 6, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
@@ -349,6 +416,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 6, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::f32,
@@ -374,6 +442,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 6, 218, 218 },
         6ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
@@ -414,6 +483,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 6, 218, 218 },
         6ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
@@ -447,6 +517,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 6, 218, 218 },
         6ul,
+        -1,
         // ActualValues
         {
             ngraph::element::f32,
@@ -472,6 +543,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 6, 218, 218 },
         6ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
@@ -497,6 +569,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 6, 218, 218 },
         6ul,
+        -1,
         // ActualValues
         {
             ngraph::element::f32,
@@ -523,6 +596,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 24, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
@@ -567,6 +641,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
         { 1, 6, 224, 224 },
         { 1, 24, 218, 218 },
         3ul,
+        -1,
         // ActualValues
         {
             ngraph::element::u8,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
index 36e81ef438d165..739897ee27350e 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
@@ -26,14 +26,40 @@ const std::vector<LayerTestsDefinitions::GroupConvolutionTransformationParam> pa
         ngraph::Shape{ 1, 6, 24, 24 },
         ngraph::Shape{ 1, 24, 18, 18 },
         3ul,
+        -1,
         { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 25.5f }, { 0.f }, { 25.5f } },
         { 255ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 254.f }, { -127.f }, { 127.f } },
+        "Convolution",
+        "U8"
+    },
+    // group convolution, tensor quantization
+    {
+        ngraph::Shape{ 1, 6, 24, 24 },
+        ngraph::Shape{ 1, 24, 18, 18 },
+        3ul,
+        0,
+        { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 25.5f }, { 0.f }, { 25.5f } },
+        { 255ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 254.f }, { -127.f }, { 127.f } },
+        "Convolution",
+        "U8"
+    },
+    // group convolution, tensor quantization
+    {
+        ngraph::Shape{ 1, 6, 24, 24 },
+        ngraph::Shape{ 1, 24, 18, 18 },
+        3ul,
+        1,
+        { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 25.5f }, { 0.f }, { 25.5f } },
+        { 255ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 254.f }, { -127.f }, { 127.f } },
+        "Convolution",
+        "U8"
     },
     // group convolution, per-channel quantization
     {
         ngraph::Shape{ 1, 6, 24, 24 },
         ngraph::Shape{ 1, 24, 18, 18 },
         3ul,
+        -1,
         {
             256ul,
             ngraph::Shape { 6, 1, 1, 1 },
@@ -49,6 +75,7 @@ const std::vector<LayerTestsDefinitions::GroupConvolutionTransformationParam> pa
         ngraph::Shape{ 1, 6, 24, 24 },
         ngraph::Shape{ 1, 6, 18, 18 },
         6ul,
+        -1,
         { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 25.5f }, { 0.f }, { 25.5f } },
         { 255ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 254.f }, { -127.f }, { 127.f } },
     },
@@ -57,6 +84,7 @@ const std::vector<LayerTestsDefinitions::GroupConvolutionTransformationParam> pa
         ngraph::Shape{ 1, 6, 24, 24 },
         ngraph::Shape{ 1, 6, 18, 18 },
         6ul,
+        -1,
         {
             256ul,
             ngraph::Shape { 6, 1, 1, 1 },
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
index 8c346ba667e75d..624d933188b7fa 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
@@ -26,14 +26,40 @@ const std::vector<LayerTestsDefinitions::GroupConvolutionTransformationParam> pa
         ngraph::Shape{ 1, 6, 24, 24 },
         ngraph::Shape{ 1, 24, 18, 18 },
         3ul,
+        -1,
         { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 25.5f }, { 0.f }, { 25.5f } },
         { 255ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 254.f }, { -127.f }, { 127.f } },
+        "Convolution",
+        "U8"
+    },
+    // group convolution, tensor quantization
+    {
+        ngraph::Shape{ 1, 6, 24, 24 },
+        ngraph::Shape{ 1, 24, 18, 18 },
+        3ul,
+        0,
+        { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 25.5f }, { 0.f }, { 25.5f } },
+        { 255ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 254.f }, { -127.f }, { 127.f } },
+        "Convolution",
+        "U8"
+    },
+    // group convolution, tensor quantization
+    {
+        ngraph::Shape{ 1, 6, 24, 24 },
+        ngraph::Shape{ 1, 24, 18, 18 },
+        3ul,
+        1,
+        { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 25.5f }, { 0.f }, { 25.5f } },
+        { 255ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 254.f }, { -127.f }, { 127.f } },
+        "Convolution",
+        "U8"
     },
     // group convolution, per-channel quantization
     {
         ngraph::Shape{ 1, 6, 24, 24 },
         ngraph::Shape{ 1, 24, 18, 18 },
         3ul,
+        -1,
         {
             256ul,
             ngraph::Shape { 6, 1, 1, 1 },
@@ -49,6 +75,7 @@ const std::vector<LayerTestsDefinitions::GroupConvolutionTransformationParam> pa
         ngraph::Shape{ 1, 6, 24, 24 },
         ngraph::Shape{ 1, 6, 18, 18 },
         6ul,
+        -1,
         { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 25.5f }, { 0.f }, { 25.5f } },
         { 255ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 254.f }, { -127.f }, { 127.f } },
     },
@@ -57,6 +84,7 @@ const std::vector<LayerTestsDefinitions::GroupConvolutionTransformationParam> pa
         ngraph::Shape{ 1, 6, 24, 24 },
         ngraph::Shape{ 1, 6, 18, 18 },
         6ul,
+        -1,
         {
             256ul,
             ngraph::Shape { 6, 1, 1, 1 },
diff --git a/inference-engine/tests/functional/plugin/shared/include/low_precision_transformations/group_convolution_transformation.hpp b/inference-engine/tests/functional/plugin/shared/include/low_precision_transformations/group_convolution_transformation.hpp
index 1910df5e46da3f..b7d92417182a2a 100644
--- a/inference-engine/tests/functional/plugin/shared/include/low_precision_transformations/group_convolution_transformation.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/low_precision_transformations/group_convolution_transformation.hpp
@@ -18,8 +18,11 @@ class GroupConvolutionTransformationParam {
     ngraph::Shape inputShape;
     ngraph::Shape outputShape;
     size_t group;
+    int groupCalculationDimention;
     ngraph::builder::subgraph::FakeQuantizeOnData fakeQuantizeOnData;
     ngraph::builder::subgraph::FakeQuantizeOnWeights fakeQuantizeOnWeights;
+    std::string layerName;
+    std::string expectedKernelType;
 };
 
 typedef std::tuple<
@@ -38,6 +41,8 @@ class GroupConvolutionTransformation :
 protected:
     void SetUp() override;
 
+    void Run() override;
+
 private:
     void validate();
 };
diff --git a/inference-engine/tests/functional/plugin/shared/src/low_precision_transformations/group_convolution_transformation.cpp b/inference-engine/tests/functional/plugin/shared/src/low_precision_transformations/group_convolution_transformation.cpp
index f56c9743defdf2..c9baa32932975c 100644
--- a/inference-engine/tests/functional/plugin/shared/src/low_precision_transformations/group_convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/shared/src/low_precision_transformations/group_convolution_transformation.cpp
@@ -33,6 +33,7 @@ std::string GroupConvolutionTransformation::getTestCaseName(testing::TestParamIn
         param.inputShape << "_" <<
         param.outputShape << "_" <<
         param.group << "_" <<
+        param.groupCalculationDimention << "_" <<
         param.fakeQuantizeOnData << "_" <<
         param.fakeQuantizeOnWeights;
     return result.str();
@@ -51,12 +52,27 @@ void GroupConvolutionTransformation::SetUp() {
         param.inputShape,
         param.outputShape,
         param.group,
+        param.groupCalculationDimention,
         param.fakeQuantizeOnData,
         param.fakeQuantizeOnWeights);
 
     validate();
 }
 
+void GroupConvolutionTransformation::Run() {
+    LayerTestsCommon::Run();
+
+    const auto param = std::get<3>(GetParam());
+    if (!param.layerName.empty()) {
+        const auto actualPrecision = getRuntimePrecisionByType(param.layerName);
+        auto expectedPrecision = param.expectedKernelType;
+        if (expectedPrecision == "FP32" && std::get<0>(GetParam()) == ngraph::element::f16) {
+            expectedPrecision = "FP16";
+        }
+        EXPECT_EQ(actualPrecision, expectedPrecision);
+    }
+}
+
 void GroupConvolutionTransformation::validate() {
     ngraph::element::Type netPrecision;
     ngraph::pass::low_precision::LayerTransformation::Params params;
diff --git a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/include/lpt_ngraph_functions/group_convolution_function.hpp b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/include/lpt_ngraph_functions/group_convolution_function.hpp
index 2273de2602c4fc..1b94400f7bc099 100644
--- a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/include/lpt_ngraph_functions/group_convolution_function.hpp
+++ b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/include/lpt_ngraph_functions/group_convolution_function.hpp
@@ -22,6 +22,7 @@ class GroupConvolutionFunction {
         const ngraph::Shape& inputShape,
         const ngraph::Shape& outputShape,
         const size_t groupCount,
+        const int groupCalculationDimention,
         const ngraph::builder::subgraph::DequantizationOperations& dequantizationBefore,
         std::shared_ptr<ngraph::opset1::Constant> weightsConst,
         const ngraph::builder::subgraph::FakeQuantizeOnWeights fakeQuantizeOnWeights);
@@ -31,6 +32,7 @@ class GroupConvolutionFunction {
         const ngraph::Shape& inputShape,
         const ngraph::Shape& outputShape,
         const size_t groupCount,
+        const int groupCalculationDimention,
         const FakeQuantizeOnData& fakeQuantizeOnData,
         const FakeQuantizeOnWeights& fakeQuantizeOnWeights);
 
@@ -39,6 +41,7 @@ class GroupConvolutionFunction {
         const ngraph::Shape& inputShape,
         const ngraph::Shape& outputShape,
         const size_t groupCount,
+        const int calculatedDimention,
         const ngraph::builder::subgraph::DequantizationOperations& dequantizationBefore,
         std::shared_ptr<ngraph::opset1::Constant> weightsConst,
         const ngraph::builder::subgraph::FakeQuantizeOnWeights& fakeQuantizeOnWeights,
diff --git a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/group_convolution_function.cpp b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/group_convolution_function.cpp
index 4d3b6153f08c56..205d57747ecebd 100644
--- a/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/group_convolution_function.cpp
+++ b/inference-engine/tests/ngraph_helpers/lpt_ngraph_functions/src/group_convolution_function.cpp
@@ -28,6 +28,7 @@ std::shared_ptr<Node> createWeightsOriginal(
     const size_t inputChannelsCount,
     const size_t outputChannelsCount,
     const size_t groupCount,
+    const int calculatedDimention,
     const size_t kernelSize,
     const std::vector<float>& weightsValues,
     const FakeQuantizeOnWeights& fakeQuantizeOnWeights,
@@ -70,7 +71,12 @@ std::shared_ptr<Node> createWeightsOriginal(
             ngraph::opset1::Constant::create(
                 element::i64,
                 Shape{ 5 },
-                std::vector<size_t>({ groupCount, outputChannelsCount / groupCount, inputChannelsPerGroup, kernelSize, kernelSize })),
+                std::vector<int64_t> {
+                    calculatedDimention == 0 ? -1 : static_cast<int64_t>(groupCount),
+                    calculatedDimention == 1 ? -1 : static_cast<int64_t>(outputChannelsCount / groupCount),
+                    static_cast<int64_t>(inputChannelsPerGroup),
+                    static_cast<int64_t>(kernelSize),
+                    static_cast<int64_t>(kernelSize) }),
             true);
     }
 
@@ -82,6 +88,7 @@ std::shared_ptr<ngraph::Function> GroupConvolutionFunction::getOriginal(
     const ngraph::Shape& inputShape,
     const ngraph::Shape& outputShape,
     const size_t groupCount,
+    const int groupCalculationDimention,
     const ngraph::builder::subgraph::DequantizationOperations& dequantizationBefore,
     std::shared_ptr<ngraph::opset1::Constant> weightsConst,
     const ngraph::builder::subgraph::FakeQuantizeOnWeights fakeQuantizeOnWeights) {
@@ -103,6 +110,7 @@ std::shared_ptr<ngraph::Function> GroupConvolutionFunction::getOriginal(
         inputChannelsCount,
         outputChannelsCount,
         groupCount,
+        groupCalculationDimention,
         kernelSize,
         weightsConst->cast_vector<float>(),
         fakeQuantizeOnWeights,
@@ -126,6 +134,7 @@ std::shared_ptr<ngraph::Function> GroupConvolutionFunction::getOriginal(
     const ngraph::Shape& inputShape,
     const ngraph::Shape& outputShape,
     const size_t groupCount,
+    const int groupCalculationDimention,
     const FakeQuantizeOnData& fakeQuantizeOnData,
     const FakeQuantizeOnWeights& fakeQuantizeOnWeights) {
     const auto input = std::make_shared<ngraph::opset1::Parameter>(precision, ngraph::Shape(inputShape));
@@ -156,6 +165,7 @@ std::shared_ptr<ngraph::Function> GroupConvolutionFunction::getOriginal(
         inputChannelsCount,
         outputChannelsCount,
         groupCount,
+        groupCalculationDimention,
         kernelSize,
         weightsValues,
         fakeQuantizeOnWeights,
@@ -178,6 +188,7 @@ std::shared_ptr<ngraph::Function> GroupConvolutionFunction::get(
     const ngraph::Shape& inputShape,
     const ngraph::Shape& outputShape,
     const size_t groupCount,
+    const int calculatedDimention,
     const ngraph::builder::subgraph::DequantizationOperations& dequantizationBefore,
     std::shared_ptr<ngraph::opset1::Constant> weightsConst,
     const ngraph::builder::subgraph::FakeQuantizeOnWeights& fakeQuantizeOnWeights,
@@ -216,6 +227,7 @@ std::shared_ptr<ngraph::Function> GroupConvolutionFunction::get(
             inputChannelsCount,
             outputChannelsCount,
             groupCount,
+            calculatedDimention,
             kernelSize,
             weightsConst->cast_vector<float>(),
             fakeQuantizeOnWeights,

From b0c508d4ff8698d80dd0187465cdaba93234e59d Mon Sep 17 00:00:00 2001
From: Alexander Shchepetov <alexander.shchepetov@intel.com>
Date: Wed, 16 Jun 2021 12:50:16 +0300
Subject: [PATCH 106/178] Add layer tests (#5789)

Co-authored-by: alexander.shchepetov <ashchepe@nnlvdp-cflr51.inn.intel.com>
---
 tests/layer_tests/.gitignore                  |   9 +
 tests/layer_tests/README.md                   |  32 +
 tests/layer_tests/common/constants.py         |  25 +
 tests/layer_tests/common/layer_test_class.py  | 155 ++++
 tests/layer_tests/common/layer_utils.py       |  58 ++
 .../common/onnx_layer_test_class.py           |  63 ++
 .../layer_tests/common/tf_layer_test_class.py |  61 ++
 .../layer_tests/common/utils/common_utils.py  |  84 +++
 .../common/utils/multiprocessing_utils.py     |  99 +++
 tests/layer_tests/common/utils/parsers.py     |  28 +
 tests/layer_tests/common/utils/tf_utils.py    | 109 +++
 tests/layer_tests/conftest.py                 |  77 ++
 tests/layer_tests/onnx_tests/conftest.py      |  12 +
 tests/layer_tests/onnx_tests/test_abs.py      | 178 +++++
 .../onnx_tests/test_add_sub_mul_div.py        | 337 +++++++++
 tests/layer_tests/onnx_tests/test_and.py      | 268 +++++++
 tests/layer_tests/onnx_tests/test_argmax.py   | 153 ++++
 tests/layer_tests/onnx_tests/test_bn.py       | 133 ++++
 tests/layer_tests/onnx_tests/test_ceil.py     | 194 +++++
 tests/layer_tests/onnx_tests/test_clip.py     | 169 +++++
 tests/layer_tests/onnx_tests/test_concat.py   | 225 ++++++
 tests/layer_tests/onnx_tests/test_conv.py     | 413 +++++++++++
 .../onnx_tests/test_conv_transpose.py         | 215 ++++++
 tests/layer_tests/onnx_tests/test_cumsum.py   | 267 +++++++
 .../onnx_tests/test_dequantize_linear.py      | 205 ++++++
 tests/layer_tests/onnx_tests/test_dropout.py  | 169 +++++
 tests/layer_tests/onnx_tests/test_elu.py      | 185 +++++
 .../onnx_tests/test_embedding_bag.py          | 131 ++++
 tests/layer_tests/onnx_tests/test_flatten.py  | 302 ++++++++
 tests/layer_tests/onnx_tests/test_floor.py    | 179 +++++
 tests/layer_tests/onnx_tests/test_gather.py   | 261 +++++++
 tests/layer_tests/onnx_tests/test_gemm.py     | 308 ++++++++
 .../onnx_tests/test_hard_sigmoid.py           | 235 ++++++
 tests/layer_tests/onnx_tests/test_identity.py | 183 +++++
 .../onnx_tests/test_image_scaler.py           | 161 +++++
 .../onnx_tests/test_instance_normalization.py | 111 +++
 .../layer_tests/onnx_tests/test_leaky_relu.py | 204 ++++++
 tests/layer_tests/onnx_tests/test_log.py      | 193 +++++
 .../layer_tests/onnx_tests/test_logsoftmax.py | 240 +++++++
 tests/layer_tests/onnx_tests/test_lrn.py      | 133 ++++
 tests/layer_tests/onnx_tests/test_lstm.py     | 160 +++++
 tests/layer_tests/onnx_tests/test_matmul.py   | 190 +++++
 .../test_mean_variance_normalization.py       |  73 ++
 tests/layer_tests/onnx_tests/test_neg.py      |  91 +++
 tests/layer_tests/onnx_tests/test_non_zero.py | 194 +++++
 tests/layer_tests/onnx_tests/test_not.py      | 190 +++++
 tests/layer_tests/onnx_tests/test_or.py       | 277 +++++++
 tests/layer_tests/onnx_tests/test_pad.py      | 212 ++++++
 tests/layer_tests/onnx_tests/test_pooling.py  | 432 +++++++++++
 tests/layer_tests/onnx_tests/test_prelu.py    | 161 +++++
 .../layer_tests/onnx_tests/test_reciprocal.py | 171 +++++
 tests/layer_tests/onnx_tests/test_reduce.py   | 179 +++++
 .../layer_tests/onnx_tests/test_reduce_lp.py  | 247 +++++++
 tests/layer_tests/onnx_tests/test_relu.py     | 183 +++++
 tests/layer_tests/onnx_tests/test_reshape.py  | 283 ++++++++
 tests/layer_tests/onnx_tests/test_resize.py   | 676 ++++++++++++++++++
 .../layer_tests/onnx_tests/test_roi_align.py  | 109 +++
 tests/layer_tests/onnx_tests/test_scale.py    | 144 ++++
 tests/layer_tests/onnx_tests/test_scatter.py  | 125 ++++
 tests/layer_tests/onnx_tests/test_sigmoid.py  | 196 +++++
 tests/layer_tests/onnx_tests/test_sign.py     | 176 +++++
 tests/layer_tests/onnx_tests/test_slice.py    | 399 +++++++++++
 tests/layer_tests/onnx_tests/test_softmax.py  | 167 +++++
 tests/layer_tests/onnx_tests/test_softplus.py | 182 +++++
 .../onnx_tests/test_split_concat.py           | 307 ++++++++
 tests/layer_tests/onnx_tests/test_sqrt.py     | 187 +++++
 tests/layer_tests/onnx_tests/test_squeeze.py  | 210 ++++++
 tests/layer_tests/onnx_tests/test_sum.py      | 322 +++++++++
 tests/layer_tests/onnx_tests/test_topk.py     | 167 +++++
 .../layer_tests/onnx_tests/test_transpose.py  | 178 +++++
 .../onnx_tests/test_trigonometry.py           | 298 ++++++++
 .../layer_tests/onnx_tests/test_unsqueeze.py  | 210 ++++++
 tests/layer_tests/onnx_tests/test_upsample.py | 188 +++++
 tests/layer_tests/onnx_tests/test_where.py    |  97 +++
 tests/layer_tests/onnx_tests/test_xor.py      | 268 +++++++
 tests/layer_tests/requirements.txt            |   2 +
 .../layer_tests/tensorflow_tests/conftest.py  |  12 +
 .../tensorflow_tests/permutation_utils.py     |  22 +
 .../tensorflow_tests/test_tf_Add.py           | 281 ++++++++
 .../tensorflow_tests/test_tf_BatchToSpace.py  |  80 +++
 .../tensorflow_tests/test_tf_BiasAdd.py       | 194 +++++
 .../tensorflow_tests/test_tf_Bucketize.py     |  83 +++
 .../tensorflow_tests/test_tf_Ceil.py          |  78 ++
 .../tensorflow_tests/test_tf_Concat.py        | 114 +++
 .../tensorflow_tests/test_tf_ELU.py           |  88 +++
 .../tensorflow_tests/test_tf_Eltwise.py       |  79 ++
 .../tensorflow_tests/test_tf_FakeQuantize.py  | 125 ++++
 .../tensorflow_tests/test_tf_Floor.py         |  78 ++
 .../tensorflow_tests/test_tf_Identity.py      |  85 +++
 .../tensorflow_tests/test_tf_Log.py           |  79 ++
 .../tensorflow_tests/test_tf_LogSoftmax.py    | 138 ++++
 .../tensorflow_tests/test_tf_Mul.py           | 279 ++++++++
 .../tensorflow_tests/test_tf_NormalizeL2.py   | 207 ++++++
 .../tensorflow_tests/test_tf_OneHot.py        | 159 ++++
 .../tensorflow_tests/test_tf_Pooling.py       | 230 ++++++
 .../tensorflow_tests/test_tf_ReLU.py          |  85 +++
 .../tensorflow_tests/test_tf_ReLU6.py         |  85 +++
 .../test_tf_Resample_pattern_new.py           |  81 +++
 .../tensorflow_tests/test_tf_Rsqrt.py         |  73 ++
 .../tensorflow_tests/test_tf_Select.py        | 107 +++
 .../tensorflow_tests/test_tf_Softplus.py      |  90 +++
 .../tensorflow_tests/test_tf_SpaceToBatch.py  |  80 +++
 .../tensorflow_tests/test_tf_Squeeze.py       | 124 ++++
 .../tensorflow_tests/test_tf_Sub.py           | 294 ++++++++
 .../tensorflow_tests/test_tf_Swish.py         |  88 +++
 .../tensorflow_tests/test_tf_TopK.py          | 143 ++++
 106 files changed, 17876 insertions(+)
 create mode 100644 tests/layer_tests/.gitignore
 create mode 100644 tests/layer_tests/README.md
 create mode 100644 tests/layer_tests/common/constants.py
 create mode 100644 tests/layer_tests/common/layer_test_class.py
 create mode 100644 tests/layer_tests/common/layer_utils.py
 create mode 100644 tests/layer_tests/common/onnx_layer_test_class.py
 create mode 100644 tests/layer_tests/common/tf_layer_test_class.py
 create mode 100644 tests/layer_tests/common/utils/common_utils.py
 create mode 100644 tests/layer_tests/common/utils/multiprocessing_utils.py
 create mode 100644 tests/layer_tests/common/utils/parsers.py
 create mode 100644 tests/layer_tests/common/utils/tf_utils.py
 create mode 100644 tests/layer_tests/conftest.py
 create mode 100644 tests/layer_tests/onnx_tests/conftest.py
 create mode 100644 tests/layer_tests/onnx_tests/test_abs.py
 create mode 100644 tests/layer_tests/onnx_tests/test_add_sub_mul_div.py
 create mode 100644 tests/layer_tests/onnx_tests/test_and.py
 create mode 100644 tests/layer_tests/onnx_tests/test_argmax.py
 create mode 100644 tests/layer_tests/onnx_tests/test_bn.py
 create mode 100644 tests/layer_tests/onnx_tests/test_ceil.py
 create mode 100644 tests/layer_tests/onnx_tests/test_clip.py
 create mode 100644 tests/layer_tests/onnx_tests/test_concat.py
 create mode 100644 tests/layer_tests/onnx_tests/test_conv.py
 create mode 100644 tests/layer_tests/onnx_tests/test_conv_transpose.py
 create mode 100644 tests/layer_tests/onnx_tests/test_cumsum.py
 create mode 100644 tests/layer_tests/onnx_tests/test_dequantize_linear.py
 create mode 100644 tests/layer_tests/onnx_tests/test_dropout.py
 create mode 100644 tests/layer_tests/onnx_tests/test_elu.py
 create mode 100644 tests/layer_tests/onnx_tests/test_embedding_bag.py
 create mode 100644 tests/layer_tests/onnx_tests/test_flatten.py
 create mode 100644 tests/layer_tests/onnx_tests/test_floor.py
 create mode 100644 tests/layer_tests/onnx_tests/test_gather.py
 create mode 100644 tests/layer_tests/onnx_tests/test_gemm.py
 create mode 100644 tests/layer_tests/onnx_tests/test_hard_sigmoid.py
 create mode 100644 tests/layer_tests/onnx_tests/test_identity.py
 create mode 100644 tests/layer_tests/onnx_tests/test_image_scaler.py
 create mode 100644 tests/layer_tests/onnx_tests/test_instance_normalization.py
 create mode 100644 tests/layer_tests/onnx_tests/test_leaky_relu.py
 create mode 100644 tests/layer_tests/onnx_tests/test_log.py
 create mode 100644 tests/layer_tests/onnx_tests/test_logsoftmax.py
 create mode 100644 tests/layer_tests/onnx_tests/test_lrn.py
 create mode 100644 tests/layer_tests/onnx_tests/test_lstm.py
 create mode 100644 tests/layer_tests/onnx_tests/test_matmul.py
 create mode 100644 tests/layer_tests/onnx_tests/test_mean_variance_normalization.py
 create mode 100644 tests/layer_tests/onnx_tests/test_neg.py
 create mode 100644 tests/layer_tests/onnx_tests/test_non_zero.py
 create mode 100644 tests/layer_tests/onnx_tests/test_not.py
 create mode 100644 tests/layer_tests/onnx_tests/test_or.py
 create mode 100644 tests/layer_tests/onnx_tests/test_pad.py
 create mode 100644 tests/layer_tests/onnx_tests/test_pooling.py
 create mode 100644 tests/layer_tests/onnx_tests/test_prelu.py
 create mode 100644 tests/layer_tests/onnx_tests/test_reciprocal.py
 create mode 100644 tests/layer_tests/onnx_tests/test_reduce.py
 create mode 100644 tests/layer_tests/onnx_tests/test_reduce_lp.py
 create mode 100644 tests/layer_tests/onnx_tests/test_relu.py
 create mode 100644 tests/layer_tests/onnx_tests/test_reshape.py
 create mode 100644 tests/layer_tests/onnx_tests/test_resize.py
 create mode 100644 tests/layer_tests/onnx_tests/test_roi_align.py
 create mode 100644 tests/layer_tests/onnx_tests/test_scale.py
 create mode 100644 tests/layer_tests/onnx_tests/test_scatter.py
 create mode 100644 tests/layer_tests/onnx_tests/test_sigmoid.py
 create mode 100644 tests/layer_tests/onnx_tests/test_sign.py
 create mode 100644 tests/layer_tests/onnx_tests/test_slice.py
 create mode 100644 tests/layer_tests/onnx_tests/test_softmax.py
 create mode 100644 tests/layer_tests/onnx_tests/test_softplus.py
 create mode 100644 tests/layer_tests/onnx_tests/test_split_concat.py
 create mode 100644 tests/layer_tests/onnx_tests/test_sqrt.py
 create mode 100644 tests/layer_tests/onnx_tests/test_squeeze.py
 create mode 100644 tests/layer_tests/onnx_tests/test_sum.py
 create mode 100644 tests/layer_tests/onnx_tests/test_topk.py
 create mode 100644 tests/layer_tests/onnx_tests/test_transpose.py
 create mode 100644 tests/layer_tests/onnx_tests/test_trigonometry.py
 create mode 100644 tests/layer_tests/onnx_tests/test_unsqueeze.py
 create mode 100644 tests/layer_tests/onnx_tests/test_upsample.py
 create mode 100644 tests/layer_tests/onnx_tests/test_where.py
 create mode 100644 tests/layer_tests/onnx_tests/test_xor.py
 create mode 100644 tests/layer_tests/requirements.txt
 create mode 100644 tests/layer_tests/tensorflow_tests/conftest.py
 create mode 100644 tests/layer_tests/tensorflow_tests/permutation_utils.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Add.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_BatchToSpace.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_BiasAdd.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Bucketize.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Ceil.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Concat.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_ELU.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Eltwise.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_FakeQuantize.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Floor.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Identity.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Log.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_LogSoftmax.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Mul.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_NormalizeL2.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_OneHot.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Pooling.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_ReLU.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_ReLU6.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Resample_pattern_new.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Rsqrt.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Select.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Softplus.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_SpaceToBatch.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Squeeze.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Sub.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Swish.py
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_TopK.py

diff --git a/tests/layer_tests/.gitignore b/tests/layer_tests/.gitignore
new file mode 100644
index 00000000000000..f7e78b45447612
--- /dev/null
+++ b/tests/layer_tests/.gitignore
@@ -0,0 +1,9 @@
+#IDE FILES
+.pytest_cache
+
+#TESTS TEMP DIRECTORY
+ir/
+out/
+tf_models/
+onnx_models/
+
diff --git a/tests/layer_tests/README.md b/tests/layer_tests/README.md
new file mode 100644
index 00000000000000..5787a19bbd21fb
--- /dev/null
+++ b/tests/layer_tests/README.md
@@ -0,0 +1,32 @@
+# Layer tests
+
+This folder layer tests framework code and test files.
+
+## Getting Started
+
+#### Pre-requisites
+
+* OpenVINO should be configured as usual.
+
+#### Setup
+
+* Install requirements:
+    ```bash
+    pip3 install -r requirements.txt
+    ```
+* Set up environment variables for layer tests:
+    ```bash
+    export MO_ROOT=PATH_TO_MO
+    ```
+    ```bash
+    export PYTHONPATH="path_to_openvino"/tests/layer_tests/:$PYTHONPATH
+    ```
+    ```bash
+    export IE_APP_PATH="path_to_IE"
+    ```
+* Add IE dependencies in LD_LIBRARY_PATH.
+
+## Run tests
+```bash
+py.test
+```
diff --git a/tests/layer_tests/common/constants.py b/tests/layer_tests/common/constants.py
new file mode 100644
index 00000000000000..f1e97d43817983
--- /dev/null
+++ b/tests/layer_tests/common/constants.py
@@ -0,0 +1,25 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+
+
+if 'MO_ROOT' in os.environ:
+    mo_bin = os.environ['MO_ROOT']
+    if not os.path.exists(mo_bin):
+        raise EnvironmentError(
+            "Environment variable MO_ROOT points to non existing path {}".format(mo_bin))
+else:
+    raise EnvironmentError("MO_ROOT variable is not set")
+
+if os.environ.get('OUTPUT_DIR') is not None:
+    out_path = os.environ['OUTPUT_DIR']
+else:
+    script_path = os.path.dirname(os.path.realpath(__file__))
+    out_path = os.path.join(script_path, 'out')
+    if not os.path.exists(out_path):
+        os.makedirs(out_path)
+
+# supported_devices : CPU, GPU, MYRIAD, FPGA
+test_device = os.environ.get('TEST_DEVICE', 'CPU;GPU').split(';')
+test_precision = os.environ.get('TEST_PRECISION', 'FP32;FP16').split(';')
diff --git a/tests/layer_tests/common/layer_test_class.py b/tests/layer_tests/common/layer_test_class.py
new file mode 100644
index 00000000000000..28d08d23090d34
--- /dev/null
+++ b/tests/layer_tests/common/layer_test_class.py
@@ -0,0 +1,155 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import itertools
+import os
+import warnings
+from pathlib import Path
+
+import numpy as np
+from common.constants import test_device, test_precision
+from common.layer_utils import IEInfer
+from mo.utils.ir_engine.ir_engine import IREngine
+
+from common.utils.common_utils import generate_ir
+from common.utils.parsers import mapping_parser
+
+
+class CommonLayerTest:
+    input_model_key = "input_model"
+
+    def produce_model_path(self, framework_model, save_path):
+        pass
+
+    def get_framework_results(self, inputs_dict, model_path):
+        pass
+
+    def _test(self, framework_model, ref_net, ie_device, precision, ir_version, temp_dir,
+              enabled_transforms='', disabled_transforms='', **kwargs):
+        """
+        :param enabled_transforms/disabled_transforms: string with idxs of transforms that should be enabled/disabled.
+                                                       Example: "transform_1,transform_2"
+        """
+        model_path = self.produce_model_path(framework_model=framework_model, save_path=temp_dir)
+
+        # TODO Pass environment variables via subprocess environment
+        os.environ['MO_ENABLED_TRANSFORMS'] = enabled_transforms
+        os.environ['MO_DISABLED_TRANSFORMS'] = disabled_transforms
+
+        mo_params = {self.input_model_key: model_path,
+                     "output_dir": temp_dir,
+                     "data_type": precision, "model_name": 'model'
+                     }
+
+        if 'input_shapes' in kwargs and len(kwargs['input_shapes']):
+            input_shapes_str = []
+            for ishape in kwargs['input_shapes']:
+                input_shapes_str.append('[' + ','.join([str(i) for i in ishape]) + ']')
+            mo_params.update(dict(input_shape=','.join(input_shapes_str)))
+
+        if 'input_names' in kwargs and len(kwargs['input_names']):
+            mo_params.update(dict(input=','.join(kwargs['input_names'])))
+
+        exit_code, stderr = generate_ir(**mo_params)
+
+        del os.environ['MO_ENABLED_TRANSFORMS']
+        del os.environ['MO_DISABLED_TRANSFORMS']
+        assert not exit_code, ("IR generation failed with {} exit code: {}".format(exit_code, stderr))
+
+        path_to_xml = Path(temp_dir, 'model.xml')
+        path_to_bin = Path(temp_dir, 'model.bin')
+
+        ir = IREngine(path_to_xml, path_to_bin, precision=precision)
+        if ref_net is not None:
+            (flag, resp) = ir.compare(ref_net)
+            assert flag, '\n'.join(resp)
+
+        # Prepare feed dict
+        if 'kwargs_to_prepare_input' in kwargs and kwargs['kwargs_to_prepare_input']:
+            inputs_dict = self._prepare_input(ir.get_inputs(), kwargs['kwargs_to_prepare_input'])
+        else:
+            inputs_dict = self._prepare_input(ir.get_inputs())
+
+        # IE infer:
+        ie_engine = IEInfer(model=path_to_xml,
+                            weights=path_to_bin,
+                            device=ie_device)
+        infer_res = ie_engine.infer(input_data=inputs_dict)
+
+        if hasattr(self, 'skip_framework') and self.skip_framework:
+            warnings.warn('Framework is skipped')
+            return
+
+        # Framework infer:
+        fw_res = self.get_framework_results(inputs_dict=inputs_dict, model_path=model_path)
+
+        if len(fw_res) == len(infer_res) == 1:
+            # match output layers directly
+            mapping_dict = {next(iter(fw_res)): next(iter(infer_res))}
+        else:
+            # Load mapping file
+            mapping_dict = mapping_parser(path_to_xml.with_suffix('.mapping'))
+
+        if 'custom_eps' in kwargs and kwargs['custom_eps'] is not None:
+            custom_eps = kwargs['custom_eps']
+        else:
+            custom_eps = 1e-4
+
+        # Compare Ie results with Framework results
+        fw_eps = custom_eps if precision == 'FP32' else 5e-2
+        assert self.compare_ie_results_with_framework(infer_res=infer_res, framework_res=fw_res,
+                                                      mapping_dict=mapping_dict, framework_eps=fw_eps), \
+            "Comparing with Framework failed: ie_res={}; framework_res={}.".format(infer_res, fw_res)
+
+    # Feed dict for each input is filled with random number.
+    # It is possible to redefine this function and generate your own input
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randint(-255, 255, inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def compare_ie_results_with_framework(self, infer_res, framework_res, mapping_dict, framework_eps):
+        is_ok = True
+        from common.utils.common_utils import allclose
+        for framework_out_name in framework_res:
+            if framework_out_name not in mapping_dict:
+                raise RuntimeError("Output {} not found in mapping file!".format(framework_out_name))
+
+            ie_out_name = mapping_dict[framework_out_name]
+
+            if not allclose(infer_res[ie_out_name], framework_res[framework_out_name], atol=framework_eps,
+                            rtol=framework_eps):
+                is_ok = False
+                print("Max diff is {}".format(
+                    np.array(abs(infer_res[ie_out_name] - framework_res[framework_out_name])).max()))
+            else:
+                print("Accuracy validation successful!\n")
+                print("absolute eps: {}, relative eps: {}".format(framework_eps, framework_eps))
+        return is_ok
+
+
+def get_params(ie_device=None, precision=None):
+    """
+    :param ie_device: list of devices
+    :param precision: list of precisions
+    """
+
+    ie_device_params = ie_device if ie_device else test_device
+    precision_params = precision if precision else test_precision
+
+    test_args = []
+    for element in itertools.product(ie_device_params, precision_params):
+        if element[0] == 'CPU' and element[1] == 'FP16':
+            continue
+        test_args.append(element)
+    return test_args
+
+
+def check_ir_version(left, right, ir_version):
+    try:
+        _ir_version = int(ir_version)
+    except ValueError:
+        raise RuntimeError("Wrong ir version type: {}, must be an integer".format(ir_version))
+    left_bound = _ir_version - 1 if left is None else left
+    right_bound = _ir_version + 1 if right is None else right
+    return left_bound <= _ir_version < right_bound
diff --git a/tests/layer_tests/common/layer_utils.py b/tests/layer_tests/common/layer_utils.py
new file mode 100644
index 00000000000000..a933ac719aa2b3
--- /dev/null
+++ b/tests/layer_tests/common/layer_utils.py
@@ -0,0 +1,58 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import subprocess
+import sys
+
+from common.utils.multiprocessing_utils import multiprocessing_run
+
+
+def shell(cmd, env=None, cwd=None):
+    if sys.platform.startswith('linux') or sys.platform == 'darwin':
+        cmd = ['/bin/bash', '-c', "".join(cmd)]
+    else:
+        cmd = "".join(cmd)
+    p = subprocess.Popen(cmd, cwd=cwd, env=env, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    (stdout, stderr) = p.communicate()
+    return p.returncode, stdout, stderr
+
+
+class BaseInfer:
+    def __init__(self, name):
+        self.name = name
+        self.res = None
+
+    def fw_infer(self, input_data):
+        raise RuntimeError("This is base class, please implement infer function for the specific framework")
+
+    def infer(self, input_data):
+        self.res = multiprocessing_run(self.fw_infer, [input_data], self.name, timeout=60)
+        return self.res
+
+
+class IEInfer(BaseInfer):
+    def __init__(self, model, weights, device):
+        super().__init__('Inference Engine')
+        self.device = device
+        self.model = model
+        self.weights = weights
+
+    def fw_infer(self, input_data):
+        from openvino.inference_engine import IECore, get_version as ie_get_version
+
+        print("Inference Engine version: {}".format(ie_get_version()))
+        print("Creating IE Core Engine...")
+        ie = IECore()
+        print("Reading network files")
+        net = ie.read_network(self.model, self.weights)
+        print("Loading network")
+        exec_net = ie.load_network(net, self.device)
+        print("Starting inference")
+        result = exec_net.infer(input_data)
+
+        if "exec_net" in locals():
+            del exec_net
+        if "ie" in locals():
+            del ie
+
+        return result
diff --git a/tests/layer_tests/common/onnx_layer_test_class.py b/tests/layer_tests/common/onnx_layer_test_class.py
new file mode 100644
index 00000000000000..ff97718f1c0417
--- /dev/null
+++ b/tests/layer_tests/common/onnx_layer_test_class.py
@@ -0,0 +1,63 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+
+from common.layer_test_class import CommonLayerTest
+from common.layer_utils import BaseInfer
+
+
+def save_to_onnx(onnx_model, path_to_saved_onnx_model):
+    import onnx
+    path = os.path.join(path_to_saved_onnx_model, 'model.onnx')
+    onnx.save(onnx_model, path)
+    assert os.path.isfile(path), "model.onnx haven't been saved here: {}".format(path_to_saved_onnx_model)
+    return path
+
+
+class Caffe2OnnxLayerTest(CommonLayerTest):
+    def produce_model_path(self, framework_model, save_path):
+        return save_to_onnx(framework_model, save_path)
+
+    def get_framework_results(self, inputs_dict, model_path):
+        # Evaluate model via Caffe2 and IE
+        # Load the ONNX model
+        import onnx
+        model = onnx.load(model_path)
+        # Run the ONNX model with Caffe2
+        import caffe2.python.onnx.backend
+        caffe2_res = caffe2.python.onnx.backend.run_model(model, inputs_dict)
+        res = dict()
+        for field in caffe2_res._fields:
+            res[field] = caffe2_res[field]
+        return res
+
+
+class OnnxRuntimeInfer(BaseInfer):
+    def __init__(self, net):
+        super().__init__('OnnxRuntime')
+        self.net = net
+
+    def fw_infer(self, input_data):
+        import onnxruntime as rt
+
+        sess = rt.InferenceSession(self.net)
+        out = sess.run(None, input_data)
+        result = dict()
+        for i, output in enumerate(sess.get_outputs()):
+            result[output.name] = out[i]
+
+        if "sess" in locals():
+            del sess
+
+        return result
+
+
+class OnnxRuntimeLayerTest(CommonLayerTest):
+    def produce_model_path(self, framework_model, save_path):
+        return save_to_onnx(framework_model, save_path)
+
+    def get_framework_results(self, inputs_dict, model_path):
+        ort = OnnxRuntimeInfer(net=model_path)
+        res = ort.infer(input_data=inputs_dict)
+        return res
diff --git a/tests/layer_tests/common/tf_layer_test_class.py b/tests/layer_tests/common/tf_layer_test_class.py
new file mode 100644
index 00000000000000..ab19c29e9701f8
--- /dev/null
+++ b/tests/layer_tests/common/tf_layer_test_class.py
@@ -0,0 +1,61 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+
+from common.layer_test_class import CommonLayerTest
+
+from common.utils.tf_utils import summarize_graph
+
+
+def save_to_pb(tf_model, path_to_saved_tf_model):
+    import tensorflow as tf
+    tf.io.write_graph(tf_model, path_to_saved_tf_model, 'model.pb', False)
+    assert os.path.isfile(os.path.join(path_to_saved_tf_model, 'model.pb')), "model.pb haven't been saved " \
+                                                                             "here: {}".format(path_to_saved_tf_model)
+    return os.path.join(path_to_saved_tf_model, 'model.pb')
+
+
+class CommonTFLayerTest(CommonLayerTest):
+    def produce_model_path(self, framework_model, save_path):
+        return save_to_pb(framework_model, save_path)
+
+    def get_framework_results(self, inputs_dict, model_path):
+        # Evaluate model via Tensorflow and IE
+        # Load the Tensorflow model
+        import tensorflow as tf
+        from tensorflow.python.platform import gfile
+
+        graph_summary = summarize_graph(model_path=model_path)
+        outputs_list = graph_summary["outputs"]
+
+        tf.compat.v1.reset_default_graph()
+
+        with tf.compat.v1.Session() as sess:
+            with gfile.FastGFile(model_path, 'rb') as f:
+                graph_def = tf.compat.v1.GraphDef()
+                graph_def.ParseFromString(f.read())
+                sess.graph.as_default()
+                tf.compat.v1.import_graph_def(graph_def, name='')
+
+                input = dict()
+                for key in inputs_dict.keys():
+                    data = inputs_dict.get(key)
+                    if len(data.shape) == 4:        # reshaping for 4D tensors
+                        input[key+':0'] = data.transpose(0, 2, 3, 1)
+                    elif len(data.shape) == 5:      # reshaping for 5D tensors
+                        input[key+':0'] = data.transpose(0, 2, 3, 4, 1)
+                    else:
+                        input[key+':0'] = data
+                tf_res = sess.run([out + ":0" for out in outputs_list], input)
+
+                result = dict()
+                for i, output in enumerate(outputs_list):
+                    _tf_res = tf_res[i]
+                    if len(_tf_res.shape) == 4:      # reshaping for 4D tensors
+                        result[output] = _tf_res.transpose(0, 3, 1, 2) # 2, 0, 1
+                    elif len(_tf_res.shape) == 5:    # reshaping for 5D tensors
+                        result[output] = _tf_res.transpose(0, 4, 1, 2, 3)  # 3, 0, 1, 2
+                    else:
+                        result[output] = _tf_res
+                return result
diff --git a/tests/layer_tests/common/utils/common_utils.py b/tests/layer_tests/common/utils/common_utils.py
new file mode 100644
index 00000000000000..a487573afd862a
--- /dev/null
+++ b/tests/layer_tests/common/utils/common_utils.py
@@ -0,0 +1,84 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import logging
+import os
+import subprocess
+import sys
+
+import numpy as np
+
+
+logger = logging.getLogger(__name__)
+
+
+def generate_ir(coverage=False, **kwargs):
+    # Get default mo args
+    mo = os.path.join(os.environ.get("MO_ROOT"), "mo.py")
+    if coverage:
+        params = [sys.executable, '-m', 'coverage', 'run', '-p', '--source={}'.format(os.environ.get("MO_ROOT")),
+                  '--omit=*_test.py', mo]
+    else:
+        params = [sys.executable, mo]
+    for key, value in kwargs.items():
+        if key == "batch":
+            params.extend(("-b", str(value)))
+        elif key == "k":
+            params.extend(("-k", str(value)))
+        elif isinstance(value, bool) and value:
+            params.append("--{}".format(key))
+        elif isinstance(value, bool) and not value:
+            continue
+        elif (isinstance(value, tuple) and value) or (isinstance(value, str)):
+            params.extend(("--{}".format(key), str('"{}"'.format(value))))
+        elif (key == "mean_values" and (' ' in value or '(' in value)):
+            params.extend(("--{}".format(key), str('"{}"'.format(value))))
+        else:
+            params.extend(("--{}".format(key), str(value)))
+    exit_code, stdout, stderr = shell(params)
+    logger.info("Model Optimizer out:\n{}".format(stdout))
+    logger.error(stderr)
+    return exit_code, stderr
+
+
+def shell(cmd, env=None, cwd=None, out_format="plain"):
+    """
+    Run command execution in specified environment
+    :param cmd: list containing command and its parameters
+    :param env: set of environment variables to set for this command
+    :param cwd: working directory from which execute call
+    :param out_format: 'plain' or 'html'. If 'html' all '\n; symbols are replaced by '<br>' tag
+    :return:
+    """
+    if sys.platform.startswith('linux') or sys.platform == 'darwin':
+        cmd = ['/bin/bash', '-c', "unset OMP_NUM_THREADS; " + " ".join(cmd)]
+    else:
+        cmd = " ".join(cmd)
+
+    sys.stdout.write("Running command:\n" + "".join(cmd) + "\n")
+    p = subprocess.Popen(cmd, cwd=cwd, env=env, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    (stdout, stderr) = p.communicate()
+    stdout = str(stdout.decode('utf-8'))
+    stderr = str(stderr.decode('utf-8'))
+    if out_format == "html":
+        stdout = "<br>\n".join(stdout.split('\n'))
+        stderr = "<br>\n".join(stderr.split('\n'))
+    return p.returncode, stdout, stderr
+
+
+def allclose(cur_array, ref_array, atol, rtol):
+    """
+    Comparison of abs_diff and rel_diff with tolerances for every values of corresponding elements.
+    If (abs_diff < atol) or (rel_diff < rtol) for every element, comparison of elements will pass, else will fail.
+    Note: if value is very small, firstly abs_diff will be used. If value is huge, abs_diff may be failed,
+    and rel_diff will be used. So if tensor has small and huge values, need to compare every value
+    with abs_diff and rel_diff instead of using one of it for the whole array.
+    :param cur_array: tensor from IE
+    :param ref_array: tensor from FW
+    :param atol: absolute tolerance (threshold for absolute difference)
+    :param rtol: relative tolerance (threshold for relative difference)
+    :return: bool value means that values of tensors are equal with tolerance or not
+    """
+    abs_diff = np.absolute(cur_array - ref_array)
+    max_val = np.maximum(np.absolute(cur_array), np.absolute(ref_array))
+    return ((abs_diff < atol) | (abs_diff < rtol * max_val)).all()
diff --git a/tests/layer_tests/common/utils/multiprocessing_utils.py b/tests/layer_tests/common/utils/multiprocessing_utils.py
new file mode 100644
index 00000000000000..8d5293476d2ad9
--- /dev/null
+++ b/tests/layer_tests/common/utils/multiprocessing_utils.py
@@ -0,0 +1,99 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import logging as log
+import os
+import platform
+import signal
+import sys
+import traceback
+from multiprocessing import Process, Queue, TimeoutError, ProcessError
+from queue import Empty as QueueEmpty
+from typing import Callable, Union
+
+if platform.system() == "Darwin":
+    # Fix for MacOS
+    import multiprocessing
+    multiprocessing.set_start_method("forkserver", True)
+
+
+def _mp_wrapped_func(func: Callable, func_args: list, queue: Queue, logger_queue: Queue):
+    """
+    Wraps callable object with exception handling. Current wrapper is a target for
+    `multiprocessing_run` function
+    :param func: see `multiprocessing_run`
+    :param func_args: see `multiprocessing_run`
+    :param queue: multiprocessing.Queue(). Used for getting callable object return values
+    :param logger_queue: multiprocessing.Queue(). Used for getting logs from child process in parent process
+    :return:
+    """
+
+    error_message = ""
+    res = None
+    try:
+        res = func(*func_args)
+    except:
+        ex_type, ex_value, tb = sys.exc_info()
+        error_message = "{tb}\n{ex_type}: {ex_value}".format(tb=''.join(traceback.format_tb(tb)),
+                                                             ex_type=ex_type.__name__, ex_value=ex_value)
+    queue.put((error_message, res))
+
+
+def multiprocessing_run(func: Callable, func_args: list, func_log_name: str, timeout: Union[int, None] = None):
+    """
+    Wraps callable object to a separate process using multiprocessing module
+    :param func: callable object
+    :param func_args: list of arguments for callable
+    :param func_log_name: name of callable used for logging
+    :param timeout: positive int to limit execution time
+    :return: return value (or values) from callable object
+    """
+    queue = Queue()
+    logger_queue = Queue(-1)
+    process = Process(target=_mp_wrapped_func, args=(func, func_args, queue, logger_queue))
+    process.start()
+    try:
+        error_message, *ret_args = queue.get(timeout=timeout)
+    except QueueEmpty:
+        raise TimeoutError("{func} running timed out!".format(func=func_log_name))
+    finally:
+        queue.close()
+
+        # Extract logs from Queue and pass to root logger
+        while not logger_queue.empty():
+            rec = logger_queue.get()
+            log.getLogger().handle(rec)
+        logger_queue.close()
+
+        if process.is_alive():
+            process.terminate()
+            process.join()
+        else:
+            exit_signal = multiprocessing_exitcode_to_signal(process.exitcode)
+            if exit_signal:
+                raise ProcessError(
+                    "{func} was killed with a signal {signal}".format(func=func_log_name, signal=exit_signal))
+
+    if error_message:
+        raise ProcessError("\n{func} running failed: \n{msg}".format(func=func_log_name, msg=error_message))
+
+    ret_args = ret_args[0] if len(ret_args) == 1 else ret_args  # unwrap from list if only 1 item is returned
+    return ret_args
+
+
+def multiprocessing_exitcode_to_signal(exitcode):
+    """
+    Map multiprocessing exitcode to signals from "signal" module
+    :param exitcode: multiprocessing exitcode
+    :return: signal from "signal" if exitcode mapped on signal or None
+    """
+    # Multiprocessing return negative values of signal of the process, but on Win they are positive.
+    # Bring the value to the positive format.
+    exit_code = exitcode if os.name == "nt" else -exitcode
+    if exit_code > 0:
+        code_map = {int(getattr(signal, sig)): str(getattr(signal, sig))
+                    for sig in dir(signal) if sig.startswith("SIG")}
+        exit_signal = code_map[exit_code] if exit_code in code_map else exit_code
+    else:
+        exit_signal = None
+    return exit_signal
diff --git a/tests/layer_tests/common/utils/parsers.py b/tests/layer_tests/common/utils/parsers.py
new file mode 100644
index 00000000000000..e5e97727c6a3b3
--- /dev/null
+++ b/tests/layer_tests/common/utils/parsers.py
@@ -0,0 +1,28 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+import xml.etree.ElementTree
+
+
+def mapping_parser(file):
+    """
+    Parse mapping file if it exists
+    :param file: Name of mapping file
+    :return: Dictionary with framework layers as keys and IR layers as values
+    """
+    mapping_dict = {}
+    if os.path.splitext(file)[1] == '.mapping' and os.path.isfile(file):
+        xml_tree = xml.etree.ElementTree.parse(file)
+        xml_root = xml_tree.getroot()
+        for child in xml_root:
+            framework_info = child.find('.//framework')
+            ir_info = child.find('.//IR')
+            if framework_info is None:
+                continue
+            framework_name = framework_info.attrib['name']
+            ir_name = ir_info.attrib['name'] if ir_info is not None else None
+            mapping_dict[framework_name] = ir_name
+    else:
+        raise FileNotFoundError("Mapping file was not found at path {}!".format(os.path.dirname(file)))
+    return mapping_dict
diff --git a/tests/layer_tests/common/utils/tf_utils.py b/tests/layer_tests/common/utils/tf_utils.py
new file mode 100644
index 00000000000000..7900586390cc1d
--- /dev/null
+++ b/tests/layer_tests/common/utils/tf_utils.py
@@ -0,0 +1,109 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+import re
+
+import tensorflow as tf
+import numpy as np
+
+
+os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
+
+
+def load_graph(model_file, output_nodes_for_freeze=None):
+    is_meta = os.path.splitext(model_file)[-1] == ".meta"
+
+    tf.compat.v1.reset_default_graph()
+    graph = tf.Graph()
+    graph_def = tf.compat.v1.GraphDef() if not is_meta else tf.compat.v1.MetaGraphDef()
+
+    with open(model_file, "rb") as f:
+        graph_def.ParseFromString(f.read())
+
+    nodes_to_clear_device = graph_def.node if isinstance(graph_def, tf.compat.v1.GraphDef) else graph_def.graph_def.node
+    for node in nodes_to_clear_device:
+        node.device = ""
+
+    if is_meta:
+        with tf.compat.v1.Session() as sess:
+            restorer = tf.compat.v1.train.import_meta_graph(graph_def)
+            restorer.restore(sess, re.sub('\.meta$', '', model_file))
+            graph_def = tf.compat.v1.graph_util.convert_variables_to_constants(sess, graph_def.graph_def, output_nodes_for_freeze)
+
+    with graph.as_default():
+        tf.import_graph_def(graph_def, name='')
+
+    return graph
+
+
+def collect_tf_references(model_path, feed_dict, out_layer, output_nodes_for_freeze=None):
+    _feed_dict = dict()
+
+    graph = load_graph(model_path, output_nodes_for_freeze)
+    output_tensors_list = list()
+    outputs_list = list()
+    for input in feed_dict:
+        input_node = [node for node in graph.as_graph_def().node if node.name == input][0]
+        if input_node.op == "Placeholder":
+            tensor = graph.get_tensor_by_name(input + ":0")
+            _feed_dict[tensor] = feed_dict[input]
+        else:
+            for parrent_input in input_node.input:
+                in_node = [node for node in graph.as_graph_def().node if node.name == parrent_input][0]
+                if in_node.op in ['Const', 'Assign', 'NoOp', 'Assert']:
+                    continue
+                else:
+                    tensor = graph.get_tensor_by_name(parrent_input + ":0")
+                _feed_dict[tensor] = feed_dict[input]
+
+    for output in out_layer:
+        tensor = graph.get_tensor_by_name(output + ":0")
+        output_tensors_list.append(tensor)
+        outputs_list.append(output)
+    with graph.as_default():
+        with tf.compat.v1.Session(graph=graph) as sess:
+            outputs = sess.run(output_tensors_list, feed_dict=_feed_dict)
+    out_dict = dict(zip(outputs_list, outputs))
+    return out_dict
+
+
+def children(op, graph):
+    op = graph.get_operation_by_name(op)
+    return set(op for out in op.outputs for op in out.consumers())
+
+
+def summarize_graph(model_path, output_nodes_for_freeze=None, reshape_net=None):
+    placeholders = dict()
+    variables = list()
+    outputs = list()
+    graph = load_graph(model_path, output_nodes_for_freeze)
+    unlikely_output_types = ['Const', 'Assign', 'NoOp', 'Placeholder', 'Assert', 'switch_t', 'switch_f']
+    for node in graph.as_graph_def().node:
+        if node.op == 'Placeholder':
+            node_dict = dict()
+            node_dict['type'] = tf.DType(node.attr['dtype'].type).name
+            node_dict['shape'] = str(node.attr['shape'].shape.dim).replace('\n', '').replace(' ', '').replace(
+                'size:', '').replace('[', '').replace(']', '')
+            node_dict['shape'] = tuple(map(lambda x: int(x), node_dict['shape'].split(',')))
+            placeholders[node.name] = node_dict
+        if node.op == "Variable" or node.op == "VariableV2":
+            variables.append(node.name)
+        if len(children(node.name, graph)) == 0:
+            if node.op not in unlikely_output_types and node.name.split('/')[-1] not in unlikely_output_types:
+                outputs.append(node.name)
+    result = dict()
+    result['inputs'] = placeholders
+    result['outputs'] = outputs
+
+    if reshape_net:
+        out_layer = list(result['inputs'].keys()) + result['outputs']
+        feed_dict = {}
+        for inputl in reshape_net:
+            feed_dict.update({inputl: np.ones(shape=reshape_net[inputl])})
+        scoring_res = collect_tf_references(model_path=model_path, feed_dict=feed_dict, out_layer=out_layer)
+        for layer in scoring_res:
+            if layer in result['inputs']:
+                result['inputs'][layer]['shape'] = scoring_res[layer].shape
+
+    return result
diff --git a/tests/layer_tests/conftest.py b/tests/layer_tests/conftest.py
new file mode 100644
index 00000000000000..350fd22a3b20f5
--- /dev/null
+++ b/tests/layer_tests/conftest.py
@@ -0,0 +1,77 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import re
+from pathlib import Path
+import tempfile
+
+import pytest
+from common import constants
+
+
+def pytest_make_parametrize_id(config, val, argname):
+    return " {0}:{1} ".format(argname, val)
+
+
+def pytest_collection_modifyitems(items):
+    def remove_ignored_attrs(ref_dict, dict_to_upd):
+        _dict_to_upd = dict_to_upd.copy()
+        for key, value in dict_to_upd.items():
+            if key not in ref_dict.keys():
+                _dict_to_upd.pop(key)
+            elif isinstance(value, dict):
+                _dict_to_upd[key] = remove_ignored_attrs(ref_dict[key], value)
+        return _dict_to_upd
+
+    for test in items:
+        special_marks = [mark for mark in test.own_markers if "special_" in mark.name]
+        for mark in special_marks:
+            if mark.name == "special_xfail":
+                params = test.callspec.params
+                # Remove items from params if key of item is not in mark.kwargs["args"].
+                # Remaining items will be used to mark test cases that contain them.
+                # It is required to specify in mark only valuable parameters
+                # (e.g. {"device": "FP16"} will mean that for all test cases with FP16 test will be marked)
+                params = remove_ignored_attrs(mark.kwargs["args"], params)
+                if mark.kwargs["args"] == params:
+                    test.add_marker(pytest.mark.xfail(reason=mark.kwargs["reason"]))
+
+
+@pytest.mark.hookwrapper
+def pytest_runtest_makereport(item, call):
+    pytest_html = item.config.pluginmanager.getplugin('html')
+    outcome = yield
+    report = outcome.get_result()
+    extra = getattr(report, 'extra', [])
+    if report.when == 'call':
+        xfail_reason = getattr(report, 'wasxfail', None)
+        if report.skipped and xfail_reason:
+            jira_ticket_nums = re.findall(r"\*-\d+", xfail_reason)
+            for ticket_num in jira_ticket_nums:
+                extra.append(pytest_html.extras.url(ticket_num))
+        report.extra = extra
+
+
+def pytest_addoption(parser):
+    """Specify command-line options for all plugins"""
+    parser.addoption(
+        "--ir_version",
+        required=True,
+        action="store",
+        help="Version of IR to generate by Model Optimizer")
+
+
+@pytest.fixture(scope="session")
+def ir_version(request):
+    """Fixture function for command-line option."""
+    return request.config.getoption('ir_version')
+
+
+@pytest.fixture(scope="function")
+def temp_dir(request):
+    """Create directory for test purposes."""
+    Path(constants.out_path).mkdir(parents=True, exist_ok=True)
+    test_name = re.sub(r"[^\w_]", "_", request.node.originalname)
+    device = request.node.funcargs["ie_device"].upper()
+    temp_dir = tempfile.TemporaryDirectory(dir=constants.out_path, prefix=f"{device}_{test_name}")
+    yield str(temp_dir.name)
diff --git a/tests/layer_tests/onnx_tests/conftest.py b/tests/layer_tests/onnx_tests/conftest.py
new file mode 100644
index 00000000000000..bbf635f56809ac
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/conftest.py
@@ -0,0 +1,12 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import inspect
+
+from common.layer_test_class import get_params
+
+
+def pytest_generate_tests(metafunc):
+    test_gen_attrs_names = list(inspect.signature(get_params).parameters)
+    params = get_params()
+    metafunc.parametrize(test_gen_attrs_names, params, scope="function")
diff --git a/tests/layer_tests/onnx_tests/test_abs.py b/tests/layer_tests/onnx_tests/test_abs.py
new file mode 100644
index 00000000000000..577965df82ad68
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_abs.py
@@ -0,0 +1,178 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestAbs(OnnxRuntimeLayerTest):
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                   IR net
+
+            Input->Abs->Output   =>    Input->Abs
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Abs',
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'Abs'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->Concat(+abs const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.randint(-127, 127, shape).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Abs',
+            inputs=['const'],
+            outputs=['abs']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'abs'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = abs(constant)
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_abs(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_abs_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_add_sub_mul_div.py b/tests/layer_tests/onnx_tests/test_add_sub_mul_div.py
new file mode 100644
index 00000000000000..9076ce8e650f11
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_add_sub_mul_div.py
@@ -0,0 +1,337 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+
+
+class TestOperations(Caffe2OnnxLayerTest):
+    def create_net(self, shape1, shape2, op, precision, ir_version, opset=None):
+        """
+            ONNX net                                  IR net
+
+            Input->Add/Mul with const->Output   =>    Input->Eltwise
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        if op not in ['Add', 'Sub', 'Mul', 'Div']:
+            raise ValueError("Operation has to be either Add or Mul or Sub or Div")
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape1)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape1)
+
+        min_val = 1 if op == 'Div' else -127
+        if shape2:
+            const = np.random.randint(min_val, 127, shape2).astype(np.float)
+        else:
+            const = np.random.randint(min_val, 127, 1).astype(np.float)
+            # TODO: add check when MO remove redundant layer (as Add/Sub if const = 0 or Mul/Div if const = 1)
+            if const in [0, 1]:
+                const = np.array([2], dtype=np.float)
+
+        node_const_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=const.shape,
+                vals=const.flatten(),
+            ),
+        )
+
+        node_def = helper.make_node(
+            op,
+            inputs=['input', 'const'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #   Create reference IR net
+        if op == 'Div':
+            const = np.power(const, -1)
+        elif op == 'Sub':
+            const = -const
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape1, shape2, op, precision, ir_version, opset=None):
+        """
+            ONNX net                                                      IR net
+
+            Input->Concat with two added/multiplied consts->Output   =>   Input->Concat
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        if op not in ['Add', 'Sub', 'Mul', 'Div']:
+            raise ValueError("op has to be either Add or Mul")
+
+        concat_axis = 0
+        output_shape = list(shape1)
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape1)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        const1 = np.random.randint(-127, 127, shape1).astype(np.float)
+        min_val = 1 if op == 'Div' else -127
+        if shape2:
+            const2 = np.random.randint(min_val, 127, shape2).astype(np.float)
+        else:
+            const2 = np.random.randint(min_val, 127, 1).astype(np.float)
+
+        node_const1_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=const1.shape,
+                vals=const1.flatten(),
+            ),
+        )
+
+        node_const2_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const2'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=const2.shape,
+                vals=const2.flatten(),
+            ),
+        )
+
+        node_def = helper.make_node(
+            op,
+            inputs=['const1', 'const2'],
+            outputs=['node_out']
+        )
+
+        node_concat_def = helper.make_node(
+            'Concat',
+            inputs=['input', 'node_out'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const1_def, node_const2_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #   Create reference IR net
+        if op == 'Add':
+            constant_calculated = np.add(const1, const2)
+        elif op == 'Sub':
+            constant_calculated = np.subtract(const1, const2)
+        elif op == 'Mul':
+            constant_calculated = np.multiply(const1, const2)
+        elif op == 'Div':
+            constant_calculated = np.divide(const1, const2)
+
+        if precision == 'FP16':
+            constant_calculated = constant_calculated.astype(np.float16)
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [dict(shape1=[2, 4], shape2=[2, 4]),
+                           # scalar cases
+                           dict(shape1=[2, 4], shape2=None)]
+
+    test_data = [dict(shape1=[4, 6], shape2=[4, 6]),
+                 dict(shape1=[4, 6, 8], shape2=[4, 6, 8]),
+                 dict(shape1=[4, 6, 8, 10], shape2=[4, 6, 8, 10]),
+                 dict(shape1=[4, 6, 8, 10, 12], shape2=[4, 6, 8, 10, 12]),
+                 # scalar cases
+                 dict(shape1=[4, 6], shape2=None),
+                 dict(shape1=[4, 6, 8], shape2=None),
+                 dict(shape1=[4, 6, 8, 10], shape2=None),
+                 dict(shape1=[4, 6, 8, 10, 12], shape2=None)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_add(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Add', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_add_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Add', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sub(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Sub', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sub_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Sub', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_mul(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Mul', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_mul_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Mul', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_div(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Div', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_div_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Div', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_add_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Add', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_add_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Add', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_sub_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Sub', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_sub_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Sub', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_mul_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Mul', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_mul_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Mul', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_div_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Div', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_div_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Div', precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_add_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Add', precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_add_const_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Add', precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sub_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Sub', precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sub_const_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Sub', precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_mul_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Mul', precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_mul_const_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Mul', precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_div_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, op='Div', precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_div_const_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, op='Div', precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_and.py b/tests/layer_tests/onnx_tests/test_and.py
new file mode 100644
index 00000000000000..115fe1e706dd1c
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_and.py
@@ -0,0 +1,268 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestAnd(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randint(0, 2, inputs_dict[input]).astype(np.bool)
+        return inputs_dict
+
+    def create_net(self, shape1, shape2, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->And with 2nd input->Output   =>    Input->LogicalAnd
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input1 = helper.make_tensor_value_info('input1', TensorProto.BOOL, shape1)
+        input2 = helper.make_tensor_value_info('input2', TensorProto.BOOL, shape2)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, shape1)
+
+        node_def = helper.make_node(
+            'And',
+            inputs=['input1', 'input2'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input1, input2],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input1': {'kind': 'op', 'type': 'Parameter'},
+                'input1_data': {'shape': shape1, 'kind': 'data'},
+                'input2': {'kind': 'op', 'type': 'Parameter'},
+                'input2_data': {'shape': shape2, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'LogicalAnd'},
+                'node_data': {'shape': shape1, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input1', 'input1_data'),
+                                   ('input2', 'input2_data'),
+                                   ('input1_data', 'node'),
+                                   ('input2_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    def create_net_one_const(self, shape1, shape2, ir_version):
+        """
+            ONNX net                              IR net
+
+            Input->And with const->Output   =>    Input->LogicalAnd
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.BOOL, shape1)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, shape1)
+
+        const = np.random.randint(0, 2, shape2).astype(np.bool)
+
+        node_const_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.BOOL,
+                dims=const.shape,
+                vals=const.flatten(),
+            ),
+        )
+
+        node_def = helper.make_node(
+            'And',
+            inputs=['input', 'const'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape1, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': const.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': const.shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'LogicalAnd'},
+                'node_data': {'shape': shape1, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'node'),
+                                   ('const_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape1, shape2, ir_version):
+        """
+            ONNX net                                          IR net
+
+            Input->Concat with const and const->Output   =>   Input->Concat
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        concat_axis = 0
+        output_shape = list(shape1)
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.BOOL, shape1)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, output_shape)
+
+        const1 = np.random.randint(0, 2, shape1).astype(np.bool)
+        const2 = np.random.randint(0, 2, shape2).astype(np.bool)
+
+        node_const1_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.BOOL,
+                dims=const1.shape,
+                vals=const1.flatten(),
+            ),
+        )
+
+        node_const2_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const2'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.BOOL,
+                dims=const2.shape,
+                vals=const2.flatten(),
+            ),
+        )
+
+        node_def = helper.make_node(
+            'And',
+            inputs=['const1', 'const2'],
+            outputs=['node_out']
+        )
+
+        node_concat_def = helper.make_node(
+            'Concat',
+            inputs=['input', 'node_out'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const1_def, node_const2_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+        constant_calculated = np.logical_and(const1, const2)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': const1.shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant_calculated.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': const1.shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape1=[4, 6], shape2=[4, 6]),
+                 dict(shape1=[4, 6, 8], shape2=[4, 6, 8]),
+                 dict(shape1=[4, 6, 8, 10], shape2=[4, 6, 8, 10]),
+                 dict(shape1=[4, 6, 8, 10, 12], shape2=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_and(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_and_one_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_one_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_and_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_argmax.py b/tests/layer_tests/onnx_tests/test_argmax.py
new file mode 100644
index 00000000000000..476a06d187b7e6
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_argmax.py
@@ -0,0 +1,153 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestArgMax(OnnxRuntimeLayerTest):
+    def create_net(self, shape, axis, keepdims, ir_version):
+        """
+            ONNX net                      IR net
+
+            Input->ArgMax->Output   =>    Input->TopK
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        output_shape = shape.copy()
+        output_shape[axis if axis is not None else 0] = 1
+        output_shape_squeeze = output_shape.copy()
+        if keepdims == 0:
+            output_shape_squeeze.remove(1)
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.INT64, output_shape_squeeze)
+
+        const = np.random.randint(-10, 10, output_shape_squeeze).astype(np.int64)
+
+        args = dict()
+        if axis is not None:
+            args['axis'] = axis
+        else:
+            axis = 0
+        if keepdims is not None:
+            args['keepdims'] = keepdims
+        node_def = onnx.helper.make_node(
+            'ArgMax',
+            inputs=['input'],
+            outputs=['argmax' if keepdims is None or keepdims == 1 else 'output'],
+            **args
+        )
+        edges = [node_def]
+
+        if keepdims is None or keepdims == 1:
+            node_flatten_def = onnx.helper.make_node(
+                'Flatten',
+                inputs=['argmax'],
+                outputs=['output']
+            )
+            edges.append(node_flatten_def)
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            edges,
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'const_indata': {'shape': [1], 'kind': 'data'},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': [], 'kind': 'data'},  # TODO shape [] or [1] ??
+                'node': {'kind': 'op', 'type': 'TopK'},
+                'node_data': {'shape': output_shape, 'kind': 'data'},
+                'indices_data': {'shape': output_shape, 'kind': 'data'},
+                'result1': {'kind': 'op', 'type': 'Result'},
+                'result2': {'kind': 'op', 'type': 'Result'}
+            }
+            edges = [('input', 'input_data'),
+                     ('const_indata', 'const'),
+                     ('const', 'const_data'),
+                     ('input_data', 'node'),
+                     ('const_data', 'node'),
+                     ('node', 'node_data'),
+                     ('node', 'indices_data'),
+                     ('node_data', 'result1')]
+            if keepdims == 0:
+                nodes_attributes.update({'squeeze_const_indata': {'shape': [1], 'kind': 'data'},
+                                         'squeeze_const': {'kind': 'op', 'type': 'Const'},
+                                         'squeeze_const_data': {'shape': [1], 'kind': 'data'},
+                                         'squeeze': {'kind': 'op', 'type': 'Squeeze'},
+                                         'squeeze_data': {'shape': output_shape_squeeze, 'kind': 'data'}
+                                         })
+                edges.extend([('squeeze_const_indata', 'squeeze_const'),
+                              ('squeeze_const', 'squeeze_const_data'),
+                              ('indices_data', 'squeeze'),
+                              ('squeeze_const_data', 'squeeze'),
+                              ('squeeze', 'squeeze_data'),
+                              ('squeeze_data', 'result2')])
+            else:
+                nodes_attributes.update(
+                    {'flatten_const_indata': {'kind': 'data', 'value': [0, -1]},
+                     'flatten_const': {'kind': 'op', 'type': 'Const'},
+                     'flatten_const_data': {'shape': [2], 'kind': 'data'},
+                     'flatten': {'kind': 'op', 'type': 'Reshape'},
+                     'flatten_data': {'shape': [output_shape_squeeze[0], np.prod(output_shape_squeeze[1:])],
+                                      'kind': 'data'}
+                     })
+                edges.extend([('indices_data', 'flatten'),
+                              ('flatten_const_indata', 'flatten_const'),
+                              ('flatten_const', 'flatten_const_data'),
+                              ('flatten_const_data', 'flatten'),
+                              ('flatten', 'flatten_data'),
+                              ('flatten_data', 'result2')])
+
+            ref_net = build_graph(nodes_attributes, edges)
+
+        return onnx_net, ref_net
+
+    test_data = [
+        dict(shape=[10, 12], axis=None),
+        dict(shape=[10, 12], axis=1),
+        dict(shape=[8, 10, 12], axis=None),
+        dict(shape=[8, 10, 12], axis=1),
+        dict(shape=[8, 10, 12], axis=2),
+        dict(shape=[6, 8, 10, 12], axis=None),
+        dict(shape=[6, 8, 10, 12], axis=1),
+        dict(shape=[6, 8, 10, 12], axis=2),
+        dict(shape=[6, 8, 10, 12], axis=3),
+        dict(shape=[4, 6, 8, 10, 12], axis=None),
+        dict(shape=[4, 6, 8, 10, 12], axis=1),
+        dict(shape=[4, 6, 8, 10, 12], axis=2),
+        dict(shape=[4, 6, 8, 10, 12], axis=3),
+        dict(shape=[4, 6, 8, 10, 12], axis=4)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("keepdims", [None, 0])
+    @pytest.mark.nightly
+    def test_argmax(self, params, keepdims, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, keepdims=keepdims),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_bn.py b/tests/layer_tests/onnx_tests/test_bn.py
new file mode 100644
index 00000000000000..fb916f4a9aaae0
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_bn.py
@@ -0,0 +1,133 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+
+
+class TestBatchNormalization(Caffe2OnnxLayerTest):
+    def create_net(self, shape, epsilon, precision, ir_version, opset=None):
+        """
+            ONNX net                                  IR net
+
+            Input->BatchNormalization->Output   =>    Input->ScaleShift(Power)
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        scale_const = np.random.randint(128, 256, shape[1]).astype(np.float32) / 128.
+        bias_const = np.random.randint(0, 128, shape[1]).astype(np.float32) / 128.
+        mean_const = np.random.randint(-127, 127, shape[1]).astype(np.float32)
+        var_const = np.random.randint(128, 256, shape[1]).astype(np.float32) / 128.
+
+        node_scale_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['scale_const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=scale_const.shape,
+                vals=scale_const.flatten(),
+            ),
+        )
+
+        node_bias_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['bias'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=bias_const.shape,
+                vals=bias_const.flatten(),
+            ),
+        )
+
+        node_mean_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['mean'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=mean_const.shape,
+                vals=mean_const.flatten(),
+            ),
+        )
+
+        node_var_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['var'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=var_const.shape,
+                vals=var_const.flatten(),
+            ),
+        )
+
+        args = dict(epsilon=epsilon)
+        if opset == 6:
+            args['is_test'] = 1;
+        node_def = helper.make_node(
+            'BatchNormalization',
+            inputs=['input', 'scale_const', 'bias', 'mean', 'var'],
+            outputs=['output'],
+            **args
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_scale_def, node_bias_def, node_mean_def, node_var_def, node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[1, 1, 4, 6], epsilon=0.001),
+                 dict(shape=[1, 2, 4, 6], epsilon=0.001),
+                 dict(shape=[2, 3, 4, 6], epsilon=0.001)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_bn(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_bn_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_bn_opset7(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, opset=7, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_ceil.py b/tests/layer_tests/onnx_tests/test_ceil.py
new file mode 100644
index 00000000000000..cde48393375ead
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_ceil.py
@@ -0,0 +1,194 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestCeil(OnnxRuntimeLayerTest):
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                    IR net
+
+            Input->Ceil->Output   =>    Input->Ceiling
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Ceil',
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'Ceiling'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                   IR net
+
+            Input->Concat(+ceil const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.randn(*shape).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Ceil',
+            inputs=['const1'],
+            outputs=['floor']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'floor'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.ceil(constant)
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [dict(shape=[2, 4, 6, 8, 10])]
+
+    test_data = [
+        dict(shape=[1, 2]),
+        dict(shape=[2, 4, 6]),
+        dict(shape=[2, 4, 6, 8]),
+        dict(shape=[2, 4, 6, 8, 10])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_ceil_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_ceil_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_ceil(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_ceil_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_clip.py b/tests/layer_tests/onnx_tests/test_clip.py
new file mode 100644
index 00000000000000..94b55bf80ef53e
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_clip.py
@@ -0,0 +1,169 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestClip(OnnxRuntimeLayerTest):
+    def create_net(self, shape, ir_version, opset, min=None, max=None):
+        """
+            ONNX net                    IR net
+
+            Input->Clip->Output   =>    Input->Clamp
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        nodes = []
+        if opset < 11:
+            args = dict()
+            if min is not None:
+                args['min'] = min
+            if max is not None:
+                args['max'] = max
+            node_def = onnx.helper.make_node(
+                'Clip',
+                inputs=['input'],
+                outputs=['output'],
+                **args
+            )
+            nodes.append(node_def)
+        else:
+            clip_inputs = ['input']
+            if min is not None:
+                node_min_def = onnx.helper.make_node(
+                    'Constant',
+                    inputs=[],
+                    outputs=['min_const'],
+                    value=helper.make_tensor(
+                        name='const_tensor',
+                        data_type=TensorProto.FLOAT,
+                        dims=[],
+                        vals=[min],
+                    ),
+                )
+                clip_inputs.append('min_const')
+                nodes.append(node_min_def)
+            else:
+                clip_inputs.append('')
+            if max is not None:
+                node_max_def = onnx.helper.make_node(
+                    'Constant',
+                    inputs=[],
+                    outputs=['max_const'],
+                    value=helper.make_tensor(
+                        name='const_tensor',
+                        data_type=TensorProto.FLOAT,
+                        dims=[],
+                        vals=[max],
+                    ),
+                )
+                clip_inputs.append('max_const')
+                nodes.append(node_max_def)
+            node_def = onnx.helper.make_node(
+                'Clip',
+                inputs=clip_inputs,
+                outputs=['output']
+            )
+            nodes.append(node_def)
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            if opset < 11 or min is not None and max is not None:
+                nodes_attributes = {
+                    'input': {'kind': 'op', 'type': 'Parameter'},
+                    'input_data': {'shape': shape, 'kind': 'data'},
+                    'node': {'kind': 'op', 'type': 'Clamp',
+                             'min': min if min is not None else -3.4028235e+38,
+                             'max': max if max is not None else 3.4028235e+38},
+                    'node_data': {'shape': shape, 'kind': 'data'},
+                    'result': {'kind': 'op', 'type': 'Result'}
+                }
+                ref_net = build_graph(nodes_attributes,
+                                      [('input', 'input_data'),
+                                       ('input_data', 'node'),
+                                       ('node', 'node_data'),
+                                       ('node_data', 'result')
+                                       ])
+            else:
+                nodes_attributes = {
+                    'input': {'kind': 'op', 'type': 'Parameter'},
+                    'input_data': {'shape': shape, 'kind': 'data'},
+                    'input_const_data': {'kind': 'data', 'value': [min] if min is not None else [max]},
+                    'const': {'kind': 'op', 'type': 'Const'},
+                    'const_data': {'shape': [], 'kind': 'data'},
+                    'node': {'kind': 'op', 'type': 'Minimum' if max is not None else 'Maximum'},
+                    'node_data': {'shape': shape, 'kind': 'data'},
+                    'result': {'kind': 'op', 'type': 'Result'}
+                }
+                ref_net = build_graph(nodes_attributes,
+                                      [('input', 'input_data'),
+                                       ('input_const_data', 'const'),
+                                       ('const', 'const_data'),
+                                       ('input_data', 'node'),
+                                       ('const_data', 'node'),
+                                       ('node', 'node_data'),
+                                       ('node_data', 'result')
+                                       ])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[12], min=-3.5),
+                 dict(shape=[12], max=3.5),
+                 dict(shape=[12], min=-3.5, max=3.5),
+                 dict(shape=[10, 12], min=-3.5),
+                 dict(shape=[10, 12], max=3.5),
+                 dict(shape=[10, 12], min=-3.5, max=3.5),
+                 dict(shape=[8, 10, 12], min=-3.5),
+                 dict(shape=[8, 10, 12], max=3.5),
+                 dict(shape=[8, 10, 12], min=-3.5, max=3.5),
+                 dict(shape=[6, 8, 10, 12], min=-3.5),
+                 dict(shape=[6, 8, 10, 12], max=3.5),
+                 dict(shape=[6, 8, 10, 12], min=-3.5, max=3.5),
+                 dict(shape=[4, 6, 8, 10, 12], min=-3.5),
+                 dict(shape=[4, 6, 8, 10, 12], max=3.5),
+                 dict(shape=[4, 6, 8, 10, 12], min=-3.5, max=3.5)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_clip_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, opset=6), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_clip_opset11(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, opset=11), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_concat.py b/tests/layer_tests/onnx_tests/test_concat.py
new file mode 100644
index 00000000000000..1321ccda5b9c1c
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_concat.py
@@ -0,0 +1,225 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestConcat(Caffe2OnnxLayerTest):
+    # TODO Add test with default values (axis=0)
+    def create_concat_net_const(self, input_shape, output_shape, axis, ir_version):
+        """
+            ONNX net                                                 IR net
+
+            Input(const)----->Concat--------->Concat->Output   =>    Input--->Concat
+            Input(const)-----'               '                        Const---'
+                                      Input-'
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        concat_output_shape = output_shape.copy()
+        concat_output_shape[concat_axis] *= 2
+
+        const_number = np.prod(input_shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, output_shape)
+
+        # Output for concat
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, concat_output_shape)
+
+        node_const1_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const1_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=input_shape,
+                vals=constant,
+            ),
+        )
+
+        node_const2_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const2'],
+            value=helper.make_tensor(
+                name='const2_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=input_shape,
+                vals=constant,
+            ),
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['const1', 'const2'],
+            outputs=['output_concat'],
+            axis=axis
+        )
+
+        node_dyn_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'output_concat'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const1_def, node_const2_def, node_concat_def, node_dyn_concat_def],
+            'test_concat_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_split_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        constant_reshape = np.reshape(constant, input_shape)
+        constant_reshape = np.concatenate([constant_reshape, constant_reshape], axis=axis)
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': output_shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant_reshape.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': output_shape, 'value': None, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': concat_output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_3D = [
+        dict(input_shape=[1, 50, 50],
+             output_shape=[2, 50, 50],
+             axis=0),
+
+        dict(input_shape=[2, 50, 50],
+             output_shape=[2, 100, 50],
+             axis=1),
+
+        dict(input_shape=[4, 50, 50],
+             output_shape=[4, 50, 100],
+             axis=2),
+    ]
+
+    test_data_4D_precommit = [
+        dict(input_shape=[1, 32, 800, 800],
+             output_shape=[2, 32, 800, 800],
+             axis=0)
+    ]
+
+    test_data_4D = [
+        dict(input_shape=[1, 32, 800, 800],
+             output_shape=[2, 32, 800, 800],
+             axis=0),
+
+        dict(input_shape=[4, 32, 80, 80],
+             output_shape=[4, 64, 80, 80],
+             axis=1),
+
+        dict(input_shape=[2, 21, 80, 80],
+             output_shape=[2, 21, 160, 80],
+             axis=2),
+
+        dict(input_shape=[3, 21, 80, 80],
+             output_shape=[3, 21, 80, 160],
+             axis=3),
+    ]
+
+    test_data_5D_precommit = [
+        dict(input_shape=[1, 50, 50, 80, 60],
+             output_shape=[2, 50, 50, 80, 60],
+             axis=0),
+
+        dict(input_shape=[1, 50, 50, 80, 60],
+             output_shape=[1, 50, 50, 80, 120],
+             axis=4),
+    ]
+
+    test_data_5D = [
+        dict(input_shape=[1, 50, 50, 80, 60],
+             output_shape=[2, 50, 50, 80, 60],
+             axis=0),
+
+        dict(input_shape=[1, 50, 50, 80, 60],
+             output_shape=[1, 100, 50, 80, 60],
+             axis=1),
+
+        dict(input_shape=[1, 50, 50, 80, 60],
+             output_shape=[1, 50, 100, 80, 60],
+             axis=2),
+
+        dict(input_shape=[1, 50, 50, 80, 60],
+             output_shape=[1, 50, 50, 160, 60],
+             axis=3),
+
+        dict(input_shape=[1, 50, 50, 80, 60],
+             output_shape=[1, 50, 50, 80, 120],
+             axis=4),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_concat_3D_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_concat_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D_precommit)
+    @pytest.mark.precommit
+    def test_concat_4D_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_concat_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_concat_4D_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_concat_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D_precommit)
+    @pytest.mark.nightly
+    def test_concat_5D_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_concat_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_concat_5D_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_concat_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_conv.py b/tests/layer_tests/onnx_tests/test_conv.py
new file mode 100644
index 00000000000000..94774be4af9345
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_conv.py
@@ -0,0 +1,413 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+
+import numpy as np
+import pytest
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestConv(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randn(*inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_net(self, shape, weights_shape, dilations, group, pads, strides, bias, ir_version, auto_pad=None):
+        """
+            ONNX net                    IR net
+
+            Input->Conv->Output   =>    Input->Convolution
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        output_shape = np.array(shape)
+        output_shape[1] = group
+        _pads = np.array(pads).reshape([2, -1])
+        kernel_extent = np.array(dilations) * (np.array(weights_shape[2:]) - 1) + 1
+        spatial_val_wo_stride = shape[2:] + np.add(_pads[0, :], _pads[1, :]) - kernel_extent
+        output_shape[2:] = (spatial_val_wo_stride.astype(np.float) / strides + 1).astype(np.int64)
+        output_shape = output_shape.astype(np.int).tolist()
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        weights_const = np.random.randn(*weights_shape).astype(np.float32)
+
+        node_weights_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['weights'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=weights_const.shape,
+                vals=weights_const.flatten(),
+            ),
+        )
+
+        conv_args = dict(kernel_shape=weights_shape[2:],
+                         dilations=dilations,
+                         group=group,
+                         strides=strides)
+        if pads and auto_pad not in ['SAME_UPPER', 'SAME_LOWER']:
+            conv_args['pads'] = pads
+        if auto_pad:
+            conv_args['auto_pad'] = auto_pad
+        if bias:
+            bias_const = np.random.randint(-10, 10, weights_shape[0]).astype(np.float32)
+
+            node_bias_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['bias'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.FLOAT,
+                    dims=bias_const.shape,
+                    vals=bias_const.flatten(),
+                ),
+            )
+            node_def = onnx.helper.make_node(
+                'Conv',
+                inputs=['input', 'weights', 'bias'],
+                outputs=['output'],
+                **conv_args
+            )
+            nodes = [node_weights_def, node_bias_def, node_def]
+        else:
+            node_def = onnx.helper.make_node(
+                'Conv',
+                inputs=['input', 'weights'],
+                outputs=['output'],
+                **conv_args
+            )
+            nodes = [node_weights_def, node_def]
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            if len(shape) == 3:
+                input_shape = shape.copy()
+                input_shape.insert(2, 1)
+                node_shape = output_shape.copy()
+                node_shape.insert(2, 1)
+                nodes_attributes = {
+                    'input': {'kind': 'op', 'type': 'Parameter'},
+                    'input_data': {'shape': shape, 'kind': 'data'},
+                    'before_shape_const_indata': {'shape': [len(input_shape)], 'value': input_shape, 'kind': 'data'},
+                    'before_shape_const': {'kind': 'op', 'type': 'Const'},
+                    'before_shape_const_data': {'shape': [len(input_shape)], 'kind': 'data'},
+                    'reshape_before': {'kind': 'op', 'type': 'Reshape'},
+                    'reshape_before_data': {'shape': input_shape, 'kind': 'data'},
+                    'kernel_indata': {'kind': 'data', 'shape': [len(weights_const.flatten())]},
+                    'kernel': {'kind': 'op', 'type': 'Const'},
+                    'kernel_data': {'kind': 'data', 'value': None},
+                    'node': {'kind': 'op', 'type': 'Convolution' if group == 1 else 'GroupConvolution',
+                             'dilations': [1, dilations[0]],
+                             'pads_begin': [0, _pads[0, 0]], 'pads_end': [0, _pads[1, 0]]},
+                    'node_data': {'shape': node_shape, 'kind': 'data'},
+                    'after_shape_const_indata': {'shape': [len(output_shape)], 'value': output_shape, 'kind': 'data'},
+                    'after_shape_const': {'kind': 'op', 'type': 'Const'},
+                    'after_shape_const_data': {'shape': [len(output_shape)], 'kind': 'data'},
+                    'reshape_after': {'kind': 'op', 'type': 'Reshape'},
+                    'reshape_after_data': {'shape': output_shape, 'kind': 'data'},
+                    'result': {'kind': 'op', 'type': 'Result'}}
+                edges = [('input', 'input_data'),
+                         ('input_data', 'reshape_before'),
+                         ('before_shape_const_indata', 'before_shape_const'),
+                         ('before_shape_const', 'before_shape_const_data'),
+                         ('before_shape_const_data', 'reshape_before'),
+                         ('reshape_before', 'reshape_before_data'),
+                         ('reshape_before_data', 'node'),
+                         ('kernel_indata', 'kernel'),
+                         ('kernel', 'kernel_data'),
+                         ('kernel_data', 'node'),
+                         ('node', 'node_data'),
+                         ('node_data', 'reshape_after'),
+                         ('after_shape_const_indata', 'after_shape_const'),
+                         ('after_shape_const', 'after_shape_const_data'),
+                         ('after_shape_const_data', 'reshape_after'),
+                         ('reshape_after', 'reshape_after_data')]
+                if bias:
+                    nodes_attributes.update({'const_indata': {'kind': 'data', 'value': bias_const.flatten()},
+                                             'const': {'kind': 'op', 'type': 'Const'},
+                                             'const_data': {'kind': 'data', 'shape': None},
+                                             'bias': {'type': 'Add', 'kind': 'op'},
+                                             'bias_data': {'kind': 'data', 'shape': output_shape}})
+                    edges += [('reshape_after_data', 'bias'),
+                              ('const_indata', 'const'),
+                              ('const', 'const_data'),
+                              ('const_data', 'bias'),
+                              ('bias', 'bias_data'),
+                              ('bias_data', 'result')]
+                else:
+                    edges += [('reshape_after_data', 'result')]
+                ref_net = build_graph(nodes_attributes, edges)
+            else:
+                _weights_shape = weights_shape.copy()
+                if group != 1:
+                    _weights_shape.insert(1, 1)
+                nodes_attributes = {
+                    'input': {'kind': 'op', 'type': 'Parameter'},
+                    'input_data': {'shape': shape, 'kind': 'data'},
+                    'kernel_indata': {'kind': 'data', 'value': weights_const.flatten()},
+                    'kernel': {'kind': 'op', 'type': 'Const'},
+                    'kernel_data': {'kind': 'data', 'shape': _weights_shape},
+                    'node': {'kind': 'op', 'type': 'Convolution' if group == 1 else 'GroupConvolution',
+                             'dilations': dilations, 'pads_begin': _pads[0, :], 'pads_end': _pads[1, :]},
+                    'node_data': {'shape': output_shape, 'kind': 'data'},
+                    'result': {'kind': 'op', 'type': 'Result'}}
+                edges = [('input', 'input_data'),
+                         ('input_data', 'node'),
+                         ('kernel_indata', 'kernel'),
+                         ('kernel', 'kernel_data'),
+                         ('kernel_data', 'node'),
+                         ('node', 'node_data')]
+
+                if bias:
+                    nodes_attributes.update({'const_indata': {'kind': 'data', 'value': bias_const.flatten()},
+                                             'const': {'kind': 'op', 'type': 'Const'},
+                                             'const_data': {'kind': 'data', 'shape': None},
+                                             'bias': {'type': 'Add', 'kind': 'op'},
+                                             'bias_data': {'kind': 'data', 'shape': output_shape}})
+                    edges += [('node_data', 'bias'),
+                              ('const_indata', 'const'),
+                              ('const', 'const_data'),
+                              ('const_data', 'bias'),
+                              ('bias', 'bias_data'),
+                              ('bias_data', 'result')]
+                else:
+                    edges += [('node_data', 'result')]
+
+                ref_net = build_graph(nodes_attributes, edges)
+
+        return onnx_net, ref_net
+
+    test_data_3D = [
+        dict(weights_shape=[1, 3, 3], group=1),
+        dict(weights_shape=[1, 3, 5], group=1),
+        dict(weights_shape=[3, 1, 3], group=3),
+        dict(weights_shape=[3, 1, 5], group=3)]
+
+    test_data_3D_autopad = [
+        dict(weights_shape=[1, 3, 3], group=1, pads=[1, 1], strides=[1], dilations=[1]),
+        dict(weights_shape=[1, 3, 3], group=1, pads=[2, 2], strides=[1], dilations=[2]),
+        dict(weights_shape=[1, 3, 3], group=1, pads=[1, 1], strides=[2], dilations=[1]),
+        dict(weights_shape=[1, 3, 3], group=1, pads=[2, 2], strides=[2], dilations=[2]),
+        dict(weights_shape=[1, 3, 5], group=1, pads=[2, 2], strides=[1], dilations=[1]),
+        dict(weights_shape=[1, 3, 5], group=1, pads=[4, 4], strides=[1], dilations=[2]),
+        dict(weights_shape=[1, 3, 5], group=1, pads=[2, 2], strides=[2], dilations=[1]),
+        dict(weights_shape=[1, 3, 5], group=1, pads=[4, 4], strides=[2], dilations=[2]),
+        dict(weights_shape=[3, 1, 3], group=3, pads=[1, 1], strides=[1], dilations=[1]),
+        dict(weights_shape=[3, 1, 3], group=3, pads=[2, 2], strides=[1], dilations=[2]),
+        dict(weights_shape=[3, 1, 3], group=3, pads=[1, 1], strides=[2], dilations=[1]),
+        dict(weights_shape=[3, 1, 3], group=3, pads=[2, 2], strides=[2], dilations=[2]),
+        dict(weights_shape=[3, 1, 5], group=3, pads=[2, 2], strides=[1], dilations=[1]),
+        dict(weights_shape=[3, 1, 5], group=3, pads=[4, 4], strides=[1], dilations=[2]),
+        dict(weights_shape=[3, 1, 5], group=3, pads=[2, 2], strides=[2], dilations=[1]),
+        dict(weights_shape=[3, 1, 5], group=3, pads=[4, 4], strides=[2], dilations=[2])]
+
+    test_data_4D_precommit = [
+        dict(weights_shape=[1, 3, 3, 3], group=1),
+        dict(weights_shape=[3, 1, 3, 3], group=3)]
+
+    test_data_4D = [
+        dict(weights_shape=[1, 3, 3, 3], group=1),
+        dict(weights_shape=[1, 3, 5, 3], group=1),
+        dict(weights_shape=[3, 1, 3, 3], group=3),
+        dict(weights_shape=[3, 1, 3, 5], group=3)]
+
+    test_data_4D_autopad = [
+        dict(weights_shape=[1, 3, 3, 3], group=1, pads=[1, 1, 1, 1], strides=[1, 1], dilations=[1, 1]),
+        dict(weights_shape=[1, 3, 3, 3], group=1, pads=[2, 2, 2, 2], strides=[1, 1], dilations=[2, 2]),
+        dict(weights_shape=[1, 3, 3, 3], group=1, pads=[3, 5, 3, 5], strides=[1, 1], dilations=[3, 5]),
+        dict(weights_shape=[1, 3, 3, 3], group=1, pads=[1, 1, 1, 1], strides=[2, 2], dilations=[1, 1]),
+        dict(weights_shape=[1, 3, 3, 3], group=1, pads=[2, 2, 2, 2], strides=[2, 2], dilations=[2, 2]),
+        dict(weights_shape=[1, 3, 3, 3], group=1, pads=[3, 5, 3, 5], strides=[2, 2], dilations=[3, 5]),
+        dict(weights_shape=[1, 3, 3, 3], group=1, pads=[1, 0, 1, 0], strides=[3, 5], dilations=[1, 1]),
+        dict(weights_shape=[1, 3, 3, 3], group=1, pads=[2, 0, 2, 0], strides=[3, 5], dilations=[2, 2]),
+        dict(weights_shape=[1, 3, 3, 3], group=1, pads=[3, 3, 3, 3], strides=[3, 5], dilations=[3, 5]),
+        dict(weights_shape=[1, 3, 5, 3], group=1, pads=[2, 1, 2, 1], strides=[1, 1], dilations=[1, 1]),
+        dict(weights_shape=[1, 3, 5, 3], group=1, pads=[4, 2, 4, 2], strides=[1, 1], dilations=[2, 2]),
+        dict(weights_shape=[1, 3, 5, 3], group=1, pads=[6, 5, 6, 5], strides=[1, 1], dilations=[3, 5]),
+        dict(weights_shape=[1, 3, 5, 3], group=1, pads=[2, 1, 2, 1], strides=[2, 2], dilations=[1, 1]),
+        dict(weights_shape=[1, 3, 5, 3], group=1, pads=[4, 2, 4, 2], strides=[2, 2], dilations=[2, 2]),
+        dict(weights_shape=[1, 3, 5, 3], group=1, pads=[6, 5, 6, 5], strides=[2, 2], dilations=[3, 5]),
+        dict(weights_shape=[1, 3, 5, 3], group=1, pads=[2, 0, 2, 0], strides=[3, 5], dilations=[1, 1]),
+        dict(weights_shape=[1, 3, 5, 3], group=1, pads=[4, 0, 4, 0], strides=[3, 5], dilations=[2, 2]),
+        dict(weights_shape=[1, 3, 5, 3], group=1, pads=[6, 3, 6, 3], strides=[3, 5], dilations=[3, 5]),
+        dict(weights_shape=[3, 1, 3, 3], group=3, pads=[1, 1, 1, 1], strides=[1, 1], dilations=[1, 1]),
+        dict(weights_shape=[3, 1, 3, 3], group=3, pads=[2, 2, 2, 2], strides=[1, 1], dilations=[2, 2]),
+        dict(weights_shape=[3, 1, 3, 3], group=3, pads=[3, 5, 3, 5], strides=[1, 1], dilations=[3, 5]),
+        dict(weights_shape=[3, 1, 3, 3], group=3, pads=[1, 1, 1, 1], strides=[2, 2], dilations=[1, 1]),
+        dict(weights_shape=[3, 1, 3, 3], group=3, pads=[2, 2, 2, 2], strides=[2, 2], dilations=[2, 2]),
+        dict(weights_shape=[3, 1, 3, 3], group=3, pads=[3, 5, 3, 5], strides=[2, 2], dilations=[3, 5]),
+        dict(weights_shape=[3, 1, 3, 3], group=3, pads=[1, 0, 1, 0], strides=[3, 5], dilations=[1, 1]),
+        dict(weights_shape=[3, 1, 3, 3], group=3, pads=[2, 0, 2, 0], strides=[3, 5], dilations=[2, 2]),
+        dict(weights_shape=[3, 1, 3, 3], group=3, pads=[3, 3, 3, 3], strides=[3, 5], dilations=[3, 5]),
+        dict(weights_shape=[3, 1, 3, 5], group=3, pads=[1, 2, 1, 2], strides=[1, 1], dilations=[1, 1]),
+        dict(weights_shape=[3, 1, 3, 5], group=3, pads=[2, 4, 2, 4], strides=[1, 1], dilations=[2, 2]),
+        dict(weights_shape=[3, 1, 3, 5], group=3, pads=[3, 10, 3, 10], strides=[1, 1], dilations=[3, 5]),
+        dict(weights_shape=[3, 1, 3, 5], group=3, pads=[1, 2, 1, 2], strides=[2, 2], dilations=[1, 1]),
+        dict(weights_shape=[3, 1, 3, 5], group=3, pads=[2, 4, 2, 4], strides=[2, 2], dilations=[2, 2]),
+        dict(weights_shape=[3, 1, 3, 5], group=3, pads=[3, 10, 3, 10], strides=[2, 2], dilations=[3, 5]),
+        dict(weights_shape=[3, 1, 3, 5], group=3, pads=[1, 0, 1, 0], strides=[3, 5], dilations=[1, 1]),
+        dict(weights_shape=[3, 1, 3, 5], group=3, pads=[2, 2, 2, 2], strides=[3, 5], dilations=[2, 2]),
+        dict(weights_shape=[3, 1, 3, 5], group=3, pads=[3, 8, 3, 8], strides=[3, 5], dilations=[3, 5])]
+
+    test_data_5D_precommit = [
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3)]
+
+    test_data_5D = [
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1),
+        dict(weights_shape=[1, 3, 3, 4, 5], group=1),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3),
+        dict(weights_shape=[3, 1, 5, 4, 3], group=3)]
+
+    test_data_5D_autopad = [
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1, pads=[1, 1, 1, 1, 1, 1], strides=[1, 1, 1], dilations=[1, 1, 1]),
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1, pads=[2, 2, 2, 2, 2, 2], strides=[1, 1, 1], dilations=[2, 2, 2]),
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1, pads=[3, 4, 5, 3, 4, 5], strides=[1, 1, 1], dilations=[3, 4, 5]),
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1, pads=[1, 1, 1, 1, 1, 1], strides=[2, 2, 2], dilations=[1, 1, 1]),
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1, pads=[2, 2, 2, 2, 2, 2], strides=[2, 2, 2], dilations=[2, 2, 2]),
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1, pads=[3, 4, 5, 3, 4, 5], strides=[2, 2, 2], dilations=[3, 4, 5]),
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1, pads=[1, 1, 0, 1, 1, 0], strides=[3, 4, 5], dilations=[1, 1, 1]),
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1, pads=[2, 2, 0, 2, 2, 0], strides=[3, 4, 5], dilations=[2, 2, 2]),
+        dict(weights_shape=[1, 3, 3, 3, 3], group=1, pads=[3, 4, 3, 3, 4, 3], strides=[3, 4, 5], dilations=[3, 4, 5]),
+        dict(weights_shape=[1, 3, 3, 4, 5], group=1, pads=[1, 1, 2, 1, 2, 2], strides=[1, 1, 1], dilations=[1, 1, 1]),
+        dict(weights_shape=[1, 3, 3, 4, 5], group=1, pads=[2, 3, 4, 2, 3, 4], strides=[1, 1, 1], dilations=[2, 2, 2]),
+        dict(weights_shape=[1, 3, 3, 4, 5], group=1, pads=[3, 6, 10, 3, 6, 10], strides=[1, 1, 1], dilations=[3, 4, 5]),
+        dict(weights_shape=[1, 3, 3, 4, 5], group=1, pads=[1, 1, 2, 1, 2, 2], strides=[2, 2, 2], dilations=[1, 1, 1]),
+        dict(weights_shape=[1, 3, 3, 4, 5], group=1, pads=[2, 3, 4, 2, 3, 4], strides=[2, 2, 2], dilations=[2, 2, 2]),
+        dict(weights_shape=[1, 3, 3, 4, 5], group=1, pads=[3, 6, 10, 3, 6, 10], strides=[2, 2, 2], dilations=[3, 4, 5]),
+        dict(weights_shape=[1, 3, 3, 4, 5], group=1, pads=[1, 1, 0, 1, 2, 0], strides=[3, 4, 5], dilations=[1, 1, 1]),
+        dict(weights_shape=[1, 3, 3, 4, 5], group=1, pads=[2, 3, 2, 2, 3, 2], strides=[3, 4, 5], dilations=[2, 2, 2]),
+        dict(weights_shape=[1, 3, 3, 4, 5], group=1, pads=[3, 6, 8, 3, 6, 8], strides=[3, 4, 5], dilations=[3, 4, 5]),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3, pads=[1, 1, 1, 1, 1, 1], strides=[1, 1, 1], dilations=[1, 1, 1]),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3, pads=[2, 2, 2, 2, 2, 2], strides=[1, 1, 1], dilations=[2, 2, 2]),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3, pads=[3, 4, 5, 3, 4, 5], strides=[1, 1, 1], dilations=[3, 4, 5]),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3, pads=[1, 1, 1, 1, 1, 1], strides=[2, 2, 2], dilations=[1, 1, 1]),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3, pads=[2, 2, 2, 2, 2, 2], strides=[2, 2, 2], dilations=[2, 2, 2]),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3, pads=[3, 4, 5, 3, 4, 5], strides=[2, 2, 2], dilations=[3, 4, 5]),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3, pads=[1, 1, 0, 1, 1, 0], strides=[3, 4, 5], dilations=[1, 1, 1]),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3, pads=[2, 2, 0, 2, 2, 0], strides=[3, 4, 5], dilations=[2, 2, 2]),
+        dict(weights_shape=[3, 1, 3, 3, 3], group=3, pads=[3, 4, 3, 3, 4, 3], strides=[3, 4, 5], dilations=[3, 4, 5]),
+        dict(weights_shape=[3, 1, 5, 4, 3], group=3, pads=[2, 1, 1, 2, 2, 1], strides=[1, 1, 1], dilations=[1, 1, 1]),
+        dict(weights_shape=[3, 1, 5, 4, 3], group=3, pads=[4, 3, 2, 4, 3, 2], strides=[1, 1, 1], dilations=[2, 2, 2]),
+        dict(weights_shape=[3, 1, 5, 4, 3], group=3, pads=[6, 6, 5, 6, 6, 5], strides=[1, 1, 1], dilations=[3, 4, 5]),
+        dict(weights_shape=[3, 1, 5, 4, 3], group=3, pads=[2, 1, 1, 2, 2, 1], strides=[2, 2, 2], dilations=[1, 1, 1]),
+        dict(weights_shape=[3, 1, 5, 4, 3], group=3, pads=[4, 3, 2, 4, 3, 2], strides=[2, 2, 2], dilations=[2, 2, 2]),
+        dict(weights_shape=[3, 1, 5, 4, 3], group=3, pads=[6, 6, 5, 6, 6, 5], strides=[2, 2, 2], dilations=[3, 4, 5]),
+        dict(weights_shape=[3, 1, 5, 4, 3], group=3, pads=[2, 1, 0, 2, 2, 0], strides=[3, 4, 5], dilations=[1, 1, 1]),
+        dict(weights_shape=[3, 1, 5, 4, 3], group=3, pads=[4, 3, 0, 4, 3, 0], strides=[3, 4, 5], dilations=[2, 2, 2]),
+        dict(weights_shape=[3, 1, 5, 4, 3], group=3, pads=[6, 6, 3, 6, 6, 3], strides=[3, 4, 5], dilations=[3, 4, 5])]
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.parametrize("dilations", [[1], [2]])
+    @pytest.mark.parametrize("pads", [[0, 0], [1, 1], [1, 2]])
+    @pytest.mark.parametrize("strides", [[1], [2]])
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.nightly
+    def test_conv_3D(self, params, dilations, pads, strides, bias, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, shape=[2, 3, 25], dilations=dilations, pads=pads, strides=strides,
+                                    bias=bias, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D_autopad[:-1])
+    @pytest.mark.parametrize("auto_pad", ['SAME_UPPER', 'SAME_LOWER'])
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.nightly
+    @pytest.mark.xfail(reason='autopad dimetions do not agree with framework')
+    def test_conv_3D_autopad(self, params, auto_pad, bias, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, shape=[2, 3, 25], bias=bias, auto_pad=auto_pad, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D_precommit)
+    @pytest.mark.parametrize("dilations", [[3, 5]])
+    @pytest.mark.parametrize("pads", [[1, 2, 3, 4]])
+    @pytest.mark.parametrize("strides", [[3, 5]])
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.precommit
+    def test_conv_4D_precommit(self, params, dilations, pads, strides, bias, ie_device, precision,
+                               ir_version, temp_dir):
+        self._test(*self.create_net(**params, shape=[2, 3, 25, 25],  dilations=dilations, pads=pads, strides=strides,
+                                    bias=bias, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.parametrize("dilations", [[1, 1], [2, 2], [3, 5]])
+    @pytest.mark.parametrize("pads", [[0, 0, 0, 0], [1, 1, 1, 1], [1, 2, 3, 4]])
+    @pytest.mark.parametrize("strides", [[1, 1], [2, 2], [3, 5]])
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.nightly
+    def test_conv_4D(self, params, dilations, pads, strides, bias, ie_device, precision, ir_version, temp_dir):
+        self._test(
+            *self.create_net(**params, shape=[2, 3, 25, 25], dilations=dilations, pads=pads, strides=strides, bias=bias,
+                             ir_version=ir_version), ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D_autopad[:-1])
+    @pytest.mark.parametrize("auto_pad", ['SAME_UPPER', 'SAME_LOWER'])
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.nightly
+    @pytest.mark.xfail(reason='autopad dimetions do not agree with framework')
+    def test_conv_4D_autopad(self, params, auto_pad, bias, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, shape=[2, 3, 25, 25], bias=bias, auto_pad=auto_pad,
+                                    ir_version=ir_version), ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D_precommit)
+    @pytest.mark.parametrize("dilations", [[3, 4, 5]])
+    @pytest.mark.parametrize("pads", [[1, 2, 3, 4, 5, 6]])
+    @pytest.mark.parametrize("strides", [[3, 4, 5]])
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.precommit
+    def test_conv_5D_precommit(self, params, dilations, pads, strides, bias, ie_device, precision,
+                               ir_version, temp_dir):
+        self._test(*self.create_net(**params, shape=[2, 3, 25, 25, 25], dilations=dilations, pads=pads, strides=strides,
+                                    bias=bias, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.parametrize("dilations", [[1, 1, 1], [2, 2, 2], [3, 4, 5]])
+    @pytest.mark.parametrize("pads", [[0, 0, 0, 0, 0, 0], [1, 1, 1, 1, 1, 1], [1, 2, 3, 4, 5, 6]])
+    @pytest.mark.parametrize("strides", [[1, 1, 1], [2, 2, 2], [3, 4, 5]])
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.nightly
+    def test_conv_5D(self, params, dilations, pads, strides, bias, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, shape=[2, 3, 25, 25, 25], dilations=dilations, pads=pads, strides=strides,
+                                    bias=bias, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D_autopad[:-1])
+    @pytest.mark.parametrize("auto_pad", ['SAME_UPPER', 'SAME_LOWER'])
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.nightly
+    @pytest.mark.xfail(reason='autopad dimetions do not agree with framework')
+    def test_conv_5D_autopad(self, params, auto_pad, bias, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, shape=[2, 3, 25, 25, 25], bias=bias, auto_pad=auto_pad,
+                                    ir_version=ir_version), ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_conv_transpose.py b/tests/layer_tests/onnx_tests/test_conv_transpose.py
new file mode 100644
index 00000000000000..1de48549f3fd1c
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_conv_transpose.py
@@ -0,0 +1,215 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestConvTranspose(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randn(*inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_conv_transpose(self, ir_version, input_shape, output_shape, kernel_shape, strides, group=1,
+                              dilations=None, pads=None, force_output_shape=False, output_padding=None, bias=False,
+                              auto_pad=None):
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        weights = np.random.randn(*kernel_shape).astype(np.float)
+
+        node_weights_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['kernel'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=weights.shape,
+                vals=weights.flatten(),
+            ),
+        )
+
+        conv_attrs = {
+            'strides': strides,
+            'group': group,
+            'kernel_shape': kernel_shape[2:],  # As we have NCHW layout
+        }
+
+        if pads is not None:
+            if not force_output_shape:
+                conv_attrs.update({'pads': pads})
+        else:
+            pads = np.zeros(2 * (len(input_shape) - 2))
+        _pads = np.array(pads).reshape([2, -1])
+        if output_padding is not None:
+            conv_attrs.update({'output_padding': output_padding})
+        if dilations is not None:
+            conv_attrs.update({'dilations': dilations})
+        else:
+            dilations = np.ones(len(input_shape) - 2)
+        if force_output_shape:
+            conv_attrs.update({'output_shape': output_shape[2:]})
+
+        if auto_pad:
+            conv_attrs.update({'auto_pad': auto_pad})
+
+        nodes = [node_weights_def]
+        if bias:
+            bias_const = np.random.randint(-10, 10, kernel_shape[0]).astype(np.float32)
+
+            node_bias_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['bias'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.FLOAT,
+                    dims=bias_const.shape,
+                    vals=bias_const.flatten(),
+                ),
+            )
+            node_conv_transpose = onnx.helper.make_node(
+                'ConvTranspose',
+                inputs=['input', 'kernel', 'bias'],
+                outputs=['output'],
+                **conv_attrs
+            )
+            nodes.extend([node_bias_def, node_conv_transpose])
+        else:
+            node_conv_transpose = onnx.helper.make_node(
+                'ConvTranspose',
+                inputs=['input', 'kernel'],
+                outputs=['output'],
+                **conv_attrs
+            )
+            nodes.append(node_conv_transpose)
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_conv_transpose_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_conv_transpose_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    common_tests_4D_precommit = [
+        pytest.param(dict(input_shape=[1, 3, 10, 10], output_shape=[1, 3, 12, 12],
+                          kernel_shape=[3, 3, 2, 2], strides=[1, 1], dilations=[2, 2]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed")),
+        pytest.param(dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 85, 85],
+                          kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2, output_padding=[1, 1]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed"))
+    ]
+
+    common_tests_4D = [
+        dict(input_shape=[1, 3, 10, 10], output_shape=[1, 3, 11, 11],
+             kernel_shape=[3, 3, 2, 2], strides=[1, 1]),
+        dict(input_shape=[1, 3, 10, 10], output_shape=[1, 3, 20, 20],
+             kernel_shape=[3, 3, 2, 2], strides=[2, 2]),
+        dict(input_shape=[1, 3, 10, 10], output_shape=[1, 3, 12, 12],
+             kernel_shape=[3, 3, 2, 2], strides=[1, 1], dilations=[2, 2]),
+        dict(input_shape=[1, 3, 10, 10], output_shape=[1, 3, 21, 21],
+             kernel_shape=[3, 3, 2, 2], strides=[2, 2], dilations=[2, 2]),
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 85, 85],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2, output_padding=[1, 1]),
+    ]
+
+    explicit_pads_tests_4D = common_tests_4D + [
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 80, 80],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2, pads=[2, 2, 2, 2]),
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 87, 87],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2, pads=[2, 2, 2, 2], dilations=[2, 2]),
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 80, 80],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2, pads=[2, 2, 2, 2], force_output_shape=True),
+    ]
+
+    valid_auto_pad_tests_4D = common_tests_4D + [
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 84, 84],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2),
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 91, 91],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2, dilations=[2, 2]),
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 80, 80],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2, force_output_shape=True),
+    ]
+
+    same_auto_pad_tests_4D = [
+        dict(input_shape=[1, 3, 10, 10], output_shape=[1, 3, 10, 10],
+             kernel_shape=[3, 3, 2, 2], strides=[1, 1]),
+        dict(input_shape=[1, 3, 10, 10], output_shape=[1, 3, 20, 20],
+             kernel_shape=[3, 3, 2, 2], strides=[2, 2]),
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 80, 80],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2),
+        dict(input_shape=[1, 3, 10, 10], output_shape=[1, 3, 10, 10],
+             kernel_shape=[3, 3, 2, 2], strides=[1, 1], dilations=[2, 2]),
+        dict(input_shape=[1, 3, 10, 10], output_shape=[1, 3, 20, 20],
+             kernel_shape=[3, 3, 2, 2], strides=[2, 2], dilations=[2, 2]),
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 80, 80],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2, dilations=[2, 2]),
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 80, 80],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2, force_output_shape=True),
+        dict(input_shape=[1, 2, 20, 20], output_shape=[1, 2, 81, 81],
+             kernel_shape=[2, 1, 8, 8], strides=[4, 4], group=2, output_padding=[1, 1]),
+    ]
+
+    @pytest.mark.parametrize("params", common_tests_4D_precommit)
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.parametrize("auto_pad", ["NOTSET"])
+    @pytest.mark.precommit
+    def test_conv_transpose_4D_precommit(self, params, bias, ie_device, precision, ir_version, auto_pad, temp_dir):
+        if ie_device == 'GPU' and 'dilations' in params:
+            pytest.xfail('dilations are not supported on GPU')
+        self._test(*self.create_conv_transpose(**params, ir_version=ir_version, bias=bias, auto_pad=auto_pad),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", explicit_pads_tests_4D)
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.parametrize("auto_pad", ["NOTSET"])
+    @pytest.mark.nightly
+    def test_conv_transpose_4D(self, params, bias, ie_device, precision, ir_version, auto_pad, temp_dir):
+        if ie_device == 'GPU' and 'dilations' in params:
+            pytest.xfail('dilations are not supported on GPU')
+        self._test(*self.create_conv_transpose(**params, ir_version=ir_version, bias=bias, auto_pad=auto_pad),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", valid_auto_pad_tests_4D)
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.parametrize("auto_pad", ["VALID"])
+    @pytest.mark.nightly
+    def test_conv_transpose_valid_auto_pad_4D(self, params, bias, ie_device, precision, ir_version, auto_pad, temp_dir):
+        if ie_device == 'GPU' and 'dilations' in params:
+            pytest.xfail('dilations are not supported on GPU')
+        self._test(*self.create_conv_transpose(**params, ir_version=ir_version, bias=bias, auto_pad=auto_pad),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", same_auto_pad_tests_4D)
+    @pytest.mark.parametrize("bias", [False, True])
+    @pytest.mark.parametrize("auto_pad", ["SAME_UPPER", "SAME_LOWER"])
+    @pytest.mark.nightly
+    def test_conv_transpose_same_auto_pad_4D(self, params, bias, ie_device, precision, ir_version, auto_pad, temp_dir):
+        if ie_device == 'GPU' and 'dilations' in params:
+            pytest.xfail('dilations are not supported on GPU')
+        self._test(*self.create_conv_transpose(**params, ir_version=ir_version, bias=bias, auto_pad=auto_pad),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_cumsum.py b/tests/layer_tests/onnx_tests/test_cumsum.py
new file mode 100644
index 00000000000000..b49f00f887c3ff
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_cumsum.py
@@ -0,0 +1,267 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+def cumsum(a, axis=None, exclusive=False, reverse=False):
+    if reverse:
+        a = np.flip(a, axis)
+    res = np.cumsum(a, axis=axis)
+    if exclusive:
+        res -= a
+    if reverse:
+        res = np.flip(res, axis)
+    return res
+
+
+class TestCumSum(OnnxRuntimeLayerTest):
+    def create_net(self, shape, ir_version, axis=None, reverse=None, exclusive=None):
+        """
+            ONNX net                      IR net
+
+            Input->CumSum->Output   =>    Input->CumSum
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        nodes = []
+        inputs = ['input']
+        if axis is not None:
+            node_axis_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['axis'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.INT64,
+                    dims=[],
+                    vals=[axis],
+                ),
+            )
+            nodes.append(node_axis_def)
+            inputs.append('axis')
+
+        args = dict()
+        if exclusive is not None:
+            args['exclusive'] = exclusive
+        if reverse is not None:
+            args['reverse'] = reverse
+        node_def = onnx.helper.make_node(
+            'CumSum',
+            inputs=inputs,
+            outputs=['output'],
+            **args
+        )
+        nodes.append(node_def)
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+        onnx.checker.check_model(onnx_net)
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'CumSum'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            if exclusive is not None:
+                nodes_attributes['node']['exclusive'] = exclusive
+            if reverse is not None:
+                nodes_attributes['node']['reverse'] = reverse
+            edges = [('input', 'input_data'),
+                     ('input_data', 'node'),
+                     ('node', 'node_data'),
+                     ('node_data', 'result')
+                     ]
+            if axis is not None:
+                nodes_attributes.update({
+                    'input_axis_data': {'kind': 'data', 'value': [axis]},
+                    'axis': {'kind': 'op', 'type': 'Const'},
+                    'axis_data': {'shape': [], 'kind': 'data'}})
+                edges.extend([('input_axis_data', 'axis'),
+                              ('axis', 'axis_data'),
+                              ('axis_data', 'node')])
+            ref_net = build_graph(nodes_attributes, edges)
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version, axis=None, reverse=None, exclusive=None):
+        """
+            ONNX net                                     IR net
+
+            Input->Concat(+cumsum const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.randn(*shape).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        nodes = [node_const_def]
+        inputs = ['const1']
+        if axis is not None:
+            node_axis_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['axis'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.INT64,
+                    dims=[],
+                    vals=[axis],
+                ),
+            )
+            nodes.append(node_axis_def)
+            inputs.append('axis')
+
+        args = dict()
+        if exclusive is not None:
+            args['exclusive'] = exclusive
+        if reverse is not None:
+            args['reverse'] = reverse
+        node_def = onnx.helper.make_node(
+            'CumSum',
+            inputs=inputs,
+            outputs=['cumsum'],
+            **args
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'cumsum'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+        nodes.extend([node_def, node_concat_def])
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+        onnx.checker.check_model(onnx_net)
+
+        #
+        #   Create reference IR net
+        #
+        constant = cumsum(constant, axis=axis, reverse=reverse, exclusive=exclusive)
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data = [
+        dict(shape=[1]),
+        dict(shape=[1, 2]),
+        dict(shape=[2, 4, 6]),
+        dict(shape=[2, 4, 6, 8]),
+        dict(shape=[2, 4, 6, 8, 10]),
+        dict(shape=[1, 2], axis=-2),
+        dict(shape=[1, 2], axis=1),
+        dict(shape=[2, 4, 6], axis=-3),
+        dict(shape=[2, 4, 6], axis=2),
+        dict(shape=[2, 4, 6, 8], axis=-4),
+        dict(shape=[2, 4, 6, 8], axis=3),
+        dict(shape=[2, 4, 6, 8, 10], axis=-1),
+        dict(shape=[2, 4, 6, 8, 10], axis=4)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("reverse", [0, 1])
+    @pytest.mark.parametrize("exclusive", [0, 1])
+    @pytest.mark.nightly
+    def test_cumsum(self, params, reverse, exclusive, ie_device, precision, ir_version, temp_dir):
+        if 'axis' not in params:
+            pytest.skip('No axis cases fail in ONNX')
+        self._test(*self.create_net(**params, exclusive=exclusive, reverse=reverse, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("reverse", [0, 1])
+    @pytest.mark.parametrize("exclusive", [0, 1])
+    @pytest.mark.nightly
+    def test_cumsum_const(self, params, reverse, exclusive, ie_device, precision, ir_version, temp_dir):
+        if 'axis' not in params:
+            pytest.skip('No axis cases fail in ONNX')
+        self._test(*self.create_net_const(**params, precision=precision, exclusive=exclusive, reverse=reverse,
+                                          ir_version=ir_version), ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_dequantize_linear.py b/tests/layer_tests/onnx_tests/test_dequantize_linear.py
new file mode 100644
index 00000000000000..1bf364accff938
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_dequantize_linear.py
@@ -0,0 +1,205 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestDequantizeLinear(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randint(np.iinfo(self.inp_type).min,
+                                                   np.iinfo(self.inp_type).max,
+                                                   inputs_dict[input],
+                                                   dtype=self.inp_type)
+        return inputs_dict
+
+    def create_dequanize_linear(self, shape, y_scale: np.array, y_zero_point=None, axis=None, opset=10, ir_version='10'):
+        """
+            ONNX net                              IR net
+
+            Input->DequantizeLinear->Output   =>    Input->Sub->Mul
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        self.inp_type = y_zero_point.dtype if y_zero_point is not None else np.uint8
+        onnx_type = TensorProto.UINT8 if self.inp_type == np.uint8 else TensorProto.INT8
+        input = helper.make_tensor_value_info('input', onnx_type, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        nodes = []
+        inputs = ['input', 'y_scale']
+        node_scale_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['y_scale'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=y_scale.shape,
+                vals=y_scale.flatten(),
+            ),
+        )
+        nodes.append(node_scale_def)
+
+        if y_zero_point is not None:
+            node_zero_point_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['y_zero_point'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=onnx_type,
+                    dims=y_zero_point.shape,
+                    vals=y_zero_point.flatten(),
+                ),
+            )
+            inputs.append('y_zero_point')
+            nodes.append(node_zero_point_def)
+        args = dict()
+        if axis is not None:
+            args['axis'] = axis
+        node_def = onnx.helper.make_node(
+            'DequantizeLinear',
+            inputs=inputs,
+            outputs=['output'],
+            **args
+        )
+        nodes.append(node_def)
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model',
+                                     opset_imports=[helper.make_opsetid("", opset)])
+        onnx.checker.check_model(onnx_net)
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        nodes_attributes = {
+            'input': {'kind': 'op', 'type': 'Parameter'},
+            'input_data': {'shape': shape, 'kind': 'data'},
+            'input_scale_data': {'kind': 'data', 'value': y_scale},
+            'scale_const': {'kind': 'op', 'type': 'Const'},
+            'scale_data': {'shape': np.ones(len(shape)), 'kind': 'data'},
+            'mul': {'kind': 'op', 'type': 'Multiply'},
+            'mul_data': {'shape': shape, 'kind': 'data'},
+            'result': {'kind': 'op', 'type': 'Result'}
+        }
+        edges = [('input', 'input_data'),
+                 ('input_data', 'mul'),
+                 ('input_scale_data', 'scale_const'),
+                 ('scale_const', 'scale_data'),
+                 ('scale_data', 'mul'),
+                 ('mul', 'mul_data')]
+        if y_zero_point is not None:
+            nodes_attributes.update({
+                'input_zero_data': {'kind': 'data', 'value': -y_scale * y_zero_point},
+                'zero_const': {'kind': 'op', 'type': 'Const'},
+                'zero_data': {'shape': np.ones(len(shape)), 'kind': 'data'},
+                'sub': {'kind': 'op', 'type': 'Add'},
+                'sub_data': {'shape': shape, 'kind': 'data'},
+            })
+            edges.extend([('mul_data', 'sub'),
+                          ('input_zero_data', 'zero_const'),
+                          ('zero_const', 'zero_data'),
+                          ('zero_data', 'sub'),
+                          ('sub', 'sub_data'),
+                          ('sub_data', 'result')])
+        else:
+            edges.append(('mul_data', 'result'))
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            ref_net = build_graph(nodes_attributes, edges)
+
+        return onnx_net, ref_net
+
+    test_data = [
+        dict(shape=[8], y_scale=np.array(2, dtype=np.float), y_zero_point=np.array(128, dtype=np.uint8)),
+        dict(shape=[8], y_scale=np.array(2, dtype=np.float), y_zero_point=np.array(1, dtype=np.int8)),
+        dict(shape=[2, 4], y_scale=np.array(2, dtype=np.float), y_zero_point=np.array(128, dtype=np.uint8)),
+        dict(shape=[2, 4], y_scale=np.array(2, dtype=np.float), y_zero_point=np.array(1, dtype=np.int8)),
+        dict(shape=[2, 4, 6], y_scale=np.array(2, dtype=np.float), y_zero_point=np.array(128, dtype=np.uint8)),
+        dict(shape=[2, 4, 6], y_scale=np.array(2, dtype=np.float), y_zero_point=np.array(1, dtype=np.int8)),
+        dict(shape=[2, 4, 6, 8], y_scale=np.array(2, dtype=np.float), y_zero_point=np.array(128, dtype=np.uint8)),
+        dict(shape=[2, 4, 6, 8], y_scale=np.array(2, dtype=np.float), y_zero_point=np.array(1, dtype=np.int8)),
+        dict(shape=[2, 4, 6, 8, 10], y_scale=np.array(2, dtype=np.float), y_zero_point=np.array(128, dtype=np.uint8)),
+        dict(shape=[2, 4, 6, 8, 10], y_scale=np.array(2, dtype=np.float), y_zero_point=np.array(1, dtype=np.int8)),
+    ]
+    test_data_def_zerop = [
+        dict(shape=[8], y_scale=np.array(2, dtype=np.float)),
+        dict(shape=[2, 4], y_scale=np.array(2, dtype=np.float)),
+        dict(shape=[2, 4, 6], y_scale=np.array(2, dtype=np.float)),
+        dict(shape=[2, 4, 6, 8], y_scale=np.array(2, dtype=np.float)),
+        dict(shape=[2, 4, 6, 8, 10], y_scale=np.array(2, dtype=np.float)),
+    ]
+
+    test_data_axis = [
+        dict(shape=[2, 4], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float), axis=1),
+        dict(shape=[2, 4], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float),
+             y_zero_point=np.array([128, 128, 128, 128], dtype=np.uint8), axis=1),
+        dict(shape=[2, 4], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float),
+             y_zero_point=np.array([1, 1, 1, 1], dtype=np.int8), axis=1),
+        dict(shape=[2, 4, 6], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float), axis=1),
+        dict(shape=[2, 4, 6], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float),
+             y_zero_point=np.array([128, 128, 128, 128], dtype=np.uint8), axis=1),
+        dict(shape=[2, 4, 6], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float),
+             y_zero_point=np.array([1, 1, 1, 1], dtype=np.int8), axis=1),
+        dict(shape=[2, 4, 6, 8], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float), axis=1),
+        dict(shape=[2, 4, 6, 8], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float),
+             y_zero_point=np.array([128, 128, 128, 128], dtype=np.uint8), axis=1),
+        dict(shape=[2, 4, 6, 8], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float),
+             y_zero_point=np.array([1, 1, 1, 1], dtype=np.int8), axis=1),
+        dict(shape=[2, 4, 6, 8, 10], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float), axis=1),
+        dict(shape=[2, 4, 6, 8, 10], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float),
+             y_zero_point=np.array([128, 128, 128, 128], dtype=np.uint8), axis=1),
+        dict(shape=[2, 4, 6, 8, 10], y_scale=np.array([2, 2.5, 3, 2.3], dtype=np.float),
+             y_zero_point=np.array([1, 1, 1, 1], dtype=np.int8), axis=1),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_def_zerop)
+    @pytest.mark.nightly
+    @pytest.mark.xfail(reason='Defualt zero_point fails on onnxruntime')
+    def test_quantize_linear_def_zerop_opset10(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_dequanize_linear(**params, ir_version=ir_version), ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_quantize_linear_opset10(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_dequanize_linear(**params, ir_version=ir_version), ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data + test_data_def_zerop)
+    @pytest.mark.nightly
+    @pytest.mark.skip(reason='DequantizeLinear-13 is unsupported in MO')
+    def test_quantize_linear_opset13(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_dequanize_linear(**params, opset=13, ir_version=ir_version), ie_device, precision,
+                   ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_axis)
+    @pytest.mark.nightly
+    @pytest.mark.skip(reason='DequantizeLinear-13 is unsupported in MO')
+    def test_quantize_linear_axis_opset13(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_dequanize_linear(**params, opset=13, ir_version=ir_version), ie_device, precision,
+                   ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_dropout.py b/tests/layer_tests/onnx_tests/test_dropout.py
new file mode 100644
index 00000000000000..618fdc02d32aac
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_dropout.py
@@ -0,0 +1,169 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+
+
+class TestDropout(Caffe2OnnxLayerTest):
+    def create_net(self, shape, ratio, ir_version, opset=None):
+        """
+            ONNX net                                IR net
+
+            Input->Dropout->Sigmoid->Output   =>    Input->sigmoid
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        args = dict()
+        if ratio:
+            args['ratio'] = ratio
+        if opset == 6:
+            args['is_test'] = 1
+        node_def = helper.make_node(
+            'Dropout',
+            inputs=['input'],
+            outputs=['dropout'],
+            **args
+        )
+
+        sigmoid_def = helper.make_node(
+            'Sigmoid',
+            inputs=['dropout'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def, sigmoid_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, ratio, ir_version, opset=None):
+        """
+            ONNX net                                           IR net
+
+            Input->Concat(+dropout with const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        constant = np.random.randint(-127, 127, shape).astype(np.float)
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_const_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        args = dict()
+        if ratio:
+            args['ratio'] = ratio
+        if opset == 6:
+            args['is_test'] = 1
+        node_def = helper.make_node(
+            'Dropout',
+            inputs=['const1'],
+            outputs=['dropout'],
+            **args
+        )
+
+        node_concat_def = helper.make_node(
+            'Concat',
+            inputs=['input', 'dropout'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[10, 12], ratio=None),
+                 dict(shape=[10, 12], ratio=0.7),
+                 dict(shape=[8, 10, 12], ratio=None),
+                 dict(shape=[8, 10, 12], ratio=0.7),
+                 dict(shape=[6, 8, 10, 12], ratio=None),
+                 dict(shape=[6, 8, 10, 12], ratio=0.7),
+                 dict(shape=[4, 6, 8, 10, 12], ratio=None),
+                 dict(shape=[4, 6, 8, 10, 12], ratio=0.7)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_dropout_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, opset=6, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_dropout(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_dropout_const_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, opset=6, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_dropout_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_elu.py b/tests/layer_tests/onnx_tests/test_elu.py
new file mode 100644
index 00000000000000..c65b000b0826f6
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_elu.py
@@ -0,0 +1,185 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestElu(Caffe2OnnxLayerTest):
+    def create_net(self, shape, alpha, ir_version):
+        """
+            ONNX net                   IR net
+
+            Input->Elu->Output   =>    Input->elu
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Elu',
+            inputs=['input'],
+            outputs=['output'],
+            alpha=alpha
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'Elu', 'alpha': alpha},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, alpha, precision, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->Concat(+elu const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        const_number = np.prod(shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+        constant = np.reshape(constant, shape)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Elu',
+            inputs=['const1'],
+            outputs=['elu1'],
+            alpha=alpha
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'elu1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.clip(constant, 0, np.inf) + (np.exp(np.clip(constant, -np.inf, 0)) - 1) * alpha
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[10, 12], alpha=0.1),
+                 dict(shape=[8, 10, 12], alpha=0.9),
+                 dict(shape=[6, 8, 10, 12], alpha=1.5),
+                 dict(shape=[4, 6, 8, 10, 12], alpha=4.5)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_elu(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_elu_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_embedding_bag.py b/tests/layer_tests/onnx_tests/test_embedding_bag.py
new file mode 100644
index 00000000000000..dce555c8261a33
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_embedding_bag.py
@@ -0,0 +1,131 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+
+import numpy as np
+import pytest
+import torch
+import torch.nn as nn
+from common.layer_test_class import CommonLayerTest, check_ir_version
+from unit_tests.utils.graph import build_graph
+
+
+class PytorchLayerTest(CommonLayerTest):
+    def produce_model_path(self, framework_model, save_path):
+        path = os.path.join(save_path, 'model.onnx')
+        self.torch_model = framework_model
+        torch.onnx.export(self.torch_model, self.var, path, output_names=['output'])
+        assert os.path.isfile(path), "model.onnx haven't been saved here: {}".format(save_path)
+        return path
+
+    def get_framework_results(self, inputs_dict, model_path):
+        return {'output': self.torch_model(*self.var).detach().numpy()}
+
+
+class EmbeddingBagModel(torch.nn.Module):
+    def __init__(self, n, m, indices_shape=None, per_sample_weights=False, mode="sum"):
+        super(EmbeddingBagModel, self).__init__()
+        EE = nn.EmbeddingBag(n, m, mode=mode, sparse=True)
+        self.W = np.random.uniform(low=-np.sqrt(1 / n), high=np.sqrt(1 / n), size=(n, m)).astype(np.float32)
+        EE.weight.data = torch.tensor(self.W, requires_grad=True)
+        self.embedding_bag = EE
+        if per_sample_weights:
+            self.per_sample_weights = torch.randn(indices_shape)
+        else:
+            self.per_sample_weights = None
+
+
+class TestPytorchEmbeddingBag(PytorchLayerTest):
+    def _prepare_input(self, inputs_dict):
+        assert 'input' in inputs_dict and 'offsets' in inputs_dict, "input and offsets should be in inputs_dict"
+        indices, offsets = self.var
+        inputs_dict['input'] = indices.numpy().astype(np.int32)
+        inputs_dict['offsets'] = offsets.numpy().astype(np.int32)
+        return inputs_dict
+
+    def create_net(self, n, m, emb_batch_size, ir_version, per_sample_weights=False, offsets=None):
+        """
+            Pytorch net                         IR net
+
+            Input->EmbeddingBag->Output   =>    Input->Gather/SparseWeightedSum
+
+        """
+        #   Create Pytorch model
+        EE = EmbeddingBagModel(n, m, indices_shape=[emb_batch_size], per_sample_weights=per_sample_weights)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input_weights_data': {'kind': 'data', 'value': EE.W.flatten()},
+                'weights': {'kind': 'op', 'type': 'Const'},
+                'weights_data': {'shape': EE.W.shape, 'kind': 'data'},
+                'indices': {'kind': 'op', 'type': 'Parameter'},
+                'indices_data': {'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'EmbeddingBagOffsetsSum'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            edges = [('input_weights_data', 'weights'),
+                     ('weights', 'weights_data'),
+                     ('indices', 'indices_data'),
+                     ('weights_data', 'node'),
+                     ('node', 'node_data'),
+                     ('node_data', 'result')
+                     ]
+            if offsets is not None:
+                nodes_attributes.update({
+                    'offsets': {'kind': 'op', 'type': 'Parameter'},
+                    'offsets_data': {'kind': 'data'},
+                    'node_data': {'shape': [len(offsets), m], 'kind': 'data'},
+                })
+                edges.extend([
+                    ('offsets', 'offsets_data'),
+                    ('indices_data', 'node'),
+                    ('offsets_data', 'node'),
+                ])
+            else:
+                nodes_attributes.update({
+                    'input_shape_data': {'kind': 'data', 'value': [-1]},
+                    'shape': {'kind': 'op', 'type': 'Const'},
+                    'shape_data': {'shape': [1], 'kind': 'data'},
+                    'reshape': {'kind': 'op', 'type': 'Reshape'},
+                    'reshape_data': {'shape': [emb_batch_size], 'kind': 'data'},
+                    'input_offsets_data': {'kind': 'data', 'value': np.arange(0, 128, 2)},
+                    'offsets': {'kind': 'op', 'type': 'Const'},
+                    'offsets_data': {'shape': [int(emb_batch_size / 2)], 'kind': 'data'},
+                    'node_data': {'shape': [int(emb_batch_size / 2), m], 'kind': 'data'},
+                })
+                edges.extend([
+                    ('input_shape_data', 'shape'),
+                    ('shape', 'shape_data'),
+                    ('indices_data', 'reshape'),
+                    ('shape_data', 'reshape'),
+                    ('reshape', 'reshape_data'),
+                    ('reshape_data', 'node'),
+                    ('input_offsets_data', 'offsets'),
+                    ('offsets', 'offsets_data'),
+                    ('offsets_data', 'node'),
+                ])
+
+            ref_net = build_graph(nodes_attributes, edges)
+        if offsets is not None:
+            self.var = (torch.from_numpy(np.random.choice(n, emb_batch_size)).long(),
+                        torch.from_numpy(np.array(offsets)).long())
+        else:
+            self.var = (
+                torch.from_numpy(np.random.choice(n, emb_batch_size).reshape(int(emb_batch_size / 2), 2)).long(),)
+        return EE, ref_net
+
+    test_data = [
+        dict(n=1460, m=16, emb_batch_size=128),
+        dict(n=1460, m=16, emb_batch_size=128, offsets=np.arange(0, 128)),
+        dict(n=1460, m=16, emb_batch_size=128, offsets=[0, 2, 6, 20, 80]),
+        # dict(n=1460, m=16, emb_batch_size=128, offsets=[0, 2, 6, 20, 80], per_sample_weights=True),
+        # per_sample_weights not supported in ONNX
+        dict(n=1460, m=16, emb_batch_size=128, offsets=[0, 2, 6, 20, 20, 80])  # empty bag case
+    ]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_pytorch_embedding_bag(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params), ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_flatten.py b/tests/layer_tests/onnx_tests/test_flatten.py
new file mode 100644
index 00000000000000..bc24e300c20693
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_flatten.py
@@ -0,0 +1,302 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestFlatten(OnnxRuntimeLayerTest):
+    def create_flatten_net(self, axis, input_shape, dim, ir_version, opset=None):
+        """
+            ONNX net                       IR net
+
+            Input->Flatten->Output   =>    Input->Reshape
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        # TODO: possible move all imports to separate func?
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, dim)
+
+        node_flatten_def = onnx.helper.make_node(
+            'Flatten',
+            inputs=['input'],
+            outputs=['output'],
+            axis=axis,
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_flatten_def],
+            'test_flatten_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #   Please, spesify 'type': 'Input' for inpit node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_flatten_net_const(self, axis, input_shape, dim, ir_version, opset=None):
+        """
+            ONNX net                               IR net
+
+            Input->Flatten->Concat->Output   =>    Input->Concat
+                     Input-'                       Const-'
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        concat_output_shape = dim.copy()
+        concat_output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, dim)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, concat_output_shape)
+
+        const_number = np.prod(input_shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=input_shape,
+                vals=constant,
+            ),
+        )
+
+        node_flatten_def = onnx.helper.make_node(
+            'Flatten',
+            inputs=['const'],
+            outputs=['flatten_output'],
+            axis=axis,
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'flatten_output'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_flatten_def, node_concat_def],
+            'test_flatten_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #   Please, spesify 'type': 'Input' for inpit node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data_3D = [
+        dict(axis=0, input_shape=[1, 3, 224], dim=[1, 672]),
+        dict(axis=-3, input_shape=[1, 3, 224], dim=[1, 672]),
+        dict(axis=1, input_shape=[1, 3, 224], dim=[1, 672]),
+        dict(axis=-2, input_shape=[1, 3, 224], dim=[1, 672]),
+        dict(axis=2, input_shape=[2, 3, 224], dim=[6, 224]),
+        dict(axis=-1, input_shape=[2, 3, 224], dim=[6, 224]),
+        dict(axis=3, input_shape=[3, 3, 224], dim=[2016, 1])
+    ]
+
+    test_data_4D_precommit = [
+        dict(axis=1, input_shape=[1, 3, 224, 224], dim=[1, 150528]),
+        dict(axis=-3, input_shape=[1, 3, 224, 224], dim=[1, 150528])
+    ]
+
+    test_data_4D = [
+        dict(axis=0, input_shape=[1, 3, 224, 224], dim=[1, 150528]),
+        dict(axis=-4, input_shape=[1, 3, 224, 224], dim=[1, 150528]),
+        dict(axis=1, input_shape=[1, 3, 224, 224], dim=[1, 150528]),
+        dict(axis=-3, input_shape=[1, 3, 224, 224], dim=[1, 150528]),
+        dict(axis=2, input_shape=[2, 3, 224, 224], dim=[6, 50176]),
+        dict(axis=-2, input_shape=[2, 3, 224, 224], dim=[6, 50176]),
+        dict(axis=3, input_shape=[3, 3, 224, 224], dim=[2016, 224]),
+        dict(axis=-1, input_shape=[3, 3, 224, 224], dim=[2016, 224]),
+        dict(axis=4, input_shape=[4, 3, 224, 224], dim=[602112, 1])
+    ]
+
+    test_data_5D_precommit = [
+        dict(axis=-5, input_shape=[1, 3, 9, 224, 224], dim=[1, 1354752]),
+        dict(axis=5, input_shape=[4, 3, 9, 224, 224], dim=[5419008, 1])]
+
+    test_data_5D = [
+        dict(axis=0, input_shape=[1, 3, 9, 224, 224], dim=[1, 1354752]),
+        dict(axis=-5, input_shape=[1, 3, 9, 224, 224], dim=[1, 1354752]),
+        dict(axis=1, input_shape=[1, 3, 9, 224, 224], dim=[1, 1354752]),
+        dict(axis=-4, input_shape=[1, 3, 9, 224, 224], dim=[1, 1354752]),
+        dict(axis=2, input_shape=[2, 3, 9, 224, 224], dim=[6, 451584]),
+        dict(axis=-3, input_shape=[2, 3, 9, 224, 224], dim=[6, 451584]),
+        dict(axis=3, input_shape=[3, 3, 9, 224, 224], dim=[81, 50176]),
+        dict(axis=-2, input_shape=[3, 3, 9, 224, 224], dim=[81, 50176]),
+        dict(axis=4, input_shape=[3, 3, 9, 224, 224], dim=[18144, 224]),
+        dict(axis=-1, input_shape=[3, 3, 9, 224, 224], dim=[18144, 224]),
+        dict(axis=5, input_shape=[4, 3, 9, 224, 224], dim=[5419008, 1])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.parametrize("opset", [6, 9])
+    @pytest.mark.nightly
+    def test_flatten_3D(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        # negative axis not allowed by onnx spec for flatten-1 and flatten-9
+        if params['axis'] < 0:
+            self.skip_framework = True
+        else:
+            self.skip_framework = False
+        self._test(*self.create_flatten_net(**params, ir_version=ir_version, opset=opset),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.parametrize("opset", [6, 9])
+    @pytest.mark.nightly
+    def test_flatten_3D_const(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        # negative axis not allowed by onnx spec for flatten-1 and flatten-9
+        if params['axis'] < 0:
+            self.skip_framework = True
+        else:
+            self.skip_framework = False
+        self._test(*self.create_flatten_net_const(**params, ir_version=ir_version, opset=opset),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.parametrize("opset", [6, 9])
+    @pytest.mark.nightly
+    def test_flatten_4D(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        # negative axis not allowed by onnx spec for flatten-1 and flatten-9
+        if params['axis'] < 0:
+            self.skip_framework = True
+        else:
+            self.skip_framework = False
+        self._test(*self.create_flatten_net(**params, ir_version=ir_version, opset=opset),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D_precommit)
+    @pytest.mark.parametrize("opset", [6, 9])
+    @pytest.mark.precommit
+    def test_flatten_4D_precommit(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        # negative axis not allowed by onnx spec for flatten-1 and flatten-9
+        if params['axis'] < 0:
+            self.skip_framework = True
+        else:
+            self.skip_framework = False
+        self._test(*self.create_flatten_net(**params, ir_version=ir_version, opset=opset),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D_precommit)
+    @pytest.mark.parametrize("opset", [6, 9])
+    @pytest.mark.nightly
+    def test_flatten_4D_const_precommit(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        # negative axis not allowed by onnx spec for flatten-1 and flatten-9
+        if params['axis'] < 0:
+            self.skip_framework = True
+        else:
+            self.skip_framework = False
+        self._test(*self.create_flatten_net_const(**params, ir_version=ir_version, opset=opset),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.parametrize("opset", [6, 9])
+    @pytest.mark.nightly
+    def test_flatten_4D_const(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        # negative axis not allowed by onnx spec for flatten-1 and flatten-9
+        if params['axis'] < 0:
+            self.skip_framework = True
+        else:
+            self.skip_framework = False
+        self._test(*self.create_flatten_net_const(**params, ir_version=ir_version, opset=opset),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D_precommit)
+    @pytest.mark.parametrize("opset", [6, 9])
+    @pytest.mark.nightly
+    def test_flatten_5D_precommit(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        # negative axis not allowed by onnx spec for flatten-1 and flatten-9
+        if params['axis'] < 0:
+            self.skip_framework = True
+        else:
+            self.skip_framework = False
+        self._test(*self.create_flatten_net(**params, ir_version=ir_version, opset=opset),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.parametrize("opset", [6, 9])
+    @pytest.mark.nightly
+    def test_flatten_5D(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        # negative axis not allowed by onnx spec for flatten-1 and flatten-9
+        if params['axis'] < 0:
+            self.skip_framework = True
+        else:
+            self.skip_framework = False
+        self._test(*self.create_flatten_net(**params, ir_version=ir_version, opset=opset),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D_precommit)
+    @pytest.mark.parametrize("opset", [6, 9])
+    @pytest.mark.nightly
+    def test_flatten_5D_const_precommit(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        # negative axis not allowed by onnx spec for flatten-1 and flatten-9
+        if params['axis'] < 0:
+            self.skip_framework = True
+        else:
+            self.skip_framework = False
+        self._test(*self.create_flatten_net_const(**params, ir_version=ir_version, opset=opset),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.parametrize("opset", [6, 9])
+    @pytest.mark.nightly
+    def test_flatten_5D_const(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        # negative axis not allowed by onnx spec for flatten-1 and flatten-9
+        if params['axis'] < 0:
+            self.skip_framework = True
+        else:
+            self.skip_framework = False
+        self._test(*self.create_flatten_net_const(**params, ir_version=ir_version, opset=opset),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_floor.py b/tests/layer_tests/onnx_tests/test_floor.py
new file mode 100644
index 00000000000000..2d6be9c7262854
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_floor.py
@@ -0,0 +1,179 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestFloor(OnnxRuntimeLayerTest):
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                     IR net
+
+            Input->Floor->Output   =>    Input->Floor
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Floor',
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'Floor'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                      IR net
+
+            Input->Concat(+floored const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.randn(*shape).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Floor',
+            inputs=['const1'],
+            outputs=['floor']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'floor'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.floor(constant)
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_floor(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_floor_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_gather.py b/tests/layer_tests/onnx_tests/test_gather.py
new file mode 100644
index 00000000000000..c4c0ee9c12e179
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_gather.py
@@ -0,0 +1,261 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestGather(OnnxRuntimeLayerTest):
+    def create_net(self, shape, axis, indices, output_shape, ir_version):
+        """
+            ONNX net                   IR net
+
+            Input->Gather->Output   =>    Input->Gather
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        indices = np.array(indices)
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_indices_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['indices'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.INT64,
+                dims=indices.shape,
+                vals=indices.flatten(),
+            ),
+        )
+
+        args = dict()
+        if axis:
+            args['axis'] = axis
+        else:
+            axis = 0
+        node_def = onnx.helper.make_node(
+            'Gather',
+            inputs=['input', 'indices'],
+            outputs=['output'],
+            **args
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_indices_def, node_def],
+            'test_model',
+            [input],
+            [output]
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': indices.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': indices.shape, 'kind': 'data', 'value': None},
+                'input_axis_const_data': {'kind': 'data', 'value': [axis]},
+                'axis_const': {'kind': 'op', 'type': 'Const'},
+                'axis_const_data': {'shape': [], 'kind': 'data', 'value': None},
+                'node': {'kind': 'op', 'type': 'Gather'},
+                'node_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_axis_const_data', 'axis_const'),
+                                   ('axis_const', 'axis_const_data'),
+                                   ('input_data', 'node'),
+                                   ('const_data', 'node'),
+                                   ('axis_const_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, axis, indices, output_shape, ir_version):
+        """
+            ONNX net                                       IR net
+
+            Input->Concat(+gathered const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        indices = np.array(indices)
+
+        concat_axis = 0
+        input_shape = output_shape.copy()
+        concat_output_shape = output_shape.copy()
+        concat_output_shape[concat_axis] = 2 * concat_output_shape[concat_axis]
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, concat_output_shape)
+
+        constant = np.random.randint(-127, 127, shape).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_indices_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['indices'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.INT64,
+                dims=indices.shape,
+                vals=indices.flatten(),
+            ),
+        )
+
+        args = dict()
+        if axis:
+            args['axis'] = axis
+        node_def = onnx.helper.make_node(
+            'Gather',
+            inputs=['const1', 'indices'],
+            outputs=['gather'],
+            **args
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'gather'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_indices_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.take(constant, indices, axis=axis if axis else 0)
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': input_shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': constant.shape, 'kind': 'data', 'value': None},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': concat_output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(shape=[6, 8, 10, 12], axis=2, indices=[[0, 2, 4], [5, 7, 9]], output_shape=[6, 8, 2, 3, 12]),
+        dict(shape=[4, 6, 8, 10, 12], axis=1, indices=[2, 5], output_shape=[4, 2, 8, 10, 12]),
+        dict(shape=[4, 6, 8, 10, 12], axis=-1, indices=[5, 8], output_shape=[4, 6, 8, 10, 2])]
+
+    test_data = [dict(shape=[10, 12], axis=0, indices=[3, 6], output_shape=[2, 12]),
+                 dict(shape=[10, 12], axis=1, indices=[4, 7], output_shape=[10, 2]),
+                 dict(shape=[10, 12], axis=-1, indices=[4, 7], output_shape=[10, 2]),
+                 dict(shape=[10, 12], axis=None, indices=[[0, 1, 3, 4], [5, 6, 8, 9]], output_shape=[2, 4, 12]),
+                 dict(shape=[10, 12], axis=1, indices=[[0, 1, 3, 4, 5], [6, 7, 9, 10, 11]], output_shape=[10, 2, 5]),
+                 dict(shape=[8, 10, 12], axis=0, indices=[3, 6], output_shape=[2, 10, 12]),
+                 dict(shape=[8, 10, 12], axis=1, indices=[4, 7], output_shape=[8, 2, 12]),
+                 dict(shape=[8, 10, 12], axis=2, indices=[5, 8], output_shape=[8, 10, 2]),
+                 dict(shape=[8, 10, 12], axis=-1, indices=[5, 8], output_shape=[8, 10, 2]),
+                 dict(shape=[8, 10, 12], axis=None, indices=[[0, 1], [3, 4], [6, 7]], output_shape=[3, 2, 10, 12]),
+                 dict(shape=[8, 10, 12], axis=1, indices=[[0, 2, 4], [5, 7, 9]], output_shape=[8, 2, 3, 12]),
+                 dict(shape=[6, 8, 10, 12], axis=0, indices=[2, 5], output_shape=[2, 8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12], axis=1, indices=[3, 6], output_shape=[6, 2, 10, 12]),
+                 dict(shape=[6, 8, 10, 12], axis=2, indices=[4, 7], output_shape=[6, 8, 2, 12]),
+                 dict(shape=[6, 8, 10, 12], axis=3, indices=[5, 8], output_shape=[6, 8, 10, 2]),
+                 dict(shape=[6, 8, 10, 12], axis=-1, indices=[5, 8], output_shape=[6, 8, 10, 2]),
+                 dict(shape=[6, 8, 10, 12], axis=None, indices=[[0, 1, 2], [3, 4, 5]], output_shape=[2, 3, 8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12], axis=2, indices=[[0, 2, 4], [5, 7, 9]], output_shape=[6, 8, 2, 3, 12]),
+                 dict(shape=[4, 6, 8, 10, 12], axis=0, indices=[1, 3], output_shape=[2, 6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12], axis=1, indices=[2, 5], output_shape=[4, 2, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12], axis=2, indices=[3, 6], output_shape=[4, 6, 2, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12], axis=3, indices=[4, 7], output_shape=[4, 6, 8, 2, 12]),
+                 dict(shape=[4, 6, 8, 10, 12], axis=4, indices=[5, 8], output_shape=[4, 6, 8, 10, 2]),
+                 dict(shape=[4, 6, 8, 10, 12], axis=-1, indices=[5, 8], output_shape=[4, 6, 8, 10, 2])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_gather(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_gather(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_gather_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_gemm.py b/tests/layer_tests/onnx_tests/test_gemm.py
new file mode 100644
index 00000000000000..83bbe411c04959
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_gemm.py
@@ -0,0 +1,308 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+
+import numpy as np
+import pytest
+import torch
+from common.layer_test_class import CommonLayerTest
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestGemm(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randn(*inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_net(self, shapeA, shapeB, shapeC, alpha, beta, trans_a, trans_b, precision, ir_version):
+        """
+            ONNX net                    IR net
+
+            Input->Gemm->Output   =>    Input->Concat
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        max_len = max([len(shapeA), len(shapeB)])
+        extended_shape1 = np.concatenate([np.ones(max_len - len(shapeA)), shapeA], axis=0)
+        extended_shape2 = np.concatenate([np.ones(max_len - len(shapeB)), shapeB], axis=0)
+        output_shape = np.concatenate(
+            [np.maximum(*[extended_shape1[0:-2], extended_shape2[0:-2]]), [shapeA[-2], shapeB[-1]]],
+            axis=0).astype(np.int).tolist()
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shapeA)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        _shapeB = shapeB.copy()
+        if trans_b:
+            _shapeB.reverse()
+        const1 = np.random.ranf(_shapeB).astype(np.float)
+        const2 = np.random.ranf(shapeC).astype(np.float)
+
+        node_const1_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=const1.shape,
+                vals=const1.flatten(),
+            ),
+        )
+
+        node_const2_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const2'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=const2.shape,
+                vals=const2.flatten(),
+            ),
+        )
+
+        attrs = dict()
+        if alpha:
+            attrs['alpha'] = alpha
+        if beta:
+            attrs['beta'] = beta
+        if trans_a:
+            attrs['transA'] = trans_a
+        if trans_b:
+            attrs['transB'] = trans_b
+        node_def = onnx.helper.make_node(
+            'Gemm',
+            inputs=['input', 'const1', 'const2'],
+            outputs=['output'],
+            **attrs
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const1_def, node_const2_def, node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        if alpha:
+            const1 *= alpha
+        if beta:
+            const2 *= beta
+        if precision == 'FP16':
+            const1 = const1.astype(np.float16)
+            const2 = const2.astype(np.float16)
+        if not trans_b:
+            const1 = const1.transpose()
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_net_double(self, shapeA, shapeB, shapeC, alpha, beta, trans_a, trans_b, precision, ir_version):
+        """
+            ONNX net                    IR net
+
+            Input->Gemm->Output   =>    Input->Concat
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        _shapeA = shapeA.copy()
+        if trans_a:
+            _shapeA.reverse()
+        _shapeB = shapeB.copy()
+        if trans_b:
+            _shapeB.reverse()
+
+        max_len = max([len(shapeA), len(shapeB)])
+        extended_shape1 = np.concatenate([np.ones(max_len - len(shapeA)), shapeA], axis=0)
+        extended_shape2 = np.concatenate([np.ones(max_len - len(shapeB)), shapeB], axis=0)
+        output_shape = np.concatenate(
+            [np.maximum(*[extended_shape1[0:-2], extended_shape2[0:-2]]), [shapeA[-2], shapeB[-1]]],
+            axis=0).astype(np.int).tolist()
+        input1 = helper.make_tensor_value_info('input1', TensorProto.FLOAT, _shapeA)
+        input2 = helper.make_tensor_value_info('input2', TensorProto.FLOAT, _shapeB)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        const = np.random.ranf(shapeC).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=const.shape,
+                vals=const.flatten(),
+            ),
+        )
+
+        attrs = dict()
+        if alpha:
+            attrs['alpha'] = alpha
+        if beta:
+            attrs['beta'] = beta
+        if trans_a:
+            attrs['transA'] = trans_a
+        if trans_b:
+            attrs['transB'] = trans_b
+        node_def = onnx.helper.make_node(
+            'Gemm',
+            inputs=['input1', 'input2', 'const'],
+            outputs=['output'],
+            **attrs
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def],
+            'test_model',
+            [input1, input2],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        if precision == 'FP16':
+            const = const.astype(np.float16)
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data = [
+        dict(shapeA=[3, 6], shapeB=[6, 4], shapeC=[3, 4])
+    ]
+
+    test_data_bc = [
+        dict(shapeA=[3, 6], shapeB=[6, 4], shapeC=[4])
+    ]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("alpha", [None, 0.1, 2.0])
+    @pytest.mark.parametrize("beta", [None, 0.1, 2.0])
+    @pytest.mark.parametrize("trans_a", [None])
+    @pytest.mark.parametrize("trans_b", [None, 1])
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_gemm(self, params, alpha, beta, trans_a, trans_b, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(params['shapeA'], params['shapeB'], params['shapeC'], alpha, beta, trans_a,
+                                    trans_b, precision, ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_bc)
+    @pytest.mark.parametrize("alpha", [None, 0.1, 2.0])
+    @pytest.mark.parametrize("beta", [None, 0.1, 2.0])
+    @pytest.mark.parametrize("trans_a", [None])  # transA is not supported
+    @pytest.mark.parametrize("trans_b", [None, 1])
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_gemm_bc(self, params, alpha, beta, trans_a, trans_b, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(params['shapeA'], params['shapeB'], params['shapeC'], alpha, beta, trans_a,
+                                    trans_b, precision, ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("alpha", [None, 0.1, 2.0])
+    @pytest.mark.parametrize("beta", [None, 0.1, 2.0])
+    @pytest.mark.parametrize("trans_a", [None, 1])
+    @pytest.mark.parametrize("trans_b", [None, 1])
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_gemm_double(self, params, alpha, beta, trans_a, trans_b, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_double(params['shapeA'], params['shapeB'], params['shapeC'], alpha, beta,
+                                           trans_a, trans_b, precision, ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_bc)
+    @pytest.mark.parametrize("alpha", [None, 0.1, 2.0])
+    @pytest.mark.parametrize("beta", [None, 0.1, 2.0])
+    @pytest.mark.parametrize("trans_a", [None, 1])
+    @pytest.mark.parametrize("trans_b", [None, 1])
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_gemm_double_bc(self, params, alpha, beta, trans_a, trans_b, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_double(params['shapeA'], params['shapeB'], params['shapeC'], alpha, beta,
+                                           trans_a, trans_b, precision, ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+
+class PytorchLayerTest(CommonLayerTest):
+    def produce_model_path(self, framework_model, save_path):
+        path = os.path.join(save_path, 'model.onnx')
+        self.torch_model = framework_model['model']
+        torch.onnx.export(self.torch_model, framework_model['var'], path, input_names=['input'],
+                          output_names=['output'])
+        assert os.path.isfile(path), "model.onnx haven't been saved here: {}".format(save_path)
+        return path
+
+    def get_framework_results(self, inputs_dict, model_path):
+        x = torch.tensor(inputs_dict['input'], dtype=torch.float32)
+        return {'output': self.torch_model(x).numpy()}
+
+
+class GemmModel(torch.nn.Module):
+    def __init__(self, weights):
+        super(GemmModel, self).__init__()
+        self.weights = torch.from_numpy(weights)
+
+
+class TestPytorchMM(PytorchLayerTest):
+    def create_net(self, precision, shape, w_shape, output_shape, ir_version):
+        """
+            Pytorch net               IR net
+
+            Input->MM->Output   =>    Input->FullyConnected
+
+        """
+
+        weights_const = np.random.randn(*w_shape).astype(np.float32)
+        #   Create Pytorch model
+        model = GemmModel(weights_const)
+
+        if precision == 'FP16':
+            weights_const = weights_const.astype(np.float16)
+
+        #   Create reference IR net
+
+        # TODO: add reference IR net. Now it is omitted since inference is more
+        #  important and needs to be checked in the first
+
+        ref_net = None
+
+        return {'model': model, 'var': torch.randn(shape)}, ref_net
+
+    test_data = [dict(shape=[1, 2048], w_shape=[2048, 3], output_shape=[1, 3])]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_pytorch_mm(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(precision, **params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_hard_sigmoid.py b/tests/layer_tests/onnx_tests/test_hard_sigmoid.py
new file mode 100644
index 00000000000000..e460d733e1c1f5
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_hard_sigmoid.py
@@ -0,0 +1,235 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestHardSigmoid(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randn(*inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_net(self, shape, alpha, beta, ir_version):
+        """
+            ONNX net                           IR net
+
+            Input->HardSigmoid->Output   =>    Input->HardSigmoid
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        args = dict()
+        if alpha is not None:
+            args['alpha'] = alpha
+        if beta is not None:
+            args['beta'] = beta
+        node_def = onnx.helper.make_node(
+            'HardSigmoid',
+            inputs=['input'],
+            outputs=['output'],
+            **args
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_alpha_data': {'kind': 'data', 'value': [alpha if alpha is not None else 0.2]},
+                'alpha': {'kind': 'op', 'type': 'Const'},
+                'alpha_data': {'shape': [], 'kind': 'data'},
+                'input_beta_data': {'kind': 'data', 'value': [beta if beta is not None else 0.5]},
+                'beta': {'kind': 'op', 'type': 'Const'},
+                'beta_data': {'shape': [], 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'HardSigmoid'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_alpha_data', 'alpha'),
+                                   ('alpha', 'alpha_data'),
+                                   ('input_beta_data', 'beta'),
+                                   ('beta', 'beta_data'),
+                                   ('input_data', 'node'),
+                                   ('alpha_data', 'node'),
+                                   ('beta_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, alpha, beta, precision, ir_version):
+        """
+            ONNX net                                      IR net
+
+            Input->Concat(+hard sigmoid const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        const_number = np.prod(shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+        constant = np.reshape(constant, shape)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        args = dict()
+        if alpha is not None:
+            args['alpha'] = alpha
+        if beta is not None:
+            args['beta'] = beta
+        node_def = onnx.helper.make_node(
+            'HardSigmoid',
+            inputs=['const1'],
+            outputs=['sigmoid1'],
+            **args
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'sigmoid1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.clip(constant * (alpha if alpha is not None else 0.2) + (beta if beta is not None else 0.5), 0, 1)
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(shape=[1, 2], alpha=None, beta=None),
+        dict(shape=[2, 3, 4, 5, 6], alpha=None, beta=0.7)]
+
+    test_data = [
+        dict(shape=[10, 12], alpha=None, beta=None),
+        dict(shape=[8, 10, 12], alpha=None, beta=None),
+        dict(shape=[6, 8, 10, 12], alpha=None, beta=None),
+        dict(shape=[4, 6, 8, 10, 12], alpha=None, beta=None),
+        dict(shape=[10, 12], alpha=0.3, beta=None),
+        dict(shape=[8, 10, 12], alpha=0.3, beta=None),
+        dict(shape=[6, 8, 10, 12], alpha=0.3, beta=None),
+        dict(shape=[4, 6, 8, 10, 12], alpha=0.3, beta=None),
+        dict(shape=[10, 12], alpha=None, beta=0.7),
+        dict(shape=[8, 10, 12], alpha=None, beta=0.7),
+        dict(shape=[6, 8, 10, 12], alpha=None, beta=0.7),
+        dict(shape=[4, 6, 8, 10, 12], alpha=None, beta=0.7),
+        dict(shape=[10, 12], alpha=0.1, beta=0.3),
+        dict(shape=[8, 10, 12], alpha=0.1, beta=0.3),
+        dict(shape=[6, 8, 10, 12], alpha=0.1, beta=0.3),
+        dict(shape=[4, 6, 8, 10, 12], alpha=0.1, beta=0.3)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_hard_sigmoid(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.nightly
+    def test_hard_sigmoid_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_hard_sigmoid_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_identity.py b/tests/layer_tests/onnx_tests/test_identity.py
new file mode 100644
index 00000000000000..e7d8d8acd65bf9
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_identity.py
@@ -0,0 +1,183 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestIdentity(Caffe2OnnxLayerTest):
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                                 IR net
+
+            Input->Identity->Sigmoid->Output   =>    Input->sigmoid
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = helper.make_node(
+            'Identity',
+            inputs=['input'],
+            outputs=['identity']
+        )
+
+        sigmoid_def = helper.make_node(
+            'Sigmoid',
+            inputs=['identity'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def, sigmoid_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'sigmoid': {'kind': 'op', 'type': 'Sigmoid'},
+                'sigmoid_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'sigmoid'),
+                                   ('sigmoid', 'sigmoid_data'),
+                                   ('sigmoid_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                          IR net
+
+            Input->Concat(+identity on const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        constant = np.random.randint(-127, 127, shape).astype(np.float)
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_const_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = helper.make_node(
+            'Identity',
+            inputs=['const1'],
+            outputs=['identity']
+        )
+
+        node_concat_def = helper.make_node(
+            'Concat',
+            inputs=['input', 'identity'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_identity(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_identity_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_image_scaler.py b/tests/layer_tests/onnx_tests/test_image_scaler.py
new file mode 100644
index 00000000000000..ee4f3490b7d5df
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_image_scaler.py
@@ -0,0 +1,161 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+
+
+class TestImageScaler(Caffe2OnnxLayerTest):
+    def create_net(self, shape, scale, ir_version):
+        """
+            ONNX net                           IR net
+
+            Input->ImageScaler->Output   =>    Input->ScaleShift(Power)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        bias = np.random.randint(-10, 10, shape[1]).astype(np.float)
+
+        node_def = onnx.helper.make_node(
+            'ImageScaler',
+            inputs=['input'],
+            outputs=['output'],
+            bias=bias,
+            scale=scale
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, scale, precision, ir_version):
+        """
+            ONNX net                                     IR net
+
+            Input->Concat(+scaled const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.randint(-127, 127, shape).astype(np.float)
+        bias = np.random.randint(-10, 10, shape[1]).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'ImageScaler',
+            inputs=['const1'],
+            outputs=['scale'],
+            bias=bias,
+            scale=scale
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'scale'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        ir_const = constant * scale + np.expand_dims(np.expand_dims([bias], 2), 3)
+        if precision == 'FP16':
+            ir_const = ir_const.astype(np.float16)
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [dict(shape=[2, 4, 6, 8], scale=4.5),
+                           dict(shape=[1, 1, 10, 12], scale=0.5)]
+
+    test_data = [dict(shape=[1, 1, 10, 12], scale=0.5),
+                 dict(shape=[1, 3, 10, 12], scale=1.5),
+                 dict(shape=[6, 8, 10, 12], scale=4.5)]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    def test_image_scaler_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_image_scaler(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    def test_image_scaler_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_image_scaler_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_instance_normalization.py b/tests/layer_tests/onnx_tests/test_instance_normalization.py
new file mode 100644
index 00000000000000..d7467f0951b4c3
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_instance_normalization.py
@@ -0,0 +1,111 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestInstanceNormalization(OnnxRuntimeLayerTest):
+    def create_net(self, shape, epsilon, precision, ir_version):
+        """
+            ONNX net                                     IR net
+
+            Input->InstanceNormalization->Output   =>    Input->MVN->ScaleShift(Power)
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        scale_const = np.random.randn(shape[1]).astype(np.float)
+        bias_const = np.random.randn(shape[1]).astype(np.float)
+
+        node_scale_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['scale'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=scale_const.shape,
+                vals=scale_const.flatten(),
+            ),
+        )
+
+        node_bias_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['bias'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=bias_const.shape,
+                vals=bias_const.flatten(),
+            ),
+        )
+
+        args = dict()
+        if epsilon:
+            args['epsilon'] = epsilon
+        node_def = helper.make_node(
+            'InstanceNormalization',
+            inputs=['input', 'scale', 'bias'],
+            outputs=['output'],
+            **args
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_scale_def, node_bias_def, node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(shape=[1, 1, 4, 6], epsilon=0.001),
+        dict(shape=[1, 1, 2, 4, 6], epsilon=0.001)]
+
+    test_data = [
+        dict(shape=[1, 1, 4, 6], epsilon=None),
+        dict(shape=[1, 1, 4, 6], epsilon=0.001),
+        dict(shape=[1, 2, 4, 6], epsilon=None),
+        dict(shape=[1, 2, 4, 6], epsilon=0.001),
+        dict(shape=[2, 3, 4, 6], epsilon=None),
+        dict(shape=[2, 3, 4, 6], epsilon=0.001),
+        dict(shape=[1, 1, 2, 4, 6], epsilon=None),
+        dict(shape=[1, 1, 2, 4, 6], epsilon=0.001),
+        dict(shape=[1, 2, 4, 6, 6], epsilon=None),
+        dict(shape=[1, 2, 4, 6, 6], epsilon=0.001),
+        dict(shape=[2, 3, 4, 6, 6], epsilon=None),
+        dict(shape=[2, 3, 4, 6, 6], epsilon=0.001)]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_instance_normalization(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_instance_normalization(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_leaky_relu.py b/tests/layer_tests/onnx_tests/test_leaky_relu.py
new file mode 100644
index 00000000000000..985ff27b801d4d
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_leaky_relu.py
@@ -0,0 +1,204 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestLeakyRelu(Caffe2OnnxLayerTest):
+    def create_net(self, shape, alpha, ir_version):
+        """
+            ONNX net                         IR net
+
+            Input->LeakyRelu->Output   =>    Input->ReLU
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'LeakyRelu',
+            inputs=['input'],
+            outputs=['output'],
+            alpha=alpha
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'alpha_indata': {'value': alpha, 'kind': 'data'},
+                'alpha': {'kind': 'op', 'type': 'Const'},
+                'alpha_data': {'shape': [1], 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'PReLU'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('alpha_indata', 'alpha'),
+                                   ('alpha', 'alpha_data'),
+                                   ('alpha_data', 'node'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, alpha, precision, ir_version):
+        """
+            ONNX net                                        IR net
+
+            Input->Concat(+LeakyRelu const)->Output   =>    Input->Concat(+ReLU const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        const_number = np.prod(shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+        constant = np.reshape(constant, shape)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'LeakyRelu',
+            inputs=['const1'],
+            outputs=['lrelu1'],
+            alpha=alpha
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'lrelu1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.clip(constant, 0, np.inf) + np.clip(constant, -np.inf, 0) * alpha
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'const_indata': {'shape': None, 'kind': 'data'},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('const_indata', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(shape=[1, 2], alpha=0.1),
+        dict(shape=[2, 3, 4, 5, 6], alpha=4.5)]
+
+    test_data = [
+        dict(shape=[10, 12], alpha=0.1),
+        dict(shape=[8, 10, 12], alpha=0.9),
+        dict(shape=[6, 8, 10, 12], alpha=1.5),
+        dict(shape=[4, 6, 8, 10, 12], alpha=4.5)]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_leaky_relu_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_leaky_relu(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_leaky_relu_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_leaky_relu_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_log.py b/tests/layer_tests/onnx_tests/test_log.py
new file mode 100644
index 00000000000000..1b284b211ef686
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_log.py
@@ -0,0 +1,193 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestLog(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.rand(*(inputs_dict[input])).astype(np.float32) * 255 + 0.5
+        return inputs_dict
+
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                   IR net
+
+            Input->Log->Output   =>    Input->Log
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Log',
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'Log'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes, [('input', 'input_data'),
+                                                     ('input_data', 'node'),
+                                                     ('node', 'node_data'),
+                                                     ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->Concat(+log const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.rand(*shape).astype(np.float) * 255 + 0.5
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Log',
+            inputs=['const'],
+            outputs=['log']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'log'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.log(constant)
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes, [('input', 'input_data'),
+                                                     ('input_const_data', 'const'),
+                                                     ('const', 'const_data'),
+                                                     ('input_data', 'concat'),
+                                                     ('const_data', 'concat'),
+                                                     ('concat', 'concat_data'),
+                                                     ('concat_data', 'result')])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [dict(shape=[1, 2]),
+                           dict(shape=[2, 3, 4, 5, 6])]
+
+    test_data = [dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_log_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_log(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.nightly
+    def test_log_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_log_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_logsoftmax.py b/tests/layer_tests/onnx_tests/test_logsoftmax.py
new file mode 100644
index 00000000000000..6371eae1a6642c
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_logsoftmax.py
@@ -0,0 +1,240 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from mo.front.common.partial_infer.utils import int64_array
+from unit_tests.utils.graph import build_graph
+
+
+def second_input_data_of_reshape(src_shape, axis):
+    if axis == 0:
+        return [1, -1]
+    if axis == 1:
+        return [0, -1]
+    if axis > 1:
+        return [int(np.prod(int64_array(src_shape[: axis]))), -1]
+    return [-1, int(np.prod(int64_array(src_shape[len(src_shape) + axis:])))]
+
+
+def get_flatten_shape(src_shape, axis):
+    flatten_axis = axis if axis >= 0 else len(src_shape) + axis
+    if flatten_axis == 0:
+        fst_dim = 1
+        snd_dim = int(np.prod(int64_array(src_shape)))
+    elif flatten_axis == 1:
+        fst_dim = src_shape[0]
+        snd_dim = int(np.prod(int64_array(src_shape[1:])))
+    else:
+        fst_dim = int(np.prod(int64_array(src_shape[: flatten_axis])))
+        snd_dim = int(np.prod(int64_array(src_shape[flatten_axis:])))
+    return [fst_dim, snd_dim]
+
+
+class TestLog(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.rand(*(inputs_dict[input])).astype(np.float32) * 255 + 0.5
+        return inputs_dict
+
+    def create_net(self, shape, logsoftmax_axis, ir_version):
+        """
+            ONNX net                   IR net
+
+            Input->LogSoftmax->Output   =>    Input->Softmax->Log->Output
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'LogSoftmax',
+            inputs=['input'],
+            outputs=['output'],
+            axis=logsoftmax_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            converted_shape = shape if len(shape) != 1 else shape[0]
+            flatten_shape = get_flatten_shape(shape, logsoftmax_axis)
+            reshape_data_val = second_input_data_of_reshape(shape, logsoftmax_axis)
+            reduce_sum_shape = np.copy(flatten_shape)
+            reduce_sum_shape[1] = 1
+
+            if len(shape) == 2 and shape == flatten_shape:
+                ref_nodes_attributes = {
+                    'input': {'kind': 'op', 'type': 'Parameter', 'shape': converted_shape},
+                    'input_data': {'shape': shape, 'kind': 'data', 'value': None},
+                    'flatten_shape_val': {'shape': int64_array(reshape_data_val).shape,
+                                          'kind': 'data',
+                                          'value': int64_array(reshape_data_val)},
+                    'flatten_shape': {'type': 'Const', 'kind': 'op', 'shape': 2},
+                    'flatten_shape_data': {'shape': int64_array([2]), 'kind': 'data', 'value': None},
+                    'reshape': {'kind': 'op', 'type': 'Reshape'},
+                    'reshape_data': {'kind': 'data', 'shape': flatten_shape, 'value': None},
+                    'reduce_max_axis_val': {'shape': int64_array([1]).shape, 'kind': 'data', 'value': int64_array([1])},
+                    'reduce_max_axis': {'type': 'Const', 'kind': 'op', 'shape': 1},
+                    'reduce_max_axis_data': {'shape': int64_array([1]), 'kind': 'data', 'value': None},
+                    'reduce_max': {'type': 'ReduceMax', 'kind': 'op', 'keep_dims': True},
+                    'reduce_max_data': {'shape': reduce_sum_shape, 'kind': 'data', 'value': None},
+                    'sub_first': {'type': 'Subtract', 'kind': 'op'},
+                    'sub_first_data': {'shape': flatten_shape, 'kind': 'data', 'value': None},
+                    'reduce_sum_axis_val': {'shape': int64_array([1]).shape, 'kind': 'data', 'value': int64_array([1])},
+                    'reduce_sum_axis': {'type': 'Const', 'kind': 'op', 'shape': 1},
+                    'reduce_sum_axis_data': {'shape': int64_array([1]), 'kind': 'data', 'value': None},
+                    'reduce_sum': {'type': 'ReduceSum', 'kind': 'op', 'keep_dims': True},
+                    'reduce_sum_data': {'shape': reduce_sum_shape, 'kind': 'data', 'value': None},
+                    'exp': {'type': 'Exp', 'kind': 'op'},
+                    'exp_data': {'shape': flatten_shape, 'kind': 'data', 'value': None},
+                    'log': {'type': 'Log', 'kind': 'op'},
+                    'log_data': {'shape': reduce_sum_shape, 'kind': 'data', 'value': None},
+                    'sub_second': {'type': 'Subtract', 'kind': 'op'},
+                    'sub_second_data': {'shape': flatten_shape, 'kind': 'data', 'value': None},
+                    'result': {'kind': 'op', 'type': 'Result'},
+                }
+
+                ref_edges = [
+                    ('input', 'input_data'),
+                    ('flatten_shape_val', 'flatten_shape'),
+                    ('flatten_shape', 'flatten_shape_data'),
+                    ('flatten_shape_data', 'reshape', {'in': 1}),
+                    ('input_data', 'reshape', {'in': 0}),
+                    ('reshape', 'reshape_data'),
+                    ('reduce_max_axis_val', 'reduce_max_axis'),
+                    ('reduce_max_axis', 'reduce_max_axis_data'),
+                    ('reduce_max_axis_data', 'reduce_max', {'in': 1}),
+                    ('reduce_max', 'reduce_max_data'),
+                    ('reshape_data', 'reduce_max', {'out': 0, 'in': 0}),
+                    ('reshape_data', 'sub_first', {'out': 0, 'in': 0}),
+                    ('reduce_max_data', 'sub_first', {'in': 1}),
+                    ('sub_first', 'sub_first_data'),
+                    ('reduce_sum_axis_val', 'reduce_sum_axis'),
+                    ('reduce_sum_axis', 'reduce_sum_axis_data'),
+                    ('reduce_sum_axis_data', 'reduce_sum', {'in': 1}),
+                    ('reduce_sum', 'reduce_sum_data'),
+                    ('sub_first_data', 'exp'),
+                    ('exp', 'exp_data'),
+                    ('exp_data', 'reduce_sum', {'in': 0}),
+                    ('reduce_sum_data', 'log'),
+                    ('log', 'log_data'),
+                    ('log_data', 'sub_second', {'in': 1}),
+                    ('sub_second', 'sub_second_data'),
+                    ('sub_first_data', 'sub_second', {'out': 0, 'in': 0}),
+                    ('sub_second_data', 'result'),
+                ]
+            else:
+                ref_nodes_attributes = {
+                    'input': {'kind': 'op', 'type': 'Parameter', 'shape': converted_shape},
+                    'input_data': {'shape': shape, 'kind': 'data', 'value': None},
+                    'flatten_shape_val': {'shape': int64_array(reshape_data_val).shape,
+                                          'kind': 'data',
+                                          'value': int64_array(reshape_data_val)},
+                    'flatten_shape': {'type': 'Const', 'kind': 'op', 'shape': 2},
+                    'flatten_shape_data': {'shape': int64_array([2]), 'kind': 'data', 'value': None},
+                    'reshape': {'kind': 'op', 'type': 'Reshape'},
+                    'reshape_data': {'kind': 'data', 'shape': flatten_shape, 'value': None},
+                    'reduce_max_axis_val': {'shape': int64_array([1]).shape, 'kind': 'data', 'value': int64_array([1])},
+                    'reduce_max_axis': {'type': 'Const', 'kind': 'op', 'shape': 1},
+                    'reduce_max_axis_data': {'shape': int64_array([1]), 'kind': 'data', 'value': None},
+                    'reduce_max': {'type': 'ReduceMax', 'kind': 'op', 'keep_dims': True},
+                    'reduce_max_data': {'shape': reduce_sum_shape, 'kind': 'data', 'value': None},
+                    'sub_first': {'type': 'Subtract', 'kind': 'op'},
+                    'sub_first_data': {'shape': flatten_shape, 'kind': 'data', 'value': None},
+                    'reduce_sum_axis_val': {'shape': int64_array([1]).shape, 'kind': 'data', 'value': int64_array([1])},
+                    'reduce_sum_axis': {'type': 'Const', 'kind': 'op', 'shape': 1},
+                    'reduce_sum_axis_data': {'shape': int64_array([1]), 'kind': 'data', 'value': None},
+                    'reduce_sum': {'type': 'ReduceSum', 'kind': 'op', 'keep_dims': True},
+                    'reduce_sum_data': {'shape': reduce_sum_shape, 'kind': 'data', 'value': None},
+                    'exp': {'type': 'Exp', 'kind': 'op'},
+                    'exp_data': {'shape': flatten_shape, 'kind': 'data', 'value': None},
+                    'log': {'type': 'Log', 'kind': 'op'},
+                    'log_data': {'shape': reduce_sum_shape, 'kind': 'data', 'value': None},
+                    'sub_second': {'type': 'Subtract', 'kind': 'op'},
+                    'sub_second_data': {'shape': flatten_shape, 'kind': 'data', 'value': None},
+                    'last_shape_val': {'shape': int64_array(shape).shape, 'kind': 'data', 'value': int64_array(shape)},
+                    'last_shape': {'type': 'Const', 'kind': 'op', 'shape': len(shape)},
+                    'last_shape_data': {'shape': int64_array([len(shape)]), 'kind': 'data', 'value': None},
+                    'last_reshape': {'kind': 'op', 'type': 'Reshape'},
+                    'last_reshape_data': {'kind': 'data', 'shape': shape, 'value': None},
+                    'result': {'kind': 'op', 'type': 'Result'},
+                }
+
+                ref_edges = [
+                    ('input', 'input_data'),
+                    ('flatten_shape_val', 'flatten_shape'),
+                    ('flatten_shape', 'flatten_shape_data'),
+                    ('flatten_shape_data', 'reshape', {'in': 1}),
+                    ('input_data', 'reshape', {'in': 0}),
+                    ('reshape', 'reshape_data'),
+                    ('reduce_max_axis_val', 'reduce_max_axis'),
+                    ('reduce_max_axis', 'reduce_max_axis_data'),
+                    ('reduce_max_axis_data', 'reduce_max', {'in': 1}),
+                    ('reduce_max', 'reduce_max_data'),
+                    ('reshape_data', 'reduce_max', {'out': 0, 'in': 0}),
+                    ('reshape_data', 'sub_first', {'out': 0, 'in': 0}),
+                    ('reduce_max_data', 'sub_first', {'in': 1}),
+                    ('sub_first', 'sub_first_data'),
+                    ('reduce_sum_axis_val', 'reduce_sum_axis'),
+                    ('reduce_sum_axis', 'reduce_sum_axis_data'),
+                    ('reduce_sum_axis_data', 'reduce_sum', {'in': 1}),
+                    ('reduce_sum', 'reduce_sum_data'),
+                    ('sub_first_data', 'exp'),
+                    ('exp', 'exp_data'),
+                    ('exp_data', 'reduce_sum', {'in': 0}),
+                    ('reduce_sum_data', 'log'),
+                    ('log', 'log_data'),
+                    ('log_data', 'sub_second', {'in': 1}),
+                    ('sub_second', 'sub_second_data'),
+                    ('sub_first_data', 'sub_second', {'out': 0, 'in': 0}),
+                    ('last_shape_val', 'last_shape'),
+                    ('last_shape', 'last_shape_data'),
+                    ('last_shape_data', 'last_reshape', {'in': 1}),
+                    ('sub_second_data', 'last_reshape', {'in': 0}),
+                    ('last_reshape', 'last_reshape_data'),
+                    ('last_reshape_data', 'result'),
+                ]
+
+            ref_net = build_graph(ref_nodes_attributes, ref_edges)
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(shape=[2, 4], logsoftmax_axis=-1),
+        dict(shape=[2, 3, 2, 5, 6], logsoftmax_axis=-2)]
+
+    test_data = [
+        dict(shape=[10, 12], logsoftmax_axis=-1),
+        dict(shape=[4, 5, 3], logsoftmax_axis=1),
+        dict(shape=[6, 8, 5, 7], logsoftmax_axis=2),
+        dict(shape=[2, 3, 2, 5, 6], logsoftmax_axis=-2)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_log(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_lrn.py b/tests/layer_tests/onnx_tests/test_lrn.py
new file mode 100644
index 00000000000000..66104f467b9c23
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_lrn.py
@@ -0,0 +1,133 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestLRN(OnnxRuntimeLayerTest):
+    def create_net(self, shape, alpha, beta, bias, size, ir_version):
+        """
+            ONNX net                   IR net
+
+            Input->LRN->Output   =>    Input->Norm->Power
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        args = dict(size=size)
+        if alpha:
+            args['alpha'] = alpha
+        if beta:
+            args['beta'] = beta
+        if bias:
+            args['bias'] = bias
+        node_def = onnx.helper.make_node(
+            'LRN',
+            inputs=['input'],
+            outputs=['output'],
+            **args
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+        if not alpha:
+            alpha = 0.0001
+        if not beta:
+            beta = 0.75
+        if not bias:
+            bias = 1.0
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'const_indata': {'value': [1], 'kind': 'data'},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': [1], 'kind': 'data'},
+                'norm': {'kind': 'op', 'type': 'LRN', 'alpha': alpha / bias, 'beta': beta, 'bias': bias,
+                         'size': size},  # 'region': 'across'
+                'norm_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            edges = [('input', 'input_data'),
+                     ('input_data', 'norm'),
+                     ('const_indata', 'const'),
+                     ('const', 'const_data'),
+                     ('const_data', 'norm'),
+                     ('norm', 'norm_data'),
+                     ('norm_data', 'result')
+                     ]
+
+            ref_net = build_graph(nodes_attributes, edges)
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(shape=[2, 12], alpha=None, beta=None, bias=None, size=1),
+        pytest.param(dict(shape=[2, 3, 12], alpha=0.0002, beta=0.5, bias=2.0, size=3),
+                     marks=pytest.mark.skip(reason="Skipped until fixed")),
+        dict(shape=[2, 3, 12], alpha=0.0002, beta=0.5, bias=2.0, size=3),
+        dict(shape=[2, 3, 12], alpha=0.0002, beta=0.5, bias=2.0, size=3)]
+
+    test_data = [
+        dict(shape=[2, 12], alpha=None, beta=None, bias=None, size=1),
+        dict(shape=[2, 12], alpha=0.0002, beta=0.5, bias=2.0, size=1),
+        dict(shape=[2, 3, 12], alpha=None, beta=None, bias=None, size=3),
+        dict(shape=[2, 3, 12], alpha=0.0002, beta=0.5, bias=2.0, size=1),
+        dict(shape=[2, 3, 12], alpha=0.0002, beta=0.5, bias=2.0, size=3),
+        dict(shape=[2, 3, 8, 10, 12], alpha=None, beta=None, bias=None, size=3),
+        dict(shape=[2, 3, 8, 10, 12], alpha=0.0002, beta=0.5, bias=2.0, size=1),
+        dict(shape=[2, 3, 8, 10, 12], alpha=0.0002, beta=0.5, bias=2.0, size=3)]
+
+    test_data_4D = [
+        dict(shape=[2, 3, 10, 12], alpha=None, beta=None, bias=None, size=3),
+        dict(shape=[2, 3, 10, 12], alpha=0.0002, beta=0.5, bias=2.0, size=1),
+        dict(shape=[2, 3, 10, 12], alpha=0.0002, beta=0.5, bias=2.0, size=3)]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_lrn_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        # onnxruntime only supports 4D tensors for LRN
+        self.skip_framework = True
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_lrn(self, params, ie_device, precision, ir_version, temp_dir):
+        # onnxruntime only supports 4D tensors for LRN
+        self.skip_framework = True
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_lrn_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self.skip_framework = False
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_lstm.py b/tests/layer_tests/onnx_tests/test_lstm.py
new file mode 100644
index 00000000000000..e142d0070a40c2
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_lstm.py
@@ -0,0 +1,160 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+
+
+class TestLSTM(Caffe2OnnxLayerTest):
+    skip_framework = True
+
+    def create_lstm(self, direction: str, cell_type: str, hidden_size=128):
+        """
+            ONNX net
+
+            Input->LSTM->Output   =>   Only accuracy check
+
+        """
+
+        #   Create ONNX model
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        assert cell_type in ['LSTM', 'RNN', 'GRU']
+        assert direction in ['forward', 'reverse', 'bidirectional']
+        n_gates = {'LSTM': 4, 'RNN': 1, 'GRU': 3}
+        M = n_gates[cell_type]
+
+        seq_len = 10
+        batch_size = 4
+        input_size = 64
+        num_direction = 1 if direction in ["forward", "reverse"] else 2
+
+        input_shape = [seq_len, batch_size, input_size]
+        output_shape = [seq_len, num_direction, batch_size, hidden_size]
+
+        w_shape = [num_direction, M * hidden_size, input_size]
+        r_shape = [num_direction, M * hidden_size, hidden_size]
+
+        init_h_shape = [num_direction, batch_size, hidden_size]
+        init_c_shape = [num_direction, batch_size, hidden_size]
+
+        init_h_value = np.ones(init_h_shape, dtype=np.float32)
+        init_c_value = np.ones(init_c_shape, dtype=np.float32)
+
+        w_value = np.ones(w_shape, dtype=np.float32)
+        r_value = np.ones(r_shape, dtype=np.float32)
+
+        # Creating LSTM Operation
+        x = helper.make_tensor_value_info('X', TensorProto.FLOAT, input_shape)
+        y = helper.make_tensor_value_info('Y', TensorProto.FLOAT, None)
+
+        w = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['W'],
+            value=onnx.helper.make_tensor(
+                name='const_tensor',
+                data_type=onnx.TensorProto.FLOAT,
+                dims=w_value.shape,
+                vals=w_value.flatten().astype(float),
+            ),
+        )
+
+        r = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['R'],
+            value=onnx.helper.make_tensor(
+                name='const_tensor',
+                data_type=onnx.TensorProto.FLOAT,
+                dims=r_value.shape,
+                vals=r_value.flatten().astype(float),
+            ),
+        )
+
+        init_h = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['init_h'],
+            value=onnx.helper.make_tensor(
+                name='const_tensor',
+                data_type=onnx.TensorProto.FLOAT,
+                dims=init_h_value.shape,
+                vals=init_h_value.flatten().astype(float),
+            ),
+        )
+
+        inputs = ['X', 'W', 'R', '', '', 'init_h']
+
+        if cell_type == 'LSTM':
+            init_c = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['init_c'],
+                value=onnx.helper.make_tensor(
+                    name='const_tensor',
+                    data_type=onnx.TensorProto.FLOAT,
+                    dims=init_c_value.shape,
+                    vals=init_c_value.flatten().astype(float),
+                ),
+            )
+
+            inputs.append('init_c')
+
+        node_lstm = onnx.helper.make_node(
+            cell_type,
+            inputs=inputs,
+            outputs=['', 'Y'],
+            hidden_size=hidden_size,
+            direction=direction,
+        )
+
+        # Create the graph (GraphProto)
+        if cell_type == 'LSTM':
+            graph_def = helper.make_graph(
+                [w, r, init_h, init_c, node_lstm],
+                'test_lstm',
+                [x],
+                [y],
+            )
+        else:
+            graph_def = helper.make_graph(
+                [w, r, init_h, node_lstm],
+                'test_lstm',
+                [x],
+                [y],
+            )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_{}_model'.format(cell_type))
+
+        # We do not create reference graph, as it's too complicated to construct it
+        # Moreover, IR reader do not support TensorIterator layers
+        # So we return None to skip IR comparision
+
+        return onnx_net, None
+
+    @pytest.mark.precommit
+    @pytest.mark.parametrize('direction', ["forward", "bidirectional", "reverse"])
+    @pytest.mark.parametrize('cell_type', ["LSTM", "GRU", "RNN"])
+    def test_lstm_simple_precommit(self, direction, cell_type, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_lstm(direction, cell_type), ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    # LSTM/RNN/GRU Sequence Generation
+    @pytest.mark.parametrize('direction', ["forward", "bidirectional", "reverse"])
+    @pytest.mark.parametrize('cell_type', ["LSTM", "GRU", "RNN"])
+    def test_lstm_sequence_generate(self, direction, cell_type, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_lstm(direction, cell_type), ie_device, precision, ir_version,
+                   disabled_transforms='lstm_to_tensor_iterator,gru_and_rnn_to_tensor_iterator', temp_dir=temp_dir)
+
+    # TODO: add more params for nightly
+    @pytest.mark.nightly
+    @pytest.mark.parametrize('direction', ["forward", "bidirectional", "reverse"])
+    @pytest.mark.parametrize('cell_type', ["LSTM", "GRU", "RNN"])
+    def test_lstm_nightly(self, direction, cell_type, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_lstm(direction, cell_type), ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_matmul.py b/tests/layer_tests/onnx_tests/test_matmul.py
new file mode 100644
index 00000000000000..d9fd24588093ed
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_matmul.py
@@ -0,0 +1,190 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestMatMul(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randn(*inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_net(self, shape1, shape2, precision, ir_version):
+        """
+            ONNX net                                 IR net
+
+            Input->MatMul with const->Output   =>    Input->FullyConnected
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        max_len = max([len(shape1), len(shape2)])
+        extended_shape1 = np.concatenate([np.ones(max_len - len(shape1)), shape1], axis=0)
+        extended_shape2 = np.concatenate([np.ones(max_len - len(shape2)), shape2], axis=0)
+        output_shape = np.concatenate(
+            [np.maximum(*[extended_shape1[0:-2], extended_shape2[0:-2]]), [shape1[-2], shape2[-1]]],
+            axis=0).astype(np.int).tolist()
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape1)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        const = np.random.randn(*shape2).astype(np.float32)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=const.shape,
+                vals=const.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'MatMul',
+            inputs=['input', 'const'],
+            outputs=['mm_output']
+        )
+
+        # to avoid mapping problems
+        node_elu_def = onnx.helper.make_node(
+            'Elu',
+            inputs=['mm_output'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_elu_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #   Please, spesify 'type': 'Input' for inpit node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        if precision == 'FP16':
+            const = const.astype(np.float16)
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_dual_net(self, shape1, shape2, ir_version):
+        """
+            ONNX net                                 IR net
+
+            Input->MatMul->Output   =>    Input->Concat
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        max_len = max([len(shape1), len(shape2)])
+        extended_shape1 = np.concatenate([np.ones(max_len - len(shape1)), shape1], axis=0)
+        extended_shape2 = np.concatenate([np.ones(max_len - len(shape2)), shape2], axis=0)
+        output_shape = np.concatenate(
+            [np.maximum(*[extended_shape1[0:-2], extended_shape2[0:-2]]), [shape1[-2], shape2[-1]]],
+            axis=0).astype(np.int).tolist()
+        input1 = helper.make_tensor_value_info('input1', TensorProto.FLOAT, shape1)
+        input2 = helper.make_tensor_value_info('input2', TensorProto.FLOAT, shape2)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_def = onnx.helper.make_node(
+            'MatMul',
+            inputs=['input1', 'input2'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input1, input2],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #   Please, spesify 'type': 'Input' for inpit node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data = [
+        dict(shape1=[4, 6], shape2=[6, 4]),
+        dict(shape1=[1, 4, 6], shape2=[1, 6, 4]),
+        dict(shape1=[2, 4, 6], shape2=[2, 6, 4]),
+        dict(shape1=[1, 1, 4, 6], shape2=[1, 1, 6, 4]),
+        dict(shape1=[1, 2, 4, 6], shape2=[1, 2, 6, 4]),
+        dict(shape1=[2, 3, 4, 6], shape2=[2, 3, 6, 4]),
+        dict(shape1=[2, 3, 4, 4, 6], shape2=[2, 3, 4, 6, 4])
+    ]
+
+    test_data_broadcasting = [
+        dict(shape1=[1, 4, 6], shape2=[6, 4]),
+        dict(shape1=[2, 4, 6], shape2=[6, 4]),
+        dict(shape1=[2, 4, 6], shape2=[1, 6, 4]),
+        dict(shape1=[1, 1, 4, 6], shape2=[6, 4]),
+        dict(shape1=[1, 1, 4, 6], shape2=[1, 6, 4]),
+        dict(shape1=[1, 2, 4, 6], shape2=[6, 4]),
+        dict(shape1=[1, 2, 4, 6], shape2=[2, 6, 4]),
+        dict(shape1=[2, 3, 4, 6], shape2=[6, 4]),
+        dict(shape1=[2, 3, 4, 6], shape2=[3, 6, 4]),
+        dict(shape1=[2, 3, 4, 6], shape2=[1, 3, 6, 4]),
+        dict(shape1=[2, 3, 4, 4, 6], shape2=[6, 4]),
+        dict(shape1=[2, 3, 4, 4, 6], shape2=[4, 6, 4]),
+        dict(shape1=[2, 3, 4, 4, 6], shape2=[3, 4, 6, 4])
+    ]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_matmul(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_broadcasting)
+    @pytest.mark.nightly
+    def test_matmul_bc(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_dual_matmul(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_dual_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_broadcasting)
+    @pytest.mark.nightly
+    def test_dual_matmul_bc(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_dual_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_mean_variance_normalization.py b/tests/layer_tests/onnx_tests/test_mean_variance_normalization.py
new file mode 100644
index 00000000000000..eb9d5e3b262dea
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_mean_variance_normalization.py
@@ -0,0 +1,73 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestMeanVarianceNormalization(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randn(*inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_net(self, shape, axes, ir_version):
+        """
+            ONNX net                                 IR net
+
+            Input->MeanVarianceNormalization->Output   =>    Input->MVN
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'MeanVarianceNormalization',
+            inputs=['input'],
+            outputs=['output'],
+            axes=axes
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output]
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data = [
+        dict(shape=[7, 2, 3, 5], axes=[2, 3]),
+        dict(shape=[7, 2, 3, 5], axes=[1, 2, 3]),
+        dict(shape=[7, 2, 3, 5, 11], axes=[2, 3, 4]),
+        dict(shape=[7, 2, 3, 5, 11], axes=[1, 2, 3, 4])
+    ]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_mvn(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_neg.py b/tests/layer_tests/onnx_tests/test_neg.py
new file mode 100644
index 00000000000000..3dd2e0106fdbbd
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_neg.py
@@ -0,0 +1,91 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestNeg(Caffe2OnnxLayerTest):
+    def create_neg(self, shape, ir_version):
+        """
+            ONNX net                   IR net
+
+            Input->Neg->Output   =>    Input->Power(scale=-1, shift=0, power=1)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_reduce_mean_def = onnx.helper.make_node(
+            'Neg',
+            inputs=['input'],
+            outputs=['output'],
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_reduce_mean_def],
+            'test_neg_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_neg_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'neg': {'kind': 'op', 'type': 'Negative'},
+                'neg_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'neg'),
+                                   ('neg', 'neg_data'),
+                                   ('neg_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [dict(shape=[2, 3, 4]),
+                           dict(shape=[1, 3, 124, 124])]
+
+    test_data = [dict(shape=[1, 64]),
+                 dict(shape=[2, 3, 4]),
+                 dict(shape=[1, 3, 124, 124]),
+                 ]
+
+    @pytest.mark.parametrize('params', test_data_precommit)
+    @pytest.mark.precommit
+    def test_neg_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_neg(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize('params', test_data)
+    @pytest.mark.nightly
+    def test_neg(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_neg(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_non_zero.py b/tests/layer_tests/onnx_tests/test_non_zero.py
new file mode 100644
index 00000000000000..da83e8e7e26bba
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_non_zero.py
@@ -0,0 +1,194 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestNonZero(Caffe2OnnxLayerTest):
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                    IR net
+
+            Input->NonZero->Output   =>    Input->NonZero->Result
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'NonZero',
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'NonZero', 'version': 'opset3', 'output_type': 'i64'},
+                'node_data': {'shape': [len(shape), np.prod(shape)], 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+        return onnx_net, ref_net
+
+    def create_net_const(self, input_value, output_value, precision, ir_version):
+        """
+            ONNX net                                   IR net
+
+            Input->Concat(+NonZero const)->Output   =>    Input->Concat(+const)->Result
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        concat_axis = 0
+        output_shape = list(output_value.shape)
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, output_value.shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=input_value.shape,
+                vals=input_value.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'NonZero',
+            inputs=['const1'],
+            outputs=['nonzero1']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'nonzero1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': output_value.shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': output_value.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': output_value.shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data = [
+        dict(shape=[10, 12]),
+        dict(shape=[8, 10, 12]),
+        dict(shape=[6, 8, 10, 12]),
+        dict(shape=[4, 6, 8, 10, 12])
+    ]
+
+    test_const_data = [
+        dict(
+            input_value=np.array([3, 0, 0, 0, 4, 0, 5, 6, 0]).reshape((3, 3)),
+            output_value=np.array([0, 1, 2, 2, 0, 1, 0, 1]).reshape(2, 4),
+        ),
+        dict(
+            input_value=np.array([0, 1, 0, 1]).reshape((4)),
+            output_value=np.array([1, 3]).reshape((1, 2)),
+        ),
+        dict(
+            input_value=np.array([0, 1, 0, 1, 1, 0, 1, 0]).reshape((2, 4)),
+            output_value=np.array([0, 0, 1, 1, 1, 3, 0, 2]).reshape((2, 4)),
+        ),
+        dict(
+            input_value=np.array([1, 0, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 0]).reshape((2, 3, 3)),
+            output_value=np.array([0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0, 1, 2, 2, 2,
+                                   0, 0, 0, 1, 1, 2, 0, 2, 1, 0, 1, 2, 0, 1, 2, 0, 2, 1]).reshape((3, 12)),
+        ),
+    ]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_non_zero(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_const_data)
+    @pytest.mark.nightly
+    def test_non_zero_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_not.py b/tests/layer_tests/onnx_tests/test_not.py
new file mode 100644
index 00000000000000..da840eef28e887
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_not.py
@@ -0,0 +1,190 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestNot(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randint(0, 2, inputs_dict[input]).astype(np.bool)
+        return inputs_dict
+
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                   IR net
+
+            Input->Not->Output   =>    Input->LogicalNot
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.BOOL, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, shape)
+
+        node_def = onnx.helper.make_node(
+            'Not',
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'LogicalNot'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->Concat(+not const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.BOOL, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, output_shape)
+
+        constant = np.random.randint(0, 2, shape).astype(np.bool)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.BOOL,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Not',
+            inputs=['const1'],
+            outputs=['not']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'not'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.logical_not(constant)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [dict(shape=[2, 3, 4]),
+                           dict(shape=[2, 4, 6, 8, 10])]
+
+    test_data = [dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_not_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_not(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_not_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_or.py b/tests/layer_tests/onnx_tests/test_or.py
new file mode 100644
index 00000000000000..8580f7e8bfd6ec
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_or.py
@@ -0,0 +1,277 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestOr(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randint(0, 2, inputs_dict[input]).astype(np.bool)
+        return inputs_dict
+
+    def create_net(self, shape1, shape2, ir_version):
+        """
+            ONNX net                                 IR net
+
+            Input->Or with 2nd input->Output   =>    Input->LogicalOr
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input1 = helper.make_tensor_value_info('input1', TensorProto.BOOL, shape1)
+        input2 = helper.make_tensor_value_info('input2', TensorProto.BOOL, shape2)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, shape1)
+
+        node_def = helper.make_node(
+            'Or',
+            inputs=['input1', 'input2'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input1, input2],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input1': {'kind': 'op', 'type': 'Parameter'},
+                'input1_data': {'shape': shape1, 'kind': 'data'},
+                'input2': {'kind': 'op', 'type': 'Parameter'},
+                'input2_data': {'shape': shape2, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'LogicalOr'},
+                'node_data': {'shape': shape1, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input1', 'input1_data'),
+                                   ('input2', 'input2_data'),
+                                   ('input1_data', 'node'),
+                                   ('input2_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    def create_net_one_const(self, shape1, shape2, ir_version):
+        """
+            ONNX net                             IR net
+
+            Input->Or with const->Output   =>    Input->LogicalOr
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.BOOL, shape1)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, shape1)
+
+        const = np.random.randint(0, 2, shape2).astype(np.bool)
+
+        node_const_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.BOOL,
+                dims=const.shape,
+                vals=const.flatten(),
+            ),
+        )
+
+        node_def = helper.make_node(
+            'Or',
+            inputs=['input', 'const'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape1, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': const.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': const.shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'LogicalOr'},
+                'node_data': {'shape': shape1, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'node'),
+                                   ('const_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape1, shape2, ir_version):
+        """
+            ONNX net                                         IR net
+
+            Input->Concat with const or const->Output   =>   Input->Concat
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        concat_axis = 0
+        output_shape = list(shape1)
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.BOOL, shape1)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, output_shape)
+
+        const1 = np.random.randint(0, 2, shape1).astype(np.bool)
+        const2 = np.random.randint(0, 2, shape2).astype(np.bool)
+
+        node_const1_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.BOOL,
+                dims=const1.shape,
+                vals=const1.flatten(),
+            ),
+        )
+
+        node_const2_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const2'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.BOOL,
+                dims=const2.shape,
+                vals=const2.flatten(),
+            ),
+        )
+
+        node_def = helper.make_node(
+            'Or',
+            inputs=['const1', 'const2'],
+            outputs=['node_out']
+        )
+
+        node_concat_def = helper.make_node(
+            'Concat',
+            inputs=['input', 'node_out'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const1_def, node_const2_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+        constant_calculated = np.logical_or(const1, const2)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': const1.shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant_calculated.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': const1.shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [dict(shape1=[2, 3, 4], shape2=[2, 3, 4]),
+                           dict(shape1=[2, 4, 6, 8, 10], shape2=[2, 4, 6, 8, 10])]
+
+    test_data = [dict(shape1=[4, 6], shape2=[4, 6]),
+                 dict(shape1=[4, 6, 8], shape2=[4, 6, 8]),
+                 dict(shape1=[4, 6, 8, 10], shape2=[4, 6, 8, 10]),
+                 dict(shape1=[4, 6, 8, 10, 12], shape2=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_or_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_or(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_or_one_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_one_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_or_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_pad.py b/tests/layer_tests/onnx_tests/test_pad.py
new file mode 100644
index 00000000000000..56c1870287156a
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_pad.py
@@ -0,0 +1,212 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestPad(OnnxRuntimeLayerTest):
+    def create_net(self, shape, mode, pads, value, ir_version, opset=None):
+        """
+            ONNX net                   IR net
+
+            Input->Pad->Output   =>    Input->Pad
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        _pads = np.array(pads).reshape([2, -1])
+        output_shape = (np.array(shape) + _pads[0, :] + _pads[1, :]).tolist()
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        nodes = []
+        if opset is not None and opset < 11:
+            args = dict(pads=pads)
+            if mode:
+                args['mode'] = mode
+            if value:
+                args['value'] = value
+            node_def = onnx.helper.make_node(
+                'Pad',
+                inputs=['input'],
+                outputs=['pad'],
+                **args
+            )
+            nodes.append(node_def)
+        else:
+            node_pads_def = helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['pads'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.INT64,
+                    dims=[len(pads)],
+                    vals=pads,
+                ),
+            )
+
+            inputs = ['input', 'pads']
+            if value is not None:
+                node_value_def = helper.make_node(
+                    'Constant',
+                    inputs=[],
+                    outputs=['value'],
+                    value=helper.make_tensor(
+                        name='const_tensor',
+                        data_type=TensorProto.FLOAT,
+                        dims=[],
+                        vals=[value],
+                    ),
+                )
+                inputs.append('value')
+                nodes.append(node_value_def)
+
+            args = dict()
+            if mode:
+                args['mode'] = mode
+            node_def = onnx.helper.make_node(
+                'Pad',
+                inputs=inputs,
+                outputs=['pad'],
+                **args
+            )
+            nodes.extend([node_pads_def, node_def])
+
+        sigmoid_def = onnx.helper.make_node(
+            'Elu',
+            inputs=['pad'],
+            outputs=['output']
+        )
+        nodes.append(sigmoid_def)
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'pads_begin_indata': {'value': _pads[0, :], 'kind': 'data'},
+                'pads_begin': {'kind': 'op', 'type': 'Const'},
+                'pads_begin_data': {'shape': [len(_pads[0, :])], 'kind': 'data'},
+                'pads_end_indata': {'value': _pads[1, :], 'kind': 'data'},
+                'pads_end': {'kind': 'op', 'type': 'Const'},
+                'pads_end_data': {'shape': [len(_pads[1, :])], 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'Pad', 'pad_mode': 'constant' if not mode else mode},
+                'node_data': {'shape': output_shape, 'kind': 'data'},
+                'elu': {'kind': 'op', 'type': 'Elu'},
+                'elu_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            edges = [('input', 'input_data'),
+                     ('input_data', 'node'),
+                     ('pads_begin_indata', 'pads_begin'),
+                     ('pads_begin', 'pads_begin_data'),
+                     ('pads_begin_data', 'node'),
+                     ('pads_end_indata', 'pads_end'),
+                     ('pads_end', 'pads_end_data'),
+                     ('pads_end_data', 'node'),
+                     ('node', 'node_data'),
+                     ('node_data', 'elu'),
+                     ('elu', 'elu_data'),
+                     ('elu_data', 'result')
+                     ]
+
+            if mode in (None, "constant"):
+                nodes_attributes.update({'const_node_indata': {'value': value, 'kind': 'data'},
+                                         'const_node': {'kind': 'op', 'type': 'Const'},
+                                         'const_node_data': {'shape': None, 'kind': 'data'}
+                                         })
+                edges += [('const_node_indata', 'const_node'),
+                          ('const_node', 'const_node_data'),
+                          ('const_node_data', 'node')
+                          ]
+
+            ref_net = build_graph(nodes_attributes, edges)
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        pytest.param(dict(shape=[6, 8, 10, 12], pads=[1, 2, 3, 4, 5, 6, 7, 8]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed")),
+        pytest.param(dict(shape=[8, 10, 12, 14, 16], pads=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed"))]
+
+    test_data = [dict(shape=[10, 12], pads=[1, 1, 1, 1]),
+                 dict(shape=[10, 12], pads=[1, 2, 3, 4]),
+                 dict(shape=[8, 10, 12], pads=[0, 0, 1, 0, 0, 1]),
+                 dict(shape=[8, 10, 12], pads=[1, 2, 3, 4, 5, 6]),
+                 dict(shape=[6, 8, 10, 12], pads=[0, 0, 1, 1, 0, 0, 1, 1]),
+                 dict(shape=[6, 8, 10, 12], pads=[0, 0, 1, 2, 0, 0, 3, 4]),
+                 dict(shape=[6, 8, 10, 12], pads=[1, 1, 1, 1, 1, 1, 1, 1]),
+                 dict(shape=[6, 8, 10, 12], pads=[1, 2, 3, 4, 5, 6, 7, 8]),
+                 dict(shape=[8, 10, 12, 14, 16], pads=[1, 1, 1, 1, 1, 1, 1, 1, 1, 1]),
+                 dict(shape=[8, 10, 12, 14, 16], pads=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("mode_value", [(None, None),
+                                            (None, 0.5),
+                                            ('constant', None),
+                                            ('constant', 0.5),
+                                            ('reflect', None),
+                                            ('edge', None)])
+    @pytest.mark.nightly
+    def test_pad_opset_9(self, params, mode_value, ie_device, precision, ir_version, temp_dir):
+        mode, value = mode_value
+        self._test(*self.create_net(**params, mode=mode, value=value, ir_version=ir_version, opset=9),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.parametrize("mode_value", [(None, None),
+                                            (None, 0.5),
+                                            ('constant', None),
+                                            ('constant', 0.5),
+                                            ('reflect', None),
+                                            ('edge', None)])
+    @pytest.mark.precommit
+    def test_pad_opset_latest_precommit(self, params, mode_value, ie_device, precision, ir_version, temp_dir):
+        mode, value = mode_value
+        self._test(*self.create_net(**params, mode=mode, value=value, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("mode_value", [(None, None),
+                                            (None, 0.5),
+                                            ('constant', None),
+                                            ('constant', 0.5),
+                                            ('reflect', None),
+                                            ('edge', None)])
+    @pytest.mark.nightly
+    def test_pad_opset_latest(self, params, mode_value, ie_device, precision, ir_version, temp_dir):
+        mode, value = mode_value
+        self._test(*self.create_net(**params, mode=mode, value=value, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_pooling.py b/tests/layer_tests/onnx_tests/test_pooling.py
new file mode 100644
index 00000000000000..3213718b9ac5de
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_pooling.py
@@ -0,0 +1,432 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+def float_array(x):
+    return np.array(x, dtype=np.float)
+
+
+class TestPooling(OnnxRuntimeLayerTest):
+    def create_net(self, shape, kernel_shape, pads, strides, op, ir_version, count_include_pad=None, auto_pad=None,
+                   storage_order=None, ceil=False, opset=None):
+        """
+            ONNX net                      IR net
+
+            Input->Pooling>Output   =>    Input->Pooling
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        node_args = dict(kernel_shape=kernel_shape)
+        if auto_pad is not None:
+            node_args['auto_pad'] = auto_pad
+            if auto_pad == 'VALID':
+                pads = np.zeros(len(shape[2:]) * 2, dtype=np.int)
+        else:
+            auto_pad = 'NOTSET'
+        if count_include_pad is not None:
+            node_args['count_include_pad'] = count_include_pad
+        else:
+            count_include_pad = 0
+        if storage_order is not None:
+            node_args['storage_order'] = storage_order
+        if pads is not None:
+            if auto_pad == 'NOTSET':
+                node_args['pads'] = pads
+            _pads = np.transpose(np.array(pads).reshape([2, -1]))
+        else:
+            _pads = np.zeros([len(kernel_shape), 2])
+        if strides is not None:
+            node_args['strides'] = strides
+        else:
+            strides = np.ones(len(kernel_shape))
+
+        if ceil:
+            node_args['ceil_mode'] = 1
+
+        if auto_pad in ['SAME_UPPER', 'SAME_LOWER']:
+            out_spacial_shape = np.ceil(np.array(shape[2:], dtype=np.float) / strides)
+        else:
+            rounding = np.ceil if ceil else np.floor
+            out_spacial_shape = rounding(
+                (float_array(shape[2:]) + np.add(_pads[:, 0], _pads[:, 1]) - float_array(kernel_shape)) / strides + 1)
+
+        out_shape = np.array(shape)
+        out_shape[2:] = out_spacial_shape
+        out_shape = out_shape.astype(np.int).tolist()
+        concat_axis = 0
+        out_concat_shape = out_shape.copy()
+        out_concat_shape[concat_axis] *= 2
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, out_concat_shape)
+
+        constant = np.random.randint(-127, 127, out_shape).astype(np.float)
+
+        node_def = onnx.helper.make_node(
+            op,
+            inputs=['input'],
+            outputs=['pool'],
+            **node_args
+        )
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['pool', 'const1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        graph_def = helper.make_graph(
+            [node_def, node_const_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': None,
+                         'pads_begin': _pads[:, 0] if len(shape) > 3 else _pads[0, 0],
+                         'pads_end': _pads[:, 1] if len(shape) > 3 else _pads[0, 1],
+                         'kernel': kernel_shape[0] if len(kernel_shape) == 1 else kernel_shape,
+                         'rounding_type': 'ceil' if auto_pad != 'NOTSET' or ceil else 'floor',
+                         'auto_pad': None},
+                'node_data': {'shape': out_shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': out_shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': out_concat_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            if op == 'AveragePool':
+                nodes_attributes['node']['type'] = 'AvgPool'
+                nodes_attributes['node']['exclude-pad'] = 'true' if count_include_pad == 0 else 'false'
+            else:
+                nodes_attributes['node']['type'] = 'MaxPool'
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('node_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_global_net(self, shape, op, ir_version):
+        """
+            ONNX net                            IR net
+
+            Input->GlobalPooling>Output   =>    Input->Pooling
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        out_shape = np.ones(len(shape))
+        out_shape[:2] = np.array(shape)[:2]
+        out_shape = out_shape.astype(np.int).tolist()
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, out_shape)
+
+        node_def = onnx.helper.make_node(
+            op,
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_axes_data': {'kind': 'data', 'value': list(range(2, len(shape)))},
+                'axes': {'kind': 'op', 'type': 'Const'},
+                'axes_data': {'shape': [len(shape) - 2], 'kind': 'data'},
+                'node': {'kind': 'op', 'type': None},
+                'node_data': {'shape': out_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            if op == 'GlobalAveragePool':
+                nodes_attributes['node']['type'] = 'ReduceMean'
+            else:
+                nodes_attributes['node']['type'] = 'ReduceMax'
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('input_axes_data', 'axes'),
+                                   ('axes', 'axes_data'),
+                                   ('axes_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(shape=[2, 3, 10], kernel_shape=[2], pads=None, strides=[3]),
+        dict(shape=[2, 3, 30, 30], kernel_shape=[5, 5], pads=None, strides=[3, 2]),
+        dict(shape=[2, 3, 28, 28, 28], kernel_shape=[5, 5, 5], pads=[2, 4, 2, 0, 0, 2], strides=None),
+        dict(shape=[2, 3, 30, 30, 30], kernel_shape=[5, 5, 5], pads=None, strides=[3, 3, 5])]
+
+    test_data = [
+        dict(shape=[2, 3, 10], kernel_shape=[2], pads=None, strides=None),
+        dict(shape=[2, 3, 10], kernel_shape=[2], pads=[2, 2], strides=None),
+        dict(shape=[2, 3, 10], kernel_shape=[2], pads=None, strides=[3]),
+        dict(shape=[2, 3, 30, 30], kernel_shape=[2, 2], pads=None, strides=None),
+        dict(shape=[2, 3, 30, 30], kernel_shape=[4, 2], pads=None, strides=None),
+        dict(shape=[2, 3, 30, 30], kernel_shape=[2, 4], pads=None, strides=None),
+        dict(shape=[2, 3, 28, 28], kernel_shape=[3, 3], pads=[2, 2, 2, 2], strides=None),
+        dict(shape=[2, 3, 28, 28], kernel_shape=[5, 5], pads=[0, 2, 0, 4], strides=None),
+        dict(shape=[2, 3, 28, 28], kernel_shape=[5, 5], pads=[2, 0, 4, 0], strides=None),
+        dict(shape=[2, 3, 30, 30], kernel_shape=[5, 5], pads=None, strides=[3, 3]),
+        dict(shape=[2, 3, 30, 30], kernel_shape=[5, 5], pads=None, strides=[2, 3]),
+        dict(shape=[2, 3, 30, 30], kernel_shape=[5, 5], pads=None, strides=[3, 2]),
+        dict(shape=[2, 3, 30, 30, 30], kernel_shape=[2, 2, 2], pads=None, strides=None),
+        dict(shape=[2, 3, 30, 30, 30], kernel_shape=[4, 2, 2], pads=None, strides=None),
+        dict(shape=[2, 3, 30, 30, 30], kernel_shape=[2, 4, 2], pads=None, strides=None),
+        dict(shape=[2, 3, 30, 30, 30], kernel_shape=[2, 2, 4], pads=None, strides=None),
+        dict(shape=[2, 3, 28, 28, 28], kernel_shape=[3, 3, 3], pads=[2, 2, 2, 2, 2, 2], strides=None),
+        dict(shape=[2, 3, 28, 28, 28], kernel_shape=[5, 5, 5], pads=[2, 4, 2, 0, 0, 2], strides=None),
+        dict(shape=[2, 3, 30, 30, 30], kernel_shape=[5, 5, 5], pads=None, strides=[3, 3, 3]),
+        dict(shape=[2, 3, 30, 30, 30], kernel_shape=[5, 5, 5], pads=None, strides=[5, 3, 3]),
+        dict(shape=[2, 3, 30, 30, 30], kernel_shape=[5, 5, 5], pads=None, strides=[3, 5, 3]),
+        dict(shape=[2, 3, 30, 30, 30], kernel_shape=[5, 5, 5], pads=None, strides=[3, 3, 5])]
+
+    test_data_autopad_precommit = [
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='VALID', kernel_shape=[2, 2, 4], pads=None, strides=None),
+        dict(shape=[2, 3, 21, 21, 21], auto_pad='VALID', kernel_shape=[3, 3, 3], pads=None, strides=[3, 2, 3]),
+        dict(shape=[2, 3, 21, 21, 21], auto_pad='VALID', kernel_shape=[3, 3, 3], pads=None, strides=[3, 3, 2])]
+
+    test_data_autopad = [
+        dict(shape=[2, 3, 10], auto_pad='SAME_UPPER', kernel_shape=[2], pads=[0, 1], strides=[3]),
+        dict(shape=[2, 3, 10], auto_pad='SAME_LOWER', kernel_shape=[2], pads=[0, 1], strides=[3]),
+        dict(shape=[2, 3, 10], auto_pad='VALID', kernel_shape=[2], pads=None, strides=[3]),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[2, 2], pads=[0, 0, 1, 1], strides=None),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[4, 2], pads=[1, 0, 2, 1], strides=None),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[2, 4], pads=[0, 1, 1, 2], strides=None),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[5, 5], pads=[1, 1, 1, 1], strides=[3, 3]),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[5, 5], pads=[1, 1, 2, 1], strides=[2, 3]),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[5, 5], pads=[1, 1, 1, 2], strides=[3, 2]),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[2, 2], pads=[0, 0, 1, 1], strides=None),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[4, 2], pads=[1, 0, 2, 1], strides=None),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[2, 4], pads=[0, 1, 1, 2], strides=None),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[5, 5], pads=[1, 1, 1, 1], strides=[3, 3]),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[5, 5], pads=[1, 1, 2, 1], strides=[2, 3]),
+        dict(shape=[2, 3, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[5, 5], pads=[1, 1, 1, 2], strides=[3, 2]),
+        dict(shape=[2, 3, 30, 30], auto_pad='VALID', kernel_shape=[2, 2], pads=None, strides=None),
+        dict(shape=[2, 3, 30, 30], auto_pad='VALID', kernel_shape=[4, 2], pads=None, strides=None),
+        dict(shape=[2, 3, 30, 30], auto_pad='VALID', kernel_shape=[2, 4], pads=None, strides=None),
+        dict(shape=[2, 3, 21, 21], auto_pad='VALID', kernel_shape=[3, 3], pads=None, strides=[3, 3]),
+        dict(shape=[2, 3, 21, 21], auto_pad='VALID', kernel_shape=[3, 3], pads=None, strides=[2, 3]),
+        dict(shape=[2, 3, 21, 21], auto_pad='VALID', kernel_shape=[3, 3], pads=None, strides=[3, 2]),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[2, 2, 2], pads=[0, 0, 0, 1, 1, 1],
+             strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[4, 2, 2], pads=[1, 0, 0, 2, 1, 1],
+             strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[2, 4, 2], pads=[0, 1, 0, 1, 2, 1],
+             strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[2, 2, 4], pads=[0, 0, 1, 1, 1, 2],
+             strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[5, 5, 5], pads=[1, 1, 1, 1, 1, 1],
+             strides=[3, 3, 3]),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[5, 5, 5], pads=[0, 1, 1, 0, 1, 1],
+             strides=[5, 3, 3]),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[5, 5, 5], pads=[1, 0, 1, 1, 0, 1],
+             strides=[3, 5, 3]),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_UPPER', kernel_shape=[5, 5, 5], pads=[1, 1, 0, 1, 1, 0],
+             strides=[3, 3, 5]),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[2, 2, 2], pads=[0, 0, 0, 1, 1, 1],
+             strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[4, 2, 2], pads=[1, 0, 0, 2, 1, 1],
+             strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[2, 4, 2], pads=[0, 1, 0, 1, 2, 1],
+             strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[2, 2, 4], pads=[0, 0, 1, 1, 1, 2],
+             strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[5, 5, 5], pads=[1, 1, 1, 1, 1, 1],
+             strides=[3, 3, 3]),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[5, 5, 5], pads=[0, 1, 1, 0, 1, 1],
+             strides=[5, 3, 3]),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[5, 5, 5], pads=[1, 0, 1, 1, 0, 1],
+             strides=[3, 5, 3]),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='SAME_LOWER', kernel_shape=[5, 5, 5], pads=[1, 1, 0, 1, 1, 0],
+             strides=[3, 3, 5]),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='VALID', kernel_shape=[2, 2, 2], pads=None, strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='VALID', kernel_shape=[4, 2, 2], pads=None, strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='VALID', kernel_shape=[2, 4, 2], pads=None, strides=None),
+        dict(shape=[2, 3, 30, 30, 30], auto_pad='VALID', kernel_shape=[2, 2, 4], pads=None, strides=None),
+        dict(shape=[2, 3, 21, 21, 21], auto_pad='VALID', kernel_shape=[3, 3, 3], pads=None, strides=[3, 3, 3]),
+        dict(shape=[2, 3, 21, 21, 21], auto_pad='VALID', kernel_shape=[3, 3, 3], pads=None, strides=[2, 3, 3]),
+        dict(shape=[2, 3, 21, 21, 21], auto_pad='VALID', kernel_shape=[3, 3, 3], pads=None, strides=[3, 2, 3]),
+        dict(shape=[2, 3, 21, 21, 21], auto_pad='VALID', kernel_shape=[3, 3, 3], pads=None, strides=[3, 3, 2])]
+
+    global_test_data = [dict(shape=[2, 3, 10]),
+                        dict(shape=[2, 3, 32, 32]),
+                        dict(shape=[2, 3, 32, 32, 32])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("incl_pad", [None, 1])
+    @pytest.mark.nightly
+    def test_avgpool_opset7(self, params, incl_pad, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(
+            *self.create_net(**params, op='AveragePool', count_include_pad=incl_pad, ir_version=ir_version, opset=7),
+            ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_autopad)
+    @pytest.mark.nightly
+    def test_avgpool_opset7_autopad(self, params, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(*self.create_net(**params, op='AveragePool', ir_version=ir_version, opset=7),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("incl_pad", [None, 1])
+    @pytest.mark.parametrize("ceil", [True, False])
+    @pytest.mark.nightly
+    def test_avgpool_opset10(self, params, incl_pad, ceil, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(
+            *self.create_net(**params, op='AveragePool', count_include_pad=incl_pad, ceil=ceil, ir_version=ir_version,
+                             opset=10), ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_autopad)
+    @pytest.mark.nightly
+    def test_avgpool_opset10_autopad(self, params, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(*self.create_net(**params, op='AveragePool', ir_version=ir_version, opset=10),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("st_order", [None, 1])
+    @pytest.mark.nightly
+    def test_maxpool_opset8(self, params, st_order, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(*self.create_net(**params, op='MaxPool', storage_order=st_order, ir_version=ir_version, opset=8),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_autopad)
+    @pytest.mark.nightly
+    def test_maxpool_opset8_autopad(self, params, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(*self.create_net(**params, op='MaxPool', ir_version=ir_version, opset=8),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("st_order", [None, 1])
+    @pytest.mark.parametrize("ceil", [True, False])
+    @pytest.mark.nightly
+    def test_maxpool_opset10(self, params, st_order, ceil, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(*self.create_net(**params, op='MaxPool', storage_order=st_order, ceil=ceil, ir_version=ir_version,
+                                    opset=10), ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_autopad_precommit)
+    @pytest.mark.precommit
+    def test_maxpool_opset10_autopad(self, params, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(*self.create_net(**params, op='MaxPool', ir_version=ir_version, opset=10),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_autopad)
+    @pytest.mark.nightly
+    def test_maxpool_opset10_autopad(self, params, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(*self.create_net(**params, op='MaxPool', ir_version=ir_version, opset=10),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", global_test_data)
+    @pytest.mark.nightly
+    def test_global_avgpool(self, params, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(*self.create_global_net(**params, op='GlobalAveragePool', ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", global_test_data)
+    @pytest.mark.nightly
+    def test_global_maxpool(self, params, ie_device, precision, ir_version, temp_dir):
+        if not len(params['shape']) in [4, 5]:
+            pytest.skip("Pooling layer support only 4D and 5D input tensors")
+        self._test(*self.create_global_net(**params, op='GlobalMaxPool', ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_prelu.py b/tests/layer_tests/onnx_tests/test_prelu.py
new file mode 100644
index 00000000000000..303ff0270ab9d6
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_prelu.py
@@ -0,0 +1,161 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestPRelu(Caffe2OnnxLayerTest):
+    def create_net(self, shape, slope_shape, precision, ir_version, opset=None):
+        """
+            ONNX net                     IR net
+
+            Input->PRelu->Output   =>    Input->PReLU
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        const = np.random.randn(*slope_shape).astype(np.float32)
+
+        node_slope_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['slope'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=const.shape,
+                vals=const.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'PRelu',
+            inputs=['input', 'slope'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_slope_def, node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'weights_indata': {'kind': 'data', 'value': const.flatten()},
+                'weights': {'kind': 'op', 'type': 'Const'},
+                'weights_data': {'kind': 'data', 'shape': [len(const.flatten())]},
+                'node': {'kind': 'op', 'type': 'PReLU'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('weights_indata', 'weights'),
+                                   ('weights', 'weights_data'),
+                                   ('weights_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    # Note: IE only support slopes of one element or of size equal to number of channels.
+    test_data_shared_channels = [
+        dict(shape=[10, 12], slope_shape=[12]),
+        dict(shape=[8, 10, 12], slope_shape=[10]),
+        dict(shape=[6, 8, 10, 12], slope_shape=[8]),
+        dict(shape=[4, 6, 8, 10, 12], slope_shape=[6])]
+
+    test_data_scalar_precommit = [
+        dict(shape=[2, 4, 6, 8], slope_shape=[1]),
+        dict(shape=[2, 4, 6, 8, 10], slope_shape=[1])
+    ]
+
+    test_data_scalar = [
+        dict(shape=[10, 12], slope_shape=[1]),
+        dict(shape=[8, 10, 12], slope_shape=[1]),
+        dict(shape=[6, 8, 10, 12], slope_shape=[1]),
+        dict(shape=[4, 6, 8, 10, 12], slope_shape=[1])]
+
+    test_data_precommit = [dict(shape=[8, 10, 12], slope_shape=[10])]
+
+    @pytest.mark.parametrize("params", test_data_scalar)
+    @pytest.mark.nightly
+    def test_prelu_opset6_scalar(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_shared_channels)
+    @pytest.mark.nightly
+    def test_prelu_opset6_shared_channels(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_scalar)
+    @pytest.mark.nightly
+    def test_prelu_opset7_scalar(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, opset=7, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_shared_channels)
+    @pytest.mark.nightly
+    def test_prelu_opset7_shared_channels(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, opset=7, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_prelu_shared_channels_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_scalar_precommit)
+    @pytest.mark.precommit
+    def test_prelu_scalar_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_scalar)
+    @pytest.mark.nightly
+    def test_prelu_scalar(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_shared_channels)
+    @pytest.mark.nightly
+    def test_prelu_shared_channels(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_reciprocal.py b/tests/layer_tests/onnx_tests/test_reciprocal.py
new file mode 100644
index 00000000000000..4161ed28860635
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_reciprocal.py
@@ -0,0 +1,171 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestReciprocal(OnnxRuntimeLayerTest):
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                              IR net
+
+            Input+258->Reciprocal->Output   =>    Input->Power
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        # adding 258 is needed to avoid division by zero
+        node_const_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=[1],
+                vals=[258],
+            ),
+        )
+
+        node_add_def = helper.make_node(
+            'Add',
+            inputs=['input', 'const'],
+            outputs=['add']
+        )
+
+        node_def = helper.make_node(
+            'Reciprocal',
+            inputs=['add'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_add_def, node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                            IR net
+
+            Input->Concat with reciprocal consts->Output   =>   Input->Concat
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        concat_axis = 0
+        output_shape = list(shape)
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        const = np.random.randint(1, 256, shape).astype(np.float)
+
+        node_const_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=const.shape,
+                vals=const.flatten(),
+            ),
+        )
+
+        node_def = helper.make_node(
+            'Reciprocal',
+            inputs=['const'],
+            outputs=['node_out']
+        )
+
+        node_concat_def = helper.make_node(
+            'Concat',
+            inputs=['input', 'node_out'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+
+        constant_calculated = 1 / const
+        if precision == 'FP16':
+            constant_calculated = constant_calculated.astype(np.float16)
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(shape=[2, 4]),
+        dict(shape=[2, 4, 6, 8])]
+
+    test_data = [
+        dict(shape=[4, 6]),
+        dict(shape=[4, 6, 8]),
+        dict(shape=[4, 6, 8, 10]),
+        dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_reciprocal_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_reciprocal(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_reciprocal_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_reciprocal_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_reduce.py b/tests/layer_tests/onnx_tests/test_reduce.py
new file mode 100644
index 00000000000000..2df16ef3d85aed
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_reduce.py
@@ -0,0 +1,179 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestReduce(OnnxRuntimeLayerTest):
+    def create_reduce(self, shape, reshapped_shape, op, axes, keep_dims, ir_version):
+        """
+            ONNX net                               IR net
+
+            Input->Reduce Operation (axes)->Output   =>    Input->Reduce Operation
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        if op not in ['ReduceMin', 'ReduceMax', 'ReduceMean', 'ReduceProd', 'ReduceSum']:
+            raise ValueError("Operation has to be either Reduce(Min or Max or Mean or Sum or Prod")
+
+        output_shape = shape.copy()
+        for axis in axes:
+            output_shape[axis] = 1
+
+        if not keep_dims:
+            output_shape = [dim for dim in output_shape if dim != 1]
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_def = onnx.helper.make_node(
+            op,
+            inputs=['input'],
+            outputs=['output'],
+            axes=axes,
+            keepdims=keep_dims
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_data_1': {'shape': [len(axes)], 'value': axes, 'kind': 'data'},
+                'const_1': {'kind': 'op', 'type': 'Const'},
+                'const_data_1': {'shape': [len(axes)], 'kind': 'data'},
+                'reduce': {'kind': 'op', 'type': op, 'keep_dims': keep_dims},
+                'reduce_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data_1', 'const_1'),
+                                   ('const_1', 'const_data_1'),
+                                   ('input_data', 'reduce'),
+                                   ('const_data_1', 'reduce'),
+                                   ('reduce', 'reduce_data'),
+                                   ('reduce_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(shape=[2, 4, 6], reshapped_shape=[2, 1, 4 * 6, 1], axes=[1, 2]),
+        dict(shape=[2, 4, 6, 8], reshapped_shape=[2, 1, 4 * 6 * 8, 1], axes=[1, 2, 3]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[2, 4, 6 * 8 * 10, 1], axes=[2, 3, 4])
+    ]
+
+    test_data = [
+        dict(shape=[2, 4, 6], reshapped_shape=[1, 1, 2, 4 * 6], axes=[0]),
+        dict(shape=[2, 4, 6], reshapped_shape=[2, 1, 4, 6], axes=[1]),
+        dict(shape=[2, 4, 6], reshapped_shape=[2, 4, 6, 1], axes=[2]),
+        dict(shape=[2, 4, 6], reshapped_shape=[1, 1, 2 * 4, 6], axes=[0, 1]),
+        dict(shape=[2, 4, 6], reshapped_shape=[2, 1, 4 * 6, 1], axes=[1, 2]),
+        dict(shape=[2, 4, 6, 8], reshapped_shape=[1, 1, 2, 4 * 6 * 8], axes=[0]),
+        dict(shape=[2, 4, 6, 8], reshapped_shape=[2, 1, 4, 6 * 8], axes=[1]),
+        dict(shape=[2, 4, 6, 8], reshapped_shape=[2, 4, 6, 8], axes=[2]),
+        dict(shape=[2, 4, 6, 8], reshapped_shape=[2, 4 * 6, 8, 1], axes=[3]),
+        dict(shape=[2, 4, 6, 8], reshapped_shape=[1, 1, 2 * 4, 6 * 8], axes=[0, 1]),
+        dict(shape=[2, 4, 6, 8], reshapped_shape=[2, 1, 4 * 6, 8], axes=[1, 2]),
+        dict(shape=[2, 4, 6, 8], reshapped_shape=[2, 4, 6 * 8, 1], axes=[2, 3]),
+        dict(shape=[2, 4, 6, 8], reshapped_shape=[1, 1, 2 * 4 * 6, 8], axes=[0, 1, 2]),
+        dict(shape=[2, 4, 6, 8], reshapped_shape=[2, 1, 4 * 6 * 8, 1], axes=[1, 2, 3]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[1, 1, 2, 4 * 6 * 8 * 10], axes=[0]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[2, 1, 4, 6 * 8 * 10], axes=[1]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[2, 4, 6, 8 * 10], axes=[2]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[2, 4 * 6, 8, 10], axes=[3]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[2, 4 * 6 * 8, 10, 1], axes=[4]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[1, 1, 2 * 4, 6 * 8 * 10], axes=[0, 1]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[2, 1, 4 * 6, 8 * 10], axes=[1, 2]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[2, 4, 6 * 8, 10], axes=[2, 3]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[2, 4 * 6, 8 * 10, 1], axes=[3, 4]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[1, 1, 2 * 4 * 6, 8 * 10], axes=[0, 1, 2]),
+        dict(shape=[2, 4, 6, 8, 10], reshapped_shape=[2, 4, 6 * 8 * 10, 1], axes=[2, 3, 4])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.precommit
+    def test_reduce_max_precommit(self, params, keep_dims, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce(**params, op='ReduceMax', keep_dims=keep_dims, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.nightly
+    def test_reduce_max(self, params, keep_dims, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce(**params, op='ReduceMax', keep_dims=keep_dims, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.nightly
+    def test_reduce_sum(self, params, keep_dims, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce(**params, op='ReduceSum', keep_dims=keep_dims, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.nightly
+    def test_reduce_prod(self, params, keep_dims, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce(**params, op='ReduceProd', keep_dims=keep_dims, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.precommit
+    def test_reduce_mean_precommit(self, params, keep_dims, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce(**params, op='ReduceMean', keep_dims=keep_dims, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_reduce_mean(self, params, keep_dims, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce(**params, op='ReduceMean', keep_dims=keep_dims, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.precommit
+    def test_reduce_min_precommit(self, params, keep_dims, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce(**params, op='ReduceMin', keep_dims=keep_dims, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.nightly
+    def test_reduce_min(self, params, keep_dims, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce(**params, op='ReduceMin', keep_dims=keep_dims, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_reduce_lp.py b/tests/layer_tests/onnx_tests/test_reduce_lp.py
new file mode 100644
index 00000000000000..4f8c42848334cb
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_reduce_lp.py
@@ -0,0 +1,247 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestReduceL1L2(OnnxRuntimeLayerTest):
+    def create_reduce_lp(self, shape, axes, keep_dims, reduce_p, ir_version):
+        """
+            ONNX net                              IR net
+
+            Input->ReduceLX(axes)->Output   =>    Input->ReduceLX
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        output_shape = shape.copy()
+        _axes = axes.copy() if axes is not None else list(range(len(shape)))
+        for axis in _axes:
+            output_shape[axis] = 1
+
+        if not keep_dims:
+            output_shape = [dim for dim in output_shape if dim != 1]
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        args = dict(keepdims=keep_dims)
+        if axes:
+            args['axes'] = axes
+        node_def = onnx.helper.make_node(
+            "ReduceL" + str(reduce_p),
+            inputs=['input'],
+            outputs=['output'],
+            **args
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_data_1': {'shape': [len(_axes)], 'value': _axes, 'kind': 'data'},
+                'const_1': {'kind': 'op', 'type': 'Const'},
+                'const_data_1': {'shape': [len(_axes)], 'kind': 'data'},
+                'reduce': {'kind': 'op', 'type': "ReduceL" + str(reduce_p), 'keep_dims': keep_dims},
+                'reduce_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data_1', 'const_1'),
+                                   ('const_1', 'const_data_1'),
+                                   ('input_data', 'reduce'),
+                                   ('const_data_1', 'reduce'),
+                                   ('reduce', 'reduce_data'),
+                                   ('reduce_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_reduce_lp_const(self, shape, axes, keep_dims, reduce_p, ir_version):
+        """
+            ONNX net                              IR net
+
+            Input->ReduceLX(axes)->Output   =>    Input->ReduceLX
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        output_shape = shape.copy()
+        _axes = axes.copy() if axes is not None else list(range(len(shape)))
+        for axis in _axes:
+            output_shape[axis] = 1
+
+        if not keep_dims:
+            output_shape = [dim for dim in output_shape if dim != 1]
+        if len(output_shape) == 0:
+            output_shape = [1]
+
+        concat_axis = 0
+        concat_output_shape = output_shape.copy()
+        concat_output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, output_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, concat_output_shape)
+
+        constant = np.random.randn(*shape).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        args = dict(keepdims=keep_dims)
+        if axes:
+            args['axes'] = axes
+        node_def = onnx.helper.make_node(
+            "ReduceL" + str(reduce_p),
+            inputs=['const1'],
+            outputs=['reduce'],
+            **args
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'reduce'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+        constant = np.power(np.sum(a=np.abs(np.power(constant, reduce_p)), axis=tuple(_axes), keepdims=keep_dims), 1 / reduce_p)
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': output_shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': constant.shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': concat_output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(shape=[2, 4, 6, 8], axes=[-3, -1, -2]),
+        dict(shape=[2, 4, 6, 8, 10], axes=[-4, -2]),
+    ]
+
+    test_data = [
+        dict(shape=[8], axes=None),
+        dict(shape=[8], axes=[0]),
+        dict(shape=[2, 4, 6], axes=None),
+        dict(shape=[2, 4, 6], axes=[1]),
+        dict(shape=[2, 4, 6], axes=[-2]),
+        dict(shape=[2, 4, 6], axes=[-2, -1]),
+        dict(shape=[2, 4, 6, 8], axes=[0]),
+        dict(shape=[2, 4, 6, 8], axes=[-3, -1, -2]),
+        dict(shape=[2, 4, 6, 8, 10], axes=None),
+        dict(shape=[2, 4, 6, 8, 10], axes=[-2]),
+        dict(shape=[2, 4, 6, 8, 10], axes=[1, 3]),
+        dict(shape=[2, 4, 6, 8, 10], axes=[-4, -2]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.parametrize("reduce_p", [1, 2])
+    @pytest.mark.precommit
+    def test_reduce_lp_precommit(self, params, keep_dims, reduce_p, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce_lp(**params, keep_dims=keep_dims, reduce_p=reduce_p, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.parametrize("reduce_p", [1, 2])
+    @pytest.mark.nightly
+    def test_reduce_lp(self, params, keep_dims, reduce_p, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce_lp(**params, keep_dims=keep_dims, reduce_p=reduce_p, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.parametrize("reduce_p", [1, 2])
+    @pytest.mark.precommit
+    def test_reduce_lp_const_precommit(self, params, keep_dims, reduce_p, ie_device, precision, ir_version, temp_dir):
+        self._test(
+            *self.create_reduce_lp_const(**params, keep_dims=keep_dims, reduce_p=reduce_p, ir_version=ir_version),
+            ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("keep_dims", [True, False])
+    @pytest.mark.parametrize("reduce_p", [1, 2])
+    @pytest.mark.nightly
+    def test_reduce_lp_const(self, params, keep_dims, reduce_p, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reduce_lp_const(**params, keep_dims=keep_dims, reduce_p=reduce_p,ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_relu.py b/tests/layer_tests/onnx_tests/test_relu.py
new file mode 100644
index 00000000000000..19cc96d1530405
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_relu.py
@@ -0,0 +1,183 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestRelu(Caffe2OnnxLayerTest):
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                    IR net
+
+            Input->Relu->Output   =>    Input->ReLU
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Relu',
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'ReLU'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                   IR net
+
+            Input->Concat(+Relu const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        const_number = np.prod(shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+        constant = np.reshape(constant, shape)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Relu',
+            inputs=['const1'],
+            outputs=['relu1']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'relu1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.clip(constant, 0, np.inf)
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])
+                 ]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_relu(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_relu_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_reshape.py b/tests/layer_tests/onnx_tests/test_reshape.py
new file mode 100644
index 00000000000000..4483d1250455fb
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_reshape.py
@@ -0,0 +1,283 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestReshape(Caffe2OnnxLayerTest):
+    def create_reshape_net(self, input_shape, output_shape, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->Reshape->Output   =>    Input->Reshape
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_shape_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['shape'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.INT64,
+                dims=[len(output_shape)],
+                vals=output_shape,
+            ),
+        )
+
+        node_reshape_def = onnx.helper.make_node(
+            'Reshape',
+            inputs=['input', 'shape'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_shape_def, node_reshape_def],
+            'test_reshape_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_reshape_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': input_shape, 'kind': 'data'},
+                'input_data_1': {'shape': [len(output_shape)], 'value': output_shape, 'kind': 'data'},
+                'const_1': {'kind': 'op', 'type': 'Const'},
+                'const_data_1': {'shape': [len(output_shape)], 'value': None, 'kind': 'data'},  # 'value': output_shape,
+                'reshape': {'kind': 'op', 'type': 'Reshape'},
+                'reshape_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data_1', 'const_1'),
+                                   ('const_1', 'const_data_1'),
+                                   ('const_data_1', 'reshape'),
+                                   ('input_data', 'reshape'),
+                                   ('reshape', 'reshape_data'),
+                                   ('reshape_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_reshape_net_const(self, input_shape, output_shape, ir_version):
+        """
+            ONNX net                                         IR net
+
+            Input->Concat(+reshaped const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        concat_output_shape = output_shape.copy()
+        concat_output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, output_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, concat_output_shape)
+
+        const_number = np.prod(input_shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=input_shape,
+                vals=constant,
+            ),
+        )
+
+        node_shape_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['shape'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.INT64,
+                dims=[len(output_shape)],
+                vals=output_shape,
+            ),
+        )
+
+        node_reshape_def = onnx.helper.make_node(
+            'Reshape',
+            inputs=['const1', 'shape'],
+            outputs=['reshape1']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'reshape1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_shape_def, node_reshape_def, node_concat_def],
+            'test_reshape_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_reshape_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': output_shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': output_shape, 'value': None, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': concat_output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result'),
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_5D = [dict(input_shape=[4, 6, 8, 10, 12], output_shape=[4, 6, 8, 120]),
+                    dict(input_shape=[4, 6, 8, 10, 12], output_shape=[4, 6, 80, 12]),
+                    dict(input_shape=[4, 6, 8, 10, 12], output_shape=[4, 48, 10, 12]),
+                    dict(input_shape=[4, 6, 8, 10, 12], output_shape=[24, 8, 10, 12]),
+                    dict(input_shape=[4, 6, 8, 10], output_shape=[2, 2, 6, 8, 10]),
+                    dict(input_shape=[4, 6, 8, 10], output_shape=[4, 2, 3, 8, 10]),
+                    dict(input_shape=[4, 6, 8, 10], output_shape=[4, 6, 2, 4, 10]),
+                    dict(input_shape=[4, 6, 8, 10], output_shape=[4, 6, 8, 2, 5])]
+
+    test_data_5D_precommit = [dict(input_shape=[2, 4, 6, 8, 10], output_shape=[8, 6, 8, 10])]
+
+    test_data_4D = [dict(input_shape=[4, 6, 8, 10], output_shape=[24, 8, 10]),
+                    dict(input_shape=[4, 6, 8, 10], output_shape=[4, 48, 10]),
+                    dict(input_shape=[4, 6, 8, 10], output_shape=[4, 6, 80]),
+                    dict(input_shape=[4, 6, 8, 10], output_shape=[192, 10]),
+                    dict(input_shape=[4, 6, 8, 10], output_shape=[4, 480]),
+                    dict(input_shape=[4, 6, 8], output_shape=[2, 2, 6, 8]),
+                    dict(input_shape=[4, 6, 8], output_shape=[4, 2, 3, 8]),
+                    dict(input_shape=[4, 6, 8], output_shape=[4, 6, 2, 4]),
+                    dict(input_shape=[4, 6], output_shape=[2, 2, 2, 3])]
+
+    test_data_4D_precommit = [dict(input_shape=[2, 4, 6, 8], output_shape=[48, 8])]
+
+    test_data_3D = [dict(input_shape=[4, 6, 8], output_shape=[24, 8]),
+                    dict(input_shape=[4, 6, 8], output_shape=[4, 48]),
+                    dict(input_shape=[4, 6], output_shape=[2, 2, 6]),
+                    dict(input_shape=[4, 6], output_shape=[4, 2, 3]),
+                    dict(input_shape=[4, 6], output_shape=[2, 4, 3])]
+
+    test_data_3D_precommit = [dict(input_shape=[2, 4, 6], output_shape=[8, 6])]
+
+    @pytest.mark.parametrize("params", test_data_5D_precommit)
+    @pytest.mark.precommit
+    def test_reshape_5D_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reshape_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D_precommit)
+    @pytest.mark.precommit
+    def test_reshape_4D_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reshape_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D_precommit)
+    @pytest.mark.precommit
+    def test_reshape_3D_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reshape_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_reshape_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reshape_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_reshape_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reshape_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_reshape_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reshape_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_reshape_const_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reshape_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_reshape_const_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reshape_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_reshape_const_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_reshape_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_resize.py b/tests/layer_tests/onnx_tests/test_resize.py
new file mode 100644
index 00000000000000..3106ed52a778a4
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_resize.py
@@ -0,0 +1,676 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from mo.front.common.partial_infer.utils import int64_array
+from mo.middle.passes.convert_data_type import data_type_str_to_np, np_data_type_to_destination_type
+from unit_tests.utils.graph import build_graph
+
+
+class TestResize(OnnxRuntimeLayerTest):
+    def create_resize_net(self, input_shape, output_shape, scales, sizes,
+                          coordinate_transformation_mode, cubic_coeff_a, mode,
+                          nearest_mode, precision, ir_version):
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input_rank = len(input_shape)
+
+        roi_node = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['roi'],
+            value=helper.make_tensor(
+                name='roi_consts',
+                data_type=TensorProto.FLOAT,
+                dims=[2 * input_rank],
+                vals=np.array([*np.zeros(input_rank), *np.ones(input_rank)])
+            )
+        )
+
+        onnx_scales = scales
+        if scales is None:
+            onnx_scales = np.array(output_shape).astype(np.float) / np.array(input_shape).astype(np.float)
+        scales_node = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['scales'],
+            value=helper.make_tensor(
+                name='scales_const',
+                data_type=TensorProto.FLOAT,
+                dims=[len(output_shape)],
+                vals=onnx_scales
+            )
+        )
+
+        nodes_list = [roi_node, scales_node]
+        inputs_list = ['input', 'roi', 'scales']
+
+        if sizes is not None:
+            sizes_node = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['sizes'],
+                value=helper.make_tensor(
+                    name='sizes_const',
+                    data_type=TensorProto.INT64,
+                    dims=[len(output_shape)],
+                    vals=sizes
+                )
+            )
+
+            nodes_list.append(sizes_node)
+            inputs_list.append('sizes')
+
+        args = dict()
+
+        onnx_mode = mode or 'nearest'
+        onnx_nearest_mode = nearest_mode or 'round_prefer_floor'
+        cube_coeff = -0.75 if cubic_coeff_a is None else cubic_coeff_a
+        onnx_coordinate_transformation_mode = coordinate_transformation_mode or 'half_pixel'
+
+        args['nearest_mode'] = onnx_nearest_mode
+        args['mode'] = onnx_mode
+        args['cubic_coeff_a'] = cube_coeff
+        args['coordinate_transformation_mode'] = onnx_coordinate_transformation_mode
+
+        x = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        y = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        resize_node = onnx.helper.make_node(
+            'Resize',
+            inputs=inputs_list,
+            outputs=['output'],
+            **args,
+        )
+
+        nodes_list.append(resize_node)
+
+        graph_def = onnx.helper.make_graph(nodes_list, 'test_model', [x], [y])
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+        onnx.checker.check_model(onnx_net)
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            if sizes is None and scales is None:
+                return onnx_net, ref_net
+
+            input_shape_as_array = int64_array(input_shape)
+
+            if sizes is not None and scales is not None:
+                shape_calculation_mode = 'sizes'
+                sizes_value = int64_array(sizes)
+                scales_value = np.array(scales).astype(np.float)
+            elif sizes is not None and scales is None:
+                shape_calculation_mode = 'sizes'
+                sizes_value = int64_array(sizes)
+                scales_value = sizes_value / input_shape_as_array
+            else:
+                shape_calculation_mode = 'scales'
+                scales_value = np.array(scales).astype(np.float)
+                sizes_value = np.floor(input_shape_as_array * scales_value + 1e-5).astype(np.int64)
+
+            if precision == 'FP16':
+                sizes_value = sizes_value.astype(np.float16)
+                scales_value = scales_value.astype(np.float16)
+
+            interp_mode = convert_onnx_mode(onnx_mode)
+
+            interp_attrs = {
+                'type': 'Interpolate',
+                'kind': 'op',
+                'mode': interp_mode,
+                'shape_calculation_mode': shape_calculation_mode,
+                'coordinate_transformation_mode': onnx_coordinate_transformation_mode,
+                'nearest_mode': onnx_nearest_mode,
+                'antialias': 0,
+                'cube_coeff': cube_coeff,
+                'pads_begin': np.zeros(input_rank).astype(np.int64),
+                'pads_end': np.zeros(input_rank).astype(np.int64),
+                'version': 'opset4'
+            }
+
+            if shape_calculation_mode == 'scales':
+                ref_net = create_ref_net_in_scales_mode(precision, input_shape_as_array, output_shape,
+                                                        sizes_value, scales_value, interp_attrs)
+            else:
+                ref_net = create_ref_net_in_sizes_mode(precision, input_shape_as_array, output_shape,
+                                                       sizes_value, scales_value, interp_attrs)
+
+        return onnx_net, ref_net
+
+    test_data = [
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 3, 3],
+             scales=[1.0, 1.0, 0.8, 0.8], sizes=None,
+             coordinate_transformation_mode='half_pixel',
+             cubic_coeff_a=None, mode='cubic', nearest_mode=None),
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 3, 3],
+             scales=[1.0, 1.0, 0.8, 0.8], sizes=None,
+             coordinate_transformation_mode='align_corners',
+             cubic_coeff_a=None, mode='cubic', nearest_mode=None),
+        dict(input_shape=[1, 1, 2, 4], output_shape=[1, 1, 1, 2],
+             scales=[1.0, 1.0, 0.6, 0.6], sizes=None,
+             coordinate_transformation_mode=None,
+             cubic_coeff_a=None, mode='linear', nearest_mode=None),
+        dict(input_shape=[1, 1, 2, 4], output_shape=[1, 1, 1, 2],
+             scales=[1.0, 1.0, 0.6, 0.6], sizes=None,
+             coordinate_transformation_mode='align_corners',
+             cubic_coeff_a=None, mode='linear', nearest_mode=None),
+        dict(input_shape=[1, 1, 2, 4], output_shape=[1, 1, 1, 2],
+             scales=[1.0, 1.0, 0.6, 0.6], sizes=None,
+             coordinate_transformation_mode=None,
+             cubic_coeff_a=None, mode='nearest', nearest_mode=None),
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 8, 8],
+             scales=[1.0, 1.0, 2.0, 2.0], sizes=None,
+             coordinate_transformation_mode=None,
+             cubic_coeff_a=None, mode='cubic', nearest_mode=None),
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 8, 8],
+             scales=[1.0, 1.0, 2.0, 2.0], sizes=None,
+             coordinate_transformation_mode='align_corners',
+             cubic_coeff_a=None, mode='cubic', nearest_mode=None),
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 8, 8],
+             scales=[1.0, 1.0, 2.0, 2.0], sizes=None,
+             coordinate_transformation_mode='asymmetric',
+             cubic_coeff_a=None, mode='cubic', nearest_mode=None),
+        dict(input_shape=[1, 1, 2, 2], output_shape=[1, 1, 4, 4],
+             scales=[1.0, 1.0, 2.0, 2.0], sizes=None,
+             coordinate_transformation_mode=None,
+             cubic_coeff_a=None, mode='linear', nearest_mode=None),
+        dict(input_shape=[1, 1, 2, 2], output_shape=[1, 1, 4, 4],
+             scales=[1.0, 1.0, 2.0, 2.0], sizes=None,
+             coordinate_transformation_mode='align_corners',
+             cubic_coeff_a=None, mode='linear', nearest_mode=None),
+        dict(input_shape=[1, 1, 2, 2], output_shape=[1, 1, 4, 4],
+             scales=[1.0, 1.0, 2.0, 2.0], sizes=None,
+             coordinate_transformation_mode=None,
+             cubic_coeff_a=None, mode='nearest', nearest_mode=None)
+    ]
+
+    @pytest.mark.parametrize("params", test_data)
+    def test_resize(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_resize_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, custom_eps=2.0e-4, temp_dir=temp_dir)
+
+    test_data_cubic = [
+        dict(input_shape=[1, 3, 100, 200], output_shape=[1, 3, 350, 150],
+             scales=[1.0, 1.0, 3.5, 150 / 200], sizes=None),
+        dict(input_shape=[16, 7, 190, 400], output_shape=[16, 7, 390, 600],
+             scales=[1.0, 1.0, 390 / 190, 600 / 400], sizes=None),
+        dict(input_shape=[4, 33, 1024, 800], output_shape=[4, 33, 512, 800],
+             scales=[1.0, 1.0, 0.5, 1.0], sizes=None),
+        dict(input_shape=[4, 33, 3, 800], output_shape=[4, 33, 1, 800],
+             scales=[1.0, 1.0, 0.3333334, 1.0], sizes=None),
+        dict(input_shape=[100, 200], output_shape=[350, 150],
+             scales=[3.5, 150 / 200], sizes=None),
+        dict(input_shape=[190, 400], output_shape=[390, 600],
+             scales=[390 / 190, 600 / 400], sizes=None),
+        dict(input_shape=[1024, 800], output_shape=[512, 800],
+             scales=[0.5, 1.0], sizes=None),
+        dict(input_shape=[3, 800], output_shape=[1, 800],
+             scales=[0.3333334, 1.0], sizes=None)
+    ]
+
+    @pytest.mark.parametrize("params", test_data_cubic)
+    @pytest.mark.parametrize("coordinate_transformation_mode",
+                             ['half_pixel', 'pytorch_half_pixel', 'align_corners',
+                              'asymmetric', 'tf_half_pixel_for_nn'])
+    @pytest.mark.parametrize("cubic_coeff_a", [-0.75])
+    @pytest.mark.parametrize("mode", ['cubic'])
+    @pytest.mark.parametrize("nearest_mode", ['round_prefer_floor'])
+    def test_resize_combined_cubic(self, params, coordinate_transformation_mode, cubic_coeff_a, mode,
+                                   nearest_mode, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_resize_net(**params,
+                                           coordinate_transformation_mode=coordinate_transformation_mode,
+                                           cubic_coeff_a=cubic_coeff_a, mode=mode, nearest_mode=nearest_mode,
+                                           precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, custom_eps=2.6e-2, temp_dir=temp_dir)
+
+    test_data_nearest = [
+        dict(input_shape=[1, 3, 100, 200], output_shape=[1, 3, 350, 150],
+             scales=[1.0, 1.0, 3.5, 150 / 200], sizes=None),
+        dict(input_shape=[16, 7, 190, 400], output_shape=[16, 7, 390, 600],
+             scales=[1.0, 1.0, 390 / 190, 600 / 400], sizes=None),
+        dict(input_shape=[4, 33, 600, 800], output_shape=[4, 33, 300, 800],
+             scales=[1.0, 1.0, 0.5, 1.0], sizes=None),
+        dict(input_shape=[4, 33, 3, 800], output_shape=[4, 33, 1, 800],
+             scales=[1.0, 1.0, 0.3333334, 1.0], sizes=None),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_nearest)
+    @pytest.mark.parametrize("coordinate_transformation_mode",
+                             ['half_pixel', 'pytorch_half_pixel', 'align_corners',
+                              'asymmetric', 'tf_half_pixel_for_nn'])
+    @pytest.mark.parametrize("cubic_coeff_a", [-0.75])
+    @pytest.mark.parametrize("mode", ['nearest'])
+    @pytest.mark.parametrize("nearest_mode", ['round_prefer_floor', 'round_prefer_ceil',
+                                              'floor', 'ceil'])
+    def test_resize_combined_nearest(self, params, coordinate_transformation_mode, cubic_coeff_a, mode,
+                                     nearest_mode, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_resize_net(**params,
+                                           coordinate_transformation_mode=coordinate_transformation_mode,
+                                           cubic_coeff_a=cubic_coeff_a, mode=mode, nearest_mode=nearest_mode,
+                                           precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_linear = [
+        dict(input_shape=[1, 3, 100, 200], output_shape=[1, 3, 350, 150],
+             scales=[1.0, 1.0, 3.5, 150 / 200], sizes=None),
+        dict(input_shape=[16, 7, 190, 400], output_shape=[16, 7, 390, 600],
+             scales=[1.0, 1.0, 390 / 190, 600 / 400], sizes=None),
+        dict(input_shape=[4, 33, 600, 800], output_shape=[4, 33, 300, 800],
+             scales=[1.0, 1.0, 0.5, 1.0], sizes=None),
+        dict(input_shape=[4, 33, 3, 800], output_shape=[4, 33, 1, 800],
+             scales=[1.0, 1.0, 0.3333334, 1.0], sizes=None),
+        dict(input_shape=[100, 200], output_shape=[350, 150],
+             scales=[3.5, 150 / 200], sizes=None),
+        dict(input_shape=[190, 400], output_shape=[390, 600],
+             scales=[390 / 190, 600 / 400], sizes=None),
+        dict(input_shape=[600, 800], output_shape=[300, 800],
+             scales=[0.5, 1.0], sizes=None),
+        dict(input_shape=[3, 800], output_shape=[1, 800],
+             scales=[0.3333334, 1.0], sizes=None),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_linear)
+    @pytest.mark.parametrize("coordinate_transformation_mode",
+                             ['half_pixel', 'pytorch_half_pixel', 'align_corners',
+                              'asymmetric', 'tf_half_pixel_for_nn'])
+    @pytest.mark.parametrize("cubic_coeff_a", [-0.75])
+    @pytest.mark.parametrize("mode", ['linear'])
+    @pytest.mark.parametrize("nearest_mode", ['round_prefer_floor'])
+    def test_resize_combined_linear(self, params, coordinate_transformation_mode, cubic_coeff_a, mode,
+                                    nearest_mode, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_resize_net(**params,
+                                           coordinate_transformation_mode=coordinate_transformation_mode,
+                                           cubic_coeff_a=cubic_coeff_a, mode=mode, nearest_mode=nearest_mode,
+                                           precision=precision, ir_version=ir_version),
+        ie_device, precision, ir_version, custom_eps=2.0e-2, temp_dir=temp_dir)
+
+    test_data_sizes = [
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 3, 3],
+             scales=None, sizes=[1, 1, 3, 3],
+             coordinate_transformation_mode=None,
+             cubic_coeff_a=None, mode='cubic', nearest_mode=None),
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 3, 1],
+             scales=None, sizes=[1, 1, 3, 1],
+             coordinate_transformation_mode='pytorch_half_pixel',
+             cubic_coeff_a=None, mode='linear', nearest_mode=None),
+        dict(input_shape=[1, 1, 2, 4], output_shape=[1, 1, 1, 3],
+             scales=None, sizes=[1, 1, 1, 3],
+             coordinate_transformation_mode=None,
+             cubic_coeff_a=None, mode='nearest', nearest_mode=None),
+        dict(input_shape=[1, 1, 2, 4], output_shape=[1, 1, 1, 2],
+             scales=None, sizes=[1, 1, 1, 2],
+             coordinate_transformation_mode=None,
+             cubic_coeff_a=None, mode='nearest', nearest_mode=None),
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 3, 2],
+             scales=None, sizes=[1, 1, 3, 2],
+             coordinate_transformation_mode='tf_half_pixel_for_nn',
+             cubic_coeff_a=None, mode='nearest', nearest_mode=None),
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 9, 10],
+             scales=None, sizes=[1, 1, 9, 10],
+             coordinate_transformation_mode=None,
+             cubic_coeff_a=None, mode='cubic', nearest_mode=None),
+        dict(input_shape=[1, 1, 2, 2], output_shape=[1, 1, 7, 8],
+             scales=None, sizes=[1, 1, 7, 8],
+             coordinate_transformation_mode=None,
+             cubic_coeff_a=None, mode='nearest', nearest_mode=None),
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 8, 8],
+             scales=None, sizes=[1, 1, 8, 8],
+             coordinate_transformation_mode='half_pixel',
+             cubic_coeff_a=None, mode='nearest', nearest_mode='ceil'),
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 8, 8],
+             scales=None, sizes=[1, 1, 8, 8],
+             coordinate_transformation_mode='align_corners',
+             cubic_coeff_a=None, mode='nearest', nearest_mode='floor'),
+        dict(input_shape=[1, 1, 4, 4], output_shape=[1, 1, 8, 8],
+             scales=None, sizes=[1, 1, 8, 8],
+             coordinate_transformation_mode='asymmetric',
+             cubic_coeff_a=None, mode='nearest', nearest_mode='round_prefer_ceil'),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_sizes)
+    def test_resize_sizes(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_resize_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_sizes_cubic = [
+        dict(input_shape=[1, 3, 100, 200], output_shape=[1, 3, 350, 150],
+             scales=None, sizes=[1, 3, 350, 150]),
+        dict(input_shape=[16, 7, 190, 400], output_shape=[16, 7, 390, 600],
+             scales=None, sizes=[16, 7, 390, 600]),
+        dict(input_shape=[4, 15, 700, 800], output_shape=[4, 15, 350, 800],
+             scales=None, sizes=[4, 15, 350, 800]),
+        dict(input_shape=[4, 15, 3, 200], output_shape=[4, 15, 1, 200],
+             scales=None, sizes=[4, 15, 1, 200]),
+        dict(input_shape=[100, 200], output_shape=[350, 150],
+             scales=None, sizes=[350, 150]),
+        dict(input_shape=[190, 400], output_shape=[390, 600],
+             scales=None, sizes=[390, 600]),
+        dict(input_shape=[700, 800], output_shape=[350, 800],
+             scales=None, sizes=[350, 800]),
+        dict(input_shape=[3, 200], output_shape=[1, 200],
+             scales=None, sizes=[1, 200]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_sizes_cubic)
+    @pytest.mark.parametrize("coordinate_transformation_mode",
+                             ['half_pixel', 'pytorch_half_pixel', 'align_corners',
+                              'asymmetric', 'tf_half_pixel_for_nn'])
+    @pytest.mark.parametrize("cubic_coeff_a", [-0.75])
+    @pytest.mark.parametrize("mode", ['cubic'])
+    @pytest.mark.parametrize("nearest_mode", ['round_prefer_floor'])
+    def test_resize_combined_sizes_cubic(self, params, coordinate_transformation_mode, cubic_coeff_a, mode,
+                                         nearest_mode, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_resize_net(**params,
+                                           coordinate_transformation_mode=coordinate_transformation_mode,
+                                           cubic_coeff_a=cubic_coeff_a, mode=mode, nearest_mode=nearest_mode,
+                                           precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, custom_eps=2.6e-2, temp_dir=temp_dir)
+
+    test_data_sizes_nearest = [
+        dict(input_shape=[1, 3, 100, 200], output_shape=[1, 3, 350, 150],
+             scales=None, sizes=[1, 3, 350, 150]),
+        dict(input_shape=[16, 7, 190, 400], output_shape=[16, 7, 390, 600],
+             scales=None, sizes=[16, 7, 390, 600]),
+        dict(input_shape=[4, 33, 600, 800], output_shape=[4, 33, 300, 800],
+             scales=None, sizes=[4, 33, 300, 800]),
+        dict(input_shape=[4, 33, 3, 800], output_shape=[4, 33, 1, 800],
+             scales=None, sizes=[4, 33, 1, 800]),
+        dict(input_shape=[3, 100, 200], output_shape=[3, 350, 150],
+             scales=None, sizes=[3, 350, 150]),
+        dict(input_shape=[7, 190, 400], output_shape=[7, 390, 600],
+             scales=None, sizes=[7, 390, 600]),
+        dict(input_shape=[33, 600, 800], output_shape=[33, 300, 800],
+             scales=None, sizes=[33, 300, 800]),
+        dict(input_shape=[33, 3, 800], output_shape=[33, 1, 800],
+             scales=None, sizes=[33, 1, 800]),
+        dict(input_shape=[100, 200], output_shape=[350, 150],
+             scales=None, sizes=[350, 150]),
+        dict(input_shape=[190, 400], output_shape=[390, 600],
+             scales=None, sizes=[390, 600]),
+        dict(input_shape=[600, 800], output_shape=[300, 800],
+             scales=None, sizes=[300, 800]),
+        dict(input_shape=[3, 800], output_shape=[1, 800],
+             scales=None, sizes=[1, 800]),
+        dict(input_shape=[100], output_shape=[350],
+             scales=None, sizes=[350]),
+        dict(input_shape=[190], output_shape=[390],
+             scales=None, sizes=[390]),
+        dict(input_shape=[600], output_shape=[300],
+             scales=None, sizes=[300]),
+        dict(input_shape=[3], output_shape=[1],
+             scales=None, sizes=[1]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_sizes_nearest)
+    @pytest.mark.parametrize("coordinate_transformation_mode",
+                             ['half_pixel', 'pytorch_half_pixel', 'align_corners',
+                              'asymmetric', 'tf_half_pixel_for_nn'])
+    @pytest.mark.parametrize("cubic_coeff_a", [-0.75])
+    @pytest.mark.parametrize("mode", ['nearest'])
+    @pytest.mark.parametrize("nearest_mode", ['round_prefer_floor', 'round_prefer_ceil',
+                                              'floor', 'ceil'])
+    def test_resize_combined_sizes_nearest(self, params, coordinate_transformation_mode, cubic_coeff_a, mode,
+                                           nearest_mode, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_resize_net(**params,
+                                           coordinate_transformation_mode=coordinate_transformation_mode,
+                                           cubic_coeff_a=cubic_coeff_a, mode=mode, nearest_mode=nearest_mode,
+                                           precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_sizes_linear = [
+        dict(input_shape=[1, 3, 100, 200], output_shape=[1, 3, 350, 150],
+             scales=None, sizes=[1, 3, 350, 150]),
+        dict(input_shape=[16, 7, 190, 400], output_shape=[16, 7, 390, 600],
+             scales=None, sizes=[16, 7, 390, 600]),
+        dict(input_shape=[4, 33, 600, 800], output_shape=[4, 33, 300, 800],
+             scales=None, sizes=[4, 33, 300, 800]),
+        dict(input_shape=[4, 33, 3, 800], output_shape=[4, 33, 1, 800],
+             scales=None, sizes=[4, 33, 1, 800]),
+        dict(input_shape=[100, 200], output_shape=[350, 150],
+             scales=None, sizes=[350, 150]),
+        dict(input_shape=[190, 400], output_shape=[390, 600],
+             scales=None, sizes=[390, 600]),
+        dict(input_shape=[600, 800], output_shape=[300, 800],
+             scales=None, sizes=[300, 800]),
+        dict(input_shape=[3, 800], output_shape=[1, 800],
+             scales=None, sizes=[1, 800]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_sizes_linear)
+    @pytest.mark.parametrize("coordinate_transformation_mode",
+                             ['half_pixel', 'pytorch_half_pixel', 'align_corners',
+                              'asymmetric', 'tf_half_pixel_for_nn'])
+    @pytest.mark.parametrize("cubic_coeff_a", [-0.75])
+    @pytest.mark.parametrize("mode", ['linear'])
+    @pytest.mark.parametrize("nearest_mode", ['round_prefer_floor'])
+    def test_resize_combined_sizes_linear(self, params, coordinate_transformation_mode, cubic_coeff_a, mode,
+                                          nearest_mode, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_resize_net(**params,
+                                           coordinate_transformation_mode=coordinate_transformation_mode,
+                                           cubic_coeff_a=cubic_coeff_a, mode=mode, nearest_mode=nearest_mode,
+                                           precision=precision, ir_version=ir_version),
+        ie_device, precision, ir_version, custom_eps=2.0e-2, temp_dir=temp_dir)
+
+
+def create_ref_net_in_sizes_mode(precision, input_shape, output_shape, sizes_value, scales_value, attrs):
+    input_data_type = np_data_type_to_destination_type(data_type_str_to_np(precision))
+    input_rank = len(input_shape)
+    epsilon = np.array([1.0e-5])
+    spatial_dims = spatial_dimensions(input_shape)
+    begin_dim = spatial_dims[0]
+    end_dim = input_rank
+
+    spatial_sizes_value = sizes_value[spatial_dims]
+
+    nodes_attrs = {
+        'input': {'kind': 'op', 'type': 'Parameter'},
+        'input_data': {'shape': input_shape, 'kind': 'data'},
+        'shape_of': {'kind': 'op', 'type': 'ShapeOf'},
+        'shape_of_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'shape_to_float': {'kind': 'op', 'type': 'Convert', 'destination_type': input_data_type},
+        'shape_to_float_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'div': {'kind': 'op', 'type': 'Divide'},
+        'div_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'div_sizes_const_data': {'kind': 'data', 'value': sizes_value},
+        'div_sizes_const': {'kind': 'op', 'type': 'Const'},
+        'div_sizes_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'eps_const_data': {'kind': 'data', 'value': epsilon},
+        'eps_const': {'kind': 'op', 'type': 'Const'},
+        'eps_data': {'shape': int64_array([1]), 'kind': 'data'},
+        'add': {'kind': 'op', 'type': 'Add'},
+        'add_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'ss_scales': {
+            'kind': 'op', 'type': 'StridedSlice', 'begin_mask': 0,
+            'end_mask': 0, 'new_axis_mask': 0,
+            'shrink_axis_mask': 0, 'ellipsis_mask': 0
+        },
+        'ss_scales_data': {'shape': int64_array([len(spatial_sizes_value)]), 'kind': 'data'},
+        'ss_scales_begin_const_data': {'kind': 'data', 'value': int64_array([begin_dim])},
+        'ss_scales_begin_const': {'kind': 'op', 'type': 'Const'},
+        'ss_scales_begin_data': {'shape': int64_array([1]), 'kind': 'data'},
+        'ss_scales_end_const_data': {'kind': 'data', 'value': int64_array([end_dim])},
+        'ss_scales_end_const': {'kind': 'op', 'type': 'Const'},
+        'ss_scales_end_data': {'shape': int64_array([1]), 'kind': 'data'},
+        'ss_scales_stride_const_data': {'kind': 'data', 'value': int64_array([1])},
+        'ss_scales_stride_const': {'kind': 'op', 'type': 'Const'},
+        'ss_scales_stride_data': {'shape': int64_array([1]), 'kind': 'data'},
+        'sizes_const_data': {'kind': 'data', 'value': spatial_sizes_value},
+        'sizes_const': {'kind': 'op', 'type': 'Const'},
+        'sizes_data': {'shape': int64_array([len(spatial_sizes_value)]), 'kind': 'data'},
+        'axes_const_data': {'kind': 'data', 'value': spatial_dims},
+        'axes_const': {'kind': 'op', 'type': 'Const'},
+        'axes_data': {'shape': int64_array([len(spatial_dims)]), 'kind': 'data'},
+        'interpolate': attrs,
+        'interpolate_data': {'shape': output_shape, 'kind': 'data'},
+        'result': {'kind': 'op', 'type': 'Result'},
+    }
+    edges = [
+        ('input', 'input_data'),
+        ('input_data', 'interpolate', {'in': 0, 'out': 0}),
+        ('input_data', 'shape_of', {'in': 0, 'out': 0}),
+        ('shape_of', 'shape_of_data'),
+        ('shape_of_data', 'shape_to_float'),
+        ('shape_to_float', 'shape_to_float_data'),
+        ('shape_to_float_data', 'div', {'in': 1}),
+        ('div_sizes_const_data', 'div_sizes_const'),
+        ('div_sizes_const', 'div_sizes_data'),
+        ('div_sizes_data', 'div', {'in': 0}),
+        ('div', 'div_data'),
+        ('eps_const_data', 'eps_const'),
+        ('eps_const', 'eps_data'),
+        ('div_data', 'add', {'in': 0}),
+        ('eps_data', 'add', {'in': 1}),
+        ('add', 'add_data'),
+        ('add_data', 'ss_scales', {'in': 0}),
+        ('ss_scales', 'ss_scales_data'),
+        ('ss_scales_begin_const_data', 'ss_scales_begin_const'),
+        ('ss_scales_begin_const', 'ss_scales_begin_data'),
+        ('ss_scales_begin_data', 'ss_scales', {'in': 1}),
+        ('ss_scales_end_const_data', 'ss_scales_end_const'),
+        ('ss_scales_end_const', 'ss_scales_end_data'),
+        ('ss_scales_end_data', 'ss_scales', {'in': 2}),
+        ('ss_scales_stride_const_data', 'ss_scales_stride_const'),
+        ('ss_scales_stride_const', 'ss_scales_stride_data'),
+        ('ss_scales_stride_data', 'ss_scales', {'in': 3}),
+        ('ss_scales_data', 'interpolate', {'in': 2}),
+        ('sizes_const_data', 'sizes_const'),
+        ('sizes_const', 'sizes_data'),
+        ('sizes_data', 'interpolate', {'in': 1}),
+        ('axes_const_data', 'axes_const'),
+        ('axes_const', 'axes_data'),
+        ('axes_data', 'interpolate', {'in': 3}),
+        ('interpolate', 'interpolate_data'),
+        ('interpolate_data', 'result')
+    ]
+
+    return build_graph(nodes_attrs, edges)
+
+
+def create_ref_net_in_scales_mode(precision, input_shape, output_shape, sizes_value, scales_value, attrs):
+    input_data_type = np_data_type_to_destination_type(data_type_str_to_np(precision))
+    input_rank = len(input_shape)
+    epsilon = np.array([1.0e-5])
+    spatial_dims = spatial_dimensions(input_shape)
+    begin_dim = spatial_dims[0]
+    end_dim = input_rank
+
+    spatial_scales_value = scales_value[spatial_dims]
+
+    nodes_attrs = {
+        'input': {'kind': 'op', 'type': 'Parameter'},
+        'input_data': {'shape': input_shape, 'kind': 'data'},
+        'shape_of': {'kind': 'op', 'type': 'ShapeOf'},
+        'shape_of_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'shape_to_float': {'kind': 'op', 'type': 'Convert', 'destination_type': input_data_type},
+        'shape_to_float_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'mul': {'kind': 'op', 'type': 'Multiply'},
+        'mul_scales_const_data': {'kind': 'data', 'value': scales_value},
+        'mul_scales_const': {'kind': 'op', 'type': 'Const'},
+        'mul_scales_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'mul_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'eps_const_data': {'kind': 'data', 'value': epsilon},
+        'eps_const': {'kind': 'op', 'type': 'Const'},
+        'eps_data': {'shape': int64_array([1]), 'kind': 'data'},
+        'add': {'kind': 'op', 'type': 'Add'},
+        'add_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'floor': {'type': 'Floor', 'kind': 'op'},
+        'floor_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'to_int': {'kind': 'op', 'type': 'Convert', 'destination_type': 'i64'},
+        'to_int_data': {'shape': int64_array([input_rank]), 'kind': 'data'},
+        'strided_slice': {
+            'kind': 'op', 'type': 'StridedSlice', 'begin_mask': 0,
+            'end_mask': 0, 'new_axis_mask': 0,
+            'shrink_axis_mask': 0, 'ellipsis_mask': 0
+        },
+        'strided_slice_data': {'shape': int64_array([len(spatial_scales_value)]), 'kind': 'data'},
+        'begin_const_data': {'kind': 'data', 'value': int64_array([begin_dim])},
+        'begin_const': {'kind': 'op', 'type': 'Const'},
+        'begin_data': {'shape': int64_array([1]), 'kind': 'data'},
+        'end_const_data': {'kind': 'data', 'value': int64_array([end_dim])},
+        'end_const': {'kind': 'op', 'type': 'Const'},
+        'end_data': {'shape': int64_array([1]), 'kind': 'data'},
+        'stride_const_data': {'kind': 'data', 'value': int64_array([1])},
+        'stride_const': {'kind': 'op', 'type': 'Const'},
+        'stride_data': {'shape': int64_array([1]), 'kind': 'data'},
+        'scales_const_data': {'kind': 'data', 'value': spatial_scales_value},
+        'scales_const': {'kind': 'op', 'type': 'Const'},
+        'scales_data': {'shape': int64_array([len(spatial_scales_value)]), 'kind': 'data'},
+        'axes_const_data': {'kind': 'data', 'value': spatial_dims},
+        'axes_const': {'kind': 'op', 'type': 'Const'},
+        'axes_data': {'shape': int64_array([len(spatial_dims)]), 'kind': 'data'},
+        'interpolate': attrs,
+        'interpolate_data': {'shape': output_shape, 'kind': 'data'},
+        'result': {'kind': 'op', 'type': 'Result'},
+    }
+    edges = [
+        ('input', 'input_data'),
+        ('input_data', 'interpolate', {'in': 0, 'out': 0}),
+        ('input_data', 'shape_of', {'in': 0, 'out': 0}),
+        ('shape_of', 'shape_of_data'),
+        ('shape_of_data', 'shape_to_float'),
+        ('shape_to_float', 'shape_to_float_data'),
+        ('shape_to_float_data', 'mul', {'in': 0}),
+        ('mul_scales_const_data', 'mul_scales_const'),
+        ('mul_scales_const', 'mul_scales_data'),
+        ('mul_scales_data', 'mul', {'in': 1}),
+        ('mul', 'mul_data'),
+        ('eps_const_data', 'eps_const'),
+        ('eps_const', 'eps_data'),
+        ('mul_data', 'add', {'in': 0}),
+        ('eps_data', 'add', {'in': 1}),
+        ('add', 'add_data'),
+        ('add_data', 'floor'),
+        ('floor', 'floor_data'),
+        ('floor_data', 'to_int'),
+        ('to_int', 'to_int_data'),
+        ('to_int_data', 'strided_slice', {'in': 0}),
+        ('strided_slice', 'strided_slice_data'),
+        ('begin_const_data', 'begin_const'),
+        ('begin_const', 'begin_data'),
+        ('begin_data', 'strided_slice', {'in': 1}),
+        ('end_const_data', 'end_const'),
+        ('end_const', 'end_data'),
+        ('end_data', 'strided_slice', {'in': 2}),
+        ('stride_const_data', 'stride_const'),
+        ('stride_const', 'stride_data'),
+        ('stride_data', 'strided_slice', {'in': 3}),
+        ('strided_slice_data', 'interpolate', {'in': 1}),
+        ('scales_const_data', 'scales_const'),
+        ('scales_const', 'scales_data'),
+        ('scales_data', 'interpolate', {'in': 2}),
+        ('axes_const_data', 'axes_const'),
+        ('axes_const', 'axes_data'),
+        ('axes_data', 'interpolate', {'in': 3}),
+        ('interpolate', 'interpolate_data'),
+        ('interpolate_data', 'result')
+    ]
+
+    return build_graph(nodes_attrs, edges)
+
+
+def spatial_dimensions(shape):
+    rank = len(shape)
+    if rank >= 4:
+        return np.arange(2, rank)
+    elif rank in [1, 2]:
+        return np.arange(0, rank)
+    else:
+        return np.arange(1, rank)
+
+
+def convert_onnx_mode(mode: str) -> str:
+    return {'nearest': 'nearest', 'linear': 'linear_onnx', 'cubic': 'cubic'}[mode]
diff --git a/tests/layer_tests/onnx_tests/test_roi_align.py b/tests/layer_tests/onnx_tests/test_roi_align.py
new file mode 100644
index 00000000000000..b7b89dd05b993d
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_roi_align.py
@@ -0,0 +1,109 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestROIAlign(OnnxRuntimeLayerTest):
+    def create_net(self, input_shape, rois_shape, indices_shape, output_shape,
+                   pooled_h, pooled_w, mode, sampling_ratio, spatial_scale, ir_version):
+        """
+            ONNX net                    IR net
+
+            Input->ROIAlign->Output   =>    Parameter->ROIAlign->Result
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        rois = helper.make_tensor_value_info('rois', TensorProto.FLOAT, rois_shape)
+        indices = helper.make_tensor_value_info('indices', TensorProto.FLOAT, indices_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_def = onnx.helper.make_node(
+            'ROIAlign',
+            inputs=['input', 'rois', 'indices'],
+            outputs=['output'],
+            **{'output_height': pooled_h, 'output_width': pooled_w, 'mode': mode,
+               'sampling_ratio': sampling_ratio, 'spatial_scale': spatial_scale},
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input, rois, indices],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                # comparison in these tests starts from input node, as we have 3 of them IREngine gets confused
+                # and takes the first input node in inputs list sorted by lexicographical order
+                '1_input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': input_shape, 'kind': 'data'},
+
+                '2_rois': {'kind': 'op', 'type': 'Parameter'},
+                'rois_data': {'shape': rois_shape, 'kind': 'data'},
+
+                '3_indices': {'kind': 'op', 'type': 'Parameter'},
+                'indices_data': {'shape': indices_shape, 'kind': 'data'},
+
+                'node': {'kind': 'op', 'type': 'ROIAlign', 'pooled_h': pooled_h, 'pooled_w': pooled_w,
+                         'mode': mode, 'sampling_ratio': sampling_ratio, 'spatial_scale': spatial_scale},
+                'node_data': {'shape': output_shape, 'kind': 'data'},
+
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [
+                                      ('1_input', 'input_data'),
+                                      ('input_data', 'node', {'in': 0}),
+                                      ('2_rois', 'rois_data'),
+                                      ('rois_data', 'node', {'in': 1}),
+                                      ('3_indices', 'indices_data'),
+                                      ('indices_data', 'node', {'in': 2}),
+
+                                      ('node', 'node_data'),
+                                      ('node_data', 'result')
+                                  ])
+        return onnx_net, ref_net
+
+    test_data = [
+        dict(input_shape=[1, 256, 200, 272], rois_shape=[1000, 4], indices_shape=[1000],
+             pooled_h=7, pooled_w=7, mode="avg", sampling_ratio=2, spatial_scale=0.25,
+             output_shape=[1000, 256, 7, 7]),
+        dict(input_shape=[7, 256, 200, 200], rois_shape=[1000, 4], indices_shape=[1000],
+             pooled_h=6, pooled_w=6, mode="max", sampling_ratio=2, spatial_scale=16.0,
+             output_shape=[1000, 256, 6, 6]),
+        dict(input_shape=[7, 256, 200, 200], rois_shape=[1000, 4], indices_shape=[1000],
+             pooled_h=5, pooled_w=6, mode="max", sampling_ratio=2, spatial_scale=16.0,
+             output_shape=[1000, 256, 5, 6]),
+
+    ]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_roi_align(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_scale.py b/tests/layer_tests/onnx_tests/test_scale.py
new file mode 100644
index 00000000000000..a624a19596b317
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_scale.py
@@ -0,0 +1,144 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+
+
+class TestScale(Caffe2OnnxLayerTest):
+    def create_net(self, shape, scale, ir_version):
+        """
+            ONNX net                     IR net
+
+            Input->Scale->Output   =>    Input->Power
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Scale',
+            inputs=['input'],
+            outputs=['output'],
+            scale=scale
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, scale, precision, ir_version):
+        """
+            ONNX net                                     IR net
+
+            Input->Concat(+scaled const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.randint(-127, 127, shape).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Scale',
+            inputs=['const1'],
+            outputs=['scale'],
+            scale=scale
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'scale'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        ir_const = constant.flatten() * scale
+        if precision == 'FP16':
+            ir_const = ir_const.astype(np.float16)
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[10, 12], scale=0.1),
+                 dict(shape=[8, 10, 12], scale=0.9),
+                 dict(shape=[6, 8, 10, 12], scale=1.5),
+                 dict(shape=[4, 6, 8, 10, 12], scale=4.5)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_scale(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_scale_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_scatter.py b/tests/layer_tests/onnx_tests/test_scatter.py
new file mode 100644
index 00000000000000..eda9fc29dccdfb
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_scatter.py
@@ -0,0 +1,125 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestScatters(Caffe2OnnxLayerTest):
+    op = None
+
+    def create_net(self, input_shape, indices_shape, updates_shape, output_shape,
+                   axis, ir_version):
+        """
+            ONNX net                    IR net
+
+            Input->Scatter->Output   =>    Parameter->ScatterElementsUpdate->Result
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        data = helper.make_tensor_value_info('data', TensorProto.FLOAT, input_shape)
+        indices = helper.make_tensor_value_info('indices', TensorProto.INT64, indices_shape)
+        updates = helper.make_tensor_value_info('updates', TensorProto.FLOAT, indices_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        params = {'axis': axis} if axis is not None else {}
+        node_def = onnx.helper.make_node(
+            self.op,
+            inputs=['data', 'indices', 'updates'],
+            outputs=['output'],
+            **params,
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [data, indices, updates],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                # comparison in these tests starts from input node, as we have 3 of them IREngine gets confused
+                # and takes the first input node in inputs list sorted by lexicographical order
+                '1_input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': input_shape, 'kind': 'data'},
+
+                '2_indices': {'kind': 'op', 'type': 'Parameter'},
+                'indices_data': {'shape': indices_shape, 'kind': 'data'},
+
+                '3_updates': {'kind': 'op', 'type': 'Parameter'},
+                'updates_data': {'shape': updates_shape, 'kind': 'data'},
+
+                'const_indata': {'kind': 'data', 'value': np.int64(axis) if axis is not None else np.int64(0)},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'kind': 'data'},
+
+                'node': {'kind': 'op', 'type': 'ScatterElementsUpdate'},
+                'node_data': {'shape': output_shape, 'kind': 'data'},
+
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [
+                                      ('1_input', 'input_data'),
+                                      ('input_data', 'node', {'in': 0}),
+                                      ('2_indices', 'indices_data'),
+                                      ('indices_data', 'node', {'in': 1}),
+                                      ('3_updates', 'updates_data'),
+                                      ('updates_data', 'node', {'in': 2}),
+                                      ('const_indata', 'const'),
+                                      ('const', 'const_data'),
+                                      ('const_data', 'node', {'in': 3}),
+
+                                      ('node', 'node_data'),
+                                      ('node_data', 'result')
+                                  ])
+        return onnx_net, ref_net
+
+
+test_data = [
+    dict(input_shape=[1, 5], indices_shape=[1, 2], updates_shape=[1, 2],
+         axis=1, output_shape=[1, 5]),
+    dict(input_shape=[1, 256, 200, 272], indices_shape=[1, 256, 200, 272], updates_shape=[1, 256, 200, 272],
+         axis=None, output_shape=[1, 256, 200, 272])]
+
+
+class TestScatter(TestScatters):
+    op = 'Scatter'
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_scatter(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+
+class TestScatterElements(TestScatters):
+    op = 'ScatterElements'
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_scatter_elements(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_sigmoid.py b/tests/layer_tests/onnx_tests/test_sigmoid.py
new file mode 100644
index 00000000000000..bf0c517f68bee3
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_sigmoid.py
@@ -0,0 +1,196 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestSigmoid(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randn(*inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                       IR net
+
+            Input->Sigmoid->Output   =>    Input->sigmoid
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Sigmoid',
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'Sigmoid'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                      IR net
+
+            Input->Concat(+sigmoid const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        const_number = np.prod(shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+        constant = np.reshape(constant, shape)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Sigmoid',
+            inputs=['const1'],
+            outputs=['sigmoid1']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'sigmoid1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = 1.0 / (1.0 + np.exp(np.negative(constant)))
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [dict(shape=[2, 4, 6, 8, 10])]
+
+    test_data = [dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_sigmoid_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sigmoid(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sigmoid_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_sign.py b/tests/layer_tests/onnx_tests/test_sign.py
new file mode 100644
index 00000000000000..90794322561f1a
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_sign.py
@@ -0,0 +1,176 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestSign(OnnxRuntimeLayerTest):
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                    IR net
+
+            Input->Sign->Output   =>    Input->Sign
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Sign',
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'Sign'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                   IR net
+
+            Input->Concat(+sign const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.randn(*shape).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Sign',
+            inputs=['const1'],
+            outputs=['sign']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'sign'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.sign(constant)
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[12]),
+                 dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sign(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sign_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_slice.py b/tests/layer_tests/onnx_tests/test_slice.py
new file mode 100644
index 00000000000000..28b1786b5c27bd
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_slice.py
@@ -0,0 +1,399 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestSlice(OnnxRuntimeLayerTest):
+    def create_net(self, shape, axes, ends, starts, ir_version, opset=6, steps=None):
+        """
+            ONNX net                    IR net
+
+            Input->Slice->Output   =>    Input->Crop
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        # calculate output shape
+        test_arr = np.zeros(shape)
+        slice_idx = [None] * len(shape)
+        for i, axis in enumerate(axes):
+            slice_idx[axis] = slice(starts[i], ends[i], steps[i] if steps is not None else 1)
+        for axis, s in enumerate(slice_idx):
+            if s is None:
+                slice_idx[axis] = slice(0, shape[axis], 1)
+        test_arr = test_arr[tuple(slice_idx)]
+
+        output_shape = list(test_arr.shape)
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        nodes = list()
+        if opset < 10:
+            node_def = onnx.helper.make_node(
+                'Slice',
+                inputs=['input'],
+                outputs=['slice'],
+                starts=starts,
+                ends=ends,
+                axes=axes
+            )
+            nodes.append(node_def)
+        else:
+            node_starts_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['starts'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.INT64,
+                    dims=[len(starts)],
+                    vals=starts
+                )
+            )
+            node_ends_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['ends'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.INT64,
+                    dims=[len(ends)],
+                    vals=ends
+                )
+            )
+            node_axes_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['axes'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.INT64,
+                    dims=[len(axes)],
+                    vals=axes
+                )
+            )
+            inputs = ['input', 'starts', 'ends', 'axes']
+            if steps:
+                node_steps_def = onnx.helper.make_node(
+                    'Constant',
+                    inputs=[],
+                    outputs=['steps'],
+                    value=helper.make_tensor(
+                        name='const_tensor',
+                        data_type=TensorProto.INT64,
+                        dims=[len(steps)],
+                        vals=steps
+                    )
+                )
+                nodes.append(node_steps_def)
+                inputs.append('steps')
+
+            node_def = onnx.helper.make_node(
+                'Slice',
+                inputs=inputs,
+                outputs=['slice']
+            )
+            nodes.extend([node_starts_def, node_ends_def, node_axes_def, node_def])
+
+        elu_def = onnx.helper.make_node(
+            'Elu',
+            inputs=['slice'],
+            outputs=['output']
+        )
+        nodes.append(elu_def)
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            [input],
+            [output]
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, axes, ends, starts, ir_version, opset=6, steps=None):
+        """
+            ONNX net                                         IR net
+
+            Input->Concat(+sliced const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        # calculate output shape
+        constant = np.random.randint(-127, 127, shape).astype(np.float)
+
+        slice_idx = [None] * len(shape)
+        for i, axis in enumerate(axes):
+            slice_idx[axis] = slice(starts[i], ends[i], steps[i] if steps is not None else 1)
+
+        for axis, s in enumerate(slice_idx):
+            if s is None:
+                slice_idx[axis] = slice(0, shape[axis], 1)
+
+        constant_after = constant[tuple(slice_idx)]
+
+        output_shape = list(constant_after.shape)
+
+        concat_axis = 0
+        concat_output_shape = output_shape.copy()
+        concat_output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, output_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, concat_output_shape)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        nodes = [node_const_def]
+        if opset < 10:
+            node_def = onnx.helper.make_node(
+                'Slice',
+                inputs=['const1'],
+                outputs=['slice'],
+                starts=starts,
+                ends=ends,
+                axes=axes
+            )
+            nodes.append(node_def)
+        else:
+            node_starts_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['starts'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.INT64,
+                    dims=[len(starts)],
+                    vals=starts
+                )
+            )
+            node_ends_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['ends'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.INT64,
+                    dims=[len(ends)],
+                    vals=ends
+                )
+            )
+            node_axes_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['axes'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.INT64,
+                    dims=[len(axes)],
+                    vals=axes
+                )
+            )
+
+            inputs = ['const1', 'starts', 'ends', 'axes']
+            if steps:
+                node_steps_def = onnx.helper.make_node(
+                    'Constant',
+                    inputs=[],
+                    outputs=['steps'],
+                    value=helper.make_tensor(
+                        name='const_tensor',
+                        data_type=TensorProto.INT64,
+                        dims=[len(steps)],
+                        vals=steps
+                    )
+                )
+                nodes.append(node_steps_def)
+                inputs.append('steps')
+
+            node_def = onnx.helper.make_node(
+                'Slice',
+                inputs=inputs,
+                outputs=['slice']
+            )
+            nodes.extend([node_starts_def, node_ends_def, node_axes_def, node_def])
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'slice'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+        nodes.append(node_concat_def)
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_reshape_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data_no_steps = [
+        dict(shape=[10, 12], axes=[0], starts=[1], ends=[9]),
+        dict(shape=[10, 12], axes=[1], starts=[1], ends=[11]),
+        dict(shape=[10, 12], axes=[0, 1], starts=[1, 1], ends=[9, 11]),
+        dict(shape=[8, 10, 12], axes=[0], starts=[1], ends=[7]),
+        dict(shape=[8, 10, 12], axes=[1], starts=[1], ends=[9]),
+        dict(shape=[8, 10, 12], axes=[2], starts=[1], ends=[11]),
+        dict(shape=[8, 10, 12], axes=[0, 1], starts=[1, 1], ends=[7, 9]),
+        dict(shape=[8, 10, 12], axes=[1, 2], starts=[1, 1], ends=[9, 11]),
+        dict(shape=[8, 10, 12], axes=[0, 2], starts=[1, 1], ends=[7, 11]),
+        dict(shape=[8, 10, 12], axes=[0, 1, 2], starts=[1, 1, 1], ends=[7, 9, 11]),
+        dict(shape=[6, 8, 10, 12], axes=[0], starts=[1], ends=[5]),
+        dict(shape=[6, 8, 10, 12], axes=[1], starts=[1], ends=[7]),
+        dict(shape=[6, 8, 10, 12], axes=[2], starts=[1], ends=[9]),
+        dict(shape=[6, 8, 10, 12], axes=[3], starts=[1], ends=[11]),
+        dict(shape=[6, 8, 10, 12], axes=[0, 1], starts=[1, 1], ends=[5, 7]),
+        dict(shape=[6, 8, 10, 12], axes=[1, 2], starts=[1, 1], ends=[7, 9]),
+        dict(shape=[6, 8, 10, 12], axes=[2, 3], starts=[1, 1], ends=[9, 11]),
+        dict(shape=[6, 8, 10, 12], axes=[0, 2], starts=[1, 1], ends=[5, 9]),
+        dict(shape=[6, 8, 10, 12], axes=[0, 3], starts=[1, 1], ends=[5, 11]),
+        dict(shape=[6, 8, 10, 12], axes=[1, 3], starts=[1, 1], ends=[7, 11]),
+        dict(shape=[6, 8, 10, 12], axes=[0, 1, 2], starts=[1, 1, 1], ends=[5, 7, 9]),
+        dict(shape=[6, 8, 10, 12], axes=[1, 2, 3], starts=[1, 1, 1], ends=[7, 9, 11]),
+        dict(shape=[6, 8, 10, 12], axes=[0, 2, 3], starts=[1, 1, 1], ends=[5, 9, 11]),
+        dict(shape=[6, 8, 10, 12], axes=[0, 1, 3], starts=[1, 1, 1], ends=[5, 7, 11]),
+        dict(shape=[6, 8, 10, 12], axes=[0, 1, 2, 3], starts=[1, 1, 1, 1], ends=[5, 7, 9, 11]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[0], starts=[1], ends=[3]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[1], starts=[1], ends=[5]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[2], starts=[1], ends=[7]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[3], starts=[1], ends=[9]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[4], starts=[1], ends=[11]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[0, 1], starts=[1, 1], ends=[3, 5]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[2, 3], starts=[1, 1], ends=[7, 9]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[3, 4], starts=[1, 1], ends=[9, 11]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[0, 1, 2], starts=[1, 1, 1], ends=[3, 5, 7]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[1, 2, 3], starts=[1, 1, 1], ends=[5, 7, 9]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[2, 3, 4], starts=[1, 1, 1], ends=[7, 9, 11]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[0, 1, 2, 3], starts=[1, 1, 1, 1], ends=[3, 5, 7, 9]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[1, 2, 3, 4], starts=[1, 1, 1, 1], ends=[5, 7, 9, 11]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[0, 1, 2, 3, 4], starts=[1, 1, 1, 1, 1], ends=[3, 5, 7, 9, 11]),
+    ]
+
+    test_data_with_steps = [
+        dict(shape=[10, 12], axes=[0, 1], starts=[1, 1], ends=[9, 11], steps=[2, 2]),
+        dict(shape=[10, 12], axes=[0, 1], starts=[9, 11], ends=[1, 1], steps=[-1, -1]),
+        dict(shape=[10, 12], axes=[0], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[10, 12], axes=[1], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[10, 12], axes=[0, 1], starts=[9, 11], ends=[1, 1], steps=[-2, -2]),
+        dict(shape=[8, 10, 12], axes=[0, 1, 2], starts=[1, 1, 1], ends=[7, 9, 11], steps=[2, 2, 2]),
+        dict(shape=[8, 10, 12], axes=[0, 1, 2], starts=[7, 9, 11], ends=[1, 1, 1], steps=[-1, -1, -1]),
+        dict(shape=[8, 10, 12], axes=[0], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[8, 10, 12], axes=[1], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[8, 10, 12], axes=[2], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[8, 10, 12], axes=[0, 1, 2], starts=[7, 9, 11], ends=[1, 1, 1], steps=[-2, -2, -2]),
+        dict(shape=[6, 8, 10, 12], axes=[0, 1, 2, 3], starts=[1, 1, 1, 1], ends=[5, 7, 9, 11], steps=[2, 2, 2, 2]),
+        dict(shape=[6, 8, 10, 12], axes=[0, 1, 2, 3], starts=[5, 7, 9, 11], ends=[1, 1, 1, 1], steps=[-1, -1, -1, -1]),
+        dict(shape=[6, 8, 10, 12], axes=[0], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[6, 8, 10, 12], axes=[1], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[6, 8, 10, 12], axes=[2], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[6, 8, 10, 12], axes=[3], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[6, 8, 10, 12], axes=[0, 1, 2, 3], starts=[5, 7, 9, 11], ends=[1, 1, 1, 1], steps=[-2, -2, -2, -2]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[0, 1, 2, 3, 4], starts=[1, 1, 1, 1, 1], ends=[3, 5, 7, 9, 11],
+             steps=[2, 2, 2, 2, 2]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[0, 1, 2, 3, 4], starts=[3, 5, 7, 9, 11], ends=[1, 1, 1, 1, 1],
+             steps=[-1, -1, -1, -1, -1]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[0], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[1], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[2], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[3], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[4], starts=[-1], ends=[-9999], steps=[-1]),
+        dict(shape=[4, 6, 8, 10, 12], axes=[0, 1, 2, 3, 4], starts=[3, 5, 7, 9, 11], ends=[1, 1, 1, 1, 1],
+             steps=[-2, -2, -2, -2, -2]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_no_steps)
+    @pytest.mark.nightly
+    def test_slice_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, opset=6, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_no_steps)
+    @pytest.mark.nightly
+    def test_slice_const_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, opset=6, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_no_steps + test_data_with_steps)
+    @pytest.mark.nightly
+    def test_slice_opset10(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(
+            *self.create_net(**params, opset=10, ir_version=ir_version), ie_device, precision, ir_version,
+            temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_no_steps + test_data_with_steps)
+    @pytest.mark.nightly
+    def test_slice_const_opset10(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, opset=10, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_no_steps + test_data_with_steps)
+    @pytest.mark.nightly
+    def test_slice_opset11(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(
+            *self.create_net(**params, opset=11, ir_version=ir_version), ie_device, precision, ir_version,
+            temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_no_steps + test_data_with_steps)
+    @pytest.mark.nightly
+    def test_slice_const_opset11(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, opset=11, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_softmax.py b/tests/layer_tests/onnx_tests/test_softmax.py
new file mode 100644
index 00000000000000..cf4a5d4d6cd8a6
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_softmax.py
@@ -0,0 +1,167 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from mo.front.common.partial_infer.utils import int64_array
+from unit_tests.utils.graph import build_graph
+
+
+def second_input_data_of_reshape(src_shape, axis):
+    if axis == 0:
+        return [1, -1]
+    if axis == 1:
+        return [0, -1]
+    if axis > 1:
+        return [int(np.prod(int64_array(src_shape[: axis]))), -1]
+    return [-1, int(np.prod(int64_array(src_shape[len(src_shape) + axis: ])))]
+
+
+def get_flatten_shape(src_shape, axis):
+    flatten_axis = axis if axis >= 0 else len(src_shape) + axis
+    if flatten_axis == 0:
+        fst_dim = 1
+        snd_dim = int(np.prod(int64_array(src_shape)))
+    elif flatten_axis == 1:
+        fst_dim = src_shape[0]
+        snd_dim = int(np.prod(int64_array(src_shape[1: ])))
+    else:
+        fst_dim = int(np.prod(int64_array(src_shape[: flatten_axis])))
+        snd_dim = int(np.prod(int64_array(src_shape[flatten_axis: ])))
+    return [fst_dim, snd_dim]
+
+
+class TestSoftmax(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randn(*inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_net(self, shape, softmax_axis, ir_version):
+        """
+            ONNX net                       IR net
+
+            Input->Softmax->Output   =>    Input->Reshape->SoftMax->Reshape
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Softmax',
+            inputs=['input'],
+            outputs=['output'],
+            axis=softmax_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        converted_shape = shape if len(shape) != 1 else shape[0]
+        flatten_shape = get_flatten_shape(shape, softmax_axis)
+        reshape_data_val = second_input_data_of_reshape(shape, softmax_axis)
+
+        if check_ir_version(10, None, ir_version):
+            if len(shape) == 2 and shape == flatten_shape:
+                ref_nodes_attributes = {
+                    'input': {'kind': 'op', 'type': 'Parameter', 'shape': converted_shape},
+                    'input_data': {'shape': shape, 'kind': 'data', 'value': None},
+                    'flatten_shape_val': {'shape': int64_array(reshape_data_val).shape,
+                                          'kind': 'data',
+                                          'value': int64_array(reshape_data_val)},
+                    'flatten_shape': {'type': 'Const', 'kind': 'op', 'shape': 2},
+                    'flatten_shape_data': {'shape': int64_array([2]), 'kind': 'data', 'value': None},
+                    'reshape': {'kind': 'op', 'type': 'Reshape'},
+                    'reshape_data': {'kind': 'data', 'shape': flatten_shape, 'value': None},
+                    'softmax': {'type': 'SoftMax', 'kind': 'op', 'axis': 1},
+                    'softmax_data': {'shape': flatten_shape, 'kind': 'data', 'value': None},
+                    'result': {'kind': 'op', 'type': 'Result'},
+                }
+
+                ref_edges = [
+                    ('input', 'input_data'),
+                    ('flatten_shape_val', 'flatten_shape'),
+                    ('flatten_shape', 'flatten_shape_data'),
+                    ('flatten_shape_data', 'reshape', {'in': 1}),
+                    ('input_data', 'reshape', {'in': 0}),
+                    ('reshape', 'reshape_data'),
+                    ('reshape_data', 'softmax'),
+                    ('softmax', 'softmax_data'),
+                    ('softmax_data', 'result'),
+                ]
+            else:
+                ref_nodes_attributes = {
+                    'input': {'kind': 'op', 'type': 'Parameter', 'shape': converted_shape},
+                    'input_data': {'shape': shape, 'kind': 'data', 'value': None},
+                    'flatten_shape_val': {'shape': int64_array(reshape_data_val).shape,
+                                          'kind': 'data',
+                                          'value': int64_array(reshape_data_val)},
+                    'flatten_shape': {'type': 'Const', 'kind': 'op', 'shape': 2},
+                    'flatten_shape_data': {'shape': int64_array([2]), 'kind': 'data', 'value': None},
+                    'reshape': {'kind': 'op', 'type': 'Reshape'},
+                    'reshape_data': {'kind': 'data', 'shape': flatten_shape, 'value': None},
+                    'softmax': {'type': 'SoftMax', 'kind': 'op', 'axis': 1},
+                    'softmax_data': {'shape': flatten_shape, 'kind': 'data', 'value': None},
+                    'last_shape_val': {'shape': int64_array(shape).shape, 'kind': 'data', 'value': int64_array(shape)},
+                    'last_shape': {'type': 'Const', 'kind': 'op', 'shape': len(shape)},
+                    'last_shape_data': {'shape': int64_array([len(shape)]), 'kind': 'data', 'value': None},
+                    'last_reshape': {'kind': 'op', 'type': 'Reshape'},
+                    'last_reshape_data': {'kind': 'data', 'shape': shape, 'value': None},
+                    'result': {'kind': 'op', 'type': 'Result'},
+                }
+
+                ref_edges = [
+                    ('input', 'input_data'),
+                    ('flatten_shape_val', 'flatten_shape'),
+                    ('flatten_shape', 'flatten_shape_data'),
+                    ('flatten_shape_data', 'reshape', {'in': 1}),
+                    ('input_data', 'reshape', {'in': 0}),
+                    ('reshape', 'reshape_data'),
+                    ('reshape_data', 'softmax'),
+                    ('softmax', 'softmax_data'),
+                    ('last_shape_val', 'last_shape'),
+                    ('last_shape', 'last_shape_data'),
+                    ('last_shape_data', 'last_reshape', {'in': 1}),
+                    ('softmax_data', 'last_reshape', {'in': 0}),
+                    ('last_reshape', 'last_reshape_data'),
+                    ('last_reshape_data', 'result'),
+                ]
+
+            ref_net = build_graph(ref_nodes_attributes, ref_edges)
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[10, 12], softmax_axis=0),
+                 dict(shape=[10, 12], softmax_axis=1)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_softmax(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_softplus.py b/tests/layer_tests/onnx_tests/test_softplus.py
new file mode 100644
index 00000000000000..ea5b56071e0dd2
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_softplus.py
@@ -0,0 +1,182 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestSoftplus(Caffe2OnnxLayerTest):
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                       IR net
+
+            Input->Softplus->Output   =>   Input->Softplus
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Softplus',
+            inputs=['input'],
+            outputs=['output'],
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'SoftPlus'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                       IR net
+
+            Input->Concat(+Softplus const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.rand(*shape).astype(np.float32) * 255 + 0.5
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Softplus',
+            inputs=['const1'],
+            outputs=['Softplus1'],
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'Softplus1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.log(np.exp(constant) + 1.0)
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')
+                                   ])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[12]),
+                 dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_softplus(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_softplus_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_split_concat.py b/tests/layer_tests/onnx_tests/test_split_concat.py
new file mode 100644
index 00000000000000..b9bac5aaf2504c
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_split_concat.py
@@ -0,0 +1,307 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+
+
+test_data_3D = [
+    dict(input_shape=[1, 50, 50], output_shapes=[[1, 50, 25], [1, 50, 25]], axis=2),
+    dict(input_shape=[2, 50, 50], output_shapes=[[2, 20, 50], [2, 15, 50], [2, 15, 50]], axis=1),
+    dict(input_shape=[4, 50, 50], output_shapes=[[1, 50, 50], [1, 50, 50], [1, 50, 50], [1, 50, 50]], axis=0)]
+
+test_data_4D = [
+    dict(input_shape=[1, 32, 800, 800], output_shapes=[[1, 16, 800, 800], [1, 16, 800, 800]], axis=1),
+    dict(input_shape=[4, 32, 80, 80], output_shapes=[[4, 8, 80, 80], [4, 8, 80, 80], [4, 8, 80, 80], [4, 8, 80, 80]],
+         axis=1),
+    dict(input_shape=[2, 21, 80, 80], output_shapes=[[2, 7, 80, 80], [2, 7, 80, 80], [2, 7, 80, 80]], axis=1),
+    dict(input_shape=[3, 21, 80, 80], output_shapes=[[3, 14, 80, 80], [3, 5, 80, 80], [3, 2, 80, 80]], axis=1),
+    dict(input_shape=[3, 21, 80, 80], output_shapes=[[1, 21, 80, 80], [1, 21, 80, 80], [1, 21, 80, 80]], axis=0),
+    dict(input_shape=[3, 21, 80, 80], output_shapes=[[3, 21, 20, 80], [3, 21, 35, 80], [3, 21, 25, 80]], axis=2),
+    dict(input_shape=[3, 21, 80, 80], output_shapes=[[3, 21, 80, 40], [3, 21, 80, 10], [3, 21, 80, 30]], axis=3)]
+
+test_data_5D = [
+    dict(input_shape=[1, 50, 50, 80, 60],
+         output_shapes=[[1, 50, 10, 80, 60],
+                        [1, 50, 10, 80, 60],
+                        [1, 50, 10, 80, 60],
+                        [1, 50, 10, 80, 60],
+                        [1, 50, 10, 80, 60]], axis=2),
+    dict(input_shape=[1, 50, 50, 80, 60], output_shapes=[[1, 25, 50, 80, 60], [1, 25, 50, 80, 60]], axis=1)]
+
+
+class TestSplitConcat(Caffe2OnnxLayerTest):
+    # TODO Add test with default values (axis=0)
+    def create_split_concat_net(self, input_shape, output_shapes, axis, ir_version):
+        """
+            ONNX net                             IR net
+
+            Input->Split->Concat->Output   =>    Input->Split->Concat
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        outputs, split = [], []
+        for id, output_shape in enumerate(output_shapes):
+            helper.make_tensor_value_info('output_{}'.format(id), TensorProto.FLOAT, output_shape)
+            outputs.append('output_{}'.format(id))
+            split.append(output_shape[axis])
+
+        # Output for concat
+        output_concat = helper.make_tensor_value_info('output_concat', TensorProto.FLOAT, input_shape)
+
+        node_split_def = onnx.helper.make_node(
+            'Split',
+            inputs=['input'],
+            outputs=outputs,
+            axis=axis,
+            split=split
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=outputs,
+            outputs=['output_concat'],
+            axis=axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_split_def, node_concat_def],
+            'test_split_model',
+            [input],
+            [output_concat],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_split_model')
+
+        #
+        #   Create reference IR net
+        #   Please, spesify 'type': 'Input' for inpit node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    # TODO Add test with default values (axis=0)
+    def create_split_concat_net_const(self, input_shape, output_shapes, axis, ir_version):
+        """
+            ONNX net                                               IR net
+
+            Input(const)->Split->Concat--->Concat->Output   =>    Input--->Concat
+                                  Input-'                         Const-'
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        concat_output_shape = input_shape.copy()
+        concat_output_shape[concat_axis] *= 2
+
+        const_number = np.prod(input_shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        outputs, split = [], []
+        for id, output_shape in enumerate(output_shapes):
+            helper.make_tensor_value_info('output_{}'.format(id), TensorProto.FLOAT, output_shape)
+            outputs.append('output_{}'.format(id))
+            split.append(output_shape[axis])
+
+        # Output for concat
+        output_concat = helper.make_tensor_value_info('output_dyn_concat', TensorProto.FLOAT, concat_output_shape)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=input_shape,
+                vals=constant,
+            ),
+        )
+
+        node_split_def = onnx.helper.make_node(
+            'Split',
+            inputs=['const1'],
+            outputs=outputs,
+            axis=axis,
+            split=split
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=outputs,
+            outputs=['output_concat'],
+            axis=axis
+        )
+
+        node_dyn_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'output_concat'],
+            outputs=['output_dyn_concat'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_split_def, node_concat_def, node_dyn_concat_def],
+            'test_split_model',
+            [input],
+            [output_concat],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_split_model')
+
+        #
+        #   Create reference IR net
+        #   Please, spesify 'type': 'Input' for inpit node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_split_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_split_concat_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_split_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_split_concat_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_split_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_split_concat_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_split_3D_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(
+            *self.create_split_concat_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+            temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_split_4D_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(
+            *self.create_split_concat_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+            temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_split_5D_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(
+            *self.create_split_concat_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+            temp_dir=temp_dir)
+
+
+class TestSplit(Caffe2OnnxLayerTest):
+    # TODO Add test with default values (axis=0)
+    def create_split_net(self, input_shape, output_shapes, axis, ir_version):
+        """
+            ONNX net                       IR net
+
+            Input->Split->Output   =>    Input->Split
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        outputs, split = [], []
+        for id, output_shape in enumerate(output_shapes):
+            out = helper.make_tensor_value_info('output_{}'.format(id), TensorProto.FLOAT, output_shape)
+            outputs.append((out, 'output_{}'.format(id)))
+            split.append(output_shape[axis])
+
+        node_split_def = onnx.helper.make_node(
+            'Split',
+            inputs=['input'],
+            outputs=['node_{}'.format(x[1]) for x in outputs],
+            axis=axis,
+            split=split
+        )
+        nodes = [node_split_def]
+
+        for x in outputs:
+            nodes.append(onnx.helper.make_node(
+                'Elu',
+                inputs=['node_{}'.format(x[1])],
+                outputs=[x[1]]
+            ))
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_split_model',
+            [input],
+            [x[0] for x in outputs],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_split_model')
+
+        #
+        #   Create reference IR net
+        #   Please, spesify 'type': 'Input' for inpit node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_split_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_split_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_split_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_split_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_split_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_split_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_sqrt.py b/tests/layer_tests/onnx_tests/test_sqrt.py
new file mode 100644
index 00000000000000..09b6df3b481bb4
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_sqrt.py
@@ -0,0 +1,187 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestSqrt(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randint(0, 255, inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_net(self, shape, ir_version):
+        """
+            ONNX net                    IR net
+
+            Input->Sqrt->Output   =>    Input->Power
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            'Sqrt',
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'const_indata': {'shape': None, 'kind': 'data'},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': np.ones(len(shape)), 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'Power'},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('const_indata', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'node'),
+                                   ('const_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, precision, ir_version):
+        """
+            ONNX net                                   IR net
+
+            Input->Concat(+sqrt const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.rand(*shape).astype(np.float) * 255
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            'Sqrt',
+            inputs=['const'],
+            outputs=['sqrt']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'sqrt'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        constant = np.sqrt(constant)
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sqrt(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sqrt_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_squeeze.py b/tests/layer_tests/onnx_tests/test_squeeze.py
new file mode 100644
index 00000000000000..413d3d86ec9fff
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_squeeze.py
@@ -0,0 +1,210 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+
+
+class TestSqueeze(Caffe2OnnxLayerTest):
+    def create_squeeze_net(self, axes, input_shape, output_shape, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->Squeeze(axes=0)->Output   =>    Input->Reshape
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_squeeze_def = onnx.helper.make_node(
+            'Squeeze',
+            inputs=['input'],
+            outputs=['output'],
+            axes=axes
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_squeeze_def],
+            'test_squeeze_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_squeeze_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_squeeze_net_const(self, axes, input_shape, output_shape, ir_version):
+        """
+            ONNX net                                         IR net
+
+            Input->Concat(+squeezed const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 0
+        concat_output_shape = output_shape.copy()
+        concat_output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, output_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, concat_output_shape)
+
+        const_number = np.prod(input_shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+        constant = np.reshape(constant, input_shape)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_squeeze_def = onnx.helper.make_node(
+            'Squeeze',
+            inputs=['const1'],
+            outputs=['squeeze1'],
+            axes=axes
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'squeeze1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_squeeze_def, node_concat_def],
+            'test_squeeze_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_squeeze_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data_5D = [
+        dict(axes=[0], input_shape=[1, 2, 3, 10, 10], output_shape=[2, 3, 10, 10]),
+        dict(axes=[1], input_shape=[2, 1, 3, 10, 10], output_shape=[2, 3, 10, 10]),
+        dict(axes=[2], input_shape=[2, 3, 1, 10, 10], output_shape=[2, 3, 10, 10]),
+        dict(axes=[3], input_shape=[2, 3, 10, 1, 10], output_shape=[2, 3, 10, 10]),
+        dict(axes=[4], input_shape=[2, 3, 10, 10, 1], output_shape=[2, 3, 10, 10]),
+        dict(axes=[0, 1], input_shape=[1, 1, 3, 10, 10], output_shape=[3, 10, 10]),
+        dict(axes=[0, 2], input_shape=[1, 3, 1, 10, 10], output_shape=[3, 10, 10]),
+        dict(axes=[0, 3], input_shape=[1, 3, 10, 1, 10], output_shape=[3, 10, 10]),
+        dict(axes=[0, 4], input_shape=[1, 3, 10, 10, 1], output_shape=[3, 10, 10]),
+        dict(axes=[1, 2], input_shape=[3, 1, 1, 10, 10], output_shape=[3, 10, 10]),
+        dict(axes=[1, 3], input_shape=[3, 1, 10, 1, 10], output_shape=[3, 10, 10]),
+        dict(axes=[1, 4], input_shape=[3, 1, 10, 10, 1], output_shape=[3, 10, 10]),
+        dict(axes=[2, 3], input_shape=[3, 10, 1, 1, 10], output_shape=[3, 10, 10]),
+        dict(axes=[2, 4], input_shape=[3, 10, 1, 10, 1], output_shape=[3, 10, 10]),
+        dict(axes=[3, 4], input_shape=[3, 10, 10, 1, 1], output_shape=[3, 10, 10]),
+        dict(axes=[0, 1, 2], input_shape=[1, 1, 1, 10, 10], output_shape=[10, 10]),
+        dict(axes=[0, 1, 3], input_shape=[1, 1, 10, 1, 10], output_shape=[10, 10]),
+        dict(axes=[0, 1, 4], input_shape=[1, 1, 10, 10, 1], output_shape=[10, 10]),
+        dict(axes=[0, 2, 3], input_shape=[1, 10, 1, 1, 10], output_shape=[10, 10]),
+        dict(axes=[0, 2, 4], input_shape=[1, 10, 1, 10, 1], output_shape=[10, 10]),
+        dict(axes=[0, 3, 4], input_shape=[1, 10, 10, 1, 1], output_shape=[10, 10]),
+        dict(axes=[1, 2, 3], input_shape=[10, 1, 1, 1, 10], output_shape=[10, 10]),
+        dict(axes=[1, 2, 4], input_shape=[10, 1, 1, 10, 1], output_shape=[10, 10]),
+        dict(axes=[1, 3, 4], input_shape=[10, 1, 10, 1, 1], output_shape=[10, 10]),
+        dict(axes=[2, 3, 4], input_shape=[10, 10, 1, 1, 1], output_shape=[10, 10])]
+
+    test_data_4D = [
+        dict(axes=[0], input_shape=[1, 3, 10, 10], output_shape=[3, 10, 10]),
+        dict(axes=[1], input_shape=[3, 1, 10, 10], output_shape=[3, 10, 10]),
+        dict(axes=[2], input_shape=[3, 10, 1, 10], output_shape=[3, 10, 10]),
+        dict(axes=[3], input_shape=[3, 10, 10, 1], output_shape=[3, 10, 10]),
+        dict(axes=[0, 1], input_shape=[1, 1, 10, 10], output_shape=[10, 10]),
+        dict(axes=[0, 2], input_shape=[1, 10, 1, 10], output_shape=[10, 10]),
+        dict(axes=[0, 3], input_shape=[1, 10, 10, 1], output_shape=[10, 10]),
+        dict(axes=[1, 2], input_shape=[10, 1, 1, 10], output_shape=[10, 10]),
+        dict(axes=[1, 3], input_shape=[10, 1, 10, 1], output_shape=[10, 10]),
+        dict(axes=[2, 3], input_shape=[10, 10, 1, 1], output_shape=[10, 10])]
+
+    test_data_3D = [
+        dict(axes=[0], input_shape=[1, 10, 10], output_shape=[10, 10]),
+        dict(axes=[1], input_shape=[10, 1, 10], output_shape=[10, 10]),
+        dict(axes=[2], input_shape=[10, 10, 1], output_shape=[10, 10])]
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_squeeze_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_squeeze_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_squeeze_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_squeeze_const_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_squeeze_const_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_squeeze_const_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_sum.py b/tests/layer_tests/onnx_tests/test_sum.py
new file mode 100644
index 00000000000000..bd4896144c0a49
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_sum.py
@@ -0,0 +1,322 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestSum(OnnxRuntimeLayerTest):
+    def create_net(self, dyn_shapes, const_shapes, precision, ir_version, opset=None):
+        """
+            ONNX net                                IR net
+
+            Inputs->Sum with consts->Output   =>    Input->Eltwise
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        inputs = list()
+        input_names = list()
+        out_shape_len = 0
+        for i, shape in enumerate(dyn_shapes):
+            input_name = 'input{}'.format(i + 1)
+            inputs.append(helper.make_tensor_value_info(input_name, TensorProto.FLOAT, shape))
+            input_names.append(input_name)
+            if len(shape) > out_shape_len:
+                out_shape_len = len(shape)
+                output_shape = shape
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        nodes = list()
+        consts = list()
+        for i, shape in enumerate(const_shapes):
+            const = np.random.randint(-127, 127, shape).astype(np.float)
+            const_name = 'const{}'.format(i + 1)
+            nodes.append(helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=[const_name],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.FLOAT,
+                    dims=const.shape,
+                    vals=const.flatten(),
+                ),
+            ))
+            input_names.append(const_name)
+            consts.append(const)
+
+        nodes.append(helper.make_node(
+            'Sum',
+            inputs=input_names,
+            outputs=['output']
+        ))
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            inputs,
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #   Create reference IR net
+
+        ref_net = None
+        # Too complicated IR to generate by hand
+
+        return onnx_net, ref_net
+
+    def create_const_net(self, const_shapes, ir_version, opset=None):
+        """
+            ONNX net                                          IR net
+
+            Inputs->Concat with Sum of consts->Output   =>    Input->Concat with consts
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        shape_len = 0
+        for shape in const_shapes:
+            if len(shape) > shape_len:
+                shape_len = len(shape)
+                input_shape = shape
+
+        concat_axis = 0
+        output_shape = input_shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        nodes = list()
+        input_names = list()
+        consts = list()
+        for i, shape in enumerate(const_shapes):
+            const = np.random.randint(-127, 127, shape).astype(np.float)
+            const_name = 'const{}'.format(i + 1)
+            nodes.append(helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=[const_name],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.FLOAT,
+                    dims=const.shape,
+                    vals=const.flatten(),
+                ),
+            ))
+            input_names.append(const_name)
+            consts.append(const)
+
+        nodes.append(helper.make_node(
+            'Sum',
+            inputs=input_names,
+            outputs=['sum']
+        ))
+
+        nodes.append(helper.make_node(
+            'Concat',
+            inputs=['input', 'sum'],
+            outputs=['output'],
+            axis=concat_axis
+        ))
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #   Create reference IR net
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [
+        dict(dyn_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8]],
+             const_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12]],
+             const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]],
+             const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]])]
+
+    test_data = [
+        # TODO: Add broadcasting tests. Note: Sum-6 doesn't support broadcasting
+        dict(dyn_shapes=[[4, 6]], const_shapes=[[4, 6]]),
+        dict(dyn_shapes=[[4, 6]], const_shapes=[[4, 6], [4, 6]]),
+        dict(dyn_shapes=[[4, 6]], const_shapes=[[4, 6], [4, 6], [4, 6]]),
+        dict(dyn_shapes=[[4, 6], [4, 6]], const_shapes=[]),
+        dict(dyn_shapes=[[4, 6], [4, 6]], const_shapes=[[4, 6]]),
+        dict(dyn_shapes=[[4, 6], [4, 6]], const_shapes=[[4, 6], [4, 6]]),
+        dict(dyn_shapes=[[4, 6], [4, 6]], const_shapes=[[4, 6], [4, 6], [4, 6]]),
+        dict(dyn_shapes=[[4, 6], [4, 6], [4, 6]], const_shapes=[]),
+        dict(dyn_shapes=[[4, 6], [4, 6], [4, 6]], const_shapes=[[4, 6]]),
+        dict(dyn_shapes=[[4, 6], [4, 6], [4, 6]], const_shapes=[[4, 6], [4, 6]]),
+        dict(dyn_shapes=[[4, 6], [4, 6], [4, 6]], const_shapes=[[4, 6], [4, 6], [4, 6]]),
+        dict(dyn_shapes=[[4, 6, 8]], const_shapes=[[4, 6, 8]]),
+        dict(dyn_shapes=[[4, 6, 8]], const_shapes=[[4, 6, 8], [4, 6, 8]]),
+        dict(dyn_shapes=[[4, 6, 8]], const_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8]]),
+        dict(dyn_shapes=[[4, 6, 8], [4, 6, 8]], const_shapes=[]),
+        dict(dyn_shapes=[[4, 6, 8], [4, 6, 8]], const_shapes=[[4, 6, 8]]),
+        dict(dyn_shapes=[[4, 6, 8], [4, 6, 8]], const_shapes=[[4, 6, 8], [4, 6, 8]]),
+        dict(dyn_shapes=[[4, 6, 8], [4, 6, 8]], const_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8]]),
+        dict(dyn_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8]], const_shapes=[]),
+        dict(dyn_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8]], const_shapes=[[4, 6, 8]]),
+        dict(dyn_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8]], const_shapes=[[4, 6, 8], [4, 6, 8]]),
+        dict(dyn_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8]], const_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8]]),
+        dict(dyn_shapes=[[4, 6, 8, 10]], const_shapes=[[4, 6, 8, 10]]),
+        dict(dyn_shapes=[[4, 6, 8, 10]], const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10]]),
+        dict(dyn_shapes=[[4, 6, 8, 10]], const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10]]),
+        dict(dyn_shapes=[[4, 6, 8, 10], [4, 6, 8, 10]], const_shapes=[]),
+        dict(dyn_shapes=[[4, 6, 8, 10], [4, 6, 8, 10]], const_shapes=[[4, 6, 8, 10]]),
+        dict(dyn_shapes=[[4, 6, 8, 10], [4, 6, 8, 10]], const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10]]),
+        dict(dyn_shapes=[[4, 6, 8, 10], [4, 6, 8, 10]], const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10]]),
+        dict(dyn_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10]], const_shapes=[]),
+        dict(dyn_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10]], const_shapes=[[4, 6, 8, 10]]),
+        dict(dyn_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10]], const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10]]),
+        dict(dyn_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10]],
+             const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12]], const_shapes=[[4, 6, 8, 10, 12]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12]], const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12]], const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12]], const_shapes=[]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12]], const_shapes=[[4, 6, 8, 10, 12]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12]], const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12]],
+             const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]], const_shapes=[]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]], const_shapes=[[4, 6, 8, 10, 12]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]],
+             const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12]]),
+        dict(dyn_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]],
+             const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]])]
+
+    const_test_data_precommit = [
+        dict(const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]])
+    ]
+
+    const_test_data = [
+        dict(const_shapes=[[4, 6], [4, 6]]),
+        dict(const_shapes=[[4, 6], [4, 6], [4, 6]]),
+        dict(const_shapes=[[4, 6], [4, 6], [4, 6], [4, 6]]),
+        dict(const_shapes=[[4, 6, 8], [4, 6, 8]]),
+        dict(const_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8]]),
+        dict(const_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8], [4, 6, 8]]),
+        dict(const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10]]),
+        dict(const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10]]),
+        dict(const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12]])
+    ]
+
+    const_test_data_broadcasting_precommit = [
+        dict(const_shapes=[[4, 6, 8, 10], [10], [10], [10]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [12]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [12]])
+    ]
+
+    const_test_data_broadcasting = [
+        dict(const_shapes=[[4, 6], [6]]),
+        dict(const_shapes=[[4, 6], [6], [6]]),
+        dict(const_shapes=[[4, 6], [4, 6], [6]]),
+        dict(const_shapes=[[4, 6], [6], [6], [6]]),
+        dict(const_shapes=[[4, 6], [4, 6], [6], [6]]),
+        dict(const_shapes=[[4, 6], [4, 6], [4, 6], [6]]),
+        dict(const_shapes=[[4, 6, 8], [8]]),
+        dict(const_shapes=[[4, 6, 8], [8], [8]]),
+        dict(const_shapes=[[4, 6, 8], [4, 6, 8], [8]]),
+        dict(const_shapes=[[4, 6, 8], [8], [8], [8]]),
+        dict(const_shapes=[[4, 6, 8], [4, 6, 8], [8], [8]]),
+        dict(const_shapes=[[4, 6, 8], [4, 6, 8], [4, 6, 8], [8]]),
+        dict(const_shapes=[[4, 6, 8, 10], [10]]),
+        dict(const_shapes=[[4, 6, 8, 10], [10], [10]]),
+        dict(const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [10]]),
+        dict(const_shapes=[[4, 6, 8, 10], [10], [10], [10]]),
+        dict(const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [10], [10]]),
+        dict(const_shapes=[[4, 6, 8, 10], [4, 6, 8, 10], [4, 6, 8, 10], [10]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [12]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [12], [12]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [12]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [12], [12], [12]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [12], [12]]),
+        dict(const_shapes=[[4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [4, 6, 8, 10, 12], [12]])
+    ]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sum_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, opset=6, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_sum_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, precision=precision, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sum(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(
+            *self.create_net(**params, precision=precision, ir_version=ir_version), ie_device, precision, ir_version,
+            temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", const_test_data)
+    @pytest.mark.nightly
+    def test_sum_const_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_const_net(**params, opset=6, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", const_test_data_precommit)
+    @pytest.mark.precommit
+    def test_sum_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_const_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", const_test_data)
+    @pytest.mark.nightly
+    def test_sum_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_const_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", const_test_data_broadcasting_precommit)
+    @pytest.mark.precommit
+    def test_sum_const_broadcasting_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_const_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", const_test_data_broadcasting)
+    @pytest.mark.nightly
+    def test_sum_const_broadcasting(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_const_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_topk.py b/tests/layer_tests/onnx_tests/test_topk.py
new file mode 100644
index 00000000000000..863bf95352eff9
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_topk.py
@@ -0,0 +1,167 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestTopK(OnnxRuntimeLayerTest):
+    def create_net(self, shape, k, axis, ir_version, largest=None, sorted=None, opset=None):
+        """
+            ONNX net                    IR net
+
+            Input->TopK->Output   =>    Input->TopK
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        output_shape = shape.copy()
+        if axis is not None:
+            output_shape[axis] = k
+        else:
+            output_shape[-1] = k
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        values = helper.make_tensor_value_info('cvalues', TensorProto.FLOAT, output_shape)
+        indices = helper.make_tensor_value_info('cindices', TensorProto.INT64, output_shape)
+
+        const1 = np.ones(output_shape).astype(np.int64)
+        const2 = np.ones(output_shape).astype(np.float)
+
+        nodes = list()
+        inputs = ['input']
+        if opset > 9:
+            node_k_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['k'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.INT64,
+                    dims=[1],
+                    vals=[k],
+                ),
+            )
+            nodes.append(node_k_def)
+            inputs.append('k')
+
+        args = dict()
+        if opset < 10:
+            args['k'] = k
+        if axis is not None:
+            args['axis'] = axis
+        if sorted is not None:
+            args['sorted'] = sorted
+        if largest is not None:
+            args['largest'] = largest
+
+        node_def = onnx.helper.make_node(
+            'TopK',
+            inputs=inputs,
+            outputs=['values', 'indices'],
+            **args
+        )
+
+        node_const1_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor2',
+                data_type=TensorProto.INT64,
+                dims=const1.shape,
+                vals=const1.flatten(),
+            ),
+        )
+
+        node_add1_def = onnx.helper.make_node(
+            'Add',
+            inputs=['indices', 'const1'],
+            outputs=['cindices']
+        )
+
+        node_const2_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const2'],
+            value=helper.make_tensor(
+                name='const_tensor3',
+                data_type=TensorProto.FLOAT,
+                dims=const2.shape,
+                vals=const2.flatten(),
+            ),
+        )
+
+        node_add2_def = onnx.helper.make_node(
+            'Add',
+            inputs=['values', 'const2'],
+            outputs=['cvalues']
+        )
+
+        nodes.extend([node_def, node_const1_def, node_add1_def, node_const2_def, node_add2_def])
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            [input],
+            [values, indices],
+        )
+
+        # Create the model (ModelProto)
+        args = dict(producer_name='test_model')
+        if opset:
+            args['opset_imports'] = [helper.make_opsetid("", opset)]
+        onnx_net = helper.make_model(graph_def, **args)
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[10, 12], k=3, axis=0),
+                 dict(shape=[10, 12], k=5, axis=1),
+                 dict(shape=[8, 10, 12], k=3, axis=0),
+                 dict(shape=[8, 10, 12], k=4, axis=1),
+                 dict(shape=[8, 10, 12], k=5, axis=2),
+                 dict(shape=[6, 8, 10, 12], k=3, axis=0),
+                 dict(shape=[6, 8, 10, 12], k=4, axis=1),
+                 dict(shape=[6, 8, 10, 12], k=5, axis=2),
+                 dict(shape=[6, 8, 10, 12], k=6, axis=3),
+                 dict(shape=[4, 6, 8, 10, 12], k=3, axis=0),
+                 dict(shape=[4, 6, 8, 10, 12], k=4, axis=1),
+                 dict(shape=[4, 6, 8, 10, 12], k=5, axis=2),
+                 dict(shape=[4, 6, 8, 10, 12], k=6, axis=3),
+                 dict(shape=[4, 6, 8, 10, 12], k=7, axis=4)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_topk_opset6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, opset=6, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_topk_opset10(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, opset=10, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("largest", [1, 0, None])
+    @pytest.mark.parametrize("sorted", [1, 0, None])
+    @pytest.mark.nightly
+    def test_topk_opset11(self, params, ie_device, precision, ir_version, largest, sorted, temp_dir):
+        self._test(*self.create_net(**params, largest=largest, sorted=sorted,
+                                    opset=11, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_transpose.py b/tests/layer_tests/onnx_tests/test_transpose.py
new file mode 100644
index 00000000000000..13e26b626c48bc
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_transpose.py
@@ -0,0 +1,178 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import itertools
+
+import numpy as np
+import pytest
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+
+
+class TestTranspose(Caffe2OnnxLayerTest):
+    def create_net(self, shape, perm, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->Transpose->Sigmoid->Output   =>    Input->Permute->sigmoid
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        output_shape = np.transpose(np.ones(shape), perm).shape
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        args = dict()
+        if perm:
+            args['perm'] = perm
+        node_def = helper.make_node(
+            'Transpose',
+            inputs=['input'],
+            outputs=['transpose'],
+            **args
+        )
+
+        sigmoid_def = helper.make_node(
+            'Sigmoid',
+            inputs=['transpose'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def, sigmoid_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+        if not perm:
+            perm = list(reversed(range(len(shape))))
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, perm, ir_version):
+        """
+            ONNX net                                         IR net
+
+            Input->Concat(+transposed const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        constant = np.random.randint(-127, 127, shape).astype(np.float)
+        constant_transposed = np.transpose(constant, perm)
+
+        concat_axis = 0
+        input_shape = list(constant_transposed.shape)
+        output_shape = input_shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_const_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        args = dict()
+        if perm:
+            args['perm'] = perm
+        node_def = helper.make_node(
+            'Transpose',
+            inputs=['const1'],
+            outputs=['transpose'],
+            **args
+        )
+
+        node_concat_def = helper.make_node(
+            'Concat',
+            inputs=['input', 'transpose'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [dict(shape=[4, 6, 8, 10, 12], perm=None),
+                           dict(shape=[8, 10, 12], perm=[2, 1, 0]),
+                           dict(shape=[6, 8, 10, 12], perm=[0, 3, 1, 2]),
+                           dict(shape=[4, 6, 8, 10, 12], perm=[1, 0, 4, 3, 2])]
+
+    test_data = [dict(shape=[10, 12], perm=None),
+                 dict(shape=[8, 10, 12], perm=None),
+                 dict(shape=[6, 8, 10, 12], perm=None),
+                 dict(shape=[4, 6, 8, 10, 12], perm=None)]
+
+    for shape in [[10, 12], [8, 10, 12], [6, 8, 10, 12], [4, 6, 8, 10, 12]]:
+        for perm in itertools.permutations(np.arange(len(shape))):
+            test_data.append(dict(shape=shape, perm=list(perm)))
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_transpose_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_transpose(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.nightly
+    def test_transpose_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_transpose_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_trigonometry.py b/tests/layer_tests/onnx_tests/test_trigonometry.py
new file mode 100644
index 00000000000000..6499f007a33da3
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_trigonometry.py
@@ -0,0 +1,298 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestTrigonomery(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.rand(*(inputs_dict[input])).astype(np.float32)
+        return inputs_dict
+
+    def create_net(self, shape, op, ir_version):
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        assert op in ['Sin', 'Sinh', 'Asin', 'Cos', 'Cosh', 'Acos', 'Tan', 'Tanh', 'Atan']
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, shape)
+
+        node_def = onnx.helper.make_node(
+            op,
+            inputs=['input'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': op},
+                'node_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape, op, precision, ir_version):
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        assert op in ['Sin', 'Sinh', 'Asin', 'Cos', 'Cosh', 'Acos', 'Tan', 'Tanh', 'Atan']
+
+        concat_axis = 0
+        output_shape = shape.copy()
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        constant = np.random.rand(*shape).astype(np.float)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_def = onnx.helper.make_node(
+            op,
+            inputs=['const'],
+            outputs=['res']
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'res'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #
+        #   Create reference IR net
+        #
+        if op == 'Sin':
+            constant = np.sin(constant)
+        elif op == 'Sinh':
+            constant = np.sinh(constant)
+        elif op == 'Asin':
+            constant = np.arcsin(constant)
+        elif op == 'Cos':
+            constant = np.cos(constant)
+        elif op == 'Cosh':
+            constant = np.cosh(constant)
+        elif op == 'Acos':
+            constant = np.arccos(constant)
+        elif op == 'Tan':
+            constant = np.tan(constant)
+        elif op == 'Tanh':
+            constant = np.tanh(constant)
+        elif op == 'Atan':
+            constant = np.arctan(constant)
+        if precision == 'FP16':
+            constant = constant.astype(np.float16)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')])
+
+        return onnx_net, ref_net
+
+    test_data_precommit = [dict(shape=[2, 4, 6, 8])]
+
+    test_data = [dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sin(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, op='Sin'), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sinh(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, op='Sinh'), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_asin(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, op='Asin'), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_cos_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, op='Cos'), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_cos(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, op='Cos'), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_cosh(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, op='Cosh'), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_acos(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, op='Acos'), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_tan(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, op='Tan'), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_tanh(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, op='Tanh'), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_atan(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version, op='Atan'), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sin_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version, precision=precision, op='Sin'),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_sinh_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version, precision=precision, op='Sinh'),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_asin_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version, precision=precision, op='Asin'),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_cos_const_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version, precision=precision, op='Cos'),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_cos_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version, precision=precision, op='Cos'),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_cosh_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version, precision=precision, op='Cosh'),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_acos_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version, precision=precision, op='Acos'),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_tan_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version, precision=precision, op='Tan'),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_tanh_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version, precision=precision, op='Tanh'),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_atan_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version, precision=precision, op='Atan'),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_unsqueeze.py b/tests/layer_tests/onnx_tests/test_unsqueeze.py
new file mode 100644
index 00000000000000..4042a2cda63971
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_unsqueeze.py
@@ -0,0 +1,210 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.onnx_layer_test_class import Caffe2OnnxLayerTest
+
+
+class TestUnsqueeze(Caffe2OnnxLayerTest):
+    def create_unsqueeze_net(self, axes, input_shape, output_shape, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->Unsqueeze(axes=0)->Output   =>    Input->Reshape
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, input_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        node_squeeze_def = onnx.helper.make_node(
+            'Unsqueeze',
+            inputs=['input'],
+            outputs=['output'],
+            axes=axes
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_squeeze_def],
+            'test_squeeze_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_squeeze_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    def create_unsqueeze_net_const(self, axes, input_shape, output_shape, ir_version):
+        """
+            ONNX net                                         IR net
+
+            Input->Concat(+unsqueezed const)->Output   =>    Input->Concat(+const)
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+        import numpy as np
+
+        concat_axis = 1
+        concat_output_shape = output_shape.copy()
+        concat_output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, output_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, concat_output_shape)
+
+        const_number = np.prod(input_shape)
+        constant = np.random.randint(-127, 127, const_number).astype(np.float)
+        constant = np.reshape(constant, input_shape)
+
+        node_const_def = onnx.helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.FLOAT,
+                dims=constant.shape,
+                vals=constant.flatten(),
+            ),
+        )
+
+        node_squeeze_def = onnx.helper.make_node(
+            'Unsqueeze',
+            inputs=['const1'],
+            outputs=['unsqueeze1'],
+            axes=axes
+        )
+
+        node_concat_def = onnx.helper.make_node(
+            'Concat',
+            inputs=['input', 'unsqueeze1'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_squeeze_def, node_concat_def],
+            'test_unsqueeze_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_unsqueeze_model')
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data_5D = [
+        dict(axes=[0], input_shape=[2, 3, 10, 10], output_shape=[1, 2, 3, 10, 10]),
+        dict(axes=[1], input_shape=[2, 3, 10, 10], output_shape=[2, 1, 3, 10, 10]),
+        dict(axes=[2], input_shape=[2, 3, 10, 10], output_shape=[2, 3, 1, 10, 10]),
+        dict(axes=[3], input_shape=[2, 3, 10, 10], output_shape=[2, 3, 10, 1, 10]),
+        dict(axes=[4], input_shape=[2, 3, 10, 10], output_shape=[2, 3, 10, 10, 1]),
+        dict(axes=[0, 1], input_shape=[3, 10, 10], output_shape=[1, 1, 3, 10, 10]),
+        dict(axes=[0, 2], input_shape=[3, 10, 10], output_shape=[1, 3, 1, 10, 10]),
+        dict(axes=[0, 3], input_shape=[3, 10, 10], output_shape=[1, 3, 10, 1, 10]),
+        dict(axes=[0, 4], input_shape=[3, 10, 10], output_shape=[1, 3, 10, 10, 1]),
+        dict(axes=[1, 2], input_shape=[3, 10, 10], output_shape=[3, 1, 1, 10, 10]),
+        dict(axes=[1, 3], input_shape=[3, 10, 10], output_shape=[3, 1, 10, 1, 10]),
+        dict(axes=[1, 4], input_shape=[3, 10, 10], output_shape=[3, 1, 10, 10, 1]),
+        dict(axes=[2, 3], input_shape=[3, 10, 10], output_shape=[3, 10, 1, 1, 10]),
+        dict(axes=[2, 4], input_shape=[3, 10, 10], output_shape=[3, 10, 1, 10, 1]),
+        dict(axes=[3, 4], input_shape=[3, 10, 10], output_shape=[3, 10, 10, 1, 1]),
+        dict(axes=[0, 1, 2], input_shape=[10, 10], output_shape=[1, 1, 1, 10, 10]),
+        dict(axes=[0, 1, 3], input_shape=[10, 10], output_shape=[1, 1, 10, 1, 10]),
+        dict(axes=[0, 1, 4], input_shape=[10, 10], output_shape=[1, 1, 10, 10, 1]),
+        dict(axes=[0, 2, 3], input_shape=[10, 10], output_shape=[1, 10, 1, 1, 10]),
+        dict(axes=[0, 2, 4], input_shape=[10, 10], output_shape=[1, 10, 1, 10, 1]),
+        dict(axes=[0, 3, 4], input_shape=[10, 10], output_shape=[1, 10, 10, 1, 1]),
+        dict(axes=[1, 2, 3], input_shape=[10, 10], output_shape=[10, 1, 1, 1, 10]),
+        dict(axes=[1, 2, 4], input_shape=[10, 10], output_shape=[10, 1, 1, 10, 1]),
+        dict(axes=[1, 3, 4], input_shape=[10, 10], output_shape=[10, 1, 10, 1, 1]),
+        dict(axes=[2, 3, 4], input_shape=[10, 10], output_shape=[10, 10, 1, 1, 1])]
+
+    test_data_4D = [
+        dict(axes=[0], input_shape=[3, 10, 10], output_shape=[1, 3, 10, 10]),
+        dict(axes=[1], input_shape=[3, 10, 10], output_shape=[3, 1, 10, 10]),
+        dict(axes=[2], input_shape=[3, 10, 10], output_shape=[3, 10, 1, 10]),
+        dict(axes=[3], input_shape=[3, 10, 10], output_shape=[3, 10, 10, 1]),
+        dict(axes=[3], input_shape=[3, 10, 10], output_shape=[3, 10, 10, 1]),
+        dict(axes=[0, 1], input_shape=[10, 10], output_shape=[1, 1, 10, 10]),
+        dict(axes=[0, 2], input_shape=[10, 10], output_shape=[1, 10, 1, 10]),
+        dict(axes=[0, 3], input_shape=[10, 10], output_shape=[1, 10, 10, 1]),
+        dict(axes=[1, 2], input_shape=[10, 10], output_shape=[10, 1, 1, 10]),
+        dict(axes=[1, 3], input_shape=[10, 10], output_shape=[10, 1, 10, 1]),
+        dict(axes=[2, 3], input_shape=[10, 10], output_shape=[10, 10, 1, 1])]
+
+    test_data_3D = [
+        dict(axes=[0], input_shape=[10, 10], output_shape=[1, 10, 10]),
+        dict(axes=[1], input_shape=[10, 10], output_shape=[10, 1, 10]),
+        dict(axes=[2], input_shape=[10, 10], output_shape=[10, 10, 1])]
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_unsqueeze_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_unsqueeze_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_unsqueeze_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_unsqueeze_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_unsqueeze_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_unsqueeze_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_unsqueeze_const_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_unsqueeze_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_unsqueeze_const_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_unsqueeze_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_unsqueeze_const_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_unsqueeze_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_upsample.py b/tests/layer_tests/onnx_tests/test_upsample.py
new file mode 100644
index 00000000000000..0a796dbbeabde0
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_upsample.py
@@ -0,0 +1,188 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import math
+import os
+
+import pytest
+import torch
+from common.layer_test_class import CommonLayerTest
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+
+
+class TestUpsample(OnnxRuntimeLayerTest):
+    def create_net(self, shape, mode, scales, opset, ir_version):
+        """
+            ONNX net                        IR net
+
+            Input->Upsample->Output   =>    Input->Resample
+
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        import onnx
+        from onnx import helper
+        from onnx import TensorProto
+
+        assert opset in [7, 9]
+
+        output_shape = shape.copy()
+        output_shape[-1] = math.floor(scales[-1] * shape[-1])
+        output_shape[-2] = math.floor(scales[-2] * shape[-2])
+        input = helper.make_tensor_value_info('input', TensorProto.FLOAT, shape)
+        output = helper.make_tensor_value_info('output', TensorProto.FLOAT, output_shape)
+
+        args = dict()
+        nodes = []
+        if opset == 7:
+            args['scales'] = scales
+        else:
+            node_scales_def = onnx.helper.make_node(
+                'Constant',
+                inputs=[],
+                outputs=['scales'],
+                value=helper.make_tensor(
+                    name='const_tensor',
+                    data_type=TensorProto.FLOAT,
+                    dims=[len(scales)],
+                    vals=scales,
+                ),
+            )
+            nodes.append(node_scales_def)
+
+        if mode:
+            args['mode'] = mode
+        node_def = helper.make_node(
+            'Upsample',
+            inputs=['input'] if opset == 7 else ['input', 'scales'],
+            outputs=['output'],
+            **args
+        )
+        nodes.append(node_def)
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            nodes,
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def,
+                                     producer_name='test_model',
+                                     opset_imports=[helper.make_opsetid("", opset)])
+
+        #   Create reference IR net
+        mode_to_resample_type = {None: 'caffe.ResampleParameter.NEAREST',
+                                 'nearest': 'caffe.ResampleParameter.NEAREST',
+                                 'linear': 'caffe.ResampleParameter.LINEAR'}
+        assert mode in mode_to_resample_type
+
+        ref_net = None
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape=[1, 3, 10, 12], scales=[1., 1., 2., 2.]),
+                 dict(shape=[1, 3, 10, 12], scales=[1., 1., 2.5, 2.5]),
+                 dict(shape=[1, 3, 10, 12], scales=[1., 1., 2.5, 2.])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("mode", [None, 'nearest'])
+    @pytest.mark.parametrize("opset", [7, 9])
+    @pytest.mark.nightly
+    def test_upsample_nearest(self, params, mode, opset, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, mode=mode, opset=opset, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("opset", [7, 9])
+    @pytest.mark.nightly
+    @pytest.mark.xfail(reason='Both onnxruntime and caffe2 calculate linear upsampling differently from IE')
+    def test_upsample_linear(self, params, opset, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, mode='linear', opset=opset, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+
+class PytorchLayerTest(CommonLayerTest):
+    def produce_model_path(self, framework_model, save_path):
+        path = os.path.join(save_path, 'model.onnx')
+        self.torch_model = framework_model['model']
+        torch.onnx.export(self.torch_model, framework_model['var'], path, output_names=['output'])
+        assert os.path.isfile(path), "model.onnx haven't been saved here: {}".format(save_path)
+        return path
+
+    def get_framework_results(self, inputs_dict, model_path):
+        x = torch.tensor(inputs_dict['input'], dtype=torch.float32)
+        return {'output': self.torch_model(x).numpy()}
+
+
+class UpsampleModel(torch.nn.Module):
+    def __init__(self, mode, size, scale_factor):
+        super(UpsampleModel, self).__init__()
+        args = dict()
+        if mode:
+            args['mode'] = mode
+        if scale_factor:
+            args['scale_factor'] = scale_factor
+        elif size:
+            args['size'] = size
+        self.upsample = torch.nn.modules.upsampling.Upsample(**args)
+
+
+class TestPytorchUpsample(PytorchLayerTest):
+    def create_net(self, shape, mode, size, scale_factor, ir_version):
+        """
+            Pytorch net                        IR net
+
+            Input->Upsample->Output   =>    Input->Resample
+
+        """
+
+        output_shape = shape.copy()
+        if size:
+            output_shape[2] = size[0]
+            output_shape[3] = size[1]
+        elif scale_factor:
+            output_shape[2] = scale_factor * output_shape[2]
+            output_shape[3] = scale_factor * output_shape[3]
+
+        #   Create Pytorch model
+        model = UpsampleModel(mode, size, scale_factor)
+
+        #   Create reference IR net
+        mode_to_resample_type = {None: 'caffe.ResampleParameter.NEAREST',
+                                 'nearest': 'caffe.ResampleParameter.NEAREST',
+                                 'bilinear': 'caffe.ResampleParameter.LINEAR'}
+        assert mode in mode_to_resample_type
+
+        ref_net = None
+
+        return {'model': model, 'var': torch.randn(shape)}, ref_net
+
+    test_data_precommit = [dict(shape=[1, 3, 10, 10], size=(25, 25), scale_factor=None),
+                           dict(shape=[1, 3, 10, 10], size=None, scale_factor=2)]
+
+    test_data = [dict(shape=[1, 3, 10, 10], size=(20, 20), scale_factor=None),
+                 dict(shape=[1, 3, 10, 10], size=(25, 25), scale_factor=None),
+                 dict(shape=[1, 3, 10, 10], size=None, scale_factor=2)]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.parametrize("mode", [None, 'nearest'])
+    def test_pytorch_upsample_precommit(self, params, mode, ie_device, precision, ir_version, temp_dir):
+        if ie_device == 'GPU':
+            pytest.skip('Linear upsampling not supported on GPU')
+        self._test(*self.create_net(**params, mode=mode, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.parametrize("mode", [None, 'nearest', 'bilinear'])
+    @pytest.mark.nightly
+    def test_pytorch_upsample(self, params, mode, ie_device, precision, ir_version, temp_dir):
+        if ie_device == 'GPU' and mode == 'bilinear':
+            pytest.skip('Linear upsampling not supported on GPU')
+        self._test(*self.create_net(**params, mode=mode, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_where.py b/tests/layer_tests/onnx_tests/test_where.py
new file mode 100644
index 00000000000000..577685d65fa2f8
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_where.py
@@ -0,0 +1,97 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestWhere(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randint(0, 2, inputs_dict[input]).astype(np.bool)
+        return inputs_dict
+
+    def create_net(self, condition_shape, shape_than, else_shape, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->Where->Output   =>    Input->Select
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input_cond = helper.make_tensor_value_info('input_cond', TensorProto.BOOL, condition_shape)
+        input_than = helper.make_tensor_value_info('input_than', TensorProto.BOOL, shape_than)
+        input_else = helper.make_tensor_value_info('input_else', TensorProto.BOOL, else_shape)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, condition_shape)
+
+        node_def = helper.make_node(
+            'Where',
+            inputs=['input_cond', 'input_than', 'input_else'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input_cond, input_than, input_else],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input_cond': {'kind': 'op', 'type': 'Parameter'},
+                'input_cond_data': {'shape': condition_shape, 'kind': 'data'},
+
+                'input_than': {'kind': 'op', 'type': 'Parameter'},
+                'input_than_data': {'shape': shape_than, 'kind': 'data'},
+
+                'input_else': {'kind': 'op', 'type': 'Parameter'},
+                'input_else_data': {'shape': else_shape, 'kind': 'data'},
+
+                'node': {'kind': 'op', 'type': 'Select'},
+                'node_data': {'shape': condition_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input_cond', 'input_cond_data'),
+                                   ('input_than', 'input_than_data'),
+                                   ('input_else', 'input_else_data'),
+                                   ('input_cond_data', 'node'),
+                                   ('input_than_data', 'node'),
+                                   ('input_else_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(condition_shape=[4, 6], shape_than=[4, 6], else_shape=[4, 6]),
+                 dict(condition_shape=[4, 6], shape_than=[4, 6], else_shape=[1, 6]),
+                 dict(condition_shape=[15, 3, 5], shape_than=[15, 1, 5], else_shape=[15, 3, 5]),
+                 dict(condition_shape=[2, 3, 4, 5], shape_than=[], else_shape=[2, 3, 4, 5]),
+                 dict(condition_shape=[2, 3, 4, 5], shape_than=[5], else_shape=[2, 3, 4, 5]),
+                 dict(condition_shape=[2, 3, 4, 5], shape_than=[2, 1, 1, 5], else_shape=[2, 3, 4, 5]),
+                 dict(condition_shape=[2, 3, 4, 5], shape_than=[2, 3, 4, 5], else_shape=[1, 3, 1, 5]),
+                ]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_where(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/onnx_tests/test_xor.py b/tests/layer_tests/onnx_tests/test_xor.py
new file mode 100644
index 00000000000000..8b6626d524181d
--- /dev/null
+++ b/tests/layer_tests/onnx_tests/test_xor.py
@@ -0,0 +1,268 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.onnx_layer_test_class import OnnxRuntimeLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestXor(OnnxRuntimeLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randint(0, 2, inputs_dict[input]).astype(np.bool)
+        return inputs_dict
+
+    def create_net(self, shape1, shape2, ir_version):
+        """
+            ONNX net                                  IR net
+
+            Input->Xor with 2nd input->Output   =>    Input->LogicalXor
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input1 = helper.make_tensor_value_info('input1', TensorProto.BOOL, shape1)
+        input2 = helper.make_tensor_value_info('input2', TensorProto.BOOL, shape2)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, shape1)
+
+        node_def = helper.make_node(
+            'Xor',
+            inputs=['input1', 'input2'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_def],
+            'test_model',
+            [input1, input2],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input1': {'kind': 'op', 'type': 'Parameter'},
+                'input1_data': {'shape': shape1, 'kind': 'data'},
+                'input2': {'kind': 'op', 'type': 'Parameter'},
+                'input2_data': {'shape': shape2, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'LogicalXor'},
+                'node_data': {'shape': shape1, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input1', 'input1_data'),
+                                   ('input2', 'input2_data'),
+                                   ('input1_data', 'node'),
+                                   ('input2_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    def create_net_one_const(self, shape1, shape2, ir_version):
+        """
+            ONNX net                              IR net
+
+            Input->Xor with const->Output   =>    Input->LogicalXor
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        input = helper.make_tensor_value_info('input', TensorProto.BOOL, shape1)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, shape1)
+
+        const = np.random.randint(0, 2, shape2).astype(np.bool)
+
+        node_const_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.BOOL,
+                dims=const.shape,
+                vals=const.flatten(),
+            ),
+        )
+
+        node_def = helper.make_node(
+            'Xor',
+            inputs=['input', 'const'],
+            outputs=['output']
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const_def, node_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape1, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': const.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': const.shape, 'kind': 'data'},
+                'node': {'kind': 'op', 'type': 'LogicalXor'},
+                'node_data': {'shape': shape1, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'node'),
+                                   ('const_data', 'node'),
+                                   ('node', 'node_data'),
+                                   ('node_data', 'result')])
+
+        return onnx_net, ref_net
+
+    def create_net_const(self, shape1, shape2, ir_version):
+        """
+            ONNX net                                          IR net
+
+            Input->Concat with const xor const->Output   =>   Input->Concat
+        """
+
+        #
+        #   Create ONNX model
+        #
+
+        from onnx import helper
+        from onnx import TensorProto
+
+        concat_axis = 0
+        output_shape = list(shape1)
+        output_shape[concat_axis] *= 2
+
+        input = helper.make_tensor_value_info('input', TensorProto.BOOL, shape1)
+        output = helper.make_tensor_value_info('output', TensorProto.BOOL, output_shape)
+
+        const1 = np.random.randint(0, 2, shape1).astype(np.bool)
+        const2 = np.random.randint(0, 2, shape2).astype(np.bool)
+
+        node_const1_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const1'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.BOOL,
+                dims=const1.shape,
+                vals=const1.flatten(),
+            ),
+        )
+
+        node_const2_def = helper.make_node(
+            'Constant',
+            inputs=[],
+            outputs=['const2'],
+            value=helper.make_tensor(
+                name='const_tensor',
+                data_type=TensorProto.BOOL,
+                dims=const2.shape,
+                vals=const2.flatten(),
+            ),
+        )
+
+        node_def = helper.make_node(
+            'Xor',
+            inputs=['const1', 'const2'],
+            outputs=['node_out']
+        )
+
+        node_concat_def = helper.make_node(
+            'Concat',
+            inputs=['input', 'node_out'],
+            outputs=['output'],
+            axis=concat_axis
+        )
+
+        # Create the graph (GraphProto)
+        graph_def = helper.make_graph(
+            [node_const1_def, node_const2_def, node_def, node_concat_def],
+            'test_model',
+            [input],
+            [output],
+        )
+
+        # Create the model (ModelProto)
+        onnx_net = helper.make_model(graph_def, producer_name='test_model')
+
+        #   Create reference IR net
+        constant_calculated = np.logical_xor(const1, const2)
+
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': const1.shape, 'kind': 'data'},
+                'input_const_data': {'kind': 'data', 'value': constant_calculated.flatten()},
+                'const': {'kind': 'op', 'type': 'Const'},
+                'const_data': {'shape': const1.shape, 'kind': 'data'},
+                'concat': {'kind': 'op', 'type': 'Concat', 'axis': concat_axis},
+                'concat_data': {'shape': output_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_const_data', 'const'),
+                                   ('const', 'const_data'),
+                                   ('input_data', 'concat'),
+                                   ('const_data', 'concat'),
+                                   ('concat', 'concat_data'),
+                                   ('concat_data', 'result')])
+
+        return onnx_net, ref_net
+
+    test_data = [dict(shape1=[4, 6], shape2=[4, 6]),
+                 dict(shape1=[4, 6, 8], shape2=[4, 6, 8]),
+                 dict(shape1=[4, 6, 8, 10], shape2=[4, 6, 8, 10]),
+                 dict(shape1=[4, 6, 8, 10, 12], shape2=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_xor(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_xor_one_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_one_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_xor_const(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_net_const(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/requirements.txt b/tests/layer_tests/requirements.txt
new file mode 100644
index 00000000000000..717c816c904b8c
--- /dev/null
+++ b/tests/layer_tests/requirements.txt
@@ -0,0 +1,2 @@
+requests>=2.25.1
+numpy~=1.19.5
diff --git a/tests/layer_tests/tensorflow_tests/conftest.py b/tests/layer_tests/tensorflow_tests/conftest.py
new file mode 100644
index 00000000000000..bbf635f56809ac
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/conftest.py
@@ -0,0 +1,12 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import inspect
+
+from common.layer_test_class import get_params
+
+
+def pytest_generate_tests(metafunc):
+    test_gen_attrs_names = list(inspect.signature(get_params).parameters)
+    params = get_params()
+    metafunc.parametrize(test_gen_attrs_names, params, scope="function")
diff --git a/tests/layer_tests/tensorflow_tests/permutation_utils.py b/tests/layer_tests/tensorflow_tests/permutation_utils.py
new file mode 100644
index 00000000000000..c26c7b690f72ec
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/permutation_utils.py
@@ -0,0 +1,22 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+
+from mo.ops.op import PermuteAttrs
+
+
+def permute_nhwc_to_nchw(shape):
+    perm = PermuteAttrs.get_nhwc_to_nchw_permutation(len(shape)).perm
+    new_shape = np.array(shape)[perm]
+    return new_shape
+
+
+def permute_nchw_to_nhwc(shape):
+    perm = PermuteAttrs.get_nchw_to_nhwc_permutation(len(shape)).perm
+    new_shape = np.array(shape)[perm]
+    return new_shape
+
+
+def permute_axis(axis, permutation_inv):
+    return permutation_inv[axis]
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Add.py b/tests/layer_tests/tensorflow_tests/test_tf_Add.py
new file mode 100644
index 00000000000000..46508f610d7a6f
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Add.py
@@ -0,0 +1,281 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestAdd(CommonTFLayerTest):
+    def create_add_placeholder_const_net(self, x_shape, y_shape, ir_version):
+        """
+            Tensorflow net                  IR net
+
+            Placeholder->Add       =>       Placeholder->Eltwise or Power or ScaleShift
+                         /                               /
+            Const-------/                   Const-------/
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            tf_x_shape = x_shape.copy()
+            tf_y_shape = y_shape.copy()
+            # reshaping
+            if len(tf_x_shape) >= 3:
+                tf_x_shape.append(tf_x_shape.pop(1))
+            if len(tf_y_shape) >= 3:
+                tf_y_shape.append(tf_y_shape.pop(1))
+
+            x = tf.compat.v1.placeholder(tf.float32, tf_x_shape, 'Input')
+            constant_value = np.random.randint(-256, 256, tf_y_shape).astype(np.float32)
+            if (constant_value == 0).all():
+                # Avoid elimination of the layer from IR
+                constant_value = constant_value + 1
+            y = tf.constant(constant_value)
+
+            add = tf.add(x, y, name="Operation")
+            add_shape = add.shape.as_list()
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        if len(add_shape) >= 3:
+            # Permute add_shape to (N,C,...) format
+            order = [0, len(add_shape) - 1] + list(range(1, len(add_shape) - 1))
+            add_shape = [add_shape[i] for i in order]
+
+        y_shape_to_compare = tf_y_shape.copy()
+        while len(y_shape_to_compare) < len(x_shape):
+            # Expand shape of constant with 1
+            y_shape_to_compare = [1] + y_shape_to_compare
+            constant_value = np.expand_dims(constant_value, axis=0)
+
+        if len(y_shape_to_compare) >= 3:
+            # Permute constant_value to (N,C,...) format for correct further reshape
+            order = [0, len(y_shape_to_compare) - 1] + list(range(1, len(y_shape_to_compare) - 1))
+            y_shape_to_compare = [y_shape_to_compare[i] for i in order]
+            constant_value = np.transpose(constant_value, order)
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    # TODO: implement tests for 2 Consts + Add
+
+    test_data_1D = [
+        # Power
+        dict(x_shape=[1], y_shape=[1]),
+        # Eltwise
+        pytest.param(dict(x_shape=[3], y_shape=[3]), marks=pytest.mark.xfail(reason="*-19180"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_1D)
+    @pytest.mark.nightly
+    def test_add_placeholder_const_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_2D = [
+        # Power
+        dict(x_shape=[1, 1], y_shape=[1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3], y_shape=[1, 3]),
+        # Eltwise
+        pytest.param(dict(x_shape=[3, 1], y_shape=[3, 1]), marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[2, 3], y_shape=[2, 3])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_2D)
+    @pytest.mark.nightly
+    def test_add_placeholder_const_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_3D = [
+        # Power
+        dict(x_shape=[1, 1, 1], y_shape=[1, 1, 1]),
+        # ScaleShift
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[1, 3, 1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 3], y_shape=[1, 1, 3]),
+                     marks=[pytest.mark.xfail(reason="*-19053"), pytest.mark.xfail(reason="*-18830")]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 3, 224], y_shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_add_placeholder_const_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_4D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1], y_shape=[1, 1, 1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1], y_shape=[1, 3, 1, 1]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 1, 3], y_shape=[1, 1, 1, 3]), marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[1, 3, 222, 224], y_shape=[1, 3, 222, 224])
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_add_placeholder_const_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1, 1], y_shape=[1, 1, 1, 1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[1, 3, 1, 1, 1]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 1, 1, 3], y_shape=[1, 1, 1, 1, 3]),
+                     marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[1, 3, 50, 100, 224], y_shape=[1, 3, 50, 100, 224])
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_add_placeholder_const_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version=ir_version, temp_dir=temp_dir)
+
+    ###############################################################################################
+    #                                                                                             #
+    #                                       Broadcast cases                                       #
+    #                                                                                             #
+    ###############################################################################################
+
+    test_data_broadcast_1D = [
+        # Power
+        dict(x_shape=[3], y_shape=[1])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_1D)
+    @pytest.mark.nightly
+    def test_add_placeholder_const_broadcast_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version=ir_version, temp_dir=temp_dir)
+
+    test_data_broadcast_2D = [
+        # Power
+        dict(x_shape=[1, 1], y_shape=[1]),
+        # Power
+        dict(x_shape=[1, 3], y_shape=[1]),
+        # ScaleShift
+        dict(x_shape=[1, 3], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[3, 1], y_shape=[3]),
+        # Eltwise
+        pytest.param(dict(x_shape=[3, 1], y_shape=[1, 3, 1, 1]), marks=pytest.mark.xfail(reason="*-19051"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_2D)
+    @pytest.mark.nightly
+    def test_add_placeholder_const_broadcast_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version=ir_version, temp_dir=temp_dir)
+
+    test_data_broadcast_3D = [
+        # Power
+        dict(x_shape=[1, 1, 1], y_shape=[1]),
+        # Power
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # ScaleShift
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[3]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[3, 1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 1], y_shape=[3, 1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[3, 1, 224], y_shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[2, 3, 1], y_shape=[1, 3, 2]), marks=pytest.mark.xfail(reason="*-19053")),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_3D)
+    @pytest.mark.nightly
+    def test_add_placeholder_const_broadcast_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version=ir_version, temp_dir=temp_dir)
+
+    test_data_broadcast_4D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1], y_shape=[1]),
+        # Power
+        dict(x_shape=[1, 3, 1, 1], y_shape=[1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1], y_shape=[3]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 100, 224], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 1, 1, 3], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 1], y_shape=[3, 1]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 2], y_shape=[3, 1, 2]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 2], y_shape=[1, 3, 2]),
+        # Eltwise
+        dict(x_shape=[1, 3, 100, 224], y_shape=[1, 1, 1, 224]),
+        # Eltwise
+        dict(x_shape=[2, 3, 1, 2], y_shape=[1, 3, 2, 1])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_4D)
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_add_placeholder_const_broadcast_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version=ir_version, temp_dir=temp_dir)
+
+    test_data_broadcast_5D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1, 1], y_shape=[1]),
+        # Power
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 1, 1, 1, 3], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[3, 1]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 1, 2], y_shape=[1, 3, 2]),
+        # Eltwise
+        dict(x_shape=[1, 3, 5, 1, 2], y_shape=[5, 3, 2, 1]),
+        # Eltwise
+        dict(x_shape=[1, 3, 50, 100, 224], y_shape=[1, 1, 1, 1, 224]),
+        # Eltwise
+        dict(x_shape=[2, 3, 1, 2, 1], y_shape=[1, 3, 2, 1, 1])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_5D)
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_add_placeholder_const_broadcast_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_add_placeholder_const_net(**params, ir_version=ir_version), ie_device, precision,
+                   ir_version=ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_BatchToSpace.py b/tests/layer_tests/tensorflow_tests/test_tf_BatchToSpace.py
new file mode 100644
index 00000000000000..f7d46dfcef4dba
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_BatchToSpace.py
@@ -0,0 +1,80 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestBatchToSpace(CommonTFLayerTest):
+    def create_batch_to_space_net(self, in_shape, crops_value, block_shape_value, out_shape, ir_version):
+        """
+            Tensorflow net               IR net
+
+            Input->BatchToSpace        =>      Input->BatchToSpace
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            x = tf.compat.v1.placeholder(tf.float32, in_shape, 'Input')
+            crops = tf.constant(crops_value)
+            block_shape = tf.constant(block_shape_value)
+            tf.batch_to_space(x, block_shape, crops, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    test_data_4D = [
+        dict(in_shape=[4, 1, 1, 3], block_shape_value=[1], crops_value=[[0, 0]],
+             out_shape=[4, 1, 1, 3]),
+        dict(in_shape=[4, 1, 1, 3], block_shape_value=[2, 2], crops_value=[[0, 0], [0, 0]],
+             out_shape=[1, 2, 2, 3]),
+        dict(in_shape=[60, 100, 30, 30], block_shape_value=[3, 2], crops_value=[[1, 5], [4, 1]],
+             out_shape=[2, 2, 1, 1]),
+        # todo: enable these tests after supporting the general case on CPU
+        # dict(in_shape=[4, 1, 1, 1], block_shape_value=[2, 1, 2], crops_value=[[0, 0], [0, 0], [0, 0]],
+        #      out_shape=[]),
+        # dict(in_shape=[12, 1, 1, 3], block_shape_value=[3, 2, 2], crops_value=[[1, 0], [0, 1], [1, 1]],
+        #      out_shape=[1, 2, 1, 4]),
+        # dict(in_shape=[36, 2, 2, 3], block_shape_value=[2, 3, 3], crops_value=[[1, 0], [0, 0], [2, 2]],
+        #      out_shape=[2, 3, 6, 5])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_batch_to_space_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_batch_to_space_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = [
+        dict(in_shape=[72, 2, 1, 4, 2], block_shape_value=[3, 4, 2], crops_value=[[1, 2], [0, 0], [3, 0]],
+             out_shape=[3, 3, 4, 5, 2]),
+        # todo: enable these tests after supporting the general case on CPU
+        # dict(in_shape=[144, 2, 1, 4, 1], block_shape_value=[3, 4, 2, 2],
+        #      crops_value=[[1, 2], [0, 0], [3, 0], [0, 0]], out_shape=[3, 3, 4, 5, 2]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_batch_to_space_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_batch_to_space_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_BiasAdd.py b/tests/layer_tests/tensorflow_tests/test_tf_BiasAdd.py
new file mode 100644
index 00000000000000..b7e997195f58ba
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_BiasAdd.py
@@ -0,0 +1,194 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestBiasAdd(CommonTFLayerTest):
+    def create_bias_add_placeholder_const_net(self, shape, ir_version):
+        """
+            Tensorflow net                      IR net
+
+            Placeholder->BiasAdd       =>       Placeholder->Power or ScaleShift
+                         /                                   /
+            Const-------/                       Const-------/
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+        import numpy as np
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            tf_x_shape = shape.copy()
+            # reshaping
+            if len(tf_x_shape) >= 3:
+                tf_x_shape.append(tf_x_shape.pop(1))
+            tf_y_shape = tf_x_shape[-1:]
+
+            x = tf.compat.v1.placeholder(tf.float32, tf_x_shape, 'Input')
+            constant_value = np.random.randint(0, 1, tf_y_shape).astype(np.float32)
+            if (constant_value == 0).all():
+                # Avoid elimination of the layer from IR
+                constant_value = constant_value + 1
+            y = tf.constant(constant_value)
+
+            tf.nn.bias_add(x, y, name="Operation")
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    def create_bias_add_2_consts_net(self, shape, ir_version):
+        """
+            Tensorflow net                         IR net
+
+            Const->BiasAdd-->Concat       =>       Const---->Concat
+                    /        /                                  /
+            Const--/        /                      Placeholder-/
+                           /
+            Placeholder---/
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+        import numpy as np
+
+        tf.compat.v1.reset_default_graph()
+
+        tf_concat_axis = -1
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            tf_x_shape = shape.copy()
+            # reshaping
+            if len(tf_x_shape) >= 3:
+                tf_x_shape.append(tf_x_shape.pop(1))
+            tf_y_shape = tf_x_shape[-1:]
+
+            constant_value_x = np.random.randint(-256, 256, tf_x_shape).astype(np.float32)
+            x = tf.constant(constant_value_x)
+            constant_value_y = np.random.randint(-256, 256, tf_y_shape).astype(np.float32)
+            y = tf.constant(constant_value_y)
+
+            add = tf.nn.bias_add(x, y, name="Operation")
+            add_shape = add.shape.as_list()
+            add_value = add.eval()
+
+            placeholder = tf.compat.v1.placeholder(tf.float32, tf_x_shape, 'Input')  # Input_1 in graph_def
+
+            concat = tf.concat([placeholder, add], axis=tf_concat_axis, name='Operation')
+            concat_shape = concat.shape.as_list()
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        # Format axis to positive value
+        concat_ax = tf_concat_axis if tf_concat_axis >= 0 else tf_concat_axis + len(shape)
+        if len(shape) >= 3:
+            # Permute shapes to (N,C,...) format
+            order = [0, len(concat_shape) - 1] + list(range(1, len(concat_shape) - 1))
+            concat_shape = [concat_shape[i] for i in order]
+            concat_ax = order.index(concat_ax)
+            add_value = np.transpose(add_value, order)
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    test_data_2D = [
+        dict(shape=[1, 1]),
+        dict(shape=[1, 224])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_2D)
+    @pytest.mark.nightly
+    def test_bias_add_placeholder_const_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_bias_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_2D)
+    @pytest.mark.nightly
+    def test_bias_add_2_consts_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_bias_add_2_consts_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_3D = [
+        pytest.param(dict(shape=[1, 1, 224]), marks=pytest.mark.xfail(reason="*-19053")),
+        pytest.param(dict(shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_bias_add_placeholder_const_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_bias_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_bias_add_2_consts_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_bias_add_2_consts_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_4D = [
+        dict(shape=[1, 1, 100, 224]),
+        dict(shape=[1, 3, 100, 224])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_bias_add_placeholder_const_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_bias_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_bias_add_2_consts_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_bias_add_2_consts_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = [
+        dict(shape=[1, 1, 50, 100, 224]),
+        dict(shape=[1, 3, 220, 222, 224])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_bias_add_placeholder_const_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_bias_add_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_bias_add_2_consts_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_bias_add_2_consts_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Bucketize.py b/tests/layer_tests/tensorflow_tests/test_tf_Bucketize.py
new file mode 100644
index 00000000000000..10d0bc6de74386
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Bucketize.py
@@ -0,0 +1,83 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+import tensorflow as tf
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestBucketize(CommonTFLayerTest):
+    def create_bucketize_net(self, input_shape, input_type, boundaries_size, ir_version):
+        """
+            Tensorflow net:                     IR net:
+                 Input            =>      Input        Boundaries
+                   |                           \       /
+               Bucketize                       Bucketize
+           {attrs: boundaries}
+        """
+
+        # create Tensorflow model
+        tf.compat.v1.reset_default_graph()
+        with tf.compat.v1.Session() as sess:
+            x = tf.compat.v1.placeholder(input_type, input_shape, 'Input')
+            constant_value = np.arange(-boundaries_size * 5, boundaries_size * 5, 10, dtype=np.float32)
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        # create reference IR net
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': input_shape, 'kind': 'data'},
+                'boundaries_input_data': {'shape': constant_value.shape, 'kind': 'data'},
+                'boundaries': {'type': 'Const', 'kind': 'op'},
+                'boundaries_data': {'kind': 'data', 'shape': constant_value.shape},
+                'bucketize': {'kind': 'op', 'type': 'Bucketize'},
+                'bucketize_data': {'shape': input_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'bucketize', {'in': 0}),
+                                   ('boundaries_input_data', 'boundaries'),
+                                   ('boundaries', 'boundaries_data'),
+                                   ('boundaries_data', 'bucketize', {'in': 1}),
+                                   ('bucketize', 'bucketize_data'),
+                                   ('bucketize_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_float32 = [
+        dict(input_shape=[5], input_type=tf.float32, boundaries_size=1),
+        dict(input_shape=[5], input_type=tf.float32, boundaries_size=3),
+        dict(input_shape=[4, 8], input_type=tf.float32, boundaries_size=5),
+        dict(input_shape=[2, 4, 7], input_type=tf.float32, boundaries_size=10),
+        dict(input_shape=[2, 4, 7, 8], input_type=tf.float32, boundaries_size=12),
+        dict(input_shape=[2, 4, 7, 8, 10], input_type=tf.float32, boundaries_size=14)]
+
+    @pytest.mark.parametrize("params", test_data_float32)
+    @pytest.mark.nightly
+    def test_bucketize_float32(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_bucketize_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_int32 = [
+        dict(input_shape=[5], input_type=tf.int32, boundaries_size=1),
+        dict(input_shape=[5], input_type=tf.int32, boundaries_size=3),
+        dict(input_shape=[4, 8], input_type=tf.int32, boundaries_size=5),
+        dict(input_shape=[2, 4, 7], input_type=tf.int32, boundaries_size=10),
+        dict(input_shape=[2, 4, 7, 8], input_type=tf.float32, boundaries_size=12),
+        dict(input_shape=[2, 4, 7, 8, 10], input_type=tf.float32, boundaries_size=14)]
+
+    @pytest.mark.parametrize("params", test_data_int32)
+    @pytest.mark.nightly
+    def test_bucketize_int32(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_bucketize_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Ceil.py b/tests/layer_tests/tensorflow_tests/test_tf_Ceil.py
new file mode 100644
index 00000000000000..c516a598b88f64
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Ceil.py
@@ -0,0 +1,78 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestCeil(CommonTFLayerTest):
+    def create_ceil_net(self, shape, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->Ceil       =>       Input->Ceil
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) >= 3:
+                shapes.append(shapes.pop(1))
+            input = tf.compat.v1.placeholder(tf.float32, shapes, 'Input')
+
+            tf.math.ceil(input, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'Ceiling': {'kind': 'op', 'type': 'Ceiling'},
+                'Ceiling_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'Ceiling'),
+                                   ('Ceiling', 'Ceiling_data'),
+                                   ('Ceiling_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_precommit = [dict(shape=[3, 2, 3, 7, 6])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_ceil_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_ceil_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[1]),
+                 dict(shape=[2, 5]),
+                 dict(shape=[5, 3, 7, 4]),
+                 dict(shape=[3, 2, 3, 7, 6])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_ceil(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_ceil_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Concat.py b/tests/layer_tests/tensorflow_tests/test_tf_Concat.py
new file mode 100644
index 00000000000000..cfa6ca0e958c35
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Concat.py
@@ -0,0 +1,114 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestConcat(CommonTFLayerTest):
+    def create_concat_net(self, shape, axis, ir_version):
+        """
+            Tensorflow net               IR net
+
+            Input->Concat        =>      Input->Concat
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+
+            ax = axis
+
+            input_shape_x = shape.copy()
+            # reshaping
+            if len(input_shape_x) >= 3:
+                input_shape_x.append(input_shape_x.pop(1))
+
+            # TODO: add concat with const inputs to check fusing (as in ONNX)
+
+            x = tf.compat.v1.placeholder(tf.float32, input_shape_x, 'Input')
+            y = tf.compat.v1.placeholder(tf.float32, input_shape_x, 'Input')  # Input_1 in graph_def
+
+            concat = tf.concat([x, y], axis=ax, name='Operation')
+            concat_shape = concat.shape.as_list()
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        # Format axis to positive value
+        concat_ax = axis if axis >= 0 else axis + len(shape)
+        if len(shape) >= 3:
+            # Permute shape to (N,C,...) format and compute correct axis value
+            order = [0, len(concat_shape) - 1] + list(range(1, len(concat_shape) - 1))
+            concat_shape = [concat_shape[i] for i in order]
+            concat_ax = order.index(concat_ax)
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    # TODO: create tests for concat with 1 input and multiple inputs
+
+    test_data_1D = [dict(shape=[1], axis=0),
+                    dict(shape=[1], axis=-1)]
+
+    @pytest.mark.parametrize("params", test_data_1D)
+    @pytest.mark.nightly
+    def test_concat_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_concat_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_2D = [dict(shape=[1, 224], axis=0),
+                    dict(shape=[1, 224], axis=-1)]
+
+    @pytest.mark.parametrize("params", test_data_2D)
+    @pytest.mark.nightly
+    def test_concat_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_concat_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_3D = [pytest.param(dict(shape=[1, 3, 224], axis=0), marks=pytest.mark.xfail(reason="*-19053")),
+                    pytest.param(dict(shape=[1, 3, 224], axis=-1), marks=pytest.mark.xfail(reason="*-19053")),
+                    pytest.param(dict(shape=[1, 3, 224], axis=2), marks=pytest.mark.xfail(reason="*-19053"))]
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_concat_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_concat_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_4D = [dict(shape=[1, 3, 100, 224], axis=0),
+                    dict(shape=[1, 3, 100, 224], axis=-1),
+                    dict(shape=[1, 3, 100, 224], axis=2)]
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_concat_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_concat_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = [dict(shape=[1, 3, 50, 100, 224], axis=0),
+                    dict(shape=[1, 3, 50, 100, 224], axis=-1),
+                    dict(shape=[1, 3, 50, 100, 224], axis=2)]
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_concat_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_concat_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_ELU.py b/tests/layer_tests/tensorflow_tests/test_tf_ELU.py
new file mode 100644
index 00000000000000..7e183cf9c3438d
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_ELU.py
@@ -0,0 +1,88 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestELU(CommonTFLayerTest):
+    def create_elu_net(self, shape, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->ELU       =>       Input->ELU
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) >= 4:
+                shapes.append(shapes.pop(1))
+            input = tf.compat.v1.placeholder(tf.float32, shapes, 'Input')
+
+            tf.nn.elu(input, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'ELU': {'kind': 'op', 'type': 'Elu'},
+                'ELU_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'ELU'),
+                                   ('ELU', 'ELU_data'),
+                                   ('ELU_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_precommit = [dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_elu_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        if ie_device == 'GPU':
+            pytest.skip("5D tensors is not supported on GPU")
+        self._test(*self.create_elu_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[10, 12]),
+                 dict(shape=[8, 10, 12]),
+                 dict(shape=[6, 8, 10, 12]),
+                 dict(shape=[4, 6, 8, 10, 12])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_elu(self, params, ie_device, precision, ir_version, temp_dir):
+        if ie_device == 'GPU':
+            pytest.skip("5D tensors is not supported on GPU")
+        self._test(*self.create_elu_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Eltwise.py b/tests/layer_tests/tensorflow_tests/test_tf_Eltwise.py
new file mode 100644
index 00000000000000..d6880f3e0a1e31
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Eltwise.py
@@ -0,0 +1,79 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestEltwise(CommonTFLayerTest):
+    def create_eltwise_net(self, shape, operation, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Inputs->Eltwise       =>       Inputs->Eltwise
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) >= 4:
+                shapes.append(shapes.pop(1))
+
+            x = tf.compat.v1.placeholder(tf.float32, shapes, 'Input')
+            y = tf.compat.v1.placeholder(tf.float32, shapes, 'Input')     # Input_1 in graph_def
+
+            if operation == 'sum':
+                tf.add(x, y, name='Operation')
+            elif operation == 'max':
+                tf.maximum(x, y, name='Operation')
+            elif operation == 'mul':
+                tf.multiply(x, y, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    test_data = []
+    for operation in ['sum', 'max', 'mul']:
+        test_data.extend([dict(shape=[1, 224], operation=operation),
+                          dict(shape=[1, 224, 224], operation=operation),
+                          dict(shape=[1, 3, 224, 224], operation=operation)])
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_eltwise(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_eltwise_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = []
+    for operation in ['sum', 'max', 'mul']:
+        test_data_5D.extend([dict(shape=[1, 3, 224, 224, 224], operation=operation)])
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.precommit
+    def test_eltwise_5D_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        if ie_device == 'GPU':
+            pytest.skip("5D tensors is not supported on GPU")
+        self._test(*self.create_eltwise_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_FakeQuantize.py b/tests/layer_tests/tensorflow_tests/test_tf_FakeQuantize.py
new file mode 100644
index 00000000000000..d4d8fe91b845cb
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_FakeQuantize.py
@@ -0,0 +1,125 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from mo.front.common.partial_infer.utils import int64_array
+from unit_tests.utils.graph import build_graph, regular_op_with_shaped_data, connect, \
+    shaped_data, connect_front, regular_op
+
+
+class TestFakeQuantize(CommonTFLayerTest):
+    def _prepare_input(self, inputs_dict, kwargs):
+        assert len(inputs_dict) == 1
+        assert np.array(list(inputs_dict.values())[0]) == np.array([11])
+        assert 'nudged_il' in kwargs and kwargs['nudged_il'] is not None
+        assert 'nudged_ih' in kwargs and kwargs['nudged_ih'] is not None
+        assert 'expected_step' in kwargs and kwargs['expected_step'] is not None
+
+        expected_nudged_input_min = kwargs['nudged_il']
+        expected_nudged_input_max = kwargs['nudged_ih']
+        expected_step = kwargs['expected_step']
+
+        return {list(inputs_dict.keys())[0]: np.array([
+            expected_nudged_input_min - expected_step,
+            expected_nudged_input_min - 0.01, expected_nudged_input_min,
+            expected_nudged_input_min + 0.01,
+            expected_nudged_input_min + expected_step - 0.01,
+            expected_nudged_input_min + expected_step,
+            expected_nudged_input_min + expected_step + 0.01,
+            expected_nudged_input_max - 0.01, expected_nudged_input_max,
+            expected_nudged_input_max + 0.01,
+            expected_nudged_input_max + expected_step
+        ])}
+
+    def create_fake_quantize_net(self, il, ih, num_bits, narrow_range, nudged_il, nudged_ih, expected_step, ir_version):
+        # original tf model
+        import tensorflow as tf
+        tf.compat.v1.reset_default_graph()
+        with tf.compat.v1.Session() as sess:
+            data = tf.compat.v1.placeholder(tf.float32, [11], 'parameter')
+            input_min = tf.constant(il, name='input_min')
+            input_max = tf.constant(ih, name='input_max')
+            tf.quantization.fake_quant_with_min_max_vars(data, input_min, input_max, num_bits, narrow_range, 'fq')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        # reference graph to compare with IR
+        ref_net = None
+        if check_ir_version(10, None, ir_version):
+            levels = 2 ** num_bits - int(narrow_range)
+
+            # data (shape, value) -> const (shape, vale) -> data (shape, no value)
+            const_for_layer_tests = lambda name, value: {
+                **{name + '_dd': {'kind': 'data', 'value': value, 'shape': value.shape}},
+                **{name: {'kind': 'op', 'type': 'Const'}},
+                **shaped_data(name + '_d', int64_array(value.shape))}
+
+            connect_const_for_layer_tests = lambda first_tensor_name, second_tensor_name: [
+                *connect_front(first_tensor_name + '_dd', first_tensor_name),
+                *connect(first_tensor_name, second_tensor_name)]
+
+            nodes = {
+                **regular_op_with_shaped_data('parameter', [11], {'type': 'Parameter'}),
+                **const_for_layer_tests('il', np.array([nudged_il], dtype=np.float32)),
+                **const_for_layer_tests('ih', np.array([nudged_ih], dtype=np.float32)),
+                **const_for_layer_tests('ol', np.array([nudged_il], dtype=np.float32)),
+                **const_for_layer_tests('oh', np.array([nudged_ih], dtype=np.float32)),
+                **regular_op_with_shaped_data('fq', [11], {'type': 'FakeQuantize', 'levels': levels}),
+                **regular_op('result', {'type': 'Result'}),
+            }
+            edges = [
+                *connect('parameter', '0:fq'),
+                *connect_const_for_layer_tests('il', '1:fq'),
+                *connect_const_for_layer_tests('ih', '2:fq'),
+                *connect_const_for_layer_tests('ol', '3:fq'),
+                *connect_const_for_layer_tests('oh', '4:fq'),
+                *connect('fq', 'result'),
+            ]
+            ref_net = build_graph(nodes, edges)
+
+        return tf_net, ref_net
+
+    test_data = [
+        # with8BitsNoScalingNoNudging
+        dict(il=0.0, ih=255.0, num_bits=8, narrow_range=False, nudged_il=0.0, nudged_ih=255.0, expected_step=1.0),
+        # with8BitsScalingAndNudgingDown
+        dict(il=0.5, ih=128.0, num_bits=8, narrow_range=False, nudged_il=0.0, nudged_ih=127.5, expected_step=0.5),
+        # with8BitsScalingAndNudgingUp
+        dict(il=-128.0, ih=-0.5, num_bits=8, narrow_range=False, nudged_il=-127.5, nudged_ih=0.0, expected_step=0.5),
+        # with8BitsScalingAndNudgingBetween
+        dict(il=-0.1, ih=127.4, num_bits=8, narrow_range=False, nudged_il=0.0, nudged_ih=127.5, expected_step=0.5),
+        # with8BitsNarrowRangeNoScalingNoNudging
+        dict(il=0.0, ih=254.0, num_bits=8, narrow_range=True, nudged_il=0.0, nudged_ih=254.0, expected_step=1.0),
+        # with8BitsNarrowRangeScalingAndNudgingDown
+        dict(il=0.1, ih=127.1, num_bits=8, narrow_range=True, nudged_il=0.0, nudged_ih=127.0, expected_step=0.5),
+        # with8BitsNarrowRangeScalingAndNudgingUp
+        dict(il=-127.1, ih=-0.1, num_bits=8, narrow_range=True, nudged_il=-127.0, nudged_ih=0.0, expected_step=0.5),
+        # with8BitsNarrowRangeScalingAndNudgingBetween
+        dict(il=-0.1, ih=126.9, num_bits=8, narrow_range=True, nudged_il=0.0, nudged_ih=127.0, expected_step=0.5),
+        # with7BitsNoScalingNoNudging
+        dict(il=0.0, ih=127.0, num_bits=7, narrow_range=False, nudged_il=0.0, nudged_ih=127.0, expected_step=1.0),
+        # with7BitsScalingAndNudgingDown
+        dict(il=0.5, ih=64.0, num_bits=7, narrow_range=False, nudged_il=0.0, nudged_ih=63.5, expected_step=0.5),
+        # with7BitsScalingAndNudgingUp
+        dict(il=-64.0, ih=-0.5, num_bits=7, narrow_range=False, nudged_il=-63.5, nudged_ih=0.0, expected_step=0.5),
+        # with7BitsScalingAndNudgingBetween
+        dict(il=-0.1, ih=63.4, num_bits=7, narrow_range=False, nudged_il=0.0, nudged_ih=63.5, expected_step=0.5),
+        # with7BitsNarrowRangeNoScalingNoNudging
+        dict(il=0.0, ih=126.0, num_bits=7, narrow_range=True, nudged_il=0.0, nudged_ih=126.0, expected_step=1.0),
+        # with7BitsNarrowRangeScalingAndNudgingDown
+        dict(il=0.1, ih=63.1, num_bits=7, narrow_range=True, nudged_il=0.0, nudged_ih=63.0, expected_step=0.5),
+        # with7BitsNarrowRangeScalingAndNudgingUp
+        dict(il=-63.1, ih=-0.1, num_bits=7, narrow_range=True, nudged_il=-63.0, nudged_ih=0.0, expected_step=0.5),
+        # with7BitsNarrowRangeScalingAndNudgingBetween
+        dict(il=-0.1, ih=62.9, num_bits=7, narrow_range=True, nudged_il=0.0, nudged_ih=63.0, expected_step=0.5)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_fake_quantize(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_fake_quantize_net(**params, ir_version=ir_version), ie_device, precision, ir_version,
+                   kwargs_to_prepare_input=params, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Floor.py b/tests/layer_tests/tensorflow_tests/test_tf_Floor.py
new file mode 100644
index 00000000000000..9c47f1f5910bcc
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Floor.py
@@ -0,0 +1,78 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestFloor(CommonTFLayerTest):
+    def create_floor_net(self, shape, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->Floor       =>       Input->Floor
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) >= 3:
+                shapes.append(shapes.pop(1))
+            input = tf.compat.v1.placeholder(tf.float32, shapes, 'Input')
+
+            tf.floor(input, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'Floor': {'kind': 'op', 'type': 'Floor'},
+                'Floor_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'Floor'),
+                                   ('Floor', 'Floor_data'),
+                                   ('Floor_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_precommit = [dict(shape=[3, 2, 3, 7, 6])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_floor_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_floor_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[1]),
+                 dict(shape=[2, 5]),
+                 dict(shape=[5, 3, 7, 4]),
+                 dict(shape=[3, 2, 3, 7, 6])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_floor(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_floor_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Identity.py b/tests/layer_tests/tensorflow_tests/test_tf_Identity.py
new file mode 100644
index 00000000000000..6e963e85c4e766
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Identity.py
@@ -0,0 +1,85 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestIdentity(CommonTFLayerTest):
+    def create_identity_net(self, shape, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->Identity->ReLU     =>     Input->ReLU
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            x_shape = shape.copy()
+            # reshaping
+            if len(x_shape) >= 3:
+                x_shape.append(x_shape.pop(1))
+
+            x = tf.compat.v1.placeholder(tf.float32, x_shape, 'Input')
+            id = tf.identity(x, name="Operation")
+            tf.nn.relu(id, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'inputX': {'kind': 'op', 'type': 'Parameter'},
+                'inputX_data': {'shape': shape, 'kind': 'data'},
+                'ReLU': {'kind': 'op', 'type': 'ReLU'},
+                'ReLU_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+            ref_net = build_graph(nodes_attributes,
+                                  [('inputX', 'inputX_data'),
+                                   ('inputX_data', 'ReLU'),
+                                   ('ReLU', 'ReLU_data'),
+                                   ('ReLU_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_precommit = [dict(shape=[1, 3, 50, 100, 224])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_identity_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_identity_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[1]),
+                 dict(shape=[1, 224]),
+                 pytest.param(dict(shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053")),
+                 dict(shape=[1, 3, 100, 224]),
+                 dict(shape=[1, 3, 50, 100, 224])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_identity(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_identity_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Log.py b/tests/layer_tests/tensorflow_tests/test_tf_Log.py
new file mode 100644
index 00000000000000..f979071395183e
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Log.py
@@ -0,0 +1,79 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestLog(CommonTFLayerTest):
+    def create_log_net(self, shape, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->Log       =>       Input->Log
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) >= 3:
+                shapes.append(shapes.pop(1))
+            input = tf.compat.v1.placeholder(tf.float32, shapes, 'Input')
+
+            tf.math.log(input, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'log': {'kind': 'op', 'type': 'Log'},
+                'log_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'log'),
+                                   ('log', 'log_data'),
+                                   ('log_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_precommit = [
+        pytest.param(dict(shape=[3, 2, 3, 7, 6]), marks=pytest.mark.skip(reason="Skipped until fixed"))]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_log_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_log_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[1]),
+                 dict(shape=[2, 5]),
+                 dict(shape=[5, 3, 7, 4]),
+                 dict(shape=[3, 2, 3, 7, 6])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_log(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_log_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_LogSoftmax.py b/tests/layer_tests/tensorflow_tests/test_tf_LogSoftmax.py
new file mode 100644
index 00000000000000..8cb3087cc3e2fc
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_LogSoftmax.py
@@ -0,0 +1,138 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+from distutils.version import LooseVersion
+
+import numpy as np
+import pytest
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from mo.front.common.partial_infer.utils import int64_array
+from unit_tests.utils.graph import build_graph
+
+
+class TestLogSoftmax(CommonTFLayerTest):
+    def create_log_softmax_net(self, shape, reduction_axis, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->LogSoftmax       =>       Input->Softmax->Log
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) >= 3:
+                shapes.append(shapes.pop(1))
+            input = tf.compat.v1.placeholder(tf.float32, shapes, 'Input')
+            if LooseVersion(tf.__version__) < LooseVersion('2.0.0'):
+                tf.nn.log_softmax(input, name='Operation', axis=reduction_axis)
+            else:
+                tf.nn.log_softmax(input, axis=reduction_axis, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        ref_net = None
+
+        reduce_sum_shape = np.copy(shape)
+        rank = len(shape)
+        if rank in {4, 5}:
+            reduction_axis = reduction_axis if reduction_axis >= 0 else rank + reduction_axis
+            if rank == 4:
+                reduction_axis = {0: 0, 1: 2, 2: 3, 3: 1}[reduction_axis]
+            else:
+                reduction_axis = {0: 0, 1: 2, 2: 3, 3: 4, 4: 1}[reduction_axis]
+
+        reduce_sum_shape[reduction_axis] = 1
+
+        converted_shape = shape if rank != 1 else shape[0]
+        if check_ir_version(10, None, ir_version):
+            ref_nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter', 'shape': converted_shape},
+                'input_data': {'shape': shape, 'kind': 'data', 'value': None},
+                'reduce_max_axis_val': {'shape': int64_array([reduction_axis]).shape,
+                                        'kind': 'data',
+                                        'value': int64_array([reduction_axis])},
+                'reduce_max_axis': {'type': 'Const', 'kind': 'op', 'shape': 1},
+                'reduce_max_axis_data': {'shape': int64_array([1]), 'kind': 'data', 'value': None},
+                'reduce_max': {'type': 'ReduceMax', 'kind': 'op', 'keep_dims': True},
+                'reduce_max_data': {'shape': reduce_sum_shape, 'kind': 'data', 'value': None},
+                'sub_first': {'type': 'Subtract', 'kind': 'op'},
+                'sub_first_data': {'shape': shape, 'kind': 'data', 'value': None},
+                'reduce_sum_axis_val': {'shape': int64_array([reduction_axis]).shape,
+                                        'kind': 'data',
+                                        'value': int64_array([reduction_axis])},
+                'reduce_sum_axis': {'type': 'Const', 'kind': 'op', 'shape': 1},
+                'reduce_sum_axis_data': {'shape': int64_array([1]), 'kind': 'data', 'value': None},
+                'reduce_sum': {'type': 'ReduceSum', 'kind': 'op', 'keep_dims': True},
+                'reduce_sum_data': {'shape': reduce_sum_shape, 'kind': 'data', 'value': None},
+                'exp': {'type': 'Exp', 'kind': 'op'},
+                'exp_data': {'shape': shape, 'kind': 'data', 'value': None},
+                'log': {'type': 'Log', 'kind': 'op'},
+                'log_data': {'shape': reduce_sum_shape, 'kind': 'data', 'value': None},
+                'sub_second': {'type': 'Subtract', 'kind': 'op'},
+                'sub_second_data': {'shape': shape, 'kind': 'data', 'value': None},
+                'result': {'kind': 'op', 'type': 'Result'},
+            }
+
+            ref_edges = [
+                ('input', 'input_data'),
+                ('reduce_max_axis_val', 'reduce_max_axis'),
+                ('reduce_max_axis', 'reduce_max_axis_data'),
+                ('reduce_max_axis_data', 'reduce_max', {'in': 1}),
+                ('reduce_max', 'reduce_max_data'),
+                ('input_data', 'reduce_max', {'out': 0, 'in': 0}),
+                ('input_data', 'sub_first', {'out': 0, 'in': 0}),
+                ('reduce_max_data', 'sub_first', {'in': 1}),
+                ('sub_first', 'sub_first_data'),
+                ('reduce_sum_axis_val', 'reduce_sum_axis'),
+                ('reduce_sum_axis', 'reduce_sum_axis_data'),
+                ('reduce_sum_axis_data', 'reduce_sum', {'in': 1}),
+                ('reduce_sum', 'reduce_sum_data'),
+                ('sub_first_data', 'exp'),
+                ('exp', 'exp_data'),
+                ('exp_data', 'reduce_sum', {'in': 0}),
+                ('reduce_sum_data', 'log'),
+                ('log', 'log_data'),
+                ('log_data', 'sub_second', {'in': 1}),
+                ('sub_second', 'sub_second_data'),
+                ('sub_first_data', 'sub_second', {'out': 0, 'in': 0}),
+                ('sub_second_data', 'result'),
+            ]
+
+            ref_net = build_graph(ref_nodes_attributes, ref_edges)
+
+        return tf_net, ref_net
+
+    test_data_precommit = [
+        pytest.param(dict(shape=[3, 2, 3, 7, 6], reduction_axis=-1),
+                     marks=pytest.mark.skip(reason="Skipped until fixed"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_log_softmax_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_log_softmax_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[1], reduction_axis=-1),
+                 dict(shape=[2, 5], reduction_axis=-1),
+                 dict(shape=[5, 3, 7, 4], reduction_axis=-1),
+                 dict(shape=[3, 2, 3, 7, 6], reduction_axis=-1)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_log_softmax(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_log_softmax_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Mul.py b/tests/layer_tests/tensorflow_tests/test_tf_Mul.py
new file mode 100644
index 00000000000000..a877e639d6571a
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Mul.py
@@ -0,0 +1,279 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestMul(CommonTFLayerTest):
+    def create_mul_placeholder_const_net(self, x_shape, y_shape, ir_version):
+        """
+            Tensorflow net                  IR net
+
+            Placeholder->Mul       =>       Placeholder->Eltwise or Power or ScaleShift
+                         /                               /
+            Const-------/                   Const-------/
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            tf_x_shape = x_shape.copy()
+            tf_y_shape = y_shape.copy()
+            # reshaping
+            if len(tf_x_shape) >= 3:
+                tf_x_shape.append(tf_x_shape.pop(1))
+            if len(tf_y_shape) >= 3:
+                tf_y_shape.append(tf_y_shape.pop(1))
+
+            x = tf.compat.v1.placeholder(tf.float32, tf_x_shape, 'Input')
+            constant_value = np.random.randint(-255, 255, tf_y_shape).astype(np.float32)
+            if (constant_value == 1).all():
+                # Avoid elimination of the layer from IR
+                constant_value = constant_value + 1
+            y = tf.constant(constant_value)
+
+            mul = tf.multiply(x, y, name="Operation")
+            mul_shape = mul.shape.as_list()
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        if len(mul_shape) >= 3:
+            # Permute mul_shape to (N,C,...) format
+            order = [0, len(mul_shape) - 1] + list(range(1, len(mul_shape) - 1))
+            mul_shape = [mul_shape[i] for i in order]
+
+        y_shape_to_compare = tf_y_shape.copy()
+        while len(y_shape_to_compare) < len(x_shape):
+            # Expand shape of constant with 1
+            y_shape_to_compare = [1] + y_shape_to_compare
+            constant_value = np.expand_dims(constant_value, axis=0)
+
+        if len(y_shape_to_compare) >= 3:
+            # Permute constant_value to (N,C,...) format for correct further reshape
+            order = [0, len(y_shape_to_compare) - 1] + list(range(1, len(y_shape_to_compare) - 1))
+            y_shape_to_compare = [y_shape_to_compare[i] for i in order]
+            constant_value = np.transpose(constant_value, order)
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    # TODO: implement tests for 2 Consts + Mul
+
+    test_data_1D = [
+        # Power
+        dict(x_shape=[1], y_shape=[1]),
+        # Eltwise
+        pytest.param(dict(x_shape=[3], y_shape=[3]), marks=pytest.mark.xfail(reason="*-19180"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_1D)
+    @pytest.mark.nightly
+    def test_mul_placeholder_const_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_mul_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_2D = [
+        # Power
+        dict(x_shape=[1, 1], y_shape=[1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3], y_shape=[1, 3]),
+        # Eltwise
+        pytest.param(dict(x_shape=[3, 1], y_shape=[3, 1]), marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[2, 3], y_shape=[2, 3])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_2D)
+    @pytest.mark.nightly
+    def test_mul_placeholder_const_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_mul_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_3D = [
+        # Power
+        dict(x_shape=[1, 1, 1], y_shape=[1, 1, 1]),
+        # ScaleShift
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[1, 3, 1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 3], y_shape=[1, 1, 3]),
+                     marks=[pytest.mark.xfail(reason="*-19053"), pytest.mark.xfail(reason="*-18830")]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 3, 224], y_shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_mul_placeholder_const_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_mul_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_4D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1], y_shape=[1, 1, 1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1], y_shape=[1, 3, 1, 1]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 1, 3], y_shape=[1, 1, 1, 3]), marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[1, 3, 222, 224], y_shape=[1, 3, 222, 224])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_mul_placeholder_const_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_mul_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1, 1], y_shape=[1, 1, 1, 1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[1, 3, 1, 1, 1]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 1, 1, 3], y_shape=[1, 1, 1, 1, 3]),
+                     marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[1, 3, 50, 100, 224], y_shape=[1, 3, 50, 100, 224])
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_mul_placeholder_const_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_mul_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    ###############################################################################################
+    #                                                                                             #
+    #                                       Broadcast cases                                       #
+    #                                                                                             #
+    ###############################################################################################
+
+    test_data_broadcast_1D = [  # Power
+        dict(x_shape=[3], y_shape=[1])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_1D)
+    @pytest.mark.nightly
+    def test_mul_placeholder_const_broadcast_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_mul_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_broadcast_2D = [
+        # Power
+        dict(x_shape=[1, 1], y_shape=[1]),
+        # Power
+        dict(x_shape=[1, 3], y_shape=[1]),
+        # ScaleShift
+        dict(x_shape=[1, 3], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[3, 1], y_shape=[3]),
+        # Eltwise
+        pytest.param(dict(x_shape=[3, 1], y_shape=[1, 3, 1, 1]), marks=pytest.mark.xfail(reason="*-19051"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_2D)
+    @pytest.mark.nightly
+    def test_mul_placeholder_const_broadcast_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_mul_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_broadcast_3D = [
+        # Power
+        dict(x_shape=[1, 1, 1], y_shape=[1]),
+        # Power
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # ScaleShift
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[3]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[3, 1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 1], y_shape=[3, 1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[3, 1, 224], y_shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[2, 3, 1], y_shape=[1, 3, 2]), marks=pytest.mark.xfail(reason="*-19053")),
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_broadcast_3D)
+    @pytest.mark.nightly
+    def test_mul_placeholder_const_broadcast_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_mul_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_broadcast_4D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1], y_shape=[1]),
+        # Power
+        dict(x_shape=[1, 3, 1, 1], y_shape=[1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1], y_shape=[3]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 100, 224], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 1, 1, 3], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 1], y_shape=[3, 1]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 2], y_shape=[3, 1, 2]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 2], y_shape=[1, 3, 2]),
+        # Eltwise
+        dict(x_shape=[1, 3, 100, 224], y_shape=[1, 1, 1, 224]),
+        # Eltwise
+        dict(x_shape=[2, 3, 1, 2], y_shape=[1, 3, 2, 1])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_4D)
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_mul_placeholder_const_broadcast_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_mul_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_broadcast_5D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1, 1], y_shape=[1]),
+        # Power
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 1, 1, 1, 3], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[3, 1]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 1, 2], y_shape=[1, 3, 2]),
+        # Eltwise
+        dict(x_shape=[1, 3, 5, 1, 2], y_shape=[5, 3, 2, 1]),
+        # Eltwise
+        dict(x_shape=[1, 3, 50, 100, 224], y_shape=[1, 1, 1, 1, 224]),
+        # Eltwise
+        dict(x_shape=[2, 3, 1, 2, 1], y_shape=[1, 3, 2, 1, 1])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_5D)
+    @pytest.mark.nightly
+    def test_mul_placeholder_const_broadcast_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_mul_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_NormalizeL2.py b/tests/layer_tests/tensorflow_tests/test_tf_NormalizeL2.py
new file mode 100644
index 00000000000000..1dc4b914172645
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_NormalizeL2.py
@@ -0,0 +1,207 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+from mo.front.common.partial_infer.utils import int64_array
+from unit_tests.utils.graph import build_graph
+from tensorflow_tests.permutation_utils import permute_nchw_to_nhwc, permute_nhwc_to_nchw
+
+
+class TestNormalizeL2(CommonTFLayerTest):
+    @staticmethod
+    def build_tf_graph(shape, axes):
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            # Permute NCHW -> NHWC for TF network creation
+            net_shape = permute_nchw_to_nhwc(shape)
+
+            data = tf.compat.v1.placeholder(tf.float32, shape=net_shape, name='data')
+
+            result = tf.math.l2_normalize(data,
+                                          axes,
+                                          name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+            return tf_net
+
+    @staticmethod
+    def create_normalize_l2_net_fusable(shape, axes, output_axes, ir_version):
+        tf_net = TestNormalizeL2.build_tf_graph(shape, axes)
+
+        nodes_attributes = {
+            'input': {'kind': 'op', 'type': 'Parameter'},
+            'input_data': {'shape': shape, 'kind': 'data'},
+            'axes_input_data': {'shape': int64_array([len(axes)]), 'kind': 'data', 'value': int64_array(output_axes)},
+            'axes': {'kind': 'op', 'type': 'Const'},
+            'axes_data': {'shape': int64_array([len(axes)]), 'kind': 'data'},
+            'normalize_l2': {'kind': 'op', 'type': 'NormalizeL2'},
+            'normalize_l2_data': {'shape': shape, 'kind': 'data'},
+            'result': {'kind': 'op', 'type': 'Result'},
+        }
+
+        ref_net = build_graph(nodes_attributes,
+                              [('input', 'input_data'),
+                               ('input_data', 'normalize_l2', {'out': 0, 'in': 0}),
+                               ('axes_input_data', 'axes'),
+                               ('axes', 'axes_data'),
+                               ('axes_data', 'normalize_l2', {'in': 1, 'out': 0}),
+                               ('normalize_l2', 'normalize_l2_data'),
+                               ('normalize_l2_data', 'result'),
+                               ])
+
+        return tf_net, ref_net
+
+    @staticmethod
+    def create_normalize_l2_net_non_fusable(shape, axes, output_axes, ir_version):
+        tf_net = TestNormalizeL2.build_tf_graph(shape, axes)
+
+        reduced_shape = permute_nchw_to_nhwc(shape).copy()
+        for axis in axes:
+            reduced_shape[axis] = 1
+        reduced_shape = permute_nhwc_to_nchw(reduced_shape)
+
+        eltwise_shapes = int64_array(np.ones(len(shape)))
+        nodes_attributes = {
+            'input': {'kind': 'op', 'type': 'Parameter'},
+            'input_data': {'shape': shape, 'kind': 'data'},
+
+            'power_const_input_data': {'shape': int64_array([1]), 'kind': 'data', 'value': np.array([2.0])},
+            'power_const': {'kind': 'op', 'type': 'Const'},
+            'power_const_data': {'shape': eltwise_shapes, 'kind': 'data'},
+            'power': {'kind': 'op', 'type': 'Power'},
+            'power_data': {'shape': shape, 'kind': 'data'},
+
+            'reduce': {'kind': 'op', 'type': 'ReduceSum', 'keep_dims': True},
+            'reduce_data': {'shape': reduced_shape, 'kind': 'data'},
+            'reduce_axes_input_data': {'shape': int64_array([len(axes)]), 'kind': 'data',
+                                       'value': int64_array(output_axes)},
+            'reduce_axes': {'kind': 'op', 'type': 'Const'},
+            'reduce_axes_data': {'shape': int64_array([len(axes)]), 'kind': 'data'},
+
+            'maximum_const_input_data': {'shape': int64_array([1]), 'kind': 'data', 'value': np.array([1e-12])},
+            'maximum_const': {'kind': 'op', 'type': 'Const'},
+            'maximum_const_data': {'shape': eltwise_shapes, 'kind': 'data'},
+            'maximum': {'kind': 'op', 'type': 'Maximum'},
+            'maximum_data': {'shape': reduced_shape, 'kind': 'data'},
+
+            'power2_const_input_data': {'shape': int64_array([1]), 'kind': 'data', 'value': np.array([-0.5])},
+            'power2_const': {'kind': 'op', 'type': 'Const'},
+            'power2_const_data': {'shape': eltwise_shapes, 'kind': 'data'},
+            'power2': {'kind': 'op', 'type': 'Power'},
+            'power2_data': {'shape': reduced_shape, 'kind': 'data'},
+
+            'multiply': {'kind': 'op', 'type': 'Multiply'},
+            'multiply_data': {'shape': shape, 'kind': 'data'},
+            'result': {'kind': 'op', 'type': 'Result'},
+        }
+
+        ref_net = build_graph(nodes_attributes,
+                              [('input', 'input_data'),
+
+                               ('input_data', 'power', {'out': 0, 'in': 0}),
+                               ('power_const_input_data', 'power_const'),
+                               ('power_const', 'power_const_data'),
+                               ('power_const_data', 'power', {'out': 0, 'in': 1}),
+                               ('power', 'power_data'),
+
+                               ('power_data', 'reduce', {'out': 0, 'in': 0}),
+                               ('reduce_axes_input_data', 'reduce_axes'),
+                               ('reduce_axes', 'reduce_axes_data'),
+                               ('reduce_axes_data', 'reduce', {'out': 0, 'in': 1}),
+                               ('reduce', 'reduce_data'),
+
+                               ('reduce_data', 'maximum', {'out': 0, 'in': 0}),
+                               ('maximum_const_input_data', 'maximum_const'),
+                               ('maximum_const', 'maximum_const_data'),
+                               ('maximum_const_data', 'maximum', {'out': 0, 'in': 1}),
+                               ('maximum', 'maximum_data'),
+
+                               ('maximum_data', 'power2', {'out': 0, 'in': 0}),
+                               ('power2_const_input_data', 'power2_const'),
+                               ('power2_const', 'power2_const_data'),
+                               ('power2_const_data', 'power2', {'out': 0, 'in': 1}),
+                               ('power2', 'power2_data'),
+
+                               ('input_data', 'multiply', {'out': 0, 'in': 0}),
+                               ('power2_data', 'multiply', {'out': 0, 'in': 1}),
+                               ('multiply', 'multiply_data'),
+                               ('multiply_data', 'result'),
+                               ])
+
+        return tf_net, ref_net
+
+    test_data_fusable_precommit = [
+        pytest.param(dict(shape=[2, 3, 5], axes=[1, -1], output_axes=[1, 2]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed")),
+        pytest.param(dict(shape=[2, 3, 5, 7], axes=[1, 2, 3], output_axes=[2, 3, 1]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_fusable_precommit)
+    @pytest.mark.precommit
+    def test_NormalizeL2_fusable_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_normalize_l2_net_fusable(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_non_fusable_precommit = [
+        pytest.param(dict(shape=[2, 3, 5], axes=[0, 1, 2], output_axes=[0, 1, 2]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed")),
+        pytest.param(dict(shape=[2, 3, 5, 7, 9], axes=[-1], output_axes=[1]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed")),
+        pytest.param(dict(shape=[2, 3, 5, 7, 9], axes=[1, 2, 3, 4], output_axes=[2, 3, 4, 1]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_non_fusable_precommit)
+    @pytest.mark.precommit
+    def test_NormalizeL2_non_fusable_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_normalize_l2_net_non_fusable(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    test_data_fusable = [
+        dict(shape=[5, 6], axes=[1], output_axes=[1]),
+        dict(shape=[2, 3, 5], axes=[1], output_axes=[1]),
+        dict(shape=[2, 3, 5], axes=[-2], output_axes=[1]),
+        dict(shape=[2, 3, 5], axes=[1, -1], output_axes=[1, 2]),
+        dict(shape=[2, 3, 5, 7], axes=[-1], output_axes=[1]),
+        dict(shape=[2, 3, 5, 7], axes=[1, 2, 3], output_axes=[2, 3, 1]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_fusable)
+    @pytest.mark.nightly
+    def test_NormalizeL2_fusable(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_normalize_l2_net_fusable(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_non_fusable = [
+        dict(shape=[5], axes=[0], output_axes=[0]),
+        dict(shape=[5, 6], axes=[0], output_axes=[0]),
+        dict(shape=[5, 6], axes=[0, 1], output_axes=[0, 1]),
+        dict(shape=[2, 3, 5], axes=[0], output_axes=[0]),
+        dict(shape=[2, 3, 5], axes=[2], output_axes=[2]),
+        dict(shape=[2, 3, 5], axes=[0, 1, 2], output_axes=[0, 1, 2]),
+        dict(shape=[2, 3, 5, 7], axes=[0], output_axes=[0]),
+        dict(shape=[2, 3, 5, 7], axes=[1], output_axes=[2]),
+        dict(shape=[2, 3, 5, 7], axes=[2], output_axes=[3]),
+        dict(shape=[2, 3, 5, 7], axes=[1, 2], output_axes=[2, 3]),
+        dict(shape=[2, 3, 5, 7], axes=[1, 3], output_axes=[2, 1]),
+        dict(shape=[2, 3, 5, 7], axes=[0, 1, 2], output_axes=[0, 2, 3]),
+        dict(shape=[2, 3, 5, 7, 9], axes=[-1], output_axes=[1]),
+        dict(shape=[2, 3, 5, 7, 9], axes=[1, 2, 3, 4], output_axes=[2, 3, 4, 1]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_non_fusable)
+    @pytest.mark.nightly
+    def test_NormalizeL2_non_fusable(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_normalize_l2_net_non_fusable(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_OneHot.py b/tests/layer_tests/tensorflow_tests/test_tf_OneHot.py
new file mode 100644
index 00000000000000..5e2d1d9131bf6b
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_OneHot.py
@@ -0,0 +1,159 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+from common.tf_layer_test_class import CommonTFLayerTest
+from tensorflow_tests.permutation_utils import permute_nchw_to_nhwc
+
+
+class TestOneHot(CommonTFLayerTest):
+    @staticmethod
+    def create_one_hot_net(shape, depth, on_value, off_value, axis, ir_version):
+        """
+            Tensorflow net
+
+            Input -> OneHot
+
+            IR net (can contain Permutes for input/output of OneHot, depending on shapes), all cases are:
+
+            Input (< 3D) -> OneHot
+
+            Input (3D) -> OneHot -> Permute (NHWC -> NCHW)
+
+            Input (> 3D)  -> Permute (NCHW -> NHWC) -> OneHot ->   Permute (NHWC -> NCHW)
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            # Permute NCHW -> NHWC for TF network creation
+            net_shape = permute_nchw_to_nhwc(shape)
+
+            indices = tf.compat.v1.placeholder(tf.int32, shape=net_shape, name='input_indices')
+
+            result = tf.one_hot(indices,
+                                 depth,
+                                 on_value,
+                                 off_value,
+                                 axis,
+                                 name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #
+
+        if on_value is None:
+            on_value = 1.0
+        if off_value is None:
+            off_value = 0.0
+
+        axis = len(shape) if axis is None else axis
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    test_data_1D = [
+        # check for default on/off value, axis params
+        dict(shape=[5], depth=7, on_value=None, off_value=None, axis=None),
+        dict(shape=[5], depth=7, on_value=2.0, off_value=-1.0, axis=0)]
+
+    @pytest.mark.parametrize("params", test_data_1D)
+    @pytest.mark.nightly
+    def test_OneHot_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_one_hot_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_2D = [
+        dict(shape=[5, 6], depth=7, on_value=None, off_value=None, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6], depth=7, on_value=5.0, off_value=None, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6], depth=7, on_value=None, off_value=-1.0, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6], depth=7, on_value=None, off_value=None, axis=1),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6], depth=7, on_value=2.0, off_value=-3.0, axis=0),
+        dict(shape=[5, 6], depth=7, on_value=2.0, off_value=-3.0, axis=1),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_2D)
+    @pytest.mark.nightly
+    def test_OneHot_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_one_hot_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_3D = [
+        dict(shape=[5, 6, 7], depth=8, on_value=None, off_value=None, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6, 7], depth=8, on_value=6.0, off_value=None, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6, 7], depth=8, on_value=None, off_value=4.0, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6, 7], depth=8, on_value=None, off_value=None, axis=1),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6, 7], depth=8, on_value=None, off_value=None, axis=0),
+        dict(shape=[5, 6, 7], depth=8, on_value=None, off_value=None, axis=1),
+        dict(shape=[5, 6, 7], depth=8, on_value=None, off_value=None, axis=2),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_OneHot_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_one_hot_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_4D = [
+        dict(shape=[5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=2),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6, 7, 8], depth=9, on_value=5.0, off_value=None, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6, 7, 8], depth=9, on_value=None, off_value=6.0, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=0),
+        dict(shape=[5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=1),
+        dict(shape=[5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=2),
+        dict(shape=[5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=3),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_OneHot_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_one_hot_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = [
+        dict(shape=[4, 5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[4, 5, 6, 7, 8], depth=9, on_value=2.0, off_value=None, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[4, 5, 6, 7, 8], depth=9, on_value=None, off_value=4.0, axis=None),
+        # check for default on/off value, axis params
+        dict(shape=[4, 5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=1),
+        # check for default on/off value, axis params
+        dict(shape=[4, 5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=0),
+        dict(shape=[4, 5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=1),
+        dict(shape=[4, 5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=2),
+        dict(shape=[4, 5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=3),
+        dict(shape=[4, 5, 6, 7, 8], depth=9, on_value=None, off_value=None, axis=4),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_OneHot_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_one_hot_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Pooling.py b/tests/layer_tests/tensorflow_tests/test_tf_Pooling.py
new file mode 100644
index 00000000000000..c0dd7f96aae193
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Pooling.py
@@ -0,0 +1,230 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestPooling(CommonTFLayerTest):
+    def create_pooling_net(self, kernel_size, strides, pads, in_shape, out_shape, method, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->Pooling        =>       Input->Pooling
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+
+            pads_begin, pads_end, padding = pads
+
+            # 4D tensors
+            if len(in_shape) == 4:
+                input_shape = [in_shape[0], in_shape[2], in_shape[3], in_shape[1]]
+                input = tf.compat.v1.placeholder(tf.float32, input_shape, 'Input')
+
+                stride = [1, strides[0], strides[1], 1]
+                kernel = [1, kernel_size[0], kernel_size[1], 1]
+
+                if method == 'max':
+                    tf.nn.max_pool2d(input=input, ksize=kernel, strides=stride, padding=padding, name='Operation')
+                elif method == 'avg':
+                    tf.nn.avg_pool2d(input=input, ksize=kernel, strides=stride, padding=padding, name='Operation')
+
+            # 5D tensors
+            elif len(in_shape) == 5:
+                input_shape = [in_shape[0], in_shape[2], in_shape[3], in_shape[4], in_shape[1]]
+                input = tf.compat.v1.placeholder(tf.float32, input_shape, 'Input')
+
+                stride = [1, strides[0], strides[1], strides[2], 1]
+                kernel = [1, kernel_size[0], kernel_size[1], kernel_size[2], 1]
+
+                if method == 'max':
+                    tf.nn.max_pool3d(input, kernel, stride, padding, name='Operation')  # , data_format='NCHW')
+                elif method == 'avg':
+                    tf.nn.avg_pool3d(input, kernel, stride, padding, name='Operation')  # , data_format='NCHW')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': in_shape, 'kind': 'data'},
+                'pooling': {'kernel': kernel_size, 'pads_begin': pads_begin, 'pads_end': pads_end,
+                            'strides': strides, 'kind': 'op', 'type': None},
+                'pooling_data': {'shape': out_shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            if method == 'avg':
+                nodes_attributes['pooling']['type'] = 'AvgPool'
+            elif method == 'max':
+                nodes_attributes['pooling']['type'] = 'MaxPool'
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'pooling'),
+                                   ('pooling', 'pooling_data'),
+                                   ('pooling_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_4D = []
+    for method in ['max', 'avg']:
+        test_data_4D.extend([dict(kernel_size=[1, 1], strides=[1, 1], pads=[[0, 0], [0, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 224, 224], method=method),
+                             dict(kernel_size=[2, 2], strides=[2, 2], pads=[[0, 0], [0, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 112, 112], method=method),
+                             dict(kernel_size=[2, 4], strides=[2, 4], pads=[[0, 0], [0, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 112, 56], method=method),
+                             dict(kernel_size=[4, 2], strides=[4, 2], pads=[[0, 0], [0, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 56, 112], method=method),
+                             dict(kernel_size=[2, 3], strides=[2, 3], pads=[[0, 0], [0, 1], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 112, 75], method=method),
+                             dict(kernel_size=[3, 2], strides=[3, 2], pads=[[0, 0], [1, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 75, 112], method=method),
+                             dict(kernel_size=[3, 3], strides=[2, 2], pads=[[0, 0], [1, 1], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 112, 112], method=method),
+                             dict(kernel_size=[3, 2], strides=[2, 2], pads=[[0, 0], [1, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 112, 112], method=method),
+                             dict(kernel_size=[2, 3], strides=[2, 3], pads=[[0, 0], [0, 1], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 112, 75], method=method),
+                             dict(kernel_size=[111, 111], strides=[111, 111], pads=[[54, 54], [55, 55], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 3, 3], method=method),
+                             dict(kernel_size=[111, 113], strides=[111, 113], pads=[[54, 1], [55, 1], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 3, 2], method=method),
+                             dict(kernel_size=[113, 113], strides=[113, 113], pads=[[1, 1], [1, 1], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 2, 2], method=method),
+                             dict(kernel_size=[113, 113], strides=[111, 111], pads=[[55, 55], [56, 56], 'SAME'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 3, 3], method=method)])
+
+        test_data_4D.extend([dict(kernel_size=[1, 1], strides=[1, 1], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 224, 224], method=method),
+                             dict(kernel_size=[2, 2], strides=[2, 2], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 112, 112], method=method),
+                             dict(kernel_size=[2, 4], strides=[2, 4], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 112, 56], method=method),
+                             dict(kernel_size=[4, 2], strides=[4, 2], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 56, 112], method=method),
+                             dict(kernel_size=[2, 3], strides=[2, 3], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 112, 74], method=method),
+                             dict(kernel_size=[3, 2], strides=[3, 2], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 74, 112], method=method),
+                             dict(kernel_size=[3, 3], strides=[2, 2], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 111, 111], method=method),
+                             dict(kernel_size=[3, 2], strides=[2, 2], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 111, 112], method=method),
+                             dict(kernel_size=[2, 3], strides=[2, 3], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 112, 74], method=method),
+                             dict(kernel_size=[111, 111], strides=[111, 111], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 2, 2], method=method),
+                             dict(kernel_size=[111, 113], strides=[111, 113], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 2, 1], method=method),
+                             dict(kernel_size=[113, 113], strides=[113, 113], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 1, 1], method=method),
+                             dict(kernel_size=[113, 113], strides=[111, 111], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 2, 2], method=method),
+                             dict(kernel_size=[224, 224], strides=[1, 1], pads=[[0, 0], [0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224], out_shape=[1, 3, 1, 1], method=method)])
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_pool_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_pooling_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = []
+    for method in ['max', 'avg']:
+        test_data_5D.extend([dict(kernel_size=[1, 1, 1], strides=[1, 1, 1], pads=[[0, 0, 0], [0, 0, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 224, 224, 224], method=method),
+                             dict(kernel_size=[2, 2, 2], strides=[2, 2, 2], pads=[[0, 0, 0], [0, 0, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 112, 112, 112], method=method),
+                             dict(kernel_size=[2, 2, 4], strides=[2, 2, 4], pads=[[0, 0, 0], [0, 0, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 112, 112, 56], method=method),
+                             dict(kernel_size=[4, 2, 2], strides=[4, 2, 2], pads=[[0, 0, 0], [0, 0, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 56, 112, 112], method=method),
+                             dict(kernel_size=[2, 2, 3], strides=[2, 2, 3], pads=[[0, 0, 0], [0, 0, 1], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 112, 112, 75], method=method),
+                             dict(kernel_size=[3, 2, 2], strides=[3, 2, 2], pads=[[0, 0, 0], [1, 0, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 75, 112, 112], method=method),
+                             dict(kernel_size=[3, 3, 3], strides=[2, 2, 2], pads=[[0, 0, 0], [1, 1, 1], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 112, 112, 112], method=method),
+                             dict(kernel_size=[3, 2, 2], strides=[2, 2, 2], pads=[[0, 0, 0], [1, 0, 0], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 112, 112, 112], method=method),
+                             dict(kernel_size=[2, 2, 3], strides=[2, 2, 3], pads=[[0, 0, 0], [0, 0, 1], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 112, 112, 75], method=method),
+                             dict(kernel_size=[111, 111, 111], strides=[111, 111, 111],
+                                  pads=[[54, 54, 54], [55, 55, 55], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 3, 3, 3], method=method),
+                             dict(kernel_size=[111, 111, 113], strides=[111, 111, 113],
+                                  pads=[[54, 54, 1], [55, 55, 1], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 3, 3, 2], method=method),
+                             dict(kernel_size=[113, 113, 113], strides=[113, 113, 113],
+                                  pads=[[1, 1, 1], [1, 1, 1], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 2, 2, 2], method=method),
+                             dict(kernel_size=[113, 113, 113], strides=[111, 111, 111],
+                                  pads=[[55, 55, 55], [56, 56, 56], 'SAME'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 3, 3, 3], method=method)])
+
+        test_data_5D.extend([dict(kernel_size=[1, 1, 1], strides=[1, 1, 1], pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 224, 224, 224], method=method),
+                             dict(kernel_size=[2, 2, 2], strides=[2, 2, 2], pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 112, 112, 112], method=method),
+                             dict(kernel_size=[2, 2, 4], strides=[2, 2, 4], pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 112, 112, 56], method=method),
+                             dict(kernel_size=[4, 2, 2], strides=[4, 2, 2], pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 56, 112, 112], method=method),
+                             dict(kernel_size=[2, 2, 3], strides=[2, 2, 3], pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 112, 112, 74], method=method),
+                             dict(kernel_size=[3, 2, 2], strides=[3, 2, 2], pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 74, 112, 112], method=method),
+                             dict(kernel_size=[3, 3, 3], strides=[2, 2, 2], pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 111, 111, 111], method=method),
+                             dict(kernel_size=[3, 2, 2], strides=[2, 2, 2], pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 111, 112, 112], method=method),
+                             dict(kernel_size=[2, 2, 3], strides=[2, 2, 3], pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 112, 112, 74], method=method),
+                             dict(kernel_size=[111, 111, 111], strides=[111, 111, 111],
+                                  pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 2, 2, 2], method=method),
+                             dict(kernel_size=[111, 111, 113], strides=[111, 111, 113],
+                                  pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 2, 2, 1], method=method),
+                             dict(kernel_size=[113, 113, 113], strides=[113, 113, 113],
+                                  pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 1, 1, 1], method=method),
+                             dict(kernel_size=[113, 113, 113], strides=[111, 111, 111],
+                                  pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 2, 2, 2], method=method),
+                             dict(kernel_size=[224, 224, 224], strides=[1, 1, 1],
+                                  pads=[[0, 0, 0], [0, 0, 0], 'VALID'],
+                                  in_shape=[1, 3, 224, 224, 224], out_shape=[1, 3, 1, 1, 1], method=method)])
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_pool_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        if ie_device == 'GPU':
+            pytest.skip("5D tensors is not supported on GPU")
+        self._test(*self.create_pooling_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_ReLU.py b/tests/layer_tests/tensorflow_tests/test_tf_ReLU.py
new file mode 100644
index 00000000000000..c178e6f8b55aef
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_ReLU.py
@@ -0,0 +1,85 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestReLU(CommonTFLayerTest):
+    def create_relu_net(self, shape, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->ReLU       =>       Input->ReLU
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) >= 3:
+                shapes.append(shapes.pop(1))
+            input = tf.compat.v1.placeholder(tf.float32, shapes, 'Input')
+
+            tf.nn.relu(input, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'ReLU': {'kind': 'op', 'type': 'ReLU'},
+                'ReLU_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'ReLU'),
+                                   ('ReLU', 'ReLU_data'),
+                                   ('ReLU_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_precommit = [dict(shape=[1, 3, 50, 100, 224])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_relu_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_relu_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[1]),
+                 dict(shape=[1, 224]),
+                 pytest.param(dict(shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053")),
+                 dict(shape=[1, 3, 100, 224]),
+                 dict(shape=[1, 3, 50, 100, 224])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_relu(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_relu_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_ReLU6.py b/tests/layer_tests/tensorflow_tests/test_tf_ReLU6.py
new file mode 100644
index 00000000000000..fd846a82358428
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_ReLU6.py
@@ -0,0 +1,85 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestReLU6(CommonTFLayerTest):
+    def create_relu6_net(self, shape, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->ReLU6       =>       Input->Clamp
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) >= 3:
+                shapes.append(shapes.pop(1))
+            input = tf.compat.v1.placeholder(tf.float32, shapes, 'Input')
+
+            tf.nn.relu6(input, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'ReLU6': {'kind': 'op', 'type': 'Clamp', "max": 6, "min": 0},
+                'ReLU6_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'ReLU6'),
+                                   ('ReLU6', 'ReLU6_data'),
+                                   ('ReLU6_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_precommit = [dict(shape=[1, 3, 50, 100, 224])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_relu6_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_relu6_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[1]),
+                 dict(shape=[1, 224]),
+                 pytest.param(dict(shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053")),
+                 dict(shape=[1, 3, 100, 224]),
+                 dict(shape=[1, 3, 50, 100, 224])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_relu6(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_relu6_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Resample_pattern_new.py b/tests/layer_tests/tensorflow_tests/test_tf_Resample_pattern_new.py
new file mode 100644
index 00000000000000..e3afdda6972d02
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Resample_pattern_new.py
@@ -0,0 +1,81 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestResamplePattern(CommonTFLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randint(1, 256, inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_resample_net(self, shape, factor):
+        """
+            The sub-graph in TF that could be expressed as a single Resample operation.
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            tf_shape = shape.copy()
+            tf_shape = np.array(tf_shape)[[0, 2, 3, 1]]
+
+            input = tf.compat.v1.placeholder(tf.float32, tf_shape, 'Input')
+
+            transpose_1 = tf.transpose(a=input, perm=[1, 2, 3, 0])
+            expand_dims = tf.expand_dims(transpose_1, 0)
+            tile = tf.tile(expand_dims, [factor * factor, 1, 1, 1, 1])
+            bts = tf.batch_to_space(tile, [factor, factor], [[0, 0], [0, 0]])
+            strided_slice = bts[0, ...]
+            tf.transpose(a=strided_slice, perm=[3, 0, 1, 2])
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        new_shape = shape.copy()
+        new_shape[2] *= factor
+        new_shape[3] *= factor
+        nodes_attributes = {
+            'input': {'kind': 'op', 'type': 'Input'},
+            'input_data': {'shape': shape, 'kind': 'data'},
+            'resample': {'kind': 'op', 'type': 'caffe.ResampleParameter.NEAREST', "factor": factor,
+                         "height": 0, "width": 0, "antialias": 0},
+            'resample_data': {'shape': new_shape, 'kind': 'data'},
+        }
+
+        ref_net = build_graph(nodes_attributes,
+                              [('input', 'input_data'),
+                               ('input_data', 'resample'),
+                               ('resample', 'resample_data')
+                               ])
+
+        return tf_net, ref_net
+
+    test_data = [dict(shape=[1, 1, 100, 200], factor=2),
+                 dict(shape=[1, 1, 200, 300], factor=3)]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    @pytest.mark.xfail(reason="*-22273")
+    def test_resample(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_resample_net(params['shape'], params['factor']),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Rsqrt.py b/tests/layer_tests/tensorflow_tests/test_tf_Rsqrt.py
new file mode 100644
index 00000000000000..953d73f6809b3a
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Rsqrt.py
@@ -0,0 +1,73 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestRsqrt(CommonTFLayerTest):
+    def _prepare_input(self, inputs_dict):
+        for input in inputs_dict.keys():
+            inputs_dict[input] = np.random.randint(1, 256, inputs_dict[input]).astype(np.float32)
+        return inputs_dict
+
+    def create_rsqrt_net(self, shape, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->Rsqrt       =>       Input->Power
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) >= 3:
+                shapes.append(shapes.pop(1))
+            input = tf.compat.v1.placeholder(tf.float32, shapes, 'Input')
+
+            tf.math.rsqrt(input, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    test_data_precommit = [dict(shape=[1, 3, 50, 100, 224])]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_rsqrt_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_rsqrt_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[1]),
+                 dict(shape=[1, 224]),
+                 pytest.param(dict(shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053")),
+                 dict(shape=[1, 3, 100, 224]),
+                 dict(shape=[1, 3, 50, 100, 224])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_rsqrt(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_rsqrt_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Select.py b/tests/layer_tests/tensorflow_tests/test_tf_Select.py
new file mode 100644
index 00000000000000..ee24964589b494
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Select.py
@@ -0,0 +1,107 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+from tensorflow_tests.permutation_utils import permute_nchw_to_nhwc
+
+
+class TestSelect(CommonTFLayerTest):
+    def create_select_net(self, shape_condition, shape_input, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Condition --|               Condition --|
+                        v                           v
+            Input_1-> Select            Input_1-> Select
+                        ^                           ^
+            Input_2-----|               Input_2-----|
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            # Permute shapes NCHW -> NHWC for TF network creation
+            shape_condition_net = permute_nchw_to_nhwc(shape_condition)
+            shape_input_net = permute_nchw_to_nhwc(shape_input)
+
+            condition = tf.compat.v1.placeholder(tf.bool, shape_condition_net, 'Input_condition')
+            input_1 = tf.compat.v1.placeholder(tf.float32, shape_input_net, 'Input_1')
+            input_2 = tf.compat.v1.placeholder(tf.float32, shape_input_net, 'Input_2')
+
+            tf.compat.v1.where(condition, input_1, input_2, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    test_data_1D = [dict(shape_condition=[2], shape_input=[2])]
+
+    @pytest.mark.parametrize("params", test_data_1D)
+    @pytest.mark.nightly
+    def test_select_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_select_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_2D = [
+        dict(shape_condition=[2], shape_input=[2, 3]),
+        dict(shape_condition=[3, 5], shape_input=[3, 5]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_2D)
+    @pytest.mark.nightly
+    def test_select_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_select_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_3D = [
+        dict(shape_condition=[3], shape_input=[3, 4, 5]),
+        dict(shape_condition=[3, 4, 5], shape_input=[3, 4, 5]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_select_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_select_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_4D = [
+        dict(shape_condition=[3], shape_input=[3, 4, 5, 6]),
+        dict(shape_condition=[3, 4, 5, 6], shape_input=[3, 4, 5, 6]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    @pytest.mark.precommit
+    def test_select_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_select_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = [
+        dict(shape_condition=[3], shape_input=[3, 4, 5, 6, 7]),
+        dict(shape_condition=[3, 4, 5, 6, 7], shape_input=[3, 4, 5, 6, 7]),
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_select_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_select_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Softplus.py b/tests/layer_tests/tensorflow_tests/test_tf_Softplus.py
new file mode 100644
index 00000000000000..a295e4660e9174
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Softplus.py
@@ -0,0 +1,90 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestSoftplus(CommonTFLayerTest):
+    def create_softplus_net(self, shape, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->Softplus       =>       Input->Softplus
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.reset_default_graph()
+
+        # Create the graph and model
+        with tf.Session() as sess:
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) > 3:
+                shapes.append(shapes.pop(1))
+            input = tf.placeholder(tf.float32, shapes, 'Input')
+
+            tf.math.softplus(input, name='Operation')
+
+            tf.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'Softplus': {'kind': 'op', 'type': 'SoftPlus'},
+                'Softplus_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'Softplus'),
+                                   ('Softplus', 'Softplus_data'),
+                                   ('Softplus_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_precommit = [
+        pytest.param(dict(shape=[1, 3, 100, 224]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed")),
+        pytest.param(dict(shape=[1, 3, 50, 100, 224]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_softplus_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_softplus_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[1]),
+                 dict(shape=[1, 224]),
+                 dict(shape=[1, 3, 224]),
+                 dict(shape=[1, 3, 100, 224]),
+                 dict(shape=[1, 3, 50, 100, 224])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_softplus(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_softplus_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_SpaceToBatch.py b/tests/layer_tests/tensorflow_tests/test_tf_SpaceToBatch.py
new file mode 100644
index 00000000000000..f86ab7dc2acf21
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_SpaceToBatch.py
@@ -0,0 +1,80 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestSpaceToBatch(CommonTFLayerTest):
+    def create_space_to_batch_net(self, in_shape, pads_value, block_shape_value, out_shape, ir_version):
+        """
+            Tensorflow net               IR net
+
+            Input->SpaceToBatch        =>      Input->SpaceToBatch
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            x = tf.compat.v1.placeholder(tf.float32, in_shape, 'Input')
+            pads = tf.constant(pads_value)
+            block_shape = tf.constant(block_shape_value)
+            tf.space_to_batch(x, block_shape, pads, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    test_data_4D = [
+        dict(in_shape=[4, 1, 1, 3], block_shape_value=[1], pads_value=[[0, 0]],
+             out_shape=[4, 1, 1, 3]),
+        dict(in_shape=[1, 2, 2, 1], block_shape_value=[2, 2], pads_value=[[0, 0], [0, 0]],
+             out_shape=[4, 1, 1, 3]),
+        dict(in_shape=[1, 2, 2, 3], block_shape_value=[2, 2], pads_value=[[0, 0], [0, 0]],
+             out_shape=[4, 1, 1, 3]),
+        dict(in_shape=[1, 2, 9, 1], block_shape_value=[4, 3], pads_value=[[1, 1], [2, 4]],
+             out_shape=[12, 1, 1, 3]),
+        # todo: enable these tests after supporting the general case on CPU
+        # dict(in_shape=[1, 2, 1, 4], block_shape_value=[3, 2, 2], pads_value=[[1, 0], [0, 1], [1, 1]],
+        #      out_shape=[12, 1, 1, 3]),
+        # dict(in_shape=[2, 3, 6, 5], block_shape_value=[2, 3, 3], pads_value=[[1, 0], [0, 0], [2, 2]],
+        #      out_shape=[36, 2, 2, 3])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_space_to_batch_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_space_to_batch_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = [
+        dict(in_shape=[3, 3, 4, 5, 2], block_shape_value=[3, 4, 2], pads_value=[[1, 2], [0, 0], [3, 0]],
+             out_shape=[72, 2, 1, 4, 2]),
+        # todo: enable these tests after supporting the general case on CPU
+        # dict(in_shape=[3, 3, 4, 5, 2], block_shape_value=[3, 4, 2, 2],
+        #      pads_value=[[1, 2], [0, 0], [3, 0], [0, 0]], out_shape=[144, 2, 1, 4, 1]),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_space_to_batch_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_space_to_batch_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Squeeze.py b/tests/layer_tests/tensorflow_tests/test_tf_Squeeze.py
new file mode 100644
index 00000000000000..0ef13751b3463e
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Squeeze.py
@@ -0,0 +1,124 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestSqueeze(CommonTFLayerTest):
+    def create_squeeze_net(self, shape, axis, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->Squeeze       =>       Input->[Permute]->Reshape
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            x_shape = shape.copy()
+            # reshaping
+            if len(x_shape) >= 3:
+                x_shape.append(x_shape.pop(1))
+
+            x = tf.compat.v1.placeholder(tf.float32, x_shape, 'Input')
+            squeeze = tf.squeeze(x, axis=axis, name="Operation")
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        unsigned_axis = [ax if ax > -1 else len(x_shape) + ax for ax in axis]
+        if not unsigned_axis:
+            unsigned_axis = [i for i, dim in enumerate(shape) if dim == 1]
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    test_data_1D = [
+        pytest.param(dict(shape=[1], axis=[]), marks=pytest.mark.xfail(reason="*-18807")),
+        pytest.param(dict(shape=[1], axis=[0]), marks=pytest.mark.xfail(reason="*-18859")),
+        pytest.param(dict(shape=[1], axis=[-1]), marks=pytest.mark.xfail(reason="*-18859"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_1D)
+    @pytest.mark.nightly
+    def test_squeeze_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_2D = [
+        pytest.param(dict(shape=[1, 1], axis=[]), marks=pytest.mark.xfail(reason="*-18807")),
+        dict(shape=[1, 1], axis=[0]),
+        dict(shape=[1, 1], axis=[-1])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_2D)
+    @pytest.mark.nightly
+    def test_squeeze_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_3D = [
+        pytest.param(dict(shape=[1, 1, 3], axis=[]),
+                     marks=[pytest.mark.xfail(reason="*-18807"), pytest.mark.xfail(reason="*-19053")]),
+        pytest.param(dict(shape=[1, 1, 3], axis=[0]), marks=pytest.mark.xfail(reason="*-19053")),
+        pytest.param(dict(shape=[1, 1, 3], axis=[-1]), marks=pytest.mark.xfail(reason="*-19053"))
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_squeeze_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_4D = [
+        pytest.param(dict(shape=[1, 1, 50, 100], axis=[]), marks=pytest.mark.xfail(reason="*-18807")),
+        dict(shape=[1, 1, 50, 100], axis=[0]),
+        dict(shape=[1, 1, 50, 100], axis=[-1]),
+        dict(shape=[1, 100, 50, 1], axis=[0, 2])
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_squeeze_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = [
+        pytest.param(dict(shape=[1, 1, 50, 100, 224], axis=[]), marks=pytest.mark.xfail(reason="*-18807")),
+        pytest.param(dict(shape=[1, 1, 50, 100, 224], axis=[0]), marks=pytest.mark.xfail(reason="*-18879")),
+        pytest.param(dict(shape=[1, 1, 50, 100, 224], axis=[-1]), marks=pytest.mark.xfail(reason="*-18879")),
+        dict(shape=[1, 224, 1, 100, 1], axis=[0, 3]),
+        dict(shape=[1, 224, 1, 100, 1], axis=[0, 1, 3]),
+        dict(shape=[1, 224, 1, 1, 100], axis=[0, 1, 2]),
+        dict(shape=[1, 224, 1, 1, 1], axis=[0, 1, 2, 3])
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.special_xfail(args={'ie_device': 'GPU', 'precision': 'FP16', 'params': {'axis': [0, 3]}},
+                               reason="*-19394")
+    @pytest.mark.special_xfail(args={'ie_device': 'GPU', 'precision': 'FP16', 'params': {'axis': [0, 1, 3]}},
+                               reason="*-19394")
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_squeeze_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_squeeze_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Sub.py b/tests/layer_tests/tensorflow_tests/test_tf_Sub.py
new file mode 100644
index 00000000000000..00f63494255d5b
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Sub.py
@@ -0,0 +1,294 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestSub(CommonTFLayerTest):
+    def create_sub_placeholder_const_net(self, x_shape, y_shape, ir_version):
+        """
+            Tensorflow net                  IR net
+
+            Placeholder->Sub       =>       Placeholder->Eltwise or Power or ScaleShift
+                         /                               /
+            Const-------/                   Const-------/
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            tf_x_shape = x_shape.copy()
+            tf_y_shape = y_shape.copy()
+            # reshaping
+            if len(tf_x_shape) >= 3:
+                tf_x_shape.append(tf_x_shape.pop(1))
+            if len(tf_y_shape) >= 3:
+                tf_y_shape.append(tf_y_shape.pop(1))
+
+            x = tf.compat.v1.placeholder(tf.float32, tf_x_shape, 'Input')
+            constant_value = np.random.randint(-256, 256, tf_y_shape).astype(np.float32)
+            if (constant_value == 0).all():
+                # Avoid elimination of the layer from IR
+                constant_value = constant_value + 1
+            y = tf.constant(constant_value)
+
+            sub = tf.subtract(x, y, name="Operation")
+            sub_shape = sub.shape.as_list()
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        if len(sub_shape) >= 3:
+            # Permute sub_shape to (N,C,...) format
+            order = [0, len(sub_shape) - 1] + list(range(1, len(sub_shape) - 1))
+            sub_shape = [sub_shape[i] for i in order]
+
+        y_shape_to_compare = tf_y_shape.copy()
+        while len(y_shape_to_compare) < len(x_shape):
+            # Expand shape of constant with 1
+            y_shape_to_compare = [1] + y_shape_to_compare
+            constant_value = np.expand_dims(constant_value, axis=0)
+
+        if len(y_shape_to_compare) >= 3:
+            # Permute constant_value to (N,C,...) format for correct further reshape
+            order = [0, len(y_shape_to_compare) - 1] + list(range(1, len(y_shape_to_compare) - 1))
+            y_shape_to_compare = [y_shape_to_compare[i] for i in order]
+            constant_value = np.transpose(constant_value, order)
+
+        ref_net = None
+
+        return tf_net, ref_net
+
+    # TODO: implement tests for 2 Consts + Sub
+
+    test_data_1D = [
+        # Power
+        dict(x_shape=[1], y_shape=[1]),
+        # Eltwise
+        pytest.param(dict(x_shape=[3], y_shape=[3]), marks=pytest.mark.xfail(reason="*-19180"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_1D)
+    @pytest.mark.nightly
+    def test_sub_placeholder_const_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_sub_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    test_data_2D = [
+        # Power
+        dict(x_shape=[1, 1], y_shape=[1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3], y_shape=[1, 3]),
+        # Eltwise
+        pytest.param(dict(x_shape=[3, 1], y_shape=[3, 1]), marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[2, 3], y_shape=[2, 3])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_2D)
+    @pytest.mark.nightly
+    def test_sub_placeholder_const_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_sub_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    test_data_3D = [
+        # Power
+        dict(x_shape=[1, 1, 1], y_shape=[1, 1, 1]),
+        # ScaleShift
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[1, 3, 1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 3], y_shape=[1, 1, 3]),
+                     marks=[pytest.mark.xfail(reason="*-19053"), pytest.mark.xfail(reason="*-18830")]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 3, 224], y_shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053"))
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_sub_placeholder_const_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_sub_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    test_data_4D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1], y_shape=[1, 1, 1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1], y_shape=[1, 3, 1, 1]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 1, 3], y_shape=[1, 1, 1, 3]), marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[1, 3, 222, 224], y_shape=[1, 3, 222, 224])
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_sub_placeholder_const_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_sub_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    test_data_5D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1, 1], y_shape=[1, 1, 1, 1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[1, 3, 1, 1, 1]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 1, 1, 3], y_shape=[1, 1, 1, 1, 3]),
+                     marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[1, 3, 50, 100, 224], y_shape=[1, 3, 50, 100, 224])
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_sub_placeholder_const_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_sub_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    ###############################################################################################
+    #                                                                                             #
+    #                                       Broadcast cases                                       #
+    #                                                                                             #
+    ###############################################################################################
+
+    test_data_broadcast_1D = [  # Power
+        dict(x_shape=[3], y_shape=[1])
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_1D)
+    @pytest.mark.nightly
+    def test_sub_placeholder_const_broadcast_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_sub_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    test_data_broadcast_2D = [
+        # Power
+        dict(x_shape=[1, 1], y_shape=[1]),
+        # Power
+        dict(x_shape=[1, 3], y_shape=[1]),
+        # ScaleShift
+        dict(x_shape=[1, 3], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[3, 1], y_shape=[3]),
+        # Eltwise
+        pytest.param(dict(x_shape=[3, 1], y_shape=[1, 3, 1, 1]), marks=pytest.mark.xfail(reason="*-19051"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_broadcast_2D)
+    @pytest.mark.nightly
+    def test_sub_placeholder_const_broadcast_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_sub_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    test_data_broadcast_3D = [
+        # Power
+        dict(x_shape=[1, 1, 1], y_shape=[1]),
+        # Power
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # ScaleShift
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[3]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 3, 1], y_shape=[3, 1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 1, 1], y_shape=[3, 1]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[3, 1, 224], y_shape=[1, 3, 224]), marks=pytest.mark.xfail(reason="*-19053")),
+        # Eltwise
+        pytest.param(dict(x_shape=[2, 3, 1], y_shape=[1, 3, 2]), marks=pytest.mark.xfail(reason="*-19053")),
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_broadcast_3D)
+    @pytest.mark.nightly
+    def test_sub_placeholder_const_broadcast_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_sub_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    test_data_broadcast_4D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1], y_shape=[1]),
+        # Power
+        dict(x_shape=[1, 3, 1, 1], y_shape=[1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 3, 100, 224], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 1, 1, 3], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 1], y_shape=[3, 1]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 2], y_shape=[3, 1, 2]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 2], y_shape=[1, 3, 2]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 3, 100, 224], y_shape=[1, 1, 1, 224]),
+                     marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[2, 3, 1, 2], y_shape=[1, 3, 2, 1])
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_broadcast_4D)
+    @pytest.mark.nightly
+    def test_sub_placeholder_const_broadcast_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_sub_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
+
+    test_data_broadcast_5D = [
+        # Power
+        dict(x_shape=[1, 1, 1, 1, 1], y_shape=[1, 1, 1, 1, 1]),
+        # Power
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[1, 1]),
+        # ScaleShift
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 1, 1, 1, 3], y_shape=[3]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 1, 1], y_shape=[3, 1]),
+        # Eltwise
+        dict(x_shape=[1, 3, 1, 1, 2], y_shape=[1, 3, 2]),
+        # Eltwise
+        dict(x_shape=[1, 3, 5, 1, 2], y_shape=[5, 3, 2, 1]),
+        # Eltwise
+        pytest.param(dict(x_shape=[1, 3, 50, 100, 224], y_shape=[1, 1, 1, 1, 224]),
+                     marks=pytest.mark.xfail(reason="*-19180")),
+        # Eltwise
+        dict(x_shape=[2, 3, 1, 2, 1], y_shape=[1, 3, 2, 1, 1])
+    ]
+
+    # TODO mark as precommit (after successfully passing in nightly)
+    @pytest.mark.parametrize("params", test_data_broadcast_5D)
+    @pytest.mark.nightly
+    def test_sub_placeholder_const_broadcast_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_sub_placeholder_const_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version,
+                   temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Swish.py b/tests/layer_tests/tensorflow_tests/test_tf_Swish.py
new file mode 100644
index 00000000000000..158a43353ac79c
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Swish.py
@@ -0,0 +1,88 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from unit_tests.utils.graph import build_graph
+
+
+class TestSwish(CommonTFLayerTest):
+    def create_swish_net(self, shape, ir_version):
+        """
+            Tensorflow net                 IR net
+
+            Input->Swish       =>       Input->Swish
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.reset_default_graph()
+
+        # Create the graph and model
+        with tf.Session() as sess:
+            shapes = shape.copy()
+            # reshaping
+            if len(shapes) > 3:
+                shapes.append(shapes.pop(1))
+            input = tf.placeholder(tf.float32, shapes, 'Input')
+
+            tf.nn.swish(input)
+
+            tf.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #   Please, specify 'type': 'Input' for input node
+        #   Moreover, do not forget to validate ALL layer attributes!!!
+        #
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'Swish': {'kind': 'op', 'type': 'Swish'},
+                'Swish_data': {'shape': shape, 'kind': 'data'},
+                'result': {'kind': 'op', 'type': 'Result'}
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'Swish'),
+                                   ('Swish', 'Swish_data'),
+                                   ('Swish_data', 'result')
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_precommit = [
+        pytest.param(dict(shape=[1, 3, 50, 100, 224]),
+                     marks=pytest.mark.skip(reason="Skipped until fixed"))
+    ]
+
+    @pytest.mark.parametrize("params", test_data_precommit)
+    @pytest.mark.precommit
+    def test_swish_precommit(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_swish_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data = [dict(shape=[1]),
+                 dict(shape=[1, 224]),
+                 dict(shape=[1, 3, 224]),
+                 dict(shape=[1, 3, 100, 224]),
+                 dict(shape=[1, 3, 50, 100, 224])]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_swish(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_swish_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_TopK.py b/tests/layer_tests/tensorflow_tests/test_tf_TopK.py
new file mode 100644
index 00000000000000..c2d2fe54c07044
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_TopK.py
@@ -0,0 +1,143 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from common.layer_test_class import check_ir_version
+from common.tf_layer_test_class import CommonTFLayerTest
+from mo.ops.op import PermuteAttrs
+from unit_tests.utils.graph import build_graph
+from tensorflow_tests.permutation_utils import permute_nchw_to_nhwc, permute_axis
+
+
+class Test_TopK(CommonTFLayerTest):
+    @staticmethod
+    def create_topK_net(shape, k, ir_version):
+        """
+            Tensorflow net:
+
+                          |-> Values
+            Input -> TopK |
+                          |-> Indices
+
+
+            IR net:
+
+                          |-> Values
+            Input -> TopK |
+                          |-> Indices
+
+        """
+
+        #
+        #   Create Tensorflow model
+        #
+
+        import tensorflow as tf
+
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            shape_net = permute_nchw_to_nhwc(shape)
+
+            input_tensor = tf.compat.v1.placeholder(tf.int32, shape=shape_net, name='Input')
+            values, indices = tf.nn.top_k(input_tensor, k=k, sorted=True, name='Operation')
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        #
+        #   Create reference IR net
+        #
+        topk_output_shape = shape.copy()
+        inverse_nhwc_nchw = PermuteAttrs.get_nhwc_to_nchw_permutation(len(topk_output_shape)).inv
+        topk_axis = permute_axis(len(topk_output_shape) - 1, inverse_nhwc_nchw)  # we need to permute axis attribute
+        topk_output_shape[topk_axis] = k
+
+        ref_net = None
+
+        if check_ir_version(10, None, ir_version):
+            nodes_attributes = {
+                'input': {'kind': 'op', 'type': 'Parameter'},
+                'input_data': {'shape': shape, 'kind': 'data'},
+                'Const_k_input_data': {'shape': [], 'kind': 'data'},
+                'Const_k': {'kind': 'op', 'type': 'Const'},
+                'Const_k_data': {'shape': [], 'kind': 'data'},
+                'TopK': {'kind': 'op', 'type': 'TopK', 'axis': topk_axis, 'mode': 'max', 'sort': 'value'},
+                'TopK_data_1': {'shape': topk_output_shape, 'kind': 'data'},
+                'TopK_data_2': {'shape': topk_output_shape, 'kind': 'data'},
+                'result_1': {'kind': 'op', 'type': 'Result'},
+                'result_2': {'kind': 'op', 'type': 'Result'},
+            }
+
+            ref_net = build_graph(nodes_attributes,
+                                  [('input', 'input_data'),
+                                   ('input_data', 'TopK', {'in': 0}),
+
+                                   ('Const_k_input_data', 'Const_k'),
+                                   ('Const_k', 'Const_k_data'),
+                                   ('Const_k_data', 'TopK', {'in': 1}),
+
+                                   ('TopK', 'TopK_data_1', {'out': 0}),
+                                   ('TopK', 'TopK_data_2', {'out': 1}),
+                                   ('TopK_data_1', 'result_1'),
+                                   ('TopK_data_2', 'result_2'),
+                                   ])
+
+        return tf_net, ref_net
+
+    test_data_1D = [
+        dict(shape=[15], k=10),
+        dict(shape=[15], k=5),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_1D)
+    @pytest.mark.nightly
+    def test_TopK_1D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_topK_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_2D = [
+        dict(shape=[14, 15], k=10),
+        dict(shape=[14, 15], k=5),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_2D)
+    @pytest.mark.nightly
+    def test_TopK_2D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_topK_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_3D = [
+        dict(shape=[13, 14, 15], k=10),
+        dict(shape=[13, 14, 15], k=5),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_3D)
+    @pytest.mark.nightly
+    def test_TopK_3D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_topK_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_4D = [
+        dict(shape=[12, 13, 14, 15], k=10),
+        dict(shape=[12, 13, 14, 15], k=5),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_4D)
+    @pytest.mark.nightly
+    def test_TopK_4D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_topK_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)
+
+    test_data_5D = [
+        dict(shape=[11, 12, 13, 14, 15], k=10),
+        dict(shape=[11, 12, 13, 14, 15], k=5),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_5D)
+    @pytest.mark.nightly
+    def test_TopK_5D(self, params, ie_device, precision, ir_version, temp_dir):
+        self._test(*self.create_topK_net(**params, ir_version=ir_version),
+                   ie_device, precision, ir_version, temp_dir=temp_dir)

From 7b2779c406bcf00c42102a1faedd459ea70a7b45 Mon Sep 17 00:00:00 2001
From: Rafal Blaczkowski <rafal.blaczkowski@intel.com>
Date: Wed, 16 Jun 2021 12:25:06 +0200
Subject: [PATCH 107/178] cleanup improvement (#6186)

---
 .ci/openvino-onnx/Jenkinsfile | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/.ci/openvino-onnx/Jenkinsfile b/.ci/openvino-onnx/Jenkinsfile
index 5fe24928798f3d..ff2e8a451253c4 100644
--- a/.ci/openvino-onnx/Jenkinsfile
+++ b/.ci/openvino-onnx/Jenkinsfile
@@ -155,10 +155,9 @@ def getConfigurationsMap() {
 
 CONFIGURATION_WORKFLOW = { configuration ->
     node("OpenVINO") {
+        String workdir = "${HOME}/workspace/${BUILD_NUMBER}_${env.CHANGE_ID}_${configuration.name}"
         try {
             PROJECT_NAME = "openvino"
-            String workdir = "${HOME}/workspace/${BUILD_NUMBER}_${env.CHANGE_ID}_${configuration.name}"
-
             stage("Clone repository") {
                     prepare_repository(workdir)
             }
@@ -185,10 +184,10 @@ CONFIGURATION_WORKFLOW = { configuration ->
         }
         finally {
             stage("Cleanup") {
-                deleteDir()
                 String docker_container_name = get_docker_container_name(configuration)
                 sh """
                     docker rm -f ${docker_container_name}
+                    rm -rf ${workdir}
                 """
             }
         }

From dc0d482c23432ba72367a84e4639faeb18c32ee5 Mon Sep 17 00:00:00 2001
From: Mateusz Tabaka <mateusz.tabaka@intel.com>
Date: Wed, 16 Jun 2021 12:45:36 +0200
Subject: [PATCH 108/178] Add support for fp64 in Sqrt's evaluate method
 (#5913)

It's required for t2t-vit models.
---
 ngraph/core/src/op/sqrt.cpp      | 4 +++-
 ngraph/test/constant_folding.cpp | 7 ++++++-
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/ngraph/core/src/op/sqrt.cpp b/ngraph/core/src/op/sqrt.cpp
index e706e4ae7c2157..339a8b74706d7c 100644
--- a/ngraph/core/src/op/sqrt.cpp
+++ b/ngraph/core/src/op/sqrt.cpp
@@ -57,6 +57,7 @@ namespace sqrtop
             NGRAPH_TYPE_CASE(evaluate_sqrt, u64, arg0, out, count);
             NGRAPH_TYPE_CASE(evaluate_sqrt, f16, arg0, out, count);
             NGRAPH_TYPE_CASE(evaluate_sqrt, f32, arg0, out, count);
+            NGRAPH_TYPE_CASE(evaluate_sqrt, f64, arg0, out, count);
         default: rc = false; break;
         }
         return rc;
@@ -79,7 +80,8 @@ bool op::Sqrt::has_evaluate() const
     case ngraph::element::u32:
     case ngraph::element::u64:
     case ngraph::element::f16:
-    case ngraph::element::f32: return true;
+    case ngraph::element::f32:
+    case ngraph::element::f64: return true;
     default: break;
     }
     return false;
diff --git a/ngraph/test/constant_folding.cpp b/ngraph/test/constant_folding.cpp
index c34dcb12c6e0b7..d6b2d98ee9f25a 100644
--- a/ngraph/test/constant_folding.cpp
+++ b/ngraph/test/constant_folding.cpp
@@ -300,6 +300,7 @@ TEST(constant_folding, constant_unary_binary)
     auto g = make_shared<op::Constant>(element::i32, Shape{2}, values_g);
     auto h = make_shared<op::Constant>(element::boolean, Shape{2, 2}, values_h);
     auto i = make_shared<op::Constant>(element::boolean, Shape{2}, values_i);
+    auto doubles = make_shared<op::Constant>(element::f64, Shape{2}, std::vector<double>{4.0, 9.0});
 
     auto add = make_shared<op::v1::Add>(a, b);
     auto sub = make_shared<op::v1::Subtract>(a, b);
@@ -328,6 +329,7 @@ TEST(constant_folding, constant_unary_binary)
     auto logical_or_autob_numpy =
         make_shared<op::v1::LogicalOr>(h, i, op::AutoBroadcastType::NUMPY);
     auto logical_xor_autob_numpy = make_shared<op::Xor>(h, i, op::AutoBroadcastType::NUMPY);
+    auto doubles_sqrt = make_shared<op::Sqrt>(doubles);
 
     auto neg_sqrt = make_shared<op::Sqrt>(c);
 
@@ -355,7 +357,8 @@ TEST(constant_folding, constant_unary_binary)
                                                  less_autob_numpy,
                                                  less_eq_autob_numpy,
                                                  logical_or_autob_numpy,
-                                                 logical_xor_autob_numpy},
+                                                 logical_xor_autob_numpy,
+                                                 doubles_sqrt},
                                       ParameterVector{});
     auto func_error = make_shared<Function>(NodeVector{neg_sqrt}, ParameterVector{});
 
@@ -388,6 +391,7 @@ TEST(constant_folding, constant_unary_binary)
     vector<char> less_eq_autob_numpy_expected{1, 1, 0, 1};
     vector<char> logical_or_autob_numpy_expected{0, 1, 1, 1};
     vector<char> logical_xor_autob_numpy_expected{0, 1, 1, 0};
+    vector<double> doubles_sqrt_expected{2.0, 3.0};
 
     ASSERT_EQ(get_result_constant<int>(func, 0), add_expected);
     ASSERT_EQ(get_result_constant<int>(func, 1), sub_expected);
@@ -414,6 +418,7 @@ TEST(constant_folding, constant_unary_binary)
     ASSERT_EQ(get_result_constant<char>(func, 22), less_eq_autob_numpy_expected);
     ASSERT_EQ(get_result_constant<char>(func, 23), logical_or_autob_numpy_expected);
     ASSERT_EQ(get_result_constant<char>(func, 24), logical_xor_autob_numpy_expected);
+    ASSERT_EQ(get_result_constant<double>(func, 25), doubles_sqrt_expected);
     ASSERT_NO_THROW(pass_manager.run_passes(func_error));
 }
 

From d76d2674782d3c3bc83be7c06cfa7db42b804326 Mon Sep 17 00:00:00 2001
From: Mateusz Tabaka <mateusz.tabaka@intel.com>
Date: Wed, 16 Jun 2021 12:46:33 +0200
Subject: [PATCH 109/178] Add support for fp64 in Convert's evaluate (#5911)

It's required for t2t-vit models.
---
 ngraph/core/src/op/convert.cpp   |  4 ++++
 ngraph/test/constant_folding.cpp | 10 ++++++++++
 2 files changed, 14 insertions(+)

diff --git a/ngraph/core/src/op/convert.cpp b/ngraph/core/src/op/convert.cpp
index 4230c3a9afbe71..a920fec7729600 100644
--- a/ngraph/core/src/op/convert.cpp
+++ b/ngraph/core/src/op/convert.cpp
@@ -104,6 +104,7 @@ namespace convert
             TYPE_OUT_CASE(bf16, arg, out);
             TYPE_OUT_CASE(f16, arg, out);
             TYPE_OUT_CASE(f32, arg, out);
+            TYPE_OUT_CASE(f64, arg, out);
             TYPE_OUT_CASE(boolean, arg, out);
         default: rc = false; break;
         }
@@ -129,6 +130,7 @@ namespace convert
             NGRAPH_TYPE_CASE(evaluate_convert, bf16, arg, out);
             NGRAPH_TYPE_CASE(evaluate_convert, f16, arg, out);
             NGRAPH_TYPE_CASE(evaluate_convert, f32, arg, out);
+            NGRAPH_TYPE_CASE(evaluate_convert, f64, arg, out);
             NGRAPH_TYPE_CASE(evaluate_convert, boolean, arg, out);
         default: rc = false; break;
         }
@@ -200,6 +202,7 @@ bool op::v0::Convert::has_evaluate() const
     case ngraph::element::bf16:
     case ngraph::element::f16:
     case ngraph::element::f32:
+    case ngraph::element::f64:
     case ngraph::element::boolean: break;
     default: return false;
     }
@@ -219,6 +222,7 @@ bool op::v0::Convert::has_evaluate() const
     case ngraph::element::bf16:
     case ngraph::element::f16:
     case ngraph::element::f32:
+    case ngraph::element::f64:
     case ngraph::element::boolean: break;
     default: return false;
     }
diff --git a/ngraph/test/constant_folding.cpp b/ngraph/test/constant_folding.cpp
index d6b2d98ee9f25a..0f5ce320531673 100644
--- a/ngraph/test/constant_folding.cpp
+++ b/ngraph/test/constant_folding.cpp
@@ -464,6 +464,16 @@ TEST(constant_folding, const_convert)
         vector<bool> expected{true, false, true, false, true, false, true};
         test_const_convert(in, expected);
     }
+    {
+        vector<int64_t> in{1, 2, 3, 4, 5};
+        vector<double> expected{1.0, 2.0, 3.0, 4.0, 5.0};
+        test_const_convert(in, expected);
+    }
+    {
+        vector<double> in{1.2, 2.1, 3.3, 4.45, 5.02};
+        vector<int64_t> expected{1, 2, 3, 4, 5};
+        test_const_convert(in, expected);
+    }
 }
 
 TEST(constant_folding, shape_of_v0)

From b05977a536bff6f618aabe8008bb43db23b3ea71 Mon Sep 17 00:00:00 2001
From: Alexandra Sidorova <alexandra.sidorova@intel.com>
Date: Wed, 16 Jun 2021 14:14:50 +0300
Subject: [PATCH 110/178] [CPU][IE TESTS] Added more input shapes for Pooling
 tests (#6083)

---
 .../plugin/cpu/single_layer_tests/pooling.cpp | 32 ++++++++++++++-----
 1 file changed, 24 insertions(+), 8 deletions(-)

diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp
index 0df9c464c4f5fd..6fb1f7d1169440 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp
@@ -105,6 +105,22 @@ const auto ref = CPUSpecificParams{{}, {}, {"ref_any"}, "ref_any"};
 const std::vector<CPUSpecificParams> vecCpuConfigs = {ref, sse42, avx, avx512};
 const std::vector<Precision> inpOutPrecision = {Precision::FP32, Precision::BF16};
 
+const std::vector<std::vector<size_t>> inputShapes4D = {
+        std::vector<size_t>{3, 4, 64, 64},
+        std::vector<size_t>{2, 8, 8, 12},
+        std::vector<size_t>{1, 16, 16, 12},
+        std::vector<size_t>{1, 21, 8, 4},
+        std::vector<size_t>{1, 32, 8, 8},
+};
+
+const std::vector<std::vector<size_t>> inputShapes5D = {
+        std::vector<size_t>{1, 4, 16, 16, 16},
+        std::vector<size_t>{2, 8, 8, 8, 8},
+        std::vector<size_t>{2, 16, 12, 16, 20},
+        std::vector<size_t>{1, 19, 16, 20, 8},
+        std::vector<size_t>{1, 32, 16, 8, 12},
+};
+
 const std::vector<poolSpecificParams> paramsMax4D = {
         poolSpecificParams{ ngraph::helpers::PoolingTypes::MAX, {2, 2}, {2, 2}, {0, 0}, {0, 0},
                             ngraph::op::RoundingType::CEIL, ngraph::op::PadType::EXPLICIT, false },
@@ -137,7 +153,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MaxPool_CPU_4D, PoolingLayerCPUTest,
                                 ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
                                 ::testing::Values(InferenceEngine::Layout::ANY),
                                 ::testing::Values(InferenceEngine::Layout::ANY),
-                                ::testing::Values(std::vector<size_t >({1, 3, 64, 64})),
+                                ::testing::ValuesIn(inputShapes4D),
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs)),
                         ::testing::Values(emptyFusingSpec)),
@@ -152,7 +168,7 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D, PoolingLayerCPUTest,
                                         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
-                                        ::testing::Values(std::vector<size_t >({1, 4, 64, 64})),
+                                        ::testing::ValuesIn(inputShapes4D),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                                 ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs)),
                                 ::testing::Values(emptyFusingSpec)),
@@ -167,7 +183,7 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D_NotOptimized, PoolingLayerCPUTest,
                                         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
-                                        ::testing::Values(std::vector<size_t >({1, 4, 64, 64})),
+                                        ::testing::ValuesIn(inputShapes4D),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                                 ::testing::Values(ref),
                                 ::testing::Values(emptyFusingSpec)),
@@ -207,7 +223,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MaxPool_CPU_5D, PoolingLayerCPUTest,
                                         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
-                                        ::testing::Values(std::vector<size_t >({1, 3, 16, 32, 32})),
+                                        ::testing::ValuesIn(inputShapes5D),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                                 ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs)),
                                 ::testing::Values(emptyFusingSpec)),
@@ -222,7 +238,7 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D, PoolingLayerCPUTest,
                                         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
-                                        ::testing::Values(std::vector<size_t >({1, 4, 32, 32, 32})),
+                                        ::testing::ValuesIn(inputShapes5D),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                                 ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigs)),
                                 ::testing::Values(emptyFusingSpec)),
@@ -237,7 +253,7 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D_NotOptimized, PoolingLayerCPUTest,
                                         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
-                                        ::testing::Values(std::vector<size_t >({1, 4, 16, 16, 16})),
+                                        ::testing::ValuesIn(inputShapes5D),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                                 ::testing::Values(ref),
                                 ::testing::Values(emptyFusingSpec)),
@@ -272,7 +288,7 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D_I8, PoolingLayerCPUTest,
                                         ::testing::Values(Precision::FP32),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
-                                        ::testing::Values(std::vector<size_t >({1, 4, 64, 64})),
+                                        ::testing::ValuesIn(inputShapes4D),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                                 ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigsFusing_4D)),
                                 ::testing::ValuesIn(fusingParamsSet)),
@@ -287,7 +303,7 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D_I8, PoolingLayerCPUTest,
                                         ::testing::Values(Precision::FP32),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
                                         ::testing::Values(InferenceEngine::Layout::ANY),
-                                        ::testing::Values(std::vector<size_t >({1, 4, 16, 16, 16})),
+                                        ::testing::ValuesIn(inputShapes5D),
                                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                                 ::testing::ValuesIn(filterCPUInfoForDevice(vecCpuConfigsFusing_5D)),
                                 ::testing::ValuesIn(fusingParamsSet)),

From 5b847fabe842a9124efbfccb39655e73f4dac19b Mon Sep 17 00:00:00 2001
From: Egor Duplensky <egor.duplenskii@intel.com>
Date: Wed, 16 Jun 2021 14:42:26 +0300
Subject: [PATCH 111/178] [CPU] Refactor load_emitter_context semantic. Update
 store emitter (#5824)

* [CPU] Refactor load_emitter_context semantic. Update store emitter

New load_emitter_context constructor arguments order
seems to be more convenient.
Store emitter now emits bf16 emu.
---
 .../mkldnn_plugin/emitters/jit_load_store_emitters.cpp   | 5 +++++
 .../mkldnn_plugin/emitters/jit_load_store_emitters.hpp   | 6 ++++--
 .../src/mkldnn_plugin/nodes/mkldnn_mvn_node.cpp          | 7 +++----
 .../src/mkldnn_plugin/nodes/mkldnn_roi_pooling_node.cpp  | 9 ++++-----
 4 files changed, 16 insertions(+), 11 deletions(-)

diff --git a/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.cpp b/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.cpp
index 83bc04c530db07..4d1e3819394fc8 100644
--- a/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.cpp
+++ b/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.cpp
@@ -510,6 +510,11 @@ size_t jit_store_emitter::aux_vecs_count() const {
 
 size_t jit_store_emitter::get_inputs_num() const { return 1; }
 
+void jit_store_emitter::emit_data() const {
+    if (emu_vcvtneps2bf16)
+        emu_vcvtneps2bf16->emit_data();
+}
+
 void jit_store_emitter::emit_impl(const std::vector<size_t> &in_idxs, const std::vector<size_t> &out_idxs,
                   const std::vector<size_t> &pool_vec_idxs, const std::vector<size_t> &pool_gpr_idxs,
                   const emitter_context *emit_context) const {
diff --git a/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.hpp b/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.hpp
index 00c2e49262d9a5..ec863d0c69e596 100644
--- a/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.hpp
+++ b/inference-engine/src/mkldnn_plugin/emitters/jit_load_store_emitters.hpp
@@ -18,8 +18,8 @@ struct load_emitter_context : public emitter_context {
     load_emitter_context() : src_prc_(Precision::FP32), dst_prc_(Precision::FP32), load_num_(8),
     offset_byte_(0), is_fill_(false), fill_value_("zero") {}
 
-    load_emitter_context(Precision src_prc, Precision dst_prc, int load_num, bool is_fill = false, std::string fill_value = "zero", int offset_byte = 0):
-    src_prc_(src_prc), dst_prc_(dst_prc), load_num_(load_num), is_fill_(is_fill), fill_value_(fill_value), offset_byte_(offset_byte) {}
+    load_emitter_context(Precision src_prc, Precision dst_prc, int load_num, int offset_byte = 0, bool is_fill = false, std::string fill_value = "zero"):
+    src_prc_(src_prc), dst_prc_(dst_prc), load_num_(load_num), offset_byte_(offset_byte), is_fill_(is_fill), fill_value_(fill_value) {}
 
     int offset_byte_;
     int load_num_;
@@ -124,6 +124,8 @@ class jit_store_emitter : public jit_emitter {
 
     size_t get_inputs_num() const override;
 
+    void emit_data() const override;
+
     std::shared_ptr<jit_emu_vcvtneps2bf16> get_emu_vcvtneps2bf16() const {
         return emu_vcvtneps2bf16;
     }
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_mvn_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_mvn_node.cpp
index f27a40e3bd255a..baff79e5d75317 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_mvn_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_mvn_node.cpp
@@ -306,8 +306,8 @@ struct jit_uni_mvn_mean_variance_kernel_f32 : public jit_uni_mvn_mean_variance_k
     inline void worker_tail_planar() {
         Precision dst_prc = isFloatCompatible(jcp_.src_prc) ? Precision::FP32 : Precision::I32;
         load_emitter->emit_code({static_cast<size_t>(reg_src.getIdx())}, {static_cast<size_t>(vmm_val.getIdx())},
-                            std::make_shared<load_emitter_context>(jcp_.src_prc, dst_prc, tail_num, true, "zero"),
-                            {}, {load_pool_gpr_idxs});
+                                std::make_shared<load_emitter_context>(jcp_.src_prc, dst_prc, tail_num, 0, true),
+                                {}, {load_pool_gpr_idxs});
 
         if (jcp_.normalize_variance) {
             if (!isFloatCompatible(jcp_.src_prc))
@@ -477,8 +477,7 @@ struct jit_uni_mvn_kernel_f32 : public jit_uni_mvn_kernel, public jit_generator
         this->postamble();
 
         load_emitter->emit_data();
-        if (!mayiuse(avx512_core_bf16) && mayiuse(avx512_core) && store_emitter != nullptr && store_emitter->get_emu_vcvtneps2bf16() != nullptr)
-            store_emitter->get_emu_vcvtneps2bf16()->emit_data();
+        store_emitter->emit_data();
 
         for (auto& inj : eltwise_injectors)
             inj->prepare_table();
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_roi_pooling_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_roi_pooling_node.cpp
index 77db762169254e..a1a7f8329a5c52 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_roi_pooling_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_roi_pooling_node.cpp
@@ -88,8 +88,7 @@ struct jit_uni_roi_pooling_kernel_f32 : public jit_uni_roi_pooling_kernel, publi
         this->postamble();
 
         load_emitter->emit_data();
-        if (!mayiuse(avx512_core_bf16) && mayiuse(avx512_core) && store_emitter != nullptr && store_emitter->get_emu_vcvtneps2bf16() != nullptr)
-            store_emitter->get_emu_vcvtneps2bf16()->emit_data();
+        store_emitter->emit_data();
     }
 
 private:
@@ -155,7 +154,7 @@ struct jit_uni_roi_pooling_kernel_f32 : public jit_uni_roi_pooling_kernel, publi
             Vmm vmm_max = get_acc_reg(i);
 
             load_emitter->emit_code({static_cast<size_t>(reg_input.getIdx())}, {static_cast<size_t>(vmm_max.getIdx())},
-                                    std::make_shared<load_emitter_context>(jpp_.src_prc, Precision::FP32, step, false, "zero", i * src_c_off),
+                                    std::make_shared<load_emitter_context>(jpp_.src_prc, Precision::FP32, step, i * src_c_off),
                                     {}, load_pool_gpr_idxs);
         }
 
@@ -169,7 +168,7 @@ struct jit_uni_roi_pooling_kernel_f32 : public jit_uni_roi_pooling_kernel, publi
                     Vmm vmm_src = get_src_reg(i);
 
                     load_emitter->emit_code({static_cast<size_t>(aux_reg_input1.getIdx())}, {static_cast<size_t>(vmm_src.getIdx())},
-                                            std::make_shared<load_emitter_context>(jpp_.src_prc, Precision::FP32, step, false, "zero", i * src_c_off),
+                                            std::make_shared<load_emitter_context>(jpp_.src_prc, Precision::FP32, step, i * src_c_off),
                                             {}, load_pool_gpr_idxs);
 
                     if (isa == cpu::x64::sse41) {
@@ -222,7 +221,7 @@ struct jit_uni_roi_pooling_kernel_f32 : public jit_uni_roi_pooling_kernel, publi
 
         for (int i = 0; i < c_blocks; i++) {
             const int src_c_off = i * jpp_.ih * jpp_.iw * jpp_.c_block * jpp_.src_data_size;
-            const auto load_context = std::make_shared<load_emitter_context>(jpp_.src_prc, Precision::FP32, step, false, "zero", src_c_off);
+            const auto load_context = std::make_shared<load_emitter_context>(jpp_.src_prc, Precision::FP32, step, src_c_off);
 
             mov(aux_reg_input, reg_input);
 

From 99ebda98f16f2290285f8c6673e039e3104c2687 Mon Sep 17 00:00:00 2001
From: Vladislav Volkov <vladislav.volkov@intel.com>
Date: Wed, 16 Jun 2021 15:54:40 +0300
Subject: [PATCH 112/178] [CPU] Memory leak in jit_uni_i8i8_pooling kernel
 (#6188)

---
 inference-engine/thirdparty/mkl-dnn | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/inference-engine/thirdparty/mkl-dnn b/inference-engine/thirdparty/mkl-dnn
index 87516e47dae71f..e0381c369fc3be 160000
--- a/inference-engine/thirdparty/mkl-dnn
+++ b/inference-engine/thirdparty/mkl-dnn
@@ -1 +1 @@
-Subproject commit 87516e47dae71fc9c326d0f3685c1572c740e127
+Subproject commit e0381c369fc3bed487b0dcfef7e9fcb2e0aea575

From b1db3448e7b72fa54dbf60d26c7155f03a0baa7c Mon Sep 17 00:00:00 2001
From: Tomasz Jankowski <tomasz1.jankowski@intel.com>
Date: Wed, 16 Jun 2021 14:57:29 +0200
Subject: [PATCH 113/178] [ONNX] Extend ONNX Importer for operation "Einsum"
 (#6074)

---
 ngraph/frontend/onnx_import/src/op/einsum.cpp | 30 ++++++++++++
 ngraph/frontend/onnx_import/src/op/einsum.hpp | 25 ++++++++++
 .../frontend/onnx_import/src/ops_bridge.cpp   |  2 +
 ngraph/python/tests/__init__.py               |  4 +-
 .../python/tests/test_ngraph/test_einsum.py   |  6 +--
 ngraph/python/tests/test_onnx/test_backend.py | 15 +++---
 ngraph/test/models/onnx/einsum_sum.prototxt   | 47 +++++++++++++++++++
 ngraph/test/onnx/onnx_import.in.cpp           | 33 +++++++++++--
 8 files changed, 142 insertions(+), 20 deletions(-)
 create mode 100644 ngraph/frontend/onnx_import/src/op/einsum.cpp
 create mode 100644 ngraph/frontend/onnx_import/src/op/einsum.hpp
 create mode 100644 ngraph/test/models/onnx/einsum_sum.prototxt

diff --git a/ngraph/frontend/onnx_import/src/op/einsum.cpp b/ngraph/frontend/onnx_import/src/op/einsum.cpp
new file mode 100644
index 00000000000000..1d2a7a4edc997a
--- /dev/null
+++ b/ngraph/frontend/onnx_import/src/op/einsum.cpp
@@ -0,0 +1,30 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "op/einsum.hpp"
+#include "default_opset.hpp"
+
+namespace ngraph
+{
+    namespace onnx_import
+    {
+        namespace op
+        {
+            namespace set_1
+            {
+                OutputVector einsum(const Node& node)
+                {
+                    const std::string& equation{node.get_attribute_value<std::string>("equation")};
+
+                    return OutputVector{
+                        std::make_shared<default_opset::Einsum>(node.get_ng_inputs(), equation)};
+                }
+
+            } // namespace set_1
+
+        } // namespace op
+
+    } // namespace onnx_import
+
+} // namespace ngraph
diff --git a/ngraph/frontend/onnx_import/src/op/einsum.hpp b/ngraph/frontend/onnx_import/src/op/einsum.hpp
new file mode 100644
index 00000000000000..c9f87479ea0b9a
--- /dev/null
+++ b/ngraph/frontend/onnx_import/src/op/einsum.hpp
@@ -0,0 +1,25 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "ngraph/node.hpp"
+#include "onnx_import/core/node.hpp"
+
+namespace ngraph
+{
+    namespace onnx_import
+    {
+        namespace op
+        {
+            namespace set_1
+            {
+                OutputVector einsum(const Node& node);
+
+            } // namespace set_1
+        }     // namespace op
+
+    } // namespace onnx_import
+
+} // namespace ngraph
diff --git a/ngraph/frontend/onnx_import/src/ops_bridge.cpp b/ngraph/frontend/onnx_import/src/ops_bridge.cpp
index f7864360ba7b43..0055b8afc4a4ca 100644
--- a/ngraph/frontend/onnx_import/src/ops_bridge.cpp
+++ b/ngraph/frontend/onnx_import/src/ops_bridge.cpp
@@ -42,6 +42,7 @@
 #include "op/dequantize_linear.hpp"
 #include "op/div.hpp"
 #include "op/dropout.hpp"
+#include "op/einsum.hpp"
 #include "op/elu.hpp"
 #include "op/equal.hpp"
 #include "op/erf.hpp"
@@ -340,6 +341,7 @@ namespace ngraph
             REGISTER_OPERATOR("Dropout", 1, dropout);
             REGISTER_OPERATOR("Dropout", 7, dropout);
             REGISTER_OPERATOR("Dropout", 12, dropout);
+            REGISTER_OPERATOR("Einsum", 1, einsum);
             REGISTER_OPERATOR("Elu", 1, elu);
             REGISTER_OPERATOR("Equal", 1, equal);
             REGISTER_OPERATOR("Erf", 1, erf);
diff --git a/ngraph/python/tests/__init__.py b/ngraph/python/tests/__init__.py
index 817d4584534934..3bcf0d961fb3db 100644
--- a/ngraph/python/tests/__init__.py
+++ b/ngraph/python/tests/__init__.py
@@ -25,8 +25,6 @@ def xfail_test(reason="Mark the test as expected to fail", strict=True):
 skip_segfault = pytest.mark.skip(reason="Segmentation fault error")
 xfail_issue_33488 = xfail_test(reason="RuntimeError: nGraph does not support the following ONNX operations:"
                                       "MaxUnpool")
-xfail_issue_33512 = xfail_test(reason="RuntimeError: nGraph does not support the following ONNX operations:"
-                                      "Einsum")
 xfail_issue_33535 = xfail_test(reason="nGraph does not support the following ONNX operations:"
                                       "DynamicQuantizeLinear")
 xfail_issue_33538 = xfail_test(reason="RuntimeError: nGraph does not support the following ONNX operations:"
@@ -145,5 +143,5 @@ def xfail_test(reason="Mark the test as expected to fail", strict=True):
 xfail_issue_49754 = xfail_test(reason="RuntimeError: Unsupported dynamic ops: v1::TopKIE")
 xfail_issue_52463 = xfail_test(reason="test_operator_add_size1_singleton_broadcast_cpu - "
                                       "Not equal to tolerance")
-xfail_issue_45432 = xfail_test(reason="Einsum is not implemented in CPU plugin.")
+xfail_issue_58033 = xfail_test(reason="Einsum operation misses support for complex ellipsis equations")
 xfail_issue_onnx_models_140 = xfail_test(reason="https://github.com/onnx/models/issues/140")
diff --git a/ngraph/python/tests/test_ngraph/test_einsum.py b/ngraph/python/tests/test_ngraph/test_einsum.py
index a89b6c3ff35d89..fb7581d9160fa9 100644
--- a/ngraph/python/tests/test_ngraph/test_einsum.py
+++ b/ngraph/python/tests/test_ngraph/test_einsum.py
@@ -3,7 +3,7 @@
 import pytest
 
 from ngraph.utils.types import get_element_type
-from tests import xfail_issue_45432
+from tests import xfail_issue_58033
 from tests.runtime import get_runtime
 
 
@@ -86,13 +86,13 @@ def test_simple_ellipsis(data_type):
     einsum_op_exec([(5, 3, 4)], "a...->...", data_type)
 
 
-@xfail_issue_45432
+@xfail_issue_58033
 @pytest.mark.parametrize("data_type", [np.float32, np.int32])
 def test_multiple_ellipsis(data_type):
     einsum_op_exec([(3, 5), 1], "a...,...->a...", data_type, with_value=True)
 
 
-@xfail_issue_45432
+@xfail_issue_58033
 @pytest.mark.parametrize("data_type", [np.float32, np.int32])
 def test_broadcasting_ellipsis(data_type):
     einsum_op_exec([(9, 1, 4, 3), (3, 11, 7, 1)], "a...b,b...->a...", data_type, with_value=True)
diff --git a/ngraph/python/tests/test_onnx/test_backend.py b/ngraph/python/tests/test_onnx/test_backend.py
index a8ee1cbb065c6e..f8a1d99c900eab 100644
--- a/ngraph/python/tests/test_onnx/test_backend.py
+++ b/ngraph/python/tests/test_onnx/test_backend.py
@@ -9,7 +9,6 @@
 
 from tests import (BACKEND_NAME,
                    xfail_issue_33488,
-                   xfail_issue_33512,
                    xfail_issue_33535,
                    xfail_issue_33538,
                    xfail_issue_33581,
@@ -59,7 +58,9 @@
                    xfail_issue_49753,
                    xfail_issue_49754,
                    xfail_issue_52463,
-                   xfail_issue_55760)
+                   xfail_issue_55760,
+                   xfail_issue_58033,
+                   )
 
 
 def expect_fail(test_case_path, xfail):  # type: (str) -> None
@@ -285,12 +286,6 @@ def expect_fail(test_case_path, xfail):  # type: (str) -> None
      "OnnxBackendNodeModelTest.test_qlinearconv_cpu"),
     (xfail_issue_38724,
      "OnnxBackendNodeModelTest.test_resize_tf_crop_and_resize_cpu"),
-    (xfail_issue_33512,
-     "OnnxBackendNodeModelTest.test_einsum_transpose_cpu",
-     "OnnxBackendNodeModelTest.test_einsum_batch_diagonal_cpu",
-     "OnnxBackendNodeModelTest.test_einsum_batch_matmul_cpu",
-     "OnnxBackendNodeModelTest.test_einsum_sum_cpu",
-     "OnnxBackendNodeModelTest.test_einsum_inner_prod_cpu"),
     (xfail_issue_33606,
      "OnnxBackendNodeModelTest.test_det_2d_cpu",
      "OnnxBackendNodeModelTest.test_det_nd_cpu"),
@@ -368,7 +363,9 @@ def expect_fail(test_case_path, xfail):  # type: (str) -> None
      "OnnxBackendNodeModelTest.test_quantizelinear_cpu"),
     (xfail_issue_33593,
      "OnnxBackendNodeModelTest.test_maxpool_with_argmax_2d_precomputed_strides_cpu",
-     "OnnxBackendNodeModelTest.test_maxpool_with_argmax_2d_precomputed_pads_cpu",)
+     "OnnxBackendNodeModelTest.test_maxpool_with_argmax_2d_precomputed_pads_cpu",),
+    (xfail_issue_58033,
+     "OnnxBackendNodeModelTest.test_einsum_batch_diagonal_cpu"),
 ]
 
 for test_group in tests_expected_to_fail:
diff --git a/ngraph/test/models/onnx/einsum_sum.prototxt b/ngraph/test/models/onnx/einsum_sum.prototxt
new file mode 100644
index 00000000000000..0f56b47d3abfe3
--- /dev/null
+++ b/ngraph/test/models/onnx/einsum_sum.prototxt
@@ -0,0 +1,47 @@
+ir_version: 7
+producer_name: "backend-test"
+graph {
+  node {
+    input: "x"
+    output: "y"
+    op_type: "Einsum"
+    attribute {
+      name: "equation"
+      s: "ij->i"
+      type: STRING
+    }
+  }
+  name: "test_einsum_sum"
+  input {
+    name: "x"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 3
+          }
+          dim {
+            dim_value: 4
+          }
+        }
+      }
+    }
+  }
+  output {
+    name: "y"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value: 3
+          }
+        }
+      }
+    }
+  }
+}
+opset_import {
+  version: 12
+}
diff --git a/ngraph/test/onnx/onnx_import.in.cpp b/ngraph/test/onnx/onnx_import.in.cpp
index 849a015c4c3807..943d6b90ca1b2e 100644
--- a/ngraph/test/onnx/onnx_import.in.cpp
+++ b/ngraph/test/onnx/onnx_import.in.cpp
@@ -4615,18 +4615,18 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_constant_sparse_tensor_float_8x17)
                                                         0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 1.f, 0.f, 1.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f,
                                                         0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 1.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f});
     test_case.run();
-}    
-                                                
+}
+
 NGRAPH_TEST(${BACKEND_NAME}, onnx_constant_sparse_tensor_float_2x3x4)
 {
     auto function = onnx_import::import_onnx_model(
     file_util::path_join(SERIALIZED_ZOO, "onnx/constant_sparse_tensor_float_2x3x4.prototxt"));
 
     auto test_case = test::TestCase<TestEngine>(function);
-    test_case.add_expected_output<float>(Shape{2, 3, 4}, {1.f, 0.f, 0.f, 8.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 3.f, 0.f, 
+    test_case.add_expected_output<float>(Shape{2, 3, 4}, {1.f, 0.f, 0.f, 8.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 3.f, 0.f,
                                                           0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 1.f, 0.f, 0.f, 0.f, 3.f, 0.f});
     test_case.run();
-}   
+}
 
 NGRAPH_TEST(${BACKEND_NAME}, onnx_constant_sparse_tensor_float_2x2x3x4)
 {
@@ -4636,7 +4636,30 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_constant_sparse_tensor_float_2x2x3x4)
     auto test_case = test::TestCase<TestEngine>(function);
     test_case.add_expected_output<float>(Shape{2, 2, 3, 4}, {0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 1.f, 0.f, 0.f, 2.f, 3.f,
                                                              0.f, 0.f, 0.f, 0.f, 0.f, 0.f, 1.f, 0.f, 0.f, 0.f, 0.f, 0.f,
-                                                             0.f, 0.f, 0.f, 8.f, 0.f, 1.f, 2.f, 0.f, 0.f, 0.f, 3.f, 0.f, 
+                                                             0.f, 0.f, 0.f, 8.f, 0.f, 1.f, 2.f, 0.f, 0.f, 0.f, 3.f, 0.f,
                                                              1.f, 0.f, 0.f, 0.f, 0.f, 2.f, 0.f, 0.f, 0.f, 0.f, 0.f, 0.f});
     test_case.run();
 }
+
+NGRAPH_TEST(${BACKEND_NAME}, onnx_einsum_sum)
+{
+    auto function = onnx_import::import_onnx_model(
+        file_util::path_join(SERIALIZED_ZOO, "onnx/einsum_sum.prototxt"));
+    auto test_case = test::TestCase<TestEngine>(function);
+    test_case.add_input<float>(Shape{3, 4},
+                               {1.764052345967664,
+                                0.4001572083672233,
+                                0.9787379841057392,
+                                2.240893199201458,
+                                1.8675579901499675,
+                                -0.977277879876411,
+                                0.9500884175255894,
+                                -0.1513572082976979,
+                                -0.10321885179355784,
+                                0.41059850193837233,
+                                0.144043571160878,
+                                1.454273506962975});
+    test_case.add_expected_output<float>(
+        Shape{3}, {5.3838407376420845, 1.689011319501448, 1.9056967282686674});
+    test_case.run();
+}

From 2c775d48b27e3148769ce9998d48e75642c019a4 Mon Sep 17 00:00:00 2001
From: Gleb Kazantaev <gleb.kazantaev@intel.com>
Date: Wed, 16 Jun 2021 16:02:57 +0300
Subject: [PATCH 114/178] Remove Pruning from MO (#6191)

---
 .../src/offline_transformations/src/moc_transformations.cpp   | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/inference-engine/src/offline_transformations/src/moc_transformations.cpp b/inference-engine/src/offline_transformations/src/moc_transformations.cpp
index 35f4a575c15c74..745e173b4d765d 100644
--- a/inference-engine/src/offline_transformations/src/moc_transformations.cpp
+++ b/inference-engine/src/offline_transformations/src/moc_transformations.cpp
@@ -12,9 +12,5 @@
 NGRAPH_RTTI_DEFINITION(ngraph::pass::MOCTransformations, "MOCTransformations", 0);
 
 bool ngraph::pass::MOCTransformations::run_on_function(std::shared_ptr<ngraph::Function> f) {
-    ngraph::pass::Manager m(get_pass_config());
-    m.register_pass<Pruning>();
-    m.run_passes(f);
-
     return false;
 }
\ No newline at end of file

From 5c55d390e82ba055bb09e7366b4a105180c46c0e Mon Sep 17 00:00:00 2001
From: Elizaveta Lobanova <elizaveta.lobanova@intel.com>
Date: Wed, 16 Jun 2021 16:19:21 +0300
Subject: [PATCH 115/178] [GNA] Allow 2d reshape of the first diagonal layer
 (#6115)

---
 inference-engine/src/gna_plugin/gna_groups.hpp             | 7 -------
 .../src/gna_plugin/optimizer/gna_pass_manager.cpp          | 5 ++---
 .../gna/pass_tests/convert_matmul_to_pointwise_conv.cpp    | 3 +--
 3 files changed, 3 insertions(+), 12 deletions(-)

diff --git a/inference-engine/src/gna_plugin/gna_groups.hpp b/inference-engine/src/gna_plugin/gna_groups.hpp
index 2449338821c109..704588a153d039 100644
--- a/inference-engine/src/gna_plugin/gna_groups.hpp
+++ b/inference-engine/src/gna_plugin/gna_groups.hpp
@@ -52,13 +52,6 @@ inline bool HasTo2DReshapeData(InferenceEngine::CNNLayerPtr layer) {
     if (!GNAPluginNS::LayerInfo(layer).isSyntheticScaleShift())
         return false;
 
-    // Don't reshape the first dnn layer since it breaks groups recognition
-    auto prevLayer = InferenceEngine::CNNNetPrevLayerSkipCertain(layer, 0, [](InferenceEngine::CNNLayerPtr ptr) {
-        return LayerInfo(ptr).isNonValuesChangable();
-    });
-    IE_ASSERT(prevLayer != nullptr);
-    if (LayerInfo(prevLayer).isInput()) return false;
-
     // Don't reshape diagonallayers with bias connection
     return !GNAPluginNS::LayerInfo(getCreatorLayer(layer->insData.front().lock()).lock()).has32BOutput();
 }
diff --git a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
index b8962cebd36e23..ef333e7e46f115 100644
--- a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
+++ b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
@@ -85,9 +85,8 @@ static void insertDiagonalLayerBetween(InferenceEngine::CNNLayerPtr prevLayer,
         return LayerInfo(ptr).isNonValuesChangable();
     });
     IE_ASSERT(inputLayer != nullptr);
-    size_t weightsSize = (LayerInfo(prevLayer).has32BOutput() || LayerInfo(inputLayer).isInput()) ?
-                         nextLayer->outData[0]->getDims().back() :
-                         Get2DReshapedData(nextLayer->outData[0], 8)->getDims()[1];
+    size_t weightsSize = LayerInfo(prevLayer).has32BOutput() ? nextLayer->outData[0]->getDims().back() :
+        Get2DReshapedData(nextLayer->outData[0], 8)->getDims()[1];
     std::vector<float> weightsValues(weightsSize, fillValue);
     IE_ASSERT(diagLayer != nullptr);
     diagLayer->_weights = make_shared_blob<float>(
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp
index 7e3d15174f3e5f..86bfe5e25c013a 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp
@@ -217,8 +217,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvertMatmulToPointwiseConvTest, ConvertMatmulToP
         ::testing::ValuesIn(inputShape)),
     ConvertMatmulToPointwiseConv::getTestCaseName);
 
-// Issue 55662
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_ConvertMatmulToPointwiseConvTest, ConvertMatmulToPointwiseConvWithFq,
+INSTANTIATE_TEST_CASE_P(smoke_ConvertMatmulToPointwiseConvTest, ConvertMatmulToPointwiseConvWithFq,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),

From b8313c4fae5a794508b0808145182fe2dd62caea Mon Sep 17 00:00:00 2001
From: Taylor Yeonbok Lee <taylor.lee@intel.com>
Date: Wed, 16 Jun 2021 23:47:08 +0900
Subject: [PATCH 116/178] [IE CLDNN] Disable crop optimization only when the
 node is inside a loop body program (#6192)

---
 .../src/graph_optimizer/prepare_buffer_fusing.cpp     |  2 +-
 .../thirdparty/clDNN/src/include/loop_inst.h          |  2 +-
 .../thirdparty/clDNN/src/include/program_impl.h       |  8 ++++++--
 inference-engine/thirdparty/clDNN/src/program.cpp     | 11 +++++++----
 4 files changed, 15 insertions(+), 8 deletions(-)

diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_buffer_fusing.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_buffer_fusing.cpp
index eebd65149ad27a..c57e72ad5f4a4e 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_buffer_fusing.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_buffer_fusing.cpp
@@ -283,7 +283,7 @@ void prepare_buffer_fusing::run(program_impl& p) {
             }
 
             if (node.get_dependencies().size() == 1 && node.get_users().size() > 0) {
-                if (node.get_dependency(0).is_type<lstm_elt>()) {
+                if (p.is_loop_body() && node.get_dependency(0).is_type<lstm_elt>()) {
                     return;
                 }
                 // optimization is available for cropping across depth(features) only
diff --git a/inference-engine/thirdparty/clDNN/src/include/loop_inst.h b/inference-engine/thirdparty/clDNN/src/include/loop_inst.h
index e39bb90f43185a..08e7d416d441ea 100644
--- a/inference-engine/thirdparty/clDNN/src/include/loop_inst.h
+++ b/inference-engine/thirdparty/clDNN/src/include/loop_inst.h
@@ -266,7 +266,7 @@ struct typed_program_node<loop> : public typed_program_node_base<loop> {
         auto opts = get_program().get_options();
         std::vector<primitive_id> output_names_vec(output_names.begin(), output_names.end());
         opts.set_option(build_option::outputs(output_names_vec));
-        body_program = program_impl::build_program(get_program().get_engine(), body, opts, false);
+        body_program = program_impl::build_program(get_program().get_engine(), body, opts, false, false, true);
     }
 
     const primitive_id& get_trip_count_id() const { return get_primitive()->trip_count_id; }
diff --git a/inference-engine/thirdparty/clDNN/src/include/program_impl.h b/inference-engine/thirdparty/clDNN/src/include/program_impl.h
index fe8c8cb55f2795..df45537f00ad91 100644
--- a/inference-engine/thirdparty/clDNN/src/include/program_impl.h
+++ b/inference-engine/thirdparty/clDNN/src/include/program_impl.h
@@ -138,7 +138,8 @@ struct program_impl {
                  topology_impl const& topology,
                  build_options const& options,
                  bool is_internal,
-                 bool no_optimizations = false);
+                 bool no_optimizations = false,
+                 bool is_body_program = false);
     /* constructor used to build a program from subset of nodes of other program (used in propagate_constants) */
     program_impl(engine& engine_ref,
                  std::set<std::shared_ptr<program_node>> const& nodes,
@@ -153,6 +154,7 @@ struct program_impl {
     std::vector<program_node*>& get_outputs() {
         return outputs;
     }  // ToDo: redesign reorder-inputs pass to make it const as_well as get_engine and get options
+    bool is_loop_body() const { return is_body_program; }
     bool is_debug_build() const { return options.get<build_option_type::debug>()->enabled(); }
     const nodes_ordering& get_processing_order() const;
     nodes_ordering& get_processing_order();
@@ -228,7 +230,8 @@ struct program_impl {
                              const topology_impl& topology,
                              const build_options& options,
                              bool is_internal = false,
-                             bool no_optimizations = false);
+                             bool no_optimizations = false,
+                             bool is_body_program = false);
     static ptr build_program(engine& engine,
                              const std::set<std::shared_ptr<program_node>>& nodes,
                              const build_options& options,
@@ -253,6 +256,7 @@ struct program_impl {
     nodes_ordering processing_order;
     std::unique_ptr<pass_manager> pm;
     std::shared_ptr<kernel_selector::TuningCache> tuning_cache;
+    bool is_body_program;
 
 
     std::map<primitive_id, std::shared_ptr<program_node>> nodes_map;
diff --git a/inference-engine/thirdparty/clDNN/src/program.cpp b/inference-engine/thirdparty/clDNN/src/program.cpp
index ec59a526be0ac8..1bd903640287cd 100644
--- a/inference-engine/thirdparty/clDNN/src/program.cpp
+++ b/inference-engine/thirdparty/clDNN/src/program.cpp
@@ -84,13 +84,15 @@ program_impl::program_impl(engine& engine_ref,
                            topology_impl const& topology,
                            build_options const& options,
                            bool is_internal,
-                           bool no_optimizations)
+                           bool no_optimizations,
+                           bool is_body_program)
     : _engine(engine_ref),
       _stream(_engine.create_stream()),
       program_state(_engine),
       options(options),
       processing_order(),
-      tuning_cache(nullptr) {
+      tuning_cache(nullptr),
+      is_body_program(is_body_program) {
     init_primitives();
     kernel_selector::KernelBase::ResetCounter();
     set_options();
@@ -163,8 +165,9 @@ program_impl::ptr program_impl::build_program(engine& engine,
                                               const topology_impl& topology,
                                               const build_options& options,
                                               bool is_internal,
-                                              bool no_optimizations) {
-    return std::make_shared<program_impl>(engine, topology, options, is_internal, no_optimizations);
+                                              bool no_optimizations,
+                                              bool is_body_program) {
+    return std::make_shared<program_impl>(engine, topology, options, is_internal, no_optimizations, is_body_program);
 }
 
 program_impl::ptr program_impl::build_program(engine& engine,

From db67c1b2b99f36f5ac46713c41b518db25cac1c0 Mon Sep 17 00:00:00 2001
From: Gabriele Galiero Casay <gabriele.galiero.casay@intel.com>
Date: Wed, 16 Jun 2021 17:11:25 +0200
Subject: [PATCH 117/178] Add reduce operations to script with trusted ops
 (#6200)

---
 .../layer_tests_summary/utils/constants.py                   | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
index 7e193c2d790df0..8309b0e85935c0 100644
--- a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
+++ b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
@@ -66,7 +66,12 @@
     'ReadValue-6',
     'ReduceL1-4',
     'ReduceL2-4',
+    'ReduceLogicalAnd-1',
+    'ReduceLogicalOr-1',
+    'ReduceMax-1',
     'ReduceMean-1',
+    'ReduceMin-1',
+    'ReduceProd-1',
     'ReduceSum-1',
     'RegionYOLO-1',
     'Relu-1',

From f9771253ae3de75bb666f8b6c2d82b542080861f Mon Sep 17 00:00:00 2001
From: Jozef Daniecki <jozef.daniecki@intel.com>
Date: Thu, 17 Jun 2021 07:47:35 +0200
Subject: [PATCH 118/178] VariadicSplit revision (#6106)

* Add visitor tests for VariadicSplit.

* Add backend tests sceleton.

* Add backend unit tests.

* Add backend tests with dynamic function.

* Remove redundand headers.

* Modified backend tests to check also scalar axis.

* Remove unused headers.

* Changed backend tests to test `-1` value in split_lenghts tensor.

* Const correctness improved in validate_and_infer_types.

* Add VariadicSplit to summarize.py.
---
 .../layer_tests_summary/utils/constants.py    |   3 +-
 .../core/include/ngraph/op/variadic_split.hpp |   2 -
 ngraph/core/src/op/variadic_split.cpp         |  29 +-
 ngraph/test/CMakeLists.txt                    |   2 +
 ngraph/test/backend/variadic_split.in.cpp     | 360 ++++++++++++++++++
 ngraph/test/runtime/ie/unit_test.manifest     |   7 +
 ngraph/test/visitors/op/variadic_split.cpp    |  32 ++
 7 files changed, 416 insertions(+), 19 deletions(-)
 create mode 100644 ngraph/test/backend/variadic_split.in.cpp
 create mode 100644 ngraph/test/visitors/op/variadic_split.cpp

diff --git a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
index 8309b0e85935c0..cdbcdcf6c54ae8 100644
--- a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
+++ b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
@@ -94,5 +94,6 @@
     'TopK-1',
     'TopK-3',
     'Transpose-1',
-    'Unsqueeze-1'
+    'Unsqueeze-1',
+    'VariadicSplit-1',
 ]
diff --git a/ngraph/core/include/ngraph/op/variadic_split.hpp b/ngraph/core/include/ngraph/op/variadic_split.hpp
index 4293270fcbe344..fbff40bc21ac62 100644
--- a/ngraph/core/include/ngraph/op/variadic_split.hpp
+++ b/ngraph/core/include/ngraph/op/variadic_split.hpp
@@ -4,9 +4,7 @@
 
 #pragma once
 
-#include "ngraph/coordinate.hpp"
 #include "ngraph/op/op.hpp"
-#include "ngraph/strides.hpp"
 
 namespace ngraph
 {
diff --git a/ngraph/core/src/op/variadic_split.cpp b/ngraph/core/src/op/variadic_split.cpp
index 505584ec6e1bc2..4b3cf710bc66a5 100644
--- a/ngraph/core/src/op/variadic_split.cpp
+++ b/ngraph/core/src/op/variadic_split.cpp
@@ -5,13 +5,9 @@
 #include <numeric>
 
 #include "itt.hpp"
-#include "ngraph/op/constant.hpp"
-#include "ngraph/op/util/op_types.hpp"
 #include "ngraph/op/variadic_split.hpp"
-#include "ngraph/validation_util.hpp"
-
-#include "ngraph/runtime/host_tensor.hpp"
 #include "ngraph/runtime/reference/slice.hpp"
+#include "ngraph/validation_util.hpp"
 
 using namespace std;
 using namespace ngraph;
@@ -49,11 +45,11 @@ void ngraph::op::v1::VariadicSplit::validate_and_infer_types()
                               split_lengths_pshape.rank(),
                               " instead.");
 
-        auto num_outputs = split_lengths_pshape[0].get_length();
-        auto data = input_value(0);
-        auto axis_source = input_value(1);
-        auto split_lengths_source = input_value(2);
-        auto data_shape = data.get_partial_shape();
+        const auto num_outputs = split_lengths_pshape[0].get_length();
+        const auto data = input_value(0);
+        const auto axis_source = input_value(1);
+        const auto split_lengths_source = input_value(2);
+        const auto data_shape = data.get_partial_shape();
         const auto& data_type = data.get_element_type();
 
         set_output_size(num_outputs);
@@ -61,9 +57,9 @@ void ngraph::op::v1::VariadicSplit::validate_and_infer_types()
         const auto& split_lengths_constant = get_constant_from_source(split_lengths_source);
         if (data_shape.rank().is_static() && axis_input_constant && split_lengths_constant)
         {
-            auto axis_val = axis_input_constant->cast_vector<int64_t>()[0];
+            const auto axis_val = axis_input_constant->cast_vector<int64_t>()[0];
             // Adjust split axis in case of negatives
-            int64_t axis = ngraph::normalize_axis(this, axis_val, data_shape.rank());
+            const int64_t axis = ngraph::normalize_axis(this, axis_val, data_shape.rank());
 
             auto split_lengths = split_lengths_constant->cast_vector<int64_t>();
             // Adjust split lengths in case of negatives
@@ -92,8 +88,8 @@ void ngraph::op::v1::VariadicSplit::validate_and_infer_types()
                     sum_of_splits += split_lengths[i];
                 }
             }
-            auto data_shape_dims = vector<Dimension>{data.get_partial_shape()};
-            auto dimension_at_axis = data_shape_dims.at(axis);
+            const auto data_shape_dims = vector<Dimension>{data.get_partial_shape()};
+            const auto dimension_at_axis = data_shape_dims.at(axis);
 
             if (negative_one >= 0 && dimension_at_axis.is_static())
             {
@@ -112,8 +108,9 @@ void ngraph::op::v1::VariadicSplit::validate_and_infer_types()
 
             for (int64_t output{0}; output < num_outputs; ++output)
             {
-                auto output_split_dim = split_lengths.at(output) == -1 ? Dimension::dynamic()
-                                                                       : split_lengths.at(output);
+                const auto output_split_dim = split_lengths.at(output) == -1
+                                                  ? Dimension::dynamic()
+                                                  : split_lengths.at(output);
                 auto tmp_shape = data_shape_dims;
                 tmp_shape.at(axis) = output_split_dim;
                 set_output_type(output, data_type, PartialShape{tmp_shape});
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index ca45c965c9ffa3..efeb68e670ec94 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -288,6 +288,7 @@ set(SRC
     visitors/op/topk.cpp
     visitors/op/transpose.cpp
     visitors/op/unsqueeze.cpp
+    visitors/op/variadic_split.cpp
     uint4.cpp
     util.cpp
 )
@@ -466,6 +467,7 @@ set(MULTI_TEST_SRC
     backend/unhandled_op.in.cpp
     backend/unsqueeze.in.cpp
     backend/validate_call.in.cpp
+    backend/variadic_split.in.cpp
     backend/zero_sized.in.cpp
 )
 
diff --git a/ngraph/test/backend/variadic_split.in.cpp b/ngraph/test/backend/variadic_split.in.cpp
new file mode 100644
index 00000000000000..d582d567550270
--- /dev/null
+++ b/ngraph/test/backend/variadic_split.in.cpp
@@ -0,0 +1,360 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/engine/test_engines.hpp"
+#include "util/test_case.hpp"
+#include "util/test_control.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+using TestEngine = test::ENGINE_CLASS_NAME(${BACKEND_NAME});
+
+namespace
+{
+    template <typename T>
+    std::shared_ptr<ngraph::Function> StaticVariadicSplit(ngraph::element::Type_t inputs_type,
+                                                          Shape data_shape,
+                                                          Shape axis_shape,
+                                                          Shape split_lenghts_shape,
+                                                          std::vector<T> axis_value,
+                                                          std::vector<T> split_lenghts_value)
+    {
+        const auto data = make_shared<op::Parameter>(inputs_type, data_shape);
+        const auto axis = op::Constant::create(inputs_type, axis_shape, axis_value);
+        const auto split_lengths =
+            op::Constant::create(inputs_type, split_lenghts_shape, split_lenghts_value);
+        const auto variadic_split = make_shared<op::v1::VariadicSplit>(data, axis, split_lengths);
+        return make_shared<Function>(variadic_split, ParameterVector{data});
+    }
+
+    std::shared_ptr<ngraph::Function> DynamicVariadicSplit(ngraph::element::Type_t inputs_type,
+                                                           Shape data_shape,
+                                                           Shape axis_shape,
+                                                           Shape split_lenghts_shape)
+    {
+        const auto data = make_shared<op::Parameter>(inputs_type, data_shape);
+        const auto axis = make_shared<op::Parameter>(inputs_type, axis_shape);
+        const auto split_lengths = make_shared<op::Parameter>(inputs_type, split_lenghts_shape);
+        const auto variadic_split = make_shared<op::v1::VariadicSplit>(data, axis, split_lengths);
+        return make_shared<Function>(variadic_split, ParameterVector{data, axis, split_lengths});
+    }
+} // namespace
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_1d_static)
+{
+    const Shape data_shape{10};
+    const std::vector<int32_t> data{1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+    const Shape axis_shape{1};
+    const std::vector<int32_t> axis_value{0};
+    const Shape split_lenghts_shape{3};
+    const std::vector<int32_t> split_lenghts{5, 3, 2};
+
+    auto test_case = test::TestCase<TestEngine>(StaticVariadicSplit(
+        element::i32, data_shape, axis_shape, split_lenghts_shape, axis_value, split_lenghts));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_expected_output<int32_t>(Shape{5}, {1, 2, 3, 4, 5});
+    test_case.add_expected_output<int32_t>(Shape{3}, {6, 7, 8});
+    test_case.add_expected_output<int32_t>(Shape{2}, {9, 10});
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_1d_dynamic)
+{
+    const Shape data_shape{10};
+    const std::vector<int32_t> data{1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+    const Shape axis_shape{1};
+    const std::vector<int32_t> axis_value{0};
+    const Shape split_lenghts_shape{3};
+    const std::vector<int32_t> split_lenghts{5, 3, 2};
+
+    const auto f = DynamicVariadicSplit(element::i32, data_shape, axis_shape, split_lenghts_shape);
+    auto test_case = test::TestCase<TestEngine, test::TestCaseType::DYNAMIC>(f);
+
+    test_case.add_input(data_shape, data);
+    test_case.add_input(axis_shape, axis_value);
+    test_case.add_input(split_lenghts_shape, split_lenghts);
+    test_case.add_expected_output<int32_t>(Shape{5}, {1, 2, 3, 4, 5});
+    test_case.add_expected_output<int32_t>(Shape{3}, {6, 7, 8});
+    test_case.add_expected_output<int32_t>(Shape{2}, {9, 10});
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_2d_axis_0_static)
+{
+    const Shape data_shape{6, 2};
+    const std::vector<int32_t> data{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12};
+    const Shape axis_shape{}; // scalar
+    const std::vector<int32_t> axis_value{0};
+    const Shape split_lenghts_shape{2};
+    const std::vector<int32_t> split_lenghts{4, 2};
+
+    auto test_case = test::TestCase<TestEngine>(StaticVariadicSplit(
+        element::i32, data_shape, axis_shape, split_lenghts_shape, axis_value, split_lenghts));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_expected_output<int32_t>(Shape{4, 2}, {1, 2, 3, 4, 5, 6, 7, 8});
+    test_case.add_expected_output<int32_t>(Shape{2, 2}, {9, 10, 11, 12});
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_2d_axis_0_dynamic)
+{
+    const Shape data_shape{6, 2};
+    const std::vector<int32_t> data{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12};
+    const Shape axis_shape{}; // scalar
+    const std::vector<int32_t> axis_value{0};
+    const Shape split_lenghts_shape{2};
+    const std::vector<int32_t> split_lenghts{4, 2};
+
+    auto test_case = test::TestCase<TestEngine, test::TestCaseType::DYNAMIC>(
+        DynamicVariadicSplit(element::i32, data_shape, axis_shape, split_lenghts_shape));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_input(axis_shape, axis_value);
+    test_case.add_input(split_lenghts_shape, split_lenghts);
+    test_case.add_expected_output<int32_t>(Shape{4, 2}, {1, 2, 3, 4, 5, 6, 7, 8});
+    test_case.add_expected_output<int32_t>(Shape{2, 2}, {9, 10, 11, 12});
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_2d_axis_1_static)
+{
+    const Shape data_shape{4, 3};
+    const std::vector<int32_t> data{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12};
+    const Shape axis_shape{}; // scalar
+    const std::vector<int32_t> axis_value{1};
+    const Shape split_lenghts_shape{2};
+    const std::vector<int32_t> split_lenghts{1, 2};
+
+    auto test_case = test::TestCase<TestEngine>(StaticVariadicSplit(
+        element::i32, data_shape, axis_shape, split_lenghts_shape, axis_value, split_lenghts));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_expected_output<int32_t>(Shape{4, 1}, {1, 4, 7, 10});
+    test_case.add_expected_output<int32_t>(Shape{4, 2}, {2, 3, 5, 6, 8, 9, 11, 12});
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_2d_axis_1_dynamic)
+{
+    const Shape data_shape{4, 3};
+    const std::vector<int32_t> data{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12};
+    const Shape axis_shape{}; // scalar
+    const std::vector<int32_t> axis_value{1};
+    const Shape split_lenghts_shape{2};
+    const std::vector<int32_t> split_lenghts{1, 2};
+
+    auto test_case = test::TestCase<TestEngine, test::TestCaseType::DYNAMIC>(
+        DynamicVariadicSplit(element::i32, data_shape, axis_shape, split_lenghts_shape));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_input(axis_shape, axis_value);
+    test_case.add_input(split_lenghts_shape, split_lenghts);
+    test_case.add_expected_output<int32_t>(Shape{4, 1}, {1, 4, 7, 10});
+    test_case.add_expected_output<int32_t>(Shape{4, 2}, {2, 3, 5, 6, 8, 9, 11, 12});
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_4d_axis_0_static)
+{
+    const Shape data_shape{6, 2, 3, 1};
+    std::vector<int32_t> data(shape_size(data_shape));
+    std::iota(data.begin(), data.end(), 0);
+    const Shape axis_shape{1};
+    const std::vector<int32_t> axis_value{0};
+    const Shape split_lenghts_shape{3};
+    const std::vector<int32_t> split_lenghts{3, 1, 2};
+
+    auto test_case = test::TestCase<TestEngine>(StaticVariadicSplit(
+        element::i32, data_shape, axis_shape, split_lenghts_shape, axis_value, split_lenghts));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_expected_output<int32_t>(
+        Shape{3, 2, 3, 1}, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17});
+    test_case.add_expected_output<int32_t>(Shape{1, 2, 3, 1}, {18, 19, 20, 21, 22, 23});
+    test_case.add_expected_output<int32_t>(Shape{2, 2, 3, 1},
+                                           {24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35});
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_4d_axis_0_dynamic)
+{
+    const Shape data_shape{6, 2, 3, 1};
+    std::vector<int32_t> data(shape_size(data_shape));
+    std::iota(data.begin(), data.end(), 0);
+    const Shape axis_shape{1};
+    const std::vector<int32_t> axis_value{0};
+    const Shape split_lenghts_shape{3};
+    const std::vector<int32_t> split_lenghts{3, 1, 2};
+
+    auto test_case = test::TestCase<TestEngine, test::TestCaseType::DYNAMIC>(
+        DynamicVariadicSplit(element::i32, data_shape, axis_shape, split_lenghts_shape));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_input(axis_shape, axis_value);
+    test_case.add_input(split_lenghts_shape, split_lenghts);
+    test_case.add_expected_output<int32_t>(
+        Shape{3, 2, 3, 1}, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17});
+    test_case.add_expected_output<int32_t>(Shape{1, 2, 3, 1}, {18, 19, 20, 21, 22, 23});
+    test_case.add_expected_output<int32_t>(Shape{2, 2, 3, 1},
+                                           {24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35});
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_4d_axis_1_static)
+{
+    const Shape data_shape{2, 8, 2, 2};
+    std::vector<int32_t> data(shape_size(data_shape));
+    std::iota(data.begin(), data.end(), 0);
+
+    const Shape axis_shape{1};
+    const std::vector<int32_t> axis_value{1};
+    const Shape split_lenghts_shape{4};
+    const std::vector<int32_t> split_lenghts{1, 3, 2, 2};
+
+    auto test_case = test::TestCase<TestEngine>(StaticVariadicSplit(
+        element::i32, data_shape, axis_shape, split_lenghts_shape, axis_value, split_lenghts));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 2, 2}, {0, 1, 2, 3, 32, 33, 34, 35});
+    test_case.add_expected_output<int32_t>(
+        Shape{2, 3, 2, 2},
+        {4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47});
+    test_case.add_expected_output<int32_t>(
+        Shape{2, 2, 2, 2}, {16, 17, 18, 19, 20, 21, 22, 23, 48, 49, 50, 51, 52, 53, 54, 55});
+    test_case.add_expected_output<int32_t>(
+        Shape{2, 2, 2, 2}, {24, 25, 26, 27, 28, 29, 30, 31, 56, 57, 58, 59, 60, 61, 62, 63});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_4d_axis_1_dynamic)
+{
+    const Shape data_shape{2, 8, 2, 2};
+    std::vector<int32_t> data(shape_size(data_shape));
+    std::iota(data.begin(), data.end(), 0);
+
+    const Shape axis_shape{1};
+    const std::vector<int32_t> axis_value{1};
+    const Shape split_lenghts_shape{4};
+    const std::vector<int32_t> split_lenghts{1, 3, 2, 2};
+
+    auto test_case = test::TestCase<TestEngine, test::TestCaseType::DYNAMIC>(
+        DynamicVariadicSplit(element::i32, data_shape, axis_shape, split_lenghts_shape));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_input(axis_shape, axis_value);
+    test_case.add_input(split_lenghts_shape, split_lenghts);
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 2, 2}, {0, 1, 2, 3, 32, 33, 34, 35});
+    test_case.add_expected_output<int32_t>(
+        Shape{2, 3, 2, 2},
+        {4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47});
+    test_case.add_expected_output<int32_t>(
+        Shape{2, 2, 2, 2}, {16, 17, 18, 19, 20, 21, 22, 23, 48, 49, 50, 51, 52, 53, 54, 55});
+    test_case.add_expected_output<int32_t>(
+        Shape{2, 2, 2, 2}, {24, 25, 26, 27, 28, 29, 30, 31, 56, 57, 58, 59, 60, 61, 62, 63});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_4d_axis_2_static)
+{
+    const Shape data_shape{2, 1, 6, 2};
+    std::vector<int32_t> data(shape_size(data_shape));
+    std::iota(data.begin(), data.end(), 0);
+
+    const Shape axis_shape{1};
+    const std::vector<int32_t> axis_value{2};
+    const Shape split_lenghts_shape{3};
+    const std::vector<int32_t> split_lenghts{3, 1, 2};
+
+    auto test_case = test::TestCase<TestEngine>(StaticVariadicSplit(
+        element::i32, data_shape, axis_shape, split_lenghts_shape, axis_value, split_lenghts));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 3, 2},
+                                           {0, 1, 2, 3, 4, 5, 12, 13, 14, 15, 16, 17});
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 1, 2}, {6, 7, 18, 19});
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 2, 2}, {8, 9, 10, 11, 20, 21, 22, 23});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_4d_axis_2_dynamic)
+{
+    const Shape data_shape{2, 1, 6, 2};
+    std::vector<int32_t> data(shape_size(data_shape));
+    std::iota(data.begin(), data.end(), 0);
+
+    const Shape axis_shape{1};
+    const std::vector<int32_t> axis_value{2};
+    const Shape split_lenghts_shape{3};
+    const std::vector<int32_t> split_lenghts{-1, 1, 2}; // -1 means "all remaining items"
+
+    auto test_case = test::TestCase<TestEngine, test::TestCaseType::DYNAMIC>(
+        DynamicVariadicSplit(element::i32, data_shape, axis_shape, split_lenghts_shape));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_input(axis_shape, axis_value);
+    test_case.add_input(split_lenghts_shape, split_lenghts);
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 3, 2},
+                                           {0, 1, 2, 3, 4, 5, 12, 13, 14, 15, 16, 17});
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 1, 2}, {6, 7, 18, 19});
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 2, 2}, {8, 9, 10, 11, 20, 21, 22, 23});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_4d_axis_3_static)
+{
+    const Shape data_shape{2, 1, 2, 6};
+    std::vector<int32_t> data(shape_size(data_shape));
+    std::iota(data.begin(), data.end(), 0);
+
+    const Shape axis_shape{1};
+    const std::vector<int32_t> axis_value{3};
+    const Shape split_lenghts_shape{3};
+    const std::vector<int32_t> split_lenghts{1, -1, 3}; // -1 means "all remaining items"
+
+    auto test_case = test::TestCase<TestEngine>(StaticVariadicSplit(
+        element::i32, data_shape, axis_shape, split_lenghts_shape, axis_value, split_lenghts));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 2, 1}, {0, 6, 12, 18});
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 2, 2}, {1, 2, 7, 8, 13, 14, 19, 20});
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 2, 3},
+                                           {3, 4, 5, 9, 10, 11, 15, 16, 17, 21, 22, 23});
+
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, variadic_split_4d_axis_3_dynamic)
+{
+    const Shape data_shape{2, 1, 2, 6};
+    std::vector<int32_t> data(shape_size(data_shape));
+    std::iota(data.begin(), data.end(), 0);
+
+    const Shape axis_shape{1};
+    const std::vector<int32_t> axis_value{3};
+    const Shape split_lenghts_shape{3};
+    const std::vector<int32_t> split_lenghts{1, 2, -1}; // -1 means "all remaining items"
+
+    auto test_case = test::TestCase<TestEngine, test::TestCaseType::DYNAMIC>(
+        DynamicVariadicSplit(element::i32, data_shape, axis_shape, split_lenghts_shape));
+
+    test_case.add_input(data_shape, data);
+    test_case.add_input(axis_shape, axis_value);
+    test_case.add_input(split_lenghts_shape, split_lenghts);
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 2, 1}, {0, 6, 12, 18});
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 2, 2}, {1, 2, 7, 8, 13, 14, 19, 20});
+    test_case.add_expected_output<int32_t>(Shape{2, 1, 2, 3},
+                                           {3, 4, 5, 9, 10, 11, 15, 16, 17, 21, 22, 23});
+
+    test_case.run();
+}
\ No newline at end of file
diff --git a/ngraph/test/runtime/ie/unit_test.manifest b/ngraph/test/runtime/ie/unit_test.manifest
index adfebe2b57aac4..aa4079eee82c89 100644
--- a/ngraph/test/runtime/ie/unit_test.manifest
+++ b/ngraph/test/runtime/ie/unit_test.manifest
@@ -1529,6 +1529,13 @@ onnx_controlflow_loop_infinite
 # unsupported dynamic ops
 onnx_dyn_shapes_reduce_max_dynamic_input_rank_negative_axis
 IE_GPU.range_v4_trunc_inputs
+variadic_split_1d_dynamic
+variadic_split_2d_axis_0_dynamic
+variadic_split_2d_axis_1_dynamic
+variadic_split_4d_axis_0_dynamic
+variadic_split_4d_axis_1_dynamic
+variadic_split_4d_axis_2_dynamic
+variadic_split_4d_axis_3_dynamic
 
 # CPU plugin does not support bool type
 IE_CPU.evaluate_2D_gather_elements_3x2_data_bool
diff --git a/ngraph/test/visitors/op/variadic_split.cpp b/ngraph/test/visitors/op/variadic_split.cpp
new file mode 100644
index 00000000000000..ae4b0c8c2d6503
--- /dev/null
+++ b/ngraph/test/visitors/op/variadic_split.cpp
@@ -0,0 +1,32 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+
+#include "ngraph/ngraph.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+#include "ngraph/opsets/opset1.hpp"
+
+#include "util/visitor.hpp"
+
+using namespace std;
+using namespace ngraph;
+using ngraph::test::NodeBuilder;
+using ngraph::test::ValueMap;
+
+TEST(attributes, variadic_split_op)
+{
+    using namespace opset1;
+
+    NodeBuilder::get_ops().register_factory<op::v1::VariadicSplit>();
+    auto data = make_shared<Parameter>(element::i32, Shape{200});
+    auto axis = make_shared<Parameter>(element::i32, Shape{1});
+    auto split_lengths = make_shared<Parameter>(element::i32, Shape{1});
+    
+    auto split = make_shared<VariadicSplit>(data, axis, split_lengths);
+    NodeBuilder builder(split);
+    const auto expected_attr_count = 0;
+
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);    
+}

From 8390f40788dcc28ad8f12f8f9ee6c07b9c83fcf8 Mon Sep 17 00:00:00 2001
From: Edward Shogulin <edward.shogulin@intel.com>
Date: Thu, 17 Jun 2021 12:36:25 +0300
Subject: [PATCH 119/178] [LPT] Empty shape on weights handling: cherry-pick to
 master (#6170)

* [LPT] empty shape on weights fix

* [LPT] SplitTransformation naming fix

* [LPT] tests

Co-authored-by: Vladislav Golubev <vladislav.golubev@intel.com>
---
 .../src/split.cpp                             |  6 +--
 .../src/weightable_layer_transformation.cpp   |  9 +++-
 ...nvolution_backprop_data_transformation.cpp | 41 +++++++++++++++++++
 ...nvolution_backprop_data_transformation.cpp |  7 ++++
 .../convolution_transformation.cpp            |  8 ++++
 ...nvolution_backprop_data_transformation.cpp |  7 ++++
 .../convolution_transformation.cpp            |  8 ++++
 7 files changed, 81 insertions(+), 5 deletions(-)

diff --git a/inference-engine/src/low_precision_transformations/src/split.cpp b/inference-engine/src/low_precision_transformations/src/split.cpp
index 486111dd73778a..919c6b5e87b185 100644
--- a/inference-engine/src/low_precision_transformations/src/split.cpp
+++ b/inference-engine/src/low_precision_transformations/src/split.cpp
@@ -111,13 +111,13 @@ void SplitTransformation::updateOutputs(
         updateOutput(context, lastNodes[0], originalNode);
     } else {
         const std::string originalName = originalNode->get_friendly_name();
-        for (auto& lastNode : lastNodes) {
+        for (size_t outIdx = 0; outIdx < lastNodes.size(); ++outIdx) {
             for (size_t i = 0; i < outputSize; ++i) {
                 std::shared_ptr<ngraph::Node> result = context.function->get_output_op(i);
                 std::shared_ptr<ngraph::Node> outputNode = result->get_input_node_shared_ptr(0);
-                if (outputNode.get() == lastNode.get()) {
+                if (outputNode.get() == lastNodes[outIdx].get()) {
                     originalNode->set_friendly_name(originalName + LayerTransformation::originalLayerPostfix);
-                    lastNode->set_friendly_name(originalName + "." + std::to_string(i));
+                    lastNodes[outIdx]->set_friendly_name(originalName + "." + std::to_string(outIdx));
                     break;
                 }
             }
diff --git a/inference-engine/src/low_precision_transformations/src/weightable_layer_transformation.cpp b/inference-engine/src/low_precision_transformations/src/weightable_layer_transformation.cpp
index ce0ae3473d92e5..babcc95303cffe 100644
--- a/inference-engine/src/low_precision_transformations/src/weightable_layer_transformation.cpp
+++ b/inference-engine/src/low_precision_transformations/src/weightable_layer_transformation.cpp
@@ -157,10 +157,15 @@ bool WeightableLayerTransformation::canBeTransformed(const TransformationContext
         }
 
         const size_t outChannelsShapeIndex = is_type<opset1::ConvolutionBackpropData>(layer) ? 1ul : 0ul;
-        if ( // Check if all dimensions of scale except the output channels are all ones
+        if (
+            // expected, it's ok: return true
+            (shape_size(constOutputShape) != 1ul) &&
+            // not expected, something wrong: return false
+            ((constOutputShape.size() <= outChannelsShapeIndex) ||
+            // Check if all dimensions of scale except the output channels are all ones
             (shape_size(constOutputShape) != constOutputShape[outChannelsShapeIndex]) ||
             ((constOutputShape[outChannelsShapeIndex] != 1ul) &&
-                (fqFromWeights->get_output_shape(0)[outChannelsShapeIndex] != constOutputShape[outChannelsShapeIndex]))) {
+                (fqFromWeights->get_output_shape(0)[outChannelsShapeIndex] != constOutputShape[outChannelsShapeIndex])))) {
             return false;
         }
     } else {
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_backprop_data_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_backprop_data_transformation.cpp
index 8e89b952526462..10ba4513f20549 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_backprop_data_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_backprop_data_transformation.cpp
@@ -116,6 +116,7 @@ class ConvolutionBackpropDataTransformation : public LayerTransformation, public
         SimpleLowPrecisionTransformer transform;
         transform.add<ngraph::pass::low_precision::ConvolutionBackpropDataTransformation, ngraph::opset1::Convolution>(testValues.params);
         transform.transform(actualFunction);
+
         std::shared_ptr<Node> refWeights = pass::low_precision::fold<opset1::Broadcast>(
                 testValues.expected.weights,
                 opset1::Constant::create(
@@ -202,6 +203,26 @@ const std::vector<ConvolutionBackpropDataTransformationTestValues> testValues =
             true
         }
     },
+    // with zero point
+    {
+        LayerTransformation::createParamsU8I8(),
+        // ActualValues
+        {
+            ngraph::element::u8,
+            {{ngraph::element::f32}, { 128.f }, { 0.02f }},
+            { 255ul, Shape({}), { 0.f }, { 254.f }, { -1.27f }, { 1.27f } },
+            op::Constant::create(ngraph::element::i8, ngraph::Shape{}, std::vector<float>{ 2.f })
+        },
+        // ExpectedValues
+        {
+            ngraph::element::u8,
+            {{}, { { 128.f }, ngraph::element::f32, {}, false }, {}},
+            {},
+            {{}, {}, {{ 0.0002f }, ngraph::element::f32, { 1 }}},
+            op::Constant::create(ngraph::element::i8, ngraph::Shape{}, std::vector<float>{ -125.f }),
+            true
+        }
+    },
     // updatePrecisions = false
     {
         LayerTransformation::createParamsU8I8().setUpdatePrecisions(false),
@@ -262,6 +283,26 @@ const std::vector<ConvolutionBackpropDataTransformationTestValues> testValues =
             true
         }
     },
+    // without zero point
+    {
+        LayerTransformation::createParamsU8I8(),
+        // ActualValues
+        {
+            ngraph::element::u8,
+            {{ngraph::element::f32}, {}, { 0.02f }},
+            { 255ul, Shape({}), { 0.f }, { 254.f }, { -1.27f }, { 1.27f } },
+            op::Constant::create(ngraph::element::i8, ngraph::Shape{}, std::vector<float>{ 2.f })
+        },
+        // ExpectedValues
+        {
+            ngraph::element::u8,
+            {},
+            {},
+            {{}, {}, {{ 0.0002f }, ngraph::element::f32, { 1 }}},
+            op::Constant::create(ngraph::element::i8, ngraph::Shape{}, std::vector<float>{ -125.f }),
+            true
+        }
+    },
     // QDq version
     {
         LayerTransformation::createParamsU8I8(),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
index a6e2e93a37ebb8..223f9743a61888 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
@@ -27,6 +27,13 @@ const std::vector<LayerTestsDefinitions::ConvolutionBackpropDataTransformationPa
         "U8"
     },
     // FQ on weights
+    {
+        {256ul, ngraph::Shape{}, { 0.f }, { 25.5f }, { 0.f }, { 25.5f }},
+        {255ul, ngraph::Shape{}, { -12.7f }, { 12.7f }, { -12.7f }, { 12.7f }},
+        "convolutionBackpropData_original",
+        "U8"
+    },
+    // FQ on weights
     // with zero point
     {
         {256ul, ngraph::Shape{1, 1, 1, 1}, { 0.f }, { 255.f }, { -12.7f }, { 12.8f }},
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
index 086a2ef6f16a29..594e645980aa74 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
@@ -46,6 +46,14 @@ const std::vector<LayerTestsDefinitions::ConvolutionTransformationParam> params
         "Convolution",
         "U8"
     },
+    {
+        { 256ul, ngraph::Shape {}, { 0.f }, { 255.f }, { 0.f }, { 25.5f } },
+        false,
+        { 255ul, ngraph::Shape {}, { 0.f }, { 254.f }, { -12.7f }, { 12.7f } },
+        false,
+        "Convolution",
+        "U8"
+    },
     {
         { 16ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 255.f }, { 0.f }, { 25.5f } },
         false,
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
index 2cf625b58628a6..697059e3ce24bf 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
@@ -28,6 +28,13 @@ const std::vector<LayerTestsDefinitions::ConvolutionBackpropDataTransformationPa
         "U8"
     },
     // FQ on weights
+    {
+        {256ul, ngraph::Shape{}, { 0.f }, { 25.5f }, { 0.f }, { 25.5f }},
+        {255ul, ngraph::Shape{}, { -12.7f }, { 12.7f }, { -12.7f }, { 12.7f }},
+        "convolutionBackpropData_original",
+        "U8"
+    },
+    // FQ on weights
     {
         {256ul, ngraph::Shape{1, 1, 1, 1}, { -12.8f }, { 12.7f }, { -12.8f }, { 12.7f }},
         {255ul, ngraph::Shape{1, 1, 1, 1}, { -12.7f }, { 12.7f }, { -12.7f }, { 12.7f }},
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
index ce0db6e1e6681d..2a0c8ff14fc7ac 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
@@ -46,6 +46,14 @@ const std::vector<LayerTestsDefinitions::ConvolutionTransformationParam> params
         "Convolution",
         "U8"
     },
+    {
+        { 256ul, ngraph::Shape {}, { 0.f }, { 255.f }, { 0.f }, { 25.5f } },
+        false,
+        { 255ul, ngraph::Shape {}, { 0.f }, { 254.f }, { -12.7f }, { 12.7f } },
+        false,
+        "Convolution",
+        "U8"
+    },
     {
         { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 255.f }, { -12.75f }, { 6.375f } },
         true,

From ccd568a2d2b0bc5b10e0b4469c1e9acb0681a26a Mon Sep 17 00:00:00 2001
From: Pavel Esir <pavel.esir@intel.com>
Date: Thu, 17 Jun 2021 13:02:07 +0300
Subject: [PATCH 120/178] [nG] Gather-8 shell (#6171)

* gather-8 nGraph shell

* added opset8 tbl; added visit_attribute test

* corrected opset tbl
---
 ngraph/core/include/ngraph/op/gather.hpp      |  27 ++
 ngraph/core/include/ngraph/opsets/opset8.hpp  |  17 +
 .../core/include/ngraph/opsets/opset8_tbl.hpp | 179 ++++++++
 ngraph/core/src/op/gather.cpp                 |  47 ++
 ngraph/test/type_prop/gather.cpp              | 413 +++++++++++++++++-
 ngraph/test/visitors/op/gather.cpp            |  16 +
 6 files changed, 691 insertions(+), 8 deletions(-)
 create mode 100644 ngraph/core/include/ngraph/opsets/opset8.hpp
 create mode 100644 ngraph/core/include/ngraph/opsets/opset8_tbl.hpp

diff --git a/ngraph/core/include/ngraph/op/gather.hpp b/ngraph/core/include/ngraph/op/gather.hpp
index 9293521a7d0047..61bbf33bc3a2be 100644
--- a/ngraph/core/include/ngraph/op/gather.hpp
+++ b/ngraph/core/include/ngraph/op/gather.hpp
@@ -61,5 +61,32 @@ namespace ngraph
                     clone_with_new_inputs(const OutputVector& new_args) const override;
             };
         } // namespace v7
+
+        namespace v8
+        {
+            /// \brief Gather slices from axis of params according to indices
+            class NGRAPH_API Gather : public op::util::GatherBase
+            {
+            public:
+                NGRAPH_RTTI_DECLARATION;
+                Gather() = default;
+
+                /// \param data The tensor from which slices are gathered
+                /// \param indices Tensor with indexes to gather
+                /// \param axis The tensor is a dimension index to gather data from
+                /// \param batch_dims The number of batch dimension in data and indices tensors.
+                Gather(const Output<Node>& data,
+                       const Output<Node>& indices,
+                       const Output<Node>& axis,
+                       const int64_t batch_dims = 0);
+
+                bool visit_attributes(AttributeVisitor& visitor) override;
+                void validate_and_infer_types() override;
+                int64_t get_batch_dims() const;
+
+                std::shared_ptr<Node>
+                    clone_with_new_inputs(const OutputVector& new_args) const override;
+            };
+        } // namespace v8
     }     // namespace op
 } // namespace ngraph
diff --git a/ngraph/core/include/ngraph/opsets/opset8.hpp b/ngraph/core/include/ngraph/opsets/opset8.hpp
new file mode 100644
index 00000000000000..f31a3142cf69a7
--- /dev/null
+++ b/ngraph/core/include/ngraph/opsets/opset8.hpp
@@ -0,0 +1,17 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "ngraph/ops.hpp"
+
+namespace ngraph
+{
+    namespace opset8
+    {
+#define NGRAPH_OP(a, b) using b::a;
+#include "ngraph/opsets/opset8_tbl.hpp"
+#undef NGRAPH_OP
+    } // namespace opset8
+} // namespace ngraph
diff --git a/ngraph/core/include/ngraph/opsets/opset8_tbl.hpp b/ngraph/core/include/ngraph/opsets/opset8_tbl.hpp
new file mode 100644
index 00000000000000..0dbe077ddae1bd
--- /dev/null
+++ b/ngraph/core/include/ngraph/opsets/opset8_tbl.hpp
@@ -0,0 +1,179 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#ifndef NGRAPH_OP
+#warning "NGRAPH_OP not defined"
+#define NGRAPH_OP(x, y)
+#endif
+
+NGRAPH_OP(Abs, ngraph::op::v0)
+NGRAPH_OP(Acos, ngraph::op::v0)
+NGRAPH_OP(Add, ngraph::op::v1)
+NGRAPH_OP(Asin, ngraph::op::v0)
+NGRAPH_OP(Atan, ngraph::op::v0)
+NGRAPH_OP(AvgPool, ngraph::op::v1)
+NGRAPH_OP(BatchNormInference, ngraph::op::v5)
+NGRAPH_OP(BinaryConvolution, ngraph::op::v1)
+NGRAPH_OP(Broadcast, ngraph::op::v3)
+NGRAPH_OP(Bucketize, ngraph::op::v3)
+NGRAPH_OP(CTCGreedyDecoder, ngraph::op::v0)
+NGRAPH_OP(Ceiling, ngraph::op::v0)
+NGRAPH_OP(Clamp, ngraph::op::v0)
+NGRAPH_OP(Concat, ngraph::op::v0)
+NGRAPH_OP(Constant, ngraph::op)
+NGRAPH_OP(Convert, ngraph::op::v0)
+NGRAPH_OP(ConvertLike, ngraph::op::v1)
+NGRAPH_OP(Convolution, ngraph::op::v1)
+NGRAPH_OP(ConvolutionBackpropData, ngraph::op::v1)
+NGRAPH_OP(Cos, ngraph::op::v0)
+NGRAPH_OP(Cosh, ngraph::op::v0)
+NGRAPH_OP(CumSum, ngraph::op::v0)
+NGRAPH_OP(DeformableConvolution, ngraph::op::v1)
+NGRAPH_OP(DeformablePSROIPooling, ngraph::op::v1)
+NGRAPH_OP(DepthToSpace, ngraph::op::v0)
+NGRAPH_OP(DetectionOutput, ngraph::op::v0)
+NGRAPH_OP(Divide, ngraph::op::v1)
+NGRAPH_OP(Elu, ngraph::op::v0)
+NGRAPH_OP(Erf, ngraph::op::v0)
+NGRAPH_OP(Equal, ngraph::op::v1)
+NGRAPH_OP(Exp, ngraph::op::v0)
+NGRAPH_OP(ExtractImagePatches, ngraph::op::v3)
+NGRAPH_OP(FakeQuantize, ngraph::op::v0)
+NGRAPH_OP(Floor, ngraph::op::v0)
+NGRAPH_OP(FloorMod, ngraph::op::v1)
+NGRAPH_OP(GatherTree, ngraph::op::v1)
+NGRAPH_OP(Greater, ngraph::op::v1)
+NGRAPH_OP(GreaterEqual, ngraph::op::v1)
+NGRAPH_OP(GroupConvolution, ngraph::op::v1)
+NGRAPH_OP(GroupConvolutionBackpropData, ngraph::op::v1)
+NGRAPH_OP(GRN, ngraph::op::v0)
+NGRAPH_OP(HardSigmoid, ngraph::op::v0)
+NGRAPH_OP(Less, ngraph::op::v1)
+NGRAPH_OP(LessEqual, ngraph::op::v1)
+NGRAPH_OP(Log, ngraph::op::v0)
+NGRAPH_OP(LogicalAnd, ngraph::op::v1)
+NGRAPH_OP(LogicalNot, ngraph::op::v1)
+NGRAPH_OP(LogicalOr, ngraph::op::v1)
+NGRAPH_OP(LogicalXor, ngraph::op::v1)
+NGRAPH_OP(LRN, ngraph::op::v0)
+NGRAPH_OP(LSTMCell, ngraph::op::v4)
+NGRAPH_OP(MatMul, ngraph::op::v0)
+NGRAPH_OP(MaxPool, ngraph::op::v1)
+NGRAPH_OP(Maximum, ngraph::op::v1)
+NGRAPH_OP(Minimum, ngraph::op::v1)
+NGRAPH_OP(Mod, ngraph::op::v1)
+NGRAPH_OP(Multiply, ngraph::op::v1)
+NGRAPH_OP(Negative, ngraph::op::v0)
+NGRAPH_OP(NormalizeL2, ngraph::op::v0)
+NGRAPH_OP(NotEqual, ngraph::op::v1)
+NGRAPH_OP(OneHot, ngraph::op::v1)
+NGRAPH_OP(PRelu, ngraph::op::v0)
+NGRAPH_OP(PSROIPooling, ngraph::op::v0)
+NGRAPH_OP(Pad, ngraph::op::v1)
+NGRAPH_OP(Parameter, ngraph::op::v0)
+NGRAPH_OP(Power, ngraph::op::v1)
+NGRAPH_OP(PriorBox, ngraph::op::v0)
+NGRAPH_OP(PriorBoxClustered, ngraph::op::v0)
+NGRAPH_OP(Proposal, ngraph::op::v4)
+NGRAPH_OP(Range, ngraph::op::v4)
+NGRAPH_OP(Relu, ngraph::op::v0)
+NGRAPH_OP(ReduceMax, ngraph::op::v1)
+NGRAPH_OP(ReduceLogicalAnd, ngraph::op::v1)
+NGRAPH_OP(ReduceLogicalOr, ngraph::op::v1)
+NGRAPH_OP(ReduceMean, ngraph::op::v1)
+NGRAPH_OP(ReduceMin, ngraph::op::v1)
+NGRAPH_OP(ReduceProd, ngraph::op::v1)
+NGRAPH_OP(ReduceSum, ngraph::op::v1)
+NGRAPH_OP(RegionYolo, ngraph::op::v0)
+NGRAPH_OP(ReorgYolo, ngraph::op::v0)
+NGRAPH_OP(Reshape, ngraph::op::v1)
+NGRAPH_OP(Result, ngraph::op::v0)
+NGRAPH_OP(ReverseSequence, ngraph::op::v0)
+NGRAPH_OP(ROIPooling, ngraph::op::v0)
+NGRAPH_OP(ScatterNDUpdate, ngraph::op::v3)
+NGRAPH_OP(Select, ngraph::op::v1)
+NGRAPH_OP(Selu, ngraph::op::v0)
+NGRAPH_OP(Sign, ngraph::op::v0)
+NGRAPH_OP(Sigmoid, ngraph::op::v0)
+NGRAPH_OP(Sin, ngraph::op::v0)
+NGRAPH_OP(Sinh, ngraph::op::v0)
+NGRAPH_OP(Softmax, ngraph::op::v1)
+NGRAPH_OP(Sqrt, ngraph::op::v0)
+NGRAPH_OP(SpaceToDepth, ngraph::op::v0)
+NGRAPH_OP(Split, ngraph::op::v1)
+NGRAPH_OP(SquaredDifference, ngraph::op::v0)
+NGRAPH_OP(Squeeze, ngraph::op::v0)
+NGRAPH_OP(StridedSlice, ngraph::op::v1)
+NGRAPH_OP(Subtract, ngraph::op::v1)
+NGRAPH_OP(Tan, ngraph::op::v0)
+NGRAPH_OP(Tanh, ngraph::op::v0)
+NGRAPH_OP(TensorIterator, ngraph::op::v0)
+NGRAPH_OP(Tile, ngraph::op::v0)
+NGRAPH_OP(Transpose, ngraph::op::v1)
+NGRAPH_OP(Unsqueeze, ngraph::op::v0)
+NGRAPH_OP(VariadicSplit, ngraph::op::v1)
+
+// New operations added in opset2
+NGRAPH_OP(BatchToSpace, ngraph::op::v1)
+NGRAPH_OP(SpaceToBatch, ngraph::op::v1)
+
+// New operations added in opset3
+NGRAPH_OP(EmbeddingBagPackedSum, ngraph::op::v3)
+NGRAPH_OP(EmbeddingSegmentsSum, ngraph::op::v3)
+NGRAPH_OP(EmbeddingBagOffsetsSum, ngraph::op::v3)
+NGRAPH_OP(GRUCell, ngraph::op::v3)
+NGRAPH_OP(NonZero, ngraph::op::v3)
+NGRAPH_OP(RNNCell, ngraph::op::v0)
+NGRAPH_OP(ROIAlign, ngraph::op::v3)
+NGRAPH_OP(ScatterElementsUpdate, ngraph::op::v3)
+NGRAPH_OP(ScatterUpdate, ngraph::op::v3)
+NGRAPH_OP(ShuffleChannels, ngraph::op::v0)
+NGRAPH_OP(ShapeOf, ngraph::op::v3)
+NGRAPH_OP(TopK, ngraph::op::v3)
+
+// New operations added in opset4
+NGRAPH_OP(Acosh, ngraph::op::v3)
+NGRAPH_OP(Asinh, ngraph::op::v3)
+NGRAPH_OP(Atanh, ngraph::op::v3)
+NGRAPH_OP(CTCLoss, ngraph::op::v4)
+NGRAPH_OP(HSwish, ngraph::op::v4)
+NGRAPH_OP(Interpolate, ngraph::op::v4)
+NGRAPH_OP(Mish, ngraph::op::v4)
+NGRAPH_OP(ReduceL1, ngraph::op::v4)
+NGRAPH_OP(ReduceL2, ngraph::op::v4)
+NGRAPH_OP(SoftPlus, ngraph::op::v4)
+NGRAPH_OP(Swish, ngraph::op::v4)
+
+// New operations added in opset5
+NGRAPH_OP(GatherND, ngraph::op::v5)
+NGRAPH_OP(GRUSequence, ngraph::op::v5)
+NGRAPH_OP(HSigmoid, ngraph::op::v5)
+NGRAPH_OP(LogSoftmax, ngraph::op::v5)
+NGRAPH_OP(Loop, ngraph::op::v5)
+NGRAPH_OP(LSTMSequence, ngraph::op::v5)
+NGRAPH_OP(NonMaxSuppression, ngraph::op::v5)
+NGRAPH_OP(RNNSequence, ngraph::op::v5)
+NGRAPH_OP(Round, ngraph::op::v5)
+
+// New operations added in opset6
+NGRAPH_OP(CTCGreedyDecoderSeqLen, ngraph::op::v6)
+NGRAPH_OP(ExperimentalDetectronDetectionOutput, ngraph::op::v6)
+NGRAPH_OP(ExperimentalDetectronGenerateProposalsSingleImage, ngraph::op::v6)
+NGRAPH_OP(ExperimentalDetectronPriorGridGenerator, ngraph::op::v6)
+NGRAPH_OP(ExperimentalDetectronROIFeatureExtractor, ngraph::op::v6)
+NGRAPH_OP(ExperimentalDetectronTopKROIs, ngraph::op::v6)
+NGRAPH_OP(GatherElements, ngraph::op::v6)
+NGRAPH_OP(MVN, ngraph::op::v6)
+NGRAPH_OP(Assign, ngraph::op::v6)    // new version
+NGRAPH_OP(ReadValue, ngraph::op::v6) // new version
+
+// New operations added in opset7
+NGRAPH_OP(DFT, ngraph::op::v7)
+NGRAPH_OP(Einsum, ngraph::op::v7)
+NGRAPH_OP(Gelu, ngraph::op::v7)
+NGRAPH_OP(IDFT, ngraph::op::v7)
+NGRAPH_OP(Roll, ngraph::op::v7)
+
+// New operations added in opset8
+NGRAPH_OP(Gather, ngraph::op::v8)
diff --git a/ngraph/core/src/op/gather.cpp b/ngraph/core/src/op/gather.cpp
index f2ddc9948b3241..aeacffc271995a 100644
--- a/ngraph/core/src/op/gather.cpp
+++ b/ngraph/core/src/op/gather.cpp
@@ -86,3 +86,50 @@ shared_ptr<Node> op::v7::Gather::clone_with_new_inputs(const OutputVector& new_a
     check_new_args_count(this, new_args);
     return make_shared<v7::Gather>(new_args.at(0), new_args.at(1), new_args.at(2), m_batch_dims);
 }
+
+NGRAPH_RTTI_DEFINITION(op::v8::Gather, "Gather", 8, op::util::GatherBase);
+
+op::v8::Gather::Gather(const Output<Node>& data,
+                       const Output<Node>& indices,
+                       const Output<Node>& axis,
+                       const int64_t batch_dims)
+    : GatherBase(data, indices, axis, batch_dims)
+{
+    constructor_validate_and_infer_types();
+}
+
+void op::v8::Gather::validate_and_infer_types()
+{
+    NGRAPH_OP_SCOPE(v8_Gather_validate_and_infer_types);
+    NODE_VALIDATION_CHECK(this,
+                          get_input_element_type(1).is_integral_number(),
+                          "Indices element type must be of an integral number type.");
+
+    NODE_VALIDATION_CHECK(this,
+                          get_input_element_type(2).is_integral_number(),
+                          "Axis element type must be of an integral number type.");
+
+    op::util::GatherBase::validate_and_infer_types();
+}
+
+int64_t op::v8::Gather::get_batch_dims() const
+{
+    if (m_batch_dims < 0 && get_input_partial_shape(1).rank().is_static())
+        return m_batch_dims + get_input_partial_shape(1).rank().get_length();
+    else
+        return m_batch_dims;
+}
+
+bool ngraph::op::v8::Gather::visit_attributes(AttributeVisitor& visitor)
+{
+    NGRAPH_OP_SCOPE(v8_Gather_visit_attributes);
+    visitor.on_attribute("batch_dims", m_batch_dims);
+    return true;
+}
+
+shared_ptr<Node> op::v8::Gather::clone_with_new_inputs(const OutputVector& new_args) const
+{
+    NGRAPH_OP_SCOPE(v8_Gather_clone_with_new_inputs);
+    check_new_args_count(this, new_args);
+    return make_shared<v8::Gather>(new_args.at(0), new_args.at(1), new_args.at(2), m_batch_dims);
+}
diff --git a/ngraph/test/type_prop/gather.cpp b/ngraph/test/type_prop/gather.cpp
index 70bfc8530ee1ff..96b2bbff832736 100644
--- a/ngraph/test/type_prop/gather.cpp
+++ b/ngraph/test/type_prop/gather.cpp
@@ -13,7 +13,7 @@ using namespace ngraph;
 
 // ------------------------------ V1 ------------------------------
 
-TEST(type_prop, gather_axis_0)
+TEST(type_prop, gather_v1_axis_0)
 {
     Shape params_shape{3, 2};
     Shape indices_shape{2, 2};
@@ -27,7 +27,7 @@ TEST(type_prop, gather_axis_0)
     ASSERT_EQ(G->get_axis(), 0);
 }
 
-TEST(type_prop, gather_7_uint8)
+TEST(type_prop, gather_v1_uint8)
 {
     // Gather_1 must allow even if indices is not int32/int64
     PartialShape data_shape{3, 2};
@@ -44,7 +44,7 @@ TEST(type_prop, gather_7_uint8)
     ASSERT_EQ(G->get_axis(), 0);
 }
 
-TEST(type_prop, gather_7_float32)
+TEST(type_prop, gather_v1_float32)
 {
     // Gather_1 should allow non int32/int64 indices
     PartialShape data_shape{3, 2};
@@ -335,7 +335,7 @@ TEST(type_prop, gather_7_axis_not_set_positive_batch_dims)
     ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
 }
 
-// --------------------- Negative tests ------------------------------
+// --------------------- V7 Negative tests ------------------------------
 
 TEST(type_prop, gather_7_incorrect_axis_shape)
 {
@@ -470,8 +470,7 @@ TEST(type_prop, gather_7_batch_dims_less_indices_rank_check)
     }
 }
 
-// disabled until decision of type constrains for gather
-TEST(type_prop, DISABLED_gather_7_indices_type_check)
+TEST(type_prop, gather_7_indices_type_check)
 {
     PartialShape data_shape{1, 20, 20, 22, 22};
     PartialShape indices_shape{1, 3};
@@ -500,8 +499,7 @@ TEST(type_prop, DISABLED_gather_7_indices_type_check)
     }
 }
 
-// disabled until decision of type constrains for gather
-TEST(type_prop, DISABLED_gather_7_axis_type_check)
+TEST(type_prop, gather_7_axis_type_check)
 {
     PartialShape data_shape{1, 20, 20, 22, 22};
     PartialShape indices_shape{1, 3};
@@ -529,3 +527,402 @@ TEST(type_prop, DISABLED_gather_7_axis_type_check)
         FAIL() << "Deduced type check failed for unexpected reason";
     }
 }
+
+// ------------------------------ V8 ------------------------------
+
+TEST(type_prop, gather_v8_axis_0)
+{
+    PartialShape data_shape{3, 2};
+    PartialShape indices_shape{2, 2};
+    PartialShape out_shape{2, 2, 2};
+    int64_t batch_dims = 0;
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i32, indices_shape);
+    auto A = op::Constant::create(element::i64, Shape{}, {0});
+    auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+    ASSERT_EQ(G->get_axis(), 0);
+}
+
+TEST(type_prop, gather_v8_axis_1)
+{
+    PartialShape data_shape{3, 3};
+    PartialShape indices_shape{1, 2};
+    PartialShape out_shape{3, 1, 2};
+    int64_t axis = 1;
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i32, indices_shape);
+    auto A = op::Constant::create(element::i64, Shape{}, {axis});
+    auto G = make_shared<op::v8::Gather>(D, I, A);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+    ASSERT_EQ(G->get_axis(), 1);
+}
+
+TEST(type_prop, gather_v8_negative_axis)
+{
+    PartialShape data_shape{5, 6, 7};
+    PartialShape indices_shape{4};
+    PartialShape out_shape{5, 4, 7};
+    int64_t axis = -2;
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    auto G = make_shared<op::v8::Gather>(D, I, A);
+
+    ASSERT_EQ(G->get_axis(), 1);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+}
+
+TEST(type_prop, gather_v8_dynamic_pshape_batch_dims_1_axis_1)
+{
+    PartialShape data_shape{Dimension(1, 7), 20, 20};
+    PartialShape indices_shape{Dimension(7, 10), 3, 8};
+    PartialShape out_shape{7, 3, 8, 20};
+    int64_t axis = 1;
+    int64_t batch_dims = 1;
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+}
+
+TEST(type_prop, gather_v8_dynamic_pshape_batch_dims_1_axis_3)
+{
+    PartialShape data_shape{Dimension(1, 7), Dimension(1, 3), 200, 400};
+    PartialShape indices_shape{Dimension(7, 10), Dimension(2, 10), 3, 8};
+    PartialShape out_shape{7, Dimension(1, 3), 200, Dimension(2, 10), 3, 8};
+    int64_t axis = 3;
+    int64_t batch_dims = 1;
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+}
+
+TEST(type_prop, gather_v8_dynamic_2d_pshape_batch_dim)
+{
+    PartialShape data_shape{Dimension(1, 7), Dimension(1, 3), 200, 400};
+    PartialShape indices_shape{Dimension(7, 10), Dimension(2, 10), 3, 8};
+    PartialShape out_shape{7, Dimension(2, 3), 3, 8, 400};
+    int64_t axis = 2;
+    int64_t batch_dims = 2;
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+}
+
+TEST(type_prop, gather_v8_dynamic_2d_pshape_batch_dim_axis_3)
+{
+    PartialShape data_shape{Dimension(1, 7), Dimension(1, 3), 200, 400};
+    PartialShape indices_shape{Dimension(7, 10), Dimension(2, 10), 3, 8};
+    PartialShape out_shape{7, Dimension(2, 3), 200, 3, 8};
+    int64_t axis = 3;
+    int64_t batch_dims = 2;
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+}
+
+TEST(type_prop, gather_v8_dynamic_rank)
+{
+    PartialShape data_shape{Dimension(1, 7), Dimension(1, 3), 200, 400};
+    PartialShape indices_shape = PartialShape::dynamic(Rank(3, 5));
+    PartialShape out_shape = PartialShape::dynamic(Rank(4, 6));
+    int64_t axis = 3;
+    int64_t batch_dims = 2;
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+}
+
+TEST(type_prop, gather_v8_axis_boundcheck_for_dynamic_data_rank)
+{
+    PartialShape data_shape = PartialShape::dynamic();
+    PartialShape indices_shape{7, 3, 8};
+    PartialShape out_shape = PartialShape::dynamic();
+    int64_t axis = 3;
+    int64_t batch_dims = 2;
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+}
+
+TEST(type_prop, gather_v8_dynamic_rank_negative_batch_dims)
+{
+    PartialShape data_shape{Dimension(1, 7), Dimension(1, 3), 200, 400};
+    PartialShape indices_shape = PartialShape::dynamic(Rank(3, 5));
+    PartialShape out_shape = PartialShape::dynamic(Rank(3, 5));
+    int64_t axis = 3;
+    int64_t batch_dims = -2;
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+}
+
+TEST(type_prop, gather_v8_axis_not_set)
+{
+    PartialShape data_shape{1, 1, 200, 400};
+    PartialShape indices_shape{2, 2};
+    // default batch_dims = 0
+    PartialShape out_shape = PartialShape::dynamic(5);  // out_rank = data_rank + indices_rank - 1 - batch_dims
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    auto A = make_shared<op::Parameter>(element::i32, Shape{1});
+    auto G = make_shared<op::v8::Gather>(D, I, A);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+}
+
+TEST(type_prop, gather_v8_axis_not_set_positive_batch_dims)
+{
+    PartialShape data_shape{2, 1, 200, 400};
+    PartialShape indices_shape{2, 2};
+    int64_t batch_dims = 1;
+    PartialShape out_shape = PartialShape({2,
+                                           Dimension::dynamic(),
+                                           Dimension::dynamic(),
+                                           Dimension::dynamic()});
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    auto A = make_shared<op::Parameter>(element::i32, Shape{1});
+    auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+
+    ASSERT_EQ(G->get_element_type(), element::f32);
+    ASSERT_EQ(G->get_output_partial_shape(0), out_shape);
+}
+
+// --------------------- V8 Negative tests ------------------------------
+
+TEST(type_prop, gather_v8_incorrect_axis_shape)
+{
+    auto D = make_shared<op::Parameter>(element::f32, Shape{5, 6});
+    auto I = make_shared<op::Parameter>(element::i64, Shape{4});
+    auto A = make_shared<op::Parameter>(element::i64, Shape{2});
+
+    try
+    {
+        auto G = make_shared<op::v8::Gather>(D, I, A);
+        // Should have thrown, so fail if it didn't
+        FAIL() << "Incorrect A input shape";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(),
+                             std::string("Axis input must be scalar or have 1 element"));
+    }
+    catch (...)
+    {
+        FAIL() << "Deduced type check failed for unexpected reason";
+    }
+}
+
+TEST(type_prop, gather_v8_axis_out_of_input_rank)
+{
+    auto D = make_shared<op::Parameter>(element::f32, Shape{5, 6});
+    auto I = make_shared<op::Parameter>(element::i64, Shape{4});
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{2});
+    int64_t batch_dims = 0;
+    try
+    {
+        auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+        // Should have thrown, so fail if it didn't
+        FAIL() << "axis check failed";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(
+                error.what(), std::string("Normalized axis must be >= 0 and < data_rank. But instead got"));
+    }
+    catch (...)
+    {
+        FAIL() << "Deduced type check failed for unexpected reason";
+    }
+}
+
+TEST(type_prop, gather_v8_dynamic_batch_dims_inconsistent)
+{
+    PartialShape data_shape{Dimension(1, 7), 20, 20};
+    PartialShape indices_shape{Dimension(8, 10), 3, 8};
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    int64_t axis = 1;
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    int64_t batch_dims = 1;
+
+    try
+    {
+        auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+        // Should have thrown, so fail if it didn't
+        FAIL() << "Shape inconsistency check for dynamic PartialShape failed";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(
+                error.what(),
+                std::string("data and indices must have equal or intersecting sizes until batch_dims"));
+    }
+    catch (...)
+    {
+        FAIL() << "Deduced type check failed for unexpected reason";
+    }
+}
+
+TEST(type_prop, gather_v8_batch_dims_less_check)
+{
+    PartialShape data_shape{1, 3, 20};
+    PartialShape indices_shape{1, 3, 8};
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    int64_t axis = 1;
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    int64_t batch_dims = 2;
+
+    try
+    {
+        auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+        // Should have thrown, so fail if it didn't
+        FAIL() << "batch_dims check failed";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(
+                error.what(),
+                std::string("After normalization batch_dims must be <= axis. But instead got: batch_dims ="));
+    }
+    catch (...)
+    {
+        FAIL() << "Deduced type check failed for unexpected reason";
+    }
+}
+
+TEST(type_prop, gather_v8_batch_dims_less_indices_rank_check)
+{
+    PartialShape data_shape{1, 20, 20, 22, 22};
+    PartialShape indices_shape{1, 3};
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i64, indices_shape);
+    int64_t axis = 4;
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    int64_t batch_dims = 3;
+
+    try
+    {
+        auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+        // Should have thrown, so fail if it didn't
+        FAIL() << "batch_dims check failed";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(
+                error.what(),
+                std::string("batch_dims must be <= indices_rank"));
+    }
+    catch (...)
+    {
+        FAIL() << "Deduced type check failed for unexpected reason";
+    }
+}
+
+TEST(type_prop, gather_v8_indices_type_check)
+{
+    PartialShape data_shape{1, 20, 20, 22, 22};
+    PartialShape indices_shape{1, 3};
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::f32, indices_shape);
+    int64_t axis = 4;
+    auto A = make_shared<op::Constant>(element::i64, Shape{1}, vector<int64_t>{axis});
+    int64_t batch_dims = 0;
+
+    try
+    {
+        auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+        // Should have thrown, so fail if it didn't
+        FAIL() << "indices element_type check failed";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(
+                error.what(),
+                std::string("Indices element type must be of an integral number type"));
+    }
+    catch (...)
+    {
+        FAIL() << "Deduced type check failed for unexpected reason";
+    }
+}
+
+TEST(type_prop, gather_v8_axis_type_check)
+{
+    PartialShape data_shape{1, 20, 20, 22, 22};
+    PartialShape indices_shape{1, 3};
+
+    auto D = make_shared<op::Parameter>(element::f32, data_shape);
+    auto I = make_shared<op::Parameter>(element::i32, indices_shape);
+    int64_t axis = 4;
+    auto A = make_shared<op::Constant>(element::f32, Shape{1}, vector<int64_t>{axis});
+    int64_t batch_dims = 0;
+
+    try
+    {
+        auto G = make_shared<op::v8::Gather>(D, I, A, batch_dims);
+        // Should have thrown, so fail if it didn't
+        FAIL() << "axis element_type check failed";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(
+                error.what(),
+                std::string("Axis element type must be of an integral number type"));
+    }
+    catch (...)
+    {
+        FAIL() << "Deduced type check failed for unexpected reason";
+    }
+}
diff --git a/ngraph/test/visitors/op/gather.cpp b/ngraph/test/visitors/op/gather.cpp
index 3e6446a07b8e89..c162d8949ffa41 100644
--- a/ngraph/test/visitors/op/gather.cpp
+++ b/ngraph/test/visitors/op/gather.cpp
@@ -7,6 +7,7 @@
 #include "ngraph/ngraph.hpp"
 #include "ngraph/opsets/opset1.hpp"
 #include "ngraph/opsets/opset7.hpp"
+#include "ngraph/opsets/opset8.hpp"
 
 #include "util/visitor.hpp"
 
@@ -29,3 +30,18 @@ TEST(attributes, gather_v7_op)
 
     EXPECT_EQ(g_gather->get_batch_dims(), gather->get_batch_dims());
 }
+
+TEST(attributes, gather_v8_op)
+{
+    NodeBuilder::get_ops().register_factory<opset8::Gather>();
+    auto data = make_shared<opset1::Parameter>(element::i32, Shape{2, 3, 4});
+    auto indices = make_shared<opset1::Parameter>(element::i32, Shape{2});
+    auto axis = make_shared<opset1::Constant>(element::i32, Shape{}, 2);
+    int64_t batch_dims = 1;
+
+    auto gather = make_shared<opset8::Gather>(data, indices, axis, batch_dims);
+    NodeBuilder builder(gather);
+    auto g_gather = as_type_ptr<opset8::Gather>(builder.create());
+
+    EXPECT_EQ(g_gather->get_batch_dims(), gather->get_batch_dims());
+}

From cadaefea8c4e9441b9e4190f1fcb6adbf10e1f9d Mon Sep 17 00:00:00 2001
From: Maria Kaglinskaya <maria.kaglinskaya@intel.com>
Date: Thu, 17 Jun 2021 16:10:58 +0300
Subject: [PATCH 121/178] Added i8, u8 precisions in Gather constant folding
 (#6195)

---
 ngraph/core/src/op/util/gather_base.cpp | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/ngraph/core/src/op/util/gather_base.cpp b/ngraph/core/src/op/util/gather_base.cpp
index a58ab3e3851d71..566d9184e33dd4 100644
--- a/ngraph/core/src/op/util/gather_base.cpp
+++ b/ngraph/core/src/op/util/gather_base.cpp
@@ -236,6 +236,8 @@ namespace gather
         {
             NGRAPH_TYPE_CASE(evaluate_gather, i32, arg0, arg1, out, axis, batch_dims);
             NGRAPH_TYPE_CASE(evaluate_gather, i64, arg0, arg1, out, axis, batch_dims);
+            NGRAPH_TYPE_CASE(evaluate_gather, i8, arg0, arg1, out, axis, batch_dims);
+            NGRAPH_TYPE_CASE(evaluate_gather, u8, arg0, arg1, out, axis, batch_dims);
             NGRAPH_TYPE_CASE(evaluate_gather, u32, arg0, arg1, out, axis, batch_dims);
             NGRAPH_TYPE_CASE(evaluate_gather, u64, arg0, arg1, out, axis, batch_dims);
             NGRAPH_TYPE_CASE(evaluate_gather, f16, arg0, arg1, out, axis, batch_dims);

From 3063afdbfc41fe0d1f9e63f2eb13c6584db287bd Mon Sep 17 00:00:00 2001
From: Nadezhda Ageeva <nadezhda.ageeva@intel.com>
Date: Thu, 17 Jun 2021 16:47:59 +0300
Subject: [PATCH 122/178] [GNA]: enable ITT couners and conditional compilation
 for GNA plugin (#5998)

---
 .../src/gna_plugin/CMakeLists.txt             | 10 +++++--
 .../gna_plugin/frontend/model_quantizer.hpp   |  2 ++
 inference-engine/src/gna_plugin/gna_itt.hpp   | 21 ++++++++++++++
 .../src/gna_plugin/gna_plugin.cpp             |  8 ++++-
 .../gna_plugin/optimizer/gna_pass_manager.cpp | 29 +++++++++++++++++++
 ...onvert_matmul_to_pointwise_convolution.cpp | 10 +++++--
 ...transpose_after_convolution_or_pooling.cpp |  2 ++
 .../insert_transpose_before_matmul.cpp        |  5 +++-
 .../transformations/remove_extra_reshapes.cpp |  5 +++-
 .../reorder_activation_and_pooling.cpp        |  5 +++-
 ...lit_convolution_with_large_buffer_size.cpp | 10 +++++--
 .../transformations/swap_input_matmul_gna.cpp |  5 +++-
 12 files changed, 98 insertions(+), 14 deletions(-)
 create mode 100644 inference-engine/src/gna_plugin/gna_itt.hpp

diff --git a/inference-engine/src/gna_plugin/CMakeLists.txt b/inference-engine/src/gna_plugin/CMakeLists.txt
index f3ce2858570733..36b9d6d5cc0b8e 100644
--- a/inference-engine/src/gna_plugin/CMakeLists.txt
+++ b/inference-engine/src/gna_plugin/CMakeLists.txt
@@ -29,12 +29,15 @@ endif()
 
 #
 # Shared plugin library
-# 
+#
 
 ie_add_plugin(NAME ${TARGET_NAME}
               DEVICE_NAME "GNA"
               SOURCES ${SOURCES} ${HEADERS})
 
+# Enable support of CC for the plugin
+ie_mark_target_as_cc(${TARGET_NAME})
+
 # saving rpath to GNA shared library be used by CI
 log_rpath_from_dir(GNA ${libGNA_LIBRARIES_BASE_PATH})
 
@@ -67,7 +70,8 @@ target_compile_definitions(${TARGET_NAME}_test_static
 
 target_link_libraries(${TARGET_NAME}_test_static PUBLIC inference_engine_preproc_s inference_engine_transformations libGNA::API)
 target_include_directories(${TARGET_NAME}_test_static PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}
-    $<TARGET_PROPERTY:inference_engine_legacy,INTERFACE_INCLUDE_DIRECTORIES>)
+    $<TARGET_PROPERTY:inference_engine_legacy,INTERFACE_INCLUDE_DIRECTORIES>
+    PRIVATE $<TARGET_PROPERTY:openvino::conditional_compilation,INTERFACE_INCLUDE_DIRECTORIES>)
 set_target_properties(${TARGET_NAME}_test_static PROPERTIES COMPILE_PDB_NAME ${TARGET_NAME}_test_static)
 
 set_target_properties(${TARGET_NAME} ${TARGET_NAME}_test_static
@@ -76,6 +80,6 @@ set_target_properties(${TARGET_NAME} ${TARGET_NAME}_test_static
 # install
 
 file(GLOB_RECURSE source_list "${libGNA_LIBRARIES_BASE_PATH}/*${CMAKE_SHARED_LIBRARY_SUFFIX}*")
-install(FILES ${source_list} 
+install(FILES ${source_list}
         DESTINATION ${IE_CPACK_IE_DIR}/external/gna/lib
         COMPONENT gna)
diff --git a/inference-engine/src/gna_plugin/frontend/model_quantizer.hpp b/inference-engine/src/gna_plugin/frontend/model_quantizer.hpp
index 1f3f125a029172..b57813858ac69a 100644
--- a/inference-engine/src/gna_plugin/frontend/model_quantizer.hpp
+++ b/inference-engine/src/gna_plugin/frontend/model_quantizer.hpp
@@ -15,6 +15,7 @@
 #include "layer_quantizer.hpp"
 #include "scale_factor_calc.hpp"
 #include "weights_converter.hpp"
+#include "gna_itt.hpp"
 
 namespace GNAPluginNS {
 
@@ -40,6 +41,7 @@ class ModelQuantizer {
 
     template <class PreQuantisationCb>
     InferenceEngine::CNNNetwork quantize(const InferenceEngine::CNNNetwork &model, const PreQuantisationCb &cb, std::vector<float> scaleFactor) const {
+        OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "ModelQuantizer::quantize");
         auto visitor = [&](InferenceEngine::CNNLayerPtr lp) {
             auto newLayer = InferenceEngine::injectData<QuantizedLayerParams>(lp);
             transformLayer(newLayer, WeightsConverter());
diff --git a/inference-engine/src/gna_plugin/gna_itt.hpp b/inference-engine/src/gna_plugin/gna_itt.hpp
new file mode 100644
index 00000000000000..3fa0211973329e
--- /dev/null
+++ b/inference-engine/src/gna_plugin/gna_itt.hpp
@@ -0,0 +1,21 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+/**
+ * @brief Defines openvino domains for tracing
+ * @file gna_itt.hpp
+ */
+
+#pragma once
+
+#include <openvino/itt.hpp>
+
+namespace GNAPluginNS {
+namespace itt {
+namespace domains {
+    OV_ITT_DOMAIN(GNAPlugin);
+    OV_ITT_DOMAIN(GNA_LT);
+}
+}
+}
diff --git a/inference-engine/src/gna_plugin/gna_plugin.cpp b/inference-engine/src/gna_plugin/gna_plugin.cpp
index f49d543def1f37..cb227304649638 100644
--- a/inference-engine/src/gna_plugin/gna_plugin.cpp
+++ b/inference-engine/src/gna_plugin/gna_plugin.cpp
@@ -37,7 +37,7 @@
 #include <layers/gna_fake_quantize_layer.hpp>
 #include "gna_graph_patterns.hpp"
 #include "gna_tensor_tools.hpp"
-#include <debug.h>
+#include "gna_itt.hpp"
 
 #include <ngraph/pass/manager.hpp>
 #include <legacy/convert_function_to_cnn_network.hpp>
@@ -391,6 +391,7 @@ GNAPlugin::GNAPlugin(const std::map<std::string, std::string>& configMap) {
 }
 
 void GNAPlugin::Init() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNAPlugin, "Init");
     dnn = std::make_shared<backend::AMIntelDNN>(backend::AMIntelDNN());
     inputsDesc = std::make_shared<GNAPluginNS::InputDesc>(GNAPluginNS::InputDesc());
     gnaFlags = std::make_shared<GNAPluginNS::GNAFlags>(GNAPluginNS::GNAFlags());
@@ -401,6 +402,7 @@ void GNAPlugin::Init() {
 }
 
 void GNAPlugin::InitGNADevice() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "InitGNADevice");
 #if GNA_LIB_VER == 1
     gnadevice = std::make_shared<GNADeviceHelper>(gnaFlags->gna_lib_async_threads_num,
                                                   gnaFlags->gna_openmp_multithreading,
@@ -419,6 +421,7 @@ void GNAPlugin::InitGNADevice() {
 }
 
 void GNAPlugin::UpdateGnaQuantModeFromNetwork(InferenceEngine::CNNNetwork & network) {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "UpdateGnaQuantModeFromNetwork");
     // fp32 emulation mode dont need any modifications to configuration
     if (config.gnaFlags.sw_fp32) return;
 
@@ -454,6 +457,7 @@ void GNAPlugin::UpdateGnaQuantModeFromNetwork(InferenceEngine::CNNNetwork & netw
 }
 
 void GNAPlugin::UpdateInputScaleFromNetwork(InferenceEngine::CNNNetwork & network) {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "UpdateInputScaleFromNetwork");
     // fp32 emulation mode dont need any modifications to configuration
     if (config.gnaFlags.sw_fp32) return;
 
@@ -561,6 +565,7 @@ bool GNAPlugin::TryToInitOutput(int portId, InferenceEngine::CNNLayerPtr layer)
 }
 
 void GNAPlugin::FillInputsAndOutputsTranspositionInfo(const InferenceEngine::CNNNetwork& net) {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "FillInputsAndOutputsTranspositionInfo");
     auto printTranspositionInfo = [](const std::vector<TranspositionInfo> &transpositionInfo) {
         for (const auto &transpositionInfoPart : transpositionInfo) {
             gnalog() << "transpose=" << transpositionInfoPart.transpose << " rows_num=" << transpositionInfoPart.num_transpose_rows
@@ -663,6 +668,7 @@ void GNAPlugin::AddDebugProperties(const InferenceEngine::CNNLayerPtr layer,
 #endif
 
 void GNAPlugin::LoadNetwork(CNNNetwork & _network) {
+    OV_ITT_SCOPED_TASK(itt::domains::GNAPlugin, "LoadNetwork");
     std::shared_ptr<InferenceEngine::details::CNNNetworkImpl> convertedNetwork;
     if (_network.getFunction()) {
         CNNNetwork clonedNetwork = InferenceEngine::cloneNetwork(_network);
diff --git a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
index ef333e7e46f115..5355a7b28f1277 100644
--- a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
+++ b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
@@ -41,6 +41,7 @@
 #include "gna_graph_patterns.hpp"
 #include "gna_data_types.hpp"
 #include "gna_tensor_tools.hpp"
+#include "gna_itt.hpp"
 
 using namespace InferenceEngine;
 using namespace InferenceEngine::details;
@@ -112,6 +113,7 @@ static void insertDiagonalLayerBetween(InferenceEngine::CNNLayerPtr prevLayer,
  */
 static CNNLayerPtr InsertCopyLayer(CNNLayerPtr prevLayer, CNNLayerPtr nextLayer, int beforeIdx,
                                    std::shared_ptr<IPassManager> passmanager,  std::string copyLayerType) {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "InsertCopyLayer");
     auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(prevLayer);
     std::string copyName = copyLayerType + std::string("_") + std::to_string(passmanager->getIntVar(copyLayersCounter)++);
     gnalog() << "Inserted " << copyName << " between: " << prevLayer->name << " and " << nextLayer->name << std::endl;
@@ -257,6 +259,7 @@ static std::vector<CNNLayerPtr> getCandidatesForIdentityInsertion(const CNNLayer
 }
 
 void InsertDiagonalLayerPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "InsertDiagonalLayerPass");
     bool lowPrecision = getPassManager()->isLowPrecision();
 
     for (auto & l : *pLayers) {
@@ -304,6 +307,7 @@ void InsertDiagonalLayerPass::run() {
 }
 
 void HandleMultipleActivationsForTheLayerPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "HandleMultipleActivationsForTheLayerPass");
     // found layer followed by multiple activations
     for (auto & l : *pLayers) {
         CNNLayerSet activations;
@@ -333,6 +337,7 @@ void HandleMultipleActivationsForTheLayerPass::run() {
 }
 
 void ForbidActivationFusingPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "ForbidActivationFusingPass");
     for (auto& l : *pLayers) {
         if (LayerInfo(l).isActivation()) {
             auto prevLayer = CNNNetPrevLayer(l);
@@ -370,6 +375,7 @@ namespace {
 } // namespace
 
 void ReorderMaxPoolPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "ReorderMaxPoolPass");
     // detecting following pattern
     // conv->activation->maxpooling
     // changing it to conv->maxpooling->activation
@@ -398,6 +404,7 @@ void ReorderMaxPoolPass::run() {
 }
 
 void SubstituteSoftSignPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "SubstituteSoftSignPass");
     //detecting following pattern
     // irv7 model:          irv10 model:
     // a layer                  a layer
@@ -501,6 +508,7 @@ void SubstituteSoftSignPass::run() {
     }
 }
 void SubstitutePReluPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "SubstitutePReluPass");
     auto getScale = [](CNNLayer* layer) {
         auto powerCandidate = LayerInfo(layer);
         if (!powerCandidate.isPower()) return 0.0f;
@@ -606,6 +614,7 @@ void SubstitutePReluPass::run() {
 }
 
 void ReversePermutationsPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "ReversePermutationsPass");
     std::function<CNNLayerPtr(CNNLayerPtr, std::function<bool(CNNLayerPtr)>)> prevLayerSkipCertain
         = [&prevLayerSkipCertain](CNNLayerPtr layer, std::function<bool(CNNLayerPtr)> shouldSkip) -> CNNLayerPtr {
         if (CNNNetHasPrevLayer(layer.get())) {
@@ -698,6 +707,7 @@ void ReversePermutationsPass::run() {
 }
 
 void RemovePermutationsNHWCToNCHWPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "RemovePermutationsNHWCToNCHWPass");
     std::set<CNNLayerPtr> permutations_to_remove;
     std::list<std::pair<CNNLayerPtr, CNNLayerPtr>> nhwc_layout_patterns;
     for (auto& l : *pLayers) {
@@ -781,6 +791,7 @@ void RemovePermutationsNHWCToNCHWPass::run() {
 }
 
 void InsertIdentityLayerPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "InsertIdentityLayerPass");
     auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(pLayers->front());
     auto createIdentityLayer = [quantized, this](const TensorDesc& tensorDesc) {
         int numOfIdentityLayers = this->getPassManager()->getIntVar(identityLayersCounterName)++;
@@ -898,6 +909,7 @@ void InsertIdentityLayerPass::run() {
 }
 
 void InsertCopyLayerPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "InsertCopyLayerPass");
     // Copy layer insertion happens in few cases:
     // Crop output goes to concat layer -> copy layer insertion
     // Splitted part of input goes to concat layer -> copy layer insertion
@@ -1020,6 +1032,7 @@ void InsertCopyLayerPass::run() {
 }
 
 void FlattenTrivialConcatPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "FlattenTrivialConcatPass");
     // change all trivial concatenations (concatenation where output buffer is a buffer made by appending input buffers)
     // by reshaping its inputs to 1 x total_input_size and its output to 1 x total_cocat_size and chaning the axis to 1
     // for example if 4D concat have unaligned inputs then ConcatAlignFilters need to be used if sizes before
@@ -1103,6 +1116,7 @@ void FlattenTrivialConcatPass::run() {
 }
 
 void InsertConcatAligningFilterPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "InsertConcatAligningFilterPass");
     auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(pLayers->front());
 
     if (getPassManager()->getPolicy().ConcatAlignmentPolicy == Policy::ConcatAlignment::DISABLED) {
@@ -1221,6 +1235,7 @@ void InsertConcatAligningFilterPass::run() {
 }
 
 void ReorderConcatInputsPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "ReorderConcatInputsPass");
     auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(pLayers->front());
     // aligning specific not required in fp32 mode
     if (getPassManager()->getPolicy().ConcatAlignmentPolicy == Policy::ConcatAlignment::DISABLED_FOR_FP32 && !quantized) {
@@ -1318,6 +1333,7 @@ void ReorderConcatInputsPass::run() {
 }
 
 void InsertSplitAligningFilterPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "InsertSplitAligningFilterPass");
     // currently split layer only supports 2 bytes in int16 and int8 mode. In fp32 mode this is not necessary but is useful for testing
     const int bytesPerSplitElement = 2;
     auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(pLayers->front());
@@ -1437,6 +1453,7 @@ static InferenceEngine::Blob::Ptr tileBlob(Blob::Ptr& blob, size_t TileTo) {
 }
 
 void EltwiseSplitOverChannelsPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "EltwiseSplitOverChannelsPass");
     if (getPassManager()->getPolicy().GNAAffineDiagonalPolicy.limitedTo == Policy::GNAAffineDiagonal::UNLIMIT) {
         return;
     }
@@ -1552,6 +1569,7 @@ void EltwiseSplitOverChannelsPass::run() {
 }
 
 void SubstituteScaleShiftBroadCastPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "SubstituteScaleShiftBroadCastPass");
     std::map<std::string, InferenceEngine::SizeVector> reshaped_data;
     auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(pLayers->front());
 
@@ -1633,6 +1651,7 @@ void SubstituteScaleShiftBroadCastPass::run() {
 }
 
 void BroadcastConstPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "BroadcastConstPass");
     for (auto constLayer : *pLayers) {
         if (!LayerInfo(constLayer).isConst()) {
             continue;
@@ -1685,6 +1704,7 @@ void BroadcastConstPass::run() {
 }
 
 void InsertIdentityToLSTMCellPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "InsertIdentityToLSTMCellPass");
     for (auto layer : *pLayers) {
         if (layer->type == "LSTMCell") {
             // This fixed the cases when both functional and non-functional outputs are mixed (or not outputs are used)
@@ -1722,6 +1742,7 @@ void InsertIdentityToLSTMCellPass::run() {
 }
 
 void BreakFusingOfOutputLayersPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "BreakFusingOfOutputLayersPass");
 #if GNA_LIB_VER == 1
     return;
 #endif
@@ -1765,6 +1786,7 @@ void BreakFusingOfOutputLayersPass::run() {
 }
 
 void UnrollLSTMCellPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "UnrollLSTMCellPass");
     InferenceEngine::NetPass::UnrollRNN_if(getPassManager()->getNetwork(), [] (const RNNCellBase& rnn) -> bool {
         if (rnn.clip != 0.0f)
             return true;
@@ -1781,6 +1803,7 @@ void UnrollLSTMCellPass::run() {
 }
 
 void UnrollTIPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "UnrollTIPass");
     auto sts = InferenceEngine::NetPass::UnrollTI(getPassManager()->getNetwork());
     if (!sts) {
         THROW_GNA_EXCEPTION << "TensorIterator layer cannot be unrolled!";
@@ -1788,6 +1811,7 @@ void UnrollTIPass::run() {
 }
 
 void RemoveConstPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "RemoveConstPass");
     auto network = getPassManager()->getNetwork();
     IE_SUPPRESS_DEPRECATED_START
     auto & icnnnet = static_cast<ICNNNetwork &>(network);
@@ -1801,6 +1825,7 @@ void RemoveConstPass::run() {
 }
 
 void RemoveSingleInputConcatPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "RemoveSingleInputConcatPass");
     for (auto &l : *pLayers) {
         if (l->type == "Concat") {
             auto concat = dynamic_cast<ConcatLayer*>(l.get());
@@ -1828,6 +1853,7 @@ void RemoveSingleInputConcatPass::run() {
 }
 
 void FuseMultipleIdentitiesPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "FuseMultipleIdentitiesPass");
     for (auto &l : *pLayers) {
         if (l->insData.empty()) continue;
 
@@ -1909,6 +1935,7 @@ void FuseMultipleIdentitiesPass::run() {
 }
 
 void FuseFQIntoWeightsPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "FuseFQIntoWeightsPass");
     auto isNonFunctional = [](CNNLayerPtr ptr) {
         return LayerInfo(ptr).isNonFunctional();
     };
@@ -2067,6 +2094,7 @@ void FuseFQIntoWeightsPass::run() {
 }
 
 void MoveFakeQuantizeLayerIntoQuantParamsPass :: run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "MoveFakeQuantizeLayerIntoQuantParamsPass");
     auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(pLayers->front());
     if (!quantized) {
         return;
@@ -2268,6 +2296,7 @@ void MoveFakeQuantizeLayerIntoQuantParamsPass :: run() {
 }
 
 void TransposeWeightsFromNCHWToNHWCPass::run() {
+    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "TransposeWeightsFromNCHWToNHWCPass");
     if (!MustBeConvertedFromNCHWToNHWC(*pLayers)) return;
 
     auto printTranspositionInfo = [](const std::vector<TranspositionInfo> &transpositionInfo) {
diff --git a/inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.cpp b/inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.cpp
index da7e6279624c81..e49d95ac2f2271 100644
--- a/inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.cpp
+++ b/inference-engine/src/gna_plugin/transformations/convert_matmul_to_pointwise_convolution.cpp
@@ -1,6 +1,7 @@
 // Copyright (C) 2021 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
+#include <openvino/cc/ngraph/itt.hpp>
 
 #include "transformations/convert_matmul_to_pointwise_convolution.hpp"
 
@@ -107,6 +108,7 @@ static bool Convert(std::shared_ptr<ngraph::Node> matmul_node,
 }
 
 ConvertMatmulToPointWiseConvolution::ConvertMatmulToPointWiseConvolution() {
+    MATCHER_SCOPE(ConvertMatmulToPointWiseConvolution);
     auto const_input = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
     auto const_fq = ngraph::pattern::wrap_type<ngraph::opset7::FakeQuantize>({const_input,
         ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
@@ -121,11 +123,12 @@ ConvertMatmulToPointWiseConvolution::ConvertMatmulToPointWiseConvolution() {
         return Convert(pattern_map.at(matmul).get_node_shared_ptr(), nullptr, nullptr, nullptr);
     };
 
-    auto m = std::make_shared<ngraph::pattern::Matcher>(matmul, "ConvertMatmulToPointWiseConvolution");
+    auto m = std::make_shared<ngraph::pattern::Matcher>(matmul, matcher_name);
     this->register_matcher(m, callback);
 }
 
 ConvertMatmulWithBiasToPointWiseConvolution::ConvertMatmulWithBiasToPointWiseConvolution() {
+    MATCHER_SCOPE(ConvertMatmulWithBiasToPointWiseConvolution);
     auto const_input = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
     auto const_fq = ngraph::pattern::wrap_type<ngraph::opset7::FakeQuantize>({const_input,
         ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
@@ -143,11 +146,12 @@ ConvertMatmulWithBiasToPointWiseConvolution::ConvertMatmulWithBiasToPointWiseCon
             pattern_map.at(bias).get_node_shared_ptr(), nullptr);
     };
 
-    auto m = std::make_shared<ngraph::pattern::Matcher>(add, "ConvertMatmulWithBiasToPointWiseConvolution");
+    auto m = std::make_shared<ngraph::pattern::Matcher>(add, matcher_name);
     this->register_matcher(m, callback);
 }
 
 ConvertMatmulWithFqToPointWiseConvolution::ConvertMatmulWithFqToPointWiseConvolution() {
+    MATCHER_SCOPE(ConvertMatmulWithFqToPointWiseConvolution);
     auto const_input = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
     auto const_fq = ngraph::pattern::wrap_type<ngraph::opset7::FakeQuantize>({const_input,
         ngraph::pattern::wrap_type<ngraph::opset7::Constant>(),
@@ -175,6 +179,6 @@ ConvertMatmulWithFqToPointWiseConvolution::ConvertMatmulWithFqToPointWiseConvolu
              pattern_map.at(out_fq).get_node_shared_ptr());
     };
 
-    auto m = std::make_shared<ngraph::pattern::Matcher>(out_fq, "ConvertMatmulWithFqToPointWiseConvolution");
+    auto m = std::make_shared<ngraph::pattern::Matcher>(out_fq, matcher_name);
     this->register_matcher(m, callback);
 }
\ No newline at end of file
diff --git a/inference-engine/src/gna_plugin/transformations/insert_transpose_after_convolution_or_pooling.cpp b/inference-engine/src/gna_plugin/transformations/insert_transpose_after_convolution_or_pooling.cpp
index 6bfef2587aeff2..4954529762d582 100644
--- a/inference-engine/src/gna_plugin/transformations/insert_transpose_after_convolution_or_pooling.cpp
+++ b/inference-engine/src/gna_plugin/transformations/insert_transpose_after_convolution_or_pooling.cpp
@@ -1,6 +1,7 @@
 // Copyright (C) 2021 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
+#include <openvino/cc/ngraph/itt.hpp>
 
 #include "transformations/insert_transpose_after_convolution_or_pooling.hpp"
 
@@ -16,6 +17,7 @@ using namespace GNAPluginNS;
 NGRAPH_RTTI_DEFINITION(InsertTransposeAfterConvOrPool, "InsertTransposeAfterConvOrPool", 0);
 
 bool InsertTransposeAfterConvOrPool::run_on_function(std::shared_ptr<ngraph::Function> f) {
+    RUN_ON_FUNCTION_SCOPE(InsertTransposeAfterConvOrPool);
     bool is_graph_modfied = false;
     for (auto& node : f->get_ordered_ops()) {
         if (std::dynamic_pointer_cast<ngraph::opset7::Convolution>(node) == nullptr &&
diff --git a/inference-engine/src/gna_plugin/transformations/insert_transpose_before_matmul.cpp b/inference-engine/src/gna_plugin/transformations/insert_transpose_before_matmul.cpp
index 4de8966d351660..3e5c579af8f14a 100644
--- a/inference-engine/src/gna_plugin/transformations/insert_transpose_before_matmul.cpp
+++ b/inference-engine/src/gna_plugin/transformations/insert_transpose_before_matmul.cpp
@@ -2,6 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+#include <openvino/cc/ngraph/itt.hpp>
+
 #include "transformations/insert_transpose_before_matmul.hpp"
 
 #include <ngraph/opsets/opset7.hpp>
@@ -13,6 +15,7 @@ using namespace GNAPluginNS;
 NGRAPH_RTTI_DEFINITION(InsertTransposeBeforeMatmul, "InsertTransposeBeforeMatmul", 0);
 
 InsertTransposeBeforeMatmul::InsertTransposeBeforeMatmul() {
+    MATCHER_SCOPE(InsertTransposeBeforeMatmul);
     auto reshape = ngraph::pattern::wrap_type<ngraph::opset7::Reshape>({ngraph::pattern::any_input(),
                                                                         ngraph::pattern::any_input()},
                                                                         ngraph::pattern::rank_equals(2));
@@ -59,6 +62,6 @@ InsertTransposeBeforeMatmul::InsertTransposeBeforeMatmul() {
         return true;
     };
 
-    auto m = std::make_shared<ngraph::pattern::Matcher>(root, "InsertTransposeBeforeMatmul");
+    auto m = std::make_shared<ngraph::pattern::Matcher>(root, matcher_name);
     this->register_matcher(m, callback);
 }
diff --git a/inference-engine/src/gna_plugin/transformations/remove_extra_reshapes.cpp b/inference-engine/src/gna_plugin/transformations/remove_extra_reshapes.cpp
index 1a7d6da2a33c8b..e1cfdefa31177c 100644
--- a/inference-engine/src/gna_plugin/transformations/remove_extra_reshapes.cpp
+++ b/inference-engine/src/gna_plugin/transformations/remove_extra_reshapes.cpp
@@ -2,6 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+#include <openvino/cc/ngraph/itt.hpp>
+
 #include "transformations/remove_extra_reshapes.hpp"
 
 #include <ngraph/opsets/opset7.hpp>
@@ -12,6 +14,7 @@ using namespace GNAPluginNS;
 NGRAPH_RTTI_DEFINITION(RemoveExtraReshapes, "RemoveExtraReshapes", 0);
 
 RemoveExtraReshapes::RemoveExtraReshapes() {
+    MATCHER_SCOPE(RemoveExtraReshapes);
     const auto reshape = ngraph::pattern::wrap_type<ngraph::opset7::Reshape>();
     const auto pooling = ngraph::pattern::wrap_type<ngraph::opset7::MaxPool>({reshape});
 
@@ -26,6 +29,6 @@ RemoveExtraReshapes::RemoveExtraReshapes() {
         return true;
     };
 
-    auto m = std::make_shared<ngraph::pattern::Matcher>(pooling, "RemoveExtraReshapes");
+    auto m = std::make_shared<ngraph::pattern::Matcher>(pooling, matcher_name);
     this->register_matcher(m, callback);
 }
diff --git a/inference-engine/src/gna_plugin/transformations/reorder_activation_and_pooling.cpp b/inference-engine/src/gna_plugin/transformations/reorder_activation_and_pooling.cpp
index 69bab295ba75e2..7e67d900e38423 100644
--- a/inference-engine/src/gna_plugin/transformations/reorder_activation_and_pooling.cpp
+++ b/inference-engine/src/gna_plugin/transformations/reorder_activation_and_pooling.cpp
@@ -2,6 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+#include <openvino/cc/ngraph/itt.hpp>
+
 #include "transformations/reorder_activation_and_pooling.hpp"
 
 #include <ngraph/opsets/opset7.hpp>
@@ -15,6 +17,7 @@ using namespace GNAPluginNS;
 NGRAPH_RTTI_DEFINITION(ReorderActivationAndPooling, "ReorderActivationAndPooling", 0);
 
 ReorderActivationAndPooling::ReorderActivationAndPooling() {
+    MATCHER_SCOPE(ReorderActivationAndPooling);
     auto conv = ngraph::pattern::wrap_type<ngraph::opset7::Convolution>({ngraph::pattern::any_input(),
                                                                          ngraph::pattern::any_input()});
     auto add = ngraph::pattern::wrap_type<ngraph::opset7::Add>({conv, ngraph::pattern::any_input()});
@@ -63,6 +66,6 @@ ReorderActivationAndPooling::ReorderActivationAndPooling() {
         return true;
     };
 
-    auto m = std::make_shared<ngraph::pattern::Matcher>(pool, "ReorderActivationAndPooling");
+    auto m = std::make_shared<ngraph::pattern::Matcher>(pool, matcher_name);
     this->register_matcher(m, callback);
 }
diff --git a/inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.cpp b/inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.cpp
index a9d79c831ab34d..2e750308e5f3d1 100644
--- a/inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.cpp
+++ b/inference-engine/src/gna_plugin/transformations/split_convolution_with_large_buffer_size.cpp
@@ -1,6 +1,7 @@
 // Copyright (C) 2021 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
+#include <openvino/cc/ngraph/itt.hpp>
 
 #include "transformations/split_convolution_with_large_buffer_size.hpp"
 
@@ -77,6 +78,7 @@ static bool Convert(std::shared_ptr<ngraph::Node> conv,
 }
 
 SplitConvolution::SplitConvolution() {
+    MATCHER_SCOPE(SplitConvolution);
     auto conv = ngraph::pattern::wrap_type<ngraph::opset7::Convolution>({ngraph::pattern::any_input(),
         ngraph::pattern::any_input()});
 
@@ -85,11 +87,12 @@ SplitConvolution::SplitConvolution() {
         return Convert(pattern_map.at(conv).get_node_shared_ptr(), nullptr, nullptr, nullptr);
     };
 
-    auto m = std::make_shared<ngraph::pattern::Matcher>(conv, "SplitConvolution");
+    auto m = std::make_shared<ngraph::pattern::Matcher>(conv, matcher_name);
     this->register_matcher(m, callback);
 }
 
 SplitConvolutionWithBias::SplitConvolutionWithBias() {
+    MATCHER_SCOPE(SplitConvolutionWithBias);
     auto conv = ngraph::pattern::wrap_type<ngraph::opset7::Convolution>({ngraph::pattern::any_input(),
         ngraph::pattern::any_input()});
     auto bias = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
@@ -101,11 +104,12 @@ SplitConvolutionWithBias::SplitConvolutionWithBias() {
             pattern_map.at(bias).get_node_shared_ptr(), nullptr);
     };
 
-    auto m = std::make_shared<ngraph::pattern::Matcher>(add, "SplitConvolutionWithBias");
+    auto m = std::make_shared<ngraph::pattern::Matcher>(add, matcher_name);
     this->register_matcher(m, callback);
 }
 
 SplitConvolutionWithFq::SplitConvolutionWithFq() {
+    MATCHER_SCOPE(SplitConvolutionWithFq);
     auto conv = ngraph::pattern::wrap_type<ngraph::opset7::Convolution>({ngraph::pattern::any_input(),
         ngraph::pattern::any_input()});
     auto bias = ngraph::pattern::wrap_type<ngraph::opset7::Constant>();
@@ -126,6 +130,6 @@ SplitConvolutionWithFq::SplitConvolutionWithFq() {
         return Convert(pattern_map.at(conv).get_node_shared_ptr(), add_node, bias_node, pattern_map.at(out_fq).get_node_shared_ptr());
     };
 
-    auto m = std::make_shared<ngraph::pattern::Matcher>(out_fq, "SplitConvolutionWithFq");
+    auto m = std::make_shared<ngraph::pattern::Matcher>(out_fq, matcher_name);
     this->register_matcher(m, callback);
 }
\ No newline at end of file
diff --git a/inference-engine/src/gna_plugin/transformations/swap_input_matmul_gna.cpp b/inference-engine/src/gna_plugin/transformations/swap_input_matmul_gna.cpp
index 9a725c33cf7151..d177b83ba40e60 100644
--- a/inference-engine/src/gna_plugin/transformations/swap_input_matmul_gna.cpp
+++ b/inference-engine/src/gna_plugin/transformations/swap_input_matmul_gna.cpp
@@ -2,6 +2,8 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+#include <openvino/cc/ngraph/itt.hpp>
+
 #include <memory>
 #include <vector>
 
@@ -19,6 +21,7 @@ using namespace GNAPluginNS;
 NGRAPH_RTTI_DEFINITION(SwapInputMatMul, "SwapInputMatMul", 0);
 
 SwapInputMatMul::SwapInputMatMul() {
+    MATCHER_SCOPE(SwapInputMatMul);
     auto matmul = ngraph::pattern::wrap_type<ngraph::opset7::MatMul>({ngraph::pattern::any_input(
             ngraph::pattern::has_static_shape()), ngraph::pattern::any_input(ngraph::pattern::has_static_shape())},
                                                                      ngraph::pattern::has_static_shape());
@@ -95,6 +98,6 @@ SwapInputMatMul::SwapInputMatMul() {
         return true;
     };
 
-    auto m = std::make_shared<ngraph::pattern::Matcher>(matmul, "SwapInputMatMul");
+    auto m = std::make_shared<ngraph::pattern::Matcher>(matmul, matcher_name);
     this->register_matcher(m, callback);
 }
\ No newline at end of file

From e61a594199ab222d8239635b1d651ce3f6bf11a4 Mon Sep 17 00:00:00 2001
From: "Gladilov, Gleb" <gleb.gladilov@intel.com>
Date: Thu, 17 Jun 2021 18:54:39 +0300
Subject: [PATCH 123/178] [IE][VPU]: Configuration options in VPU plugins
 refactoring (#3211)

* [IE]: Enables Abstract class -> Parameter conversion support

Parameter has templated constructor allowing to write code

```
Parameter p = i; // i of type int for example
```

This constructor uses SFINAE to resolve ambiguity with
move-constructor, so checks that argument is not of the same type.
In case it's not the same type it calls std::tuple constructors that
constructs an instance of argument type. In the following case:

```
Parameter p = static_cast<Parameter>(abstractRef);
// abstractRef is a reference to abstract class
```

We have a reference to some abstract class that defines explicit
cast operator to Parameter. In contrast with expectations,
instead of cast operator, Parameter constructor is instantiated,
since template type deduction for Parameter constructor didn't fail
(abstract class has not the same type as Parameter). Instantiation
of tuple constructor used inside failed: it's impossible to create an
instance of abstract class what lead to compile-time error. To resolve
the issue additional condition introduced to check if argument type is
abstract.

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE]: Enables PrintTo method for Parameter and tests on it

Inference Engine API for configuration options uses Parameter
type as a return type of GetConfig method. Parameter is intended
to store object associated with configuration option.
To support objects of different types its constructor is templated.
Parameter overloads cast operators which are templated
as well. Both constructor and cast operators are implicit, which
makes it possible to implicitly convert any type to Parameter
and vice versa.

Since Parameter is a part of Inference Engine configuration API it's
essential google tests on API contain Parameter as tests parameter.
For each test parameter Google Test framework tries to print it to
an output stream. For that purpose, Google Test checks if test
parameter has output stream operator or PrintTo method. If not, it
checks if it could be implicitly converted to integral type and,
in this case, prints it as a long integer.

InferenceEngine::Parameter does not define output stream operator,
but could be implicitly converted to an integer, according cast
operators mentioned above, so Google Test tries to convert to
integer. Since Parameter not necessarily contains integer, this
conversion throws an exception of type mismatch, which makes it
impossible to use Parameter in Google Test framework as is.

In order to resolve that issue Parameter should define either
output stream operator or PrintTo method. If Parameter will
define output stream operator it will make it possible to compile
streaming almost any object to an output stream. The reason for it
is C++ checks if object could be implicitly converted to other type
which defines output stream operator, if objects itself doesn't do it
(e.g. `stream << "text";` calls std::string::operator<<, since
char const* is implicitly convertible to std::string).

Taking this into consideration the only way to support Parameter in
Google Test without breaking backward compatibility is define PrintTo
method.

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE]: Fixes ill-formed extending std names

According to the standard:

The behavior of a C++ program is undefined if
it adds declarations or definitions to namespace
std or to a namespace within namespace std unless
otherwise specified. A program may add a template
specialization for any standard library template
to namespace std only if the declaration depends
on a user-defined type and the specialization meets
the standard library requirements for the original
template and is not explicitly prohibited.

As as an unexpected result, InferenceEngine::Parameter
that contains std::vector<std::string> can be printed
via PrintTo. In that case operator<< version from
Inference Engine is picked up.

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Moves CompilationConfig out of GT header

Keeping config in a separate header simplifies migration
to new interface.

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Removes Platform enum

Since there is enum from MVNC for the same purpose
there is no need in Platform anyway

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Introduces containers utility header

Contains some helpers to work with C++ maps

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Introduces new configuration API

The main ideas are separate option-specific logic
from common container, automate logic processing
public vs private, deprecated, compile-time vs
runtime-time options and remove code duplication.

Since IE defines configuration API using std::string
and Parameter, options have to provide ways to be
represented as Parameter (ex.: GetConfig is called)
and be defined using std::string (ex.: SetConfig is
called). Keeping information about actual key value
is useful for error reporting.

New API fallbacks to previous version in case of
unsupported options are requested. This way migration
becomes iterative and looks simpler.

Options containers are related to corresponding components:
CompilationConfig (name to be changed) - GraphTransformer,
PluginConfiguration - base class for plugins configurations,
MyriadConfiguration - Myriad plugin configuration,
HDDLConfiguration - HDDL plugin configuration (to be
introduced in a separate request)

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Replaces CompilationConfig with PluginConfiguration

Some of options to be refactored are stored inside CompilationConfig.
CompilationConfig is passed to graph transformer as a compiler to be
processed. Since it's separate data structure and migration process
is iterative we need a mechanism to provide some of compilation
options from new interface and some from old. It cannot be done via
plugin specific class (MyriadConfiguration), since there are others
plugins as graph transformer users. Plugin specific class
(MyriadConfiguration) already inherits from old version (MyriadConfig),
which in turn inherits from ParsedConfig containing CompilationConfig.

To resolve the issue MyriadConfig inheritance from ParsedConfig is made
virtual to make it possible for PluginConfiguration to virtually inherit
from ParsedConfig as well an so make PluginConfiguration data structure
for configuration options for graph transformer. Since
PluginConfiguration is base class of MyriadConfiguration as well as
MyriadConfig and inheritance is virtual plugin just casts its specific
configuration to base one passing to graph transformer.

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Enables new tests on configuration API

* Enables following new shared tests on configuration API
  * Can load network with empty configuration
  * Check default value for configuration option
  * Can load network with correct configuration
  * Check custom value for configuration option (set and compare)
  * Check public configuration options are visible through API
  * Check private configuration options are invisible through API
  * Check GetConfig throws an exception on incorrect key
* Refactors myriad plugin instantiations for shared tests

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Extracts LogLevel enum to a separate header

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Refactors LOG_LEVEL configuration option

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Refactors COPY_OPTIMIZATION configuration option

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Fixes behavior tests build

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Updates tests on new exception class

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Removes unused variable from mvnc test

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>

* [IE][VPU]: Removes SizeVector streaming call

New assertion macro IE_ASSERT implementation uses
output streaming operator with r-value reference
argument as a stream. This prevents the compiler
from picking up overload from InferenceEngine::details,
since our version takes stream by non-const l-value
reference.

Since there is no simple solution to provide output
streaming operator overload for r-value references as
well and this call is just a message for assert in
test utilities, it was decided just to remove call
for now.

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>
---
 inference-engine/include/ie_parameter.hpp     |  51 +-
 inference-engine/src/plugin_api/debug.h       |   7 +-
 .../vpu/common/include/vpu/configuration.hpp  |  99 ++++
 .../configuration/as_parameter_enabler.hpp    |  18 +
 .../options/copy_optimization.hpp             |  34 ++
 .../vpu/configuration/options/log_level.hpp   |  36 ++
 .../configuration/plugin_configuration.hpp    | 142 +++++
 .../vpu/configuration/switch_converters.hpp   |  15 +
 .../include/vpu/parsed_config.hpp             |  12 +-
 .../common/include/vpu/parsed_config_base.hpp |  11 +-
 .../include/vpu/private_plugin_config.hpp     |   0
 .../common/include/vpu/utils/containers.hpp   |  40 ++
 .../vpu/common/include/vpu/utils/error.hpp    |  14 +-
 .../common/include/vpu/utils/log_level.hpp    |  21 +
 .../vpu/common/include/vpu/utils/logger.hpp   |  15 +-
 .../configuration/as_parameter_enabler.cpp    |  10 +
 .../options/copy_optimization.cpp             |  45 ++
 .../src/configuration/options/log_level.cpp   |  64 +++
 .../configuration/plugin_configuration.cpp    | 114 ++++
 .../src/configuration/switch_converters.cpp   |  25 +
 .../src/parsed_config.cpp                     |   1 -
 .../src/vpu/common/src/parsed_config_base.cpp |  37 +-
 .../src/vpu/graph_transformer/CMakeLists.txt  |   9 +-
 .../include/vpu/compile_env.hpp               |  25 +-
 .../include/vpu/graph_transformer.hpp         | 110 +---
 .../vpu/graph_transformer_internal.hpp        |   4 +-
 .../graph_transformer/src/backend/backend.cpp |  10 +-
 .../src/frontend/detect_network_batch.cpp     |   2 +-
 .../src/frontend/frontend.cpp                 |  16 +-
 .../src/frontend/in_out_convert.cpp           |  18 +-
 .../src/frontend/parse_data.cpp               |   4 +-
 .../src/frontend/unroll_loops.cpp             |   6 +-
 .../src/graph_transformer.cpp                 |  71 +--
 .../src/middleend/pass_manager.cpp            |  33 +-
 .../middleend/passes/adjust_data_location.cpp |   4 +-
 .../src/middleend/passes/eliminate_copy.cpp   |   3 +-
 .../src/middleend/passes/inject_sw.cpp        |   2 +-
 .../passes/merge_eltwise_and_relu.cpp         |   2 +-
 .../src/middleend/passes/merge_hw_stages.cpp  |   2 +-
 .../passes/replace_deconv_by_conv.cpp         |   2 +-
 .../src/middleend/passes/weights_analysis.cpp |   4 +-
 .../vpu/graph_transformer/src/model/stage.cpp |   2 +-
 .../src/stages/activation.cpp                 |   2 +-
 .../src/stages/convolution.cpp                |  12 +-
 .../vpu/graph_transformer/src/stages/fc.cpp   |   4 +-
 .../graph_transformer/src/stages/mtcnn.cpp    |   2 +-
 .../graph_transformer/src/stages/permute.cpp  |   2 +-
 .../graph_transformer/src/stages/pooling.cpp  |   8 +-
 .../configuration/myriad_configuration.cpp    |  31 ++
 .../configuration/myriad_configuration.hpp    |  21 +
 .../src/vpu/myriad_plugin/myriad_config.h     |   2 +-
 .../myriad_executable_network.cpp             |  33 +-
 .../myriad_plugin/myriad_executable_network.h |  32 +-
 .../src/vpu/myriad_plugin/myriad_executor.cpp |   5 +-
 .../src/vpu/myriad_plugin/myriad_executor.h   |  10 +-
 .../vpu/myriad_plugin/myriad_infer_request.h  |   1 +
 .../src/vpu/myriad_plugin/myriad_metrics.cpp  |   2 +-
 .../src/vpu/myriad_plugin/myriad_plugin.cpp   |  70 ++-
 .../src/vpu/myriad_plugin/myriad_plugin.h     |   3 +-
 .../inference_engine/parameter_tests.cpp      | 105 +++-
 .../functional/plugin/myriad/CMakeLists.txt   |  13 +
 .../behavior/config.cpp                       | 496 +++++++++++-------
 .../include/base/behavior_test_utils.hpp      | 126 ++++-
 .../plugin/shared/include/behavior/config.hpp |  97 +++-
 .../common_test_utils/data_utils.cpp          |   6 +-
 .../unit/vpu/base/graph_transformer_tests.cpp |  23 +-
 .../unit/vpu/base/graph_transformer_tests.hpp |   6 +-
 .../tests/unit/vpu/blob_reader_tests.cpp      |   5 +-
 .../passes_tests/annotate_memory_types.cpp    |   2 +-
 .../plugin_tests/behavior_test_plugin.h       |   3 +-
 .../myriad_tests/helpers/myriad_devices.cpp   |   1 -
 .../myriad_tests/helpers/myriad_devices.hpp   |   5 +-
 .../helpers/myriad_protocol_case.cpp          |   1 +
 .../vpu/myriad_tests/vpu_watchdog_tests.cpp   |   1 +
 .../plugin_tests/vpu_test_data.hpp            |   1 +
 .../common/layers/myriad_layers_blob_test.cpp |  11 +-
 .../layers/myriad_layers_concat_test.hpp      |   1 +
 .../graph_transformer/gt_functional_tests.cpp |  16 +-
 .../graph_transformer/gt_functional_tests.hpp |  10 +-
 .../merge_permute_and_reorder_test.cpp        |   4 +-
 .../helpers/tests_vpu_common.hpp              |   2 +-
 .../tests_deprecated/unit/CMakeLists.txt      |   1 +
 .../engines/gna/layers/gna_squeeze_test.cpp   |   2 +
 .../vpu/adjust_data_location_tests.cpp        |   4 +-
 .../vpu/get_vpu_scale_from_ir_tests.cpp       |   8 +-
 .../engines/vpu/graph_transformer_tests.cpp   |  22 +-
 .../engines/vpu/graph_transformer_tests.hpp   |   6 +-
 .../vpu/replace_deconv_by_conv_tests.cpp      |   2 +-
 .../get_num_iterations_test.cpp               |   2 +
 .../thirdparty/movidius/XLink/CMakeLists.txt  |   3 +
 .../thirdparty/movidius/mvnc/CMakeLists.txt   |   3 +
 .../movidius/mvnc/tests/mvnc_tests_common.cpp |   1 -
 92 files changed, 1773 insertions(+), 596 deletions(-)
 create mode 100644 inference-engine/src/vpu/common/include/vpu/configuration.hpp
 create mode 100644 inference-engine/src/vpu/common/include/vpu/configuration/as_parameter_enabler.hpp
 create mode 100644 inference-engine/src/vpu/common/include/vpu/configuration/options/copy_optimization.hpp
 create mode 100644 inference-engine/src/vpu/common/include/vpu/configuration/options/log_level.hpp
 create mode 100644 inference-engine/src/vpu/common/include/vpu/configuration/plugin_configuration.hpp
 create mode 100644 inference-engine/src/vpu/common/include/vpu/configuration/switch_converters.hpp
 rename inference-engine/src/vpu/{graph_transformer => common}/include/vpu/parsed_config.hpp (80%)
 rename inference-engine/src/vpu/{graph_transformer => common}/include/vpu/private_plugin_config.hpp (100%)
 create mode 100644 inference-engine/src/vpu/common/include/vpu/utils/containers.hpp
 create mode 100644 inference-engine/src/vpu/common/include/vpu/utils/log_level.hpp
 create mode 100644 inference-engine/src/vpu/common/src/configuration/as_parameter_enabler.cpp
 create mode 100644 inference-engine/src/vpu/common/src/configuration/options/copy_optimization.cpp
 create mode 100644 inference-engine/src/vpu/common/src/configuration/options/log_level.cpp
 create mode 100644 inference-engine/src/vpu/common/src/configuration/plugin_configuration.cpp
 create mode 100644 inference-engine/src/vpu/common/src/configuration/switch_converters.cpp
 rename inference-engine/src/vpu/{graph_transformer => common}/src/parsed_config.cpp (99%)
 create mode 100644 inference-engine/src/vpu/myriad_plugin/configuration/myriad_configuration.cpp
 create mode 100644 inference-engine/src/vpu/myriad_plugin/configuration/myriad_configuration.hpp

diff --git a/inference-engine/include/ie_parameter.hpp b/inference-engine/include/ie_parameter.hpp
index 1343f89db32939..683f02fd6a552f 100644
--- a/inference-engine/include/ie_parameter.hpp
+++ b/inference-engine/include/ie_parameter.hpp
@@ -86,7 +86,8 @@ class INFERENCE_ENGINE_API_CLASS(Parameter) {
      * @param parameter object
      */
     template <class T,
-              typename = typename std::enable_if<!std::is_same<typename std::decay<T>::type, Parameter>::value>::type>
+              typename = typename std::enable_if<!std::is_same<typename std::decay<T>::type, Parameter>::value &&
+                                                 !std::is_abstract<typename std::decay<T>::type>::value>::type>
     Parameter(T&& parameter) {  // NOLINT
         static_assert(!std::is_same<typename std::decay<T>::type, Parameter>::value, "To prevent recursion");
         ptr = new RealData<typename std::decay<T>::type>(std::forward<T>(parameter));
@@ -254,6 +255,21 @@ class INFERENCE_ENGINE_API_CLASS(Parameter) {
         return !(*this == rhs);
     }
 
+    /**
+     * @brief Prints underlying object to the given output stream.
+     * Uses operator<< if it is defined, leaves stream unchanged otherwise.
+     * In case of empty parameter or nullptr stream immediately returns.
+     *
+     * @param object Object to be printed to the given output stream.
+     * @param stream Output stream object will be printed to.
+     */
+    friend void PrintTo(const Parameter& object, std::ostream* stream) {
+        if (object.empty() || !stream) {
+            return;
+        }
+        object.ptr->print(*stream);
+    }
+
 private:
     template <class T, class EqualTo>
     struct CheckOperatorEqual {
@@ -273,6 +289,24 @@ class INFERENCE_ENGINE_API_CLASS(Parameter) {
     template <class T, class EqualTo = T>
     struct HasOperatorEqual : CheckOperatorEqual<T, EqualTo>::type {};
 
+    template <class T, class U>
+    struct CheckOutputStreamOperator {
+        template <class V, class W>
+        static auto test(W*) -> decltype(std::declval<V&>() << std::declval<W>(), std::true_type()) {
+            return {};
+        }
+
+        template <typename, typename>
+        static auto test(...) -> std::false_type {
+            return {};
+        }
+
+        using type = typename std::is_same<std::true_type, decltype(test<T, U>(nullptr))>::type;
+    };
+
+    template <class T>
+    struct HasOutputStreamOperator : CheckOutputStreamOperator<std::ostream, T>::type {};
+
     struct Any {
 #ifdef __ANDROID__
         virtual ~Any();
@@ -282,6 +316,7 @@ class INFERENCE_ENGINE_API_CLASS(Parameter) {
         virtual bool is(const std::type_info&) const = 0;
         virtual Any* copy() const = 0;
         virtual bool operator==(const Any& rhs) const = 0;
+        virtual void print(std::ostream&) const = 0;
     };
 
     template <class T>
@@ -318,6 +353,20 @@ class INFERENCE_ENGINE_API_CLASS(Parameter) {
         bool operator==(const Any& rhs) const override {
             return rhs.is(typeid(T)) && equal<T>(*this, rhs);
         }
+
+        template <class U>
+        typename std::enable_if<!HasOutputStreamOperator<U>::value, void>::type
+        print(std::ostream& stream, const U& object) const {}
+
+        template <class U>
+        typename std::enable_if<HasOutputStreamOperator<U>::value, void>::type
+        print(std::ostream& stream, const U& object) const {
+            stream << object;
+        }
+
+        void print(std::ostream& stream) const override {
+            print<T>(stream, get());
+        }
     };
 
     template <typename T>
diff --git a/inference-engine/src/plugin_api/debug.h b/inference-engine/src/plugin_api/debug.h
index 838c5b02941f00..d52c7b0b942038 100644
--- a/inference-engine/src/plugin_api/debug.h
+++ b/inference-engine/src/plugin_api/debug.h
@@ -25,6 +25,9 @@
 
 #include "ie_algorithm.hpp"
 
+namespace InferenceEngine {
+namespace details {
+
 /**
  * @brief Serializes a `std::vector` to a `std::ostream`
  * @ingroup ie_dev_api_error_debug
@@ -32,7 +35,6 @@
  * @param vec A vector to serialize
  * @return A reference to a `std::stream`
  */
-namespace std {
 template <typename T>
 inline std::ostream& operator<<(std::ostream& out, const std::vector<T>& vec) {
     if (vec.empty()) return std::operator<<(out, "[]");
@@ -42,10 +44,7 @@ inline std::ostream& operator<<(std::ostream& out, const std::vector<T>& vec) {
     }
     return out << "]";
 }
-}  // namespace std
 
-namespace InferenceEngine {
-namespace details {
 /**
  * @brief trim from start (in place)
  * @ingroup ie_dev_api_error_debug
diff --git a/inference-engine/src/vpu/common/include/vpu/configuration.hpp b/inference-engine/src/vpu/common/include/vpu/configuration.hpp
new file mode 100644
index 00000000000000..4ed6f77b91fdaf
--- /dev/null
+++ b/inference-engine/src/vpu/common/include/vpu/configuration.hpp
@@ -0,0 +1,99 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <string>
+#include <vector>
+#include <map>
+
+#include "caseless.hpp"
+
+#include "vpu/utils/optional.hpp"
+
+namespace vpu {
+
+struct CompilationConfig {
+    int numSHAVEs = -1;
+    int numCMXSlices = -1;
+    int numExecutors = -1;
+    int tilingCMXLimitKB = -1;
+
+    bool hwOptimization = true;
+    bool hwExtraSplit = false;
+
+    std::string irWithVpuScalesDir;
+
+    std::string customLayers;
+
+    bool detectBatch = true;
+
+    Optional<bool> injectSwOps;
+    Optional<bool> packDataInCmx;
+    bool mergeHwPoolToConv = true;
+    bool hwDilation = false;
+    bool forceDeprecatedCnnConversion = false;
+    bool enableEarlyEltwiseReLUFusion = true;
+
+    std::map<std::string, std::vector<int>> ioStrides;
+
+    //
+    // Debug options
+    //
+
+    InferenceEngine::details::caseless_set<std::string> hwWhiteList;
+    InferenceEngine::details::caseless_set<std::string> hwBlackList;
+
+    bool hwDisabled(const std::string& layerName) const {
+        if (!hwWhiteList.empty()) {
+            return hwWhiteList.count(layerName) == 0;
+        }
+
+        if (!hwBlackList.empty()) {
+            return hwBlackList.count(layerName) != 0;
+        }
+
+        return false;
+    }
+
+    InferenceEngine::details::caseless_set<std::string> noneLayers;
+
+    bool skipAllLayers() const {
+        if (noneLayers.size() == 1) {
+            const auto& val = *noneLayers.begin();
+            return val == "*";
+        }
+        return false;
+    }
+
+    bool skipLayerType(const std::string& layerType) const {
+        return noneLayers.count(layerType) != 0;
+    }
+    bool ignoreUnknownLayers = false;
+
+    std::string dumpInternalGraphFileName;
+    std::string dumpInternalGraphDirectory;
+    bool dumpAllPasses;
+
+    bool disableReorder = false;  // TODO: rename to enableReorder and switch logic.
+    bool disableConvertStages = false;
+    bool enablePermuteMerging = true;
+    bool enableReplWithSCRelu = false;
+    bool enableReplaceWithReduceMean = true;
+    bool enableTensorIteratorUnrolling = false;
+    bool forcePureTensorIterator = false;
+    bool enableMemoryTypesAnnotation = false;
+    bool enableWeightsAnalysis = true;
+    bool checkPreprocessingInsideModel = true;
+    bool enableCustomReshapeParam = false;
+
+    //
+    // Deprecated options
+    //
+
+    float inputScale = 1.0f;
+    float inputBias = 0.0f;
+};
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/common/include/vpu/configuration/as_parameter_enabler.hpp b/inference-engine/src/vpu/common/include/vpu/configuration/as_parameter_enabler.hpp
new file mode 100644
index 00000000000000..4d480461ec7e71
--- /dev/null
+++ b/inference-engine/src/vpu/common/include/vpu/configuration/as_parameter_enabler.hpp
@@ -0,0 +1,18 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <string>
+
+#include "ie_parameter.hpp"
+
+template<class OptionConcept>
+struct AsParsedParameterEnabler {
+    static InferenceEngine::Parameter asParameter(const std::string& value) { return {OptionConcept::parse(value)}; }
+};
+
+struct AsParameterEnabler {
+    static InferenceEngine::Parameter asParameter(const std::string& value);
+};
diff --git a/inference-engine/src/vpu/common/include/vpu/configuration/options/copy_optimization.hpp b/inference-engine/src/vpu/common/include/vpu/configuration/options/copy_optimization.hpp
new file mode 100644
index 00000000000000..5777774fdf40cd
--- /dev/null
+++ b/inference-engine/src/vpu/common/include/vpu/configuration/options/copy_optimization.hpp
@@ -0,0 +1,34 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <string>
+
+#include "vpu/configuration/as_parameter_enabler.hpp"
+
+namespace vpu {
+
+namespace details {
+
+enum class Access;
+enum class Category;
+
+}  // namespace details
+
+class PluginConfiguration;
+
+struct CopyOptimizationOption : public AsParsedParameterEnabler<CopyOptimizationOption> {
+    using value_type = bool;
+
+    static std::string key();
+    static void validate(const std::string&);
+    static void validate(const PluginConfiguration&);
+    static std::string defaultValue();
+    static value_type parse(const std::string&);
+    static details::Access access();
+    static details::Category category();
+};
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/common/include/vpu/configuration/options/log_level.hpp b/inference-engine/src/vpu/common/include/vpu/configuration/options/log_level.hpp
new file mode 100644
index 00000000000000..48ced0d2fbbdd7
--- /dev/null
+++ b/inference-engine/src/vpu/common/include/vpu/configuration/options/log_level.hpp
@@ -0,0 +1,36 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <string>
+
+#include "vpu/configuration/as_parameter_enabler.hpp"
+
+namespace vpu {
+
+enum class LogLevel;
+
+namespace details {
+
+enum class Access;
+enum class Category;
+
+}  // namespace details
+
+class PluginConfiguration;
+
+struct LogLevelOption : public AsParameterEnabler {
+    using value_type = LogLevel;
+
+    static std::string key();
+    static void validate(const std::string&);
+    static void validate(const PluginConfiguration&);
+    static std::string defaultValue();
+    static value_type parse(const std::string&);
+    static details::Access access();
+    static details::Category category();
+};
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/common/include/vpu/configuration/plugin_configuration.hpp b/inference-engine/src/vpu/common/include/vpu/configuration/plugin_configuration.hpp
new file mode 100644
index 00000000000000..ced07f36477ee9
--- /dev/null
+++ b/inference-engine/src/vpu/common/include/vpu/configuration/plugin_configuration.hpp
@@ -0,0 +1,142 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <string>
+#include <map>
+#include <unordered_map>
+#include <unordered_set>
+#include <memory>
+#include <vpu/parsed_config.hpp>
+
+#include "ie_parameter.hpp"
+
+#include "vpu/utils/logger.hpp"
+
+namespace vpu {
+
+class PluginConfiguration;
+
+struct ConfigurationOptionConcept {
+    virtual std::string key() const = 0;
+    virtual void validate(const std::string&) const = 0;
+    virtual void validate(const PluginConfiguration&) const = 0;
+    virtual InferenceEngine::Parameter asParameter(const std::string&) const = 0;
+};
+
+namespace details {
+
+template<class Option>
+struct ConfigurationOptionModel : public ConfigurationOptionConcept {
+    std::string key() const override { return Option::key(); }
+    void validate(const std::string& value) const override { return Option::validate(value); }
+    void validate(const PluginConfiguration& options) const override { Option::validate(options); }
+    InferenceEngine::Parameter asParameter(const std::string& value) const override { return Option::asParameter(value); }
+};
+
+enum class Deprecation {
+    Off,
+    On
+};
+
+enum class Access {
+    Private,
+    Public
+};
+
+enum class Category {
+    CompileTime,
+    RunTime
+};
+
+class ConfigurationEntry {
+public:
+    template<class Option>
+    ConfigurationEntry(Option, details::Deprecation deprecation)
+        : m_access(Option::access())
+        , m_deprecation(deprecation)
+        , m_category(Option::category())
+        , m_value(std::make_shared<ConfigurationOptionModel<Option>>())
+        {}
+
+    ConfigurationOptionConcept& get();
+    const ConfigurationOptionConcept& get() const;
+
+    std::string key() const;
+    bool isPrivate() const;
+    bool isDeprecated() const;
+    Category getCategory() const;
+
+private:
+    Access m_access = Access::Public;
+    Deprecation m_deprecation = Deprecation::Off;
+    Category m_category = Category::CompileTime;
+    std::shared_ptr<ConfigurationOptionConcept> m_value;
+};
+
+}  // namespace details
+
+// TODO: remove virtual inheritance once all options are migrated
+// it's needed to pass updated compilation config to graph transformer
+class PluginConfiguration : public virtual ParsedConfig {
+public:
+    PluginConfiguration();
+
+    void from(const std::map<std::string, std::string>& config);
+    void fromAtRuntime(const std::map<std::string, std::string>& config);
+    std::unordered_set<std::string> getPublicKeys() const;
+    bool supports(const std::string& key) const;
+
+    template<class Option>
+    void registerOption() {
+        const auto& key = Option::key();
+        concepts.emplace(key, details::ConfigurationEntry(Option{}, details::Deprecation::Off));
+        if (values.count(key) == 0) {
+            // option could be registered more than once if there are deprecated versions of it
+            values.emplace(key, Option::defaultValue());
+        }
+    }
+
+    template<class Option>
+    void registerDeprecatedOption(const std::string& deprecatedKey) {
+        const auto& key = Option::key();
+        concepts.emplace(deprecatedKey, details::ConfigurationEntry(Option{}, details::Deprecation::On));
+        if (values.count(key) == 0) {
+            // option could be registered more than once if there are deprecated versions of it
+            values.emplace(key, Option::defaultValue());
+        }
+    }
+
+    template<class Option>
+    typename Option::value_type get() const {
+        const auto& key = Option::key();
+        validate(key);
+        return Option::parse(values.at(key));
+    }
+
+    void set(const std::string& key, const std::string& value);
+
+    const std::string& operator[](const std::string& key) const;
+
+    InferenceEngine::Parameter asParameter(const std::string& key) const;
+
+    virtual void validate() const;
+
+private:
+    std::unordered_map<std::string, details::ConfigurationEntry> concepts;
+    std::unordered_map<std::string, std::string> values;
+
+    Logger::Ptr logger;
+
+    enum class Mode {
+        Default,
+        RunTime
+    };
+    void create(const std::map<std::string, std::string>& config, Mode mode = Mode::Default);
+
+    void validate(const std::string& key) const;
+};
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/common/include/vpu/configuration/switch_converters.hpp b/inference-engine/src/vpu/common/include/vpu/configuration/switch_converters.hpp
new file mode 100644
index 00000000000000..0348286871a708
--- /dev/null
+++ b/inference-engine/src/vpu/common/include/vpu/configuration/switch_converters.hpp
@@ -0,0 +1,15 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <string>
+#include <unordered_map>
+
+namespace vpu {
+
+const std::unordered_map<std::string, bool>& string2switch();
+const std::unordered_map<bool, std::string>& switch2string();
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/graph_transformer/include/vpu/parsed_config.hpp b/inference-engine/src/vpu/common/include/vpu/parsed_config.hpp
similarity index 80%
rename from inference-engine/src/vpu/graph_transformer/include/vpu/parsed_config.hpp
rename to inference-engine/src/vpu/common/include/vpu/parsed_config.hpp
index d220cda3c3ba56..aa1793806e945d 100644
--- a/inference-engine/src/vpu/graph_transformer/include/vpu/parsed_config.hpp
+++ b/inference-engine/src/vpu/common/include/vpu/parsed_config.hpp
@@ -10,11 +10,11 @@
 #include <string>
 
 #include <vpu/myriad_config.hpp>
+#include <vpu/configuration.hpp>
 #include <vpu/private_plugin_config.hpp>
 
 #include <vpu/parsed_config_base.hpp>
 
-#include <vpu/graph_transformer.hpp>
 #include <vpu/utils/perf_report.hpp>
 #include <vpu/utils/logger.hpp>
 #include <vpu/utils/enums.hpp>
@@ -23,6 +23,12 @@ namespace vpu {
 
 class ParsedConfig : public ParsedConfigBase {
 public:
+    ParsedConfig() = default;
+    ParsedConfig(const ParsedConfig&) = default;
+    ParsedConfig& operator=(const ParsedConfig&) = default;
+    ParsedConfig(ParsedConfig&&) = delete;
+    ParsedConfig& operator=(ParsedConfig&&) = delete;
+
     const std::string& compilerLogFilePath() const {
         return _compilerLogFilePath;
     }
@@ -31,6 +37,10 @@ class ParsedConfig : public ParsedConfigBase {
         return _compileConfig;
     }
 
+    CompilationConfig& compileConfig() {
+        return _compileConfig;
+    }
+
     bool printReceiveTensorTime() const {
         return _printReceiveTensorTime;
     }
diff --git a/inference-engine/src/vpu/common/include/vpu/parsed_config_base.hpp b/inference-engine/src/vpu/common/include/vpu/parsed_config_base.hpp
index 6c7e776461d139..fc57ea8ed240f3 100644
--- a/inference-engine/src/vpu/common/include/vpu/parsed_config_base.hpp
+++ b/inference-engine/src/vpu/common/include/vpu/parsed_config_base.hpp
@@ -25,10 +25,6 @@ VPU_DECLARE_ENUM(ConfigMode,
 
 class ParsedConfigBase {
 public:
-    LogLevel logLevel() const {
-        return _logLevel;
-    }
-
     bool exclusiveAsyncRequests() const {
         return _exclusiveAsyncRequests;
     }
@@ -37,11 +33,9 @@ class ParsedConfigBase {
     ParsedConfigBase();
     virtual ~ParsedConfigBase();
 
-    void update(
-            const std::map<std::string, std::string>& config,
-            ConfigMode mode = ConfigMode::Any);
-
 protected:
+    void update(const std::map<std::string, std::string>& config, ConfigMode mode = ConfigMode::Any);
+
     virtual const std::unordered_set<std::string>& getCompileOptions() const;
     virtual const std::unordered_set<std::string>& getRunTimeOptions() const;
     virtual const std::unordered_set<std::string>& getDeprecatedOptions() const;
@@ -130,7 +124,6 @@ class ParsedConfigBase {
     Logger::Ptr _log;
 
 private:
-    LogLevel _logLevel = LogLevel::None;
     bool _exclusiveAsyncRequests = false;
 };
 
diff --git a/inference-engine/src/vpu/graph_transformer/include/vpu/private_plugin_config.hpp b/inference-engine/src/vpu/common/include/vpu/private_plugin_config.hpp
similarity index 100%
rename from inference-engine/src/vpu/graph_transformer/include/vpu/private_plugin_config.hpp
rename to inference-engine/src/vpu/common/include/vpu/private_plugin_config.hpp
diff --git a/inference-engine/src/vpu/common/include/vpu/utils/containers.hpp b/inference-engine/src/vpu/common/include/vpu/utils/containers.hpp
new file mode 100644
index 00000000000000..745613c977ece8
--- /dev/null
+++ b/inference-engine/src/vpu/common/include/vpu/utils/containers.hpp
@@ -0,0 +1,40 @@
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <vector>
+#include <algorithm>
+
+#include "error.hpp"
+
+namespace vpu {
+
+template<class Key, class Value, template<class...> class Map>
+inline std::vector<Key> getKeys(const Map<Key, Value>& map) {
+    auto keys = std::vector<Key>{};
+    keys.reserve(map.size());
+    std::transform(map.cbegin(), map.cend(), std::back_inserter(keys), [](const std::pair<Key, Value>& entry) { return entry.first; });
+    return keys;
+}
+
+template<class Key, class Value, template<class...> class Map>
+inline std::vector<Value> getValues(const Map<Key, Value>& map) {
+    auto values = std::vector<Value>{};
+    values.reserve(map.size());
+    std::transform(map.cbegin(), map.cend(), std::back_inserter(values), [](const std::pair<Key, Value>& entry) { return entry.second; });
+    return values;
+}
+
+template<class Key, class Value, template<class...> class Map>
+inline Map<Value, Key> inverse(const Map<Key, Value>& map) {
+    auto inverted = Map<Value, Key>{};
+    for (const auto& entry : map) {
+        const auto& insertion = inverted.emplace(entry.second, entry.first);
+        VPU_THROW_UNLESS(insertion.second, "Could not invert map {} due to duplicated value \"{}\"", map, entry.second);
+    }
+    return inverted;
+}
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/common/include/vpu/utils/error.hpp b/inference-engine/src/vpu/common/include/vpu/utils/error.hpp
index 2180e6c86c0320..1c36e1f7dd41d5 100644
--- a/inference-engine/src/vpu/common/include/vpu/utils/error.hpp
+++ b/inference-engine/src/vpu/common/include/vpu/utils/error.hpp
@@ -29,6 +29,11 @@ class UnsupportedLayerException : public VPUException {
     using VPUException::VPUException;
 };
 
+class UnsupportedConfigurationOptionException : public VPUException {
+public:
+    using VPUException::VPUException;
+};
+
 template <class Exception, typename... Args>
 void throwFormat(const char* fileName, int lineNumber, const char* messageFormat, Args&&... args) {
     IE_THROW(GeneralError) << '\n' << fileName  << ':' << lineNumber << ' '
@@ -47,13 +52,20 @@ void throwFormat(const char* fileName, int lineNumber, const char* messageFormat
         }                                                                                               \
     } while (false)
 
-#define VPU_THROW_UNSUPPORTED_UNLESS(condition, ...)                                                                    \
+#define VPU_THROW_UNSUPPORTED_LAYER_UNLESS(condition, ...)                                                              \
     do {                                                                                                                \
         if (!(condition)) {                                                                                             \
             ::vpu::details::throwFormat<::vpu::details::UnsupportedLayerException>(__FILE__, __LINE__, __VA_ARGS__);    \
         }                                                                                                               \
     } while (false)
 
+#define VPU_THROW_UNSUPPORTED_OPTION_UNLESS(condition, ...)                                                                           \
+    do {                                                                                                                              \
+        if (!(condition)) {                                                                                                           \
+            ::vpu::details::throwFormat<::vpu::details::UnsupportedConfigurationOptionException>(__FILE__, __LINE__, __VA_ARGS__);    \
+        }                                                                                                                             \
+    } while (false)
+
 #ifdef NDEBUG
 #   define VPU_INTERNAL_CHECK(condition, ...)                               \
         do {                                                                \
diff --git a/inference-engine/src/vpu/common/include/vpu/utils/log_level.hpp b/inference-engine/src/vpu/common/include/vpu/utils/log_level.hpp
new file mode 100644
index 00000000000000..51eb753150815d
--- /dev/null
+++ b/inference-engine/src/vpu/common/include/vpu/utils/log_level.hpp
@@ -0,0 +1,21 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "vpu/utils/enums.hpp"
+
+namespace vpu {
+
+VPU_DECLARE_ENUM(LogLevel,
+    None,
+    Fatal,    /* used for very severe error events that will most probably cause the application to terminate */
+    Error,    /* reporting events which are not expected during normal execution, containing probable reason */
+    Warning,  /* indicating events which are not usual and might lead to errors later */
+    Info,     /* short enough messages about ongoing activity in the process */
+    Debug,    /* more fine-grained messages with references to particular data and explanations */
+    Trace     /* involved and detailed information about execution, helps to trace the execution flow, produces huge output */
+)
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/common/include/vpu/utils/logger.hpp b/inference-engine/src/vpu/common/include/vpu/utils/logger.hpp
index de06c8511adcfd..931d02a1648ce4 100644
--- a/inference-engine/src/vpu/common/include/vpu/utils/logger.hpp
+++ b/inference-engine/src/vpu/common/include/vpu/utils/logger.hpp
@@ -13,6 +13,7 @@
 #include <vpu/utils/enums.hpp>
 #include <vpu/utils/auto_scope.hpp>
 #include <vpu/utils/io.hpp>
+#include <vpu/utils/log_level.hpp>
 
 namespace vpu {
 
@@ -39,20 +40,6 @@ OutputStream::Ptr fileOutput(const std::string& fileName);
 
 OutputStream::Ptr defaultOutput(const std::string& fileName = std::string());
 
-//
-// Logger
-//
-
-VPU_DECLARE_ENUM(LogLevel,
-    None,
-    Fatal,    /* used for very severe error events that will most probably cause the application to terminate */
-    Error,    /* reporting events which are not expected during normal execution, containing probable reason */
-    Warning,  /* indicating events which are not usual and might lead to errors later */
-    Info,     /* short enough messages about ongoing activity in the process */
-    Debug,    /* more fine-grained messages with references to particular data and explanations */
-    Trace     /* involved and detailed information about execution, helps to trace the execution flow, produces huge output */
-)
-
 class Logger final {
 public:
     using Ptr = std::shared_ptr<Logger>;
diff --git a/inference-engine/src/vpu/common/src/configuration/as_parameter_enabler.cpp b/inference-engine/src/vpu/common/src/configuration/as_parameter_enabler.cpp
new file mode 100644
index 00000000000000..04c9d525971e72
--- /dev/null
+++ b/inference-engine/src/vpu/common/src/configuration/as_parameter_enabler.cpp
@@ -0,0 +1,10 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <vpu/configuration/as_parameter_enabler.hpp>
+
+InferenceEngine::Parameter AsParameterEnabler::asParameter(const std::string& value) {
+    return {value};
+}
+
diff --git a/inference-engine/src/vpu/common/src/configuration/options/copy_optimization.cpp b/inference-engine/src/vpu/common/src/configuration/options/copy_optimization.cpp
new file mode 100644
index 00000000000000..7fc384aac01d6b
--- /dev/null
+++ b/inference-engine/src/vpu/common/src/configuration/options/copy_optimization.cpp
@@ -0,0 +1,45 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "vpu/private_plugin_config.hpp"
+#include "vpu/utils/containers.hpp"
+#include "vpu/configuration/options/copy_optimization.hpp"
+#include "vpu/configuration/switch_converters.hpp"
+#include "vpu/configuration/plugin_configuration.hpp"
+
+namespace vpu {
+
+void CopyOptimizationOption::validate(const std::string& value) {
+    const auto& converters = string2switch();
+    VPU_THROW_UNLESS(converters.count(value) != 0, R"(unexpected copy optimization option value "{}", only {} are supported)", value, getKeys(converters));
+}
+
+void CopyOptimizationOption::validate(const PluginConfiguration& configuration) {
+    validate(configuration[key()]);
+}
+
+std::string CopyOptimizationOption::key() {
+    return InferenceEngine::MYRIAD_COPY_OPTIMIZATION;
+}
+
+details::Access CopyOptimizationOption::access() {
+    return details::Access::Private;
+}
+
+details::Category CopyOptimizationOption::category() {
+    return details::Category::CompileTime;
+}
+
+std::string CopyOptimizationOption::defaultValue() {
+    return InferenceEngine::PluginConfigParams::YES;
+}
+
+CopyOptimizationOption::value_type CopyOptimizationOption::parse(const std::string& value) {
+    const auto& converters = string2switch();
+    VPU_THROW_UNSUPPORTED_OPTION_UNLESS(converters.count(value) != 0, R"(unexpected copy optimization option value "{}", only {} are supported)",
+        value, getKeys(converters));
+    return converters.at(value);
+}
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/common/src/configuration/options/log_level.cpp b/inference-engine/src/vpu/common/src/configuration/options/log_level.cpp
new file mode 100644
index 00000000000000..79729aebfef8d3
--- /dev/null
+++ b/inference-engine/src/vpu/common/src/configuration/options/log_level.cpp
@@ -0,0 +1,64 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "vpu/configuration/options/log_level.hpp"
+#include "vpu/utils/log_level.hpp"
+#include "vpu/utils/containers.hpp"
+#include "vpu/configuration/plugin_configuration.hpp"
+
+#include "ie_plugin_config.hpp"
+
+#include <unordered_map>
+
+namespace vpu {
+
+namespace {
+
+const std::unordered_map<std::string, LogLevel>& string2level() {
+    static const std::unordered_map<std::string, LogLevel> converters = {
+        {CONFIG_VALUE(LOG_NONE),    LogLevel::None},
+        {CONFIG_VALUE(LOG_ERROR),   LogLevel::Error},
+        {CONFIG_VALUE(LOG_WARNING), LogLevel::Warning},
+        {CONFIG_VALUE(LOG_INFO),    LogLevel::Info},
+        {CONFIG_VALUE(LOG_DEBUG),   LogLevel::Debug},
+        {CONFIG_VALUE(LOG_TRACE),   LogLevel::Trace},
+    };
+    return converters;
+}
+
+}  // namespace
+
+void LogLevelOption::validate(const std::string& value) {
+    const auto& converters = string2level();
+    VPU_THROW_UNLESS(converters.count(value) != 0, R"(unexpected log level option value "{}", only {} are supported)", value, getKeys(converters));
+}
+
+void LogLevelOption::validate(const PluginConfiguration& configuration) {
+    validate(configuration[key()]);
+}
+
+std::string LogLevelOption::key() {
+    return InferenceEngine::PluginConfigParams::KEY_LOG_LEVEL;
+}
+
+details::Access LogLevelOption::access() {
+    return details::Access::Public;
+}
+
+details::Category LogLevelOption::category() {
+    return details::Category::CompileTime;
+}
+
+std::string LogLevelOption::defaultValue() {
+    return InferenceEngine::PluginConfigParams::LOG_NONE;
+}
+
+LogLevelOption::value_type LogLevelOption::parse(const std::string& value) {
+    const auto& converters = string2level();
+    VPU_THROW_UNSUPPORTED_OPTION_UNLESS(converters.count(value) != 0, R"(unexpected log level option value "{}", only {} are supported)",
+        value, getKeys(converters));
+    return converters.at(value);
+}
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/common/src/configuration/plugin_configuration.cpp b/inference-engine/src/vpu/common/src/configuration/plugin_configuration.cpp
new file mode 100644
index 00000000000000..6e051606c675b1
--- /dev/null
+++ b/inference-engine/src/vpu/common/src/configuration/plugin_configuration.cpp
@@ -0,0 +1,114 @@
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "vpu/utils/error.hpp"
+#include "vpu/configuration/plugin_configuration.hpp"
+
+#include "ie_plugin_config.hpp"
+
+namespace vpu {
+
+namespace details {
+
+ConfigurationOptionConcept& ConfigurationEntry::get() {
+    return *m_value;
+}
+
+const ConfigurationOptionConcept& ConfigurationEntry::get() const {
+    return *m_value;
+}
+
+bool ConfigurationEntry::isPrivate() const {
+    return m_access == Access::Private;
+}
+
+bool ConfigurationEntry::isDeprecated() const {
+    return m_deprecation == Deprecation::On;
+}
+
+Category ConfigurationEntry::getCategory() const {
+    return m_category;
+}
+
+std::string ConfigurationEntry::key() const {
+    return m_value->key();
+}
+
+}  // namespace details
+
+PluginConfiguration::PluginConfiguration() : logger(std::make_shared<Logger>("Configuration", LogLevel::Warning, consoleOutput())) {}
+
+
+std::unordered_set<std::string> PluginConfiguration::getPublicKeys() const {
+    auto publicKeys = std::unordered_set<std::string>{};
+    for (const auto& entry : concepts) {
+        const auto& key    = entry.first;
+        const auto& option = entry.second;
+        if (option.isPrivate()) {
+            continue;
+        }
+        publicKeys.insert(key);
+    }
+    return publicKeys;
+}
+
+bool PluginConfiguration::supports(const std::string& key) const {
+    return concepts.count(key) != 0;
+}
+
+void PluginConfiguration::from(const std::map<std::string, std::string>& config) {
+    create(config);
+}
+
+void PluginConfiguration::fromAtRuntime(const std::map<std::string, std::string>& config) {
+    create(config, Mode::RunTime);
+}
+
+void PluginConfiguration::validate() const {
+    for (const auto& option : concepts) {
+        option.second.get().validate(*this);
+    }
+}
+
+void PluginConfiguration::create(const std::map<std::string, std::string>& config, Mode mode) {
+    for (const auto& entry : config) {
+        const auto& key = entry.first;
+        validate(key);
+
+        const auto& optionConcept = concepts.at(key);
+        if (mode == Mode::RunTime && optionConcept.getCategory() == details::Category::CompileTime) {
+            logger->warning("Configuration option \"{}\" is used after network is loaded. Its value is going to be ignored.", key);
+            continue;
+        }
+
+        const auto& value = entry.second;
+        set(key, value);
+    }
+}
+
+InferenceEngine::Parameter PluginConfiguration::asParameter(const std::string& key) const {
+    const auto& value = operator[](key);
+    return concepts.at(key).get().asParameter(value);
+}
+
+void PluginConfiguration::validate(const std::string& key) const {
+    VPU_THROW_UNSUPPORTED_OPTION_UNLESS(supports(key), "Encountered an unsupported key {}, supported keys are {}", key, getPublicKeys());
+    if (concepts.at(key).isDeprecated()) {
+        logger->warning("Encountered deprecated option {} usage, consider replacing it with {} option", key, concepts.at(key).key());
+    }
+}
+
+const std::string& PluginConfiguration::operator[](const std::string& key) const {
+    validate(key);
+    return values.at(concepts.at(key).key());
+}
+
+void PluginConfiguration::set(const std::string& key, const std::string& value) {
+    validate(key);
+    const auto& optionConcept = concepts.at(key).get();
+    optionConcept.validate(value);
+    values[optionConcept.key()] = value;
+}
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/common/src/configuration/switch_converters.cpp b/inference-engine/src/vpu/common/src/configuration/switch_converters.cpp
new file mode 100644
index 00000000000000..232b184cf8dfb2
--- /dev/null
+++ b/inference-engine/src/vpu/common/src/configuration/switch_converters.cpp
@@ -0,0 +1,25 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "vpu/utils/containers.hpp"
+#include "vpu/configuration/switch_converters.hpp"
+
+#include "ie_plugin_config.hpp"
+
+namespace vpu {
+
+const std::unordered_map<std::string, bool>& string2switch() {
+    static const std::unordered_map<std::string, bool> converters = {
+        {CONFIG_VALUE(NO), false},
+        {CONFIG_VALUE(YES), true}
+    };
+    return converters;
+}
+
+const std::unordered_map<bool, std::string>& switch2string() {
+    static const auto converters = inverse(string2switch());
+    return converters;
+}
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/graph_transformer/src/parsed_config.cpp b/inference-engine/src/vpu/common/src/parsed_config.cpp
similarity index 99%
rename from inference-engine/src/vpu/graph_transformer/src/parsed_config.cpp
rename to inference-engine/src/vpu/common/src/parsed_config.cpp
index 5e8779a73127ee..b0a2ec46c9392a 100644
--- a/inference-engine/src/vpu/graph_transformer/src/parsed_config.cpp
+++ b/inference-engine/src/vpu/common/src/parsed_config.cpp
@@ -169,7 +169,6 @@ void ParsedConfig::parse(const std::map<std::string, std::string>& config) {
     setOption(_compileConfig.dumpAllPasses,                  switches, config, ie::MYRIAD_DUMP_ALL_PASSES);
 
     setOption(_compileConfig.detectBatch,                    switches, config, ie::MYRIAD_DETECT_NETWORK_BATCH);
-    setOption(_compileConfig.copyOptimization,               switches, config, ie::MYRIAD_COPY_OPTIMIZATION);
     setOption(_compileConfig.packDataInCmx,                  switches, config, ie::MYRIAD_PACK_DATA_IN_CMX);
     setOption(_compileConfig.ignoreUnknownLayers,            switches, config, ie::MYRIAD_IGNORE_UNKNOWN_LAYERS);
     setOption(_compileConfig.hwOptimization,                 switches, config, ie::MYRIAD_ENABLE_HW_ACCELERATION);
diff --git a/inference-engine/src/vpu/common/src/parsed_config_base.cpp b/inference-engine/src/vpu/common/src/parsed_config_base.cpp
index 876232bd15e820..c9e8a36f4a7c45 100644
--- a/inference-engine/src/vpu/common/src/parsed_config_base.cpp
+++ b/inference-engine/src/vpu/common/src/parsed_config_base.cpp
@@ -59,13 +59,7 @@ void ParsedConfigBase::update(
 }
 
 const std::unordered_set<std::string>& ParsedConfigBase::getCompileOptions() const {
-IE_SUPPRESS_DEPRECATED_START
-    static const std::unordered_set<std::string> options = {
-        CONFIG_KEY(LOG_LEVEL),
-        VPU_CONFIG_KEY(LOG_LEVEL)
-    };
-IE_SUPPRESS_DEPRECATED_END
-
+    static const std::unordered_set<std::string> options;
     return options;
 }
 
@@ -73,8 +67,6 @@ const std::unordered_set<std::string>& ParsedConfigBase::getRunTimeOptions() con
 IE_SUPPRESS_DEPRECATED_START
     static const std::unordered_set<std::string> options = {
         CONFIG_KEY(EXCLUSIVE_ASYNC_REQUESTS),
-        CONFIG_KEY(LOG_LEVEL),
-        VPU_CONFIG_KEY(LOG_LEVEL)
     };
 IE_SUPPRESS_DEPRECATED_END
 
@@ -82,37 +74,12 @@ IE_SUPPRESS_DEPRECATED_END
 }
 
 const std::unordered_set<std::string>& ParsedConfigBase::getDeprecatedOptions() const {
-    IE_SUPPRESS_DEPRECATED_START
-    static const std::unordered_set<std::string> options = {
-        VPU_CONFIG_KEY(LOG_LEVEL)
-    };
-    IE_SUPPRESS_DEPRECATED_END
-
+    static const std::unordered_set<std::string> options;
     return options;
 }
 
 void ParsedConfigBase::parse(const std::map<std::string, std::string>& config) {
-    static const std::unordered_map<std::string, LogLevel> logLevels = {
-        { CONFIG_VALUE(LOG_NONE), LogLevel::None },
-        { CONFIG_VALUE(LOG_ERROR), LogLevel::Error },
-        { CONFIG_VALUE(LOG_WARNING), LogLevel::Warning },
-        { CONFIG_VALUE(LOG_INFO), LogLevel::Info },
-        { CONFIG_VALUE(LOG_DEBUG), LogLevel::Debug },
-        { CONFIG_VALUE(LOG_TRACE), LogLevel::Trace }
-    };
-
-    setOption(_logLevel, logLevels, config, CONFIG_KEY(LOG_LEVEL));
     setOption(_exclusiveAsyncRequests, switches, config, CONFIG_KEY(EXCLUSIVE_ASYNC_REQUESTS));
-
-IE_SUPPRESS_DEPRECATED_START
-    setOption(_logLevel, logLevels, config, VPU_CONFIG_KEY(LOG_LEVEL));
-IE_SUPPRESS_DEPRECATED_END
-
-#ifndef NDEBUG
-    if (const auto envVar = std::getenv("IE_VPU_LOG_LEVEL")) {
-        _logLevel = logLevels.at(envVar);
-    }
-#endif
 }
 
 std::unordered_set<std::string> ParsedConfigBase::merge(
diff --git a/inference-engine/src/vpu/graph_transformer/CMakeLists.txt b/inference-engine/src/vpu/graph_transformer/CMakeLists.txt
index bc73ab5b155696..1de08e8cae3be4 100644
--- a/inference-engine/src/vpu/graph_transformer/CMakeLists.txt
+++ b/inference-engine/src/vpu/graph_transformer/CMakeLists.txt
@@ -48,8 +48,13 @@ function(add_graph_transformer_target TARGET_NAME STATIC_IE)
         target_link_libraries(${TARGET_NAME} PUBLIC pugixml vpu_common_lib)
     endif()
 
-    target_link_libraries(${TARGET_NAME} PUBLIC ${NGRAPH_LIBRARIES}
-                                         PRIVATE openvino::itt)
+    target_link_libraries(${TARGET_NAME}
+        PUBLIC
+            ${NGRAPH_LIBRARIES}
+        PRIVATE
+            openvino::itt
+            mvnc # TODO: remove once all options are migrated
+    )
 
     if(WIN32)
         target_compile_definitions(${TARGET_NAME} PRIVATE NOMINMAX)
diff --git a/inference-engine/src/vpu/graph_transformer/include/vpu/compile_env.hpp b/inference-engine/src/vpu/graph_transformer/include/vpu/compile_env.hpp
index 157b13a0b59b03..3228c7e1a4cedf 100644
--- a/inference-engine/src/vpu/graph_transformer/include/vpu/compile_env.hpp
+++ b/inference-engine/src/vpu/graph_transformer/include/vpu/compile_env.hpp
@@ -8,28 +8,29 @@
 #include <vpu/model/model.hpp>
 #include <vpu/utils/logger.hpp>
 #include <vpu/utils/profiling.hpp>
+#include <mvnc.h>
 
 namespace vpu {
 
 struct DeviceResources {
-    static int numShaves(const Platform& platform);
-    static int numSlices(const Platform& platform);
+    static int numShaves(const ncDevicePlatform_t& platform);
+    static int numSlices(const ncDevicePlatform_t& platform);
     static int numStreams();
 };
 
 struct DefaultAllocation {
-    static int numStreams(const Platform& platform, const CompilationConfig& configuration);
-    static int numSlices(const Platform& platform, int numStreams);
-    static int numShaves(const Platform& platform, int numStreams, int numSlices);
+    static int numStreams(const ncDevicePlatform_t& platform, const PluginConfiguration& configuration);
+    static int numSlices(const ncDevicePlatform_t& platform, int numStreams);
+    static int numShaves(const ncDevicePlatform_t& platform, int numStreams, int numSlices);
     static int tilingCMXLimit(int numSlices);
 };
 
 struct CompileEnv final {
 public:
-    Platform platform;
+    ncDevicePlatform_t platform;
     Resources resources;
 
-    CompilationConfig config;
+    PluginConfiguration config;
 
     Logger::Ptr log;
 
@@ -49,14 +50,14 @@ struct CompileEnv final {
     static const CompileEnv* getOrNull();
 
     static void init(
-            Platform platform,
-            const CompilationConfig& config,
-            const Logger::Ptr& log);
-    static void updateConfig(const CompilationConfig& config);
+        ncDevicePlatform_t platform,
+        const PluginConfiguration& config,
+        const Logger::Ptr& log);
+    static void updateConfig(const PluginConfiguration& config);
     static void free();
 
 private:
-    explicit CompileEnv(Platform platform);
+    explicit CompileEnv(ncDevicePlatform_t platform);
 };
 
 }  // namespace vpu
diff --git a/inference-engine/src/vpu/graph_transformer/include/vpu/graph_transformer.hpp b/inference-engine/src/vpu/graph_transformer/include/vpu/graph_transformer.hpp
index c0881b9b0c6ddc..afe4974502ccb3 100644
--- a/inference-engine/src/vpu/graph_transformer/include/vpu/graph_transformer.hpp
+++ b/inference-engine/src/vpu/graph_transformer/include/vpu/graph_transformer.hpp
@@ -21,108 +21,14 @@
 #include <vpu/utils/perf_report.hpp>
 #include <vpu/utils/logger.hpp>
 #include <vpu/utils/optional.hpp>
+#include <vpu/configuration/plugin_configuration.hpp>
+
+#include "mvnc.h"
 
 namespace vpu {
 
 namespace ie = InferenceEngine;
 
-//
-// CompilationConfig
-//
-
-VPU_DECLARE_ENUM(Platform,
-    MYRIAD_2 = 2450,
-    MYRIAD_X = 2480,
-)
-
-struct CompilationConfig final {
-    //
-    // Compilation options
-    //
-
-    int numSHAVEs = -1;
-    int numCMXSlices = -1;
-    int numExecutors = -1;
-    int tilingCMXLimitKB = -1;
-
-    bool hwOptimization = true;
-    bool hwExtraSplit = false;
-
-    std::string irWithVpuScalesDir;
-
-    std::string customLayers;
-
-    bool detectBatch = true;
-
-    Optional<bool> copyOptimization;
-    Optional<bool> injectSwOps;
-    Optional<bool> packDataInCmx;
-    bool mergeHwPoolToConv = true;
-    bool hwDilation = false;
-    bool forceDeprecatedCnnConversion = false;
-    bool enableEarlyEltwiseReLUFusion = true;
-
-    std::map<std::string, std::vector<int>> ioStrides;
-
-    //
-    // Debug options
-    //
-
-    ie::details::caseless_set<std::string> hwWhiteList;
-    ie::details::caseless_set<std::string> hwBlackList;
-
-    bool hwDisabled(const std::string& layerName) const {
-        if (!hwWhiteList.empty()) {
-            return hwWhiteList.count(layerName) == 0;
-        }
-
-        if (!hwBlackList.empty()) {
-            return hwBlackList.count(layerName) != 0;
-        }
-
-        return false;
-    }
-
-    ie::details::caseless_set<std::string> noneLayers;
-
-    bool skipAllLayers() const {
-        if (noneLayers.size() == 1) {
-            const auto& val = *noneLayers.begin();
-            return val == "*";
-        }
-        return false;
-    }
-
-    bool skipLayerType(const std::string& layerType) const {
-        return noneLayers.count(layerType) != 0;
-    }
-    bool ignoreUnknownLayers = false;
-
-    std::string dumpInternalGraphFileName;
-    std::string dumpInternalGraphDirectory;
-    bool dumpAllPasses;
-
-    bool disableReorder = false;  // TODO: rename to enableReorder and switch logic.
-    bool disableConvertStages = false;
-    bool enablePermuteMerging = true;
-    bool enableReplWithSCRelu = false;
-    bool enableReplaceWithReduceMean = true;
-    bool enableTensorIteratorUnrolling = false;
-    bool forcePureTensorIterator = false;
-    bool enableMemoryTypesAnnotation = false;
-    bool enableWeightsAnalysis = true;
-    bool checkPreprocessingInsideModel = true;
-    bool enableCustomReshapeParam = false;
-
-    //
-    // Deprecated options
-    //
-
-    float inputScale = 1.0f;
-    float inputBias = 0.0f;
-};
-
-
 //
 // DataInfo
 //
@@ -165,17 +71,17 @@ struct CompiledGraph final {
 // compileNetwork
 //
 
-CompiledGraph::Ptr compileNetwork(const ie::CNNNetwork& network, Platform platform, const CompilationConfig& config, const Logger::Ptr& log,
-    const ie::ICore* core);
+CompiledGraph::Ptr compileNetwork(const ie::CNNNetwork& network, ncDevicePlatform_t platform, const PluginConfiguration& config, const Logger::Ptr& log,
+                                  const ie::ICore* core);
 
-CompiledGraph::Ptr compileSubNetwork(const ie::CNNNetwork& network, const CompilationConfig& subConfig, const ie::ICore* core);
+CompiledGraph::Ptr compileSubNetwork(const ie::CNNNetwork& network, const PluginConfiguration& subConfig, const ie::ICore* core);
 
 //
 // getSupportedLayers
 //
 
-std::set<std::string> getSupportedLayers(const ie::CNNNetwork& network, Platform platform, const CompilationConfig& config, const Logger::Ptr& log,
-    const ie::ICore* core);
+std::set<std::string> getSupportedLayers(const ie::CNNNetwork& network, ncDevicePlatform_t platform, const PluginConfiguration& config, const Logger::Ptr& log,
+                                         const ie::ICore* core);
 
 //
 // Blob version and checks
diff --git a/inference-engine/src/vpu/graph_transformer/include/vpu/graph_transformer_internal.hpp b/inference-engine/src/vpu/graph_transformer/include/vpu/graph_transformer_internal.hpp
index 141512bd40cecc..99cddb1be791c9 100644
--- a/inference-engine/src/vpu/graph_transformer/include/vpu/graph_transformer_internal.hpp
+++ b/inference-engine/src/vpu/graph_transformer/include/vpu/graph_transformer_internal.hpp
@@ -12,8 +12,8 @@ namespace vpu {
 
 CompiledGraph::Ptr compileModel(
         const Model& model,
-        Platform platform,
-        const CompilationConfig& config,
+        ncDevicePlatform_t platform,
+        const PluginConfiguration& config,
         const Logger::Ptr& log);
 
 }  // namespace vpu
diff --git a/inference-engine/src/vpu/graph_transformer/src/backend/backend.cpp b/inference-engine/src/vpu/graph_transformer/src/backend/backend.cpp
index fdb632993b6f6b..27a9d40794e7b0 100644
--- a/inference-engine/src/vpu/graph_transformer/src/backend/backend.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/backend/backend.cpp
@@ -85,12 +85,12 @@ void BackEnd::dumpModel(
 
     std::string fileName;
 
-    if (!env.config.dumpInternalGraphFileName.empty()) {
-        fileName = fileNameNoExt(env.config.dumpInternalGraphFileName);
-    } else if (!env.config.dumpInternalGraphDirectory.empty()) {
+    if (!env.config.compileConfig().dumpInternalGraphFileName.empty()) {
+        fileName = fileNameNoExt(env.config.compileConfig().dumpInternalGraphFileName);
+    } else if (!env.config.compileConfig().dumpInternalGraphDirectory.empty()) {
         fileName = formatString(
             "%s/vpu_graph_%f%f%i_%s",
-            env.config.dumpInternalGraphDirectory,
+            env.config.compileConfig().dumpInternalGraphDirectory,
             std::setw(2), std::setfill('0'),
             model->attrs().get<int>("index"),
             replaceBadCharacters(model->name()));
@@ -99,7 +99,7 @@ void BackEnd::dumpModel(
     }
 
     if (!postfix.empty()) {
-        if (!env.config.dumpAllPasses) {
+        if (!env.config.compileConfig().dumpAllPasses) {
             return;
         }
 
diff --git a/inference-engine/src/vpu/graph_transformer/src/frontend/detect_network_batch.cpp b/inference-engine/src/vpu/graph_transformer/src/frontend/detect_network_batch.cpp
index 9e3a3f7eab542e..48de8bb9f87dc2 100644
--- a/inference-engine/src/vpu/graph_transformer/src/frontend/detect_network_batch.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/frontend/detect_network_batch.cpp
@@ -29,7 +29,7 @@ void FrontEnd::detectNetworkBatch(
     using PrecisionsMap = std::map<std::string, ie::Precision>;
     const auto& env = CompileEnv::get();
 
-    if (!env.config.detectBatch) {
+    if (!env.config.compileConfig().detectBatch) {
         // skip batch extraction step and go as is
         return;
     }
diff --git a/inference-engine/src/vpu/graph_transformer/src/frontend/frontend.cpp b/inference-engine/src/vpu/graph_transformer/src/frontend/frontend.cpp
index 5ee49dd4ed2905..c399cdc8d2d5b6 100644
--- a/inference-engine/src/vpu/graph_transformer/src/frontend/frontend.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/frontend/frontend.cpp
@@ -436,7 +436,7 @@ void FrontEnd::processTrivialCases(const Model& model) {
 void FrontEnd::defaultOnUnsupportedLayerCallback(const Model& model, const ie::CNNLayerPtr& layer, const DataVector& inputs, const DataVector& outputs,
                                                  const std::string& extraMessage) {
     const auto& env = CompileEnv::get();
-    VPU_THROW_UNSUPPORTED_UNLESS(env.config.ignoreUnknownLayers, "Failed to compile layer \"%v\": %v", layer->name, extraMessage);
+    VPU_THROW_UNSUPPORTED_LAYER_UNLESS(env.config.compileConfig().ignoreUnknownLayers, "Failed to compile layer \"%v\": %v", layer->name, extraMessage);
     _stageBuilder->addNoneStage(model, layer->name, layer, inputs, outputs);
 }
 
@@ -466,15 +466,15 @@ ModelPtr FrontEnd::runCommonPasses(ie::CNNNetwork network,
     // Parse custom layers
     //
 
-    if (!env.config.customLayers.empty()) {
-        env.log->trace("Parse custom layers : %s", env.config.customLayers);
+    if (!env.config.compileConfig().customLayers.empty()) {
+        env.log->trace("Parse custom layers : %s", env.config.compileConfig().customLayers);
         VPU_LOGGER_SECTION(env.log);
 
-        if (env.platform != Platform::MYRIAD_X) {
+        if (env.platform != ncDevicePlatform_t::NC_MYRIAD_X) {
             VPU_THROW_FORMAT("Custom layers are not supported for %v platforms", env.platform);
         }
 
-        _customLayers = CustomLayer::loadFromFile(env.config.customLayers);
+        _customLayers = CustomLayer::loadFromFile(env.config.compileConfig().customLayers);
     }
 
     //
@@ -494,7 +494,7 @@ ModelPtr FrontEnd::runCommonPasses(ie::CNNNetwork network,
         env.log->trace("Update IE Network");
         VPU_LOGGER_SECTION(env.log);
 
-        if (network.getFunction() && env.config.forceDeprecatedCnnConversion) {
+        if (network.getFunction() && env.config.compileConfig().forceDeprecatedCnnConversion) {
             network = convertNetwork(network);
         }
 
@@ -545,7 +545,7 @@ ModelPtr FrontEnd::runCommonPasses(ie::CNNNetwork network,
 
         processTrivialCases(model);
 
-        if (!CompileEnv::get().config.disableConvertStages) {
+        if (!CompileEnv::get().config.compileConfig().disableConvertStages) {
             addDataTypeConvertStages(model);
         }
 
@@ -567,7 +567,7 @@ ModelPtr FrontEnd::runCommonPasses(ie::CNNNetwork network,
 
         getInputAndOutputData(model, layer, inputs, outputs);
 
-        if (env.config.skipAllLayers() || env.config.skipLayerType(layer->type)) {
+        if (env.config.compileConfig().skipAllLayers() || env.config.compileConfig().skipLayerType(layer->type)) {
             _stageBuilder->addNoneStage(model, layer->name, layer, inputs, outputs);
             supportedLayer(layer);
             continue;
diff --git a/inference-engine/src/vpu/graph_transformer/src/frontend/in_out_convert.cpp b/inference-engine/src/vpu/graph_transformer/src/frontend/in_out_convert.cpp
index 6308504cfcc3db..159bd3430a4c01 100644
--- a/inference-engine/src/vpu/graph_transformer/src/frontend/in_out_convert.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/frontend/in_out_convert.cpp
@@ -22,7 +22,7 @@ void FrontEnd::addDataTypeConvertStages(const Model& model) {
     env.log->trace("Add Data type conversion stages");
     VPU_LOGGER_SECTION(env.log);
 
-    const bool hasScaleBias = env.config.inputScale != 1.0f || env.config.inputBias != 0.0f;
+    const bool hasScaleBias = env.config.compileConfig().inputScale != 1.0f || env.config.compileConfig().inputBias != 0.0f;
 
     for (const auto& input : model->datas()) {
         if (input->usage() != DataUsage::Input) {
@@ -38,11 +38,11 @@ void FrontEnd::addDataTypeConvertStages(const Model& model) {
                     env.log->trace("Apply deprecated scale/bias parameters");
 
                     std::ostringstream postfix;
-                    if (env.config.inputScale != 1.0f) {
-                        postfix << "@SCALE=" << InferenceEngine::CNNLayer::ie_serialize_float(env.config.inputScale);
+                    if (env.config.compileConfig().inputScale != 1.0f) {
+                        postfix << "@SCALE=" << InferenceEngine::CNNLayer::ie_serialize_float(env.config.compileConfig().inputScale);
                     }
-                    if (env.config.inputBias != 0.0f) {
-                        postfix << "@BIAS=" << InferenceEngine::CNNLayer::ie_serialize_float(env.config.inputBias);
+                    if (env.config.compileConfig().inputBias != 0.0f) {
+                        postfix << "@BIAS=" << InferenceEngine::CNNLayer::ie_serialize_float(env.config.compileConfig().inputBias);
                     }
 
                     const auto scaledInput = model->duplicateData(
@@ -55,9 +55,9 @@ void FrontEnd::addDataTypeConvertStages(const Model& model) {
                             model,
                             scaledInput->name(),
                             nullptr,
-                            env.config.inputScale,
+                            env.config.compileConfig().inputScale,
                             1.0f,
-                            env.config.inputBias,
+                            env.config.compileConfig().inputBias,
                             input,
                             scaledInput);
                 }
@@ -89,8 +89,8 @@ void FrontEnd::addDataTypeConvertStages(const Model& model) {
                         inputFP16->name(),
                         input,
                         inputFP16,
-                        env.config.inputScale,
-                        env.config.inputBias);
+                        env.config.compileConfig().inputScale,
+                        env.config.compileConfig().inputBias);
 
                 break;
             }
diff --git a/inference-engine/src/vpu/graph_transformer/src/frontend/parse_data.cpp b/inference-engine/src/vpu/graph_transformer/src/frontend/parse_data.cpp
index 00ab79700603cd..392c0b86df84a5 100644
--- a/inference-engine/src/vpu/graph_transformer/src/frontend/parse_data.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/frontend/parse_data.cpp
@@ -25,8 +25,8 @@ void FrontEnd::parseInputAndOutputData(const Model& model) {
     VPU_LOGGER_SECTION(env.log);
 
     const auto parseIOStrides = [&env](const std::string& name, const Data& data) {
-        const auto& match = env.config.ioStrides.find(name);
-        if (match == env.config.ioStrides.end()) {
+        const auto& match = env.config.compileConfig().ioStrides.find(name);
+        if (match == env.config.compileConfig().ioStrides.end()) {
             return;
         }
 
diff --git a/inference-engine/src/vpu/graph_transformer/src/frontend/unroll_loops.cpp b/inference-engine/src/vpu/graph_transformer/src/frontend/unroll_loops.cpp
index 28386c243a289e..34ca4835a9210a 100644
--- a/inference-engine/src/vpu/graph_transformer/src/frontend/unroll_loops.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/frontend/unroll_loops.cpp
@@ -21,7 +21,7 @@ void FrontEnd::unrollLoops(ie::CNNNetwork& network) {
     env.log->trace("Unroll TensorIterator loops");
     VPU_LOGGER_SECTION(env.log);
 
-    if (!env.config.irWithVpuScalesDir.empty()) {
+    if (!env.config.compileConfig().irWithVpuScalesDir.empty()) {
         // TODO: Scale dumps does not work with IR, which contain Tensor Iterator layers, because we cannot serialize them. #-23429
         for (auto iterator = ie::details::CNNNetworkIterator(network); iterator != ie::details::CNNNetworkIterator(); ++iterator) {
             const auto& layer = *iterator;
@@ -30,11 +30,11 @@ void FrontEnd::unrollLoops(ie::CNNNetwork& network) {
         }
     }
 
-    if (env.config.forcePureTensorIterator) {
+    if (env.config.compileConfig().forcePureTensorIterator) {
         return;
     }
 
-    if (env.config.enableTensorIteratorUnrolling) {
+    if (env.config.compileConfig().enableTensorIteratorUnrolling) {
         ie::NetPass::UnrollTI(network);
     } else {
         // Try to convert network to a RNN sequence due to performance reasons
diff --git a/inference-engine/src/vpu/graph_transformer/src/graph_transformer.cpp b/inference-engine/src/vpu/graph_transformer/src/graph_transformer.cpp
index 644529866ba949..180632e2846871 100644
--- a/inference-engine/src/vpu/graph_transformer/src/graph_transformer.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/graph_transformer.cpp
@@ -42,6 +42,7 @@
 #include <vpu/utils/auto_scope.hpp>
 #include <vpu/utils/dot_io.hpp>
 #include <vpu/utils/file_system.hpp>
+#include <mvnc.h>
 
 namespace vpu {
 
@@ -55,7 +56,7 @@ thread_local CompileEnv* g_compileEnv = nullptr;
 
 }  // namespace
 
-CompileEnv::CompileEnv(Platform platform) : platform(platform) {}
+CompileEnv::CompileEnv(ncDevicePlatform_t platform) : platform(platform) {}
 
 const CompileEnv& CompileEnv::get() {
     IE_ASSERT(g_compileEnv != nullptr);
@@ -70,7 +71,7 @@ const CompileEnv* CompileEnv::getOrNull() {
     return g_compileEnv;
 }
 
-void CompileEnv::init(Platform platform, const CompilationConfig& config, const Logger::Ptr& log) {
+void CompileEnv::init(ncDevicePlatform_t platform, const PluginConfiguration& config, const Logger::Ptr& log) {
     g_compileEnv = new CompileEnv(platform);
     g_compileEnv->config = config;
     g_compileEnv->log = log;
@@ -79,31 +80,37 @@ void CompileEnv::init(Platform platform, const CompilationConfig& config, const
     g_compileEnv->profile.setLogger(log);
 #endif
 
-    if (platform == Platform::MYRIAD_2) {
-        g_compileEnv->config.hwOptimization = false;
+    if (platform == ncDevicePlatform_t::NC_MYRIAD_2) {
+        g_compileEnv->config.compileConfig().hwOptimization = false;
     }
 
-    VPU_THROW_UNLESS(g_compileEnv->config.numSHAVEs <= g_compileEnv->config.numCMXSlices,
+    VPU_THROW_UNLESS(g_compileEnv->config.compileConfig().numSHAVEs <= g_compileEnv->config.compileConfig().numCMXSlices,
         R"(Value of configuration option ("{}") must be not greater than value of configuration option ("{}"), but {} > {} are provided)",
-        ie::MYRIAD_NUMBER_OF_SHAVES, ie::MYRIAD_NUMBER_OF_CMX_SLICES, config.numSHAVEs, config.numCMXSlices);
+        ie::MYRIAD_NUMBER_OF_SHAVES, ie::MYRIAD_NUMBER_OF_CMX_SLICES, config.compileConfig().numSHAVEs, config.compileConfig().numCMXSlices);
 
-    const auto numExecutors = config.numExecutors != -1 ? config.numExecutors : DefaultAllocation::numStreams(platform, config);
+    const auto numExecutors = config.compileConfig().numExecutors != -1 ? config.compileConfig().numExecutors : DefaultAllocation::numStreams(platform, config);
     VPU_THROW_UNLESS(numExecutors >= 1 && numExecutors <= DeviceResources::numStreams(),
         R"(Value of configuration option ("{}") must be in the range [{}, {}], actual is "{}")",
         ie::MYRIAD_THROUGHPUT_STREAMS, 1, DeviceResources::numStreams(), numExecutors);
 
-    const auto numSlices  = config.numCMXSlices != -1 ? config.numCMXSlices : DefaultAllocation::numSlices(platform, numExecutors);
+    const auto numSlices  = config.compileConfig().numCMXSlices != -1
+        ? config.compileConfig().numCMXSlices
+        : DefaultAllocation::numSlices(platform, numExecutors);
     VPU_THROW_UNLESS(numSlices >= 1 && numSlices <= DeviceResources::numSlices(platform),
         R"(Value of configuration option ("{}") must be in the range [{}, {}], actual is "{}")",
         ie::MYRIAD_NUMBER_OF_CMX_SLICES, 1, DeviceResources::numSlices(platform), numSlices);
 
     int defaultCmxLimit = DefaultAllocation::tilingCMXLimit(numSlices);
-    const auto tilingCMXLimit  = config.tilingCMXLimitKB != -1 ? std::min(config.tilingCMXLimitKB * 1024, defaultCmxLimit) : defaultCmxLimit;
+    const auto tilingCMXLimit  = config.compileConfig().tilingCMXLimitKB != -1
+        ? std::min(config.compileConfig().tilingCMXLimitKB * 1024, defaultCmxLimit)
+        : defaultCmxLimit;
     VPU_THROW_UNLESS(tilingCMXLimit >= 0,
         R"(Value of configuration option ("{}") must be greater than {}, actual is "{}")",
         ie::MYRIAD_TILING_CMX_LIMIT_KB, 0, tilingCMXLimit);
 
-    const auto numShaves = config.numSHAVEs != -1 ? config.numSHAVEs : DefaultAllocation::numShaves(platform, numExecutors, numSlices);
+    const auto numShaves = config.compileConfig().numSHAVEs != -1
+        ? config.compileConfig().numSHAVEs
+        : DefaultAllocation::numShaves(platform, numExecutors, numSlices);
     VPU_THROW_UNLESS(numShaves >= 1 && numShaves <= DeviceResources::numShaves(platform),
         R"(Value of configuration option ("{}") must be in the range [{}, {}], actual is "{}")",
         ie::MYRIAD_NUMBER_OF_SHAVES, 1, DeviceResources::numShaves(platform), numShaves);
@@ -123,7 +130,7 @@ void CompileEnv::init(Platform platform, const CompilationConfig& config, const
     g_compileEnv->initialized = true;
 }
 
-void CompileEnv::updateConfig(const CompilationConfig& config) {
+void CompileEnv::updateConfig(const PluginConfiguration& config) {
     IE_ASSERT(g_compileEnv != nullptr);
     IE_ASSERT(g_compileEnv->initialized);
 
@@ -165,9 +172,9 @@ CompiledGraph::Ptr compileImpl(const ie::CNNNetwork& network, const ie::ICore* c
 
     middleEnd->run(model);
 
-    if (!env.config.irWithVpuScalesDir.empty()) {
-        network.serialize(env.config.irWithVpuScalesDir + "/" + network.getName() + "_scales.xml",
-                          env.config.irWithVpuScalesDir + "/" + network.getName() + "_scales.bin");
+    if (!env.config.compileConfig().irWithVpuScalesDir.empty()) {
+        network.serialize(env.config.compileConfig().irWithVpuScalesDir + "/" + network.getName() + "_scales.xml",
+                          env.config.compileConfig().irWithVpuScalesDir + "/" + network.getName() + "_scales.bin");
     }
 
     return backEnd->build(model, frontEnd->origLayers());
@@ -191,8 +198,8 @@ CompiledGraph::Ptr compileImpl(const Model& model) {
 
 }  // namespace
 
-CompiledGraph::Ptr compileNetwork(const ie::CNNNetwork& network, Platform platform, const CompilationConfig& config, const Logger::Ptr& log,
-    const ie::ICore* core) {
+CompiledGraph::Ptr compileNetwork(const ie::CNNNetwork& network, ncDevicePlatform_t platform, const PluginConfiguration& config, const Logger::Ptr& log,
+                                  const ie::ICore* core) {
     CompileEnv::init(platform, config, log);
     AutoScope autoDeinit([] {
         CompileEnv::free();
@@ -205,8 +212,8 @@ CompiledGraph::Ptr compileNetwork(const ie::CNNNetwork& network, Platform platfo
 
 CompiledGraph::Ptr compileModel(
         const Model& model,
-        Platform platform,
-        const CompilationConfig& config,
+        ncDevicePlatform_t platform,
+        const PluginConfiguration& config,
         const Logger::Ptr& log) {
     CompileEnv::init(platform, config, log);
     AutoScope autoDeinit([] {
@@ -218,7 +225,7 @@ CompiledGraph::Ptr compileModel(
     return compileImpl(model);
 }
 
-CompiledGraph::Ptr compileSubNetwork(const ie::CNNNetwork& network, const CompilationConfig& subConfig, const ie::ICore* core) {
+CompiledGraph::Ptr compileSubNetwork(const ie::CNNNetwork& network, const PluginConfiguration& subConfig, const ie::ICore* core) {
     VPU_PROFILE(compileSubNetwork);
 
     const auto& env = CompileEnv::get();
@@ -238,11 +245,11 @@ CompiledGraph::Ptr compileSubNetwork(const ie::CNNNetwork& network, const Compil
 //
 
 std::set<std::string> getSupportedLayers(
-        const ie::CNNNetwork& network,
-        Platform platform,
-        const CompilationConfig& config,
-        const Logger::Ptr& log,
-        const ie::ICore* core) {
+    const ie::CNNNetwork& network,
+    ncDevicePlatform_t platform,
+    const PluginConfiguration& config,
+    const Logger::Ptr& log,
+    const ie::ICore* core) {
     CompileEnv::init(platform, config, log);
     AutoScope autoDeinit([] {
         CompileEnv::free();
@@ -255,28 +262,28 @@ std::set<std::string> getSupportedLayers(
     return frontEnd->checkSupportedLayers(network);
 }
 
-int DeviceResources::numShaves(const Platform& platform) {
-    return platform == Platform::MYRIAD_2 ? 12 : 16;
+int DeviceResources::numShaves(const ncDevicePlatform_t& platform) {
+    return platform == ncDevicePlatform_t::NC_MYRIAD_2 ? 12 : 16;
 }
 
-int DeviceResources::numSlices(const Platform& platform) {
-    return platform == Platform::MYRIAD_2 ? 12 : 19;
+int DeviceResources::numSlices(const ncDevicePlatform_t& platform) {
+    return platform == ncDevicePlatform_t::NC_MYRIAD_2 ? 12 : 19;
 }
 
 int DeviceResources::numStreams() {
     return 3;
 }
 
-int DefaultAllocation::numStreams(const Platform& platform, const CompilationConfig& configuration) {
-    return platform == Platform::MYRIAD_X && configuration.hwOptimization ? 2 : 1;
+int DefaultAllocation::numStreams(const ncDevicePlatform_t& platform, const PluginConfiguration& configuration) {
+    return platform == ncDevicePlatform_t::NC_MYRIAD_X && configuration.compileConfig().hwOptimization ? 2 : 1;
 }
 
-int DefaultAllocation::numSlices(const Platform& platform, int numStreams) {
+int DefaultAllocation::numSlices(const ncDevicePlatform_t& platform, int numStreams) {
     const auto capabilities = DeviceResources::numSlices(platform);
     return capabilities / numStreams;
 }
 
-int DefaultAllocation::numShaves(const Platform& platform, int numStreams, int numSlices) {
+int DefaultAllocation::numShaves(const ncDevicePlatform_t& platform, int numStreams, int numSlices) {
     const auto numAvailableShaves = DeviceResources::numShaves(platform);
     if (numStreams == 1) {
         return numAvailableShaves;
diff --git a/inference-engine/src/vpu/graph_transformer/src/middleend/pass_manager.cpp b/inference-engine/src/vpu/graph_transformer/src/middleend/pass_manager.cpp
index c40588acb93a81..f5e5d6ff2dfb96 100644
--- a/inference-engine/src/vpu/graph_transformer/src/middleend/pass_manager.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/middleend/pass_manager.cpp
@@ -10,6 +10,7 @@
 #include <string>
 
 #include <vpu/compile_env.hpp>
+#include <vpu/configuration/options/copy_optimization.hpp>
 
 namespace vpu {
 
@@ -93,7 +94,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     ADD_PASS(convertShapeNotation);
     ADD_DUMP_PASS("convertShapeNotation");
 
-    if (!env.config.disableReorder && !env.config.hwOptimization) {
+    if (!env.config.compileConfig().disableReorder && !env.config.compileConfig().hwOptimization) {
         ADD_PASS(reorderInputsToChannelMinor);
         ADD_DUMP_PASS("reorderInputsToChannelMinor");
     }
@@ -125,7 +126,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     // To overcome fp16 limitations
     //
 
-    if (env.config.hwOptimization && env.config.enableWeightsAnalysis) {
+    if (env.config.compileConfig().hwOptimization && env.config.compileConfig().enableWeightsAnalysis) {
         ADD_PASS(analyzeWeightableLayers);
         ADD_DUMP_PASS("analyzeWeightableLayers");
     }
@@ -150,7 +151,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     // Model HW-specific optimizations
     //
 
-    if (env.config.hwOptimization) {
+    if (env.config.compileConfig().hwOptimization) {
         ADD_PASS(replaceFCbyConv);
         ADD_DUMP_PASS("replaceFCbyConv");
 
@@ -161,7 +162,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
         ADD_PASS(replaceDeconvByConv);
         ADD_DUMP_PASS("replaceDeconvByConv");
 
-        if (env.config.hwDilation) {
+        if (env.config.compileConfig().hwDilation) {
             ADD_PASS(reshapeDilationConv);
             ADD_DUMP_PASS("reshapeDilationConv");
         }
@@ -173,7 +174,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
         // Pass should be located before "adjustDataBatch" because "adjustDataBatch" specifies "origConvOutput" attribute
         // for convolution in order to provide that information to "hwConvTiling" pass.
         // Otherwise, "hwConvTiling" will see incorrect values in "origConvOutput" attribute.
-        if (env.config.enableCustomReshapeParam) {
+        if (env.config.compileConfig().enableCustomReshapeParam) {
             ADD_PASS(reshapeBeforeConvTiling);
             ADD_DUMP_PASS("reshapeBeforeConvTiling");
         }
@@ -197,7 +198,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     ADD_PASS(hwPadding);
     ADD_DUMP_PASS("hwPadding");
 
-    if (env.config.hwOptimization) {
+    if (env.config.compileConfig().hwOptimization) {
         ADD_PASS(splitLargeKernelConv);
         ADD_DUMP_PASS("splitLargeKernelConv");
     }
@@ -209,7 +210,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     ADD_PASS(adjustDataBatch);
     ADD_DUMP_PASS("adjustDataBatch");
 
-    if (env.config.enableReplWithSCRelu) {
+    if (env.config.compileConfig().enableReplWithSCRelu) {
         ADD_PASS(replaceWithSCReLU);
         ADD_DUMP_PASS("replaceWithSCReLU");
     }
@@ -218,13 +219,13 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     // HW stages tiling
     //
 
-    if (env.config.hwOptimization) {
+    if (env.config.compileConfig().hwOptimization) {
         ADD_PASS(hwConvTiling);
         ADD_PASS(hwPoolTiling);
         ADD_PASS(hwFullyConnectedTiling);
         ADD_DUMP_PASS("hwTiling");
 
-        if (env.config.hwExtraSplit) {
+        if (env.config.compileConfig().hwExtraSplit) {
             ADD_PASS(hwExtraSplit);
             ADD_DUMP_PASS("hwExtraSplit");
         }
@@ -242,7 +243,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     //
     // this stage should be executed after "hwPoolTiling"
     // and before "swPoolAdaptation"
-    if (env.config.enableReplaceWithReduceMean) {
+    if (env.config.compileConfig().enableReplaceWithReduceMean) {
         ADD_PASS(replaceWithReduceMean);
         ADD_DUMP_PASS("replaceWithReduceMean");
     }
@@ -261,7 +262,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     ADD_PASS(mergeReLUAndBias);
     ADD_DUMP_PASS("mergeReLUAndBias");
 
-    if (env.config.enableEarlyEltwiseReLUFusion) {
+    if (env.config.compileConfig().enableEarlyEltwiseReLUFusion) {
         ADD_PASS(mergeEltwiseAndReLUDynamic);
         ADD_DUMP_PASS("mergeEltwiseAndReLUDynamic");
     }
@@ -279,7 +280,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
 
     // TODO: mergePermute support for reorder stage too.
     // TODO: pass that will swap Permute and per-element operations.
-    if (env.config.enablePermuteMerging) {
+    if (env.config.compileConfig().enablePermuteMerging) {
         ADD_PASS(mergePermuteStages);
         ADD_DUMP_PASS("mergePermuteStages");
     }
@@ -326,7 +327,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     // Model common optimizations
     //
 
-    if (env.config.copyOptimization.getOrDefault(true)) {
+    if (env.config.get<CopyOptimizationOption>()) {
         ADD_PASS(eliminateCopyStages);
         ADD_DUMP_PASS("eliminateCopyStages");
     }
@@ -334,7 +335,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     //
     // HW/SW injection
 
-    if (env.config.hwOptimization && env.config.injectSwOps.getOrDefault(true)) {
+    if (env.config.compileConfig().hwOptimization && env.config.compileConfig().injectSwOps.getOrDefault(true)) {
         ADD_PASS(injectSw);
         ADD_DUMP_PASS("injectSw");
     }
@@ -350,7 +351,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     // HW stages finalization
     //
 
-    if (env.config.hwOptimization) {
+    if (env.config.compileConfig().hwOptimization) {
         ADD_PASS(finalizeHwOps);
         ADD_DUMP_PASS("hwFinalization");
     }
@@ -361,7 +362,7 @@ PassSet::Ptr PassManager::buildMiddleEnd() {
     ADD_PASS(markFastStages);
     ADD_DUMP_PASS("markFastStages");
 
-    if (env.config.enableMemoryTypesAnnotation) {
+    if (env.config.compileConfig().enableMemoryTypesAnnotation) {
         ADD_PASS(annotateMemoryTypes);
         ADD_DUMP_PASS("annotateMemoryTypes");
     }
diff --git a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/adjust_data_location.cpp b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/adjust_data_location.cpp
index c734ef2a06bbef..4990b8153ca0aa 100644
--- a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/adjust_data_location.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/adjust_data_location.cpp
@@ -48,7 +48,7 @@ void PassImpl::run(const Model& model) {
     allocNonIntermediateData(model);
     adjustModelForMemReqs(model);
     copyHwMisalignedInput(model);
-    if (env.config.packDataInCmx.getOrDefault(true)) {
+    if (env.config.compileConfig().packDataInCmx.getOrDefault(true)) {
         packDataInCmx(model);
     }
 }
@@ -147,7 +147,7 @@ void PassImpl::collectMemReqs(const Model& model) {
 }
 
 void PassImpl::resetStageOrder(const Model& model) {
-    if (!CompileEnv::get().config.hwOptimization)
+    if (!CompileEnv::get().config.compileConfig().hwOptimization)
         return;
 
     static const std::string s_expectCMXOutput {"expectCMXOutput"};
diff --git a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/eliminate_copy.cpp b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/eliminate_copy.cpp
index 74a42d8a13e6cc..9d9bf482e75a06 100644
--- a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/eliminate_copy.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/eliminate_copy.cpp
@@ -14,6 +14,7 @@
 
 #include <vpu/middleend/allocator/allocator.hpp>
 #include <vpu/compile_env.hpp>
+#include <vpu/configuration/options/copy_optimization.hpp>
 
 namespace vpu {
 
@@ -78,7 +79,7 @@ void PassImpl::run(const Model& model) {
 
     std::queue<Stage> copyToRemove;
 
-    if (!env.config.copyOptimization.hasValue()) {
+    if (!env.config.get<CopyOptimizationOption>()) {
         int nCopyStages = 0;
         for (const auto& stage : model->getStages()) {
             if (stage->type() == StageType::Copy) {
diff --git a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/inject_sw.cpp b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/inject_sw.cpp
index d8827d522d276d..4fb8c5c01232a4 100644
--- a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/inject_sw.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/inject_sw.cpp
@@ -68,7 +68,7 @@ void PassImpl::run(const Model& model) {
     // Collect HW and SW candidates
     //
 
-    if (!env.config.injectSwOps.hasValue() &&
+    if (!env.config.compileConfig().injectSwOps.hasValue() &&
         model->numStages() > nMaxStagesForInjectSw) {
         env.log->warning(
             "Pass [injectSw] SKIPPED : number of stages (%d) is larger than threshold %d",
diff --git a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/merge_eltwise_and_relu.cpp b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/merge_eltwise_and_relu.cpp
index 4684cc50ee6691..548557b594b1dc 100644
--- a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/merge_eltwise_and_relu.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/merge_eltwise_and_relu.cpp
@@ -30,7 +30,7 @@ class PassImpl final : public Pass {
 };
 
 void PassImpl::run(const Model& model) {
-    const bool enableEarlyEltwiseReLUFusion = CompileEnv::get().config.enableEarlyEltwiseReLUFusion;
+    const bool enableEarlyEltwiseReLUFusion = CompileEnv::get().config.compileConfig().enableEarlyEltwiseReLUFusion;
     if (enableEarlyEltwiseReLUFusion) {
         if (m_mode == MergeMode::DYNAMIC_NETWORK) {
             VPU_PROFILE(mergeEltwiseAndReLUDynamic);
diff --git a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/merge_hw_stages.cpp b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/merge_hw_stages.cpp
index 5bade8b8bbce8d..5b83fd0e49b475 100644
--- a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/merge_hw_stages.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/merge_hw_stages.cpp
@@ -170,7 +170,7 @@ void PassImpl::run(const Model& model) {
         // Try to merge next Pooling layer
         //
 
-        if (env.config.mergeHwPoolToConv) {
+        if (env.config.compileConfig().mergeHwPoolToConv) {
             if (stage->type() == StageType::StubConv) {
                 if (auto nextPoolStage = getNextPoolStage(stage, output)) {
                     output = nextPoolStage->output(0);
diff --git a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/replace_deconv_by_conv.cpp b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/replace_deconv_by_conv.cpp
index beb1e55cbb84bd..4ca08a4f65f100 100644
--- a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/replace_deconv_by_conv.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/replace_deconv_by_conv.cpp
@@ -148,7 +148,7 @@ void PassImpl::run(const Model& model) {
         auto output = stage->output(0);
         const auto& env = CompileEnv::get();
 
-        if (env.config.hwDisabled(stage->origLayer()->name)) {
+        if (env.config.compileConfig().hwDisabled(stage->origLayer()->name)) {
             continue;
         }
 
diff --git a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/weights_analysis.cpp b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/weights_analysis.cpp
index 27e703ec4f53cc..24efadc78574b5 100644
--- a/inference-engine/src/vpu/graph_transformer/src/middleend/passes/weights_analysis.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/middleend/passes/weights_analysis.cpp
@@ -88,7 +88,7 @@ bool isScalable(const Stage& stage) {
 
 bool checkGrowingOutput(const Model& model) {
     const auto& env = CompileEnv::get();
-    if (!env.config.checkPreprocessingInsideModel) {
+    if (!env.config.compileConfig().checkPreprocessingInsideModel) {
         return false;
     }
 
@@ -258,7 +258,7 @@ void PassImpl::run(const Model& model) {
                 scale = static_cast<float>(1ULL << static_cast<std::uint32_t>(shift));
             }
 
-            if (!env.config.irWithVpuScalesDir.empty()) {
+            if (!env.config.compileConfig().irWithVpuScalesDir.empty()) {
                 stage->origLayer()->params["vpu_scale"] = toString(scale);
             }
         }
diff --git a/inference-engine/src/vpu/graph_transformer/src/model/stage.cpp b/inference-engine/src/vpu/graph_transformer/src/model/stage.cpp
index 7323587ce9f38b..cad7da95ca8e41 100644
--- a/inference-engine/src/vpu/graph_transformer/src/model/stage.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/model/stage.cpp
@@ -199,7 +199,7 @@ StageSHAVEsRequirements StageNode::getSHAVEsRequirements() const {
 
     // return max for Myriad2
     const auto& compileEnv = CompileEnv::get();
-    if (compileEnv.platform == Platform::MYRIAD_2) {
+    if (compileEnv.platform == ncDevicePlatform_t::NC_MYRIAD_2) {
         return StageSHAVEsRequirements::NeedMax;
     }
 
diff --git a/inference-engine/src/vpu/graph_transformer/src/stages/activation.cpp b/inference-engine/src/vpu/graph_transformer/src/stages/activation.cpp
index 04c4923fcae189..7f666964414e59 100644
--- a/inference-engine/src/vpu/graph_transformer/src/stages/activation.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/stages/activation.cpp
@@ -24,7 +24,7 @@ void FrontEnd::parseActivation(const Model& model, const ie::CNNLayerPtr& layer,
     const auto type = layer->GetParamAsString("type");
 
     const auto activationParserIt = activationParsers.find(type);
-    VPU_THROW_UNSUPPORTED_UNLESS(activationParserIt != activationParsers.end(),
+    VPU_THROW_UNSUPPORTED_LAYER_UNLESS(activationParserIt != activationParsers.end(),
                                  "Failed to compile layer \"%v\"(type = %v) ", layer->name, type);
 
     activationParserIt->second(model, layer, inputs, outputs);
diff --git a/inference-engine/src/vpu/graph_transformer/src/stages/convolution.cpp b/inference-engine/src/vpu/graph_transformer/src/stages/convolution.cpp
index 529456f919ebcd..695bd43f3c9bf5 100644
--- a/inference-engine/src/vpu/graph_transformer/src/stages/convolution.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/stages/convolution.cpp
@@ -163,9 +163,9 @@ void parseConv2D(const Model      & model,
                           kernelStrideY,
                           dilationX,
                           dilationY,
-                          env.config.hwOptimization,
-                          env.config.hwDilation,
-                          env.config.hwDisabled(layer->name));
+                          env.config.compileConfig().hwOptimization,
+                          env.config.compileConfig().hwDilation,
+                          env.config.compileConfig().hwDisabled(layer->name));
 
     //
     // Create const datas
@@ -476,9 +476,9 @@ void parseConvND(const Model      & model,
                           strides[1],
                           dilations[0],
                           dilations[1],
-                          env.config.hwOptimization,
-                          env.config.hwDilation,
-                          env.config.hwDisabled(layer->name));
+                          env.config.compileConfig().hwOptimization,
+                          env.config.compileConfig().hwDilation,
+                          env.config.compileConfig().hwDisabled(layer->name));
 
     int try_hw = tryHW ? 1 : 0;
 
diff --git a/inference-engine/src/vpu/graph_transformer/src/stages/fc.cpp b/inference-engine/src/vpu/graph_transformer/src/stages/fc.cpp
index c2994273ca7401..96da14dc941a12 100644
--- a/inference-engine/src/vpu/graph_transformer/src/stages/fc.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/stages/fc.cpp
@@ -37,13 +37,13 @@ void FrontEnd::parseFullyConnected(const Model& model, const ie::CNNLayerPtr& _l
     // Check if HW is applicable
     //
 
-    auto tryHW = env.config.hwOptimization;
+    auto tryHW = env.config.compileConfig().hwOptimization;
 
     if (output->desc().dim(Dim::W, 1) != 1 || output->desc().dim(Dim::H, 1) != 1) {
         tryHW = false;
     }
 
-    if (env.config.hwDisabled(layer->name)) {
+    if (env.config.compileConfig().hwDisabled(layer->name)) {
         tryHW = false;
     }
 
diff --git a/inference-engine/src/vpu/graph_transformer/src/stages/mtcnn.cpp b/inference-engine/src/vpu/graph_transformer/src/stages/mtcnn.cpp
index 795b45c8a04486..e9d6899f7abc6a 100644
--- a/inference-engine/src/vpu/graph_transformer/src/stages/mtcnn.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/stages/mtcnn.cpp
@@ -162,7 +162,7 @@ void FrontEnd::parseMTCNN(const Model& model, const ie::CNNLayerPtr& layer, cons
     IE_ASSERT(inputs.size() == 1);
     IE_ASSERT(outputs.size() == 1);
 
-    if (!env.config.hwOptimization) {
+    if (!env.config.compileConfig().hwOptimization) {
         VPU_THROW_EXCEPTION << "MTCNN layer supports Myriad X with NCE only";
     }
 
diff --git a/inference-engine/src/vpu/graph_transformer/src/stages/permute.cpp b/inference-engine/src/vpu/graph_transformer/src/stages/permute.cpp
index 411cc2de0d7438..071c2fb0328d3e 100644
--- a/inference-engine/src/vpu/graph_transformer/src/stages/permute.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/stages/permute.cpp
@@ -124,7 +124,7 @@ Stage StageBuilder::addReorderStage(
         const Data& output) {
     const auto* env = CompileEnv::getOrNull();
     VPU_THROW_UNLESS(
-        env == nullptr || !env->config.disableReorder,
+        env == nullptr || !env->config.compileConfig().disableReorder,
         "Tried to add Reorder Stage %v, while DISABLE_REORDER option was set",
         name);
 
diff --git a/inference-engine/src/vpu/graph_transformer/src/stages/pooling.cpp b/inference-engine/src/vpu/graph_transformer/src/stages/pooling.cpp
index 5551e72231bd44..2823d0c9165424 100644
--- a/inference-engine/src/vpu/graph_transformer/src/stages/pooling.cpp
+++ b/inference-engine/src/vpu/graph_transformer/src/stages/pooling.cpp
@@ -221,8 +221,8 @@ void parsePool2D(const     Model      & model,
     //
 
     const auto& env = CompileEnv::get();
-    bool hwOptimization = env.config.hwOptimization;
-    bool hwDisabled = env.config.hwDisabled(layer->name);
+    bool hwOptimization = env.config.compileConfig().hwOptimization;
+    bool hwDisabled = env.config.compileConfig().hwDisabled(layer->name);
 
     int inputWidth = input->desc().dim(Dim::W);
     int inputHeight = input->desc().dim(Dim::H);
@@ -480,8 +480,8 @@ void parsePoolND(const     Model      & model,
     //
 
     const auto& env = CompileEnv::get();
-    bool hwOptimization = env.config.hwOptimization;
-    bool hwDisabled = env.config.hwDisabled(layer->name);
+    bool hwOptimization = env.config.compileConfig().hwOptimization;
+    bool hwDisabled = env.config.compileConfig().hwDisabled(layer->name);
 
     bool tryHW = canTryHW(poolLayer->_type,
                           input_shape[0],
diff --git a/inference-engine/src/vpu/myriad_plugin/configuration/myriad_configuration.cpp b/inference-engine/src/vpu/myriad_plugin/configuration/myriad_configuration.cpp
new file mode 100644
index 00000000000000..41a53f0619eb4b
--- /dev/null
+++ b/inference-engine/src/vpu/myriad_plugin/configuration/myriad_configuration.cpp
@@ -0,0 +1,31 @@
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "configuration/myriad_configuration.hpp"
+
+namespace vpu {
+
+MyriadConfiguration::MyriadConfiguration() {}
+
+void MyriadConfiguration::from(const std::map<std::string, std::string>& configuration) {
+    std::map<std::string, std::string> migratedOptions, notMigratedOptions;
+    for (const auto& entry : configuration) {
+        auto& destination = PluginConfiguration::supports(entry.first) ? migratedOptions : notMigratedOptions;
+        destination.emplace(entry);
+    }
+    PluginConfiguration::from(migratedOptions);
+    update(notMigratedOptions);
+}
+
+void MyriadConfiguration::fromAtRuntime(const std::map<std::string, std::string>& configuration) {
+    std::map<std::string, std::string> migratedOptions, notMigratedOptions;
+    for (const auto& entry : configuration) {
+        auto& destination = PluginConfiguration::supports(entry.first) ? migratedOptions : notMigratedOptions;
+        destination.emplace(entry);
+    }
+    PluginConfiguration::fromAtRuntime(migratedOptions);
+    update(notMigratedOptions, ConfigMode::RunTime);
+}
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/myriad_plugin/configuration/myriad_configuration.hpp b/inference-engine/src/vpu/myriad_plugin/configuration/myriad_configuration.hpp
new file mode 100644
index 00000000000000..080d262c7037cb
--- /dev/null
+++ b/inference-engine/src/vpu/myriad_plugin/configuration/myriad_configuration.hpp
@@ -0,0 +1,21 @@
+// Copyright (C) 2020 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "vpu/configuration/plugin_configuration.hpp"
+#include "myriad_config.h"
+
+namespace vpu {
+
+class MyriadConfiguration final : public PluginConfiguration, public MyriadPlugin::MyriadConfig {
+public:
+    MyriadConfiguration();
+
+    // TODO: remove once all options are migrated
+    void from(const std::map<std::string, std::string>& configuration);
+    void fromAtRuntime(const std::map<std::string, std::string>& configuration);
+};
+
+}  // namespace vpu
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_config.h b/inference-engine/src/vpu/myriad_plugin/myriad_config.h
index 176e677a9b6f1a..5745516e179189 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_config.h
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_config.h
@@ -34,7 +34,7 @@ VPU_DECLARE_ENUM(MovidiusDdrType,
     MICRON_1GB  = 4,
 )
 
-class MyriadConfig final : public ParsedConfig {
+class MyriadConfig : public virtual ParsedConfig {
 public:
     const std::string& pluginLogFilePath() const {
         return _pluginLogFilePath;
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.cpp b/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.cpp
index 7a0a22499c1942..f8806435188aa0 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.cpp
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.cpp
@@ -14,6 +14,7 @@
 #include <vpu/utils/runtime_graph.hpp>
 #include <legacy/net_pass.h>
 #include <vpu/compile_env.hpp>
+#include <vpu/configuration/options/log_level.hpp>
 
 using namespace InferenceEngine;
 
@@ -25,23 +26,24 @@ namespace MyriadPlugin {
 ExecutableNetwork::ExecutableNetwork(
         std::shared_ptr<IMvnc> mvnc,
         std::vector<DevicePtr>& devicePool,
-        const MyriadConfig& config,
+        const MyriadConfiguration& config,
         const ie::ICore* core) :
             _config(config),
             _core(core) {
     VPU_PROFILE(ExecutableNetwork);
 
+    const auto& logLevel = _config.get<LogLevelOption>();
+
     _log = std::make_shared<Logger>(
         "MyriadPlugin",
-        _config.logLevel(),
+        logLevel,
         defaultOutput(_config.pluginLogFilePath()));
 
-    _executor = std::make_shared<MyriadExecutor>(_config.forceReset(), std::move(mvnc), _config.logLevel(), _log);
+    _executor = std::make_shared<MyriadExecutor>(_config.forceReset(), std::move(mvnc), logLevel, _log);
     _device = _executor->openDevice(devicePool, _config);
 
-    const auto& compileConfig = config.compileConfig();
     const auto& revision = _device->revision();
-    _actualNumExecutors = compileConfig.numExecutors != -1 ? compileConfig.numExecutors : DefaultAllocation::numStreams(revision, compileConfig);
+    _actualNumExecutors = config.compileConfig().numExecutors != -1 ? config.compileConfig().numExecutors : DefaultAllocation::numStreams(revision, config);
 
     _supportedMetrics = {
         METRIC_KEY(NETWORK_NAME),
@@ -56,22 +58,22 @@ ExecutableNetwork::ExecutableNetwork(
         const ie::CNNNetwork& network,
         std::shared_ptr<IMvnc> mvnc,
         std::vector<DevicePtr>& devicePool,
-        const MyriadConfig& config,
+        const MyriadConfiguration& config,
         const ie::ICore* core) :
             ExecutableNetwork(std::move(mvnc), devicePool, config, core) {
     VPU_PROFILE(ExecutableNetwork);
 
     const auto compilerLog = std::make_shared<Logger>(
         "GraphCompiler",
-        _config.logLevel(),
+        _config.get<LogLevelOption>(),
         defaultOutput(_config.compilerLogFilePath()));
 
     if (_device == nullptr)
         IE_THROW() << "No device was detected";
     auto compiledGraph = compileNetwork(
         network,
-        static_cast<Platform>(_device->_platform),
-        _config.compileConfig(),
+        _device->_platform,
+        _config,
         compilerLog,
         _core);
 
@@ -100,9 +102,7 @@ ExecutableNetwork::ExecutableNetwork(
     }
 }
 
-void ExecutableNetwork::Import(std::istream& strm,
-                               std::vector<DevicePtr> &devicePool,
-                               const MyriadConfig& config) {
+void ExecutableNetwork::Import(std::istream& strm, std::vector<DevicePtr> &devicePool, const MyriadConfiguration& configuration) {
     auto currentPos = strm.tellg();
     strm.seekg(0, strm.end);
     auto blobSize = strm.tellg() - currentPos;
@@ -147,11 +147,8 @@ void ExecutableNetwork::Import(std::istream& strm,
     }
 }
 
-ExecutableNetwork::ExecutableNetwork(std::istream& strm,
-                               std::shared_ptr<IMvnc> mvnc,
-                               std::vector<DevicePtr> &devicePool,
-                               const MyriadConfig& config,
-                               const ie::ICore* core) :
+ExecutableNetwork::ExecutableNetwork(std::istream& strm, std::shared_ptr<IMvnc> mvnc, std::vector<DevicePtr> &devicePool,
+    const MyriadConfiguration& config, const ie::ICore* core) :
     ExecutableNetwork(std::move(mvnc), devicePool, config, core) {
     VPU_PROFILE(ExecutableNetwork);
     Import(strm, devicePool, config);
@@ -161,7 +158,7 @@ ExecutableNetwork::ExecutableNetwork(
         const std::string& blobFilename,
         std::shared_ptr<IMvnc> mvnc,
         std::vector<DevicePtr>& devicePool,
-        const MyriadConfig& config,
+        const MyriadConfiguration& config,
         const ie::ICore* core) :
     ExecutableNetwork(std::move(mvnc), devicePool, config, core) {
     VPU_PROFILE(ExecutableNetwork);
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.h b/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.h
index 22824ee5ec1c47..0b58bf2f4802e9 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.h
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_executable_network.h
@@ -32,23 +32,14 @@ class ExecutableNetwork : public ie::ExecutableNetworkThreadSafeDefault {
 public:
     typedef std::shared_ptr<ExecutableNetwork> Ptr;
 
-    explicit ExecutableNetwork(const ie::CNNNetwork& network,
-                               std::shared_ptr<IMvnc> mvnc,
-                               std::vector<DevicePtr> &devicePool,
-                               const MyriadConfig& config,
-                               const ie::ICore* core);
+    ExecutableNetwork(const InferenceEngine::CNNNetwork& network, std::shared_ptr<IMvnc> mvnc, std::vector<DevicePtr> &devicePool,
+                      const MyriadConfiguration& configuration, const ie::ICore* core);
 
-    explicit ExecutableNetwork(std::istream& strm,
-                               std::shared_ptr<IMvnc> mvnc,
-                               std::vector<DevicePtr> &devicePool,
-                               const MyriadConfig& config,
-                               const ie::ICore* core);
+    ExecutableNetwork(std::istream& strm, std::shared_ptr<IMvnc> mvnc, std::vector<DevicePtr> &devicePool, const MyriadConfiguration& configuration,
+                      const ie::ICore* core);
 
-    explicit ExecutableNetwork(const std::string &blobFilename,
-                               std::shared_ptr<IMvnc> mvnc,
-                               std::vector<DevicePtr> &devicePool,
-                               const MyriadConfig& config,
-                               const ie::ICore* core);
+    ExecutableNetwork(const std::string &blobFilename, std::shared_ptr<IMvnc> mvnc, std::vector<DevicePtr> &devicePool,
+                      const MyriadConfiguration& configuration, const ie::ICore* core);
 
 
     virtual ~ExecutableNetwork() {
@@ -97,9 +88,7 @@ class ExecutableNetwork : public ie::ExecutableNetworkThreadSafeDefault {
 
     ie::CNNNetwork GetExecGraphInfo() override;
 
-    void Import(std::istream& strm,
-                std::vector<DevicePtr> &devicePool,
-                const MyriadConfig& config);
+    void Import(std::istream& strm, std::vector<DevicePtr> &devicePool, const MyriadConfiguration& configuration);
 
 private:
     Logger::Ptr _log;
@@ -108,7 +97,7 @@ class ExecutableNetwork : public ie::ExecutableNetworkThreadSafeDefault {
     GraphDesc _graphDesc;
     DevicePtr _device;
     GraphMetaInfo _graphMetaData;
-    MyriadConfig _config;
+    MyriadConfiguration _config;
     const ie::ICore* _core = nullptr;
     int _actualNumExecutors = 0;
     std::vector<std::string> _supportedMetrics;
@@ -119,10 +108,7 @@ class ExecutableNetwork : public ie::ExecutableNetworkThreadSafeDefault {
     const size_t _maxTaskExecutorGetResultCount = 1;
     std::queue<std::string> _taskExecutorGetResultIds;
 
-    ExecutableNetwork(std::shared_ptr<IMvnc> mvnc,
-        std::vector<DevicePtr> &devicePool,
-        const MyriadConfig& config,
-        const ie::ICore* core);
+    ExecutableNetwork(std::shared_ptr<IMvnc> mvnc, std::vector<DevicePtr> &devicePool, const MyriadConfiguration& config, const ie::ICore* core);
 
     ie::ITaskExecutor::Ptr getNextTaskExecutor() {
         std::string id = _taskExecutorGetResultIds.front();
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_executor.cpp b/inference-engine/src/vpu/myriad_plugin/myriad_executor.cpp
index d92a948ce5727b..49cf63c4e636c5 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_executor.cpp
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_executor.cpp
@@ -73,8 +73,7 @@ MyriadExecutor::MyriadExecutor(bool forceReset, std::shared_ptr<IMvnc> mvnc,
 /*
  * @brief Boot available device
  */
-ncStatus_t MyriadExecutor::bootNextDevice(std::vector<DevicePtr> &devicePool,
-                                          const MyriadConfig& config) {
+ncStatus_t MyriadExecutor::bootNextDevice(std::vector<DevicePtr> &devicePool, const MyriadConfiguration& config) {
     VPU_PROFILE(bootNextDevice);
 // #-17972, #-16790
 #if defined(NO_BOOT)
@@ -221,7 +220,7 @@ ncStatus_t MyriadExecutor::bootNextDevice(std::vector<DevicePtr> &devicePool,
 }
 
 DevicePtr MyriadExecutor::openDevice(std::vector<DevicePtr>& devicePool,
-                                     const MyriadConfig& config) {
+                                     const MyriadConfiguration& config) {
     VPU_PROFILE(openDevice);
     std::lock_guard<std::mutex> lock(device_mutex);
 
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_executor.h b/inference-engine/src/vpu/myriad_plugin/myriad_executor.h
index e7c9c8614ee91b..d9883d8501e843 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_executor.h
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_executor.h
@@ -13,6 +13,7 @@
 
 #include <mvnc.h>
 #include "myriad_mvnc_wrapper.h"
+#include "configuration/myriad_configuration.hpp"
 
 #include <ie_parameter.hpp>
 
@@ -63,9 +64,9 @@ struct DeviceDesc {
                 ((config.protocol() == NC_ANY_PROTOCOL) || (_protocol == config.protocol()));
     }
 
-    Platform revision() const {
+    ncDevicePlatform_t revision() const {
         VPU_THROW_UNLESS(_platform != NC_ANY_PLATFORM, "Cannot get a revision from not booted device");
-        return _platform == NC_MYRIAD_2 ? Platform::MYRIAD_2 : Platform::MYRIAD_X;
+        return _platform;
     }
 };
 
@@ -86,7 +87,7 @@ class MyriadExecutor {
      * @brief Get myriad device
      * @return Already booted and empty device or new booted device
      */
-    DevicePtr openDevice(std::vector<DevicePtr> &devicePool, const MyriadConfig& config);
+    DevicePtr openDevice(std::vector<DevicePtr> &devicePool, const MyriadConfiguration& config);
 
     static void closeDevices(std::vector<DevicePtr> &devicePool, std::shared_ptr<IMvnc> mvnc);
 
@@ -134,8 +135,7 @@ class MyriadExecutor {
      * @param configPlatform Boot the selected platform
      * @param configProtocol Boot device with selected protocol
      */
-    ncStatus_t bootNextDevice(std::vector<DevicePtr> &devicePool,
-                              const MyriadConfig& config);
+    ncStatus_t bootNextDevice(std::vector<DevicePtr> &devicePool, const MyriadConfiguration& config);
 };
 
 typedef std::shared_ptr<MyriadExecutor> MyriadExecutorPtr;
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_infer_request.h b/inference-engine/src/vpu/myriad_plugin/myriad_infer_request.h
index 667b93b6aae1b5..9373f771fb927a 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_infer_request.h
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_infer_request.h
@@ -14,6 +14,7 @@
 
 #include <vpu/utils/logger.hpp>
 #include <vpu/utils/ie_helpers.hpp>
+#include <vpu/graph_transformer.hpp>
 
 #include "myriad_executor.h"
 #include "myriad_config.h"
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_metrics.cpp b/inference-engine/src/vpu/myriad_plugin/myriad_metrics.cpp
index fc022ff4184eae..e91ef8333fe83d 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_metrics.cpp
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_metrics.cpp
@@ -31,6 +31,7 @@ MyriadMetrics::MyriadMetrics() {
     };
 
 IE_SUPPRESS_DEPRECATED_START
+    // TODO: remove once all options are migrated
     _supportedConfigKeys = {
         MYRIAD_ENABLE_HW_ACCELERATION,
         MYRIAD_ENABLE_RECEIVING_TENSOR_TIME,
@@ -45,7 +46,6 @@ IE_SUPPRESS_DEPRECATED_START
         KEY_VPU_MYRIAD_FORCE_RESET,
         KEY_VPU_MYRIAD_PLATFORM,
 
-        CONFIG_KEY(LOG_LEVEL),
         CONFIG_KEY(EXCLUSIVE_ASYNC_REQUESTS),
         CONFIG_KEY(PERF_COUNT),
         CONFIG_KEY(CONFIG_FILE),
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_plugin.cpp b/inference-engine/src/vpu/myriad_plugin/myriad_plugin.cpp
index 75e7ef395d9266..7771cc8bfbf88c 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_plugin.cpp
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_plugin.cpp
@@ -18,8 +18,9 @@
 #include <vpu/utils/profiling.hpp>
 #include <vpu/utils/error.hpp>
 #include <vpu/ngraph/query_network.hpp>
-#include <transformations/common_optimizations/common_optimizations.hpp>
-#include <ngraph/pass/manager.hpp>
+
+#include <vpu/configuration/options/log_level.hpp>
+#include <vpu/configuration/options/copy_optimization.hpp>
 
 #include "myriad_plugin.h"
 
@@ -34,31 +35,40 @@ IExecutableNetworkInternal::Ptr Engine::LoadExeNetworkImpl(
         const std::map<std::string, std::string>& config) {
     VPU_PROFILE(LoadExeNetworkImpl);
 
-    auto parsedConfigCopy = _parsedConfig;
-    parsedConfigCopy.update(config);
+    auto executableNetworkConfiguration = _parsedConfig;
+    executableNetworkConfiguration.from(config);
+    executableNetworkConfiguration.validate();
 
-    return std::make_shared<ExecutableNetwork>(network, _mvnc, _devicePool, parsedConfigCopy, GetCore());
+    return std::make_shared<ExecutableNetwork>(network, _mvnc, _devicePool, executableNetworkConfiguration, GetCore());
 }
 
 void Engine::SetConfig(const std::map<std::string, std::string> &config) {
-    _parsedConfig.update(config);
+    _parsedConfig.from(config);
 
+    // TODO: remove once all options are migrated
     for (const auto& entry : config) {
         _config[entry.first] = entry.second;
     }
+
+#ifndef NDEBUG
+    if (const auto envVar = std::getenv("IE_VPU_LOG_LEVEL")) {
+        _parsedConfig.set(LogLevelOption::key(), envVar);
+    }
+#endif
 }
 
 Parameter Engine::GetConfig(const std::string& name, const std::map<std::string, Parameter>& options) const {
-    auto supported_keys = _metrics->SupportedConfigKeys();
-    if (std::find(supported_keys.begin(),
-        supported_keys.end(), name) == supported_keys.end()) {
-        IE_THROW() << "Unsupported config key : " << name;
-    }
+    // TODO: remove once all options are migrated
+    const auto& supportedKeys = _metrics->SupportedConfigKeys();
+    VPU_THROW_UNSUPPORTED_OPTION_UNLESS(supportedKeys.count(name) == 1 || _parsedConfig.supports(name), "Unsupported configuration key: {}", name);
 
     Parameter result;
-    auto option = _config.find(name);
-    if (option != _config.end())
-        result = option->second;
+    if (_parsedConfig.supports(name)) {
+        result = _parsedConfig.asParameter(name);
+    } else if (_config.count(name)) {
+        // TODO: remove once all options are migrated
+        result = _config.at(name);
+    }
 
     return result;
 }
@@ -70,7 +80,7 @@ QueryNetworkResult Engine::QueryNetwork(
     QueryNetworkResult res;
 
     auto parsedConfigCopy = _parsedConfig;
-    parsedConfigCopy.update(config);
+    parsedConfigCopy.from(config);
 
     const auto deviceName = parsedConfigCopy.deviceName();
     if (!deviceName.empty()) {
@@ -80,13 +90,13 @@ QueryNetworkResult Engine::QueryNetwork(
 
     const auto log = std::make_shared<Logger>(
             "GraphCompiler",
-            parsedConfigCopy.logLevel(),
+            _parsedConfig.get<LogLevelOption>(),
             defaultOutput(parsedConfigCopy.compilerLogFilePath()));
 
     const auto supportedLayers = getSupportedLayers(
             network,
-            static_cast<Platform>(parsedConfigCopy.platform()),
-            parsedConfigCopy.compileConfig(),
+            parsedConfigCopy.platform(),
+            parsedConfigCopy,
             log,
             GetCore());
 
@@ -111,6 +121,7 @@ Engine::Engine(std::shared_ptr<IMvnc> mvnc) :
 
     _pluginName = "MYRIAD";
 
+    // TODO: remove once all options are migrated
 IE_SUPPRESS_DEPRECATED_START
     _config = {
         { MYRIAD_ENABLE_HW_ACCELERATION, CONFIG_VALUE(YES) },
@@ -126,13 +137,19 @@ IE_SUPPRESS_DEPRECATED_START
         { KEY_VPU_MYRIAD_FORCE_RESET, CONFIG_VALUE(NO) },
         { KEY_VPU_MYRIAD_PLATFORM, "" },
 
-        { KEY_LOG_LEVEL, CONFIG_VALUE(LOG_NONE) },
         { KEY_EXCLUSIVE_ASYNC_REQUESTS, CONFIG_VALUE(NO) },
         { KEY_PERF_COUNT, CONFIG_VALUE(NO) },
         { KEY_CONFIG_FILE, "" },
         { KEY_DEVICE_ID, "" },
     };
 IE_SUPPRESS_DEPRECATED_END
+
+    _parsedConfig.registerOption<LogLevelOption>();
+    _parsedConfig.registerOption<CopyOptimizationOption>();
+
+IE_SUPPRESS_DEPRECATED_START
+    _parsedConfig.registerDeprecatedOption<LogLevelOption>(VPU_CONFIG_KEY(LOG_LEVEL));
+IE_SUPPRESS_DEPRECATED_END
 }
 
 InferenceEngine::IExecutableNetworkInternal::Ptr Engine::ImportNetwork(
@@ -140,14 +157,12 @@ InferenceEngine::IExecutableNetworkInternal::Ptr Engine::ImportNetwork(
         const std::map<std::string, std::string>& config) {
     VPU_PROFILE(ImportNetwork);
 
-    auto parsedConfigCopy = _parsedConfig;
-    parsedConfigCopy.update(config, ConfigMode::RunTime);
+    auto executableNetworkConfiguration = _parsedConfig;
+    executableNetworkConfiguration.fromAtRuntime(config);
+    executableNetworkConfiguration.validate();
 
-    const auto executableNetwork =
-            std::make_shared<ExecutableNetwork>(
-                model, _mvnc, _devicePool, parsedConfigCopy, GetCore());
+    const auto executableNetwork = std::make_shared<ExecutableNetwork>(model, _mvnc, _devicePool, executableNetworkConfiguration, GetCore());
     executableNetwork->SetPointerToPlugin(shared_from_this());
-
     return executableNetwork;
 }
 
@@ -186,7 +201,10 @@ InferenceEngine::Parameter Engine::GetMetric(const std::string& name,
         const auto& supportedMetrics = _metrics->SupportedMetrics();
         IE_SET_METRIC_RETURN(SUPPORTED_METRICS, std::vector<std::string>{supportedMetrics.cbegin(), supportedMetrics.cend()});
     } else if (name == METRIC_KEY(SUPPORTED_CONFIG_KEYS)) {
-        const auto& supportedConfigKeys = _metrics->SupportedConfigKeys();
+        // TODO: remove once all options are migrated
+        auto supportedConfigKeys = _metrics->SupportedConfigKeys();
+        const auto& publicKeys = _parsedConfig.getPublicKeys();
+        supportedConfigKeys.insert(publicKeys.cbegin(), publicKeys.cend());
         IE_SET_METRIC_RETURN(SUPPORTED_CONFIG_KEYS, std::vector<std::string>{supportedConfigKeys.cbegin(), supportedConfigKeys.cend()});
     } else if (name == METRIC_KEY(OPTIMIZATION_CAPABILITIES)) {
         const auto& optimizationCapabilities = _metrics->OptimizationCapabilities();
diff --git a/inference-engine/src/vpu/myriad_plugin/myriad_plugin.h b/inference-engine/src/vpu/myriad_plugin/myriad_plugin.h
index 9fb074b5ac174c..1d7536600c265b 100644
--- a/inference-engine/src/vpu/myriad_plugin/myriad_plugin.h
+++ b/inference-engine/src/vpu/myriad_plugin/myriad_plugin.h
@@ -8,6 +8,7 @@
 #include "myriad_executable_network.h"
 #include "myriad_mvnc_wrapper.h"
 #include "myriad_metrics.h"
+#include "configuration/myriad_configuration.hpp"
 #include <memory>
 #include <string>
 #include <vector>
@@ -50,7 +51,7 @@ class Engine : public ie::IInferencePlugin {
             const std::map<std::string, ie::Parameter>& options) const override;
 
 private:
-    MyriadConfig _parsedConfig;
+    MyriadConfiguration _parsedConfig;
     std::vector<DevicePtr> _devicePool;
     std::shared_ptr<IMvnc> _mvnc;
     std::shared_ptr<MyriadMetrics> _metrics;
diff --git a/inference-engine/tests/functional/inference_engine/parameter_tests.cpp b/inference-engine/tests/functional/inference_engine/parameter_tests.cpp
index 85496e9170715d..49c72cad916c49 100644
--- a/inference-engine/tests/functional/inference_engine/parameter_tests.cpp
+++ b/inference-engine/tests/functional/inference_engine/parameter_tests.cpp
@@ -246,6 +246,18 @@ TEST_F(ParameterTests, ParametersCStringEqual) {
     ASSERT_FALSE(p1 != p2);
 }
 
+TEST_F(ParameterTests, MapOfParametersEqual) {
+    std::map<std::string, Parameter> map0;
+    map0["testParamInt"] = 4;
+    map0["testParamString"] = "test";
+    const auto map1 = map0;
+
+    Parameter p0 = map0;
+    Parameter p1 = map1;
+    ASSERT_TRUE(p0 == p1);
+    ASSERT_FALSE(p0 != p1);
+}
+
 TEST_F(ParameterTests, CompareParametersWithoutEqualOperator) {
     class TestClass {
     public:
@@ -312,4 +324,95 @@ TEST_F(ParameterTests, ParameterRemovedRealObjectPointerWithDuplication) {
     }
     ASSERT_EQ(1, DestructorTest::constructorCount);
     ASSERT_EQ(1, DestructorTest::destructorCount);
-}
\ No newline at end of file
+}
+
+TEST_F(ParameterTests, PrintToEmptyParameterDoesNothing) {
+    Parameter p;
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::string{});
+}
+
+TEST_F(ParameterTests, PrintToIntParameter) {
+    int value = -5;
+    Parameter p = value;
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::to_string(value));
+}
+
+TEST_F(ParameterTests, PrintToUIntParameter) {
+    unsigned int value = 5;
+    Parameter p = value;
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::to_string(value));
+}
+
+TEST_F(ParameterTests, PrintToSize_tParameter) {
+    std::size_t value = 5;
+    Parameter p = value;
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::to_string(value));
+}
+
+TEST_F(ParameterTests, PrintToFloatParameter) {
+    Parameter p = 5.5f;
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::string{"5.5"});
+}
+
+TEST_F(ParameterTests, PrintToStringParameter) {
+    std::string value = "some text";
+    Parameter p = value;
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), value);
+}
+
+TEST_F(ParameterTests, PrintToVectorOfIntsParameterDoesNothing) {
+    Parameter p = std::vector<int>{-5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5};
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::string{});
+}
+
+TEST_F(ParameterTests, PrintToVectorOfUIntsParameterDoesNothing) {
+    Parameter p = std::vector<unsigned int>{0, 1, 2, 3, 4, 5};
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::string{});
+}
+
+TEST_F(ParameterTests, PrintToVectorOfSize_tParameterDoesNothing) {
+    Parameter p = std::vector<std::size_t>{0, 1, 2, 3, 4, 5};
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::string{});
+}
+
+TEST_F(ParameterTests, PrintToVectorOfFloatsParameterDoesNothing) {
+    Parameter p = std::vector<float>{0.0f, 1.1f, 2.2f, 3.3f, 4.4f, 5.5f};
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::string{});
+}
+
+TEST_F(ParameterTests, PrintToVectorOfStringsParameterDoesNothing) {
+    Parameter p = std::vector<std::string>{"zero", "one", "two", "three", "four", "five"};
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::string{});
+}
+
+TEST_F(ParameterTests, PrintToMapOfParametersDoesNothing) {
+    std::map<std::string, Parameter> refMap;
+    refMap["testParamInt"] = 4;
+    refMap["testParamString"] = "test";
+    Parameter p = refMap;
+    std::stringstream stream;
+    ASSERT_NO_THROW(PrintTo(p, &stream));
+    ASSERT_EQ(stream.str(), std::string{});
+}
diff --git a/inference-engine/tests/functional/plugin/myriad/CMakeLists.txt b/inference-engine/tests/functional/plugin/myriad/CMakeLists.txt
index 6a5b0f90e90f05..14f387fa7bae8a 100644
--- a/inference-engine/tests/functional/plugin/myriad/CMakeLists.txt
+++ b/inference-engine/tests/functional/plugin/myriad/CMakeLists.txt
@@ -4,19 +4,32 @@
 
 set(TARGET_NAME myriadFuncTests)
 
+disable_deprecated_warnings()
+
+include(${XLINK_DIR}/XLink.cmake)
+
 addIeTargetTest(
         NAME ${TARGET_NAME}
         ROOT ${CMAKE_CURRENT_SOURCE_DIR}
         INCLUDES
             ${CMAKE_CURRENT_SOURCE_DIR}
             ${IE_MAIN_SOURCE_DIR}/src/vpu/graph_transformer/include
+            ${IE_MAIN_SOURCE_DIR}/tests_deprecated/behavior/vpu/myriad_tests/helpers
+            ${XLINK_INCLUDE}
+            ${XLINK_PLATFORM_INCLUDE}
         DEPENDENCIES
             myriadPlugin
         LINK_LIBRARIES
             vpu_common_lib
             vpu_graph_transformer
             funcSharedTests
+            mvnc
         ADD_CPPLINT
+        DEFINES
+            __PC__
+        OBJECT_FILES
+            ${IE_MAIN_SOURCE_DIR}/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_devices.hpp
+            ${IE_MAIN_SOURCE_DIR}/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_devices.cpp
         LABELS
             VPU
             MYRIAD
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/config.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/config.cpp
index 8198792fd836b3..773419d78f18e9 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/config.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/config.cpp
@@ -5,209 +5,353 @@
 #include "vpu/vpu_plugin_config.hpp"
 #include "vpu/private_plugin_config.hpp"
 #include "behavior/config.hpp"
+#include "myriad_devices.hpp"
 
 IE_SUPPRESS_DEPRECATED_START
 
-using namespace BehaviorTestsDefinitions;
 namespace {
-    const std::vector<InferenceEngine::Precision> netPrecisions = {
-            InferenceEngine::Precision::FP32,
-            InferenceEngine::Precision::FP16
-    };
-
-    const std::vector<std::map<std::string, std::string>> Configs = {
-            {{InferenceEngine::MYRIAD_ENABLE_FORCE_RESET, CONFIG_VALUE(YES)}},
-            {{InferenceEngine::MYRIAD_ENABLE_FORCE_RESET, CONFIG_VALUE(NO)}},
-
-            {{CONFIG_KEY(LOG_LEVEL), CONFIG_VALUE(LOG_NONE)}},
-            {{CONFIG_KEY(LOG_LEVEL), CONFIG_VALUE(LOG_ERROR)}},
-            {{CONFIG_KEY(LOG_LEVEL), CONFIG_VALUE(LOG_WARNING)}},
-            {{CONFIG_KEY(LOG_LEVEL), CONFIG_VALUE(LOG_INFO)}},
-            {{CONFIG_KEY(LOG_LEVEL), CONFIG_VALUE(LOG_DEBUG)}},
-            {{CONFIG_KEY(LOG_LEVEL), CONFIG_VALUE(LOG_TRACE)}},
-
-            {{InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, CONFIG_VALUE(YES)}},
-            {{InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, CONFIG_VALUE(NO)}},
-
-            {{InferenceEngine::MYRIAD_TILING_CMX_LIMIT_KB, "-1"}},
-            {{InferenceEngine::MYRIAD_TILING_CMX_LIMIT_KB, "0"}},
-            {{InferenceEngine::MYRIAD_TILING_CMX_LIMIT_KB, "10"}},
-
-            {{InferenceEngine::MYRIAD_ENABLE_RECEIVING_TENSOR_TIME, CONFIG_VALUE(YES)}},
-            {{InferenceEngine::MYRIAD_ENABLE_RECEIVING_TENSOR_TIME, CONFIG_VALUE(NO)}},
-            {{InferenceEngine::MYRIAD_PROTOCOL, InferenceEngine::MYRIAD_USB}},
-            {{InferenceEngine::MYRIAD_PROTOCOL, InferenceEngine::MYRIAD_PCIE}},
-
-            {{InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "1"}},
-            {{InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "2"}},
-            {{InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "3"}},
-
-            {{InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, CONFIG_VALUE(YES)}},
-            {{InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, CONFIG_VALUE(NO)}},
 
-            // Deprecated
-            {{VPU_MYRIAD_CONFIG_KEY(FORCE_RESET), CONFIG_VALUE(YES)}},
-            {{VPU_MYRIAD_CONFIG_KEY(FORCE_RESET), CONFIG_VALUE(NO)}},
-
-            {{VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), CONFIG_VALUE(YES)}},
-            {{VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), CONFIG_VALUE(NO)}},
-
-            {{VPU_CONFIG_KEY(PRINT_RECEIVE_TENSOR_TIME), CONFIG_VALUE(YES)}},
-            {{VPU_CONFIG_KEY(PRINT_RECEIVE_TENSOR_TIME), CONFIG_VALUE(NO)}},
-
-            {{VPU_MYRIAD_CONFIG_KEY(PROTOCOL), VPU_MYRIAD_CONFIG_VALUE(USB)}},
-            {{VPU_MYRIAD_CONFIG_KEY(PROTOCOL), VPU_MYRIAD_CONFIG_VALUE(PCIE)}},
+using namespace BehaviorTestsDefinitions;
+using namespace InferenceEngine::PluginConfigParams;
 
-            {{VPU_MYRIAD_CONFIG_KEY(PLATFORM), VPU_MYRIAD_CONFIG_VALUE(2450)}},
-            {{VPU_MYRIAD_CONFIG_KEY(PLATFORM), VPU_MYRIAD_CONFIG_VALUE(2480)}}
+const std::vector<InferenceEngine::Precision>& getPrecisions() {
+    static const std::vector<InferenceEngine::Precision> precisions = {
+        InferenceEngine::Precision::FP32,
+        InferenceEngine::Precision::FP16,
     };
-
-    const std::vector<std::map<std::string, std::string>> MultiConfigs = {
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
-             {CONFIG_KEY(LOG_LEVEL), CONFIG_VALUE(LOG_DEBUG)}},
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
-             {InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, CONFIG_VALUE(YES)}},
-
-            // Deprecated
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
-             {VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), CONFIG_VALUE(YES)}}
+    return precisions;
+}
+
+std::vector<std::map<std::string, std::string>> getCorrectConfigs() {
+    std::vector<std::map<std::string, std::string>> correctConfigs = {
+        {{KEY_LOG_LEVEL, LOG_NONE}},
+        {{KEY_LOG_LEVEL, LOG_ERROR}},
+        {{KEY_LOG_LEVEL, LOG_WARNING}},
+        {{KEY_LOG_LEVEL, LOG_INFO}},
+        {{KEY_LOG_LEVEL, LOG_DEBUG}},
+        {{KEY_LOG_LEVEL, LOG_TRACE}},
+
+        {{InferenceEngine::MYRIAD_ENABLE_FORCE_RESET, CONFIG_VALUE(YES)}},
+        {{InferenceEngine::MYRIAD_ENABLE_FORCE_RESET, CONFIG_VALUE(NO)}},
+
+        {{InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, CONFIG_VALUE(YES)}},
+        {{InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, CONFIG_VALUE(NO)}},
+
+        {{InferenceEngine::MYRIAD_TILING_CMX_LIMIT_KB, "-1"}},
+        {{InferenceEngine::MYRIAD_TILING_CMX_LIMIT_KB, "0"}},
+        {{InferenceEngine::MYRIAD_TILING_CMX_LIMIT_KB, "10"}},
+
+        {{InferenceEngine::MYRIAD_ENABLE_RECEIVING_TENSOR_TIME, CONFIG_VALUE(YES)}},
+        {{InferenceEngine::MYRIAD_ENABLE_RECEIVING_TENSOR_TIME, CONFIG_VALUE(NO)}},
+
+        {{InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "1"}},
+        {{InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "2"}},
+        {{InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "3"}},
+
+        {{InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, CONFIG_VALUE(YES)}},
+        {{InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, CONFIG_VALUE(NO)}},
+
+        // Deprecated
+        {{VPU_CONFIG_KEY(LOG_LEVEL), LOG_NONE}},
+        {{VPU_CONFIG_KEY(LOG_LEVEL), LOG_ERROR}},
+        {{VPU_CONFIG_KEY(LOG_LEVEL), LOG_WARNING}},
+        {{VPU_CONFIG_KEY(LOG_LEVEL), LOG_INFO}},
+        {{VPU_CONFIG_KEY(LOG_LEVEL), LOG_DEBUG}},
+        {{VPU_CONFIG_KEY(LOG_LEVEL), LOG_TRACE}},
+
+        {{VPU_MYRIAD_CONFIG_KEY(FORCE_RESET), CONFIG_VALUE(YES)}},
+        {{VPU_MYRIAD_CONFIG_KEY(FORCE_RESET), CONFIG_VALUE(NO)}},
+
+        {{VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), CONFIG_VALUE(YES)}},
+        {{VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), CONFIG_VALUE(NO)}},
+
+        {{VPU_CONFIG_KEY(PRINT_RECEIVE_TENSOR_TIME), CONFIG_VALUE(YES)}},
+        {{VPU_CONFIG_KEY(PRINT_RECEIVE_TENSOR_TIME), CONFIG_VALUE(NO)}},
+
+        {{VPU_MYRIAD_CONFIG_KEY(PLATFORM), VPU_MYRIAD_CONFIG_VALUE(2480)}},
+
+        {
+            {KEY_LOG_LEVEL, LOG_INFO},
+            {InferenceEngine::MYRIAD_COPY_OPTIMIZATION, InferenceEngine::PluginConfigParams::NO},
+            {InferenceEngine::MYRIAD_ENABLE_FORCE_RESET, CONFIG_VALUE(YES)},
+            {InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, CONFIG_VALUE(YES)},
+            {InferenceEngine::MYRIAD_TILING_CMX_LIMIT_KB, "10"},
+            {InferenceEngine::MYRIAD_ENABLE_RECEIVING_TENSOR_TIME, CONFIG_VALUE(YES)},
+            {InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "1"},
+            {InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, CONFIG_VALUE(YES)},
+        },
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigTests,
-                            ::testing::Combine(
-                                    ::testing::ValuesIn(netPrecisions),
-                                    ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
-                                    ::testing::ValuesIn(Configs)),
-                            CorrectConfigTests::getTestCaseName);
-
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigTests,
-                            ::testing::Combine(
-                                    ::testing::ValuesIn(netPrecisions),
-                                    ::testing::Values(CommonTestUtils::DEVICE_MULTI),
-                                    ::testing::ValuesIn(MultiConfigs)),
-                            CorrectConfigTests::getTestCaseName);
-
-    const std::vector<std::map<std::string, std::string>> inconfigs = {
-            {{InferenceEngine::MYRIAD_PROTOCOL, "BLUETOOTH"}},
-            {{InferenceEngine::MYRIAD_PROTOCOL, "LAN"}},
-
-            {{InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, "ON"}},
-            {{InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, "OFF"}},
-
-            {{InferenceEngine::MYRIAD_ENABLE_FORCE_RESET, "ON"}},
-            {{InferenceEngine::MYRIAD_ENABLE_FORCE_RESET, "OFF"}},
-
-            {{CONFIG_KEY(LOG_LEVEL), "VERBOSE"}},
-
-            {{InferenceEngine::MYRIAD_TILING_CMX_LIMIT_KB, "-10"}},
-
-            {{InferenceEngine::MYRIAD_ENABLE_RECEIVING_TENSOR_TIME, "ON"}},
-            {{InferenceEngine::MYRIAD_ENABLE_RECEIVING_TENSOR_TIME, "OFF"}},
-
-            {{InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "Two"}},
-            {{InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "SINGLE"}},
-
-            {{InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, "ON"}},
-            {{InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, "OFF"}},
-
-            // Deprecated
-            {{VPU_MYRIAD_CONFIG_KEY(PROTOCOL), "BLUETOOTH"}},
-            {{VPU_MYRIAD_CONFIG_KEY(PROTOCOL), "LAN"}},
-
-            {{VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), "ON"}},
-            {{VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), "OFF"}},
+    MyriadDevicesInfo info;
+    if (info.getAmountOfDevices(ncDeviceProtocol_t::NC_PCIE) > 0) {
+        correctConfigs.emplace_back(std::map<std::string, std::string>{{VPU_MYRIAD_CONFIG_KEY(PROTOCOL), VPU_MYRIAD_CONFIG_VALUE(PCIE)}});
+        correctConfigs.emplace_back(std::map<std::string, std::string>{{InferenceEngine::MYRIAD_PROTOCOL, InferenceEngine::MYRIAD_PCIE}});
+    }
+
+    if (info.getAmountOfDevices(ncDeviceProtocol_t::NC_USB) > 0) {
+        correctConfigs.emplace_back(std::map<std::string, std::string>{{VPU_MYRIAD_CONFIG_KEY(PROTOCOL), VPU_MYRIAD_CONFIG_VALUE(USB)}});
+        correctConfigs.emplace_back(std::map<std::string, std::string>{{InferenceEngine::MYRIAD_PROTOCOL, InferenceEngine::MYRIAD_USB}});
+    }
+
+    return correctConfigs;
+}
+
+INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigTests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
+        ::testing::ValuesIn(getCorrectConfigs())),
+    CorrectConfigTests::getTestCaseName);
+
+const std::vector<std::map<std::string, std::string>>& getCorrectMultiConfigs() {
+    static const std::vector<std::map<std::string, std::string>> correctMultiConfigs = {
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+        },
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {KEY_LOG_LEVEL, LOG_DEBUG},
+        },
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {InferenceEngine::MYRIAD_COPY_OPTIMIZATION, InferenceEngine::PluginConfigParams::NO},
+        },
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, YES},
+        },
+
+        // Deprecated
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {VPU_CONFIG_KEY(LOG_LEVEL), LOG_DEBUG},
+        },
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), CONFIG_VALUE(YES)},
+        },
+    };
+    return correctMultiConfigs;
+}
+
+INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigTests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MULTI),
+        ::testing::ValuesIn(getCorrectMultiConfigs())),
+    CorrectConfigTests::getTestCaseName);
+
+const std::vector<std::pair<std::string, InferenceEngine::Parameter>>& getDefaultEntries() {
+    static const std::vector<std::pair<std::string, InferenceEngine::Parameter>> defaultEntries = {
+        {KEY_LOG_LEVEL, {LOG_NONE}},
+    };
+    return defaultEntries;
+}
+
+INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectSingleOptionDefaultValueConfigTests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
+        ::testing::ValuesIn(getDefaultEntries())));
+
+const std::vector<std::tuple<std::string, std::string, InferenceEngine::Parameter>>& getCustomEntries() {
+    static const std::vector<std::tuple<std::string, std::string, InferenceEngine::Parameter>> customEntries = {
+        std::make_tuple(KEY_LOG_LEVEL, LOG_NONE,    InferenceEngine::Parameter{LOG_NONE}),
+        std::make_tuple(KEY_LOG_LEVEL, LOG_ERROR,   InferenceEngine::Parameter{LOG_ERROR}),
+        std::make_tuple(KEY_LOG_LEVEL, LOG_WARNING, InferenceEngine::Parameter{LOG_WARNING}),
+        std::make_tuple(KEY_LOG_LEVEL, LOG_INFO,    InferenceEngine::Parameter{LOG_INFO}),
+        std::make_tuple(KEY_LOG_LEVEL, LOG_DEBUG,   InferenceEngine::Parameter{LOG_DEBUG}),
+        std::make_tuple(KEY_LOG_LEVEL, LOG_TRACE,   InferenceEngine::Parameter{LOG_TRACE}),
+
+        std::make_tuple(VPU_CONFIG_KEY(LOG_LEVEL), LOG_NONE,    InferenceEngine::Parameter{LOG_NONE}),
+        std::make_tuple(VPU_CONFIG_KEY(LOG_LEVEL), LOG_ERROR,   InferenceEngine::Parameter{LOG_ERROR}),
+        std::make_tuple(VPU_CONFIG_KEY(LOG_LEVEL), LOG_WARNING, InferenceEngine::Parameter{LOG_WARNING}),
+        std::make_tuple(VPU_CONFIG_KEY(LOG_LEVEL), LOG_INFO,    InferenceEngine::Parameter{LOG_INFO}),
+        std::make_tuple(VPU_CONFIG_KEY(LOG_LEVEL), LOG_DEBUG,   InferenceEngine::Parameter{LOG_DEBUG}),
+        std::make_tuple(VPU_CONFIG_KEY(LOG_LEVEL), LOG_TRACE,   InferenceEngine::Parameter{LOG_TRACE}),
+
+        std::make_tuple(InferenceEngine::MYRIAD_COPY_OPTIMIZATION, InferenceEngine::PluginConfigParams::YES, InferenceEngine::Parameter{true}),
+        std::make_tuple(InferenceEngine::MYRIAD_COPY_OPTIMIZATION, InferenceEngine::PluginConfigParams::NO,  InferenceEngine::Parameter{false}),
+    };
+    return customEntries;
+}
+
+INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectSingleOptionCustomValueConfigTests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
+        ::testing::ValuesIn(getCustomEntries())));
+
+const std::vector<std::string>& getPublicOptions() {
+    static const std::vector<std::string> publicOptions = {
+        KEY_LOG_LEVEL,
+        VPU_CONFIG_KEY(LOG_LEVEL),
+    };
+    return publicOptions;
+}
+
+INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigPublicOptionsTests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
+        ::testing::ValuesIn(getPublicOptions())));
+
+const std::vector<std::string>& getPrivateOptions() {
+    static const std::vector<std::string> privateOptions = {
+        InferenceEngine::MYRIAD_COPY_OPTIMIZATION,
+    };
+    return privateOptions;
+}
 
-            {{VPU_MYRIAD_CONFIG_KEY(FORCE_RESET), "ON"}},
-            {{VPU_MYRIAD_CONFIG_KEY(FORCE_RESET), "OFF"}},
+INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigPrivateOptionsTests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
+        ::testing::ValuesIn(getPrivateOptions())));
 
-            {{VPU_CONFIG_KEY(PRINT_RECEIVE_TENSOR_TIME), "ON"}},
-            {{VPU_CONFIG_KEY(PRINT_RECEIVE_TENSOR_TIME), "OFF"}},
+const std::vector<std::map<std::string, std::string>>& getIncorrectConfigs() {
+    static const std::vector<std::map<std::string, std::string>> incorrectConfigs = {
+        {{KEY_LOG_LEVEL, "INCORRECT_LOG_LEVEL"}},
 
-            {{VPU_MYRIAD_CONFIG_KEY(PLATFORM), "-1"}},
-            {{VPU_MYRIAD_CONFIG_KEY(PLATFORM), "0"}},
-            {{VPU_MYRIAD_CONFIG_KEY(PLATFORM), "1"}},
-    };
+        {{InferenceEngine::MYRIAD_COPY_OPTIMIZATION, "ON"}},
+        {{InferenceEngine::MYRIAD_COPY_OPTIMIZATION, "OFF"}},
 
-    const std::vector<std::map<std::string, std::string>> multiinconfigs = {
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
-             {InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, "ON"}},
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
-             {CONFIG_KEY(LOG_LEVEL), "VERBOSE"}},
-
-            // Deprecated
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
-             {VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), "ON"}},
-
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
-             {VPU_MYRIAD_CONFIG_KEY(PLATFORM), "-1"}},
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
-             {VPU_MYRIAD_CONFIG_KEY(PLATFORM), "0"}},
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
-             {VPU_MYRIAD_CONFIG_KEY(PLATFORM), "1"}},
-    };
+        {{InferenceEngine::MYRIAD_PROTOCOL, "BLUETOOTH"}},
+        {{InferenceEngine::MYRIAD_PROTOCOL, "LAN"}},
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigTests,
-                            ::testing::Combine(
-                                    ::testing::ValuesIn(netPrecisions),
-                                    ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
-                                    ::testing::ValuesIn(inconfigs)),
-                            IncorrectConfigTests::getTestCaseName);
+        {{InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, "ON"}},
+        {{InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, "OFF"}},
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, IncorrectConfigTests,
-                            ::testing::Combine(
-                                    ::testing::ValuesIn(netPrecisions),
-                                    ::testing::Values(CommonTestUtils::DEVICE_MULTI),
-                                    ::testing::ValuesIn(multiinconfigs)),
-                            IncorrectConfigTests::getTestCaseName);
+        {{InferenceEngine::MYRIAD_ENABLE_FORCE_RESET, "ON"}},
+        {{InferenceEngine::MYRIAD_ENABLE_FORCE_RESET, "OFF"}},
 
+        {{InferenceEngine::MYRIAD_TILING_CMX_LIMIT_KB, "-10"}},
 
+        {{InferenceEngine::MYRIAD_ENABLE_RECEIVING_TENSOR_TIME, "ON"}},
+        {{InferenceEngine::MYRIAD_ENABLE_RECEIVING_TENSOR_TIME, "OFF"}},
 
-    const std::vector<std::map<std::string, std::string>> Inconf = {
-            {{"some_nonexistent_key", "some_unknown_value"}}
-    };
+        {{InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "Two"}},
+        {{InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "SINGLE"}},
 
-    const std::vector<std::map<std::string, std::string>> multiInconf = {
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES , CommonTestUtils::DEVICE_MYRIAD},
-             {"some_nonexistent_key", "some_unknown_value"}}
-    };
+        {{InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, "ON"}},
+        {{InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, "OFF"}},
 
+        // Deprecated
+        {{VPU_CONFIG_KEY(LOG_LEVEL), "INCORRECT_LOG_LEVEL"}},
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
-                            ::testing::Combine(
-                                    ::testing::ValuesIn(netPrecisions),
-                                    ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
-                                    ::testing::ValuesIn(Inconf)),
-                            IncorrectConfigAPITests::getTestCaseName);
+        {{VPU_MYRIAD_CONFIG_KEY(PROTOCOL), "BLUETOOTH"}},
+        {{VPU_MYRIAD_CONFIG_KEY(PROTOCOL), "LAN"}},
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, IncorrectConfigAPITests,
-                            ::testing::Combine(
-                                    ::testing::ValuesIn(netPrecisions),
-                                    ::testing::Values(CommonTestUtils::DEVICE_MULTI),
-                                    ::testing::ValuesIn(multiInconf)),
-                            IncorrectConfigAPITests::getTestCaseName);
+        {{VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), "ON"}},
+        {{VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), "OFF"}},
 
+        {{VPU_MYRIAD_CONFIG_KEY(FORCE_RESET), "ON"}},
+        {{VPU_MYRIAD_CONFIG_KEY(FORCE_RESET), "OFF"}},
 
+        {{VPU_CONFIG_KEY(PRINT_RECEIVE_TENSOR_TIME), "ON"}},
+        {{VPU_CONFIG_KEY(PRINT_RECEIVE_TENSOR_TIME), "OFF"}},
 
+        {{VPU_MYRIAD_CONFIG_KEY(PLATFORM), "-1"}},
+        {{VPU_MYRIAD_CONFIG_KEY(PLATFORM), "0"}},
+        {{VPU_MYRIAD_CONFIG_KEY(PLATFORM), "1"}},
 
-    const std::vector<std::map<std::string, std::string>> conf = {
-            {}
+        {
+            {KEY_LOG_LEVEL, LOG_INFO},
+            {InferenceEngine::MYRIAD_COPY_OPTIMIZATION, "ON"},
+            {InferenceEngine::MYRIAD_PROTOCOL, "BLUETOOTH"},
+            {InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, CONFIG_VALUE(YES)},
+            {InferenceEngine::MYRIAD_ENABLE_FORCE_RESET, "ON"},
+            {InferenceEngine::MYRIAD_TILING_CMX_LIMIT_KB, "10"},
+            {InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, "OFF"},
+            {InferenceEngine::MYRIAD_THROUGHPUT_STREAMS, "1"},
+            {InferenceEngine::MYRIAD_ENABLE_WEIGHTS_ANALYSIS, "ON"},
+        },
     };
-
-    const std::vector<std::map<std::string, std::string>> multiconf = {
-            {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES , CommonTestUtils::DEVICE_MYRIAD}}
+    return incorrectConfigs;
+}
+
+INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigTests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
+        ::testing::ValuesIn(getIncorrectConfigs())),
+    IncorrectConfigTests::getTestCaseName);
+
+const std::vector<std::map<std::string, std::string>>& getIncorrectMultiConfigs() {
+    static const std::vector<std::map<std::string, std::string>> incorrectMultiConfigs = {
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {KEY_LOG_LEVEL, "INCORRECT_LOG_LEVEL"},
+        },
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {InferenceEngine::MYRIAD_ENABLE_HW_ACCELERATION, "ON"},
+        },
+
+        // Deprecated
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {VPU_CONFIG_KEY(LOG_LEVEL), "INCORRECT_LOG_LEVEL"},
+        },
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), "ON"},
+        },
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {VPU_MYRIAD_CONFIG_KEY(PLATFORM), "-1"},
+        },
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {VPU_MYRIAD_CONFIG_KEY(PLATFORM), "0"},
+        },
+        {
+            {InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD},
+            {VPU_MYRIAD_CONFIG_KEY(PLATFORM), "1"},
+        },
     };
+    return incorrectMultiConfigs;
+}
+
+INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, IncorrectConfigTests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MULTI),
+        ::testing::ValuesIn(getIncorrectMultiConfigs())),
+    IncorrectConfigTests::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigSingleOptionTests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
+        ::testing::Values("INCORRECT_KEY")));
+
+INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigAPITests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
+        ::testing::Values(std::map<std::string, std::string>{})),
+    CorrectConfigAPITests::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigAPITests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MULTI),
+        ::testing::ValuesIn(getCorrectMultiConfigs())),
+    CorrectConfigAPITests::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
+        ::testing::Values(std::map<std::string, std::string>{{"INCORRECT_KEY", "INCORRECT_VALUE"}})),
+    IncorrectConfigAPITests::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, IncorrectConfigAPITests,
+    ::testing::Combine(
+        ::testing::ValuesIn(getPrecisions()),
+        ::testing::Values(CommonTestUtils::DEVICE_MULTI),
+        ::testing::ValuesIn(getIncorrectMultiConfigs())),
+    IncorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigAPITests,
-                            ::testing::Combine(
-                                    ::testing::ValuesIn(netPrecisions),
-                                    ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
-                                    ::testing::ValuesIn(conf)),
-                            CorrectConfigAPITests::getTestCaseName);
-
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigAPITests,
-                            ::testing::Combine(
-                                    ::testing::ValuesIn(netPrecisions),
-                                    ::testing::Values(CommonTestUtils::DEVICE_MULTI),
-                                    ::testing::ValuesIn(multiconf)),
-                            CorrectConfigAPITests::getTestCaseName);
 } // namespace
diff --git a/inference-engine/tests/functional/plugin/shared/include/base/behavior_test_utils.hpp b/inference-engine/tests/functional/plugin/shared/include/base/behavior_test_utils.hpp
index e21eb11624b20d..5705bdc8ae37c6 100644
--- a/inference-engine/tests/functional/plugin/shared/include/base/behavior_test_utils.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/base/behavior_test_utils.hpp
@@ -27,11 +27,45 @@
 #include "ngraph_functions/pass/convert_prc.hpp"
 
 namespace BehaviorTestsUtils {
-    typedef std::tuple<
-            InferenceEngine::Precision,         // Network precision
-            std::string,                        // Device name
-            std::map<std::string, std::string>  // Config
-    > BehaviorParams;
+
+using BehaviorParamsEmptyConfig = std::tuple<
+    InferenceEngine::Precision,         // Network precision
+    std::string                         // Device name
+>;
+
+class BehaviorTestsEmptyConfig : public testing::WithParamInterface<BehaviorParamsEmptyConfig>,
+                                 public CommonTestUtils::TestsCommon {
+public:
+    static std::string getTestCaseName(testing::TestParamInfo<BehaviorParamsEmptyConfig> obj) {
+        InferenceEngine::Precision  netPrecision;
+        std::string targetDevice;
+        std::tie(netPrecision, targetDevice) = obj.param;
+        std::ostringstream result;
+        result << "netPRC=" << netPrecision.name() << "_";
+        result << "targetDevice=" << targetDevice;
+        return result.str();
+    }
+
+    void SetUp()  override {
+        std::tie(netPrecision, targetDevice) = this->GetParam();
+        function = ngraph::builder::subgraph::makeConvPoolRelu();
+    }
+
+    void TearDown() override {
+        function.reset();
+    }
+
+    std::shared_ptr<InferenceEngine::Core> ie = PluginCache::get().ie();
+    std::shared_ptr<ngraph::Function> function;
+    InferenceEngine::Precision netPrecision;
+    std::string targetDevice;
+};
+
+typedef std::tuple<
+        InferenceEngine::Precision,         // Network precision
+        std::string,                        // Device name
+        std::map<std::string, std::string>  // Config
+> BehaviorParams;
 
 class BehaviorTestsBasic : public testing::WithParamInterface<BehaviorParams>,
                            public CommonTestUtils::TestsCommon {
@@ -71,4 +105,86 @@ class BehaviorTestsBasic : public testing::WithParamInterface<BehaviorParams>,
     std::map<std::string, std::string> configuration;
 };
 
+using BehaviorParamsSingleOption = std::tuple<
+    InferenceEngine::Precision,         // Network precision
+    std::string,                        // Device name
+    std::string                         // Key
+>;
+
+class BehaviorTestsSingleOption : public testing::WithParamInterface<BehaviorParamsSingleOption>,
+                                  public CommonTestUtils::TestsCommon {
+public:
+    void SetUp()  override {
+        std::tie(netPrecision, targetDevice, key) = this->GetParam();
+        function = ngraph::builder::subgraph::makeConvPoolRelu();
+    }
+
+    void TearDown() override {
+        function.reset();
+    }
+
+    std::shared_ptr<InferenceEngine::Core> ie = PluginCache::get().ie();
+    std::shared_ptr<ngraph::Function> function;
+    InferenceEngine::Precision netPrecision;
+    std::string targetDevice;
+    std::string key;
+};
+
+using BehaviorParamsSingleOptionDefault = std::tuple<
+    InferenceEngine::Precision,                        // Network precision
+    std::string,                                       // Device name
+    std::pair<std::string, InferenceEngine::Parameter> // Configuration key and its default value
+>;
+
+class BehaviorTestsSingleOptionDefault : public testing::WithParamInterface<BehaviorParamsSingleOptionDefault>,
+                                         public CommonTestUtils::TestsCommon {
+public:
+    void SetUp()  override {
+        std::pair<std::string, InferenceEngine::Parameter> entry;
+        std::tie(netPrecision, targetDevice, entry) = this->GetParam();
+        std::tie(key, value) = entry;
+        function = ngraph::builder::subgraph::makeConvPoolRelu();
+    }
+
+    void TearDown() override {
+        function.reset();
+    }
+
+    std::shared_ptr<InferenceEngine::Core> ie = PluginCache::get().ie();
+    std::shared_ptr<ngraph::Function> function;
+    InferenceEngine::Precision netPrecision;
+    std::string targetDevice;
+    std::string key;
+    InferenceEngine::Parameter value;
+};
+
+using BehaviorParamsSingleOptionCustom = std::tuple<
+    InferenceEngine::Precision,                                      // Network precision
+    std::string,                                                     // Device name
+    std::tuple<std::string, std::string, InferenceEngine::Parameter> // Configuration key, value and reference
+>;
+
+class BehaviorTestsSingleOptionCustom : public testing::WithParamInterface<BehaviorParamsSingleOptionCustom>,
+                                        public CommonTestUtils::TestsCommon {
+public:
+    void SetUp()  override {
+        std::tuple<std::string, std::string, InferenceEngine::Parameter> entry;
+        std::tie(netPrecision, targetDevice, entry) = this->GetParam();
+        std::tie(key, value, reference) = entry;
+        function = ngraph::builder::subgraph::makeConvPoolRelu();
+    }
+
+    void TearDown() override {
+        function.reset();
+    }
+
+    std::shared_ptr<InferenceEngine::Core> ie = PluginCache::get().ie();
+    std::shared_ptr<ngraph::Function> function;
+    InferenceEngine::Precision netPrecision;
+    std::string targetDevice;
+    std::string key;
+    std::string value;
+    InferenceEngine::Parameter reference;
+};
+
 }  // namespace BehaviorTestsUtils
diff --git a/inference-engine/tests/functional/plugin/shared/include/behavior/config.hpp b/inference-engine/tests/functional/plugin/shared/include/behavior/config.hpp
index e13fe679b2a82c..efdd24187ae31b 100644
--- a/inference-engine/tests/functional/plugin/shared/include/behavior/config.hpp
+++ b/inference-engine/tests/functional/plugin/shared/include/behavior/config.hpp
@@ -27,9 +27,11 @@
 #include "ngraph_functions/subgraph_builders.hpp"
 
 namespace BehaviorTestsDefinitions {
-    using CorrectConfigTests = BehaviorTestsUtils::BehaviorTestsBasic;
+
+    using EmptyConfigTests = BehaviorTestsUtils::BehaviorTestsEmptyConfig;
+
     // Setting empty config doesn't throw
-    TEST_P(CorrectConfigTests, SetEmptyConfig) {
+    TEST_P(EmptyConfigTests, SetEmptyConfig) {
         // Skip test according to plugin specific disabledTestPatterns() (if any)
         SKIP_IF_CURRENT_TEST_IS_DISABLED()
         // Create CNNNetwork from ngrpah::Function
@@ -39,6 +41,29 @@ namespace BehaviorTestsDefinitions {
         ASSERT_NO_THROW(ie->SetConfig(config, targetDevice));
     }
 
+    TEST_P(EmptyConfigTests, CanLoadNetworkWithEmptyConfig) {
+        // Skip test according to plugin specific disabledTestPatterns() (if any)
+        SKIP_IF_CURRENT_TEST_IS_DISABLED()
+        // Create CNNNetwork from ngrpah::Function
+        InferenceEngine::CNNNetwork cnnNet(function);
+        std::map<std::string, std::string> config;
+        ASSERT_NO_THROW(ie->GetMetric(targetDevice, METRIC_KEY(SUPPORTED_CONFIG_KEYS)));
+        ASSERT_NO_THROW(ie->LoadNetwork(cnnNet, targetDevice, config));
+    }
+
+    using CorrectSingleOptionDefaultValueConfigTests = BehaviorTestsUtils::BehaviorTestsSingleOptionDefault;
+
+    TEST_P(CorrectSingleOptionDefaultValueConfigTests, CheckDefaultValueOfConfig) {
+        // Skip test according to plugin specific disabledTestPatterns() (if any)
+        SKIP_IF_CURRENT_TEST_IS_DISABLED()
+        // Create CNNNetwork from ngrpah::Function
+        InferenceEngine::CNNNetwork cnnNet(function);
+        ASSERT_NO_THROW(ie->GetMetric(targetDevice, METRIC_KEY(SUPPORTED_CONFIG_KEYS)));
+        ASSERT_EQ(ie->GetConfig(targetDevice, key), value);
+    }
+
+    using CorrectConfigTests = BehaviorTestsUtils::BehaviorTestsBasic;
+
     // Setting correct config doesn't throw
     TEST_P(CorrectConfigTests, SetCorrectConfig) {
         // Skip test according to plugin specific disabledTestPatterns() (if any)
@@ -49,6 +74,53 @@ namespace BehaviorTestsDefinitions {
         ASSERT_NO_THROW(ie->SetConfig(configuration, targetDevice));
     }
 
+    TEST_P(CorrectConfigTests, CanLoadNetworkWithCorrectConfig) {
+        // Skip test according to plugin specific disabledTestPatterns() (if any)
+        SKIP_IF_CURRENT_TEST_IS_DISABLED()
+        // Create CNNNetwork from ngrpah::Function
+        InferenceEngine::CNNNetwork cnnNet(function);
+        ASSERT_NO_THROW(ie->LoadNetwork(cnnNet, targetDevice, configuration));
+    }
+
+    using CorrectSingleOptionCustomValueConfigTests = BehaviorTestsUtils::BehaviorTestsSingleOptionCustom;
+
+    TEST_P(CorrectSingleOptionCustomValueConfigTests, CheckCustomValueOfConfig) {
+        // Skip test according to plugin specific disabledTestPatterns() (if any)
+        SKIP_IF_CURRENT_TEST_IS_DISABLED()
+        // Create CNNNetwork from ngrpah::Function
+        InferenceEngine::CNNNetwork cnnNet(function);
+        ASSERT_NO_THROW(ie->GetMetric(targetDevice, METRIC_KEY(SUPPORTED_CONFIG_KEYS)));
+        std::map<std::string, std::string> configuration = {{key, value}};
+        ASSERT_NO_THROW(ie->SetConfig(configuration, targetDevice));
+        ASSERT_EQ(ie->GetConfig(targetDevice, key), reference);
+    }
+
+    using CorrectConfigPublicOptionsTests = BehaviorTestsUtils::BehaviorTestsSingleOption;
+
+    TEST_P(CorrectConfigPublicOptionsTests, CanSeePublicOption) {
+        // Skip test according to plugin specific disabledTestPatterns() (if any)
+        SKIP_IF_CURRENT_TEST_IS_DISABLED()
+        // Create CNNNetwork from ngrpah::Function
+        InferenceEngine::CNNNetwork cnnNet(function);
+        InferenceEngine::Parameter metric;
+        ASSERT_NO_THROW(metric = ie->GetMetric(targetDevice, METRIC_KEY(SUPPORTED_CONFIG_KEYS)));
+        const auto& supportedOptions = metric.as<std::vector<std::string>>();
+        ASSERT_NE(std::find(supportedOptions.cbegin(), supportedOptions.cend(), key), supportedOptions.cend());
+    }
+
+    using CorrectConfigPrivateOptionsTests = BehaviorTestsUtils::BehaviorTestsSingleOption;
+
+    TEST_P(CorrectConfigPrivateOptionsTests, CanNotSeePrivateOption) {
+        // Skip test according to plugin specific disabledTestPatterns() (if any)
+        SKIP_IF_CURRENT_TEST_IS_DISABLED()
+        // Create CNNNetwork from ngrpah::Function
+        InferenceEngine::CNNNetwork cnnNet(function);
+        InferenceEngine::Parameter metric;
+        ASSERT_NO_THROW(metric = ie->GetMetric(targetDevice, METRIC_KEY(SUPPORTED_CONFIG_KEYS)));
+        const auto& supportedOptions = metric.as<std::vector<std::string>>();
+        ASSERT_EQ(std::find(supportedOptions.cbegin(), supportedOptions.cend(), key), supportedOptions.cend());
+    }
+
     using IncorrectConfigTests = BehaviorTestsUtils::BehaviorTestsBasic;
 
     TEST_P(IncorrectConfigTests, SetConfigWithIncorrectKey) {
@@ -67,7 +139,7 @@ namespace BehaviorTestsDefinitions {
         }
     }
 
-    TEST_P(IncorrectConfigTests, canNotLoadNetworkWithIncorrectConfig) {
+    TEST_P(IncorrectConfigTests, CanNotLoadNetworkWithIncorrectConfig) {
         // Skip test according to plugin specific disabledTestPatterns() (if any)
         SKIP_IF_CURRENT_TEST_IS_DISABLED()
         // Create CNNNetwork from ngrpah::Function
@@ -80,6 +152,17 @@ namespace BehaviorTestsDefinitions {
         }
     }
 
+    using IncorrectConfigSingleOptionTests = BehaviorTestsUtils::BehaviorTestsSingleOption;
+
+    TEST_P(IncorrectConfigSingleOptionTests, CanNotGetConfigWithIncorrectConfig) {
+        // Skip test according to plugin specific disabledTestPatterns() (if any)
+        SKIP_IF_CURRENT_TEST_IS_DISABLED()
+        // Create CNNNetwork from ngrpah::Function
+        InferenceEngine::CNNNetwork cnnNet(function);
+        ASSERT_NO_THROW(ie->GetMetric(targetDevice, METRIC_KEY(SUPPORTED_CONFIG_KEYS)));
+        ASSERT_THROW(ie->GetConfig(targetDevice, key), InferenceEngine::Exception);
+    }
+
     using IncorrectConfigAPITests = BehaviorTestsUtils::BehaviorTestsBasic;
 
     TEST_P(IncorrectConfigAPITests, SetConfigWithNoExistingKey) {
@@ -99,7 +182,7 @@ namespace BehaviorTestsDefinitions {
 
     using CorrectConfigAPITests = BehaviorTestsUtils::BehaviorTestsBasic;
 
-    TEST_P(CorrectConfigAPITests, canSetExclusiveAsyncRequests) {
+    TEST_P(CorrectConfigAPITests, CanSetExclusiveAsyncRequests) {
         // Skip test according to plugin specific disabledTestPatterns() (if any)
         SKIP_IF_CURRENT_TEST_IS_DISABLED()
         // Create CNNNetwork from ngrpah::Function
@@ -130,7 +213,7 @@ namespace BehaviorTestsDefinitions {
         }
     }
 
-    TEST_P(CorrectConfigAPITests, withoutExclusiveAsyncRequests) {
+    TEST_P(CorrectConfigAPITests, WithoutExclusiveAsyncRequests) {
         // Skip test according to plugin specific disabledTestPatterns() (if any)
         SKIP_IF_CURRENT_TEST_IS_DISABLED()
         // Create CNNNetwork from ngrpah::Function
@@ -159,7 +242,7 @@ namespace BehaviorTestsDefinitions {
         }
     }
 
-    TEST_P(CorrectConfigAPITests, reusableCPUStreamsExecutor) {
+    TEST_P(CorrectConfigAPITests, ReusableCPUStreamsExecutor) {
         // Skip test according to plugin specific disabledTestPatterns() (if any)
         SKIP_IF_CURRENT_TEST_IS_DISABLED()
         ASSERT_EQ(0u, InferenceEngine::ExecutorManager::getInstance()->getExecutorsNumber());
@@ -200,4 +283,4 @@ namespace BehaviorTestsDefinitions {
             ASSERT_EQ(0u, InferenceEngine::ExecutorManager::getInstance()->getIdleCPUStreamsExecutorsNumber());
         }
     }
-}  // namespace BehaviorTestsDefinitions
\ No newline at end of file
+}  // namespace BehaviorTestsDefinitions
diff --git a/inference-engine/tests/ie_test_utils/common_test_utils/data_utils.cpp b/inference-engine/tests/ie_test_utils/common_test_utils/data_utils.cpp
index 68d3af73a39d04..dbd552d6da8bdb 100644
--- a/inference-engine/tests/ie_test_utils/common_test_utils/data_utils.cpp
+++ b/inference-engine/tests/ie_test_utils/common_test_utils/data_utils.cpp
@@ -9,6 +9,8 @@
 #include <ie_blob.h>
 #include <blob_factory.hpp>
 
+using namespace InferenceEngine::details;
+
 namespace CommonTestUtils {
 
 bool isDenseBlob(const InferenceEngine::Blob::Ptr& blob) {
@@ -69,8 +71,8 @@ void fill_data_with_broadcast(InferenceEngine::Blob::Ptr& blob, InferenceEngine:
         if (src_dims[i] != dst_dims[i] && src_dims[i] != 1)
             compatible = false;
     }
-    IE_ASSERT(compatible) << "fill_data_with_broadcast error: Tensor shape " << values_dims
-                          << " can not be broadcasted to shape " << blob_dims;
+
+    IE_ASSERT(compatible);
 
     auto fill_strides_like_plain = [] (SizeVector dims) {
         SizeVector str(dims.size());
diff --git a/inference-engine/tests/unit/vpu/base/graph_transformer_tests.cpp b/inference-engine/tests/unit/vpu/base/graph_transformer_tests.cpp
index 5f018914a5ee8e..52ff0425928522 100644
--- a/inference-engine/tests/unit/vpu/base/graph_transformer_tests.cpp
+++ b/inference-engine/tests/unit/vpu/base/graph_transformer_tests.cpp
@@ -6,6 +6,9 @@
 
 #include <vpu/utils/io.hpp>
 
+#include <vpu/configuration/options/log_level.hpp>
+#include <vpu/configuration/options/copy_optimization.hpp>
+
 #include <atomic>
 #include <iomanip>
 
@@ -287,6 +290,8 @@ void GraphTransformerTest::SetUp() {
     frontEnd = std::make_shared<FrontEnd>(stageBuilder, &_mockCore);
     backEnd = std::make_shared<BackEnd>();
     passManager = std::make_shared<PassManager>(stageBuilder, backEnd);
+
+    config = createConfiguration();
 }
 
 void GraphTransformerTest::TearDown() {
@@ -301,13 +306,13 @@ void GraphTransformerTest::TearDown() {
 
 void GraphTransformerTest::InitCompileEnv() {
     if (const auto envVar = std::getenv("IE_VPU_DUMP_INTERNAL_GRAPH_FILE_NAME")) {
-        config.dumpInternalGraphFileName = envVar;
+        config.compileConfig().dumpInternalGraphFileName = envVar;
     }
     if (const auto envVar = std::getenv("IE_VPU_DUMP_INTERNAL_GRAPH_DIRECTORY")) {
-        config.dumpInternalGraphDirectory = envVar;
+        config.compileConfig().dumpInternalGraphDirectory = envVar;
     }
     if (const auto envVar = std::getenv("IE_VPU_DUMP_ALL_PASSES")) {
-        config.dumpAllPasses = std::stoi(envVar) != 0;
+        config.compileConfig().dumpAllPasses = std::stoi(envVar) != 0;
     }
 
     CompileEnv::init(platform, config, _log);
@@ -342,4 +347,16 @@ TestModel GraphTransformerTest::CreateTestModel() {
     return TestModel(CreateModel());
 }
 
+PluginConfiguration createConfiguration() {
+    PluginConfiguration configuration;
+    configuration.registerOption<LogLevelOption>();
+    configuration.registerOption<CopyOptimizationOption>();
+
+IE_SUPPRESS_DEPRECATED_START
+    configuration.registerDeprecatedOption<LogLevelOption>(VPU_CONFIG_KEY(LOG_LEVEL));
+IE_SUPPRESS_DEPRECATED_END
+
+    return configuration;
+}
+
 } // namespace vpu
diff --git a/inference-engine/tests/unit/vpu/base/graph_transformer_tests.hpp b/inference-engine/tests/unit/vpu/base/graph_transformer_tests.hpp
index ea2fa0de07b694..fd007cca4f40bc 100644
--- a/inference-engine/tests/unit/vpu/base/graph_transformer_tests.hpp
+++ b/inference-engine/tests/unit/vpu/base/graph_transformer_tests.hpp
@@ -130,10 +130,12 @@ void checkStageTestInds(const StageRange& stageRange, std::initializer_list<int>
 
 bool checkExecutionOrder(const Model& model, const std::vector<int>& execOrder);
 
+PluginConfiguration createConfiguration();
+
 class GraphTransformerTest : public ::testing::Test {
 public:
-    Platform platform = Platform::MYRIAD_X;
-    CompilationConfig config;
+    ncDevicePlatform_t platform = ncDevicePlatform_t::NC_MYRIAD_X;
+    PluginConfiguration config;
 
     StageBuilder::Ptr stageBuilder;
     FrontEnd::Ptr frontEnd;
diff --git a/inference-engine/tests/unit/vpu/blob_reader_tests.cpp b/inference-engine/tests/unit/vpu/blob_reader_tests.cpp
index 91115fc522a8dd..0092c540885287 100644
--- a/inference-engine/tests/unit/vpu/blob_reader_tests.cpp
+++ b/inference-engine/tests/unit/vpu/blob_reader_tests.cpp
@@ -12,6 +12,8 @@
 #include <vpu/graph_transformer.hpp>
 #include <vpu/utils/logger.hpp>
 
+#include "graph_transformer_tests.hpp"
+
 #include <ngraph/op/util/attr_types.hpp>
 #include <ngraph_functions/subgraph_builders.hpp>
 
@@ -48,9 +50,8 @@ class VPUBlobReaderHeaderTests: public ::testing::Test, public testing::WithPara
         auto fn_ptr = ngraph::builder::subgraph::makeSplitConvConcat();
         ASSERT_NO_THROW(_network = InferenceEngine::CNNNetwork(fn_ptr));
 
-        CompilationConfig compileConfig;
         auto log = std::make_shared<Logger>("GraphCompiler", LogLevel::None, consoleOutput());
-        _compiledGraph = compileNetwork(_network, Platform::MYRIAD_X, compileConfig, log, &_mockCore);
+        _compiledGraph = compileNetwork(_network, ncDevicePlatform_t::NC_MYRIAD_X, createConfiguration(), log, &_mockCore);
     }
 
     CNNNetwork _network;
diff --git a/inference-engine/tests/unit/vpu/middleend_tests/passes_tests/annotate_memory_types.cpp b/inference-engine/tests/unit/vpu/middleend_tests/passes_tests/annotate_memory_types.cpp
index f1f841101c56bf..14e73c28cc7cec 100644
--- a/inference-engine/tests/unit/vpu/middleend_tests/passes_tests/annotate_memory_types.cpp
+++ b/inference-engine/tests/unit/vpu/middleend_tests/passes_tests/annotate_memory_types.cpp
@@ -24,7 +24,7 @@ class AnnotateMemoryTypes : public GraphTransformerTest, public testing::WithPar
 protected:
     void SetUp() override {
         ASSERT_NO_FATAL_FAILURE(GraphTransformerTest::SetUp());
-        config.enableMemoryTypesAnnotation = true;
+        config.compileConfig().enableMemoryTypesAnnotation = true;
 
         ASSERT_NO_FATAL_FAILURE(InitCompileEnv());
         ASSERT_NO_FATAL_FAILURE(InitPipeline());
diff --git a/inference-engine/tests_deprecated/behavior/shared_tests/plugin_tests/behavior_test_plugin.h b/inference-engine/tests_deprecated/behavior/shared_tests/plugin_tests/behavior_test_plugin.h
index ef1b9e308110c9..e5c4c17679efaa 100644
--- a/inference-engine/tests_deprecated/behavior/shared_tests/plugin_tests/behavior_test_plugin.h
+++ b/inference-engine/tests_deprecated/behavior/shared_tests/plugin_tests/behavior_test_plugin.h
@@ -10,7 +10,6 @@
 #include <inference_engine.hpp>
 #include <ie_plugin_config.hpp>
 #include <vpu/vpu_plugin_config.hpp>
-#include <vpu/private_plugin_config.hpp>
 #include <gna/gna_config.hpp>
 #include <common_test_utils/test_assertions.hpp>
 #include <memory>
@@ -111,4 +110,4 @@ const TestModel convReluNormPoolFcModelQ78 = getConvReluNormPoolFcModel(Inferenc
 class FPGAHangingTest : public BehaviorPluginTest {
 };
 
-#endif
\ No newline at end of file
+#endif
diff --git a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_devices.cpp b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_devices.cpp
index d5adac10397fc5..5016a4a3b9d8b3 100644
--- a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_devices.cpp
+++ b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_devices.cpp
@@ -30,7 +30,6 @@ std::vector<std::string> MyriadDevicesInfo::getDevicesList(
                     const ncDeviceProtocol_t deviceProtocol,
                     const ncDevicePlatform_t devicePlatform,
                     const XLinkDeviceState_t state) {
-
         deviceDesc_t req_deviceDesc = {};
         req_deviceDesc.protocol = convertProtocolToXlink(deviceProtocol);
         req_deviceDesc.platform = convertPlatformToXlink(devicePlatform);
diff --git a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_devices.hpp b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_devices.hpp
index a5628fcd0e60fb..ed9265f868cc5a 100644
--- a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_devices.hpp
+++ b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_devices.hpp
@@ -30,8 +30,7 @@ class MyriadDevicesInfo {
     std::vector<std::string> getDevicesList(
             const ncDeviceProtocol_t deviceProtocol = NC_ANY_PROTOCOL,
             const ncDevicePlatform_t devicePlatform = NC_ANY_PLATFORM,
-            const XLinkDeviceState_t state = X_LINK_ANY_STATE
-            );
+            const XLinkDeviceState_t state = X_LINK_ANY_STATE);
 
     inline bool isMyriadXDevice(const std::string &device_name);
     inline bool isMyriad2Device(const std::string &device_name);
@@ -77,4 +76,4 @@ long MyriadDevicesInfo::getAmountOfUnbootedDevices(const ncDeviceProtocol_t devi
 
 long MyriadDevicesInfo::getAmountOfBootedDevices(const ncDeviceProtocol_t deviceProtocol) {
     return getAmountOfDevices(deviceProtocol, NC_ANY_PLATFORM, X_LINK_BOOTED);
-}
\ No newline at end of file
+}
diff --git a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_protocol_case.cpp b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_protocol_case.cpp
index 48c8a413b84feb..a6b77a6b523a04 100644
--- a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_protocol_case.cpp
+++ b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/helpers/myriad_protocol_case.cpp
@@ -4,6 +4,7 @@
 
 #include "myriad_protocol_case.hpp"
 #include "mvnc_ext.h"
+#include "vpu/myriad_config.hpp"
 
 void MyriadProtocolTests::SetUp() {
     protocol = GetParam();
diff --git a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_watchdog_tests.cpp b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_watchdog_tests.cpp
index c2ecf468b4638e..f55dd04473594a 100644
--- a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_watchdog_tests.cpp
+++ b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_watchdog_tests.cpp
@@ -15,6 +15,7 @@
 #include "helpers/myriad_devices.hpp"
 
 #include <cpp/ie_plugin.hpp>
+#include <vpu/private_plugin_config.hpp>
 
 using namespace std;
 using namespace ::testing;
diff --git a/inference-engine/tests_deprecated/behavior/vpu/shared_tests_instances/plugin_tests/vpu_test_data.hpp b/inference-engine/tests_deprecated/behavior/vpu/shared_tests_instances/plugin_tests/vpu_test_data.hpp
index b1bb32646b6529..6a1ab38d36a4e2 100644
--- a/inference-engine/tests_deprecated/behavior/vpu/shared_tests_instances/plugin_tests/vpu_test_data.hpp
+++ b/inference-engine/tests_deprecated/behavior/vpu/shared_tests_instances/plugin_tests/vpu_test_data.hpp
@@ -3,6 +3,7 @@
 //
 
 #include "behavior_test_plugin.h"
+#include "vpu/myriad_config.hpp"
 
 // correct params
 #define BEH_MYRIAD BehTestParams("MYRIAD", \
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_blob_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_blob_test.cpp
index fc210dee5b7d26..7a2c027be03113 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_blob_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_blob_test.cpp
@@ -169,10 +169,15 @@ TEST_F(myriadConfigsWithBlobImportTests_smoke, TryingToSetCompileOptionPrintsWar
 
     std::string content = redirectCoutStream.str();
     for (auto &&elem : config) {
+        // TODO: remove once all options are migrated
+        std::stringstream deprecatedExpectedMsgStream;
+        deprecatedExpectedMsgStream << "[Warning][VPU][Config] " << elem.first;
+        const auto& deprecatedMsg = deprecatedExpectedMsgStream.str();
+
         std::stringstream expectedMsgStream;
-        expectedMsgStream << "[Warning][VPU][Config] " << elem.first;
-        std::string msg = expectedMsgStream.str();
-        ASSERT_TRUE(content.find(msg) != std::string::npos) << msg;
+        expectedMsgStream << "[Warning][VPU][Configuration] Configuration option \"" << elem.first;
+        const auto& msg = expectedMsgStream.str();
+        ASSERT_TRUE(content.find(msg) != std::string::npos || content.find(deprecatedMsg) != std::string::npos) << msg;
     }
 }
 
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_concat_test.hpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_concat_test.hpp
index 775708c6389c7f..5d6074334f810f 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_concat_test.hpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_concat_test.hpp
@@ -5,6 +5,7 @@
 #include "myriad_layers_tests.hpp"
 
 using namespace InferenceEngine;
+using namespace InferenceEngine::details;
 
 using myriadConcatTestParams = std::tuple<InferenceEngine::SizeVector, int32_t, InferenceEngine::SizeVector, int32_t, int32_t >;
 typedef myriadLayerTestBaseWithParam<myriadConcatTestParams> myriadLayersTestsConcat_smoke;
diff --git a/inference-engine/tests_deprecated/functional/vpu/graph_transformer/gt_functional_tests.cpp b/inference-engine/tests_deprecated/functional/vpu/graph_transformer/gt_functional_tests.cpp
index b5c98e0b368a97..ca07c26f03a92e 100644
--- a/inference-engine/tests_deprecated/functional/vpu/graph_transformer/gt_functional_tests.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/graph_transformer/gt_functional_tests.cpp
@@ -7,6 +7,8 @@
 #include <vpu/utils/logger.hpp>
 #include <vpu/compile_env.hpp>
 #include <vpu/graph_transformer_internal.hpp>
+#include <vpu/configuration/options/log_level.hpp>
+#include <vpu/configuration/options/copy_optimization.hpp>
 
 using namespace InferenceEngine;
 using namespace vpu;
@@ -19,12 +21,12 @@ void graphTransformerFunctionalTests::SetUp() {
     vpuLayersTests::SetUp();
 
     _stageBuilder = std::make_shared<StageBuilder>();
-    _platform = CheckMyriadX() ? Platform::MYRIAD_X : Platform::MYRIAD_2;
+    _platform = CheckMyriadX() ? ncDevicePlatform_t::NC_MYRIAD_X : ncDevicePlatform_t::NC_MYRIAD_2;
 }
 
 void graphTransformerFunctionalTests::CreateModel() {
     const auto compilerLog = std::make_shared<Logger>("Test", LogLevel::Info, consoleOutput());
-    CompileEnv::init(_platform, _compilationConfig, compilerLog);
+    CompileEnv::init(_platform, _configuration, compilerLog);
     AutoScope autoDeinit([] {
         CompileEnv::free();
     });
@@ -43,7 +45,13 @@ void graphTransformerFunctionalTests::CreateModel() {
 
 void graphTransformerFunctionalTests::PrepareGraphCompilation() {
     SetSeed(DEFAULT_SEED_VALUE);
-    _compilationConfig = CompilationConfig();
+
+    _configuration.registerOption<LogLevelOption>();
+    _configuration.registerOption<CopyOptimizationOption>();
+IE_SUPPRESS_DEPRECATED_START
+    _configuration.registerDeprecatedOption<LogLevelOption>(VPU_CONFIG_KEY(LOG_LEVEL));
+IE_SUPPRESS_DEPRECATED_END
+
     _inputsInfo.clear();
     _outputsInfo.clear();
     _inputMap.clear();
@@ -87,7 +95,7 @@ int64_t graphTransformerFunctionalTests::CompileAndInfer(Blob::Ptr& inputBlob, B
     auto compiledGraph = compileModel(
                 _gtModel,
                 _platform,
-                _compilationConfig,
+                _configuration,
                 compilerLog);
 
     std::istringstream instream(std::string(compiledGraph->blob.data(), compiledGraph->blob.size()));
diff --git a/inference-engine/tests_deprecated/functional/vpu/graph_transformer/gt_functional_tests.hpp b/inference-engine/tests_deprecated/functional/vpu/graph_transformer/gt_functional_tests.hpp
index 155b9f21ea3683..ff3063b217f7f2 100644
--- a/inference-engine/tests_deprecated/functional/vpu/graph_transformer/gt_functional_tests.hpp
+++ b/inference-engine/tests_deprecated/functional/vpu/graph_transformer/gt_functional_tests.hpp
@@ -25,12 +25,12 @@ class graphTransformerFunctionalTests : public vpuLayersTests {
                             bool lockLayout = false);
 
 protected:
-   vpu::ModelPtr          _gtModel;
-   vpu::CompilationConfig _compilationConfig;
-   vpu::StageBuilder::Ptr _stageBuilder;
-   vpu::Data              _dataIntermediate;
+   vpu::ModelPtr            _gtModel;
+   vpu::PluginConfiguration _configuration;
+   vpu::StageBuilder::Ptr   _stageBuilder;
+   vpu::Data                _dataIntermediate;
 
 private:
-   vpu::Platform                      _platform = vpu::Platform::MYRIAD_X;
+   ncDevicePlatform_t _platform = ncDevicePlatform_t::NC_MYRIAD_X;
    InferenceEngine::ExecutableNetwork _executableNetwork;
 };
diff --git a/inference-engine/tests_deprecated/functional/vpu/graph_transformer/merge_permute_and_reorder_test.cpp b/inference-engine/tests_deprecated/functional/vpu/graph_transformer/merge_permute_and_reorder_test.cpp
index abcf7b062c6133..ec6332dcc98231 100644
--- a/inference-engine/tests_deprecated/functional/vpu/graph_transformer/merge_permute_and_reorder_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/graph_transformer/merge_permute_and_reorder_test.cpp
@@ -62,8 +62,8 @@ class myriadGTMergePermuteNDTests_nightly:
                                  const bool usePermuteMerging,
                                  Blob::Ptr& outputBlob) {
         PrepareGraphCompilation();
-        _compilationConfig.detectBatch = false;
-        _compilationConfig.enablePermuteMerging = usePermuteMerging;
+        _configuration.compileConfig().detectBatch = false;
+        _configuration.compileConfig().enablePermuteMerging = usePermuteMerging;
 
         IE_ASSERT(permutationVectors.size() >= 2);
 
diff --git a/inference-engine/tests_deprecated/helpers/tests_vpu_common.hpp b/inference-engine/tests_deprecated/helpers/tests_vpu_common.hpp
index 12417a62a0027b..5cfc76f825b486 100644
--- a/inference-engine/tests_deprecated/helpers/tests_vpu_common.hpp
+++ b/inference-engine/tests_deprecated/helpers/tests_vpu_common.hpp
@@ -13,7 +13,7 @@
 
 #include "single_layer_common.hpp"
 #include "vpu/vpu_plugin_config.hpp"
-#include <graph_transformer/include/vpu/private_plugin_config.hpp>
+#include <vpu/private_plugin_config.hpp>
 
 
 using config_t = std::map<std::string, std::string>;
diff --git a/inference-engine/tests_deprecated/unit/CMakeLists.txt b/inference-engine/tests_deprecated/unit/CMakeLists.txt
index 18d7724add571a..c222e01f2b30b3 100644
--- a/inference-engine/tests_deprecated/unit/CMakeLists.txt
+++ b/inference-engine/tests_deprecated/unit/CMakeLists.txt
@@ -63,6 +63,7 @@ if (ENABLE_MYRIAD)
             engines/vpu/myriad_tests/helpers/*cpp
             engines/vpu/myriad_tests/helpers/*h
             ${IE_MAIN_SOURCE_DIR}/src/vpu/myriad_plugin/*.cpp
+            ${IE_MAIN_SOURCE_DIR}/src/vpu/myriad_plugin/configuration/*.cpp
             )
     include_directories(
             engines/vpu/myriad_tests/helpers
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_squeeze_test.cpp b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_squeeze_test.cpp
index ee3b84c8c83d84..603efb6578e9cd 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_squeeze_test.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_squeeze_test.cpp
@@ -12,6 +12,8 @@
 #include <debug.h>
 #include "../gna_matcher.hpp"
 
+using namespace InferenceEngine::details;
+
 typedef struct {
     std::vector<size_t> input_shape;
     std::vector<size_t> squeeze_indices;
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/adjust_data_location_tests.cpp b/inference-engine/tests_deprecated/unit/engines/vpu/adjust_data_location_tests.cpp
index 5e5ebf2908222c..77ce658129ffbe 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/adjust_data_location_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/adjust_data_location_tests.cpp
@@ -25,8 +25,8 @@ using VPU_AdjustDataLocationTest = GraphTransformerTest;
 //
 
 TEST_F(VPU_AdjustDataLocationTest, FlushCMX_TwoSpecialConsumers) {
-    config.numSHAVEs = 1;
-    config.numCMXSlices = 1;
+    config.compileConfig().numSHAVEs = 1;
+    config.compileConfig().numCMXSlices = 1;
     InitCompileEnv();
 
     DataDesc dataDesc1(DataType::FP16, DimsOrder::NCHW, {CMX_SLICE_SIZE / (2 * 2), 1, 2, 1});
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/get_vpu_scale_from_ir_tests.cpp b/inference-engine/tests_deprecated/unit/engines/vpu/get_vpu_scale_from_ir_tests.cpp
index aac3ac3ee97b2d..d0c2855f6794ca 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/get_vpu_scale_from_ir_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/get_vpu_scale_from_ir_tests.cpp
@@ -24,8 +24,8 @@ TEST_F(VPU_AddVpuScaleTest, CanAddVpuScaleToNetwork) {
     InitCompileEnv();
 
     auto& env = CompileEnv::get();
-    CompilationConfig config{};
-    config.irWithVpuScalesDir = "/";
+    auto config = createConfiguration();
+    config.compileConfig().irWithVpuScalesDir = "/";
     env.updateConfig(config);
 
     std::shared_ptr<ngraph::Function> function;
@@ -69,8 +69,8 @@ TEST_F(VPU_AddVpuScaleTest, CanAddVpuScaleToNetwork) {
 TEST_F(VPU_AddVpuScaleTest, VpuScaleFromIrChangesWeights) {
     InitCompileEnv();
     const auto& env = CompileEnv::get();
-    CompilationConfig config{};
-    config.irWithVpuScalesDir = "/";
+    auto config = createConfiguration();
+    config.compileConfig().irWithVpuScalesDir = "/";
     env.updateConfig(config);
 
     std::shared_ptr<ngraph::Function> function;
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/graph_transformer_tests.cpp b/inference-engine/tests_deprecated/unit/engines/vpu/graph_transformer_tests.cpp
index cd4b1619bdfc93..d0cd3a315ccc80 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/graph_transformer_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/graph_transformer_tests.cpp
@@ -8,6 +8,8 @@
 #include <iomanip>
 
 #include <vpu/utils/io.hpp>
+#include <vpu/configuration/options/log_level.hpp>
+#include <vpu/configuration/options/copy_optimization.hpp>
 
 namespace vpu {
 
@@ -158,6 +160,18 @@ void TestModel::setStageBatchInfo(
     }
 }
 
+PluginConfiguration createConfiguration() {
+    PluginConfiguration configuration;
+    configuration.registerOption<LogLevelOption>();
+    configuration.registerOption<CopyOptimizationOption>();
+
+IE_SUPPRESS_DEPRECATED_START
+    configuration.registerDeprecatedOption<LogLevelOption>(VPU_CONFIG_KEY(LOG_LEVEL));
+IE_SUPPRESS_DEPRECATED_END
+
+    return configuration;
+}
+
 void GraphTransformerTest::SetUp() {
     ASSERT_NO_FATAL_FAILURE(TestsCommon::SetUp());
 
@@ -170,6 +184,8 @@ void GraphTransformerTest::SetUp() {
     frontEnd = std::make_shared<FrontEnd>(stageBuilder, &_mockCore);
     backEnd = std::make_shared<BackEnd>();
     passManager = std::make_shared<PassManager>(stageBuilder, backEnd);
+
+    config = createConfiguration();
 }
 
 void GraphTransformerTest::TearDown() {
@@ -186,13 +202,13 @@ void GraphTransformerTest::TearDown() {
 
 void GraphTransformerTest::InitCompileEnv() {
     if (const auto envVar = std::getenv("IE_VPU_DUMP_INTERNAL_GRAPH_FILE_NAME")) {
-        config.dumpInternalGraphFileName = envVar;
+        config.compileConfig().dumpInternalGraphFileName = envVar;
     }
     if (const auto envVar = std::getenv("IE_VPU_DUMP_INTERNAL_GRAPH_DIRECTORY")) {
-        config.dumpInternalGraphDirectory = envVar;
+        config.compileConfig().dumpInternalGraphDirectory = envVar;
     }
     if (const auto envVar = std::getenv("IE_VPU_DUMP_ALL_PASSES")) {
-        config.dumpAllPasses = std::stoi(envVar) != 0;
+        config.compileConfig().dumpAllPasses = std::stoi(envVar) != 0;
     }
 
     CompileEnv::init(platform, config, _log);
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/graph_transformer_tests.hpp b/inference-engine/tests_deprecated/unit/engines/vpu/graph_transformer_tests.hpp
index dc9768ea529ce1..6a656bc94d2f37 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/graph_transformer_tests.hpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/graph_transformer_tests.hpp
@@ -176,10 +176,12 @@ void CheckStageTestInds(const StageRange& stageRange, std::initializer_list<int>
     }
 }
 
+PluginConfiguration createConfiguration();
+
 class GraphTransformerTest : public TestsCommon {
 public:
-    Platform platform = Platform::MYRIAD_X;
-    CompilationConfig config;
+    ncDevicePlatform_t platform = ncDevicePlatform_t::NC_MYRIAD_X;
+    PluginConfiguration config;
 
     StageBuilder::Ptr stageBuilder;
     FrontEnd::Ptr frontEnd;
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/replace_deconv_by_conv_tests.cpp b/inference-engine/tests_deprecated/unit/engines/vpu/replace_deconv_by_conv_tests.cpp
index ef18901be416d1..59e9c77dadc023 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/replace_deconv_by_conv_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/replace_deconv_by_conv_tests.cpp
@@ -91,7 +91,7 @@ TEST_F(VPU_ReplaceDeconvByConvTest, deconvReplacedByConvIfKernelSizeFitsHWUnit)
 }
 
 TEST_F(VPU_ReplaceDeconvByConvTest, deconvCannotBeReplacedByConvIfDisabledInConfig) {
-    config.hwBlackList.insert("deconv");
+    config.compileConfig().hwBlackList.insert("deconv");
     InitCompileEnv();
     InitDeconvStage(16, 15);
 
diff --git a/inference-engine/tests_deprecated/unit/inference_engine_tests/get_num_iterations_test.cpp b/inference-engine/tests_deprecated/unit/inference_engine_tests/get_num_iterations_test.cpp
index 36990430195bba..01fb9f61163ba5 100644
--- a/inference-engine/tests_deprecated/unit/inference_engine_tests/get_num_iterations_test.cpp
+++ b/inference-engine/tests_deprecated/unit/inference_engine_tests/get_num_iterations_test.cpp
@@ -10,6 +10,8 @@
 
 namespace {
 
+using InferenceEngine::details::operator<<;
+
 struct From {
     explicit From(int new_value) : value(new_value) {}
     int value;
diff --git a/inference-engine/thirdparty/movidius/XLink/CMakeLists.txt b/inference-engine/thirdparty/movidius/XLink/CMakeLists.txt
index 6e7f904227afec..9b365d8da3961f 100644
--- a/inference-engine/thirdparty/movidius/XLink/CMakeLists.txt
+++ b/inference-engine/thirdparty/movidius/XLink/CMakeLists.txt
@@ -40,3 +40,6 @@ if (ENABLE_MYRIAD_NO_BOOT)
 endif()
 
 set_property(TARGET ${TARGET_NAME} PROPERTY C_STANDARD 99)
+
+# TODO: remove once all options are migrated
+ie_developer_export_targets(${TARGET_NAME})
diff --git a/inference-engine/thirdparty/movidius/mvnc/CMakeLists.txt b/inference-engine/thirdparty/movidius/mvnc/CMakeLists.txt
index 37fd33e136bfe5..31a22c740351a8 100644
--- a/inference-engine/thirdparty/movidius/mvnc/CMakeLists.txt
+++ b/inference-engine/thirdparty/movidius/mvnc/CMakeLists.txt
@@ -71,6 +71,9 @@ if(NOT WIN32)
             ${LIBUSB_LIBRARY})
 endif()
 
+# TODO: remove once all options are migrated
+ie_developer_export_targets(${TARGET_NAME})
+
 if(ENABLE_TESTS AND ENABLE_MYRIAD_MVNC_TESTS)
     add_subdirectory(tests)
 endif()
diff --git a/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_common.cpp b/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_common.cpp
index 0f7b69d1b4935a..1a3bca457f9a0e 100644
--- a/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_common.cpp
+++ b/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_common.cpp
@@ -300,7 +300,6 @@ TEST_P(MvncOpenDevice, WatchdogShouldResetDeviceWithoutConnection) {
     if (availableDevices_ == 0)
         GTEST_SKIP() << ncProtocolToStr(_deviceProtocol) << " devices not found";
 
-    ncDeviceHandle_t*   deviceHandle = nullptr;
     std::string         deviceName;
     deviceDesc_t deviceDescToBoot = {};
     deviceDesc_t in_deviceDesc = {};

From 6d4704ba9dc953d9135bd4ffe9b39378a9e00836 Mon Sep 17 00:00:00 2001
From: Maria Kaglinskaya <maria.kaglinskaya@intel.com>
Date: Thu, 17 Jun 2021 19:59:33 +0300
Subject: [PATCH 124/178] Fix klocwork master (#6219)

* Fix klockwork issues in pruning transformation

* Fixed tabs
---
 .../src/offline_transformations/include/mask_attribute.hpp  | 6 ++++--
 .../offline_transformations/src/pruning/propagate_masks.cpp | 6 ++++++
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/inference-engine/src/offline_transformations/include/mask_attribute.hpp b/inference-engine/src/offline_transformations/include/mask_attribute.hpp
index 48c5b4ee9f0864..282f81b054e9f3 100644
--- a/inference-engine/src/offline_transformations/include/mask_attribute.hpp
+++ b/inference-engine/src/offline_transformations/include/mask_attribute.hpp
@@ -96,7 +96,8 @@ class Mask : public std::vector<std::set<uint64_t>>,
         auto mask_2_iter = mask->rbegin();
 
         while (mask_1_iter != rend() &&
-               mask_2_iter != mask->rend()) {
+               mask_2_iter != mask->rend() &&
+               result_iter != result_mask->rend()) {
             // Merge mask dimension values for both masks
             // Example: (MaskValue[1,2,3,4], MaskValue[2,3]) -> MaskValue[2,3]
             for (const auto & value : *mask_1_iter) {
@@ -119,7 +120,8 @@ class Mask : public std::vector<std::set<uint64_t>>,
         auto mask_2_iter = mask->rbegin();
 
         while (mask_1_iter != rend() &&
-               mask_2_iter != mask->rend()) {
+               mask_2_iter != mask->rend() &&
+               result_iter != result_mask->rend()) {
             // Union mask dimension values for both masks
             // Example: (MaskValue[1,2,3,4], MaskValue[2, 5]) -> MaskValue[1, 2, 3, 4, 5]
             for (const auto & value : *mask_1_iter) {
diff --git a/inference-engine/src/offline_transformations/src/pruning/propagate_masks.cpp b/inference-engine/src/offline_transformations/src/pruning/propagate_masks.cpp
index 424b6ae958317d..271b200f31b5bc 100644
--- a/inference-engine/src/offline_transformations/src/pruning/propagate_masks.cpp
+++ b/inference-engine/src/offline_transformations/src/pruning/propagate_masks.cpp
@@ -246,6 +246,9 @@ class ngraph::pass::mask_propagation::Reshape : public MatcherPass {
             // To allow pruning on weights (allow reshape input Group (0) dim changing) replace Reshape Shape constant
             // [G, 1, 1, X, Y, Z] by [-1, 1, 1, X, Y, Z].
             auto old_shape_const = std::dynamic_pointer_cast<opset6::Constant>(m_shape.get_node_shared_ptr());
+            if (!old_shape_const) {
+                return false;
+            }
             auto shape_value = old_shape_const.get()->cast_vector<int64_t>();
             shape_value[0] = -1;
             auto new_const = opset6::Constant::create(old_shape_const->get_element_type(),
@@ -462,6 +465,9 @@ class ngraph::pass::mask_propagation::Concat : public MatcherPass{
             const auto & pattern_map = m.get_pattern_value_map();
             const auto & m_output = pattern_map.at(concat);
             auto concat_ptr = std::dynamic_pointer_cast<opset6::Concat>(m_output.get_node_shared_ptr());
+            if (!concat_ptr) {
+                return false;
+            }
             auto axis = concat_ptr->get_concatenation_axis();
 
             auto inputs = concat_ptr->inputs();

From 03b00f63bc06003cae49a1bf9d3d2c793592fb88 Mon Sep 17 00:00:00 2001
From: Egor Duplensky <egor.duplenskii@intel.com>
Date: Thu, 17 Jun 2021 21:18:14 +0300
Subject: [PATCH 125/178] [CPU] Extend blob dumping (#5939)

* [CPU] Extend blob dumping

* Allow multiple node types
* Allow dump only input / output blobs
* Allow dump all blobs using "*"
---
 .../src/mkldnn_plugin/utils/README.md         | 38 +++++++++---
 .../mkldnn_plugin/utils/debug_capabilities.h  |  2 +
 .../src/mkldnn_plugin/utils/node_dumper.cpp   | 62 +++++++++++++------
 .../src/mkldnn_plugin/utils/node_dumper.h     | 13 ++--
 4 files changed, 83 insertions(+), 32 deletions(-)

diff --git a/inference-engine/src/mkldnn_plugin/utils/README.md b/inference-engine/src/mkldnn_plugin/utils/README.md
index d3b98f1cb48f2f..bd3f1329a5e785 100644
--- a/inference-engine/src/mkldnn_plugin/utils/README.md
+++ b/inference-engine/src/mkldnn_plugin/utils/README.md
@@ -6,7 +6,7 @@ Use the following cmake option to enable debug capabilities:
 ## Blob dumping
 Blob dumping is controlled by environment variables (filters).
 
-The variables define conditions of the node which input, output and internal blobs
+The variables define conditions of the node which input and output blobs
 should be dumped for.
 
 > **NOTE**: Nothing is dumped by default
@@ -15,11 +15,13 @@ should be dumped for.
 
 Environment variables can be set per execution, for example:
 ```sh
-    OV_CPU_BLOB_DUMP_DIR=dump_dir binary ...
+    OV_CPU_BLOB_DUMP_DIR=dump_dir OV_CPU_BLOB_DUMP_FORMAT=TEXT OV_CPU_BLOB_DUMP_NODE_PORTS=OUT binary ...
 ```
 or for shell session (bash example):
 ```sh
     export OV_CPU_BLOB_DUMP_DIR=dump_dir
+    export OV_CPU_BLOB_DUMP_FORMAT=TEXT
+    export OV_CPU_BLOB_DUMP_NODE_PORTS=OUT
     binary ...
 ```
 ### Specify dump directory
@@ -35,8 +37,22 @@ Options are:
 * BIN (default)
 * TEXT
 
+### Filter input / output blobs
+To dump only input / output blobs:
+```sh
+    OV_CPU_BLOB_DUMP_NODE_PORTS='<ports_kind>' binary ...
+```
+Example:
+```sh
+    OV_CPU_BLOB_DUMP_NODE_PORTS=OUT binary ...
+```
+Options are:
+* IN
+* OUT
+* ALL
+
 ### Filter by execution ID
-To dump blobs only for node with specified execution IDs:
+To dump blobs only for nodes with specified execution IDs:
 ```sh
     OV_CPU_BLOB_DUMP_NODE_EXEC_ID='<space_separated_list_of_ids>' binary ...
 ```
@@ -46,19 +62,19 @@ Example:
 ```
 
 ### Filter by type
-To dump blobs only for node with specified type:
+To dump blobs only for nodes with specified types:
 ```sh
-    OV_CPU_BLOB_DUMP_NODE_TYPE=<type> binary ...
+    OV_CPU_BLOB_DUMP_NODE_TYPE=<space_separated_list_of_types> binary ...
 ```
 Example:
 ```sh
-    OV_CPU_BLOB_DUMP_NODE_TYPE=Convolution binary ...
+    OV_CPU_BLOB_DUMP_NODE_TYPE='Convolution Reorder' binary ...
 ```
 
 > **NOTE**: see **enum Type** in [mkldnn_node.h](../mkldnn_node.h) for list of the types
 
 ### Filter by name
-To dump blobs only for node with name matching specified regex:
+To dump blobs only for nodes with name matching specified regex:
 ```sh
     OV_CPU_BLOB_DUMP_NODE_NAME=<regex> binary ...
 ```
@@ -68,9 +84,17 @@ Example:
 ```
 
 ### Dump all the blobs
+```sh
+    OV_CPU_BLOB_DUMP_NODE_NAME="*" binary ...
+```
+    or
 ```sh
     OV_CPU_BLOB_DUMP_NODE_NAME=".+" binary ...
 ```
+    or
+```sh
+    OV_CPU_BLOB_DUMP_NODE_PORTS=ALL binary ...
+```
 
 ## Graph serialization
 The functionality allows to serialize execution graph using environment variable:
diff --git a/inference-engine/src/mkldnn_plugin/utils/debug_capabilities.h b/inference-engine/src/mkldnn_plugin/utils/debug_capabilities.h
index be6e7a830c2c63..c2784f8a46772f 100644
--- a/inference-engine/src/mkldnn_plugin/utils/debug_capabilities.h
+++ b/inference-engine/src/mkldnn_plugin/utils/debug_capabilities.h
@@ -20,6 +20,7 @@ class Config {
         readParam(blobDumpDir, "OV_CPU_BLOB_DUMP_DIR");
         readParam(blobDumpFormat, "OV_CPU_BLOB_DUMP_FORMAT");
         readParam(blobDumpNodeExecId, "OV_CPU_BLOB_DUMP_NODE_EXEC_ID");
+        readParam(blobDumpNodePorts, "OV_CPU_BLOB_DUMP_NODE_PORTS");
         readParam(blobDumpNodeType, "OV_CPU_BLOB_DUMP_NODE_TYPE");
         readParam(blobDumpNodeName, "OV_CPU_BLOB_DUMP_NODE_NAME");
         readParam(execGraphPath, "OV_CPU_EXEC_GRAPH_PATH");
@@ -28,6 +29,7 @@ class Config {
     std::string blobDumpDir;
     std::string blobDumpFormat;
     std::string blobDumpNodeExecId;
+    std::string blobDumpNodePorts;
     std::string blobDumpNodeType;
     std::string blobDumpNodeName;
     std::string execGraphPath;
diff --git a/inference-engine/src/mkldnn_plugin/utils/node_dumper.cpp b/inference-engine/src/mkldnn_plugin/utils/node_dumper.cpp
index 9f3af44a66a26a..1cfbae1ab5ff09 100644
--- a/inference-engine/src/mkldnn_plugin/utils/node_dumper.cpp
+++ b/inference-engine/src/mkldnn_plugin/utils/node_dumper.cpp
@@ -20,7 +20,7 @@ using namespace InferenceEngine;
 namespace MKLDNNPlugin {
 
 NodeDumper::NodeDumper(const DebugCaps::Config& config, const int _count)
-    : dumpFormat(DUMP_FORMAT::BIN)
+    : dumpFormat(FORMAT::BIN)
     , dumpDirName("mkldnn_dump")
     , count(_count) {
     if (!config.blobDumpDir.empty())
@@ -32,6 +32,9 @@ NodeDumper::NodeDumper(const DebugCaps::Config& config, const int _count)
     if (!config.blobDumpNodeExecId.empty())
         dumpFilters[FILTER::BY_EXEC_ID] = config.blobDumpNodeExecId;
 
+    if (!config.blobDumpNodePorts.empty())
+        dumpFilters[FILTER::BY_PORTS] = config.blobDumpNodePorts;
+
     if (!config.blobDumpNodeType.empty())
         dumpFilters[FILTER::BY_TYPE] = config.blobDumpNodeType;
 
@@ -40,7 +43,7 @@ NodeDumper::NodeDumper(const DebugCaps::Config& config, const int _count)
 }
 
 void NodeDumper::dumpInputBlobs(const MKLDNNNodePtr& node) const {
-    if (!shouldBeDumped(node))
+    if (!shouldBeDumped(node, "IN"))
         return;
 
     auto exec_order = std::to_string(node->getExecIndex());
@@ -60,7 +63,7 @@ void NodeDumper::dumpInputBlobs(const MKLDNNNodePtr& node) const {
             file_name = file_name.substr(file_name.size() - 240);
 
         auto dump_file = dumpDirName + "/#" + exec_order + "_" + file_name;
-        std::cout << "Dump before: " << dump_file << std::endl;
+        std::cout << "Dump inputs: " << dump_file << std::endl;
 
         TensorDesc desc = prEdge->getDesc();
         if (desc.getPrecision() == Precision::BIN)
@@ -77,7 +80,7 @@ void NodeDumper::dumpInputBlobs(const MKLDNNNodePtr& node) const {
 }
 
 void NodeDumper::dumpOutputBlobs(const MKLDNNNodePtr& node) const {
-    if (!shouldBeDumped(node))
+    if (!shouldBeDumped(node, "OUT"))
         return;
 
     auto exec_order = std::to_string(node->getExecIndex());
@@ -96,7 +99,7 @@ void NodeDumper::dumpOutputBlobs(const MKLDNNNodePtr& node) const {
             file_name = file_name.substr(file_name.size() - 240);
 
         auto dump_file = dumpDirName + "/#" + exec_order + "_" + file_name;
-        std::cout << "Dump after:  " << dump_file << std::endl;
+        std::cout << "Dump outputs:  " << dump_file << std::endl;
 
         TensorDesc desc = childEdge->getDesc();
         if (desc.getPrecision() == Precision::BIN)
@@ -130,56 +133,77 @@ void NodeDumper::dumpInternalBlobs(const MKLDNNNodePtr& node) const {
 
 void NodeDumper::dump(const BlobDumper& bd, const std::string& file) const {
     switch (dumpFormat) {
-    case DUMP_FORMAT::BIN: {
+    case FORMAT::BIN: {
         bd.dump(file);
         break;
     }
-    case DUMP_FORMAT::TEXT: {
+    case FORMAT::TEXT: {
         bd.dumpAsTxt(file);
         break;
     }
     default:
-        IE_THROW() << "Unknown dump format";
+        IE_THROW() << "NodeDumper: Unknown dump format";
     }
 }
 
-bool NodeDumper::shouldBeDumped(const MKLDNNNodePtr& node) const {
+bool NodeDumper::shouldBeDumped(const MKLDNNNodePtr& node, const std::string& portsKind) const {
     if (dumpFilters.empty())
         return false;
 
-    if (dumpFilters.count(FILTER::BY_EXEC_ID)) { // filter by exec id env set
+    if (dumpFilters.count(FILTER::BY_PORTS)) { // filter by ports configured
+        if (dumpFilters.at(FILTER::BY_PORTS) != "ALL" &&
+            portsKind != dumpFilters.at(FILTER::BY_PORTS))
+            return false;
+    }
+
+    if (dumpFilters.count(FILTER::BY_EXEC_ID)) { // filter by exec id configured
         std::stringstream ss(dumpFilters.at(FILTER::BY_EXEC_ID));
         int id;
         bool matched = false;
+
         while (ss >> id) {
-            if (node->getExecIndex() == id) // exec id matches
+            if (node->getExecIndex() == id) {// exec id matches
                 matched = true;
+                break;
+            }
         }
 
         if (!matched)
             return false;
     }
 
-    if (dumpFilters.count(FILTER::BY_TYPE)) { // filter by type env set
-        if (NameFromType(node->getType()) != dumpFilters.at(FILTER::BY_TYPE)) // type does not match
+    if (dumpFilters.count(FILTER::BY_TYPE)) { // filter by type configured
+        std::stringstream ss(dumpFilters.at(FILTER::BY_TYPE));
+        std::string type;
+        bool matched = false;
+
+        while (ss >> type) {
+            if (NameFromType(node->getType()) == type) {// type does not match
+                matched = true;
+                break;
+            }
+        }
+
+        if (!matched)
             return false;
     }
 
-    if (dumpFilters.count(FILTER::BY_NAME)) { // filter by name env set
-        if (!std::regex_match(node->getName(), std::regex(dumpFilters.at(FILTER::BY_NAME)))) // name does not match
+    if (dumpFilters.count(FILTER::BY_NAME)) { // filter by name configured
+        if (dumpFilters.at(FILTER::BY_NAME) != "*" && // to have 'single char' option for matching all the names
+            !std::regex_match(node->getName(), std::regex(dumpFilters.at(FILTER::BY_NAME)))) // name does not match
             return false;
     }
 
     return true;
 }
 
-NodeDumper::DUMP_FORMAT NodeDumper::parseDumpFormat(const std::string& format) const {
+NodeDumper::FORMAT NodeDumper::parseDumpFormat(const std::string& format) const {
     if (format == "BIN")
-        return DUMP_FORMAT::BIN;
+        return FORMAT::BIN;
     else if (format == "TEXT")
-        return DUMP_FORMAT::TEXT;
+        return FORMAT::TEXT;
     else
-        IE_THROW() << "Unknown dump format";
+        IE_THROW() << "NodeDumper: Unknown dump format";
 }
 
 void NodeDumper::formatNodeName(std::string& name) const {
diff --git a/inference-engine/src/mkldnn_plugin/utils/node_dumper.h b/inference-engine/src/mkldnn_plugin/utils/node_dumper.h
index 0580bee4731b91..bac237c8883249 100644
--- a/inference-engine/src/mkldnn_plugin/utils/node_dumper.h
+++ b/inference-engine/src/mkldnn_plugin/utils/node_dumper.h
@@ -31,28 +31,29 @@ class NodeDumper {
 private:
     void dumpInternalBlobs(const MKLDNNNodePtr& node) const;
     void dump(const BlobDumper& bd, const std::string& file) const;
-    bool shouldBeDumped(const MKLDNNNodePtr &node) const;
+    bool shouldBeDumped(const MKLDNNNodePtr &node, const std::string& portsKind) const;
 
-    enum class DUMP_FORMAT {
+    enum class FORMAT {
         BIN,
         TEXT,
     };
 
-    DUMP_FORMAT parseDumpFormat(const std::string& format) const;
+    FORMAT parseDumpFormat(const std::string& format) const;
     void formatNodeName(std::string& name) const;
 
-    DUMP_FORMAT dumpFormat;
+    FORMAT dumpFormat;
     std::string dumpDirName;
     int count;
 
     enum FILTER {
+        BY_PORTS,
         BY_EXEC_ID,
         BY_TYPE,
         BY_NAME,
-        COUNT,
     };
 
-    std::unordered_map<FILTER, std::string> dumpFilters;
+    // std::hash<int> is necessary for Ubuntu-16.04 (gcc-5.4 and defect in C++11 standart)
+    std::unordered_map<FILTER, std::string, std::hash<int>> dumpFilters;
 };
 } // namespace MKLDNNPlugin
 #endif // CPU_DEBUG_CAPS

From 1fadfdafe74a268a0ee287eaf83eef6ca78e8481 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Tomasz=20Do=C5=82bniak?= <tomasz.dolbniak@intel.com>
Date: Fri, 18 Jun 2021 04:41:26 +0200
Subject: [PATCH 126/178] Deprecation of Node::get_default_value and some
 cleanup (#6203)

* Deprecated the get_default_value method of Node class

* Cleanup of the obsolete get_version methods

* Deprecation warnings suppression

* Deprecation of get_default_value for v0 ops
---
 ngraph/core/include/ngraph/node.hpp                  | 2 ++
 ngraph/core/include/ngraph/op/add.hpp                | 1 -
 ngraph/core/include/ngraph/op/avg_pool.hpp           | 3 ++-
 ngraph/core/include/ngraph/op/binary_convolution.hpp | 1 -
 ngraph/core/include/ngraph/op/convolution.hpp        | 2 ++
 ngraph/core/include/ngraph/op/cum_sum.hpp            | 2 ++
 ngraph/core/include/ngraph/op/divide.hpp             | 1 -
 ngraph/core/include/ngraph/op/group_conv.hpp         | 2 ++
 ngraph/core/include/ngraph/op/log_softmax.hpp        | 1 -
 ngraph/core/include/ngraph/op/max_pool.hpp           | 3 ++-
 ngraph/core/include/ngraph/op/pad.hpp                | 1 -
 ngraph/core/include/ngraph/op/reduce_l1.hpp          | 3 ++-
 ngraph/core/include/ngraph/op/reduce_l2.hpp          | 3 ++-
 ngraph/core/include/ngraph/op/reduce_mean.hpp        | 1 -
 ngraph/core/include/ngraph/op/reduce_prod.hpp        | 3 ++-
 ngraph/core/include/ngraph/op/reduce_sum.hpp         | 4 ++--
 ngraph/core/include/ngraph/op/reshape.hpp            | 1 -
 ngraph/core/include/ngraph/op/reverse.hpp            | 5 ++---
 ngraph/core/include/ngraph/op/softmax.hpp            | 5 ++---
 ngraph/core/include/ngraph/op/strided_slice.hpp      | 1 -
 ngraph/core/include/ngraph/op/topk.hpp               | 5 ++---
 ngraph/core/src/op/reverse.cpp                       | 2 +-
 ngraph/core/src/op/softmax.cpp                       | 2 +-
 ngraph/core/src/op/topk.cpp                          | 2 +-
 ngraph/test/runtime/op/avg_pool.hpp                  | 2 ++
 ngraph/test/runtime/op/convolution.hpp               | 2 ++
 26 files changed, 33 insertions(+), 27 deletions(-)

diff --git a/ngraph/core/include/ngraph/node.hpp b/ngraph/core/include/ngraph/node.hpp
index 51a0cccfa25ba4..3d4880298b9d10 100644
--- a/ngraph/core/include/ngraph/node.hpp
+++ b/ngraph/core/include/ngraph/node.hpp
@@ -434,6 +434,8 @@ namespace ngraph
 
         /// \return Version of this node
         virtual size_t get_version() const { return get_type_info().version; }
+
+        NGRAPH_DEPRECATED("This method is deprecated and will be removed soon.")
         virtual std::shared_ptr<Node> get_default_value() const { return nullptr; }
         /// Use instance ids for comparison instead of memory addresses to improve determinism
         bool operator<(const Node& other) const { return m_instance_id < other.m_instance_id; }
diff --git a/ngraph/core/include/ngraph/op/add.hpp b/ngraph/core/include/ngraph/op/add.hpp
index c16c8baac7b0f8..1cf09af76cf6cf 100644
--- a/ngraph/core/include/ngraph/op/add.hpp
+++ b/ngraph/core/include/ngraph/op/add.hpp
@@ -48,7 +48,6 @@ namespace ngraph
 
                 bool visit_attributes(AttributeVisitor& visitor) override;
 
-                size_t get_version() const override { return 1; }
                 bool evaluate(const HostTensorVector& outputs,
                               const HostTensorVector& inputs) const override;
                 bool has_evaluate() const override;
diff --git a/ngraph/core/include/ngraph/op/avg_pool.hpp b/ngraph/core/include/ngraph/op/avg_pool.hpp
index 8f414a6274d6c0..a46250885f33af 100644
--- a/ngraph/core/include/ngraph/op/avg_pool.hpp
+++ b/ngraph/core/include/ngraph/op/avg_pool.hpp
@@ -49,7 +49,6 @@ namespace ngraph
                         op::RoundingType rounding_type = op::RoundingType::FLOOR,
                         const PadType& auto_pad = op::PadType::EXPLICIT);
 
-                size_t get_version() const override { return 1; }
                 void validate_and_infer_types() override;
                 bool visit_attributes(AttributeVisitor& visitor) override;
 
@@ -76,7 +75,9 @@ namespace ngraph
                 op::RoundingType get_rounding_type() const;
                 void set_rounding_type(op::RoundingType rounding_type);
                 /// \return The default value for AvgPool.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
 
             protected:
                 Shape m_kernel;
diff --git a/ngraph/core/include/ngraph/op/binary_convolution.hpp b/ngraph/core/include/ngraph/op/binary_convolution.hpp
index 12e2c448f54036..3d4bf98496e620 100644
--- a/ngraph/core/include/ngraph/op/binary_convolution.hpp
+++ b/ngraph/core/include/ngraph/op/binary_convolution.hpp
@@ -59,7 +59,6 @@ namespace ngraph
                                   float pad_value,
                                   const PadType& auto_pad = PadType::EXPLICIT);
 
-                size_t get_version() const override { return 1; }
                 void validate_and_infer_types() override;
 
                 bool visit_attributes(AttributeVisitor& visitor) override;
diff --git a/ngraph/core/include/ngraph/op/convolution.hpp b/ngraph/core/include/ngraph/op/convolution.hpp
index 72a365be533705..31f82c3c390fe9 100644
--- a/ngraph/core/include/ngraph/op/convolution.hpp
+++ b/ngraph/core/include/ngraph/op/convolution.hpp
@@ -72,7 +72,9 @@ namespace ngraph
                 const PadType& get_auto_pad() const { return m_auto_pad; }
                 void set_auto_pad(const PadType& auto_pad) { m_auto_pad = auto_pad; }
                 /// \return The default value for Convolution.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
 
             protected:
                 Strides m_strides;
diff --git a/ngraph/core/include/ngraph/op/cum_sum.hpp b/ngraph/core/include/ngraph/op/cum_sum.hpp
index 8a133a2502b26c..72cfe89225642d 100644
--- a/ngraph/core/include/ngraph/op/cum_sum.hpp
+++ b/ngraph/core/include/ngraph/op/cum_sum.hpp
@@ -91,7 +91,9 @@ namespace ngraph
                 void validate_and_infer_types() override;
 
                 /// \return The default value for CumSum.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
                 bool is_exclusive() const { return m_exclusive; }
                 bool is_reverse() const { return m_reverse; }
 
diff --git a/ngraph/core/include/ngraph/op/divide.hpp b/ngraph/core/include/ngraph/op/divide.hpp
index 0c33fe6774ee98..f2e40dc55320b1 100644
--- a/ngraph/core/include/ngraph/op/divide.hpp
+++ b/ngraph/core/include/ngraph/op/divide.hpp
@@ -50,7 +50,6 @@ namespace ngraph
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
 
-                size_t get_version() const override { return 1; }
                 bool evaluate(const HostTensorVector& outputs,
                               const HostTensorVector& inputs) const override;
                 bool has_evaluate() const override;
diff --git a/ngraph/core/include/ngraph/op/group_conv.hpp b/ngraph/core/include/ngraph/op/group_conv.hpp
index ebda0392d96571..e55e3fd3457c0b 100644
--- a/ngraph/core/include/ngraph/op/group_conv.hpp
+++ b/ngraph/core/include/ngraph/op/group_conv.hpp
@@ -71,7 +71,9 @@ namespace ngraph
                 const PadType& get_auto_pad() const { return m_auto_pad; }
                 void set_auto_pad(const PadType& auto_pad) { m_auto_pad = auto_pad; }
                 /// \return The default value for Convolution.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
 
             protected:
                 Strides m_strides;
diff --git a/ngraph/core/include/ngraph/op/log_softmax.hpp b/ngraph/core/include/ngraph/op/log_softmax.hpp
index 8bc4cec8e06694..f43bb82c5dff21 100644
--- a/ngraph/core/include/ngraph/op/log_softmax.hpp
+++ b/ngraph/core/include/ngraph/op/log_softmax.hpp
@@ -30,7 +30,6 @@ namespace ngraph
                 bool visit_attributes(AttributeVisitor& visitor) override;
                 void validate_and_infer_types() override;
 
-                size_t get_version() const override { return 1; }
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
 
diff --git a/ngraph/core/include/ngraph/op/max_pool.hpp b/ngraph/core/include/ngraph/op/max_pool.hpp
index 9ca644cae75513..fcfbf69132d84d 100644
--- a/ngraph/core/include/ngraph/op/max_pool.hpp
+++ b/ngraph/core/include/ngraph/op/max_pool.hpp
@@ -41,7 +41,6 @@ namespace ngraph
                         const PadType& auto_pad = op::PadType::EXPLICIT);
 
                 bool visit_attributes(AttributeVisitor& visitor) override;
-                size_t get_version() const override { return 1; }
                 void validate_and_infer_types() override;
 
                 virtual std::shared_ptr<Node>
@@ -69,7 +68,9 @@ namespace ngraph
                     m_rounding_type = rounding_mode;
                 }
                 /// \return The default value for MaxPool.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
 
                 bool evaluate(const HostTensorVector& outputs,
                               const HostTensorVector& inputs) const override;
diff --git a/ngraph/core/include/ngraph/op/pad.hpp b/ngraph/core/include/ngraph/op/pad.hpp
index 93ef2740b60bac..9f0b94cf16c91b 100644
--- a/ngraph/core/include/ngraph/op/pad.hpp
+++ b/ngraph/core/include/ngraph/op/pad.hpp
@@ -57,7 +57,6 @@ namespace ngraph
                 Pad() = default;
 
                 bool visit_attributes(AttributeVisitor& visitor) override;
-                size_t get_version() const override { return 1; }
                 void validate_and_infer_types() override;
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
diff --git a/ngraph/core/include/ngraph/op/reduce_l1.hpp b/ngraph/core/include/ngraph/op/reduce_l1.hpp
index 5760f855727fe3..5dd82a2b1f834d 100644
--- a/ngraph/core/include/ngraph/op/reduce_l1.hpp
+++ b/ngraph/core/include/ngraph/op/reduce_l1.hpp
@@ -31,9 +31,10 @@ namespace ngraph
                          const Output<Node>& reduction_axes,
                          bool keep_dims = false);
 
-                size_t get_version() const override { return 4; }
                 /// \return The default value for Reduce.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
 
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
diff --git a/ngraph/core/include/ngraph/op/reduce_l2.hpp b/ngraph/core/include/ngraph/op/reduce_l2.hpp
index 4cd629f22aaaf4..35c8419c3dd5c5 100644
--- a/ngraph/core/include/ngraph/op/reduce_l2.hpp
+++ b/ngraph/core/include/ngraph/op/reduce_l2.hpp
@@ -30,9 +30,10 @@ namespace ngraph
                          const Output<Node>& reduction_axes,
                          bool keep_dims = false);
 
-                size_t get_version() const override { return 4; }
                 /// \return The default value for Reduce.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
 
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
diff --git a/ngraph/core/include/ngraph/op/reduce_mean.hpp b/ngraph/core/include/ngraph/op/reduce_mean.hpp
index fb4c393b3d5252..dd64240981032b 100644
--- a/ngraph/core/include/ngraph/op/reduce_mean.hpp
+++ b/ngraph/core/include/ngraph/op/reduce_mean.hpp
@@ -26,7 +26,6 @@ namespace ngraph
                            const Output<Node>& reduction_axes,
                            bool keep_dims = false);
 
-                size_t get_version() const override { return 1; }
                 std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
 
diff --git a/ngraph/core/include/ngraph/op/reduce_prod.hpp b/ngraph/core/include/ngraph/op/reduce_prod.hpp
index 12ded1656d1aaf..2b1b73d3927cb2 100644
--- a/ngraph/core/include/ngraph/op/reduce_prod.hpp
+++ b/ngraph/core/include/ngraph/op/reduce_prod.hpp
@@ -30,9 +30,10 @@ namespace ngraph
                            const Output<Node>& reduction_axes,
                            bool keep_dims = false);
 
-                size_t get_version() const override { return 1; }
                 /// \return The default value for Product.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
 
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
diff --git a/ngraph/core/include/ngraph/op/reduce_sum.hpp b/ngraph/core/include/ngraph/op/reduce_sum.hpp
index 5bef40536b1c43..83b663c10e6a12 100644
--- a/ngraph/core/include/ngraph/op/reduce_sum.hpp
+++ b/ngraph/core/include/ngraph/op/reduce_sum.hpp
@@ -77,13 +77,13 @@ namespace ngraph
                           const Output<Node>& reduction_axes,
                           bool keep_dims = false);
 
-                size_t get_version() const override { return 1; }
-
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
 
                 /// \return The default value for Sum.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
 
                 bool evaluate(const HostTensorVector& outputs,
                               const HostTensorVector& inputs) const override;
diff --git a/ngraph/core/include/ngraph/op/reshape.hpp b/ngraph/core/include/ngraph/op/reshape.hpp
index 6be77386cd014c..da2725247c0df4 100644
--- a/ngraph/core/include/ngraph/op/reshape.hpp
+++ b/ngraph/core/include/ngraph/op/reshape.hpp
@@ -46,7 +46,6 @@ namespace ngraph
                 bool visit_attributes(AttributeVisitor& visitor) override;
                 void validate_and_infer_types() override;
 
-                size_t get_version() const override { return 1; }
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
 
diff --git a/ngraph/core/include/ngraph/op/reverse.hpp b/ngraph/core/include/ngraph/op/reverse.hpp
index 2a71ec75df2b13..012d02879df1d7 100644
--- a/ngraph/core/include/ngraph/op/reverse.hpp
+++ b/ngraph/core/include/ngraph/op/reverse.hpp
@@ -15,14 +15,14 @@ namespace ngraph
             class NGRAPH_API Reverse : public Op
             {
             public:
+                NGRAPH_RTTI_DECLARATION;
+
                 enum class Mode
                 {
                     INDEX,
                     MASK
                 };
 
-                static constexpr NodeTypeInfo type_info{"Reverse", 1};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
                 Reverse() = default;
                 /// \brief Constructs a reverse operation.
                 ///
@@ -47,7 +47,6 @@ namespace ngraph
                 /// \return The second input data interpretation mode.
                 Mode get_mode() const { return m_mode; }
                 void set_mode(const Mode mode) { m_mode = mode; }
-                virtual size_t get_version() const override { return 1; }
                 bool evaluate(const HostTensorVector& outputs,
                               const HostTensorVector& inputs) const override;
                 bool has_evaluate() const override;
diff --git a/ngraph/core/include/ngraph/op/softmax.hpp b/ngraph/core/include/ngraph/op/softmax.hpp
index 34537618153b70..33201cc843c344 100644
--- a/ngraph/core/include/ngraph/op/softmax.hpp
+++ b/ngraph/core/include/ngraph/op/softmax.hpp
@@ -15,8 +15,8 @@ namespace ngraph
             class NGRAPH_API Softmax : public Op
             {
             public:
-                static constexpr NodeTypeInfo type_info{"Softmax", 1};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
+                NGRAPH_RTTI_DECLARATION;
+
                 Softmax()
                     : m_axis(0)
                 {
@@ -34,7 +34,6 @@ namespace ngraph
                 bool visit_attributes(AttributeVisitor& visitor) override;
                 void validate_and_infer_types() override;
 
-                size_t get_version() const override { return 1; }
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
 
diff --git a/ngraph/core/include/ngraph/op/strided_slice.hpp b/ngraph/core/include/ngraph/op/strided_slice.hpp
index 33144706b74b28..10e7e00feff779 100644
--- a/ngraph/core/include/ngraph/op/strided_slice.hpp
+++ b/ngraph/core/include/ngraph/op/strided_slice.hpp
@@ -90,7 +90,6 @@ namespace ngraph
                 std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
                 void validate_and_infer_types() override;
-                size_t get_version() const override { return 1; }
                 bool evaluate(const HostTensorVector& output_values,
                               const HostTensorVector& input_values) const override;
                 bool has_evaluate() const override;
diff --git a/ngraph/core/include/ngraph/op/topk.hpp b/ngraph/core/include/ngraph/op/topk.hpp
index 6d03190b6f63ab..e5ad952c46aff3 100644
--- a/ngraph/core/include/ngraph/op/topk.hpp
+++ b/ngraph/core/include/ngraph/op/topk.hpp
@@ -21,11 +21,11 @@ namespace ngraph
             class NGRAPH_API TopK : public Op
             {
             public:
+                NGRAPH_RTTI_DECLARATION;
+
                 using SortType = TopKSortType;
                 using Mode = TopKMode;
 
-                static constexpr NodeTypeInfo type_info{"TopK", 1};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
                 /// \brief Constructs a TopK operation
                 TopK() = default;
                 /// \brief Constructs a TopK operation with two outputs: values and indices.
@@ -60,7 +60,6 @@ namespace ngraph
                 virtual std::shared_ptr<Node>
                     clone_with_new_inputs(const OutputVector& new_args) const override;
 
-                virtual size_t get_version() const override { return 1; }
                 /// \brief Returns axis value after normalization
                 /// \note If input rank required to normalization is dynamic, the exception is
                 /// thrown
diff --git a/ngraph/core/src/op/reverse.cpp b/ngraph/core/src/op/reverse.cpp
index 68571a94a95100..903e9eea6c2531 100644
--- a/ngraph/core/src/op/reverse.cpp
+++ b/ngraph/core/src/op/reverse.cpp
@@ -18,7 +18,7 @@
 using namespace std;
 using namespace ngraph;
 
-constexpr NodeTypeInfo op::v1::Reverse::type_info;
+NGRAPH_RTTI_DEFINITION(op::v1::Reverse, "Reverse", 1);
 
 op::v1::Reverse::Reverse(const Output<Node>& data,
                          const Output<Node>& reversed_axes,
diff --git a/ngraph/core/src/op/softmax.cpp b/ngraph/core/src/op/softmax.cpp
index ae4317a9c12b11..097d46dd1594d1 100644
--- a/ngraph/core/src/op/softmax.cpp
+++ b/ngraph/core/src/op/softmax.cpp
@@ -51,7 +51,7 @@ namespace
 } // namespace
 
 // *** SOFTMAX OP SET V1 ***
-constexpr NodeTypeInfo op::v1::Softmax::type_info;
+NGRAPH_RTTI_DEFINITION(op::v1::Softmax, "Softmax", 1);
 
 op::v1::Softmax::Softmax(const Output<Node>& arg, const size_t axis)
     : Op({arg})
diff --git a/ngraph/core/src/op/topk.cpp b/ngraph/core/src/op/topk.cpp
index 6b91fd13339e43..b346c5d238f60b 100644
--- a/ngraph/core/src/op/topk.cpp
+++ b/ngraph/core/src/op/topk.cpp
@@ -205,7 +205,7 @@ namespace topk
 } // namespace topk
 
 // v1 version starts
-constexpr NodeTypeInfo op::v1::TopK::type_info;
+NGRAPH_RTTI_DEFINITION(op::v1::TopK, "TopK", 1);
 
 static const std::uint64_t UNKNOWN_NORMALIZED_AXIS = std::numeric_limits<uint64_t>::max();
 
diff --git a/ngraph/test/runtime/op/avg_pool.hpp b/ngraph/test/runtime/op/avg_pool.hpp
index 1fab6fc25de2ec..701b3e174f3f46 100644
--- a/ngraph/test/runtime/op/avg_pool.hpp
+++ b/ngraph/test/runtime/op/avg_pool.hpp
@@ -146,7 +146,9 @@ namespace ngraph
                 bool get_ceil_mode() const;
                 void set_ceil_mode(bool ceil_mode);
                 /// \return The default value for AvgPool.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
 
             protected:
                 Shape m_window_shape;
diff --git a/ngraph/test/runtime/op/convolution.hpp b/ngraph/test/runtime/op/convolution.hpp
index 5ddea7b0f7de6f..f5d45d269183af 100644
--- a/ngraph/test/runtime/op/convolution.hpp
+++ b/ngraph/test/runtime/op/convolution.hpp
@@ -178,7 +178,9 @@ namespace ngraph
                 const PadType& get_pad_type() const { return m_pad_type; }
                 void set_pad_type(const PadType& pad_type) { m_pad_type = pad_type; }
                 /// \return The default value for Convolution.
+                NGRAPH_SUPPRESS_DEPRECATED_START
                 virtual std::shared_ptr<Node> get_default_value() const override;
+                NGRAPH_SUPPRESS_DEPRECATED_END
 
             protected:
                 Strides m_window_movement_strides;

From 94352874a56793a5f379fe6cbb90a686377ed361 Mon Sep 17 00:00:00 2001
From: Piotr Szmelczynski <piotr.szmelczynski@intel.com>
Date: Fri, 18 Jun 2021 04:50:02 +0200
Subject: [PATCH 127/178] Revise ceiling (#6124)

* update spec

* add RTTI macro

* clean backend test file

* create visitor test

* remove Ceiling cpu functional tests from skip_test_config

* fix skip_test_config conflict

* Add type_prop test for Ceiling.

* Fix failing ceiling type_prop tests.

* Replace unary_ops.cpp with single test op files.

* Enable integer tests.

Co-authored-by: Szymon Durawa <szymon.durawa@intel.com>
---
 docs/ops/arithmetic/Ceiling_1.md              | 20 +++++++-------
 .../single_layer_tests/activation.cpp         |  1 +
 .../skip_tests_config.cpp                     |  2 --
 .../layer_tests_summary/utils/constants.py    |  1 +
 ngraph/core/include/ngraph/op/ceiling.hpp     |  3 +--
 ngraph/core/src/op/ceiling.cpp                |  2 +-
 ngraph/test/CMakeLists.txt                    | 10 ++++++-
 ngraph/test/backend/ceiling.in.cpp            |  7 -----
 ngraph/test/runtime/ie/unit_test.manifest     |  3 ---
 ngraph/test/type_prop/abs.cpp                 |  9 +++++++
 ngraph/test/type_prop/acos.cpp                |  9 +++++++
 ngraph/test/type_prop/asin.cpp                |  9 +++++++
 ngraph/test/type_prop/ceiling.cpp             |  9 +++++++
 ngraph/test/type_prop/exp.cpp                 |  9 +++++++
 ngraph/test/type_prop/floor.cpp               |  9 +++++++
 ngraph/test/type_prop/sin.cpp                 |  9 +++++++
 ngraph/test/type_prop/sqrt.cpp                |  9 +++++++
 .../{unary_ops.cpp => unary_ops.hpp}          |  6 +----
 ngraph/test/visitors/op/ceiling.cpp           | 26 +++++++++++++++++++
 19 files changed, 122 insertions(+), 31 deletions(-)
 create mode 100644 ngraph/test/type_prop/abs.cpp
 create mode 100644 ngraph/test/type_prop/acos.cpp
 create mode 100644 ngraph/test/type_prop/asin.cpp
 create mode 100644 ngraph/test/type_prop/ceiling.cpp
 create mode 100644 ngraph/test/type_prop/exp.cpp
 create mode 100644 ngraph/test/type_prop/floor.cpp
 create mode 100644 ngraph/test/type_prop/sin.cpp
 create mode 100644 ngraph/test/type_prop/sqrt.cpp
 rename ngraph/test/type_prop/{unary_ops.cpp => unary_ops.hpp} (94%)
 create mode 100644 ngraph/test/visitors/op/ceiling.cpp

diff --git a/docs/ops/arithmetic/Ceiling_1.md b/docs/ops/arithmetic/Ceiling_1.md
index 588b5ff6842f55..4d4cfeb945002e 100644
--- a/docs/ops/arithmetic/Ceiling_1.md
+++ b/docs/ops/arithmetic/Ceiling_1.md
@@ -2,31 +2,31 @@
 
 **Versioned name**: *Ceiling-1*
 
-**Category**: Arithmetic unary operation 
+**Category**: Arithmetic unary operation
 
 **Short description**: *Ceiling* performs element-wise ceiling operation with given tensor.
 
-**Attributes**:
+**Detailed description**: For each element from the input tensor calculates corresponding
+element in the output tensor with the following formula:
 
-    No attributes available.
+\f[
+a_{i} = ceiling(a_{i})
+\f]
+
+**Attributes**: *Ceiling* operation has no attributes.
 
 **Inputs**
 
-* **1**: An tensor of type T. **Required.**
+* **1**: A tensor of type *T* and arbitrary shape. **Required.**
 
 **Outputs**
 
-* **1**: The result of element-wise ceiling operation. A tensor of type T.
+* **1**: The result of element-wise ceiling operation. A tensor of type *T*.
 
 **Types**
 
 * *T*: any numeric type.
 
-*Ceiling* does the following with the input tensor *a*:
-
-\f[
-a_{i} = ceiling(a_{i})
-\f]
 
 **Examples**
 
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
index 6e762ed562c004..510e4039e2bfd0 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
@@ -64,6 +64,7 @@ const std::map<ActivationTypes, std::vector<std::vector<float>>> activationTypes
 
 // List of operations that should be tested also with integer precision
 const std::map<ActivationTypes, std::vector<std::vector<float>>> intActivationTypes = {
+        {Ceiling,               {}},
         {Sqrt,                  {}},
         {Tanh,                  {}},
 };
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
index 3d4678bf5892b4..aa4c6dafb02c4b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/skip_tests_config.cpp
@@ -33,8 +33,6 @@ std::vector<std::string> disabledTestPatterns() {
         // TODO: Issue: 34055
         R"(.*ShapeOfLayerTest.*)",
         R"(.*ReluShapeOfSubgraphTest.*)",
-        // TODO: Issue: 34805
-        R"(.*ActivationLayerTest.*Ceiling.*)",
         // TODO: Issue: 43314
         R"(.*Broadcast.*mode=BIDIRECTIONAL.*inNPrec=BOOL.*)",
         // TODO: Issue 43417 sporadic issue, looks like an issue in test, reproducible only on Windows platform
diff --git a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
index cdbcdcf6c54ae8..7db26eb289542f 100644
--- a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
+++ b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
@@ -13,6 +13,7 @@
     'Broadcast-1',
     'Broadcast-3',
     'Bucketize-3',
+    'Ceiling-1',
     'CTCGreedyDecoder-1',
     'CTCGreedyDecoderSeqLen-6',
     'Concat-1',
diff --git a/ngraph/core/include/ngraph/op/ceiling.hpp b/ngraph/core/include/ngraph/op/ceiling.hpp
index 491ca099b2ac47..7ac74ad3dd407c 100644
--- a/ngraph/core/include/ngraph/op/ceiling.hpp
+++ b/ngraph/core/include/ngraph/op/ceiling.hpp
@@ -16,8 +16,7 @@ namespace ngraph
             class NGRAPH_API Ceiling : public util::UnaryElementwiseArithmetic
             {
             public:
-                static constexpr NodeTypeInfo type_info{"Ceiling", 0};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
+                NGRAPH_RTTI_DECLARATION;
                 /// \brief Constructs a ceiling operation.
                 Ceiling() = default;
                 /// \brief Constructs a ceiling operation.
diff --git a/ngraph/core/src/op/ceiling.cpp b/ngraph/core/src/op/ceiling.cpp
index 7ef0e8a6bece10..22abc45eba3ac2 100644
--- a/ngraph/core/src/op/ceiling.cpp
+++ b/ngraph/core/src/op/ceiling.cpp
@@ -12,7 +12,7 @@
 using namespace std;
 using namespace ngraph;
 
-constexpr NodeTypeInfo op::Ceiling::type_info;
+NGRAPH_RTTI_DEFINITION(op::v0::Ceiling, "Ceiling", 0, util::UnaryElementwiseArithmetic);
 
 op::Ceiling::Ceiling(const Output<Node>& arg)
     : UnaryElementwiseArithmetic(arg)
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index efeb68e670ec94..fcb4f013775139 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -99,6 +99,9 @@ set(SRC
     span.cpp
     specialize_function.cpp
     tensor.cpp
+    type_prop/abs.cpp
+    type_prop/acos.cpp
+    type_prop/asin.cpp
     type_prop/assign.cpp
     type_prop/avg_pool.cpp
     type_prop/batch_norm.cpp
@@ -107,6 +110,7 @@ set(SRC
     type_prop/binary_elementwise.cpp
     type_prop/broadcast.cpp
     type_prop/bucketize.cpp
+    type_prop/ceiling.cpp
     type_prop/clamp.cpp
     type_prop/concat.cpp
     type_prop/constant.cpp
@@ -123,6 +127,7 @@ set(SRC
     type_prop/dft.cpp
     type_prop/dyn_reshape.cpp
     type_prop/einsum.cpp
+    type_prop/exp.cpp
     type_prop/experimental_detectron_generate_proposals.cpp
     type_prop/experimental_detectron_roi_feature_extractor.cpp
     type_prop/experimental_detectron_topkrois.cpp
@@ -135,6 +140,7 @@ set(SRC
     type_prop/embeddingbag_packedsum.cpp
     type_prop/embedding_segments_sum.cpp
     type_prop/fake_quantize.cpp
+    type_prop/floor.cpp
     type_prop/floor_mod.cpp
     type_prop/gather.cpp
     type_prop/gather_elements.cpp
@@ -201,11 +207,13 @@ set(SRC
     type_prop/selu.cpp
     type_prop/shape_of.cpp
     type_prop/shuffle_channels.cpp
+    type_prop/sin.cpp
     type_prop/softmax.cpp
     type_prop/softplus.cpp
     type_prop/space_to_batch.cpp
     type_prop/space_to_depth.cpp
     type_prop/split.cpp
+    type_prop/sqrt.cpp
     type_prop/squared_difference.cpp
     type_prop/squeeze.cpp
     type_prop/swish.cpp
@@ -214,7 +222,6 @@ set(SRC
     type_prop/top_k.cpp
     type_prop/transpose.cpp
     type_prop/unary_elementwise.cpp
-    type_prop/unary_ops.cpp
     type_prop/unsqueeze.cpp
     type_prop/variadic_split.cpp
     type_prop_layers.cpp
@@ -224,6 +231,7 @@ set(SRC
     visitors/op/batch_norm.cpp
     visitors/op/broadcast.cpp
     visitors/op/bucketize.cpp
+    visitors/op/ceiling.cpp
     visitors/op/constant.cpp
     visitors/op/convert.cpp
     visitors/op/convolution_backprop.cpp
diff --git a/ngraph/test/backend/ceiling.in.cpp b/ngraph/test/backend/ceiling.in.cpp
index cfd5d71e3e1315..2cf7b5489491b0 100644
--- a/ngraph/test/backend/ceiling.in.cpp
+++ b/ngraph/test/backend/ceiling.in.cpp
@@ -2,13 +2,6 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <algorithm>
-#include <cinttypes>
-#include <cmath>
-#include <cstdlib>
-#include <random>
-#include <string>
-
 // clang-format off
 #ifdef ${BACKEND_NAME}_FLOAT_TOLERANCE_BITS
 #define DEFAULT_FLOAT_TOLERANCE_BITS ${BACKEND_NAME}_FLOAT_TOLERANCE_BITS
diff --git a/ngraph/test/runtime/ie/unit_test.manifest b/ngraph/test/runtime/ie/unit_test.manifest
index aa4079eee82c89..03d0185ad8f310 100644
--- a/ngraph/test/runtime/ie/unit_test.manifest
+++ b/ngraph/test/runtime/ie/unit_test.manifest
@@ -723,9 +723,6 @@ cum_sum_2dim
 cum_sum_3d
 cum_sum_2dim_allmodes
 
-# Unsupported primitive of type: Ceiling
-ceiling
-
 # Incorrect dimensions for broadcasting for Add
 auto_bcast_binary_elementwise_pdpd_dynamic
 
diff --git a/ngraph/test/type_prop/abs.cpp b/ngraph/test/type_prop/abs.cpp
new file mode 100644
index 00000000000000..7ef727bdee09f5
--- /dev/null
+++ b/ngraph/test/type_prop/abs.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Type = ::testing::Types<ngraph::op::Abs>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_abs, UnaryOperator, Type);
diff --git a/ngraph/test/type_prop/acos.cpp b/ngraph/test/type_prop/acos.cpp
new file mode 100644
index 00000000000000..8bff6e06dd968a
--- /dev/null
+++ b/ngraph/test/type_prop/acos.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Type = ::testing::Types<ngraph::op::Acos>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_acos, UnaryOperator, Type);
diff --git a/ngraph/test/type_prop/asin.cpp b/ngraph/test/type_prop/asin.cpp
new file mode 100644
index 00000000000000..96293580d57dc4
--- /dev/null
+++ b/ngraph/test/type_prop/asin.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Type = ::testing::Types<ngraph::op::Asin>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_asin, UnaryOperator, Type);
diff --git a/ngraph/test/type_prop/ceiling.cpp b/ngraph/test/type_prop/ceiling.cpp
new file mode 100644
index 00000000000000..f173a031491074
--- /dev/null
+++ b/ngraph/test/type_prop/ceiling.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Type = ::testing::Types<ngraph::op::Ceiling>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_ceiling, UnaryOperator, Type);
diff --git a/ngraph/test/type_prop/exp.cpp b/ngraph/test/type_prop/exp.cpp
new file mode 100644
index 00000000000000..f25c335504ef7a
--- /dev/null
+++ b/ngraph/test/type_prop/exp.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Type = ::testing::Types<ngraph::op::Exp>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_exp, UnaryOperator, Type);
diff --git a/ngraph/test/type_prop/floor.cpp b/ngraph/test/type_prop/floor.cpp
new file mode 100644
index 00000000000000..3eca60835f0b1a
--- /dev/null
+++ b/ngraph/test/type_prop/floor.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Type = ::testing::Types<ngraph::op::Floor>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_floor, UnaryOperator, Type);
diff --git a/ngraph/test/type_prop/sin.cpp b/ngraph/test/type_prop/sin.cpp
new file mode 100644
index 00000000000000..ab777d9a4429bd
--- /dev/null
+++ b/ngraph/test/type_prop/sin.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Type = ::testing::Types<ngraph::op::Sin>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_sin, UnaryOperator, Type);
diff --git a/ngraph/test/type_prop/sqrt.cpp b/ngraph/test/type_prop/sqrt.cpp
new file mode 100644
index 00000000000000..48a5ce13a85300
--- /dev/null
+++ b/ngraph/test/type_prop/sqrt.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Type = ::testing::Types<ngraph::op::Sqrt>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_sqrt, UnaryOperator, Type);
diff --git a/ngraph/test/type_prop/unary_ops.cpp b/ngraph/test/type_prop/unary_ops.hpp
similarity index 94%
rename from ngraph/test/type_prop/unary_ops.cpp
rename to ngraph/test/type_prop/unary_ops.hpp
index 788b50917fa6a2..e8babcba4067d5 100644
--- a/ngraph/test/type_prop/unary_ops.cpp
+++ b/ngraph/test/type_prop/unary_ops.hpp
@@ -1,4 +1,4 @@
-// Copyright (C) 2018-2021 Intel Corporation
+// Copyright (C) 2021 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
 
@@ -95,7 +95,3 @@ REGISTER_TYPED_TEST_CASE_P(UnaryOperator,
                            dynamic_rank_input_shape_2D,
                            dynamic_rank_input_shape_3D,
                            dynamic_rank_input_shape_full);
-
-using Types = ::testing::Types<op::Acos, op::Asin, op::Abs, op::Sqrt, op::Sin, op::Exp, op::Floor, op::Log>;
-
-INSTANTIATE_TYPED_TEST_CASE_P(type_prop, UnaryOperator, Types);
diff --git a/ngraph/test/visitors/op/ceiling.cpp b/ngraph/test/visitors/op/ceiling.cpp
new file mode 100644
index 00000000000000..34ed3b653b36ef
--- /dev/null
+++ b/ngraph/test/visitors/op/ceiling.cpp
@@ -0,0 +1,26 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+
+#include "ngraph/ngraph.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+#include "ngraph/opsets/opset1.hpp"
+#include "util/visitor.hpp"
+
+using namespace std;
+using namespace ngraph;
+using ngraph::test::NodeBuilder;
+
+TEST(attributes, ceiling_op)
+{
+    NodeBuilder::get_ops().register_factory<opset1::Ceiling>();
+    const auto A = make_shared<op::Parameter>(element::f32, Shape{5, 2});
+
+    const auto ceiling = make_shared<opset1::Ceiling>(A);
+    NodeBuilder builder(ceiling);
+
+    const auto expected_attr_count = 0;
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
+}

From 73ef2ffb415bea856472ddec61679f3af75ee6a2 Mon Sep 17 00:00:00 2001
From: cecilia peng <cecilia.peng@intel.com>
Date: Fri, 18 Jun 2021 14:20:00 +0800
Subject: [PATCH 128/178] MulticlassNms-8 spec. (#5907)

* MulticlassNms-8 spec.

This Op functionally extends NonMaxSuppression-5, to perform more post-processing phases, and lay out
the detection outputs in the way of PaddlePaddle detection.

* Update MulticlassNMS_8.md

Clarify the meaning of "nms_top_k".

* Update MulticlassNMS_8.md

* Update MulticlassNMS_8.md

* Update MulticlassNMS_8.md

* Update MulticlassNMS_8.md

* Update MulticlassNMS_8.md
---
 docs/doxygen/ie_docs.xml         |   1 +
 docs/ops/sort/MulticlassNMS_8.md | 161 +++++++++++++++++++++++++++++++
 2 files changed, 162 insertions(+)
 create mode 100644 docs/ops/sort/MulticlassNMS_8.md

diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index 503f9f3898601b..c12cc4d22be918 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -194,6 +194,7 @@ limitations under the License.
                 <tab type="user" title="Minimum-1" url="@ref openvino_docs_ops_arithmetic_Minimum_1"/>
                 <tab type="user" title="Mish-4" url="@ref openvino_docs_ops_activation_Mish_4"/>
                 <tab type="user" title="Mod-1" url="@ref openvino_docs_ops_arithmetic_Mod_1"/>
+                <tab type="user" title="MulticlassNonMaxSuppression-8" url="@ref openvino_docs_ops_sort_MulticlassNonMaxSuppression_8"/>
                 <tab type="user" title="Multiply-1" url="@ref openvino_docs_ops_arithmetic_Multiply_1"/>
                 <tab type="user" title="Negative-1" url="@ref openvino_docs_ops_arithmetic_Negative_1"/>
                 <tab type="user" title="NonMaxSuppression-1" url="@ref openvino_docs_ops_sort_NonMaxSuppression_1"/>
diff --git a/docs/ops/sort/MulticlassNMS_8.md b/docs/ops/sort/MulticlassNMS_8.md
new file mode 100644
index 00000000000000..16997a81397ef0
--- /dev/null
+++ b/docs/ops/sort/MulticlassNMS_8.md
@@ -0,0 +1,161 @@
+##  MulticlassNonMaxSuppression<a name="MulticlassNonMaxSuppression"></a> {#openvino_docs_ops_sort_MulticlassNonMaxSuppression_8}
+
+**Versioned name**: *MulticlassNonMaxSuppression-8*
+
+**Category**: *Sorting and maximization*
+
+**Short description**: *MulticlassNonMaxSuppression* performs multi-class non-maximum suppression of the boxes with predicted scores.
+
+**Detailed description**: *MulticlassNonMaxSuppression* is a multi-phase operation. It implements non-maximum suppression algorithm as described below:
+
+1.  Let `B = [b_0,...,b_n]` be the list of initial detection boxes, `S = [s_0,...,s_N]` be  the list of corresponding scores.
+2.  Let `D = []` be an initial collection of resulting boxes. Let `adaptive_threshold = iou_threshold`.
+3.  If `B` is empty, go to step 9.
+4.  Take the box with highest score. Suppose that it is the box `b` with the score `s`.
+5.  Delete `b` from `B`.
+6.  If the score `s` is greater than or equal to `score_threshold`,  add `b` to `D`, else go to step 9.
+7.  If `nms_eta < 1` and `adaptive_threshold > 0.5`, update `adaptive_threshold *= nms_eta`.
+8.  For each input box `b_i` from `B` and the corresponding score `s_i`, set `s_i = 0` when `iou(b, b_i) > adaptive_threshold`, and go to step 3.
+9.  Return `D`, a collection of the corresponding scores `S`, and the number of elements in `D`.
+
+This algorithm is applied independently to each class of each batch element. The operation feeds at most `nms_top_k` scoring candidate boxes to this algorithm.
+The total number of output boxes of each batch element must not exceed `keep_top_k`.
+Boxes of `background_class` are skipped and thus eliminated.
+
+**Attributes**:
+
+* *sort_result*
+
+  * **Description**: *sort_result* specifies the order of output elements.
+  * **Range of values**: `class`, `score`, `none`
+    * *class* - sort selected boxes by class id (ascending).
+    * *score* - sort selected boxes by score (descending).
+    * *none* - do not guarantee the order.
+  * **Type**: `string`
+  * **Default value**: `none`
+  * **Required**: *No*
+
+* *sort_result_across_batch*
+
+  * **Description**: *sort_result_across_batch* is a flag that specifies whenever it is necessary to sort selected boxes across batches or not.
+  * **Range of values**: true or false
+    * *true* - sort selected boxes across batches.
+    * *false* - do not sort selected boxes across batches (boxes are sorted per batch element).
+  * **Type**: boolean
+  * **Default value**: false
+  * **Required**: *No*
+
+* *output_type*
+
+  * **Description**: the tensor type of outputs `selected_indices` and `valid_outputs`.
+  * **Range of values**: `i64` or `i32`
+  * **Type**: `string`
+  * **Default value**: `i64`
+  * **Required**: *No*
+
+* *iou_threshold*
+
+  * **Description**: intersection over union threshold.
+  * **Range of values**: a floating-point number
+  * **Type**: `float`
+  * **Default value**: `0`
+  * **Required**: *No*
+
+* *score_threshold*
+
+  * **Description**: minimum score to consider box for the processing.
+  * **Range of values**: a floating-point number
+  * **Type**: `float`
+  * **Default value**: `0`
+  * **Required**: *No*
+
+* *nms_top_k*
+
+  * **Description**: maximum number of boxes to be selected per class.
+  * **Range of values**: an integer
+  * **Type**: `int`
+  * **Default value**: `-1` meaning to keep all boxes
+  * **Required**: *No*
+
+* *keep_top_k*
+
+  * **Description**: maximum number of boxes to be selected per batch element.
+  * **Range of values**: an integer
+  * **Type**: `int`
+  * **Default value**: `-1` meaning to keep all boxes
+  * **Required**: *No*
+
+* *background_class*
+
+  * **Description**: the background class id.
+  * **Range of values**: an integer
+  * **Type**: `int`
+  * **Default value**: `-1` meaning to keep all classes.
+  * **Required**: *No*
+
+* *nms_eta*
+
+  * **Description**: eta parameter for adaptive NMS.
+  * **Range of values**: a floating-point number in close range `[0, 1.0]`.
+  * **Type**: `float`
+  * **Default value**: `1.0`
+  * **Required**: *No*
+
+**Inputs**:
+
+*   **1**: `boxes` - tensor of type *T* and shape `[num_batches, num_boxes, 4]` with box coordinates. The box coordinates are layout as `[xmin, ymin, xmax, ymax]`. **Required.**
+
+*   **2**: `scores` - tensor of type *T* and shape `[num_batches, num_classes, num_boxes]` with box scores. **Required.**
+
+**Outputs**:
+
+*   **1**: `selected_outputs` - tensor of type *T_THRESHOLDS* and shape `[number of selected boxes, 6]` containing the selected boxes with score and class as tuples `[class_id, box_score, xmin, ymin, xmax, ymax]`.
+
+*   **2**: `selected_indices` - tensor of type *T_IND* and shape `[number of selected boxes, 1]` the selected indices in the flattened `boxes`, which are absolute values cross batches. Therefore possible valid values are in the range `[0, num_batches * num_boxes - 1]`.
+
+*   **3**: `selected_num` - 1D tensor of type *T_IND* and shape `[num_batches]` representing the number of selected boxes for each batch element.
+
+When there is no box selected, `selected_num` is filled with `0`. `selected_outputs` is an empty tensor of shape `[0, 6]`, and `selected_indices` is an empty tensor of shape `[0, 1]`.
+
+**Types**
+
+* *T*: floating point type.
+
+* *T_MAX_BOXES*: integer type.
+
+* *T_THRESHOLDS*: floating point type.
+
+* *T_IND*: `int64` or `int32`.
+
+**Example**
+
+```xml
+<layer ... type="MulticlassNonMaxSuppression" ... >
+    <data sort_result="score" output_type="i64"/>
+    <input>
+        <port id="0">
+            <dim>3</dim>
+            <dim>100</dim>
+            <dim>4</dim>
+        </port>
+        <port id="1">
+            <dim>3</dim>
+            <dim>5</dim>
+            <dim>100</dim>
+        </port>
+    </input>
+    <output>
+        <port id="5" precision="FP32">
+            <dim>-1</dim> <!-- "-1" means a undefined dimension calculated during the model inference -->
+            <dim>6</dim>
+        </port>
+        <port id="6" precision="I64">
+            <dim>-1</dim>
+            <dim>1</dim>         
+        </port>
+        <port id="7" precision="I64">
+            <dim>3</dim>
+        </port>
+    </output>
+</layer>
+```

From 9dd0531599d74fdf281c83179e9e10bcd3ce37ac Mon Sep 17 00:00:00 2001
From: Gabriele Galiero Casay <gabriele.galiero.casay@intel.com>
Date: Fri, 18 Jun 2021 08:20:40 +0200
Subject: [PATCH 129/178] Negative operation specification refactoring (#6204)

* update spec

* update supported types in spec

* Minor changes in spec

Co-authored-by: pszmel <piotr.szmelczynski@intel.com>
---
 docs/ops/arithmetic/Negative_1.md | 32 +++++++++++++++----------------
 1 file changed, 15 insertions(+), 17 deletions(-)

diff --git a/docs/ops/arithmetic/Negative_1.md b/docs/ops/arithmetic/Negative_1.md
index 2e17112e7bcc51..997342c2d05da1 100644
--- a/docs/ops/arithmetic/Negative_1.md
+++ b/docs/ops/arithmetic/Negative_1.md
@@ -2,35 +2,33 @@
 
 **Versioned name**: *Negative-1*
 
-**Category**: Arithmetic unary operation 
+**Category**: Arithmetic unary operation
 
-**Short description**: *Negative* performs element-wise negative operation with given tensor.
+**Short description**: *Negative* performs element-wise negative operation on a given input tensor.
 
-**Attributes**:
+**Detailed description**
 
-    No attributes available.
+*Negative* performs element-wise negative operation on a given input tensor, based on the following mathematical formula:
+
+\f[
+a_{i} = -a_{i}
+\f]
+
+**Attributes**: *Negative* operation has no attributes.
 
 **Inputs**
 
-* **1**: An tensor of type T. **Required.**
+* **1**: A tensor of type *T* and arbitrary shape. **Required.**
 
 **Outputs**
 
-* **1**: The result of element-wise negative operation. A tensor of type T.
+* **1**: The result of element-wise *Negative* operation applied to the input tensor. A tensor of type *T* and the same shape as input tensor.
 
 **Types**
 
-* *T*: any numeric type.
-
-*Negative* does the following with the input tensor *a*:
-
-\f[
-a_{i} = -a_{i}
-\f]
-
-**Examples**
+* *T*: any supported signed numeric type.
 
-*Example 1*
+**Example**
 
 ```xml
 <layer ... type="Negative">
@@ -47,4 +45,4 @@ a_{i} = -a_{i}
         </port>
     </output>
 </layer>
-```
\ No newline at end of file
+```

From 8543b42ac0a50877c4816dcae8ae15933b49c506 Mon Sep 17 00:00:00 2001
From: Ivan Tikhonov <ivan.tikhonov@intel.com>
Date: Fri, 18 Jun 2021 09:50:29 +0300
Subject: [PATCH 130/178] DeformableConvolution v8 specification (#5684)

* ModulatedDeformableConvolution spec

* add bilinear_interpolation_mode attribute

* updated formulas for DeformableConv v1 and v8, applied review remarks

* fix doxygen formulas

* Update Deformable conv v8 spec, update docs
---
 docs/doxygen/ie_docs.xml                      |   2 +-
 .../convolution/DeformableConvolution_1.md    |  20 +++
 .../convolution/DeformableConvolution_8.md    | 168 ++++++++++++++++++
 docs/ops/opset8.md                            |   2 +-
 4 files changed, 190 insertions(+), 2 deletions(-)
 create mode 100644 docs/ops/convolution/DeformableConvolution_8.md

diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index c12cc4d22be918..faabfab9238054 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -131,7 +131,7 @@ limitations under the License.
                 <tab type="user" title="Cosh-1" url="@ref openvino_docs_ops_arithmetic_Cosh_1"/>
                 <tab type="user" title="CTCLoss-4" url="@ref openvino_docs_ops_sequence_CTCLoss_4"/>
                 <tab type="user" title="CumSum" url="@ref openvino_docs_ops_arithmetic_CumSum_3"/>
-                <tab type="user" title="DeformableConvolution-1" url="@ref openvino_docs_ops_convolution_DeformableConvolution_1"/>
+                <tab type="user" title="DeformableConvolution-8" url="@ref openvino_docs_ops_convolution_DeformableConvolution_8"/>
                 <tab type="user" title="DeformablePSROIPooling-1" url="@ref openvino_docs_ops_detection_DeformablePSROIPooling_1"/>
                 <tab type="user" title="DepthToSpace-1" url="@ref openvino_docs_ops_movement_DepthToSpace_1"/>
                 <tab type="user" title="DetectionOutput-1" url="@ref openvino_docs_ops_detection_DetectionOutput_1"/>
diff --git a/docs/ops/convolution/DeformableConvolution_1.md b/docs/ops/convolution/DeformableConvolution_1.md
index 2cba8d84039fe9..612d3c419d4ec2 100644
--- a/docs/ops/convolution/DeformableConvolution_1.md
+++ b/docs/ops/convolution/DeformableConvolution_1.md
@@ -8,6 +8,26 @@
 
 **Detailed description**: *Deformable Convolution* is similar to regular *Convolution* but its receptive field is deformed because of additional spatial offsets used during input sampling. More thorough explanation can be found in [Deformable Convolutions Demystified](https://towardsdatascience.com/deformable-convolutions-demystified-2a77498699e8) and [Deformable Convolutional Networks](https://arxiv.org/abs/1703.06211).
 
+Output is calculated using the following formula: 
+
+  \f[
+
+  y(p) = \sum_{k = 1}^{K}w_{k}x(p + p_{k} + {\Delta}p_{k})
+  
+  \f]
+
+Where 
+* K is a number of sampling locations, e.g. for kernel 3x3 and dilation = 1, K = 9
+
+* \f$x(p)\f$ and \f$y(p)\f$ denote the features at location p from the input feature maps x and output feature maps y
+
+* \f$w_{k}\f$ is the weight for k-th location.
+
+* \f$p_{k}\f$ is pre-specified offset for the k-th location, e.g. K = 9 and
+\f$p_{k} \in \{(-1, -1),(-1, 0), . . . ,(1, 1)\}\f$
+
+* \f${\Delta}p_{k}\f$ is the learnable offset for the k-th location.
+
 **Attributes**:
 
 * *strides*
diff --git a/docs/ops/convolution/DeformableConvolution_8.md b/docs/ops/convolution/DeformableConvolution_8.md
new file mode 100644
index 00000000000000..cf59584a5f4104
--- /dev/null
+++ b/docs/ops/convolution/DeformableConvolution_8.md
@@ -0,0 +1,168 @@
+## DeformableConvolution<a name="DeformableConvolution"></a> {#openvino_docs_ops_convolution_DeformableConvolution_8}
+
+**Versioned name**: *DeformableConvolution-8*
+
+**Category**: Convolution
+
+**Short description**: Computes 2D deformable convolution of input and kernel tensors.
+
+**Detailed description**: *Deformable Convolution* is similar to regular *Convolution* but its receptive field is deformed because of additional spatial offsets used during input sampling. More thorough explanation can be found in [Deformable Convolutions Demystified](https://towardsdatascience.com/deformable-convolutions-demystified-2a77498699e8), [Deformable Convolutional Networks](https://arxiv.org/abs/1703.06211).
+
+Modification of DeformableConvolution using modulating scalars is also supported. Please refer to [Deformable ConvNets v2: More Deformable, Better Results](https://arxiv.org/pdf/1811.11168.pdf).
+
+Output is calculated using the following formula: 
+
+  \f[
+
+  y(p) = \sum_{k = 1}^{K}w_{k}x(p + p_{k} + {\Delta}p_{k}) * {\Delta}m_{k}
+  
+  \f]
+Where 
+* K is a number of sampling locations, e.g. for kernel 3x3 and dilation = 1, K = 9
+
+* \f$x(p)\f$ and \f$y(p)\f$ denote the features at location p from the input feature maps x and output feature maps y
+
+* \f$w_{k}\f$ is the weight for k-th location.
+
+* \f$p_{k}\f$ is pre-specified offset for the k-th location, e.g. K = 9 and
+\f$p_{k} \in \{(-1, -1),(-1, 0), . . . ,(1, 1)\}\f$
+
+* \f${\Delta}p_{k}\f$ is the learnable offset for the k-th location.
+
+* \f${\Delta}m_{k}\f$ is the modulation scalar from 0 to 1 for the k-th location.
+
+**Attributes**:
+
+* *strides*
+
+  * **Description**: *strides* is a distance (in pixels) to slide the filter on the feature map over the `(y,x)` axes. For example, *strides* equal `2,1` means sliding the filter 2 pixel at a time over height dimension and 1 over width dimension.
+  * **Range of values**: integer values starting from `0`
+  * **Type**: `int[]`
+  * **Default value**: None
+  * **Required**: *yes*
+
+* *pads_begin*
+
+  * **Description**: *pads_begin* is a number of pixels to add to the beginning along each axis. For example, *pads_begin* equal `1,2` means adding 1 pixel to the top of the input and 2 to the left of the input.
+  * **Range of values**: integer values starting from `0`
+  * **Type**: `int[]`
+  * **Default value**: None
+  * **Required**: *yes*
+  * **Note**: the attribute is ignored when *auto_pad* attribute is specified.
+
+* *pads_end*
+
+  * **Description**: *pads_end* is a number of pixels to add to the ending along each axis. For example, *pads_end* equal `1,2` means adding 1 pixel to the bottom of the input and 2 to the right of the input.
+  * **Range of values**: integer values starting from `0`
+  * **Type**: `int[]`
+  * **Default value**: None
+  * **Required**: *yes*
+  * **Note**: the attribute is ignored when *auto_pad* attribute is specified.
+
+* *dilations*
+
+  * **Description**: *dilations* denotes the distance in width and height between elements (weights) in the filter. For example, *dilation* equal `1,1` means that all the elements in the filter are neighbors, so it is the same as for the usual convolution. *dilation* equal `2,2` means that all the elements in the filter are matched not to adjacent elements in the input matrix, but to those that are adjacent with distance 1.
+  * **Range of values**: integer value starting from `0`
+  * **Type**: `int[]`
+  * **Default value**: None
+  * **Required**: *yes*
+
+* *auto_pad*
+
+  * **Description**: *auto_pad* how the padding is calculated. Possible values:
+    * *explicit* - use explicit padding values from *pads_begin* and *pads_end*.
+    * *same_upper* - the input is padded to match the output size. In case of odd padding value an extra padding is added at the end.
+    * *same_lower* - the input is padded to match the output size. In case of odd padding value an extra padding is added at the beginning.
+    * *valid* - do not use padding.
+  * **Type**: `string`
+  * **Default value**: explicit
+  * **Required**: *no*
+  * **Note**: *pads_begin* and *pads_end* attributes are ignored when *auto_pad* is specified.
+
+
+* *group*
+
+  * **Description**: *group* is the number of groups which *output* and *input* should be split into. For example, *group* equal to 1 means that all filters are applied to the whole input (usual convolution), *group* equal to 2 means that both *input* and *output* channels are separated into two groups and the *i-th output* group is connected to the *i-th input* group channel. *group* equal to a number of output feature maps implies depth-wise separable convolution.
+  * **Range of values**: integer value starting from `1`
+  * **Type**: `int`
+  * **Default value**: `1`
+  * **Required**: *no*
+
+* *deformable_group*
+
+  * **Description**: *deformable_group* is the number of groups in which *offsets* input and *output* should be split into along the channel axis. Apply the deformable convolution using the i-th part of the offsets part on the i-th out.
+  * **Range of values**: integer value starting from `1`
+  * **Type**: `int`
+  * **Default value**: `1`
+  * **Required**: *no*
+
+* *bilinear_interpolation_padding*
+
+  * **Description**: *bilinear_interpolation_padding* is the number of pixels outside of the feature map boundary to apply bilinear interpolation.
+  * **Range of values**: non-negative integer value
+  * **Type**: `int`
+  * **Default value**: `0`
+  * **Required**: *no*
+  
+**Inputs**:
+
+*   **1**: Input tensor of type *T* and rank 4. Layout is `NCYX` (number of batches, number of channels, spatial axes Y and X). **Required.**
+
+*   **2**: Offsets tensor of type *T* and rank 4. Layout is `NCYX` (number of batches, *deformable_group* \* kernel_Y \* kernel_X \* 2, spatial axes Y and X). **Required.**
+
+*   **3**: Kernel tensor of type *T* and rank 4. Layout is `OIYX` (number of output channels, number of input channels, spatial axes Y and X). **Required.**
+
+*   **4**: ModulationScalars tensor of type *T2* and rank 4, the values are within [0, 1]. Layout is `NCYX` (number of batches, *deformable_group* \* kernel_Y \* kernel_X, spatial axes Y and X). If the input is not provided, the values are assumed to be equal to 1. **Optional.**
+
+
+**Outputs**:
+
+*   **1**: Output tensor of type *T* and rank 4. Layout is `NOYX` (number of batches, number of kernel output channels, spatial axes Y and X).
+
+**Types**:
+
+* *T*: Any numeric type.
+* *T2*: Any supported floating point.
+ 
+**Example**
+
+2D DeformableConvolution (deformable_group=1)
+```xml
+<layer type="DeformableConvolution" ...>
+    <data dilations="1,1" pads_begin="0,0" pads_end="0,0" strides="1,1" auto_pad="explicit" group="1" deformable_group="1"/>
+    <input>
+        <port id="0">
+            <dim>1</dim>
+            <dim>4</dim>
+            <dim>224</dim>
+            <dim>224</dim>
+        </port>
+        <port id="1">
+            <dim>1</dim>
+            <dim>50</dim>
+            <dim>220</dim>
+            <dim>220</dim>
+        </port>
+        <port id="2">
+            <dim>64</dim>
+            <dim>4</dim>
+            <dim>5</dim>
+            <dim>5</dim>
+        </port>
+        <port id="3">
+            <dim>1</dim>
+            <dim>25</dim>
+            <dim>220</dim>
+            <dim>220</dim>
+        </port>
+    </input>
+    <output>
+        <port id="4" precision="FP32">
+            <dim>1</dim>
+            <dim>64</dim>
+            <dim>220</dim>
+            <dim>220</dim>
+        </port>
+    </output>
+</layer>
+```
diff --git a/docs/ops/opset8.md b/docs/ops/opset8.md
index 8f43927b5ec360..fc68d6f32e07ea 100644
--- a/docs/ops/opset8.md
+++ b/docs/ops/opset8.md
@@ -40,7 +40,7 @@ declared in `namespace opset8`.
 * [Cos](arithmetic/Cos_1.md)
 * [Cosh](arithmetic/Cosh_1.md)
 * [CumSum](arithmetic/CumSum_3.md)
-* [DeformableConvolution](convolution/DeformableConvolution_1.md)
+* [DeformableConvolution](convolution/DeformableConvolution_8.md)
 * [DeformablePSROIPooling](detection/DeformablePSROIPooling_1.md)
 * [DepthToSpace](movement/DepthToSpace_1.md)
 * [DetectionOutput](detection/DetectionOutput_1.md)

From 70f9d8564e5aced17ca975e5b543e5f7666b0127 Mon Sep 17 00:00:00 2001
From: Vladimir Gavrilov <vladimir.gavrilov@intel.com>
Date: Fri, 18 Jun 2021 09:54:41 +0300
Subject: [PATCH 131/178] Written reading the operation CTCGreedyDecoder for
 ONNX (as a custom operation). (#6215)

---
 model-optimizer/automation/package_BOM.txt    |  7 ++++---
 .../{tf => }/CTCGreedyDecoderReplacement.py   |  0
 .../front/{tf => }/CTCLossReplacement.py      |  2 +-
 .../front/onnx/CTCGreedyDecoder_ext.py        | 19 +++++++++++++++++++
 .../{tf => }/sparse_to_dense_replacer.py      |  4 ++--
 .../CTCGreedyDecoderReplacement_test.py       |  2 +-
 .../front/{tf => }/CTCLossReplacement_test.py |  2 +-
 .../{tf => }/sparse_to_dense_replacer_test.py |  2 +-
 8 files changed, 29 insertions(+), 9 deletions(-)
 rename model-optimizer/extensions/front/{tf => }/CTCGreedyDecoderReplacement.py (100%)
 rename model-optimizer/extensions/front/{tf => }/CTCLossReplacement.py (98%)
 create mode 100644 model-optimizer/extensions/front/onnx/CTCGreedyDecoder_ext.py
 rename model-optimizer/extensions/front/{tf => }/sparse_to_dense_replacer.py (93%)
 rename model-optimizer/unit_tests/extensions/front/{tf => }/CTCGreedyDecoderReplacement_test.py (97%)
 rename model-optimizer/unit_tests/extensions/front/{tf => }/CTCLossReplacement_test.py (99%)
 rename model-optimizer/unit_tests/extensions/front/{tf => }/sparse_to_dense_replacer_test.py (96%)

diff --git a/model-optimizer/automation/package_BOM.txt b/model-optimizer/automation/package_BOM.txt
index e7c779543cdda8..668eb02325ff97 100644
--- a/model-optimizer/automation/package_BOM.txt
+++ b/model-optimizer/automation/package_BOM.txt
@@ -134,6 +134,8 @@ extensions/front/caffe/split_to_identity.py
 extensions/front/caffe/tanh.py
 extensions/front/ChangePlaceholderTypes.py
 extensions/front/create_tensor_nodes.py
+extensions/front/CTCGreedyDecoderReplacement.py
+extensions/front/CTCLossReplacement.py
 extensions/front/disable_weights_quantize_value_propagation.py
 extensions/front/div.py
 extensions/front/DropoutWithRandomUniformReplacer.py
@@ -274,6 +276,7 @@ extensions/front/onnx/constant_of_shape_ext.py
 extensions/front/onnx/constant_of_shape_to_broadcast.py
 extensions/front/onnx/conv_ext.py
 extensions/front/onnx/crop_ext.py
+extensions/front/onnx/CTCGreedyDecoder_ext.py
 extensions/front/onnx/cumsum_ext.py
 extensions/front/onnx/deformable_conv_ext.py
 extensions/front/onnx/depth_to_space_ext.py
@@ -370,6 +373,7 @@ extensions/front/SizeReplacer.py
 extensions/front/softmax.py
 extensions/front/Softplus_fusion.py
 extensions/front/softsign_replacer.py
+extensions/front/sparse_to_dense_replacer.py
 extensions/front/split_normalizer.py
 extensions/front/SqueezeNormalize.py
 extensions/front/sub.py
@@ -401,9 +405,7 @@ extensions/front/tf/CorrectRollAxes.py
 extensions/front/tf/crop_and_resize_ext.py
 extensions/front/tf/CropAndResizeReplacement.py
 extensions/front/tf/CTCGreedyDecoder_ext.py
-extensions/front/tf/CTCGreedyDecoderReplacement.py
 extensions/front/tf/CTCLoss_ext.py
-extensions/front/tf/CTCLossReplacement.py
 extensions/front/tf/cumsum_ext.py
 extensions/front/tf/deconv_ext.py
 extensions/front/tf/depth_to_space.py
@@ -497,7 +499,6 @@ extensions/front/tf/sparse_fill_empty_rows_ext.py
 extensions/front/tf/sparse_segment_mean_ext.py
 extensions/front/tf/sparse_segment_sqrtn_ext.py
 extensions/front/tf/sparse_segment_sum_ext.py
-extensions/front/tf/sparse_to_dense_replacer.py
 extensions/front/tf/split_ext.py
 extensions/front/tf/ssd_support.json
 extensions/front/tf/ssd_support_api_v1.14.json
diff --git a/model-optimizer/extensions/front/tf/CTCGreedyDecoderReplacement.py b/model-optimizer/extensions/front/CTCGreedyDecoderReplacement.py
similarity index 100%
rename from model-optimizer/extensions/front/tf/CTCGreedyDecoderReplacement.py
rename to model-optimizer/extensions/front/CTCGreedyDecoderReplacement.py
diff --git a/model-optimizer/extensions/front/tf/CTCLossReplacement.py b/model-optimizer/extensions/front/CTCLossReplacement.py
similarity index 98%
rename from model-optimizer/extensions/front/tf/CTCLossReplacement.py
rename to model-optimizer/extensions/front/CTCLossReplacement.py
index c47296db98e39e..ea2e6cbd3bc135 100644
--- a/model-optimizer/extensions/front/tf/CTCLossReplacement.py
+++ b/model-optimizer/extensions/front/CTCLossReplacement.py
@@ -20,7 +20,7 @@ class CTCLossReplacement(FrontReplacementSubgraph):
     enabled = True
 
     def run_before(self):
-        from extensions.front.tf.CTCGreedyDecoderReplacement import CTCGreedyDecoderReplacement
+        from extensions.front.CTCGreedyDecoderReplacement import CTCGreedyDecoderReplacement
         return [CTCGreedyDecoderReplacement]
 
     def pattern(self):
diff --git a/model-optimizer/extensions/front/onnx/CTCGreedyDecoder_ext.py b/model-optimizer/extensions/front/onnx/CTCGreedyDecoder_ext.py
new file mode 100644
index 00000000000000..31e0799876c4c2
--- /dev/null
+++ b/model-optimizer/extensions/front/onnx/CTCGreedyDecoder_ext.py
@@ -0,0 +1,19 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+from extensions.ops.ctc_greedy_decoder_seq_len import CTCGreedyDecoderSeqLenOp
+from mo.front.extractor import FrontExtractorOp
+from mo.front.onnx.extractors.utils import onnx_attr
+
+
+class CTCCGreedyDecoderFrontExtractor(FrontExtractorOp):
+    op = 'CTCGreedyDecoder'
+    enabled = True
+
+    @classmethod
+    def extract(cls, node):
+        attrs = {
+            'merge_repeated': bool(onnx_attr(node, 'merge_repeated', 'i', default=1)),
+        }
+        CTCGreedyDecoderSeqLenOp.update_node_stat(node, attrs)
+        return cls.enabled
diff --git a/model-optimizer/extensions/front/tf/sparse_to_dense_replacer.py b/model-optimizer/extensions/front/sparse_to_dense_replacer.py
similarity index 93%
rename from model-optimizer/extensions/front/tf/sparse_to_dense_replacer.py
rename to model-optimizer/extensions/front/sparse_to_dense_replacer.py
index 1440b150696a94..f285192a5a009c 100644
--- a/model-optimizer/extensions/front/tf/sparse_to_dense_replacer.py
+++ b/model-optimizer/extensions/front/sparse_to_dense_replacer.py
@@ -21,8 +21,8 @@ class SparseToDenseReplacer(FrontReplacementOp):
     enabled = True
 
     def run_after(self):
-        from extensions.front.tf.CTCGreedyDecoderReplacement import CTCGreedyDecoderReplacement
-        from extensions.front.tf.CTCLossReplacement import CTCLossReplacement
+        from extensions.front.CTCGreedyDecoderReplacement import CTCGreedyDecoderReplacement
+        from extensions.front.CTCLossReplacement import CTCLossReplacement
         return [CTCGreedyDecoderReplacement, CTCLossReplacement]
 
     def replace_op(self, graph: Graph, node: Node):
diff --git a/model-optimizer/unit_tests/extensions/front/tf/CTCGreedyDecoderReplacement_test.py b/model-optimizer/unit_tests/extensions/front/CTCGreedyDecoderReplacement_test.py
similarity index 97%
rename from model-optimizer/unit_tests/extensions/front/tf/CTCGreedyDecoderReplacement_test.py
rename to model-optimizer/unit_tests/extensions/front/CTCGreedyDecoderReplacement_test.py
index c4d1d467df4160..063d71173e1f7e 100644
--- a/model-optimizer/unit_tests/extensions/front/tf/CTCGreedyDecoderReplacement_test.py
+++ b/model-optimizer/unit_tests/extensions/front/CTCGreedyDecoderReplacement_test.py
@@ -3,7 +3,7 @@
 
 import unittest
 
-from extensions.front.tf.CTCGreedyDecoderReplacement import CTCGreedyDecoderReplacement, CTCGreedyDecoderWithSparseToDenseShapeReplacement
+from extensions.front.CTCGreedyDecoderReplacement import CTCGreedyDecoderReplacement, CTCGreedyDecoderWithSparseToDenseShapeReplacement
 from mo.front.common.partial_infer.utils import int64_array
 from mo.utils.ir_engine.compare_graphs import compare_graphs
 from unit_tests.utils.graph import build_graph, const
diff --git a/model-optimizer/unit_tests/extensions/front/tf/CTCLossReplacement_test.py b/model-optimizer/unit_tests/extensions/front/CTCLossReplacement_test.py
similarity index 99%
rename from model-optimizer/unit_tests/extensions/front/tf/CTCLossReplacement_test.py
rename to model-optimizer/unit_tests/extensions/front/CTCLossReplacement_test.py
index f173ec15c17367..77ecfc8fa1afd9 100644
--- a/model-optimizer/unit_tests/extensions/front/tf/CTCLossReplacement_test.py
+++ b/model-optimizer/unit_tests/extensions/front/CTCLossReplacement_test.py
@@ -5,7 +5,7 @@
 import unittest
 from argparse import Namespace
 
-from extensions.front.tf.CTCLossReplacement import CTCLossReplacement
+from extensions.front.CTCLossReplacement import CTCLossReplacement
 from mo.front.common.partial_infer.utils import int64_array
 from mo.utils.ir_engine.compare_graphs import compare_graphs
 from unit_tests.utils.graph import build_graph, const
diff --git a/model-optimizer/unit_tests/extensions/front/tf/sparse_to_dense_replacer_test.py b/model-optimizer/unit_tests/extensions/front/sparse_to_dense_replacer_test.py
similarity index 96%
rename from model-optimizer/unit_tests/extensions/front/tf/sparse_to_dense_replacer_test.py
rename to model-optimizer/unit_tests/extensions/front/sparse_to_dense_replacer_test.py
index 4e624dfd069c1e..588d52198c9c16 100644
--- a/model-optimizer/unit_tests/extensions/front/tf/sparse_to_dense_replacer_test.py
+++ b/model-optimizer/unit_tests/extensions/front/sparse_to_dense_replacer_test.py
@@ -3,7 +3,7 @@
 
 import unittest
 
-from extensions.front.tf.sparse_to_dense_replacer import SparseToDenseReplacer
+from extensions.front.sparse_to_dense_replacer import SparseToDenseReplacer
 from mo.front.common.partial_infer.utils import int64_array
 from mo.utils.ir_engine.compare_graphs import compare_graphs
 from unit_tests.utils.graph import build_graph, const

From 1048e6f69b3590f91d83c17c40a426d6b5a16e43 Mon Sep 17 00:00:00 2001
From: Krzysztof Bruniecki <krzysztof.bruniecki@intel.com>
Date: Fri, 18 Jun 2021 11:42:18 +0200
Subject: [PATCH 132/178] [GNA] Introduce an option to invoke the QoS feature
 (#5827)

* [GNA] [WIP] Introduce an option to invoke the QoS feature 56759

* Apply remaining revew (typos) for PR 5741

* Introduce HW_WITH_SW_FBACK

* Add unit test for HW_WITH_SW_FBACK

* Enable HW_WITH_SW_FBACK in speech_sample cpp

* Use perf counters to report number of HW delivered frames to the user (eg speech_sample)

* Update speech_sample.hpp

based on review

* Update gna_config.hpp

Describe special perf counter

* lint fix

* Apply review

  * Remove special performance counter
  * Add GNA frequency for 6/151 CPU family/model

* Update inference-engine/samples/speech_sample/main.cpp

Co-authored-by: Mikhail Ryzhov <mikhail.ryzhov@intel.com>

* Update main.cpp

Co-authored-by: Mikhail Ryzhov <mikhail.ryzhov@intel.com>
---
 inference-engine/include/gna/gna_config.hpp   |  1 +
 .../samples/speech_sample/main.cpp            | 45 +++++++++++--------
 .../samples/speech_sample/speech_sample.hpp   |  5 +--
 .../src/gna_plugin/gna_device.cpp             |  8 ++--
 .../src/gna_plugin/gna_device.hpp             |  6 ---
 .../src/gna_plugin/gna_plugin_config.cpp      | 26 ++++++-----
 .../convolution_negative.cpp                  | 26 +++++------
 .../tests/unit/gna/gna_plugin_config_test.cpp |  5 +++
 8 files changed, 66 insertions(+), 56 deletions(-)

diff --git a/inference-engine/include/gna/gna_config.hpp b/inference-engine/include/gna/gna_config.hpp
index 958227696a1c98..3433ab58887f7a 100644
--- a/inference-engine/include/gna/gna_config.hpp
+++ b/inference-engine/include/gna/gna_config.hpp
@@ -65,6 +65,7 @@ DECLARE_GNA_CONFIG_KEY(DEVICE_MODE);
 
 DECLARE_GNA_CONFIG_VALUE(AUTO);
 DECLARE_GNA_CONFIG_VALUE(HW);
+DECLARE_GNA_CONFIG_VALUE(HW_WITH_SW_FBACK);
 DECLARE_GNA_CONFIG_VALUE(SW);
 DECLARE_GNA_CONFIG_VALUE(SW_EXACT);
 DECLARE_GNA_CONFIG_VALUE(SW_FP32);
diff --git a/inference-engine/samples/speech_sample/main.cpp b/inference-engine/samples/speech_sample/main.cpp
index 2b9131774ad557..57db61a8e9e8cd 100644
--- a/inference-engine/samples/speech_sample/main.cpp
+++ b/inference-engine/samples/speech_sample/main.cpp
@@ -236,7 +236,8 @@ float getGnaFrequencyMHz() {
     const uint8_t cannon_lake_model = 102;
     const uint8_t gemini_lake_model = 122;
     const uint8_t ice_lake_model = 126;
-    const uint8_t next_model = 140;
+    const uint8_t tgl_model = 140;
+    const uint8_t next_model = 151;
 
     native_cpuid(&eax, &ebx, &ecx, &edx);
     family = (eax >> 8) & 0xF;
@@ -254,6 +255,7 @@ float getGnaFrequencyMHz() {
         switch (model) {
         case cannon_lake_model:
         case ice_lake_model:
+        case tgl_model:
         case next_model:
             return 400;
         case gemini_lake_model:
@@ -287,13 +289,14 @@ void printReferenceCompareResults(score_error_t const& totalError, size_t frames
 /**
  * @brief Print a report on the performance counts
  * @param utterancePerfMap reference to a map to store performance counters
- * @param callsNum frame index
+ * @param numberOfFrames number of frames
  * @param stream output stream
  * @param fullDeviceName full device name string
+ * @param numberOfFramesOnHw number of frames delivered to GNA HW
  * @return none.
  */
-void printPerformanceCounters(std::map<std::string, InferenceEngine::InferenceEngineProfileInfo> const& utterancePerfMap, size_t callsNum, std::ostream& stream,
-                              std::string fullDeviceName) {
+void printPerformanceCounters(std::map<std::string, InferenceEngine::InferenceEngineProfileInfo> const& utterancePerfMap, size_t numberOfFrames,
+                              std::ostream& stream, std::string fullDeviceName, const uint64_t numberOfFramesOnHw) {
 #if !defined(__arm__) && !defined(_M_ARM) && !defined(__aarch64__) && !defined(_M_ARM64)
     stream << std::endl << "Performance counts:" << std::endl;
     stream << std::setw(10) << std::right << ""
@@ -305,29 +308,29 @@ void printPerformanceCounters(std::map<std::string, InferenceEngine::InferenceEn
     stream << std::setw(46) << "(ms)";
     stream << std::setw(24) << "(us per call)";
     stream << std::endl;
-
+    // if GNA HW counters
+    // get frequency of GNA module
+    float freq = getGnaFrequencyMHz();
     for (const auto& it : utterancePerfMap) {
         std::string const& counter_name = it.first;
-        float current_units = static_cast<float>(it.second.realTime_uSec);
-        float call_units = current_units / callsNum;
-        // if GNA HW counters
-        // get frequency of GNA module
-        float freq = getGnaFrequencyMHz();
-        current_units /= freq * 1000;
-        call_units /= freq;
+        float current_units_us = static_cast<float>(it.second.realTime_uSec) / freq;
+        float call_units_us = current_units_us / numberOfFrames;
         if (FLAGS_d.find("GNA") != std::string::npos) {
             stream << std::setw(30) << std::left << counter_name.substr(4, counter_name.size() - 1);
         } else {
             stream << std::setw(30) << std::left << counter_name;
         }
-        stream << std::setw(16) << std::right << current_units;
-        stream << std::setw(21) << std::right << call_units;
+        stream << std::setw(16) << std::right << current_units_us / 1000;
+        stream << std::setw(21) << std::right << call_units_us;
         stream << std::endl;
     }
     stream << std::endl;
     std::cout << std::endl;
     std::cout << "Full device name: " << fullDeviceName << std::endl;
     std::cout << std::endl;
+    stream << "Number of frames delivered to GNA HW: " << numberOfFramesOnHw;
+    stream << "/" << numberOfFrames;
+    stream << std::endl;
 #endif
 }
 
@@ -346,16 +349,20 @@ void getPerformanceCounters(InferenceEngine::InferRequest& request, std::map<std
 }
 
 /**
- * @brief Summarize performance counts
+ * @brief Summarize performance counts and total number of frames executed on the GNA HW device
  * @param perfCounters reference to a map to get performance counters
  * @param totalPerfCounters reference to a map to save total performance counters
+ * @param totalRunsOnHw reference to a total number of frames computed on GNA HW
  * @return none.
  */
 void sumPerformanceCounters(std::map<std::string, InferenceEngine::InferenceEngineProfileInfo> const& perfCounters,
-                            std::map<std::string, InferenceEngine::InferenceEngineProfileInfo>& totalPerfCounters) {
+                            std::map<std::string, InferenceEngine::InferenceEngineProfileInfo>& totalPerfCounters, uint64_t& totalRunsOnHw) {
+    auto runOnHw = false;
     for (const auto& pair : perfCounters) {
         totalPerfCounters[pair.first].realTime_uSec += pair.second.realTime_uSec;
+        runOnHw |= pair.second.realTime_uSec > 0;  // if realTime is above zero, that means that a primitive was executed on the device
     }
+    totalRunsOnHw += runOnHw;
 }
 
 /**
@@ -443,6 +450,7 @@ bool ParseAndCheckCommandLine(int argc, char* argv[]) {
                                                  "GPU",
                                                  "GNA_AUTO",
                                                  "GNA_HW",
+                                                 "GNA_HW_WITH_SW_FBACK",
                                                  "GNA_SW_EXACT",
                                                  "GNA_SW",
                                                  "GNA_SW_FP32",
@@ -829,6 +837,7 @@ int main(int argc, char* argv[]) {
             /** Work with each utterance **/
             for (uint32_t utteranceIndex = 0; utteranceIndex < numUtterances; ++utteranceIndex) {
                 std::map<std::string, InferenceEngine::InferenceEngineProfileInfo> utterancePerfMap;
+                uint64_t totalNumberOfRunsOnHw = 0;
                 std::string uttName;
                 uint32_t numFrames(0), n(0);
                 std::vector<uint32_t> numFrameElementsInput;
@@ -984,7 +993,7 @@ int main(int argc, char* argv[]) {
                                     // retrieve new counters
                                     getPerformanceCounters(inferRequest.inferRequest, callPerfMap);
                                     // summarize retrieved counters with all previous
-                                    sumPerformanceCounters(callPerfMap, utterancePerfMap);
+                                    sumPerformanceCounters(callPerfMap, utterancePerfMap, totalNumberOfRunsOnHw);
                                 }
                             }
                             // -----------------------------------------------------------------------------------------------------
@@ -1092,7 +1101,7 @@ int main(int argc, char* argv[]) {
                 std::cout << "Average Infer time per frame:\t\t" << totalTime / static_cast<double>(numFrames) << " ms" << std::endl;
                 if (FLAGS_pc) {
                     // print performance results
-                    printPerformanceCounters(utterancePerfMap, frameIndex, std::cout, getFullDeviceName(ie, FLAGS_d));
+                    printPerformanceCounters(utterancePerfMap, frameIndex, std::cout, getFullDeviceName(ie, FLAGS_d), totalNumberOfRunsOnHw);
                 }
                 if (!FLAGS_r.empty()) {
                     // print statistical score error
diff --git a/inference-engine/samples/speech_sample/speech_sample.hpp b/inference-engine/samples/speech_sample/speech_sample.hpp
index cafe4db5c61758..66d3b24a4c5c47 100644
--- a/inference-engine/samples/speech_sample/speech_sample.hpp
+++ b/inference-engine/samples/speech_sample/speech_sample.hpp
@@ -21,10 +21,9 @@ static const char model_message[] = "Required. Path to an .xml file with a train
 
 /// @brief message for assigning cnn calculation to device
 static const char target_device_message[] = "Optional. Specify a target device to infer on. CPU, GPU, MYRIAD, GNA_AUTO, GNA_HW, "
-                                            "GNA_SW_FP32, "
+                                            "GNA_HW_WITH_SW_FBACK, GNA_SW_FP32, "
                                             "GNA_SW_EXACT and HETERO with combination of GNA as the primary device and CPU"
-                                            " as a secondary (e.g. HETERO:GNA,CPU) are supported. The list of available devices is shown "
-                                            "below. "
+                                            " as a secondary (e.g. HETERO:GNA,CPU) are supported. "
                                             "The sample will look for a suitable plugin for device specified.";
 
 /// @brief message for execution target
diff --git a/inference-engine/src/gna_plugin/gna_device.cpp b/inference-engine/src/gna_plugin/gna_device.cpp
index cbfc47f57aab0b..85a246ea34f134 100644
--- a/inference-engine/src/gna_plugin/gna_device.cpp
+++ b/inference-engine/src/gna_plugin/gna_device.cpp
@@ -96,14 +96,12 @@ void GNADeviceHelper::setUpActiveList(const uint32_t requestConfigId, uint32_t l
     const auto status = Gna2RequestConfigEnableActiveList(requestConfigId, layerIndex, num_active_indices, ptr_active_indices);
     checkGna2Status(status, "Gna2RequestConfigEnableActiveList");
 }
-void GNADeviceHelper::propagateSync(const uint32_t requestConfigId, Gna2AccelerationMode gna2AccelerationMode) {
-    wait(propagate(requestConfigId, gna2AccelerationMode));
-}
 
 uint32_t GNADeviceHelper::propagate(const uint32_t requestConfigId, Gna2AccelerationMode gna2AccelerationMode) {
     std::unique_lock<std::mutex> lockGnaCalls{ acrossPluginsSync };
     uint32_t reqId{};
-    if (gna2AccelerationMode == Gna2AccelerationModeHardware &&
+    if ((gna2AccelerationMode == Gna2AccelerationModeHardware ||
+         gna2AccelerationMode == Gna2AccelerationModeHardwareWithSoftwareFallback) &&
         detectedGnaDevVersion == Gna2DeviceVersionSoftwareEmulation) {
         gnawarn() << "GNA Device not detected, consider using other mode of acceleration";
     }
@@ -541,6 +539,8 @@ void GNADeviceHelper::updateGnaPerfCounters() {
 #if GNA_LIB_VER == 2
     instrumentationTotal[0] = instrumentationResults[0];
     instrumentationTotal[1] = instrumentationResults[1];
+    instrumentationResults[0] = 0;
+    instrumentationResults[1] = 0;
 #else
     nGNAPerfResultsTotal.hw.stall = nGNAPerfResults.hw.stall;
     nGNAPerfResultsTotal.hw.total = nGNAPerfResults.hw.total;
diff --git a/inference-engine/src/gna_plugin/gna_device.hpp b/inference-engine/src/gna_plugin/gna_device.hpp
index e032e5532dafc3..cae32c70b1de3e 100644
--- a/inference-engine/src/gna_plugin/gna_device.hpp
+++ b/inference-engine/src/gna_plugin/gna_device.hpp
@@ -117,18 +117,12 @@ class GNADeviceHelper {
     uint8_t *alloc(uint32_t size_requested, uint32_t *size_granted);
 
 #if GNA_LIB_VER == 1
-    void propagateSync(const intel_nnet_type_t *pNeuralNetwork,
-                       const uint32_t *pActiveIndices,
-                       uint32_t nActiveIndices,
-                       intel_gna_proc_t nGNAProcType);
-
     uint32_t propagate(const intel_nnet_type_t *pNeuralNetwork,
                        const uint32_t *pActiveIndices,
                        uint32_t nActiveIndices,
                        intel_gna_proc_t nGNAProcType);
 #else
     void setUpActiveList(unsigned req_config_id, uint32_t layerIndex, uint32_t* ptr_active_indices, uint32_t num_active_indices);
-    void propagateSync(const uint32_t requestConfigId, Gna2AccelerationMode gna2AccelerationMode);
     uint32_t propagate(const uint32_t requestConfigId, Gna2AccelerationMode gna2AccelerationMode);
     uint32_t createModel(Gna2Model& gnaModel) const;
     void releaseModel(const uint32_t model_id);
diff --git a/inference-engine/src/gna_plugin/gna_plugin_config.cpp b/inference-engine/src/gna_plugin/gna_plugin_config.cpp
index 2dcb05d6ab8a3f..766e7d2d52c609 100644
--- a/inference-engine/src/gna_plugin/gna_plugin_config.cpp
+++ b/inference-engine/src/gna_plugin/gna_plugin_config.cpp
@@ -23,6 +23,7 @@ static const caseless_unordered_map<std::string, uint32_t> supported_values = {
         {GNAConfigParams::GNA_SW_EXACT, GNA_SOFTWARE & GNA_HARDWARE}
 };
 static const  std::vector<std::string> supported_values_on_gna2 = {
+        GNAConfigParams::GNA_HW_WITH_SW_FBACK,
         GNAConfigParams::GNA_GEN,
         GNAConfigParams::GNA_GEN_EXACT,
         GNAConfigParams::GNA_SSE,
@@ -34,18 +35,19 @@ static const  std::vector<std::string> supported_values_on_gna2 = {
 };
 #else
 static const caseless_unordered_map <std::string, std::pair<Gna2AccelerationMode, bool>> supported_values = {
-                {GNAConfigParams::GNA_AUTO,       {Gna2AccelerationModeAuto,     false}},
-                {GNAConfigParams::GNA_HW,         {Gna2AccelerationModeHardware, false}},
-                {GNAConfigParams::GNA_SW,         {Gna2AccelerationModeSoftware, false}},
-                {GNAConfigParams::GNA_SW_EXACT,   {Gna2AccelerationModeSoftware, true}},
-                {GNAConfigParams::GNA_GEN,        {Gna2AccelerationModeGeneric,  false}},
-                {GNAConfigParams::GNA_GEN_EXACT,  {Gna2AccelerationModeGeneric,  true}},
-                {GNAConfigParams::GNA_SSE,        {Gna2AccelerationModeSse4x2,   false}},
-                {GNAConfigParams::GNA_SSE_EXACT,  {Gna2AccelerationModeSse4x2,   true}},
-                {GNAConfigParams::GNA_AVX1,       {Gna2AccelerationModeAvx1,     false}},
-                {GNAConfigParams::GNA_AVX1_EXACT, {Gna2AccelerationModeAvx1,     true}},
-                {GNAConfigParams::GNA_AVX2,       {Gna2AccelerationModeAvx2,     false}},
-                {GNAConfigParams::GNA_AVX2_EXACT, {Gna2AccelerationModeAvx2,     true}},
+                {GNAConfigParams::GNA_AUTO,             {Gna2AccelerationModeAuto,                         false}},
+                {GNAConfigParams::GNA_HW,               {Gna2AccelerationModeHardware,                     false}},
+                {GNAConfigParams::GNA_HW_WITH_SW_FBACK, {Gna2AccelerationModeHardwareWithSoftwareFallback, false}},
+                {GNAConfigParams::GNA_SW,               {Gna2AccelerationModeSoftware,                     false}},
+                {GNAConfigParams::GNA_SW_EXACT,         {Gna2AccelerationModeSoftware,                     true}},
+                {GNAConfigParams::GNA_GEN,              {Gna2AccelerationModeGeneric,                      false}},
+                {GNAConfigParams::GNA_GEN_EXACT,        {Gna2AccelerationModeGeneric,                      true}},
+                {GNAConfigParams::GNA_SSE,              {Gna2AccelerationModeSse4x2,                       false}},
+                {GNAConfigParams::GNA_SSE_EXACT,        {Gna2AccelerationModeSse4x2,                       true}},
+                {GNAConfigParams::GNA_AVX1,             {Gna2AccelerationModeAvx1,                         false}},
+                {GNAConfigParams::GNA_AVX1_EXACT,       {Gna2AccelerationModeAvx1,                         true}},
+                {GNAConfigParams::GNA_AVX2,             {Gna2AccelerationModeAvx2,                         false}},
+                {GNAConfigParams::GNA_AVX2_EXACT,       {Gna2AccelerationModeAvx2,                         true}},
         };
 #endif
 
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution_negative.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution_negative.cpp
index e25236aafabf7e..aa4975e602facd 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution_negative.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution_negative.cpp
@@ -24,7 +24,7 @@ const std::vector<std::vector<size_t >> kernels2D = {
     {3, 3},
 };
 
-const std::vector<std::vector<size_t >> InvalidKernels2D = {
+const std::vector<std::vector<size_t >> kernels2DInvalid = {
     {1, 4},
     {2, 3},
     {3, 2},
@@ -50,8 +50,8 @@ const std::vector<std::vector<size_t >> dilations2D = { {1, 1},
 };
 const std::vector<std::vector<size_t >> dilations2DInvalid = { {2, 2},
 };
-const std::vector<size_t> numOutCannels2D = { 32 };
-const std::vector<size_t> numOutCannels2DInvalid = { 1, 7, 9, 400 };
+const std::vector<size_t> numOutChannels2D = { 32 };
+const std::vector<size_t> numOutChannels2DInvalid = { 1, 7, 9, 400 };
 
 const std::vector<std::vector<size_t>> input2DNCHWFine = { { 1, 8, 20, 16 } };
 
@@ -68,16 +68,16 @@ const auto conv2DParametersFine = ::testing::Combine(
     ::testing::ValuesIn(padBegins2D),
     ::testing::ValuesIn(padEnds2D),
     ::testing::ValuesIn(dilations2D),
-    ::testing::ValuesIn(numOutCannels2D),
+    ::testing::ValuesIn(numOutChannels2D),
     ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 const auto conv2DParametersInvalidKernel = ::testing::Combine(
-    ::testing::ValuesIn(InvalidKernels2D),
+    ::testing::ValuesIn(kernels2DInvalid),
     ::testing::ValuesIn(strides2D),
     ::testing::ValuesIn(padBegins2D),
     ::testing::ValuesIn(padEnds2D),
     ::testing::ValuesIn(dilations2D),
-    ::testing::ValuesIn(numOutCannels2D),
+    ::testing::ValuesIn(numOutChannels2D),
     ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 const auto conv2DParametersInvalidFilterNumber = ::testing::Combine(
@@ -86,7 +86,7 @@ const auto conv2DParametersInvalidFilterNumber = ::testing::Combine(
     ::testing::ValuesIn(padBegins2D),
     ::testing::ValuesIn(padEnds2D),
     ::testing::ValuesIn(dilations2D),
-    ::testing::ValuesIn(numOutCannels2DInvalid),
+    ::testing::ValuesIn(numOutChannels2DInvalid),
     ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 const auto conv2DParametersInvalidPadding = ::testing::Combine(
@@ -95,7 +95,7 @@ const auto conv2DParametersInvalidPadding = ::testing::Combine(
     ::testing::ValuesIn(padBegins2DInvalid),
     ::testing::ValuesIn(padEnds2DInvalid),
     ::testing::ValuesIn(dilations2D),
-    ::testing::ValuesIn(numOutCannels2D),
+    ::testing::ValuesIn(numOutChannels2D),
     ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 const auto conv2DParametersInvalidStride = ::testing::Combine(
@@ -104,7 +104,7 @@ const auto conv2DParametersInvalidStride = ::testing::Combine(
     ::testing::ValuesIn(padBegins2D),
     ::testing::ValuesIn(padEnds2D),
     ::testing::ValuesIn(dilations2D),
-    ::testing::ValuesIn(numOutCannels2D),
+    ::testing::ValuesIn(numOutChannels2D),
     ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 const auto conv2DParametersInvalidDilation = ::testing::Combine(
@@ -113,7 +113,7 @@ const auto conv2DParametersInvalidDilation = ::testing::Combine(
     ::testing::ValuesIn(padBegins2D),
     ::testing::ValuesIn(padEnds2D),
     ::testing::ValuesIn(dilations2DInvalid),
-    ::testing::ValuesIn(numOutCannels2D),
+    ::testing::ValuesIn(numOutChannels2D),
     ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 
@@ -142,7 +142,7 @@ class GnaConv2DNegativeTest : public ConvolutionLayerTest, protected GnaLayerTes
     }
 };
 
-#define GNA_NEG_INSTANTIATE(whats_wrong, sufix_params, sufix_input, error_message)                              \
+#define GNA_NEG_INSTANTIATE(whats_wrong, suffix_params, suffix_input, error_message)                            \
 struct GnaConv2DNegativeTest##whats_wrong : GnaConv2DNegativeTest {                                             \
     std::string expectedSubstring() override {                                                                  \
         return error_message;                                                                                   \
@@ -153,13 +153,13 @@ TEST_P(GnaConv2DNegativeTest##whats_wrong, ThrowAsNotSupported) {
 }                                                                                                               \
 INSTANTIATE_TEST_CASE_P(smoke_GnaConv2DNegativeTestInvalid##whats_wrong, GnaConv2DNegativeTest##whats_wrong,    \
 ::testing::Combine(                                                                                             \
-    conv2DParameters##sufix_params,                                                                             \
+    conv2DParameters##suffix_params,                                                                            \
     ::testing::ValuesIn(netPrecisions),                                                                         \
     ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),                                                 \
     ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),                                                 \
     ::testing::Values(InferenceEngine::Layout::ANY),                                                            \
     ::testing::Values(InferenceEngine::Layout::ANY),                                                            \
-    ::testing::ValuesIn(input2DNCHW##sufix_input),                                                              \
+    ::testing::ValuesIn(input2DNCHW##suffix_input),                                                             \
     ::testing::Values(CommonTestUtils::DEVICE_GNA)),                                                            \
     GnaConv2DNegativeTest##whats_wrong::getTestCaseName);
 
diff --git a/inference-engine/tests/unit/gna/gna_plugin_config_test.cpp b/inference-engine/tests/unit/gna/gna_plugin_config_test.cpp
index bdfd50ba03700d..7fa12a428251ff 100644
--- a/inference-engine/tests/unit/gna/gna_plugin_config_test.cpp
+++ b/inference-engine/tests/unit/gna/gna_plugin_config_test.cpp
@@ -107,6 +107,11 @@ TEST_F(GNAPluginConfigTest, GnaConfigDeviceModeTest) {
 #else
     EXPECT_EQ(config.pluginGna2AccMode, Gna2AccelerationModeHardware);
     EXPECT_EQ(config.swExactMode, false);
+#endif
+#if GNA_LIB_VER == 2
+    SetAndCompare(GNA_CONFIG_KEY(DEVICE_MODE), GNAConfigParams::GNA_HW_WITH_SW_FBACK);
+    EXPECT_EQ(config.pluginGna2AccMode, Gna2AccelerationModeHardwareWithSoftwareFallback);
+    EXPECT_EQ(config.swExactMode, false);
 #endif
     SetAndCompare(GNA_CONFIG_KEY(DEVICE_MODE), GNAConfigParams::GNA_SW);
 #if GNA_LIB_VER == 1

From b6e62e7a0b1ddb8d598a2edfc6d97cad59e2eefe Mon Sep 17 00:00:00 2001
From: Rafal Blaczkowski <rafal.blaczkowski@intel.com>
Date: Fri, 18 Jun 2021 11:58:42 +0200
Subject: [PATCH 133/178] Update Openvino ONNX CI Azure check (#6227)

* Update

* update models

* update models

* Fix models path
---
 .ci/azure/linux_ngraph_onnx.yml | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/.ci/azure/linux_ngraph_onnx.yml b/.ci/azure/linux_ngraph_onnx.yml
index e11e72e102d33f..1e13710f2c2b1e 100644
--- a/.ci/azure/linux_ngraph_onnx.yml
+++ b/.ci/azure/linux_ngraph_onnx.yml
@@ -17,6 +17,8 @@ jobs:
     WORK_DIR: $(Pipeline.Workspace)/_w
     MODELS_DIR: /mount/cinfsshare/onnxtestdata
     TMP_DIR: /mnt/tmp
+    ONNX_MODEL_ZOO_SHA: "d58213534f2a4d1c4b19ba62b3bb5f544353256e"
+
 
   steps:
   - script: |
@@ -55,7 +57,7 @@ jobs:
   - script: docker build --tag=openvino-onnx-ci-image --file=.ci/openvino-onnx/Dockerfile .
     displayName: 'Docker build'
 
-  - script: ngraph/python/tests/test_onnx/model_zoo_preprocess.sh -d $(TMP_DIR) -o
+  - script: ngraph/python/tests/test_onnx/model_zoo_preprocess.sh -d $(TMP_DIR) -o -s "$(ONNX_MODEL_ZOO_SHA)"
     displayName: 'Get models'
 
   - script: |
@@ -77,6 +79,6 @@ jobs:
     displayName: 'Create swap'
 
   - script: |
-      docker run --name openvino-onnx-ci-container --volume $(TMP_DIR)/model_zoo:/root/.onnx/model_zoo --volume $(MODELS_DIR)/msft:/root/.onnx/model_zoo/MSFT openvino-onnx-ci-image
+      docker run --name openvino-onnx-ci-container --volume $(TMP_DIR)/model_zoo/onnx_model_zoo_$(ONNX_MODEL_ZOO_SHA):/root/.onnx/model_zoo/onnx_model_zoo --volume $(MODELS_DIR)/msft:/root/.onnx/model_zoo/MSFT openvino-onnx-ci-image /bin/bash -c "tox && tox -e zoo_models"
     displayName: 'Docker run'
 

From 63211d548f5ed960e269ba6fa9e0015885cc96ba Mon Sep 17 00:00:00 2001
From: "Gladilov, Gleb" <gleb.gladilov@intel.com>
Date: Fri, 18 Jun 2021 13:34:30 +0300
Subject: [PATCH 134/178] Fixes skip tests config after tests renaming (#6231)

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>
---
 docs/template_plugin/tests/functional/skip_tests_config.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/template_plugin/tests/functional/skip_tests_config.cpp b/docs/template_plugin/tests/functional/skip_tests_config.cpp
index 252ed7c9a712c9..8d100118a9d248 100644
--- a/docs/template_plugin/tests/functional/skip_tests_config.cpp
+++ b/docs/template_plugin/tests/functional/skip_tests_config.cpp
@@ -10,10 +10,10 @@
 std::vector<std::string> disabledTestPatterns() {
     return {
         ".*ExclusiveAsyncRequests.*",
-        ".*reusableCPUStreamsExecutor.*",
+        ".*ReusableCPUStreamsExecutor.*",
         R"(.*SplitLayerTest.*numSplits\=30.*)",
         // CVS-51758
         ".*PreprocessConversionTest.*oLT=NHWC.*",
         ".*PreprocessDynamicallyInSetBlobTest.*oPRC=0.*oLT=1.*",
     };
-}
\ No newline at end of file
+}

From b5f2383c1c7115d1209289e722066e2807b1f889 Mon Sep 17 00:00:00 2001
From: Gleb Kazantaev <gleb.kazantaev@intel.com>
Date: Fri, 18 Jun 2021 17:27:12 +0300
Subject: [PATCH 135/178] Add std::shared_ptr<Variable> attribute comparison
 support (#6222)

* Add std::shared_ptr<Variable> attribute comparision support

* Added unit test
---
 .../transformations/low_latency_test.cpp      | 12 ++++++---
 .../common_test_utils/ngraph_test_utils.cpp   |  4 +++
 .../common_test_utils/ngraph_test_utils.hpp   | 27 +++++++++++++++++--
 .../core/include/ngraph/op/util/variable.hpp  |  6 +++++
 ngraph/test/type_prop/assign.cpp              | 23 ++++++++++++++++
 5 files changed, 67 insertions(+), 5 deletions(-)

diff --git a/inference-engine/tests/functional/inference_engine/transformations/low_latency_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/low_latency_test.cpp
index 2d1594d38411db..6e151972e47cdb 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/low_latency_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/low_latency_test.cpp
@@ -25,12 +25,18 @@ TEST(TransformationTests, LowLatencyLSTM) {
     std::shared_ptr<ngraph::Function> f(nullptr), f_ref(nullptr);
     {
         auto X = std::make_shared<opset6::Parameter>(element::f32, Shape{1, 1, 16});
+        X->set_friendly_name("X");
         auto H_init = std::make_shared<opset6::Parameter>(element::f32, Shape{1, 128});
+        H_init->set_friendly_name("H_init");
         auto C_init = std::make_shared<opset6::Parameter>(element::f32, Shape{1, 128});
+        C_init->set_friendly_name("C_init");
 
         auto Xi = std::make_shared<opset6::Parameter>(element::f32, Shape{1, 1, 16});
+        Xi->set_friendly_name("Xi");
         auto H_t = std::make_shared<opset6::Parameter>(element::f32, Shape{1, 128});
+        H_t->set_friendly_name("H_t");
         auto C_t = std::make_shared<opset6::Parameter>(element::f32, Shape{1, 128});
+        C_t->set_friendly_name("C_t");
 
         // Body
         auto axis = ngraph::opset6::Constant::create(ngraph::element::i64, ngraph::Shape{}, {0});
@@ -79,8 +85,8 @@ TEST(TransformationTests, LowLatencyLSTM) {
         auto H_t = std::make_shared<opset6::Parameter>(element::f32, Shape{1, 128});
         auto C_t = std::make_shared<opset6::Parameter>(element::f32, Shape{1, 128});
 
-        const std::string variable_name_H("LSTMTensorIterator/variable0");
-        const std::string variable_name_C("LSTMTensorIterator/variable1");
+        const std::string variable_name_H("LSTMTensorIterator/H_t/variable_2");
+        const std::string variable_name_C("LSTMTensorIterator/C_t/variable_0");
         auto variable_H = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_H});
         auto variable_C = std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, variable_name_C});
         auto read_value_H = std::make_shared<opset6::ReadValue>(H_t, variable_H);
@@ -107,7 +113,7 @@ TEST(TransformationTests, LowLatencyLSTM) {
         assign_H->add_control_dependency(read_value_H);
         assign_C->add_control_dependency(read_value_C);
     }
-    auto res = compare_functions(f, f_ref);
+    auto res = compare_functions(f, f_ref, true, false, false, true, true);
     ASSERT_TRUE(res.first) << res.second;
 }
 
diff --git a/inference-engine/tests/ie_test_utils/common_test_utils/ngraph_test_utils.cpp b/inference-engine/tests/ie_test_utils/common_test_utils/ngraph_test_utils.cpp
index eee6e9613c8ce3..3d0ec47531d326 100644
--- a/inference-engine/tests/ie_test_utils/common_test_utils/ngraph_test_utils.cpp
+++ b/inference-engine/tests/ie_test_utils/common_test_utils/ngraph_test_utils.cpp
@@ -803,6 +803,8 @@ void ReadAndStoreAttributes::on_adapter(const std::string& name, ngraph::ValueAc
         insert(name, storage::MemoryChunk{storage::MemoryChunk::Data(beg, end)});
     } else if (auto framework_node_attr = ngraph::as_type<ngraph::AttributeAdapter<ngraph::op::FrameworkNodeAttrs>>(&adapter)) {
         insert(name, framework_node_attr->get());
+    } else if (auto variable_ptr = ngraph::as_type<ngraph::AttributeAdapter<std::shared_ptr<ngraph::Variable>>>(&adapter)) {
+        insert(name, variable_ptr->get());
     } else {
         m_read_result += "store   attr [ ERR ]: " + name +
                          " [drop `void` comparison which is '" + adapter.get_type_info().name +
@@ -882,6 +884,8 @@ void ReadAndCompareAttributes::verify_others(const std::string &name, ngraph::Va
         verify_mem_buf(name, a->get());
     } else if (auto attrs = ngraph::as_type<ngraph::AttributeAdapter<ngraph::op::FrameworkNodeAttrs>>(&adapter)) {
         verify(name, attrs->get());
+    } else if (auto variable_ptr = ngraph::as_type<ngraph::AttributeAdapter<std::shared_ptr<ngraph::Variable>>>(&adapter)) {
+        verify(name, variable_ptr->get());
     } else {
         m_cmp_result += "compare attr [ ERR ]: " + name +
                         " [drop `void` comparison which is '" + adapter.get_type_info().name +
diff --git a/inference-engine/tests/ie_test_utils/common_test_utils/ngraph_test_utils.hpp b/inference-engine/tests/ie_test_utils/common_test_utils/ngraph_test_utils.hpp
index 5e7220b7872f0f..ab636060af1264 100644
--- a/inference-engine/tests/ie_test_utils/common_test_utils/ngraph_test_utils.hpp
+++ b/inference-engine/tests/ie_test_utils/common_test_utils/ngraph_test_utils.hpp
@@ -322,7 +322,8 @@ class Storage : private AttributeStorage<MemoryChunk>,
                 private AttributeStorage<std::shared_ptr<ngraph::Function>>,
                 private AttributeStorage<SubGraphOpInputDescription>,
                 private AttributeStorage<SubGraphOpOutputDescription>,
-                private AttributeStorage<ngraph::op::FrameworkNodeAttrs> {
+                private AttributeStorage<ngraph::op::FrameworkNodeAttrs>,
+                private AttributeStorage<std::shared_ptr<ngraph::Variable>> {
 public:
     template <typename AttrValue>
     const AttributeStorage<AttrValue>& storage() const {
@@ -361,7 +362,8 @@ class Storage : private AttributeStorage<MemoryChunk>,
                storage<std::shared_ptr<ngraph::Function>>().get_attributes_number() +
                storage<SubGraphOpInputDescription>().get_attributes_number() +
                storage<SubGraphOpOutputDescription>().get_attributes_number() +
-               storage<ngraph::op::FrameworkNodeAttrs>().get_attributes_number();
+               storage<ngraph::op::FrameworkNodeAttrs>().get_attributes_number() +
+               storage<std::shared_ptr<ngraph::Variable>>().get_attributes_number();
     }
 };
 
@@ -562,6 +564,14 @@ struct Equal<SubGraphOpOutputDescription> {
     }
 };
 
+template <>
+struct Equal<std::shared_ptr<ngraph::Variable>> {
+    static bool equal_value(
+            const std::shared_ptr<ngraph::Variable>& lhs, const std::shared_ptr<ngraph::Variable>& rhs) {
+        return lhs->get_info() == rhs->get_info();
+    }
+};
+
 template <>
 struct Equal<uint8_t*> {
     static constexpr uint8_t BITS_IN_BYTE_COUNT = 8;
@@ -705,6 +715,19 @@ struct Get<ngraph::op::FrameworkNodeAttrs, void> {
     }
 };
 
+template <>
+struct Get<std::shared_ptr<ngraph::Variable>, void> {
+    static std::string value(const std::shared_ptr<ngraph::Variable>& variable) {
+        std::stringstream oss;
+        const auto variable_info = variable->get_info();
+        oss << "[";
+        oss << "data_shape=" << variable_info.data_shape << ", ";
+        oss << "data_type=" << variable_info.data_type << ", ";
+        oss << "variable_id=" << variable_info.variable_id;
+        oss << "]";
+        return oss.str();
+    }
+};
 
 }  // namespace str
 
diff --git a/ngraph/core/include/ngraph/op/util/variable.hpp b/ngraph/core/include/ngraph/op/util/variable.hpp
index 6d0f48dafd40b6..3390de9026e3d2 100644
--- a/ngraph/core/include/ngraph/op/util/variable.hpp
+++ b/ngraph/core/include/ngraph/op/util/variable.hpp
@@ -14,6 +14,12 @@ namespace ngraph
         PartialShape data_shape;
         element::Type data_type;
         std::string variable_id;
+
+        inline bool operator==(const VariableInfo& other) const
+        {
+            return data_shape == other.data_shape && data_type == other.data_type &&
+                   variable_id == other.variable_id;
+        }
     };
 
     class NGRAPH_API Variable
diff --git a/ngraph/test/type_prop/assign.cpp b/ngraph/test/type_prop/assign.cpp
index 9f8169ae71f3c2..996231175e3e77 100644
--- a/ngraph/test/type_prop/assign.cpp
+++ b/ngraph/test/type_prop/assign.cpp
@@ -64,3 +64,26 @@ TEST(type_prop, assign_read_value_new_shape)
     ASSERT_EQ(variable->get_info().data_type, element::f16);
     ASSERT_EQ(variable->get_info().data_shape, (PartialShape{3, {4, 5}, 8}));
 }
+
+TEST(type_prop, variable_comparison)
+{
+    auto variable1 =
+        std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "ID"});
+
+    auto variable2 =
+        std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "ID"});
+
+    auto variable3 =
+        std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::dynamic, "ID1"});
+
+    auto variable4 =
+        std::make_shared<Variable>(VariableInfo{PartialShape::dynamic(), element::f32, "ID"});
+
+    auto variable5 =
+        std::make_shared<Variable>(VariableInfo{Shape{1}, element::dynamic, "ID"});
+
+    ASSERT_TRUE(variable1->get_info() == variable2->get_info());
+    ASSERT_FALSE(variable1->get_info() == variable3->get_info());
+    ASSERT_FALSE(variable1->get_info() == variable4->get_info());
+    ASSERT_FALSE(variable1->get_info() == variable5->get_info());
+}
\ No newline at end of file

From 020af3ddddc371e12bc591b81157a85cd62d6dd9 Mon Sep 17 00:00:00 2001
From: Alexander Shchepetov <alexander.shchepetov@intel.com>
Date: Fri, 18 Jun 2021 17:36:08 +0300
Subject: [PATCH 136/178] add TF Roll op test (#6228)

---
 .../tensorflow_tests/test_tf_Roll.py          | 42 +++++++++++++++++++
 1 file changed, 42 insertions(+)
 create mode 100644 tests/layer_tests/tensorflow_tests/test_tf_Roll.py

diff --git a/tests/layer_tests/tensorflow_tests/test_tf_Roll.py b/tests/layer_tests/tensorflow_tests/test_tf_Roll.py
new file mode 100644
index 00000000000000..340adcc1eca60a
--- /dev/null
+++ b/tests/layer_tests/tensorflow_tests/test_tf_Roll.py
@@ -0,0 +1,42 @@
+import pytest
+import tensorflow as tf
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestTFRoll(CommonTFLayerTest):
+    def create_tf_roll_net(self, shift, axis, x_shape, input_type, ir_version):
+        tf.compat.v1.reset_default_graph()
+
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            tf_x_shape = x_shape.copy()
+            # reshaping
+            if len(tf_x_shape) >= 3:
+                tf_x_shape.append(tf_x_shape.pop(1))
+
+            x = tf.compat.v1.placeholder(input_type, tf_x_shape, 'Input')
+            roll = tf.roll(x, shift=shift, axis=axis)
+
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        # TODO: add reference IR net. Now it is omitted and tests only inference result that is more important
+        ref_net = None
+
+        return tf_net, ref_net
+
+    test_data = [dict(shift=[1], axis=[-1], x_shape=[4, 3], input_type=tf.float32),
+                 dict(shift=[1, 5, -7], axis=[0, 1, 1], x_shape=[2, 3, 5], input_type=tf.float16),
+                 dict(shift=[11, -8], axis=[-1, -2], x_shape=[3, 4, 3, 1], input_type=tf.int32),
+                 dict(shift=[7, -2, 5], axis=[0, -1, -1], x_shape=[5, 2, 3, 7], input_type=tf.int64),
+                 dict(shift=[3, 7], axis=[0, 1], x_shape=[2, 4, 3, 5, 4], input_type=tf.half),
+                 pytest.param(dict(shift=[1, -2], axis=[0, 1], x_shape=[2, 4, 3, 5], input_type=tf.float32),
+                              marks=pytest.mark.precommit)]
+
+    @pytest.mark.parametrize("params", test_data)
+    @pytest.mark.nightly
+    def test_tf_roll(self, params, ie_device, precision, ir_version, temp_dir):
+        if ie_device == 'GPU':
+            pytest.skip("Roll is not supported on GPU")
+        self._test(*self.create_tf_roll_net(**params, ir_version=ir_version), ie_device, precision,
+                   temp_dir=temp_dir, ir_version=ir_version, **params)

From ac01777166c03948b9b378655ad22bd37dfd8850 Mon Sep 17 00:00:00 2001
From: cecilia peng <cecilia.peng@intel.com>
Date: Fri, 18 Jun 2021 22:39:23 +0800
Subject: [PATCH 137/178] MatrixNMS-8 spec. (#5948)

* Specification of a new operator called MatrixNMS.

This Op functionally equivalent to NonMaxSuppression-5, with an non max suppression algorithm called matrix_nms, and perform more post-processing phases, and lay out
the detection outputs in the way of PaddlePaddle detection.

* Update docs/ops/sort/MatrixNMS_8.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Update docs/ops/sort/MatrixNMS_8.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Update docs/ops/sort/MatrixNMS_8.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Update docs/ops/sort/MatrixNMS_8.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Update docs/ops/sort/MatrixNMS_8.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Update docs/ops/sort/MatrixNMS_8.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Update docs/ops/sort/MatrixNMS_8.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Update docs/ops/sort/MatrixNMS_8.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Update MatrixNMS_8.md

* Update MatrixNMS_8.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
---
 docs/doxygen/ie_docs.xml     |   1 +
 docs/ops/sort/MatrixNMS_8.md | 168 +++++++++++++++++++++++++++++++++++
 2 files changed, 169 insertions(+)
 create mode 100644 docs/ops/sort/MatrixNMS_8.md

diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index faabfab9238054..120492baef0c2d 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -189,6 +189,7 @@ limitations under the License.
                 <tab type="user" title="MVN-1" url="@ref openvino_docs_ops_normalization_MVN_1"/>
                 <tab type="user" title="MVN-6" url="@ref openvino_docs_ops_normalization_MVN_6"/>
                 <tab type="user" title="MatMul-1" url="@ref openvino_docs_ops_matrix_MatMul_1"/>
+                <tab type="user" title="MatrixNonMaxSuppression-8" url="@ref openvino_docs_ops_sort_MatrixNonMaxSuppression_8"/>
                 <tab type="user" title="MaxPool-1" url="@ref openvino_docs_ops_pooling_MaxPool_1"/>
                 <tab type="user" title="Maximum-1" url="@ref openvino_docs_ops_arithmetic_Maximum_1"/>
                 <tab type="user" title="Minimum-1" url="@ref openvino_docs_ops_arithmetic_Minimum_1"/>
diff --git a/docs/ops/sort/MatrixNMS_8.md b/docs/ops/sort/MatrixNMS_8.md
new file mode 100644
index 00000000000000..d3a422307223c9
--- /dev/null
+++ b/docs/ops/sort/MatrixNMS_8.md
@@ -0,0 +1,168 @@
+##  MatrixNonMaxSuppression<a name="MatrixNonMaxSuppression"></a> {#openvino_docs_ops_sort_MatrixNms_8}
+
+**Versioned name**: *MatrixNonMaxSuppression-8*
+
+**Category**: *Sorting and maximization*
+
+**Short description**: *MatrixNonMaxSuppression* performs matrix non-maximum suppression (NMS) of the boxes with predicted scores.
+
+**Detailed description**: The operation performs the following:
+
+1. Selects candidate bounding boxes with scores higher than `score_threshold`.
+2. For each class, selects at most `nms_top_k` candidate boxes.
+3. Decays scores of the candidate boxes according to the Matrix NMS algorithm [Wang et al](https://arxiv.org/abs/2003.10152.pdf). This algorithm is applied independently to each class and each batch element. Boxes of `background_class` are skipped and thus eliminated during the process.
+4. Selects boxes with the decayed scores higher than `post_threshold`, and selects at most `keep_top_k` scoring candidate boxes per batch element.
+
+The Matrix NMS algorithm is described below:
+1.  Sort descending the candidate boxes by score, and compute `n*n` pairwise IOU (IntersectionOverUnion) matrix `X` for the top `n` boxes. Suppose `n` is the number of candidate boxes.
+2.  Set the lower triangle and diagonal of `X` to 0. Therefore get the upper triangular matrix `X`.
+3.  Take the column-wise max of `X` to compute a vector `K` of maximum IOU for each candidate box.
+4.  Repeat element value of `K` along axis 1. Suppose this gets a matrix `X_cmax`.
+5.  Compute the decay factor: `decay_factor = exp((X_cmax**2 - X**2) * gaussian_sigma)` if `decay_function` is `guassian`, else `decay_factor = (1 - X) / (1 - X_cmax)`.
+6.  Take the column-wise min of `decay_factor`, and element-wise multiply with scores to decay them.
+
+**Attributes**:
+
+* *sort_result*
+
+  * **Description**: *sort_result* specifies the order of output elements.
+  * **Range of values**: `class`, `score`, `none`
+    * *class* - sort selected boxes by class id (ascending).
+    * *score* - sort selected boxes by score (descending).
+    * *none* - do not guarantee the order.
+  * **Type**: `string`
+  * **Default value**: `none`
+  * **Required**: *No*
+
+* *sort_result_across_batch*
+
+  * **Description**: *sort_result_across_batch* is a flag that specifies whenever it is necessary to sort selected boxes across batches or not.
+  * **Range of values**: true or false
+    * *true* - sort selected boxes across batches.
+    * *false* - do not sort selected boxes across batches (boxes are sorted per batch element).
+  * **Type**: boolean
+  * **Default value**: false
+  * **Required**: *No*
+
+* *output_type*
+
+  * **Description**: the tensor type of outputs `selected_indices` and `valid_outputs`.
+  * **Range of values**: `i64` or `i32`
+  * **Type**: `string`
+  * **Default value**: `i64`
+  * **Required**: *No*
+
+* *score_threshold*
+
+  * **Description**: minimum score to consider box for the processing.
+  * **Range of values**: a floating-point number
+  * **Type**: `float`
+  * **Default value**: `0`
+  * **Required**: *No*
+
+* *nms_top_k*
+
+  * **Description**: maximum number of boxes to be selected per class.
+  * **Range of values**: an integer
+  * **Type**: `int`
+  * **Default value**: `-1` meaning to keep all boxes
+  * **Required**: *No*
+
+* *keep_top_k*
+
+  * **Description**: maximum number of boxes to be selected per batch element.
+  * **Range of values**: an integer
+  * **Type**: `int`
+  * **Default value**: `-1` meaning to keep all boxes
+  * **Required**: *No*
+
+* *background_class*
+
+  * **Description**: the background class id.
+  * **Range of values**: an integer
+  * **Type**: `int`
+  * **Default value**: `-1` meaning to keep all classes
+  * **Required**: *No*
+
+* *decay_function*
+
+  * **Description**: decay function used to decay scores.
+  * **Range of values**: `gaussian`, `linear`
+  * **Type**: `string`
+  * **Default value**: `linear`
+  * **Required**: *No*
+
+* *gaussian_sigma*
+
+  * **Description**: gaussian_sigma parameter for gaussian decay_function.
+  * **Range of values**: a floating-point number
+  * **Type**: `float`
+  * **Default value**: `2.0`
+  * **Required**: *No*
+
+* *post_threshold*
+
+  * **Description**: threshold to filter out boxes with low confidence score after decaying.
+  * **Range of values**: a floating-point number
+  * **Type**: `float`
+  * **Default value**: `0`
+  * **Required**: *No*
+
+**Inputs**:
+
+*   **1**: `boxes` - tensor of type *T* and shape `[num_batches, num_boxes, 4]` with box coordinates. The box cooridnates are layout as `[xmin, ymin, xmax, ymax]`. **Required.**
+
+*   **2**: `scores` - tensor of type *T* and shape `[num_batches, num_classes, num_boxes]` with box scores. **Required.**
+
+**Outputs**:
+
+*   **1**: `selected_outputs` - tensor of type *T_THRESHOLDS* and shape `[number of selected boxes, 6]` containing the selected boxes with score and class as tuples `[class_id, box_score, xmin, ymin, xmax, ymax]`.
+
+*   **2**: `selected_indices` - tensor of type *T_IND* and shape `[number of selected boxes, 1]` the selected indices in the flattened input `boxes`, which are absolute values cross batches. Therefore possible valid values are in the range `[0, num_batches * num_boxes - 1]`.
+
+*   **3**: `selected_num` - 1D tensor of type *T_IND* and shape `[num_batches]` representing the number of selected boxes for each batch element.
+
+When there is no box selected, `selected_num` is filled with `0`. `selected_outputs` is an empty tensor of shape `[0, 6]`, and `selected_indices` is an empty tensor of shape `[0, 1]`.
+
+**Types**
+
+* *T*: floating point type.
+
+* *T_MAX_BOXES*: integer type.
+
+* *T_THRESHOLDS*: floating point type.
+
+* *T_IND*: `int64` or `int32`.
+
+**Example**
+
+```xml
+<layer ... type="MatrixNonMaxSuppression" ... >
+    <data decay_function="guassian" sort_result="score" output_type="i64"/>
+    <input>
+        <port id="0">
+            <dim>3</dim>
+            <dim>100</dim>
+            <dim>4</dim>
+        </port>
+        <port id="1">
+            <dim>3</dim>
+            <dim>5</dim>
+            <dim>100</dim>
+        </port>
+    </input>
+    <output>
+        <port id="5" precision="FP32">
+            <dim>-1</dim> <!-- "-1" means a undefined dimension calculated during the model inference -->
+            <dim>6</dim>
+        </port>
+        <port id="6" precision="I64">
+            <dim>-1</dim>
+            <dim>1</dim>         
+        </port>
+        <port id="7" precision="I64">
+            <dim>3</dim>
+        </port>
+    </output>
+</layer>
+```

From 72c8743a7128488aab77e7c9c662b3f5aa4e1381 Mon Sep 17 00:00:00 2001
From: Gleb Kazantaev <gleb.kazantaev@intel.com>
Date: Fri, 18 Jun 2021 17:59:43 +0300
Subject: [PATCH 138/178] Add NGRAPH_UNIT_TEST_ENABLE cmake key to control
 backends build in unit-test target (#6232)

---
 ngraph/CMakeLists.txt      | 2 ++
 ngraph/test/CMakeLists.txt | 6 +++---
 2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/ngraph/CMakeLists.txt b/ngraph/CMakeLists.txt
index a05ca3b15c624d..55e218daa2c924 100644
--- a/ngraph/CMakeLists.txt
+++ b/ngraph/CMakeLists.txt
@@ -84,6 +84,7 @@ if (APPLE)
 endif()
 
 option(NGRAPH_UNIT_TEST_ENABLE "Control the building of unit tests" ON)
+option(NGRAPH_UNIT_TEST_BACKENDS_ENABLE "Control the building of unit tests using backends" ON)
 option(NGRAPH_INTERPRETER_ENABLE "Control the building of the INTERPRETER backend" ON)
 option(NGRAPH_DEBUG_ENABLE "Enable output for NGRAPH_DEBUG statements" OFF)
 option(NGRAPH_ONNX_IMPORT_ENABLE "Enable ONNX importer" OFF)
@@ -117,6 +118,7 @@ message(STATUS "NGRAPH_THREAD_SANITIZER_ENABLE:       ${NGRAPH_THREAD_SANITIZER_
 message(STATUS "NGRAPH_UB_SANITIZER_ENABLE:           ${NGRAPH_UB_SANITIZER_ENABLE}")
 message(STATUS "NGRAPH_USE_PROTOBUF_LITE:             ${NGRAPH_USE_PROTOBUF_LITE}")
 message(STATUS "NGRAPH_UNIT_TEST_ENABLE:              ${NGRAPH_UNIT_TEST_ENABLE}")
+message(STATUS "NGRAPH_UNIT_TEST_BACKENDS_ENABLE:     ${NGRAPH_UNIT_TEST_BACKENDS_ENABLE}")
 
 # Setup CMAKE_ARGS to be forwarded to External Projects
 set(NGRAPH_FORWARD_CMAKE_ARGS
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index fcb4f013775139..278121669a83c1 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -304,7 +304,7 @@ set(SRC
 set_source_files_properties(includes.cpp PROPERTIES COMPILE_DEFINITIONS
     NGRAPH_INCLUDES="${PROJECT_SOURCE_DIR}/src/ngraph")
 
-if (ENABLE_MKL_DNN)
+if (ENABLE_MKL_DNN AND NGRAPH_UNIT_TEST_BACKENDS_ENABLE)
     message(STATUS "NGRAPH_TESTS: IE:CPU enabled")
     set(ACTIVE_BACKEND_LIST ${ACTIVE_BACKEND_LIST} "IE:CPU")
     if (ENABLE_STRICT_DEPENDENCIES)
@@ -314,7 +314,7 @@ if (ENABLE_MKL_DNN)
     endif()
 endif()
 
-if (ENABLE_CLDNN)
+if (ENABLE_CLDNN AND NGRAPH_UNIT_TEST_BACKENDS_ENABLE)
     message(STATUS "NGRAPH_TESTS: IE:GPU enabled")
     set(ACTIVE_BACKEND_LIST ${ACTIVE_BACKEND_LIST} "IE:GPU")
     if (ENABLE_STRICT_DEPENDENCIES)
@@ -324,7 +324,7 @@ if (ENABLE_CLDNN)
     endif()
 endif()
 
-if (NGRAPH_INTERPRETER_ENABLE)
+if (NGRAPH_INTERPRETER_ENABLE AND NGRAPH_UNIT_TEST_BACKENDS_ENABLE)
     list(APPEND SRC
         builder.cpp
         backend_api.cpp)

From 7e664119281eb60d24dd30ea39a8b163404ab5db Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Sun, 20 Jun 2021 13:01:32 +0300
Subject: [PATCH 139/178] Remove deprecated api (#6256)

* Revert "Fixed creation of C++ wrappers from old API (#5805)"

This reverts commit ffe03b6ed35a6cf5912282ee0dc08dfc59555acf.

* Removed deprecated API

* Fixes 2
---
 .../include/cpp/ie_executable_network.hpp     |  15 --
 .../include/cpp/ie_infer_request.hpp          |  16 --
 .../include/cpp/ie_memory_state.hpp           |  48 +---
 .../include/ie_iexecutable_network.hpp        |  17 --
 .../include/ie_iinfer_request.hpp             |  16 --
 inference-engine/include/ie_imemory_state.hpp |  95 --------
 inference-engine/include/ie_parameter.hpp     |  42 ----
 inference-engine/include/ie_unicode.hpp       |  69 ------
 .../cpp/ie_executable_network.cpp             | 103 +--------
 .../cpp/ie_executable_network_base.hpp        |  20 --
 .../cpp/ie_infer_async_request_base.hpp       |  19 +-
 .../inference_engine/cpp/ie_infer_request.cpp | 205 +-----------------
 .../cpp/ie_variable_state.cpp                 |  39 ----
 .../cpp/ie_variable_state_base.hpp            |  59 -----
 .../interface/ie_ivariable_state_internal.cpp |   3 -
 .../src/inference_engine/ie_parameter.cpp     |  45 ----
 .../interface/ie_ivariable_state_internal.hpp |   8 -
 .../async_infer_request_test.cpp              |  14 --
 .../inference_engine/executable_network.cpp   |  35 +--
 .../inference_engine/variable_state.cpp       |  20 --
 ...ntization_during_memory_requantization.cpp |   4 +-
 .../src/subgraph/memory_LSTMCell.cpp          |   4 +-
 .../subgraph/negative_memory_layer_offset.cpp |   2 +-
 .../ie_test_utils/unit_test_utils/mock.cpp    |   1 -
 .../mocks/mock_ie_ivariable_state.hpp         |  25 ---
 .../mocks/mock_iexecutable_network.hpp        |   1 -
 .../mocks/mock_iinfer_request.hpp             |   1 -
 .../ie_executable_network_test.cpp            |   1 -
 28 files changed, 30 insertions(+), 897 deletions(-)
 delete mode 100644 inference-engine/include/ie_imemory_state.hpp
 delete mode 100644 inference-engine/include/ie_unicode.hpp
 delete mode 100644 inference-engine/src/inference_engine/cpp/ie_variable_state_base.hpp
 delete mode 100644 inference-engine/src/inference_engine/ie_parameter.cpp
 delete mode 100644 inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_ie_ivariable_state.hpp

diff --git a/inference-engine/include/cpp/ie_executable_network.hpp b/inference-engine/include/cpp/ie_executable_network.hpp
index eb1824f9da041e..81d5b10e7dd4be 100644
--- a/inference-engine/include/cpp/ie_executable_network.hpp
+++ b/inference-engine/include/cpp/ie_executable_network.hpp
@@ -32,9 +32,6 @@ class IExecutableNetworkInternal;
 class INFERENCE_ENGINE_API_CLASS(ExecutableNetwork) {
     details::SharedObjectLoader                  _so;
     std::shared_ptr<IExecutableNetworkInternal>  _impl;
-    IE_SUPPRESS_DEPRECATED_START
-    std::shared_ptr<IExecutableNetwork>          actual;
-    IE_SUPPRESS_DEPRECATED_END
 
     /**
      * @brief Constructs ExecutableNetwork from the initialized std::shared_ptr
@@ -51,18 +48,6 @@ class INFERENCE_ENGINE_API_CLASS(ExecutableNetwork) {
      */
     ExecutableNetwork() = default;
 
-    IE_SUPPRESS_DEPRECATED_START
-    /**
-     * @deprecated This ctor will be removed in 2022.1
-     * @brief Constructs ExecutableNetwork from the initialized std::shared_ptr
-     * @param exec Initialized shared pointer
-     * @param splg Plugin to use. This is required to ensure that ExecutableNetwork can work properly even if plugin object is destroyed.
-     */
-    INFERENCE_ENGINE_DEPRECATED("This ctor will be removed in 2022.1")
-    explicit ExecutableNetwork(std::shared_ptr<IExecutableNetwork> exec,
-                               std::shared_ptr<details::SharedObjectLoader> splg = {});
-    IE_SUPPRESS_DEPRECATED_END
-
     /**
      * @brief Gets the Executable network output Data node information.
      *
diff --git a/inference-engine/include/cpp/ie_infer_request.hpp b/inference-engine/include/cpp/ie_infer_request.hpp
index fd71bf18bc2cc1..c5d52ec6fc0905 100644
--- a/inference-engine/include/cpp/ie_infer_request.hpp
+++ b/inference-engine/include/cpp/ie_infer_request.hpp
@@ -35,10 +35,6 @@ class ICompletionCallbackWrapper;
 class INFERENCE_ENGINE_API_CLASS(InferRequest) {
     details::SharedObjectLoader                          _so;
     std::shared_ptr<IInferRequestInternal>               _impl;
-    IE_SUPPRESS_DEPRECATED_START
-    IInferRequest::Ptr                                   actual;
-    std::shared_ptr<details::ICompletionCallbackWrapper> callback;
-    IE_SUPPRESS_DEPRECATED_END
 
     /**
      * @brief Constructs InferRequest from the initialized std::shared_ptr
@@ -71,18 +67,6 @@ class INFERENCE_ENGINE_API_CLASS(InferRequest) {
      */
     InferRequest() = default;
 
-    IE_SUPPRESS_DEPRECATED_START
-    /**
-     * @deprecated This ctor will be removed in 2022.1
-     * @brief Constructs InferRequest from the initialized std::shared_ptr
-     * @param request Initialized shared pointer
-     * @param splg Plugin to use. This is required to ensure that InferRequest can work properly even if plugin object is destroyed.
-     */
-    INFERENCE_ENGINE_DEPRECATED("This ctor will be removed in 2022.1")
-    explicit InferRequest(IInferRequest::Ptr request,
-                          std::shared_ptr<details::SharedObjectLoader> splg = {});
-    IE_SUPPRESS_DEPRECATED_END
-
     /**
      * @brief Sets input/output data to infer
      *
diff --git a/inference-engine/include/cpp/ie_memory_state.hpp b/inference-engine/include/cpp/ie_memory_state.hpp
index 0c055cec40cc79..8d54f79f06ce6a 100644
--- a/inference-engine/include/cpp/ie_memory_state.hpp
+++ b/inference-engine/include/cpp/ie_memory_state.hpp
@@ -3,7 +3,7 @@
 //
 
 /**
- * @brief A header file that provides wrapper classes for IVariableState
+ * @brief A header file that provides VariableState
  *
  * @file ie_memory_state.hpp
  */
@@ -16,21 +16,17 @@
 #include "ie_api.h"
 #include "ie_blob.h"
 #include "details/ie_so_loader.h"
-#include "ie_imemory_state.hpp"
 
 namespace InferenceEngine {
 
 class IVariableStateInternal;
 
 /**
- * @brief C++ exception based error reporting wrapper of API class IVariableState
+ * @brief VariableState class
  */
 class INFERENCE_ENGINE_API_CLASS(VariableState) {
     details::SharedObjectLoader              _so;
     std::shared_ptr<IVariableStateInternal>  _impl;
-    IE_SUPPRESS_DEPRECATED_START
-    std::shared_ptr<IVariableState>          actual;
-    IE_SUPPRESS_DEPRECATED_END
 
     /**
      * @brief Constructs VariableState from the initialized std::shared_ptr
@@ -48,55 +44,27 @@ class INFERENCE_ENGINE_API_CLASS(VariableState) {
      */
     VariableState() = default;
 
-    IE_SUPPRESS_DEPRECATED_START
     /**
-     * @deprecated This ctor will be removed in 2022.1
-     * @brief constructs VariableState from the initialized std::shared_ptr
-     * @param pState Initialized shared pointer
-     * @param plg Optional: Plugin to use. This is required to ensure that VariableState can work properly even if plugin object is destroyed.
-     */
-    INFERENCE_ENGINE_DEPRECATED("This ctor will be removed in 2022.1")
-    explicit VariableState(std::shared_ptr<IVariableState> pState,
-                           std::shared_ptr<details::SharedObjectLoader> plg = {});
-    IE_SUPPRESS_DEPRECATED_END
-
-    /**
-     * @copybrief IVariableState::Reset
-     *
-     * Wraps IVariableState::Reset
+     * @brief Reset internal variable state for relevant infer request,
+     * to a value specified as default for according ReadValue node
      */
     void Reset();
 
     /**
-     * @copybrief IVariableState::GetName
-     *
-     * Wraps IVariableState::GetName
+     * @brief Gets name of current variable state, if length of array is not enough name is truncated by len, null
+     * terminator is inserted as well. As variable state name `variable_id` from according `ReadValue` used. 
      * @return A string representing a state name
      */
     std::string GetName() const;
 
     /**
-     * @copybrief IVariableState::GetState
-     *
-     * Wraps IVariableState::GetState
+     * @brief Returns the value of the variable state.
      * @return A blob representing a state
      */
     Blob::CPtr GetState() const;
 
     /**
-     * @copybrief IVariableState::GetLastState
-     * @deprecated Use IVariableState::SetState instead
-     *
-     * Wraps IVariableState::GetLastState
-     * @return A blob representing a last state 
-     */
-    INFERENCE_ENGINE_DEPRECATED("Use VariableState::GetState function instead")
-    Blob::CPtr GetLastState() const;
-
-    /**
-     * @copybrief IVariableState::SetState
-     *
-     * Wraps IVariableState::SetState
+     * @brief Sets the new state for the next inference.
      * @param state The current state to set
      */
     void SetState(Blob::Ptr state);
diff --git a/inference-engine/include/ie_iexecutable_network.hpp b/inference-engine/include/ie_iexecutable_network.hpp
index caef9bb95b95bd..bb0a6f71c4ac3f 100644
--- a/inference-engine/include/ie_iexecutable_network.hpp
+++ b/inference-engine/include/ie_iexecutable_network.hpp
@@ -18,7 +18,6 @@
 #include "ie_common.h"
 #include "ie_icnn_network.hpp"
 #include "ie_iinfer_request.hpp"
-#include "ie_imemory_state.hpp"
 #include "ie_input_info.hpp"
 #include "ie_parameter.hpp"
 #include "ie_remote_context.hpp"
@@ -113,22 +112,6 @@ class INFERENCE_ENGINE_DEPRECATED("Use InferenceEngine::ExecutableNetwork instea
     INFERENCE_ENGINE_DEPRECATED("Use InferenceEngine::ExecutableNetwork::GetExecGraphInfo instead")
     virtual StatusCode GetExecGraphInfo(ICNNNetwork::Ptr& graphPtr, ResponseDesc* resp) noexcept = 0;
 
-    /**
-     * @deprecated Use InferRequest::QueryState instead
-     * @brief Gets state control interface for given executable network.
-     *
-     * State control essential for recurrent networks
-     *
-     * @param pState reference to a pointer that receives internal states
-     * @param idx requested index for receiving memory state
-     * @param resp Optional: pointer to an already allocated object to contain information in case of failure
-     * @return Status code of the operation: InferenceEngine::OK (0) for success, OUT_OF_BOUNDS (-6) no memory state for
-     * given index
-     */
-    INFERENCE_ENGINE_DEPRECATED("Use InferRequest::QueryState instead")
-    virtual StatusCode QueryState(IVariableState::Ptr& pState, size_t idx, ResponseDesc* resp) noexcept = 0;
-    IE_SUPPRESS_DEPRECATED_END
-
     /**
      * @brief Sets configuration for current executable network
      *
diff --git a/inference-engine/include/ie_iinfer_request.hpp b/inference-engine/include/ie_iinfer_request.hpp
index 7d762d96a11305..4fd200c0252062 100644
--- a/inference-engine/include/ie_iinfer_request.hpp
+++ b/inference-engine/include/ie_iinfer_request.hpp
@@ -17,7 +17,6 @@
 #include "ie_blob.h"
 #include "ie_common.h"
 #include "ie_preprocess.hpp"
-#include "ie_imemory_state.hpp"
 
 namespace InferenceEngine {
 
@@ -195,21 +194,6 @@ class INFERENCE_ENGINE_DEPRECATED("Use InferenceEngine::InferRequest C++ wrapper
      */
     virtual InferenceEngine::StatusCode SetBatch(int batch_size, ResponseDesc* resp) noexcept = 0;
 
-    IE_SUPPRESS_DEPRECATED_START
-    /**
-     * @brief Gets state control interface for given infer request.
-     *
-     * State control essential for recurrent networks
-     *
-     * @param pState reference to a pointer that receives internal states
-     * @param idx requested index for receiving memory state
-     * @param resp Optional: pointer to an already allocated object to contain information in case of failure
-     * @return Status code of the operation: InferenceEngine::OK (0) for success, OUT_OF_BOUNDS (-6) no memory state for
-     * given index
-     */
-    virtual StatusCode QueryState(IVariableState::Ptr& pState, size_t idx, ResponseDesc* resp) noexcept = 0;
-    IE_SUPPRESS_DEPRECATED_END
-
 protected:
     ~IInferRequest() = default;
 };
diff --git a/inference-engine/include/ie_imemory_state.hpp b/inference-engine/include/ie_imemory_state.hpp
deleted file mode 100644
index 7f3ef99cbd11d3..00000000000000
--- a/inference-engine/include/ie_imemory_state.hpp
+++ /dev/null
@@ -1,95 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-/**
- * @brief a header file for IVariableState interface
- *
- * @file ie_imemory_state.hpp
- */
-
-#pragma once
-
-#include <memory>
-
-#include "ie_blob.h"
-#include "ie_common.h"
-
-namespace InferenceEngine {
-
-/**
- * @deprecated Use InferenceEngine::VariableState C++ wrapper instead
- * @interface IVariableState
- * @brief Manages data for reset operations
- */
-class INFERENCE_ENGINE_DEPRECATED("InferenceEngine::") IVariableState {
-public:
-    IE_SUPPRESS_DEPRECATED_START
-    /**
-     * @brief A shared pointer to the IVariableState interface
-     */
-    using Ptr = std::shared_ptr<IVariableState>;
-    IE_SUPPRESS_DEPRECATED_END
-
-    /**
-     * @brief Gets name of current variable state, if length of array is not enough name is truncated by len, null
-     * terminator is inserted as well. As variable state name `variable_id` from according `ReadValue` used. 
-     *
-     * @param name preallocated buffer for receiving name
-     * @param len Length of the buffer
-     * @param  resp Optional: pointer to an already allocated object to contain information in case of failure
-     * @return Status code of the operation: InferenceEngine::OK (0) for success
-     */
-    virtual StatusCode GetName(char* name, size_t len, ResponseDesc* resp) const noexcept = 0;
-
-    /**
-     * @brief Reset internal variable state for relevant infer request, to a value specified as default for according ReadValue node
-     *
-     * @param  resp Optional: pointer to an already allocated object to contain information in case of failure
-     * @return Status code of the operation: InferenceEngine::OK (0) for success*
-     */
-    virtual StatusCode Reset(ResponseDesc* resp) noexcept = 0;
-
-    /**
-     * @brief  Sets the new state for the next inference.
-     *
-     * This method can fail if Blob size does not match the internal state size or precision
-     *
-     * @param  newState The data to use as new state
-     * @param  resp Optional: pointer to an already allocated object to contain information in case of failure
-     * @return Status code of the operation: InferenceEngine::OK (0) for success
-     */
-    virtual StatusCode SetState(Blob::Ptr newState, ResponseDesc* resp) noexcept = 0;
-
-    /**
-     * @brief Returns the value of the variable state.
-     *
-     * @param state A reference to a blob containing a variable state
-     * @param  resp Optional: pointer to an already allocated object to contain information in case of failure
-     * @return Status code of the operation: InferenceEngine::OK (0) for success
-     */
-    INFERENCE_ENGINE_DEPRECATED("Use GetState function instead")
-    virtual StatusCode GetLastState(Blob::CPtr& state, ResponseDesc* resp) const noexcept {
-        return GetState(state, resp);
-    }
-
-    /**
-     * @brief Returns the value of the variable state.
-     *
-     * @param state A reference to a blob containing a variable state
-     * @param  resp Optional: pointer to an already allocated object to contain information in case of failure
-     * @return Status code of the operation: InferenceEngine::OK (0) for success
-     */
-    virtual StatusCode GetState(Blob::CPtr& state, ResponseDesc* resp) const noexcept = 0;
-};
-
-IE_SUPPRESS_DEPRECATED_START
-
-/**
- * @brief For compatibility reasons.
- */
-using IMemoryState = IVariableState;
-
-IE_SUPPRESS_DEPRECATED_END
-
-}  // namespace InferenceEngine
\ No newline at end of file
diff --git a/inference-engine/include/ie_parameter.hpp b/inference-engine/include/ie_parameter.hpp
index 683f02fd6a552f..4aa6760d474874 100644
--- a/inference-engine/include/ie_parameter.hpp
+++ b/inference-engine/include/ie_parameter.hpp
@@ -49,26 +49,6 @@ class INFERENCE_ENGINE_API_CLASS(Parameter) {
         std::swap(ptr, parameter.ptr);
     }
 
-    /**
-     * @deprecated Use ngraph::Variant directly
-     * @brief Creates parameter from variant.
-     * This method creates empty parameter if variant doesn't contain Parameter
-     *
-     * @param var ngraph variant
-     */
-    INFERENCE_ENGINE_DEPRECATED("Use ngraph::Variant directly")
-    Parameter(const std::shared_ptr<ngraph::Variant>& var);
-
-    /**
-     * @deprecated Use ngraph::Variant directly
-     * @brief Creates parameter from variant.
-     * This method creates empty parameter if variant doesn't contain Parameter
-     *
-     * @param var ngraph variant
-     */
-    INFERENCE_ENGINE_DEPRECATED("Use ngraph::Variant directly")
-    Parameter(std::shared_ptr<ngraph::Variant>& var);
-
     /**
      * @brief Copy constructor
      *
@@ -204,28 +184,6 @@ class INFERENCE_ENGINE_API_CLASS(Parameter) {
         return dyn_cast<typename std::remove_cv<T>::type>(ptr);
     }
 
-    /**
-     * @deprecated Use ngraph::Variant directly
-     * @brief Converts parameter to shared pointer on ngraph::Variant
-     *
-     * @return shared pointer on ngraph::Variant
-     */
-    INFERENCE_ENGINE_DEPRECATED("Use ngraph::Variant directly")
-    std::shared_ptr<ngraph::Variant> asVariant() const;
-
-    /**
-     * @deprecated Use ngraph::Variant directly
-     * @brief Casts to shared pointer on ngraph::Variant
-     *
-     * @return shared pointer on ngraph::Variant
-     */
-    INFERENCE_ENGINE_DEPRECATED("Use ngraph::Variant directly")
-    operator std::shared_ptr<ngraph::Variant>() const {
-        IE_SUPPRESS_DEPRECATED_START
-        return asVariant();
-        IE_SUPPRESS_DEPRECATED_END
-    }
-
     /**
      * Dynamic cast to specified type
      * @tparam T type
diff --git a/inference-engine/include/ie_unicode.hpp b/inference-engine/include/ie_unicode.hpp
deleted file mode 100644
index dc943d6f558be5..00000000000000
--- a/inference-engine/include/ie_unicode.hpp
+++ /dev/null
@@ -1,69 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-/**
- * @brief This is a header file with common inference engine definitions
- * 
- * @file ie_unicode.hpp
- */
-#pragma once
-
-#include <algorithm>
-#include <cstdlib>
-#include <memory>
-#include <ostream>
-#include <string>
-#include <vector>
-
-#ifdef UNICODE
-typedef wchar_t tchar;
-typedef std::wstring file_name_t;
-#else
-typedef char tchar;
-typedef std::string file_name_t;
-#endif
-
-namespace InferenceEngine {
-
-/**
- * @deprecated Use OS-native conversion utilities
- * @brief Conversion from possibly-wide character string to a single-byte chain.
- * @param str A possibly-wide character string
- * @return A single-byte character string
- */
-INFERENCE_ENGINE_DEPRECATED("Use OS-native conversion utilities")
-inline std::string fileNameToString(const file_name_t& str) {
-#ifdef UNICODE
-    size_t maxlen = (str.length() + 1) * sizeof(wchar_t) / sizeof(char);
-    std::vector<char> mbstr(maxlen);
-    mbstr[0] = 0;
-    std::wcstombs(&mbstr[0], str.c_str(), maxlen);
-    std::string res = std::string(&mbstr[0]);
-    return res;
-#else
-    return str;
-#endif
-}
-
-/**
- * @deprecated Use OS-native conversion utilities
- * @brief Conversion from single-byte character string to a possibly-wide one
- * @param str A single-byte character string
- * @return A possibly-wide character string
- */
-INFERENCE_ENGINE_DEPRECATED("Use OS-native conversion utilities")
-inline file_name_t stringToFileName(const std::string& str) {
-#ifdef UNICODE
-    size_t maxlen = str.length() + 1;
-    std::vector<wchar_t> wcstr(maxlen);
-    wcstr[0] = 0;
-    std::mbstowcs(&wcstr[0], str.c_str(), maxlen);
-    file_name_t res = file_name_t(&wcstr[0]);
-    return res;
-#else
-    return str;
-#endif
-}
-
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/inference_engine/cpp/ie_executable_network.cpp b/inference-engine/src/inference_engine/cpp/ie_executable_network.cpp
index 6de781d11e62ec..a4afee5a28ba2b 100644
--- a/inference-engine/src/inference_engine/cpp/ie_executable_network.cpp
+++ b/inference-engine/src/inference_engine/cpp/ie_executable_network.cpp
@@ -25,47 +25,15 @@ ExecutableNetwork::ExecutableNetwork(const details::SharedObjectLoader&      so,
 
 IE_SUPPRESS_DEPRECATED_START
 
-ExecutableNetwork::ExecutableNetwork(IExecutableNetwork::Ptr exec,
-                                     std::shared_ptr<details::SharedObjectLoader> splg)
-    : _so(), _impl(), actual(exec) {
-    if (splg) {
-        _so = *splg;
-    }
-
-    //  plg can be null, but not the actual
-    if (actual == nullptr)
-        IE_THROW(NotAllocated) << "ExecutableNetwork was not initialized.";
-}
-
 ConstOutputsDataMap ExecutableNetwork::GetOutputsInfo() const {
-    if (actual) {
-        ConstOutputsDataMap data;
-        CALL_STATUS_FNC(GetOutputsInfo, data);
-        return data;
-    }
-
     EXEC_NET_CALL_STATEMENT(return _impl->GetOutputsInfo());
 }
 
 ConstInputsDataMap ExecutableNetwork::GetInputsInfo() const {
-    if (actual) {
-        ConstInputsDataMap info;
-        CALL_STATUS_FNC(GetInputsInfo, info);
-        return info;
-    }
-
     EXEC_NET_CALL_STATEMENT(return _impl->GetInputsInfo());
 }
 
 void ExecutableNetwork::reset(IExecutableNetwork::Ptr newActual) {
-    if (actual) {
-        if (newActual == nullptr) {
-            THROW_IE_EXCEPTION << "ExecutableNetwork wrapper used for reset was not initialized.";
-        }
-        this->actual.swap(newActual);
-        return;
-    }
-
     if (_impl == nullptr) IE_THROW() << "ExecutableNetwork was not initialized.";
     if (newActual == nullptr) IE_THROW() << "ExecutableNetwork wrapper used for reset was not initialized.";
     auto newBase = std::dynamic_pointer_cast<ExecutableNetworkBase>(newActual);
@@ -76,36 +44,10 @@ void ExecutableNetwork::reset(IExecutableNetwork::Ptr newActual) {
 }
 
 ExecutableNetwork::operator IExecutableNetwork::Ptr() {
-    if (actual) {
-        return actual;
-    }
-
     return std::make_shared<ExecutableNetworkBase>(_impl);
 }
 
 std::vector<VariableState> ExecutableNetwork::QueryState() {
-    if (actual) {
-        if (actual == nullptr) THROW_IE_EXCEPTION << "ExecutableNetwork was not initialized.";
-        IVariableState::Ptr pState = nullptr;
-        auto res = OK;
-        std::vector<VariableState> controller;
-        for (size_t idx = 0; res == OK; ++idx) {
-            ResponseDesc resp;
-            IE_SUPPRESS_DEPRECATED_START
-            res = actual->QueryState(pState, idx, &resp);
-            IE_SUPPRESS_DEPRECATED_END
-            if (res != OK && res != OUT_OF_BOUNDS) {
-                THROW_IE_EXCEPTION << resp.msg;
-            }
-            if (res != OUT_OF_BOUNDS) {
-                controller.push_back(VariableState(pState,
-                    std::make_shared<details::SharedObjectLoader>(_so)));
-            }
-        }
-
-        return controller;
-    }
-
     std::vector<VariableState> controller;
     EXEC_NET_CALL_STATEMENT(
         for (auto&& state : _impl->QueryState()) {
@@ -115,13 +57,6 @@ std::vector<VariableState> ExecutableNetwork::QueryState() {
 }
 
 InferRequest ExecutableNetwork::CreateInferRequest() {
-    if (actual) {
-        IInferRequest::Ptr req;
-        CALL_STATUS_FNC(CreateInferRequest, req);
-        if (req.get() == nullptr) THROW_IE_EXCEPTION << "Internal error: pointer to infer request is null";
-        return InferRequest(req, std::make_shared<details::SharedObjectLoader>(_so));
-    }
-
     EXEC_NET_CALL_STATEMENT(return {_so, _impl->CreateInferRequest()});
 }
 
@@ -130,72 +65,38 @@ InferRequest::Ptr ExecutableNetwork::CreateInferRequestPtr() {
 }
 
 void ExecutableNetwork::Export(const std::string& modelFileName) {
-    if (actual) {
-        CALL_STATUS_FNC(Export, modelFileName);
-        return;
-    }
     EXEC_NET_CALL_STATEMENT(_impl->Export(modelFileName));
 }
 
 void ExecutableNetwork::Export(std::ostream& networkModel) {
-    if (actual) {
-        CALL_STATUS_FNC(Export, networkModel);
-        return;
-    }
     EXEC_NET_CALL_STATEMENT(_impl->Export(networkModel));
 }
 
 CNNNetwork ExecutableNetwork::GetExecGraphInfo() {
-    if (actual) {
-        IE_SUPPRESS_DEPRECATED_START
-        ICNNNetwork::Ptr ptr = nullptr;
-        CALL_STATUS_FNC(GetExecGraphInfo, ptr);
-        return CNNNetwork(ptr);
-        IE_SUPPRESS_DEPRECATED_END
-    }
     EXEC_NET_CALL_STATEMENT(return _impl->GetExecGraphInfo());
 }
 
 void ExecutableNetwork::SetConfig(const std::map<std::string, Parameter>& config) {
-    if (actual) {
-        CALL_STATUS_FNC(SetConfig, config);
-        return;
-    }
     EXEC_NET_CALL_STATEMENT(_impl->SetConfig(config));
 }
 
 Parameter ExecutableNetwork::GetConfig(const std::string& name) const {
-    if (actual) {
-        Parameter configValue;
-        CALL_STATUS_FNC(GetConfig, name, configValue);
-        return configValue;
-    }
     EXEC_NET_CALL_STATEMENT(return _impl->GetConfig(name));
 }
 
 Parameter ExecutableNetwork::GetMetric(const std::string& name) const {
-    if (actual) {
-        Parameter metricValue;
-        CALL_STATUS_FNC(GetMetric, name, metricValue);
-        return metricValue;
-    }
     EXEC_NET_CALL_STATEMENT(return _impl->GetMetric(name));
 }
 
 RemoteContext::Ptr ExecutableNetwork::GetContext() const {
-    if (actual) {
-        RemoteContext::Ptr pContext;
-        CALL_STATUS_FNC(GetContext, pContext);
-        return pContext;
-    }
     EXEC_NET_CALL_STATEMENT(return _impl->GetContext());
 }
 
 bool ExecutableNetwork::operator!() const noexcept {
-    return !_impl || !actual;
+    return !_impl;
 }
 
 ExecutableNetwork::operator bool() const noexcept {
-    return !!_impl || !!actual;
+    return !!_impl;
 }
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/inference_engine/cpp/ie_executable_network_base.hpp b/inference-engine/src/inference_engine/cpp/ie_executable_network_base.hpp
index 2f813c0b783560..c87b1fc7098a2c 100644
--- a/inference-engine/src/inference_engine/cpp/ie_executable_network_base.hpp
+++ b/inference-engine/src/inference_engine/cpp/ie_executable_network_base.hpp
@@ -18,7 +18,6 @@
 #include <cpp_interfaces/interface/ie_ivariable_state_internal.hpp>
 #include <cpp_interfaces/interface/ie_iexecutable_network_internal.hpp>
 #include "cpp/exception2status.hpp"
-#include "ie_variable_state_base.hpp"
 #include "ie_infer_async_request_base.hpp"
 
 namespace InferenceEngine {
@@ -64,29 +63,10 @@ class ExecutableNetworkBase : public IExecutableNetwork {
         TO_STATUS(_impl->Export(networkModel));
     }
 
-    IE_SUPPRESS_DEPRECATED_START
     StatusCode GetExecGraphInfo(ICNNNetwork::Ptr& graphPtr, ResponseDesc* resp) noexcept override {
-        // should be refactored together with ExecutableNetwork interface
         TO_STATUS(graphPtr = _impl->GetExecGraphInfo());
     }
 
-    INFERENCE_ENGINE_DEPRECATED("Use InferRequest::QueryState instead")
-    StatusCode QueryState(IVariableState::Ptr& pState, size_t idx, ResponseDesc* resp) noexcept override {
-        try {
-            auto v = _impl->QueryState();
-            if (idx >= v.size()) {
-                return OUT_OF_BOUNDS;
-            }
-            pState = std::make_shared<VariableStateBase>(v[idx]);
-            return OK;
-        } catch (const std::exception& ex) {
-            return InferenceEngine::DescriptionBuffer(GENERAL_ERROR, resp) << ex.what();
-        } catch (...) {
-            return InferenceEngine::DescriptionBuffer(UNEXPECTED);
-        }
-    }
-    IE_SUPPRESS_DEPRECATED_END
-
     StatusCode SetConfig(const std::map<std::string, Parameter>& config, ResponseDesc* resp) noexcept override {
         TO_STATUS(_impl->SetConfig(config));
     }
diff --git a/inference-engine/src/inference_engine/cpp/ie_infer_async_request_base.hpp b/inference-engine/src/inference_engine/cpp/ie_infer_async_request_base.hpp
index 1253947eeafd15..6ede78f720e3a2 100644
--- a/inference-engine/src/inference_engine/cpp/ie_infer_async_request_base.hpp
+++ b/inference-engine/src/inference_engine/cpp/ie_infer_async_request_base.hpp
@@ -10,10 +10,10 @@
 
 #include "cpp/exception2status.hpp"
 #include "cpp_interfaces/plugin_itt.hpp"
-#include "ie_variable_state_base.hpp"
 #include <cpp_interfaces/interface/ie_iinfer_request_internal.hpp>
 #include "ie_iinfer_request.hpp"
 #include "ie_preprocess.hpp"
+
 namespace InferenceEngine {
 
 #define CATCH_IE_EXCEPTION_TO_STATUS_NO_RESP(StatusCode, ExceptionType) catch (const ExceptionType& ex) {       \
@@ -169,23 +169,6 @@ class InferRequestBase : public IInferRequest {
     StatusCode SetBatch(int batch_size, ResponseDesc* resp) noexcept override {
         TO_STATUS(_impl->SetBatch(batch_size));
     }
-
-    IE_SUPPRESS_DEPRECATED_START
-    StatusCode QueryState(IVariableState::Ptr& pState, size_t idx, ResponseDesc* resp) noexcept override {
-        try {
-            auto v = _impl->QueryState();
-            if (idx >= v.size()) {
-                return OUT_OF_BOUNDS;
-            }
-            pState = std::make_shared<VariableStateBase>(v[idx]);
-            return OK;
-        } catch (const std::exception& ex) {
-            return InferenceEngine::DescriptionBuffer(GENERAL_ERROR, resp) << ex.what();
-        } catch (...) {
-            return InferenceEngine::DescriptionBuffer(UNEXPECTED);
-        }
-    }
-    IE_SUPPRESS_DEPRECATED_END
 };
 
 IE_SUPPRESS_DEPRECATED_END
diff --git a/inference-engine/src/inference_engine/cpp/ie_infer_request.cpp b/inference-engine/src/inference_engine/cpp/ie_infer_request.cpp
index 97fba9af7f924f..9e68666b7a36f6 100644
--- a/inference-engine/src/inference_engine/cpp/ie_infer_request.cpp
+++ b/inference-engine/src/inference_engine/cpp/ie_infer_request.cpp
@@ -23,44 +23,17 @@ namespace InferenceEngine {
 
 InferRequest::InferRequest(const details::SharedObjectLoader& so,
                            const IInferRequestInternal::Ptr&  impl)
-    : _so(so), _impl(impl), actual() {
+    : _so(so), _impl(impl) {
     IE_ASSERT(_impl != nullptr);
 }
 
 IE_SUPPRESS_DEPRECATED_START
 
-InferRequest::InferRequest(IInferRequest::Ptr request,
-                           std::shared_ptr<details::SharedObjectLoader> splg)
-    : _so(), _impl(), actual(request) {
-    if (splg) {
-        _so = *splg;
-    }
-
-    //  plg can be null, but not the actual
-    if (actual == nullptr)
-        IE_THROW(NotAllocated) << "InferRequest was not initialized.";
-}
-
 void InferRequest::SetBlob(const std::string& name, const Blob::Ptr& data) {
-    if (actual) {
-        CALL_STATUS_FNC(SetBlob, name.c_str(), data);
-        return;
-    }
     INFER_REQ_CALL_STATEMENT(_impl->SetBlob(name, data);)
 }
 
 Blob::Ptr InferRequest::GetBlob(const std::string& name) {
-    if (actual) {
-        Blob::Ptr data;
-        CALL_STATUS_FNC(GetBlob, name.c_str(), data);
-        std::string error = "Internal error: blob with name `" + name + "` is not allocated!";
-        auto blobPtr = data.get();
-        const bool remoteBlobPassed = blobPtr->is<RemoteBlob>();
-        if (blobPtr == nullptr) IE_THROW() << error;
-        if (!remoteBlobPassed && blobPtr->buffer() == nullptr) IE_THROW() << error;
-        return data;
-    }
-
     Blob::Ptr blobPtr;
     INFER_REQ_CALL_STATEMENT(blobPtr = _impl->GetBlob(name);)
     std::string error = "Internal error: blob with name `" + name + "` is not allocated!";
@@ -71,60 +44,26 @@ Blob::Ptr InferRequest::GetBlob(const std::string& name) {
 }
 
 void InferRequest::SetBlob(const std::string &name, const Blob::Ptr &data, const PreProcessInfo& info) {
-    if (actual) {
-        CALL_STATUS_FNC(SetBlob, name.c_str(), data, info);
-        return;
-    }
-
     INFER_REQ_CALL_STATEMENT(_impl->SetBlob(name, data, info);)
 }
 
 const PreProcessInfo& InferRequest::GetPreProcess(const std::string& name) const {
-    if (actual) {
-        const PreProcessInfo* info = nullptr;
-        CALL_STATUS_FNC(GetPreProcess, name.c_str(), &info);
-        return *info;
-    }
-
     INFER_REQ_CALL_STATEMENT(return _impl->GetPreProcess(name);)
 }
 
 void InferRequest::Infer() {
-    if (actual) {
-        CALL_STATUS_FNC_NO_ARGS(Infer);
-        return;
-    }
-
     INFER_REQ_CALL_STATEMENT(_impl->Infer();)
 }
 
 void InferRequest::Cancel() {
-    if (actual) {
-        CALL_STATUS_FNC_NO_ARGS(Cancel);
-        return;
-    }
-
     INFER_REQ_CALL_STATEMENT(_impl->Cancel();)
 }
 
 std::map<std::string, InferenceEngineProfileInfo> InferRequest::GetPerformanceCounts() const {
-    if (actual) {
-        std::map<std::string, InferenceEngineProfileInfo> perfMap;
-        CALL_STATUS_FNC(GetPerformanceCounts, perfMap);
-        return perfMap;
-    }
-
     INFER_REQ_CALL_STATEMENT(return _impl->GetPerformanceCounts();)
 }
 
 void InferRequest::SetInput(const BlobMap& inputs) {
-    if (actual) {
-        for (auto&& input : inputs) {
-            CALL_STATUS_FNC(SetBlob, input.first.c_str(), input.second);
-        }
-        return;
-    }
-
     INFER_REQ_CALL_STATEMENT(
         for (auto&& input : inputs) {
             _impl->SetBlob(input.first, input.second);
@@ -133,13 +72,6 @@ void InferRequest::SetInput(const BlobMap& inputs) {
 }
 
 void InferRequest::SetOutput(const BlobMap& results) {
-    if (actual) {
-        for (auto&& result : results) {
-            CALL_STATUS_FNC(SetBlob, result.first.c_str(), result.second);
-        }
-        return;
-    }
-
     INFER_REQ_CALL_STATEMENT(
         for (auto&& result : results) {
             _impl->SetBlob(result.first, result.second);
@@ -148,106 +80,19 @@ void InferRequest::SetOutput(const BlobMap& results) {
 }
 
 void InferRequest::SetBatch(const int batch) {
-    if (actual) {
-        CALL_STATUS_FNC(SetBatch, batch);
-        return;
-    }
-
     INFER_REQ_CALL_STATEMENT(_impl->SetBatch(batch);)
 }
 
 void InferRequest::StartAsync() {
-    if (actual) {
-        CALL_STATUS_FNC_NO_ARGS(StartAsync);
-        return;
-    }
-
     INFER_REQ_CALL_STATEMENT(_impl->StartAsync();)
 }
 
 
 StatusCode InferRequest::Wait(int64_t millis_timeout) {
-    if (actual) {
-        ResponseDesc resp;
-        if (actual == nullptr) IE_THROW() << "InferRequest was not initialized.";
-        auto res = actual->Wait(millis_timeout, &resp);
-        if (res != OK && res != RESULT_NOT_READY &&
-            res != INFER_NOT_STARTED && res != INFER_CANCELLED) {
-            IE_EXCEPTION_SWITCH(res, ExceptionType,
-                InferenceEngine::details::ThrowNow<ExceptionType>{}
-                    <<= std::stringstream{} << IE_LOCATION << resp.msg)
-        }
-        return res;
-    }
-
     INFER_REQ_CALL_STATEMENT(return _impl->Wait(millis_timeout);)
 }
 
-namespace details {
-
-class ICompletionCallbackWrapper {
-public:
-    virtual ~ICompletionCallbackWrapper() = default;
-
-    virtual void call(InferenceEngine::IInferRequest::Ptr request, InferenceEngine::StatusCode code) const noexcept = 0;
-};
-
-template <class T>
-class CompletionCallbackWrapper : public ICompletionCallbackWrapper {
-    T lambda;
-
-public:
-    explicit CompletionCallbackWrapper(const T& lambda): lambda(lambda) {}
-
-    void call(InferenceEngine::IInferRequest::Ptr /*request*/, InferenceEngine::StatusCode /*code*/) const
-        noexcept override {
-        lambda();
-    }
-};
-
-template <>
-class CompletionCallbackWrapper<IInferRequest::CompletionCallback> : public ICompletionCallbackWrapper {
-    IInferRequest::CompletionCallback callBack;
-
-public:
-    explicit CompletionCallbackWrapper(const IInferRequest::CompletionCallback& callBack): callBack(callBack) {}
-
-    void call(InferenceEngine::IInferRequest::Ptr request, InferenceEngine::StatusCode code) const noexcept override {
-        callBack(request, code);
-    }
-};
-
-template <>
-class CompletionCallbackWrapper<std::function<void(InferRequest, StatusCode)>> : public ICompletionCallbackWrapper {
-    std::function<void(InferRequest, StatusCode)> lambda;
-
-public:
-    explicit CompletionCallbackWrapper(const std::function<void(InferRequest, InferenceEngine::StatusCode)>& lambda)
-        : lambda(lambda) {}
-
-    void call(InferenceEngine::IInferRequest::Ptr request, InferenceEngine::StatusCode code) const noexcept override {
-        lambda(InferRequest(request), code);
-    }
-};
-
-void callWrapper(InferenceEngine::IInferRequest::Ptr request, InferenceEngine::StatusCode code) {
-    details::ICompletionCallbackWrapper* pWrapper = nullptr;
-    ResponseDesc dsc;
-    request->GetUserData(reinterpret_cast<void**>(&pWrapper), &dsc);
-    pWrapper->call(request, code);
-}
-
-}  // namespace details
-
 void InferRequest::SetCompletionCallbackImpl(std::function<void()> callbackToSet) {
-    if (actual) {
-        using T = std::function<void()>;
-        callback.reset(new details::CompletionCallbackWrapper<T>(callbackToSet));
-        CALL_STATUS_FNC(SetUserData, callback.get());
-        actual->SetCompletionCallback(InferenceEngine::details::callWrapper);
-        return;
-    }
-
     INFER_REQ_CALL_STATEMENT(
         _impl->SetCallback([callbackToSet] (std::exception_ptr) {
             callbackToSet();
@@ -274,14 +119,6 @@ void InferRequest::SetCompletionCallbackImpl(std::function<void()> callbackToSet
 
 
 void InferRequest::SetCompletionCallbackImpl(std::function<void(InferRequest, StatusCode)> callbackToSet) {
-    if (actual) {
-        using T = std::function<void(InferRequest, StatusCode)>;
-        callback.reset(new details::CompletionCallbackWrapper<T>(callbackToSet));
-        CALL_STATUS_FNC(SetUserData, callback.get());
-        actual->SetCompletionCallback(InferenceEngine::details::callWrapper);
-        return;
-    }
-
     INFER_REQ_CALL_STATEMENT(
         auto weakThis = InferRequest{_so, std::shared_ptr<IInferRequestInternal>{_impl.get(), [](IInferRequestInternal*){}}};
         _impl->SetCallback([callbackToSet, weakThis] (std::exception_ptr exceptionPtr) {
@@ -303,14 +140,6 @@ void InferRequest::SetCompletionCallbackImpl(std::function<void(InferRequest, St
 }
 
 void InferRequest::SetCompletionCallbackImpl(IInferRequest::CompletionCallback callbackToSet) {
-    if (actual) {
-        using T = IInferRequest::CompletionCallback;
-        callback.reset(new details::CompletionCallbackWrapper<T>(callbackToSet));
-        CALL_STATUS_FNC(SetUserData, callback.get());
-        actual->SetCompletionCallback(InferenceEngine::details::callWrapper);
-        return;
-    }
-
     INFER_REQ_CALL_STATEMENT(
         IInferRequest::Ptr weakThis = InferRequest{_so, std::shared_ptr<IInferRequestInternal>{_impl.get(), [](IInferRequestInternal*){}}};
         _impl->SetCallback([callbackToSet, weakThis] (std::exception_ptr exceptionPtr) {
@@ -332,38 +161,12 @@ void InferRequest::SetCompletionCallbackImpl(IInferRequest::CompletionCallback c
 }
 
 InferRequest::operator IInferRequest::Ptr () {
-    if (actual) {
-        return actual;
-    }
-
     INFER_REQ_CALL_STATEMENT(
         return std::make_shared<InferRequestBase>(_impl);
     )
 }
 
 std::vector<VariableState> InferRequest::QueryState() {
-    if (actual) {
-        IE_SUPPRESS_DEPRECATED_START
-        if (actual == nullptr) IE_THROW() << "ExecutableNetwork was not initialized.";
-        IVariableState::Ptr pState = nullptr;
-        auto res = OK;
-        std::vector<VariableState> controller;
-        for (size_t idx = 0; res == OK; ++idx) {
-            ResponseDesc resp;
-            res = actual->QueryState(pState, idx, &resp);
-            if (res != OK && res != OUT_OF_BOUNDS) {
-                IE_THROW() << resp.msg;
-            }
-            if (res != OUT_OF_BOUNDS) {
-                controller.push_back(VariableState(pState,
-                    std::make_shared<details::SharedObjectLoader>(_so)));
-            }
-        }
-        IE_SUPPRESS_DEPRECATED_END
-
-        return controller;
-    }
-
     std::vector<VariableState> controller;
     INFER_REQ_CALL_STATEMENT(
         for (auto&& state : _impl->QueryState()) {
@@ -374,11 +177,11 @@ std::vector<VariableState> InferRequest::QueryState() {
 }
 
 bool InferRequest::operator!() const noexcept {
-    return !_impl || !actual;
+    return !_impl;
 }
 
 InferRequest::operator bool() const noexcept {
-    return (!!_impl) || (!!actual);
+    return (!!_impl);
 }
 
 bool InferRequest::operator!=(const InferRequest& r) const noexcept {
@@ -386,7 +189,7 @@ bool InferRequest::operator!=(const InferRequest& r) const noexcept {
 }
 
 bool InferRequest::operator==(const InferRequest& r) const noexcept {
-    return r._impl == _impl && r.actual == actual;
+    return r._impl == _impl;
 }
 
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/inference_engine/cpp/ie_variable_state.cpp b/inference-engine/src/inference_engine/cpp/ie_variable_state.cpp
index 46f99d3fc6c1d0..63f7305e8b2b38 100644
--- a/inference-engine/src/inference_engine/cpp/ie_variable_state.cpp
+++ b/inference-engine/src/inference_engine/cpp/ie_variable_state.cpp
@@ -4,7 +4,6 @@
 
 #include "details/ie_so_loader.h"
 #include "cpp/ie_memory_state.hpp"
-#include "ie_imemory_state.hpp"
 #include "cpp_interfaces/interface/ie_ivariable_state_internal.hpp"
 #include "exception2status.hpp"
 
@@ -24,57 +23,19 @@ VariableState::VariableState(const details::SharedObjectLoader& so,
 
 IE_SUPPRESS_DEPRECATED_START
 
-VariableState::VariableState(std::shared_ptr<IVariableState> state,
-                             std::shared_ptr<details::SharedObjectLoader> splg)
-    : _so(), _impl(), actual(state) {
-    if (splg) {
-        _so = *splg;
-    }
-
-    //  plg can be null, but not the actual
-    if (actual == nullptr)
-        IE_THROW(NotAllocated) << "VariableState was not initialized.";
-}
-
-Blob::CPtr VariableState::GetLastState() const {
-    return GetState();
-}
-
 void VariableState::Reset() {
-    if (actual) {
-        CALL_STATUS_FNC_NO_ARGS(Reset);
-        return;
-    }
-
     VARIABLE_CALL_STATEMENT(_impl->Reset());
 }
 
 std::string VariableState::GetName() const {
-    if (actual) {
-        char name[256];
-        CALL_STATUS_FNC(GetName, name, sizeof(name));
-        return name;
-    }
-
     VARIABLE_CALL_STATEMENT(return _impl->GetName());
 }
 
 Blob::CPtr VariableState::GetState() const {
-    if (actual) {
-        Blob::CPtr stateBlob;
-        CALL_STATUS_FNC(GetState, stateBlob);
-        return stateBlob;
-    }
-
     VARIABLE_CALL_STATEMENT(return _impl->GetState());
 }
 
 void VariableState::SetState(Blob::Ptr state) {
-    if (actual) {
-        CALL_STATUS_FNC(SetState, state);
-        return;
-    }
-
     VARIABLE_CALL_STATEMENT(_impl->SetState(state));
 }
 
diff --git a/inference-engine/src/inference_engine/cpp/ie_variable_state_base.hpp b/inference-engine/src/inference_engine/cpp/ie_variable_state_base.hpp
deleted file mode 100644
index 2481ca67852ddb..00000000000000
--- a/inference-engine/src/inference_engine/cpp/ie_variable_state_base.hpp
+++ /dev/null
@@ -1,59 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-
-#include <memory>
-
-#include "cpp/exception2status.hpp"
-#include "cpp_interfaces/interface/ie_ivariable_state_internal.hpp"
-#include "ie_imemory_state.hpp"
-
-namespace InferenceEngine {
-
-IE_SUPPRESS_DEPRECATED_START
-
-/**
- * @brief Default implementation for IVariableState
-  * @ingroup ie_dev_api_variable_state_api
- */
-class VariableStateBase : public IVariableState {
-    std::shared_ptr<IVariableStateInternal> impl;
-
-public:
-    /**
-     * @brief Constructor with actual underlying implementation.
-     * @param impl Underlying implementation of type IVariableStateInternal
-     */
-    explicit VariableStateBase(std::shared_ptr<IVariableStateInternal> impl): impl(impl) {
-        if (impl == nullptr) {
-            IE_THROW() << "VariableStateBase implementation is not defined";
-        }
-    }
-
-    StatusCode GetName(char* name, size_t len, ResponseDesc* resp) const noexcept override {
-        for (size_t i = 0; i != len; i++) {
-            name[i] = 0;
-        }
-        DescriptionBuffer buf(name, len);
-        TO_STATUS(buf << impl->GetName());
-        return OK;
-    }
-
-    StatusCode Reset(ResponseDesc* resp) noexcept override {
-        TO_STATUS(impl->Reset());
-    }
-
-    StatusCode SetState(Blob::Ptr newState, ResponseDesc* resp) noexcept override {
-        TO_STATUS(impl->SetState(newState));
-    }
-
-    StatusCode GetState(Blob::CPtr& state, ResponseDesc* resp) const noexcept override {
-        TO_STATUS(state = impl->GetState());
-    }
-};
-
-IE_SUPPRESS_DEPRECATED_END
-
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_ivariable_state_internal.cpp b/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_ivariable_state_internal.cpp
index 0171292d36bbdc..a499e816ee0b2c 100644
--- a/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_ivariable_state_internal.cpp
+++ b/inference-engine/src/inference_engine/cpp_interfaces/interface/ie_ivariable_state_internal.cpp
@@ -23,7 +23,4 @@ Blob::CPtr IVariableStateInternal::GetState() const {
     return state;
 }
 
-Blob::CPtr IVariableStateInternal::GetLastState() const {
-    return GetState();
-}
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/inference_engine/ie_parameter.cpp b/inference-engine/src/inference_engine/ie_parameter.cpp
deleted file mode 100644
index 61fbf54c37dcc2..00000000000000
--- a/inference-engine/src/inference_engine/ie_parameter.cpp
+++ /dev/null
@@ -1,45 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include <ie_parameter.hpp>
-#include <memory>
-
-#include <ngraph/variant.hpp>
-
-namespace ngraph {
-
-template class INFERENCE_ENGINE_API_CLASS(VariantImpl<InferenceEngine::Parameter>);
-
-template <>
-class INFERENCE_ENGINE_API_CLASS(VariantWrapper<InferenceEngine::Parameter>) : public VariantImpl<InferenceEngine::Parameter> {
-public:
-    static constexpr VariantTypeInfo type_info {"Variant::InferenceEngine::Parameter", 0};
-    const VariantTypeInfo& get_type_info() const override {
-        return type_info;
-    }
-    VariantWrapper(const value_type& value): VariantImpl<value_type>(value) {}  // NOLINT
-};
-
-}  // namespace ngraph
-
-constexpr ngraph::VariantTypeInfo ngraph::VariantWrapper<InferenceEngine::Parameter>::type_info;
-
-InferenceEngine::Parameter::Parameter(const std::shared_ptr<ngraph::Variant>& var) {
-    if (auto paramWrapper = std::dynamic_pointer_cast<ngraph::VariantWrapper<InferenceEngine::Parameter>>(var)) {
-        auto param = paramWrapper->get();
-        if (!param.empty()) ptr = param.ptr->copy();
-    }
-}
-
-InferenceEngine::Parameter::Parameter(std::shared_ptr<ngraph::Variant>& var) {
-    if (auto paramWrapper = std::dynamic_pointer_cast<ngraph::VariantWrapper<InferenceEngine::Parameter>>(var)) {
-        auto param = paramWrapper->get();
-        if (!param.empty()) ptr = param.ptr->copy();
-    }
-}
-
-
-std::shared_ptr<ngraph::Variant> InferenceEngine::Parameter::asVariant() const {
-    return std::make_shared<ngraph::VariantWrapper<InferenceEngine::Parameter>>(*this);
-}
diff --git a/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_ivariable_state_internal.hpp b/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_ivariable_state_internal.hpp
index f92fd556f00f45..d34af53631a556 100644
--- a/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_ivariable_state_internal.hpp
+++ b/inference-engine/src/plugin_api/cpp_interfaces/interface/ie_ivariable_state_internal.hpp
@@ -50,14 +50,6 @@ class INFERENCE_ENGINE_API_CLASS(IVariableStateInternal) : public std::enable_sh
      */
     virtual Blob::CPtr GetState() const;
 
-    /**
-     * @deprecated Use IVariableStateInternal::GetState method instead
-     * @brief Returns the value of the variable state.
-     * @return The value of the variable state
-     */
-    INFERENCE_ENGINE_DEPRECATED("Use IVariableStateInternal::GetState method instead")
-    virtual Blob::CPtr GetLastState() const;
-
 protected:
     /**
      * @brief A default dtor
diff --git a/inference-engine/tests/functional/inference_engine/async_infer_request_test.cpp b/inference-engine/tests/functional/inference_engine/async_infer_request_test.cpp
index cc5cc1fe4eb826..6393aa69d16199 100644
--- a/inference-engine/tests/functional/inference_engine/async_infer_request_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/async_infer_request_test.cpp
@@ -5,26 +5,12 @@
 #include <gtest/gtest.h>
 
 #include <cpp/ie_infer_request.hpp>
-#include "unit_test_utils/mocks/mock_iinfer_request.hpp"
 
 using namespace ::testing;
 using namespace std;
 using namespace InferenceEngine;
 using namespace InferenceEngine::details;
 
-IE_SUPPRESS_DEPRECATED_START
-
-TEST(InferRequestCPPTests, throwsOnUninitialized) {
-    std::shared_ptr<IInferRequest> ptr;
-    ASSERT_THROW(InferRequest req(ptr), InferenceEngine::NotAllocated);
-}
-
-TEST(InferRequestCPPTests, nothrowOnInitialized) {
-    std::shared_ptr<IInferRequest> ptr = std::make_shared<MockIInferRequest>();
-    ASSERT_NO_THROW(InferRequest req(ptr));
-}
-
-IE_SUPPRESS_DEPRECATED_END
 
 TEST(InferRequestCPPTests, throwsOnUninitializedSetBlob) {
     InferRequest req;
diff --git a/inference-engine/tests/functional/inference_engine/executable_network.cpp b/inference-engine/tests/functional/inference_engine/executable_network.cpp
index 8db2bf40ed2402..89653bdc5ee7f1 100644
--- a/inference-engine/tests/functional/inference_engine/executable_network.cpp
+++ b/inference-engine/tests/functional/inference_engine/executable_network.cpp
@@ -4,75 +4,60 @@
 
 #include <gtest/gtest.h>
 #include <cpp/ie_executable_network.hpp>
-#include "unit_test_utils/mocks/mock_iexecutable_network.hpp"
 
 using namespace ::testing;
 using namespace std;
 using namespace InferenceEngine;
 using namespace InferenceEngine::details;
 
-IE_SUPPRESS_DEPRECATED_START
-
-TEST(ExecutableNetworkTests, throwsOnUninitialized) {
-    std::shared_ptr<IExecutableNetwork> ptr;
-    ASSERT_THROW(ExecutableNetwork req(ptr), InferenceEngine::NotAllocated);
-}
-
-TEST(ExecutableNetworkTests, nothrowOnInitialized) {
-    std::shared_ptr<IExecutableNetwork> ptr = std::make_shared<MockIExecutableNetwork>();
-    ASSERT_NO_THROW(ExecutableNetwork req(ptr));
-}
-
-IE_SUPPRESS_DEPRECATED_END
-
 TEST(ExecutableNetworkTests, throwsOnUninitializedGetOutputsInfo) {
     ExecutableNetwork exec;
-    ASSERT_THROW(exec.GetOutputsInfo(), InferenceEngine::Exception);
+    ASSERT_THROW(exec.GetOutputsInfo(), InferenceEngine::NotAllocated);
 }
 
 TEST(ExecutableNetworkTests, throwsOnUninitializedGetInputsInfo) {
     ExecutableNetwork exec;
-    ASSERT_THROW(exec.GetInputsInfo(), InferenceEngine::Exception);
+    ASSERT_THROW(exec.GetInputsInfo(), InferenceEngine::NotAllocated);
 }
 
 TEST(ExecutableNetworkTests, throwsOnUninitializedExport) {
     ExecutableNetwork exec;
-    ASSERT_THROW(exec.Export(std::string()), InferenceEngine::Exception);
+    ASSERT_THROW(exec.Export(std::string()), InferenceEngine::NotAllocated);
 }
 
 TEST(ExecutableNetworkTests, throwsOnUninitializedExportStream) {
     ExecutableNetwork exec;
-    ASSERT_THROW(exec.Export(std::cout), InferenceEngine::Exception);
+    ASSERT_THROW(exec.Export(std::cout), InferenceEngine::NotAllocated);
 }
 
 TEST(ExecutableNetworkTests, throwsOnUninitializedGetExecGraphInfo) {
     ExecutableNetwork exec;
-    ASSERT_THROW(exec.GetExecGraphInfo(), InferenceEngine::Exception);
+    ASSERT_THROW(exec.GetExecGraphInfo(), InferenceEngine::NotAllocated);
 }
 
 TEST(ExecutableNetworkTests, throwsOnUninitializedQueryState) {
     IE_SUPPRESS_DEPRECATED_START
     ExecutableNetwork exec;
-    ASSERT_THROW(exec.QueryState(), InferenceEngine::Exception);
+    ASSERT_THROW(exec.QueryState(), InferenceEngine::NotAllocated);
     IE_SUPPRESS_DEPRECATED_END
 }
 
 TEST(ExecutableNetworkTests, throwsOnUninitializedSetConfig) {
     ExecutableNetwork exec;
-    ASSERT_THROW(exec.SetConfig({{}}), InferenceEngine::Exception);
+    ASSERT_THROW(exec.SetConfig({{}}), InferenceEngine::NotAllocated);
 }
 
 TEST(ExecutableNetworkTests, throwsOnUninitializedGetConfig) {
     ExecutableNetwork exec;
-    ASSERT_THROW(exec.GetConfig({}), InferenceEngine::Exception);
+    ASSERT_THROW(exec.GetConfig({}), InferenceEngine::NotAllocated);
 }
 
 TEST(ExecutableNetworkTests, throwsOnUninitializedGetMetric) {
     ExecutableNetwork exec;
-    ASSERT_THROW(exec.GetMetric({}), InferenceEngine::Exception);
+    ASSERT_THROW(exec.GetMetric({}), InferenceEngine::NotAllocated);
 }
 
 TEST(ExecutableNetworkTests, throwsOnUninitializedGetContext) {
     ExecutableNetwork exec;
-    ASSERT_THROW(exec.GetContext(), InferenceEngine::Exception);
+    ASSERT_THROW(exec.GetContext(), InferenceEngine::NotAllocated);
 }
\ No newline at end of file
diff --git a/inference-engine/tests/functional/inference_engine/variable_state.cpp b/inference-engine/tests/functional/inference_engine/variable_state.cpp
index 5073a155fca2e9..f17bd599bb6026 100644
--- a/inference-engine/tests/functional/inference_engine/variable_state.cpp
+++ b/inference-engine/tests/functional/inference_engine/variable_state.cpp
@@ -5,32 +5,12 @@
 #include <gtest/gtest.h>
 
 #include <cpp/ie_infer_request.hpp>
-#include "unit_test_utils/mocks/mock_ie_ivariable_state.hpp"
 
 using namespace ::testing;
 using namespace std;
 using namespace InferenceEngine;
 using namespace InferenceEngine::details;
 
-IE_SUPPRESS_DEPRECATED_START
-
-TEST(VariableStateCPPTests, throwsOnUninitialized) {
-    std::shared_ptr<IVariableState> ptr;
-    ASSERT_THROW(VariableState var(ptr), InferenceEngine::NotAllocated);
-}
-
-TEST(VariableStateCPPTests, nothrowOnInitialized) {
-    std::shared_ptr<IVariableState> ptr = std::make_shared<MockIVariableState>();
-    ASSERT_NO_THROW(VariableState var(ptr));
-}
-
-TEST(VariableStateCPPTests, throwsOnUninitializedGetLastState) {
-    VariableState req;
-    ASSERT_THROW(req.GetLastState(), InferenceEngine::NotAllocated);
-}
-
-IE_SUPPRESS_DEPRECATED_END
-
 TEST(VariableStateCPPTests, throwsOnUninitializedReset) {
     VariableState req;
     ASSERT_THROW(req.Reset(), InferenceEngine::NotAllocated);
diff --git a/inference-engine/tests/functional/shared_test_classes/src/subgraph/concat_quantization_during_memory_requantization.cpp b/inference-engine/tests/functional/shared_test_classes/src/subgraph/concat_quantization_during_memory_requantization.cpp
index b02dc5b86b7f54..729dfdb4e508ef 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/subgraph/concat_quantization_during_memory_requantization.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/subgraph/concat_quantization_during_memory_requantization.cpp
@@ -104,11 +104,11 @@ namespace SubgraphTestsDefinitions {
         for (auto& state : states) {
             auto name = state.GetName();
             if (name == "memory_1") {
-                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetLastState()->getTensorDesc(),
+                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetState()->getTensorDesc(),
                                                                            memory_1_init.data(), memory_1_init.size());
                 state.SetState(blob);
             } else if (name == "memory_2") {
-                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetLastState()->getTensorDesc(),
+                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetState()->getTensorDesc(),
                                                                            memory_2_init.data(), memory_2_init.size());
                 state.SetState(blob);
             } else {
diff --git a/inference-engine/tests/functional/shared_test_classes/src/subgraph/memory_LSTMCell.cpp b/inference-engine/tests/functional/shared_test_classes/src/subgraph/memory_LSTMCell.cpp
index bac0c293add651..291774616714a7 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/subgraph/memory_LSTMCell.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/subgraph/memory_LSTMCell.cpp
@@ -280,11 +280,11 @@ namespace SubgraphTestsDefinitions {
         for (auto& state : states) {
             auto name = state.GetName();
             if (name.find("cell_state_1") != std::string::npos) {
-                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetLastState()->getTensorDesc(),
+                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetState()->getTensorDesc(),
                                                                            cell_memory_init.data(), cell_memory_init.size());
                 state.SetState(blob);
             } else if (name.find("hidden_state_1") != std::string::npos) {
-                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetLastState()->getTensorDesc(),
+                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetState()->getTensorDesc(),
                                                                            hidden_memory_init.data(), hidden_memory_init.size());
                 state.SetState(blob);
             } else {
diff --git a/inference-engine/tests/functional/shared_test_classes/src/subgraph/negative_memory_layer_offset.cpp b/inference-engine/tests/functional/shared_test_classes/src/subgraph/negative_memory_layer_offset.cpp
index c11eff780f7275..9375b7b6baf31f 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/subgraph/negative_memory_layer_offset.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/subgraph/negative_memory_layer_offset.cpp
@@ -77,7 +77,7 @@ namespace SubgraphTestsDefinitions {
         for (auto& state : states) {
             auto name = state.GetName();
             if (name == "memory") {
-                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetLastState()->getTensorDesc(),
+                auto blob = FuncTestUtils::createAndFillBlobWithFloatArray(state.GetState()->getTensorDesc(),
                                                                            memory_init.data(), memory_init.size());
                 state.SetState(blob);
             } else {
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mock.cpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mock.cpp
index 676bf59b340061..980044f8d0265d 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mock.cpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mock.cpp
@@ -4,7 +4,6 @@
 
 #include "unit_test_utils/mocks/mock_allocator.hpp"
 #include "unit_test_utils/mocks/mock_icnn_network.hpp"
-#include "unit_test_utils/mocks/mock_ie_ivariable_state.hpp"
 #include "unit_test_utils/mocks/mock_iexecutable_network.hpp"
 #include "unit_test_utils/mocks/mock_iinfer_request.hpp"
 #include "unit_test_utils/mocks/mock_not_empty_icnn_network.hpp"
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_ie_ivariable_state.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_ie_ivariable_state.hpp
deleted file mode 100644
index eac6fcbfa3f246..00000000000000
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_ie_ivariable_state.hpp
+++ /dev/null
@@ -1,25 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-
-#include <gmock/gmock.h>
-#include <string>
-#include <vector>
-
-#include "ie_imemory_state.hpp"
-
-using namespace InferenceEngine;
-
-IE_SUPPRESS_DEPRECATED_START
-
-class MockIVariableState : public InferenceEngine::IVariableState {
-public:
-    MOCK_QUALIFIED_METHOD3(GetName, const noexcept, StatusCode(char * , size_t, ResponseDesc *));
-    MOCK_QUALIFIED_METHOD1(Reset, noexcept, StatusCode(ResponseDesc *));
-    MOCK_QUALIFIED_METHOD2(SetState, noexcept, StatusCode(Blob::Ptr, ResponseDesc *));
-    MOCK_QUALIFIED_METHOD2(GetState, const noexcept, StatusCode(Blob::CPtr &, ResponseDesc *));
-};
-
-IE_SUPPRESS_DEPRECATED_END
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iexecutable_network.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iexecutable_network.hpp
index 147ee892a8f1e0..a5187953206ddd 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iexecutable_network.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iexecutable_network.hpp
@@ -32,7 +32,6 @@ class MockIExecutableNetwork : public IExecutableNetwork {
     MOCK_QUALIFIED_METHOD3(GetConfig, const noexcept, StatusCode(const std::string &name, Parameter &result, ResponseDesc *resp));
     MOCK_QUALIFIED_METHOD3(GetMetric, const noexcept, StatusCode(const std::string &name, Parameter &result, ResponseDesc *resp));
     MOCK_QUALIFIED_METHOD2(GetContext, const noexcept, StatusCode(RemoteContext::Ptr &pContext, ResponseDesc *resp));
-    MOCK_QUALIFIED_METHOD3(QueryState, noexcept, StatusCode(IVariableState::Ptr &, size_t, ResponseDesc *));
 };
 
 IE_SUPPRESS_DEPRECATED_END
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iinfer_request.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iinfer_request.hpp
index a56ad790a979ea..bac5fc7773e024 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iinfer_request.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iinfer_request.hpp
@@ -35,7 +35,6 @@ class MockIInferRequest : public IInferRequest {
     MOCK_QUALIFIED_METHOD3(SetBlob, noexcept, StatusCode(const char*, const Blob::Ptr&, ResponseDesc*));
     MOCK_QUALIFIED_METHOD4(SetBlob, noexcept, StatusCode(const char*, const Blob::Ptr&, const PreProcessInfo&, ResponseDesc*));
     MOCK_QUALIFIED_METHOD2(SetBatch, noexcept, StatusCode(int batch, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD3(QueryState, noexcept, StatusCode(IVariableState::Ptr &, size_t, ResponseDesc *));
     MOCK_QUALIFIED_METHOD1(Cancel, noexcept, InferenceEngine::StatusCode(ResponseDesc*));
 };
 
diff --git a/inference-engine/tests/unit/inference_engine/ie_executable_network_test.cpp b/inference-engine/tests/unit/inference_engine/ie_executable_network_test.cpp
index 2c35e8652dfcc0..998620c026da7f 100644
--- a/inference-engine/tests/unit/inference_engine/ie_executable_network_test.cpp
+++ b/inference-engine/tests/unit/inference_engine/ie_executable_network_test.cpp
@@ -13,7 +13,6 @@
 
 #include "unit_test_utils/mocks/mock_iexecutable_network.hpp"
 #include "unit_test_utils/mocks/mock_iinfer_request.hpp"
-#include "unit_test_utils/mocks/mock_ie_ivariable_state.hpp"
 #include "unit_test_utils/mocks/cpp_interfaces/impl/mock_inference_plugin_internal.hpp"
 #include "unit_test_utils/mocks/cpp_interfaces/interface/mock_iexecutable_network_internal.hpp"
 #include "unit_test_utils/mocks/cpp_interfaces/interface/mock_ivariable_state_internal.hpp"

From 0361e7ca73f6dc6241eedcb4623c84b46d1a0129 Mon Sep 17 00:00:00 2001
From: LuweiZhou <luwei.zhou@intel.com>
Date: Mon, 21 Jun 2021 12:01:11 +0800
Subject: [PATCH 140/178] Migrate the unary OP visitor to typed template test.
 (#6181)

* Migrate the unary OP visitor to typed template test.
The unaray OP list is :Floor, Log, Result, Sqrt, Squeeze.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>

* Add element precision type into GTEST typed parameter.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>

* Update the license to new version.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>

* Update the OP Ceiling visitor test to use unary typed template.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>
---
 ngraph/test/visitors/op/ceiling.cpp   | 27 +++---------
 ngraph/test/visitors/op/floor.cpp     | 27 ++++--------
 ngraph/test/visitors/op/log.cpp       | 34 ++++-----------
 ngraph/test/visitors/op/result.cpp    | 26 ++++--------
 ngraph/test/visitors/op/sqrt.cpp      | 33 ++++-----------
 ngraph/test/visitors/op/squeeze.cpp   | 32 ++++-----------
 ngraph/test/visitors/op/unary_ops.hpp | 59 +++++++++++++++++++++++++++
 7 files changed, 103 insertions(+), 135 deletions(-)
 create mode 100644 ngraph/test/visitors/op/unary_ops.hpp

diff --git a/ngraph/test/visitors/op/ceiling.cpp b/ngraph/test/visitors/op/ceiling.cpp
index 34ed3b653b36ef..11e71e1e9c604e 100644
--- a/ngraph/test/visitors/op/ceiling.cpp
+++ b/ngraph/test/visitors/op/ceiling.cpp
@@ -2,25 +2,10 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gtest/gtest.h"
+#include "unary_ops.hpp"
+using Type = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Ceiling, element::f32>>;
 
-#include "ngraph/ngraph.hpp"
-#include "ngraph/op/util/attr_types.hpp"
-#include "ngraph/opsets/opset1.hpp"
-#include "util/visitor.hpp"
-
-using namespace std;
-using namespace ngraph;
-using ngraph::test::NodeBuilder;
-
-TEST(attributes, ceiling_op)
-{
-    NodeBuilder::get_ops().register_factory<opset1::Ceiling>();
-    const auto A = make_shared<op::Parameter>(element::f32, Shape{5, 2});
-
-    const auto ceiling = make_shared<opset1::Ceiling>(A);
-    NodeBuilder builder(ceiling);
-
-    const auto expected_attr_count = 0;
-    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
-}
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+                              UnaryOperatorVisitor,
+                              Type,
+                              UnaryOperatorTypeName);
\ No newline at end of file
diff --git a/ngraph/test/visitors/op/floor.cpp b/ngraph/test/visitors/op/floor.cpp
index 57cd6011c67239..98731229010b1b 100644
--- a/ngraph/test/visitors/op/floor.cpp
+++ b/ngraph/test/visitors/op/floor.cpp
@@ -2,25 +2,12 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gtest/gtest.h"
+#include "unary_ops.hpp"
 
-#include "ngraph/ngraph.hpp"
-#include "ngraph/op/util/attr_types.hpp"
-#include "ngraph/opsets/opset1.hpp"
-#include "util/visitor.hpp"
+using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Floor, element::f32>,
+                               UnaryOperatorType<ngraph::op::v0::Floor, element::f16>>;
 
-using namespace std;
-using namespace ngraph;
-using ngraph::test::NodeBuilder;
-
-TEST(attributes, floor_op)
-{
-    NodeBuilder::get_ops().register_factory<opset1::Floor>();
-    const auto A = make_shared<op::Parameter>(element::f32, Shape{5, 2});
-
-    const auto floor = make_shared<opset1::Floor>(A);
-    NodeBuilder builder(floor);
-
-    const auto expected_attr_count = 0;
-    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
-}
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+                              UnaryOperatorVisitor,
+                              Types,
+                              UnaryOperatorTypeName);
\ No newline at end of file
diff --git a/ngraph/test/visitors/op/log.cpp b/ngraph/test/visitors/op/log.cpp
index 02d6293871c028..5acb4971ba81f2 100644
--- a/ngraph/test/visitors/op/log.cpp
+++ b/ngraph/test/visitors/op/log.cpp
@@ -1,29 +1,11 @@
 // Copyright (C) 2021 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
-
-#include "gtest/gtest.h"
-
-#include "ngraph/ngraph.hpp"
-#include "ngraph/op/util/attr_types.hpp"
-#include "ngraph/opsets/opset1.hpp"
-
-#include "util/visitor.hpp"
-
-using namespace std;
-using namespace ngraph;
-using ngraph::test::NodeBuilder;
-
-TEST(attributes, log_op)
-{
-    using namespace opset1;
-
-    NodeBuilder::get_ops().register_factory<Log>();
-    const auto data_input = make_shared<Parameter>(element::f32, Shape{1, 2, 3});
-    const auto op = make_shared<Sqrt>(data_input);
-
-    NodeBuilder builder(op);
-    const auto expected_attr_count = 0;
-
-    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
-}
\ No newline at end of file
+#include "unary_ops.hpp"
+using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Log, element::f32>,
+                               UnaryOperatorType<ngraph::op::v0::Log, element::f16>>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+                              UnaryOperatorVisitor,
+                              Types,
+                              UnaryOperatorTypeName);
\ No newline at end of file
diff --git a/ngraph/test/visitors/op/result.cpp b/ngraph/test/visitors/op/result.cpp
index 528d1ee50cfe54..df01ad5f913371 100644
--- a/ngraph/test/visitors/op/result.cpp
+++ b/ngraph/test/visitors/op/result.cpp
@@ -2,24 +2,12 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gtest/gtest.h"
-#include "ngraph/ngraph.hpp"
-#include "ngraph/opsets/opset1.hpp"
-#include "util/visitor.hpp"
+#include "unary_ops.hpp"
 
-using namespace std;
-using namespace ngraph;
-using ngraph::test::NodeBuilder;
-using ngraph::test::ValueMap;
+using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Result, element::f32>,
+                               UnaryOperatorType<ngraph::op::v0::Result, element::f16>>;
 
-TEST(attributes, result)
-{
-    NodeBuilder::get_ops().register_factory<opset1::Squeeze>();
-    const auto data_node = make_shared<op::Parameter>(element::f32, Shape{1});
-    const auto result = make_shared<op::Result>(data_node);
-
-    NodeBuilder builder(result);
-    const auto expected_attr_count = 0;
-
-    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
-}
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+                              UnaryOperatorVisitor,
+                              Types,
+                              UnaryOperatorTypeName);
\ No newline at end of file
diff --git a/ngraph/test/visitors/op/sqrt.cpp b/ngraph/test/visitors/op/sqrt.cpp
index 23bc0035390c81..5e26c542348dbf 100644
--- a/ngraph/test/visitors/op/sqrt.cpp
+++ b/ngraph/test/visitors/op/sqrt.cpp
@@ -2,28 +2,11 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gtest/gtest.h"
-
-#include "ngraph/ngraph.hpp"
-#include "ngraph/op/util/attr_types.hpp"
-#include "ngraph/opsets/opset1.hpp"
-
-#include "util/visitor.hpp"
-
-using namespace std;
-using namespace ngraph;
-using ngraph::test::NodeBuilder;
-
-TEST(attributes, sqrt_op)
-{
-    using namespace opset1;
-
-    NodeBuilder::get_ops().register_factory<Sqrt>();
-    const auto data_input = make_shared<Parameter>(element::f32, Shape{1, 2, 3});
-    const auto op = make_shared<Sqrt>(data_input);
-
-    NodeBuilder builder(op);
-    const auto expected_attr_count = 0;
-
-    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
-}
+#include "unary_ops.hpp"
+using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Sqrt, element::f32>,
+                               UnaryOperatorType<ngraph::op::v0::Sqrt, element::f16>>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+                              UnaryOperatorVisitor,
+                              Types,
+                              UnaryOperatorTypeName);
\ No newline at end of file
diff --git a/ngraph/test/visitors/op/squeeze.cpp b/ngraph/test/visitors/op/squeeze.cpp
index feed9fcdf9be81..a8b0c2c488c1ab 100644
--- a/ngraph/test/visitors/op/squeeze.cpp
+++ b/ngraph/test/visitors/op/squeeze.cpp
@@ -2,27 +2,11 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gtest/gtest.h"
-
-#include "ngraph/ngraph.hpp"
-#include "ngraph/op/util/attr_types.hpp"
-#include "ngraph/opsets/opset1.hpp"
-
-#include "util/visitor.hpp"
-
-using namespace std;
-using namespace ngraph;
-using ngraph::test::NodeBuilder;
-using ngraph::test::ValueMap;
-
-TEST(attributes, squeeze_op)
-{
-    NodeBuilder::get_ops().register_factory<opset1::Squeeze>();
-    const auto data_node = make_shared<op::Parameter>(element::f32, Shape{1});
-    const auto squeeze = make_shared<op::Squeeze>(data_node);
-
-    NodeBuilder builder(squeeze);
-    const auto expected_attr_count = 0;
-
-    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
-}
+#include "unary_ops.hpp"
+using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Squeeze, element::f32>,
+                              UnaryOperatorType<ngraph::op::v0::Squeeze, element::f16>>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+                              UnaryOperatorVisitor,
+                              Types,
+                              UnaryOperatorTypeName);
\ No newline at end of file
diff --git a/ngraph/test/visitors/op/unary_ops.hpp b/ngraph/test/visitors/op/unary_ops.hpp
new file mode 100644
index 00000000000000..16764c42a7c2cc
--- /dev/null
+++ b/ngraph/test/visitors/op/unary_ops.hpp
@@ -0,0 +1,59 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <vector>
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+
+#include "ngraph/op/util/attr_types.hpp"
+#include "util/visitor.hpp"
+
+using namespace ngraph;
+using ngraph::test::NodeBuilder;
+template <typename T, element::Type_t ELEMENT_TYPE>
+class UnaryOperatorType
+{
+public:
+    using op_type = T;
+    static constexpr element::Type_t element_type = ELEMENT_TYPE;
+};
+template <typename T>
+class UnaryOperatorVisitor : public testing::Test
+{
+};
+
+class UnaryOperatorTypeName
+{
+public:
+    template <typename T>
+    static std::string GetName(int)
+    {
+        using OP_Type = typename T::op_type;
+        constexpr element::Type precision(T::element_type);
+        const ngraph::Node::type_info_t typeinfo = OP_Type::get_type_info_static();
+        std::string op_name{typeinfo.name};
+        op_name.append("_");
+        return (op_name.append(precision.get_type_name()));
+    }
+};
+
+TYPED_TEST_CASE_P(UnaryOperatorVisitor);
+
+TYPED_TEST_P(UnaryOperatorVisitor, No_Attribute_4D)
+{
+    using OP_Type = typename TypeParam::op_type;
+    const element::Type_t element_type = TypeParam::element_type;
+
+    NodeBuilder::get_ops().register_factory<OP_Type>();
+    const auto A = std::make_shared<op::Parameter>(element_type, PartialShape{2, 2, 2, 2});
+
+    const auto op_func = std::make_shared<OP_Type>(A);
+    NodeBuilder builder(op_func);
+    const auto expected_attr_count = 0;
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
+}
+
+REGISTER_TYPED_TEST_CASE_P(UnaryOperatorVisitor, No_Attribute_4D);

From 6239ff8a1c4146d8c2ed2dffc87ac46613bb6387 Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Mon, 21 Jun 2021 08:27:15 +0300
Subject: [PATCH 141/178] Removed deprecated documentation (#6253)

* Removed deprecated documentation

* Removed snippets
---
 docs/IE_DG/Migration_CoreAPI.md         | 70 -------------------------
 docs/IE_DG/OnnxImporterTutorial.md      | 67 -----------------------
 docs/IE_DG/supported_plugins/FPGA.md    | 22 --------
 docs/doxygen/doxygen-ignore.txt         |  2 +
 docs/doxygen/ie_docs.xml                |  3 --
 docs/snippets/CMakeLists.txt            | 11 ----
 docs/snippets/Migration_CoreAPI.cpp     | 48 -----------------
 docs/snippets/OnnxImporterTutorial0.cpp | 19 -------
 docs/snippets/OnnxImporterTutorial1.cpp | 15 ------
 docs/snippets/OnnxImporterTutorial2.cpp | 29 ----------
 docs/snippets/OnnxImporterTutorial3.cpp | 12 -----
 11 files changed, 2 insertions(+), 296 deletions(-)
 delete mode 100644 docs/IE_DG/Migration_CoreAPI.md
 delete mode 100644 docs/IE_DG/OnnxImporterTutorial.md
 delete mode 100644 docs/IE_DG/supported_plugins/FPGA.md
 delete mode 100644 docs/snippets/Migration_CoreAPI.cpp
 delete mode 100644 docs/snippets/OnnxImporterTutorial0.cpp
 delete mode 100644 docs/snippets/OnnxImporterTutorial1.cpp
 delete mode 100644 docs/snippets/OnnxImporterTutorial2.cpp
 delete mode 100644 docs/snippets/OnnxImporterTutorial3.cpp

diff --git a/docs/IE_DG/Migration_CoreAPI.md b/docs/IE_DG/Migration_CoreAPI.md
deleted file mode 100644
index d49bd425bc87c6..00000000000000
--- a/docs/IE_DG/Migration_CoreAPI.md
+++ /dev/null
@@ -1,70 +0,0 @@
-[DEPRECATED] Migration from Inference Engine Plugin API to Core API {#openvino_docs_IE_DG_Migration_CoreAPI}
-===============================
-
-For 2019 R2 Release, the new Inference Engine Core API is introduced. This guide is updated to reflect the new API approach. The Inference Engine Plugin API is still supported, but is going to be deprecated in future releases.
-
-This section provides common steps to migrate your application written using the Inference Engine Plugin API (`InferenceEngine::InferencePlugin`) to the Inference Engine Core API (`InferenceEngine::Core`). 
-
-To learn how to write a new application using the Inference Engine, refer to [Integrate the Inference Engine Request API with Your Application](Integrate_with_customer_application_new_API.md) and [Inference Engine Samples Overview](Samples_Overview.md).
-
-## Inference Engine Core Class
-
-The Inference Engine Core class is implemented on top existing Inference Engine Plugin API and handles plugins internally. 
-The main responsibility of the `InferenceEngine::Core` class is to hide plugin specifics inside and provide a new layer of abstraction that works with devices (`InferenceEngine::Core::GetAvailableDevices`). Almost all methods of this class accept `deviceName` as an additional parameter that denotes an actual device you are working with. Plugins are listed in the `plugins.xml` file, which is loaded during constructing `InferenceEngine::Core` objects:
-
-```bash
-<ie>
-    <plugins>
-        <plugin name="CPU" location="libMKLDNNPlugin.so">
-        </plugin>
-        ...
-</ie>
-```
-
-## Migration Steps
-
-Common migration process includes the following steps:
-
-1. Migrate from the `InferenceEngine::InferencePlugin` initialization:
-
-@snippet snippets/Migration_CoreAPI.cpp part0
-
-to the `InferenceEngine::Core` class initialization:
-
-@snippet snippets/Migration_CoreAPI.cpp part1
-
-2. Instead of using `InferenceEngine::CNNNetReader` to read IR:
-
-@snippet snippets/Migration_CoreAPI.cpp part2
-
-read networks using the Core class:
-
-@snippet snippets/Migration_CoreAPI.cpp part3
-
-The Core class also allows reading models from the ONNX format (more information is [here](./ONNX_Support.md)):
-
-@snippet snippets/Migration_CoreAPI.cpp part4
-
-3. Instead of adding CPU device extensions to the plugin:
-
-@snippet snippets/Migration_CoreAPI.cpp part5
-
-add extensions to CPU device using the Core class:
-
-@snippet snippets/Migration_CoreAPI.cpp part6
-
-4. Instead of setting configuration keys to a particular plugin, set (key, value) pairs via `InferenceEngine::Core::SetConfig`
-
-@snippet snippets/Migration_CoreAPI.cpp part7
-
-> **NOTE**: If `deviceName` is omitted as the last argument, configuration is set for all Inference Engine devices.
-
-5. Migrate from loading the network to a particular plugin:
-
-@snippet snippets/Migration_CoreAPI.cpp part8
-
-to `InferenceEngine::Core::LoadNetwork` to a particular device:
-
-@snippet snippets/Migration_CoreAPI.cpp part9
-
-After you have an instance of `InferenceEngine::ExecutableNetwork`, all other steps are as usual.
diff --git a/docs/IE_DG/OnnxImporterTutorial.md b/docs/IE_DG/OnnxImporterTutorial.md
deleted file mode 100644
index f4538633a7e805..00000000000000
--- a/docs/IE_DG/OnnxImporterTutorial.md
+++ /dev/null
@@ -1,67 +0,0 @@
-# ONNX* Importer API Tutorial {#openvino_docs_IE_DG_OnnxImporterTutorial}
-
-> **NOTE**: This tutorial is deprecated. Since OpenVINO™ 2020.4 version, Inference Engine enables reading ONNX models via the Inference Engine Core API
-> and there is no need to use directly the low-level ONNX* Importer API anymore. 
-> To read ONNX\* models, it's recommended to use the `Core::ReadNetwork()` method that provide a uniform way to read models from IR or ONNX format.
-
-This tutorial demonstrates how to use the ONNX\* Importer API.
-This API makes it possible to create an nGraph `Function` object from an imported ONNX model.
-
-All functions of the ONNX Importer API are in the [onnx.hpp][onnx_header] header file.
-
-Two categories of API functions:
-* Helper functions that check which ONNX ops are supported in a current version of the ONNX Importer
-* Functions that read ONNX models from a stream or file and result in an nGraph function, which can be executed using the Inference Engine
-
-## Check Which ONNX Ops Are Supported
-
-To list all supported ONNX ops in a specific version and domain, use the `get_supported_operators` 
-as shown in the example below:
-
-@snippet snippets/OnnxImporterTutorial0.cpp part0
-
-The above code produces a list of all the supported operators for the `version` and `domain` you specified and outputs a list similar to this:
-```cpp
-Abs
-Acos
-...
-Xor
-```
-
-To determine whether a specific ONNX operator in a particular version and domain is supported by the importer, use the `is_operator_supported` function as shown in the example below:
-
-@snippet snippets/OnnxImporterTutorial1.cpp part1
-
-## Import ONNX Model
-
-To import an ONNX model, use the `import_onnx_model` function.
-The method has two overloads:
-* <a href="#stream">`import_onnx_model` takes a stream as an input</a>, for example, file stream, memory stream
-* <a href="#path">`import_onnx_model` takes a file path as an input</a>
-
-Refer to the sections below for details.
-
-> **NOTE**: The examples below use the ONNX ResNet50 model, which is available at the [ONNX Model Zoo][onnx_model_zoo]:
-> ```bash
-> $ wget https://s3.amazonaws.com/download.onnx/models/opset_8/resnet50.tar.gz
-> $ tar -xzvf resnet50.tar.gz
-> ```
-
-Once you create the `ng_function`, you can use it to run computation on the Inference Engine.
-As it was shown in [Build a Model with nGraph Library](../nGraph_DG/build_function.md), `std::shared_ptr<ngraph::Function>` can be transformed into a `CNNNetwork`.
-
-
-### <a name="stream">Stream as Input</a>
-
-The code below shows how to convert the ONNX ResNet50 model to the nGraph function using `import_onnx_model` with the stream as an input:
-
-@snippet snippets/OnnxImporterTutorial2.cpp part2
-
-### <a name="path">Filepath as Input</a>
-
-The code below shows how to convert the ONNX ResNet50 model to the nGraph function using `import_onnx_model` with the filepath as an input:
-
-@snippet snippets/OnnxImporterTutorial3.cpp part3
-
-[onnx_header]: https://github.com/NervanaSystems/ngraph/blob/master/src/ngraph/frontend/onnx_import/onnx.hpp
-[onnx_model_zoo]: https://github.com/onnx/models
diff --git a/docs/IE_DG/supported_plugins/FPGA.md b/docs/IE_DG/supported_plugins/FPGA.md
deleted file mode 100644
index 63ae6e62ed7be0..00000000000000
--- a/docs/IE_DG/supported_plugins/FPGA.md
+++ /dev/null
@@ -1,22 +0,0 @@
-FPGA Plugin {#openvino_docs_IE_DG_supported_plugins_FPGA}
-===========
-
-## Product Change Notice
-Intel® Distribution of OpenVINO™ toolkit for Intel® Vision Accelerator Design with an Intel® Arria® 10 FPGA and the Intel® Programmable Acceleration Card with Intel® Arria® 10 GX FPGA
-
-<table>
-  <tr>
-    <td><strong>Change Notice Begins</strong></td>
-    <td>July 2020</td>
-  </tr>
-  <tr>
-    <td><strong>Change Date</strong></td>
-    <td>October 2020</td>
-  </tr>
-</table> 
-
-Intel will be transitioning to the next-generation programmable deep-learning solution based on FPGAs in order to increase the level of customization possible in FPGA deep-learning. As part of this transition, future standard releases (i.e., non-LTS releases) of Intel® Distribution of OpenVINO™ toolkit will no longer include the Intel® Vision Accelerator Design with an Intel® Arria® 10 FPGA and the Intel® Programmable Acceleration Card with Intel® Arria® 10 GX FPGA.
-
-Intel® Distribution of OpenVINO™ toolkit 2020.3.X LTS release will continue to support Intel® Vision Accelerator Design with an Intel® Arria® 10 FPGA and the Intel® Programmable Acceleration Card with Intel® Arria® 10 GX FPGA. For questions about next-generation programmable deep-learning solutions based on FPGAs, please talk to your sales representative or contact us to get the latest FPGA updates.
-
-For documentation for the FPGA plugin available in previous releases of Intel® Distribution of OpenVINO™ toolkit with FPGA Support, see documentation for the [2020.4 version](https://docs.openvinotoolkit.org/2020.4/openvino_docs_IE_DG_supported_plugins_FPGA.html) and lower.
\ No newline at end of file
diff --git a/docs/doxygen/doxygen-ignore.txt b/docs/doxygen/doxygen-ignore.txt
index 0be7a70dc06342..7f963ac63e71de 100644
--- a/docs/doxygen/doxygen-ignore.txt
+++ b/docs/doxygen/doxygen-ignore.txt
@@ -22,6 +22,8 @@ inference-engine/include/vpu/vpu_config.hpp
 inference-engine/include/vpu/vpu_plugin_config.hpp
 openvino/docs/benchmarks/performance_int8_vs_fp32.md
 openvino/docs/get_started/get_started_macos.md
+openvino/docs/optimization_guide/dldt_optimization_guide.md
+openvino/docs/IE_DG/ShapeInference.md
 inference-engine/include/details/ie_so_pointer.hpp
 inference-engine/include/ie_compound_blob.h
 inference-engine/include/ie_data.h
diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index 120492baef0c2d..b8581444627662 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -291,11 +291,9 @@ limitations under the License.
                     <tab type="user" title="Custom ONNX operators" url="@ref openvino_docs_IE_DG_Extensibility_DG_Custom_ONNX_Ops"/>
                 </tab>
                 <tab type="user" title="Integrate the Inference Engine with Your Application" url="@ref openvino_docs_IE_DG_Integrate_with_customer_application_new_API"/>
-                <tab type="user" title="[DEPRECATED] Migration from Inference Engine Plugin API to Core API" url="@ref openvino_docs_IE_DG_Migration_CoreAPI"/>
                 <tab type="user" title="Introduction to Performance Topics" url="@ref openvino_docs_IE_DG_Intro_to_Performance"/>
                 <tab type="user" title="Inference Engine Python* API Overview" url="@ref openvino_inference_engine_ie_bridges_python_docs_api_overview"/>
                 <tab type="user" title="Read an ONNX model" url="@ref openvino_docs_IE_DG_ONNX_Support"/>
-                <tab type="user" title="[DEPRECATED] Import an ONNX model" url="@ref openvino_docs_IE_DG_OnnxImporterTutorial"/>
                 <tab type="user" title="Using Dynamic Batching Feature" url="@ref openvino_docs_IE_DG_DynamicBatching"/>
                 <tab type="user" title="Using Static Shape Infer Feature" url="@ref openvino_docs_IE_DG_ShapeInference"/>
                 <tab type="usergroup" title="Using Bfloat16 Inference" url="@ref openvino_docs_IE_DG_Bfloat16Inference">
@@ -311,7 +309,6 @@ limitations under the License.
                         <tab type="user" title="RemoteBlob API of GPU Plugin" url="@ref openvino_docs_IE_DG_supported_plugins_GPU_RemoteBlob_API"/>
                     </tab>
                     <tab type="user" title="CPU Plugin" url="@ref openvino_docs_IE_DG_supported_plugins_CPU"/>
-                    <tab type="user" title="[DEPRECATED] FPGA Plugin" url="@ref openvino_docs_IE_DG_supported_plugins_FPGA"/>
                     <tab type="usergroup" title="VPU Plugins" url="@ref openvino_docs_IE_DG_supported_plugins_VPU">
                         <tab type="user" title="MYRIAD Plugin " url="@ref openvino_docs_IE_DG_supported_plugins_MYRIAD"/>
                         <tab type="user" title="HDDL Plugin " url="@ref openvino_docs_IE_DG_supported_plugins_HDDL"/>
diff --git a/docs/snippets/CMakeLists.txt b/docs/snippets/CMakeLists.txt
index 1d2a20eea0a3d7..48edae1e832547 100644
--- a/docs/snippets/CMakeLists.txt
+++ b/docs/snippets/CMakeLists.txt
@@ -26,17 +26,6 @@ if(NOT OpenCV_FOUND)
                              "${CMAKE_CURRENT_SOURCE_DIR}/ShapeInference.cpp")
 endif()
 
-# ONNX importer related files
-if(NOT NGRAPH_ONNX_IMPORT_ENABLE)
-    list(REMOVE_ITEM SOURCES "${CMAKE_CURRENT_SOURCE_DIR}/OnnxImporterTutorial0.cpp"
-                             "${CMAKE_CURRENT_SOURCE_DIR}/OnnxImporterTutorial1.cpp"
-                             "${CMAKE_CURRENT_SOURCE_DIR}/OnnxImporterTutorial2.cpp"
-                             "${CMAKE_CURRENT_SOURCE_DIR}/OnnxImporterTutorial3.cpp")
-endif()
-
-# remove snippets for deprecated / removed API
-list(REMOVE_ITEM SOURCES "${CMAKE_CURRENT_SOURCE_DIR}/Migration_CoreAPI.cpp")
-
 # requires mfxFrameSurface1 and MSS API
 list(REMOVE_ITEM SOURCES "${CMAKE_CURRENT_SOURCE_DIR}/dldt_optimization_guide2.cpp"
                          "${CMAKE_CURRENT_SOURCE_DIR}/dldt_optimization_guide3.cpp"
diff --git a/docs/snippets/Migration_CoreAPI.cpp b/docs/snippets/Migration_CoreAPI.cpp
deleted file mode 100644
index fd89803093b307..00000000000000
--- a/docs/snippets/Migration_CoreAPI.cpp
+++ /dev/null
@@ -1,48 +0,0 @@
-#include <ie_core.hpp>
-
-int main() {
-std::string deviceName = "Device name";
-//! [part0]
-InferenceEngine::InferencePlugin plugin = InferenceEngine::PluginDispatcher({ FLAGS_pp }).getPluginByDevice(FLAGS_d);
-//! [part0]
-
-//! [part1]
-InferenceEngine::Core core;
-//! [part1]
-
-//! [part2]
-InferenceEngine::CNNNetReader network_reader;
-network_reader.ReadNetwork(fileNameToString(input_model));
-network_reader.ReadWeights(fileNameToString(input_model).substr(0, input_model.size() - 4) + ".bin");
-InferenceEngine::CNNNetwork network = network_reader.getNetwork();
-//! [part2]
-
-//! [part3]
-InferenceEngine::CNNNetwork network = core.ReadNetwork(input_model);
-//! [part3]
-
-//! [part4]
-InferenceEngine::CNNNetwork network = core.ReadNetwork("model.onnx");
-//! [part4]
-
-//! [part5]
-plugin.AddExtension(std::make_shared<Extensions::Cpu::CpuExtensions>());
-//! [part5]
-
-//! [part6]
-core.AddExtension(std::make_shared<Extensions::Cpu::CpuExtensions>(), "CPU");
-//! [part6]
-
-//! [part7]
-core.SetConfig({{PluginConfigParams::KEY_CONFIG_FILE, FLAGS_c}}, "GPU");
-//! [part7]
-
-//! [part8]
-auto execNetwork = plugin.LoadNetwork(network, { });
-//! [part8]
-
-//! [part9]
-auto execNetwork = core.LoadNetwork(network, deviceName, { });
-//! [part9]
-return 0;
-}
diff --git a/docs/snippets/OnnxImporterTutorial0.cpp b/docs/snippets/OnnxImporterTutorial0.cpp
deleted file mode 100644
index cf434622cb9395..00000000000000
--- a/docs/snippets/OnnxImporterTutorial0.cpp
+++ /dev/null
@@ -1,19 +0,0 @@
-#include <ie_core.hpp>
-#include <ngraph/ngraph.hpp>
-#include "onnx_import/onnx.hpp"
-#include <iostream>
-#include <set>
-
-int main() {
-//! [part0]
-const std::int64_t version = 12;
-const std::string domain = "ai.onnx";
-const std::set<std::string> supported_ops = ngraph::onnx_import::get_supported_operators(version, domain);
-
-for(const auto& op : supported_ops)
-{
-    std::cout << op << std::endl;
-}
-//! [part0]
-return 0;
-}
diff --git a/docs/snippets/OnnxImporterTutorial1.cpp b/docs/snippets/OnnxImporterTutorial1.cpp
deleted file mode 100644
index 60122f1a1ea025..00000000000000
--- a/docs/snippets/OnnxImporterTutorial1.cpp
+++ /dev/null
@@ -1,15 +0,0 @@
-#include <ie_core.hpp>
-#include <ngraph/ngraph.hpp>
-#include "onnx_import/onnx.hpp"
-
-int main() {
-//! [part1]
-const std::string op_name = "Abs";
-const std::int64_t version = 12;
-const std::string domain = "ai.onnx";
-const bool is_abs_op_supported = ngraph::onnx_import::is_operator_supported(op_name, version, domain);
-
-std::cout << "Abs in version 12, domain `ai.onnx`is supported: " << (is_abs_op_supported ? "true" : "false") << std::endl;
-//! [part1]
-return 0;
-}
diff --git a/docs/snippets/OnnxImporterTutorial2.cpp b/docs/snippets/OnnxImporterTutorial2.cpp
deleted file mode 100644
index 00ce2949a1d163..00000000000000
--- a/docs/snippets/OnnxImporterTutorial2.cpp
+++ /dev/null
@@ -1,29 +0,0 @@
-#include <ie_core.hpp>
-#include <ngraph/ngraph.hpp>
-#include "onnx_import/onnx.hpp"
-#include <iostream>
-#include <fstream>
-
-int main() {
-//! [part2]
- const char * resnet50_path = "resnet50/model.onnx";
- std::ifstream resnet50_stream(resnet50_path);
- if (resnet50_stream.is_open())
- {
-     try
-     {
-         const std::shared_ptr<ngraph::Function> ng_function = ngraph::onnx_import::import_onnx_model(resnet50_stream);
-
-         // Check shape of the first output, for example
-         std::cout << ng_function->get_output_shape(0) << std::endl;
-         // The output is Shape{1, 1000}
-     }
-     catch (const ngraph::ngraph_error& error)
-     {
-         std::cout << "Error when importing ONNX model: " << error.what() << std::endl;
-     }
- }
- resnet50_stream.close();
-//! [part2]
-return 0;
-}
diff --git a/docs/snippets/OnnxImporterTutorial3.cpp b/docs/snippets/OnnxImporterTutorial3.cpp
deleted file mode 100644
index 6fc1e1b59de907..00000000000000
--- a/docs/snippets/OnnxImporterTutorial3.cpp
+++ /dev/null
@@ -1,12 +0,0 @@
-#include <ie_core.hpp>
-#include <ngraph/ngraph.hpp>
-#include "onnx_import/onnx.hpp"
-#include <iostream>
-
-int main() {
-//! [part3]
-const char * resnet50_path = "resnet50/model.onnx";
-const std::shared_ptr<ngraph::Function> ng_function = ngraph::onnx_import::import_onnx_model(resnet50_path);
-//! [part3]
-return 0;
-}

From 487aff7d7ed98bcc0684ada35cdd3a04e120a94f Mon Sep 17 00:00:00 2001
From: Eugeny Volosenkov <eugeny.volosenkov@intel.com>
Date: Mon, 21 Jun 2021 09:53:09 +0300
Subject: [PATCH 142/178] [Attention OCR] Document model support (#6153)

* Add doc How to convert AttentionOCR

* Add converting

* Add converting 2

* Add converting 3

* Fix document

* Fix document1

* Fix document1

* Fix document1

* Fix ie_docs

* Fix model/path

* Add link to Convert_Model_From_TensorFlow.md

* fix doc

* Fix documentation
---
 .../Convert_Model_From_TensorFlow.md          |  2 +-
 .../Convert_AttentionOCR_From_Tensorflow.md   | 35 +++++++++++++++++++
 docs/doxygen/ie_docs.xml                      |  1 +
 3 files changed, 37 insertions(+), 1 deletion(-)
 create mode 100644 docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_AttentionOCR_From_Tensorflow.md

diff --git a/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_TensorFlow.md b/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_TensorFlow.md
index c4721cdead07ee..7e29a7668b2f24 100644
--- a/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_TensorFlow.md
+++ b/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_TensorFlow.md
@@ -161,7 +161,7 @@ Where `HEIGHT` and `WIDTH` are the input images height and width for which the m
 * [GNMT](https://github.com/tensorflow/nmt) topology can be converted using [these instructions](tf_specific/Convert_GNMT_From_Tensorflow.md).
 * [BERT](https://github.com/google-research/bert) topology can be converted using [these instructions](tf_specific/Convert_BERT_From_Tensorflow.md).
 * [XLNet](https://github.com/zihangdai/xlnet) topology can be converted using [these instructions](tf_specific/Convert_XLNet_From_Tensorflow.md).
-
+* [Attention OCR](https://github.com/emedvedev/attention-ocr) topology can be converted using [these instructions](tf_specific/Convert_AttentionOCR_From_Tensorflow.md).
   
 
 ## Loading Non-Frozen Models to the Model Optimizer <a name="loading-nonfrozen-models"></a>
diff --git a/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_AttentionOCR_From_Tensorflow.md b/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_AttentionOCR_From_Tensorflow.md
new file mode 100644
index 00000000000000..90e94677dd7f33
--- /dev/null
+++ b/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_AttentionOCR_From_Tensorflow.md
@@ -0,0 +1,35 @@
+# Convert TensorFlow* Attention OCR Model to Intermediate Representation {#openvino_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_AttentionOCR_From_Tensorflow}
+
+This tutorial explains how to convert the Attention OCR (AOCR) model from the [TensorFlow* Attention OCR repository](https://github.com/emedvedev/attention-ocr) to the Intermediate Representation (IR).
+
+## Extract Model from `aocr` Library
+
+The easiest way to get an AOCR model is to download `aocr` Python\* library:
+```
+pip install git+https://github.com/emedvedev/attention-ocr.git@master#egg=aocr
+```
+This library contains a pretrained model and allows to train and run AOCR using the command line. After installing `aocr`, you can extract the model:
+```
+aocr export --format=frozengraph model/path/
+```
+After this step you can find the model in model/path/ folder.
+
+## Convert the TensorFlow* AOCR Model to IR
+
+The original AOCR model contains data preprocessing which consists of the following steps:
+* Decoding input data to binary format where input data is an image represented as a string.
+* Resizing binary image to working resolution.
+
+After that, the resized image is sent to the convolution neural network (CNN). The Model Optimizer does not support image decoding so you should cut of preprocessing part of the model using '--input' command line parameter. 
+```sh
+python3 path/to/model_optimizer/mo_tf.py \
+--input_model=model/path/frozen_graph.pb \
+--input="map/TensorArrayStack/TensorArrayGatherV3:0[1 32 86 1]" \
+--output "transpose_1,transpose_2" \
+--output_dir path/to/ir/
+```
+
+Where:
+* `map/TensorArrayStack/TensorArrayGatherV3:0[1 32 86 1]` - name of node producing tensor after preprocessing.
+* `transpose_1` - name of the node producing tensor with predicted characters.
+* `transpose_2` - name of the node producing tensor with predicted characters probabilties
\ No newline at end of file
diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index b8581444627662..8841672b430623 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -41,6 +41,7 @@ limitations under the License.
                             <tab type="user" title="Convert TensorFlow* XLNet Model to the Intermediate Representation" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_XLNet_From_Tensorflow"/>
                             <tab type="user" title="Converting TensorFlow* Wide and Deep Models from TensorFlow" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_WideAndDeep_Family_Models"/>
                             <tab type="user" title="Converting EfficientDet Models from TensorFlow" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_EfficientDet_Models"/>
+                            <tab type="user" title="Converting Attention OCR Model from TensorFlow" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_AttentionOCR_From_Tensorflow"/>
                         </tab>
                         <tab type="usergroup" title="Converting a MXNet* Model" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_MxNet">
                             <tab type="user" title="Converting a Style Transfer Model from MXNet" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_mxnet_specific_Convert_Style_Transfer_From_MXNet"/>

From cd0ded7ee39ad47f42ad204f910b5d6871fc643a Mon Sep 17 00:00:00 2001
From: Alina Kladieva <alina.kladieva@intel.com>
Date: Mon, 21 Jun 2021 12:38:49 +0300
Subject: [PATCH 143/178] Exclude test failing on Ubuntu20 precommit scope
 (#6202)

---
 .../myriad/shared_tests_instances/skip_tests_config.cpp       | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/skip_tests_config.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/skip_tests_config.cpp
index fc2b4b2d69b686..7badf1a9677a63 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/skip_tests_config.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/skip_tests_config.cpp
@@ -44,6 +44,8 @@ std::vector<std::string> disabledTestPatterns() {
         R"(.*IS=\(16\.16\.96\)\(96\)_eltwiseOpType=FloorMod_secondaryInputType=PARAMETER_opType=VECTOR_netPRC=FP32.*)",
         // TODO: Issue CVS-57108
         R"(.*QueryNetworkHETEROWithMULTINoThrow_V10.*)",
-        R"(.*QueryNetworkMULTIWithHETERONoThrow_V10.*)"
+        R"(.*QueryNetworkMULTIWithHETERONoThrow_V10.*)",
+        // TODO: Issue CVS-58162
+        R"(.*HoldersTestOnImportedNetwork\.CreateRequestWithCoreRemoved.*)"
     };
 }

From d71ac2e2c151557a740d87cb2e575a420a617994 Mon Sep 17 00:00:00 2001
From: Irina Burgun <irinax.burgun@intel.com>
Date: Mon, 21 Jun 2021 13:07:34 +0300
Subject: [PATCH 144/178] Added filters block to report template (#6201)

---
 .../layer_tests_summary/summarize.py          |   1 +
 .../layer_tests_summary/template/filters.js   | 209 ++++++++++++++++
 .../template/report_template.html             | 236 +++++++++++-------
 .../layer_tests_summary/template/style.css    |  88 +++++++
 4 files changed, 439 insertions(+), 95 deletions(-)
 create mode 100644 inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/filters.js
 create mode 100644 inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/style.css

diff --git a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/summarize.py b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/summarize.py
index 4419c4ffeb89ad..1245ae8a0fb9bf 100644
--- a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/summarize.py
+++ b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/summarize.py
@@ -126,6 +126,7 @@ def collect_statistic(root: ET.Element, is_conformance_mode: bool):
         pass_rate_avg[device.tag] = round(float(pass_rate_avg[device.tag]), 1)
         general_pass_rate[device.tag] = general_passed_tests[device.tag] * 100 / general_test_count[device.tag]
         general_pass_rate[device.tag] = round(float(general_pass_rate[device.tag]), 1)
+        trusted_ops[device.tag] = round(float(trusted_ops[device.tag] * 100 / len(results[device.tag])), 1)
 
     logger.info("Test number comparison between devices is started")
     for op in op_res:
diff --git a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/filters.js b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/filters.js
new file mode 100644
index 00000000000000..af832efa3cfb3b
--- /dev/null
+++ b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/filters.js
@@ -0,0 +1,209 @@
+deviceList = [];
+$(document).ready(function () {
+    var opsets = {};
+    LoadOpsetNumbers();
+    LoadDevices();
+
+    $("#filters").submit(function (event) {
+        event.preventDefault();
+        filterTable();
+    });
+    $('#reset').click(function () {
+        $('#opsetNumber').val(0);
+        $('#operationName').val('');
+        $('#status').prop("disabled", true).val(0);
+        $('#devices').val(0);
+        $('#references').val(0);
+        filterTable();
+    });
+    $('#devices').on('change', function () {
+        if (this.value == 0) {
+            $('#status').prop("disabled", true).val(0);
+        } else {
+            $('#status').prop("disabled", false);
+        };
+    });
+
+});
+
+function LoadOpsetNumbers() {
+    var data = [];
+
+    $('#data th[scope="row"]').each(function () {
+
+        num = $(this).text().split("-")[1];
+        if (data.indexOf(num) < 0) {
+            data.push(num);
+        }
+    });
+    data.sort();
+    data = $.map(data, function (item) {
+        return "<option value=" + item + ">" + item + "</option>";
+    });
+    $("#opsetNumber").html('<option value="0">All</option>');
+    $("#opsetNumber").append(data.join(""));
+}
+
+function LoadDevices() {
+    var data = [];
+
+    $('.table-dark.device').each(function () {
+        if (data.indexOf($(this).text()) < 0) {
+            data.push($(this).text());
+        }
+    });
+    data.sort();
+    deviceList = data;
+    data = $.map(data, function (item) {
+        return "<option value=" + item + ">" + item + "</option>";
+    });
+    $("#devices").html('<option value="0">All</option>');
+    $("#devices").append(data.join(""));
+}
+
+function filterTable() {
+    device = $("#devices").val();
+    if (device == 0) {
+        $("#report td.value, #report td.table-secondary, #report td.table-primary, #report th.table-dark.device").show();
+    } else {
+        $("#report td.value, #report td.table-secondary, #report td.table-primary, #report th.table-dark.device").filter(function () {
+            $(this).toggle($(this).hasClass(device))
+        });
+    }
+    opsetNumber = $("#opsetNumber").val();
+    operationName = $('#operationName').val().trim();
+    status = $('#status').val();
+    references = $('#references').val();
+
+    $("#report #data tr").show();
+    $('#report').show();
+    $('#message').hide();
+    if (opsetNumber != 0) {
+        $("#report #data tr").filter(function () {
+            $(this).toggle(checkVersion($(this), opsetNumber));
+        });
+    }
+
+    if (operationName) {
+        $("#report #data tr:not(:hidden)").filter(function () {
+            $(this).toggle($(this).find('th').text().split("-")[0].toLowerCase().indexOf(operationName.toLowerCase()) > -1);
+        });
+    }
+
+    if (references != 0) {
+        if (references == 'nv') {
+            $("#report #data tr:not(:hidden)").filter(function () {
+                $(this).toggle($(this).find('th').hasClass("colorRed"))
+            });
+        } else {
+            $("#report #data tr:not(:hidden)").filter(function () {
+                $(this).toggle(!$(this).find('th').hasClass("colorRed"));
+            });
+        }
+    }
+    if (status != 0) {
+        if (status == 'p') {
+            $("#report #data tr:not(:hidden)").filter(function () {
+                $(this).toggle($(this).find('.value:visible[crashed="0"][failed="0"][skipped="0"]').length > 0)
+            });
+        } else if (status == 'f') {
+            $("#report #data tr:not(:hidden)").filter(function () {
+                $(this).toggle($(this).find('.value:visible[passed="0"][crashed="0"][skipped="0"]').length > 0)
+            });
+        } else if (status == 'c') {
+            $("#report #data tr:not(:hidden)").filter(function () {
+                $(this).toggle($(this).find('.value:visible[passed="0"][failed="0"][skipped="0"]').length > 0)
+            });
+        } else if (status == 's') {
+            $("#report #data tr:not(:hidden)").filter(function () {
+                $(this).toggle($(this).find('.value:visible[passed="0"][failed="0"][crashed="0"]').length > 0)
+            });
+        } else { // No tests
+
+            $("#report #data tr:not(:hidden)").filter(function () {
+                $(this).toggle($(this).find('.table-secondary:visible').length > 0)
+            });
+        }
+    }
+
+
+    if ($("#report #data tr").length == $("#report #data tr:hidden").length) {
+        $('#report').hide();
+        $('#message').show();
+    } else {
+        calculateStatistics(device);
+    }
+}
+
+function checkVersion(element, opsetNumber) {
+    var name = element.find('th').text().split("-")[0];
+    var version = element.find('th').text().split("-")[1];
+    if (version > opsetNumber) {
+        return false;
+    } else {
+        var versions = [];
+        $('#report #data tr th[name^="' + name + '-"]').each(function () {
+            if ($(this).text().split('-')[1] <= opsetNumber) {
+                versions.push(+$(this).text().split('-')[1]);
+            }
+        });
+        return version == Math.max.apply(null, versions);
+    }
+}
+
+function calculateStatistics() {
+    if (device != 0) {
+        calculateColumnStatistics(device);
+    } else {
+        deviceList.map((el) => calculateColumnStatistics(el))
+    }
+}
+
+function calculateColumnStatistics(device) {
+    // total
+    total = $("#report #data tr:not(:hidden)").length;
+    $('#statistic .table-primary[scope="row"] i').text(total);
+    // trusted op
+    count_trasted_op = $("#report #data tr:not(:hidden) ." + device + ".value[value^='100'][crashed='0'][failed='0'][skipped='0']").length;
+    all_operations = $("#report #data tr:not(:hidden) .value." + device).length;
+    if (!all_operations) {
+        trasted_op = "---";
+    } else {
+        trasted_op = (count_trasted_op * 100 / all_operations).toFixed(1) + ' %';
+    }
+    $('#statistic .table-primary.' + device + '.trusted-ops').text(trasted_op);
+    $('#statistic .table-primary.' + device + '.test_total').text(all_operations || 0);
+
+    // tested op_counter
+    tested_op_count = 0;
+    passed_tested_op_count = 0;
+    $("#report #data tr:not(:hidden) ." + device + ".value span").each(function () {
+        text = $(this).text().split(':')[1];
+        if ($(this).hasClass('green')) {
+            passed_tested_op_count += +text;
+        }
+        tested_op_count += +text;
+    });
+
+    // General Pass Rate
+    if (tested_op_count == 0) {
+        $('#statistic .table-primary.' + device + '.general_pass_rate').text('---');
+
+    } else {
+        general_pass_rate = (passed_tested_op_count * 100 / tested_op_count).toFixed(1) + ' %';
+        $('#statistic .table-primary.' + device + '.general_pass_rate').text(general_pass_rate);
+    }
+    $('#statistic .table-primary.' + device + '.tested-ops_count').text(tested_op_count);
+
+    // AVG Pass Rate
+    sum_pass_rate = 0;
+    $("#report #data tr:not(:hidden) ." + device + ".value").each(function () {
+        sum_pass_rate += +$(this).attr('value');
+    });
+    if (all_operations == 0) {
+        $('#statistic .table-primary.' + device + '.avg_pass_rate').text('---');
+    } else {
+        avg_pass_rate = (sum_pass_rate / all_operations).toFixed(1) + ' %';
+        $('#statistic .table-primary.' + device + '.avg_pass_rate').text(avg_pass_rate);
+    }
+}
diff --git a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/report_template.html b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/report_template.html
index 7d6f751f91701a..12bb09fcd8167c 100644
--- a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/report_template.html
+++ b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/report_template.html
@@ -1,5 +1,6 @@
 <!doctype html>
 <html lang="en">
+
 <head>
     <!-- Required meta tags -->
     <meta charset="utf-8">
@@ -7,106 +8,151 @@
 
     <!-- Bootstrap CSS -->
     <link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/bootstrap/4.0.0/css/bootstrap.min.css"
-          integrity="sha384-Gn5384xqQ1aoWXA+058RXPxPg6fy4IWvTNh0E263XmFcJlSAwiGgFAW/dAiS6JXm" crossorigin="anonymous">
-    <link rel="stylesheet" type="text/css" href="extensions/filter-control/bootstrap-table-filter-control.css">
+        integrity="sha384-Gn5384xqQ1aoWXA+058RXPxPg6fy4IWvTNh0E263XmFcJlSAwiGgFAW/dAiS6JXm" crossorigin="anonymous">
+    <link rel="stylesheet" href="template/style.css" />
+    <script src="https://code.jquery.com/jquery-3.2.1.slim.min.js"
+        integrity="sha384-KJ3o2DKtIkvYIK3UENzmM7KCkRr/rE9/Qpg6aAZGJwFDMVNA/GpGFF93hXpG5KkN"
+        crossorigin="anonymous"></script>
+    <script src="template/filters.js"></script>
     <title>Report</title>
 </head>
+
 <body>
-<script src="extensions/filter-control/bootstrap-table-filter-control.js"></script>
-<!-- Optional JavaScript -->
-<!-- jQuery first, then Popper.js, then Bootstrap JS -->
-<!--
-<script src="https://code.jquery.com/jquery-3.2.1.slim.min.js"
-        integrity="sha384-KJ3o2DKtIkvYIK3UENzmM7KCkRr/rE9/Qpg6aAZGJwFDMVNA/GpGFF93hXpG5KkN"
-        crossorigin="anonymous"></script>
-<script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.12.9/umd/popper.min.js"
-        integrity="sha384-ApNbgh9B+Y1QKtv3Rn7W3mgPxhU9K/ScQsAP7hUibX39j7fakFPskvXusvfa0b4Q"
-        crossorigin="anonymous"></script>
-<script src="https://maxcdn.bootstrapcdn.com/bootstrap/4.0.0/js/bootstrap.min.js"
-        integrity="sha384-JZR6Spejh4U02d8jOt6vLEHfe/JQGiRRSQQxSfFWpi1MquVdAyjUar5+76PVCmYl"
-        crossorigin="anonymous"></script>
--->
+    <div class="main">
+        <h2>Operations coverage summary: {{report_tag}} {{ timestamp }}</h2>
+        <div class="legend">
+            <div>
+                <span class="border colorRed">Acosh-4</span><span>Not verified Ngraph references</span>
+            </div>
+
+            <div>
+                <span class="table-primary border"></span><span>Collected statistic info</span>
+            </div>
+            <div>
+                <span class="table-secondary border">N/A</span><span>No Tests</span>
+            </div>
+            <div>
+                <span><b>Status:</b></span>
+                <span class="green">P:85</span><span>Passed</span>
+                <span class="red">F:0</span><span>Failed</span>
+                <span class="grey">S:2</span><span>Skipped</span>
+                <span class="dark">C:0</span><span>Crashed</span>
+            </div>
+        </div>
+    </div>
+    <!-- Filters block -->
+    <div class="filters">
+        <form id="filters">
+            <div class="form-group">
+                <label for="operationName"><b>Operation Name</b></label>
+                <input id="operationName" type="text" class="form-control" />
+            </div>
+            <div class="form-group">
+                <label for="opsetNumber"><b>Opset Number</b></label>
+                <select id="opsetNumber" class="form-control"></select>
+            </div>
+            <div class="form-group">
+                <label for="references"><b>Ngraph references</b></label>
+                <select id="references" class="form-control">
+                    <option value="0">All</option>
+                    <option value="v">Verified</option>
+                    <option value="nv">Not verified</option>
+                </select>
+            </div>
+            <div class="form-group">
+                <label for="devices"><b>Devices</b></label>
+                <select id="devices" class="form-control"></select>
+            </div>
+
 
-<hr class="my-4">
-<h2>Operations coverage summary: {{report_tag}} {{ timestamp }}</h2>
-<hr class="my-4">
-<table class="table table-hover" id="legend">
-    <thead>
-    <tr>
-        <th scope="col"><span style="color: red">"Operation_name"-"opset_version"</span></th>
-        <th><span style="color: black"> Not verified Ngraph references</span></th>
-    </tr>
-    <tr>
-        <th class="table-secondary" scope="col"></th>
-        <th>Collected statistic info</th>
-    </tr>
-    </thead>
-</table>
-<table class="table table-hover" id="report">
-    <thead>
-    <tr>
-        <th class="table-primary" scope="col" style="position: sticky; top: 0">Operation</th>
-        {% for d in devices -%}
-        <th class="table-primary" style="position: sticky; top: 0">{{ d }}</th>
-        {% endfor %}
-    </tr>
-    </thead>
-    <tbody>
-        <tr>
-        <th class="table-secondary" scope="row">Total: {{ordered_ops|length}}</th>
-        {% for d in devices -%}
-        <td class="table-secondary" >{{results[d]|length}}</td>
-        {% endfor %}
-    </tr>
-        <tr>
-        <th class="table-secondary" scope="row">Trusted op counter (passrate=100%):</th>
-        {% for d in devices -%}
-        <td class="table-secondary" >{{trusted_ops[d]}}</td>
-        {% endfor %}
-    </tr>
-        <tr>
-        <th class="table-secondary" scope="row">Tested op counter:</th>
-        {% for d in devices -%}
-        <td class="table-secondary" >{{general_test_count[d]}}</td>
-        {% endfor %}
-    </tr>
-        <tr>
-        <th class="table-secondary" scope="row">AVG passrate per op (=sum_pass_rates/covered_ops_num):</th>
-        {% for d in devices -%}
-        <td class="table-secondary" >{{pass_rate_avg[d]}}%</td>
-        {% endfor %}
-    </tr>
-        <tr>
-        <th class="table-secondary" scope="row">General passrate (=passed_tests/all_tests):</th>
-        {% for d in devices -%}
-        <td class="table-secondary" >{{general_pass_rate[d]}}%</td>
-        {% endfor %}
-    </tr>
-    </tbody>
-    <tbody>
-    {% for op in ordered_ops -%}
-    <tr>
-        {% if op in verified_operations -%}
-            <th scope="row">{{ op }}</th>
-        {% else -%}
-        <th scope="row"><span style="color: red">{{ op }}</span></th>
-        {% endif -%}
-        {% for d in devices -%}
-        {% if op in results[d] -%}
-            <td>
-            {{ results[d][op].passrate }}% (<span style="color: green">p:{{ results[d][op].passed }}</span>,
-               <span style="color: red">f:{{ results[d][op].failed }}</span>,s:{{ results[d][op].skipped }}</span>,
-               <span style="color: red">c:{{ results[d][op].crashed }}</span>)
-            </td>
-        {% else -%}
-        <td class="table-warning">No tests</td>
-        {% endif -%}
+            <div class="form-group">
+                <label for="status"><b>Status</b></label>
+                <select id="status" class="form-control" disabled>
+                    <option value="0">All</option>
+                    <option value="p">100% Passed</option>
+                    <option value="f">100% Failed</option>
+                    <option value="s">100% Skipped</option>
+                    <option value="c">100% Crashed</option>
+                    <option value="na">No tests</option>
+                </select>
+            </div>
+            <button type="submit" class="btn btn-primary">Apply filters</button>
+            <button type="button" class="btn btn-secondary" id="reset">Reset filters</button>
+        </form>
+    </div>
 
-        {% endfor %}
-    </tr>
-    {% endfor -%}
+    <!-- Results -->
+    <table class="table table-hover" id="report">
+        <thead>
+            <tr>
+                <th class="table-dark" scope="col" style="position: sticky; top: 97px">Operation</th>
+                {% for d in devices -%}
+                <th class="table-dark device {{ d }}" style="position: sticky; top: 97px">{{ d }}</th>
+                {% endfor %}
+            </tr>
+        </thead>
+        <tbody id="statistic">
+            <tr>
+                <th class="table-primary" scope="row">Total: <i>{{ordered_ops|length}}</i></th>
+                {% for d in devices -%}
+                <td class="table-primary {{ d }} test_total">{{results[d]|length}}</td>
+                {% endfor %}
+            </tr>
+            <tr>
+                <th class="table-primary" scope="row">Trusted op (passrate=100%):</th>
+                {% for d in devices -%}
+                <td class="table-primary  {{ d }} trusted-ops">{{trusted_ops[d]}} %</td>
+                {% endfor %}
+            </tr>
+            <tr>
+                <th class="table-primary" scope="row">Tested op counter:</th>
+                {% for d in devices -%}
+                <td class="table-primary  {{ d }} tested-ops_count">{{general_test_count[d]}}</td>
+                {% endfor %}
+            </tr>
+            <tr>
+                <th class="table-primary" scope="row">AVG passrate per op (=sum_pass_rates/covered_ops_num):</th>
+                {% for d in devices -%}
+                <td class="table-primary  {{ d }} avg_pass_rate">{{pass_rate_avg[d]}} %</td>
+                {% endfor %}
+            </tr>
+            <tr>
+                <th class="table-primary" scope="row">General passrate (=passed_tests/all_tests):</th>
+                {% for d in devices -%}
+                <td class="table-primary  {{ d }} general_pass_rate">{{general_pass_rate[d]}} %</td>
+                {% endfor %}
+            </tr>
+        </tbody>
+        <tbody id="data">
+            {% for op in ordered_ops -%}
+            <tr>
+                <th scope="row" {% if op not in verified_operations -%} class="colorRed" {% endif -%} name="{{ op }}">{{
+                    op }}</th>
 
-    </tbody>
-</table>
+                {% for d in devices -%}
+                {% if op in results[d] -%}
+                <td class="value {{ d }}" passed="{{ results[d][op].passed }}" failed="{{ results[d][op].failed }}"
+                    skipped="{{ results[d][op].skipped }}" crashed="{{ results[d][op].crashed }}"
+                    value="{{ results[d][op].passrate }}">
+                    {{ results[d][op].passrate }} %<br />
+                    <span class="green" title="Passed">P:{{ results[d][op].passed }}</span>
+                    <span class="red" title="Failed">F:{{ results[d][op].failed }}</span>
+                    <span class="grey" title="Skipped">S:{{ results[d][op].skipped }}</span>
+                    <span class="dark" title="Crashed">C:{{ results[d][op].crashed }}</span>
+                </td>
+                {% else -%}
+                <td class="table-secondary {{ d }}">N/A</td>
+                {% endif -%}
+
+                {% endfor %}
+            </tr>
+            {% endfor -%}
+
+        </tbody>
+    </table>
+    <div id="message" style="display:none">
+        There is no data related to selected filters. Please set new filters.
+    </div>
 </body>
+
 </html>
diff --git a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/style.css b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/style.css
new file mode 100644
index 00000000000000..11831d83b83e05
--- /dev/null
+++ b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/template/style.css
@@ -0,0 +1,88 @@
+body {
+    font-size: 14px;
+}
+.table td, .table th {
+    padding: .5em .75em;
+}
+.table td span {
+    font-size: 0.8em;
+}
+.value {
+    font-weight: 500;
+}
+.value span {
+   display:inline-block;
+   font-weight: 400;
+   padding: 1px 5px;
+   border-radius: 2px;
+   cursor: default;
+}
+ .green {
+    background: #0080002e;
+}
+.red {
+    background: #ff000038;
+}
+.grey {
+    background: #8080803d;
+}
+.dark {
+    background: #8b000040;
+}
+.filters {
+    background: #FFF;
+    padding: 5px 10px;
+    position: sticky;
+    top: 0;
+}
+.filters form {
+    display: flex;
+    background: #efefef;
+    padding: 10px;
+    border-radius: 5px;
+}
+form div{
+    margin-right: 10px;
+}
+form button {
+    align-self: center;
+    margin-top: 26px;
+    margin-left: 20px;
+}
+.main {
+    margin: 10px;
+}
+.legend {
+    display: flex;
+}
+.legend div {
+    display:flex;
+    align-items: center;
+    margin-right: 20px;
+}
+.legend span{
+    display: inline-block;
+    padding: 3px 5px;
+    min-height: 25px;
+    min-width: 25px;
+}
+.colorRed {
+    color:#cf1d1d;
+    font-weight: bold;
+}
+.form-group {
+    margin-bottom: 0;
+}
+#message {
+    font-weight: 500;
+    font-size: 20px;
+    margin: 20px;
+    text-align: center;
+    color: #cf1d1d;
+}
+.table-dark:hover {
+    background: #212529!important;
+}
+h2 {
+    margin-bottom: 2rem;
+}

From 8c5fcdc5c9839838c16a44853414da649fad6729 Mon Sep 17 00:00:00 2001
From: Ilya Churaev <ilya.churaev@intel.com>
Date: Mon, 21 Jun 2021 15:05:16 +0300
Subject: [PATCH 145/178] Fixed incorrect check in IE backend (#6241)

---
 ngraph/test/runtime/ie/ie_tensor.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ngraph/test/runtime/ie/ie_tensor.cpp b/ngraph/test/runtime/ie/ie_tensor.cpp
index fd4e8468f92206..a7be07005e2275 100644
--- a/ngraph/test/runtime/ie/ie_tensor.cpp
+++ b/ngraph/test/runtime/ie/ie_tensor.cpp
@@ -36,7 +36,7 @@ void runtime::ie::IETensor::write(const void* src, size_t bytes)
     {
         m_data = AlignedBuffer(bytes);
     }
-    NGRAPH_CHECK(m_data.size() <= bytes,
+    NGRAPH_CHECK(bytes <= m_data.size(),
                  "Buffer over-write. The buffer size: ",
                  m_data.size(),
                  " is lower than the number of bytes to write: ",

From b2d25da04de5b6396ef98d4c2131adddb918bfcb Mon Sep 17 00:00:00 2001
From: Pavel Esir <pavel.esir@intel.com>
Date: Mon, 21 Jun 2021 15:33:13 +0300
Subject: [PATCH 146/178] [spec] add Gather-8 with negative indices (#6172)

* add Gather-8 with negative indices

* incremented version

* updated opset8.md
---
 docs/doxygen/ie_docs.xml      |   1 +
 docs/ops/movement/Gather_8.md | 200 ++++++++++++++++++++++++++++++++++
 docs/ops/opset8.md            |   2 +-
 3 files changed, 202 insertions(+), 1 deletion(-)
 create mode 100644 docs/ops/movement/Gather_8.md

diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index 8841672b430623..008b0d051646b1 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -161,6 +161,7 @@ limitations under the License.
                 <tab type="user" title="GatherTree-1" url="@ref openvino_docs_ops_movement_GatherTree_1"/>
                 <tab type="user" title="Gather-1" url="@ref openvino_docs_ops_movement_Gather_1"/>
                 <tab type="user" title="Gather-7" url="@ref openvino_docs_ops_movement_Gather_7"/>
+                <tab type="user" title="Gather-8" url="@ref openvino_docs_ops_movement_Gather_8"/>
                 <tab type="user" title="GatherElements-6" url="@ref openvino_docs_ops_movement_GatherElements_6"/>
                 <tab type="user" title="GatherND-5" url="@ref openvino_docs_ops_movement_GatherND_5"/>
                 <tab type="user" title="Gelu-2" url="@ref openvino_docs_ops_activation_GELU_2"/>
diff --git a/docs/ops/movement/Gather_8.md b/docs/ops/movement/Gather_8.md
new file mode 100644
index 00000000000000..bdb687bd467c02
--- /dev/null
+++ b/docs/ops/movement/Gather_8.md
@@ -0,0 +1,200 @@
+## Gather <a name="Gather"></a> {#openvino_docs_ops_movement_Gather_8}
+
+**Versioned name**: *Gather-8*
+
+**Category**: Data movement operations
+
+**Short description**: *Gather* operation takes slices of data of the first input tensor according to the indices
+ specified with the second input tensor and axis from the third input. Semantics of this operation is identical to 
+TensorFlow\* [Gather](https://www.tensorflow.org/api_docs/python/tf/gather) operation.
+
+**Detailed description**
+
+    output[p_0, p_1, ..., p_{axis-1}, i_b, ..., i_{M-1}, p_{axis+1}, ..., p_{N-1}] = 
+       data[p_0, p_1, ..., p_{axis-1}, indices[p_0, p_1, ..., p_{b-1}, i_b, ..., i_{M-1}], p_{axis+1}, ..., p_{N-1}]
+
+Where `data`, `indices` and `axis` are tensors from first, second and third inputs correspondingly, `b` is 
+the number of batch dimensions. `N` and `M` are numbers of dimensions of `data` and `indices` tensors, respectively.
+
+**Attributes**:
+* *batch_dims*
+  * **Description**: *batch_dims* (also denoted as `b`) is a leading number of dimensions of `data` tensor and `indices` 
+  representing the batches, and *Gather* starts to gather from the `b` dimension. It requires the first `b` 
+  dimensions in `data` and `indices` tensors to be equal. If `batch_dims` is less than zero, normalized value is used 
+  `batch_dims = indices.rank + batch_dims`.
+  * **Range of values**: `[-min(data.rank, indices.rank); min(data.rank, indices.rank)]` and `batch_dims' <= axis'`.
+  Where `batch_dims'` and `axis'` stand for normalized `batch_dims` and `axis` values.
+  * **Type**: *T_AXIS*
+  * **Default value**: 0
+  * **Required**: *no*
+
+Example 1 with default *batch_dims* value:
+```
+batch_dims = 0
+axis = 0
+
+indices = [0, 0, 4] 
+data    = [1, 2, 3, 4, 5]
+output  = [1, 1, 5]
+```
+
+Example 2 with non-default *batch_dims* value:
+```
+batch_dims = 1
+axis = 1
+
+indices = [[0, 0, 4], <-- this is applied to the first batch 
+           [4, 0, 0]]  <-- this is applied to the second batch
+indices_shape = (2, 3)
+
+data    = [[1, 2, 3, 4, 5],  <-- the first batch
+           [6, 7, 8, 9, 10]]  <-- the second batch 
+data_shape = (2, 5)
+
+output  = [[ 1, 1, 5],
+           [10, 6, 6]]
+output_shape = (2, 3)
+```
+
+Example 3 with non-default *batch_dims* value:
+```
+batch_dims = 2
+axis = 2
+
+indices = [[[0, 0, 4],  <-- this is applied to the first batch, index = (0, 0)
+            [4, 0, 0]],  <-- this is applied to the second batch, index = (0, 1)
+          
+           [[1, 2, 4],  <-- this is applied to the third batch, index = (1, 0)
+            [4, 3, 2]]]  <-- this is applied to the fourth batch, index = (1, 1) 
+indices_shape = (2, 2, 3)
+
+data    = [[[1, 2, 3, 4, 5],  <-- the first batch, index = (0, 0)
+            [6, 7, 8, 9, 10]],  <-- the second batch, index = (0, 1)
+          
+           [[11, 12, 13, 14, 15],  <-- the third batch, index = (1, 0)
+            [16, 17, 18, 19, 20]]]  <-- the fourth batch, index = (1, 1)
+data_shape = (2, 2, 5)
+
+output  = [[[ 1, 1, 5],
+            [10, 6, 6]],
+
+           [[12, 13, 15],
+            [20, 19, 18]]] 
+output_shape = (2, 2, 3)
+```
+Example 4 with *axis* > *batch_dims*:
+```
+batch_dims = 1
+axis = 2
+
+indices = [[1, 2, 4],  <-- this is applied to the first batch 
+           [4, 3, 2]]  <-- this is applied to the second batch
+indices_shape = (2, 3)
+
+data = [[[[ 1,  2,  3,  4], <-- first batch
+          [ 5,  6,  7,  8],
+          [ 9, 10, 11, 12],
+          [13, 14, 15, 16],
+          [17, 18, 19, 20]]],
+  
+        [[[21, 22, 23, 24], <-- second batch
+          [25, 26, 27, 28],
+          [29, 30, 31, 32],
+          [33, 34, 35, 36],
+          [37, 38, 39, 40]]]]
+data_shape = (2, 1, 5, 4)
+
+output = [[[[ 5,  6,  7,  8],
+            [ 9, 10, 11, 12],
+            [17, 18, 19, 20]]],
+
+          [[[37, 38, 39, 40],
+            [33, 34, 35, 36],
+            [29, 30, 31, 32]]]]
+output_shape = (2, 1, 3, 4)
+```
+
+Example 5 with negative *batch_dims* value:
+```
+batch_dims = -1  <-- normalized value will be indices.rank + batch_dims = 2 - 1 = 1
+axis = 1
+
+indices = [[0, 0, 4], <-- this is applied to the first batch 
+           [4, 0, 0]]  <-- this is applied to the second batch
+indices_shape = (2, 3)
+
+data    = [[1, 2, 3, 4, 5],  <-- the first batch
+           [6, 7, 8, 9, 10]]  <-- the second batch 
+data_shape = (2, 5)
+
+output  = [[ 1, 1, 5],
+           [10, 6, 6]]
+output_shape = (2, 3)
+```
+
+Example 6 with negative indices:
+```
+batch_dims = 0
+axis = 0
+
+indices = [0, -2, -1] 
+data    = [1, 2, 3, 4, 5]
+output  = [1, 4, 5]
+```
+
+**Inputs**
+
+* **1**:  `data` tensor of type *T* with arbitrary data. **Required**.
+
+* **2**:  `indices` tensor of type *T_IND* with indices to gather. 0D tensor (scalar) for indices is also allowed. 
+  The values for indices are in the range `[-data[axis], data[axis] - 1]`.
+  Negative values of indices indicate reverse indexing from `data[axis]`. 
+  **Required**.
+
+* **3**:  Scalar or 1D tensor `axis` of *T_AXIS* type is a dimension index to gather data from. For example, 
+*axis* equal to 1 means that gathering is performed over the first dimension. Negative `axis` means reverse indexing and 
+  will be normalized to value `axis = data.rank + axis`. Allowed values are from `[-len(data.shape), len(data.shape) - 1]` 
+  and `axis' >= batch_dims'`. Where `axis'` and `batch_dims'` stand for normalized `batch_dims` and `axis` values.
+**Required**.
+
+**Outputs**
+
+* **1**: The resulting tensor of type *T* that consists of elements from `data` tensor gathered by `indices`. The shape 
+of the output tensor is `data.shape[:axis] + indices.shape[batch_dims:] + data.shape[axis + 1:]`
+
+**Types**
+
+* *T*: any supported type.
+
+* *T_IND*: any supported integer types.
+
+* *T_AXIS*: any supported integer types.
+
+**Example**
+
+```xml
+<layer ... type="Gather" version="opset8">
+    <data batch_dims="1" />
+    <input>
+        <port id="0">
+            <dim>2</dim>
+            <dim>64</dim>
+            <dim>128</dim>
+        </port>
+        <port id="1">
+            <dim>2</dim>
+            <dim>32</dim>
+            <dim>21</dim>
+        </port>
+        <port id="2"/>   <!--  axis = 1  -->
+    </input>
+    <output>
+        <port id="2">
+            <dim>2</dim>
+            <dim>32</dim>
+            <dim>21</dim>
+            <dim>128</dim>
+        </port>
+    </output>
+</layer>
+```
diff --git a/docs/ops/opset8.md b/docs/ops/opset8.md
index fc68d6f32e07ea..02e97eab4e42f6 100644
--- a/docs/ops/opset8.md
+++ b/docs/ops/opset8.md
@@ -63,7 +63,7 @@ declared in `namespace opset8`.
 * [FakeQuantize](quantization/FakeQuantize_1.md)
 * [Floor](arithmetic/Floor_1.md)
 * [FloorMod](arithmetic/FloorMod_1.md)
-* [Gather](movement/Gather_7.md)
+* [Gather](movement/Gather_8.md)
 * [GatherElements](movement/GatherElements_6.md)
 * [GatherND_5](movement/GatherND_5.md)
 * [GatherTree](movement/GatherTree_1.md)

From c5b940ecdb86d742844c399c4451bbca21c5f14e Mon Sep 17 00:00:00 2001
From: Egor Duplensky <egor.duplenskii@intel.com>
Date: Mon, 21 Jun 2021 16:13:19 +0300
Subject: [PATCH 147/178] Correct output nodes naming in scope of NMS
 transformation (#6119)

* Correct output nodes naming in scope of NMS transformation

Details:
NMS transformation changes name of NMS node output (adds '/convert' suffix)
In case NMS transformation is the last node before Output, it leads to
the situation when InferRequest's network Outputs do not match any
graph's Output nodes.

Tickets:
- 57695

* Add test
---
 .../convert_nms_to_nms_ie_internal.cpp        |  5 +-
 .../nms_transformation_for_last_node.cpp      | 13 +++
 .../nms_transformation_for_last_node.hpp      | 14 +++
 .../nms_transformation_for_last_node.cpp      | 88 +++++++++++++++++++
 4 files changed, 118 insertions(+), 2 deletions(-)
 create mode 100644 inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/nms_transformation_for_last_node.cpp
 create mode 100644 inference-engine/tests/functional/plugin/shared/include/execution_graph_tests/nms_transformation_for_last_node.hpp
 create mode 100644 inference-engine/tests/functional/plugin/shared/src/execution_graph_tests/nms_transformation_for_last_node.cpp

diff --git a/inference-engine/src/transformations/src/transformations/op_conversions/convert_nms_to_nms_ie_internal.cpp b/inference-engine/src/transformations/src/transformations/op_conversions/convert_nms_to_nms_ie_internal.cpp
index 3460985788e796..07b1d26ebe1a02 100644
--- a/inference-engine/src/transformations/src/transformations/op_conversions/convert_nms_to_nms_ie_internal.cpp
+++ b/inference-engine/src/transformations/src/transformations/op_conversions/convert_nms_to_nms_ie_internal.cpp
@@ -14,6 +14,7 @@
 
 #include "ngraph_ops/nms_ie_internal.hpp"
 #include "transformations/op_conversions/convert_nms_to_nms_ie_internal.hpp"
+#include "transformations/utils/utils.hpp"
 
 NGRAPH_RTTI_DEFINITION(ngraph::pass::ConvertNMSToNMSIEInternal, "ConvertNMSToNMSIEInternal", 0);
 
@@ -103,14 +104,14 @@ ngraph::pass::ConvertNMSToNMSIEInternal::ConvertNMSToNMSIEInternal() {
         Output<Node> output_0 = nms_legacy->output(0);
         if (nms_5->output(0).get_element_type() != output_0.get_element_type()) {
             output_0 = std::make_shared<opset1::Convert>(output_0, nms_5->output(0).get_element_type());
-            output_0.get_node_shared_ptr()->set_friendly_name(nms_5->get_friendly_name() + "/convert.0");
+            output_0.get_node_shared_ptr()->set_friendly_name(op::util::create_ie_output_name(nms_5->output(0)));
             new_ops.emplace_back(output_0.get_node_shared_ptr());
         }
 
         Output<Node> output_2 = nms_legacy->output(2);
         if (nms_5->output(2).get_element_type() != output_2.get_element_type()) {
             output_2 = std::make_shared<opset1::Convert>(output_2, nms_5->output(2).get_element_type());
-            output_2.get_node_shared_ptr()->set_friendly_name(nms_5->get_friendly_name() + "/convert.2");
+            output_2.get_node_shared_ptr()->set_friendly_name(op::util::create_ie_output_name(nms_5->output(2)));
             new_ops.emplace_back(output_2.get_node_shared_ptr());
         }
 
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/nms_transformation_for_last_node.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/nms_transformation_for_last_node.cpp
new file mode 100644
index 00000000000000..5d5fbe597f8398
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/nms_transformation_for_last_node.cpp
@@ -0,0 +1,13 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "execution_graph_tests/nms_transformation_for_last_node.hpp"
+#include "common_test_utils/test_constants.hpp"
+
+namespace {
+using namespace ExecutionGraphTests;
+
+INSTANTIATE_TEST_CASE_P(smoke_NmsTransformationLastNode, ExecGraphNmsTransformLastNode, ::testing::Values(CommonTestUtils::DEVICE_CPU),
+                        ExecGraphNmsTransformLastNode::getTestCaseName);
+} // namespace
diff --git a/inference-engine/tests/functional/plugin/shared/include/execution_graph_tests/nms_transformation_for_last_node.hpp b/inference-engine/tests/functional/plugin/shared/include/execution_graph_tests/nms_transformation_for_last_node.hpp
new file mode 100644
index 00000000000000..4f6bee8a16b200
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/shared/include/execution_graph_tests/nms_transformation_for_last_node.hpp
@@ -0,0 +1,14 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+
+namespace ExecutionGraphTests {
+
+class ExecGraphNmsTransformLastNode : public testing::TestWithParam<std::string> {
+public:
+    static std::string getTestCaseName(testing::TestParamInfo<std::string> obj);
+};
+
+}  // namespace ExecutionGraphTests
diff --git a/inference-engine/tests/functional/plugin/shared/src/execution_graph_tests/nms_transformation_for_last_node.cpp b/inference-engine/tests/functional/plugin/shared/src/execution_graph_tests/nms_transformation_for_last_node.cpp
new file mode 100644
index 00000000000000..5b67f9a468004e
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/shared/src/execution_graph_tests/nms_transformation_for_last_node.cpp
@@ -0,0 +1,88 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "execution_graph_tests/nms_transformation_for_last_node.hpp"
+
+#include <gtest/gtest.h>
+
+#include <ngraph/function.hpp>
+#include <ngraph/opsets/opset5.hpp>
+
+#include <inference_engine.hpp>
+
+#include "functional_test_utils/skip_tests_config.hpp"
+#include "common_test_utils/ngraph_test_utils.hpp"
+
+#include <memory>
+#include <string>
+#include <algorithm>
+#include <utility>
+
+namespace ExecutionGraphTests {
+
+std::string ExecGraphNmsTransformLastNode::getTestCaseName(
+    testing::TestParamInfo<std::string> obj) {
+  std::string targetDevice = obj.param;
+  return "Dev=" + targetDevice;
+}
+
+/**
+ * Infer simple graph with just NMS node.
+ * Verify that after NMS transformation network can be inferred
+ * especially, that NMS transformation does not change name
+ * of the output (Result) node
+ */
+TEST_P(ExecGraphNmsTransformLastNode, CheckIfCanBeInfered) {
+  SKIP_IF_CURRENT_TEST_IS_DISABLED()
+
+  using namespace ngraph;
+
+  auto device_name = this->GetParam();
+  ngraph::Shape boxes_shape = {1, 2, 4};
+  ngraph::Shape scores_shape = {1, 1, 2};
+  float in_boxes[8] = {1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0};
+  float in_scores[8] = {1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0};
+
+  auto boxes = std::make_shared<opset5::Parameter>(element::f32, boxes_shape);
+  auto scores = std::make_shared<opset5::Parameter>(element::f32, scores_shape);
+  auto max_output_boxes_per_class = opset5::Constant::create(element::i64, Shape{}, {10});
+  auto iou_threshold = opset5::Constant::create(element::f32, Shape{}, {0.75});
+  auto score_threshold = opset5::Constant::create(element::f32, Shape{}, {0.7});
+  auto nms = std::make_shared<opset5::NonMaxSuppression>(boxes, scores, max_output_boxes_per_class,
+                                                                                  iou_threshold, score_threshold,
+                                                                                  opset5::NonMaxSuppression::BoxEncodingType::CORNER, true, element::i64);
+  ngraph::ResultVector results {
+      std::make_shared<opset5::Result>(nms->output(0)),
+  };
+
+  auto f = std::make_shared<Function>(results, ParameterVector{boxes, scores}, "NMS");
+
+  auto ie = InferenceEngine::Core();
+  auto net = InferenceEngine::CNNNetwork(f);
+  auto exec_net = ie.LoadNetwork(net, device_name);
+  auto infer_req = exec_net.CreateInferRequest();
+
+  InferenceEngine::TensorDesc tDesc1(InferenceEngine::Precision::FP32, boxes_shape,
+                                    InferenceEngine::Layout::CHW);
+  InferenceEngine::TensorDesc tDesc2(InferenceEngine::Precision::FP32, scores_shape,
+                                    InferenceEngine::Layout::CHW);
+
+  InferenceEngine::Blob::Ptr inBlob1 = InferenceEngine::make_shared_blob<float>(tDesc1, in_boxes);
+  infer_req.SetBlob(boxes->get_name(), inBlob1);
+  InferenceEngine::Blob::Ptr inBlob2 = InferenceEngine::make_shared_blob<float>(tDesc2, in_scores);
+  infer_req.SetBlob(scores->get_name(), inBlob2);
+
+  infer_req.Infer();
+
+  const auto& initial_outputs = net.getOutputsInfo();
+  const auto& final_outputs = exec_net.GetOutputsInfo();
+
+  auto compareOutputNames = [] (const std::pair<std::string, InferenceEngine::CDataPtr>& lhs,
+                                const std::pair<std::string, InferenceEngine::CDataPtr>& rhs)
+  { return lhs.first == rhs.first; };
+
+  ASSERT_TRUE(std::equal(initial_outputs.begin(), initial_outputs.end(), final_outputs.begin(), compareOutputNames));
+}
+
+} // namespace ExecutionGraphTests

From 9e7d98fca9e79ad63fa3c4be80e69678960e252e Mon Sep 17 00:00:00 2001
From: Dmitrii Khurtin <dmitrii.khurtin@intel.com>
Date: Mon, 21 Jun 2021 17:53:38 +0300
Subject: [PATCH 148/178] Additional checks are added to prevent errors from
 occurring on the GNA library (#5953)

* additional checks are added to prevent errors from occurring on the GNA library side

* fixing after review

* move AreLayersSupported from Cnn2D namespace

* GNA_DEBUG is commented

* batch size <= 8 test was added
---
 .../gna_plugin/backend/gna_limitations.cpp    |  77 ++++++++++--
 .../gna_plugin/backend/gna_limitations.hpp    |   4 +
 .../src/gna_plugin/gna_plugin.cpp             |   2 +-
 .../src/gna_plugin/layers/gna_layer_info.hpp  |  26 ++++
 .../src/gna_plugin/layers/gna_layer_type.cpp  |  49 --------
 .../src/gna_plugin/layers/gna_layer_type.hpp  |   1 -
 .../gna_plugin/optimizer/gna_pass_manager.cpp |   7 ++
 .../gna/pass_tests/layers_restrictions.cpp    | 116 ++++++++++++++++++
 8 files changed, 224 insertions(+), 58 deletions(-)
 create mode 100644 inference-engine/tests/functional/plugin/gna/pass_tests/layers_restrictions.cpp

diff --git a/inference-engine/src/gna_plugin/backend/gna_limitations.cpp b/inference-engine/src/gna_plugin/backend/gna_limitations.cpp
index 8443856a449946..6afe55bd043d93 100644
--- a/inference-engine/src/gna_plugin/backend/gna_limitations.cpp
+++ b/inference-engine/src/gna_plugin/backend/gna_limitations.cpp
@@ -5,14 +5,15 @@
 #include "gna_limitations.hpp"
 
 #include <cstdint>
+#include <unordered_set>
+#include <legacy/ie_layers.h>
+#include <legacy/graph_tools.hpp>
+#include <layers/gna_layer_type.hpp>
+#include <layers/gna_layer_info.hpp>
 
-using GNAPluginNS::GNALimitations::Cnn2D::Validator;
-using GNAPluginNS::GNALimitations::Cnn2D::VectorOrSquareLimit;
-using GNAPluginNS::GNALimitations::Cnn2D::VectorOrSquareLimitByChannels;
-using GNAPluginNS::GNALimitations::Cnn2D::VectorOrSquareLimitByChannelsAndPrecision;
-using GNAPluginNS::GNALimitations::Cnn2D::RangeLimit;
-using GNAPluginNS::GNALimitations::Cnn2D::RangeLimit2D;
-using GNAPluginNS::GNALimitations::Cnn2D::RangeMultipleLimit;
+namespace GNAPluginNS {
+namespace GNALimitations {
+namespace Cnn2D {
 
 bool RangeLimit::isValid(const uint32_t val) const {
     return val >= min && val <= max;
@@ -127,3 +128,65 @@ void Validator::ThrowIfNotEmpty(const std::string prefix, const std::string erro
         THROW_GNA_EXCEPTION << prefix << error;
     }
 }
+
+} // namespace Cnn2D
+
+bool AreLayersSupported(InferenceEngine::CNNNetwork& network, std::string& errMessage) {
+    IE_SUPPRESS_DEPRECATED_START
+    InferenceEngine::InputsDataMap inputs = network.getInputsInfo();
+    std::unordered_set<InferenceEngine::CNNLayer *> allLayers;
+    InferenceEngine::CNNLayerPtr startLayer;
+    if (inputs.empty()) {
+        auto outputs = network.getOutputsInfo();
+        IE_ASSERT(!outputs.empty());
+        // If there are no inputs start search from an output
+        startLayer = getCreatorLayer(outputs.begin()->second).lock();
+    } else {
+        auto network_input_precision = inputs.begin()->second->getPrecision();
+
+        if (network_input_precision != InferenceEngine::Precision::FP32 &&
+            network_input_precision != InferenceEngine::Precision::I16 &&
+            network_input_precision != InferenceEngine::Precision::U8) {
+            errMessage = "The plugin does not support input precision with " +
+                         std::string(network_input_precision.name()) +
+                         " format. Supported  input precisions FP32, I16, U8\n";
+            return false;
+        }
+
+        auto & secondLayers = getInputTo(inputs.begin()->second->getInputData());
+        if (secondLayers.empty()) {
+            errMessage = "Network consists of input layer only (GNA)\n";
+            return false;
+        }
+        startLayer = secondLayers.begin()->second;
+    }
+    auto batch_size = network.getBatchSize();
+
+    bool check_result = true;
+    InferenceEngine::details::UnorderedDFS(allLayers,
+                                           startLayer,
+                                           [&](const InferenceEngine::CNNLayerPtr layer) {
+                                               if (GNAPluginNS::LayerTypeFromStr(layer->type) == GNAPluginNS::LayerType::NO_TYPE) {
+                                                   errMessage = "The plugin does not support layer: " + layer->name + ":" + layer->type + "\n";
+                                                   check_result =  false;
+                                               }
+                                               if (batch_size != 1 && GNAPluginNS::LayerInfo::isBatchSizeConstrained(layer->type)) {
+                                                   errMessage = "topology with layer: " + layer->name + ", type: " + layer->type +
+                                                                ", and batch size(" + std::to_string(batch_size) + ") != 1 not supported";
+                                                   check_result =  false;
+                                               }
+                                               if (GNAPluginNS::LayerInfo(layer).isFullyConnected()) {
+                                                   size_t output_batch_size = GNAPluginNS::LayerInfo(layer).getOutputBatchSize();
+                                                   if (output_batch_size > 8) {
+                                                        errMessage = "topology with layer: " + layer->name + ", type: " + layer->type +
+                                                                     ", and batch size(" + std::to_string(output_batch_size) + ") not supported";
+                                                        check_result =  false;
+                                                   }
+                                               }
+                                           }, false);
+    IE_SUPPRESS_DEPRECATED_END
+    return check_result;
+}
+
+} // namespace GNALimitations
+} // namespace GNAPluginNS
diff --git a/inference-engine/src/gna_plugin/backend/gna_limitations.hpp b/inference-engine/src/gna_plugin/backend/gna_limitations.hpp
index 3a283cae895d15..59dd0478cfa900 100644
--- a/inference-engine/src/gna_plugin/backend/gna_limitations.hpp
+++ b/inference-engine/src/gna_plugin/backend/gna_limitations.hpp
@@ -6,6 +6,7 @@
 
 #include "dnn_types.h"
 #include <cstdint>
+#include <cpp/ie_cnn_network.h>
 
 namespace GNAPluginNS {
 namespace GNALimitations {
@@ -94,5 +95,8 @@ class Validator {
         const uint32_t strideH, const uint32_t strideW) const;
 };
 } // namespace Cnn2D
+
+bool AreLayersSupported(InferenceEngine::CNNNetwork& network, std::string& errMessage);
+
 } // namespace GNALimitations
 } // namespace GNAPluginNS
diff --git a/inference-engine/src/gna_plugin/gna_plugin.cpp b/inference-engine/src/gna_plugin/gna_plugin.cpp
index cb227304649638..01ba39a638a0b4 100644
--- a/inference-engine/src/gna_plugin/gna_plugin.cpp
+++ b/inference-engine/src/gna_plugin/gna_plugin.cpp
@@ -723,7 +723,7 @@ void GNAPlugin::LoadNetwork(CNNNetwork & _network) {
 
     //  Check the input network
     std::string error;
-    if (!AreLayersSupported(network, error)) {
+    if (!GNAPluginNS::GNALimitations::AreLayersSupported(network, error)) {
         THROW_GNA_EXCEPTION << error.c_str();
     }
 
diff --git a/inference-engine/src/gna_plugin/layers/gna_layer_info.hpp b/inference-engine/src/gna_plugin/layers/gna_layer_info.hpp
index a4681bdff41e01..93fb4417dc7296 100644
--- a/inference-engine/src/gna_plugin/layers/gna_layer_info.hpp
+++ b/inference-engine/src/gna_plugin/layers/gna_layer_info.hpp
@@ -90,6 +90,32 @@ class LayerInfo {
         static InferenceEngine::details::caseless_set<std::string> layersWithConstrains = {"memory", "convolution"};
         return layersWithConstrains.find(name) != layersWithConstrains.end();
     }
+    size_t getOutputBatchSize() const {
+        if (!layer) {
+            THROW_GNA_EXCEPTION << "layer is null";
+        }
+        if (!layer->outData[0]) {
+            THROW_GNA_EXCEPTION << "output data of layer '" << layer->name << "' is null";
+        }
+        auto& dims = layer->outData[0]->getDims();
+        auto layout = layer->outData[0]->getLayout();
+        switch (dims.size()) {
+        case 1:
+            return 1;
+        case 2:
+            if (layout == InferenceEngine::Layout::NC) {
+                return dims[0];
+            } else if (layout == InferenceEngine::Layout::CN) {
+                return dims[1];
+            } else {
+                THROW_GNA_EXCEPTION << "batch size is not define in layer '" << layer->name << "'";
+            }
+        case 4:
+            return dims[0];
+        default:
+            THROW_GNA_EXCEPTION << "batch size is not define in layer '" << layer->name << "'";
+        }
+    }
     bool isActivation() const noexcept {
         IS_VALID();
         static InferenceEngine::details::caseless_set<std::string> activations =
diff --git a/inference-engine/src/gna_plugin/layers/gna_layer_type.cpp b/inference-engine/src/gna_plugin/layers/gna_layer_type.cpp
index a333d47c48aac3..7a3de49a2de052 100644
--- a/inference-engine/src/gna_plugin/layers/gna_layer_type.cpp
+++ b/inference-engine/src/gna_plugin/layers/gna_layer_type.cpp
@@ -15,52 +15,3 @@ GNAPluginNS::LayerType GNAPluginNS::LayerTypeFromStr(const std::string &str) {
     else
         return NO_TYPE;
 }
-
-bool GNAPluginNS::AreLayersSupported(InferenceEngine::CNNNetwork& network, std::string& errMessage) {
-    IE_SUPPRESS_DEPRECATED_START
-    InferenceEngine::InputsDataMap inputs = network.getInputsInfo();
-    std::unordered_set<InferenceEngine::CNNLayer *> allLayers;
-    InferenceEngine::CNNLayerPtr startLayer;
-    if (inputs.empty()) {
-        auto outputs = network.getOutputsInfo();
-        IE_ASSERT(!outputs.empty());
-        // If there are no inputs start search from an output
-        startLayer = getCreatorLayer(outputs.begin()->second).lock();
-    } else {
-        auto network_input_precision = inputs.begin()->second->getPrecision();
-
-        if (network_input_precision != InferenceEngine::Precision::FP32 &&
-            network_input_precision != InferenceEngine::Precision::I16 &&
-            network_input_precision != InferenceEngine::Precision::U8) {
-            errMessage = "The plugin does not support input precision with " +
-                         std::string(network_input_precision.name()) +
-                         " format. Supported  input precisions FP32, I16, U8\n";
-            return false;
-        }
-
-        auto & secondLayers = getInputTo(inputs.begin()->second->getInputData());
-        if (secondLayers.empty()) {
-            errMessage = "Network consists of input layer only (GNA)\n";
-            return false;
-        }
-        startLayer = secondLayers.begin()->second;
-    }
-    auto batch_size = network.getBatchSize();
-
-    bool check_result = true;
-    InferenceEngine::details::UnorderedDFS(allLayers,
-                                           startLayer,
-                                           [&](const InferenceEngine::CNNLayerPtr layer) {
-                                               if (LayerTypeFromStr(layer->type) == LayerType::NO_TYPE) {
-                                                   errMessage = "The plugin does not support layer: " + layer->name + ":" + layer->type + "\n";
-                                                   check_result =  false;
-                                               }
-                                               if (batch_size != 1 && LayerInfo::isBatchSizeConstrained(layer->type)) {
-                                                   errMessage = "topology with layer: " + layer->name + ", type: " + layer->type +
-                                                                ", and batch size(" + std::to_string(batch_size) + ") != 1 not supported";
-                                                   check_result =  false;
-                                               }
-                                           }, false);
-    IE_SUPPRESS_DEPRECATED_END
-    return check_result;
-}
diff --git a/inference-engine/src/gna_plugin/layers/gna_layer_type.hpp b/inference-engine/src/gna_plugin/layers/gna_layer_type.hpp
index 7fdbf7e576d24b..266590dba68e43 100644
--- a/inference-engine/src/gna_plugin/layers/gna_layer_type.hpp
+++ b/inference-engine/src/gna_plugin/layers/gna_layer_type.hpp
@@ -91,5 +91,4 @@ static const InferenceEngine::details::caseless_map<std::string, GNAPluginNS::La
 };
 
 GNAPluginNS::LayerType LayerTypeFromStr(const std::string &str);
-bool AreLayersSupported(InferenceEngine::CNNNetwork& network, std::string& errMessage);
 }  // namespace GNAPluginNS
diff --git a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
index 5355a7b28f1277..ac01a2324c997a 100644
--- a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
+++ b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
@@ -1216,6 +1216,13 @@ void InsertConcatAligningFilterPass::run() {
                 // modifying output rows to be used - to avoid modification to original concat we are store num of elements in params
                 dims[1] = num_rows_out;
 
+                if (concatInput->getLayout() == Layout::NC && dims[0] > 8 ||
+                    concatInput->getLayout() == Layout::CN && dims[1] > 8) {
+                    THROW_GNA_EXCEPTION << "unsupported batch number '" <<
+                        (concatInput->getLayout() == Layout::NC ? dims[0] : dims[1]) <<
+                        "' in layer '" << concatLayer->name << "'";
+                }
+
                 auto outData = std::make_shared<Data>(filterName,
                                                       TensorDesc(concatInput->getPrecision(),
                                                                  dims,
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/layers_restrictions.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/layers_restrictions.cpp
new file mode 100644
index 00000000000000..fc7f15994d2044
--- /dev/null
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/layers_restrictions.cpp
@@ -0,0 +1,116 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+//
+#include <vector>
+#include <tuple>
+#include <string>
+
+#include <ie_core.hpp>
+
+#include "common_test_utils/common_utils.hpp"
+#include "functional_test_utils/plugin_cache.hpp"
+#include "shared_test_classes/base/layer_test_utils.hpp"
+#include "functional_test_utils/blob_utils.hpp"
+#include "ngraph_functions/utils/ngraph_helpers.hpp"
+#include "ngraph_functions/builders.hpp"
+#include "ngraph_functions/pass/convert_prc.hpp"
+
+using LayersRestrictionsParamsTuple = typename std::tuple<
+        InferenceEngine::Precision,         // Network Precision
+        std::map<std::string, std::string>, // Configuration
+        std::string>;                       // Device name
+
+namespace LayerTestsDefinitions {
+
+struct FullyConnectedBatchSizeMoreThan8 {
+    static const char* getName() { return "FullyConnectedBatchSizeMoreThan8"; }
+    static std::shared_ptr<ngraph::Function> createTopology(const InferenceEngine::Precision& netPrecision) {
+        auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
+        std::vector<size_t> inputShape = {9, 1};
+        auto params = ngraph::builder::makeParams(ngPrc, {inputShape});
+        auto weights = CommonTestUtils::generate_float_numbers(inputShape[1] * inputShape[1], -0.0001f, 0.0001f);
+        auto fullyConnected = ngraph::builder::makeFullyConnected(params[0], ngPrc, inputShape[1], false, {}, weights);
+        ngraph::ResultVector results{ std::make_shared<ngraph::opset1::Result>(fullyConnected) };
+        return std::make_shared<ngraph::Function>(results, params, getName());
+    }
+    static const char* getMatch() { return "and batch size(9) not supported"; }
+};
+
+struct FullyConnectedBatchSizeLessThanOrEqual8 {
+    static const char* getName() { return "FullyConnectedBatchSizeLessThanOrEqual8"; }
+    static std::shared_ptr<ngraph::Function> createTopology(const InferenceEngine::Precision& netPrecision) {
+        auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
+        std::vector<size_t> inputShape = {7, 1};
+        auto params = ngraph::builder::makeParams(ngPrc, {inputShape});
+        auto weights = CommonTestUtils::generate_float_numbers(inputShape[1] * inputShape[1], -0.0001f, 0.0001f);
+        auto fullyConnected = ngraph::builder::makeFullyConnected(params[0], ngPrc, inputShape[1], false, {}, weights);
+        ngraph::ResultVector results{ std::make_shared<ngraph::opset1::Result>(fullyConnected) };
+        return std::make_shared<ngraph::Function>(results, params, getName());
+    }
+};
+
+template<typename T>
+class LayersRestrictions : public testing::WithParamInterface<LayersRestrictionsParamsTuple>,
+    public LayerTestsUtils::LayerTestsCommon {
+public:
+    static std::string getTestCaseName(testing::TestParamInfo<LayersRestrictionsParamsTuple> obj) {
+        InferenceEngine::Precision netPrecision;
+        std::map<std::string, std::string> configuration;
+        std::string targetDevice;
+        std::tie(netPrecision, configuration, targetDevice) = obj.param;
+        std::ostringstream result;
+        result << T::getName() << "_";
+        result << "netPRC=" << netPrecision.name() << "_";
+        result << "targetDevice=" << targetDevice << "_";
+        for (auto const& configItem : configuration) {
+            result << "configItem=" << configItem.first << "_" << configItem.second << "_";
+        }
+        return result.str();
+    }
+    static const char* getMatch() { return T::getMatch(); }
+
+protected:
+    void SetUp() override {
+        InferenceEngine::Precision netPrecision;
+        std::tie(netPrecision, configuration, targetDevice) = this->GetParam();
+        function = T::createTopology(netPrecision);
+    }
+};
+
+using LayersRestrictionsFullyConnectedBatchSizeMoreThan8 = LayersRestrictions<FullyConnectedBatchSizeMoreThan8>;
+using LayersRestrictionsFullyConnectedBatchSizeLessThanOrEqual8 = LayersRestrictions<FullyConnectedBatchSizeLessThanOrEqual8>;
+
+TEST_P(LayersRestrictionsFullyConnectedBatchSizeMoreThan8, CompareWithRefImpl) {
+    std::string what;
+    try {
+        LoadNetwork();
+    } catch (const std::exception& e) {
+        what.assign(e.what());
+    }
+    EXPECT_TRUE(what.find(getMatch()) != std::string::npos);
+}
+
+TEST_P(LayersRestrictionsFullyConnectedBatchSizeLessThanOrEqual8, CompareWithRefImpl) {
+    Run();
+};
+
+const std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32};
+const std::vector<std::map<std::string, std::string>> configs = {
+    { {"GNA_DEVICE_MODE", "GNA_SW_EXACT"} }
+};
+
+INSTANTIATE_TEST_CASE_P(smoke_layers_restrictions, LayersRestrictionsFullyConnectedBatchSizeMoreThan8,
+                        ::testing::Combine(
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::ValuesIn(configs),
+                                ::testing::Values(CommonTestUtils::DEVICE_GNA)),
+                        LayersRestrictionsFullyConnectedBatchSizeMoreThan8::getTestCaseName);
+
+INSTANTIATE_TEST_CASE_P(smoke_layers_restrictions, LayersRestrictionsFullyConnectedBatchSizeLessThanOrEqual8,
+                        ::testing::Combine(
+                                ::testing::ValuesIn(netPrecisions),
+                                ::testing::ValuesIn(configs),
+                                ::testing::Values(CommonTestUtils::DEVICE_GNA)),
+                        LayersRestrictionsFullyConnectedBatchSizeLessThanOrEqual8::getTestCaseName);
+} // namespace LayerTestsDefinitions

From 6799a319112fce5bf5edfa1d36a395c22f143669 Mon Sep 17 00:00:00 2001
From: Vladislav Golubev <vladislav.golubev@intel.com>
Date: Mon, 21 Jun 2021 18:02:23 +0300
Subject: [PATCH 149/178] LPT fixes (#6214)

* [LPT] StridedSlice fix

* [LPT] separateInStandaloneBranch fix
---
 .../src/network_helper.cpp                    |  6 +-
 .../src/strided_slice.cpp                     |  9 +-
 ...te_in_standalone_branch_transformation.cpp |  6 +-
 .../strided_slice_transformation.cpp          | 93 ++++++++++++++++++-
 4 files changed, 106 insertions(+), 8 deletions(-)

diff --git a/inference-engine/src/low_precision_transformations/src/network_helper.cpp b/inference-engine/src/low_precision_transformations/src/network_helper.cpp
index 90995bc0ffb22d..346034e02abd89 100644
--- a/inference-engine/src/low_precision_transformations/src/network_helper.cpp
+++ b/inference-engine/src/low_precision_transformations/src/network_helper.cpp
@@ -549,7 +549,11 @@ std::shared_ptr<ngraph::Node> NetworkHelper::separateInStandaloneBranch(std::sha
         }
 
         std::vector<Output<Node>> inputs = node->input_values();
-        const size_t inputIndex = NetworkHelper::getChildInputIndex(dequantization.multiply, node);
+        const auto originalParent = dequantization.multiply ?
+            dequantization.multiply->shared_from_this() :
+            dequantization.subtract->shared_from_this();
+
+        const size_t inputIndex = NetworkHelper::getChildInputIndex(originalParent, node);
         inputs[inputIndex] = parent;
         const std::shared_ptr<Node> newNode = node->clone_with_new_inputs(inputs);
 
diff --git a/inference-engine/src/low_precision_transformations/src/strided_slice.cpp b/inference-engine/src/low_precision_transformations/src/strided_slice.cpp
index a269e392302ce4..e3d420c849068f 100644
--- a/inference-engine/src/low_precision_transformations/src/strided_slice.cpp
+++ b/inference-engine/src/low_precision_transformations/src/strided_slice.cpp
@@ -17,13 +17,12 @@ std::shared_ptr<Node> stridedSliceDeqConstant(
     const std::shared_ptr<ngraph::Node> strSlice,
     const std::shared_ptr<ngraph::Node> dequantizaitonConstant) {
     auto constant = as_type_ptr<ngraph::opset1::Constant>(dequantizaitonConstant);
-    // issue #48857: constant is mistakenly recognized as a scalar. Uncomment after fix
-    //if (NetworkHelper::isScalarLike(constant)) {
-    //    return NetworkHelper::toScalar(constant);
-    //}
+    auto constantShape = constant->get_shape();
+    if (ngraph::shape_size(constantShape) == 1ul) {
+        return NetworkHelper::toScalar(constant);
+    }
 
     const auto stridedSliceShape = strSlice->get_input_shape(0);
-    auto constantShape = constant->get_shape();
     if (stridedSliceShape.size() != constantShape.size()) {
         ngraph::Shape newConstantShape;
         if (ngraph::shape_size(constantShape) == 1) {
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/separate_in_standalone_branch_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/separate_in_standalone_branch_transformation.cpp
index ec6176375b0493..04832b837fe4bc 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/separate_in_standalone_branch_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/separate_in_standalone_branch_transformation.cpp
@@ -81,7 +81,6 @@ class SeparateInStandaloneBranchTransformation :
                 "SeparateInStandaloneBranchTransformation");
         };
         actualFunction = createActualFunction(testValues.precisionBefore, shape, testValues.dequantization);
-
         const auto result = actualFunction->get_results()[0];
         ngraph::pass::low_precision::NetworkHelper::separateInStandaloneBranch(result->get_input_node_shared_ptr(0));
 
@@ -143,6 +142,11 @@ std::vector<SeparateInStandaloneBranchTransformationTestValues> testValues = {
         ngraph::element::u8,
         { ngraph::element::f32, { 127.f }, { 0.02f } }
     },
+    {
+        LayerTransformation::createParamsU8U8(),
+        ngraph::element::u8,
+        { ngraph::element::f32, { 127.f }, {} }
+    },
     {
         LayerTransformation::createParamsU8U8().setSupportAsymmetricQuantization(true),
         ngraph::element::u8,
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/strided_slice_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/strided_slice_transformation.cpp
index 8b16ce99d75eda..ca8f4a68518ace 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/strided_slice_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/strided_slice_transformation.cpp
@@ -117,7 +117,8 @@ class StridedSliceTransformation : public LayerTransformation, public testing::W
             testValues.inputShape << testValues.actual.inputPrecision << "_" << toString(testValues.params) <<
             testValues.actual.dequantization << "_strided_slice_params_" << testValues.layerParams.begin <<
             testValues.layerParams.end << testValues.layerParams.beginMask <<
-            testValues.layerParams.endMask << testValues.layerParams.strides;
+            testValues.layerParams.endMask << testValues.layerParams.strides <<
+            testValues.layerParams.shrinkAxisMask << testValues.layerParams.newAxisMask;
         return result.str();
     }
 };
@@ -161,6 +162,28 @@ StridedSliceTransformationTestValues::LayerParams specialDimensionEndSlice = {
     {}
 };
 
+StridedSliceTransformationTestValues::LayerParams sliceWithRemovedAxis = {
+    { 0, 1, 0, 0 }, // begin
+    { 1, 2, 1, 1 }, // end
+    { 1, 1, 1, 1 }, // strided
+    { 1, 0, 1, 1 }, // beginMask
+    { 1, 0, 1, 1 }, // endMask
+    { 0, 0, 0, 0 }, // newAxisMask
+    { 0, 1, 0, 0 }, // shrinkAxisMask
+    { 0, 0, 0, 0 } // elipsisMask
+};
+
+StridedSliceTransformationTestValues::LayerParams sliceWithAdditionalAxis = {
+    { 0, 1, 0, 0 }, // begin
+    { 1, 2, 1, 1 }, // end
+    { 1, 1, 1, 1 }, // strided
+    { 1, 0, 1, 1 }, // beginMask
+    { 1, 0, 1, 1 }, // endMask
+    { 0, 1, 0, 0 }, // newAxisMask
+    { 0, 0, 0, 0 }, // shrinkAxisMask
+    { 0, 0, 0, 0 } // elipsisMask
+};
+
 const std::vector<StridedSliceTransformationTestValues> stridedSliceTransformationTestValues = {
     // U8: channel slice, per-tensor quantization
     {
@@ -442,6 +465,74 @@ const std::vector<StridedSliceTransformationTestValues> stridedSliceTransformati
             {{ngraph::element::f32}, {}, { {0.1f, 0.01f}, ngraph::element::f32, {1, 2, 1, 1} }}
         }
     },
+    // U8: channel slice, per-tensor quantization
+    {
+        ngraph::Shape{1, 3, 16, 1200},
+        LayerTransformation::createParamsU8I8(),
+        sliceWithRemovedAxis,
+        {
+            ngraph::element::u8,
+            {{ngraph::element::f32}, {128.f}, {0.1f}}
+        },
+        {
+            ngraph::element::u8,
+            {},
+            ngraph::element::u8,
+            {{ngraph::element::f32}, {128.f}, {0.1f}}
+        }
+    },
+    // U8: channel slice, per-channel quantization
+    {
+        ngraph::Shape{1, 3, 16, 1200},
+        LayerTransformation::createParamsU8I8(),
+        sliceWithRemovedAxis,
+        {
+            ngraph::element::u8,
+            {{ngraph::element::f32}, { {128.f, 64.f, 32.f} }, { {0.1f, 0.2f, 0.3f} }}
+        },
+        {
+            ngraph::element::u8,
+            {},
+            ngraph::element::u8,
+            {{ngraph::element::f32}, {64.f}, {0.2f}},
+        }
+    },
+    // U8: channel slice, per-tensor quantization
+    {
+        ngraph::Shape{1, 3, 16, 1200},
+        LayerTransformation::createParamsU8I8(),
+        sliceWithAdditionalAxis,
+        {
+            ngraph::element::u8,
+            {{ngraph::element::f32}, {128.f}, {0.1f}}
+        },
+        {
+            ngraph::element::u8,
+            {},
+            ngraph::element::u8,
+            {{ngraph::element::f32}, {128.f}, {0.1f}}
+        }
+    },
+    // U8: channel slice, per-channel quantization
+    {
+        ngraph::Shape{1, 3, 16, 1200},
+        LayerTransformation::createParamsU8I8(),
+        sliceWithAdditionalAxis,
+        {
+            ngraph::element::u8,
+            {{ngraph::element::f32}, { {128.f, 64.f, 32.f} }, { {0.1f, 0.2f, 0.3f} }}
+        },
+        {
+            ngraph::element::u8,
+            {},
+            ngraph::element::u8,
+            {
+                {ngraph::element::f32},
+                { {128.f, 64.f, 32.f}, ngraph::element::f32, {1, 1, 3, 1, 1} },
+                { {0.1f, 0.2f, 0.3f}, ngraph::element::f32, {1, 1, 3, 1, 1} }
+            },
+        }
+    },
 };
 
 INSTANTIATE_TEST_CASE_P(

From 3b2da7de3084374122e57f1c1636f1b22ad469c1 Mon Sep 17 00:00:00 2001
From: Pavel Esir <pavel.esir@intel.com>
Date: Mon, 21 Jun 2021 19:14:37 +0300
Subject: [PATCH 150/178] [docs] corrected DeepSpeech conversion (#6011)

* corrected output names in DeepSpeech conversion doc

* mo args correction

* changed instruction for DeepSpeech version 0.8.2

* added venv activate; removed redundant ending

* added picture and squashed MO graph input args into one

* Apply suggestions from code review

Co-authored-by: Tatiana Savina <tatiana.savina@intel.com>

* applied review comments

Co-authored-by: Tatiana Savina <tatiana.savina@intel.com>
---
 docs/MO_DG/img/DeepSpeech-0.8.2.png           |  3 +
 docs/MO_DG/img/DeepSpeech.png                 |  3 -
 .../Convert_DeepSpeech_From_Tensorflow.md     | 89 +++++++++++--------
 3 files changed, 55 insertions(+), 40 deletions(-)
 create mode 100644 docs/MO_DG/img/DeepSpeech-0.8.2.png
 delete mode 100644 docs/MO_DG/img/DeepSpeech.png

diff --git a/docs/MO_DG/img/DeepSpeech-0.8.2.png b/docs/MO_DG/img/DeepSpeech-0.8.2.png
new file mode 100644
index 00000000000000..ddab04ac34ac29
--- /dev/null
+++ b/docs/MO_DG/img/DeepSpeech-0.8.2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdff3768930f683b81ca466be4f947af3172933a702cd38201a254df27a68556
+size 62498
diff --git a/docs/MO_DG/img/DeepSpeech.png b/docs/MO_DG/img/DeepSpeech.png
deleted file mode 100644
index b6f1ca96486850..00000000000000
--- a/docs/MO_DG/img/DeepSpeech.png
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:7ed2c9052f631055090ef3744117ca5a8e8314e0717ba0fdc984e295caa5b925
-size 112455
diff --git a/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_DeepSpeech_From_Tensorflow.md b/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_DeepSpeech_From_Tensorflow.md
index 74833cf3ad3332..29df0e4695d330 100644
--- a/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_DeepSpeech_From_Tensorflow.md
+++ b/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_DeepSpeech_From_Tensorflow.md
@@ -2,66 +2,81 @@
 
 [DeepSpeech project](https://github.com/mozilla/DeepSpeech) provides an engine to train speech-to-text models.
 
-## Download the Pre-Trained DeepSpeech Model
+## Download the Pretrained DeepSpeech Model
 
-[Pre-trained English speech-to-text model](https://github.com/mozilla/DeepSpeech#getting-the-pre-trained-model)
-is publicly available. To download the model, please follow the instruction below:
+Create a directory where model and metagraph with pretrained weights will be stored:
+```
+mkdir deepspeech
+cd deepspeech
+```
+[Pretrained English speech-to-text model](https://github.com/mozilla/DeepSpeech/releases/tag/v0.8.2) is publicly available. 
+To download the model, follow the instruction below:
 
 * For UNIX*-like systems, run the following command:
 ```
-wget -O - https://github.com/mozilla/DeepSpeech/releases/download/v0.3.0/deepspeech-0.3.0-models.tar.gz | tar xvfz -
+wget -O - https://github.com/mozilla/DeepSpeech/archive/v0.8.2.tar.gz | tar xvfz -
+wget -O - https://github.com/mozilla/DeepSpeech/releases/download/v0.8.2/deepspeech-0.8.2-checkpoint.tar.gz | tar xvfz -
 ```
 * For Windows* systems:
-  1. Download the archive from the DeepSpeech project repository: [https://github.com/mozilla/DeepSpeech/releases/download/v0.3.0/deepspeech-0.3.0-models.tar.gz](https://github.com/mozilla/DeepSpeech/releases/download/v0.3.0/deepspeech-0.3.0-models.tar.gz).
-  2. Unpack it with a file archiver application.
+  1. Download the archive with the model: [https://github.com/mozilla/DeepSpeech/archive/v0.8.2.tar.gz](https://github.com/mozilla/DeepSpeech/archive/v0.8.2.tar.gz).
+  2. Download the TensorFlow\* MetaGraph with pretrained weights: [https://github.com/mozilla/DeepSpeech/releases/download/v0.8.2/deepspeech-0.8.2-checkpoint.tar.gz](https://github.com/mozilla/DeepSpeech/releases/download/v0.8.2/deepspeech-0.8.2-checkpoint.tar.gz).
+  3. Unpack it with a file archiver application.
+
+## Freeze the Model into a *.pb File
 
-After you unpack the archive with the pre-trained model, you will have the new `models` directory with the
-following files:
+After unpacking the archives above, you have to freeze the model. Note that this requires 
+TensorFlow* version 1 which is not available under Python 3.8, so you need Python 3.7 or lower.
+Before freezing, deploy a virtual environment and install the required packages:
 ```
-alphabet.txt  
-lm.binary
-output_graph.pb  
-output_graph.pbmm  
-output_graph.rounded.pb  
-output_graph.rounded.pbmm  
-trie
+virtualenv --python=python3.7 venv-deep-speech
+source venv-deep-speech/bin/activate
+cd DeepSpeech-0.8.2
+pip3 install -e .
 ```
+Freeze the model with the following command:
+```
+python3 DeepSpeech.py --checkpoint_dir ../deepspeech-0.8.2-checkpoint --export_dir ../
+```
+After that, you will get the pretrained frozen model file `output_graph.pb` in the directory `deepspeech` created at 
+the beginning. The model contains the preprocessing and main parts. The first preprocessing part performs conversion of input 
+spectrogram into a form useful for speech recognition (mel). This part of the model is not convertible into 
+IR because it contains unsupported operations `AudioSpectrogram` and `Mfcc`.
 
-Pre-trained frozen model file is `output_graph.pb`.
-
-![DeepSpeech model view](../../../img/DeepSpeech.png)
+The main and most computationally expensive part of the model converts the preprocessed audio into text. 
+There are two specificities with the supported part of the model. 
 
-As you can see, the frozen model still has two variables: `previous_state_c` and
-`previous_state_h`. It means that the model keeps training those variables at each inference.
+The first is that the model contains an input with sequence length. So the model can be converted with 
+a fixed input length shape, thus the model is not reshapeable. 
+Refer to the [Using Shape Inference](../../../../IE_DG/ShapeInference.md).
 
-At the first inference of this graph, the variables are initialized by zero tensors. After executing the `lstm_fused_cell` nodes, cell state and hidden state, which are the results of the `BlockLSTM` execution, are assigned to these two variables.
+The second is that the frozen model still has two variables: `previous_state_c` and `previous_state_h`, figure 
+with the frozen *.pb model is below. It means that the model keeps training these variables at each inference. 
 
-With each inference of the DeepSpeech graph, initial cell state and hidden state data for `BlockLSTM` is taken from previous inference from variables. Outputs (cell state and hidden state) of `BlockLSTM` are reassigned to the same variables.
+![DeepSpeech model view](../../../img/DeepSpeech-0.8.2.png)
 
-It helps the model to remember the context of the words that it takes as input.
+At the first inference the variables are initialized with zero tensors. After executing, the results of the `BlockLSTM` 
+are assigned to cell state and hidden state, which are these two variables.
 
-## Convert the TensorFlow* DeepSpeech Model to IR
+## Convert the Main Part of DeepSpeech Model into IR
 
-The Model Optimizer assumes that the output model is for inference only. That is why you should cut those variables off and resolve keeping cell and hidden states on the application level.
+Model Optimizer assumes that the output model is for inference only. That is why you should cut `previous_state_c` 
+and `previous_state_h` variables off and resolve keeping cell and hidden states on the application level.
 
 There are certain limitations for the model conversion:
 - Time length (`time_len`) and sequence length (`seq_len`) are equal.
 - Original model cannot be reshaped, so you should keep original shapes.
 
-To generate the DeepSpeech Intermediate Representation (IR), provide the TensorFlow DeepSpeech model to the Model Optimizer with the following parameters:
+To generate the IR, run the Model Optimizer with the following parameters:
 ```sh
-python3 ./mo_tf.py                                                      \
---input_model path_to_model/output_graph.pb                             \
---freeze_placeholder_with_value input_lengths->[16]                     \
---input input_node,previous_state_h/read,previous_state_c/read          \
---input_shape [1,16,19,26],[1,2048],[1,2048]                            \
---output raw_logits,lstm_fused_cell/GatherNd,lstm_fused_cell/GatherNd_1 \
+python3 {path_to_mo}/mo_tf.py                            \
+--input_model output_graph.pb                            \
+--input "input_lengths->[16],input_node[1 16 19 26],previous_state_h[1 2048],previous_state_c[1 2048]"   \
+--output "cudnn_lstm/rnn/multi_rnn_cell/cell_0/cudnn_compatible_lstm_cell/GatherNd_1,cudnn_lstm/rnn/multi_rnn_cell/cell_0/cudnn_compatible_lstm_cell/GatherNd,logits" \
 --disable_nhwc_to_nchw
 ```
 
 Where:
-* `--freeze_placeholder_with_value input_lengths->[16]` freezes sequence length
-* `--input input_node,previous_state_h/read,previous_state_c/read` and
-`--input_shape [1,16,19,26],[1,2048],[1,2048]` replace the variables with a placeholder
-* `--output raw_logits,lstm_fused_cell/GatherNd,lstm_fused_cell/GatherNd_1` gets data for the next model
-execution.
+* `input_lengths->[16]` Replaces the input node with name "input_lengths" with a constant tensor of shape [1] with a 
+  single integer value 16. This means that the model now can consume input sequences of length 16 only.
+* `input_node[1 16 19 26],previous_state_h[1 2048],previous_state_c[1 2048]` replaces the variables with a placeholder.
+* `--output ".../GatherNd_1,.../GatherNd,logits" ` output node names.

From 4dc9c03511692fd8d71f660f495cf4953236b348 Mon Sep 17 00:00:00 2001
From: Ilya Naumov <ilya.naumov@intel.com>
Date: Mon, 21 Jun 2021 21:08:21 +0300
Subject: [PATCH 151/178] Fix demo scripts (#6280)

* Fix unknown command line flag errors if -sample-options is used

* Fix raw escape sequences output with dashes

* Fix raw escape sequences output with run_again message

* try change sampleoptions to array
---
 scripts/demo/demo_benchmark_app.sh            | 28 +++++++++----------
 scripts/demo/demo_security_barrier_camera.sh  | 20 ++++++-------
 .../demo_squeezenet_download_convert_run.sh   | 24 ++++++++--------
 3 files changed, 36 insertions(+), 36 deletions(-)

diff --git a/scripts/demo/demo_benchmark_app.sh b/scripts/demo/demo_benchmark_app.sh
index 25d1f94ae8eca7..a287c406d2b2e7 100755
--- a/scripts/demo/demo_benchmark_app.sh
+++ b/scripts/demo/demo_benchmark_app.sh
@@ -33,8 +33,8 @@ case $key in
     shift
     ;;
     -sample-options)
-    sampleoptions="$2 $3 $4 $5 $6"
-    echo sample-options = "${sampleoptions}"
+    sampleoptions=("${@:2}")
+    echo sample-options = "${sampleoptions[*]}"
     shift
     ;;
     *)
@@ -44,8 +44,8 @@ esac
 shift
 done
 
-if ([ -z "$sampleoptions" ]); then
-    sampleoptions="-niter 1000"
+if [ -z "${sampleoptions[*]}" ]; then
+    sampleoptions=( -niter 1000 )
 fi
 
 target_precision="FP16"
@@ -71,12 +71,12 @@ else
 fi
 
 if ! . "$setupvars_path" ; then
-    printf "Unable to run ./setupvars.sh. Please check its presence. %s" "${run_again}"
+    echo -ne "Unable to run ./setupvars.sh. Please check its presence. ${run_again}"
     exit 1
 fi
 
 # Step 1. Download the Caffe model and the prototxt of the model
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "\n\nDownloading the Caffe model and the prototxt"
 
 cur_path=$PWD
@@ -143,7 +143,7 @@ elif [[ "$OSTYPE" == "darwin"* ]]; then
 fi
 
 if ! command -v $python_binary &>/dev/null; then
-    printf "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. %s" "${run_again}"
+    echo -ne "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. ${run_again}"
     exit 1
 fi
 
@@ -166,14 +166,14 @@ ir_dir="${irs_path}/${model_dir}/${target_precision}"
 
 if [ ! -e "$ir_dir" ]; then
     # Step 2. Configure Model Optimizer
-    printf "%s" "${dashes}"
+    echo -ne "${dashes}"
     printf "Install Model Optimizer dependencies\n\n"
     cd "${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/install_prerequisites"
     . ./install_prerequisites.sh caffe
     cd "$cur_path"
 
     # Step 3. Convert a model with Model Optimizer
-    printf "%s" "${dashes}"
+    echo -ne "${dashes}"
     printf "Convert a model with Model Optimizer\n\n"
 
     mo_path="${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/mo.py"
@@ -182,11 +182,11 @@ if [ ! -e "$ir_dir" ]; then
     print_and_run "$python_binary" "$downloader_dir/converter.py" --mo "$mo_path" --name "$model_name" -d "$models_path" -o "$irs_path" --precisions "$target_precision"
 else
     printf "\n\nTarget folder %s already exists. Skipping IR generation  with Model Optimizer." "${ir_dir}"
-    printf "If you want to convert a model again, remove the entire %s folder. %s" "${ir_dir}" "${run_again}"
+    echo -ne "If you want to convert a model again, remove the entire ${ir_dir} folder. ${run_again}"
 fi
 
 # Step 4. Build samples
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "Build Inference Engine samples\n\n"
 
 OS_PATH=$(uname -m)
@@ -211,15 +211,15 @@ cmake -DCMAKE_BUILD_TYPE=Release "$samples_path"
 make $NUM_THREADS benchmark_app
 
 # Step 5. Run samples
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "Run Inference Engine benchmark app\n\n"
 
 cd "$binaries_dir"
 
 cp -f "$ROOT_DIR/${model_name}.labels" "${ir_dir}/"
 
-print_and_run ./benchmark_app -d "$target" -i "$target_image_path" -m "${ir_dir}/${model_name}.xml" -pc "${sampleoptions}"
+print_and_run ./benchmark_app -d "$target" -i "$target_image_path" -m "${ir_dir}/${model_name}.xml" -pc "${sampleoptions[@]}"
 
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 
 printf "Inference Engine benchmark app completed successfully.\n\n"
diff --git a/scripts/demo/demo_security_barrier_camera.sh b/scripts/demo/demo_security_barrier_camera.sh
index eaf6cd6784e094..36e09cab7ba9fe 100755
--- a/scripts/demo/demo_security_barrier_camera.sh
+++ b/scripts/demo/demo_security_barrier_camera.sh
@@ -33,8 +33,8 @@ case $key in
     shift
     ;;
     -sample-options)
-    sampleoptions="$2 $3 $4 $5 $6"
-    echo sample-options = "${sampleoptions}"
+    sampleoptions=("${@:2}")
+    echo sample-options = "${sampleoptions[*]}"
     shift
     ;;
     *)
@@ -112,7 +112,7 @@ elif [[ "$OSTYPE" == "darwin"* ]]; then
 fi
 
 if ! command -v $python_binary &>/dev/null; then
-    printf "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. %s" "${run_again}"
+    echo -ne "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. ${run_again}"
     exit 1
 fi
 
@@ -128,12 +128,12 @@ else
     printf "Error: setupvars.sh is not found\n"
 fi
 if ! . "$setupvars_path" ; then
-    printf "Unable to run ./setupvars.sh. Please check its presence. %s" "${run_again}"
+    echo -ne "Unable to run ./setupvars.sh. Please check its presence. ${run_again}"
     exit 1
 fi
 
 # Step 1. Downloading Intel models
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "Downloading Intel models\n\n"
 
 
@@ -161,13 +161,13 @@ while read -r model_opt model_name; do
 done < "$ROOT_DIR/demo_security_barrier_camera.conf"
 
 # Step 2. Build samples
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "Build Inference Engine demos\n\n"
 
 demos_path="${INTEL_OPENVINO_DIR}/deployment_tools/open_model_zoo/demos"
 
 if ! command -v cmake &>/dev/null; then
-    printf "\n\nCMAKE is not installed. It is required to build Inference Engine demos. Please install it. %s" "${run_again}"
+    echo -ne "\n\nCMAKE is not installed. It is required to build Inference Engine demos. Please install it. ${run_again}"
     exit 1
 fi
 
@@ -189,13 +189,13 @@ cmake -DCMAKE_BUILD_TYPE=Release "$demos_path"
 make $NUM_THREADS security_barrier_camera_demo
 
 # Step 3. Run samples
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "Run Inference Engine security_barrier_camera demo\n\n"
 
 binaries_dir="${build_dir}/${OS_PATH}/Release"
 cd "$binaries_dir"
 
-print_and_run ./security_barrier_camera_demo -d "$target" -d_va "$target" -d_lpr "$target" -i "$target_image_path" "${model_args[@]}" "${sampleoptions}"
+print_and_run ./security_barrier_camera_demo -d "$target" -d_va "$target" -d_lpr "$target" -i "$target_image_path" "${model_args[@]}" "${sampleoptions[@]}"
 
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "Demo completed successfully.\n\n"
diff --git a/scripts/demo/demo_squeezenet_download_convert_run.sh b/scripts/demo/demo_squeezenet_download_convert_run.sh
index 91a9055375b7ef..24fc26c335c9bd 100755
--- a/scripts/demo/demo_squeezenet_download_convert_run.sh
+++ b/scripts/demo/demo_squeezenet_download_convert_run.sh
@@ -33,8 +33,8 @@ case $key in
     shift
     ;;
     -sample-options)
-    sampleoptions="$2 $3 $4 $5 $6"
-    echo sample-options = "${sampleoptions}"
+    sampleoptions=("${@:2}")
+    echo sample-options = "${sampleoptions[*]}"
     shift
     ;;
     *)
@@ -67,12 +67,12 @@ else
 fi
 
 if ! . "$setupvars_path" ; then
-    printf "Unable to run ./setupvars.sh. Please check its presence. %s" "${run_again}"
+    echo -ne "Unable to run ./setupvars.sh. Please check its presence. ${run_again}"
     exit 1
 fi
 
 # Step 1. Download the Caffe model and the prototxt of the model
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "\n\nDownloading the Caffe model and the prototxt"
 
 cur_path=$PWD
@@ -139,7 +139,7 @@ elif [[ "$OSTYPE" == "darwin"* ]]; then
 fi
 
 if ! command -v $python_binary &>/dev/null; then
-    printf "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. %s" "${run_again}"
+    echo -ne "\n\nPython 3.5 (x64) or higher is not installed. It is required to run Model Optimizer, please install it. ${run_again}"
     exit 1
 fi
 
@@ -162,14 +162,14 @@ ir_dir="${irs_path}/${model_dir}/${target_precision}"
 
 if [ ! -e "$ir_dir" ]; then
     # Step 2. Configure Model Optimizer
-    printf "%s" "${dashes}"
+    echo -ne "${dashes}"
     printf "Install Model Optimizer dependencies\n\n"
     cd "${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/install_prerequisites"
     . ./install_prerequisites.sh caffe
     cd "$cur_path"
 
     # Step 3. Convert a model with Model Optimizer
-    printf "%s" "${dashes}"
+    echo -ne "${dashes}"
     printf "Convert a model with Model Optimizer\n\n"
 
     mo_path="${INTEL_OPENVINO_DIR}/deployment_tools/model_optimizer/mo.py"
@@ -178,11 +178,11 @@ if [ ! -e "$ir_dir" ]; then
     print_and_run "$python_binary" "$downloader_dir/converter.py" --mo "$mo_path" --name "$model_name" -d "$models_path" -o "$irs_path" --precisions "$target_precision"
 else
     printf "\n\nTarget folder %s already exists. Skipping IR generation  with Model Optimizer." "${ir_dir}"
-    printf "If you want to convert a model again, remove the entire %s folder. %s" "${ir_dir}" "${run_again}"
+    echo -ne "If you want to convert a model again, remove the entire ${ir_dir} folder. ${run_again}"
 fi
 
 # Step 4. Build samples
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "Build Inference Engine samples\n\n"
 
 OS_PATH=$(uname -m)
@@ -207,14 +207,14 @@ cmake -DCMAKE_BUILD_TYPE=Release "$samples_path"
 make $NUM_THREADS classification_sample_async
 
 # Step 5. Run samples
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "Run Inference Engine classification sample\n\n"
 
 cd "$binaries_dir"
 
 cp -f "$ROOT_DIR/${model_name}.labels" "${ir_dir}/"
 
-print_and_run ./classification_sample_async -d "$target" -i "$target_image_path" -m "${ir_dir}/${model_name}.xml" "${sampleoptions}"
+print_and_run ./classification_sample_async -d "$target" -i "$target_image_path" -m "${ir_dir}/${model_name}.xml" "${sampleoptions[@]}"
 
-printf "%s" "${dashes}"
+echo -ne "${dashes}"
 printf "Demo completed successfully.\n\n"

From 6926d5d6398ffd56725ffa57e38f57a3a1e1cae8 Mon Sep 17 00:00:00 2001
From: Mikhail Nosov <mikhail.nosov@intel.com>
Date: Tue, 22 Jun 2021 08:24:47 +0300
Subject: [PATCH 152/178] [nGraph FE] Python bindings for nGraph
 FrontEndManager API (#5995)

* [NG Frontend] Introduce classes and macros for Frontend-specific exceptions

* [nGraph Frontend] Python bindings

Exposing Python API for nGraph FrontEndManager introduced in PR #5470

* [nGraph Frontend] Python bindings

Exposing Python API for nGraph FrontEndManager introduced in PR #5470

* Added docs + moved sources to 'pyngraph/frontend'

* Fix issues after rebase

* Fix clang

* Moving ngraph/python/tests_cpp to ngraph/python/tests/mock

* Fix cmakelists-related comments

* Apply review comments

* Added dots to docstrings

* Cmake - Set target properties for ARCHIVE/PDB/COMPILE_PDB as well as for LIBRARY_OUTPUT_DIRECTORY
---
 .../frontend/frontend_manager/CMakeLists.txt  |  15 +
 ngraph/python/CMakeLists.txt                  |  17 +-
 ngraph/python/setup.py                        |   1 +
 ngraph/python/src/ngraph/__init__.py          |  16 +-
 ngraph/python/src/ngraph/frontend/__init__.py |  23 +
 .../python/src/pyngraph/frontend/frontend.cpp | 124 ++++
 .../python/src/pyngraph/frontend/frontend.hpp |  11 +
 .../pyngraph/frontend/frontend_manager.cpp    | 163 ++++++
 .../pyngraph/frontend/frontend_manager.hpp    |  18 +
 .../src/pyngraph/frontend/inputmodel.cpp      | 364 ++++++++++++
 .../src/pyngraph/frontend/inputmodel.hpp      |  11 +
 ngraph/python/src/pyngraph/frontend/place.cpp | 319 ++++++++++
 ngraph/python/src/pyngraph/frontend/place.hpp |  11 +
 ngraph/python/src/pyngraph/pyngraph.cpp       |  14 +
 .../mock_py_ngraph_frontend/CMakeLists.txt    |  22 +
 .../mock_py_frontend.cpp                      |  25 +
 .../mock_py_frontend.hpp                      | 542 +++++++++++++++++
 .../mock/pyngraph_fe_mock_api/CMakeLists.txt  |  19 +
 .../pyngraph_mock_frontend_api.cpp            | 136 +++++
 .../tests/test_ngraph/test_frontendmanager.py | 545 ++++++++++++++++++
 20 files changed, 2393 insertions(+), 3 deletions(-)
 create mode 100644 ngraph/python/src/ngraph/frontend/__init__.py
 create mode 100644 ngraph/python/src/pyngraph/frontend/frontend.cpp
 create mode 100644 ngraph/python/src/pyngraph/frontend/frontend.hpp
 create mode 100644 ngraph/python/src/pyngraph/frontend/frontend_manager.cpp
 create mode 100644 ngraph/python/src/pyngraph/frontend/frontend_manager.hpp
 create mode 100644 ngraph/python/src/pyngraph/frontend/inputmodel.cpp
 create mode 100644 ngraph/python/src/pyngraph/frontend/inputmodel.hpp
 create mode 100644 ngraph/python/src/pyngraph/frontend/place.cpp
 create mode 100644 ngraph/python/src/pyngraph/frontend/place.hpp
 create mode 100644 ngraph/python/tests/mock/mock_py_ngraph_frontend/CMakeLists.txt
 create mode 100644 ngraph/python/tests/mock/mock_py_ngraph_frontend/mock_py_frontend.cpp
 create mode 100644 ngraph/python/tests/mock/mock_py_ngraph_frontend/mock_py_frontend.hpp
 create mode 100644 ngraph/python/tests/mock/pyngraph_fe_mock_api/CMakeLists.txt
 create mode 100644 ngraph/python/tests/mock/pyngraph_fe_mock_api/pyngraph_mock_frontend_api.cpp
 create mode 100644 ngraph/python/tests/test_ngraph/test_frontendmanager.py

diff --git a/ngraph/frontend/frontend_manager/CMakeLists.txt b/ngraph/frontend/frontend_manager/CMakeLists.txt
index 5efd560357b357..d11be1439114ff 100644
--- a/ngraph/frontend/frontend_manager/CMakeLists.txt
+++ b/ngraph/frontend/frontend_manager/CMakeLists.txt
@@ -33,3 +33,18 @@ target_include_directories(${TARGET_NAME} PUBLIC $<BUILD_INTERFACE:${FRONTEND_IN
 target_include_directories(${TARGET_NAME} PRIVATE ${NGRAPH_INCLUDE_PATH} ${FRONTEND_INCLUDE_DIR})
 
 target_include_directories(${TARGET_NAME} PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/src)
+
+# Installation rules
+install(TARGETS ${TARGET_NAME} EXPORT ngraphTargets
+        RUNTIME DESTINATION ${NGRAPH_INSTALL_LIB} COMPONENT ngraph
+        ARCHIVE DESTINATION ${NGRAPH_INSTALL_LIB} COMPONENT ngraph
+        LIBRARY DESTINATION ${NGRAPH_INSTALL_LIB} COMPONENT ngraph)
+
+install(DIRECTORY ${FRONTEND_INCLUDE_DIR}/frontend_manager
+    DESTINATION ${FRONTEND_INSTALL_INCLUDE}
+    COMPONENT ngraph
+    FILES_MATCHING
+        PATTERN "*.hpp"
+)
+
+export(TARGETS ${TARGET_NAME} NAMESPACE ngraph:: APPEND FILE "${NGRAPH_TARGETS_FILE}")
diff --git a/ngraph/python/CMakeLists.txt b/ngraph/python/CMakeLists.txt
index 58d30c300fdf3b..5f30c92fef4a31 100644
--- a/ngraph/python/CMakeLists.txt
+++ b/ngraph/python/CMakeLists.txt
@@ -33,6 +33,7 @@ endif()
 set(PYTHON_VERSION python${PYTHON_VERSION_MAJOR}.${PYTHON_VERSION_MINOR})
 message("Python version=${PYTHON_VERSION}")
 
+set(LIBRARY_OUTPUT_DIRECTORY_BIN ${CMAKE_LIBRARY_OUTPUT_DIRECTORY})
 if(OpenVINO_MAIN_SOURCE_DIR)
     if(WIN32)
         set(PYTHON_BRIDGE_OUTPUT_DIRECTORY ${CMAKE_LIBRARY_OUTPUT_DIRECTORY}/$<CONFIG>/python_api/${PYTHON_VERSION}/)
@@ -73,11 +74,25 @@ file(GLOB_RECURSE SOURCES src/pyngraph/*.cpp)
 pybind11_add_module(_${PROJECT_NAME} MODULE ${SOURCES})
 
 target_include_directories(_${PROJECT_NAME} PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/src")
-target_link_libraries(_${PROJECT_NAME} PRIVATE ngraph::ngraph)
+
+target_link_libraries(_${PROJECT_NAME} PRIVATE ngraph::ngraph ngraph::frontend_manager)
+
 if (TARGET ngraph::onnx_importer)
     add_dependencies(_${PROJECT_NAME} ngraph::onnx_importer)
 endif()
 
+if(NGRAPH_UNIT_TEST_ENABLE)
+    add_subdirectory(tests/mock/mock_py_ngraph_frontend)
+    add_dependencies(_${PROJECT_NAME} mock_py_ngraph_frontend)
+    set_target_properties(mock_py_ngraph_frontend PROPERTIES LIBRARY_OUTPUT_DIRECTORY ${LIBRARY_OUTPUT_DIRECTORY_BIN}
+            ARCHIVE_OUTPUT_DIRECTORY ${LIBRARY_OUTPUT_DIRECTORY_BIN}
+            COMPILE_PDB_OUTPUT_DIRECTORY ${LIBRARY_OUTPUT_DIRECTORY_BIN}
+            PDB_OUTPUT_DIRECTORY ${LIBRARY_OUTPUT_DIRECTORY_BIN})
+
+    add_subdirectory(tests/mock/pyngraph_fe_mock_api)
+    add_dependencies(_${PROJECT_NAME} pybind_mock_frontend)
+endif()
+
 # perform copy
 if(OpenVINO_MAIN_SOURCE_DIR)
     add_custom_command(TARGET _${PROJECT_NAME}
diff --git a/ngraph/python/setup.py b/ngraph/python/setup.py
index dfbb03fad98c00..e90cc66d6d173d 100644
--- a/ngraph/python/setup.py
+++ b/ngraph/python/setup.py
@@ -39,6 +39,7 @@
     "ngraph.impl.op",
     "ngraph.impl.op.util",
     "ngraph.impl.passes",
+    "ngraph.frontend",
 ]
 
 data_files = []
diff --git a/ngraph/python/src/ngraph/__init__.py b/ngraph/python/src/ngraph/__init__.py
index 0b276049d33ea8..f51c5cea130510 100644
--- a/ngraph/python/src/ngraph/__init__.py
+++ b/ngraph/python/src/ngraph/__init__.py
@@ -11,11 +11,23 @@
 except DistributionNotFound:
     __version__ = "0.0.0.dev0"
 
-from ngraph.impl import Node
+
+from ngraph.impl import Dimension
 from ngraph.impl import Function
+from ngraph.impl import Node
+from ngraph.impl import PartialShape
+from ngraph.frontend import FrontEnd
+from ngraph.frontend import FrontEndCapabilities
+from ngraph.frontend import FrontEndManager
+from ngraph.frontend import GeneralFailure
+from ngraph.frontend import NotImplementedFailure
+from ngraph.frontend import InitializationFailure
+from ngraph.frontend import InputModel
+from ngraph.frontend import OpConversionFailure
+from ngraph.frontend import OpValidationFailure
+from ngraph.frontend import Place
 from ngraph.helpers import function_from_cnn
 from ngraph.helpers import function_to_cnn
-
 from ngraph.opset7 import absolute
 from ngraph.opset7 import absolute as abs
 from ngraph.opset7 import acos
diff --git a/ngraph/python/src/ngraph/frontend/__init__.py b/ngraph/python/src/ngraph/frontend/__init__.py
new file mode 100644
index 00000000000000..0ea21ad7c8827b
--- /dev/null
+++ b/ngraph/python/src/ngraph/frontend/__init__.py
@@ -0,0 +1,23 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+"""
+Package: ngraph
+Low level wrappers for the FrontEnd c++ api.
+"""
+
+# flake8: noqa
+
+# main classes
+from _pyngraph import FrontEndManager
+from _pyngraph import FrontEnd
+from _pyngraph import FrontEndCapabilities
+from _pyngraph import InputModel
+from _pyngraph import Place
+
+# exceptions
+from _pyngraph import NotImplementedFailure
+from _pyngraph import InitializationFailure
+from _pyngraph import OpConversionFailure
+from _pyngraph import OpValidationFailure
+from _pyngraph import GeneralFailure
diff --git a/ngraph/python/src/pyngraph/frontend/frontend.cpp b/ngraph/python/src/pyngraph/frontend/frontend.cpp
new file mode 100644
index 00000000000000..ecc736b37f0269
--- /dev/null
+++ b/ngraph/python/src/pyngraph/frontend/frontend.cpp
@@ -0,0 +1,124 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+#include <pybind11/stl_bind.h>
+
+#include "frontend_manager.hpp"
+#include "frontend_manager/frontend_exceptions.hpp"
+#include "frontend_manager/frontend_manager.hpp"
+#include "pyngraph/function.hpp"
+
+namespace py = pybind11;
+
+void regclass_pyngraph_FrontEnd(py::module m)
+{
+    py::class_<ngraph::frontend::FrontEnd, std::shared_ptr<ngraph::frontend::FrontEnd>> fem(
+        m, "FrontEnd", py::dynamic_attr());
+    fem.doc() = "ngraph.impl.FrontEnd wraps ngraph::frontend::FrontEnd";
+
+    fem.def("load_from_file",
+            &ngraph::frontend::FrontEnd::load_from_file,
+            py::arg("path"),
+            R"(
+                Loads an input model by specified model file path.
+
+                Parameters
+                ----------
+                path : str
+                    Main model file path.
+
+                Returns
+                ----------
+                load_from_file : InputModel
+                    Loaded input model.
+             )");
+
+    fem.def("convert",
+            static_cast<std::shared_ptr<ngraph::Function> (ngraph::frontend::FrontEnd::*)(
+                ngraph::frontend::InputModel::Ptr) const>(&ngraph::frontend::FrontEnd::convert),
+            py::arg("model"),
+            R"(
+                Completely convert and normalize entire function, throws if it is not possible.
+
+                Parameters
+                ----------
+                model : InputModel
+                    Input model.
+
+                Returns
+                ----------
+                convert : Function
+                    Fully converted nGraph function.
+             )");
+
+    fem.def("convert",
+            static_cast<std::shared_ptr<ngraph::Function> (ngraph::frontend::FrontEnd::*)(
+                std::shared_ptr<ngraph::Function>) const>(&ngraph::frontend::FrontEnd::convert),
+            py::arg("function"),
+            R"(
+                Completely convert the remaining, not converted part of a function.
+
+                Parameters
+                ----------
+                function : Function
+                    Partially converted nGraph function.
+
+                Returns
+                ----------
+                convert : Function
+                    Fully converted nGraph function.
+             )");
+
+    fem.def("convert_partially",
+            &ngraph::frontend::FrontEnd::convert_partially,
+            py::arg("model"),
+            R"(
+                Convert only those parts of the model that can be converted leaving others as-is.
+                Converted parts are not normalized by additional transformations; normalize function or
+                another form of convert function should be called to finalize the conversion process.
+
+                Parameters
+                ----------
+                model : InputModel
+                    Input model.
+
+                Returns
+                ----------
+                convert_partially : Function
+                    Partially converted nGraph function.
+             )");
+
+    fem.def("decode",
+            &ngraph::frontend::FrontEnd::decode,
+            py::arg("model"),
+            R"(
+                Convert operations with one-to-one mapping with decoding nodes.
+                Each decoding node is an nGraph node representing a single FW operation node with
+                all attributes represented in FW-independent way.
+
+                Parameters
+                ----------
+                model : InputModel
+                    Input model.
+
+                Returns
+                ----------
+                decode : Function
+                    nGraph function after decoding.
+             )");
+
+    fem.def("normalize",
+            &ngraph::frontend::FrontEnd::normalize,
+            py::arg("function"),
+            R"(
+                Runs normalization passes on function that was loaded with partial conversion.
+
+                Parameters
+                ----------
+                function : Function
+                    Partially converted nGraph function.
+             )");
+}
diff --git a/ngraph/python/src/pyngraph/frontend/frontend.hpp b/ngraph/python/src/pyngraph/frontend/frontend.hpp
new file mode 100644
index 00000000000000..de28e950bb5c97
--- /dev/null
+++ b/ngraph/python/src/pyngraph/frontend/frontend.hpp
@@ -0,0 +1,11 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <pybind11/pybind11.h>
+
+namespace py = pybind11;
+
+void regclass_pyngraph_FrontEnd(py::module m);
diff --git a/ngraph/python/src/pyngraph/frontend/frontend_manager.cpp b/ngraph/python/src/pyngraph/frontend/frontend_manager.cpp
new file mode 100644
index 00000000000000..a550f22e516257
--- /dev/null
+++ b/ngraph/python/src/pyngraph/frontend/frontend_manager.cpp
@@ -0,0 +1,163 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+#include <pybind11/stl_bind.h>
+
+#include "frontend_manager.hpp"
+#include "frontend_manager/frontend_exceptions.hpp"
+#include "frontend_manager/frontend_manager.hpp"
+
+namespace py = pybind11;
+
+void regclass_pyngraph_FrontEndManager(py::module m)
+{
+    py::class_<ngraph::frontend::FrontEndManager,
+               std::shared_ptr<ngraph::frontend::FrontEndManager>>
+        fem(m, "FrontEndManager", py::dynamic_attr());
+    fem.doc() = "ngraph.impl.FrontEndManager wraps ngraph::frontend::FrontEndManager";
+
+    fem.def(py::init<>());
+
+    fem.def("get_available_front_ends",
+            &ngraph::frontend::FrontEndManager::get_available_front_ends,
+            R"(
+                Gets list of registered frontends.
+
+                Returns
+                ----------
+                get_available_front_ends : List[str]
+                    List of available frontend names.
+             )");
+    fem.def("load_by_framework",
+            &ngraph::frontend::FrontEndManager::load_by_framework,
+            py::arg("framework"),
+            py::arg("capabilities") = ngraph::frontend::FrontEndCapabilities::FEC_DEFAULT,
+            R"(
+                Loads frontend by name of framework and capabilities.
+
+                Parameters
+                ----------
+                framework : str
+                    Framework name. Throws exception if name is not in list of available frontends.
+
+                capabilities : int
+                    Frontend capabilities. Default is FrontEndCapabilities.FEC_DEFAULT. It is recommended to use only
+                    those capabilities which are needed to minimize load time.
+
+                Returns
+                ----------
+                load_by_framework : FrontEnd
+                    Frontend interface for further loading of models.
+             )");
+}
+
+void regclass_pyngraph_FEC(py::module m)
+{
+    class FeCaps
+    {
+    public:
+        int get_caps() const { return m_caps; }
+
+    private:
+        int m_caps;
+    };
+
+    py::class_<FeCaps, std::shared_ptr<FeCaps>> type(m, "FrontEndCapabilities");
+    // type.doc() = "FrontEndCapabilities";
+    type.attr("DEFAULT") = ngraph::frontend::FrontEndCapabilities::FEC_DEFAULT;
+    type.attr("CUT") = ngraph::frontend::FrontEndCapabilities::FEC_CUT;
+    type.attr("NAMES") = ngraph::frontend::FrontEndCapabilities::FEC_NAMES;
+    type.attr("WILDCARDS") = ngraph::frontend::FrontEndCapabilities::FEC_WILDCARDS;
+
+    type.def(
+        "__eq__",
+        [](const FeCaps& a, const FeCaps& b) { return a.get_caps() == b.get_caps(); },
+        py::is_operator());
+}
+
+void regclass_pyngraph_GeneralFailureFrontEnd(py::module m)
+{
+    static py::exception<ngraph::frontend::GeneralFailure> exc(std::move(m), "GeneralFailure");
+    py::register_exception_translator([](std::exception_ptr p) {
+        try
+        {
+            if (p)
+                std::rethrow_exception(p);
+        }
+        catch (const ngraph::frontend::GeneralFailure& e)
+        {
+            exc(e.what());
+        }
+    });
+}
+
+void regclass_pyngraph_OpValidationFailureFrontEnd(py::module m)
+{
+    static py::exception<ngraph::frontend::OpValidationFailure> exc(std::move(m),
+                                                                    "OpValidationFailure");
+    py::register_exception_translator([](std::exception_ptr p) {
+        try
+        {
+            if (p)
+                std::rethrow_exception(p);
+        }
+        catch (const ngraph::frontend::OpValidationFailure& e)
+        {
+            exc(e.what());
+        }
+    });
+}
+
+void regclass_pyngraph_OpConversionFailureFrontEnd(py::module m)
+{
+    static py::exception<ngraph::frontend::OpConversionFailure> exc(std::move(m),
+                                                                    "OpConversionFailure");
+    py::register_exception_translator([](std::exception_ptr p) {
+        try
+        {
+            if (p)
+                std::rethrow_exception(p);
+        }
+        catch (const ngraph::frontend::OpConversionFailure& e)
+        {
+            exc(e.what());
+        }
+    });
+}
+
+void regclass_pyngraph_InitializationFailureFrontEnd(py::module m)
+{
+    static py::exception<ngraph::frontend::InitializationFailure> exc(std::move(m),
+                                                                      "InitializationFailure");
+    py::register_exception_translator([](std::exception_ptr p) {
+        try
+        {
+            if (p)
+                std::rethrow_exception(p);
+        }
+        catch (const ngraph::frontend::InitializationFailure& e)
+        {
+            exc(e.what());
+        }
+    });
+}
+
+void regclass_pyngraph_NotImplementedFailureFrontEnd(py::module m)
+{
+    static py::exception<ngraph::frontend::NotImplementedFailure> exc(std::move(m),
+                                                                      "NotImplementedFailure");
+    py::register_exception_translator([](std::exception_ptr p) {
+        try
+        {
+            if (p)
+                std::rethrow_exception(p);
+        }
+        catch (const ngraph::frontend::NotImplementedFailure& e)
+        {
+            exc(e.what());
+        }
+    });
+}
diff --git a/ngraph/python/src/pyngraph/frontend/frontend_manager.hpp b/ngraph/python/src/pyngraph/frontend/frontend_manager.hpp
new file mode 100644
index 00000000000000..35caa7e5dd1def
--- /dev/null
+++ b/ngraph/python/src/pyngraph/frontend/frontend_manager.hpp
@@ -0,0 +1,18 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <pybind11/pybind11.h>
+
+namespace py = pybind11;
+
+void regclass_pyngraph_FrontEndManager(py::module m);
+void regclass_pyngraph_FEC(py::module m);
+void regclass_pyngraph_NotImplementedFailureFrontEnd(py::module m);
+void regclass_pyngraph_InitializationFailureFrontEnd(py::module m);
+void regclass_pyngraph_OpConversionFailureFrontEnd(py::module m);
+void regclass_pyngraph_OpValidationFailureFrontEnd(py::module m);
+void regclass_pyngraph_GeneralFailureFrontEnd(py::module m);
+
diff --git a/ngraph/python/src/pyngraph/frontend/inputmodel.cpp b/ngraph/python/src/pyngraph/frontend/inputmodel.cpp
new file mode 100644
index 00000000000000..f8775763806d97
--- /dev/null
+++ b/ngraph/python/src/pyngraph/frontend/inputmodel.cpp
@@ -0,0 +1,364 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+#include <pybind11/stl_bind.h>
+
+#include "frontend_manager/frontend_exceptions.hpp"
+#include "frontend_manager/frontend_manager.hpp"
+
+namespace py = pybind11;
+
+void regclass_pyngraph_InputModel(py::module m)
+{
+    py::class_<ngraph::frontend::InputModel, std::shared_ptr<ngraph::frontend::InputModel>> im(
+        m, "InputModel", py::dynamic_attr());
+    im.doc() = "ngraph.impl.InputModel wraps ngraph::frontend::InputModel";
+
+    im.def("get_place_by_tensor_name",
+           &ngraph::frontend::InputModel::get_place_by_tensor_name,
+           py::arg("tensorName"),
+           R"(
+                Returns a tensor place by a tensor name following framework conventions, or
+                nullptr if a tensor with this name doesn't exist.
+
+                Parameters
+                ----------
+                tensorName : str
+                    Name of tensor.
+
+                Returns
+                ----------
+                get_place_by_tensor_name : Place
+                    Tensor place corresponding to specified tensor name.
+             )");
+
+    im.def("get_place_by_operation_name",
+           &ngraph::frontend::InputModel::get_place_by_operation_name,
+           py::arg("operationName"),
+           R"(
+                Returns an operation place by an operation name following framework conventions, or
+                nullptr if an operation with this name doesn't exist.
+
+                Parameters
+                ----------
+                operationName : str
+                    Name of operation.
+
+                Returns
+                ----------
+                get_place_by_operation_name : Place
+                    Place representing operation.
+             )");
+
+    im.def("get_place_by_operation_name_and_input_port",
+           &ngraph::frontend::InputModel::get_place_by_operation_name_and_input_port,
+           py::arg("operationName"),
+           py::arg("inputPortIndex"),
+           R"(
+                Returns an input port place by operation name and appropriate port index.
+
+                Parameters
+                ----------
+                operationName : str
+                    Name of operation.
+
+                inputPortIndex : int
+                    Index of input port for this operation.
+
+                Returns
+                ----------
+                get_place_by_operation_name_and_input_port : Place
+                    Place representing input port of operation.
+             )");
+
+    im.def("get_place_by_operation_name_and_output_port",
+           &ngraph::frontend::InputModel::get_place_by_operation_name_and_output_port,
+           py::arg("operationName"),
+           py::arg("outputPortIndex"),
+           R"(
+                Returns an output port place by operation name and appropriate port index.
+
+                Parameters
+                ----------
+                operationName : str
+                    Name of operation.
+
+                outputPortIndex : int
+                    Index of output port for this operation.
+
+                Returns
+                ----------
+                get_place_by_operation_name_and_output_port : Place
+                    Place representing output port of operation.
+             )");
+
+    im.def("set_name_for_tensor",
+           &ngraph::frontend::InputModel::set_name_for_tensor,
+           py::arg("tensor"),
+           py::arg("newName"),
+           R"(
+                Sets name for tensor. Overwrites existing names of this place.
+
+                Parameters
+                ----------
+                tensor : Place
+                    Tensor place.
+
+                newName : str
+                    New name for this tensor.
+            )");
+
+    im.def("add_name_for_tensor",
+           &ngraph::frontend::InputModel::add_name_for_tensor,
+           py::arg("tensor"),
+           py::arg("newName"),
+           R"(
+                Adds new name for tensor
+
+                Parameters
+                ----------
+                tensor : Place
+                    Tensor place.
+
+                newName : str
+                    New name to be added to this place.
+            )");
+
+    im.def("set_name_for_operation",
+           &ngraph::frontend::InputModel::set_name_for_operation,
+           py::arg("operation"),
+           py::arg("newName"),
+           R"(
+                Adds new name for tensor.
+
+                Parameters
+                ----------
+                operation : Place
+                    Operation place.
+
+                newName : str
+                    New name for this operation.
+            )");
+
+    im.def("free_name_for_tensor",
+           &ngraph::frontend::InputModel::free_name_for_tensor,
+           py::arg("name"),
+           R"(
+                Unassign specified name from tensor place(s).
+
+                Parameters
+                ----------
+                name : str
+                    Name of tensor.
+            )");
+
+    im.def("free_name_for_operation",
+           &ngraph::frontend::InputModel::free_name_for_operation,
+           py::arg("name"),
+           R"(
+                Unassign specified name from operation place(s).
+
+                Parameters
+                ----------
+                name : str
+                    Name of operation.
+            )");
+
+    im.def("set_name_for_dimension",
+           &ngraph::frontend::InputModel::set_name_for_dimension,
+           py::arg("place"),
+           py::arg("dimIndex"),
+           py::arg("dimName"),
+           R"(
+                Set name for a particular dimension of a place (e.g. batch dimension).
+
+                Parameters
+                ----------
+                place : Place
+                    Model's place.
+
+                shapeDimIndex : int
+                    Dimension index.
+
+                dimName : str
+                    Name to assign on this dimension.
+            )");
+
+    im.def("cut_and_add_new_input",
+           &ngraph::frontend::InputModel::cut_and_add_new_input,
+           py::arg("place"),
+           py::arg("newName") = std::string(),
+           R"(
+                Cut immediately before this place and assign this place as new input; prune
+                all nodes that don't contribute to any output.
+
+                Parameters
+                ----------
+                place : Place
+                    New place to be assigned as input.
+
+                newNameOptional : str
+                    Optional new name assigned to this input place.
+            )");
+
+    im.def("cut_and_add_new_output",
+           &ngraph::frontend::InputModel::cut_and_add_new_output,
+           py::arg("place"),
+           py::arg("newName") = std::string(),
+           R"(
+                Cut immediately before this place and assign this place as new output; prune
+                all nodes that don't contribute to any output.
+
+                Parameters
+                ----------
+                place : Place
+                    New place to be assigned as output.
+
+                newNameOptional : str
+                    Optional new name assigned to this output place.
+            )");
+
+    im.def("add_output",
+           &ngraph::frontend::InputModel::add_output,
+           py::arg("place"),
+           R"(
+                Assign this place as new output or add necessary nodes to represent a new output.
+
+                Parameters
+                ----------
+                place : Place
+                    Anchor point to add an output.
+            )");
+
+    im.def("remove_output",
+           &ngraph::frontend::InputModel::remove_output,
+           py::arg("place"),
+           R"(
+                Removes any sinks directly attached to this place with all inbound data flow
+                if it is not required by any other output.
+
+                Parameters
+                ----------
+                place : Place
+                    Model place
+            )");
+
+    im.def("set_partial_shape",
+           &ngraph::frontend::InputModel::set_partial_shape,
+           py::arg("place"),
+           py::arg("shape"),
+           R"(
+                Defines all possible shape that may be used for this place; place should be
+                uniquely refer to some data. This partial shape will be converted to corresponding
+                shape of results ngraph nodes and will define shape inference when the model is
+                converted to ngraph.
+
+                Parameters
+                ----------
+                place : Place
+                    Model place.
+
+                shape : PartialShape
+                    Partial shape for this place.
+            )");
+
+    im.def("get_partial_shape",
+           &ngraph::frontend::InputModel::get_partial_shape,
+           py::arg("place"),
+           R"(
+                Returns current partial shape used for this place.
+
+                Parameters
+                ----------
+                place : Place
+                    Model place
+
+                Returns
+                ----------
+                get_partial_shape : PartialShape
+                    Partial shape for this place.
+            )");
+
+    im.def("get_inputs",
+           &ngraph::frontend::InputModel::get_inputs,
+           R"(
+                Returns all inputs for a model.
+
+                Returns
+                ----------
+                get_inputs : List[Place]
+                    A list of input places.
+            )");
+
+    im.def("get_outputs",
+           &ngraph::frontend::InputModel::get_outputs,
+           R"(
+                Returns all outputs for a model. An output is a terminal place in a graph where data escapes the flow.
+
+                Returns
+                ----------
+                get_outputs : List[Place]
+                    A list of output places
+            )");
+
+    im.def("extract_subgraph",
+           &ngraph::frontend::InputModel::extract_subgraph,
+           py::arg("inputs"),
+           py::arg("outputs"),
+           R"(
+                Leaves only subgraph that are defined by new inputs and new outputs.
+
+                Parameters
+                ----------
+                inputs : List[Place]
+                    Array of new input places.
+
+                outputs : List[Place]
+                    Array of new output places.
+            )");
+
+    im.def("override_all_inputs",
+           &ngraph::frontend::InputModel::override_all_inputs,
+           py::arg("inputs"),
+           R"(
+                Modifies the graph to use new inputs instead of existing ones. New inputs
+                should completely satisfy all existing outputs.
+
+                Parameters
+                ----------
+                inputs : List[Place]
+                    Array of new input places.
+            )");
+
+    im.def("override_all_outputs",
+           &ngraph::frontend::InputModel::override_all_outputs,
+           py::arg("outputs"),
+           R"(
+                Replaces all existing outputs with new ones removing all data flow that
+                is not required for new outputs.
+
+                Parameters
+                ----------
+                outputs : List[Place]
+                    Vector with places that will become new outputs; may intersect existing outputs.
+            )");
+
+    im.def("set_element_type",
+           &ngraph::frontend::InputModel::set_element_type,
+           py::arg("place"),
+           py::arg("type"),
+           R"(
+                Sets new element type for a place.
+
+                Parameters
+                ----------
+                place : Place
+                    Model place.
+
+                type : ngraph.Type
+                    New element type.
+            )");
+}
diff --git a/ngraph/python/src/pyngraph/frontend/inputmodel.hpp b/ngraph/python/src/pyngraph/frontend/inputmodel.hpp
new file mode 100644
index 00000000000000..e95179c4a8cddc
--- /dev/null
+++ b/ngraph/python/src/pyngraph/frontend/inputmodel.hpp
@@ -0,0 +1,11 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <pybind11/pybind11.h>
+
+namespace py = pybind11;
+
+void regclass_pyngraph_InputModel(py::module m);
diff --git a/ngraph/python/src/pyngraph/frontend/place.cpp b/ngraph/python/src/pyngraph/frontend/place.cpp
new file mode 100644
index 00000000000000..19bc36ef13a3b0
--- /dev/null
+++ b/ngraph/python/src/pyngraph/frontend/place.cpp
@@ -0,0 +1,319 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+#include <pybind11/stl_bind.h>
+
+#include "frontend_manager.hpp"
+#include "frontend_manager/frontend_exceptions.hpp"
+#include "frontend_manager/frontend_manager.hpp"
+#include "pyngraph/function.hpp"
+
+namespace py = pybind11;
+
+void regclass_pyngraph_Place(py::module m)
+{
+    py::class_<ngraph::frontend::Place, std::shared_ptr<ngraph::frontend::Place>> place(
+        m, "Place", py::dynamic_attr());
+    place.doc() = "ngraph.impl.Place wraps ngraph::frontend::Place";
+
+    place.def("is_input",
+              &ngraph::frontend::Place::is_input,
+              R"(
+                Returns true if this place is input for a model.
+
+                Returns
+                ----------
+                is_input : bool
+                    True if this place is input for a model
+             )");
+
+    place.def("is_output",
+              &ngraph::frontend::Place::is_output,
+              R"(
+                Returns true if this place is output for a model.
+
+                Returns
+                ----------
+                is_output : bool
+                    True if this place is output for a model.
+             )");
+
+    place.def("get_names",
+              &ngraph::frontend::Place::get_names,
+              R"(
+                All associated names (synonyms) that identify this place in the graph in a framework specific way.
+
+                Returns
+                ----------
+                get_names : List[str]
+                    A vector of strings each representing a name that identifies this place in the graph.
+                    Can be empty if there are no names associated with this place or name cannot be attached.
+             )");
+
+    place.def("is_equal",
+              &ngraph::frontend::Place::is_equal,
+              py::arg("other"),
+              R"(
+                Returns true if another place is the same as this place.
+
+                Parameters
+                ----------
+                other : Place
+                    Another place object.
+
+                Returns
+                ----------
+                is_equal : bool
+                    True if another place is the same as this place.
+             )");
+
+    place.def("is_equal_data",
+              &ngraph::frontend::Place::is_equal_data,
+              py::arg("other"),
+              R"(
+                Returns true if another place points to the same data.
+                Note: The same data means all places on path:
+                      output port -> output edge -> tensor -> input edge -> input port.
+
+                Parameters
+                ----------
+                other : Place
+                    Another place object.
+
+                Returns
+                ----------
+                is_equal_data : bool
+                    True if another place points to the same data.
+             )");
+
+    place.def(
+        "get_consuming_operations",
+        [](const ngraph::frontend::Place& self, py::object outputPortIndex) {
+            if (outputPortIndex == py::none())
+            {
+                return self.get_consuming_operations();
+            }
+            else
+            {
+                return self.get_consuming_operations(py::cast<int>(outputPortIndex));
+            }
+        },
+        py::arg("outputPortIndex") = py::none(),
+        R"(
+                Returns references to all operation nodes that consume data from this place for specified output port.
+                Note: It can be called for any kind of graph place searching for the first consuming operations.
+
+                Parameters
+                ----------
+                outputPortIndex : int
+                    If place is an operational node it specifies which output port should be considered
+                    May not be set if node has only one output port.
+
+                Returns
+                ----------
+                get_consuming_operations : List[Place]
+                    A list with all operation node references that consumes data from this place
+             )");
+
+    place.def(
+        "get_target_tensor",
+        [](const ngraph::frontend::Place& self, py::object outputPortIndex) {
+            if (outputPortIndex == py::none())
+            {
+                return self.get_target_tensor();
+            }
+            else
+            {
+                return self.get_target_tensor(py::cast<int>(outputPortIndex));
+            }
+        },
+        py::arg("outputPortIndex") = py::none(),
+        R"(
+                Returns a tensor place that gets data from this place; applicable for operations,
+                output ports and output edges.
+
+                Parameters
+                ----------
+                outputPortIndex : int
+                    Output port index if the current place is an operation node and has multiple output ports.
+                    May not be set if place has only one output port.
+
+                Returns
+                ----------
+                get_consuming_operations : Place
+                    A tensor place which hold the resulting value for this place.
+             )");
+
+    place.def(
+        "get_producing_operation",
+        [](const ngraph::frontend::Place& self, py::object inputPortIndex) {
+            if (inputPortIndex == py::none())
+            {
+                return self.get_producing_operation();
+            }
+            else
+            {
+                return self.get_producing_operation(py::cast<int>(inputPortIndex));
+            }
+        },
+        py::arg("inputPortIndex") = py::none(),
+        R"(
+                Get an operation node place that immediately produces data for this place.
+
+                Parameters
+                ----------
+                inputPortIndex : int
+                    If a given place is itself an operation node, this specifies a port index.
+                    May not be set if place has only one input port.
+
+                Returns
+                ----------
+                get_producing_operation : Place
+                    An operation place that produces data for this place.
+             )");
+
+    place.def("get_producing_port",
+              &ngraph::frontend::Place::get_producing_port,
+              R"(
+                Returns a port that produces data for this place.
+
+                Returns
+                ----------
+                get_producing_port : Place
+                    A port place that produces data for this place.
+             )");
+
+    place.def(
+        "get_input_port",
+        [](const ngraph::frontend::Place& self, py::object inputName, py::object inputPortIndex) {
+            if (inputName == py::none())
+            {
+                if (inputPortIndex == py::none())
+                {
+                    return self.get_input_port();
+                }
+                else
+                {
+                    return self.get_input_port(py::cast<int>(inputPortIndex));
+                }
+            }
+            else
+            {
+                if (inputPortIndex == py::none())
+                {
+                    return self.get_input_port(py::cast<std::string>(inputName));
+                }
+                else
+                {
+                    return self.get_input_port(py::cast<std::string>(inputName),
+                                               py::cast<int>(inputPortIndex));
+                }
+            }
+        },
+        py::arg("inputName") = py::none(),
+        py::arg("inputPortIndex") = py::none(),
+        R"(
+                For operation node returns reference to an input port with specified name and index.
+
+                Parameters
+                ----------
+                inputName : str
+                    Name of port group. May not be set if node has one input port group.
+
+                inputPortIndex : int
+                    Input port index in a group. May not be set if node has one input port in a group.
+
+                Returns
+                ----------
+                get_input_port : Place
+                    Appropriate input port place.
+             )");
+
+    place.def(
+        "get_output_port",
+        [](const ngraph::frontend::Place& self, py::object outputName, py::object outputPortIndex) {
+            if (outputName == py::none())
+            {
+                if (outputPortIndex == py::none())
+                {
+                    return self.get_output_port();
+                }
+                else
+                {
+                    return self.get_output_port(py::cast<int>(outputPortIndex));
+                }
+            }
+            else
+            {
+                if (outputPortIndex == py::none())
+                {
+                    return self.get_output_port(py::cast<std::string>(outputName));
+                }
+                else
+                {
+                    return self.get_output_port(py::cast<std::string>(outputName),
+                                                py::cast<int>(outputPortIndex));
+                }
+            }
+        },
+        py::arg("outputName") = py::none(),
+        py::arg("outputPortIndex") = py::none(),
+        R"(
+                For operation node returns reference to an output port with specified name and index.
+
+                Parameters
+                ----------
+                outputName : str
+                    Name of output port group. May not be set if node has one output port group.
+
+                outputPortIndex : int
+                    Output port index. May not be set if node has one output port in a group.
+
+                Returns
+                ----------
+                get_output_port : Place
+                    Appropriate output port place.
+             )");
+
+    place.def("get_consuming_ports",
+              &ngraph::frontend::Place::get_consuming_ports,
+              R"(
+                Returns all input ports that consume data flows through this place.
+
+                Returns
+                ----------
+                get_consuming_ports : List[Place]
+                    Input ports that consume data flows through this place.
+             )");
+
+    place.def(
+        "get_source_tensor",
+        [](const ngraph::frontend::Place& self, py::object inputPortIndex) {
+            if (inputPortIndex == py::none())
+            {
+                return self.get_source_tensor();
+            }
+            else
+            {
+                return self.get_source_tensor(py::cast<int>(inputPortIndex));
+            }
+        },
+        py::arg("inputPortIndex") = py::none(),
+        R"(
+                Returns a tensor place that supplies data for this place; applicable for operations,
+                input ports and input edges.
+
+                Parameters
+                ----------
+                inputPortIndex : int
+                    Input port index for operational node. May not be specified if place has only one input port.
+
+                Returns
+                ----------
+                get_source_tensor : Place
+                    A tensor place which supplies data for this place.
+             )");
+}
diff --git a/ngraph/python/src/pyngraph/frontend/place.hpp b/ngraph/python/src/pyngraph/frontend/place.hpp
new file mode 100644
index 00000000000000..1a52e21f028e24
--- /dev/null
+++ b/ngraph/python/src/pyngraph/frontend/place.hpp
@@ -0,0 +1,11 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <pybind11/pybind11.h>
+
+namespace py = pybind11;
+
+void regclass_pyngraph_Place(py::module m);
diff --git a/ngraph/python/src/pyngraph/pyngraph.cpp b/ngraph/python/src/pyngraph/pyngraph.cpp
index 92b507b64834e1..0849de45f58554 100644
--- a/ngraph/python/src/pyngraph/pyngraph.cpp
+++ b/ngraph/python/src/pyngraph/pyngraph.cpp
@@ -17,6 +17,10 @@
 #include "pyngraph/onnx_import/onnx_import.hpp"
 #endif
 #include "pyngraph/dimension.hpp"
+#include "pyngraph/frontend/frontend.hpp"
+#include "pyngraph/frontend/frontend_manager.hpp"
+#include "pyngraph/frontend/inputmodel.hpp"
+#include "pyngraph/frontend/place.hpp"
 #include "pyngraph/ops/constant.hpp"
 #include "pyngraph/ops/parameter.hpp"
 #include "pyngraph/ops/result.hpp"
@@ -41,6 +45,16 @@ PYBIND11_MODULE(_pyngraph, m)
     regclass_pyngraph_Shape(m);
     regclass_pyngraph_PartialShape(m);
     regclass_pyngraph_Node(m);
+    regclass_pyngraph_Place(m);
+    regclass_pyngraph_InitializationFailureFrontEnd(m);
+    regclass_pyngraph_GeneralFailureFrontEnd(m);
+    regclass_pyngraph_OpConversionFailureFrontEnd(m);
+    regclass_pyngraph_OpValidationFailureFrontEnd(m);
+    regclass_pyngraph_NotImplementedFailureFrontEnd(m);
+    regclass_pyngraph_FEC(m);
+    regclass_pyngraph_FrontEndManager(m);
+    regclass_pyngraph_FrontEnd(m);
+    regclass_pyngraph_InputModel(m);
     regclass_pyngraph_Input(m);
     regclass_pyngraph_Output(m);
     regclass_pyngraph_NodeFactory(m);
diff --git a/ngraph/python/tests/mock/mock_py_ngraph_frontend/CMakeLists.txt b/ngraph/python/tests/mock/mock_py_ngraph_frontend/CMakeLists.txt
new file mode 100644
index 00000000000000..d39827b0f18830
--- /dev/null
+++ b/ngraph/python/tests/mock/mock_py_ngraph_frontend/CMakeLists.txt
@@ -0,0 +1,22 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+#
+
+set(TARGET_FE_NAME "mock_py_ngraph_frontend")
+
+file(GLOB_RECURSE LIBRARY_SRC ${CMAKE_CURRENT_SOURCE_DIR}/*.cpp)
+file(GLOB_RECURSE LIBRARY_HEADERS ${CMAKE_CURRENT_SOURCE_DIR}/*.hpp)
+
+source_group("src" FILES ${LIBRARY_SRC})
+source_group("include" FILES ${LIBRARY_HEADERS})
+
+# Create shared library
+add_library(${TARGET_FE_NAME} SHARED ${LIBRARY_SRC} ${LIBRARY_HEADERS})
+
+target_include_directories(${TARGET_FE_NAME} PRIVATE ".")
+
+target_include_directories(${TARGET_FE_NAME} PRIVATE ${FRONTEND_INCLUDE_PATH} ${NGRAPH_INCLUDE_PATH})
+target_link_libraries(${TARGET_FE_NAME} PRIVATE frontend_manager)
+target_link_libraries(${TARGET_FE_NAME} PUBLIC ngraph  PRIVATE ngraph::builder)
+
+add_clang_format_target(${TARGET_FE_NAME}_clang FOR_TARGETS ${TARGET_FE_NAME})
diff --git a/ngraph/python/tests/mock/mock_py_ngraph_frontend/mock_py_frontend.cpp b/ngraph/python/tests/mock/mock_py_ngraph_frontend/mock_py_frontend.cpp
new file mode 100644
index 00000000000000..22a6e23a2b0a36
--- /dev/null
+++ b/ngraph/python/tests/mock/mock_py_ngraph_frontend/mock_py_frontend.cpp
@@ -0,0 +1,25 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "mock_py_frontend.hpp"
+#include "frontend_manager/frontend_manager.hpp"
+#include "frontend_manager/frontend_manager_defs.hpp"
+#include "ngraph/visibility.hpp"
+
+using namespace ngraph;
+using namespace ngraph::frontend;
+
+extern "C" MOCK_API FrontEndVersion GetAPIVersion()
+{
+    return OV_FRONTEND_API_VERSION;
+}
+
+extern "C" MOCK_API void* GetFrontEndData()
+{
+    FrontEndPluginInfo* res = new FrontEndPluginInfo();
+    res->m_name = "mock_py";
+    res->m_creator = [](FrontEndCapFlags flags) { return std::make_shared<FrontEndMockPy>(flags); };
+
+    return res;
+}
diff --git a/ngraph/python/tests/mock/mock_py_ngraph_frontend/mock_py_frontend.hpp b/ngraph/python/tests/mock/mock_py_ngraph_frontend/mock_py_frontend.hpp
new file mode 100644
index 00000000000000..651e9e53809683
--- /dev/null
+++ b/ngraph/python/tests/mock/mock_py_ngraph_frontend/mock_py_frontend.hpp
@@ -0,0 +1,542 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "frontend_manager/frontend_manager.hpp"
+#include "frontend_manager/frontend_manager_defs.hpp"
+#include "ngraph/visibility.hpp"
+
+// Defined if we are building the plugin DLL (instead of using it)
+#ifdef mock_py_ngraph_frontend_EXPORTS
+#define MOCK_API NGRAPH_HELPER_DLL_EXPORT
+#else
+#define MOCK_API NGRAPH_HELPER_DLL_IMPORT
+#endif // mock1_ngraph_frontend_EXPORTS
+
+// OK to have 'using' in mock header
+
+using namespace ngraph;
+using namespace ngraph::frontend;
+
+////////////////////////////////
+
+struct MOCK_API PlaceStat
+{
+    int m_get_names = 0;
+    int m_get_consuming_operations = 0;
+    int m_get_target_tensor = 0;
+    int m_get_producing_operation = 0;
+    int m_get_producing_port = 0;
+    int m_get_input_port = 0;
+    int m_get_output_port = 0;
+    int m_get_consuming_ports = 0;
+    int m_is_input = 0;
+    int m_is_output = 0;
+    int m_is_equal = 0;
+    int m_is_equal_data = 0;
+    int m_get_source_tensor = 0;
+
+    // Arguments tracking
+    std::string m_lastArgString;
+    int m_lastArgInt;
+    Place::Ptr m_lastArgPlace = nullptr;
+
+    // Getters
+    int get_names() const { return m_get_names; }
+    int get_consuming_operations() const { return m_get_consuming_operations; }
+    int get_target_tensor() const { return m_get_target_tensor; }
+    int get_producing_operation() const { return m_get_producing_operation; }
+    int get_producing_port() const { return m_get_producing_port; }
+    int get_input_port() const { return m_get_input_port; }
+    int get_output_port() const { return m_get_output_port; }
+    int get_consuming_ports() const { return m_get_consuming_ports; }
+    int is_input() const { return m_is_input; }
+    int is_output() const { return m_is_output; }
+    int is_equal() const { return m_is_equal; }
+    int is_equal_data() const { return m_is_equal_data; }
+    int get_source_tensor() const { return m_get_source_tensor; }
+
+    // Arguments getters
+    std::string get_lastArgString() const { return m_lastArgString; }
+    int get_lastArgInt() const { return m_lastArgInt; }
+    Place::Ptr get_lastArgPlace() const { return m_lastArgPlace; }
+};
+
+class MOCK_API PlaceMockPy : public Place
+{
+    mutable PlaceStat m_stat;
+
+public:
+    std::vector<std::string> get_names() const override
+    {
+        m_stat.m_get_names++;
+        return {};
+    }
+
+    std::vector<Place::Ptr> get_consuming_operations() const override
+    {
+        m_stat.m_get_consuming_operations++;
+        m_stat.m_lastArgInt = -1;
+        return {std::make_shared<PlaceMockPy>()};
+    }
+
+    std::vector<Place::Ptr> get_consuming_operations(int outputPortIndex) const override
+    {
+        m_stat.m_get_consuming_operations++;
+        m_stat.m_lastArgInt = outputPortIndex;
+        return {std::make_shared<PlaceMockPy>()};
+    }
+
+    Place::Ptr get_target_tensor() const override
+    {
+        m_stat.m_get_target_tensor++;
+        m_stat.m_lastArgInt = -1;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_target_tensor(int outputPortIndex) const override
+    {
+        m_stat.m_get_target_tensor++;
+        m_stat.m_lastArgInt = outputPortIndex;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_producing_operation() const override
+    {
+        m_stat.m_get_producing_operation++;
+        m_stat.m_lastArgInt = -1;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_producing_operation(int inputPortIndex) const override
+    {
+        m_stat.m_get_producing_operation++;
+        m_stat.m_lastArgInt = inputPortIndex;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_producing_port() const override
+    {
+        m_stat.m_get_producing_port++;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_input_port() const override
+    {
+        m_stat.m_get_input_port++;
+        m_stat.m_lastArgInt = -1;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_input_port(int inputPortIndex) const override
+    {
+        m_stat.m_get_input_port++;
+        m_stat.m_lastArgInt = inputPortIndex;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_input_port(const std::string& inputName) const override
+    {
+        m_stat.m_get_input_port++;
+        m_stat.m_lastArgInt = -1;
+        m_stat.m_lastArgString = inputName;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_input_port(const std::string& inputName, int inputPortIndex) const override
+    {
+        m_stat.m_get_input_port++;
+        m_stat.m_lastArgInt = inputPortIndex;
+        m_stat.m_lastArgString = inputName;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_output_port() const override
+    {
+        m_stat.m_get_output_port++;
+        m_stat.m_lastArgInt = -1;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_output_port(int outputPortIndex) const override
+    {
+        m_stat.m_get_output_port++;
+        m_stat.m_lastArgInt = outputPortIndex;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_output_port(const std::string& outputName) const override
+    {
+        m_stat.m_get_output_port++;
+        m_stat.m_lastArgInt = -1;
+        m_stat.m_lastArgString = outputName;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_output_port(const std::string& outputName, int outputPortIndex) const override
+    {
+        m_stat.m_get_output_port++;
+        m_stat.m_lastArgInt = outputPortIndex;
+        m_stat.m_lastArgString = outputName;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    std::vector<Place::Ptr> get_consuming_ports() const override
+    {
+        m_stat.m_get_consuming_ports++;
+        return {std::make_shared<PlaceMockPy>()};
+    }
+
+    bool is_input() const override
+    {
+        m_stat.m_is_input++;
+        return false;
+    }
+
+    bool is_output() const override
+    {
+        m_stat.m_is_output++;
+        return false;
+    }
+
+    bool is_equal(Ptr another) const override
+    {
+        m_stat.m_is_equal++;
+        m_stat.m_lastArgPlace = another;
+        return false;
+    }
+
+    bool is_equal_data(Ptr another) const override
+    {
+        m_stat.m_is_equal_data++;
+        m_stat.m_lastArgPlace = another;
+        return false;
+    }
+
+    Place::Ptr get_source_tensor(int inputPortIndex) const override
+    {
+        m_stat.m_get_source_tensor++;
+        m_stat.m_lastArgInt = inputPortIndex;
+        return {std::make_shared<PlaceMockPy>()};
+    }
+
+    Place::Ptr get_source_tensor() const override
+    {
+        m_stat.m_get_source_tensor++;
+        m_stat.m_lastArgInt = -1;
+        return {std::make_shared<PlaceMockPy>()};
+    }
+
+    //---------------Stat--------------------
+    PlaceStat get_stat() const { return m_stat; }
+};
+
+////////////////////////////////
+
+struct MOCK_API ModelStat
+{
+    int m_get_inputs = 0;
+    int m_get_outputs = 0;
+    int m_get_place_by_tensor_name = 0;
+    int m_get_place_by_operation_name = 0;
+    int m_get_place_by_operation_and_input_port = 0;
+    int m_get_place_by_operation_and_output_port = 0;
+    int m_set_name_for_tensor = 0;
+    int m_add_name_for_tensor = 0;
+    int m_set_name_for_operation = 0;
+    int m_free_name_for_tensor = 0;
+    int m_free_name_for_operation = 0;
+    int m_set_name_for_dimension = 0;
+    int m_cut_and_add_new_input = 0;
+    int m_cut_and_add_new_output = 0;
+    int m_add_output = 0;
+    int m_remove_output = 0;
+    int m_set_partial_shape = 0;
+    int m_get_partial_shape = 0;
+    int m_set_element_type = 0;
+
+    int m_extract_subgraph = 0;
+    int m_override_all_inputs = 0;
+    int m_override_all_outputs = 0;
+
+    // Arguments tracking
+    std::string m_lastArgString;
+    int m_lastArgInt;
+    Place::Ptr m_lastArgPlace = nullptr;
+    std::vector<Place::Ptr> m_lastArgInputPlaces;
+    std::vector<Place::Ptr> m_lastArgOutputPlaces;
+    ngraph::element::Type m_lastArgElementType;
+    ngraph::PartialShape m_lastArgPartialShape;
+
+    // Getters
+    int get_inputs() const { return m_get_inputs; }
+    int get_outputs() const { return m_get_outputs; }
+    int extract_subgraph() const { return m_extract_subgraph; }
+    int override_all_inputs() const { return m_override_all_inputs; }
+    int override_all_outputs() const { return m_override_all_outputs; }
+    int get_place_by_tensor_name() const { return m_get_place_by_tensor_name; }
+    int get_place_by_operation_name() const { return m_get_place_by_operation_name; }
+    int get_place_by_operation_and_input_port() const
+    {
+        return m_get_place_by_operation_and_input_port;
+    }
+    int get_place_by_operation_and_output_port() const
+    {
+        return m_get_place_by_operation_and_output_port;
+    }
+    int set_name_for_tensor() const { return m_set_name_for_tensor; }
+    int add_name_for_tensor() const { return m_add_name_for_tensor; }
+    int set_name_for_operation() const { return m_set_name_for_operation; }
+    int free_name_for_tensor() const { return m_free_name_for_tensor; }
+    int free_name_for_operation() const { return m_free_name_for_operation; }
+    int set_name_for_dimension() const { return m_set_name_for_dimension; }
+    int cut_and_add_new_input() const { return m_cut_and_add_new_input; }
+    int cut_and_add_new_output() const { return m_cut_and_add_new_output; }
+    int add_output() const { return m_add_output; }
+    int remove_output() const { return m_remove_output; }
+    int set_partial_shape() const { return m_set_partial_shape; }
+    int get_partial_shape() const { return m_get_partial_shape; }
+    int set_element_type() const { return m_set_element_type; }
+
+    // Arguments getters
+    std::string get_lastArgString() const { return m_lastArgString; }
+    int get_lastArgInt() const { return m_lastArgInt; }
+    Place::Ptr get_lastArgPlace() const { return m_lastArgPlace; }
+    std::vector<Place::Ptr> get_lastArgInputPlaces() const { return m_lastArgInputPlaces; }
+    std::vector<Place::Ptr> get_lastArgOutputPlaces() const { return m_lastArgOutputPlaces; }
+    ngraph::element::Type get_lastArgElementType() const { return m_lastArgElementType; }
+    ngraph::PartialShape get_lastArgPartialShape() const { return m_lastArgPartialShape; }
+};
+
+class MOCK_API InputModelMockPy : public InputModel
+{
+    mutable ModelStat m_stat;
+
+public:
+    std::vector<Place::Ptr> get_inputs() const override
+    {
+        m_stat.m_get_inputs++;
+        return {std::make_shared<PlaceMockPy>()};
+    }
+
+    std::vector<Place::Ptr> get_outputs() const override
+    {
+        m_stat.m_get_outputs++;
+        return {std::make_shared<PlaceMockPy>()};
+    }
+
+    Place::Ptr get_place_by_tensor_name(const std::string& tensorName) const override
+    {
+        m_stat.m_get_place_by_tensor_name++;
+        m_stat.m_lastArgString = tensorName;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_place_by_operation_name(const std::string& operationName) override
+    {
+        m_stat.m_get_place_by_operation_name++;
+        m_stat.m_lastArgString = operationName;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_place_by_operation_name_and_input_port(const std::string& operationName,
+                                                          int inputPortIndex) override
+    {
+        m_stat.m_get_place_by_operation_and_input_port++;
+        m_stat.m_lastArgInt = inputPortIndex;
+        m_stat.m_lastArgString = operationName;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    Place::Ptr get_place_by_operation_name_and_output_port(const std::string& operationName,
+                                                           int outputPortIndex) override
+    {
+        m_stat.m_get_place_by_operation_and_output_port++;
+        m_stat.m_lastArgInt = outputPortIndex;
+        m_stat.m_lastArgString = operationName;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    void set_name_for_tensor(Place::Ptr tensor, const std::string& newName) override
+    {
+        m_stat.m_set_name_for_tensor++;
+        m_stat.m_lastArgPlace = tensor;
+        m_stat.m_lastArgString = newName;
+    }
+
+    void add_name_for_tensor(Place::Ptr tensor, const std::string& newName) override
+    {
+        m_stat.m_add_name_for_tensor++;
+        m_stat.m_lastArgPlace = tensor;
+        m_stat.m_lastArgString = newName;
+    }
+
+    void set_name_for_operation(Place::Ptr operation, const std::string& newName) override
+    {
+        m_stat.m_set_name_for_operation++;
+        m_stat.m_lastArgPlace = operation;
+        m_stat.m_lastArgString = newName;
+    }
+
+    void free_name_for_tensor(const std::string& name) override
+    {
+        m_stat.m_free_name_for_tensor++;
+        m_stat.m_lastArgString = name;
+    }
+
+    void free_name_for_operation(const std::string& name) override
+    {
+        m_stat.m_free_name_for_operation++;
+        m_stat.m_lastArgString = name;
+    }
+
+    void set_name_for_dimension(Place::Ptr place,
+                                size_t shapeDimIndex,
+                                const std::string& dimName) override
+    {
+        m_stat.m_set_name_for_dimension++;
+        m_stat.m_lastArgPlace = place;
+        m_stat.m_lastArgInt = static_cast<int>(shapeDimIndex);
+        m_stat.m_lastArgString = dimName;
+    }
+
+    void cut_and_add_new_input(Place::Ptr place, const std::string& newNameOptional) override
+    {
+        m_stat.m_cut_and_add_new_input++;
+        m_stat.m_lastArgPlace = place;
+        m_stat.m_lastArgString = newNameOptional;
+    }
+
+    void cut_and_add_new_output(Place::Ptr place, const std::string& newNameOptional) override
+    {
+        m_stat.m_cut_and_add_new_output++;
+        m_stat.m_lastArgPlace = place;
+        m_stat.m_lastArgString = newNameOptional;
+    }
+
+    Place::Ptr add_output(Place::Ptr place) override
+    {
+        m_stat.m_add_output++;
+        m_stat.m_lastArgPlace = place;
+        return std::make_shared<PlaceMockPy>();
+    }
+
+    void remove_output(Place::Ptr place) override
+    {
+        m_stat.m_remove_output++;
+        m_stat.m_lastArgPlace = place;
+    }
+
+    void override_all_outputs(const std::vector<Place::Ptr>& outputs) override
+    {
+        m_stat.m_override_all_outputs++;
+        m_stat.m_lastArgOutputPlaces = outputs;
+    }
+
+    void override_all_inputs(const std::vector<Place::Ptr>& inputs) override
+    {
+        m_stat.m_override_all_inputs++;
+        m_stat.m_lastArgInputPlaces = inputs;
+    }
+
+    void extract_subgraph(const std::vector<Place::Ptr>& inputs,
+                          const std::vector<Place::Ptr>& outputs) override
+    {
+        m_stat.m_extract_subgraph++;
+        m_stat.m_lastArgInputPlaces = inputs;
+        m_stat.m_lastArgOutputPlaces = outputs;
+    }
+
+    // Setting tensor properties
+    void set_partial_shape(Place::Ptr place, const ngraph::PartialShape& shape) override
+    {
+        m_stat.m_set_partial_shape++;
+        m_stat.m_lastArgPlace = place;
+        m_stat.m_lastArgPartialShape = shape;
+    }
+
+    ngraph::PartialShape get_partial_shape(Place::Ptr place) const override
+    {
+        m_stat.m_get_partial_shape++;
+        m_stat.m_lastArgPlace = place;
+        return {};
+    }
+
+    void set_element_type(Place::Ptr place, const ngraph::element::Type& type) override
+    {
+        m_stat.m_set_element_type++;
+        m_stat.m_lastArgPlace = place;
+        m_stat.m_lastArgElementType = type;
+    }
+
+    //---------------Stat--------------------
+    ModelStat get_stat() const { return m_stat; }
+};
+
+/////////////////////////////////////////////////////////
+
+struct MOCK_API FeStat
+{
+    FrontEndCapFlags m_load_flags;
+    std::vector<std::string> m_load_paths;
+    int m_convert_model = 0;
+    int m_convert = 0;
+    int m_convert_partially = 0;
+    int m_decode = 0;
+    int m_normalize = 0;
+    // Getters
+    FrontEndCapFlags load_flags() const { return m_load_flags; }
+    std::vector<std::string> load_paths() const { return m_load_paths; }
+    int convert_model() const { return m_convert_model; }
+    int convert() const { return m_convert; }
+    int convert_partially() const { return m_convert_partially; }
+    int decode() const { return m_decode; }
+    int normalize() const { return m_normalize; }
+};
+
+class MOCK_API FrontEndMockPy : public FrontEnd
+{
+    mutable FeStat m_stat;
+
+public:
+    FrontEndMockPy(FrontEndCapFlags flags) { m_stat.m_load_flags = flags; }
+
+    InputModel::Ptr load_from_file(const std::string& path) const override
+    {
+        m_stat.m_load_paths.push_back(path);
+        return std::make_shared<InputModelMockPy>();
+    }
+
+    std::shared_ptr<ngraph::Function> convert(InputModel::Ptr model) const override
+    {
+        m_stat.m_convert_model++;
+        return std::make_shared<ngraph::Function>(NodeVector{}, ParameterVector{});
+    }
+
+    std::shared_ptr<ngraph::Function> convert(std::shared_ptr<ngraph::Function> func) const override
+    {
+        m_stat.m_convert++;
+        return func;
+    }
+
+    std::shared_ptr<ngraph::Function> convert_partially(InputModel::Ptr model) const override
+    {
+        m_stat.m_convert_partially++;
+        return std::make_shared<ngraph::Function>(NodeVector{}, ParameterVector{});
+    }
+
+    std::shared_ptr<ngraph::Function> decode(InputModel::Ptr model) const override
+    {
+        m_stat.m_decode++;
+        return std::make_shared<ngraph::Function>(NodeVector{}, ParameterVector{});
+    }
+
+    void normalize(std::shared_ptr<ngraph::Function> function) const override
+    {
+        m_stat.m_normalize++;
+    }
+
+    FeStat get_stat() const { return m_stat; }
+};
diff --git a/ngraph/python/tests/mock/pyngraph_fe_mock_api/CMakeLists.txt b/ngraph/python/tests/mock/pyngraph_fe_mock_api/CMakeLists.txt
new file mode 100644
index 00000000000000..7d2e4a3077acc0
--- /dev/null
+++ b/ngraph/python/tests/mock/pyngraph_fe_mock_api/CMakeLists.txt
@@ -0,0 +1,19 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+#
+
+set(TARGET_FE_NAME "mock_py_ngraph_frontend")
+set(PYBIND_FE_NAME "pybind_mock_frontend")
+
+set(PYBIND_FE_SRC ${CMAKE_CURRENT_SOURCE_DIR}/pyngraph_mock_frontend_api.cpp)
+
+source_group("src" FILES ${PYBIND_FE_SRC})
+
+pybind11_add_module(${PYBIND_FE_NAME} MODULE ${PYBIND_FE_SRC})
+
+target_link_libraries(${PYBIND_FE_NAME} PRIVATE ngraph::ngraph ngraph::frontend_manager)
+target_link_libraries(${PYBIND_FE_NAME} PRIVATE ${TARGET_FE_NAME})
+
+add_dependencies(${PYBIND_FE_NAME} ${TARGET_FE_NAME})
+
+add_clang_format_target(${PYBIND_FE_NAME}_clang FOR_TARGETS ${PYBIND_FE_NAME})
diff --git a/ngraph/python/tests/mock/pyngraph_fe_mock_api/pyngraph_mock_frontend_api.cpp b/ngraph/python/tests/mock/pyngraph_fe_mock_api/pyngraph_mock_frontend_api.cpp
new file mode 100644
index 00000000000000..ec87842d417330
--- /dev/null
+++ b/ngraph/python/tests/mock/pyngraph_fe_mock_api/pyngraph_mock_frontend_api.cpp
@@ -0,0 +1,136 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+
+#include "../mock_py_ngraph_frontend/mock_py_frontend.hpp"
+
+namespace py = pybind11;
+using namespace ngraph;
+using namespace ngraph::frontend;
+
+static void register_mock_frontend_stat(py::module m)
+{
+    m.def(
+        "get_fe_stat",
+        [](const std::shared_ptr<FrontEnd>& fe) {
+            std::shared_ptr<FrontEndMockPy> ptr = std::dynamic_pointer_cast<FrontEndMockPy>(fe);
+            if (ptr)
+            {
+                auto stat = ptr->get_stat();
+                return stat;
+            }
+            return FeStat();
+        },
+        py::arg("frontend"));
+
+    py::class_<FeStat> feStat(m, "FeStat", py::dynamic_attr());
+    feStat.def_property_readonly("load_flags", &FeStat::load_flags);
+    feStat.def_property_readonly("load_paths", &FeStat::load_paths);
+    feStat.def_property_readonly("convert_model", &FeStat::convert_model);
+    feStat.def_property_readonly("convert", &FeStat::convert);
+    feStat.def_property_readonly("convert_partially", &FeStat::convert_partially);
+    feStat.def_property_readonly("decode", &FeStat::decode);
+    feStat.def_property_readonly("normalize", &FeStat::normalize);
+}
+
+static void register_mock_model_stat(py::module m)
+{
+    m.def(
+        "get_mdl_stat",
+        [](const std::shared_ptr<InputModel>& mdl) {
+            std::shared_ptr<InputModelMockPy> ptr =
+                std::dynamic_pointer_cast<InputModelMockPy>(mdl);
+            if (ptr)
+            {
+                auto stat = ptr->get_stat();
+                return stat;
+            }
+            return ModelStat();
+        },
+        py::arg("model"));
+
+    py::class_<ModelStat> mdlStat(m, "ModelStat", py::dynamic_attr());
+    mdlStat.def_property_readonly("get_inputs", &ModelStat::get_inputs);
+    mdlStat.def_property_readonly("get_outputs", &ModelStat::get_outputs);
+    mdlStat.def_property_readonly("get_place_by_tensor_name", &ModelStat::get_place_by_tensor_name);
+    mdlStat.def_property_readonly("get_place_by_operation_name",
+                                  &ModelStat::get_place_by_operation_name);
+    mdlStat.def_property_readonly("get_place_by_operation_and_input_port",
+                                  &ModelStat::get_place_by_operation_and_input_port);
+    mdlStat.def_property_readonly("get_place_by_operation_and_output_port",
+                                  &ModelStat::get_place_by_operation_and_output_port);
+
+    mdlStat.def_property_readonly("set_name_for_tensor", &ModelStat::set_name_for_tensor);
+    mdlStat.def_property_readonly("add_name_for_tensor", &ModelStat::add_name_for_tensor);
+    mdlStat.def_property_readonly("set_name_for_operation", &ModelStat::set_name_for_operation);
+    mdlStat.def_property_readonly("free_name_for_tensor", &ModelStat::free_name_for_tensor);
+    mdlStat.def_property_readonly("free_name_for_operation", &ModelStat::free_name_for_operation);
+    mdlStat.def_property_readonly("set_name_for_dimension", &ModelStat::set_name_for_dimension);
+    mdlStat.def_property_readonly("cut_and_add_new_input", &ModelStat::cut_and_add_new_input);
+    mdlStat.def_property_readonly("cut_and_add_new_output", &ModelStat::cut_and_add_new_output);
+    mdlStat.def_property_readonly("add_output", &ModelStat::add_output);
+    mdlStat.def_property_readonly("remove_output", &ModelStat::remove_output);
+    mdlStat.def_property_readonly("set_partial_shape", &ModelStat::set_partial_shape);
+    mdlStat.def_property_readonly("get_partial_shape", &ModelStat::get_partial_shape);
+    mdlStat.def_property_readonly("set_element_type", &ModelStat::set_element_type);
+    mdlStat.def_property_readonly("extract_subgraph", &ModelStat::extract_subgraph);
+    mdlStat.def_property_readonly("override_all_inputs", &ModelStat::override_all_inputs);
+    mdlStat.def_property_readonly("override_all_outputs", &ModelStat::override_all_outputs);
+
+    // Arguments tracking
+    mdlStat.def_property_readonly("lastArgString", &ModelStat::get_lastArgString);
+    mdlStat.def_property_readonly("lastArgInt", &ModelStat::get_lastArgInt);
+    mdlStat.def_property_readonly("lastArgPlace", &ModelStat::get_lastArgPlace);
+    mdlStat.def_property_readonly("lastArgInputPlaces", &ModelStat::get_lastArgInputPlaces);
+    mdlStat.def_property_readonly("lastArgOutputPlaces", &ModelStat::get_lastArgOutputPlaces);
+    mdlStat.def_property_readonly("lastArgElementType", &ModelStat::get_lastArgElementType);
+    mdlStat.def_property_readonly("lastArgPartialShape", &ModelStat::get_lastArgPartialShape);
+}
+
+static void register_mock_place_stat(py::module m)
+{
+    m.def(
+        "get_place_stat",
+        [](const Place::Ptr& fe) {
+            std::shared_ptr<PlaceMockPy> ptr = std::dynamic_pointer_cast<PlaceMockPy>(fe);
+            if (ptr)
+            {
+                auto stat = ptr->get_stat();
+                return stat;
+            }
+            return PlaceStat();
+        },
+        py::arg("place"));
+
+    py::class_<PlaceStat> placeStat(m, "PlaceStat", py::dynamic_attr());
+
+    placeStat.def_property_readonly("lastArgString", &PlaceStat::get_lastArgString);
+    placeStat.def_property_readonly("lastArgInt", &PlaceStat::get_lastArgInt);
+    placeStat.def_property_readonly("lastArgPlace", &PlaceStat::get_lastArgPlace);
+
+    placeStat.def_property_readonly("get_names", &PlaceStat::get_names);
+    placeStat.def_property_readonly("get_consuming_operations",
+                                    &PlaceStat::get_consuming_operations);
+    placeStat.def_property_readonly("get_target_tensor", &PlaceStat::get_target_tensor);
+    placeStat.def_property_readonly("get_producing_operation", &PlaceStat::get_producing_operation);
+    placeStat.def_property_readonly("get_producing_port", &PlaceStat::get_producing_port);
+    placeStat.def_property_readonly("get_input_port", &PlaceStat::get_input_port);
+    placeStat.def_property_readonly("get_output_port", &PlaceStat::get_output_port);
+    placeStat.def_property_readonly("get_consuming_ports", &PlaceStat::get_consuming_ports);
+    placeStat.def_property_readonly("is_input", &PlaceStat::is_input);
+    placeStat.def_property_readonly("is_output", &PlaceStat::is_output);
+    placeStat.def_property_readonly("is_equal", &PlaceStat::is_equal);
+    placeStat.def_property_readonly("is_equal_data", &PlaceStat::is_equal_data);
+    placeStat.def_property_readonly("get_source_tensor", &PlaceStat::get_source_tensor);
+}
+
+PYBIND11_MODULE(pybind_mock_frontend, m)
+{
+    m.doc() = "Mock frontend call counters for testing Pyngraph frontend bindings";
+    register_mock_frontend_stat(m);
+    register_mock_model_stat(m);
+    register_mock_place_stat(m);
+}
diff --git a/ngraph/python/tests/test_ngraph/test_frontendmanager.py b/ngraph/python/tests/test_ngraph/test_frontendmanager.py
new file mode 100644
index 00000000000000..6c3d080657cc6f
--- /dev/null
+++ b/ngraph/python/tests/test_ngraph/test_frontendmanager.py
@@ -0,0 +1,545 @@
+# Copyright (C) 2018-2021 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import numpy as np
+import pytest
+
+from ngraph import PartialShape
+from ngraph.frontend import FrontEndCapabilities, FrontEndManager, InitializationFailure
+from ngraph.utils.types import get_element_type
+
+mock_available = True
+try:
+    from pybind_mock_frontend import get_fe_stat, get_mdl_stat, get_place_stat
+except Exception:
+    print("No mock frontend available")
+    mock_available = False
+
+# FrontEndManager shall be initialized and destroyed after all tests finished
+# This is because destroy of FrontEndManager will unload all plugins, no objects shall exist after this
+fem = FrontEndManager()
+
+mock_needed = pytest.mark.skipif(not mock_available,
+                                 reason="mock fe is not available")
+
+
+# ---------- FrontEnd tests ---------------
+@mock_needed
+def test_load_by_framework_caps():
+    frontEnds = fem.get_available_front_ends()
+    assert frontEnds is not None
+    assert "mock_py" in frontEnds
+    caps = [FrontEndCapabilities.DEFAULT,
+            FrontEndCapabilities.CUT,
+            FrontEndCapabilities.NAMES,
+            FrontEndCapabilities.WILDCARDS,
+            FrontEndCapabilities.CUT | FrontEndCapabilities.NAMES | FrontEndCapabilities.WILDCARDS]
+    for cap in caps:
+        fe = fem.load_by_framework(framework="mock_py", capabilities=cap)
+        stat = get_fe_stat(fe)
+        assert cap == stat.load_flags
+    for i in range(len(caps) - 1):
+        for j in range(i + 1, len(caps)):
+            assert caps[i] != caps[j]
+
+
+def test_load_by_unknown_framework():
+    frontEnds = fem.get_available_front_ends()
+    assert not("UnknownFramework" in frontEnds)
+    try:
+        fem.load_by_framework("UnknownFramework")
+    except InitializationFailure as exc:
+        print(exc)
+    else:
+        raise AssertionError("Unexpected exception.")
+
+
+@mock_needed
+def test_load_from_file():
+    fe = fem.load_by_framework(framework="mock_py")
+    assert fe is not None
+    model = fe.load_from_file("abc.bin")
+    assert model is not None
+    stat = get_fe_stat(fe)
+    assert "abc.bin" in stat.load_paths
+
+
+@mock_needed
+def test_convert_model():
+    fe = fem.load_by_framework(framework="mock_py")
+    assert fe is not None
+    model = fe.load_from_file(path="")
+    func = fe.convert(model=model)
+    assert func is not None
+    stat = get_fe_stat(fe)
+    assert stat.convert_model == 1
+
+
+@mock_needed
+def test_convert_partially():
+    fe = fem.load_by_framework(framework="mock_py")
+    assert fe is not None
+    model = fe.load_from_file(path="")
+    func = fe.convert_partially(model=model)
+    stat = get_fe_stat(fe)
+    assert stat.convert_partially == 1
+    fe.convert(function=func)
+    stat = get_fe_stat(fe)
+    assert stat.convert == 1
+
+
+@mock_needed
+def test_decode_and_normalize():
+    fe = fem.load_by_framework(framework="mock_py")
+    assert fe is not None
+    model = fe.load_from_file(path="")
+    func = fe.decode(model=model)
+    stat = get_fe_stat(fe)
+    assert stat.decode == 1
+    fe.normalize(function=func)
+    stat = get_fe_stat(fe)
+    assert stat.normalize == 1
+    assert stat.decode == 1
+
+
+# --------InputModel tests-----------------
+@mock_needed
+def init_model():
+    fe = fem.load_by_framework(framework="mock_py")
+    model = fe.load_from_file(path="")
+    return model
+
+
+@mock_needed
+def test_model_get_inputs():
+    model = init_model()
+    for i in range(1, 10):
+        model.get_inputs()
+        stat = get_mdl_stat(model)
+        assert stat.get_inputs == i
+
+
+@mock_needed
+def test_model_get_outputs():
+    model = init_model()
+    for i in range(1, 10):
+        model.get_outputs()
+        stat = get_mdl_stat(model)
+        assert stat.get_outputs == i
+
+
+@mock_needed
+def test_model_get_place_by_tensor_name():
+    model = init_model()
+    for i in range(1, 10):
+        name = str(i)
+        model.get_place_by_tensor_name(tensorName=name)
+        stat = get_mdl_stat(model)
+        assert stat.get_place_by_tensor_name == i
+        assert stat.lastArgString == name
+
+
+@mock_needed
+def test_model_get_place_by_operation_name():
+    model = init_model()
+    for i in range(1, 10):
+        name = str(i)
+        model.get_place_by_operation_name(operationName=name)
+        stat = get_mdl_stat(model)
+        assert stat.get_place_by_operation_name == i
+        assert stat.lastArgString == name
+
+
+@mock_needed
+def test_model_get_place_by_operation_name_and_input_port():
+    model = init_model()
+    for i in range(1, 10):
+        name = str(i)
+        model.get_place_by_operation_name_and_input_port(operationName=name, inputPortIndex=i * 2)
+        stat = get_mdl_stat(model)
+        assert stat.get_place_by_operation_and_input_port == i
+        assert stat.lastArgString == name
+        assert stat.lastArgInt == i * 2
+
+
+@mock_needed
+def test_model_get_place_by_operation_name_and_output_port():
+    model = init_model()
+    for i in range(1, 10):
+        name = str(i)
+        model.get_place_by_operation_name_and_output_port(operationName=name, outputPortIndex=i * 2)
+        stat = get_mdl_stat(model)
+        assert stat.get_place_by_operation_and_output_port == i
+        assert stat.lastArgString == name
+        assert stat.lastArgInt == i * 2
+
+
+@mock_needed
+def test_model_set_name_for_tensor():
+    model = init_model()
+    place = model.get_place_by_tensor_name(tensorName="")
+    model.set_name_for_tensor(tensor=place, newName="1234")
+    stat = get_mdl_stat(model)
+    assert stat.set_name_for_tensor == 1
+    assert stat.lastArgString == "1234"
+    assert stat.lastArgPlace == place
+
+
+@mock_needed
+def test_model_add_name_for_tensor():
+    model = init_model()
+    place = model.get_place_by_tensor_name(tensorName="")
+    model.add_name_for_tensor(tensor=place, newName="1234")
+    stat = get_mdl_stat(model)
+    assert stat.add_name_for_tensor == 1
+    assert stat.lastArgString == "1234"
+    assert stat.lastArgPlace == place
+
+
+@mock_needed
+def test_model_set_name_for_operation():
+    model = init_model()
+    place = model.get_place_by_operation_name(operationName="")
+    model.set_name_for_operation(operation=place, newName="1111")
+    stat = get_mdl_stat(model)
+    assert stat.set_name_for_operation == 1
+    assert stat.lastArgString == "1111"
+    assert stat.lastArgPlace == place
+
+
+@mock_needed
+def test_model_free_name_for_tensor():
+    model = init_model()
+    model.free_name_for_tensor(name="2222")
+    stat = get_mdl_stat(model)
+    assert stat.free_name_for_tensor == 1
+    assert stat.lastArgString == "2222"
+
+
+@mock_needed
+def test_model_free_name_for_operation():
+    model = init_model()
+    model.free_name_for_operation(name="3333")
+    stat = get_mdl_stat(model)
+    assert stat.free_name_for_operation == 1
+    assert stat.lastArgString == "3333"
+
+
+@mock_needed
+def test_model_set_name_for_dimension():
+    model = init_model()
+    place = model.get_place_by_operation_name(operationName="")
+    model.set_name_for_dimension(place=place, dimIndex=123, dimName="4444")
+    stat = get_mdl_stat(model)
+    assert stat.set_name_for_dimension == 1
+    assert stat.lastArgString == "4444"
+    assert stat.lastArgInt == 123
+    assert stat.lastArgPlace == place
+
+
+@mock_needed
+def test_model_cut_and_add_new_input():
+    model = init_model()
+    place = model.get_place_by_operation_name("")
+    model.cut_and_add_new_input(place=place, newName="5555")
+    stat = get_mdl_stat(model)
+    assert stat.cut_and_add_new_input == 1
+    assert stat.lastArgString == "5555"
+    assert stat.lastArgPlace == place
+    model.cut_and_add_new_input(place=place)
+    stat = get_mdl_stat(model)
+    assert stat.cut_and_add_new_input == 2
+    assert stat.lastArgString == ""
+    assert stat.lastArgPlace == place
+
+
+@mock_needed
+def test_model_cut_and_add_new_output():
+    model = init_model()
+    place = model.get_place_by_operation_name("")
+    model.cut_and_add_new_output(place=place, newName="5555")
+    stat = get_mdl_stat(model)
+    assert stat.cut_and_add_new_output == 1
+    assert stat.lastArgString == "5555"
+    assert stat.lastArgPlace == place
+    model.cut_and_add_new_output(place=place)
+    stat = get_mdl_stat(model)
+    assert stat.cut_and_add_new_output == 2
+    assert stat.lastArgString == ""
+    assert stat.lastArgPlace == place
+
+
+@mock_needed
+def test_model_add_output():
+    model = init_model()
+    place = model.get_place_by_operation_name("")
+    place2 = model.add_output(place=place)
+    assert place2 is not None
+    stat = get_mdl_stat(model)
+    assert stat.add_output == 1
+    assert stat.lastArgPlace == place
+
+
+@mock_needed
+def test_model_remove_output():
+    model = init_model()
+    place = model.get_place_by_operation_name("")
+    model.remove_output(place=place)
+    stat = get_mdl_stat(model)
+    assert stat.remove_output == 1
+    assert stat.lastArgPlace == place
+
+
+@mock_needed
+def test_model_set_partial_shape():
+    model = init_model()
+    place = model.get_place_by_tensor_name(tensorName="")
+    test_shape = PartialShape([1, 2, 3, 4])
+    model.set_partial_shape(place=place, shape=test_shape)
+    stat = get_mdl_stat(model)
+    assert stat.set_partial_shape == 1
+    assert stat.lastArgPlace == place
+    assert stat.lastArgPartialShape == test_shape
+
+
+@mock_needed
+def test_model_get_partial_shape():
+    model = init_model()
+    place = model.get_place_by_tensor_name(tensorName="")
+    shape = model.get_partial_shape(place=place)
+    assert shape is not None
+    stat = get_mdl_stat(model)
+    assert stat.get_partial_shape == 1
+    assert stat.lastArgPlace == place
+
+
+@mock_needed
+def test_model_override_all_inputs():
+    model = init_model()
+    place1 = model.get_place_by_tensor_name(tensorName="p1")
+    place2 = model.get_place_by_tensor_name(tensorName="p2")
+    model.override_all_inputs(inputs=[place1, place2])
+    stat = get_mdl_stat(model)
+    assert stat.override_all_inputs == 1
+    assert len(stat.lastArgInputPlaces) == 2
+    assert stat.lastArgInputPlaces[0] == place1
+    assert stat.lastArgInputPlaces[1] == place2
+
+
+@mock_needed
+def test_model_override_all_outputs():
+    model = init_model()
+    place1 = model.get_place_by_tensor_name(tensorName="p1")
+    place2 = model.get_place_by_tensor_name(tensorName="p2")
+    model.override_all_outputs(outputs=[place1, place2])
+    stat = get_mdl_stat(model)
+    assert stat.override_all_outputs == 1
+    assert len(stat.lastArgOutputPlaces) == 2
+    assert stat.lastArgOutputPlaces[0] == place1
+    assert stat.lastArgOutputPlaces[1] == place2
+
+
+@mock_needed
+def test_model_extract_subgraph():
+    model = init_model()
+    place1 = model.get_place_by_tensor_name(tensorName="p1")
+    place2 = model.get_place_by_tensor_name(tensorName="p2")
+    place3 = model.get_place_by_tensor_name(tensorName="p3")
+    place4 = model.get_place_by_tensor_name(tensorName="p4")
+    model.extract_subgraph(inputs=[place1, place2], outputs=[place3, place4])
+    stat = get_mdl_stat(model)
+    assert stat.extract_subgraph == 1
+    assert len(stat.lastArgInputPlaces) == 2
+    assert stat.lastArgInputPlaces[0] == place1
+    assert stat.lastArgInputPlaces[1] == place2
+    assert len(stat.lastArgOutputPlaces) == 2
+    assert stat.lastArgOutputPlaces[0] == place3
+    assert stat.lastArgOutputPlaces[1] == place4
+
+
+@mock_needed
+def test_model_set_element_type():
+    model = init_model()
+    place = model.get_place_by_tensor_name(tensorName="")
+    model.set_element_type(place=place, type=get_element_type(np.int32))
+    stat = get_mdl_stat(model)
+    assert stat.set_element_type == 1
+    assert stat.lastArgPlace == place
+    assert stat.lastArgElementType == get_element_type(np.int32)
+
+
+# ----------- Place test ------------
+@mock_needed
+def init_place():
+    fe = fem.load_by_framework(framework="mock_py")
+    model = fe.load_from_file(path="")
+    place = model.get_place_by_tensor_name(tensorName="")
+    return model, place
+
+
+@mock_needed
+def test_place_is_input():
+    _, place = init_place()
+    assert place.is_input() is not None
+    stat = get_place_stat(place)
+    assert stat.is_input == 1
+
+
+@mock_needed
+def test_place_is_output():
+    _, place = init_place()
+    assert place.is_output() is not None
+    stat = get_place_stat(place)
+    assert stat.is_output == 1
+
+
+@mock_needed
+def test_place_get_names():
+    _, place = init_place()
+    assert place.get_names() is not None
+    stat = get_place_stat(place)
+    assert stat.get_names == 1
+
+
+@mock_needed
+def test_place_is_equal():
+    model, place = init_place()
+    place2 = model.get_place_by_tensor_name("2")
+    assert place.is_equal(other=place2) is not None
+    stat = get_place_stat(place)
+    assert stat.is_equal == 1
+    assert stat.lastArgPlace == place2
+
+
+@mock_needed
+def test_place_is_equal_data():
+    model, place = init_place()
+    place2 = model.get_place_by_tensor_name("2")
+    assert place.is_equal_data(other=place2) is not None
+    stat = get_place_stat(place)
+    assert stat.is_equal_data == 1
+    assert stat.lastArgPlace == place2
+
+
+@mock_needed
+def test_place_get_consuming_operations():
+    _, place = init_place()
+    assert place.get_consuming_operations(outputPortIndex=22) is not None
+    stat = get_place_stat(place)
+    assert stat.get_consuming_operations == 1
+    assert stat.lastArgInt == 22
+    assert place.get_consuming_operations() is not None
+    stat = get_place_stat(place)
+    assert stat.get_consuming_operations == 2
+    assert stat.lastArgInt == -1
+
+
+@mock_needed
+def test_place_get_target_tensor():
+    _, place = init_place()
+    assert place.get_target_tensor(outputPortIndex=22) is not None
+    stat = get_place_stat(place)
+    assert stat.get_target_tensor == 1
+    assert stat.lastArgInt == 22
+    assert place.get_target_tensor() is not None
+    stat = get_place_stat(place)
+    assert stat.get_target_tensor == 2
+    assert stat.lastArgInt == -1
+
+
+@mock_needed
+def test_place_get_producing_operation():
+    _, place = init_place()
+    assert place.get_producing_operation(inputPortIndex=22) is not None
+    stat = get_place_stat(place)
+    assert stat.get_producing_operation == 1
+    assert stat.lastArgInt == 22
+    assert place.get_producing_operation() is not None
+    stat = get_place_stat(place)
+    assert stat.get_producing_operation == 2
+    assert stat.lastArgInt == -1
+
+
+@mock_needed
+def test_place_get_producing_port():
+    _, place = init_place()
+    assert place.get_producing_port() is not None
+    stat = get_place_stat(place)
+    assert stat.get_producing_port == 1
+
+
+@mock_needed
+def test_place_get_input_port():
+    _, place = init_place()
+    assert place.get_input_port() is not None
+    stat = get_place_stat(place)
+    assert stat.get_input_port == 1
+    assert stat.lastArgInt == -1
+    assert place.get_input_port(inputPortIndex=22) is not None
+    stat = get_place_stat(place)
+    assert stat.get_input_port == 2
+    assert stat.lastArgInt == 22
+
+
+@mock_needed
+def test_place_get_input_port2():
+    _, place = init_place()
+    assert place.get_input_port(inputName="abc") is not None
+    stat = get_place_stat(place)
+    assert stat.get_input_port == 1
+    assert stat.lastArgInt == -1
+    assert stat.lastArgString == "abc"
+    assert place.get_input_port(inputName="abcd", inputPortIndex=22) is not None
+    stat = get_place_stat(place)
+    assert stat.get_input_port == 2
+    assert stat.lastArgInt == 22
+    assert stat.lastArgString == "abcd"
+
+
+@mock_needed
+def test_place_get_output_port():
+    _, place = init_place()
+    assert place.get_output_port() is not None
+    stat = get_place_stat(place)
+    assert stat.get_output_port == 1
+    assert stat.lastArgInt == -1
+    assert place.get_output_port(outputPortIndex=22) is not None
+    stat = get_place_stat(place)
+    assert stat.get_output_port == 2
+    assert stat.lastArgInt == 22
+
+
+@mock_needed
+def test_place_get_output_port2():
+    _, place = init_place()
+    assert place.get_output_port(outputName="abc") is not None
+    stat = get_place_stat(place)
+    assert stat.get_output_port == 1
+    assert stat.lastArgInt == -1
+    assert stat.lastArgString == "abc"
+    assert place.get_output_port(outputName="abcd", outputPortIndex=22) is not None
+    stat = get_place_stat(place)
+    assert stat.get_output_port == 2
+    assert stat.lastArgInt == 22
+    assert stat.lastArgString == "abcd"
+
+
+@mock_needed
+def test_place_get_consuming_ports():
+    _, place = init_place()
+    assert place.get_consuming_ports() is not None
+    stat = get_place_stat(place)
+    assert stat.get_consuming_ports == 1
+
+
+@mock_needed
+def test_place_get_source_tensor():
+    _, place = init_place()
+    assert place.get_source_tensor() is not None
+    stat = get_place_stat(place)
+    assert stat.get_source_tensor == 1
+    assert stat.lastArgInt == -1
+    assert place.get_source_tensor(inputPortIndex=22) is not None
+    stat = get_place_stat(place)
+    assert stat.get_source_tensor == 2
+    assert stat.lastArgInt == 22

From 77912ca06e50da435b1793e00b014ab14ad63459 Mon Sep 17 00:00:00 2001
From: Piotr Szmelczynski <piotr.szmelczynski@intel.com>
Date: Tue, 22 Jun 2021 08:44:31 +0200
Subject: [PATCH 153/178] Revise negative (#5955)

* update spec

* add RTTI macro

* create visitor test

* clean backend test file

* add negative to type_prop tests of unary_ops

* add validate_and_infer_types method

* remove negative from type_propp unary_ops

* create negative type_prop tests

* remove boolean check from validate_and_infer_types

* update supported types in spec

* Remove check for signed numeric types

* Remove type_prop test cases with non-signed type

* Add zero sized negative test to interpreter manifest

* Revert "update supported types in spec"

This reverts commit a7ac3e4dc85e6d0408858c25d651e5699b6c5f5d.

* Revert "update spec"

This reverts commit 4c431d753df465e5b8a7bdc1925ed316f6c56a82.

* Add minor changes to review op class

 * Add validation of inputs and outputs in evaluate
 * Use shape from host tensor to calculate element count

* Minor changes in variable names of backend tests

* Add SLT case with integer precision

* Add operation to list of trusted ops

* Address review comments:

 * Add bf16 precision to evaluate method
 * Use unary ops typed case tests for type_prop tests
 * Add end of line to interpreter manifest

* Refactored visitor unit test

* Fix typo in visitor test name

Co-authored-by: ggalieroc <gabriele.galiero.casay@intel.com>
---
 .../single_layer_tests/activation.cpp          |  1 +
 .../layer_tests_summary/utils/constants.py     |  1 +
 ngraph/core/include/ngraph/op/negative.hpp     |  3 +--
 ngraph/core/src/op/negative.cpp                | 15 +++++++--------
 ngraph/test/CMakeLists.txt                     |  2 ++
 ngraph/test/backend/negative.in.cpp            | 18 ++++++------------
 .../runtime/interpreter/unit_test.manifest     |  3 +++
 ngraph/test/type_prop/negative.cpp             |  9 +++++++++
 ngraph/test/visitors/op/ceiling.cpp            |  4 ++--
 ngraph/test/visitors/op/floor.cpp              |  4 ++--
 ngraph/test/visitors/op/log.cpp                |  4 ++--
 ngraph/test/visitors/op/negative.cpp           | 13 +++++++++++++
 ngraph/test/visitors/op/result.cpp             |  4 ++--
 ngraph/test/visitors/op/sqrt.cpp               |  4 ++--
 ngraph/test/visitors/op/squeeze.cpp            |  4 ++--
 15 files changed, 55 insertions(+), 34 deletions(-)
 create mode 100644 ngraph/test/type_prop/negative.cpp
 create mode 100644 ngraph/test/visitors/op/negative.cpp

diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
index 510e4039e2bfd0..51e379a67f128b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
@@ -64,6 +64,7 @@ const std::map<ActivationTypes, std::vector<std::vector<float>>> activationTypes
 
 // List of operations that should be tested also with integer precision
 const std::map<ActivationTypes, std::vector<std::vector<float>>> intActivationTypes = {
+        {Negative,              {}},
         {Ceiling,               {}},
         {Sqrt,                  {}},
         {Tanh,                  {}},
diff --git a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
index 7db26eb289542f..24e42cd5e04d81 100644
--- a/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
+++ b/inference-engine/tests/ie_test_utils/functional_test_utils/layer_tests_summary/utils/constants.py
@@ -54,6 +54,7 @@
     'MaxPool-1',
     'Mish-4',
     'Multiply-1',
+    'Negative-1',
     'NonMaxSuppression-4',
     'NonMaxSuppression-5',
     'PSROIPooling-1',
diff --git a/ngraph/core/include/ngraph/op/negative.hpp b/ngraph/core/include/ngraph/op/negative.hpp
index 69c7d39bdc0da4..1768a58c01012b 100644
--- a/ngraph/core/include/ngraph/op/negative.hpp
+++ b/ngraph/core/include/ngraph/op/negative.hpp
@@ -16,8 +16,7 @@ namespace ngraph
             class NGRAPH_API Negative : public util::UnaryElementwiseArithmetic
             {
             public:
-                static constexpr NodeTypeInfo type_info{"Negative", 0};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
+                NGRAPH_RTTI_DECLARATION;
                 /// \brief Constructs a negative operation.
                 Negative() = default;
                 /// \brief Constructs a negative operation.
diff --git a/ngraph/core/src/op/negative.cpp b/ngraph/core/src/op/negative.cpp
index f670ac19a2e125..4ed3c44e53b508 100644
--- a/ngraph/core/src/op/negative.cpp
+++ b/ngraph/core/src/op/negative.cpp
@@ -6,11 +6,12 @@
 #include "itt.hpp"
 #include "ngraph/runtime/host_tensor.hpp"
 #include "ngraph/runtime/reference/negate.hpp"
+#include "ngraph/validation_util.hpp"
 
 using namespace std;
 using namespace ngraph;
 
-constexpr NodeTypeInfo op::Negative::type_info;
+NGRAPH_RTTI_DEFINITION(op::v0::Negative, "Negative", 0, util::UnaryElementwiseArithmetic);
 
 op::Negative::Negative(const Output<Node>& arg)
     : UnaryElementwiseArithmetic(arg)
@@ -48,11 +49,9 @@ namespace negativeop
 
         switch (arg0->get_element_type())
         {
-            NGRAPH_TYPE_CASE(evaluate_negative, boolean, arg0, out, count);
             NGRAPH_TYPE_CASE(evaluate_negative, i32, arg0, out, count);
             NGRAPH_TYPE_CASE(evaluate_negative, i64, arg0, out, count);
-            NGRAPH_TYPE_CASE(evaluate_negative, u32, arg0, out, count);
-            NGRAPH_TYPE_CASE(evaluate_negative, u64, arg0, out, count);
+            NGRAPH_TYPE_CASE(evaluate_negative, bf16, arg0, out, count);
             NGRAPH_TYPE_CASE(evaluate_negative, f16, arg0, out, count);
             NGRAPH_TYPE_CASE(evaluate_negative, f32, arg0, out, count);
         default: rc = false; break;
@@ -64,7 +63,10 @@ namespace negativeop
 bool op::Negative::evaluate(const HostTensorVector& outputs, const HostTensorVector& inputs) const
 {
     NGRAPH_OP_SCOPE(v0_Negative_evaluate);
-    return negativeop::evaluate_negative(inputs[0], outputs[0], shape_size(get_output_shape(0)));
+    NGRAPH_CHECK(validate_host_tensor_vector(inputs, 1));
+    NGRAPH_CHECK(validate_host_tensor_vector(outputs, 1));
+    return negativeop::evaluate_negative(
+        inputs[0], outputs[0], shape_size(outputs[0]->get_shape()));
 }
 
 bool op::Negative::has_evaluate() const
@@ -72,11 +74,8 @@ bool op::Negative::has_evaluate() const
     NGRAPH_OP_SCOPE(v0_Negative_has_evaluate);
     switch (get_input_element_type(0))
     {
-    case ngraph::element::boolean:
     case ngraph::element::i32:
     case ngraph::element::i64:
-    case ngraph::element::u32:
-    case ngraph::element::u64:
     case ngraph::element::f16:
     case ngraph::element::f32: return true;
     default: break;
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 278121669a83c1..2b7757f0527b81 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -167,6 +167,7 @@ set(SRC
     type_prop/mish.cpp
     type_prop/mod.cpp
     type_prop/mvn.cpp
+    type_prop/negative.cpp
     type_prop/non_max_suppression.cpp
     type_prop/non_zero.cpp
     type_prop/normalize.cpp
@@ -257,6 +258,7 @@ set(SRC
     visitors/op/max_pool.cpp
     visitors/op/mod.cpp
     visitors/op/mvn.cpp
+    visitors/op/negative.cpp
     visitors/op/non_max_suppression.cpp
     visitors/op/normalize_l2.cpp
     visitors/op/one_hot.cpp
diff --git a/ngraph/test/backend/negative.in.cpp b/ngraph/test/backend/negative.in.cpp
index a092fe6f7607a4..01aca08610dce9 100644
--- a/ngraph/test/backend/negative.in.cpp
+++ b/ngraph/test/backend/negative.in.cpp
@@ -2,13 +2,6 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include <algorithm>
-#include <cinttypes>
-#include <cmath>
-#include <cstdlib>
-#include <random>
-#include <string>
-
 // clang-format off
 #ifdef ${BACKEND_NAME}_FLOAT_TOLERANCE_BITS
 #define DEFAULT_FLOAT_TOLERANCE_BITS ${BACKEND_NAME}_FLOAT_TOLERANCE_BITS
@@ -49,15 +42,16 @@ NGRAPH_TEST(${BACKEND_NAME}, negative_i32)
 {
     auto shape_a = Shape{2, 5};
     auto A = make_shared<op::Parameter>(element::i32, shape_a);
-    auto relu = make_shared<op::Negative>(A);
+    auto negative = make_shared<op::Negative>(A);
     auto shape_rt = Shape{2, 5};
-    auto f = make_shared<Function>(relu, ParameterVector{A});
+    auto f = make_shared<Function>(negative, ParameterVector{A});
 
     std::vector<int32_t> a{1, 8, -8, 17, -2, 1, 8, -8, 17, -1};
+    std::vector<int32_t> r{-1, -8, 8, -17, 2, -1, -8, 8, -17, 1};
 
     auto test_case = test::TestCase<TestEngine>(f);
     test_case.add_input<int32_t>(shape_a, a);
-    test_case.add_expected_output<int32_t>(shape_rt, {-1, -8, 8, -17, 2, -1, -8, 8, -17, 1});
+    test_case.add_expected_output<int32_t>(shape_rt, r);
     test_case.run();
 }
 
@@ -65,9 +59,9 @@ NGRAPH_TEST(${BACKEND_NAME}, negative_f32)
 {
     auto shape_a = Shape{2, 5};
     auto A = make_shared<op::Parameter>(element::f32, shape_a);
-    auto relu = make_shared<op::Negative>(A);
+    auto negative = make_shared<op::Negative>(A);
     auto shape_rt = Shape{2, 5};
-    auto f = make_shared<Function>(relu, ParameterVector{A});
+    auto f = make_shared<Function>(negative, ParameterVector{A});
 
     std::vector<float> a{1.35f, 8.76f, -8.0f, 17.234f, -2.121f, 1.0f, 8.7f, -8.92f, 17.0f, -1.0f};
 
diff --git a/ngraph/test/runtime/interpreter/unit_test.manifest b/ngraph/test/runtime/interpreter/unit_test.manifest
index 799025f3b92752..a76df269cddff2 100644
--- a/ngraph/test/runtime/interpreter/unit_test.manifest
+++ b/ngraph/test/runtime/interpreter/unit_test.manifest
@@ -158,3 +158,6 @@ INTERPRETER.onnx_model_experimental_detectron_roi_feature_extractor
 
 # No evaluator for DeformableConv2D
 onnx_model_deformable_conv_2d
+
+# No support for unsigned types
+INTERPRETER.zero_sized_negative
diff --git a/ngraph/test/type_prop/negative.cpp b/ngraph/test/type_prop/negative.cpp
new file mode 100644
index 00000000000000..68313c2f5442b7
--- /dev/null
+++ b/ngraph/test/type_prop/negative.cpp
@@ -0,0 +1,9 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Type = ::testing::Types<ngraph::op::Negative>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(type_prop_negative, UnaryOperator, Type);
diff --git a/ngraph/test/visitors/op/ceiling.cpp b/ngraph/test/visitors/op/ceiling.cpp
index 11e71e1e9c604e..327fb4ac4b3569 100644
--- a/ngraph/test/visitors/op/ceiling.cpp
+++ b/ngraph/test/visitors/op/ceiling.cpp
@@ -5,7 +5,7 @@
 #include "unary_ops.hpp"
 using Type = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Ceiling, element::f32>>;
 
-INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_attribute,
                               UnaryOperatorVisitor,
                               Type,
-                              UnaryOperatorTypeName);
\ No newline at end of file
+                              UnaryOperatorTypeName);
diff --git a/ngraph/test/visitors/op/floor.cpp b/ngraph/test/visitors/op/floor.cpp
index 98731229010b1b..be2fb04372b7cf 100644
--- a/ngraph/test/visitors/op/floor.cpp
+++ b/ngraph/test/visitors/op/floor.cpp
@@ -7,7 +7,7 @@
 using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Floor, element::f32>,
                                UnaryOperatorType<ngraph::op::v0::Floor, element::f16>>;
 
-INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_attribute,
                               UnaryOperatorVisitor,
                               Types,
-                              UnaryOperatorTypeName);
\ No newline at end of file
+                              UnaryOperatorTypeName);
diff --git a/ngraph/test/visitors/op/log.cpp b/ngraph/test/visitors/op/log.cpp
index 5acb4971ba81f2..84e87eb6db3edc 100644
--- a/ngraph/test/visitors/op/log.cpp
+++ b/ngraph/test/visitors/op/log.cpp
@@ -5,7 +5,7 @@
 using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Log, element::f32>,
                                UnaryOperatorType<ngraph::op::v0::Log, element::f16>>;
 
-INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_attribute,
                               UnaryOperatorVisitor,
                               Types,
-                              UnaryOperatorTypeName);
\ No newline at end of file
+                              UnaryOperatorTypeName);
diff --git a/ngraph/test/visitors/op/negative.cpp b/ngraph/test/visitors/op/negative.cpp
new file mode 100644
index 00000000000000..769a629fad66ca
--- /dev/null
+++ b/ngraph/test/visitors/op/negative.cpp
@@ -0,0 +1,13 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Negative, element::f32>,
+                               UnaryOperatorType<ngraph::op::v0::Negative, element::i32>>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_attribute,
+                              UnaryOperatorVisitor,
+                              Types,
+                              UnaryOperatorTypeName);
diff --git a/ngraph/test/visitors/op/result.cpp b/ngraph/test/visitors/op/result.cpp
index df01ad5f913371..27e0e45e891b0e 100644
--- a/ngraph/test/visitors/op/result.cpp
+++ b/ngraph/test/visitors/op/result.cpp
@@ -7,7 +7,7 @@
 using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Result, element::f32>,
                                UnaryOperatorType<ngraph::op::v0::Result, element::f16>>;
 
-INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_attribute,
                               UnaryOperatorVisitor,
                               Types,
-                              UnaryOperatorTypeName);
\ No newline at end of file
+                              UnaryOperatorTypeName);
diff --git a/ngraph/test/visitors/op/sqrt.cpp b/ngraph/test/visitors/op/sqrt.cpp
index 5e26c542348dbf..1df6ae6420a97d 100644
--- a/ngraph/test/visitors/op/sqrt.cpp
+++ b/ngraph/test/visitors/op/sqrt.cpp
@@ -6,7 +6,7 @@
 using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Sqrt, element::f32>,
                                UnaryOperatorType<ngraph::op::v0::Sqrt, element::f16>>;
 
-INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_attribute,
                               UnaryOperatorVisitor,
                               Types,
-                              UnaryOperatorTypeName);
\ No newline at end of file
+                              UnaryOperatorTypeName);
diff --git a/ngraph/test/visitors/op/squeeze.cpp b/ngraph/test/visitors/op/squeeze.cpp
index a8b0c2c488c1ab..9c6813af16fece 100644
--- a/ngraph/test/visitors/op/squeeze.cpp
+++ b/ngraph/test/visitors/op/squeeze.cpp
@@ -6,7 +6,7 @@
 using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v0::Squeeze, element::f32>,
                               UnaryOperatorType<ngraph::op::v0::Squeeze, element::f16>>;
 
-INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_attribute,
                               UnaryOperatorVisitor,
                               Types,
-                              UnaryOperatorTypeName);
\ No newline at end of file
+                              UnaryOperatorTypeName);

From 5ce5f9e0c86944ca9acbcc08f60a048757835ee3 Mon Sep 17 00:00:00 2001
From: Duncan Ding <duncan.ding@intel.com>
Date: Tue, 22 Jun 2021 15:04:11 +0800
Subject: [PATCH 154/178] Revise SoftPlus operation reference implementation
 37559 (#5938)

* change threshold to 20 instead of max limit of data type

* add invalid input type test case

* add invalid input data check

* add input type dynamic check

* add backend test case

* add more clarity on spec and align with real implementation

* Revert "[CPU] Fix for  CoreThreadingTestsWithIterations tests (#5892)"

This reverts commit 115aa143ef8d8a0a17aed705db2f6d6e1dcb4164.

* Revert "Revert "[CPU] Fix for  CoreThreadingTestsWithIterations tests (#5892)""

This reverts commit 95afa50d94b7164968753194557888ecf16a9e69.

* Revert "change threshold to 20 instead of max limit of data type"

This reverts commit 91af8250567f843161f920be2b44f219a127efb4.

* Revert "add more clarity on spec and align with real implementation"

This reverts commit a3b232a8fbfe69cb35f96a011588102a79fb12b0.

* add visitor attribute test case

* Revert "add visitor attribute test case"

This reverts commit 610728f1abbe89178a514fecf2e02443e6739c74.

* add attribute test case

* revise the attribute visitor test per parametrized visitor API PR: #6181
---
 ngraph/core/src/op/softplus.cpp      |  7 +++++
 ngraph/test/CMakeLists.txt           |  2 ++
 ngraph/test/backend/softplus.in.cpp  | 45 ++++++++++++++++++++++++++++
 ngraph/test/type_prop/softplus.cpp   | 20 +++++++++++++
 ngraph/test/visitors/op/softplus.cpp | 12 ++++++++
 5 files changed, 86 insertions(+)
 create mode 100644 ngraph/test/backend/softplus.in.cpp
 create mode 100644 ngraph/test/visitors/op/softplus.cpp

diff --git a/ngraph/core/src/op/softplus.cpp b/ngraph/core/src/op/softplus.cpp
index f37008220a4fe0..23f356ae74f232 100644
--- a/ngraph/core/src/op/softplus.cpp
+++ b/ngraph/core/src/op/softplus.cpp
@@ -29,6 +29,13 @@ bool op::v4::SoftPlus::visit_attributes(AttributeVisitor& visitor)
 void op::v4::SoftPlus::validate_and_infer_types()
 {
     NGRAPH_OP_SCOPE(v4_SoftPlus_validate_and_infer_types);
+    const element::Type& input_et = get_input_element_type(0);
+
+    NODE_VALIDATION_CHECK(this,
+                          input_et.is_dynamic() || input_et.is_real(),
+                          "Input element type must be float. Got: ",
+                          input_et);
+
     set_output_size(1);
     set_output_type(0, get_input_element_type(0), get_input_partial_shape(0));
 }
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 2b7757f0527b81..6f4e2bb0db47a2 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -287,6 +287,7 @@ set(SRC
     visitors/op/selu.cpp
     visitors/op/shuffle_channels.cpp
     visitors/op/softmax.cpp
+    visitors/op/softplus.cpp
     visitors/op/space_to_batch.cpp
     visitors/op/space_to_depth.cpp
     visitors/op/split.cpp
@@ -463,6 +464,7 @@ set(MULTI_TEST_SRC
     backend/sin.in.cpp
     backend/sinh.in.cpp
     backend/softmax.in.cpp
+    backend/softplus.in.cpp
     backend/space_to_batch.in.cpp
     backend/split.in.cpp
     backend/sqrt.in.cpp
diff --git a/ngraph/test/backend/softplus.in.cpp b/ngraph/test/backend/softplus.in.cpp
new file mode 100644
index 00000000000000..c1e45a10c23422
--- /dev/null
+++ b/ngraph/test/backend/softplus.in.cpp
@@ -0,0 +1,45 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <algorithm>
+#include <cinttypes>
+#include <cmath>
+#include <cstdlib>
+#include <random>
+#include <string>
+
+// clang-format off
+#ifdef ${BACKEND_NAME}_FLOAT_TOLERANCE_BITS
+#define DEFAULT_FLOAT_TOLERANCE_BITS ${BACKEND_NAME}_FLOAT_TOLERANCE_BITS
+#endif
+
+#ifdef ${BACKEND_NAME}_DOUBLE_TOLERANCE_BITS
+#define DEFAULT_DOUBLE_TOLERANCE_BITS ${BACKEND_NAME}_DOUBLE_TOLERANCE_BITS
+#endif
+// clang-format on
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/engine/test_engines.hpp"
+#include "util/test_case.hpp"
+#include "util/test_control.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+using TestEngine = test::ENGINE_CLASS_NAME(${BACKEND_NAME});
+
+NGRAPH_TEST(${BACKEND_NAME}, softplus)
+{
+    auto A = make_shared<op::Parameter>(element::f32, Shape{4});
+    auto softplus = make_shared<op::v4::SoftPlus>(A);
+    auto function = make_shared<Function>(NodeVector{softplus}, ParameterVector{A});
+
+    auto test_case = test::TestCase<TestEngine>(function);
+    test_case.add_input(vector<float>{-1.0, 0.0, 1.0, 20.0});
+    test_case.add_expected_output(
+        vector<float>{0.31326166, 0.69314718, 1.3132616, 20.0});
+    test_case.run();
+}
diff --git a/ngraph/test/type_prop/softplus.cpp b/ngraph/test/type_prop/softplus.cpp
index 7d6ddb49144b0e..d0cec9fdd27d6d 100644
--- a/ngraph/test/type_prop/softplus.cpp
+++ b/ngraph/test/type_prop/softplus.cpp
@@ -40,3 +40,23 @@ TEST(type_prop, softplus_partial_static_rank)
         (PartialShape{1, Dimension::dynamic(), 6})));
     ASSERT_TRUE(softplus_func->get_output_partial_shape(0).rank().is_static());
 }
+
+TEST(type_prop, softplus_invalid_element_type)
+{
+    auto data = make_shared<op::Parameter>(element::i32, Shape{2, 2});
+
+    try
+    {
+        auto softplus = make_shared<op::v4::SoftPlus>(data);
+        // Input element type is boolean
+        FAIL() << "Invalid int element type for input not detected";
+    }
+    catch (const NodeValidationFailure& error)
+    {
+        EXPECT_HAS_SUBSTRING(error.what(), "Input element type must be float");
+    }
+    catch (...)
+    {
+        FAIL() << "Numeric element type node validation check failed for unexpected reason";
+    }
+}
diff --git a/ngraph/test/visitors/op/softplus.cpp b/ngraph/test/visitors/op/softplus.cpp
new file mode 100644
index 00000000000000..e5ff77e3a4e86e
--- /dev/null
+++ b/ngraph/test/visitors/op/softplus.cpp
@@ -0,0 +1,12 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "unary_ops.hpp"
+
+using Types = ::testing::Types<UnaryOperatorType<ngraph::op::v4::SoftPlus, element::f32>>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+                              UnaryOperatorVisitor,
+                              Types,
+                              UnaryOperatorTypeName);

From b68166fc3cb97116dc5685539f852d5174601b21 Mon Sep 17 00:00:00 2001
From: Yuan Hu <yuan2.hu@intel.com>
Date: Tue, 22 Jun 2021 16:05:26 +0800
Subject: [PATCH 155/178] Revise Round operation reference implementation
 (#6287)

* Revise OP Round

improve the input type error check
add attribute test case

Signed-off-by: Hu, Yuan2 <yuan2.hu@intel.com>

* fix clang code style issue

Signed-off-by: Hu, Yuan2 <yuan2.hu@intel.com>
---
 ngraph/core/src/op/round.cpp      |  2 ++
 ngraph/test/CMakeLists.txt        |  1 +
 ngraph/test/backend/round.in.cpp  |  2 +-
 ngraph/test/visitors/op/round.cpp | 42 +++++++++++++++++++++++++++++++
 4 files changed, 46 insertions(+), 1 deletion(-)
 create mode 100644 ngraph/test/visitors/op/round.cpp

diff --git a/ngraph/core/src/op/round.cpp b/ngraph/core/src/op/round.cpp
index dde7a19d81bff1..83ed1fff66bb88 100644
--- a/ngraph/core/src/op/round.cpp
+++ b/ngraph/core/src/op/round.cpp
@@ -83,6 +83,8 @@ bool ngraph::op::v5::Round::visit_attributes(AttributeVisitor& visitor)
 void op::v5::Round::validate_and_infer_types()
 {
     NGRAPH_OP_SCOPE(v5_Round_validate_and_infer_types);
+    NODE_VALIDATION_CHECK(
+        this, get_input_size() == 1, "Only accepts one argument. Got: ", get_input_size());
     set_output_size(1);
     set_output_type(0, get_input_element_type(0), get_input_partial_shape(0));
 }
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index 6f4e2bb0db47a2..e25b6410ba09dc 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -284,6 +284,7 @@ set(SRC
     visitors/op/reverse_sequence.cpp
     visitors/op/rnn_cell.cpp
     visitors/op/roi_pooling.cpp
+    visitors/op/round.cpp
     visitors/op/selu.cpp
     visitors/op/shuffle_channels.cpp
     visitors/op/softmax.cpp
diff --git a/ngraph/test/backend/round.in.cpp b/ngraph/test/backend/round.in.cpp
index c58ecc7d8dcbbf..d8fa657b0101d9 100644
--- a/ngraph/test/backend/round.in.cpp
+++ b/ngraph/test/backend/round.in.cpp
@@ -19,7 +19,7 @@ using namespace ngraph;
 
 static string s_manifest = "${MANIFEST}";
 
-NGRAPH_TEST(${BACKEND_NAME}, round)
+NGRAPH_TEST(${BACKEND_NAME}, round_half_to_even)
 {
     Shape shape{5};
     auto A = make_shared<op::Parameter>(element::f32, shape);
diff --git a/ngraph/test/visitors/op/round.cpp b/ngraph/test/visitors/op/round.cpp
new file mode 100644
index 00000000000000..159d07edd332d7
--- /dev/null
+++ b/ngraph/test/visitors/op/round.cpp
@@ -0,0 +1,42 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+
+#include "ngraph/ngraph.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+#include "ngraph/opsets/opset1.hpp"
+#include "ngraph/opsets/opset3.hpp"
+#include "ngraph/opsets/opset4.hpp"
+#include "ngraph/opsets/opset5.hpp"
+
+#include "util/visitor.hpp"
+
+using namespace std;
+using namespace ngraph;
+using ngraph::test::NodeBuilder;
+using ngraph::test::ValueMap;
+
+void static test_mode(opset5::Round::RoundMode mode)
+{
+    NodeBuilder::get_ops().register_factory<opset5::Round>();
+    auto data = make_shared<op::Parameter>(element::f32, Shape{200});
+    auto round = make_shared<opset5::Round>(data, opset5::Round::RoundMode::HALF_TO_EVEN);
+    NodeBuilder builder(round);
+    auto g_round = as_type_ptr<opset5::Round>(builder.create());
+
+    EXPECT_EQ(g_round->get_mode(), round->get_mode());
+}
+
+TEST(attributes, round_op_enum_mode_half_to_even)
+{
+    test_mode(opset5::Round::RoundMode::HALF_TO_EVEN);
+}
+
+TEST(attributes, round_op_enum_mode_half_away_from_zero)
+{
+    test_mode(opset5::Round::RoundMode::HALF_AWAY_FROM_ZERO);
+}
+
+

From 38bb6dc8741a085c727bc37d7fd0f8455601af5f Mon Sep 17 00:00:00 2001
From: cecilia peng <cecilia.peng@intel.com>
Date: Tue, 22 Jun 2021 17:34:34 +0800
Subject: [PATCH 156/178] spec revise: add attribute "normalize" to Paddle NMS
 operators. (#6290)

---
 docs/ops/sort/MatrixNMS_8.md     | 10 ++++++++++
 docs/ops/sort/MulticlassNMS_8.md | 10 ++++++++++
 2 files changed, 20 insertions(+)

diff --git a/docs/ops/sort/MatrixNMS_8.md b/docs/ops/sort/MatrixNMS_8.md
index d3a422307223c9..b578822d6ebd81 100644
--- a/docs/ops/sort/MatrixNMS_8.md
+++ b/docs/ops/sort/MatrixNMS_8.md
@@ -84,6 +84,16 @@ The Matrix NMS algorithm is described below:
   * **Default value**: `-1` meaning to keep all classes
   * **Required**: *No*
 
+* *normalized*
+
+  * **Description**: *normalized* is a flag that indicates whether `boxes` are normalized or not.
+  * **Range of values**: true or false
+    * *true* - the box coordinates are normalized.
+    * *false* - the box coordinates are not normalized.
+  * **Type**: boolean
+  * **Default value**: True
+  * **Required**: *No*
+
 * *decay_function*
 
   * **Description**: decay function used to decay scores.
diff --git a/docs/ops/sort/MulticlassNMS_8.md b/docs/ops/sort/MulticlassNMS_8.md
index 16997a81397ef0..866254963fdfa8 100644
--- a/docs/ops/sort/MulticlassNMS_8.md
+++ b/docs/ops/sort/MulticlassNMS_8.md
@@ -93,6 +93,16 @@ Boxes of `background_class` are skipped and thus eliminated.
   * **Default value**: `-1` meaning to keep all classes.
   * **Required**: *No*
 
+* *normalized*
+
+  * **Description**: *normalized* is a flag that indicates whether `boxes` are normalized or not.
+  * **Range of values**: true or false
+    * *true* - the box coordinates are normalized.
+    * *false* - the box coordinates are not normalized.
+  * **Type**: boolean
+  * **Default value**: True
+  * **Required**: *No*
+
 * *nms_eta*
 
   * **Description**: eta parameter for adaptive NMS.

From 15ee515a88085a3209dfdef36f2ce4d97e2153e9 Mon Sep 17 00:00:00 2001
From: Edward Shogulin <edward.shogulin@intel.com>
Date: Tue, 22 Jun 2021 13:05:27 +0300
Subject: [PATCH 157/178] [Runtime] INT8 inference documentation update (#5949)

---
 docs/IE_DG/Int8Inference.md      | 47 ++++++++++++++++----------------
 docs/IE_DG/img/cpu_int8_flow.png |  3 --
 2 files changed, 23 insertions(+), 27 deletions(-)
 delete mode 100644 docs/IE_DG/img/cpu_int8_flow.png

diff --git a/docs/IE_DG/Int8Inference.md b/docs/IE_DG/Int8Inference.md
index 917c7836de293b..889af6a53278b1 100644
--- a/docs/IE_DG/Int8Inference.md
+++ b/docs/IE_DG/Int8Inference.md
@@ -1,6 +1,13 @@
 # Low-Precision 8-bit Integer Inference {#openvino_docs_IE_DG_Int8Inference}
 
-## Disclaimer
+## Table of Contents
+1. [Supported devices](#supported-devices)
+2. [Low-Precision 8-bit Integer Inference Workflow](#low-precision-8-bit-integer-inference-workflow)
+3. [Prerequisites](#prerequisites)
+4. [Inference](#inference)
+5. [Results analysis](#results-analysis)
+
+## Supported devices
 
 Low-precision 8-bit inference is optimized for:
 - Intel® architecture processors with the following instruction set architecture extensions:  
@@ -12,16 +19,22 @@ Low-precision 8-bit inference is optimized for:
   - Intel® Iris® Xe Graphics
   - Intel® Iris® Xe MAX Graphics
 - A model must be quantized. You can use a quantized model from [OpenVINO™ Toolkit Intel's Pre-Trained Models](@ref omz_models_group_intel) or quantize a model yourself. For quantization, you can use the:
-  - [Post-Training Optimization Tool](@ref pot_README) delivered with the Intel® Distribution of OpenVINO™ toolkit release package.
+  - [Post-Training Optimization Tool](@ref pot_docs_LowPrecisionOptimizationGuide) delivered with the Intel® Distribution of OpenVINO™ toolkit release package.
   - [Neural Network Compression Framework](https://www.intel.com/content/www/us/en/artificial-intelligence/posts/openvino-nncf.html) available on GitHub: https://github.com/openvinotoolkit/nncf
 
-## Introduction
-
-A lot of investigation was made in the field of deep learning with the idea of using low precision computations during inference in order to boost deep learning pipelines and gather higher performance. For example, one of the popular approaches is to shrink the precision of activations and weights values from `fp32` precision to smaller ones, for example, to `fp11` or `int8`. For more information about this approach, refer to 
-**Brief History of Lower Precision in Deep Learning** section in [this whitepaper](https://software.intel.com/en-us/articles/lower-numerical-precision-deep-learning-inference-and-training).
+## Low-Precision 8-bit Integer Inference Workflow
 
 8-bit computations (referred to as `int8`) offer better performance compared to the results of inference in higher precision (for example, `fp32`), because they allow loading more data into a single processor instruction. Usually the cost for significant boost is a reduced accuracy. However, it is proved that an accuracy drop can be negligible and depends on task requirements, so that the application engineer can set up the maximum accuracy drop that is acceptable.
 
+For 8-bit integer computations, a model must be quantized. Quantized models can be downloaded from [Overview of OpenVINO™ Toolkit Intel's Pre-Trained Models](@ref omz_models_group_intel). If the model is not quantized, you can use the [Post-Training Optimization Tool](@ref pot_README) to quantize the model. The quantization process adds [FakeQuantize](../ops/quantization/FakeQuantize_1.md) layers on activations and weights for most layers. Read more about mathematical computations in the [Uniform Quantization with Fine-Tuning](https://github.com/openvinotoolkit/nncf/blob/develop/docs/compression_algorithms/Quantization.md).
+
+When you pass the quantized IR to the OpenVINO™ plugin, the plugin automatically recognizes it as a quantized model and performs 8-bit inference. Note, if you pass a quantized model to another plugin that does not support 8-bit inference but supports all operations from the model, the model is inferred in precision that this plugin supports.
+
+In *Runtime stage* stage, the quantized model is loaded to the plugin. The plugin uses `Low Precision Transformation` component to update the model to infer it in low precision:
+   - Update `FakeQuantize` layers to have quantized output tensors in low precision range and add dequantization layers to compensate the update. Dequantization layers are pushed through as many layers as possible to have more layers in low precision. After that, most layers have quantized input tensors in low precision range and can be inferred in low precision. Ideally, dequantization layers should be fused in the next `FakeQuantize` layer.
+   - Weights are quantized and stored in `Constant` layers. 
+
+## Prerequisites
 
 Let's explore quantized [TensorFlow* implementation of ResNet-50](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/resnet-50-tf) model. Use [Model Downloader](@ref omz_tools_downloader) tool to download the `fp16` model from [OpenVINO™ Toolkit - Open Model Zoo repository](https://github.com/openvinotoolkit/open_model_zoo):
 ```sh
@@ -31,28 +44,16 @@ After that you should quantize model by the [Model Quantizer](@ref omz_tools_dow
 ```sh
 ./quantizer.py --model_dir public/resnet-50-tf --dataset_dir <DATASET_DIR> --precisions=FP16-INT8
 ```
+
+## Inference
+
 The simplest way to infer the model and collect performance counters is [C++ Benchmark Application](../../inference-engine/samples/benchmark_app/README.md). 
 ```sh
 ./benchmark_app -m resnet-50-tf.xml -d CPU -niter 1 -api sync -report_type average_counters  -report_folder pc_report_dir
 ```
 If you infer the model with the OpenVINO™ CPU plugin and collect performance counters, all operations (except last not quantized SoftMax) are executed in INT8 precision.  
 
-## Low-Precision 8-bit Integer Inference Workflow
-
-For 8-bit integer computations, a model must be quantized. Quantized models can be downloaded from [Overview of OpenVINO™ Toolkit Intel's Pre-Trained Models](@ref omz_models_group_intel). If the model is not quantized, you can use the [Post-Training Optimization Tool](@ref pot_README) to quantize the model. The quantization process adds [FakeQuantize](../ops/quantization/FakeQuantize_1.md) layers on activations and weights for most layers. Read more about mathematical computations in the [Uniform Quantization with Fine-Tuning](https://github.com/openvinotoolkit/nncf/blob/develop/docs/compression_algorithms/Quantization.md).
-
-8-bit inference pipeline includes two stages (also refer to the figure below):
-1. *Offline stage*, or *model quantization*. During this stage, [FakeQuantize](../ops/quantization/FakeQuantize_1.md) layers are added before most layers to have quantized tensors before layers in a way that low-precision accuracy drop for 8-bit integer inference satisfies the specified threshold. The output of this stage is a quantized model. Quantized model precision is not changed, quantized tensors are in original precision range (`fp32`). `FakeQuantize` layer has `levels` attribute which defines quants count. Quants count defines precision which is used during inference. For `int8` range `levels` attribute value has to be 255 or 256. To quantize the model, you can use the [Post-Training Optimization Tool](@ref pot_README) delivered with the Intel® Distribution of OpenVINO™ toolkit release package.
-
-   When you pass the quantized IR to the OpenVINO™ plugin, the plugin automatically recognizes it as a quantized model and performs 8-bit inference. Note, if you pass a quantized model to another plugin that does not support 8-bit inference but supports all operations from the model, the model is inferred in precision that this plugin supports.
-
-2. *Runtime stage*. This stage is an internal procedure of the OpenVINO™ plugin. During this stage, the quantized model is loaded to the plugin. The plugin uses `Low Precision Transformation` component to update the model to infer it in low precision:
-   - Update `FakeQuantize` layers to have quantized output tensors in low precision range and add dequantization layers to compensate the update. Dequantization layers are pushed through as many layers as possible to have more layers in low precision. After that, most layers have quantized input tensors in low precision range and can be inferred in low precision. Ideally, dequantization layers should be fused in the next `FakeQuantize` layer.
-   - Weights are quantized and stored in `Constant` layers. 
-
-![int8_flow]
-
-## Performance Counters
+## Results analysis
 
 Information about layer precision is stored in the performance counters that are
 available from the Inference Engine API. For example, the part of performance counters table for quantized [TensorFlow* implementation of ResNet-50](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/resnet-50-tf) model inference on [CPU Plugin](supported_plugins/CPU.md) looks as follows:
@@ -79,5 +80,3 @@ available from the Inference Engine API. For example, the part of performance co
 > * Suffix `FP32` for layers computed in 32-bit precision 
 
 All `Convolution` layers are executed in int8 precision. Rest layers are fused into Convolutions using post operations optimization technique, which is described in [Internal CPU Plugin Optimizations](supported_plugins/CPU.md).
-
-[int8_flow]: img/cpu_int8_flow.png
diff --git a/docs/IE_DG/img/cpu_int8_flow.png b/docs/IE_DG/img/cpu_int8_flow.png
deleted file mode 100644
index 794430126b2877..00000000000000
--- a/docs/IE_DG/img/cpu_int8_flow.png
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:83bcd7888d3843ddfd9a601288627e98f5874290c00b9988bf1beac9209f2e8d
-size 79741

From b0e932567d72557123ac49fc6d1ccd8cef53ad9a Mon Sep 17 00:00:00 2001
From: Evgenya Stepyreva <evgenya.stepyreva@intel.com>
Date: Tue, 22 Jun 2021 14:35:30 +0300
Subject: [PATCH 158/178] ShuffleChannels shape propagation unified (#6269)

---
 ngraph/core/src/op/shuffle_channels.cpp | 11 ++---------
 1 file changed, 2 insertions(+), 9 deletions(-)

diff --git a/ngraph/core/src/op/shuffle_channels.cpp b/ngraph/core/src/op/shuffle_channels.cpp
index 71683af4030d61..d38a560f72b9c3 100644
--- a/ngraph/core/src/op/shuffle_channels.cpp
+++ b/ngraph/core/src/op/shuffle_channels.cpp
@@ -63,11 +63,10 @@ void op::ShuffleChannels::validate_and_infer_types()
     if (get_input_partial_shape(0).is_static())
     {
         const auto shape = get_input_shape(0);
-
         NODE_VALIDATION_CHECK(
             this, shape.size() >= 1, "The input tensor's shape is expected to be at least 1D.");
-        size_t axis_zb = get_zero_based_axis();
 
+        size_t axis_zb = get_zero_based_axis();
         NODE_VALIDATION_CHECK(this,
                               axis_zb < shape.size(),
                               "The 'axis' parameter for ShuffleChannels has to point to one of the "
@@ -81,14 +80,8 @@ void op::ShuffleChannels::validate_and_infer_types()
             this,
             channel_dim_size % m_group == 0,
             "The channel dimension size has to be a multiple of the groups parameter value.");
-        set_output_size(1);
-        set_output_type(0, data_type, shape);
-    }
-    else
-    {
-        const auto shape = get_input_partial_shape(0);
-        set_output_type(0, data_type, shape);
     }
+    set_output_type(0, data_type, get_input_partial_shape(0));
 }
 
 shared_ptr<Node> op::ShuffleChannels::clone_with_new_inputs(const OutputVector& new_args) const

From 1bc18d8f4e8bd9b98e367b1becd3fd678853d5a6 Mon Sep 17 00:00:00 2001
From: Maxim Vafin <maxim.vafin@intel.com>
Date: Tue, 22 Jun 2021 15:00:27 +0300
Subject: [PATCH 159/178] Add ngraph shell for AdaptivePool operations (#6126)

Apply review feedback

Fix code style

Add visitors tests

Check visitor for type attribute
---
 .../include/ngraph/op/adaptive_avg_pool.hpp   |  43 ++++++
 .../include/ngraph/op/adaptive_max_pool.hpp   |  54 ++++++++
 ngraph/core/include/ngraph/ops.hpp            |   2 +
 ngraph/core/include/ngraph/opsets/opset.hpp   |   1 +
 .../core/include/ngraph/opsets/opset8_tbl.hpp |   2 +
 ngraph/core/src/op/adaptive_avg_pool.cpp      |  73 ++++++++++
 ngraph/core/src/op/adaptive_max_pool.cpp      |  83 ++++++++++++
 ngraph/test/CMakeLists.txt                    |   4 +
 ngraph/test/type_prop/adaptive_avg_pool.cpp   | 107 +++++++++++++++
 ngraph/test/type_prop/adaptive_max_pool.cpp   | 128 ++++++++++++++++++
 ngraph/test/visitors/op/adaptive_avg_pool.cpp |  27 ++++
 ngraph/test/visitors/op/adaptive_max_pool.cpp |  29 ++++
 12 files changed, 553 insertions(+)
 create mode 100644 ngraph/core/include/ngraph/op/adaptive_avg_pool.hpp
 create mode 100644 ngraph/core/include/ngraph/op/adaptive_max_pool.hpp
 create mode 100644 ngraph/core/src/op/adaptive_avg_pool.cpp
 create mode 100644 ngraph/core/src/op/adaptive_max_pool.cpp
 create mode 100644 ngraph/test/type_prop/adaptive_avg_pool.cpp
 create mode 100644 ngraph/test/type_prop/adaptive_max_pool.cpp
 create mode 100644 ngraph/test/visitors/op/adaptive_avg_pool.cpp
 create mode 100644 ngraph/test/visitors/op/adaptive_max_pool.cpp

diff --git a/ngraph/core/include/ngraph/op/adaptive_avg_pool.hpp b/ngraph/core/include/ngraph/op/adaptive_avg_pool.hpp
new file mode 100644
index 00000000000000..e030cd8852d174
--- /dev/null
+++ b/ngraph/core/include/ngraph/op/adaptive_avg_pool.hpp
@@ -0,0 +1,43 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "ngraph/op/op.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+
+namespace ngraph
+{
+    namespace op
+    {
+        namespace v8
+        {
+            /// \brief Adaptive average pooling operation.
+            ///
+            class NGRAPH_API AdaptiveAvgPool : public Op
+            {
+            public:
+                NGRAPH_RTTI_DECLARATION;
+
+                AdaptiveAvgPool() = default;
+
+                ///
+                /// \brief    Constructs adaptive average pooling operation.
+                ///
+                /// \param    data            Input data
+                ///
+                /// \param    output_shape    1D tensor describing output shape for spatial
+                ///                           dimensions.
+                ///
+                AdaptiveAvgPool(const Output<Node>& data, const Output<Node>& output_shape);
+
+                void validate_and_infer_types() override;
+                bool visit_attributes(AttributeVisitor& visitor) override;
+
+                std::shared_ptr<Node>
+                    clone_with_new_inputs(const OutputVector& new_args) const override;
+            };
+        } // namespace v8
+    }     // namespace op
+} // namespace ngraph
diff --git a/ngraph/core/include/ngraph/op/adaptive_max_pool.hpp b/ngraph/core/include/ngraph/op/adaptive_max_pool.hpp
new file mode 100644
index 00000000000000..2e528c5ed87d3b
--- /dev/null
+++ b/ngraph/core/include/ngraph/op/adaptive_max_pool.hpp
@@ -0,0 +1,54 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "ngraph/op/op.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+
+namespace ngraph
+{
+    namespace op
+    {
+        namespace v8
+        {
+            /// \brief Adaptive max pooling operation.
+            ///
+            class NGRAPH_API AdaptiveMaxPool : public Op
+            {
+            public:
+                NGRAPH_RTTI_DECLARATION;
+
+                AdaptiveMaxPool() = default;
+
+                ///
+                /// \brief    Constructs adaptive max pooling operation.
+                ///
+                /// \param    data                  Input data
+                ///
+                /// \param    output_shape          1D tensor describing output shape for spatial
+                ///                                 dimensions.
+                ///
+                /// \param    index_element_type    Specifies the output tensor type for indices
+                /// output
+                ///
+                AdaptiveMaxPool(
+                    const Output<Node>& data,
+                    const Output<Node>& output_shape,
+                    const ngraph::element::Type& index_element_type = ngraph::element::i64);
+
+                void validate_and_infer_types() override;
+                bool visit_attributes(AttributeVisitor& visitor) override;
+
+                std::shared_ptr<Node>
+                    clone_with_new_inputs(const OutputVector& new_args) const override;
+
+                element::Type get_index_element_type() const { return m_index_element_type; }
+
+            protected:
+                ngraph::element::Type m_index_element_type = ngraph::element::i64;
+            };
+        } // namespace v8
+    }     // namespace op
+} // namespace ngraph
diff --git a/ngraph/core/include/ngraph/ops.hpp b/ngraph/core/include/ngraph/ops.hpp
index 5a4ad078616db7..6999dc93c9de71 100644
--- a/ngraph/core/include/ngraph/ops.hpp
+++ b/ngraph/core/include/ngraph/ops.hpp
@@ -9,6 +9,8 @@
 #include "ngraph/op/abs.hpp"
 #include "ngraph/op/acos.hpp"
 #include "ngraph/op/acosh.hpp"
+#include "ngraph/op/adaptive_avg_pool.hpp"
+#include "ngraph/op/adaptive_max_pool.hpp"
 #include "ngraph/op/add.hpp"
 #include "ngraph/op/and.hpp"
 #include "ngraph/op/asin.hpp"
diff --git a/ngraph/core/include/ngraph/opsets/opset.hpp b/ngraph/core/include/ngraph/opsets/opset.hpp
index f0929ed839da55..9ba03cbcd89a35 100644
--- a/ngraph/core/include/ngraph/opsets/opset.hpp
+++ b/ngraph/core/include/ngraph/opsets/opset.hpp
@@ -124,4 +124,5 @@ namespace ngraph
     const NGRAPH_API OpSet& get_opset5();
     const NGRAPH_API OpSet& get_opset6();
     const NGRAPH_API OpSet& get_opset7();
+    const NGRAPH_API OpSet& get_opset8();
 } // namespace ngraph
diff --git a/ngraph/core/include/ngraph/opsets/opset8_tbl.hpp b/ngraph/core/include/ngraph/opsets/opset8_tbl.hpp
index 0dbe077ddae1bd..627441eb30033b 100644
--- a/ngraph/core/include/ngraph/opsets/opset8_tbl.hpp
+++ b/ngraph/core/include/ngraph/opsets/opset8_tbl.hpp
@@ -177,3 +177,5 @@ NGRAPH_OP(Roll, ngraph::op::v7)
 
 // New operations added in opset8
 NGRAPH_OP(Gather, ngraph::op::v8)
+NGRAPH_OP(AdaptiveAvgPool, ngraph::op::v8)
+NGRAPH_OP(AdaptiveMaxPool, ngraph::op::v8)
diff --git a/ngraph/core/src/op/adaptive_avg_pool.cpp b/ngraph/core/src/op/adaptive_avg_pool.cpp
new file mode 100644
index 00000000000000..ef0321fad1769a
--- /dev/null
+++ b/ngraph/core/src/op/adaptive_avg_pool.cpp
@@ -0,0 +1,73 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "ngraph/op/adaptive_avg_pool.hpp"
+#include "itt.hpp"
+#include "ngraph/attribute_visitor.hpp"
+#include "ngraph/graph_util.hpp"
+#include "ngraph/validation_util.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+NGRAPH_RTTI_DEFINITION(op::v8::AdaptiveAvgPool, "AdaptiveAvgPool", 8);
+
+op::v8::AdaptiveAvgPool::AdaptiveAvgPool(const Output<Node>& data, const Output<Node>& output_shape)
+    : Op({data, output_shape})
+{
+    constructor_validate_and_infer_types();
+}
+
+bool op::v8::AdaptiveAvgPool::visit_attributes(AttributeVisitor& visitor)
+{
+    NGRAPH_OP_SCOPE(v8_AdaptiveAvgPool_visit_attributes);
+    return true;
+}
+
+void op::v8::AdaptiveAvgPool::validate_and_infer_types()
+{
+    NGRAPH_OP_SCOPE(v8_AdaptiveAvgPool_validate_and_infer_types);
+
+    const PartialShape& data_shape = get_input_partial_shape(0);
+
+    NODE_VALIDATION_CHECK(this,
+                          data_shape.rank().compatible(3) || data_shape.rank().compatible(4) ||
+                              data_shape.rank().compatible(5),
+                          "Expected a 3D, 4D or 5D tensor for the input. Got: ",
+                          data_shape);
+
+    auto output_shape = PartialShape::dynamic(data_shape.rank());
+    if (data_shape.rank().is_static())
+    {
+        if (data_shape[0].is_static())
+        {
+            output_shape[0] = data_shape[0]; // batch size
+        }
+        if (data_shape[1].is_static())
+        {
+            output_shape[1] = data_shape[1]; // channel size
+        }
+        if (const auto& const_output_shape = get_constant_from_source(input_value(1)))
+        {
+            auto output_spatial_shape = const_output_shape->cast_vector<int64_t>();
+            NODE_VALIDATION_CHECK(this,
+                                  (size_t)data_shape.rank().get_length() ==
+                                      2 + output_spatial_shape.size(),
+                                  "Output shape is not compatible with input data rank");
+            int i = 2;
+            for (auto& dim : output_spatial_shape)
+            {
+                output_shape[i++] = dim;
+            }
+        }
+    }
+    set_output_type(0, get_input_element_type(0), output_shape);
+}
+
+shared_ptr<Node> op::v8::AdaptiveAvgPool::clone_with_new_inputs(const OutputVector& new_args) const
+{
+    NGRAPH_OP_SCOPE(v8_AdaptiveAvgPool_clone_with_new_inputs);
+    check_new_args_count(this, new_args);
+    return make_shared<v8::AdaptiveAvgPool>(new_args.at(0), new_args.at(1));
+}
diff --git a/ngraph/core/src/op/adaptive_max_pool.cpp b/ngraph/core/src/op/adaptive_max_pool.cpp
new file mode 100644
index 00000000000000..88f4fc1e230799
--- /dev/null
+++ b/ngraph/core/src/op/adaptive_max_pool.cpp
@@ -0,0 +1,83 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "ngraph/op/adaptive_max_pool.hpp"
+#include "itt.hpp"
+#include "ngraph/attribute_visitor.hpp"
+#include "ngraph/graph_util.hpp"
+#include "ngraph/validation_util.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+NGRAPH_RTTI_DEFINITION(op::v8::AdaptiveMaxPool, "AdaptiveMaxPool", 8);
+
+op::v8::AdaptiveMaxPool::AdaptiveMaxPool(const Output<Node>& data,
+                                         const Output<Node>& output_shape,
+                                         const ngraph::element::Type& index_element_type)
+    : Op({data, output_shape})
+    , m_index_element_type{index_element_type}
+{
+    constructor_validate_and_infer_types();
+}
+
+bool op::v8::AdaptiveMaxPool::visit_attributes(AttributeVisitor& visitor)
+{
+    NGRAPH_OP_SCOPE(v8_AdaptiveMaxPool_visit_attributes);
+    visitor.on_attribute("index_element_type", m_index_element_type);
+    return true;
+}
+
+void op::v8::AdaptiveMaxPool::validate_and_infer_types()
+{
+    NGRAPH_OP_SCOPE(v8_AdaptiveMaxPool_validate_and_infer_types);
+
+    NODE_VALIDATION_CHECK(this,
+                          m_index_element_type == element::i64 ||
+                              m_index_element_type == element::i32,
+                          "Index element type must be i32 or i64");
+
+    const PartialShape& data_shape = get_input_partial_shape(0);
+
+    NODE_VALIDATION_CHECK(this,
+                          data_shape.rank().compatible(3) || data_shape.rank().compatible(4) ||
+                              data_shape.rank().compatible(5),
+                          "Expected a 3D, 4D or 5D tensor for the input. Got: ",
+                          data_shape);
+
+    auto output_shape = PartialShape::dynamic(data_shape.rank());
+    if (data_shape.rank().is_static())
+    {
+        if (data_shape[0].is_static())
+        {
+            output_shape[0] = data_shape[0]; // batch size
+        }
+        if (data_shape[1].is_static())
+        {
+            output_shape[1] = data_shape[1]; // channel size
+        }
+        if (const auto& const_output_shape = get_constant_from_source(input_value(1)))
+        {
+            auto output_spatial_shape = const_output_shape->cast_vector<int64_t>();
+            NODE_VALIDATION_CHECK(this,
+                                  (size_t)data_shape.rank().get_length() ==
+                                      2 + output_spatial_shape.size(),
+                                  "Output shape is not compatible with input data rank");
+            int i = 2;
+            for (auto& dim : output_spatial_shape)
+            {
+                output_shape[i++] = dim;
+            }
+        }
+    }
+    set_output_type(0, get_input_element_type(0), output_shape);
+    set_output_type(1, m_index_element_type, output_shape);
+}
+
+shared_ptr<Node> op::v8::AdaptiveMaxPool::clone_with_new_inputs(const OutputVector& new_args) const
+{
+    NGRAPH_OP_SCOPE(v8_AdaptiveMaxPool_clone_with_new_inputs);
+    check_new_args_count(this, new_args);
+    return make_shared<v8::AdaptiveMaxPool>(new_args.at(0), new_args.at(1), m_index_element_type);
+}
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index e25b6410ba09dc..f6d12b6483d4f6 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -101,6 +101,8 @@ set(SRC
     tensor.cpp
     type_prop/abs.cpp
     type_prop/acos.cpp
+    type_prop/adaptive_avg_pool.cpp
+    type_prop/adaptive_max_pool.cpp
     type_prop/asin.cpp
     type_prop/assign.cpp
     type_prop/avg_pool.cpp
@@ -229,6 +231,8 @@ set(SRC
     visitors/partial_shape.cpp
     visitors/user_op.cpp
     visitors/value_map.cpp
+    visitors/op/adaptive_avg_pool.cpp
+    visitors/op/adaptive_max_pool.cpp
     visitors/op/batch_norm.cpp
     visitors/op/broadcast.cpp
     visitors/op/bucketize.cpp
diff --git a/ngraph/test/type_prop/adaptive_avg_pool.cpp b/ngraph/test/type_prop/adaptive_avg_pool.cpp
new file mode 100644
index 00000000000000..09923756bc3fd7
--- /dev/null
+++ b/ngraph/test/type_prop/adaptive_avg_pool.cpp
@@ -0,0 +1,107 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/type_prop.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+TEST(type_prop, adaptive_avg_pool)
+{
+    const PartialShape arg_shape{1, 6, 8, 9};
+    const vector<int64_t> output_shape{5, 7};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, output_shape);
+    auto adaptive_pool = make_shared<op::v8::AdaptiveAvgPool>(data, out_shape);
+
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(0).same_scheme({1, 6, 5, 7}));
+}
+
+TEST(type_prop, adaptive_avg_pool_dyn_batch)
+{
+    const PartialShape arg_shape{Dimension::dynamic(), 6, 8, 9};
+    const vector<int64_t> output_shape{5, 7};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, output_shape);
+    auto adaptive_pool = make_shared<op::v8::AdaptiveAvgPool>(data, out_shape);
+
+    ASSERT_TRUE(
+        adaptive_pool->get_output_partial_shape(0).same_scheme({Dimension::dynamic(), 6, 5, 7}));
+}
+
+TEST(type_prop, adaptive_avg_pool_dyn_channels)
+{
+    const PartialShape arg_shape{1, Dimension::dynamic(), 8, 9};
+    const vector<int64_t> output_shape{5, 7};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, output_shape);
+    auto adaptive_pool = make_shared<op::v8::AdaptiveAvgPool>(data, out_shape);
+
+    ASSERT_TRUE(
+        adaptive_pool->get_output_partial_shape(0).same_scheme({1, Dimension::dynamic(), 5, 7}));
+}
+
+TEST(type_prop, adaptive_avg_pool_dyn_spatial)
+{
+    const PartialShape arg_shape{1, 6, Dimension::dynamic(), Dimension::dynamic()};
+    const vector<int64_t> output_shape{5, 7};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, output_shape);
+    auto adaptive_pool = make_shared<op::v8::AdaptiveAvgPool>(data, out_shape);
+
+    ASSERT_TRUE(
+        adaptive_pool->get_output_partial_shape(0).same_scheme({1, 6, 5, 7}));
+}
+
+TEST(type_prop, adaptive_avg_pool_dyn_output_shape)
+{
+    const PartialShape arg_shape{1, 6, 8, 9};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = make_shared<op::Parameter>(element::i64, Shape{2});
+    auto adaptive_pool = make_shared<op::v8::AdaptiveAvgPool>(data, out_shape);
+
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(0).same_scheme(
+        {1, 6, Dimension::dynamic(), Dimension::dynamic()}));
+}
+
+TEST(type_prop, adaptive_avg_pool_dyn_rank)
+{
+    const PartialShape arg_shape = PartialShape::dynamic();
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = make_shared<op::Parameter>(element::i64, Shape{2});
+    auto adaptive_pool = make_shared<op::v8::AdaptiveAvgPool>(data, out_shape);
+
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(0).same_scheme(PartialShape::dynamic()));
+}
+
+TEST(type_prop, adaptive_avg_pool_unsupported_input_shape)
+{
+    const PartialShape arg_shape{1, 6};
+    const vector<int64_t> output_shape{1};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{}, output_shape);
+
+    EXPECT_THROW(make_shared<op::v8::AdaptiveAvgPool>(data, out_shape), NodeValidationFailure);
+}
+
+TEST(type_prop, adaptive_avg_pool_wrong_out_shape)
+{
+    const PartialShape arg_shape{1, 6, 8, 9};
+    const vector<int64_t> output_shape{5, 7, 8};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{3}, output_shape);
+
+    EXPECT_THROW(make_shared<op::v8::AdaptiveAvgPool>(data, out_shape),
+                 NodeValidationFailure);
+}
diff --git a/ngraph/test/type_prop/adaptive_max_pool.cpp b/ngraph/test/type_prop/adaptive_max_pool.cpp
new file mode 100644
index 00000000000000..0526be23c446a1
--- /dev/null
+++ b/ngraph/test/type_prop/adaptive_max_pool.cpp
@@ -0,0 +1,128 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/type_prop.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+TEST(type_prop, adaptive_max_pool)
+{
+    const PartialShape arg_shape{1, 6, 8, 9};
+    const vector<int64_t> output_shape{5, 7};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, output_shape);
+    auto adaptive_pool = make_shared<op::v8::AdaptiveMaxPool>(data, out_shape);
+
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(0).same_scheme({1, 6, 5, 7}));
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(1).same_scheme({1, 6, 5, 7}));
+}
+
+TEST(type_prop, adaptive_max_pool_i32_indices)
+{
+    const PartialShape arg_shape{1, 6, 8, 9};
+    const vector<int64_t> output_shape{5, 7};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, output_shape);
+    auto adaptive_pool = make_shared<op::v8::AdaptiveMaxPool>(data, out_shape, element::i32);
+
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(0).same_scheme({1, 6, 5, 7}));
+    ASSERT_EQ(adaptive_pool->output(1).get_element_type(), element::i32);
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(1).same_scheme({1, 6, 5, 7}));
+}
+
+TEST(type_prop, adaptive_max_pool_dyn_batch)
+{
+    const PartialShape arg_shape{Dimension::dynamic(), 6, 8, 9};
+    const vector<int64_t> output_shape{5, 7};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, output_shape);
+    auto adaptive_pool = make_shared<op::v8::AdaptiveMaxPool>(data, out_shape);
+
+    ASSERT_TRUE(
+        adaptive_pool->get_output_partial_shape(0).same_scheme({Dimension::dynamic(), 6, 5, 7}));
+    ASSERT_TRUE(
+        adaptive_pool->get_output_partial_shape(1).same_scheme({Dimension::dynamic(), 6, 5, 7}));
+}
+
+TEST(type_prop, adaptive_max_pool_dyn_channels)
+{
+    const PartialShape arg_shape{1, Dimension::dynamic(), 8, 9};
+    const vector<int64_t> output_shape{5, 7};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, output_shape);
+    auto adaptive_pool = make_shared<op::v8::AdaptiveMaxPool>(data, out_shape);
+
+    ASSERT_TRUE(
+        adaptive_pool->get_output_partial_shape(0).same_scheme({1, Dimension::dynamic(), 5, 7}));
+    ASSERT_TRUE(
+        adaptive_pool->get_output_partial_shape(1).same_scheme({1, Dimension::dynamic(), 5, 7}));
+}
+
+TEST(type_prop, adaptive_max_pool_dyn_spatial)
+{
+    const PartialShape arg_shape{1, 6, Dimension::dynamic(), Dimension::dynamic()};
+    const vector<int64_t> output_shape{5, 7};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, output_shape);
+    auto adaptive_pool = make_shared<op::v8::AdaptiveMaxPool>(data, out_shape);
+
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(0).same_scheme({1, 6, 5, 7}));
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(1).same_scheme({1, 6, 5, 7}));
+}
+
+TEST(type_prop, adaptive_max_pool_dyn_output_shape)
+{
+    const PartialShape arg_shape{1, 6, 8, 9};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = make_shared<op::Parameter>(element::i64, Shape{2});
+    auto adaptive_pool = make_shared<op::v8::AdaptiveMaxPool>(data, out_shape);
+
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(0).same_scheme(
+        {1, 6, Dimension::dynamic(), Dimension::dynamic()}));
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(1).same_scheme(
+        {1, 6, Dimension::dynamic(), Dimension::dynamic()}));
+}
+
+TEST(type_prop, adaptive_max_pool_dyn_rank)
+{
+    const PartialShape arg_shape = PartialShape::dynamic();
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = make_shared<op::Parameter>(element::i64, Shape{2});
+    auto adaptive_pool = make_shared<op::v8::AdaptiveMaxPool>(data, out_shape);
+
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(0).same_scheme(PartialShape::dynamic()));
+    ASSERT_TRUE(adaptive_pool->get_output_partial_shape(1).same_scheme(PartialShape::dynamic()));
+}
+
+TEST(type_prop, adaptive_max_pool_unsupported_input_shape)
+{
+    const PartialShape arg_shape{1, 6};
+    const vector<int64_t> output_shape{1};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{}, output_shape);
+
+    EXPECT_THROW(make_shared<op::v8::AdaptiveMaxPool>(data, out_shape), NodeValidationFailure);
+}
+
+TEST(type_prop, adaptive_max_pool_wrong_out_shape)
+{
+    const PartialShape arg_shape{1, 6, 8, 9};
+    const vector<int64_t> output_shape{5, 7, 8};
+
+    auto data = make_shared<op::Parameter>(element::f32, arg_shape);
+    auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{3}, output_shape);
+
+    EXPECT_THROW(make_shared<op::v8::AdaptiveMaxPool>(data, out_shape), NodeValidationFailure);
+}
diff --git a/ngraph/test/visitors/op/adaptive_avg_pool.cpp b/ngraph/test/visitors/op/adaptive_avg_pool.cpp
new file mode 100644
index 00000000000000..377b955e1f01e0
--- /dev/null
+++ b/ngraph/test/visitors/op/adaptive_avg_pool.cpp
@@ -0,0 +1,27 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+
+#include "ngraph/ngraph.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+#include "ngraph/opsets/opset8.hpp"
+#include "util/visitor.hpp"
+
+using namespace std;
+using namespace ngraph;
+using ngraph::test::NodeBuilder;
+
+TEST(attributes, adaptive_avg_pool_op)
+{
+    NodeBuilder::get_ops().register_factory<opset8::AdaptiveAvgPool>();
+    const auto A = make_shared<op::Parameter>(element::f32, Shape{1, 3, 5, 4});
+    const auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, {4, 3});
+
+    const auto adaptive_pool = make_shared<opset8::AdaptiveAvgPool>(A, out_shape);
+    NodeBuilder builder(adaptive_pool);
+
+    const auto expected_attr_count = 0;
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
+}
diff --git a/ngraph/test/visitors/op/adaptive_max_pool.cpp b/ngraph/test/visitors/op/adaptive_max_pool.cpp
new file mode 100644
index 00000000000000..6210235c7d2342
--- /dev/null
+++ b/ngraph/test/visitors/op/adaptive_max_pool.cpp
@@ -0,0 +1,29 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "gtest/gtest.h"
+
+#include "ngraph/ngraph.hpp"
+#include "ngraph/op/util/attr_types.hpp"
+#include "ngraph/opsets/opset8.hpp"
+#include "util/visitor.hpp"
+
+using namespace std;
+using namespace ngraph;
+using ngraph::test::NodeBuilder;
+
+TEST(attributes, adaptive_max_pool_op)
+{
+    NodeBuilder::get_ops().register_factory<opset8::AdaptiveMaxPool>();
+    const auto A = make_shared<op::Parameter>(element::f32, Shape{1, 3, 5, 4});
+    const auto out_shape = op::Constant::create<int64_t>(element::i64, Shape{2}, {4, 3});
+
+    const auto adaptive_pool = make_shared<opset8::AdaptiveMaxPool>(A, out_shape);
+    NodeBuilder builder(adaptive_pool);
+    auto g_adaptive_pool = as_type_ptr<opset8::AdaptiveMaxPool>(builder.create());
+
+    const auto expected_attr_count = 1;
+    EXPECT_EQ(builder.get_value_map_size(), expected_attr_count);
+    EXPECT_EQ(g_adaptive_pool->get_index_element_type(), adaptive_pool->get_index_element_type());
+}

From 3426a239081f5eebed8b521778b0e90b59280a41 Mon Sep 17 00:00:00 2001
From: Evgenya Stepyreva <evgenya.stepyreva@intel.com>
Date: Tue, 22 Jun 2021 15:04:55 +0300
Subject: [PATCH 160/178] Range for convert precision (#6265)

* ConvertPrecision for opset4::Range

* ConvertPrecision test for Range operation
---
 .../transformations/convert_precision.hpp     |  1 +
 .../transformations/convert_precision.cpp     | 26 +++++++++++++++++++
 ngraph/core/include/ngraph/op/range.hpp       |  3 +++
 ngraph/core/src/pass/convert_precision.cpp    | 19 +++++++++++++-
 4 files changed, 48 insertions(+), 1 deletion(-)

diff --git a/inference-engine/src/transformations/include/transformations/convert_precision.hpp b/inference-engine/src/transformations/include/transformations/convert_precision.hpp
index 56ea4dba7b9391..e7b8eab2fd1124 100644
--- a/inference-engine/src/transformations/include/transformations/convert_precision.hpp
+++ b/inference-engine/src/transformations/include/transformations/convert_precision.hpp
@@ -49,6 +49,7 @@ class NGRAPH_API ConvertPrecision;
  *     opset4::Parameter
  *     opset4::Convert
  *     opset4::ShapeOf
+ *     opset4::Range
  *     opset3::NonMaxSuppression
  *     opset4::NonMaxSuppression
  *     opset4::TopK
diff --git a/inference-engine/tests/functional/inference_engine/transformations/convert_precision.cpp b/inference-engine/tests/functional/inference_engine/transformations/convert_precision.cpp
index a52b677037d56f..0604d062886204 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/convert_precision.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/convert_precision.cpp
@@ -119,6 +119,32 @@ TEST(TransformationTests, ConvertPrecision_ShapeOf) {
     ASSERT_FALSE(has_type<ngraph::element::Type_t::f16>(f));
 }
 
+
+TEST(TransformationTests, ConvertPrecision_Range) {
+    std::shared_ptr<Function> f(nullptr);
+    {
+        auto start = std::make_shared<opset4::Parameter>(element::f16, Shape{});
+        auto stop = std::make_shared<opset4::Parameter>(element::f16, Shape{});
+        auto shift = std::make_shared<opset4::Parameter>(element::f16, Shape{});
+        auto range = std::make_shared<opset4::Range>(start, stop, shift, element::i64);
+
+        f = std::make_shared<Function>(NodeVector{range}, ParameterVector{start, stop, shift});
+
+        pass::Manager manager;
+
+        static const precisions_array precisions = {
+            { ngraph::element::i64, ngraph::element::i32 },
+            { ngraph::element::f16, ngraph::element::f32 }
+        };
+
+        manager.register_pass<ngraph::pass::ConvertPrecision>(precisions);
+        manager.run_passes(f);
+    }
+
+    ASSERT_FALSE(has_type<ngraph::element::Type_t::i64>(f));
+    ASSERT_FALSE(has_type<ngraph::element::Type_t::f16>(f));
+}
+
 TEST(TransformationTests, ConvertPrecision_ConstantRelu) {
     std::shared_ptr<Function> f(nullptr);
     {
diff --git a/ngraph/core/include/ngraph/op/range.hpp b/ngraph/core/include/ngraph/op/range.hpp
index fe09eecdf777f8..a63d6fac7c5e5c 100644
--- a/ngraph/core/include/ngraph/op/range.hpp
+++ b/ngraph/core/include/ngraph/op/range.hpp
@@ -43,6 +43,9 @@ namespace ngraph
                 bool evaluate(const HostTensorVector& outputs,
                               const HostTensorVector& inputs) const override;
                 bool has_evaluate() const override;
+                void set_output_type(element::Type output_type) { m_output_type = output_type; }
+                // Overload collision with method on Node
+                using Node::set_output_type;
 
             private:
                 element::Type m_output_type;
diff --git a/ngraph/core/src/pass/convert_precision.cpp b/ngraph/core/src/pass/convert_precision.cpp
index c5bfe5e95899cb..1efd710aa01bcf 100644
--- a/ngraph/core/src/pass/convert_precision.cpp
+++ b/ngraph/core/src/pass/convert_precision.cpp
@@ -28,6 +28,9 @@ bool fuse_type_to_shapeof(const std::shared_ptr<ngraph::Node>& node,
 bool fuse_type_to_shapeof_v0(const std::shared_ptr<ngraph::Node>& node,
                              ngraph::element::Type to,
                              size_t idx);
+bool fuse_type_to_range_v4(const std::shared_ptr<ngraph::Node>& node,
+                           ngraph::element::Type to,
+                           size_t idx);
 bool fuse_type_to_parameter(const std::shared_ptr<ngraph::Node>& node,
                             ngraph::element::Type to,
                             size_t idx);
@@ -341,7 +344,8 @@ bool ngraph::pass::ConvertPrecision::run_on_function(std::shared_ptr<ngraph::Fun
         {opset4::ReduceLogicalAnd::type_info,
          fuse_type_to_reduce_logical<opset4::ReduceLogicalAnd>},
         {opset4::ReduceLogicalOr::type_info, fuse_type_to_reduce_logical<opset4::ReduceLogicalOr>},
-        {opset1::ShapeOf::type_info, fuse_type_to_shapeof_v0}};
+        {opset1::ShapeOf::type_info, fuse_type_to_shapeof_v0},
+        {opset4::Range::type_info, fuse_type_to_range_v4}};
 
     type_to_fuse.insert(m_additional_type_to_fuse_map.begin(), m_additional_type_to_fuse_map.end());
 
@@ -383,6 +387,19 @@ bool fuse_type_to_shapeof(const std::shared_ptr<ngraph::Node>& node, element::Ty
     return false;
 }
 
+bool fuse_type_to_range_v4(const std::shared_ptr<ngraph::Node>& node, element::Type to, size_t idx)
+{
+    if (auto range = as_type_ptr<opset4::Range>(node))
+    {
+        if (to.is_integral_number() || to.is_real())
+        {
+            range->set_output_type(to);
+            return true;
+        }
+    }
+    return false;
+}
+
 bool fuse_type_to_parameter(const std::shared_ptr<ngraph::Node>& node, element::Type to, size_t idx)
 {
     if (auto param = as_type_ptr<opset4::Parameter>(node))

From eabaa260eea15a8cf898d9017a7f70e8b3369e65 Mon Sep 17 00:00:00 2001
From: Egor Duplensky <egor.duplenskii@intel.com>
Date: Tue, 22 Jun 2021 15:54:24 +0300
Subject: [PATCH 161/178] [Python Tools] Fix cross_check_tool device config
 parsing (#6150)

---
 inference-engine/tools/cross_check_tool/utils.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/inference-engine/tools/cross_check_tool/utils.py b/inference-engine/tools/cross_check_tool/utils.py
index 088f5b7973996b..8d5d778b41cb1a 100644
--- a/inference-engine/tools/cross_check_tool/utils.py
+++ b/inference-engine/tools/cross_check_tool/utils.py
@@ -295,7 +295,7 @@ def get_config_dictionary(config_file):
         return config
     with open(config_file) as f:
         config_line = f.readline()
-        key = config_line.split(config_file)[0]
+        key = config_line.split()[0]
         value = config_line[len(key):].strip()
         config[key] = value
     return config

From fa2f9c5201983d7ce247d1723083ca033f3c672d Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Tue, 22 Jun 2021 17:39:39 +0300
Subject: [PATCH 162/178] Migrated to official GoogleTest repo (#6286)

* Move gtest to <root>/wqthirdparty

* Fixed inference engine tests

* Fixed unit-tests

* Fixed GPU

* Fixed clDNN_unit_tests compilation

* Reverted ngraph changes

* Fixed VPU tests

* Fixed deprecated tests compilation

* Removed OpenVINO clone

* Added public submodule

* Removed

* Fixed Windows

* More updates for INSTANTIATE_TEST_SUITE_P

* Use release-1.10.0

* Removed ngraph copy of gtest

* Revert "Removed ngraph copy of gtest"

This reverts commit ec9fe08d79acf77b0a77f1b255c1b39cc9ce93cb.

* Fixes for TYPED_TEST_CASE

* Fixed GNA tests

* Updated submodule

* Updaed index

* TMP disabled tests

* Revert changes and fix compilation errors

* Fixed caching tests

* Revert back

* Revert back all

Co-authored-by: Michael Nosov <mikhail.nosov@intel.com>
---
 .gitmodules                                   |     8 +-
 docs/IE_PLUGIN_DG/PluginTesting.md            |     2 +-
 .../behavior/caching_tests.cpp                |     2 +-
 .../behavior/config.cpp                       |     8 +-
 .../behavior/core_integration.cpp             |    50 +-
 .../behavior/cpp_holders.cpp                  |     6 +-
 .../behavior/exec_graph_info.cpp              |     2 +-
 .../behavior/infer_request.cpp                |     2 +-
 .../behavior/infer_request_callback.cpp       |     2 +-
 .../behavior/infer_request_config.cpp         |     2 +-
 .../behavior/infer_request_input.cpp          |     2 +-
 .../behavior/infer_request_output.cpp         |     2 +-
 .../behavior/layout.cpp                       |     2 +-
 .../behavior/preprocessing.cpp                |     4 +-
 .../behavior/set_preprocess.cpp               |    10 +-
 .../behavior/test_plugin.cpp                  |     6 +-
 .../behavior/version.cpp                      |     2 +-
 .../hetero/query_network.cpp                  |     2 +-
 .../hetero/synthetic.cpp                      |     4 +-
 .../single_layer_tests/convolution.cpp        |     8 +-
 .../single_layer_tests/reshape.cpp            |     6 +-
 .../single_layer_tests/softmax.cpp            |     4 +-
 .../single_layer_tests/split.cpp              |     2 +-
 .../gna_plugin/optimizer/gna_pass_manager.cpp |     6 +-
 .../inference_engine/blob_copy_test.cpp       |     4 +-
 .../inference_engine/caching_test.cpp         |     6 +-
 .../cnn_network/matmul_sr_tests.cpp           |     2 +-
 .../ir_serialization/serialize.cpp            |     4 +-
 .../lp_transformations/add_transformation.cpp |     2 +-
 .../avg_pool_transformation.cpp               |     2 +-
 .../clamp_transformation.cpp                  |     2 +-
 .../compose_fake_quantize_transformation.cpp  |     2 +-
 ...ction_with_intermediate_transformation.cpp |     2 +-
 .../concat_transformation.cpp                 |     2 +-
 ...t_with_different_precision_on_children.cpp |     2 +-
 ...ate_precision_selection_transformation.cpp |     2 +-
 ...th_intermediate_reshape_transformation.cpp |     2 +-
 ...oncat_with_intermediate_transformation.cpp |     2 +-
 ...ermediate_with_constant_transformation.cpp |     2 +-
 .../concat_with_neighbors_transformation.cpp  |     2 +-
 ...with_reshape_at_the_end_transformation.cpp |     2 +-
 .../concat_with_split_transformation.cpp      |     4 +-
 ...ncat_with_strided_slice_transformation.cpp |     2 +-
 ...lly_transformation_with_dequantization.cpp |     2 +-
 ...nvert_subtract_constant_transformation.cpp |     2 +-
 ...nvolution_backprop_data_transformation.cpp |     2 +-
 .../convolution_qdq_transformation.cpp        |     2 +-
 .../convolution_transformation.cpp            |     2 +-
 .../convolution_with_incorrect_weights.cpp    |     2 +-
 .../depth_to_space_transformation.cpp         |     2 +-
 ...e_convert_on_const_path_transformation.cpp |     2 +-
 ...i_parent_dequantization_transformation.cpp |     2 +-
 ...d_two_output_branches_with_convolution.cpp |     2 +-
 ...tize_on_weights_with_unsupported_child.cpp |     2 +-
 ...ize_precision_selection_transformation.cpp |     2 +-
 .../fake_quantize_transformation.cpp          |     2 +-
 ...ize_with_dq_not_optimal_transformation.cpp |     2 +-
 ..._with_dynamic_intervals_transformation.cpp |     2 +-
 .../fold_convert_transformation.cpp           |     2 +-
 .../fold_fake_quantize_in_transformations.cpp |     2 +-
 .../fuse_convert_transformation.cpp           |     2 +-
 .../fuse_fake_quantize_transformation.cpp     |     2 +-
 ...ntize_with_multi_inputs_transformation.cpp |     2 +-
 ...ltiply_to_fake_quantize_transformation.cpp |     2 +-
 ...btract_to_fake_quantize_transformation.cpp |     2 +-
 ...et_dequantization_below_transformation.cpp |     2 +-
 .../get_dequantization_test.cpp               |     2 +-
 .../get_dequantization_transformation.cpp     |     2 +-
 .../group_convolution_transformation.cpp      |     2 +-
 .../interpolate_transformation.cpp            |     2 +-
 .../is_function_quantized_transformation.cpp  |     2 +-
 .../mat_mul_transformation.cpp                |     2 +-
 .../mat_mul_with_constant_transformation.cpp  |     2 +-
 .../max_pool_transformation.cpp               |     2 +-
 ...ve_dequantization_after_transformation.cpp |     2 +-
 ..._to_scaleshift_or_power_transformation.cpp |     2 +-
 ...ly_to_group_convolution_transformation.cpp |     2 +-
 .../multiply_transformation.cpp               |     2 +-
 .../lp_transformations/mvn_transformation.cpp |     2 +-
 ...ormalize_dequantization_transformation.cpp |     2 +-
 .../normalize_l2_transformation.cpp           |     2 +-
 .../prelu_transformation.cpp                  |     2 +-
 ..._through_dequantization_transformation.cpp |     2 +-
 ..._through_dequantization_transformation.cpp |     2 +-
 .../reduce_max_transformation.cpp             |     2 +-
 .../reduce_mean_transformation.cpp            |     2 +-
 .../reduce_min_transformation.cpp             |     2 +-
 .../reduce_sum_transformation.cpp             |     2 +-
 .../relu_transformation.cpp                   |     2 +-
 ...reshape_fully_connected_transformation.cpp |     2 +-
 .../reshape_transformation.cpp                |     2 +-
 .../round_transformation.cpp                  |     2 +-
 ...te_in_standalone_branch_transformation.cpp |     2 +-
 .../shuffle_channels_transformation.cpp       |     2 +-
 .../split_transformation.cpp                  |     2 +-
 .../squeeze_transformation.cpp                |     2 +-
 .../strided_slice_transformation.cpp          |     2 +-
 ...ultiply_to_multiply_add_transformation.cpp |     2 +-
 ...sformations_after_split_transformation.cpp |     2 +-
 .../transformer_is_function_quantized.cpp     |     2 +-
 .../transpose_transformation.cpp              |     2 +-
 .../unsqueeze_transformation.cpp              |     2 +-
 .../variadic_split_transformation.cpp         |     2 +-
 .../ngraph_reader/net_reader_test.cpp         |     2 +-
 .../serialization/single_layer/activation.cpp |     8 +-
 .../serialization/single_layer/batch_norm.cpp |     2 +-
 .../single_layer/binary_convolution.cpp       |     4 +-
 .../serialization/single_layer/broadcast.cpp  |     2 +-
 .../serialization/single_layer/clamp.cpp      |     2 +-
 .../serialization/single_layer/constant.cpp   |     4 +-
 .../serialization/single_layer/convert.cpp    |     2 +-
 .../single_layer/convolution.cpp              |     4 +-
 .../single_layer/convolution_backprop.cpp     |     2 +-
 .../single_layer/deformable_convolution.cpp   |     4 +-
 .../single_layer/deformable_psroi_pooling.cpp |     2 +-
 .../single_layer/detection_output.cpp         |     2 +-
 .../single_layer/elementwise.cpp              |     2 +-
 .../single_layer/extractimagepatches.cpp      |     2 +-
 .../single_layer/group_convolution.cpp        |     2 +-
 .../group_convolution_backprop_data.cpp       |     2 +-
 .../single_layer/minimum_maximum.cpp          |     2 +-
 .../serialization/single_layer/mvn.cpp        |     4 +-
 .../single_layer/non_max_suppression.cpp      |     2 +-
 .../serialization/single_layer/nonzero.cpp    |     2 +-
 .../single_layer/normalize_l2.cpp             |     2 +-
 .../serialization/single_layer/one_hot.cpp    |     2 +-
 .../serialization/single_layer/pad.cpp        |     2 +-
 .../serialization/single_layer/pooling.cpp    |     4 +-
 .../serialization/single_layer/prelu.cpp      |     2 +-
 .../single_layer/prior_box_clustered.cpp      |     2 +-
 .../serialization/single_layer/reduce_ops.cpp |    10 +-
 .../single_layer/region_yolo.cpp              |     2 +-
 .../serialization/single_layer/reshape.cpp    |     2 +-
 .../serialization/single_layer/result.cpp     |     2 +-
 .../single_layer/scatter_nd_update.cpp        |     2 +-
 .../single_layer/shuffle_channels.cpp         |     2 +-
 .../single_layer/space_to_batch.cpp           |     2 +-
 .../serialization/single_layer/split.cpp      |     2 +-
 .../serialization/single_layer/squeeze.cpp    |     2 +-
 .../single_layer/tensor_iterator.cpp          |     2 +-
 .../serialization/single_layer/transpose.cpp  |     6 +-
 .../serialization/single_layer/unsqueeze.cpp  |     2 +-
 .../single_layer/variadic_split.cpp           |     2 +-
 .../inference_engine/task_executor_tests.cpp  |     4 +-
 .../broadcast_elementwise_fusion_test.cpp     |    10 +-
 .../transformations/conv_fusion_test.cpp      |     8 +-
 .../convert_broadcast3_test.cpp               |    10 +-
 .../convert_convolution_test.cpp              |     2 +-
 .../convert_deconvolution_test.cpp            |     2 +-
 .../convert_reduce_to_pooling_test.cpp        |     6 +-
 .../eliminate_unsqueeze_gather.cpp            |     2 +-
 .../transformations/fq_decomposition_test.cpp |    16 +-
 .../transformations/fq_mul_fusion_test.cpp    |    24 +-
 .../transformations/fq_reshape_fusion.cpp     |     2 +-
 .../mul_add_conversion_test.cpp               |    18 +-
 .../shuffle_channels_fusion_test.cpp          |     2 +-
 .../transformations/softmax_fusion.cpp        |     4 +-
 .../transpose_sinking_test.cpp                |     6 +-
 .../transpose_to_reshape_test.cpp             |     8 +-
 .../weights_dequantize_to_fake_quantize.cpp   |     2 +-
 .../test_runner/src/read_ir/read_ir.cpp       |     2 +-
 .../cpu/bfloat16/bf16_network_restoring.cpp   |     2 +-
 .../plugin/cpu/bfloat16/bfloat16_helpers.hpp  |     2 +-
 .../plugin/cpu/bfloat16/concat_in_place.cpp   |     4 +-
 .../plugin/cpu/bfloat16/conv_add.cpp          |     4 +-
 .../plugin/cpu/bfloat16/conv_conv.cpp         |     4 +-
 .../plugin/cpu/bfloat16/conv_dwconv_relu.cpp  |     4 +-
 .../cpu/bfloat16/conv_eltwise_depthwise.cpp   |     6 +-
 .../conv_relu_pool_conv_relu_pool.cpp         |     4 +-
 .../plugin/cpu/bfloat16/elt_max.cpp           |     4 +-
 .../functional/plugin/cpu/bfloat16/elt_x3.cpp |     4 +-
 .../cpu/bfloat16/faster_100_5_1_1_conv.cpp    |     4 +-
 .../plugin/cpu/bfloat16/gather_multiply.cpp   |     4 +-
 .../gather_x2_add_mul_relu_concat_matmul.cpp  |     4 +-
 .../plugin/cpu/bfloat16/memory_conv.cpp       |     2 +-
 .../bfloat16/mobilenet_ssd_with_branching.cpp |     4 +-
 .../bfloat16/scaleshift_conv_eltwise_conv.cpp |     4 +-
 .../scaleshift_conv_eltwise_relu_conv.cpp     |     4 +-
 .../scaleshift_conv_eltwise_scaleshift.cpp    |     4 +-
 .../cpu/bfloat16/scaleshift_conv_elu_conv.cpp |     4 +-
 .../cpu/bfloat16/scaleshift_conv_relu.cpp     |     4 +-
 .../scaleshift_conv_x2_concat_relu.cpp        |     4 +-
 .../bfloat16/scaleshift_conv_x2_eltwise.cpp   |     4 +-
 .../scaleshift_conv_x2_mixed1_eltwise.cpp     |     4 +-
 .../scaleshift_conv_x2_mixed2_eltwise.cpp     |     4 +-
 .../bfloat16/scaleshift_conv_x3_eltwise.cpp   |     4 +-
 .../scaleshift_x2_conv_x2_eltwise.cpp         |     4 +-
 .../scaleshift_x3_conv_eltwise_relu.cpp       |     4 +-
 .../cpu/bfloat16/tail_fp32_optimization.cpp   |     4 +-
 .../plugin/cpu/bfloat16/topk_inputs_i32.cpp   |     4 +-
 .../behavior/add_output.cpp                   |     2 +-
 .../behavior/caching_tests.cpp                |     2 +-
 .../behavior/config.cpp                       |    22 +-
 .../behavior/core_integration.cpp             |    54 +-
 .../behavior/core_threading_tests.cpp         |     6 +-
 .../behavior/cpp_holders.cpp                  |     2 +-
 .../behavior/exec_graph_info.cpp              |     6 +-
 .../behavior/infer_request.cpp                |     6 +-
 .../behavior/infer_request_callback.cpp       |     6 +-
 .../behavior/infer_request_cancellation.cpp   |     2 +-
 .../behavior/infer_request_config.cpp         |    12 +-
 .../behavior/infer_request_input.cpp          |     6 +-
 .../behavior/infer_request_output.cpp         |     6 +-
 .../behavior/invalid_cases/proposal.cpp       |     2 +-
 .../behavior/layout.cpp                       |     2 +-
 .../behavior/perf_counters.cpp                |     4 +-
 .../behavior/preprocessing.cpp                |     4 +-
 .../behavior/set_blob.cpp                     |     2 +-
 .../behavior/set_blob_of_kind.cpp             |     8 +-
 .../behavior/set_preprocess.cpp               |    24 +-
 .../behavior/test_plugin.cpp                  |    18 +-
 .../behavior/version.cpp                      |     8 +-
 .../configuration_tests.cpp                   |     2 +-
 .../configuration_tests/dynamic_batch.cpp     |     4 +-
 .../exec_graph_serialization.cpp              |     2 +-
 .../execution_graph_tests/keep_assign.cpp     |     2 +-
 .../nms_transformation_for_last_node.cpp      |     2 +-
 .../num_inputs_fusing_bin_conv.cpp            |     2 +-
 .../remove_parameter.cpp                      |     2 +-
 .../runtime_precision.cpp                     |     2 +-
 .../unique_node_names.cpp                     |     2 +-
 .../hetero/query_network.cpp                  |     2 +-
 .../hetero/synthetic.cpp                      |     4 +-
 .../add_transformation.cpp                    |     2 +-
 .../clamp_transformation.cpp                  |     2 +-
 .../concat_transformation.cpp                 |     2 +-
 .../concat_with_child_and_output.cpp          |     2 +-
 ...t_with_different_precision_on_children.cpp |     2 +-
 ...oncat_with_intermediate_transformation.cpp |     2 +-
 .../concat_with_neighbors_transformation.cpp  |     2 +-
 .../concat_with_split_transformation.cpp      |     2 +-
 ...nvolution_backprop_data_transformation.cpp |     2 +-
 .../convolution_qdq_transformation.cpp        |     2 +-
 .../convolution_transformation.cpp            |     4 +-
 .../depth_to_space_transformation.cpp         |     4 +-
 ...e_quantize_and_avg_pool_transformation.cpp |     2 +-
 ...e_quantize_and_max_pool_transformation.cpp |     2 +-
 ...d_two_output_branches_with_convolution.cpp |     2 +-
 ...ize_precision_selection_transformation.cpp |     2 +-
 .../fake_quantize_transformation.cpp          |     2 +-
 ...ize_with_dq_not_optimal_transformation.cpp |     2 +-
 .../fully_connected_transformation.cpp        |     2 +-
 .../fuse_convert_transformation.cpp           |     2 +-
 ...uantize_and_scale_shift_transformation.cpp |     2 +-
 .../fuse_fake_quantize_transformation.cpp     |     2 +-
 ...ltiply_to_fake_quantize_transformation.cpp |     2 +-
 ...btract_to_fake_quantize_transformation.cpp |     2 +-
 .../gemm_transformation.cpp                   |     2 +-
 .../group_convolution_transformation.cpp      |     2 +-
 .../groupconvolution_qdq_transformation.cpp   |     2 +-
 .../interpolate_transformation.cpp            |     2 +-
 .../mat_mul_transformation.cpp                |     2 +-
 .../mat_mul_with_constant_transformation.cpp  |     2 +-
 ..._constant_fake_quantize_transformation.cpp |     2 +-
 ...ly_to_group_convolution_transformation.cpp |     2 +-
 .../multiply_transformation.cpp               |     2 +-
 ...ultiply_with_one_parent_transformation.cpp |     2 +-
 .../mvn_transformation.cpp                    |     2 +-
 .../normalize_transformation.cpp              |     2 +-
 ...put_layers_handling_in_transformations.cpp |     2 +-
 ...handling_in_transformations_for_concat.cpp |     2 +-
 ...ansformations_for_concat_multi_channel.cpp |     2 +-
 .../prelu_transformation.cpp                  |     2 +-
 ..._through_dequantization_transformation.cpp |     2 +-
 .../reduce_max_transformation.cpp             |     2 +-
 .../reduce_mean_transformation.cpp            |     2 +-
 .../reduce_min_transformation.cpp             |     2 +-
 .../reduce_sum_transformation.cpp             |     2 +-
 .../relu_transformation.cpp                   |     2 +-
 .../reshape_transformation.cpp                |     2 +-
 .../shuffle_channels_transformation.cpp       |     2 +-
 .../split_transformation.cpp                  |     2 +-
 .../squeeze_transformation.cpp                |     2 +-
 .../strided_slice_transformation.cpp          |     2 +-
 ...ultiply_to_multiply_add_transformation.cpp |     2 +-
 .../subtract_transformation.cpp               |     2 +-
 .../transpose_after_matmul_transformation.cpp |     2 +-
 .../transpose_transformation.cpp              |     2 +-
 .../unsqueeze_transformation.cpp              |     2 +-
 .../variadic_split_transformation.cpp         |     2 +-
 .../multi/cpu_remote_blob_tests.cpp           |     2 +-
 .../conv_bias_fusion.cpp                      |     2 +-
 .../plugin_specific_ngraph_conversion.cpp     |     2 +-
 .../single_layer_tests/activation.cpp         |    10 +-
 .../single_layer_tests/batch_norm.cpp         |     2 +-
 .../single_layer_tests/batch_to_space.cpp     |     2 +-
 .../single_layer_tests/binary_convolution.cpp |     4 +-
 .../single_layer_tests/broadcast.cpp          |    14 +-
 .../single_layer_tests/bucketize.cpp          |     4 +-
 .../single_layer_tests/clamp.cpp              |     4 +-
 .../single_layer_tests/comparison.cpp         |     2 +-
 .../single_layer_tests/concat.cpp             |     2 +-
 .../single_layer_tests/constant.cpp           |     4 +-
 .../single_layer_tests/convert.cpp            |     2 +-
 .../single_layer_tests/convert_like.cpp       |     2 +-
 .../single_layer_tests/convolution.cpp        |    14 +-
 .../convolution_backprop_data.cpp             |    20 +-
 .../single_layer_tests/ctc_greedy_decoder.cpp |     2 +-
 .../ctc_greedy_decoder_seq_len.cpp            |     4 +-
 .../single_layer_tests/ctc_loss.cpp           |     4 +-
 .../single_layer_tests/cum_sum.cpp            |    16 +-
 .../deformable_convolution.cpp                |     6 +-
 .../deformable_psroi_pooling.cpp              |     4 +-
 .../single_layer_tests/depth_to_space.cpp     |     4 +-
 .../single_layer_tests/detection_output.cpp   |     4 +-
 .../single_layer_tests/dft.cpp                |     8 +-
 .../single_layer_tests/eltwise.cpp            |     4 +-
 .../embedding_bag_offsets_sum.cpp             |     2 +-
 .../embedding_bag_packed_sum.cpp              |     2 +-
 .../embedding_segments_sum.cpp                |     2 +-
 .../extract_image_patches.cpp                 |     2 +-
 .../single_layer_tests/fake_quantize.cpp      |     8 +-
 .../single_layer_tests/gather.cpp             |    12 +-
 .../single_layer_tests/gather_elements.cpp    |    10 +-
 .../single_layer_tests/gather_nd.cpp          |     4 +-
 .../single_layer_tests/gather_tree.cpp        |     2 +-
 .../single_layer_tests/group_convolution.cpp  |    12 +-
 .../group_convolution_backprop_data.cpp       |    12 +-
 .../single_layer_tests/gru_cell.cpp           |     2 +-
 .../single_layer_tests/gru_sequence.cpp       |     4 +-
 .../single_layer_tests/interpolate.cpp        |     8 +-
 .../single_layer_tests/log_softmax.cpp        |     4 +-
 .../single_layer_tests/logical.cpp            |     4 +-
 .../single_layer_tests/loop.cpp               |     6 +-
 .../single_layer_tests/lrn.cpp                |     8 +-
 .../single_layer_tests/lstm_cell.cpp          |     2 +-
 .../single_layer_tests/lstm_sequence.cpp      |     4 +-
 .../single_layer_tests/mat_mul.cpp            |     2 +-
 .../single_layer_tests/memory.cpp             |     2 +-
 .../single_layer_tests/minimum_maximum.cpp    |     2 +-
 .../single_layer_tests/mvn.cpp                |    18 +-
 .../non_max_suppression.cpp                   |     2 +-
 .../single_layer_tests/normalize_l2.cpp       |     2 +-
 .../single_layer_tests/one_hot.cpp            |     6 +-
 .../single_layer_tests/pad.cpp                |    12 +-
 .../single_layer_tests/pooling.cpp            |    28 +-
 .../single_layer_tests/proposal.cpp           |     2 +-
 .../single_layer_tests/psroi_pooling.cpp      |     4 +-
 .../single_layer_tests/range.cpp              |     2 +-
 .../single_layer_tests/reduce_ops.cpp         |    16 +-
 .../single_layer_tests/region_yolo.cpp        |     6 +-
 .../single_layer_tests/reorg_yolo.cpp         |    12 +-
 .../single_layer_tests/reshape.cpp            |     6 +-
 .../single_layer_tests/result.cpp             |     2 +-
 .../single_layer_tests/reverse_sequence.cpp   |     2 +-
 .../single_layer_tests/rnn_cell.cpp           |     2 +-
 .../single_layer_tests/rnn_sequence.cpp       |     4 +-
 .../single_layer_tests/roi_align.cpp          |     4 +-
 .../single_layer_tests/roi_pooling.cpp        |     4 +-
 .../single_layer_tests/roll.cpp               |    16 +-
 .../single_layer_tests/scatter_ND_update.cpp  |     2 +-
 .../scatter_elements_update.cpp               |     2 +-
 .../single_layer_tests/scatter_update.cpp     |     2 +-
 .../single_layer_tests/select.cpp             |     4 +-
 .../single_layer_tests/shape_of.cpp           |     2 +-
 .../single_layer_tests/shuffle_channels.cpp   |    12 +-
 .../single_layer_tests/softmax.cpp            |     4 +-
 .../single_layer_tests/space_to_batch.cpp     |     4 +-
 .../single_layer_tests/space_to_depth.cpp     |     4 +-
 .../single_layer_tests/split.cpp              |     4 +-
 .../single_layer_tests/squeeze_unsqueeze.cpp  |     4 +-
 .../single_layer_tests/strided_slice.cpp      |     2 +-
 .../single_layer_tests/tensor_iterator.cpp    |     4 +-
 .../single_layer_tests/tile.cpp               |     4 +-
 .../single_layer_tests/topk.cpp               |     2 +-
 .../single_layer_tests/transpose.cpp          |     8 +-
 .../single_layer_tests/variadic_split.cpp     |     2 +-
 .../subgraph_tests/codegen_add.cpp            |     2 +-
 .../subgraph_tests/codegen_bert.cpp           |     2 +-
 .../subgraph_tests/codegen_gelu.cpp           |     2 +-
 .../subgraph_tests/constant_result.cpp        |     2 +-
 .../subgraph_tests/conv_eltwise_fusion.cpp    |    24 +-
 .../convert_pad_to_group_conv.cpp             |     4 +-
 .../get_output_before_activation.cpp          |     2 +-
 .../subgraph_tests/matmul_squeeze_add.cpp     |     2 +-
 .../subgraph_tests/memory_LSTMCell.cpp        |     2 +-
 .../subgraph_tests/multiple_LSTMCell.cpp      |     2 +-
 .../subgraph_tests/multiply_add.cpp           |     2 +-
 .../subgraph_tests/mvn_multiply_add.cpp       |     6 +-
 .../subgraph_tests/parameter_result.cpp       |     2 +-
 .../subgraph_tests/perm_conv_perm_concat.cpp  |     2 +-
 .../quantized_convolution_backprop_data.cpp   |     4 +-
 .../quantized_group_convolution.cpp           |     4 +-
 ...ntized_group_convolution_backprop_data.cpp |     4 +-
 .../subgraph_tests/quantized_mat_mul.cpp      |     6 +-
 .../subgraph_tests/range_add.cpp              |    10 +-
 .../subgraph_tests/relu_shape_of.cpp          |     2 +-
 .../reshape_squeeze_reshape_relu.cpp          |     2 +-
 .../subgraph_tests/split_concat_memory.cpp    |     2 +-
 .../subgraph_tests/split_conv_concat.cpp      |     2 +-
 .../subgraph_tests/tensor_names.cpp           |     2 +-
 .../subgraph_tests/variadic_split_pad.cpp     |     2 +-
 .../cpu/single_layer_tests/activation.cpp     |     4 +-
 .../cpu/single_layer_tests/batch_to_space.cpp |     8 +-
 .../plugin/cpu/single_layer_tests/concat.cpp  |    22 +-
 .../plugin/cpu/single_layer_tests/convert.cpp |     4 +-
 .../convert_to_plugin_specific_node.cpp       |     4 +-
 .../cpu/single_layer_tests/convolution.cpp    |    46 +-
 .../convolution_backprop_data.cpp             |    20 +-
 .../cpu/single_layer_tests/depth_to_space.cpp |     8 +-
 .../plugin/cpu/single_layer_tests/eltwise.cpp |    20 +-
 .../extract_image_patches.cpp                 |     2 +-
 .../cpu/single_layer_tests/fake_quantize.cpp  |    10 +-
 .../single_layer_tests/gather_elements.cpp    |     2 +-
 .../single_layer_tests/group_convolution.cpp  |    38 +-
 .../group_convolution_backprop_data.cpp       |    20 +-
 .../cpu/single_layer_tests/gru_cell.cpp       |     2 +-
 .../cpu/single_layer_tests/gru_sequence.cpp   |     4 +-
 .../cpu/single_layer_tests/interpolate.cpp    |    12 +-
 .../plugin/cpu/single_layer_tests/logical.cpp |     4 +-
 .../cpu/single_layer_tests/lstm_cell.cpp      |     2 +-
 .../cpu/single_layer_tests/lstm_sequence.cpp  |     4 +-
 .../plugin/cpu/single_layer_tests/mat_mul.cpp |     6 +-
 .../plugin/cpu/single_layer_tests/mvn.cpp     |    12 +-
 .../cpu/single_layer_tests/normalize.cpp      |     6 +-
 .../plugin/cpu/single_layer_tests/one_hot.cpp |    10 +-
 .../plugin/cpu/single_layer_tests/pad.cpp     |    16 +-
 .../plugin/cpu/single_layer_tests/pooling.cpp |    16 +-
 .../cpu/single_layer_tests/psroi_pooling.cpp  |     4 +-
 .../cpu/single_layer_tests/reduce_ops.cpp     |    16 +-
 .../cpu/single_layer_tests/region_yolo.cpp    |     6 +-
 .../cpu/single_layer_tests/rnn_cell.cpp       |     2 +-
 .../cpu/single_layer_tests/rnn_sequence.cpp   |     4 +-
 .../cpu/single_layer_tests/roi_pooling.cpp    |     4 +-
 .../cpu/single_layer_tests/roialign.cpp       |     2 +-
 .../single_layer_tests/shuffle_channels.cpp   |     8 +-
 .../plugin/cpu/single_layer_tests/softmax.cpp |     4 +-
 .../cpu/single_layer_tests/space_to_batch.cpp |     6 +-
 .../cpu/single_layer_tests/space_to_depth.cpp |     8 +-
 .../plugin/cpu/single_layer_tests/split.cpp   |    26 +-
 .../cpu/single_layer_tests/strided_slice.cpp  |    10 +-
 .../cpu/single_layer_tests/transpose.cpp      |     8 +-
 .../cpu/subgraph_tests/src/conv3d_reshape.cpp |     2 +-
 .../cpu/subgraph_tests/src/conv_concat.cpp    |    26 +-
 .../subgraph_tests/src/conv_maxpool_activ.cpp |     2 +-
 .../cpu/subgraph_tests/src/eltwise_chain.cpp  |     4 +-
 .../src/fuse_muladd_ewsimple.cpp              |     6 +-
 .../src/fuse_scaleshift_and_fakequantize.cpp  |     2 +-
 .../src/fuse_transpose_reorder.cpp            |     6 +-
 .../src/param_result_custom_blob.cpp          |     4 +-
 .../cpu/subgraph_tests/src/reshape_fc.cpp     |     2 +-
 ...shape_permute_conv_permute_reshape_act.cpp |     2 +-
 .../import_export_act_conv_act.cpp            |     2 +-
 .../import_export_memory_layer.cpp            |     2 +-
 .../plugin/gna/pass_tests/4d_eltwise.cpp      |     4 +-
 .../gna/pass_tests/act_maxpool_reordering.cpp |     2 +-
 .../pass_tests/broadcast_const_with_fq.cpp    |     2 +-
 .../convert_matmul_to_fullyconnected.cpp      |     2 +-
 .../convert_matmul_to_pointwise_conv.cpp      |     4 +-
 .../eltwise_split_over_channels_pass.cpp      |     2 +-
 .../plugin/gna/pass_tests/fq_activation.cpp   |     2 +-
 .../fq_fusion_with_multiple_weights.cpp       |     2 +-
 .../gna/pass_tests/fq_maxpool_reordering.cpp  |     2 +-
 .../insert_copy_layer_before_self_concat.cpp  |     2 +-
 .../insert_transpose_before_matmul.cpp        |     2 +-
 .../insert_transpose_between_convs.cpp        |     4 +-
 .../gna/pass_tests/layers_restrictions.cpp    |     4 +-
 .../remove_permutations_NHWC_to_NCHW_pass.cpp |    10 +-
 .../behavior/add_output.cpp                   |     2 +-
 .../behavior/caching_tests.cpp                |     2 +-
 .../behavior/config.cpp                       |     6 +-
 .../behavior/core_integration.cpp             |    56 +-
 .../behavior/core_threading_tests.cpp         |     4 +-
 .../behavior/cpp_holders.cpp                  |     6 +-
 .../behavior/exec_graph_info.cpp              |     2 +-
 .../behavior/infer_request.cpp                |     4 +-
 .../behavior/infer_request_callback.cpp       |     4 +-
 .../behavior/infer_request_config.cpp         |     4 +-
 .../behavior/infer_request_input.cpp          |     2 +-
 .../behavior/infer_request_output.cpp         |     2 +-
 .../behavior/layout.cpp                       |     2 +-
 .../behavior/memory_states.cpp                |     2 +-
 .../behavior/test_plugin.cpp                  |     6 +-
 .../behavior/version.cpp                      |     4 +-
 .../import_reshape_permute_conv.cpp           |     2 +-
 .../single_layer_tests/activation.cpp         |     2 +-
 .../single_layer_tests/concat.cpp             |     2 +-
 .../single_layer_tests/convolution.cpp        |    12 +-
 .../convolution_negative.cpp                  |     2 +-
 .../single_layer_tests/eltwise.cpp            |     2 +-
 .../single_layer_tests/fake_quantize.cpp      |     2 +-
 .../single_layer_tests/low_precision.cpp      |     2 +-
 .../single_layer_tests/mat_mul.cpp            |     2 +-
 .../single_layer_tests/memory.cpp             |     2 +-
 .../single_layer_tests/pooling.cpp            |    10 +-
 .../single_layer_tests/power.cpp              |     2 +-
 .../single_layer_tests/split.cpp              |     2 +-
 .../single_layer_tests/tensor_iterator.cpp    |     2 +-
 .../activation_concats_eltwise.cpp            |     2 +-
 .../subgraph_tests/activation_fq.cpp          |     2 +-
 .../subgraph_tests/basic_lstm.cpp             |     2 +-
 .../subgraph_tests/broadcast_power.cpp        |     2 +-
 .../subgraph_tests/cascade_concat.cpp         |     4 +-
 .../subgraph_tests/clamp_fq.cpp               |     2 +-
 .../subgraph_tests/concat_conv.cpp            |     2 +-
 .../subgraph_tests/concat_multi_input.cpp     |     2 +-
 .../subgraph_tests/concat_quantization.cpp    |     2 +-
 ...ntization_during_memory_requantization.cpp |     2 +-
 .../subgraph_tests/connect_split_concat.cpp   |     4 +-
 .../subgraph_tests/const_conv_concat.cpp      |     2 +-
 .../const_strided_slice_concat.cpp            |     2 +-
 .../subgraph_tests/constant_result.cpp        |     2 +-
 .../subgraph_tests/conv_fq_eltwise.cpp        |     2 +-
 .../subgraph_tests/conv_fq_relu.cpp           |     2 +-
 .../convolution_relu_sequence.cpp             |     2 +-
 .../subgraph_tests/copy_before_squeeze.cpp    |     2 +-
 .../subgraph_tests/crop4d.cpp                 |     2 +-
 .../subgraph_tests/delayed_copy_layer.cpp     |     2 +-
 .../subgraph_tests/eltwise_conv_eltwise.cpp   |     6 +-
 .../eltwise_reshape_activation.cpp            |     2 +-
 .../subgraph_tests/fc_conv_fc.cpp             |     6 +-
 .../first_connect_input_concat.cpp            |     2 +-
 .../subgraph_tests/fq_conv_fq_affine.cpp      |     2 +-
 .../get_output_before_activation.cpp          |     2 +-
 .../handling_orientation_conv.cpp             |     2 +-
 .../subgraph_tests/input_conv.cpp             |     2 +-
 .../subgraph_tests/input_split_concat.cpp     |     2 +-
 .../subgraph_tests/matmul_squeeze_add.cpp     |     2 +-
 .../subgraph_tests/memory_LSTMCell.cpp        |     2 +-
 .../memory_eltwise_reshape_concat.cpp         |     2 +-
 .../multioutput_eltwise_squeeze_eltwise.cpp   |     2 +-
 .../subgraph_tests/multiple_LSTMCell.cpp      |     2 +-
 .../subgraph_tests/multiple_concat.cpp        |     8 +-
 .../negative_memory_layer_offset.cpp          |     2 +-
 .../subgraph_tests/perm_conv_perm_concat.cpp  |     2 +-
 ...shape_permute_conv_permute_reshape_act.cpp |     2 +-
 .../reshape_squeeze_reshape_relu.cpp          |     2 +-
 .../reshapre_permute_reshape.cpp              |     2 +-
 .../subgraph_tests/scale_shift.cpp            |     2 +-
 .../scaleshift_conv_scaleshift.cpp            |     4 +-
 .../subgraph_tests/softsign.cpp               |     2 +-
 .../subgraph_tests/split_conv.cpp             |     2 +-
 .../subgraph_tests/split_conv_concat.cpp      |     2 +-
 .../subgraph_tests/split_relu.cpp             |     2 +-
 .../split_trivial_permute_concat.cpp          |     2 +-
 .../subgraph_tests/stridedslice_conv.cpp      |     2 +-
 .../subgraph_tests/tensor_names.cpp           |     2 +-
 .../subgraph_tests/trivial_concat.cpp         |     2 +-
 .../two_fake_quantize_to_fullyconnected.cpp   |     6 +-
 .../cldnn_remote_blob_tests.cpp               |     4 +-
 .../behavior/caching_tests.cpp                |     2 +-
 .../behavior/config.cpp                       |    24 +-
 .../behavior/core_integration.cpp             |    64 +-
 .../behavior/core_threading_tests.cpp         |     4 +-
 .../behavior/cpp_holders.cpp                  |     2 +-
 .../behavior/exec_graph_info.cpp              |     2 +-
 .../behavior/infer_request.cpp                |     8 +-
 .../behavior/infer_request_callback.cpp       |     8 +-
 .../behavior/infer_request_config.cpp         |     8 +-
 .../behavior/infer_request_input.cpp          |     8 +-
 .../behavior/infer_request_output.cpp         |     8 +-
 .../behavior/layout.cpp                       |     2 +-
 .../behavior/perf_counters.cpp                |     4 +-
 .../behavior/set_blob_of_kind.cpp             |     2 +-
 .../behavior/set_preprocess.cpp               |    12 +-
 .../behavior/test_plugin.cpp                  |    24 +-
 .../behavior/version.cpp                      |    10 +-
 .../configuration_tests/dynamic_batch.cpp     |     4 +-
 .../unique_node_names.cpp                     |     2 +-
 .../hetero/query_network.cpp                  |     2 +-
 .../add_transformation.cpp                    |     2 +-
 .../clamp_transformation.cpp                  |     2 +-
 .../concat_transformation.cpp                 |     2 +-
 .../concat_with_child_and_output.cpp          |     2 +-
 ...t_with_different_precision_on_children.cpp |     2 +-
 ...oncat_with_intermediate_transformation.cpp |     2 +-
 ...at_with_neighbors_graph_transformation.cpp |     2 +-
 .../concat_with_split_transformation.cpp      |     2 +-
 ...nvolution_backprop_data_transformation.cpp |     2 +-
 .../convolution_qdq_transformation.cpp        |     2 +-
 .../convolution_transformation.cpp            |     4 +-
 .../depth_to_space_transformation.cpp         |     4 +-
 ...e_quantize_and_avg_pool_transformation.cpp |     2 +-
 ...e_quantize_and_max_pool_transformation.cpp |     2 +-
 ...d_two_output_branches_with_convolution.cpp |     2 +-
 ...ize_precision_selection_transformation.cpp |     2 +-
 .../fake_quantize_transformation.cpp          |     2 +-
 ...ize_with_dq_not_optimal_transformation.cpp |     2 +-
 .../fully_connected_transformation.cpp        |     2 +-
 .../fuse_convert_transformation.cpp           |     2 +-
 ...uantize_and_scale_shift_transformation.cpp |     2 +-
 .../fuse_fake_quantize_transformation.cpp     |     2 +-
 ...ltiply_to_fake_quantize_transformation.cpp |     2 +-
 ...btract_to_fake_quantize_transformation.cpp |     2 +-
 .../gemm_transformation.cpp                   |     2 +-
 .../group_convolution_transformation.cpp      |     2 +-
 .../groupconvolution_qdq_transformation.cpp   |     2 +-
 .../interpolate_transformation.cpp            |     2 +-
 .../mat_mul_transformation.cpp                |     2 +-
 .../mat_mul_with_constant_transformation.cpp  |     2 +-
 ..._constant_fake_quantize_transformation.cpp |     2 +-
 ...ly_to_group_convolution_transformation.cpp |     2 +-
 .../multiply_transformation.cpp               |     2 +-
 ...ultiply_with_one_parent_transformation.cpp |     2 +-
 .../mvn_transformation.cpp                    |     2 +-
 .../normalize_transformation.cpp              |     2 +-
 ...put_layers_handling_in_transformations.cpp |     2 +-
 ...handling_in_transformations_for_concat.cpp |     2 +-
 ...ansformations_for_concat_multi_channel.cpp |     2 +-
 .../prelu_transformation.cpp                  |     2 +-
 ..._through_dequantization_transformation.cpp |     2 +-
 .../reduce_max_transformation.cpp             |     2 +-
 .../reduce_mean_transformation.cpp            |     2 +-
 .../reduce_min_transformation.cpp             |     2 +-
 .../reduce_sum_transformation.cpp             |     2 +-
 .../relu_transformation.cpp                   |     2 +-
 .../reshape_transformation.cpp                |     2 +-
 .../shuffle_channels_transformation.cpp       |     2 +-
 .../split_transformation.cpp                  |     2 +-
 .../squeeze_transformation.cpp                |     2 +-
 .../strided_slice_transformation.cpp          |     2 +-
 ...ultiply_to_multiply_add_transformation.cpp |     2 +-
 .../subtract_transformation.cpp               |     2 +-
 .../transpose_after_matmul_transformation.cpp |     2 +-
 .../transpose_transformation.cpp              |     2 +-
 .../unsqueeze_transformation.cpp              |     2 +-
 .../variadic_split_transformation.cpp         |     2 +-
 .../multi/gpu_remote_blob_tests.cpp           |     4 +-
 .../conv_bias_fusion.cpp                      |     2 +-
 .../plugin_specific_ngraph_conversion.cpp     |     2 +-
 .../single_layer_tests/activation.cpp         |     6 +-
 .../single_layer_tests/batch_norm.cpp         |     2 +-
 .../single_layer_tests/batch_to_space.cpp     |     2 +-
 .../single_layer_tests/broadcast.cpp          |    14 +-
 .../single_layer_tests/comparison.cpp         |     2 +-
 .../single_layer_tests/concat.cpp             |     2 +-
 .../single_layer_tests/convert.cpp            |     2 +-
 .../single_layer_tests/convert_like.cpp       |     2 +-
 .../single_layer_tests/convolution.cpp        |     6 +-
 .../convolution_backprop_data.cpp             |    16 +-
 .../single_layer_tests/ctc_greedy_decoder.cpp |     2 +-
 .../ctc_greedy_decoder_seq_len.cpp            |     4 +-
 .../single_layer_tests/cum_sum.cpp            |     4 +-
 .../single_layer_tests/depth_to_space.cpp     |     4 +-
 .../single_layer_tests/detection_output.cpp   |     4 +-
 .../single_layer_tests/eltwise.cpp            |     2 +-
 .../embedding_bag_offsets_sum.cpp             |     2 +-
 .../embedding_bag_packed_sum.cpp              |     2 +-
 .../embedding_segments_sum.cpp                |     2 +-
 .../extract_image_patches.cpp                 |     2 +-
 .../single_layer_tests/fake_quantize.cpp      |     2 +-
 .../single_layer_tests/gather.cpp             |    28 +-
 .../single_layer_tests/gather_nd.cpp          |     6 +-
 .../single_layer_tests/gather_tree.cpp        |     2 +-
 .../single_layer_tests/grn.cpp                |     2 +-
 .../single_layer_tests/group_convolution.cpp  |     8 +-
 .../group_convolution_backprop_data.cpp       |     8 +-
 .../single_layer_tests/gru_cell.cpp           |     2 +-
 .../single_layer_tests/gru_sequence.cpp       |     4 +-
 .../single_layer_tests/interpolate.cpp        |     4 +-
 .../single_layer_tests/log_softmax.cpp        |     4 +-
 .../single_layer_tests/logical.cpp            |     4 +-
 .../single_layer_tests/lrn.cpp                |     2 +-
 .../single_layer_tests/lstm_cell.cpp          |     4 +-
 .../single_layer_tests/lstm_sequence.cpp      |     6 +-
 .../single_layer_tests/mat_mul.cpp            |     2 +-
 .../single_layer_tests/minimum_maximum.cpp    |     2 +-
 .../single_layer_tests/mvn.cpp                |    18 +-
 .../non_max_suppression.cpp                   |     2 +-
 .../single_layer_tests/normalize_l2.cpp       |     2 +-
 .../single_layer_tests/pad.cpp                |     8 +-
 .../single_layer_tests/pooling.cpp            |    10 +-
 .../single_layer_tests/power.cpp              |     2 +-
 .../prior_box_clustered.cpp                   |     2 +-
 .../single_layer_tests/proposal.cpp           |     2 +-
 .../single_layer_tests/reduce_ops.cpp         |    16 +-
 .../single_layer_tests/region_yolo.cpp        |     6 +-
 .../single_layer_tests/reorg_yolo.cpp         |    12 +-
 .../single_layer_tests/reshape.cpp            |     6 +-
 .../single_layer_tests/reverse_sequence.cpp   |     2 +-
 .../single_layer_tests/rnn_cell.cpp           |     2 +-
 .../single_layer_tests/rnn_sequence.cpp       |     4 +-
 .../scatter_elements_update.cpp               |     2 +-
 .../single_layer_tests/scatter_nd_update.cpp  |     2 +-
 .../single_layer_tests/scatter_update.cpp     |     2 +-
 .../single_layer_tests/select.cpp             |     4 +-
 .../single_layer_tests/shuffle_channels.cpp   |     8 +-
 .../single_layer_tests/softmax.cpp            |     4 +-
 .../single_layer_tests/space_to_batch.cpp     |     2 +-
 .../single_layer_tests/space_to_depth.cpp     |     4 +-
 .../single_layer_tests/split.cpp              |     4 +-
 .../single_layer_tests/squeeze_unsqueeze.cpp  |     4 +-
 .../single_layer_tests/strided_slice.cpp      |     4 +-
 .../single_layer_tests/tile.cpp               |     4 +-
 .../single_layer_tests/topk.cpp               |     2 +-
 .../single_layer_tests/transpose.cpp          |     6 +-
 .../single_layer_tests/variadic_split.cpp     |     2 +-
 .../subgraph_tests/constant_result.cpp        |     2 +-
 .../get_output_before_activation.cpp          |     2 +-
 .../subgraph_tests/matmul_squeeze_add.cpp     |     2 +-
 .../subgraph_tests/multiply_add.cpp           |     2 +-
 .../subgraph_tests/parameter_result.cpp       |     2 +-
 .../subgraph_tests/perm_conv_perm_concat.cpp  |     2 +-
 .../quantized_convolution_backprop_data.cpp   |     4 +-
 ...ntized_group_convolution_backprop_data.cpp |     4 +-
 .../subgraph_tests/reduce_eltwise.cpp         |     6 +-
 ...shape_permute_conv_permute_reshape_act.cpp |     2 +-
 .../reshape_squeeze_reshape_relu.cpp          |     2 +-
 .../subgraph_tests/scale_shift.cpp            |     2 +-
 .../subgraph_tests/split_conv_concat.cpp      |     2 +-
 .../subgraph_tests/tensor_names.cpp           |     2 +-
 .../single_layer_tests/tensor_iterator.cpp    |     2 +-
 .../operations/dynamic_shape_resolver.cpp     |    12 +-
 .../operations/out_shape_of_reshape.cpp       |    10 +-
 .../operations/static_shape_broadcast.cpp     |    14 +-
 .../operations/static_shape_nonzero.cpp       |     8 +-
 ...xtract_image_patches_to_reorg_yolo_vpu.cpp |    10 +-
 ...mic_to_static_shape_binary_elementwise.cpp |     2 +-
 .../dynamic_to_static_shape_broadcast.cpp     |     4 +-
 .../dynamic_to_static_shape_clamp.cpp         |     2 +-
 .../dynamic_to_static_shape_concat.cpp        |     2 +-
 .../dynamic_to_static_shape_convert.cpp       |     2 +-
 .../dynamic_to_static_shape_gather.cpp        |     6 +-
 ...ynamic_to_static_shape_gather_elements.cpp |     2 +-
 .../dynamic_to_static_shape_gather_nd.cpp     |     2 +-
 .../dynamic_to_static_shape_matmul.cpp        |     2 +-
 ...ic_to_static_shape_non_max_suppression.cpp |     2 +-
 .../dynamic_to_static_shape_nonzero.cpp       |     2 +-
 .../dynamic_to_static_shape_reduce.cpp        |     4 +-
 .../dynamic_to_static_shape_reshape.cpp       |     2 +-
 .../dynamic_to_static_shape_roialign.cpp      |     6 +-
 .../dynamic_to_static_shape_scatter.cpp       |     2 +-
 .../dynamic_to_static_shape_split.cpp         |     2 +-
 .../dynamic_to_static_shape_squeeze.cpp       |     2 +-
 .../dynamic_to_static_shape_topk.cpp          |     4 +-
 .../dynamic_to_static_shape_transpose.cpp     |     2 +-
 ...amic_to_static_shape_unary_elementwise.cpp |     2 +-
 .../dynamic_to_static_shape_unsqueeze.cpp     |     2 +-
 ...dynamic_to_static_shape_variadic_split.cpp |     2 +-
 .../eliminate_shapeof_after_dsr.cpp           |     6 +-
 .../behavior/caching_tests.cpp                |     4 +-
 .../behavior/config.cpp                       |    26 +-
 .../behavior/core_integration.cpp             |    42 +-
 .../behavior/core_threading_tests.cpp         |     4 +-
 .../behavior/cpp_holders.cpp                  |     6 +-
 .../behavior/detect_network_batch_test.cpp    |     2 +-
 .../behavior/exec_graph_info.cpp              |     2 +-
 .../behavior/infer_request.cpp                |     4 +-
 .../behavior/infer_request_callback.cpp       |     4 +-
 .../behavior/infer_request_config.cpp         |     8 +-
 .../behavior/infer_request_input.cpp          |     4 +-
 .../behavior/infer_request_output.cpp         |     4 +-
 .../behavior/layout.cpp                       |     2 +-
 .../behavior/query_network.cpp                |     2 +-
 .../behavior/set_preprocess.cpp               |     8 +-
 .../behavior/stress_tests.cpp                 |     2 +-
 .../behavior/test_plugin.cpp                  |    12 +-
 .../behavior/version.cpp                      |     6 +-
 .../import_export_tests/import_nonzero.cpp    |     2 +-
 .../myriad_remote_blobs_tests.cpp             |     2 +-
 .../single_layer_tests/activation.cpp         |     2 +-
 .../single_layer_tests/comparison.cpp         |     4 +-
 .../single_layer_tests/concat.cpp             |     2 +-
 .../single_layer_tests/convolution.cpp        |     6 +-
 .../ctc_greedy_decoder_seq_len.cpp            |     2 +-
 .../single_layer_tests/eltwise.cpp            |     4 +-
 .../single_layer_tests/gather_elements.cpp    |     4 +-
 .../single_layer_tests/gather_nd.cpp          |     2 +-
 .../single_layer_tests/interpolate.cpp        |    14 +-
 .../single_layer_tests/logical.cpp            |     4 +-
 .../single_layer_tests/mat_mul.cpp            |     2 +-
 .../single_layer_tests/minimum_maximum.cpp    |     2 +-
 .../single_layer_tests/mvn.cpp                |     4 +-
 .../single_layer_tests/nonzero.cpp            |     2 +-
 .../single_layer_tests/pad.cpp                |     8 +-
 .../single_layer_tests/pooling.cpp            |    12 +-
 .../single_layer_tests/proposal.cpp           |     2 +-
 .../single_layer_tests/roi_pooling.cpp        |     4 +-
 .../single_layer_tests/split.cpp              |     4 +-
 .../single_layer_tests/strided_slice.cpp      |     2 +-
 .../single_layer_tests/topk.cpp               |     4 +-
 .../subgraph_tests/constant_result.cpp        |     2 +-
 .../reshape_squeeze_reshape_relu.cpp          |     2 +-
 .../subgraph_tests/split_conv_concat.cpp      |     2 +-
 .../subgraph_tests/tensor_names.cpp           |     2 +-
 .../myriad/single_layer_tests/gather.cpp      |     2 +-
 .../out_shape_of_reshape.cpp                  |     2 +-
 .../static_shape_broadcast.cpp                |     2 +-
 .../single_layer_tests/static_shape_nms.cpp   |     8 +-
 .../static_shape_nonzero.cpp                  |     2 +-
 .../subgraph_tests/concat_split_transpose.cpp |     2 +-
 .../subgraph_tests/dsr_binary_elementwise.cpp |    12 +-
 .../myriad/subgraph_tests/dsr_clamp.cpp       |     2 +-
 .../myriad/subgraph_tests/dsr_concat.cpp      |     2 +-
 .../myriad/subgraph_tests/dsr_convert.cpp     |     2 +-
 .../myriad/subgraph_tests/dsr_gather.cpp      |     6 +-
 .../subgraph_tests/dsr_gather_elements.cpp    |     2 +-
 .../myriad/subgraph_tests/dsr_gather_nd.cpp   |     6 +-
 .../myriad/subgraph_tests/dsr_matmul.cpp      |     2 +-
 .../dsr_non_max_suppression.cpp               |     2 +-
 .../myriad/subgraph_tests/dsr_reduce.cpp      |     4 +-
 .../myriad/subgraph_tests/dsr_reshape.cpp     |     4 +-
 .../myriad/subgraph_tests/dsr_roialign.cpp    |     6 +-
 .../myriad/subgraph_tests/dsr_scatter.cpp     |     2 +-
 .../myriad/subgraph_tests/dsr_split.cpp       |     2 +-
 .../myriad/subgraph_tests/dsr_squeeze.cpp     |     2 +-
 .../subgraph_tests/dsr_strided_slice.cpp      |     2 +-
 .../plugin/myriad/subgraph_tests/dsr_topk.cpp |     4 +-
 .../subgraph_tests/dsr_unary_elementwise.cpp  |     2 +-
 .../myriad/subgraph_tests/dsr_unsqueeze.cpp   |     2 +-
 .../subgraph_tests/dsr_variadic_split.cpp     |     2 +-
 .../subgraph_tests/gather_gather_elements.cpp |     2 +-
 .../myriad/subgraph_tests/nms_nonzero.cpp     |     2 +-
 .../subgraph_tests/nonzero_broadcast.cpp      |     2 +-
 .../subgraph_tests/nonzero_transpose.cpp      |     2 +-
 .../subgraph_tests/parameter_result.cpp       |     2 +-
 .../parameter_shapeof_result.cpp              |     2 +-
 .../subgraph_tests/topk_k_propagation.cpp     |    12 +-
 .../subgraph_tests/unsqueeze_gather.cpp       |     2 +-
 .../shared/src/behavior/caching_tests.cpp     |    10 +-
 .../single_layer/non_max_suppression.hpp      |    10 +
 .../tests/ie_test_utils/CMakeLists.txt        |     4 +-
 .../common_test_utils/CMakeLists.txt          |    62 +-
 .../common_test_utils/common_utils.cpp        |     5 +
 .../common_test_utils/common_utils.hpp        |     5 +
 .../ie_test_utils/common_test_utils/gtest     |     1 -
 .../ie_test_utils/unit_test_utils/mock.cpp    |     1 +
 .../cpp_interfaces/interface/mock_icore.hpp   |    14 +-
 .../interface/mock_iinference_plugin.hpp      |    20 +-
 .../unit_test_utils/mocks/mock_allocator.hpp  |     8 +-
 .../mocks/mock_icnn_network.hpp               |    40 +-
 .../mocks/mock_iexecutable_network.hpp        |    24 +-
 .../mocks/mock_iinfer_request.hpp             |    30 +-
 .../mocks/mock_not_empty_icnn_network.hpp     |    17 +-
 .../unit/cpu/mkldnn_memory_desc_test.cpp      |    10 +-
 .../tests/unit/cpu/mkldnn_memory_test.cpp     |     4 +-
 .../ie_infer_async_request_base_test.cpp      |     3 +-
 .../ie_memory_state_internal_test.cpp         |     9 +-
 .../tests/unit/vpu/blob_reader_tests.cpp      |     6 +-
 .../passes_tests/annotate_memory_types.cpp    |     2 +-
 .../vpu/myriad_tests/aot_behavior_tests.cpp   |     2 +-
 .../vpu/myriad_tests/vpu_boot_tests.cpp       |     2 +-
 .../vpu/myriad_tests/vpu_protocol_tests.cpp   |     2 +-
 .../vpu/myriad_tests/vpu_watchdog_tests.cpp   |     2 +-
 .../fluid_preproc/cpu/fluid_tests_cpu.cpp     |    70 +-
 .../backward_compatibility.cpp                |     2 +-
 .../input_tests/parser_tests.cpp              |     2 +-
 .../io_blob_tests/cropResize_tests.hpp        |     4 +-
 .../shared_tests/io_blob_tests/dims_tests.hpp |     4 +-
 .../io_blob_tests/layout_tests.hpp            |     6 +-
 .../functional/shared_tests/lstm/plg_test.hpp |     4 +-
 .../shared_tests/lstm/rnn_seq_test.hpp        |     2 +-
 .../layers/myriad_layers_CTCDecoder_test.cpp  |     2 +-
 ...myriad_layers_batch_normalization_test.cpp |     2 +-
 .../common/layers/myriad_layers_bias_test.cpp |     2 +-
 .../common/layers/myriad_layers_blob_test.cpp |     6 +-
 .../layers/myriad_layers_clamp_test.cpp       |     2 +-
 .../layers/myriad_layers_concat_test.cpp      |     2 +-
 .../layers/myriad_layers_conv_nd_test.cpp     |    30 +-
 .../layers/myriad_layers_convert_test.cpp     |     6 +-
 .../layers/myriad_layers_convolution1x1.cpp   |     2 +-
 .../layers/myriad_layers_convolution3x3.cpp   |     2 +-
 .../layers/myriad_layers_convolution_test.cpp |   102 +-
 .../common/layers/myriad_layers_copy_test.cpp |     2 +-
 .../common/layers/myriad_layers_crop_test.cpp |     6 +-
 .../layers/myriad_layers_custom_test.cpp      |    14 +-
 .../myriad_layers_deconvolution_test.cpp      |    14 +-
 .../layers/myriad_layers_eltwise_test.cpp     |    48 +-
 .../common/layers/myriad_layers_elu_test.cpp  |     2 +-
 .../common/layers/myriad_layers_erf_test.cpp  |     2 +-
 ...myriad_layers_exp_detectionoutput_test.cpp |     2 +-
 .../myriad_layers_exp_generateproposals.cpp   |     2 +-
 ...iad_layers_exp_priorgridgenerator_test.cpp |     6 +-
 .../common/layers/myriad_layers_exp_test.cpp  |     2 +-
 .../myriad_layers_exp_topkrois_test.cpp       |     2 +-
 .../layers/myriad_layers_flatten_test.cpp     |     2 +-
 .../layers/myriad_layers_floor_test.cpp       |     2 +-
 .../myriad_layers_fully_connected_tests.cpp   |     6 +-
 .../common/layers/myriad_layers_gemm_test.cpp |     2 +-
 .../common/layers/myriad_layers_grn_test.cpp  |     2 +-
 .../layers/myriad_layers_interp_test.cpp      |     2 +-
 .../common/layers/myriad_layers_log_test.cpp  |     2 +-
 .../common/layers/myriad_layers_lrn_test.cpp  |     2 +-
 .../common/layers/myriad_layers_lstm_cell.cpp |     4 +-
 .../common/layers/myriad_layers_mvn_test.cpp  |     2 +-
 .../layers/myriad_layers_nonzero_test.cpp     |     2 +-
 .../layers/myriad_layers_normalize_test.cpp   |     4 +-
 .../layers/myriad_layers_oneHot_test.cpp      |     4 +-
 .../common/layers/myriad_layers_pad_test.cpp  |     2 +-
 .../layers/myriad_layers_permute_nd_test.cpp  |    14 +-
 .../layers/myriad_layers_permute_test.cpp     |     4 +-
 .../layers/myriad_layers_pool_nd_test.cpp     |    44 +-
 .../layers/myriad_layers_pooling_test.cpp     |    48 +-
 .../layers/myriad_layers_power_test.cpp       |     2 +-
 .../layers/myriad_layers_prelu_test.cpp       |    16 +-
 .../myriad_layers_psroipooling_test.cpp       |     2 +-
 .../myriad_layers_psroipooling_test.hpp       |     2 +-
 .../layers/myriad_layers_reduce_test.cpp      |    10 +-
 .../layers/myriad_layers_region_test.cpp      |     4 +-
 .../common/layers/myriad_layers_relu_test.cpp |    16 +-
 .../layers/myriad_layers_reorg_test.cpp       |     4 +-
 .../layers/myriad_layers_resample_test.cpp    |     4 +-
 .../layers/myriad_layers_reshape_test.cpp     |     8 +-
 .../myriad_layers_reverse_sequence_test.cpp   |     2 +-
 .../layers/myriad_layers_roi_align_test.cpp   |     8 +-
 ...riad_layers_roi_feature_extractor_test.cpp |     4 +-
 .../layers/myriad_layers_roi_pooling_test.cpp |     4 +-
 .../layers/myriad_layers_scale_test.cpp       |     2 +-
 ...ad_layers_scatter_elements_update_test.cpp |     4 +-
 .../myriad_layers_scatter_update_test.cpp     |     4 +-
 .../layers/myriad_layers_select_test.cpp      |     2 +-
 .../layers/myriad_layers_sigmoid_test.cpp     |    10 +-
 .../layers/myriad_layers_slice_test.cpp       |     2 +-
 .../layers/myriad_layers_softmax_test.cpp     |     2 +-
 .../layers/myriad_layers_split_test.cpp       |     2 +-
 .../layers/myriad_layers_squeeze_test.cpp     |    10 +-
 .../myriad_layers_strided_slice_test.cpp      |     2 +-
 .../common/layers/myriad_layers_tanh_test.cpp |    10 +-
 .../common/layers/myriad_layers_tile_test.cpp |     4 +-
 .../common/layers/myriad_layers_topk_test.cpp |     2 +-
 .../layers/myriad_layers_unsqueeze_test.cpp   |     2 +-
 .../vpu/common/myriad_get_output_tests.cpp    |     2 +-
 .../vpu/common/myriad_hw_conv_tests.hpp       |   130 +-
 .../vpu/common/myriad_hw_extra_tests.hpp      |    20 +-
 .../vpu/common/myriad_hw_fc_tests.hpp         |    28 +-
 .../vpu/common/myriad_hw_network_tests.hpp    |     4 +-
 .../vpu/common/myriad_hw_pool_tests.hpp       |    60 +-
 .../vpu/common/myriad_merge_permute_tests.cpp |     6 +-
 .../regression/helpers/vpu_case_common.hpp    |     8 +-
 .../merge_permute_and_reorder_test.cpp        |     4 +-
 .../vpu/myriad_tests/myriad_configs_tests.cpp |     8 +-
 .../myriad_multiple_graph_tests.cpp           |     2 +-
 .../myriad_streams_configuration_tests.cpp    |     2 +-
 .../single_layer_tests.cpp                    |     6 +-
 .../input_tests/parser_tests.cpp              |     2 +-
 .../vpu/vpu_base/vpu_tests_config.hpp         |     4 +-
 .../helpers/tests_vpu_common.hpp              |     2 +-
 .../engines/gna/fp32_non_quantized_tests.cpp  |     2 +-
 .../engines/gna/gna_input_precision_test.cpp  |     2 +-
 .../unit/engines/gna/gna_matcher.cpp          |     2 +-
 .../unit/engines/gna/gna_proc_type_test.cpp   |     2 +-
 .../engines/gna/layers/activation_test.cpp    |     2 +-
 .../gna/layers/gna_align_filter2_tests.cpp    |     2 +-
 .../engines/gna/layers/gna_conv1d_test.cpp    |     2 +-
 .../engines/gna/layers/gna_eltwise_test.cpp   |     2 +-
 .../layers/gna_multi_input_to_concat_test.cpp |     2 +-
 .../engines/gna/layers/gna_permute_test.cpp   |     2 +-
 .../gna/layers/gna_split_to_concat_test.cpp   |     2 +-
 .../engines/gna/layers/gna_squeeze_test.cpp   |     4 +-
 .../unit/engines/vpu/merge_parallel_fc.cpp    |     4 +-
 .../vpu/mvnc/pthread_semaphore_tests.cpp      |     2 +-
 .../unit/engines/vpu/mvnc/watchdog_tests.cpp  |     8 +-
 .../engines/vpu/mvnc/xlink_device_tests.cpp   |     2 +-
 .../myriad_tests/helpers/myriad_mvnc_stub.h   |     4 +-
 .../vpu/myriad_tests/myriad_engine_tests.cpp  |    18 +-
 .../vpu/myriad_tests/myriad_metrics_tests.cpp |     6 +-
 .../unit/graph_tools/graph_copy_tests.cpp     |     4 +-
 .../unit/graph_tools/graph_tools_test.cpp     |     4 +-
 .../get_num_iterations_test.cpp               |    18 +-
 .../thirdparty/clDNN/CMakeLists.txt           |     3 -
 .../clDNN/common/googletest-fused/License.txt |    28 -
 .../googletest-fused/gmock-gtest-all.cc       | 13278 ----------
 .../common/googletest-fused/gmock/gmock.h     | 16697 ------------
 .../common/googletest-fused/gtest/gtest.h     | 21575 ----------------
 .../clDNN/common/googletest-fused/version.txt |     4 -
 .../thirdparty/clDNN/tests/CMakeLists.txt     |    18 +-
 .../module_tests/test_uqr_distribution.cpp    |     2 +-
 .../tests/module_tests/usm_memory_test.cpp    |    12 +-
 .../binary_convolution_gpu_test.cpp           |     4 +-
 .../clDNN/tests/test_cases/cache_test.cpp     |     2 +-
 .../test_cases/concatenation_gpu_test.cpp     |    10 +-
 .../tests/test_cases/convolution_gpu_test.cpp |    24 +-
 .../clDNN/tests/test_cases/crop_gpu_test.cpp  |     4 +-
 .../tests/test_cases/cum_sum_gpu_test.cpp     |    24 +-
 .../test_cases/deconvolution_gpu_test.cpp     |     6 +-
 .../test_cases/depth_concatenate_gpu_test.cpp |     2 +-
 .../test_cases/detection_output_test.cpp      |     2 +-
 .../tests/test_cases/eltwise_gpu_test.cpp     |    24 +-
 .../test_cases/fully_connected_gpu_test.cpp   |    52 +-
 .../tests/test_cases/fusings_gpu_test.cpp     |   544 +-
 .../clDNN/tests/test_cases/gemm_gpu_test.cpp  |    60 +-
 .../test_cases/lstm_dynamic_gpu_test.cpp      |     4 +-
 .../clDNN/tests/test_cases/mvn_gpu_test.cpp   |    12 +-
 .../test_cases/non_max_suppression_test.cpp   |     2 +-
 .../tests/test_cases/permute_gpu_test.cpp     |     8 +-
 .../tests/test_cases/pooling_gpu_test.cpp     |    10 +-
 .../test_cases/pyramid_roi_align_gpu_test.cpp |     2 +-
 .../tests/test_cases/reduce_gpu_test.cpp      |    14 +-
 .../tests/test_cases/reorder_gpu_test.cpp     |     2 +-
 .../tests/test_cases/resample_gpu_test.cpp    |    12 +-
 .../tests/test_cases/softmax_gpu_test.cpp     |     2 +-
 .../clDNN/tests/test_cases/topology_test.cpp  |     2 +-
 .../clDNN/tests/test_utils/network_test.h     |     2 +-
 .../movidius/mvnc/tests/mvnc_stress_tests.cpp |     2 +-
 .../movidius/mvnc/tests/mvnc_tests_common.cpp |     8 +-
 .../movidius/mvnc/tests/mvnc_tests_usb.cpp    |     2 +-
 .../tests/XLink/XLink_common_tests.cpp        |    16 +-
 tests/stress_tests/memcheck_tests/tests.cpp   |     2 +-
 tests/stress_tests/memleaks_tests/tests.cpp   |     6 +-
 tests/stress_tests/unittests/tests.cpp        |     6 +-
 thirdparty/CMakeLists.txt                     |    56 +-
 thirdparty/gtest                              |     1 +
 992 files changed, 2921 insertions(+), 54490 deletions(-)
 delete mode 160000 inference-engine/tests/ie_test_utils/common_test_utils/gtest
 delete mode 100644 inference-engine/thirdparty/clDNN/common/googletest-fused/License.txt
 delete mode 100644 inference-engine/thirdparty/clDNN/common/googletest-fused/gmock-gtest-all.cc
 delete mode 100644 inference-engine/thirdparty/clDNN/common/googletest-fused/gmock/gmock.h
 delete mode 100644 inference-engine/thirdparty/clDNN/common/googletest-fused/gtest/gtest.h
 delete mode 100644 inference-engine/thirdparty/clDNN/common/googletest-fused/version.txt
 create mode 160000 thirdparty/gtest

diff --git a/.gitmodules b/.gitmodules
index dad6d39352d7b0..586502884302ac 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -6,10 +6,6 @@
 	path = inference-engine/thirdparty/mkl-dnn
 	url = https://github.com/openvinotoolkit/oneDNN.git
 	ignore = dirty
-[submodule "inference-engine/tests/ie_test_utils/common_test_utils/gtest"]
-	path = inference-engine/tests/ie_test_utils/common_test_utils/gtest
-	url = https://github.com/openvinotoolkit/googletest.git
-	ignore = dirty
 [submodule "inference-engine/samples/thirdparty/gflags"]
 	path = inference-engine/samples/thirdparty/gflags
 	url = https://github.com/gflags/gflags.git
@@ -22,3 +18,7 @@
 	path = thirdparty/zlib/zlib
 	url = https://github.com/madler/zlib.git
 	ignore = dirty
+[submodule "thirdparty/gtest"]
+	path = thirdparty/gtest
+	url = https://github.com/openvinotoolkit/googletest.git
+	ignore = dirty
diff --git a/docs/IE_PLUGIN_DG/PluginTesting.md b/docs/IE_PLUGIN_DG/PluginTesting.md
index 0aae6280601ca1..e2aabeff32e497 100644
--- a/docs/IE_PLUGIN_DG/PluginTesting.md
+++ b/docs/IE_PLUGIN_DG/PluginTesting.md
@@ -21,7 +21,7 @@ Engine concepts: plugin creation, multiple executable networks support, multiple
 
     @snippet single_layer_tests/convolution.cpp test_convolution:declare_parameters
 
-    - Instantiate the test itself using standard GoogleTest macro `INSTANTIATE_TEST_CASE_P`:
+    - Instantiate the test itself using standard GoogleTest macro `INSTANTIATE_TEST_SUITE_P`:
 
     @snippet single_layer_tests/convolution.cpp test_convolution:instantiate
 
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/caching_tests.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/caching_tests.cpp
index f61e4c54d7ec81..a698d72e52610d 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/caching_tests.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/caching_tests.cpp
@@ -15,7 +15,7 @@ namespace {
             1, 2
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_CachingSupportCase_Template, LoadNetworkCacheTestBase,
+    INSTANTIATE_TEST_SUITE_P(smoke_CachingSupportCase_Template, LoadNetworkCacheTestBase,
                             ::testing::Combine(
                                     ::testing::ValuesIn(LoadNetworkCacheTestBase::getStandardFunctions()),
                                     ::testing::ValuesIn(precisionsTemplate),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/config.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/config.cpp
index fad461c814da58..34e0eb46eace0c 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/config.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/config.cpp
@@ -24,14 +24,14 @@ const std::vector<std::map<std::string, std::string>> inconfigs = {
     {{TEMPLATE_CONFIG_KEY(THROUGHPUT_STREAMS), CONFIG_VALUE(NO)}},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigTests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
                                 ::testing::ValuesIn(inconfigs)),
                         IncorrectConfigTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
@@ -39,14 +39,14 @@ INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
                         IncorrectConfigAPITests::getTestCaseName);
 
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigAPITests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectConfigAPITests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
                                 ::testing::ValuesIn(configs)),
                         CorrectConfigAPITests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigTests,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, CorrectConfigTests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/core_integration.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/core_integration.cpp
index 2b669a6520a376..2c067aaf7b68ab 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/core_integration.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/core_integration.cpp
@@ -16,11 +16,11 @@ namespace {
 // IE Class Common tests with <pluginName, deviceName params>
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassBasicTestP, IEClassBasicTestP,
         ::testing::Values(std::make_pair("templatePlugin", CommonTestUtils::DEVICE_TEMPLATE)));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassNetworkTestP, IEClassNetworkTestP,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
@@ -28,39 +28,39 @@ INSTANTIATE_TEST_CASE_P(
 // IE Class GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_AVAILABLE_DEVICES,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_FULL_DEVICE_NAME,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_OPTIMIZATION_CAPABILITIES,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_RANGE_FOR_ASYNC_INFER_REQUESTS,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_ThrowUnsupported,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetConfigTest, IEClassGetConfigTest_ThrowUnsupported,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetAvailableDevices, IEClassGetAvailableDevices,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
@@ -111,7 +111,7 @@ TEST_F(IEClassSetConfigTestHETERO, smoke_SetConfigNoThrow) {
 // IE Class GetConfig
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetConfigTest, IEClassGetConfigTest,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
@@ -143,46 +143,46 @@ TEST_F(IEClassGetConfigTestTEMPLATE, smoke_GetConfigNoThrow) {
 // Executable Network GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE, "MULTI:TEMPLATE", "HETERO:TEMPLATE"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE, "MULTI:TEMPLATE", "HETERO:TEMPLATE"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_NETWORK_NAME,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE, "MULTI:TEMPLATE", "HETERO:TEMPLATE"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_OPTIMAL_NUMBER_OF_INFER_REQUESTS,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE, "MULTI:TEMPLATE", "HETERO:TEMPLATE"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetMetricTest_ThrowsUnsupported, IEClassExecutableNetworkGetMetricTest,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE, "MULTI:TEMPLATE", "HETERO:TEMPLATE"));
 //
 // Executable Network GetConfig / SetConfig
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetConfigTest, IEClassExecutableNetworkGetConfigTest,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkSetConfigTest, IEClassExecutableNetworkSetConfigTest,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
 // IE Class Query network
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassQueryNetworkTest, IEClassQueryNetworkTest,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
 // IE Class Load network
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassLoadNetworkTest, IEClassLoadNetworkTest,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
@@ -192,19 +192,19 @@ INSTANTIATE_TEST_CASE_P(
 
 #ifdef ENABLE_MKL_DNN
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_NETWORK_NAME,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_TARGET_FALLBACK,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE));
 
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/cpp_holders.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/cpp_holders.cpp
index 8a2ee657b7917a..7c372c708ad3bb 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/cpp_holders.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/cpp_holders.cpp
@@ -20,19 +20,19 @@ const std::vector<std::vector<int >> orders = {
     {2, 1, 0}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTest,
         ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
         ::testing::ValuesIn(orders)),
         HoldersTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTestImportNetwork,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTestImportNetwork,
         ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE, "HETERO:TEMPLATE"),
         ::testing::ValuesIn(orders)),
         HoldersTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTestOnImportedNetwork,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTestOnImportedNetwork,
         ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE, "HETERO:TEMPLATE"),
         HoldersTestOnImportedNetwork::getTestCaseName);
 
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/exec_graph_info.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/exec_graph_info.cpp
index a0aa412fb0d6e9..597e029fc5be53 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/exec_graph_info.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/exec_graph_info.cpp
@@ -19,7 +19,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     {}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, ExecGraphTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, ExecGraphTests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request.cpp
index 2b39ef540057b0..edb803a614f689 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request.cpp
@@ -19,7 +19,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     {}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestTests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_callback.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_callback.cpp
index 8fd635ae8bf481..3bc9d1f6f02d3f 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_callback.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_callback.cpp
@@ -19,7 +19,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     {}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CallbackTests,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_config.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_config.cpp
index 3ee70738e5a0d5..a6ffbe80bbcec9 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_config.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_config.cpp
@@ -19,7 +19,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     {}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferConfigTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferConfigTests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_input.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_input.cpp
index dde2fa6d2cd3f2..23d5cb7136faf4 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_input.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_input.cpp
@@ -17,7 +17,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     {}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestInputTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestInputTests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_output.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_output.cpp
index 9c8a24413fa647..eae52cec60c14e 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_output.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/infer_request_output.cpp
@@ -17,7 +17,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     {}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestOutputTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestOutputTests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/layout.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/layout.cpp
index c5770ede2a0956..f4d8091b20e088 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/layout.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/layout.cpp
@@ -26,7 +26,7 @@ const std::vector<std::vector<size_t>> inputShapes = {
     { 3 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, LayoutTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, LayoutTest,
                         ::testing::Combine(
                                 ::testing::Values(InferenceEngine::Precision::FP32),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/preprocessing.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/preprocessing.cpp
index 29651f6da34f92..4aa9439b1ad891 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/preprocessing.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/preprocessing.cpp
@@ -17,7 +17,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     {}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_PreprocessingPrecisionConvertTestsViaSetInput, PreprocessingPrecisionConvertTest,
+INSTANTIATE_TEST_SUITE_P(smoke_PreprocessingPrecisionConvertTestsViaSetInput, PreprocessingPrecisionConvertTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inputPrecisions),
                                 ::testing::Values(4),   // Number of input tensor channels
@@ -26,7 +26,7 @@ INSTANTIATE_TEST_CASE_P(smoke_PreprocessingPrecisionConvertTestsViaSetInput, Pre
                                 ::testing::ValuesIn(configs)),
                         PreprocessingPrecisionConvertTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_PreprocessingPrecisionConvertTestsViaGetBlob, PreprocessingPrecisionConvertTest,
+INSTANTIATE_TEST_SUITE_P(smoke_PreprocessingPrecisionConvertTestsViaGetBlob, PreprocessingPrecisionConvertTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inputPrecisions),
                                 ::testing::Values(4),       // Number of input tensor channels (blob_copy only supports 4d and 5d tensors)
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/set_preprocess.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/set_preprocess.cpp
index a33bcd7638cc8a..19841bb64da643 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/set_preprocess.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/set_preprocess.cpp
@@ -26,21 +26,21 @@ const std::vector<std::map<std::string, std::string>> heteroConfigs = {
         {{ "TARGET_FALLBACK", CommonTestUtils::DEVICE_TEMPLATE }}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
                                 ::testing::ValuesIn(configs)),
                         PreprocessTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, PreprocessTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, PreprocessTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                 ::testing::ValuesIn(multiConfigs)),
                         PreprocessTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Hetero_BehaviorTests, PreprocessTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Hetero_BehaviorTests, PreprocessTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_HETERO),
@@ -61,7 +61,7 @@ const std::vector<InferenceEngine::Layout> ioLayouts = {
     InferenceEngine::Layout::NHWC
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessConversionTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessConversionTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::ValuesIn(ioPrecisions),
@@ -75,7 +75,7 @@ INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessConversionTest,
                                 ::testing::ValuesIn(configs)),
                         PreprocessConversionTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Bool(),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/test_plugin.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/test_plugin.cpp
index 8d65895bc36b04..b8d5e9bc9bdb5a 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/test_plugin.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/test_plugin.cpp
@@ -17,21 +17,21 @@ const std::vector<std::map<std::string, std::string>> configs = {
     {}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTests,
                         ::testing::Combine(
                                 ::testing::Values(InferenceEngine::Precision::FP32),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
                                 ::testing::ValuesIn(configs)),
                         BehaviorTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTestInput,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTestInput,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
                                 ::testing::ValuesIn(configs)),
                         BehaviorTestInput::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTestOutput,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTestOutput,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/version.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/version.cpp
index 818a627a51973a..8b22d2eb24094e 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/behavior/version.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/behavior/version.cpp
@@ -12,7 +12,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     {}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, VersionTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, VersionTest,
                         ::testing::Combine(
                                 ::testing::Values(InferenceEngine::Precision::FP32),
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/hetero/query_network.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/hetero/query_network.cpp
index b369e55adf225e..64b88ea714f88f 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/hetero/query_network.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/hetero/query_network.cpp
@@ -13,7 +13,7 @@ using namespace HeteroTests;
 
 auto ConvBias = ngraph::builder::subgraph::makeConvBias();
 
-INSTANTIATE_TEST_CASE_P(smoke_FullySupportedTopologies, QueryNetworkTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FullySupportedTopologies, QueryNetworkTest,
                         ::testing::Combine(
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE, "HETERO:TEMPLATE", "MULTI:TEMPLATE"),
                                 ::testing::Values(ConvBias)),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/hetero/synthetic.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/hetero/synthetic.cpp
index 15f940780ddb64..164d959e408c08 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/hetero/synthetic.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/hetero/synthetic.cpp
@@ -11,13 +11,13 @@
 namespace {
 using namespace HeteroTests;
 
-INSTANTIATE_TEST_CASE_P(smoke_SingleMajorNode, HeteroSyntheticTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SingleMajorNode, HeteroSyntheticTest,
                         ::testing::Combine(
                                 ::testing::Values(std::vector<PluginParameter>{{"TEMPLATE0", "templatePlugin"}, {"TEMPLATE1", "templatePlugin"}}),
                                 ::testing::ValuesIn(HeteroTests::HeteroSyntheticTest::_singleMajorNodeFunctions)),
                         HeteroSyntheticTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(nightly_RandomMajorNodes, HeteroSyntheticTest,
+INSTANTIATE_TEST_SUITE_P(nightly_RandomMajorNodes, HeteroSyntheticTest,
                         ::testing::Combine(
                                 ::testing::Values(std::vector<PluginParameter>{{"TEMPLATE0", "templatePlugin"}, {"TEMPLATE1", "templatePlugin"}}),
                                 ::testing::ValuesIn(HeteroTests::HeteroSyntheticTest::_randomMajorNodeFunctions)),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/convolution.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/convolution.cpp
index 3f6d9c494066c7..c099cb3d5d4ffc 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/convolution.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/convolution.cpp
@@ -57,7 +57,7 @@ const auto conv2DParams_AutoPadValid = ::testing::Combine(
 );
 
 // ! [test_convolution:instantiate]
-INSTANTIATE_TEST_CASE_P(Convolution2D_ExplicitPadding, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(Convolution2D_ExplicitPadding, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -70,7 +70,7 @@ INSTANTIATE_TEST_CASE_P(Convolution2D_ExplicitPadding, ConvolutionLayerTest,
                         ConvolutionLayerTest::getTestCaseName);
 // ! [test_convolution:instantiate]
 
-INSTANTIATE_TEST_CASE_P(Convolution2D_AutoPadValid, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(Convolution2D_AutoPadValid, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -113,7 +113,7 @@ const auto conv3DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::VALID)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution3D_ExplicitPadding, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution3D_ExplicitPadding, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv3DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -125,7 +125,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Convolution3D_ExplicitPadding, ConvolutionLayerTes
                                 ::testing::Values(CommonTestUtils::DEVICE_TEMPLATE)),
                         ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(nightly_Convolution3D_AutoPadValid, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(nightly_Convolution3D_AutoPadValid, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv3DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/reshape.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/reshape.cpp
index e0c986ad8b52e5..d95ecdd81c826f 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/reshape.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/reshape.cpp
@@ -15,7 +15,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
     InferenceEngine::Precision::FP32,
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_ReshapeCheckDynBatch, ReshapeLayerTestRevise,
     ::testing::Combine(
         ::testing::Values(true), ::testing::ValuesIn(netPrecisions),
@@ -29,7 +29,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(std::map<std::string, std::string>({}))),
     ReshapeLayerTestRevise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_ReshapeCheck, ReshapeLayerTestRevise,
     ::testing::Combine(
         ::testing::Values(true), ::testing::ValuesIn(netPrecisions),
@@ -43,7 +43,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(std::map<std::string, std::string>({}))),
     ReshapeLayerTestRevise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_ReshapeCheckNegative, ReshapeLayerTestRevise,
     ::testing::Combine(
         ::testing::Values(true), ::testing::ValuesIn(netPrecisions),
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/softmax.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/softmax.cpp
index 3ad1b07c736e5e..583d2d7b81e727 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/softmax.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/softmax.cpp
@@ -41,7 +41,7 @@ const auto params2D = testing::Combine(
     testing::Values(std::map<std::string, std::string>())
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_SoftMax2D,
         SoftMaxLayerTest,
         params2D,
@@ -68,7 +68,7 @@ const auto params4D = testing::Combine(
     testing::Values(std::map<std::string, std::string>())
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_SoftMax4D,
         SoftMaxLayerTest,
         params4D,
diff --git a/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/split.cpp b/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/split.cpp
index 44f2c6737785b4..f4ea3db8af9525 100644
--- a/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/split.cpp
+++ b/docs/template_plugin/tests/functional/shared_tests_instances/single_layer_tests/split.cpp
@@ -11,7 +11,7 @@ using namespace LayerTestsDefinitions;
 
 namespace {
 
-INSTANTIATE_TEST_CASE_P(smoke_NumSplitsCheck, SplitLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NumSplitsCheck, SplitLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(1, 2, 3, 5, 6, 10, 30),
                                 ::testing::Values(0, 1, 2, 3),
diff --git a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
index ac01a2324c997a..5b0fbcd297d400 100644
--- a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
+++ b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
@@ -1216,8 +1216,8 @@ void InsertConcatAligningFilterPass::run() {
                 // modifying output rows to be used - to avoid modification to original concat we are store num of elements in params
                 dims[1] = num_rows_out;
 
-                if (concatInput->getLayout() == Layout::NC && dims[0] > 8 ||
-                    concatInput->getLayout() == Layout::CN && dims[1] > 8) {
+                if ((concatInput->getLayout() == Layout::NC && dims[0] > 8) ||
+                    (concatInput->getLayout() == Layout::CN && dims[1] > 8)) {
                     THROW_GNA_EXCEPTION << "unsupported batch number '" <<
                         (concatInput->getLayout() == Layout::NC ? dims[0] : dims[1]) <<
                         "' in layer '" << concatLayer->name << "'";
@@ -1669,7 +1669,7 @@ void BroadcastConstPass::run() {
         };
 
         auto nextLayer = CNNNetCheckNextLayerSkipCertain(constLayer, 0, 0, true, isNonFunctional).first;
-        if (!nextLayer || !LayerInfo(nextLayer).isEltwise() && !LayerInfo(nextLayer).isFakeQuantize()) {
+        if (!nextLayer || (!LayerInfo(nextLayer).isEltwise() && !LayerInfo(nextLayer).isFakeQuantize())) {
             continue;
         }
 
diff --git a/inference-engine/tests/functional/inference_engine/blob_copy_test.cpp b/inference-engine/tests/functional/inference_engine/blob_copy_test.cpp
index 1968c4cb91da9e..d6356f154d0c6d 100644
--- a/inference-engine/tests/functional/inference_engine/blob_copy_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/blob_copy_test.cpp
@@ -320,7 +320,7 @@ std::vector<PrecisionType> BlobCopy_PrecisionParams = {
 
 }  // namespace
 
-INSTANTIATE_TEST_CASE_P(accuracy, BlobCopyTest,
+INSTANTIATE_TEST_SUITE_P(accuracy, BlobCopyTest,
                         ::testing::Combine(::testing::ValuesIn(BlobCopy_srcLayoutParam),
                            ::testing::ValuesIn(BlobCopy_dstLayoutParam),
                            ::testing::ValuesIn(BlobCopy_BatchNum),
@@ -468,7 +468,7 @@ TEST_P(BlobCopySetLayoutTest, BlobCopyWithNCHW_To_NHWC_After_setLayout) {
     ASSERT_TRUE(IsEqualBlobCopy(ref, dst)) << "'blob_copy' after setLayout function is not correct";
 }
 
-INSTANTIATE_TEST_CASE_P(accuracy, BlobCopySetLayoutTest,
+INSTANTIATE_TEST_SUITE_P(accuracy, BlobCopySetLayoutTest,
     ::testing::Combine(::testing::ValuesIn(BlobCopySetLayout_Dims),
                        ::testing::ValuesIn(BlobCopySetLayout_Precisions)));
 
diff --git a/inference-engine/tests/functional/inference_engine/caching_test.cpp b/inference-engine/tests/functional/inference_engine/caching_test.cpp
index cad8bd4428e3ba..4f27b8abaf99fa 100644
--- a/inference-engine/tests/functional/inference_engine/caching_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/caching_test.cpp
@@ -67,7 +67,7 @@ class MockRemoteContext : public RemoteContext {
     MockRemoteContext(std::string name): m_name(std::move(name)) {}
     std::string getDeviceName() const noexcept { return m_name; }
     MOCK_METHOD2(CreateBlob, RemoteBlob::Ptr(const TensorDesc&, const ParamMap&));
-    MOCK_QUALIFIED_METHOD0(getParams, const, ParamMap());
+    MOCK_CONST_METHOD0(getParams, ParamMap());
 };
 
 class MockCachingInferencePluginBase : public InferenceEngine::IInferencePlugin {
@@ -1075,7 +1075,7 @@ TEST_P(CachingTest, TestCacheFileOldVersion) {
             if (index != std::string::npos) {
                 content.replace(index, buildNum.size(), zeroBuild);
             } else {
-                SKIP();
+                return; // skip test
             }
             std::ofstream out(fileName, std::ios_base::binary);
             out.write(content.c_str(), content.size());
@@ -1505,7 +1505,7 @@ TEST_P(CachingTest, LoadMulti_NoCachingOnDevice) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(CachingTest, CachingTest,
+INSTANTIATE_TEST_SUITE_P(CachingTest, CachingTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(loadVariants),
                             ::testing::ValuesIn(cacheFolders)),
diff --git a/inference-engine/tests/functional/inference_engine/cnn_network/matmul_sr_tests.cpp b/inference-engine/tests/functional/inference_engine/cnn_network/matmul_sr_tests.cpp
index a013ac8c902ba4..33940bfab6de15 100644
--- a/inference-engine/tests/functional/inference_engine/cnn_network/matmul_sr_tests.cpp
+++ b/inference-engine/tests/functional/inference_engine/cnn_network/matmul_sr_tests.cpp
@@ -70,7 +70,7 @@ class SmartReshapeMatMulTests : public CommonTestUtils::TestsCommon, public test
 TEST_P(SmartReshapeMatMulTests, ReshapeMatMul) {
 }
 
-INSTANTIATE_TEST_CASE_P(NGraph, SmartReshapeMatMulTests, testing::Values(
+INSTANTIATE_TEST_SUITE_P(NGraph, SmartReshapeMatMulTests, testing::Values(
         ReshapeMatMulTestCase{true, {1, 20, 30}, {30, 40}, {20, -1}, false, false, {{"input_A", {2, 20, 30}}}},
         ReshapeMatMulTestCase{true, {1, 20, 30}, {40, 30}, {20, -1}, false, true, {{"input_A", {2, 20, 30}}}},
         ReshapeMatMulTestCase{true, {1, 30, 20}, {30, 20}, {-1, 20}, true, false, {{"input_A", {2, 30, 20}}}},
diff --git a/inference-engine/tests/functional/inference_engine/ir_serialization/serialize.cpp b/inference-engine/tests/functional/inference_engine/ir_serialization/serialize.cpp
index 990fa74c183460..c9efc087a318c2 100644
--- a/inference-engine/tests/functional/inference_engine/ir_serialization/serialize.cpp
+++ b/inference-engine/tests/functional/inference_engine/ir_serialization/serialize.cpp
@@ -53,7 +53,7 @@ TEST_P(SerializationTest, CompareFunctions) {
     ASSERT_TRUE(success) << message;
 }
 
-INSTANTIATE_TEST_CASE_P(IRSerialization, SerializationTest,
+INSTANTIATE_TEST_SUITE_P(IRSerialization, SerializationTest,
         testing::Values(std::make_tuple("add_abc.xml", "add_abc.bin"),
                         std::make_tuple("add_abc_f64.xml", ""),
                         std::make_tuple("add_abc_bin.xml", ""),
@@ -76,7 +76,7 @@ INSTANTIATE_TEST_CASE_P(IRSerialization, SerializationTest,
 
 #ifdef NGRAPH_ONNX_IMPORT_ENABLE
 
-INSTANTIATE_TEST_CASE_P(ONNXSerialization, SerializationTest,
+INSTANTIATE_TEST_SUITE_P(ONNXSerialization, SerializationTest,
         testing::Values(std::make_tuple("add_abc.prototxt", ""),
                         std::make_tuple("split_equal_parts_2d.prototxt", ""),
                         std::make_tuple("addmul_abc.prototxt", ""),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/add_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/add_transformation.cpp
index 8098b89e73c1b2..2644d58470fc43 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/add_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/add_transformation.cpp
@@ -807,7 +807,7 @@ const std::vector<AddTransformationTestValues> addTransformationTestValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     AddTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/avg_pool_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/avg_pool_transformation.cpp
index 94d0abed429aeb..1e200cba65525c 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/avg_pool_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/avg_pool_transformation.cpp
@@ -336,7 +336,7 @@ const std::vector<AvgPoolTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     AvgPoolTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/clamp_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/clamp_transformation.cpp
index d2785e355a4d7b..994fe9a644eed1 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/clamp_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/clamp_transformation.cpp
@@ -451,7 +451,7 @@ const std::vector<ClampTransformationTestValues> testValues = {
         true // non dequantization multiply
     },
 };
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ClampTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/compose_fake_quantize_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/compose_fake_quantize_transformation.cpp
index ce66f1848e0010..247569a9573bf9 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/compose_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/compose_fake_quantize_transformation.cpp
@@ -140,7 +140,7 @@ const std::vector<ComposeFakeQuantizeTransformationParams> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ComposeFakeQuantizeTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_selection_with_intermediate_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_selection_with_intermediate_transformation.cpp
index 8f553ac5fd64ea..cadf373676c82d 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_selection_with_intermediate_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_selection_with_intermediate_transformation.cpp
@@ -173,7 +173,7 @@ const std::vector<TestValues> testValues = {
     }
 };
 
- INSTANTIATE_TEST_CASE_P(
+ INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatSelectionWithIntermediateTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_transformation.cpp
index c199e60fbfd143..fd50c03c1c55f1 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_transformation.cpp
@@ -737,7 +737,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 9, 9 }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_different_precision_on_children.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_different_precision_on_children.cpp
index dc2567fb70dbac..53fddd1e39d123 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_different_precision_on_children.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_different_precision_on_children.cpp
@@ -237,7 +237,7 @@ const std::vector<ConcatTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatWithDifferentChildrenTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_precision_selection_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_precision_selection_transformation.cpp
index 0d6b29d5fe55d4..82afd2c7fe3954 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_precision_selection_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_precision_selection_transformation.cpp
@@ -306,7 +306,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 9, 9 }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatWithIntermediatePrecisionSelectionTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_reshape_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_reshape_transformation.cpp
index ea537db49cfc98..a56c34ce08aefd 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_reshape_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_reshape_transformation.cpp
@@ -133,7 +133,7 @@ const std::vector<TestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatWithIntermediateReshapeTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_transformation.cpp
index 974111bdae8015..5963cbb8fd37a5 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_transformation.cpp
@@ -292,7 +292,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 9, 9 }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatWithIntermediateTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_with_constant_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_with_constant_transformation.cpp
index 7ddf74cd52a0ea..0a6b0cbdc7ccdd 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_with_constant_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_intermediate_with_constant_transformation.cpp
@@ -309,7 +309,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 9, 9 }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatWithIntermediateWithConstantTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_neighbors_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_neighbors_transformation.cpp
index c099b523b39140..7a8098a996b40e 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_neighbors_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_neighbors_transformation.cpp
@@ -344,7 +344,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 9, 9 }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatWithNeighborsTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_reshape_at_the_end_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_reshape_at_the_end_transformation.cpp
index 8f2f17a00f8b77..599e4f2c0bc5af 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_reshape_at_the_end_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_reshape_at_the_end_transformation.cpp
@@ -206,7 +206,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 9, 9 }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatWithReshapeAtTheEndTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_split_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_split_transformation.cpp
index 5f966576594b3d..de12ada6591b87 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_split_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_split_transformation.cpp
@@ -300,7 +300,7 @@ const std::vector<ConcatTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatWithSplitTransformation,
     ::testing::Combine(
@@ -334,7 +334,7 @@ const std::vector<ConcatTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatWithSplitTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_strided_slice_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_strided_slice_transformation.cpp
index f11f20da124404..72dd960dbc406e 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_strided_slice_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/concat_with_strided_slice_transformation.cpp
@@ -273,7 +273,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 4, 9, 9 }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConcatWithStridedSliceTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/convert_mul_or_add_finally_transformation_with_dequantization.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/convert_mul_or_add_finally_transformation_with_dequantization.cpp
index 585f57c707e58b..edb4e813b7fe7a 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/convert_mul_or_add_finally_transformation_with_dequantization.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/convert_mul_or_add_finally_transformation_with_dequantization.cpp
@@ -109,7 +109,7 @@ std::vector<ConvertMulOrAddFinallyTransformationWithDequantizationTestValues> te
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConvertMulOrAddFinallyTransformationWithDequantization,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/convert_subtract_constant_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/convert_subtract_constant_transformation.cpp
index d908a6c6c9f064..235c6f77e26e62 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/convert_subtract_constant_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/convert_subtract_constant_transformation.cpp
@@ -324,7 +324,7 @@ const std::vector<ConvertSubtractConstantTransformationTestValues> testValues =
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConvertSubtractConstantTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_backprop_data_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_backprop_data_transformation.cpp
index 10ba4513f20549..eb62cd89b24436 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_backprop_data_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_backprop_data_transformation.cpp
@@ -445,7 +445,7 @@ const std::vector<ConvolutionBackpropDataTransformationTestValues> testValues =
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConvolutionBackpropDataTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_qdq_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_qdq_transformation.cpp
index 75b1d965e53416..6812b00c1da4cc 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_qdq_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_qdq_transformation.cpp
@@ -495,7 +495,7 @@ const std::vector<ConvolutionQDqTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConvolutionQDqTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_transformation.cpp
index 8c2d42dfbf3c98..bad2201a65dcc9 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_transformation.cpp
@@ -424,7 +424,7 @@ const std::vector<ConvolutionTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConvolutionTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_with_incorrect_weights.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_with_incorrect_weights.cpp
index b85c333928f1d8..7a6b43bc54788c 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_with_incorrect_weights.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/convolution_with_incorrect_weights.cpp
@@ -130,7 +130,7 @@ const std::vector<ConvolutionWIthIncorrectWeightsTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ConvolutionWIthIncorrectWeightsTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/depth_to_space_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/depth_to_space_transformation.cpp
index afbc2346c0f811..8ba97cb1e0ed34 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/depth_to_space_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/depth_to_space_transformation.cpp
@@ -244,7 +244,7 @@ const std::vector<DepthToSpaceTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     DepthToSpaceTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/disable_convert_on_const_path_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/disable_convert_on_const_path_transformation.cpp
index 23396b71cda2bb..5a2fbe0f39aee9 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/disable_convert_on_const_path_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/disable_convert_on_const_path_transformation.cpp
@@ -161,7 +161,7 @@ const std::vector<DisableConvertOnConstPathTransformationValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     DisableConvertOnConstPathTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/elementwise_with_multi_parent_dequantization_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/elementwise_with_multi_parent_dequantization_transformation.cpp
index 6945e3ed1c0243..415301ec850816 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/elementwise_with_multi_parent_dequantization_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/elementwise_with_multi_parent_dequantization_transformation.cpp
@@ -154,7 +154,7 @@ const std::vector<ElementwiseWithMultiParentDequantizationTransformationTestValu
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ElementwiseWithMultiParentDequantizationTransformation,
     ::testing::ValuesIn(addTransformationTestValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp
index 2799cde32cca64..ca4bc3e294c5d2 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp
@@ -161,7 +161,7 @@ const std::vector<ngraph::Shape> shapes = {
     // TODO: 3D tensor
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FakeQuantizeAndTwoOutputBranchesWithConvolutionTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_on_weights_with_unsupported_child.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_on_weights_with_unsupported_child.cpp
index f430adb5974318..b29eff3926e379 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_on_weights_with_unsupported_child.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_on_weights_with_unsupported_child.cpp
@@ -119,7 +119,7 @@ const std::vector<FakeQuantizeOnWeightsWithUnsupportedChildTestValues> testValue
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FakeQuantizeOnWeightsWithUnsupportedChild,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_precision_selection_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_precision_selection_transformation.cpp
index 5bbf9363792d69..c0db76cfd24d7b 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_precision_selection_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_precision_selection_transformation.cpp
@@ -193,7 +193,7 @@ const std::vector<ngraph::Shape> shapes = {
     // TODO: 3D tensor
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FakeQuantizePrecisionSelectionTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_transformation.cpp
index 9ec593b6371502..7154009ff6fd25 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_transformation.cpp
@@ -222,7 +222,7 @@ const std::vector<ngraph::Shape> shapes = {
     // TODO: 3D tensor
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FakeQuantizeTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp
index b057615dcad5d4..a67d10ed8bf164 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp
@@ -205,7 +205,7 @@ const std::vector<ngraph::Shape> shapes = {
     // TODO: 3D tensor
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FakeQuantizeWithNotOptimalTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_with_dynamic_intervals_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_with_dynamic_intervals_transformation.cpp
index e7955f9987153f..77513ca6b92148 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_with_dynamic_intervals_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fake_quantize_with_dynamic_intervals_transformation.cpp
@@ -157,7 +157,7 @@ const std::vector<FakeQuantizeWithDynamicIntervalsTransformationTestValues> fake
 
 const std::vector<ngraph::Shape> shapes = { { 1, 32, 72, 48 } };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FakeQuantizeWithDynamicIntervalsTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fold_convert_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fold_convert_transformation.cpp
index 68b6e16b9a68a4..94a978e582588e 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fold_convert_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fold_convert_transformation.cpp
@@ -165,7 +165,7 @@ const std::vector<FoldConvertTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FoldConvertTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fold_fake_quantize_in_transformations.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fold_fake_quantize_in_transformations.cpp
index fe2b958d2164b7..3be09188b8cdb3 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fold_fake_quantize_in_transformations.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fold_fake_quantize_in_transformations.cpp
@@ -206,7 +206,7 @@ const std::vector<FoldFakeQuantizeInTransformationsTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FoldFakeQuantizeInTransformations,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_convert_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_convert_transformation.cpp
index 626b251ebfc66f..e3854ad12cd990 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_convert_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_convert_transformation.cpp
@@ -180,7 +180,7 @@ TEST_P(FuseConvertTransformation, CompareFunctions) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FuseConvertTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_fake_quantize_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_fake_quantize_transformation.cpp
index 100b166e92e403..6b063c008757dc 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_fake_quantize_transformation.cpp
@@ -342,7 +342,7 @@ const std::vector<FuseFakeQuantizeTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FuseFakeQuantizeTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_fake_quantize_with_multi_inputs_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_fake_quantize_with_multi_inputs_transformation.cpp
index fcbb532a6f1cc8..0f51338a464600 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_fake_quantize_with_multi_inputs_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_fake_quantize_with_multi_inputs_transformation.cpp
@@ -135,7 +135,7 @@ const std::vector<FuseFakeQuantizeTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FuseFakeQuantizeWithMultiInputsTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_multiply_to_fake_quantize_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_multiply_to_fake_quantize_transformation.cpp
index bf0ce1b4484b84..2fe1093cd9420f 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_multiply_to_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_multiply_to_fake_quantize_transformation.cpp
@@ -140,7 +140,7 @@ const std::vector<FuseMultiplyToFakeQuantizeTransformationTestValues> testValues
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FuseMultiplyToFakeQuantizeTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_subtract_to_fake_quantize_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_subtract_to_fake_quantize_transformation.cpp
index 94016eb98f8705..d65c313ebabfc8 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_subtract_to_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/fuse_subtract_to_fake_quantize_transformation.cpp
@@ -217,7 +217,7 @@ const std::vector<FuseSubtractToFakeQuantizeTransformationTestValues> testValues
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     FuseSubtractToFakeQuantizeTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_below_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_below_transformation.cpp
index c92126b3252d12..53a803e88a9852 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_below_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_below_transformation.cpp
@@ -124,7 +124,7 @@ const std::vector<ngraph::Shape> shapes = {
     // TODO: 3D tensor
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     GetDequantizationBelowTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_test.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_test.cpp
index 0260a46467a419..3e6af62a9376fd 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_test.cpp
@@ -78,7 +78,7 @@ TEST_P(GetDequantizationTestTransformation, CompareFunctions) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, GetDequantizationTestTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, GetDequantizationTestTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(isConvert),
         ::testing::ValuesIn(isSubtract),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_transformation.cpp
index 7f527e210929a5..1e69ae86bc8157 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/get_dequantization_transformation.cpp
@@ -155,7 +155,7 @@ const std::vector<ngraph::Shape> shapes = {
     // TODO: 3D tensor
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     GetDequantizationTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/group_convolution_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/group_convolution_transformation.cpp
index 22844037c98d75..e8ba7d4e5f1c26 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/group_convolution_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/group_convolution_transformation.cpp
@@ -702,7 +702,7 @@ const std::vector<GroupConvolutionTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     GroupConvolutionTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/interpolate_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/interpolate_transformation.cpp
index 0e90edef6d7e5f..74a78d350e04a3 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/interpolate_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/interpolate_transformation.cpp
@@ -446,7 +446,7 @@ TEST_P(InterpolateTransformation, CompareFunctions) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     InterpolateTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/is_function_quantized_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/is_function_quantized_transformation.cpp
index 860f7931cf64ca..b66b02eba31e2d 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/is_function_quantized_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/is_function_quantized_transformation.cpp
@@ -109,7 +109,7 @@ const std::vector<IsFunctionQuantizedTransformationValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     IsFunctionQuantizedTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/mat_mul_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/mat_mul_transformation.cpp
index 934326f9573f63..090c46bbd9a444 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/mat_mul_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/mat_mul_transformation.cpp
@@ -357,7 +357,7 @@ std::vector<MatMullTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     MatMulTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/mat_mul_with_constant_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/mat_mul_with_constant_transformation.cpp
index 5c4e171d504847..35e2806fd63bb6 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/mat_mul_with_constant_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/mat_mul_with_constant_transformation.cpp
@@ -592,7 +592,7 @@ std::vector<MatMullTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     MatMulWithConstantTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/max_pool_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/max_pool_transformation.cpp
index c7c3bae73fcf9e..f3c641b9af19b9 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/max_pool_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/max_pool_transformation.cpp
@@ -208,7 +208,7 @@ const std::vector<MaxPoolTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     MaxPoolTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/move_dequantization_after_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/move_dequantization_after_transformation.cpp
index c50375d14df92b..a9106994aa7320 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/move_dequantization_after_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/move_dequantization_after_transformation.cpp
@@ -256,7 +256,7 @@ const std::vector<MoveDequantizationAfterTransformationParams> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     MoveDequantizationAfterTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/mul_add_to_scaleshift_or_power_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/mul_add_to_scaleshift_or_power_transformation.cpp
index 14da96bd21bf94..0b32cedb3515f6 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/mul_add_to_scaleshift_or_power_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/mul_add_to_scaleshift_or_power_transformation.cpp
@@ -135,7 +135,7 @@ const std::vector<MulAddToScaleshiftOrPowerParams> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     MulAddToScaleshiftOrPowerTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/multiply_to_group_convolution_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/multiply_to_group_convolution_transformation.cpp
index a7f4013c315d88..858041bba68897 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/multiply_to_group_convolution_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/multiply_to_group_convolution_transformation.cpp
@@ -312,7 +312,7 @@ TEST_P(MultiplyToGroupConvolutionTransformation, CompareFunctions) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     MultiplyToGroupConvolutionTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/multiply_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/multiply_transformation.cpp
index a17839f8d3c4cd..7441cf2916026b 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/multiply_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/multiply_transformation.cpp
@@ -727,7 +727,7 @@ const std::vector<MultiplyTransformationTestValues> multiplyTransformationTestVa
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     MultiplyTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/mvn_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/mvn_transformation.cpp
index ba90f89d3f4a42..40cb46151dc4a0 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/mvn_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/mvn_transformation.cpp
@@ -298,7 +298,7 @@ TEST_P(MVNTransformation, CompareFunctions) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     MVNTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/normalize_dequantization_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/normalize_dequantization_transformation.cpp
index 63592e5b35ce99..17a8b601f12c6c 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/normalize_dequantization_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/normalize_dequantization_transformation.cpp
@@ -163,7 +163,7 @@ const std::vector<NormalizeDequantizationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     NormalizeDequantizationTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/normalize_l2_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/normalize_l2_transformation.cpp
index 72e1c715ac6de5..6e34ea092d92ea 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/normalize_l2_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/normalize_l2_transformation.cpp
@@ -269,7 +269,7 @@ const std::vector<NormalizeL2TransformationTestValues> normalizeL2Transformation
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     NormalizeL2Transformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/prelu_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/prelu_transformation.cpp
index bbe0e3bd9a2668..66e6a161a7a8de 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/prelu_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/prelu_transformation.cpp
@@ -159,7 +159,7 @@ const std::vector<PReluTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     PReluTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/pull_reshape_through_dequantization_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/pull_reshape_through_dequantization_transformation.cpp
index f00dcc62173953..33a9b90cd54c78 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/pull_reshape_through_dequantization_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/pull_reshape_through_dequantization_transformation.cpp
@@ -349,7 +349,7 @@ const std::vector<PullReshapeThroughDequantizationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     PullReshapeThroughDequantizationTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/pull_transpose_through_dequantization_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/pull_transpose_through_dequantization_transformation.cpp
index 5794ffc55e5d7c..fd459eeb1d1a6a 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/pull_transpose_through_dequantization_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/pull_transpose_through_dequantization_transformation.cpp
@@ -237,7 +237,7 @@ const std::vector<PullTransposeThroughDequantizationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     PullTransposeThroughDequantizationTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_max_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_max_transformation.cpp
index 4d2ab19bdb3e83..23569ae8f8043b 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_max_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_max_transformation.cpp
@@ -301,7 +301,7 @@ const std::vector<ReduceTransformationTestValues> reduceMaxTransformationTestVal
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ReduceMaxTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_mean_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_mean_transformation.cpp
index 5a61b55f21b6fa..1fbfbb4362d373 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_mean_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_mean_transformation.cpp
@@ -269,7 +269,7 @@ const std::vector<ReduceTransformationTestValues> reduceMeanTransformationTestVa
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ReduceMeanTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_min_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_min_transformation.cpp
index 65a01088ab55f0..563a30ebdc6f1d 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_min_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_min_transformation.cpp
@@ -301,7 +301,7 @@ const std::vector<ReduceTransformationTestValues> reduceMinTransformationTestVal
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ReduceMinTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_sum_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_sum_transformation.cpp
index 0dcf43331f00c9..95c9d8b57078d2 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_sum_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/reduce_sum_transformation.cpp
@@ -293,7 +293,7 @@ const std::vector<ReduceTransformationTestValues> reduceSumTransformationTestVal
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ReduceSumTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/relu_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/relu_transformation.cpp
index 090bf1b6aeab16..92ee87caa61123 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/relu_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/relu_transformation.cpp
@@ -234,7 +234,7 @@ const std::vector<ReluTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ReluTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/reshape_fully_connected_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/reshape_fully_connected_transformation.cpp
index a0ad21362ed6d9..9c78d2178973b7 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/reshape_fully_connected_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/reshape_fully_connected_transformation.cpp
@@ -99,7 +99,7 @@ std::vector<ReshapeFullyConnectedTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ReshapeFullyConnectedTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/reshape_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/reshape_transformation.cpp
index c2ff5b59048e44..2ea6cb9f088292 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/reshape_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/reshape_transformation.cpp
@@ -647,7 +647,7 @@ TEST_P(ReshapeTransformation, CompareFunctions) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ReshapeTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/round_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/round_transformation.cpp
index 04ff6e9f0d3700..0612881bff3646 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/round_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/round_transformation.cpp
@@ -102,7 +102,7 @@ TEST_P(RoundTransformation, CompareFunctions) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     LPT,
     RoundTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/separate_in_standalone_branch_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/separate_in_standalone_branch_transformation.cpp
index 04832b837fe4bc..a62be54c87f860 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/separate_in_standalone_branch_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/separate_in_standalone_branch_transformation.cpp
@@ -158,7 +158,7 @@ std::vector<SeparateInStandaloneBranchTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     SeparateInStandaloneBranchTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/shuffle_channels_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/shuffle_channels_transformation.cpp
index 88a205089d6a5d..d70f611b91d5de 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/shuffle_channels_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/shuffle_channels_transformation.cpp
@@ -280,7 +280,7 @@ const std::vector<ShuffleChannelsTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     ShuffleChannelsTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/split_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/split_transformation.cpp
index b08d154c17cbab..89b5b166e8c7d6 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/split_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/split_transformation.cpp
@@ -456,7 +456,7 @@ const std::vector<SplitTransformationTestValues> testValues = {
         { }
     },
 };
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     SplitTransformation,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/squeeze_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/squeeze_transformation.cpp
index a753c972ba3fd8..ffac08817e352f 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/squeeze_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/squeeze_transformation.cpp
@@ -223,7 +223,7 @@ const std::vector<SqueezeTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     SqueezeTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/strided_slice_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/strided_slice_transformation.cpp
index ca8f4a68518ace..028fa55a2cb468 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/strided_slice_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/strided_slice_transformation.cpp
@@ -535,7 +535,7 @@ const std::vector<StridedSliceTransformationTestValues> stridedSliceTransformati
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     StridedSliceTransformation,
     ::testing::ValuesIn(stridedSliceTransformationTestValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/subtract_multiply_to_multiply_add_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/subtract_multiply_to_multiply_add_transformation.cpp
index e7a7ca81fbc679..d854afe64286af 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/subtract_multiply_to_multiply_add_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/subtract_multiply_to_multiply_add_transformation.cpp
@@ -273,7 +273,7 @@ const std::vector<SubtractMultiplyToMultiplyAddTransformationTestValues> testVal
     },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     SubtractMultiplyToMultiplyAddTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/transformations_after_split_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/transformations_after_split_transformation.cpp
index d4e5b3773055be..0795c8d5101697 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/transformations_after_split_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/transformations_after_split_transformation.cpp
@@ -213,7 +213,7 @@ const std::vector<std::string> transformationNames = {
     "SubtractMultiplyToMultiplyAddTransformation",
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     TransformationsAfterSplitTransformation,
     ::testing::ValuesIn(transformationNames),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/transformer_is_function_quantized.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/transformer_is_function_quantized.cpp
index 1ad9e702d182b2..5a00bbc015cb58 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/transformer_is_function_quantized.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/transformer_is_function_quantized.cpp
@@ -79,7 +79,7 @@ const std::vector<TestValues> testValues = {
     { {}, {} }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     TransformerIsFunctionQuantized,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/transpose_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/transpose_transformation.cpp
index de8d2f780b828a..ad07bcd6bc7d2f 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/transpose_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/transpose_transformation.cpp
@@ -202,7 +202,7 @@ TEST_P(TransposeTransformation, CompareFunctions) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     TransposeTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/unsqueeze_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/unsqueeze_transformation.cpp
index 00359326fb0a93..2aa00c5cefdf20 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/unsqueeze_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/unsqueeze_transformation.cpp
@@ -223,7 +223,7 @@ const std::vector<UnsqueezeTransformationTestValues> testValues = {
 },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     UnsqueezeTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/lp_transformations/variadic_split_transformation.cpp b/inference-engine/tests/functional/inference_engine/lp_transformations/variadic_split_transformation.cpp
index 990de4d98d7a56..b5075dfc7f429f 100644
--- a/inference-engine/tests/functional/inference_engine/lp_transformations/variadic_split_transformation.cpp
+++ b/inference-engine/tests/functional/inference_engine/lp_transformations/variadic_split_transformation.cpp
@@ -464,7 +464,7 @@ const std::vector<VariadicSplitTransformationTestValues> testValues = {
         { }
     },
 };
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_LPT,
     VariadicSplitTransformation,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/ngraph_reader/net_reader_test.cpp b/inference-engine/tests/functional/inference_engine/ngraph_reader/net_reader_test.cpp
index 3944a8853823e0..d32b6ab07de619 100644
--- a/inference-engine/tests/functional/inference_engine/ngraph_reader/net_reader_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/ngraph_reader/net_reader_test.cpp
@@ -198,7 +198,7 @@ static const auto params = testing::Combine(
         testing::Values(InferenceEngine::SizeVector{1, 3, 227, 227}),
         testing::Values(InferenceEngine::Precision::FP32, InferenceEngine::Precision::FP16));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         NetReaderTest,
         NetReaderTest,
         params,
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/activation.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/activation.cpp
index ff6a32786c8a6a..fb0c8acd71b806 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/activation.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/activation.cpp
@@ -99,11 +99,11 @@ const auto basicPreluCases = ::testing::Combine(
 );
 
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic_Prelu, ActivationLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic_Prelu, ActivationLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationParamLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic, ActivationParamLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationDynamicLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic, ActivationDynamicLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/batch_norm.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/batch_norm.cpp
index cc1cbf7dff2a73..3f78aea438be00 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/batch_norm.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/batch_norm.cpp
@@ -45,7 +45,7 @@ const auto batchNormParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_BatchNorm_Serialization,
     BatchNormLayerTest,
     batchNormParams,
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/binary_convolution.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/binary_convolution.cpp
index 4ffae87a5706df..aa4f5c9eb3f5a5 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/binary_convolution.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/binary_convolution.cpp
@@ -37,7 +37,7 @@ const auto binConv2DParams_AutoPadValid = ::testing::Combine(
     ::testing::Values(ngraph::op::PadType::VALID),
     ::testing::ValuesIn(pad_values));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_BinaryConvolution2D_Serialization_ExplicitPadding, BinaryConvolutionLayerTest,
     ::testing::Combine(
         binConv2DParams_ExplicitPadding, ::testing::ValuesIn(netPrecisions),
@@ -49,7 +49,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     BinaryConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_BinaryConvolution2D__Serialization_AutoPadValid, BinaryConvolutionLayerTest,
     ::testing::Combine(
         binConv2DParams_AutoPadValid, ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/broadcast.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/broadcast.cpp
index 1e101775319f01..0bfc51fa9831f9 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/broadcast.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/broadcast.cpp
@@ -27,7 +27,7 @@ const auto numpyBroadcastParams1 = ::testing::Combine(
     ::testing::Values(inShapesNumpy[0]), ::testing::ValuesIn(inputPrecisions),
     ::testing::Values(CommonTestUtils::DEVICE_CPU));
 
-INSTANTIATE_TEST_CASE_P(smoke_Broadcast1Serialization, BroadcastLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Broadcast1Serialization, BroadcastLayerTest,
                         numpyBroadcastParams1,
                         BroadcastLayerTest::getTestCaseName);
 } // namespace
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/clamp.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/clamp.cpp
index 607989781e6f72..7bff8a3359a1cb 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/clamp.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/clamp.cpp
@@ -22,7 +22,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 const std::vector<std::pair<float, float>> intervals = {
     {-20.1, -10.5}, {-10.0, 10.0}, {10.3, 20.4}};
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Clamp_Serialization, ClampLayerTest,
     ::testing::Combine(
         ::testing::ValuesIn(inShapes),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/constant.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/constant.cpp
index 4d4cce07bf0d0e..0fc324f2bd6492 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/constant.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/constant.cpp
@@ -41,13 +41,13 @@ std::vector<InferenceEngine::Precision> precisionsWithNegativeValues = {
 std::vector<std::string> dataWithNegativeValues = {"1", "-2", "3", "-4", "5", "-6",
                                                    "7", "-1", "2", "-3", "4", "-5"};
 
-INSTANTIATE_TEST_CASE_P(smoke_Constant_Serialization, ConstantLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Constant_Serialization, ConstantLayerTest,
                         ::testing::Combine(::testing::ValuesIn(shapes),
                                            ::testing::ValuesIn(precisions), ::testing::Values(data),
                                            ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConstantLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Constant_Serialization_with_negative_values, ConstantLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Constant_Serialization_with_negative_values, ConstantLayerTest,
                         ::testing::Combine(::testing::ValuesIn(shapes),
                                            ::testing::ValuesIn(precisionsWithNegativeValues),
                                            ::testing::Values(dataWithNegativeValues),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/convert.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/convert.cpp
index e6d6bc19ed7d6d..7cc308c41e287c 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/convert.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/convert.cpp
@@ -25,7 +25,7 @@ TEST_P(ConvertLayerTest, Serialize) {
     Serialize();
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Serialization_ConvertLayerTest, ConvertLayerTest,
     ::testing::Combine(::testing::Values(inShape),
                        ::testing::ValuesIn(precisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution.cpp
index 7e22ba3e388df6..f68db9c3b08a05 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution.cpp
@@ -36,7 +36,7 @@ const auto conv2DParams_AutoPadValid = ::testing::Combine(
     ::testing::ValuesIn(dilations), ::testing::ValuesIn(numOutChannels),
     ::testing::Values(ngraph::op::PadType::VALID));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Convolution2D_Serialization_ExplicitPadding, ConvolutionLayerTest,
     ::testing::Combine(
         conv2DParams_ExplicitPadding, ::testing::ValuesIn(netPrecisions),
@@ -48,7 +48,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Convolution2D__Serialization_AutoPadValid, ConvolutionLayerTest,
     ::testing::Combine(
         conv2DParams_AutoPadValid, ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution_backprop.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution_backprop.cpp
index 2f9383ad914475..0e29deb207e8f1 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution_backprop.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/convolution_backprop.cpp
@@ -41,7 +41,7 @@ const auto convolutionBackpropData2DParams = ::testing::Combine(
     ::testing::ValuesIn(dilations), ::testing::ValuesIn(numOutChannels),
     ::testing::ValuesIn(pad_types), ::testing::ValuesIn(outPadding));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_convolutionBackpropData2D_Serialization, ConvolutionBackpropLayerTest,
     ::testing::Combine(
         convolutionBackpropData2DParams,
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/deformable_convolution.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/deformable_convolution.cpp
index a73f583f8b87c7..09107a33cd576d 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/deformable_convolution.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/deformable_convolution.cpp
@@ -42,7 +42,7 @@ const auto conv2DParams_AutoPadValid = ::testing::Combine(
     ::testing::ValuesIn(defor_groups), ::testing::ValuesIn(numOutChannels),
     ::testing::Values(ngraph::op::PadType::VALID));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_DeformableConvolution2D_Serialization_ExplicitPadding, DeformableConvolutionLayerTest,
     ::testing::Combine(
         conv2DParams_ExplicitPadding, ::testing::ValuesIn(netPrecisions),
@@ -54,7 +54,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     DeformableConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_DeformableConvolution2D__Serialization_AutoPadValid, DeformableConvolutionLayerTest,
     ::testing::Combine(
         conv2DParams_AutoPadValid, ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/deformable_psroi_pooling.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/deformable_psroi_pooling.cpp
index ffdedd39905a7b..c4606cf7dec44c 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/deformable_psroi_pooling.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/deformable_psroi_pooling.cpp
@@ -30,6 +30,6 @@ namespace {
         ::testing::Values(InferenceEngine::Precision::FP32),   // Net precision
         ::testing::Values(CommonTestUtils::DEVICE_CPU));       // Device name
 
-    INSTANTIATE_TEST_CASE_P(smoke_TestsDeformablePSROIPooling, DeformablePSROIPoolingLayerTest, deformablePSROICases_test_params,
+    INSTANTIATE_TEST_SUITE_P(smoke_TestsDeformablePSROIPooling, DeformablePSROIPoolingLayerTest, deformablePSROICases_test_params,
                             DeformablePSROIPoolingLayerTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/detection_output.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/detection_output.cpp
index e1f66b84c22276..00d5382d6fe32f 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/detection_output.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/detection_output.cpp
@@ -66,7 +66,7 @@ namespace {
         ::testing::Values(objectnessScore),
         ::testing::Values(CommonTestUtils::DEVICE_CPU));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_DetectionOutput3In,
     DetectionOutputLayerTest,
     params3Inputs,
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/elementwise.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/elementwise.cpp
index 7d2ac7f47a1957..1a131fffaf52ec 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/elementwise.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/elementwise.cpp
@@ -57,7 +57,7 @@ const auto elementiwiseParams = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(additionalConfig));
 
-INSTANTIATE_TEST_CASE_P(smoke_ElementwiseSerialization, EltwiseLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ElementwiseSerialization, EltwiseLayerTest,
                         elementiwiseParams,
                         EltwiseLayerTest::getTestCaseName);
 } // namespace
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/extractimagepatches.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/extractimagepatches.cpp
index 7e48a0ad0e93a7..6eda8ccde991be 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/extractimagepatches.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/extractimagepatches.cpp
@@ -27,7 +27,7 @@ const std::vector<InferenceEngine::Precision> netPrecision = {
     InferenceEngine::Precision::FP32
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ExtractImagePatchesLayerTest, ExtractImagePatchesTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ExtractImagePatchesLayerTest, ExtractImagePatchesTest,
         ::testing::Combine(::testing::ValuesIn(inShapes),
                             ::testing::ValuesIn(kSizes),
                             ::testing::ValuesIn(strides),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/group_convolution.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/group_convolution.cpp
index ca080890ee37b0..66b5ec2684feb3 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/group_convolution.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/group_convolution.cpp
@@ -40,7 +40,7 @@ const auto groupConv2DParams = ::testing::Combine(
     ::testing::ValuesIn(dilations), ::testing::ValuesIn(numOutChannels),
     ::testing::ValuesIn(numGroups), ::testing::ValuesIn(pad_types));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_GroupConvolution2D_Serialization, GroupConvolutionLayerTest,
     ::testing::Combine(
         groupConv2DParams, ::testing::ValuesIn(precisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/group_convolution_backprop_data.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/group_convolution_backprop_data.cpp
index d98ff4016e4bd0..df5596e2363d6d 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/group_convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/group_convolution_backprop_data.cpp
@@ -40,7 +40,7 @@ const auto groupConvBackpropData2DParams = ::testing::Combine(
     ::testing::ValuesIn(dilations), ::testing::ValuesIn(numOutChannels),
     ::testing::ValuesIn(numGroups), ::testing::ValuesIn(pad_types));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_GroupConvBackpropData2D_Serialization, GroupConvBackpropDataLayerTest,
     ::testing::Combine(
         groupConvBackpropData2DParams,
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/minimum_maximum.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/minimum_maximum.cpp
index 19417dc8e0f64c..d64b7cf4e61915 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/minimum_maximum.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/minimum_maximum.cpp
@@ -36,7 +36,7 @@ const std::vector<ngraph::helpers::InputLayerType> inputType = {
         ngraph::helpers::InputLayerType::PARAMETER,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_maximum, MaxMinLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_maximum, MaxMinLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inShapes),
                                 ::testing::ValuesIn(opType),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/mvn.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/mvn.cpp
index cc8f5728d31322..b5a3cf828c79b7 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/mvn.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/mvn.cpp
@@ -31,7 +31,7 @@ TEST_P(MvnLayerTest, Serialize) {
     Serialize();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsMVN, MvnLayerTest, MvnCases,
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsMVN, MvnLayerTest, MvnCases,
                         MvnLayerTest::getTestCaseName);
 
 // ------------------- MVN-6 -------------------------------------------------
@@ -47,7 +47,7 @@ TEST_P(Mvn6LayerTest, Serialize) {
     Serialize();
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_MVN_5D, Mvn6LayerTest,
     ::testing::Combine(::testing::ValuesIn(std::vector<std::vector<size_t>>{
                            {1, 10, 5, 7, 8}, {1, 3, 8, 9, 49}}),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/non_max_suppression.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/non_max_suppression.cpp
index 633b69b02bdd54..510d8b860e6dd3 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/non_max_suppression.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/non_max_suppression.cpp
@@ -52,6 +52,6 @@ namespace {
             ::testing::ValuesIn(outType),
             ::testing::Values(CommonTestUtils::DEVICE_CPU));
 
-    INSTANTIATE_TEST_CASE_P(smoke_NmsLayerTest, NmsLayerTest, nmsParams, NmsLayerTest::getTestCaseName);
+    INSTANTIATE_TEST_SUITE_P(smoke_NmsLayerTest, NmsLayerTest, nmsParams, NmsLayerTest::getTestCaseName);
 }  // namespace
 
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/nonzero.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/nonzero.cpp
index 2d0c7453ec9fd2..a10a2c52510b66 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/nonzero.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/nonzero.cpp
@@ -25,7 +25,7 @@ std::vector<InferenceEngine::Precision> inputPrecisions = {
 
 ConfigMap config;
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_NonZeroLayerTest, NonZeroLayerTest,
     ::testing::Combine(::testing::ValuesIn(inputDims),
                        ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/normalize_l2.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/normalize_l2.cpp
index 68b5ce21bab004..cf56b7cbacdd4d 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/normalize_l2.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/normalize_l2.cpp
@@ -30,7 +30,7 @@ const std::vector<Precision> netPrecisions = {
     Precision::BF16
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NormalizeL2Serialization, NormalizeL2LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NormalizeL2Serialization, NormalizeL2LayerTest,
         testing::Combine(
             testing::ValuesIn(axes),
             testing::ValuesIn(eps),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/one_hot.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/one_hot.cpp
index 7a013aa7b55d13..42a80dd60b7756 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/one_hot.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/one_hot.cpp
@@ -34,6 +34,6 @@ const auto oneHotParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_OneHotConstSerialization, OneHotLayerTest, oneHotParams,
+INSTANTIATE_TEST_SUITE_P(smoke_OneHotConstSerialization, OneHotLayerTest, oneHotParams,
                         OneHotLayerTest::getTestCaseName);
 } // namespace
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/pad.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/pad.cpp
index 6a3a9a484ac815..6035d38f1752cd 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/pad.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/pad.cpp
@@ -35,6 +35,6 @@ const auto pad2DConstparams = testing::Combine(
     testing::Values(std::vector<size_t>{13, 5}),
     testing::Values(CommonTestUtils::DEVICE_CPU));
 
-INSTANTIATE_TEST_CASE_P(smoke_Pad2DConstSerialization, PadLayerTest, pad2DConstparams,
+INSTANTIATE_TEST_SUITE_P(smoke_Pad2DConstSerialization, PadLayerTest, pad2DConstparams,
                         PadLayerTest::getTestCaseName);
 } // namespace
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/pooling.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/pooling.cpp
index c1a05395e17759..df14f0bac5d541 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/pooling.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/pooling.cpp
@@ -55,7 +55,7 @@ namespace {
             ::testing::ValuesIn(padTypes),
             ::testing::Values(excludePad[0]));
 
-    INSTANTIATE_TEST_CASE_P(smoke_AvgPoolExcluePad, PoolingLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_AvgPoolExcluePad, PoolingLayerTest,
             ::testing::Combine(
                 avgExcludePadParams,
                 ::testing::ValuesIn(netPrecisions),
@@ -77,7 +77,7 @@ namespace {
             ::testing::ValuesIn(padTypes),
             ::testing::Values(excludePad[1]));
 
-    INSTANTIATE_TEST_CASE_P(smoke_AvgPool, PoolingLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_AvgPool, PoolingLayerTest,
             ::testing::Combine(
                 avgPadParams,
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/prelu.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/prelu.cpp
index d00123b4eb13ef..35d708bd9a1031 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/prelu.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/prelu.cpp
@@ -38,6 +38,6 @@ const auto basicCases = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ActivationPreluSerialization,
+INSTANTIATE_TEST_SUITE_P(smoke_ActivationPreluSerialization,
                         ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/prior_box_clustered.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/prior_box_clustered.cpp
index dd0f2d5144d5a9..f59e469660b56d 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/prior_box_clustered.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/prior_box_clustered.cpp
@@ -61,7 +61,7 @@ namespace {
         ::testing::ValuesIn(offsets),
         ::testing::ValuesIn(variances));
 
-    INSTANTIATE_TEST_CASE_P(smoke_PriorBoxClustered_Basic, PriorBoxClusteredLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_PriorBoxClustered_Basic, PriorBoxClusteredLayerTest,
             ::testing::Combine(
                 layerSpeficParams,
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/reduce_ops.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/reduce_ops.cpp
index 8f28fc7f72b2af..61b0353c83650a 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/reduce_ops.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/reduce_ops.cpp
@@ -140,35 +140,35 @@ const auto params_ReductionTypesLogical = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceOneAxis_Serialization,
         ReduceOpsLayerTest,
         paramsOneAxis,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceLogicalOneAxis_Serialization,
         ReduceOpsLayerTest,
         paramsOneAxisLogical,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceAxes_Serialization,
         ReduceOpsLayerTest,
         params_Axes,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_ReductionTypes_Serialization,
         ReduceOpsLayerTest,
         params_ReductionTypes,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceLogical_ReductionTypes_Serialization,
         ReduceOpsLayerTest,
         params_ReductionTypesLogical,
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/region_yolo.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/region_yolo.cpp
index 228fbb33192ca1..c5bb56e763bed1 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/region_yolo.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/region_yolo.cpp
@@ -33,7 +33,7 @@ namespace {
     const int start_axis = 1;
     const int end_axis = 3;
 
-    INSTANTIATE_TEST_CASE_P(smoke_RegionYolov3Serialization, RegionYoloLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_RegionYolov3Serialization, RegionYoloLayerTest,
     ::testing::Combine(
         ::testing::ValuesIn(inShapes_v3),
         ::testing::Values(classes[0]),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/reshape.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/reshape.cpp
index 9ba9aec704b19d..96338ae40de2f2 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/reshape.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/reshape.cpp
@@ -19,7 +19,7 @@ namespace {
             InferenceEngine::Precision::FP16
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_ReshapeSerialization, ReshapeLayerTestRevise,
+    INSTANTIATE_TEST_SUITE_P(smoke_ReshapeSerialization, ReshapeLayerTestRevise,
             ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/result.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/result.cpp
index aa7e365bc79cec..174779da42a9b2 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/result.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/result.cpp
@@ -25,7 +25,7 @@ std::vector<InferenceEngine::Precision> inputPrecisions = {
 
 ConfigMap config;
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_ResultLayerTest, ResultLayerTest,
     ::testing::Combine(::testing::ValuesIn(inputDims),
                        ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/scatter_nd_update.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/scatter_nd_update.cpp
index 668eeadcd8a43e..95fcf54112337a 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/scatter_nd_update.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/scatter_nd_update.cpp
@@ -42,7 +42,7 @@ const auto ScatterNDUpdateCases = ::testing::Combine(
     ::testing::ValuesIn(idxPrecisions),
     ::testing::Values(CommonTestUtils::DEVICE_CPU));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_ScatterNDUpdateLayerTestSerialization,
     ScatterNDUpdateLayerTest,
     ScatterNDUpdateCases,
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/shuffle_channels.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/shuffle_channels.cpp
index e66352547e4d70..0dc0411d1d749c 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/shuffle_channels.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/shuffle_channels.cpp
@@ -35,7 +35,7 @@ const auto shuffleChannelsParamsNegativeAxis4D = ::testing::Combine(
         ::testing::ValuesIn(groups)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannelsSerialization, ShuffleChannelsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannelsSerialization, ShuffleChannelsLayerTest,
         ::testing::Combine(
                 shuffleChannelsParams4D,
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/space_to_batch.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/space_to_batch.cpp
index 1b55e2fbac71f1..d5293dc50500ac 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/space_to_batch.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/space_to_batch.cpp
@@ -31,7 +31,7 @@ const auto SpaceToBatch4D = ::testing::Combine(
     ::testing::Values(InferenceEngine::Layout::ANY),
     ::testing::Values(CommonTestUtils::DEVICE_CPU));
 
-INSTANTIATE_TEST_CASE_P(smoke_spacetobatch4D_Serialization,
+INSTANTIATE_TEST_SUITE_P(smoke_spacetobatch4D_Serialization,
                         SpaceToBatchLayerTest, SpaceToBatch4D,
                         SpaceToBatchLayerTest::getTestCaseName);
 
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/split.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/split.cpp
index 80b1ca17ecacad..56e56900beb39b 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/split.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/split.cpp
@@ -20,7 +20,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::U16,
         InferenceEngine::Precision::BOOL};
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Split_Serialization, SplitLayerTest,
     ::testing::Combine(
         ::testing::Values(1, 2, 5, 10),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/squeeze.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/squeeze.cpp
index 126f01f2a0d4c7..7164d4da8be974 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/squeeze.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/squeeze.cpp
@@ -32,7 +32,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 const std::vector<ngraph::helpers::SqueezeOpType> opTypes = {
     ngraph::helpers::SqueezeOpType::SQUEEZE};
 
-INSTANTIATE_TEST_CASE_P(smoke_Squeeze_Basic, SqueezeUnsqueezeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Squeeze_Basic, SqueezeUnsqueezeLayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(CommonTestUtils::combineParams(axesVectors)),
                             ::testing::ValuesIn(opTypes),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/tensor_iterator.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/tensor_iterator.cpp
index 5c2ba098b0567f..b76879ef12fcee 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/tensor_iterator.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/tensor_iterator.cpp
@@ -27,7 +27,7 @@ const std::vector<float> clip = {0.f};
 const std::vector<ngraph::op::RecurrentSequenceDirection> direction = {
     ngraph::op::RecurrentSequenceDirection::FORWARD, ngraph::op::RecurrentSequenceDirection::REVERSE};
 
-INSTANTIATE_TEST_CASE_P(smoke_TensorIterator, TensorIteratorTest,
+INSTANTIATE_TEST_SUITE_P(smoke_TensorIterator, TensorIteratorTest,
     ::testing::Combine(
         ::testing::ValuesIn(decompose),
         ::testing::ValuesIn(sequenceLength),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/transpose.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/transpose.cpp
index 4fd458c12441b1..8b9b2be7bd1b47 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/transpose.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/transpose.cpp
@@ -23,7 +23,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 std::vector<std::vector<size_t>> inputShape2D = {{2, 10}, {10, 2}, {10, 10}};
 std::vector<std::vector<size_t>> order2D      = {{}, {0, 1}, {1, 0}};
 
-INSTANTIATE_TEST_CASE_P(smoke_Transpose2D, TransposeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Transpose2D, TransposeLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(order2D),
                 ::testing::ValuesIn(netPrecisions),
@@ -40,7 +40,7 @@ std::vector<std::vector<size_t>> order4D      = {
         {}, {0, 1, 2, 3}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Transpose4D, TransposeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Transpose4D, TransposeLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(order4D),
                 ::testing::ValuesIn(netPrecisions),
@@ -57,7 +57,7 @@ std::vector<std::vector<size_t>> order5D      = {
         {}, {0, 1, 2, 3, 4}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Transpose5D, TransposeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Transpose5D, TransposeLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(order5D),
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/unsqueeze.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/unsqueeze.cpp
index 2e26a70d6274c7..bd62fd2dcc4c1c 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/unsqueeze.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/unsqueeze.cpp
@@ -32,7 +32,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 const std::vector<ngraph::helpers::SqueezeOpType> opTypes = {
     ngraph::helpers::SqueezeOpType::UNSQUEEZE};
 
-INSTANTIATE_TEST_CASE_P(smoke_Squeeze_Basic, SqueezeUnsqueezeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Squeeze_Basic, SqueezeUnsqueezeLayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(CommonTestUtils::combineParams(axesVectors)),
                             ::testing::ValuesIn(opTypes),
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/variadic_split.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/variadic_split.cpp
index 117d7ccbb52f4e..ed8b1edea41317 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/variadic_split.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/variadic_split.cpp
@@ -24,7 +24,7 @@ namespace {
             {1, 16, 5, 8},
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_VariadicSplitSerialization, VariadicSplitLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_VariadicSplitSerialization, VariadicSplitLayerTest,
             ::testing::Combine(
             ::testing::ValuesIn(numSplits),
             ::testing::Values(0, 1, 2, 3),
diff --git a/inference-engine/tests/functional/inference_engine/task_executor_tests.cpp b/inference-engine/tests/functional/inference_engine/task_executor_tests.cpp
index d3ffda26614b47..c4fe68a87b7e15 100644
--- a/inference-engine/tests/functional/inference_engine/task_executor_tests.cpp
+++ b/inference-engine/tests/functional/inference_engine/task_executor_tests.cpp
@@ -198,7 +198,7 @@ static auto Executors = ::testing::Values(
     }
 );
 
-INSTANTIATE_TEST_CASE_P(TaskExecutorTests, TaskExecutorTests, Executors);
+INSTANTIATE_TEST_SUITE_P(TaskExecutorTests, TaskExecutorTests, Executors);
 
 static auto AsyncExecutors = ::testing::Values(
     [] {
@@ -209,5 +209,5 @@ static auto AsyncExecutors = ::testing::Values(
     }
 );
 
-INSTANTIATE_TEST_CASE_P(ASyncTaskExecutorTests, ASyncTaskExecutorTests, AsyncExecutors);
+INSTANTIATE_TEST_SUITE_P(ASyncTaskExecutorTests, ASyncTaskExecutorTests, AsyncExecutors);
 
diff --git a/inference-engine/tests/functional/inference_engine/transformations/broadcast_elementwise_fusion_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/broadcast_elementwise_fusion_test.cpp
index 324bfae029b890..a9667226068d27 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/broadcast_elementwise_fusion_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/broadcast_elementwise_fusion_test.cpp
@@ -238,32 +238,32 @@ TEST_P(NoEliminateDynamicBroadcastTest, CompareFunctions) {
     eliminate_broadcast_test(f, f_ref);
 }
 
-INSTANTIATE_TEST_CASE_P(EliminateBroadcast, EliminateBroadcastTest,
+INSTANTIATE_TEST_SUITE_P(EliminateBroadcast, EliminateBroadcastTest,
                         testing::Values(std::make_tuple(InputShape{1, 2, 3}, InputShape{1, 2, 3}, TargetShape{1, 2, 3}),
                                         std::make_tuple(InputShape{DYN, 2, 3}, InputShape{1, 2, 3}, TargetShape{1, 2, 3}),
                                         std::make_tuple(InputShape{DYN, DYN, DYN}, InputShape{1, 1, 1}, TargetShape{1, 1, 1}),
                                         std::make_tuple(InputShape{1, 2, 3}, InputShape{2, 3}, TargetShape{2, 3}),
                                         std::make_tuple(InputShape{1, 2, 1}, InputShape{1}, TargetShape{1})));
 
-INSTANTIATE_TEST_CASE_P(EliminateBroadcastSwapInputs, EliminateBroadcastSwapInputsTest,
+INSTANTIATE_TEST_SUITE_P(EliminateBroadcastSwapInputs, EliminateBroadcastSwapInputsTest,
                         testing::Values(std::make_tuple(InputShape{1, 2, 3}, InputShape{1, 2, 3}, TargetShape{1, 2, 3}),
                                         std::make_tuple(InputShape{DYN, 2, 3}, InputShape{1, 2, 3}, TargetShape{1, 2, 3}),
                                         std::make_tuple(InputShape{DYN, DYN, DYN}, InputShape{1, 1, 1}, TargetShape{1, 1, 1}),
                                         std::make_tuple(InputShape{1, 2, 3}, InputShape{2, 3}, TargetShape{2, 3}),
                                         std::make_tuple(InputShape{1, 2, 1}, InputShape{1}, TargetShape{1})));
 
-INSTANTIATE_TEST_CASE_P(NoEliminateBroadcast, NoEliminateBroadcastTest,
+INSTANTIATE_TEST_SUITE_P(NoEliminateBroadcast, NoEliminateBroadcastTest,
                         testing::Values(std::make_tuple(InputShape{1, 2, 1}, InputShape{3}, TargetShape{3}),
                                         std::make_tuple(InputShape{DYN, 2, 3}, InputShape{3, 2, 3}, TargetShape{3, 2, 3}),
                                         std::make_tuple(InputShape{DYN, DYN, DYN}, InputShape{3, 2, 1}, TargetShape{3, 2, 1}),
                                         std::make_tuple(ngraph::PartialShape::dynamic(), InputShape{1, 2, 3}, TargetShape{1, 2, 3}),
                                         std::make_tuple(ngraph::PartialShape::dynamic(), ngraph::PartialShape::dynamic(), TargetShape{1, 2, 3})));
 
-INSTANTIATE_TEST_CASE_P(EliminateDynamicBroadcast, EliminateDynamicBroadcastTest,
+INSTANTIATE_TEST_SUITE_P(EliminateDynamicBroadcast, EliminateDynamicBroadcastTest,
                         testing::Values(std::make_tuple(InputShape{2, 2, 4}, InputShape{2, DYN, 4}, InputShape{2, DYN, 4}, InputShape{2, DYN, 4}),
                                         std::make_tuple(InputShape{2, 2, 4}, InputShape{DYN, DYN, DYN}, InputShape{DYN, DYN, DYN}, InputShape{DYN, DYN, DYN}),
                                         std::make_tuple(InputShape{2, 2, 4}, InputShape{2, 2, 4}, InputShape{2, DYN, 4}, InputShape{2, 2, 4})));
 
-INSTANTIATE_TEST_CASE_P(NoEliminateDynamicBroadcast, NoEliminateDynamicBroadcastTest,
+INSTANTIATE_TEST_SUITE_P(NoEliminateDynamicBroadcast, NoEliminateDynamicBroadcastTest,
                         testing::Values(std::make_tuple(InputShape{2, 1, 4}, InputShape{2, DYN, 4}, InputShape{2, DYN, 4}),
                                         std::make_tuple(InputShape{2, DYN, 4}, InputShape{2, DYN, 4}, InputShape{2, DYN, 4})));
diff --git a/inference-engine/tests/functional/inference_engine/transformations/conv_fusion_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/conv_fusion_test.cpp
index b12a0f6f9de0e6..9610df3e6ba018 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/conv_fusion_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/conv_fusion_test.cpp
@@ -125,7 +125,7 @@ TEST_P(ConvFusionTests, CompareFunctions) {
 using add = ngraph::opset5::Add;
 using mul = ngraph::opset5::Multiply;
 
-INSTANTIATE_TEST_CASE_P(ConvAddFusion, ConvFusionTests,
+INSTANTIATE_TEST_SUITE_P(ConvAddFusion, ConvFusionTests,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, WeightsShape{8, 3, 1, 2, 3}, add::type_info, EltwiseShape{8, 1, 1, 1}, false),
                         std::make_tuple(InputShape{DYN, 3, 64, 64, 64}, WeightsShape{8, 3, 1, 2, 3}, add::type_info, EltwiseShape{8, 1, 1, 1}, false),
                         std::make_tuple(InputShape{2, DYN, 64, 64, 64}, WeightsShape{9, 3, 2, 3, 1}, add::type_info, EltwiseShape{9, 1, 1, 1}, false),
@@ -143,12 +143,12 @@ INSTANTIATE_TEST_CASE_P(ConvAddFusion, ConvFusionTests,
                         std::make_tuple(InputShape{2, DYN, 9},          WeightsShape{2, 3, 2},       add::type_info, EltwiseShape{2, 1}, false),
                         std::make_tuple(InputShape{3, 3, DYN},          WeightsShape{1, 3, 3},       add::type_info, EltwiseShape{1, 1}, false)));
 
-INSTANTIATE_TEST_CASE_P(DISABLED_ConvAddFusionNegative, ConvFusionTests,
+INSTANTIATE_TEST_SUITE_P(DISABLED_ConvAddFusionNegative, ConvFusionTests,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, WeightsShape{8, 3, 1, 2, 3}, add::type_info, EltwiseShape{2, 1}, true),
                         std::make_tuple(InputShape{DYN, 3, 64, 64, 64}, WeightsShape{8, 3, 1, 2, 3}, add::type_info, EltwiseShape{8, 1, 1}, true),
                         std::make_tuple(InputShape{2, DYN, 64, 64, 64}, WeightsShape{9, 3, 2, 3, 1}, add::type_info, EltwiseShape{9, 1, 1, 1, 1}, true)));
 
-INSTANTIATE_TEST_CASE_P(DISABLED_ConvMulFusion, ConvFusionTests,
+INSTANTIATE_TEST_SUITE_P(DISABLED_ConvMulFusion, ConvFusionTests,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, WeightsShape{8, 3, 1, 2, 3}, mul::type_info, EltwiseShape{8, 1, 1, 1}, false),
                         std::make_tuple(InputShape{DYN, 3, 64, 64, 64}, WeightsShape{8, 3, 1, 2, 3}, mul::type_info, EltwiseShape{8, 1, 1, 1}, false),
                         std::make_tuple(InputShape{2, DYN, 64, 64, 64}, WeightsShape{9, 3, 2, 3, 1}, mul::type_info, EltwiseShape{9, 1, 1, 1}, false),
@@ -166,7 +166,7 @@ INSTANTIATE_TEST_CASE_P(DISABLED_ConvMulFusion, ConvFusionTests,
                         std::make_tuple(InputShape{2, DYN, 9},          WeightsShape{2, 3, 2},       mul::type_info, EltwiseShape{2, 1}, false),
                         std::make_tuple(InputShape{3, 3, DYN},          WeightsShape{1, 3, 3},       mul::type_info, EltwiseShape{1, 1}, false)));
 
-INSTANTIATE_TEST_CASE_P(DISABLED_ConvMulFusionNegative, ConvFusionTests,
+INSTANTIATE_TEST_SUITE_P(DISABLED_ConvMulFusionNegative, ConvFusionTests,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, WeightsShape{8, 3, 1, 2, 3}, mul::type_info, EltwiseShape{2, 1}, true),
                         std::make_tuple(InputShape{DYN, 3, 64, 64}, WeightsShape{8, 3, 1, 2, 3}, mul::type_info, EltwiseShape{8, 1, 1}, true),
                         std::make_tuple(InputShape{2, DYN, 64, 64}, WeightsShape{9, 3, 2, 3, 1}, mul::type_info, EltwiseShape{9, 1, 1, 1, 1}, true)));
diff --git a/inference-engine/tests/functional/inference_engine/transformations/convert_broadcast3_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/convert_broadcast3_test.cpp
index 7058f52eb5ce42..95b54fe0461174 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/convert_broadcast3_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/convert_broadcast3_test.cpp
@@ -223,7 +223,7 @@ TEST_P(ConvertBroadcast3BIDIRECTBroadcastLogicalOrTest, CompareFunctions) {
     convert_broadcast3_test(f, f_ref);
 }
 
-INSTANTIATE_TEST_CASE_P(ConvertBroadcast3NUMPY, ConvertBroadcast3NUMPYTest,
+INSTANTIATE_TEST_SUITE_P(ConvertBroadcast3NUMPY, ConvertBroadcast3NUMPYTest,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, TargetShape{1, 2, 3, 4, 5}),
                         std::make_tuple(InputShape{DYN, 3, 64, 64, 64},      TargetShape{8, 3, 64, 64, 64}),
                         std::make_tuple(InputShape{2, DYN, 64, 64, 64},      TargetShape{2, 3, 64, 64, 64}),
@@ -241,7 +241,7 @@ INSTANTIATE_TEST_CASE_P(ConvertBroadcast3NUMPY, ConvertBroadcast3NUMPYTest,
                         std::make_tuple(InputShape{2, DYN, 9},          TargetShape{2, 3, 9}),
                         std::make_tuple(InputShape{3, 3, DYN},          TargetShape{3, 3, 3})));
 
-INSTANTIATE_TEST_CASE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTMulTest,
+INSTANTIATE_TEST_SUITE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTMulTest,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, TargetShape{1, 2, 3, 4, 5}),
                         std::make_tuple(InputShape{DYN, 3, 64, 64, 64},      TargetShape{1, 3, 64, 64, 64}),
                         std::make_tuple(InputShape{2, DYN, 64, 64, 64},      TargetShape{2, 1, 64, 64, 64}),
@@ -261,7 +261,7 @@ INSTANTIATE_TEST_CASE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTMulT
                         std::make_tuple(InputShape{2, DYN, 9},          TargetShape{2, 1, 9}),
                         std::make_tuple(InputShape{3, 3, DYN},          TargetShape{3, 3, 1})));
 
-INSTANTIATE_TEST_CASE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTBroadcastTest,
+INSTANTIATE_TEST_SUITE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTBroadcastTest,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, TargetShape{2, 2, 3, 4, 5},    TargetShape{2, 2, 3, 4, 5}),
                         std::make_tuple(InputShape{DYN, 3, 64, 64, 64},      TargetShape{3, 3, 64, 64, 64}, TargetShape{3, 3, 64, 64, 64}),
                         std::make_tuple(InputShape{2, DYN, 64, 64, 64},      TargetShape{2, 3, 64, 64, 1},  TargetShape{2, 3, 64, 64, 64}),
@@ -280,7 +280,7 @@ INSTANTIATE_TEST_CASE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTBroa
                         std::make_tuple(InputShape{2, DYN, 9},          TargetShape{2, 2, 1},       TargetShape{2, 2, 9}),
                         std::make_tuple(InputShape{3, 3, DYN},          TargetShape{3},             TargetShape{3, 3, 3})));
 
-INSTANTIATE_TEST_CASE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTBroadcastMultiplyTest,
+INSTANTIATE_TEST_SUITE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTBroadcastMultiplyTest,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, TargetShape{5}),
                         std::make_tuple(InputShape{DYN, 3, 64, 64, 64},      TargetShape{4}),
                         std::make_tuple(InputShape{2, DYN, 64, 64, 64},      TargetShape{3}),
@@ -296,7 +296,7 @@ INSTANTIATE_TEST_CASE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTBroa
                         std::make_tuple(InputShape{2, DYN, 9},          TargetShape{3}),
                         std::make_tuple(InputShape{3, 3, DYN},          TargetShape{2})));
 
-INSTANTIATE_TEST_CASE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTBroadcastLogicalOrTest,
+INSTANTIATE_TEST_SUITE_P(ConvertBroadcast3BIDIRECT, ConvertBroadcast3BIDIRECTBroadcastLogicalOrTest,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, TargetShape{5}),
                         std::make_tuple(InputShape{DYN, 3, 64, 64, 64},      TargetShape{4}),
                         std::make_tuple(InputShape{2, DYN, 64, 64, 64},      TargetShape{3}),
diff --git a/inference-engine/tests/functional/inference_engine/transformations/convert_convolution_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/convert_convolution_test.cpp
index 43bd10eccd1aa4..1a8e196be5950e 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/convert_convolution_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/convert_convolution_test.cpp
@@ -80,7 +80,7 @@ TEST_P(ConvertConvolutionTest, CompareFunctions) {
     ASSERT_TRUE(orig_shape.same_scheme(f->get_output_partial_shape(0))) << "Shape " << orig_shape << " is not equal to " << f->get_output_partial_shape(0);
 }
 
-INSTANTIATE_TEST_CASE_P(ConvertConvolution, ConvertConvolutionTest,
+INSTANTIATE_TEST_SUITE_P(ConvertConvolution, ConvertConvolutionTest,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, WeightsShape{8, 3, 1, 2, 3}),
                         std::make_tuple(InputShape{DYN, 3, 64, 64, 64}, WeightsShape{8, 3, 1, 2, 3}),
                         std::make_tuple(InputShape{2, DYN, 64, 64, 64}, WeightsShape{9, 3, 2, 3, 1}),
diff --git a/inference-engine/tests/functional/inference_engine/transformations/convert_deconvolution_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/convert_deconvolution_test.cpp
index 51f60318f1824a..3c456c7f083c6d 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/convert_deconvolution_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/convert_deconvolution_test.cpp
@@ -76,7 +76,7 @@ TEST_P(ConvertDeconvolutionTest, CompareFunctions) {
     ASSERT_TRUE(orig_shape.same_scheme(f->get_output_partial_shape(0))) << "Shape " << orig_shape << " is not equal to " << f->get_output_partial_shape(0);
 }
 
-INSTANTIATE_TEST_CASE_P(ConvertDeconvolution, ConvertDeconvolutionTest,
+INSTANTIATE_TEST_SUITE_P(ConvertDeconvolution, ConvertDeconvolutionTest,
         testing::Values(std::make_tuple(InputShape{DYN, DYN, DYN, DYN, DYN}, WeightsShape{3, 8, 1, 2, 3}),
                         std::make_tuple(InputShape{DYN, 3, 64, 64, 64}, WeightsShape{3, 8, 1, 2, 3}),
                         std::make_tuple(InputShape{2, DYN, 64, 64, 64}, WeightsShape{3, 9, 2, 3, 1}),
diff --git a/inference-engine/tests/functional/inference_engine/transformations/convert_reduce_to_pooling_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/convert_reduce_to_pooling_test.cpp
index 813fe6a7dbf182..34b287a47bc7dc 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/convert_reduce_to_pooling_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/convert_reduce_to_pooling_test.cpp
@@ -121,17 +121,17 @@ TEST_P(ConvertReduceToPoolingTests, CompareFunctions) {
 
 #define MAX std::make_shared<ngraph::opset1::ReduceMax>()
 
-INSTANTIATE_TEST_CASE_P(ReduceToMaxPooling, ConvertReduceToPoolingTests,
+INSTANTIATE_TEST_SUITE_P(ReduceToMaxPooling, ConvertReduceToPoolingTests,
         testing::Values(std::make_tuple(MAX, InputShape{2, 3, 64, 64},  ReduceAxes{3},    KeepDims{true}, ReduceToPoolParams({}, {1, 64}, {})),
                         std::make_tuple(MAX, InputShape{2, 3, 64, 64},  ReduceAxes{3, 2}, KeepDims{true}, ReduceToPoolParams({}, {64, 64}, {}))));
 
-INSTANTIATE_TEST_CASE_P(ReduceToReshape, ConvertReduceToPoolingTests,
+INSTANTIATE_TEST_SUITE_P(ReduceToReshape, ConvertReduceToPoolingTests,
         testing::Values(std::make_tuple(MAX, InputShape{2, 3, 64, 1}, ReduceAxes{3},       KeepDims{false}, ReduceToPoolParams({1, 3, 64}, {}, {})),
                         std::make_tuple(MAX, InputShape{2, 3},        ReduceAxes{-2},      KeepDims{false}, ReduceToPoolParams({3}, {}, {})),
                         std::make_tuple(MAX, InputShape{2, 3, 1},     ReduceAxes{2},       KeepDims{false}, ReduceToPoolParams({1, 3}, {}, {})),
                         std::make_tuple(MAX, InputShape{2, 3, 1, 1},  ReduceAxes{0, 3, 2}, KeepDims{false}, ReduceToPoolParams({3}, {}, {}))));
 
-INSTANTIATE_TEST_CASE_P(ReduceToReshapePoolReshape, ConvertReduceToPoolingTests,
+INSTANTIATE_TEST_SUITE_P(ReduceToReshapePoolReshape, ConvertReduceToPoolingTests,
         testing::Values(std::make_tuple(MAX, InputShape{2, 3, 3},    ReduceAxes{1, 2},    KeepDims{false}, ReduceToPoolParams({1, 1, 9, 1}, {9, 1}, {1})),
                         std::make_tuple(MAX, InputShape{2, 9},       ReduceAxes{-1},      KeepDims{true},  ReduceToPoolParams({1, 1, 9, 1}, {9, 1}, {1, 1})),
                         std::make_tuple(MAX, InputShape{2, 3, 4, 1}, ReduceAxes{1, 3, 2}, KeepDims{false}, ReduceToPoolParams({1, 1, 12, 1}, {12, 1}, {1}))));
diff --git a/inference-engine/tests/functional/inference_engine/transformations/eliminate_unsqueeze_gather.cpp b/inference-engine/tests/functional/inference_engine/transformations/eliminate_unsqueeze_gather.cpp
index 526695879d8908..cc91018e1555a9 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/eliminate_unsqueeze_gather.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/eliminate_unsqueeze_gather.cpp
@@ -71,7 +71,7 @@ class EliminateUnsqueezeGatherTest : public CommonTestUtils::TestsCommon,
 TEST_P(EliminateUnsqueezeGatherTest, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, EliminateUnsqueezeGatherTest, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, EliminateUnsqueezeGatherTest, testing::Combine(
         testing::Values(
                 ngraph::element::f16,
                 ngraph::element::f32,
diff --git a/inference-engine/tests/functional/inference_engine/transformations/fq_decomposition_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/fq_decomposition_test.cpp
index d5b49446232faf..91d218924db641 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/fq_decomposition_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/fq_decomposition_test.cpp
@@ -186,28 +186,28 @@ const auto broadcast_6D_fq_basic = ::testing::Combine(::testing::ValuesIn(precis
                                                       ::testing::Values(ngraph::Shape{1, 1, 1, 5, 6, 7}),
                                                       ::testing::ValuesIn(levels));
 
-INSTANTIATE_TEST_CASE_P(SimpleFakeQuantize_Decomposition, FakeQuantizeDecompositionTest,
+INSTANTIATE_TEST_SUITE_P(SimpleFakeQuantize_Decomposition, FakeQuantizeDecompositionTest,
                         ::testing::Combine(
                             simple_fq_basic,
                             ::testing::ValuesIn(input_ranges_supported),
                             ::testing::Values(true)),
                         FakeQuantizeDecompositionTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(BroadcastFakeQuantize_Decomposition, FakeQuantizeDecompositionTest,
+INSTANTIATE_TEST_SUITE_P(BroadcastFakeQuantize_Decomposition, FakeQuantizeDecompositionTest,
                         ::testing::Combine(
                             broadcast_fq_basic,
                             ::testing::ValuesIn(input_ranges_supported),
                             ::testing::Values(true)),
                         FakeQuantizeDecompositionTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(ElementwiseFakeQuantize_Decomposition, FakeQuantizeDecompositionTest,
+INSTANTIATE_TEST_SUITE_P(ElementwiseFakeQuantize_Decomposition, FakeQuantizeDecompositionTest,
                         ::testing::Combine(
                             elementwise_fq_basic,
                             ::testing::ValuesIn(input_ranges_supported),
                             ::testing::Values(true)),
                         FakeQuantizeDecompositionTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(FakeQuantize6D_Decomposition, FakeQuantizeDecompositionTest,
+INSTANTIATE_TEST_SUITE_P(FakeQuantize6D_Decomposition, FakeQuantizeDecompositionTest,
                         ::testing::Combine(
                             broadcast_6D_fq_basic,
                             ::testing::ValuesIn(input_ranges_supported),
@@ -220,28 +220,28 @@ const std::vector<std::pair<float, float>> input_ranges_unsupported = {
     {-5.0f, -5.0f}
 };
 
-INSTANTIATE_TEST_CASE_P(SimpleFakeQuantize_NoDecomposition, FakeQuantizeDecompositionTest,
+INSTANTIATE_TEST_SUITE_P(SimpleFakeQuantize_NoDecomposition, FakeQuantizeDecompositionTest,
                         ::testing::Combine(
                             simple_fq_basic,
                             ::testing::ValuesIn(input_ranges_unsupported),
                             ::testing::Values(false)),
                         FakeQuantizeDecompositionTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(BroadcastFakeQuantize_NoDecomposition, FakeQuantizeDecompositionTest,
+INSTANTIATE_TEST_SUITE_P(BroadcastFakeQuantize_NoDecomposition, FakeQuantizeDecompositionTest,
                         ::testing::Combine(
                             broadcast_fq_basic,
                             ::testing::ValuesIn(input_ranges_unsupported),
                             ::testing::Values(false)),
                         FakeQuantizeDecompositionTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(ElementwiseFakeQuantize_NoDecomposition, FakeQuantizeDecompositionTest,
+INSTANTIATE_TEST_SUITE_P(ElementwiseFakeQuantize_NoDecomposition, FakeQuantizeDecompositionTest,
                         ::testing::Combine(
                             elementwise_fq_basic,
                             ::testing::ValuesIn(input_ranges_unsupported),
                             ::testing::Values(false)),
                         FakeQuantizeDecompositionTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(FakeQuantize6D_NoDecomposition, FakeQuantizeDecompositionTest,
+INSTANTIATE_TEST_SUITE_P(FakeQuantize6D_NoDecomposition, FakeQuantizeDecompositionTest,
                         ::testing::Combine(
                             broadcast_6D_fq_basic,
                             ::testing::ValuesIn(input_ranges_unsupported),
diff --git a/inference-engine/tests/functional/inference_engine/transformations/fq_mul_fusion_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/fq_mul_fusion_test.cpp
index 920f2aa8954eb5..57135f0285a3b6 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/fq_mul_fusion_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/fq_mul_fusion_test.cpp
@@ -92,84 +92,84 @@ TEST_P(FQMulFusion, ExpectFusion) {
 };
 
 namespace {
-INSTANTIATE_TEST_CASE_P(ScalarFQParams_C6_4D_channel_0, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(ScalarFQParams_C6_4D_channel_0, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{64, 3, 7, 7}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{64, 1, 1, 1}),
                                            ::testing::Values(ngraph::Shape{64, 1, 1, 1})));
 
-INSTANTIATE_TEST_CASE_P(ScalarFQParams_C6_4D_channel_1, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(ScalarFQParams_C6_4D_channel_1, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{64, 3, 7, 7}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{1, 3, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 3, 1, 1})));
 
-INSTANTIATE_TEST_CASE_P(ScalarFQParams_C6_scalar, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(ScalarFQParams_C6_scalar, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{64, 3, 7, 7}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{})));
 
-INSTANTIATE_TEST_CASE_P(FQOutputs1D_C6_scalar, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(FQOutputs1D_C6_scalar, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{64, 3, 7, 7}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{1}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{1})));
 
-INSTANTIATE_TEST_CASE_P(FQOutputs_NHWC_C6_scalar, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(FQOutputs_NHWC_C6_scalar, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{1, 7, 7, 3}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{1, 1, 1, 3}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{1, 1, 1, 3})));
 
-INSTANTIATE_TEST_CASE_P(FQOutputs_NCHW_C6_scalar, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(FQOutputs_NCHW_C6_scalar, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{1, 3, 7, 7}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{1, 3, 1, 1}),
                                            ::testing::Values(ngraph::Shape{}),
                                            ::testing::Values(ngraph::Shape{1, 3, 1, 1})));
 
-INSTANTIATE_TEST_CASE_P(FQInputs_4D_with_channel_dimension, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(FQInputs_4D_with_channel_dimension, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{1, 64, 3, 3}),
                                            ::testing::Values(ngraph::Shape{1, 1, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 64, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 64, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 64, 1, 1})));
 
-INSTANTIATE_TEST_CASE_P(FQInputs_4D_per__multiplier_with_channel, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(FQInputs_4D_per__multiplier_with_channel, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{1, 64, 3, 3}),
                                            ::testing::Values(ngraph::Shape{1, 1, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 1, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 64, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 64, 1, 1})));
 
-INSTANTIATE_TEST_CASE_P(FQInputs_4D_with_channel__multiplier_4D_per_tensor, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(FQInputs_4D_with_channel__multiplier_4D_per_tensor, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{1, 64, 3, 3}),
                                            ::testing::Values(ngraph::Shape{1, 1, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 64, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 1, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 64, 1, 1})));
 
-INSTANTIATE_TEST_CASE_P(FQInputs_4D__multiplier_channel_3rd_dim, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(FQInputs_4D__multiplier_channel_3rd_dim, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{1, 64, 3, 3}),
                                            ::testing::Values(ngraph::Shape{1, 1, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 64, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 1, 3, 1}),
                                            ::testing::Values(ngraph::Shape{1, 64, 3, 1})));
 
-INSTANTIATE_TEST_CASE_P(FQOutputs_1D__multiplier_3D, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(FQOutputs_1D__multiplier_3D, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{1, 64, 3, 3}),
                                            ::testing::Values(ngraph::Shape{1, 64, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1}),
                                            ::testing::Values(ngraph::Shape{1, 3, 1}),
                                            ::testing::Values(ngraph::Shape{1, 3, 1})));
 
-INSTANTIATE_TEST_CASE_P(FQInOUt_ones__multiplier_4D_with_channel, FQMulFusion,
+INSTANTIATE_TEST_SUITE_P(FQInOUt_ones__multiplier_4D_with_channel, FQMulFusion,
                         ::testing::Combine(::testing::Values(ngraph::Shape{1, 64, 3, 3}),
                                            ::testing::Values(ngraph::Shape{1, 1, 1, 1}),
                                            ::testing::Values(ngraph::Shape{1, 1, 1, 1}),
diff --git a/inference-engine/tests/functional/inference_engine/transformations/fq_reshape_fusion.cpp b/inference-engine/tests/functional/inference_engine/transformations/fq_reshape_fusion.cpp
index 5fe78681594042..4f0760923471e6 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/fq_reshape_fusion.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/fq_reshape_fusion.cpp
@@ -113,7 +113,7 @@ TEST_P(nGraphFQReshapeFusionTests, ReshapeMatMul) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(NGraph, nGraphFQReshapeFusionTests, testing::Values(
+INSTANTIATE_TEST_SUITE_P(NGraph, nGraphFQReshapeFusionTests, testing::Values(
     // positive
     FQReshapeFusionTestCase{{1, 2, 1, 3}, {2, 1, 1}, {1}, {1, 1}, {1, 2, 1, 1}, {2, 3}, {2, 1}, {1, 1}, DO_NOT_RESHAPE, {2, 1}, false},
     FQReshapeFusionTestCase{{1, 2, 1, 3}, {2, 1, 1}, {1}, {1, 1}, {1, 2, 1, 1}, {1, 2, 1, 3}, {1, 2, 1, 1}, {1, 1, 1, 1},  {1, 1, 1, 1}, DO_NOT_RESHAPE, false},
diff --git a/inference-engine/tests/functional/inference_engine/transformations/mul_add_conversion_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/mul_add_conversion_test.cpp
index 22c18407a0743c..8da572786cd121 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/mul_add_conversion_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/mul_add_conversion_test.cpp
@@ -176,7 +176,7 @@ TEST_P(MulOrAddConversionTests, CompareFunctions) {
 #define ELTWISE_SUM MulAddConversionTests::get_eltwise_add_reference
 #define ELTWISE_PROD MulAddConversionTests::get_eltwise_mul_reference
 
-INSTANTIATE_TEST_CASE_P(MulAddToScaleShift, MulAddConversionTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(MulAddToScaleShift, MulAddConversionTests, testing::Combine(
         testing::Values(std::make_tuple(InputShape{DYN, 3, 64, 64},
                                         CONST(ngraph::Shape({1, 3, 1, 1}), 0.5),
                                         CONST(ngraph::Shape({1, 3, 1, 1}), 0.5), false),
@@ -188,7 +188,7 @@ INSTANTIATE_TEST_CASE_P(MulAddToScaleShift, MulAddConversionTests, testing::Comb
                                         CONST(ngraph::Shape({1, 3, 1, 1}), 0.5), false)),
         testing::Values(SCALESHIFT)));
 
-INSTANTIATE_TEST_CASE_P(MulToScaleShift, MulOrAddConversionTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(MulToScaleShift, MulOrAddConversionTests, testing::Combine(
         testing::Values(std::make_tuple(InputShape{DYN, 3, 64, 64},
                                         CONST(ngraph::Shape({1, 3, 1, 1}), 0.5),
                                         NONE, false),
@@ -200,7 +200,7 @@ INSTANTIATE_TEST_CASE_P(MulToScaleShift, MulOrAddConversionTests, testing::Combi
                                         NONE, false)),
         testing::Values(SCALESHIFT)));
 
-INSTANTIATE_TEST_CASE_P(AddToScaleShift, MulOrAddConversionTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(AddToScaleShift, MulOrAddConversionTests, testing::Combine(
         testing::Values(std::make_tuple(InputShape{DYN, 3, 64, 64},
                                         NONE,
                                         CONST(ngraph::Shape({1, 3, 1, 1}), 0.5), false),
@@ -212,7 +212,7 @@ INSTANTIATE_TEST_CASE_P(AddToScaleShift, MulOrAddConversionTests, testing::Combi
                                         CONST(ngraph::Shape({1, 3, 1, 1}), 0.5), false)),
         testing::Values(SCALESHIFT)));
 
-INSTANTIATE_TEST_CASE_P(MulAddToPower, MulAddConversionTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(MulAddToPower, MulAddConversionTests, testing::Combine(
         testing::Values(std::make_tuple(InputShape{DYN, 3, 64, 64},
                                         CONST(ngraph::Shape({1}), 0.5),
                                         CONST(ngraph::Shape({1}), 0.5), false),
@@ -224,7 +224,7 @@ INSTANTIATE_TEST_CASE_P(MulAddToPower, MulAddConversionTests, testing::Combine(
                                         CONST(ngraph::Shape({1}), 0.5), false)),
         testing::Values(POWER)));
 
-INSTANTIATE_TEST_CASE_P(MulToPower, MulOrAddConversionTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(MulToPower, MulOrAddConversionTests, testing::Combine(
         testing::Values(std::make_tuple(InputShape{DYN, 3, 64, 64},
                                         CONST(ngraph::Shape({1}), 0.5),
                                         NONE, false),
@@ -236,7 +236,7 @@ INSTANTIATE_TEST_CASE_P(MulToPower, MulOrAddConversionTests, testing::Combine(
                                         NONE, false)),
         testing::Values(POWER)));
 
-INSTANTIATE_TEST_CASE_P(AddToPower, MulOrAddConversionTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(AddToPower, MulOrAddConversionTests, testing::Combine(
         testing::Values(std::make_tuple(InputShape{DYN, 3, 64, 64},
                                         NONE,
                                         CONST(ngraph::Shape({1}), 0.5), false),
@@ -249,7 +249,7 @@ INSTANTIATE_TEST_CASE_P(AddToPower, MulOrAddConversionTests, testing::Combine(
         testing::Values(POWER)));
 
 
-INSTANTIATE_TEST_CASE_P(MulAddNegative, MulAddConversionTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(MulAddNegative, MulAddConversionTests, testing::Combine(
         testing::Values(std::make_tuple(InputShape{DYN, 3, DYN},
                                         CONST(ngraph::Shape({1, 1, 3, 1}), 0.5),
                                         CONST(ngraph::Shape({3, 1}), 0.5)/*detect broadcast case*/, false),
@@ -270,7 +270,7 @@ INSTANTIATE_TEST_CASE_P(MulAddNegative, MulAddConversionTests, testing::Combine(
                                         CONST(ngraph::Shape({1, 3, 1, 1}), 0.5), false)),
         testing::Values(SAME)));
 
-INSTANTIATE_TEST_CASE_P(MulToEltwise, MulOrAddConversionTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(MulToEltwise, MulOrAddConversionTests, testing::Combine(
         testing::Values(std::make_tuple(InputShape{DYN, 3, 64},
                                         CONST(ngraph::Shape({1, 1, 64}), 0.5),
                                         NONE, false),
@@ -303,7 +303,7 @@ INSTANTIATE_TEST_CASE_P(MulToEltwise, MulOrAddConversionTests, testing::Combine(
                                         NONE, true)),
         testing::Values(ELTWISE_PROD)));
 
-INSTANTIATE_TEST_CASE_P(AddToEltwise, MulOrAddConversionTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(AddToEltwise, MulOrAddConversionTests, testing::Combine(
         testing::Values(std::make_tuple(InputShape{DYN, 3, 64},
                                         NONE,
                                         CONST(ngraph::Shape({1, 1, 64}), 0.5), false),
diff --git a/inference-engine/tests/functional/inference_engine/transformations/shuffle_channels_fusion_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/shuffle_channels_fusion_test.cpp
index db379df386f8e5..5819423f1766b0 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/shuffle_channels_fusion_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/shuffle_channels_fusion_test.cpp
@@ -127,7 +127,7 @@ const std::vector<ShuffleChannelsFusionTestValues> testValues = {
     { {128, 720 * 480}, {1, 2, 64, 720 * 480}, {0, 2, 1, 3},  {1, -1, 720, 480}, false, false },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     TransformationTests,
     ShuffleChannelsFusion,
     ::testing::ValuesIn(testValues),
diff --git a/inference-engine/tests/functional/inference_engine/transformations/softmax_fusion.cpp b/inference-engine/tests/functional/inference_engine/transformations/softmax_fusion.cpp
index f8f982cd48192e..e8d5ad211165a0 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/softmax_fusion.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/softmax_fusion.cpp
@@ -55,7 +55,7 @@ TEST_P(SoftmaxFusionFixture, SoftmaxFusion) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(SoftmaxFusionTests, SoftmaxFusionFixture,
+INSTANTIATE_TEST_SUITE_P(SoftmaxFusionTests, SoftmaxFusionFixture,
         ::testing::Values(
             std::make_tuple(1, 1),
             std::make_tuple(1, -2),
@@ -109,7 +109,7 @@ TEST_P(NegativeSoftmaxFusionFixture, NegativeSoftmaxFusion) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(NegativeSoftmaxFusionTests, NegativeSoftmaxFusionFixture,
+INSTANTIATE_TEST_SUITE_P(NegativeSoftmaxFusionTests, NegativeSoftmaxFusionFixture,
         ::testing::ValuesIn(std::vector<std::tuple<std::vector<int64_t>, std::vector<int64_t>>>{
                                 std::make_tuple<std::vector<int64_t>, std::vector<int64_t>>({2}, {1}),
                                 std::make_tuple<std::vector<int64_t>, std::vector<int64_t>>({1}, {-1}),
diff --git a/inference-engine/tests/functional/inference_engine/transformations/transpose_sinking_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/transpose_sinking_test.cpp
index 7e8db994280f75..9d0c6fdf715dea 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/transpose_sinking_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/transpose_sinking_test.cpp
@@ -94,7 +94,7 @@ TEST_P(TransposeSinkingFQ, TransposeFQReduce) {
 }
 
 
-INSTANTIATE_TEST_CASE_P(TransformationTest, TransposeSinkingFQ, testing::Values(
+INSTANTIATE_TEST_SUITE_P(TransformationTest, TransposeSinkingFQ, testing::Values(
         TransposeFQReduceParams{{1, 3, 240, 140}, {0, 2, 3, 1}, {1}, {3}, {1, 1, 1, 1}, {1, 1, 1, 3}, {1, 2}, true,
                                 {1, 1, 1, 1}, {1, 3, 1, 1}, {1, 1, 1, 1}, {1, 3, 1, 1}, {2, 3}, {0, 2, 3, 1}},
         TransposeFQReduceParams{{1, 3, 240, 140}, {0, 2, 3, 1}, {1}, {3}, {1, 1, 1, 1}, {1, 1, 1, 3}, {1, 2}, false,
@@ -175,7 +175,7 @@ ASSERT_TRUE(res.first) << res.second;
 }
 
 
-INSTANTIATE_TEST_CASE_P(TransposeSinkingReduces, TransposeSinking, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(TransposeSinkingReduces, TransposeSinking, testing::Combine(
         testing::Values(
             TransposeReduceParams{{1, 3, 240, 140}, {0, 2, 3, 1}, {1, 2}, true, {2, 3}, {0, 2, 3, 1}},
             TransposeReduceParams{{10, 20, 30, 40, 50, 60, 70}, {0, 6, 1, 5, 2, 4, 3}, {1, 3, 6}, true, {6, 5, 3}, {0, 6, 1, 5, 2, 4, 3}},
@@ -194,7 +194,7 @@ INSTANTIATE_TEST_CASE_P(TransposeSinkingReduces, TransposeSinking, testing::Comb
             ngraph::opset6::ReduceLogicalAnd::type_info,
             ngraph::opset6::ReduceLogicalOr::type_info)));
 
-INSTANTIATE_TEST_CASE_P(TransposeSinkingSqueeze, TransposeSinking, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(TransposeSinkingSqueeze, TransposeSinking, testing::Combine(
         testing::Values(
             TransposeReduceParams{{2, 3, 1, 1}, {0, 2, 3, 1}, {1, 2}, false, {2, 3}, {0, 1}},
             TransposeReduceParams{{10, 20, 30, 1, 50, 1, 1}, {0, 6, 1, 5, 2, 4, 3}, {1, 3, 6}, false, {6, 5, 3}, {0, 1, 2, 3}}),
diff --git a/inference-engine/tests/functional/inference_engine/transformations/transpose_to_reshape_test.cpp b/inference-engine/tests/functional/inference_engine/transformations/transpose_to_reshape_test.cpp
index bd07dd7ef99f66..154f11e00c10fd 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/transpose_to_reshape_test.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/transpose_to_reshape_test.cpp
@@ -112,7 +112,7 @@ TEST_P(TransposeToReshapeTests, CompareFunctions) {
 #define EMPTY_FUNCTION   ReferenceParams(false, true)
 #define SHAPE_OF_GATHER  ReferenceParams()
 
-INSTANTIATE_TEST_CASE_P(KeepTranspose, TransposeToReshapeTests,
+INSTANTIATE_TEST_SUITE_P(KeepTranspose, TransposeToReshapeTests,
         testing::Values(std::make_tuple(InputShape{1, 3, 64, 64},  TransposeOrder{0, 1, 3, 2}, SAME_FUNCTION),
                         std::make_tuple(InputShape{1, 3, 1, 64},   TransposeOrder{2, 0, 3, 1}, SAME_FUNCTION),
                         std::make_tuple(InputShape{1, 3, 1, 3},    TransposeOrder{3, 0, 2, 1}, SAME_FUNCTION),
@@ -121,19 +121,19 @@ INSTANTIATE_TEST_CASE_P(KeepTranspose, TransposeToReshapeTests,
                         std::make_tuple(InputShape{DYN, DYN, 1},   TransposeOrder{2, 1, 0},    SAME_FUNCTION),
                         std::make_tuple(InputShape{DYN, DYN},      TransposeOrder{1, 0},       SAME_FUNCTION)));
 
-INSTANTIATE_TEST_CASE_P(EliminateTranspose, TransposeToReshapeTests,
+INSTANTIATE_TEST_SUITE_P(EliminateTranspose, TransposeToReshapeTests,
         testing::Values(std::make_tuple(InputShape{1, 3, 64, 64}, TransposeOrder{0, 1, 2, 3}, EMPTY_FUNCTION),
                         std::make_tuple(InputShape{1, 1, 1},      TransposeOrder{2, 0, 1},    EMPTY_FUNCTION),
                         std::make_tuple(InputShape{DYN, DYN},     TransposeOrder{0, 1},       EMPTY_FUNCTION)));
 
-INSTANTIATE_TEST_CASE_P(ReshapeWithConstant, TransposeToReshapeTests,
+INSTANTIATE_TEST_SUITE_P(ReshapeWithConstant, TransposeToReshapeTests,
         testing::Values(std::make_tuple(InputShape{1, 3, 64, 1},   TransposeOrder{0, 1, 3, 2}, ReferenceParams({1, 3, 1, 64})),
                         std::make_tuple(InputShape{1, 3, 1, 64},   TransposeOrder{1, 0, 3, 2}, ReferenceParams({3, 1, 64, 1})),
                         std::make_tuple(InputShape{DYN, DYN, 1},   TransposeOrder{0, 2, 1},    ReferenceParams({0, 1, -1})),
                         std::make_tuple(InputShape{1, 1, DYN},     TransposeOrder{2, 1, 0},    ReferenceParams({-1, 0, 1})),
                         std::make_tuple(InputShape{DYN, 1, 64, 1}, TransposeOrder{1, 0, 3, 2}, ReferenceParams({1, -1, 1, 64}))));
 
-INSTANTIATE_TEST_CASE_P(ReshapeWithGather, TransposeToReshapeTests,
+INSTANTIATE_TEST_SUITE_P(ReshapeWithGather, TransposeToReshapeTests,
         testing::Values(std::make_tuple(InputShape{DYN, 1, DYN, 1},   TransposeOrder{1, 0, 3, 2}, SHAPE_OF_GATHER),
                         std::make_tuple(InputShape{1, DYN, DYN, DYN}, TransposeOrder{1, 2, 3, 0}, SHAPE_OF_GATHER)));
 
diff --git a/inference-engine/tests/functional/inference_engine/transformations/weights_dequantize_to_fake_quantize.cpp b/inference-engine/tests/functional/inference_engine/transformations/weights_dequantize_to_fake_quantize.cpp
index a860565d53281c..944456e01f2c3c 100644
--- a/inference-engine/tests/functional/inference_engine/transformations/weights_dequantize_to_fake_quantize.cpp
+++ b/inference-engine/tests/functional/inference_engine/transformations/weights_dequantize_to_fake_quantize.cpp
@@ -84,7 +84,7 @@ TEST_P(TranslateNewWeightFormatToOldOne, ReshapeMatMul) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(NGraph, TranslateNewWeightFormatToOldOne, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(NGraph, TranslateNewWeightFormatToOldOne, testing::Combine(
         testing::Values(
             FQ_as_Mul_Sub_dequantize{-128, 127, 1, 2, (-128 - 1) * 2, (127 - 1) * 2, 256},
             FQ_as_Mul_Sub_dequantize{-127, 127, 1, 2, (-127 - 1) * 2, (127 - 1) * 2, 255},
diff --git a/inference-engine/tests/functional/plugin/conformance/test_runner/src/read_ir/read_ir.cpp b/inference-engine/tests/functional/plugin/conformance/test_runner/src/read_ir/read_ir.cpp
index 922573884a3d68..a3e70abb4873f8 100644
--- a/inference-engine/tests/functional/plugin/conformance/test_runner/src/read_ir/read_ir.cpp
+++ b/inference-engine/tests/functional/plugin/conformance/test_runner/src/read_ir/read_ir.cpp
@@ -16,7 +16,7 @@ std::vector<std::string> IRFolderPaths = {};
 std::vector<std::string> disabledTests = {};
 
 namespace {
-INSTANTIATE_TEST_CASE_P(conformance,
+INSTANTIATE_TEST_SUITE_P(conformance,
                         ReadIRTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(CommonTestUtils::getFileListByPatternRecursive(IRFolderPaths,  std::regex(R"(.*\.xml)"))),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/bf16_network_restoring.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/bf16_network_restoring.cpp
index 6f41e879e7ac81..3355659619837d 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/bf16_network_restoring.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/bf16_network_restoring.cpp
@@ -201,7 +201,7 @@ TEST_P(BF16NetworkRestore1, CompareWithRefImpl) {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, BF16NetworkRestore1,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, BF16NetworkRestore1,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/bfloat16_helpers.hpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/bfloat16_helpers.hpp
index 4c9414e08658be..f319b2baf122d7 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/bfloat16_helpers.hpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/bfloat16_helpers.hpp
@@ -120,7 +120,7 @@ typedef std::tuple<
  *  TEST_P(ScaleshiftConv_x3_Eltwise, CompareWithRefImpl) {
     test();
 };
- *  3. INSTANTIATE_TEST_CASE_P(smoke_bfloat16_NoReshape, ScaleshiftConv_x3_Eltwise,
+ *  3. INSTANTIATE_TEST_SUITE_P(smoke_bfloat16_NoReshape, ScaleshiftConv_x3_Eltwise,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::FP32),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/concat_in_place.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/concat_in_place.cpp
index 6e5f1161d4cb08..d81033f0d2a0d5 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/concat_in_place.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/concat_in_place.cpp
@@ -140,7 +140,7 @@ TEST_P(Concat_in_place, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Concat_in_place,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, Concat_in_place,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -149,7 +149,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Concat_in_place,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Concat_in_place::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, Concat_in_place,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, Concat_in_place,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_add.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_add.cpp
index 7ba4ff64729701..510345ca562268 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_add.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_add.cpp
@@ -121,7 +121,7 @@ TEST_P(ConvAdd, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ConvAdd,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ConvAdd,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -130,7 +130,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ConvAdd,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvAdd::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ConvAdd,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ConvAdd,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_conv.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_conv.cpp
index 8695fdc3e536f3..bccdaef98625f6 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_conv.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_conv.cpp
@@ -110,7 +110,7 @@ TEST_P(ConvConv, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ConvConv,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ConvConv,
                         ::testing::Combine(
                         ::testing::Values(Precision::FP32),
                         ::testing::Values(Precision::FP32),
@@ -119,7 +119,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ConvConv,
                         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvConv::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ConvConv,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ConvConv,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::FP32),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_dwconv_relu.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_dwconv_relu.cpp
index 0f6350b57170ed..6c9180b78ba015 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_dwconv_relu.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_dwconv_relu.cpp
@@ -129,7 +129,7 @@ TEST_P(ConvDWConvReLU, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ConvDWConvReLU,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ConvDWConvReLU,
                             ::testing::Combine(
                                     ::testing::Values(Precision::FP32),
                                     ::testing::Values(Precision::FP32),
@@ -138,7 +138,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ConvDWConvReLU,
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvDWConvReLU::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ConvDWConvReLU,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ConvDWConvReLU,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_eltwise_depthwise.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_eltwise_depthwise.cpp
index 224eb3b1c880fd..a673912fac04ee 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_eltwise_depthwise.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_eltwise_depthwise.cpp
@@ -226,7 +226,7 @@ TEST_P(ConvEltwiseDepthwise, CompareWithRefImpl) {
     Run_test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_1x1_depthwise_BF16, ConvEltwiseDepthwise,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_1x1_depthwise_BF16, ConvEltwiseDepthwise,
     ::testing::Combine(
         ::testing::Values(Precision::FP32),
         ::testing::Values(SizeVector({ 1, 5, 1, 1 })),
@@ -236,7 +236,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_1x1_depthwise_BF16, ConvEltwiseDepth
         ::testing::Values(std::string("jit_avx512_1x1_BF16"))),
     ConvEltwiseDepthwise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_gemm_depthwise_BF16, ConvEltwiseDepthwise,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_gemm_depthwise_BF16, ConvEltwiseDepthwise,
     ::testing::Combine(
         ::testing::Values(Precision::FP32),
         ::testing::Values(SizeVector({ 1, 3, 10, 10 })),
@@ -246,7 +246,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_gemm_depthwise_BF16, ConvEltwiseDept
         ::testing::Values(std::string("jit_avx512_BF16"))),
     ConvEltwiseDepthwise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_conv_depthwise_BF16, ConvEltwiseDepthwise,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_conv_depthwise_BF16, ConvEltwiseDepthwise,
     ::testing::Combine(
         ::testing::Values(Precision::FP32),
         ::testing::Values(SizeVector({ 1, 5, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_relu_pool_conv_relu_pool.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_relu_pool_conv_relu_pool.cpp
index 536e92a142ba58..1b83146fca7c58 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_relu_pool_conv_relu_pool.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/conv_relu_pool_conv_relu_pool.cpp
@@ -179,7 +179,7 @@ TEST_P(ConvReLUPoolConvReLUPool, CompareWithRefImpl) {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ConvReLUPoolConvReLUPool,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ConvReLUPoolConvReLUPool,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::FP32),
@@ -188,7 +188,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ConvReLUPoolConvReLUPool,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvReLUPoolConvReLUPool::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ConvReLUPoolConvReLUPool,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ConvReLUPoolConvReLUPool,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/elt_max.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/elt_max.cpp
index 95eedfc6d814f7..150c3f71580d01 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/elt_max.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/elt_max.cpp
@@ -132,7 +132,7 @@ TEST_P(Elt_max, CompareWithRefImpl) {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Elt_max,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, Elt_max,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -141,7 +141,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Elt_max,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Elt_max::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, Elt_max,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, Elt_max,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/elt_x3.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/elt_x3.cpp
index 53516f32e57871..5e71b92b19ddcf 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/elt_x3.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/elt_x3.cpp
@@ -189,7 +189,7 @@ TEST_P(Elt_x3, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Elt_x3,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, Elt_x3,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -198,7 +198,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Elt_x3,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Elt_x3::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, Elt_x3,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, Elt_x3,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/faster_100_5_1_1_conv.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/faster_100_5_1_1_conv.cpp
index 04928f1eed9fcc..93141614abd451 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/faster_100_5_1_1_conv.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/faster_100_5_1_1_conv.cpp
@@ -117,7 +117,7 @@ TEST_P(Faster100_5_1_1_Conv, CompareWithRefImpl) {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_bfloat16_NoReshape, Faster100_5_1_1_Conv,
+INSTANTIATE_TEST_SUITE_P(smoke_bfloat16_NoReshape, Faster100_5_1_1_Conv,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::FP32),
@@ -126,7 +126,7 @@ INSTANTIATE_TEST_CASE_P(smoke_bfloat16_NoReshape, Faster100_5_1_1_Conv,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             Faster100_5_1_1_Conv::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, Faster100_5_1_1_Conv,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, Faster100_5_1_1_Conv,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/gather_multiply.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/gather_multiply.cpp
index 8436ce732074a6..e4283a18931b18 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/gather_multiply.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/gather_multiply.cpp
@@ -110,7 +110,7 @@ TEST_P(Gather_multiply, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, Gather_multiply,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, Gather_multiply,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::BF16),
@@ -119,7 +119,7 @@ INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, Gather_multiply,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Gather_multiply::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Gather_multiply,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, Gather_multiply,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/gather_x2_add_mul_relu_concat_matmul.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/gather_x2_add_mul_relu_concat_matmul.cpp
index 492faef53148cc..ee0ce7ae308676 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/gather_x2_add_mul_relu_concat_matmul.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/gather_x2_add_mul_relu_concat_matmul.cpp
@@ -137,7 +137,7 @@ TEST_P(Gather_x2_add_mul_relu_concat_matmul, CompareWithRefImpl) {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Gather_x2_add_mul_relu_concat_matmul,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, Gather_x2_add_mul_relu_concat_matmul,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -146,7 +146,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Gather_x2_add_mul_relu_co
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Gather_x2_add_mul_relu_concat_matmul::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, Gather_x2_add_mul_relu_concat_matmul,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, Gather_x2_add_mul_relu_concat_matmul,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/memory_conv.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/memory_conv.cpp
index ec754661f9dd6d..fc2cebf349340e 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/memory_conv.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/memory_conv.cpp
@@ -105,7 +105,7 @@ TEST_P(MemoryConv, CheckTypeConversion) {
     ASSERT_EQ(ngraph::element::bf16, mem_w->input(0).get_element_type());
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_CPU, MemoryConv,
+INSTANTIATE_TEST_SUITE_P(smoke_CPU, MemoryConv,
                         ::testing::Combine(
                                 ::testing::Values<Precision>(Precision::BF16, Precision::FP32),
                                 ::testing::Values(SizeVector{1, 200}),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/mobilenet_ssd_with_branching.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/mobilenet_ssd_with_branching.cpp
index 3215e553e57c06..1dadf97341b42f 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/mobilenet_ssd_with_branching.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/mobilenet_ssd_with_branching.cpp
@@ -164,7 +164,7 @@ TEST_P(MobileNet_ssd_with_branching, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, MobileNet_ssd_with_branching,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, MobileNet_ssd_with_branching,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -173,7 +173,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, MobileNet_ssd_with_branch
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         MobileNet_ssd_with_branching::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, MobileNet_ssd_with_branching,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, MobileNet_ssd_with_branching,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_conv.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_conv.cpp
index 98f7ee4361414b..8b39519ed1bade 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_conv.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_conv.cpp
@@ -136,7 +136,7 @@ TEST_P(ScaleshiftConvEltwiseConv, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEltwiseConv,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEltwiseConv,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -145,7 +145,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEltwiseConv
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ScaleshiftConvEltwiseConv::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConvEltwiseConv,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConvEltwiseConv,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_relu_conv.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_relu_conv.cpp
index 014ab08e5f0c63..854ba27949bbb6 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_relu_conv.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_relu_conv.cpp
@@ -141,7 +141,7 @@ TEST_P(ScaleshiftConvEltwiseReluConv, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEltwiseReluConv,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEltwiseReluConv,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -150,7 +150,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEltwiseRelu
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ScaleshiftConvEltwiseReluConv::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConvEltwiseReluConv,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConvEltwiseReluConv,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_scaleshift.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_scaleshift.cpp
index 8abb82066de81b..c47afc8c24d058 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_scaleshift.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_eltwise_scaleshift.cpp
@@ -133,7 +133,7 @@ TEST_P(ScaleshiftConvEltwiseScaleshift, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEltwiseScaleshift,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEltwiseScaleshift,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -142,7 +142,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEltwiseScal
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ScaleshiftConvEltwiseScaleshift::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConvEltwiseScaleshift,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConvEltwiseScaleshift,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_elu_conv.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_elu_conv.cpp
index f9735e60bc359b..825501350d1163 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_elu_conv.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_elu_conv.cpp
@@ -121,7 +121,7 @@ TEST_P(ScaleshiftConvEluConv, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEluConv,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEluConv,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -130,7 +130,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvEluConv,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ScaleshiftConvEluConv::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConvEluConv,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConvEluConv,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_relu.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_relu.cpp
index 33d494658babda..bd725c29b903c5 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_relu.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_relu.cpp
@@ -109,7 +109,7 @@ TEST_P(ScaleshiftConvRelu, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvRelu,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvRelu,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -118,7 +118,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConvRelu,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ScaleshiftConvRelu::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConvRelu,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConvRelu,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_concat_relu.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_concat_relu.cpp
index d00b5144b0f157..aee55af910a13e 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_concat_relu.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_concat_relu.cpp
@@ -126,7 +126,7 @@ TEST_P(ScaleshiftConv_x2_ConcatRelu, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_ConcatRelu,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_ConcatRelu,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -135,7 +135,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_ConcatR
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ScaleshiftConv_x2_ConcatRelu::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConv_x2_ConcatRelu,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConv_x2_ConcatRelu,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_eltwise.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_eltwise.cpp
index 8a46347e5a7f1a..97d29dd540e5c4 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_eltwise.cpp
@@ -119,7 +119,7 @@ TEST_P(ScaleshiftConv_x2_Eltwise, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_Eltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_Eltwise,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -128,7 +128,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_Eltwise
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ScaleshiftConv_x2_Eltwise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConv_x2_Eltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConv_x2_Eltwise,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_mixed1_eltwise.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_mixed1_eltwise.cpp
index b517c66a9e0dc3..df17ac9fc026f9 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_mixed1_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_mixed1_eltwise.cpp
@@ -119,7 +119,7 @@ TEST_P(ScaleshiftConv_x2_mixed1_Eltwise, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_mixed1_Eltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_mixed1_Eltwise,
                             ::testing::Combine(
                                     ::testing::Values(Precision::FP32),
                                     ::testing::Values(Precision::FP32),
@@ -128,7 +128,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_mixed1_
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ScaleshiftConv_x2_mixed1_Eltwise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConv_x2_mixed1_Eltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConv_x2_mixed1_Eltwise,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_mixed2_eltwise.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_mixed2_eltwise.cpp
index d8601d6f9dd1ca..58f1be475f6f26 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_mixed2_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x2_mixed2_eltwise.cpp
@@ -122,7 +122,7 @@ TEST_P(ScaleshiftConv_x2_mixed2_Eltwise, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_mixed2_Eltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_mixed2_Eltwise,
                             ::testing::Combine(
                                     ::testing::Values(Precision::FP32),
                                     ::testing::Values(Precision::FP32),
@@ -131,7 +131,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x2_mixed2_
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ScaleshiftConv_x2_mixed2_Eltwise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConv_x2_mixed2_Eltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConv_x2_mixed2_Eltwise,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x3_eltwise.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x3_eltwise.cpp
index aba2d02a111ca8..36b8145cb17f70 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x3_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_conv_x3_eltwise.cpp
@@ -155,7 +155,7 @@ TEST_P(ScaleshiftConv_x3_Eltwise, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x3_Eltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x3_Eltwise,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::FP32),
@@ -164,7 +164,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, ScaleshiftConv_x3_Eltwise
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ScaleshiftConv_x3_Eltwise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConv_x3_Eltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, ScaleshiftConv_x3_Eltwise,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_x2_conv_x2_eltwise.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_x2_conv_x2_eltwise.cpp
index 043aa6b87a01f8..c710be46f10199 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_x2_conv_x2_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_x2_conv_x2_eltwise.cpp
@@ -141,7 +141,7 @@ TEST_P(Scaleshift_x2_Conv_x2_Eltwise, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Scaleshift_x2_Conv_x2_Eltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, Scaleshift_x2_Conv_x2_Eltwise,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -150,7 +150,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Scaleshift_x2_Conv_x2_Elt
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Scaleshift_x2_Conv_x2_Eltwise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, Scaleshift_x2_Conv_x2_Eltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, Scaleshift_x2_Conv_x2_Eltwise,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_x3_conv_eltwise_relu.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_x3_conv_eltwise_relu.cpp
index a9916754d1ea12..dbed0feef7d65e 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_x3_conv_eltwise_relu.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/scaleshift_x3_conv_eltwise_relu.cpp
@@ -161,7 +161,7 @@ TEST_P(Scaleshift_x3_ConvEltwiseRelu, CompareWithRefImpl) {
     test();
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Scaleshift_x3_ConvEltwiseRelu,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, Scaleshift_x3_ConvEltwiseRelu,
                         ::testing::Combine(
                                 ::testing::Values(Precision::FP32),
                                 ::testing::Values(Precision::FP32),
@@ -170,7 +170,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, Scaleshift_x3_ConvEltwise
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Scaleshift_x3_ConvEltwiseRelu::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, Scaleshift_x3_ConvEltwiseRelu,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, Scaleshift_x3_ConvEltwiseRelu,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/tail_fp32_optimization.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/tail_fp32_optimization.cpp
index a4dcc9713ca8b4..0d3cea9eef2ddc 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/tail_fp32_optimization.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/tail_fp32_optimization.cpp
@@ -124,7 +124,7 @@ TEST_P(PoolingAfterConv, CompareWithRefImpl) {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, PoolingAfterConv,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, PoolingAfterConv,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::FP32),
@@ -133,7 +133,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, PoolingAfterConv,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             PoolingAfterConv::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, PoolingAfterConv,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, PoolingAfterConv,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/bfloat16/topk_inputs_i32.cpp b/inference-engine/tests/functional/plugin/cpu/bfloat16/topk_inputs_i32.cpp
index 32912caca0b25c..fb8d97c2981f78 100644
--- a/inference-engine/tests/functional/plugin/cpu/bfloat16/topk_inputs_i32.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/bfloat16/topk_inputs_i32.cpp
@@ -146,7 +146,7 @@ TEST_P(TopKInputsI32, CompareWithRefImpl) {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, TopKInputsI32,
+INSTANTIATE_TEST_SUITE_P(smoke_FP32_bfloat16_NoReshape, TopKInputsI32,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::FP32),
@@ -155,7 +155,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FP32_bfloat16_NoReshape, TopKInputsI32,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         TopKInputsI32::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BF16_bfloat16_NoReshape, TopKInputsI32,
+INSTANTIATE_TEST_SUITE_P(smoke_BF16_bfloat16_NoReshape, TopKInputsI32,
                         ::testing::Combine(
                             ::testing::Values(Precision::FP32),
                             ::testing::Values(Precision::BF16),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/add_output.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/add_output.cpp
index 030da02964b20a..e197301c5342ed 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/add_output.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/add_output.cpp
@@ -32,6 +32,6 @@ std::vector<addOutputsParams> testCases = {
         addOutputsParams(getTargetNetwork(), {"Memory"}, CommonTestUtils::DEVICE_CPU)
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_AddOutputBasic, AddOutputsTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AddOutputBasic, AddOutputsTest,
         ::testing::ValuesIn(testCases),
         AddOutputsTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/caching_tests.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/caching_tests.cpp
index c9e4d9c4294fe5..c541f4e21bafde 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/caching_tests.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/caching_tests.cpp
@@ -22,7 +22,7 @@ namespace {
             1, 2
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_CachingSupportCase_CPU, LoadNetworkCacheTestBase,
+    INSTANTIATE_TEST_SUITE_P(smoke_CachingSupportCase_CPU, LoadNetworkCacheTestBase,
                             ::testing::Combine(
                                     ::testing::ValuesIn(LoadNetworkCacheTestBase::getStandardFunctions()),
                                     ::testing::ValuesIn(precisionsCPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/config.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/config.cpp
index e3f0adb1cb6117..ffc91dee6d1366 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/config.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/config.cpp
@@ -45,21 +45,21 @@ namespace {
             {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectConfigTests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_CPU),
             ::testing::ValuesIn(Configs)),
             CorrectConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, CorrectConfigTests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_MULTI),
             ::testing::ValuesIn(MultiConfigs)),
             CorrectConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, CorrectConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, CorrectConfigTests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_AUTO),
@@ -90,56 +90,56 @@ namespace {
             {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES , CommonTestUtils::DEVICE_CPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectConfigAPITests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_CPU),
             ::testing::ValuesIn(conf)),
             CorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, CorrectConfigAPITests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_MULTI),
             ::testing::ValuesIn(multiconf)),
             CorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, CorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, CorrectConfigAPITests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_AUTO),
             ::testing::ValuesIn(AutoConfigs)),
             CorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigTests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_CPU),
             ::testing::ValuesIn(inconfigs)),
             IncorrectConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, IncorrectConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, IncorrectConfigTests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_MULTI),
             ::testing::ValuesIn(multiinconfigs)),
             IncorrectConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_CPU),
             ::testing::ValuesIn(inconfigs)),
             IncorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, IncorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, IncorrectConfigAPITests,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_MULTI),
             ::testing::ValuesIn(multiinconfigs)),
             IncorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, IncorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, IncorrectConfigAPITests,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/core_integration.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/core_integration.cpp
index 54fbba129b5d15..040911489e5e69 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/core_integration.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/core_integration.cpp
@@ -13,15 +13,15 @@ namespace {
 // IE Class Common tests with <pluginName, deviceName params>
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassCommon, IEClassBasicTestP,
         ::testing::Values(std::make_pair("MKLDNNPlugin", "CPU")));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassNetworkTestP, IEClassNetworkTestP,
         ::testing::Values("CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassImportExportTestP, IEClassImportExportTestP,
         ::testing::Values("HETERO:CPU"));
 
@@ -29,43 +29,43 @@ INSTANTIATE_TEST_CASE_P(
 // IE Class GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values("CPU", "MULTI", "HETERO", "AUTO"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values("CPU", "MULTI", "HETERO", "AUTO"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_AVAILABLE_DEVICES,
         ::testing::Values("CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_FULL_DEVICE_NAME,
         ::testing::Values("CPU", "MULTI", "HETERO", "AUTO"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_OPTIMIZATION_CAPABILITIES,
         ::testing::Values("CPU", "AUTO"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_RANGE_FOR_ASYNC_INFER_REQUESTS,
         ::testing::Values("CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_RANGE_FOR_STREAMS,
         ::testing::Values("CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricTest, IEClassGetMetricTest_ThrowUnsupported,
         ::testing::Values("CPU", "MULTI", "HETERO", "AUTO"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetConfigTest, IEClassGetConfigTest_ThrowUnsupported,
         ::testing::Values("CPU", "MULTI", "HETERO", "AUTO"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetAvailableDevices, IEClassGetAvailableDevices,
         ::testing::Values("CPU"));
 
@@ -73,7 +73,7 @@ INSTANTIATE_TEST_CASE_P(
 // IE Class GetConfig
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetConfigTest, IEClassGetConfigTest,
         ::testing::Values("CPU"));
 
@@ -81,23 +81,23 @@ INSTANTIATE_TEST_CASE_P(
 // Executable Network GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values("CPU", "MULTI:CPU", "HETERO:CPU", "AUTO:CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values("CPU", "MULTI:CPU", "HETERO:CPU", "AUTO:CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_NETWORK_NAME,
         ::testing::Values("CPU", "MULTI:CPU", "HETERO:CPU", "AUTO:CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_OPTIMAL_NUMBER_OF_INFER_REQUESTS,
         ::testing::Values("CPU", "MULTI:CPU", "HETERO:CPU", "AUTO:CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_ThrowsUnsupported,
         ::testing::Values("CPU", "MULTI:CPU", "HETERO:CPU", "AUTO:CPU"));
 
@@ -105,11 +105,11 @@ INSTANTIATE_TEST_CASE_P(
 // Executable Network GetConfig / SetConfig
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkGetConfigTest, IEClassExecutableNetworkGetConfigTest,
         ::testing::Values("CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassExecutableNetworkSetConfigTest, IEClassExecutableNetworkSetConfigTest,
         ::testing::Values("CPU"));
 
@@ -117,19 +117,19 @@ INSTANTIATE_TEST_CASE_P(
 // Hetero Executable Network GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassHeteroExecutableNetworkGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values("CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassHeteroExecutableNetworkGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values("CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassHeteroExecutableNetworkGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_NETWORK_NAME,
         ::testing::Values("CPU"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassHeteroExecutableNetworkGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_TARGET_FALLBACK,
         ::testing::Values("CPU"));
 
@@ -150,13 +150,13 @@ TEST(IEClassBasicTest, smoke_SetConfigAfterCreatedThrow) {
 
 // IE Class Query network
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassQueryNetworkTest, IEClassQueryNetworkTest,
         ::testing::Values("CPU"));
 
 // IE Class Load network
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassLoadNetworkTest, IEClassLoadNetworkTest,
         ::testing::Values("CPU"));
 } // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/core_threading_tests.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/core_threading_tests.cpp
index 87dc75d3be8a32..fe294e97075b5f 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/core_threading_tests.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/core_threading_tests.cpp
@@ -18,16 +18,16 @@ const Params paramsStreams[] = {
 };
 }  // namespace
 
-INSTANTIATE_TEST_CASE_P(CPU, CoreThreadingTests, testing::ValuesIn(params), CoreThreadingTests::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(CPU, CoreThreadingTests, testing::ValuesIn(params), CoreThreadingTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(CPU, CoreThreadingTestsWithIterations,
+INSTANTIATE_TEST_SUITE_P(CPU, CoreThreadingTestsWithIterations,
     testing::Combine(testing::ValuesIn(params),
                      testing::Values(4),
                      testing::Values(50),
                      testing::Values(ModelClass::Default)),
     CoreThreadingTestsWithIterations::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(CPU_Streams, CoreThreadingTestsWithIterations,
+INSTANTIATE_TEST_SUITE_P(CPU_Streams, CoreThreadingTestsWithIterations,
     testing::Combine(testing::ValuesIn(paramsStreams),
                      testing::Values(4),
                      testing::Values(50),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/cpp_holders.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/cpp_holders.cpp
index b71d2e3d0a1e67..22546c9b74ac7d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/cpp_holders.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/cpp_holders.cpp
@@ -21,7 +21,7 @@ namespace {
             {0, 1, 3, 2}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTest,
             ::testing::Combine(
             ::testing::Values(CommonTestUtils::DEVICE_CPU),
             ::testing::ValuesIn(orders)),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/exec_graph_info.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/exec_graph_info.cpp
index c54175304f52c1..ee0bad4bb1f051 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/exec_graph_info.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/exec_graph_info.cpp
@@ -22,21 +22,21 @@ namespace {
             {{ InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, ExecGraphTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, ExecGraphTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configs)),
                             ExecGraphTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, ExecGraphTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, ExecGraphTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(multiConfigs)),
                             ExecGraphTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, ExecGraphTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, ExecGraphTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request.cpp
index ced3a3d22104be..b5c9180ef8461a 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request.cpp
@@ -28,21 +28,21 @@ namespace {
             {{ AUTO_CONFIG_KEY(DEVICE_LIST) , CommonTestUtils::DEVICE_CPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configs)),
                             InferRequestTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferRequestTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferRequestTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(Multiconfigs)),
                             InferRequestTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoPlugin_BehaviorTests, InferRequestTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoPlugin_BehaviorTests, InferRequestTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_callback.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_callback.cpp
index 3033c404bbfcec..8483b82c3db437 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_callback.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_callback.cpp
@@ -25,21 +25,21 @@ const std::vector<std::map<std::string, std::string>> autoConfigs = {
         {{ AUTO_CONFIG_KEY(DEVICE_LIST) , CommonTestUtils::DEVICE_CPU}}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CallbackTests,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_CPU),
             ::testing::ValuesIn(configs)),
         CallbackTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, CallbackTests,
         ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                 ::testing::ValuesIn(multiConfigs)),
         CallbackTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, CallbackTests,
         ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_cancellation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_cancellation.cpp
index 612d9047208635..0d2aa9a72fd19d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_cancellation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_cancellation.cpp
@@ -15,7 +15,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
         {},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CancellationTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CancellationTests,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_config.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_config.cpp
index 3db325234cf1a6..1f5e502818a32a 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_config.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_config.cpp
@@ -54,42 +54,42 @@ namespace {
             {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferConfigTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configs)),
                             InferConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferConfigTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(multiConfigs)),
                             InferConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, InferConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, InferConfigTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(autoConfigs)),
                             InferConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferConfigInTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferConfigInTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(InConfigs)),
                             InferConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferConfigInTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferConfigInTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(MultiInConfigs)),
                             InferConfigInTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, InferConfigInTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, InferConfigInTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_input.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_input.cpp
index eb988c0043a289..73bc9fb8f97ffa 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_input.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_input.cpp
@@ -29,21 +29,21 @@ namespace {
             {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestInputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestInputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configs)),
                             InferRequestInputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferRequestInputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferRequestInputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(multiConfigs)),
                             InferRequestInputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, InferRequestInputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, InferRequestInputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_output.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_output.cpp
index a77253f8900fa5..36ad2752244d6f 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_output.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/infer_request_output.cpp
@@ -25,21 +25,21 @@ namespace {
             {{InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestOutputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestOutputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configs)),
                             InferRequestOutputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferRequestOutputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferRequestOutputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(multiConfigs)),
                             InferRequestOutputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, InferRequestOutputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, InferRequestOutputTests,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/invalid_cases/proposal.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/invalid_cases/proposal.cpp
index dd8c3167e121e5..48fe63eb398a70 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/invalid_cases/proposal.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/invalid_cases/proposal.cpp
@@ -44,7 +44,7 @@ const auto proposalParams = ::testing::Combine(
         ::testing::ValuesIn(framework_)
 );
 
-INSTANTIATE_TEST_CASE_P(invalid, ProposalBehTest,
+INSTANTIATE_TEST_SUITE_P(invalid, ProposalBehTest,
                         ::testing::Combine(
                                 proposalParams,
                                 ::testing::ValuesIn(img_info_invalid),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/layout.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/layout.cpp
index 908ec597e0d340..f3854504d81cce 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/layout.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/layout.cpp
@@ -26,7 +26,7 @@ namespace {
             { 3 }
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, LayoutTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, LayoutTest,
             ::testing::Combine(
                     ::testing::Values(InferenceEngine::Precision::FP32),
                     ::testing::Values(CommonTestUtils::DEVICE_CPU, "HETERO:CPU"),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/perf_counters.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/perf_counters.cpp
index 8bf1a8f95ba766..39aff76e6fd230 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/perf_counters.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/perf_counters.cpp
@@ -14,14 +14,14 @@ namespace {
             {{ MULTI_CONFIG_KEY(DEVICE_PRIORITIES) , CommonTestUtils::DEVICE_CPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PerfCountersTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PerfCountersTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configs)),
                             PerfCountersTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, PerfCountersTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, PerfCountersTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/preprocessing.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/preprocessing.cpp
index f4059b4688283a..657dfe5e947b56 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/preprocessing.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/preprocessing.cpp
@@ -17,7 +17,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     {}
 };
 
-INSTANTIATE_TEST_CASE_P(BehaviourPreprocessingTestsViaSetInput, PreprocessingPrecisionConvertTest,
+INSTANTIATE_TEST_SUITE_P(BehaviourPreprocessingTestsViaSetInput, PreprocessingPrecisionConvertTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inputPrecisions),
                                 ::testing::Values(1, 2, 3, 4, 5),   // Number of input tensor channels
@@ -26,7 +26,7 @@ INSTANTIATE_TEST_CASE_P(BehaviourPreprocessingTestsViaSetInput, PreprocessingPre
                                 ::testing::ValuesIn(configs)),
                         PreprocessingPrecisionConvertTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(BehaviourPreprocessingTestsViaGetBlob, PreprocessingPrecisionConvertTest,
+INSTANTIATE_TEST_SUITE_P(BehaviourPreprocessingTestsViaGetBlob, PreprocessingPrecisionConvertTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inputPrecisions),
                                 ::testing::Values(4, 5),       // Number of input tensor channels (blob_copy only supports 4d and 5d tensors)
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_blob.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_blob.cpp
index 923cf62a61cd5d..214e9dc9cb49a5 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_blob.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_blob.cpp
@@ -18,4 +18,4 @@ const auto params = ::testing::Combine(::testing::ValuesIn(precisionSet),
                                        ::testing::ValuesIn(typeSet),
                                        ::testing::Values(CommonTestUtils::DEVICE_CPU));
 
-INSTANTIATE_TEST_CASE_P(smoke_SetBlobCPU, SetBlobTest, params, SetBlobTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_SetBlobCPU, SetBlobTest, params, SetBlobTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_blob_of_kind.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_blob_of_kind.cpp
index c6814674d3fb23..0ce335bcfada2c 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_blob_of_kind.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_blob_of_kind.cpp
@@ -19,26 +19,26 @@ const SetBlobOfKindConfig autoConfig{};
 const SetBlobOfKindConfig multiConfig{{ MULTI_CONFIG_KEY(DEVICE_PRIORITIES) , CommonTestUtils::DEVICE_CPU}};
 const SetBlobOfKindConfig heteroConfig{{ "TARGET_FALLBACK", CommonTestUtils::DEVICE_CPU }};
 
-INSTANTIATE_TEST_CASE_P(smoke_SetBlobOfKindCPU, SetBlobOfKindTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SetBlobOfKindCPU, SetBlobOfKindTest,
     ::testing::Combine(::testing::ValuesIn(blobKinds),
                        ::testing::Values(CommonTestUtils::DEVICE_CPU),
                        ::testing::Values(cpuConfig)),
     SetBlobOfKindTest::getTestCaseName);
 
 
-INSTANTIATE_TEST_CASE_P(smoke_SetBlobOfKindMULTI, SetBlobOfKindTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SetBlobOfKindMULTI, SetBlobOfKindTest,
     ::testing::Combine(::testing::ValuesIn(blobKinds),
                        ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                        ::testing::Values(multiConfig)),
     SetBlobOfKindTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_SetBlobOfKindAUTO, SetBlobOfKindTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SetBlobOfKindAUTO, SetBlobOfKindTest,
     ::testing::Combine(::testing::ValuesIn(blobKinds),
                        ::testing::Values(CommonTestUtils::DEVICE_AUTO + std::string(":") + CommonTestUtils::DEVICE_CPU),
                        ::testing::Values(autoConfig)),
     SetBlobOfKindTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_SetBlobOfKindHETERO, SetBlobOfKindTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SetBlobOfKindHETERO, SetBlobOfKindTest,
     ::testing::Combine(::testing::ValuesIn(blobKinds),
                        ::testing::Values(CommonTestUtils::DEVICE_HETERO),
                        ::testing::Values(heteroConfig)),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_preprocess.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_preprocess.cpp
index 3db386c50c7294..d3450e08ee27b2 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_preprocess.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/set_preprocess.cpp
@@ -31,28 +31,28 @@ namespace {
             {{ InferenceEngine::KEY_AUTO_DEVICE_LIST , CommonTestUtils::DEVICE_CPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configs)),
                             PreprocessTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Hetero_BehaviorTests, PreprocessTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Hetero_BehaviorTests, PreprocessTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_HETERO),
                                     ::testing::ValuesIn(heteroConfigs)),
                             PreprocessTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, PreprocessTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, PreprocessTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(multiConfigs)),
                             PreprocessTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, PreprocessTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, PreprocessTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
@@ -74,7 +74,7 @@ namespace {
         InferenceEngine::Layout::NHWC
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessConversionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessConversionTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::ValuesIn(ioPrecisions),
@@ -88,7 +88,7 @@ namespace {
                                 ::testing::ValuesIn(configs)),
                         PreprocessConversionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Bool(),
@@ -102,7 +102,7 @@ namespace {
                                 ::testing::ValuesIn(configs)),
                         PreprocessDynamicallyInSetBlobTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Hetero_BehaviorTests, PreprocessConversionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Hetero_BehaviorTests, PreprocessConversionTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::ValuesIn(ioPrecisions),
@@ -116,7 +116,7 @@ namespace {
                                 ::testing::ValuesIn(heteroConfigs)),
                         PreprocessConversionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Hetero_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Hetero_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Bool(),
@@ -130,7 +130,7 @@ namespace {
                                 ::testing::ValuesIn(heteroConfigs)),
                         PreprocessDynamicallyInSetBlobTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, PreprocessConversionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, PreprocessConversionTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::ValuesIn(ioPrecisions),
@@ -144,7 +144,7 @@ namespace {
                                 ::testing::ValuesIn(multiConfigs)),
                         PreprocessConversionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Bool(),
@@ -158,7 +158,7 @@ namespace {
                                 ::testing::ValuesIn(multiConfigs)),
                         PreprocessDynamicallyInSetBlobTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, PreprocessConversionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, PreprocessConversionTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::ValuesIn(ioPrecisions),
@@ -172,7 +172,7 @@ namespace {
                                 ::testing::ValuesIn(autoConfigs)),
                         PreprocessConversionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Bool(),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/test_plugin.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/test_plugin.cpp
index da76bc26498b1e..f8728390c9ed9b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/test_plugin.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/test_plugin.cpp
@@ -46,63 +46,63 @@ namespace {
     };
 
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTestOutput,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configsOutput)),
                             BehaviorTestOutput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, BehaviorTestOutput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(MultiConfigsInputOutput)),
                             BehaviorTestOutput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, BehaviorTestOutput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(AutoConfigsInputOutput)),
                             BehaviorTestOutput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTests,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configs)),
                             BehaviorTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, BehaviorTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, BehaviorTests,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(MultiConfigs)),
                             BehaviorTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, BehaviorTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, BehaviorTests,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(AutoConfigs)),
                             BehaviorTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTestInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configsInput)),
                             BehaviorTestInput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, BehaviorTestInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(MultiConfigsInputOutput)),
                             BehaviorTestInput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, BehaviorTestInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/version.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/version.cpp
index bcbe9dfaf08fdf..66b5a3c06685cd 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/version.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/behavior/version.cpp
@@ -22,28 +22,28 @@ namespace {
             {{ HETERO_CONFIG_KEY(DUMP_GRAPH_DOT) , CommonTestUtils::DEVICE_CPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
                                     ::testing::ValuesIn(configs)),
                             VersionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(Multiconfigs)),
                             VersionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(Autoconfigs)),
                             VersionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Hetero_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Hetero_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_HETERO),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/configuration_tests/configuration_tests.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/configuration_tests/configuration_tests.cpp
index dc1e42f28e2ede..48f4b0bc4d274c 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/configuration_tests/configuration_tests.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/configuration_tests/configuration_tests.cpp
@@ -13,7 +13,7 @@ auto defaultBindThreadParameter = InferenceEngine::Parameter{std::string{CONFIG_
 #else
 auto defaultBindThreadParameter = InferenceEngine::Parameter{std::string{CONFIG_VALUE(YES)}};
 #endif
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Basic,
     DefaultConfigurationTest,
     ::testing::Combine(
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/configuration_tests/dynamic_batch.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/configuration_tests/dynamic_batch.cpp
index 6c882e13ab71cb..0b6ff2ae24e74b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/configuration_tests/dynamic_batch.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/configuration_tests/dynamic_batch.cpp
@@ -19,7 +19,7 @@ std::map<std::string, std::string> additional_config = {
 } // namespace
 
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicBatchTest_async, DynamicBatchTest,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicBatchTest_async, DynamicBatchTest,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(InferenceEngine::Precision::FP32),
@@ -28,7 +28,7 @@ INSTANTIATE_TEST_CASE_P(smoke_DynamicBatchTest_async, DynamicBatchTest,
         ::testing::Values(additional_config)),
     DynamicBatchTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicBatchTest_sync, DynamicBatchTest,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicBatchTest_sync, DynamicBatchTest,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(InferenceEngine::Precision::FP32),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/exec_graph_serialization.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/exec_graph_serialization.cpp
index f8bcd35e9a469b..5d0346424a1bca 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/exec_graph_serialization.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/exec_graph_serialization.cpp
@@ -9,7 +9,7 @@ namespace {
 
 using namespace ExecutionGraphTests;
 
-INSTANTIATE_TEST_CASE_P(smoke_serialization, ExecGraphSerializationTest,
+INSTANTIATE_TEST_SUITE_P(smoke_serialization, ExecGraphSerializationTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU),
                         ExecGraphSerializationTest::getTestCaseName);
 
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/keep_assign.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/keep_assign.cpp
index 83f9808d267b2b..de0da5d16078f1 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/keep_assign.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/keep_assign.cpp
@@ -9,7 +9,7 @@ using namespace ExecutionGraphTests;
 
 namespace {
 
-INSTANTIATE_TEST_CASE_P(smoke_KeepAssign, ExecGraphKeepAssignNode,
+INSTANTIATE_TEST_SUITE_P(smoke_KeepAssign, ExecGraphKeepAssignNode,
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ExecGraphKeepAssignNode::getTestCaseName);
 
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/nms_transformation_for_last_node.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/nms_transformation_for_last_node.cpp
index 5d5fbe597f8398..8d4d11972d9fb0 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/nms_transformation_for_last_node.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/nms_transformation_for_last_node.cpp
@@ -8,6 +8,6 @@
 namespace {
 using namespace ExecutionGraphTests;
 
-INSTANTIATE_TEST_CASE_P(smoke_NmsTransformationLastNode, ExecGraphNmsTransformLastNode, ::testing::Values(CommonTestUtils::DEVICE_CPU),
+INSTANTIATE_TEST_SUITE_P(smoke_NmsTransformationLastNode, ExecGraphNmsTransformLastNode, ::testing::Values(CommonTestUtils::DEVICE_CPU),
                         ExecGraphNmsTransformLastNode::getTestCaseName);
 } // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/num_inputs_fusing_bin_conv.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/num_inputs_fusing_bin_conv.cpp
index 4ff4c385747dd3..4b31128536189b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/num_inputs_fusing_bin_conv.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/num_inputs_fusing_bin_conv.cpp
@@ -8,6 +8,6 @@
 namespace {
 using namespace ExecutionGraphTests;
 
-INSTANTIATE_TEST_CASE_P(smoke_inputsNumFusingBinConv, ExecGraphInputsFusingBinConv, ::testing::Values(CommonTestUtils::DEVICE_CPU),
+INSTANTIATE_TEST_SUITE_P(smoke_inputsNumFusingBinConv, ExecGraphInputsFusingBinConv, ::testing::Values(CommonTestUtils::DEVICE_CPU),
                         ExecGraphInputsFusingBinConv::getTestCaseName);
 } // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/remove_parameter.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/remove_parameter.cpp
index 790a8f91aeda61..2706e2c3578ace 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/remove_parameter.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/remove_parameter.cpp
@@ -9,7 +9,7 @@ using namespace ExecutionGraphTests;
 
 namespace {
 
-INSTANTIATE_TEST_CASE_P(smoke_removeParameter, ExecGraphRemoveParameterNode,
+INSTANTIATE_TEST_SUITE_P(smoke_removeParameter, ExecGraphRemoveParameterNode,
                         ::testing::Values(CommonTestUtils::DEVICE_CPU),
                         ExecGraphRemoveParameterNode::getTestCaseName);
 
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/runtime_precision.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/runtime_precision.cpp
index f95ba68b777ddf..6d1272c2c8f175 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/runtime_precision.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/runtime_precision.cpp
@@ -24,7 +24,7 @@ const std::vector<RuntimePrecisionSpecificParams> params = {
         {makeFakeQuantizeBinaryConvolutionFunction, {Precision::FP32}, {{"FakeQuantize", Precision::FP32}, {"BinaryConvolution", Precision::BIN}}},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ExecGraph, ExecGraphRuntimePrecision,
+INSTANTIATE_TEST_SUITE_P(smoke_ExecGraph, ExecGraphRuntimePrecision,
                         ::testing::Combine(
                                 ::testing::ValuesIn(params),
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/unique_node_names.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/unique_node_names.cpp
index ebd1d57c3ef107..d470fc7bc04730 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/unique_node_names.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/execution_graph_tests/unique_node_names.cpp
@@ -15,7 +15,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP32
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, ExecGraphUniqueNodeNames,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, ExecGraphUniqueNodeNames,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(InferenceEngine::SizeVector({1, 2, 5, 5})),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/hetero/query_network.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/hetero/query_network.cpp
index ff05d70af9cad7..0ba1d74c899913 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/hetero/query_network.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/hetero/query_network.cpp
@@ -13,7 +13,7 @@ using namespace HeteroTests;
 
 auto ConvBias = ngraph::builder::subgraph::makeConvBias();
 
-INSTANTIATE_TEST_CASE_P(smoke_FullySupportedTopologies, QueryNetworkTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FullySupportedTopologies, QueryNetworkTest,
                         ::testing::Combine(
                                 ::testing::Values("CPU", "HETERO:CPU", "MULTI:CPU"),
                                 ::testing::Values(ConvBias)),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/hetero/synthetic.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/hetero/synthetic.cpp
index 49f4ddda323f55..8c531cb2c151fb 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/hetero/synthetic.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/hetero/synthetic.cpp
@@ -11,13 +11,13 @@
 namespace {
 using namespace HeteroTests;
 
-INSTANTIATE_TEST_CASE_P(smoke_SingleMajorNode, HeteroSyntheticTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SingleMajorNode, HeteroSyntheticTest,
                         ::testing::Combine(
                                 ::testing::Values(std::vector<PluginParameter>{{"CPU0", "MKLDNNPlugin"}, {"CPU1", "MKLDNNPlugin"}}),
                                 ::testing::ValuesIn(HeteroTests::HeteroSyntheticTest::_singleMajorNodeFunctions)),
                         HeteroSyntheticTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(nightly_RandomMajorNodes, HeteroSyntheticTest,
+INSTANTIATE_TEST_SUITE_P(nightly_RandomMajorNodes, HeteroSyntheticTest,
                         ::testing::Combine(
                                 ::testing::Values(std::vector<PluginParameter>{{"CPU0", "MKLDNNPlugin"}, {"CPU1", "MKLDNNPlugin"}}),
                                 ::testing::ValuesIn(HeteroTests::HeteroSyntheticTest::_randomMajorNodeFunctions)),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/add_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/add_transformation.cpp
index c90dc6545d6a11..f575a6cc2ec472 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/add_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/add_transformation.cpp
@@ -68,7 +68,7 @@ const std::vector<LayerTestsDefinitions::AddTestValues> params = {
     { {}, {}, false }, { {}, {}, true },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, AddTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, AddTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/clamp_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/clamp_transformation.cpp
index a657058d1feb1f..6f45a17b35214e 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/clamp_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/clamp_transformation.cpp
@@ -76,7 +76,7 @@ const std::vector<LayerTestsDefinitions::ClampTransformationParam> params{
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ClampTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ClampTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_transformation.cpp
index be35416515a9d7..60d5e1ff730589 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_transformation.cpp
@@ -49,7 +49,7 @@ const std::vector<ngraph::Shape> shapes = {
     ngraph::Shape({ 4, 3, 16, 16 })
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_child_and_output.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_child_and_output.cpp
index 5688483d58b110..0141d37d23a5ef 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_child_and_output.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_child_and_output.cpp
@@ -42,7 +42,7 @@ const std::vector<ConcatWithChildAndOutputTransformationParam> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatWithChildAndOutputTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatWithChildAndOutputTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 6, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_different_precision_on_children.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_different_precision_on_children.cpp
index fe0cd0b5465477..dc7d9936e6af76 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_different_precision_on_children.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_different_precision_on_children.cpp
@@ -45,7 +45,7 @@ const std::vector<ConcatWithDifferentChildrenTransformationParam> testValues = {
 
 const std::vector<bool> multiChannel = { true/*, false*/ };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatWithDifferentChildrenTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatWithDifferentChildrenTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 6, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_intermediate_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_intermediate_transformation.cpp
index c4f984b8e94a65..8c24e9cec2091d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_intermediate_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_intermediate_transformation.cpp
@@ -31,7 +31,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 16, 16 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatWithIntermediateTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatWithIntermediateTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_neighbors_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_neighbors_transformation.cpp
index 9ece832517f48c..a0724e1bfe667c 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_neighbors_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_neighbors_transformation.cpp
@@ -28,7 +28,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 16, 16 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatWithNeighborsGraphTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatWithNeighborsGraphTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_split_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_split_transformation.cpp
index 4c9d43c124f847..f6d2f1c16b78fd 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_split_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/concat_with_split_transformation.cpp
@@ -45,7 +45,7 @@ const std::vector<ConcatWithSplitTransformationParam> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatWithSplitTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatWithSplitTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 6, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
index 223f9743a61888..180ef107e02ebb 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
@@ -108,7 +108,7 @@ const std::vector<ngraph::Shape> outputShapes = {
     { 16, 16 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConvolutionBackpropDataTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConvolutionBackpropDataTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_qdq_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_qdq_transformation.cpp
index fdaae7a35f7c88..5cb46fc2ee868c 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_qdq_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_qdq_transformation.cpp
@@ -238,7 +238,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 4, 4 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConvolutionQDqTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConvolutionQDqTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
index 594e645980aa74..0a574422ae8469 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
@@ -101,7 +101,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 16, 16 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConvolutionTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConvolutionTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(shapes),
@@ -125,7 +125,7 @@ const std::vector<LayerTestsDefinitions::ConvolutionWIthIncorrectWeightsParam> i
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConvolutionWIthIncorrectWeightsTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConvolutionWIthIncorrectWeightsTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/depth_to_space_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/depth_to_space_transformation.cpp
index baf025aefacca9..d9057f66a92565 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/depth_to_space_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/depth_to_space_transformation.cpp
@@ -34,7 +34,7 @@ const auto DepthToSpaceBS2 = ::testing::Combine(
     ::testing::Values(2)
 );
 
-INSTANTIATE_TEST_CASE_P(LPT_BS2, DepthToSpaceTransformation, DepthToSpaceBS2, DepthToSpaceTransformation::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(LPT_BS2, DepthToSpaceTransformation, DepthToSpaceBS2, DepthToSpaceTransformation::getTestCaseName);
 
 const std::vector<ngraph::Shape> inputShapesBS3 = {
         {1, 9, 3, 3}, {2, 27, 5, 4}
@@ -48,5 +48,5 @@ const auto DepthToSpaceBS3 = ::testing::Combine(
     ::testing::Values(3)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT_BS3, DepthToSpaceTransformation, DepthToSpaceBS3, DepthToSpaceTransformation::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_LPT_BS3, DepthToSpaceTransformation, DepthToSpaceBS3, DepthToSpaceTransformation::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_avg_pool_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_avg_pool_transformation.cpp
index cf86089bfd68f5..d0f2b614f43a89 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_avg_pool_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_avg_pool_transformation.cpp
@@ -26,7 +26,7 @@ const std::vector<ngraph::builder::subgraph::FakeQuantizeOnData> fakeQuantizes =
 
 // FakeQuantizeOnData
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizeAndAvgPoolTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizeAndAvgPoolTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 32, 72, 48 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_max_pool_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_max_pool_transformation.cpp
index fe184eef51659a..78df11d2f7853d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_max_pool_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_max_pool_transformation.cpp
@@ -26,7 +26,7 @@ const std::vector<ngraph::builder::subgraph::FakeQuantizeOnData> fakeQuantizes =
 
 // FakeQuantizeOnData
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizeAndMaxPoolTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizeAndMaxPoolTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 32, 72, 48 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp
index 29124d9503b9c8..adb79c659ecd32 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp
@@ -28,7 +28,7 @@ const std::vector<FakeQuantizeAndTwoOutputBranchesWithConvolution> testValues =
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizeAndTwoOutputBranchesWithConvolutionTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizeAndTwoOutputBranchesWithConvolutionTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 32, 72, 48 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_precision_selection_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_precision_selection_transformation.cpp
index 3b897406957e18..f2e7df2acde38b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_precision_selection_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_precision_selection_transformation.cpp
@@ -56,7 +56,7 @@ const std::vector<FakeQuantizePrecisionSelectionTransformationTestValues> testVa
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizePrecisionSelectionTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizePrecisionSelectionTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 32, 72, 48 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_transformation.cpp
index 0208a6b0e7217b..69ce1065c206c1 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_transformation.cpp
@@ -59,7 +59,7 @@ const std::vector<FakeQuantizeTransformationParam> fakeQuantizeOnDataValues = {
     // { 256ul, { 1ul }, { -1.28f} , { 1.27f } }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizeTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 32, 72, 48 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp
index 073e13da0b78d1..af1a82dcd41bfe 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp
@@ -101,7 +101,7 @@ const std::vector<FakeQuantizeWithNotOptimalTransformationTestValues> fakeQuanti
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizeWithNotOptimalTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizeWithNotOptimalTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fully_connected_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fully_connected_transformation.cpp
index e439c53abcefd4..a285b41a53675f 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fully_connected_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fully_connected_transformation.cpp
@@ -41,7 +41,7 @@ const std::vector<ngraph::pass::low_precision::LayerTransformation::Params> tras
     LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParams()
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FullyConnectedTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FullyConnectedTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_convert_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_convert_transformation.cpp
index f59e9bcf8e8ffe..8082ecfe96d7ec 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_convert_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_convert_transformation.cpp
@@ -31,7 +31,7 @@ const std::vector<ngraph::builder::subgraph::DequantizationOperations> deqOperat
 
 const std::vector<bool> constInput = { true, false };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FuseConvertTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FuseConvertTransformation,
     ::testing::Combine(
             ::testing::ValuesIn(precisions),
             ::testing::ValuesIn(inputAndQuantizationShapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_and_scale_shift_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_and_scale_shift_transformation.cpp
index 189cc02d14b1cd..8d4b127eb48ae0 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_and_scale_shift_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_and_scale_shift_transformation.cpp
@@ -33,7 +33,7 @@ const std::vector<ngraph::builder::subgraph::FakeQuantizeOnData> fakeQuantizeOnD
 //    },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FuseFakeQuantizeAndScaleShiftTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FuseFakeQuantizeAndScaleShiftTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 9, 9 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_transformation.cpp
index 39a32db4ceeafc..d1d38ed94d72fd 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_transformation.cpp
@@ -104,7 +104,7 @@ const std::vector<FuseFakeQuantizeTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FuseFakeQuantizeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FuseFakeQuantizeTransformation,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::ValuesIn(testValues)),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_multiply_to_fake_quantize_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_multiply_to_fake_quantize_transformation.cpp
index 927c64cad19e2a..81674818d6bce2 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_multiply_to_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_multiply_to_fake_quantize_transformation.cpp
@@ -31,7 +31,7 @@ const std::vector<FuseMultiplyToFakeQuantizeTransformationTestValues> testValues
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FuseMultiplyToFakeQuantizeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FuseMultiplyToFakeQuantizeTransformation,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::ValuesIn(testValues)),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_subtract_to_fake_quantize_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_subtract_to_fake_quantize_transformation.cpp
index 252015a0a42c9a..4a65f1eb27bc4a 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_subtract_to_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/fuse_subtract_to_fake_quantize_transformation.cpp
@@ -31,7 +31,7 @@ const std::vector<FuseSubtractToFakeQuantizeTransformationTestValues> testValues
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FuseSubtractToFakeQuantizeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FuseSubtractToFakeQuantizeTransformation,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::ValuesIn(testValues)),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/gemm_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/gemm_transformation.cpp
index 52e9223f208e5a..242df59cfff1ae 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/gemm_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/gemm_transformation.cpp
@@ -27,7 +27,7 @@ const std::vector<LayerTransformation::Params> trasformationParamValues = {
     LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParamsU8I8()
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, GemmTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, GemmTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(dimensions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
index 739897ee27350e..d8a81076d61ce9 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
@@ -97,7 +97,7 @@ const std::vector<LayerTestsDefinitions::GroupConvolutionTransformationParam> pa
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, GroupConvolutionTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, GroupConvolutionTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/groupconvolution_qdq_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/groupconvolution_qdq_transformation.cpp
index 54f14fb037d20e..74c60def9fc5c5 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/groupconvolution_qdq_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/groupconvolution_qdq_transformation.cpp
@@ -492,7 +492,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 1, 4, 7, 7 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, GroupConvolutionQDqTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, GroupConvolutionQDqTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/interpolate_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/interpolate_transformation.cpp
index d8e2854699ea91..7954fac1a232a0 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/interpolate_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/interpolate_transformation.cpp
@@ -47,5 +47,5 @@ const auto combineValues = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU),
     ::testing::ValuesIn(interpAttrs));
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, InterpolateTransformation, combineValues, InterpolateTransformation::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, InterpolateTransformation, combineValues, InterpolateTransformation::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_transformation.cpp
index 58caafc62f5dd9..2df46fe0bfc5c5 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_transformation.cpp
@@ -42,7 +42,7 @@ std::vector<MatMulTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MatMulTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MatMulTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 384, 1024 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_with_constant_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_with_constant_transformation.cpp
index 34e2dbf542d46b..776be2284259d2 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_with_constant_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_with_constant_transformation.cpp
@@ -96,7 +96,7 @@ std::vector<MatMulWithConstantTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MatMulWithConstantTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MatMulWithConstantTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_with_optimized_constant_fake_quantize_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_with_optimized_constant_fake_quantize_transformation.cpp
index 47d7d6eda915b4..76ec3422c0638d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_with_optimized_constant_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mat_mul_with_optimized_constant_fake_quantize_transformation.cpp
@@ -26,7 +26,7 @@ const std::vector<std::pair<ngraph::Shape, ngraph::Shape>> inputShapes = {
     std::pair<ngraph::Shape, ngraph::Shape>({ 1, 16 }, { 16, 10 })
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MatMulWithOptimizedConstantFakeQuantizeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MatMulWithOptimizedConstantFakeQuantizeTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_to_group_convolution_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_to_group_convolution_transformation.cpp
index 0a7baea851642d..39dc081f40265a 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_to_group_convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_to_group_convolution_transformation.cpp
@@ -22,7 +22,7 @@ const std::vector<builder::subgraph::FakeQuantizeOnData> fqOnData = {
     { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 25.5f }, { -12.8f }, { 12.7f } }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MultiplyToGroupConvolutionTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MultiplyToGroupConvolutionTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_transformation.cpp
index 1814b6830eea09..8f791dc2e7cf5f 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_transformation.cpp
@@ -67,7 +67,7 @@ const std::vector<LayerTestsDefinitions::MultiplyTestValues> params = {
     { {}, {}, false }, { {}, {}, true },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MultiplyTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MultiplyTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_with_one_parent_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_with_one_parent_transformation.cpp
index efafab0e36ec6b..2c790eb7023473 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_with_one_parent_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/multiply_with_one_parent_transformation.cpp
@@ -21,7 +21,7 @@ const std::vector<MultiplyWithOneParentTransformationValues> values = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MultiplyWithOneParentTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MultiplyWithOneParentTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mvn_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mvn_transformation.cpp
index fd3811e909eb56..505fffa8d54424 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mvn_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/mvn_transformation.cpp
@@ -20,7 +20,7 @@ const std::vector<AxisSet> reductionAxes = { { 2, 3 }, { 1, 2, 3 } };
 
 const std::vector<bool> normalizeVariance = { true, false };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MVNTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MVNTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::ValuesIn(inputAndQuantizationShapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/normalize_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/normalize_transformation.cpp
index ed312302d5ec1a..6a88af8938d97b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/normalize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/normalize_transformation.cpp
@@ -29,7 +29,7 @@ const std::vector<bool> fuseMultiplyValues = { true, false };
 
 const std::vector<bool> shiftValues = { true, false };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, NormalizeL2Transformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, NormalizeL2Transformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::ValuesIn(inputAndQuantizationShapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations.cpp
index d2ae271b277c67..828d9f852bf53b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations.cpp
@@ -23,7 +23,7 @@ const std::vector<LayerTransformation::Params> trasformationParamValues = {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, OutputLayersHandlingInTransformations,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, OutputLayersHandlingInTransformations,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat.cpp
index ea42b1c8ef336b..a8ee6f581f5467 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat.cpp
@@ -22,7 +22,7 @@ const std::vector<LayerTransformation::Params> trasformationParamValues = {
     LayerTestsUtils::LayerTransformationParamsFactory::createParamsU8I8()
 };
 
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_LPT, OutputLayersHandlingInTransformationsForConcat,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_LPT, OutputLayersHandlingInTransformationsForConcat,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat_multi_channel.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat_multi_channel.cpp
index aa09761dd87519..6c0c6ad3fd1bfb 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat_multi_channel.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat_multi_channel.cpp
@@ -23,7 +23,7 @@ const std::vector<LayerTransformation::Params> trasformationParamValues = {
 };
 
 // TODO: issue #41231: enable previous LPT version tests
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_LPT, OutputLayersHandlingInTransformationsForConcatMultiChannel,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_LPT, OutputLayersHandlingInTransformationsForConcatMultiChannel,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/prelu_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/prelu_transformation.cpp
index ac0af7702391b0..bb35f882e628c8 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/prelu_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/prelu_transformation.cpp
@@ -23,7 +23,7 @@ std::vector<PReluTestValues> testValues = {
     { { 256ul, ngraph::Shape({}), {-12.8f / 2.f}, {12.7f}, {-12.8f / 2.f}, {12.7f} }, true }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, PReluTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, PReluTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/pull_reshape_through_dequantization_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/pull_reshape_through_dequantization_transformation.cpp
index d0c33423c175d6..0d8ca9c22840b9 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/pull_reshape_through_dequantization_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/pull_reshape_through_dequantization_transformation.cpp
@@ -70,7 +70,7 @@ const std::vector<ngraph::Shape> dequantizationOnWeightElementwiseConstantShapes
     { ngraph::Shape({1, 16}) }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, PullReshapeThroughDequantizationTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, PullReshapeThroughDequantizationTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_max_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_max_transformation.cpp
index 6a29b84fb0ebe3..a24ef762ea5db9 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_max_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_max_transformation.cpp
@@ -104,7 +104,7 @@ const std::vector<LayerTestsDefinitions::ReduceMaxTransformationParam> params =
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReduceMaxTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReduceMaxTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_mean_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_mean_transformation.cpp
index 7310ca764e671e..7c6d1b60c6ba7c 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_mean_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_mean_transformation.cpp
@@ -104,7 +104,7 @@ const std::vector<LayerTestsDefinitions::ReduceMeanTransformationParam> params =
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReduceMeanTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReduceMeanTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_min_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_min_transformation.cpp
index 5bc4f1f3ba0b60..af56b057a33084 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_min_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_min_transformation.cpp
@@ -104,7 +104,7 @@ const std::vector<LayerTestsDefinitions::ReduceMinTransformationParam> params =
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReduceMinTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReduceMinTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_sum_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_sum_transformation.cpp
index 2ab45ce1fb92b8..cf767b71e4d860 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_sum_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reduce_sum_transformation.cpp
@@ -90,7 +90,7 @@ const std::vector<LayerTestsDefinitions::ReduceSumTransformationParam> params =
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReduceSumTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReduceSumTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/relu_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/relu_transformation.cpp
index 25d7bd0398e267..460d050f6aaa79 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/relu_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/relu_transformation.cpp
@@ -24,7 +24,7 @@ std::vector<ReluTestValues> testValues = {
     { { 256ul, ngraph::Shape({}), {-12.8f / 2.f}, {12.7f}, {-12.8f / 2.f}, {12.7f} }, true }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReluTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReluTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reshape_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reshape_transformation.cpp
index 5a9342ca718655..599b3de32e6d2c 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reshape_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/reshape_transformation.cpp
@@ -52,7 +52,7 @@ const std::vector<ReshapeTransformationParam> params = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReshapeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReshapeTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/shuffle_channels_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/shuffle_channels_transformation.cpp
index 6d2e97a0008ea9..2a214de6d3a2f1 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/shuffle_channels_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/shuffle_channels_transformation.cpp
@@ -90,7 +90,7 @@ const std::vector<LayerTestsDefinitions::ShuffleChannelsTransformationParam> par
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ShuffleChannelsTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ShuffleChannelsTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/split_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/split_transformation.cpp
index be380eeac0963f..56f3c7e2f256f9 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/split_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/split_transformation.cpp
@@ -84,7 +84,7 @@ const std::vector<LayerTestsDefinitions::SplitTransformationParam> params = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, SplitTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, SplitTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/squeeze_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/squeeze_transformation.cpp
index efe0109ee55192..0e608b55c10fa2 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/squeeze_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/squeeze_transformation.cpp
@@ -45,7 +45,7 @@ namespace {
         }
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_LPT, SqueezeTransformation,
+    INSTANTIATE_TEST_SUITE_P(smoke_LPT, SqueezeTransformation,
         ::testing::Combine(
             ::testing::ValuesIn(precisions),
             ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/strided_slice_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/strided_slice_transformation.cpp
index 273daf6f7102e8..a66106064a1a44 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/strided_slice_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/strided_slice_transformation.cpp
@@ -89,7 +89,7 @@ const std::vector<LayerTestsDefinitions::StridedSliceTransformationParam> params
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, StridedSliceTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, StridedSliceTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 24, 24 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/subtract_multiply_to_multiply_add_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/subtract_multiply_to_multiply_add_transformation.cpp
index db33013688268e..6f2c4136e1a159 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/subtract_multiply_to_multiply_add_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/subtract_multiply_to_multiply_add_transformation.cpp
@@ -58,7 +58,7 @@ const std::vector<SubtractMultiplyToMultiplyAddTransformationTestValues> testVal
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, SubtractMultiplyToMultiplyAddTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, SubtractMultiplyToMultiplyAddTransformation,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::ValuesIn(testValues)),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/subtract_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/subtract_transformation.cpp
index 4b3c6a792c3efb..3a5d5d6a32f607 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/subtract_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/subtract_transformation.cpp
@@ -22,7 +22,7 @@ const std::vector<LayerTransformation::Params> trasformationParamValues = {
     LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParamsU8I8()
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, SubtractTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, SubtractTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/transpose_after_matmul_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/transpose_after_matmul_transformation.cpp
index f378f4ff487639..98eb3cdde77033 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/transpose_after_matmul_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/transpose_after_matmul_transformation.cpp
@@ -26,7 +26,7 @@ const std::vector<bool> perTensorValues = { true, false };
 
 const std::vector<bool> transposeChannelDimValues = { true, false };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, TransposeAfterMatMulTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, TransposeAfterMatMulTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/transpose_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/transpose_transformation.cpp
index 28c09a394543dc..390daa853c254c 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/transpose_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/transpose_transformation.cpp
@@ -41,7 +41,7 @@ const std::vector<TransposeTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, TransposeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, TransposeTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/unsqueeze_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/unsqueeze_transformation.cpp
index b245edb9bec4f4..eb8e47fe08c36a 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/unsqueeze_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/unsqueeze_transformation.cpp
@@ -50,7 +50,7 @@ namespace {
         }
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_LPT, UnsqueezeTransformation,
+    INSTANTIATE_TEST_SUITE_P(smoke_LPT, UnsqueezeTransformation,
         ::testing::Combine(
             ::testing::ValuesIn(precisions),
             ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/variadic_split_transformation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/variadic_split_transformation.cpp
index 5778caa6fb8687..247dce334444f3 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/variadic_split_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/low_precision_transformations/variadic_split_transformation.cpp
@@ -90,7 +90,7 @@ const std::vector<LayerTestsDefinitions::VariadicSplitTransformationParam> param
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, VariadicSplitTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, VariadicSplitTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/multi/cpu_remote_blob_tests.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/multi/cpu_remote_blob_tests.cpp
index 3b0ba475eae223..ba5442e158e815 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/multi/cpu_remote_blob_tests.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/multi/cpu_remote_blob_tests.cpp
@@ -11,5 +11,5 @@ const std::vector<DevicesNamesAndSupportPair> device_names_and_support_for_remot
         {{CPU}, false}, // CPU via MULTI
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_RemoteBlobMultiCPU, MultiDevice_SupportTest,
+INSTANTIATE_TEST_SUITE_P(smoke_RemoteBlobMultiCPU, MultiDevice_SupportTest,
         ::testing::ValuesIn(device_names_and_support_for_remote_blobs), MultiDevice_SupportTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/ngraph_conversion_tests/conv_bias_fusion.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/ngraph_conversion_tests/conv_bias_fusion.cpp
index 5f8562bfa555d3..53d8627119c85d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/ngraph_conversion_tests/conv_bias_fusion.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/ngraph_conversion_tests/conv_bias_fusion.cpp
@@ -9,6 +9,6 @@ using namespace NGraphConversionTestsDefinitions;
 
 namespace {
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, ConvBiasFusion, ::testing::Values("CPU"), ConvBiasFusion::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, ConvBiasFusion, ::testing::Values("CPU"), ConvBiasFusion::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/ngraph_conversion_tests/plugin_specific_ngraph_conversion.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/ngraph_conversion_tests/plugin_specific_ngraph_conversion.cpp
index 5ba015166dd6af..d4b2e16b6ee316 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/ngraph_conversion_tests/plugin_specific_ngraph_conversion.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/ngraph_conversion_tests/plugin_specific_ngraph_conversion.cpp
@@ -9,6 +9,6 @@ using namespace NGraphConversionTestsDefinitions;
 
 namespace {
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, PluginSpecificConversion, ::testing::Values("CPU"), PluginSpecificConversion::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, PluginSpecificConversion, ::testing::Values("CPU"), PluginSpecificConversion::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
index 51e379a67f128b..87f986f980fffc 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
@@ -124,10 +124,10 @@ const auto basicIntegerOperations = ::testing::Combine(
             ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationDynamicLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_Integer_Activation_Basic, ActivationLayerTest, basicIntegerOperations, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic, ActivationDynamicLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Integer_Activation_Basic, ActivationLayerTest, basicIntegerOperations, ActivationLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic_Prelu_Const, ActivationLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic_Prelu_Param, ActivationParamLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic_Prelu_Const, ActivationLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic_Prelu_Param, ActivationParamLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/batch_norm.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/batch_norm.cpp
index 753efd3acaa84c..f765e2d0afeadb 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/batch_norm.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/batch_norm.cpp
@@ -40,7 +40,7 @@ const auto batchNormParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_BatchNorm,
         BatchNormLayerTest,
         batchNormParams,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/batch_to_space.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/batch_to_space.cpp
index 18a099484fffb6..ddf930f34b72b5 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/batch_to_space.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/batch_to_space.cpp
@@ -49,7 +49,7 @@ batchToSpaceParamsTuple bts_only_test_cases[] = {
                                 CommonTestUtils::DEVICE_CPU),
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN, BatchToSpaceLayerTest, ::testing::ValuesIn(bts_only_test_cases),
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN, BatchToSpaceLayerTest, ::testing::ValuesIn(bts_only_test_cases),
                         BatchToSpaceLayerTest::getTestCaseName);
 
 
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/binary_convolution.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/binary_convolution.cpp
index d451e80b292067..369015510d567a 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/binary_convolution.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/binary_convolution.cpp
@@ -44,7 +44,7 @@ const auto binConv2DParams_ValidPadding = ::testing::Combine(
     ::testing::Values(ngraph::op::PadType::VALID),
     ::testing::ValuesIn(padValues));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_BinaryConvolution2D_ExplicitPadding, BinaryConvolutionLayerTest,
     ::testing::Combine(
         binConv2DParams_ExplicitPadding,
@@ -57,7 +57,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     BinaryConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_BinaryConvolution2D_AutoPadValid, BinaryConvolutionLayerTest,
     ::testing::Combine(
         binConv2DParams_ValidPadding,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/broadcast.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/broadcast.cpp
index 8f69fd6ee4bcef..d30ba6cd215ba5 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/broadcast.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/broadcast.cpp
@@ -38,7 +38,7 @@ const auto numpyBroadcastParams1 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestNumpyBroadcast1,
         BroadcastLayerTest,
         numpyBroadcastParams1,
@@ -54,7 +54,7 @@ const auto numpyBroadcastParams2 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestNumpyBroadcast2,
         BroadcastLayerTest,
         numpyBroadcastParams2,
@@ -84,7 +84,7 @@ const auto bidirectionalBroadcastParams1 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestBidirectionalBroadcast1,
         BroadcastLayerTest,
         bidirectionalBroadcastParams1,
@@ -100,7 +100,7 @@ const auto bidirectionalBroadcastParams2 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestBidirectionalBroadcast2,
         BroadcastLayerTest,
         bidirectionalBroadcastParams2,
@@ -116,7 +116,7 @@ const auto bidirectionalBroadcastParams3 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestBidirectionalBroadcast3,
         BroadcastLayerTest,
         bidirectionalBroadcastParams3,
@@ -149,7 +149,7 @@ const auto explicitBroadcastParams1 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestExplicitBroadcast1,
         BroadcastLayerTest,
         explicitBroadcastParams1,
@@ -165,7 +165,7 @@ const auto explicitBroadcastParams2 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestExplicitBroadcast2,
         BroadcastLayerTest,
         explicitBroadcastParams2,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/bucketize.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/bucketize.cpp
index 6d75dfe3717a2d..6398fa83ef0fb0 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/bucketize.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/bucketize.cpp
@@ -52,5 +52,5 @@ const auto test_Bucketize_left_edge = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsBucketize_right, BucketizeLayerTest, test_Bucketize_right_edge, BucketizeLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsBucketize_left, BucketizeLayerTest, test_Bucketize_left_edge, BucketizeLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsBucketize_right, BucketizeLayerTest, test_Bucketize_right_edge, BucketizeLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsBucketize_left, BucketizeLayerTest, test_Bucketize_left_edge, BucketizeLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/clamp.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/clamp.cpp
index 200350fd316067..9f146ae0dbf12b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/clamp.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/clamp.cpp
@@ -48,5 +48,5 @@ const auto test_Clamp_unsigned = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsClamp_signed, ClampLayerTest, test_Clamp_signed, ClampLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsClamp_unsigned, ClampLayerTest, test_Clamp_unsigned, ClampLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsClamp_signed, ClampLayerTest, test_Clamp_signed, ClampLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsClamp_unsigned, ClampLayerTest, test_Clamp_unsigned, ClampLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/comparison.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/comparison.cpp
index d1602d698a07e3..ad6c1c897b1c42 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/comparison.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/comparison.cpp
@@ -50,6 +50,6 @@ const auto ComparisonTestParams = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(additional_config));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs, ComparisonLayerTest, ComparisonTestParams, ComparisonLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs, ComparisonLayerTest, ComparisonTestParams, ComparisonLayerTest::getTestCaseName);
 
 }  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/concat.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/concat.cpp
index 7c75e31a3548b9..e983d4a65ec428 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/concat.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/concat.cpp
@@ -24,7 +24,7 @@ std::vector<std::vector<std::vector<size_t>>> inShapes = {
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                          InferenceEngine::Precision::FP16};
 
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, ConcatLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, ConcatLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(axes),
                                 ::testing::ValuesIn(inShapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/constant.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/constant.cpp
index 06ee157d39108f..5a06250dd6ed08 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/constant.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/constant.cpp
@@ -39,13 +39,13 @@ std::vector<InferenceEngine::Precision> precisionsWithNegativeValues{
 std::vector<std::string> dataWithNegativeValues{"1", "-2", "3", "-4", "5", "-6",
                                                 "7", "-1", "2", "-3", "4", "-5"};
 
-INSTANTIATE_TEST_CASE_P(smoke_Constant, ConstantLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Constant, ConstantLayerTest,
                         ::testing::Combine(::testing::ValuesIn(shapes),
                                            ::testing::ValuesIn(precisions), ::testing::Values(data),
                                            ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConstantLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Constant_with_negative_values, ConstantLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Constant_with_negative_values, ConstantLayerTest,
                         ::testing::Combine(::testing::ValuesIn(shapes),
                                            ::testing::ValuesIn(precisionsWithNegativeValues),
                                            ::testing::Values(dataWithNegativeValues),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convert.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convert.cpp
index 3a024a65965bc1..70ed6c6512fc41 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convert.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convert.cpp
@@ -24,7 +24,7 @@ const std::vector<Precision> precisions = {
         Precision::FP32
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvertLayerTest, ConvertLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvertLayerTest, ConvertLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(inShape),
                                 ::testing::ValuesIn(precisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convert_like.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convert_like.cpp
index abb8d43057e3b1..f9e20dcf0de0a2 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convert_like.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convert_like.cpp
@@ -19,7 +19,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::I8,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, ConvertLikeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, ConvertLikeLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(inShape),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution.cpp
index 325636314d7bd5..fbaba32e8d8b87 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution.cpp
@@ -35,7 +35,7 @@ const auto conv1DParams_AutoPadValid = ::testing::Combine(
     ::testing::ValuesIn(dilations1D), ::testing::ValuesIn(numOutChannels1D),
     ::testing::Values(ngraph::op::PadType::VALID));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Convolution1D_ExplicitPadding, ConvolutionLayerTest,
     ::testing::Combine(
         conv1DParams_ExplicitPadding, ::testing::ValuesIn(netPrecisions),
@@ -47,7 +47,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Convolution1D_AutoPadValid, ConvolutionLayerTest,
     ::testing::Combine(
         conv1DParams_AutoPadValid, ::testing::ValuesIn(netPrecisions),
@@ -79,7 +79,7 @@ const auto conv2DParams_AutoPadValid = ::testing::Combine(
     ::testing::ValuesIn(dilations), ::testing::ValuesIn(numOutChannels),
     ::testing::Values(ngraph::op::PadType::VALID));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Convolution2D_ExplicitPadding, ConvolutionLayerTest,
     ::testing::Combine(
         conv2DParams_ExplicitPadding, ::testing::ValuesIn(netPrecisions),
@@ -91,7 +91,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Convolution2D_AutoPadValid, ConvolutionLayerTest,
     ::testing::Combine(
         conv2DParams_AutoPadValid, ::testing::ValuesIn(netPrecisions),
@@ -120,7 +120,7 @@ namespace specificWeightLayout {
                                                               ::testing::Values(numOutChannels),
                                                               ::testing::Values(ngraph::op::PadType::EXPLICIT));
 
-    INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_SpecificWeightLayout, ConvolutionLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D_SpecificWeightLayout, ConvolutionLayerTest,
                                 ::testing::Combine(conv2DParams_WeightLayout,
                                                    ::testing::ValuesIn(netPrecisions),
                                                    ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -151,7 +151,7 @@ const auto conv3DParams_AutoPadValid = ::testing::Combine(
     ::testing::ValuesIn(dilations3d), ::testing::ValuesIn(numOutChannels3D),
     ::testing::Values(ngraph::op::PadType::VALID));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Convolution3D_ExplicitPadding, ConvolutionLayerTest,
     ::testing::Combine(
         conv3DParams_ExplicitPadding, ::testing::ValuesIn(netPrecisions),
@@ -163,7 +163,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Convolution3D_AutoPadValid, ConvolutionLayerTest,
     ::testing::Combine(
         conv3DParams_AutoPadValid, ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
index 1a5f3885c9364b..7fe868eccba9ac 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
@@ -51,7 +51,7 @@ const auto conv2DParams_AutoPadValid = ::testing::Combine(
         ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -64,7 +64,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, Convolu
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -80,7 +80,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, Convolutio
 const std::vector<std::vector<size_t >> inputShape2D = {{1, 3, 9, 12}};
 const std::vector<std::vector<size_t >> outputShapes2D = {{6, 6}, {4, 9}};
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_OutputShapeDefined, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D_OutputShapeDefined, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -117,7 +117,7 @@ const auto conv2DParams_AutoPadValid_output_padding = ::testing::Combine(
         ::testing::ValuesIn(outputPadding2D)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid_output_padding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -130,7 +130,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding_OutputPa
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding_output_padding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -174,7 +174,7 @@ const auto conv3DParams_AutoPadValid = ::testing::Combine(
         ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -187,7 +187,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, Convolu
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -203,7 +203,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, Convolutio
 const std::vector<std::vector<size_t >> inputShape3D = {{1, 3, 10, 10, 10}};
 const std::vector<std::vector<size_t >> outputShapes3D = {{8, 8, 8}, {10, 10, 10}};
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_OutputShapeDefined, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData3D_OutputShapeDefined, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -240,7 +240,7 @@ const auto conv3DParams_AutoPadValid_output_padding = ::testing::Combine(
         ::testing::ValuesIn(outputPadding3D)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_AutoPadValid_output_padding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -253,7 +253,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding_OutputPa
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData3D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_ExplicitPadding_output_padding,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder.cpp
index b0ddcd67c89d10..4b349fc5978e6a 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder.cpp
@@ -34,7 +34,7 @@ const auto basicCases = ::testing::Combine(
     ::testing::ValuesIn(mergeRepeated),
     ::testing::Values(CommonTestUtils::DEVICE_CPU));
 
-INSTANTIATE_TEST_CASE_P(smoke_CtcGreedyDecoderBasic, CTCGreedyDecoderLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_CtcGreedyDecoderBasic, CTCGreedyDecoderLayerTest,
                         basicCases,
                         CTCGreedyDecoderLayerTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp
index 2247dfd7821fd7..f25df60e99af0e 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp
@@ -33,11 +33,11 @@ const auto basicCases = ::testing::Combine(
     ::testing::ValuesIn(mergeRepeated),
     ::testing::Values(CommonTestUtils::DEVICE_CPU));
 
-INSTANTIATE_TEST_CASE_P(smoke_set1, CTCGreedyDecoderSeqLenLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_set1, CTCGreedyDecoderSeqLenLayerTest,
                         basicCases,
                         CTCGreedyDecoderSeqLenLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_set2, CTCGreedyDecoderSeqLenLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_set2, CTCGreedyDecoderSeqLenLayerTest,
         ::testing::Combine(
                         ::testing::ValuesIn(std::vector<std::vector<size_t>>{{2, 8, 11}, {4, 10, 55}}),
                         ::testing::ValuesIn(std::vector<int>{5, 100}),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_loss.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_loss.cpp
index e0609683606a92..7203f49950c6e8 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_loss.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/ctc_loss.cpp
@@ -35,7 +35,7 @@ const auto ctcLossArgsSubset1 = ::testing::Combine(
         ::testing::ValuesIn(unique)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Set1, CTCLossLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Set1, CTCLossLayerTest,
                         ::testing::Combine(
                             ctcLossArgsSubset1,
                             ::testing::ValuesIn(fPrecisions),
@@ -56,7 +56,7 @@ const auto ctcLossArgsSubset2 = ::testing::Combine(
         ::testing::ValuesIn(unique)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Set2, CTCLossLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Set2, CTCLossLayerTest,
                         ::testing::Combine(
                             ctcLossArgsSubset2,
                             ::testing::ValuesIn(fPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/cum_sum.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/cum_sum.cpp
index 9a8325041a6785..eef6c6551216cb 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/cum_sum.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/cum_sum.cpp
@@ -105,11 +105,11 @@ const auto testCasesAxis_6 = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsCumSum_negative_axis, CumSumLayerTest, testCasesNegativeAxis, CumSumLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsCumSum_axis_0, CumSumLayerTest, testCasesAxis_0, CumSumLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsCumSum_axis_1, CumSumLayerTest, testCasesAxis_1, CumSumLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsCumSum_axis_2, CumSumLayerTest, testCasesAxis_2, CumSumLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsCumSum_axis_3, CumSumLayerTest, testCasesAxis_3, CumSumLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsCumSum_axis_4, CumSumLayerTest, testCasesAxis_4, CumSumLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsCumSum_axis_5, CumSumLayerTest, testCasesAxis_5, CumSumLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsCumSum_axis_6, CumSumLayerTest, testCasesAxis_6, CumSumLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsCumSum_negative_axis, CumSumLayerTest, testCasesNegativeAxis, CumSumLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsCumSum_axis_0, CumSumLayerTest, testCasesAxis_0, CumSumLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsCumSum_axis_1, CumSumLayerTest, testCasesAxis_1, CumSumLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsCumSum_axis_2, CumSumLayerTest, testCasesAxis_2, CumSumLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsCumSum_axis_3, CumSumLayerTest, testCasesAxis_3, CumSumLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsCumSum_axis_4, CumSumLayerTest, testCasesAxis_4, CumSumLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsCumSum_axis_5, CumSumLayerTest, testCasesAxis_5, CumSumLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsCumSum_axis_6, CumSumLayerTest, testCasesAxis_6, CumSumLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/deformable_convolution.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/deformable_convolution.cpp
index e54812fe516a5c..7bf74fb3a82cd3 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/deformable_convolution.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/deformable_convolution.cpp
@@ -54,7 +54,7 @@ const auto deformableConv2DParams_DeformableGroups_AutoPadExplicit = ::testing::
     ::testing::ValuesIn(multiple_defor_groups), ::testing::ValuesIn(numOutChannels),
     ::testing::Values(ngraph::op::PadType::EXPLICIT));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_DeformableConvolution2D_ExplicitPadding, DeformableConvolutionLayerTest,
     ::testing::Combine(
         deformableConv2DParams_ExplicitPadding, ::testing::ValuesIn(netPrecisions),
@@ -66,7 +66,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     DeformableConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_DeformableConvolution2D_AutoPadValid, DeformableConvolutionLayerTest,
     ::testing::Combine(
         deformableConv2DParams_AutoPadValid, ::testing::ValuesIn(netPrecisions),
@@ -78,7 +78,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     DeformableConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_DeformableConvolution2D_DeformableGroups_ExplicitPadding, DeformableConvolutionLayerTest,
     ::testing::Combine(
         deformableConv2DParams_DeformableGroups_AutoPadExplicit, ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/deformable_psroi_pooling.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/deformable_psroi_pooling.cpp
index 4a0a16c3822004..8cf39c201df8ab 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/deformable_psroi_pooling.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/deformable_psroi_pooling.cpp
@@ -27,7 +27,7 @@ namespace {
         ::testing::Values(InferenceEngine::Precision::FP32), // Net precision
         ::testing::Values(CommonTestUtils::DEVICE_CPU));     // Device name
 
-    INSTANTIATE_TEST_CASE_P(smoke_TestsDeformablePSROIPooling, DeformablePSROIPoolingLayerTest, deformablePSROICases_test_params,
+    INSTANTIATE_TEST_SUITE_P(smoke_TestsDeformablePSROIPooling, DeformablePSROIPoolingLayerTest, deformablePSROICases_test_params,
                             DeformablePSROIPoolingLayerTest::getTestCaseName);
 
 
@@ -47,6 +47,6 @@ namespace {
         ::testing::Values(InferenceEngine::Precision::FP32), // Net precision
         ::testing::Values(CommonTestUtils::DEVICE_CPU));     // Device name
 
-    INSTANTIATE_TEST_CASE_P(smoke_TestsDeformablePSROIPooling_advanced, DeformablePSROIPoolingLayerTest, deformablePSROICases_test_params_advanced,
+    INSTANTIATE_TEST_SUITE_P(smoke_TestsDeformablePSROIPooling_advanced, DeformablePSROIPoolingLayerTest, deformablePSROICases_test_params_advanced,
                             DeformablePSROIPoolingLayerTest::getTestCaseName);
 }  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/depth_to_space.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/depth_to_space.cpp
index d8785579ed6409..33955e2e4a160d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/depth_to_space.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/depth_to_space.cpp
@@ -36,7 +36,7 @@ const auto DepthToSpaceBS2 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DepthToSpaceBS2, DepthToSpaceLayerTest, DepthToSpaceBS2, DepthToSpaceLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DepthToSpaceBS2, DepthToSpaceLayerTest, DepthToSpaceBS2, DepthToSpaceLayerTest::getTestCaseName);
 
 const std::vector<std::vector<size_t >> inputShapesBS3 = {
         {1, 9, 1, 1}, {1, 9, 2, 2}, {1, 9, 3, 3}, {2, 36, 3, 3}, {2, 27, 5, 4},
@@ -51,6 +51,6 @@ const auto DepthToSpaceBS3 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DepthToSpaceBS3, DepthToSpaceLayerTest, DepthToSpaceBS3, DepthToSpaceLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DepthToSpaceBS3, DepthToSpaceLayerTest, DepthToSpaceBS3, DepthToSpaceLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/detection_output.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/detection_output.cpp
index e30c577241fbba..c40ab0ca365209 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/detection_output.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/detection_output.cpp
@@ -56,7 +56,7 @@ const auto params3Inputs = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DetectionOutput3In, DetectionOutputLayerTest, params3Inputs, DetectionOutputLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DetectionOutput3In, DetectionOutputLayerTest, params3Inputs, DetectionOutputLayerTest::getTestCaseName);
 
 /* =============== 5 inputs cases =============== */
 
@@ -80,6 +80,6 @@ const auto params5Inputs = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DetectionOutput5In, DetectionOutputLayerTest, params5Inputs, DetectionOutputLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DetectionOutput5In, DetectionOutputLayerTest, params5Inputs, DetectionOutputLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/dft.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/dft.cpp
index 27e5f5051dcae6..c917b1bd887a2c 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/dft.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/dft.cpp
@@ -101,7 +101,7 @@ const auto testCase4D = ::testing::Combine(
 );
 
 
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsDFT_1d, DFTLayerTest, testCase1D, DFTLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsDFT_2d, DFTLayerTest, testCase2D, DFTLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsDFT_3d, DFTLayerTest, testCase3D, DFTLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsDFT_4d, DFTLayerTest, testCase4D, DFTLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsDFT_1d, DFTLayerTest, testCase1D, DFTLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsDFT_2d, DFTLayerTest, testCase2D, DFTLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsDFT_3d, DFTLayerTest, testCase3D, DFTLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsDFT_4d, DFTLayerTest, testCase4D, DFTLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/eltwise.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/eltwise.cpp
index ea9c438abe2722..b0933abc379763 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/eltwise.cpp
@@ -66,7 +66,7 @@ const auto multiply_params = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(additional_config));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs, EltwiseLayerTest, multiply_params, EltwiseLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs, EltwiseLayerTest, multiply_params, EltwiseLayerTest::getTestCaseName);
 
 
 std::vector<std::vector<std::vector<size_t>>> inShapesSingleThread = {
@@ -96,7 +96,7 @@ const auto single_thread_params = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(additional_config_single_thread));
 
-INSTANTIATE_TEST_CASE_P(smoke_SingleThread, EltwiseLayerTest, single_thread_params, EltwiseLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_SingleThread, EltwiseLayerTest, single_thread_params, EltwiseLayerTest::getTestCaseName);
 
 
 }  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_bag_offsets_sum.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_bag_offsets_sum.cpp
index 7bd296bca76df5..79081a70dfc692 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_bag_offsets_sum.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_bag_offsets_sum.cpp
@@ -39,7 +39,7 @@ const auto embBagOffsetSumArgSet = ::testing::Combine(
         ::testing::ValuesIn(with_default_index)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke, EmbeddingBagOffsetsSumLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke, EmbeddingBagOffsetsSumLayerTest,
                         ::testing::Combine(
                                 embBagOffsetSumArgSet,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_bag_packed_sum.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_bag_packed_sum.cpp
index 805a766f49110d..47f5686637835d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_bag_packed_sum.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_bag_packed_sum.cpp
@@ -33,7 +33,7 @@ const auto embBagPackedSumArgSet = ::testing::Combine(
         ::testing::ValuesIn(with_weights)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke, EmbeddingBagPackedSumLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke, EmbeddingBagPackedSumLayerTest,
                         ::testing::Combine(
                                 embBagPackedSumArgSet,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_segments_sum.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_segments_sum.cpp
index 70d982c5f33ef4..ad80292245c0fa 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_segments_sum.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/embedding_segments_sum.cpp
@@ -41,7 +41,7 @@ const auto embSegmentsSumArgSet = ::testing::Combine(
         ::testing::ValuesIn(with_default_index)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke, EmbeddingSegmentsSumLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke, EmbeddingSegmentsSumLayerTest,
                         ::testing::Combine(
                                 embSegmentsSumArgSet,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/extract_image_patches.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/extract_image_patches.cpp
index 14ca4d3e6194cc..32278fb9ade5e5 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/extract_image_patches.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/extract_image_patches.cpp
@@ -25,7 +25,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
     InferenceEngine::Precision::I64
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_layers_CPU, ExtractImagePatchesTest,
+INSTANTIATE_TEST_SUITE_P(smoke_layers_CPU, ExtractImagePatchesTest,
         ::testing::Combine(
             ::testing::ValuesIn(inDataShape),
             ::testing::ValuesIn(kernels),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/fake_quantize.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/fake_quantize.cpp
index 3dbb6f5fd539b5..bc0440b0d87f4a 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/fake_quantize.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/fake_quantize.cpp
@@ -33,7 +33,7 @@ const auto fqParams = ::testing::Combine(
         ::testing::Values(inputParams)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantize, FakeQuantizeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantize, FakeQuantizeLayerTest,
                         ::testing::Combine(
                                 fqParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -64,7 +64,7 @@ const auto fqParamsPerChannelAxis1 = ::testing::Combine(
         ::testing::Values(inputParams)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantizePerChannelAxis0, FakeQuantizeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantizePerChannelAxis0, FakeQuantizeLayerTest,
                         ::testing::Combine(
                                 fqParamsPerChannelAxis0,
                                 ::testing::ValuesIn(netPrecisions),
@@ -77,7 +77,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FakeQuantizePerChannelAxis0, FakeQuantizeLayerTest
                                 ::testing::Values(config)),
                         FakeQuantizeLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantizePerChannelAxis1, FakeQuantizeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantizePerChannelAxis1, FakeQuantizeLayerTest,
                         ::testing::Combine(
                                 fqParamsPerChannelAxis1,
                                 ::testing::ValuesIn(netPrecisions),
@@ -99,7 +99,7 @@ const auto fqParamsPerChannel2D = ::testing::Combine(
     ::testing::Values(inputParams)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantizePerChannel2D, FakeQuantizeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantizePerChannel2D, FakeQuantizeLayerTest,
     ::testing::Combine(
         fqParamsPerChannel2D,
         ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather.cpp
index 32038696be6f68..d187d5d86cc918 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather.cpp
@@ -44,7 +44,7 @@ const auto gather7Params_1D = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Gather7_1D, Gather7LayerTest, gather7Params_1D, Gather7LayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Gather7_1D, Gather7LayerTest, gather7Params_1D, Gather7LayerTest::getTestCaseName);
 
 const std::vector<std::vector<size_t>> inputShapes_2D = {
         std::vector<size_t>{4, 19},
@@ -74,7 +74,7 @@ const auto gather7Params_2D = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Gather7_2D, Gather7LayerTest, gather7Params_2D, Gather7LayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Gather7_2D, Gather7LayerTest, gather7Params_2D, Gather7LayerTest::getTestCaseName);
 
 const std::vector<std::vector<size_t>> inputShapes4D = {
         std::vector<size_t>{4, 5, 6, 7},
@@ -105,7 +105,7 @@ const auto gather7ParamsSubset_BD0 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Gather7_BD0, Gather7LayerTest, gather7ParamsSubset_BD0, Gather7LayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Gather7_BD0, Gather7LayerTest, gather7ParamsSubset_BD0, Gather7LayerTest::getTestCaseName);
 
 const std::vector<std::vector<size_t>> indicesShapes_BD1 = {
         std::vector<size_t>{4, 2},
@@ -132,7 +132,7 @@ const auto gather7ParamsSubset_BD1 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Gather7_BD1, Gather7LayerTest, gather7ParamsSubset_BD1, Gather7LayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Gather7_BD1, Gather7LayerTest, gather7ParamsSubset_BD1, Gather7LayerTest::getTestCaseName);
 
 const std::vector<std::vector<size_t>> indicesShapes_BD2 = {
         std::vector<size_t>{4, 5, 4, 3},
@@ -158,7 +158,7 @@ const auto gather7ParamsSubset_BD2 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Gather7_BD2, Gather7LayerTest, gather7ParamsSubset_BD2, Gather7LayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Gather7_BD2, Gather7LayerTest, gather7ParamsSubset_BD2, Gather7LayerTest::getTestCaseName);
 
 const std::vector<std::vector<size_t>> indicesShapes_NegativeBD = {
         std::vector<size_t>{4, 5, 4},
@@ -186,6 +186,6 @@ const auto gather7ParamsSubset_NegativeBD = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Gather7_NegativeBD, Gather7LayerTest, gather7ParamsSubset_NegativeBD, Gather7LayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Gather7_NegativeBD, Gather7LayerTest, gather7ParamsSubset_NegativeBD, Gather7LayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_elements.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_elements.cpp
index b07dfe252e1370..1ad8bbd0d4c335 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_elements.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_elements.cpp
@@ -24,7 +24,7 @@ const std::vector<InferenceEngine::Precision> iPrecisions = {
         InferenceEngine::Precision::I64
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_set1, GatherElementsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_set1, GatherElementsLayerTest,
                         ::testing::Combine(
                             ::testing::Values(std::vector<size_t>({2, 2})),     // Data shape
                             ::testing::Values(std::vector<size_t>({2, 2})),     // Indices shape
@@ -34,7 +34,7 @@ INSTANTIATE_TEST_CASE_P(smoke_set1, GatherElementsLayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         GatherElementsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_set2, GatherElementsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_set2, GatherElementsLayerTest,
                         ::testing::Combine(
                             ::testing::Values(std::vector<size_t>({2, 2, 1})),  // Data shape
                             ::testing::Values(std::vector<size_t>({4, 2, 1})),  // Indices shape
@@ -44,7 +44,7 @@ INSTANTIATE_TEST_CASE_P(smoke_set2, GatherElementsLayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         GatherElementsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_set3, GatherElementsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_set3, GatherElementsLayerTest,
                         ::testing::Combine(
                             ::testing::Values(std::vector<size_t>({2, 2, 3, 5})),   // Data shape
                             ::testing::Values(std::vector<size_t>({2, 2, 3, 7})),   // Indices shape
@@ -54,7 +54,7 @@ INSTANTIATE_TEST_CASE_P(smoke_set3, GatherElementsLayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         GatherElementsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_set4, GatherElementsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_set4, GatherElementsLayerTest,
                         ::testing::Combine(
                             ::testing::Values(std::vector<size_t>({3, 2, 3, 8})),   // Data shape
                             ::testing::Values(std::vector<size_t>({2, 2, 3, 8})),   // Indices shape
@@ -64,7 +64,7 @@ INSTANTIATE_TEST_CASE_P(smoke_set4, GatherElementsLayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         GatherElementsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_set5, GatherElementsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_set5, GatherElementsLayerTest,
                         ::testing::Combine(
                             ::testing::Values(std::vector<size_t>({3, 2, 3, 4, 8})),   // Data shape
                             ::testing::Values(std::vector<size_t>({3, 2, 3, 5, 8})),   // Indices shape
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_nd.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_nd.cpp
index a467f96ec22fb8..1639dabcb557d6 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_nd.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_nd.cpp
@@ -31,7 +31,7 @@ const auto gatherNDArgsSubset1 = ::testing::Combine(
             {{2, 1}, {2, 1, 1}})),                                // Indices shape
         ::testing::ValuesIn(std::vector<int>({0, 1}))             // Batch dims
 );
-INSTANTIATE_TEST_CASE_P(smoke_Set1, GatherNDLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Set1, GatherNDLayerTest,
                         ::testing::Combine(
                             gatherNDArgsSubset1,
                             ::testing::ValuesIn(dPrecisions),
@@ -47,7 +47,7 @@ const auto gatherNDArgsSubset2 = ::testing::Combine(
             {{15, 12, 2}, {15, 12, 5, 9, 1, 3}})),                 // Indices shape
         ::testing::ValuesIn(std::vector<int>({0, 1, 2}))           // Batch dims
 );
-INSTANTIATE_TEST_CASE_P(smoke_Set2, GatherNDLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Set2, GatherNDLayerTest,
                         ::testing::Combine(
                             gatherNDArgsSubset2,
                             ::testing::ValuesIn(dPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_tree.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_tree.cpp
index cb19f2e2b3bbdc..0076fd7b8fe588 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_tree.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gather_tree.cpp
@@ -23,7 +23,7 @@ const std::vector<ngraph::helpers::InputLayerType> secondaryInputTypes = {
         ngraph::helpers::InputLayerType::PARAMETER
 };
 
-INSTANTIATE_TEST_CASE_P(Basic_smoke, GatherTreeLayerTest,
+INSTANTIATE_TEST_SUITE_P(Basic_smoke, GatherTreeLayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(inputShapes),
                             ::testing::ValuesIn(secondaryInputTypes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/group_convolution.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/group_convolution.cpp
index 417ad9d0130096..10da188bcb7e3e 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/group_convolution.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/group_convolution.cpp
@@ -39,7 +39,7 @@ const auto groupConv1DParams_AutoPadValid = ::testing::Combine(
     ::testing::ValuesIn(numGroups1d),
     ::testing::Values(ngraph::op::PadType::VALID));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_GroupConvolution1D_ExplicitPadding, GroupConvolutionLayerTest,
     ::testing::Combine(
         groupConv1DParams_ExplicitPadding, ::testing::ValuesIn(netPrecisions),
@@ -51,7 +51,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     GroupConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_GroupConvolution1D_AutoPadValid, GroupConvolutionLayerTest,
     ::testing::Combine(
         groupConv1DParams_AutoPadValid, ::testing::ValuesIn(netPrecisions),
@@ -87,7 +87,7 @@ const auto groupConv2DParams_AutoPadValid = ::testing::Combine(
     ::testing::ValuesIn(numGroups),
     ::testing::Values(ngraph::op::PadType::VALID));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_GroupConvolution2D_ExplicitPadding, GroupConvolutionLayerTest,
     ::testing::Combine(
         groupConv2DParams_ExplicitPadding, ::testing::ValuesIn(netPrecisions),
@@ -99,7 +99,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     GroupConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_GroupConvolution2D_AutoPadValid, GroupConvolutionLayerTest,
     ::testing::Combine(
         groupConv2DParams_AutoPadValid, ::testing::ValuesIn(netPrecisions),
@@ -130,7 +130,7 @@ const auto groupConv3DParams_AutoPadValid = ::testing::Combine(
     ::testing::ValuesIn(dilations3d), ::testing::Values(4),
     ::testing::Values(2), ::testing::Values(ngraph::op::PadType::VALID));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_GroupConvolution3D_ExplicitPadding, GroupConvolutionLayerTest,
     ::testing::Combine(
         groupConv3DParams_ExplicitPadding, ::testing::ValuesIn(netPrecisions),
@@ -142,7 +142,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)),
     GroupConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_GroupConvolution3D_AutoPadValid, GroupConvolutionLayerTest,
     ::testing::Combine(
         groupConv3DParams_AutoPadValid, ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/group_convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/group_convolution_backprop_data.cpp
index 70a1151d9bae1d..2f08be54039156 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/group_convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/group_convolution_backprop_data.cpp
@@ -50,7 +50,7 @@ const auto groupConvBackpropData1DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::VALID)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData1D_ExplicitPadding, GroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvBackpropData1D_ExplicitPadding, GroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 groupConvBackpropData1DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -62,7 +62,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData1D_ExplicitPadding, GroupConv
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         GroupConvBackpropDataLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData1D_AutoPadValid, GroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvBackpropData1D_AutoPadValid, GroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 groupConvBackpropData1DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -104,7 +104,7 @@ const auto groupConvBackpropData2DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::VALID)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData2D_ExplicitPadding, GroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvBackpropData2D_ExplicitPadding, GroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 groupConvBackpropData2DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -116,7 +116,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData2D_ExplicitPadding, GroupConv
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         GroupConvBackpropDataLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData2D_AutoPadValid, GroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvBackpropData2D_AutoPadValid, GroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 groupConvBackpropData2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -158,7 +158,7 @@ const auto groupConvBackpropData3DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::VALID)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData3D_ExplicitPadding, GroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvBackpropData3D_ExplicitPadding, GroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 groupConvBackpropData3DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -170,7 +170,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData3D_ExplicitPadding, GroupConv
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         GroupConvBackpropDataLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData3D_AutoPadValid, GroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvBackpropData3D_AutoPadValid, GroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 groupConvBackpropData3DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gru_cell.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gru_cell.cpp
index 51e7dad83ca208..350d6dc74f5198 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gru_cell.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gru_cell.cpp
@@ -21,7 +21,7 @@ namespace {
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                              InferenceEngine::Precision::FP16};
 
-    INSTANTIATE_TEST_CASE_P(smoke_GRUCellCommon, GRUCellTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_GRUCellCommon, GRUCellTest,
             ::testing::Combine(
             ::testing::ValuesIn(should_decompose),
             ::testing::ValuesIn(batch),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gru_sequence.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gru_sequence.cpp
index 5521023a192fd2..884a6e3bd2897d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gru_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gru_sequence.cpp
@@ -34,7 +34,7 @@ namespace {
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                              InferenceEngine::Precision::FP16};
 
-    INSTANTIATE_TEST_CASE_P(smoke_GRUSequenceCommonZeroClip, GRUSequenceTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_GRUSequenceCommonZeroClip, GRUSequenceTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(mode),
                                     ::testing::ValuesIn(seq_lengths_zero_clip),
@@ -49,7 +49,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             GRUSequenceTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_GRUSequenceCommonClip, GRUSequenceTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_GRUSequenceCommonClip, GRUSequenceTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(mode),
                                     ::testing::ValuesIn(seq_lengths_clip_non_zero),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/interpolate.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/interpolate.cpp
index 08c529379767a7..5d617ce9c90ea0 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/interpolate.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/interpolate.cpp
@@ -108,7 +108,7 @@ const auto interpolateCases = ::testing::Combine(
         ::testing::ValuesIn(defaultAxes),
         ::testing::ValuesIn(defaultScales));
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_Basic, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_Basic, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesWithoutNearest,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -121,7 +121,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Interpolate_Basic, InterpolateLayerTest, ::testing
         ::testing::Values(additional_config)),
     InterpolateLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_Nearest, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_Nearest, InterpolateLayerTest, ::testing::Combine(
         interpolateCases,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -166,7 +166,7 @@ const auto interpolateCasesTail = ::testing::Combine(
         ::testing::ValuesIn(defaultAxes),
         ::testing::ValuesIn(defaultScalesTailTest));
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_Basic_Down_Sample_Tail, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_Basic_Down_Sample_Tail, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesWithoutNearestTail,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -179,7 +179,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Interpolate_Basic_Down_Sample_Tail, InterpolateLay
         ::testing::Values(additional_config)),
     InterpolateLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_Nearest_Down_Sample_Tail, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_Nearest_Down_Sample_Tail, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesTail,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/log_softmax.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/log_softmax.cpp
index 37042c8ad7d9cb..2b9a71ddfeffe0 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/log_softmax.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/log_softmax.cpp
@@ -37,7 +37,7 @@ const auto params2D = testing::Combine(
     testing::Values(std::map<std::string, std::string>())
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_LogSoftmax2D,
         LogSoftmaxLayerTest,
         params2D,
@@ -66,7 +66,7 @@ const auto params4D = testing::Combine(
     testing::Values(std::map<std::string, std::string>())
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_LogSoftmax4D,
         LogSoftmaxLayerTest,
         params4D,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/logical.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/logical.cpp
index 0e252f6753873c..b90ed72b717c50 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/logical.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/logical.cpp
@@ -74,8 +74,8 @@ const auto LogicalTestParamsNot = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(additional_config));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs, LogicalLayerTest, LogicalTestParams, LogicalLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs, LogicalLayerTest, LogicalTestParams, LogicalLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefsNot, LogicalLayerTest, LogicalTestParamsNot, LogicalLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefsNot, LogicalLayerTest, LogicalTestParamsNot, LogicalLayerTest::getTestCaseName);
 
 }  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/loop.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/loop.cpp
index 1a14f9a97e4cbe..505d5f7f29d617 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/loop.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/loop.cpp
@@ -20,7 +20,7 @@ namespace {
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                              InferenceEngine::Precision::FP16};
 
-    INSTANTIATE_TEST_CASE_P(smoke_LoopCommonZeroClip, LoopTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_LoopCommonZeroClip, LoopTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(execute_first_iteration),
                                     ::testing::ValuesIn(is_body_condition_const),
@@ -48,7 +48,7 @@ namespace {
     using namespace testing;
     using namespace InferenceEngine;
 
-    INSTANTIATE_TEST_CASE_P(smoke_StaticShapeLoop, StaticShapeLoopTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_StaticShapeLoop, StaticShapeLoopTest,
                             Combine(
                                     ValuesIn(std::vector<bool>{true, false}),
                                     Values(true),
@@ -58,7 +58,7 @@ namespace {
                                     Values<InferenceEngine::Precision>(Precision::FP32, Precision::I32),
                                     Values(CommonTestUtils::DEVICE_CPU)));
     using namespace testing;
-    INSTANTIATE_TEST_CASE_P(smoke_TrivialLoop, TrivialLoopTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_TrivialLoop, TrivialLoopTest,
                             Combine(
                                     Values<InferenceEngine::Precision>(Precision::FP32, Precision::I32),
                                     Values<InferenceEngine::SizeVector>({2, 3, 4}),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lrn.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lrn.cpp
index 916abf39f4cd8d..4863b9934486ad 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lrn.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lrn.cpp
@@ -22,7 +22,7 @@ namespace LRN2D {
 
 const std::vector<std::vector<int64_t>> axes = {{1}};
 
-INSTANTIATE_TEST_CASE_P(smoke_LrnCheck_2D, LrnLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_LrnCheck_2D, LrnLayerTest,
                         ::testing::Combine(::testing::Values(alpha),
                                            ::testing::Values(beta),
                                            ::testing::Values(bias),
@@ -41,7 +41,7 @@ namespace LRN3D {
 
 const std::vector<std::vector<int64_t>> axes = {{1}, {2}};
 
-INSTANTIATE_TEST_CASE_P(smoke_LrnCheck_3D, LrnLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_LrnCheck_3D, LrnLayerTest,
                         ::testing::Combine(::testing::Values(alpha),
                                            ::testing::Values(beta),
                                            ::testing::Values(bias),
@@ -60,7 +60,7 @@ namespace LRN4D {
 
 const std::vector<std::vector<int64_t>> axes = {{1}, {2, 3}, {3, 2}};
 
-INSTANTIATE_TEST_CASE_P(smoke_LrnCheck_4D, LrnLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_LrnCheck_4D, LrnLayerTest,
                         ::testing::Combine(::testing::Values(alpha),
                                            ::testing::Values(beta),
                                            ::testing::Values(bias),
@@ -79,7 +79,7 @@ namespace LRN5D {
 
 const std::vector<std::vector<int64_t>> axes = {{1}, {2, 3, 4}, {4, 2, 3}};
 
-INSTANTIATE_TEST_CASE_P(smoke_LrnCheck_5D, LrnLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_LrnCheck_5D, LrnLayerTest,
                         ::testing::Combine(::testing::Values(alpha),
                                            ::testing::Values(beta),
                                            ::testing::Values(bias),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lstm_cell.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lstm_cell.cpp
index cc1f0a4fa783bf..5affa8d67d3331 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lstm_cell.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lstm_cell.cpp
@@ -21,7 +21,7 @@ namespace {
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                              InferenceEngine::Precision::FP16};
 
-    INSTANTIATE_TEST_CASE_P(smoke_LSTMCellCommon, LSTMCellTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_LSTMCellCommon, LSTMCellTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(should_decompose),
                                     ::testing::ValuesIn(batch),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lstm_sequence.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lstm_sequence.cpp
index 3ccbddc8f31c35..aea380561198ef 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lstm_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lstm_sequence.cpp
@@ -34,7 +34,7 @@ namespace {
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                              InferenceEngine::Precision::FP16};
 
-    INSTANTIATE_TEST_CASE_P(smoke_LSTMSequenceCommonZeroClip, LSTMSequenceTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_LSTMSequenceCommonZeroClip, LSTMSequenceTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(mode),
                                     ::testing::ValuesIn(seq_lengths_zero_clip),
@@ -48,7 +48,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             LSTMSequenceTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_LSTMSequenceCommonClip, LSTMSequenceTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_LSTMSequenceCommonClip, LSTMSequenceTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(mode),
                                     ::testing::ValuesIn(seq_lengths_clip_non_zero),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/mat_mul.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/mat_mul.cpp
index 4a0d5dcdcbd556..680276b2f721e1 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/mat_mul.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/mat_mul.cpp
@@ -44,7 +44,7 @@ std::vector<ngraph::helpers::InputLayerType> secondaryInputTypes = {
 
 std::map<std::string, std::string> additional_config = {};
 
-INSTANTIATE_TEST_CASE_P(smoke_MatMul, MatMulTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MatMul, MatMulTest,
         ::testing::Combine(
                 ::testing::ValuesIn(shapeRelatedParams),
                 ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp
index 062ea0cad9194e..c0a0fa3b430e89 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/memory.cpp
@@ -32,7 +32,7 @@ const std::vector<int64_t> iterationCount {
         10
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MemoryTest, MemoryTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MemoryTest, MemoryTest,
         ::testing::Combine(
                 ::testing::ValuesIn(transformation),
                 ::testing::ValuesIn(iterationCount),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/minimum_maximum.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/minimum_maximum.cpp
index 64c3878736eb72..e2f5f72497e15d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/minimum_maximum.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/minimum_maximum.cpp
@@ -35,7 +35,7 @@ const std::vector<ngraph::helpers::InputLayerType> inputType = {
         ngraph::helpers::InputLayerType::PARAMETER,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_maximum, MaxMinLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_maximum, MaxMinLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inShapes),
                                 ::testing::ValuesIn(opType),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/mvn.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/mvn.cpp
index c633c1b6bc82dd..73fdce483efd20 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/mvn.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/mvn.cpp
@@ -50,7 +50,7 @@ const auto MvnCases = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsMVN, MvnLayerTest, MvnCases, MvnLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsMVN, MvnLayerTest, MvnCases, MvnLayerTest::getTestCaseName);
 
 
 
@@ -73,7 +73,7 @@ const std::vector<float> epsilonF = {
     0.0001
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_5D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_5D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 10, 5, 7, 8}, {1, 3, 8, 9, 49}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -85,7 +85,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVN_5D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_4D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_4D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 10, 5, 17}, {1, 3, 8, 9}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -97,7 +97,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVN_4D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_3D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_3D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 32, 17}, {1, 37, 9}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -109,7 +109,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVN_3D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_2D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_2D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{3, 5}, {2, 55}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -121,7 +121,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVN_2D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_1D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_1D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{3}, {9}, {55}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -133,7 +133,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVN_1D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Decomposition_3D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Decomposition_3D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 32, 17}, {1, 37, 9}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -145,7 +145,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Decomposition_3D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Decomposition_4D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Decomposition_4D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 16, 5, 8}, {2, 19, 5, 10}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -157,7 +157,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Decomposition_4D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Decomposition_10D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Decomposition_10D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 3, 5, 4, 2, 6, 5, 3, 2, 1}}),
                             ::testing::ValuesIn(dataPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/non_max_suppression.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/non_max_suppression.cpp
index a8d10d65bdf329..4bc28cf1b5b747 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/non_max_suppression.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/non_max_suppression.cpp
@@ -39,4 +39,4 @@ const auto nmsParams = ::testing::Combine(::testing::ValuesIn(inShapeParams),
                                           ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_NmsLayerTest, NmsLayerTest, nmsParams, NmsLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_NmsLayerTest, NmsLayerTest, nmsParams, NmsLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/normalize_l2.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/normalize_l2.cpp
index 75b2c0a5b492cd..843609102e44ce 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/normalize_l2.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/normalize_l2.cpp
@@ -34,7 +34,7 @@ const auto normL2params = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         NormalizeL2,
         NormalizeL2LayerTest,
         normL2params,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/one_hot.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/one_hot.cpp
index 9ab687d76f2834..7a14b67f0ea183 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/one_hot.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/one_hot.cpp
@@ -32,7 +32,7 @@ const auto oneHotParams_IC = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_OneHotIntConst,
         OneHotLayerTest,
         oneHotParams_IC,
@@ -60,7 +60,7 @@ const auto oneHotParams_Ax = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_OneHotAxrng,
         OneHotLayerTest,
         oneHotParams_Ax,
@@ -89,7 +89,7 @@ const auto oneHotParams_T = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_OneHotArgType,
         OneHotLayerTest,
         oneHotParams_T,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/pad.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/pad.cpp
index 0a760ed69e59e6..17e9f4e17dfaf8 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/pad.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/pad.cpp
@@ -43,7 +43,7 @@ const auto pad1DConstparams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad1DConst,
         PadLayerTest,
         pad1DConstparams,
@@ -63,7 +63,7 @@ const auto pad1Dparams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad1D,
         PadLayerTest,
         pad1Dparams,
@@ -86,7 +86,7 @@ const auto pad2DConstparams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad2DConst,
         PadLayerTest,
         pad2DConstparams,
@@ -106,7 +106,7 @@ const auto pad2Dparams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad2D,
         PadLayerTest,
         pad2Dparams,
@@ -129,7 +129,7 @@ const auto pad4DConstparams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad4DConst,
         PadLayerTest,
         pad4DConstparams,
@@ -149,7 +149,7 @@ const auto pad4Dparams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad4D,
         PadLayerTest,
         pad4Dparams,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/pooling.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/pooling.cpp
index 0e71de6d4c33af..a3053fb6809820 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/pooling.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/pooling.cpp
@@ -54,7 +54,7 @@ const auto maxPool_ExplicitPad_FloorRounding_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_ExplicitPad_FloorRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_ExplicitPad_FloorRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_ExplicitPad_FloorRounding_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -78,7 +78,7 @@ const auto maxPool_SameUpperPad_FloorRounding_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_SameUpperPad_FloorRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_SameUpperPad_FloorRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_SameUpperPad_FloorRounding_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -102,7 +102,7 @@ const auto maxPool_SameLowerPad_FloorRounding_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_SameLowerPad_FloorRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_SameLowerPad_FloorRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_SameUpperPad_FloorRounding_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -126,7 +126,7 @@ const auto maxPool_ExplicitPad_FloorRounding_5Dinput_Params = ::testing::Combine
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_ExplicitPad_FloorRounding_5Dinput, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_ExplicitPad_FloorRounding_5Dinput, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_ExplicitPad_FloorRounding_5Dinput_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -150,7 +150,7 @@ const auto maxPool_SameUpperPad_FloorRounding_5Dinput_Params = ::testing::Combin
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_SameUpperPad_FloorRounding_5Dinput, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_SameUpperPad_FloorRounding_5Dinput, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_SameUpperPad_FloorRounding_5Dinput_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -174,7 +174,7 @@ const auto maxPool_SameLowerPad_CeilRounding_5Dinput_Params = ::testing::Combine
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_SameLowerPad_CeilRounding_5Dinput, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_SameLowerPad_CeilRounding_5Dinput, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_SameUpperPad_FloorRounding_5Dinput_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -198,7 +198,7 @@ const auto maxPool_ExplicitPad_CeilRounding_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_ExplicitPad_CeilRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_ExplicitPad_CeilRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_ExplicitPad_CeilRounding_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -225,7 +225,7 @@ const auto avgPoolExplicitPadCeilRoundingParams = ::testing::Combine(
         ::testing::Values(true, false)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_ExplicitPad_CeilRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_ExplicitPad_CeilRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 avgPoolExplicitPadCeilRoundingParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -242,7 +242,7 @@ std::vector<poolSpecificParams> psParams({poolSpecificParams(ngraph::helpers::Po
     poolSpecificParams(ngraph::helpers::PoolingTypes::AVG, {7, 7}, {1, 1}, {0, 0}, {1, 1},
                         ngraph::op::RoundingType::CEIL, ngraph::op::PadType::EXPLICIT, false)});
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_ExplicitPad_CeilRounding_corner, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_ExplicitPad_CeilRounding_corner, PoolingLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(psParams),
                                 ::testing::ValuesIn(netPrecisions),
@@ -267,7 +267,7 @@ const auto avgPoolExplicitPadFloorRoundingParams = ::testing::Combine(
 );
 
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_ExplicitPad_FloorRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_ExplicitPad_FloorRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 avgPoolExplicitPadFloorRoundingParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -291,7 +291,7 @@ const auto avgPool_ExplicitPad_FloorRounding_5Dinput_Params = ::testing::Combine
         ::testing::Values(true, false)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_ExplicitPad_FloorRounding_5Dinput, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_ExplicitPad_FloorRounding_5Dinput, PoolingLayerTest,
                         ::testing::Combine(
                                 avgPool_ExplicitPad_FloorRounding_5Dinput_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -315,7 +315,7 @@ const auto avgPool_SameUpperPad_FloorRounding_5Dinput_Params = ::testing::Combin
         ::testing::Values(true)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_SameUpperPad_FloorRounding_5Dinput, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_SameUpperPad_FloorRounding_5Dinput, PoolingLayerTest,
                         ::testing::Combine(
                                 avgPool_SameUpperPad_FloorRounding_5Dinput_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -339,7 +339,7 @@ const auto avgPool_SameLowerPad_CeilRounding_5Dinput_Params = ::testing::Combine
         ::testing::Values(true)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_SameLowerPad_CeilRounding_5Dinput, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_SameLowerPad_CeilRounding_5Dinput, PoolingLayerTest,
                         ::testing::Combine(
                                 avgPool_SameLowerPad_CeilRounding_5Dinput_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -365,7 +365,7 @@ const auto allPools_ValidPad_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MAX_and_AVGPool_ValidPad, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MAX_and_AVGPool_ValidPad, PoolingLayerTest,
                         ::testing::Combine(
                                 allPools_ValidPad_Params,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/proposal.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/proposal.cpp
index df5352605093ba..a225800b954cd4 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/proposal.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/proposal.cpp
@@ -39,7 +39,7 @@ const auto proposalParams = ::testing::Combine(
         ::testing::ValuesIn(framework_)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Proposal_tests, ProposalLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Proposal_tests, ProposalLayerTest,
                         ::testing::Combine(
                                 proposalParams,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/psroi_pooling.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/psroi_pooling.cpp
index 6c896b56aa8845..cca6d83a882ac9 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/psroi_pooling.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/psroi_pooling.cpp
@@ -24,7 +24,7 @@ const auto PSROICases_average = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsPSROIPooling_average, PSROIPoolingLayerTest, PSROICases_average, PSROIPoolingLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsPSROIPooling_average, PSROIPoolingLayerTest, PSROICases_average, PSROIPoolingLayerTest::getTestCaseName);
 
 
 const auto PSROICases_bilinear = ::testing::Combine(
@@ -40,4 +40,4 @@ const auto PSROICases_bilinear = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsPSROIPooling_bilinear, PSROIPoolingLayerTest, PSROICases_bilinear, PSROIPoolingLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsPSROIPooling_bilinear, PSROIPoolingLayerTest, PSROICases_bilinear, PSROIPoolingLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/range.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/range.cpp
index acc17b1142eed3..c4bfa1f1567f4e 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/range.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/range.cpp
@@ -20,7 +20,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP16
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, RangeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, RangeLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(start),
                                 ::testing::ValuesIn(stop),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reduce_ops.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reduce_ops.cpp
index f5d96aa0d5d330..2608c1bb4584f2 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reduce_ops.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reduce_ops.cpp
@@ -166,56 +166,56 @@ const auto params_ReductionTypesLogical = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceOneAxis,
         ReduceOpsLayerTest,
         paramsOneAxis,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceLogicalOneAxis,
         ReduceOpsLayerTest,
         paramsOneAxisLogical,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_Precisions,
         ReduceOpsLayerTest,
         params_Precisions,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_InputShapes,
         ReduceOpsLayerTest,
         params_InputShapes,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_Axes,
         ReduceOpsLayerTest,
         params_Axes,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_ReductionTypes,
         ReduceOpsLayerTest,
         params_ReductionTypes,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceLogical_ReductionTypes,
         ReduceOpsLayerTest,
         params_ReductionTypesLogical,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce,
         ReduceOpsLayerWithSpecificInputTest,
         testing::Combine(
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/region_yolo.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/region_yolo.cpp
index ac4e3b384039b9..8961b4a91882f9 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/region_yolo.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/region_yolo.cpp
@@ -80,6 +80,6 @@ const auto testCase_yolov2_caffe = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsRegionYolov3, RegionYoloLayerTest, testCase_yolov3, RegionYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsRegionYoloMxnet, RegionYoloLayerTest, testCase_yolov3_mxnet, RegionYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsRegionYoloCaffe, RegionYoloLayerTest, testCase_yolov2_caffe, RegionYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsRegionYolov3, RegionYoloLayerTest, testCase_yolov3, RegionYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsRegionYoloMxnet, RegionYoloLayerTest, testCase_yolov3_mxnet, RegionYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsRegionYoloCaffe, RegionYoloLayerTest, testCase_yolov2_caffe, RegionYoloLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reorg_yolo.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reorg_yolo.cpp
index 32a1f934968d43..c288688cd548ea 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reorg_yolo.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reorg_yolo.cpp
@@ -67,9 +67,9 @@ const auto testCase_batch_2 = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_caffe_YoloV2, ReorgYoloLayerTest, testCase_caffe_yolov2, ReorgYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_stride_2_smallest, ReorgYoloLayerTest, testCase_smallest, ReorgYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_stride_2, ReorgYoloLayerTest, testCase_stride_2, ReorgYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_stride_3, ReorgYoloLayerTest, testCase_stride_3, ReorgYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_smaller_h, ReorgYoloLayerTest, testCase_smaller_h, ReorgYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_batch_2, ReorgYoloLayerTest, testCase_batch_2, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_caffe_YoloV2, ReorgYoloLayerTest, testCase_caffe_yolov2, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_stride_2_smallest, ReorgYoloLayerTest, testCase_smallest, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_stride_2, ReorgYoloLayerTest, testCase_stride_2, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_stride_3, ReorgYoloLayerTest, testCase_stride_3, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_smaller_h, ReorgYoloLayerTest, testCase_smaller_h, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_batch_2, ReorgYoloLayerTest, testCase_batch_2, ReorgYoloLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reshape.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reshape.cpp
index 28419ec971fca7..6baed1f21fc1ac 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reshape.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reshape.cpp
@@ -15,7 +15,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP16
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheckDynBatch, ReshapeLayerTestRevise,
+INSTANTIATE_TEST_SUITE_P(smoke_ReshapeCheckDynBatch, ReshapeLayerTestRevise,
         ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
@@ -29,7 +29,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheckDynBatch, ReshapeLayerTestRevise,
                 ::testing::Values(std::map<std::string, std::string>({{CONFIG_KEY(DYN_BATCH_ENABLED), CONFIG_VALUE(YES)}}))),
                 ReshapeLayerTestRevise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTestRevise,
+INSTANTIATE_TEST_SUITE_P(smoke_ReshapeCheck, ReshapeLayerTestRevise,
         ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
@@ -43,7 +43,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTestRevise,
                 ::testing::Values(std::map<std::string, std::string>({}))),
                 ReshapeLayerTestRevise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheckNegative, ReshapeLayerTestRevise,
+INSTANTIATE_TEST_SUITE_P(smoke_ReshapeCheckNegative, ReshapeLayerTestRevise,
         ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/result.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/result.cpp
index fd30ef30a0ab30..f352eacaa39661 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/result.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/result.cpp
@@ -21,7 +21,7 @@ std::vector<InferenceEngine::Precision> inputPrecisions = {
 
 ConfigMap config;
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_ResultLayerTest, ResultLayerTest,
     ::testing::Combine(::testing::ValuesIn(inputDims),
                        ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reverse_sequence.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reverse_sequence.cpp
index 91e0fc5bf5f1ba..b1f864d12ae14f 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reverse_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/reverse_sequence.cpp
@@ -33,7 +33,7 @@ const std::vector<ngraph::helpers::InputLayerType> secondaryInputTypes = {
         ngraph::helpers::InputLayerType::PARAMETER
 };
 
-INSTANTIATE_TEST_CASE_P(Basic_smoke, ReverseSequenceLayerTest,
+INSTANTIATE_TEST_SUITE_P(Basic_smoke, ReverseSequenceLayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(batchAxisIndices),
                             ::testing::ValuesIn(seqAxisIndices),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/rnn_cell.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/rnn_cell.cpp
index db9c450dba8006..e730cc0d2f9bfb 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/rnn_cell.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/rnn_cell.cpp
@@ -19,7 +19,7 @@ namespace {
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                              InferenceEngine::Precision::FP16};
 
-    INSTANTIATE_TEST_CASE_P(smoke_RNNCellCommon, RNNCellTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_RNNCellCommon, RNNCellTest,
             ::testing::Combine(
             ::testing::ValuesIn(should_decompose),
             ::testing::ValuesIn(batch),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/rnn_sequence.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/rnn_sequence.cpp
index 3db31d471c532c..e5f7b821564ade 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/rnn_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/rnn_sequence.cpp
@@ -31,7 +31,7 @@ namespace {
     };
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32};
 
-    INSTANTIATE_TEST_CASE_P(smoke_RNNSequenceCommonZeroClip, RNNSequenceTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_RNNSequenceCommonZeroClip, RNNSequenceTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(mode),
                                     ::testing::ValuesIn(seq_lengths_zero_clip),
@@ -45,7 +45,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             RNNSequenceTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_RNNSequenceCommonClip, RNNSequenceTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_RNNSequenceCommonClip, RNNSequenceTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(mode),
                                     ::testing::ValuesIn(seq_lengths_clip_non_zero),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roi_align.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roi_align.cpp
index 38c97de0ffbd15..b4cb181519c902 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roi_align.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roi_align.cpp
@@ -31,7 +31,7 @@ const auto ROIAlignCases_average = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsROIAlign_average, ROIAlignLayerTest, ROIAlignCases_average, ROIAlignLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsROIAlign_average, ROIAlignLayerTest, ROIAlignCases_average, ROIAlignLayerTest::getTestCaseName);
 
 const auto ROIAlignCases_max = ::testing::Combine(
         ::testing::ValuesIn(
@@ -49,4 +49,4 @@ const auto ROIAlignCases_max = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsROIAlign_max, ROIAlignLayerTest, ROIAlignCases_max, ROIAlignLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsROIAlign_max, ROIAlignLayerTest, ROIAlignCases_max, ROIAlignLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roi_pooling.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roi_pooling.cpp
index 8387a754b43afe..279f5333e22f6a 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roi_pooling.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roi_pooling.cpp
@@ -61,5 +61,5 @@ const auto test_ROIPooling_bilinear = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsROIPooling_max, ROIPoolingLayerTest, test_ROIPooling_max, ROIPoolingLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsROIPooling_bilinear, ROIPoolingLayerTest, test_ROIPooling_bilinear, ROIPoolingLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsROIPooling_max, ROIPoolingLayerTest, test_ROIPooling_max, ROIPoolingLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsROIPooling_bilinear, ROIPoolingLayerTest, test_ROIPooling_bilinear, ROIPoolingLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roll.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roll.cpp
index f47d29704ded90..af67f1eb431990 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roll.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/roll.cpp
@@ -84,13 +84,13 @@ const auto testCaseUnordNegAxesAndShifts10D = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsRoll_2d_zero_shifts, RollLayerTest, testCase2DZeroShifts, RollLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsRoll_1d, RollLayerTest, testCase1D, RollLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsRoll_2d, RollLayerTest, testCase2D, RollLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsRoll_3d, RollLayerTest, testCase3D, RollLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsRoll_negative_unordered_axes_4d, RollLayerTest, testCaseNegativeUnorderedAxes4D, RollLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsRoll_negative_unordered_axes_5d, RollLayerTest, testCaseRepeatingAxes5D, RollLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsRoll_negative_shifts_6d, RollLayerTest, testCaseNegativeShifts6D, RollLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsRoll_unord_neg_shifts_and_axes_10d, RollLayerTest, testCaseUnordNegAxesAndShifts10D, RollLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsRoll_2d_zero_shifts, RollLayerTest, testCase2DZeroShifts, RollLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsRoll_1d, RollLayerTest, testCase1D, RollLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsRoll_2d, RollLayerTest, testCase2D, RollLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsRoll_3d, RollLayerTest, testCase3D, RollLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsRoll_negative_unordered_axes_4d, RollLayerTest, testCaseNegativeUnorderedAxes4D, RollLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsRoll_negative_unordered_axes_5d, RollLayerTest, testCaseRepeatingAxes5D, RollLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsRoll_negative_shifts_6d, RollLayerTest, testCaseNegativeShifts6D, RollLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsRoll_unord_neg_shifts_and_axes_10d, RollLayerTest, testCaseUnordNegAxesAndShifts10D, RollLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_ND_update.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_ND_update.cpp
index 081d07345347cc..cbccbd3793b2ed 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_ND_update.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_ND_update.cpp
@@ -36,6 +36,6 @@ const auto ScatterNDUpdateCases = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ScatterNDUpdate, ScatterNDUpdateLayerTest, ScatterNDUpdateCases, ScatterNDUpdateLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_ScatterNDUpdate, ScatterNDUpdateLayerTest, ScatterNDUpdateCases, ScatterNDUpdateLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_elements_update.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_elements_update.cpp
index 15b8852b101392..64c808b260aa18 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_elements_update.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_elements_update.cpp
@@ -42,7 +42,7 @@ const auto ScatterEltUpdateCases = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ScatterEltsUpdate, ScatterElementsUpdateLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ScatterEltsUpdate, ScatterElementsUpdateLayerTest,
     ScatterEltUpdateCases, ScatterElementsUpdateLayerTest::getTestCaseName);
 
 }  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_update.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_update.cpp
index b9ef4bbbfc9614..31362e0b2646b2 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_update.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/scatter_update.cpp
@@ -41,6 +41,6 @@ const auto ScatterUpdateCase = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ScatterUpdate, ScatterUpdateLayerTest, ScatterUpdateCase, ScatterUpdateLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_ScatterUpdate, ScatterUpdateLayerTest, ScatterUpdateCase, ScatterUpdateLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/select.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/select.cpp
index 315c3331db9800..424f692b1ef6aa 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/select.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/select.cpp
@@ -81,6 +81,6 @@ const auto numpyCases = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsSelect_none, SelectLayerTest, noneCases, SelectLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsSelect_none, SelectLayerTest, noneCases, SelectLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MKLDNN_TestsSelect_numpy, SelectLayerTest, numpyCases, SelectLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_MKLDNN_TestsSelect_numpy, SelectLayerTest, numpyCases, SelectLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shape_of.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shape_of.cpp
index 9dbd9a6abbff27..814d3c64f7f5a9 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shape_of.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shape_of.cpp
@@ -14,7 +14,7 @@ namespace {
             InferenceEngine::Precision::I32
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_Check, ShapeOfLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Check, ShapeOfLayerTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(std::vector<size_t>({10, 10, 10})),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
index b338d02efc14bd..dfc51edd9ca298 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
@@ -24,7 +24,7 @@ const auto shuffleChannelsParams4D = ::testing::Combine(
         ::testing::ValuesIn(groups)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels4D, ShuffleChannelsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels4D, ShuffleChannelsLayerTest,
         ::testing::Combine(
                 shuffleChannelsParams4D,
                 ::testing::ValuesIn(netPrecisions),
@@ -37,7 +37,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels4D, ShuffleChannelsLayerTest,
         ShuffleChannelsLayerTest::getTestCaseName);
 
 // ND support tests
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels6D, ShuffleChannelsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels6D, ShuffleChannelsLayerTest,
         ::testing::Combine(
                 ::testing::Values(std::tuple<int, int>(2, 3)),
                 ::testing::ValuesIn(netPrecisions),
@@ -49,7 +49,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels6D, ShuffleChannelsLayerTest,
                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ShuffleChannelsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels5D, ShuffleChannelsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels5D, ShuffleChannelsLayerTest,
         ::testing::Combine(
                 ::testing::Values(std::tuple<int, int>(2, 3)),
                 ::testing::ValuesIn(netPrecisions),
@@ -61,7 +61,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels5D, ShuffleChannelsLayerTest,
                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ShuffleChannelsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels3D, ShuffleChannelsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels3D, ShuffleChannelsLayerTest,
         ::testing::Combine(
                 ::testing::Values(std::tuple<int, int>(1, 3)),
                 ::testing::ValuesIn(netPrecisions),
@@ -73,7 +73,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels3D, ShuffleChannelsLayerTest,
                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ShuffleChannelsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels2D, ShuffleChannelsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels2D, ShuffleChannelsLayerTest,
         ::testing::Combine(
                 ::testing::Values(std::tuple<int, int>(1, 3)),
                 ::testing::ValuesIn(netPrecisions),
@@ -85,7 +85,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels2D, ShuffleChannelsLayerTest,
                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ShuffleChannelsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels1D, ShuffleChannelsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels1D, ShuffleChannelsLayerTest,
         ::testing::Combine(
                 ::testing::Values(std::tuple<int, int>(0, 3)),
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/softmax.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/softmax.cpp
index 4c90f92d98b117..23d4c6555da928 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/softmax.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/softmax.cpp
@@ -41,7 +41,7 @@ const auto params2D = testing::Combine(
     testing::Values(std::map<std::string, std::string>())
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_SoftMax2D,
         SoftMaxLayerTest,
         params2D,
@@ -68,7 +68,7 @@ const auto params4D = testing::Combine(
     testing::Values(std::map<std::string, std::string>())
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_SoftMax4D,
         SoftMaxLayerTest,
         params4D,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp
index 884440c7bdedbc..f6afb616f60655 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp
@@ -37,7 +37,7 @@ const auto SpaceToBatch4D = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_spacetobatch4D, SpaceToBatchLayerTest, SpaceToBatch4D,
         SpaceToBatchLayerTest::getTestCaseName);
 
@@ -67,7 +67,7 @@ const auto SpaceToBatch5D = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_spacetobatch5D, SpaceToBatchLayerTest, SpaceToBatch5D,
         SpaceToBatchLayerTest::getTestCaseName);
 
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_depth.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_depth.cpp
index ff7e33e495ec26..29d2af87ea4f29 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_depth.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/space_to_depth.cpp
@@ -36,7 +36,7 @@ const auto SpaceToDepthBS2 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_SpaceToDepthBS2, SpaceToDepthLayerTest, SpaceToDepthBS2, SpaceToDepthLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_SpaceToDepthBS2, SpaceToDepthLayerTest, SpaceToDepthBS2, SpaceToDepthLayerTest::getTestCaseName);
 
 const std::vector<std::vector<size_t >> inputShapesBS3 = {
         {1, 1, 3, 3}, {1, 1, 6, 6}, {1, 1, 9, 9}, {2, 4, 9, 9}, {2, 3, 15, 12},
@@ -51,6 +51,6 @@ const auto SpaceToDepthBS3 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_SpaceToDepthBS3, SpaceToDepthLayerTest, SpaceToDepthBS3, SpaceToDepthLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_SpaceToDepthBS3, SpaceToDepthLayerTest, SpaceToDepthBS3, SpaceToDepthLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/split.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/split.cpp
index 35d3149dcc361c..7c12638fd6d4ad 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/split.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/split.cpp
@@ -18,7 +18,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::U8
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NumSplitsCheck, SplitLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NumSplitsCheck, SplitLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(1, 2, 3, 5),
                                 ::testing::Values(0, 1, 2, 3),
@@ -32,7 +32,7 @@ INSTANTIATE_TEST_CASE_P(smoke_NumSplitsCheck, SplitLayerTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         SplitLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_splitWithUnusedOutputsTest, SplitLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_splitWithUnusedOutputsTest, SplitLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(5),
                                 ::testing::Values(0),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/squeeze_unsqueeze.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/squeeze_unsqueeze.cpp
index a5a8f35a35080b..2145c5d37eaaa9 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/squeeze_unsqueeze.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/squeeze_unsqueeze.cpp
@@ -38,7 +38,7 @@ const std::vector<ngraph::helpers::SqueezeOpType> opTypes = {
         ngraph::helpers::SqueezeOpType::UNSQUEEZE
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, SqueezeUnsqueezeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, SqueezeUnsqueezeLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(CommonTestUtils::combineParams(axesVectors)),
                                 ::testing::ValuesIn(opTypes),
@@ -50,7 +50,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Basic, SqueezeUnsqueezeLayerTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         SqueezeUnsqueezeLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic_emptyAxes, SqueezeUnsqueezeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Basic_emptyAxes, SqueezeUnsqueezeLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(CommonTestUtils::combineParams(emptyAxesVectors)),
                                 ::testing::Values(ngraph::helpers::SqueezeOpType::SQUEEZE),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/strided_slice.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/strided_slice.cpp
index 9897df6a8303ed..bd749a0b51c7d9 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/strided_slice.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/strided_slice.cpp
@@ -116,7 +116,7 @@ std::vector<StridedSliceSpecificParams> ss_only_test_cases = {
                             { 1, 1, 1, 0, 1 }, { 1, 1, 1, 0, 1 },  { },  { 0, 1, 0, 0, 0 },  { } },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_MKLDNN, StridedSliceLayerTest,
         ::testing::Combine(
             ::testing::ValuesIn(ss_only_test_cases),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/tensor_iterator.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/tensor_iterator.cpp
index 2a551b695f9685..279e4e98495eef 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/tensor_iterator.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/tensor_iterator.cpp
@@ -28,7 +28,7 @@ namespace {
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                              InferenceEngine::Precision::FP16};
 
-    INSTANTIATE_TEST_CASE_P(smoke_TensorIteratorCommon, TensorIteratorTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_TensorIteratorCommon, TensorIteratorTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(should_decompose),
                                     ::testing::ValuesIn(seq_lengths_zero_clip),
@@ -43,7 +43,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             TensorIteratorTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_TensorIteratorCommonClip, TensorIteratorTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_TensorIteratorCommonClip, TensorIteratorTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(should_decompose),
                                     ::testing::ValuesIn(seq_lengths_clip_non_zero),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/tile.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/tile.cpp
index 5894b1c1ec6724..bff906dd1019ec 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/tile.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/tile.cpp
@@ -22,7 +22,7 @@ const std::vector<std::vector<int64_t>> repeats3D = {
         {1, 1, 1}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Tile, TileLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Tile, TileLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(repeats3D),
                 ::testing::ValuesIn(netPrecisions),
@@ -39,7 +39,7 @@ const std::vector<std::vector<int64_t>> repeats6D = {
         {1, 1, 1, 1, 1, 1}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Tile6d, TileLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Tile6d, TileLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(repeats6D),
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/topk.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/topk.cpp
index d3da99f63690dc..dfb717a6b64df6 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/topk.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/topk.cpp
@@ -38,7 +38,7 @@ const std::vector<ngraph::opset4::TopK::SortType> sortTypes = {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_TopK, TopKLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_TopK, TopKLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(k),
                 ::testing::ValuesIn(axes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/transpose.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/transpose.cpp
index 044448b7843968..0575fc345b79dc 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/transpose.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/transpose.cpp
@@ -23,7 +23,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 std::vector<std::vector<size_t>> inputShape2D = {{2, 10}, {10, 2}, {10, 10}};
 std::vector<std::vector<size_t>> order2D      = {{}, {0, 1}, {1, 0}};
 
-INSTANTIATE_TEST_CASE_P(smoke_Transpose2D, TransposeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Transpose2D, TransposeLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(order2D),
                 ::testing::ValuesIn(netPrecisions),
@@ -43,7 +43,7 @@ std::vector<std::vector<size_t>> order4D      = {
         {3, 0, 1, 2}, {3, 0, 2, 1}, {3, 1, 0, 2}, {3, 1, 2, 0}, {3, 2, 0, 1}, {3, 2, 1, 0}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Transpose4D, TransposeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Transpose4D, TransposeLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(order4D),
                 ::testing::ValuesIn(netPrecisions),
@@ -61,7 +61,7 @@ std::vector<std::vector<size_t>> order5D      = {
         {1, 4, 2, 3, 0}, {2, 4, 1, 0, 3}, {3, 0, 2, 1, 4}, {4, 1, 0, 3, 2}, {0, 4, 1, 2, 3},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Transpose5D, TransposeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Transpose5D, TransposeLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(order5D),
                 ::testing::ValuesIn(netPrecisions),
@@ -79,7 +79,7 @@ std::vector<std::vector<size_t>> order6D      = {
         {1, 5, 4, 2, 3, 0}, {2, 5, 4, 1, 0, 3}, {3, 0, 2, 1, 4, 5}, {5, 1, 0, 4, 3, 2}, {0, 5, 1, 2, 3, 4},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Transpose6D, TransposeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Transpose6D, TransposeLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(order6D),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/variadic_split.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/variadic_split.cpp
index db0c9e156e6af5..7b9af2bbae2c29 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/variadic_split.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/variadic_split.cpp
@@ -25,7 +25,7 @@ namespace {
             {4, 11, 6, 9}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_NumSplitsCheck, VariadicSplitLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_NumSplitsCheck, VariadicSplitLayerTest,
             ::testing::Combine(
             ::testing::ValuesIn(numSplits),
             ::testing::Values(0, 1, 2, 3),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_add.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_add.cpp
index 72c50bce2a4f7f..940aa4bf5c81b0 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_add.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_add.cpp
@@ -16,7 +16,7 @@ namespace {
             InferenceEngine::Precision::FP32
     };
 
-    INSTANTIATE_TEST_CASE_P(NoReshape, CodegenAdd,
+    INSTANTIATE_TEST_SUITE_P(NoReshape, CodegenAdd,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(InferenceEngine::SizeVector({1, 42, 16, 64})),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_bert.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_bert.cpp
index 97b3737d69366d..ea76ba4d7dfb49 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_bert.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_bert.cpp
@@ -16,7 +16,7 @@ namespace {
             InferenceEngine::Precision::FP32
     };
 
-    INSTANTIATE_TEST_CASE_P(NoReshape, CodegenBert,
+    INSTANTIATE_TEST_SUITE_P(NoReshape, CodegenBert,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(InferenceEngine::SizeVector({1, 42, 16, 64})),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_gelu.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_gelu.cpp
index 81b03911dabe41..c7e9a5664b74ff 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_gelu.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/codegen_gelu.cpp
@@ -16,7 +16,7 @@ namespace {
             InferenceEngine::Precision::FP32
     };
 
-    INSTANTIATE_TEST_CASE_P(NoReshape, CodegenGelu,
+    INSTANTIATE_TEST_SUITE_P(NoReshape, CodegenGelu,
             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(InferenceEngine::SizeVector({1, 384, 4096})),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/constant_result.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/constant_result.cpp
index 19dcaacecda54e..10fb3563eee56d 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/constant_result.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/constant_result.cpp
@@ -34,7 +34,7 @@ const std::vector<Precision> precisions = {
     Precision::BOOL
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Check, ConstantResultSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Check, ConstantResultSubgraphTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(types),
                             ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/conv_eltwise_fusion.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/conv_eltwise_fusion.cpp
index 920891a726ae51..fe598de71bac73 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/conv_eltwise_fusion.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/conv_eltwise_fusion.cpp
@@ -16,7 +16,7 @@ namespace {
 #define ADD(X) std::tuple<ngraph::NodeTypeInfo, int64_t>(ngraph::opset4::Add::type_info, X)
 #define IN std::vector<std::tuple<ngraph::NodeTypeInfo, int64_t>>
 
-    INSTANTIATE_TEST_CASE_P(smoke_Convolution_1D, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_Convolution_1D, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::Convolution::type_info),
                                     ::testing::ValuesIn(IN({MUL(4), ADD(5)})),
@@ -27,7 +27,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             ConvEltwiseFusion::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution_1D, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolution_1D, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::GroupConvolution::type_info),
                                     ::testing::ValuesIn(IN({MUL(4), ADD(5)})),
@@ -38,7 +38,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             ConvEltwiseFusion::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData_1D, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData_1D, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::ConvolutionBackpropData::type_info),
                                     ::testing::ValuesIn(IN({MUL(4), ADD(5)})),
@@ -49,7 +49,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             ConvEltwiseFusion::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_GroupConvolutionBackpropData_1D, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolutionBackpropData_1D, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::GroupConvolutionBackpropData::type_info),
                                     ::testing::ValuesIn(IN({MUL(4), ADD(5)})),
@@ -68,7 +68,7 @@ namespace {
         {1, 1, 1, 1}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_Convolution_2D, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_Convolution_2D, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::Convolution::type_info),
                                     ::testing::ValuesIn(IN({MUL(4), ADD(5)})),
@@ -79,7 +79,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             ConvEltwiseFusion::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution_2D, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolution_2D, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::GroupConvolution::type_info),
                                     ::testing::ValuesIn(IN({MUL(4), ADD(5)})),
@@ -90,7 +90,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             ConvEltwiseFusion::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData_2D, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData_2D, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::ConvolutionBackpropData::type_info),
                                     ::testing::ValuesIn(IN({MUL(4), ADD(5)})),
@@ -101,7 +101,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             ConvEltwiseFusion::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_GroupConvolutionBackpropData_2D, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolutionBackpropData_2D, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::GroupConvolutionBackpropData::type_info),
                                     ::testing::ValuesIn(IN({MUL(4), ADD(5)})),
@@ -117,7 +117,7 @@ namespace {
         {3}, {3, 1}, {3, 1, 1, 1}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_Convolution_2D_Negative, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_Convolution_2D_Negative, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::Convolution::type_info),
                                     ::testing::ValuesIn(IN({MUL(6), ADD(6)})),
@@ -128,7 +128,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             ConvEltwiseFusion::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution_2D_Negative, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolution_2D_Negative, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::GroupConvolution::type_info),
                                     ::testing::ValuesIn(IN({MUL(6), ADD(6)})),
@@ -139,7 +139,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             ConvEltwiseFusion::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData_2D_Negative, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData_2D_Negative, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::ConvolutionBackpropData::type_info),
                                     ::testing::ValuesIn(IN({MUL(6), ADD(6)})),
@@ -150,7 +150,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                             ConvEltwiseFusion::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_GroupConvolutionBackpropData_2D_Negative, ConvEltwiseFusion,
+    INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolutionBackpropData_2D_Negative, ConvEltwiseFusion,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::opset4::GroupConvolutionBackpropData::type_info),
                                     ::testing::ValuesIn(IN({MUL(6), ADD(6)})),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/convert_pad_to_group_conv.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/convert_pad_to_group_conv.cpp
index 2ab868e8ca04ad..556eafd6aaf239 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/convert_pad_to_group_conv.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/convert_pad_to_group_conv.cpp
@@ -16,7 +16,7 @@ namespace {
 
     const std::vector<float> values{0., 1.};
 
-    INSTANTIATE_TEST_CASE_P(smoke_Pad_1D, ConvertPadToConvTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Pad_1D, ConvertPadToConvTests,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::Shape{1, 8, 64}),
                                     ::testing::ValuesIn(pads_1d),
@@ -31,7 +31,7 @@ namespace {
             {0, 0, 10, 10}, {0, 0, 0, 4}, {0, 0, 4, 0}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_Pad_2D, ConvertPadToConvTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Pad_2D, ConvertPadToConvTests,
                             ::testing::Combine(
                                     ::testing::Values(ngraph::Shape{1, 8, 64, 16}),
                                     ::testing::ValuesIn(pads_2d),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp
index 6f47cf9d7d9dff..337490706795b6 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp
@@ -23,7 +23,7 @@ namespace {
     std::map<std::string, std::string> additional_config = {};
 } // namespace
 
-INSTANTIATE_TEST_CASE_P(OutputBeforeActivation, OutputBeforeActivation,
+INSTANTIATE_TEST_SUITE_P(OutputBeforeActivation, OutputBeforeActivation,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(InferenceEngine::Precision::FP32),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp
index 9cab6afd9fb68d..ac3192476002cd 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp
@@ -36,7 +36,7 @@ std::vector<size_t> output_sizes = {
     8
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MatmulSqueezeAdd, MatmulSqueezeAddTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MatmulSqueezeAdd, MatmulSqueezeAddTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
index 8e9754343474cb..2d7af0c03d32e0 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
@@ -33,7 +33,7 @@ namespace SubgraphTestsDefinitions {
     std::map<std::string, std::string> additional_config = {
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_MemoryLSTMCellTest, MemoryLSTMCellTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_MemoryLSTMCellTest, MemoryLSTMCellTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(transformation),
                                     ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
index 4c18ee6ea054bb..f0f8e6c6895399 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
@@ -35,7 +35,7 @@ std::map<std::string, std::string> additional_config = {
 } // namespace
 
 
-INSTANTIATE_TEST_CASE_P(MultipleLSTMCellTest, MultipleLSTMCellTest,
+INSTANTIATE_TEST_SUITE_P(MultipleLSTMCellTest, MultipleLSTMCellTest,
     ::testing::Combine(
         ::testing::ValuesIn(transformation),
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiply_add.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiply_add.cpp
index 37df66d8b6bbbf..73145968aa718e 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiply_add.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/multiply_add.cpp
@@ -22,7 +22,7 @@ const std::vector<std::vector<size_t>> inputShapes = {
         {1, 3, 2, 2, 2, 2, 2, 2, 4, 5},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MultipleAdd_Nd, MultiplyAddLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MultipleAdd_Nd, MultiplyAddLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inputShapes),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/mvn_multiply_add.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/mvn_multiply_add.cpp
index 3150926b2a9ea9..58087839db20a9 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/mvn_multiply_add.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/mvn_multiply_add.cpp
@@ -55,7 +55,7 @@ const std::vector<std::pair<SizeVector, SizeVector>> shapes_3D = {
         std::pair<SizeVector, SizeVector>{ SizeVector{2, 3, 3}, SizeVector{2, 3, 3}},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MVNMultiplyAdd_1D, MVNMultiplyAdd,
+INSTANTIATE_TEST_SUITE_P(smoke_MVNMultiplyAdd_1D, MVNMultiplyAdd,
                         ::testing::Combine(
                                 ::testing::ValuesIn(shapes_1D),
                                 ::testing::ValuesIn(netPrecision),
@@ -67,7 +67,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVNMultiplyAdd_1D, MVNMultiplyAdd,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         MVNMultiplyAdd::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MVNMultiplyAdd_2D, MVNMultiplyAdd,
+INSTANTIATE_TEST_SUITE_P(smoke_MVNMultiplyAdd_2D, MVNMultiplyAdd,
                         ::testing::Combine(
                                 ::testing::ValuesIn(shapes_2D),
                                 ::testing::ValuesIn(netPrecision),
@@ -79,7 +79,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVNMultiplyAdd_2D, MVNMultiplyAdd,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         MVNMultiplyAdd::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MVNMultiplyAdd_3D, MVNMultiplyAdd,
+INSTANTIATE_TEST_SUITE_P(smoke_MVNMultiplyAdd_3D, MVNMultiplyAdd,
                         ::testing::Combine(
                                 ::testing::ValuesIn(shapes_3D),
                                 ::testing::ValuesIn(netPrecision),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/parameter_result.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/parameter_result.cpp
index 7eb8265150c599..3da2e776949a58 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/parameter_result.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/parameter_result.cpp
@@ -8,7 +8,7 @@
 using namespace SubgraphTestsDefinitions;
 
 namespace {
-    INSTANTIATE_TEST_CASE_P(smoke_Check, ParameterResultSubgraphTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Check, ParameterResultSubgraphTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU),
                             ParameterResultSubgraphTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp
index da0b67cd8c4f03..01192fdffd5386 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp
@@ -31,7 +31,7 @@ std::map<std::string, std::string> additional_config = {
 } // namespace
 
 namespace SubgraphTestsDefinitions {
-INSTANTIATE_TEST_CASE_P(smoke_basic, PermConvPermConcat,
+INSTANTIATE_TEST_SUITE_P(smoke_basic, PermConvPermConcat,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_convolution_backprop_data.cpp
index 0413002c0967e3..2d35a470247bd3 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_convolution_backprop_data.cpp
@@ -42,7 +42,7 @@ const auto quantConvBackpropData2DParams = ::testing::Combine(
         ::testing::ValuesIn(granularity)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantConvBackpropData2D, QuantConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantConvBackpropData2D, QuantConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 quantConvBackpropData2DParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -70,7 +70,7 @@ const auto quantConvBackpropData3DParams = ::testing::Combine(
         ::testing::ValuesIn(granularity)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantConvBackpropData3D, QuantConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantConvBackpropData3D, QuantConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 quantConvBackpropData3DParams,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_group_convolution.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_group_convolution.cpp
index 33c010dc36ca29..7627c3730ddf9b 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_group_convolution.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_group_convolution.cpp
@@ -46,7 +46,7 @@ const auto quantGroupConv2DParams = ::testing::Combine(
         ::testing::ValuesIn(quantizeWeights)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantGroupConv2D, QuantGroupConvLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantGroupConv2D, QuantGroupConvLayerTest,
                         ::testing::Combine(
                                 quantGroupConv2DParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -75,7 +75,7 @@ const auto quantGroupConv3DParams = ::testing::Combine(
         ::testing::ValuesIn(quantizeWeights)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantGroupConv3D, QuantGroupConvLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantGroupConv3D, QuantGroupConvLayerTest,
                         ::testing::Combine(
                                 quantGroupConv3DParams,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_group_convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_group_convolution_backprop_data.cpp
index bc64dd43041553..ef002cc8ee5e6e 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_group_convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_group_convolution_backprop_data.cpp
@@ -44,7 +44,7 @@ const auto quantGroupConvBackpropData2DParams = ::testing::Combine(
         ::testing::ValuesIn(granularity)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantGroupConvBackpropData2D, QuantGroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantGroupConvBackpropData2D, QuantGroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 quantGroupConvBackpropData2DParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -73,7 +73,7 @@ const auto quantGroupConvBackpropData3DParams = ::testing::Combine(
         ::testing::ValuesIn(granularity)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantGroupConvBackpropData3D, QuantGroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantGroupConvBackpropData3D, QuantGroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 quantGroupConvBackpropData3DParams,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_mat_mul.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_mat_mul.cpp
index 3af9da8f49c961..7914371a463d30 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_mat_mul.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/quantized_mat_mul.cpp
@@ -77,7 +77,7 @@ const auto quantParams_i32 = ::testing::Combine(
         ::testing::Values(InferenceEngine::Precision::I32)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantMatMul_i8i8, QuantMatMulTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantMatMul_i8i8, QuantMatMulTest,
                         ::testing::Combine(
                                 quantParams_i8,
                                 quantParams_i8,
@@ -87,7 +87,7 @@ INSTANTIATE_TEST_CASE_P(smoke_QuantMatMul_i8i8, QuantMatMulTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         QuantMatMulTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantMatMul_u8i8, QuantMatMulTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantMatMul_u8i8, QuantMatMulTest,
                         ::testing::Combine(
                                 quantParams_u8,
                                 quantParams_i8,
@@ -97,7 +97,7 @@ INSTANTIATE_TEST_CASE_P(smoke_QuantMatMul_u8i8, QuantMatMulTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         QuantMatMulTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantMatMul_i16i32, QuantMatMulTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantMatMul_i16i32, QuantMatMulTest,
                         ::testing::Combine(
                                 quantParams_i16,
                                 quantParams_i32,
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/range_add.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/range_add.cpp
index 9a53c864dd99bc..35e388f328ec9f 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/range_add.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/range_add.cpp
@@ -30,7 +30,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 
 // ------------------------------ V0 ------------------------------
 
-INSTANTIATE_TEST_CASE_P(smoke_BasicPositive, RangeAddSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BasicPositive, RangeAddSubgraphTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(positiveStart),
                                 ::testing::ValuesIn(positiveStop),
@@ -43,7 +43,7 @@ INSTANTIATE_TEST_CASE_P(smoke_BasicPositive, RangeAddSubgraphTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         RangeAddSubgraphTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BasicNegative, RangeAddSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BasicNegative, RangeAddSubgraphTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(negativeStart),
                                 ::testing::ValuesIn(negativeStop),
@@ -57,7 +57,7 @@ INSTANTIATE_TEST_CASE_P(smoke_BasicNegative, RangeAddSubgraphTest,
                         RangeAddSubgraphTest::getTestCaseName);
 
 // ------------------------------ V4 ------------------------------
-INSTANTIATE_TEST_CASE_P(smoke_BasicPositive, RangeNumpyAddSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BasicPositive, RangeNumpyAddSubgraphTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(positiveStart),
                                 ::testing::ValuesIn(positiveStop),
@@ -70,7 +70,7 @@ INSTANTIATE_TEST_CASE_P(smoke_BasicPositive, RangeNumpyAddSubgraphTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         RangeNumpyAddSubgraphTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BasicNegative, RangeNumpyAddSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BasicNegative, RangeNumpyAddSubgraphTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(negativeStart),
                                 ::testing::ValuesIn(negativeStop),
@@ -83,7 +83,7 @@ INSTANTIATE_TEST_CASE_P(smoke_BasicNegative, RangeNumpyAddSubgraphTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         RangeNumpyAddSubgraphTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BasicTruncateInputs, RangeNumpyAddSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BasicTruncateInputs, RangeNumpyAddSubgraphTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(trunc_start),
                                 ::testing::ValuesIn(trunc_stop),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/relu_shape_of.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/relu_shape_of.cpp
index 917cb6932c0cab..a1618c3d2fb01c 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/relu_shape_of.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/relu_shape_of.cpp
@@ -14,7 +14,7 @@ namespace {
             InferenceEngine::Precision::I32
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_Check, ReluShapeOfSubgraphTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Check, ReluShapeOfSubgraphTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(std::vector<size_t>({20, 10, 10, 10})),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
index 1e4666b523b5e0..8645cb95325671 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
@@ -38,7 +38,7 @@ namespace {
             ngraph::helpers::SqueezeOpType::UNSQUEEZE
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_reshape_squeeze_reshape_relu, ReshapeSqueezeReshapeRelu,
+    INSTANTIATE_TEST_SUITE_P(smoke_reshape_squeeze_reshape_relu, ReshapeSqueezeReshapeRelu,
                             ::testing::Combine(
                                     ::testing::ValuesIn(inputs),
                                     ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/split_concat_memory.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/split_concat_memory.cpp
index a5c93c9ac9ca51..8fdf19bcb5894f 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/split_concat_memory.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/split_concat_memory.cpp
@@ -26,7 +26,7 @@ const std::vector<InferenceEngine::SizeVector> shapes = {
     {3, 8},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_CPU, SplitConcatMemory,
+INSTANTIATE_TEST_SUITE_P(smoke_CPU, SplitConcatMemory,
                         ::testing::Combine(
                                 ::testing::ValuesIn(shapes),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/split_conv_concat.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
index 12dedadc265072..b156873f470a26 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
@@ -16,7 +16,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP16
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, SplitConvConcat,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, SplitConvConcat,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(InferenceEngine::SizeVector({1, 6, 40, 40})),
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/tensor_names.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/tensor_names.cpp
index e5d104920cb3f2..8e4e93ceb63d95 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/tensor_names.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/tensor_names.cpp
@@ -10,7 +10,7 @@
 using namespace SubgraphTestsDefinitions;
 
 namespace {
-    INSTANTIATE_TEST_CASE_P(smoke_Check, TensorNamesTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Check, TensorNamesTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU),
                             TensorNamesTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/variadic_split_pad.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/variadic_split_pad.cpp
index 3e73dbeccf8a1e..8f188589344462 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/variadic_split_pad.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/subgraph_tests/variadic_split_pad.cpp
@@ -48,7 +48,7 @@ const std::vector<ngraph::helpers::PadMode> padMode = {
         ngraph::helpers::PadMode::SYMMETRIC
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_CPU, VariadicSplitPad,
+INSTANTIATE_TEST_SUITE_P(smoke_CPU, VariadicSplitPad,
                         ::testing::Combine(
                                 ::testing::ValuesIn(shapes),
                                 ::testing::Values(1),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/activation.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/activation.cpp
index d473bbde6e2c62..f2e5a76e33c483 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/activation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/activation.cpp
@@ -128,7 +128,7 @@ const auto basicCases4D = ::testing::Combine(
         ::testing::ValuesIn(filterCPUSpecificParams(cpuParams_4D))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation4D_Eltwise_CPU_BF16, ActivationLayerCPUTest, basicCases4D, ActivationLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation4D_Eltwise_CPU_BF16, ActivationLayerCPUTest, basicCases4D, ActivationLayerCPUTest::getTestCaseName);
 
 std::vector<CPUSpecificParams> cpuParams_5D = {
         CPUSpecificParams({nCdhw16c}, {nCdhw16c}, {}, {}),
@@ -154,6 +154,6 @@ const auto basicCases5D = ::testing::Combine(
         ::testing::ValuesIn(filterCPUSpecificParams(cpuParams_5D))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation5D_Eltwise_CPU_BF16, ActivationLayerCPUTest, basicCases5D, ActivationLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation5D_Eltwise_CPU_BF16, ActivationLayerCPUTest, basicCases5D, ActivationLayerCPUTest::getTestCaseName);
 } // namespace
 } // namespace CPULayerTestsDefinitions
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/batch_to_space.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/batch_to_space.cpp
index 82c97a51166ca6..a1c583adfcc1b2 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/batch_to_space.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/batch_to_space.cpp
@@ -120,10 +120,10 @@ const auto batchToSpaceParamsSet4D2 = ::testing::Combine(
                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(cpuParams_4D));
 
-INSTANTIATE_TEST_CASE_P(smoke_BatchToSpaceCPULayerTestCase1_4D, BatchToSpaceCPULayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BatchToSpaceCPULayerTestCase1_4D, BatchToSpaceCPULayerTest,
                             batchToSpaceParamsSet4D1, BatchToSpaceCPULayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BatchToSpaceCPULayerTestCase2_4D, BatchToSpaceCPULayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BatchToSpaceCPULayerTestCase2_4D, BatchToSpaceCPULayerTest,
                             batchToSpaceParamsSet4D2, BatchToSpaceCPULayerTest::getTestCaseName);
 
 const std::vector<std::vector<int64_t>> blockShape5D1  = {{1, 1, 2, 2, 1}, {1, 2, 1, 2, 2}};
@@ -169,10 +169,10 @@ const auto batchToSpaceParamsSet5D2 = ::testing::Combine(
                 ::testing::Values(InferenceEngine::Layout::ANY),
                 ::testing::Values(CommonTestUtils::DEVICE_CPU)), ::testing::ValuesIn(cpuParams_5D));
 
-INSTANTIATE_TEST_CASE_P(smoke_BatchToSpaceCPULayerTestCase1_5D, BatchToSpaceCPULayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BatchToSpaceCPULayerTestCase1_5D, BatchToSpaceCPULayerTest,
                             batchToSpaceParamsSet5D1, BatchToSpaceCPULayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BatchToSpaceCPULayerTestCase2_5D, BatchToSpaceCPULayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BatchToSpaceCPULayerTestCase2_5D, BatchToSpaceCPULayerTest,
                             batchToSpaceParamsSet5D2, BatchToSpaceCPULayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/concat.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/concat.cpp
index 32b7b6d60910d0..11c20c7cd01fcd 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/concat.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/concat.cpp
@@ -96,7 +96,7 @@ const std::vector<Precision> netPrecisions = {
         Precision::BF16
 };
 
-INSTANTIATE_TEST_CASE_P(concat_Concat4D_CPU_Block8inPlace, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(concat_Concat4D_CPU_Block8inPlace, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(1),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{1, 8,  3, 5},
@@ -106,7 +106,7 @@ INSTANTIATE_TEST_CASE_P(concat_Concat4D_CPU_Block8inPlace, ConcatLayerCPUTest,
                                 ::testing::Values(planar_4D, planarChannels_4D, blocked8_4D)),
                         ConcatLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Concat4D_CPU_Block8, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Concat4D_CPU_Block8, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(0, 2, 3),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{2, 16, 3, 5},
@@ -116,7 +116,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Concat4D_CPU_Block8, ConcatLayerCPUTest,
                                 ::testing::Values(planar_4D_ref, planarChannels_4D, blocked8_4D_ref)),
                         ConcatLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Concat4D_CPU_Block16inPlace, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Concat4D_CPU_Block16inPlace, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(1),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{2, 16, 3, 5},
@@ -126,7 +126,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Concat4D_CPU_Block16inPlace, ConcatLayerCPUTest,
                                 ::testing::Values(blocked16_4D)),
                         ConcatLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Concat4D_CPU_Block16, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Concat4D_CPU_Block16, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(0, 2, 3),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{2, 32, 3, 5},
@@ -136,7 +136,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Concat4D_CPU_Block16, ConcatLayerCPUTest,
                                 ::testing::Values(blocked16_4D_ref)),
                         ConcatLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(concat_Concat5D_CPU_Block8inPlace, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(concat_Concat5D_CPU_Block8inPlace, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(1),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{1, 8,  3, 5, 7},
@@ -146,7 +146,7 @@ INSTANTIATE_TEST_CASE_P(concat_Concat5D_CPU_Block8inPlace, ConcatLayerCPUTest,
                                 ::testing::Values(planar_5D, planarChannels_5D, blocked8_5D)),
                         ConcatLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Concat5D_CPU_Block8, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Concat5D_CPU_Block8, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(0, 2, 3, 4),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{2, 16, 3, 5, 7},
@@ -156,7 +156,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Concat5D_CPU_Block8, ConcatLayerCPUTest,
                                 ::testing::Values(planar_5D_ref, planarChannels_5D, blocked8_5D_ref)),
                         ConcatLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Concat5D_CPU_Block16inPlace, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Concat5D_CPU_Block16inPlace, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(1),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{2, 16, 3, 5, 7},
@@ -166,7 +166,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Concat5D_CPU_Block16inPlace, ConcatLayerCPUTest,
                                 ::testing::Values(blocked16_5D)),
                         ConcatLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Concat5D_CPU_Block16, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Concat5D_CPU_Block16, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(0, 2, 3, 4),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{2, 32, 3, 5, 7},
@@ -177,7 +177,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Concat5D_CPU_Block16, ConcatLayerCPUTest,
                         ConcatLayerCPUTest::getTestCaseName);
 
 
-INSTANTIATE_TEST_CASE_P(smoke_Concat_inPlace, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Concat_inPlace, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(1),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{2, 3, 5},
@@ -189,7 +189,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Concat_inPlace, ConcatLayerCPUTest,
                                 ::testing::Values(CPUSpecificParams{{}, {}, {}, "unknown"})),
                         ConcatLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Concat3D, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Concat3D, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(0, 2),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{2, 4, 5},
@@ -199,7 +199,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Concat3D, ConcatLayerCPUTest,
                                 ::testing::Values(CPUSpecificParams{{}, {}, {}, "ref"})),
                         ConcatLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Concat_1D_2D, ConcatLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Concat_1D_2D, ConcatLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(0),
                                 ::testing::Values(std::vector<std::vector<size_t>>{{2, 4},
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convert.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convert.cpp
index e0ef6bfee15ce6..3a8d3a920e20c3 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convert.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convert.cpp
@@ -33,7 +33,7 @@ const std::vector<Precision> precisions = {
         Precision::BF16
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvertLayerTest_From_BF16, ConvertCPULayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvertLayerTest_From_BF16, ConvertCPULayerTest,
                         ::testing::Combine(
                                 ::testing::Values(inShape),
                                 ::testing::Values(Precision::BF16),
@@ -43,7 +43,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvertLayerTest_From_BF16, ConvertCPULayerTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvertLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvertLayerTest_To_BF16, ConvertCPULayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvertLayerTest_To_BF16, ConvertCPULayerTest,
                         ::testing::Combine(
                                 ::testing::Values(inShape),
                                 ::testing::ValuesIn(precisions),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convert_to_plugin_specific_node.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convert_to_plugin_specific_node.cpp
index a59b7f6fc3b5ce..204cf031f02783 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convert_to_plugin_specific_node.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convert_to_plugin_specific_node.cpp
@@ -104,7 +104,7 @@ const auto testParamsEltwise = ::testing::Combine(::testing::ValuesIn(nonConstIS
                                                   ::testing::ValuesIn(port),
                                                   ::testing::Values(0));
 
-INSTANTIATE_TEST_CASE_P(smoke_CheckEltwise, ConvertToPluginSpecificNode, testParamsEltwise, ConvertToPluginSpecificNode::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CheckEltwise, ConvertToPluginSpecificNode, testParamsEltwise, ConvertToPluginSpecificNode::getTestCaseName);
 
 const auto testParamsPower = ::testing::Combine(::testing::ValuesIn(nonConstIS),
                                                 ::testing::ValuesIn(constIS),
@@ -113,7 +113,7 @@ const auto testParamsPower = ::testing::Combine(::testing::ValuesIn(nonConstIS),
                                                 ::testing::Values(1),
                                                 ::testing::Values(0));
 
-INSTANTIATE_TEST_CASE_P(smoke_CheckPower, ConvertToPluginSpecificNode, testParamsPower, ConvertToPluginSpecificNode::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CheckPower, ConvertToPluginSpecificNode, testParamsPower, ConvertToPluginSpecificNode::getTestCaseName);
 
 } // namespace
 
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution.cpp
index 9be9a270ec701e..b8d447d5a2737f 100755
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution.cpp
@@ -220,7 +220,7 @@ const std::vector<CPUSpecificParams> CPUParams_GEMM_2D = {
         conv_gemm_2D_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_GEMM_FP32, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_2D_GEMM_FP32, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_GEMM_2D,
@@ -236,7 +236,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_GEMM_FP32, ConvolutionLayerCPUTest,
         ::testing::Values(cpuEmptyPluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_GEMM_BF16, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_2D_GEMM_BF16, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_GEMM_2D,
@@ -252,7 +252,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_GEMM_BF16, ConvolutionLayerCPUTest,
         ::testing::Values(cpuBF16PluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_GEMM_I8, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_2D_GEMM_I8, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_GEMM_2D,
@@ -284,7 +284,7 @@ const std::vector<CPUSpecificParams> CPUParams_GEMM_3D = {
         conv_gemm_3D_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_3D_GEMM_FP32, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_3D_GEMM_FP32, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_GEMM_3D,
@@ -300,7 +300,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_3D_GEMM_FP32, ConvolutionLayerCPUTest,
         ::testing::Values(cpuEmptyPluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_3D_GEMM_BF16, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_3D_GEMM_BF16, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_GEMM_3D,
@@ -316,7 +316,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_3D_GEMM_BF16, ConvolutionLayerCPUTest,
         ::testing::Values(cpuBF16PluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_3D_GEMM_I8, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_3D_GEMM_I8, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_GEMM_3D,
@@ -352,7 +352,7 @@ const std::vector<CPUSpecificParams> CPUParams_2D = {
         conv_avx512_2D_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_FP32, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_2D_FP32, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_2D,
@@ -368,7 +368,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_FP32, ConvolutionLayerCPUTest,
         ::testing::Values(cpuEmptyPluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_BF16, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_2D_BF16, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_2D,
@@ -384,7 +384,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_BF16, ConvolutionLayerCPUTest,
         ::testing::Values(cpuBF16PluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_I8, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_2D_I8, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_2D,
@@ -406,7 +406,7 @@ const std::vector<CPUSpecificParams> CPUParams_2D_plain_to_blocked = {
         conv_avx512_plain_to_blocked_2D,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_PlainToBlocked_2D_FP32, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_PlainToBlocked_2D_FP32, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_2D,
@@ -422,7 +422,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_PlainToBlocked_2D_FP32, ConvolutionLayerCPUTe
         ::testing::Values(cpuEmptyPluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_PlainToBlocked_2D_BF16, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_PlainToBlocked_2D_BF16, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_2D,
@@ -457,7 +457,7 @@ const std::vector<CPUSpecificParams> CPUParams_3D = {
             conv_avx512_3D_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_3D_FP32, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_3D_FP32, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_3D,
@@ -473,7 +473,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_3D_FP32, ConvolutionLayerCPUTest,
         ::testing::Values(cpuEmptyPluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_3D_BF16, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_3D_BF16, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_3D,
@@ -489,7 +489,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_3D_BF16, ConvolutionLayerCPUTest,
         ::testing::Values(cpuBF16PluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_3D_I8, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_3D_I8, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_3D,
@@ -510,7 +510,7 @@ const std::vector<CPUSpecificParams> CPUParams_3D_plain_to_blocked = {
         conv_avx512_plain_to_blocked_3D,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_PlainToBlocked_3D_FP32, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_PlainToBlocked_3D_FP32, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_3D,
@@ -526,7 +526,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_PlainToBlocked_3D_FP32, ConvolutionLayerCPUTe
         ::testing::Values(cpuEmptyPluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_PlainToBlocked_3D_BF16, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_PlainToBlocked_3D_BF16, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_3D,
@@ -563,7 +563,7 @@ const std::vector<CPUSpecificParams> CPUParams_1x1_2D = {
     conv_avx512_2D_1x1_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_1x1_FP32, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_2D_1x1_FP32, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_1x1_2D,
@@ -579,7 +579,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_1x1_FP32, ConvolutionLayerCPUTest,
         ::testing::Values(cpuEmptyPluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_1x1_BF16, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_2D_1x1_BF16, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_1x1_2D,
@@ -595,7 +595,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_1x1_BF16, ConvolutionLayerCPUTest,
         ::testing::Values(cpuBF16PluginConfig)),
     ConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_2D_1x1_I8, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_2D_1x1_I8, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_1x1_2D,
@@ -635,7 +635,7 @@ const std::vector<CPUSpecificParams> CPUParams_1D = {
         conv_avx512_1D
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_1D, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_1D, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_1D,
@@ -670,7 +670,7 @@ const auto convParams_Planar_ExplicitPadding_2D = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_Jit_Planar_2D_FP32, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_Jit_Planar_2D_FP32, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_Planar_ExplicitPadding_2D,
@@ -703,7 +703,7 @@ const auto convParams_Planar_ExplicitPadding_3D = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_Jit_Planar_3D_FP32, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_Jit_Planar_3D_FP32, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_Planar_ExplicitPadding_3D,
@@ -752,7 +752,7 @@ const auto convParams_2D = ::testing::Combine(
     ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv_winograd, ConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Conv_winograd, ConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_2D,
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution_backprop_data.cpp
index f1d144f666ebf9..69544f12e1f665 100755
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/convolution_backprop_data.cpp
@@ -152,7 +152,7 @@ const auto convParams_ExplicitPadding_Planar_2D = ::testing::Combine(
     ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Planar_FP32, DeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_2D_Planar_FP32, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_Planar_2D,
@@ -169,7 +169,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Planar_FP32, DeconvolutionLayerCPUTest,
         ::testing::Values(cpuEmptyPluginConfig)),
     DeconvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Planar_BF16, DeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_2D_Planar_BF16, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_Planar_2D,
@@ -198,7 +198,7 @@ const auto convParams_ExplicitPadding_Planar_3D = ::testing::Combine(
     ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Planar_FP32, DeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_3D_Planar_FP32, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_Planar_3D,
@@ -215,7 +215,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Planar_FP32, DeconvolutionLayerCPUTest,
         ::testing::Values(cpuEmptyPluginConfig)),
     DeconvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Planar_BF16, DeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_3D_Planar_BF16, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_Planar_3D,
@@ -244,7 +244,7 @@ const auto convParams_ExplicitPadding_Blocked_2D = ::testing::Combine(
     ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Blocked_FP32, DeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_2D_Blocked_FP32, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_Blocked_2D,
@@ -261,7 +261,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Blocked_FP32, DeconvolutionLayerCPUTest,
         ::testing::Values(cpuEmptyPluginConfig)),
     DeconvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_Blocked_BF16, DeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_2D_Blocked_BF16, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_Blocked_2D,
@@ -290,7 +290,7 @@ const auto convParams_ExplicitPadding_Blocked_3D = ::testing::Combine(
     ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Blocked_FP32, DeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_3D_Blocked_FP32, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_Blocked_3D,
@@ -307,7 +307,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Blocked_FP32, DeconvolutionLayerCPUTest,
         ::testing::Values(cpuEmptyPluginConfig)),
     DeconvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Deconv_3D_Blocked_BF16, DeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_3D_Blocked_BF16, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_Blocked_3D,
@@ -337,7 +337,7 @@ const auto convParams_ExplicitPadding_1x1_2D = ::testing::Combine(
         ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_1x1_FP32, DeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_2D_1x1_FP32, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_1x1_2D,
@@ -354,7 +354,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_1x1_FP32, DeconvolutionLayerCPUTest,
         ::testing::Values(cpuEmptyPluginConfig)),
     DeconvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Deconv_2D_1x1_BF16, DeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_2D_1x1_BF16, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_ExplicitPadding_1x1_2D,
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/depth_to_space.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/depth_to_space.cpp
index 8588d9d2658632..a5e54532001ca5 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/depth_to_space.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/depth_to_space.cpp
@@ -124,7 +124,7 @@ const auto depthToSpaceBS2_4DParams = testing::Combine(
         testing::ValuesIn(filterCPUInfoForDevice(CPUParamsBS2_4D))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_CPUDepthToSpaceBS2_4D, DepthToSpaceLayerCPUTest, depthToSpaceBS2_4DParams, DepthToSpaceLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CPUDepthToSpaceBS2_4D, DepthToSpaceLayerCPUTest, depthToSpaceBS2_4DParams, DepthToSpaceLayerCPUTest::getTestCaseName);
 
 const std::vector<CPUSpecificParams> CPUParamsBS3_4D = {
         cpuParams_nhwc_avx2,
@@ -142,7 +142,7 @@ const auto depthToSpaceBS3_4DParams = testing::Combine(
         ::testing::ValuesIn(filterCPUInfoForDevice(CPUParamsBS3_4D))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_CPUDepthToSpaceBS3_4D, DepthToSpaceLayerCPUTest, depthToSpaceBS3_4DParams, DepthToSpaceLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CPUDepthToSpaceBS3_4D, DepthToSpaceLayerCPUTest, depthToSpaceBS3_4DParams, DepthToSpaceLayerCPUTest::getTestCaseName);
 
 const std::vector<std::vector<size_t >> inputShapesBS2_5D = {
         {1, 128, 1, 1, 1}, {1, 128, 2, 1, 2}, {1, 256, 2, 1, 3}, {2, 256, 3, 1, 1}, {1, 384, 1, 2, 2}, {2, 512, 1, 2, 1}
@@ -171,7 +171,7 @@ const auto depthToSpaceBS2_5DParams = testing::Combine(
         ::testing::ValuesIn(filterCPUInfoForDevice(CPUParamsBS2_5D))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_CPUDepthToSpaceBS2_5D, DepthToSpaceLayerCPUTest, depthToSpaceBS2_5DParams, DepthToSpaceLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CPUDepthToSpaceBS2_5D, DepthToSpaceLayerCPUTest, depthToSpaceBS2_5DParams, DepthToSpaceLayerCPUTest::getTestCaseName);
 
 const std::vector<CPUSpecificParams> CPUParamsBS3_5D = {
         cpuParams_ndhwc_avx2,
@@ -189,7 +189,7 @@ const auto depthToSpaceBS3_5DParams = testing::Combine(
         ::testing::ValuesIn(filterCPUInfoForDevice(CPUParamsBS3_5D))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_CPUDepthToSpaceBS3_5D, DepthToSpaceLayerCPUTest, depthToSpaceBS3_5DParams, DepthToSpaceLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CPUDepthToSpaceBS3_5D, DepthToSpaceLayerCPUTest, depthToSpaceBS3_5DParams, DepthToSpaceLayerCPUTest::getTestCaseName);
 
 } // namespace
 } // namespace CPULayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/eltwise.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/eltwise.cpp
index 0deb0efc300965..3932b9349bb212 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/eltwise.cpp
@@ -167,7 +167,7 @@ const auto params_4D = ::testing::Combine(
             ::testing::Values(additional_config)),
         ::testing::ValuesIn(filterCPUSpecificParams(cpuParams_4D)));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_4D_MemOrder, EltwiseLayerCPUTest, params_4D, EltwiseLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_4D_MemOrder, EltwiseLayerCPUTest, params_4D, EltwiseLayerCPUTest::getTestCaseName);
 
 const auto params_4D_emptyCPUSpec = ::testing::Combine(
         ::testing::Combine(
@@ -183,7 +183,7 @@ const auto params_4D_emptyCPUSpec = ::testing::Combine(
                 ::testing::Values(additional_config)),
         ::testing::Values(emptyCPUSpec));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_4D_emptyCPUSpec, EltwiseLayerCPUTest, params_4D_emptyCPUSpec, EltwiseLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_4D_emptyCPUSpec, EltwiseLayerCPUTest, params_4D_emptyCPUSpec, EltwiseLayerCPUTest::getTestCaseName);
 
 std::vector<std::vector<std::vector<size_t>>> inShapes_5D = {
         {{2, 4, 3, 4, 1}},
@@ -212,7 +212,7 @@ const auto params_5D = ::testing::Combine(
             ::testing::Values(additional_config)),
         ::testing::ValuesIn(filterCPUSpecificParams(cpuParams_5D)));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_5D_MemOrder, EltwiseLayerCPUTest, params_5D, EltwiseLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_5D_MemOrder, EltwiseLayerCPUTest, params_5D, EltwiseLayerCPUTest::getTestCaseName);
 
 const auto params_5D_emptyCPUSpec = ::testing::Combine(
         ::testing::Combine(
@@ -228,7 +228,7 @@ const auto params_5D_emptyCPUSpec = ::testing::Combine(
                 ::testing::Values(additional_config)),
         ::testing::Values(emptyCPUSpec));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_5D, EltwiseLayerCPUTest, params_5D_emptyCPUSpec, EltwiseLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_5D, EltwiseLayerCPUTest, params_5D_emptyCPUSpec, EltwiseLayerCPUTest::getTestCaseName);
 
 std::vector<std::vector<std::vector<size_t>>> inShapes_4D_Blocked_Planar = {
         {{2, 17, 31, 3}, {2, 1, 31, 3}},
@@ -253,7 +253,7 @@ const auto params_4D_Blocked_Planar = ::testing::Combine(
             ::testing::Values(additional_config)),
         ::testing::ValuesIn(filterCPUSpecificParams(cpuParams_4D_Blocked_Planar)));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_4D_Blocked_Planar, EltwiseLayerCPUTest, params_4D_Blocked_Planar, EltwiseLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_4D_Blocked_Planar, EltwiseLayerCPUTest, params_4D_Blocked_Planar, EltwiseLayerCPUTest::getTestCaseName);
 
 
 std::vector<std::vector<std::vector<size_t>>> inShapes_4D_Planar_Blocked = {
@@ -279,7 +279,7 @@ const auto params_4D_Planar_Blocked = ::testing::Combine(
             ::testing::Values(additional_config)),
         ::testing::ValuesIn(filterCPUSpecificParams(cpuParams_4D_Planar_Blocked)));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_4D_Planar_Blocked, EltwiseLayerCPUTest, params_4D_Planar_Blocked, EltwiseLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_4D_Planar_Blocked, EltwiseLayerCPUTest, params_4D_Planar_Blocked, EltwiseLayerCPUTest::getTestCaseName);
 
 
 std::vector<std::vector<std::vector<size_t>>> inShapes_5D_Blocked_Planar = {
@@ -305,7 +305,7 @@ const auto params_5D_Blocked_Planar = ::testing::Combine(
             ::testing::Values(additional_config)),
         ::testing::ValuesIn(filterCPUSpecificParams(cpuParams_5D_Blocked_Planar)));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_5D_Blocked_Planar, EltwiseLayerCPUTest, params_5D_Blocked_Planar, EltwiseLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_5D_Blocked_Planar, EltwiseLayerCPUTest, params_5D_Blocked_Planar, EltwiseLayerCPUTest::getTestCaseName);
 
 
 std::vector<std::vector<std::vector<size_t>>> inShapes_5D_Planar_Blocked = {
@@ -331,7 +331,7 @@ const auto params_5D_Planar_Blocked = ::testing::Combine(
                 ::testing::Values(additional_config)),
         ::testing::ValuesIn(filterCPUSpecificParams(cpuParams_5D_Planar_Blocked)));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_5D_Planar_Blocked, EltwiseLayerCPUTest, params_5D_Planar_Blocked, EltwiseLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_5D_Planar_Blocked, EltwiseLayerCPUTest, params_5D_Planar_Blocked, EltwiseLayerCPUTest::getTestCaseName);
 
 
 std::vector<std::vector<std::vector<size_t>>> inShapes_4D_1D = {
@@ -359,7 +359,7 @@ const auto params_4D_1D = ::testing::Combine(
                 ::testing::Values(additional_config)),
         ::testing::ValuesIn(filterCPUSpecificParams(cpuParams_4D_1D)));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_4D_1D, EltwiseLayerCPUTest, params_4D_1D, EltwiseLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_4D_1D, EltwiseLayerCPUTest, params_4D_1D, EltwiseLayerCPUTest::getTestCaseName);
 
 std::vector<std::vector<std::vector<size_t>>> inShapes_5D_1D = {
         {{2, 17, 5, 4, 10}, {10}},
@@ -386,7 +386,7 @@ const auto params_5D_1D = ::testing::Combine(
                 ::testing::Values(additional_config)),
         ::testing::ValuesIn(filterCPUSpecificParams(cpuParams_5D_1D)));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_5D_1D, EltwiseLayerCPUTest, params_5D_1D, EltwiseLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_5D_1D, EltwiseLayerCPUTest, params_5D_1D, EltwiseLayerCPUTest::getTestCaseName);
 
 
 } // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/extract_image_patches.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/extract_image_patches.cpp
index 028cd9118d8c3d..20edee2aad70d4 100755
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/extract_image_patches.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/extract_image_patches.cpp
@@ -84,7 +84,7 @@ const auto Layer_params = ::testing::Combine(
         ::testing::Values(InferenceEngine::Layout::ANY),
         ::testing::Values(CommonTestUtils::DEVICE_CPU));
 
-INSTANTIATE_TEST_CASE_P(smoke_ExtractImagePatches_CPU, ExtractImagePatchesLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ExtractImagePatches_CPU, ExtractImagePatchesLayerCPUTest,
                         ::testing::Combine(Layer_params, ::testing::Values(CPUParams)),
                         ExtractImagePatchesLayerCPUTest::getTestCaseName);
 
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/fake_quantize.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/fake_quantize.cpp
index a5832b40835d49..09bc5f27a62195 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/fake_quantize.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/fake_quantize.cpp
@@ -174,7 +174,7 @@ const auto testParams4D_jit = ::testing::Combine(specificParams4D_jit,
                                                  ::testing::ValuesIn(input_ranges),
                                                  ::testing::Values(false),
                                                  ::testing::ValuesIn(filterCPUSpecificParams(memForm4D_jit)));
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantizeLayerCPUTest_4D_jit, FakeQuantizeLayerCPUTest, testParams4D_jit, FakeQuantizeLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantizeLayerCPUTest_4D_jit, FakeQuantizeLayerCPUTest, testParams4D_jit, FakeQuantizeLayerCPUTest::getTestCaseName);
 
 
 std::vector<CPUSpecificParams> memForm4D_ref = {
@@ -197,7 +197,7 @@ const auto testParams4D_ref = ::testing::Combine(specificParams4D_ref,
                                                  ::testing::ValuesIn(input_ranges),
                                                  ::testing::Values(false),
                                                  ::testing::ValuesIn(memForm4D_ref));
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantizeLayerCPUTest_4D_ref, FakeQuantizeLayerCPUTest, testParams4D_ref, FakeQuantizeLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantizeLayerCPUTest_4D_ref, FakeQuantizeLayerCPUTest, testParams4D_ref, FakeQuantizeLayerCPUTest::getTestCaseName);
 
 
 std::vector<CPUSpecificParams> memForm5D_jit = {
@@ -224,7 +224,7 @@ const auto testParams5D_jit = ::testing::Combine(specificParams5D_jit,
                                                  ::testing::Values(false),
                                                  ::testing::ValuesIn(filterCPUSpecificParams(memForm5D_jit)));
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantizeLayerCPUTest_5D_jit, FakeQuantizeLayerCPUTest, testParams5D_jit, FakeQuantizeLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantizeLayerCPUTest_5D_jit, FakeQuantizeLayerCPUTest, testParams5D_jit, FakeQuantizeLayerCPUTest::getTestCaseName);
 
 
 std::vector<CPUSpecificParams> memForm5D_ref = {
@@ -248,7 +248,7 @@ const auto testParams5D_ref = ::testing::Combine(specificParams5D_ref,
                                                  ::testing::Values(false),
                                                  ::testing::ValuesIn(memForm5D_ref));
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantizeLayerCPUTest_5D_ref, FakeQuantizeLayerCPUTest, testParams5D_ref, FakeQuantizeLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantizeLayerCPUTest_5D_ref, FakeQuantizeLayerCPUTest, testParams5D_ref, FakeQuantizeLayerCPUTest::getTestCaseName);
 
 } // namespace fqImpl
 
@@ -281,7 +281,7 @@ const auto testParams = ::testing::Combine(specificParams,
                                            ::testing::Values(true),
                                            ::testing::Values(CPUSpecificParams{}));
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantizeLayerCPUTest_Decompos, FakeQuantizeLayerCPUTest, testParams, FakeQuantizeLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantizeLayerCPUTest_Decompos, FakeQuantizeLayerCPUTest, testParams, FakeQuantizeLayerCPUTest::getTestCaseName);
 
 } // namespace fqDecompos
 
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gather_elements.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gather_elements.cpp
index 7ea6cb505bac1c..32b6b04ad69072 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gather_elements.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gather_elements.cpp
@@ -76,7 +76,7 @@ std::vector<CPUSpecificParams> cpuParams_4D = {
         CPUSpecificParams({nchw}, {nchw}, {}, {})
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_set1, GatherElementsCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_set1, GatherElementsCPUTest,
             ::testing::Combine(
                 ::testing::Combine(
                     ::testing::Values(std::vector<size_t>({2, 3, 5, 7})),     // Data shape
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/group_convolution.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/group_convolution.cpp
index 5e18a0f3ba099a..1f75554b1b49ea 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/group_convolution.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/group_convolution.cpp
@@ -218,7 +218,7 @@ const std::vector<CPUSpecificParams> CPUParams_Gemm_2D = {
         conv_gemm_2D_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_2D_Gemm_FP32, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_2D_Gemm_FP32, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_Gemm_2D,
@@ -233,7 +233,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConv_2D_Gemm_FP32, GroupConvolutionLayerCPUTe
                                 ::testing::ValuesIn(fusingParamsSet)),
                         GroupConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_2D_Gemm_BF16, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_2D_Gemm_BF16, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_Gemm_2D,
@@ -265,7 +265,7 @@ const std::vector<CPUSpecificParams> CPUParams_Gemm_3D = {
         conv_gemm_3D_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_3D_Gemm_FP32, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_3D_Gemm_FP32, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_Gemm_3D,
@@ -280,7 +280,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConv_3D_Gemm_FP32, GroupConvolutionLayerCPUTe
                                 ::testing::ValuesIn(fusingParamsSet)),
                         GroupConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_3D_Gemm_BF16, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_3D_Gemm_BF16, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_Gemm_3D,
@@ -316,7 +316,7 @@ const std::vector<CPUSpecificParams> CPUParams_2D = {
         conv_avx512_2D_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_2D_FP32, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_2D_FP32, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_2D,
@@ -331,7 +331,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConv_2D_FP32, GroupConvolutionLayerCPUTest,
                                 ::testing::ValuesIn(fusingParamsSet)),
                         GroupConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_2D_BF16, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_2D_BF16, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_2D,
@@ -366,7 +366,7 @@ const std::vector<CPUSpecificParams> CPUParams_3D = {
         conv_avx512_3D_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_3D_FP32, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_3D_FP32, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_3D,
@@ -381,7 +381,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConv_3D_FP32, GroupConvolutionLayerCPUTest,
                                 ::testing::ValuesIn(fusingParamsSet)),
                         GroupConvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_3D_BF16, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_3D_BF16, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_3D,
@@ -417,7 +417,7 @@ const std::vector<CPUSpecificParams> CPUParams_DW_2D = {
         conv_avx512_dw_2D_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_2D_DW_FP32, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_2D_DW_FP32, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_DW_2D,
@@ -433,7 +433,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConv_2D_DW_FP32, GroupConvolutionLayerCPUTest
                         GroupConvolutionLayerCPUTest::getTestCaseName);
 
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_2D_DW_BF16, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_2D_DW_BF16, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_DW_2D,
@@ -469,7 +469,7 @@ const std::vector<CPUSpecificParams> CPUParams_DW_3D = {
         conv_avx512_dw_3D_nspc
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_3D_DW_FP32, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_3D_DW_FP32, GroupConvolutionLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         groupConvParams_ExplicitPadding_DW_3D,
@@ -586,7 +586,7 @@ const std::vector<groupConvLayerCPUTestParamsSet> gemmGroupConvTestCases = gener
                                         2, 1, {10, 10, 10}, 3, 3, CPUParams_Gemm_3D, vecPrcConnectParams)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GEMM_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice(gemmGroupConvTestCases)),
+INSTANTIATE_TEST_SUITE_P(smoke_GEMM_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice(gemmGroupConvTestCases)),
                         GroupConvolutionLayerCPUTest::getTestCaseName);
 
 /* ============= JIT SSE42 GroupConvolution ============= */
@@ -637,7 +637,7 @@ const std::vector<groupConvLayerCPUTestParamsSet> JIT_SSE42_GroupConvTestCases =
         //                              2, 1, {10, 10, 10}, 8, 8, cpuParams_sse42_3D),
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_JIT_SSE42_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice(JIT_SSE42_GroupConvTestCases)),
+INSTANTIATE_TEST_SUITE_P(smoke_JIT_SSE42_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice(JIT_SSE42_GroupConvTestCases)),
                         GroupConvolutionLayerCPUTest::getTestCaseName);
 
 /* ============= JIT AVX2 GroupConvolution ============= */
@@ -687,7 +687,7 @@ const std::vector<groupConvLayerCPUTestParamsSet> JIT_AVX2_GroupConvTestCases =
                                         2, 1, {10, 10, 10}, 8, 8, avx2_GroupConv_3D, vecPrcConnectParamsFP32)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_JIT_AVX2_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice(JIT_AVX2_GroupConvTestCases)),
+INSTANTIATE_TEST_SUITE_P(smoke_JIT_AVX2_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice(JIT_AVX2_GroupConvTestCases)),
                         GroupConvolutionLayerCPUTest::getTestCaseName);
 
 /* ============= JIT AVX512 GroupConvolution ============= */
@@ -718,7 +718,7 @@ const std::vector<groupConvLayerCPUTestParamsSet> JIT_AVX512_GroupConvTestCases
                                         2, 1, {10, 10, 10}, 16, 16, avx512_GroupConv_3D, vecPrcConnectParams)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_JIT_AVX512_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice(JIT_AVX512_GroupConvTestCases)),
+INSTANTIATE_TEST_SUITE_P(smoke_JIT_AVX512_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice(JIT_AVX512_GroupConvTestCases)),
                         GroupConvolutionLayerCPUTest::getTestCaseName);
 
 /* ============= JIT SSE42 DW GroupConvolution ============= */
@@ -754,7 +754,7 @@ const std::vector<groupConvLayerCPUTestParamsSet> JIT_SSE42_DW_GroupConvTestCase
                                         8, 1, {10, 10, 10}, 1, 1, sse42_DW_3D, vecPrcConnectParamsFP32)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_JIT_SSE42_DW_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice
+INSTANTIATE_TEST_SUITE_P(smoke_JIT_SSE42_DW_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice
 (JIT_SSE42_DW_GroupConvTestCases)), GroupConvolutionLayerCPUTest::getTestCaseName);
 
 /* ============= JIT AVX2 DW GroupConvolution ============= */
@@ -790,7 +790,7 @@ const std::vector<groupConvLayerCPUTestParamsSet> JIT_AVX2_DW_GroupConvTestCases
                                         8, 1, {10, 10, 10}, 1, 1, avx2_DW_3D, vecPrcConnectParamsFP32)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_JIT_AVX2_DW_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice
+INSTANTIATE_TEST_SUITE_P(smoke_JIT_AVX2_DW_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice
 (JIT_AVX2_DW_GroupConvTestCases)), GroupConvolutionLayerCPUTest::getTestCaseName);
 
 /* ============= JIT AVX512 DW GroupConvolution ============= */
@@ -825,7 +825,7 @@ const std::vector<groupConvLayerCPUTestParamsSet> JIT_AVX512_DW_GroupConvTestCas
                                         16, 1, {10, 10, 10}, 1, 1, avx512_DW_3D, vecPrcConnectParams)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_JIT_AVX512_DW_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice
+INSTANTIATE_TEST_SUITE_P(smoke_JIT_AVX512_DW_GroupConv, GroupConvolutionLayerCPUTest, ::testing::ValuesIn(filterParamsSetForDevice
 (JIT_AVX512_DW_GroupConvTestCases)), GroupConvolutionLayerCPUTest::getTestCaseName);
 
 /* ============= JIT SSE42 1x1 Convolution (not supported with groups) ============= */
@@ -860,7 +860,7 @@ const std::vector<CPUSpecificParams> CPUParams_1D = {
         conv_avx512_1D
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConv_1D, GroupConvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConv_1D, GroupConvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             convParams_1D,
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/group_convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/group_convolution_backprop_data.cpp
index 8645f0e21be869..8ceff946d147d1 100755
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/group_convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/group_convolution_backprop_data.cpp
@@ -169,7 +169,7 @@ const auto groupConvParams_ExplicitPadding_Planar_2D = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_2D_Planar_FP32, GroupDeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupDeconv_2D_Planar_FP32, GroupDeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             groupConvParams_ExplicitPadding_Planar_2D,
@@ -185,7 +185,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_2D_Planar_FP32, GroupDeconvolutionLaye
         ::testing::Values(cpuEmptyPluginConfig)),
     GroupDeconvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_2D_Planar_BF16, GroupDeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupDeconv_2D_Planar_BF16, GroupDeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             groupConvParams_ExplicitPadding_Planar_2D,
@@ -213,7 +213,7 @@ const auto groupConvParams_ExplicitPadding_Planar_3D = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_3D_Planar_FP32, GroupDeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupDeconv_3D_Planar_FP32, GroupDeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             groupConvParams_ExplicitPadding_Planar_3D,
@@ -229,7 +229,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_3D_Planar_FP32, GroupDeconvolutionLaye
         ::testing::Values(cpuEmptyPluginConfig)),
     GroupDeconvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_3D_Planar_BF16, GroupDeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupDeconv_3D_Planar_BF16, GroupDeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             groupConvParams_ExplicitPadding_Planar_3D,
@@ -257,7 +257,7 @@ const auto groupConvParams_ExplicitPadding_Blocked_2D = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_2D_Blocked_FP32, GroupDeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupDeconv_2D_Blocked_FP32, GroupDeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             groupConvParams_ExplicitPadding_Blocked_2D,
@@ -273,7 +273,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_2D_Blocked_FP32, GroupDeconvolutionLay
         ::testing::Values(cpuEmptyPluginConfig)),
     GroupDeconvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_2D_Blocked_BF16, GroupDeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupDeconv_2D_Blocked_BF16, GroupDeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             groupConvParams_ExplicitPadding_Blocked_2D,
@@ -301,7 +301,7 @@ const auto groupConvParams_ExplicitPadding_Blocked_3D = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_3D_Blocked_FP32, GroupDeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupDeconv_3D_Blocked_FP32, GroupDeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             groupConvParams_ExplicitPadding_Blocked_3D,
@@ -317,7 +317,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_3D_Blocked_FP32, GroupDeconvolutionLay
         ::testing::Values(cpuEmptyPluginConfig)),
     GroupDeconvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_3D_Blocked_BF16, GroupDeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupDeconv_3D_Blocked_BF16, GroupDeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             groupConvParams_ExplicitPadding_Blocked_3D,
@@ -345,7 +345,7 @@ const auto groupConvParams_ExplicitPadding_DW_2D = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_2D_DW_FP32, GroupDeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupDeconv_2D_DW_FP32, GroupDeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             groupConvParams_ExplicitPadding_DW_2D,
@@ -361,7 +361,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_2D_DW_FP32, GroupDeconvolutionLayerCPU
         ::testing::Values(cpuEmptyPluginConfig)),
     GroupDeconvolutionLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupDeconv_2D_DW_BF16, GroupDeconvolutionLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupDeconv_2D_DW_BF16, GroupDeconvolutionLayerCPUTest,
     ::testing::Combine(
         ::testing::Combine(
             groupConvParams_ExplicitPadding_DW_2D,
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gru_cell.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gru_cell.cpp
index 6f58eeda2f42e0..1e5a3602765a16 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gru_cell.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gru_cell.cpp
@@ -117,7 +117,7 @@ std::vector<float> clip = {0.f};
 std::vector<bool> linear_before_reset = {true, false};
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32};
 
-INSTANTIATE_TEST_CASE_P(smoke_GRUCellCPU,
+INSTANTIATE_TEST_SUITE_P(smoke_GRUCellCPU,
                         GRUCellCPUTest,
                         ::testing::Combine(::testing::Combine(::testing::ValuesIn(should_decompose),
                                                               ::testing::ValuesIn(batch),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gru_sequence.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gru_sequence.cpp
index 55ece3f0a663be..799f7516887429 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gru_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/gru_sequence.cpp
@@ -186,7 +186,7 @@ std::vector<ngraph::op::RecurrentSequenceDirection> direction = {ngraph::op::Rec
 
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32};
 
-INSTANTIATE_TEST_CASE_P(smoke_GRUSequenceCPU,
+INSTANTIATE_TEST_SUITE_P(smoke_GRUSequenceCPU,
                         GRUSequenceCPUTest,
                         ::testing::Combine(::testing::Combine(::testing::ValuesIn(mode),
                                                               ::testing::ValuesIn(seq_lengths_zero_clip),
@@ -202,7 +202,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GRUSequenceCPU,
                                            ::testing::ValuesIn(additionalConfig)),
                         GRUSequenceCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GRUSequenceCPUBatchSizeOne,
+INSTANTIATE_TEST_SUITE_P(smoke_GRUSequenceCPUBatchSizeOne,
                         GRUSequenceCPUTest,
                         ::testing::Combine(::testing::Combine(::testing::ValuesIn(mode),
                                                               ::testing::ValuesIn(seq_lengths_zero_clip),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/interpolate.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/interpolate.cpp
index 55863095e75acb..05783b6b122096 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/interpolate.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/interpolate.cpp
@@ -260,7 +260,7 @@ std::vector<std::map<std::string, std::string>> filterAdditionalConfig() {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_InterpolateNN_Layout_Test, InterpolateLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_InterpolateNN_Layout_Test, InterpolateLayerCPUTest,
         ::testing::Combine(
             ::testing::Combine(
                 interpolateCasesNN,
@@ -278,7 +278,7 @@ INSTANTIATE_TEST_CASE_P(smoke_InterpolateNN_Layout_Test, InterpolateLayerCPUTest
             ::testing::ValuesIn(filterAdditionalConfig())),
     InterpolateLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_InterpolateLinearOnnx_Layout_Test, InterpolateLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_InterpolateLinearOnnx_Layout_Test, InterpolateLayerCPUTest,
         ::testing::Combine(
             ::testing::Combine(
                 interpolateCasesLinearOnnx,
@@ -296,7 +296,7 @@ INSTANTIATE_TEST_CASE_P(smoke_InterpolateLinearOnnx_Layout_Test, InterpolateLaye
             ::testing::ValuesIn(filterAdditionalConfig())),
     InterpolateLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_InterpolateLinear_Layout_Test, InterpolateLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_InterpolateLinear_Layout_Test, InterpolateLayerCPUTest,
         ::testing::Combine(
             ::testing::Combine(
                 interpolateCasesLinear,
@@ -314,7 +314,7 @@ INSTANTIATE_TEST_CASE_P(smoke_InterpolateLinear_Layout_Test, InterpolateLayerCPU
             ::testing::ValuesIn(filterAdditionalConfig())),
     InterpolateLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_InterpolateCubic_Layout_Test, InterpolateLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_InterpolateCubic_Layout_Test, InterpolateLayerCPUTest,
         ::testing::Combine(
             ::testing::Combine(
                 interpolateCasesCubic,
@@ -387,7 +387,7 @@ const auto interpolateCasesNN5D = ::testing::Combine(
         ::testing::ValuesIn(defaultAxes5D),
         ::testing::ValuesIn(defaultScales5D));
 
-INSTANTIATE_TEST_CASE_P(smoke_InterpolateLinearOnnx5D_Layout_Test, InterpolateLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_InterpolateLinearOnnx5D_Layout_Test, InterpolateLayerCPUTest,
         ::testing::Combine(
             ::testing::Combine(
                 interpolateCasesLinearOnnx5D,
@@ -405,7 +405,7 @@ INSTANTIATE_TEST_CASE_P(smoke_InterpolateLinearOnnx5D_Layout_Test, InterpolateLa
             ::testing::ValuesIn(filterAdditionalConfig())),
     InterpolateLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_InterpolateNN5D_Layout_Test, InterpolateLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_InterpolateNN5D_Layout_Test, InterpolateLayerCPUTest,
         ::testing::Combine(
             ::testing::Combine(
                 interpolateCasesNN5D,
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/logical.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/logical.cpp
index 6e423ca669f855..d9bcddab858e16 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/logical.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/logical.cpp
@@ -150,9 +150,9 @@ const auto LogicalTestParamsNot = ::testing::Combine(
         ::testing::Values(emptyCPUSpec));
 
 
-INSTANTIATE_TEST_CASE_P(smoke_Logical_Eltwise_CPU_BF16, LogicalLayerCPUTest, LogicalTestParams, LogicalLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Logical_Eltwise_CPU_BF16, LogicalLayerCPUTest, LogicalTestParams, LogicalLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Logical_Not_Eltwise_CPU_BF16, LogicalLayerCPUTest, LogicalTestParamsNot, LogicalLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Logical_Not_Eltwise_CPU_BF16, LogicalLayerCPUTest, LogicalTestParamsNot, LogicalLayerCPUTest::getTestCaseName);
 
 } // namespace
 } // namespace CPULayerTestsDefinitions
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/lstm_cell.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/lstm_cell.cpp
index 4ee12f62ea9353..32ab7da1004790 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/lstm_cell.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/lstm_cell.cpp
@@ -115,7 +115,7 @@ std::vector<std::vector<std::string>> activations = {{"sigmoid", "tanh", "tanh"}
 std::vector<float> clip{0.f};
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32, InferenceEngine::Precision::BF16};
 
-INSTANTIATE_TEST_CASE_P(smoke_LSTMCellCPU,
+INSTANTIATE_TEST_SUITE_P(smoke_LSTMCellCPU,
                         LSTMCellLayerCPUTest,
                         ::testing::Combine(::testing::Combine(::testing::ValuesIn(should_decompose),
                                                               ::testing::ValuesIn(batch),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/lstm_sequence.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/lstm_sequence.cpp
index 234a579085cd01..959d5e0c89d6e2 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/lstm_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/lstm_sequence.cpp
@@ -195,7 +195,7 @@ std::vector<float> clip{0.f};
 std::vector<ngraph::op::RecurrentSequenceDirection> direction = {ngraph::op::RecurrentSequenceDirection::FORWARD};
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32};
 
-INSTANTIATE_TEST_CASE_P(smoke_LSTMSequenceCPU,
+INSTANTIATE_TEST_SUITE_P(smoke_LSTMSequenceCPU,
                         LSTMSequenceCPUTest,
                         ::testing::Combine(::testing::Combine(::testing::ValuesIn(mode),
                                                               ::testing::ValuesIn(seq_lengths_zero_clip),
@@ -211,7 +211,7 @@ INSTANTIATE_TEST_CASE_P(smoke_LSTMSequenceCPU,
                                            ::testing::ValuesIn(additionalConfig)),
                         LSTMSequenceCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_LSTMSequenceCPUbatchSizeOne,
+INSTANTIATE_TEST_SUITE_P(smoke_LSTMSequenceCPUbatchSizeOne,
                         LSTMSequenceCPUTest,
                         ::testing::Combine(::testing::Combine(::testing::ValuesIn(mode),
                                                               ::testing::ValuesIn(seq_lengths_zero_clip),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/mat_mul.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/mat_mul.cpp
index 020d9ffc3bf4f8..c2285e5f8cf0a8 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/mat_mul.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/mat_mul.cpp
@@ -149,7 +149,7 @@ const auto testParams2D = ::testing::Combine(fullyConnectedParams2D,
                                              ::testing::Values(MatMulNodeType::FullyConnected),
                                              ::testing::ValuesIn(fusingParamsSet2D));
 
-INSTANTIATE_TEST_CASE_P(smoke_Check_2D, MatMulLayerCPUTest, testParams2D, MatMulLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Check_2D, MatMulLayerCPUTest, testParams2D, MatMulLayerCPUTest::getTestCaseName);
 
 const std::vector<std::pair<SizeVector, SizeVector>> IS3D = {
     {{1, 32, 120}, {120, 5}},
@@ -171,7 +171,7 @@ const auto testParams3D = ::testing::Combine(fullyConnectedParams3D,
                                              ::testing::Values(MatMulNodeType::FullyConnected),
                                              ::testing::ValuesIn(fusingParamsSet3D));
 
-INSTANTIATE_TEST_CASE_P(smoke_Check_3D, MatMulLayerCPUTest, testParams3D, MatMulLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Check_3D, MatMulLayerCPUTest, testParams3D, MatMulLayerCPUTest::getTestCaseName);
 
 }; // namespace fullyConnected
 
@@ -196,7 +196,7 @@ const auto testParams = ::testing::Combine(gemmParams,
                                            ::testing::Values(MatMulNodeType::MatMul),
                                            ::testing::Values(emptyFusingSpec));
 
-INSTANTIATE_TEST_CASE_P(smoke_Check, MatMulLayerCPUTest, testParams, MatMulLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Check, MatMulLayerCPUTest, testParams, MatMulLayerCPUTest::getTestCaseName);
 
 }; // namespace gemm
 
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/mvn.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/mvn.cpp
index f89e302c539e50..6b877960e7f2f6 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/mvn.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/mvn.cpp
@@ -171,7 +171,7 @@ const auto Mvn3D = ::testing::Combine(
         ::testing::ValuesIn(inpPrc),
         ::testing::ValuesIn(outPrc));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_Mvn3D, MvnLayerCPUTest, Mvn3D, MvnLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Mvn3D, MvnLayerCPUTest, Mvn3D, MvnLayerCPUTest::getTestCaseName);
 
 const auto Mvn4D = ::testing::Combine(
         ::testing::Combine(
@@ -186,7 +186,7 @@ const auto Mvn4D = ::testing::Combine(
         ::testing::ValuesIn(inpPrc),
         ::testing::ValuesIn(outPrc));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_Mvn4D, MvnLayerCPUTest, Mvn4D, MvnLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Mvn4D, MvnLayerCPUTest, Mvn4D, MvnLayerCPUTest::getTestCaseName);
 
 const auto Mvn5D = ::testing::Combine(
         ::testing::Combine(
@@ -201,7 +201,7 @@ const auto Mvn5D = ::testing::Combine(
         ::testing::ValuesIn(inpPrc),
         ::testing::ValuesIn(outPrc));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_Mvn5D, MvnLayerCPUTest, Mvn5D, MvnLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Mvn5D, MvnLayerCPUTest, Mvn5D, MvnLayerCPUTest::getTestCaseName);
 
 // 1D 2D case
 std::vector<fusingSpecificParams> fusingUnaryEltwiseParamsSet {
@@ -225,7 +225,7 @@ const auto Mvn1D = ::testing::Combine(
         ::testing::ValuesIn(inpPrc),
         ::testing::ValuesIn(outPrc));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_Mvn1D, MvnLayerCPUTest, Mvn1D, MvnLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Mvn1D, MvnLayerCPUTest, Mvn1D, MvnLayerCPUTest::getTestCaseName);
 
 // 2D no transformed
 const auto Mvn2D = ::testing::Combine(
@@ -241,7 +241,7 @@ const auto Mvn2D = ::testing::Combine(
         ::testing::ValuesIn(inpPrc),
         ::testing::ValuesIn(outPrc));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_Mvn2D, MvnLayerCPUTest, Mvn2D, MvnLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Mvn2D, MvnLayerCPUTest, Mvn2D, MvnLayerCPUTest::getTestCaseName);
 
 // 2d transformed
 const auto Mvn2DTrans = ::testing::Combine(
@@ -257,7 +257,7 @@ const auto Mvn2DTrans = ::testing::Combine(
         ::testing::ValuesIn(inpPrc),
         ::testing::ValuesIn(outPrc));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_MVN2DTrans, MvnLayerCPUTest, Mvn2DTrans, MvnLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_MVN2DTrans, MvnLayerCPUTest, Mvn2DTrans, MvnLayerCPUTest::getTestCaseName);
 
 } // namespace
 } // namespace CPULayerTestsDefinitions
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/normalize.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/normalize.cpp
index 74b265d1935dcf..17e9b24f76b26c 100755
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/normalize.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/normalize.cpp
@@ -119,7 +119,7 @@ const auto testParams_2D = ::testing::Combine(normalizeParams_2D,
                                               ::testing::Values(CPUSpecificParams{}),
                                               ::testing::ValuesIn(fusingParamsSet));
 
-INSTANTIATE_TEST_CASE_P(smoke_2D, NormalizeL2LayerCPUTest, testParams_2D, NormalizeL2LayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_2D, NormalizeL2LayerCPUTest, testParams_2D, NormalizeL2LayerCPUTest::getTestCaseName);
 
 /* ============= 3D ============= */
 const std::vector<std::vector<size_t>> inputShape_3D = {
@@ -144,7 +144,7 @@ const auto testParams_3D = ::testing::Combine(normalizeParams_3D,
                                               ::testing::Values(CPUSpecificParams{}),
                                               ::testing::ValuesIn(fusingParamsSet));
 
-INSTANTIATE_TEST_CASE_P(smoke_3D, NormalizeL2LayerCPUTest, testParams_3D, NormalizeL2LayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_3D, NormalizeL2LayerCPUTest, testParams_3D, NormalizeL2LayerCPUTest::getTestCaseName);
 
 /* ============= 4D ============= */
 const std::vector<std::vector<size_t>> inputShape_4D = {
@@ -183,7 +183,7 @@ const auto testParams_4D = ::testing::Combine(normalizeParams_4D,
                                               ::testing::ValuesIn(getCPUSpecificParams()),
                                               ::testing::ValuesIn(fusingParamsSet));
 
-INSTANTIATE_TEST_CASE_P(smoke_4D, NormalizeL2LayerCPUTest, testParams_4D, NormalizeL2LayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_4D, NormalizeL2LayerCPUTest, testParams_4D, NormalizeL2LayerCPUTest::getTestCaseName);
 
 } // namespace
 
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/one_hot.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/one_hot.cpp
index 4e966d6c0f91b9..e3f083e7460763 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/one_hot.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/one_hot.cpp
@@ -100,7 +100,7 @@ const auto testCase_1d = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(emptyCPUSpec)
 );
-INSTANTIATE_TEST_CASE_P(smoke_OneHotCPU_1D, OneHotLayerCPUTest, testCase_1d, OneHotLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_OneHotCPU_1D, OneHotLayerCPUTest, testCase_1d, OneHotLayerCPUTest::getTestCaseName);
 
 
 // 1d -> 2d, axis default
@@ -116,7 +116,7 @@ const auto testCase_2d = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(emptyCPUSpec)
 );
-INSTANTIATE_TEST_CASE_P(smoke_OneHotCPU_2D, OneHotLayerCPUTest, testCase_2d, OneHotLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_OneHotCPU_2D, OneHotLayerCPUTest, testCase_2d, OneHotLayerCPUTest::getTestCaseName);
 
 // 2d -> 3d, on_value, off_value
 const auto testCase_3d = ::testing::Combine(
@@ -131,7 +131,7 @@ const auto testCase_3d = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(emptyCPUSpec)
 );
-INSTANTIATE_TEST_CASE_P(smoke_OneHotCPU_3D, OneHotLayerCPUTest, testCase_3d, OneHotLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_OneHotCPU_3D, OneHotLayerCPUTest, testCase_3d, OneHotLayerCPUTest::getTestCaseName);
 
 // 3d -> 4d
 const auto testCase_4d = ::testing::Combine(
@@ -146,7 +146,7 @@ const auto testCase_4d = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(emptyCPUSpec)
 );
-INSTANTIATE_TEST_CASE_P(smoke_OneHotCPU_4D, OneHotLayerCPUTest, testCase_4d, OneHotLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_OneHotCPU_4D, OneHotLayerCPUTest, testCase_4d, OneHotLayerCPUTest::getTestCaseName);
 
 // 4d -> 5d
 const auto testCase_5d = ::testing::Combine(
@@ -161,7 +161,7 @@ const auto testCase_5d = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU),
         ::testing::Values(emptyCPUSpec)
 );
-INSTANTIATE_TEST_CASE_P(smoke_OneHotCPU_5D, OneHotLayerCPUTest, testCase_5d, OneHotLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_OneHotCPU_5D, OneHotLayerCPUTest, testCase_5d, OneHotLayerCPUTest::getTestCaseName);
 
 } // namespace
 } // namespace CPULayerTestsDefinitions
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pad.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pad.cpp
index 82bdc0971a6dba..b528e9f644f42b 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pad.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pad.cpp
@@ -122,7 +122,7 @@ const auto pad4DConstParamsBlocked = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_CPUPad4DConstBlocked,
         PadLayerCPUTest,
         ::testing::Combine(
@@ -145,7 +145,7 @@ const auto pad4DConstParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_CPUPad4DConst,
         PadLayerCPUTest,
         ::testing::Combine(
@@ -167,7 +167,7 @@ const auto pad4DParamsBlocked = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_CPUPad4DBlocked,
         PadLayerCPUTest,
         ::testing::Combine(
@@ -189,7 +189,7 @@ const auto pad4DParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_CPUPad4D,
         PadLayerCPUTest,
         ::testing::Combine(
@@ -225,7 +225,7 @@ const auto pad5DConstParamsBlocked = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_CPUPad5DConstBlocked,
         PadLayerCPUTest,
         ::testing::Combine(
@@ -247,7 +247,7 @@ const auto pad5DConstParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_CPUPad5DConst,
         PadLayerCPUTest,
         ::testing::Combine(
@@ -269,7 +269,7 @@ const auto pad5DParamsBlocked = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_CPUPad5DBlocked,
         PadLayerCPUTest,
         ::testing::Combine(
@@ -291,7 +291,7 @@ const auto pad5DParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_CPUPad5D,
         PadLayerCPUTest,
         ::testing::Combine(
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp
index 6fb1f7d1169440..0021a43dbb9d8b 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/pooling.cpp
@@ -144,7 +144,7 @@ const std::vector<poolSpecificParams> paramsAvg4D_RefOnly = {
                             ngraph::op::RoundingType::CEIL, ngraph::op::PadType::EXPLICIT, false },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_CPU_4D, PoolingLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_CPU_4D, PoolingLayerCPUTest,
                         ::testing::Combine(
                             ::testing::Combine(
                                 ::testing::ValuesIn(paramsMax4D),
@@ -159,7 +159,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MaxPool_CPU_4D, PoolingLayerCPUTest,
                         ::testing::Values(emptyFusingSpec)),
                         PoolingLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D, PoolingLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_CPU_4D, PoolingLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         ::testing::ValuesIn(paramsAvg4D),
@@ -174,7 +174,7 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D, PoolingLayerCPUTest,
                                 ::testing::Values(emptyFusingSpec)),
                         PoolingLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D_NotOptimized, PoolingLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_CPU_4D_NotOptimized, PoolingLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         ::testing::ValuesIn(paramsAvg4D_RefOnly),
@@ -214,7 +214,7 @@ const std::vector<poolSpecificParams> paramsAvg5D_RefOnly = {
                             ngraph::op::RoundingType::CEIL, ngraph::op::PadType::EXPLICIT, false },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_CPU_5D, PoolingLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_CPU_5D, PoolingLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         ::testing::ValuesIn(paramsMax5D),
@@ -229,7 +229,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MaxPool_CPU_5D, PoolingLayerCPUTest,
                                 ::testing::Values(emptyFusingSpec)),
                         PoolingLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D, PoolingLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_CPU_5D, PoolingLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         ::testing::ValuesIn(paramsAvg5D),
@@ -244,7 +244,7 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D, PoolingLayerCPUTest,
                                 ::testing::Values(emptyFusingSpec)),
                         PoolingLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D_NotOptimized, PoolingLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_CPU_5D_NotOptimized, PoolingLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         ::testing::ValuesIn(paramsAvg5D_RefOnly),
@@ -279,7 +279,7 @@ std::vector<fusingSpecificParams> fusingParamsSet {
     fusingFakeQuantizePerChannel,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D_I8, PoolingLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_CPU_4D_I8, PoolingLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         ::testing::ValuesIn(paramsAvg4D),
@@ -294,7 +294,7 @@ INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_4D_I8, PoolingLayerCPUTest,
                                 ::testing::ValuesIn(fusingParamsSet)),
                         PoolingLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_CPU_5D_I8, PoolingLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_CPU_5D_I8, PoolingLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         ::testing::ValuesIn(paramsAvg5D),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/psroi_pooling.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/psroi_pooling.cpp
index 5327e72a6a2281..01131bf0b24e9c 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/psroi_pooling.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/psroi_pooling.cpp
@@ -163,7 +163,7 @@ const auto psroiPoolingBilinearParams = ::testing::Combine(
         ::testing::Values("bilinear")
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_PSROIPoolingAverageLayoutTest, PSROIPoolingLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_PSROIPoolingAverageLayoutTest, PSROIPoolingLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         psroiPoolingAverageParams,
@@ -172,7 +172,7 @@ INSTANTIATE_TEST_CASE_P(smoke_PSROIPoolingAverageLayoutTest, PSROIPoolingLayerCP
                                 ::testing::ValuesIn(filterCPUSpecificParams(resCPUParams))),
                         PSROIPoolingLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_PSROIPoolingBilinearLayoutTest, PSROIPoolingLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_PSROIPoolingBilinearLayoutTest, PSROIPoolingLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Combine(
                                         psroiPoolingBilinearParams,
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/reduce_ops.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/reduce_ops.cpp
index dc528b82fad1c6..222651e51ff564 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/reduce_ops.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/reduce_ops.cpp
@@ -282,56 +282,56 @@ const auto params_MultiAxisLogical5D = testing::Combine(
                 testing::Values(CommonTestUtils::DEVICE_CPU)),
         testing::ValuesIn(filterCPUSpecificParams(cpuParams_5D)));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceOneAxis_CPU,
         ReduceCPULayerTest,
         paramsOneAxis,
         ReduceCPULayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceLogicalOneAxis_CPU,
         ReduceCPULayerTest,
         paramsOneAxisLogical,
         ReduceCPULayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_ReductionTypes_CPU,
         ReduceCPULayerTest,
         params_MultiAxis,
         ReduceCPULayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_ReductionTypes4D_CPU,
         ReduceCPULayerTest,
         params_MultiAxis_4D,
         ReduceCPULayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_ReductionTypes5D_CPU,
         ReduceCPULayerTest,
         params_MultiAxis_5D,
         ReduceCPULayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceLogical_ReductionTypes_CPU,
         ReduceCPULayerTest,
         params_MultiAxisLogical,
         ReduceCPULayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceLogical4D_ReductionTypes_CPU,
         ReduceCPULayerTest,
         params_MultiAxisLogical4D,
         ReduceCPULayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceLogical5D_ReductionTypes_CPU,
         ReduceCPULayerTest,
         params_MultiAxisLogical5D,
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/region_yolo.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/region_yolo.cpp
index 4c88cf8d884873..e5a00a813f765d 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/region_yolo.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/region_yolo.cpp
@@ -160,8 +160,8 @@ const auto testCase_yolov2_caffe = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_CPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsRegionYolov3CPU, RegionYoloCPULayerTest, testCase_yolov3, RegionYoloCPULayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsRegionYoloMxnetCPU, RegionYoloCPULayerTest, testCase_yolov3_mxnet, RegionYoloCPULayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsRegionYoloCaffeCPU, RegionYoloCPULayerTest, testCase_yolov2_caffe, RegionYoloCPULayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsRegionYolov3CPU, RegionYoloCPULayerTest, testCase_yolov3, RegionYoloCPULayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsRegionYoloMxnetCPU, RegionYoloCPULayerTest, testCase_yolov3_mxnet, RegionYoloCPULayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsRegionYoloCaffeCPU, RegionYoloCPULayerTest, testCase_yolov2_caffe, RegionYoloCPULayerTest::getTestCaseName);
 } // namespace
 } // namespace CPULayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/rnn_cell.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/rnn_cell.cpp
index 381bdfecf369a6..59b23396c38396 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/rnn_cell.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/rnn_cell.cpp
@@ -107,7 +107,7 @@ std::vector<std::vector<std::string>> activations = {{"relu"}, {"sigmoid"}, {"ta
 std::vector<float> clip = {0.f};
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32};
 
-INSTANTIATE_TEST_CASE_P(smoke_RNNCellCPU,
+INSTANTIATE_TEST_SUITE_P(smoke_RNNCellCPU,
                         RNNCellCPUTest,
                         ::testing::Combine(::testing::Combine(::testing::ValuesIn(should_decompose),
                                                               ::testing::ValuesIn(batch),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/rnn_sequence.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/rnn_sequence.cpp
index 6c7f873cc95c9d..e1e66a660501e8 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/rnn_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/rnn_sequence.cpp
@@ -167,7 +167,7 @@ std::vector<ngraph::op::RecurrentSequenceDirection> direction{ngraph::op::Recurr
 
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32};
 
-INSTANTIATE_TEST_CASE_P(smoke_RNNSequenceCPU,
+INSTANTIATE_TEST_SUITE_P(smoke_RNNSequenceCPU,
                         RNNSequenceCPUTest,
                         ::testing::Combine(::testing::Combine(::testing::ValuesIn(mode),
                                                               ::testing::ValuesIn(seq_lengths_zero_clip),
@@ -183,7 +183,7 @@ INSTANTIATE_TEST_CASE_P(smoke_RNNSequenceCPU,
                                            ::testing::ValuesIn(additionalConfig)),
                         RNNSequenceCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_RNNSequenceCPUBatchSizeOne,
+INSTANTIATE_TEST_SUITE_P(smoke_RNNSequenceCPUBatchSizeOne,
                         RNNSequenceCPUTest,
                         ::testing::Combine(::testing::Combine(::testing::ValuesIn(mode),
                                                               ::testing::ValuesIn(seq_lengths_zero_clip),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/roi_pooling.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/roi_pooling.cpp
index 7422befc949160..094a9f4b507fc1 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/roi_pooling.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/roi_pooling.cpp
@@ -186,14 +186,14 @@ const auto test_ROIPooling_bilinear = ::testing::Combine(::testing::ValuesIn(inS
                                                          ::testing::ValuesIn(netPRCs),
                                                          ::testing::Values(CommonTestUtils::DEVICE_CPU));
 
-INSTANTIATE_TEST_CASE_P(smoke_ROIPoolingCPU_max,
+INSTANTIATE_TEST_SUITE_P(smoke_ROIPoolingCPU_max,
                         ROIPoolingCPULayerTest,
                         ::testing::Combine(test_ROIPooling_max,
                                            ::testing::ValuesIn(selectCPUInfoForDevice()),
                                            ::testing::ValuesIn(additionalConfig)),
                         ROIPoolingCPULayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ROIPoolingCPU_bilinear,
+INSTANTIATE_TEST_SUITE_P(smoke_ROIPoolingCPU_bilinear,
                         ROIPoolingCPULayerTest,
                         ::testing::Combine(test_ROIPooling_bilinear,
                                            ::testing::ValuesIn(selectCPUInfoForDevice()),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/roialign.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/roialign.cpp
index aac49d69274c50..78a6fdcd0cba00 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/roialign.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/roialign.cpp
@@ -163,7 +163,7 @@ const auto roiAlignParams = ::testing::Combine(
         ::testing::ValuesIn(inputShapeVector)         // feature map shape
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ROIAlignLayoutTest, ROIAlignLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ROIAlignLayoutTest, ROIAlignLayerCPUTest,
         ::testing::Combine(
                 ::testing::Combine(
                         roiAlignParams,
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/shuffle_channels.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/shuffle_channels.cpp
index 0548aad8c5e7be..bb06d40a551153 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/shuffle_channels.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/shuffle_channels.cpp
@@ -151,7 +151,7 @@ const auto shuffleChannelsParams5D = ::testing::Combine(
         ::testing::ValuesIn(std::vector<int>{1, 2, 3, 6})
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels4D, ShuffleChannelsLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels4D, ShuffleChannelsLayerCPUTest,
         ::testing::Combine(
             ::testing::Combine(
                 shuffleChannelsParams4D,
@@ -165,7 +165,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels4D, ShuffleChannelsLayerCPUTest,
             ::testing::ValuesIn(filterCPUInfoForDevice4D())),
         ShuffleChannelsLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels5D, ShuffleChannelsLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels5D, ShuffleChannelsLayerCPUTest,
         ::testing::Combine(
             ::testing::Combine(
                 shuffleChannelsParams5D,
@@ -189,7 +189,7 @@ const auto shuffleChannelsParams5DBlock = ::testing::Combine(
         ::testing::ValuesIn(std::vector<int>{1, 2, 3, 6})
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels4DBlock, ShuffleChannelsLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels4DBlock, ShuffleChannelsLayerCPUTest,
         ::testing::Combine(
             ::testing::Combine(
                 shuffleChannelsParams4DBlock,
@@ -203,7 +203,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels4DBlock, ShuffleChannelsLayerCPUTes
             ::testing::ValuesIn(filterCPUInfoForDevice4DBlock())),
         ShuffleChannelsLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels5DBlock, ShuffleChannelsLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels5DBlock, ShuffleChannelsLayerCPUTest,
         ::testing::Combine(
             ::testing::Combine(
                 shuffleChannelsParams5DBlock,
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/softmax.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/softmax.cpp
index d672629ab0ca69..2877a64c44a457 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/softmax.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/softmax.cpp
@@ -118,7 +118,7 @@ const auto OptimizedParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU),
         testing::Values(emptyCPUSpec));
 
-INSTANTIATE_TEST_CASE_P(smoke_SoftMax_Optimized_CPU, SoftMaxLayerCPUTest, OptimizedParams, SoftMaxLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_SoftMax_Optimized_CPU, SoftMaxLayerCPUTest, OptimizedParams, SoftMaxLayerCPUTest::getTestCaseName);
 
 const auto NotOptimizedParams = testing::Combine(
         testing::Values(Precision::FP32, Precision::BF16),
@@ -126,7 +126,7 @@ const auto NotOptimizedParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_CPU),
         testing::Values(notOptimizedCPUSpec));
 
-INSTANTIATE_TEST_CASE_P(smoke_SoftMax_CPU, SoftMaxLayerCPUTest, NotOptimizedParams, SoftMaxLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_SoftMax_CPU, SoftMaxLayerCPUTest, NotOptimizedParams, SoftMaxLayerCPUTest::getTestCaseName);
 
 } // namespace
 } // namespace CPULayerTestsDefinitions
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/space_to_batch.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/space_to_batch.cpp
index 64dd742000c88e..61301aebba7846 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/space_to_batch.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/space_to_batch.cpp
@@ -125,10 +125,10 @@ const auto spaceToBatchParamsSet4D2 = ::testing::Combine(
                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(cpuParams_4D));
 
-INSTANTIATE_TEST_CASE_P(smoke_SpaceToBatchCPULayerTestCase1_4D, SpaceToBatchCPULayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SpaceToBatchCPULayerTestCase1_4D, SpaceToBatchCPULayerTest,
                         spaceToBatchParamsSet4D1, SpaceToBatchCPULayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_SpaceToBatchCPULayerTestCase2_4D, SpaceToBatchCPULayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SpaceToBatchCPULayerTestCase2_4D, SpaceToBatchCPULayerTest,
                         spaceToBatchParamsSet4D2, SpaceToBatchCPULayerTest::getTestCaseName);
 
 const std::vector<std::vector<int64_t>> blockShape5D = {{1, 1, 2, 2, 1}, {1, 2, 4, 1, 3}};
@@ -157,7 +157,7 @@ const auto spaceToBatchParamsSet5D = ::testing::Combine(
                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
         ::testing::ValuesIn(cpuParams_5D));
 
-INSTANTIATE_TEST_CASE_P(smoke_SpaceToBatchCPULayerTestCase_5D, SpaceToBatchCPULayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SpaceToBatchCPULayerTestCase_5D, SpaceToBatchCPULayerTest,
                         spaceToBatchParamsSet5D, SpaceToBatchCPULayerTest::getTestCaseName);
 
 } // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/space_to_depth.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/space_to_depth.cpp
index a080146064f453..4c3bdbbba8d516 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/space_to_depth.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/space_to_depth.cpp
@@ -124,7 +124,7 @@ const auto spaceToDepthBS2_4DParams = testing::Combine(
         testing::ValuesIn(filterCPUInfoForDevice(CPUParamsBS2_4D))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_CPUSpaceToDepthBS2_4D, SpaceToDepthLayerCPUTest, spaceToDepthBS2_4DParams, SpaceToDepthLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CPUSpaceToDepthBS2_4D, SpaceToDepthLayerCPUTest, spaceToDepthBS2_4DParams, SpaceToDepthLayerCPUTest::getTestCaseName);
 
 const std::vector<CPUSpecificParams> CPUParamsBS3_4D = {
         cpuParams_nhwc_avx2,
@@ -142,7 +142,7 @@ const auto spaceToDepthBS3_4DParams = testing::Combine(
         ::testing::ValuesIn(filterCPUInfoForDevice(CPUParamsBS3_4D))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_CPUSpaceToDepthBS3_4D, SpaceToDepthLayerCPUTest, spaceToDepthBS3_4DParams, SpaceToDepthLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CPUSpaceToDepthBS3_4D, SpaceToDepthLayerCPUTest, spaceToDepthBS3_4DParams, SpaceToDepthLayerCPUTest::getTestCaseName);
 
 const std::vector<std::vector<size_t >> inputShapesBS2_5D = {
         {1, 16, 2, 2, 2}, {1, 16, 4, 4, 2}, {1, 32, 2, 6, 2}, {2, 32, 4, 2, 2}, {1, 48, 6, 2, 2}, {2, 64, 2, 2, 6}
@@ -171,7 +171,7 @@ const auto spaceToDepthBS2_5DParams = testing::Combine(
         ::testing::ValuesIn(filterCPUInfoForDevice(CPUParamsBS2_5D))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_CPUSpaceToDepthBS2_5D, SpaceToDepthLayerCPUTest, spaceToDepthBS2_5DParams, SpaceToDepthLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CPUSpaceToDepthBS2_5D, SpaceToDepthLayerCPUTest, spaceToDepthBS2_5DParams, SpaceToDepthLayerCPUTest::getTestCaseName);
 
 const std::vector<CPUSpecificParams> CPUParamsBS3_5D = {
         cpuParams_ndhwc_avx2,
@@ -189,7 +189,7 @@ const auto spaceToDepthBS3_5DParams = testing::Combine(
         ::testing::ValuesIn(filterCPUInfoForDevice(CPUParamsBS3_5D))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_CPUSpaceToDepthBS3_5D, SpaceToDepthLayerCPUTest, spaceToDepthBS3_5DParams, SpaceToDepthLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CPUSpaceToDepthBS3_5D, SpaceToDepthLayerCPUTest, spaceToDepthBS3_5DParams, SpaceToDepthLayerCPUTest::getTestCaseName);
 
 } // namespace
 } // namespace CPULayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/split.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/split.cpp
index 7129ee4c4aea51..14b83c08ae07a6 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/split.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/split.cpp
@@ -126,7 +126,7 @@ const std::vector<std::vector<size_t>> outIndices3 = {{0, 1, 2}, {0, 1, 1, 0, 2}
 const std::vector<std::vector<size_t>> outIndices4 = {{0, 1, 2, 3}, {0, 1, 1, 0, 2, 3}, {0, 0, 0, 2, 3}};
 
 
-INSTANTIATE_TEST_CASE_P(smoke_Split4D_CPU_Nspc2NcspSpecial, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split4D_CPU_Nspc2NcspSpecial, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(4),
                                 ::testing::Values(1),
@@ -137,7 +137,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split4D_CPU_Nspc2NcspSpecial, SplitLayerCPUTest,
                                 ::testing::Values(perChannelsToPlanar_4D)),
                         SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split5D_CPU_Nspc2NcspSpecial, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split5D_CPU_Nspc2NcspSpecial, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(3),
                                 ::testing::Values(1),
@@ -148,7 +148,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split5D_CPU_Nspc2NcspSpecial, SplitLayerCPUTest,
                                 ::testing::Values(perChannelsToPlanar_5D)),
                         SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split4D_CPU_Block8inPlace, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split4D_CPU_Block8inPlace, SplitLayerCPUTest,
                     ::testing::Combine(
                             ::testing::Values(3),
                             ::testing::Values(0, 1),
@@ -159,7 +159,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split4D_CPU_Block8inPlace, SplitLayerCPUTest,
                             ::testing::Values(planar_4D, planar_4D_ref, perChannels_4D, blocked8_4D)),
                     SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split4D_CPU_Block8, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split4D_CPU_Block8, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(3),
                                 ::testing::Values(2, 3),
@@ -170,7 +170,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split4D_CPU_Block8, SplitLayerCPUTest,
                                 ::testing::Values(planar_4D, planar_4D_ref, perChannels_4D, blocked8_4D_ref)),
                         SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split4D_CPU_Block16inPlace, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split4D_CPU_Block16inPlace, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(4),
                                 ::testing::Values(0, 1),
@@ -181,7 +181,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split4D_CPU_Block16inPlace, SplitLayerCPUTest,
                                 ::testing::Values(blocked16_4D)),
                         SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split4D_CPU_Block16, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split4D_CPU_Block16, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(4),
                                 ::testing::Values(2, 3),
@@ -192,7 +192,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split4D_CPU_Block16, SplitLayerCPUTest,
                                 ::testing::Values(blocked16_4D_ref)),
                         SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split5D_CPU_Block8inPlace, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split5D_CPU_Block8inPlace, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(3),
                                 ::testing::Values(0, 1),
@@ -203,7 +203,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split5D_CPU_Block8inPlace, SplitLayerCPUTest,
                                 ::testing::Values(planar_5D, planar_5D_ref, perChannels_5D, blocked8_5D)),
                         SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split5D_CPU_Block8, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split5D_CPU_Block8, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(3),
                                 ::testing::Values(2, 3, 4),
@@ -214,7 +214,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split5D_CPU_Block8, SplitLayerCPUTest,
                                 ::testing::Values(planar_5D, planar_5D_ref, perChannels_5D, blocked8_5D_ref)),
                         SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split5D_CPU_Block16inPlace, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split5D_CPU_Block16inPlace, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(4),
                                 ::testing::Values(0, 1),
@@ -225,7 +225,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split5D_CPU_Block16inPlace, SplitLayerCPUTest,
                                 ::testing::Values(blocked16_5D)),
                         SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split5D_CPU_Block16, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split5D_CPU_Block16, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(4),
                                 ::testing::Values(2, 3, 4),
@@ -236,7 +236,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split5D_CPU_Block16, SplitLayerCPUTest,
                                 ::testing::Values(blocked16_5D_ref)),
                         SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split3D, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split3D, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(7),
                                 ::testing::Values(0, 1, 2),
@@ -247,7 +247,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split3D, SplitLayerCPUTest,
                                 ::testing::Values(CPUSpecificParams{{}, {}, {}, "unknown"}, CPUSpecificParams{{}, {}, {"ref"}, "ref"})),
                                 SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split2D, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split2D, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(2),
                                 ::testing::Values(0, 1),
@@ -258,7 +258,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Split2D, SplitLayerCPUTest,
                                 ::testing::Values(CPUSpecificParams{{}, {}, {}, "unknown"}, CPUSpecificParams{{}, {}, {"ref"}, "ref"})),
                         SplitLayerCPUTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Split1D, SplitLayerCPUTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Split1D, SplitLayerCPUTest,
                         ::testing::Combine(
                                 ::testing::Values(5),
                                 ::testing::Values(0),
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/strided_slice.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/strided_slice.cpp
index c4746897552ef2..b01e2455a2904e 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/strided_slice.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/strided_slice.cpp
@@ -121,7 +121,7 @@ const auto StridedSliceParamsPlain2D = ::testing::Combine(
         ::testing::Values(additional_config),
         ::testing::Values(emptyCPUSpec));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_Plain_2D, StridedSliceLayerCPUTest, StridedSliceParamsPlain2D, StridedSliceLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Plain_2D, StridedSliceLayerCPUTest, StridedSliceParamsPlain2D, StridedSliceLayerCPUTest::getTestCaseName);
 
 const std::vector<StridedSliceSpecificParams> testCasesCommon4D = {
         StridedSliceSpecificParams{ { 1, 5, 32, 32 }, { 0, 2, 5, 4 }, { 1, 4, 28, 27 }, { 1, 1, 1, 1 },
@@ -162,7 +162,7 @@ const auto StridedSliceParamsCommon4D = ::testing::Combine(
         ::testing::Values(additional_config),
         ::testing::ValuesIn(CPUParamsCommon4D));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_Common_4D, StridedSliceLayerCPUTest, StridedSliceParamsCommon4D, StridedSliceLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Common_4D, StridedSliceLayerCPUTest, StridedSliceParamsCommon4D, StridedSliceLayerCPUTest::getTestCaseName);
 
 const std::vector<StridedSliceSpecificParams> testCasesBlocked4D = {
         StridedSliceSpecificParams{ { 1, 16, 32, 32 }, { 0, 0, 5, 4 }, { 1, 16, 28, 27 }, { 1, 1, 1, 1 },
@@ -207,7 +207,7 @@ const auto StridedSliceParamsBlocked4D = ::testing::Combine(
         ::testing::Values(additional_config),
         ::testing::ValuesIn(CPUParamsBlocked4D));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_Blocked_4D, StridedSliceLayerCPUTest, StridedSliceParamsBlocked4D, StridedSliceLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Blocked_4D, StridedSliceLayerCPUTest, StridedSliceParamsBlocked4D, StridedSliceLayerCPUTest::getTestCaseName);
 
 const std::vector<StridedSliceSpecificParams> testCasesCommon5D = {
         StridedSliceSpecificParams{ { 1, 5, 20, 32, 32 }, { 0, 2, 0, 5, 4 }, { 1, 4, 5, 28, 27 }, { 1, 1, 1, 1, 1 },
@@ -250,7 +250,7 @@ const auto StridedSliceParamsCommon5D = ::testing::Combine(
         ::testing::Values(additional_config),
         ::testing::ValuesIn(CPUParamsCommon5D));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_Common_5D, StridedSliceLayerCPUTest, StridedSliceParamsCommon5D, StridedSliceLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Common_5D, StridedSliceLayerCPUTest, StridedSliceParamsCommon5D, StridedSliceLayerCPUTest::getTestCaseName);
 
 const std::vector<StridedSliceSpecificParams> testCasesBlocked5D = {
         StridedSliceSpecificParams{ { 1, 16, 20, 32, 32 }, { 0, 0, 0, 5, 4 }, { 1, 16, 5, 28, 27 }, { 1, 1, 1, 1, 1 },
@@ -297,7 +297,7 @@ const auto StridedSliceParamsBlocked5D = ::testing::Combine(
         ::testing::Values(additional_config),
         ::testing::ValuesIn(CPUParamsBlocked5D));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs_Blocked_5D, StridedSliceLayerCPUTest, StridedSliceParamsBlocked5D, StridedSliceLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Blocked_5D, StridedSliceLayerCPUTest, StridedSliceParamsBlocked5D, StridedSliceLayerCPUTest::getTestCaseName);
 
 } // namespace
 } // namespace CPULayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/transpose.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/transpose.cpp
index 053ba0ebeae69c..9708b8d50ec48d 100644
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/transpose.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/transpose.cpp
@@ -136,7 +136,7 @@ const auto params4D = ::testing::Combine(
         ::testing::Values(additional_config),
         ::testing::ValuesIn(CPUParams4D));
 
-INSTANTIATE_TEST_CASE_P(smoke_Transpose4D_CPU, TransposeLayerCPUTest, params4D, TransposeLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Transpose4D_CPU, TransposeLayerCPUTest, params4D, TransposeLayerCPUTest::getTestCaseName);
 
 const auto paramsPerChannels4D = ::testing::Combine(
         ::testing::ValuesIn(inputOrderPerChannels4D),
@@ -146,7 +146,7 @@ const auto paramsPerChannels4D = ::testing::Combine(
         ::testing::Values(additional_config),
         ::testing::Values(cpuParams_nhwc));
 
-INSTANTIATE_TEST_CASE_P(smoke_PermutePerChannels4D_CPU, TransposeLayerCPUTest, paramsPerChannels4D, TransposeLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_PermutePerChannels4D_CPU, TransposeLayerCPUTest, paramsPerChannels4D, TransposeLayerCPUTest::getTestCaseName);
 
 const std::vector<std::vector<size_t>> inputShapes5D = {
         {2, 32, 5, 10, 20}
@@ -189,7 +189,7 @@ const auto params5D = ::testing::Combine(
         ::testing::Values(additional_config),
         ::testing::ValuesIn(CPUParams5D));
 
-INSTANTIATE_TEST_CASE_P(smoke_Transpose5D_CPU, TransposeLayerCPUTest, params5D, TransposeLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Transpose5D_CPU, TransposeLayerCPUTest, params5D, TransposeLayerCPUTest::getTestCaseName);
 
 const auto paramsPerChannels5D = ::testing::Combine(
         ::testing::ValuesIn(inputOrderPerChannels5D),
@@ -199,7 +199,7 @@ const auto paramsPerChannels5D = ::testing::Combine(
         ::testing::Values(additional_config),
         ::testing::Values(cpuParams_ndhwc));
 
-INSTANTIATE_TEST_CASE_P(smoke_PermutePerChannels5D_CPU, TransposeLayerCPUTest, paramsPerChannels5D, TransposeLayerCPUTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_PermutePerChannels5D_CPU, TransposeLayerCPUTest, paramsPerChannels5D, TransposeLayerCPUTest::getTestCaseName);
 
 } // namespace
 } // namespace CPULayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv3d_reshape.cpp b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv3d_reshape.cpp
index 80ff36aeeae034..fd65019e6acaa0 100644
--- a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv3d_reshape.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv3d_reshape.cpp
@@ -91,7 +91,7 @@ const std::vector<size_t> numOut = { 1, 2, 5 };
 const auto conv3dReshapeParams = ::testing::Combine(::testing::ValuesIn(convType),
                                                     ::testing::ValuesIn(numOut));
 
-INSTANTIATE_TEST_CASE_P(smoke_Conv3dReshapeTest, Conv3dReshapeTest, conv3dReshapeParams, Conv3dReshapeTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Conv3dReshapeTest, Conv3dReshapeTest, conv3dReshapeParams, Conv3dReshapeTest::getTestCaseName);
 
 } // namespace
 
diff --git a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv_concat.cpp b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv_concat.cpp
index be03883ed6e8e0..aef68679e83ac0 100644
--- a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv_concat.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv_concat.cpp
@@ -160,7 +160,7 @@ const auto params2DConv = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D1x1, ConvConcatSubgraphTest, params2DConv, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D1x1, ConvConcatSubgraphTest, params2DConv, ConvConcatSubgraphTest::getTestCaseName);
 
 const std::vector<CPUSpecificParams> CPUParams2DDeconv = {
     conv_avx2_2D_1x1,
@@ -175,7 +175,7 @@ const auto params2DDeconv = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D1x1, ConvConcatSubgraphTest, params2DDeconv, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D1x1, ConvConcatSubgraphTest, params2DDeconv, ConvConcatSubgraphTest::getTestCaseName);
 
 }  // namespace Kernel_1x1
 
@@ -198,7 +198,7 @@ const auto params2D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DWGroupConvolutionBackpropData2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DWGroupConvolutionBackpropData2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
 
 }  // namespace GroupConvolutionBackpropDataDWConcat
 
@@ -221,7 +221,7 @@ const auto params2D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DWGroupConvolution2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DWGroupConvolution2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
 
 /* ============= GroupConvolution (DW 3D) ============= */
 commonConvParams dwConvParams3D = commonConvParams{kernelSize3D, strides3D, padBegin3D, padEnd3D, dilation3D, numOutChannels, paddingType, numOutChannels};
@@ -240,7 +240,7 @@ const auto params3D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DWGroupConvolution3D, ConvConcatSubgraphTest, params3D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DWGroupConvolution3D, ConvConcatSubgraphTest, params3D, ConvConcatSubgraphTest::getTestCaseName);
 
 }  // namespace GroupConvolutionDWConcat
 
@@ -261,7 +261,7 @@ const auto params2D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
 
 /* ============= ConvolutionBackpropData (3D) ============= */
 const std::vector<CPUSpecificParams> CPUParams3D = {
@@ -278,7 +278,7 @@ const auto params3D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D, ConvConcatSubgraphTest, params3D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData3D, ConvConcatSubgraphTest, params3D, ConvConcatSubgraphTest::getTestCaseName);
 
 }  // namespace ConvolutionBackpropDataConcat
 
@@ -301,7 +301,7 @@ const auto params2D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
 
 /* ============= Convolution (3D) ============= */
 const std::vector<CPUSpecificParams> CPUParams3D = {
@@ -319,7 +319,7 @@ const auto params3D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution3D, ConvConcatSubgraphTest, params3D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution3D, ConvConcatSubgraphTest, params3D, ConvConcatSubgraphTest::getTestCaseName);
 
 }  // namespace ConvolutionConact
 
@@ -342,7 +342,7 @@ const auto params2D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolution2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
 
 /* ============= GroupConvolution (3D) ============= */
 const std::vector<CPUSpecificParams> CPUParams3D = {
@@ -360,7 +360,7 @@ const auto params3D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution3D, ConvConcatSubgraphTest, params3D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolution3D, ConvConcatSubgraphTest, params3D, ConvConcatSubgraphTest::getTestCaseName);
 
 }  // namespace GroupConvolutionConcat
 
@@ -382,7 +382,7 @@ const auto params2D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvolutionBackpropData2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolutionBackpropData2D, ConvConcatSubgraphTest, params2D, ConvConcatSubgraphTest::getTestCaseName);
 
 /* ============= GroupConvolutionBackpropData (3D) ============= */
 const std::vector<CPUSpecificParams> CPUParams3D = {
@@ -399,7 +399,7 @@ const auto params3D = ::testing::Combine(
     ::testing::Values(axis)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvolutionBackpropData3D, ConvConcatSubgraphTest, params3D, ConvConcatSubgraphTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolutionBackpropData3D, ConvConcatSubgraphTest, params3D, ConvConcatSubgraphTest::getTestCaseName);
 
 }  // namespace GroupConvolutionBackpropDataConcat
 
diff --git a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv_maxpool_activ.cpp b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv_maxpool_activ.cpp
index 632fc25d1dd2a8..bc8600fd853f16 100644
--- a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv_maxpool_activ.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/conv_maxpool_activ.cpp
@@ -78,7 +78,7 @@ const std::vector<fusingSpecificParams> fusingParamsSet {
         fusingSigmoid
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Check, ConvPoolActivTest, ::testing::ValuesIn(fusingParamsSet), ConvPoolActivTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Check, ConvPoolActivTest, ::testing::ValuesIn(fusingParamsSet), ConvPoolActivTest::getTestCaseName);
 
 } // namespace
 
diff --git a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/eltwise_chain.cpp b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/eltwise_chain.cpp
index 25c109b5c051c9..8179b0867b4cde 100644
--- a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/eltwise_chain.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/eltwise_chain.cpp
@@ -140,7 +140,7 @@ std::vector<std::vector<EltwiseTypes>> eltwiseOps = {
         { EltwiseTypes::DIVIDE, EltwiseTypes::SQUARED_DIFF, EltwiseTypes::ADD },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseChain, EltwiseChainTest,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseChain, EltwiseChainTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inputShapes),
                                 ::testing::ValuesIn(inputPrecisions),
@@ -172,7 +172,7 @@ std::vector<std::vector<InferenceEngine::Precision>> inputPrecisionsFQ {
         { Precision::FP32, Precision::FP32, Precision::FP32, Precision::FP32 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseChainWithFQ, EltwiseChainTest,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseChainWithFQ, EltwiseChainTest,
                     ::testing::Combine(
                             ::testing::ValuesIn(inputShapesFQ),
                             ::testing::ValuesIn(inputPrecisionsFQ),
diff --git a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_muladd_ewsimple.cpp b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_muladd_ewsimple.cpp
index c48947dfa707a7..76e9e453ce40c1 100644
--- a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_muladd_ewsimple.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_muladd_ewsimple.cpp
@@ -59,7 +59,7 @@ TEST_P(FuseMulAddAndEwSimpleTest1, CompareWithRefs) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, FuseMulAddAndEwSimpleTest1, mulAddAndEwSimpleCommonParams, FuseMulAddAndEwSimpleTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, FuseMulAddAndEwSimpleTest1, mulAddAndEwSimpleCommonParams, FuseMulAddAndEwSimpleTest::getTestCaseName);
 
 
 // Fused EltwiseAndSimple comes on the 2nd input into MulAdd
@@ -84,7 +84,7 @@ TEST_P(FuseMulAddAndEwSimpleTest2, CompareWithRefs) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, FuseMulAddAndEwSimpleTest2, mulAddAndEwSimpleCommonParams, FuseMulAddAndEwSimpleTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, FuseMulAddAndEwSimpleTest2, mulAddAndEwSimpleCommonParams, FuseMulAddAndEwSimpleTest::getTestCaseName);
 
 
 // Fused MulAdd with more than 3 inputs
@@ -108,5 +108,5 @@ TEST_P(FuseMulAddAndEwSimpleTest3, CompareWithRefs) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, FuseMulAddAndEwSimpleTest3, mulAddAndEwSimpleCommonParams, FuseMulAddAndEwSimpleTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, FuseMulAddAndEwSimpleTest3, mulAddAndEwSimpleCommonParams, FuseMulAddAndEwSimpleTest::getTestCaseName);
 }  // namespace SubgraphTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_scaleshift_and_fakequantize.cpp b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_scaleshift_and_fakequantize.cpp
index 871d93ab68860f..7c2dadd9f5262f 100644
--- a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_scaleshift_and_fakequantize.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_scaleshift_and_fakequantize.cpp
@@ -109,7 +109,7 @@ std::vector<std::vector<std::vector<float>>> quantizes {
     { {0.f}, {2.55f}, {0.f}, {2.55f} },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FuseScaleShiftAndFakeQuantize, FuseScaleShiftAndFakeQuantizeTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FuseScaleShiftAndFakeQuantize, FuseScaleShiftAndFakeQuantizeTest,
     ::testing::Combine(
         ::testing::ValuesIn(inputShapes),
         ::testing::Values(element::f32),
diff --git a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_transpose_reorder.cpp b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_transpose_reorder.cpp
index 3c07843a39c7b1..3f8eab0dd936de 100644
--- a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_transpose_reorder.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/fuse_transpose_reorder.cpp
@@ -97,7 +97,7 @@ TEST_P(FuseTransposeAndReorderTest, CompareWithRefs) {
     CheckTransposeCount(0);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, FuseTransposeAndReorderTest, fuseTransposeAndReorderCommonParams, FuseTransposeAndReorderTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, FuseTransposeAndReorderTest, fuseTransposeAndReorderCommonParams, FuseTransposeAndReorderTest::getTestCaseName);
 
 
 /*  FuseTransposeAndReorderTest1 graph
@@ -174,7 +174,7 @@ TEST_P(FuseTransposeAndReorderTest1, DISABLED_CompareWithRefs) {
     CheckTransposeCount(2);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, FuseTransposeAndReorderTest1, fuseTransposeAndReorderCommonParams, FuseTransposeAndReorderTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, FuseTransposeAndReorderTest1, fuseTransposeAndReorderCommonParams, FuseTransposeAndReorderTest::getTestCaseName);
 
 
 /*  FuseTransposeAndReorderTest2 graph
@@ -236,6 +236,6 @@ TEST_P(FuseTransposeAndReorderTest2, CompareWithRefs) {
     CheckTransposeCount(1);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, FuseTransposeAndReorderTest2, fuseTransposeAndReorderCommonParams, FuseTransposeAndReorderTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, FuseTransposeAndReorderTest2, fuseTransposeAndReorderCommonParams, FuseTransposeAndReorderTest::getTestCaseName);
 
 }  // namespace SubgraphTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/param_result_custom_blob.cpp b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/param_result_custom_blob.cpp
index 85d7e4787cec82..020bcac57565ba 100644
--- a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/param_result_custom_blob.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/param_result_custom_blob.cpp
@@ -53,7 +53,7 @@ TEST_P(ParameterResultCustomBlobTest, CompareWithRefs) {
     Run();
 }
 namespace {
-    INSTANTIATE_TEST_CASE_P(smoke_Check_Custom_Blob, ParameterResultCustomBlobTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Check_Custom_Blob, ParameterResultCustomBlobTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU),
                             ParameterResultSubgraphTest::getTestCaseName);
 } // namespace
@@ -79,7 +79,7 @@ TEST_P(ParameterResultSameBlobTest, CompareWithRefs) {
     Run();
 }
 namespace {
-    INSTANTIATE_TEST_CASE_P(smoke_Check_Same_Blob, ParameterResultSameBlobTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Check_Same_Blob, ParameterResultSameBlobTest,
                             ::testing::Values(CommonTestUtils::DEVICE_CPU),
                             ParameterResultSubgraphTest::getTestCaseName);
 } // namespace
diff --git a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/reshape_fc.cpp b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/reshape_fc.cpp
index 1ed41cb5b2b996..50a6218f63f9e3 100644
--- a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/reshape_fc.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/reshape_fc.cpp
@@ -91,7 +91,7 @@ const auto reshapeFCParams = ::testing::Combine(::testing::ValuesIn(isFC),
                                                 ::testing::ValuesIn(transpose),
                                                 ::testing::ValuesIn(fusingParamsSet));
 
-INSTANTIATE_TEST_CASE_P(smoke_Check, ReshapeFCTest, reshapeFCParams, ReshapeFCTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Check, ReshapeFCTest, reshapeFCParams, ReshapeFCTest::getTestCaseName);
 
 } // namespace
 
diff --git a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/reshape_permute_conv_permute_reshape_act.cpp b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/reshape_permute_conv_permute_reshape_act.cpp
index b389e2bc3fb938..a54bd62b4297e1 100644
--- a/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/reshape_permute_conv_permute_reshape_act.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/subgraph_tests/src/reshape_permute_conv_permute_reshape_act.cpp
@@ -32,7 +32,7 @@ std::vector<InferenceEngine::Precision> netPrecisions = {
 std::map<std::string, std::string> additional_config = { };
 
 namespace SubgraphTestsDefinitions {
-    INSTANTIATE_TEST_CASE_P(smoke_basic, ConvReshapeAct,
+    INSTANTIATE_TEST_SUITE_P(smoke_basic, ConvReshapeAct,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_CPU),
diff --git a/inference-engine/tests/functional/plugin/gna/Import_export_tests/import_export_act_conv_act.cpp b/inference-engine/tests/functional/plugin/gna/Import_export_tests/import_export_act_conv_act.cpp
index e5abadfb1cc67f..18feccd8cbc98f 100644
--- a/inference-engine/tests/functional/plugin/gna/Import_export_tests/import_export_act_conv_act.cpp
+++ b/inference-engine/tests/functional/plugin/gna/Import_export_tests/import_export_act_conv_act.cpp
@@ -165,7 +165,7 @@ const std::vector<std::map<std::string, std::string>> importConfigs = {
         }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ImportActConvAct, ImportActConvActTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ImportActConvAct, ImportActConvActTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inputShape),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/Import_export_tests/import_export_memory_layer.cpp b/inference-engine/tests/functional/plugin/gna/Import_export_tests/import_export_memory_layer.cpp
index c3f7a0c78ae176..f0deafaf07140d 100644
--- a/inference-engine/tests/functional/plugin/gna/Import_export_tests/import_export_memory_layer.cpp
+++ b/inference-engine/tests/functional/plugin/gna/Import_export_tests/import_export_memory_layer.cpp
@@ -130,7 +130,7 @@ const std::vector<std::map<std::string, std::string>> importConfigs = {
         },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ImportNetworkMemoryCase, ImportMemoryTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ImportNetworkMemoryCase, ImportMemoryTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/4d_eltwise.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/4d_eltwise.cpp
index 37cb4a825690bd..00f32649eb9229 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/4d_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/4d_eltwise.cpp
@@ -157,7 +157,7 @@ class Eltwise4dMultipleInput : public testing::WithParamInterface<eltwiseParams>
         ngraph::helpers::EltwiseTypes::ADD
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_Eltwise4d, Eltwise4dBroadcast,
+    INSTANTIATE_TEST_SUITE_P(smoke_Eltwise4d, Eltwise4dBroadcast,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -165,7 +165,7 @@ class Eltwise4dMultipleInput : public testing::WithParamInterface<eltwiseParams>
             ::testing::ValuesIn(eltwiseOpTypes)),
         Eltwise4dBroadcast::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Eltwise4d, Eltwise4dMultipleInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Eltwise4d, Eltwise4dMultipleInput,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/act_maxpool_reordering.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/act_maxpool_reordering.cpp
index 36b7d209aa2ffe..eea2de7d6e4d2c 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/act_maxpool_reordering.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/act_maxpool_reordering.cpp
@@ -125,7 +125,7 @@ const std::vector<ngraph::helpers::ActivationTypes> activationTypes = {
     ngraph::helpers::ActivationTypes::Abs
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_act_maxpool_reordering, ActMaxpoolReordering,
+INSTANTIATE_TEST_SUITE_P(smoke_act_maxpool_reordering, ActMaxpoolReordering,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/broadcast_const_with_fq.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/broadcast_const_with_fq.cpp
index 7c5f24cf3e84e9..085d69999f2bee 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/broadcast_const_with_fq.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/broadcast_const_with_fq.cpp
@@ -80,7 +80,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     { {"GNA_DEVICE_MODE", "GNA_SW_EXACT"} }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_broadcast_const_with_fq, BroadcastConstWithFq,
+INSTANTIATE_TEST_SUITE_P(smoke_broadcast_const_with_fq, BroadcastConstWithFq,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::ValuesIn(inputShapes1),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_fullyconnected.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_fullyconnected.cpp
index f4a74fe58c8b11..4cd9a227cbf105 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_fullyconnected.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_fullyconnected.cpp
@@ -103,7 +103,7 @@ const std::vector<std::vector<std::vector<size_t>>> input_shapes = {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_convert_matmul_to_fc, ConvertMatmulToFcPass,
+INSTANTIATE_TEST_SUITE_P(smoke_convert_matmul_to_fc, ConvertMatmulToFcPass,
                         ::testing::Combine(
                                 ::testing::ValuesIn(input_shapes),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp
index 86bfe5e25c013a..2065ed394d2706 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/convert_matmul_to_pointwise_conv.cpp
@@ -209,7 +209,7 @@ const std::vector<std::pair<float, float>> fqStats = {
     {-0.5, 0.5}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvertMatmulToPointwiseConvTest, ConvertMatmulToPointwiseConv,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvertMatmulToPointwiseConvTest, ConvertMatmulToPointwiseConv,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -217,7 +217,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvertMatmulToPointwiseConvTest, ConvertMatmulToP
         ::testing::ValuesIn(inputShape)),
     ConvertMatmulToPointwiseConv::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvertMatmulToPointwiseConvTest, ConvertMatmulToPointwiseConvWithFq,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvertMatmulToPointwiseConvTest, ConvertMatmulToPointwiseConvWithFq,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/eltwise_split_over_channels_pass.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/eltwise_split_over_channels_pass.cpp
index f4c6cc98d34136..5f69ab0261521c 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/eltwise_split_over_channels_pass.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/eltwise_split_over_channels_pass.cpp
@@ -80,7 +80,7 @@ const std::vector<std::vector<size_t>> inputShape = {
     {1, 500000}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseSplitOverChennels, EltwiseSplitOverChannelsPassTest,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseSplitOverChennels, EltwiseSplitOverChannelsPassTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/fq_activation.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/fq_activation.cpp
index ff0495f756eb12..9299fceb643527 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/fq_activation.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/fq_activation.cpp
@@ -122,7 +122,7 @@ const std::vector<size_t> levels = {
     65535,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_fq_activation, FQActivation,
+INSTANTIATE_TEST_SUITE_P(smoke_fq_activation, FQActivation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/fq_fusion_with_multiple_weights.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/fq_fusion_with_multiple_weights.cpp
index d65a94eaca6e5c..a4a26ba18d043f 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/fq_fusion_with_multiple_weights.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/fq_fusion_with_multiple_weights.cpp
@@ -129,7 +129,7 @@ const std::vector<size_t> levels = {
     65535,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_fq_fusion, FQFusionWithMultipleWeights,
+INSTANTIATE_TEST_SUITE_P(smoke_fq_fusion, FQFusionWithMultipleWeights,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/fq_maxpool_reordering.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/fq_maxpool_reordering.cpp
index 316df2ca9d7665..a8c5e52d30d4e4 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/fq_maxpool_reordering.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/fq_maxpool_reordering.cpp
@@ -136,7 +136,7 @@ const std::vector<size_t> levels = {
     65535,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_fq_maxpool_reordering, FQMaxpoolReordering,
+INSTANTIATE_TEST_SUITE_P(smoke_fq_maxpool_reordering, FQMaxpoolReordering,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/insert_copy_layer_before_self_concat.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/insert_copy_layer_before_self_concat.cpp
index 5c01537be7dc36..421280a7189cc5 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/insert_copy_layer_before_self_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/insert_copy_layer_before_self_concat.cpp
@@ -103,7 +103,7 @@ std::vector<size_t> concatsNum = {1, 2, 3, 4};
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                          InferenceEngine::Precision::FP16};
 
-INSTANTIATE_TEST_CASE_P(smoke_InsertCopy, InsertCopyBeforeSelfConcatTest,
+INSTANTIATE_TEST_SUITE_P(smoke_InsertCopy, InsertCopyBeforeSelfConcatTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(axes),
                                 ::testing::ValuesIn(inShapes),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/insert_transpose_before_matmul.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/insert_transpose_before_matmul.cpp
index eb867e05e31000..2c4298c9e97d74 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/insert_transpose_before_matmul.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/insert_transpose_before_matmul.cpp
@@ -120,7 +120,7 @@ const std::vector<size_t> inputShape = {
 
 const std::vector<bool> firstInputConst = {false, true};
 
-INSTANTIATE_TEST_CASE_P(smoke_InsertTransposeBeforeMatmulTest, InsertTransposeBeforeMatmul,
+INSTANTIATE_TEST_SUITE_P(smoke_InsertTransposeBeforeMatmulTest, InsertTransposeBeforeMatmul,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/insert_transpose_between_convs.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/insert_transpose_between_convs.cpp
index 6e63962504c41c..578eecb33ac3d9 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/insert_transpose_between_convs.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/insert_transpose_between_convs.cpp
@@ -209,7 +209,7 @@ const std::vector<std::vector<size_t>> inputShape = {
     {1, 4, 1, 128}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_InsertTransposeBetweenConvsTest, InsertTransposeBetweenConvs,
+INSTANTIATE_TEST_SUITE_P(smoke_InsertTransposeBetweenConvsTest, InsertTransposeBetweenConvs,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -217,7 +217,7 @@ INSTANTIATE_TEST_CASE_P(smoke_InsertTransposeBetweenConvsTest, InsertTransposeBe
         ::testing::ValuesIn(inputShape)),
     InsertTransposeBetweenConvs::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_InsertTransposeBetweenConvsWithPoolTest, InsertTransposeBetweenConvsWithPool,
+INSTANTIATE_TEST_SUITE_P(smoke_InsertTransposeBetweenConvsWithPoolTest, InsertTransposeBetweenConvsWithPool,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/layers_restrictions.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/layers_restrictions.cpp
index fc7f15994d2044..15d2075f5baac3 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/layers_restrictions.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/layers_restrictions.cpp
@@ -100,14 +100,14 @@ const std::vector<std::map<std::string, std::string>> configs = {
     { {"GNA_DEVICE_MODE", "GNA_SW_EXACT"} }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_layers_restrictions, LayersRestrictionsFullyConnectedBatchSizeMoreThan8,
+INSTANTIATE_TEST_SUITE_P(smoke_layers_restrictions, LayersRestrictionsFullyConnectedBatchSizeMoreThan8,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::ValuesIn(configs),
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA)),
                         LayersRestrictionsFullyConnectedBatchSizeMoreThan8::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_layers_restrictions, LayersRestrictionsFullyConnectedBatchSizeLessThanOrEqual8,
+INSTANTIATE_TEST_SUITE_P(smoke_layers_restrictions, LayersRestrictionsFullyConnectedBatchSizeLessThanOrEqual8,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::ValuesIn(configs),
diff --git a/inference-engine/tests/functional/plugin/gna/pass_tests/remove_permutations_NHWC_to_NCHW_pass.cpp b/inference-engine/tests/functional/plugin/gna/pass_tests/remove_permutations_NHWC_to_NCHW_pass.cpp
index d5cd3fc58f8d6d..326d2cbb8ef949 100644
--- a/inference-engine/tests/functional/plugin/gna/pass_tests/remove_permutations_NHWC_to_NCHW_pass.cpp
+++ b/inference-engine/tests/functional/plugin/gna/pass_tests/remove_permutations_NHWC_to_NCHW_pass.cpp
@@ -471,7 +471,7 @@ class RemovePermutationsWithEltwiseTest : public testing::WithParamInterface<rem
         {1, 32, 1, 8}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_PermutationPass, RemovePermutationsNHWCToNCHWPassTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_PermutationPass, RemovePermutationsNHWCToNCHWPassTest,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -479,7 +479,7 @@ class RemovePermutationsWithEltwiseTest : public testing::WithParamInterface<rem
             ::testing::ValuesIn(inputShapes)),
         RemovePermutationsNHWCToNCHWPassTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_PermutationPass, RemovePermutationsNHWCToNCHWPass4DOutputTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_PermutationPass, RemovePermutationsNHWCToNCHWPass4DOutputTest,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -487,7 +487,7 @@ class RemovePermutationsWithEltwiseTest : public testing::WithParamInterface<rem
             ::testing::ValuesIn(inputShapes)),
         RemovePermutationsNHWCToNCHWPass4DOutputTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_PermutationPass, RemovePermutationsWithPoolAndActTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_PermutationPass, RemovePermutationsWithPoolAndActTest,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -496,7 +496,7 @@ class RemovePermutationsWithEltwiseTest : public testing::WithParamInterface<rem
             ::testing::ValuesIn(std::vector<bool>{false, true})), // with activation
         RemovePermutationsWithPoolAndActTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_PermutationPass, RemovePermutationsWithTwoConvTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_PermutationPass, RemovePermutationsWithTwoConvTest,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -504,7 +504,7 @@ class RemovePermutationsWithEltwiseTest : public testing::WithParamInterface<rem
             ::testing::ValuesIn(inputShapes)),
         RemovePermutationsWithTwoConvTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_PermutationPass, RemovePermutationsWithEltwiseTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_PermutationPass, RemovePermutationsWithEltwiseTest,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/add_output.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/add_output.cpp
index 15c35dfa433c91..ad8c4b2bc78011 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/add_output.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/add_output.cpp
@@ -89,6 +89,6 @@ std::vector<addOutputsParams> testCases = {
         addOutputsParams(getTargetNetwork(), {"Memory_1"}, CommonTestUtils::DEVICE_GNA)
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_AddOutputBasic, AddOutputsTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AddOutputBasic, AddOutputsTest,
         ::testing::ValuesIn(testCases),
         AddOutputsTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/caching_tests.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/caching_tests.cpp
index 5d956898897025..620e6285cd9584 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/caching_tests.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/caching_tests.cpp
@@ -17,7 +17,7 @@ namespace {
             1, 2
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_CachingSupportCase_GNA, LoadNetworkCacheTestBase,
+    INSTANTIATE_TEST_SUITE_P(smoke_CachingSupportCase_GNA, LoadNetworkCacheTestBase,
                             ::testing::Combine(
                                     ::testing::ValuesIn(LoadNetworkCacheTestBase::getStandardFunctions()),
                                     ::testing::ValuesIn(precisionsGNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/config.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/config.cpp
index 2c74871bd324ea..13b8f62c966092 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/config.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/config.cpp
@@ -22,7 +22,7 @@ namespace {
             {{InferenceEngine::GNAConfigParams::KEY_GNA_COMPACT_MODE, "ON"}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -38,7 +38,7 @@ namespace {
     };
 
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -51,7 +51,7 @@ namespace {
             {}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectConfigAPITests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/core_integration.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/core_integration.cpp
index 52efa7af27bea1..c2b9a62b0492d9 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/core_integration.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/core_integration.cpp
@@ -13,12 +13,12 @@ namespace {
 // IE Class Common tests with <pluginName, deviceName params>
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassBasicTestP, IEClassBasicTestP,
         ::testing::Values(std::make_pair("GNAPlugin", "GNA")));
 
 // TODO
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassNetworkTestP, IEClassNetworkTestP,
         ::testing::Values("GNA"));
 
@@ -26,41 +26,41 @@ INSTANTIATE_TEST_CASE_P(
 // IE Class GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values("GNA", "MULTI", "HETERO"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values("GNA", "MULTI", "HETERO"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_AVAILABLE_DEVICES,
         ::testing::Values("GNA"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_FULL_DEVICE_NAME,
         ::testing::Values("GNA", "MULTI", "HETERO"));
 
 // TODO: Issue: 30198
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassGetMetricTest, IEClassGetMetricTest_OPTIMIZATION_CAPABILITIES,
         ::testing::Values("GNA"));
 
 // TODO: Issue: 30199
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassGetMetricTest, IEClassGetMetricTest_RANGE_FOR_ASYNC_INFER_REQUESTS,
         ::testing::Values("GNA"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_ThrowUnsupported,
         ::testing::Values("GNA", "MULTI", "HETERO"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetConfigTest, IEClassGetConfigTest_ThrowUnsupported,
         ::testing::Values("GNA", "MULTI", "HETERO"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetAvailableDevices, IEClassGetAvailableDevices,
         ::testing::Values("GNA"));
 
@@ -68,7 +68,7 @@ INSTANTIATE_TEST_CASE_P(
 // IE Class GetConfig
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetConfigTest, IEClassGetConfigTest,
         ::testing::Values("GNA"));
 
@@ -77,27 +77,27 @@ INSTANTIATE_TEST_CASE_P(
 //
 
 // TODO: Convolution with 3D input is not supported on GNA
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values("GNA" /*, "MULTI:GNA", "HETERO:GNA" */));
 
 // TODO: Convolution with 3D input is not supported on GNA
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values("GNA" /*, "MULTI:GNA",  "HETERO:GNA" */));
 
 // TODO: this metric is not supported by the plugin
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_NETWORK_NAME,
         ::testing::Values("GNA", "MULTI:GNA", "HETERO:GNA"));
 
 // TODO: Convolution with 3D input is not supported on GNA
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_OPTIMAL_NUMBER_OF_INFER_REQUESTS,
         ::testing::Values("GNA"/*, "MULTI:GNA", "HETERO:GNA" */));
 
 // TODO: Convolution with 3D input is not supported on GNA
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_ThrowsUnsupported,
         ::testing::Values("GNA", /* "MULTI:GNA", */ "HETERO:GNA"));
 
@@ -106,17 +106,17 @@ INSTANTIATE_TEST_CASE_P(
 //
 
 // TODO: Convolution with 3D input is not supported on GNA
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassExecutableNetworkGetConfigTest, IEClassExecutableNetworkGetConfigTest,
         ::testing::Values("GNA"));
 
 // TODO: Convolution with 3D input is not supported on GNA
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassExecutableNetworkSetConfigTest, IEClassExecutableNetworkSetConfigTest,
         ::testing::Values("GNA"));
 
 // TODO: Convolution with 3D input is not supported on GNA
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassExecutableNetworkSupportedConfigTest, IEClassExecutableNetworkSupportedConfigTest,
         ::testing::Combine(::testing::Values("GNA"),
                            ::testing::Values(std::make_pair(GNA_CONFIG_KEY(DEVICE_MODE), GNAConfigParams::GNA_HW),
@@ -125,7 +125,7 @@ INSTANTIATE_TEST_CASE_P(
                                              std::make_pair(GNA_CONFIG_KEY(DEVICE_MODE), GNAConfigParams::GNA_AUTO))));
 
 // TODO: Convolution with 3D input is not supported on GNA
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassExecutableNetworkUnsupportedConfigTest, IEClassExecutableNetworkUnsupportedConfigTest,
         ::testing::Combine(::testing::Values("GNA"),
                            ::testing::Values(std::make_pair(GNA_CONFIG_KEY(DEVICE_MODE), GNAConfigParams::GNA_SW_FP32),
@@ -146,7 +146,7 @@ TEST_P(IEClassExecutableNetworkSetConfigFromFp32Test, SetConfigFromFp32Throws) {
 }
 
 // TODO: Convolution with 3D input is not supported on GNA
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassExecutableNetworkSetConfigFromFp32Test, IEClassExecutableNetworkSetConfigFromFp32Test,
         ::testing::Combine(::testing::Values("GNA"),
                            ::testing::Values(std::make_pair(GNA_CONFIG_KEY(DEVICE_MODE), GNAConfigParams::GNA_HW),
@@ -157,13 +157,13 @@ INSTANTIATE_TEST_CASE_P(
 
 // IE Class Query network
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassQueryNetworkTest, IEClassQueryNetworkTest,
         ::testing::Values("GNA"));
 
 // IE Class Load network
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassLoadNetworkTest, IEClassLoadNetworkTest,
         ::testing::Values("GNA"));
 
@@ -172,21 +172,21 @@ INSTANTIATE_TEST_CASE_P(
 //
 
 // TODO: verify hetero interop
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values("GNA"));
 
 // TODO: verify hetero interop
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values("GNA"));
 
 // TODO: verify hetero interop
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_smoke_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_NETWORK_NAME,
         ::testing::Values("GNA"));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_TARGET_FALLBACK,
         ::testing::Values("GNA"));
 } // namespace
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/core_threading_tests.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/core_threading_tests.cpp
index f1cb21ac6b5f0e..690ec0a0bfb688 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/core_threading_tests.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/core_threading_tests.cpp
@@ -14,9 +14,9 @@ Params params[] = {
 std::vector< std::tuple<Device, Config> > paramsWithIterations{ params[0], params[2] };
 }  // namespace
 
-INSTANTIATE_TEST_CASE_P(GNA, CoreThreadingTests, testing::ValuesIn(params), CoreThreadingTests::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(GNA, CoreThreadingTests, testing::ValuesIn(params), CoreThreadingTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(GNA, CoreThreadingTestsWithIterations,
+INSTANTIATE_TEST_SUITE_P(GNA, CoreThreadingTestsWithIterations,
     testing::Combine(testing::ValuesIn(paramsWithIterations),
                      testing::Values(3),
                      testing::Values(4),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/cpp_holders.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/cpp_holders.cpp
index 433d4adf9336cc..38ac8bf5f13ab6 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/cpp_holders.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/cpp_holders.cpp
@@ -21,19 +21,19 @@ namespace {
             {0, 1, 3, 2}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTest,
             ::testing::Combine(
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
             ::testing::ValuesIn(orders)),
             HoldersTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTestImportNetwork,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTestImportNetwork,
             ::testing::Combine(
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
             ::testing::ValuesIn(orders)),
             HoldersTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTestOnImportedNetwork,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTestOnImportedNetwork,
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
             HoldersTestOnImportedNetwork::getTestCaseName);
 
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/exec_graph_info.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/exec_graph_info.cpp
index d8cb26b4499588..84b86f47f0477a 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/exec_graph_info.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/exec_graph_info.cpp
@@ -15,7 +15,7 @@ namespace {
             {},
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, ExecGraphTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, ExecGraphTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request.cpp
index 9d67255fafc429..055c23f047a1de 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request.cpp
@@ -10,14 +10,14 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP32
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestTests,
         ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_GNA),
                 ::testing::Values(std::map<std::string, std::string>({}))),
                 InferRequestTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestTestsResultNotReady,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestTestsResultNotReady,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_callback.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_callback.cpp
index 6b7c64fe39f8b1..9336cb7efa315e 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_callback.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_callback.cpp
@@ -20,14 +20,14 @@ const std::vector<std::map<std::string, std::string>> multiConfigs = {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CallbackTests,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
             ::testing::ValuesIn(configs)),
         CallbackTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, CallbackTests,
         ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_MULTI),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_config.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_config.cpp
index 62d2a94ffa81da..c21747bd33506e 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_config.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_config.cpp
@@ -29,14 +29,14 @@ namespace {
             {{InferenceEngine::GNAConfigParams::KEY_GNA_COMPACT_MODE, InferenceEngine::PluginConfigParams::NO}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferConfigTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
                                     ::testing::ValuesIn(configs)),
                             InferConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferConfigInTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferConfigInTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_input.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_input.cpp
index 81620660e0f250..6aa0825d34df4b 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_input.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_input.cpp
@@ -15,7 +15,7 @@ namespace {
             {}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestInputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestInputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_output.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_output.cpp
index da5b0b42611ae2..a8c4ed86f3e245 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_output.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/infer_request_output.cpp
@@ -14,7 +14,7 @@ namespace {
             {}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestOutputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestOutputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/layout.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/layout.cpp
index 2d5f773bf41e83..ccbab8aecb4891 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/layout.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/layout.cpp
@@ -22,7 +22,7 @@ namespace {
             { 1, 3 }
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, LayoutTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, LayoutTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/memory_states.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/memory_states.cpp
index 475b7693c41ba7..c2db36ff9f3ed7 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/memory_states.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/memory_states.cpp
@@ -127,6 +127,6 @@ std::vector<memoryStateParams> memoryStateTestCases = {
         memoryStateParams(getNetwork(), {"c_1-3", "r_1-3"}, CommonTestUtils::DEVICE_GNA)
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_VariableStateBasic, VariableStateTest,
+INSTANTIATE_TEST_SUITE_P(smoke_VariableStateBasic, VariableStateTest,
         ::testing::ValuesIn(memoryStateTestCases),
         VariableStateTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/test_plugin.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/test_plugin.cpp
index 4ee7a5a8761abe..8c5e79825f55f1 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/test_plugin.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/test_plugin.cpp
@@ -16,21 +16,21 @@ namespace {
             {}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTests,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
                                     ::testing::ValuesIn(configs)),
                             BehaviorTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTestInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
                                     ::testing::ValuesIn(configs)),
                             BehaviorTestInput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTestOutput,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/version.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/version.cpp
index 74850f0129ba6e..e9c9c12d292b8e 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/version.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/behavior/version.cpp
@@ -14,14 +14,14 @@ namespace {
             {{ HETERO_CONFIG_KEY(DUMP_GRAPH_DOT) , CommonTestUtils::DEVICE_GNA}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_GNA),
                                     ::testing::ValuesIn(configs)),
                             VersionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Hetero_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Hetero_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_HETERO),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/import_export_tests/import_reshape_permute_conv.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/import_export_tests/import_reshape_permute_conv.cpp
index 442cb116f5d387..0d824402305da1 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/import_export_tests/import_reshape_permute_conv.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/import_export_tests/import_reshape_permute_conv.cpp
@@ -74,7 +74,7 @@ const std::vector<std::string> appHeaders = {
         "APPLICATION_HEADER"
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ImportNetworkCase, ImportReshapePermuteConvGNA,
+INSTANTIATE_TEST_SUITE_P(smoke_ImportNetworkCase, ImportReshapePermuteConvGNA,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/activation.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/activation.cpp
index 8e800d1b1c86c2..9de08e5f84ffd8 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/activation.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/activation.cpp
@@ -61,6 +61,6 @@ const auto basicCases = ::testing::Combine(
 );
 
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/concat.cpp
index 022898dd3c7bac..68107baee29bd3 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/concat.cpp
@@ -22,7 +22,7 @@ std::vector<std::vector<std::vector<size_t>>> inShapes = {
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                          InferenceEngine::Precision::FP16};
 // TODO: Issue:  26421
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, ConcatLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, ConcatLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(axes),
                                 ::testing::ValuesIn(inShapes),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution.cpp
index 6605857612aa49..7231ec975f4efe 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution.cpp
@@ -147,7 +147,7 @@ const auto conv2DParams_AutoPadValid_MapTo1d = ::testing::Combine(
 );
 
 // TODO: padding isn't currently supported in GNA
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_Convolution2D_ExplicitPadding_Height1, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_Convolution2D_ExplicitPadding_Height1, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding_Height1,
                                 ::testing::ValuesIn(netPrecisions),
@@ -159,7 +159,7 @@ INSTANTIATE_TEST_CASE_P(DISABLED_smoke_Convolution2D_ExplicitPadding_Height1, Co
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA)),
                         ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_Convolution2D_ExplicitPadding_Width1, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_Convolution2D_ExplicitPadding_Width1, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding_Width1,
                                 ::testing::ValuesIn(netPrecisions),
@@ -171,7 +171,7 @@ INSTANTIATE_TEST_CASE_P(DISABLED_smoke_Convolution2D_ExplicitPadding_Width1, Con
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA)),
                         ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_AutoPadValid_Height1, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D_AutoPadValid_Height1, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid_Height1,
                                 ::testing::ValuesIn(netPrecisions),
@@ -183,7 +183,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_AutoPadValid_Height1, ConvolutionLay
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA)),
                         ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_AutoPadValid_Width1, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D_AutoPadValid_Width1, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid_Width1,
                                 ::testing::ValuesIn(netPrecisions),
@@ -195,7 +195,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_AutoPadValid_Width1, ConvolutionLaye
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA)),
                         ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_AutoPadValid_MapTo1d, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D_AutoPadValid_MapTo1d, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid_MapTo1d,
                                 ::testing::ValuesIn(netPrecisions),
@@ -207,7 +207,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_AutoPadValid_MapTo1d, ConvolutionLay
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA)),
                         ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_Kernels2D, GnaConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D_Kernels2D, GnaConvolutionLayerTest,
     ::testing::Combine(
         conv2DParams_Kernels2D,
         ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution_negative.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution_negative.cpp
index aa4975e602facd..2235a2b90fa5d0 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution_negative.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/convolution_negative.cpp
@@ -151,7 +151,7 @@ struct GnaConv2DNegativeTest##whats_wrong : GnaConv2DNegativeTest {
 TEST_P(GnaConv2DNegativeTest##whats_wrong, ThrowAsNotSupported) {                                               \
     Run();                                                                                                      \
 }                                                                                                               \
-INSTANTIATE_TEST_CASE_P(smoke_GnaConv2DNegativeTestInvalid##whats_wrong, GnaConv2DNegativeTest##whats_wrong,    \
+INSTANTIATE_TEST_SUITE_P(smoke_GnaConv2DNegativeTestInvalid##whats_wrong, GnaConv2DNegativeTest##whats_wrong,    \
 ::testing::Combine(                                                                                             \
     conv2DParameters##suffix_params,                                                                            \
     ::testing::ValuesIn(netPrecisions),                                                                         \
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/eltwise.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/eltwise.cpp
index 13430d921e7b16..6c13b86e434dc7 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/eltwise.cpp
@@ -59,5 +59,5 @@ const auto multiply_params = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
         ::testing::Values(additional_config));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs, EltwiseLayerTest, multiply_params, EltwiseLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs, EltwiseLayerTest, multiply_params, EltwiseLayerTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/fake_quantize.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/fake_quantize.cpp
index 31ec8bc4a62cf8..4a50e05ce089d4 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/fake_quantize.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/fake_quantize.cpp
@@ -84,7 +84,7 @@ const auto fqParams = ::testing::Combine(
     ::testing::ValuesIn(inputParams)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantize, FakeQuantizeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantize, FakeQuantizeLayerTest,
     ::testing::Combine(
     fqParams,
     ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/low_precision.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/low_precision.cpp
index 04853e098605f5..9d68c1cea942a5 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/low_precision.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/low_precision.cpp
@@ -58,7 +58,7 @@ const std::vector<std::pair<std::string, std::map<std::string, std::string>>> co
     {"sw_exact_i8", config_i8},
 };
 
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_LowPrecision, GnaLowPrecisionTest,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_LowPrecision, GnaLowPrecisionTest,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/mat_mul.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/mat_mul.cpp
index f8ad9761689918..f502d0ed3cbd02 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/mat_mul.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/mat_mul.cpp
@@ -27,7 +27,7 @@ std::vector<ngraph::helpers::InputLayerType> secondaryInputTypes = {
 
 std::map<std::string, std::string> additional_config = {{"GNA_DEVICE_MODE", "GNA_SW_EXACT"}};
 
-INSTANTIATE_TEST_CASE_P(smoke_MatMul, MatMulTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MatMul, MatMulTest,
         ::testing::Combine(
                 ::testing::ValuesIn(shapeRelatedParams),
                 ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/memory.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/memory.cpp
index c04b76705ccac0..3e3c053f192c80 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/memory.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/memory.cpp
@@ -34,7 +34,7 @@ const std::vector<int64_t> iterationCount {
     10
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MemoryTest, MemoryTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MemoryTest, MemoryTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(transformation),
                                 ::testing::ValuesIn(iterationCount),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/pooling.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/pooling.cpp
index cd5af6385d2ac8..880a3f95bb6bc5 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/pooling.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/pooling.cpp
@@ -40,7 +40,7 @@ const auto maxPool_ExplicitPad_FloorRounding_Params = ::testing::Combine(
 );
 
 // TODO: Issue:  26503
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_MaxPool_ExplicitPad_FloorRpunding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_MaxPool_ExplicitPad_FloorRpunding, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_ExplicitPad_FloorRounding_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -65,7 +65,7 @@ const auto maxPool_ExplicitPad_CeilRounding_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 // TODO: Issue:  26503
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_MaxPool_ExplicitPad_CeilRpunding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_MaxPool_ExplicitPad_CeilRpunding, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_ExplicitPad_CeilRounding_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -93,7 +93,7 @@ const auto avgPoolExplicitPadCeilRoundingParams = ::testing::Combine(
         ::testing::Values(true, false)
 );
 // TODO: Issue:  26503
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_AvgPool_ExplicitPad_CeilRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_AvgPool_ExplicitPad_CeilRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 avgPoolExplicitPadCeilRoundingParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -118,7 +118,7 @@ const auto avgPoolExplicitPadFloorRoundingParams = ::testing::Combine(
 );
 
 // TODO: Issue:  26503
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_AvgPool_ExplicitPad_FloorRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_AvgPool_ExplicitPad_FloorRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 avgPoolExplicitPadFloorRoundingParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -144,7 +144,7 @@ const auto allPools_ValidPad_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 // TODO: Issue:  26503
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_MAX_and_AVGPool_ValidPad, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_MAX_and_AVGPool_ValidPad, PoolingLayerTest,
                         ::testing::Combine(
                                 allPools_ValidPad_Params,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/power.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/power.cpp
index 1c67b1b35ce791..892e96670f6b85 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/power.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/power.cpp
@@ -36,7 +36,7 @@ namespace {
                                                              InferenceEngine::Precision::FP16,
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_power, PowerLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_power, PowerLayerTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(inShapes),
                                     ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/split.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/split.cpp
index a2740dd8ea7867..80f784e5cf4568 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/split.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/split.cpp
@@ -16,7 +16,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 };
 
 // TODO: Issue:  26411
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_NumSplitsCheck, SplitLayerTest,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_NumSplitsCheck, SplitLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(1),
                                 ::testing::Values(0, 1),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/tensor_iterator.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/tensor_iterator.cpp
index 64cb0d1d7760d8..55e72e2b819834 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/tensor_iterator.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/tensor_iterator.cpp
@@ -21,7 +21,7 @@ namespace {
         InferenceEngine::Precision::FP16
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_TensorIterator, TensorIteratorTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_TensorIterator, TensorIteratorTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(should_decompose),
                                     ::testing::ValuesIn(seqLengths),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/activation_concats_eltwise.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/activation_concats_eltwise.cpp
index f08ef63d0b9842..e3c190cc176981 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/activation_concats_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/activation_concats_eltwise.cpp
@@ -31,7 +31,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 
 std::map<std::string, std::string>  additional_config = {};
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareRefs, ActivationConcatsEltwise,
+INSTANTIATE_TEST_SUITE_P(smoke_CompareRefs, ActivationConcatsEltwise,
                         ::testing::Combine(
                                 ::testing::ValuesIn(input_sizes),
                                 ::testing::ValuesIn(concat_const_sizes),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/activation_fq.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/activation_fq.cpp
index 223b243768b971..e3f51c28d76ed1 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/activation_fq.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/activation_fq.cpp
@@ -67,7 +67,7 @@ const std::vector<ngraph::helpers::ActivationTypes> activations = {
     ngraph::helpers::ActivationTypes::Exp
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ActivationFQSubgraph, ActivationFakeQuantizeSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ActivationFQSubgraph, ActivationFakeQuantizeSubgraphTest,
                         ::testing::Combine(
                                 fqParams,
                                 ::testing::ValuesIn(activations),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/basic_lstm.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/basic_lstm.cpp
index ff5633c0d865a1..5b6530090c0f3a 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/basic_lstm.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/basic_lstm.cpp
@@ -29,7 +29,7 @@ const std::vector<std::pair<size_t, size_t>> size_params = {
 
 const std::vector<bool> decompose = { false, true };
 
-INSTANTIATE_TEST_CASE_P(smoke_BasicLSTM, Basic_LSTM_S,
+INSTANTIATE_TEST_SUITE_P(smoke_BasicLSTM, Basic_LSTM_S,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/broadcast_power.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/broadcast_power.cpp
index 252e64cb2753c7..5cabf80e6add53 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/broadcast_power.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/broadcast_power.cpp
@@ -39,7 +39,7 @@ const std::vector<std::vector<std::vector<size_t>>> input_shapes {
 };
 
 
-INSTANTIATE_TEST_CASE_P(PowerBroadcast, BroadcastPowerTest,
+INSTANTIATE_TEST_SUITE_P(PowerBroadcast, BroadcastPowerTest,
         ::testing::Combine(
         ::testing::ValuesIn(input_shapes),
         ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/cascade_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/cascade_concat.cpp
index fc408535b6d235..f4a68244df0690 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/cascade_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/cascade_concat.cpp
@@ -43,7 +43,7 @@ namespace {
             {"GNA_SCALE_FACTOR_2", "1"}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_cascade_concat, CascadeConcat,
+    INSTANTIATE_TEST_SUITE_P(smoke_cascade_concat, CascadeConcat,
                             ::testing::Combine(
                                     ::testing::ValuesIn(shape1),
                                     ::testing::ValuesIn(shape2),
@@ -54,7 +54,7 @@ namespace {
                                     ::testing::Values(additional_config)),
                             CascadeConcat::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_cascade_concat_multioutput, CascadeConcat,
+    INSTANTIATE_TEST_SUITE_P(smoke_cascade_concat_multioutput, CascadeConcat,
                             ::testing::Combine(
                                     ::testing::ValuesIn(shape1),
                                     ::testing::ValuesIn(shape2),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/clamp_fq.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/clamp_fq.cpp
index 2a53f88a83aa5d..bf0a38ee4bc3ca 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/clamp_fq.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/clamp_fq.cpp
@@ -72,7 +72,7 @@ const auto fqParams = ::testing::Combine(
 );
 
 
-INSTANTIATE_TEST_CASE_P(smoke_Clamp_FQ_subgraph, ClampFakeQuantizeSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Clamp_FQ_subgraph, ClampFakeQuantizeSubgraphTest,
                         ::testing::Combine(
                                 fqParams,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_conv.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_conv.cpp
index 4b657aec21e3fe..2087e98dfe1954 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_conv.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_conv.cpp
@@ -39,7 +39,7 @@ std::vector<size_t> outputChannels = {
     8
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ConcatConvTest, ConcatConvTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConcatConvTest, ConcatConvTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_multi_input.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_multi_input.cpp
index c748d53a83fa3e..0a153590268bb2 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_multi_input.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_multi_input.cpp
@@ -32,7 +32,7 @@ std::map<std::string, std::string> additional_config = {
     {"GNA_PRECISION", "I16"},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_concat_multi_input, ConcatMultiInput,
+INSTANTIATE_TEST_SUITE_P(smoke_concat_multi_input, ConcatMultiInput,
     ::testing::Combine(
         ::testing::ValuesIn(inShapes),
         ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_quantization.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_quantization.cpp
index 01db9ae551f0ea..90e6961b404fc1 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_quantization.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_quantization.cpp
@@ -21,7 +21,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ConcatQuantization, ConcatQuantization,
+INSTANTIATE_TEST_SUITE_P(smoke_ConcatQuantization, ConcatQuantization,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_quantization_during_memory_requantization.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_quantization_during_memory_requantization.cpp
index 32df9109d6af06..0357c545f57872 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_quantization_during_memory_requantization.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/concat_quantization_during_memory_requantization.cpp
@@ -31,7 +31,7 @@ namespace {
         32
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_concat_quant_memory_requant, ConcatQuantDuringMemoryRequantTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_concat_quant_memory_requant, ConcatQuantDuringMemoryRequantTest,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/connect_split_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/connect_split_concat.cpp
index 47bb9adda01581..3322d3f0800ba6 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/connect_split_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/connect_split_concat.cpp
@@ -17,14 +17,14 @@ std::map<std::string, std::string> additional_config = {
         {"GNA_COMPACT_MODE", "NO"}
 };
 
-INSTANTIATE_TEST_CASE_P(multiple_connect_split_concat, MultipleConnectSplitConcatTest,
+INSTANTIATE_TEST_SUITE_P(multiple_connect_split_concat, MultipleConnectSplitConcatTest,
         ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
         ::testing::Values(additional_config)),
         MultipleConnectSplitConcatTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(connect_split_concat_concat, SplitConcatConcatTest,
+INSTANTIATE_TEST_SUITE_P(connect_split_concat_concat, SplitConcatConcatTest,
         ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/const_conv_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/const_conv_concat.cpp
index 48224037b2ae82..3a03f5c7cbc455 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/const_conv_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/const_conv_concat.cpp
@@ -44,7 +44,7 @@ std::vector<size_t> outputChannels = {
     64
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ConstConvConcatTest, ConstConvConcatTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConstConvConcatTest, ConstConvConcatTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/const_strided_slice_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/const_strided_slice_concat.cpp
index 06fec972d17231..35d42981e8fd85 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/const_strided_slice_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/const_strided_slice_concat.cpp
@@ -39,7 +39,7 @@ std::vector<uint32_t> constChunksNumber = {
     1, 3
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ConstStridedSliceConcatTest, ConstStridedSliceConcatTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConstStridedSliceConcatTest, ConstStridedSliceConcatTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/constant_result.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/constant_result.cpp
index 3c3c7978fe24bd..8c9f9ab1d03672 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/constant_result.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/constant_result.cpp
@@ -34,7 +34,7 @@ const std::vector<Precision> precisions = {
     Precision::BOOL
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Check, ConstantResultSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Check, ConstantResultSubgraphTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(types),
                             ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/conv_fq_eltwise.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/conv_fq_eltwise.cpp
index 0a09c33f6df8f6..f9358fee19fa28 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/conv_fq_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/conv_fq_eltwise.cpp
@@ -46,7 +46,7 @@ const auto convParams = ::testing::Combine(
         ::testing::ValuesIn(outputChannels)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvFqEltwiseTest, ConvFqEltwiseTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvFqEltwiseTest, ConvFqEltwiseTest,
                         ::testing::Combine(
                                 fqParams,
                                 convParams,
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/conv_fq_relu.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/conv_fq_relu.cpp
index 2a2296950e628d..ebb96d8aa17fb8 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/conv_fq_relu.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/conv_fq_relu.cpp
@@ -46,7 +46,7 @@ const auto convParams = ::testing::Combine(
         ::testing::ValuesIn(outputChannels)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvFqReluTest, ConvFqReluTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvFqReluTest, ConvFqReluTest,
                         ::testing::Combine(
                                 fqParams,
                                 convParams,
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/convolution_relu_sequence.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/convolution_relu_sequence.cpp
index 7600e34a7fb095..8169ac44bee6a3 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/convolution_relu_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/convolution_relu_sequence.cpp
@@ -190,7 +190,7 @@ const std::vector<std::map<std::string, std::string> > configs = {
 };
 
 // Enable when using GNA 2.1 library
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionReluSequenceTest, GnaConvolutionReluSequenceTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionReluSequenceTest, GnaConvolutionReluSequenceTest,
     ::testing::Combine(
         ::testing::ValuesIn(convReluSpecificParamsAllAll),
         ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/copy_before_squeeze.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/copy_before_squeeze.cpp
index 3e82d4f0e8b43e..6e3d94864708bb 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/copy_before_squeeze.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/copy_before_squeeze.cpp
@@ -26,7 +26,7 @@ namespace {
         {1, 640}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_copy_before_squeeze, CopyBeforeSqueezeTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_copy_before_squeeze, CopyBeforeSqueezeTest,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/crop4d.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/crop4d.cpp
index 7e390c381585a7..335c15ff2e8275 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/crop4d.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/crop4d.cpp
@@ -30,7 +30,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
         }
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_crop4d_gna, Crop4dTest,
         ::testing::Combine(
         ::testing::ValuesIn(ss_only_test_cases),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/delayed_copy_layer.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/delayed_copy_layer.cpp
index 41a8c1d68e0d7e..3f3cfa1f063ed4 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/delayed_copy_layer.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/delayed_copy_layer.cpp
@@ -17,7 +17,7 @@ namespace {
             {"GNA_COMPACT_MODE", "NO"}
     };
 
-    INSTANTIATE_TEST_CASE_P(delayed_copy_layer, DelayedCopyTest,
+    INSTANTIATE_TEST_SUITE_P(delayed_copy_layer, DelayedCopyTest,
                             ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/eltwise_conv_eltwise.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/eltwise_conv_eltwise.cpp
index bacdd0d90061bc..ae29ad27baed8e 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/eltwise_conv_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/eltwise_conv_eltwise.cpp
@@ -45,7 +45,7 @@ std::vector<size_t> outputChannels = {
     8
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseAfterConvTest, EltwiseAfterConvTest,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseAfterConvTest, EltwiseAfterConvTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -55,7 +55,7 @@ INSTANTIATE_TEST_CASE_P(smoke_EltwiseAfterConvTest, EltwiseAfterConvTest,
                             ::testing::ValuesIn(outputChannels)),
                         EltwiseAfterConvTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseBeforeConvTest, EltwiseBeforeConvTest,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseBeforeConvTest, EltwiseBeforeConvTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -65,7 +65,7 @@ INSTANTIATE_TEST_CASE_P(smoke_EltwiseBeforeConvTest, EltwiseBeforeConvTest,
                             ::testing::ValuesIn(outputChannels)),
                         EltwiseBeforeConvTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseWithTwoConvsAsInputsTest, EltwiseWithTwoConvsAsInputsTest,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseWithTwoConvsAsInputsTest, EltwiseWithTwoConvsAsInputsTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/eltwise_reshape_activation.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/eltwise_reshape_activation.cpp
index 332da49fb38b41..8e7af93c892d00 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/eltwise_reshape_activation.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/eltwise_reshape_activation.cpp
@@ -30,7 +30,7 @@ std::vector<std::map<std::string, std::string>>  additional_config = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseReshapeActivationTest, EltwiseReshapeActivation,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseReshapeActivationTest, EltwiseReshapeActivation,
                         ::testing::Combine(
                                 ::testing::ValuesIn(shapes),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/fc_conv_fc.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/fc_conv_fc.cpp
index db70b15fd5c01e..069398cd7d5c4c 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/fc_conv_fc.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/fc_conv_fc.cpp
@@ -45,7 +45,7 @@ std::vector<size_t> outputChannels = {
     8
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_FcAfterConvTest, FcAfterConvTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FcAfterConvTest, FcAfterConvTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -55,7 +55,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FcAfterConvTest, FcAfterConvTest,
                             ::testing::ValuesIn(outputChannels)),
                         FcAfterConvTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_FcBeforeConvTest, FcBeforeConvTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FcBeforeConvTest, FcBeforeConvTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -65,7 +65,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FcBeforeConvTest, FcBeforeConvTest,
                             ::testing::ValuesIn(outputChannels)),
                         FcBeforeConvTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_FcBetweenConvsTest, FcBetweenConvsTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FcBetweenConvsTest, FcBetweenConvsTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/first_connect_input_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/first_connect_input_concat.cpp
index 3a7bdf8935d524..e62bdd99cfcb72 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/first_connect_input_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/first_connect_input_concat.cpp
@@ -27,7 +27,7 @@ std::map<std::string, std::string> additional_config = {
         {"GNA_COMPACT_MODE", "NO"},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_concat_first_input, ConcatFirstInputTest,
+INSTANTIATE_TEST_SUITE_P(smoke_concat_first_input, ConcatFirstInputTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inShapes),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/fq_conv_fq_affine.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/fq_conv_fq_affine.cpp
index e48d4ad12c0c8e..34957dc3da754e 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/fq_conv_fq_affine.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/fq_conv_fq_affine.cpp
@@ -48,7 +48,7 @@ const auto convParams = ::testing::Combine(
 
 const std::vector<bool> permute = {false, true};
 
-INSTANTIATE_TEST_CASE_P(smoke_FqConvFqAffineTest, FqConvFqAffineTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FqConvFqAffineTest, FqConvFqAffineTest,
                         ::testing::Combine(
                                 fqParams,
                                 convParams,
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp
index 7588776f73b64c..f7b2e5fd5eb018 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp
@@ -28,7 +28,7 @@ std::map<std::string, std::string> additional_config = {
 };
 } // namespace
 
-INSTANTIATE_TEST_CASE_P(OutputBeforeActivation, OutputBeforeActivation,
+INSTANTIATE_TEST_SUITE_P(OutputBeforeActivation, OutputBeforeActivation,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
         ::testing::Values(InferenceEngine::Precision::FP32),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/handling_orientation_conv.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/handling_orientation_conv.cpp
index f752ba54f1d6cc..ed51ceb4c59b63 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/handling_orientation_conv.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/handling_orientation_conv.cpp
@@ -21,7 +21,7 @@ const std::vector<std::map<std::string, std::string>> configs = {
         }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_handling_orientation, HandlingOrientationClass,
+INSTANTIATE_TEST_SUITE_P(smoke_handling_orientation, HandlingOrientationClass,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/input_conv.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/input_conv.cpp
index 3fad3b60d27d46..9b15275b91d1e9 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/input_conv.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/input_conv.cpp
@@ -46,7 +46,7 @@ std::vector<bool> addReshape = {
     false
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_InputConv, InputConvTest,
+INSTANTIATE_TEST_SUITE_P(smoke_InputConv, InputConvTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/input_split_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/input_split_concat.cpp
index 2e667336fecbe9..873b8518a795e9 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/input_split_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/input_split_concat.cpp
@@ -30,7 +30,7 @@ const std::vector<std::vector<size_t>> inputShapes = {
     {1, 320}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_InputSplitConcat, InputSplitConcatTest,
+INSTANTIATE_TEST_SUITE_P(smoke_InputSplitConcat, InputSplitConcatTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp
index 17afb8cd5a8af3..2b638fad70e8b7 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp
@@ -39,7 +39,7 @@ std::vector<size_t> output_sizes = {
     8
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MatmulSqueezeAdd, MatmulSqueezeAddTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MatmulSqueezeAdd, MatmulSqueezeAddTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
index 5818a40d4ecaa2..db2ca0e8f743b6 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_LSTMCell.cpp
@@ -36,7 +36,7 @@ namespace SubgraphTestsDefinitions {
         {"GNA_SCALE_FACTOR_0", "1638.4"},
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_MemoryLSTMCellTest, MemoryLSTMCellTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_MemoryLSTMCellTest, MemoryLSTMCellTest,
         ::testing::Combine(
             ::testing::ValuesIn(transformation),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_eltwise_reshape_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_eltwise_reshape_concat.cpp
index bf287286ac59db..13264c36b312c6 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_eltwise_reshape_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/memory_eltwise_reshape_concat.cpp
@@ -26,7 +26,7 @@ std::map<std::string, std::string> additional_config = {
 };
 } // namespace
 
-INSTANTIATE_TEST_CASE_P(smoke_MemoryEltwiseReshapeConcatTest, MemoryEltwiseReshapeConcatTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MemoryEltwiseReshapeConcatTest, MemoryEltwiseReshapeConcatTest,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
         ::testing::Values(InferenceEngine::Precision::FP32),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multioutput_eltwise_squeeze_eltwise.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multioutput_eltwise_squeeze_eltwise.cpp
index 0055e18c95bafc..b937e07b93a1e5 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multioutput_eltwise_squeeze_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multioutput_eltwise_squeeze_eltwise.cpp
@@ -31,7 +31,7 @@ namespace {
                                                              InferenceEngine::Precision::FP16,
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_multioutput_eltwise_identity, MultioutputEltwiseReshapeEltwise,
+    INSTANTIATE_TEST_SUITE_P(smoke_multioutput_eltwise_identity, MultioutputEltwiseReshapeEltwise,
                             ::testing::Combine(
                                     ::testing::ValuesIn(inputs),
                                     ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
index d59c023773bb38..3f695bd1bffefa 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_LSTMCell.cpp
@@ -38,7 +38,7 @@ std::map<std::string, std::string> additional_config = {
 } // namespace
 
 
-INSTANTIATE_TEST_CASE_P(MultipleLSTMCellTest, MultipleLSTMCellTest,
+INSTANTIATE_TEST_SUITE_P(MultipleLSTMCellTest, MultipleLSTMCellTest,
     ::testing::Combine(
         ::testing::ValuesIn(transformation),
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_concat.cpp
index 0cb64714ffb702..2cd3e96ee992e3 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/multiple_concat.cpp
@@ -34,7 +34,7 @@ std::map<std::string, std::string> additional_config = {
 };
 } // namespace
 
-INSTANTIATE_TEST_CASE_P(I_aligned_C_aligned, MultipleConcatTest,
+INSTANTIATE_TEST_SUITE_P(I_aligned_C_aligned, MultipleConcatTest,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
         ::testing::Values(InferenceEngine::Precision::FP32),
@@ -43,7 +43,7 @@ INSTANTIATE_TEST_CASE_P(I_aligned_C_aligned, MultipleConcatTest,
         ::testing::Values(additional_config)),
     MultipleConcatTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(I_aligned_C_unaligned, MultipleConcatTest,
+INSTANTIATE_TEST_SUITE_P(I_aligned_C_unaligned, MultipleConcatTest,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
         ::testing::Values(InferenceEngine::Precision::FP32),
@@ -52,7 +52,7 @@ INSTANTIATE_TEST_CASE_P(I_aligned_C_unaligned, MultipleConcatTest,
         ::testing::Values(additional_config)),
     MultipleConcatTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(I_unaligned_C_aligned, MultipleConcatTest,
+INSTANTIATE_TEST_SUITE_P(I_unaligned_C_aligned, MultipleConcatTest,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
         ::testing::Values(InferenceEngine::Precision::FP32),
@@ -61,7 +61,7 @@ INSTANTIATE_TEST_CASE_P(I_unaligned_C_aligned, MultipleConcatTest,
         ::testing::Values(additional_config)),
     MultipleConcatTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(I_unaligned_C_unaligned, MultipleConcatTest,
+INSTANTIATE_TEST_SUITE_P(I_unaligned_C_unaligned, MultipleConcatTest,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GNA),
         ::testing::Values(InferenceEngine::Precision::FP32),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/negative_memory_layer_offset.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/negative_memory_layer_offset.cpp
index 36b530e4469501..c1d038f0463219 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/negative_memory_layer_offset.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/negative_memory_layer_offset.cpp
@@ -32,7 +32,7 @@ namespace {
         100
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_negative_memory_layer_offset, NegativeMemoryOffsetTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_negative_memory_layer_offset, NegativeMemoryOffsetTest,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp
index e6fd385f5f825e..f5a812e40ea614 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp
@@ -32,7 +32,7 @@ std::map<std::string, std::string> additional_config = {
 } // namespace
 
 namespace SubgraphTestsDefinitions {
-    INSTANTIATE_TEST_CASE_P(smoke_basic, PermConvPermConcat,
+    INSTANTIATE_TEST_SUITE_P(smoke_basic, PermConvPermConcat,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshape_permute_conv_permute_reshape_act.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshape_permute_conv_permute_reshape_act.cpp
index a07908927c1d0c..08a665e4e06c91 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshape_permute_conv_permute_reshape_act.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshape_permute_conv_permute_reshape_act.cpp
@@ -35,7 +35,7 @@ std::map<std::string, std::string> additional_config = {
 };
 
 namespace SubgraphTestsDefinitions {
-    INSTANTIATE_TEST_CASE_P(smoke_basic, ConvReshapeAct,
+    INSTANTIATE_TEST_SUITE_P(smoke_basic, ConvReshapeAct,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
index 5cf777fe654281..e9163914719484 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
@@ -38,7 +38,7 @@ namespace {
             ngraph::helpers::SqueezeOpType::UNSQUEEZE
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_reshape_squeeze_reshape_relu, ReshapeSqueezeReshapeRelu,
+    INSTANTIATE_TEST_SUITE_P(smoke_reshape_squeeze_reshape_relu, ReshapeSqueezeReshapeRelu,
                             ::testing::Combine(
                                     ::testing::ValuesIn(inputs),
                                     ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshapre_permute_reshape.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshapre_permute_reshape.cpp
index 27c9b22c0e87e0..f74add45bfb896 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshapre_permute_reshape.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/reshapre_permute_reshape.cpp
@@ -22,7 +22,7 @@ namespace {
                                                              InferenceEngine::Precision::FP16,
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_reshape_permute_reshape, ReshapePermuteReshape,
+    INSTANTIATE_TEST_SUITE_P(smoke_reshape_permute_reshape, ReshapePermuteReshape,
                             ::testing::Combine(
                                     ::testing::ValuesIn(inputs),
                                     ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/scale_shift.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/scale_shift.cpp
index 759413f8bd5b19..c5856c9fca5ce4 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/scale_shift.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/scale_shift.cpp
@@ -42,7 +42,7 @@ namespace {
                                                              InferenceEngine::Precision::FP16,
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_scale_shift, ScaleShiftLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_scale_shift, ScaleShiftLayerTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(inShapes),
                                     ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/scaleshift_conv_scaleshift.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/scaleshift_conv_scaleshift.cpp
index 2e476ac1f876cf..2a823babc5ee8e 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/scaleshift_conv_scaleshift.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/scaleshift_conv_scaleshift.cpp
@@ -45,7 +45,7 @@ std::vector<size_t> outputChannels = {
     8
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ScaleShiftAfterConvTest, ScaleShiftAfterConvTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ScaleShiftAfterConvTest, ScaleShiftAfterConvTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
@@ -55,7 +55,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ScaleShiftAfterConvTest, ScaleShiftAfterConvTest,
                             ::testing::ValuesIn(outputChannels)),
                         ScaleShiftAfterConvTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ScaleShiftBeforeConvTest, ScaleShiftBeforeConvTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ScaleShiftBeforeConvTest, ScaleShiftBeforeConvTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/softsign.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/softsign.cpp
index 58c987151c060c..00333131e3b1cb 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/softsign.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/softsign.cpp
@@ -33,7 +33,7 @@ std::vector<std::vector<size_t>> input_shapes = {
     {1, 1, 1, 100}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Softsign, SoftsignTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Softsign, SoftsignTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_conv.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_conv.cpp
index bf71c5e2c9de58..0b508c3b1fd794 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_conv.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_conv.cpp
@@ -42,7 +42,7 @@ std::vector<size_t> outputChannels = {
     8
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_SplitConvTest, SplitConvTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SplitConvTest, SplitConvTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_conv_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
index 3a3c4e3bbba30d..755613335b30bd 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
@@ -19,7 +19,7 @@ std::vector<std::vector<size_t>> inputShapes = {
     {1, 32, 1, 1026}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_SplitConvConcat, SplitConvConcat,
+INSTANTIATE_TEST_SUITE_P(smoke_SplitConvConcat, SplitConvConcat,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_relu.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_relu.cpp
index b9764b919fc821..69a5e053f726aa 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_relu.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_relu.cpp
@@ -47,7 +47,7 @@ namespace {
             {std::string(GNA_CONFIG_KEY(COMPACT_MODE)), "NO"}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_split_connected, SplitRelu,
+    INSTANTIATE_TEST_SUITE_P(smoke_split_connected, SplitRelu,
                             ::testing::Combine(
                                     ::testing::ValuesIn(inputs),
                                     ::testing::ValuesIn(connect_index),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_trivial_permute_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_trivial_permute_concat.cpp
index 49047d3cf8341f..1982001ed9dcd8 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_trivial_permute_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/split_trivial_permute_concat.cpp
@@ -27,7 +27,7 @@ namespace {
     std::vector<size_t> split_axes = { 1 }; // only channels split is currently supported by gna for 4d inputs
     std::vector<size_t> concat_axes = { 1 }; // only channels concat is currently supported by gna for 4d inputs
 
-    INSTANTIATE_TEST_CASE_P(smoke_split_trivial_permute_concat, SplitTrivialPermuteConcatTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_split_trivial_permute_concat, SplitTrivialPermuteConcatTest,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/stridedslice_conv.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/stridedslice_conv.cpp
index 4517005f3d7afa..c0894a2e6b406a 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/stridedslice_conv.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/stridedslice_conv.cpp
@@ -36,7 +36,7 @@ std::vector<size_t> outputChannels = {
     8
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_SliceConvTest, SliceConvTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SliceConvTest, SliceConvTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/tensor_names.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/tensor_names.cpp
index 5ef074f3496285..8a1489aebbb509 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/tensor_names.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/tensor_names.cpp
@@ -10,7 +10,7 @@
 using namespace SubgraphTestsDefinitions;
 
 namespace {
-    INSTANTIATE_TEST_CASE_P(smoke_Check, TensorNamesTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Check, TensorNamesTest,
                             ::testing::Values(CommonTestUtils::DEVICE_GNA),
                             TensorNamesTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/trivial_concat.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/trivial_concat.cpp
index 866080faa2a38e..aed6bf9b9ebc61 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/trivial_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/trivial_concat.cpp
@@ -27,7 +27,7 @@ std::map<std::string, std::string> additional_config = {
     {"GNA_SCALE_FACTOR_0", "2000.0"},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_trivial_concat_Basic, TrivialConcatLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_trivial_concat_Basic, TrivialConcatLayerTest,
     ::testing::Combine(
         ::testing::ValuesIn(inShapes),
         ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/two_fake_quantize_to_fullyconnected.cpp b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/two_fake_quantize_to_fullyconnected.cpp
index 5de8e1813aab9a..fb85c4334ca9a0 100644
--- a/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/two_fake_quantize_to_fullyconnected.cpp
+++ b/inference-engine/tests/functional/plugin/gna/shared_tests_instances/subgraph_tests/two_fake_quantize_to_fullyconnected.cpp
@@ -80,7 +80,7 @@ const auto fqParams_I16 = ::testing::Combine(
         ::testing::ValuesIn(inputParams_I8)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantize_subgraph, FakeQuantizeSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantize_subgraph, FakeQuantizeSubgraphTest,
                         ::testing::Combine(
                                 fqParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -94,7 +94,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FakeQuantize_subgraph, FakeQuantizeSubgraphTest,
                                 ::testing::ValuesIn(biases)),
                         FakeQuantizeSubgraphTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantize_subgraph_U8, FakeQuantizeSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantize_subgraph_U8, FakeQuantizeSubgraphTest,
                         ::testing::Combine(
                                 fqParams_I8,
                                 ::testing::ValuesIn(netPrecisions),
@@ -108,7 +108,7 @@ INSTANTIATE_TEST_CASE_P(smoke_FakeQuantize_subgraph_U8, FakeQuantizeSubgraphTest
                                 ::testing::ValuesIn(biases)),
                         FakeQuantizeSubgraphTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantize_subgraph_I16, FakeQuantizeSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantize_subgraph_I16, FakeQuantizeSubgraphTest,
                         ::testing::Combine(
                                 fqParams_I16,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/remote_blob_tests/cldnn_remote_blob_tests.cpp b/inference-engine/tests/functional/plugin/gpu/remote_blob_tests/cldnn_remote_blob_tests.cpp
index 6837c0b84c33a8..30b740464a62f3 100644
--- a/inference-engine/tests/functional/plugin/gpu/remote_blob_tests/cldnn_remote_blob_tests.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/remote_blob_tests/cldnn_remote_blob_tests.cpp
@@ -269,7 +269,7 @@ TEST_P(BatchedBlob_Test, canInputNV12) {
 
 const std::vector<size_t> num_batches{1, 2, 4};
 
-INSTANTIATE_TEST_CASE_P(smoke_RemoteBlob, BatchedBlob_Test, ::testing::ValuesIn(num_batches), BatchedBlob_Test::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_RemoteBlob, BatchedBlob_Test, ::testing::ValuesIn(num_batches), BatchedBlob_Test::getTestCaseName);
 
 class TwoNets_Test : public CommonTestUtils::TestsCommon, public testing::WithParamInterface<size_t> {
     void SetUp() override {
@@ -357,4 +357,4 @@ TEST_P(TwoNets_Test, canInferTwoExecNets) {
 
 const std::vector<size_t> num_streams{1, 2};
 
-INSTANTIATE_TEST_CASE_P(smoke_RemoteBlob, TwoNets_Test, ::testing::ValuesIn(num_streams), TwoNets_Test::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_RemoteBlob, TwoNets_Test, ::testing::ValuesIn(num_streams), TwoNets_Test::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/caching_tests.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/caching_tests.cpp
index e840f2aa797f75..247d1482e765cd 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/caching_tests.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/caching_tests.cpp
@@ -22,7 +22,7 @@ namespace {
             1, 2
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_CachingSupportCase_GPU, LoadNetworkCacheTestBase,
+    INSTANTIATE_TEST_SUITE_P(smoke_CachingSupportCase_GPU, LoadNetworkCacheTestBase,
                             ::testing::Combine(
                                     ::testing::ValuesIn(LoadNetworkCacheTestBase::getStandardFunctions()),
                                     ::testing::ValuesIn(precisionsGPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp
index a8c039e43915a4..235e82475c117b 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/config.cpp
@@ -50,21 +50,21 @@ namespace {
     };
     IE_SUPPRESS_DEPRECATED_END
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigTests,
             ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_GPU),
                 ::testing::ValuesIn(inconfigs)),
             IncorrectConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, IncorrectConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, IncorrectConfigTests,
             ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                 ::testing::ValuesIn(multiinconfigs)),
             IncorrectConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, IncorrectConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, IncorrectConfigTests,
             ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
@@ -111,63 +111,63 @@ namespace {
         {{InferenceEngine::KEY_AUTO_DEVICE_LIST , std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectConfigAPITests,
             ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_GPU),
                 ::testing::ValuesIn(conf)),
             CorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_GPU_BehaviorTests, CorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_GPU_BehaviorTests, CorrectConfigAPITests,
             ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_GPU),
                 ::testing::ValuesIn(conf_gpu)),
             CorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, CorrectConfigAPITests,
             ::testing::Combine(
                     ::testing::ValuesIn(netPrecisions),
                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                     ::testing::ValuesIn(multiconf)),
             CorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, CorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, CorrectConfigAPITests,
             ::testing::Combine(
                     ::testing::ValuesIn(netPrecisions),
                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                     ::testing::ValuesIn(autoconf)),
             CorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, CorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, CorrectConfigAPITests,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                 ::testing::ValuesIn(auto_cpu_gpu_conf)),
                             CorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
             ::testing::Combine(
                     ::testing::ValuesIn(netPrecisions),
                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                     ::testing::ValuesIn(conf)),
              IncorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, IncorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, IncorrectConfigAPITests,
             ::testing::Combine(
                     ::testing::ValuesIn(netPrecisions),
                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                     ::testing::ValuesIn(multiconf)),
             IncorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, IncorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, IncorrectConfigAPITests,
             ::testing::Combine(
                     ::testing::ValuesIn(netPrecisions),
                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                     ::testing::ValuesIn(autoinconfigs)),
             IncorrectConfigAPITests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, IncorrectConfigAPITests,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, IncorrectConfigAPITests,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_integration.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_integration.cpp
index 68b23831e47944..a260d3a28d334a 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_integration.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_integration.cpp
@@ -19,12 +19,12 @@ namespace {
 // IE Class Common tests with <pluginName, deviceName params>
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassCommon, IEClassBasicTestP,
         ::testing::Values(std::make_pair("clDNNPlugin", "GPU"))
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassNetworkTestP, IEClassNetworkTestP,
         ::testing::Values("GPU")
 );
@@ -33,62 +33,62 @@ INSTANTIATE_TEST_CASE_P(
 // IE Class GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values("GPU", "MULTI", "HETERO", "AUTO")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values("GPU", "MULTI", "HETERO", "AUTO")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_AVAILABLE_DEVICES,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_FULL_DEVICE_NAME,
         ::testing::Values("GPU", "MULTI", "HETERO", "AUTO")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_OPTIMIZATION_CAPABILITIES,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_DEVICE_GOPS,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_DEVICE_TYPE,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_RANGE_FOR_ASYNC_INFER_REQUESTS,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_RANGE_FOR_STREAMS,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_ThrowUnsupported,
         ::testing::Values("GPU", "MULTI", "HETERO", "AUTO")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetConfigTest, IEClassGetConfigTest_ThrowUnsupported,
         ::testing::Values("GPU", "MULTI", "HETERO", "AUTO")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetAvailableDevices, IEClassGetAvailableDevices,
         ::testing::Values("GPU")
 );
@@ -110,7 +110,7 @@ TEST_P(IEClassGetMetricTest_GPU_DEVICE_TOTAL_MEM_SIZE, GetMetricAndPrintNoThrow)
     ASSERT_METRIC_SUPPORTED(GPU_METRIC_KEY(DEVICE_TOTAL_MEM_SIZE));
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_GPU_DEVICE_TOTAL_MEM_SIZE,
         ::testing::Values("GPU")
 );
@@ -129,7 +129,7 @@ TEST_P(IEClassGetMetricTest_GPU_UARCH_VERSION, GetMetricAndPrintNoThrow) {
     ASSERT_METRIC_SUPPORTED(GPU_METRIC_KEY(UARCH_VERSION));
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_GPU_UARCH_VERSION,
         ::testing::Values("GPU")
 );
@@ -148,7 +148,7 @@ TEST_P(IEClassGetMetricTest_GPU_EXECUTION_UNITS_COUNT, GetMetricAndPrintNoThrow)
     ASSERT_METRIC_SUPPORTED(GPU_METRIC_KEY(EXECUTION_UNITS_COUNT));
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetMetricTest, IEClassGetMetricTest_GPU_EXECUTION_UNITS_COUNT,
         ::testing::Values("GPU")
 );
@@ -157,7 +157,7 @@ INSTANTIATE_TEST_CASE_P(
 // IE Class GetConfig
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassGetConfigTest, IEClassGetConfigTest,
         ::testing::Values("GPU")
 );
@@ -166,27 +166,27 @@ INSTANTIATE_TEST_CASE_P(
 // Executable Network GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_OPTIMAL_NUMBER_OF_INFER_REQUESTS,
         ::testing::Values("GPU", "MULTI:GPU", "HETERO:GPU", "AUTO:GPU,CPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values("GPU", "MULTI:GPU", "HETERO:GPU", "AUTO:GPU,CPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values("GPU", "MULTI:GPU", "HETERO:GPU", "AUTO:GPU,CPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_NETWORK_NAME,
         ::testing::Values("GPU", "MULTI:GPU", "HETERO:GPU", "AUTO:GPU,CPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassExecutableNetworkGetMetricTest, IEClassExecutableNetworkGetMetricTest_ThrowsUnsupported,
         ::testing::Values("GPU", "MULTI:GPU", "HETERO:GPU", "AUTO:GPU,CPU")
 );
@@ -195,12 +195,12 @@ INSTANTIATE_TEST_CASE_P(
 // Executable Network GetConfig / SetConfig
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassExecutableNetworkGetConfigTest, IEClassExecutableNetworkGetConfigTest,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassExecutableNetworkSetConfigTest, IEClassExecutableNetworkSetConfigTest,
         ::testing::Values("GPU")
 );
@@ -209,41 +209,41 @@ INSTANTIATE_TEST_CASE_P(
 // Hetero Executable Network GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_SUPPORTED_METRICS,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_NETWORK_NAME,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         nightly_IEClassHeteroExecutableNetworlGetMetricTest, IEClassHeteroExecutableNetworkGetMetricTest_TARGET_FALLBACK,
         ::testing::Values("GPU")
 );
 
 // IE Class Query network
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassQueryNetworkTest, IEClassQueryNetworkTest,
         ::testing::Values("GPU")
 );
 
 // IE Class Load network
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassLoadNetworkTest, IEClassLoadNetworkTest,
         ::testing::Values("GPU")
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassHeteroExecutableNetworkGetMetricTest, IEClassLoadNetworkAfterCoreRecreateTest,
         ::testing::Values("GPU")
 );
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_threading_tests.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_threading_tests.cpp
index db06501a896e3b..779e0212d0fd33 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_threading_tests.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/core_threading_tests.cpp
@@ -39,9 +39,9 @@ TEST_P(CoreThreadingTestsWithIterations, smoke_LoadNetwork_RemoteContext) {
     }, numIterations, numThreads);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_GPU, CoreThreadingTests, testing::ValuesIn(params), CoreThreadingTests::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_GPU, CoreThreadingTests, testing::ValuesIn(params), CoreThreadingTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GPU, CoreThreadingTestsWithIterations,
+INSTANTIATE_TEST_SUITE_P(smoke_GPU, CoreThreadingTestsWithIterations,
     testing::Combine(testing::ValuesIn(params),
                      testing::Values(4),
                      testing::Values(20),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/cpp_holders.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/cpp_holders.cpp
index 329db9f216d115..12ad85d05be141 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/cpp_holders.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/cpp_holders.cpp
@@ -18,7 +18,7 @@ namespace {
             {2, 1, 0}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTest,
             ::testing::Combine(
             ::testing::Values(CommonTestUtils::DEVICE_GPU),
             ::testing::ValuesIn(orders)),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/exec_graph_info.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/exec_graph_info.cpp
index 1a3ca13252c030..8b220fad0326a6 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/exec_graph_info.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/exec_graph_info.cpp
@@ -15,7 +15,7 @@ namespace {
             {},
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, ExecGraphTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, ExecGraphTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request.cpp
index 866f4b1cb68ac5..c04aeffd48bc97 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request.cpp
@@ -26,28 +26,28 @@ namespace {
         {{InferenceEngine::KEY_AUTO_DEVICE_LIST , std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                                     ::testing::Values(std::map<std::string, std::string>({}))),
                             InferRequestTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferRequestTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferRequestTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(configs)),
                             InferRequestTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, InferRequestTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, InferRequestTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(autoconfigs)),
                             InferRequestTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, InferRequestTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, InferRequestTests,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_callback.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_callback.cpp
index 3d9963a41d791f..a0108e84339d4a 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_callback.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_callback.cpp
@@ -27,28 +27,28 @@ const std::vector<std::map<std::string, std::string>> auto_cpu_gpu_conf = {
     {{InferenceEngine::KEY_AUTO_DEVICE_LIST , std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CallbackTests,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GPU),
             ::testing::ValuesIn(configs)),
         CallbackTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, CallbackTests,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_MULTI),
             ::testing::ValuesIn(multiConfigs)),
         CallbackTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, CallbackTests,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                             ::testing::ValuesIn(autoConfigs)),
                         CallbackTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, CallbackTests,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_config.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_config.cpp
index 0072369d920734..f48cd8451a71f3 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_config.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_config.cpp
@@ -27,28 +27,28 @@ namespace {
         {{InferenceEngine::KEY_AUTO_DEVICE_LIST , std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferConfigTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                                     ::testing::ValuesIn(configs)),
                             InferConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferConfigTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(multiConfigs)),
                             InferConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, InferConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, InferConfigTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(autoConfigs)),
                             InferConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, InferConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, InferConfigTests,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp
index 59f4dd21677c24..f63be4c8854326 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_input.cpp
@@ -33,28 +33,28 @@ namespace {
         {{InferenceEngine::KEY_AUTO_DEVICE_LIST , std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestInputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestInputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                                     ::testing::ValuesIn(configs)),
                             InferRequestInputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferRequestInputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferRequestInputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(multiConfigs)),
                             InferRequestInputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, InferRequestInputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, InferRequestInputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(autoConfigs)),
                             InferRequestInputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, InferRequestInputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, InferRequestInputTests,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp
index 1135f6d9f7d8e4..638106e6a75a52 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/infer_request_output.cpp
@@ -29,28 +29,28 @@ namespace {
         {{InferenceEngine::KEY_AUTO_DEVICE_LIST , std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestOutputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestOutputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                                     ::testing::ValuesIn(configs)),
                             InferRequestOutputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferRequestOutputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferRequestOutputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(multiConfigs)),
                             InferRequestOutputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, InferRequestOutputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, InferRequestOutputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(autoConfigs)),
                             InferRequestOutputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, InferRequestOutputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, InferRequestOutputTests,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/layout.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/layout.cpp
index 6748fc58c50c32..c705c3a9d76d73 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/layout.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/layout.cpp
@@ -20,7 +20,7 @@ namespace {
             { 1, 3 },
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, LayoutTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, LayoutTest,
             ::testing::Combine(
                     ::testing::Values(InferenceEngine::Precision::FP32),
                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/perf_counters.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/perf_counters.cpp
index d8a89ef317d425..cc6a2a14a24fa7 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/perf_counters.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/perf_counters.cpp
@@ -14,14 +14,14 @@ namespace {
             {{ MULTI_CONFIG_KEY(DEVICE_PRIORITIES) , CommonTestUtils::DEVICE_GPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PerfCountersTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PerfCountersTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                                     ::testing::ValuesIn(configs)),
                             PerfCountersTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, PerfCountersTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, PerfCountersTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/set_blob_of_kind.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/set_blob_of_kind.cpp
index 936abfe826dd8d..f055c0b0ec9c8c 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/set_blob_of_kind.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/set_blob_of_kind.cpp
@@ -17,7 +17,7 @@ const std::vector<FuncTestUtils::BlobKind> blobKinds = {
 
 const SetBlobOfKindConfig gpuConfig{}; //nothing special
 
-INSTANTIATE_TEST_CASE_P(smoke_SetBlobOfKindGPU, SetBlobOfKindTest,
+INSTANTIATE_TEST_SUITE_P(smoke_SetBlobOfKindGPU, SetBlobOfKindTest,
     ::testing::Combine(::testing::ValuesIn(blobKinds),
                        ::testing::Values(CommonTestUtils::DEVICE_GPU),
                        ::testing::Values(gpuConfig)),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/set_preprocess.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/set_preprocess.cpp
index 80c1fdccca9800..136cb96436ca61 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/set_preprocess.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/set_preprocess.cpp
@@ -30,28 +30,28 @@ namespace {
         {{InferenceEngine::KEY_AUTO_DEVICE_LIST , std::string(CommonTestUtils::DEVICE_CPU) + "," + CommonTestUtils::DEVICE_GPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                                     ::testing::ValuesIn(configs)),
                             PreprocessTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, PreprocessTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, PreprocessTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(multiConfigs)),
                             PreprocessTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, PreprocessTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, PreprocessTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(autoConfigs)),
                             PreprocessTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, PreprocessTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, PreprocessTest,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
@@ -73,7 +73,7 @@ namespace {
         InferenceEngine::Layout::NHWC
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessConversionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessConversionTest,
                                 ::testing::Combine(
                                         ::testing::ValuesIn(netPrecisions),
                                         ::testing::ValuesIn(ioPrecisions),
@@ -87,7 +87,7 @@ namespace {
                                         ::testing::ValuesIn(configs)),
                                 PreprocessConversionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Bool(),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp
index 729bf57c64ac8a..36790623b489ab 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/test_plugin.cpp
@@ -47,84 +47,84 @@ namespace {
             {{InferenceEngine::PluginConfigParams::KEY_GPU_THROUGHPUT_STREAMS, InferenceEngine::PluginConfigParams::GPU_THROUGHPUT_AUTO}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTestOutput,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                                     ::testing::ValuesIn(configsOutput)),
                             BehaviorTestOutput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, BehaviorTestOutput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(MultiConfigsInputOutput)),
                             BehaviorTestOutput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, BehaviorTestOutput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(AutoConfigs)),
                             BehaviorTestOutput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, BehaviorTestOutput,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                 ::testing::ValuesIn(auto_cpu_gpu_conf)),
                             BehaviorTestOutput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTests,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                                     ::testing::ValuesIn(configs)),
                             BehaviorTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, BehaviorTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, BehaviorTests,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(MultiConfigs)),
                             BehaviorTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, BehaviorTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, BehaviorTests,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(AutoConfigs)),
                             BehaviorTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, BehaviorTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, BehaviorTests,
                             ::testing::Combine(
                                 ::testing::Values(InferenceEngine::Precision::FP32),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                 ::testing::ValuesIn(auto_cpu_gpu_conf)),
                             BehaviorTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTestInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                                     ::testing::ValuesIn(configsInput)),
                             BehaviorTestInput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, BehaviorTestInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(MultiConfigsInputOutput)),
                             BehaviorTestInput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, BehaviorTestInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(AutoConfigs)),
                             BehaviorTestInput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, BehaviorTestInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/version.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/version.cpp
index 81ec5b808d3116..6b69c22855ecee 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/version.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/behavior/version.cpp
@@ -26,35 +26,35 @@ namespace {
             {{ HETERO_CONFIG_KEY(DUMP_GRAPH_DOT) , CommonTestUtils::DEVICE_GPU}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU),
                                     ::testing::ValuesIn(configs)),
                             VersionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(Multiconfigs)),
                             VersionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Auto_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Auto_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                     ::testing::ValuesIn(Autoconfigs)),
                             VersionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_AutoCG_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_AutoCG_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                 ::testing::Values(InferenceEngine::Precision::FP32),
                                 ::testing::Values(CommonTestUtils::DEVICE_AUTO),
                                 ::testing::ValuesIn(auto_cpu_gpu_conf)),
                             VersionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Hetero_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Hetero_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_HETERO),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/configuration_tests/dynamic_batch.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/configuration_tests/dynamic_batch.cpp
index 6f3cac60cfb74c..36a13d4ea5f733 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/configuration_tests/dynamic_batch.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/configuration_tests/dynamic_batch.cpp
@@ -20,7 +20,7 @@ std::map<std::string, std::string> additional_config = {
 } // namespace
 
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicBatchTest_async, DynamicBatchTest,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicBatchTest_async, DynamicBatchTest,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
         ::testing::Values(InferenceEngine::Precision::FP32),
@@ -29,7 +29,7 @@ INSTANTIATE_TEST_CASE_P(smoke_DynamicBatchTest_async, DynamicBatchTest,
         ::testing::Values(additional_config)),
     DynamicBatchTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicBatchTest_sync, DynamicBatchTest,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicBatchTest_sync, DynamicBatchTest,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
         ::testing::Values(InferenceEngine::Precision::FP32),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/execution_graph_tests/unique_node_names.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/execution_graph_tests/unique_node_names.cpp
index d3689b4492c894..275d8f179fbf9e 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/execution_graph_tests/unique_node_names.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/execution_graph_tests/unique_node_names.cpp
@@ -15,7 +15,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP32
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, ExecGraphUniqueNodeNames,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, ExecGraphUniqueNodeNames,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(InferenceEngine::SizeVector({1, 2, 5, 5})),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/hetero/query_network.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/hetero/query_network.cpp
index a23cf6b936b24a..893f1fce09bdf0 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/hetero/query_network.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/hetero/query_network.cpp
@@ -15,7 +15,7 @@ auto ConvBias = ngraph::builder::subgraph::makeConvBias();
 auto SplitConvConcat = ngraph::builder::subgraph::makeNestedSplitConvConcat();
 auto BranchSplitConvConcat = ngraph::builder::subgraph::makeSplitConvConcatNestedInBranch();
 
-INSTANTIATE_TEST_CASE_P(smoke_FullySupportedTopologies, QueryNetworkTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FullySupportedTopologies, QueryNetworkTest,
                         ::testing::Combine(
                                 ::testing::Values("GPU", "HETERO:GPU,CPU", "MULTI:GPU,CPU"),
                                 ::testing::Values(ConvBias, SplitConvConcat, BranchSplitConvConcat)),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/add_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/add_transformation.cpp
index 8a33dacde43fea..7b51e051339c3c 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/add_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/add_transformation.cpp
@@ -68,7 +68,7 @@ const std::vector<LayerTestsDefinitions::AddTestValues> params = {
     { {}, {}, false }, { {}, {}, true },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, AddTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, AddTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/clamp_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/clamp_transformation.cpp
index 078146f1819c10..408cbb550714c0 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/clamp_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/clamp_transformation.cpp
@@ -76,7 +76,7 @@ const std::vector<LayerTestsDefinitions::ClampTransformationParam> params = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ClampTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ClampTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_transformation.cpp
index 57cca3618123ae..f22eb9f63a142f 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_transformation.cpp
@@ -39,7 +39,7 @@ const std::vector<ConcatTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_child_and_output.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_child_and_output.cpp
index b5702fe243114e..f53eb289438546 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_child_and_output.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_child_and_output.cpp
@@ -42,7 +42,7 @@ const std::vector<ConcatWithChildAndOutputTransformationParam> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatWithChildAndOutputTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatWithChildAndOutputTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 6, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_different_precision_on_children.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_different_precision_on_children.cpp
index 0b2834837164b1..d7605440afe6f5 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_different_precision_on_children.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_different_precision_on_children.cpp
@@ -45,7 +45,7 @@ const std::vector<ConcatWithDifferentChildrenTransformationParam> testValues = {
 
 const std::vector<bool> multiChannel = { true/*, false*/ };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatWithDifferentChildrenTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatWithDifferentChildrenTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_intermediate_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_intermediate_transformation.cpp
index 16daf1cbc1d377..00f5467b33c714 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_intermediate_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_intermediate_transformation.cpp
@@ -26,7 +26,7 @@ const std::vector<ngraph::pass::low_precision::LayerTransformation::Params> tras
 const std::vector<bool> transparentIntermediateValues = { true, false };
 const std::vector<bool> multiChannelValues = { /*true,*/ false };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatWithIntermediateTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatWithIntermediateTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_neighbors_graph_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_neighbors_graph_transformation.cpp
index c009d2694f02d4..2dc3a2b4fdecd7 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_neighbors_graph_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_neighbors_graph_transformation.cpp
@@ -23,7 +23,7 @@ const std::vector<ngraph::pass::low_precision::LayerTransformation::Params> tras
     LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParamsU8I8()
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatWithNeighborsGraphTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatWithNeighborsGraphTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_split_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_split_transformation.cpp
index cf789286f7f97f..2c969a1775fd4d 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_split_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/concat_with_split_transformation.cpp
@@ -45,7 +45,7 @@ const std::vector<ConcatWithSplitTransformationParam> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConcatWithSplitTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConcatWithSplitTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 6, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
index 697059e3ce24bf..e656b6f3ff54e6 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_backprop_data_transformation.cpp
@@ -111,7 +111,7 @@ const std::vector<ngraph::Shape> outputShapes = {
         { 16, 16 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConvolutionBackpropDataTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConvolutionBackpropDataTransformation,
     ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_qdq_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_qdq_transformation.cpp
index 43928a7ec1caee..bb1bd0cb8dcb5e 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_qdq_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_qdq_transformation.cpp
@@ -238,7 +238,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 4, 3, 4, 4 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConvolutionQDqTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConvolutionQDqTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
index 2a0c8ff14fc7ac..6b76d8a9eaafa2 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/convolution_transformation.cpp
@@ -64,7 +64,7 @@ const std::vector<LayerTestsDefinitions::ConvolutionTransformationParam> params
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConvolutionTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConvolutionTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
@@ -88,7 +88,7 @@ const std::vector<LayerTestsDefinitions::ConvolutionWIthIncorrectWeightsParam> i
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ConvolutionWIthIncorrectWeightsTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ConvolutionWIthIncorrectWeightsTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/depth_to_space_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/depth_to_space_transformation.cpp
index 2075bfe18d76d2..85e076155cc443 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/depth_to_space_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/depth_to_space_transformation.cpp
@@ -34,7 +34,7 @@ const auto DepthToSpaceBS2 = ::testing::Combine(
     ::testing::Values(2)
 );
 
-INSTANTIATE_TEST_CASE_P(LPT_BS2, DepthToSpaceTransformation, DepthToSpaceBS2, DepthToSpaceTransformation::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(LPT_BS2, DepthToSpaceTransformation, DepthToSpaceBS2, DepthToSpaceTransformation::getTestCaseName);
 
 const std::vector<ngraph::Shape> inputShapesBS3 = {
         {1, 9, 3, 3}, {2, 27, 5, 4}
@@ -48,5 +48,5 @@ const auto DepthToSpaceBS3 = ::testing::Combine(
     ::testing::Values(3)
 );
 
-INSTANTIATE_TEST_CASE_P(LPT_BS3, DepthToSpaceTransformation, DepthToSpaceBS3, DepthToSpaceTransformation::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(LPT_BS3, DepthToSpaceTransformation, DepthToSpaceBS3, DepthToSpaceTransformation::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_avg_pool_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_avg_pool_transformation.cpp
index 45ca8e01a7d7c9..ea2587340fff7b 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_avg_pool_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_avg_pool_transformation.cpp
@@ -27,7 +27,7 @@ const std::vector<ngraph::builder::subgraph::FakeQuantizeOnData> fakeQuantizes =
 
 // FakeQuantizeOnData
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizeAndAvgPoolTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizeAndAvgPoolTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 32, 72, 48 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_max_pool_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_max_pool_transformation.cpp
index 56e2ecd29a9ac3..ba1bc5bb20a674 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_max_pool_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_max_pool_transformation.cpp
@@ -27,7 +27,7 @@ const std::vector<ngraph::builder::subgraph::FakeQuantizeOnData> fakeQuantizes =
 
 // FakeQuantizeOnData
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizeAndMaxPoolTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizeAndMaxPoolTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 32, 72, 48 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp
index dc48bcb9941561..fa47c845378a6d 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_and_two_output_branches_with_convolution.cpp
@@ -28,7 +28,7 @@ const std::vector<FakeQuantizeAndTwoOutputBranchesWithConvolution> testValues =
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizeAndTwoOutputBranchesWithConvolutionTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizeAndTwoOutputBranchesWithConvolutionTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 32, 72, 48 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_precision_selection_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_precision_selection_transformation.cpp
index f70b1cdde5fdad..2b67ce0720cd28 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_precision_selection_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_precision_selection_transformation.cpp
@@ -57,7 +57,7 @@ const std::vector<FakeQuantizePrecisionSelectionTransformationTestValues> testVa
 };
 
 // GPU issue
-INSTANTIATE_TEST_CASE_P(DISABLED_LPT, FakeQuantizePrecisionSelectionTransformation,
+INSTANTIATE_TEST_SUITE_P(DISABLED_LPT, FakeQuantizePrecisionSelectionTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 32, 72, 48 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_transformation.cpp
index 62fb144c785e6b..fe2ae3c55d6d7c 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_transformation.cpp
@@ -59,7 +59,7 @@ const std::vector<FakeQuantizeTransformationParam> fakeQuantizeOnDataValues = {
     // { 256ul, { 1ul }, { -1.28f} , { 1.27f } }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizeTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 32, 72, 48 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp
index 963b6e6f28a7a3..6e780a94ed2e09 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fake_quantize_with_dq_not_optimal_transformation.cpp
@@ -101,7 +101,7 @@ const std::vector<FakeQuantizeWithNotOptimalTransformationTestValues> fakeQuanti
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FakeQuantizeWithNotOptimalTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FakeQuantizeWithNotOptimalTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fully_connected_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fully_connected_transformation.cpp
index db3665f6aaf317..45bf73121d2bfa 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fully_connected_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fully_connected_transformation.cpp
@@ -41,7 +41,7 @@ const std::vector<ngraph::pass::low_precision::LayerTransformation::Params> tras
     LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParams()
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FullyConnectedTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FullyConnectedTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_convert_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_convert_transformation.cpp
index 01013dd414758b..537f95363e17df 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_convert_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_convert_transformation.cpp
@@ -32,7 +32,7 @@ const std::vector<ngraph::builder::subgraph::DequantizationOperations> deqOperat
 
 const std::vector<bool> constInput = { true, false };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FuseConvertTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FuseConvertTransformation,
     ::testing::Combine(
             ::testing::ValuesIn(precisions),
             ::testing::ValuesIn(inputAndQuantizationShapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_and_scale_shift_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_and_scale_shift_transformation.cpp
index 175141ef885d45..f78feda5ce6443 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_and_scale_shift_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_and_scale_shift_transformation.cpp
@@ -34,7 +34,7 @@ const std::vector<ngraph::builder::subgraph::FakeQuantizeOnData> fakeQuantizeOnD
 //    },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FuseFakeQuantizeAndScaleShiftTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FuseFakeQuantizeAndScaleShiftTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 9, 9 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_transformation.cpp
index 3125575ec5f1ab..07b0f0a8ad8cd5 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_fake_quantize_transformation.cpp
@@ -104,7 +104,7 @@ const std::vector<FuseFakeQuantizeTransformationTestValues> testValues = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, FuseFakeQuantizeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, FuseFakeQuantizeTransformation,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
         ::testing::ValuesIn(testValues)),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_multiply_to_fake_quantize_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_multiply_to_fake_quantize_transformation.cpp
index 2cc508be12b9dd..58fee1528c2ad1 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_multiply_to_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_multiply_to_fake_quantize_transformation.cpp
@@ -30,7 +30,7 @@ namespace {
         },
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_LPT, FuseMultiplyToFakeQuantizeTransformation,
+    INSTANTIATE_TEST_SUITE_P(smoke_LPT, FuseMultiplyToFakeQuantizeTransformation,
         ::testing::Combine(
             ::testing::Values(CommonTestUtils::DEVICE_GPU),
             ::testing::ValuesIn(testValues)),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_subtract_to_fake_quantize_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_subtract_to_fake_quantize_transformation.cpp
index 35861a9830aeb1..0b25acfc569d04 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_subtract_to_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/fuse_subtract_to_fake_quantize_transformation.cpp
@@ -30,7 +30,7 @@ namespace {
         },
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_LPT, FuseSubtractToFakeQuantizeTransformation,
+    INSTANTIATE_TEST_SUITE_P(smoke_LPT, FuseSubtractToFakeQuantizeTransformation,
         ::testing::Combine(
             ::testing::Values(CommonTestUtils::DEVICE_GPU),
             ::testing::ValuesIn(testValues)),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/gemm_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/gemm_transformation.cpp
index ca0f60b35f4e91..cd2b40ceb26e41 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/gemm_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/gemm_transformation.cpp
@@ -24,7 +24,7 @@ const std::vector<LayerTransformation::Params> trasformationParamValues = {
     LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParamsI8I8()
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, GemmTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, GemmTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(dimensions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
index 624d933188b7fa..18325c824f4f49 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/group_convolution_transformation.cpp
@@ -97,7 +97,7 @@ const std::vector<LayerTestsDefinitions::GroupConvolutionTransformationParam> pa
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, GroupConvolutionTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, GroupConvolutionTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/groupconvolution_qdq_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/groupconvolution_qdq_transformation.cpp
index 5f07e42d4b2ef2..b95d0e7f8c6521 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/groupconvolution_qdq_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/groupconvolution_qdq_transformation.cpp
@@ -492,7 +492,7 @@ const std::vector<ngraph::Shape> shapes = {
     { 1, 4, 7, 7 }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, GroupConvolutionQDqTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, GroupConvolutionQDqTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/interpolate_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/interpolate_transformation.cpp
index c00f6bf695e56a..33135bb451a019 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/interpolate_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/interpolate_transformation.cpp
@@ -48,5 +48,5 @@ const auto combineValues = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_GPU),
     ::testing::ValuesIn(interpAttrs));
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, InterpolateTransformation, combineValues, InterpolateTransformation::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, InterpolateTransformation, combineValues, InterpolateTransformation::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_transformation.cpp
index 5afaefb92be6ec..6d0df41c7f48ab 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_transformation.cpp
@@ -30,7 +30,7 @@ std::vector<MatMulTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MatMulTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MatMulTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 384, 1024 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_with_constant_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_with_constant_transformation.cpp
index bff6d09467bd9a..919f2dd9388b68 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_with_constant_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_with_constant_transformation.cpp
@@ -91,7 +91,7 @@ std::vector<MatMulWithConstantTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MatMulWithConstantTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MatMulWithConstantTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_with_optimized_constant_fake_quantize_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_with_optimized_constant_fake_quantize_transformation.cpp
index a3d2c4eac12975..8759d2903bc228 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_with_optimized_constant_fake_quantize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mat_mul_with_optimized_constant_fake_quantize_transformation.cpp
@@ -27,7 +27,7 @@ const std::vector<std::pair<ngraph::Shape, ngraph::Shape>> inputShapes = {
     std::pair<ngraph::Shape, ngraph::Shape>({ 1, 16 }, { 16, 10 })
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MatMulWithOptimizedConstantFakeQuantizeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MatMulWithOptimizedConstantFakeQuantizeTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_to_group_convolution_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_to_group_convolution_transformation.cpp
index 2bb1abfb9ed980..4f60cc1a3bb765 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_to_group_convolution_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_to_group_convolution_transformation.cpp
@@ -23,7 +23,7 @@ const std::vector<builder::subgraph::FakeQuantizeOnData> fqOnData = {
     { 256ul, ngraph::Shape { 1, 1, 1, 1 }, { 0.f }, { 25.5f }, { -12.8f }, { 12.7f } }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MultiplyToGroupConvolutionTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MultiplyToGroupConvolutionTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_transformation.cpp
index b107cff023fd83..6c294cb6d9c1d2 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_transformation.cpp
@@ -67,7 +67,7 @@ const std::vector<LayerTestsDefinitions::MultiplyTestValues> params = {
     { {}, {}, false }, { {}, {}, true },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MultiplyTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MultiplyTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_with_one_parent_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_with_one_parent_transformation.cpp
index dba8fb64ad7848..083b4e8357b7ff 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_with_one_parent_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/multiply_with_one_parent_transformation.cpp
@@ -21,7 +21,7 @@ const std::vector<MultiplyWithOneParentTransformationValues> values = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, MultiplyWithOneParentTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, MultiplyWithOneParentTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mvn_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mvn_transformation.cpp
index 7599e34484d56b..da5fbff9e91836 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mvn_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/mvn_transformation.cpp
@@ -21,7 +21,7 @@ namespace {
 
     const std::vector<bool> normalizeVariance = { true, false };
 
-    INSTANTIATE_TEST_CASE_P(smoke_LPT, MVNTransformation,
+    INSTANTIATE_TEST_SUITE_P(smoke_LPT, MVNTransformation,
         ::testing::Combine(
             ::testing::ValuesIn(precisions),
             ::testing::ValuesIn(inputAndQuantizationShapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/normalize_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/normalize_transformation.cpp
index 0fa6be4ada6a15..fe30318107a6b6 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/normalize_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/normalize_transformation.cpp
@@ -29,7 +29,7 @@ const std::vector<bool> fuseMultiplyValues = { true, false };
 
 const std::vector<bool> shiftValues = { true, false };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, NormalizeL2Transformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, NormalizeL2Transformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::ValuesIn(inputAndQuantizationShapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations.cpp
index 487a99a379d74d..ebbc0ce20030b6 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations.cpp
@@ -19,7 +19,7 @@ const std::vector<LayerTransformation::Params> trasformationParamValues = {
     LayerTestsUtils::LayerTransformationParamsFactory::createParams()
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, OutputLayersHandlingInTransformations,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, OutputLayersHandlingInTransformations,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat.cpp
index 1901cc4a219bb8..dfc77a5cc4c557 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat.cpp
@@ -19,7 +19,7 @@ const std::vector<LayerTransformation::Params> trasformationParamValues = {
     LayerTestsUtils::LayerTransformationParamsFactory::createParams()
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, OutputLayersHandlingInTransformationsForConcat,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, OutputLayersHandlingInTransformationsForConcat,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat_multi_channel.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat_multi_channel.cpp
index cbf5ccb423c59d..98ea4e3340518d 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat_multi_channel.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/output_layers_handling_in_transformations_for_concat_multi_channel.cpp
@@ -19,7 +19,7 @@ const std::vector<LayerTransformation::Params> trasformationParamValues = {
     LayerTestsUtils::LayerTransformationParamsFactory::createParams()
 };
 
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_LPT, OutputLayersHandlingInTransformationsForConcatMultiChannel,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_LPT, OutputLayersHandlingInTransformationsForConcatMultiChannel,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::SizeVector({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/prelu_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/prelu_transformation.cpp
index e65229a9f81d55..aea8e1be969c10 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/prelu_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/prelu_transformation.cpp
@@ -25,7 +25,7 @@ std::vector<PReluTestValues> testValues = {
 };
 
 // PRelu in low precision is not supported in GPU
-INSTANTIATE_TEST_CASE_P(DISABLED_LPT, PReluTransformation,
+INSTANTIATE_TEST_SUITE_P(DISABLED_LPT, PReluTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/pull_reshape_through_dequantization_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/pull_reshape_through_dequantization_transformation.cpp
index aecad435daafe5..04c82dfaff6cca 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/pull_reshape_through_dequantization_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/pull_reshape_through_dequantization_transformation.cpp
@@ -70,7 +70,7 @@ const std::vector<ngraph::Shape> dequantizationOnWeightElementwiseConstantShapes
     { ngraph::Shape({1, 16}) }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, PullReshapeThroughDequantizationTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, PullReshapeThroughDequantizationTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_max_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_max_transformation.cpp
index 8e4d6d49682e23..f53880f5df795a 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_max_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_max_transformation.cpp
@@ -88,7 +88,7 @@ const std::vector<LayerTestsDefinitions::ReduceMaxTransformationParam> params =
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReduceMaxTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReduceMaxTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_mean_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_mean_transformation.cpp
index 138117498749fc..5b343f98747ee1 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_mean_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_mean_transformation.cpp
@@ -88,7 +88,7 @@ const std::vector<LayerTestsDefinitions::ReduceMeanTransformationParam> params =
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReduceMeanTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReduceMeanTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_min_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_min_transformation.cpp
index 3fcc88fdecd7f4..89aeb83f875b27 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_min_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_min_transformation.cpp
@@ -88,7 +88,7 @@ const std::vector<LayerTestsDefinitions::ReduceMinTransformationParam> params =
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReduceMinTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReduceMinTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_sum_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_sum_transformation.cpp
index 9b2d2fa9341b23..46cd094fcc13fc 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_sum_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reduce_sum_transformation.cpp
@@ -78,7 +78,7 @@ const std::vector<LayerTestsDefinitions::ReduceSumTransformationParam> params =
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReduceSumTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReduceSumTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 10, 10 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/relu_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/relu_transformation.cpp
index 219a64f9c005cf..a1d65bbf658c07 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/relu_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/relu_transformation.cpp
@@ -24,7 +24,7 @@ std::vector<ReluTestValues> testValues = {
     { { 256ul, ngraph::Shape({}), {-12.8f / 2.f}, {12.7f}, {-12.8f / 2.f}, {12.7f} }, true }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReluTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReluTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reshape_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reshape_transformation.cpp
index 46ca94e9cb7e97..a8d40ca957f76f 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reshape_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/reshape_transformation.cpp
@@ -52,7 +52,7 @@ const std::vector<ReshapeTransformationParam> params = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ReshapeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ReshapeTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/shuffle_channels_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/shuffle_channels_transformation.cpp
index e1fae4213e9c0b..19a0cfbd01503a 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/shuffle_channels_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/shuffle_channels_transformation.cpp
@@ -77,7 +77,7 @@ const std::vector<LayerTestsDefinitions::ShuffleChannelsTransformationParam> par
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, ShuffleChannelsTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, ShuffleChannelsTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::ValuesIn(inputShapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/split_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/split_transformation.cpp
index 9f9ce9f544b8cc..18535a437dafee 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/split_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/split_transformation.cpp
@@ -84,7 +84,7 @@ const std::vector<LayerTestsDefinitions::SplitTransformationParam> params = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, SplitTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, SplitTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/squeeze_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/squeeze_transformation.cpp
index 36512fd4e58258..586a1ac9695b18 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/squeeze_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/squeeze_transformation.cpp
@@ -46,7 +46,7 @@ namespace {
         }
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_LPT, SqueezeTransformation,
+    INSTANTIATE_TEST_SUITE_P(smoke_LPT, SqueezeTransformation,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/strided_slice_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/strided_slice_transformation.cpp
index bc974c50570f9f..477fb83a1587c3 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/strided_slice_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/strided_slice_transformation.cpp
@@ -86,7 +86,7 @@ const std::vector<LayerTestsDefinitions::StridedSliceTransformationParam> params
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, StridedSliceTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, StridedSliceTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 24, 24 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/subtract_multiply_to_multiply_add_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/subtract_multiply_to_multiply_add_transformation.cpp
index 41d4109cb5bc90..b9a63e37cfbf85 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/subtract_multiply_to_multiply_add_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/subtract_multiply_to_multiply_add_transformation.cpp
@@ -57,7 +57,7 @@ const std::vector<SubtractMultiplyToMultiplyAddTransformationTestValues> testVal
      },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, SubtractMultiplyToMultiplyAddTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, SubtractMultiplyToMultiplyAddTransformation,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
         ::testing::ValuesIn(testValues)),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/subtract_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/subtract_transformation.cpp
index 4bdb067f4b9116..f58f01d8fc5724 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/subtract_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/subtract_transformation.cpp
@@ -20,7 +20,7 @@ const std::vector<LayerTransformation::Params> trasformationParamValues = {
     LayerTestsUtils::LayerTransformationParamsNGraphFactory::createParams()
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, SubtractTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, SubtractTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/transpose_after_matmul_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/transpose_after_matmul_transformation.cpp
index e3cdb1f93ff173..7377d54b6cd9e0 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/transpose_after_matmul_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/transpose_after_matmul_transformation.cpp
@@ -24,7 +24,7 @@ const std::vector<bool> perTensorValues = { true, false };
 
 const std::vector<bool> transposeChannelDimValues = { true, false };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, TransposeAfterMatMulTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, TransposeAfterMatMulTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/transpose_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/transpose_transformation.cpp
index e35f3cde45ab43..b1913f592d1ff0 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/transpose_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/transpose_transformation.cpp
@@ -41,7 +41,7 @@ const std::vector<TransposeTransformationTestValues> testValues = {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, TransposeTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, TransposeTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(precisions),
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/unsqueeze_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/unsqueeze_transformation.cpp
index 9397d7482ebf74..5bb19861240c52 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/unsqueeze_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/unsqueeze_transformation.cpp
@@ -51,7 +51,7 @@ namespace {
         }
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_LPT, UnsqueezeTransformation,
+    INSTANTIATE_TEST_SUITE_P(smoke_LPT, UnsqueezeTransformation,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/variadic_split_transformation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/variadic_split_transformation.cpp
index 6b83a4a0bf426d..95f4e4487aa4d9 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/variadic_split_transformation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/low_precision_transformations/variadic_split_transformation.cpp
@@ -90,7 +90,7 @@ const std::vector<LayerTestsDefinitions::VariadicSplitTransformationParam> param
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_LPT, VariadicSplitTransformation,
+INSTANTIATE_TEST_SUITE_P(smoke_LPT, VariadicSplitTransformation,
     ::testing::Combine(
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(ngraph::Shape({ 1, 3, 16, 16 })),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/multi/gpu_remote_blob_tests.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/multi/gpu_remote_blob_tests.cpp
index 4fffb2cad6e157..26e2655607d283 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/multi/gpu_remote_blob_tests.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/multi/gpu_remote_blob_tests.cpp
@@ -16,7 +16,7 @@ const std::vector<DevicesNamesAndSupportPair> device_names_and_support_for_remot
 #endif
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_RemoteBlobMultiGPU, MultiDevice_SupportTest,
+INSTANTIATE_TEST_SUITE_P(smoke_RemoteBlobMultiGPU, MultiDevice_SupportTest,
                         ::testing::ValuesIn(device_names_and_support_for_remote_blobs), MultiDevice_SupportTest::getTestCaseName);
 
 TEST_P(MultiDevice_Test, cannotInferRemoteBlobIfNotInitializedForDevice) {
@@ -53,5 +53,5 @@ const std::vector<DevicesNames> device_names_and_support_for_remote_blobs2 {
         {"GPU.1"},  // another GPU (the test will test its presence), different OCL contexts
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_RemoteBlobMultiInitializedWithoutGPU, MultiDevice_Test,
+INSTANTIATE_TEST_SUITE_P(smoke_RemoteBlobMultiInitializedWithoutGPU, MultiDevice_Test,
                         ::testing::ValuesIn(device_names_and_support_for_remote_blobs2), MultiDevice_Test::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/ngraph_conversion_tests/conv_bias_fusion.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/ngraph_conversion_tests/conv_bias_fusion.cpp
index 37471e8cc67041..fe6a2dda21374b 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/ngraph_conversion_tests/conv_bias_fusion.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/ngraph_conversion_tests/conv_bias_fusion.cpp
@@ -9,6 +9,6 @@ using namespace NGraphConversionTestsDefinitions;
 
 namespace {
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, ConvBiasFusion, ::testing::Values(CommonTestUtils::DEVICE_GPU), ConvBiasFusion::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, ConvBiasFusion, ::testing::Values(CommonTestUtils::DEVICE_GPU), ConvBiasFusion::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/ngraph_conversion_tests/plugin_specific_ngraph_conversion.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/ngraph_conversion_tests/plugin_specific_ngraph_conversion.cpp
index bbfc01e3f1877f..3774936fe78725 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/ngraph_conversion_tests/plugin_specific_ngraph_conversion.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/ngraph_conversion_tests/plugin_specific_ngraph_conversion.cpp
@@ -9,6 +9,6 @@ using namespace NGraphConversionTestsDefinitions;
 
 namespace {
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, PluginSpecificConversion, ::testing::Values("GPU"), PluginSpecificConversion::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, PluginSpecificConversion, ::testing::Values("GPU"), PluginSpecificConversion::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/activation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/activation.cpp
index e7e1f8a63440a1..c62641c0b4cab7 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/activation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/activation.cpp
@@ -95,9 +95,9 @@ const auto basicPreluCases = ::testing::Combine(
 );
 
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic_Prelu, ActivationLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic_Prelu, ActivationLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationParamLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic, ActivationParamLayerTest, basicPreluCases, ActivationLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/batch_norm.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/batch_norm.cpp
index 94c15c93018ce4..fe89b264d9a3ea 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/batch_norm.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/batch_norm.cpp
@@ -39,7 +39,7 @@ const auto batchNormParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_BatchNorm,
         BatchNormLayerTest,
         batchNormParams,
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/batch_to_space.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/batch_to_space.cpp
index 7194d632344120..1a05db98a6a4f8 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/batch_to_space.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/batch_to_space.cpp
@@ -119,6 +119,6 @@ batchToSpaceParamsTuple bts_only_test_cases[] = {
                                 CommonTestUtils::DEVICE_GPU)
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_CLDNN, BatchToSpaceLayerTest, ::testing::ValuesIn(bts_only_test_cases),
+INSTANTIATE_TEST_SUITE_P(smoke_CLDNN, BatchToSpaceLayerTest, ::testing::ValuesIn(bts_only_test_cases),
                         BatchToSpaceLayerTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/broadcast.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/broadcast.cpp
index aa055e9b5282f0..979552b9933b54 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/broadcast.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/broadcast.cpp
@@ -38,7 +38,7 @@ const auto numpyBroadcastParams1 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestNumpyBroadcast1,
         BroadcastLayerTest,
         numpyBroadcastParams1,
@@ -54,7 +54,7 @@ const auto numpyBroadcastParams2 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestNumpyBroadcast2,
         BroadcastLayerTest,
         numpyBroadcastParams2,
@@ -84,7 +84,7 @@ const auto bidirectionalBroadcastParams1 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestBidirectionalBroadcast1,
         BroadcastLayerTest,
         bidirectionalBroadcastParams1,
@@ -100,7 +100,7 @@ const auto bidirectionalBroadcastParams2 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestBidirectionalBroadcast2,
         BroadcastLayerTest,
         bidirectionalBroadcastParams2,
@@ -116,7 +116,7 @@ const auto bidirectionalBroadcastParams3 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestBidirectionalBroadcast3,
         BroadcastLayerTest,
         bidirectionalBroadcastParams3,
@@ -149,7 +149,7 @@ const auto explicitBroadcastParams1 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestExplicitBroadcast1,
         BroadcastLayerTest,
         explicitBroadcastParams1,
@@ -165,7 +165,7 @@ const auto explicitBroadcastParams2 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_TestExplicitBroadcast2,
         BroadcastLayerTest,
         explicitBroadcastParams2,
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/comparison.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/comparison.cpp
index 877bd366e0e9a6..e92cdf4f41763b 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/comparison.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/comparison.cpp
@@ -50,6 +50,6 @@ const auto ComparisonTestParams = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
         ::testing::Values(additional_config));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs, ComparisonLayerTest, ComparisonTestParams, ComparisonLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs, ComparisonLayerTest, ComparisonTestParams, ComparisonLayerTest::getTestCaseName);
 
 }  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/concat.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/concat.cpp
index bde5e2fc19bbe7..a57887e900fbab 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/concat.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/concat.cpp
@@ -24,7 +24,7 @@ std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precis
                                                          InferenceEngine::Precision::I64};
 
 
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, ConcatLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, ConcatLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(axes),
                                 ::testing::ValuesIn(inShapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convert.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convert.cpp
index 130360d4cdf472..e404be39af6039 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convert.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convert.cpp
@@ -19,7 +19,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::I8,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, ConvertLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, ConvertLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(inShape),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convert_like.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convert_like.cpp
index 8b0208ea8cc51c..c55dc9d85bd6a6 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convert_like.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convert_like.cpp
@@ -19,7 +19,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::I8,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, ConvertLikeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, ConvertLikeLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(inShape),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution.cpp
index 6b4c1bea5cf523..94e77a1a4a7201 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution.cpp
@@ -50,7 +50,7 @@ const auto conv2DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::VALID)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_ExplicitPadding, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D_ExplicitPadding, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -62,7 +62,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_ExplicitPadding, ConvolutionLayerTes
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_AutoPadValid, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D_AutoPadValid, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -97,7 +97,7 @@ const auto conv3DParams = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::EXPLICIT)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution3D_Basic1, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution3D_Basic1, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv3DParams,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
index f404d932bc86a0..a34a5b75e1b711 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/convolution_backprop_data.cpp
@@ -56,7 +56,7 @@ const auto conv2DParams_AutoPadValid = ::testing::Combine(
         ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions2D),
@@ -69,7 +69,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding, Convolu
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D_AutoPadValid, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions2D),
@@ -106,7 +106,7 @@ const auto conv2DParams_AutoPadValid_output_padding = ::testing::Combine(
         ::testing::ValuesIn(outputPadding2D)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid_output_padding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -119,7 +119,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_ExplicitPadding_OutputPa
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData2D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData2D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding_output_padding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -166,7 +166,7 @@ const auto conv3DParams_AutoPadValid = ::testing::Combine(
         ::testing::ValuesIn(emptyOutputPadding)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions3D),
@@ -179,7 +179,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding, Convolu
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData3D_AutoPadValid, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions3D),
@@ -216,7 +216,7 @@ const auto conv3DParams_AutoPadValid_output_padding = ::testing::Combine(
         ::testing::ValuesIn(outputPadding3D)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_AutoPadValid_output_padding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -229,7 +229,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_ExplicitPadding_OutputPa
                                 ::testing::Values(CommonTestUtils::DEVICE_CPU)),
                         ConvolutionBackpropLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ConvolutionBackpropData3D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ConvolutionBackpropData3D_AutoPadding_OutputPaddingDefined, ConvolutionBackpropLayerTest,
                         ::testing::Combine(
                                 conv3DParams_ExplicitPadding_output_padding,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder.cpp
index a9faeddaa46325..3c423614217ab0 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder.cpp
@@ -34,7 +34,7 @@ const auto basicCases = ::testing::Combine(
     ::testing::ValuesIn(mergeRepeated),
     ::testing::Values(CommonTestUtils::DEVICE_GPU));
 
-INSTANTIATE_TEST_CASE_P(smoke_CtcGreedyDecoderBasic, CTCGreedyDecoderLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_CtcGreedyDecoderBasic, CTCGreedyDecoderLayerTest,
                         basicCases,
                         CTCGreedyDecoderLayerTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp
index 497686924b2af4..04042aa4de5020 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp
@@ -33,11 +33,11 @@ const auto basicCases = ::testing::Combine(
     ::testing::ValuesIn(mergeRepeated),
     ::testing::Values(CommonTestUtils::DEVICE_GPU));
 
-INSTANTIATE_TEST_CASE_P(smoke_set1, CTCGreedyDecoderSeqLenLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_set1, CTCGreedyDecoderSeqLenLayerTest,
                         basicCases,
                         CTCGreedyDecoderSeqLenLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_set2, CTCGreedyDecoderSeqLenLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_set2, CTCGreedyDecoderSeqLenLayerTest,
         ::testing::Combine(
                         ::testing::ValuesIn(std::vector<std::vector<size_t>>{{2, 8, 11}, {4, 10, 55}}),
                         ::testing::ValuesIn(std::vector<int>{5, 100}),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/cum_sum.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/cum_sum.cpp
index b92a453693f27e..12cebb9f8d4565 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/cum_sum.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/cum_sum.cpp
@@ -26,7 +26,7 @@ std::vector<InferenceEngine::Precision> precisions = {InferenceEngine::Precision
 
 std::vector<std::vector<size_t>> shape1d = {{10}};
 std::vector<int64_t> axis1d = {0};
-INSTANTIATE_TEST_CASE_P(smoke_CumSum1D, CumSumLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_CumSum1D, CumSumLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(shape1d),
                                 ::testing::ValuesIn(precisions),
@@ -36,7 +36,7 @@ INSTANTIATE_TEST_CASE_P(smoke_CumSum1D, CumSumLayerTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         CumSumLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_CumSum, CumSumLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_CumSum, CumSumLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inShapes),
                                 ::testing::ValuesIn(precisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/depth_to_space.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/depth_to_space.cpp
index 4e42493d567978..89b8bc7de845d1 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/depth_to_space.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/depth_to_space.cpp
@@ -34,7 +34,7 @@ const auto DepthToSpaceBS2 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DepthToSpaceBS2, DepthToSpaceLayerTest, DepthToSpaceBS2, DepthToSpaceLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DepthToSpaceBS2, DepthToSpaceLayerTest, DepthToSpaceBS2, DepthToSpaceLayerTest::getTestCaseName);
 
 const std::vector<std::vector<size_t >> inputShapesBS3 = {
         {1, 9, 1, 1}, {1, 9, 2, 2}, {1, 9, 3, 3}, {2, 36, 3, 3}, {2, 27, 5, 4},
@@ -48,6 +48,6 @@ const auto DepthToSpaceBS3 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DepthToSpaceBS3, DepthToSpaceLayerTest, DepthToSpaceBS3, DepthToSpaceLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DepthToSpaceBS3, DepthToSpaceLayerTest, DepthToSpaceBS3, DepthToSpaceLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/detection_output.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/detection_output.cpp
index e12fdbc6ea2779..7be886fbaadc42 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/detection_output.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/detection_output.cpp
@@ -56,7 +56,7 @@ const auto params3Inputs = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DetectionOutput3In, DetectionOutputLayerTest, params3Inputs, DetectionOutputLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DetectionOutput3In, DetectionOutputLayerTest, params3Inputs, DetectionOutputLayerTest::getTestCaseName);
 
 /* =============== 5 inputs cases =============== */
 
@@ -80,6 +80,6 @@ const auto params5Inputs = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_DetectionOutput5In, DetectionOutputLayerTest, params5Inputs, DetectionOutputLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_DetectionOutput5In, DetectionOutputLayerTest, params5Inputs, DetectionOutputLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/eltwise.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/eltwise.cpp
index 3cd2feebf58399..fc411847cb9159 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/eltwise.cpp
@@ -66,7 +66,7 @@ const auto multiply_params = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
         ::testing::Values(additional_config));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs, EltwiseLayerTest, multiply_params, EltwiseLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs, EltwiseLayerTest, multiply_params, EltwiseLayerTest::getTestCaseName);
 
 
 std::vector<std::vector<std::vector<size_t>>> inShapesSingleThread = {
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_bag_offsets_sum.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_bag_offsets_sum.cpp
index afb5945fa19cd9..cc88c9527bd580 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_bag_offsets_sum.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_bag_offsets_sum.cpp
@@ -49,7 +49,7 @@ const auto embBagOffsetSumArgSet = ::testing::Combine(
     ::testing::ValuesIn(offsets), ::testing::ValuesIn(default_index),
     ::testing::ValuesIn(with_weights), ::testing::ValuesIn(with_default_index));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke, EmbeddingBagOffsetsSumLayerTest,
     ::testing::Combine(embBagOffsetSumArgSet,
                        ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_bag_packed_sum.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_bag_packed_sum.cpp
index 2981ac862663d2..c7379c8bb8ff86 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_bag_packed_sum.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_bag_packed_sum.cpp
@@ -39,7 +39,7 @@ const auto embBagPackedSumArgSet = ::testing::Combine(
     ::testing::ValuesIn(emb_table_shape), ::testing::ValuesIn(indices),
     ::testing::ValuesIn(with_weights));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke, EmbeddingBagPackedSumLayerTest,
     ::testing::Combine(embBagPackedSumArgSet,
                        ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_segments_sum.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_segments_sum.cpp
index 34caed6822dce9..5239daf682057b 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_segments_sum.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/embedding_segments_sum.cpp
@@ -45,7 +45,7 @@ const auto embSegmentsSumArgSet = ::testing::Combine(
     ::testing::ValuesIn(default_index), ::testing::ValuesIn(with_weights),
     ::testing::ValuesIn(with_default_index));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke, EmbeddingSegmentsSumLayerTest,
     ::testing::Combine(embSegmentsSumArgSet, ::testing::ValuesIn(netPrecisions),
                        ::testing::ValuesIn(indPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/extract_image_patches.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/extract_image_patches.cpp
index 1749c0db0963c7..ca6db6370842c0 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/extract_image_patches.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/extract_image_patches.cpp
@@ -56,7 +56,7 @@ const auto extractImagePatchesParamsSet = ::testing::Combine(
         ::testing::ValuesIn(autoPads)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_layers_GPU, ExtractImagePatchesTest,
+INSTANTIATE_TEST_SUITE_P(smoke_layers_GPU, ExtractImagePatchesTest,
         ::testing::Combine(
             ::testing::ValuesIn(inDataShape),
             ::testing::ValuesIn(kernels),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/fake_quantize.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/fake_quantize.cpp
index ba1e5a6ca67136..a5223cb641c697 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/fake_quantize.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/fake_quantize.cpp
@@ -32,7 +32,7 @@ const auto fqParams = ::testing::Combine(
         ::testing::Values(inputParams)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_FakeQuantize, FakeQuantizeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_FakeQuantize, FakeQuantizeLayerTest,
                         ::testing::Combine(
                                 fqParams,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather.cpp
index 2d98eba92b5aa4..6af03cf397d85c 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather.cpp
@@ -126,28 +126,28 @@ const auto GatherAxes4i8b2 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Gather7Axes4i4b1,
         Gather7LayerTest,
         GatherAxes4i4b1,
         Gather7LayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Gather7Axes4i4b2,
         Gather7LayerTest,
         GatherAxes4i4b1,
         Gather7LayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Gather7Axes4i8b1,
         Gather7LayerTest,
         GatherAxes4i8b1,
         Gather7LayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Gather7Axes4i8b2,
         Gather7LayerTest,
         GatherAxes4i8b2,
@@ -232,7 +232,7 @@ const auto GatherAxes4 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_GatherAxes4,
         GatherLayerTest,
         GatherAxes4,
@@ -252,7 +252,7 @@ const auto Gather6dAxes4 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Gather6dAxes4,
         GatherLayerTest,
         Gather6dAxes4,
@@ -300,7 +300,7 @@ const auto GatherAxes3 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_GatherAxes3,
         GatherLayerTest,
         GatherAxes3,
@@ -320,7 +320,7 @@ const auto Gather6dAxes3 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Gather6dAxes3,
         GatherLayerTest,
         Gather6dAxes3,
@@ -369,7 +369,7 @@ const auto GatherAxes2 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_GatherAxes2,
         GatherLayerTest,
         GatherAxes2,
@@ -389,7 +389,7 @@ const auto Gather6dAxes2 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Gather6dAxes2,
         GatherLayerTest,
         Gather6dAxes2,
@@ -439,7 +439,7 @@ const auto GatherAxes1 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_GatherAxes1,
         GatherLayerTest,
         GatherAxes1,
@@ -459,7 +459,7 @@ const auto Gather6dAxes1 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Gather6dAxes1,
         GatherLayerTest,
         Gather6dAxes1,
@@ -510,7 +510,7 @@ const auto GatherAxes0 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_GatherAxes0,
         GatherLayerTest,
         GatherAxes0,
@@ -530,7 +530,7 @@ const auto Gather6dAxes0 = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Gather6dAxes0,
         GatherLayerTest,
         Gather6dAxes0,
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather_nd.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather_nd.cpp
index be599acaef41ca..83cd36196537de 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather_nd.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather_nd.cpp
@@ -33,7 +33,7 @@ const auto gatherNDArgsSubset1 = ::testing::Combine(
     ::testing::ValuesIn(std::vector<int>({ 0, 1 }))             // Batch dims
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GatherND_set1, GatherNDLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GatherND_set1, GatherNDLayerTest,
     ::testing::Combine(
         gatherNDArgsSubset1,
         ::testing::ValuesIn(inputPrecisions),
@@ -51,7 +51,7 @@ const auto gatherNDArgsSubset2 = ::testing::Combine(
     ::testing::ValuesIn(std::vector<int>({ 1, 2 }))             // Batch dims
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GatherND_set2, GatherNDLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GatherND_set2, GatherNDLayerTest,
     ::testing::Combine(
         gatherNDArgsSubset2,
         ::testing::ValuesIn(inputPrecisions),
@@ -69,7 +69,7 @@ const auto gatherNDArgsSubset3 = ::testing::Combine(
     ::testing::ValuesIn(std::vector<int>({ 3, 4 }))             // Batch dims
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GatherND_set3, GatherNDLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GatherND_set3, GatherNDLayerTest,
     ::testing::Combine(
         gatherNDArgsSubset3,
         ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather_tree.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather_tree.cpp
index ee09b1fa36a73a..a1b5488a1b670d 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather_tree.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gather_tree.cpp
@@ -23,7 +23,7 @@ const std::vector<ngraph::helpers::InputLayerType> secondaryInputTypes = {
         ngraph::helpers::InputLayerType::PARAMETER
 };
 
-INSTANTIATE_TEST_CASE_P(Basic_smoke, GatherTreeLayerTest,
+INSTANTIATE_TEST_SUITE_P(Basic_smoke, GatherTreeLayerTest,
         ::testing::Combine(
         ::testing::ValuesIn(inputShapes),
         ::testing::ValuesIn(secondaryInputTypes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/grn.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/grn.cpp
index 52c0ed6c59b140..84dd2994e85675 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/grn.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/grn.cpp
@@ -27,7 +27,7 @@ namespace {
         ::testing::Values(0.33f, 1.1f),
         ::testing::Values(CommonTestUtils::DEVICE_GPU));
 
-    INSTANTIATE_TEST_CASE_P(smoke_Grn_Basic, GrnLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Grn_Basic, GrnLayerTest,
                             basicCases,
                             GrnLayerTest::getTestCaseName);
 
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/group_convolution.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/group_convolution.cpp
index 75a28f94315714..603433d1bfdf43 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/group_convolution.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/group_convolution.cpp
@@ -45,7 +45,7 @@ const auto groupConv2DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::VALID)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution2D_ExplicitPadding, GroupConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolution2D_ExplicitPadding, GroupConvolutionLayerTest,
                         ::testing::Combine(
                                 groupConv2DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -57,7 +57,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution2D_ExplicitPadding, GroupConvoluti
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         GroupConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution2D_AutoPadValid, GroupConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolution2D_AutoPadValid, GroupConvolutionLayerTest,
                         ::testing::Combine(
                                 groupConv2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -97,7 +97,7 @@ const auto groupConv3DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::VALID)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution3D_ExplicitPadding, GroupConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolution3D_ExplicitPadding, GroupConvolutionLayerTest,
                         ::testing::Combine(
                                 groupConv3DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -109,7 +109,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution3D_ExplicitPadding, GroupConvoluti
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         GroupConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvolution3D_AutoPadValid, GroupConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvolution3D_AutoPadValid, GroupConvolutionLayerTest,
                         ::testing::Combine(
                                 groupConv3DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/group_convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/group_convolution_backprop_data.cpp
index 40e751e817a899..5e4b6db1501810 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/group_convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/group_convolution_backprop_data.cpp
@@ -48,7 +48,7 @@ const auto groupConvBackpropData2DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::VALID)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData2D_ExplicitPadding, GroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvBackpropData2D_ExplicitPadding, GroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 groupConvBackpropData2DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -60,7 +60,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData2D_ExplicitPadding, GroupConv
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         GroupConvBackpropDataLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData2D_AutoPadValid, GroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvBackpropData2D_AutoPadValid, GroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 groupConvBackpropData2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -102,7 +102,7 @@ const auto groupConvBackpropData3DParams_AutoPadValid = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::VALID)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData3D_ExplicitPadding, GroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvBackpropData3D_ExplicitPadding, GroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 groupConvBackpropData3DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -114,7 +114,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData3D_ExplicitPadding, GroupConv
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         GroupConvBackpropDataLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GroupConvBackpropData3D_AutoPadValid, GroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GroupConvBackpropData3D_AutoPadValid, GroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 groupConvBackpropData3DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gru_cell.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gru_cell.cpp
index 7224b098c9fb98..0188040bfbdb28 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gru_cell.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gru_cell.cpp
@@ -21,7 +21,7 @@ namespace {
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                              InferenceEngine::Precision::FP16};
 
-    INSTANTIATE_TEST_CASE_P(GRUCellCommon, GRUCellTest,
+    INSTANTIATE_TEST_SUITE_P(GRUCellCommon, GRUCellTest,
             ::testing::Combine(
             ::testing::ValuesIn(should_decompose),
             ::testing::ValuesIn(batch),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gru_sequence.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gru_sequence.cpp
index d0033f4f4738c3..7b613dadc91c38 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gru_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/gru_sequence.cpp
@@ -34,7 +34,7 @@ namespace {
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                              InferenceEngine::Precision::FP16};
 
-    INSTANTIATE_TEST_CASE_P(GRUSequenceCommonZeroClip, GRUSequenceTest,
+    INSTANTIATE_TEST_SUITE_P(GRUSequenceCommonZeroClip, GRUSequenceTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(mode),
                                     ::testing::ValuesIn(seq_lengths_zero_clip),
@@ -49,7 +49,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                             GRUSequenceTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(GRUSequenceCommonClip, GRUSequenceTest,
+    INSTANTIATE_TEST_SUITE_P(GRUSequenceCommonClip, GRUSequenceTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(mode),
                                     ::testing::ValuesIn(seq_lengths_clip_non_zero),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/interpolate.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/interpolate.cpp
index 0b210ebef2e4db..c26f168d62848a 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/interpolate.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/interpolate.cpp
@@ -108,7 +108,7 @@ const auto interpolateCasesNearesMode = ::testing::Combine(
         ::testing::ValuesIn(defaultAxes),
         ::testing::ValuesIn(defaultScales));
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_Basic, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_Basic, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesWithoutNearest,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -121,7 +121,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Interpolate_Basic, InterpolateLayerTest, ::testing
         ::testing::Values(additional_config)),
     InterpolateLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_Nearest, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_Nearest, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesNearesMode,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/log_softmax.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/log_softmax.cpp
index eb62a64128029e..b84f9736077590 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/log_softmax.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/log_softmax.cpp
@@ -37,7 +37,7 @@ const auto params2D = testing::Combine(
     testing::Values(std::map<std::string, std::string>())
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_LogSoftmax2D,
         LogSoftmaxLayerTest,
         params2D,
@@ -66,7 +66,7 @@ const auto params4D = testing::Combine(
     testing::Values(std::map<std::string, std::string>())
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_LogSoftmax4D,
         LogSoftmaxLayerTest,
         params4D,
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/logical.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/logical.cpp
index c8dd9db4c2e1ad..588cebf8a3966b 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/logical.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/logical.cpp
@@ -74,8 +74,8 @@ const auto LogicalTestParamsNot = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
         ::testing::Values(additional_config));
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefs, LogicalLayerTest, LogicalTestParams, LogicalLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs, LogicalLayerTest, LogicalTestParams, LogicalLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_CompareWithRefsNot, LogicalLayerTest, LogicalTestParamsNot, LogicalLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefsNot, LogicalLayerTest, LogicalTestParamsNot, LogicalLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lrn.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lrn.cpp
index b513e61a137421..d9b7099399d4c3 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lrn.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lrn.cpp
@@ -21,7 +21,7 @@ const double beta = 2;
 const double bias = 1.0;
 const size_t size = 5;
 
-INSTANTIATE_TEST_CASE_P(smoke_LrnCheck, LrnLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_LrnCheck, LrnLayerTest,
                         ::testing::Combine(::testing::Values(alpha),
                                            ::testing::Values(beta),
                                            ::testing::Values(bias),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lstm_cell.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lstm_cell.cpp
index b9267eb552d08a..855a0e432621de 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lstm_cell.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lstm_cell.cpp
@@ -23,7 +23,7 @@ std::vector<float> clip{0.f, 0.7f};
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                         InferenceEngine::Precision::FP16};
 
-INSTANTIATE_TEST_CASE_P(LSTMCellCommon, LSTMCellTest,
+INSTANTIATE_TEST_SUITE_P(LSTMCellCommon, LSTMCellTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(should_decompose),
                                 ::testing::ValuesIn(batch),
@@ -35,7 +35,7 @@ INSTANTIATE_TEST_CASE_P(LSTMCellCommon, LSTMCellTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         LSTMCellTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_LSTMCellCommon, LSTMCellTest,
+INSTANTIATE_TEST_SUITE_P(smoke_LSTMCellCommon, LSTMCellTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(should_decompose),
                                 ::testing::ValuesIn(batch),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lstm_sequence.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lstm_sequence.cpp
index d79cee69eea5d5..604931349eca9a 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lstm_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lstm_sequence.cpp
@@ -36,7 +36,7 @@ std::vector<ngraph::op::RecurrentSequenceDirection> direction = {ngraph::op::Rec
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                          InferenceEngine::Precision::FP16};
 
-INSTANTIATE_TEST_CASE_P(LSTMSequenceCommonZeroClip, LSTMSequenceTest,
+INSTANTIATE_TEST_SUITE_P(LSTMSequenceCommonZeroClip, LSTMSequenceTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(mode),
                                 ::testing::ValuesIn(seq_lengths_zero_clip),
@@ -50,7 +50,7 @@ INSTANTIATE_TEST_CASE_P(LSTMSequenceCommonZeroClip, LSTMSequenceTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         LSTMSequenceTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(LSTMSequenceCommonClip, LSTMSequenceTest,
+INSTANTIATE_TEST_SUITE_P(LSTMSequenceCommonClip, LSTMSequenceTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(mode),
                                 ::testing::ValuesIn(seq_lengths_clip_non_zero),
@@ -64,7 +64,7 @@ INSTANTIATE_TEST_CASE_P(LSTMSequenceCommonClip, LSTMSequenceTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         LSTMSequenceTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_LSTMSequenceCommonClip, LSTMSequenceTest,
+INSTANTIATE_TEST_SUITE_P(smoke_LSTMSequenceCommonClip, LSTMSequenceTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(mode),
                                 ::testing::ValuesIn(seq_lengths_clip_non_zero),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/mat_mul.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/mat_mul.cpp
index 88a05601e2243f..757bd2955b264e 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/mat_mul.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/mat_mul.cpp
@@ -26,7 +26,7 @@ std::vector<ngraph::helpers::InputLayerType> secondaryInputTypes = {
 
 std::map<std::string, std::string> additional_config = {};
 
-INSTANTIATE_TEST_CASE_P(smoke_MatMul, MatMulTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MatMul, MatMulTest,
         ::testing::Combine(
                 ::testing::ValuesIn(shapeRelatedParams),
                 ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/minimum_maximum.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/minimum_maximum.cpp
index 8c1ba19aa05f15..e2af2e0d970c85 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/minimum_maximum.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/minimum_maximum.cpp
@@ -35,7 +35,7 @@ const std::vector<ngraph::helpers::InputLayerType> inputType = {
         ngraph::helpers::InputLayerType::PARAMETER,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_maximum, MaxMinLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_maximum, MaxMinLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inShapes),
                                 ::testing::ValuesIn(opType),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/mvn.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/mvn.cpp
index 97d1335c813d70..a4e83dc98732c4 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/mvn.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/mvn.cpp
@@ -47,7 +47,7 @@ const auto MvnCases = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_CLDNN_TestsMVN, MvnLayerTest, MvnCases, MvnLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CLDNN_TestsMVN, MvnLayerTest, MvnCases, MvnLayerTest::getTestCaseName);
 
 std::vector<InferenceEngine::Precision> dataPrecisions = {
     InferenceEngine::Precision::FP32,
@@ -68,7 +68,7 @@ const std::vector<float> epsilonF = {
     0.0001
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_5D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_5D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 10, 5, 7, 8}, {1, 3, 8, 9, 49}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -80,7 +80,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVN_5D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_4D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_4D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 10, 5, 17}, {1, 3, 8, 9}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -92,7 +92,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVN_4D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_3D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_3D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 32, 17}, {1, 37, 9}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -104,7 +104,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVN_3D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_2D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_2D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{3, 5}, {2, 55}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -116,7 +116,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVN_2D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Decomposition_1D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Decomposition_1D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{3}, {9}, {55}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -128,7 +128,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Decomposition_1D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Decomposition_3D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Decomposition_3D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 32, 17}, {1, 37, 9}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -140,7 +140,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Decomposition_3D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Decomposition_4D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Decomposition_4D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 16, 5, 8}, {2, 19, 5, 10}}),
                             ::testing::ValuesIn(dataPrecisions),
@@ -152,7 +152,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Decomposition_4D, Mvn6LayerTest,
                             ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Decomposition_6D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Decomposition_6D, Mvn6LayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(std::vector<std::vector<size_t>>{{1, 3, 5, 4, 2, 6}}),
                             ::testing::ValuesIn(dataPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/non_max_suppression.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/non_max_suppression.cpp
index 47abf07e3ce53c..c402ba66ec441f 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/non_max_suppression.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/non_max_suppression.cpp
@@ -39,4 +39,4 @@ const auto nmsParams = ::testing::Combine(::testing::ValuesIn(inShapeParams),
                                           ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_NmsLayerTest, NmsLayerTest, nmsParams, NmsLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_NmsLayerTest, NmsLayerTest, nmsParams, NmsLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/normalize_l2.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/normalize_l2.cpp
index 438e26ed039fd1..6a3632f00f0a03 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/normalize_l2.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/normalize_l2.cpp
@@ -34,7 +34,7 @@ const auto normL2params = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_NormalizeL2,
         NormalizeL2LayerTest,
         normL2params,
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/pad.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/pad.cpp
index 1d2d3ab67913e2..8e8a3f018aaf0d 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/pad.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/pad.cpp
@@ -36,7 +36,7 @@ const auto pad2DConstparams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad2DConst,
         PadLayerTest,
         pad2DConstparams,
@@ -56,7 +56,7 @@ const auto pad2Dparams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad2D,
         PadLayerTest,
         pad2Dparams,
@@ -79,7 +79,7 @@ const auto pad4DConstparams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad4DConst,
         PadLayerTest,
         pad4DConstparams,
@@ -99,7 +99,7 @@ const auto pad4Dparams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad4D,
         PadLayerTest,
         pad4Dparams,
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/pooling.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/pooling.cpp
index ff326a2d0a4c78..bf4c365ae954e9 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/pooling.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/pooling.cpp
@@ -40,7 +40,7 @@ const auto maxPool_ExplicitPad_FloorRounding_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_ExplicitPad_FloorRpunding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_ExplicitPad_FloorRpunding, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_ExplicitPad_FloorRounding_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -65,7 +65,7 @@ const auto maxPool_ExplicitPad_CeilRounding_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_ExplicitPad_CeilRpunding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_ExplicitPad_CeilRpunding, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_ExplicitPad_CeilRounding_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -92,7 +92,7 @@ const auto avgPoolExplicitPadCeilRoundingParams = ::testing::Combine(
         ::testing::Values(true, false)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_ExplicitPad_CeilRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_ExplicitPad_CeilRounding, PoolingLayerTest,
                        ::testing::Combine(
                                avgPoolExplicitPadCeilRoundingParams,
                                ::testing::ValuesIn(netPrecisions),
@@ -117,7 +117,7 @@ const auto avgPoolExplicitPadFloorRoundingParams = ::testing::Combine(
 );
 
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_ExplicitPad_FloorRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_ExplicitPad_FloorRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 avgPoolExplicitPadFloorRoundingParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -143,7 +143,7 @@ const auto allPools_ValidPad_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MAX_and_AVGPool_ValidPad, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MAX_and_AVGPool_ValidPad, PoolingLayerTest,
                         ::testing::Combine(
                                 allPools_ValidPad_Params,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/power.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/power.cpp
index 288b641754eaa1..f66634b7782624 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/power.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/power.cpp
@@ -34,7 +34,7 @@ namespace {
                                                              InferenceEngine::Precision::FP16,
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_power, PowerLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_power, PowerLayerTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(inShapes),
                                     ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/prior_box_clustered.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/prior_box_clustered.cpp
index 171b85bc7740b1..108deb8d3f206c 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/prior_box_clustered.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/prior_box_clustered.cpp
@@ -56,7 +56,7 @@ const auto layerSpeficParams = ::testing::Combine(
     ::testing::ValuesIn(variances)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_PriorBoxClustered_Basic, PriorBoxClusteredLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_PriorBoxClustered_Basic, PriorBoxClusteredLayerTest,
                         ::testing::Combine(
                             layerSpeficParams,
                             ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/proposal.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/proposal.cpp
index 7562964b5e8498..4983c5a4b8fd3b 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/proposal.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/proposal.cpp
@@ -39,7 +39,7 @@ const auto proposalParams = ::testing::Combine(
         ::testing::ValuesIn(framework_)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Proposal_tests, ProposalLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Proposal_tests, ProposalLayerTest,
                         ::testing::Combine(
                                 proposalParams,
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reduce_ops.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reduce_ops.cpp
index 041f1040403741..0641408588c111 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reduce_ops.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reduce_ops.cpp
@@ -169,56 +169,56 @@ const auto params_ReductionTypesLogical = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceOneAxis,
         ReduceOpsLayerTest,
         paramsOneAxis,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceLogicalOneAxis,
         ReduceOpsLayerTest,
         paramsOneAxisLogical,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_Precisions,
         ReduceOpsLayerTest,
         params_Precisions,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_InputShapes,
         ReduceOpsLayerTest,
         params_InputShapes,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_Axes,
         ReduceOpsLayerTest,
         params_Axes,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce_ReductionTypes,
         ReduceOpsLayerTest,
         params_ReductionTypes,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_ReduceLogical_ReductionTypes,
         ReduceOpsLayerTest,
         params_ReductionTypesLogical,
         ReduceOpsLayerTest::getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Reduce,
         ReduceOpsLayerWithSpecificInputTest,
         testing::Combine(
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/region_yolo.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/region_yolo.cpp
index 395210d020e265..cbcba098db08a3 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/region_yolo.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/region_yolo.cpp
@@ -80,6 +80,6 @@ const auto testCase_yolov2_caffe = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsRegionYolov3, RegionYoloLayerTest, testCase_yolov3, RegionYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsRegionYoloMxnet, RegionYoloLayerTest, testCase_yolov3_mxnet, RegionYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsRegionYoloCaffe, RegionYoloLayerTest, testCase_yolov2_caffe, RegionYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsRegionYolov3, RegionYoloLayerTest, testCase_yolov3, RegionYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsRegionYoloMxnet, RegionYoloLayerTest, testCase_yolov3_mxnet, RegionYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsRegionYoloCaffe, RegionYoloLayerTest, testCase_yolov2_caffe, RegionYoloLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reorg_yolo.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reorg_yolo.cpp
index e0294010774b84..566e4f36169eae 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reorg_yolo.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reorg_yolo.cpp
@@ -67,9 +67,9 @@ const auto testCase_batch_2 = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_caffe_YoloV2, ReorgYoloLayerTest, testCase_caffe_yolov2, ReorgYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_stride_2_smallest, ReorgYoloLayerTest, testCase_smallest, ReorgYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_stride_2, ReorgYoloLayerTest, testCase_stride_2, ReorgYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_stride_3, ReorgYoloLayerTest, testCase_stride_3, ReorgYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_smaller_h, ReorgYoloLayerTest, testCase_smaller_h, ReorgYoloLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsReorgYolo_batch_2, ReorgYoloLayerTest, testCase_batch_2, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_caffe_YoloV2, ReorgYoloLayerTest, testCase_caffe_yolov2, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_stride_2_smallest, ReorgYoloLayerTest, testCase_smallest, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_stride_2, ReorgYoloLayerTest, testCase_stride_2, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_stride_3, ReorgYoloLayerTest, testCase_stride_3, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_smaller_h, ReorgYoloLayerTest, testCase_smaller_h, ReorgYoloLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsReorgYolo_batch_2, ReorgYoloLayerTest, testCase_batch_2, ReorgYoloLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reshape.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reshape.cpp
index 84d30807341593..831c4a4d7e5e56 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reshape.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reshape.cpp
@@ -16,7 +16,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
 };
 
 //TODO: Issue : - 28981
-INSTANTIATE_TEST_CASE_P(DISABLE_smoke_ReshapeCheckDynBatch, ReshapeLayerTestRevise,
+INSTANTIATE_TEST_SUITE_P(DISABLE_smoke_ReshapeCheckDynBatch, ReshapeLayerTestRevise,
         ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
@@ -30,7 +30,7 @@ INSTANTIATE_TEST_CASE_P(DISABLE_smoke_ReshapeCheckDynBatch, ReshapeLayerTestRevi
                 ::testing::Values(std::map<std::string, std::string>({{CONFIG_KEY(DYN_BATCH_ENABLED), CONFIG_VALUE(YES)}}))),
                 ReshapeLayerTestRevise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTestRevise,
+INSTANTIATE_TEST_SUITE_P(smoke_ReshapeCheck, ReshapeLayerTestRevise,
         ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
@@ -44,7 +44,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheck, ReshapeLayerTestRevise,
                 ::testing::Values(std::map<std::string, std::string>({}))),
                 ReshapeLayerTestRevise::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ReshapeCheckNegative, ReshapeLayerTestRevise,
+INSTANTIATE_TEST_SUITE_P(smoke_ReshapeCheckNegative, ReshapeLayerTestRevise,
         ::testing::Combine(
                 ::testing::Values(true),
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reverse_sequence.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reverse_sequence.cpp
index aa4054030a696b..fb3d9786c83e92 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reverse_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reverse_sequence.cpp
@@ -33,7 +33,7 @@ const std::vector<ngraph::helpers::InputLayerType> secondaryInputTypes = {
         ngraph::helpers::InputLayerType::PARAMETER
 };
 
-INSTANTIATE_TEST_CASE_P(Basic_smoke, ReverseSequenceLayerTest,
+INSTANTIATE_TEST_SUITE_P(Basic_smoke, ReverseSequenceLayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(batchAxisIndices),
                             ::testing::ValuesIn(seqAxisIndices),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/rnn_cell.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/rnn_cell.cpp
index b353a97e318be8..9d0f3a58f3c0bb 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/rnn_cell.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/rnn_cell.cpp
@@ -19,7 +19,7 @@ namespace {
     std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32,
                                                              InferenceEngine::Precision::FP16};
 
-    INSTANTIATE_TEST_CASE_P(RNNCellCommon, RNNCellTest,
+    INSTANTIATE_TEST_SUITE_P(RNNCellCommon, RNNCellTest,
             ::testing::Combine(
             ::testing::ValuesIn(should_decompose),
             ::testing::ValuesIn(batch),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/rnn_sequence.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/rnn_sequence.cpp
index 718c1b6e771540..ba5f3fa50ac614 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/rnn_sequence.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/rnn_sequence.cpp
@@ -31,7 +31,7 @@ std::vector<ngraph::op::RecurrentSequenceDirection> direction = {ngraph::op::Rec
 };
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP32};
 
-INSTANTIATE_TEST_CASE_P(RNNSequenceCommonZeroClip, RNNSequenceTest,
+INSTANTIATE_TEST_SUITE_P(RNNSequenceCommonZeroClip, RNNSequenceTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(mode),
                                 ::testing::ValuesIn(seq_lengths_zero_clip),
@@ -45,7 +45,7 @@ INSTANTIATE_TEST_CASE_P(RNNSequenceCommonZeroClip, RNNSequenceTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         RNNSequenceTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(RNNSequenceCommonClip, RNNSequenceTest,
+INSTANTIATE_TEST_SUITE_P(RNNSequenceCommonClip, RNNSequenceTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(mode),
                                 ::testing::ValuesIn(seq_lengths_clip_non_zero),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_elements_update.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_elements_update.cpp
index 325ace5c88ad7e..ffa19fcc4fca70 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_elements_update.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_elements_update.cpp
@@ -43,6 +43,6 @@ const auto ScatterEltUpdateCases = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ScatterEltsUpdate, ScatterElementsUpdateLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ScatterEltsUpdate, ScatterElementsUpdateLayerTest,
     ScatterEltUpdateCases, ScatterElementsUpdateLayerTest::getTestCaseName);
 }  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_nd_update.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_nd_update.cpp
index 39767aaa50cdea..70800dd664c4db 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_nd_update.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_nd_update.cpp
@@ -49,5 +49,5 @@ const auto ScatterNDUpdateCases = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ScatterNDUpdate, ScatterNDUpdateLayerTest, ScatterNDUpdateCases, ScatterNDUpdateLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_ScatterNDUpdate, ScatterNDUpdateLayerTest, ScatterNDUpdateCases, ScatterNDUpdateLayerTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_update.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_update.cpp
index 6f48dd1be4268e..ed4250538d90ae 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_update.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/scatter_update.cpp
@@ -42,6 +42,6 @@ const auto ScatterUpdateCase = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_ScatterUpdate, ScatterUpdateLayerTest, ScatterUpdateCase, ScatterUpdateLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_ScatterUpdate, ScatterUpdateLayerTest, ScatterUpdateCase, ScatterUpdateLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/select.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/select.cpp
index f9a3e9b8c3ac03..dd0f75e78bab38 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/select.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/select.cpp
@@ -63,6 +63,6 @@ const auto numpyCases = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_CLDNN_TestsSelect_none, SelectLayerTest, noneCases, SelectLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CLDNN_TestsSelect_none, SelectLayerTest, noneCases, SelectLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_CLDNN_TestsSelect_numpy, SelectLayerTest, numpyCases, SelectLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_CLDNN_TestsSelect_numpy, SelectLayerTest, numpyCases, SelectLayerTest::getTestCaseName);
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
index 0eb593e16f45e4..74821cf0cc4d25 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/shuffle_channels.cpp
@@ -34,10 +34,10 @@ const auto testCases = ::testing::Combine(::testing::ValuesIn(shuffleParameters)
                                           ::testing::Values(CommonTestUtils::DEVICE_GPU));
 
 
-INSTANTIATE_TEST_CASE_P(smoke_GPU_ShuffleChannels, ShuffleChannelsLayerTest, testCases, ShuffleChannelsLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_GPU_ShuffleChannels, ShuffleChannelsLayerTest, testCases, ShuffleChannelsLayerTest::getTestCaseName);
 
 // ND support tests
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels3D, ShuffleChannelsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels3D, ShuffleChannelsLayerTest,
         ::testing::Combine(
                 ::testing::Values(std::tuple<int, int>(1, 3)),
                 ::testing::ValuesIn(netPrecisions),
@@ -49,7 +49,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels3D, ShuffleChannelsLayerTest,
                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
         ShuffleChannelsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels2D, ShuffleChannelsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels2D, ShuffleChannelsLayerTest,
         ::testing::Combine(
                 ::testing::Values(std::tuple<int, int>(1, 3)),
                 ::testing::ValuesIn(netPrecisions),
@@ -61,7 +61,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels2D, ShuffleChannelsLayerTest,
                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
         ShuffleChannelsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ShuffleChannels1D, ShuffleChannelsLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ShuffleChannels1D, ShuffleChannelsLayerTest,
         ::testing::Combine(
                 ::testing::Values(std::tuple<int, int>(0, 3)),
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/softmax.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/softmax.cpp
index 98727447a883b3..ad797508c6f3ec 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/softmax.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/softmax.cpp
@@ -41,7 +41,7 @@ const auto params2D = testing::Combine(
     testing::Values(std::map<std::string, std::string>())
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_SoftMax2D,
         SoftMaxLayerTest,
         params2D,
@@ -68,7 +68,7 @@ const auto params4D = testing::Combine(
     testing::Values(std::map<std::string, std::string>())
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_SoftMax4D,
         SoftMaxLayerTest,
         params4D,
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp
index 149908a1d7dd06..2da3b2bb72985e 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/space_to_batch.cpp
@@ -119,6 +119,6 @@ spaceToBatchParamsTuple stb_only_test_cases[] = {
                                 CommonTestUtils::DEVICE_GPU),
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_CLDNN, SpaceToBatchLayerTest, ::testing::ValuesIn(stb_only_test_cases),
+INSTANTIATE_TEST_SUITE_P(smoke_CLDNN, SpaceToBatchLayerTest, ::testing::ValuesIn(stb_only_test_cases),
                         SpaceToBatchLayerTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/space_to_depth.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/space_to_depth.cpp
index 4ad127c4d4c75f..e40127f312e58b 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/space_to_depth.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/space_to_depth.cpp
@@ -34,7 +34,7 @@ const auto SpaceToDepthBS2 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(SpaceToDepthBS2, SpaceToDepthLayerTest, SpaceToDepthBS2, SpaceToDepthLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(SpaceToDepthBS2, SpaceToDepthLayerTest, SpaceToDepthBS2, SpaceToDepthLayerTest::getTestCaseName);
 
 const std::vector<std::vector<size_t >> inputShapesBS3 = {
         {1, 1, 3, 3}, {1, 1, 6, 6}, {1, 1, 9, 9}, {2, 4, 9, 9}, {2, 3, 15, 12},
@@ -48,6 +48,6 @@ const auto SpaceToDepthBS3 = ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(SpaceToDepthBS3, SpaceToDepthLayerTest, SpaceToDepthBS3, SpaceToDepthLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(SpaceToDepthBS3, SpaceToDepthLayerTest, SpaceToDepthBS3, SpaceToDepthLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/split.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/split.cpp
index c7fb37ec7488e8..a456325ad7e7d6 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/split.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/split.cpp
@@ -17,7 +17,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::I64
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NumSplitsCheck, SplitLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NumSplitsCheck, SplitLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(1),
                                 ::testing::Values(0, 1, 2, 3),
@@ -31,7 +31,7 @@ INSTANTIATE_TEST_CASE_P(smoke_NumSplitsCheck, SplitLayerTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         SplitLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_splitWithUnusedOutputsTest, SplitLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_splitWithUnusedOutputsTest, SplitLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(5),
                                 ::testing::Values(0),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/squeeze_unsqueeze.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/squeeze_unsqueeze.cpp
index ca5b6ece34a9d0..895439d7d548f8 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/squeeze_unsqueeze.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/squeeze_unsqueeze.cpp
@@ -38,7 +38,7 @@ const std::vector<ngraph::helpers::SqueezeOpType> opTypes = {
         ngraph::helpers::SqueezeOpType::UNSQUEEZE
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic, SqueezeUnsqueezeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Basic, SqueezeUnsqueezeLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(CommonTestUtils::combineParams(axesVectors)),
                                 ::testing::ValuesIn(opTypes),
@@ -50,7 +50,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Basic, SqueezeUnsqueezeLayerTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
                         SqueezeUnsqueezeLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Basic_emptyAxes, SqueezeUnsqueezeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Basic_emptyAxes, SqueezeUnsqueezeLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(CommonTestUtils::combineParams(emptyAxesVectors)),
                                 ::testing::Values(ngraph::helpers::SqueezeOpType::SQUEEZE),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/strided_slice.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/strided_slice.cpp
index 7fb8dd1d8c6e13..1b4447a75dac13 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/strided_slice.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/strided_slice.cpp
@@ -69,7 +69,7 @@ std::vector<StridedSliceSpecificParams> ss_only_test_cases_i64 = {
                                        {0, 0, 0, 0}, {0, 0, 0, 0},  {},  {},  {} },
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_CLDNN_FP32, StridedSliceLayerTest,
         ::testing::Combine(
             ::testing::ValuesIn(ss_only_test_cases_fp32),
@@ -82,7 +82,7 @@ INSTANTIATE_TEST_CASE_P(
             ::testing::Values(std::map<std::string, std::string>())),
         StridedSliceLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_CLDNN_I64, StridedSliceLayerTest,
         ::testing::Combine(
             ::testing::ValuesIn(ss_only_test_cases_i64),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/tile.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/tile.cpp
index 9ddeabf4789b94..8758fc208e8028 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/tile.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/tile.cpp
@@ -21,7 +21,7 @@ const std::vector<std::vector<int64_t>> repeats = {
         {2, 2, 2},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Tile, TileLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Tile, TileLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(repeats),
                 ::testing::ValuesIn(netPrecisions),
@@ -33,7 +33,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Tile, TileLayerTest,
                 ::testing::Values(CommonTestUtils::DEVICE_GPU)),
         TileLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Tile6d, TileLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Tile6d, TileLayerTest,
         ::testing::Combine(
                 ::testing::Values(std::vector<int64_t>({1, 1, 1, 2, 1, 2})),
                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/topk.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/topk.cpp
index 2ac2c1a2373d5e..b34287c5bab924 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/topk.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/topk.cpp
@@ -38,7 +38,7 @@ const std::vector<ngraph::opset4::TopK::SortType> sortTypes = {
 };
 
 
-INSTANTIATE_TEST_CASE_P(smoke_TopK, TopKLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_TopK, TopKLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(k),
                 ::testing::ValuesIn(axes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/transpose.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/transpose.cpp
index 265b742e6f2afa..2fc92324ef5f09 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/transpose.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/transpose.cpp
@@ -46,7 +46,7 @@ const auto params = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Transpose,
         TransposeLayerTest,
         params,
@@ -82,7 +82,7 @@ const auto params5D = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Transpose_5D,
         TransposeLayerTest,
         params5D,
@@ -118,7 +118,7 @@ const auto params6D = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_GPU)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Transpose_6D,
         TransposeLayerTest,
         params6D,
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/variadic_split.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/variadic_split.cpp
index b5e91b031e7269..d6fc5e6f8309f1 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/variadic_split.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/variadic_split.cpp
@@ -25,7 +25,7 @@ namespace {
             {4, 11, 6, 9}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_NumSplitsCheck, VariadicSplitLayerTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_NumSplitsCheck, VariadicSplitLayerTest,
             ::testing::Combine(
             ::testing::ValuesIn(numSplits),
             ::testing::Values(0, 1, 2, 3),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/constant_result.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/constant_result.cpp
index bf4ca2f13523da..35c31e30577465 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/constant_result.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/constant_result.cpp
@@ -34,7 +34,7 @@ const std::vector<Precision> precisions = {
     Precision::BOOL
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Check, ConstantResultSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Check, ConstantResultSubgraphTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(types),
                             ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp
index 7c4be794741700..98027ac7bb64d6 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/get_output_before_activation.cpp
@@ -23,7 +23,7 @@ namespace {
     std::map<std::string, std::string> additional_config = {};
 } // namespace
 
-INSTANTIATE_TEST_CASE_P(OutputBeforeActivation, OutputBeforeActivation,
+INSTANTIATE_TEST_SUITE_P(OutputBeforeActivation, OutputBeforeActivation,
     ::testing::Combine(
         ::testing::Values(CommonTestUtils::DEVICE_GPU),
         ::testing::Values(InferenceEngine::Precision::FP32),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp
index 9afd80f0af9216..eb6d89436b707f 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/matmul_squeeze_add.cpp
@@ -36,7 +36,7 @@ std::vector<size_t> output_sizes = {
     8
 };
 
-INSTANTIATE_TEST_CASE_P(MatmulSqueezeAdd, MatmulSqueezeAddTest,
+INSTANTIATE_TEST_SUITE_P(MatmulSqueezeAdd, MatmulSqueezeAddTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/multiply_add.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/multiply_add.cpp
index 642cb25d4cf150..7dae838e78d53a 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/multiply_add.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/multiply_add.cpp
@@ -23,7 +23,7 @@ const std::vector<std::vector<size_t>> inputShapes = {
         {1, 3, 2, 2, 4, 5},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MultipleAdd_Nd, MultiplyAddLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MultipleAdd_Nd, MultiplyAddLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inputShapes),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/parameter_result.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/parameter_result.cpp
index 2819577a7710b5..9b3de7ad2b7439 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/parameter_result.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/parameter_result.cpp
@@ -10,7 +10,7 @@
 using namespace SubgraphTestsDefinitions;
 
 namespace {
-    INSTANTIATE_TEST_CASE_P(smoke_Check, ParameterResultSubgraphTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Check, ParameterResultSubgraphTest,
                             ::testing::Values(CommonTestUtils::DEVICE_GPU),
                             ParameterResultSubgraphTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp
index 88da94c4ddb6c1..6f757c2bfdaf50 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/perm_conv_perm_concat.cpp
@@ -31,7 +31,7 @@ std::map<std::string, std::string> additional_config = {
 } // namespace
 
 namespace SubgraphTestsDefinitions {
-    INSTANTIATE_TEST_CASE_P(smoke_basic, PermConvPermConcat,
+    INSTANTIATE_TEST_SUITE_P(smoke_basic, PermConvPermConcat,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/quantized_convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/quantized_convolution_backprop_data.cpp
index 48f6635bdae026..5f13a31053d0a4 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/quantized_convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/quantized_convolution_backprop_data.cpp
@@ -41,7 +41,7 @@ const auto quantConvBackpropData2DParams = ::testing::Combine(
         ::testing::ValuesIn(granularity)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantConvBackpropData2D, QuantConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantConvBackpropData2D, QuantConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 quantConvBackpropData2DParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -69,7 +69,7 @@ const auto quantConvBackpropData3DParams = ::testing::Combine(
         ::testing::ValuesIn(granularity)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantConvBackpropData3D, QuantConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantConvBackpropData3D, QuantConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 quantConvBackpropData3DParams,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/quantized_group_convolution_backprop_data.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/quantized_group_convolution_backprop_data.cpp
index b310104ff97e0f..f965da23574218 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/quantized_group_convolution_backprop_data.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/quantized_group_convolution_backprop_data.cpp
@@ -43,7 +43,7 @@ const auto quantGroupConvBackpropData2DParams = ::testing::Combine(
         ::testing::ValuesIn(granularity)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantGroupConvBackpropData2D, QuantGroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantGroupConvBackpropData2D, QuantGroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 quantGroupConvBackpropData2DParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -72,7 +72,7 @@ const auto quantGroupConvBackpropData3DParams = ::testing::Combine(
         ::testing::ValuesIn(granularity)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_QuantGroupConvBackpropData3D, QuantGroupConvBackpropDataLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_QuantGroupConvBackpropData3D, QuantGroupConvBackpropDataLayerTest,
                         ::testing::Combine(
                                 quantGroupConvBackpropData3DParams,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reduce_eltwise.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reduce_eltwise.cpp
index 04dbc6e800551e..62a6097184e27f 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reduce_eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reduce_eltwise.cpp
@@ -14,7 +14,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP32,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ReduceEltwise6D, ReduceEltwiseTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ReduceEltwise6D, ReduceEltwiseTest,
                         testing::Combine(
                                 testing::Values(std::vector<size_t>{2, 3, 4, 5, 6, 7}),
                                 testing::Values(std::vector<int>{2, 3, 4}),
@@ -24,7 +24,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ReduceEltwise6D, ReduceEltwiseTest,
                                 testing::Values(CommonTestUtils::DEVICE_GPU)),
                         ReduceEltwiseTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ReduceEltwise5D, ReduceEltwiseTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ReduceEltwise5D, ReduceEltwiseTest,
                         testing::Combine(
                                 testing::Values(std::vector<size_t>{2, 3, 4, 5, 6}),
                                 testing::Values(std::vector<int>{2, 3}),
@@ -34,7 +34,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ReduceEltwise5D, ReduceEltwiseTest,
                                 testing::Values(CommonTestUtils::DEVICE_GPU)),
                         ReduceEltwiseTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ReduceEltwise4D, ReduceEltwiseTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ReduceEltwise4D, ReduceEltwiseTest,
                         testing::Combine(
                                 testing::Values(std::vector<size_t>{2, 3, 4, 5}),
                                 testing::Values(std::vector<int>{2}),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reshape_permute_conv_permute_reshape_act.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reshape_permute_conv_permute_reshape_act.cpp
index 05fcaf36ecd9b7..88ac469ef1cf78 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reshape_permute_conv_permute_reshape_act.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reshape_permute_conv_permute_reshape_act.cpp
@@ -32,7 +32,7 @@ std::vector<InferenceEngine::Precision> netPrecisions = {
 std::map<std::string, std::string> additional_config = {};
 
 namespace SubgraphTestsDefinitions {
-    INSTANTIATE_TEST_CASE_P(smoke_basic, ConvReshapeAct,
+    INSTANTIATE_TEST_SUITE_P(smoke_basic, ConvReshapeAct,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_GPU),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
index 75d90a5fa2913d..dbf28170fdd1bd 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
@@ -39,7 +39,7 @@ namespace {
             ngraph::helpers::SqueezeOpType::UNSQUEEZE
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_reshape_squeeze_reshape_relu, ReshapeSqueezeReshapeRelu,
+    INSTANTIATE_TEST_SUITE_P(smoke_reshape_squeeze_reshape_relu, ReshapeSqueezeReshapeRelu,
                             ::testing::Combine(
                                     ::testing::ValuesIn(inputs_squeeze),
                                     ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/scale_shift.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/scale_shift.cpp
index 928a9f19df6028..5c1de701dee96f 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/scale_shift.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/scale_shift.cpp
@@ -47,7 +47,7 @@ std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precis
 
 }  // namespace
 
-INSTANTIATE_TEST_CASE_P(smoke_ScaleShift, ScaleShiftLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_ScaleShift, ScaleShiftLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inShapes),
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/split_conv_concat.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
index 40364a93cd2c62..adafa4adbee17b 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
@@ -15,7 +15,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP16
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, SplitConvConcat,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, SplitConvConcat,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(std::vector<size_t >({1, 6, 40, 40})),
diff --git a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/tensor_names.cpp b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/tensor_names.cpp
index a1f5fc59981012..405966554288bd 100644
--- a/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/tensor_names.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/shared_tests_instances/subgraph_tests/tensor_names.cpp
@@ -10,7 +10,7 @@
 using namespace SubgraphTestsDefinitions;
 
 namespace {
-    INSTANTIATE_TEST_CASE_P(smoke_Check, TensorNamesTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Check, TensorNamesTest,
                             ::testing::Values(CommonTestUtils::DEVICE_GPU),
                             TensorNamesTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/gpu/single_layer_tests/tensor_iterator.cpp b/inference-engine/tests/functional/plugin/gpu/single_layer_tests/tensor_iterator.cpp
index bce1ef106917f2..f35942099c9239 100644
--- a/inference-engine/tests/functional/plugin/gpu/single_layer_tests/tensor_iterator.cpp
+++ b/inference-engine/tests/functional/plugin/gpu/single_layer_tests/tensor_iterator.cpp
@@ -268,7 +268,7 @@ TEST_P(TensorIteratorWithConfigTest, CompareWithRefs) {
 using namespace LayerTestsDefinitions;
 
 namespace {
-    INSTANTIATE_TEST_CASE_P(smoke_TensorIteratorCommon, TensorIteratorWithConfigTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_TensorIteratorCommon, TensorIteratorWithConfigTest,
         ::testing::Combine(
             ::testing::ValuesIn(std::vector<size_t> {2, 4}), // seq lengths
             ::testing::ValuesIn(std::vector<size_t> {1}), // only single batch supported
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/operations/dynamic_shape_resolver.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/operations/dynamic_shape_resolver.cpp
index e3b5e4c137c2d9..650d39238e01cc 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/operations/dynamic_shape_resolver.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/operations/dynamic_shape_resolver.cpp
@@ -87,7 +87,7 @@ std::set<ngraph::element::Type_t> allNGraphStaticTypes() {
     return staticTypes;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicShapeResolverTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicShapeResolverTests, testing::Combine(
     testing::ValuesIn(allNGraphStaticTypes()),
     testing::Values(DataShape{1, 800}, DataShape{1, 1})));
 
@@ -119,7 +119,7 @@ TEST_P(DynamicShapeResolverNegativeTestsDataType, ThrowsOnInvalidDimsType) {
     ASSERT_THROW(auto dynamicShapeResolver = std::make_shared<ngraph::vpu::op::DynamicShapeResolver>(data, dims), ngraph::ngraph_error);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicShapeResolverNegativeTestsDataType, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicShapeResolverNegativeTestsDataType, testing::Combine(
     testing::Values(ngraph::element::dynamic),
     testing::Values(ngraph::element::i64),
     testing::Values(DataPartialShape{1, 800}),
@@ -139,7 +139,7 @@ std::set<ngraph::element::Type_t> allNGraphNotIntegralTypes() {
     return notIntegralTypes;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicShapeResolverNegativeTestsDimsType, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicShapeResolverNegativeTestsDimsType, testing::Combine(
     testing::ValuesIn(allNGraphStaticTypes()),
     testing::ValuesIn(allNGraphNotIntegralTypes()),
     testing::Values(DataPartialShape{1, 800}),
@@ -150,7 +150,7 @@ TEST_P(DynamicShapeResolverNegativeTestsDataShape, ThrowsOnInvalidDimsType) {
     ASSERT_THROW(auto dynamicShapeResolver = std::make_shared<ngraph::vpu::op::DynamicShapeResolver>(data, dims), ngraph::ngraph_error);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicShapeResolverNegativeTestsDataShape, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicShapeResolverNegativeTestsDataShape, testing::Combine(
     testing::ValuesIn(allNGraphStaticTypes()),
     testing::Values(ngraph::element::i64),
     testing::Values(
@@ -165,7 +165,7 @@ TEST_P(DynamicShapeResolverNegativeTestsDimsShape, ThrowsOnInvalidDimsType) {
     ASSERT_THROW(auto dynamicShapeResolver = std::make_shared<ngraph::vpu::op::DynamicShapeResolver>(data, dims), ngraph::ngraph_error);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicShapeResolverNegativeTestsDimsShape, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicShapeResolverNegativeTestsDimsShape, testing::Combine(
     testing::ValuesIn(allNGraphTypes()),
     testing::Values(ngraph::element::i64),
     testing::Values(DataShape{1, 800}),
@@ -275,7 +275,7 @@ const auto basicCases = ::testing::Combine(
 );
 
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicShapeResolverPluginTests, DynamicShapeResolverPluginTests,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicShapeResolverPluginTests, DynamicShapeResolverPluginTests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inputDatas),
                                 ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)),
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/operations/out_shape_of_reshape.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/operations/out_shape_of_reshape.cpp
index fd5687af017bc1..9128463dd6dd14 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/operations/out_shape_of_reshape.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/operations/out_shape_of_reshape.cpp
@@ -98,7 +98,7 @@ TEST_P(OutShapeOfReshapeTests, CanValidateAndInferTypes) {
             ngraph::ParameterVector{m_inDataShapeParam, m_outShapeDescriptorParam}));
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, OutShapeOfReshapeTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, OutShapeOfReshapeTests, testing::Combine(
         testing::ValuesIn(tensorShapes),
         testing::ValuesIn(allNGraphIntegralNumberTypes()),
         testing::ValuesIn(tensorShapes),
@@ -125,7 +125,7 @@ TEST_P(OutShapeOfReshapeTestsNegativeDataType, ThrowsOnInvalidDataType) {
             m_inDataShapeParam, m_outShapeDescriptorParam, true),
                  ngraph::NodeValidationFailure);
 }
-INSTANTIATE_TEST_CASE_P(smoke_InvalidInDataShapeTensorType, OutShapeOfReshapeTestsNegativeDataType,
+INSTANTIATE_TEST_SUITE_P(smoke_InvalidInDataShapeTensorType, OutShapeOfReshapeTestsNegativeDataType,
         testing::Combine(
             testing::Values(TensorShape{4}),
             testing::ValuesIn(allNGraphNotIntegralTypes()),
@@ -133,7 +133,7 @@ INSTANTIATE_TEST_CASE_P(smoke_InvalidInDataShapeTensorType, OutShapeOfReshapeTes
             testing::Values(ngraph::element::i64))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_InvalidOutShapeDescriptorTensorType, OutShapeOfReshapeTestsNegativeDataType,
+INSTANTIATE_TEST_SUITE_P(smoke_InvalidOutShapeDescriptorTensorType, OutShapeOfReshapeTestsNegativeDataType,
         testing::Combine(
             testing::Values(TensorShape{4}),
             testing::Values(ngraph::element::i64),
@@ -155,7 +155,7 @@ TEST_P(OutShapeOfReshapeTestsNegativeDataShape, ThrowsOnInvalidDataShape) {
                  ngraph::NodeValidationFailure);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_InvalidInDataShapeTensorShape, OutShapeOfReshapeTestsNegativeDataShape,
+INSTANTIATE_TEST_SUITE_P(smoke_InvalidInDataShapeTensorShape, OutShapeOfReshapeTestsNegativeDataShape,
         testing::Combine(
             testing::ValuesIn(invalidTensorShapes),
             testing::Values(ngraph::element::i64),
@@ -163,7 +163,7 @@ INSTANTIATE_TEST_CASE_P(smoke_InvalidInDataShapeTensorShape, OutShapeOfReshapeTe
             testing::Values(ngraph::element::i64))
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_InvalidOutShapeDescriptorTensorShape, OutShapeOfReshapeTestsNegativeDataShape,
+INSTANTIATE_TEST_SUITE_P(smoke_InvalidOutShapeDescriptorTensorShape, OutShapeOfReshapeTestsNegativeDataShape,
         testing::Combine(
             testing::ValuesIn(tensorShapes),
             testing::Values(ngraph::element::i64),
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/operations/static_shape_broadcast.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/operations/static_shape_broadcast.cpp
index 79371a30a42c6a..79bc42c66e62b0 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/operations/static_shape_broadcast.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/operations/static_shape_broadcast.cpp
@@ -151,7 +151,7 @@ TEST_P(StaticShapeBroadcastNumpyTests, CanValidateAndInferTypes) {
     ASSERT_EQ(m_tensorWithTargetShape->get_shape(), op->output(0).get_shape());
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeBroadcastNumpyTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeBroadcastNumpyTests, testing::Combine(
         testing::ValuesIn(testNGraphNumericTypes),
         testing::ValuesIn(testNumpyStaticShapes))
 );
@@ -167,7 +167,7 @@ TEST_P(StaticShapeBroadcastExplicitTests, CanValidateAndInferTypes) {
     ASSERT_EQ(m_tensorWithTargetShape->get_shape(), op->get_output_shape(0));
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeBroadcastExplicitTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeBroadcastExplicitTests, testing::Combine(
         testing::ValuesIn(testNGraphNumericTypes),
         testing::ValuesIn(testExplicitStaticShapes))
 );
@@ -183,7 +183,7 @@ TEST_P(StaticShapeBroadcastBidirectionalTests, CanValidateAndInferTypes) {
     ASSERT_EQ(m_tensorWithOutput->get_shape(), op->output(0).get_shape());
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeBroadcastBidirectionalTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeBroadcastBidirectionalTests, testing::Combine(
         testing::ValuesIn(testNGraphNumericTypes),
         testing::ValuesIn(testBidirectionalStaticShapes))
 );
@@ -204,7 +204,7 @@ TEST_P(StaticShapeBroadcastNumpyTestsNegativeNumInputs, ThrowsOnInvalidNumInputs
                  ngraph::NodeValidationFailure);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeBroadcastNumpyTestsNegativeNumInputs, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeBroadcastNumpyTestsNegativeNumInputs, testing::Combine(
         testing::Values(ngraph::element::f16),
         testing::Values(testNumpyStaticShapes[0]))
 );
@@ -218,7 +218,7 @@ TEST_P(StaticShapeBroadcastExplicitTestsNegativeNumInputs, ThrowsOnInvalidNumInp
                  ngraph::NodeValidationFailure);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeBroadcastExplicitTestsNegativeNumInputs, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeBroadcastExplicitTestsNegativeNumInputs, testing::Combine(
         testing::Values(ngraph::element::f16),
         testing::Values(testExplicitStaticShapes[0]))
 );
@@ -235,7 +235,7 @@ TEST_P(StaticShapeBroadcastBidirectionalTestsNegativeNumInputs, ThrowsOnInvalidN
 }
 
 using StaticShapeBroadcastTestsNegativeMode = StaticShapeBroadcastNumpyTests;
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeBroadcastTestsNegativeMode, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeBroadcastTestsNegativeMode, testing::Combine(
         testing::Values(ngraph::element::f16),
         testing::Values(testNumpyStaticShapes[0]))
 );
@@ -249,7 +249,7 @@ TEST_P(StaticShapeBroadcastTestsNegativeEvaluate, ThrowsOnInvalidMode) {
             m_tensor, targetShape), ngraph::NodeValidationFailure);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeBroadcastTestsNegativeEvaluate, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeBroadcastTestsNegativeEvaluate, testing::Combine(
         testing::Values(ngraph::element::f16),
         testing::Values(testNumpyStaticShapes[0]))
 );
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/operations/static_shape_nonzero.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/operations/static_shape_nonzero.cpp
index 245bee742a33bb..80a9d9b64f6d9c 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/operations/static_shape_nonzero.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/operations/static_shape_nonzero.cpp
@@ -88,7 +88,7 @@ TEST_P(StaticShapeNonZeroTests, CanValidateAndInferTypes) {
             ngraph::ParameterVector{m_param}));
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeNonZeroTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeNonZeroTests, testing::Combine(
         testing::ValuesIn(testNGraphNumericTypes),
         testing::ValuesIn(testStaticShapes),
         testing::ValuesIn(outputTypes))
@@ -105,7 +105,7 @@ TEST_P(StaticShapeNonZeroTestsNegativeInputDataType, ThrowsOnInvalidInputType) {
                  ngraph::NodeValidationFailure);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeNonZeroTestsNegativeInputDataType, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeNonZeroTestsNegativeInputDataType, testing::Combine(
         testing::Values(ngraph::element::dynamic),
         testing::ValuesIn(testStaticShapes),
         testing::ValuesIn(outputTypes))
@@ -118,7 +118,7 @@ TEST_P(StaticShapeNonZeroTestsNegativeDataShape, ThrowsOnInvalidDataShape) {
                  ngraph::NodeValidationFailure);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeNonZeroTestsNegativeDataShape, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeNonZeroTestsNegativeDataShape, testing::Combine(
         testing::ValuesIn(testNGraphNumericTypes),
         testing::ValuesIn(testDynamicShapes),
         testing::ValuesIn(outputTypes))
@@ -131,7 +131,7 @@ TEST_P(StaticShapeNonZeroTestsNegativeOutputDataType, ThrowsOnInvalidOutputType)
                  ngraph::NodeValidationFailure);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, StaticShapeNonZeroTestsNegativeOutputDataType, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, StaticShapeNonZeroTestsNegativeOutputDataType, testing::Combine(
         testing::ValuesIn(testNGraphNumericTypes),
         testing::ValuesIn(testStaticShapes),
         testing::Values(ngraph::element::boolean))
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/convert_extract_image_patches_to_reorg_yolo_vpu.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/convert_extract_image_patches_to_reorg_yolo_vpu.cpp
index e61a1765880c6d..f3bb5bc2d63fac 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/convert_extract_image_patches_to_reorg_yolo_vpu.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/convert_extract_image_patches_to_reorg_yolo_vpu.cpp
@@ -85,7 +85,7 @@ TEST_P(ConvertEIPToReorgYoloPositiveTest, CompareFunctions) {
     ASSERT_TRUE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, ConvertEIPToReorgYoloPositiveTest, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, ConvertEIPToReorgYoloPositiveTest, testing::Combine(
         testing::Values(ngraph::Shape{1, 64, 500, 500}),
         testing::Values(ngraph::Shape{5, 5}),
         testing::Values(ngraph::Strides{5, 5}),
@@ -106,7 +106,7 @@ TEST_P(DoNotConvertEIPToReorgYoloOnDiffSizeAndStride, CompareFunctions) {
     ASSERT_FALSE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DoNotConvertEIPToReorgYoloOnDiffSizeAndStride, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DoNotConvertEIPToReorgYoloOnDiffSizeAndStride, testing::Combine(
         testing::Values(ngraph::PartialShape{1, 64, 500, 500}),
         testing::Values(ngraph::Shape{5, 5}),
         testing::Values(ngraph::Strides{4, 4}),
@@ -126,7 +126,7 @@ TEST_P(DoNotConvertEIPToReorgYoloOnNot4DInput, CompareFunctions) {
     EXPECT_ANY_THROW(transform(dataShape, sizes, strides, rates, padMode));
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DoNotConvertEIPToReorgYoloOnNot4DInput, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DoNotConvertEIPToReorgYoloOnNot4DInput, testing::Combine(
         testing::Values(ngraph::PartialShape{1, 1, 64, 500, 500},
                         ngraph::PartialShape{64, 500, 500},
                         ngraph::PartialShape{500, 500},
@@ -154,7 +154,7 @@ TEST_P(DoNotConvertEIPToReorgYoloOnNotStaticInput, CompareFunctions) {
     ASSERT_TRUE(reorgIt == ops.end());
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DoNotConvertEIPToReorgYoloOnNotStaticInput, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DoNotConvertEIPToReorgYoloOnNotStaticInput, testing::Combine(
         testing::Values(ngraph::PartialShape{1, 64, ngraph::Dimension::dynamic(), 500},
                         ngraph::PartialShape{1, 64, 500, ngraph::Dimension::dynamic()}),
         testing::Values(ngraph::Shape{5, 5}),
@@ -169,7 +169,7 @@ TEST_P(DoNotConvertEIPToReorgYoloOnNonSingleRates, CompareFunctions) {
     ASSERT_FALSE(res.first) << res.second;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DoNotConvertEIPToReorgYoloOnNonSingleRates, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DoNotConvertEIPToReorgYoloOnNonSingleRates, testing::Combine(
         testing::Values(ngraph::Shape{1, 64, 500, 500}),
         testing::Values(ngraph::Shape{5, 5}),
         testing::Values(ngraph::Strides{5, 5}),
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_binary_elementwise.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_binary_elementwise.cpp
index 24dcef17e604cb..d1d87c77d071fa 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_binary_elementwise.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_binary_elementwise.cpp
@@ -266,7 +266,7 @@ class DynamicToStaticShapeEltwise: public CommonTestUtils::TestsCommon, public t
 TEST_P(DynamicToStaticShapeEltwise, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseBroadcast, DynamicToStaticShapeEltwise, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseBroadcast, DynamicToStaticShapeEltwise, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_broadcast.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_broadcast.cpp
index dae6706a399b83..78de9260c2db61 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_broadcast.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_broadcast.cpp
@@ -151,7 +151,7 @@ class DynamicToStaticShapeBroadcastExplicitTests
 TEST_P(DynamicToStaticShapeBroadcastExplicitTests, compareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeBroadcastExplicitTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeBroadcastExplicitTests, testing::Combine(
         testing::Values(
             ngraph::element::f16,
             ngraph::element::f32,
@@ -297,7 +297,7 @@ class DynamicToStaticShapeBroadcastBidirectionalTests : public CommonTestUtils::
 TEST_P(DynamicToStaticShapeBroadcastBidirectionalTests, compareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeBroadcastBidirectionalTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeBroadcastBidirectionalTests, testing::Combine(
         testing::Values(
             ngraph::element::f16,
             ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_clamp.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_clamp.cpp
index 15ecbfc951e37d..97d3d79238f375 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_clamp.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_clamp.cpp
@@ -72,7 +72,7 @@ class DynamicToStaticShapeClamp : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeClamp, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeClamp, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeClamp, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_concat.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_concat.cpp
index d18710a0bd9fae..90b62ee451c3fd 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_concat.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_concat.cpp
@@ -134,7 +134,7 @@ std::vector<ConcatParam> concatParams = {
         {DataShapes{DataShape{3, 64, 128}, DataShape{3, 64, 256}, DataShape{3, 64, 512}}, -1},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeConcatTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeConcatTests, testing::Combine(
         testing::ValuesIn(dataTypes),
         testing::ValuesIn(concatParams)));
 
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_convert.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_convert.cpp
index 851702d95752ff..e738d1d49c29c9 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_convert.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_convert.cpp
@@ -72,7 +72,7 @@ class DynamicToStaticShapeConvert : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeConvert, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeConvert, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeConvert, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather.cpp
index 13b1c44893af91..94463c74b717d2 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather.cpp
@@ -125,7 +125,7 @@ class DynamicToStaticShapeGatherDataDSR : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeGatherDataDSR, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeGatherDataDSR, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeGatherDataDSR, combinations);
 
 class DynamicToStaticShapeGatherIdxDSR : public CommonTestUtils::TestsCommon,
         public testing::WithParamInterface<std::tuple<DataType, DataType, GatherTestCase>> {
@@ -206,7 +206,7 @@ class DynamicToStaticShapeGatherIdxDSR : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeGatherIdxDSR, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeGatherIdxDSR, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeGatherIdxDSR, combinations);
 
 class DynamicToStaticShapeGather : public CommonTestUtils::TestsCommon,
         public testing::WithParamInterface<std::tuple<DataType, DataType, GatherTestCase>> {
@@ -289,6 +289,6 @@ class DynamicToStaticShapeGather : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeGather, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeGather, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeGather, combinations);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather_elements.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather_elements.cpp
index c60be44e9e5d10..baf216e0101943 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather_elements.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather_elements.cpp
@@ -128,6 +128,6 @@ class DynamicToStaticShapeGatherElements : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeGatherElements, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeGatherElements, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeGatherElements, combinations);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather_nd.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather_nd.cpp
index f4f91dd2bf6235..16fbf15ea3c2bb 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather_nd.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_gather_nd.cpp
@@ -186,6 +186,6 @@ class DynamicToStaticShapeGatherND : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeGatherND, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeGatherND, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeGatherND, combinations);
 
 } // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_matmul.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_matmul.cpp
index fbab95ffbd4f9c..998b640d62d138 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_matmul.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_matmul.cpp
@@ -204,6 +204,6 @@ class DynamicToStaticShapeMatMul: public CommonTestUtils::TestsCommon,
 
 TEST_P(DynamicToStaticShapeMatMul, CompareFunctions) {
 }
-INSTANTIATE_TEST_CASE_P(smoke_MatMul, DynamicToStaticShapeMatMul, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_MatMul, DynamicToStaticShapeMatMul, combinations);
 
 }  // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_non_max_suppression.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_non_max_suppression.cpp
index c69468741d1dd9..b5e8cb06442411 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_non_max_suppression.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_non_max_suppression.cpp
@@ -111,7 +111,7 @@ class DynamicToStaticShapeNonMaxSuppression : public CommonTestUtils::TestsCommo
 TEST_P(DynamicToStaticShapeNonMaxSuppression, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeNonMaxSuppression, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeNonMaxSuppression, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32),
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_nonzero.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_nonzero.cpp
index b62ce98c88b150..9da9b5072d373d 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_nonzero.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_nonzero.cpp
@@ -91,7 +91,7 @@ TEST_P(DynamicToStaticShapeNonZeroTests, CompareFunctions) {
     compareFunctions();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeNonZeroTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeNonZeroTests, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_reduce.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_reduce.cpp
index 13fbaee293a7ec..09bd745769c5f7 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_reduce.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_reduce.cpp
@@ -158,7 +158,7 @@ class DynamicToStaticShapeReduce: public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeReduce, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_Arithmetic, DynamicToStaticShapeReduce, arithmetic_combinations);
-INSTANTIATE_TEST_CASE_P(smoke_Logical, DynamicToStaticShapeReduce, logical_combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_Arithmetic, DynamicToStaticShapeReduce, arithmetic_combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_Logical, DynamicToStaticShapeReduce, logical_combinations);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_reshape.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_reshape.cpp
index a0abc4e5c9ad41..1200660372ab8a 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_reshape.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_reshape.cpp
@@ -114,7 +114,7 @@ std::shared_ptr<ngraph::op::Op> generateDynamicReshapePattern(std::shared_ptr<ng
         0);
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeReshapeTests, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeReshapeTests, testing::Combine(
     testing::Values(
         DataShape{4, 1000},
         DataShape{3, 128, 256},
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_roialign.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_roialign.cpp
index 4551bfb1378b81..1730bc9336260a 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_roialign.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_roialign.cpp
@@ -101,7 +101,7 @@ class DynamicToStaticShapeROIAlignDataDSR : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeROIAlignDataDSR, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeROIAlignDataDSR, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeROIAlignDataDSR, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32),
@@ -188,7 +188,7 @@ class DynamicToStaticShapeROIAlignROIDSR : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeROIAlignROIDSR, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeROIAlignROIDSR, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeROIAlignROIDSR, testing::Combine(
         testing::Values(
                 ngraph::element::f16,
                 ngraph::element::f32),
@@ -281,7 +281,7 @@ class DynamicToStaticShapeROIAlign : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeROIAlign, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeROIAlign, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeROIAlign, testing::Combine(
         testing::Values(
                 ngraph::element::f16,
                 ngraph::element::f32),
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_scatter.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_scatter.cpp
index a846600340696f..068eab69e7114c 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_scatter.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_scatter.cpp
@@ -134,7 +134,7 @@ class DynamicToStaticShapeScatter : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeScatter, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeScatter, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeScatter, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_split.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_split.cpp
index b61464a91405c9..fdb9c033f512be 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_split.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_split.cpp
@@ -115,6 +115,6 @@ std::shared_ptr<const ngraph::Function> reference(
 TEST_P(DynamicToStaticShapeSplit, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeSplit, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeSplit, combinations);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_squeeze.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_squeeze.cpp
index 65200017df2d9b..757a8fd511310a 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_squeeze.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_squeeze.cpp
@@ -98,7 +98,7 @@ public testing::WithParamInterface<std::tuple<DataType, SqueezeTestCase>> {
 TEST_P(DynamicToStaticShapeSqueeze, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeSqueeze, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeSqueeze, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_topk.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_topk.cpp
index e761ed6d94b645..b2674397fa8490 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_topk.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_topk.cpp
@@ -135,7 +135,7 @@ class DynamicToStaticShapeTopKConst : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeTopKConst, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeTopKConst, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeTopKConst, combinations);
 
 
 class DynamicToStaticShapeTopK : public CommonTestUtils::TestsCommon,
@@ -239,6 +239,6 @@ class DynamicToStaticShapeTopK : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeTopK, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeTopK, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeTopK, combinations);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_transpose.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_transpose.cpp
index 71e6d7cf7e5f44..81214680410cde 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_transpose.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_transpose.cpp
@@ -104,7 +104,7 @@ class DynamicToStaticShapeTranspose : public CommonTestUtils::TestsCommon, publi
 TEST_P(DynamicToStaticShapeTranspose, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeTranspose, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeTranspose, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_unary_elementwise.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_unary_elementwise.cpp
index 55bef3bb2ceed8..317b8153344e0f 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_unary_elementwise.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_unary_elementwise.cpp
@@ -76,7 +76,7 @@ class DynamicToStaticShapeUnaryElementwise : public CommonTestUtils::TestsCommon
 TEST_P(DynamicToStaticShapeUnaryElementwise, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeUnaryElementwise, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeUnaryElementwise, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_unsqueeze.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_unsqueeze.cpp
index 8684b1e6bbabe4..9be7cfc2a688f3 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_unsqueeze.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_unsqueeze.cpp
@@ -111,7 +111,7 @@ public testing::WithParamInterface<std::tuple<DataType, UnsqueezeTestCase>> {
 TEST_P(DynamicToStaticShapeUnsqueeze, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeUnsqueeze, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeUnsqueeze, testing::Combine(
     testing::Values(
         ngraph::element::f16,
         ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_variadic_split.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_variadic_split.cpp
index 8d871bdd6c5757..75ed78c0880f7a 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_variadic_split.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/dynamic_to_static_shape_variadic_split.cpp
@@ -135,6 +135,6 @@ class DynamicToStaticShapeVeriadicSplit : public CommonTestUtils::TestsCommon,
 TEST_P(DynamicToStaticShapeVeriadicSplit, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticShapeVeriadicSplit, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticShapeVeriadicSplit, combinations);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/eliminate_shapeof_after_dsr.cpp b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/eliminate_shapeof_after_dsr.cpp
index 15b7f6f434ecc3..15d863b446fcfc 100644
--- a/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/eliminate_shapeof_after_dsr.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/ngraph/transformations/eliminate_shapeof_after_dsr.cpp
@@ -69,7 +69,7 @@ class EliminateShapeOfAfterDSRTest : public CommonTestUtils::TestsCommon,
 TEST_P(EliminateShapeOfAfterDSRTest, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, EliminateShapeOfAfterDSRTest, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, EliminateShapeOfAfterDSRTest, testing::Combine(
         testing::Values(
                 ngraph::element::f16,
                 ngraph::element::f32,
@@ -134,7 +134,7 @@ class EliminateShapeOfAfterDSRWithoutOutputDSR : public CommonTestUtils::TestsCo
 TEST_P(EliminateShapeOfAfterDSRWithoutOutputDSR, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, EliminateShapeOfAfterDSRWithoutOutputDSR, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, EliminateShapeOfAfterDSRWithoutOutputDSR, testing::Combine(
         testing::Values(
                 ngraph::element::f16,
                 ngraph::element::f32,
@@ -203,7 +203,7 @@ class EliminateShapeOfAfterDSRKeepDSR : public CommonTestUtils::TestsCommon,
 TEST_P(EliminateShapeOfAfterDSRKeepDSR, CompareFunctions) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, EliminateShapeOfAfterDSRKeepDSR, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, EliminateShapeOfAfterDSRKeepDSR, testing::Combine(
         testing::Values(
                 ngraph::element::f16,
                 ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/caching_tests.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/caching_tests.cpp
index c7b1194c2940fc..6c17496e7c2f19 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/caching_tests.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/caching_tests.cpp
@@ -31,7 +31,7 @@ namespace {
         return funcs;
     }
 
-    INSTANTIATE_TEST_CASE_P(smoke_CachingSupportCase_Myriad, LoadNetworkCacheTestBase,
+    INSTANTIATE_TEST_SUITE_P(smoke_CachingSupportCase_Myriad, LoadNetworkCacheTestBase,
                             ::testing::Combine(
                                     ::testing::ValuesIn(smoke_functions()),
                                     ::testing::ValuesIn(smoke_precisionsMyriad),
@@ -39,7 +39,7 @@ namespace {
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)),
                             LoadNetworkCacheTestBase::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(nightly_CachingSupportCase_Myriad, LoadNetworkCacheTestBase,
+    INSTANTIATE_TEST_SUITE_P(nightly_CachingSupportCase_Myriad, LoadNetworkCacheTestBase,
                             ::testing::Combine(
                                     ::testing::ValuesIn(LoadNetworkCacheTestBase::getStandardFunctions()),
                                     ::testing::ValuesIn(nightly_precisionsMyriad),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/config.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/config.cpp
index 773419d78f18e9..a045692d932c2d 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/config.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/config.cpp
@@ -96,7 +96,7 @@ std::vector<std::map<std::string, std::string>> getCorrectConfigs() {
     return correctConfigs;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectConfigTests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
@@ -134,7 +134,7 @@ const std::vector<std::map<std::string, std::string>>& getCorrectMultiConfigs()
     return correctMultiConfigs;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigTests,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, CorrectConfigTests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MULTI),
@@ -148,7 +148,7 @@ const std::vector<std::pair<std::string, InferenceEngine::Parameter>>& getDefaul
     return defaultEntries;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectSingleOptionDefaultValueConfigTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectSingleOptionDefaultValueConfigTests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
@@ -176,7 +176,7 @@ const std::vector<std::tuple<std::string, std::string, InferenceEngine::Paramete
     return customEntries;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectSingleOptionCustomValueConfigTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectSingleOptionCustomValueConfigTests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
@@ -190,7 +190,7 @@ const std::vector<std::string>& getPublicOptions() {
     return publicOptions;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigPublicOptionsTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectConfigPublicOptionsTests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
@@ -203,7 +203,7 @@ const std::vector<std::string>& getPrivateOptions() {
     return privateOptions;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigPrivateOptionsTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectConfigPrivateOptionsTests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
@@ -270,7 +270,7 @@ const std::vector<std::map<std::string, std::string>>& getIncorrectConfigs() {
     return incorrectConfigs;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigTests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
@@ -313,41 +313,41 @@ const std::vector<std::map<std::string, std::string>>& getIncorrectMultiConfigs(
     return incorrectMultiConfigs;
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, IncorrectConfigTests,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, IncorrectConfigTests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MULTI),
         ::testing::ValuesIn(getIncorrectMultiConfigs())),
     IncorrectConfigTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigSingleOptionTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigSingleOptionTests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
         ::testing::Values("INCORRECT_KEY")));
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CorrectConfigAPITests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CorrectConfigAPITests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
         ::testing::Values(std::map<std::string, std::string>{})),
     CorrectConfigAPITests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CorrectConfigAPITests,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, CorrectConfigAPITests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MULTI),
         ::testing::ValuesIn(getCorrectMultiConfigs())),
     CorrectConfigAPITests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, IncorrectConfigAPITests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
         ::testing::Values(std::map<std::string, std::string>{{"INCORRECT_KEY", "INCORRECT_VALUE"}})),
     IncorrectConfigAPITests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, IncorrectConfigAPITests,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, IncorrectConfigAPITests,
     ::testing::Combine(
         ::testing::ValuesIn(getPrecisions()),
         ::testing::Values(CommonTestUtils::DEVICE_MULTI),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_integration.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_integration.cpp
index 2eaa15ec86612f..0897d13c2d102e 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_integration.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_integration.cpp
@@ -26,11 +26,11 @@ std::pair<std::string, std::string> plugins[] = {
 // IE Class Common tests with <pluginName, deviceName params>
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassBasicTestP_smoke, IEClassBasicTestP,
         ::testing::ValuesIn(plugins));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassNetworkTestP_smoke, IEClassNetworkTestP,
         ::testing::ValuesIn(devices));
 
@@ -52,17 +52,17 @@ TEST_P(IEClassNetworkTestP_VPU_GetMetric, smoke_OptimizationCapabilitiesReturnsF
     ASSERT_EQ(optimizationCapabilities.front(), METRIC_VALUE(FP16));
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassGetMetricP, IEClassNetworkTestP_VPU_GetMetric,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassImportExportTestP, IEClassImportExportTestP,
         ::testing::Values(std::string(CommonTestUtils::DEVICE_MYRIAD), "HETERO:" + std::string(CommonTestUtils::DEVICE_MYRIAD)));
 
 #if defined(ENABLE_MKL_DNN) && ENABLE_MKL_DNN
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_IEClassImportExportTestP_HETERO_CPU, IEClassImportExportTestP,
         ::testing::Values("HETERO:" + std::string(CommonTestUtils::DEVICE_MYRIAD) + ",CPU"));
 #endif
@@ -71,27 +71,27 @@ INSTANTIATE_TEST_CASE_P(
 // Executable Network GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassExecutableNetworkGetMetricTest_nightly,
         IEClassExecutableNetworkGetMetricTest_ThrowsUnsupported,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassExecutableNetworkGetMetricTest_nightly,
         IEClassExecutableNetworkGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassExecutableNetworkGetMetricTest_nightly,
         IEClassExecutableNetworkGetMetricTest_SUPPORTED_METRICS,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassExecutableNetworkGetMetricTest_nightly,
         IEClassExecutableNetworkGetMetricTest_NETWORK_NAME,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassExecutableNetworkGetMetricTest_nightly,
         IEClassExecutableNetworkGetMetricTest_OPTIMAL_NUMBER_OF_INFER_REQUESTS,
         ::testing::ValuesIn(devices));
@@ -100,7 +100,7 @@ INSTANTIATE_TEST_CASE_P(
 // Executable Network GetConfig
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassExecutableNetworkGetConfigTest_nightly,
         IEClassExecutableNetworkGetConfigTest,
         ::testing::ValuesIn(devices));
@@ -109,32 +109,32 @@ INSTANTIATE_TEST_CASE_P(
 // IE Class GetMetric
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassGetMetricTest_nightly,
         IEClassGetMetricTest_ThrowUnsupported,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassGetMetricTest_nightly,
         IEClassGetMetricTest_AVAILABLE_DEVICES,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassGetMetricTest_nightly,
         IEClassGetMetricTest_SUPPORTED_METRICS,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassGetMetricTest_nightly,
         IEClassGetMetricTest_SUPPORTED_CONFIG_KEYS,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassGetMetricTest_nightly,
         IEClassGetMetricTest_OPTIMIZATION_CAPABILITIES,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassGetMetricTest_nightly,
         IEClassGetMetricTest_RANGE_FOR_ASYNC_INFER_REQUESTS,
         ::testing::ValuesIn(devices));
@@ -143,26 +143,26 @@ INSTANTIATE_TEST_CASE_P(
 // IE Class GetConfig
 //
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassGetConfigTest_nightly,
         IEClassGetConfigTest,
         ::testing::ValuesIn(devices));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassGetConfigTest_nightly,
         IEClassGetConfigTest_ThrowUnsupported,
         ::testing::ValuesIn(devices));
 
 // IE Class Query network
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         DISABLED_IEClassQueryNetworkTest_smoke,
         IEClassQueryNetworkTest,
         ::testing::ValuesIn(devices));
 
 // IE Class Load network
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         IEClassLoadNetworkTest_smoke,
         IEClassLoadNetworkTest,
         ::testing::ValuesIn(devices));
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_threading_tests.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_threading_tests.cpp
index 9ca3ccb37dd15d..9a8a54b6991bb6 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_threading_tests.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/core_threading_tests.cpp
@@ -14,9 +14,9 @@ Params params[] = {
 
 }  // namespace
 
-INSTANTIATE_TEST_CASE_P(MYRIAD, CoreThreadingTests, testing::ValuesIn(params), CoreThreadingTests::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(MYRIAD, CoreThreadingTests, testing::ValuesIn(params), CoreThreadingTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(DISABLED_MYRIAD, CoreThreadingTestsWithIterations,
+INSTANTIATE_TEST_SUITE_P(DISABLED_MYRIAD, CoreThreadingTestsWithIterations,
     testing::Combine(testing::ValuesIn(params),
                      testing::Values(2),
                      testing::Values(2),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/cpp_holders.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/cpp_holders.cpp
index 08de6a3269e129..d64be9c7ac0ba7 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/cpp_holders.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/cpp_holders.cpp
@@ -18,19 +18,19 @@ namespace {
             {2, 1, 0}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTest,
             ::testing::Combine(
             ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
             ::testing::ValuesIn(orders)),
             HoldersTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTestImportNetwork,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTestImportNetwork,
             ::testing::Combine(
             ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
             ::testing::ValuesIn(orders)),
             HoldersTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, HoldersTestOnImportedNetwork,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, HoldersTestOnImportedNetwork,
             ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
             HoldersTestOnImportedNetwork::getTestCaseName);
 
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/detect_network_batch_test.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/detect_network_batch_test.cpp
index d4d6ef76525934..16bf3b8462dd6e 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/detect_network_batch_test.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/detect_network_batch_test.cpp
@@ -13,7 +13,7 @@ const std::vector<unsigned int> batchSizes = {
 };
 
 namespace {
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, DetectNetworkBatch,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, DetectNetworkBatch,
                             ::testing::Combine(
                                 ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                 ::testing::ValuesIn(batchSizes)),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/exec_graph_info.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/exec_graph_info.cpp
index e644dbc140bd93..1e62bb2da87034 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/exec_graph_info.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/exec_graph_info.cpp
@@ -15,7 +15,7 @@ namespace {
             {}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, ExecGraphTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, ExecGraphTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request.cpp
index 931fefbf391f8e..5e515c8b562e91 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request.cpp
@@ -17,14 +17,14 @@ const std::vector<std::map<std::string, std::string>> configs = {
         {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD}}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestTests,
         ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                 ::testing::Values(std::map<std::string, std::string>({}))),
         InferRequestTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferRequestTests,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferRequestTests,
         ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_MULTI),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_callback.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_callback.cpp
index 71c5b891cce86b..4bb24fca393597 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_callback.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_callback.cpp
@@ -18,14 +18,14 @@ const std::vector<std::map<std::string, std::string>> multiConfigs = {
         {{ MULTI_CONFIG_KEY(DEVICE_PRIORITIES) , CommonTestUtils::DEVICE_MYRIAD}}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, CallbackTests,
         ::testing::Combine(
             ::testing::ValuesIn(netPrecisions),
             ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
             ::testing::ValuesIn(configs)),
         CallbackTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, CallbackTests,
+INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, CallbackTests,
         ::testing::Combine(
                 ::testing::ValuesIn(netPrecisions),
                 ::testing::Values(CommonTestUtils::DEVICE_MULTI),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_config.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_config.cpp
index d3ff02668a3a2a..eba711f7b209fd 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_config.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_config.cpp
@@ -76,28 +76,28 @@ namespace {
              {VPU_CONFIG_KEY(HW_STAGES_OPTIMIZATION), CONFIG_VALUE(YES)}},
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferConfigTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                     ::testing::ValuesIn(configs)),
                             InferConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferConfigTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferConfigTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(multiConfigs)),
                             InferConfigTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferConfigInTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferConfigInTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                     ::testing::ValuesIn(inferConfigs)),
                             InferConfigInTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferConfigInTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferConfigInTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_input.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_input.cpp
index 201f76cf0c41a7..8b97210ca01939 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_input.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_input.cpp
@@ -19,14 +19,14 @@ namespace {
             {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestInputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestInputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                     ::testing::ValuesIn(configs)),
                             InferRequestInputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferRequestInputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferRequestInputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_output.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_output.cpp
index dfa9d4dfe3c8ac..4ef86349b9aa71 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_output.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/infer_request_output.cpp
@@ -19,14 +19,14 @@ namespace {
             {{InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES, CommonTestUtils::DEVICE_MYRIAD}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, InferRequestOutputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, InferRequestOutputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                     ::testing::ValuesIn(configs)),
                             InferRequestOutputTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, InferRequestOutputTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, InferRequestOutputTests,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/layout.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/layout.cpp
index b0f9d38c93b3a8..33f02720fef002 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/layout.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/layout.cpp
@@ -24,7 +24,7 @@ namespace {
             { 3 }
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, LayoutTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, LayoutTest,
             ::testing::Combine(
                     ::testing::Values(InferenceEngine::Precision::FP16),
                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/query_network.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/query_network.cpp
index bacbb6e4de2787..01d9aec269419c 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/query_network.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/query_network.cpp
@@ -16,7 +16,7 @@ auto TIwithLSTMcell = ngraph::builder::subgraph::makeTIwithLSTMcell();
 auto SplitConvConcat = ngraph::builder::subgraph::makeNestedSplitConvConcat();
 auto BranchSplitConvConcat = ngraph::builder::subgraph::makeSplitConvConcatNestedInBranch();
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, QueryNetworkTest,
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, QueryNetworkTest,
                         ::testing::Combine(
                                 ::testing::Values("MYRIAD", "HETERO:MYRIAD,CPU", "MULTI:MYRIAD,CPU"),
                                 ::testing::Values(ConvBias, TIwithLSTMcell, SplitConvConcat, BranchSplitConvConcat)),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/set_preprocess.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/set_preprocess.cpp
index 1071fa66da8691..7ef46d71e7e913 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/set_preprocess.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/set_preprocess.cpp
@@ -20,14 +20,14 @@ namespace {
             {{ InferenceEngine::MultiDeviceConfigParams::KEY_MULTI_DEVICE_PRIORITIES , CommonTestUtils::DEVICE_MYRIAD}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                     ::testing::ValuesIn(configs)),
                             PreprocessTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, PreprocessTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, PreprocessTest,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
@@ -48,7 +48,7 @@ namespace {
         InferenceEngine::Layout::NHWC
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessConversionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessConversionTest,
                                 ::testing::Combine(
                                         ::testing::ValuesIn(netPrecisions),
                                         ::testing::ValuesIn(ioPrecisions),
@@ -62,7 +62,7 @@ namespace {
                                         ::testing::ValuesIn(configs)),
                                 PreprocessConversionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, PreprocessDynamicallyInSetBlobTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Bool(),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/stress_tests.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/stress_tests.cpp
index 8655b88d39b564..8b65e6d3f5156b 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/stress_tests.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/stress_tests.cpp
@@ -9,7 +9,7 @@ using namespace LayerTestsDefinitions;
 const unsigned int g_BugAllocationLimit = 10000;
 
 namespace {
-    INSTANTIATE_TEST_CASE_P(nightly_BehaviorTests, MultipleAllocations,
+    INSTANTIATE_TEST_SUITE_P(nightly_BehaviorTests, MultipleAllocations,
                             ::testing::Combine(
                                 ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                 ::testing::Values(g_BugAllocationLimit)),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/test_plugin.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/test_plugin.cpp
index 7e7353aebb5ffa..c41206a5b43e53 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/test_plugin.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/test_plugin.cpp
@@ -29,42 +29,42 @@ namespace {
             InferenceEngine::Precision::FP16
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTestOutput,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecision),
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                     ::testing::ValuesIn(configs)),
                             BehaviorTestOutput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, BehaviorTestOutput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, BehaviorTestOutput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecision),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(MultiConfigsInputOutput)),
                             BehaviorTestOutput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTests,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                     ::testing::ValuesIn(configs)),
                             BehaviorTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, BehaviorTests,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, BehaviorTests,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(MultiConfigs)),
                             BehaviorTests::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, BehaviorTestInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                     ::testing::ValuesIn(configs)),
                             BehaviorTestInput::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, BehaviorTestInput,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, BehaviorTestInput,
                             ::testing::Combine(
                                     ::testing::ValuesIn(netPrecisions),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/version.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/version.cpp
index 8c7613862357d0..d5e9f9d9162702 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/version.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/behavior/version.cpp
@@ -18,21 +18,21 @@ namespace {
             {{ HETERO_CONFIG_KEY(DUMP_GRAPH_DOT) , CommonTestUtils::DEVICE_MYRIAD}}
     };
 
-    INSTANTIATE_TEST_CASE_P(smoke_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                                     ::testing::ValuesIn(configs)),
                             VersionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Multi_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Multi_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_MULTI),
                                     ::testing::ValuesIn(Multiconfigs)),
                             VersionTest::getTestCaseName);
 
-    INSTANTIATE_TEST_CASE_P(smoke_Hetero_BehaviorTests, VersionTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Hetero_BehaviorTests, VersionTest,
                             ::testing::Combine(
                                     ::testing::Values(InferenceEngine::Precision::FP32),
                                     ::testing::Values(CommonTestUtils::DEVICE_HETERO),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/import_export_tests/import_nonzero.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/import_export_tests/import_nonzero.cpp
index ccdbb4182af9a3..4555ac1ec9d0c1 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/import_export_tests/import_nonzero.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/import_export_tests/import_nonzero.cpp
@@ -26,7 +26,7 @@ const std::vector<std::string> appHeaders = {
         "APPLICATION_HEADER"
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ImportNetworkCase, ImportNonZero,
+INSTANTIATE_TEST_SUITE_P(smoke_ImportNetworkCase, ImportNonZero,
                         ::testing::Combine(
                             ::testing::ValuesIn(netPrecisions),
                             ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/myriad_remote_blobs_tests.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/myriad_remote_blobs_tests.cpp
index 7a6948044cb797..15324128621e0e 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/myriad_remote_blobs_tests.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/myriad_remote_blobs_tests.cpp
@@ -14,5 +14,5 @@ const std::vector<DevicesNamesAndSupportPair> device_names_and_support_for_remot
 #endif
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_RemoteBlobMultiMyriad, MultiDevice_SupportTest,
+INSTANTIATE_TEST_SUITE_P(smoke_RemoteBlobMultiMyriad, MultiDevice_SupportTest,
                         ::testing::ValuesIn(device_names_and_support_for_remote_blobs), MultiDevice_SupportTest::getTestCaseName);
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/activation.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/activation.cpp
index e87e73f6e6f47c..7fff88eb0bbaa6 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/activation.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/activation.cpp
@@ -48,6 +48,6 @@ const auto basicCases = ::testing::Combine(
 );
 
 
-INSTANTIATE_TEST_CASE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_Activation_Basic, ActivationLayerTest, basicCases, ActivationLayerTest::getTestCaseName);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/comparison.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/comparison.cpp
index f4cfc820195b7e..4485bbdea68516 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/comparison.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/comparison.cpp
@@ -50,7 +50,7 @@ std::vector<ngraph::helpers::ComparisonTypes> comparisonOpTypesFpToInt = {
         ngraph::helpers::ComparisonTypes::LESS,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_ComparisonFp,
+INSTANTIATE_TEST_SUITE_P(smoke_ComparisonFp,
                         ComparisonLayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(CommonTestUtils::combineParams(inputShapes)),
@@ -63,7 +63,7 @@ INSTANTIATE_TEST_CASE_P(smoke_ComparisonFp,
                             ::testing::Values(Config{{InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO)}})),
                         ComparisonLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_ComparisonInt,
+INSTANTIATE_TEST_SUITE_P(smoke_ComparisonInt,
                         ComparisonLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(CommonTestUtils::combineParams(inputShapes)),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/concat.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/concat.cpp
index f9500dad7e6226..228c892fa3ac3e 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/concat.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/concat.cpp
@@ -20,7 +20,7 @@ std::vector<std::vector<std::vector<size_t>>> inShapes = {
 std::vector<InferenceEngine::Precision> netPrecisions = {InferenceEngine::Precision::FP16};
 
 
-INSTANTIATE_TEST_CASE_P(smoke_Concat_Basic, ConcatLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Concat_Basic, ConcatLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(axes),
                                 ::testing::ValuesIn(inShapes),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/convolution.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/convolution.cpp
index ac5285cad1bad9..43a582b89fe71f 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/convolution.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/convolution.cpp
@@ -60,7 +60,7 @@ const auto conv2DParams_BigDimensionValid = ::testing::Combine(
         ::testing::Values(ngraph::op::PadType::VALID)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_ExplicitPadding, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D_ExplicitPadding, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_ExplicitPadding,
                                 ::testing::ValuesIn(netPrecisions),
@@ -72,7 +72,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_ExplicitPadding, ConvolutionLayerTes
                                 ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)),
                         ConvolutionLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_AutoPadValid, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D_AutoPadValid, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_AutoPadValid,
                                 ::testing::ValuesIn(netPrecisions),
@@ -83,7 +83,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_AutoPadValid, ConvolutionLayerTest,
                                 ::testing::Values(std::vector<size_t >({1, 3, 30, 30})),
                                 ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)),
                         ConvolutionLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_Convolution2D_BigDimensionValid, ConvolutionLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Convolution2D_BigDimensionValid, ConvolutionLayerTest,
                         ::testing::Combine(
                                 conv2DParams_BigDimensionValid,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp
index c3866b9621ccc2..5025da6e91f61a 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/ctc_greedy_decoder_seq_len.cpp
@@ -29,7 +29,7 @@ const auto sequenceLengths = std::vector<int>{1, 10, 50, 100};
 
 const auto blankIndexes = std::vector<int>{0, 10, 100};
 
-INSTANTIATE_TEST_CASE_P(smoke, CTCGreedyDecoderSeqLenLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke, CTCGreedyDecoderSeqLenLayerTest,
         ::testing::Combine(
                         ::testing::ValuesIn(inputShape),
                         ::testing::ValuesIn(sequenceLengths),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/eltwise.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/eltwise.cpp
index dc258dec53825e..6821c49494a185 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/eltwise.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/eltwise.cpp
@@ -55,7 +55,7 @@ std::vector<ngraph::helpers::EltwiseTypes> eltwiseMathTypesINT = {
         ngraph::helpers::EltwiseTypes::DIVIDE,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseMathFP,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseMathFP,
                         EltwiseLayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(inShapes),
@@ -70,7 +70,7 @@ INSTANTIATE_TEST_CASE_P(smoke_EltwiseMathFP,
                             ::testing::Values(Config{{InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO)}})),
                         EltwiseLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseMathInt,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseMathInt,
                         EltwiseLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inShapes),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/gather_elements.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/gather_elements.cpp
index e6dae4781080fa..55843e804bf32d 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/gather_elements.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/gather_elements.cpp
@@ -33,7 +33,7 @@ const std::vector<InferenceEngine::Precision> iPrecisions = {
         InferenceEngine::Precision::I32
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_GatherElements1, GatherElementsLayerTestVPU,
+INSTANTIATE_TEST_SUITE_P(smoke_GatherElements1, GatherElementsLayerTestVPU,
                         ::testing::Combine(
                             ::testing::Values(std::vector<size_t>({2, 2})),   // Data shape
                             ::testing::Values(std::vector<size_t>({2, 2})),   // Indices shape
@@ -43,7 +43,7 @@ INSTANTIATE_TEST_CASE_P(smoke_GatherElements1, GatherElementsLayerTestVPU,
                             ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)),
                         GatherElementsLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_GatherElements2, GatherElementsLayerTestVPU,
+INSTANTIATE_TEST_SUITE_P(smoke_GatherElements2, GatherElementsLayerTestVPU,
                         ::testing::Combine(
                             ::testing::Values(std::vector<size_t>({2, 65, 300})),  // Data shape
                             ::testing::Values(std::vector<size_t>({2, 65, 64})),   // Indices shape
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/gather_nd.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/gather_nd.cpp
index d6dbf6f534913c..dc01dde5d878a3 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/gather_nd.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/gather_nd.cpp
@@ -39,7 +39,7 @@ const std::vector<GatherNDParamsSubset> layerParams = {
     GatherNDParamsSubset{{2, 2, 2, 2}, {2, 2, 1}, 2},
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_GatherND,
     GatherNDLayerTest,
     testing::Combine(
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/interpolate.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/interpolate.cpp
index 3c7d701c09ef75..31d17ca51ba607 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/interpolate.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/interpolate.cpp
@@ -179,7 +179,7 @@ const auto interpolateCasesWithoutNearestMode = ::testing::Combine(
         ::testing::ValuesIn(defaultAxes),
         ::testing::ValuesIn(defaultScales));
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_nearest_mode_no_batch, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_nearest_mode_no_batch, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesNearestMode,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -192,7 +192,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Interpolate_nearest_mode_no_batch, InterpolateLaye
         ::testing::Values(Config{{InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO)}})),
     InterpolateLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_without_nearest_mode_no_batch, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_without_nearest_mode_no_batch, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesWithoutNearestMode,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -205,7 +205,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Interpolate_without_nearest_mode_no_batch, Interpo
         ::testing::Values(Config{{InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO)}})),
     InterpolateLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_nearest_mode, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_nearest_mode, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesNearestMode,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -218,7 +218,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Interpolate_nearest_mode, InterpolateLayerTest, ::
         ::testing::Values(Config{{InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO)}})),
     InterpolateLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_nearest_mode_more, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_nearest_mode_more, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesNearestModeMore,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -231,7 +231,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Interpolate_nearest_mode_more, InterpolateLayerTes
         ::testing::Values(Config{{InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO)}})),
     InterpolateLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_Interpolate_without_nearest, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_Interpolate_without_nearest, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesWithoutNearestMode,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -244,7 +244,7 @@ INSTANTIATE_TEST_CASE_P(smoke_Interpolate_without_nearest, InterpolateLayerTest,
         ::testing::Values(Config{{InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO)}})),
     InterpolateLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(nightly_Interpolate_without_nearest, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(nightly_Interpolate_without_nearest, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesWithoutNearestMode,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
@@ -257,7 +257,7 @@ INSTANTIATE_TEST_CASE_P(nightly_Interpolate_without_nearest, InterpolateLayerTes
         ::testing::Values(Config{{InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO)}})),
     InterpolateLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(nightly_Interpolate_nearest_mode_2x, InterpolateLayerTest, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(nightly_Interpolate_nearest_mode_2x, InterpolateLayerTest, ::testing::Combine(
         interpolateCasesNearestMode2x,
         ::testing::ValuesIn(netPrecisions),
         ::testing::Values(InferenceEngine::Precision::UNSPECIFIED),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/logical.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/logical.cpp
index 8c1202dba66551..d4dd50db3e18b7 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/logical.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/logical.cpp
@@ -66,7 +66,7 @@ std::map<std::string, std::string> additional_config = {
     {InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO)}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseLogicalInt,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseLogicalInt,
                         LogicalLayerTestVPU,
                         ::testing::Combine(
                                 ::testing::ValuesIn(LogicalLayerTest::combineShapes(inputShapes)),
@@ -81,7 +81,7 @@ INSTANTIATE_TEST_CASE_P(smoke_EltwiseLogicalInt,
                                 ::testing::Values(additional_config)),
                         LogicalLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_EltwiseLogicalNotInt,
+INSTANTIATE_TEST_SUITE_P(smoke_EltwiseLogicalNotInt,
                         LogicalLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(LogicalLayerTest::combineShapes(inputShapesNot)),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/mat_mul.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/mat_mul.cpp
index f947367633bba5..fcfa981c0e827f 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/mat_mul.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/mat_mul.cpp
@@ -31,7 +31,7 @@ Config additionalConfig = {
         {InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO)}
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MatMul, MatMulTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MatMul, MatMulTest,
         ::testing::Combine(
             ::testing::ValuesIn(shapeRelatedParams),
             ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/minimum_maximum.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/minimum_maximum.cpp
index 8db8e1297d32df..d174e19756d1ef 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/minimum_maximum.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/minimum_maximum.cpp
@@ -33,7 +33,7 @@ const std::vector<ngraph::helpers::InputLayerType> inputType = {
         ngraph::helpers::InputLayerType::PARAMETER,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_maximum, MaxMinLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_maximum, MaxMinLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inShapes),
                                 ::testing::ValuesIn(opType),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/mvn.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/mvn.cpp
index 1dc99bdd642978..2c78d5c6698a36 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/mvn.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/mvn.cpp
@@ -35,7 +35,7 @@ const std::vector<float> eps = {
         1.0e-10, 1.0e-8, 1.0e-7, 1.0e-5, 1.0e-3
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_4D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_4D, Mvn6LayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(input_shape_4D),
                                 ::testing::Values(InferenceEngine::Precision::FP16),
@@ -47,7 +47,7 @@ INSTANTIATE_TEST_CASE_P(smoke_MVN_4D, Mvn6LayerTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)),
                         Mvn6LayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_MVN_3D, Mvn6LayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MVN_3D, Mvn6LayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(input_shape_3D),
                                 ::testing::Values(InferenceEngine::Precision::FP16),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/nonzero.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/nonzero.cpp
index 592236bb34bd23..b418a50da73f07 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/nonzero.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/nonzero.cpp
@@ -38,7 +38,7 @@ const std::vector<InferenceEngine::Precision> inputPrecisions = {
         InferenceEngine::Precision::U8,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_nonzero, NonZeroLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_nonzero, NonZeroLayerTest,
         ::testing::Combine(
                 ::testing::ValuesIn(inShapes),
                 ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/pad.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/pad.cpp
index ce681365a1c2b2..a4b5d32c220f24 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/pad.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/pad.cpp
@@ -37,7 +37,7 @@ const auto pad4DConstParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_MYRIAD)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad4DConst,
         PadLayerTest,
         pad4DConstParams,
@@ -57,7 +57,7 @@ const auto pad4DParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_MYRIAD)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad4D,
         PadLayerTest,
         pad4DParams,
@@ -77,7 +77,7 @@ const auto pad3DConstParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_MYRIAD)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad3DConst,
         PadLayerTest,
         pad3DConstParams,
@@ -97,7 +97,7 @@ const auto pad3DParams = testing::Combine(
         testing::Values(CommonTestUtils::DEVICE_MYRIAD)
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         smoke_Pad3D,
         PadLayerTest,
         pad3DParams,
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/pooling.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/pooling.cpp
index 0909880261ba90..83f4c60039dc46 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/pooling.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/pooling.cpp
@@ -40,7 +40,7 @@ const auto maxPool_ExplicitPad_FloorRounding_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_ExplicitPad_FloorRpunding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_ExplicitPad_FloorRpunding, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_ExplicitPad_FloorRounding_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -67,7 +67,7 @@ const auto maxPool_ExplicitPad_CeilRounding_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MaxPool_ExplicitPad_CeilRpunding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MaxPool_ExplicitPad_CeilRpunding, PoolingLayerTest,
                         ::testing::Combine(
                                 maxPool_ExplicitPad_CeilRounding_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -96,7 +96,7 @@ const auto avgPoolExplicitPadCeilRoundingParams = ::testing::Combine(
         ::testing::Values(true, false)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_ExplicitPad_CeilRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_ExplicitPad_CeilRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 avgPoolExplicitPadCeilRoundingParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -122,7 +122,7 @@ const auto avgPoolExplicitPadFloorRoundingParams = ::testing::Combine(
 );
 
 
-INSTANTIATE_TEST_CASE_P(smoke_AvgPool_ExplicitPad_FloorRounding, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_AvgPool_ExplicitPad_FloorRounding, PoolingLayerTest,
                         ::testing::Combine(
                                 avgPoolExplicitPadFloorRoundingParams,
                                 ::testing::ValuesIn(netPrecisions),
@@ -149,7 +149,7 @@ const auto allPools_ValidPad_Params = ::testing::Combine(
         ::testing::Values(false)  // placeholder value - exclude pad not applicable for max pooling
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_MAX_and_AVGPool_ValidPad, PoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_MAX_and_AVGPool_ValidPad, PoolingLayerTest,
                         ::testing::Combine(
                                 allPools_ValidPad_Params,
                                 ::testing::ValuesIn(netPrecisions),
@@ -175,7 +175,7 @@ const auto avgGlobalPoolParams = ::testing::Combine(
         ::testing::Values(false)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_GlobalAvgPool, GlobalPoolingLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_GlobalAvgPool, GlobalPoolingLayerTest,
                         ::testing::Combine(
                                 avgGlobalPoolParams,
                                 ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/proposal.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/proposal.cpp
index 103dff03d6f15d..438274a55f1d3f 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/proposal.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/proposal.cpp
@@ -40,7 +40,7 @@ const auto proposalParams = ::testing::Combine(
         ::testing::ValuesIn(framework_)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_Proposal_tests, ProposalLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Proposal_tests, ProposalLayerTest,
                         ::testing::Combine(
                                 proposalParams,
                                 ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/roi_pooling.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/roi_pooling.cpp
index 66baf439e144bf..d28509dbe33bf2 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/roi_pooling.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/roi_pooling.cpp
@@ -52,7 +52,7 @@ const auto test_ROIPooling_bilinear = ::testing::Combine(
     ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)
 );
 
-INSTANTIATE_TEST_CASE_P(smoke_TestsROIPooling_max, ROIPoolingLayerTest, test_ROIPooling_max, ROIPoolingLayerTest::getTestCaseName);
-INSTANTIATE_TEST_CASE_P(smoke_TestsROIPooling_bilinear, ROIPoolingLayerTest, test_ROIPooling_bilinear, ROIPoolingLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsROIPooling_max, ROIPoolingLayerTest, test_ROIPooling_max, ROIPoolingLayerTest::getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_TestsROIPooling_bilinear, ROIPoolingLayerTest, test_ROIPooling_bilinear, ROIPoolingLayerTest::getTestCaseName);
 
 } // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/split.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/split.cpp
index 11afafbc42ac4f..5e69eebc5032c4 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/split.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/split.cpp
@@ -14,7 +14,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP16
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NumSplitsCheck, SplitLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_NumSplitsCheck, SplitLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(5),
                                 // TODO: 0-axis excluded
@@ -32,7 +32,7 @@ INSTANTIATE_TEST_CASE_P(smoke_NumSplitsCheck, SplitLayerTest,
                                 ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)),
                         SplitLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_splitWithUnusedOutputsTest, SplitLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_splitWithUnusedOutputsTest, SplitLayerTest,
                         ::testing::Combine(
                                 ::testing::Values(5),
                                 // TODO: 0-axis excluded
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/strided_slice.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/strided_slice.cpp
index 6c4003a3f3a08f..33893ae4241933 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/strided_slice.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/strided_slice.cpp
@@ -55,7 +55,7 @@ Config getConfig() {
     return Config{ { InferenceEngine::MYRIAD_DETECT_NETWORK_BATCH, CONFIG_VALUE(NO) } };
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_StridedSlice_tests, StridedSliceLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_StridedSlice_tests, StridedSliceLayerTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(testCases),
                             ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/topk.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/topk.cpp
index 9b3e4c7239d4ae..722f528d876e62 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/topk.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/single_layer_tests/topk.cpp
@@ -40,7 +40,7 @@ const std::vector<ngraph::opset5::TopK::SortType> sortTypes = {
         ngraph::opset5::TopK::SortType::SORT_VALUES,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_TopK_IndicesValuesSort, TopKLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_TopK_IndicesValuesSort, TopKLayerTest,
         ::testing::Combine(
             ::testing::ValuesIn(k),
             ::testing::ValuesIn(axes),
@@ -54,7 +54,7 @@ INSTANTIATE_TEST_CASE_P(smoke_TopK_IndicesValuesSort, TopKLayerTest,
             ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)),
         TopKLayerTest::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(smoke_TopK_NoneSort, TopKLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_TopK_NoneSort, TopKLayerTest,
         ::testing::Combine(
             ::testing::Values(1),
             ::testing::ValuesIn(axes),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/constant_result.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/constant_result.cpp
index 19fdc4ba9bf34f..919fcdef3d05a1 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/constant_result.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/constant_result.cpp
@@ -34,7 +34,7 @@ const std::vector<Precision> precisions = {
     Precision::BOOL
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_Check, ConstantResultSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Check, ConstantResultSubgraphTest,
                         ::testing::Combine(
                             ::testing::ValuesIn(types),
                             ::testing::ValuesIn(shapes),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
index bdc85b7ae9ff43..6636ef971ff19f 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/reshape_squeeze_reshape_relu.cpp
@@ -32,7 +32,7 @@ namespace {
             ngraph::helpers::SqueezeOpType::UNSQUEEZE
     };
 
-    INSTANTIATE_TEST_CASE_P(DISABLED_reshape_squeeze_reshape_relu, ReshapeSqueezeReshapeRelu,
+    INSTANTIATE_TEST_SUITE_P(DISABLED_reshape_squeeze_reshape_relu, ReshapeSqueezeReshapeRelu,
                             ::testing::Combine(
                                     ::testing::ValuesIn(inputs),
                                     ::testing::ValuesIn(netPrecisions),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/split_conv_concat.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
index 2688eed42fc06c..82224e56375d20 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/split_conv_concat.cpp
@@ -15,7 +15,7 @@ const std::vector<InferenceEngine::Precision> netPrecisions = {
         InferenceEngine::Precision::FP16
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_NoReshape, SplitConvConcat,
+INSTANTIATE_TEST_SUITE_P(smoke_NoReshape, SplitConvConcat,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values(std::vector<size_t >({1, 6, 40, 40})),
diff --git a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/tensor_names.cpp b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/tensor_names.cpp
index b89622fb297f82..76fbf656b6f808 100644
--- a/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/tensor_names.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/shared_tests_instances/subgraph_tests/tensor_names.cpp
@@ -10,7 +10,7 @@
 using namespace SubgraphTestsDefinitions;
 
 namespace {
-    INSTANTIATE_TEST_CASE_P(smoke_Check, TensorNamesTest,
+    INSTANTIATE_TEST_SUITE_P(smoke_Check, TensorNamesTest,
                             ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                             TensorNamesTest::getTestCaseName);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/single_layer_tests/gather.cpp b/inference-engine/tests/functional/plugin/myriad/single_layer_tests/gather.cpp
index d882c3ba7c2a82..96e245fad63fce 100644
--- a/inference-engine/tests/functional/plugin/myriad/single_layer_tests/gather.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/single_layer_tests/gather.cpp
@@ -108,7 +108,7 @@ TEST_P(MyriadGatherLayerTest, accuracy) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_Gather,
     MyriadGatherLayerTest,
     testing::Combine(
diff --git a/inference-engine/tests/functional/plugin/myriad/single_layer_tests/out_shape_of_reshape.cpp b/inference-engine/tests/functional/plugin/myriad/single_layer_tests/out_shape_of_reshape.cpp
index 61b27f909e758e..681c14d3ce61ab 100644
--- a/inference-engine/tests/functional/plugin/myriad/single_layer_tests/out_shape_of_reshape.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/single_layer_tests/out_shape_of_reshape.cpp
@@ -111,7 +111,7 @@ std::vector<OutShapeOfReshapeParam> shapeParams = {
         std::make_tuple(InputShape{ 2, 5,   5,   0 }, ShapeDescriptor{ 0,  4          }, false),
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_accuracy, OutShapeOfReshapeLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_accuracy, OutShapeOfReshapeLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(shapeParams),
                                 ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)),
diff --git a/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_broadcast.cpp b/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_broadcast.cpp
index fb1f38f1ae46ea..a1f62c1ce1a8af 100644
--- a/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_broadcast.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_broadcast.cpp
@@ -124,7 +124,7 @@ std::vector<InferenceEngine::Precision> broadcastPrecisions = {
         InferenceEngine::Precision::I32,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_accuracy, StaticShapeBroadcastLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_accuracy, StaticShapeBroadcastLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(broadcastParam),
                                 ::testing::ValuesIn(broadcastPrecisions),
diff --git a/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_nms.cpp b/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_nms.cpp
index ea5a44ed98817f..83dd6dd2b04fe4 100644
--- a/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_nms.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_nms.cpp
@@ -115,7 +115,7 @@ std::vector<InferenceEngine::Precision> NMSPrecisions = {
 };
 
 // #-30919
-INSTANTIATE_TEST_CASE_P(DISABLED_accuracy, StaticShapeNMSLayerTest,
+INSTANTIATE_TEST_SUITE_P(DISABLED_accuracy, StaticShapeNMSLayerTest,
         ::testing::Combine(
         ::testing::ValuesIn(NMSParams),
         ::testing::ValuesIn(NMSPrecisions),
@@ -170,7 +170,7 @@ TEST_P(NMS1toStaticShapeNMS, PreviousNMSCanBeLoaded) {
     ASSERT_NO_THROW(LoadNetwork());
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NetworkLoad, NMS1toStaticShapeNMS,
+INSTANTIATE_TEST_SUITE_P(smoke_NetworkLoad, NMS1toStaticShapeNMS,
                         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD));
 
 class NMS3toStaticShapeNMS : public PreviousNMStoStaticShapeNMS {
@@ -190,7 +190,7 @@ TEST_P(NMS3toStaticShapeNMS, PreviousNMSCanBeLoaded) {
     ASSERT_NO_THROW(LoadNetwork());
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NetworkLoad, NMS3toStaticShapeNMS,
+INSTANTIATE_TEST_SUITE_P(smoke_NetworkLoad, NMS3toStaticShapeNMS,
                             ::testing::Values(CommonTestUtils::DEVICE_MYRIAD));
 
 class NMS4toStaticShapeNMS : public PreviousNMStoStaticShapeNMS {
@@ -211,7 +211,7 @@ TEST_P(NMS4toStaticShapeNMS, PreviousNMSCanBeLoaded) {
     ASSERT_NO_THROW(LoadNetwork());
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NetworkLoad, NMS4toStaticShapeNMS,
+INSTANTIATE_TEST_SUITE_P(smoke_NetworkLoad, NMS4toStaticShapeNMS,
                         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD));
 
 }  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_nonzero.cpp b/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_nonzero.cpp
index 6833f2b0f2d406..c268a97753c363 100644
--- a/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_nonzero.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/single_layer_tests/static_shape_nonzero.cpp
@@ -111,7 +111,7 @@ std::vector<InferenceEngine::Precision> inputPrecisions = {
         InferenceEngine::Precision::I32,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_accuracy, StaticShapeNonZeroLayerTest,
+INSTANTIATE_TEST_SUITE_P(smoke_accuracy, StaticShapeNonZeroLayerTest,
                         ::testing::Combine(
                                 ::testing::ValuesIn(inputDims),
                                 ::testing::ValuesIn(inputPrecisions),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/concat_split_transpose.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/concat_split_transpose.cpp
index 7845ee434f5507..1293b7e9609c7f 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/concat_split_transpose.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/concat_split_transpose.cpp
@@ -74,7 +74,7 @@ std::vector<std::vector<std::size_t>> length = {
         {500, 500}
 };
 
-INSTANTIATE_TEST_CASE_P(SpecialStages, Concat_Split_Transpose,
+INSTANTIATE_TEST_SUITE_P(SpecialStages, Concat_Split_Transpose,
                         ::testing::Combine(
                                 ::testing::Values(ngraph::element::i32),
                                 ::testing::ValuesIn(dims),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_binary_elementwise.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_binary_elementwise.cpp
index 279dc7a783d291..003799dba8452b 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_binary_elementwise.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_binary_elementwise.cpp
@@ -92,7 +92,7 @@ TEST_P(DSR_BinaryElementwiseBothDSR, CompareWithReference) {
     const auto& eltwiseType = std::get<2>(GetParam());
 
     if (doNotSupportI32.count(eltwiseType) && inDataType == ngraph::element::i32) {
-        SKIP() << eltwiseType.name << " doesn't support int32_t inputs" << std::endl;
+        GTEST_SKIP() << eltwiseType.name << " doesn't support int32_t inputs" << std::endl;
     }
 
     Run();
@@ -105,7 +105,7 @@ std::vector<BinaryEltwiseShapes> dataShapesWithUpperBound = {
         },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicBinaryElementwise, DSR_BinaryElementwiseBothDSR,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicBinaryElementwise, DSR_BinaryElementwiseBothDSR,
     ::testing::Combine(
         ::testing::Values(ngraph::element::f16, ngraph::element::f32, ngraph::element::i32),
         ::testing::ValuesIn(dataShapesWithUpperBound),
@@ -117,7 +117,7 @@ TEST_P(DSR_BinaryElementwiseSingleDSR, CompareWithReference) {
     const auto& eltwiseType = std::get<2>(GetParam());
 
     if (doNotSupportI32.count(eltwiseType) && inDataType == ngraph::element::i32) {
-        SKIP() << eltwiseType.name << " doesn't support int32_t inputs" << std::endl;
+        GTEST_SKIP() << eltwiseType.name << " doesn't support int32_t inputs" << std::endl;
     }
 
     Run();
@@ -130,7 +130,7 @@ std::vector<BinaryEltwiseShapes> dataShapesWithUpperBoundSingleDSR = {
         },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicBinaryElementwiseSingleDSR, DSR_BinaryElementwiseSingleDSR,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicBinaryElementwiseSingleDSR, DSR_BinaryElementwiseSingleDSR,
     ::testing::Combine(
         ::testing::Values(ngraph::element::f16, ngraph::element::f32, ngraph::element::i32),
         ::testing::ValuesIn(dataShapesWithUpperBoundSingleDSR),
@@ -169,7 +169,7 @@ TEST_P(DSR_BinaryElementwiseBothDSRCheckOutputShape, CheckOutputShape) {
     const auto& eltwiseType = std::get<2>(GetParam());
 
     if (doNotSupportI32.count(eltwiseType) && inDataType == ngraph::element::i32) {
-        SKIP() << eltwiseType.name << " doesn't support int32_t inputs" << std::endl;
+        GTEST_SKIP() << eltwiseType.name << " doesn't support int32_t inputs" << std::endl;
     }
 
     Run();
@@ -334,7 +334,7 @@ std::vector<BinaryEltwiseShapes> dataShapesWithUpperBoundBothDSREmpty = {
     },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BinaryElementwiseBothDSRCheckOutputShape, DSR_BinaryElementwiseBothDSRCheckOutputShape,
+INSTANTIATE_TEST_SUITE_P(smoke_BinaryElementwiseBothDSRCheckOutputShape, DSR_BinaryElementwiseBothDSRCheckOutputShape,
     ::testing::Combine(
         ::testing::Values(ngraph::element::f16, ngraph::element::f32, ngraph::element::i32),
         ::testing::ValuesIn(dataShapesWithUpperBoundBothDSREmpty),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_clamp.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_clamp.cpp
index 1cc7abf4ca9bb2..6d80220fbccee2 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_clamp.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_clamp.cpp
@@ -38,7 +38,7 @@ TEST_P(DSR_Clamp, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicClamp, DSR_Clamp,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicClamp, DSR_Clamp,
     ::testing::Combine(
         ::testing::Values(ngraph::element::f16, ngraph::element::f32),
         ::testing::Values(DataShapeWithUpperBound{DataShape{1, 800}, DataShape{2, 1000}}),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_concat.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_concat.cpp
index 58d3c5d4357970..81d1f3219f30a2 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_concat.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_concat.cpp
@@ -98,7 +98,7 @@ std::vector<ConcatParam> concatParams = {
         },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicConcat, DSR_Concat, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicConcat, DSR_Concat, ::testing::Combine(
         ::testing::ValuesIn(dataTypes),
         ::testing::ValuesIn(concatParams),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)));
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_convert.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_convert.cpp
index 7ba067a7e7494c..99af45c332378b 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_convert.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_convert.cpp
@@ -49,7 +49,7 @@ std::vector<DataTypeConversionPair> dataTypeConversionPairVector {
     {ngraph::element::f16, ngraph::element::i32},
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicConvert, DSR_Convert,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicConvert, DSR_Convert,
     ::testing::Combine(
         ::testing::ValuesIn(dataTypeConversionPairVector),
         ::testing::Values(DataShapeWithUpperBound{ngraph::Shape{1, 800}, ngraph::Shape{2, 1000}},
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather.cpp
index 628383b5bf6fa3..3d4b5be8dfb83c 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather.cpp
@@ -50,7 +50,7 @@ TEST_P(DSR_GatherDynamicDataStaticIdx, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicGatherData, DSR_GatherDynamicDataStaticIdx, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicGatherData, DSR_GatherDynamicDataStaticIdx, testing::Combine(
         testing::ValuesIn(dataTypeVector),
         testing::ValuesIn(idxTypeVector),
         testing::Values(
@@ -88,7 +88,7 @@ TEST_P(DSR_GatherStaticDataDynamicIdx, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicGatherIdx, DSR_GatherStaticDataDynamicIdx, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicGatherIdx, DSR_GatherStaticDataDynamicIdx, testing::Combine(
         testing::ValuesIn(dataTypeVector),
         testing::ValuesIn(idxTypeVector),
         testing::Values(
@@ -125,7 +125,7 @@ TEST_P(DSR_GatherDynamicDataDynamicIdx, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicGather, DSR_GatherDynamicDataDynamicIdx, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicGather, DSR_GatherDynamicDataDynamicIdx, testing::Combine(
         testing::ValuesIn(dataTypeVector),
         testing::ValuesIn(idxTypeVector),
         testing::Values(
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather_elements.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather_elements.cpp
index fd27c196162d53..0529858d58112f 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather_elements.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather_elements.cpp
@@ -49,7 +49,7 @@ TEST_P(DSR_GatherElements, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicGatherElements, DSR_GatherElements,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicGatherElements, DSR_GatherElements,
     testing::Combine(
         testing::ValuesIn(dataTypeVector),
         testing::ValuesIn(idxTypeVector),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather_nd.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather_nd.cpp
index 0b0b2a7db16f8c..5003ea60685f88 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather_nd.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_gather_nd.cpp
@@ -82,7 +82,7 @@ TEST_P(DSR_GatherNDDynamicDataStaticIdx, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicGatherData, DSR_GatherNDDynamicDataStaticIdx, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicGatherData, DSR_GatherNDDynamicDataStaticIdx, testing::Combine(
     testing::ValuesIn(dataTypeVector),
     testing::ValuesIn(idxTypeVector),
     testing::Values(
@@ -112,7 +112,7 @@ TEST_P(DSR_GatherNDStaticDataDynamicIdx, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicGatherIdx, DSR_GatherNDStaticDataDynamicIdx, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicGatherIdx, DSR_GatherNDStaticDataDynamicIdx, testing::Combine(
     testing::ValuesIn(dataTypeVector),
     testing::ValuesIn(idxTypeVector),
     testing::Values(
@@ -142,7 +142,7 @@ TEST_P(DSR_GatherNDDynamicDataDynamicIdx, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicGather, DSR_GatherNDDynamicDataDynamicIdx, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicGather, DSR_GatherNDDynamicDataDynamicIdx, testing::Combine(
     testing::ValuesIn(dataTypeVector),
     testing::ValuesIn(idxTypeVector),
     testing::Values(
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_matmul.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_matmul.cpp
index 536da208e72d41..2bf60b55441c65 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_matmul.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_matmul.cpp
@@ -99,5 +99,5 @@ TEST_P(DSR_MatMul, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicMatMul, DSR_MatMul, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicMatMul, DSR_MatMul, combinations);
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_non_max_suppression.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_non_max_suppression.cpp
index a764cb07c338e5..a3e86b211e1f5b 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_non_max_suppression.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_non_max_suppression.cpp
@@ -91,7 +91,7 @@ TEST_P(DSR_NonMaxSuppression, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicNonMaxSupression, DSR_NonMaxSuppression,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicNonMaxSupression, DSR_NonMaxSuppression,
     ::testing::Combine(
         ::testing::Values(
             ngraph::element::f16,
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_reduce.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_reduce.cpp
index 20fb3c905a92c1..2a0a89351ce895 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_reduce.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_reduce.cpp
@@ -101,7 +101,7 @@ TEST_P(DSR_Reduce, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicArithmeticReduce, DSR_Reduce, arithmeticCombinations);
-INSTANTIATE_TEST_CASE_P(smoke_DynamicLogicalReduce, DSR_Reduce, logicalCombinations);
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicArithmeticReduce, DSR_Reduce, arithmeticCombinations);
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicLogicalReduce, DSR_Reduce, logicalCombinations);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_reshape.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_reshape.cpp
index 2aa408c7a5f6a2..d416dd7b357d0d 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_reshape.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_reshape.cpp
@@ -88,13 +88,13 @@ const std::vector<ngraph::element::Type> dataTypesVector = {
         ngraph::element::i32,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicReshape, DSR_ReshapeWithStaticDescriptor,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicReshape, DSR_ReshapeWithStaticDescriptor,
     ::testing::Combine(
         ::testing::ValuesIn(dataTypesVector),
         ::testing::ValuesIn(reshapeTestParams),
         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD)));
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicReshape, DSR_ReshapeWithDynamicDescriptor,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicReshape, DSR_ReshapeWithDynamicDescriptor,
     ::testing::Combine(
         ::testing::ValuesIn(dataTypesVector),
         ::testing::ValuesIn(reshapeTestParams),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_roialign.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_roialign.cpp
index c3e55b5ce93ad2..19f07085ace75b 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_roialign.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_roialign.cpp
@@ -54,7 +54,7 @@ TEST_P(DSR_ROIAlignDataDSR, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(DISABLED_DynamicROIAlignDataDSR, DSR_ROIAlignDataDSR,
+INSTANTIATE_TEST_SUITE_P(DISABLED_DynamicROIAlignDataDSR, DSR_ROIAlignDataDSR,
     ::testing::Combine(
         ::testing::Values(
                     ngraph::element::f16,
@@ -99,7 +99,7 @@ TEST_P(DSR_ROIAlignROIDSR, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_DynamicROIAlign, DSR_ROIAlignROIDSR,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_DynamicROIAlign, DSR_ROIAlignROIDSR,
     ::testing::Combine(
         ::testing::Values(
                     ngraph::element::f16,
@@ -148,7 +148,7 @@ TEST_P(DSR_ROIAlign, CompareWithReference) {
 }
 
 // #-30909
-INSTANTIATE_TEST_CASE_P(DISABLED_smoke_DynamicROIAlign, DSR_ROIAlign,
+INSTANTIATE_TEST_SUITE_P(DISABLED_smoke_DynamicROIAlign, DSR_ROIAlign,
     ::testing::Combine(
         ::testing::Values(
                     ngraph::element::f16,
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_scatter.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_scatter.cpp
index 0aa93b16d50ba4..2871505b726355 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_scatter.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_scatter.cpp
@@ -44,7 +44,7 @@ TEST_P(DSR_Scatter, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicScatter, DSR_Scatter,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicScatter, DSR_Scatter,
     ::testing::Combine(
             testing::Values(
                     ngraph::element::f16),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_split.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_split.cpp
index 0f8164bc831d8e..2378fdcc5185fa 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_split.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_split.cpp
@@ -56,6 +56,6 @@ TEST_P(DSR_Split, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicSplit, DSR_Split, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicSplit, DSR_Split, combinations);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_squeeze.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_squeeze.cpp
index ef6f8cfad1e722..8399251b60d321 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_squeeze.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_squeeze.cpp
@@ -49,7 +49,7 @@ TEST_P(DSR_Squeeze, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicSqueeze, DSR_Squeeze,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicSqueeze, DSR_Squeeze,
     ::testing::Combine(
         ::testing::Values(ngraph::element::f16, ngraph::element::i32),
         ::testing::Values(
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_strided_slice.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_strided_slice.cpp
index da7c6bd5530d21..0cb0bcb4211a7f 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_strided_slice.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_strided_slice.cpp
@@ -67,7 +67,7 @@ std::vector<DataType> precisions = {
         ngraph::element::i32
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_StridedSlice, DSR_StridedSlice,
+INSTANTIATE_TEST_SUITE_P(smoke_StridedSlice, DSR_StridedSlice,
                         ::testing::Combine(
                             ::testing::ValuesIn(testCases),
                             ::testing::ValuesIn(precisions),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_topk.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_topk.cpp
index bafc3d9212cb69..554ff1774bf1d6 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_topk.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_topk.cpp
@@ -53,7 +53,7 @@ TEST_P(DSR_TopK_Const, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicTopKConst, DSR_TopK_Const, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicTopKConst, DSR_TopK_Const, combinations);
 
 class DSR_TopK : public testing::WithParamInterface<Parameters>, public DSR_TestsCommon {
 protected:
@@ -83,6 +83,6 @@ TEST_P(DSR_TopK, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicTopKConst, DSR_TopK, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicTopKConst, DSR_TopK, combinations);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_unary_elementwise.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_unary_elementwise.cpp
index 589bbaae849128..472cf02ddef715 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_unary_elementwise.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_unary_elementwise.cpp
@@ -40,7 +40,7 @@ TEST_P(DSR_UnaryElementwise, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicUnaryElementwise, DSR_UnaryElementwise,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicUnaryElementwise, DSR_UnaryElementwise,
     ::testing::Combine(
         ::testing::Values(ngraph::element::f16, ngraph::element::f32),
         ::testing::Values(DataShapeWithUpperBound{ngraph::Shape{8, 800}, ngraph::Shape{10, 1000}}),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_unsqueeze.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_unsqueeze.cpp
index 998e7f3878c97e..014c61fbb93f03 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_unsqueeze.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_unsqueeze.cpp
@@ -44,7 +44,7 @@ TEST_P(DSR_Unsqueeze, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicUnsqueeze, DSR_Unsqueeze,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicUnsqueeze, DSR_Unsqueeze,
     ::testing::Combine(
         ::testing::Values(ngraph::element::f16, ngraph::element::i32),
         ::testing::Values(
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_variadic_split.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_variadic_split.cpp
index aead15031984f8..77bb741be24921 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_variadic_split.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/dsr_variadic_split.cpp
@@ -57,6 +57,6 @@ TEST_P(DSR_VariadicSplit, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicGatherData, DSR_VariadicSplit, combinations);
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicGatherData, DSR_VariadicSplit, combinations);
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/gather_gather_elements.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/gather_gather_elements.cpp
index e8abae995bb344..09630d00540698 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/gather_gather_elements.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/gather_gather_elements.cpp
@@ -131,6 +131,6 @@ TEST_P(Gather_GatherElements, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_Gather_GatherElements, Gather_GatherElements, testing::Values(CommonTestUtils::DEVICE_MYRIAD));
+INSTANTIATE_TEST_SUITE_P(smoke_Gather_GatherElements, Gather_GatherElements, testing::Values(CommonTestUtils::DEVICE_MYRIAD));
 
 } // namespace
\ No newline at end of file
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nms_nonzero.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nms_nonzero.cpp
index c71eeaceaaa1d2..344e3acb3ea045 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nms_nonzero.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nms_nonzero.cpp
@@ -61,7 +61,7 @@ TEST_P(NMS_NonZero, CompareWithReference) {
 }
 
 // #-30919
-INSTANTIATE_TEST_CASE_P(DISABLED_DynamicNonMaxSupression, NMS_NonZero,
+INSTANTIATE_TEST_SUITE_P(DISABLED_DynamicNonMaxSupression, NMS_NonZero,
                         ::testing::Combine(
                                 ::testing::Values(
                                         ngraph::element::f16,
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nonzero_broadcast.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nonzero_broadcast.cpp
index d739c1279cea8f..70d4a2fcd57b34 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nonzero_broadcast.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nonzero_broadcast.cpp
@@ -136,7 +136,7 @@ std::vector<BroadcastInputParams> broadcastTestParams = {
         { DataShapeWithUpperBound{ {128, 256}, {} }, DataShapeWithUpperBound{ {1, 128, 256}, {3, 128, 256} }, {1, 2} },
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicBroadcast, NonZero_Broadcast,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicBroadcast, NonZero_Broadcast,
         ::testing::Combine(
             ::testing::ValuesIn(broadcastTestParams),
             ::testing::Values(ngraph::element::f16, ngraph::element::f32, ngraph::element::i32),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nonzero_transpose.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nonzero_transpose.cpp
index 9a446b4f727429..290838ec3fa622 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nonzero_transpose.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/nonzero_transpose.cpp
@@ -43,7 +43,7 @@ TEST_P(NonZero_Transpose, CompareWithReference) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_DynamicTranspose, NonZero_Transpose,
+INSTANTIATE_TEST_SUITE_P(smoke_DynamicTranspose, NonZero_Transpose,
     ::testing::Combine(
         ::testing::Values(ngraph::element::f16, ngraph::element::f32, ngraph::element::i32),
         ::testing::Values(ngraph::Shape{1, 800}),
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/parameter_result.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/parameter_result.cpp
index 57866f28ddeda4..3001e5b7e1b94f 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/parameter_result.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/parameter_result.cpp
@@ -8,7 +8,7 @@ using namespace SubgraphTestsDefinitions;
 
 namespace {
 
-INSTANTIATE_TEST_CASE_P(smoke_Check, ParameterResultSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Check, ParameterResultSubgraphTest,
                         ::testing::Values(CommonTestUtils::DEVICE_MYRIAD),
                         ParameterResultSubgraphTest::getTestCaseName);
 
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/parameter_shapeof_result.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/parameter_shapeof_result.cpp
index d47af641a3c1ea..bb484899b3f8b9 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/parameter_shapeof_result.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/parameter_shapeof_result.cpp
@@ -8,7 +8,7 @@ using namespace SubgraphTestsDefinitions;
 
 namespace {
 
-INSTANTIATE_TEST_CASE_P(smoke_Check, ParameterShapeOfResultSubgraphTest,
+INSTANTIATE_TEST_SUITE_P(smoke_Check, ParameterShapeOfResultSubgraphTest,
                         ::testing::Combine(
                             ::testing::Values(
                                 ngraph::element::f32,
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/topk_k_propagation.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/topk_k_propagation.cpp
index e5a6c520a4b75e..9c5199b0adb90b 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/topk_k_propagation.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/topk_k_propagation.cpp
@@ -77,7 +77,7 @@ const std::vector<int64_t> kVec = {0, 10, 100, 200, 500};
 TEST_P(DynamicToStaticTopKPropagationConcatBased, KPropagation) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticTopKPropagationConcatBased, ::testing::ValuesIn(kVec));
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticTopKPropagationConcatBased, ::testing::ValuesIn(kVec));
 
 class DynamicToStaticTopKPropagationConcatReshape : public DynamicToStaticTopKPropagationConcatBased {
 protected:
@@ -89,7 +89,7 @@ class DynamicToStaticTopKPropagationConcatReshape : public DynamicToStaticTopKPr
 TEST_P(DynamicToStaticTopKPropagationConcatReshape, KPropagation) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticTopKPropagationConcatReshape, ::testing::ValuesIn(kVec));
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticTopKPropagationConcatReshape, ::testing::ValuesIn(kVec));
 
 class DynamicToStaticTopKPropagationConcatSqueezeUnsqueeze : public DynamicToStaticTopKPropagationConcatBased {
 protected:
@@ -104,7 +104,7 @@ class DynamicToStaticTopKPropagationConcatSqueezeUnsqueeze : public DynamicToSta
 TEST_P(DynamicToStaticTopKPropagationConcatSqueezeUnsqueeze, KPropagation) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticTopKPropagationConcatSqueezeUnsqueeze, ::testing::ValuesIn(kVec));
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticTopKPropagationConcatSqueezeUnsqueeze, ::testing::ValuesIn(kVec));
 
 class DynamicToStaticTopKPropagationConcatConvert : public DynamicToStaticTopKPropagationConcatBased {
 protected:
@@ -117,7 +117,7 @@ class DynamicToStaticTopKPropagationConcatConvert : public DynamicToStaticTopKPr
 TEST_P(DynamicToStaticTopKPropagationConcatConvert, KPropagation) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticTopKPropagationConcatConvert, ::testing::ValuesIn(kVec));
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticTopKPropagationConcatConvert, ::testing::ValuesIn(kVec));
 
 class DynamicToStaticTopKPropagationShapeOfBased : public DynamicToStaticTopKPropagationBase {
 public:
@@ -163,7 +163,7 @@ class DynamicToStaticTopKPropagationShapeOfGather : public DynamicToStaticTopKPr
 TEST_P(DynamicToStaticTopKPropagationShapeOfGather, KPropagation) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, DynamicToStaticTopKPropagationShapeOfGather, ::testing::ValuesIn(kVec));
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, DynamicToStaticTopKPropagationShapeOfGather, ::testing::ValuesIn(kVec));
 
 class KPropagationAfterShapeOfElimination : public DynamicToStaticTopKPropagationShapeOfBased {
     void SetUp() override {
@@ -209,6 +209,6 @@ class KPropagationAfterShapeOfElimination : public DynamicToStaticTopKPropagatio
 TEST_P(KPropagationAfterShapeOfElimination, KPropagation) {
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, KPropagationAfterShapeOfElimination, ::testing::ValuesIn(kVec));
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, KPropagationAfterShapeOfElimination, ::testing::ValuesIn(kVec));
 
 }  // namespace
diff --git a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/unsqueeze_gather.cpp b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/unsqueeze_gather.cpp
index c3d3e5ba9e734b..65e1d0def46578 100644
--- a/inference-engine/tests/functional/plugin/myriad/subgraph_tests/unsqueeze_gather.cpp
+++ b/inference-engine/tests/functional/plugin/myriad/subgraph_tests/unsqueeze_gather.cpp
@@ -48,7 +48,7 @@ TEST_P(UnsqueezeGather, CompareWithRefs) {
     Run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_NGraph, UnsqueezeGather, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(smoke_NGraph, UnsqueezeGather, testing::Combine(
         testing::Values(
                 ngraph::element::f16,
                 ngraph::element::f32),
diff --git a/inference-engine/tests/functional/plugin/shared/src/behavior/caching_tests.cpp b/inference-engine/tests/functional/plugin/shared/src/behavior/caching_tests.cpp
index 0f8f97c653451b..167afc80b6e17e 100644
--- a/inference-engine/tests/functional/plugin/shared/src/behavior/caching_tests.cpp
+++ b/inference-engine/tests/functional/plugin/shared/src/behavior/caching_tests.cpp
@@ -139,7 +139,7 @@ void LoadNetworkCacheTestBase::SetUp() {
     try {
         function = fGen(m_precision, m_batchSize);
     } catch (...) {
-        SKIP();
+        GTEST_SKIP();
     }
 
     std::stringstream ss;
@@ -168,11 +168,11 @@ void LoadNetworkCacheTestBase::Run() {
     };
     if (!function) {
         GTEST_COUT << "Can't create function " << m_functionName << " with precision " << m_precision.get_type_name() << std::endl;
-        SKIP();
+        GTEST_SKIP();
     }
     if (!importExportSupported(*core)) {
         GTEST_COUT << "Plugin doesn't support import and export - skipping test" << std::endl;
-        SKIP();
+        GTEST_SKIP();
     }
     cnnNetwork = CNNNetwork{function};
     ConfigureNetwork();
@@ -183,10 +183,10 @@ void LoadNetworkCacheTestBase::Run() {
     } catch (const Exception &ex) {
         GTEST_COUT << "Can't loadNetwork without cache for " << m_functionName << " with precision " << m_precision.get_type_name() << std::endl;
         GTEST_COUT << "Exception [" << ex.what() << "]" << std::endl;
-        SKIP();
+        GTEST_SKIP();
     } catch (...) {
         GTEST_COUT << "Can't loadNetwork without cache for " << m_functionName << " with precision " << m_precision.get_type_name() << std::endl;
-        SKIP(); // skip caching test if such network is not supported by device at all
+        GTEST_SKIP(); // skip caching test if such network is not supported by device at all
     }
     auto originalOutputs = GetOutputs();
 
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/non_max_suppression.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/non_max_suppression.hpp
index d608a7e7526974..67b0f5951782af 100644
--- a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/non_max_suppression.hpp
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/non_max_suppression.hpp
@@ -10,6 +10,16 @@
 #include "shared_test_classes/base/layer_test_utils.hpp"
 #include "ngraph_functions/builders.hpp"
 
+namespace testing {
+namespace internal {
+
+template <> inline void
+PrintTo(const ::ngraph::op::v5::NonMaxSuppression::BoxEncodingType& value,
+    ::std::ostream* os) { }
+
+}
+}
+
 namespace LayerTestsDefinitions {
 
 using InputShapeParams = std::tuple<size_t,  // Number of batches
diff --git a/inference-engine/tests/ie_test_utils/CMakeLists.txt b/inference-engine/tests/ie_test_utils/CMakeLists.txt
index cb82ae840199b3..6b9543b9f7beba 100644
--- a/inference-engine/tests/ie_test_utils/CMakeLists.txt
+++ b/inference-engine/tests/ie_test_utils/CMakeLists.txt
@@ -2,11 +2,11 @@
 # SPDX-License-Identifier: Apache-2.0
 #
 
-if (ENABLE_TESTS)
+if(ENABLE_TESTS)
     add_subdirectory(common_test_utils)
     add_subdirectory(unit_test_utils)
 endif()
 
-if (ENABLE_FUNCTIONAL_TESTS)
+if(ENABLE_FUNCTIONAL_TESTS)
     add_subdirectory(functional_test_utils)
 endif()
diff --git a/inference-engine/tests/ie_test_utils/common_test_utils/CMakeLists.txt b/inference-engine/tests/ie_test_utils/common_test_utils/CMakeLists.txt
index 5e8643dc5db0f2..edff009d5e474a 100644
--- a/inference-engine/tests/ie_test_utils/common_test_utils/CMakeLists.txt
+++ b/inference-engine/tests/ie_test_utils/common_test_utils/CMakeLists.txt
@@ -2,63 +2,15 @@
 # SPDX-License-Identifier: Apache-2.0
 #
 
-#####################################################################################################
-#                                     SETUP GOOGLE TESTS                                            #
-#####################################################################################################
-set(gtest_force_shared_crt ON CACHE BOOL "disable static CRT for google test")
-function(add_gtest_libraries)
-    if (UNIX)
-        ie_add_compiler_flags(-Wno-undef)
-        if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
-            ie_add_compiler_flags(-Wno-deprecated-copy)
-        endif()
-    endif ()
-
-    set(BUILD_SHARED_LIBS OFF)
-    add_subdirectory(gtest EXCLUDE_FROM_ALL)
-
-    get_target_property(gtest_include_dirs gtest INTERFACE_INCLUDE_DIRECTORIES)
-    set_target_properties(gtest PROPERTIES INTERFACE_SYSTEM_INCLUDE_DIRECTORIES "${gtest_include_dirs}")
-
-    get_target_property(gmock_include_dirs gtest INTERFACE_INCLUDE_DIRECTORIES)
-    set_target_properties(gmock PROPERTIES INTERFACE_SYSTEM_INCLUDE_DIRECTORIES "${gmock_include_dirs};${gmock_SOURCE_DIR}/include")
+function(add_common_utils ADD_TARGET_NAME)
+    list(APPEND EXPORT_DEPENDENCIES gtest gtest_main)
 
-    # If we have specified /Z7 option, remove -Zi option which comes from gtest
-    if (WIN32)
-        set(gtest_targets "gtest;gtest_main;gmock;gmock_main")
-        foreach(target_name ${gtest_targets})
-            if(TARGET "${target_name}")
-                get_target_property(_target_cxx_flags ${target_name} COMPILE_FLAGS)
-                if(_target_cxx_flags)
-                    if(CMAKE_CXX_FLAGS_DEBUG MATCHES ".+/Z7.+" OR CMAKE_CXX_FLAGS_RELWITHDEBINFO MATCHES ".+/Z7.+")
-                        string(REPLACE "-Zi" " " _target_cxx_flags ${_target_cxx_flags})
-                        message(STATUS "Removing -Zi flag from target " ${target_name})
-                        set_target_properties(${target_name} PROPERTIES COMPILE_FLAGS "${_target_cxx_flags}")
-                    endif()
-                endif()
-            endif()
-        endforeach()
+    if(MSVC)
+        set(PUGI pugixml_mt)
+    else()
+        set(PUGI pugixml)
     endif()
-    set_target_properties(gtest gtest_main gmock gmock_main
-                          PROPERTIES FOLDER thirdparty)
-endfunction()
-
-add_gtest_libraries()
 
-if (MSVC)
-    set(PUGI pugixml_mt)
-else ()
-    set(PUGI pugixml)
-endif ()
-
-list(APPEND EXPORT_DEPENDENCIES
-        gtest
-        gtest_main
-        )
-
-set(TARGET_NAME commonTestUtils)
-
-function(add_common_utils ADD_TARGET_NAME)
     # create target
     addIeTarget(
             NAME ${ADD_TARGET_NAME}
@@ -121,5 +73,7 @@ function(add_common_utils ADD_TARGET_NAME)
     target_compile_definitions(${ADD_TARGET_NAME} PUBLIC ${ARGN})
 endfunction()
 
+set(TARGET_NAME commonTestUtils)
+
 add_common_utils(${TARGET_NAME})
 add_common_utils(${TARGET_NAME}_s USE_STATIC_IE)
diff --git a/inference-engine/tests/ie_test_utils/common_test_utils/common_utils.cpp b/inference-engine/tests/ie_test_utils/common_test_utils/common_utils.cpp
index 738191890cfab9..3a34ffc73984f2 100644
--- a/inference-engine/tests/ie_test_utils/common_test_utils/common_utils.cpp
+++ b/inference-engine/tests/ie_test_utils/common_test_utils/common_utils.cpp
@@ -5,6 +5,11 @@
 #include "common_test_utils/common_utils.hpp"
 #include <legacy/details/ie_cnn_network_iterator.hpp>
 
+::std::ostream& ngraph::operator << (::std::ostream & os, const Function&) {
+    throw std::runtime_error("should not be called");
+    return os;
+}
+
 namespace CommonTestUtils {
 
 IE_SUPPRESS_DEPRECATED_START
diff --git a/inference-engine/tests/ie_test_utils/common_test_utils/common_utils.hpp b/inference-engine/tests/ie_test_utils/common_test_utils/common_utils.hpp
index dc6aaf2e0518c5..39ea049ba1a612 100644
--- a/inference-engine/tests/ie_test_utils/common_test_utils/common_utils.hpp
+++ b/inference-engine/tests/ie_test_utils/common_test_utils/common_utils.hpp
@@ -15,6 +15,11 @@
 #include <memory>
 
 #include <cpp/ie_cnn_network.h>
+#include <ngraph/function.hpp>
+
+namespace ngraph {
+::std::ostream& operator << (::std::ostream &, const Function&);
+}
 
 namespace InferenceEngine {
 class CNNLayer;
diff --git a/inference-engine/tests/ie_test_utils/common_test_utils/gtest b/inference-engine/tests/ie_test_utils/common_test_utils/gtest
deleted file mode 160000
index 9bd163b993459b..00000000000000
--- a/inference-engine/tests/ie_test_utils/common_test_utils/gtest
+++ /dev/null
@@ -1 +0,0 @@
-Subproject commit 9bd163b993459b2ca6ba2dc508577bbc8774c851
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mock.cpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mock.cpp
index 980044f8d0265d..9cc7309c1c9666 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mock.cpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mock.cpp
@@ -14,6 +14,7 @@
 #include "unit_test_utils/mocks/cpp_interfaces/impl/mock_executable_thread_safe_default.hpp"
 #include "unit_test_utils/mocks/cpp_interfaces/impl/mock_inference_plugin_internal.hpp"
 
+#include "unit_test_utils/mocks/cpp_interfaces/interface/mock_icore.hpp"
 #include "unit_test_utils/mocks/cpp_interfaces/interface/mock_iexecutable_network_internal.hpp"
 #include "unit_test_utils/mocks/cpp_interfaces/interface/mock_iinfer_request_internal.hpp"
 #include "unit_test_utils/mocks/cpp_interfaces/interface/mock_ivariable_state_internal.hpp"
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_icore.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_icore.hpp
index f84f4086e34e2f..b061c4c3bdc090 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_icore.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_icore.hpp
@@ -9,10 +9,10 @@
 
 class MockICore : public InferenceEngine::ICore {
 public:
-    MOCK_QUALIFIED_METHOD0(GetTaskExecutor, const, std::shared_ptr<InferenceEngine::ITaskExecutor>());
+    MOCK_CONST_METHOD0(GetTaskExecutor, std::shared_ptr<InferenceEngine::ITaskExecutor>());
 
-    MOCK_QUALIFIED_METHOD2(ReadNetwork, const, InferenceEngine::CNNNetwork(const std::string&, const InferenceEngine::Blob::CPtr&));
-    MOCK_QUALIFIED_METHOD2(ReadNetwork, const, InferenceEngine::CNNNetwork(const std::string&, const std::string&));
+    MOCK_CONST_METHOD2(ReadNetwork, InferenceEngine::CNNNetwork(const std::string&, const InferenceEngine::Blob::CPtr&));
+    MOCK_CONST_METHOD2(ReadNetwork, InferenceEngine::CNNNetwork(const std::string&, const std::string&));
 
     MOCK_METHOD3(LoadNetwork, InferenceEngine::SoExecutableNetworkInternal(
         const InferenceEngine::CNNNetwork&, const std::string&, const std::map<std::string, std::string>&));
@@ -26,12 +26,12 @@ class MockICore : public InferenceEngine::ICore {
     MOCK_METHOD3(ImportNetwork, InferenceEngine::SoExecutableNetworkInternal(
         std::istream&, const InferenceEngine::RemoteContext::Ptr&, const std::map<std::string, std::string>&));
 
-    MOCK_QUALIFIED_METHOD3(QueryNetwork, const, InferenceEngine::QueryNetworkResult(
+    MOCK_CONST_METHOD3(QueryNetwork, InferenceEngine::QueryNetworkResult(
         const InferenceEngine::CNNNetwork&, const std::string&, const std::map<std::string, std::string>&));
 
-    MOCK_QUALIFIED_METHOD2(GetMetric, const, InferenceEngine::Parameter(const std::string&, const std::string&));
-    MOCK_QUALIFIED_METHOD0(GetAvailableDevices, const, std::vector<std::string>());
-    MOCK_QUALIFIED_METHOD1(DeviceSupportsImportExport, const, bool(const std::string&)); // NOLINT not a cast to bool
+    MOCK_CONST_METHOD2(GetMetric, InferenceEngine::Parameter(const std::string&, const std::string&));
+    MOCK_CONST_METHOD0(GetAvailableDevices, std::vector<std::string>());
+    MOCK_CONST_METHOD1(DeviceSupportsImportExport, bool(const std::string&)); // NOLINT not a cast to bool
 
     ~MockICore() = default;
 };
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_iinference_plugin.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_iinference_plugin.hpp
index 083a529e121493..f6bc2a47953c30 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_iinference_plugin.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/cpp_interfaces/interface/mock_iinference_plugin.hpp
@@ -7,8 +7,8 @@
 #include <map>
 #include <string>
 
-#include "cpp_interfaces/interface/ie_iplugin_internal.hpp"
 #include <gmock/gmock.h>
+#include "cpp_interfaces/interface/ie_iplugin_internal.hpp"
 
 class MockIInferencePlugin : public InferenceEngine::IInferencePlugin {
 public:
@@ -21,13 +21,13 @@ class MockIInferencePlugin : public InferenceEngine::IInferencePlugin {
                 const std::string&, const std::map<std::string, std::string>&));
     MOCK_METHOD1(SetConfig, void(const std::map<std::string, std::string> &));
 
-    MOCK_QUALIFIED_METHOD1(SetName, noexcept, void(const std::string&));
-    MOCK_QUALIFIED_METHOD0(GetName, const noexcept, std::string(void));
-    MOCK_QUALIFIED_METHOD1(SetCore, noexcept, void(InferenceEngine::ICore*));
-    MOCK_QUALIFIED_METHOD0(GetCore, const noexcept, InferenceEngine::ICore *(void));
-    MOCK_QUALIFIED_METHOD2(GetConfig, const, InferenceEngine::Parameter(
+    MOCK_METHOD(void, SetName, (const std::string&), (noexcept));
+    MOCK_METHOD(std::string, GetName, (), (const, noexcept));
+    MOCK_METHOD(void, SetCore, (InferenceEngine::ICore*), (noexcept));
+    MOCK_METHOD(InferenceEngine::ICore *, GetCore, (), (const, noexcept));
+    MOCK_CONST_METHOD2(GetConfig, InferenceEngine::Parameter(
                 const std::string&, const std::map<std::string, InferenceEngine::Parameter>&));
-    MOCK_QUALIFIED_METHOD2(GetMetric, const, InferenceEngine::Parameter(
+    MOCK_CONST_METHOD2(GetMetric, InferenceEngine::Parameter(
                 const std::string&, const std::map<std::string, InferenceEngine::Parameter>&));
     MOCK_METHOD1(CreateContext,
                 InferenceEngine::RemoteContext::Ptr(const InferenceEngine::ParamMap&));
@@ -40,7 +40,7 @@ class MockIInferencePlugin : public InferenceEngine::IInferencePlugin {
     MOCK_METHOD3(ImportNetwork, std::shared_ptr<InferenceEngine::IExecutableNetworkInternal>(
                 std::istream&, const InferenceEngine::RemoteContext::Ptr&,
                 const std::map<std::string, std::string>&));
-    MOCK_QUALIFIED_METHOD2(QueryNetwork, const,
-                           InferenceEngine::QueryNetworkResult(const InferenceEngine::CNNNetwork&,
-                                                               const std::map<std::string, std::string>&));
+    MOCK_CONST_METHOD2(QueryNetwork,
+                       InferenceEngine::QueryNetworkResult(const InferenceEngine::CNNNetwork&,
+                                                           const std::map<std::string, std::string>&));
 };
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_allocator.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_allocator.hpp
index 47b77d999e0f8c..83af019e4343ab 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_allocator.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_allocator.hpp
@@ -14,8 +14,8 @@
 
 class MockAllocator : public InferenceEngine::IAllocator {
 public:
-    MOCK_QUALIFIED_METHOD2(lock, noexcept, void*(void*, InferenceEngine::LockOp));
-    MOCK_QUALIFIED_METHOD1(unlock, noexcept, void(void *));
-    MOCK_QUALIFIED_METHOD1(alloc, noexcept, void*(size_t));
-    MOCK_QUALIFIED_METHOD1(free, noexcept, bool(void*));  // NOLINT(readability/casting)
+    MOCK_METHOD(void*, lock, (void*, InferenceEngine::LockOp), (noexcept));
+    MOCK_METHOD(void, unlock, (void *), (noexcept));
+    MOCK_METHOD(void*, alloc, (size_t), (noexcept));
+    MOCK_METHOD(bool, free, (void*), (noexcept));  // NOLINT(readability/casting)
 };
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_icnn_network.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_icnn_network.hpp
index 618122505a4409..a3f7d337dd0bba 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_icnn_network.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_icnn_network.hpp
@@ -27,23 +27,25 @@ class CNNLayer;
  */
 class MockICNNNetwork final : public InferenceEngine::ICNNNetwork {
  public:
-    MOCK_QUALIFIED_METHOD0(getFunction, const noexcept, std::shared_ptr<const ngraph::Function> ());
-    MOCK_QUALIFIED_METHOD0(getFunction, noexcept, std::shared_ptr<ngraph::Function>());
-    MOCK_QUALIFIED_METHOD1(getOutputsInfo, const noexcept, void(InferenceEngine::OutputsDataMap& out));
-    MOCK_QUALIFIED_METHOD1(getInputsInfo, const noexcept, void(InferenceEngine::InputsDataMap &inputs));
-    MOCK_QUALIFIED_METHOD1(getInput, const noexcept, InferenceEngine::InputInfo::Ptr(const std::string &inputName));
-    MOCK_QUALIFIED_METHOD0(layerCount, const noexcept, size_t());
-    MOCK_QUALIFIED_METHOD0(getName, const noexcept, const std::string&());
-    MOCK_QUALIFIED_METHOD3(addOutput, noexcept, InferenceEngine::StatusCode(const std::string &, size_t, InferenceEngine::ResponseDesc*));
-     MOCK_QUALIFIED_METHOD3(getLayerByName, const noexcept, InferenceEngine::StatusCode(const char* ,
-            std::shared_ptr<InferenceEngine::CNNLayer>&,
-            InferenceEngine::ResponseDesc*));
-    MOCK_QUALIFIED_METHOD2(setBatchSize, noexcept, InferenceEngine::StatusCode(const size_t size, InferenceEngine::ResponseDesc*));
-    MOCK_QUALIFIED_METHOD0(getBatchSize, const noexcept, size_t());
-    MOCK_QUALIFIED_METHOD1(getInputShapes, const noexcept, void(InferenceEngine::ICNNNetwork::InputShapes&));
-    MOCK_QUALIFIED_METHOD2(reshape, noexcept, InferenceEngine::StatusCode(const InferenceEngine::ICNNNetwork::InputShapes &, InferenceEngine::ResponseDesc *));
-    MOCK_QUALIFIED_METHOD3(serialize, const noexcept, InferenceEngine::StatusCode(
-            const std::string &,
-            const std::string &,
-            InferenceEngine::ResponseDesc*));
+    MOCK_METHOD(std::shared_ptr<const ngraph::Function>, getFunction, (), (const, noexcept));
+    MOCK_METHOD(std::shared_ptr<ngraph::Function>, getFunction, (), (noexcept));
+    MOCK_METHOD(void, getOutputsInfo, (InferenceEngine::OutputsDataMap& out), (const, noexcept));
+    MOCK_METHOD(void, getInputsInfo, (InferenceEngine::InputsDataMap &inputs), (const, noexcept));
+    MOCK_METHOD(InferenceEngine::InputInfo::Ptr, getInput, (const std::string &inputName), (const, noexcept));
+    MOCK_METHOD(size_t, layerCount, (), (const, noexcept));
+    MOCK_METHOD(const std::string&, getName, (), (const, noexcept));
+    MOCK_METHOD(InferenceEngine::StatusCode, addOutput,
+        (const std::string &, size_t, InferenceEngine::ResponseDesc*), (noexcept));
+    MOCK_METHOD(InferenceEngine::StatusCode, getLayerByName,
+        (const char* , std::shared_ptr<InferenceEngine::CNNLayer>&, InferenceEngine::ResponseDesc*),
+        (const, noexcept));
+    MOCK_METHOD(InferenceEngine::StatusCode, setBatchSize,
+        (const size_t size, InferenceEngine::ResponseDesc*), (noexcept));
+    MOCK_METHOD(size_t, getBatchSize, (), (const, noexcept));
+    MOCK_METHOD(void, getInputShapes, (InferenceEngine::ICNNNetwork::InputShapes&), (const, noexcept));
+    MOCK_METHOD(InferenceEngine::StatusCode, reshape,
+        (const InferenceEngine::ICNNNetwork::InputShapes &, InferenceEngine::ResponseDesc *),
+        (const, noexcept));
+    MOCK_METHOD(InferenceEngine::StatusCode, serialize,
+        (const std::string &, const std::string &, InferenceEngine::ResponseDesc*), (const, noexcept));
 };
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iexecutable_network.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iexecutable_network.hpp
index a5187953206ddd..5a6779f12ab521 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iexecutable_network.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iexecutable_network.hpp
@@ -22,16 +22,20 @@ IE_SUPPRESS_DEPRECATED_START
 
 class MockIExecutableNetwork : public IExecutableNetwork {
 public:
-    MOCK_QUALIFIED_METHOD2(GetOutputsInfo, const noexcept, StatusCode(ConstOutputsDataMap  &, ResponseDesc *));
-    MOCK_QUALIFIED_METHOD2(GetInputsInfo, const noexcept, StatusCode(ConstInputsDataMap &, ResponseDesc *));
-    MOCK_QUALIFIED_METHOD2(CreateInferRequest, noexcept, StatusCode(IInferRequest::Ptr &, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD2(Export, noexcept, StatusCode(const std::string &, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD2(Export, noexcept, StatusCode(std::ostream &, ResponseDesc *));
-    MOCK_QUALIFIED_METHOD2(GetExecGraphInfo, noexcept, StatusCode(ICNNNetwork::Ptr &, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD2(SetConfig, noexcept, StatusCode(const std::map<std::string, Parameter> &config, ResponseDesc *resp));
-    MOCK_QUALIFIED_METHOD3(GetConfig, const noexcept, StatusCode(const std::string &name, Parameter &result, ResponseDesc *resp));
-    MOCK_QUALIFIED_METHOD3(GetMetric, const noexcept, StatusCode(const std::string &name, Parameter &result, ResponseDesc *resp));
-    MOCK_QUALIFIED_METHOD2(GetContext, const noexcept, StatusCode(RemoteContext::Ptr &pContext, ResponseDesc *resp));
+    MOCK_METHOD(StatusCode, GetOutputsInfo, (ConstOutputsDataMap  &, ResponseDesc *), (const, noexcept));
+    MOCK_METHOD(StatusCode, GetInputsInfo, (ConstInputsDataMap &, ResponseDesc *), (const, noexcept));
+    MOCK_METHOD(StatusCode, CreateInferRequest, (IInferRequest::Ptr &, ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, Export, (const std::string &, ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, Export, (std::ostream &, ResponseDesc *), (noexcept));
+    MOCK_METHOD(StatusCode, GetExecGraphInfo, (ICNNNetwork::Ptr &, ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, SetConfig,
+        ((const std::map<std::string, Parameter> &config), ResponseDesc *resp), (noexcept));
+    MOCK_METHOD(StatusCode, GetConfig,
+        (const std::string &name, Parameter &result, ResponseDesc *resp), (const, noexcept));
+    MOCK_METHOD(StatusCode, GetMetric,
+        (const std::string &name, Parameter &result, ResponseDesc *resp), (const, noexcept));
+    MOCK_METHOD(StatusCode, GetContext,
+        (RemoteContext::Ptr &pContext, ResponseDesc *resp), (const, noexcept));
 };
 
 IE_SUPPRESS_DEPRECATED_END
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iinfer_request.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iinfer_request.hpp
index bac5fc7773e024..32060d0138a804 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iinfer_request.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_iinfer_request.hpp
@@ -22,20 +22,22 @@ IE_SUPPRESS_DEPRECATED_START
 
 class MockIInferRequest : public IInferRequest {
 public:
-    MOCK_QUALIFIED_METHOD1(StartAsync, noexcept, StatusCode(ResponseDesc*));
-    MOCK_QUALIFIED_METHOD2(Wait, noexcept, StatusCode(int64_t millis_timeout, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD2(GetUserData, noexcept, StatusCode(void**, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD2(SetUserData, noexcept, StatusCode(void*, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD1(SetCompletionCallback, noexcept, StatusCode(IInferRequest::CompletionCallback));
-    MOCK_QUALIFIED_METHOD1(Infer, noexcept, StatusCode(ResponseDesc*));
-    MOCK_QUALIFIED_METHOD2(GetPerformanceCounts, const noexcept,
-                           StatusCode(std::map<std::string, InferenceEngineProfileInfo> &perfMap, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD3(GetBlob, noexcept, StatusCode(const char*, Blob::Ptr&, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD3(GetPreProcess, const noexcept, StatusCode(const char*, const PreProcessInfo**, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD3(SetBlob, noexcept, StatusCode(const char*, const Blob::Ptr&, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD4(SetBlob, noexcept, StatusCode(const char*, const Blob::Ptr&, const PreProcessInfo&, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD2(SetBatch, noexcept, StatusCode(int batch, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD1(Cancel, noexcept, InferenceEngine::StatusCode(ResponseDesc*));
+    MOCK_METHOD(StatusCode, StartAsync, (ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, Wait, (int64_t, ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, GetUserData, (void**, ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, SetUserData, (void*, ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, SetCompletionCallback, (IInferRequest::CompletionCallback), (noexcept));
+    MOCK_METHOD(StatusCode, Infer, (ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, GetPerformanceCounts,
+        ((std::map<std::string, InferenceEngineProfileInfo> &), ResponseDesc*), (const, noexcept));
+    MOCK_METHOD(StatusCode, GetBlob, (const char*, Blob::Ptr&, ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, GetPreProcess,
+        (const char*, const PreProcessInfo**, ResponseDesc*), (const, noexcept));
+    MOCK_METHOD(StatusCode, SetBlob, (const char*, const Blob::Ptr&, ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, SetBlob,
+        (const char*, const Blob::Ptr&, const PreProcessInfo&, ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, SetBatch, (int batch, ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, Cancel, (ResponseDesc*), (noexcept));
 };
 
 IE_SUPPRESS_DEPRECATED_END
diff --git a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_not_empty_icnn_network.hpp b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_not_empty_icnn_network.hpp
index 814336b887b872..f9ffeab2f2ddad 100644
--- a/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_not_empty_icnn_network.hpp
+++ b/inference-engine/tests/ie_test_utils/unit_test_utils/mocks/mock_not_empty_icnn_network.hpp
@@ -34,14 +34,15 @@ class MockNotEmptyICNNNetwork final : public ICNNNetwork {
     std::shared_ptr<const ngraph::Function> getFunction() const noexcept override {
         return nullptr;
     }
-    MOCK_QUALIFIED_METHOD1(getInput, const noexcept, InputInfo::Ptr(const std::string &inputName));
-    MOCK_QUALIFIED_METHOD0(layerCount, const noexcept, size_t());
-    MOCK_QUALIFIED_METHOD3(addOutput, noexcept, StatusCode(const std::string &, size_t , ResponseDesc*));
-    MOCK_QUALIFIED_METHOD2(setBatchSize, noexcept, StatusCode(const size_t size, ResponseDesc*));
-    MOCK_QUALIFIED_METHOD0(getBatchSize, const noexcept, size_t());
-    MOCK_QUALIFIED_METHOD1(getInputShapes, const noexcept, void(ICNNNetwork::InputShapes &));
-    MOCK_QUALIFIED_METHOD2(reshape, noexcept, StatusCode(const ICNNNetwork::InputShapes &, ResponseDesc *));
-    MOCK_QUALIFIED_METHOD3(serialize, const noexcept, StatusCode(const std::string &, const std::string &, InferenceEngine::ResponseDesc*));
+    MOCK_METHOD(InputInfo::Ptr, getInput, (const std::string &inputName), (const, noexcept));
+    MOCK_METHOD(size_t, layerCount, (), (const, noexcept));
+    MOCK_METHOD(StatusCode, addOutput, (const std::string &, size_t , ResponseDesc*), (noexcept));
+    MOCK_METHOD(StatusCode, setBatchSize, (const size_t size, ResponseDesc*), (noexcept));
+    MOCK_METHOD(size_t, getBatchSize, (), (const, noexcept));
+    MOCK_METHOD(void, getInputShapes, (ICNNNetwork::InputShapes &), (const, noexcept));
+    MOCK_METHOD(StatusCode, reshape, (const ICNNNetwork::InputShapes &, ResponseDesc *), (noexcept));
+    MOCK_METHOD(StatusCode, serialize,
+        (const std::string &, const std::string &, InferenceEngine::ResponseDesc*), (const, noexcept));
 };
 
 IE_SUPPRESS_DEPRECATED_END
diff --git a/inference-engine/tests/unit/cpu/mkldnn_memory_desc_test.cpp b/inference-engine/tests/unit/cpu/mkldnn_memory_desc_test.cpp
index ee7684cca24200..fd828df6dd9a92 100644
--- a/inference-engine/tests/unit/cpu/mkldnn_memory_desc_test.cpp
+++ b/inference-engine/tests/unit/cpu/mkldnn_memory_desc_test.cpp
@@ -131,23 +131,23 @@ TEST(MemDescTest, isTailCCheck) {
 }
 
 TEST(MemDescTest, constructWithPlainFormat) {
-    SKIP();
+    GTEST_SKIP();
 }
 
 TEST(MemDescTest, CheckScalar) {
-    SKIP();
+    GTEST_SKIP();
 }
 
 TEST(MemDescTest, UpperBound) {
-    SKIP();
+    GTEST_SKIP();
 }
 
 TEST(MemDescTest, BlockedConversion) {
-    SKIP();
+    GTEST_SKIP();
 }
 
 TEST(MemDescTest, ComaptibleWithFormat) {
-    SKIP();
+    GTEST_SKIP();
 }
 
 TEST(isSameMethodTest, CheckTensorWithSameStrides) {
diff --git a/inference-engine/tests/unit/cpu/mkldnn_memory_test.cpp b/inference-engine/tests/unit/cpu/mkldnn_memory_test.cpp
index 41e2d42cf7bada..7b46e154617e9c 100644
--- a/inference-engine/tests/unit/cpu/mkldnn_memory_test.cpp
+++ b/inference-engine/tests/unit/cpu/mkldnn_memory_test.cpp
@@ -11,9 +11,9 @@ using namespace MKLDNNPlugin;
 using namespace InferenceEngine;
 
 TEST(MemoryTest, SedDataCheck) {
-    SKIP();
+    GTEST_SKIP();
 }
 
 TEST(MemoryTest, SedDataWithAutoPadCheck) {
-    SKIP();
+    GTEST_SKIP();
 }
diff --git a/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_infer_async_request_base_test.cpp b/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_infer_async_request_base_test.cpp
index d0cde74580cb1b..6fbb50abe3e52c 100644
--- a/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_infer_async_request_base_test.cpp
+++ b/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_infer_async_request_base_test.cpp
@@ -3,8 +3,7 @@
 //
 
 #include <gtest/gtest.h>
-#include <gmock/gmock-spec-builders.h>
-#include <gmock/gmock-generated-actions.h>
+#include <gmock/gmock.h>
 
 #include <cpp/ie_infer_request.hpp>
 #include <cpp/ie_executable_network.hpp>
diff --git a/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_memory_state_internal_test.cpp b/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_memory_state_internal_test.cpp
index 83712bf6fd5a8f..f35afd674e4f55 100644
--- a/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_memory_state_internal_test.cpp
+++ b/inference-engine/tests/unit/inference_engine/cpp_interfaces/ie_memory_state_internal_test.cpp
@@ -184,18 +184,16 @@ TEST_F(VariableStateTests, VariableStateCanPropagateSetState) {
 TEST_F(VariableStateTests, VariableStateCanPropagateGetLastState) {
     IE_SUPPRESS_DEPRECATED_START
     std::vector<IVariableStateInternal::Ptr> toReturn;
+    toReturn.push_back(mockVariableStateInternal);
 
     float data[] = {123, 124, 125};
     auto stateBlob = make_shared_blob<float>({ Precision::FP32, {3}, C }, data, sizeof(data) / sizeof(*data));
 
-
-    toReturn.push_back(mockVariableStateInternal);
-
     EXPECT_CALL(*mockExeNetworkInternal.get(), QueryState()).WillRepeatedly(Return(toReturn));
     EXPECT_CALL(*mockVariableStateInternal.get(), GetState()).WillOnce(Return(stateBlob));
 
-
     auto saver = net->QueryState().front()->GetState();
+    ASSERT_NE(saver, nullptr);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float*>()[0], 123);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float*>()[1], 124);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float*>()[2], 125);
@@ -222,6 +220,7 @@ TEST_F(VariableStateTests, VariableStateInternalCanSaveState) {
     pState->SetState(stateBlob);
     auto saver = pState->GetState();
 
+    ASSERT_NE(saver, nullptr);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float *>()[0], 123);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float *>()[1], 124);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float *>()[2], 125);
@@ -240,6 +239,7 @@ TEST_F(VariableStateTests, VariableStateInternalCanSaveStateByReference) {
     data[2] = 123;
     auto saver = pState->GetState();
 
+    ASSERT_NE(saver, nullptr);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float *>()[0], 121);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float *>()[1], 122);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float *>()[2], 123);
@@ -338,6 +338,7 @@ TEST_F(VariableStateTests, InfReqVariableStateCanPropagateGetLastState) {
     EXPECT_CALL(*mockVariableStateInternal.get(), GetState()).WillOnce(Return(stateBlob));
 
     auto saver = req->QueryState().front()->GetState();
+    ASSERT_NE(saver, nullptr);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float*>()[0], 123);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float*>()[1], 124);
     ASSERT_FLOAT_EQ(saver->cbuffer().as<const float*>()[2], 125);
diff --git a/inference-engine/tests/unit/vpu/blob_reader_tests.cpp b/inference-engine/tests/unit/vpu/blob_reader_tests.cpp
index 0092c540885287..5cb73484281d33 100644
--- a/inference-engine/tests/unit/vpu/blob_reader_tests.cpp
+++ b/inference-engine/tests/unit/vpu/blob_reader_tests.cpp
@@ -199,6 +199,6 @@ TEST_P(VPUBlobReaderOutputTests, canGetCorrectOutputNamesFromImportedNetwork) {
 
 const std::vector<size_t> inputShape = {{1, 4, 10, 10}};
 
-INSTANTIATE_TEST_CASE_P(myriadBlobReader_nightly, VPUBlobReaderHeaderTests, ::testing::Values(inputShape));
-INSTANTIATE_TEST_CASE_P(myriadBlobReader_nightly, VPUBlobReaderInputTests, ::testing::Values(inputShape));
-INSTANTIATE_TEST_CASE_P(myriadBlobReader_nightly, VPUBlobReaderOutputTests, ::testing::Values(inputShape));
+INSTANTIATE_TEST_SUITE_P(myriadBlobReader_nightly, VPUBlobReaderHeaderTests, ::testing::Values(inputShape));
+INSTANTIATE_TEST_SUITE_P(myriadBlobReader_nightly, VPUBlobReaderInputTests, ::testing::Values(inputShape));
+INSTANTIATE_TEST_SUITE_P(myriadBlobReader_nightly, VPUBlobReaderOutputTests, ::testing::Values(inputShape));
diff --git a/inference-engine/tests/unit/vpu/middleend_tests/passes_tests/annotate_memory_types.cpp b/inference-engine/tests/unit/vpu/middleend_tests/passes_tests/annotate_memory_types.cpp
index 14e73c28cc7cec..cb4bd63f44c7bf 100644
--- a/inference-engine/tests/unit/vpu/middleend_tests/passes_tests/annotate_memory_types.cpp
+++ b/inference-engine/tests/unit/vpu/middleend_tests/passes_tests/annotate_memory_types.cpp
@@ -103,7 +103,7 @@ class AnnotateMemoryTypes : public GraphTransformerTest, public testing::WithPar
 TEST_P(AnnotateMemoryTypes, SubgraphOf3Stages) {
 }
 
-INSTANTIATE_TEST_CASE_P(unit, AnnotateMemoryTypes, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(unit, AnnotateMemoryTypes, testing::Combine(
     testing::Combine(
         testing::Values(MemoryType::DDR, MemoryType::CMX),
         testing::Values(MemoryType::DDR, MemoryType::CMX),
diff --git a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/aot_behavior_tests.cpp b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/aot_behavior_tests.cpp
index 99617e16f3bf46..b67029dbedcc2e 100644
--- a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/aot_behavior_tests.cpp
+++ b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/aot_behavior_tests.cpp
@@ -202,6 +202,6 @@ const BehTestParams vpuValues[] = {
     BEH_MYRIAD,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTest, AOTBehaviorTests, ValuesIn(vpuValues), getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTest, AOTBehaviorTests, ValuesIn(vpuValues), getTestCaseName);
 
 #endif
diff --git a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_boot_tests.cpp b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_boot_tests.cpp
index e96bcbe408c19d..0998c48e25cf96 100644
--- a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_boot_tests.cpp
+++ b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_boot_tests.cpp
@@ -96,4 +96,4 @@ const BehTestParams vpuValues[] = {
         BEH_MYRIAD,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTest, MYRIADBoot, ValuesIn(vpuValues), getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTest, MYRIADBoot, ValuesIn(vpuValues), getTestCaseName);
diff --git a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_protocol_tests.cpp b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_protocol_tests.cpp
index 8f987c7c399da1..50b345866c9552 100644
--- a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_protocol_tests.cpp
+++ b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_protocol_tests.cpp
@@ -48,7 +48,7 @@ TEST_P(MyriadProtocolTests, NoErrorsMessagesWhenLoadNetworkSuccessful) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_VPUConfigProtocolTests,
+INSTANTIATE_TEST_SUITE_P(smoke_VPUConfigProtocolTests,
                         MyriadProtocolTests,
                         ::testing::ValuesIn(myriadProtocols),
                         MyriadProtocolTests::getTestCaseName);
\ No newline at end of file
diff --git a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_watchdog_tests.cpp b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_watchdog_tests.cpp
index f55dd04473594a..a3dda5c82000db 100644
--- a/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_watchdog_tests.cpp
+++ b/inference-engine/tests_deprecated/behavior/vpu/myriad_tests/vpu_watchdog_tests.cpp
@@ -261,4 +261,4 @@ const BehTestParams vpuValues[] = {
     BEH_MYRIAD,
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_BehaviorTest, MYRIADWatchdog, ValuesIn(vpuValues), getTestCaseName);
+INSTANTIATE_TEST_SUITE_P(smoke_BehaviorTest, MYRIADWatchdog, ValuesIn(vpuValues), getTestCaseName);
diff --git a/inference-engine/tests_deprecated/fluid_preproc/cpu/fluid_tests_cpu.cpp b/inference-engine/tests_deprecated/fluid_preproc/cpu/fluid_tests_cpu.cpp
index e20fa1f066d7ba..f06ba5025df1e1 100644
--- a/inference-engine/tests_deprecated/fluid_preproc/cpu/fluid_tests_cpu.cpp
+++ b/inference-engine/tests_deprecated/fluid_preproc/cpu/fluid_tests_cpu.cpp
@@ -98,57 +98,57 @@
 
 using namespace testing;
 #if defined(__arm__) || defined(__aarch64__)
-INSTANTIATE_TEST_CASE_P(ResizeTestFluid_U8, ResizeTestGAPI,
+INSTANTIATE_TEST_SUITE_P(ResizeTestFluid_U8, ResizeTestGAPI,
                         Combine(Values(CV_8UC1, CV_8UC3),
                                 Values(cv::INTER_LINEAR, cv::INTER_AREA),
                                 Values(TEST_RESIZE_PAIRS),
                                 Values(4))); // error not more than 4 unit
 
-INSTANTIATE_TEST_CASE_P(ResizeRGB8UTestFluid_U8, ResizeRGB8UTestGAPI,
+INSTANTIATE_TEST_SUITE_P(ResizeRGB8UTestFluid_U8, ResizeRGB8UTestGAPI,
                         Combine(Values(CV_8UC3, CV_8UC4),
                                 Values(cv::INTER_LINEAR),
                                 Values(TEST_RESIZE_PAIRS),
                                 Values(4))); // error not more than 4 unit
 #else
-INSTANTIATE_TEST_CASE_P(ResizeTestFluid_U8, ResizeTestGAPI,
+INSTANTIATE_TEST_SUITE_P(ResizeTestFluid_U8, ResizeTestGAPI,
                         Combine(Values(CV_8UC1, CV_8UC3),
                                 Values(cv::INTER_LINEAR, cv::INTER_AREA),
                                 Values(TEST_RESIZE_PAIRS),
                                 Values(1))); // error not more than 1 unit
 
-INSTANTIATE_TEST_CASE_P(ResizeRGB8UTestFluid_U8, ResizeRGB8UTestGAPI,
+INSTANTIATE_TEST_SUITE_P(ResizeRGB8UTestFluid_U8, ResizeRGB8UTestGAPI,
                         Combine(Values(CV_8UC3, CV_8UC4),
                                 Values(cv::INTER_LINEAR),
                                 Values(TEST_RESIZE_PAIRS),
                                 Values(1))); // error not more than 1 unit
 #endif
 
-INSTANTIATE_TEST_CASE_P(ResizeTestFluid_F32, ResizeTestGAPI,
+INSTANTIATE_TEST_SUITE_P(ResizeTestFluid_F32, ResizeTestGAPI,
                         Combine(Values(CV_32FC1, CV_32FC3),
                                 Values(cv::INTER_LINEAR, cv::INTER_AREA),
                                 Values(TEST_RESIZE_PAIRS),
                                 Values(0.015))); // accuracy like ~1.5%
 
 
-INSTANTIATE_TEST_CASE_P(SplitTestFluid, SplitTestGAPI,
+INSTANTIATE_TEST_SUITE_P(SplitTestFluid, SplitTestGAPI,
                         Combine(Values(2, 3, 4),
                                 Values(CV_8U, CV_8S, CV_16U, CV_16S, CV_16F, CV_32F, CV_32S),
                                 Values(TEST_SIZES),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(ChanToPlaneTestFluid, ChanToPlaneTestGAPI,
+INSTANTIATE_TEST_SUITE_P(ChanToPlaneTestFluid, ChanToPlaneTestGAPI,
                         Combine(Values(1, 3),
                                 Values(CV_8U, CV_32F),
                                 Values(TEST_SIZES),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(MergeTestFluid, MergeTestGAPI,
+INSTANTIATE_TEST_SUITE_P(MergeTestFluid, MergeTestGAPI,
                         Combine(Values(2, 3, 4),
                                 Values(CV_8U, CV_8S, CV_16U, CV_16S, CV_16F, CV_32F, CV_32S),
                                 Values(TEST_SIZES),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(NV12toRGBTestFluid, NV12toRGBTestGAPI,
+INSTANTIATE_TEST_SUITE_P(NV12toRGBTestFluid, NV12toRGBTestGAPI,
                         Combine(Values(cv::Size(3840, 2160),
                                        cv::Size(1920, 1080),
                                        cv::Size(1280,  720),
@@ -159,7 +159,7 @@ INSTANTIATE_TEST_CASE_P(NV12toRGBTestFluid, NV12toRGBTestGAPI,
                                        cv::Size( 320,  200)),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(I420toRGBTestFluid, I420toRGBTestGAPI,
+INSTANTIATE_TEST_SUITE_P(I420toRGBTestFluid, I420toRGBTestGAPI,
                         Combine(Values(cv::Size(3840, 2160),
                                        cv::Size(1920, 1080),
                                        cv::Size(1280,  720),
@@ -170,7 +170,7 @@ INSTANTIATE_TEST_CASE_P(I420toRGBTestFluid, I420toRGBTestGAPI,
                                        cv::Size( 320,  200)),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(ConvertDepthFluid, ConvertDepthTestGAPI,
+INSTANTIATE_TEST_SUITE_P(ConvertDepthFluid, ConvertDepthTestGAPI,
                         Combine(Values(CV_16U, CV_32F, CV_8U),
                                 Values(CV_32F, CV_16U, CV_8U),
                                 Values(cv::Size(3840, 2160),
@@ -183,21 +183,21 @@ INSTANTIATE_TEST_CASE_P(ConvertDepthFluid, ConvertDepthTestGAPI,
                                        cv::Size( 320,  200)),
                                 Values(1)));
 
-INSTANTIATE_TEST_CASE_P(DivCFluid, DivCTestGAPI,
+INSTANTIATE_TEST_SUITE_P(DivCFluid, DivCTestGAPI,
                         Combine(Values(CV_32F),
                                 Values(1),      //channels
                                 Values(TEST_SIZES),
                                 Values(cv::Scalar{0.229}),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(SubCFluid, SubCTestGAPI,
+INSTANTIATE_TEST_SUITE_P(SubCFluid, SubCTestGAPI,
                         Combine(Values(CV_32F),
                                 Values(1),      //channels
                                 Values(TEST_SIZES),
                                 Values(cv::Scalar{0.229}),
                                 Values(0.00001)));
 
-INSTANTIATE_TEST_CASE_P(ResizeRoiTestFluid, ResizeRoiTestGAPI,
+INSTANTIATE_TEST_SUITE_P(ResizeRoiTestFluid, ResizeRoiTestGAPI,
                         Combine(Values(CV_8UC1, CV_8UC3),
                                 Values(cv::INTER_LINEAR),
                                 Values(std::make_pair(cv::Size(24, 24), cv::Size(12, 12))),
@@ -207,7 +207,7 @@ INSTANTIATE_TEST_CASE_P(ResizeRoiTestFluid, ResizeRoiTestGAPI,
                                        cv::Rect{0, 9, 12, 3}),
                                 Values(1))); // error not more than 1 unit
 
-INSTANTIATE_TEST_CASE_P(ResizeRGB8URoiTestFluid, ResizeRGB8URoiTestGAPI,
+INSTANTIATE_TEST_SUITE_P(ResizeRGB8URoiTestFluid, ResizeRGB8URoiTestGAPI,
                         Combine(Values(CV_8UC3, CV_8UC4),
                                 Values(cv::INTER_LINEAR),
                                 Values(std::make_pair(cv::Size(24, 24), cv::Size(12, 12))),
@@ -220,38 +220,38 @@ INSTANTIATE_TEST_CASE_P(ResizeRGB8URoiTestFluid, ResizeRGB8URoiTestGAPI,
 //----------------------------------------------------------------------
 
 #if defined(__arm__) || defined(__aarch64__)
-INSTANTIATE_TEST_CASE_P(ResizeTestFluid_U8, ResizeTestIE,
+INSTANTIATE_TEST_SUITE_P(ResizeTestFluid_U8, ResizeTestIE,
                         Combine(Values(CV_8UC1, CV_8UC3),
                                 Values(cv::INTER_LINEAR, cv::INTER_AREA),
                                 Values(TEST_RESIZE_PAIRS),
                                 Values(4))); // error not more than 4 unit
 #else
-INSTANTIATE_TEST_CASE_P(ResizeTestFluid_U8, ResizeTestIE,
+INSTANTIATE_TEST_SUITE_P(ResizeTestFluid_U8, ResizeTestIE,
                         Combine(Values(CV_8UC1, CV_8UC3),
                                 Values(cv::INTER_LINEAR, cv::INTER_AREA),
                                 Values(TEST_RESIZE_PAIRS),
                                 Values(1))); // error not more than 1 unit
 #endif
 
-INSTANTIATE_TEST_CASE_P(ResizeTestFluid_F32, ResizeTestIE,
+INSTANTIATE_TEST_SUITE_P(ResizeTestFluid_F32, ResizeTestIE,
                         Combine(Values(CV_32FC1, CV_32FC3),
                                 Values(cv::INTER_LINEAR, cv::INTER_AREA),
                                 Values(TEST_RESIZE_PAIRS),
                                 Values(0.05))); // error within 0.05 units
 
-INSTANTIATE_TEST_CASE_P(SplitTestFluid, SplitTestIE,
+INSTANTIATE_TEST_SUITE_P(SplitTestFluid, SplitTestIE,
                         Combine(Values(CV_8UC2, CV_8UC3, CV_8UC4,
                                        CV_32FC2, CV_32FC3, CV_32FC4),
                                 Values(TEST_SIZES),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(MergeTestFluid, MergeTestIE,
+INSTANTIATE_TEST_SUITE_P(MergeTestFluid, MergeTestIE,
                         Combine(Values(CV_8UC2, CV_8UC3, CV_8UC4,
                                        CV_32FC2, CV_32FC3, CV_32FC4),
                                 Values(TEST_SIZES),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(ColorConvertFluid_3ch, ColorConvertTestIE,
+INSTANTIATE_TEST_SUITE_P(ColorConvertFluid_3ch, ColorConvertTestIE,
                         Combine(Values(CV_8U, CV_32F),
                                 Values(InferenceEngine::ColorFormat::RGB),
                                 Values(InferenceEngine::NHWC, InferenceEngine::NCHW),
@@ -259,7 +259,7 @@ INSTANTIATE_TEST_CASE_P(ColorConvertFluid_3ch, ColorConvertTestIE,
                                 Values(TEST_SIZES),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(ColorConvertFluid_4ch, ColorConvertTestIE,
+INSTANTIATE_TEST_SUITE_P(ColorConvertFluid_4ch, ColorConvertTestIE,
                         Combine(Values(CV_8U, CV_32F),
                                 Values(InferenceEngine::ColorFormat::RGBX,
                                        InferenceEngine::ColorFormat::BGRX),
@@ -268,7 +268,7 @@ INSTANTIATE_TEST_CASE_P(ColorConvertFluid_4ch, ColorConvertTestIE,
                                 Values(TEST_SIZES),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(ColorConvertYUV420Fluid, ColorConvertYUV420TestIE,
+INSTANTIATE_TEST_SUITE_P(ColorConvertYUV420Fluid, ColorConvertYUV420TestIE,
                         Combine(Values(InferenceEngine::NV12, InferenceEngine::I420),
                                 Values(InferenceEngine::NHWC, InferenceEngine::NCHW),
                                 Values(cv::Size(3840, 2160),
@@ -282,7 +282,7 @@ INSTANTIATE_TEST_CASE_P(ColorConvertYUV420Fluid, ColorConvertYUV420TestIE,
                                        cv::Size( 150,  150)),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(Reorder_HWC2CHW, ColorConvertTestIE,
+INSTANTIATE_TEST_SUITE_P(Reorder_HWC2CHW, ColorConvertTestIE,
                         Combine(Values(CV_8U, CV_32F, CV_16S, CV_16F),
                                 Values(InferenceEngine::ColorFormat::BGR),
                                 Values(InferenceEngine::NHWC),
@@ -290,7 +290,7 @@ INSTANTIATE_TEST_CASE_P(Reorder_HWC2CHW, ColorConvertTestIE,
                                 Values(TEST_SIZES),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(Reorder_CHW2HWC, ColorConvertTestIE,
+INSTANTIATE_TEST_SUITE_P(Reorder_CHW2HWC, ColorConvertTestIE,
                         Combine(Values(CV_8U, CV_32F, CV_16S, CV_16F),
                                 Values(InferenceEngine::ColorFormat::BGR),
                                 Values(InferenceEngine::NCHW),
@@ -298,7 +298,7 @@ INSTANTIATE_TEST_CASE_P(Reorder_CHW2HWC, ColorConvertTestIE,
                                 Values(TEST_SIZES),
                                 Values(0)));
 
-INSTANTIATE_TEST_CASE_P(MeanValueGAPI32F, MeanValueGAPI,
+INSTANTIATE_TEST_SUITE_P(MeanValueGAPI32F, MeanValueGAPI,
                         Combine(Values(TEST_SIZES),
                                 Values(0.00001)));
 
@@ -345,7 +345,7 @@ static const auto U8toU8 = std::make_pair(IE::Precision::U8,IE::Precision::U8);
 
 static const auto PRECISIONS = Values(U8toU8, std::make_pair(IE::Precision::FP32,IE::Precision::FP32));
 
-INSTANTIATE_TEST_CASE_P(ReorderResize_Frame, PreprocTest,
+INSTANTIATE_TEST_SUITE_P(ReorderResize_Frame, PreprocTest,
                         Combine(PRECISIONS,
                                 Values(IE::ResizeAlgorithm::RESIZE_BILINEAR), // AREA is not there yet
                                 Values(IE::ColorFormat::RAW),
@@ -354,7 +354,7 @@ INSTANTIATE_TEST_CASE_P(ReorderResize_Frame, PreprocTest,
                                 Values(std::make_pair(1, 1), std::make_pair(3, 3)),
                                 FRAME_SIZES));
 
-INSTANTIATE_TEST_CASE_P(Scale3ch_Frame, PreprocTest,
+INSTANTIATE_TEST_SUITE_P(Scale3ch_Frame, PreprocTest,
                         Combine(PRECISIONS,
                                 Values(IE::ResizeAlgorithm::RESIZE_BILINEAR), // AREA is not there yet
                                 Values(IE::ColorFormat::RAW),
@@ -363,7 +363,7 @@ INSTANTIATE_TEST_CASE_P(Scale3ch_Frame, PreprocTest,
                                 Values(std::make_pair(3, 3)),
                                 FRAME_SIZES));
 
-INSTANTIATE_TEST_CASE_P(ReorderResize_Patch, PreprocTest,
+INSTANTIATE_TEST_SUITE_P(ReorderResize_Patch, PreprocTest,
                         Combine(PRECISIONS,
                                 Values(IE::ResizeAlgorithm::RESIZE_BILINEAR), // AREA is not there yet
                                 Values(IE::ColorFormat::RAW),
@@ -372,7 +372,7 @@ INSTANTIATE_TEST_CASE_P(ReorderResize_Patch, PreprocTest,
                                 Values(std::make_pair(1, 1), std::make_pair(3, 3)),
                                 PATCH_SIZES));
 
-INSTANTIATE_TEST_CASE_P(Everything_Resize, PreprocTest,
+INSTANTIATE_TEST_SUITE_P(Everything_Resize, PreprocTest,
                         Combine(PRECISIONS,
                                 Values(IE::ResizeAlgorithm::RESIZE_BILINEAR, IE::ResizeAlgorithm::RESIZE_AREA),
                                 Values(IE::ColorFormat::RAW),
@@ -384,7 +384,7 @@ INSTANTIATE_TEST_CASE_P(Everything_Resize, PreprocTest,
                                        std::make_pair(4, 4)),
                                 Values(TEST_SIZES_PREPROC)));
 
-INSTANTIATE_TEST_CASE_P(ColorFormats_3ch, PreprocTest,
+INSTANTIATE_TEST_SUITE_P(ColorFormats_3ch, PreprocTest,
                         Combine(PRECISIONS,
                                 Values(IE::ResizeAlgorithm::RESIZE_BILINEAR, IE::ResizeAlgorithm::RESIZE_AREA),
                                 Values(IE::ColorFormat::RGB),
@@ -393,7 +393,7 @@ INSTANTIATE_TEST_CASE_P(ColorFormats_3ch, PreprocTest,
                                 Values(std::make_pair(3, 3)),
                                 Values(TEST_SIZES_PREPROC)));
 
-INSTANTIATE_TEST_CASE_P(ColorFormats_4ch, PreprocTest,
+INSTANTIATE_TEST_SUITE_P(ColorFormats_4ch, PreprocTest,
                         Combine(PRECISIONS,
                                 Values(IE::ResizeAlgorithm::RESIZE_BILINEAR, IE::ResizeAlgorithm::RESIZE_AREA),
                                 Values(IE::ColorFormat::BGRX, IE::ColorFormat::RGBX),
@@ -402,7 +402,7 @@ INSTANTIATE_TEST_CASE_P(ColorFormats_4ch, PreprocTest,
                                 Values(std::make_pair(4, 3)),
                                 Values(TEST_SIZES_PREPROC)));
 
-INSTANTIATE_TEST_CASE_P(ColorFormat_NV12, PreprocTest,
+INSTANTIATE_TEST_SUITE_P(ColorFormat_NV12, PreprocTest,
                         Combine(Values(U8toU8),
                                 Values(IE::ResizeAlgorithm::RESIZE_BILINEAR, IE::ResizeAlgorithm::RESIZE_AREA),
                                 Values(IE::ColorFormat::NV12),
@@ -412,7 +412,7 @@ INSTANTIATE_TEST_CASE_P(ColorFormat_NV12, PreprocTest,
                                 Values(TEST_SIZES_PREPROC)));
 
 
-INSTANTIATE_TEST_CASE_P(PlainPrecisionConversions, PreprocTest,
+INSTANTIATE_TEST_SUITE_P(PlainPrecisionConversions, PreprocTest,
                         Combine(Values(std::make_pair(IE::Precision::U16,IE::Precision::FP32),
                                        std::make_pair(IE::Precision::FP32,IE::Precision::U16)
                                 ),
@@ -424,7 +424,7 @@ INSTANTIATE_TEST_CASE_P(PlainPrecisionConversions, PreprocTest,
                                 Values(std::make_pair(cv::Size(640,480), cv::Size(640,480)))));
 
 
-INSTANTIATE_TEST_CASE_P(PrecisionConversionsPipelines, PreprocTest,
+INSTANTIATE_TEST_SUITE_P(PrecisionConversionsPipelines, PreprocTest,
                         Combine(Values(std::make_pair(IE::Precision::U16, IE::Precision::FP32),
                                        std::make_pair(IE::Precision::FP32,IE::Precision::U8),
                                        std::make_pair(IE::Precision::U8,  IE::Precision::FP32)
diff --git a/inference-engine/tests_deprecated/functional/gna/backward_compatibility/backward_compatibility.cpp b/inference-engine/tests_deprecated/functional/gna/backward_compatibility/backward_compatibility.cpp
index 618c857c7b5a30..21ddf424726415 100644
--- a/inference-engine/tests_deprecated/functional/gna/backward_compatibility/backward_compatibility.cpp
+++ b/inference-engine/tests_deprecated/functional/gna/backward_compatibility/backward_compatibility.cpp
@@ -121,7 +121,7 @@ const std::vector<std::map<std::string, std::string>> importConfigs = {
 
 const std::vector<std::string> nameExportModel = {"export2dot1.blob", "export2dot2.blob", "export2dot3.blob", "export2dot4.blob"};
 
-INSTANTIATE_TEST_CASE_P(smoke_OldVersion, BackwardCompatibilityTests,
+INSTANTIATE_TEST_SUITE_P(smoke_OldVersion, BackwardCompatibilityTests,
                         ::testing::Combine(
                                 ::testing::ValuesIn(netPrecisions),
                                 ::testing::Values("GNA"),
diff --git a/inference-engine/tests_deprecated/functional/gna/shared_tests_instance/input_tests/parser_tests.cpp b/inference-engine/tests_deprecated/functional/gna/shared_tests_instance/input_tests/parser_tests.cpp
index d54817b53ed7ce..d53446b6a2e800 100644
--- a/inference-engine/tests_deprecated/functional/gna/shared_tests_instance/input_tests/parser_tests.cpp
+++ b/inference-engine/tests_deprecated/functional/gna/shared_tests_instance/input_tests/parser_tests.cpp
@@ -30,6 +30,6 @@ ir_test_params ir_test_cases[] = {
         ir_test_params("GNA", "FP32", negative_norm_k_case)
 };
 
-INSTANTIATE_TEST_CASE_P(FunctionalTest_smoke, IncorrectIRTests,
+INSTANTIATE_TEST_SUITE_P(FunctionalTest_smoke, IncorrectIRTests,
         ::testing::ValuesIn(ir_test_cases),
         getTestName);
\ No newline at end of file
diff --git a/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/cropResize_tests.hpp b/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/cropResize_tests.hpp
index 7000bedee2596b..957f47b31563f4 100644
--- a/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/cropResize_tests.hpp
+++ b/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/cropResize_tests.hpp
@@ -1186,9 +1186,9 @@ TEST_P(NV12ColorConvertTest, NV12Test) {
     ColorFormat::BGRX, ColorFormat::RGBX
 
 // #define PLUGING_CASE(_plugin, _test, _params) \
-//     INSTANTIATE_TEST_CASE_P(_plugin##_run, _test, Combine(Values(#_plugin "Plugin"), _params) )
+//     INSTANTIATE_TEST_SUITE_P(_plugin##_run, _test, Combine(Values(#_plugin "Plugin"), _params) )
 
 #define PLUGING_CASE_WITH_SUFFIX(_device, _suffix, _test, _params) \
-    INSTANTIATE_TEST_CASE_P(_device##_run##_suffix, _test, Combine(Values(#_device), _params) )
+    INSTANTIATE_TEST_SUITE_P(_device##_run##_suffix, _test, Combine(Values(#_device), _params) )
 
 #endif  // USE_OPENCV
diff --git a/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/dims_tests.hpp b/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/dims_tests.hpp
index 866542377bd7cf..b89ed7f1c7dea4 100644
--- a/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/dims_tests.hpp
+++ b/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/dims_tests.hpp
@@ -192,7 +192,7 @@ static auto params = ::testing::Values(Precision::FP32);  // network precision
 static auto params_myriad = ::testing::Values(Precision::FP16);  // network precision
 
 #define PLUGING_CASE(_device, _test, _params) \
-    INSTANTIATE_TEST_CASE_P(_device##_run, _test, ::testing::Combine(::testing::Values(#_device), _params) )
+    INSTANTIATE_TEST_SUITE_P(_device##_run, _test, ::testing::Combine(::testing::Values(#_device), _params) )
 
 #define PLUGING_CASE_WITH_SUFFIX(_device, _suffix, _test, _params) \
-    INSTANTIATE_TEST_CASE_P(_device##_run##_suffix, _test, ::testing::Combine(::testing::Values(#_device), _params) )
+    INSTANTIATE_TEST_SUITE_P(_device##_run##_suffix, _test, ::testing::Combine(::testing::Values(#_device), _params) )
diff --git a/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/layout_tests.hpp b/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/layout_tests.hpp
index 502cefab902003..06f0a48d71e501 100644
--- a/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/layout_tests.hpp
+++ b/inference-engine/tests_deprecated/functional/shared_tests/io_blob_tests/layout_tests.hpp
@@ -217,10 +217,10 @@ conv_param conv_p = { 3,3, // kernel
                       {2,16,13,13}}; // out shape
 
 #define PLUGING_CASE(_device, _test, _params) \
-    INSTANTIATE_TEST_CASE_P(_device##_run, _test, ::testing::Combine(::testing::Values(#_device), _params) )
+    INSTANTIATE_TEST_SUITE_P(_device##_run, _test, ::testing::Combine(::testing::Values(#_device), _params) )
 
 #define PLUGING_CASE_WITH_SUFFIX(_device, _suffix, _test, _params) \
-    INSTANTIATE_TEST_CASE_P(_device##_run##_suffix, _test, ::testing::Combine(::testing::Values(#_device), _params) )
+    INSTANTIATE_TEST_SUITE_P(_device##_run##_suffix, _test, ::testing::Combine(::testing::Values(#_device), _params) )
 
 #define PLUGING_CASE_WITH_PREFIX(_device, _prefix, _test, _params) \
-    INSTANTIATE_TEST_CASE_P(_prefix##_device##_run, _test, ::testing::Combine(::testing::Values(#_device), _params) )
+    INSTANTIATE_TEST_SUITE_P(_prefix##_device##_run, _test, ::testing::Combine(::testing::Values(#_device), _params) )
diff --git a/inference-engine/tests_deprecated/functional/shared_tests/lstm/plg_test.hpp b/inference-engine/tests_deprecated/functional/shared_tests/lstm/plg_test.hpp
index c7e5039b327fb3..9e5b63d76a3753 100644
--- a/inference-engine/tests_deprecated/functional/shared_tests/lstm/plg_test.hpp
+++ b/inference-engine/tests_deprecated/functional/shared_tests/lstm/plg_test.hpp
@@ -64,10 +64,10 @@ class Named {
  * @brief Macros to specify Per Plugin Run Test Case with parameters.
  */
 #define RUN_CASE_P_WITH_SUFFIX(_plugin, _suffix, _test, _params) \
-    INSTANTIATE_TEST_CASE_P(_plugin##_run##_suffix, _test, ::testing::Combine(::testing::Values(#_plugin), ::testing::ValuesIn(_params) ))
+    INSTANTIATE_TEST_SUITE_P(_plugin##_run##_suffix, _test, ::testing::Combine(::testing::Values(#_plugin), ::testing::ValuesIn(_params) ))
 
 /**
  * @brief Macros to specify Per Plugin Run Test Case with Cartesian Product of parameters.
  */
 #define RUN_CASE_CP_WITH_SUFFIX(_plugin, _suffix, _test, _params, ...) \
-    INSTANTIATE_TEST_CASE_P(_plugin##_run##_suffix, _test, ::testing::Combine(::testing::Values(#_plugin), _params), __VA_ARGS__ )
+    INSTANTIATE_TEST_SUITE_P(_plugin##_run##_suffix, _test, ::testing::Combine(::testing::Values(#_plugin), _params), __VA_ARGS__ )
diff --git a/inference-engine/tests_deprecated/functional/shared_tests/lstm/rnn_seq_test.hpp b/inference-engine/tests_deprecated/functional/shared_tests/lstm/rnn_seq_test.hpp
index 4e83949cf7389f..0c4db4ef993520 100644
--- a/inference-engine/tests_deprecated/functional/shared_tests/lstm/rnn_seq_test.hpp
+++ b/inference-engine/tests_deprecated/functional/shared_tests/lstm/rnn_seq_test.hpp
@@ -75,7 +75,7 @@ TEST_P(RNNSeqTest, DISABLED_SingleRNN) {
     auto resh = std::get<7>(p);
 
     if (device_name == "GPU" && cell.type != LSTM)
-        SKIP();
+        GTEST_SKIP();
 
     cell.clip = clip;
 
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_CTCDecoder_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_CTCDecoder_test.cpp
index d645f5627ea3d2..c95b62f8b3e83e 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_CTCDecoder_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_CTCDecoder_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_CTCDecoder_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
 	DISABLED_accuracy, myriadCTCDecoderLayerTests_smoke,
 	::testing::Combine(
 		::testing::Values<Dims>({{1, 88, 1, 71}}),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_batch_normalization_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_batch_normalization_test.cpp
index 7e0378afebfaa0..d64f8d4eeea5f0 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_batch_normalization_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_batch_normalization_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_batch_normalization_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayersTestsBatchNormalization_smoke,
         ::testing::Values(
                 bn_test_params{{1, 1, 16, 8}, 0.001f},
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_bias_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_bias_test.cpp
index a54a50efbe0303..8757631cb8e9bb 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_bias_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_bias_test.cpp
@@ -4,6 +4,6 @@
 
 #include "myriad_layers_bias_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsBias_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsBias_smoke,
         ::testing::ValuesIn(s_biasDims)
 );
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_blob_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_blob_test.cpp
index 7a2c027be03113..6128ac1ee1bd17 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_blob_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_blob_test.cpp
@@ -62,7 +62,7 @@ TEST_P(myriadBlobTests_smoke, CanGetSameBlobsOnSameIR) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadBlobTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadBlobTests_smoke,
     ::testing::Values(CONFIG_VALUE(YES), CONFIG_VALUE(NO))
 );
 
@@ -356,7 +356,7 @@ using myriadExtraTests_smoke = myriadLayersTests_nightly;
 
 TEST_F(myriadExtraTests_smoke, ThereIsNoSegfaultOnZeroConvolutionWeights) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     tensor_test_params input_dims = { 1, 3, 25, 25 };
@@ -413,5 +413,5 @@ static const std::vector<InferenceEngine::Precision> inputPrecisions = {Inferenc
 static const std::vector<InferenceEngine::Precision> outputPrecisions = {InferenceEngine::Precision::FP16, InferenceEngine::Precision::FP32};
 
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadBlobExportAccuracyDifferentPrecisionOfInAndOutTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadBlobExportAccuracyDifferentPrecisionOfInAndOutTests_smoke,
                         ::testing::Combine(::testing::ValuesIn(inputPrecisions), ::testing::ValuesIn(outputPrecisions)));
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_clamp_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_clamp_test.cpp
index 4d031eed5dcd2c..5f5e64c36156a3 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_clamp_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_clamp_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_clamp_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsClampParams_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsClampParams_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_clampTensors),
         ::testing::ValuesIn(s_clampParams))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_concat_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_concat_test.cpp
index 5872049504a145..7b7ad34fdce67b 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_concat_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_concat_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_concat_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsConcat_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsConcat_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_concatCores),
         ::testing::ValuesIn(s_axis),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_conv_nd_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_conv_nd_test.cpp
index 472403fa6f283a..f5c59174bd12b6 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_conv_nd_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_conv_nd_test.cpp
@@ -12,7 +12,7 @@ using namespace testing;
 //
 //----------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(tricky_ncdhw_userpad, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(tricky_ncdhw_userpad, myriadLayersConvNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 19, 65, 47}),
         Values(KernelShape {1, 3, 5}),
@@ -26,7 +26,7 @@ INSTANTIATE_TEST_CASE_P(tricky_ncdhw_userpad, myriadLayersConvNDTest_smoke,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(tricky_ncdhw_autopad, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(tricky_ncdhw_autopad, myriadLayersConvNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 19, 65, 47}),
         Values(KernelShape {1, 3, 5}),
@@ -48,7 +48,7 @@ INSTANTIATE_TEST_CASE_P(tricky_ncdhw_autopad, myriadLayersConvNDTest_smoke,
 //
 //----------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(simple_ncdhw_userpad, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(simple_ncdhw_userpad, myriadLayersConvNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 20, 64, 48}),
         Values(KernelShape {3, 3, 3}),
@@ -62,7 +62,7 @@ INSTANTIATE_TEST_CASE_P(simple_ncdhw_userpad, myriadLayersConvNDTest_smoke,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(simple_ncdhw_autopad, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(simple_ncdhw_autopad, myriadLayersConvNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 20, 64, 48}),
         Values(KernelShape {3, 3, 3}),
@@ -84,7 +84,7 @@ INSTANTIATE_TEST_CASE_P(simple_ncdhw_autopad, myriadLayersConvNDTest_smoke,
 //
 //----------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(tricky_nchw_userpad, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(tricky_nchw_userpad, myriadLayersConvNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 65, 47}),
         Values(KernelShape {1, 3}),
@@ -98,7 +98,7 @@ INSTANTIATE_TEST_CASE_P(tricky_nchw_userpad, myriadLayersConvNDTest_smoke,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(tricky_nchw_autopad, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(tricky_nchw_autopad, myriadLayersConvNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 65, 47}),
         Values(KernelShape {1, 3}),
@@ -120,7 +120,7 @@ INSTANTIATE_TEST_CASE_P(tricky_nchw_autopad, myriadLayersConvNDTest_smoke,
 //
 //----------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(simple_nchw_userpad, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(simple_nchw_userpad, myriadLayersConvNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 64, 48}),
         Values(KernelShape {3, 3}),
@@ -134,7 +134,7 @@ INSTANTIATE_TEST_CASE_P(simple_nchw_userpad, myriadLayersConvNDTest_smoke,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(simple_nchw_autopad, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(simple_nchw_autopad, myriadLayersConvNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 64, 48}),
         Values(KernelShape {3, 3}),
@@ -158,7 +158,7 @@ INSTANTIATE_TEST_CASE_P(simple_nchw_autopad, myriadLayersConvNDTest_smoke,
 
 // NB: requires 1GB of RAM on device (e.g. ma2085 board)
 // Stress test: large image with large depth, large kernel
-INSTANTIATE_TEST_CASE_P(i3d_id6, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(i3d_id6, myriadLayersConvNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 3, 79, 224, 224}),
                                 Values(KernelShape {7, 7, 7}),
@@ -171,7 +171,7 @@ INSTANTIATE_TEST_CASE_P(i3d_id6, myriadLayersConvNDTest_smoke,
                                 Values(Groups(1))));
 
 // Like `i3d_id6` test but with smaller image (so must fit in Myriad X)
-INSTANTIATE_TEST_CASE_P(i3d_id6_shrink, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(i3d_id6_shrink, myriadLayersConvNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 3, 39, 112, 112}),
                                 Values(KernelShape {7, 7, 7}),
@@ -184,7 +184,7 @@ INSTANTIATE_TEST_CASE_P(i3d_id6_shrink, myriadLayersConvNDTest_smoke,
                                 Values(Groups(1))));
 
 // Average-size image, trivial kernel 1x1x1
-INSTANTIATE_TEST_CASE_P(i3d_id12, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(i3d_id12, myriadLayersConvNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 64, 40, 56, 56}),
                                 Values(KernelShape {1, 1, 1}),
@@ -197,7 +197,7 @@ INSTANTIATE_TEST_CASE_P(i3d_id12, myriadLayersConvNDTest_smoke,
                                 Values(Groups(1))));
 
 // Average-size image, non-trivial kernel 3x3x3
-INSTANTIATE_TEST_CASE_P(i3d_id17, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(i3d_id17, myriadLayersConvNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 64, 40, 56, 56}),
                                 Values(KernelShape {3, 3, 3}),
@@ -210,7 +210,7 @@ INSTANTIATE_TEST_CASE_P(i3d_id17, myriadLayersConvNDTest_smoke,
                                 Values(Groups(1))));
 
 // Small image (7x7), trivial kernel
-INSTANTIATE_TEST_CASE_P(i3d_id249, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(i3d_id249, myriadLayersConvNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 832, 10, 7, 7}),
                                 Values(KernelShape {1, 1, 1}),
@@ -223,7 +223,7 @@ INSTANTIATE_TEST_CASE_P(i3d_id249, myriadLayersConvNDTest_smoke,
                                 Values(Groups(1))));
 
 // Small image (7x7), non-trivial kernel
-INSTANTIATE_TEST_CASE_P(i3d_id301, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(i3d_id301, myriadLayersConvNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 48, 10, 7, 7}),
                                 Values(KernelShape {3, 3, 3}),
@@ -236,7 +236,7 @@ INSTANTIATE_TEST_CASE_P(i3d_id301, myriadLayersConvNDTest_smoke,
                                 Values(Groups(1))));
 
 // Trivial image (1x1), trivial kernel
-INSTANTIATE_TEST_CASE_P(i3d_id314, myriadLayersConvNDTest_smoke,
+INSTANTIATE_TEST_SUITE_P(i3d_id314, myriadLayersConvNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 1024, 9, 1, 1}),
                                 Values(KernelShape {1, 1, 1}),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convert_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convert_test.cpp
index 582311afe2289a..0b2fe34caeaa35 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convert_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convert_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_convert_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     accuracy, myriadLayersTestsIOConvert_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(inputsDims),
@@ -13,7 +13,7 @@ INSTANTIATE_TEST_CASE_P(
     )
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy_customu8f16, myriadLayersTestsIOConvert_smoke,
         ::testing::Combine(
                 ::testing::ValuesIn(inputsDims4D),
@@ -22,7 +22,7 @@ INSTANTIATE_TEST_CASE_P(
         )
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     accuracy, myriadLayersTestsConvertWithFP16_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(inputsDims),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution1x1.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution1x1.cpp
index b25f1d80d38e12..eaf3e8f3052d64 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution1x1.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution1x1.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_convolution1x1.hpp"
 
-INSTANTIATE_TEST_CASE_P(myriad, myriadConvolution1x1LayerTests_smoke,
+INSTANTIATE_TEST_SUITE_P(myriad, myriadConvolution1x1LayerTests_smoke,
         ::testing::Combine(
         ::testing::Values(CONFIG_VALUE(NO)),
         ::testing::ValuesIn(s_isHWC),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution3x3.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution3x3.cpp
index 537bef657fdf3b..97257fd76c51f3 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution3x3.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution3x3.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_convolution3x3.hpp"
 
-INSTANTIATE_TEST_CASE_P(myriad, myriadConvolution3x3LayerTests_smoke,
+INSTANTIATE_TEST_SUITE_P(myriad, myriadConvolution3x3LayerTests_smoke,
         ::testing::Combine(
         ::testing::Values(CONFIG_VALUE(NO)),
         ::testing::ValuesIn(s_DimsConfig)));
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution_test.cpp
index 9093504eaf65ea..8bea8f5d995f7d 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_convolution_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_convolution_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy_chw_dilation, myriadLayerConvolution_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_chw_dilation, myriadLayerConvolution_smoke,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 37, 43, 43)
                                        , MAKE_STRUCT(tensor_test_params, 1, 37, 19, 19))
@@ -21,7 +21,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_chw_dilation, myriadLayerConvolution_smoke,
         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayers_IR3_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayers_IR3_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({1, 3, 32, 24})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -33,7 +33,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayers_IR3_ConvTests_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_0, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_0, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 5, 1, 1})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -45,7 +45,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_0, myriadLayers_BatchTest_ConvTests_smoke
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_1, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_1, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 576, 7, 7})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -57,7 +57,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_1, myriadLayers_BatchTest_ConvTests_smoke
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_2, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_2, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 128, 7, 7})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -69,7 +69,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_2, myriadLayers_BatchTest_ConvTests_smoke
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_3, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_3, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 4, 7, 7})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -81,7 +81,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_3, myriadLayers_BatchTest_ConvTests_smoke
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_4, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_4, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 256, 7, 7})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -93,7 +93,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_4, myriadLayers_BatchTest_ConvTests_smoke
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_5, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_5, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 1024, 4, 4})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -105,7 +105,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_5, myriadLayers_BatchTest_ConvTests_smoke
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_6, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_6, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 192, 4, 4})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -117,7 +117,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_6, myriadLayers_BatchTest_ConvTests_smoke
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_7, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_7, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 160, 4, 4})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -129,7 +129,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_7, myriadLayers_BatchTest_ConvTests_smoke
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_8, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_8, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 224, 4, 4})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -141,7 +141,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_8, myriadLayers_BatchTest_ConvTests_smoke
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_9, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_9, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 1024, 4, 4})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -153,7 +153,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_9, myriadLayers_BatchTest_ConvTests_smoke
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_10, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_10, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({1, 64, 56, 56})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -165,7 +165,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_10, myriadLayers_BatchTest_ConvTests_smok
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_11, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_11, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 192, 7, 7})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -177,7 +177,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_11, myriadLayers_BatchTest_ConvTests_smok
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_12, myriadLayers_BatchTest_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_12, myriadLayers_BatchTest_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 576, 7, 7})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -189,7 +189,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_12, myriadLayers_BatchTest_ConvTests_smok
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_Batch_1, myriadLayers_BatchTest2_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_Batch_1, myriadLayers_BatchTest2_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({10, 576, 7, 7})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -201,7 +201,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_Batch_1, myriadLayers_BatchTest2_ConvTests_smok
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_3X3, myriadLayers_IR3_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_3X3, myriadLayers_IR3_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({1, 3, 32, 24})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -213,7 +213,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_3X3, myriadLayers_IR3_ConvTests_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_3X1, myriadLayers_IR3_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_3X1, myriadLayers_IR3_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({1, 3, 32, 24})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 1))
@@ -225,7 +225,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_3X1, myriadLayers_IR3_ConvTests_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_1X3, myriadLayers_IR3_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_1X3, myriadLayers_IR3_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({1, 4, 16, 16})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 3))
@@ -237,7 +237,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_1X3, myriadLayers_IR3_ConvTests_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayers_3X3X3_ConstInput_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayers_3X3X3_ConstInput_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({1, 3, 10, 10})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -249,7 +249,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayers_3X3X3_ConstInput_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_crossroad_spatialConv, myriadLayerConvolutionTensorFlow,
+INSTANTIATE_TEST_SUITE_P(accuracy_crossroad_spatialConv, myriadLayerConvolutionTensorFlow,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 3, 1024, 1024))
           , ::testing::Values<DimsOutput>(MAKE_STRUCT(tensor_test_params, 1, 3, 512, 512))
@@ -262,7 +262,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_crossroad_spatialConv, myriadLayerConvolutionTe
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_inception_v2, myriadLayerConvolutionTensorFlow,
+INSTANTIATE_TEST_SUITE_P(accuracy_inception_v2, myriadLayerConvolutionTensorFlow,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 16, 28, 28))
           , ::testing::Values<DimsOutput>(MAKE_STRUCT(tensor_test_params, 1, 64, 14, 14))
@@ -275,7 +275,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_inception_v2, myriadLayerConvolutionTensorFlow,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_inception_v1, myriadLayerConvolutionTensorFlow,
+INSTANTIATE_TEST_SUITE_P(accuracy_inception_v1, myriadLayerConvolutionTensorFlow,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 3, 224, 224),
                                          MAKE_STRUCT(tensor_test_params, 1, 32, 224, 224)
@@ -290,7 +290,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_inception_v1, myriadLayerConvolutionTensorFlow,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(test_3x3_SSD_dilation, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(test_3x3_SSD_dilation, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 512, 19, 19))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -303,7 +303,7 @@ INSTANTIATE_TEST_CASE_P(test_3x3_SSD_dilation, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(test_TF_Resnet_50, myriadLayers_IR3_ConvTests_smoke,
+INSTANTIATE_TEST_SUITE_P(test_TF_Resnet_50, myriadLayers_IR3_ConvTests_smoke,
         ::testing::Combine(
             ::testing::Values<InferenceEngine::SizeVector>({1, 512, 38, 38})
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -315,7 +315,7 @@ INSTANTIATE_TEST_CASE_P(test_TF_Resnet_50, myriadLayers_IR3_ConvTests_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(test_3x3_icvnet_dilation, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(test_3x3_icvnet_dilation, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 24, 20, 20))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -330,7 +330,7 @@ INSTANTIATE_TEST_CASE_P(test_3x3_icvnet_dilation, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(test_5x5_with_dilation, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(test_5x5_with_dilation, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 64,  77))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 5, 5))
@@ -346,7 +346,7 @@ INSTANTIATE_TEST_CASE_P(test_5x5_with_dilation, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(test_7x7_with_dilation, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(test_7x7_with_dilation, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 64,  77))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 7, 7))
@@ -363,7 +363,7 @@ INSTANTIATE_TEST_CASE_P(test_7x7_with_dilation, myriadLayerConvolution,
 );
 
 
-INSTANTIATE_TEST_CASE_P(test_conv1x1, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(test_conv1x1, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 10, 13, 13))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -377,7 +377,7 @@ INSTANTIATE_TEST_CASE_P(test_conv1x1, myriadLayerConvolution,
            )
 );
 
-INSTANTIATE_TEST_CASE_P(test_yolo_tiny_2_512x13x13_use_3x3_convolution, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(test_yolo_tiny_2_512x13x13_use_3x3_convolution, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 512, 13, 13))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -390,7 +390,7 @@ INSTANTIATE_TEST_CASE_P(test_yolo_tiny_2_512x13x13_use_3x3_convolution, myriadLa
            )
 );
 
-INSTANTIATE_TEST_CASE_P(test_yolo_tiny_2_512x13x13_use_1x1_convolution, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(test_yolo_tiny_2_512x13x13_use_1x1_convolution, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 4608, 13, 13))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -403,7 +403,7 @@ INSTANTIATE_TEST_CASE_P(test_yolo_tiny_2_512x13x13_use_1x1_convolution, myriadLa
            )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_group, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_group, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 64,  77)
                                        , MAKE_STRUCT(tensor_test_params, 1, 32, 112, 96))
@@ -422,7 +422,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_group, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_group_large_input, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_group_large_input, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 192, 336))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -437,7 +437,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_group_large_input, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_any_group, myriadLayerConvolution_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_any_group, myriadLayerConvolution_smoke,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 64,  77))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -453,7 +453,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_any_group, myriadLayerConvolution_smoke,
         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_any_group, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_any_group, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 64,  77))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -469,7 +469,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_any_group, myriadLayerConvolution,
         )
 );
 
-INSTANTIATE_TEST_CASE_P(set_optimization_for_3x3_with_group, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(set_optimization_for_3x3_with_group, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 24, 80, 80)
                                        , MAKE_STRUCT(tensor_test_params, 1, 36, 80, 80))
@@ -484,7 +484,7 @@ INSTANTIATE_TEST_CASE_P(set_optimization_for_3x3_with_group, myriadLayerConvolut
           )
 );
 
-INSTANTIATE_TEST_CASE_P(set_optimization_for_3x3s1, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(set_optimization_for_3x3s1, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 24, 80, 80))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -497,7 +497,7 @@ INSTANTIATE_TEST_CASE_P(set_optimization_for_3x3s1, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_1x1, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_1x1, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 16, 64, 64)
                                        , MAKE_STRUCT(tensor_test_params, 1, 32, 1, 1))
@@ -514,7 +514,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_1x1, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_3x3, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_3x3, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 8, 16, 16)
                                        , MAKE_STRUCT(tensor_test_params, 1, 8, 59, 73))
@@ -530,7 +530,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_3x3, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_1x3, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_1x3, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 8, 59, 73))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 3), MAKE_STRUCT(param_size, 3, 1))
@@ -544,7 +544,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_1x3, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_5x5, myriadLayerConvolution_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_5x5, myriadLayerConvolution_smoke,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 16, 32, 32)
                                      /*, MAKE_STRUCT(tensor_test_params, 1, 8, 511, 399) failed*/)
@@ -561,7 +561,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_5x5, myriadLayerConvolution_smoke,
         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_5x5, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_5x5, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 16, 32, 32)
                                      /*, MAKE_STRUCT(tensor_test_params, 1, 8, 511, 399) failed*/)
@@ -578,7 +578,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_5x5, myriadLayerConvolution,
         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_7x7, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_7x7, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 8, 32, 32)
                                      /*, MAKE_STRUCT(tensor_test_params, 1, 8, 511, 399) failed*/)
@@ -595,7 +595,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_7x7, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_3x3_large_input_1, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_3x3_large_input_1, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 3, 720, 1280))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -608,7 +608,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_3x3_large_input_1, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_3x3_large_input_2, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_3x3_large_input_2, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 24, 357, 637))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -622,7 +622,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_3x3_large_input_2, myriadLayerConvolution,
 );
 
 
-INSTANTIATE_TEST_CASE_P(accuracy_3x3_large_input_3, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_3x3_large_input_3, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 16, 359, 639))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -635,7 +635,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_3x3_large_input_3, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_1x1_large_input, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_1x1_large_input, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 24, 355, 635))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -648,7 +648,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_1x1_large_input, myriadLayerConvolution,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_small_input_0, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_small_input_0, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 128, 38, 38))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -660,7 +660,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_small_input_0, myriadLayerConvolution,
           , ::testing::Values<layoutPreference>(vpu::LayoutPreference::ChannelMinor)
           )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_small_input_1, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_small_input_1, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 2, 3))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -672,7 +672,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_small_input_1, myriadLayerConvolution,
           , ::testing::Values<layoutPreference>(vpu::LayoutPreference::ChannelMinor)
           )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_small_input_2, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_small_input_2, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 2, 2))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -684,7 +684,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_small_input_2, myriadLayerConvolution,
           , ::testing::Values<layoutPreference>(vpu::LayoutPreference::ChannelMinor)
           )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_small_input_3, myriadLayerConvolution,
+INSTANTIATE_TEST_SUITE_P(accuracy_small_input_3, myriadLayerConvolution,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 1, 1))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_copy_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_copy_test.cpp
index 2dafed621277ba..2f42d4d9dfff13 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_copy_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_copy_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_copy_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerCopy_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerCopy_smoke,
         ::testing::Combine(
             ::testing::Values<NDims>(MAKE_STRUCT(nd_tensor_test_params, {36, 19, 20, 21})
                                    , MAKE_STRUCT(nd_tensor_test_params, {7, 8, 5, 12})
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_crop_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_crop_test.cpp
index 8df5595df1786b..c7bcc255f3ca6e 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_crop_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_crop_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_crop_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     accuracy_Crop, myriadLayerCropOneInputAndDim_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_tileTensors1),
@@ -14,7 +14,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::ValuesIn(s_tileDim))
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     accuracy_Crop1, myriadLayerCropOneInput_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_tileTensors1),
@@ -24,7 +24,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::ValuesIn(s_tileCropEnd))
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     accuracy_Crop2, myriadLayerCropTwoInputs_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_tileTensors1),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_custom_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_custom_test.cpp
index 106f2ea52532a6..ceebce481f85bb 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_custom_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_custom_test.cpp
@@ -4,27 +4,27 @@
 
 #include "myriad_layers_custom_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsShuffleChannel_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsShuffleChannel_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_ShuffleChannelTensors),
         ::testing::ValuesIn(s_ShuffleChannelGroup),
         ::testing::ValuesIn(s_CustomConfig)));
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsFakeQuantize_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsFakeQuantize_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_QuantizeTensors),
         ::testing::ValuesIn(s_QuantizeLevels),
         ::testing::Values(IRVersion::v7, IRVersion::v10),
         ::testing::ValuesIn(s_CustomConfig)));
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsQuantizeBinarize_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsQuantizeBinarize_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_QuantizeTensors),
         ::testing::Values(2),
         ::testing::ValuesIn(s_QuantizeSwitchOut),
         ::testing::ValuesIn(s_CustomConfig)));
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsBinaryConvolution_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsBinaryConvolution_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_BinaryConvolutionTensors),
         ::testing::ValuesIn(s_BinaryConvolutionDilations),
@@ -33,17 +33,17 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsBinaryConvolution_smoke,
         ::testing::ValuesIn(s_BinaryConvolutionStrides),
         ::testing::ValuesIn(s_CustomConfig)));
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsExperimentalDetectronPriorGridGenerator_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsExperimentalDetectronPriorGridGenerator_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_ExperimentalDetectronPriorGridGeneratorImageDims),
         ::testing::ValuesIn(s_CustomConfig)));
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsCorrelate_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsCorrelate_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_CorrelateParams),
         ::testing::ValuesIn(s_CustomConfig)));
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsSpatialTransform_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsSpatialTransform_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_SpatialTransformInputs),
         ::testing::ValuesIn(s_SpatialTransformTheta),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_deconvolution_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_deconvolution_test.cpp
index 6dd93d3b1460e6..c7412b5760c193 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_deconvolution_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_deconvolution_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_deconvolution_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy_deconv_to_conv, myriadLayerDeconvolution_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_deconv_to_conv, myriadLayerDeconvolution_smoke,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 6, 5, 6))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 1), MAKE_STRUCT(param_size, 3, 3))
@@ -17,7 +17,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_deconv_to_conv, myriadLayerDeconvolution_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_deconv_to_conv_2, myriadLayerDeconvolution_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_deconv_to_conv_2, myriadLayerDeconvolution_smoke,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 2, 256, 14, 14))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 2, 2), MAKE_STRUCT(param_size, 3, 3))
@@ -30,7 +30,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_deconv_to_conv_2, myriadLayerDeconvolution_smok
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_group, myriadLayerDeconvolution_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_group, myriadLayerDeconvolution_smoke,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 384, 4, 2))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 2, 2)
@@ -47,7 +47,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_group, myriadLayerDeconvolution_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_depthDeconv, myriadLayerDeconvolution_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_depthDeconv, myriadLayerDeconvolution_smoke,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 384, 4, 2))
           , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 2, 2)
@@ -65,7 +65,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_depthDeconv, myriadLayerDeconvolution_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerDeconvolution_asymm_pad,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerDeconvolution_asymm_pad,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 128, 60, 80),
                                          MAKE_STRUCT(tensor_test_params, 1,   2, 37, 59))
@@ -84,7 +84,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerDeconvolution_asymm_pad,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerDeconvolution_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerDeconvolution_smoke,
         ::testing::Combine(
             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 2, 37, 59)
                                        , MAKE_STRUCT(tensor_test_params, 1, 21, 16, 16)
@@ -109,7 +109,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerDeconvolution_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(extra3x3s1, myriadLayerDeconvolution_smoke,
+INSTANTIATE_TEST_SUITE_P(extra3x3s1, myriadLayerDeconvolution_smoke,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 1, 1))
                               , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_eltwise_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_eltwise_test.cpp
index 9b307b8d91e438..8945b73fdd4fb2 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_eltwise_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_eltwise_test.cpp
@@ -4,168 +4,168 @@
 
 #include "myriad_layers_eltwise_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseMax_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseMax_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseSum_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseSum_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseSub_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseSub_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseMul_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseMul_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseSumWithCoeff_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseSumWithCoeff_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseSumWithBroadcast_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseSumWithBroadcast_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseInputs),
         ::testing::Values<int>(4))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseSubWithCoeff_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseSubWithCoeff_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseSubWithBroadcast_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseSubWithBroadcast_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::Values<int>(4))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseDiv_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseDiv_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseMin_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseMin_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseSqDiff_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseSqDiff_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwisePow_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwisePow_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseFloorMod_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseFloorMod_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseEqual_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseEqual_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseNotEqual_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseNotEqual_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseGreater_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseGreater_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseGreaterEqual_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseGreaterEqual_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseLess_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseLess_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseLessEqual_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseLessEqual_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseLogicalNot_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseLogicalNot_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyOneInput),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseLogicalAnd_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseLogicalAnd_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseLogicalOr_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseLogicalOr_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseLogicalXor_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseLogicalXor_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseInputs),
         ::testing::ValuesIn(s_eltwiseDims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsEltwiseMean_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsEltwiseMean_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseOnlyTwoInputs),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_elu_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_elu_test.cpp
index 16d211a1467917..5e99e6795b445e 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_elu_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_elu_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_elu_test.hpp"
 
-INSTANTIATE_TEST_CASE_P( accuracy, myriadLayersTestsELUParams_smoke,
+INSTANTIATE_TEST_SUITE_P( accuracy, myriadLayersTestsELUParams_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_powerTensors),
         ::testing::ValuesIn(s_powerParams))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_erf_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_erf_test.cpp
index 609d76fa4c227f..4730fa0a2e9fd1 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_erf_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_erf_test.cpp
@@ -4,6 +4,6 @@
 
 #include "myriad_layers_erf_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayersTestsErf_smoke,
         ::testing::ValuesIn(s_ErfDims));
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_detectionoutput_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_detectionoutput_test.cpp
index 23f6b46deaca60..2ac5cb77e043a5 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_detectionoutput_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_detectionoutput_test.cpp
@@ -18,7 +18,7 @@ static const std::vector<ExpDetectionOutputParams> s_layerParams_list =
     {{ 10.0, 10.0, 5.0, 5.0 }, 4.135166645050049, 0.5, 0.05, _MaxDetections, _NumClasses, 2000, 0 },
 };
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsExpDetectionOutput_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsExpDetectionOutput_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_sizeParams_list),
         ::testing::ValuesIn(s_layerParams_list))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_generateproposals.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_generateproposals.cpp
index 63175d7ce9d843..027198d3f01be0 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_generateproposals.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_generateproposals.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_exp_generateproposals_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsExpGenerateProposals_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsExpGenerateProposals_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_ExpGenerateProposalsLayerScores),
         ::testing::ValuesIn(s_ExpGenerateProposalsLayerImInfo),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_priorgridgenerator_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_priorgridgenerator_test.cpp
index 1036a4bdf7da45..3c63109a214141 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_priorgridgenerator_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_priorgridgenerator_test.cpp
@@ -4,19 +4,19 @@
 
 #include "myriad_layers_exp_priorgridgenerator_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsExpPriorGridGeneratorAllInputs_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsExpPriorGridGeneratorAllInputs_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_ExpPriorGridGeneratorLayerInputs),
         ::testing::ValuesIn(s_ExpPriorGridGeneratorLayerParam))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsExpPriorGridGeneratorNoFeatureMap_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsExpPriorGridGeneratorNoFeatureMap_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_ExpPriorGridGenLayerNoFMInputs),
         ::testing::ValuesIn(s_ExpPriorGridGenLayerNoFMParam))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsExpPriorGridGeneratorNoInputImage_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsExpPriorGridGeneratorNoInputImage_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_ExpPriorGridGenLayerNoInputImage),
         ::testing::ValuesIn(s_ExpPriorGridGenLayerNoInputImageParam))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_test.cpp
index b78906e928d640..def1e2174980fa 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_test.cpp
@@ -4,6 +4,6 @@
 
 #include "myriad_layers_exp_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayersTestsExp_smoke,
         ::testing::ValuesIn(s_expParams));
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_topkrois_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_topkrois_test.cpp
index ff0094c4f20462..8ed4022ba138dd 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_topkrois_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_exp_topkrois_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_exp_topkrois_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsExpTopKROIs_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsExpTopKROIs_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_ExpTopKROIsInputRoisNum),
         ::testing::ValuesIn(s_ExpTopKROIsMaxRoisNum))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_flatten_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_flatten_test.cpp
index 98978ecfd11650..260f01d1b8f155 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_flatten_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_flatten_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_flatten_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsFlatten_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsFlatten_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_flattenTensors),
         ::testing::ValuesIn(s_flattenAxis)
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_floor_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_floor_test.cpp
index cdd5dc7422ac10..046608b0d5e039 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_floor_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_floor_test.cpp
@@ -4,6 +4,6 @@
 
 #include "myriad_layers_floor_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayersTestsFloor_smoke,
         ::testing::ValuesIn(s_FloorParams));
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_fully_connected_tests.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_fully_connected_tests.cpp
index 8ec8ad6378ce77..7e65bc26ec03f3 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_fully_connected_tests.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_fully_connected_tests.cpp
@@ -3,19 +3,19 @@
 //
 
 #include "myriad_layers_fully_connected_tests.hpp"
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayersTestsFullyConnected_smoke,
         ::testing::ValuesIn(s_fcTestParams)
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsFullyConnectedBatch_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsFullyConnectedBatch_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(s_fcTestBatchParams)
           , ::testing::ValuesIn(s_fcTestBatchOutSizes)
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsFullyConnectedPVA_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsFullyConnectedPVA_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(s_fcTestPVAParams)
           , ::testing::ValuesIn(s_fcTestPVAOutSizes)
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_gemm_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_gemm_test.cpp
index 7cb12c87748490..bd4c7483665e78 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_gemm_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_gemm_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_gemm_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerGEMM_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerGEMM_smoke,
         ::testing::Combine(
         ::testing::Values<gemm_parameters>(
                 MAKE_STRUCT(gemm_parameters, 4.7f, 2.3f, 5,  7,   11,  1, 2,   3, 4,  5, 6,  7, 8),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_grn_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_grn_test.cpp
index 5cccc5f5be32d4..d8e94ac8dc899b 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_grn_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_grn_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_grn_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
 	accuracy, myriadLayersTestsGRN_smoke,
 	::testing::Combine(
 		::testing::ValuesIn(s_GRNInputs),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_interp_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_interp_test.cpp
index bfedb4600671d1..3fff0275d7bab7 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_interp_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_interp_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_interp_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayersTestsInterp_smoke,
         ::testing::Combine(
 
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_log_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_log_test.cpp
index 490a762997e805..0e709247015aab 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_log_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_log_test.cpp
@@ -4,6 +4,6 @@
 
 #include "myriad_layers_log_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayersTestsLog_smoke,
         ::testing::ValuesIn(s_logParams));
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_lrn_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_lrn_test.cpp
index a7b3172ff43149..e3a0a66af6624a 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_lrn_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_lrn_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_lrn_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsLRN_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsLRN_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_LRNTensors),
         ::testing::ValuesIn(s_LRNlocal_size),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_lstm_cell.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_lstm_cell.cpp
index 819de8087dbbc6..1b7ad35d89261a 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_lstm_cell.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_lstm_cell.cpp
@@ -162,6 +162,6 @@ TEST_F(myriadLayersTests_nightly, LSTMCellSequenceNet) {
     CompareCommonAbsolute(output, refOut0, ERROR_BOUND);
 }
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsLSTMCell_smoke,
-                        ::testing::Values<lstmcell_test_params>(MAKE_STRUCT(lstmcell_test_params, 512, 128)),
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsLSTMCell_smoke,
+        ::testing::Values<lstmcell_test_params>(MAKE_STRUCT(lstmcell_test_params, 512, 128))
 );
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_mvn_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_mvn_test.cpp
index 702c3688e560dc..16bf3cdc6655d2 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_mvn_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_mvn_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_mvn_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMVN_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsMVN_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(s_MVNTensors),
                                 ::testing::ValuesIn(s_MVN_acrossChannels),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_nonzero_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_nonzero_test.cpp
index 0075457ce6a4db..7a330009890329 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_nonzero_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_nonzero_test.cpp
@@ -4,5 +4,5 @@
 
 #include "myriad_layers_nonzero_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerTestNonZero_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerTestNonZero_smoke,
                         ::testing::ValuesIn(inputDims));
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_normalize_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_normalize_test.cpp
index 34b9e4600cd86b..37f5948e5c0cdf 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_normalize_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_normalize_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_normalize_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsNormalize_smoke, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsNormalize_smoke, ::testing::Combine(
     ::testing::Values<Dims>(
         // small size, num_channels is not divisible by 8
         MAKE_STRUCT(tensor_test_params, 1, 33, 1, 1),
@@ -21,7 +21,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsNormalize_smoke, ::testing::C
 ));
 
 
-INSTANTIATE_TEST_CASE_P(accuracy_more, myriadLayersTestsNormalize_smoke, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(accuracy_more, myriadLayersTestsNormalize_smoke, ::testing::Combine(
     ::testing::Values<Dims>(
         //more tests
         MAKE_STRUCT(tensor_test_params, 1, 1, 38, 38),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_oneHot_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_oneHot_test.cpp
index 7233519144730b..b3041a99167667 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_oneHot_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_oneHot_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_oneHot_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerTestOneHot_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerTestOneHot_smoke,
                         ::testing::Values<oneHot_test_params>(
                                 MAKE_STRUCT(OneHotParams, {64}, 2, {0}, {}, {}),
                                 MAKE_STRUCT(OneHotParams, {64}, 2, {-1}, {}, {}),
@@ -22,7 +22,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerTestOneHot_smoke,
                                 MAKE_STRUCT(OneHotParams, {4, 8, 16, 32, 64}, 2, {-1}, {}, {})
                         ));
 
-INSTANTIATE_TEST_CASE_P(accuracy_add, myriadLayerTestOneHot_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_add, myriadLayerTestOneHot_smoke,
                         ::testing::Values<oneHot_test_params>(
                                 MAKE_STRUCT(OneHotParams, {16, 32, 64}, 2, {2}, {}, {}),
                                 MAKE_STRUCT(OneHotParams, {8, 16, 32,64}, 2, {2}, {}, {}),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pad_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pad_test.cpp
index e45d93d121fb15..c68044dad2afd8 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pad_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pad_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_pad_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerPad_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerPad_smoke,
                         ::testing::Combine(
                             ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 16, 16)),
                             ::testing::Values<pad_parameters>(MAKE_STRUCT(pad_parameters, 0, 32, 1, 2,  0, 32, 3, 4)),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_permute_nd_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_permute_nd_test.cpp
index 463163e8c561fb..96c50476c84b14 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_permute_nd_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_permute_nd_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_permute_nd_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy_2D, myriadLayersPermuteNDTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2D, myriadLayersPermuteNDTests_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(s_inTensors_2D)
           , ::testing::ValuesIn(s_permuteTensors_2D)
@@ -12,7 +12,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_2D, myriadLayersPermuteNDTests_smoke,
           , ::testing::ValuesIn(s_permutePrecisions)
 ));
 
-INSTANTIATE_TEST_CASE_P(accuracy_3D, myriadLayersPermuteNDTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_3D, myriadLayersPermuteNDTests_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(s_inTensors_3D)
           , ::testing::ValuesIn(s_permuteTensors_3D)
@@ -20,7 +20,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_3D, myriadLayersPermuteNDTests_smoke,
           , ::testing::ValuesIn(s_permutePrecisions)
 ));
 
-INSTANTIATE_TEST_CASE_P(accuracy_4D, myriadLayersPermuteNDTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_4D, myriadLayersPermuteNDTests_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(s_inTensors_4D)
           , ::testing::ValuesIn(s_permuteTensors_4D)
@@ -28,7 +28,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_4D, myriadLayersPermuteNDTests_smoke,
           , ::testing::ValuesIn(s_permutePrecisions)
 ));
 
-INSTANTIATE_TEST_CASE_P(accuracy_5D, myriadLayersPermuteNDTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_5D, myriadLayersPermuteNDTests_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(s_inTensors_5D)
           , ::testing::ValuesIn(s_permuteTensors_5D)
@@ -36,7 +36,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_5D, myriadLayersPermuteNDTests_smoke,
           , ::testing::ValuesIn(s_permutePrecisions)
 ));
 
-INSTANTIATE_TEST_CASE_P(fc_to_conv_case, myriadLayersPermuteNDTests_smoke,
+INSTANTIATE_TEST_SUITE_P(fc_to_conv_case, myriadLayersPermuteNDTests_smoke,
     ::testing::Values(
         std::make_tuple(
             SizeVector{8, 50, 256, 7, 7},
@@ -53,7 +53,7 @@ INSTANTIATE_TEST_CASE_P(fc_to_conv_case, myriadLayersPermuteNDTests_smoke,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_FasterRCNN, myriadLayersPermuteNDTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_FasterRCNN, myriadLayersPermuteNDTests_smoke,
         ::testing::Combine(
              ::testing::Values<InferenceEngine::SizeVector>({1, 24, 14, 14})
             ,::testing::Values<InferenceEngine::SizeVector>({0, 2, 3, 1})
@@ -61,7 +61,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_FasterRCNN, myriadLayersPermuteNDTests_smoke,
             ,::testing::ValuesIn(s_permutePrecisions)
             ));
 
-INSTANTIATE_TEST_CASE_P(accuracy_MaskRCNN, myriadLayersPermuteNDTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_MaskRCNN, myriadLayersPermuteNDTests_smoke,
         ::testing::Combine(
              ::testing::Values<InferenceEngine::SizeVector>({4, 3, 1, 88, 120})
             ,::testing::Values<InferenceEngine::SizeVector>({0, 3, 4, 1, 2})
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_permute_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_permute_test.cpp
index a8738e300b51c6..0ccef5858f03e4 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_permute_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_permute_test.cpp
@@ -4,13 +4,13 @@
 
 #include "myriad_layers_permute_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersPermuteTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersPermuteTests_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(s_inTensors)
           , ::testing::ValuesIn(s_permuteTensors)
 ));
 
-INSTANTIATE_TEST_CASE_P(accuracyFasterRCNN, myriadLayersPermuteTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracyFasterRCNN, myriadLayersPermuteTests_smoke,
         ::testing::Combine(
              ::testing::Values<InferenceEngine::SizeVector>({1, 24, 14, 14})
             ,::testing::Values<InferenceEngine::SizeVector>({0, 2, 3, 1})
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pool_nd_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pool_nd_test.cpp
index b6b791c108a08d..094ef5c4994df4 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pool_nd_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pool_nd_test.cpp
@@ -12,7 +12,7 @@ using namespace testing;
 //
 //======================================================================
 
-INSTANTIATE_TEST_CASE_P(tricky_ncdhw_avg_userpad,
+INSTANTIATE_TEST_SUITE_P(tricky_ncdhw_avg_userpad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 19, 65, 47}),
@@ -28,7 +28,7 @@ INSTANTIATE_TEST_CASE_P(tricky_ncdhw_avg_userpad,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(tricky_ncdhw_max_userpad,
+INSTANTIATE_TEST_SUITE_P(tricky_ncdhw_max_userpad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 19, 65, 47}),
@@ -43,7 +43,7 @@ INSTANTIATE_TEST_CASE_P(tricky_ncdhw_max_userpad,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(tricky_ncdhw_avg_autopad,
+INSTANTIATE_TEST_SUITE_P(tricky_ncdhw_avg_autopad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 19, 65, 47}),
@@ -61,7 +61,7 @@ INSTANTIATE_TEST_CASE_P(tricky_ncdhw_avg_autopad,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(tricky_ncdhw_max_autopad,
+INSTANTIATE_TEST_SUITE_P(tricky_ncdhw_max_autopad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 19, 65, 47}),
@@ -84,7 +84,7 @@ INSTANTIATE_TEST_CASE_P(tricky_ncdhw_max_autopad,
 //
 //======================================================================
 
-INSTANTIATE_TEST_CASE_P(simple_ncdhw_avg_userpad,
+INSTANTIATE_TEST_SUITE_P(simple_ncdhw_avg_userpad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 20, 64, 48}),
@@ -100,7 +100,7 @@ INSTANTIATE_TEST_CASE_P(simple_ncdhw_avg_userpad,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(simple_ncdhw_max_userpad,
+INSTANTIATE_TEST_SUITE_P(simple_ncdhw_max_userpad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 20, 64, 48}),
@@ -124,7 +124,7 @@ INSTANTIATE_TEST_CASE_P(simple_ncdhw_max_userpad,
 // Issue-15146 HW AvgPool doesn't support excludePad parameter
 //----------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(simple_ncdhw_avg_autopad_1,
+INSTANTIATE_TEST_SUITE_P(simple_ncdhw_avg_autopad_1,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 20, 64, 48}),
@@ -141,7 +141,7 @@ INSTANTIATE_TEST_CASE_P(simple_ncdhw_avg_autopad_1,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(simple_ncdhw_avg_autopad_2,
+INSTANTIATE_TEST_SUITE_P(simple_ncdhw_avg_autopad_2,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 20, 64, 48}),
@@ -158,7 +158,7 @@ INSTANTIATE_TEST_CASE_P(simple_ncdhw_avg_autopad_2,
 
 //----------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(simple_ncdhw_max_autopad,
+INSTANTIATE_TEST_SUITE_P(simple_ncdhw_max_autopad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 20, 64, 48}),
@@ -181,7 +181,7 @@ INSTANTIATE_TEST_CASE_P(simple_ncdhw_max_autopad,
 //
 //======================================================================
 
-INSTANTIATE_TEST_CASE_P(tricky_nchw_avg_userpad,
+INSTANTIATE_TEST_SUITE_P(tricky_nchw_avg_userpad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 65, 47}),
@@ -197,7 +197,7 @@ INSTANTIATE_TEST_CASE_P(tricky_nchw_avg_userpad,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(tricky_nchw_max_userpad,
+INSTANTIATE_TEST_SUITE_P(tricky_nchw_max_userpad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 65, 47}),
@@ -212,7 +212,7 @@ INSTANTIATE_TEST_CASE_P(tricky_nchw_max_userpad,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(tricky_nchw_avg_autopad,
+INSTANTIATE_TEST_SUITE_P(tricky_nchw_avg_autopad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 65, 47}),
@@ -230,7 +230,7 @@ INSTANTIATE_TEST_CASE_P(tricky_nchw_avg_autopad,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(tricky_nchw_max_autopad,
+INSTANTIATE_TEST_SUITE_P(tricky_nchw_max_autopad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 65, 47}),
@@ -253,7 +253,7 @@ INSTANTIATE_TEST_CASE_P(tricky_nchw_max_autopad,
 //
 //======================================================================
 
-INSTANTIATE_TEST_CASE_P(simple_nchw_avg_userpad,
+INSTANTIATE_TEST_SUITE_P(simple_nchw_avg_userpad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 64, 48}),
@@ -269,7 +269,7 @@ INSTANTIATE_TEST_CASE_P(simple_nchw_avg_userpad,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(simple_nchw_max_userpad,
+INSTANTIATE_TEST_SUITE_P(simple_nchw_max_userpad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 64, 48}),
@@ -293,7 +293,7 @@ INSTANTIATE_TEST_CASE_P(simple_nchw_max_userpad,
 // Issue-15146 HW AvgPool doesn't support excludePad parameter
 //----------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(simple_nchw_avg_autopad_1,
+INSTANTIATE_TEST_SUITE_P(simple_nchw_avg_autopad_1,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 64, 48}),
@@ -310,7 +310,7 @@ INSTANTIATE_TEST_CASE_P(simple_nchw_avg_autopad_1,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(simple_nchw_avg_autopad_2,
+INSTANTIATE_TEST_SUITE_P(simple_nchw_avg_autopad_2,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 64, 48}),
@@ -327,7 +327,7 @@ INSTANTIATE_TEST_CASE_P(simple_nchw_avg_autopad_2,
 
 //----------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(simple_nchw_max_autopad,
+INSTANTIATE_TEST_SUITE_P(simple_nchw_max_autopad,
                         myriadLayersPoolNDTest_smoke,
     Combine(
         Values(InputShape {1, 3, 64, 48}),
@@ -350,7 +350,7 @@ INSTANTIATE_TEST_CASE_P(simple_nchw_max_autopad,
 //
 //======================================================================
 
-INSTANTIATE_TEST_CASE_P(i3d_id10,
+INSTANTIATE_TEST_SUITE_P(i3d_id10,
                         myriadLayersPoolNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 64, 40, 112, 112}),
@@ -363,7 +363,7 @@ INSTANTIATE_TEST_CASE_P(i3d_id10,
                                 Values(RoundingType("")),
                                 Values(ExcludePad(true))));
 
-INSTANTIATE_TEST_CASE_P(i3d_id47,
+INSTANTIATE_TEST_SUITE_P(i3d_id47,
                         myriadLayersPoolNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 192, 40, 28, 28}),
@@ -376,7 +376,7 @@ INSTANTIATE_TEST_CASE_P(i3d_id47,
                                 Values(RoundingType("")),
                                 Values(ExcludePad(true))));
 
-INSTANTIATE_TEST_CASE_P(i3d_id247,
+INSTANTIATE_TEST_SUITE_P(i3d_id247,
                         myriadLayersPoolNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 832, 20, 14, 14}),
@@ -389,7 +389,7 @@ INSTANTIATE_TEST_CASE_P(i3d_id247,
                                 Values(RoundingType("")),
                                 Values(ExcludePad(true))));
 
-INSTANTIATE_TEST_CASE_P(i3d_id312,
+INSTANTIATE_TEST_SUITE_P(i3d_id312,
                         myriadLayersPoolNDTest_smoke,
                         Combine(
                                 Values(InputShape {1, 1024, 10, 7, 7}),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pooling_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pooling_test.cpp
index 1fda0995a0b357..5bd7ddd3df49dc 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pooling_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_pooling_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_pooling_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({10, 192, 56, 56})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 3, 3)) /* kernel     */
@@ -17,7 +17,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayers_IR3_BatchPoolingTests_smoke,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_1, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_1, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({10, 576, 14, 14})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -31,7 +31,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_1, myriadLayers_IR3_BatchPoolingTests_smoke,
 );
 
 
-INSTANTIATE_TEST_CASE_P(accuracy_4X4, myriadLayers_IR3_PoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_4X4, myriadLayers_IR3_PoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({10, 1024, 4, 4})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 4, 4)) /* kernel     */
@@ -44,7 +44,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_4X4, myriadLayers_IR3_PoolingTests_smoke,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_1X1, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_1X1, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 3, 5, 7})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 1, 1)) /* kernel     */
@@ -57,7 +57,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_1X1, myriadLayers_IR3_BatchPoolingTests_smoke,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_2X2p0000, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2X2p0000, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 512, 26, 26})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -70,7 +70,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_2X2p0000, myriadLayers_IR3_BatchPoolingTests_sm
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_2X2p0001, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2X2p0001, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 512, 26, 26})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -82,7 +82,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_2X2p0001, myriadLayers_IR3_BatchPoolingTests_sm
                                 , ::testing::ValuesIn(s_poolingMethod)
                         )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_2X2p0011, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2X2p0011, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 512, 26, 26})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -94,7 +94,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_2X2p0011, myriadLayers_IR3_BatchPoolingTests_sm
                                 , ::testing::ValuesIn(s_poolingMethod)
                         )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_2X2p0111, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2X2p0111, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 512, 26, 26})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -106,7 +106,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_2X2p0111, myriadLayers_IR3_BatchPoolingTests_sm
                                 , ::testing::ValuesIn(s_poolingMethod)
                         )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_2X2p1111, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2X2p1111, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 512, 26, 26})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -118,7 +118,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_2X2p1111, myriadLayers_IR3_BatchPoolingTests_sm
                                 , ::testing::ValuesIn(s_poolingMethod)
                         )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_2X2p1110, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2X2p1110, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 512, 26, 26})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -130,7 +130,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_2X2p1110, myriadLayers_IR3_BatchPoolingTests_sm
                                 , ::testing::ValuesIn(s_poolingMethod)
                         )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_2X2p1100, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2X2p1100, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 512, 26, 26})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -142,7 +142,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_2X2p1100, myriadLayers_IR3_BatchPoolingTests_sm
                                 , ::testing::ValuesIn(s_poolingMethod)
                         )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_2X2p1000, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2X2p1000, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 512, 26, 26})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -154,7 +154,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_2X2p1000, myriadLayers_IR3_BatchPoolingTests_sm
                                 , ::testing::ValuesIn(s_poolingMethod)
                         )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_2X2p1101, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2X2p1101, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 512, 26, 26})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -166,7 +166,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_2X2p1101, myriadLayers_IR3_BatchPoolingTests_sm
                                 , ::testing::ValuesIn(s_poolingMethod)
                         )
 );
-INSTANTIATE_TEST_CASE_P(accuracy_2X2p1011, myriadLayers_IR3_BatchPoolingTests_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_2X2p1011, myriadLayers_IR3_BatchPoolingTests_smoke,
                         ::testing::Combine(
                                 ::testing::Values<InferenceEngine::SizeVector>({1, 512, 26, 26})
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 2, 2)) /* kernel     */
@@ -179,14 +179,14 @@ INSTANTIATE_TEST_CASE_P(accuracy_2X2p1011, myriadLayers_IR3_BatchPoolingTests_sm
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMax_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsMax_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInput),
                                 ::testing::ValuesIn(g_poolingLayerParamsFull),
                                 ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMaxOverlappedByKernel_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsMaxOverlappedByKernel_smoke,
                         ::testing::Combine(
                             ::testing::Values<InferenceEngine::SizeVector>({1, 1024, 6, 6}),
                             ::testing::Values<param_size>(MAKE_STRUCT(param_size, 7, 7)),
@@ -195,7 +195,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMaxOverlappedByKernel_smoke,
                             ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMaxPad4_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsMaxPad4_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInputPad4),
                                 ::testing::ValuesIn(g_poolingKernelPad4),
@@ -204,7 +204,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMaxPad4_smoke,
                                 ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvgPad4_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsAvgPad4_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInputPad4),
                                 ::testing::ValuesIn(g_poolingKernelPad4),
@@ -213,24 +213,24 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvgPad4_smoke,
                                 ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsGlobalMax_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsGlobalMax_smoke,
                         ::testing::ValuesIn(g_GlobalPoolingInput ));
 
-INSTANTIATE_TEST_CASE_P(accuracy_3x3, myriadLayersTestsMax_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_3x3, myriadLayersTestsMax_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInput),
                                 ::testing::ValuesIn(s_poolingLayerParams_k3x3),
                                 ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvg_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsAvg_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInput),
                                 ::testing::ValuesIn(g_poolingLayerParamsFull),
                                 ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvgOverlappedByKernel_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsAvgOverlappedByKernel_smoke,
                         ::testing::Combine(
                             ::testing::Values<InferenceEngine::SizeVector>({1, 1024, 6, 6}),
                             ::testing::Values<param_size>(MAKE_STRUCT(param_size, 7, 7)),
@@ -239,12 +239,12 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvgOverlappedByKernel_smoke,
                             ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_3x3, myriadLayersTestsAvg_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_3x3, myriadLayersTestsAvg_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInput),
                                 ::testing::ValuesIn(s_poolingLayerParams_k3x3),
                                 ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsGlobalAvg_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsGlobalAvg_smoke,
                         ::testing::ValuesIn(g_GlobalPoolingInput));
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_power_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_power_test.cpp
index a8557aa10ed013..8585729e5292fd 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_power_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_power_test.cpp
@@ -3,7 +3,7 @@
 //
 
 #include "myriad_layers_power_test.hpp"
-INSTANTIATE_TEST_CASE_P( accuracy, myriadLayersTestsPowerParams_smoke,
+INSTANTIATE_TEST_SUITE_P( accuracy, myriadLayersTestsPowerParams_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_powerTensors),
         ::testing::ValuesIn(s_powerParams))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_prelu_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_prelu_test.cpp
index 7a6f469a06acee..35f41cbe28c32f 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_prelu_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_prelu_test.cpp
@@ -4,14 +4,14 @@
 
 #include "myriad_layers_prelu_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy_PReLU, myriadLayerPReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_PReLU, myriadLayerPReLU_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_PReLUTensors)
       , ::testing::Values<ChannelSharedPrelu>(0, 1)
     )
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     accuracy, myriadLayerFullyConnectedWithPReLU_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_fcTestParamsSubset),
@@ -21,7 +21,7 @@ INSTANTIATE_TEST_CASE_P(
     )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMaxPoolingWithPReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsMaxPoolingWithPReLU_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_poolingInput),
         ::testing::ValuesIn(g_poolingLayerParamsLite),
@@ -29,7 +29,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMaxPoolingWithPReLU_smoke,
         ::testing::ValuesIn(s_PReluLayerParams))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvgPoolingWithPReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsAvgPoolingWithPReLU_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_poolingInput),
         ::testing::ValuesIn(g_poolingLayerParamsLite),
@@ -37,7 +37,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvgPoolingWithPReLU_smoke,
         ::testing::ValuesIn(s_PReluLayerParams))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_postop, myriadLayersTestsMaxPoolingWithPReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_postop, myriadLayersTestsMaxPoolingWithPReLU_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_poolingInput_postOp),
         ::testing::Values<pooling_layer_params>(MAKE_STRUCT(pooling_layer_params, {3, 3}, {1, 1}, {1, 1})),
@@ -45,7 +45,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_postop, myriadLayersTestsMaxPoolingWithPReLU_sm
         ::testing::Values<PReLULayerDef>(MAKE_STRUCT(PReLULayerDef, {{{PRELU_PARAM, "0"}}})))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_postop, myriadLayersTestsAvgPoolingWithPReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_postop, myriadLayersTestsAvgPoolingWithPReLU_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_poolingInput_postOp),
         ::testing::Values<pooling_layer_params>(MAKE_STRUCT(pooling_layer_params, {3, 3}, {1, 1}, {1, 1})),
@@ -53,7 +53,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_postop, myriadLayersTestsAvgPoolingWithPReLU_sm
         ::testing::Values<PReLULayerDef>(MAKE_STRUCT(PReLULayerDef, {{{PRELU_PARAM, "0"}}})))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerConvolutionWithPReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerConvolutionWithPReLU_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(g_convolutionTensors)
           , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 3, 3))
@@ -65,7 +65,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerConvolutionWithPReLU_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_postop, myriadLayerConvolutionWithPReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_postop, myriadLayerConvolutionWithPReLU_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(g_poolingInput_postOp)
           , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 3, 3))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_psroipooling_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_psroipooling_test.cpp
index 1955910aa7f78f..4a5196bef7e81b 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_psroipooling_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_psroipooling_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_psroipooling_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsPSROIPooling_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsPSROIPooling_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_PSROIPoolingLayerInput),
         ::testing::ValuesIn(s_PSROIPoolingLayerParam),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_psroipooling_test.hpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_psroipooling_test.hpp
index 9f25bb1df37fd7..0aadb311e02402 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_psroipooling_test.hpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_psroipooling_test.hpp
@@ -192,7 +192,7 @@ static std::vector<uint32_t> s_PSROIPoolingNumROIs = {
 
 TEST_P(myriadLayersTestsPSROIPooling_smoke, PSROIPooling) {
 #ifdef _WIN32
-    SKIP() << "Disabled for Windows. Issue-13239";
+    GTEST_SKIP() << "Disabled for Windows. Issue-13239";
 #endif
     tensor_test_params dims_layer_in = std::get<0>(GetParam());
     PSROIPoolingParams test_params = std::get<1>(GetParam());
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reduce_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reduce_test.cpp
index eeb62e51ca0e27..689233138776cf 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reduce_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reduce_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_reduce_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsReduceAnd_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsReduceAnd_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_input_pair),
         ::testing::ValuesIn(s_axes_list),
@@ -12,7 +12,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsReduceAnd_smoke,
         ::testing::ValuesIn(s_keep_dims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsReduceMin_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsReduceMin_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_input_pair),
         ::testing::ValuesIn(s_axes_list),
@@ -20,7 +20,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsReduceMin_smoke,
         ::testing::ValuesIn(s_keep_dims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsReduceMax_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsReduceMax_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_input_pair),
         ::testing::ValuesIn(s_axes_list),
@@ -28,7 +28,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsReduceMax_smoke,
         ::testing::ValuesIn(s_keep_dims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsReduceSum_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsReduceSum_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_input_pair),
         ::testing::ValuesIn(s_axes_list),
@@ -36,7 +36,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsReduceSum_smoke,
         ::testing::ValuesIn(s_keep_dims))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsReduceMean_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsReduceMean_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_input_pair),
         ::testing::ValuesIn(s_axes_list),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_region_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_region_test.cpp
index a5470c4ac956f1..80c2a4c1e11a76 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_region_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_region_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_region_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
 	accuracy, myriadLayersTestsRegionYolo_smoke,
 	::testing::Combine(
 		::testing::Values<Coords>(4),
@@ -19,7 +19,7 @@ INSTANTIATE_TEST_CASE_P(
 
 #ifdef VPU_HAS_CUSTOM_KERNELS
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
 	accuracy_custom, myriadLayersTestsRegionYolo_smoke,
 	::testing::Combine(
 		::testing::Values<Coords>(4),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_relu_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_relu_test.cpp
index 84c9a828040b65..da43b0b7922efd 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_relu_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_relu_test.cpp
@@ -4,14 +4,14 @@
 
 #include "myriad_layers_relu_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerReLU_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(s_copyTensors),
                                 ::testing::ValuesIn(s_reluLayerParams)
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayerFullyConnectedWithReLU_smoke,
         ::testing::Combine(
                 ::testing::ValuesIn(g_fcTestParamsSubset),
@@ -21,7 +21,7 @@ INSTANTIATE_TEST_CASE_P(
         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMaxPoolingWithReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsMaxPoolingWithReLU_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInput),
                                 ::testing::ValuesIn(g_poolingLayerParamsLite),
@@ -29,7 +29,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMaxPoolingWithReLU_smoke,
                                 ::testing::ValuesIn(s_reluLayerParams))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvgPoolingWithReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsAvgPoolingWithReLU_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInput),
                                 ::testing::ValuesIn(g_poolingLayerParamsLite),
@@ -37,7 +37,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvgPoolingWithReLU_smoke,
                                 ::testing::ValuesIn(s_reluLayerParams))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_postop, myriadLayersTestsMaxPoolingWithReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_postop, myriadLayersTestsMaxPoolingWithReLU_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInput_postOp),
                                 ::testing::Values<pooling_layer_params>(MAKE_STRUCT(pooling_layer_params, {3, 3}, {1, 1}, {1, 1})),
@@ -45,7 +45,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_postop, myriadLayersTestsMaxPoolingWithReLU_smo
                                 ::testing::Values<ReLULayerDef>(MAKE_STRUCT(ReLULayerDef, {{{"negative_slope", "0.0"}}})))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_postop, myriadLayersTestsAvgPoolingWithReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_postop, myriadLayersTestsAvgPoolingWithReLU_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInput_postOp),
                                 ::testing::Values<pooling_layer_params>(MAKE_STRUCT(pooling_layer_params, {3, 3}, {1, 1}, {1, 1})),
@@ -53,7 +53,7 @@ INSTANTIATE_TEST_CASE_P(accuracy_postop, myriadLayersTestsAvgPoolingWithReLU_smo
                                 ::testing::Values<ReLULayerDef>(MAKE_STRUCT(ReLULayerDef, {{{"negative_slope", "0.0"}}})))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerConvolutionWithReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerConvolutionWithReLU_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_convolutionTensors)
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 3, 3))
@@ -65,7 +65,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerConvolutionWithReLU_smoke,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_postop, myriadLayerConvolutionWithReLU_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_postop, myriadLayerConvolutionWithReLU_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(g_poolingInput_postOp)
                                 , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 3, 3))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reorg_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reorg_test.cpp
index 6b5e88de88d17f..964170e4f21ba6 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reorg_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reorg_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_reorg_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsReorg_smoke, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsReorg_smoke, ::testing::Combine(
     ::testing::ValuesIn(s_ReorgInputs),
     ::testing::Values<Stride>(2),
     ::testing::Values(vpu::LayoutPreference::ChannelMinor, vpu::LayoutPreference::ChannelMajor),
@@ -14,7 +14,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsReorg_smoke, ::testing::Combi
 
 #ifdef VPU_HAS_CUSTOM_KERNELS
 
-INSTANTIATE_TEST_CASE_P(accuracy_custom, myriadLayersTestsReorg_smoke, ::testing::Combine(
+INSTANTIATE_TEST_SUITE_P(accuracy_custom, myriadLayersTestsReorg_smoke, ::testing::Combine(
     ::testing::ValuesIn(s_ReorgInputs_CustomLayer),
     ::testing::Values<Stride>(2),
     ::testing::Values(vpu::LayoutPreference::ChannelMinor, vpu::LayoutPreference::ChannelMajor),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_resample_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_resample_test.cpp
index c111327edd8614..f9f57089c10a8a 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_resample_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_resample_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_resample_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
 	accuracy, myriadResampleLayerTests_smoke,
 	::testing::Combine(
 		::testing::ValuesIn(s_ResampleInput),
@@ -16,7 +16,7 @@ INSTANTIATE_TEST_CASE_P(
 
 #ifdef VPU_HAS_CUSTOM_KERNELS
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
 	accuracy_custom, myriadResampleLayerTests_smoke,
 	::testing::Combine(
 		::testing::ValuesIn(s_ResampleInput),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reshape_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reshape_test.cpp
index 721a9c6820155e..3dde909a85efa5 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reshape_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reshape_test.cpp
@@ -264,13 +264,13 @@ TEST_F(myriadLayerReshapeFasterRCNN_smoke, Reshape) {
     ASSERT_TRUE(generateNetAndInfer(NetworkInitParams().useHWOpt( CheckMyriadX())));
 }
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerReshape_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerReshape_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_reshapeInParams),
         ::testing::ValuesIn(s_reshapeOutParams))
 );
 
-INSTANTIATE_TEST_CASE_P(fc_to_conv_case, myriadLayerReshape_smoke,
+INSTANTIATE_TEST_SUITE_P(fc_to_conv_case, myriadLayerReshape_smoke,
     ::testing::Values(
         std::make_tuple(
             SizeVector{400, 12544},
@@ -291,11 +291,11 @@ INSTANTIATE_TEST_CASE_P(fc_to_conv_case, myriadLayerReshape_smoke,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsReshapeBeforeFC_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsReshapeBeforeFC_smoke,
         ::testing::Values(CONFIG_VALUE(YES), CONFIG_VALUE(NO))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsReshapeFasterRCNN_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsReshapeFasterRCNN_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(s_convTensor)
           , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 1, 1))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reverse_sequence_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reverse_sequence_test.cpp
index 21938c5e595ba1..2685d602f842f8 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reverse_sequence_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_reverse_sequence_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_reverse_sequence_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerReverseSequence_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerReverseSequence_smoke,
     ::testing::Combine(
         ::testing::Values<ReverseSequence>(
                 MAKE_STRUCT(reverse_sequence_test_params, {5, 6, 18}, 0, 0)
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_align_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_align_test.cpp
index 87c467e3dfc6bd..a19af42ecf3425 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_align_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_align_test.cpp
@@ -4,18 +4,18 @@
 
 #include "myriad_layers_roi_align_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsROIAlign_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsROIAlign_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_ROIAlignLayerInput),
         ::testing::ValuesIn(s_ROIAlignLayerParam),
         ::testing::ValuesIn(s_ROIAlignNumROIs),
-        ::testing::ValuesIn(s_ROIAlignMode)),
+        ::testing::ValuesIn(s_ROIAlignMode))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy_faster, myriadLayersTestsROIAlign_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy_faster, myriadLayersTestsROIAlign_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_ROIAlignLayerInput_Faster),
         ::testing::ValuesIn(s_ROIAlignLayerParam_Faster),
         ::testing::ValuesIn(s_ROIAlignNumROIs_Faster),
-        ::testing::ValuesIn(s_ROIAlignMode_Faster)),
+        ::testing::ValuesIn(s_ROIAlignMode_Faster))
 );
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_feature_extractor_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_feature_extractor_test.cpp
index e047f00ce0c47a..2c30e1b1bc4e47 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_feature_extractor_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_feature_extractor_test.cpp
@@ -4,9 +4,9 @@
 
 #include "myriad_layers_roi_feature_extractor_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsROIFeatureExtractor_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsROIFeatureExtractor_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_ROIFeatureExtractorLayerInput),
         ::testing::ValuesIn(s_ROIFeatureExtractorLayerParam),
-        ::testing::ValuesIn(s_ROIFeatureExtractorNumROIs)),
+        ::testing::ValuesIn(s_ROIFeatureExtractorNumROIs))
 );
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_pooling_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_pooling_test.cpp
index dc88686743cd4d..6e8bf111eae12e 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_pooling_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_roi_pooling_test.cpp
@@ -4,11 +4,11 @@
 
 #include "myriad_layers_roi_pooling_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsROIPooling_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsROIPooling_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_ROIPoolingLayerInput),
         ::testing::ValuesIn(s_ROIPoolingLayerParam),
         ::testing::ValuesIn(s_ROIPoolingNumRois),
         ::testing::ValuesIn(s_ROIPoolingMethod),
-        ::testing::Values<IRVersion>(IRVersion::v7, IRVersion::v10)),
+        ::testing::Values<IRVersion>(IRVersion::v7, IRVersion::v10))
 );
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scale_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scale_test.cpp
index 0bee6c57c432ee..66a2df071366dc 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scale_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scale_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_scale_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayersTestsScale_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(s_inputScaleTensors),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scatter_elements_update_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scatter_elements_update_test.cpp
index e344d9fac8a151..e156cc2559a876 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scatter_elements_update_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scatter_elements_update_test.cpp
@@ -53,12 +53,12 @@ static const std::vector<DataShape> dataShapeList_useCases = {
     { 16, 512, 56, 56 },
 };
 
-INSTANTIATE_TEST_CASE_P(nd_tensors, myriadLayersScatterElementsUpdateTest_smoke,
+INSTANTIATE_TEST_SUITE_P(nd_tensors, myriadLayersScatterElementsUpdateTest_smoke,
                         Combine(
                             ValuesIn(dataShapeList_ndTensors),
                             ValuesIn(dataTypeList)));
 
-INSTANTIATE_TEST_CASE_P(use_cases, myriadLayersScatterElementsUpdateTest_smoke,
+INSTANTIATE_TEST_SUITE_P(use_cases, myriadLayersScatterElementsUpdateTest_smoke,
                         Combine(
                             ValuesIn(dataShapeList_useCases),
                             ValuesIn(dataTypeList)));
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scatter_update_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scatter_update_test.cpp
index 4412c1d1cf6248..640598f5fb55ff 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scatter_update_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_scatter_update_test.cpp
@@ -12,7 +12,7 @@ using namespace testing;
 //
 //----------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     nd_tensors,
     myriadLayersScatterUpdateTest_smoke,
     Values(
@@ -49,7 +49,7 @@ INSTANTIATE_TEST_CASE_P(
 //
 //----------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     use_cases,
     myriadLayersScatterUpdateTest_smoke,
     Values(
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_select_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_select_test.cpp
index 76939b7c23e7b4..ef886903419dd0 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_select_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_select_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_select_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsSelect_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsSelect_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_eltwiseTensors),
         ::testing::ValuesIn(s_eltwiseDims))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_sigmoid_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_sigmoid_test.cpp
index 4eca99fa2ff43f..091740c25c65af 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_sigmoid_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_sigmoid_test.cpp
@@ -4,25 +4,25 @@
 
 #include "myriad_layers_sigmoid_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayersTestsSigmoid_smoke,
         ::testing::ValuesIn(s_sigmoidParams));
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMaxPoolingWithSigmoid_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsMaxPoolingWithSigmoid_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_poolingInput),
         ::testing::ValuesIn(g_poolingLayerParamsLite),
         ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvgPoolingWithSigmoid_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsAvgPoolingWithSigmoid_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_poolingInput),
         ::testing::ValuesIn(g_poolingLayerParamsLite),
         ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerConvolutionWithSigmoid_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerConvolutionWithSigmoid_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(g_convolutionTensors)
           , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 3, 3))
@@ -34,7 +34,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerConvolutionWithSigmoid_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     accuracy, myriadLayerFullyConnectedWithSigmoid_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_fcTestParamsSubset),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_slice_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_slice_test.cpp
index f710365f2e02c3..c99bef79a167c2 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_slice_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_slice_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_slice_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsSlice_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsSlice_smoke,
                         ::testing::Values<SliceTestParams>(
                                 MAKE_STRUCT(SliceParams, {4, 8, 16, 32, 64}, {{4, 8, 16, 10, 64}, {4, 8, 16, 22, 64}}, 3),
                                 MAKE_STRUCT(SliceParams, {4, 8, 16, 32}, {{4, 8, 2, 32}, {4, 8, 14, 32}}, 2))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_softmax_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_softmax_test.cpp
index 75cea33ed2b3b2..3d64c5c37a1168 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_softmax_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_softmax_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_softmax_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     accuracy, myriadLayersTestsSoftMax_smoke,    
     ::testing::Combine(
         ::testing::ValuesIn(s_softMaxTensors)
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_split_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_split_test.cpp
index cc28aa41976701..c75d3b375924b4 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_split_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_split_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_split_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsSplit_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsSplit_smoke,
                         ::testing::Values<SplitTestParams>(
                                 MAKE_STRUCT(SplitParams, {4, 8, 16, 32, 64}, 2, 6),
                                 MAKE_STRUCT(SplitParams, {4, 8, 16, 32}, 2, 6),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_squeeze_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_squeeze_test.cpp
index be4f155363970d..170aab55518f39 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_squeeze_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_squeeze_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_squeeze_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsSqueezeTC1_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsSqueezeTC1_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_squeezeTensorsTC1),
         ::testing::ValuesIn(s_squeezeIndicesTC1),
@@ -13,7 +13,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsSqueezeTC1_smoke,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsSqueezeTC2_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsSqueezeTC2_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_squeezeTensorsTC2),
         ::testing::ValuesIn(s_squeezeIndicesTC2),
@@ -23,7 +23,7 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsSqueezeTC2_smoke,
 );
 
 // TODO: rewrite to ngraph to have reshape functionality
-INSTANTIATE_TEST_CASE_P(DISABLED_accuracy, myriadLayersTestsSqueezeTC3_smoke,
+INSTANTIATE_TEST_SUITE_P(DISABLED_accuracy, myriadLayersTestsSqueezeTC3_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_squeezeTensorsTC3),
         ::testing::ValuesIn(s_squeezeIndicesTC3),
@@ -33,7 +33,7 @@ INSTANTIATE_TEST_CASE_P(DISABLED_accuracy, myriadLayersTestsSqueezeTC3_smoke,
 );
 
 // TODO: rewrite to ngraph to have reshape functionality
-INSTANTIATE_TEST_CASE_P(DISABLED_accuracy, myriadLayersTestsSqueezeTC4_smoke,
+INSTANTIATE_TEST_SUITE_P(DISABLED_accuracy, myriadLayersTestsSqueezeTC4_smoke,
         ::testing::Combine(
         ::testing::ValuesIn(s_squeezeTensorsTC4),
         ::testing::ValuesIn(s_squeezeIndicesTC4),
@@ -42,7 +42,7 @@ INSTANTIATE_TEST_CASE_P(DISABLED_accuracy, myriadLayersTestsSqueezeTC4_smoke,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsSqueezeTC5_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsSqueezeTC5_smoke,
                         ::testing::Combine(
                                 ::testing::ValuesIn(s_squeezeTensorsTC5),
                                 ::testing::ValuesIn(s_squeezeIndicesTC5),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_strided_slice_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_strided_slice_test.cpp
index b3936b10db02e8..8df50c44271b17 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_strided_slice_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_strided_slice_test.cpp
@@ -4,6 +4,6 @@
 
 #include "myriad_layers_strided_slice_test.h"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     accuracy, myriadLayersTestsStridedSlice_smoke,
     ::testing::ValuesIn(s_stridedSliceParams));
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_tanh_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_tanh_test.cpp
index 60f816b0842c25..5a7b5be8a46298 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_tanh_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_tanh_test.cpp
@@ -4,11 +4,11 @@
 
 #include "myriad_layers_tanh_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         accuracy, myriadLayersTestsTanh_smoke,
         ::testing::ValuesIn(s_tanhParams));
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerConvolutionWithTanH_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerConvolutionWithTanH_smoke,
         ::testing::Combine(
             ::testing::ValuesIn(g_convolutionTensors)
           , ::testing::Values<param_size>(MAKE_STRUCT(param_size, 3, 3))
@@ -20,21 +20,21 @@ INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerConvolutionWithTanH_smoke,
           )
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsMaxPoolingWithTanh_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsMaxPoolingWithTanh_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_poolingInput),
         ::testing::ValuesIn(g_poolingLayerParamsLite),
         ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsAvgPoolingWithTanh_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsAvgPoolingWithTanh_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_poolingInput),
         ::testing::ValuesIn(g_poolingLayerParamsLite),
         ::testing::ValuesIn(g_poolingLayout))
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     accuracy, myriadLayerFullyConnectedWithTanH_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(g_fcTestParamsSubset),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_tile_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_tile_test.cpp
index ad6d40376a515e..7d68de1b77ba3d 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_tile_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_tile_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_tile_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracyAdd, myriadLayerTestTile_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracyAdd, myriadLayerTestTile_smoke,
                         ::testing::Combine(
                                 ::testing::Values<test_params>(
                                         MAKE_STRUCT(tile_test::nd_tensor_test_params, {4, 5, 6}, 0)
@@ -21,7 +21,7 @@ INSTANTIATE_TEST_CASE_P(accuracyAdd, myriadLayerTestTile_smoke,
                                 , ::testing::Values<tiles>(2, 3, 5)
                         ));
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayerTestTile_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayerTestTile_smoke,
                         ::testing::Combine(
                                 ::testing::Values<test_params>(
                                         MAKE_STRUCT(tile_test::nd_tensor_test_params, {4, 5, 6}, 1)
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_topk_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_topk_test.cpp
index 23a36df4f58559..1cf3edc8642d1a 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_topk_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_topk_test.cpp
@@ -32,7 +32,7 @@ static const std::vector<std::string> s_sorts_list =
 //    "none", // currently is not supported by firmware
 };
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadTestsTopK_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadTestsTopK_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_geometries_list),
         ::testing::ValuesIn(s_modes_list),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_unsqueeze_test.cpp b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_unsqueeze_test.cpp
index 070a99e64051fe..d5384668bf99bf 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_unsqueeze_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/layers/myriad_layers_unsqueeze_test.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_layers_unsqueeze_test.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy, myriadLayersTestsUnsqueeze_smoke,
+INSTANTIATE_TEST_SUITE_P(accuracy, myriadLayersTestsUnsqueeze_smoke,
     ::testing::Combine(
         ::testing::ValuesIn(s_squeezeTensors),
         ::testing::ValuesIn(s_squeezeIndices)
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/myriad_get_output_tests.cpp b/inference-engine/tests_deprecated/functional/vpu/common/myriad_get_output_tests.cpp
index 4d45cc5a9a2795..41540403574d9e 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/myriad_get_output_tests.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/myriad_get_output_tests.cpp
@@ -4,7 +4,7 @@
 
 #include "myriad_get_output_tests.hpp"
 
-INSTANTIATE_TEST_CASE_P(Test_params_pool, myriadGetOutput_nightly,
+INSTANTIATE_TEST_SUITE_P(Test_params_pool, myriadGetOutput_nightly,
         testing::Values(
                 std::make_tuple(std::make_tuple(&full_model, &poolModel), "pool1_3x3_s2"),
                 std::make_tuple(std::make_tuple(&full_model, &convModel), "conv1_7x7_s2"),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_conv_tests.hpp b/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_conv_tests.hpp
index f911e8e3fe7cf7..a8fed6cea07012 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_conv_tests.hpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_conv_tests.hpp
@@ -105,7 +105,7 @@ class MyriadX_HW_Convolution_Tests_nightly
 
 TEST_P(MyriadX_HW_Convolution_Tests_nightly, Single) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddInitialCopyLayer();
@@ -117,7 +117,7 @@ TEST_P(MyriadX_HW_Convolution_Tests_nightly, Single) {
 
 TEST_P(MyriadX_HW_Convolution_Tests_nightly, WithReLU) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddInitialCopyLayer();
@@ -130,7 +130,7 @@ TEST_P(MyriadX_HW_Convolution_Tests_nightly, WithReLU) {
 
 TEST_P(MyriadX_HW_Convolution_Tests_nightly, WithLeakyReLU) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddConvolutionLayer();
@@ -142,7 +142,7 @@ TEST_P(MyriadX_HW_Convolution_Tests_nightly, WithLeakyReLU) {
 
 TEST_P(MyriadX_HW_Convolution_Tests_nightly, WithClamp) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddConvolutionLayer();
@@ -154,7 +154,7 @@ TEST_P(MyriadX_HW_Convolution_Tests_nightly, WithClamp) {
 
 TEST_P(MyriadX_HW_Convolution_Tests_nightly, MultipleInfer) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddConvolutionLayer();
@@ -162,7 +162,7 @@ TEST_P(MyriadX_HW_Convolution_Tests_nightly, MultipleInfer) {
     CompareWithItself(100);
 }
 
-INSTANTIATE_TEST_CASE_P(MaskRcnn101_DILATION, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(MaskRcnn101_DILATION, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 50, 50),
                                                              MAKE_STRUCT(tensor_test_params, 1, 256, 100, 171))
@@ -175,7 +175,7 @@ INSTANTIATE_TEST_CASE_P(MaskRcnn101_DILATION, MyriadX_HW_Convolution_Tests_night
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(kernel_7x7_DILATION, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(kernel_7x7_DILATION, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 90, 90))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 7, 7))
@@ -187,7 +187,7 @@ INSTANTIATE_TEST_CASE_P(kernel_7x7_DILATION, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(Unequal_hw_pad_dilationfactor_DILATION, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(Unequal_hw_pad_dilationfactor_DILATION, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 128, 128))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3),
@@ -201,7 +201,7 @@ INSTANTIATE_TEST_CASE_P(Unequal_hw_pad_dilationfactor_DILATION, MyriadX_HW_Convo
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(Strides_DILATION, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(Strides_DILATION, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 64, 64))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -215,7 +215,7 @@ INSTANTIATE_TEST_CASE_P(Strides_DILATION, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s1p0_extra1, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s1p0_extra1, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 180, 320))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -227,7 +227,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s1p0_extra1, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s1p0_extra2, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s1p0_extra2, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 512, 45, 80))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -239,7 +239,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s1p0_extra2, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s1p0_extra3, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s1p0_extra3, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 45, 80))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -251,7 +251,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s1p0_extra3, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s1p0_extra4, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s1p0_extra4, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 512, 45, 80))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -263,7 +263,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s1p0_extra4, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s2p0_extra1, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s2p0_extra1, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 180, 320))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -275,7 +275,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s2p0_extra1, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s2p0_extra2, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s2p0_extra2, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 128, 90, 160))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -287,7 +287,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s2p0_extra2, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s2p0_extra3, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s2p0_extra3, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 128, 90, 160))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -299,7 +299,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s2p0_extra3, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_extra1, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s2p1_extra1, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 180, 320))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -311,7 +311,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_extra1, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_extra2, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s2p1_extra2, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 128, 90, 160))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -323,7 +323,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_extra2, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_extra3, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s2p1_extra3, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 45, 80))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -335,7 +335,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_extra3, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_extra4, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s2p1_extra4, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 180, 320))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -347,7 +347,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_extra4, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s1p0, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s1p0, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 56, 56))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -359,7 +359,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s1p0, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s1p0_resnet50, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s1p0_resnet50, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 2048, 7, 7))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -372,7 +372,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s1p0_resnet50, MyriadX_HW_Convolution_Tests_nigh
 );
 
 // This case adds extra CopyMakeBorder stage
-INSTANTIATE_TEST_CASE_P(conv_1x1s1p1, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s1p1, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 512, 13, 13))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -384,7 +384,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s1p1, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s2p0, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s2p0, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 56, 56),
                                                              MAKE_STRUCT(tensor_test_params, 1, 512, 28, 28),
@@ -398,7 +398,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s2p0, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s1p1, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s1p1, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 56, 56))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -410,7 +410,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s1p1, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_yolo_tiny_v1_conv1, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s1p1_yolo_tiny_v1_conv1, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 3, 448, 448))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -422,7 +422,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_yolo_tiny_v1_conv1, MyriadX_HW_Convolution_
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_yolo_tiny_v1_conv7, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s1p1_yolo_tiny_v1_conv7, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 512, 7, 7))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -434,7 +434,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_yolo_tiny_v1_conv7, MyriadX_HW_Convolution_
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_yolo_tiny_v1_conv8, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s1p1_yolo_tiny_v1_conv8, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 1024, 7, 7))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -446,7 +446,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_yolo_tiny_v1_conv8, MyriadX_HW_Convolution_
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_vgg, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s1p1_vgg, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 224, 224))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -458,7 +458,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_vgg, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_7x7s2p3, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_7x7s2p3, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 3, 224, 224))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 7, 7))
@@ -472,7 +472,7 @@ INSTANTIATE_TEST_CASE_P(conv_7x7s2p3, MyriadX_HW_Convolution_Tests_nightly,
 
 //  This case for unsymmetric convolution
 
-INSTANTIATE_TEST_CASE_P(conv_3x1s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x1s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 22, 92))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 1))
@@ -484,7 +484,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x1s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x3s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x3s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 22, 92))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 3))
@@ -496,7 +496,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x3s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x5s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x5s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 5, 88))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 5))
@@ -508,7 +508,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x5s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_13x1s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_13x1s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 128, 1, 88))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 13, 1))
@@ -520,7 +520,7 @@ INSTANTIATE_TEST_CASE_P(conv_13x1s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_5x1s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_5x1s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 128, 1, 28))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 5, 1))
@@ -532,7 +532,7 @@ INSTANTIATE_TEST_CASE_P(conv_5x1s1_LPR, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_4x4s2p1, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_4x4s2p1, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 3, 256, 416))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 4, 4))
@@ -544,7 +544,7 @@ INSTANTIATE_TEST_CASE_P(conv_4x4s2p1, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_5x5s2p1, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_5x5s2p1, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 3, 256, 416))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 5, 5))
@@ -556,7 +556,7 @@ INSTANTIATE_TEST_CASE_P(conv_5x5s2p1, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_5x5s2p2, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_5x5s2p2, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 3, 256, 416))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 5, 5))
@@ -568,7 +568,7 @@ INSTANTIATE_TEST_CASE_P(conv_5x5s2p2, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_group1, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s1p1_group1, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 150, 150))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -580,7 +580,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_group1, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_group1, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s2p1_group1, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 150, 150))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -592,7 +592,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_group1, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_group2, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s2p1_group2, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 128, 75, 75))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -604,7 +604,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_group2, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_group3, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s2p1_group3, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 38, 38))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -616,7 +616,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s2p1_group3, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_pva_pvd, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s1p1_pva_pvd, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 6, 208, 368))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -628,7 +628,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_pva_pvd, MyriadX_HW_Convolution_Tests_night
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_1x1s2p0_pva_pvd, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_1x1s2p0_pva_pvd, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 128, 208))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 1, 1))
@@ -640,7 +640,7 @@ INSTANTIATE_TEST_CASE_P(conv_1x1s2p0_pva_pvd, MyriadX_HW_Convolution_Tests_night
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_ssd, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s1p1_ssd, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 128, 75, 75))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -652,7 +652,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s1p1_ssd, MyriadX_HW_Convolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_unequal_hw_pad, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_unequal_hw_pad, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 128, 128))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 5, 5),
@@ -665,7 +665,7 @@ INSTANTIATE_TEST_CASE_P(conv_unequal_hw_pad, MyriadX_HW_Convolution_Tests_nightl
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(conv_3x3s3p1_resnet34, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(conv_3x3s3p1_resnet34, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 512, 75, 75))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -677,7 +677,7 @@ INSTANTIATE_TEST_CASE_P(conv_3x3s3p1_resnet34, MyriadX_HW_Convolution_Tests_nigh
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(fc_to_conv_case, MyriadX_HW_Convolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(fc_to_conv_case, MyriadX_HW_Convolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 56, 350))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 7, 7))
@@ -794,7 +794,7 @@ class MyriadX_HW_ConvPoolMerged_Tests_nightly
 
 TEST_P(MyriadX_HW_ConvPoolMerged_Tests_nightly, WithReLU) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddConvLayer();
@@ -807,7 +807,7 @@ TEST_P(MyriadX_HW_ConvPoolMerged_Tests_nightly, WithReLU) {
 
 TEST_P(MyriadX_HW_ConvPoolMerged_Tests_nightly, WithLeakyReLU) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddConvLayer();
@@ -818,7 +818,7 @@ TEST_P(MyriadX_HW_ConvPoolMerged_Tests_nightly, WithLeakyReLU) {
     CompareWithSW(maxerr);
 }
 
-INSTANTIATE_TEST_CASE_P(yolo_conv1, MyriadX_HW_ConvPoolMerged_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(yolo_conv1, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 3, 448, 448)),
                                 ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3)),
@@ -831,7 +831,7 @@ INSTANTIATE_TEST_CASE_P(yolo_conv1, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(yolov2_tf_conv, MyriadX_HW_ConvPoolMerged_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(yolov2_tf_conv, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 304, 304),
                                                              MAKE_STRUCT(tensor_test_params, 1, 64, 152, 152)),
@@ -845,7 +845,7 @@ INSTANTIATE_TEST_CASE_P(yolov2_tf_conv, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(yolo_conv2, MyriadX_HW_ConvPoolMerged_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(yolo_conv2, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 16, 224, 224)),
                                 ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3)),
@@ -858,7 +858,7 @@ INSTANTIATE_TEST_CASE_P(yolo_conv2, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(yolo_conv4, MyriadX_HW_ConvPoolMerged_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(yolo_conv4, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 56, 56)),
                                 ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3)),
@@ -871,7 +871,7 @@ INSTANTIATE_TEST_CASE_P(yolo_conv4, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(ssd_case1, MyriadX_HW_ConvPoolMerged_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(ssd_case1, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 98, 150)),
                                 ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3)),
@@ -884,7 +884,7 @@ INSTANTIATE_TEST_CASE_P(ssd_case1, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(vgg16_case1, MyriadX_HW_ConvPoolMerged_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(vgg16_case1, MyriadX_HW_ConvPoolMerged_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 512, 28, 28)),
                                 ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3)),
@@ -951,7 +951,7 @@ class MyriadX_HW_ConvTF_Tests_nightly :
 
 TEST_P(MyriadX_HW_ConvTF_Tests_nightly, Single) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddConvolutionLayer();
@@ -960,7 +960,7 @@ TEST_P(MyriadX_HW_ConvTF_Tests_nightly, Single) {
     CompareWithSW(maxerr);
 }
 
-INSTANTIATE_TEST_CASE_P(tf, MyriadX_HW_ConvTF_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(tf, MyriadX_HW_ConvTF_Tests_nightly,
     ::testing::Values(
         std::make_tuple(
             MAKE_STRUCT(tensor_test_params, 1, 3, 224, 224),    // input
@@ -1124,7 +1124,7 @@ class MyriadX_HW_Deconvolution_Tests_nightly
 
 TEST_P(MyriadX_HW_Deconvolution_Tests_nightly, Single) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddInitialCopyLayer();
@@ -1136,7 +1136,7 @@ TEST_P(MyriadX_HW_Deconvolution_Tests_nightly, Single) {
 
 TEST_P(MyriadX_HW_Deconvolution_Tests_nightly, ScaleTests) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddInitialCopyLayer();
@@ -1146,7 +1146,7 @@ TEST_P(MyriadX_HW_Deconvolution_Tests_nightly, ScaleTests) {
     CompareWithSW(maxerr);
 }
 
-INSTANTIATE_TEST_CASE_P(deconv_tf_ssd, MyriadX_HW_Deconvolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(deconv_tf_ssd, MyriadX_HW_Deconvolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 2, 3, 3))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -1157,7 +1157,7 @@ INSTANTIATE_TEST_CASE_P(deconv_tf_ssd, MyriadX_HW_Deconvolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(deconv_3x3_str1, MyriadX_HW_Deconvolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(deconv_3x3_str1, MyriadX_HW_Deconvolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 3, 3),
                                                              MAKE_STRUCT(tensor_test_params, 1, 128, 5, 5),
@@ -1170,7 +1170,7 @@ INSTANTIATE_TEST_CASE_P(deconv_3x3_str1, MyriadX_HW_Deconvolution_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(hw_accuracy_deconv_3x3, MyriadX_HW_Deconvolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(hw_accuracy_deconv_3x3, MyriadX_HW_Deconvolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 5, 5),
                                                              MAKE_STRUCT(tensor_test_params, 1, 128, 11, 11),
@@ -1185,7 +1185,7 @@ INSTANTIATE_TEST_CASE_P(hw_accuracy_deconv_3x3, MyriadX_HW_Deconvolution_Tests_n
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(hw_accuracy_deconv, MyriadX_HW_Deconvolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(hw_accuracy_deconv, MyriadX_HW_Deconvolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 120, 36, 36),
                                                              MAKE_STRUCT(tensor_test_params, 1, 73, 40, 54),
@@ -1200,7 +1200,7 @@ INSTANTIATE_TEST_CASE_P(hw_accuracy_deconv, MyriadX_HW_Deconvolution_Tests_night
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(hw_accuracy_scale_deconv, MyriadX_HW_Deconvolution_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(hw_accuracy_scale_deconv, MyriadX_HW_Deconvolution_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 120, 36, 36))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 5, 5))
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_extra_tests.hpp b/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_extra_tests.hpp
index f7c5b2bca0645d..41a4370feb3709 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_extra_tests.hpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_extra_tests.hpp
@@ -8,7 +8,7 @@
 
 TEST_F(MyriadX_HW_Tests_nightly, SeveralLayers) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     tensor_test_params dims1{1, 3, 224, 224};
@@ -78,7 +78,7 @@ TEST_F(MyriadX_HW_Tests_nightly, SeveralLayers) {
 
 TEST_F(MyriadX_HW_Tests_nightly, LargePoolWithConv) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     tensor_test_params dims1{1, 16, 448, 448};
@@ -139,7 +139,7 @@ TEST_F(MyriadX_HW_Tests_nightly, LargePoolWithConv) {
 
 TEST_F(MyriadX_HW_Tests_nightly, ConvWithPool) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     tensor_test_params dims1{1, 16, 4, 4};
@@ -202,7 +202,7 @@ TEST_F(MyriadX_HW_Tests_nightly, ConvWithPool) {
 
 TEST_F(MyriadX_HW_Tests_nightly, WithConcat) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     const std::string model = R"V0G0N(
@@ -391,7 +391,7 @@ TEST_F(MyriadX_HW_Tests_nightly, WithConcat) {
 
 TEST_F(MyriadX_HW_Tests_nightly, WithConcatMisaligned) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     const std::string model = R"V0G0N(
@@ -548,7 +548,7 @@ TEST_F(MyriadX_HW_Tests_nightly, WithConcatMisaligned) {
 
 TEST_F(MyriadX_HW_Tests_nightly, With_3_FC_Layers) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     const std::string model = R"V0G0N(
@@ -693,7 +693,7 @@ TEST_F(MyriadX_HW_Tests_nightly, With_3_FC_Layers) {
 
 TEST_F(MyriadX_HW_Tests_nightly, WithEltwise) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     const std::string model = R"V0G0N(
@@ -913,7 +913,7 @@ TEST_F(MyriadX_HW_Tests_nightly, WithEltwise) {
 
 TEST_F(MyriadX_HW_Tests_nightly, WithEltwiseReLU) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     const std::string model = R"V0G0N(
@@ -1152,7 +1152,7 @@ TEST_F(MyriadX_HW_Tests_nightly, WithEltwiseReLU) {
 
 TEST_F(MyriadX_HW_Tests_nightly, PermuteFlattenConcat) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     const std::string model = R"V0G0N(
@@ -1378,7 +1378,7 @@ TEST_F(MyriadX_HW_Tests_nightly, PermuteFlattenConcat) {
 
 TEST_F(MyriadX_HW_Tests_nightly, VGG_FirstTwoConvs) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     IN_OUT_desc in_tensor, out_tensor;
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_fc_tests.hpp b/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_fc_tests.hpp
index 5542538947785c..dd8cdaafb54975 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_fc_tests.hpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_fc_tests.hpp
@@ -57,7 +57,7 @@ class MyriadX_HW_FullyConnected_Tests_nightly
 
 TEST_P(MyriadX_HW_FullyConnected_Tests_nightly, Single) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddFCLayer();
@@ -67,11 +67,11 @@ TEST_P(MyriadX_HW_FullyConnected_Tests_nightly, Single) {
 
 TEST_P(MyriadX_HW_FullyConnected_Tests_nightly, Single_NC) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     if (p.in.h != 1 || p.in.w != 1) {
-        SKIP() << "Non NC case";
+        GTEST_SKIP() << "Non NC case";
     }
 
     in_tensor.clear();
@@ -84,7 +84,7 @@ TEST_P(MyriadX_HW_FullyConnected_Tests_nightly, Single_NC) {
 
 TEST_P(MyriadX_HW_FullyConnected_Tests_nightly, WithReLU) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddFCLayer();
@@ -95,7 +95,7 @@ TEST_P(MyriadX_HW_FullyConnected_Tests_nightly, WithReLU) {
 
 TEST_P(MyriadX_HW_FullyConnected_Tests_nightly, MultipleInfer) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddFCLayer();
@@ -103,38 +103,38 @@ TEST_P(MyriadX_HW_FullyConnected_Tests_nightly, MultipleInfer) {
     CompareWithItself(100);
 }
 
-INSTANTIATE_TEST_CASE_P(fc_1024to1000, MyriadX_HW_FullyConnected_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(fc_1024to1000, MyriadX_HW_FullyConnected_Tests_nightly,
                         ::testing::Values(MAKE_STRUCT(fcon_test_params, {1, 1024, 1, 1}, 1000, 0.25f))
 );
 
-INSTANTIATE_TEST_CASE_P(fc_4096to1000, MyriadX_HW_FullyConnected_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(fc_4096to1000, MyriadX_HW_FullyConnected_Tests_nightly,
                         ::testing::Values(MAKE_STRUCT(fcon_test_params, {1, 4096, 1, 1}, 1000, 0.82f))
 );
 
-INSTANTIATE_TEST_CASE_P(fc_4096to4096, MyriadX_HW_FullyConnected_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(fc_4096to4096, MyriadX_HW_FullyConnected_Tests_nightly,
                         ::testing::Values(MAKE_STRUCT(fcon_test_params, {1, 4096, 1, 1}, 4096, 0.9f))
 );
 
-INSTANTIATE_TEST_CASE_P(fc_16x16x16to16, MyriadX_HW_FullyConnected_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(fc_16x16x16to16, MyriadX_HW_FullyConnected_Tests_nightly,
                         ::testing::Values(MAKE_STRUCT(fcon_test_params, {1, 16, 16, 16}, 16, 0.71f))
 );
 
-INSTANTIATE_TEST_CASE_P(fc_512x7x7to4096, MyriadX_HW_FullyConnected_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(fc_512x7x7to4096, MyriadX_HW_FullyConnected_Tests_nightly,
                         ::testing::Values(MAKE_STRUCT(fcon_test_params, {1, 512, 7, 7}, 4096, 4.38f))
 );
 
-INSTANTIATE_TEST_CASE_P(fc_256x7x7to1470, MyriadX_HW_FullyConnected_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(fc_256x7x7to1470, MyriadX_HW_FullyConnected_Tests_nightly,
                         ::testing::Values(MAKE_STRUCT(fcon_test_params, {1, 256, 7, 7}, 1470, 2.375f))
 );
 
-INSTANTIATE_TEST_CASE_P(fc_576to128, MyriadX_HW_FullyConnected_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(fc_576to128, MyriadX_HW_FullyConnected_Tests_nightly,
                         ::testing::Values(MAKE_STRUCT(fcon_test_params, {1, 576, 1, 1}, 128, 0.76f))
 );
 
-INSTANTIATE_TEST_CASE_P(fc_1152to128, MyriadX_HW_FullyConnected_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(fc_1152to128, MyriadX_HW_FullyConnected_Tests_nightly,
                         ::testing::Values(MAKE_STRUCT(fcon_test_params, {1, 1152, 1, 1}, 128, 0.76f))
 );
 
-INSTANTIATE_TEST_CASE_P(fc_batch, MyriadX_HW_FullyConnected_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(fc_batch, MyriadX_HW_FullyConnected_Tests_nightly,
                         ::testing::Values(MAKE_STRUCT(fcon_test_params, {100, 256, 1, 1}, 1024, 0.1f))
 );
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_network_tests.hpp b/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_network_tests.hpp
index 7d91bd5ee22a90..1df7a1307c82c3 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_network_tests.hpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_network_tests.hpp
@@ -58,7 +58,7 @@ class MyriadX_HW_Networks_Tests_nightly :
 
     void RunAsyncTest(int numIters = 20) {
         if (!CheckMyriadX()) {
-            SKIP() << "Non-MyriadX device";
+            GTEST_SKIP() << "Non-MyriadX device";
         }
 
         auto fnPtr = ngraph::builder::subgraph::makeSplitMultiConvConcat();
@@ -145,7 +145,7 @@ inline std::string getTestCaseName(const testing::TestParamInfo<HwNetworkParams>
            std::string((std::get<1>(param.param)).name());
 }
 
-INSTANTIATE_TEST_CASE_P(Input_Output_ExecMode, MyriadX_HW_Networks_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(Input_Output_ExecMode, MyriadX_HW_Networks_Tests_nightly,
     testing::Values(
           std::make_tuple(Precision::FP16, Precision::FP16)
     ),
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_pool_tests.hpp b/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_pool_tests.hpp
index 10f982b6945a65..e4fcf81a66a131 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_pool_tests.hpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/myriad_hw_pool_tests.hpp
@@ -70,7 +70,7 @@ class MyriadX_HW_Pooling_Tests_nightly
 
     void RunSingleTest(const std::string& poolMethod, float tolerance) {
         if (!CheckMyriadX()) {
-            SKIP() << "Non-MyriadX device";
+            GTEST_SKIP() << "Non-MyriadX device";
         }
 
         AddPoolingLayer(poolMethod);
@@ -80,7 +80,7 @@ class MyriadX_HW_Pooling_Tests_nightly
 
     void RunWithReLUTest(const std::string& poolMethod, float tolerance) {
         if (!CheckMyriadX()) {
-            SKIP() << "Non-MyriadX device";
+            GTEST_SKIP() << "Non-MyriadX device";
         }
 
         AddPoolingLayer(poolMethod);
@@ -91,7 +91,7 @@ class MyriadX_HW_Pooling_Tests_nightly
 
     void RunMultipleInferTest(const std::string& poolMethod) {
         if (!CheckMyriadX()) {
-            SKIP() << "Non-MyriadX device";
+            GTEST_SKIP() << "Non-MyriadX device";
         }
 
         AddPoolingLayer(poolMethod);
@@ -108,11 +108,11 @@ TEST_P(MyriadX_HW_Pooling_Tests_nightly, Avg_Single) {
     // this case is not supported by HW
     if (kernel.x == 3 && kernel.y == 3 &&
         stride.x == 2 && stride.y == 2) {
-        SKIP() << "Unsupported case";
+        GTEST_SKIP() << "Unsupported case";
     }
     if ((kernel.x % 2 == 0 || kernel.y % 2 == 0) &&
         (in_dims.w % 2 == 1 || in_dims.h % 2 == 1)) {
-        SKIP() << "Unsupported case";
+        GTEST_SKIP() << "Unsupported case";
     }
 
     RunSingleTest("avg", 0.0015f);
@@ -126,11 +126,11 @@ TEST_P(MyriadX_HW_Pooling_Tests_nightly, Avg_WithReLU) {
     // this case is not supported by HW
     if (kernel.x == 3 && kernel.y == 3 &&
         stride.x == 2 && stride.y == 2) {
-        SKIP() << "Unsupported case";
+        GTEST_SKIP() << "Unsupported case";
     }
     if ((kernel.x % 2 == 0 || kernel.y % 2 == 0) &&
         (in_dims.w % 2 == 1 || in_dims.h % 2 == 1)) {
-        SKIP() << "Unsupported case";
+        GTEST_SKIP() << "Unsupported case";
     }
 
     RunWithReLUTest("avg", 0.0015f);
@@ -144,17 +144,17 @@ TEST_P(MyriadX_HW_Pooling_Tests_nightly, Avg_MultipleInfer) {
     // this case is not supported by HW
     if (kernel.x == 3 && kernel.y == 3 &&
         stride.x == 2 && stride.y == 2) {
-        SKIP() << "Unsupported case";
+        GTEST_SKIP() << "Unsupported case";
     }
     if ((kernel.x % 2 == 0 || kernel.y % 2 == 0) &&
         (in_dims.w % 2 == 1 || in_dims.h % 2 == 1)) {
-        SKIP() << "Unsupported case";
+        GTEST_SKIP() << "Unsupported case";
     }
 
     RunMultipleInferTest("avg");
 }
 
-INSTANTIATE_TEST_CASE_P(pool_2x2s1p0, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_2x2s1p0, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 112, 112))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 2, 2))
@@ -163,7 +163,7 @@ INSTANTIATE_TEST_CASE_P(pool_2x2s1p0, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_2x2s2p0, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_2x2s2p0, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 224, 224),
                                                              MAKE_STRUCT(tensor_test_params, 1, 128, 112, 112),
@@ -176,7 +176,7 @@ INSTANTIATE_TEST_CASE_P(pool_2x2s2p0, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_2x2s2p0_yolo_tiny_v1, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_2x2s2p0_yolo_tiny_v1, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 16, 448, 448))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 2, 2))
@@ -185,7 +185,7 @@ INSTANTIATE_TEST_CASE_P(pool_2x2s2p0_yolo_tiny_v1, MyriadX_HW_Pooling_Tests_nigh
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_3x3s1p0, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_3x3s1p0, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 192, 28, 28),
                                                              MAKE_STRUCT(tensor_test_params, 1, 100, 28, 28))
@@ -195,7 +195,7 @@ INSTANTIATE_TEST_CASE_P(pool_3x3s1p0, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_3x3s1p1, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_3x3s1p1, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 192, 28, 28))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -205,7 +205,7 @@ INSTANTIATE_TEST_CASE_P(pool_3x3s1p1, MyriadX_HW_Pooling_Tests_nightly,
 );
 
 // TODO : 3x3s2p0 HW seems to work only for Max Pooling
-INSTANTIATE_TEST_CASE_P(pool_3x3s2p0, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_3x3s2p0, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 64, 112, 112))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -214,7 +214,7 @@ INSTANTIATE_TEST_CASE_P(pool_3x3s2p0, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_3x3s2p1, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_3x3s2p1, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 576, 7, 7),
                                                              MAKE_STRUCT(tensor_test_params, 1, 16, 35, 35),
@@ -226,7 +226,7 @@ INSTANTIATE_TEST_CASE_P(pool_3x3s2p1, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_7x7s1p0, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_7x7s1p0, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 1024, 7, 7))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 7, 7))
@@ -235,7 +235,7 @@ INSTANTIATE_TEST_CASE_P(pool_7x7s1p0, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_14x14s1p0, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_14x14s1p0, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 1024, 14, 14),
                                                              MAKE_STRUCT(tensor_test_params, 1, 1000, 14, 14))
@@ -245,7 +245,7 @@ INSTANTIATE_TEST_CASE_P(pool_14x14s1p0, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_15x15s1p0, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_15x15s1p0, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 1024, 15, 15),
                                                              MAKE_STRUCT(tensor_test_params, 1, 1000, 15, 15))
@@ -255,7 +255,7 @@ INSTANTIATE_TEST_CASE_P(pool_15x15s1p0, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_2x2s1p1_odd, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_2x2s1p1_odd, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 512, 13, 13))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 2, 2))
@@ -264,7 +264,7 @@ INSTANTIATE_TEST_CASE_P(pool_2x2s1p1_odd, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_2x2s2p0_odd, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_2x2s2p0_odd, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 256, 75, 75),
                                                              MAKE_STRUCT(tensor_test_params, 2, 64, 75, 75),
@@ -276,7 +276,7 @@ INSTANTIATE_TEST_CASE_P(pool_2x2s2p0_odd, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_3x3s1p0_odd, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_3x3s1p0_odd, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 192, 37, 37),
                                                              MAKE_STRUCT(tensor_test_params, 1, 832, 9, 9),
@@ -287,7 +287,7 @@ INSTANTIATE_TEST_CASE_P(pool_3x3s1p0_odd, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_3x3s2p0_odd, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_3x3s2p0_odd, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 96, 93, 93),
                                                              MAKE_STRUCT(tensor_test_params, 1, 512, 23, 23),
@@ -299,7 +299,7 @@ INSTANTIATE_TEST_CASE_P(pool_3x3s2p0_odd, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_3x3s2p0_extra, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_3x3s2p0_extra, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 96, 32, 52))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 3, 3))
@@ -308,7 +308,7 @@ INSTANTIATE_TEST_CASE_P(pool_3x3s2p0_extra, MyriadX_HW_Pooling_Tests_nightly,
                         )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_7x7s7p0_rfcn_batch, MyriadX_HW_Pooling_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_7x7s7p0_rfcn_batch, MyriadX_HW_Pooling_Tests_nightly,
                         ::testing::Combine(
                                 ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 300, 5, 7, 7))
                                 , ::testing::Values<kernel>(MAKE_STRUCT(param_size, 7, 7))
@@ -368,7 +368,7 @@ class MyriadX_HW_PoolTF_Tests_nightly :
 
 TEST_P(MyriadX_HW_PoolTF_Tests_nightly, Single) {
     if (!CheckMyriadX()) {
-        SKIP() << "Non-MyriadX device";
+        GTEST_SKIP() << "Non-MyriadX device";
     }
 
     AddPoolingLayer();
@@ -376,7 +376,7 @@ TEST_P(MyriadX_HW_PoolTF_Tests_nightly, Single) {
     CompareWithSW(0.0f);
 }
 
-INSTANTIATE_TEST_CASE_P(pool_2x2_3x3, MyriadX_HW_PoolTF_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_2x2_3x3, MyriadX_HW_PoolTF_Tests_nightly,
     ::testing::Combine(
         ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 128, 128)),
         ::testing::Values<kernel>(MAKE_STRUCT(param_size, 2, 2),
@@ -391,7 +391,7 @@ INSTANTIATE_TEST_CASE_P(pool_2x2_3x3, MyriadX_HW_PoolTF_Tests_nightly,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(pool4x4, MyriadX_HW_PoolTF_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool4x4, MyriadX_HW_PoolTF_Tests_nightly,
     ::testing::Combine(
         ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 32, 128, 128)),
         ::testing::Values<kernel>(MAKE_STRUCT(param_size, 4, 4)),
@@ -409,7 +409,7 @@ INSTANTIATE_TEST_CASE_P(pool4x4, MyriadX_HW_PoolTF_Tests_nightly,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(pool_with_large_width, MyriadX_HW_PoolTF_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(pool_with_large_width, MyriadX_HW_PoolTF_Tests_nightly,
     ::testing::Combine(
         ::testing::Values<DimsInput>(MAKE_STRUCT(tensor_test_params, 1, 8, 640, 960),
                                      MAKE_STRUCT(tensor_test_params, 1, 64, 6, 1000)),
@@ -422,7 +422,7 @@ INSTANTIATE_TEST_CASE_P(pool_with_large_width, MyriadX_HW_PoolTF_Tests_nightly,
     )
 );
 
-INSTANTIATE_TEST_CASE_P(tf, MyriadX_HW_PoolTF_Tests_nightly,
+INSTANTIATE_TEST_SUITE_P(tf, MyriadX_HW_PoolTF_Tests_nightly,
     ::testing::Values(
         std::make_tuple(
             MAKE_STRUCT(tensor_test_params, 1, 64, 112, 112),   // input
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/myriad_merge_permute_tests.cpp b/inference-engine/tests_deprecated/functional/vpu/common/myriad_merge_permute_tests.cpp
index 0c11efcf6eabe2..becb06c53853ff 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/myriad_merge_permute_tests.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/myriad_merge_permute_tests.cpp
@@ -4,19 +4,19 @@
 
 #include "myriad_merge_permute_tests.hpp"
 
-INSTANTIATE_TEST_CASE_P(accuracy_3D, myriadLayersMergePermuteNDTests_nightly,
+INSTANTIATE_TEST_SUITE_P(accuracy_3D, myriadLayersMergePermuteNDTests_nightly,
         ::testing::Combine(
             ::testing::ValuesIn(s_inTensors_3D)
           , ::testing::ValuesIn(s_permuteParams_3D)
 ));
 
-INSTANTIATE_TEST_CASE_P(accuracy_4D, myriadLayersMergePermuteNDTests_nightly,
+INSTANTIATE_TEST_SUITE_P(accuracy_4D, myriadLayersMergePermuteNDTests_nightly,
         ::testing::Combine(
             ::testing::ValuesIn(s_inTensors_4D)
           , ::testing::ValuesIn(s_permuteParams_4D)
 ));
 
-INSTANTIATE_TEST_CASE_P(accuracy_5D, myriadLayersMergePermuteNDTests_nightly,
+INSTANTIATE_TEST_SUITE_P(accuracy_5D, myriadLayersMergePermuteNDTests_nightly,
         ::testing::Combine(
             ::testing::ValuesIn(s_inTensors_5D)
           , ::testing::ValuesIn(s_permuteParams_5D)
diff --git a/inference-engine/tests_deprecated/functional/vpu/common/regression/helpers/vpu_case_common.hpp b/inference-engine/tests_deprecated/functional/vpu/common/regression/helpers/vpu_case_common.hpp
index 82217999d188f2..b3f3b15b3c7e19 100644
--- a/inference-engine/tests_deprecated/functional/vpu/common/regression/helpers/vpu_case_common.hpp
+++ b/inference-engine/tests_deprecated/functional/vpu/common/regression/helpers/vpu_case_common.hpp
@@ -21,7 +21,7 @@ using namespace InferenceEngine;
 #define DISABLE_IF(expr) \
     do { \
         if (expr) { \
-            SKIP() << "Disabled since " << #expr << std::endl; \
+            GTEST_SKIP() << "Disabled since " << #expr << std::endl; \
         } \
     }while(false)
 
@@ -33,7 +33,7 @@ using namespace InferenceEngine;
 #endif
 
 #if defined(__arm__) || defined(_M_ARM) || defined(__aarch64__) || defined(_M_ARM64)
-#   define DISABLE_ON_ARM      SKIP() << "Disabled on ARM" << std::endl;
+#   define DISABLE_ON_ARM      GTEST_SKIP() << "Disabled on ARM" << std::endl;
 #   define VPU_REG_TEST_ARM_PLATFORM
 #else
 #   define DISABLE_ON_ARM
@@ -42,9 +42,9 @@ using namespace InferenceEngine;
 #define ENABLE_IF_MA2085 \
     do { \
         if (!CheckMA2085()) { \
-        SKIP() << "Disabled since not on MA2085" << std::endl; \
+            GTEST_SKIP() << "Disabled since not on MA2085" << std::endl; \
         }\
-    }while(false)
+    } while(false)
 
 extern bool CheckMyriadX();
 extern bool CheckMA2085();
diff --git a/inference-engine/tests_deprecated/functional/vpu/graph_transformer/merge_permute_and_reorder_test.cpp b/inference-engine/tests_deprecated/functional/vpu/graph_transformer/merge_permute_and_reorder_test.cpp
index ec6332dcc98231..f6105b9979cb58 100644
--- a/inference-engine/tests_deprecated/functional/vpu/graph_transformer/merge_permute_and_reorder_test.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/graph_transformer/merge_permute_and_reorder_test.cpp
@@ -147,13 +147,13 @@ TEST_P(myriadGTMergePermuteNDTests_nightly, Permute) {
     std::cout << "Myriad time = non-optimized: " << executionMicroseconds << " us., optimized: " << executionMicrosecondsOptimized << " us.\n";
 }
 
-INSTANTIATE_TEST_CASE_P(accuracy_3D, myriadGTMergePermuteNDTests_nightly,
+INSTANTIATE_TEST_SUITE_P(accuracy_3D, myriadGTMergePermuteNDTests_nightly,
         ::testing::Combine(
             ::testing::ValuesIn(s_inTensors_3D)
           , ::testing::ValuesIn(s_permuteParams_3D)
 ));
 
-INSTANTIATE_TEST_CASE_P(accuracy_5D, myriadGTMergePermuteNDTests_nightly,
+INSTANTIATE_TEST_SUITE_P(accuracy_5D, myriadGTMergePermuteNDTests_nightly,
         ::testing::Combine(
             ::testing::ValuesIn(s_inTensors_5D)
           , ::testing::ValuesIn(s_permuteParams_5D)
diff --git a/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_configs_tests.cpp b/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_configs_tests.cpp
index 074ed1a78ef2cb..8fc795322c815c 100644
--- a/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_configs_tests.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_configs_tests.cpp
@@ -100,14 +100,14 @@ static const std::vector<config_t> myriadIncorrectPackageTypeConfigValues = {
     {{VPU_MYRIAD_CONFIG_KEY(MOVIDIUS_DDR_TYPE), "-MICRON_1GB"}},
 };
 
-INSTANTIATE_TEST_CASE_P(MyriadConfigs, myriadCorrectModelsConfigsTests_nightly,
+INSTANTIATE_TEST_SUITE_P(MyriadConfigs, myriadCorrectModelsConfigsTests_nightly,
                         ::testing::ValuesIn(myriadCorrectPlatformConfigValues));
 
-INSTANTIATE_TEST_CASE_P(MyriadConfigs, myriadIncorrectModelsConfigsTests_nightly,
+INSTANTIATE_TEST_SUITE_P(MyriadConfigs, myriadIncorrectModelsConfigsTests_nightly,
                         ::testing::ValuesIn(myriadIncorrectPlatformConfigValues));
 
-INSTANTIATE_TEST_CASE_P(MyriadPackageConfigs, myriadCorrectModelsConfigsTests_nightly,
+INSTANTIATE_TEST_SUITE_P(MyriadPackageConfigs, myriadCorrectModelsConfigsTests_nightly,
     ::testing::ValuesIn(myriadCorrectPackageTypeConfigValues));
 
-INSTANTIATE_TEST_CASE_P(MyriadPackageConfigs, myriadIncorrectModelsConfigsTests_nightly,
+INSTANTIATE_TEST_SUITE_P(MyriadPackageConfigs, myriadIncorrectModelsConfigsTests_nightly,
     ::testing::ValuesIn(myriadIncorrectPackageTypeConfigValues));
diff --git a/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_multiple_graph_tests.cpp b/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_multiple_graph_tests.cpp
index 19a626a90bb2bd..59dee4b72808eb 100644
--- a/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_multiple_graph_tests.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_multiple_graph_tests.cpp
@@ -23,6 +23,6 @@ TEST_P(myriadMultipleGraphsTests_nightly, LoadGraphsOnDevice) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(numerOfGraphs, myriadMultipleGraphsTests_nightly,
+INSTANTIATE_TEST_SUITE_P(numerOfGraphs, myriadMultipleGraphsTests_nightly,
     ::testing::Values(2, 4, 10)
 );
diff --git a/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_streams_configuration_tests.cpp b/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_streams_configuration_tests.cpp
index a6d99e2367a3ea..93f4b4cc8d01f6 100644
--- a/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_streams_configuration_tests.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/myriad_tests/myriad_streams_configuration_tests.cpp
@@ -24,6 +24,6 @@ TEST_P(myriadCorrectStreamsConfiguration_nightly, InfersWithConfiguredStreams) {
     ASSERT_TRUE(Infer());
 }
 
-INSTANTIATE_TEST_CASE_P(StreamsConfiguration, myriadCorrectStreamsConfiguration_nightly, testing::Values(1, 2, 3));
+INSTANTIATE_TEST_SUITE_P(StreamsConfiguration, myriadCorrectStreamsConfiguration_nightly, testing::Values(1, 2, 3));
 
 }
diff --git a/inference-engine/tests_deprecated/functional/vpu/shared_tests_instance/common_single_layer_tests/single_layer_tests.cpp b/inference-engine/tests_deprecated/functional/vpu/shared_tests_instance/common_single_layer_tests/single_layer_tests.cpp
index 0b59f755bb1b80..6587fb3eaf96c2 100644
--- a/inference-engine/tests_deprecated/functional/vpu/shared_tests_instance/common_single_layer_tests/single_layer_tests.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/shared_tests_instance/common_single_layer_tests/single_layer_tests.cpp
@@ -44,7 +44,7 @@ getTestCaseName(testing::TestParamInfo<std::tuple<InitialShapes, NewShapes, Plug
     return device2FilterName[pluginParams.deviceName] + helper->getType();
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         // TODO: rewrite to ngraph to have reshape functionality
         DISABLED_Conv_nightly, CommonSingleLayerTest,
         ::testing::Combine(
@@ -61,7 +61,7 @@ INSTANTIATE_TEST_CASE_P(
 ), getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         // TODO: rewrite to ngraph to have reshape functionality
         DISABLED_Deconv_nightly, CommonSingleLayerTest,
         ::testing::Combine(
@@ -78,7 +78,7 @@ INSTANTIATE_TEST_CASE_P(
 ), getTestCaseName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         // TODO: rewrite to ngraph to have reshape functionality
         DISABLED_Pool_nightly, CommonSingleLayerTest,
         ::testing::Combine(
diff --git a/inference-engine/tests_deprecated/functional/vpu/shared_tests_instance/input_tests/parser_tests.cpp b/inference-engine/tests_deprecated/functional/vpu/shared_tests_instance/input_tests/parser_tests.cpp
index 616e2214011bed..14e48bcd2e4bf6 100644
--- a/inference-engine/tests_deprecated/functional/vpu/shared_tests_instance/input_tests/parser_tests.cpp
+++ b/inference-engine/tests_deprecated/functional/vpu/shared_tests_instance/input_tests/parser_tests.cpp
@@ -30,6 +30,6 @@ ir_test_params ir_test_cases[] = {
     ir_test_params("MYRIAD", "FP16", negative_norm_k_case)
 };
 
-INSTANTIATE_TEST_CASE_P(FunctionalTest_nightly, IncorrectIRTests,
+INSTANTIATE_TEST_SUITE_P(FunctionalTest_nightly, IncorrectIRTests,
         ::testing::ValuesIn(ir_test_cases),
         getTestName);
diff --git a/inference-engine/tests_deprecated/functional/vpu/vpu_base/vpu_tests_config.hpp b/inference-engine/tests_deprecated/functional/vpu/vpu_base/vpu_tests_config.hpp
index ab402644b3f3e0..bddbf9dd20efec 100644
--- a/inference-engine/tests_deprecated/functional/vpu/vpu_base/vpu_tests_config.hpp
+++ b/inference-engine/tests_deprecated/functional/vpu/vpu_base/vpu_tests_config.hpp
@@ -25,7 +25,7 @@ bool deviceForceReset();
 // IE macro forcing gave us no ability to pass device name as variable.
 // So we create this two replacements to PLUGING_CASE_WITH_SUFFIX.
 #define VPU_PLUGING_CASE_WITH_SUFFIX(_suffix, _test, _params) \
-    INSTANTIATE_TEST_CASE_P(VPU_run##_suffix, _test, ::testing::Combine(::testing::Values(::vpu::tests::deviceName()), _params) )
+    INSTANTIATE_TEST_SUITE_P(VPU_run##_suffix, _test, ::testing::Combine(::testing::Values(::vpu::tests::deviceName()), _params) )
 
 #define DISABLED_VPU_PLUGING_CASE_WITH_SUFFIX(_suffix, _test, _params) \
-    INSTANTIATE_TEST_CASE_P(DISABLED_VPU_run##_suffix, _test, ::testing::Combine(::testing::Values(::vpu::tests::deviceName()), _params) )
+    INSTANTIATE_TEST_SUITE_P(DISABLED_VPU_run##_suffix, _test, ::testing::Combine(::testing::Values(::vpu::tests::deviceName()), _params) )
diff --git a/inference-engine/tests_deprecated/helpers/tests_vpu_common.hpp b/inference-engine/tests_deprecated/helpers/tests_vpu_common.hpp
index 5cfc76f825b486..18d17be9337dfc 100644
--- a/inference-engine/tests_deprecated/helpers/tests_vpu_common.hpp
+++ b/inference-engine/tests_deprecated/helpers/tests_vpu_common.hpp
@@ -24,7 +24,7 @@ static constexpr char ENV_HDDL_R[]  = "IE_VPU_ENABLE_PER_LAYER_TESTS_HDDL";
 #define DISABLE_IF(expression)                                   \
 {                                                                \
     if (expression) {                                            \
-        SKIP() << "Disabled since " << #expression << std::endl; \
+        GTEST_SKIP() << "Disabled since " << #expression << std::endl; \
     }                                                            \
 }
 
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/fp32_non_quantized_tests.cpp b/inference-engine/tests_deprecated/unit/engines/gna/fp32_non_quantized_tests.cpp
index e321c50a77c9f7..3570511851511d 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/fp32_non_quantized_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/fp32_non_quantized_tests.cpp
@@ -66,7 +66,7 @@ FP32TestParams gna_fp32_test_params[] = {
     {32, FP32TestParams::eSumm}
 };
 
-INSTANTIATE_TEST_CASE_P(GNAFP32Tests, GNAFP32ParametricTest,
+INSTANTIATE_TEST_SUITE_P(GNAFP32Tests, GNAFP32ParametricTest,
     ::testing::ValuesIn(gna_fp32_test_params), getTestName);
 
 TEST_F(FP32NonQuantizedTest, SplitFollowedByFCAndEltwiseOnCPU) {
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/gna_input_precision_test.cpp b/inference-engine/tests_deprecated/unit/engines/gna/gna_input_precision_test.cpp
index 34baad4349b68d..0efbeb83a72dd5 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/gna_input_precision_test.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/gna_input_precision_test.cpp
@@ -3,7 +3,7 @@
 //
 
 #include <gtest/gtest.h>
-#include <gmock/gmock-generated-actions.h>
+#include <gmock/gmock.h>
 #include "gna_matcher.hpp"
 #include "matchers/input_data_matcher.hpp"
 #include "test_irs.hpp"
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.cpp b/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.cpp
index 70f8c2d7ef0897..f6d611faffe9a9 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.cpp
@@ -17,7 +17,7 @@
 #include "matchers/pool_matcher.hpp"
 #include "matchers/fill_with_data.hpp"
 #include "matchers/weights_matcher.hpp"
-#include <gmock/gmock-generated-actions.h>
+#include <gmock/gmock.h>
 #include <debug.h>
 
 #include <gmock/gmock-more-actions.h>
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/gna_proc_type_test.cpp b/inference-engine/tests_deprecated/unit/engines/gna/gna_proc_type_test.cpp
index 8eed7282212857..0f8102468226ed 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/gna_proc_type_test.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/gna_proc_type_test.cpp
@@ -5,7 +5,7 @@
 #include <vector>
 
 #include <gtest/gtest.h>
-#include <gmock/gmock-generated-actions.h>
+#include <gmock/gmock.h>
 #include <gna/gna_config.hpp>
 #include "gna_plugin.hpp"
 #include "gna_mock_api.hpp"
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/layers/activation_test.cpp b/inference-engine/tests_deprecated/unit/engines/gna/layers/activation_test.cpp
index 34b3c20761eff2..0d768f823a1c1c 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/layers/activation_test.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/layers/activation_test.cpp
@@ -89,7 +89,7 @@ static const ActivationCaseParam gna_activation_test_params[] = {
     {"softsign", 200, {-10, 10}},
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     GNALayerTests, GNAActivationTest,
     ::testing::Combine(
         ::testing::Values(InferenceEngine::Precision::FP32, InferenceEngine::Precision::I16, InferenceEngine::Precision::I8),
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_align_filter2_tests.cpp b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_align_filter2_tests.cpp
index 71db4e90376bb8..75fdc98b934934 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_align_filter2_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_align_filter2_tests.cpp
@@ -170,7 +170,7 @@ TEST_P(GNAAlignFilterTest, concatWith_2_Inputs_accurate) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     GNALayerTests,
     GNAAlignFilterTest,
     testing::Combine(
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_conv1d_test.cpp b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_conv1d_test.cpp
index 310cf46d0f165e..9761edd803b793 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_conv1d_test.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_conv1d_test.cpp
@@ -100,7 +100,7 @@ TEST_P(GNAConv1DTest, SplitToConcatWith2Inputs) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         GNALayerTests,
         GNAConv1DTest,
         testing::Combine(
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_eltwise_test.cpp b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_eltwise_test.cpp
index 28c26e21af1924..c1e2ea788c3aa3 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_eltwise_test.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_eltwise_test.cpp
@@ -106,7 +106,7 @@ TEST_P(GNAEltwiseTest, FourBytesInputsViaReshape) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     GNALayerTests,
     GNAEltwiseTest,
     ::testing::Combine(
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_multi_input_to_concat_test.cpp b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_multi_input_to_concat_test.cpp
index 068b0f681e91b9..dac88c92cc7f62 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_multi_input_to_concat_test.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_multi_input_to_concat_test.cpp
@@ -106,7 +106,7 @@ TEST_P(GNAMultiInputToConcatTest, InputsToConcat) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         GNALayerTests,
         GNAMultiInputToConcatTest,
         ::testing::Combine(
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_permute_test.cpp b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_permute_test.cpp
index 3c0c7ffe33e386..8392de59122978 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_permute_test.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_permute_test.cpp
@@ -71,5 +71,5 @@ const Permute3dimTestParam gna_permute3d_test_params[] = {
         {{{2, 1, 0}, {2, 1, 4}}, false}
 };
 
-INSTANTIATE_TEST_CASE_P(GNALayerTests, GNAPermute3dTest,
+INSTANTIATE_TEST_SUITE_P(GNALayerTests, GNAPermute3dTest,
         ::testing::ValuesIn(gna_permute3d_test_params), getPermute3dTestName);
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_split_to_concat_test.cpp b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_split_to_concat_test.cpp
index 9ac1011ca6cb54..7e429f9edca85f 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_split_to_concat_test.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_split_to_concat_test.cpp
@@ -84,7 +84,7 @@ TEST_P(GNASplitToConcatTest, SplitToConcatWith2Inputs) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     GNALayerTests,
     GNASplitToConcatTest,
     testing::Combine(
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_squeeze_test.cpp b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_squeeze_test.cpp
index 603efb6578e9cd..ce015857859bbf 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_squeeze_test.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_squeeze_test.cpp
@@ -147,7 +147,7 @@ static const SqueezeCaseParam gna_squeeze_test_params[] = {
         {{1, 1, 1, 1, 1, 3}, {1, 3}}
 };
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     GNALayerTests, GNASqueezeTest,
     ::testing::Combine(
         ::testing::Values(InferenceEngine::Precision::FP32, InferenceEngine::Precision::I16),
@@ -155,7 +155,7 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::ValuesIn(gna_squeeze_test_params)),
         GNAUnsqueezeTest::getTestName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     GNALayerTests, GNAUnsqueezeTest,
     ::testing::Combine(
         ::testing::Values(InferenceEngine::Precision::FP32, InferenceEngine::Precision::I16),
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/merge_parallel_fc.cpp b/inference-engine/tests_deprecated/unit/engines/vpu/merge_parallel_fc.cpp
index 3ab1ab9dec0e76..45da2af742c9cc 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/merge_parallel_fc.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/merge_parallel_fc.cpp
@@ -346,12 +346,12 @@ TEST_P(VPU_NoMergeParallelFCTest, NoMergeParallelFCMergeCases) {
     AssertNotMergedCorrectly(testInputDimensions, testFCDescriptors);
 }
 
-INSTANTIATE_TEST_CASE_P(MergeParallelFCTest, VPU_MergeParallelFCTest, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(MergeParallelFCTest, VPU_MergeParallelFCTest, testing::Combine(
     testing::ValuesIn(inputDimensions),
     testing::ValuesIn(FCDescriptorsMerge)
 ));
 
-INSTANTIATE_TEST_CASE_P(NoMergeParallelFCTest, VPU_NoMergeParallelFCTest, testing::Combine(
+INSTANTIATE_TEST_SUITE_P(NoMergeParallelFCTest, VPU_NoMergeParallelFCTest, testing::Combine(
     testing::ValuesIn(inputDimensions),
     testing::ValuesIn(FCDescriptorsNoMerge)
 ));
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/pthread_semaphore_tests.cpp b/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/pthread_semaphore_tests.cpp
index 8a2f6282aa4230..6f6f851ae4ecbc 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/pthread_semaphore_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/pthread_semaphore_tests.cpp
@@ -366,7 +366,7 @@ TEST_P(PThreadBinSemaphoreTest, PostWakeUpOnlyOneWaiterOfMany) {
     th3.join();
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     PThreadParametrizedTests,
     PThreadBinSemaphoreTest,
     ::testing::Values(
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/watchdog_tests.cpp b/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/watchdog_tests.cpp
index a4c2f50aa8b20c..8ae4110fc1a1ea 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/watchdog_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/watchdog_tests.cpp
@@ -17,10 +17,10 @@ using ms = std::chrono::milliseconds;
 class MockWatchdogDevice : public Watchdog::IDevice {
  public:
     using time_point  = Watchdog::IDevice::time_point;
-    MOCK_QUALIFIED_METHOD1(keepAlive, noexcept, void(const time_point &));
-    MOCK_QUALIFIED_METHOD1(dueIn, const noexcept, std::chrono::milliseconds (const time_point &current_time));
-    MOCK_QUALIFIED_METHOD0(isTimeout, const noexcept, bool ());
-    MOCK_QUALIFIED_METHOD0(getHandle, const noexcept, void* ());
+    MOCK_METHOD(void, keepAlive, (const time_point &), (noexcept));
+    MOCK_METHOD(std::chrono::milliseconds, dueIn, (const time_point &current_time), (const, noexcept));
+    MOCK_METHOD(bool, isTimeout, (), (const, noexcept));
+    MOCK_METHOD(void *, getHandle, (), (const, noexcept));
 };
 
 class MVNCWatchdogTests: public TestsCommon {
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/xlink_device_tests.cpp b/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/xlink_device_tests.cpp
index 555caa04a95456..400c84e1cd0055 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/xlink_device_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/mvnc/xlink_device_tests.cpp
@@ -34,6 +34,6 @@ TEST_P(XLinkDeviceTestsWithParam, shouldNotCreateXlinkDeviceWithInvalidInterval)
     xlink_device_destroy(deviceHndl);
 }
 
-INSTANTIATE_TEST_CASE_P(WatchdogDevice,
+INSTANTIATE_TEST_SUITE_P(WatchdogDevice,
     XLinkDeviceTestsWithParam,
     testing::Values(0, -1, -WATCHDOG_MAX_PING_INTERVAL_MS));
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/helpers/myriad_mvnc_stub.h b/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/helpers/myriad_mvnc_stub.h
index 2a52228fe938c4..b9541c789593cb 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/helpers/myriad_mvnc_stub.h
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/helpers/myriad_mvnc_stub.h
@@ -17,8 +17,8 @@ using namespace vpu::MyriadPlugin;
 class MvncStub : public IMvnc {
 public:
     //Operations
-    MOCK_QUALIFIED_METHOD0(AvailableDevicesNames, const, std::vector<std::string>());
-    MOCK_QUALIFIED_METHOD0(AvailableDevicesDesc, const, std::vector<ncDeviceDescr_t>());
+    MOCK_CONST_METHOD0(AvailableDevicesNames, std::vector<std::string>());
+    MOCK_CONST_METHOD0(AvailableDevicesDesc, std::vector<ncDeviceDescr_t>());
 
     MOCK_METHOD0(watchdogHndl, WatchdogHndl_t*());
 
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/myriad_engine_tests.cpp b/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/myriad_engine_tests.cpp
index ab437ab8774455..5fc90e2e678241 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/myriad_engine_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/myriad_engine_tests.cpp
@@ -104,32 +104,32 @@ static const std::vector<config_t> myriadIncorrectPackageTypeConfigValues = {
 IE_SUPPRESS_DEPRECATED_END
 
 /// Platform
-INSTANTIATE_TEST_CASE_P(MyriadPlatformConfigs, MyriadEngineSetCorrectConfigTest,
+INSTANTIATE_TEST_SUITE_P(MyriadPlatformConfigs, MyriadEngineSetCorrectConfigTest,
                         ::testing::ValuesIn(myriadCorrectPlatformConfigValues));
 
-INSTANTIATE_TEST_CASE_P(MyriadPlatformConfigs, MyriadEngineSetIncorrectConfigTest,
+INSTANTIATE_TEST_SUITE_P(MyriadPlatformConfigs, MyriadEngineSetIncorrectConfigTest,
                         ::testing::ValuesIn(myriadIncorrectPlatformConfigValues));
 
 /// Protocol
-INSTANTIATE_TEST_CASE_P(MyriadProtocolConfigs, MyriadEngineSetCorrectConfigTest,
+INSTANTIATE_TEST_SUITE_P(MyriadProtocolConfigs, MyriadEngineSetCorrectConfigTest,
                         ::testing::ValuesIn(myriadCorrectProtocolConfigValues));
 
-INSTANTIATE_TEST_CASE_P(MyriadProtocolConfigs, MyriadEngineSetIncorrectConfigTest,
+INSTANTIATE_TEST_SUITE_P(MyriadProtocolConfigs, MyriadEngineSetIncorrectConfigTest,
                         ::testing::ValuesIn(myriadIncorrectProtocolConfigValues));
 
 /// Config combinations
-INSTANTIATE_TEST_CASE_P(MyriadConfigOptionsCombination, MyriadEngineSetCorrectConfigTest,
+INSTANTIATE_TEST_SUITE_P(MyriadConfigOptionsCombination, MyriadEngineSetCorrectConfigTest,
                         ::testing::ValuesIn(myriadCorrectConfigCombinationValues));
 
 /// Power Config
-INSTANTIATE_TEST_CASE_P(MyriadPowerConfigs, MyriadEngineSetCorrectConfigTest,
+INSTANTIATE_TEST_SUITE_P(MyriadPowerConfigs, MyriadEngineSetCorrectConfigTest,
                         ::testing::ValuesIn(myriadCorrectPowerConfigValues));
 
-INSTANTIATE_TEST_CASE_P(MyriadPowerConfigs, MyriadEngineSetIncorrectConfigTest,
+INSTANTIATE_TEST_SUITE_P(MyriadPowerConfigs, MyriadEngineSetIncorrectConfigTest,
                         ::testing::ValuesIn(myriadIncorrectPowerConfigValues));
 /// Package Config
-INSTANTIATE_TEST_CASE_P(MyriadPackageConfigs, MyriadEngineSetCorrectConfigTest,
+INSTANTIATE_TEST_SUITE_P(MyriadPackageConfigs, MyriadEngineSetCorrectConfigTest,
                         ::testing::ValuesIn(myriadCorrectPackageTypeConfigValues));
 
-INSTANTIATE_TEST_CASE_P(MyriadPackageConfigs, MyriadEngineSetIncorrectConfigTest,
+INSTANTIATE_TEST_SUITE_P(MyriadPackageConfigs, MyriadEngineSetIncorrectConfigTest,
                         ::testing::ValuesIn(myriadIncorrectPackageTypeConfigValues));
diff --git a/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/myriad_metrics_tests.cpp b/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/myriad_metrics_tests.cpp
index 55879599166831..2c27aea933f142 100644
--- a/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/myriad_metrics_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/vpu/myriad_tests/myriad_metrics_tests.cpp
@@ -193,14 +193,14 @@ TEST_P(MyriadRangeInferMetricsTestWithParam, CheckValues) {
     ASSERT_TRUE(act_res == exp_range_);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     MetricsTest,
     MyriadGetMetricsTestCaseWithParam,
     ::testing::Values(std::vector<std::string> {},
                       std::vector<std::string> {"2.1-ma2480"},
                       std::vector<std::string> {"2.1-ma2480", "3.1-ma2085"}));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     MetricsTest,
     MyriadDeviceMetricsTestWithParam,
     Combine(::testing::Values(std::vector<std::string> {},
@@ -210,7 +210,7 @@ INSTANTIATE_TEST_CASE_P(
                                 std::vector<std::string> {"1.4-ma2455"},
                                 std::vector<std::string> {"1.1-ma2080", "3.3-ma2085"})));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     MetricsTest,
     MyriadRangeInferMetricsTestWithParam,
     ::testing::Values(std::tuple<range_type, std::string>(range_type(3,6,1), "-1"),
diff --git a/inference-engine/tests_deprecated/unit/graph_tools/graph_copy_tests.cpp b/inference-engine/tests_deprecated/unit/graph_tools/graph_copy_tests.cpp
index c9a1e8bb2187eb..3dfe1293624717 100644
--- a/inference-engine/tests_deprecated/unit/graph_tools/graph_copy_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/graph_tools/graph_copy_tests.cpp
@@ -7,9 +7,7 @@
 #include <common_test_utils/test_assertions.hpp>
 #include <common_test_utils/common_utils.hpp>
 #include <unordered_set>
-#include <gmock/gmock-generated-function-mockers.h>
-#include <gmock/gmock-generated-matchers.h>
-#include <gmock/gmock-more-actions.h>
+#include <gmock/gmock.h>
 #include "ie_common.h"
 #include <ie_core.hpp>
 #include "graph_test_base.hpp"
diff --git a/inference-engine/tests_deprecated/unit/graph_tools/graph_tools_test.cpp b/inference-engine/tests_deprecated/unit/graph_tools/graph_tools_test.cpp
index 71b1f6ec46d2b1..f96bf8e8abd75f 100644
--- a/inference-engine/tests_deprecated/unit/graph_tools/graph_tools_test.cpp
+++ b/inference-engine/tests_deprecated/unit/graph_tools/graph_tools_test.cpp
@@ -7,9 +7,7 @@
 #include <gna_graph_tools.hpp>
 #include "graph_test_base.hpp"
 #include <unordered_set>
-#include <gmock/gmock-generated-function-mockers.h>
-#include <gmock/gmock-generated-matchers.h>
-#include <gmock/gmock-more-actions.h>
+#include <gmock/gmock.h>
 #include "ie_common.h"
 #include <memory>
 #include <legacy/details/ie_cnn_network_iterator.hpp>
diff --git a/inference-engine/tests_deprecated/unit/inference_engine_tests/get_num_iterations_test.cpp b/inference-engine/tests_deprecated/unit/inference_engine_tests/get_num_iterations_test.cpp
index 01fb9f61163ba5..830cab7cc4cc88 100644
--- a/inference-engine/tests_deprecated/unit/inference_engine_tests/get_num_iterations_test.cpp
+++ b/inference-engine/tests_deprecated/unit/inference_engine_tests/get_num_iterations_test.cpp
@@ -147,7 +147,7 @@ using GetNumIterationsInvalidAxisTests = GetNumIterationsNegativeTests;
 TEST_P(GetNumIterationsInvalidAxisTests, InvalidAxisThrowsAnException) {
     ASSERT_ANY_THROW(Run());
 }
-INSTANTIATE_TEST_CASE_P(NumIterationsTest, GetNumIterationsInvalidAxisTests, testing::Values<NegativeTestParams>(
+INSTANTIATE_TEST_SUITE_P(NumIterationsTest, GetNumIterationsInvalidAxisTests, testing::Values<NegativeTestParams>(
     NegativeTestParams{{{1}}, {{From{0}, Axis{-2}, Start{0}, Stride{1}, End{1}}}, {}, {}},
     NegativeTestParams{{{1}}, {{From{0}, Axis{10}, Start{0}, Stride{1}, End{1}}}, {}, {}},
     NegativeTestParams{{{1}, {1, 2}}, {{From{0}, Axis{-2}, Start{0}, Stride{1}, End{1}}}, {}, {}},
@@ -177,7 +177,7 @@ using GetNumIterationsInvalidStartTests = GetNumIterationsNegativeTests;
 TEST_P(GetNumIterationsInvalidStartTests, InvalidStartThrowsAnException) {
     ASSERT_ANY_THROW(Run());
 }
-INSTANTIATE_TEST_CASE_P(NumIterationsTest, GetNumIterationsInvalidStartTests, testing::Values<NegativeTestParams>(
+INSTANTIATE_TEST_SUITE_P(NumIterationsTest, GetNumIterationsInvalidStartTests, testing::Values<NegativeTestParams>(
     NegativeTestParams{{{1}}, {{From{0}, Axis{0}, Start{2}, Stride{1}, End{1}}}, {}, {}},
     NegativeTestParams{{{1}, {1, 2}}, {{From{0}, Axis{0}, Start{2}, Stride{1}, End{1}}}, {}, {}},
     NegativeTestParams{{{1}, {1, 2}}, {{From{1}, Axis{0}, Start{2}, Stride{1}, End{1}}}, {}, {}},
@@ -193,7 +193,7 @@ using GetNumIterationsInvalidEndTests = GetNumIterationsNegativeTests;
 TEST_P(GetNumIterationsInvalidEndTests, InvalidEndThrowsAnException) {
     ASSERT_ANY_THROW(Run());
 }
-INSTANTIATE_TEST_CASE_P(NumIterationsTest, GetNumIterationsInvalidEndTests, testing::Values<NegativeTestParams>(
+INSTANTIATE_TEST_SUITE_P(NumIterationsTest, GetNumIterationsInvalidEndTests, testing::Values<NegativeTestParams>(
     NegativeTestParams{{}, {}, {{1}}, {{From{0}, Axis{0}, Start{0}, Stride{1}, End{2}}}},
     NegativeTestParams{{}, {}, {{1}, {1, 2}}, {{From{0}, Axis{0}, Start{0}, Stride{1}, End{2}}}},
     NegativeTestParams{{}, {}, {{1}, {1, 2}}, {{From{1}, Axis{0}, Start{0}, Stride{1}, End{2}}}}
@@ -203,7 +203,7 @@ using GetNumIterationsInvalidStrideTests = GetNumIterationsNegativeTests;
 TEST_P(GetNumIterationsInvalidStrideTests, InvalidStrideThrowsAnException) {
     ASSERT_ANY_THROW(Run());
 }
-INSTANTIATE_TEST_CASE_P(NumIterationsTest, GetNumIterationsInvalidStrideTests, testing::Values<NegativeTestParams>(
+INSTANTIATE_TEST_SUITE_P(NumIterationsTest, GetNumIterationsInvalidStrideTests, testing::Values<NegativeTestParams>(
     NegativeTestParams{{{1}}, {{From{0}, Axis{0}, Start{0}, Stride{0}, End{1}}}, {}, {}},
     NegativeTestParams{{}, {}, {{1}}, {{From{0}, Axis{0}, Start{0}, Stride{0}, End{1}}}}
 ));
@@ -212,7 +212,7 @@ using GetNumIterationsInvalidFromTests = GetNumIterationsNegativeTests;
 TEST_P(GetNumIterationsInvalidFromTests, InvalidFromThrowsAnException) {
     ASSERT_ANY_THROW(Run());
 }
-INSTANTIATE_TEST_CASE_P(NumIterationsTest, GetNumIterationsInvalidFromTests, testing::Values<NegativeTestParams>(
+INSTANTIATE_TEST_SUITE_P(NumIterationsTest, GetNumIterationsInvalidFromTests, testing::Values<NegativeTestParams>(
     NegativeTestParams{{{1}}, {{From{-1}, Axis{0}, Start{0}, Stride{1}, End{1}}}, {}, {}},
     NegativeTestParams{{{1}}, {{From{1}, Axis{0}, Start{0}, Stride{1}, End{1}}}, {}, {}},
 
@@ -224,7 +224,7 @@ using GetNumIterationsInvalidDirectionTests = GetNumIterationsNegativeTests;
 TEST_P(GetNumIterationsInvalidDirectionTests, InvalidDirectionThrowsAnException) {
     ASSERT_ANY_THROW(Run());
 }
-INSTANTIATE_TEST_CASE_P(NumIterationsTest, GetNumIterationsInvalidDirectionTests, testing::Values<NegativeTestParams>(
+INSTANTIATE_TEST_SUITE_P(NumIterationsTest, GetNumIterationsInvalidDirectionTests, testing::Values<NegativeTestParams>(
     NegativeTestParams{{{10}}, {{From{0}, Axis{0}, Start{8}, Stride{1}, End{2}}}, {}, {}},
     NegativeTestParams{{{10}}, {{From{0}, Axis{0}, Start{2}, Stride{-1}, End{8}}}, {}, {}},
 
@@ -236,7 +236,7 @@ using GetNumIterationsInvalidStepTests = GetNumIterationsNegativeTests;
 TEST_P(GetNumIterationsInvalidStepTests, InvalidStepThrowsAnException) {
     ASSERT_ANY_THROW(Run());
 }
-INSTANTIATE_TEST_CASE_P(NumIterationsTest, GetNumIterationsInvalidStepTests, testing::Values<NegativeTestParams>(
+INSTANTIATE_TEST_SUITE_P(NumIterationsTest, GetNumIterationsInvalidStepTests, testing::Values<NegativeTestParams>(
     NegativeTestParams{{{10}}, {{From{0}, Axis{0}, Start{2}, Stride{3}, End{6}}}, {}, {}},
     NegativeTestParams{{{10}}, {{From{0}, Axis{0}, Start{2}, Stride{8}, End{6}}}, {}, {}},
     NegativeTestParams{{{10}}, {{From{0}, Axis{0}, Start{6}, Stride{-3}, End{2}}}, {}, {}},
@@ -252,7 +252,7 @@ using GetNumIterationsInvalidIterationNumbersTests = GetNumIterationsNegativeTes
 TEST_P(GetNumIterationsInvalidIterationNumbersTests, InvalidInterationNumbersThrowAnException) {
     ASSERT_ANY_THROW(Run());
 }
-INSTANTIATE_TEST_CASE_P(NumIterationsTest, GetNumIterationsInvalidIterationNumbersTests, testing::Values<NegativeTestParams>(
+INSTANTIATE_TEST_SUITE_P(NumIterationsTest, GetNumIterationsInvalidIterationNumbersTests, testing::Values<NegativeTestParams>(
     NegativeTestParams{
         {
             {1, 3, 24, 24},
@@ -473,6 +473,6 @@ std::vector<PositiveTestParams> g_positiveTestParameters = {
     {{{13}}, {{From{0}, Axis{0}, Start{ 10}, Stride{-1}, End{-11}}}, {}, {}, 7},
     {{{13}}, {{From{0}, Axis{0}, Start{  9}, Stride{-2}, End{-11}}}, {}, {}, 3}
 };
-INSTANTIATE_TEST_CASE_P(NumIterationsTest, GetNumIterationsPositiveTests, testing::ValuesIn(g_positiveTestParameters));
+INSTANTIATE_TEST_SUITE_P(NumIterationsTest, GetNumIterationsPositiveTests, testing::ValuesIn(g_positiveTestParameters));
 
 }
diff --git a/inference-engine/thirdparty/clDNN/CMakeLists.txt b/inference-engine/thirdparty/clDNN/CMakeLists.txt
index 41833b58dc58f2..4106e0522cb9d2 100644
--- a/inference-engine/thirdparty/clDNN/CMakeLists.txt
+++ b/inference-engine/thirdparty/clDNN/CMakeLists.txt
@@ -92,9 +92,6 @@ set(CLDNN__OCL_ICD_DIR  "${CLDNN__COMMON_DIR}/khronos_ocl_icd")
 # Path which points to directory with C++ bindings for OpenCL (header files + wrapper that disables specific warnings).
 set(CLDNN__KHR_CLHPP_DIR "${CLDNN__COMMON_DIR}/include")
 
-# Path which points to directory with fused version of googletest framework (with fused googlemock as well).
-set(CLDNN__GTEST_DIR     "${CLDNN__COMMON_DIR}/googletest-fused")
-
 # Build targets settings.
 
 # Path which points to default root directory for compilation output.
diff --git a/inference-engine/thirdparty/clDNN/common/googletest-fused/License.txt b/inference-engine/thirdparty/clDNN/common/googletest-fused/License.txt
deleted file mode 100644
index 65c76c50ce257c..00000000000000
--- a/inference-engine/thirdparty/clDNN/common/googletest-fused/License.txt
+++ /dev/null
@@ -1,28 +0,0 @@
-Copyright 2008, Google Inc.
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are
-met:
-
-    * Redistributions of source code must retain the above copyright
-notice, this list of conditions and the following disclaimer.
-    * Redistributions in binary form must reproduce the above
-copyright notice, this list of conditions and the following disclaimer
-in the documentation and/or other materials provided with the
-distribution.
-    * Neither the name of Google Inc. nor the names of its
-contributors may be used to endorse or promote products derived from
-this software without specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
\ No newline at end of file
diff --git a/inference-engine/thirdparty/clDNN/common/googletest-fused/gmock-gtest-all.cc b/inference-engine/thirdparty/clDNN/common/googletest-fused/gmock-gtest-all.cc
deleted file mode 100644
index 272563d970d039..00000000000000
--- a/inference-engine/thirdparty/clDNN/common/googletest-fused/gmock-gtest-all.cc
+++ /dev/null
@@ -1,13278 +0,0 @@
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: mheule@google.com (Markus Heule)
-//
-// Google C++ Testing and Mocking Framework (Google Test)
-//
-// Sometimes it's desirable to build Google Test by compiling a single file.
-// This file serves this purpose.
-
-// This line ensures that gtest.h can be compiled on its own, even
-// when it's fused.
-#include "gtest/gtest.h"
-
-// The following lines pull in the real gtest *.cc files.
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-//
-// The Google C++ Testing and Mocking Framework (Google Test)
-
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-//
-// Utilities for testing Google Test itself and code that uses Google Test
-// (e.g. frameworks built on top of Google Test).
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_SPI_H_
-#define GTEST_INCLUDE_GTEST_GTEST_SPI_H_
-
-
-namespace testing {
-
-// This helper class can be used to mock out Google Test failure reporting
-// so that we can test Google Test or code that builds on Google Test.
-//
-// An object of this class appends a TestPartResult object to the
-// TestPartResultArray object given in the constructor whenever a Google Test
-// failure is reported. It can either intercept only failures that are
-// generated in the same thread that created this object or it can intercept
-// all generated failures. The scope of this mock object can be controlled with
-// the second argument to the two arguments constructor.
-class GTEST_API_ ScopedFakeTestPartResultReporter
-    : public TestPartResultReporterInterface {
- public:
-  // The two possible mocking modes of this object.
-  enum InterceptMode {
-    INTERCEPT_ONLY_CURRENT_THREAD,  // Intercepts only thread local failures.
-    INTERCEPT_ALL_THREADS           // Intercepts all failures.
-  };
-
-  // The c'tor sets this object as the test part result reporter used
-  // by Google Test.  The 'result' parameter specifies where to report the
-  // results. This reporter will only catch failures generated in the current
-  // thread. DEPRECATED
-  explicit ScopedFakeTestPartResultReporter(TestPartResultArray* result);
-
-  // Same as above, but you can choose the interception scope of this object.
-  ScopedFakeTestPartResultReporter(InterceptMode intercept_mode,
-                                   TestPartResultArray* result);
-
-  // The d'tor restores the previous test part result reporter.
-  virtual ~ScopedFakeTestPartResultReporter();
-
-  // Appends the TestPartResult object to the TestPartResultArray
-  // received in the constructor.
-  //
-  // This method is from the TestPartResultReporterInterface
-  // interface.
-  virtual void ReportTestPartResult(const TestPartResult& result);
- private:
-  void Init();
-
-  const InterceptMode intercept_mode_;
-  TestPartResultReporterInterface* old_reporter_;
-  TestPartResultArray* const result_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ScopedFakeTestPartResultReporter);
-};
-
-namespace internal {
-
-// A helper class for implementing EXPECT_FATAL_FAILURE() and
-// EXPECT_NONFATAL_FAILURE().  Its destructor verifies that the given
-// TestPartResultArray contains exactly one failure that has the given
-// type and contains the given substring.  If that's not the case, a
-// non-fatal failure will be generated.
-class GTEST_API_ SingleFailureChecker {
- public:
-  // The constructor remembers the arguments.
-  SingleFailureChecker(const TestPartResultArray* results,
-                       TestPartResult::Type type, const std::string& substr);
-  ~SingleFailureChecker();
- private:
-  const TestPartResultArray* const results_;
-  const TestPartResult::Type type_;
-  const std::string substr_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(SingleFailureChecker);
-};
-
-}  // namespace internal
-
-}  // namespace testing
-
-// A set of macros for testing Google Test assertions or code that's expected
-// to generate Google Test fatal failures.  It verifies that the given
-// statement will cause exactly one fatal Google Test failure with 'substr'
-// being part of the failure message.
-//
-// There are two different versions of this macro. EXPECT_FATAL_FAILURE only
-// affects and considers failures generated in the current thread and
-// EXPECT_FATAL_FAILURE_ON_ALL_THREADS does the same but for all threads.
-//
-// The verification of the assertion is done correctly even when the statement
-// throws an exception or aborts the current function.
-//
-// Known restrictions:
-//   - 'statement' cannot reference local non-static variables or
-//     non-static members of the current object.
-//   - 'statement' cannot return a value.
-//   - You cannot stream a failure message to this macro.
-//
-// Note that even though the implementations of the following two
-// macros are much alike, we cannot refactor them to use a common
-// helper macro, due to some peculiarity in how the preprocessor
-// works.  The AcceptsMacroThatExpandsToUnprotectedComma test in
-// gtest_unittest.cc will fail to compile if we do that.
-#define EXPECT_FATAL_FAILURE(statement, substr) \
-  do { \
-    class GTestExpectFatalFailureHelper {\
-     public:\
-      static void Execute() { statement; }\
-    };\
-    ::testing::TestPartResultArray gtest_failures;\
-    ::testing::internal::SingleFailureChecker gtest_checker(\
-        &gtest_failures, ::testing::TestPartResult::kFatalFailure, (substr));\
-    {\
-      ::testing::ScopedFakeTestPartResultReporter gtest_reporter(\
-          ::testing::ScopedFakeTestPartResultReporter:: \
-          INTERCEPT_ONLY_CURRENT_THREAD, &gtest_failures);\
-      GTestExpectFatalFailureHelper::Execute();\
-    }\
-  } while (::testing::internal::AlwaysFalse())
-
-#define EXPECT_FATAL_FAILURE_ON_ALL_THREADS(statement, substr) \
-  do { \
-    class GTestExpectFatalFailureHelper {\
-     public:\
-      static void Execute() { statement; }\
-    };\
-    ::testing::TestPartResultArray gtest_failures;\
-    ::testing::internal::SingleFailureChecker gtest_checker(\
-        &gtest_failures, ::testing::TestPartResult::kFatalFailure, (substr));\
-    {\
-      ::testing::ScopedFakeTestPartResultReporter gtest_reporter(\
-          ::testing::ScopedFakeTestPartResultReporter:: \
-          INTERCEPT_ALL_THREADS, &gtest_failures);\
-      GTestExpectFatalFailureHelper::Execute();\
-    }\
-  } while (::testing::internal::AlwaysFalse())
-
-// A macro for testing Google Test assertions or code that's expected to
-// generate Google Test non-fatal failures.  It asserts that the given
-// statement will cause exactly one non-fatal Google Test failure with 'substr'
-// being part of the failure message.
-//
-// There are two different versions of this macro. EXPECT_NONFATAL_FAILURE only
-// affects and considers failures generated in the current thread and
-// EXPECT_NONFATAL_FAILURE_ON_ALL_THREADS does the same but for all threads.
-//
-// 'statement' is allowed to reference local variables and members of
-// the current object.
-//
-// The verification of the assertion is done correctly even when the statement
-// throws an exception or aborts the current function.
-//
-// Known restrictions:
-//   - You cannot stream a failure message to this macro.
-//
-// Note that even though the implementations of the following two
-// macros are much alike, we cannot refactor them to use a common
-// helper macro, due to some peculiarity in how the preprocessor
-// works.  If we do that, the code won't compile when the user gives
-// EXPECT_NONFATAL_FAILURE() a statement that contains a macro that
-// expands to code containing an unprotected comma.  The
-// AcceptsMacroThatExpandsToUnprotectedComma test in gtest_unittest.cc
-// catches that.
-//
-// For the same reason, we have to write
-//   if (::testing::internal::AlwaysTrue()) { statement; }
-// instead of
-//   GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(statement)
-// to avoid an MSVC warning on unreachable code.
-#define EXPECT_NONFATAL_FAILURE(statement, substr) \
-  do {\
-    ::testing::TestPartResultArray gtest_failures;\
-    ::testing::internal::SingleFailureChecker gtest_checker(\
-        &gtest_failures, ::testing::TestPartResult::kNonFatalFailure, \
-        (substr));\
-    {\
-      ::testing::ScopedFakeTestPartResultReporter gtest_reporter(\
-          ::testing::ScopedFakeTestPartResultReporter:: \
-          INTERCEPT_ONLY_CURRENT_THREAD, &gtest_failures);\
-      if (::testing::internal::AlwaysTrue()) { statement; }\
-    }\
-  } while (::testing::internal::AlwaysFalse())
-
-#define EXPECT_NONFATAL_FAILURE_ON_ALL_THREADS(statement, substr) \
-  do {\
-    ::testing::TestPartResultArray gtest_failures;\
-    ::testing::internal::SingleFailureChecker gtest_checker(\
-        &gtest_failures, ::testing::TestPartResult::kNonFatalFailure, \
-        (substr));\
-    {\
-      ::testing::ScopedFakeTestPartResultReporter gtest_reporter(\
-          ::testing::ScopedFakeTestPartResultReporter::INTERCEPT_ALL_THREADS, \
-          &gtest_failures);\
-      if (::testing::internal::AlwaysTrue()) { statement; }\
-    }\
-  } while (::testing::internal::AlwaysFalse())
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_SPI_H_
-
-#include <ctype.h>
-#include <math.h>
-#include <stdarg.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <time.h>
-#include <wchar.h>
-#include <wctype.h>
-
-#include <algorithm>
-#include <iomanip>
-#include <limits>
-#include <list>
-#include <map>
-#include <ostream>  // NOLINT
-#include <sstream>
-#include <vector>
-
-#if GTEST_OS_LINUX
-
-// TODO(kenton@google.com): Use autoconf to detect availability of
-// gettimeofday().
-# define GTEST_HAS_GETTIMEOFDAY_ 1
-
-# include <fcntl.h>  // NOLINT
-# include <limits.h>  // NOLINT
-# include <sched.h>  // NOLINT
-// Declares vsnprintf().  This header is not available on Windows.
-# include <strings.h>  // NOLINT
-# include <sys/mman.h>  // NOLINT
-# include <sys/time.h>  // NOLINT
-# include <unistd.h>  // NOLINT
-# include <string>
-
-#elif GTEST_OS_SYMBIAN
-# define GTEST_HAS_GETTIMEOFDAY_ 1
-# include <sys/time.h>  // NOLINT
-
-#elif GTEST_OS_ZOS
-# define GTEST_HAS_GETTIMEOFDAY_ 1
-# include <sys/time.h>  // NOLINT
-
-// On z/OS we additionally need strings.h for strcasecmp.
-# include <strings.h>  // NOLINT
-
-#elif GTEST_OS_WINDOWS_MOBILE  // We are on Windows CE.
-
-# include <windows.h>  // NOLINT
-# undef min
-
-#elif GTEST_OS_WINDOWS  // We are on Windows proper.
-
-# include <io.h>  // NOLINT
-# include <sys/timeb.h>  // NOLINT
-# include <sys/types.h>  // NOLINT
-# include <sys/stat.h>  // NOLINT
-
-# if GTEST_OS_WINDOWS_MINGW
-// MinGW has gettimeofday() but not _ftime64().
-// TODO(kenton@google.com): Use autoconf to detect availability of
-//   gettimeofday().
-// TODO(kenton@google.com): There are other ways to get the time on
-//   Windows, like GetTickCount() or GetSystemTimeAsFileTime().  MinGW
-//   supports these.  consider using them instead.
-#  define GTEST_HAS_GETTIMEOFDAY_ 1
-#  include <sys/time.h>  // NOLINT
-# endif  // GTEST_OS_WINDOWS_MINGW
-
-// cpplint thinks that the header is already included, so we want to
-// silence it.
-# include <windows.h>  // NOLINT
-# undef min
-
-#else
-
-// Assume other platforms have gettimeofday().
-// TODO(kenton@google.com): Use autoconf to detect availability of
-//   gettimeofday().
-# define GTEST_HAS_GETTIMEOFDAY_ 1
-
-// cpplint thinks that the header is already included, so we want to
-// silence it.
-# include <sys/time.h>  // NOLINT
-# include <unistd.h>  // NOLINT
-
-#endif  // GTEST_OS_LINUX
-
-#if GTEST_HAS_EXCEPTIONS
-# include <stdexcept>
-#endif
-
-#if GTEST_CAN_STREAM_RESULTS_
-# include <arpa/inet.h>  // NOLINT
-# include <netdb.h>  // NOLINT
-# include <sys/socket.h>  // NOLINT
-# include <sys/types.h>  // NOLINT
-#endif
-
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-// Utility functions and classes used by the Google C++ testing framework.
-//
-// Author: wan@google.com (Zhanyong Wan)
-//
-// This file contains purely Google Test's internal implementation.  Please
-// DO NOT #INCLUDE IT IN A USER PROGRAM.
-
-#ifndef GTEST_SRC_GTEST_INTERNAL_INL_H_
-#define GTEST_SRC_GTEST_INTERNAL_INL_H_
-
-#ifndef _WIN32_WCE
-# include <errno.h>
-#endif  // !_WIN32_WCE
-#include <stddef.h>
-#include <stdlib.h>  // For strtoll/_strtoul64/malloc/free.
-#include <string.h>  // For memmove.
-
-#include <algorithm>
-#include <string>
-#include <vector>
-
-
-#if GTEST_CAN_STREAM_RESULTS_
-# include <arpa/inet.h>  // NOLINT
-# include <netdb.h>  // NOLINT
-#endif
-
-#if GTEST_OS_WINDOWS
-# include <windows.h>  // NOLINT
-#endif  // GTEST_OS_WINDOWS
-
-
-namespace testing {
-
-// Declares the flags.
-//
-// We don't want the users to modify this flag in the code, but want
-// Google Test's own unit tests to be able to access it. Therefore we
-// declare it here as opposed to in gtest.h.
-GTEST_DECLARE_bool_(death_test_use_fork);
-
-namespace internal {
-
-// The value of GetTestTypeId() as seen from within the Google Test
-// library.  This is solely for testing GetTestTypeId().
-GTEST_API_ extern const TypeId kTestTypeIdInGoogleTest;
-
-// Names of the flags (needed for parsing Google Test flags).
-const char kAlsoRunDisabledTestsFlag[] = "also_run_disabled_tests";
-const char kBreakOnFailureFlag[] = "break_on_failure";
-const char kCatchExceptionsFlag[] = "catch_exceptions";
-const char kColorFlag[] = "color";
-const char kFilterFlag[] = "filter";
-const char kListTestsFlag[] = "list_tests";
-const char kOutputFlag[] = "output";
-const char kPrintTimeFlag[] = "print_time";
-const char kPrintUTF8Flag[] = "print_utf8";
-const char kRandomSeedFlag[] = "random_seed";
-const char kRepeatFlag[] = "repeat";
-const char kShuffleFlag[] = "shuffle";
-const char kStackTraceDepthFlag[] = "stack_trace_depth";
-const char kStreamResultToFlag[] = "stream_result_to";
-const char kThrowOnFailureFlag[] = "throw_on_failure";
-const char kFlagfileFlag[] = "flagfile";
-
-// A valid random seed must be in [1, kMaxRandomSeed].
-const int kMaxRandomSeed = 99999;
-
-// g_help_flag is true iff the --help flag or an equivalent form is
-// specified on the command line.
-GTEST_API_ extern bool g_help_flag;
-
-// Returns the current time in milliseconds.
-GTEST_API_ TimeInMillis GetTimeInMillis();
-
-// Returns true iff Google Test should use colors in the output.
-GTEST_API_ bool ShouldUseColor(bool stdout_is_tty);
-
-// Formats the given time in milliseconds as seconds.
-GTEST_API_ std::string FormatTimeInMillisAsSeconds(TimeInMillis ms);
-
-// Converts the given time in milliseconds to a date string in the ISO 8601
-// format, without the timezone information.  N.B.: due to the use the
-// non-reentrant localtime() function, this function is not thread safe.  Do
-// not use it in any code that can be called from multiple threads.
-GTEST_API_ std::string FormatEpochTimeInMillisAsIso8601(TimeInMillis ms);
-
-// Parses a string for an Int32 flag, in the form of "--flag=value".
-//
-// On success, stores the value of the flag in *value, and returns
-// true.  On failure, returns false without changing *value.
-GTEST_API_ bool ParseInt32Flag(
-    const char* str, const char* flag, Int32* value);
-
-// Returns a random seed in range [1, kMaxRandomSeed] based on the
-// given --gtest_random_seed flag value.
-inline int GetRandomSeedFromFlag(Int32 random_seed_flag) {
-  const unsigned int raw_seed = (random_seed_flag == 0) ?
-      static_cast<unsigned int>(GetTimeInMillis()) :
-      static_cast<unsigned int>(random_seed_flag);
-
-  // Normalizes the actual seed to range [1, kMaxRandomSeed] such that
-  // it's easy to type.
-  const int normalized_seed =
-      static_cast<int>((raw_seed - 1U) %
-                       static_cast<unsigned int>(kMaxRandomSeed)) + 1;
-  return normalized_seed;
-}
-
-// Returns the first valid random seed after 'seed'.  The behavior is
-// undefined if 'seed' is invalid.  The seed after kMaxRandomSeed is
-// considered to be 1.
-inline int GetNextRandomSeed(int seed) {
-  GTEST_CHECK_(1 <= seed && seed <= kMaxRandomSeed)
-      << "Invalid random seed " << seed << " - must be in [1, "
-      << kMaxRandomSeed << "].";
-  const int next_seed = seed + 1;
-  return (next_seed > kMaxRandomSeed) ? 1 : next_seed;
-}
-
-// This class saves the values of all Google Test flags in its c'tor, and
-// restores them in its d'tor.
-class GTestFlagSaver {
- public:
-  // The c'tor.
-  GTestFlagSaver() {
-    also_run_disabled_tests_ = GTEST_FLAG(also_run_disabled_tests);
-    break_on_failure_ = GTEST_FLAG(break_on_failure);
-    catch_exceptions_ = GTEST_FLAG(catch_exceptions);
-    color_ = GTEST_FLAG(color);
-    death_test_style_ = GTEST_FLAG(death_test_style);
-    death_test_use_fork_ = GTEST_FLAG(death_test_use_fork);
-    filter_ = GTEST_FLAG(filter);
-    internal_run_death_test_ = GTEST_FLAG(internal_run_death_test);
-    list_tests_ = GTEST_FLAG(list_tests);
-    output_ = GTEST_FLAG(output);
-    print_time_ = GTEST_FLAG(print_time);
-    print_utf8_ = GTEST_FLAG(print_utf8);
-    random_seed_ = GTEST_FLAG(random_seed);
-    repeat_ = GTEST_FLAG(repeat);
-    shuffle_ = GTEST_FLAG(shuffle);
-    stack_trace_depth_ = GTEST_FLAG(stack_trace_depth);
-    stream_result_to_ = GTEST_FLAG(stream_result_to);
-    throw_on_failure_ = GTEST_FLAG(throw_on_failure);
-  }
-
-  // The d'tor is not virtual.  DO NOT INHERIT FROM THIS CLASS.
-  ~GTestFlagSaver() {
-    GTEST_FLAG(also_run_disabled_tests) = also_run_disabled_tests_;
-    GTEST_FLAG(break_on_failure) = break_on_failure_;
-    GTEST_FLAG(catch_exceptions) = catch_exceptions_;
-    GTEST_FLAG(color) = color_;
-    GTEST_FLAG(death_test_style) = death_test_style_;
-    GTEST_FLAG(death_test_use_fork) = death_test_use_fork_;
-    GTEST_FLAG(filter) = filter_;
-    GTEST_FLAG(internal_run_death_test) = internal_run_death_test_;
-    GTEST_FLAG(list_tests) = list_tests_;
-    GTEST_FLAG(output) = output_;
-    GTEST_FLAG(print_time) = print_time_;
-    GTEST_FLAG(print_utf8) = print_utf8_;
-    GTEST_FLAG(random_seed) = random_seed_;
-    GTEST_FLAG(repeat) = repeat_;
-    GTEST_FLAG(shuffle) = shuffle_;
-    GTEST_FLAG(stack_trace_depth) = stack_trace_depth_;
-    GTEST_FLAG(stream_result_to) = stream_result_to_;
-    GTEST_FLAG(throw_on_failure) = throw_on_failure_;
-  }
-
- private:
-  // Fields for saving the original values of flags.
-  bool also_run_disabled_tests_;
-  bool break_on_failure_;
-  bool catch_exceptions_;
-  std::string color_;
-  std::string death_test_style_;
-  bool death_test_use_fork_;
-  std::string filter_;
-  std::string internal_run_death_test_;
-  bool list_tests_;
-  std::string output_;
-  bool print_time_;
-  bool print_utf8_;
-  internal::Int32 random_seed_;
-  internal::Int32 repeat_;
-  bool shuffle_;
-  internal::Int32 stack_trace_depth_;
-  std::string stream_result_to_;
-  bool throw_on_failure_;
-} GTEST_ATTRIBUTE_UNUSED_;
-
-// Converts a Unicode code point to a narrow string in UTF-8 encoding.
-// code_point parameter is of type UInt32 because wchar_t may not be
-// wide enough to contain a code point.
-// If the code_point is not a valid Unicode code point
-// (i.e. outside of Unicode range U+0 to U+10FFFF) it will be converted
-// to "(Invalid Unicode 0xXXXXXXXX)".
-GTEST_API_ std::string CodePointToUtf8(UInt32 code_point);
-
-// Converts a wide string to a narrow string in UTF-8 encoding.
-// The wide string is assumed to have the following encoding:
-//   UTF-16 if sizeof(wchar_t) == 2 (on Windows, Cygwin, Symbian OS)
-//   UTF-32 if sizeof(wchar_t) == 4 (on Linux)
-// Parameter str points to a null-terminated wide string.
-// Parameter num_chars may additionally limit the number
-// of wchar_t characters processed. -1 is used when the entire string
-// should be processed.
-// If the string contains code points that are not valid Unicode code points
-// (i.e. outside of Unicode range U+0 to U+10FFFF) they will be output
-// as '(Invalid Unicode 0xXXXXXXXX)'. If the string is in UTF16 encoding
-// and contains invalid UTF-16 surrogate pairs, values in those pairs
-// will be encoded as individual Unicode characters from Basic Normal Plane.
-GTEST_API_ std::string WideStringToUtf8(const wchar_t* str, int num_chars);
-
-// Reads the GTEST_SHARD_STATUS_FILE environment variable, and creates the file
-// if the variable is present. If a file already exists at this location, this
-// function will write over it. If the variable is present, but the file cannot
-// be created, prints an error and exits.
-void WriteToShardStatusFileIfNeeded();
-
-// Checks whether sharding is enabled by examining the relevant
-// environment variable values. If the variables are present,
-// but inconsistent (e.g., shard_index >= total_shards), prints
-// an error and exits. If in_subprocess_for_death_test, sharding is
-// disabled because it must only be applied to the original test
-// process. Otherwise, we could filter out death tests we intended to execute.
-GTEST_API_ bool ShouldShard(const char* total_shards_str,
-                            const char* shard_index_str,
-                            bool in_subprocess_for_death_test);
-
-// Parses the environment variable var as an Int32. If it is unset,
-// returns default_val. If it is not an Int32, prints an error and
-// and aborts.
-GTEST_API_ Int32 Int32FromEnvOrDie(const char* env_var, Int32 default_val);
-
-// Given the total number of shards, the shard index, and the test id,
-// returns true iff the test should be run on this shard. The test id is
-// some arbitrary but unique non-negative integer assigned to each test
-// method. Assumes that 0 <= shard_index < total_shards.
-GTEST_API_ bool ShouldRunTestOnShard(
-    int total_shards, int shard_index, int test_id);
-
-// STL container utilities.
-
-// Returns the number of elements in the given container that satisfy
-// the given predicate.
-template <class Container, typename Predicate>
-inline int CountIf(const Container& c, Predicate predicate) {
-  // Implemented as an explicit loop since std::count_if() in libCstd on
-  // Solaris has a non-standard signature.
-  int count = 0;
-  for (typename Container::const_iterator it = c.begin(); it != c.end(); ++it) {
-    if (predicate(*it))
-      ++count;
-  }
-  return count;
-}
-
-// Applies a function/functor to each element in the container.
-template <class Container, typename Functor>
-void ForEach(const Container& c, Functor functor) {
-  std::for_each(c.begin(), c.end(), functor);
-}
-
-// Returns the i-th element of the vector, or default_value if i is not
-// in range [0, v.size()).
-template <typename E>
-inline E GetElementOr(const std::vector<E>& v, int i, E default_value) {
-  return (i < 0 || i >= static_cast<int>(v.size())) ? default_value : v[i];
-}
-
-// Performs an in-place shuffle of a range of the vector's elements.
-// 'begin' and 'end' are element indices as an STL-style range;
-// i.e. [begin, end) are shuffled, where 'end' == size() means to
-// shuffle to the end of the vector.
-template <typename E>
-void ShuffleRange(internal::Random* random, int begin, int end,
-                  std::vector<E>* v) {
-  const int size = static_cast<int>(v->size());
-  GTEST_CHECK_(0 <= begin && begin <= size)
-      << "Invalid shuffle range start " << begin << ": must be in range [0, "
-      << size << "].";
-  GTEST_CHECK_(begin <= end && end <= size)
-      << "Invalid shuffle range finish " << end << ": must be in range ["
-      << begin << ", " << size << "].";
-
-  // Fisher-Yates shuffle, from
-  // http://en.wikipedia.org/wiki/Fisher-Yates_shuffle
-  for (int range_width = end - begin; range_width >= 2; range_width--) {
-    const int last_in_range = begin + range_width - 1;
-    const int selected = begin + random->Generate(range_width);
-    std::swap((*v)[selected], (*v)[last_in_range]);
-  }
-}
-
-// Performs an in-place shuffle of the vector's elements.
-template <typename E>
-inline void Shuffle(internal::Random* random, std::vector<E>* v) {
-  ShuffleRange(random, 0, static_cast<int>(v->size()), v);
-}
-
-// A function for deleting an object.  Handy for being used as a
-// functor.
-template <typename T>
-static void Delete(T* x) {
-  delete x;
-}
-
-// A predicate that checks the key of a TestProperty against a known key.
-//
-// TestPropertyKeyIs is copyable.
-class TestPropertyKeyIs {
- public:
-  // Constructor.
-  //
-  // TestPropertyKeyIs has NO default constructor.
-  explicit TestPropertyKeyIs(const std::string& key) : key_(key) {}
-
-  // Returns true iff the test name of test property matches on key_.
-  bool operator()(const TestProperty& test_property) const {
-    return test_property.key() == key_;
-  }
-
- private:
-  std::string key_;
-};
-
-// Class UnitTestOptions.
-//
-// This class contains functions for processing options the user
-// specifies when running the tests.  It has only static members.
-//
-// In most cases, the user can specify an option using either an
-// environment variable or a command line flag.  E.g. you can set the
-// test filter using either GTEST_FILTER or --gtest_filter.  If both
-// the variable and the flag are present, the latter overrides the
-// former.
-class GTEST_API_ UnitTestOptions {
- public:
-  // Functions for processing the gtest_output flag.
-
-  // Returns the output format, or "" for normal printed output.
-  static std::string GetOutputFormat();
-
-  // Returns the absolute path of the requested output file, or the
-  // default (test_detail.xml in the original working directory) if
-  // none was explicitly specified.
-  static std::string GetAbsolutePathToOutputFile();
-
-  // Functions for processing the gtest_filter flag.
-
-  // Returns true iff the wildcard pattern matches the string.  The
-  // first ':' or '\0' character in pattern marks the end of it.
-  //
-  // This recursive algorithm isn't very efficient, but is clear and
-  // works well enough for matching test names, which are short.
-  static bool PatternMatchesString(const char *pattern, const char *str);
-
-  // Returns true iff the user-specified filter matches the test case
-  // name and the test name.
-  static bool FilterMatchesTest(const std::string &test_case_name,
-                                const std::string &test_name);
-
-#if GTEST_OS_WINDOWS
-  // Function for supporting the gtest_catch_exception flag.
-
-  // Returns EXCEPTION_EXECUTE_HANDLER if Google Test should handle the
-  // given SEH exception, or EXCEPTION_CONTINUE_SEARCH otherwise.
-  // This function is useful as an __except condition.
-  static int GTestShouldProcessSEH(DWORD exception_code);
-#endif  // GTEST_OS_WINDOWS
-
-  // Returns true if "name" matches the ':' separated list of glob-style
-  // filters in "filter".
-  static bool MatchesFilter(const std::string& name, const char* filter);
-};
-
-// Returns the current application's name, removing directory path if that
-// is present.  Used by UnitTestOptions::GetOutputFile.
-GTEST_API_ FilePath GetCurrentExecutableName();
-
-// The role interface for getting the OS stack trace as a string.
-class OsStackTraceGetterInterface {
- public:
-  OsStackTraceGetterInterface() {}
-  virtual ~OsStackTraceGetterInterface() {}
-
-  // Returns the current OS stack trace as an std::string.  Parameters:
-  //
-  //   max_depth  - the maximum number of stack frames to be included
-  //                in the trace.
-  //   skip_count - the number of top frames to be skipped; doesn't count
-  //                against max_depth.
-  virtual std::string CurrentStackTrace(int max_depth, int skip_count) = 0;
-
-  // UponLeavingGTest() should be called immediately before Google Test calls
-  // user code. It saves some information about the current stack that
-  // CurrentStackTrace() will use to find and hide Google Test stack frames.
-  virtual void UponLeavingGTest() = 0;
-
-  // This string is inserted in place of stack frames that are part of
-  // Google Test's implementation.
-  static const char* const kElidedFramesMarker;
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(OsStackTraceGetterInterface);
-};
-
-// A working implementation of the OsStackTraceGetterInterface interface.
-class OsStackTraceGetter : public OsStackTraceGetterInterface {
- public:
-  OsStackTraceGetter() {}
-
-  virtual std::string CurrentStackTrace(int max_depth, int skip_count);
-  virtual void UponLeavingGTest();
-
- private:
-#if GTEST_HAS_ABSL
-  Mutex mutex_;  // Protects all internal state.
-
-  // We save the stack frame below the frame that calls user code.
-  // We do this because the address of the frame immediately below
-  // the user code changes between the call to UponLeavingGTest()
-  // and any calls to the stack trace code from within the user code.
-  void* caller_frame_ = nullptr;
-#endif  // GTEST_HAS_ABSL
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(OsStackTraceGetter);
-};
-
-// Information about a Google Test trace point.
-struct TraceInfo {
-  const char* file;
-  int line;
-  std::string message;
-};
-
-// This is the default global test part result reporter used in UnitTestImpl.
-// This class should only be used by UnitTestImpl.
-class DefaultGlobalTestPartResultReporter
-  : public TestPartResultReporterInterface {
- public:
-  explicit DefaultGlobalTestPartResultReporter(UnitTestImpl* unit_test);
-  // Implements the TestPartResultReporterInterface. Reports the test part
-  // result in the current test.
-  virtual void ReportTestPartResult(const TestPartResult& result);
-
- private:
-  UnitTestImpl* const unit_test_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(DefaultGlobalTestPartResultReporter);
-};
-
-// This is the default per thread test part result reporter used in
-// UnitTestImpl. This class should only be used by UnitTestImpl.
-class DefaultPerThreadTestPartResultReporter
-    : public TestPartResultReporterInterface {
- public:
-  explicit DefaultPerThreadTestPartResultReporter(UnitTestImpl* unit_test);
-  // Implements the TestPartResultReporterInterface. The implementation just
-  // delegates to the current global test part result reporter of *unit_test_.
-  virtual void ReportTestPartResult(const TestPartResult& result);
-
- private:
-  UnitTestImpl* const unit_test_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(DefaultPerThreadTestPartResultReporter);
-};
-
-// The private implementation of the UnitTest class.  We don't protect
-// the methods under a mutex, as this class is not accessible by a
-// user and the UnitTest class that delegates work to this class does
-// proper locking.
-class GTEST_API_ UnitTestImpl {
- public:
-  explicit UnitTestImpl(UnitTest* parent);
-  virtual ~UnitTestImpl();
-
-  // There are two different ways to register your own TestPartResultReporter.
-  // You can register your own repoter to listen either only for test results
-  // from the current thread or for results from all threads.
-  // By default, each per-thread test result repoter just passes a new
-  // TestPartResult to the global test result reporter, which registers the
-  // test part result for the currently running test.
-
-  // Returns the global test part result reporter.
-  TestPartResultReporterInterface* GetGlobalTestPartResultReporter();
-
-  // Sets the global test part result reporter.
-  void SetGlobalTestPartResultReporter(
-      TestPartResultReporterInterface* reporter);
-
-  // Returns the test part result reporter for the current thread.
-  TestPartResultReporterInterface* GetTestPartResultReporterForCurrentThread();
-
-  // Sets the test part result reporter for the current thread.
-  void SetTestPartResultReporterForCurrentThread(
-      TestPartResultReporterInterface* reporter);
-
-  // Gets the number of successful test cases.
-  int successful_test_case_count() const;
-
-  // Gets the number of failed test cases.
-  int failed_test_case_count() const;
-
-  // Gets the number of all test cases.
-  int total_test_case_count() const;
-
-  // Gets the number of all test cases that contain at least one test
-  // that should run.
-  int test_case_to_run_count() const;
-
-  // Gets the number of successful tests.
-  int successful_test_count() const;
-
-  // Gets the number of failed tests.
-  int failed_test_count() const;
-
-  // Gets the number of disabled tests that will be reported in the XML report.
-  int reportable_disabled_test_count() const;
-
-  // Gets the number of disabled tests.
-  int disabled_test_count() const;
-
-  // Gets the number of tests to be printed in the XML report.
-  int reportable_test_count() const;
-
-  // Gets the number of all tests.
-  int total_test_count() const;
-
-  // Gets the number of tests that should run.
-  int test_to_run_count() const;
-
-  // Gets the time of the test program start, in ms from the start of the
-  // UNIX epoch.
-  TimeInMillis start_timestamp() const { return start_timestamp_; }
-
-  // Gets the elapsed time, in milliseconds.
-  TimeInMillis elapsed_time() const { return elapsed_time_; }
-
-  // Returns true iff the unit test passed (i.e. all test cases passed).
-  bool Passed() const { return !Failed(); }
-
-  // Returns true iff the unit test failed (i.e. some test case failed
-  // or something outside of all tests failed).
-  bool Failed() const {
-    return failed_test_case_count() > 0 || ad_hoc_test_result()->Failed();
-  }
-
-  // Gets the i-th test case among all the test cases. i can range from 0 to
-  // total_test_case_count() - 1. If i is not in that range, returns NULL.
-  const TestCase* GetTestCase(int i) const {
-    const int index = GetElementOr(test_case_indices_, i, -1);
-    return index < 0 ? NULL : test_cases_[i];
-  }
-
-  // Gets the i-th test case among all the test cases. i can range from 0 to
-  // total_test_case_count() - 1. If i is not in that range, returns NULL.
-  TestCase* GetMutableTestCase(int i) {
-    const int index = GetElementOr(test_case_indices_, i, -1);
-    return index < 0 ? NULL : test_cases_[index];
-  }
-
-  // Provides access to the event listener list.
-  TestEventListeners* listeners() { return &listeners_; }
-
-  // Returns the TestResult for the test that's currently running, or
-  // the TestResult for the ad hoc test if no test is running.
-  TestResult* current_test_result();
-
-  // Returns the TestResult for the ad hoc test.
-  const TestResult* ad_hoc_test_result() const { return &ad_hoc_test_result_; }
-
-  // Sets the OS stack trace getter.
-  //
-  // Does nothing if the input and the current OS stack trace getter
-  // are the same; otherwise, deletes the old getter and makes the
-  // input the current getter.
-  void set_os_stack_trace_getter(OsStackTraceGetterInterface* getter);
-
-  // Returns the current OS stack trace getter if it is not NULL;
-  // otherwise, creates an OsStackTraceGetter, makes it the current
-  // getter, and returns it.
-  OsStackTraceGetterInterface* os_stack_trace_getter();
-
-  // Returns the current OS stack trace as an std::string.
-  //
-  // The maximum number of stack frames to be included is specified by
-  // the gtest_stack_trace_depth flag.  The skip_count parameter
-  // specifies the number of top frames to be skipped, which doesn't
-  // count against the number of frames to be included.
-  //
-  // For example, if Foo() calls Bar(), which in turn calls
-  // CurrentOsStackTraceExceptTop(1), Foo() will be included in the
-  // trace but Bar() and CurrentOsStackTraceExceptTop() won't.
-  std::string CurrentOsStackTraceExceptTop(int skip_count) GTEST_NO_INLINE_;
-
-  // Finds and returns a TestCase with the given name.  If one doesn't
-  // exist, creates one and returns it.
-  //
-  // Arguments:
-  //
-  //   test_case_name: name of the test case
-  //   type_param:     the name of the test's type parameter, or NULL if
-  //                   this is not a typed or a type-parameterized test.
-  //   set_up_tc:      pointer to the function that sets up the test case
-  //   tear_down_tc:   pointer to the function that tears down the test case
-  TestCase* GetTestCase(const char* test_case_name,
-                        const char* type_param,
-                        Test::SetUpTestCaseFunc set_up_tc,
-                        Test::TearDownTestCaseFunc tear_down_tc);
-
-  // Adds a TestInfo to the unit test.
-  //
-  // Arguments:
-  //
-  //   set_up_tc:    pointer to the function that sets up the test case
-  //   tear_down_tc: pointer to the function that tears down the test case
-  //   test_info:    the TestInfo object
-  void AddTestInfo(Test::SetUpTestCaseFunc set_up_tc,
-                   Test::TearDownTestCaseFunc tear_down_tc,
-                   TestInfo* test_info) {
-    // In order to support thread-safe death tests, we need to
-    // remember the original working directory when the test program
-    // was first invoked.  We cannot do this in RUN_ALL_TESTS(), as
-    // the user may have changed the current directory before calling
-    // RUN_ALL_TESTS().  Therefore we capture the current directory in
-    // AddTestInfo(), which is called to register a TEST or TEST_F
-    // before main() is reached.
-    if (original_working_dir_.IsEmpty()) {
-      original_working_dir_.Set(FilePath::GetCurrentDir());
-      GTEST_CHECK_(!original_working_dir_.IsEmpty())
-          << "Failed to get the current working directory.";
-    }
-
-    GetTestCase(test_info->test_case_name(),
-                test_info->type_param(),
-                set_up_tc,
-                tear_down_tc)->AddTestInfo(test_info);
-  }
-
-  // Returns ParameterizedTestCaseRegistry object used to keep track of
-  // value-parameterized tests and instantiate and register them.
-  internal::ParameterizedTestCaseRegistry& parameterized_test_registry() {
-    return parameterized_test_registry_;
-  }
-
-  // Sets the TestCase object for the test that's currently running.
-  void set_current_test_case(TestCase* a_current_test_case) {
-    current_test_case_ = a_current_test_case;
-  }
-
-  // Sets the TestInfo object for the test that's currently running.  If
-  // current_test_info is NULL, the assertion results will be stored in
-  // ad_hoc_test_result_.
-  void set_current_test_info(TestInfo* a_current_test_info) {
-    current_test_info_ = a_current_test_info;
-  }
-
-  // Registers all parameterized tests defined using TEST_P and
-  // INSTANTIATE_TEST_CASE_P, creating regular tests for each test/parameter
-  // combination. This method can be called more then once; it has guards
-  // protecting from registering the tests more then once.  If
-  // value-parameterized tests are disabled, RegisterParameterizedTests is
-  // present but does nothing.
-  void RegisterParameterizedTests();
-
-  // Runs all tests in this UnitTest object, prints the result, and
-  // returns true if all tests are successful.  If any exception is
-  // thrown during a test, this test is considered to be failed, but
-  // the rest of the tests will still be run.
-  bool RunAllTests();
-
-  // Clears the results of all tests, except the ad hoc tests.
-  void ClearNonAdHocTestResult() {
-    ForEach(test_cases_, TestCase::ClearTestCaseResult);
-  }
-
-  // Clears the results of ad-hoc test assertions.
-  void ClearAdHocTestResult() {
-    ad_hoc_test_result_.Clear();
-  }
-
-  // Adds a TestProperty to the current TestResult object when invoked in a
-  // context of a test or a test case, or to the global property set. If the
-  // result already contains a property with the same key, the value will be
-  // updated.
-  void RecordProperty(const TestProperty& test_property);
-
-  enum ReactionToSharding {
-    HONOR_SHARDING_PROTOCOL,
-    IGNORE_SHARDING_PROTOCOL
-  };
-
-  // Matches the full name of each test against the user-specified
-  // filter to decide whether the test should run, then records the
-  // result in each TestCase and TestInfo object.
-  // If shard_tests == HONOR_SHARDING_PROTOCOL, further filters tests
-  // based on sharding variables in the environment.
-  // Returns the number of tests that should run.
-  int FilterTests(ReactionToSharding shard_tests);
-
-  // Prints the names of the tests matching the user-specified filter flag.
-  void ListTestsMatchingFilter();
-
-  const TestCase* current_test_case() const { return current_test_case_; }
-  TestInfo* current_test_info() { return current_test_info_; }
-  const TestInfo* current_test_info() const { return current_test_info_; }
-
-  // Returns the vector of environments that need to be set-up/torn-down
-  // before/after the tests are run.
-  std::vector<Environment*>& environments() { return environments_; }
-
-  // Getters for the per-thread Google Test trace stack.
-  std::vector<TraceInfo>& gtest_trace_stack() {
-    return *(gtest_trace_stack_.pointer());
-  }
-  const std::vector<TraceInfo>& gtest_trace_stack() const {
-    return gtest_trace_stack_.get();
-  }
-
-#if GTEST_HAS_DEATH_TEST
-  void InitDeathTestSubprocessControlInfo() {
-    internal_run_death_test_flag_.reset(ParseInternalRunDeathTestFlag());
-  }
-  // Returns a pointer to the parsed --gtest_internal_run_death_test
-  // flag, or NULL if that flag was not specified.
-  // This information is useful only in a death test child process.
-  // Must not be called before a call to InitGoogleTest.
-  const InternalRunDeathTestFlag* internal_run_death_test_flag() const {
-    return internal_run_death_test_flag_.get();
-  }
-
-  // Returns a pointer to the current death test factory.
-  internal::DeathTestFactory* death_test_factory() {
-    return death_test_factory_.get();
-  }
-
-  void SuppressTestEventsIfInSubprocess();
-
-  friend class ReplaceDeathTestFactory;
-#endif  // GTEST_HAS_DEATH_TEST
-
-  // Initializes the event listener performing XML output as specified by
-  // UnitTestOptions. Must not be called before InitGoogleTest.
-  void ConfigureXmlOutput();
-
-#if GTEST_CAN_STREAM_RESULTS_
-  // Initializes the event listener for streaming test results to a socket.
-  // Must not be called before InitGoogleTest.
-  void ConfigureStreamingOutput();
-#endif
-
-  // Performs initialization dependent upon flag values obtained in
-  // ParseGoogleTestFlagsOnly.  Is called from InitGoogleTest after the call to
-  // ParseGoogleTestFlagsOnly.  In case a user neglects to call InitGoogleTest
-  // this function is also called from RunAllTests.  Since this function can be
-  // called more than once, it has to be idempotent.
-  void PostFlagParsingInit();
-
-  // Gets the random seed used at the start of the current test iteration.
-  int random_seed() const { return random_seed_; }
-
-  // Gets the random number generator.
-  internal::Random* random() { return &random_; }
-
-  // Shuffles all test cases, and the tests within each test case,
-  // making sure that death tests are still run first.
-  void ShuffleTests();
-
-  // Restores the test cases and tests to their order before the first shuffle.
-  void UnshuffleTests();
-
-  // Returns the value of GTEST_FLAG(catch_exceptions) at the moment
-  // UnitTest::Run() starts.
-  bool catch_exceptions() const { return catch_exceptions_; }
-
- private:
-  friend class ::testing::UnitTest;
-
-  // Used by UnitTest::Run() to capture the state of
-  // GTEST_FLAG(catch_exceptions) at the moment it starts.
-  void set_catch_exceptions(bool value) { catch_exceptions_ = value; }
-
-  // The UnitTest object that owns this implementation object.
-  UnitTest* const parent_;
-
-  // The working directory when the first TEST() or TEST_F() was
-  // executed.
-  internal::FilePath original_working_dir_;
-
-  // The default test part result reporters.
-  DefaultGlobalTestPartResultReporter default_global_test_part_result_reporter_;
-  DefaultPerThreadTestPartResultReporter
-      default_per_thread_test_part_result_reporter_;
-
-  // Points to (but doesn't own) the global test part result reporter.
-  TestPartResultReporterInterface* global_test_part_result_repoter_;
-
-  // Protects read and write access to global_test_part_result_reporter_.
-  internal::Mutex global_test_part_result_reporter_mutex_;
-
-  // Points to (but doesn't own) the per-thread test part result reporter.
-  internal::ThreadLocal<TestPartResultReporterInterface*>
-      per_thread_test_part_result_reporter_;
-
-  // The vector of environments that need to be set-up/torn-down
-  // before/after the tests are run.
-  std::vector<Environment*> environments_;
-
-  // The vector of TestCases in their original order.  It owns the
-  // elements in the vector.
-  std::vector<TestCase*> test_cases_;
-
-  // Provides a level of indirection for the test case list to allow
-  // easy shuffling and restoring the test case order.  The i-th
-  // element of this vector is the index of the i-th test case in the
-  // shuffled order.
-  std::vector<int> test_case_indices_;
-
-  // ParameterizedTestRegistry object used to register value-parameterized
-  // tests.
-  internal::ParameterizedTestCaseRegistry parameterized_test_registry_;
-
-  // Indicates whether RegisterParameterizedTests() has been called already.
-  bool parameterized_tests_registered_;
-
-  // Index of the last death test case registered.  Initially -1.
-  int last_death_test_case_;
-
-  // This points to the TestCase for the currently running test.  It
-  // changes as Google Test goes through one test case after another.
-  // When no test is running, this is set to NULL and Google Test
-  // stores assertion results in ad_hoc_test_result_.  Initially NULL.
-  TestCase* current_test_case_;
-
-  // This points to the TestInfo for the currently running test.  It
-  // changes as Google Test goes through one test after another.  When
-  // no test is running, this is set to NULL and Google Test stores
-  // assertion results in ad_hoc_test_result_.  Initially NULL.
-  TestInfo* current_test_info_;
-
-  // Normally, a user only writes assertions inside a TEST or TEST_F,
-  // or inside a function called by a TEST or TEST_F.  Since Google
-  // Test keeps track of which test is current running, it can
-  // associate such an assertion with the test it belongs to.
-  //
-  // If an assertion is encountered when no TEST or TEST_F is running,
-  // Google Test attributes the assertion result to an imaginary "ad hoc"
-  // test, and records the result in ad_hoc_test_result_.
-  TestResult ad_hoc_test_result_;
-
-  // The list of event listeners that can be used to track events inside
-  // Google Test.
-  TestEventListeners listeners_;
-
-  // The OS stack trace getter.  Will be deleted when the UnitTest
-  // object is destructed.  By default, an OsStackTraceGetter is used,
-  // but the user can set this field to use a custom getter if that is
-  // desired.
-  OsStackTraceGetterInterface* os_stack_trace_getter_;
-
-  // True iff PostFlagParsingInit() has been called.
-  bool post_flag_parse_init_performed_;
-
-  // The random number seed used at the beginning of the test run.
-  int random_seed_;
-
-  // Our random number generator.
-  internal::Random random_;
-
-  // The time of the test program start, in ms from the start of the
-  // UNIX epoch.
-  TimeInMillis start_timestamp_;
-
-  // How long the test took to run, in milliseconds.
-  TimeInMillis elapsed_time_;
-
-#if GTEST_HAS_DEATH_TEST
-  // The decomposed components of the gtest_internal_run_death_test flag,
-  // parsed when RUN_ALL_TESTS is called.
-  internal::scoped_ptr<InternalRunDeathTestFlag> internal_run_death_test_flag_;
-  internal::scoped_ptr<internal::DeathTestFactory> death_test_factory_;
-#endif  // GTEST_HAS_DEATH_TEST
-
-  // A per-thread stack of traces created by the SCOPED_TRACE() macro.
-  internal::ThreadLocal<std::vector<TraceInfo> > gtest_trace_stack_;
-
-  // The value of GTEST_FLAG(catch_exceptions) at the moment RunAllTests()
-  // starts.
-  bool catch_exceptions_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(UnitTestImpl);
-};  // class UnitTestImpl
-
-// Convenience function for accessing the global UnitTest
-// implementation object.
-inline UnitTestImpl* GetUnitTestImpl() {
-  return UnitTest::GetInstance()->impl();
-}
-
-#if GTEST_USES_SIMPLE_RE
-
-// Internal helper functions for implementing the simple regular
-// expression matcher.
-GTEST_API_ bool IsInSet(char ch, const char* str);
-GTEST_API_ bool IsAsciiDigit(char ch);
-GTEST_API_ bool IsAsciiPunct(char ch);
-GTEST_API_ bool IsRepeat(char ch);
-GTEST_API_ bool IsAsciiWhiteSpace(char ch);
-GTEST_API_ bool IsAsciiWordChar(char ch);
-GTEST_API_ bool IsValidEscape(char ch);
-GTEST_API_ bool AtomMatchesChar(bool escaped, char pattern, char ch);
-GTEST_API_ bool ValidateRegex(const char* regex);
-GTEST_API_ bool MatchRegexAtHead(const char* regex, const char* str);
-GTEST_API_ bool MatchRepetitionAndRegexAtHead(
-    bool escaped, char ch, char repeat, const char* regex, const char* str);
-GTEST_API_ bool MatchRegexAnywhere(const char* regex, const char* str);
-
-#endif  // GTEST_USES_SIMPLE_RE
-
-// Parses the command line for Google Test flags, without initializing
-// other parts of Google Test.
-GTEST_API_ void ParseGoogleTestFlagsOnly(int* argc, char** argv);
-GTEST_API_ void ParseGoogleTestFlagsOnly(int* argc, wchar_t** argv);
-
-#if GTEST_HAS_DEATH_TEST
-
-// Returns the message describing the last system error, regardless of the
-// platform.
-GTEST_API_ std::string GetLastErrnoDescription();
-
-// Attempts to parse a string into a positive integer pointed to by the
-// number parameter.  Returns true if that is possible.
-// GTEST_HAS_DEATH_TEST implies that we have ::std::string, so we can use
-// it here.
-template <typename Integer>
-bool ParseNaturalNumber(const ::std::string& str, Integer* number) {
-  // Fail fast if the given string does not begin with a digit;
-  // this bypasses strtoXXX's "optional leading whitespace and plus
-  // or minus sign" semantics, which are undesirable here.
-  if (str.empty() || !IsDigit(str[0])) {
-    return false;
-  }
-  errno = 0;
-
-  char* end;
-  // BiggestConvertible is the largest integer type that system-provided
-  // string-to-number conversion routines can return.
-
-# if GTEST_OS_WINDOWS && !defined(__GNUC__)
-
-  // MSVC and C++ Builder define __int64 instead of the standard long long.
-  typedef unsigned __int64 BiggestConvertible;
-  const BiggestConvertible parsed = _strtoui64(str.c_str(), &end, 10);
-
-# else
-
-  typedef unsigned long long BiggestConvertible;  // NOLINT
-  const BiggestConvertible parsed = strtoull(str.c_str(), &end, 10);
-
-# endif  // GTEST_OS_WINDOWS && !defined(__GNUC__)
-
-  const bool parse_success = *end == '\0' && errno == 0;
-
-  // TODO(vladl@google.com): Convert this to compile time assertion when it is
-  // available.
-  GTEST_CHECK_(sizeof(Integer) <= sizeof(parsed));
-
-  const Integer result = static_cast<Integer>(parsed);
-  if (parse_success && static_cast<BiggestConvertible>(result) == parsed) {
-    *number = result;
-    return true;
-  }
-  return false;
-}
-#endif  // GTEST_HAS_DEATH_TEST
-
-// TestResult contains some private methods that should be hidden from
-// Google Test user but are required for testing. This class allow our tests
-// to access them.
-//
-// This class is supplied only for the purpose of testing Google Test's own
-// constructs. Do not use it in user tests, either directly or indirectly.
-class TestResultAccessor {
- public:
-  static void RecordProperty(TestResult* test_result,
-                             const std::string& xml_element,
-                             const TestProperty& property) {
-    test_result->RecordProperty(xml_element, property);
-  }
-
-  static void ClearTestPartResults(TestResult* test_result) {
-    test_result->ClearTestPartResults();
-  }
-
-  static const std::vector<testing::TestPartResult>& test_part_results(
-      const TestResult& test_result) {
-    return test_result.test_part_results();
-  }
-};
-
-#if GTEST_CAN_STREAM_RESULTS_
-
-// Streams test results to the given port on the given host machine.
-class StreamingListener : public EmptyTestEventListener {
- public:
-  // Abstract base class for writing strings to a socket.
-  class AbstractSocketWriter {
-   public:
-    virtual ~AbstractSocketWriter() {}
-
-    // Sends a string to the socket.
-    virtual void Send(const std::string& message) = 0;
-
-    // Closes the socket.
-    virtual void CloseConnection() {}
-
-    // Sends a string and a newline to the socket.
-    void SendLn(const std::string& message) { Send(message + "\n"); }
-  };
-
-  // Concrete class for actually writing strings to a socket.
-  class SocketWriter : public AbstractSocketWriter {
-   public:
-    SocketWriter(const std::string& host, const std::string& port)
-        : sockfd_(-1), host_name_(host), port_num_(port) {
-      MakeConnection();
-    }
-
-    virtual ~SocketWriter() {
-      if (sockfd_ != -1)
-        CloseConnection();
-    }
-
-    // Sends a string to the socket.
-    virtual void Send(const std::string& message) {
-      GTEST_CHECK_(sockfd_ != -1)
-          << "Send() can be called only when there is a connection.";
-
-      const int len = static_cast<int>(message.length());
-      if (write(sockfd_, message.c_str(), len) != len) {
-        GTEST_LOG_(WARNING)
-            << "stream_result_to: failed to stream to "
-            << host_name_ << ":" << port_num_;
-      }
-    }
-
-   private:
-    // Creates a client socket and connects to the server.
-    void MakeConnection();
-
-    // Closes the socket.
-    void CloseConnection() {
-      GTEST_CHECK_(sockfd_ != -1)
-          << "CloseConnection() can be called only when there is a connection.";
-
-      close(sockfd_);
-      sockfd_ = -1;
-    }
-
-    int sockfd_;  // socket file descriptor
-    const std::string host_name_;
-    const std::string port_num_;
-
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(SocketWriter);
-  };  // class SocketWriter
-
-  // Escapes '=', '&', '%', and '\n' characters in str as "%xx".
-  static std::string UrlEncode(const char* str);
-
-  StreamingListener(const std::string& host, const std::string& port)
-      : socket_writer_(new SocketWriter(host, port)) {
-    Start();
-  }
-
-  explicit StreamingListener(AbstractSocketWriter* socket_writer)
-      : socket_writer_(socket_writer) { Start(); }
-
-  void OnTestProgramStart(const UnitTest& /* unit_test */) {
-    SendLn("event=TestProgramStart");
-  }
-
-  void OnTestProgramEnd(const UnitTest& unit_test) {
-    // Note that Google Test current only report elapsed time for each
-    // test iteration, not for the entire test program.
-    SendLn("event=TestProgramEnd&passed=" + FormatBool(unit_test.Passed()));
-
-    // Notify the streaming server to stop.
-    socket_writer_->CloseConnection();
-  }
-
-  void OnTestIterationStart(const UnitTest& /* unit_test */, int iteration) {
-    SendLn("event=TestIterationStart&iteration=" +
-           StreamableToString(iteration));
-  }
-
-  void OnTestIterationEnd(const UnitTest& unit_test, int /* iteration */) {
-    SendLn("event=TestIterationEnd&passed=" +
-           FormatBool(unit_test.Passed()) + "&elapsed_time=" +
-           StreamableToString(unit_test.elapsed_time()) + "ms");
-  }
-
-  void OnTestCaseStart(const TestCase& test_case) {
-    SendLn(std::string("event=TestCaseStart&name=") + test_case.name());
-  }
-
-  void OnTestCaseEnd(const TestCase& test_case) {
-    SendLn("event=TestCaseEnd&passed=" + FormatBool(test_case.Passed())
-           + "&elapsed_time=" + StreamableToString(test_case.elapsed_time())
-           + "ms");
-  }
-
-  void OnTestStart(const TestInfo& test_info) {
-    SendLn(std::string("event=TestStart&name=") + test_info.name());
-  }
-
-  void OnTestEnd(const TestInfo& test_info) {
-    SendLn("event=TestEnd&passed=" +
-           FormatBool((test_info.result())->Passed()) +
-           "&elapsed_time=" +
-           StreamableToString((test_info.result())->elapsed_time()) + "ms");
-  }
-
-  void OnTestPartResult(const TestPartResult& test_part_result) {
-    const char* file_name = test_part_result.file_name();
-    if (file_name == NULL)
-      file_name = "";
-    SendLn("event=TestPartResult&file=" + UrlEncode(file_name) +
-           "&line=" + StreamableToString(test_part_result.line_number()) +
-           "&message=" + UrlEncode(test_part_result.message()));
-  }
-
- private:
-  // Sends the given message and a newline to the socket.
-  void SendLn(const std::string& message) { socket_writer_->SendLn(message); }
-
-  // Called at the start of streaming to notify the receiver what
-  // protocol we are using.
-  void Start() { SendLn("gtest_streaming_protocol_version=1.0"); }
-
-  std::string FormatBool(bool value) { return value ? "1" : "0"; }
-
-  const scoped_ptr<AbstractSocketWriter> socket_writer_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(StreamingListener);
-};  // class StreamingListener
-
-#endif  // GTEST_CAN_STREAM_RESULTS_
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GTEST_SRC_GTEST_INTERNAL_INL_H_
-
-#if GTEST_OS_WINDOWS
-# define vsnprintf _vsnprintf
-#endif  // GTEST_OS_WINDOWS
-
-#if GTEST_HAS_ABSL
-#include "absl/debugging/failure_signal_handler.h"
-#include "absl/debugging/stacktrace.h"
-#include "absl/debugging/symbolize.h"
-#include "absl/strings/str_cat.h"
-#endif  // GTEST_HAS_ABSL
-
-namespace testing {
-
-using internal::CountIf;
-using internal::ForEach;
-using internal::GetElementOr;
-using internal::Shuffle;
-
-// Constants.
-
-// A test whose test case name or test name matches this filter is
-// disabled and not run.
-static const char kDisableTestFilter[] = "DISABLED_*:*/DISABLED_*";
-
-// A test case whose name matches this filter is considered a death
-// test case and will be run before test cases whose name doesn't
-// match this filter.
-static const char kDeathTestCaseFilter[] = "*DeathTest:*DeathTest/*";
-
-// A test filter that matches everything.
-static const char kUniversalFilter[] = "*";
-
-// The default output format.
-static const char kDefaultOutputFormat[] = "xml";
-// The default output file.
-static const char kDefaultOutputFile[] = "test_detail";
-
-// The environment variable name for the test shard index.
-static const char kTestShardIndex[] = "GTEST_SHARD_INDEX";
-// The environment variable name for the total number of test shards.
-static const char kTestTotalShards[] = "GTEST_TOTAL_SHARDS";
-// The environment variable name for the test shard status file.
-static const char kTestShardStatusFile[] = "GTEST_SHARD_STATUS_FILE";
-
-namespace internal {
-
-// The text used in failure messages to indicate the start of the
-// stack trace.
-const char kStackTraceMarker[] = "\nStack trace:\n";
-
-// g_help_flag is true iff the --help flag or an equivalent form is
-// specified on the command line.
-bool g_help_flag = false;
-
-}  // namespace internal
-
-// Bazel passes in the argument to '--test_filter' via the TESTBRIDGE_TEST_ONLY
-// environment variable.
-static const char* GetDefaultFilter() {
-  const char* const testbridge_test_only =
-      internal::posix::GetEnv("TESTBRIDGE_TEST_ONLY");
-  if (testbridge_test_only != NULL) {
-    return testbridge_test_only;
-  }
-  return kUniversalFilter;
-}
-
-GTEST_DEFINE_bool_(
-    also_run_disabled_tests,
-    internal::BoolFromGTestEnv("also_run_disabled_tests", false),
-    "Run disabled tests too, in addition to the tests normally being run.");
-
-GTEST_DEFINE_bool_(
-    break_on_failure,
-    internal::BoolFromGTestEnv("break_on_failure", false),
-    "True iff a failed assertion should be a debugger break-point.");
-
-GTEST_DEFINE_bool_(
-    catch_exceptions,
-    internal::BoolFromGTestEnv("catch_exceptions", true),
-    "True iff " GTEST_NAME_
-    " should catch exceptions and treat them as test failures.");
-
-GTEST_DEFINE_string_(
-    color,
-    internal::StringFromGTestEnv("color", "auto"),
-    "Whether to use colors in the output.  Valid values: yes, no, "
-    "and auto.  'auto' means to use colors if the output is "
-    "being sent to a terminal and the TERM environment variable "
-    "is set to a terminal type that supports colors.");
-
-GTEST_DEFINE_string_(
-    filter,
-    internal::StringFromGTestEnv("filter", GetDefaultFilter()),
-    "A colon-separated list of glob (not regex) patterns "
-    "for filtering the tests to run, optionally followed by a "
-    "'-' and a : separated list of negative patterns (tests to "
-    "exclude).  A test is run if it matches one of the positive "
-    "patterns and does not match any of the negative patterns.");
-
-GTEST_DEFINE_bool_(
-    install_failure_signal_handler,
-    internal::BoolFromGTestEnv("install_failure_signal_handler", false),
-    "If true and supported on the current platform, " GTEST_NAME_ " should "
-    "install a signal handler that dumps debugging information when fatal "
-    "signals are raised.");
-
-GTEST_DEFINE_bool_(list_tests, false,
-                   "List all tests without running them.");
-
-// The net priority order after flag processing is thus:
-//   --gtest_output command line flag
-//   GTEST_OUTPUT environment variable
-//   XML_OUTPUT_FILE environment variable
-//   ''
-GTEST_DEFINE_string_(
-    output,
-    internal::StringFromGTestEnv("output",
-      internal::OutputFlagAlsoCheckEnvVar().c_str()),
-    "A format (defaults to \"xml\" but can be specified to be \"json\"), "
-    "optionally followed by a colon and an output file name or directory. "
-    "A directory is indicated by a trailing pathname separator. "
-    "Examples: \"xml:filename.xml\", \"xml::directoryname/\". "
-    "If a directory is specified, output files will be created "
-    "within that directory, with file-names based on the test "
-    "executable's name and, if necessary, made unique by adding "
-    "digits.");
-
-GTEST_DEFINE_bool_(
-    print_time,
-    internal::BoolFromGTestEnv("print_time", true),
-    "True iff " GTEST_NAME_
-    " should display elapsed time in text output.");
-
-GTEST_DEFINE_bool_(
-    print_utf8,
-    internal::BoolFromGTestEnv("print_utf8", true),
-    "True iff " GTEST_NAME_
-    " prints UTF8 characters as text.");
-
-GTEST_DEFINE_int32_(
-    random_seed,
-    internal::Int32FromGTestEnv("random_seed", 0),
-    "Random number seed to use when shuffling test orders.  Must be in range "
-    "[1, 99999], or 0 to use a seed based on the current time.");
-
-GTEST_DEFINE_int32_(
-    repeat,
-    internal::Int32FromGTestEnv("repeat", 1),
-    "How many times to repeat each test.  Specify a negative number "
-    "for repeating forever.  Useful for shaking out flaky tests.");
-
-GTEST_DEFINE_bool_(
-    show_internal_stack_frames, false,
-    "True iff " GTEST_NAME_ " should include internal stack frames when "
-    "printing test failure stack traces.");
-
-GTEST_DEFINE_bool_(
-    shuffle,
-    internal::BoolFromGTestEnv("shuffle", false),
-    "True iff " GTEST_NAME_
-    " should randomize tests' order on every run.");
-
-GTEST_DEFINE_int32_(
-    stack_trace_depth,
-    internal::Int32FromGTestEnv("stack_trace_depth", kMaxStackTraceDepth),
-    "The maximum number of stack frames to print when an "
-    "assertion fails.  The valid range is 0 through 100, inclusive.");
-
-GTEST_DEFINE_string_(
-    stream_result_to,
-    internal::StringFromGTestEnv("stream_result_to", ""),
-    "This flag specifies the host name and the port number on which to stream "
-    "test results. Example: \"localhost:555\". The flag is effective only on "
-    "Linux.");
-
-GTEST_DEFINE_bool_(
-    throw_on_failure,
-    internal::BoolFromGTestEnv("throw_on_failure", false),
-    "When this flag is specified, a failed assertion will throw an exception "
-    "if exceptions are enabled or exit the program with a non-zero code "
-    "otherwise. For use with an external test framework.");
-
-#if GTEST_USE_OWN_FLAGFILE_FLAG_
-GTEST_DEFINE_string_(
-    flagfile,
-    internal::StringFromGTestEnv("flagfile", ""),
-    "This flag specifies the flagfile to read command-line flags from.");
-#endif  // GTEST_USE_OWN_FLAGFILE_FLAG_
-
-namespace internal {
-
-// Generates a random number from [0, range), using a Linear
-// Congruential Generator (LCG).  Crashes if 'range' is 0 or greater
-// than kMaxRange.
-UInt32 Random::Generate(UInt32 range) {
-  // These constants are the same as are used in glibc's rand(3).
-  // Use wider types than necessary to prevent unsigned overflow diagnostics.
-  state_ = static_cast<UInt32>(1103515245ULL*state_ + 12345U) % kMaxRange;
-
-  GTEST_CHECK_(range > 0)
-      << "Cannot generate a number in the range [0, 0).";
-  GTEST_CHECK_(range <= kMaxRange)
-      << "Generation of a number in [0, " << range << ") was requested, "
-      << "but this can only generate numbers in [0, " << kMaxRange << ").";
-
-  // Converting via modulus introduces a bit of downward bias, but
-  // it's simple, and a linear congruential generator isn't too good
-  // to begin with.
-  return state_ % range;
-}
-
-// GTestIsInitialized() returns true iff the user has initialized
-// Google Test.  Useful for catching the user mistake of not initializing
-// Google Test before calling RUN_ALL_TESTS().
-static bool GTestIsInitialized() { return GetArgvs().size() > 0; }
-
-// Iterates over a vector of TestCases, keeping a running sum of the
-// results of calling a given int-returning method on each.
-// Returns the sum.
-static int SumOverTestCaseList(const std::vector<TestCase*>& case_list,
-                               int (TestCase::*method)() const) {
-  int sum = 0;
-  for (size_t i = 0; i < case_list.size(); i++) {
-    sum += (case_list[i]->*method)();
-  }
-  return sum;
-}
-
-// Returns true iff the test case passed.
-static bool TestCasePassed(const TestCase* test_case) {
-  return test_case->should_run() && test_case->Passed();
-}
-
-// Returns true iff the test case failed.
-static bool TestCaseFailed(const TestCase* test_case) {
-  return test_case->should_run() && test_case->Failed();
-}
-
-// Returns true iff test_case contains at least one test that should
-// run.
-static bool ShouldRunTestCase(const TestCase* test_case) {
-  return test_case->should_run();
-}
-
-// AssertHelper constructor.
-AssertHelper::AssertHelper(TestPartResult::Type type,
-                           const char* file,
-                           int line,
-                           const char* message)
-    : data_(new AssertHelperData(type, file, line, message)) {
-}
-
-AssertHelper::~AssertHelper() {
-  delete data_;
-}
-
-// Message assignment, for assertion streaming support.
-void AssertHelper::operator=(const Message& message) const {
-  UnitTest::GetInstance()->
-    AddTestPartResult(data_->type, data_->file, data_->line,
-                      AppendUserMessage(data_->message, message),
-                      UnitTest::GetInstance()->impl()
-                      ->CurrentOsStackTraceExceptTop(1)
-                      // Skips the stack frame for this function itself.
-                      );  // NOLINT
-}
-
-// Mutex for linked pointers.
-GTEST_API_ GTEST_DEFINE_STATIC_MUTEX_(g_linked_ptr_mutex);
-
-// A copy of all command line arguments.  Set by InitGoogleTest().
-::std::vector<std::string> g_argvs;
-
-::std::vector<std::string> GetArgvs() {
-#if defined(GTEST_CUSTOM_GET_ARGVS_)
-  // GTEST_CUSTOM_GET_ARGVS_() may return a container of std::string or
-  // ::string. This code converts it to the appropriate type.
-  const auto& custom = GTEST_CUSTOM_GET_ARGVS_();
-  return ::std::vector<std::string>(custom.begin(), custom.end());
-#else   // defined(GTEST_CUSTOM_GET_ARGVS_)
-  return g_argvs;
-#endif  // defined(GTEST_CUSTOM_GET_ARGVS_)
-}
-
-// Returns the current application's name, removing directory path if that
-// is present.
-FilePath GetCurrentExecutableName() {
-  FilePath result;
-
-#if GTEST_OS_WINDOWS
-  result.Set(FilePath(GetArgvs()[0]).RemoveExtension("exe"));
-#else
-  result.Set(FilePath(GetArgvs()[0]));
-#endif  // GTEST_OS_WINDOWS
-
-  return result.RemoveDirectoryName();
-}
-
-// Functions for processing the gtest_output flag.
-
-// Returns the output format, or "" for normal printed output.
-std::string UnitTestOptions::GetOutputFormat() {
-  const char* const gtest_output_flag = GTEST_FLAG(output).c_str();
-  if (gtest_output_flag == NULL) return std::string("");
-
-  const char* const colon = strchr(gtest_output_flag, ':');
-  return (colon == NULL) ?
-      std::string(gtest_output_flag) :
-      std::string(gtest_output_flag, colon - gtest_output_flag);
-}
-
-// Returns the name of the requested output file, or the default if none
-// was explicitly specified.
-std::string UnitTestOptions::GetAbsolutePathToOutputFile() {
-  const char* const gtest_output_flag = GTEST_FLAG(output).c_str();
-  if (gtest_output_flag == NULL)
-    return "";
-
-  std::string format = GetOutputFormat();
-  if (format.empty())
-    format = std::string(kDefaultOutputFormat);
-
-  const char* const colon = strchr(gtest_output_flag, ':');
-  if (colon == NULL)
-    return internal::FilePath::MakeFileName(
-        internal::FilePath(
-            UnitTest::GetInstance()->original_working_dir()),
-        internal::FilePath(kDefaultOutputFile), 0,
-        format.c_str()).string();
-
-  internal::FilePath output_name(colon + 1);
-  if (!output_name.IsAbsolutePath())
-    // TODO(wan@google.com): on Windows \some\path is not an absolute
-    // path (as its meaning depends on the current drive), yet the
-    // following logic for turning it into an absolute path is wrong.
-    // Fix it.
-    output_name = internal::FilePath::ConcatPaths(
-        internal::FilePath(UnitTest::GetInstance()->original_working_dir()),
-        internal::FilePath(colon + 1));
-
-  if (!output_name.IsDirectory())
-    return output_name.string();
-
-  internal::FilePath result(internal::FilePath::GenerateUniqueFileName(
-      output_name, internal::GetCurrentExecutableName(),
-      GetOutputFormat().c_str()));
-  return result.string();
-}
-
-// Returns true iff the wildcard pattern matches the string.  The
-// first ':' or '\0' character in pattern marks the end of it.
-//
-// This recursive algorithm isn't very efficient, but is clear and
-// works well enough for matching test names, which are short.
-bool UnitTestOptions::PatternMatchesString(const char *pattern,
-                                           const char *str) {
-  switch (*pattern) {
-    case '\0':
-    case ':':  // Either ':' or '\0' marks the end of the pattern.
-      return *str == '\0';
-    case '?':  // Matches any single character.
-      return *str != '\0' && PatternMatchesString(pattern + 1, str + 1);
-    case '*':  // Matches any string (possibly empty) of characters.
-      return (*str != '\0' && PatternMatchesString(pattern, str + 1)) ||
-          PatternMatchesString(pattern + 1, str);
-    default:  // Non-special character.  Matches itself.
-      return *pattern == *str &&
-          PatternMatchesString(pattern + 1, str + 1);
-  }
-}
-
-bool UnitTestOptions::MatchesFilter(
-    const std::string& name, const char* filter) {
-  const char *cur_pattern = filter;
-  for (;;) {
-    if (PatternMatchesString(cur_pattern, name.c_str())) {
-      return true;
-    }
-
-    // Finds the next pattern in the filter.
-    cur_pattern = strchr(cur_pattern, ':');
-
-    // Returns if no more pattern can be found.
-    if (cur_pattern == NULL) {
-      return false;
-    }
-
-    // Skips the pattern separater (the ':' character).
-    cur_pattern++;
-  }
-}
-
-// Returns true iff the user-specified filter matches the test case
-// name and the test name.
-bool UnitTestOptions::FilterMatchesTest(const std::string &test_case_name,
-                                        const std::string &test_name) {
-  const std::string& full_name = test_case_name + "." + test_name.c_str();
-
-  // Split --gtest_filter at '-', if there is one, to separate into
-  // positive filter and negative filter portions
-  const char* const p = GTEST_FLAG(filter).c_str();
-  const char* const dash = strchr(p, '-');
-  std::string positive;
-  std::string negative;
-  if (dash == NULL) {
-    positive = GTEST_FLAG(filter).c_str();  // Whole string is a positive filter
-    negative = "";
-  } else {
-    positive = std::string(p, dash);   // Everything up to the dash
-    negative = std::string(dash + 1);  // Everything after the dash
-    if (positive.empty()) {
-      // Treat '-test1' as the same as '*-test1'
-      positive = kUniversalFilter;
-    }
-  }
-
-  // A filter is a colon-separated list of patterns.  It matches a
-  // test if any pattern in it matches the test.
-  return (MatchesFilter(full_name, positive.c_str()) &&
-          !MatchesFilter(full_name, negative.c_str()));
-}
-
-#if GTEST_HAS_SEH
-// Returns EXCEPTION_EXECUTE_HANDLER if Google Test should handle the
-// given SEH exception, or EXCEPTION_CONTINUE_SEARCH otherwise.
-// This function is useful as an __except condition.
-int UnitTestOptions::GTestShouldProcessSEH(DWORD exception_code) {
-  // Google Test should handle a SEH exception if:
-  //   1. the user wants it to, AND
-  //   2. this is not a breakpoint exception, AND
-  //   3. this is not a C++ exception (VC++ implements them via SEH,
-  //      apparently).
-  //
-  // SEH exception code for C++ exceptions.
-  // (see http://support.microsoft.com/kb/185294 for more information).
-  const DWORD kCxxExceptionCode = 0xe06d7363;
-
-  bool should_handle = true;
-
-  if (!GTEST_FLAG(catch_exceptions))
-    should_handle = false;
-  else if (exception_code == EXCEPTION_BREAKPOINT)
-    should_handle = false;
-  else if (exception_code == kCxxExceptionCode)
-    should_handle = false;
-
-  return should_handle ? EXCEPTION_EXECUTE_HANDLER : EXCEPTION_CONTINUE_SEARCH;
-}
-#endif  // GTEST_HAS_SEH
-
-}  // namespace internal
-
-// The c'tor sets this object as the test part result reporter used by
-// Google Test.  The 'result' parameter specifies where to report the
-// results. Intercepts only failures from the current thread.
-ScopedFakeTestPartResultReporter::ScopedFakeTestPartResultReporter(
-    TestPartResultArray* result)
-    : intercept_mode_(INTERCEPT_ONLY_CURRENT_THREAD),
-      result_(result) {
-  Init();
-}
-
-// The c'tor sets this object as the test part result reporter used by
-// Google Test.  The 'result' parameter specifies where to report the
-// results.
-ScopedFakeTestPartResultReporter::ScopedFakeTestPartResultReporter(
-    InterceptMode intercept_mode, TestPartResultArray* result)
-    : intercept_mode_(intercept_mode),
-      result_(result) {
-  Init();
-}
-
-void ScopedFakeTestPartResultReporter::Init() {
-  internal::UnitTestImpl* const impl = internal::GetUnitTestImpl();
-  if (intercept_mode_ == INTERCEPT_ALL_THREADS) {
-    old_reporter_ = impl->GetGlobalTestPartResultReporter();
-    impl->SetGlobalTestPartResultReporter(this);
-  } else {
-    old_reporter_ = impl->GetTestPartResultReporterForCurrentThread();
-    impl->SetTestPartResultReporterForCurrentThread(this);
-  }
-}
-
-// The d'tor restores the test part result reporter used by Google Test
-// before.
-ScopedFakeTestPartResultReporter::~ScopedFakeTestPartResultReporter() {
-  internal::UnitTestImpl* const impl = internal::GetUnitTestImpl();
-  if (intercept_mode_ == INTERCEPT_ALL_THREADS) {
-    impl->SetGlobalTestPartResultReporter(old_reporter_);
-  } else {
-    impl->SetTestPartResultReporterForCurrentThread(old_reporter_);
-  }
-}
-
-// Increments the test part result count and remembers the result.
-// This method is from the TestPartResultReporterInterface interface.
-void ScopedFakeTestPartResultReporter::ReportTestPartResult(
-    const TestPartResult& result) {
-  result_->Append(result);
-}
-
-namespace internal {
-
-// Returns the type ID of ::testing::Test.  We should always call this
-// instead of GetTypeId< ::testing::Test>() to get the type ID of
-// testing::Test.  This is to work around a suspected linker bug when
-// using Google Test as a framework on Mac OS X.  The bug causes
-// GetTypeId< ::testing::Test>() to return different values depending
-// on whether the call is from the Google Test framework itself or
-// from user test code.  GetTestTypeId() is guaranteed to always
-// return the same value, as it always calls GetTypeId<>() from the
-// gtest.cc, which is within the Google Test framework.
-TypeId GetTestTypeId() {
-  return GetTypeId<Test>();
-}
-
-// The value of GetTestTypeId() as seen from within the Google Test
-// library.  This is solely for testing GetTestTypeId().
-extern const TypeId kTestTypeIdInGoogleTest = GetTestTypeId();
-
-// This predicate-formatter checks that 'results' contains a test part
-// failure of the given type and that the failure message contains the
-// given substring.
-static AssertionResult HasOneFailure(const char* /* results_expr */,
-                                     const char* /* type_expr */,
-                                     const char* /* substr_expr */,
-                                     const TestPartResultArray& results,
-                                     TestPartResult::Type type,
-                                     const std::string& substr) {
-  const std::string expected(type == TestPartResult::kFatalFailure ?
-                        "1 fatal failure" :
-                        "1 non-fatal failure");
-  Message msg;
-  if (results.size() != 1) {
-    msg << "Expected: " << expected << "\n"
-        << "  Actual: " << results.size() << " failures";
-    for (int i = 0; i < results.size(); i++) {
-      msg << "\n" << results.GetTestPartResult(i);
-    }
-    return AssertionFailure() << msg;
-  }
-
-  const TestPartResult& r = results.GetTestPartResult(0);
-  if (r.type() != type) {
-    return AssertionFailure() << "Expected: " << expected << "\n"
-                              << "  Actual:\n"
-                              << r;
-  }
-
-  if (strstr(r.message(), substr.c_str()) == NULL) {
-    return AssertionFailure() << "Expected: " << expected << " containing \""
-                              << substr << "\"\n"
-                              << "  Actual:\n"
-                              << r;
-  }
-
-  return AssertionSuccess();
-}
-
-// The constructor of SingleFailureChecker remembers where to look up
-// test part results, what type of failure we expect, and what
-// substring the failure message should contain.
-SingleFailureChecker::SingleFailureChecker(const TestPartResultArray* results,
-                                           TestPartResult::Type type,
-                                           const std::string& substr)
-    : results_(results), type_(type), substr_(substr) {}
-
-// The destructor of SingleFailureChecker verifies that the given
-// TestPartResultArray contains exactly one failure that has the given
-// type and contains the given substring.  If that's not the case, a
-// non-fatal failure will be generated.
-SingleFailureChecker::~SingleFailureChecker() {
-  EXPECT_PRED_FORMAT3(HasOneFailure, *results_, type_, substr_);
-}
-
-DefaultGlobalTestPartResultReporter::DefaultGlobalTestPartResultReporter(
-    UnitTestImpl* unit_test) : unit_test_(unit_test) {}
-
-void DefaultGlobalTestPartResultReporter::ReportTestPartResult(
-    const TestPartResult& result) {
-  unit_test_->current_test_result()->AddTestPartResult(result);
-  unit_test_->listeners()->repeater()->OnTestPartResult(result);
-}
-
-DefaultPerThreadTestPartResultReporter::DefaultPerThreadTestPartResultReporter(
-    UnitTestImpl* unit_test) : unit_test_(unit_test) {}
-
-void DefaultPerThreadTestPartResultReporter::ReportTestPartResult(
-    const TestPartResult& result) {
-  unit_test_->GetGlobalTestPartResultReporter()->ReportTestPartResult(result);
-}
-
-// Returns the global test part result reporter.
-TestPartResultReporterInterface*
-UnitTestImpl::GetGlobalTestPartResultReporter() {
-  internal::MutexLock lock(&global_test_part_result_reporter_mutex_);
-  return global_test_part_result_repoter_;
-}
-
-// Sets the global test part result reporter.
-void UnitTestImpl::SetGlobalTestPartResultReporter(
-    TestPartResultReporterInterface* reporter) {
-  internal::MutexLock lock(&global_test_part_result_reporter_mutex_);
-  global_test_part_result_repoter_ = reporter;
-}
-
-// Returns the test part result reporter for the current thread.
-TestPartResultReporterInterface*
-UnitTestImpl::GetTestPartResultReporterForCurrentThread() {
-  return per_thread_test_part_result_reporter_.get();
-}
-
-// Sets the test part result reporter for the current thread.
-void UnitTestImpl::SetTestPartResultReporterForCurrentThread(
-    TestPartResultReporterInterface* reporter) {
-  per_thread_test_part_result_reporter_.set(reporter);
-}
-
-// Gets the number of successful test cases.
-int UnitTestImpl::successful_test_case_count() const {
-  return CountIf(test_cases_, TestCasePassed);
-}
-
-// Gets the number of failed test cases.
-int UnitTestImpl::failed_test_case_count() const {
-  return CountIf(test_cases_, TestCaseFailed);
-}
-
-// Gets the number of all test cases.
-int UnitTestImpl::total_test_case_count() const {
-  return static_cast<int>(test_cases_.size());
-}
-
-// Gets the number of all test cases that contain at least one test
-// that should run.
-int UnitTestImpl::test_case_to_run_count() const {
-  return CountIf(test_cases_, ShouldRunTestCase);
-}
-
-// Gets the number of successful tests.
-int UnitTestImpl::successful_test_count() const {
-  return SumOverTestCaseList(test_cases_, &TestCase::successful_test_count);
-}
-
-// Gets the number of failed tests.
-int UnitTestImpl::failed_test_count() const {
-  return SumOverTestCaseList(test_cases_, &TestCase::failed_test_count);
-}
-
-// Gets the number of disabled tests that will be reported in the XML report.
-int UnitTestImpl::reportable_disabled_test_count() const {
-  return SumOverTestCaseList(test_cases_,
-                             &TestCase::reportable_disabled_test_count);
-}
-
-// Gets the number of disabled tests.
-int UnitTestImpl::disabled_test_count() const {
-  return SumOverTestCaseList(test_cases_, &TestCase::disabled_test_count);
-}
-
-// Gets the number of tests to be printed in the XML report.
-int UnitTestImpl::reportable_test_count() const {
-  return SumOverTestCaseList(test_cases_, &TestCase::reportable_test_count);
-}
-
-// Gets the number of all tests.
-int UnitTestImpl::total_test_count() const {
-  return SumOverTestCaseList(test_cases_, &TestCase::total_test_count);
-}
-
-// Gets the number of tests that should run.
-int UnitTestImpl::test_to_run_count() const {
-  return SumOverTestCaseList(test_cases_, &TestCase::test_to_run_count);
-}
-
-// Returns the current OS stack trace as an std::string.
-//
-// The maximum number of stack frames to be included is specified by
-// the gtest_stack_trace_depth flag.  The skip_count parameter
-// specifies the number of top frames to be skipped, which doesn't
-// count against the number of frames to be included.
-//
-// For example, if Foo() calls Bar(), which in turn calls
-// CurrentOsStackTraceExceptTop(1), Foo() will be included in the
-// trace but Bar() and CurrentOsStackTraceExceptTop() won't.
-std::string UnitTestImpl::CurrentOsStackTraceExceptTop(int skip_count) {
-  return os_stack_trace_getter()->CurrentStackTrace(
-      static_cast<int>(GTEST_FLAG(stack_trace_depth)),
-      skip_count + 1
-      // Skips the user-specified number of frames plus this function
-      // itself.
-      );  // NOLINT
-}
-
-// Returns the current time in milliseconds.
-TimeInMillis GetTimeInMillis() {
-#if GTEST_OS_WINDOWS_MOBILE || defined(__BORLANDC__)
-  // Difference between 1970-01-01 and 1601-01-01 in milliseconds.
-  // http://analogous.blogspot.com/2005/04/epoch.html
-  const TimeInMillis kJavaEpochToWinFileTimeDelta =
-    static_cast<TimeInMillis>(116444736UL) * 100000UL;
-  const DWORD kTenthMicrosInMilliSecond = 10000;
-
-  SYSTEMTIME now_systime;
-  FILETIME now_filetime;
-  ULARGE_INTEGER now_int64;
-  // TODO(kenton@google.com): Shouldn't this just use
-  //   GetSystemTimeAsFileTime()?
-  GetSystemTime(&now_systime);
-  if (SystemTimeToFileTime(&now_systime, &now_filetime)) {
-    now_int64.LowPart = now_filetime.dwLowDateTime;
-    now_int64.HighPart = now_filetime.dwHighDateTime;
-    now_int64.QuadPart = (now_int64.QuadPart / kTenthMicrosInMilliSecond) -
-      kJavaEpochToWinFileTimeDelta;
-    return now_int64.QuadPart;
-  }
-  return 0;
-#elif GTEST_OS_WINDOWS && !GTEST_HAS_GETTIMEOFDAY_
-  __timeb64 now;
-
-  // MSVC 8 deprecates _ftime64(), so we want to suppress warning 4996
-  // (deprecated function) there.
-  // TODO(kenton@google.com): Use GetTickCount()?  Or use
-  //   SystemTimeToFileTime()
-  GTEST_DISABLE_MSC_WARNINGS_PUSH_(4996)
-  _ftime64(&now);
-  GTEST_DISABLE_MSC_WARNINGS_POP_()
-
-  return static_cast<TimeInMillis>(now.time) * 1000 + now.millitm;
-#elif GTEST_HAS_GETTIMEOFDAY_
-  struct timeval now;
-  gettimeofday(&now, NULL);
-  return static_cast<TimeInMillis>(now.tv_sec) * 1000 + now.tv_usec / 1000;
-#else
-# error "Don't know how to get the current time on your system."
-#endif
-}
-
-// Utilities
-
-// class String.
-
-#if GTEST_OS_WINDOWS_MOBILE
-// Creates a UTF-16 wide string from the given ANSI string, allocating
-// memory using new. The caller is responsible for deleting the return
-// value using delete[]. Returns the wide string, or NULL if the
-// input is NULL.
-LPCWSTR String::AnsiToUtf16(const char* ansi) {
-  if (!ansi) return NULL;
-  const int length = strlen(ansi);
-  const int unicode_length =
-      MultiByteToWideChar(CP_ACP, 0, ansi, length,
-                          NULL, 0);
-  WCHAR* unicode = new WCHAR[unicode_length + 1];
-  MultiByteToWideChar(CP_ACP, 0, ansi, length,
-                      unicode, unicode_length);
-  unicode[unicode_length] = 0;
-  return unicode;
-}
-
-// Creates an ANSI string from the given wide string, allocating
-// memory using new. The caller is responsible for deleting the return
-// value using delete[]. Returns the ANSI string, or NULL if the
-// input is NULL.
-const char* String::Utf16ToAnsi(LPCWSTR utf16_str)  {
-  if (!utf16_str) return NULL;
-  const int ansi_length =
-      WideCharToMultiByte(CP_ACP, 0, utf16_str, -1,
-                          NULL, 0, NULL, NULL);
-  char* ansi = new char[ansi_length + 1];
-  WideCharToMultiByte(CP_ACP, 0, utf16_str, -1,
-                      ansi, ansi_length, NULL, NULL);
-  ansi[ansi_length] = 0;
-  return ansi;
-}
-
-#endif  // GTEST_OS_WINDOWS_MOBILE
-
-// Compares two C strings.  Returns true iff they have the same content.
-//
-// Unlike strcmp(), this function can handle NULL argument(s).  A NULL
-// C string is considered different to any non-NULL C string,
-// including the empty string.
-bool String::CStringEquals(const char * lhs, const char * rhs) {
-  if ( lhs == NULL ) return rhs == NULL;
-
-  if ( rhs == NULL ) return false;
-
-  return strcmp(lhs, rhs) == 0;
-}
-
-#if GTEST_HAS_STD_WSTRING || GTEST_HAS_GLOBAL_WSTRING
-
-// Converts an array of wide chars to a narrow string using the UTF-8
-// encoding, and streams the result to the given Message object.
-static void StreamWideCharsToMessage(const wchar_t* wstr, size_t length,
-                                     Message* msg) {
-  for (size_t i = 0; i != length; ) {  // NOLINT
-    if (wstr[i] != L'\0') {
-      *msg << WideStringToUtf8(wstr + i, static_cast<int>(length - i));
-      while (i != length && wstr[i] != L'\0')
-        i++;
-    } else {
-      *msg << '\0';
-      i++;
-    }
-  }
-}
-
-#endif  // GTEST_HAS_STD_WSTRING || GTEST_HAS_GLOBAL_WSTRING
-
-void SplitString(const ::std::string& str, char delimiter,
-                 ::std::vector< ::std::string>* dest) {
-  ::std::vector< ::std::string> parsed;
-  ::std::string::size_type pos = 0;
-  while (::testing::internal::AlwaysTrue()) {
-    const ::std::string::size_type colon = str.find(delimiter, pos);
-    if (colon == ::std::string::npos) {
-      parsed.push_back(str.substr(pos));
-      break;
-    } else {
-      parsed.push_back(str.substr(pos, colon - pos));
-      pos = colon + 1;
-    }
-  }
-  dest->swap(parsed);
-}
-
-}  // namespace internal
-
-// Constructs an empty Message.
-// We allocate the stringstream separately because otherwise each use of
-// ASSERT/EXPECT in a procedure adds over 200 bytes to the procedure's
-// stack frame leading to huge stack frames in some cases; gcc does not reuse
-// the stack space.
-Message::Message() : ss_(new ::std::stringstream) {
-  // By default, we want there to be enough precision when printing
-  // a double to a Message.
-  *ss_ << std::setprecision(std::numeric_limits<double>::digits10 + 2);
-}
-
-// These two overloads allow streaming a wide C string to a Message
-// using the UTF-8 encoding.
-Message& Message::operator <<(const wchar_t* wide_c_str) {
-  return *this << internal::String::ShowWideCString(wide_c_str);
-}
-Message& Message::operator <<(wchar_t* wide_c_str) {
-  return *this << internal::String::ShowWideCString(wide_c_str);
-}
-
-#if GTEST_HAS_STD_WSTRING
-// Converts the given wide string to a narrow string using the UTF-8
-// encoding, and streams the result to this Message object.
-Message& Message::operator <<(const ::std::wstring& wstr) {
-  internal::StreamWideCharsToMessage(wstr.c_str(), wstr.length(), this);
-  return *this;
-}
-#endif  // GTEST_HAS_STD_WSTRING
-
-#if GTEST_HAS_GLOBAL_WSTRING
-// Converts the given wide string to a narrow string using the UTF-8
-// encoding, and streams the result to this Message object.
-Message& Message::operator <<(const ::wstring& wstr) {
-  internal::StreamWideCharsToMessage(wstr.c_str(), wstr.length(), this);
-  return *this;
-}
-#endif  // GTEST_HAS_GLOBAL_WSTRING
-
-// Gets the text streamed to this object so far as an std::string.
-// Each '\0' character in the buffer is replaced with "\\0".
-std::string Message::GetString() const {
-  return internal::StringStreamToString(ss_.get());
-}
-
-// AssertionResult constructors.
-// Used in EXPECT_TRUE/FALSE(assertion_result).
-AssertionResult::AssertionResult(const AssertionResult& other)
-    : success_(other.success_),
-      message_(other.message_.get() != NULL ?
-               new ::std::string(*other.message_) :
-               static_cast< ::std::string*>(NULL)) {
-}
-
-// Swaps two AssertionResults.
-void AssertionResult::swap(AssertionResult& other) {
-  using std::swap;
-  swap(success_, other.success_);
-  swap(message_, other.message_);
-}
-
-// Returns the assertion's negation. Used with EXPECT/ASSERT_FALSE.
-AssertionResult AssertionResult::operator!() const {
-  AssertionResult negation(!success_);
-  if (message_.get() != NULL)
-    negation << *message_;
-  return negation;
-}
-
-// Makes a successful assertion result.
-AssertionResult AssertionSuccess() {
-  return AssertionResult(true);
-}
-
-// Makes a failed assertion result.
-AssertionResult AssertionFailure() {
-  return AssertionResult(false);
-}
-
-// Makes a failed assertion result with the given failure message.
-// Deprecated; use AssertionFailure() << message.
-AssertionResult AssertionFailure(const Message& message) {
-  return AssertionFailure() << message;
-}
-
-namespace internal {
-
-namespace edit_distance {
-std::vector<EditType> CalculateOptimalEdits(const std::vector<size_t>& left,
-                                            const std::vector<size_t>& right) {
-  std::vector<std::vector<double> > costs(
-      left.size() + 1, std::vector<double>(right.size() + 1));
-  std::vector<std::vector<EditType> > best_move(
-      left.size() + 1, std::vector<EditType>(right.size() + 1));
-
-  // Populate for empty right.
-  for (size_t l_i = 0; l_i < costs.size(); ++l_i) {
-    costs[l_i][0] = static_cast<double>(l_i);
-    best_move[l_i][0] = kRemove;
-  }
-  // Populate for empty left.
-  for (size_t r_i = 1; r_i < costs[0].size(); ++r_i) {
-    costs[0][r_i] = static_cast<double>(r_i);
-    best_move[0][r_i] = kAdd;
-  }
-
-  for (size_t l_i = 0; l_i < left.size(); ++l_i) {
-    for (size_t r_i = 0; r_i < right.size(); ++r_i) {
-      if (left[l_i] == right[r_i]) {
-        // Found a match. Consume it.
-        costs[l_i + 1][r_i + 1] = costs[l_i][r_i];
-        best_move[l_i + 1][r_i + 1] = kMatch;
-        continue;
-      }
-
-      const double add = costs[l_i + 1][r_i];
-      const double remove = costs[l_i][r_i + 1];
-      const double replace = costs[l_i][r_i];
-      if (add < remove && add < replace) {
-        costs[l_i + 1][r_i + 1] = add + 1;
-        best_move[l_i + 1][r_i + 1] = kAdd;
-      } else if (remove < add && remove < replace) {
-        costs[l_i + 1][r_i + 1] = remove + 1;
-        best_move[l_i + 1][r_i + 1] = kRemove;
-      } else {
-        // We make replace a little more expensive than add/remove to lower
-        // their priority.
-        costs[l_i + 1][r_i + 1] = replace + 1.00001;
-        best_move[l_i + 1][r_i + 1] = kReplace;
-      }
-    }
-  }
-
-  // Reconstruct the best path. We do it in reverse order.
-  std::vector<EditType> best_path;
-  for (size_t l_i = left.size(), r_i = right.size(); l_i > 0 || r_i > 0;) {
-    EditType move = best_move[l_i][r_i];
-    best_path.push_back(move);
-    l_i -= move != kAdd;
-    r_i -= move != kRemove;
-  }
-  std::reverse(best_path.begin(), best_path.end());
-  return best_path;
-}
-
-namespace {
-
-// Helper class to convert string into ids with deduplication.
-class InternalStrings {
- public:
-  size_t GetId(const std::string& str) {
-    IdMap::iterator it = ids_.find(str);
-    if (it != ids_.end()) return it->second;
-    size_t id = ids_.size();
-    return ids_[str] = id;
-  }
-
- private:
-  typedef std::map<std::string, size_t> IdMap;
-  IdMap ids_;
-};
-
-}  // namespace
-
-std::vector<EditType> CalculateOptimalEdits(
-    const std::vector<std::string>& left,
-    const std::vector<std::string>& right) {
-  std::vector<size_t> left_ids, right_ids;
-  {
-    InternalStrings intern_table;
-    for (size_t i = 0; i < left.size(); ++i) {
-      left_ids.push_back(intern_table.GetId(left[i]));
-    }
-    for (size_t i = 0; i < right.size(); ++i) {
-      right_ids.push_back(intern_table.GetId(right[i]));
-    }
-  }
-  return CalculateOptimalEdits(left_ids, right_ids);
-}
-
-namespace {
-
-// Helper class that holds the state for one hunk and prints it out to the
-// stream.
-// It reorders adds/removes when possible to group all removes before all
-// adds. It also adds the hunk header before printint into the stream.
-class Hunk {
- public:
-  Hunk(size_t left_start, size_t right_start)
-      : left_start_(left_start),
-        right_start_(right_start),
-        adds_(),
-        removes_(),
-        common_() {}
-
-  void PushLine(char edit, const char* line) {
-    switch (edit) {
-      case ' ':
-        ++common_;
-        FlushEdits();
-        hunk_.push_back(std::make_pair(' ', line));
-        break;
-      case '-':
-        ++removes_;
-        hunk_removes_.push_back(std::make_pair('-', line));
-        break;
-      case '+':
-        ++adds_;
-        hunk_adds_.push_back(std::make_pair('+', line));
-        break;
-    }
-  }
-
-  void PrintTo(std::ostream* os) {
-    PrintHeader(os);
-    FlushEdits();
-    for (std::list<std::pair<char, const char*> >::const_iterator it =
-             hunk_.begin();
-         it != hunk_.end(); ++it) {
-      *os << it->first << it->second << "\n";
-    }
-  }
-
-  bool has_edits() const { return adds_ || removes_; }
-
- private:
-  void FlushEdits() {
-    hunk_.splice(hunk_.end(), hunk_removes_);
-    hunk_.splice(hunk_.end(), hunk_adds_);
-  }
-
-  // Print a unified diff header for one hunk.
-  // The format is
-  //   "@@ -<left_start>,<left_length> +<right_start>,<right_length> @@"
-  // where the left/right parts are omitted if unnecessary.
-  void PrintHeader(std::ostream* ss) const {
-    *ss << "@@ ";
-    if (removes_) {
-      *ss << "-" << left_start_ << "," << (removes_ + common_);
-    }
-    if (removes_ && adds_) {
-      *ss << " ";
-    }
-    if (adds_) {
-      *ss << "+" << right_start_ << "," << (adds_ + common_);
-    }
-    *ss << " @@\n";
-  }
-
-  size_t left_start_, right_start_;
-  size_t adds_, removes_, common_;
-  std::list<std::pair<char, const char*> > hunk_, hunk_adds_, hunk_removes_;
-};
-
-}  // namespace
-
-// Create a list of diff hunks in Unified diff format.
-// Each hunk has a header generated by PrintHeader above plus a body with
-// lines prefixed with ' ' for no change, '-' for deletion and '+' for
-// addition.
-// 'context' represents the desired unchanged prefix/suffix around the diff.
-// If two hunks are close enough that their contexts overlap, then they are
-// joined into one hunk.
-std::string CreateUnifiedDiff(const std::vector<std::string>& left,
-                              const std::vector<std::string>& right,
-                              size_t context) {
-  const std::vector<EditType> edits = CalculateOptimalEdits(left, right);
-
-  size_t l_i = 0, r_i = 0, edit_i = 0;
-  std::stringstream ss;
-  while (edit_i < edits.size()) {
-    // Find first edit.
-    while (edit_i < edits.size() && edits[edit_i] == kMatch) {
-      ++l_i;
-      ++r_i;
-      ++edit_i;
-    }
-
-    // Find the first line to include in the hunk.
-    const size_t prefix_context = std::min(l_i, context);
-    Hunk hunk(l_i - prefix_context + 1, r_i - prefix_context + 1);
-    for (size_t i = prefix_context; i > 0; --i) {
-      hunk.PushLine(' ', left[l_i - i].c_str());
-    }
-
-    // Iterate the edits until we found enough suffix for the hunk or the input
-    // is over.
-    size_t n_suffix = 0;
-    for (; edit_i < edits.size(); ++edit_i) {
-      if (n_suffix >= context) {
-        // Continue only if the next hunk is very close.
-        std::vector<EditType>::const_iterator it = edits.begin() + edit_i;
-        while (it != edits.end() && *it == kMatch) ++it;
-        if (it == edits.end() || (it - edits.begin()) - edit_i >= context) {
-          // There is no next edit or it is too far away.
-          break;
-        }
-      }
-
-      EditType edit = edits[edit_i];
-      // Reset count when a non match is found.
-      n_suffix = edit == kMatch ? n_suffix + 1 : 0;
-
-      if (edit == kMatch || edit == kRemove || edit == kReplace) {
-        hunk.PushLine(edit == kMatch ? ' ' : '-', left[l_i].c_str());
-      }
-      if (edit == kAdd || edit == kReplace) {
-        hunk.PushLine('+', right[r_i].c_str());
-      }
-
-      // Advance indices, depending on edit type.
-      l_i += edit != kAdd;
-      r_i += edit != kRemove;
-    }
-
-    if (!hunk.has_edits()) {
-      // We are done. We don't want this hunk.
-      break;
-    }
-
-    hunk.PrintTo(&ss);
-  }
-  return ss.str();
-}
-
-}  // namespace edit_distance
-
-namespace {
-
-// The string representation of the values received in EqFailure() are already
-// escaped. Split them on escaped '\n' boundaries. Leave all other escaped
-// characters the same.
-std::vector<std::string> SplitEscapedString(const std::string& str) {
-  std::vector<std::string> lines;
-  size_t start = 0, end = str.size();
-  if (end > 2 && str[0] == '"' && str[end - 1] == '"') {
-    ++start;
-    --end;
-  }
-  bool escaped = false;
-  for (size_t i = start; i + 1 < end; ++i) {
-    if (escaped) {
-      escaped = false;
-      if (str[i] == 'n') {
-        lines.push_back(str.substr(start, i - start - 1));
-        start = i + 1;
-      }
-    } else {
-      escaped = str[i] == '\\';
-    }
-  }
-  lines.push_back(str.substr(start, end - start));
-  return lines;
-}
-
-}  // namespace
-
-// Constructs and returns the message for an equality assertion
-// (e.g. ASSERT_EQ, EXPECT_STREQ, etc) failure.
-//
-// The first four parameters are the expressions used in the assertion
-// and their values, as strings.  For example, for ASSERT_EQ(foo, bar)
-// where foo is 5 and bar is 6, we have:
-//
-//   lhs_expression: "foo"
-//   rhs_expression: "bar"
-//   lhs_value:      "5"
-//   rhs_value:      "6"
-//
-// The ignoring_case parameter is true iff the assertion is a
-// *_STRCASEEQ*.  When it's true, the string "Ignoring case" will
-// be inserted into the message.
-AssertionResult EqFailure(const char* lhs_expression,
-                          const char* rhs_expression,
-                          const std::string& lhs_value,
-                          const std::string& rhs_value,
-                          bool ignoring_case) {
-  Message msg;
-  msg << "Expected equality of these values:";
-  msg << "\n  " << lhs_expression;
-  if (lhs_value != lhs_expression) {
-    msg << "\n    Which is: " << lhs_value;
-  }
-  msg << "\n  " << rhs_expression;
-  if (rhs_value != rhs_expression) {
-    msg << "\n    Which is: " << rhs_value;
-  }
-
-  if (ignoring_case) {
-    msg << "\nIgnoring case";
-  }
-
-  if (!lhs_value.empty() && !rhs_value.empty()) {
-    const std::vector<std::string> lhs_lines =
-        SplitEscapedString(lhs_value);
-    const std::vector<std::string> rhs_lines =
-        SplitEscapedString(rhs_value);
-    if (lhs_lines.size() > 1 || rhs_lines.size() > 1) {
-      msg << "\nWith diff:\n"
-          << edit_distance::CreateUnifiedDiff(lhs_lines, rhs_lines);
-    }
-  }
-
-  return AssertionFailure() << msg;
-}
-
-// Constructs a failure message for Boolean assertions such as EXPECT_TRUE.
-std::string GetBoolAssertionFailureMessage(
-    const AssertionResult& assertion_result,
-    const char* expression_text,
-    const char* actual_predicate_value,
-    const char* expected_predicate_value) {
-  const char* actual_message = assertion_result.message();
-  Message msg;
-  msg << "Value of: " << expression_text
-      << "\n  Actual: " << actual_predicate_value;
-  if (actual_message[0] != '\0')
-    msg << " (" << actual_message << ")";
-  msg << "\nExpected: " << expected_predicate_value;
-  return msg.GetString();
-}
-
-// Helper function for implementing ASSERT_NEAR.
-AssertionResult DoubleNearPredFormat(const char* expr1,
-                                     const char* expr2,
-                                     const char* abs_error_expr,
-                                     double val1,
-                                     double val2,
-                                     double abs_error) {
-  const double diff = fabs(val1 - val2);
-  if (diff <= abs_error) return AssertionSuccess();
-
-  // TODO(wan): do not print the value of an expression if it's
-  // already a literal.
-  return AssertionFailure()
-      << "The difference between " << expr1 << " and " << expr2
-      << " is " << diff << ", which exceeds " << abs_error_expr << ", where\n"
-      << expr1 << " evaluates to " << val1 << ",\n"
-      << expr2 << " evaluates to " << val2 << ", and\n"
-      << abs_error_expr << " evaluates to " << abs_error << ".";
-}
-
-
-// Helper template for implementing FloatLE() and DoubleLE().
-template <typename RawType>
-AssertionResult FloatingPointLE(const char* expr1,
-                                const char* expr2,
-                                RawType val1,
-                                RawType val2) {
-  // Returns success if val1 is less than val2,
-  if (val1 < val2) {
-    return AssertionSuccess();
-  }
-
-  // or if val1 is almost equal to val2.
-  const FloatingPoint<RawType> lhs(val1), rhs(val2);
-  if (lhs.AlmostEquals(rhs)) {
-    return AssertionSuccess();
-  }
-
-  // Note that the above two checks will both fail if either val1 or
-  // val2 is NaN, as the IEEE floating-point standard requires that
-  // any predicate involving a NaN must return false.
-
-  ::std::stringstream val1_ss;
-  val1_ss << std::setprecision(std::numeric_limits<RawType>::digits10 + 2)
-          << val1;
-
-  ::std::stringstream val2_ss;
-  val2_ss << std::setprecision(std::numeric_limits<RawType>::digits10 + 2)
-          << val2;
-
-  return AssertionFailure()
-      << "Expected: (" << expr1 << ") <= (" << expr2 << ")\n"
-      << "  Actual: " << StringStreamToString(&val1_ss) << " vs "
-      << StringStreamToString(&val2_ss);
-}
-
-}  // namespace internal
-
-// Asserts that val1 is less than, or almost equal to, val2.  Fails
-// otherwise.  In particular, it fails if either val1 or val2 is NaN.
-AssertionResult FloatLE(const char* expr1, const char* expr2,
-                        float val1, float val2) {
-  return internal::FloatingPointLE<float>(expr1, expr2, val1, val2);
-}
-
-// Asserts that val1 is less than, or almost equal to, val2.  Fails
-// otherwise.  In particular, it fails if either val1 or val2 is NaN.
-AssertionResult DoubleLE(const char* expr1, const char* expr2,
-                         double val1, double val2) {
-  return internal::FloatingPointLE<double>(expr1, expr2, val1, val2);
-}
-
-namespace internal {
-
-// The helper function for {ASSERT|EXPECT}_EQ with int or enum
-// arguments.
-AssertionResult CmpHelperEQ(const char* lhs_expression,
-                            const char* rhs_expression,
-                            BiggestInt lhs,
-                            BiggestInt rhs) {
-  if (lhs == rhs) {
-    return AssertionSuccess();
-  }
-
-  return EqFailure(lhs_expression,
-                   rhs_expression,
-                   FormatForComparisonFailureMessage(lhs, rhs),
-                   FormatForComparisonFailureMessage(rhs, lhs),
-                   false);
-}
-
-// A macro for implementing the helper functions needed to implement
-// ASSERT_?? and EXPECT_?? with integer or enum arguments.  It is here
-// just to avoid copy-and-paste of similar code.
-#define GTEST_IMPL_CMP_HELPER_(op_name, op)\
-AssertionResult CmpHelper##op_name(const char* expr1, const char* expr2, \
-                                   BiggestInt val1, BiggestInt val2) {\
-  if (val1 op val2) {\
-    return AssertionSuccess();\
-  } else {\
-    return AssertionFailure() \
-        << "Expected: (" << expr1 << ") " #op " (" << expr2\
-        << "), actual: " << FormatForComparisonFailureMessage(val1, val2)\
-        << " vs " << FormatForComparisonFailureMessage(val2, val1);\
-  }\
-}
-
-// Implements the helper function for {ASSERT|EXPECT}_NE with int or
-// enum arguments.
-GTEST_IMPL_CMP_HELPER_(NE, !=)
-// Implements the helper function for {ASSERT|EXPECT}_LE with int or
-// enum arguments.
-GTEST_IMPL_CMP_HELPER_(LE, <=)
-// Implements the helper function for {ASSERT|EXPECT}_LT with int or
-// enum arguments.
-GTEST_IMPL_CMP_HELPER_(LT, < )
-// Implements the helper function for {ASSERT|EXPECT}_GE with int or
-// enum arguments.
-GTEST_IMPL_CMP_HELPER_(GE, >=)
-// Implements the helper function for {ASSERT|EXPECT}_GT with int or
-// enum arguments.
-GTEST_IMPL_CMP_HELPER_(GT, > )
-
-#undef GTEST_IMPL_CMP_HELPER_
-
-// The helper function for {ASSERT|EXPECT}_STREQ.
-AssertionResult CmpHelperSTREQ(const char* lhs_expression,
-                               const char* rhs_expression,
-                               const char* lhs,
-                               const char* rhs) {
-  if (String::CStringEquals(lhs, rhs)) {
-    return AssertionSuccess();
-  }
-
-  return EqFailure(lhs_expression,
-                   rhs_expression,
-                   PrintToString(lhs),
-                   PrintToString(rhs),
-                   false);
-}
-
-// The helper function for {ASSERT|EXPECT}_STRCASEEQ.
-AssertionResult CmpHelperSTRCASEEQ(const char* lhs_expression,
-                                   const char* rhs_expression,
-                                   const char* lhs,
-                                   const char* rhs) {
-  if (String::CaseInsensitiveCStringEquals(lhs, rhs)) {
-    return AssertionSuccess();
-  }
-
-  return EqFailure(lhs_expression,
-                   rhs_expression,
-                   PrintToString(lhs),
-                   PrintToString(rhs),
-                   true);
-}
-
-// The helper function for {ASSERT|EXPECT}_STRNE.
-AssertionResult CmpHelperSTRNE(const char* s1_expression,
-                               const char* s2_expression,
-                               const char* s1,
-                               const char* s2) {
-  if (!String::CStringEquals(s1, s2)) {
-    return AssertionSuccess();
-  } else {
-    return AssertionFailure() << "Expected: (" << s1_expression << ") != ("
-                              << s2_expression << "), actual: \""
-                              << s1 << "\" vs \"" << s2 << "\"";
-  }
-}
-
-// The helper function for {ASSERT|EXPECT}_STRCASENE.
-AssertionResult CmpHelperSTRCASENE(const char* s1_expression,
-                                   const char* s2_expression,
-                                   const char* s1,
-                                   const char* s2) {
-  if (!String::CaseInsensitiveCStringEquals(s1, s2)) {
-    return AssertionSuccess();
-  } else {
-    return AssertionFailure()
-        << "Expected: (" << s1_expression << ") != ("
-        << s2_expression << ") (ignoring case), actual: \""
-        << s1 << "\" vs \"" << s2 << "\"";
-  }
-}
-
-}  // namespace internal
-
-namespace {
-
-// Helper functions for implementing IsSubString() and IsNotSubstring().
-
-// This group of overloaded functions return true iff needle is a
-// substring of haystack.  NULL is considered a substring of itself
-// only.
-
-bool IsSubstringPred(const char* needle, const char* haystack) {
-  if (needle == NULL || haystack == NULL)
-    return needle == haystack;
-
-  return strstr(haystack, needle) != NULL;
-}
-
-bool IsSubstringPred(const wchar_t* needle, const wchar_t* haystack) {
-  if (needle == NULL || haystack == NULL)
-    return needle == haystack;
-
-  return wcsstr(haystack, needle) != NULL;
-}
-
-// StringType here can be either ::std::string or ::std::wstring.
-template <typename StringType>
-bool IsSubstringPred(const StringType& needle,
-                     const StringType& haystack) {
-  return haystack.find(needle) != StringType::npos;
-}
-
-// This function implements either IsSubstring() or IsNotSubstring(),
-// depending on the value of the expected_to_be_substring parameter.
-// StringType here can be const char*, const wchar_t*, ::std::string,
-// or ::std::wstring.
-template <typename StringType>
-AssertionResult IsSubstringImpl(
-    bool expected_to_be_substring,
-    const char* needle_expr, const char* haystack_expr,
-    const StringType& needle, const StringType& haystack) {
-  if (IsSubstringPred(needle, haystack) == expected_to_be_substring)
-    return AssertionSuccess();
-
-  const bool is_wide_string = sizeof(needle[0]) > 1;
-  const char* const begin_string_quote = is_wide_string ? "L\"" : "\"";
-  return AssertionFailure()
-      << "Value of: " << needle_expr << "\n"
-      << "  Actual: " << begin_string_quote << needle << "\"\n"
-      << "Expected: " << (expected_to_be_substring ? "" : "not ")
-      << "a substring of " << haystack_expr << "\n"
-      << "Which is: " << begin_string_quote << haystack << "\"";
-}
-
-}  // namespace
-
-// IsSubstring() and IsNotSubstring() check whether needle is a
-// substring of haystack (NULL is considered a substring of itself
-// only), and return an appropriate error message when they fail.
-
-AssertionResult IsSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const char* needle, const char* haystack) {
-  return IsSubstringImpl(true, needle_expr, haystack_expr, needle, haystack);
-}
-
-AssertionResult IsSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const wchar_t* needle, const wchar_t* haystack) {
-  return IsSubstringImpl(true, needle_expr, haystack_expr, needle, haystack);
-}
-
-AssertionResult IsNotSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const char* needle, const char* haystack) {
-  return IsSubstringImpl(false, needle_expr, haystack_expr, needle, haystack);
-}
-
-AssertionResult IsNotSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const wchar_t* needle, const wchar_t* haystack) {
-  return IsSubstringImpl(false, needle_expr, haystack_expr, needle, haystack);
-}
-
-AssertionResult IsSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const ::std::string& needle, const ::std::string& haystack) {
-  return IsSubstringImpl(true, needle_expr, haystack_expr, needle, haystack);
-}
-
-AssertionResult IsNotSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const ::std::string& needle, const ::std::string& haystack) {
-  return IsSubstringImpl(false, needle_expr, haystack_expr, needle, haystack);
-}
-
-#if GTEST_HAS_STD_WSTRING
-AssertionResult IsSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const ::std::wstring& needle, const ::std::wstring& haystack) {
-  return IsSubstringImpl(true, needle_expr, haystack_expr, needle, haystack);
-}
-
-AssertionResult IsNotSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const ::std::wstring& needle, const ::std::wstring& haystack) {
-  return IsSubstringImpl(false, needle_expr, haystack_expr, needle, haystack);
-}
-#endif  // GTEST_HAS_STD_WSTRING
-
-namespace internal {
-
-#if GTEST_OS_WINDOWS
-
-namespace {
-
-// Helper function for IsHRESULT{SuccessFailure} predicates
-AssertionResult HRESULTFailureHelper(const char* expr,
-                                     const char* expected,
-                                     long hr) {  // NOLINT
-# if GTEST_OS_WINDOWS_MOBILE || GTEST_OS_WINDOWS_TV_TITLE
-
-  // Windows CE doesn't support FormatMessage.
-  const char error_text[] = "";
-
-# else
-
-  // Looks up the human-readable system message for the HRESULT code
-  // and since we're not passing any params to FormatMessage, we don't
-  // want inserts expanded.
-  const DWORD kFlags = FORMAT_MESSAGE_FROM_SYSTEM |
-                       FORMAT_MESSAGE_IGNORE_INSERTS;
-  const DWORD kBufSize = 4096;
-  // Gets the system's human readable message string for this HRESULT.
-  char error_text[kBufSize] = { '\0' };
-  DWORD message_length = ::FormatMessageA(kFlags,
-                                          0,  // no source, we're asking system
-                                          hr,  // the error
-                                          0,  // no line width restrictions
-                                          error_text,  // output buffer
-                                          kBufSize,  // buf size
-                                          NULL);  // no arguments for inserts
-  // Trims tailing white space (FormatMessage leaves a trailing CR-LF)
-  for (; message_length && IsSpace(error_text[message_length - 1]);
-          --message_length) {
-    error_text[message_length - 1] = '\0';
-  }
-
-# endif  // GTEST_OS_WINDOWS_MOBILE
-
-  const std::string error_hex("0x" + String::FormatHexInt(hr));
-  return ::testing::AssertionFailure()
-      << "Expected: " << expr << " " << expected << ".\n"
-      << "  Actual: " << error_hex << " " << error_text << "\n";
-}
-
-}  // namespace
-
-AssertionResult IsHRESULTSuccess(const char* expr, long hr) {  // NOLINT
-  if (SUCCEEDED(hr)) {
-    return AssertionSuccess();
-  }
-  return HRESULTFailureHelper(expr, "succeeds", hr);
-}
-
-AssertionResult IsHRESULTFailure(const char* expr, long hr) {  // NOLINT
-  if (FAILED(hr)) {
-    return AssertionSuccess();
-  }
-  return HRESULTFailureHelper(expr, "fails", hr);
-}
-
-#endif  // GTEST_OS_WINDOWS
-
-// Utility functions for encoding Unicode text (wide strings) in
-// UTF-8.
-
-// A Unicode code-point can have up to 21 bits, and is encoded in UTF-8
-// like this:
-//
-// Code-point length   Encoding
-//   0 -  7 bits       0xxxxxxx
-//   8 - 11 bits       110xxxxx 10xxxxxx
-//  12 - 16 bits       1110xxxx 10xxxxxx 10xxxxxx
-//  17 - 21 bits       11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
-
-// The maximum code-point a one-byte UTF-8 sequence can represent.
-const UInt32 kMaxCodePoint1 = (static_cast<UInt32>(1) <<  7) - 1;
-
-// The maximum code-point a two-byte UTF-8 sequence can represent.
-const UInt32 kMaxCodePoint2 = (static_cast<UInt32>(1) << (5 + 6)) - 1;
-
-// The maximum code-point a three-byte UTF-8 sequence can represent.
-const UInt32 kMaxCodePoint3 = (static_cast<UInt32>(1) << (4 + 2*6)) - 1;
-
-// The maximum code-point a four-byte UTF-8 sequence can represent.
-const UInt32 kMaxCodePoint4 = (static_cast<UInt32>(1) << (3 + 3*6)) - 1;
-
-// Chops off the n lowest bits from a bit pattern.  Returns the n
-// lowest bits.  As a side effect, the original bit pattern will be
-// shifted to the right by n bits.
-inline UInt32 ChopLowBits(UInt32* bits, int n) {
-  const UInt32 low_bits = *bits & ((static_cast<UInt32>(1) << n) - 1);
-  *bits >>= n;
-  return low_bits;
-}
-
-// Converts a Unicode code point to a narrow string in UTF-8 encoding.
-// code_point parameter is of type UInt32 because wchar_t may not be
-// wide enough to contain a code point.
-// If the code_point is not a valid Unicode code point
-// (i.e. outside of Unicode range U+0 to U+10FFFF) it will be converted
-// to "(Invalid Unicode 0xXXXXXXXX)".
-std::string CodePointToUtf8(UInt32 code_point) {
-  if (code_point > kMaxCodePoint4) {
-    return "(Invalid Unicode 0x" + String::FormatHexInt(code_point) + ")";
-  }
-
-  char str[5];  // Big enough for the largest valid code point.
-  if (code_point <= kMaxCodePoint1) {
-    str[1] = '\0';
-    str[0] = static_cast<char>(code_point);                          // 0xxxxxxx
-  } else if (code_point <= kMaxCodePoint2) {
-    str[2] = '\0';
-    str[1] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
-    str[0] = static_cast<char>(0xC0 | code_point);                   // 110xxxxx
-  } else if (code_point <= kMaxCodePoint3) {
-    str[3] = '\0';
-    str[2] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
-    str[1] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
-    str[0] = static_cast<char>(0xE0 | code_point);                   // 1110xxxx
-  } else {  // code_point <= kMaxCodePoint4
-    str[4] = '\0';
-    str[3] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
-    str[2] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
-    str[1] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
-    str[0] = static_cast<char>(0xF0 | code_point);                   // 11110xxx
-  }
-  return str;
-}
-
-// The following two functions only make sense if the system
-// uses UTF-16 for wide string encoding. All supported systems
-// with 16 bit wchar_t (Windows, Cygwin, Symbian OS) do use UTF-16.
-
-// Determines if the arguments constitute UTF-16 surrogate pair
-// and thus should be combined into a single Unicode code point
-// using CreateCodePointFromUtf16SurrogatePair.
-inline bool IsUtf16SurrogatePair(wchar_t first, wchar_t second) {
-  return sizeof(wchar_t) == 2 &&
-      (first & 0xFC00) == 0xD800 && (second & 0xFC00) == 0xDC00;
-}
-
-// Creates a Unicode code point from UTF16 surrogate pair.
-inline UInt32 CreateCodePointFromUtf16SurrogatePair(wchar_t first,
-                                                    wchar_t second) {
-  const UInt32 mask = (1 << 10) - 1;
-  return (sizeof(wchar_t) == 2) ?
-      (((first & mask) << 10) | (second & mask)) + 0x10000 :
-      // This function should not be called when the condition is
-      // false, but we provide a sensible default in case it is.
-      static_cast<UInt32>(first);
-}
-
-// Converts a wide string to a narrow string in UTF-8 encoding.
-// The wide string is assumed to have the following encoding:
-//   UTF-16 if sizeof(wchar_t) == 2 (on Windows, Cygwin, Symbian OS)
-//   UTF-32 if sizeof(wchar_t) == 4 (on Linux)
-// Parameter str points to a null-terminated wide string.
-// Parameter num_chars may additionally limit the number
-// of wchar_t characters processed. -1 is used when the entire string
-// should be processed.
-// If the string contains code points that are not valid Unicode code points
-// (i.e. outside of Unicode range U+0 to U+10FFFF) they will be output
-// as '(Invalid Unicode 0xXXXXXXXX)'. If the string is in UTF16 encoding
-// and contains invalid UTF-16 surrogate pairs, values in those pairs
-// will be encoded as individual Unicode characters from Basic Normal Plane.
-std::string WideStringToUtf8(const wchar_t* str, int num_chars) {
-  if (num_chars == -1)
-    num_chars = static_cast<int>(wcslen(str));
-
-  ::std::stringstream stream;
-  for (int i = 0; i < num_chars; ++i) {
-    UInt32 unicode_code_point;
-
-    if (str[i] == L'\0') {
-      break;
-    } else if (i + 1 < num_chars && IsUtf16SurrogatePair(str[i], str[i + 1])) {
-      unicode_code_point = CreateCodePointFromUtf16SurrogatePair(str[i],
-                                                                 str[i + 1]);
-      i++;
-    } else {
-      unicode_code_point = static_cast<UInt32>(str[i]);
-    }
-
-    stream << CodePointToUtf8(unicode_code_point);
-  }
-  return StringStreamToString(&stream);
-}
-
-// Converts a wide C string to an std::string using the UTF-8 encoding.
-// NULL will be converted to "(null)".
-std::string String::ShowWideCString(const wchar_t * wide_c_str) {
-  if (wide_c_str == NULL)  return "(null)";
-
-  return internal::WideStringToUtf8(wide_c_str, -1);
-}
-
-// Compares two wide C strings.  Returns true iff they have the same
-// content.
-//
-// Unlike wcscmp(), this function can handle NULL argument(s).  A NULL
-// C string is considered different to any non-NULL C string,
-// including the empty string.
-bool String::WideCStringEquals(const wchar_t * lhs, const wchar_t * rhs) {
-  if (lhs == NULL) return rhs == NULL;
-
-  if (rhs == NULL) return false;
-
-  return wcscmp(lhs, rhs) == 0;
-}
-
-// Helper function for *_STREQ on wide strings.
-AssertionResult CmpHelperSTREQ(const char* lhs_expression,
-                               const char* rhs_expression,
-                               const wchar_t* lhs,
-                               const wchar_t* rhs) {
-  if (String::WideCStringEquals(lhs, rhs)) {
-    return AssertionSuccess();
-  }
-
-  return EqFailure(lhs_expression,
-                   rhs_expression,
-                   PrintToString(lhs),
-                   PrintToString(rhs),
-                   false);
-}
-
-// Helper function for *_STRNE on wide strings.
-AssertionResult CmpHelperSTRNE(const char* s1_expression,
-                               const char* s2_expression,
-                               const wchar_t* s1,
-                               const wchar_t* s2) {
-  if (!String::WideCStringEquals(s1, s2)) {
-    return AssertionSuccess();
-  }
-
-  return AssertionFailure() << "Expected: (" << s1_expression << ") != ("
-                            << s2_expression << "), actual: "
-                            << PrintToString(s1)
-                            << " vs " << PrintToString(s2);
-}
-
-// Compares two C strings, ignoring case.  Returns true iff they have
-// the same content.
-//
-// Unlike strcasecmp(), this function can handle NULL argument(s).  A
-// NULL C string is considered different to any non-NULL C string,
-// including the empty string.
-bool String::CaseInsensitiveCStringEquals(const char * lhs, const char * rhs) {
-  if (lhs == NULL)
-    return rhs == NULL;
-  if (rhs == NULL)
-    return false;
-  return posix::StrCaseCmp(lhs, rhs) == 0;
-}
-
-  // Compares two wide C strings, ignoring case.  Returns true iff they
-  // have the same content.
-  //
-  // Unlike wcscasecmp(), this function can handle NULL argument(s).
-  // A NULL C string is considered different to any non-NULL wide C string,
-  // including the empty string.
-  // NB: The implementations on different platforms slightly differ.
-  // On windows, this method uses _wcsicmp which compares according to LC_CTYPE
-  // environment variable. On GNU platform this method uses wcscasecmp
-  // which compares according to LC_CTYPE category of the current locale.
-  // On MacOS X, it uses towlower, which also uses LC_CTYPE category of the
-  // current locale.
-bool String::CaseInsensitiveWideCStringEquals(const wchar_t* lhs,
-                                              const wchar_t* rhs) {
-  if (lhs == NULL) return rhs == NULL;
-
-  if (rhs == NULL) return false;
-
-#if GTEST_OS_WINDOWS
-  return _wcsicmp(lhs, rhs) == 0;
-#elif GTEST_OS_LINUX && !GTEST_OS_LINUX_ANDROID
-  return wcscasecmp(lhs, rhs) == 0;
-#else
-  // Android, Mac OS X and Cygwin don't define wcscasecmp.
-  // Other unknown OSes may not define it either.
-  wint_t left, right;
-  do {
-    left = towlower(*lhs++);
-    right = towlower(*rhs++);
-  } while (left && left == right);
-  return left == right;
-#endif  // OS selector
-}
-
-// Returns true iff str ends with the given suffix, ignoring case.
-// Any string is considered to end with an empty suffix.
-bool String::EndsWithCaseInsensitive(
-    const std::string& str, const std::string& suffix) {
-  const size_t str_len = str.length();
-  const size_t suffix_len = suffix.length();
-  return (str_len >= suffix_len) &&
-         CaseInsensitiveCStringEquals(str.c_str() + str_len - suffix_len,
-                                      suffix.c_str());
-}
-
-// Formats an int value as "%02d".
-std::string String::FormatIntWidth2(int value) {
-  std::stringstream ss;
-  ss << std::setfill('0') << std::setw(2) << value;
-  return ss.str();
-}
-
-// Formats an int value as "%X".
-std::string String::FormatHexInt(int value) {
-  std::stringstream ss;
-  ss << std::hex << std::uppercase << value;
-  return ss.str();
-}
-
-// Formats a byte as "%02X".
-std::string String::FormatByte(unsigned char value) {
-  std::stringstream ss;
-  ss << std::setfill('0') << std::setw(2) << std::hex << std::uppercase
-     << static_cast<unsigned int>(value);
-  return ss.str();
-}
-
-// Converts the buffer in a stringstream to an std::string, converting NUL
-// bytes to "\\0" along the way.
-std::string StringStreamToString(::std::stringstream* ss) {
-  const ::std::string& str = ss->str();
-  const char* const start = str.c_str();
-  const char* const end = start + str.length();
-
-  std::string result;
-  result.reserve(2 * (end - start));
-  for (const char* ch = start; ch != end; ++ch) {
-    if (*ch == '\0') {
-      result += "\\0";  // Replaces NUL with "\\0";
-    } else {
-      result += *ch;
-    }
-  }
-
-  return result;
-}
-
-// Appends the user-supplied message to the Google-Test-generated message.
-std::string AppendUserMessage(const std::string& gtest_msg,
-                              const Message& user_msg) {
-  // Appends the user message if it's non-empty.
-  const std::string user_msg_string = user_msg.GetString();
-  if (user_msg_string.empty()) {
-    return gtest_msg;
-  }
-
-  return gtest_msg + "\n" + user_msg_string;
-}
-
-}  // namespace internal
-
-// class TestResult
-
-// Creates an empty TestResult.
-TestResult::TestResult()
-    : death_test_count_(0),
-      elapsed_time_(0) {
-}
-
-// D'tor.
-TestResult::~TestResult() {
-}
-
-// Returns the i-th test part result among all the results. i can
-// range from 0 to total_part_count() - 1. If i is not in that range,
-// aborts the program.
-const TestPartResult& TestResult::GetTestPartResult(int i) const {
-  if (i < 0 || i >= total_part_count())
-    internal::posix::Abort();
-  return test_part_results_.at(i);
-}
-
-// Returns the i-th test property. i can range from 0 to
-// test_property_count() - 1. If i is not in that range, aborts the
-// program.
-const TestProperty& TestResult::GetTestProperty(int i) const {
-  if (i < 0 || i >= test_property_count())
-    internal::posix::Abort();
-  return test_properties_.at(i);
-}
-
-// Clears the test part results.
-void TestResult::ClearTestPartResults() {
-  test_part_results_.clear();
-}
-
-// Adds a test part result to the list.
-void TestResult::AddTestPartResult(const TestPartResult& test_part_result) {
-  test_part_results_.push_back(test_part_result);
-}
-
-// Adds a test property to the list. If a property with the same key as the
-// supplied property is already represented, the value of this test_property
-// replaces the old value for that key.
-void TestResult::RecordProperty(const std::string& xml_element,
-                                const TestProperty& test_property) {
-  if (!ValidateTestProperty(xml_element, test_property)) {
-    return;
-  }
-  internal::MutexLock lock(&test_properites_mutex_);
-  const std::vector<TestProperty>::iterator property_with_matching_key =
-      std::find_if(test_properties_.begin(), test_properties_.end(),
-                   internal::TestPropertyKeyIs(test_property.key()));
-  if (property_with_matching_key == test_properties_.end()) {
-    test_properties_.push_back(test_property);
-    return;
-  }
-  property_with_matching_key->SetValue(test_property.value());
-}
-
-// The list of reserved attributes used in the <testsuites> element of XML
-// output.
-static const char* const kReservedTestSuitesAttributes[] = {
-  "disabled",
-  "errors",
-  "failures",
-  "name",
-  "random_seed",
-  "tests",
-  "time",
-  "timestamp"
-};
-
-// The list of reserved attributes used in the <testsuite> element of XML
-// output.
-static const char* const kReservedTestSuiteAttributes[] = {
-  "disabled",
-  "errors",
-  "failures",
-  "name",
-  "tests",
-  "time"
-};
-
-// The list of reserved attributes used in the <testcase> element of XML output.
-static const char* const kReservedTestCaseAttributes[] = {
-  "classname",
-  "name",
-  "status",
-  "time",
-  "type_param",
-  "value_param"
-};
-
-template <int kSize>
-std::vector<std::string> ArrayAsVector(const char* const (&array)[kSize]) {
-  return std::vector<std::string>(array, array + kSize);
-}
-
-static std::vector<std::string> GetReservedAttributesForElement(
-    const std::string& xml_element) {
-  if (xml_element == "testsuites") {
-    return ArrayAsVector(kReservedTestSuitesAttributes);
-  } else if (xml_element == "testsuite") {
-    return ArrayAsVector(kReservedTestSuiteAttributes);
-  } else if (xml_element == "testcase") {
-    return ArrayAsVector(kReservedTestCaseAttributes);
-  } else {
-    GTEST_CHECK_(false) << "Unrecognized xml_element provided: " << xml_element;
-  }
-  // This code is unreachable but some compilers may not realizes that.
-  return std::vector<std::string>();
-}
-
-static std::string FormatWordList(const std::vector<std::string>& words) {
-  Message word_list;
-  for (size_t i = 0; i < words.size(); ++i) {
-    if (i > 0 && words.size() > 2) {
-      word_list << ", ";
-    }
-    if (i == words.size() - 1) {
-      word_list << "and ";
-    }
-    word_list << "'" << words[i] << "'";
-  }
-  return word_list.GetString();
-}
-
-static bool ValidateTestPropertyName(
-    const std::string& property_name,
-    const std::vector<std::string>& reserved_names) {
-  if (std::find(reserved_names.begin(), reserved_names.end(), property_name) !=
-          reserved_names.end()) {
-    ADD_FAILURE() << "Reserved key used in RecordProperty(): " << property_name
-                  << " (" << FormatWordList(reserved_names)
-                  << " are reserved by " << GTEST_NAME_ << ")";
-    return false;
-  }
-  return true;
-}
-
-// Adds a failure if the key is a reserved attribute of the element named
-// xml_element.  Returns true if the property is valid.
-bool TestResult::ValidateTestProperty(const std::string& xml_element,
-                                      const TestProperty& test_property) {
-  return ValidateTestPropertyName(test_property.key(),
-                                  GetReservedAttributesForElement(xml_element));
-}
-
-// Clears the object.
-void TestResult::Clear() {
-  test_part_results_.clear();
-  test_properties_.clear();
-  death_test_count_ = 0;
-  elapsed_time_ = 0;
-}
-
-// Returns true iff the test failed.
-bool TestResult::Failed() const {
-  for (int i = 0; i < total_part_count(); ++i) {
-    if (GetTestPartResult(i).failed())
-      return true;
-  }
-  return false;
-}
-
-// Returns true iff the test part fatally failed.
-static bool TestPartFatallyFailed(const TestPartResult& result) {
-  return result.fatally_failed();
-}
-
-// Returns true iff the test fatally failed.
-bool TestResult::HasFatalFailure() const {
-  return CountIf(test_part_results_, TestPartFatallyFailed) > 0;
-}
-
-// Returns true iff the test part non-fatally failed.
-static bool TestPartNonfatallyFailed(const TestPartResult& result) {
-  return result.nonfatally_failed();
-}
-
-// Returns true iff the test has a non-fatal failure.
-bool TestResult::HasNonfatalFailure() const {
-  return CountIf(test_part_results_, TestPartNonfatallyFailed) > 0;
-}
-
-// Gets the number of all test parts.  This is the sum of the number
-// of successful test parts and the number of failed test parts.
-int TestResult::total_part_count() const {
-  return static_cast<int>(test_part_results_.size());
-}
-
-// Returns the number of the test properties.
-int TestResult::test_property_count() const {
-  return static_cast<int>(test_properties_.size());
-}
-
-// class Test
-
-// Creates a Test object.
-
-// The c'tor saves the states of all flags.
-Test::Test()
-    : gtest_flag_saver_(new GTEST_FLAG_SAVER_) {
-}
-
-// The d'tor restores the states of all flags.  The actual work is
-// done by the d'tor of the gtest_flag_saver_ field, and thus not
-// visible here.
-Test::~Test() {
-}
-
-// Sets up the test fixture.
-//
-// A sub-class may override this.
-void Test::SetUp() {
-}
-
-// Tears down the test fixture.
-//
-// A sub-class may override this.
-void Test::TearDown() {
-}
-
-// Allows user supplied key value pairs to be recorded for later output.
-void Test::RecordProperty(const std::string& key, const std::string& value) {
-  UnitTest::GetInstance()->RecordProperty(key, value);
-}
-
-// Allows user supplied key value pairs to be recorded for later output.
-void Test::RecordProperty(const std::string& key, int value) {
-  Message value_message;
-  value_message << value;
-  RecordProperty(key, value_message.GetString().c_str());
-}
-
-namespace internal {
-
-void ReportFailureInUnknownLocation(TestPartResult::Type result_type,
-                                    const std::string& message) {
-  // This function is a friend of UnitTest and as such has access to
-  // AddTestPartResult.
-  UnitTest::GetInstance()->AddTestPartResult(
-      result_type,
-      NULL,  // No info about the source file where the exception occurred.
-      -1,    // We have no info on which line caused the exception.
-      message,
-      "");   // No stack trace, either.
-}
-
-}  // namespace internal
-
-// Google Test requires all tests in the same test case to use the same test
-// fixture class.  This function checks if the current test has the
-// same fixture class as the first test in the current test case.  If
-// yes, it returns true; otherwise it generates a Google Test failure and
-// returns false.
-bool Test::HasSameFixtureClass() {
-  internal::UnitTestImpl* const impl = internal::GetUnitTestImpl();
-  const TestCase* const test_case = impl->current_test_case();
-
-  // Info about the first test in the current test case.
-  const TestInfo* const first_test_info = test_case->test_info_list()[0];
-  const internal::TypeId first_fixture_id = first_test_info->fixture_class_id_;
-  const char* const first_test_name = first_test_info->name();
-
-  // Info about the current test.
-  const TestInfo* const this_test_info = impl->current_test_info();
-  const internal::TypeId this_fixture_id = this_test_info->fixture_class_id_;
-  const char* const this_test_name = this_test_info->name();
-
-  if (this_fixture_id != first_fixture_id) {
-    // Is the first test defined using TEST?
-    const bool first_is_TEST = first_fixture_id == internal::GetTestTypeId();
-    // Is this test defined using TEST?
-    const bool this_is_TEST = this_fixture_id == internal::GetTestTypeId();
-
-    if (first_is_TEST || this_is_TEST) {
-      // Both TEST and TEST_F appear in same test case, which is incorrect.
-      // Tell the user how to fix this.
-
-      // Gets the name of the TEST and the name of the TEST_F.  Note
-      // that first_is_TEST and this_is_TEST cannot both be true, as
-      // the fixture IDs are different for the two tests.
-      const char* const TEST_name =
-          first_is_TEST ? first_test_name : this_test_name;
-      const char* const TEST_F_name =
-          first_is_TEST ? this_test_name : first_test_name;
-
-      ADD_FAILURE()
-          << "All tests in the same test case must use the same test fixture\n"
-          << "class, so mixing TEST_F and TEST in the same test case is\n"
-          << "illegal.  In test case " << this_test_info->test_case_name()
-          << ",\n"
-          << "test " << TEST_F_name << " is defined using TEST_F but\n"
-          << "test " << TEST_name << " is defined using TEST.  You probably\n"
-          << "want to change the TEST to TEST_F or move it to another test\n"
-          << "case.";
-    } else {
-      // Two fixture classes with the same name appear in two different
-      // namespaces, which is not allowed. Tell the user how to fix this.
-      ADD_FAILURE()
-          << "All tests in the same test case must use the same test fixture\n"
-          << "class.  However, in test case "
-          << this_test_info->test_case_name() << ",\n"
-          << "you defined test " << first_test_name
-          << " and test " << this_test_name << "\n"
-          << "using two different test fixture classes.  This can happen if\n"
-          << "the two classes are from different namespaces or translation\n"
-          << "units and have the same name.  You should probably rename one\n"
-          << "of the classes to put the tests into different test cases.";
-    }
-    return false;
-  }
-
-  return true;
-}
-
-#if GTEST_HAS_SEH
-
-// Adds an "exception thrown" fatal failure to the current test.  This
-// function returns its result via an output parameter pointer because VC++
-// prohibits creation of objects with destructors on stack in functions
-// using __try (see error C2712).
-static std::string* FormatSehExceptionMessage(DWORD exception_code,
-                                              const char* location) {
-  Message message;
-  message << "SEH exception with code 0x" << std::setbase(16) <<
-    exception_code << std::setbase(10) << " thrown in " << location << ".";
-
-  return new std::string(message.GetString());
-}
-
-#endif  // GTEST_HAS_SEH
-
-namespace internal {
-
-#if GTEST_HAS_EXCEPTIONS
-
-// Adds an "exception thrown" fatal failure to the current test.
-static std::string FormatCxxExceptionMessage(const char* description,
-                                             const char* location) {
-  Message message;
-  if (description != NULL) {
-    message << "C++ exception with description \"" << description << "\"";
-  } else {
-    message << "Unknown C++ exception";
-  }
-  message << " thrown in " << location << ".";
-
-  return message.GetString();
-}
-
-static std::string PrintTestPartResultToString(
-    const TestPartResult& test_part_result);
-
-GoogleTestFailureException::GoogleTestFailureException(
-    const TestPartResult& failure)
-    : ::std::runtime_error(PrintTestPartResultToString(failure).c_str()) {}
-
-#endif  // GTEST_HAS_EXCEPTIONS
-
-// We put these helper functions in the internal namespace as IBM's xlC
-// compiler rejects the code if they were declared static.
-
-// Runs the given method and handles SEH exceptions it throws, when
-// SEH is supported; returns the 0-value for type Result in case of an
-// SEH exception.  (Microsoft compilers cannot handle SEH and C++
-// exceptions in the same function.  Therefore, we provide a separate
-// wrapper function for handling SEH exceptions.)
-template <class T, typename Result>
-Result HandleSehExceptionsInMethodIfSupported(
-    T* object, Result (T::*method)(), const char* location) {
-#if GTEST_HAS_SEH
-  __try {
-    return (object->*method)();
-  } __except (internal::UnitTestOptions::GTestShouldProcessSEH(  // NOLINT
-      GetExceptionCode())) {
-    // We create the exception message on the heap because VC++ prohibits
-    // creation of objects with destructors on stack in functions using __try
-    // (see error C2712).
-    std::string* exception_message = FormatSehExceptionMessage(
-        GetExceptionCode(), location);
-    internal::ReportFailureInUnknownLocation(TestPartResult::kFatalFailure,
-                                             *exception_message);
-    delete exception_message;
-    return static_cast<Result>(0);
-  }
-#else
-  (void)location;
-  return (object->*method)();
-#endif  // GTEST_HAS_SEH
-}
-
-// Runs the given method and catches and reports C++ and/or SEH-style
-// exceptions, if they are supported; returns the 0-value for type
-// Result in case of an SEH exception.
-template <class T, typename Result>
-Result HandleExceptionsInMethodIfSupported(
-    T* object, Result (T::*method)(), const char* location) {
-  // NOTE: The user code can affect the way in which Google Test handles
-  // exceptions by setting GTEST_FLAG(catch_exceptions), but only before
-  // RUN_ALL_TESTS() starts. It is technically possible to check the flag
-  // after the exception is caught and either report or re-throw the
-  // exception based on the flag's value:
-  //
-  // try {
-  //   // Perform the test method.
-  // } catch (...) {
-  //   if (GTEST_FLAG(catch_exceptions))
-  //     // Report the exception as failure.
-  //   else
-  //     throw;  // Re-throws the original exception.
-  // }
-  //
-  // However, the purpose of this flag is to allow the program to drop into
-  // the debugger when the exception is thrown. On most platforms, once the
-  // control enters the catch block, the exception origin information is
-  // lost and the debugger will stop the program at the point of the
-  // re-throw in this function -- instead of at the point of the original
-  // throw statement in the code under test.  For this reason, we perform
-  // the check early, sacrificing the ability to affect Google Test's
-  // exception handling in the method where the exception is thrown.
-  if (internal::GetUnitTestImpl()->catch_exceptions()) {
-#if GTEST_HAS_EXCEPTIONS
-    try {
-      return HandleSehExceptionsInMethodIfSupported(object, method, location);
-    } catch (const AssertionException&) {  // NOLINT
-      // This failure was reported already.
-    } catch (const internal::GoogleTestFailureException&) {  // NOLINT
-      // This exception type can only be thrown by a failed Google
-      // Test assertion with the intention of letting another testing
-      // framework catch it.  Therefore we just re-throw it.
-      throw;
-    } catch (const std::exception& e) {  // NOLINT
-      internal::ReportFailureInUnknownLocation(
-          TestPartResult::kFatalFailure,
-          FormatCxxExceptionMessage(e.what(), location));
-    } catch (...) {  // NOLINT
-      internal::ReportFailureInUnknownLocation(
-          TestPartResult::kFatalFailure,
-          FormatCxxExceptionMessage(NULL, location));
-    }
-    return static_cast<Result>(0);
-#else
-    return HandleSehExceptionsInMethodIfSupported(object, method, location);
-#endif  // GTEST_HAS_EXCEPTIONS
-  } else {
-    return (object->*method)();
-  }
-}
-
-}  // namespace internal
-
-// Runs the test and updates the test result.
-void Test::Run() {
-  if (!HasSameFixtureClass()) return;
-
-  internal::UnitTestImpl* const impl = internal::GetUnitTestImpl();
-  impl->os_stack_trace_getter()->UponLeavingGTest();
-  internal::HandleExceptionsInMethodIfSupported(this, &Test::SetUp, "SetUp()");
-  // We will run the test only if SetUp() was successful.
-  if (!HasFatalFailure()) {
-    impl->os_stack_trace_getter()->UponLeavingGTest();
-    internal::HandleExceptionsInMethodIfSupported(
-        this, &Test::TestBody, "the test body");
-  }
-
-  // However, we want to clean up as much as possible.  Hence we will
-  // always call TearDown(), even if SetUp() or the test body has
-  // failed.
-  impl->os_stack_trace_getter()->UponLeavingGTest();
-  internal::HandleExceptionsInMethodIfSupported(
-      this, &Test::TearDown, "TearDown()");
-}
-
-// Returns true iff the current test has a fatal failure.
-bool Test::HasFatalFailure() {
-  return internal::GetUnitTestImpl()->current_test_result()->HasFatalFailure();
-}
-
-// Returns true iff the current test has a non-fatal failure.
-bool Test::HasNonfatalFailure() {
-  return internal::GetUnitTestImpl()->current_test_result()->
-      HasNonfatalFailure();
-}
-
-// class TestInfo
-
-// Constructs a TestInfo object. It assumes ownership of the test factory
-// object.
-TestInfo::TestInfo(const std::string& a_test_case_name,
-                   const std::string& a_name,
-                   const char* a_type_param,
-                   const char* a_value_param,
-                   internal::CodeLocation a_code_location,
-                   internal::TypeId fixture_class_id,
-                   internal::TestFactoryBase* factory)
-    : test_case_name_(a_test_case_name),
-      name_(a_name),
-      type_param_(a_type_param ? new std::string(a_type_param) : NULL),
-      value_param_(a_value_param ? new std::string(a_value_param) : NULL),
-      location_(a_code_location),
-      fixture_class_id_(fixture_class_id),
-      should_run_(false),
-      is_disabled_(false),
-      matches_filter_(false),
-      factory_(factory),
-      result_() {}
-
-// Destructs a TestInfo object.
-TestInfo::~TestInfo() { delete factory_; }
-
-namespace internal {
-
-// Creates a new TestInfo object and registers it with Google Test;
-// returns the created object.
-//
-// Arguments:
-//
-//   test_case_name:   name of the test case
-//   name:             name of the test
-//   type_param:       the name of the test's type parameter, or NULL if
-//                     this is not a typed or a type-parameterized test.
-//   value_param:      text representation of the test's value parameter,
-//                     or NULL if this is not a value-parameterized test.
-//   code_location:    code location where the test is defined
-//   fixture_class_id: ID of the test fixture class
-//   set_up_tc:        pointer to the function that sets up the test case
-//   tear_down_tc:     pointer to the function that tears down the test case
-//   factory:          pointer to the factory that creates a test object.
-//                     The newly created TestInfo instance will assume
-//                     ownership of the factory object.
-TestInfo* MakeAndRegisterTestInfo(
-    const char* test_case_name,
-    const char* name,
-    const char* type_param,
-    const char* value_param,
-    CodeLocation code_location,
-    TypeId fixture_class_id,
-    SetUpTestCaseFunc set_up_tc,
-    TearDownTestCaseFunc tear_down_tc,
-    TestFactoryBase* factory) {
-  TestInfo* const test_info =
-      new TestInfo(test_case_name, name, type_param, value_param,
-                   code_location, fixture_class_id, factory);
-  GetUnitTestImpl()->AddTestInfo(set_up_tc, tear_down_tc, test_info);
-  return test_info;
-}
-
-void ReportInvalidTestCaseType(const char* test_case_name,
-                               CodeLocation code_location) {
-  Message errors;
-  errors
-      << "Attempted redefinition of test case " << test_case_name << ".\n"
-      << "All tests in the same test case must use the same test fixture\n"
-      << "class.  However, in test case " << test_case_name << ", you tried\n"
-      << "to define a test using a fixture class different from the one\n"
-      << "used earlier. This can happen if the two fixture classes are\n"
-      << "from different namespaces and have the same name. You should\n"
-      << "probably rename one of the classes to put the tests into different\n"
-      << "test cases.";
-
-  GTEST_LOG_(ERROR) << FormatFileLocation(code_location.file.c_str(),
-                                          code_location.line)
-                    << " " << errors.GetString();
-}
-}  // namespace internal
-
-namespace {
-
-// A predicate that checks the test name of a TestInfo against a known
-// value.
-//
-// This is used for implementation of the TestCase class only.  We put
-// it in the anonymous namespace to prevent polluting the outer
-// namespace.
-//
-// TestNameIs is copyable.
-class TestNameIs {
- public:
-  // Constructor.
-  //
-  // TestNameIs has NO default constructor.
-  explicit TestNameIs(const char* name)
-      : name_(name) {}
-
-  // Returns true iff the test name of test_info matches name_.
-  bool operator()(const TestInfo * test_info) const {
-    return test_info && test_info->name() == name_;
-  }
-
- private:
-  std::string name_;
-};
-
-}  // namespace
-
-namespace internal {
-
-// This method expands all parameterized tests registered with macros TEST_P
-// and INSTANTIATE_TEST_CASE_P into regular tests and registers those.
-// This will be done just once during the program runtime.
-void UnitTestImpl::RegisterParameterizedTests() {
-  if (!parameterized_tests_registered_) {
-    parameterized_test_registry_.RegisterTests();
-    parameterized_tests_registered_ = true;
-  }
-}
-
-}  // namespace internal
-
-// Creates the test object, runs it, records its result, and then
-// deletes it.
-void TestInfo::Run() {
-  if (!should_run_) return;
-
-  // Tells UnitTest where to store test result.
-  internal::UnitTestImpl* const impl = internal::GetUnitTestImpl();
-  impl->set_current_test_info(this);
-
-  TestEventListener* repeater = UnitTest::GetInstance()->listeners().repeater();
-
-  // Notifies the unit test event listeners that a test is about to start.
-  repeater->OnTestStart(*this);
-
-  const TimeInMillis start = internal::GetTimeInMillis();
-
-  impl->os_stack_trace_getter()->UponLeavingGTest();
-
-  // Creates the test object.
-  Test* const test = internal::HandleExceptionsInMethodIfSupported(
-      factory_, &internal::TestFactoryBase::CreateTest,
-      "the test fixture's constructor");
-
-  // Runs the test only if the test object was created and its
-  // constructor didn't generate a fatal failure.
-  if ((test != NULL) && !Test::HasFatalFailure()) {
-    // This doesn't throw as all user code that can throw are wrapped into
-    // exception handling code.
-    test->Run();
-  }
-
-  // Deletes the test object.
-  impl->os_stack_trace_getter()->UponLeavingGTest();
-  internal::HandleExceptionsInMethodIfSupported(
-      test, &Test::DeleteSelf_, "the test fixture's destructor");
-
-  result_.set_elapsed_time(internal::GetTimeInMillis() - start);
-
-  // Notifies the unit test event listener that a test has just finished.
-  repeater->OnTestEnd(*this);
-
-  // Tells UnitTest to stop associating assertion results to this
-  // test.
-  impl->set_current_test_info(NULL);
-}
-
-// class TestCase
-
-// Gets the number of successful tests in this test case.
-int TestCase::successful_test_count() const {
-  return CountIf(test_info_list_, TestPassed);
-}
-
-// Gets the number of failed tests in this test case.
-int TestCase::failed_test_count() const {
-  return CountIf(test_info_list_, TestFailed);
-}
-
-// Gets the number of disabled tests that will be reported in the XML report.
-int TestCase::reportable_disabled_test_count() const {
-  return CountIf(test_info_list_, TestReportableDisabled);
-}
-
-// Gets the number of disabled tests in this test case.
-int TestCase::disabled_test_count() const {
-  return CountIf(test_info_list_, TestDisabled);
-}
-
-// Gets the number of tests to be printed in the XML report.
-int TestCase::reportable_test_count() const {
-  return CountIf(test_info_list_, TestReportable);
-}
-
-// Get the number of tests in this test case that should run.
-int TestCase::test_to_run_count() const {
-  return CountIf(test_info_list_, ShouldRunTest);
-}
-
-// Gets the number of all tests.
-int TestCase::total_test_count() const {
-  return static_cast<int>(test_info_list_.size());
-}
-
-// Creates a TestCase with the given name.
-//
-// Arguments:
-//
-//   name:         name of the test case
-//   a_type_param: the name of the test case's type parameter, or NULL if
-//                 this is not a typed or a type-parameterized test case.
-//   set_up_tc:    pointer to the function that sets up the test case
-//   tear_down_tc: pointer to the function that tears down the test case
-TestCase::TestCase(const char* a_name, const char* a_type_param,
-                   Test::SetUpTestCaseFunc set_up_tc,
-                   Test::TearDownTestCaseFunc tear_down_tc)
-    : name_(a_name),
-      type_param_(a_type_param ? new std::string(a_type_param) : NULL),
-      set_up_tc_(set_up_tc),
-      tear_down_tc_(tear_down_tc),
-      should_run_(false),
-      elapsed_time_(0) {
-}
-
-// Destructor of TestCase.
-TestCase::~TestCase() {
-  // Deletes every Test in the collection.
-  ForEach(test_info_list_, internal::Delete<TestInfo>);
-}
-
-// Returns the i-th test among all the tests. i can range from 0 to
-// total_test_count() - 1. If i is not in that range, returns NULL.
-const TestInfo* TestCase::GetTestInfo(int i) const {
-  const int index = GetElementOr(test_indices_, i, -1);
-  return index < 0 ? NULL : test_info_list_[index];
-}
-
-// Returns the i-th test among all the tests. i can range from 0 to
-// total_test_count() - 1. If i is not in that range, returns NULL.
-TestInfo* TestCase::GetMutableTestInfo(int i) {
-  const int index = GetElementOr(test_indices_, i, -1);
-  return index < 0 ? NULL : test_info_list_[index];
-}
-
-// Adds a test to this test case.  Will delete the test upon
-// destruction of the TestCase object.
-void TestCase::AddTestInfo(TestInfo * test_info) {
-  test_info_list_.push_back(test_info);
-  test_indices_.push_back(static_cast<int>(test_indices_.size()));
-}
-
-// Runs every test in this TestCase.
-void TestCase::Run() {
-  if (!should_run_) return;
-
-  internal::UnitTestImpl* const impl = internal::GetUnitTestImpl();
-  impl->set_current_test_case(this);
-
-  TestEventListener* repeater = UnitTest::GetInstance()->listeners().repeater();
-
-  repeater->OnTestCaseStart(*this);
-  impl->os_stack_trace_getter()->UponLeavingGTest();
-  internal::HandleExceptionsInMethodIfSupported(
-      this, &TestCase::RunSetUpTestCase, "SetUpTestCase()");
-
-  const internal::TimeInMillis start = internal::GetTimeInMillis();
-  for (int i = 0; i < total_test_count(); i++) {
-    GetMutableTestInfo(i)->Run();
-  }
-  elapsed_time_ = internal::GetTimeInMillis() - start;
-
-  impl->os_stack_trace_getter()->UponLeavingGTest();
-  internal::HandleExceptionsInMethodIfSupported(
-      this, &TestCase::RunTearDownTestCase, "TearDownTestCase()");
-
-  repeater->OnTestCaseEnd(*this);
-  impl->set_current_test_case(NULL);
-}
-
-// Clears the results of all tests in this test case.
-void TestCase::ClearResult() {
-  ad_hoc_test_result_.Clear();
-  ForEach(test_info_list_, TestInfo::ClearTestResult);
-}
-
-// Shuffles the tests in this test case.
-void TestCase::ShuffleTests(internal::Random* random) {
-  Shuffle(random, &test_indices_);
-}
-
-// Restores the test order to before the first shuffle.
-void TestCase::UnshuffleTests() {
-  for (size_t i = 0; i < test_indices_.size(); i++) {
-    test_indices_[i] = static_cast<int>(i);
-  }
-}
-
-// Formats a countable noun.  Depending on its quantity, either the
-// singular form or the plural form is used. e.g.
-//
-// FormatCountableNoun(1, "formula", "formuli") returns "1 formula".
-// FormatCountableNoun(5, "book", "books") returns "5 books".
-static std::string FormatCountableNoun(int count,
-                                       const char * singular_form,
-                                       const char * plural_form) {
-  return internal::StreamableToString(count) + " " +
-      (count == 1 ? singular_form : plural_form);
-}
-
-// Formats the count of tests.
-static std::string FormatTestCount(int test_count) {
-  return FormatCountableNoun(test_count, "test", "tests");
-}
-
-// Formats the count of test cases.
-static std::string FormatTestCaseCount(int test_case_count) {
-  return FormatCountableNoun(test_case_count, "test case", "test cases");
-}
-
-// Converts a TestPartResult::Type enum to human-friendly string
-// representation.  Both kNonFatalFailure and kFatalFailure are translated
-// to "Failure", as the user usually doesn't care about the difference
-// between the two when viewing the test result.
-static const char * TestPartResultTypeToString(TestPartResult::Type type) {
-  switch (type) {
-    case TestPartResult::kSuccess:
-      return "Success";
-
-    case TestPartResult::kNonFatalFailure:
-    case TestPartResult::kFatalFailure:
-#ifdef _MSC_VER
-      return "error: ";
-#else
-      return "Failure\n";
-#endif
-    default:
-      return "Unknown result type";
-  }
-}
-
-namespace internal {
-
-// Prints a TestPartResult to an std::string.
-static std::string PrintTestPartResultToString(
-    const TestPartResult& test_part_result) {
-  return (Message()
-          << internal::FormatFileLocation(test_part_result.file_name(),
-                                          test_part_result.line_number())
-          << " " << TestPartResultTypeToString(test_part_result.type())
-          << test_part_result.message()).GetString();
-}
-
-// Prints a TestPartResult.
-static void PrintTestPartResult(const TestPartResult& test_part_result) {
-  const std::string& result =
-      PrintTestPartResultToString(test_part_result);
-  printf("%s\n", result.c_str());
-  fflush(stdout);
-  // If the test program runs in Visual Studio or a debugger, the
-  // following statements add the test part result message to the Output
-  // window such that the user can double-click on it to jump to the
-  // corresponding source code location; otherwise they do nothing.
-#if GTEST_OS_WINDOWS && !GTEST_OS_WINDOWS_MOBILE
-  // We don't call OutputDebugString*() on Windows Mobile, as printing
-  // to stdout is done by OutputDebugString() there already - we don't
-  // want the same message printed twice.
-  ::OutputDebugStringA(result.c_str());
-  ::OutputDebugStringA("\n");
-#endif
-}
-
-// class PrettyUnitTestResultPrinter
-
-enum GTestColor {
-  COLOR_DEFAULT,
-  COLOR_RED,
-  COLOR_GREEN,
-  COLOR_YELLOW
-};
-
-#if GTEST_OS_WINDOWS && !GTEST_OS_WINDOWS_MOBILE && \
-    !GTEST_OS_WINDOWS_PHONE && !GTEST_OS_WINDOWS_RT && !GTEST_OS_WINDOWS_MINGW
-
-// Returns the character attribute for the given color.
-static WORD GetColorAttribute(GTestColor color) {
-  switch (color) {
-    case COLOR_RED:    return FOREGROUND_RED;
-    case COLOR_GREEN:  return FOREGROUND_GREEN;
-    case COLOR_YELLOW: return FOREGROUND_RED | FOREGROUND_GREEN;
-    default:           return 0;
-  }
-}
-
-static int GetBitOffset(WORD color_mask) {
-  if (color_mask == 0) return 0;
-
-  int bitOffset = 0;
-  while ((color_mask & 1) == 0) {
-    color_mask >>= 1;
-    ++bitOffset;
-  }
-  return bitOffset;
-}
-
-static WORD GetNewColor(GTestColor color, WORD old_color_attrs) {
-  // Let's reuse the BG
-  static const WORD background_mask = BACKGROUND_BLUE | BACKGROUND_GREEN |
-                                      BACKGROUND_RED | BACKGROUND_INTENSITY;
-  static const WORD foreground_mask = FOREGROUND_BLUE | FOREGROUND_GREEN |
-                                      FOREGROUND_RED | FOREGROUND_INTENSITY;
-  const WORD existing_bg = old_color_attrs & background_mask;
-
-  WORD new_color =
-      GetColorAttribute(color) | existing_bg | FOREGROUND_INTENSITY;
-  static const int bg_bitOffset = GetBitOffset(background_mask);
-  static const int fg_bitOffset = GetBitOffset(foreground_mask);
-
-  if (((new_color & background_mask) >> bg_bitOffset) ==
-      ((new_color & foreground_mask) >> fg_bitOffset)) {
-    new_color ^= FOREGROUND_INTENSITY;  // invert intensity
-  }
-  return new_color;
-}
-
-#else
-
-// Returns the ANSI color code for the given color.  COLOR_DEFAULT is
-// an invalid input.
-static const char* GetAnsiColorCode(GTestColor color) {
-  switch (color) {
-    case COLOR_RED:     return "1";
-    case COLOR_GREEN:   return "2";
-    case COLOR_YELLOW:  return "3";
-    default:            return NULL;
-  };
-}
-
-#endif  // GTEST_OS_WINDOWS && !GTEST_OS_WINDOWS_MOBILE
-
-// Returns true iff Google Test should use colors in the output.
-bool ShouldUseColor(bool stdout_is_tty) {
-  const char* const gtest_color = GTEST_FLAG(color).c_str();
-
-  if (String::CaseInsensitiveCStringEquals(gtest_color, "auto")) {
-#if GTEST_OS_WINDOWS && !GTEST_OS_WINDOWS_MINGW
-    // On Windows the TERM variable is usually not set, but the
-    // console there does support colors.
-    return stdout_is_tty;
-#else
-    // On non-Windows platforms, we rely on the TERM variable.
-    const char* const term = posix::GetEnv("TERM");
-    const bool term_supports_color =
-        String::CStringEquals(term, "xterm") ||
-        String::CStringEquals(term, "xterm-color") ||
-        String::CStringEquals(term, "xterm-256color") ||
-        String::CStringEquals(term, "screen") ||
-        String::CStringEquals(term, "screen-256color") ||
-        String::CStringEquals(term, "tmux") ||
-        String::CStringEquals(term, "tmux-256color") ||
-        String::CStringEquals(term, "rxvt-unicode") ||
-        String::CStringEquals(term, "rxvt-unicode-256color") ||
-        String::CStringEquals(term, "linux") ||
-        String::CStringEquals(term, "cygwin");
-    return stdout_is_tty && term_supports_color;
-#endif  // GTEST_OS_WINDOWS
-  }
-
-  return String::CaseInsensitiveCStringEquals(gtest_color, "yes") ||
-      String::CaseInsensitiveCStringEquals(gtest_color, "true") ||
-      String::CaseInsensitiveCStringEquals(gtest_color, "t") ||
-      String::CStringEquals(gtest_color, "1");
-  // We take "yes", "true", "t", and "1" as meaning "yes".  If the
-  // value is neither one of these nor "auto", we treat it as "no" to
-  // be conservative.
-}
-
-// Helpers for printing colored strings to stdout. Note that on Windows, we
-// cannot simply emit special characters and have the terminal change colors.
-// This routine must actually emit the characters rather than return a string
-// that would be colored when printed, as can be done on Linux.
-static void ColoredPrintf(GTestColor color, const char* fmt, ...) {
-  va_list args;
-  va_start(args, fmt);
-
-#if GTEST_OS_WINDOWS_MOBILE || GTEST_OS_SYMBIAN || GTEST_OS_ZOS || \
-    GTEST_OS_IOS || GTEST_OS_WINDOWS_PHONE || GTEST_OS_WINDOWS_RT
-  const bool use_color = AlwaysFalse();
-#else
-  static const bool in_color_mode =
-      ShouldUseColor(posix::IsATTY(posix::FileNo(stdout)) != 0);
-  const bool use_color = in_color_mode && (color != COLOR_DEFAULT);
-#endif  // GTEST_OS_WINDOWS_MOBILE || GTEST_OS_SYMBIAN || GTEST_OS_ZOS
-  // The '!= 0' comparison is necessary to satisfy MSVC 7.1.
-
-  if (!use_color) {
-    vprintf(fmt, args);
-    va_end(args);
-    return;
-  }
-
-#if GTEST_OS_WINDOWS && !GTEST_OS_WINDOWS_MOBILE && \
-    !GTEST_OS_WINDOWS_PHONE && !GTEST_OS_WINDOWS_RT && !GTEST_OS_WINDOWS_MINGW
-  const HANDLE stdout_handle = GetStdHandle(STD_OUTPUT_HANDLE);
-
-  // Gets the current text color.
-  CONSOLE_SCREEN_BUFFER_INFO buffer_info;
-  GetConsoleScreenBufferInfo(stdout_handle, &buffer_info);
-  const WORD old_color_attrs = buffer_info.wAttributes;
-  const WORD new_color = GetNewColor(color, old_color_attrs);
-
-  // We need to flush the stream buffers into the console before each
-  // SetConsoleTextAttribute call lest it affect the text that is already
-  // printed but has not yet reached the console.
-  fflush(stdout);
-  SetConsoleTextAttribute(stdout_handle, new_color);
-
-  vprintf(fmt, args);
-
-  fflush(stdout);
-  // Restores the text color.
-  SetConsoleTextAttribute(stdout_handle, old_color_attrs);
-#else
-  printf("\033[0;3%sm", GetAnsiColorCode(color));
-  vprintf(fmt, args);
-  printf("\033[m");  // Resets the terminal to default.
-#endif  // GTEST_OS_WINDOWS && !GTEST_OS_WINDOWS_MOBILE
-  va_end(args);
-}
-
-// Text printed in Google Test's text output and --gtest_list_tests
-// output to label the type parameter and value parameter for a test.
-static const char kTypeParamLabel[] = "TypeParam";
-static const char kValueParamLabel[] = "GetParam()";
-
-static void PrintFullTestCommentIfPresent(const TestInfo& test_info) {
-  const char* const type_param = test_info.type_param();
-  const char* const value_param = test_info.value_param();
-
-  if (type_param != NULL || value_param != NULL) {
-    printf(", where ");
-    if (type_param != NULL) {
-      printf("%s = %s", kTypeParamLabel, type_param);
-      if (value_param != NULL)
-        printf(" and ");
-    }
-    if (value_param != NULL) {
-      printf("%s = %s", kValueParamLabel, value_param);
-    }
-  }
-}
-
-// This class implements the TestEventListener interface.
-//
-// Class PrettyUnitTestResultPrinter is copyable.
-class PrettyUnitTestResultPrinter : public TestEventListener {
- public:
-  PrettyUnitTestResultPrinter() {}
-  static void PrintTestName(const char * test_case, const char * test) {
-    printf("%s.%s", test_case, test);
-  }
-
-  // The following methods override what's in the TestEventListener class.
-  virtual void OnTestProgramStart(const UnitTest& /*unit_test*/) {}
-  virtual void OnTestIterationStart(const UnitTest& unit_test, int iteration);
-  virtual void OnEnvironmentsSetUpStart(const UnitTest& unit_test);
-  virtual void OnEnvironmentsSetUpEnd(const UnitTest& /*unit_test*/) {}
-  virtual void OnTestCaseStart(const TestCase& test_case);
-  virtual void OnTestStart(const TestInfo& test_info);
-  virtual void OnTestPartResult(const TestPartResult& result);
-  virtual void OnTestEnd(const TestInfo& test_info);
-  virtual void OnTestCaseEnd(const TestCase& test_case);
-  virtual void OnEnvironmentsTearDownStart(const UnitTest& unit_test);
-  virtual void OnEnvironmentsTearDownEnd(const UnitTest& /*unit_test*/) {}
-  virtual void OnTestIterationEnd(const UnitTest& unit_test, int iteration);
-  virtual void OnTestProgramEnd(const UnitTest& /*unit_test*/) {}
-
- private:
-  static void PrintFailedTests(const UnitTest& unit_test);
-};
-
-  // Fired before each iteration of tests starts.
-void PrettyUnitTestResultPrinter::OnTestIterationStart(
-    const UnitTest& unit_test, int iteration) {
-  if (GTEST_FLAG(repeat) != 1)
-    printf("\nRepeating all tests (iteration %d) . . .\n\n", iteration + 1);
-
-  const char* const filter = GTEST_FLAG(filter).c_str();
-
-  // Prints the filter if it's not *.  This reminds the user that some
-  // tests may be skipped.
-  if (!String::CStringEquals(filter, kUniversalFilter)) {
-    ColoredPrintf(COLOR_YELLOW,
-                  "Note: %s filter = %s\n", GTEST_NAME_, filter);
-  }
-
-  if (internal::ShouldShard(kTestTotalShards, kTestShardIndex, false)) {
-    const Int32 shard_index = Int32FromEnvOrDie(kTestShardIndex, -1);
-    ColoredPrintf(COLOR_YELLOW,
-                  "Note: This is test shard %d of %s.\n",
-                  static_cast<int>(shard_index) + 1,
-                  internal::posix::GetEnv(kTestTotalShards));
-  }
-
-  if (GTEST_FLAG(shuffle)) {
-    ColoredPrintf(COLOR_YELLOW,
-                  "Note: Randomizing tests' orders with a seed of %d .\n",
-                  unit_test.random_seed());
-  }
-  ColoredPrintf(COLOR_GREEN,  "[==========] ");
-  printf("Running %s from %s.\n",
-         FormatTestCount(unit_test.test_to_run_count()).c_str(),
-         FormatTestCaseCount(unit_test.test_case_to_run_count()).c_str());
-  fflush(stdout);
-}
-
-void PrettyUnitTestResultPrinter::OnEnvironmentsSetUpStart(
-    const UnitTest& /*unit_test*/) {
-  ColoredPrintf(COLOR_GREEN,  "[----------] ");
-  printf("Global test environment set-up.\n");
-  fflush(stdout);
-}
-
-void PrettyUnitTestResultPrinter::OnTestCaseStart(const TestCase& test_case) {
-  const std::string counts =
-      FormatCountableNoun(test_case.test_to_run_count(), "test", "tests");
-  ColoredPrintf(COLOR_GREEN, "[----------] ");
-  printf("%s from %s", counts.c_str(), test_case.name());
-  if (test_case.type_param() == NULL) {
-    printf("\n");
-  } else {
-    printf(", where %s = %s\n", kTypeParamLabel, test_case.type_param());
-  }
-  fflush(stdout);
-}
-
-void PrettyUnitTestResultPrinter::OnTestStart(const TestInfo& test_info) {
-  ColoredPrintf(COLOR_GREEN,  "[ RUN      ] ");
-  PrintTestName(test_info.test_case_name(), test_info.name());
-  printf("\n");
-  fflush(stdout);
-}
-
-// Called after an assertion failure.
-void PrettyUnitTestResultPrinter::OnTestPartResult(
-    const TestPartResult& result) {
-  // If the test part succeeded, we don't need to do anything.
-  if (result.type() == TestPartResult::kSuccess)
-    return;
-
-  // Print failure message from the assertion (e.g. expected this and got that).
-  PrintTestPartResult(result);
-  fflush(stdout);
-}
-
-void PrettyUnitTestResultPrinter::OnTestEnd(const TestInfo& test_info) {
-  if (test_info.result()->Passed()) {
-    ColoredPrintf(COLOR_GREEN, "[       OK ] ");
-  } else {
-    ColoredPrintf(COLOR_RED, "[  FAILED  ] ");
-  }
-  PrintTestName(test_info.test_case_name(), test_info.name());
-  if (test_info.result()->Failed())
-    PrintFullTestCommentIfPresent(test_info);
-
-  if (GTEST_FLAG(print_time)) {
-    printf(" (%s ms)\n", internal::StreamableToString(
-           test_info.result()->elapsed_time()).c_str());
-  } else {
-    printf("\n");
-  }
-  fflush(stdout);
-}
-
-void PrettyUnitTestResultPrinter::OnTestCaseEnd(const TestCase& test_case) {
-  if (!GTEST_FLAG(print_time)) return;
-
-  const std::string counts =
-      FormatCountableNoun(test_case.test_to_run_count(), "test", "tests");
-  ColoredPrintf(COLOR_GREEN, "[----------] ");
-  printf("%s from %s (%s ms total)\n\n",
-         counts.c_str(), test_case.name(),
-         internal::StreamableToString(test_case.elapsed_time()).c_str());
-  fflush(stdout);
-}
-
-void PrettyUnitTestResultPrinter::OnEnvironmentsTearDownStart(
-    const UnitTest& /*unit_test*/) {
-  ColoredPrintf(COLOR_GREEN,  "[----------] ");
-  printf("Global test environment tear-down\n");
-  fflush(stdout);
-}
-
-// Internal helper for printing the list of failed tests.
-void PrettyUnitTestResultPrinter::PrintFailedTests(const UnitTest& unit_test) {
-  const int failed_test_count = unit_test.failed_test_count();
-  if (failed_test_count == 0) {
-    return;
-  }
-
-  for (int i = 0; i < unit_test.total_test_case_count(); ++i) {
-    const TestCase& test_case = *unit_test.GetTestCase(i);
-    if (!test_case.should_run() || (test_case.failed_test_count() == 0)) {
-      continue;
-    }
-    for (int j = 0; j < test_case.total_test_count(); ++j) {
-      const TestInfo& test_info = *test_case.GetTestInfo(j);
-      if (!test_info.should_run() || test_info.result()->Passed()) {
-        continue;
-      }
-      ColoredPrintf(COLOR_RED, "[  FAILED  ] ");
-      printf("%s.%s", test_case.name(), test_info.name());
-      PrintFullTestCommentIfPresent(test_info);
-      printf("\n");
-    }
-  }
-}
-
-void PrettyUnitTestResultPrinter::OnTestIterationEnd(const UnitTest& unit_test,
-                                                     int /*iteration*/) {
-  ColoredPrintf(COLOR_GREEN,  "[==========] ");
-  printf("%s from %s ran.",
-         FormatTestCount(unit_test.test_to_run_count()).c_str(),
-         FormatTestCaseCount(unit_test.test_case_to_run_count()).c_str());
-  if (GTEST_FLAG(print_time)) {
-    printf(" (%s ms total)",
-           internal::StreamableToString(unit_test.elapsed_time()).c_str());
-  }
-  printf("\n");
-  ColoredPrintf(COLOR_GREEN,  "[  PASSED  ] ");
-  printf("%s.\n", FormatTestCount(unit_test.successful_test_count()).c_str());
-
-  int num_failures = unit_test.failed_test_count();
-  if (!unit_test.Passed()) {
-    const int failed_test_count = unit_test.failed_test_count();
-    ColoredPrintf(COLOR_RED,  "[  FAILED  ] ");
-    printf("%s, listed below:\n", FormatTestCount(failed_test_count).c_str());
-    PrintFailedTests(unit_test);
-    printf("\n%2d FAILED %s\n", num_failures,
-                        num_failures == 1 ? "TEST" : "TESTS");
-  }
-
-  int num_disabled = unit_test.reportable_disabled_test_count();
-  if (num_disabled && !GTEST_FLAG(also_run_disabled_tests)) {
-    if (!num_failures) {
-      printf("\n");  // Add a spacer if no FAILURE banner is displayed.
-    }
-    ColoredPrintf(COLOR_YELLOW,
-                  "  YOU HAVE %d DISABLED %s\n\n",
-                  num_disabled,
-                  num_disabled == 1 ? "TEST" : "TESTS");
-  }
-  // Ensure that Google Test output is printed before, e.g., heapchecker output.
-  fflush(stdout);
-}
-
-// End PrettyUnitTestResultPrinter
-
-// class TestEventRepeater
-//
-// This class forwards events to other event listeners.
-class TestEventRepeater : public TestEventListener {
- public:
-  TestEventRepeater() : forwarding_enabled_(true) {}
-  virtual ~TestEventRepeater();
-  void Append(TestEventListener *listener);
-  TestEventListener* Release(TestEventListener* listener);
-
-  // Controls whether events will be forwarded to listeners_. Set to false
-  // in death test child processes.
-  bool forwarding_enabled() const { return forwarding_enabled_; }
-  void set_forwarding_enabled(bool enable) { forwarding_enabled_ = enable; }
-
-  virtual void OnTestProgramStart(const UnitTest& unit_test);
-  virtual void OnTestIterationStart(const UnitTest& unit_test, int iteration);
-  virtual void OnEnvironmentsSetUpStart(const UnitTest& unit_test);
-  virtual void OnEnvironmentsSetUpEnd(const UnitTest& unit_test);
-  virtual void OnTestCaseStart(const TestCase& test_case);
-  virtual void OnTestStart(const TestInfo& test_info);
-  virtual void OnTestPartResult(const TestPartResult& result);
-  virtual void OnTestEnd(const TestInfo& test_info);
-  virtual void OnTestCaseEnd(const TestCase& test_case);
-  virtual void OnEnvironmentsTearDownStart(const UnitTest& unit_test);
-  virtual void OnEnvironmentsTearDownEnd(const UnitTest& unit_test);
-  virtual void OnTestIterationEnd(const UnitTest& unit_test, int iteration);
-  virtual void OnTestProgramEnd(const UnitTest& unit_test);
-
- private:
-  // Controls whether events will be forwarded to listeners_. Set to false
-  // in death test child processes.
-  bool forwarding_enabled_;
-  // The list of listeners that receive events.
-  std::vector<TestEventListener*> listeners_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(TestEventRepeater);
-};
-
-TestEventRepeater::~TestEventRepeater() {
-  ForEach(listeners_, Delete<TestEventListener>);
-}
-
-void TestEventRepeater::Append(TestEventListener *listener) {
-  listeners_.push_back(listener);
-}
-
-// TODO(vladl@google.com): Factor the search functionality into Vector::Find.
-TestEventListener* TestEventRepeater::Release(TestEventListener *listener) {
-  for (size_t i = 0; i < listeners_.size(); ++i) {
-    if (listeners_[i] == listener) {
-      listeners_.erase(listeners_.begin() + i);
-      return listener;
-    }
-  }
-
-  return NULL;
-}
-
-// Since most methods are very similar, use macros to reduce boilerplate.
-// This defines a member that forwards the call to all listeners.
-#define GTEST_REPEATER_METHOD_(Name, Type) \
-void TestEventRepeater::Name(const Type& parameter) { \
-  if (forwarding_enabled_) { \
-    for (size_t i = 0; i < listeners_.size(); i++) { \
-      listeners_[i]->Name(parameter); \
-    } \
-  } \
-}
-// This defines a member that forwards the call to all listeners in reverse
-// order.
-#define GTEST_REVERSE_REPEATER_METHOD_(Name, Type) \
-void TestEventRepeater::Name(const Type& parameter) { \
-  if (forwarding_enabled_) { \
-    for (int i = static_cast<int>(listeners_.size()) - 1; i >= 0; i--) { \
-      listeners_[i]->Name(parameter); \
-    } \
-  } \
-}
-
-GTEST_REPEATER_METHOD_(OnTestProgramStart, UnitTest)
-GTEST_REPEATER_METHOD_(OnEnvironmentsSetUpStart, UnitTest)
-GTEST_REPEATER_METHOD_(OnTestCaseStart, TestCase)
-GTEST_REPEATER_METHOD_(OnTestStart, TestInfo)
-GTEST_REPEATER_METHOD_(OnTestPartResult, TestPartResult)
-GTEST_REPEATER_METHOD_(OnEnvironmentsTearDownStart, UnitTest)
-GTEST_REVERSE_REPEATER_METHOD_(OnEnvironmentsSetUpEnd, UnitTest)
-GTEST_REVERSE_REPEATER_METHOD_(OnEnvironmentsTearDownEnd, UnitTest)
-GTEST_REVERSE_REPEATER_METHOD_(OnTestEnd, TestInfo)
-GTEST_REVERSE_REPEATER_METHOD_(OnTestCaseEnd, TestCase)
-GTEST_REVERSE_REPEATER_METHOD_(OnTestProgramEnd, UnitTest)
-
-#undef GTEST_REPEATER_METHOD_
-#undef GTEST_REVERSE_REPEATER_METHOD_
-
-void TestEventRepeater::OnTestIterationStart(const UnitTest& unit_test,
-                                             int iteration) {
-  if (forwarding_enabled_) {
-    for (size_t i = 0; i < listeners_.size(); i++) {
-      listeners_[i]->OnTestIterationStart(unit_test, iteration);
-    }
-  }
-}
-
-void TestEventRepeater::OnTestIterationEnd(const UnitTest& unit_test,
-                                           int iteration) {
-  if (forwarding_enabled_) {
-    for (int i = static_cast<int>(listeners_.size()) - 1; i >= 0; i--) {
-      listeners_[i]->OnTestIterationEnd(unit_test, iteration);
-    }
-  }
-}
-
-// End TestEventRepeater
-
-// This class generates an XML output file.
-class XmlUnitTestResultPrinter : public EmptyTestEventListener {
- public:
-  explicit XmlUnitTestResultPrinter(const char* output_file);
-
-  virtual void OnTestIterationEnd(const UnitTest& unit_test, int iteration);
-
- private:
-  // Is c a whitespace character that is normalized to a space character
-  // when it appears in an XML attribute value?
-  static bool IsNormalizableWhitespace(char c) {
-    return c == 0x9 || c == 0xA || c == 0xD;
-  }
-
-  // May c appear in a well-formed XML document?
-  static bool IsValidXmlCharacter(char c) {
-    return IsNormalizableWhitespace(c) || c >= 0x20;
-  }
-
-  // Returns an XML-escaped copy of the input string str.  If
-  // is_attribute is true, the text is meant to appear as an attribute
-  // value, and normalizable whitespace is preserved by replacing it
-  // with character references.
-  static std::string EscapeXml(const std::string& str, bool is_attribute);
-
-  // Returns the given string with all characters invalid in XML removed.
-  static std::string RemoveInvalidXmlCharacters(const std::string& str);
-
-  // Convenience wrapper around EscapeXml when str is an attribute value.
-  static std::string EscapeXmlAttribute(const std::string& str) {
-    return EscapeXml(str, true);
-  }
-
-  // Convenience wrapper around EscapeXml when str is not an attribute value.
-  static std::string EscapeXmlText(const char* str) {
-    return EscapeXml(str, false);
-  }
-
-  // Verifies that the given attribute belongs to the given element and
-  // streams the attribute as XML.
-  static void OutputXmlAttribute(std::ostream* stream,
-                                 const std::string& element_name,
-                                 const std::string& name,
-                                 const std::string& value);
-
-  // Streams an XML CDATA section, escaping invalid CDATA sequences as needed.
-  static void OutputXmlCDataSection(::std::ostream* stream, const char* data);
-
-  // Streams an XML representation of a TestInfo object.
-  static void OutputXmlTestInfo(::std::ostream* stream,
-                                const char* test_case_name,
-                                const TestInfo& test_info);
-
-  // Prints an XML representation of a TestCase object
-  static void PrintXmlTestCase(::std::ostream* stream,
-                               const TestCase& test_case);
-
-  // Prints an XML summary of unit_test to output stream out.
-  static void PrintXmlUnitTest(::std::ostream* stream,
-                               const UnitTest& unit_test);
-
-  // Produces a string representing the test properties in a result as space
-  // delimited XML attributes based on the property key="value" pairs.
-  // When the std::string is not empty, it includes a space at the beginning,
-  // to delimit this attribute from prior attributes.
-  static std::string TestPropertiesAsXmlAttributes(const TestResult& result);
-
-  // Streams an XML representation of the test properties of a TestResult
-  // object.
-  static void OutputXmlTestProperties(std::ostream* stream,
-                                      const TestResult& result);
-
-  // The output file.
-  const std::string output_file_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(XmlUnitTestResultPrinter);
-};
-
-// Creates a new XmlUnitTestResultPrinter.
-XmlUnitTestResultPrinter::XmlUnitTestResultPrinter(const char* output_file)
-    : output_file_(output_file) {
-  if (output_file_.c_str() == NULL || output_file_.empty()) {
-    GTEST_LOG_(FATAL) << "XML output file may not be null";
-  }
-}
-
-// Called after the unit test ends.
-void XmlUnitTestResultPrinter::OnTestIterationEnd(const UnitTest& unit_test,
-                                                  int /*iteration*/) {
-  FILE* xmlout = NULL;
-  FilePath output_file(output_file_);
-  FilePath output_dir(output_file.RemoveFileName());
-
-  if (output_dir.CreateDirectoriesRecursively()) {
-    xmlout = posix::FOpen(output_file_.c_str(), "w");
-  }
-  if (xmlout == NULL) {
-    // TODO(wan): report the reason of the failure.
-    //
-    // We don't do it for now as:
-    //
-    //   1. There is no urgent need for it.
-    //   2. It's a bit involved to make the errno variable thread-safe on
-    //      all three operating systems (Linux, Windows, and Mac OS).
-    //   3. To interpret the meaning of errno in a thread-safe way,
-    //      we need the strerror_r() function, which is not available on
-    //      Windows.
-
-    GTEST_LOG_(FATAL) << "Unable to open file \"" << output_file_ << "\"";
-  }
-  std::stringstream stream;
-  PrintXmlUnitTest(&stream, unit_test);
-  fprintf(xmlout, "%s", StringStreamToString(&stream).c_str());
-  fclose(xmlout);
-}
-
-// Returns an XML-escaped copy of the input string str.  If is_attribute
-// is true, the text is meant to appear as an attribute value, and
-// normalizable whitespace is preserved by replacing it with character
-// references.
-//
-// Invalid XML characters in str, if any, are stripped from the output.
-// It is expected that most, if not all, of the text processed by this
-// module will consist of ordinary English text.
-// If this module is ever modified to produce version 1.1 XML output,
-// most invalid characters can be retained using character references.
-// TODO(wan): It might be nice to have a minimally invasive, human-readable
-// escaping scheme for invalid characters, rather than dropping them.
-std::string XmlUnitTestResultPrinter::EscapeXml(
-    const std::string& str, bool is_attribute) {
-  Message m;
-
-  for (size_t i = 0; i < str.size(); ++i) {
-    const char ch = str[i];
-    switch (ch) {
-      case '<':
-        m << "&lt;";
-        break;
-      case '>':
-        m << "&gt;";
-        break;
-      case '&':
-        m << "&amp;";
-        break;
-      case '\'':
-        if (is_attribute)
-          m << "&apos;";
-        else
-          m << '\'';
-        break;
-      case '"':
-        if (is_attribute)
-          m << "&quot;";
-        else
-          m << '"';
-        break;
-      default:
-        if (IsValidXmlCharacter(ch)) {
-          if (is_attribute && IsNormalizableWhitespace(ch))
-            m << "&#x" << String::FormatByte(static_cast<unsigned char>(ch))
-              << ";";
-          else
-            m << ch;
-        }
-        break;
-    }
-  }
-
-  return m.GetString();
-}
-
-// Returns the given string with all characters invalid in XML removed.
-// Currently invalid characters are dropped from the string. An
-// alternative is to replace them with certain characters such as . or ?.
-std::string XmlUnitTestResultPrinter::RemoveInvalidXmlCharacters(
-    const std::string& str) {
-  std::string output;
-  output.reserve(str.size());
-  for (std::string::const_iterator it = str.begin(); it != str.end(); ++it)
-    if (IsValidXmlCharacter(*it))
-      output.push_back(*it);
-
-  return output;
-}
-
-// The following routines generate an XML representation of a UnitTest
-// object.
-//
-// This is how Google Test concepts map to the DTD:
-//
-// <testsuites name="AllTests">        <-- corresponds to a UnitTest object
-//   <testsuite name="testcase-name">  <-- corresponds to a TestCase object
-//     <testcase name="test-name">     <-- corresponds to a TestInfo object
-//       <failure message="...">...</failure>
-//       <failure message="...">...</failure>
-//       <failure message="...">...</failure>
-//                                     <-- individual assertion failures
-//     </testcase>
-//   </testsuite>
-// </testsuites>
-
-// Formats the given time in milliseconds as seconds.
-std::string FormatTimeInMillisAsSeconds(TimeInMillis ms) {
-  ::std::stringstream ss;
-  ss << (static_cast<double>(ms) * 1e-3);
-  return ss.str();
-}
-
-static bool PortableLocaltime(time_t seconds, struct tm* out) {
-#if defined(_MSC_VER)
-  return localtime_s(out, &seconds) == 0;
-#elif defined(__MINGW32__) || defined(__MINGW64__)
-  // MINGW <time.h> provides neither localtime_r nor localtime_s, but uses
-  // Windows' localtime(), which has a thread-local tm buffer.
-  struct tm* tm_ptr = localtime(&seconds);  // NOLINT
-  if (tm_ptr == NULL)
-    return false;
-  *out = *tm_ptr;
-  return true;
-#else
-  return localtime_r(&seconds, out) != NULL;
-#endif
-}
-
-// Converts the given epoch time in milliseconds to a date string in the ISO
-// 8601 format, without the timezone information.
-std::string FormatEpochTimeInMillisAsIso8601(TimeInMillis ms) {
-  struct tm time_struct;
-  if (!PortableLocaltime(static_cast<time_t>(ms / 1000), &time_struct))
-    return "";
-  // YYYY-MM-DDThh:mm:ss
-  return StreamableToString(time_struct.tm_year + 1900) + "-" +
-      String::FormatIntWidth2(time_struct.tm_mon + 1) + "-" +
-      String::FormatIntWidth2(time_struct.tm_mday) + "T" +
-      String::FormatIntWidth2(time_struct.tm_hour) + ":" +
-      String::FormatIntWidth2(time_struct.tm_min) + ":" +
-      String::FormatIntWidth2(time_struct.tm_sec);
-}
-
-// Streams an XML CDATA section, escaping invalid CDATA sequences as needed.
-void XmlUnitTestResultPrinter::OutputXmlCDataSection(::std::ostream* stream,
-                                                     const char* data) {
-  const char* segment = data;
-  *stream << "<![CDATA[";
-  for (;;) {
-    const char* const next_segment = strstr(segment, "]]>");
-    if (next_segment != NULL) {
-      stream->write(
-          segment, static_cast<std::streamsize>(next_segment - segment));
-      *stream << "]]>]]&gt;<![CDATA[";
-      segment = next_segment + strlen("]]>");
-    } else {
-      *stream << segment;
-      break;
-    }
-  }
-  *stream << "]]>";
-}
-
-void XmlUnitTestResultPrinter::OutputXmlAttribute(
-    std::ostream* stream,
-    const std::string& element_name,
-    const std::string& name,
-    const std::string& value) {
-  const std::vector<std::string>& allowed_names =
-      GetReservedAttributesForElement(element_name);
-
-  GTEST_CHECK_(std::find(allowed_names.begin(), allowed_names.end(), name) !=
-                   allowed_names.end())
-      << "Attribute " << name << " is not allowed for element <" << element_name
-      << ">.";
-
-  *stream << " " << name << "=\"" << EscapeXmlAttribute(value) << "\"";
-}
-
-// Prints an XML representation of a TestInfo object.
-// TODO(wan): There is also value in printing properties with the plain printer.
-void XmlUnitTestResultPrinter::OutputXmlTestInfo(::std::ostream* stream,
-                                                 const char* test_case_name,
-                                                 const TestInfo& test_info) {
-  const TestResult& result = *test_info.result();
-  const std::string kTestcase = "testcase";
-
-  if (test_info.is_in_another_shard()) {
-    return;
-  }
-
-  *stream << "    <testcase";
-  OutputXmlAttribute(stream, kTestcase, "name", test_info.name());
-
-  if (test_info.value_param() != NULL) {
-    OutputXmlAttribute(stream, kTestcase, "value_param",
-                       test_info.value_param());
-  }
-  if (test_info.type_param() != NULL) {
-    OutputXmlAttribute(stream, kTestcase, "type_param", test_info.type_param());
-  }
-
-  OutputXmlAttribute(stream, kTestcase, "status",
-                     test_info.should_run() ? "run" : "notrun");
-  OutputXmlAttribute(stream, kTestcase, "time",
-                     FormatTimeInMillisAsSeconds(result.elapsed_time()));
-  OutputXmlAttribute(stream, kTestcase, "classname", test_case_name);
-
-  int failures = 0;
-  for (int i = 0; i < result.total_part_count(); ++i) {
-    const TestPartResult& part = result.GetTestPartResult(i);
-    if (part.failed()) {
-      if (++failures == 1) {
-        *stream << ">\n";
-      }
-      const std::string location =
-          internal::FormatCompilerIndependentFileLocation(part.file_name(),
-                                                          part.line_number());
-      const std::string summary = location + "\n" + part.summary();
-      *stream << "      <failure message=\""
-              << EscapeXmlAttribute(summary.c_str())
-              << "\" type=\"\">";
-      const std::string detail = location + "\n" + part.message();
-      OutputXmlCDataSection(stream, RemoveInvalidXmlCharacters(detail).c_str());
-      *stream << "</failure>\n";
-    }
-  }
-
-  if (failures == 0 && result.test_property_count() == 0) {
-    *stream << " />\n";
-  } else {
-    if (failures == 0) {
-      *stream << ">\n";
-    }
-    OutputXmlTestProperties(stream, result);
-    *stream << "    </testcase>\n";
-  }
-}
-
-// Prints an XML representation of a TestCase object
-void XmlUnitTestResultPrinter::PrintXmlTestCase(std::ostream* stream,
-                                                const TestCase& test_case) {
-  const std::string kTestsuite = "testsuite";
-  *stream << "  <" << kTestsuite;
-  OutputXmlAttribute(stream, kTestsuite, "name", test_case.name());
-  OutputXmlAttribute(stream, kTestsuite, "tests",
-                     StreamableToString(test_case.reportable_test_count()));
-  OutputXmlAttribute(stream, kTestsuite, "failures",
-                     StreamableToString(test_case.failed_test_count()));
-  OutputXmlAttribute(
-      stream, kTestsuite, "disabled",
-      StreamableToString(test_case.reportable_disabled_test_count()));
-  OutputXmlAttribute(stream, kTestsuite, "errors", "0");
-  OutputXmlAttribute(stream, kTestsuite, "time",
-                     FormatTimeInMillisAsSeconds(test_case.elapsed_time()));
-  *stream << TestPropertiesAsXmlAttributes(test_case.ad_hoc_test_result())
-          << ">\n";
-
-  for (int i = 0; i < test_case.total_test_count(); ++i) {
-    if (test_case.GetTestInfo(i)->is_reportable())
-      OutputXmlTestInfo(stream, test_case.name(), *test_case.GetTestInfo(i));
-  }
-  *stream << "  </" << kTestsuite << ">\n";
-}
-
-// Prints an XML summary of unit_test to output stream out.
-void XmlUnitTestResultPrinter::PrintXmlUnitTest(std::ostream* stream,
-                                                const UnitTest& unit_test) {
-  const std::string kTestsuites = "testsuites";
-
-  *stream << "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n";
-  *stream << "<" << kTestsuites;
-
-  OutputXmlAttribute(stream, kTestsuites, "tests",
-                     StreamableToString(unit_test.reportable_test_count()));
-  OutputXmlAttribute(stream, kTestsuites, "failures",
-                     StreamableToString(unit_test.failed_test_count()));
-  OutputXmlAttribute(
-      stream, kTestsuites, "disabled",
-      StreamableToString(unit_test.reportable_disabled_test_count()));
-  OutputXmlAttribute(stream, kTestsuites, "errors", "0");
-  OutputXmlAttribute(
-      stream, kTestsuites, "timestamp",
-      FormatEpochTimeInMillisAsIso8601(unit_test.start_timestamp()));
-  OutputXmlAttribute(stream, kTestsuites, "time",
-                     FormatTimeInMillisAsSeconds(unit_test.elapsed_time()));
-
-  if (GTEST_FLAG(shuffle)) {
-    OutputXmlAttribute(stream, kTestsuites, "random_seed",
-                       StreamableToString(unit_test.random_seed()));
-  }
-  *stream << TestPropertiesAsXmlAttributes(unit_test.ad_hoc_test_result());
-
-  OutputXmlAttribute(stream, kTestsuites, "name", "AllTests");
-  *stream << ">\n";
-
-  for (int i = 0; i < unit_test.total_test_case_count(); ++i) {
-    if (unit_test.GetTestCase(i)->reportable_test_count() > 0)
-      PrintXmlTestCase(stream, *unit_test.GetTestCase(i));
-  }
-  *stream << "</" << kTestsuites << ">\n";
-}
-
-// Produces a string representing the test properties in a result as space
-// delimited XML attributes based on the property key="value" pairs.
-std::string XmlUnitTestResultPrinter::TestPropertiesAsXmlAttributes(
-    const TestResult& result) {
-  Message attributes;
-  for (int i = 0; i < result.test_property_count(); ++i) {
-    const TestProperty& property = result.GetTestProperty(i);
-    attributes << " " << property.key() << "="
-        << "\"" << EscapeXmlAttribute(property.value()) << "\"";
-  }
-  return attributes.GetString();
-}
-
-void XmlUnitTestResultPrinter::OutputXmlTestProperties(
-    std::ostream* stream, const TestResult& result) {
-  const std::string kProperties = "properties";
-  const std::string kProperty = "property";
-
-  if (result.test_property_count() <= 0) {
-    return;
-  }
-
-  *stream << "<" << kProperties << ">\n";
-  for (int i = 0; i < result.test_property_count(); ++i) {
-    const TestProperty& property = result.GetTestProperty(i);
-    *stream << "<" << kProperty;
-    *stream << " name=\"" << EscapeXmlAttribute(property.key()) << "\"";
-    *stream << " value=\"" << EscapeXmlAttribute(property.value()) << "\"";
-    *stream << "/>\n";
-  }
-  *stream << "</" << kProperties << ">\n";
-}
-
-// End XmlUnitTestResultPrinter
-
-
-// This class generates an JSON output file.
-class JsonUnitTestResultPrinter : public EmptyTestEventListener {
- public:
-  explicit JsonUnitTestResultPrinter(const char* output_file);
-
-  virtual void OnTestIterationEnd(const UnitTest& unit_test, int iteration);
-
- private:
-  // Returns an JSON-escaped copy of the input string str.
-  static std::string EscapeJson(const std::string& str);
-
-  //// Verifies that the given attribute belongs to the given element and
-  //// streams the attribute as JSON.
-  static void OutputJsonKey(std::ostream* stream,
-                            const std::string& element_name,
-                            const std::string& name,
-                            const std::string& value,
-                            const std::string& indent,
-                            bool comma = true);
-  static void OutputJsonKey(std::ostream* stream,
-                            const std::string& element_name,
-                            const std::string& name,
-                            int value,
-                            const std::string& indent,
-                            bool comma = true);
-
-  // Streams a JSON representation of a TestInfo object.
-  static void OutputJsonTestInfo(::std::ostream* stream,
-                                 const char* test_case_name,
-                                 const TestInfo& test_info);
-
-  // Prints a JSON representation of a TestCase object
-  static void PrintJsonTestCase(::std::ostream* stream,
-                                const TestCase& test_case);
-
-  // Prints a JSON summary of unit_test to output stream out.
-  static void PrintJsonUnitTest(::std::ostream* stream,
-                                const UnitTest& unit_test);
-
-  // Produces a string representing the test properties in a result as
-  // a JSON dictionary.
-  static std::string TestPropertiesAsJson(const TestResult& result,
-                                          const std::string& indent);
-
-  // The output file.
-  const std::string output_file_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(JsonUnitTestResultPrinter);
-};
-
-// Creates a new JsonUnitTestResultPrinter.
-JsonUnitTestResultPrinter::JsonUnitTestResultPrinter(const char* output_file)
-    : output_file_(output_file) {
-  if (output_file_.empty()) {
-    GTEST_LOG_(FATAL) << "JSON output file may not be null";
-  }
-}
-
-void JsonUnitTestResultPrinter::OnTestIterationEnd(const UnitTest& unit_test,
-                                                  int /*iteration*/) {
-  FILE* jsonout = NULL;
-  FilePath output_file(output_file_);
-  FilePath output_dir(output_file.RemoveFileName());
-
-  if (output_dir.CreateDirectoriesRecursively()) {
-    jsonout = posix::FOpen(output_file_.c_str(), "w");
-  }
-  if (jsonout == NULL) {
-    // TODO(phosek): report the reason of the failure.
-    //
-    // We don't do it for now as:
-    //
-    //   1. There is no urgent need for it.
-    //   2. It's a bit involved to make the errno variable thread-safe on
-    //      all three operating systems (Linux, Windows, and Mac OS).
-    //   3. To interpret the meaning of errno in a thread-safe way,
-    //      we need the strerror_r() function, which is not available on
-    //      Windows.
-    GTEST_LOG_(FATAL) << "Unable to open file \""
-                      << output_file_ << "\"";
-  }
-  std::stringstream stream;
-  PrintJsonUnitTest(&stream, unit_test);
-  fprintf(jsonout, "%s", StringStreamToString(&stream).c_str());
-  fclose(jsonout);
-}
-
-// Returns an JSON-escaped copy of the input string str.
-std::string JsonUnitTestResultPrinter::EscapeJson(const std::string& str) {
-  Message m;
-
-  for (size_t i = 0; i < str.size(); ++i) {
-    const char ch = str[i];
-    switch (ch) {
-      case '\\':
-      case '"':
-      case '/':
-        m << '\\' << ch;
-        break;
-      case '\b':
-        m << "\\b";
-        break;
-      case '\t':
-        m << "\\t";
-        break;
-      case '\n':
-        m << "\\n";
-        break;
-      case '\f':
-        m << "\\f";
-        break;
-      case '\r':
-        m << "\\r";
-        break;
-      default:
-        if (ch < ' ') {
-          m << "\\u00" << String::FormatByte(static_cast<unsigned char>(ch));
-        } else {
-          m << ch;
-        }
-        break;
-    }
-  }
-
-  return m.GetString();
-}
-
-// The following routines generate an JSON representation of a UnitTest
-// object.
-
-// Formats the given time in milliseconds as seconds.
-static std::string FormatTimeInMillisAsDuration(TimeInMillis ms) {
-  ::std::stringstream ss;
-  ss << (static_cast<double>(ms) * 1e-3) << "s";
-  return ss.str();
-}
-
-// Converts the given epoch time in milliseconds to a date string in the
-// RFC3339 format, without the timezone information.
-static std::string FormatEpochTimeInMillisAsRFC3339(TimeInMillis ms) {
-  struct tm time_struct;
-  if (!PortableLocaltime(static_cast<time_t>(ms / 1000), &time_struct))
-    return "";
-  // YYYY-MM-DDThh:mm:ss
-  return StreamableToString(time_struct.tm_year + 1900) + "-" +
-      String::FormatIntWidth2(time_struct.tm_mon + 1) + "-" +
-      String::FormatIntWidth2(time_struct.tm_mday) + "T" +
-      String::FormatIntWidth2(time_struct.tm_hour) + ":" +
-      String::FormatIntWidth2(time_struct.tm_min) + ":" +
-      String::FormatIntWidth2(time_struct.tm_sec) + "Z";
-}
-
-static inline std::string Indent(int width) {
-  return std::string(width, ' ');
-}
-
-void JsonUnitTestResultPrinter::OutputJsonKey(
-    std::ostream* stream,
-    const std::string& element_name,
-    const std::string& name,
-    const std::string& value,
-    const std::string& indent,
-    bool comma) {
-  const std::vector<std::string>& allowed_names =
-      GetReservedAttributesForElement(element_name);
-
-  GTEST_CHECK_(std::find(allowed_names.begin(), allowed_names.end(), name) !=
-                   allowed_names.end())
-      << "Key \"" << name << "\" is not allowed for value \"" << element_name
-      << "\".";
-
-  *stream << indent << "\"" << name << "\": \"" << EscapeJson(value) << "\"";
-  if (comma)
-    *stream << ",\n";
-}
-
-void JsonUnitTestResultPrinter::OutputJsonKey(
-    std::ostream* stream,
-    const std::string& element_name,
-    const std::string& name,
-    int value,
-    const std::string& indent,
-    bool comma) {
-  const std::vector<std::string>& allowed_names =
-      GetReservedAttributesForElement(element_name);
-
-  GTEST_CHECK_(std::find(allowed_names.begin(), allowed_names.end(), name) !=
-                   allowed_names.end())
-      << "Key \"" << name << "\" is not allowed for value \"" << element_name
-      << "\".";
-
-  *stream << indent << "\"" << name << "\": " << StreamableToString(value);
-  if (comma)
-    *stream << ",\n";
-}
-
-// Prints a JSON representation of a TestInfo object.
-void JsonUnitTestResultPrinter::OutputJsonTestInfo(::std::ostream* stream,
-                                                   const char* test_case_name,
-                                                   const TestInfo& test_info) {
-  const TestResult& result = *test_info.result();
-  const std::string kTestcase = "testcase";
-  const std::string kIndent = Indent(10);
-
-  *stream << Indent(8) << "{\n";
-  OutputJsonKey(stream, kTestcase, "name", test_info.name(), kIndent);
-
-  if (test_info.value_param() != NULL) {
-    OutputJsonKey(stream, kTestcase, "value_param",
-                  test_info.value_param(), kIndent);
-  }
-  if (test_info.type_param() != NULL) {
-    OutputJsonKey(stream, kTestcase, "type_param", test_info.type_param(),
-                  kIndent);
-  }
-
-  OutputJsonKey(stream, kTestcase, "status",
-                test_info.should_run() ? "RUN" : "NOTRUN", kIndent);
-  OutputJsonKey(stream, kTestcase, "time",
-                FormatTimeInMillisAsDuration(result.elapsed_time()), kIndent);
-  OutputJsonKey(stream, kTestcase, "classname", test_case_name, kIndent, false);
-  *stream << TestPropertiesAsJson(result, kIndent);
-
-  int failures = 0;
-  for (int i = 0; i < result.total_part_count(); ++i) {
-    const TestPartResult& part = result.GetTestPartResult(i);
-    if (part.failed()) {
-      *stream << ",\n";
-      if (++failures == 1) {
-        *stream << kIndent << "\"" << "failures" << "\": [\n";
-      }
-      const std::string location =
-          internal::FormatCompilerIndependentFileLocation(part.file_name(),
-                                                          part.line_number());
-      const std::string message = EscapeJson(location + "\n" + part.message());
-      *stream << kIndent << "  {\n"
-              << kIndent << "    \"failure\": \"" << message << "\",\n"
-              << kIndent << "    \"type\": \"\"\n"
-              << kIndent << "  }";
-    }
-  }
-
-  if (failures > 0)
-    *stream << "\n" << kIndent << "]";
-  *stream << "\n" << Indent(8) << "}";
-}
-
-// Prints an JSON representation of a TestCase object
-void JsonUnitTestResultPrinter::PrintJsonTestCase(std::ostream* stream,
-                                                  const TestCase& test_case) {
-  const std::string kTestsuite = "testsuite";
-  const std::string kIndent = Indent(6);
-
-  *stream << Indent(4) << "{\n";
-  OutputJsonKey(stream, kTestsuite, "name", test_case.name(), kIndent);
-  OutputJsonKey(stream, kTestsuite, "tests", test_case.reportable_test_count(),
-                kIndent);
-  OutputJsonKey(stream, kTestsuite, "failures", test_case.failed_test_count(),
-                kIndent);
-  OutputJsonKey(stream, kTestsuite, "disabled",
-                test_case.reportable_disabled_test_count(), kIndent);
-  OutputJsonKey(stream, kTestsuite, "errors", 0, kIndent);
-  OutputJsonKey(stream, kTestsuite, "time",
-                FormatTimeInMillisAsDuration(test_case.elapsed_time()), kIndent,
-                false);
-  *stream << TestPropertiesAsJson(test_case.ad_hoc_test_result(), kIndent)
-          << ",\n";
-
-  *stream << kIndent << "\"" << kTestsuite << "\": [\n";
-
-  bool comma = false;
-  for (int i = 0; i < test_case.total_test_count(); ++i) {
-    if (test_case.GetTestInfo(i)->is_reportable()) {
-      if (comma) {
-        *stream << ",\n";
-      } else {
-        comma = true;
-      }
-      OutputJsonTestInfo(stream, test_case.name(), *test_case.GetTestInfo(i));
-    }
-  }
-  *stream << "\n" << kIndent << "]\n" << Indent(4) << "}";
-}
-
-// Prints a JSON summary of unit_test to output stream out.
-void JsonUnitTestResultPrinter::PrintJsonUnitTest(std::ostream* stream,
-                                                  const UnitTest& unit_test) {
-  const std::string kTestsuites = "testsuites";
-  const std::string kIndent = Indent(2);
-  *stream << "{\n";
-
-  OutputJsonKey(stream, kTestsuites, "tests", unit_test.reportable_test_count(),
-                kIndent);
-  OutputJsonKey(stream, kTestsuites, "failures", unit_test.failed_test_count(),
-                kIndent);
-  OutputJsonKey(stream, kTestsuites, "disabled",
-                unit_test.reportable_disabled_test_count(), kIndent);
-  OutputJsonKey(stream, kTestsuites, "errors", 0, kIndent);
-  if (GTEST_FLAG(shuffle)) {
-    OutputJsonKey(stream, kTestsuites, "random_seed", unit_test.random_seed(),
-                  kIndent);
-  }
-  OutputJsonKey(stream, kTestsuites, "timestamp",
-                FormatEpochTimeInMillisAsRFC3339(unit_test.start_timestamp()),
-                kIndent);
-  OutputJsonKey(stream, kTestsuites, "time",
-                FormatTimeInMillisAsDuration(unit_test.elapsed_time()), kIndent,
-                false);
-
-  *stream << TestPropertiesAsJson(unit_test.ad_hoc_test_result(), kIndent)
-          << ",\n";
-
-  OutputJsonKey(stream, kTestsuites, "name", "AllTests", kIndent);
-  *stream << kIndent << "\"" << kTestsuites << "\": [\n";
-
-  bool comma = false;
-  for (int i = 0; i < unit_test.total_test_case_count(); ++i) {
-    if (unit_test.GetTestCase(i)->reportable_test_count() > 0) {
-      if (comma) {
-        *stream << ",\n";
-      } else {
-        comma = true;
-      }
-      PrintJsonTestCase(stream, *unit_test.GetTestCase(i));
-    }
-  }
-
-  *stream << "\n" << kIndent << "]\n" << "}\n";
-}
-
-// Produces a string representing the test properties in a result as
-// a JSON dictionary.
-std::string JsonUnitTestResultPrinter::TestPropertiesAsJson(
-    const TestResult& result, const std::string& indent) {
-  Message attributes;
-  for (int i = 0; i < result.test_property_count(); ++i) {
-    const TestProperty& property = result.GetTestProperty(i);
-    attributes << ",\n" << indent << "\"" << property.key() << "\": "
-               << "\"" << EscapeJson(property.value()) << "\"";
-  }
-  return attributes.GetString();
-}
-
-// End JsonUnitTestResultPrinter
-
-#if GTEST_CAN_STREAM_RESULTS_
-
-// Checks if str contains '=', '&', '%' or '\n' characters. If yes,
-// replaces them by "%xx" where xx is their hexadecimal value. For
-// example, replaces "=" with "%3D".  This algorithm is O(strlen(str))
-// in both time and space -- important as the input str may contain an
-// arbitrarily long test failure message and stack trace.
-std::string StreamingListener::UrlEncode(const char* str) {
-  std::string result;
-  result.reserve(strlen(str) + 1);
-  for (char ch = *str; ch != '\0'; ch = *++str) {
-    switch (ch) {
-      case '%':
-      case '=':
-      case '&':
-      case '\n':
-        result.append("%" + String::FormatByte(static_cast<unsigned char>(ch)));
-        break;
-      default:
-        result.push_back(ch);
-        break;
-    }
-  }
-  return result;
-}
-
-void StreamingListener::SocketWriter::MakeConnection() {
-  GTEST_CHECK_(sockfd_ == -1)
-      << "MakeConnection() can't be called when there is already a connection.";
-
-  addrinfo hints;
-  memset(&hints, 0, sizeof(hints));
-  hints.ai_family = AF_UNSPEC;    // To allow both IPv4 and IPv6 addresses.
-  hints.ai_socktype = SOCK_STREAM;
-  addrinfo* servinfo = NULL;
-
-  // Use the getaddrinfo() to get a linked list of IP addresses for
-  // the given host name.
-  const int error_num = getaddrinfo(
-      host_name_.c_str(), port_num_.c_str(), &hints, &servinfo);
-  if (error_num != 0) {
-    GTEST_LOG_(WARNING) << "stream_result_to: getaddrinfo() failed: "
-                        << gai_strerror(error_num);
-  }
-
-  // Loop through all the results and connect to the first we can.
-  for (addrinfo* cur_addr = servinfo; sockfd_ == -1 && cur_addr != NULL;
-       cur_addr = cur_addr->ai_next) {
-    sockfd_ = socket(
-        cur_addr->ai_family, cur_addr->ai_socktype, cur_addr->ai_protocol);
-    if (sockfd_ != -1) {
-      // Connect the client socket to the server socket.
-      if (connect(sockfd_, cur_addr->ai_addr, cur_addr->ai_addrlen) == -1) {
-        close(sockfd_);
-        sockfd_ = -1;
-      }
-    }
-  }
-
-  freeaddrinfo(servinfo);  // all done with this structure
-
-  if (sockfd_ == -1) {
-    GTEST_LOG_(WARNING) << "stream_result_to: failed to connect to "
-                        << host_name_ << ":" << port_num_;
-  }
-}
-
-// End of class Streaming Listener
-#endif  // GTEST_CAN_STREAM_RESULTS__
-
-// class OsStackTraceGetter
-
-const char* const OsStackTraceGetterInterface::kElidedFramesMarker =
-    "... " GTEST_NAME_ " internal frames ...";
-
-std::string OsStackTraceGetter::CurrentStackTrace(int max_depth, int skip_count)
-    GTEST_LOCK_EXCLUDED_(mutex_) {
-#if GTEST_HAS_ABSL
-  std::string result;
-
-  if (max_depth <= 0) {
-    return result;
-  }
-
-  max_depth = std::min(max_depth, kMaxStackTraceDepth);
-
-  std::vector<void*> raw_stack(max_depth);
-  // Skips the frames requested by the caller, plus this function.
-  const int raw_stack_size =
-      absl::GetStackTrace(&raw_stack[0], max_depth, skip_count + 1);
-
-  void* caller_frame = nullptr;
-  {
-    MutexLock lock(&mutex_);
-    caller_frame = caller_frame_;
-  }
-
-  for (int i = 0; i < raw_stack_size; ++i) {
-    if (raw_stack[i] == caller_frame &&
-        !GTEST_FLAG(show_internal_stack_frames)) {
-      // Add a marker to the trace and stop adding frames.
-      absl::StrAppend(&result, kElidedFramesMarker, "\n");
-      break;
-    }
-
-    char tmp[1024];
-    const char* symbol = "(unknown)";
-    if (absl::Symbolize(raw_stack[i], tmp, sizeof(tmp))) {
-      symbol = tmp;
-    }
-
-    char line[1024];
-    snprintf(line, sizeof(line), "  %p: %s\n", raw_stack[i], symbol);
-    result += line;
-  }
-
-  return result;
-
-#else  // !GTEST_HAS_ABSL
-  static_cast<void>(max_depth);
-  static_cast<void>(skip_count);
-  return "";
-#endif  // GTEST_HAS_ABSL
-}
-
-void OsStackTraceGetter::UponLeavingGTest() GTEST_LOCK_EXCLUDED_(mutex_) {
-#if GTEST_HAS_ABSL
-  void* caller_frame = nullptr;
-  if (absl::GetStackTrace(&caller_frame, 1, 3) <= 0) {
-    caller_frame = nullptr;
-  }
-
-  MutexLock lock(&mutex_);
-  caller_frame_ = caller_frame;
-#endif  // GTEST_HAS_ABSL
-}
-
-// A helper class that creates the premature-exit file in its
-// constructor and deletes the file in its destructor.
-class ScopedPrematureExitFile {
- public:
-  explicit ScopedPrematureExitFile(const char* premature_exit_filepath)
-      : premature_exit_filepath_(premature_exit_filepath ?
-                                 premature_exit_filepath : "") {
-    // If a path to the premature-exit file is specified...
-    if (!premature_exit_filepath_.empty()) {
-      // create the file with a single "0" character in it.  I/O
-      // errors are ignored as there's nothing better we can do and we
-      // don't want to fail the test because of this.
-      FILE* pfile = posix::FOpen(premature_exit_filepath, "w");
-      fwrite("0", 1, 1, pfile);
-      fclose(pfile);
-    }
-  }
-
-  ~ScopedPrematureExitFile() {
-    if (!premature_exit_filepath_.empty()) {
-      int retval = remove(premature_exit_filepath_.c_str());
-      if (retval) {
-        GTEST_LOG_(ERROR) << "Failed to remove premature exit filepath \""
-                          << premature_exit_filepath_ << "\" with error "
-                          << retval;
-      }
-    }
-  }
-
- private:
-  const std::string premature_exit_filepath_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ScopedPrematureExitFile);
-};
-
-}  // namespace internal
-
-// class TestEventListeners
-
-TestEventListeners::TestEventListeners()
-    : repeater_(new internal::TestEventRepeater()),
-      default_result_printer_(NULL),
-      default_xml_generator_(NULL) {
-}
-
-TestEventListeners::~TestEventListeners() { delete repeater_; }
-
-// Returns the standard listener responsible for the default console
-// output.  Can be removed from the listeners list to shut down default
-// console output.  Note that removing this object from the listener list
-// with Release transfers its ownership to the user.
-void TestEventListeners::Append(TestEventListener* listener) {
-  repeater_->Append(listener);
-}
-
-// Removes the given event listener from the list and returns it.  It then
-// becomes the caller's responsibility to delete the listener. Returns
-// NULL if the listener is not found in the list.
-TestEventListener* TestEventListeners::Release(TestEventListener* listener) {
-  if (listener == default_result_printer_)
-    default_result_printer_ = NULL;
-  else if (listener == default_xml_generator_)
-    default_xml_generator_ = NULL;
-  return repeater_->Release(listener);
-}
-
-// Returns repeater that broadcasts the TestEventListener events to all
-// subscribers.
-TestEventListener* TestEventListeners::repeater() { return repeater_; }
-
-// Sets the default_result_printer attribute to the provided listener.
-// The listener is also added to the listener list and previous
-// default_result_printer is removed from it and deleted. The listener can
-// also be NULL in which case it will not be added to the list. Does
-// nothing if the previous and the current listener objects are the same.
-void TestEventListeners::SetDefaultResultPrinter(TestEventListener* listener) {
-  if (default_result_printer_ != listener) {
-    // It is an error to pass this method a listener that is already in the
-    // list.
-    delete Release(default_result_printer_);
-    default_result_printer_ = listener;
-    if (listener != NULL)
-      Append(listener);
-  }
-}
-
-// Sets the default_xml_generator attribute to the provided listener.  The
-// listener is also added to the listener list and previous
-// default_xml_generator is removed from it and deleted. The listener can
-// also be NULL in which case it will not be added to the list. Does
-// nothing if the previous and the current listener objects are the same.
-void TestEventListeners::SetDefaultXmlGenerator(TestEventListener* listener) {
-  if (default_xml_generator_ != listener) {
-    // It is an error to pass this method a listener that is already in the
-    // list.
-    delete Release(default_xml_generator_);
-    default_xml_generator_ = listener;
-    if (listener != NULL)
-      Append(listener);
-  }
-}
-
-// Controls whether events will be forwarded by the repeater to the
-// listeners in the list.
-bool TestEventListeners::EventForwardingEnabled() const {
-  return repeater_->forwarding_enabled();
-}
-
-void TestEventListeners::SuppressEventForwarding() {
-  repeater_->set_forwarding_enabled(false);
-}
-
-// class UnitTest
-
-// Gets the singleton UnitTest object.  The first time this method is
-// called, a UnitTest object is constructed and returned.  Consecutive
-// calls will return the same object.
-//
-// We don't protect this under mutex_ as a user is not supposed to
-// call this before main() starts, from which point on the return
-// value will never change.
-UnitTest* UnitTest::GetInstance() {
-  // When compiled with MSVC 7.1 in optimized mode, destroying the
-  // UnitTest object upon exiting the program messes up the exit code,
-  // causing successful tests to appear failed.  We have to use a
-  // different implementation in this case to bypass the compiler bug.
-  // This implementation makes the compiler happy, at the cost of
-  // leaking the UnitTest object.
-
-  // CodeGear C++Builder insists on a public destructor for the
-  // default implementation.  Use this implementation to keep good OO
-  // design with private destructor.
-
-#if (_MSC_VER == 1310 && !defined(_DEBUG)) || defined(__BORLANDC__)
-  static UnitTest* const instance = new UnitTest;
-  return instance;
-#else
-  static UnitTest instance;
-  return &instance;
-#endif  // (_MSC_VER == 1310 && !defined(_DEBUG)) || defined(__BORLANDC__)
-}
-
-// Gets the number of successful test cases.
-int UnitTest::successful_test_case_count() const {
-  return impl()->successful_test_case_count();
-}
-
-// Gets the number of failed test cases.
-int UnitTest::failed_test_case_count() const {
-  return impl()->failed_test_case_count();
-}
-
-// Gets the number of all test cases.
-int UnitTest::total_test_case_count() const {
-  return impl()->total_test_case_count();
-}
-
-// Gets the number of all test cases that contain at least one test
-// that should run.
-int UnitTest::test_case_to_run_count() const {
-  return impl()->test_case_to_run_count();
-}
-
-// Gets the number of successful tests.
-int UnitTest::successful_test_count() const {
-  return impl()->successful_test_count();
-}
-
-// Gets the number of failed tests.
-int UnitTest::failed_test_count() const { return impl()->failed_test_count(); }
-
-// Gets the number of disabled tests that will be reported in the XML report.
-int UnitTest::reportable_disabled_test_count() const {
-  return impl()->reportable_disabled_test_count();
-}
-
-// Gets the number of disabled tests.
-int UnitTest::disabled_test_count() const {
-  return impl()->disabled_test_count();
-}
-
-// Gets the number of tests to be printed in the XML report.
-int UnitTest::reportable_test_count() const {
-  return impl()->reportable_test_count();
-}
-
-// Gets the number of all tests.
-int UnitTest::total_test_count() const { return impl()->total_test_count(); }
-
-// Gets the number of tests that should run.
-int UnitTest::test_to_run_count() const { return impl()->test_to_run_count(); }
-
-// Gets the time of the test program start, in ms from the start of the
-// UNIX epoch.
-internal::TimeInMillis UnitTest::start_timestamp() const {
-    return impl()->start_timestamp();
-}
-
-// Gets the elapsed time, in milliseconds.
-internal::TimeInMillis UnitTest::elapsed_time() const {
-  return impl()->elapsed_time();
-}
-
-// Returns true iff the unit test passed (i.e. all test cases passed).
-bool UnitTest::Passed() const { return impl()->Passed(); }
-
-// Returns true iff the unit test failed (i.e. some test case failed
-// or something outside of all tests failed).
-bool UnitTest::Failed() const { return impl()->Failed(); }
-
-// Gets the i-th test case among all the test cases. i can range from 0 to
-// total_test_case_count() - 1. If i is not in that range, returns NULL.
-const TestCase* UnitTest::GetTestCase(int i) const {
-  return impl()->GetTestCase(i);
-}
-
-// Returns the TestResult containing information on test failures and
-// properties logged outside of individual test cases.
-const TestResult& UnitTest::ad_hoc_test_result() const {
-  return *impl()->ad_hoc_test_result();
-}
-
-// Gets the i-th test case among all the test cases. i can range from 0 to
-// total_test_case_count() - 1. If i is not in that range, returns NULL.
-TestCase* UnitTest::GetMutableTestCase(int i) {
-  return impl()->GetMutableTestCase(i);
-}
-
-// Returns the list of event listeners that can be used to track events
-// inside Google Test.
-TestEventListeners& UnitTest::listeners() {
-  return *impl()->listeners();
-}
-
-// Registers and returns a global test environment.  When a test
-// program is run, all global test environments will be set-up in the
-// order they were registered.  After all tests in the program have
-// finished, all global test environments will be torn-down in the
-// *reverse* order they were registered.
-//
-// The UnitTest object takes ownership of the given environment.
-//
-// We don't protect this under mutex_, as we only support calling it
-// from the main thread.
-Environment* UnitTest::AddEnvironment(Environment* env) {
-  if (env == NULL) {
-    return NULL;
-  }
-
-  impl_->environments().push_back(env);
-  return env;
-}
-
-// Adds a TestPartResult to the current TestResult object.  All Google Test
-// assertion macros (e.g. ASSERT_TRUE, EXPECT_EQ, etc) eventually call
-// this to report their results.  The user code should use the
-// assertion macros instead of calling this directly.
-void UnitTest::AddTestPartResult(
-    TestPartResult::Type result_type,
-    const char* file_name,
-    int line_number,
-    const std::string& message,
-    const std::string& os_stack_trace) GTEST_LOCK_EXCLUDED_(mutex_) {
-  Message msg;
-  msg << message;
-
-  internal::MutexLock lock(&mutex_);
-  if (impl_->gtest_trace_stack().size() > 0) {
-    msg << "\n" << GTEST_NAME_ << " trace:";
-
-    for (int i = static_cast<int>(impl_->gtest_trace_stack().size());
-         i > 0; --i) {
-      const internal::TraceInfo& trace = impl_->gtest_trace_stack()[i - 1];
-      msg << "\n" << internal::FormatFileLocation(trace.file, trace.line)
-          << " " << trace.message;
-    }
-  }
-
-  if (os_stack_trace.c_str() != NULL && !os_stack_trace.empty()) {
-    msg << internal::kStackTraceMarker << os_stack_trace;
-  }
-
-  const TestPartResult result =
-    TestPartResult(result_type, file_name, line_number,
-                   msg.GetString().c_str());
-  impl_->GetTestPartResultReporterForCurrentThread()->
-      ReportTestPartResult(result);
-
-  if (result_type != TestPartResult::kSuccess) {
-    // gtest_break_on_failure takes precedence over
-    // gtest_throw_on_failure.  This allows a user to set the latter
-    // in the code (perhaps in order to use Google Test assertions
-    // with another testing framework) and specify the former on the
-    // command line for debugging.
-    if (GTEST_FLAG(break_on_failure)) {
-#if GTEST_OS_WINDOWS && !GTEST_OS_WINDOWS_PHONE && !GTEST_OS_WINDOWS_RT
-      // Using DebugBreak on Windows allows gtest to still break into a debugger
-      // when a failure happens and both the --gtest_break_on_failure and
-      // the --gtest_catch_exceptions flags are specified.
-      DebugBreak();
-#elif (!defined(__native_client__)) &&            \
-    ((defined(__clang__) || defined(__GNUC__)) && \
-     (defined(__x86_64__) || defined(__i386__)))
-      // with clang/gcc we can achieve the same effect on x86 by invoking int3
-      asm("int3");
-#else
-      // Dereference NULL through a volatile pointer to prevent the compiler
-      // from removing. We use this rather than abort() or __builtin_trap() for
-      // portability: Symbian doesn't implement abort() well, and some debuggers
-      // don't correctly trap abort().
-      *static_cast<volatile int*>(NULL) = 1;
-#endif  // GTEST_OS_WINDOWS
-    } else if (GTEST_FLAG(throw_on_failure)) {
-#if GTEST_HAS_EXCEPTIONS
-      throw internal::GoogleTestFailureException(result);
-#else
-      // We cannot call abort() as it generates a pop-up in debug mode
-      // that cannot be suppressed in VC 7.1 or below.
-      exit(1);
-#endif
-    }
-  }
-}
-
-// Adds a TestProperty to the current TestResult object when invoked from
-// inside a test, to current TestCase's ad_hoc_test_result_ when invoked
-// from SetUpTestCase or TearDownTestCase, or to the global property set
-// when invoked elsewhere.  If the result already contains a property with
-// the same key, the value will be updated.
-void UnitTest::RecordProperty(const std::string& key,
-                              const std::string& value) {
-  impl_->RecordProperty(TestProperty(key, value));
-}
-
-// Runs all tests in this UnitTest object and prints the result.
-// Returns 0 if successful, or 1 otherwise.
-//
-// We don't protect this under mutex_, as we only support calling it
-// from the main thread.
-int UnitTest::Run() {
-  const bool in_death_test_child_process =
-      internal::GTEST_FLAG(internal_run_death_test).length() > 0;
-
-  // Google Test implements this protocol for catching that a test
-  // program exits before returning control to Google Test:
-  //
-  //   1. Upon start, Google Test creates a file whose absolute path
-  //      is specified by the environment variable
-  //      TEST_PREMATURE_EXIT_FILE.
-  //   2. When Google Test has finished its work, it deletes the file.
-  //
-  // This allows a test runner to set TEST_PREMATURE_EXIT_FILE before
-  // running a Google-Test-based test program and check the existence
-  // of the file at the end of the test execution to see if it has
-  // exited prematurely.
-
-  // If we are in the child process of a death test, don't
-  // create/delete the premature exit file, as doing so is unnecessary
-  // and will confuse the parent process.  Otherwise, create/delete
-  // the file upon entering/leaving this function.  If the program
-  // somehow exits before this function has a chance to return, the
-  // premature-exit file will be left undeleted, causing a test runner
-  // that understands the premature-exit-file protocol to report the
-  // test as having failed.
-  const internal::ScopedPrematureExitFile premature_exit_file(
-      in_death_test_child_process ?
-      NULL : internal::posix::GetEnv("TEST_PREMATURE_EXIT_FILE"));
-
-  // Captures the value of GTEST_FLAG(catch_exceptions).  This value will be
-  // used for the duration of the program.
-  impl()->set_catch_exceptions(GTEST_FLAG(catch_exceptions));
-
-#if GTEST_HAS_SEH
-  // Either the user wants Google Test to catch exceptions thrown by the
-  // tests or this is executing in the context of death test child
-  // process. In either case the user does not want to see pop-up dialogs
-  // about crashes - they are expected.
-  if (impl()->catch_exceptions() || in_death_test_child_process) {
-# if !GTEST_OS_WINDOWS_MOBILE && !GTEST_OS_WINDOWS_PHONE && !GTEST_OS_WINDOWS_RT
-    // SetErrorMode doesn't exist on CE.
-    SetErrorMode(SEM_FAILCRITICALERRORS | SEM_NOALIGNMENTFAULTEXCEPT |
-                 SEM_NOGPFAULTERRORBOX | SEM_NOOPENFILEERRORBOX);
-# endif  // !GTEST_OS_WINDOWS_MOBILE
-
-# if (defined(_MSC_VER) || GTEST_OS_WINDOWS_MINGW) && !GTEST_OS_WINDOWS_MOBILE
-    // Death test children can be terminated with _abort().  On Windows,
-    // _abort() can show a dialog with a warning message.  This forces the
-    // abort message to go to stderr instead.
-    _set_error_mode(_OUT_TO_STDERR);
-# endif
-
-# if _MSC_VER >= 1400 && !GTEST_OS_WINDOWS_MOBILE
-    // In the debug version, Visual Studio pops up a separate dialog
-    // offering a choice to debug the aborted program. We need to suppress
-    // this dialog or it will pop up for every EXPECT/ASSERT_DEATH statement
-    // executed. Google Test will notify the user of any unexpected
-    // failure via stderr.
-    //
-    // VC++ doesn't define _set_abort_behavior() prior to the version 8.0.
-    // Users of prior VC versions shall suffer the agony and pain of
-    // clicking through the countless debug dialogs.
-    // TODO(vladl@google.com): find a way to suppress the abort dialog() in the
-    // debug mode when compiled with VC 7.1 or lower.
-    if (!GTEST_FLAG(break_on_failure))
-      _set_abort_behavior(
-          0x0,                                    // Clear the following flags:
-          _WRITE_ABORT_MSG | _CALL_REPORTFAULT);  // pop-up window, core dump.
-# endif
-  }
-#endif  // GTEST_HAS_SEH
-
-  return internal::HandleExceptionsInMethodIfSupported(
-      impl(),
-      &internal::UnitTestImpl::RunAllTests,
-      "auxiliary test code (environments or event listeners)") ? 0 : 1;
-}
-
-// Returns the working directory when the first TEST() or TEST_F() was
-// executed.
-const char* UnitTest::original_working_dir() const {
-  return impl_->original_working_dir_.c_str();
-}
-
-// Returns the TestCase object for the test that's currently running,
-// or NULL if no test is running.
-const TestCase* UnitTest::current_test_case() const
-    GTEST_LOCK_EXCLUDED_(mutex_) {
-  internal::MutexLock lock(&mutex_);
-  return impl_->current_test_case();
-}
-
-// Returns the TestInfo object for the test that's currently running,
-// or NULL if no test is running.
-const TestInfo* UnitTest::current_test_info() const
-    GTEST_LOCK_EXCLUDED_(mutex_) {
-  internal::MutexLock lock(&mutex_);
-  return impl_->current_test_info();
-}
-
-// Returns the random seed used at the start of the current test run.
-int UnitTest::random_seed() const { return impl_->random_seed(); }
-
-// Returns ParameterizedTestCaseRegistry object used to keep track of
-// value-parameterized tests and instantiate and register them.
-internal::ParameterizedTestCaseRegistry&
-    UnitTest::parameterized_test_registry()
-        GTEST_LOCK_EXCLUDED_(mutex_) {
-  return impl_->parameterized_test_registry();
-}
-
-// Creates an empty UnitTest.
-UnitTest::UnitTest() {
-  impl_ = new internal::UnitTestImpl(this);
-}
-
-// Destructor of UnitTest.
-UnitTest::~UnitTest() {
-  delete impl_;
-}
-
-// Pushes a trace defined by SCOPED_TRACE() on to the per-thread
-// Google Test trace stack.
-void UnitTest::PushGTestTrace(const internal::TraceInfo& trace)
-    GTEST_LOCK_EXCLUDED_(mutex_) {
-  internal::MutexLock lock(&mutex_);
-  impl_->gtest_trace_stack().push_back(trace);
-}
-
-// Pops a trace from the per-thread Google Test trace stack.
-void UnitTest::PopGTestTrace()
-    GTEST_LOCK_EXCLUDED_(mutex_) {
-  internal::MutexLock lock(&mutex_);
-  impl_->gtest_trace_stack().pop_back();
-}
-
-namespace internal {
-
-UnitTestImpl::UnitTestImpl(UnitTest* parent)
-    : parent_(parent),
-      GTEST_DISABLE_MSC_WARNINGS_PUSH_(4355 /* using this in initializer */)
-      default_global_test_part_result_reporter_(this),
-      default_per_thread_test_part_result_reporter_(this),
-      GTEST_DISABLE_MSC_WARNINGS_POP_()
-      global_test_part_result_repoter_(
-          &default_global_test_part_result_reporter_),
-      per_thread_test_part_result_reporter_(
-          &default_per_thread_test_part_result_reporter_),
-      parameterized_test_registry_(),
-      parameterized_tests_registered_(false),
-      last_death_test_case_(-1),
-      current_test_case_(NULL),
-      current_test_info_(NULL),
-      ad_hoc_test_result_(),
-      os_stack_trace_getter_(NULL),
-      post_flag_parse_init_performed_(false),
-      random_seed_(0),  // Will be overridden by the flag before first use.
-      random_(0),  // Will be reseeded before first use.
-      start_timestamp_(0),
-      elapsed_time_(0),
-#if GTEST_HAS_DEATH_TEST
-      death_test_factory_(new DefaultDeathTestFactory),
-#endif
-      // Will be overridden by the flag before first use.
-      catch_exceptions_(false) {
-  listeners()->SetDefaultResultPrinter(new PrettyUnitTestResultPrinter);
-}
-
-UnitTestImpl::~UnitTestImpl() {
-  // Deletes every TestCase.
-  ForEach(test_cases_, internal::Delete<TestCase>);
-
-  // Deletes every Environment.
-  ForEach(environments_, internal::Delete<Environment>);
-
-  delete os_stack_trace_getter_;
-}
-
-// Adds a TestProperty to the current TestResult object when invoked in a
-// context of a test, to current test case's ad_hoc_test_result when invoke
-// from SetUpTestCase/TearDownTestCase, or to the global property set
-// otherwise.  If the result already contains a property with the same key,
-// the value will be updated.
-void UnitTestImpl::RecordProperty(const TestProperty& test_property) {
-  std::string xml_element;
-  TestResult* test_result;  // TestResult appropriate for property recording.
-
-  if (current_test_info_ != NULL) {
-    xml_element = "testcase";
-    test_result = &(current_test_info_->result_);
-  } else if (current_test_case_ != NULL) {
-    xml_element = "testsuite";
-    test_result = &(current_test_case_->ad_hoc_test_result_);
-  } else {
-    xml_element = "testsuites";
-    test_result = &ad_hoc_test_result_;
-  }
-  test_result->RecordProperty(xml_element, test_property);
-}
-
-#if GTEST_HAS_DEATH_TEST
-// Disables event forwarding if the control is currently in a death test
-// subprocess. Must not be called before InitGoogleTest.
-void UnitTestImpl::SuppressTestEventsIfInSubprocess() {
-  if (internal_run_death_test_flag_.get() != NULL)
-    listeners()->SuppressEventForwarding();
-}
-#endif  // GTEST_HAS_DEATH_TEST
-
-// Initializes event listeners performing XML output as specified by
-// UnitTestOptions. Must not be called before InitGoogleTest.
-void UnitTestImpl::ConfigureXmlOutput() {
-  const std::string& output_format = UnitTestOptions::GetOutputFormat();
-  if (output_format == "xml") {
-    listeners()->SetDefaultXmlGenerator(new XmlUnitTestResultPrinter(
-        UnitTestOptions::GetAbsolutePathToOutputFile().c_str()));
-  } else if (output_format == "json") {
-    listeners()->SetDefaultXmlGenerator(new JsonUnitTestResultPrinter(
-        UnitTestOptions::GetAbsolutePathToOutputFile().c_str()));
-  } else if (output_format != "") {
-    GTEST_LOG_(WARNING) << "WARNING: unrecognized output format \""
-                        << output_format << "\" ignored.";
-  }
-}
-
-#if GTEST_CAN_STREAM_RESULTS_
-// Initializes event listeners for streaming test results in string form.
-// Must not be called before InitGoogleTest.
-void UnitTestImpl::ConfigureStreamingOutput() {
-  const std::string& target = GTEST_FLAG(stream_result_to);
-  if (!target.empty()) {
-    const size_t pos = target.find(':');
-    if (pos != std::string::npos) {
-      listeners()->Append(new StreamingListener(target.substr(0, pos),
-                                                target.substr(pos+1)));
-    } else {
-      GTEST_LOG_(WARNING) << "unrecognized streaming target \"" << target
-                          << "\" ignored.";
-    }
-  }
-}
-#endif  // GTEST_CAN_STREAM_RESULTS_
-
-// Performs initialization dependent upon flag values obtained in
-// ParseGoogleTestFlagsOnly.  Is called from InitGoogleTest after the call to
-// ParseGoogleTestFlagsOnly.  In case a user neglects to call InitGoogleTest
-// this function is also called from RunAllTests.  Since this function can be
-// called more than once, it has to be idempotent.
-void UnitTestImpl::PostFlagParsingInit() {
-  // Ensures that this function does not execute more than once.
-  if (!post_flag_parse_init_performed_) {
-    post_flag_parse_init_performed_ = true;
-
-#if defined(GTEST_CUSTOM_TEST_EVENT_LISTENER_)
-    // Register to send notifications about key process state changes.
-    listeners()->Append(new GTEST_CUSTOM_TEST_EVENT_LISTENER_());
-#endif  // defined(GTEST_CUSTOM_TEST_EVENT_LISTENER_)
-
-#if GTEST_HAS_DEATH_TEST
-    InitDeathTestSubprocessControlInfo();
-    SuppressTestEventsIfInSubprocess();
-#endif  // GTEST_HAS_DEATH_TEST
-
-    // Registers parameterized tests. This makes parameterized tests
-    // available to the UnitTest reflection API without running
-    // RUN_ALL_TESTS.
-    RegisterParameterizedTests();
-
-    // Configures listeners for XML output. This makes it possible for users
-    // to shut down the default XML output before invoking RUN_ALL_TESTS.
-    ConfigureXmlOutput();
-
-#if GTEST_CAN_STREAM_RESULTS_
-    // Configures listeners for streaming test results to the specified server.
-    ConfigureStreamingOutput();
-#endif  // GTEST_CAN_STREAM_RESULTS_
-
-#if GTEST_HAS_ABSL
-    if (GTEST_FLAG(install_failure_signal_handler)) {
-      absl::FailureSignalHandlerOptions options;
-      absl::InstallFailureSignalHandler(options);
-    }
-#endif  // GTEST_HAS_ABSL
-  }
-}
-
-// A predicate that checks the name of a TestCase against a known
-// value.
-//
-// This is used for implementation of the UnitTest class only.  We put
-// it in the anonymous namespace to prevent polluting the outer
-// namespace.
-//
-// TestCaseNameIs is copyable.
-class TestCaseNameIs {
- public:
-  // Constructor.
-  explicit TestCaseNameIs(const std::string& name)
-      : name_(name) {}
-
-  // Returns true iff the name of test_case matches name_.
-  bool operator()(const TestCase* test_case) const {
-    return test_case != NULL && strcmp(test_case->name(), name_.c_str()) == 0;
-  }
-
- private:
-  std::string name_;
-};
-
-// Finds and returns a TestCase with the given name.  If one doesn't
-// exist, creates one and returns it.  It's the CALLER'S
-// RESPONSIBILITY to ensure that this function is only called WHEN THE
-// TESTS ARE NOT SHUFFLED.
-//
-// Arguments:
-//
-//   test_case_name: name of the test case
-//   type_param:     the name of the test case's type parameter, or NULL if
-//                   this is not a typed or a type-parameterized test case.
-//   set_up_tc:      pointer to the function that sets up the test case
-//   tear_down_tc:   pointer to the function that tears down the test case
-TestCase* UnitTestImpl::GetTestCase(const char* test_case_name,
-                                    const char* type_param,
-                                    Test::SetUpTestCaseFunc set_up_tc,
-                                    Test::TearDownTestCaseFunc tear_down_tc) {
-  // Can we find a TestCase with the given name?
-  const std::vector<TestCase*>::const_reverse_iterator test_case =
-      std::find_if(test_cases_.rbegin(), test_cases_.rend(),
-                   TestCaseNameIs(test_case_name));
-
-  if (test_case != test_cases_.rend())
-    return *test_case;
-
-  // No.  Let's create one.
-  TestCase* const new_test_case =
-      new TestCase(test_case_name, type_param, set_up_tc, tear_down_tc);
-
-  // Is this a death test case?
-  if (internal::UnitTestOptions::MatchesFilter(test_case_name,
-                                               kDeathTestCaseFilter)) {
-    // Yes.  Inserts the test case after the last death test case
-    // defined so far.  This only works when the test cases haven't
-    // been shuffled.  Otherwise we may end up running a death test
-    // after a non-death test.
-    ++last_death_test_case_;
-    test_cases_.insert(test_cases_.begin() + last_death_test_case_,
-                       new_test_case);
-  } else {
-    // No.  Appends to the end of the list.
-    test_cases_.push_back(new_test_case);
-  }
-
-  test_case_indices_.push_back(static_cast<int>(test_case_indices_.size()));
-  return new_test_case;
-}
-
-// Helpers for setting up / tearing down the given environment.  They
-// are for use in the ForEach() function.
-static void SetUpEnvironment(Environment* env) { env->SetUp(); }
-static void TearDownEnvironment(Environment* env) { env->TearDown(); }
-
-// Runs all tests in this UnitTest object, prints the result, and
-// returns true if all tests are successful.  If any exception is
-// thrown during a test, the test is considered to be failed, but the
-// rest of the tests will still be run.
-//
-// When parameterized tests are enabled, it expands and registers
-// parameterized tests first in RegisterParameterizedTests().
-// All other functions called from RunAllTests() may safely assume that
-// parameterized tests are ready to be counted and run.
-bool UnitTestImpl::RunAllTests() {
-  // True iff Google Test is initialized before RUN_ALL_TESTS() is called.
-  const bool gtest_is_initialized_before_run_all_tests = GTestIsInitialized();
-
-  // Do not run any test if the --help flag was specified.
-  if (g_help_flag)
-    return true;
-
-  // Repeats the call to the post-flag parsing initialization in case the
-  // user didn't call InitGoogleTest.
-  PostFlagParsingInit();
-
-  // Even if sharding is not on, test runners may want to use the
-  // GTEST_SHARD_STATUS_FILE to query whether the test supports the sharding
-  // protocol.
-  internal::WriteToShardStatusFileIfNeeded();
-
-  // True iff we are in a subprocess for running a thread-safe-style
-  // death test.
-  bool in_subprocess_for_death_test = false;
-
-#if GTEST_HAS_DEATH_TEST
-  in_subprocess_for_death_test = (internal_run_death_test_flag_.get() != NULL);
-# if defined(GTEST_EXTRA_DEATH_TEST_CHILD_SETUP_)
-  if (in_subprocess_for_death_test) {
-    GTEST_EXTRA_DEATH_TEST_CHILD_SETUP_();
-  }
-# endif  // defined(GTEST_EXTRA_DEATH_TEST_CHILD_SETUP_)
-#endif  // GTEST_HAS_DEATH_TEST
-
-  const bool should_shard = ShouldShard(kTestTotalShards, kTestShardIndex,
-                                        in_subprocess_for_death_test);
-
-  // Compares the full test names with the filter to decide which
-  // tests to run.
-  const bool has_tests_to_run = FilterTests(should_shard
-                                              ? HONOR_SHARDING_PROTOCOL
-                                              : IGNORE_SHARDING_PROTOCOL) > 0;
-
-  // Lists the tests and exits if the --gtest_list_tests flag was specified.
-  if (GTEST_FLAG(list_tests)) {
-    // This must be called *after* FilterTests() has been called.
-    ListTestsMatchingFilter();
-    return true;
-  }
-
-  random_seed_ = GTEST_FLAG(shuffle) ?
-      GetRandomSeedFromFlag(GTEST_FLAG(random_seed)) : 0;
-
-  // True iff at least one test has failed.
-  bool failed = false;
-
-  TestEventListener* repeater = listeners()->repeater();
-
-  start_timestamp_ = GetTimeInMillis();
-  repeater->OnTestProgramStart(*parent_);
-
-  // How many times to repeat the tests?  We don't want to repeat them
-  // when we are inside the subprocess of a death test.
-  const int repeat = in_subprocess_for_death_test ? 1 : GTEST_FLAG(repeat);
-  // Repeats forever if the repeat count is negative.
-  const bool forever = repeat < 0;
-  for (int i = 0; forever || i != repeat; i++) {
-    // We want to preserve failures generated by ad-hoc test
-    // assertions executed before RUN_ALL_TESTS().
-    ClearNonAdHocTestResult();
-
-    const TimeInMillis start = GetTimeInMillis();
-
-    // Shuffles test cases and tests if requested.
-    if (has_tests_to_run && GTEST_FLAG(shuffle)) {
-      random()->Reseed(random_seed_);
-      // This should be done before calling OnTestIterationStart(),
-      // such that a test event listener can see the actual test order
-      // in the event.
-      ShuffleTests();
-    }
-
-    // Tells the unit test event listeners that the tests are about to start.
-    repeater->OnTestIterationStart(*parent_, i);
-
-    // Runs each test case if there is at least one test to run.
-    if (has_tests_to_run) {
-      // Sets up all environments beforehand.
-      repeater->OnEnvironmentsSetUpStart(*parent_);
-      ForEach(environments_, SetUpEnvironment);
-      repeater->OnEnvironmentsSetUpEnd(*parent_);
-
-      // Runs the tests only if there was no fatal failure during global
-      // set-up.
-      if (!Test::HasFatalFailure()) {
-        for (int test_index = 0; test_index < total_test_case_count();
-             test_index++) {
-          GetMutableTestCase(test_index)->Run();
-        }
-      }
-
-      // Tears down all environments in reverse order afterwards.
-      repeater->OnEnvironmentsTearDownStart(*parent_);
-      std::for_each(environments_.rbegin(), environments_.rend(),
-                    TearDownEnvironment);
-      repeater->OnEnvironmentsTearDownEnd(*parent_);
-    }
-
-    elapsed_time_ = GetTimeInMillis() - start;
-
-    // Tells the unit test event listener that the tests have just finished.
-    repeater->OnTestIterationEnd(*parent_, i);
-
-    // Gets the result and clears it.
-    if (!Passed()) {
-      failed = true;
-    }
-
-    // Restores the original test order after the iteration.  This
-    // allows the user to quickly repro a failure that happens in the
-    // N-th iteration without repeating the first (N - 1) iterations.
-    // This is not enclosed in "if (GTEST_FLAG(shuffle)) { ... }", in
-    // case the user somehow changes the value of the flag somewhere
-    // (it's always safe to unshuffle the tests).
-    UnshuffleTests();
-
-    if (GTEST_FLAG(shuffle)) {
-      // Picks a new random seed for each iteration.
-      random_seed_ = GetNextRandomSeed(random_seed_);
-    }
-  }
-
-  repeater->OnTestProgramEnd(*parent_);
-
-  if (!gtest_is_initialized_before_run_all_tests) {
-    ColoredPrintf(
-        COLOR_RED,
-        "\nIMPORTANT NOTICE - DO NOT IGNORE:\n"
-        "This test program did NOT call " GTEST_INIT_GOOGLE_TEST_NAME_
-        "() before calling RUN_ALL_TESTS(). This is INVALID. Soon " GTEST_NAME_
-        " will start to enforce the valid usage. "
-        "Please fix it ASAP, or IT WILL START TO FAIL.\n");  // NOLINT
-#if GTEST_FOR_GOOGLE_
-    ColoredPrintf(COLOR_RED,
-                  "For more details, see http://wiki/Main/ValidGUnitMain.\n");
-#endif  // GTEST_FOR_GOOGLE_
-  }
-
-  return !failed;
-}
-
-// Reads the GTEST_SHARD_STATUS_FILE environment variable, and creates the file
-// if the variable is present. If a file already exists at this location, this
-// function will write over it. If the variable is present, but the file cannot
-// be created, prints an error and exits.
-void WriteToShardStatusFileIfNeeded() {
-  const char* const test_shard_file = posix::GetEnv(kTestShardStatusFile);
-  if (test_shard_file != NULL) {
-    FILE* const file = posix::FOpen(test_shard_file, "w");
-    if (file == NULL) {
-      ColoredPrintf(COLOR_RED,
-                    "Could not write to the test shard status file \"%s\" "
-                    "specified by the %s environment variable.\n",
-                    test_shard_file, kTestShardStatusFile);
-      fflush(stdout);
-      exit(EXIT_FAILURE);
-    }
-    fclose(file);
-  }
-}
-
-// Checks whether sharding is enabled by examining the relevant
-// environment variable values. If the variables are present,
-// but inconsistent (i.e., shard_index >= total_shards), prints
-// an error and exits. If in_subprocess_for_death_test, sharding is
-// disabled because it must only be applied to the original test
-// process. Otherwise, we could filter out death tests we intended to execute.
-bool ShouldShard(const char* total_shards_env,
-                 const char* shard_index_env,
-                 bool in_subprocess_for_death_test) {
-  if (in_subprocess_for_death_test) {
-    return false;
-  }
-
-  const Int32 total_shards = Int32FromEnvOrDie(total_shards_env, -1);
-  const Int32 shard_index = Int32FromEnvOrDie(shard_index_env, -1);
-
-  if (total_shards == -1 && shard_index == -1) {
-    return false;
-  } else if (total_shards == -1 && shard_index != -1) {
-    const Message msg = Message()
-      << "Invalid environment variables: you have "
-      << kTestShardIndex << " = " << shard_index
-      << ", but have left " << kTestTotalShards << " unset.\n";
-    ColoredPrintf(COLOR_RED, msg.GetString().c_str());
-    fflush(stdout);
-    exit(EXIT_FAILURE);
-  } else if (total_shards != -1 && shard_index == -1) {
-    const Message msg = Message()
-      << "Invalid environment variables: you have "
-      << kTestTotalShards << " = " << total_shards
-      << ", but have left " << kTestShardIndex << " unset.\n";
-    ColoredPrintf(COLOR_RED, msg.GetString().c_str());
-    fflush(stdout);
-    exit(EXIT_FAILURE);
-  } else if (shard_index < 0 || shard_index >= total_shards) {
-    const Message msg = Message()
-      << "Invalid environment variables: we require 0 <= "
-      << kTestShardIndex << " < " << kTestTotalShards
-      << ", but you have " << kTestShardIndex << "=" << shard_index
-      << ", " << kTestTotalShards << "=" << total_shards << ".\n";
-    ColoredPrintf(COLOR_RED, msg.GetString().c_str());
-    fflush(stdout);
-    exit(EXIT_FAILURE);
-  }
-
-  return total_shards > 1;
-}
-
-// Parses the environment variable var as an Int32. If it is unset,
-// returns default_val. If it is not an Int32, prints an error
-// and aborts.
-Int32 Int32FromEnvOrDie(const char* var, Int32 default_val) {
-  const char* str_val = posix::GetEnv(var);
-  if (str_val == NULL) {
-    return default_val;
-  }
-
-  Int32 result;
-  if (!ParseInt32(Message() << "The value of environment variable " << var,
-                  str_val, &result)) {
-    exit(EXIT_FAILURE);
-  }
-  return result;
-}
-
-// Given the total number of shards, the shard index, and the test id,
-// returns true iff the test should be run on this shard. The test id is
-// some arbitrary but unique non-negative integer assigned to each test
-// method. Assumes that 0 <= shard_index < total_shards.
-bool ShouldRunTestOnShard(int total_shards, int shard_index, int test_id) {
-  return (test_id % total_shards) == shard_index;
-}
-
-// Compares the name of each test with the user-specified filter to
-// decide whether the test should be run, then records the result in
-// each TestCase and TestInfo object.
-// If shard_tests == true, further filters tests based on sharding
-// variables in the environment - see
-// https://github.com/google/googletest/blob/master/googletest/docs/advanced.md
-// . Returns the number of tests that should run.
-int UnitTestImpl::FilterTests(ReactionToSharding shard_tests) {
-  const Int32 total_shards = shard_tests == HONOR_SHARDING_PROTOCOL ?
-      Int32FromEnvOrDie(kTestTotalShards, -1) : -1;
-  const Int32 shard_index = shard_tests == HONOR_SHARDING_PROTOCOL ?
-      Int32FromEnvOrDie(kTestShardIndex, -1) : -1;
-
-  // num_runnable_tests are the number of tests that will
-  // run across all shards (i.e., match filter and are not disabled).
-  // num_selected_tests are the number of tests to be run on
-  // this shard.
-  int num_runnable_tests = 0;
-  int num_selected_tests = 0;
-  for (size_t i = 0; i < test_cases_.size(); i++) {
-    TestCase* const test_case = test_cases_[i];
-    const std::string &test_case_name = test_case->name();
-    test_case->set_should_run(false);
-
-    for (size_t j = 0; j < test_case->test_info_list().size(); j++) {
-      TestInfo* const test_info = test_case->test_info_list()[j];
-      const std::string test_name(test_info->name());
-      // A test is disabled if test case name or test name matches
-      // kDisableTestFilter.
-      const bool is_disabled =
-          internal::UnitTestOptions::MatchesFilter(test_case_name,
-                                                   kDisableTestFilter) ||
-          internal::UnitTestOptions::MatchesFilter(test_name,
-                                                   kDisableTestFilter);
-      test_info->is_disabled_ = is_disabled;
-
-      const bool matches_filter =
-          internal::UnitTestOptions::FilterMatchesTest(test_case_name,
-                                                       test_name);
-      test_info->matches_filter_ = matches_filter;
-
-      const bool is_runnable =
-          (GTEST_FLAG(also_run_disabled_tests) || !is_disabled) &&
-          matches_filter;
-
-      const bool is_in_another_shard =
-          shard_tests != IGNORE_SHARDING_PROTOCOL &&
-          !ShouldRunTestOnShard(total_shards, shard_index, num_runnable_tests);
-      test_info->is_in_another_shard_ = is_in_another_shard;
-      const bool is_selected = is_runnable && !is_in_another_shard;
-
-      num_runnable_tests += is_runnable;
-      num_selected_tests += is_selected;
-
-      test_info->should_run_ = is_selected;
-      test_case->set_should_run(test_case->should_run() || is_selected);
-    }
-  }
-  return num_selected_tests;
-}
-
-// Prints the given C-string on a single line by replacing all '\n'
-// characters with string "\\n".  If the output takes more than
-// max_length characters, only prints the first max_length characters
-// and "...".
-static void PrintOnOneLine(const char* str, int max_length) {
-  if (str != NULL) {
-    for (int i = 0; *str != '\0'; ++str) {
-      if (i >= max_length) {
-        printf("...");
-        break;
-      }
-      if (*str == '\n') {
-        printf("\\n");
-        i += 2;
-      } else {
-        printf("%c", *str);
-        ++i;
-      }
-    }
-  }
-}
-
-// Prints the names of the tests matching the user-specified filter flag.
-void UnitTestImpl::ListTestsMatchingFilter() {
-  // Print at most this many characters for each type/value parameter.
-  const int kMaxParamLength = 250;
-
-  for (size_t i = 0; i < test_cases_.size(); i++) {
-    const TestCase* const test_case = test_cases_[i];
-    bool printed_test_case_name = false;
-
-    for (size_t j = 0; j < test_case->test_info_list().size(); j++) {
-      const TestInfo* const test_info =
-          test_case->test_info_list()[j];
-      if (test_info->matches_filter_) {
-        if (!printed_test_case_name) {
-          printed_test_case_name = true;
-          printf("%s.", test_case->name());
-          if (test_case->type_param() != NULL) {
-            printf("  # %s = ", kTypeParamLabel);
-            // We print the type parameter on a single line to make
-            // the output easy to parse by a program.
-            PrintOnOneLine(test_case->type_param(), kMaxParamLength);
-          }
-          printf("\n");
-        }
-        printf("  %s", test_info->name());
-        if (test_info->value_param() != NULL) {
-          printf("  # %s = ", kValueParamLabel);
-          // We print the value parameter on a single line to make the
-          // output easy to parse by a program.
-          PrintOnOneLine(test_info->value_param(), kMaxParamLength);
-        }
-        printf("\n");
-      }
-    }
-  }
-  fflush(stdout);
-}
-
-// Sets the OS stack trace getter.
-//
-// Does nothing if the input and the current OS stack trace getter are
-// the same; otherwise, deletes the old getter and makes the input the
-// current getter.
-void UnitTestImpl::set_os_stack_trace_getter(
-    OsStackTraceGetterInterface* getter) {
-  if (os_stack_trace_getter_ != getter) {
-    delete os_stack_trace_getter_;
-    os_stack_trace_getter_ = getter;
-  }
-}
-
-// Returns the current OS stack trace getter if it is not NULL;
-// otherwise, creates an OsStackTraceGetter, makes it the current
-// getter, and returns it.
-OsStackTraceGetterInterface* UnitTestImpl::os_stack_trace_getter() {
-  if (os_stack_trace_getter_ == NULL) {
-#ifdef GTEST_OS_STACK_TRACE_GETTER_
-    os_stack_trace_getter_ = new GTEST_OS_STACK_TRACE_GETTER_;
-#else
-    os_stack_trace_getter_ = new OsStackTraceGetter;
-#endif  // GTEST_OS_STACK_TRACE_GETTER_
-  }
-
-  return os_stack_trace_getter_;
-}
-
-// Returns the most specific TestResult currently running.
-TestResult* UnitTestImpl::current_test_result() {
-  if (current_test_info_ != NULL) {
-    return &current_test_info_->result_;
-  }
-  if (current_test_case_ != NULL) {
-    return &current_test_case_->ad_hoc_test_result_;
-  }
-  return &ad_hoc_test_result_;
-}
-
-// Shuffles all test cases, and the tests within each test case,
-// making sure that death tests are still run first.
-void UnitTestImpl::ShuffleTests() {
-  // Shuffles the death test cases.
-  ShuffleRange(random(), 0, last_death_test_case_ + 1, &test_case_indices_);
-
-  // Shuffles the non-death test cases.
-  ShuffleRange(random(), last_death_test_case_ + 1,
-               static_cast<int>(test_cases_.size()), &test_case_indices_);
-
-  // Shuffles the tests inside each test case.
-  for (size_t i = 0; i < test_cases_.size(); i++) {
-    test_cases_[i]->ShuffleTests(random());
-  }
-}
-
-// Restores the test cases and tests to their order before the first shuffle.
-void UnitTestImpl::UnshuffleTests() {
-  for (size_t i = 0; i < test_cases_.size(); i++) {
-    // Unshuffles the tests in each test case.
-    test_cases_[i]->UnshuffleTests();
-    // Resets the index of each test case.
-    test_case_indices_[i] = static_cast<int>(i);
-  }
-}
-
-// Returns the current OS stack trace as an std::string.
-//
-// The maximum number of stack frames to be included is specified by
-// the gtest_stack_trace_depth flag.  The skip_count parameter
-// specifies the number of top frames to be skipped, which doesn't
-// count against the number of frames to be included.
-//
-// For example, if Foo() calls Bar(), which in turn calls
-// GetCurrentOsStackTraceExceptTop(..., 1), Foo() will be included in
-// the trace but Bar() and GetCurrentOsStackTraceExceptTop() won't.
-std::string GetCurrentOsStackTraceExceptTop(UnitTest* /*unit_test*/,
-                                            int skip_count) {
-  // We pass skip_count + 1 to skip this wrapper function in addition
-  // to what the user really wants to skip.
-  return GetUnitTestImpl()->CurrentOsStackTraceExceptTop(skip_count + 1);
-}
-
-// Used by the GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_ macro to
-// suppress unreachable code warnings.
-namespace {
-class ClassUniqueToAlwaysTrue {};
-}
-
-bool IsTrue(bool condition) { return condition; }
-
-bool AlwaysTrue() {
-#if GTEST_HAS_EXCEPTIONS
-  // This condition is always false so AlwaysTrue() never actually throws,
-  // but it makes the compiler think that it may throw.
-  if (IsTrue(false))
-    throw ClassUniqueToAlwaysTrue();
-#endif  // GTEST_HAS_EXCEPTIONS
-  return true;
-}
-
-// If *pstr starts with the given prefix, modifies *pstr to be right
-// past the prefix and returns true; otherwise leaves *pstr unchanged
-// and returns false.  None of pstr, *pstr, and prefix can be NULL.
-bool SkipPrefix(const char* prefix, const char** pstr) {
-  const size_t prefix_len = strlen(prefix);
-  if (strncmp(*pstr, prefix, prefix_len) == 0) {
-    *pstr += prefix_len;
-    return true;
-  }
-  return false;
-}
-
-// Parses a string as a command line flag.  The string should have
-// the format "--flag=value".  When def_optional is true, the "=value"
-// part can be omitted.
-//
-// Returns the value of the flag, or NULL if the parsing failed.
-static const char* ParseFlagValue(const char* str, const char* flag,
-                                  bool def_optional) {
-  // str and flag must not be NULL.
-  if (str == NULL || flag == NULL) return NULL;
-
-  // The flag must start with "--" followed by GTEST_FLAG_PREFIX_.
-  const std::string flag_str = std::string("--") + GTEST_FLAG_PREFIX_ + flag;
-  const size_t flag_len = flag_str.length();
-  if (strncmp(str, flag_str.c_str(), flag_len) != 0) return NULL;
-
-  // Skips the flag name.
-  const char* flag_end = str + flag_len;
-
-  // When def_optional is true, it's OK to not have a "=value" part.
-  if (def_optional && (flag_end[0] == '\0')) {
-    return flag_end;
-  }
-
-  // If def_optional is true and there are more characters after the
-  // flag name, or if def_optional is false, there must be a '=' after
-  // the flag name.
-  if (flag_end[0] != '=') return NULL;
-
-  // Returns the string after "=".
-  return flag_end + 1;
-}
-
-// Parses a string for a bool flag, in the form of either
-// "--flag=value" or "--flag".
-//
-// In the former case, the value is taken as true as long as it does
-// not start with '0', 'f', or 'F'.
-//
-// In the latter case, the value is taken as true.
-//
-// On success, stores the value of the flag in *value, and returns
-// true.  On failure, returns false without changing *value.
-static bool ParseBoolFlag(const char* str, const char* flag, bool* value) {
-  // Gets the value of the flag as a string.
-  const char* const value_str = ParseFlagValue(str, flag, true);
-
-  // Aborts if the parsing failed.
-  if (value_str == NULL) return false;
-
-  // Converts the string value to a bool.
-  *value = !(*value_str == '0' || *value_str == 'f' || *value_str == 'F');
-  return true;
-}
-
-// Parses a string for an Int32 flag, in the form of
-// "--flag=value".
-//
-// On success, stores the value of the flag in *value, and returns
-// true.  On failure, returns false without changing *value.
-bool ParseInt32Flag(const char* str, const char* flag, Int32* value) {
-  // Gets the value of the flag as a string.
-  const char* const value_str = ParseFlagValue(str, flag, false);
-
-  // Aborts if the parsing failed.
-  if (value_str == NULL) return false;
-
-  // Sets *value to the value of the flag.
-  return ParseInt32(Message() << "The value of flag --" << flag,
-                    value_str, value);
-}
-
-// Parses a string for a string flag, in the form of
-// "--flag=value".
-//
-// On success, stores the value of the flag in *value, and returns
-// true.  On failure, returns false without changing *value.
-template <typename String>
-static bool ParseStringFlag(const char* str, const char* flag, String* value) {
-  // Gets the value of the flag as a string.
-  const char* const value_str = ParseFlagValue(str, flag, false);
-
-  // Aborts if the parsing failed.
-  if (value_str == NULL) return false;
-
-  // Sets *value to the value of the flag.
-  *value = value_str;
-  return true;
-}
-
-// Determines whether a string has a prefix that Google Test uses for its
-// flags, i.e., starts with GTEST_FLAG_PREFIX_ or GTEST_FLAG_PREFIX_DASH_.
-// If Google Test detects that a command line flag has its prefix but is not
-// recognized, it will print its help message. Flags starting with
-// GTEST_INTERNAL_PREFIX_ followed by "internal_" are considered Google Test
-// internal flags and do not trigger the help message.
-static bool HasGoogleTestFlagPrefix(const char* str) {
-  return (SkipPrefix("--", &str) ||
-          SkipPrefix("-", &str) ||
-          SkipPrefix("/", &str)) &&
-         !SkipPrefix(GTEST_FLAG_PREFIX_ "internal_", &str) &&
-         (SkipPrefix(GTEST_FLAG_PREFIX_, &str) ||
-          SkipPrefix(GTEST_FLAG_PREFIX_DASH_, &str));
-}
-
-// Prints a string containing code-encoded text.  The following escape
-// sequences can be used in the string to control the text color:
-//
-//   @@    prints a single '@' character.
-//   @R    changes the color to red.
-//   @G    changes the color to green.
-//   @Y    changes the color to yellow.
-//   @D    changes to the default terminal text color.
-//
-// TODO(wan@google.com): Write tests for this once we add stdout
-// capturing to Google Test.
-static void PrintColorEncoded(const char* str) {
-  GTestColor color = COLOR_DEFAULT;  // The current color.
-
-  // Conceptually, we split the string into segments divided by escape
-  // sequences.  Then we print one segment at a time.  At the end of
-  // each iteration, the str pointer advances to the beginning of the
-  // next segment.
-  for (;;) {
-    const char* p = strchr(str, '@');
-    if (p == NULL) {
-      ColoredPrintf(color, "%s", str);
-      return;
-    }
-
-    ColoredPrintf(color, "%s", std::string(str, p).c_str());
-
-    const char ch = p[1];
-    str = p + 2;
-    if (ch == '@') {
-      ColoredPrintf(color, "@");
-    } else if (ch == 'D') {
-      color = COLOR_DEFAULT;
-    } else if (ch == 'R') {
-      color = COLOR_RED;
-    } else if (ch == 'G') {
-      color = COLOR_GREEN;
-    } else if (ch == 'Y') {
-      color = COLOR_YELLOW;
-    } else {
-      --str;
-    }
-  }
-}
-
-static const char kColorEncodedHelpMessage[] =
-"This program contains tests written using " GTEST_NAME_ ". You can use the\n"
-"following command line flags to control its behavior:\n"
-"\n"
-"Test Selection:\n"
-"  @G--" GTEST_FLAG_PREFIX_ "list_tests@D\n"
-"      List the names of all tests instead of running them. The name of\n"
-"      TEST(Foo, Bar) is \"Foo.Bar\".\n"
-"  @G--" GTEST_FLAG_PREFIX_ "filter=@YPOSTIVE_PATTERNS"
-    "[@G-@YNEGATIVE_PATTERNS]@D\n"
-"      Run only the tests whose name matches one of the positive patterns but\n"
-"      none of the negative patterns. '?' matches any single character; '*'\n"
-"      matches any substring; ':' separates two patterns.\n"
-"  @G--" GTEST_FLAG_PREFIX_ "also_run_disabled_tests@D\n"
-"      Run all disabled tests too.\n"
-"\n"
-"Test Execution:\n"
-"  @G--" GTEST_FLAG_PREFIX_ "repeat=@Y[COUNT]@D\n"
-"      Run the tests repeatedly; use a negative count to repeat forever.\n"
-"  @G--" GTEST_FLAG_PREFIX_ "shuffle@D\n"
-"      Randomize tests' orders on every iteration.\n"
-"  @G--" GTEST_FLAG_PREFIX_ "random_seed=@Y[NUMBER]@D\n"
-"      Random number seed to use for shuffling test orders (between 1 and\n"
-"      99999, or 0 to use a seed based on the current time).\n"
-"\n"
-"Test Output:\n"
-"  @G--" GTEST_FLAG_PREFIX_ "color=@Y(@Gyes@Y|@Gno@Y|@Gauto@Y)@D\n"
-"      Enable/disable colored output. The default is @Gauto@D.\n"
-"  -@G-" GTEST_FLAG_PREFIX_ "print_time=0@D\n"
-"      Don't print the elapsed time of each test.\n"
-"  @G--" GTEST_FLAG_PREFIX_ "output=@Y(@Gjson@Y|@Gxml@Y)[@G:@YDIRECTORY_PATH@G"
-    GTEST_PATH_SEP_ "@Y|@G:@YFILE_PATH]@D\n"
-"      Generate a JSON or XML report in the given directory or with the given\n"
-"      file name. @YFILE_PATH@D defaults to @Gtest_details.xml@D.\n"
-# if GTEST_CAN_STREAM_RESULTS_
-"  @G--" GTEST_FLAG_PREFIX_ "stream_result_to=@YHOST@G:@YPORT@D\n"
-"      Stream test results to the given server.\n"
-# endif  // GTEST_CAN_STREAM_RESULTS_
-"\n"
-"Assertion Behavior:\n"
-# if GTEST_HAS_DEATH_TEST && !GTEST_OS_WINDOWS
-"  @G--" GTEST_FLAG_PREFIX_ "death_test_style=@Y(@Gfast@Y|@Gthreadsafe@Y)@D\n"
-"      Set the default death test style.\n"
-# endif  // GTEST_HAS_DEATH_TEST && !GTEST_OS_WINDOWS
-"  @G--" GTEST_FLAG_PREFIX_ "break_on_failure@D\n"
-"      Turn assertion failures into debugger break-points.\n"
-"  @G--" GTEST_FLAG_PREFIX_ "throw_on_failure@D\n"
-"      Turn assertion failures into C++ exceptions for use by an external\n"
-"      test framework.\n"
-"  @G--" GTEST_FLAG_PREFIX_ "catch_exceptions=0@D\n"
-"      Do not report exceptions as test failures. Instead, allow them\n"
-"      to crash the program or throw a pop-up (on Windows).\n"
-"\n"
-"Except for @G--" GTEST_FLAG_PREFIX_ "list_tests@D, you can alternatively set "
-    "the corresponding\n"
-"environment variable of a flag (all letters in upper-case). For example, to\n"
-"disable colored text output, you can either specify @G--" GTEST_FLAG_PREFIX_
-    "color=no@D or set\n"
-"the @G" GTEST_FLAG_PREFIX_UPPER_ "COLOR@D environment variable to @Gno@D.\n"
-"\n"
-"For more information, please read the " GTEST_NAME_ " documentation at\n"
-"@G" GTEST_PROJECT_URL_ "@D. If you find a bug in " GTEST_NAME_ "\n"
-"(not one in your own code or tests), please report it to\n"
-"@G<" GTEST_DEV_EMAIL_ ">@D.\n";
-
-static bool ParseGoogleTestFlag(const char* const arg) {
-  return ParseBoolFlag(arg, kAlsoRunDisabledTestsFlag,
-                       &GTEST_FLAG(also_run_disabled_tests)) ||
-      ParseBoolFlag(arg, kBreakOnFailureFlag,
-                    &GTEST_FLAG(break_on_failure)) ||
-      ParseBoolFlag(arg, kCatchExceptionsFlag,
-                    &GTEST_FLAG(catch_exceptions)) ||
-      ParseStringFlag(arg, kColorFlag, &GTEST_FLAG(color)) ||
-      ParseStringFlag(arg, kDeathTestStyleFlag,
-                      &GTEST_FLAG(death_test_style)) ||
-      ParseBoolFlag(arg, kDeathTestUseFork,
-                    &GTEST_FLAG(death_test_use_fork)) ||
-      ParseStringFlag(arg, kFilterFlag, &GTEST_FLAG(filter)) ||
-      ParseStringFlag(arg, kInternalRunDeathTestFlag,
-                      &GTEST_FLAG(internal_run_death_test)) ||
-      ParseBoolFlag(arg, kListTestsFlag, &GTEST_FLAG(list_tests)) ||
-      ParseStringFlag(arg, kOutputFlag, &GTEST_FLAG(output)) ||
-      ParseBoolFlag(arg, kPrintTimeFlag, &GTEST_FLAG(print_time)) ||
-      ParseBoolFlag(arg, kPrintUTF8Flag, &GTEST_FLAG(print_utf8)) ||
-      ParseInt32Flag(arg, kRandomSeedFlag, &GTEST_FLAG(random_seed)) ||
-      ParseInt32Flag(arg, kRepeatFlag, &GTEST_FLAG(repeat)) ||
-      ParseBoolFlag(arg, kShuffleFlag, &GTEST_FLAG(shuffle)) ||
-      ParseInt32Flag(arg, kStackTraceDepthFlag,
-                     &GTEST_FLAG(stack_trace_depth)) ||
-      ParseStringFlag(arg, kStreamResultToFlag,
-                      &GTEST_FLAG(stream_result_to)) ||
-      ParseBoolFlag(arg, kThrowOnFailureFlag,
-                    &GTEST_FLAG(throw_on_failure));
-}
-
-#if GTEST_USE_OWN_FLAGFILE_FLAG_
-static void LoadFlagsFromFile(const std::string& path) {
-  FILE* flagfile = posix::FOpen(path.c_str(), "r");
-  if (!flagfile) {
-    GTEST_LOG_(FATAL) << "Unable to open file \"" << GTEST_FLAG(flagfile)
-                      << "\"";
-  }
-  std::string contents(ReadEntireFile(flagfile));
-  posix::FClose(flagfile);
-  std::vector<std::string> lines;
-  SplitString(contents, '\n', &lines);
-  for (size_t i = 0; i < lines.size(); ++i) {
-    if (lines[i].empty())
-      continue;
-    if (!ParseGoogleTestFlag(lines[i].c_str()))
-      g_help_flag = true;
-  }
-}
-#endif  // GTEST_USE_OWN_FLAGFILE_FLAG_
-
-// Parses the command line for Google Test flags, without initializing
-// other parts of Google Test.  The type parameter CharType can be
-// instantiated to either char or wchar_t.
-template <typename CharType>
-void ParseGoogleTestFlagsOnlyImpl(int* argc, CharType** argv) {
-  for (int i = 1; i < *argc; i++) {
-    const std::string arg_string = StreamableToString(argv[i]);
-    const char* const arg = arg_string.c_str();
-
-    using internal::ParseBoolFlag;
-    using internal::ParseInt32Flag;
-    using internal::ParseStringFlag;
-
-    bool remove_flag = false;
-    if (ParseGoogleTestFlag(arg)) {
-      remove_flag = true;
-#if GTEST_USE_OWN_FLAGFILE_FLAG_
-    } else if (ParseStringFlag(arg, kFlagfileFlag, &GTEST_FLAG(flagfile))) {
-      LoadFlagsFromFile(GTEST_FLAG(flagfile));
-      remove_flag = true;
-#endif  // GTEST_USE_OWN_FLAGFILE_FLAG_
-    } else if (arg_string == "--help" || arg_string == "-h" ||
-               arg_string == "-?" || arg_string == "/?" ||
-               HasGoogleTestFlagPrefix(arg)) {
-      // Both help flag and unrecognized Google Test flags (excluding
-      // internal ones) trigger help display.
-      g_help_flag = true;
-    }
-
-    if (remove_flag) {
-      // Shift the remainder of the argv list left by one.  Note
-      // that argv has (*argc + 1) elements, the last one always being
-      // NULL.  The following loop moves the trailing NULL element as
-      // well.
-      for (int j = i; j != *argc; j++) {
-        argv[j] = argv[j + 1];
-      }
-
-      // Decrements the argument count.
-      (*argc)--;
-
-      // We also need to decrement the iterator as we just removed
-      // an element.
-      i--;
-    }
-  }
-
-  if (g_help_flag) {
-    // We print the help here instead of in RUN_ALL_TESTS(), as the
-    // latter may not be called at all if the user is using Google
-    // Test with another testing framework.
-    PrintColorEncoded(kColorEncodedHelpMessage);
-  }
-}
-
-// Parses the command line for Google Test flags, without initializing
-// other parts of Google Test.
-void ParseGoogleTestFlagsOnly(int* argc, char** argv) {
-  ParseGoogleTestFlagsOnlyImpl(argc, argv);
-}
-void ParseGoogleTestFlagsOnly(int* argc, wchar_t** argv) {
-  ParseGoogleTestFlagsOnlyImpl(argc, argv);
-}
-
-// The internal implementation of InitGoogleTest().
-//
-// The type parameter CharType can be instantiated to either char or
-// wchar_t.
-template <typename CharType>
-void InitGoogleTestImpl(int* argc, CharType** argv) {
-  // We don't want to run the initialization code twice.
-  if (GTestIsInitialized()) return;
-
-  if (*argc <= 0) return;
-
-  g_argvs.clear();
-  for (int i = 0; i != *argc; i++) {
-    g_argvs.push_back(StreamableToString(argv[i]));
-  }
-
-#if GTEST_HAS_ABSL
-  absl::InitializeSymbolizer(g_argvs[0].c_str());
-#endif  // GTEST_HAS_ABSL
-
-  ParseGoogleTestFlagsOnly(argc, argv);
-  GetUnitTestImpl()->PostFlagParsingInit();
-}
-
-}  // namespace internal
-
-// Initializes Google Test.  This must be called before calling
-// RUN_ALL_TESTS().  In particular, it parses a command line for the
-// flags that Google Test recognizes.  Whenever a Google Test flag is
-// seen, it is removed from argv, and *argc is decremented.
-//
-// No value is returned.  Instead, the Google Test flag variables are
-// updated.
-//
-// Calling the function for the second time has no user-visible effect.
-void InitGoogleTest(int* argc, char** argv) {
-#if defined(GTEST_CUSTOM_INIT_GOOGLE_TEST_FUNCTION_)
-  GTEST_CUSTOM_INIT_GOOGLE_TEST_FUNCTION_(argc, argv);
-#else  // defined(GTEST_CUSTOM_INIT_GOOGLE_TEST_FUNCTION_)
-  internal::InitGoogleTestImpl(argc, argv);
-#endif  // defined(GTEST_CUSTOM_INIT_GOOGLE_TEST_FUNCTION_)
-}
-
-// This overloaded version can be used in Windows programs compiled in
-// UNICODE mode.
-void InitGoogleTest(int* argc, wchar_t** argv) {
-#if defined(GTEST_CUSTOM_INIT_GOOGLE_TEST_FUNCTION_)
-  GTEST_CUSTOM_INIT_GOOGLE_TEST_FUNCTION_(argc, argv);
-#else  // defined(GTEST_CUSTOM_INIT_GOOGLE_TEST_FUNCTION_)
-  internal::InitGoogleTestImpl(argc, argv);
-#endif  // defined(GTEST_CUSTOM_INIT_GOOGLE_TEST_FUNCTION_)
-}
-
-std::string TempDir() {
-#if defined(GTEST_CUSTOM_TEMPDIR_FUNCTION_)
-  return GTEST_CUSTOM_TEMPDIR_FUNCTION_();
-#endif
-
-#if GTEST_OS_WINDOWS_MOBILE
-  return "\\temp\\";
-#elif GTEST_OS_WINDOWS
-  const char* temp_dir = internal::posix::GetEnv("TEMP");
-  if (temp_dir == NULL || temp_dir[0] == '\0')
-    return "\\temp\\";
-  else if (temp_dir[strlen(temp_dir) - 1] == '\\')
-    return temp_dir;
-  else
-    return std::string(temp_dir) + "\\";
-#elif GTEST_OS_LINUX_ANDROID
-  return "/sdcard/";
-#else
-  return "/tmp/";
-#endif  // GTEST_OS_WINDOWS_MOBILE
-}
-
-// Class ScopedTrace
-
-// Pushes the given source file location and message onto a per-thread
-// trace stack maintained by Google Test.
-void ScopedTrace::PushTrace(const char* file, int line, std::string message) {
-  internal::TraceInfo trace;
-  trace.file = file;
-  trace.line = line;
-  trace.message.swap(message);
-
-  UnitTest::GetInstance()->PushGTestTrace(trace);
-}
-
-// Pops the info pushed by the c'tor.
-ScopedTrace::~ScopedTrace()
-    GTEST_LOCK_EXCLUDED_(&UnitTest::mutex_) {
-  UnitTest::GetInstance()->PopGTestTrace();
-}
-
-}  // namespace testing
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan), vladl@google.com (Vlad Losev)
-//
-// This file implements death tests.
-
-
-#if GTEST_HAS_DEATH_TEST
-
-# if GTEST_OS_MAC
-#  include <crt_externs.h>
-# endif  // GTEST_OS_MAC
-
-# include <errno.h>
-# include <fcntl.h>
-# include <limits.h>
-
-# if GTEST_OS_LINUX
-#  include <signal.h>
-# endif  // GTEST_OS_LINUX
-
-# include <stdarg.h>
-
-# if GTEST_OS_WINDOWS
-#  include <windows.h>
-# else
-#  include <sys/mman.h>
-#  include <sys/wait.h>
-# endif  // GTEST_OS_WINDOWS
-
-# if GTEST_OS_QNX
-#  include <spawn.h>
-# endif  // GTEST_OS_QNX
-
-# if GTEST_OS_FUCHSIA
-#  include <lib/fdio/io.h>
-#  include <lib/fdio/spawn.h>
-#  include <zircon/processargs.h>
-#  include <zircon/syscalls.h>
-# endif  // GTEST_OS_FUCHSIA
-
-#endif  // GTEST_HAS_DEATH_TEST
-
-
-namespace testing {
-
-// Constants.
-
-// The default death test style.
-//
-// This is defined in internal/gtest-port.h as "fast", but can be overridden by
-// a definition in internal/custom/gtest-port.h. The recommended value, which is
-// used internally at Google, is "threadsafe".
-static const char kDefaultDeathTestStyle[] = GTEST_DEFAULT_DEATH_TEST_STYLE;
-
-GTEST_DEFINE_string_(
-    death_test_style,
-    internal::StringFromGTestEnv("death_test_style", kDefaultDeathTestStyle),
-    "Indicates how to run a death test in a forked child process: "
-    "\"threadsafe\" (child process re-executes the test binary "
-    "from the beginning, running only the specific death test) or "
-    "\"fast\" (child process runs the death test immediately "
-    "after forking).");
-
-GTEST_DEFINE_bool_(
-    death_test_use_fork,
-    internal::BoolFromGTestEnv("death_test_use_fork", false),
-    "Instructs to use fork()/_exit() instead of clone() in death tests. "
-    "Ignored and always uses fork() on POSIX systems where clone() is not "
-    "implemented. Useful when running under valgrind or similar tools if "
-    "those do not support clone(). Valgrind 3.3.1 will just fail if "
-    "it sees an unsupported combination of clone() flags. "
-    "It is not recommended to use this flag w/o valgrind though it will "
-    "work in 99% of the cases. Once valgrind is fixed, this flag will "
-    "most likely be removed.");
-
-namespace internal {
-GTEST_DEFINE_string_(
-    internal_run_death_test, "",
-    "Indicates the file, line number, temporal index of "
-    "the single death test to run, and a file descriptor to "
-    "which a success code may be sent, all separated by "
-    "the '|' characters.  This flag is specified if and only if the current "
-    "process is a sub-process launched for running a thread-safe "
-    "death test.  FOR INTERNAL USE ONLY.");
-}  // namespace internal
-
-#if GTEST_HAS_DEATH_TEST
-
-namespace internal {
-
-// Valid only for fast death tests. Indicates the code is running in the
-// child process of a fast style death test.
-# if !GTEST_OS_WINDOWS && !GTEST_OS_FUCHSIA
-static bool g_in_fast_death_test_child = false;
-# endif
-
-// Returns a Boolean value indicating whether the caller is currently
-// executing in the context of the death test child process.  Tools such as
-// Valgrind heap checkers may need this to modify their behavior in death
-// tests.  IMPORTANT: This is an internal utility.  Using it may break the
-// implementation of death tests.  User code MUST NOT use it.
-bool InDeathTestChild() {
-# if GTEST_OS_WINDOWS || GTEST_OS_FUCHSIA
-
-  // On Windows and Fuchsia, death tests are thread-safe regardless of the value
-  // of the death_test_style flag.
-  return !GTEST_FLAG(internal_run_death_test).empty();
-
-# else
-
-  if (GTEST_FLAG(death_test_style) == "threadsafe")
-    return !GTEST_FLAG(internal_run_death_test).empty();
-  else
-    return g_in_fast_death_test_child;
-#endif
-}
-
-}  // namespace internal
-
-// ExitedWithCode constructor.
-ExitedWithCode::ExitedWithCode(int exit_code) : exit_code_(exit_code) {
-}
-
-// ExitedWithCode function-call operator.
-bool ExitedWithCode::operator()(int exit_status) const {
-# if GTEST_OS_WINDOWS || GTEST_OS_FUCHSIA
-
-  return exit_status == exit_code_;
-
-# else
-
-  return WIFEXITED(exit_status) && WEXITSTATUS(exit_status) == exit_code_;
-
-# endif  // GTEST_OS_WINDOWS || GTEST_OS_FUCHSIA
-}
-
-# if !GTEST_OS_WINDOWS && !GTEST_OS_FUCHSIA
-// KilledBySignal constructor.
-KilledBySignal::KilledBySignal(int signum) : signum_(signum) {
-}
-
-// KilledBySignal function-call operator.
-bool KilledBySignal::operator()(int exit_status) const {
-#  if defined(GTEST_KILLED_BY_SIGNAL_OVERRIDE_)
-  {
-    bool result;
-    if (GTEST_KILLED_BY_SIGNAL_OVERRIDE_(signum_, exit_status, &result)) {
-      return result;
-    }
-  }
-#  endif  // defined(GTEST_KILLED_BY_SIGNAL_OVERRIDE_)
-  return WIFSIGNALED(exit_status) && WTERMSIG(exit_status) == signum_;
-}
-# endif  // !GTEST_OS_WINDOWS && !GTEST_OS_FUCHSIA
-
-namespace internal {
-
-// Utilities needed for death tests.
-
-// Generates a textual description of a given exit code, in the format
-// specified by wait(2).
-static std::string ExitSummary(int exit_code) {
-  Message m;
-
-# if GTEST_OS_WINDOWS || GTEST_OS_FUCHSIA
-
-  m << "Exited with exit status " << exit_code;
-
-# else
-
-  if (WIFEXITED(exit_code)) {
-    m << "Exited with exit status " << WEXITSTATUS(exit_code);
-  } else if (WIFSIGNALED(exit_code)) {
-    m << "Terminated by signal " << WTERMSIG(exit_code);
-  }
-#  ifdef WCOREDUMP
-  if (WCOREDUMP(exit_code)) {
-    m << " (core dumped)";
-  }
-#  endif
-# endif  // GTEST_OS_WINDOWS || GTEST_OS_FUCHSIA
-
-  return m.GetString();
-}
-
-// Returns true if exit_status describes a process that was terminated
-// by a signal, or exited normally with a nonzero exit code.
-bool ExitedUnsuccessfully(int exit_status) {
-  return !ExitedWithCode(0)(exit_status);
-}
-
-# if !GTEST_OS_WINDOWS && !GTEST_OS_FUCHSIA
-// Generates a textual failure message when a death test finds more than
-// one thread running, or cannot determine the number of threads, prior
-// to executing the given statement.  It is the responsibility of the
-// caller not to pass a thread_count of 1.
-static std::string DeathTestThreadWarning(size_t thread_count) {
-  Message msg;
-  msg << "Death tests use fork(), which is unsafe particularly"
-      << " in a threaded context. For this test, " << GTEST_NAME_ << " ";
-  if (thread_count == 0)
-    msg << "couldn't detect the number of threads.";
-  else
-    msg << "detected " << thread_count << " threads.";
-  return msg.GetString();
-}
-# endif  // !GTEST_OS_WINDOWS && !GTEST_OS_FUCHSIA
-
-// Flag characters for reporting a death test that did not die.
-static const char kDeathTestLived = 'L';
-static const char kDeathTestReturned = 'R';
-static const char kDeathTestThrew = 'T';
-static const char kDeathTestInternalError = 'I';
-
-#if GTEST_OS_FUCHSIA
-
-// File descriptor used for the pipe in the child process.
-static const int kFuchsiaReadPipeFd = 3;
-
-#endif
-
-// An enumeration describing all of the possible ways that a death test can
-// conclude.  DIED means that the process died while executing the test
-// code; LIVED means that process lived beyond the end of the test code;
-// RETURNED means that the test statement attempted to execute a return
-// statement, which is not allowed; THREW means that the test statement
-// returned control by throwing an exception.  IN_PROGRESS means the test
-// has not yet concluded.
-// TODO(vladl@google.com): Unify names and possibly values for
-// AbortReason, DeathTestOutcome, and flag characters above.
-enum DeathTestOutcome { IN_PROGRESS, DIED, LIVED, RETURNED, THREW };
-
-// Routine for aborting the program which is safe to call from an
-// exec-style death test child process, in which case the error
-// message is propagated back to the parent process.  Otherwise, the
-// message is simply printed to stderr.  In either case, the program
-// then exits with status 1.
-static void DeathTestAbort(const std::string& message) {
-  // On a POSIX system, this function may be called from a threadsafe-style
-  // death test child process, which operates on a very small stack.  Use
-  // the heap for any additional non-minuscule memory requirements.
-  const InternalRunDeathTestFlag* const flag =
-      GetUnitTestImpl()->internal_run_death_test_flag();
-  if (flag != NULL) {
-    FILE* parent = posix::FDOpen(flag->write_fd(), "w");
-    fputc(kDeathTestInternalError, parent);
-    fprintf(parent, "%s", message.c_str());
-    fflush(parent);
-    _exit(1);
-  } else {
-    fprintf(stderr, "%s", message.c_str());
-    fflush(stderr);
-    posix::Abort();
-  }
-}
-
-// A replacement for CHECK that calls DeathTestAbort if the assertion
-// fails.
-# define GTEST_DEATH_TEST_CHECK_(expression) \
-  do { \
-    if (!::testing::internal::IsTrue(expression)) { \
-      DeathTestAbort( \
-          ::std::string("CHECK failed: File ") + __FILE__ +  ", line " \
-          + ::testing::internal::StreamableToString(__LINE__) + ": " \
-          + #expression); \
-    } \
-  } while (::testing::internal::AlwaysFalse())
-
-// This macro is similar to GTEST_DEATH_TEST_CHECK_, but it is meant for
-// evaluating any system call that fulfills two conditions: it must return
-// -1 on failure, and set errno to EINTR when it is interrupted and
-// should be tried again.  The macro expands to a loop that repeatedly
-// evaluates the expression as long as it evaluates to -1 and sets
-// errno to EINTR.  If the expression evaluates to -1 but errno is
-// something other than EINTR, DeathTestAbort is called.
-# define GTEST_DEATH_TEST_CHECK_SYSCALL_(expression) \
-  do { \
-    int gtest_retval; \
-    do { \
-      gtest_retval = (expression); \
-    } while (gtest_retval == -1 && errno == EINTR); \
-    if (gtest_retval == -1) { \
-      DeathTestAbort( \
-          ::std::string("CHECK failed: File ") + __FILE__ + ", line " \
-          + ::testing::internal::StreamableToString(__LINE__) + ": " \
-          + #expression + " != -1"); \
-    } \
-  } while (::testing::internal::AlwaysFalse())
-
-// Returns the message describing the last system error in errno.
-std::string GetLastErrnoDescription() {
-    return errno == 0 ? "" : posix::StrError(errno);
-}
-
-// This is called from a death test parent process to read a failure
-// message from the death test child process and log it with the FATAL
-// severity. On Windows, the message is read from a pipe handle. On other
-// platforms, it is read from a file descriptor.
-static void FailFromInternalError(int fd) {
-  Message error;
-  char buffer[256];
-  int num_read;
-
-  do {
-    while ((num_read = posix::Read(fd, buffer, 255)) > 0) {
-      buffer[num_read] = '\0';
-      error << buffer;
-    }
-  } while (num_read == -1 && errno == EINTR);
-
-  if (num_read == 0) {
-    GTEST_LOG_(FATAL) << error.GetString();
-  } else {
-    const int last_error = errno;
-    GTEST_LOG_(FATAL) << "Error while reading death test internal: "
-                      << GetLastErrnoDescription() << " [" << last_error << "]";
-  }
-}
-
-// Death test constructor.  Increments the running death test count
-// for the current test.
-DeathTest::DeathTest() {
-  TestInfo* const info = GetUnitTestImpl()->current_test_info();
-  if (info == NULL) {
-    DeathTestAbort("Cannot run a death test outside of a TEST or "
-                   "TEST_F construct");
-  }
-}
-
-// Creates and returns a death test by dispatching to the current
-// death test factory.
-bool DeathTest::Create(const char* statement, const RE* regex,
-                       const char* file, int line, DeathTest** test) {
-  return GetUnitTestImpl()->death_test_factory()->Create(
-      statement, regex, file, line, test);
-}
-
-const char* DeathTest::LastMessage() {
-  return last_death_test_message_.c_str();
-}
-
-void DeathTest::set_last_death_test_message(const std::string& message) {
-  last_death_test_message_ = message;
-}
-
-std::string DeathTest::last_death_test_message_;
-
-// Provides cross platform implementation for some death functionality.
-class DeathTestImpl : public DeathTest {
- protected:
-  DeathTestImpl(const char* a_statement, const RE* a_regex)
-      : statement_(a_statement),
-        regex_(a_regex),
-        spawned_(false),
-        status_(-1),
-        outcome_(IN_PROGRESS),
-        read_fd_(-1),
-        write_fd_(-1) {}
-
-  // read_fd_ is expected to be closed and cleared by a derived class.
-  ~DeathTestImpl() { GTEST_DEATH_TEST_CHECK_(read_fd_ == -1); }
-
-  void Abort(AbortReason reason);
-  virtual bool Passed(bool status_ok);
-
-  const char* statement() const { return statement_; }
-  const RE* regex() const { return regex_; }
-  bool spawned() const { return spawned_; }
-  void set_spawned(bool is_spawned) { spawned_ = is_spawned; }
-  int status() const { return status_; }
-  void set_status(int a_status) { status_ = a_status; }
-  DeathTestOutcome outcome() const { return outcome_; }
-  void set_outcome(DeathTestOutcome an_outcome) { outcome_ = an_outcome; }
-  int read_fd() const { return read_fd_; }
-  void set_read_fd(int fd) { read_fd_ = fd; }
-  int write_fd() const { return write_fd_; }
-  void set_write_fd(int fd) { write_fd_ = fd; }
-
-  // Called in the parent process only. Reads the result code of the death
-  // test child process via a pipe, interprets it to set the outcome_
-  // member, and closes read_fd_.  Outputs diagnostics and terminates in
-  // case of unexpected codes.
-  void ReadAndInterpretStatusByte();
-
- private:
-  // The textual content of the code this object is testing.  This class
-  // doesn't own this string and should not attempt to delete it.
-  const char* const statement_;
-  // The regular expression which test output must match.  DeathTestImpl
-  // doesn't own this object and should not attempt to delete it.
-  const RE* const regex_;
-  // True if the death test child process has been successfully spawned.
-  bool spawned_;
-  // The exit status of the child process.
-  int status_;
-  // How the death test concluded.
-  DeathTestOutcome outcome_;
-  // Descriptor to the read end of the pipe to the child process.  It is
-  // always -1 in the child process.  The child keeps its write end of the
-  // pipe in write_fd_.
-  int read_fd_;
-  // Descriptor to the child's write end of the pipe to the parent process.
-  // It is always -1 in the parent process.  The parent keeps its end of the
-  // pipe in read_fd_.
-  int write_fd_;
-};
-
-// Called in the parent process only. Reads the result code of the death
-// test child process via a pipe, interprets it to set the outcome_
-// member, and closes read_fd_.  Outputs diagnostics and terminates in
-// case of unexpected codes.
-void DeathTestImpl::ReadAndInterpretStatusByte() {
-  char flag;
-  int bytes_read;
-
-  // The read() here blocks until data is available (signifying the
-  // failure of the death test) or until the pipe is closed (signifying
-  // its success), so it's okay to call this in the parent before
-  // the child process has exited.
-  do {
-    bytes_read = posix::Read(read_fd(), &flag, 1);
-  } while (bytes_read == -1 && errno == EINTR);
-
-  if (bytes_read == 0) {
-    set_outcome(DIED);
-  } else if (bytes_read == 1) {
-    switch (flag) {
-      case kDeathTestReturned:
-        set_outcome(RETURNED);
-        break;
-      case kDeathTestThrew:
-        set_outcome(THREW);
-        break;
-      case kDeathTestLived:
-        set_outcome(LIVED);
-        break;
-      case kDeathTestInternalError:
-        FailFromInternalError(read_fd());  // Does not return.
-        break;
-      default:
-        GTEST_LOG_(FATAL) << "Death test child process reported "
-                          << "unexpected status byte ("
-                          << static_cast<unsigned int>(flag) << ")";
-    }
-  } else {
-    GTEST_LOG_(FATAL) << "Read from death test child process failed: "
-                      << GetLastErrnoDescription();
-  }
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(posix::Close(read_fd()));
-  set_read_fd(-1);
-}
-
-// Signals that the death test code which should have exited, didn't.
-// Should be called only in a death test child process.
-// Writes a status byte to the child's status file descriptor, then
-// calls _exit(1).
-void DeathTestImpl::Abort(AbortReason reason) {
-  // The parent process considers the death test to be a failure if
-  // it finds any data in our pipe.  So, here we write a single flag byte
-  // to the pipe, then exit.
-  const char status_ch =
-      reason == TEST_DID_NOT_DIE ? kDeathTestLived :
-      reason == TEST_THREW_EXCEPTION ? kDeathTestThrew : kDeathTestReturned;
-
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(posix::Write(write_fd(), &status_ch, 1));
-  // We are leaking the descriptor here because on some platforms (i.e.,
-  // when built as Windows DLL), destructors of global objects will still
-  // run after calling _exit(). On such systems, write_fd_ will be
-  // indirectly closed from the destructor of UnitTestImpl, causing double
-  // close if it is also closed here. On debug configurations, double close
-  // may assert. As there are no in-process buffers to flush here, we are
-  // relying on the OS to close the descriptor after the process terminates
-  // when the destructors are not run.
-  _exit(1);  // Exits w/o any normal exit hooks (we were supposed to crash)
-}
-
-// Returns an indented copy of stderr output for a death test.
-// This makes distinguishing death test output lines from regular log lines
-// much easier.
-static ::std::string FormatDeathTestOutput(const ::std::string& output) {
-  ::std::string ret;
-  for (size_t at = 0; ; ) {
-    const size_t line_end = output.find('\n', at);
-    ret += "[  DEATH   ] ";
-    if (line_end == ::std::string::npos) {
-      ret += output.substr(at);
-      break;
-    }
-    ret += output.substr(at, line_end + 1 - at);
-    at = line_end + 1;
-  }
-  return ret;
-}
-
-// Assesses the success or failure of a death test, using both private
-// members which have previously been set, and one argument:
-//
-// Private data members:
-//   outcome:  An enumeration describing how the death test
-//             concluded: DIED, LIVED, THREW, or RETURNED.  The death test
-//             fails in the latter three cases.
-//   status:   The exit status of the child process. On *nix, it is in the
-//             in the format specified by wait(2). On Windows, this is the
-//             value supplied to the ExitProcess() API or a numeric code
-//             of the exception that terminated the program.
-//   regex:    A regular expression object to be applied to
-//             the test's captured standard error output; the death test
-//             fails if it does not match.
-//
-// Argument:
-//   status_ok: true if exit_status is acceptable in the context of
-//              this particular death test, which fails if it is false
-//
-// Returns true iff all of the above conditions are met.  Otherwise, the
-// first failing condition, in the order given above, is the one that is
-// reported. Also sets the last death test message string.
-bool DeathTestImpl::Passed(bool status_ok) {
-  if (!spawned())
-    return false;
-
-  const std::string error_message = GetCapturedStderr();
-
-  bool success = false;
-  Message buffer;
-
-  buffer << "Death test: " << statement() << "\n";
-  switch (outcome()) {
-    case LIVED:
-      buffer << "    Result: failed to die.\n"
-             << " Error msg:\n" << FormatDeathTestOutput(error_message);
-      break;
-    case THREW:
-      buffer << "    Result: threw an exception.\n"
-             << " Error msg:\n" << FormatDeathTestOutput(error_message);
-      break;
-    case RETURNED:
-      buffer << "    Result: illegal return in test statement.\n"
-             << " Error msg:\n" << FormatDeathTestOutput(error_message);
-      break;
-    case DIED:
-      if (status_ok) {
-# if GTEST_USES_PCRE
-        // PCRE regexes support embedded NULs.
-        // GTEST_USES_PCRE is defined only in google3 mode
-        const bool matched = RE::PartialMatch(error_message, *regex());
-# else
-        const bool matched = RE::PartialMatch(error_message.c_str(), *regex());
-# endif  // GTEST_USES_PCRE
-        if (matched) {
-          success = true;
-        } else {
-          buffer << "    Result: died but not with expected error.\n"
-                 << "  Expected: " << regex()->pattern() << "\n"
-                 << "Actual msg:\n" << FormatDeathTestOutput(error_message);
-        }
-      } else {
-        buffer << "    Result: died but not with expected exit code:\n"
-               << "            " << ExitSummary(status()) << "\n"
-               << "Actual msg:\n" << FormatDeathTestOutput(error_message);
-      }
-      break;
-    case IN_PROGRESS:
-    default:
-      GTEST_LOG_(FATAL)
-          << "DeathTest::Passed somehow called before conclusion of test";
-  }
-
-  DeathTest::set_last_death_test_message(buffer.GetString());
-  return success;
-}
-
-# if GTEST_OS_WINDOWS
-// WindowsDeathTest implements death tests on Windows. Due to the
-// specifics of starting new processes on Windows, death tests there are
-// always threadsafe, and Google Test considers the
-// --gtest_death_test_style=fast setting to be equivalent to
-// --gtest_death_test_style=threadsafe there.
-//
-// A few implementation notes:  Like the Linux version, the Windows
-// implementation uses pipes for child-to-parent communication. But due to
-// the specifics of pipes on Windows, some extra steps are required:
-//
-// 1. The parent creates a communication pipe and stores handles to both
-//    ends of it.
-// 2. The parent starts the child and provides it with the information
-//    necessary to acquire the handle to the write end of the pipe.
-// 3. The child acquires the write end of the pipe and signals the parent
-//    using a Windows event.
-// 4. Now the parent can release the write end of the pipe on its side. If
-//    this is done before step 3, the object's reference count goes down to
-//    0 and it is destroyed, preventing the child from acquiring it. The
-//    parent now has to release it, or read operations on the read end of
-//    the pipe will not return when the child terminates.
-// 5. The parent reads child's output through the pipe (outcome code and
-//    any possible error messages) from the pipe, and its stderr and then
-//    determines whether to fail the test.
-//
-// Note: to distinguish Win32 API calls from the local method and function
-// calls, the former are explicitly resolved in the global namespace.
-//
-class WindowsDeathTest : public DeathTestImpl {
- public:
-  WindowsDeathTest(const char* a_statement,
-                   const RE* a_regex,
-                   const char* file,
-                   int line)
-      : DeathTestImpl(a_statement, a_regex), file_(file), line_(line) {}
-
-  // All of these virtual functions are inherited from DeathTest.
-  virtual int Wait();
-  virtual TestRole AssumeRole();
-
- private:
-  // The name of the file in which the death test is located.
-  const char* const file_;
-  // The line number on which the death test is located.
-  const int line_;
-  // Handle to the write end of the pipe to the child process.
-  AutoHandle write_handle_;
-  // Child process handle.
-  AutoHandle child_handle_;
-  // Event the child process uses to signal the parent that it has
-  // acquired the handle to the write end of the pipe. After seeing this
-  // event the parent can release its own handles to make sure its
-  // ReadFile() calls return when the child terminates.
-  AutoHandle event_handle_;
-};
-
-// Waits for the child in a death test to exit, returning its exit
-// status, or 0 if no child process exists.  As a side effect, sets the
-// outcome data member.
-int WindowsDeathTest::Wait() {
-  if (!spawned())
-    return 0;
-
-  // Wait until the child either signals that it has acquired the write end
-  // of the pipe or it dies.
-  const HANDLE wait_handles[2] = { child_handle_.Get(), event_handle_.Get() };
-  switch (::WaitForMultipleObjects(2,
-                                   wait_handles,
-                                   FALSE,  // Waits for any of the handles.
-                                   INFINITE)) {
-    case WAIT_OBJECT_0:
-    case WAIT_OBJECT_0 + 1:
-      break;
-    default:
-      GTEST_DEATH_TEST_CHECK_(false);  // Should not get here.
-  }
-
-  // The child has acquired the write end of the pipe or exited.
-  // We release the handle on our side and continue.
-  write_handle_.Reset();
-  event_handle_.Reset();
-
-  ReadAndInterpretStatusByte();
-
-  // Waits for the child process to exit if it haven't already. This
-  // returns immediately if the child has already exited, regardless of
-  // whether previous calls to WaitForMultipleObjects synchronized on this
-  // handle or not.
-  GTEST_DEATH_TEST_CHECK_(
-      WAIT_OBJECT_0 == ::WaitForSingleObject(child_handle_.Get(),
-                                             INFINITE));
-  DWORD status_code;
-  GTEST_DEATH_TEST_CHECK_(
-      ::GetExitCodeProcess(child_handle_.Get(), &status_code) != FALSE);
-  child_handle_.Reset();
-  set_status(static_cast<int>(status_code));
-  return status();
-}
-
-// The AssumeRole process for a Windows death test.  It creates a child
-// process with the same executable as the current process to run the
-// death test.  The child process is given the --gtest_filter and
-// --gtest_internal_run_death_test flags such that it knows to run the
-// current death test only.
-DeathTest::TestRole WindowsDeathTest::AssumeRole() {
-  const UnitTestImpl* const impl = GetUnitTestImpl();
-  const InternalRunDeathTestFlag* const flag =
-      impl->internal_run_death_test_flag();
-  const TestInfo* const info = impl->current_test_info();
-  const int death_test_index = info->result()->death_test_count();
-
-  if (flag != NULL) {
-    // ParseInternalRunDeathTestFlag() has performed all the necessary
-    // processing.
-    set_write_fd(flag->write_fd());
-    return EXECUTE_TEST;
-  }
-
-  // WindowsDeathTest uses an anonymous pipe to communicate results of
-  // a death test.
-  SECURITY_ATTRIBUTES handles_are_inheritable = {
-    sizeof(SECURITY_ATTRIBUTES), NULL, TRUE };
-  HANDLE read_handle, write_handle;
-  GTEST_DEATH_TEST_CHECK_(
-      ::CreatePipe(&read_handle, &write_handle, &handles_are_inheritable,
-                   0)  // Default buffer size.
-      != FALSE);
-  set_read_fd(::_open_osfhandle(reinterpret_cast<intptr_t>(read_handle),
-                                O_RDONLY));
-  write_handle_.Reset(write_handle);
-  event_handle_.Reset(::CreateEvent(
-      &handles_are_inheritable,
-      TRUE,    // The event will automatically reset to non-signaled state.
-      FALSE,   // The initial state is non-signalled.
-      NULL));  // The even is unnamed.
-  GTEST_DEATH_TEST_CHECK_(event_handle_.Get() != NULL);
-  const std::string filter_flag =
-      std::string("--") + GTEST_FLAG_PREFIX_ + kFilterFlag + "=" +
-      info->test_case_name() + "." + info->name();
-  const std::string internal_flag =
-      std::string("--") + GTEST_FLAG_PREFIX_ + kInternalRunDeathTestFlag +
-      "=" + file_ + "|" + StreamableToString(line_) + "|" +
-      StreamableToString(death_test_index) + "|" +
-      StreamableToString(static_cast<unsigned int>(::GetCurrentProcessId())) +
-      // size_t has the same width as pointers on both 32-bit and 64-bit
-      // Windows platforms.
-      // See http://msdn.microsoft.com/en-us/library/tcxf1dw6.aspx.
-      "|" + StreamableToString(reinterpret_cast<size_t>(write_handle)) +
-      "|" + StreamableToString(reinterpret_cast<size_t>(event_handle_.Get()));
-
-  char executable_path[_MAX_PATH + 1];  // NOLINT
-  GTEST_DEATH_TEST_CHECK_(
-      _MAX_PATH + 1 != ::GetModuleFileNameA(NULL,
-                                            executable_path,
-                                            _MAX_PATH));
-
-  std::string command_line =
-      std::string(::GetCommandLineA()) + " " + filter_flag + " \"" +
-      internal_flag + "\"";
-
-  DeathTest::set_last_death_test_message("");
-
-  CaptureStderr();
-  // Flush the log buffers since the log streams are shared with the child.
-  FlushInfoLog();
-
-  // The child process will share the standard handles with the parent.
-  STARTUPINFOA startup_info;
-  memset(&startup_info, 0, sizeof(STARTUPINFO));
-  startup_info.dwFlags = STARTF_USESTDHANDLES;
-  startup_info.hStdInput = ::GetStdHandle(STD_INPUT_HANDLE);
-  startup_info.hStdOutput = ::GetStdHandle(STD_OUTPUT_HANDLE);
-  startup_info.hStdError = ::GetStdHandle(STD_ERROR_HANDLE);
-
-  PROCESS_INFORMATION process_info;
-  GTEST_DEATH_TEST_CHECK_(::CreateProcessA(
-      executable_path,
-      const_cast<char*>(command_line.c_str()),
-      NULL,   // Retuned process handle is not inheritable.
-      NULL,   // Retuned thread handle is not inheritable.
-      TRUE,   // Child inherits all inheritable handles (for write_handle_).
-      0x0,    // Default creation flags.
-      NULL,   // Inherit the parent's environment.
-      UnitTest::GetInstance()->original_working_dir(),
-      &startup_info,
-      &process_info) != FALSE);
-  child_handle_.Reset(process_info.hProcess);
-  ::CloseHandle(process_info.hThread);
-  set_spawned(true);
-  return OVERSEE_TEST;
-}
-
-# elif GTEST_OS_FUCHSIA
-
-class FuchsiaDeathTest : public DeathTestImpl {
- public:
-  FuchsiaDeathTest(const char* a_statement,
-                   const RE* a_regex,
-                   const char* file,
-                   int line)
-      : DeathTestImpl(a_statement, a_regex), file_(file), line_(line) {}
-
-  // All of these virtual functions are inherited from DeathTest.
-  virtual int Wait();
-  virtual TestRole AssumeRole();
-
- private:
-  // The name of the file in which the death test is located.
-  const char* const file_;
-  // The line number on which the death test is located.
-  const int line_;
-
-  zx_handle_t child_process_;
-};
-
-// Utility class for accumulating command-line arguments.
-class Arguments {
- public:
-  Arguments() {
-    args_.push_back(NULL);
-  }
-
-  ~Arguments() {
-    for (std::vector<char*>::iterator i = args_.begin(); i != args_.end();
-         ++i) {
-      free(*i);
-    }
-  }
-  void AddArgument(const char* argument) {
-    args_.insert(args_.end() - 1, posix::StrDup(argument));
-  }
-
-  template <typename Str>
-  void AddArguments(const ::std::vector<Str>& arguments) {
-    for (typename ::std::vector<Str>::const_iterator i = arguments.begin();
-         i != arguments.end();
-         ++i) {
-      args_.insert(args_.end() - 1, posix::StrDup(i->c_str()));
-    }
-  }
-  char* const* Argv() {
-    return &args_[0];
-  }
-
-  int size() {
-    return args_.size() - 1;
-  }
-
- private:
-  std::vector<char*> args_;
-};
-
-// Waits for the child in a death test to exit, returning its exit
-// status, or 0 if no child process exists.  As a side effect, sets the
-// outcome data member.
-int FuchsiaDeathTest::Wait() {
-  if (!spawned())
-    return 0;
-
-  // Wait for child process to terminate.
-  zx_status_t status_zx;
-  zx_signals_t signals;
-  status_zx = zx_object_wait_one(
-      child_process_,
-      ZX_PROCESS_TERMINATED,
-      ZX_TIME_INFINITE,
-      &signals);
-  GTEST_DEATH_TEST_CHECK_(status_zx == ZX_OK);
-
-  ReadAndInterpretStatusByte();
-
-  zx_info_process_t buffer;
-  status_zx = zx_object_get_info(
-      child_process_,
-      ZX_INFO_PROCESS,
-      &buffer,
-      sizeof(buffer),
-      nullptr,
-      nullptr);
-  GTEST_DEATH_TEST_CHECK_(status_zx == ZX_OK);
-
-  GTEST_DEATH_TEST_CHECK_(buffer.exited);
-  set_status(buffer.return_code);
-  return status();
-}
-
-// The AssumeRole process for a Fuchsia death test.  It creates a child
-// process with the same executable as the current process to run the
-// death test.  The child process is given the --gtest_filter and
-// --gtest_internal_run_death_test flags such that it knows to run the
-// current death test only.
-DeathTest::TestRole FuchsiaDeathTest::AssumeRole() {
-  const UnitTestImpl* const impl = GetUnitTestImpl();
-  const InternalRunDeathTestFlag* const flag =
-      impl->internal_run_death_test_flag();
-  const TestInfo* const info = impl->current_test_info();
-  const int death_test_index = info->result()->death_test_count();
-
-  if (flag != NULL) {
-    // ParseInternalRunDeathTestFlag() has performed all the necessary
-    // processing.
-    set_write_fd(kFuchsiaReadPipeFd);
-    return EXECUTE_TEST;
-  }
-
-  CaptureStderr();
-  // Flush the log buffers since the log streams are shared with the child.
-  FlushInfoLog();
-
-  // Build the child process command line.
-  const std::string filter_flag =
-      std::string("--") + GTEST_FLAG_PREFIX_ + kFilterFlag + "="
-      + info->test_case_name() + "." + info->name();
-  const std::string internal_flag =
-      std::string("--") + GTEST_FLAG_PREFIX_ + kInternalRunDeathTestFlag + "="
-      + file_ + "|"
-      + StreamableToString(line_) + "|"
-      + StreamableToString(death_test_index);
-  Arguments args;
-  args.AddArguments(GetInjectableArgvs());
-  args.AddArgument(filter_flag.c_str());
-  args.AddArgument(internal_flag.c_str());
-
-  // Build the pipe for communication with the child.
-  zx_status_t status;
-  zx_handle_t child_pipe_handle;
-  uint32_t type;
-  status = fdio_pipe_half(&child_pipe_handle, &type);
-  GTEST_DEATH_TEST_CHECK_(status >= 0);
-  set_read_fd(status);
-
-  // Set the pipe handle for the child.
-  fdio_spawn_action_t add_handle_action = {
-    .action = FDIO_SPAWN_ACTION_ADD_HANDLE,
-    .h = {
-      .id = PA_HND(type, kFuchsiaReadPipeFd),
-      .handle = child_pipe_handle
-    }
-  };
-
-  // Spawn the child process.
-  status = fdio_spawn_etc(ZX_HANDLE_INVALID, FDIO_SPAWN_CLONE_ALL,
-                          args.Argv()[0], args.Argv(), nullptr, 1,
-                          &add_handle_action, &child_process_, nullptr);
-  GTEST_DEATH_TEST_CHECK_(status == ZX_OK);
-
-  set_spawned(true);
-  return OVERSEE_TEST;
-}
-
-#else  // We are neither on Windows, nor on Fuchsia.
-
-// ForkingDeathTest provides implementations for most of the abstract
-// methods of the DeathTest interface.  Only the AssumeRole method is
-// left undefined.
-class ForkingDeathTest : public DeathTestImpl {
- public:
-  ForkingDeathTest(const char* statement, const RE* regex);
-
-  // All of these virtual functions are inherited from DeathTest.
-  virtual int Wait();
-
- protected:
-  void set_child_pid(pid_t child_pid) { child_pid_ = child_pid; }
-
- private:
-  // PID of child process during death test; 0 in the child process itself.
-  pid_t child_pid_;
-};
-
-// Constructs a ForkingDeathTest.
-ForkingDeathTest::ForkingDeathTest(const char* a_statement, const RE* a_regex)
-    : DeathTestImpl(a_statement, a_regex),
-      child_pid_(-1) {}
-
-// Waits for the child in a death test to exit, returning its exit
-// status, or 0 if no child process exists.  As a side effect, sets the
-// outcome data member.
-int ForkingDeathTest::Wait() {
-  if (!spawned())
-    return 0;
-
-  ReadAndInterpretStatusByte();
-
-  int status_value;
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(waitpid(child_pid_, &status_value, 0));
-  set_status(status_value);
-  return status_value;
-}
-
-// A concrete death test class that forks, then immediately runs the test
-// in the child process.
-class NoExecDeathTest : public ForkingDeathTest {
- public:
-  NoExecDeathTest(const char* a_statement, const RE* a_regex) :
-      ForkingDeathTest(a_statement, a_regex) { }
-  virtual TestRole AssumeRole();
-};
-
-// The AssumeRole process for a fork-and-run death test.  It implements a
-// straightforward fork, with a simple pipe to transmit the status byte.
-DeathTest::TestRole NoExecDeathTest::AssumeRole() {
-  const size_t thread_count = GetThreadCount();
-  if (thread_count != 1) {
-    GTEST_LOG_(WARNING) << DeathTestThreadWarning(thread_count);
-  }
-
-  int pipe_fd[2];
-  GTEST_DEATH_TEST_CHECK_(pipe(pipe_fd) != -1);
-
-  DeathTest::set_last_death_test_message("");
-  CaptureStderr();
-  // When we fork the process below, the log file buffers are copied, but the
-  // file descriptors are shared.  We flush all log files here so that closing
-  // the file descriptors in the child process doesn't throw off the
-  // synchronization between descriptors and buffers in the parent process.
-  // This is as close to the fork as possible to avoid a race condition in case
-  // there are multiple threads running before the death test, and another
-  // thread writes to the log file.
-  FlushInfoLog();
-
-  const pid_t child_pid = fork();
-  GTEST_DEATH_TEST_CHECK_(child_pid != -1);
-  set_child_pid(child_pid);
-  if (child_pid == 0) {
-    GTEST_DEATH_TEST_CHECK_SYSCALL_(close(pipe_fd[0]));
-    set_write_fd(pipe_fd[1]);
-    // Redirects all logging to stderr in the child process to prevent
-    // concurrent writes to the log files.  We capture stderr in the parent
-    // process and append the child process' output to a log.
-    LogToStderr();
-    // Event forwarding to the listeners of event listener API mush be shut
-    // down in death test subprocesses.
-    GetUnitTestImpl()->listeners()->SuppressEventForwarding();
-    g_in_fast_death_test_child = true;
-    return EXECUTE_TEST;
-  } else {
-    GTEST_DEATH_TEST_CHECK_SYSCALL_(close(pipe_fd[1]));
-    set_read_fd(pipe_fd[0]);
-    set_spawned(true);
-    return OVERSEE_TEST;
-  }
-}
-
-// A concrete death test class that forks and re-executes the main
-// program from the beginning, with command-line flags set that cause
-// only this specific death test to be run.
-class ExecDeathTest : public ForkingDeathTest {
- public:
-  ExecDeathTest(const char* a_statement, const RE* a_regex,
-                const char* file, int line) :
-      ForkingDeathTest(a_statement, a_regex), file_(file), line_(line) { }
-  virtual TestRole AssumeRole();
- private:
-  static ::std::vector<std::string> GetArgvsForDeathTestChildProcess() {
-    ::std::vector<std::string> args = GetInjectableArgvs();
-#  if defined(GTEST_EXTRA_DEATH_TEST_COMMAND_LINE_ARGS_)
-    ::std::vector<std::string> extra_args =
-        GTEST_EXTRA_DEATH_TEST_COMMAND_LINE_ARGS_();
-    args.insert(args.end(), extra_args.begin(), extra_args.end());
-#  endif  // defined(GTEST_EXTRA_DEATH_TEST_COMMAND_LINE_ARGS_)
-    return args;
-  }
-  // The name of the file in which the death test is located.
-  const char* const file_;
-  // The line number on which the death test is located.
-  const int line_;
-};
-
-// Utility class for accumulating command-line arguments.
-class Arguments {
- public:
-  Arguments() {
-    args_.push_back(NULL);
-  }
-
-  ~Arguments() {
-    for (std::vector<char*>::iterator i = args_.begin(); i != args_.end();
-         ++i) {
-      free(*i);
-    }
-  }
-  void AddArgument(const char* argument) {
-    args_.insert(args_.end() - 1, posix::StrDup(argument));
-  }
-
-  template <typename Str>
-  void AddArguments(const ::std::vector<Str>& arguments) {
-    for (typename ::std::vector<Str>::const_iterator i = arguments.begin();
-         i != arguments.end();
-         ++i) {
-      args_.insert(args_.end() - 1, posix::StrDup(i->c_str()));
-    }
-  }
-  char* const* Argv() {
-    return &args_[0];
-  }
-
- private:
-  std::vector<char*> args_;
-};
-
-// A struct that encompasses the arguments to the child process of a
-// threadsafe-style death test process.
-struct ExecDeathTestArgs {
-  char* const* argv;  // Command-line arguments for the child's call to exec
-  int close_fd;       // File descriptor to close; the read end of a pipe
-};
-
-#  if GTEST_OS_MAC
-inline char** GetEnviron() {
-  // When Google Test is built as a framework on MacOS X, the environ variable
-  // is unavailable. Apple's documentation (man environ) recommends using
-  // _NSGetEnviron() instead.
-  return *_NSGetEnviron();
-}
-#  else
-// Some POSIX platforms expect you to declare environ. extern "C" makes
-// it reside in the global namespace.
-extern "C" char** environ;
-inline char** GetEnviron() { return environ; }
-#  endif  // GTEST_OS_MAC
-
-#  if !GTEST_OS_QNX
-// The main function for a threadsafe-style death test child process.
-// This function is called in a clone()-ed process and thus must avoid
-// any potentially unsafe operations like malloc or libc functions.
-static int ExecDeathTestChildMain(void* child_arg) {
-  ExecDeathTestArgs* const args = static_cast<ExecDeathTestArgs*>(child_arg);
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(close(args->close_fd));
-
-  // We need to execute the test program in the same environment where
-  // it was originally invoked.  Therefore we change to the original
-  // working directory first.
-  const char* const original_dir =
-      UnitTest::GetInstance()->original_working_dir();
-  // We can safely call chdir() as it's a direct system call.
-  if (chdir(original_dir) != 0) {
-    DeathTestAbort(std::string("chdir(\"") + original_dir + "\") failed: " +
-                   GetLastErrnoDescription());
-    return EXIT_FAILURE;
-  }
-
-  // We can safely call execve() as it's a direct system call.  We
-  // cannot use execvp() as it's a libc function and thus potentially
-  // unsafe.  Since execve() doesn't search the PATH, the user must
-  // invoke the test program via a valid path that contains at least
-  // one path separator.
-  execve(args->argv[0], args->argv, GetEnviron());
-  DeathTestAbort(std::string("execve(") + args->argv[0] + ", ...) in " +
-                 original_dir + " failed: " +
-                 GetLastErrnoDescription());
-  return EXIT_FAILURE;
-}
-#  endif  // !GTEST_OS_QNX
-
-#  if GTEST_HAS_CLONE
-// Two utility routines that together determine the direction the stack
-// grows.
-// This could be accomplished more elegantly by a single recursive
-// function, but we want to guard against the unlikely possibility of
-// a smart compiler optimizing the recursion away.
-//
-// GTEST_NO_INLINE_ is required to prevent GCC 4.6 from inlining
-// StackLowerThanAddress into StackGrowsDown, which then doesn't give
-// correct answer.
-static void StackLowerThanAddress(const void* ptr,
-                                  bool* result) GTEST_NO_INLINE_;
-static void StackLowerThanAddress(const void* ptr, bool* result) {
-  int dummy;
-  *result = (&dummy < ptr);
-}
-
-// Make sure AddressSanitizer does not tamper with the stack here.
-GTEST_ATTRIBUTE_NO_SANITIZE_ADDRESS_
-static bool StackGrowsDown() {
-  int dummy;
-  bool result;
-  StackLowerThanAddress(&dummy, &result);
-  return result;
-}
-#  endif  // GTEST_HAS_CLONE
-
-// Spawns a child process with the same executable as the current process in
-// a thread-safe manner and instructs it to run the death test.  The
-// implementation uses fork(2) + exec.  On systems where clone(2) is
-// available, it is used instead, being slightly more thread-safe.  On QNX,
-// fork supports only single-threaded environments, so this function uses
-// spawn(2) there instead.  The function dies with an error message if
-// anything goes wrong.
-static pid_t ExecDeathTestSpawnChild(char* const* argv, int close_fd) {
-  ExecDeathTestArgs args = { argv, close_fd };
-  pid_t child_pid = -1;
-
-#  if GTEST_OS_QNX
-  // Obtains the current directory and sets it to be closed in the child
-  // process.
-  const int cwd_fd = open(".", O_RDONLY);
-  GTEST_DEATH_TEST_CHECK_(cwd_fd != -1);
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(fcntl(cwd_fd, F_SETFD, FD_CLOEXEC));
-  // We need to execute the test program in the same environment where
-  // it was originally invoked.  Therefore we change to the original
-  // working directory first.
-  const char* const original_dir =
-      UnitTest::GetInstance()->original_working_dir();
-  // We can safely call chdir() as it's a direct system call.
-  if (chdir(original_dir) != 0) {
-    DeathTestAbort(std::string("chdir(\"") + original_dir + "\") failed: " +
-                   GetLastErrnoDescription());
-    return EXIT_FAILURE;
-  }
-
-  int fd_flags;
-  // Set close_fd to be closed after spawn.
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(fd_flags = fcntl(close_fd, F_GETFD));
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(fcntl(close_fd, F_SETFD,
-                                        fd_flags | FD_CLOEXEC));
-  struct inheritance inherit = {0};
-  // spawn is a system call.
-  child_pid = spawn(args.argv[0], 0, NULL, &inherit, args.argv, GetEnviron());
-  // Restores the current working directory.
-  GTEST_DEATH_TEST_CHECK_(fchdir(cwd_fd) != -1);
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(close(cwd_fd));
-
-#  else   // GTEST_OS_QNX
-#   if GTEST_OS_LINUX
-  // When a SIGPROF signal is received while fork() or clone() are executing,
-  // the process may hang. To avoid this, we ignore SIGPROF here and re-enable
-  // it after the call to fork()/clone() is complete.
-  struct sigaction saved_sigprof_action;
-  struct sigaction ignore_sigprof_action;
-  memset(&ignore_sigprof_action, 0, sizeof(ignore_sigprof_action));
-  sigemptyset(&ignore_sigprof_action.sa_mask);
-  ignore_sigprof_action.sa_handler = SIG_IGN;
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(sigaction(
-      SIGPROF, &ignore_sigprof_action, &saved_sigprof_action));
-#   endif  // GTEST_OS_LINUX
-
-#   if GTEST_HAS_CLONE
-  const bool use_fork = GTEST_FLAG(death_test_use_fork);
-
-  if (!use_fork) {
-    static const bool stack_grows_down = StackGrowsDown();
-    const size_t stack_size = getpagesize();
-    // MMAP_ANONYMOUS is not defined on Mac, so we use MAP_ANON instead.
-    void* const stack = mmap(NULL, stack_size, PROT_READ | PROT_WRITE,
-                             MAP_ANON | MAP_PRIVATE, -1, 0);
-    GTEST_DEATH_TEST_CHECK_(stack != MAP_FAILED);
-
-    // Maximum stack alignment in bytes:  For a downward-growing stack, this
-    // amount is subtracted from size of the stack space to get an address
-    // that is within the stack space and is aligned on all systems we care
-    // about.  As far as I know there is no ABI with stack alignment greater
-    // than 64.  We assume stack and stack_size already have alignment of
-    // kMaxStackAlignment.
-    const size_t kMaxStackAlignment = 64;
-    void* const stack_top =
-        static_cast<char*>(stack) +
-            (stack_grows_down ? stack_size - kMaxStackAlignment : 0);
-    GTEST_DEATH_TEST_CHECK_(stack_size > kMaxStackAlignment &&
-        reinterpret_cast<intptr_t>(stack_top) % kMaxStackAlignment == 0);
-
-    child_pid = clone(&ExecDeathTestChildMain, stack_top, SIGCHLD, &args);
-
-    GTEST_DEATH_TEST_CHECK_(munmap(stack, stack_size) != -1);
-  }
-#   else
-  const bool use_fork = true;
-#   endif  // GTEST_HAS_CLONE
-
-  if (use_fork && (child_pid = fork()) == 0) {
-      ExecDeathTestChildMain(&args);
-      _exit(0);
-  }
-#  endif  // GTEST_OS_QNX
-#  if GTEST_OS_LINUX
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(
-      sigaction(SIGPROF, &saved_sigprof_action, NULL));
-#  endif  // GTEST_OS_LINUX
-
-  GTEST_DEATH_TEST_CHECK_(child_pid != -1);
-  return child_pid;
-}
-
-// The AssumeRole process for a fork-and-exec death test.  It re-executes the
-// main program from the beginning, setting the --gtest_filter
-// and --gtest_internal_run_death_test flags to cause only the current
-// death test to be re-run.
-DeathTest::TestRole ExecDeathTest::AssumeRole() {
-  const UnitTestImpl* const impl = GetUnitTestImpl();
-  const InternalRunDeathTestFlag* const flag =
-      impl->internal_run_death_test_flag();
-  const TestInfo* const info = impl->current_test_info();
-  const int death_test_index = info->result()->death_test_count();
-
-  if (flag != NULL) {
-    set_write_fd(flag->write_fd());
-    return EXECUTE_TEST;
-  }
-
-  int pipe_fd[2];
-  GTEST_DEATH_TEST_CHECK_(pipe(pipe_fd) != -1);
-  // Clear the close-on-exec flag on the write end of the pipe, lest
-  // it be closed when the child process does an exec:
-  GTEST_DEATH_TEST_CHECK_(fcntl(pipe_fd[1], F_SETFD, 0) != -1);
-
-  const std::string filter_flag =
-      std::string("--") + GTEST_FLAG_PREFIX_ + kFilterFlag + "="
-      + info->test_case_name() + "." + info->name();
-  const std::string internal_flag =
-      std::string("--") + GTEST_FLAG_PREFIX_ + kInternalRunDeathTestFlag + "="
-      + file_ + "|" + StreamableToString(line_) + "|"
-      + StreamableToString(death_test_index) + "|"
-      + StreamableToString(pipe_fd[1]);
-  Arguments args;
-  args.AddArguments(GetArgvsForDeathTestChildProcess());
-  args.AddArgument(filter_flag.c_str());
-  args.AddArgument(internal_flag.c_str());
-
-  DeathTest::set_last_death_test_message("");
-
-  CaptureStderr();
-  // See the comment in NoExecDeathTest::AssumeRole for why the next line
-  // is necessary.
-  FlushInfoLog();
-
-  const pid_t child_pid = ExecDeathTestSpawnChild(args.Argv(), pipe_fd[0]);
-  GTEST_DEATH_TEST_CHECK_SYSCALL_(close(pipe_fd[1]));
-  set_child_pid(child_pid);
-  set_read_fd(pipe_fd[0]);
-  set_spawned(true);
-  return OVERSEE_TEST;
-}
-
-# endif  // !GTEST_OS_WINDOWS
-
-// Creates a concrete DeathTest-derived class that depends on the
-// --gtest_death_test_style flag, and sets the pointer pointed to
-// by the "test" argument to its address.  If the test should be
-// skipped, sets that pointer to NULL.  Returns true, unless the
-// flag is set to an invalid value.
-bool DefaultDeathTestFactory::Create(const char* statement, const RE* regex,
-                                     const char* file, int line,
-                                     DeathTest** test) {
-  UnitTestImpl* const impl = GetUnitTestImpl();
-  const InternalRunDeathTestFlag* const flag =
-      impl->internal_run_death_test_flag();
-  const int death_test_index = impl->current_test_info()
-      ->increment_death_test_count();
-
-  if (flag != NULL) {
-    if (death_test_index > flag->index()) {
-      DeathTest::set_last_death_test_message(
-          "Death test count (" + StreamableToString(death_test_index)
-          + ") somehow exceeded expected maximum ("
-          + StreamableToString(flag->index()) + ")");
-      return false;
-    }
-
-    if (!(flag->file() == file && flag->line() == line &&
-          flag->index() == death_test_index)) {
-      *test = NULL;
-      return true;
-    }
-  }
-
-# if GTEST_OS_WINDOWS
-
-  if (GTEST_FLAG(death_test_style) == "threadsafe" ||
-      GTEST_FLAG(death_test_style) == "fast") {
-    *test = new WindowsDeathTest(statement, regex, file, line);
-  }
-
-# elif GTEST_OS_FUCHSIA
-
-  if (GTEST_FLAG(death_test_style) == "threadsafe" ||
-      GTEST_FLAG(death_test_style) == "fast") {
-    *test = new FuchsiaDeathTest(statement, regex, file, line);
-  }
-
-# else
-
-  if (GTEST_FLAG(death_test_style) == "threadsafe") {
-    *test = new ExecDeathTest(statement, regex, file, line);
-  } else if (GTEST_FLAG(death_test_style) == "fast") {
-    *test = new NoExecDeathTest(statement, regex);
-  }
-
-# endif  // GTEST_OS_WINDOWS
-
-  else {  // NOLINT - this is more readable than unbalanced brackets inside #if.
-    DeathTest::set_last_death_test_message(
-        "Unknown death test style \"" + GTEST_FLAG(death_test_style)
-        + "\" encountered");
-    return false;
-  }
-
-  return true;
-}
-
-# if GTEST_OS_WINDOWS
-// Recreates the pipe and event handles from the provided parameters,
-// signals the event, and returns a file descriptor wrapped around the pipe
-// handle. This function is called in the child process only.
-static int GetStatusFileDescriptor(unsigned int parent_process_id,
-                            size_t write_handle_as_size_t,
-                            size_t event_handle_as_size_t) {
-  AutoHandle parent_process_handle(::OpenProcess(PROCESS_DUP_HANDLE,
-                                                   FALSE,  // Non-inheritable.
-                                                   parent_process_id));
-  if (parent_process_handle.Get() == INVALID_HANDLE_VALUE) {
-    DeathTestAbort("Unable to open parent process " +
-                   StreamableToString(parent_process_id));
-  }
-
-  // TODO(vladl@google.com): Replace the following check with a
-  // compile-time assertion when available.
-  GTEST_CHECK_(sizeof(HANDLE) <= sizeof(size_t));
-
-  const HANDLE write_handle =
-      reinterpret_cast<HANDLE>(write_handle_as_size_t);
-  HANDLE dup_write_handle;
-
-  // The newly initialized handle is accessible only in the parent
-  // process. To obtain one accessible within the child, we need to use
-  // DuplicateHandle.
-  if (!::DuplicateHandle(parent_process_handle.Get(), write_handle,
-                         ::GetCurrentProcess(), &dup_write_handle,
-                         0x0,    // Requested privileges ignored since
-                                 // DUPLICATE_SAME_ACCESS is used.
-                         FALSE,  // Request non-inheritable handler.
-                         DUPLICATE_SAME_ACCESS)) {
-    DeathTestAbort("Unable to duplicate the pipe handle " +
-                   StreamableToString(write_handle_as_size_t) +
-                   " from the parent process " +
-                   StreamableToString(parent_process_id));
-  }
-
-  const HANDLE event_handle = reinterpret_cast<HANDLE>(event_handle_as_size_t);
-  HANDLE dup_event_handle;
-
-  if (!::DuplicateHandle(parent_process_handle.Get(), event_handle,
-                         ::GetCurrentProcess(), &dup_event_handle,
-                         0x0,
-                         FALSE,
-                         DUPLICATE_SAME_ACCESS)) {
-    DeathTestAbort("Unable to duplicate the event handle " +
-                   StreamableToString(event_handle_as_size_t) +
-                   " from the parent process " +
-                   StreamableToString(parent_process_id));
-  }
-
-  const int write_fd =
-      ::_open_osfhandle(reinterpret_cast<intptr_t>(dup_write_handle), O_APPEND);
-  if (write_fd == -1) {
-    DeathTestAbort("Unable to convert pipe handle " +
-                   StreamableToString(write_handle_as_size_t) +
-                   " to a file descriptor");
-  }
-
-  // Signals the parent that the write end of the pipe has been acquired
-  // so the parent can release its own write end.
-  ::SetEvent(dup_event_handle);
-
-  return write_fd;
-}
-# endif  // GTEST_OS_WINDOWS
-
-// Returns a newly created InternalRunDeathTestFlag object with fields
-// initialized from the GTEST_FLAG(internal_run_death_test) flag if
-// the flag is specified; otherwise returns NULL.
-InternalRunDeathTestFlag* ParseInternalRunDeathTestFlag() {
-  if (GTEST_FLAG(internal_run_death_test) == "") return NULL;
-
-  // GTEST_HAS_DEATH_TEST implies that we have ::std::string, so we
-  // can use it here.
-  int line = -1;
-  int index = -1;
-  ::std::vector< ::std::string> fields;
-  SplitString(GTEST_FLAG(internal_run_death_test).c_str(), '|', &fields);
-  int write_fd = -1;
-
-# if GTEST_OS_WINDOWS
-
-  unsigned int parent_process_id = 0;
-  size_t write_handle_as_size_t = 0;
-  size_t event_handle_as_size_t = 0;
-
-  if (fields.size() != 6
-      || !ParseNaturalNumber(fields[1], &line)
-      || !ParseNaturalNumber(fields[2], &index)
-      || !ParseNaturalNumber(fields[3], &parent_process_id)
-      || !ParseNaturalNumber(fields[4], &write_handle_as_size_t)
-      || !ParseNaturalNumber(fields[5], &event_handle_as_size_t)) {
-    DeathTestAbort("Bad --gtest_internal_run_death_test flag: " +
-                   GTEST_FLAG(internal_run_death_test));
-  }
-  write_fd = GetStatusFileDescriptor(parent_process_id,
-                                     write_handle_as_size_t,
-                                     event_handle_as_size_t);
-
-# elif GTEST_OS_FUCHSIA
-
-  if (fields.size() != 3
-      || !ParseNaturalNumber(fields[1], &line)
-      || !ParseNaturalNumber(fields[2], &index)) {
-    DeathTestAbort("Bad --gtest_internal_run_death_test flag: "
-        + GTEST_FLAG(internal_run_death_test));
-  }
-
-# else
-
-  if (fields.size() != 4
-      || !ParseNaturalNumber(fields[1], &line)
-      || !ParseNaturalNumber(fields[2], &index)
-      || !ParseNaturalNumber(fields[3], &write_fd)) {
-    DeathTestAbort("Bad --gtest_internal_run_death_test flag: "
-        + GTEST_FLAG(internal_run_death_test));
-  }
-
-# endif  // GTEST_OS_WINDOWS
-
-  return new InternalRunDeathTestFlag(fields[0], line, index, write_fd);
-}
-
-}  // namespace internal
-
-#endif  // GTEST_HAS_DEATH_TEST
-
-}  // namespace testing
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-
-#include <stdlib.h>
-
-#if GTEST_OS_WINDOWS_MOBILE
-# include <windows.h>
-#elif GTEST_OS_WINDOWS
-# include <direct.h>
-# include <io.h>
-#elif GTEST_OS_SYMBIAN
-// Symbian OpenC has PATH_MAX in sys/syslimits.h
-# include <sys/syslimits.h>
-#else
-# include <limits.h>
-# include <climits>  // Some Linux distributions define PATH_MAX here.
-#endif  // GTEST_OS_WINDOWS_MOBILE
-
-
-#if GTEST_OS_WINDOWS
-# define GTEST_PATH_MAX_ _MAX_PATH
-#elif defined(PATH_MAX)
-# define GTEST_PATH_MAX_ PATH_MAX
-#elif defined(_XOPEN_PATH_MAX)
-# define GTEST_PATH_MAX_ _XOPEN_PATH_MAX
-#else
-# define GTEST_PATH_MAX_ _POSIX_PATH_MAX
-#endif  // GTEST_OS_WINDOWS
-
-namespace testing {
-namespace internal {
-
-#if GTEST_OS_WINDOWS
-// On Windows, '\\' is the standard path separator, but many tools and the
-// Windows API also accept '/' as an alternate path separator. Unless otherwise
-// noted, a file path can contain either kind of path separators, or a mixture
-// of them.
-const char kPathSeparator = '\\';
-const char kAlternatePathSeparator = '/';
-const char kAlternatePathSeparatorString[] = "/";
-# if GTEST_OS_WINDOWS_MOBILE
-// Windows CE doesn't have a current directory. You should not use
-// the current directory in tests on Windows CE, but this at least
-// provides a reasonable fallback.
-const char kCurrentDirectoryString[] = "\\";
-// Windows CE doesn't define INVALID_FILE_ATTRIBUTES
-const DWORD kInvalidFileAttributes = 0xffffffff;
-# else
-const char kCurrentDirectoryString[] = ".\\";
-# endif  // GTEST_OS_WINDOWS_MOBILE
-#else
-const char kPathSeparator = '/';
-const char kCurrentDirectoryString[] = "./";
-#endif  // GTEST_OS_WINDOWS
-
-// Returns whether the given character is a valid path separator.
-static bool IsPathSeparator(char c) {
-#if GTEST_HAS_ALT_PATH_SEP_
-  return (c == kPathSeparator) || (c == kAlternatePathSeparator);
-#else
-  return c == kPathSeparator;
-#endif
-}
-
-// Returns the current working directory, or "" if unsuccessful.
-FilePath FilePath::GetCurrentDir() {
-#if GTEST_OS_WINDOWS_MOBILE || GTEST_OS_WINDOWS_PHONE || GTEST_OS_WINDOWS_RT
-  // Windows CE doesn't have a current directory, so we just return
-  // something reasonable.
-  return FilePath(kCurrentDirectoryString);
-#elif GTEST_OS_WINDOWS
-  char cwd[GTEST_PATH_MAX_ + 1] = { '\0' };
-  return FilePath(_getcwd(cwd, sizeof(cwd)) == NULL ? "" : cwd);
-#else
-  char cwd[GTEST_PATH_MAX_ + 1] = { '\0' };
-  char* result = getcwd(cwd, sizeof(cwd));
-# if GTEST_OS_NACL
-  // getcwd will likely fail in NaCl due to the sandbox, so return something
-  // reasonable. The user may have provided a shim implementation for getcwd,
-  // however, so fallback only when failure is detected.
-  return FilePath(result == NULL ? kCurrentDirectoryString : cwd);
-# endif  // GTEST_OS_NACL
-  return FilePath(result == NULL ? "" : cwd);
-#endif  // GTEST_OS_WINDOWS_MOBILE
-}
-
-// Returns a copy of the FilePath with the case-insensitive extension removed.
-// Example: FilePath("dir/file.exe").RemoveExtension("EXE") returns
-// FilePath("dir/file"). If a case-insensitive extension is not
-// found, returns a copy of the original FilePath.
-FilePath FilePath::RemoveExtension(const char* extension) const {
-  const std::string dot_extension = std::string(".") + extension;
-  if (String::EndsWithCaseInsensitive(pathname_, dot_extension)) {
-    return FilePath(pathname_.substr(
-        0, pathname_.length() - dot_extension.length()));
-  }
-  return *this;
-}
-
-// Returns a pointer to the last occurrence of a valid path separator in
-// the FilePath. On Windows, for example, both '/' and '\' are valid path
-// separators. Returns NULL if no path separator was found.
-const char* FilePath::FindLastPathSeparator() const {
-  const char* const last_sep = strrchr(c_str(), kPathSeparator);
-#if GTEST_HAS_ALT_PATH_SEP_
-  const char* const last_alt_sep = strrchr(c_str(), kAlternatePathSeparator);
-  // Comparing two pointers of which only one is NULL is undefined.
-  if (last_alt_sep != NULL &&
-      (last_sep == NULL || last_alt_sep > last_sep)) {
-    return last_alt_sep;
-  }
-#endif
-  return last_sep;
-}
-
-// Returns a copy of the FilePath with the directory part removed.
-// Example: FilePath("path/to/file").RemoveDirectoryName() returns
-// FilePath("file"). If there is no directory part ("just_a_file"), it returns
-// the FilePath unmodified. If there is no file part ("just_a_dir/") it
-// returns an empty FilePath ("").
-// On Windows platform, '\' is the path separator, otherwise it is '/'.
-FilePath FilePath::RemoveDirectoryName() const {
-  const char* const last_sep = FindLastPathSeparator();
-  return last_sep ? FilePath(last_sep + 1) : *this;
-}
-
-// RemoveFileName returns the directory path with the filename removed.
-// Example: FilePath("path/to/file").RemoveFileName() returns "path/to/".
-// If the FilePath is "a_file" or "/a_file", RemoveFileName returns
-// FilePath("./") or, on Windows, FilePath(".\\"). If the filepath does
-// not have a file, like "just/a/dir/", it returns the FilePath unmodified.
-// On Windows platform, '\' is the path separator, otherwise it is '/'.
-FilePath FilePath::RemoveFileName() const {
-  const char* const last_sep = FindLastPathSeparator();
-  std::string dir;
-  if (last_sep) {
-    dir = std::string(c_str(), last_sep + 1 - c_str());
-  } else {
-    dir = kCurrentDirectoryString;
-  }
-  return FilePath(dir);
-}
-
-// Helper functions for naming files in a directory for xml output.
-
-// Given directory = "dir", base_name = "test", number = 0,
-// extension = "xml", returns "dir/test.xml". If number is greater
-// than zero (e.g., 12), returns "dir/test_12.xml".
-// On Windows platform, uses \ as the separator rather than /.
-FilePath FilePath::MakeFileName(const FilePath& directory,
-                                const FilePath& base_name,
-                                int number,
-                                const char* extension) {
-  std::string file;
-  if (number == 0) {
-    file = base_name.string() + "." + extension;
-  } else {
-    file = base_name.string() + "_" + StreamableToString(number)
-        + "." + extension;
-  }
-  return ConcatPaths(directory, FilePath(file));
-}
-
-// Given directory = "dir", relative_path = "test.xml", returns "dir/test.xml".
-// On Windows, uses \ as the separator rather than /.
-FilePath FilePath::ConcatPaths(const FilePath& directory,
-                               const FilePath& relative_path) {
-  if (directory.IsEmpty())
-    return relative_path;
-  const FilePath dir(directory.RemoveTrailingPathSeparator());
-  return FilePath(dir.string() + kPathSeparator + relative_path.string());
-}
-
-// Returns true if pathname describes something findable in the file-system,
-// either a file, directory, or whatever.
-bool FilePath::FileOrDirectoryExists() const {
-#if GTEST_OS_WINDOWS_MOBILE
-  LPCWSTR unicode = String::AnsiToUtf16(pathname_.c_str());
-  const DWORD attributes = GetFileAttributes(unicode);
-  delete [] unicode;
-  return attributes != kInvalidFileAttributes;
-#else
-  posix::StatStruct file_stat;
-  return posix::Stat(pathname_.c_str(), &file_stat) == 0;
-#endif  // GTEST_OS_WINDOWS_MOBILE
-}
-
-// Returns true if pathname describes a directory in the file-system
-// that exists.
-bool FilePath::DirectoryExists() const {
-  bool result = false;
-#if GTEST_OS_WINDOWS
-  // Don't strip off trailing separator if path is a root directory on
-  // Windows (like "C:\\").
-  const FilePath& path(IsRootDirectory() ? *this :
-                                           RemoveTrailingPathSeparator());
-#else
-  const FilePath& path(*this);
-#endif
-
-#if GTEST_OS_WINDOWS_MOBILE
-  LPCWSTR unicode = String::AnsiToUtf16(path.c_str());
-  const DWORD attributes = GetFileAttributes(unicode);
-  delete [] unicode;
-  if ((attributes != kInvalidFileAttributes) &&
-      (attributes & FILE_ATTRIBUTE_DIRECTORY)) {
-    result = true;
-  }
-#else
-  posix::StatStruct file_stat;
-  result = posix::Stat(path.c_str(), &file_stat) == 0 &&
-      posix::IsDir(file_stat);
-#endif  // GTEST_OS_WINDOWS_MOBILE
-
-  return result;
-}
-
-// Returns true if pathname describes a root directory. (Windows has one
-// root directory per disk drive.)
-bool FilePath::IsRootDirectory() const {
-#if GTEST_OS_WINDOWS
-  // TODO(wan@google.com): on Windows a network share like
-  // \\server\share can be a root directory, although it cannot be the
-  // current directory.  Handle this properly.
-  return pathname_.length() == 3 && IsAbsolutePath();
-#else
-  return pathname_.length() == 1 && IsPathSeparator(pathname_.c_str()[0]);
-#endif
-}
-
-// Returns true if pathname describes an absolute path.
-bool FilePath::IsAbsolutePath() const {
-  const char* const name = pathname_.c_str();
-#if GTEST_OS_WINDOWS
-  return pathname_.length() >= 3 &&
-     ((name[0] >= 'a' && name[0] <= 'z') ||
-      (name[0] >= 'A' && name[0] <= 'Z')) &&
-     name[1] == ':' &&
-     IsPathSeparator(name[2]);
-#else
-  return IsPathSeparator(name[0]);
-#endif
-}
-
-// Returns a pathname for a file that does not currently exist. The pathname
-// will be directory/base_name.extension or
-// directory/base_name_<number>.extension if directory/base_name.extension
-// already exists. The number will be incremented until a pathname is found
-// that does not already exist.
-// Examples: 'dir/foo_test.xml' or 'dir/foo_test_1.xml'.
-// There could be a race condition if two or more processes are calling this
-// function at the same time -- they could both pick the same filename.
-FilePath FilePath::GenerateUniqueFileName(const FilePath& directory,
-                                          const FilePath& base_name,
-                                          const char* extension) {
-  FilePath full_pathname;
-  int number = 0;
-  do {
-    full_pathname.Set(MakeFileName(directory, base_name, number++, extension));
-  } while (full_pathname.FileOrDirectoryExists());
-  return full_pathname;
-}
-
-// Returns true if FilePath ends with a path separator, which indicates that
-// it is intended to represent a directory. Returns false otherwise.
-// This does NOT check that a directory (or file) actually exists.
-bool FilePath::IsDirectory() const {
-  return !pathname_.empty() &&
-         IsPathSeparator(pathname_.c_str()[pathname_.length() - 1]);
-}
-
-// Create directories so that path exists. Returns true if successful or if
-// the directories already exist; returns false if unable to create directories
-// for any reason.
-bool FilePath::CreateDirectoriesRecursively() const {
-  if (!this->IsDirectory()) {
-    return false;
-  }
-
-  if (pathname_.length() == 0 || this->DirectoryExists()) {
-    return true;
-  }
-
-  const FilePath parent(this->RemoveTrailingPathSeparator().RemoveFileName());
-  return parent.CreateDirectoriesRecursively() && this->CreateFolder();
-}
-
-// Create the directory so that path exists. Returns true if successful or
-// if the directory already exists; returns false if unable to create the
-// directory for any reason, including if the parent directory does not
-// exist. Not named "CreateDirectory" because that's a macro on Windows.
-bool FilePath::CreateFolder() const {
-#if GTEST_OS_WINDOWS_MOBILE
-  FilePath removed_sep(this->RemoveTrailingPathSeparator());
-  LPCWSTR unicode = String::AnsiToUtf16(removed_sep.c_str());
-  int result = CreateDirectory(unicode, NULL) ? 0 : -1;
-  delete [] unicode;
-#elif GTEST_OS_WINDOWS
-  int result = _mkdir(pathname_.c_str());
-#else
-  int result = mkdir(pathname_.c_str(), 0777);
-#endif  // GTEST_OS_WINDOWS_MOBILE
-
-  if (result == -1) {
-    return this->DirectoryExists();  // An error is OK if the directory exists.
-  }
-  return true;  // No error.
-}
-
-// If input name has a trailing separator character, remove it and return the
-// name, otherwise return the name string unmodified.
-// On Windows platform, uses \ as the separator, other platforms use /.
-FilePath FilePath::RemoveTrailingPathSeparator() const {
-  return IsDirectory()
-      ? FilePath(pathname_.substr(0, pathname_.length() - 1))
-      : *this;
-}
-
-// Removes any redundant separators that might be in the pathname.
-// For example, "bar///foo" becomes "bar/foo". Does not eliminate other
-// redundancies that might be in a pathname involving "." or "..".
-// TODO(wan@google.com): handle Windows network shares (e.g. \\server\share).
-void FilePath::Normalize() {
-  if (pathname_.c_str() == NULL) {
-    pathname_ = "";
-    return;
-  }
-  const char* src = pathname_.c_str();
-  char* const dest = new char[pathname_.length() + 1];
-  char* dest_ptr = dest;
-  memset(dest_ptr, 0, pathname_.length() + 1);
-
-  while (*src != '\0') {
-    *dest_ptr = *src;
-    if (!IsPathSeparator(*src)) {
-      src++;
-    } else {
-#if GTEST_HAS_ALT_PATH_SEP_
-      if (*dest_ptr == kAlternatePathSeparator) {
-        *dest_ptr = kPathSeparator;
-      }
-#endif
-      while (IsPathSeparator(*src))
-        src++;
-    }
-    dest_ptr++;
-  }
-  *dest_ptr = '\0';
-  pathname_ = dest;
-  delete[] dest;
-}
-
-}  // namespace internal
-}  // namespace testing
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-
-#include <limits.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-#include <fstream>
-
-#if GTEST_OS_WINDOWS
-# include <windows.h>
-# include <io.h>
-# include <sys/stat.h>
-# include <map>  // Used in ThreadLocal.
-#else
-# include <unistd.h>
-#endif  // GTEST_OS_WINDOWS
-
-#if GTEST_OS_MAC
-# include <mach/mach_init.h>
-# include <mach/task.h>
-# include <mach/vm_map.h>
-#endif  // GTEST_OS_MAC
-
-#if GTEST_OS_QNX
-# include <devctl.h>
-# include <fcntl.h>
-# include <sys/procfs.h>
-#endif  // GTEST_OS_QNX
-
-#if GTEST_OS_AIX
-# include <procinfo.h>
-# include <sys/types.h>
-#endif  // GTEST_OS_AIX
-
-#if GTEST_OS_FUCHSIA
-# include <zircon/process.h>
-# include <zircon/syscalls.h>
-#endif  // GTEST_OS_FUCHSIA
-
-
-namespace testing {
-namespace internal {
-
-#if defined(_MSC_VER) || defined(__BORLANDC__)
-// MSVC and C++Builder do not provide a definition of STDERR_FILENO.
-const int kStdOutFileno = 1;
-const int kStdErrFileno = 2;
-#else
-const int kStdOutFileno = STDOUT_FILENO;
-const int kStdErrFileno = STDERR_FILENO;
-#endif  // _MSC_VER
-
-#if GTEST_OS_LINUX
-
-namespace {
-template <typename T>
-T ReadProcFileField(const std::string& filename, int field) {
-  std::string dummy;
-  std::ifstream file(filename.c_str());
-  while (field-- > 0) {
-    file >> dummy;
-  }
-  T output = 0;
-  file >> output;
-  return output;
-}
-}  // namespace
-
-// Returns the number of active threads, or 0 when there is an error.
-size_t GetThreadCount() {
-  const std::string filename =
-      (Message() << "/proc/" << getpid() << "/stat").GetString();
-  return ReadProcFileField<int>(filename, 19);
-}
-
-#elif GTEST_OS_MAC
-
-size_t GetThreadCount() {
-  const task_t task = mach_task_self();
-  mach_msg_type_number_t thread_count;
-  thread_act_array_t thread_list;
-  const kern_return_t status = task_threads(task, &thread_list, &thread_count);
-  if (status == KERN_SUCCESS) {
-    // task_threads allocates resources in thread_list and we need to free them
-    // to avoid leaks.
-    vm_deallocate(task,
-                  reinterpret_cast<vm_address_t>(thread_list),
-                  sizeof(thread_t) * thread_count);
-    return static_cast<size_t>(thread_count);
-  } else {
-    return 0;
-  }
-}
-
-#elif GTEST_OS_QNX
-
-// Returns the number of threads running in the process, or 0 to indicate that
-// we cannot detect it.
-size_t GetThreadCount() {
-  const int fd = open("/proc/self/as", O_RDONLY);
-  if (fd < 0) {
-    return 0;
-  }
-  procfs_info process_info;
-  const int status =
-      devctl(fd, DCMD_PROC_INFO, &process_info, sizeof(process_info), NULL);
-  close(fd);
-  if (status == EOK) {
-    return static_cast<size_t>(process_info.num_threads);
-  } else {
-    return 0;
-  }
-}
-
-#elif GTEST_OS_AIX
-
-size_t GetThreadCount() {
-  struct procentry64 entry;
-  pid_t pid = getpid();
-  int status = getprocs64(&entry, sizeof(entry), NULL, 0, &pid, 1);
-  if (status == 1) {
-    return entry.pi_thcount;
-  } else {
-    return 0;
-  }
-}
-
-#elif GTEST_OS_FUCHSIA
-
-size_t GetThreadCount() {
-  int dummy_buffer;
-  size_t avail;
-  zx_status_t status = zx_object_get_info(
-      zx_process_self(),
-      ZX_INFO_PROCESS_THREADS,
-      &dummy_buffer,
-      0,
-      nullptr,
-      &avail);
-  if (status == ZX_OK) {
-    return avail;
-  } else {
-    return 0;
-  }
-}
-
-#else
-
-size_t GetThreadCount() {
-  // There's no portable way to detect the number of threads, so we just
-  // return 0 to indicate that we cannot detect it.
-  return 0;
-}
-
-#endif  // GTEST_OS_LINUX
-
-#if GTEST_IS_THREADSAFE && GTEST_OS_WINDOWS
-
-void SleepMilliseconds(int n) {
-  ::Sleep(n);
-}
-
-AutoHandle::AutoHandle()
-    : handle_(INVALID_HANDLE_VALUE) {}
-
-AutoHandle::AutoHandle(Handle handle)
-    : handle_(handle) {}
-
-AutoHandle::~AutoHandle() {
-  Reset();
-}
-
-AutoHandle::Handle AutoHandle::Get() const {
-  return handle_;
-}
-
-void AutoHandle::Reset() {
-  Reset(INVALID_HANDLE_VALUE);
-}
-
-void AutoHandle::Reset(HANDLE handle) {
-  // Resetting with the same handle we already own is invalid.
-  if (handle_ != handle) {
-    if (IsCloseable()) {
-      ::CloseHandle(handle_);
-    }
-    handle_ = handle;
-  } else {
-    GTEST_CHECK_(!IsCloseable())
-        << "Resetting a valid handle to itself is likely a programmer error "
-            "and thus not allowed.";
-  }
-}
-
-bool AutoHandle::IsCloseable() const {
-  // Different Windows APIs may use either of these values to represent an
-  // invalid handle.
-  return handle_ != NULL && handle_ != INVALID_HANDLE_VALUE;
-}
-
-Notification::Notification()
-    : event_(::CreateEvent(NULL,   // Default security attributes.
-                           TRUE,   // Do not reset automatically.
-                           FALSE,  // Initially unset.
-                           NULL)) {  // Anonymous event.
-  GTEST_CHECK_(event_.Get() != NULL);
-}
-
-void Notification::Notify() {
-  GTEST_CHECK_(::SetEvent(event_.Get()) != FALSE);
-}
-
-void Notification::WaitForNotification() {
-  GTEST_CHECK_(
-      ::WaitForSingleObject(event_.Get(), INFINITE) == WAIT_OBJECT_0);
-}
-
-Mutex::Mutex()
-    : owner_thread_id_(0),
-      type_(kDynamic),
-      critical_section_init_phase_(0),
-      critical_section_(new CRITICAL_SECTION) {
-  ::InitializeCriticalSection(critical_section_);
-}
-
-Mutex::~Mutex() {
-  // Static mutexes are leaked intentionally. It is not thread-safe to try
-  // to clean them up.
-  // TODO(yukawa): Switch to Slim Reader/Writer (SRW) Locks, which requires
-  // nothing to clean it up but is available only on Vista and later.
-  // http://msdn.microsoft.com/en-us/library/windows/desktop/aa904937.aspx
-  if (type_ == kDynamic) {
-    ::DeleteCriticalSection(critical_section_);
-    delete critical_section_;
-    critical_section_ = NULL;
-  }
-}
-
-void Mutex::Lock() {
-  ThreadSafeLazyInit();
-  ::EnterCriticalSection(critical_section_);
-  owner_thread_id_ = ::GetCurrentThreadId();
-}
-
-void Mutex::Unlock() {
-  ThreadSafeLazyInit();
-  // We don't protect writing to owner_thread_id_ here, as it's the
-  // caller's responsibility to ensure that the current thread holds the
-  // mutex when this is called.
-  owner_thread_id_ = 0;
-  ::LeaveCriticalSection(critical_section_);
-}
-
-// Does nothing if the current thread holds the mutex. Otherwise, crashes
-// with high probability.
-void Mutex::AssertHeld() {
-  ThreadSafeLazyInit();
-  GTEST_CHECK_(owner_thread_id_ == ::GetCurrentThreadId())
-      << "The current thread is not holding the mutex @" << this;
-}
-
-// Initializes owner_thread_id_ and critical_section_ in static mutexes.
-void Mutex::ThreadSafeLazyInit() {
-  // Dynamic mutexes are initialized in the constructor.
-  if (type_ == kStatic) {
-    switch (
-        ::InterlockedCompareExchange(&critical_section_init_phase_, 1L, 0L)) {
-      case 0:
-        // If critical_section_init_phase_ was 0 before the exchange, we
-        // are the first to test it and need to perform the initialization.
-        owner_thread_id_ = 0;
-        critical_section_ = new CRITICAL_SECTION;
-        ::InitializeCriticalSection(critical_section_);
-        // Updates the critical_section_init_phase_ to 2 to signal
-        // initialization complete.
-        GTEST_CHECK_(::InterlockedCompareExchange(
-                          &critical_section_init_phase_, 2L, 1L) ==
-                      1L);
-        break;
-      case 1:
-        // Somebody else is already initializing the mutex; spin until they
-        // are done.
-        while (::InterlockedCompareExchange(&critical_section_init_phase_,
-                                            2L,
-                                            2L) != 2L) {
-          // Possibly yields the rest of the thread's time slice to other
-          // threads.
-          ::Sleep(0);
-        }
-        break;
-
-      case 2:
-        break;  // The mutex is already initialized and ready for use.
-
-      default:
-        GTEST_CHECK_(false)
-            << "Unexpected value of critical_section_init_phase_ "
-            << "while initializing a static mutex.";
-    }
-  }
-}
-
-namespace {
-
-class ThreadWithParamSupport : public ThreadWithParamBase {
- public:
-  static HANDLE CreateThread(Runnable* runnable,
-                             Notification* thread_can_start) {
-    ThreadMainParam* param = new ThreadMainParam(runnable, thread_can_start);
-    DWORD thread_id;
-    // TODO(yukawa): Consider to use _beginthreadex instead.
-    HANDLE thread_handle = ::CreateThread(
-        NULL,    // Default security.
-        0,       // Default stack size.
-        &ThreadWithParamSupport::ThreadMain,
-        param,   // Parameter to ThreadMainStatic
-        0x0,     // Default creation flags.
-        &thread_id);  // Need a valid pointer for the call to work under Win98.
-    GTEST_CHECK_(thread_handle != NULL) << "CreateThread failed with error "
-                                        << ::GetLastError() << ".";
-    if (thread_handle == NULL) {
-      delete param;
-    }
-    return thread_handle;
-  }
-
- private:
-  struct ThreadMainParam {
-    ThreadMainParam(Runnable* runnable, Notification* thread_can_start)
-        : runnable_(runnable),
-          thread_can_start_(thread_can_start) {
-    }
-    scoped_ptr<Runnable> runnable_;
-    // Does not own.
-    Notification* thread_can_start_;
-  };
-
-  static DWORD WINAPI ThreadMain(void* ptr) {
-    // Transfers ownership.
-    scoped_ptr<ThreadMainParam> param(static_cast<ThreadMainParam*>(ptr));
-    if (param->thread_can_start_ != NULL)
-      param->thread_can_start_->WaitForNotification();
-    param->runnable_->Run();
-    return 0;
-  }
-
-  // Prohibit instantiation.
-  ThreadWithParamSupport();
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ThreadWithParamSupport);
-};
-
-}  // namespace
-
-ThreadWithParamBase::ThreadWithParamBase(Runnable *runnable,
-                                         Notification* thread_can_start)
-      : thread_(ThreadWithParamSupport::CreateThread(runnable,
-                                                     thread_can_start)) {
-}
-
-ThreadWithParamBase::~ThreadWithParamBase() {
-  Join();
-}
-
-void ThreadWithParamBase::Join() {
-  GTEST_CHECK_(::WaitForSingleObject(thread_.Get(), INFINITE) == WAIT_OBJECT_0)
-      << "Failed to join the thread with error " << ::GetLastError() << ".";
-}
-
-// Maps a thread to a set of ThreadIdToThreadLocals that have values
-// instantiated on that thread and notifies them when the thread exits.  A
-// ThreadLocal instance is expected to persist until all threads it has
-// values on have terminated.
-class ThreadLocalRegistryImpl {
- public:
-  // Registers thread_local_instance as having value on the current thread.
-  // Returns a value that can be used to identify the thread from other threads.
-  static ThreadLocalValueHolderBase* GetValueOnCurrentThread(
-      const ThreadLocalBase* thread_local_instance) {
-    DWORD current_thread = ::GetCurrentThreadId();
-    MutexLock lock(&mutex_);
-    ThreadIdToThreadLocals* const thread_to_thread_locals =
-        GetThreadLocalsMapLocked();
-    ThreadIdToThreadLocals::iterator thread_local_pos =
-        thread_to_thread_locals->find(current_thread);
-    if (thread_local_pos == thread_to_thread_locals->end()) {
-      thread_local_pos = thread_to_thread_locals->insert(
-          std::make_pair(current_thread, ThreadLocalValues())).first;
-      StartWatcherThreadFor(current_thread);
-    }
-    ThreadLocalValues& thread_local_values = thread_local_pos->second;
-    ThreadLocalValues::iterator value_pos =
-        thread_local_values.find(thread_local_instance);
-    if (value_pos == thread_local_values.end()) {
-      value_pos =
-          thread_local_values
-              .insert(std::make_pair(
-                  thread_local_instance,
-                  linked_ptr<ThreadLocalValueHolderBase>(
-                      thread_local_instance->NewValueForCurrentThread())))
-              .first;
-    }
-    return value_pos->second.get();
-  }
-
-  static void OnThreadLocalDestroyed(
-      const ThreadLocalBase* thread_local_instance) {
-    std::vector<linked_ptr<ThreadLocalValueHolderBase> > value_holders;
-    // Clean up the ThreadLocalValues data structure while holding the lock, but
-    // defer the destruction of the ThreadLocalValueHolderBases.
-    {
-      MutexLock lock(&mutex_);
-      ThreadIdToThreadLocals* const thread_to_thread_locals =
-          GetThreadLocalsMapLocked();
-      for (ThreadIdToThreadLocals::iterator it =
-          thread_to_thread_locals->begin();
-          it != thread_to_thread_locals->end();
-          ++it) {
-        ThreadLocalValues& thread_local_values = it->second;
-        ThreadLocalValues::iterator value_pos =
-            thread_local_values.find(thread_local_instance);
-        if (value_pos != thread_local_values.end()) {
-          value_holders.push_back(value_pos->second);
-          thread_local_values.erase(value_pos);
-          // This 'if' can only be successful at most once, so theoretically we
-          // could break out of the loop here, but we don't bother doing so.
-        }
-      }
-    }
-    // Outside the lock, let the destructor for 'value_holders' deallocate the
-    // ThreadLocalValueHolderBases.
-  }
-
-  static void OnThreadExit(DWORD thread_id) {
-    GTEST_CHECK_(thread_id != 0) << ::GetLastError();
-    std::vector<linked_ptr<ThreadLocalValueHolderBase> > value_holders;
-    // Clean up the ThreadIdToThreadLocals data structure while holding the
-    // lock, but defer the destruction of the ThreadLocalValueHolderBases.
-    {
-      MutexLock lock(&mutex_);
-      ThreadIdToThreadLocals* const thread_to_thread_locals =
-          GetThreadLocalsMapLocked();
-      ThreadIdToThreadLocals::iterator thread_local_pos =
-          thread_to_thread_locals->find(thread_id);
-      if (thread_local_pos != thread_to_thread_locals->end()) {
-        ThreadLocalValues& thread_local_values = thread_local_pos->second;
-        for (ThreadLocalValues::iterator value_pos =
-            thread_local_values.begin();
-            value_pos != thread_local_values.end();
-            ++value_pos) {
-          value_holders.push_back(value_pos->second);
-        }
-        thread_to_thread_locals->erase(thread_local_pos);
-      }
-    }
-    // Outside the lock, let the destructor for 'value_holders' deallocate the
-    // ThreadLocalValueHolderBases.
-  }
-
- private:
-  // In a particular thread, maps a ThreadLocal object to its value.
-  typedef std::map<const ThreadLocalBase*,
-                   linked_ptr<ThreadLocalValueHolderBase> > ThreadLocalValues;
-  // Stores all ThreadIdToThreadLocals having values in a thread, indexed by
-  // thread's ID.
-  typedef std::map<DWORD, ThreadLocalValues> ThreadIdToThreadLocals;
-
-  // Holds the thread id and thread handle that we pass from
-  // StartWatcherThreadFor to WatcherThreadFunc.
-  typedef std::pair<DWORD, HANDLE> ThreadIdAndHandle;
-
-  static void StartWatcherThreadFor(DWORD thread_id) {
-    // The returned handle will be kept in thread_map and closed by
-    // watcher_thread in WatcherThreadFunc.
-    HANDLE thread = ::OpenThread(SYNCHRONIZE | THREAD_QUERY_INFORMATION,
-                                 FALSE,
-                                 thread_id);
-    GTEST_CHECK_(thread != NULL);
-    // We need to pass a valid thread ID pointer into CreateThread for it
-    // to work correctly under Win98.
-    DWORD watcher_thread_id;
-    HANDLE watcher_thread = ::CreateThread(
-        NULL,   // Default security.
-        0,      // Default stack size
-        &ThreadLocalRegistryImpl::WatcherThreadFunc,
-        reinterpret_cast<LPVOID>(new ThreadIdAndHandle(thread_id, thread)),
-        CREATE_SUSPENDED,
-        &watcher_thread_id);
-    GTEST_CHECK_(watcher_thread != NULL);
-    // Give the watcher thread the same priority as ours to avoid being
-    // blocked by it.
-    ::SetThreadPriority(watcher_thread,
-                        ::GetThreadPriority(::GetCurrentThread()));
-    ::ResumeThread(watcher_thread);
-    ::CloseHandle(watcher_thread);
-  }
-
-  // Monitors exit from a given thread and notifies those
-  // ThreadIdToThreadLocals about thread termination.
-  static DWORD WINAPI WatcherThreadFunc(LPVOID param) {
-    const ThreadIdAndHandle* tah =
-        reinterpret_cast<const ThreadIdAndHandle*>(param);
-    GTEST_CHECK_(
-        ::WaitForSingleObject(tah->second, INFINITE) == WAIT_OBJECT_0);
-    OnThreadExit(tah->first);
-    ::CloseHandle(tah->second);
-    delete tah;
-    return 0;
-  }
-
-  // Returns map of thread local instances.
-  static ThreadIdToThreadLocals* GetThreadLocalsMapLocked() {
-    mutex_.AssertHeld();
-    static ThreadIdToThreadLocals* map = new ThreadIdToThreadLocals;
-    return map;
-  }
-
-  // Protects access to GetThreadLocalsMapLocked() and its return value.
-  static Mutex mutex_;
-  // Protects access to GetThreadMapLocked() and its return value.
-  static Mutex thread_map_mutex_;
-};
-
-Mutex ThreadLocalRegistryImpl::mutex_(Mutex::kStaticMutex);
-Mutex ThreadLocalRegistryImpl::thread_map_mutex_(Mutex::kStaticMutex);
-
-ThreadLocalValueHolderBase* ThreadLocalRegistry::GetValueOnCurrentThread(
-      const ThreadLocalBase* thread_local_instance) {
-  return ThreadLocalRegistryImpl::GetValueOnCurrentThread(
-      thread_local_instance);
-}
-
-void ThreadLocalRegistry::OnThreadLocalDestroyed(
-      const ThreadLocalBase* thread_local_instance) {
-  ThreadLocalRegistryImpl::OnThreadLocalDestroyed(thread_local_instance);
-}
-
-#endif  // GTEST_IS_THREADSAFE && GTEST_OS_WINDOWS
-
-#if GTEST_USES_POSIX_RE
-
-// Implements RE.  Currently only needed for death tests.
-
-RE::~RE() {
-  if (is_valid_) {
-    // regfree'ing an invalid regex might crash because the content
-    // of the regex is undefined. Since the regex's are essentially
-    // the same, one cannot be valid (or invalid) without the other
-    // being so too.
-    regfree(&partial_regex_);
-    regfree(&full_regex_);
-  }
-  free(const_cast<char*>(pattern_));
-}
-
-// Returns true iff regular expression re matches the entire str.
-bool RE::FullMatch(const char* str, const RE& re) {
-  if (!re.is_valid_) return false;
-
-  regmatch_t match;
-  return regexec(&re.full_regex_, str, 1, &match, 0) == 0;
-}
-
-// Returns true iff regular expression re matches a substring of str
-// (including str itself).
-bool RE::PartialMatch(const char* str, const RE& re) {
-  if (!re.is_valid_) return false;
-
-  regmatch_t match;
-  return regexec(&re.partial_regex_, str, 1, &match, 0) == 0;
-}
-
-// Initializes an RE from its string representation.
-void RE::Init(const char* regex) {
-  pattern_ = posix::StrDup(regex);
-
-  // Reserves enough bytes to hold the regular expression used for a
-  // full match.
-  const size_t full_regex_len = strlen(regex) + 10;
-  char* const full_pattern = new char[full_regex_len];
-
-  snprintf(full_pattern, full_regex_len, "^(%s)$", regex);
-  is_valid_ = regcomp(&full_regex_, full_pattern, REG_EXTENDED) == 0;
-  // We want to call regcomp(&partial_regex_, ...) even if the
-  // previous expression returns false.  Otherwise partial_regex_ may
-  // not be properly initialized can may cause trouble when it's
-  // freed.
-  //
-  // Some implementation of POSIX regex (e.g. on at least some
-  // versions of Cygwin) doesn't accept the empty string as a valid
-  // regex.  We change it to an equivalent form "()" to be safe.
-  if (is_valid_) {
-    const char* const partial_regex = (*regex == '\0') ? "()" : regex;
-    is_valid_ = regcomp(&partial_regex_, partial_regex, REG_EXTENDED) == 0;
-  }
-  EXPECT_TRUE(is_valid_)
-      << "Regular expression \"" << regex
-      << "\" is not a valid POSIX Extended regular expression.";
-
-  delete[] full_pattern;
-}
-
-#elif GTEST_USES_SIMPLE_RE
-
-// Returns true iff ch appears anywhere in str (excluding the
-// terminating '\0' character).
-bool IsInSet(char ch, const char* str) {
-  return ch != '\0' && strchr(str, ch) != NULL;
-}
-
-// Returns true iff ch belongs to the given classification.  Unlike
-// similar functions in <ctype.h>, these aren't affected by the
-// current locale.
-bool IsAsciiDigit(char ch) { return '0' <= ch && ch <= '9'; }
-bool IsAsciiPunct(char ch) {
-  return IsInSet(ch, "^-!\"#$%&'()*+,./:;<=>?@[\\]_`{|}~");
-}
-bool IsRepeat(char ch) { return IsInSet(ch, "?*+"); }
-bool IsAsciiWhiteSpace(char ch) { return IsInSet(ch, " \f\n\r\t\v"); }
-bool IsAsciiWordChar(char ch) {
-  return ('a' <= ch && ch <= 'z') || ('A' <= ch && ch <= 'Z') ||
-      ('0' <= ch && ch <= '9') || ch == '_';
-}
-
-// Returns true iff "\\c" is a supported escape sequence.
-bool IsValidEscape(char c) {
-  return (IsAsciiPunct(c) || IsInSet(c, "dDfnrsStvwW"));
-}
-
-// Returns true iff the given atom (specified by escaped and pattern)
-// matches ch.  The result is undefined if the atom is invalid.
-bool AtomMatchesChar(bool escaped, char pattern_char, char ch) {
-  if (escaped) {  // "\\p" where p is pattern_char.
-    switch (pattern_char) {
-      case 'd': return IsAsciiDigit(ch);
-      case 'D': return !IsAsciiDigit(ch);
-      case 'f': return ch == '\f';
-      case 'n': return ch == '\n';
-      case 'r': return ch == '\r';
-      case 's': return IsAsciiWhiteSpace(ch);
-      case 'S': return !IsAsciiWhiteSpace(ch);
-      case 't': return ch == '\t';
-      case 'v': return ch == '\v';
-      case 'w': return IsAsciiWordChar(ch);
-      case 'W': return !IsAsciiWordChar(ch);
-    }
-    return IsAsciiPunct(pattern_char) && pattern_char == ch;
-  }
-
-  return (pattern_char == '.' && ch != '\n') || pattern_char == ch;
-}
-
-// Helper function used by ValidateRegex() to format error messages.
-static std::string FormatRegexSyntaxError(const char* regex, int index) {
-  return (Message() << "Syntax error at index " << index
-          << " in simple regular expression \"" << regex << "\": ").GetString();
-}
-
-// Generates non-fatal failures and returns false if regex is invalid;
-// otherwise returns true.
-bool ValidateRegex(const char* regex) {
-  if (regex == NULL) {
-    // TODO(wan@google.com): fix the source file location in the
-    // assertion failures to match where the regex is used in user
-    // code.
-    ADD_FAILURE() << "NULL is not a valid simple regular expression.";
-    return false;
-  }
-
-  bool is_valid = true;
-
-  // True iff ?, *, or + can follow the previous atom.
-  bool prev_repeatable = false;
-  for (int i = 0; regex[i]; i++) {
-    if (regex[i] == '\\') {  // An escape sequence
-      i++;
-      if (regex[i] == '\0') {
-        ADD_FAILURE() << FormatRegexSyntaxError(regex, i - 1)
-                      << "'\\' cannot appear at the end.";
-        return false;
-      }
-
-      if (!IsValidEscape(regex[i])) {
-        ADD_FAILURE() << FormatRegexSyntaxError(regex, i - 1)
-                      << "invalid escape sequence \"\\" << regex[i] << "\".";
-        is_valid = false;
-      }
-      prev_repeatable = true;
-    } else {  // Not an escape sequence.
-      const char ch = regex[i];
-
-      if (ch == '^' && i > 0) {
-        ADD_FAILURE() << FormatRegexSyntaxError(regex, i)
-                      << "'^' can only appear at the beginning.";
-        is_valid = false;
-      } else if (ch == '$' && regex[i + 1] != '\0') {
-        ADD_FAILURE() << FormatRegexSyntaxError(regex, i)
-                      << "'$' can only appear at the end.";
-        is_valid = false;
-      } else if (IsInSet(ch, "()[]{}|")) {
-        ADD_FAILURE() << FormatRegexSyntaxError(regex, i)
-                      << "'" << ch << "' is unsupported.";
-        is_valid = false;
-      } else if (IsRepeat(ch) && !prev_repeatable) {
-        ADD_FAILURE() << FormatRegexSyntaxError(regex, i)
-                      << "'" << ch << "' can only follow a repeatable token.";
-        is_valid = false;
-      }
-
-      prev_repeatable = !IsInSet(ch, "^$?*+");
-    }
-  }
-
-  return is_valid;
-}
-
-// Matches a repeated regex atom followed by a valid simple regular
-// expression.  The regex atom is defined as c if escaped is false,
-// or \c otherwise.  repeat is the repetition meta character (?, *,
-// or +).  The behavior is undefined if str contains too many
-// characters to be indexable by size_t, in which case the test will
-// probably time out anyway.  We are fine with this limitation as
-// std::string has it too.
-bool MatchRepetitionAndRegexAtHead(
-    bool escaped, char c, char repeat, const char* regex,
-    const char* str) {
-  const size_t min_count = (repeat == '+') ? 1 : 0;
-  const size_t max_count = (repeat == '?') ? 1 :
-      static_cast<size_t>(-1) - 1;
-  // We cannot call numeric_limits::max() as it conflicts with the
-  // max() macro on Windows.
-
-  for (size_t i = 0; i <= max_count; ++i) {
-    // We know that the atom matches each of the first i characters in str.
-    if (i >= min_count && MatchRegexAtHead(regex, str + i)) {
-      // We have enough matches at the head, and the tail matches too.
-      // Since we only care about *whether* the pattern matches str
-      // (as opposed to *how* it matches), there is no need to find a
-      // greedy match.
-      return true;
-    }
-    if (str[i] == '\0' || !AtomMatchesChar(escaped, c, str[i]))
-      return false;
-  }
-  return false;
-}
-
-// Returns true iff regex matches a prefix of str.  regex must be a
-// valid simple regular expression and not start with "^", or the
-// result is undefined.
-bool MatchRegexAtHead(const char* regex, const char* str) {
-  if (*regex == '\0')  // An empty regex matches a prefix of anything.
-    return true;
-
-  // "$" only matches the end of a string.  Note that regex being
-  // valid guarantees that there's nothing after "$" in it.
-  if (*regex == '$')
-    return *str == '\0';
-
-  // Is the first thing in regex an escape sequence?
-  const bool escaped = *regex == '\\';
-  if (escaped)
-    ++regex;
-  if (IsRepeat(regex[1])) {
-    // MatchRepetitionAndRegexAtHead() calls MatchRegexAtHead(), so
-    // here's an indirect recursion.  It terminates as the regex gets
-    // shorter in each recursion.
-    return MatchRepetitionAndRegexAtHead(
-        escaped, regex[0], regex[1], regex + 2, str);
-  } else {
-    // regex isn't empty, isn't "$", and doesn't start with a
-    // repetition.  We match the first atom of regex with the first
-    // character of str and recurse.
-    return (*str != '\0') && AtomMatchesChar(escaped, *regex, *str) &&
-        MatchRegexAtHead(regex + 1, str + 1);
-  }
-}
-
-// Returns true iff regex matches any substring of str.  regex must be
-// a valid simple regular expression, or the result is undefined.
-//
-// The algorithm is recursive, but the recursion depth doesn't exceed
-// the regex length, so we won't need to worry about running out of
-// stack space normally.  In rare cases the time complexity can be
-// exponential with respect to the regex length + the string length,
-// but usually it's must faster (often close to linear).
-bool MatchRegexAnywhere(const char* regex, const char* str) {
-  if (regex == NULL || str == NULL)
-    return false;
-
-  if (*regex == '^')
-    return MatchRegexAtHead(regex + 1, str);
-
-  // A successful match can be anywhere in str.
-  do {
-    if (MatchRegexAtHead(regex, str))
-      return true;
-  } while (*str++ != '\0');
-  return false;
-}
-
-// Implements the RE class.
-
-RE::~RE() {
-  free(const_cast<char*>(pattern_));
-  free(const_cast<char*>(full_pattern_));
-}
-
-// Returns true iff regular expression re matches the entire str.
-bool RE::FullMatch(const char* str, const RE& re) {
-  return re.is_valid_ && MatchRegexAnywhere(re.full_pattern_, str);
-}
-
-// Returns true iff regular expression re matches a substring of str
-// (including str itself).
-bool RE::PartialMatch(const char* str, const RE& re) {
-  return re.is_valid_ && MatchRegexAnywhere(re.pattern_, str);
-}
-
-// Initializes an RE from its string representation.
-void RE::Init(const char* regex) {
-  pattern_ = full_pattern_ = NULL;
-  if (regex != NULL) {
-    pattern_ = posix::StrDup(regex);
-  }
-
-  is_valid_ = ValidateRegex(regex);
-  if (!is_valid_) {
-    // No need to calculate the full pattern when the regex is invalid.
-    return;
-  }
-
-  const size_t len = strlen(regex);
-  // Reserves enough bytes to hold the regular expression used for a
-  // full match: we need space to prepend a '^', append a '$', and
-  // terminate the string with '\0'.
-  char* buffer = static_cast<char*>(malloc(len + 3));
-  full_pattern_ = buffer;
-
-  if (*regex != '^')
-    *buffer++ = '^';  // Makes sure full_pattern_ starts with '^'.
-
-  // We don't use snprintf or strncpy, as they trigger a warning when
-  // compiled with VC++ 8.0.
-  memcpy(buffer, regex, len);
-  buffer += len;
-
-  if (len == 0 || regex[len - 1] != '$')
-    *buffer++ = '$';  // Makes sure full_pattern_ ends with '$'.
-
-  *buffer = '\0';
-}
-
-#endif  // GTEST_USES_POSIX_RE
-
-const char kUnknownFile[] = "unknown file";
-
-// Formats a source file path and a line number as they would appear
-// in an error message from the compiler used to compile this code.
-GTEST_API_ ::std::string FormatFileLocation(const char* file, int line) {
-  const std::string file_name(file == NULL ? kUnknownFile : file);
-
-  if (line < 0) {
-    return file_name + ":";
-  }
-#ifdef _MSC_VER
-  return file_name + "(" + StreamableToString(line) + "):";
-#else
-  return file_name + ":" + StreamableToString(line) + ":";
-#endif  // _MSC_VER
-}
-
-// Formats a file location for compiler-independent XML output.
-// Although this function is not platform dependent, we put it next to
-// FormatFileLocation in order to contrast the two functions.
-// Note that FormatCompilerIndependentFileLocation() does NOT append colon
-// to the file location it produces, unlike FormatFileLocation().
-GTEST_API_ ::std::string FormatCompilerIndependentFileLocation(
-    const char* file, int line) {
-  const std::string file_name(file == NULL ? kUnknownFile : file);
-
-  if (line < 0)
-    return file_name;
-  else
-    return file_name + ":" + StreamableToString(line);
-}
-
-GTestLog::GTestLog(GTestLogSeverity severity, const char* file, int line)
-    : severity_(severity) {
-  const char* const marker =
-      severity == GTEST_INFO ?    "[  INFO ]" :
-      severity == GTEST_WARNING ? "[WARNING]" :
-      severity == GTEST_ERROR ?   "[ ERROR ]" : "[ FATAL ]";
-  GetStream() << ::std::endl << marker << " "
-              << FormatFileLocation(file, line).c_str() << ": ";
-}
-
-// Flushes the buffers and, if severity is GTEST_FATAL, aborts the program.
-GTestLog::~GTestLog() {
-  GetStream() << ::std::endl;
-  if (severity_ == GTEST_FATAL) {
-    fflush(stderr);
-    posix::Abort();
-  }
-}
-
-// Disable Microsoft deprecation warnings for POSIX functions called from
-// this class (creat, dup, dup2, and close)
-GTEST_DISABLE_MSC_WARNINGS_PUSH_(4996)
-
-#if GTEST_HAS_STREAM_REDIRECTION
-
-// Object that captures an output stream (stdout/stderr).
-class CapturedStream {
- public:
-  // The ctor redirects the stream to a temporary file.
-  explicit CapturedStream(int fd) : fd_(fd), uncaptured_fd_(dup(fd)) {
-# if GTEST_OS_WINDOWS
-    char temp_dir_path[MAX_PATH + 1] = { '\0' };  // NOLINT
-    char temp_file_path[MAX_PATH + 1] = { '\0' };  // NOLINT
-
-    ::GetTempPathA(sizeof(temp_dir_path), temp_dir_path);
-    const UINT success = ::GetTempFileNameA(temp_dir_path,
-                                            "gtest_redir",
-                                            0,  // Generate unique file name.
-                                            temp_file_path);
-    GTEST_CHECK_(success != 0)
-        << "Unable to create a temporary file in " << temp_dir_path;
-    const int captured_fd = creat(temp_file_path, _S_IREAD | _S_IWRITE);
-    GTEST_CHECK_(captured_fd != -1) << "Unable to open temporary file "
-                                    << temp_file_path;
-    filename_ = temp_file_path;
-# else
-    // There's no guarantee that a test has write access to the current
-    // directory, so we create the temporary file in the /tmp directory
-    // instead. We use /tmp on most systems, and /sdcard on Android.
-    // That's because Android doesn't have /tmp.
-#  if GTEST_OS_LINUX_ANDROID
-    // Note: Android applications are expected to call the framework's
-    // Context.getExternalStorageDirectory() method through JNI to get
-    // the location of the world-writable SD Card directory. However,
-    // this requires a Context handle, which cannot be retrieved
-    // globally from native code. Doing so also precludes running the
-    // code as part of a regular standalone executable, which doesn't
-    // run in a Dalvik process (e.g. when running it through 'adb shell').
-    //
-    // The location /sdcard is directly accessible from native code
-    // and is the only location (unofficially) supported by the Android
-    // team. It's generally a symlink to the real SD Card mount point
-    // which can be /mnt/sdcard, /mnt/sdcard0, /system/media/sdcard, or
-    // other OEM-customized locations. Never rely on these, and always
-    // use /sdcard.
-    char name_template[] = "/sdcard/gtest_captured_stream.XXXXXX";
-#  else
-    char name_template[] = "/tmp/captured_stream.XXXXXX";
-#  endif  // GTEST_OS_LINUX_ANDROID
-    const int captured_fd = mkstemp(name_template);
-    filename_ = name_template;
-# endif  // GTEST_OS_WINDOWS
-    fflush(NULL);
-    dup2(captured_fd, fd_);
-    close(captured_fd);
-  }
-
-  ~CapturedStream() {
-    remove(filename_.c_str());
-  }
-
-  std::string GetCapturedString() {
-    if (uncaptured_fd_ != -1) {
-      // Restores the original stream.
-      fflush(NULL);
-      dup2(uncaptured_fd_, fd_);
-      close(uncaptured_fd_);
-      uncaptured_fd_ = -1;
-    }
-
-    FILE* const file = posix::FOpen(filename_.c_str(), "r");
-    const std::string content = ReadEntireFile(file);
-    posix::FClose(file);
-    return content;
-  }
-
- private:
-  const int fd_;  // A stream to capture.
-  int uncaptured_fd_;
-  // Name of the temporary file holding the stderr output.
-  ::std::string filename_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(CapturedStream);
-};
-
-GTEST_DISABLE_MSC_WARNINGS_POP_()
-
-static CapturedStream* g_captured_stderr = NULL;
-static CapturedStream* g_captured_stdout = NULL;
-
-// Starts capturing an output stream (stdout/stderr).
-static void CaptureStream(int fd, const char* stream_name,
-                          CapturedStream** stream) {
-  if (*stream != NULL) {
-    GTEST_LOG_(FATAL) << "Only one " << stream_name
-                      << " capturer can exist at a time.";
-  }
-  *stream = new CapturedStream(fd);
-}
-
-// Stops capturing the output stream and returns the captured string.
-static std::string GetCapturedStream(CapturedStream** captured_stream) {
-  const std::string content = (*captured_stream)->GetCapturedString();
-
-  delete *captured_stream;
-  *captured_stream = NULL;
-
-  return content;
-}
-
-// Starts capturing stdout.
-void CaptureStdout() {
-  CaptureStream(kStdOutFileno, "stdout", &g_captured_stdout);
-}
-
-// Starts capturing stderr.
-void CaptureStderr() {
-  CaptureStream(kStdErrFileno, "stderr", &g_captured_stderr);
-}
-
-// Stops capturing stdout and returns the captured string.
-std::string GetCapturedStdout() {
-  return GetCapturedStream(&g_captured_stdout);
-}
-
-// Stops capturing stderr and returns the captured string.
-std::string GetCapturedStderr() {
-  return GetCapturedStream(&g_captured_stderr);
-}
-
-#endif  // GTEST_HAS_STREAM_REDIRECTION
-
-
-
-
-
-size_t GetFileSize(FILE* file) {
-  fseek(file, 0, SEEK_END);
-  return static_cast<size_t>(ftell(file));
-}
-
-std::string ReadEntireFile(FILE* file) {
-  const size_t file_size = GetFileSize(file);
-  char* const buffer = new char[file_size];
-
-  size_t bytes_last_read = 0;  // # of bytes read in the last fread()
-  size_t bytes_read = 0;       // # of bytes read so far
-
-  fseek(file, 0, SEEK_SET);
-
-  // Keeps reading the file until we cannot read further or the
-  // pre-determined file size is reached.
-  do {
-    bytes_last_read = fread(buffer+bytes_read, 1, file_size-bytes_read, file);
-    bytes_read += bytes_last_read;
-  } while (bytes_last_read > 0 && bytes_read < file_size);
-
-  const std::string content(buffer, bytes_read);
-  delete[] buffer;
-
-  return content;
-}
-
-#if GTEST_HAS_DEATH_TEST
-static const std::vector<std::string>* g_injected_test_argvs = NULL;  // Owned.
-
-std::vector<std::string> GetInjectableArgvs() {
-  if (g_injected_test_argvs != NULL) {
-    return *g_injected_test_argvs;
-  }
-  return GetArgvs();
-}
-
-void SetInjectableArgvs(const std::vector<std::string>* new_argvs) {
-  if (g_injected_test_argvs != new_argvs) delete g_injected_test_argvs;
-  g_injected_test_argvs = new_argvs;
-}
-
-void SetInjectableArgvs(const std::vector<std::string>& new_argvs) {
-  SetInjectableArgvs(
-      new std::vector<std::string>(new_argvs.begin(), new_argvs.end()));
-}
-
-#if GTEST_HAS_GLOBAL_STRING
-void SetInjectableArgvs(const std::vector< ::string>& new_argvs) {
-  SetInjectableArgvs(
-      new std::vector<std::string>(new_argvs.begin(), new_argvs.end()));
-}
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-void ClearInjectableArgvs() {
-  delete g_injected_test_argvs;
-  g_injected_test_argvs = NULL;
-}
-#endif  // GTEST_HAS_DEATH_TEST
-
-#if GTEST_OS_WINDOWS_MOBILE
-namespace posix {
-void Abort() {
-  DebugBreak();
-  TerminateProcess(GetCurrentProcess(), 1);
-}
-}  // namespace posix
-#endif  // GTEST_OS_WINDOWS_MOBILE
-
-// Returns the name of the environment variable corresponding to the
-// given flag.  For example, FlagToEnvVar("foo") will return
-// "GTEST_FOO" in the open-source version.
-static std::string FlagToEnvVar(const char* flag) {
-  const std::string full_flag =
-      (Message() << GTEST_FLAG_PREFIX_ << flag).GetString();
-
-  Message env_var;
-  for (size_t i = 0; i != full_flag.length(); i++) {
-    env_var << ToUpper(full_flag.c_str()[i]);
-  }
-
-  return env_var.GetString();
-}
-
-// Parses 'str' for a 32-bit signed integer.  If successful, writes
-// the result to *value and returns true; otherwise leaves *value
-// unchanged and returns false.
-bool ParseInt32(const Message& src_text, const char* str, Int32* value) {
-  // Parses the environment variable as a decimal integer.
-  char* end = NULL;
-  const long long_value = strtol(str, &end, 10);  // NOLINT
-
-  // Has strtol() consumed all characters in the string?
-  if (*end != '\0') {
-    // No - an invalid character was encountered.
-    Message msg;
-    msg << "WARNING: " << src_text
-        << " is expected to be a 32-bit integer, but actually"
-        << " has value \"" << str << "\".\n";
-    printf("%s", msg.GetString().c_str());
-    fflush(stdout);
-    return false;
-  }
-
-  // Is the parsed value in the range of an Int32?
-  const Int32 result = static_cast<Int32>(long_value);
-  if (long_value == LONG_MAX || long_value == LONG_MIN ||
-      // The parsed value overflows as a long.  (strtol() returns
-      // LONG_MAX or LONG_MIN when the input overflows.)
-      result != long_value
-      // The parsed value overflows as an Int32.
-      ) {
-    Message msg;
-    msg << "WARNING: " << src_text
-        << " is expected to be a 32-bit integer, but actually"
-        << " has value " << str << ", which overflows.\n";
-    printf("%s", msg.GetString().c_str());
-    fflush(stdout);
-    return false;
-  }
-
-  *value = result;
-  return true;
-}
-
-// Reads and returns the Boolean environment variable corresponding to
-// the given flag; if it's not set, returns default_value.
-//
-// The value is considered true iff it's not "0".
-bool BoolFromGTestEnv(const char* flag, bool default_value) {
-#if defined(GTEST_GET_BOOL_FROM_ENV_)
-  return GTEST_GET_BOOL_FROM_ENV_(flag, default_value);
-#else
-  const std::string env_var = FlagToEnvVar(flag);
-  const char* const string_value = posix::GetEnv(env_var.c_str());
-  return string_value == NULL ?
-      default_value : strcmp(string_value, "0") != 0;
-#endif  // defined(GTEST_GET_BOOL_FROM_ENV_)
-}
-
-// Reads and returns a 32-bit integer stored in the environment
-// variable corresponding to the given flag; if it isn't set or
-// doesn't represent a valid 32-bit integer, returns default_value.
-Int32 Int32FromGTestEnv(const char* flag, Int32 default_value) {
-#if defined(GTEST_GET_INT32_FROM_ENV_)
-  return GTEST_GET_INT32_FROM_ENV_(flag, default_value);
-#else
-  const std::string env_var = FlagToEnvVar(flag);
-  const char* const string_value = posix::GetEnv(env_var.c_str());
-  if (string_value == NULL) {
-    // The environment variable is not set.
-    return default_value;
-  }
-
-  Int32 result = default_value;
-  if (!ParseInt32(Message() << "Environment variable " << env_var,
-                  string_value, &result)) {
-    printf("The default value %s is used.\n",
-           (Message() << default_value).GetString().c_str());
-    fflush(stdout);
-    return default_value;
-  }
-
-  return result;
-#endif  // defined(GTEST_GET_INT32_FROM_ENV_)
-}
-
-// As a special case for the 'output' flag, if GTEST_OUTPUT is not
-// set, we look for XML_OUTPUT_FILE, which is set by the Bazel build
-// system.  The value of XML_OUTPUT_FILE is a filename without the
-// "xml:" prefix of GTEST_OUTPUT.
-// Note that this is meant to be called at the call site so it does
-// not check that the flag is 'output'
-// In essence this checks an env variable called XML_OUTPUT_FILE
-// and if it is set we prepend "xml:" to its value, if it not set we return ""
-std::string OutputFlagAlsoCheckEnvVar(){
-  std::string default_value_for_output_flag = "";
-  const char* xml_output_file_env = posix::GetEnv("XML_OUTPUT_FILE");
-  if (NULL != xml_output_file_env) {
-    default_value_for_output_flag = std::string("xml:") + xml_output_file_env;
-  }
-  return default_value_for_output_flag;
-}
-
-// Reads and returns the string environment variable corresponding to
-// the given flag; if it's not set, returns default_value.
-const char* StringFromGTestEnv(const char* flag, const char* default_value) {
-#if defined(GTEST_GET_STRING_FROM_ENV_)
-  return GTEST_GET_STRING_FROM_ENV_(flag, default_value);
-#else
-  const std::string env_var = FlagToEnvVar(flag);
-  const char* const value = posix::GetEnv(env_var.c_str());
-  return value == NULL ? default_value : value;
-#endif  // defined(GTEST_GET_STRING_FROM_ENV_)
-}
-
-}  // namespace internal
-}  // namespace testing
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Test - The Google C++ Testing and Mocking Framework
-//
-// This file implements a universal value printer that can print a
-// value of any type T:
-//
-//   void ::testing::internal::UniversalPrinter<T>::Print(value, ostream_ptr);
-//
-// It uses the << operator when possible, and prints the bytes in the
-// object otherwise.  A user can override its behavior for a class
-// type Foo by defining either operator<<(::std::ostream&, const Foo&)
-// or void PrintTo(const Foo&, ::std::ostream*) in the namespace that
-// defines Foo.
-
-#include <stdio.h>
-#include <cctype>
-#include <cwchar>
-#include <ostream>  // NOLINT
-#include <string>
-
-namespace testing {
-
-namespace {
-
-using ::std::ostream;
-
-// Prints a segment of bytes in the given object.
-GTEST_ATTRIBUTE_NO_SANITIZE_MEMORY_
-GTEST_ATTRIBUTE_NO_SANITIZE_ADDRESS_
-GTEST_ATTRIBUTE_NO_SANITIZE_THREAD_
-void PrintByteSegmentInObjectTo(const unsigned char* obj_bytes, size_t start,
-                                size_t count, ostream* os) {
-  char text[5] = "";
-  for (size_t i = 0; i != count; i++) {
-    const size_t j = start + i;
-    if (i != 0) {
-      // Organizes the bytes into groups of 2 for easy parsing by
-      // human.
-      if ((j % 2) == 0)
-        *os << ' ';
-      else
-        *os << '-';
-    }
-    GTEST_SNPRINTF_(text, sizeof(text), "%02X", obj_bytes[j]);
-    *os << text;
-  }
-}
-
-// Prints the bytes in the given value to the given ostream.
-void PrintBytesInObjectToImpl(const unsigned char* obj_bytes, size_t count,
-                              ostream* os) {
-  // Tells the user how big the object is.
-  *os << count << "-byte object <";
-
-  const size_t kThreshold = 132;
-  const size_t kChunkSize = 64;
-  // If the object size is bigger than kThreshold, we'll have to omit
-  // some details by printing only the first and the last kChunkSize
-  // bytes.
-  // TODO(wan): let the user control the threshold using a flag.
-  if (count < kThreshold) {
-    PrintByteSegmentInObjectTo(obj_bytes, 0, count, os);
-  } else {
-    PrintByteSegmentInObjectTo(obj_bytes, 0, kChunkSize, os);
-    *os << " ... ";
-    // Rounds up to 2-byte boundary.
-    const size_t resume_pos = (count - kChunkSize + 1)/2*2;
-    PrintByteSegmentInObjectTo(obj_bytes, resume_pos, count - resume_pos, os);
-  }
-  *os << ">";
-}
-
-}  // namespace
-
-namespace internal2 {
-
-// Delegates to PrintBytesInObjectToImpl() to print the bytes in the
-// given object.  The delegation simplifies the implementation, which
-// uses the << operator and thus is easier done outside of the
-// ::testing::internal namespace, which contains a << operator that
-// sometimes conflicts with the one in STL.
-void PrintBytesInObjectTo(const unsigned char* obj_bytes, size_t count,
-                          ostream* os) {
-  PrintBytesInObjectToImpl(obj_bytes, count, os);
-}
-
-}  // namespace internal2
-
-namespace internal {
-
-// Depending on the value of a char (or wchar_t), we print it in one
-// of three formats:
-//   - as is if it's a printable ASCII (e.g. 'a', '2', ' '),
-//   - as a hexadecimal escape sequence (e.g. '\x7F'), or
-//   - as a special escape sequence (e.g. '\r', '\n').
-enum CharFormat {
-  kAsIs,
-  kHexEscape,
-  kSpecialEscape
-};
-
-// Returns true if c is a printable ASCII character.  We test the
-// value of c directly instead of calling isprint(), which is buggy on
-// Windows Mobile.
-inline bool IsPrintableAscii(wchar_t c) {
-  return 0x20 <= c && c <= 0x7E;
-}
-
-// Prints a wide or narrow char c as a character literal without the
-// quotes, escaping it when necessary; returns how c was formatted.
-// The template argument UnsignedChar is the unsigned version of Char,
-// which is the type of c.
-template <typename UnsignedChar, typename Char>
-static CharFormat PrintAsCharLiteralTo(Char c, ostream* os) {
-  switch (static_cast<wchar_t>(c)) {
-    case L'\0':
-      *os << "\\0";
-      break;
-    case L'\'':
-      *os << "\\'";
-      break;
-    case L'\\':
-      *os << "\\\\";
-      break;
-    case L'\a':
-      *os << "\\a";
-      break;
-    case L'\b':
-      *os << "\\b";
-      break;
-    case L'\f':
-      *os << "\\f";
-      break;
-    case L'\n':
-      *os << "\\n";
-      break;
-    case L'\r':
-      *os << "\\r";
-      break;
-    case L'\t':
-      *os << "\\t";
-      break;
-    case L'\v':
-      *os << "\\v";
-      break;
-    default:
-      if (IsPrintableAscii(c)) {
-        *os << static_cast<char>(c);
-        return kAsIs;
-      } else {
-        ostream::fmtflags flags = os->flags();
-        *os << "\\x" << std::hex << std::uppercase
-            << static_cast<int>(static_cast<UnsignedChar>(c));
-        os->flags(flags);
-        return kHexEscape;
-      }
-  }
-  return kSpecialEscape;
-}
-
-// Prints a wchar_t c as if it's part of a string literal, escaping it when
-// necessary; returns how c was formatted.
-static CharFormat PrintAsStringLiteralTo(wchar_t c, ostream* os) {
-  switch (c) {
-    case L'\'':
-      *os << "'";
-      return kAsIs;
-    case L'"':
-      *os << "\\\"";
-      return kSpecialEscape;
-    default:
-      return PrintAsCharLiteralTo<wchar_t>(c, os);
-  }
-}
-
-// Prints a char c as if it's part of a string literal, escaping it when
-// necessary; returns how c was formatted.
-static CharFormat PrintAsStringLiteralTo(char c, ostream* os) {
-  return PrintAsStringLiteralTo(
-      static_cast<wchar_t>(static_cast<unsigned char>(c)), os);
-}
-
-// Prints a wide or narrow character c and its code.  '\0' is printed
-// as "'\\0'", other unprintable characters are also properly escaped
-// using the standard C++ escape sequence.  The template argument
-// UnsignedChar is the unsigned version of Char, which is the type of c.
-template <typename UnsignedChar, typename Char>
-void PrintCharAndCodeTo(Char c, ostream* os) {
-  // First, print c as a literal in the most readable form we can find.
-  *os << ((sizeof(c) > 1) ? "L'" : "'");
-  const CharFormat format = PrintAsCharLiteralTo<UnsignedChar>(c, os);
-  *os << "'";
-
-  // To aid user debugging, we also print c's code in decimal, unless
-  // it's 0 (in which case c was printed as '\\0', making the code
-  // obvious).
-  if (c == 0)
-    return;
-  *os << " (" << static_cast<int>(c);
-
-  // For more convenience, we print c's code again in hexadecimal,
-  // unless c was already printed in the form '\x##' or the code is in
-  // [1, 9].
-  if (format == kHexEscape || (1 <= c && c <= 9)) {
-    // Do nothing.
-  } else {
-    *os << ", 0x" << String::FormatHexInt(static_cast<UnsignedChar>(c));
-  }
-  *os << ")";
-}
-
-void PrintTo(unsigned char c, ::std::ostream* os) {
-  PrintCharAndCodeTo<unsigned char>(c, os);
-}
-void PrintTo(signed char c, ::std::ostream* os) {
-  PrintCharAndCodeTo<unsigned char>(c, os);
-}
-
-// Prints a wchar_t as a symbol if it is printable or as its internal
-// code otherwise and also as its code.  L'\0' is printed as "L'\\0'".
-void PrintTo(wchar_t wc, ostream* os) {
-  PrintCharAndCodeTo<wchar_t>(wc, os);
-}
-
-// Prints the given array of characters to the ostream.  CharType must be either
-// char or wchar_t.
-// The array starts at begin, the length is len, it may include '\0' characters
-// and may not be NUL-terminated.
-template <typename CharType>
-GTEST_ATTRIBUTE_NO_SANITIZE_MEMORY_
-GTEST_ATTRIBUTE_NO_SANITIZE_ADDRESS_
-GTEST_ATTRIBUTE_NO_SANITIZE_THREAD_
-static CharFormat PrintCharsAsStringTo(
-    const CharType* begin, size_t len, ostream* os) {
-  const char* const kQuoteBegin = sizeof(CharType) == 1 ? "\"" : "L\"";
-  *os << kQuoteBegin;
-  bool is_previous_hex = false;
-  CharFormat print_format = kAsIs;
-  for (size_t index = 0; index < len; ++index) {
-    const CharType cur = begin[index];
-    if (is_previous_hex && IsXDigit(cur)) {
-      // Previous character is of '\x..' form and this character can be
-      // interpreted as another hexadecimal digit in its number. Break string to
-      // disambiguate.
-      *os << "\" " << kQuoteBegin;
-    }
-    is_previous_hex = PrintAsStringLiteralTo(cur, os) == kHexEscape;
-    // Remember if any characters required hex escaping.
-    if (is_previous_hex) {
-      print_format = kHexEscape;
-    }
-  }
-  *os << "\"";
-  return print_format;
-}
-
-// Prints a (const) char/wchar_t array of 'len' elements, starting at address
-// 'begin'.  CharType must be either char or wchar_t.
-template <typename CharType>
-GTEST_ATTRIBUTE_NO_SANITIZE_MEMORY_
-GTEST_ATTRIBUTE_NO_SANITIZE_ADDRESS_
-GTEST_ATTRIBUTE_NO_SANITIZE_THREAD_
-static void UniversalPrintCharArray(
-    const CharType* begin, size_t len, ostream* os) {
-  // The code
-  //   const char kFoo[] = "foo";
-  // generates an array of 4, not 3, elements, with the last one being '\0'.
-  //
-  // Therefore when printing a char array, we don't print the last element if
-  // it's '\0', such that the output matches the string literal as it's
-  // written in the source code.
-  if (len > 0 && begin[len - 1] == '\0') {
-    PrintCharsAsStringTo(begin, len - 1, os);
-    return;
-  }
-
-  // If, however, the last element in the array is not '\0', e.g.
-  //    const char kFoo[] = { 'f', 'o', 'o' };
-  // we must print the entire array.  We also print a message to indicate
-  // that the array is not NUL-terminated.
-  PrintCharsAsStringTo(begin, len, os);
-  *os << " (no terminating NUL)";
-}
-
-// Prints a (const) char array of 'len' elements, starting at address 'begin'.
-void UniversalPrintArray(const char* begin, size_t len, ostream* os) {
-  UniversalPrintCharArray(begin, len, os);
-}
-
-// Prints a (const) wchar_t array of 'len' elements, starting at address
-// 'begin'.
-void UniversalPrintArray(const wchar_t* begin, size_t len, ostream* os) {
-  UniversalPrintCharArray(begin, len, os);
-}
-
-// Prints the given C string to the ostream.
-void PrintTo(const char* s, ostream* os) {
-  if (s == NULL) {
-    *os << "NULL";
-  } else {
-    *os << ImplicitCast_<const void*>(s) << " pointing to ";
-    PrintCharsAsStringTo(s, strlen(s), os);
-  }
-}
-
-// MSVC compiler can be configured to define whar_t as a typedef
-// of unsigned short. Defining an overload for const wchar_t* in that case
-// would cause pointers to unsigned shorts be printed as wide strings,
-// possibly accessing more memory than intended and causing invalid
-// memory accesses. MSVC defines _NATIVE_WCHAR_T_DEFINED symbol when
-// wchar_t is implemented as a native type.
-#if !defined(_MSC_VER) || defined(_NATIVE_WCHAR_T_DEFINED)
-// Prints the given wide C string to the ostream.
-void PrintTo(const wchar_t* s, ostream* os) {
-  if (s == NULL) {
-    *os << "NULL";
-  } else {
-    *os << ImplicitCast_<const void*>(s) << " pointing to ";
-    PrintCharsAsStringTo(s, std::wcslen(s), os);
-  }
-}
-#endif  // wchar_t is native
-
-namespace {
-
-bool ContainsUnprintableControlCodes(const char* str, size_t length) {
-  const unsigned char *s = reinterpret_cast<const unsigned char *>(str);
-
-  for (size_t i = 0; i < length; i++) {
-    unsigned char ch = *s++;
-    if (std::iscntrl(ch)) {
-        switch (ch) {
-        case '\t':
-        case '\n':
-        case '\r':
-          break;
-        default:
-          return true;
-        }
-      }
-  }
-  return false;
-}
-
-bool IsUTF8TrailByte(unsigned char t) { return 0x80 <= t && t<= 0xbf; }
-
-bool IsValidUTF8(const char* str, size_t length) {
-  const unsigned char *s = reinterpret_cast<const unsigned char *>(str);
-
-  for (size_t i = 0; i < length;) {
-    unsigned char lead = s[i++];
-
-    if (lead <= 0x7f) {
-      continue;  // single-byte character (ASCII) 0..7F
-    }
-    if (lead < 0xc2) {
-      return false;  // trail byte or non-shortest form
-    } else if (lead <= 0xdf && (i + 1) <= length && IsUTF8TrailByte(s[i])) {
-      ++i;  // 2-byte character
-    } else if (0xe0 <= lead && lead <= 0xef && (i + 2) <= length &&
-               IsUTF8TrailByte(s[i]) &&
-               IsUTF8TrailByte(s[i + 1]) &&
-               // check for non-shortest form and surrogate
-               (lead != 0xe0 || s[i] >= 0xa0) &&
-               (lead != 0xed || s[i] < 0xa0)) {
-      i += 2;  // 3-byte character
-    } else if (0xf0 <= lead && lead <= 0xf4 && (i + 3) <= length &&
-               IsUTF8TrailByte(s[i]) &&
-               IsUTF8TrailByte(s[i + 1]) &&
-               IsUTF8TrailByte(s[i + 2]) &&
-               // check for non-shortest form
-               (lead != 0xf0 || s[i] >= 0x90) &&
-               (lead != 0xf4 || s[i] < 0x90)) {
-      i += 3;  // 4-byte character
-    } else {
-      return false;
-    }
-  }
-  return true;
-}
-
-void ConditionalPrintAsText(const char* str, size_t length, ostream* os) {
-  if (!ContainsUnprintableControlCodes(str, length) &&
-      IsValidUTF8(str, length)) {
-    *os << "\n    As Text: \"" << str << "\"";
-  }
-}
-
-}  // anonymous namespace
-
-// Prints a ::string object.
-#if GTEST_HAS_GLOBAL_STRING
-void PrintStringTo(const ::string& s, ostream* os) {
-  if (PrintCharsAsStringTo(s.data(), s.size(), os) == kHexEscape) {
-    if (GTEST_FLAG(print_utf8)) {
-      ConditionalPrintAsText(s.data(), s.size(), os);
-    }
-  }
-}
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-void PrintStringTo(const ::std::string& s, ostream* os) {
-  if (PrintCharsAsStringTo(s.data(), s.size(), os) == kHexEscape) {
-    if (GTEST_FLAG(print_utf8)) {
-      ConditionalPrintAsText(s.data(), s.size(), os);
-    }
-  }
-}
-
-// Prints a ::wstring object.
-#if GTEST_HAS_GLOBAL_WSTRING
-void PrintWideStringTo(const ::wstring& s, ostream* os) {
-  PrintCharsAsStringTo(s.data(), s.size(), os);
-}
-#endif  // GTEST_HAS_GLOBAL_WSTRING
-
-#if GTEST_HAS_STD_WSTRING
-void PrintWideStringTo(const ::std::wstring& s, ostream* os) {
-  PrintCharsAsStringTo(s.data(), s.size(), os);
-}
-#endif  // GTEST_HAS_STD_WSTRING
-
-}  // namespace internal
-
-}  // namespace testing
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: mheule@google.com (Markus Heule)
-//
-// The Google C++ Testing and Mocking Framework (Google Test)
-
-
-namespace testing {
-
-using internal::GetUnitTestImpl;
-
-// Gets the summary of the failure message by omitting the stack trace
-// in it.
-std::string TestPartResult::ExtractSummary(const char* message) {
-  const char* const stack_trace = strstr(message, internal::kStackTraceMarker);
-  return stack_trace == NULL ? message :
-      std::string(message, stack_trace);
-}
-
-// Prints a TestPartResult object.
-std::ostream& operator<<(std::ostream& os, const TestPartResult& result) {
-  return os
-      << result.file_name() << ":" << result.line_number() << ": "
-      << (result.type() == TestPartResult::kSuccess ? "Success" :
-          result.type() == TestPartResult::kFatalFailure ? "Fatal failure" :
-          "Non-fatal failure") << ":\n"
-      << result.message() << std::endl;
-}
-
-// Appends a TestPartResult to the array.
-void TestPartResultArray::Append(const TestPartResult& result) {
-  array_.push_back(result);
-}
-
-// Returns the TestPartResult at the given index (0-based).
-const TestPartResult& TestPartResultArray::GetTestPartResult(int index) const {
-  if (index < 0 || index >= size()) {
-    printf("\nInvalid index (%d) into TestPartResultArray.\n", index);
-    internal::posix::Abort();
-  }
-
-  return array_[index];
-}
-
-// Returns the number of TestPartResult objects in the array.
-int TestPartResultArray::size() const {
-  return static_cast<int>(array_.size());
-}
-
-namespace internal {
-
-HasNewFatalFailureHelper::HasNewFatalFailureHelper()
-    : has_new_fatal_failure_(false),
-      original_reporter_(GetUnitTestImpl()->
-                         GetTestPartResultReporterForCurrentThread()) {
-  GetUnitTestImpl()->SetTestPartResultReporterForCurrentThread(this);
-}
-
-HasNewFatalFailureHelper::~HasNewFatalFailureHelper() {
-  GetUnitTestImpl()->SetTestPartResultReporterForCurrentThread(
-      original_reporter_);
-}
-
-void HasNewFatalFailureHelper::ReportTestPartResult(
-    const TestPartResult& result) {
-  if (result.fatally_failed())
-    has_new_fatal_failure_ = true;
-  original_reporter_->ReportTestPartResult(result);
-}
-
-}  // namespace internal
-
-}  // namespace testing
-// Copyright 2008 Google Inc.
-// All Rights Reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-
-
-namespace testing {
-namespace internal {
-
-#if GTEST_HAS_TYPED_TEST_P
-
-// Skips to the first non-space char in str. Returns an empty string if str
-// contains only whitespace characters.
-static const char* SkipSpaces(const char* str) {
-  while (IsSpace(*str))
-    str++;
-  return str;
-}
-
-static std::vector<std::string> SplitIntoTestNames(const char* src) {
-  std::vector<std::string> name_vec;
-  src = SkipSpaces(src);
-  for (; src != NULL; src = SkipComma(src)) {
-    name_vec.push_back(StripTrailingSpaces(GetPrefixUntilComma(src)));
-  }
-  return name_vec;
-}
-
-// Verifies that registered_tests match the test names in
-// registered_tests_; returns registered_tests if successful, or
-// aborts the program otherwise.
-const char* TypedTestCasePState::VerifyRegisteredTestNames(
-    const char* file, int line, const char* registered_tests) {
-  typedef RegisteredTestsMap::const_iterator RegisteredTestIter;
-  registered_ = true;
-
-  std::vector<std::string> name_vec = SplitIntoTestNames(registered_tests);
-
-  Message errors;
-
-  std::set<std::string> tests;
-  for (std::vector<std::string>::const_iterator name_it = name_vec.begin();
-       name_it != name_vec.end(); ++name_it) {
-    const std::string& name = *name_it;
-    if (tests.count(name) != 0) {
-      errors << "Test " << name << " is listed more than once.\n";
-      continue;
-    }
-
-    bool found = false;
-    for (RegisteredTestIter it = registered_tests_.begin();
-         it != registered_tests_.end();
-         ++it) {
-      if (name == it->first) {
-        found = true;
-        break;
-      }
-    }
-
-    if (found) {
-      tests.insert(name);
-    } else {
-      errors << "No test named " << name
-             << " can be found in this test case.\n";
-    }
-  }
-
-  for (RegisteredTestIter it = registered_tests_.begin();
-       it != registered_tests_.end();
-       ++it) {
-    if (tests.count(it->first) == 0) {
-      errors << "You forgot to list test " << it->first << ".\n";
-    }
-  }
-
-  const std::string& errors_str = errors.GetString();
-  if (errors_str != "") {
-    fprintf(stderr, "%s %s", FormatFileLocation(file, line).c_str(),
-            errors_str.c_str());
-    fflush(stderr);
-    posix::Abort();
-  }
-
-  return registered_tests;
-}
-
-#endif  // GTEST_HAS_TYPED_TEST_P
-
-}  // namespace internal
-}  // namespace testing
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-//
-// Google C++ Mocking Framework (Google Mock)
-//
-// This file #includes all Google Mock implementation .cc files.  The
-// purpose is to allow a user to build Google Mock by compiling this
-// file alone.
-
-// This line ensures that gmock.h can be compiled on its own, even
-// when it's fused.
-#include "gmock/gmock.h"
-
-// The following lines pull in the real gmock *.cc files.
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements cardinalities.
-
-
-#include <limits.h>
-#include <ostream>  // NOLINT
-#include <sstream>
-#include <string>
-
-namespace testing {
-
-namespace {
-
-// Implements the Between(m, n) cardinality.
-class BetweenCardinalityImpl : public CardinalityInterface {
- public:
-  BetweenCardinalityImpl(int min, int max)
-      : min_(min >= 0 ? min : 0),
-        max_(max >= min_ ? max : min_) {
-    std::stringstream ss;
-    if (min < 0) {
-      ss << "The invocation lower bound must be >= 0, "
-         << "but is actually " << min << ".";
-      internal::Expect(false, __FILE__, __LINE__, ss.str());
-    } else if (max < 0) {
-      ss << "The invocation upper bound must be >= 0, "
-         << "but is actually " << max << ".";
-      internal::Expect(false, __FILE__, __LINE__, ss.str());
-    } else if (min > max) {
-      ss << "The invocation upper bound (" << max
-         << ") must be >= the invocation lower bound (" << min
-         << ").";
-      internal::Expect(false, __FILE__, __LINE__, ss.str());
-    }
-  }
-
-  // Conservative estimate on the lower/upper bound of the number of
-  // calls allowed.
-  virtual int ConservativeLowerBound() const { return min_; }
-  virtual int ConservativeUpperBound() const { return max_; }
-
-  virtual bool IsSatisfiedByCallCount(int call_count) const {
-    return min_ <= call_count && call_count <= max_;
-  }
-
-  virtual bool IsSaturatedByCallCount(int call_count) const {
-    return call_count >= max_;
-  }
-
-  virtual void DescribeTo(::std::ostream* os) const;
-
- private:
-  const int min_;
-  const int max_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(BetweenCardinalityImpl);
-};
-
-// Formats "n times" in a human-friendly way.
-inline std::string FormatTimes(int n) {
-  if (n == 1) {
-    return "once";
-  } else if (n == 2) {
-    return "twice";
-  } else {
-    std::stringstream ss;
-    ss << n << " times";
-    return ss.str();
-  }
-}
-
-// Describes the Between(m, n) cardinality in human-friendly text.
-void BetweenCardinalityImpl::DescribeTo(::std::ostream* os) const {
-  if (min_ == 0) {
-    if (max_ == 0) {
-      *os << "never called";
-    } else if (max_ == INT_MAX) {
-      *os << "called any number of times";
-    } else {
-      *os << "called at most " << FormatTimes(max_);
-    }
-  } else if (min_ == max_) {
-    *os << "called " << FormatTimes(min_);
-  } else if (max_ == INT_MAX) {
-    *os << "called at least " << FormatTimes(min_);
-  } else {
-    // 0 < min_ < max_ < INT_MAX
-    *os << "called between " << min_ << " and " << max_ << " times";
-  }
-}
-
-}  // Unnamed namespace
-
-// Describes the given call count to an ostream.
-void Cardinality::DescribeActualCallCountTo(int actual_call_count,
-                                            ::std::ostream* os) {
-  if (actual_call_count > 0) {
-    *os << "called " << FormatTimes(actual_call_count);
-  } else {
-    *os << "never called";
-  }
-}
-
-// Creates a cardinality that allows at least n calls.
-GTEST_API_ Cardinality AtLeast(int n) { return Between(n, INT_MAX); }
-
-// Creates a cardinality that allows at most n calls.
-GTEST_API_ Cardinality AtMost(int n) { return Between(0, n); }
-
-// Creates a cardinality that allows any number of calls.
-GTEST_API_ Cardinality AnyNumber() { return AtLeast(0); }
-
-// Creates a cardinality that allows between min and max calls.
-GTEST_API_ Cardinality Between(int min, int max) {
-  return Cardinality(new BetweenCardinalityImpl(min, max));
-}
-
-// Creates a cardinality that allows exactly n calls.
-GTEST_API_ Cardinality Exactly(int n) { return Between(n, n); }
-
-}  // namespace testing
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file defines some utilities useful for implementing Google
-// Mock.  They are subject to change without notice, so please DO NOT
-// USE THEM IN USER CODE.
-
-
-#include <ctype.h>
-#include <ostream>  // NOLINT
-#include <string>
-
-namespace testing {
-namespace internal {
-
-// Joins a vector of strings as if they are fields of a tuple; returns
-// the joined string.
-GTEST_API_ std::string JoinAsTuple(const Strings& fields) {
-  switch (fields.size()) {
-    case 0:
-      return "";
-    case 1:
-      return fields[0];
-    default:
-      std::string result = "(" + fields[0];
-      for (size_t i = 1; i < fields.size(); i++) {
-        result += ", ";
-        result += fields[i];
-      }
-      result += ")";
-      return result;
-  }
-}
-
-// Converts an identifier name to a space-separated list of lower-case
-// words.  Each maximum substring of the form [A-Za-z][a-z]*|\d+ is
-// treated as one word.  For example, both "FooBar123" and
-// "foo_bar_123" are converted to "foo bar 123".
-GTEST_API_ std::string ConvertIdentifierNameToWords(const char* id_name) {
-  std::string result;
-  char prev_char = '\0';
-  for (const char* p = id_name; *p != '\0'; prev_char = *(p++)) {
-    // We don't care about the current locale as the input is
-    // guaranteed to be a valid C++ identifier name.
-    const bool starts_new_word = IsUpper(*p) ||
-        (!IsAlpha(prev_char) && IsLower(*p)) ||
-        (!IsDigit(prev_char) && IsDigit(*p));
-
-    if (IsAlNum(*p)) {
-      if (starts_new_word && result != "")
-        result += ' ';
-      result += ToLower(*p);
-    }
-  }
-  return result;
-}
-
-// This class reports Google Mock failures as Google Test failures.  A
-// user can define another class in a similar fashion if they intend to
-// use Google Mock with a testing framework other than Google Test.
-class GoogleTestFailureReporter : public FailureReporterInterface {
- public:
-  virtual void ReportFailure(FailureType type, const char* file, int line,
-                             const std::string& message) {
-    AssertHelper(type == kFatal ?
-                 TestPartResult::kFatalFailure :
-                 TestPartResult::kNonFatalFailure,
-                 file,
-                 line,
-                 message.c_str()) = Message();
-    if (type == kFatal) {
-      posix::Abort();
-    }
-  }
-};
-
-// Returns the global failure reporter.  Will create a
-// GoogleTestFailureReporter and return it the first time called.
-GTEST_API_ FailureReporterInterface* GetFailureReporter() {
-  // Points to the global failure reporter used by Google Mock.  gcc
-  // guarantees that the following use of failure_reporter is
-  // thread-safe.  We may need to add additional synchronization to
-  // protect failure_reporter if we port Google Mock to other
-  // compilers.
-  static FailureReporterInterface* const failure_reporter =
-      new GoogleTestFailureReporter();
-  return failure_reporter;
-}
-
-// Protects global resources (stdout in particular) used by Log().
-static GTEST_DEFINE_STATIC_MUTEX_(g_log_mutex);
-
-// Returns true iff a log with the given severity is visible according
-// to the --gmock_verbose flag.
-GTEST_API_ bool LogIsVisible(LogSeverity severity) {
-  if (GMOCK_FLAG(verbose) == kInfoVerbosity) {
-    // Always show the log if --gmock_verbose=info.
-    return true;
-  } else if (GMOCK_FLAG(verbose) == kErrorVerbosity) {
-    // Always hide it if --gmock_verbose=error.
-    return false;
-  } else {
-    // If --gmock_verbose is neither "info" nor "error", we treat it
-    // as "warning" (its default value).
-    return severity == kWarning;
-  }
-}
-
-// Prints the given message to stdout iff 'severity' >= the level
-// specified by the --gmock_verbose flag.  If stack_frames_to_skip >=
-// 0, also prints the stack trace excluding the top
-// stack_frames_to_skip frames.  In opt mode, any positive
-// stack_frames_to_skip is treated as 0, since we don't know which
-// function calls will be inlined by the compiler and need to be
-// conservative.
-GTEST_API_ void Log(LogSeverity severity, const std::string& message,
-                    int stack_frames_to_skip) {
-  if (!LogIsVisible(severity))
-    return;
-
-  // Ensures that logs from different threads don't interleave.
-  MutexLock l(&g_log_mutex);
-
-  // "using ::std::cout;" doesn't work with Symbian's STLport, where cout is a
-  // macro.
-
-  if (severity == kWarning) {
-    // Prints a GMOCK WARNING marker to make the warnings easily searchable.
-    std::cout << "\nGMOCK WARNING:";
-  }
-  // Pre-pends a new-line to message if it doesn't start with one.
-  if (message.empty() || message[0] != '\n') {
-    std::cout << "\n";
-  }
-  std::cout << message;
-  if (stack_frames_to_skip >= 0) {
-#ifdef NDEBUG
-    // In opt mode, we have to be conservative and skip no stack frame.
-    const int actual_to_skip = 0;
-#else
-    // In dbg mode, we can do what the caller tell us to do (plus one
-    // for skipping this function's stack frame).
-    const int actual_to_skip = stack_frames_to_skip + 1;
-#endif  // NDEBUG
-
-    // Appends a new-line to message if it doesn't end with one.
-    if (!message.empty() && *message.rbegin() != '\n') {
-      std::cout << "\n";
-    }
-    std::cout << "Stack trace:\n"
-         << ::testing::internal::GetCurrentOsStackTraceExceptTop(
-             ::testing::UnitTest::GetInstance(), actual_to_skip);
-  }
-  std::cout << ::std::flush;
-}
-
-GTEST_API_ WithoutMatchers GetWithoutMatchers() { return WithoutMatchers(); }
-
-GTEST_API_ void IllegalDoDefault(const char* file, int line) {
-  internal::Assert(
-      false, file, line,
-      "You are using DoDefault() inside a composite action like "
-      "DoAll() or WithArgs().  This is not supported for technical "
-      "reasons.  Please instead spell out the default action, or "
-      "assign the default action to an Action variable and use "
-      "the variable in various places.");
-}
-
-}  // namespace internal
-}  // namespace testing
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements Matcher<const string&>, Matcher<string>, and
-// utilities for defining matchers.
-
-
-#include <string.h>
-#include <iostream>
-#include <sstream>
-#include <string>
-
-namespace testing {
-
-// Constructs a matcher that matches a const std::string& whose value is
-// equal to s.
-Matcher<const std::string&>::Matcher(const std::string& s) { *this = Eq(s); }
-
-#if GTEST_HAS_GLOBAL_STRING
-// Constructs a matcher that matches a const std::string& whose value is
-// equal to s.
-Matcher<const std::string&>::Matcher(const ::string& s) {
-  *this = Eq(static_cast<std::string>(s));
-}
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-// Constructs a matcher that matches a const std::string& whose value is
-// equal to s.
-Matcher<const std::string&>::Matcher(const char* s) {
-  *this = Eq(std::string(s));
-}
-
-// Constructs a matcher that matches a std::string whose value is equal to
-// s.
-Matcher<std::string>::Matcher(const std::string& s) { *this = Eq(s); }
-
-#if GTEST_HAS_GLOBAL_STRING
-// Constructs a matcher that matches a std::string whose value is equal to
-// s.
-Matcher<std::string>::Matcher(const ::string& s) {
-  *this = Eq(static_cast<std::string>(s));
-}
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-// Constructs a matcher that matches a std::string whose value is equal to
-// s.
-Matcher<std::string>::Matcher(const char* s) { *this = Eq(std::string(s)); }
-
-#if GTEST_HAS_GLOBAL_STRING
-// Constructs a matcher that matches a const ::string& whose value is
-// equal to s.
-Matcher<const ::string&>::Matcher(const std::string& s) {
-  *this = Eq(static_cast<::string>(s));
-}
-
-// Constructs a matcher that matches a const ::string& whose value is
-// equal to s.
-Matcher<const ::string&>::Matcher(const ::string& s) { *this = Eq(s); }
-
-// Constructs a matcher that matches a const ::string& whose value is
-// equal to s.
-Matcher<const ::string&>::Matcher(const char* s) { *this = Eq(::string(s)); }
-
-// Constructs a matcher that matches a ::string whose value is equal to s.
-Matcher<::string>::Matcher(const std::string& s) {
-  *this = Eq(static_cast<::string>(s));
-}
-
-// Constructs a matcher that matches a ::string whose value is equal to s.
-Matcher<::string>::Matcher(const ::string& s) { *this = Eq(s); }
-
-// Constructs a matcher that matches a string whose value is equal to s.
-Matcher<::string>::Matcher(const char* s) { *this = Eq(::string(s)); }
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-#if GTEST_HAS_ABSL
-// Constructs a matcher that matches a const absl::string_view& whose value is
-// equal to s.
-Matcher<const absl::string_view&>::Matcher(const std::string& s) {
-  *this = Eq(s);
-}
-
-#if GTEST_HAS_GLOBAL_STRING
-// Constructs a matcher that matches a const absl::string_view& whose value is
-// equal to s.
-Matcher<const absl::string_view&>::Matcher(const ::string& s) { *this = Eq(s); }
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-// Constructs a matcher that matches a const absl::string_view& whose value is
-// equal to s.
-Matcher<const absl::string_view&>::Matcher(const char* s) {
-  *this = Eq(std::string(s));
-}
-
-// Constructs a matcher that matches a const absl::string_view& whose value is
-// equal to s.
-Matcher<const absl::string_view&>::Matcher(absl::string_view s) {
-  *this = Eq(std::string(s));
-}
-
-// Constructs a matcher that matches a absl::string_view whose value is equal to
-// s.
-Matcher<absl::string_view>::Matcher(const std::string& s) { *this = Eq(s); }
-
-#if GTEST_HAS_GLOBAL_STRING
-// Constructs a matcher that matches a absl::string_view whose value is equal to
-// s.
-Matcher<absl::string_view>::Matcher(const ::string& s) { *this = Eq(s); }
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-// Constructs a matcher that matches a absl::string_view whose value is equal to
-// s.
-Matcher<absl::string_view>::Matcher(const char* s) {
-  *this = Eq(std::string(s));
-}
-
-// Constructs a matcher that matches a absl::string_view whose value is equal to
-// s.
-Matcher<absl::string_view>::Matcher(absl::string_view s) {
-  *this = Eq(std::string(s));
-}
-#endif  // GTEST_HAS_ABSL
-
-namespace internal {
-
-// Returns the description for a matcher defined using the MATCHER*()
-// macro where the user-supplied description string is "", if
-// 'negation' is false; otherwise returns the description of the
-// negation of the matcher.  'param_values' contains a list of strings
-// that are the print-out of the matcher's parameters.
-GTEST_API_ std::string FormatMatcherDescription(bool negation,
-                                                const char* matcher_name,
-                                                const Strings& param_values) {
-  std::string result = ConvertIdentifierNameToWords(matcher_name);
-  if (param_values.size() >= 1) result += " " + JoinAsTuple(param_values);
-  return negation ? "not (" + result + ")" : result;
-}
-
-// FindMaxBipartiteMatching and its helper class.
-//
-// Uses the well-known Ford-Fulkerson max flow method to find a maximum
-// bipartite matching. Flow is considered to be from left to right.
-// There is an implicit source node that is connected to all of the left
-// nodes, and an implicit sink node that is connected to all of the
-// right nodes. All edges have unit capacity.
-//
-// Neither the flow graph nor the residual flow graph are represented
-// explicitly. Instead, they are implied by the information in 'graph' and
-// a vector<int> called 'left_' whose elements are initialized to the
-// value kUnused. This represents the initial state of the algorithm,
-// where the flow graph is empty, and the residual flow graph has the
-// following edges:
-//   - An edge from source to each left_ node
-//   - An edge from each right_ node to sink
-//   - An edge from each left_ node to each right_ node, if the
-//     corresponding edge exists in 'graph'.
-//
-// When the TryAugment() method adds a flow, it sets left_[l] = r for some
-// nodes l and r. This induces the following changes:
-//   - The edges (source, l), (l, r), and (r, sink) are added to the
-//     flow graph.
-//   - The same three edges are removed from the residual flow graph.
-//   - The reverse edges (l, source), (r, l), and (sink, r) are added
-//     to the residual flow graph, which is a directional graph
-//     representing unused flow capacity.
-//
-// When the method augments a flow (moving left_[l] from some r1 to some
-// other r2), this can be thought of as "undoing" the above steps with
-// respect to r1 and "redoing" them with respect to r2.
-//
-// It bears repeating that the flow graph and residual flow graph are
-// never represented explicitly, but can be derived by looking at the
-// information in 'graph' and in left_.
-//
-// As an optimization, there is a second vector<int> called right_ which
-// does not provide any new information. Instead, it enables more
-// efficient queries about edges entering or leaving the right-side nodes
-// of the flow or residual flow graphs. The following invariants are
-// maintained:
-//
-// left[l] == kUnused or right[left[l]] == l
-// right[r] == kUnused or left[right[r]] == r
-//
-// . [ source ]                                        .
-// .   |||                                             .
-// .   |||                                             .
-// .   ||\--> left[0]=1  ---\    right[0]=-1 ----\     .
-// .   ||                   |                    |     .
-// .   |\---> left[1]=-1    \--> right[1]=0  ---\|     .
-// .   |                                        ||     .
-// .   \----> left[2]=2  ------> right[2]=2  --\||     .
-// .                                           |||     .
-// .         elements           matchers       vvv     .
-// .                                         [ sink ]  .
-//
-// See Also:
-//   [1] Cormen, et al (2001). "Section 26.2: The Ford-Fulkerson method".
-//       "Introduction to Algorithms (Second ed.)", pp. 651-664.
-//   [2] "Ford-Fulkerson algorithm", Wikipedia,
-//       'http://en.wikipedia.org/wiki/Ford%E2%80%93Fulkerson_algorithm'
-class MaxBipartiteMatchState {
- public:
-  explicit MaxBipartiteMatchState(const MatchMatrix& graph)
-      : graph_(&graph),
-        left_(graph_->LhsSize(), kUnused),
-        right_(graph_->RhsSize(), kUnused) {}
-
-  // Returns the edges of a maximal match, each in the form {left, right}.
-  ElementMatcherPairs Compute() {
-    // 'seen' is used for path finding { 0: unseen, 1: seen }.
-    ::std::vector<char> seen;
-    // Searches the residual flow graph for a path from each left node to
-    // the sink in the residual flow graph, and if one is found, add flow
-    // to the graph. It's okay to search through the left nodes once. The
-    // edge from the implicit source node to each previously-visited left
-    // node will have flow if that left node has any path to the sink
-    // whatsoever. Subsequent augmentations can only add flow to the
-    // network, and cannot take away that previous flow unit from the source.
-    // Since the source-to-left edge can only carry one flow unit (or,
-    // each element can be matched to only one matcher), there is no need
-    // to visit the left nodes more than once looking for augmented paths.
-    // The flow is known to be possible or impossible by looking at the
-    // node once.
-    for (size_t ilhs = 0; ilhs < graph_->LhsSize(); ++ilhs) {
-      // Reset the path-marking vector and try to find a path from
-      // source to sink starting at the left_[ilhs] node.
-      GTEST_CHECK_(left_[ilhs] == kUnused)
-          << "ilhs: " << ilhs << ", left_[ilhs]: " << left_[ilhs];
-      // 'seen' initialized to 'graph_->RhsSize()' copies of 0.
-      seen.assign(graph_->RhsSize(), 0);
-      TryAugment(ilhs, &seen);
-    }
-    ElementMatcherPairs result;
-    for (size_t ilhs = 0; ilhs < left_.size(); ++ilhs) {
-      size_t irhs = left_[ilhs];
-      if (irhs == kUnused) continue;
-      result.push_back(ElementMatcherPair(ilhs, irhs));
-    }
-    return result;
-  }
-
- private:
-  static const size_t kUnused = static_cast<size_t>(-1);
-
-  // Perform a depth-first search from left node ilhs to the sink.  If a
-  // path is found, flow is added to the network by linking the left and
-  // right vector elements corresponding each segment of the path.
-  // Returns true if a path to sink was found, which means that a unit of
-  // flow was added to the network. The 'seen' vector elements correspond
-  // to right nodes and are marked to eliminate cycles from the search.
-  //
-  // Left nodes will only be explored at most once because they
-  // are accessible from at most one right node in the residual flow
-  // graph.
-  //
-  // Note that left_[ilhs] is the only element of left_ that TryAugment will
-  // potentially transition from kUnused to another value. Any other
-  // left_ element holding kUnused before TryAugment will be holding it
-  // when TryAugment returns.
-  //
-  bool TryAugment(size_t ilhs, ::std::vector<char>* seen) {
-    for (size_t irhs = 0; irhs < graph_->RhsSize(); ++irhs) {
-      if ((*seen)[irhs]) continue;
-      if (!graph_->HasEdge(ilhs, irhs)) continue;
-      // There's an available edge from ilhs to irhs.
-      (*seen)[irhs] = 1;
-      // Next a search is performed to determine whether
-      // this edge is a dead end or leads to the sink.
-      //
-      // right_[irhs] == kUnused means that there is residual flow from
-      // right node irhs to the sink, so we can use that to finish this
-      // flow path and return success.
-      //
-      // Otherwise there is residual flow to some ilhs. We push flow
-      // along that path and call ourselves recursively to see if this
-      // ultimately leads to sink.
-      if (right_[irhs] == kUnused || TryAugment(right_[irhs], seen)) {
-        // Add flow from left_[ilhs] to right_[irhs].
-        left_[ilhs] = irhs;
-        right_[irhs] = ilhs;
-        return true;
-      }
-    }
-    return false;
-  }
-
-  const MatchMatrix* graph_;  // not owned
-  // Each element of the left_ vector represents a left hand side node
-  // (i.e. an element) and each element of right_ is a right hand side
-  // node (i.e. a matcher). The values in the left_ vector indicate
-  // outflow from that node to a node on the right_ side. The values
-  // in the right_ indicate inflow, and specify which left_ node is
-  // feeding that right_ node, if any. For example, left_[3] == 1 means
-  // there's a flow from element #3 to matcher #1. Such a flow would also
-  // be redundantly represented in the right_ vector as right_[1] == 3.
-  // Elements of left_ and right_ are either kUnused or mutually
-  // referent. Mutually referent means that left_[right_[i]] = i and
-  // right_[left_[i]] = i.
-  ::std::vector<size_t> left_;
-  ::std::vector<size_t> right_;
-
-  GTEST_DISALLOW_ASSIGN_(MaxBipartiteMatchState);
-};
-
-const size_t MaxBipartiteMatchState::kUnused;
-
-GTEST_API_ ElementMatcherPairs FindMaxBipartiteMatching(const MatchMatrix& g) {
-  return MaxBipartiteMatchState(g).Compute();
-}
-
-static void LogElementMatcherPairVec(const ElementMatcherPairs& pairs,
-                                     ::std::ostream* stream) {
-  typedef ElementMatcherPairs::const_iterator Iter;
-  ::std::ostream& os = *stream;
-  os << "{";
-  const char* sep = "";
-  for (Iter it = pairs.begin(); it != pairs.end(); ++it) {
-    os << sep << "\n  ("
-       << "element #" << it->first << ", "
-       << "matcher #" << it->second << ")";
-    sep = ",";
-  }
-  os << "\n}";
-}
-
-bool MatchMatrix::NextGraph() {
-  for (size_t ilhs = 0; ilhs < LhsSize(); ++ilhs) {
-    for (size_t irhs = 0; irhs < RhsSize(); ++irhs) {
-      char& b = matched_[SpaceIndex(ilhs, irhs)];
-      if (!b) {
-        b = 1;
-        return true;
-      }
-      b = 0;
-    }
-  }
-  return false;
-}
-
-void MatchMatrix::Randomize() {
-  for (size_t ilhs = 0; ilhs < LhsSize(); ++ilhs) {
-    for (size_t irhs = 0; irhs < RhsSize(); ++irhs) {
-      char& b = matched_[SpaceIndex(ilhs, irhs)];
-      b = static_cast<char>(rand() & 1);  // NOLINT
-    }
-  }
-}
-
-std::string MatchMatrix::DebugString() const {
-  ::std::stringstream ss;
-  const char* sep = "";
-  for (size_t i = 0; i < LhsSize(); ++i) {
-    ss << sep;
-    for (size_t j = 0; j < RhsSize(); ++j) {
-      ss << HasEdge(i, j);
-    }
-    sep = ";";
-  }
-  return ss.str();
-}
-
-void UnorderedElementsAreMatcherImplBase::DescribeToImpl(
-    ::std::ostream* os) const {
-  switch (match_flags()) {
-    case UnorderedMatcherRequire::ExactMatch:
-      if (matcher_describers_.empty()) {
-        *os << "is empty";
-        return;
-      }
-      if (matcher_describers_.size() == 1) {
-        *os << "has " << Elements(1) << " and that element ";
-        matcher_describers_[0]->DescribeTo(os);
-        return;
-      }
-      *os << "has " << Elements(matcher_describers_.size())
-          << " and there exists some permutation of elements such that:\n";
-      break;
-    case UnorderedMatcherRequire::Superset:
-      *os << "a surjection from elements to requirements exists such that:\n";
-      break;
-    case UnorderedMatcherRequire::Subset:
-      *os << "an injection from elements to requirements exists such that:\n";
-      break;
-  }
-
-  const char* sep = "";
-  for (size_t i = 0; i != matcher_describers_.size(); ++i) {
-    *os << sep;
-    if (match_flags() == UnorderedMatcherRequire::ExactMatch) {
-      *os << " - element #" << i << " ";
-    } else {
-      *os << " - an element ";
-    }
-    matcher_describers_[i]->DescribeTo(os);
-    if (match_flags() == UnorderedMatcherRequire::ExactMatch) {
-      sep = ", and\n";
-    } else {
-      sep = "\n";
-    }
-  }
-}
-
-void UnorderedElementsAreMatcherImplBase::DescribeNegationToImpl(
-    ::std::ostream* os) const {
-  switch (match_flags()) {
-    case UnorderedMatcherRequire::ExactMatch:
-      if (matcher_describers_.empty()) {
-        *os << "isn't empty";
-        return;
-      }
-      if (matcher_describers_.size() == 1) {
-        *os << "doesn't have " << Elements(1) << ", or has " << Elements(1)
-            << " that ";
-        matcher_describers_[0]->DescribeNegationTo(os);
-        return;
-      }
-      *os << "doesn't have " << Elements(matcher_describers_.size())
-          << ", or there exists no permutation of elements such that:\n";
-      break;
-    case UnorderedMatcherRequire::Superset:
-      *os << "no surjection from elements to requirements exists such that:\n";
-      break;
-    case UnorderedMatcherRequire::Subset:
-      *os << "no injection from elements to requirements exists such that:\n";
-      break;
-  }
-  const char* sep = "";
-  for (size_t i = 0; i != matcher_describers_.size(); ++i) {
-    *os << sep;
-    if (match_flags() == UnorderedMatcherRequire::ExactMatch) {
-      *os << " - element #" << i << " ";
-    } else {
-      *os << " - an element ";
-    }
-    matcher_describers_[i]->DescribeTo(os);
-    if (match_flags() == UnorderedMatcherRequire::ExactMatch) {
-      sep = ", and\n";
-    } else {
-      sep = "\n";
-    }
-  }
-}
-
-// Checks that all matchers match at least one element, and that all
-// elements match at least one matcher. This enables faster matching
-// and better error reporting.
-// Returns false, writing an explanation to 'listener', if and only
-// if the success criteria are not met.
-bool UnorderedElementsAreMatcherImplBase::VerifyMatchMatrix(
-    const ::std::vector<std::string>& element_printouts,
-    const MatchMatrix& matrix, MatchResultListener* listener) const {
-  bool result = true;
-  ::std::vector<char> element_matched(matrix.LhsSize(), 0);
-  ::std::vector<char> matcher_matched(matrix.RhsSize(), 0);
-
-  for (size_t ilhs = 0; ilhs < matrix.LhsSize(); ilhs++) {
-    for (size_t irhs = 0; irhs < matrix.RhsSize(); irhs++) {
-      char matched = matrix.HasEdge(ilhs, irhs);
-      element_matched[ilhs] |= matched;
-      matcher_matched[irhs] |= matched;
-    }
-  }
-
-  if (match_flags() & UnorderedMatcherRequire::Superset) {
-    const char* sep =
-        "where the following matchers don't match any elements:\n";
-    for (size_t mi = 0; mi < matcher_matched.size(); ++mi) {
-      if (matcher_matched[mi]) continue;
-      result = false;
-      if (listener->IsInterested()) {
-        *listener << sep << "matcher #" << mi << ": ";
-        matcher_describers_[mi]->DescribeTo(listener->stream());
-        sep = ",\n";
-      }
-    }
-  }
-
-  if (match_flags() & UnorderedMatcherRequire::Subset) {
-    const char* sep =
-        "where the following elements don't match any matchers:\n";
-    const char* outer_sep = "";
-    if (!result) {
-      outer_sep = "\nand ";
-    }
-    for (size_t ei = 0; ei < element_matched.size(); ++ei) {
-      if (element_matched[ei]) continue;
-      result = false;
-      if (listener->IsInterested()) {
-        *listener << outer_sep << sep << "element #" << ei << ": "
-                  << element_printouts[ei];
-        sep = ",\n";
-        outer_sep = "";
-      }
-    }
-  }
-  return result;
-}
-
-bool UnorderedElementsAreMatcherImplBase::FindPairing(
-    const MatchMatrix& matrix, MatchResultListener* listener) const {
-  ElementMatcherPairs matches = FindMaxBipartiteMatching(matrix);
-
-  size_t max_flow = matches.size();
-  if ((match_flags() & UnorderedMatcherRequire::Superset) &&
-      max_flow < matrix.RhsSize()) {
-    if (listener->IsInterested()) {
-      *listener << "where no permutation of the elements can satisfy all "
-                   "matchers, and the closest match is "
-                << max_flow << " of " << matrix.RhsSize()
-                << " matchers with the pairings:\n";
-      LogElementMatcherPairVec(matches, listener->stream());
-    }
-    return false;
-  }
-  if ((match_flags() & UnorderedMatcherRequire::Subset) &&
-      max_flow < matrix.LhsSize()) {
-    if (listener->IsInterested()) {
-      *listener
-          << "where not all elements can be matched, and the closest match is "
-          << max_flow << " of " << matrix.RhsSize()
-          << " matchers with the pairings:\n";
-      LogElementMatcherPairVec(matches, listener->stream());
-    }
-    return false;
-  }
-
-  if (matches.size() > 1) {
-    if (listener->IsInterested()) {
-      const char* sep = "where:\n";
-      for (size_t mi = 0; mi < matches.size(); ++mi) {
-        *listener << sep << " - element #" << matches[mi].first
-                  << " is matched by matcher #" << matches[mi].second;
-        sep = ",\n";
-      }
-    }
-  }
-  return true;
-}
-
-}  // namespace internal
-}  // namespace testing
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements the spec builder syntax (ON_CALL and
-// EXPECT_CALL).
-
-
-#include <stdlib.h>
-#include <iostream>  // NOLINT
-#include <map>
-#include <set>
-#include <string>
-#include <vector>
-
-#if GTEST_OS_CYGWIN || GTEST_OS_LINUX || GTEST_OS_MAC
-# include <unistd.h>  // NOLINT
-#endif
-
-// Silence C4800 (C4800: 'int *const ': forcing value
-// to bool 'true' or 'false') for MSVC 14,15
-#ifdef _MSC_VER
-#if _MSC_VER <= 1900
-#  pragma warning(push)
-#  pragma warning(disable:4800)
-#endif
-#endif
-
-namespace testing {
-namespace internal {
-
-// Protects the mock object registry (in class Mock), all function
-// mockers, and all expectations.
-GTEST_API_ GTEST_DEFINE_STATIC_MUTEX_(g_gmock_mutex);
-
-// Logs a message including file and line number information.
-GTEST_API_ void LogWithLocation(testing::internal::LogSeverity severity,
-                                const char* file, int line,
-                                const std::string& message) {
-  ::std::ostringstream s;
-  s << file << ":" << line << ": " << message << ::std::endl;
-  Log(severity, s.str(), 0);
-}
-
-// Constructs an ExpectationBase object.
-ExpectationBase::ExpectationBase(const char* a_file, int a_line,
-                                 const std::string& a_source_text)
-    : file_(a_file),
-      line_(a_line),
-      source_text_(a_source_text),
-      cardinality_specified_(false),
-      cardinality_(Exactly(1)),
-      call_count_(0),
-      retired_(false),
-      extra_matcher_specified_(false),
-      repeated_action_specified_(false),
-      retires_on_saturation_(false),
-      last_clause_(kNone),
-      action_count_checked_(false) {}
-
-// Destructs an ExpectationBase object.
-ExpectationBase::~ExpectationBase() {}
-
-// Explicitly specifies the cardinality of this expectation.  Used by
-// the subclasses to implement the .Times() clause.
-void ExpectationBase::SpecifyCardinality(const Cardinality& a_cardinality) {
-  cardinality_specified_ = true;
-  cardinality_ = a_cardinality;
-}
-
-// Retires all pre-requisites of this expectation.
-void ExpectationBase::RetireAllPreRequisites()
-    GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-  if (is_retired()) {
-    // We can take this short-cut as we never retire an expectation
-    // until we have retired all its pre-requisites.
-    return;
-  }
-
-  ::std::vector<ExpectationBase*> expectations(1, this);
-  while (!expectations.empty()) {
-    ExpectationBase* exp = expectations.back();
-    expectations.pop_back();
-
-    for (ExpectationSet::const_iterator it =
-             exp->immediate_prerequisites_.begin();
-         it != exp->immediate_prerequisites_.end(); ++it) {
-      ExpectationBase* next = it->expectation_base().get();
-      if (!next->is_retired()) {
-        next->Retire();
-        expectations.push_back(next);
-      }
-    }
-  }
-}
-
-// Returns true iff all pre-requisites of this expectation have been
-// satisfied.
-bool ExpectationBase::AllPrerequisitesAreSatisfied() const
-    GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-  g_gmock_mutex.AssertHeld();
-  ::std::vector<const ExpectationBase*> expectations(1, this);
-  while (!expectations.empty()) {
-    const ExpectationBase* exp = expectations.back();
-    expectations.pop_back();
-
-    for (ExpectationSet::const_iterator it =
-             exp->immediate_prerequisites_.begin();
-         it != exp->immediate_prerequisites_.end(); ++it) {
-      const ExpectationBase* next = it->expectation_base().get();
-      if (!next->IsSatisfied()) return false;
-      expectations.push_back(next);
-    }
-  }
-  return true;
-}
-
-// Adds unsatisfied pre-requisites of this expectation to 'result'.
-void ExpectationBase::FindUnsatisfiedPrerequisites(ExpectationSet* result) const
-    GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-  g_gmock_mutex.AssertHeld();
-  ::std::vector<const ExpectationBase*> expectations(1, this);
-  while (!expectations.empty()) {
-    const ExpectationBase* exp = expectations.back();
-    expectations.pop_back();
-
-    for (ExpectationSet::const_iterator it =
-             exp->immediate_prerequisites_.begin();
-         it != exp->immediate_prerequisites_.end(); ++it) {
-      const ExpectationBase* next = it->expectation_base().get();
-
-      if (next->IsSatisfied()) {
-        // If *it is satisfied and has a call count of 0, some of its
-        // pre-requisites may not be satisfied yet.
-        if (next->call_count_ == 0) {
-          expectations.push_back(next);
-        }
-      } else {
-        // Now that we know next is unsatisfied, we are not so interested
-        // in whether its pre-requisites are satisfied.  Therefore we
-        // don't iterate into it here.
-        *result += *it;
-      }
-    }
-  }
-}
-
-// Describes how many times a function call matching this
-// expectation has occurred.
-void ExpectationBase::DescribeCallCountTo(::std::ostream* os) const
-    GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-  g_gmock_mutex.AssertHeld();
-
-  // Describes how many times the function is expected to be called.
-  *os << "         Expected: to be ";
-  cardinality().DescribeTo(os);
-  *os << "\n           Actual: ";
-  Cardinality::DescribeActualCallCountTo(call_count(), os);
-
-  // Describes the state of the expectation (e.g. is it satisfied?
-  // is it active?).
-  *os << " - " << (IsOverSaturated() ? "over-saturated" :
-                   IsSaturated() ? "saturated" :
-                   IsSatisfied() ? "satisfied" : "unsatisfied")
-      << " and "
-      << (is_retired() ? "retired" : "active");
-}
-
-// Checks the action count (i.e. the number of WillOnce() and
-// WillRepeatedly() clauses) against the cardinality if this hasn't
-// been done before.  Prints a warning if there are too many or too
-// few actions.
-void ExpectationBase::CheckActionCountIfNotDone() const
-    GTEST_LOCK_EXCLUDED_(mutex_) {
-  bool should_check = false;
-  {
-    MutexLock l(&mutex_);
-    if (!action_count_checked_) {
-      action_count_checked_ = true;
-      should_check = true;
-    }
-  }
-
-  if (should_check) {
-    if (!cardinality_specified_) {
-      // The cardinality was inferred - no need to check the action
-      // count against it.
-      return;
-    }
-
-    // The cardinality was explicitly specified.
-    const int action_count = static_cast<int>(untyped_actions_.size());
-    const int upper_bound = cardinality().ConservativeUpperBound();
-    const int lower_bound = cardinality().ConservativeLowerBound();
-    bool too_many;  // True if there are too many actions, or false
-    // if there are too few.
-    if (action_count > upper_bound ||
-        (action_count == upper_bound && repeated_action_specified_)) {
-      too_many = true;
-    } else if (0 < action_count && action_count < lower_bound &&
-               !repeated_action_specified_) {
-      too_many = false;
-    } else {
-      return;
-    }
-
-    ::std::stringstream ss;
-    DescribeLocationTo(&ss);
-    ss << "Too " << (too_many ? "many" : "few")
-       << " actions specified in " << source_text() << "...\n"
-       << "Expected to be ";
-    cardinality().DescribeTo(&ss);
-    ss << ", but has " << (too_many ? "" : "only ")
-       << action_count << " WillOnce()"
-       << (action_count == 1 ? "" : "s");
-    if (repeated_action_specified_) {
-      ss << " and a WillRepeatedly()";
-    }
-    ss << ".";
-    Log(kWarning, ss.str(), -1);  // -1 means "don't print stack trace".
-  }
-}
-
-// Implements the .Times() clause.
-void ExpectationBase::UntypedTimes(const Cardinality& a_cardinality) {
-  if (last_clause_ == kTimes) {
-    ExpectSpecProperty(false,
-                       ".Times() cannot appear "
-                       "more than once in an EXPECT_CALL().");
-  } else {
-    ExpectSpecProperty(last_clause_ < kTimes,
-                       ".Times() cannot appear after "
-                       ".InSequence(), .WillOnce(), .WillRepeatedly(), "
-                       "or .RetiresOnSaturation().");
-  }
-  last_clause_ = kTimes;
-
-  SpecifyCardinality(a_cardinality);
-}
-
-// Points to the implicit sequence introduced by a living InSequence
-// object (if any) in the current thread or NULL.
-GTEST_API_ ThreadLocal<Sequence*> g_gmock_implicit_sequence;
-
-// Reports an uninteresting call (whose description is in msg) in the
-// manner specified by 'reaction'.
-void ReportUninterestingCall(CallReaction reaction, const std::string& msg) {
-  // Include a stack trace only if --gmock_verbose=info is specified.
-  const int stack_frames_to_skip =
-      GMOCK_FLAG(verbose) == kInfoVerbosity ? 3 : -1;
-  switch (reaction) {
-    case kAllow:
-      Log(kInfo, msg, stack_frames_to_skip);
-      break;
-    case kWarn:
-      Log(kWarning,
-          msg +
-              "\nNOTE: You can safely ignore the above warning unless this "
-              "call should not happen.  Do not suppress it by blindly adding "
-              "an EXPECT_CALL() if you don't mean to enforce the call.  "
-              "See "
-              "https://github.com/google/googletest/blob/master/googlemock/"
-              "docs/CookBook.md#"
-              "knowing-when-to-expect for details.\n",
-          stack_frames_to_skip);
-      break;
-    default:  // FAIL
-      Expect(false, NULL, -1, msg);
-  }
-}
-
-UntypedFunctionMockerBase::UntypedFunctionMockerBase()
-    : mock_obj_(NULL), name_("") {}
-
-UntypedFunctionMockerBase::~UntypedFunctionMockerBase() {}
-
-// Sets the mock object this mock method belongs to, and registers
-// this information in the global mock registry.  Will be called
-// whenever an EXPECT_CALL() or ON_CALL() is executed on this mock
-// method.
-void UntypedFunctionMockerBase::RegisterOwner(const void* mock_obj)
-    GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-  {
-    MutexLock l(&g_gmock_mutex);
-    mock_obj_ = mock_obj;
-  }
-  Mock::Register(mock_obj, this);
-}
-
-// Sets the mock object this mock method belongs to, and sets the name
-// of the mock function.  Will be called upon each invocation of this
-// mock function.
-void UntypedFunctionMockerBase::SetOwnerAndName(const void* mock_obj,
-                                                const char* name)
-    GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-  // We protect name_ under g_gmock_mutex in case this mock function
-  // is called from two threads concurrently.
-  MutexLock l(&g_gmock_mutex);
-  mock_obj_ = mock_obj;
-  name_ = name;
-}
-
-// Returns the name of the function being mocked.  Must be called
-// after RegisterOwner() or SetOwnerAndName() has been called.
-const void* UntypedFunctionMockerBase::MockObject() const
-    GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-  const void* mock_obj;
-  {
-    // We protect mock_obj_ under g_gmock_mutex in case this mock
-    // function is called from two threads concurrently.
-    MutexLock l(&g_gmock_mutex);
-    Assert(mock_obj_ != NULL, __FILE__, __LINE__,
-           "MockObject() must not be called before RegisterOwner() or "
-           "SetOwnerAndName() has been called.");
-    mock_obj = mock_obj_;
-  }
-  return mock_obj;
-}
-
-// Returns the name of this mock method.  Must be called after
-// SetOwnerAndName() has been called.
-const char* UntypedFunctionMockerBase::Name() const
-    GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-  const char* name;
-  {
-    // We protect name_ under g_gmock_mutex in case this mock
-    // function is called from two threads concurrently.
-    MutexLock l(&g_gmock_mutex);
-    Assert(name_ != NULL, __FILE__, __LINE__,
-           "Name() must not be called before SetOwnerAndName() has "
-           "been called.");
-    name = name_;
-  }
-  return name;
-}
-
-// Calculates the result of invoking this mock function with the given
-// arguments, prints it, and returns it.  The caller is responsible
-// for deleting the result.
-UntypedActionResultHolderBase* UntypedFunctionMockerBase::UntypedInvokeWith(
-    void* const untyped_args) GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-  // See the definition of untyped_expectations_ for why access to it
-  // is unprotected here.
-  if (untyped_expectations_.size() == 0) {
-    // No expectation is set on this mock method - we have an
-    // uninteresting call.
-
-    // We must get Google Mock's reaction on uninteresting calls
-    // made on this mock object BEFORE performing the action,
-    // because the action may DELETE the mock object and make the
-    // following expression meaningless.
-    const CallReaction reaction =
-        Mock::GetReactionOnUninterestingCalls(MockObject());
-
-    // True iff we need to print this call's arguments and return
-    // value.  This definition must be kept in sync with
-    // the behavior of ReportUninterestingCall().
-    const bool need_to_report_uninteresting_call =
-        // If the user allows this uninteresting call, we print it
-        // only when they want informational messages.
-        reaction == kAllow ? LogIsVisible(kInfo) :
-                           // If the user wants this to be a warning, we print
-                           // it only when they want to see warnings.
-            reaction == kWarn
-                ? LogIsVisible(kWarning)
-                :
-                // Otherwise, the user wants this to be an error, and we
-                // should always print detailed information in the error.
-                true;
-
-    if (!need_to_report_uninteresting_call) {
-      // Perform the action without printing the call information.
-      return this->UntypedPerformDefaultAction(
-          untyped_args, "Function call: " + std::string(Name()));
-    }
-
-    // Warns about the uninteresting call.
-    ::std::stringstream ss;
-    this->UntypedDescribeUninterestingCall(untyped_args, &ss);
-
-    // Calculates the function result.
-    UntypedActionResultHolderBase* const result =
-        this->UntypedPerformDefaultAction(untyped_args, ss.str());
-
-    // Prints the function result.
-    if (result != NULL)
-      result->PrintAsActionResult(&ss);
-
-    ReportUninterestingCall(reaction, ss.str());
-    return result;
-  }
-
-  bool is_excessive = false;
-  ::std::stringstream ss;
-  ::std::stringstream why;
-  ::std::stringstream loc;
-  const void* untyped_action = NULL;
-
-  // The UntypedFindMatchingExpectation() function acquires and
-  // releases g_gmock_mutex.
-  const ExpectationBase* const untyped_expectation =
-      this->UntypedFindMatchingExpectation(
-          untyped_args, &untyped_action, &is_excessive,
-          &ss, &why);
-  const bool found = untyped_expectation != NULL;
-
-  // True iff we need to print the call's arguments and return value.
-  // This definition must be kept in sync with the uses of Expect()
-  // and Log() in this function.
-  const bool need_to_report_call =
-      !found || is_excessive || LogIsVisible(kInfo);
-  if (!need_to_report_call) {
-    // Perform the action without printing the call information.
-    return
-        untyped_action == NULL ?
-        this->UntypedPerformDefaultAction(untyped_args, "") :
-        this->UntypedPerformAction(untyped_action, untyped_args);
-  }
-
-  ss << "    Function call: " << Name();
-  this->UntypedPrintArgs(untyped_args, &ss);
-
-  // In case the action deletes a piece of the expectation, we
-  // generate the message beforehand.
-  if (found && !is_excessive) {
-    untyped_expectation->DescribeLocationTo(&loc);
-  }
-
-  UntypedActionResultHolderBase* const result =
-      untyped_action == NULL ?
-      this->UntypedPerformDefaultAction(untyped_args, ss.str()) :
-      this->UntypedPerformAction(untyped_action, untyped_args);
-  if (result != NULL)
-    result->PrintAsActionResult(&ss);
-  ss << "\n" << why.str();
-
-  if (!found) {
-    // No expectation matches this call - reports a failure.
-    Expect(false, NULL, -1, ss.str());
-  } else if (is_excessive) {
-    // We had an upper-bound violation and the failure message is in ss.
-    Expect(false, untyped_expectation->file(),
-           untyped_expectation->line(), ss.str());
-  } else {
-    // We had an expected call and the matching expectation is
-    // described in ss.
-    Log(kInfo, loc.str() + ss.str(), 2);
-  }
-
-  return result;
-}
-
-// Returns an Expectation object that references and co-owns exp,
-// which must be an expectation on this mock function.
-Expectation UntypedFunctionMockerBase::GetHandleOf(ExpectationBase* exp) {
-  // See the definition of untyped_expectations_ for why access to it
-  // is unprotected here.
-  for (UntypedExpectations::const_iterator it =
-           untyped_expectations_.begin();
-       it != untyped_expectations_.end(); ++it) {
-    if (it->get() == exp) {
-      return Expectation(*it);
-    }
-  }
-
-  Assert(false, __FILE__, __LINE__, "Cannot find expectation.");
-  return Expectation();
-  // The above statement is just to make the code compile, and will
-  // never be executed.
-}
-
-// Verifies that all expectations on this mock function have been
-// satisfied.  Reports one or more Google Test non-fatal failures
-// and returns false if not.
-bool UntypedFunctionMockerBase::VerifyAndClearExpectationsLocked()
-    GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-  g_gmock_mutex.AssertHeld();
-  bool expectations_met = true;
-  for (UntypedExpectations::const_iterator it =
-           untyped_expectations_.begin();
-       it != untyped_expectations_.end(); ++it) {
-    ExpectationBase* const untyped_expectation = it->get();
-    if (untyped_expectation->IsOverSaturated()) {
-      // There was an upper-bound violation.  Since the error was
-      // already reported when it occurred, there is no need to do
-      // anything here.
-      expectations_met = false;
-    } else if (!untyped_expectation->IsSatisfied()) {
-      expectations_met = false;
-      ::std::stringstream ss;
-      ss  << "Actual function call count doesn't match "
-          << untyped_expectation->source_text() << "...\n";
-      // No need to show the source file location of the expectation
-      // in the description, as the Expect() call that follows already
-      // takes care of it.
-      untyped_expectation->MaybeDescribeExtraMatcherTo(&ss);
-      untyped_expectation->DescribeCallCountTo(&ss);
-      Expect(false, untyped_expectation->file(),
-             untyped_expectation->line(), ss.str());
-    }
-  }
-
-  // Deleting our expectations may trigger other mock objects to be deleted, for
-  // example if an action contains a reference counted smart pointer to that
-  // mock object, and that is the last reference. So if we delete our
-  // expectations within the context of the global mutex we may deadlock when
-  // this method is called again. Instead, make a copy of the set of
-  // expectations to delete, clear our set within the mutex, and then clear the
-  // copied set outside of it.
-  UntypedExpectations expectations_to_delete;
-  untyped_expectations_.swap(expectations_to_delete);
-
-  g_gmock_mutex.Unlock();
-  expectations_to_delete.clear();
-  g_gmock_mutex.Lock();
-
-  return expectations_met;
-}
-
-CallReaction intToCallReaction(int mock_behavior) {
-  if (mock_behavior >= kAllow && mock_behavior <= kFail) {
-    return static_cast<internal::CallReaction>(mock_behavior);
-  }
-  return kWarn;
-}
-
-}  // namespace internal
-
-// Class Mock.
-
-namespace {
-
-typedef std::set<internal::UntypedFunctionMockerBase*> FunctionMockers;
-
-// The current state of a mock object.  Such information is needed for
-// detecting leaked mock objects and explicitly verifying a mock's
-// expectations.
-struct MockObjectState {
-  MockObjectState()
-      : first_used_file(NULL), first_used_line(-1), leakable(false) {}
-
-  // Where in the source file an ON_CALL or EXPECT_CALL is first
-  // invoked on this mock object.
-  const char* first_used_file;
-  int first_used_line;
-  ::std::string first_used_test_case;
-  ::std::string first_used_test;
-  bool leakable;  // true iff it's OK to leak the object.
-  FunctionMockers function_mockers;  // All registered methods of the object.
-};
-
-// A global registry holding the state of all mock objects that are
-// alive.  A mock object is added to this registry the first time
-// Mock::AllowLeak(), ON_CALL(), or EXPECT_CALL() is called on it.  It
-// is removed from the registry in the mock object's destructor.
-class MockObjectRegistry {
- public:
-  // Maps a mock object (identified by its address) to its state.
-  typedef std::map<const void*, MockObjectState> StateMap;
-
-  // This destructor will be called when a program exits, after all
-  // tests in it have been run.  By then, there should be no mock
-  // object alive.  Therefore we report any living object as test
-  // failure, unless the user explicitly asked us to ignore it.
-  ~MockObjectRegistry() {
-    // "using ::std::cout;" doesn't work with Symbian's STLport, where cout is
-    // a macro.
-
-    if (!GMOCK_FLAG(catch_leaked_mocks))
-      return;
-
-    int leaked_count = 0;
-    for (StateMap::const_iterator it = states_.begin(); it != states_.end();
-         ++it) {
-      if (it->second.leakable)  // The user said it's fine to leak this object.
-        continue;
-
-      // TODO(wan@google.com): Print the type of the leaked object.
-      // This can help the user identify the leaked object.
-      std::cout << "\n";
-      const MockObjectState& state = it->second;
-      std::cout << internal::FormatFileLocation(state.first_used_file,
-                                                state.first_used_line);
-      std::cout << " ERROR: this mock object";
-      if (state.first_used_test != "") {
-        std::cout << " (used in test " << state.first_used_test_case << "."
-             << state.first_used_test << ")";
-      }
-      std::cout << " should be deleted but never is. Its address is @"
-           << it->first << ".";
-      leaked_count++;
-    }
-    if (leaked_count > 0) {
-      std::cout << "\nERROR: " << leaked_count << " leaked mock "
-                << (leaked_count == 1 ? "object" : "objects")
-                << " found at program exit. Expectations on a mock object is "
-                   "verified when the object is destructed. Leaking a mock "
-                   "means that its expectations aren't verified, which is "
-                   "usually a test bug. If you really intend to leak a mock, "
-                   "you can suppress this error using "
-                   "testing::Mock::AllowLeak(mock_object), or you may use a "
-                   "fake or stub instead of a mock.\n";
-      std::cout.flush();
-      ::std::cerr.flush();
-      // RUN_ALL_TESTS() has already returned when this destructor is
-      // called.  Therefore we cannot use the normal Google Test
-      // failure reporting mechanism.
-      _exit(1);  // We cannot call exit() as it is not reentrant and
-                 // may already have been called.
-    }
-  }
-
-  StateMap& states() { return states_; }
-
- private:
-  StateMap states_;
-};
-
-// Protected by g_gmock_mutex.
-MockObjectRegistry g_mock_object_registry;
-
-// Maps a mock object to the reaction Google Mock should have when an
-// uninteresting method is called.  Protected by g_gmock_mutex.
-std::map<const void*, internal::CallReaction> g_uninteresting_call_reaction;
-
-// Sets the reaction Google Mock should have when an uninteresting
-// method of the given mock object is called.
-void SetReactionOnUninterestingCalls(const void* mock_obj,
-                                     internal::CallReaction reaction)
-    GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  internal::MutexLock l(&internal::g_gmock_mutex);
-  g_uninteresting_call_reaction[mock_obj] = reaction;
-}
-
-}  // namespace
-
-// Tells Google Mock to allow uninteresting calls on the given mock
-// object.
-void Mock::AllowUninterestingCalls(const void* mock_obj)
-    GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  SetReactionOnUninterestingCalls(mock_obj, internal::kAllow);
-}
-
-// Tells Google Mock to warn the user about uninteresting calls on the
-// given mock object.
-void Mock::WarnUninterestingCalls(const void* mock_obj)
-    GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  SetReactionOnUninterestingCalls(mock_obj, internal::kWarn);
-}
-
-// Tells Google Mock to fail uninteresting calls on the given mock
-// object.
-void Mock::FailUninterestingCalls(const void* mock_obj)
-    GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  SetReactionOnUninterestingCalls(mock_obj, internal::kFail);
-}
-
-// Tells Google Mock the given mock object is being destroyed and its
-// entry in the call-reaction table should be removed.
-void Mock::UnregisterCallReaction(const void* mock_obj)
-    GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  internal::MutexLock l(&internal::g_gmock_mutex);
-  g_uninteresting_call_reaction.erase(mock_obj);
-}
-
-// Returns the reaction Google Mock will have on uninteresting calls
-// made on the given mock object.
-internal::CallReaction Mock::GetReactionOnUninterestingCalls(
-    const void* mock_obj)
-        GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  internal::MutexLock l(&internal::g_gmock_mutex);
-  return (g_uninteresting_call_reaction.count(mock_obj) == 0) ?
-      internal::intToCallReaction(GMOCK_FLAG(default_mock_behavior)) :
-      g_uninteresting_call_reaction[mock_obj];
-}
-
-// Tells Google Mock to ignore mock_obj when checking for leaked mock
-// objects.
-void Mock::AllowLeak(const void* mock_obj)
-    GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  internal::MutexLock l(&internal::g_gmock_mutex);
-  g_mock_object_registry.states()[mock_obj].leakable = true;
-}
-
-// Verifies and clears all expectations on the given mock object.  If
-// the expectations aren't satisfied, generates one or more Google
-// Test non-fatal failures and returns false.
-bool Mock::VerifyAndClearExpectations(void* mock_obj)
-    GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  internal::MutexLock l(&internal::g_gmock_mutex);
-  return VerifyAndClearExpectationsLocked(mock_obj);
-}
-
-// Verifies all expectations on the given mock object and clears its
-// default actions and expectations.  Returns true iff the
-// verification was successful.
-bool Mock::VerifyAndClear(void* mock_obj)
-    GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  internal::MutexLock l(&internal::g_gmock_mutex);
-  ClearDefaultActionsLocked(mock_obj);
-  return VerifyAndClearExpectationsLocked(mock_obj);
-}
-
-// Verifies and clears all expectations on the given mock object.  If
-// the expectations aren't satisfied, generates one or more Google
-// Test non-fatal failures and returns false.
-bool Mock::VerifyAndClearExpectationsLocked(void* mock_obj)
-    GTEST_EXCLUSIVE_LOCK_REQUIRED_(internal::g_gmock_mutex) {
-  internal::g_gmock_mutex.AssertHeld();
-  if (g_mock_object_registry.states().count(mock_obj) == 0) {
-    // No EXPECT_CALL() was set on the given mock object.
-    return true;
-  }
-
-  // Verifies and clears the expectations on each mock method in the
-  // given mock object.
-  bool expectations_met = true;
-  FunctionMockers& mockers =
-      g_mock_object_registry.states()[mock_obj].function_mockers;
-  for (FunctionMockers::const_iterator it = mockers.begin();
-       it != mockers.end(); ++it) {
-    if (!(*it)->VerifyAndClearExpectationsLocked()) {
-      expectations_met = false;
-    }
-  }
-
-  // We don't clear the content of mockers, as they may still be
-  // needed by ClearDefaultActionsLocked().
-  return expectations_met;
-}
-
-// Registers a mock object and a mock method it owns.
-void Mock::Register(const void* mock_obj,
-                    internal::UntypedFunctionMockerBase* mocker)
-    GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  internal::MutexLock l(&internal::g_gmock_mutex);
-  g_mock_object_registry.states()[mock_obj].function_mockers.insert(mocker);
-}
-
-// Tells Google Mock where in the source code mock_obj is used in an
-// ON_CALL or EXPECT_CALL.  In case mock_obj is leaked, this
-// information helps the user identify which object it is.
-void Mock::RegisterUseByOnCallOrExpectCall(const void* mock_obj,
-                                           const char* file, int line)
-    GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex) {
-  internal::MutexLock l(&internal::g_gmock_mutex);
-  MockObjectState& state = g_mock_object_registry.states()[mock_obj];
-  if (state.first_used_file == NULL) {
-    state.first_used_file = file;
-    state.first_used_line = line;
-    const TestInfo* const test_info =
-        UnitTest::GetInstance()->current_test_info();
-    if (test_info != NULL) {
-      // TODO(wan@google.com): record the test case name when the
-      // ON_CALL or EXPECT_CALL is invoked from SetUpTestCase() or
-      // TearDownTestCase().
-      state.first_used_test_case = test_info->test_case_name();
-      state.first_used_test = test_info->name();
-    }
-  }
-}
-
-// Unregisters a mock method; removes the owning mock object from the
-// registry when the last mock method associated with it has been
-// unregistered.  This is called only in the destructor of
-// FunctionMockerBase.
-void Mock::UnregisterLocked(internal::UntypedFunctionMockerBase* mocker)
-    GTEST_EXCLUSIVE_LOCK_REQUIRED_(internal::g_gmock_mutex) {
-  internal::g_gmock_mutex.AssertHeld();
-  for (MockObjectRegistry::StateMap::iterator it =
-           g_mock_object_registry.states().begin();
-       it != g_mock_object_registry.states().end(); ++it) {
-    FunctionMockers& mockers = it->second.function_mockers;
-    if (mockers.erase(mocker) > 0) {
-      // mocker was in mockers and has been just removed.
-      if (mockers.empty()) {
-        g_mock_object_registry.states().erase(it);
-      }
-      return;
-    }
-  }
-}
-
-// Clears all ON_CALL()s set on the given mock object.
-void Mock::ClearDefaultActionsLocked(void* mock_obj)
-    GTEST_EXCLUSIVE_LOCK_REQUIRED_(internal::g_gmock_mutex) {
-  internal::g_gmock_mutex.AssertHeld();
-
-  if (g_mock_object_registry.states().count(mock_obj) == 0) {
-    // No ON_CALL() was set on the given mock object.
-    return;
-  }
-
-  // Clears the default actions for each mock method in the given mock
-  // object.
-  FunctionMockers& mockers =
-      g_mock_object_registry.states()[mock_obj].function_mockers;
-  for (FunctionMockers::const_iterator it = mockers.begin();
-       it != mockers.end(); ++it) {
-    (*it)->ClearDefaultActionsLocked();
-  }
-
-  // We don't clear the content of mockers, as they may still be
-  // needed by VerifyAndClearExpectationsLocked().
-}
-
-Expectation::Expectation() {}
-
-Expectation::Expectation(
-    const internal::linked_ptr<internal::ExpectationBase>& an_expectation_base)
-    : expectation_base_(an_expectation_base) {}
-
-Expectation::~Expectation() {}
-
-// Adds an expectation to a sequence.
-void Sequence::AddExpectation(const Expectation& expectation) const {
-  if (*last_expectation_ != expectation) {
-    if (last_expectation_->expectation_base() != NULL) {
-      expectation.expectation_base()->immediate_prerequisites_
-          += *last_expectation_;
-    }
-    *last_expectation_ = expectation;
-  }
-}
-
-// Creates the implicit sequence if there isn't one.
-InSequence::InSequence() {
-  if (internal::g_gmock_implicit_sequence.get() == NULL) {
-    internal::g_gmock_implicit_sequence.set(new Sequence);
-    sequence_created_ = true;
-  } else {
-    sequence_created_ = false;
-  }
-}
-
-// Deletes the implicit sequence if it was created by the constructor
-// of this object.
-InSequence::~InSequence() {
-  if (sequence_created_) {
-    delete internal::g_gmock_implicit_sequence.get();
-    internal::g_gmock_implicit_sequence.set(NULL);
-  }
-}
-
-}  // namespace testing
-
-#ifdef _MSC_VER
-#if _MSC_VER <= 1900
-#  pragma warning(pop)
-#endif
-#endif
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-
-namespace testing {
-
-// TODO(wan@google.com): support using environment variables to
-// control the flag values, like what Google Test does.
-
-GMOCK_DEFINE_bool_(catch_leaked_mocks, true,
-                   "true iff Google Mock should report leaked mock objects "
-                   "as failures.");
-
-GMOCK_DEFINE_string_(verbose, internal::kWarningVerbosity,
-                     "Controls how verbose Google Mock's output is."
-                     "  Valid values:\n"
-                     "  info    - prints all messages.\n"
-                     "  warning - prints warnings and errors.\n"
-                     "  error   - prints errors only.");
-
-GMOCK_DEFINE_int32_(default_mock_behavior, 1,
-                    "Controls the default behavior of mocks."
-                    "  Valid values:\n"
-                    "  0 - by default, mocks act as NiceMocks.\n"
-                    "  1 - by default, mocks act as NaggyMocks.\n"
-                    "  2 - by default, mocks act as StrictMocks.");
-
-namespace internal {
-
-// Parses a string as a command line flag.  The string should have the
-// format "--gmock_flag=value".  When def_optional is true, the
-// "=value" part can be omitted.
-//
-// Returns the value of the flag, or NULL if the parsing failed.
-static const char* ParseGoogleMockFlagValue(const char* str,
-                                            const char* flag,
-                                            bool def_optional) {
-  // str and flag must not be NULL.
-  if (str == NULL || flag == NULL) return NULL;
-
-  // The flag must start with "--gmock_".
-  const std::string flag_str = std::string("--gmock_") + flag;
-  const size_t flag_len = flag_str.length();
-  if (strncmp(str, flag_str.c_str(), flag_len) != 0) return NULL;
-
-  // Skips the flag name.
-  const char* flag_end = str + flag_len;
-
-  // When def_optional is true, it's OK to not have a "=value" part.
-  if (def_optional && (flag_end[0] == '\0')) {
-    return flag_end;
-  }
-
-  // If def_optional is true and there are more characters after the
-  // flag name, or if def_optional is false, there must be a '=' after
-  // the flag name.
-  if (flag_end[0] != '=') return NULL;
-
-  // Returns the string after "=".
-  return flag_end + 1;
-}
-
-// Parses a string for a Google Mock bool flag, in the form of
-// "--gmock_flag=value".
-//
-// On success, stores the value of the flag in *value, and returns
-// true.  On failure, returns false without changing *value.
-static bool ParseGoogleMockBoolFlag(const char* str, const char* flag,
-                                    bool* value) {
-  // Gets the value of the flag as a string.
-  const char* const value_str = ParseGoogleMockFlagValue(str, flag, true);
-
-  // Aborts if the parsing failed.
-  if (value_str == NULL) return false;
-
-  // Converts the string value to a bool.
-  *value = !(*value_str == '0' || *value_str == 'f' || *value_str == 'F');
-  return true;
-}
-
-// Parses a string for a Google Mock string flag, in the form of
-// "--gmock_flag=value".
-//
-// On success, stores the value of the flag in *value, and returns
-// true.  On failure, returns false without changing *value.
-template <typename String>
-static bool ParseGoogleMockStringFlag(const char* str, const char* flag,
-                                      String* value) {
-  // Gets the value of the flag as a string.
-  const char* const value_str = ParseGoogleMockFlagValue(str, flag, false);
-
-  // Aborts if the parsing failed.
-  if (value_str == NULL) return false;
-
-  // Sets *value to the value of the flag.
-  *value = value_str;
-  return true;
-}
-
-static bool ParseGoogleMockIntFlag(const char* str, const char* flag,
-                                   int* value) {
-  // Gets the value of the flag as a string.
-  const char* const value_str = ParseGoogleMockFlagValue(str, flag, true);
-
-  // Aborts if the parsing failed.
-  if (value_str == NULL) return false;
-
-  // Sets *value to the value of the flag.
-  return ParseInt32(Message() << "The value of flag --" << flag,
-                    value_str, value);
-}
-
-// The internal implementation of InitGoogleMock().
-//
-// The type parameter CharType can be instantiated to either char or
-// wchar_t.
-template <typename CharType>
-void InitGoogleMockImpl(int* argc, CharType** argv) {
-  // Makes sure Google Test is initialized.  InitGoogleTest() is
-  // idempotent, so it's fine if the user has already called it.
-  InitGoogleTest(argc, argv);
-  if (*argc <= 0) return;
-
-  for (int i = 1; i != *argc; i++) {
-    const std::string arg_string = StreamableToString(argv[i]);
-    const char* const arg = arg_string.c_str();
-
-    // Do we see a Google Mock flag?
-    if (ParseGoogleMockBoolFlag(arg, "catch_leaked_mocks",
-                                &GMOCK_FLAG(catch_leaked_mocks)) ||
-        ParseGoogleMockStringFlag(arg, "verbose", &GMOCK_FLAG(verbose)) ||
-        ParseGoogleMockIntFlag(arg, "default_mock_behavior",
-                               &GMOCK_FLAG(default_mock_behavior))) {
-      // Yes.  Shift the remainder of the argv list left by one.  Note
-      // that argv has (*argc + 1) elements, the last one always being
-      // NULL.  The following loop moves the trailing NULL element as
-      // well.
-      for (int j = i; j != *argc; j++) {
-        argv[j] = argv[j + 1];
-      }
-
-      // Decrements the argument count.
-      (*argc)--;
-
-      // We also need to decrement the iterator as we just removed
-      // an element.
-      i--;
-    }
-  }
-}
-
-}  // namespace internal
-
-// Initializes Google Mock.  This must be called before running the
-// tests.  In particular, it parses a command line for the flags that
-// Google Mock recognizes.  Whenever a Google Mock flag is seen, it is
-// removed from argv, and *argc is decremented.
-//
-// No value is returned.  Instead, the Google Mock flag variables are
-// updated.
-//
-// Since Google Test is needed for Google Mock to work, this function
-// also initializes Google Test and parses its flags, if that hasn't
-// been done.
-GTEST_API_ void InitGoogleMock(int* argc, char** argv) {
-  internal::InitGoogleMockImpl(argc, argv);
-}
-
-// This overloaded version can be used in Windows programs compiled in
-// UNICODE mode.
-GTEST_API_ void InitGoogleMock(int* argc, wchar_t** argv) {
-  internal::InitGoogleMockImpl(argc, argv);
-}
-
-}  // namespace testing
diff --git a/inference-engine/thirdparty/clDNN/common/googletest-fused/gmock/gmock.h b/inference-engine/thirdparty/clDNN/common/googletest-fused/gmock/gmock.h
deleted file mode 100644
index b10aae379bfdba..00000000000000
--- a/inference-engine/thirdparty/clDNN/common/googletest-fused/gmock/gmock.h
+++ /dev/null
@@ -1,16697 +0,0 @@
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This is the main header file a user should include.
-
-#ifndef GMOCK_INCLUDE_GMOCK_GMOCK_H_
-#define GMOCK_INCLUDE_GMOCK_GMOCK_H_
-
-// This file implements the following syntax:
-//
-//   ON_CALL(mock_object.Method(...))
-//     .With(...) ?
-//     .WillByDefault(...);
-//
-// where With() is optional and WillByDefault() must appear exactly
-// once.
-//
-//   EXPECT_CALL(mock_object.Method(...))
-//     .With(...) ?
-//     .Times(...) ?
-//     .InSequence(...) *
-//     .WillOnce(...) *
-//     .WillRepeatedly(...) ?
-//     .RetiresOnSaturation() ? ;
-//
-// where all clauses are optional and WillOnce() can be repeated.
-
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements some commonly used actions.
-
-#ifndef GMOCK_INCLUDE_GMOCK_GMOCK_ACTIONS_H_
-#define GMOCK_INCLUDE_GMOCK_GMOCK_ACTIONS_H_
-
-#ifndef _WIN32_WCE
-# include <errno.h>
-#endif
-
-#include <algorithm>
-#include <string>
-
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file defines some utilities useful for implementing Google
-// Mock.  They are subject to change without notice, so please DO NOT
-// USE THEM IN USER CODE.
-
-#ifndef GMOCK_INCLUDE_GMOCK_INTERNAL_GMOCK_INTERNAL_UTILS_H_
-#define GMOCK_INCLUDE_GMOCK_INTERNAL_GMOCK_INTERNAL_UTILS_H_
-
-#include <stdio.h>
-#include <ostream>  // NOLINT
-#include <string>
-// This file was GENERATED by command:
-//     pump.py gmock-generated-internal-utils.h.pump
-// DO NOT EDIT BY HAND!!!
-
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file contains template meta-programming utility classes needed
-// for implementing Google Mock.
-
-#ifndef GMOCK_INCLUDE_GMOCK_INTERNAL_GMOCK_GENERATED_INTERNAL_UTILS_H_
-#define GMOCK_INCLUDE_GMOCK_INTERNAL_GMOCK_GENERATED_INTERNAL_UTILS_H_
-
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: vadimb@google.com (Vadim Berman)
-//
-// Low-level types and utilities for porting Google Mock to various
-// platforms.  All macros ending with _ and symbols defined in an
-// internal namespace are subject to change without notice.  Code
-// outside Google Mock MUST NOT USE THEM DIRECTLY.  Macros that don't
-// end with _ are part of Google Mock's public API and can be used by
-// code outside Google Mock.
-
-#ifndef GMOCK_INCLUDE_GMOCK_INTERNAL_GMOCK_PORT_H_
-#define GMOCK_INCLUDE_GMOCK_INTERNAL_GMOCK_PORT_H_
-
-#include <assert.h>
-#include <stdlib.h>
-#include <iostream>
-
-// Most of the utilities needed for porting Google Mock are also
-// required for Google Test and are defined in gtest-port.h.
-//
-// Note to maintainers: to reduce code duplication, prefer adding
-// portability utilities to Google Test's gtest-port.h instead of
-// here, as Google Mock depends on Google Test.  Only add a utility
-// here if it's truly specific to Google Mock.
-
-#include "gtest/gtest.h"
-// Copyright 2015, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Injection point for custom user configurations.
-// The following macros can be defined:
-//
-//   Flag related macros:
-//     GMOCK_DECLARE_bool_(name)
-//     GMOCK_DECLARE_int32_(name)
-//     GMOCK_DECLARE_string_(name)
-//     GMOCK_DEFINE_bool_(name, default_val, doc)
-//     GMOCK_DEFINE_int32_(name, default_val, doc)
-//     GMOCK_DEFINE_string_(name, default_val, doc)
-//
-// ** Custom implementation starts here **
-
-#ifndef GMOCK_INCLUDE_GMOCK_INTERNAL_CUSTOM_GMOCK_PORT_H_
-#define GMOCK_INCLUDE_GMOCK_INTERNAL_CUSTOM_GMOCK_PORT_H_
-
-#endif  // GMOCK_INCLUDE_GMOCK_INTERNAL_CUSTOM_GMOCK_PORT_H_
-
-// For MS Visual C++, check the compiler version. At least VS 2003 is
-// required to compile Google Mock.
-#if defined(_MSC_VER) && _MSC_VER < 1310
-# error "At least Visual C++ 2003 (7.1) is required to compile Google Mock."
-#endif
-
-// Macro for referencing flags.  This is public as we want the user to
-// use this syntax to reference Google Mock flags.
-#define GMOCK_FLAG(name) FLAGS_gmock_##name
-
-#if !defined(GMOCK_DECLARE_bool_)
-
-// Macros for declaring flags.
-# define GMOCK_DECLARE_bool_(name) extern GTEST_API_ bool GMOCK_FLAG(name)
-# define GMOCK_DECLARE_int32_(name) \
-    extern GTEST_API_ ::testing::internal::Int32 GMOCK_FLAG(name)
-# define GMOCK_DECLARE_string_(name) \
-    extern GTEST_API_ ::std::string GMOCK_FLAG(name)
-
-// Macros for defining flags.
-# define GMOCK_DEFINE_bool_(name, default_val, doc) \
-    GTEST_API_ bool GMOCK_FLAG(name) = (default_val)
-# define GMOCK_DEFINE_int32_(name, default_val, doc) \
-    GTEST_API_ ::testing::internal::Int32 GMOCK_FLAG(name) = (default_val)
-# define GMOCK_DEFINE_string_(name, default_val, doc) \
-    GTEST_API_ ::std::string GMOCK_FLAG(name) = (default_val)
-
-#endif  // !defined(GMOCK_DECLARE_bool_)
-
-#endif  // GMOCK_INCLUDE_GMOCK_INTERNAL_GMOCK_PORT_H_
-
-namespace testing {
-
-template <typename T>
-class Matcher;
-
-namespace internal {
-
-// An IgnoredValue object can be implicitly constructed from ANY value.
-// This is used in implementing the IgnoreResult(a) action.
-class IgnoredValue {
- public:
-  // This constructor template allows any value to be implicitly
-  // converted to IgnoredValue.  The object has no data member and
-  // doesn't try to remember anything about the argument.  We
-  // deliberately omit the 'explicit' keyword in order to allow the
-  // conversion to be implicit.
-  template <typename T>
-  IgnoredValue(const T& /* ignored */) {}  // NOLINT(runtime/explicit)
-};
-
-// MatcherTuple<T>::type is a tuple type where each field is a Matcher
-// for the corresponding field in tuple type T.
-template <typename Tuple>
-struct MatcherTuple;
-
-template <>
-struct MatcherTuple< ::testing::tuple<> > {
-  typedef ::testing::tuple< > type;
-};
-
-template <typename A1>
-struct MatcherTuple< ::testing::tuple<A1> > {
-  typedef ::testing::tuple<Matcher<A1> > type;
-};
-
-template <typename A1, typename A2>
-struct MatcherTuple< ::testing::tuple<A1, A2> > {
-  typedef ::testing::tuple<Matcher<A1>, Matcher<A2> > type;
-};
-
-template <typename A1, typename A2, typename A3>
-struct MatcherTuple< ::testing::tuple<A1, A2, A3> > {
-  typedef ::testing::tuple<Matcher<A1>, Matcher<A2>, Matcher<A3> > type;
-};
-
-template <typename A1, typename A2, typename A3, typename A4>
-struct MatcherTuple< ::testing::tuple<A1, A2, A3, A4> > {
-  typedef ::testing::tuple<Matcher<A1>, Matcher<A2>, Matcher<A3>, Matcher<A4> >
-      type;
-};
-
-template <typename A1, typename A2, typename A3, typename A4, typename A5>
-struct MatcherTuple< ::testing::tuple<A1, A2, A3, A4, A5> > {
-  typedef ::testing::tuple<Matcher<A1>, Matcher<A2>, Matcher<A3>, Matcher<A4>,
-                           Matcher<A5> >
-      type;
-};
-
-template <typename A1, typename A2, typename A3, typename A4, typename A5,
-    typename A6>
-struct MatcherTuple< ::testing::tuple<A1, A2, A3, A4, A5, A6> > {
-  typedef ::testing::tuple<Matcher<A1>, Matcher<A2>, Matcher<A3>, Matcher<A4>,
-                           Matcher<A5>, Matcher<A6> >
-      type;
-};
-
-template <typename A1, typename A2, typename A3, typename A4, typename A5,
-    typename A6, typename A7>
-struct MatcherTuple< ::testing::tuple<A1, A2, A3, A4, A5, A6, A7> > {
-  typedef ::testing::tuple<Matcher<A1>, Matcher<A2>, Matcher<A3>, Matcher<A4>,
-                           Matcher<A5>, Matcher<A6>, Matcher<A7> >
-      type;
-};
-
-template <typename A1, typename A2, typename A3, typename A4, typename A5,
-    typename A6, typename A7, typename A8>
-struct MatcherTuple< ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8> > {
-  typedef ::testing::tuple<Matcher<A1>, Matcher<A2>, Matcher<A3>, Matcher<A4>,
-                           Matcher<A5>, Matcher<A6>, Matcher<A7>, Matcher<A8> >
-      type;
-};
-
-template <typename A1, typename A2, typename A3, typename A4, typename A5,
-    typename A6, typename A7, typename A8, typename A9>
-struct MatcherTuple< ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8, A9> > {
-  typedef ::testing::tuple<Matcher<A1>, Matcher<A2>, Matcher<A3>, Matcher<A4>,
-                           Matcher<A5>, Matcher<A6>, Matcher<A7>, Matcher<A8>,
-                           Matcher<A9> >
-      type;
-};
-
-template <typename A1, typename A2, typename A3, typename A4, typename A5,
-    typename A6, typename A7, typename A8, typename A9, typename A10>
-struct MatcherTuple< ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8, A9,
-    A10> > {
-  typedef ::testing::tuple<Matcher<A1>, Matcher<A2>, Matcher<A3>, Matcher<A4>,
-                           Matcher<A5>, Matcher<A6>, Matcher<A7>, Matcher<A8>,
-                           Matcher<A9>, Matcher<A10> >
-      type;
-};
-
-// Template struct Function<F>, where F must be a function type, contains
-// the following typedefs:
-//
-//   Result:               the function's return type.
-//   ArgumentN:            the type of the N-th argument, where N starts with 1.
-//   ArgumentTuple:        the tuple type consisting of all parameters of F.
-//   ArgumentMatcherTuple: the tuple type consisting of Matchers for all
-//                         parameters of F.
-//   MakeResultVoid:       the function type obtained by substituting void
-//                         for the return type of F.
-//   MakeResultIgnoredValue:
-//                         the function type obtained by substituting Something
-//                         for the return type of F.
-template <typename F>
-struct Function;
-
-template <typename R>
-struct Function<R()> {
-  typedef R Result;
-  typedef ::testing::tuple<> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid();
-  typedef IgnoredValue MakeResultIgnoredValue();
-};
-
-template <typename R, typename A1>
-struct Function<R(A1)>
-    : Function<R()> {
-  typedef A1 Argument1;
-  typedef ::testing::tuple<A1> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid(A1);
-  typedef IgnoredValue MakeResultIgnoredValue(A1);
-};
-
-template <typename R, typename A1, typename A2>
-struct Function<R(A1, A2)>
-    : Function<R(A1)> {
-  typedef A2 Argument2;
-  typedef ::testing::tuple<A1, A2> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid(A1, A2);
-  typedef IgnoredValue MakeResultIgnoredValue(A1, A2);
-};
-
-template <typename R, typename A1, typename A2, typename A3>
-struct Function<R(A1, A2, A3)>
-    : Function<R(A1, A2)> {
-  typedef A3 Argument3;
-  typedef ::testing::tuple<A1, A2, A3> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid(A1, A2, A3);
-  typedef IgnoredValue MakeResultIgnoredValue(A1, A2, A3);
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4>
-struct Function<R(A1, A2, A3, A4)>
-    : Function<R(A1, A2, A3)> {
-  typedef A4 Argument4;
-  typedef ::testing::tuple<A1, A2, A3, A4> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid(A1, A2, A3, A4);
-  typedef IgnoredValue MakeResultIgnoredValue(A1, A2, A3, A4);
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5>
-struct Function<R(A1, A2, A3, A4, A5)>
-    : Function<R(A1, A2, A3, A4)> {
-  typedef A5 Argument5;
-  typedef ::testing::tuple<A1, A2, A3, A4, A5> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid(A1, A2, A3, A4, A5);
-  typedef IgnoredValue MakeResultIgnoredValue(A1, A2, A3, A4, A5);
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6>
-struct Function<R(A1, A2, A3, A4, A5, A6)>
-    : Function<R(A1, A2, A3, A4, A5)> {
-  typedef A6 Argument6;
-  typedef ::testing::tuple<A1, A2, A3, A4, A5, A6> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid(A1, A2, A3, A4, A5, A6);
-  typedef IgnoredValue MakeResultIgnoredValue(A1, A2, A3, A4, A5, A6);
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7>
-struct Function<R(A1, A2, A3, A4, A5, A6, A7)>
-    : Function<R(A1, A2, A3, A4, A5, A6)> {
-  typedef A7 Argument7;
-  typedef ::testing::tuple<A1, A2, A3, A4, A5, A6, A7> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid(A1, A2, A3, A4, A5, A6, A7);
-  typedef IgnoredValue MakeResultIgnoredValue(A1, A2, A3, A4, A5, A6, A7);
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7, typename A8>
-struct Function<R(A1, A2, A3, A4, A5, A6, A7, A8)>
-    : Function<R(A1, A2, A3, A4, A5, A6, A7)> {
-  typedef A8 Argument8;
-  typedef ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid(A1, A2, A3, A4, A5, A6, A7, A8);
-  typedef IgnoredValue MakeResultIgnoredValue(A1, A2, A3, A4, A5, A6, A7, A8);
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7, typename A8, typename A9>
-struct Function<R(A1, A2, A3, A4, A5, A6, A7, A8, A9)>
-    : Function<R(A1, A2, A3, A4, A5, A6, A7, A8)> {
-  typedef A9 Argument9;
-  typedef ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8, A9> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid(A1, A2, A3, A4, A5, A6, A7, A8, A9);
-  typedef IgnoredValue MakeResultIgnoredValue(A1, A2, A3, A4, A5, A6, A7, A8,
-      A9);
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7, typename A8, typename A9,
-    typename A10>
-struct Function<R(A1, A2, A3, A4, A5, A6, A7, A8, A9, A10)>
-    : Function<R(A1, A2, A3, A4, A5, A6, A7, A8, A9)> {
-  typedef A10 Argument10;
-  typedef ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8, A9,
-      A10> ArgumentTuple;
-  typedef typename MatcherTuple<ArgumentTuple>::type ArgumentMatcherTuple;
-  typedef void MakeResultVoid(A1, A2, A3, A4, A5, A6, A7, A8, A9, A10);
-  typedef IgnoredValue MakeResultIgnoredValue(A1, A2, A3, A4, A5, A6, A7, A8,
-      A9, A10);
-};
-
-}  // namespace internal
-
-}  // namespace testing
-
-#endif  // GMOCK_INCLUDE_GMOCK_INTERNAL_GMOCK_GENERATED_INTERNAL_UTILS_H_
-
-namespace testing {
-namespace internal {
-
-// Silence MSVC C4100 (unreferenced formal parameter) and
-// C4805('==': unsafe mix of type 'const int' and type 'const bool')
-#ifdef _MSC_VER
-# pragma warning(push)
-# pragma warning(disable:4100)
-# pragma warning(disable:4805)
-#endif
-
-// Joins a vector of strings as if they are fields of a tuple; returns
-// the joined string.
-GTEST_API_ std::string JoinAsTuple(const Strings& fields);
-
-// Converts an identifier name to a space-separated list of lower-case
-// words.  Each maximum substring of the form [A-Za-z][a-z]*|\d+ is
-// treated as one word.  For example, both "FooBar123" and
-// "foo_bar_123" are converted to "foo bar 123".
-GTEST_API_ std::string ConvertIdentifierNameToWords(const char* id_name);
-
-// PointeeOf<Pointer>::type is the type of a value pointed to by a
-// Pointer, which can be either a smart pointer or a raw pointer.  The
-// following default implementation is for the case where Pointer is a
-// smart pointer.
-template <typename Pointer>
-struct PointeeOf {
-  // Smart pointer classes define type element_type as the type of
-  // their pointees.
-  typedef typename Pointer::element_type type;
-};
-// This specialization is for the raw pointer case.
-template <typename T>
-struct PointeeOf<T*> { typedef T type; };  // NOLINT
-
-// GetRawPointer(p) returns the raw pointer underlying p when p is a
-// smart pointer, or returns p itself when p is already a raw pointer.
-// The following default implementation is for the smart pointer case.
-template <typename Pointer>
-inline const typename Pointer::element_type* GetRawPointer(const Pointer& p) {
-  return p.get();
-}
-// This overloaded version is for the raw pointer case.
-template <typename Element>
-inline Element* GetRawPointer(Element* p) { return p; }
-
-// This comparator allows linked_ptr to be stored in sets.
-template <typename T>
-struct LinkedPtrLessThan {
-  bool operator()(const ::testing::internal::linked_ptr<T>& lhs,
-                  const ::testing::internal::linked_ptr<T>& rhs) const {
-    return lhs.get() < rhs.get();
-  }
-};
-
-// Symbian compilation can be done with wchar_t being either a native
-// type or a typedef.  Using Google Mock with OpenC without wchar_t
-// should require the definition of _STLP_NO_WCHAR_T.
-//
-// MSVC treats wchar_t as a native type usually, but treats it as the
-// same as unsigned short when the compiler option /Zc:wchar_t- is
-// specified.  It defines _NATIVE_WCHAR_T_DEFINED symbol when wchar_t
-// is a native type.
-#if (GTEST_OS_SYMBIAN && defined(_STLP_NO_WCHAR_T)) || \
-    (defined(_MSC_VER) && !defined(_NATIVE_WCHAR_T_DEFINED))
-// wchar_t is a typedef.
-#else
-# define GMOCK_WCHAR_T_IS_NATIVE_ 1
-#endif
-
-// signed wchar_t and unsigned wchar_t are NOT in the C++ standard.
-// Using them is a bad practice and not portable.  So DON'T use them.
-//
-// Still, Google Mock is designed to work even if the user uses signed
-// wchar_t or unsigned wchar_t (obviously, assuming the compiler
-// supports them).
-//
-// To gcc,
-//   wchar_t == signed wchar_t != unsigned wchar_t == unsigned int
-#ifdef __GNUC__
-#if !defined(__WCHAR_UNSIGNED__)
-// signed/unsigned wchar_t are valid types.
-# define GMOCK_HAS_SIGNED_WCHAR_T_ 1
-#endif
-#endif
-
-// In what follows, we use the term "kind" to indicate whether a type
-// is bool, an integer type (excluding bool), a floating-point type,
-// or none of them.  This categorization is useful for determining
-// when a matcher argument type can be safely converted to another
-// type in the implementation of SafeMatcherCast.
-enum TypeKind {
-  kBool, kInteger, kFloatingPoint, kOther
-};
-
-// KindOf<T>::value is the kind of type T.
-template <typename T> struct KindOf {
-  enum { value = kOther };  // The default kind.
-};
-
-// This macro declares that the kind of 'type' is 'kind'.
-#define GMOCK_DECLARE_KIND_(type, kind) \
-  template <> struct KindOf<type> { enum { value = kind }; }
-
-GMOCK_DECLARE_KIND_(bool, kBool);
-
-// All standard integer types.
-GMOCK_DECLARE_KIND_(char, kInteger);
-GMOCK_DECLARE_KIND_(signed char, kInteger);
-GMOCK_DECLARE_KIND_(unsigned char, kInteger);
-GMOCK_DECLARE_KIND_(short, kInteger);  // NOLINT
-GMOCK_DECLARE_KIND_(unsigned short, kInteger);  // NOLINT
-GMOCK_DECLARE_KIND_(int, kInteger);
-GMOCK_DECLARE_KIND_(unsigned int, kInteger);
-GMOCK_DECLARE_KIND_(long, kInteger);  // NOLINT
-GMOCK_DECLARE_KIND_(unsigned long, kInteger);  // NOLINT
-
-#if GMOCK_WCHAR_T_IS_NATIVE_
-GMOCK_DECLARE_KIND_(wchar_t, kInteger);
-#endif
-
-// Non-standard integer types.
-GMOCK_DECLARE_KIND_(Int64, kInteger);
-GMOCK_DECLARE_KIND_(UInt64, kInteger);
-
-// All standard floating-point types.
-GMOCK_DECLARE_KIND_(float, kFloatingPoint);
-GMOCK_DECLARE_KIND_(double, kFloatingPoint);
-GMOCK_DECLARE_KIND_(long double, kFloatingPoint);
-
-#undef GMOCK_DECLARE_KIND_
-
-// Evaluates to the kind of 'type'.
-#define GMOCK_KIND_OF_(type) \
-  static_cast< ::testing::internal::TypeKind>( \
-      ::testing::internal::KindOf<type>::value)
-
-// Evaluates to true iff integer type T is signed.
-#define GMOCK_IS_SIGNED_(T) (static_cast<T>(-1) < 0)
-
-// LosslessArithmeticConvertibleImpl<kFromKind, From, kToKind, To>::value
-// is true iff arithmetic type From can be losslessly converted to
-// arithmetic type To.
-//
-// It's the user's responsibility to ensure that both From and To are
-// raw (i.e. has no CV modifier, is not a pointer, and is not a
-// reference) built-in arithmetic types, kFromKind is the kind of
-// From, and kToKind is the kind of To; the value is
-// implementation-defined when the above pre-condition is violated.
-template <TypeKind kFromKind, typename From, TypeKind kToKind, typename To>
-struct LosslessArithmeticConvertibleImpl : public false_type {};
-
-// Converting bool to bool is lossless.
-template <>
-struct LosslessArithmeticConvertibleImpl<kBool, bool, kBool, bool>
-    : public true_type {};  // NOLINT
-
-// Converting bool to any integer type is lossless.
-template <typename To>
-struct LosslessArithmeticConvertibleImpl<kBool, bool, kInteger, To>
-    : public true_type {};  // NOLINT
-
-// Converting bool to any floating-point type is lossless.
-template <typename To>
-struct LosslessArithmeticConvertibleImpl<kBool, bool, kFloatingPoint, To>
-    : public true_type {};  // NOLINT
-
-// Converting an integer to bool is lossy.
-template <typename From>
-struct LosslessArithmeticConvertibleImpl<kInteger, From, kBool, bool>
-    : public false_type {};  // NOLINT
-
-// Converting an integer to another non-bool integer is lossless iff
-// the target type's range encloses the source type's range.
-template <typename From, typename To>
-struct LosslessArithmeticConvertibleImpl<kInteger, From, kInteger, To>
-    : public bool_constant<
-      // When converting from a smaller size to a larger size, we are
-      // fine as long as we are not converting from signed to unsigned.
-      ((sizeof(From) < sizeof(To)) &&
-       (!GMOCK_IS_SIGNED_(From) || GMOCK_IS_SIGNED_(To))) ||
-      // When converting between the same size, the signedness must match.
-      ((sizeof(From) == sizeof(To)) &&
-       (GMOCK_IS_SIGNED_(From) == GMOCK_IS_SIGNED_(To)))> {};  // NOLINT
-
-#undef GMOCK_IS_SIGNED_
-
-// Converting an integer to a floating-point type may be lossy, since
-// the format of a floating-point number is implementation-defined.
-template <typename From, typename To>
-struct LosslessArithmeticConvertibleImpl<kInteger, From, kFloatingPoint, To>
-    : public false_type {};  // NOLINT
-
-// Converting a floating-point to bool is lossy.
-template <typename From>
-struct LosslessArithmeticConvertibleImpl<kFloatingPoint, From, kBool, bool>
-    : public false_type {};  // NOLINT
-
-// Converting a floating-point to an integer is lossy.
-template <typename From, typename To>
-struct LosslessArithmeticConvertibleImpl<kFloatingPoint, From, kInteger, To>
-    : public false_type {};  // NOLINT
-
-// Converting a floating-point to another floating-point is lossless
-// iff the target type is at least as big as the source type.
-template <typename From, typename To>
-struct LosslessArithmeticConvertibleImpl<
-  kFloatingPoint, From, kFloatingPoint, To>
-    : public bool_constant<sizeof(From) <= sizeof(To)> {};  // NOLINT
-
-// LosslessArithmeticConvertible<From, To>::value is true iff arithmetic
-// type From can be losslessly converted to arithmetic type To.
-//
-// It's the user's responsibility to ensure that both From and To are
-// raw (i.e. has no CV modifier, is not a pointer, and is not a
-// reference) built-in arithmetic types; the value is
-// implementation-defined when the above pre-condition is violated.
-template <typename From, typename To>
-struct LosslessArithmeticConvertible
-    : public LosslessArithmeticConvertibleImpl<
-  GMOCK_KIND_OF_(From), From, GMOCK_KIND_OF_(To), To> {};  // NOLINT
-
-// This interface knows how to report a Google Mock failure (either
-// non-fatal or fatal).
-class FailureReporterInterface {
- public:
-  // The type of a failure (either non-fatal or fatal).
-  enum FailureType {
-    kNonfatal, kFatal
-  };
-
-  virtual ~FailureReporterInterface() {}
-
-  // Reports a failure that occurred at the given source file location.
-  virtual void ReportFailure(FailureType type, const char* file, int line,
-                             const std::string& message) = 0;
-};
-
-// Returns the failure reporter used by Google Mock.
-GTEST_API_ FailureReporterInterface* GetFailureReporter();
-
-// Asserts that condition is true; aborts the process with the given
-// message if condition is false.  We cannot use LOG(FATAL) or CHECK()
-// as Google Mock might be used to mock the log sink itself.  We
-// inline this function to prevent it from showing up in the stack
-// trace.
-inline void Assert(bool condition, const char* file, int line,
-                   const std::string& msg) {
-  if (!condition) {
-    GetFailureReporter()->ReportFailure(FailureReporterInterface::kFatal,
-                                        file, line, msg);
-  }
-}
-inline void Assert(bool condition, const char* file, int line) {
-  Assert(condition, file, line, "Assertion failed.");
-}
-
-// Verifies that condition is true; generates a non-fatal failure if
-// condition is false.
-inline void Expect(bool condition, const char* file, int line,
-                   const std::string& msg) {
-  if (!condition) {
-    GetFailureReporter()->ReportFailure(FailureReporterInterface::kNonfatal,
-                                        file, line, msg);
-  }
-}
-inline void Expect(bool condition, const char* file, int line) {
-  Expect(condition, file, line, "Expectation failed.");
-}
-
-// Severity level of a log.
-enum LogSeverity {
-  kInfo = 0,
-  kWarning = 1
-};
-
-// Valid values for the --gmock_verbose flag.
-
-// All logs (informational and warnings) are printed.
-const char kInfoVerbosity[] = "info";
-// Only warnings are printed.
-const char kWarningVerbosity[] = "warning";
-// No logs are printed.
-const char kErrorVerbosity[] = "error";
-
-// Returns true iff a log with the given severity is visible according
-// to the --gmock_verbose flag.
-GTEST_API_ bool LogIsVisible(LogSeverity severity);
-
-// Prints the given message to stdout iff 'severity' >= the level
-// specified by the --gmock_verbose flag.  If stack_frames_to_skip >=
-// 0, also prints the stack trace excluding the top
-// stack_frames_to_skip frames.  In opt mode, any positive
-// stack_frames_to_skip is treated as 0, since we don't know which
-// function calls will be inlined by the compiler and need to be
-// conservative.
-GTEST_API_ void Log(LogSeverity severity, const std::string& message,
-                    int stack_frames_to_skip);
-
-// A marker class that is used to resolve parameterless expectations to the
-// correct overload. This must not be instantiable, to prevent client code from
-// accidentally resolving to the overload; for example:
-//
-//    ON_CALL(mock, Method({}, nullptr))...
-//
-class WithoutMatchers {
- private:
-  WithoutMatchers() {}
-  friend GTEST_API_ WithoutMatchers GetWithoutMatchers();
-};
-
-// Internal use only: access the singleton instance of WithoutMatchers.
-GTEST_API_ WithoutMatchers GetWithoutMatchers();
-
-// TODO(wan@google.com): group all type utilities together.
-
-// Type traits.
-
-// is_reference<T>::value is non-zero iff T is a reference type.
-template <typename T> struct is_reference : public false_type {};
-template <typename T> struct is_reference<T&> : public true_type {};
-
-// type_equals<T1, T2>::value is non-zero iff T1 and T2 are the same type.
-template <typename T1, typename T2> struct type_equals : public false_type {};
-template <typename T> struct type_equals<T, T> : public true_type {};
-
-// remove_reference<T>::type removes the reference from type T, if any.
-template <typename T> struct remove_reference { typedef T type; };  // NOLINT
-template <typename T> struct remove_reference<T&> { typedef T type; }; // NOLINT
-
-// DecayArray<T>::type turns an array type U[N] to const U* and preserves
-// other types.  Useful for saving a copy of a function argument.
-template <typename T> struct DecayArray { typedef T type; };  // NOLINT
-template <typename T, size_t N> struct DecayArray<T[N]> {
-  typedef const T* type;
-};
-// Sometimes people use arrays whose size is not available at the use site
-// (e.g. extern const char kNamePrefix[]).  This specialization covers that
-// case.
-template <typename T> struct DecayArray<T[]> {
-  typedef const T* type;
-};
-
-// Disable MSVC warnings for infinite recursion, since in this case the
-// the recursion is unreachable.
-#ifdef _MSC_VER
-# pragma warning(push)
-# pragma warning(disable:4717)
-#endif
-
-// Invalid<T>() is usable as an expression of type T, but will terminate
-// the program with an assertion failure if actually run.  This is useful
-// when a value of type T is needed for compilation, but the statement
-// will not really be executed (or we don't care if the statement
-// crashes).
-template <typename T>
-inline T Invalid() {
-  Assert(false, "", -1, "Internal error: attempt to return invalid value");
-  // This statement is unreachable, and would never terminate even if it
-  // could be reached. It is provided only to placate compiler warnings
-  // about missing return statements.
-  return Invalid<T>();
-}
-
-#ifdef _MSC_VER
-# pragma warning(pop)
-#endif
-
-// Given a raw type (i.e. having no top-level reference or const
-// modifier) RawContainer that's either an STL-style container or a
-// native array, class StlContainerView<RawContainer> has the
-// following members:
-//
-//   - type is a type that provides an STL-style container view to
-//     (i.e. implements the STL container concept for) RawContainer;
-//   - const_reference is a type that provides a reference to a const
-//     RawContainer;
-//   - ConstReference(raw_container) returns a const reference to an STL-style
-//     container view to raw_container, which is a RawContainer.
-//   - Copy(raw_container) returns an STL-style container view of a
-//     copy of raw_container, which is a RawContainer.
-//
-// This generic version is used when RawContainer itself is already an
-// STL-style container.
-template <class RawContainer>
-class StlContainerView {
- public:
-  typedef RawContainer type;
-  typedef const type& const_reference;
-
-  static const_reference ConstReference(const RawContainer& container) {
-    // Ensures that RawContainer is not a const type.
-    testing::StaticAssertTypeEq<RawContainer,
-        GTEST_REMOVE_CONST_(RawContainer)>();
-    return container;
-  }
-  static type Copy(const RawContainer& container) { return container; }
-};
-
-// This specialization is used when RawContainer is a native array type.
-template <typename Element, size_t N>
-class StlContainerView<Element[N]> {
- public:
-  typedef GTEST_REMOVE_CONST_(Element) RawElement;
-  typedef internal::NativeArray<RawElement> type;
-  // NativeArray<T> can represent a native array either by value or by
-  // reference (selected by a constructor argument), so 'const type'
-  // can be used to reference a const native array.  We cannot
-  // 'typedef const type& const_reference' here, as that would mean
-  // ConstReference() has to return a reference to a local variable.
-  typedef const type const_reference;
-
-  static const_reference ConstReference(const Element (&array)[N]) {
-    // Ensures that Element is not a const type.
-    testing::StaticAssertTypeEq<Element, RawElement>();
-#if GTEST_OS_SYMBIAN
-    // The Nokia Symbian compiler confuses itself in template instantiation
-    // for this call without the cast to Element*:
-    // function call '[testing::internal::NativeArray<char *>].NativeArray(
-    //     {lval} const char *[4], long, testing::internal::RelationToSource)'
-    //     does not match
-    // 'testing::internal::NativeArray<char *>::NativeArray(
-    //     char *const *, unsigned int, testing::internal::RelationToSource)'
-    // (instantiating: 'testing::internal::ContainsMatcherImpl
-    //     <const char * (&)[4]>::Matches(const char * (&)[4]) const')
-    // (instantiating: 'testing::internal::StlContainerView<char *[4]>::
-    //     ConstReference(const char * (&)[4])')
-    // (and though the N parameter type is mismatched in the above explicit
-    // conversion of it doesn't help - only the conversion of the array).
-    return type(const_cast<Element*>(&array[0]), N,
-                RelationToSourceReference());
-#else
-    return type(array, N, RelationToSourceReference());
-#endif  // GTEST_OS_SYMBIAN
-  }
-  static type Copy(const Element (&array)[N]) {
-#if GTEST_OS_SYMBIAN
-    return type(const_cast<Element*>(&array[0]), N, RelationToSourceCopy());
-#else
-    return type(array, N, RelationToSourceCopy());
-#endif  // GTEST_OS_SYMBIAN
-  }
-};
-
-// This specialization is used when RawContainer is a native array
-// represented as a (pointer, size) tuple.
-template <typename ElementPointer, typename Size>
-class StlContainerView< ::testing::tuple<ElementPointer, Size> > {
- public:
-  typedef GTEST_REMOVE_CONST_(
-      typename internal::PointeeOf<ElementPointer>::type) RawElement;
-  typedef internal::NativeArray<RawElement> type;
-  typedef const type const_reference;
-
-  static const_reference ConstReference(
-      const ::testing::tuple<ElementPointer, Size>& array) {
-    return type(get<0>(array), get<1>(array), RelationToSourceReference());
-  }
-  static type Copy(const ::testing::tuple<ElementPointer, Size>& array) {
-    return type(get<0>(array), get<1>(array), RelationToSourceCopy());
-  }
-};
-
-// The following specialization prevents the user from instantiating
-// StlContainer with a reference type.
-template <typename T> class StlContainerView<T&>;
-
-// A type transform to remove constness from the first part of a pair.
-// Pairs like that are used as the value_type of associative containers,
-// and this transform produces a similar but assignable pair.
-template <typename T>
-struct RemoveConstFromKey {
-  typedef T type;
-};
-
-// Partially specialized to remove constness from std::pair<const K, V>.
-template <typename K, typename V>
-struct RemoveConstFromKey<std::pair<const K, V> > {
-  typedef std::pair<K, V> type;
-};
-
-// Mapping from booleans to types. Similar to boost::bool_<kValue> and
-// std::integral_constant<bool, kValue>.
-template <bool kValue>
-struct BooleanConstant {};
-
-// Emit an assertion failure due to incorrect DoDefault() usage. Out-of-lined to
-// reduce code size.
-GTEST_API_ void IllegalDoDefault(const char* file, int line);
-
-#if GTEST_LANG_CXX11
-// Helper types for Apply() below.
-template <size_t... Is> struct int_pack { typedef int_pack type; };
-
-template <class Pack, size_t I> struct append;
-template <size_t... Is, size_t I>
-struct append<int_pack<Is...>, I> : int_pack<Is..., I> {};
-
-template <size_t C>
-struct make_int_pack : append<typename make_int_pack<C - 1>::type, C - 1> {};
-template <> struct make_int_pack<0> : int_pack<> {};
-
-template <typename F, typename Tuple, size_t... Idx>
-auto ApplyImpl(F&& f, Tuple&& args, int_pack<Idx...>) -> decltype(
-    std::forward<F>(f)(std::get<Idx>(std::forward<Tuple>(args))...)) {
-  return std::forward<F>(f)(std::get<Idx>(std::forward<Tuple>(args))...);
-}
-
-// Apply the function to a tuple of arguments.
-template <typename F, typename Tuple>
-auto Apply(F&& f, Tuple&& args)
-    -> decltype(ApplyImpl(std::forward<F>(f), std::forward<Tuple>(args),
-                          make_int_pack<std::tuple_size<Tuple>::value>())) {
-  return ApplyImpl(std::forward<F>(f), std::forward<Tuple>(args),
-                   make_int_pack<std::tuple_size<Tuple>::value>());
-}
-#endif
-
-
-#ifdef _MSC_VER
-# pragma warning(pop)
-#endif
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GMOCK_INCLUDE_GMOCK_INTERNAL_GMOCK_INTERNAL_UTILS_H_
-
-#if GTEST_LANG_CXX11  // Defined by gtest-port.h via gmock-port.h.
-#include <functional>
-#include <type_traits>
-#endif  // GTEST_LANG_CXX11
-
-namespace testing {
-
-// To implement an action Foo, define:
-//   1. a class FooAction that implements the ActionInterface interface, and
-//   2. a factory function that creates an Action object from a
-//      const FooAction*.
-//
-// The two-level delegation design follows that of Matcher, providing
-// consistency for extension developers.  It also eases ownership
-// management as Action objects can now be copied like plain values.
-
-namespace internal {
-
-template <typename F1, typename F2>
-class ActionAdaptor;
-
-// BuiltInDefaultValueGetter<T, true>::Get() returns a
-// default-constructed T value.  BuiltInDefaultValueGetter<T,
-// false>::Get() crashes with an error.
-//
-// This primary template is used when kDefaultConstructible is true.
-template <typename T, bool kDefaultConstructible>
-struct BuiltInDefaultValueGetter {
-  static T Get() { return T(); }
-};
-template <typename T>
-struct BuiltInDefaultValueGetter<T, false> {
-  static T Get() {
-    Assert(false, __FILE__, __LINE__,
-           "Default action undefined for the function return type.");
-    return internal::Invalid<T>();
-    // The above statement will never be reached, but is required in
-    // order for this function to compile.
-  }
-};
-
-// BuiltInDefaultValue<T>::Get() returns the "built-in" default value
-// for type T, which is NULL when T is a raw pointer type, 0 when T is
-// a numeric type, false when T is bool, or "" when T is string or
-// std::string.  In addition, in C++11 and above, it turns a
-// default-constructed T value if T is default constructible.  For any
-// other type T, the built-in default T value is undefined, and the
-// function will abort the process.
-template <typename T>
-class BuiltInDefaultValue {
- public:
-#if GTEST_LANG_CXX11
-  // This function returns true iff type T has a built-in default value.
-  static bool Exists() {
-    return ::std::is_default_constructible<T>::value;
-  }
-
-  static T Get() {
-    return BuiltInDefaultValueGetter<
-        T, ::std::is_default_constructible<T>::value>::Get();
-  }
-
-#else  // GTEST_LANG_CXX11
-  // This function returns true iff type T has a built-in default value.
-  static bool Exists() {
-    return false;
-  }
-
-  static T Get() {
-    return BuiltInDefaultValueGetter<T, false>::Get();
-  }
-
-#endif  // GTEST_LANG_CXX11
-};
-
-// This partial specialization says that we use the same built-in
-// default value for T and const T.
-template <typename T>
-class BuiltInDefaultValue<const T> {
- public:
-  static bool Exists() { return BuiltInDefaultValue<T>::Exists(); }
-  static T Get() { return BuiltInDefaultValue<T>::Get(); }
-};
-
-// This partial specialization defines the default values for pointer
-// types.
-template <typename T>
-class BuiltInDefaultValue<T*> {
- public:
-  static bool Exists() { return true; }
-  static T* Get() { return NULL; }
-};
-
-// The following specializations define the default values for
-// specific types we care about.
-#define GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(type, value) \
-  template <> \
-  class BuiltInDefaultValue<type> { \
-   public: \
-    static bool Exists() { return true; } \
-    static type Get() { return value; } \
-  }
-
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(void, );  // NOLINT
-#if GTEST_HAS_GLOBAL_STRING
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(::string, "");
-#endif  // GTEST_HAS_GLOBAL_STRING
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(::std::string, "");
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(bool, false);
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(unsigned char, '\0');
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(signed char, '\0');
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(char, '\0');
-
-// There's no need for a default action for signed wchar_t, as that
-// type is the same as wchar_t for gcc, and invalid for MSVC.
-//
-// There's also no need for a default action for unsigned wchar_t, as
-// that type is the same as unsigned int for gcc, and invalid for
-// MSVC.
-#if GMOCK_WCHAR_T_IS_NATIVE_
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(wchar_t, 0U);  // NOLINT
-#endif
-
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(unsigned short, 0U);  // NOLINT
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(signed short, 0);     // NOLINT
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(unsigned int, 0U);
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(signed int, 0);
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(unsigned long, 0UL);  // NOLINT
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(signed long, 0L);     // NOLINT
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(UInt64, 0);
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(Int64, 0);
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(float, 0);
-GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_(double, 0);
-
-#undef GMOCK_DEFINE_DEFAULT_ACTION_FOR_RETURN_TYPE_
-
-}  // namespace internal
-
-// When an unexpected function call is encountered, Google Mock will
-// let it return a default value if the user has specified one for its
-// return type, or if the return type has a built-in default value;
-// otherwise Google Mock won't know what value to return and will have
-// to abort the process.
-//
-// The DefaultValue<T> class allows a user to specify the
-// default value for a type T that is both copyable and publicly
-// destructible (i.e. anything that can be used as a function return
-// type).  The usage is:
-//
-//   // Sets the default value for type T to be foo.
-//   DefaultValue<T>::Set(foo);
-template <typename T>
-class DefaultValue {
- public:
-  // Sets the default value for type T; requires T to be
-  // copy-constructable and have a public destructor.
-  static void Set(T x) {
-    delete producer_;
-    producer_ = new FixedValueProducer(x);
-  }
-
-  // Provides a factory function to be called to generate the default value.
-  // This method can be used even if T is only move-constructible, but it is not
-  // limited to that case.
-  typedef T (*FactoryFunction)();
-  static void SetFactory(FactoryFunction factory) {
-    delete producer_;
-    producer_ = new FactoryValueProducer(factory);
-  }
-
-  // Unsets the default value for type T.
-  static void Clear() {
-    delete producer_;
-    producer_ = NULL;
-  }
-
-  // Returns true iff the user has set the default value for type T.
-  static bool IsSet() { return producer_ != NULL; }
-
-  // Returns true if T has a default return value set by the user or there
-  // exists a built-in default value.
-  static bool Exists() {
-    return IsSet() || internal::BuiltInDefaultValue<T>::Exists();
-  }
-
-  // Returns the default value for type T if the user has set one;
-  // otherwise returns the built-in default value. Requires that Exists()
-  // is true, which ensures that the return value is well-defined.
-  static T Get() {
-    return producer_ == NULL ?
-        internal::BuiltInDefaultValue<T>::Get() : producer_->Produce();
-  }
-
- private:
-  class ValueProducer {
-   public:
-    virtual ~ValueProducer() {}
-    virtual T Produce() = 0;
-  };
-
-  class FixedValueProducer : public ValueProducer {
-   public:
-    explicit FixedValueProducer(T value) : value_(value) {}
-    virtual T Produce() { return value_; }
-
-   private:
-    const T value_;
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(FixedValueProducer);
-  };
-
-  class FactoryValueProducer : public ValueProducer {
-   public:
-    explicit FactoryValueProducer(FactoryFunction factory)
-        : factory_(factory) {}
-    virtual T Produce() { return factory_(); }
-
-   private:
-    const FactoryFunction factory_;
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(FactoryValueProducer);
-  };
-
-  static ValueProducer* producer_;
-};
-
-// This partial specialization allows a user to set default values for
-// reference types.
-template <typename T>
-class DefaultValue<T&> {
- public:
-  // Sets the default value for type T&.
-  static void Set(T& x) {  // NOLINT
-    address_ = &x;
-  }
-
-  // Unsets the default value for type T&.
-  static void Clear() {
-    address_ = NULL;
-  }
-
-  // Returns true iff the user has set the default value for type T&.
-  static bool IsSet() { return address_ != NULL; }
-
-  // Returns true if T has a default return value set by the user or there
-  // exists a built-in default value.
-  static bool Exists() {
-    return IsSet() || internal::BuiltInDefaultValue<T&>::Exists();
-  }
-
-  // Returns the default value for type T& if the user has set one;
-  // otherwise returns the built-in default value if there is one;
-  // otherwise aborts the process.
-  static T& Get() {
-    return address_ == NULL ?
-        internal::BuiltInDefaultValue<T&>::Get() : *address_;
-  }
-
- private:
-  static T* address_;
-};
-
-// This specialization allows DefaultValue<void>::Get() to
-// compile.
-template <>
-class DefaultValue<void> {
- public:
-  static bool Exists() { return true; }
-  static void Get() {}
-};
-
-// Points to the user-set default value for type T.
-template <typename T>
-typename DefaultValue<T>::ValueProducer* DefaultValue<T>::producer_ = NULL;
-
-// Points to the user-set default value for type T&.
-template <typename T>
-T* DefaultValue<T&>::address_ = NULL;
-
-// Implement this interface to define an action for function type F.
-template <typename F>
-class ActionInterface {
- public:
-  typedef typename internal::Function<F>::Result Result;
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  ActionInterface() {}
-  virtual ~ActionInterface() {}
-
-  // Performs the action.  This method is not const, as in general an
-  // action can have side effects and be stateful.  For example, a
-  // get-the-next-element-from-the-collection action will need to
-  // remember the current element.
-  virtual Result Perform(const ArgumentTuple& args) = 0;
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ActionInterface);
-};
-
-// An Action<F> is a copyable and IMMUTABLE (except by assignment)
-// object that represents an action to be taken when a mock function
-// of type F is called.  The implementation of Action<T> is just a
-// linked_ptr to const ActionInterface<T>, so copying is fairly cheap.
-// Don't inherit from Action!
-//
-// You can view an object implementing ActionInterface<F> as a
-// concrete action (including its current state), and an Action<F>
-// object as a handle to it.
-template <typename F>
-class Action {
- public:
-  typedef typename internal::Function<F>::Result Result;
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  // Constructs a null Action.  Needed for storing Action objects in
-  // STL containers.
-  Action() {}
-
-#if GTEST_LANG_CXX11
-  // Construct an Action from a specified callable.
-  // This cannot take std::function directly, because then Action would not be
-  // directly constructible from lambda (it would require two conversions).
-  template <typename G,
-            typename = typename ::std::enable_if<
-                ::std::is_constructible<::std::function<F>, G>::value>::type>
-  Action(G&& fun) : fun_(::std::forward<G>(fun)) {}  // NOLINT
-#endif
-
-  // Constructs an Action from its implementation.
-  explicit Action(ActionInterface<F>* impl) : impl_(impl) {}
-
-  // This constructor allows us to turn an Action<Func> object into an
-  // Action<F>, as long as F's arguments can be implicitly converted
-  // to Func's and Func's return type can be implicitly converted to
-  // F's.
-  template <typename Func>
-  explicit Action(const Action<Func>& action);
-
-  // Returns true iff this is the DoDefault() action.
-  bool IsDoDefault() const {
-#if GTEST_LANG_CXX11
-    return impl_ == nullptr && fun_ == nullptr;
-#else
-    return impl_ == NULL;
-#endif
-  }
-
-  // Performs the action.  Note that this method is const even though
-  // the corresponding method in ActionInterface is not.  The reason
-  // is that a const Action<F> means that it cannot be re-bound to
-  // another concrete action, not that the concrete action it binds to
-  // cannot change state.  (Think of the difference between a const
-  // pointer and a pointer to const.)
-  Result Perform(ArgumentTuple args) const {
-    if (IsDoDefault()) {
-      internal::IllegalDoDefault(__FILE__, __LINE__);
-    }
-#if GTEST_LANG_CXX11
-    if (fun_ != nullptr) {
-      return internal::Apply(fun_, ::std::move(args));
-    }
-#endif
-    return impl_->Perform(args);
-  }
-
- private:
-  template <typename F1, typename F2>
-  friend class internal::ActionAdaptor;
-
-  template <typename G>
-  friend class Action;
-
-  // In C++11, Action can be implemented either as a generic functor (through
-  // std::function), or legacy ActionInterface. In C++98, only ActionInterface
-  // is available. The invariants are as follows:
-  // * in C++98, impl_ is null iff this is the default action
-  // * in C++11, at most one of fun_ & impl_ may be nonnull; both are null iff
-  //   this is the default action
-#if GTEST_LANG_CXX11
-  ::std::function<F> fun_;
-#endif
-  internal::linked_ptr<ActionInterface<F> > impl_;
-};
-
-// The PolymorphicAction class template makes it easy to implement a
-// polymorphic action (i.e. an action that can be used in mock
-// functions of than one type, e.g. Return()).
-//
-// To define a polymorphic action, a user first provides a COPYABLE
-// implementation class that has a Perform() method template:
-//
-//   class FooAction {
-//    public:
-//     template <typename Result, typename ArgumentTuple>
-//     Result Perform(const ArgumentTuple& args) const {
-//       // Processes the arguments and returns a result, using
-//       // tr1::get<N>(args) to get the N-th (0-based) argument in the tuple.
-//     }
-//     ...
-//   };
-//
-// Then the user creates the polymorphic action using
-// MakePolymorphicAction(object) where object has type FooAction.  See
-// the definition of Return(void) and SetArgumentPointee<N>(value) for
-// complete examples.
-template <typename Impl>
-class PolymorphicAction {
- public:
-  explicit PolymorphicAction(const Impl& impl) : impl_(impl) {}
-
-  template <typename F>
-  operator Action<F>() const {
-    return Action<F>(new MonomorphicImpl<F>(impl_));
-  }
-
- private:
-  template <typename F>
-  class MonomorphicImpl : public ActionInterface<F> {
-   public:
-    typedef typename internal::Function<F>::Result Result;
-    typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-    explicit MonomorphicImpl(const Impl& impl) : impl_(impl) {}
-
-    virtual Result Perform(const ArgumentTuple& args) {
-      return impl_.template Perform<Result>(args);
-    }
-
-   private:
-    Impl impl_;
-
-    GTEST_DISALLOW_ASSIGN_(MonomorphicImpl);
-  };
-
-  Impl impl_;
-
-  GTEST_DISALLOW_ASSIGN_(PolymorphicAction);
-};
-
-// Creates an Action from its implementation and returns it.  The
-// created Action object owns the implementation.
-template <typename F>
-Action<F> MakeAction(ActionInterface<F>* impl) {
-  return Action<F>(impl);
-}
-
-// Creates a polymorphic action from its implementation.  This is
-// easier to use than the PolymorphicAction<Impl> constructor as it
-// doesn't require you to explicitly write the template argument, e.g.
-//
-//   MakePolymorphicAction(foo);
-// vs
-//   PolymorphicAction<TypeOfFoo>(foo);
-template <typename Impl>
-inline PolymorphicAction<Impl> MakePolymorphicAction(const Impl& impl) {
-  return PolymorphicAction<Impl>(impl);
-}
-
-namespace internal {
-
-// Allows an Action<F2> object to pose as an Action<F1>, as long as F2
-// and F1 are compatible.
-template <typename F1, typename F2>
-class ActionAdaptor : public ActionInterface<F1> {
- public:
-  typedef typename internal::Function<F1>::Result Result;
-  typedef typename internal::Function<F1>::ArgumentTuple ArgumentTuple;
-
-  explicit ActionAdaptor(const Action<F2>& from) : impl_(from.impl_) {}
-
-  virtual Result Perform(const ArgumentTuple& args) {
-    return impl_->Perform(args);
-  }
-
- private:
-  const internal::linked_ptr<ActionInterface<F2> > impl_;
-
-  GTEST_DISALLOW_ASSIGN_(ActionAdaptor);
-};
-
-// Helper struct to specialize ReturnAction to execute a move instead of a copy
-// on return. Useful for move-only types, but could be used on any type.
-template <typename T>
-struct ByMoveWrapper {
-  explicit ByMoveWrapper(T value) : payload(internal::move(value)) {}
-  T payload;
-};
-
-// Implements the polymorphic Return(x) action, which can be used in
-// any function that returns the type of x, regardless of the argument
-// types.
-//
-// Note: The value passed into Return must be converted into
-// Function<F>::Result when this action is cast to Action<F> rather than
-// when that action is performed. This is important in scenarios like
-//
-// MOCK_METHOD1(Method, T(U));
-// ...
-// {
-//   Foo foo;
-//   X x(&foo);
-//   EXPECT_CALL(mock, Method(_)).WillOnce(Return(x));
-// }
-//
-// In the example above the variable x holds reference to foo which leaves
-// scope and gets destroyed.  If copying X just copies a reference to foo,
-// that copy will be left with a hanging reference.  If conversion to T
-// makes a copy of foo, the above code is safe. To support that scenario, we
-// need to make sure that the type conversion happens inside the EXPECT_CALL
-// statement, and conversion of the result of Return to Action<T(U)> is a
-// good place for that.
-//
-// The real life example of the above scenario happens when an invocation
-// of gtl::Container() is passed into Return.
-//
-template <typename R>
-class ReturnAction {
- public:
-  // Constructs a ReturnAction object from the value to be returned.
-  // 'value' is passed by value instead of by const reference in order
-  // to allow Return("string literal") to compile.
-  explicit ReturnAction(R value) : value_(new R(internal::move(value))) {}
-
-  // This template type conversion operator allows Return(x) to be
-  // used in ANY function that returns x's type.
-  template <typename F>
-  operator Action<F>() const {
-    // Assert statement belongs here because this is the best place to verify
-    // conditions on F. It produces the clearest error messages
-    // in most compilers.
-    // Impl really belongs in this scope as a local class but can't
-    // because MSVC produces duplicate symbols in different translation units
-    // in this case. Until MS fixes that bug we put Impl into the class scope
-    // and put the typedef both here (for use in assert statement) and
-    // in the Impl class. But both definitions must be the same.
-    typedef typename Function<F>::Result Result;
-    GTEST_COMPILE_ASSERT_(
-        !is_reference<Result>::value,
-        use_ReturnRef_instead_of_Return_to_return_a_reference);
-    return Action<F>(new Impl<R, F>(value_));
-  }
-
- private:
-  // Implements the Return(x) action for a particular function type F.
-  template <typename R_, typename F>
-  class Impl : public ActionInterface<F> {
-   public:
-    typedef typename Function<F>::Result Result;
-    typedef typename Function<F>::ArgumentTuple ArgumentTuple;
-
-    // The implicit cast is necessary when Result has more than one
-    // single-argument constructor (e.g. Result is std::vector<int>) and R
-    // has a type conversion operator template.  In that case, value_(value)
-    // won't compile as the compiler doesn't known which constructor of
-    // Result to call.  ImplicitCast_ forces the compiler to convert R to
-    // Result without considering explicit constructors, thus resolving the
-    // ambiguity. value_ is then initialized using its copy constructor.
-    explicit Impl(const linked_ptr<R>& value)
-        : value_before_cast_(*value),
-          value_(ImplicitCast_<Result>(value_before_cast_)) {}
-
-    virtual Result Perform(const ArgumentTuple&) { return value_; }
-
-   private:
-    GTEST_COMPILE_ASSERT_(!is_reference<Result>::value,
-                          Result_cannot_be_a_reference_type);
-    // We save the value before casting just in case it is being cast to a
-    // wrapper type.
-    R value_before_cast_;
-    Result value_;
-
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(Impl);
-  };
-
-  // Partially specialize for ByMoveWrapper. This version of ReturnAction will
-  // move its contents instead.
-  template <typename R_, typename F>
-  class Impl<ByMoveWrapper<R_>, F> : public ActionInterface<F> {
-   public:
-    typedef typename Function<F>::Result Result;
-    typedef typename Function<F>::ArgumentTuple ArgumentTuple;
-
-    explicit Impl(const linked_ptr<R>& wrapper)
-        : performed_(false), wrapper_(wrapper) {}
-
-    virtual Result Perform(const ArgumentTuple&) {
-      GTEST_CHECK_(!performed_)
-          << "A ByMove() action should only be performed once.";
-      performed_ = true;
-      return internal::move(wrapper_->payload);
-    }
-
-   private:
-    bool performed_;
-    const linked_ptr<R> wrapper_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
-  const linked_ptr<R> value_;
-
-  GTEST_DISALLOW_ASSIGN_(ReturnAction);
-};
-
-// Implements the ReturnNull() action.
-class ReturnNullAction {
- public:
-  // Allows ReturnNull() to be used in any pointer-returning function. In C++11
-  // this is enforced by returning nullptr, and in non-C++11 by asserting a
-  // pointer type on compile time.
-  template <typename Result, typename ArgumentTuple>
-  static Result Perform(const ArgumentTuple&) {
-#if GTEST_LANG_CXX11
-    return nullptr;
-#else
-    GTEST_COMPILE_ASSERT_(internal::is_pointer<Result>::value,
-                          ReturnNull_can_be_used_to_return_a_pointer_only);
-    return NULL;
-#endif  // GTEST_LANG_CXX11
-  }
-};
-
-// Implements the Return() action.
-class ReturnVoidAction {
- public:
-  // Allows Return() to be used in any void-returning function.
-  template <typename Result, typename ArgumentTuple>
-  static void Perform(const ArgumentTuple&) {
-    CompileAssertTypesEqual<void, Result>();
-  }
-};
-
-// Implements the polymorphic ReturnRef(x) action, which can be used
-// in any function that returns a reference to the type of x,
-// regardless of the argument types.
-template <typename T>
-class ReturnRefAction {
- public:
-  // Constructs a ReturnRefAction object from the reference to be returned.
-  explicit ReturnRefAction(T& ref) : ref_(ref) {}  // NOLINT
-
-  // This template type conversion operator allows ReturnRef(x) to be
-  // used in ANY function that returns a reference to x's type.
-  template <typename F>
-  operator Action<F>() const {
-    typedef typename Function<F>::Result Result;
-    // Asserts that the function return type is a reference.  This
-    // catches the user error of using ReturnRef(x) when Return(x)
-    // should be used, and generates some helpful error message.
-    GTEST_COMPILE_ASSERT_(internal::is_reference<Result>::value,
-                          use_Return_instead_of_ReturnRef_to_return_a_value);
-    return Action<F>(new Impl<F>(ref_));
-  }
-
- private:
-  // Implements the ReturnRef(x) action for a particular function type F.
-  template <typename F>
-  class Impl : public ActionInterface<F> {
-   public:
-    typedef typename Function<F>::Result Result;
-    typedef typename Function<F>::ArgumentTuple ArgumentTuple;
-
-    explicit Impl(T& ref) : ref_(ref) {}  // NOLINT
-
-    virtual Result Perform(const ArgumentTuple&) {
-      return ref_;
-    }
-
-   private:
-    T& ref_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
-  T& ref_;
-
-  GTEST_DISALLOW_ASSIGN_(ReturnRefAction);
-};
-
-// Implements the polymorphic ReturnRefOfCopy(x) action, which can be
-// used in any function that returns a reference to the type of x,
-// regardless of the argument types.
-template <typename T>
-class ReturnRefOfCopyAction {
- public:
-  // Constructs a ReturnRefOfCopyAction object from the reference to
-  // be returned.
-  explicit ReturnRefOfCopyAction(const T& value) : value_(value) {}  // NOLINT
-
-  // This template type conversion operator allows ReturnRefOfCopy(x) to be
-  // used in ANY function that returns a reference to x's type.
-  template <typename F>
-  operator Action<F>() const {
-    typedef typename Function<F>::Result Result;
-    // Asserts that the function return type is a reference.  This
-    // catches the user error of using ReturnRefOfCopy(x) when Return(x)
-    // should be used, and generates some helpful error message.
-    GTEST_COMPILE_ASSERT_(
-        internal::is_reference<Result>::value,
-        use_Return_instead_of_ReturnRefOfCopy_to_return_a_value);
-    return Action<F>(new Impl<F>(value_));
-  }
-
- private:
-  // Implements the ReturnRefOfCopy(x) action for a particular function type F.
-  template <typename F>
-  class Impl : public ActionInterface<F> {
-   public:
-    typedef typename Function<F>::Result Result;
-    typedef typename Function<F>::ArgumentTuple ArgumentTuple;
-
-    explicit Impl(const T& value) : value_(value) {}  // NOLINT
-
-    virtual Result Perform(const ArgumentTuple&) {
-      return value_;
-    }
-
-   private:
-    T value_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
-  const T value_;
-
-  GTEST_DISALLOW_ASSIGN_(ReturnRefOfCopyAction);
-};
-
-// Implements the polymorphic DoDefault() action.
-class DoDefaultAction {
- public:
-  // This template type conversion operator allows DoDefault() to be
-  // used in any function.
-  template <typename F>
-  operator Action<F>() const { return Action<F>(); }  // NOLINT
-};
-
-// Implements the Assign action to set a given pointer referent to a
-// particular value.
-template <typename T1, typename T2>
-class AssignAction {
- public:
-  AssignAction(T1* ptr, T2 value) : ptr_(ptr), value_(value) {}
-
-  template <typename Result, typename ArgumentTuple>
-  void Perform(const ArgumentTuple& /* args */) const {
-    *ptr_ = value_;
-  }
-
- private:
-  T1* const ptr_;
-  const T2 value_;
-
-  GTEST_DISALLOW_ASSIGN_(AssignAction);
-};
-
-#if !GTEST_OS_WINDOWS_MOBILE
-
-// Implements the SetErrnoAndReturn action to simulate return from
-// various system calls and libc functions.
-template <typename T>
-class SetErrnoAndReturnAction {
- public:
-  SetErrnoAndReturnAction(int errno_value, T result)
-      : errno_(errno_value),
-        result_(result) {}
-  template <typename Result, typename ArgumentTuple>
-  Result Perform(const ArgumentTuple& /* args */) const {
-    errno = errno_;
-    return result_;
-  }
-
- private:
-  const int errno_;
-  const T result_;
-
-  GTEST_DISALLOW_ASSIGN_(SetErrnoAndReturnAction);
-};
-
-#endif  // !GTEST_OS_WINDOWS_MOBILE
-
-// Implements the SetArgumentPointee<N>(x) action for any function
-// whose N-th argument (0-based) is a pointer to x's type.  The
-// template parameter kIsProto is true iff type A is ProtocolMessage,
-// proto2::Message, or a sub-class of those.
-template <size_t N, typename A, bool kIsProto>
-class SetArgumentPointeeAction {
- public:
-  // Constructs an action that sets the variable pointed to by the
-  // N-th function argument to 'value'.
-  explicit SetArgumentPointeeAction(const A& value) : value_(value) {}
-
-  template <typename Result, typename ArgumentTuple>
-  void Perform(const ArgumentTuple& args) const {
-    CompileAssertTypesEqual<void, Result>();
-    *::testing::get<N>(args) = value_;
-  }
-
- private:
-  const A value_;
-
-  GTEST_DISALLOW_ASSIGN_(SetArgumentPointeeAction);
-};
-
-template <size_t N, typename Proto>
-class SetArgumentPointeeAction<N, Proto, true> {
- public:
-  // Constructs an action that sets the variable pointed to by the
-  // N-th function argument to 'proto'.  Both ProtocolMessage and
-  // proto2::Message have the CopyFrom() method, so the same
-  // implementation works for both.
-  explicit SetArgumentPointeeAction(const Proto& proto) : proto_(new Proto) {
-    proto_->CopyFrom(proto);
-  }
-
-  template <typename Result, typename ArgumentTuple>
-  void Perform(const ArgumentTuple& args) const {
-    CompileAssertTypesEqual<void, Result>();
-    ::testing::get<N>(args)->CopyFrom(*proto_);
-  }
-
- private:
-  const internal::linked_ptr<Proto> proto_;
-
-  GTEST_DISALLOW_ASSIGN_(SetArgumentPointeeAction);
-};
-
-// Implements the InvokeWithoutArgs(f) action.  The template argument
-// FunctionImpl is the implementation type of f, which can be either a
-// function pointer or a functor.  InvokeWithoutArgs(f) can be used as an
-// Action<F> as long as f's type is compatible with F (i.e. f can be
-// assigned to a tr1::function<F>).
-template <typename FunctionImpl>
-class InvokeWithoutArgsAction {
- public:
-  // The c'tor makes a copy of function_impl (either a function
-  // pointer or a functor).
-  explicit InvokeWithoutArgsAction(FunctionImpl function_impl)
-      : function_impl_(function_impl) {}
-
-  // Allows InvokeWithoutArgs(f) to be used as any action whose type is
-  // compatible with f.
-  template <typename Result, typename ArgumentTuple>
-  Result Perform(const ArgumentTuple&) { return function_impl_(); }
-
- private:
-  FunctionImpl function_impl_;
-
-  GTEST_DISALLOW_ASSIGN_(InvokeWithoutArgsAction);
-};
-
-// Implements the InvokeWithoutArgs(object_ptr, &Class::Method) action.
-template <class Class, typename MethodPtr>
-class InvokeMethodWithoutArgsAction {
- public:
-  InvokeMethodWithoutArgsAction(Class* obj_ptr, MethodPtr method_ptr)
-      : obj_ptr_(obj_ptr), method_ptr_(method_ptr) {}
-
-  template <typename Result, typename ArgumentTuple>
-  Result Perform(const ArgumentTuple&) const {
-    return (obj_ptr_->*method_ptr_)();
-  }
-
- private:
-  Class* const obj_ptr_;
-  const MethodPtr method_ptr_;
-
-  GTEST_DISALLOW_ASSIGN_(InvokeMethodWithoutArgsAction);
-};
-
-// Implements the InvokeWithoutArgs(callback) action.
-template <typename CallbackType>
-class InvokeCallbackWithoutArgsAction {
- public:
-  // The c'tor takes ownership of the callback.
-  explicit InvokeCallbackWithoutArgsAction(CallbackType* callback)
-      : callback_(callback) {
-    callback->CheckIsRepeatable();  // Makes sure the callback is permanent.
-  }
-
-  // This type conversion operator template allows Invoke(callback) to
-  // be used wherever the callback's return type can be implicitly
-  // converted to that of the mock function.
-  template <typename Result, typename ArgumentTuple>
-  Result Perform(const ArgumentTuple&) const { return callback_->Run(); }
-
- private:
-  const internal::linked_ptr<CallbackType> callback_;
-
-  GTEST_DISALLOW_ASSIGN_(InvokeCallbackWithoutArgsAction);
-};
-
-// Implements the IgnoreResult(action) action.
-template <typename A>
-class IgnoreResultAction {
- public:
-  explicit IgnoreResultAction(const A& action) : action_(action) {}
-
-  template <typename F>
-  operator Action<F>() const {
-    // Assert statement belongs here because this is the best place to verify
-    // conditions on F. It produces the clearest error messages
-    // in most compilers.
-    // Impl really belongs in this scope as a local class but can't
-    // because MSVC produces duplicate symbols in different translation units
-    // in this case. Until MS fixes that bug we put Impl into the class scope
-    // and put the typedef both here (for use in assert statement) and
-    // in the Impl class. But both definitions must be the same.
-    typedef typename internal::Function<F>::Result Result;
-
-    // Asserts at compile time that F returns void.
-    CompileAssertTypesEqual<void, Result>();
-
-    return Action<F>(new Impl<F>(action_));
-  }
-
- private:
-  template <typename F>
-  class Impl : public ActionInterface<F> {
-   public:
-    typedef typename internal::Function<F>::Result Result;
-    typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-    explicit Impl(const A& action) : action_(action) {}
-
-    virtual void Perform(const ArgumentTuple& args) {
-      // Performs the action and ignores its result.
-      action_.Perform(args);
-    }
-
-   private:
-    // Type OriginalFunction is the same as F except that its return
-    // type is IgnoredValue.
-    typedef typename internal::Function<F>::MakeResultIgnoredValue
-        OriginalFunction;
-
-    const Action<OriginalFunction> action_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
-  const A action_;
-
-  GTEST_DISALLOW_ASSIGN_(IgnoreResultAction);
-};
-
-// A ReferenceWrapper<T> object represents a reference to type T,
-// which can be either const or not.  It can be explicitly converted
-// from, and implicitly converted to, a T&.  Unlike a reference,
-// ReferenceWrapper<T> can be copied and can survive template type
-// inference.  This is used to support by-reference arguments in the
-// InvokeArgument<N>(...) action.  The idea was from "reference
-// wrappers" in tr1, which we don't have in our source tree yet.
-template <typename T>
-class ReferenceWrapper {
- public:
-  // Constructs a ReferenceWrapper<T> object from a T&.
-  explicit ReferenceWrapper(T& l_value) : pointer_(&l_value) {}  // NOLINT
-
-  // Allows a ReferenceWrapper<T> object to be implicitly converted to
-  // a T&.
-  operator T&() const { return *pointer_; }
- private:
-  T* pointer_;
-};
-
-// Allows the expression ByRef(x) to be printed as a reference to x.
-template <typename T>
-void PrintTo(const ReferenceWrapper<T>& ref, ::std::ostream* os) {
-  T& value = ref;
-  UniversalPrinter<T&>::Print(value, os);
-}
-
-// Does two actions sequentially.  Used for implementing the DoAll(a1,
-// a2, ...) action.
-template <typename Action1, typename Action2>
-class DoBothAction {
- public:
-  DoBothAction(Action1 action1, Action2 action2)
-      : action1_(action1), action2_(action2) {}
-
-  // This template type conversion operator allows DoAll(a1, ..., a_n)
-  // to be used in ANY function of compatible type.
-  template <typename F>
-  operator Action<F>() const {
-    return Action<F>(new Impl<F>(action1_, action2_));
-  }
-
- private:
-  // Implements the DoAll(...) action for a particular function type F.
-  template <typename F>
-  class Impl : public ActionInterface<F> {
-   public:
-    typedef typename Function<F>::Result Result;
-    typedef typename Function<F>::ArgumentTuple ArgumentTuple;
-    typedef typename Function<F>::MakeResultVoid VoidResult;
-
-    Impl(const Action<VoidResult>& action1, const Action<F>& action2)
-        : action1_(action1), action2_(action2) {}
-
-    virtual Result Perform(const ArgumentTuple& args) {
-      action1_.Perform(args);
-      return action2_.Perform(args);
-    }
-
-   private:
-    const Action<VoidResult> action1_;
-    const Action<F> action2_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
-  Action1 action1_;
-  Action2 action2_;
-
-  GTEST_DISALLOW_ASSIGN_(DoBothAction);
-};
-
-}  // namespace internal
-
-// An Unused object can be implicitly constructed from ANY value.
-// This is handy when defining actions that ignore some or all of the
-// mock function arguments.  For example, given
-//
-//   MOCK_METHOD3(Foo, double(const string& label, double x, double y));
-//   MOCK_METHOD3(Bar, double(int index, double x, double y));
-//
-// instead of
-//
-//   double DistanceToOriginWithLabel(const string& label, double x, double y) {
-//     return sqrt(x*x + y*y);
-//   }
-//   double DistanceToOriginWithIndex(int index, double x, double y) {
-//     return sqrt(x*x + y*y);
-//   }
-//   ...
-//   EXPECT_CALL(mock, Foo("abc", _, _))
-//       .WillOnce(Invoke(DistanceToOriginWithLabel));
-//   EXPECT_CALL(mock, Bar(5, _, _))
-//       .WillOnce(Invoke(DistanceToOriginWithIndex));
-//
-// you could write
-//
-//   // We can declare any uninteresting argument as Unused.
-//   double DistanceToOrigin(Unused, double x, double y) {
-//     return sqrt(x*x + y*y);
-//   }
-//   ...
-//   EXPECT_CALL(mock, Foo("abc", _, _)).WillOnce(Invoke(DistanceToOrigin));
-//   EXPECT_CALL(mock, Bar(5, _, _)).WillOnce(Invoke(DistanceToOrigin));
-typedef internal::IgnoredValue Unused;
-
-// This constructor allows us to turn an Action<From> object into an
-// Action<To>, as long as To's arguments can be implicitly converted
-// to From's and From's return type cann be implicitly converted to
-// To's.
-template <typename To>
-template <typename From>
-Action<To>::Action(const Action<From>& from)
-    :
-#if GTEST_LANG_CXX11
-      fun_(from.fun_),
-#endif
-      impl_(from.impl_ == NULL ? NULL
-                               : new internal::ActionAdaptor<To, From>(from)) {
-}
-
-// Creates an action that returns 'value'.  'value' is passed by value
-// instead of const reference - otherwise Return("string literal")
-// will trigger a compiler error about using array as initializer.
-template <typename R>
-internal::ReturnAction<R> Return(R value) {
-  return internal::ReturnAction<R>(internal::move(value));
-}
-
-// Creates an action that returns NULL.
-inline PolymorphicAction<internal::ReturnNullAction> ReturnNull() {
-  return MakePolymorphicAction(internal::ReturnNullAction());
-}
-
-// Creates an action that returns from a void function.
-inline PolymorphicAction<internal::ReturnVoidAction> Return() {
-  return MakePolymorphicAction(internal::ReturnVoidAction());
-}
-
-// Creates an action that returns the reference to a variable.
-template <typename R>
-inline internal::ReturnRefAction<R> ReturnRef(R& x) {  // NOLINT
-  return internal::ReturnRefAction<R>(x);
-}
-
-// Creates an action that returns the reference to a copy of the
-// argument.  The copy is created when the action is constructed and
-// lives as long as the action.
-template <typename R>
-inline internal::ReturnRefOfCopyAction<R> ReturnRefOfCopy(const R& x) {
-  return internal::ReturnRefOfCopyAction<R>(x);
-}
-
-// Modifies the parent action (a Return() action) to perform a move of the
-// argument instead of a copy.
-// Return(ByMove()) actions can only be executed once and will assert this
-// invariant.
-template <typename R>
-internal::ByMoveWrapper<R> ByMove(R x) {
-  return internal::ByMoveWrapper<R>(internal::move(x));
-}
-
-// Creates an action that does the default action for the give mock function.
-inline internal::DoDefaultAction DoDefault() {
-  return internal::DoDefaultAction();
-}
-
-// Creates an action that sets the variable pointed by the N-th
-// (0-based) function argument to 'value'.
-template <size_t N, typename T>
-PolymorphicAction<
-  internal::SetArgumentPointeeAction<
-    N, T, internal::IsAProtocolMessage<T>::value> >
-SetArgPointee(const T& x) {
-  return MakePolymorphicAction(internal::SetArgumentPointeeAction<
-      N, T, internal::IsAProtocolMessage<T>::value>(x));
-}
-
-#if !((GTEST_GCC_VER_ && GTEST_GCC_VER_ < 40000) || GTEST_OS_SYMBIAN)
-// This overload allows SetArgPointee() to accept a string literal.
-// GCC prior to the version 4.0 and Symbian C++ compiler cannot distinguish
-// this overload from the templated version and emit a compile error.
-template <size_t N>
-PolymorphicAction<
-  internal::SetArgumentPointeeAction<N, const char*, false> >
-SetArgPointee(const char* p) {
-  return MakePolymorphicAction(internal::SetArgumentPointeeAction<
-      N, const char*, false>(p));
-}
-
-template <size_t N>
-PolymorphicAction<
-  internal::SetArgumentPointeeAction<N, const wchar_t*, false> >
-SetArgPointee(const wchar_t* p) {
-  return MakePolymorphicAction(internal::SetArgumentPointeeAction<
-      N, const wchar_t*, false>(p));
-}
-#endif
-
-// The following version is DEPRECATED.
-template <size_t N, typename T>
-PolymorphicAction<
-  internal::SetArgumentPointeeAction<
-    N, T, internal::IsAProtocolMessage<T>::value> >
-SetArgumentPointee(const T& x) {
-  return MakePolymorphicAction(internal::SetArgumentPointeeAction<
-      N, T, internal::IsAProtocolMessage<T>::value>(x));
-}
-
-// Creates an action that sets a pointer referent to a given value.
-template <typename T1, typename T2>
-PolymorphicAction<internal::AssignAction<T1, T2> > Assign(T1* ptr, T2 val) {
-  return MakePolymorphicAction(internal::AssignAction<T1, T2>(ptr, val));
-}
-
-#if !GTEST_OS_WINDOWS_MOBILE
-
-// Creates an action that sets errno and returns the appropriate error.
-template <typename T>
-PolymorphicAction<internal::SetErrnoAndReturnAction<T> >
-SetErrnoAndReturn(int errval, T result) {
-  return MakePolymorphicAction(
-      internal::SetErrnoAndReturnAction<T>(errval, result));
-}
-
-#endif  // !GTEST_OS_WINDOWS_MOBILE
-
-// Various overloads for InvokeWithoutArgs().
-
-// Creates an action that invokes 'function_impl' with no argument.
-template <typename FunctionImpl>
-PolymorphicAction<internal::InvokeWithoutArgsAction<FunctionImpl> >
-InvokeWithoutArgs(FunctionImpl function_impl) {
-  return MakePolymorphicAction(
-      internal::InvokeWithoutArgsAction<FunctionImpl>(function_impl));
-}
-
-// Creates an action that invokes the given method on the given object
-// with no argument.
-template <class Class, typename MethodPtr>
-PolymorphicAction<internal::InvokeMethodWithoutArgsAction<Class, MethodPtr> >
-InvokeWithoutArgs(Class* obj_ptr, MethodPtr method_ptr) {
-  return MakePolymorphicAction(
-      internal::InvokeMethodWithoutArgsAction<Class, MethodPtr>(
-          obj_ptr, method_ptr));
-}
-
-// Creates an action that performs an_action and throws away its
-// result.  In other words, it changes the return type of an_action to
-// void.  an_action MUST NOT return void, or the code won't compile.
-template <typename A>
-inline internal::IgnoreResultAction<A> IgnoreResult(const A& an_action) {
-  return internal::IgnoreResultAction<A>(an_action);
-}
-
-// Creates a reference wrapper for the given L-value.  If necessary,
-// you can explicitly specify the type of the reference.  For example,
-// suppose 'derived' is an object of type Derived, ByRef(derived)
-// would wrap a Derived&.  If you want to wrap a const Base& instead,
-// where Base is a base class of Derived, just write:
-//
-//   ByRef<const Base>(derived)
-template <typename T>
-inline internal::ReferenceWrapper<T> ByRef(T& l_value) {  // NOLINT
-  return internal::ReferenceWrapper<T>(l_value);
-}
-
-}  // namespace testing
-
-#endif  // GMOCK_INCLUDE_GMOCK_GMOCK_ACTIONS_H_
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements some commonly used cardinalities.  More
-// cardinalities can be defined by the user implementing the
-// CardinalityInterface interface if necessary.
-
-#ifndef GMOCK_INCLUDE_GMOCK_GMOCK_CARDINALITIES_H_
-#define GMOCK_INCLUDE_GMOCK_GMOCK_CARDINALITIES_H_
-
-#include <limits.h>
-#include <ostream>  // NOLINT
-
-namespace testing {
-
-// To implement a cardinality Foo, define:
-//   1. a class FooCardinality that implements the
-//      CardinalityInterface interface, and
-//   2. a factory function that creates a Cardinality object from a
-//      const FooCardinality*.
-//
-// The two-level delegation design follows that of Matcher, providing
-// consistency for extension developers.  It also eases ownership
-// management as Cardinality objects can now be copied like plain values.
-
-// The implementation of a cardinality.
-class CardinalityInterface {
- public:
-  virtual ~CardinalityInterface() {}
-
-  // Conservative estimate on the lower/upper bound of the number of
-  // calls allowed.
-  virtual int ConservativeLowerBound() const { return 0; }
-  virtual int ConservativeUpperBound() const { return INT_MAX; }
-
-  // Returns true iff call_count calls will satisfy this cardinality.
-  virtual bool IsSatisfiedByCallCount(int call_count) const = 0;
-
-  // Returns true iff call_count calls will saturate this cardinality.
-  virtual bool IsSaturatedByCallCount(int call_count) const = 0;
-
-  // Describes self to an ostream.
-  virtual void DescribeTo(::std::ostream* os) const = 0;
-};
-
-// A Cardinality is a copyable and IMMUTABLE (except by assignment)
-// object that specifies how many times a mock function is expected to
-// be called.  The implementation of Cardinality is just a linked_ptr
-// to const CardinalityInterface, so copying is fairly cheap.
-// Don't inherit from Cardinality!
-class GTEST_API_ Cardinality {
- public:
-  // Constructs a null cardinality.  Needed for storing Cardinality
-  // objects in STL containers.
-  Cardinality() {}
-
-  // Constructs a Cardinality from its implementation.
-  explicit Cardinality(const CardinalityInterface* impl) : impl_(impl) {}
-
-  // Conservative estimate on the lower/upper bound of the number of
-  // calls allowed.
-  int ConservativeLowerBound() const { return impl_->ConservativeLowerBound(); }
-  int ConservativeUpperBound() const { return impl_->ConservativeUpperBound(); }
-
-  // Returns true iff call_count calls will satisfy this cardinality.
-  bool IsSatisfiedByCallCount(int call_count) const {
-    return impl_->IsSatisfiedByCallCount(call_count);
-  }
-
-  // Returns true iff call_count calls will saturate this cardinality.
-  bool IsSaturatedByCallCount(int call_count) const {
-    return impl_->IsSaturatedByCallCount(call_count);
-  }
-
-  // Returns true iff call_count calls will over-saturate this
-  // cardinality, i.e. exceed the maximum number of allowed calls.
-  bool IsOverSaturatedByCallCount(int call_count) const {
-    return impl_->IsSaturatedByCallCount(call_count) &&
-        !impl_->IsSatisfiedByCallCount(call_count);
-  }
-
-  // Describes self to an ostream
-  void DescribeTo(::std::ostream* os) const { impl_->DescribeTo(os); }
-
-  // Describes the given actual call count to an ostream.
-  static void DescribeActualCallCountTo(int actual_call_count,
-                                        ::std::ostream* os);
-
- private:
-  internal::linked_ptr<const CardinalityInterface> impl_;
-};
-
-// Creates a cardinality that allows at least n calls.
-GTEST_API_ Cardinality AtLeast(int n);
-
-// Creates a cardinality that allows at most n calls.
-GTEST_API_ Cardinality AtMost(int n);
-
-// Creates a cardinality that allows any number of calls.
-GTEST_API_ Cardinality AnyNumber();
-
-// Creates a cardinality that allows between min and max calls.
-GTEST_API_ Cardinality Between(int min, int max);
-
-// Creates a cardinality that allows exactly n calls.
-GTEST_API_ Cardinality Exactly(int n);
-
-// Creates a cardinality from its implementation.
-inline Cardinality MakeCardinality(const CardinalityInterface* c) {
-  return Cardinality(c);
-}
-
-}  // namespace testing
-
-#endif  // GMOCK_INCLUDE_GMOCK_GMOCK_CARDINALITIES_H_
-// This file was GENERATED by command:
-//     pump.py gmock-generated-actions.h.pump
-// DO NOT EDIT BY HAND!!!
-
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements some commonly used variadic actions.
-
-#ifndef GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_ACTIONS_H_
-#define GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_ACTIONS_H_
-
-
-namespace testing {
-namespace internal {
-
-// InvokeHelper<F> knows how to unpack an N-tuple and invoke an N-ary
-// function, method, or callback with the unpacked values, where F is
-// a function type that takes N arguments.
-template <typename Result, typename ArgumentTuple>
-class InvokeHelper;
-
-template <typename R>
-class InvokeHelper<R, ::testing::tuple<> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<>&) {
-           return function();
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<>&) {
-           return (obj_ptr->*method_ptr)();
-  }
-
-  template <typename CallbackType>
-  static R InvokeCallback(CallbackType* callback,
-                          const ::testing::tuple<>&) {
-           return callback->Run();
-  }
-};
-
-template <typename R, typename A1>
-class InvokeHelper<R, ::testing::tuple<A1> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<A1>& args) {
-           return function(get<0>(args));
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<A1>& args) {
-           return (obj_ptr->*method_ptr)(get<0>(args));
-  }
-
-  template <typename CallbackType>
-  static R InvokeCallback(CallbackType* callback,
-                          const ::testing::tuple<A1>& args) {
-           return callback->Run(get<0>(args));
-  }
-};
-
-template <typename R, typename A1, typename A2>
-class InvokeHelper<R, ::testing::tuple<A1, A2> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<A1, A2>& args) {
-           return function(get<0>(args), get<1>(args));
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<A1, A2>& args) {
-           return (obj_ptr->*method_ptr)(get<0>(args), get<1>(args));
-  }
-
-  template <typename CallbackType>
-  static R InvokeCallback(CallbackType* callback,
-                          const ::testing::tuple<A1, A2>& args) {
-           return callback->Run(get<0>(args), get<1>(args));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3>
-class InvokeHelper<R, ::testing::tuple<A1, A2, A3> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<A1, A2, A3>& args) {
-           return function(get<0>(args), get<1>(args), get<2>(args));
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<A1, A2, A3>& args) {
-           return (obj_ptr->*method_ptr)(get<0>(args), get<1>(args),
-               get<2>(args));
-  }
-
-  template <typename CallbackType>
-  static R InvokeCallback(CallbackType* callback,
-                          const ::testing::tuple<A1, A2, A3>& args) {
-           return callback->Run(get<0>(args), get<1>(args), get<2>(args));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4>
-class InvokeHelper<R, ::testing::tuple<A1, A2, A3, A4> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<A1, A2, A3,
-      A4>& args) {
-           return function(get<0>(args), get<1>(args), get<2>(args),
-               get<3>(args));
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<A1, A2, A3, A4>& args) {
-           return (obj_ptr->*method_ptr)(get<0>(args), get<1>(args),
-               get<2>(args), get<3>(args));
-  }
-
-  template <typename CallbackType>
-  static R InvokeCallback(CallbackType* callback,
-                          const ::testing::tuple<A1, A2, A3, A4>& args) {
-           return callback->Run(get<0>(args), get<1>(args), get<2>(args),
-               get<3>(args));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5>
-class InvokeHelper<R, ::testing::tuple<A1, A2, A3, A4, A5> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<A1, A2, A3, A4,
-      A5>& args) {
-           return function(get<0>(args), get<1>(args), get<2>(args),
-               get<3>(args), get<4>(args));
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<A1, A2, A3, A4, A5>& args) {
-           return (obj_ptr->*method_ptr)(get<0>(args), get<1>(args),
-               get<2>(args), get<3>(args), get<4>(args));
-  }
-
-  template <typename CallbackType>
-  static R InvokeCallback(CallbackType* callback,
-                          const ::testing::tuple<A1, A2, A3, A4, A5>& args) {
-           return callback->Run(get<0>(args), get<1>(args), get<2>(args),
-               get<3>(args), get<4>(args));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6>
-class InvokeHelper<R, ::testing::tuple<A1, A2, A3, A4, A5, A6> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<A1, A2, A3, A4, A5,
-      A6>& args) {
-           return function(get<0>(args), get<1>(args), get<2>(args),
-               get<3>(args), get<4>(args), get<5>(args));
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<A1, A2, A3, A4, A5, A6>& args) {
-           return (obj_ptr->*method_ptr)(get<0>(args), get<1>(args),
-               get<2>(args), get<3>(args), get<4>(args), get<5>(args));
-  }
-
-  // There is no InvokeCallback() for 6-tuples, as google3 callbacks
-  // support 5 arguments at most.
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7>
-class InvokeHelper<R, ::testing::tuple<A1, A2, A3, A4, A5, A6, A7> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<A1, A2, A3, A4, A5,
-      A6, A7>& args) {
-           return function(get<0>(args), get<1>(args), get<2>(args),
-               get<3>(args), get<4>(args), get<5>(args), get<6>(args));
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<A1, A2, A3, A4, A5, A6,
-                            A7>& args) {
-           return (obj_ptr->*method_ptr)(get<0>(args), get<1>(args),
-               get<2>(args), get<3>(args), get<4>(args), get<5>(args),
-               get<6>(args));
-  }
-
-  // There is no InvokeCallback() for 7-tuples, as google3 callbacks
-  // support 5 arguments at most.
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7, typename A8>
-class InvokeHelper<R, ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<A1, A2, A3, A4, A5,
-      A6, A7, A8>& args) {
-           return function(get<0>(args), get<1>(args), get<2>(args),
-               get<3>(args), get<4>(args), get<5>(args), get<6>(args),
-               get<7>(args));
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<A1, A2, A3, A4, A5, A6, A7,
-                            A8>& args) {
-           return (obj_ptr->*method_ptr)(get<0>(args), get<1>(args),
-               get<2>(args), get<3>(args), get<4>(args), get<5>(args),
-               get<6>(args), get<7>(args));
-  }
-
-  // There is no InvokeCallback() for 8-tuples, as google3 callbacks
-  // support 5 arguments at most.
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7, typename A8, typename A9>
-class InvokeHelper<R, ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8, A9> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<A1, A2, A3, A4, A5,
-      A6, A7, A8, A9>& args) {
-           return function(get<0>(args), get<1>(args), get<2>(args),
-               get<3>(args), get<4>(args), get<5>(args), get<6>(args),
-               get<7>(args), get<8>(args));
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8,
-                            A9>& args) {
-           return (obj_ptr->*method_ptr)(get<0>(args), get<1>(args),
-               get<2>(args), get<3>(args), get<4>(args), get<5>(args),
-               get<6>(args), get<7>(args), get<8>(args));
-  }
-
-  // There is no InvokeCallback() for 9-tuples, as google3 callbacks
-  // support 5 arguments at most.
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7, typename A8, typename A9,
-    typename A10>
-class InvokeHelper<R, ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8, A9,
-    A10> > {
- public:
-  template <typename Function>
-  static R Invoke(Function function, const ::testing::tuple<A1, A2, A3, A4, A5,
-      A6, A7, A8, A9, A10>& args) {
-           return function(get<0>(args), get<1>(args), get<2>(args),
-               get<3>(args), get<4>(args), get<5>(args), get<6>(args),
-               get<7>(args), get<8>(args), get<9>(args));
-  }
-
-  template <class Class, typename MethodPtr>
-  static R InvokeMethod(Class* obj_ptr,
-                        MethodPtr method_ptr,
-                        const ::testing::tuple<A1, A2, A3, A4, A5, A6, A7, A8,
-                            A9, A10>& args) {
-           return (obj_ptr->*method_ptr)(get<0>(args), get<1>(args),
-               get<2>(args), get<3>(args), get<4>(args), get<5>(args),
-               get<6>(args), get<7>(args), get<8>(args), get<9>(args));
-  }
-
-  // There is no InvokeCallback() for 10-tuples, as google3 callbacks
-  // support 5 arguments at most.
-};
-
-// Implements the Invoke(callback) action.
-template <typename CallbackType>
-class InvokeCallbackAction {
- public:
-  // The c'tor takes ownership of the callback.
-  explicit InvokeCallbackAction(CallbackType* callback)
-      : callback_(callback) {
-    callback->CheckIsRepeatable();  // Makes sure the callback is permanent.
-  }
-
-  // This type conversion operator template allows Invoke(callback) to
-  // be used wherever the callback's type is compatible with that of
-  // the mock function, i.e. if the mock function's arguments can be
-  // implicitly converted to the callback's arguments and the
-  // callback's result can be implicitly converted to the mock
-  // function's result.
-  template <typename Result, typename ArgumentTuple>
-  Result Perform(const ArgumentTuple& args) const {
-    return InvokeHelper<Result, ArgumentTuple>::InvokeCallback(
-        callback_.get(), args);
-  }
- private:
-  const linked_ptr<CallbackType> callback_;
-};
-
-// An INTERNAL macro for extracting the type of a tuple field.  It's
-// subject to change without notice - DO NOT USE IN USER CODE!
-#define GMOCK_FIELD_(Tuple, N) \
-    typename ::testing::tuple_element<N, Tuple>::type
-
-// SelectArgs<Result, ArgumentTuple, k1, k2, ..., k_n>::type is the
-// type of an n-ary function whose i-th (1-based) argument type is the
-// k{i}-th (0-based) field of ArgumentTuple, which must be a tuple
-// type, and whose return type is Result.  For example,
-//   SelectArgs<int, ::testing::tuple<bool, char, double, long>, 0, 3>::type
-// is int(bool, long).
-//
-// SelectArgs<Result, ArgumentTuple, k1, k2, ..., k_n>::Select(args)
-// returns the selected fields (k1, k2, ..., k_n) of args as a tuple.
-// For example,
-//   SelectArgs<int, tuple<bool, char, double>, 2, 0>::Select(
-//       ::testing::make_tuple(true, 'a', 2.5))
-// returns tuple (2.5, true).
-//
-// The numbers in list k1, k2, ..., k_n must be >= 0, where n can be
-// in the range [0, 10].  Duplicates are allowed and they don't have
-// to be in an ascending or descending order.
-
-template <typename Result, typename ArgumentTuple, int k1, int k2, int k3,
-    int k4, int k5, int k6, int k7, int k8, int k9, int k10>
-class SelectArgs {
- public:
-  typedef Result type(GMOCK_FIELD_(ArgumentTuple, k1),
-      GMOCK_FIELD_(ArgumentTuple, k2), GMOCK_FIELD_(ArgumentTuple, k3),
-      GMOCK_FIELD_(ArgumentTuple, k4), GMOCK_FIELD_(ArgumentTuple, k5),
-      GMOCK_FIELD_(ArgumentTuple, k6), GMOCK_FIELD_(ArgumentTuple, k7),
-      GMOCK_FIELD_(ArgumentTuple, k8), GMOCK_FIELD_(ArgumentTuple, k9),
-      GMOCK_FIELD_(ArgumentTuple, k10));
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& args) {
-    return SelectedArgs(get<k1>(args), get<k2>(args), get<k3>(args),
-        get<k4>(args), get<k5>(args), get<k6>(args), get<k7>(args),
-        get<k8>(args), get<k9>(args), get<k10>(args));
-  }
-};
-
-template <typename Result, typename ArgumentTuple>
-class SelectArgs<Result, ArgumentTuple,
-                 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1> {
- public:
-  typedef Result type();
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& /* args */) {
-    return SelectedArgs();
-  }
-};
-
-template <typename Result, typename ArgumentTuple, int k1>
-class SelectArgs<Result, ArgumentTuple,
-                 k1, -1, -1, -1, -1, -1, -1, -1, -1, -1> {
- public:
-  typedef Result type(GMOCK_FIELD_(ArgumentTuple, k1));
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& args) {
-    return SelectedArgs(get<k1>(args));
-  }
-};
-
-template <typename Result, typename ArgumentTuple, int k1, int k2>
-class SelectArgs<Result, ArgumentTuple,
-                 k1, k2, -1, -1, -1, -1, -1, -1, -1, -1> {
- public:
-  typedef Result type(GMOCK_FIELD_(ArgumentTuple, k1),
-      GMOCK_FIELD_(ArgumentTuple, k2));
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& args) {
-    return SelectedArgs(get<k1>(args), get<k2>(args));
-  }
-};
-
-template <typename Result, typename ArgumentTuple, int k1, int k2, int k3>
-class SelectArgs<Result, ArgumentTuple,
-                 k1, k2, k3, -1, -1, -1, -1, -1, -1, -1> {
- public:
-  typedef Result type(GMOCK_FIELD_(ArgumentTuple, k1),
-      GMOCK_FIELD_(ArgumentTuple, k2), GMOCK_FIELD_(ArgumentTuple, k3));
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& args) {
-    return SelectedArgs(get<k1>(args), get<k2>(args), get<k3>(args));
-  }
-};
-
-template <typename Result, typename ArgumentTuple, int k1, int k2, int k3,
-    int k4>
-class SelectArgs<Result, ArgumentTuple,
-                 k1, k2, k3, k4, -1, -1, -1, -1, -1, -1> {
- public:
-  typedef Result type(GMOCK_FIELD_(ArgumentTuple, k1),
-      GMOCK_FIELD_(ArgumentTuple, k2), GMOCK_FIELD_(ArgumentTuple, k3),
-      GMOCK_FIELD_(ArgumentTuple, k4));
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& args) {
-    return SelectedArgs(get<k1>(args), get<k2>(args), get<k3>(args),
-        get<k4>(args));
-  }
-};
-
-template <typename Result, typename ArgumentTuple, int k1, int k2, int k3,
-    int k4, int k5>
-class SelectArgs<Result, ArgumentTuple,
-                 k1, k2, k3, k4, k5, -1, -1, -1, -1, -1> {
- public:
-  typedef Result type(GMOCK_FIELD_(ArgumentTuple, k1),
-      GMOCK_FIELD_(ArgumentTuple, k2), GMOCK_FIELD_(ArgumentTuple, k3),
-      GMOCK_FIELD_(ArgumentTuple, k4), GMOCK_FIELD_(ArgumentTuple, k5));
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& args) {
-    return SelectedArgs(get<k1>(args), get<k2>(args), get<k3>(args),
-        get<k4>(args), get<k5>(args));
-  }
-};
-
-template <typename Result, typename ArgumentTuple, int k1, int k2, int k3,
-    int k4, int k5, int k6>
-class SelectArgs<Result, ArgumentTuple,
-                 k1, k2, k3, k4, k5, k6, -1, -1, -1, -1> {
- public:
-  typedef Result type(GMOCK_FIELD_(ArgumentTuple, k1),
-      GMOCK_FIELD_(ArgumentTuple, k2), GMOCK_FIELD_(ArgumentTuple, k3),
-      GMOCK_FIELD_(ArgumentTuple, k4), GMOCK_FIELD_(ArgumentTuple, k5),
-      GMOCK_FIELD_(ArgumentTuple, k6));
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& args) {
-    return SelectedArgs(get<k1>(args), get<k2>(args), get<k3>(args),
-        get<k4>(args), get<k5>(args), get<k6>(args));
-  }
-};
-
-template <typename Result, typename ArgumentTuple, int k1, int k2, int k3,
-    int k4, int k5, int k6, int k7>
-class SelectArgs<Result, ArgumentTuple,
-                 k1, k2, k3, k4, k5, k6, k7, -1, -1, -1> {
- public:
-  typedef Result type(GMOCK_FIELD_(ArgumentTuple, k1),
-      GMOCK_FIELD_(ArgumentTuple, k2), GMOCK_FIELD_(ArgumentTuple, k3),
-      GMOCK_FIELD_(ArgumentTuple, k4), GMOCK_FIELD_(ArgumentTuple, k5),
-      GMOCK_FIELD_(ArgumentTuple, k6), GMOCK_FIELD_(ArgumentTuple, k7));
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& args) {
-    return SelectedArgs(get<k1>(args), get<k2>(args), get<k3>(args),
-        get<k4>(args), get<k5>(args), get<k6>(args), get<k7>(args));
-  }
-};
-
-template <typename Result, typename ArgumentTuple, int k1, int k2, int k3,
-    int k4, int k5, int k6, int k7, int k8>
-class SelectArgs<Result, ArgumentTuple,
-                 k1, k2, k3, k4, k5, k6, k7, k8, -1, -1> {
- public:
-  typedef Result type(GMOCK_FIELD_(ArgumentTuple, k1),
-      GMOCK_FIELD_(ArgumentTuple, k2), GMOCK_FIELD_(ArgumentTuple, k3),
-      GMOCK_FIELD_(ArgumentTuple, k4), GMOCK_FIELD_(ArgumentTuple, k5),
-      GMOCK_FIELD_(ArgumentTuple, k6), GMOCK_FIELD_(ArgumentTuple, k7),
-      GMOCK_FIELD_(ArgumentTuple, k8));
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& args) {
-    return SelectedArgs(get<k1>(args), get<k2>(args), get<k3>(args),
-        get<k4>(args), get<k5>(args), get<k6>(args), get<k7>(args),
-        get<k8>(args));
-  }
-};
-
-template <typename Result, typename ArgumentTuple, int k1, int k2, int k3,
-    int k4, int k5, int k6, int k7, int k8, int k9>
-class SelectArgs<Result, ArgumentTuple,
-                 k1, k2, k3, k4, k5, k6, k7, k8, k9, -1> {
- public:
-  typedef Result type(GMOCK_FIELD_(ArgumentTuple, k1),
-      GMOCK_FIELD_(ArgumentTuple, k2), GMOCK_FIELD_(ArgumentTuple, k3),
-      GMOCK_FIELD_(ArgumentTuple, k4), GMOCK_FIELD_(ArgumentTuple, k5),
-      GMOCK_FIELD_(ArgumentTuple, k6), GMOCK_FIELD_(ArgumentTuple, k7),
-      GMOCK_FIELD_(ArgumentTuple, k8), GMOCK_FIELD_(ArgumentTuple, k9));
-  typedef typename Function<type>::ArgumentTuple SelectedArgs;
-  static SelectedArgs Select(const ArgumentTuple& args) {
-    return SelectedArgs(get<k1>(args), get<k2>(args), get<k3>(args),
-        get<k4>(args), get<k5>(args), get<k6>(args), get<k7>(args),
-        get<k8>(args), get<k9>(args));
-  }
-};
-
-#undef GMOCK_FIELD_
-
-// Implements the WithArgs action.
-template <typename InnerAction, int k1 = -1, int k2 = -1, int k3 = -1,
-    int k4 = -1, int k5 = -1, int k6 = -1, int k7 = -1, int k8 = -1,
-    int k9 = -1, int k10 = -1>
-class WithArgsAction {
- public:
-  explicit WithArgsAction(const InnerAction& action) : action_(action) {}
-
-  template <typename F>
-  operator Action<F>() const { return MakeAction(new Impl<F>(action_)); }
-
- private:
-  template <typename F>
-  class Impl : public ActionInterface<F> {
-   public:
-    typedef typename Function<F>::Result Result;
-    typedef typename Function<F>::ArgumentTuple ArgumentTuple;
-
-    explicit Impl(const InnerAction& action) : action_(action) {}
-
-    virtual Result Perform(const ArgumentTuple& args) {
-      return action_.Perform(SelectArgs<Result, ArgumentTuple, k1, k2, k3, k4,
-          k5, k6, k7, k8, k9, k10>::Select(args));
-    }
-
-   private:
-    typedef typename SelectArgs<Result, ArgumentTuple,
-        k1, k2, k3, k4, k5, k6, k7, k8, k9, k10>::type InnerFunctionType;
-
-    Action<InnerFunctionType> action_;
-  };
-
-  const InnerAction action_;
-
-  GTEST_DISALLOW_ASSIGN_(WithArgsAction);
-};
-
-// A macro from the ACTION* family (defined later in this file)
-// defines an action that can be used in a mock function.  Typically,
-// these actions only care about a subset of the arguments of the mock
-// function.  For example, if such an action only uses the second
-// argument, it can be used in any mock function that takes >= 2
-// arguments where the type of the second argument is compatible.
-//
-// Therefore, the action implementation must be prepared to take more
-// arguments than it needs.  The ExcessiveArg type is used to
-// represent those excessive arguments.  In order to keep the compiler
-// error messages tractable, we define it in the testing namespace
-// instead of testing::internal.  However, this is an INTERNAL TYPE
-// and subject to change without notice, so a user MUST NOT USE THIS
-// TYPE DIRECTLY.
-struct ExcessiveArg {};
-
-// A helper class needed for implementing the ACTION* macros.
-template <typename Result, class Impl>
-class ActionHelper {
- public:
-  static Result Perform(Impl* impl, const ::testing::tuple<>& args) {
-    return impl->template gmock_PerformImpl<>(args, ExcessiveArg(),
-        ExcessiveArg(), ExcessiveArg(), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg(), ExcessiveArg(), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg());
-  }
-
-  template <typename A0>
-  static Result Perform(Impl* impl, const ::testing::tuple<A0>& args) {
-    return impl->template gmock_PerformImpl<A0>(args, get<0>(args),
-        ExcessiveArg(), ExcessiveArg(), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg(), ExcessiveArg(), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg());
-  }
-
-  template <typename A0, typename A1>
-  static Result Perform(Impl* impl, const ::testing::tuple<A0, A1>& args) {
-    return impl->template gmock_PerformImpl<A0, A1>(args, get<0>(args),
-        get<1>(args), ExcessiveArg(), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg(), ExcessiveArg(), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg());
-  }
-
-  template <typename A0, typename A1, typename A2>
-  static Result Perform(Impl* impl, const ::testing::tuple<A0, A1, A2>& args) {
-    return impl->template gmock_PerformImpl<A0, A1, A2>(args, get<0>(args),
-        get<1>(args), get<2>(args), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg(), ExcessiveArg(), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg());
-  }
-
-  template <typename A0, typename A1, typename A2, typename A3>
-  static Result Perform(Impl* impl, const ::testing::tuple<A0, A1, A2,
-      A3>& args) {
-    return impl->template gmock_PerformImpl<A0, A1, A2, A3>(args, get<0>(args),
-        get<1>(args), get<2>(args), get<3>(args), ExcessiveArg(),
-        ExcessiveArg(), ExcessiveArg(), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg());
-  }
-
-  template <typename A0, typename A1, typename A2, typename A3, typename A4>
-  static Result Perform(Impl* impl, const ::testing::tuple<A0, A1, A2, A3,
-      A4>& args) {
-    return impl->template gmock_PerformImpl<A0, A1, A2, A3, A4>(args,
-        get<0>(args), get<1>(args), get<2>(args), get<3>(args), get<4>(args),
-        ExcessiveArg(), ExcessiveArg(), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg());
-  }
-
-  template <typename A0, typename A1, typename A2, typename A3, typename A4,
-      typename A5>
-  static Result Perform(Impl* impl, const ::testing::tuple<A0, A1, A2, A3, A4,
-      A5>& args) {
-    return impl->template gmock_PerformImpl<A0, A1, A2, A3, A4, A5>(args,
-        get<0>(args), get<1>(args), get<2>(args), get<3>(args), get<4>(args),
-        get<5>(args), ExcessiveArg(), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg());
-  }
-
-  template <typename A0, typename A1, typename A2, typename A3, typename A4,
-      typename A5, typename A6>
-  static Result Perform(Impl* impl, const ::testing::tuple<A0, A1, A2, A3, A4,
-      A5, A6>& args) {
-    return impl->template gmock_PerformImpl<A0, A1, A2, A3, A4, A5, A6>(args,
-        get<0>(args), get<1>(args), get<2>(args), get<3>(args), get<4>(args),
-        get<5>(args), get<6>(args), ExcessiveArg(), ExcessiveArg(),
-        ExcessiveArg());
-  }
-
-  template <typename A0, typename A1, typename A2, typename A3, typename A4,
-      typename A5, typename A6, typename A7>
-  static Result Perform(Impl* impl, const ::testing::tuple<A0, A1, A2, A3, A4,
-      A5, A6, A7>& args) {
-    return impl->template gmock_PerformImpl<A0, A1, A2, A3, A4, A5, A6,
-        A7>(args, get<0>(args), get<1>(args), get<2>(args), get<3>(args),
-        get<4>(args), get<5>(args), get<6>(args), get<7>(args), ExcessiveArg(),
-        ExcessiveArg());
-  }
-
-  template <typename A0, typename A1, typename A2, typename A3, typename A4,
-      typename A5, typename A6, typename A7, typename A8>
-  static Result Perform(Impl* impl, const ::testing::tuple<A0, A1, A2, A3, A4,
-      A5, A6, A7, A8>& args) {
-    return impl->template gmock_PerformImpl<A0, A1, A2, A3, A4, A5, A6, A7,
-        A8>(args, get<0>(args), get<1>(args), get<2>(args), get<3>(args),
-        get<4>(args), get<5>(args), get<6>(args), get<7>(args), get<8>(args),
-        ExcessiveArg());
-  }
-
-  template <typename A0, typename A1, typename A2, typename A3, typename A4,
-      typename A5, typename A6, typename A7, typename A8, typename A9>
-  static Result Perform(Impl* impl, const ::testing::tuple<A0, A1, A2, A3, A4,
-      A5, A6, A7, A8, A9>& args) {
-    return impl->template gmock_PerformImpl<A0, A1, A2, A3, A4, A5, A6, A7, A8,
-        A9>(args, get<0>(args), get<1>(args), get<2>(args), get<3>(args),
-        get<4>(args), get<5>(args), get<6>(args), get<7>(args), get<8>(args),
-        get<9>(args));
-  }
-};
-
-}  // namespace internal
-
-// Various overloads for Invoke().
-
-// WithArgs<N1, N2, ..., Nk>(an_action) creates an action that passes
-// the selected arguments of the mock function to an_action and
-// performs it.  It serves as an adaptor between actions with
-// different argument lists.  C++ doesn't support default arguments for
-// function templates, so we have to overload it.
-template <int k1, typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k1>
-WithArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k1>(action);
-}
-
-template <int k1, int k2, typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k1, k2>
-WithArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k1, k2>(action);
-}
-
-template <int k1, int k2, int k3, typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k1, k2, k3>
-WithArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k1, k2, k3>(action);
-}
-
-template <int k1, int k2, int k3, int k4, typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k1, k2, k3, k4>
-WithArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k1, k2, k3, k4>(action);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5>
-WithArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5>(action);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, int k6, typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5, k6>
-WithArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5, k6>(action);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, int k6, int k7,
-    typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5, k6, k7>
-WithArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5, k6,
-      k7>(action);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, int k6, int k7, int k8,
-    typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5, k6, k7, k8>
-WithArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5, k6, k7,
-      k8>(action);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, int k6, int k7, int k8,
-    int k9, typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5, k6, k7, k8, k9>
-WithArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5, k6, k7, k8,
-      k9>(action);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, int k6, int k7, int k8,
-    int k9, int k10, typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5, k6, k7, k8,
-    k9, k10>
-WithArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k1, k2, k3, k4, k5, k6, k7, k8,
-      k9, k10>(action);
-}
-
-// Creates an action that does actions a1, a2, ..., sequentially in
-// each invocation.
-template <typename Action1, typename Action2>
-inline internal::DoBothAction<Action1, Action2>
-DoAll(Action1 a1, Action2 a2) {
-  return internal::DoBothAction<Action1, Action2>(a1, a2);
-}
-
-template <typename Action1, typename Action2, typename Action3>
-inline internal::DoBothAction<Action1, internal::DoBothAction<Action2,
-    Action3> >
-DoAll(Action1 a1, Action2 a2, Action3 a3) {
-  return DoAll(a1, DoAll(a2, a3));
-}
-
-template <typename Action1, typename Action2, typename Action3,
-    typename Action4>
-inline internal::DoBothAction<Action1, internal::DoBothAction<Action2,
-    internal::DoBothAction<Action3, Action4> > >
-DoAll(Action1 a1, Action2 a2, Action3 a3, Action4 a4) {
-  return DoAll(a1, DoAll(a2, a3, a4));
-}
-
-template <typename Action1, typename Action2, typename Action3,
-    typename Action4, typename Action5>
-inline internal::DoBothAction<Action1, internal::DoBothAction<Action2,
-    internal::DoBothAction<Action3, internal::DoBothAction<Action4,
-    Action5> > > >
-DoAll(Action1 a1, Action2 a2, Action3 a3, Action4 a4, Action5 a5) {
-  return DoAll(a1, DoAll(a2, a3, a4, a5));
-}
-
-template <typename Action1, typename Action2, typename Action3,
-    typename Action4, typename Action5, typename Action6>
-inline internal::DoBothAction<Action1, internal::DoBothAction<Action2,
-    internal::DoBothAction<Action3, internal::DoBothAction<Action4,
-    internal::DoBothAction<Action5, Action6> > > > >
-DoAll(Action1 a1, Action2 a2, Action3 a3, Action4 a4, Action5 a5, Action6 a6) {
-  return DoAll(a1, DoAll(a2, a3, a4, a5, a6));
-}
-
-template <typename Action1, typename Action2, typename Action3,
-    typename Action4, typename Action5, typename Action6, typename Action7>
-inline internal::DoBothAction<Action1, internal::DoBothAction<Action2,
-    internal::DoBothAction<Action3, internal::DoBothAction<Action4,
-    internal::DoBothAction<Action5, internal::DoBothAction<Action6,
-    Action7> > > > > >
-DoAll(Action1 a1, Action2 a2, Action3 a3, Action4 a4, Action5 a5, Action6 a6,
-    Action7 a7) {
-  return DoAll(a1, DoAll(a2, a3, a4, a5, a6, a7));
-}
-
-template <typename Action1, typename Action2, typename Action3,
-    typename Action4, typename Action5, typename Action6, typename Action7,
-    typename Action8>
-inline internal::DoBothAction<Action1, internal::DoBothAction<Action2,
-    internal::DoBothAction<Action3, internal::DoBothAction<Action4,
-    internal::DoBothAction<Action5, internal::DoBothAction<Action6,
-    internal::DoBothAction<Action7, Action8> > > > > > >
-DoAll(Action1 a1, Action2 a2, Action3 a3, Action4 a4, Action5 a5, Action6 a6,
-    Action7 a7, Action8 a8) {
-  return DoAll(a1, DoAll(a2, a3, a4, a5, a6, a7, a8));
-}
-
-template <typename Action1, typename Action2, typename Action3,
-    typename Action4, typename Action5, typename Action6, typename Action7,
-    typename Action8, typename Action9>
-inline internal::DoBothAction<Action1, internal::DoBothAction<Action2,
-    internal::DoBothAction<Action3, internal::DoBothAction<Action4,
-    internal::DoBothAction<Action5, internal::DoBothAction<Action6,
-    internal::DoBothAction<Action7, internal::DoBothAction<Action8,
-    Action9> > > > > > > >
-DoAll(Action1 a1, Action2 a2, Action3 a3, Action4 a4, Action5 a5, Action6 a6,
-    Action7 a7, Action8 a8, Action9 a9) {
-  return DoAll(a1, DoAll(a2, a3, a4, a5, a6, a7, a8, a9));
-}
-
-template <typename Action1, typename Action2, typename Action3,
-    typename Action4, typename Action5, typename Action6, typename Action7,
-    typename Action8, typename Action9, typename Action10>
-inline internal::DoBothAction<Action1, internal::DoBothAction<Action2,
-    internal::DoBothAction<Action3, internal::DoBothAction<Action4,
-    internal::DoBothAction<Action5, internal::DoBothAction<Action6,
-    internal::DoBothAction<Action7, internal::DoBothAction<Action8,
-    internal::DoBothAction<Action9, Action10> > > > > > > > >
-DoAll(Action1 a1, Action2 a2, Action3 a3, Action4 a4, Action5 a5, Action6 a6,
-    Action7 a7, Action8 a8, Action9 a9, Action10 a10) {
-  return DoAll(a1, DoAll(a2, a3, a4, a5, a6, a7, a8, a9, a10));
-}
-
-}  // namespace testing
-
-// The ACTION* family of macros can be used in a namespace scope to
-// define custom actions easily.  The syntax:
-//
-//   ACTION(name) { statements; }
-//
-// will define an action with the given name that executes the
-// statements.  The value returned by the statements will be used as
-// the return value of the action.  Inside the statements, you can
-// refer to the K-th (0-based) argument of the mock function by
-// 'argK', and refer to its type by 'argK_type'.  For example:
-//
-//   ACTION(IncrementArg1) {
-//     arg1_type temp = arg1;
-//     return ++(*temp);
-//   }
-//
-// allows you to write
-//
-//   ...WillOnce(IncrementArg1());
-//
-// You can also refer to the entire argument tuple and its type by
-// 'args' and 'args_type', and refer to the mock function type and its
-// return type by 'function_type' and 'return_type'.
-//
-// Note that you don't need to specify the types of the mock function
-// arguments.  However rest assured that your code is still type-safe:
-// you'll get a compiler error if *arg1 doesn't support the ++
-// operator, or if the type of ++(*arg1) isn't compatible with the
-// mock function's return type, for example.
-//
-// Sometimes you'll want to parameterize the action.   For that you can use
-// another macro:
-//
-//   ACTION_P(name, param_name) { statements; }
-//
-// For example:
-//
-//   ACTION_P(Add, n) { return arg0 + n; }
-//
-// will allow you to write:
-//
-//   ...WillOnce(Add(5));
-//
-// Note that you don't need to provide the type of the parameter
-// either.  If you need to reference the type of a parameter named
-// 'foo', you can write 'foo_type'.  For example, in the body of
-// ACTION_P(Add, n) above, you can write 'n_type' to refer to the type
-// of 'n'.
-//
-// We also provide ACTION_P2, ACTION_P3, ..., up to ACTION_P10 to support
-// multi-parameter actions.
-//
-// For the purpose of typing, you can view
-//
-//   ACTION_Pk(Foo, p1, ..., pk) { ... }
-//
-// as shorthand for
-//
-//   template <typename p1_type, ..., typename pk_type>
-//   FooActionPk<p1_type, ..., pk_type> Foo(p1_type p1, ..., pk_type pk) { ... }
-//
-// In particular, you can provide the template type arguments
-// explicitly when invoking Foo(), as in Foo<long, bool>(5, false);
-// although usually you can rely on the compiler to infer the types
-// for you automatically.  You can assign the result of expression
-// Foo(p1, ..., pk) to a variable of type FooActionPk<p1_type, ...,
-// pk_type>.  This can be useful when composing actions.
-//
-// You can also overload actions with different numbers of parameters:
-//
-//   ACTION_P(Plus, a) { ... }
-//   ACTION_P2(Plus, a, b) { ... }
-//
-// While it's tempting to always use the ACTION* macros when defining
-// a new action, you should also consider implementing ActionInterface
-// or using MakePolymorphicAction() instead, especially if you need to
-// use the action a lot.  While these approaches require more work,
-// they give you more control on the types of the mock function
-// arguments and the action parameters, which in general leads to
-// better compiler error messages that pay off in the long run.  They
-// also allow overloading actions based on parameter types (as opposed
-// to just based on the number of parameters).
-//
-// CAVEAT:
-//
-// ACTION*() can only be used in a namespace scope.  The reason is
-// that C++ doesn't yet allow function-local types to be used to
-// instantiate templates.  The up-coming C++0x standard will fix this.
-// Once that's done, we'll consider supporting using ACTION*() inside
-// a function.
-//
-// MORE INFORMATION:
-//
-// To learn more about using these macros, please search for 'ACTION'
-// on https://github.com/google/googletest/blob/master/googlemock/docs/CookBook.md
-
-// An internal macro needed for implementing ACTION*().
-#define GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_\
-    const args_type& args GTEST_ATTRIBUTE_UNUSED_, \
-    arg0_type arg0 GTEST_ATTRIBUTE_UNUSED_, \
-    arg1_type arg1 GTEST_ATTRIBUTE_UNUSED_, \
-    arg2_type arg2 GTEST_ATTRIBUTE_UNUSED_, \
-    arg3_type arg3 GTEST_ATTRIBUTE_UNUSED_, \
-    arg4_type arg4 GTEST_ATTRIBUTE_UNUSED_, \
-    arg5_type arg5 GTEST_ATTRIBUTE_UNUSED_, \
-    arg6_type arg6 GTEST_ATTRIBUTE_UNUSED_, \
-    arg7_type arg7 GTEST_ATTRIBUTE_UNUSED_, \
-    arg8_type arg8 GTEST_ATTRIBUTE_UNUSED_, \
-    arg9_type arg9 GTEST_ATTRIBUTE_UNUSED_
-
-// Sometimes you want to give an action explicit template parameters
-// that cannot be inferred from its value parameters.  ACTION() and
-// ACTION_P*() don't support that.  ACTION_TEMPLATE() remedies that
-// and can be viewed as an extension to ACTION() and ACTION_P*().
-//
-// The syntax:
-//
-//   ACTION_TEMPLATE(ActionName,
-//                   HAS_m_TEMPLATE_PARAMS(kind1, name1, ..., kind_m, name_m),
-//                   AND_n_VALUE_PARAMS(p1, ..., p_n)) { statements; }
-//
-// defines an action template that takes m explicit template
-// parameters and n value parameters.  name_i is the name of the i-th
-// template parameter, and kind_i specifies whether it's a typename,
-// an integral constant, or a template.  p_i is the name of the i-th
-// value parameter.
-//
-// Example:
-//
-//   // DuplicateArg<k, T>(output) converts the k-th argument of the mock
-//   // function to type T and copies it to *output.
-//   ACTION_TEMPLATE(DuplicateArg,
-//                   HAS_2_TEMPLATE_PARAMS(int, k, typename, T),
-//                   AND_1_VALUE_PARAMS(output)) {
-//     *output = T(::testing::get<k>(args));
-//   }
-//   ...
-//     int n;
-//     EXPECT_CALL(mock, Foo(_, _))
-//         .WillOnce(DuplicateArg<1, unsigned char>(&n));
-//
-// To create an instance of an action template, write:
-//
-//   ActionName<t1, ..., t_m>(v1, ..., v_n)
-//
-// where the ts are the template arguments and the vs are the value
-// arguments.  The value argument types are inferred by the compiler.
-// If you want to explicitly specify the value argument types, you can
-// provide additional template arguments:
-//
-//   ActionName<t1, ..., t_m, u1, ..., u_k>(v1, ..., v_n)
-//
-// where u_i is the desired type of v_i.
-//
-// ACTION_TEMPLATE and ACTION/ACTION_P* can be overloaded on the
-// number of value parameters, but not on the number of template
-// parameters.  Without the restriction, the meaning of the following
-// is unclear:
-//
-//   OverloadedAction<int, bool>(x);
-//
-// Are we using a single-template-parameter action where 'bool' refers
-// to the type of x, or are we using a two-template-parameter action
-// where the compiler is asked to infer the type of x?
-//
-// Implementation notes:
-//
-// GMOCK_INTERNAL_*_HAS_m_TEMPLATE_PARAMS and
-// GMOCK_INTERNAL_*_AND_n_VALUE_PARAMS are internal macros for
-// implementing ACTION_TEMPLATE.  The main trick we use is to create
-// new macro invocations when expanding a macro.  For example, we have
-//
-//   #define ACTION_TEMPLATE(name, template_params, value_params)
-//       ... GMOCK_INTERNAL_DECL_##template_params ...
-//
-// which causes ACTION_TEMPLATE(..., HAS_1_TEMPLATE_PARAMS(typename, T), ...)
-// to expand to
-//
-//       ... GMOCK_INTERNAL_DECL_HAS_1_TEMPLATE_PARAMS(typename, T) ...
-//
-// Since GMOCK_INTERNAL_DECL_HAS_1_TEMPLATE_PARAMS is a macro, the
-// preprocessor will continue to expand it to
-//
-//       ... typename T ...
-//
-// This technique conforms to the C++ standard and is portable.  It
-// allows us to implement action templates using O(N) code, where N is
-// the maximum number of template/value parameters supported.  Without
-// using it, we'd have to devote O(N^2) amount of code to implement all
-// combinations of m and n.
-
-// Declares the template parameters.
-#define GMOCK_INTERNAL_DECL_HAS_1_TEMPLATE_PARAMS(kind0, name0) kind0 name0
-#define GMOCK_INTERNAL_DECL_HAS_2_TEMPLATE_PARAMS(kind0, name0, kind1, \
-    name1) kind0 name0, kind1 name1
-#define GMOCK_INTERNAL_DECL_HAS_3_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2) kind0 name0, kind1 name1, kind2 name2
-#define GMOCK_INTERNAL_DECL_HAS_4_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3) kind0 name0, kind1 name1, kind2 name2, \
-    kind3 name3
-#define GMOCK_INTERNAL_DECL_HAS_5_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3, kind4, name4) kind0 name0, kind1 name1, \
-    kind2 name2, kind3 name3, kind4 name4
-#define GMOCK_INTERNAL_DECL_HAS_6_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3, kind4, name4, kind5, name5) kind0 name0, \
-    kind1 name1, kind2 name2, kind3 name3, kind4 name4, kind5 name5
-#define GMOCK_INTERNAL_DECL_HAS_7_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3, kind4, name4, kind5, name5, kind6, \
-    name6) kind0 name0, kind1 name1, kind2 name2, kind3 name3, kind4 name4, \
-    kind5 name5, kind6 name6
-#define GMOCK_INTERNAL_DECL_HAS_8_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3, kind4, name4, kind5, name5, kind6, name6, \
-    kind7, name7) kind0 name0, kind1 name1, kind2 name2, kind3 name3, \
-    kind4 name4, kind5 name5, kind6 name6, kind7 name7
-#define GMOCK_INTERNAL_DECL_HAS_9_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3, kind4, name4, kind5, name5, kind6, name6, \
-    kind7, name7, kind8, name8) kind0 name0, kind1 name1, kind2 name2, \
-    kind3 name3, kind4 name4, kind5 name5, kind6 name6, kind7 name7, \
-    kind8 name8
-#define GMOCK_INTERNAL_DECL_HAS_10_TEMPLATE_PARAMS(kind0, name0, kind1, \
-    name1, kind2, name2, kind3, name3, kind4, name4, kind5, name5, kind6, \
-    name6, kind7, name7, kind8, name8, kind9, name9) kind0 name0, \
-    kind1 name1, kind2 name2, kind3 name3, kind4 name4, kind5 name5, \
-    kind6 name6, kind7 name7, kind8 name8, kind9 name9
-
-// Lists the template parameters.
-#define GMOCK_INTERNAL_LIST_HAS_1_TEMPLATE_PARAMS(kind0, name0) name0
-#define GMOCK_INTERNAL_LIST_HAS_2_TEMPLATE_PARAMS(kind0, name0, kind1, \
-    name1) name0, name1
-#define GMOCK_INTERNAL_LIST_HAS_3_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2) name0, name1, name2
-#define GMOCK_INTERNAL_LIST_HAS_4_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3) name0, name1, name2, name3
-#define GMOCK_INTERNAL_LIST_HAS_5_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3, kind4, name4) name0, name1, name2, name3, \
-    name4
-#define GMOCK_INTERNAL_LIST_HAS_6_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3, kind4, name4, kind5, name5) name0, name1, \
-    name2, name3, name4, name5
-#define GMOCK_INTERNAL_LIST_HAS_7_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3, kind4, name4, kind5, name5, kind6, \
-    name6) name0, name1, name2, name3, name4, name5, name6
-#define GMOCK_INTERNAL_LIST_HAS_8_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3, kind4, name4, kind5, name5, kind6, name6, \
-    kind7, name7) name0, name1, name2, name3, name4, name5, name6, name7
-#define GMOCK_INTERNAL_LIST_HAS_9_TEMPLATE_PARAMS(kind0, name0, kind1, name1, \
-    kind2, name2, kind3, name3, kind4, name4, kind5, name5, kind6, name6, \
-    kind7, name7, kind8, name8) name0, name1, name2, name3, name4, name5, \
-    name6, name7, name8
-#define GMOCK_INTERNAL_LIST_HAS_10_TEMPLATE_PARAMS(kind0, name0, kind1, \
-    name1, kind2, name2, kind3, name3, kind4, name4, kind5, name5, kind6, \
-    name6, kind7, name7, kind8, name8, kind9, name9) name0, name1, name2, \
-    name3, name4, name5, name6, name7, name8, name9
-
-// Declares the types of value parameters.
-#define GMOCK_INTERNAL_DECL_TYPE_AND_0_VALUE_PARAMS()
-#define GMOCK_INTERNAL_DECL_TYPE_AND_1_VALUE_PARAMS(p0) , typename p0##_type
-#define GMOCK_INTERNAL_DECL_TYPE_AND_2_VALUE_PARAMS(p0, p1) , \
-    typename p0##_type, typename p1##_type
-#define GMOCK_INTERNAL_DECL_TYPE_AND_3_VALUE_PARAMS(p0, p1, p2) , \
-    typename p0##_type, typename p1##_type, typename p2##_type
-#define GMOCK_INTERNAL_DECL_TYPE_AND_4_VALUE_PARAMS(p0, p1, p2, p3) , \
-    typename p0##_type, typename p1##_type, typename p2##_type, \
-    typename p3##_type
-#define GMOCK_INTERNAL_DECL_TYPE_AND_5_VALUE_PARAMS(p0, p1, p2, p3, p4) , \
-    typename p0##_type, typename p1##_type, typename p2##_type, \
-    typename p3##_type, typename p4##_type
-#define GMOCK_INTERNAL_DECL_TYPE_AND_6_VALUE_PARAMS(p0, p1, p2, p3, p4, p5) , \
-    typename p0##_type, typename p1##_type, typename p2##_type, \
-    typename p3##_type, typename p4##_type, typename p5##_type
-#define GMOCK_INTERNAL_DECL_TYPE_AND_7_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6) , typename p0##_type, typename p1##_type, typename p2##_type, \
-    typename p3##_type, typename p4##_type, typename p5##_type, \
-    typename p6##_type
-#define GMOCK_INTERNAL_DECL_TYPE_AND_8_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6, p7) , typename p0##_type, typename p1##_type, typename p2##_type, \
-    typename p3##_type, typename p4##_type, typename p5##_type, \
-    typename p6##_type, typename p7##_type
-#define GMOCK_INTERNAL_DECL_TYPE_AND_9_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6, p7, p8) , typename p0##_type, typename p1##_type, typename p2##_type, \
-    typename p3##_type, typename p4##_type, typename p5##_type, \
-    typename p6##_type, typename p7##_type, typename p8##_type
-#define GMOCK_INTERNAL_DECL_TYPE_AND_10_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6, p7, p8, p9) , typename p0##_type, typename p1##_type, \
-    typename p2##_type, typename p3##_type, typename p4##_type, \
-    typename p5##_type, typename p6##_type, typename p7##_type, \
-    typename p8##_type, typename p9##_type
-
-// Initializes the value parameters.
-#define GMOCK_INTERNAL_INIT_AND_0_VALUE_PARAMS()\
-    ()
-#define GMOCK_INTERNAL_INIT_AND_1_VALUE_PARAMS(p0)\
-    (p0##_type gmock_p0) : p0(::testing::internal::move(gmock_p0))
-#define GMOCK_INTERNAL_INIT_AND_2_VALUE_PARAMS(p0, p1)\
-    (p0##_type gmock_p0, \
-        p1##_type gmock_p1) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1))
-#define GMOCK_INTERNAL_INIT_AND_3_VALUE_PARAMS(p0, p1, p2)\
-    (p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2))
-#define GMOCK_INTERNAL_INIT_AND_4_VALUE_PARAMS(p0, p1, p2, p3)\
-    (p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-        p3##_type gmock_p3) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3))
-#define GMOCK_INTERNAL_INIT_AND_5_VALUE_PARAMS(p0, p1, p2, p3, p4)\
-    (p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-        p3##_type gmock_p3, \
-        p4##_type gmock_p4) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4))
-#define GMOCK_INTERNAL_INIT_AND_6_VALUE_PARAMS(p0, p1, p2, p3, p4, p5)\
-    (p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-        p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)), \
-        p5(::testing::internal::move(gmock_p5))
-#define GMOCK_INTERNAL_INIT_AND_7_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6)\
-    (p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-        p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-        p6##_type gmock_p6) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)), \
-        p5(::testing::internal::move(gmock_p5)), \
-        p6(::testing::internal::move(gmock_p6))
-#define GMOCK_INTERNAL_INIT_AND_8_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, p7)\
-    (p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-        p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-        p6##_type gmock_p6, \
-        p7##_type gmock_p7) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)), \
-        p5(::testing::internal::move(gmock_p5)), \
-        p6(::testing::internal::move(gmock_p6)), \
-        p7(::testing::internal::move(gmock_p7))
-#define GMOCK_INTERNAL_INIT_AND_9_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7, p8)\
-    (p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-        p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-        p6##_type gmock_p6, p7##_type gmock_p7, \
-        p8##_type gmock_p8) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)), \
-        p5(::testing::internal::move(gmock_p5)), \
-        p6(::testing::internal::move(gmock_p6)), \
-        p7(::testing::internal::move(gmock_p7)), \
-        p8(::testing::internal::move(gmock_p8))
-#define GMOCK_INTERNAL_INIT_AND_10_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7, p8, p9)\
-    (p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-        p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-        p6##_type gmock_p6, p7##_type gmock_p7, p8##_type gmock_p8, \
-        p9##_type gmock_p9) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)), \
-        p5(::testing::internal::move(gmock_p5)), \
-        p6(::testing::internal::move(gmock_p6)), \
-        p7(::testing::internal::move(gmock_p7)), \
-        p8(::testing::internal::move(gmock_p8)), \
-        p9(::testing::internal::move(gmock_p9))
-
-// Declares the fields for storing the value parameters.
-#define GMOCK_INTERNAL_DEFN_AND_0_VALUE_PARAMS()
-#define GMOCK_INTERNAL_DEFN_AND_1_VALUE_PARAMS(p0) p0##_type p0;
-#define GMOCK_INTERNAL_DEFN_AND_2_VALUE_PARAMS(p0, p1) p0##_type p0; \
-    p1##_type p1;
-#define GMOCK_INTERNAL_DEFN_AND_3_VALUE_PARAMS(p0, p1, p2) p0##_type p0; \
-    p1##_type p1; p2##_type p2;
-#define GMOCK_INTERNAL_DEFN_AND_4_VALUE_PARAMS(p0, p1, p2, p3) p0##_type p0; \
-    p1##_type p1; p2##_type p2; p3##_type p3;
-#define GMOCK_INTERNAL_DEFN_AND_5_VALUE_PARAMS(p0, p1, p2, p3, \
-    p4) p0##_type p0; p1##_type p1; p2##_type p2; p3##_type p3; p4##_type p4;
-#define GMOCK_INTERNAL_DEFN_AND_6_VALUE_PARAMS(p0, p1, p2, p3, p4, \
-    p5) p0##_type p0; p1##_type p1; p2##_type p2; p3##_type p3; p4##_type p4; \
-    p5##_type p5;
-#define GMOCK_INTERNAL_DEFN_AND_7_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6) p0##_type p0; p1##_type p1; p2##_type p2; p3##_type p3; p4##_type p4; \
-    p5##_type p5; p6##_type p6;
-#define GMOCK_INTERNAL_DEFN_AND_8_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7) p0##_type p0; p1##_type p1; p2##_type p2; p3##_type p3; p4##_type p4; \
-    p5##_type p5; p6##_type p6; p7##_type p7;
-#define GMOCK_INTERNAL_DEFN_AND_9_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7, p8) p0##_type p0; p1##_type p1; p2##_type p2; p3##_type p3; \
-    p4##_type p4; p5##_type p5; p6##_type p6; p7##_type p7; p8##_type p8;
-#define GMOCK_INTERNAL_DEFN_AND_10_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7, p8, p9) p0##_type p0; p1##_type p1; p2##_type p2; p3##_type p3; \
-    p4##_type p4; p5##_type p5; p6##_type p6; p7##_type p7; p8##_type p8; \
-    p9##_type p9;
-
-// Lists the value parameters.
-#define GMOCK_INTERNAL_LIST_AND_0_VALUE_PARAMS()
-#define GMOCK_INTERNAL_LIST_AND_1_VALUE_PARAMS(p0) p0
-#define GMOCK_INTERNAL_LIST_AND_2_VALUE_PARAMS(p0, p1) p0, p1
-#define GMOCK_INTERNAL_LIST_AND_3_VALUE_PARAMS(p0, p1, p2) p0, p1, p2
-#define GMOCK_INTERNAL_LIST_AND_4_VALUE_PARAMS(p0, p1, p2, p3) p0, p1, p2, p3
-#define GMOCK_INTERNAL_LIST_AND_5_VALUE_PARAMS(p0, p1, p2, p3, p4) p0, p1, \
-    p2, p3, p4
-#define GMOCK_INTERNAL_LIST_AND_6_VALUE_PARAMS(p0, p1, p2, p3, p4, p5) p0, \
-    p1, p2, p3, p4, p5
-#define GMOCK_INTERNAL_LIST_AND_7_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6) p0, p1, p2, p3, p4, p5, p6
-#define GMOCK_INTERNAL_LIST_AND_8_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7) p0, p1, p2, p3, p4, p5, p6, p7
-#define GMOCK_INTERNAL_LIST_AND_9_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7, p8) p0, p1, p2, p3, p4, p5, p6, p7, p8
-#define GMOCK_INTERNAL_LIST_AND_10_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7, p8, p9) p0, p1, p2, p3, p4, p5, p6, p7, p8, p9
-
-// Lists the value parameter types.
-#define GMOCK_INTERNAL_LIST_TYPE_AND_0_VALUE_PARAMS()
-#define GMOCK_INTERNAL_LIST_TYPE_AND_1_VALUE_PARAMS(p0) , p0##_type
-#define GMOCK_INTERNAL_LIST_TYPE_AND_2_VALUE_PARAMS(p0, p1) , p0##_type, \
-    p1##_type
-#define GMOCK_INTERNAL_LIST_TYPE_AND_3_VALUE_PARAMS(p0, p1, p2) , p0##_type, \
-    p1##_type, p2##_type
-#define GMOCK_INTERNAL_LIST_TYPE_AND_4_VALUE_PARAMS(p0, p1, p2, p3) , \
-    p0##_type, p1##_type, p2##_type, p3##_type
-#define GMOCK_INTERNAL_LIST_TYPE_AND_5_VALUE_PARAMS(p0, p1, p2, p3, p4) , \
-    p0##_type, p1##_type, p2##_type, p3##_type, p4##_type
-#define GMOCK_INTERNAL_LIST_TYPE_AND_6_VALUE_PARAMS(p0, p1, p2, p3, p4, p5) , \
-    p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, p5##_type
-#define GMOCK_INTERNAL_LIST_TYPE_AND_7_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6) , p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, p5##_type, \
-    p6##_type
-#define GMOCK_INTERNAL_LIST_TYPE_AND_8_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6, p7) , p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-    p5##_type, p6##_type, p7##_type
-#define GMOCK_INTERNAL_LIST_TYPE_AND_9_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6, p7, p8) , p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-    p5##_type, p6##_type, p7##_type, p8##_type
-#define GMOCK_INTERNAL_LIST_TYPE_AND_10_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6, p7, p8, p9) , p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-    p5##_type, p6##_type, p7##_type, p8##_type, p9##_type
-
-// Declares the value parameters.
-#define GMOCK_INTERNAL_DECL_AND_0_VALUE_PARAMS()
-#define GMOCK_INTERNAL_DECL_AND_1_VALUE_PARAMS(p0) p0##_type p0
-#define GMOCK_INTERNAL_DECL_AND_2_VALUE_PARAMS(p0, p1) p0##_type p0, \
-    p1##_type p1
-#define GMOCK_INTERNAL_DECL_AND_3_VALUE_PARAMS(p0, p1, p2) p0##_type p0, \
-    p1##_type p1, p2##_type p2
-#define GMOCK_INTERNAL_DECL_AND_4_VALUE_PARAMS(p0, p1, p2, p3) p0##_type p0, \
-    p1##_type p1, p2##_type p2, p3##_type p3
-#define GMOCK_INTERNAL_DECL_AND_5_VALUE_PARAMS(p0, p1, p2, p3, \
-    p4) p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, p4##_type p4
-#define GMOCK_INTERNAL_DECL_AND_6_VALUE_PARAMS(p0, p1, p2, p3, p4, \
-    p5) p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, p4##_type p4, \
-    p5##_type p5
-#define GMOCK_INTERNAL_DECL_AND_7_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, \
-    p6) p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, p4##_type p4, \
-    p5##_type p5, p6##_type p6
-#define GMOCK_INTERNAL_DECL_AND_8_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7) p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, p4##_type p4, \
-    p5##_type p5, p6##_type p6, p7##_type p7
-#define GMOCK_INTERNAL_DECL_AND_9_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7, p8) p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, \
-    p4##_type p4, p5##_type p5, p6##_type p6, p7##_type p7, p8##_type p8
-#define GMOCK_INTERNAL_DECL_AND_10_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7, p8, p9) p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, \
-    p4##_type p4, p5##_type p5, p6##_type p6, p7##_type p7, p8##_type p8, \
-    p9##_type p9
-
-// The suffix of the class template implementing the action template.
-#define GMOCK_INTERNAL_COUNT_AND_0_VALUE_PARAMS()
-#define GMOCK_INTERNAL_COUNT_AND_1_VALUE_PARAMS(p0) P
-#define GMOCK_INTERNAL_COUNT_AND_2_VALUE_PARAMS(p0, p1) P2
-#define GMOCK_INTERNAL_COUNT_AND_3_VALUE_PARAMS(p0, p1, p2) P3
-#define GMOCK_INTERNAL_COUNT_AND_4_VALUE_PARAMS(p0, p1, p2, p3) P4
-#define GMOCK_INTERNAL_COUNT_AND_5_VALUE_PARAMS(p0, p1, p2, p3, p4) P5
-#define GMOCK_INTERNAL_COUNT_AND_6_VALUE_PARAMS(p0, p1, p2, p3, p4, p5) P6
-#define GMOCK_INTERNAL_COUNT_AND_7_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6) P7
-#define GMOCK_INTERNAL_COUNT_AND_8_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7) P8
-#define GMOCK_INTERNAL_COUNT_AND_9_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7, p8) P9
-#define GMOCK_INTERNAL_COUNT_AND_10_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, \
-    p7, p8, p9) P10
-
-// The name of the class template implementing the action template.
-#define GMOCK_ACTION_CLASS_(name, value_params)\
-    GTEST_CONCAT_TOKEN_(name##Action, GMOCK_INTERNAL_COUNT_##value_params)
-
-#define ACTION_TEMPLATE(name, template_params, value_params)\
-  template <GMOCK_INTERNAL_DECL_##template_params\
-            GMOCK_INTERNAL_DECL_TYPE_##value_params>\
-  class GMOCK_ACTION_CLASS_(name, value_params) {\
-   public:\
-    explicit GMOCK_ACTION_CLASS_(name, value_params)\
-        GMOCK_INTERNAL_INIT_##value_params {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      explicit gmock_Impl GMOCK_INTERNAL_INIT_##value_params {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      GMOCK_INTERNAL_DEFN_##value_params\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(\
-          new gmock_Impl<F>(GMOCK_INTERNAL_LIST_##value_params));\
-    }\
-    GMOCK_INTERNAL_DEFN_##value_params\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(GMOCK_ACTION_CLASS_(name, value_params));\
-  };\
-  template <GMOCK_INTERNAL_DECL_##template_params\
-            GMOCK_INTERNAL_DECL_TYPE_##value_params>\
-  inline GMOCK_ACTION_CLASS_(name, value_params)<\
-      GMOCK_INTERNAL_LIST_##template_params\
-      GMOCK_INTERNAL_LIST_TYPE_##value_params> name(\
-          GMOCK_INTERNAL_DECL_##value_params) {\
-    return GMOCK_ACTION_CLASS_(name, value_params)<\
-        GMOCK_INTERNAL_LIST_##template_params\
-        GMOCK_INTERNAL_LIST_TYPE_##value_params>(\
-            GMOCK_INTERNAL_LIST_##value_params);\
-  }\
-  template <GMOCK_INTERNAL_DECL_##template_params\
-            GMOCK_INTERNAL_DECL_TYPE_##value_params>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      GMOCK_ACTION_CLASS_(name, value_params)<\
-          GMOCK_INTERNAL_LIST_##template_params\
-          GMOCK_INTERNAL_LIST_TYPE_##value_params>::gmock_Impl<F>::\
-              gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION(name)\
-  class name##Action {\
-   public:\
-    name##Action() {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      gmock_Impl() {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>());\
-    }\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##Action);\
-  };\
-  inline name##Action name() {\
-    return name##Action();\
-  }\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##Action::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION_P(name, p0)\
-  template <typename p0##_type>\
-  class name##ActionP {\
-   public:\
-    explicit name##ActionP(p0##_type gmock_p0) : \
-        p0(::testing::internal::forward<p0##_type>(gmock_p0)) {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      explicit gmock_Impl(p0##_type gmock_p0) : \
-          p0(::testing::internal::forward<p0##_type>(gmock_p0)) {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      p0##_type p0;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>(p0));\
-    }\
-    p0##_type p0;\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##ActionP);\
-  };\
-  template <typename p0##_type>\
-  inline name##ActionP<p0##_type> name(p0##_type p0) {\
-    return name##ActionP<p0##_type>(p0);\
-  }\
-  template <typename p0##_type>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##ActionP<p0##_type>::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION_P2(name, p0, p1)\
-  template <typename p0##_type, typename p1##_type>\
-  class name##ActionP2 {\
-   public:\
-    name##ActionP2(p0##_type gmock_p0, \
-        p1##_type gmock_p1) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-        p1(::testing::internal::forward<p1##_type>(gmock_p1)) {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      gmock_Impl(p0##_type gmock_p0, \
-          p1##_type gmock_p1) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-          p1(::testing::internal::forward<p1##_type>(gmock_p1)) {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      p0##_type p0;\
-      p1##_type p1;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>(p0, p1));\
-    }\
-    p0##_type p0;\
-    p1##_type p1;\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##ActionP2);\
-  };\
-  template <typename p0##_type, typename p1##_type>\
-  inline name##ActionP2<p0##_type, p1##_type> name(p0##_type p0, \
-      p1##_type p1) {\
-    return name##ActionP2<p0##_type, p1##_type>(p0, p1);\
-  }\
-  template <typename p0##_type, typename p1##_type>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##ActionP2<p0##_type, p1##_type>::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION_P3(name, p0, p1, p2)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type>\
-  class name##ActionP3 {\
-   public:\
-    name##ActionP3(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-        p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-        p2(::testing::internal::forward<p2##_type>(gmock_p2)) {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, \
-          p2##_type gmock_p2) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-          p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-          p2(::testing::internal::forward<p2##_type>(gmock_p2)) {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      p0##_type p0;\
-      p1##_type p1;\
-      p2##_type p2;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>(p0, p1, p2));\
-    }\
-    p0##_type p0;\
-    p1##_type p1;\
-    p2##_type p2;\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##ActionP3);\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type>\
-  inline name##ActionP3<p0##_type, p1##_type, p2##_type> name(p0##_type p0, \
-      p1##_type p1, p2##_type p2) {\
-    return name##ActionP3<p0##_type, p1##_type, p2##_type>(p0, p1, p2);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##ActionP3<p0##_type, p1##_type, \
-          p2##_type>::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION_P4(name, p0, p1, p2, p3)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type>\
-  class name##ActionP4 {\
-   public:\
-    name##ActionP4(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, \
-        p3##_type gmock_p3) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-        p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-        p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-        p3(::testing::internal::forward<p3##_type>(gmock_p3)) {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-          p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-          p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-          p3(::testing::internal::forward<p3##_type>(gmock_p3)) {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      p0##_type p0;\
-      p1##_type p1;\
-      p2##_type p2;\
-      p3##_type p3;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>(p0, p1, p2, p3));\
-    }\
-    p0##_type p0;\
-    p1##_type p1;\
-    p2##_type p2;\
-    p3##_type p3;\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##ActionP4);\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type>\
-  inline name##ActionP4<p0##_type, p1##_type, p2##_type, \
-      p3##_type> name(p0##_type p0, p1##_type p1, p2##_type p2, \
-      p3##_type p3) {\
-    return name##ActionP4<p0##_type, p1##_type, p2##_type, p3##_type>(p0, p1, \
-        p2, p3);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##ActionP4<p0##_type, p1##_type, p2##_type, \
-          p3##_type>::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION_P5(name, p0, p1, p2, p3, p4)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type>\
-  class name##ActionP5 {\
-   public:\
-    name##ActionP5(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, \
-        p4##_type gmock_p4) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-        p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-        p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-        p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-        p4(::testing::internal::forward<p4##_type>(gmock_p4)) {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, \
-          p4##_type gmock_p4) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-          p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-          p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-          p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-          p4(::testing::internal::forward<p4##_type>(gmock_p4)) {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      p0##_type p0;\
-      p1##_type p1;\
-      p2##_type p2;\
-      p3##_type p3;\
-      p4##_type p4;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>(p0, p1, p2, p3, p4));\
-    }\
-    p0##_type p0;\
-    p1##_type p1;\
-    p2##_type p2;\
-    p3##_type p3;\
-    p4##_type p4;\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##ActionP5);\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type>\
-  inline name##ActionP5<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type> name(p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, \
-      p4##_type p4) {\
-    return name##ActionP5<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type>(p0, p1, p2, p3, p4);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##ActionP5<p0##_type, p1##_type, p2##_type, p3##_type, \
-          p4##_type>::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION_P6(name, p0, p1, p2, p3, p4, p5)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type>\
-  class name##ActionP6 {\
-   public:\
-    name##ActionP6(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-        p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-        p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-        p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-        p4(::testing::internal::forward<p4##_type>(gmock_p4)), \
-        p5(::testing::internal::forward<p5##_type>(gmock_p5)) {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4, \
-          p5##_type gmock_p5) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-          p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-          p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-          p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-          p4(::testing::internal::forward<p4##_type>(gmock_p4)), \
-          p5(::testing::internal::forward<p5##_type>(gmock_p5)) {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      p0##_type p0;\
-      p1##_type p1;\
-      p2##_type p2;\
-      p3##_type p3;\
-      p4##_type p4;\
-      p5##_type p5;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>(p0, p1, p2, p3, p4, p5));\
-    }\
-    p0##_type p0;\
-    p1##_type p1;\
-    p2##_type p2;\
-    p3##_type p3;\
-    p4##_type p4;\
-    p5##_type p5;\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##ActionP6);\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type>\
-  inline name##ActionP6<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type> name(p0##_type p0, p1##_type p1, p2##_type p2, \
-      p3##_type p3, p4##_type p4, p5##_type p5) {\
-    return name##ActionP6<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type, p5##_type>(p0, p1, p2, p3, p4, p5);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##ActionP6<p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-          p5##_type>::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION_P7(name, p0, p1, p2, p3, p4, p5, p6)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type>\
-  class name##ActionP7 {\
-   public:\
-    name##ActionP7(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5, \
-        p6##_type gmock_p6) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-        p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-        p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-        p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-        p4(::testing::internal::forward<p4##_type>(gmock_p4)), \
-        p5(::testing::internal::forward<p5##_type>(gmock_p5)), \
-        p6(::testing::internal::forward<p6##_type>(gmock_p6)) {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-          p6##_type gmock_p6) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-          p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-          p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-          p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-          p4(::testing::internal::forward<p4##_type>(gmock_p4)), \
-          p5(::testing::internal::forward<p5##_type>(gmock_p5)), \
-          p6(::testing::internal::forward<p6##_type>(gmock_p6)) {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      p0##_type p0;\
-      p1##_type p1;\
-      p2##_type p2;\
-      p3##_type p3;\
-      p4##_type p4;\
-      p5##_type p5;\
-      p6##_type p6;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>(p0, p1, p2, p3, p4, p5, \
-          p6));\
-    }\
-    p0##_type p0;\
-    p1##_type p1;\
-    p2##_type p2;\
-    p3##_type p3;\
-    p4##_type p4;\
-    p5##_type p5;\
-    p6##_type p6;\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##ActionP7);\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type>\
-  inline name##ActionP7<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type, p6##_type> name(p0##_type p0, p1##_type p1, \
-      p2##_type p2, p3##_type p3, p4##_type p4, p5##_type p5, \
-      p6##_type p6) {\
-    return name##ActionP7<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type, p5##_type, p6##_type>(p0, p1, p2, p3, p4, p5, p6);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##ActionP7<p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-          p5##_type, p6##_type>::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION_P8(name, p0, p1, p2, p3, p4, p5, p6, p7)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type>\
-  class name##ActionP8 {\
-   public:\
-    name##ActionP8(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5, p6##_type gmock_p6, \
-        p7##_type gmock_p7) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-        p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-        p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-        p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-        p4(::testing::internal::forward<p4##_type>(gmock_p4)), \
-        p5(::testing::internal::forward<p5##_type>(gmock_p5)), \
-        p6(::testing::internal::forward<p6##_type>(gmock_p6)), \
-        p7(::testing::internal::forward<p7##_type>(gmock_p7)) {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-          p6##_type gmock_p6, \
-          p7##_type gmock_p7) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-          p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-          p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-          p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-          p4(::testing::internal::forward<p4##_type>(gmock_p4)), \
-          p5(::testing::internal::forward<p5##_type>(gmock_p5)), \
-          p6(::testing::internal::forward<p6##_type>(gmock_p6)), \
-          p7(::testing::internal::forward<p7##_type>(gmock_p7)) {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      p0##_type p0;\
-      p1##_type p1;\
-      p2##_type p2;\
-      p3##_type p3;\
-      p4##_type p4;\
-      p5##_type p5;\
-      p6##_type p6;\
-      p7##_type p7;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>(p0, p1, p2, p3, p4, p5, \
-          p6, p7));\
-    }\
-    p0##_type p0;\
-    p1##_type p1;\
-    p2##_type p2;\
-    p3##_type p3;\
-    p4##_type p4;\
-    p5##_type p5;\
-    p6##_type p6;\
-    p7##_type p7;\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##ActionP8);\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type>\
-  inline name##ActionP8<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type, p6##_type, p7##_type> name(p0##_type p0, \
-      p1##_type p1, p2##_type p2, p3##_type p3, p4##_type p4, p5##_type p5, \
-      p6##_type p6, p7##_type p7) {\
-    return name##ActionP8<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type, p5##_type, p6##_type, p7##_type>(p0, p1, p2, p3, p4, p5, \
-        p6, p7);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##ActionP8<p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-          p5##_type, p6##_type, \
-          p7##_type>::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION_P9(name, p0, p1, p2, p3, p4, p5, p6, p7, p8)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type>\
-  class name##ActionP9 {\
-   public:\
-    name##ActionP9(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5, p6##_type gmock_p6, p7##_type gmock_p7, \
-        p8##_type gmock_p8) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-        p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-        p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-        p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-        p4(::testing::internal::forward<p4##_type>(gmock_p4)), \
-        p5(::testing::internal::forward<p5##_type>(gmock_p5)), \
-        p6(::testing::internal::forward<p6##_type>(gmock_p6)), \
-        p7(::testing::internal::forward<p7##_type>(gmock_p7)), \
-        p8(::testing::internal::forward<p8##_type>(gmock_p8)) {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-          p6##_type gmock_p6, p7##_type gmock_p7, \
-          p8##_type gmock_p8) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-          p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-          p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-          p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-          p4(::testing::internal::forward<p4##_type>(gmock_p4)), \
-          p5(::testing::internal::forward<p5##_type>(gmock_p5)), \
-          p6(::testing::internal::forward<p6##_type>(gmock_p6)), \
-          p7(::testing::internal::forward<p7##_type>(gmock_p7)), \
-          p8(::testing::internal::forward<p8##_type>(gmock_p8)) {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      p0##_type p0;\
-      p1##_type p1;\
-      p2##_type p2;\
-      p3##_type p3;\
-      p4##_type p4;\
-      p5##_type p5;\
-      p6##_type p6;\
-      p7##_type p7;\
-      p8##_type p8;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>(p0, p1, p2, p3, p4, p5, \
-          p6, p7, p8));\
-    }\
-    p0##_type p0;\
-    p1##_type p1;\
-    p2##_type p2;\
-    p3##_type p3;\
-    p4##_type p4;\
-    p5##_type p5;\
-    p6##_type p6;\
-    p7##_type p7;\
-    p8##_type p8;\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##ActionP9);\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type>\
-  inline name##ActionP9<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type, p6##_type, p7##_type, \
-      p8##_type> name(p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, \
-      p4##_type p4, p5##_type p5, p6##_type p6, p7##_type p7, \
-      p8##_type p8) {\
-    return name##ActionP9<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type, p5##_type, p6##_type, p7##_type, p8##_type>(p0, p1, p2, \
-        p3, p4, p5, p6, p7, p8);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##ActionP9<p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-          p5##_type, p6##_type, p7##_type, \
-          p8##_type>::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-#define ACTION_P10(name, p0, p1, p2, p3, p4, p5, p6, p7, p8, p9)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type, \
-      typename p9##_type>\
-  class name##ActionP10 {\
-   public:\
-    name##ActionP10(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5, p6##_type gmock_p6, p7##_type gmock_p7, \
-        p8##_type gmock_p8, \
-        p9##_type gmock_p9) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-        p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-        p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-        p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-        p4(::testing::internal::forward<p4##_type>(gmock_p4)), \
-        p5(::testing::internal::forward<p5##_type>(gmock_p5)), \
-        p6(::testing::internal::forward<p6##_type>(gmock_p6)), \
-        p7(::testing::internal::forward<p7##_type>(gmock_p7)), \
-        p8(::testing::internal::forward<p8##_type>(gmock_p8)), \
-        p9(::testing::internal::forward<p9##_type>(gmock_p9)) {}\
-    template <typename F>\
-    class gmock_Impl : public ::testing::ActionInterface<F> {\
-     public:\
-      typedef F function_type;\
-      typedef typename ::testing::internal::Function<F>::Result return_type;\
-      typedef typename ::testing::internal::Function<F>::ArgumentTuple\
-          args_type;\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-          p6##_type gmock_p6, p7##_type gmock_p7, p8##_type gmock_p8, \
-          p9##_type gmock_p9) : p0(::testing::internal::forward<p0##_type>(gmock_p0)), \
-          p1(::testing::internal::forward<p1##_type>(gmock_p1)), \
-          p2(::testing::internal::forward<p2##_type>(gmock_p2)), \
-          p3(::testing::internal::forward<p3##_type>(gmock_p3)), \
-          p4(::testing::internal::forward<p4##_type>(gmock_p4)), \
-          p5(::testing::internal::forward<p5##_type>(gmock_p5)), \
-          p6(::testing::internal::forward<p6##_type>(gmock_p6)), \
-          p7(::testing::internal::forward<p7##_type>(gmock_p7)), \
-          p8(::testing::internal::forward<p8##_type>(gmock_p8)), \
-          p9(::testing::internal::forward<p9##_type>(gmock_p9)) {}\
-      virtual return_type Perform(const args_type& args) {\
-        return ::testing::internal::ActionHelper<return_type, gmock_Impl>::\
-            Perform(this, args);\
-      }\
-      template <typename arg0_type, typename arg1_type, typename arg2_type, \
-          typename arg3_type, typename arg4_type, typename arg5_type, \
-          typename arg6_type, typename arg7_type, typename arg8_type, \
-          typename arg9_type>\
-      return_type gmock_PerformImpl(const args_type& args, arg0_type arg0, \
-          arg1_type arg1, arg2_type arg2, arg3_type arg3, arg4_type arg4, \
-          arg5_type arg5, arg6_type arg6, arg7_type arg7, arg8_type arg8, \
-          arg9_type arg9) const;\
-      p0##_type p0;\
-      p1##_type p1;\
-      p2##_type p2;\
-      p3##_type p3;\
-      p4##_type p4;\
-      p5##_type p5;\
-      p6##_type p6;\
-      p7##_type p7;\
-      p8##_type p8;\
-      p9##_type p9;\
-     private:\
-      GTEST_DISALLOW_ASSIGN_(gmock_Impl);\
-    };\
-    template <typename F> operator ::testing::Action<F>() const {\
-      return ::testing::Action<F>(new gmock_Impl<F>(p0, p1, p2, p3, p4, p5, \
-          p6, p7, p8, p9));\
-    }\
-    p0##_type p0;\
-    p1##_type p1;\
-    p2##_type p2;\
-    p3##_type p3;\
-    p4##_type p4;\
-    p5##_type p5;\
-    p6##_type p6;\
-    p7##_type p7;\
-    p8##_type p8;\
-    p9##_type p9;\
-   private:\
-    GTEST_DISALLOW_ASSIGN_(name##ActionP10);\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type, \
-      typename p9##_type>\
-  inline name##ActionP10<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type, p6##_type, p7##_type, p8##_type, \
-      p9##_type> name(p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, \
-      p4##_type p4, p5##_type p5, p6##_type p6, p7##_type p7, p8##_type p8, \
-      p9##_type p9) {\
-    return name##ActionP10<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type, p5##_type, p6##_type, p7##_type, p8##_type, p9##_type>(p0, \
-        p1, p2, p3, p4, p5, p6, p7, p8, p9);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type, \
-      typename p9##_type>\
-  template <typename F>\
-  template <typename arg0_type, typename arg1_type, typename arg2_type, \
-      typename arg3_type, typename arg4_type, typename arg5_type, \
-      typename arg6_type, typename arg7_type, typename arg8_type, \
-      typename arg9_type>\
-  typename ::testing::internal::Function<F>::Result\
-      name##ActionP10<p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-          p5##_type, p6##_type, p7##_type, p8##_type, \
-          p9##_type>::gmock_Impl<F>::gmock_PerformImpl(\
-          GMOCK_ACTION_ARG_TYPES_AND_NAMES_UNUSED_) const
-
-namespace testing {
-
-
-// The ACTION*() macros trigger warning C4100 (unreferenced formal
-// parameter) in MSVC with -W4.  Unfortunately they cannot be fixed in
-// the macro definition, as the warnings are generated when the macro
-// is expanded and macro expansion cannot contain #pragma.  Therefore
-// we suppress them here.
-#ifdef _MSC_VER
-# pragma warning(push)
-# pragma warning(disable:4100)
-#endif
-
-// Various overloads for InvokeArgument<N>().
-//
-// The InvokeArgument<N>(a1, a2, ..., a_k) action invokes the N-th
-// (0-based) argument, which must be a k-ary callable, of the mock
-// function, with arguments a1, a2, ..., a_k.
-//
-// Notes:
-//
-//   1. The arguments are passed by value by default.  If you need to
-//   pass an argument by reference, wrap it inside ByRef().  For
-//   example,
-//
-//     InvokeArgument<1>(5, string("Hello"), ByRef(foo))
-//
-//   passes 5 and string("Hello") by value, and passes foo by
-//   reference.
-//
-//   2. If the callable takes an argument by reference but ByRef() is
-//   not used, it will receive the reference to a copy of the value,
-//   instead of the original value.  For example, when the 0-th
-//   argument of the mock function takes a const string&, the action
-//
-//     InvokeArgument<0>(string("Hello"))
-//
-//   makes a copy of the temporary string("Hello") object and passes a
-//   reference of the copy, instead of the original temporary object,
-//   to the callable.  This makes it easy for a user to define an
-//   InvokeArgument action from temporary values and have it performed
-//   later.
-
-namespace internal {
-namespace invoke_argument {
-
-// Appears in InvokeArgumentAdl's argument list to help avoid
-// accidental calls to user functions of the same name.
-struct AdlTag {};
-
-// InvokeArgumentAdl - a helper for InvokeArgument.
-// The basic overloads are provided here for generic functors.
-// Overloads for other custom-callables are provided in the
-// internal/custom/callback-actions.h header.
-
-template <typename R, typename F>
-R InvokeArgumentAdl(AdlTag, F f) {
-  return f();
-}
-template <typename R, typename F, typename A1>
-R InvokeArgumentAdl(AdlTag, F f, A1 a1) {
-  return f(a1);
-}
-template <typename R, typename F, typename A1, typename A2>
-R InvokeArgumentAdl(AdlTag, F f, A1 a1, A2 a2) {
-  return f(a1, a2);
-}
-template <typename R, typename F, typename A1, typename A2, typename A3>
-R InvokeArgumentAdl(AdlTag, F f, A1 a1, A2 a2, A3 a3) {
-  return f(a1, a2, a3);
-}
-template <typename R, typename F, typename A1, typename A2, typename A3,
-    typename A4>
-R InvokeArgumentAdl(AdlTag, F f, A1 a1, A2 a2, A3 a3, A4 a4) {
-  return f(a1, a2, a3, a4);
-}
-template <typename R, typename F, typename A1, typename A2, typename A3,
-    typename A4, typename A5>
-R InvokeArgumentAdl(AdlTag, F f, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5) {
-  return f(a1, a2, a3, a4, a5);
-}
-template <typename R, typename F, typename A1, typename A2, typename A3,
-    typename A4, typename A5, typename A6>
-R InvokeArgumentAdl(AdlTag, F f, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6) {
-  return f(a1, a2, a3, a4, a5, a6);
-}
-template <typename R, typename F, typename A1, typename A2, typename A3,
-    typename A4, typename A5, typename A6, typename A7>
-R InvokeArgumentAdl(AdlTag, F f, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6,
-    A7 a7) {
-  return f(a1, a2, a3, a4, a5, a6, a7);
-}
-template <typename R, typename F, typename A1, typename A2, typename A3,
-    typename A4, typename A5, typename A6, typename A7, typename A8>
-R InvokeArgumentAdl(AdlTag, F f, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6,
-    A7 a7, A8 a8) {
-  return f(a1, a2, a3, a4, a5, a6, a7, a8);
-}
-template <typename R, typename F, typename A1, typename A2, typename A3,
-    typename A4, typename A5, typename A6, typename A7, typename A8,
-    typename A9>
-R InvokeArgumentAdl(AdlTag, F f, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6,
-    A7 a7, A8 a8, A9 a9) {
-  return f(a1, a2, a3, a4, a5, a6, a7, a8, a9);
-}
-template <typename R, typename F, typename A1, typename A2, typename A3,
-    typename A4, typename A5, typename A6, typename A7, typename A8,
-    typename A9, typename A10>
-R InvokeArgumentAdl(AdlTag, F f, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6,
-    A7 a7, A8 a8, A9 a9, A10 a10) {
-  return f(a1, a2, a3, a4, a5, a6, a7, a8, a9, a10);
-}
-}  // namespace invoke_argument
-}  // namespace internal
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_0_VALUE_PARAMS()) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args));
-}
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_1_VALUE_PARAMS(p0)) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args), p0);
-}
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_2_VALUE_PARAMS(p0, p1)) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args), p0, p1);
-}
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_3_VALUE_PARAMS(p0, p1, p2)) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args), p0, p1, p2);
-}
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_4_VALUE_PARAMS(p0, p1, p2, p3)) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args), p0, p1, p2, p3);
-}
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_5_VALUE_PARAMS(p0, p1, p2, p3, p4)) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args), p0, p1, p2, p3, p4);
-}
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_6_VALUE_PARAMS(p0, p1, p2, p3, p4, p5)) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args), p0, p1, p2, p3, p4, p5);
-}
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_7_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6)) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args), p0, p1, p2, p3, p4, p5, p6);
-}
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_8_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, p7)) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args), p0, p1, p2, p3, p4, p5, p6, p7);
-}
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_9_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, p7, p8)) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args), p0, p1, p2, p3, p4, p5, p6, p7, p8);
-}
-
-ACTION_TEMPLATE(InvokeArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_10_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9)) {
-  using internal::invoke_argument::InvokeArgumentAdl;
-  return InvokeArgumentAdl<return_type>(
-      internal::invoke_argument::AdlTag(),
-      ::testing::get<k>(args), p0, p1, p2, p3, p4, p5, p6, p7, p8, p9);
-}
-
-// Various overloads for ReturnNew<T>().
-//
-// The ReturnNew<T>(a1, a2, ..., a_k) action returns a pointer to a new
-// instance of type T, constructed on the heap with constructor arguments
-// a1, a2, ..., and a_k. The caller assumes ownership of the returned value.
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_0_VALUE_PARAMS()) {
-  return new T();
-}
-
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_1_VALUE_PARAMS(p0)) {
-  return new T(p0);
-}
-
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_2_VALUE_PARAMS(p0, p1)) {
-  return new T(p0, p1);
-}
-
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_3_VALUE_PARAMS(p0, p1, p2)) {
-  return new T(p0, p1, p2);
-}
-
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_4_VALUE_PARAMS(p0, p1, p2, p3)) {
-  return new T(p0, p1, p2, p3);
-}
-
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_5_VALUE_PARAMS(p0, p1, p2, p3, p4)) {
-  return new T(p0, p1, p2, p3, p4);
-}
-
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_6_VALUE_PARAMS(p0, p1, p2, p3, p4, p5)) {
-  return new T(p0, p1, p2, p3, p4, p5);
-}
-
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_7_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6)) {
-  return new T(p0, p1, p2, p3, p4, p5, p6);
-}
-
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_8_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, p7)) {
-  return new T(p0, p1, p2, p3, p4, p5, p6, p7);
-}
-
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_9_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, p7, p8)) {
-  return new T(p0, p1, p2, p3, p4, p5, p6, p7, p8);
-}
-
-ACTION_TEMPLATE(ReturnNew,
-                HAS_1_TEMPLATE_PARAMS(typename, T),
-                AND_10_VALUE_PARAMS(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9)) {
-  return new T(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9);
-}
-
-#ifdef _MSC_VER
-# pragma warning(pop)
-#endif
-
-}  // namespace testing
-
-// Include any custom callback actions added by the local installation.
-// We must include this header at the end to make sure it can use the
-// declarations from this file.
-// This file was GENERATED by command:
-//     pump.py gmock-generated-actions.h.pump
-// DO NOT EDIT BY HAND!!!
-
-#ifndef GMOCK_INCLUDE_GMOCK_INTERNAL_CUSTOM_GMOCK_GENERATED_ACTIONS_H_
-#define GMOCK_INCLUDE_GMOCK_INTERNAL_CUSTOM_GMOCK_GENERATED_ACTIONS_H_
-
-#endif  // GMOCK_INCLUDE_GMOCK_INTERNAL_CUSTOM_GMOCK_GENERATED_ACTIONS_H_
-
-#endif  // GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_ACTIONS_H_
-// This file was GENERATED by command:
-//     pump.py gmock-generated-function-mockers.h.pump
-// DO NOT EDIT BY HAND!!!
-
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements function mockers of various arities.
-
-#ifndef GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_FUNCTION_MOCKERS_H_
-#define GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_FUNCTION_MOCKERS_H_
-
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements the ON_CALL() and EXPECT_CALL() macros.
-//
-// A user can use the ON_CALL() macro to specify the default action of
-// a mock method.  The syntax is:
-//
-//   ON_CALL(mock_object, Method(argument-matchers))
-//       .With(multi-argument-matcher)
-//       .WillByDefault(action);
-//
-//  where the .With() clause is optional.
-//
-// A user can use the EXPECT_CALL() macro to specify an expectation on
-// a mock method.  The syntax is:
-//
-//   EXPECT_CALL(mock_object, Method(argument-matchers))
-//       .With(multi-argument-matchers)
-//       .Times(cardinality)
-//       .InSequence(sequences)
-//       .After(expectations)
-//       .WillOnce(action)
-//       .WillRepeatedly(action)
-//       .RetiresOnSaturation();
-//
-// where all clauses are optional, and .InSequence()/.After()/
-// .WillOnce() can appear any number of times.
-
-#ifndef GMOCK_INCLUDE_GMOCK_GMOCK_SPEC_BUILDERS_H_
-#define GMOCK_INCLUDE_GMOCK_GMOCK_SPEC_BUILDERS_H_
-
-#include <map>
-#include <set>
-#include <sstream>
-#include <string>
-#include <vector>
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements some commonly used argument matchers.  More
-// matchers can be defined by the user implementing the
-// MatcherInterface<T> interface if necessary.
-
-#ifndef GMOCK_INCLUDE_GMOCK_GMOCK_MATCHERS_H_
-#define GMOCK_INCLUDE_GMOCK_GMOCK_MATCHERS_H_
-
-#include <math.h>
-#include <algorithm>
-#include <iterator>
-#include <limits>
-#include <ostream>  // NOLINT
-#include <sstream>
-#include <string>
-#include <utility>
-#include <vector>
-
-#if GTEST_HAS_STD_INITIALIZER_LIST_
-# include <initializer_list>  // NOLINT -- must be after gtest.h
-#endif
-
-namespace testing {
-
-// To implement a matcher Foo for type T, define:
-//   1. a class FooMatcherImpl that implements the
-//      MatcherInterface<T> interface, and
-//   2. a factory function that creates a Matcher<T> object from a
-//      FooMatcherImpl*.
-//
-// The two-level delegation design makes it possible to allow a user
-// to write "v" instead of "Eq(v)" where a Matcher is expected, which
-// is impossible if we pass matchers by pointers.  It also eases
-// ownership management as Matcher objects can now be copied like
-// plain values.
-
-// MatchResultListener is an abstract class.  Its << operator can be
-// used by a matcher to explain why a value matches or doesn't match.
-//
-// TODO(wan@google.com): add method
-//   bool InterestedInWhy(bool result) const;
-// to indicate whether the listener is interested in why the match
-// result is 'result'.
-class MatchResultListener {
- public:
-  // Creates a listener object with the given underlying ostream.  The
-  // listener does not own the ostream, and does not dereference it
-  // in the constructor or destructor.
-  explicit MatchResultListener(::std::ostream* os) : stream_(os) {}
-  virtual ~MatchResultListener() = 0;  // Makes this class abstract.
-
-  // Streams x to the underlying ostream; does nothing if the ostream
-  // is NULL.
-  template <typename T>
-  MatchResultListener& operator<<(const T& x) {
-    if (stream_ != NULL)
-      *stream_ << x;
-    return *this;
-  }
-
-  // Returns the underlying ostream.
-  ::std::ostream* stream() { return stream_; }
-
-  // Returns true iff the listener is interested in an explanation of
-  // the match result.  A matcher's MatchAndExplain() method can use
-  // this information to avoid generating the explanation when no one
-  // intends to hear it.
-  bool IsInterested() const { return stream_ != NULL; }
-
- private:
-  ::std::ostream* const stream_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MatchResultListener);
-};
-
-inline MatchResultListener::~MatchResultListener() {
-}
-
-// An instance of a subclass of this knows how to describe itself as a
-// matcher.
-class MatcherDescriberInterface {
- public:
-  virtual ~MatcherDescriberInterface() {}
-
-  // Describes this matcher to an ostream.  The function should print
-  // a verb phrase that describes the property a value matching this
-  // matcher should have.  The subject of the verb phrase is the value
-  // being matched.  For example, the DescribeTo() method of the Gt(7)
-  // matcher prints "is greater than 7".
-  virtual void DescribeTo(::std::ostream* os) const = 0;
-
-  // Describes the negation of this matcher to an ostream.  For
-  // example, if the description of this matcher is "is greater than
-  // 7", the negated description could be "is not greater than 7".
-  // You are not required to override this when implementing
-  // MatcherInterface, but it is highly advised so that your matcher
-  // can produce good error messages.
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    *os << "not (";
-    DescribeTo(os);
-    *os << ")";
-  }
-};
-
-// The implementation of a matcher.
-template <typename T>
-class MatcherInterface : public MatcherDescriberInterface {
- public:
-  // Returns true iff the matcher matches x; also explains the match
-  // result to 'listener' if necessary (see the next paragraph), in
-  // the form of a non-restrictive relative clause ("which ...",
-  // "whose ...", etc) that describes x.  For example, the
-  // MatchAndExplain() method of the Pointee(...) matcher should
-  // generate an explanation like "which points to ...".
-  //
-  // Implementations of MatchAndExplain() should add an explanation of
-  // the match result *if and only if* they can provide additional
-  // information that's not already present (or not obvious) in the
-  // print-out of x and the matcher's description.  Whether the match
-  // succeeds is not a factor in deciding whether an explanation is
-  // needed, as sometimes the caller needs to print a failure message
-  // when the match succeeds (e.g. when the matcher is used inside
-  // Not()).
-  //
-  // For example, a "has at least 10 elements" matcher should explain
-  // what the actual element count is, regardless of the match result,
-  // as it is useful information to the reader; on the other hand, an
-  // "is empty" matcher probably only needs to explain what the actual
-  // size is when the match fails, as it's redundant to say that the
-  // size is 0 when the value is already known to be empty.
-  //
-  // You should override this method when defining a new matcher.
-  //
-  // It's the responsibility of the caller (Google Mock) to guarantee
-  // that 'listener' is not NULL.  This helps to simplify a matcher's
-  // implementation when it doesn't care about the performance, as it
-  // can talk to 'listener' without checking its validity first.
-  // However, in order to implement dummy listeners efficiently,
-  // listener->stream() may be NULL.
-  virtual bool MatchAndExplain(T x, MatchResultListener* listener) const = 0;
-
-  // Inherits these methods from MatcherDescriberInterface:
-  //   virtual void DescribeTo(::std::ostream* os) const = 0;
-  //   virtual void DescribeNegationTo(::std::ostream* os) const;
-};
-
-namespace internal {
-
-// Converts a MatcherInterface<T> to a MatcherInterface<const T&>.
-template <typename T>
-class MatcherInterfaceAdapter : public MatcherInterface<const T&> {
- public:
-  explicit MatcherInterfaceAdapter(const MatcherInterface<T>* impl)
-      : impl_(impl) {}
-  virtual ~MatcherInterfaceAdapter() { delete impl_; }
-
-  virtual void DescribeTo(::std::ostream* os) const { impl_->DescribeTo(os); }
-
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    impl_->DescribeNegationTo(os);
-  }
-
-  virtual bool MatchAndExplain(const T& x,
-                               MatchResultListener* listener) const {
-    return impl_->MatchAndExplain(x, listener);
-  }
-
- private:
-  const MatcherInterface<T>* const impl_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MatcherInterfaceAdapter);
-};
-
-}  // namespace internal
-
-// A match result listener that stores the explanation in a string.
-class StringMatchResultListener : public MatchResultListener {
- public:
-  StringMatchResultListener() : MatchResultListener(&ss_) {}
-
-  // Returns the explanation accumulated so far.
-  std::string str() const { return ss_.str(); }
-
-  // Clears the explanation accumulated so far.
-  void Clear() { ss_.str(""); }
-
- private:
-  ::std::stringstream ss_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(StringMatchResultListener);
-};
-
-namespace internal {
-
-struct AnyEq {
-  template <typename A, typename B>
-  bool operator()(const A& a, const B& b) const { return a == b; }
-};
-struct AnyNe {
-  template <typename A, typename B>
-  bool operator()(const A& a, const B& b) const { return a != b; }
-};
-struct AnyLt {
-  template <typename A, typename B>
-  bool operator()(const A& a, const B& b) const { return a < b; }
-};
-struct AnyGt {
-  template <typename A, typename B>
-  bool operator()(const A& a, const B& b) const { return a > b; }
-};
-struct AnyLe {
-  template <typename A, typename B>
-  bool operator()(const A& a, const B& b) const { return a <= b; }
-};
-struct AnyGe {
-  template <typename A, typename B>
-  bool operator()(const A& a, const B& b) const { return a >= b; }
-};
-
-// A match result listener that ignores the explanation.
-class DummyMatchResultListener : public MatchResultListener {
- public:
-  DummyMatchResultListener() : MatchResultListener(NULL) {}
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(DummyMatchResultListener);
-};
-
-// A match result listener that forwards the explanation to a given
-// ostream.  The difference between this and MatchResultListener is
-// that the former is concrete.
-class StreamMatchResultListener : public MatchResultListener {
- public:
-  explicit StreamMatchResultListener(::std::ostream* os)
-      : MatchResultListener(os) {}
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(StreamMatchResultListener);
-};
-
-// An internal class for implementing Matcher<T>, which will derive
-// from it.  We put functionalities common to all Matcher<T>
-// specializations here to avoid code duplication.
-template <typename T>
-class MatcherBase {
- public:
-  // Returns true iff the matcher matches x; also explains the match
-  // result to 'listener'.
-  bool MatchAndExplain(GTEST_REFERENCE_TO_CONST_(T) x,
-                       MatchResultListener* listener) const {
-    return impl_->MatchAndExplain(x, listener);
-  }
-
-  // Returns true iff this matcher matches x.
-  bool Matches(GTEST_REFERENCE_TO_CONST_(T) x) const {
-    DummyMatchResultListener dummy;
-    return MatchAndExplain(x, &dummy);
-  }
-
-  // Describes this matcher to an ostream.
-  void DescribeTo(::std::ostream* os) const { impl_->DescribeTo(os); }
-
-  // Describes the negation of this matcher to an ostream.
-  void DescribeNegationTo(::std::ostream* os) const {
-    impl_->DescribeNegationTo(os);
-  }
-
-  // Explains why x matches, or doesn't match, the matcher.
-  void ExplainMatchResultTo(GTEST_REFERENCE_TO_CONST_(T) x,
-                            ::std::ostream* os) const {
-    StreamMatchResultListener listener(os);
-    MatchAndExplain(x, &listener);
-  }
-
-  // Returns the describer for this matcher object; retains ownership
-  // of the describer, which is only guaranteed to be alive when
-  // this matcher object is alive.
-  const MatcherDescriberInterface* GetDescriber() const {
-    return impl_.get();
-  }
-
- protected:
-  MatcherBase() {}
-
-  // Constructs a matcher from its implementation.
-  explicit MatcherBase(
-      const MatcherInterface<GTEST_REFERENCE_TO_CONST_(T)>* impl)
-      : impl_(impl) {}
-
-  template <typename U>
-  explicit MatcherBase(
-      const MatcherInterface<U>* impl,
-      typename internal::EnableIf<
-          !internal::IsSame<U, GTEST_REFERENCE_TO_CONST_(U)>::value>::type* =
-          NULL)
-      : impl_(new internal::MatcherInterfaceAdapter<U>(impl)) {}
-
-  virtual ~MatcherBase() {}
-
- private:
-  // shared_ptr (util/gtl/shared_ptr.h) and linked_ptr have similar
-  // interfaces.  The former dynamically allocates a chunk of memory
-  // to hold the reference count, while the latter tracks all
-  // references using a circular linked list without allocating
-  // memory.  It has been observed that linked_ptr performs better in
-  // typical scenarios.  However, shared_ptr can out-perform
-  // linked_ptr when there are many more uses of the copy constructor
-  // than the default constructor.
-  //
-  // If performance becomes a problem, we should see if using
-  // shared_ptr helps.
-  ::testing::internal::linked_ptr<
-      const MatcherInterface<GTEST_REFERENCE_TO_CONST_(T)> >
-      impl_;
-};
-
-}  // namespace internal
-
-// A Matcher<T> is a copyable and IMMUTABLE (except by assignment)
-// object that can check whether a value of type T matches.  The
-// implementation of Matcher<T> is just a linked_ptr to const
-// MatcherInterface<T>, so copying is fairly cheap.  Don't inherit
-// from Matcher!
-template <typename T>
-class Matcher : public internal::MatcherBase<T> {
- public:
-  // Constructs a null matcher.  Needed for storing Matcher objects in STL
-  // containers.  A default-constructed matcher is not yet initialized.  You
-  // cannot use it until a valid value has been assigned to it.
-  explicit Matcher() {}  // NOLINT
-
-  // Constructs a matcher from its implementation.
-  explicit Matcher(const MatcherInterface<GTEST_REFERENCE_TO_CONST_(T)>* impl)
-      : internal::MatcherBase<T>(impl) {}
-
-  template <typename U>
-  explicit Matcher(const MatcherInterface<U>* impl,
-                   typename internal::EnableIf<!internal::IsSame<
-                       U, GTEST_REFERENCE_TO_CONST_(U)>::value>::type* = NULL)
-      : internal::MatcherBase<T>(impl) {}
-
-  // Implicit constructor here allows people to write
-  // EXPECT_CALL(foo, Bar(5)) instead of EXPECT_CALL(foo, Bar(Eq(5))) sometimes
-  Matcher(T value);  // NOLINT
-};
-
-// The following two specializations allow the user to write str
-// instead of Eq(str) and "foo" instead of Eq("foo") when a std::string
-// matcher is expected.
-template <>
-class GTEST_API_ Matcher<const std::string&>
-    : public internal::MatcherBase<const std::string&> {
- public:
-  Matcher() {}
-
-  explicit Matcher(const MatcherInterface<const std::string&>* impl)
-      : internal::MatcherBase<const std::string&>(impl) {}
-
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a std::string object.
-  Matcher(const std::string& s);  // NOLINT
-
-#if GTEST_HAS_GLOBAL_STRING
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a ::string object.
-  Matcher(const ::string& s);  // NOLINT
-#endif                         // GTEST_HAS_GLOBAL_STRING
-
-  // Allows the user to write "foo" instead of Eq("foo") sometimes.
-  Matcher(const char* s);  // NOLINT
-};
-
-template <>
-class GTEST_API_ Matcher<std::string>
-    : public internal::MatcherBase<std::string> {
- public:
-  Matcher() {}
-
-  explicit Matcher(const MatcherInterface<const std::string&>* impl)
-      : internal::MatcherBase<std::string>(impl) {}
-  explicit Matcher(const MatcherInterface<std::string>* impl)
-      : internal::MatcherBase<std::string>(impl) {}
-
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a string object.
-  Matcher(const std::string& s);  // NOLINT
-
-#if GTEST_HAS_GLOBAL_STRING
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a ::string object.
-  Matcher(const ::string& s);  // NOLINT
-#endif                         // GTEST_HAS_GLOBAL_STRING
-
-  // Allows the user to write "foo" instead of Eq("foo") sometimes.
-  Matcher(const char* s);  // NOLINT
-};
-
-#if GTEST_HAS_GLOBAL_STRING
-// The following two specializations allow the user to write str
-// instead of Eq(str) and "foo" instead of Eq("foo") when a ::string
-// matcher is expected.
-template <>
-class GTEST_API_ Matcher<const ::string&>
-    : public internal::MatcherBase<const ::string&> {
- public:
-  Matcher() {}
-
-  explicit Matcher(const MatcherInterface<const ::string&>* impl)
-      : internal::MatcherBase<const ::string&>(impl) {}
-
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a std::string object.
-  Matcher(const std::string& s);  // NOLINT
-
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a ::string object.
-  Matcher(const ::string& s);  // NOLINT
-
-  // Allows the user to write "foo" instead of Eq("foo") sometimes.
-  Matcher(const char* s);  // NOLINT
-};
-
-template <>
-class GTEST_API_ Matcher< ::string>
-    : public internal::MatcherBase< ::string> {
- public:
-  Matcher() {}
-
-  explicit Matcher(const MatcherInterface<const ::string&>* impl)
-      : internal::MatcherBase< ::string>(impl) {}
-  explicit Matcher(const MatcherInterface< ::string>* impl)
-      : internal::MatcherBase< ::string>(impl) {}
-
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a std::string object.
-  Matcher(const std::string& s);  // NOLINT
-
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a ::string object.
-  Matcher(const ::string& s);  // NOLINT
-
-  // Allows the user to write "foo" instead of Eq("foo") sometimes.
-  Matcher(const char* s);  // NOLINT
-};
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-#if GTEST_HAS_ABSL
-// The following two specializations allow the user to write str
-// instead of Eq(str) and "foo" instead of Eq("foo") when a absl::string_view
-// matcher is expected.
-template <>
-class GTEST_API_ Matcher<const absl::string_view&>
-    : public internal::MatcherBase<const absl::string_view&> {
- public:
-  Matcher() {}
-
-  explicit Matcher(const MatcherInterface<const absl::string_view&>* impl)
-      : internal::MatcherBase<const absl::string_view&>(impl) {}
-
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a std::string object.
-  Matcher(const std::string& s);  // NOLINT
-
-#if GTEST_HAS_GLOBAL_STRING
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a ::string object.
-  Matcher(const ::string& s);  // NOLINT
-#endif                         // GTEST_HAS_GLOBAL_STRING
-
-  // Allows the user to write "foo" instead of Eq("foo") sometimes.
-  Matcher(const char* s);  // NOLINT
-
-  // Allows the user to pass absl::string_views directly.
-  Matcher(absl::string_view s);  // NOLINT
-};
-
-template <>
-class GTEST_API_ Matcher<absl::string_view>
-    : public internal::MatcherBase<absl::string_view> {
- public:
-  Matcher() {}
-
-  explicit Matcher(const MatcherInterface<const absl::string_view&>* impl)
-      : internal::MatcherBase<absl::string_view>(impl) {}
-  explicit Matcher(const MatcherInterface<absl::string_view>* impl)
-      : internal::MatcherBase<absl::string_view>(impl) {}
-
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a std::string object.
-  Matcher(const std::string& s);  // NOLINT
-
-#if GTEST_HAS_GLOBAL_STRING
-  // Allows the user to write str instead of Eq(str) sometimes, where
-  // str is a ::string object.
-  Matcher(const ::string& s);  // NOLINT
-#endif                         // GTEST_HAS_GLOBAL_STRING
-
-  // Allows the user to write "foo" instead of Eq("foo") sometimes.
-  Matcher(const char* s);  // NOLINT
-
-  // Allows the user to pass absl::string_views directly.
-  Matcher(absl::string_view s);  // NOLINT
-};
-#endif  // GTEST_HAS_ABSL
-
-// Prints a matcher in a human-readable format.
-template <typename T>
-std::ostream& operator<<(std::ostream& os, const Matcher<T>& matcher) {
-  matcher.DescribeTo(&os);
-  return os;
-}
-
-// The PolymorphicMatcher class template makes it easy to implement a
-// polymorphic matcher (i.e. a matcher that can match values of more
-// than one type, e.g. Eq(n) and NotNull()).
-//
-// To define a polymorphic matcher, a user should provide an Impl
-// class that has a DescribeTo() method and a DescribeNegationTo()
-// method, and define a member function (or member function template)
-//
-//   bool MatchAndExplain(const Value& value,
-//                        MatchResultListener* listener) const;
-//
-// See the definition of NotNull() for a complete example.
-template <class Impl>
-class PolymorphicMatcher {
- public:
-  explicit PolymorphicMatcher(const Impl& an_impl) : impl_(an_impl) {}
-
-  // Returns a mutable reference to the underlying matcher
-  // implementation object.
-  Impl& mutable_impl() { return impl_; }
-
-  // Returns an immutable reference to the underlying matcher
-  // implementation object.
-  const Impl& impl() const { return impl_; }
-
-  template <typename T>
-  operator Matcher<T>() const {
-    return Matcher<T>(new MonomorphicImpl<GTEST_REFERENCE_TO_CONST_(T)>(impl_));
-  }
-
- private:
-  template <typename T>
-  class MonomorphicImpl : public MatcherInterface<T> {
-   public:
-    explicit MonomorphicImpl(const Impl& impl) : impl_(impl) {}
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      impl_.DescribeTo(os);
-    }
-
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      impl_.DescribeNegationTo(os);
-    }
-
-    virtual bool MatchAndExplain(T x, MatchResultListener* listener) const {
-      return impl_.MatchAndExplain(x, listener);
-    }
-
-   private:
-    const Impl impl_;
-
-    GTEST_DISALLOW_ASSIGN_(MonomorphicImpl);
-  };
-
-  Impl impl_;
-
-  GTEST_DISALLOW_ASSIGN_(PolymorphicMatcher);
-};
-
-// Creates a matcher from its implementation.  This is easier to use
-// than the Matcher<T> constructor as it doesn't require you to
-// explicitly write the template argument, e.g.
-//
-//   MakeMatcher(foo);
-// vs
-//   Matcher<const string&>(foo);
-template <typename T>
-inline Matcher<T> MakeMatcher(const MatcherInterface<T>* impl) {
-  return Matcher<T>(impl);
-}
-
-// Creates a polymorphic matcher from its implementation.  This is
-// easier to use than the PolymorphicMatcher<Impl> constructor as it
-// doesn't require you to explicitly write the template argument, e.g.
-//
-//   MakePolymorphicMatcher(foo);
-// vs
-//   PolymorphicMatcher<TypeOfFoo>(foo);
-template <class Impl>
-inline PolymorphicMatcher<Impl> MakePolymorphicMatcher(const Impl& impl) {
-  return PolymorphicMatcher<Impl>(impl);
-}
-
-// Anything inside the 'internal' namespace IS INTERNAL IMPLEMENTATION
-// and MUST NOT BE USED IN USER CODE!!!
-namespace internal {
-
-// The MatcherCastImpl class template is a helper for implementing
-// MatcherCast().  We need this helper in order to partially
-// specialize the implementation of MatcherCast() (C++ allows
-// class/struct templates to be partially specialized, but not
-// function templates.).
-
-// This general version is used when MatcherCast()'s argument is a
-// polymorphic matcher (i.e. something that can be converted to a
-// Matcher but is not one yet; for example, Eq(value)) or a value (for
-// example, "hello").
-template <typename T, typename M>
-class MatcherCastImpl {
- public:
-  static Matcher<T> Cast(const M& polymorphic_matcher_or_value) {
-    // M can be a polymorphic matcher, in which case we want to use
-    // its conversion operator to create Matcher<T>.  Or it can be a value
-    // that should be passed to the Matcher<T>'s constructor.
-    //
-    // We can't call Matcher<T>(polymorphic_matcher_or_value) when M is a
-    // polymorphic matcher because it'll be ambiguous if T has an implicit
-    // constructor from M (this usually happens when T has an implicit
-    // constructor from any type).
-    //
-    // It won't work to unconditionally implict_cast
-    // polymorphic_matcher_or_value to Matcher<T> because it won't trigger
-    // a user-defined conversion from M to T if one exists (assuming M is
-    // a value).
-    return CastImpl(
-        polymorphic_matcher_or_value,
-        BooleanConstant<
-            internal::ImplicitlyConvertible<M, Matcher<T> >::value>(),
-        BooleanConstant<
-            internal::ImplicitlyConvertible<M, T>::value>());
-  }
-
- private:
-  template <bool Ignore>
-  static Matcher<T> CastImpl(const M& polymorphic_matcher_or_value,
-                             BooleanConstant<true> /* convertible_to_matcher */,
-                             BooleanConstant<Ignore>) {
-    // M is implicitly convertible to Matcher<T>, which means that either
-    // M is a polymorphic matcher or Matcher<T> has an implicit constructor
-    // from M.  In both cases using the implicit conversion will produce a
-    // matcher.
-    //
-    // Even if T has an implicit constructor from M, it won't be called because
-    // creating Matcher<T> would require a chain of two user-defined conversions
-    // (first to create T from M and then to create Matcher<T> from T).
-    return polymorphic_matcher_or_value;
-  }
-
-  // M can't be implicitly converted to Matcher<T>, so M isn't a polymorphic
-  // matcher. It's a value of a type implicitly convertible to T. Use direct
-  // initialization to create a matcher.
-  static Matcher<T> CastImpl(
-      const M& value, BooleanConstant<false> /* convertible_to_matcher */,
-      BooleanConstant<true> /* convertible_to_T */) {
-    return Matcher<T>(ImplicitCast_<T>(value));
-  }
-
-  // M can't be implicitly converted to either Matcher<T> or T. Attempt to use
-  // polymorphic matcher Eq(value) in this case.
-  //
-  // Note that we first attempt to perform an implicit cast on the value and
-  // only fall back to the polymorphic Eq() matcher afterwards because the
-  // latter calls bool operator==(const Lhs& lhs, const Rhs& rhs) in the end
-  // which might be undefined even when Rhs is implicitly convertible to Lhs
-  // (e.g. std::pair<const int, int> vs. std::pair<int, int>).
-  //
-  // We don't define this method inline as we need the declaration of Eq().
-  static Matcher<T> CastImpl(
-      const M& value, BooleanConstant<false> /* convertible_to_matcher */,
-      BooleanConstant<false> /* convertible_to_T */);
-};
-
-// This more specialized version is used when MatcherCast()'s argument
-// is already a Matcher.  This only compiles when type T can be
-// statically converted to type U.
-template <typename T, typename U>
-class MatcherCastImpl<T, Matcher<U> > {
- public:
-  static Matcher<T> Cast(const Matcher<U>& source_matcher) {
-    return Matcher<T>(new Impl(source_matcher));
-  }
-
- private:
-  class Impl : public MatcherInterface<T> {
-   public:
-    explicit Impl(const Matcher<U>& source_matcher)
-        : source_matcher_(source_matcher) {}
-
-    // We delegate the matching logic to the source matcher.
-    virtual bool MatchAndExplain(T x, MatchResultListener* listener) const {
-#if GTEST_LANG_CXX11
-      using FromType = typename std::remove_cv<typename std::remove_pointer<
-          typename std::remove_reference<T>::type>::type>::type;
-      using ToType = typename std::remove_cv<typename std::remove_pointer<
-          typename std::remove_reference<U>::type>::type>::type;
-      // Do not allow implicitly converting base*/& to derived*/&.
-      static_assert(
-          // Do not trigger if only one of them is a pointer. That implies a
-          // regular conversion and not a down_cast.
-          (std::is_pointer<typename std::remove_reference<T>::type>::value !=
-           std::is_pointer<typename std::remove_reference<U>::type>::value) ||
-              std::is_same<FromType, ToType>::value ||
-              !std::is_base_of<FromType, ToType>::value,
-          "Can't implicitly convert from <base> to <derived>");
-#endif  // GTEST_LANG_CXX11
-
-      return source_matcher_.MatchAndExplain(static_cast<U>(x), listener);
-    }
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      source_matcher_.DescribeTo(os);
-    }
-
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      source_matcher_.DescribeNegationTo(os);
-    }
-
-   private:
-    const Matcher<U> source_matcher_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-};
-
-// This even more specialized version is used for efficiently casting
-// a matcher to its own type.
-template <typename T>
-class MatcherCastImpl<T, Matcher<T> > {
- public:
-  static Matcher<T> Cast(const Matcher<T>& matcher) { return matcher; }
-};
-
-}  // namespace internal
-
-// In order to be safe and clear, casting between different matcher
-// types is done explicitly via MatcherCast<T>(m), which takes a
-// matcher m and returns a Matcher<T>.  It compiles only when T can be
-// statically converted to the argument type of m.
-template <typename T, typename M>
-inline Matcher<T> MatcherCast(const M& matcher) {
-  return internal::MatcherCastImpl<T, M>::Cast(matcher);
-}
-
-// Implements SafeMatcherCast().
-//
-// We use an intermediate class to do the actual safe casting as Nokia's
-// Symbian compiler cannot decide between
-// template <T, M> ... (M) and
-// template <T, U> ... (const Matcher<U>&)
-// for function templates but can for member function templates.
-template <typename T>
-class SafeMatcherCastImpl {
- public:
-  // This overload handles polymorphic matchers and values only since
-  // monomorphic matchers are handled by the next one.
-  template <typename M>
-  static inline Matcher<T> Cast(const M& polymorphic_matcher_or_value) {
-    return internal::MatcherCastImpl<T, M>::Cast(polymorphic_matcher_or_value);
-  }
-
-  // This overload handles monomorphic matchers.
-  //
-  // In general, if type T can be implicitly converted to type U, we can
-  // safely convert a Matcher<U> to a Matcher<T> (i.e. Matcher is
-  // contravariant): just keep a copy of the original Matcher<U>, convert the
-  // argument from type T to U, and then pass it to the underlying Matcher<U>.
-  // The only exception is when U is a reference and T is not, as the
-  // underlying Matcher<U> may be interested in the argument's address, which
-  // is not preserved in the conversion from T to U.
-  template <typename U>
-  static inline Matcher<T> Cast(const Matcher<U>& matcher) {
-    // Enforce that T can be implicitly converted to U.
-    GTEST_COMPILE_ASSERT_((internal::ImplicitlyConvertible<T, U>::value),
-                          T_must_be_implicitly_convertible_to_U);
-    // Enforce that we are not converting a non-reference type T to a reference
-    // type U.
-    GTEST_COMPILE_ASSERT_(
-        internal::is_reference<T>::value || !internal::is_reference<U>::value,
-        cannot_convert_non_reference_arg_to_reference);
-    // In case both T and U are arithmetic types, enforce that the
-    // conversion is not lossy.
-    typedef GTEST_REMOVE_REFERENCE_AND_CONST_(T) RawT;
-    typedef GTEST_REMOVE_REFERENCE_AND_CONST_(U) RawU;
-    const bool kTIsOther = GMOCK_KIND_OF_(RawT) == internal::kOther;
-    const bool kUIsOther = GMOCK_KIND_OF_(RawU) == internal::kOther;
-    GTEST_COMPILE_ASSERT_(
-        kTIsOther || kUIsOther ||
-        (internal::LosslessArithmeticConvertible<RawT, RawU>::value),
-        conversion_of_arithmetic_types_must_be_lossless);
-    return MatcherCast<T>(matcher);
-  }
-};
-
-template <typename T, typename M>
-inline Matcher<T> SafeMatcherCast(const M& polymorphic_matcher) {
-  return SafeMatcherCastImpl<T>::Cast(polymorphic_matcher);
-}
-
-// A<T>() returns a matcher that matches any value of type T.
-template <typename T>
-Matcher<T> A();
-
-// Anything inside the 'internal' namespace IS INTERNAL IMPLEMENTATION
-// and MUST NOT BE USED IN USER CODE!!!
-namespace internal {
-
-// If the explanation is not empty, prints it to the ostream.
-inline void PrintIfNotEmpty(const std::string& explanation,
-                            ::std::ostream* os) {
-  if (explanation != "" && os != NULL) {
-    *os << ", " << explanation;
-  }
-}
-
-// Returns true if the given type name is easy to read by a human.
-// This is used to decide whether printing the type of a value might
-// be helpful.
-inline bool IsReadableTypeName(const std::string& type_name) {
-  // We consider a type name readable if it's short or doesn't contain
-  // a template or function type.
-  return (type_name.length() <= 20 ||
-          type_name.find_first_of("<(") == std::string::npos);
-}
-
-// Matches the value against the given matcher, prints the value and explains
-// the match result to the listener. Returns the match result.
-// 'listener' must not be NULL.
-// Value cannot be passed by const reference, because some matchers take a
-// non-const argument.
-template <typename Value, typename T>
-bool MatchPrintAndExplain(Value& value, const Matcher<T>& matcher,
-                          MatchResultListener* listener) {
-  if (!listener->IsInterested()) {
-    // If the listener is not interested, we do not need to construct the
-    // inner explanation.
-    return matcher.Matches(value);
-  }
-
-  StringMatchResultListener inner_listener;
-  const bool match = matcher.MatchAndExplain(value, &inner_listener);
-
-  UniversalPrint(value, listener->stream());
-#if GTEST_HAS_RTTI
-  const std::string& type_name = GetTypeName<Value>();
-  if (IsReadableTypeName(type_name))
-    *listener->stream() << " (of type " << type_name << ")";
-#endif
-  PrintIfNotEmpty(inner_listener.str(), listener->stream());
-
-  return match;
-}
-
-// An internal helper class for doing compile-time loop on a tuple's
-// fields.
-template <size_t N>
-class TuplePrefix {
- public:
-  // TuplePrefix<N>::Matches(matcher_tuple, value_tuple) returns true
-  // iff the first N fields of matcher_tuple matches the first N
-  // fields of value_tuple, respectively.
-  template <typename MatcherTuple, typename ValueTuple>
-  static bool Matches(const MatcherTuple& matcher_tuple,
-                      const ValueTuple& value_tuple) {
-    return TuplePrefix<N - 1>::Matches(matcher_tuple, value_tuple)
-        && get<N - 1>(matcher_tuple).Matches(get<N - 1>(value_tuple));
-  }
-
-  // TuplePrefix<N>::ExplainMatchFailuresTo(matchers, values, os)
-  // describes failures in matching the first N fields of matchers
-  // against the first N fields of values.  If there is no failure,
-  // nothing will be streamed to os.
-  template <typename MatcherTuple, typename ValueTuple>
-  static void ExplainMatchFailuresTo(const MatcherTuple& matchers,
-                                     const ValueTuple& values,
-                                     ::std::ostream* os) {
-    // First, describes failures in the first N - 1 fields.
-    TuplePrefix<N - 1>::ExplainMatchFailuresTo(matchers, values, os);
-
-    // Then describes the failure (if any) in the (N - 1)-th (0-based)
-    // field.
-    typename tuple_element<N - 1, MatcherTuple>::type matcher =
-        get<N - 1>(matchers);
-    typedef typename tuple_element<N - 1, ValueTuple>::type Value;
-    GTEST_REFERENCE_TO_CONST_(Value) value = get<N - 1>(values);
-    StringMatchResultListener listener;
-    if (!matcher.MatchAndExplain(value, &listener)) {
-      // TODO(wan): include in the message the name of the parameter
-      // as used in MOCK_METHOD*() when possible.
-      *os << "  Expected arg #" << N - 1 << ": ";
-      get<N - 1>(matchers).DescribeTo(os);
-      *os << "\n           Actual: ";
-      // We remove the reference in type Value to prevent the
-      // universal printer from printing the address of value, which
-      // isn't interesting to the user most of the time.  The
-      // matcher's MatchAndExplain() method handles the case when
-      // the address is interesting.
-      internal::UniversalPrint(value, os);
-      PrintIfNotEmpty(listener.str(), os);
-      *os << "\n";
-    }
-  }
-};
-
-// The base case.
-template <>
-class TuplePrefix<0> {
- public:
-  template <typename MatcherTuple, typename ValueTuple>
-  static bool Matches(const MatcherTuple& /* matcher_tuple */,
-                      const ValueTuple& /* value_tuple */) {
-    return true;
-  }
-
-  template <typename MatcherTuple, typename ValueTuple>
-  static void ExplainMatchFailuresTo(const MatcherTuple& /* matchers */,
-                                     const ValueTuple& /* values */,
-                                     ::std::ostream* /* os */) {}
-};
-
-// TupleMatches(matcher_tuple, value_tuple) returns true iff all
-// matchers in matcher_tuple match the corresponding fields in
-// value_tuple.  It is a compiler error if matcher_tuple and
-// value_tuple have different number of fields or incompatible field
-// types.
-template <typename MatcherTuple, typename ValueTuple>
-bool TupleMatches(const MatcherTuple& matcher_tuple,
-                  const ValueTuple& value_tuple) {
-  // Makes sure that matcher_tuple and value_tuple have the same
-  // number of fields.
-  GTEST_COMPILE_ASSERT_(tuple_size<MatcherTuple>::value ==
-                        tuple_size<ValueTuple>::value,
-                        matcher_and_value_have_different_numbers_of_fields);
-  return TuplePrefix<tuple_size<ValueTuple>::value>::
-      Matches(matcher_tuple, value_tuple);
-}
-
-// Describes failures in matching matchers against values.  If there
-// is no failure, nothing will be streamed to os.
-template <typename MatcherTuple, typename ValueTuple>
-void ExplainMatchFailureTupleTo(const MatcherTuple& matchers,
-                                const ValueTuple& values,
-                                ::std::ostream* os) {
-  TuplePrefix<tuple_size<MatcherTuple>::value>::ExplainMatchFailuresTo(
-      matchers, values, os);
-}
-
-// TransformTupleValues and its helper.
-//
-// TransformTupleValuesHelper hides the internal machinery that
-// TransformTupleValues uses to implement a tuple traversal.
-template <typename Tuple, typename Func, typename OutIter>
-class TransformTupleValuesHelper {
- private:
-  typedef ::testing::tuple_size<Tuple> TupleSize;
-
- public:
-  // For each member of tuple 't', taken in order, evaluates '*out++ = f(t)'.
-  // Returns the final value of 'out' in case the caller needs it.
-  static OutIter Run(Func f, const Tuple& t, OutIter out) {
-    return IterateOverTuple<Tuple, TupleSize::value>()(f, t, out);
-  }
-
- private:
-  template <typename Tup, size_t kRemainingSize>
-  struct IterateOverTuple {
-    OutIter operator() (Func f, const Tup& t, OutIter out) const {
-      *out++ = f(::testing::get<TupleSize::value - kRemainingSize>(t));
-      return IterateOverTuple<Tup, kRemainingSize - 1>()(f, t, out);
-    }
-  };
-  template <typename Tup>
-  struct IterateOverTuple<Tup, 0> {
-    OutIter operator() (Func /* f */, const Tup& /* t */, OutIter out) const {
-      return out;
-    }
-  };
-};
-
-// Successively invokes 'f(element)' on each element of the tuple 't',
-// appending each result to the 'out' iterator. Returns the final value
-// of 'out'.
-template <typename Tuple, typename Func, typename OutIter>
-OutIter TransformTupleValues(Func f, const Tuple& t, OutIter out) {
-  return TransformTupleValuesHelper<Tuple, Func, OutIter>::Run(f, t, out);
-}
-
-// Implements A<T>().
-template <typename T>
-class AnyMatcherImpl : public MatcherInterface<GTEST_REFERENCE_TO_CONST_(T)> {
- public:
-  virtual bool MatchAndExplain(GTEST_REFERENCE_TO_CONST_(T) /* x */,
-                               MatchResultListener* /* listener */) const {
-    return true;
-  }
-  virtual void DescribeTo(::std::ostream* os) const { *os << "is anything"; }
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    // This is mostly for completeness' safe, as it's not very useful
-    // to write Not(A<bool>()).  However we cannot completely rule out
-    // such a possibility, and it doesn't hurt to be prepared.
-    *os << "never matches";
-  }
-};
-
-// Implements _, a matcher that matches any value of any
-// type.  This is a polymorphic matcher, so we need a template type
-// conversion operator to make it appearing as a Matcher<T> for any
-// type T.
-class AnythingMatcher {
- public:
-  template <typename T>
-  operator Matcher<T>() const { return A<T>(); }
-};
-
-// Implements a matcher that compares a given value with a
-// pre-supplied value using one of the ==, <=, <, etc, operators.  The
-// two values being compared don't have to have the same type.
-//
-// The matcher defined here is polymorphic (for example, Eq(5) can be
-// used to match an int, a short, a double, etc).  Therefore we use
-// a template type conversion operator in the implementation.
-//
-// The following template definition assumes that the Rhs parameter is
-// a "bare" type (i.e. neither 'const T' nor 'T&').
-template <typename D, typename Rhs, typename Op>
-class ComparisonBase {
- public:
-  explicit ComparisonBase(const Rhs& rhs) : rhs_(rhs) {}
-  template <typename Lhs>
-  operator Matcher<Lhs>() const {
-    return MakeMatcher(new Impl<Lhs>(rhs_));
-  }
-
- private:
-  template <typename Lhs>
-  class Impl : public MatcherInterface<Lhs> {
-   public:
-    explicit Impl(const Rhs& rhs) : rhs_(rhs) {}
-    virtual bool MatchAndExplain(
-        Lhs lhs, MatchResultListener* /* listener */) const {
-      return Op()(lhs, rhs_);
-    }
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << D::Desc() << " ";
-      UniversalPrint(rhs_, os);
-    }
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << D::NegatedDesc() <<  " ";
-      UniversalPrint(rhs_, os);
-    }
-   private:
-    Rhs rhs_;
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-  Rhs rhs_;
-  GTEST_DISALLOW_ASSIGN_(ComparisonBase);
-};
-
-template <typename Rhs>
-class EqMatcher : public ComparisonBase<EqMatcher<Rhs>, Rhs, AnyEq> {
- public:
-  explicit EqMatcher(const Rhs& rhs)
-      : ComparisonBase<EqMatcher<Rhs>, Rhs, AnyEq>(rhs) { }
-  static const char* Desc() { return "is equal to"; }
-  static const char* NegatedDesc() { return "isn't equal to"; }
-};
-template <typename Rhs>
-class NeMatcher : public ComparisonBase<NeMatcher<Rhs>, Rhs, AnyNe> {
- public:
-  explicit NeMatcher(const Rhs& rhs)
-      : ComparisonBase<NeMatcher<Rhs>, Rhs, AnyNe>(rhs) { }
-  static const char* Desc() { return "isn't equal to"; }
-  static const char* NegatedDesc() { return "is equal to"; }
-};
-template <typename Rhs>
-class LtMatcher : public ComparisonBase<LtMatcher<Rhs>, Rhs, AnyLt> {
- public:
-  explicit LtMatcher(const Rhs& rhs)
-      : ComparisonBase<LtMatcher<Rhs>, Rhs, AnyLt>(rhs) { }
-  static const char* Desc() { return "is <"; }
-  static const char* NegatedDesc() { return "isn't <"; }
-};
-template <typename Rhs>
-class GtMatcher : public ComparisonBase<GtMatcher<Rhs>, Rhs, AnyGt> {
- public:
-  explicit GtMatcher(const Rhs& rhs)
-      : ComparisonBase<GtMatcher<Rhs>, Rhs, AnyGt>(rhs) { }
-  static const char* Desc() { return "is >"; }
-  static const char* NegatedDesc() { return "isn't >"; }
-};
-template <typename Rhs>
-class LeMatcher : public ComparisonBase<LeMatcher<Rhs>, Rhs, AnyLe> {
- public:
-  explicit LeMatcher(const Rhs& rhs)
-      : ComparisonBase<LeMatcher<Rhs>, Rhs, AnyLe>(rhs) { }
-  static const char* Desc() { return "is <="; }
-  static const char* NegatedDesc() { return "isn't <="; }
-};
-template <typename Rhs>
-class GeMatcher : public ComparisonBase<GeMatcher<Rhs>, Rhs, AnyGe> {
- public:
-  explicit GeMatcher(const Rhs& rhs)
-      : ComparisonBase<GeMatcher<Rhs>, Rhs, AnyGe>(rhs) { }
-  static const char* Desc() { return "is >="; }
-  static const char* NegatedDesc() { return "isn't >="; }
-};
-
-// Implements the polymorphic IsNull() matcher, which matches any raw or smart
-// pointer that is NULL.
-class IsNullMatcher {
- public:
-  template <typename Pointer>
-  bool MatchAndExplain(const Pointer& p,
-                       MatchResultListener* /* listener */) const {
-#if GTEST_LANG_CXX11
-    return p == nullptr;
-#else  // GTEST_LANG_CXX11
-    return GetRawPointer(p) == NULL;
-#endif  // GTEST_LANG_CXX11
-  }
-
-  void DescribeTo(::std::ostream* os) const { *os << "is NULL"; }
-  void DescribeNegationTo(::std::ostream* os) const {
-    *os << "isn't NULL";
-  }
-};
-
-// Implements the polymorphic NotNull() matcher, which matches any raw or smart
-// pointer that is not NULL.
-class NotNullMatcher {
- public:
-  template <typename Pointer>
-  bool MatchAndExplain(const Pointer& p,
-                       MatchResultListener* /* listener */) const {
-#if GTEST_LANG_CXX11
-    return p != nullptr;
-#else  // GTEST_LANG_CXX11
-    return GetRawPointer(p) != NULL;
-#endif  // GTEST_LANG_CXX11
-  }
-
-  void DescribeTo(::std::ostream* os) const { *os << "isn't NULL"; }
-  void DescribeNegationTo(::std::ostream* os) const {
-    *os << "is NULL";
-  }
-};
-
-// Ref(variable) matches any argument that is a reference to
-// 'variable'.  This matcher is polymorphic as it can match any
-// super type of the type of 'variable'.
-//
-// The RefMatcher template class implements Ref(variable).  It can
-// only be instantiated with a reference type.  This prevents a user
-// from mistakenly using Ref(x) to match a non-reference function
-// argument.  For example, the following will righteously cause a
-// compiler error:
-//
-//   int n;
-//   Matcher<int> m1 = Ref(n);   // This won't compile.
-//   Matcher<int&> m2 = Ref(n);  // This will compile.
-template <typename T>
-class RefMatcher;
-
-template <typename T>
-class RefMatcher<T&> {
-  // Google Mock is a generic framework and thus needs to support
-  // mocking any function types, including those that take non-const
-  // reference arguments.  Therefore the template parameter T (and
-  // Super below) can be instantiated to either a const type or a
-  // non-const type.
- public:
-  // RefMatcher() takes a T& instead of const T&, as we want the
-  // compiler to catch using Ref(const_value) as a matcher for a
-  // non-const reference.
-  explicit RefMatcher(T& x) : object_(x) {}  // NOLINT
-
-  template <typename Super>
-  operator Matcher<Super&>() const {
-    // By passing object_ (type T&) to Impl(), which expects a Super&,
-    // we make sure that Super is a super type of T.  In particular,
-    // this catches using Ref(const_value) as a matcher for a
-    // non-const reference, as you cannot implicitly convert a const
-    // reference to a non-const reference.
-    return MakeMatcher(new Impl<Super>(object_));
-  }
-
- private:
-  template <typename Super>
-  class Impl : public MatcherInterface<Super&> {
-   public:
-    explicit Impl(Super& x) : object_(x) {}  // NOLINT
-
-    // MatchAndExplain() takes a Super& (as opposed to const Super&)
-    // in order to match the interface MatcherInterface<Super&>.
-    virtual bool MatchAndExplain(
-        Super& x, MatchResultListener* listener) const {
-      *listener << "which is located @" << static_cast<const void*>(&x);
-      return &x == &object_;
-    }
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "references the variable ";
-      UniversalPrinter<Super&>::Print(object_, os);
-    }
-
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << "does not reference the variable ";
-      UniversalPrinter<Super&>::Print(object_, os);
-    }
-
-   private:
-    const Super& object_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
-  T& object_;
-
-  GTEST_DISALLOW_ASSIGN_(RefMatcher);
-};
-
-// Polymorphic helper functions for narrow and wide string matchers.
-inline bool CaseInsensitiveCStringEquals(const char* lhs, const char* rhs) {
-  return String::CaseInsensitiveCStringEquals(lhs, rhs);
-}
-
-inline bool CaseInsensitiveCStringEquals(const wchar_t* lhs,
-                                         const wchar_t* rhs) {
-  return String::CaseInsensitiveWideCStringEquals(lhs, rhs);
-}
-
-// String comparison for narrow or wide strings that can have embedded NUL
-// characters.
-template <typename StringType>
-bool CaseInsensitiveStringEquals(const StringType& s1,
-                                 const StringType& s2) {
-  // Are the heads equal?
-  if (!CaseInsensitiveCStringEquals(s1.c_str(), s2.c_str())) {
-    return false;
-  }
-
-  // Skip the equal heads.
-  const typename StringType::value_type nul = 0;
-  const size_t i1 = s1.find(nul), i2 = s2.find(nul);
-
-  // Are we at the end of either s1 or s2?
-  if (i1 == StringType::npos || i2 == StringType::npos) {
-    return i1 == i2;
-  }
-
-  // Are the tails equal?
-  return CaseInsensitiveStringEquals(s1.substr(i1 + 1), s2.substr(i2 + 1));
-}
-
-// String matchers.
-
-// Implements equality-based string matchers like StrEq, StrCaseNe, and etc.
-template <typename StringType>
-class StrEqualityMatcher {
- public:
-  StrEqualityMatcher(const StringType& str, bool expect_eq,
-                     bool case_sensitive)
-      : string_(str), expect_eq_(expect_eq), case_sensitive_(case_sensitive) {}
-
-#if GTEST_HAS_ABSL
-  bool MatchAndExplain(const absl::string_view& s,
-                       MatchResultListener* listener) const {
-    if (s.data() == NULL) {
-      return !expect_eq_;
-    }
-    // This should fail to compile if absl::string_view is used with wide
-    // strings.
-    const StringType& str = string(s);
-    return MatchAndExplain(str, listener);
-  }
-#endif  // GTEST_HAS_ABSL
-
-  // Accepts pointer types, particularly:
-  //   const char*
-  //   char*
-  //   const wchar_t*
-  //   wchar_t*
-  template <typename CharType>
-  bool MatchAndExplain(CharType* s, MatchResultListener* listener) const {
-    if (s == NULL) {
-      return !expect_eq_;
-    }
-    return MatchAndExplain(StringType(s), listener);
-  }
-
-  // Matches anything that can convert to StringType.
-  //
-  // This is a template, not just a plain function with const StringType&,
-  // because absl::string_view has some interfering non-explicit constructors.
-  template <typename MatcheeStringType>
-  bool MatchAndExplain(const MatcheeStringType& s,
-                       MatchResultListener* /* listener */) const {
-    const StringType& s2(s);
-    const bool eq = case_sensitive_ ? s2 == string_ :
-        CaseInsensitiveStringEquals(s2, string_);
-    return expect_eq_ == eq;
-  }
-
-  void DescribeTo(::std::ostream* os) const {
-    DescribeToHelper(expect_eq_, os);
-  }
-
-  void DescribeNegationTo(::std::ostream* os) const {
-    DescribeToHelper(!expect_eq_, os);
-  }
-
- private:
-  void DescribeToHelper(bool expect_eq, ::std::ostream* os) const {
-    *os << (expect_eq ? "is " : "isn't ");
-    *os << "equal to ";
-    if (!case_sensitive_) {
-      *os << "(ignoring case) ";
-    }
-    UniversalPrint(string_, os);
-  }
-
-  const StringType string_;
-  const bool expect_eq_;
-  const bool case_sensitive_;
-
-  GTEST_DISALLOW_ASSIGN_(StrEqualityMatcher);
-};
-
-// Implements the polymorphic HasSubstr(substring) matcher, which
-// can be used as a Matcher<T> as long as T can be converted to a
-// string.
-template <typename StringType>
-class HasSubstrMatcher {
- public:
-  explicit HasSubstrMatcher(const StringType& substring)
-      : substring_(substring) {}
-
-#if GTEST_HAS_ABSL
-  bool MatchAndExplain(const absl::string_view& s,
-                       MatchResultListener* listener) const {
-    if (s.data() == NULL) {
-      return false;
-    }
-    // This should fail to compile if absl::string_view is used with wide
-    // strings.
-    const StringType& str = string(s);
-    return MatchAndExplain(str, listener);
-  }
-#endif  // GTEST_HAS_ABSL
-
-  // Accepts pointer types, particularly:
-  //   const char*
-  //   char*
-  //   const wchar_t*
-  //   wchar_t*
-  template <typename CharType>
-  bool MatchAndExplain(CharType* s, MatchResultListener* listener) const {
-    return s != NULL && MatchAndExplain(StringType(s), listener);
-  }
-
-  // Matches anything that can convert to StringType.
-  //
-  // This is a template, not just a plain function with const StringType&,
-  // because absl::string_view has some interfering non-explicit constructors.
-  template <typename MatcheeStringType>
-  bool MatchAndExplain(const MatcheeStringType& s,
-                       MatchResultListener* /* listener */) const {
-    const StringType& s2(s);
-    return s2.find(substring_) != StringType::npos;
-  }
-
-  // Describes what this matcher matches.
-  void DescribeTo(::std::ostream* os) const {
-    *os << "has substring ";
-    UniversalPrint(substring_, os);
-  }
-
-  void DescribeNegationTo(::std::ostream* os) const {
-    *os << "has no substring ";
-    UniversalPrint(substring_, os);
-  }
-
- private:
-  const StringType substring_;
-
-  GTEST_DISALLOW_ASSIGN_(HasSubstrMatcher);
-};
-
-// Implements the polymorphic StartsWith(substring) matcher, which
-// can be used as a Matcher<T> as long as T can be converted to a
-// string.
-template <typename StringType>
-class StartsWithMatcher {
- public:
-  explicit StartsWithMatcher(const StringType& prefix) : prefix_(prefix) {
-  }
-
-#if GTEST_HAS_ABSL
-  bool MatchAndExplain(const absl::string_view& s,
-                       MatchResultListener* listener) const {
-    if (s.data() == NULL) {
-      return false;
-    }
-    // This should fail to compile if absl::string_view is used with wide
-    // strings.
-    const StringType& str = string(s);
-    return MatchAndExplain(str, listener);
-  }
-#endif  // GTEST_HAS_ABSL
-
-  // Accepts pointer types, particularly:
-  //   const char*
-  //   char*
-  //   const wchar_t*
-  //   wchar_t*
-  template <typename CharType>
-  bool MatchAndExplain(CharType* s, MatchResultListener* listener) const {
-    return s != NULL && MatchAndExplain(StringType(s), listener);
-  }
-
-  // Matches anything that can convert to StringType.
-  //
-  // This is a template, not just a plain function with const StringType&,
-  // because absl::string_view has some interfering non-explicit constructors.
-  template <typename MatcheeStringType>
-  bool MatchAndExplain(const MatcheeStringType& s,
-                       MatchResultListener* /* listener */) const {
-    const StringType& s2(s);
-    return s2.length() >= prefix_.length() &&
-        s2.substr(0, prefix_.length()) == prefix_;
-  }
-
-  void DescribeTo(::std::ostream* os) const {
-    *os << "starts with ";
-    UniversalPrint(prefix_, os);
-  }
-
-  void DescribeNegationTo(::std::ostream* os) const {
-    *os << "doesn't start with ";
-    UniversalPrint(prefix_, os);
-  }
-
- private:
-  const StringType prefix_;
-
-  GTEST_DISALLOW_ASSIGN_(StartsWithMatcher);
-};
-
-// Implements the polymorphic EndsWith(substring) matcher, which
-// can be used as a Matcher<T> as long as T can be converted to a
-// string.
-template <typename StringType>
-class EndsWithMatcher {
- public:
-  explicit EndsWithMatcher(const StringType& suffix) : suffix_(suffix) {}
-
-#if GTEST_HAS_ABSL
-  bool MatchAndExplain(const absl::string_view& s,
-                       MatchResultListener* listener) const {
-    if (s.data() == NULL) {
-      return false;
-    }
-    // This should fail to compile if absl::string_view is used with wide
-    // strings.
-    const StringType& str = string(s);
-    return MatchAndExplain(str, listener);
-  }
-#endif  // GTEST_HAS_ABSL
-
-  // Accepts pointer types, particularly:
-  //   const char*
-  //   char*
-  //   const wchar_t*
-  //   wchar_t*
-  template <typename CharType>
-  bool MatchAndExplain(CharType* s, MatchResultListener* listener) const {
-    return s != NULL && MatchAndExplain(StringType(s), listener);
-  }
-
-  // Matches anything that can convert to StringType.
-  //
-  // This is a template, not just a plain function with const StringType&,
-  // because absl::string_view has some interfering non-explicit constructors.
-  template <typename MatcheeStringType>
-  bool MatchAndExplain(const MatcheeStringType& s,
-                       MatchResultListener* /* listener */) const {
-    const StringType& s2(s);
-    return s2.length() >= suffix_.length() &&
-        s2.substr(s2.length() - suffix_.length()) == suffix_;
-  }
-
-  void DescribeTo(::std::ostream* os) const {
-    *os << "ends with ";
-    UniversalPrint(suffix_, os);
-  }
-
-  void DescribeNegationTo(::std::ostream* os) const {
-    *os << "doesn't end with ";
-    UniversalPrint(suffix_, os);
-  }
-
- private:
-  const StringType suffix_;
-
-  GTEST_DISALLOW_ASSIGN_(EndsWithMatcher);
-};
-
-// Implements polymorphic matchers MatchesRegex(regex) and
-// ContainsRegex(regex), which can be used as a Matcher<T> as long as
-// T can be converted to a string.
-class MatchesRegexMatcher {
- public:
-  MatchesRegexMatcher(const RE* regex, bool full_match)
-      : regex_(regex), full_match_(full_match) {}
-
-#if GTEST_HAS_ABSL
-  bool MatchAndExplain(const absl::string_view& s,
-                       MatchResultListener* listener) const {
-    return s.data() && MatchAndExplain(string(s), listener);
-  }
-#endif  // GTEST_HAS_ABSL
-
-  // Accepts pointer types, particularly:
-  //   const char*
-  //   char*
-  //   const wchar_t*
-  //   wchar_t*
-  template <typename CharType>
-  bool MatchAndExplain(CharType* s, MatchResultListener* listener) const {
-    return s != NULL && MatchAndExplain(std::string(s), listener);
-  }
-
-  // Matches anything that can convert to std::string.
-  //
-  // This is a template, not just a plain function with const std::string&,
-  // because absl::string_view has some interfering non-explicit constructors.
-  template <class MatcheeStringType>
-  bool MatchAndExplain(const MatcheeStringType& s,
-                       MatchResultListener* /* listener */) const {
-    const std::string& s2(s);
-    return full_match_ ? RE::FullMatch(s2, *regex_) :
-        RE::PartialMatch(s2, *regex_);
-  }
-
-  void DescribeTo(::std::ostream* os) const {
-    *os << (full_match_ ? "matches" : "contains")
-        << " regular expression ";
-    UniversalPrinter<std::string>::Print(regex_->pattern(), os);
-  }
-
-  void DescribeNegationTo(::std::ostream* os) const {
-    *os << "doesn't " << (full_match_ ? "match" : "contain")
-        << " regular expression ";
-    UniversalPrinter<std::string>::Print(regex_->pattern(), os);
-  }
-
- private:
-  const internal::linked_ptr<const RE> regex_;
-  const bool full_match_;
-
-  GTEST_DISALLOW_ASSIGN_(MatchesRegexMatcher);
-};
-
-// Implements a matcher that compares the two fields of a 2-tuple
-// using one of the ==, <=, <, etc, operators.  The two fields being
-// compared don't have to have the same type.
-//
-// The matcher defined here is polymorphic (for example, Eq() can be
-// used to match a tuple<int, short>, a tuple<const long&, double>,
-// etc).  Therefore we use a template type conversion operator in the
-// implementation.
-template <typename D, typename Op>
-class PairMatchBase {
- public:
-  template <typename T1, typename T2>
-  operator Matcher< ::testing::tuple<T1, T2> >() const {
-    return MakeMatcher(new Impl< ::testing::tuple<T1, T2> >);
-  }
-  template <typename T1, typename T2>
-  operator Matcher<const ::testing::tuple<T1, T2>&>() const {
-    return MakeMatcher(new Impl<const ::testing::tuple<T1, T2>&>);
-  }
-
- private:
-  static ::std::ostream& GetDesc(::std::ostream& os) {  // NOLINT
-    return os << D::Desc();
-  }
-
-  template <typename Tuple>
-  class Impl : public MatcherInterface<Tuple> {
-   public:
-    virtual bool MatchAndExplain(
-        Tuple args,
-        MatchResultListener* /* listener */) const {
-      return Op()(::testing::get<0>(args), ::testing::get<1>(args));
-    }
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "are " << GetDesc;
-    }
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << "aren't " << GetDesc;
-    }
-  };
-};
-
-class Eq2Matcher : public PairMatchBase<Eq2Matcher, AnyEq> {
- public:
-  static const char* Desc() { return "an equal pair"; }
-};
-class Ne2Matcher : public PairMatchBase<Ne2Matcher, AnyNe> {
- public:
-  static const char* Desc() { return "an unequal pair"; }
-};
-class Lt2Matcher : public PairMatchBase<Lt2Matcher, AnyLt> {
- public:
-  static const char* Desc() { return "a pair where the first < the second"; }
-};
-class Gt2Matcher : public PairMatchBase<Gt2Matcher, AnyGt> {
- public:
-  static const char* Desc() { return "a pair where the first > the second"; }
-};
-class Le2Matcher : public PairMatchBase<Le2Matcher, AnyLe> {
- public:
-  static const char* Desc() { return "a pair where the first <= the second"; }
-};
-class Ge2Matcher : public PairMatchBase<Ge2Matcher, AnyGe> {
- public:
-  static const char* Desc() { return "a pair where the first >= the second"; }
-};
-
-// Implements the Not(...) matcher for a particular argument type T.
-// We do not nest it inside the NotMatcher class template, as that
-// will prevent different instantiations of NotMatcher from sharing
-// the same NotMatcherImpl<T> class.
-template <typename T>
-class NotMatcherImpl : public MatcherInterface<GTEST_REFERENCE_TO_CONST_(T)> {
- public:
-  explicit NotMatcherImpl(const Matcher<T>& matcher)
-      : matcher_(matcher) {}
-
-  virtual bool MatchAndExplain(GTEST_REFERENCE_TO_CONST_(T) x,
-                               MatchResultListener* listener) const {
-    return !matcher_.MatchAndExplain(x, listener);
-  }
-
-  virtual void DescribeTo(::std::ostream* os) const {
-    matcher_.DescribeNegationTo(os);
-  }
-
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    matcher_.DescribeTo(os);
-  }
-
- private:
-  const Matcher<T> matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(NotMatcherImpl);
-};
-
-// Implements the Not(m) matcher, which matches a value that doesn't
-// match matcher m.
-template <typename InnerMatcher>
-class NotMatcher {
- public:
-  explicit NotMatcher(InnerMatcher matcher) : matcher_(matcher) {}
-
-  // This template type conversion operator allows Not(m) to be used
-  // to match any type m can match.
-  template <typename T>
-  operator Matcher<T>() const {
-    return Matcher<T>(new NotMatcherImpl<T>(SafeMatcherCast<T>(matcher_)));
-  }
-
- private:
-  InnerMatcher matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(NotMatcher);
-};
-
-// Implements the AllOf(m1, m2) matcher for a particular argument type
-// T. We do not nest it inside the BothOfMatcher class template, as
-// that will prevent different instantiations of BothOfMatcher from
-// sharing the same BothOfMatcherImpl<T> class.
-template <typename T>
-class AllOfMatcherImpl
-    : public MatcherInterface<GTEST_REFERENCE_TO_CONST_(T)> {
- public:
-  explicit AllOfMatcherImpl(std::vector<Matcher<T> > matchers)
-      : matchers_(internal::move(matchers)) {}
-
-  virtual void DescribeTo(::std::ostream* os) const {
-    *os << "(";
-    for (size_t i = 0; i < matchers_.size(); ++i) {
-      if (i != 0) *os << ") and (";
-      matchers_[i].DescribeTo(os);
-    }
-    *os << ")";
-  }
-
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    *os << "(";
-    for (size_t i = 0; i < matchers_.size(); ++i) {
-      if (i != 0) *os << ") or (";
-      matchers_[i].DescribeNegationTo(os);
-    }
-    *os << ")";
-  }
-
-  virtual bool MatchAndExplain(GTEST_REFERENCE_TO_CONST_(T) x,
-                               MatchResultListener* listener) const {
-    // If either matcher1_ or matcher2_ doesn't match x, we only need
-    // to explain why one of them fails.
-    std::string all_match_result;
-
-    for (size_t i = 0; i < matchers_.size(); ++i) {
-      StringMatchResultListener slistener;
-      if (matchers_[i].MatchAndExplain(x, &slistener)) {
-        if (all_match_result.empty()) {
-          all_match_result = slistener.str();
-        } else {
-          std::string result = slistener.str();
-          if (!result.empty()) {
-            all_match_result += ", and ";
-            all_match_result += result;
-          }
-        }
-      } else {
-        *listener << slistener.str();
-        return false;
-      }
-    }
-
-    // Otherwise we need to explain why *both* of them match.
-    *listener << all_match_result;
-    return true;
-  }
-
- private:
-  const std::vector<Matcher<T> > matchers_;
-
-  GTEST_DISALLOW_ASSIGN_(AllOfMatcherImpl);
-};
-
-#if GTEST_LANG_CXX11
-// VariadicMatcher is used for the variadic implementation of
-// AllOf(m_1, m_2, ...) and AnyOf(m_1, m_2, ...).
-// CombiningMatcher<T> is used to recursively combine the provided matchers
-// (of type Args...).
-template <template <typename T> class CombiningMatcher, typename... Args>
-class VariadicMatcher {
- public:
-  VariadicMatcher(const Args&... matchers)  // NOLINT
-      : matchers_(matchers...) {
-    static_assert(sizeof...(Args) > 0, "Must have at least one matcher.");
-  }
-
-  // This template type conversion operator allows an
-  // VariadicMatcher<Matcher1, Matcher2...> object to match any type that
-  // all of the provided matchers (Matcher1, Matcher2, ...) can match.
-  template <typename T>
-  operator Matcher<T>() const {
-    std::vector<Matcher<T> > values;
-    CreateVariadicMatcher<T>(&values, std::integral_constant<size_t, 0>());
-    return Matcher<T>(new CombiningMatcher<T>(internal::move(values)));
-  }
-
- private:
-  template <typename T, size_t I>
-  void CreateVariadicMatcher(std::vector<Matcher<T> >* values,
-                             std::integral_constant<size_t, I>) const {
-    values->push_back(SafeMatcherCast<T>(std::get<I>(matchers_)));
-    CreateVariadicMatcher<T>(values, std::integral_constant<size_t, I + 1>());
-  }
-
-  template <typename T>
-  void CreateVariadicMatcher(
-      std::vector<Matcher<T> >*,
-      std::integral_constant<size_t, sizeof...(Args)>) const {}
-
-  tuple<Args...> matchers_;
-
-  GTEST_DISALLOW_ASSIGN_(VariadicMatcher);
-};
-
-template <typename... Args>
-using AllOfMatcher = VariadicMatcher<AllOfMatcherImpl, Args...>;
-
-#endif  // GTEST_LANG_CXX11
-
-// Used for implementing the AllOf(m_1, ..., m_n) matcher, which
-// matches a value that matches all of the matchers m_1, ..., and m_n.
-template <typename Matcher1, typename Matcher2>
-class BothOfMatcher {
- public:
-  BothOfMatcher(Matcher1 matcher1, Matcher2 matcher2)
-      : matcher1_(matcher1), matcher2_(matcher2) {}
-
-  // This template type conversion operator allows a
-  // BothOfMatcher<Matcher1, Matcher2> object to match any type that
-  // both Matcher1 and Matcher2 can match.
-  template <typename T>
-  operator Matcher<T>() const {
-    std::vector<Matcher<T> > values;
-    values.push_back(SafeMatcherCast<T>(matcher1_));
-    values.push_back(SafeMatcherCast<T>(matcher2_));
-    return Matcher<T>(new AllOfMatcherImpl<T>(internal::move(values)));
-  }
-
- private:
-  Matcher1 matcher1_;
-  Matcher2 matcher2_;
-
-  GTEST_DISALLOW_ASSIGN_(BothOfMatcher);
-};
-
-// Implements the AnyOf(m1, m2) matcher for a particular argument type
-// T.  We do not nest it inside the AnyOfMatcher class template, as
-// that will prevent different instantiations of AnyOfMatcher from
-// sharing the same EitherOfMatcherImpl<T> class.
-template <typename T>
-class AnyOfMatcherImpl
-    : public MatcherInterface<GTEST_REFERENCE_TO_CONST_(T)> {
- public:
-  explicit AnyOfMatcherImpl(std::vector<Matcher<T> > matchers)
-      : matchers_(internal::move(matchers)) {}
-
-  virtual void DescribeTo(::std::ostream* os) const {
-    *os << "(";
-    for (size_t i = 0; i < matchers_.size(); ++i) {
-      if (i != 0) *os << ") or (";
-      matchers_[i].DescribeTo(os);
-    }
-    *os << ")";
-  }
-
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    *os << "(";
-    for (size_t i = 0; i < matchers_.size(); ++i) {
-      if (i != 0) *os << ") and (";
-      matchers_[i].DescribeNegationTo(os);
-    }
-    *os << ")";
-  }
-
-  virtual bool MatchAndExplain(GTEST_REFERENCE_TO_CONST_(T) x,
-                               MatchResultListener* listener) const {
-    std::string no_match_result;
-
-    // If either matcher1_ or matcher2_ matches x, we just need to
-    // explain why *one* of them matches.
-    for (size_t i = 0; i < matchers_.size(); ++i) {
-      StringMatchResultListener slistener;
-      if (matchers_[i].MatchAndExplain(x, &slistener)) {
-        *listener << slistener.str();
-        return true;
-      } else {
-        if (no_match_result.empty()) {
-          no_match_result = slistener.str();
-        } else {
-          std::string result = slistener.str();
-          if (!result.empty()) {
-            no_match_result += ", and ";
-            no_match_result += result;
-          }
-        }
-      }
-    }
-
-    // Otherwise we need to explain why *both* of them fail.
-    *listener << no_match_result;
-    return false;
-  }
-
- private:
-  const std::vector<Matcher<T> > matchers_;
-
-  GTEST_DISALLOW_ASSIGN_(AnyOfMatcherImpl);
-};
-
-#if GTEST_LANG_CXX11
-// AnyOfMatcher is used for the variadic implementation of AnyOf(m_1, m_2, ...).
-template <typename... Args>
-using AnyOfMatcher = VariadicMatcher<AnyOfMatcherImpl, Args...>;
-
-#endif  // GTEST_LANG_CXX11
-
-// Used for implementing the AnyOf(m_1, ..., m_n) matcher, which
-// matches a value that matches at least one of the matchers m_1, ...,
-// and m_n.
-template <typename Matcher1, typename Matcher2>
-class EitherOfMatcher {
- public:
-  EitherOfMatcher(Matcher1 matcher1, Matcher2 matcher2)
-      : matcher1_(matcher1), matcher2_(matcher2) {}
-
-  // This template type conversion operator allows a
-  // EitherOfMatcher<Matcher1, Matcher2> object to match any type that
-  // both Matcher1 and Matcher2 can match.
-  template <typename T>
-  operator Matcher<T>() const {
-    std::vector<Matcher<T> > values;
-    values.push_back(SafeMatcherCast<T>(matcher1_));
-    values.push_back(SafeMatcherCast<T>(matcher2_));
-    return Matcher<T>(new AnyOfMatcherImpl<T>(internal::move(values)));
-  }
-
- private:
-  Matcher1 matcher1_;
-  Matcher2 matcher2_;
-
-  GTEST_DISALLOW_ASSIGN_(EitherOfMatcher);
-};
-
-// Used for implementing Truly(pred), which turns a predicate into a
-// matcher.
-template <typename Predicate>
-class TrulyMatcher {
- public:
-  explicit TrulyMatcher(Predicate pred) : predicate_(pred) {}
-
-  // This method template allows Truly(pred) to be used as a matcher
-  // for type T where T is the argument type of predicate 'pred'.  The
-  // argument is passed by reference as the predicate may be
-  // interested in the address of the argument.
-  template <typename T>
-  bool MatchAndExplain(T& x,  // NOLINT
-                       MatchResultListener* /* listener */) const {
-    // Without the if-statement, MSVC sometimes warns about converting
-    // a value to bool (warning 4800).
-    //
-    // We cannot write 'return !!predicate_(x);' as that doesn't work
-    // when predicate_(x) returns a class convertible to bool but
-    // having no operator!().
-    if (predicate_(x))
-      return true;
-    return false;
-  }
-
-  void DescribeTo(::std::ostream* os) const {
-    *os << "satisfies the given predicate";
-  }
-
-  void DescribeNegationTo(::std::ostream* os) const {
-    *os << "doesn't satisfy the given predicate";
-  }
-
- private:
-  Predicate predicate_;
-
-  GTEST_DISALLOW_ASSIGN_(TrulyMatcher);
-};
-
-// Used for implementing Matches(matcher), which turns a matcher into
-// a predicate.
-template <typename M>
-class MatcherAsPredicate {
- public:
-  explicit MatcherAsPredicate(M matcher) : matcher_(matcher) {}
-
-  // This template operator() allows Matches(m) to be used as a
-  // predicate on type T where m is a matcher on type T.
-  //
-  // The argument x is passed by reference instead of by value, as
-  // some matcher may be interested in its address (e.g. as in
-  // Matches(Ref(n))(x)).
-  template <typename T>
-  bool operator()(const T& x) const {
-    // We let matcher_ commit to a particular type here instead of
-    // when the MatcherAsPredicate object was constructed.  This
-    // allows us to write Matches(m) where m is a polymorphic matcher
-    // (e.g. Eq(5)).
-    //
-    // If we write Matcher<T>(matcher_).Matches(x) here, it won't
-    // compile when matcher_ has type Matcher<const T&>; if we write
-    // Matcher<const T&>(matcher_).Matches(x) here, it won't compile
-    // when matcher_ has type Matcher<T>; if we just write
-    // matcher_.Matches(x), it won't compile when matcher_ is
-    // polymorphic, e.g. Eq(5).
-    //
-    // MatcherCast<const T&>() is necessary for making the code work
-    // in all of the above situations.
-    return MatcherCast<const T&>(matcher_).Matches(x);
-  }
-
- private:
-  M matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(MatcherAsPredicate);
-};
-
-// For implementing ASSERT_THAT() and EXPECT_THAT().  The template
-// argument M must be a type that can be converted to a matcher.
-template <typename M>
-class PredicateFormatterFromMatcher {
- public:
-  explicit PredicateFormatterFromMatcher(M m) : matcher_(internal::move(m)) {}
-
-  // This template () operator allows a PredicateFormatterFromMatcher
-  // object to act as a predicate-formatter suitable for using with
-  // Google Test's EXPECT_PRED_FORMAT1() macro.
-  template <typename T>
-  AssertionResult operator()(const char* value_text, const T& x) const {
-    // We convert matcher_ to a Matcher<const T&> *now* instead of
-    // when the PredicateFormatterFromMatcher object was constructed,
-    // as matcher_ may be polymorphic (e.g. NotNull()) and we won't
-    // know which type to instantiate it to until we actually see the
-    // type of x here.
-    //
-    // We write SafeMatcherCast<const T&>(matcher_) instead of
-    // Matcher<const T&>(matcher_), as the latter won't compile when
-    // matcher_ has type Matcher<T> (e.g. An<int>()).
-    // We don't write MatcherCast<const T&> either, as that allows
-    // potentially unsafe downcasting of the matcher argument.
-    const Matcher<const T&> matcher = SafeMatcherCast<const T&>(matcher_);
-    StringMatchResultListener listener;
-    if (MatchPrintAndExplain(x, matcher, &listener))
-      return AssertionSuccess();
-
-    ::std::stringstream ss;
-    ss << "Value of: " << value_text << "\n"
-       << "Expected: ";
-    matcher.DescribeTo(&ss);
-    ss << "\n  Actual: " << listener.str();
-    return AssertionFailure() << ss.str();
-  }
-
- private:
-  const M matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(PredicateFormatterFromMatcher);
-};
-
-// A helper function for converting a matcher to a predicate-formatter
-// without the user needing to explicitly write the type.  This is
-// used for implementing ASSERT_THAT() and EXPECT_THAT().
-// Implementation detail: 'matcher' is received by-value to force decaying.
-template <typename M>
-inline PredicateFormatterFromMatcher<M>
-MakePredicateFormatterFromMatcher(M matcher) {
-  return PredicateFormatterFromMatcher<M>(internal::move(matcher));
-}
-
-// Implements the polymorphic floating point equality matcher, which matches
-// two float values using ULP-based approximation or, optionally, a
-// user-specified epsilon.  The template is meant to be instantiated with
-// FloatType being either float or double.
-template <typename FloatType>
-class FloatingEqMatcher {
- public:
-  // Constructor for FloatingEqMatcher.
-  // The matcher's input will be compared with expected.  The matcher treats two
-  // NANs as equal if nan_eq_nan is true.  Otherwise, under IEEE standards,
-  // equality comparisons between NANs will always return false.  We specify a
-  // negative max_abs_error_ term to indicate that ULP-based approximation will
-  // be used for comparison.
-  FloatingEqMatcher(FloatType expected, bool nan_eq_nan) :
-    expected_(expected), nan_eq_nan_(nan_eq_nan), max_abs_error_(-1) {
-  }
-
-  // Constructor that supports a user-specified max_abs_error that will be used
-  // for comparison instead of ULP-based approximation.  The max absolute
-  // should be non-negative.
-  FloatingEqMatcher(FloatType expected, bool nan_eq_nan,
-                    FloatType max_abs_error)
-      : expected_(expected),
-        nan_eq_nan_(nan_eq_nan),
-        max_abs_error_(max_abs_error) {
-    GTEST_CHECK_(max_abs_error >= 0)
-        << ", where max_abs_error is" << max_abs_error;
-  }
-
-  // Implements floating point equality matcher as a Matcher<T>.
-  template <typename T>
-  class Impl : public MatcherInterface<T> {
-   public:
-    Impl(FloatType expected, bool nan_eq_nan, FloatType max_abs_error)
-        : expected_(expected),
-          nan_eq_nan_(nan_eq_nan),
-          max_abs_error_(max_abs_error) {}
-
-    virtual bool MatchAndExplain(T value,
-                                 MatchResultListener* listener) const {
-      const FloatingPoint<FloatType> actual(value), expected(expected_);
-
-      // Compares NaNs first, if nan_eq_nan_ is true.
-      if (actual.is_nan() || expected.is_nan()) {
-        if (actual.is_nan() && expected.is_nan()) {
-          return nan_eq_nan_;
-        }
-        // One is nan; the other is not nan.
-        return false;
-      }
-      if (HasMaxAbsError()) {
-        // We perform an equality check so that inf will match inf, regardless
-        // of error bounds.  If the result of value - expected_ would result in
-        // overflow or if either value is inf, the default result is infinity,
-        // which should only match if max_abs_error_ is also infinity.
-        if (value == expected_) {
-          return true;
-        }
-
-        const FloatType diff = value - expected_;
-        if (fabs(diff) <= max_abs_error_) {
-          return true;
-        }
-
-        if (listener->IsInterested()) {
-          *listener << "which is " << diff << " from " << expected_;
-        }
-        return false;
-      } else {
-        return actual.AlmostEquals(expected);
-      }
-    }
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      // os->precision() returns the previously set precision, which we
-      // store to restore the ostream to its original configuration
-      // after outputting.
-      const ::std::streamsize old_precision = os->precision(
-          ::std::numeric_limits<FloatType>::digits10 + 2);
-      if (FloatingPoint<FloatType>(expected_).is_nan()) {
-        if (nan_eq_nan_) {
-          *os << "is NaN";
-        } else {
-          *os << "never matches";
-        }
-      } else {
-        *os << "is approximately " << expected_;
-        if (HasMaxAbsError()) {
-          *os << " (absolute error <= " << max_abs_error_ << ")";
-        }
-      }
-      os->precision(old_precision);
-    }
-
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      // As before, get original precision.
-      const ::std::streamsize old_precision = os->precision(
-          ::std::numeric_limits<FloatType>::digits10 + 2);
-      if (FloatingPoint<FloatType>(expected_).is_nan()) {
-        if (nan_eq_nan_) {
-          *os << "isn't NaN";
-        } else {
-          *os << "is anything";
-        }
-      } else {
-        *os << "isn't approximately " << expected_;
-        if (HasMaxAbsError()) {
-          *os << " (absolute error > " << max_abs_error_ << ")";
-        }
-      }
-      // Restore original precision.
-      os->precision(old_precision);
-    }
-
-   private:
-    bool HasMaxAbsError() const {
-      return max_abs_error_ >= 0;
-    }
-
-    const FloatType expected_;
-    const bool nan_eq_nan_;
-    // max_abs_error will be used for value comparison when >= 0.
-    const FloatType max_abs_error_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
-  // The following 3 type conversion operators allow FloatEq(expected) and
-  // NanSensitiveFloatEq(expected) to be used as a Matcher<float>, a
-  // Matcher<const float&>, or a Matcher<float&>, but nothing else.
-  // (While Google's C++ coding style doesn't allow arguments passed
-  // by non-const reference, we may see them in code not conforming to
-  // the style.  Therefore Google Mock needs to support them.)
-  operator Matcher<FloatType>() const {
-    return MakeMatcher(
-        new Impl<FloatType>(expected_, nan_eq_nan_, max_abs_error_));
-  }
-
-  operator Matcher<const FloatType&>() const {
-    return MakeMatcher(
-        new Impl<const FloatType&>(expected_, nan_eq_nan_, max_abs_error_));
-  }
-
-  operator Matcher<FloatType&>() const {
-    return MakeMatcher(
-        new Impl<FloatType&>(expected_, nan_eq_nan_, max_abs_error_));
-  }
-
- private:
-  const FloatType expected_;
-  const bool nan_eq_nan_;
-  // max_abs_error will be used for value comparison when >= 0.
-  const FloatType max_abs_error_;
-
-  GTEST_DISALLOW_ASSIGN_(FloatingEqMatcher);
-};
-
-// A 2-tuple ("binary") wrapper around FloatingEqMatcher:
-// FloatingEq2Matcher() matches (x, y) by matching FloatingEqMatcher(x, false)
-// against y, and FloatingEq2Matcher(e) matches FloatingEqMatcher(x, false, e)
-// against y. The former implements "Eq", the latter "Near". At present, there
-// is no version that compares NaNs as equal.
-template <typename FloatType>
-class FloatingEq2Matcher {
- public:
-  FloatingEq2Matcher() { Init(-1, false); }
-
-  explicit FloatingEq2Matcher(bool nan_eq_nan) { Init(-1, nan_eq_nan); }
-
-  explicit FloatingEq2Matcher(FloatType max_abs_error) {
-    Init(max_abs_error, false);
-  }
-
-  FloatingEq2Matcher(FloatType max_abs_error, bool nan_eq_nan) {
-    Init(max_abs_error, nan_eq_nan);
-  }
-
-  template <typename T1, typename T2>
-  operator Matcher< ::testing::tuple<T1, T2> >() const {
-    return MakeMatcher(
-        new Impl< ::testing::tuple<T1, T2> >(max_abs_error_, nan_eq_nan_));
-  }
-  template <typename T1, typename T2>
-  operator Matcher<const ::testing::tuple<T1, T2>&>() const {
-    return MakeMatcher(
-        new Impl<const ::testing::tuple<T1, T2>&>(max_abs_error_, nan_eq_nan_));
-  }
-
- private:
-  static ::std::ostream& GetDesc(::std::ostream& os) {  // NOLINT
-    return os << "an almost-equal pair";
-  }
-
-  template <typename Tuple>
-  class Impl : public MatcherInterface<Tuple> {
-   public:
-    Impl(FloatType max_abs_error, bool nan_eq_nan) :
-        max_abs_error_(max_abs_error),
-        nan_eq_nan_(nan_eq_nan) {}
-
-    virtual bool MatchAndExplain(Tuple args,
-                                 MatchResultListener* listener) const {
-      if (max_abs_error_ == -1) {
-        FloatingEqMatcher<FloatType> fm(::testing::get<0>(args), nan_eq_nan_);
-        return static_cast<Matcher<FloatType> >(fm).MatchAndExplain(
-            ::testing::get<1>(args), listener);
-      } else {
-        FloatingEqMatcher<FloatType> fm(::testing::get<0>(args), nan_eq_nan_,
-                                        max_abs_error_);
-        return static_cast<Matcher<FloatType> >(fm).MatchAndExplain(
-            ::testing::get<1>(args), listener);
-      }
-    }
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "are " << GetDesc;
-    }
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << "aren't " << GetDesc;
-    }
-
-   private:
-    FloatType max_abs_error_;
-    const bool nan_eq_nan_;
-  };
-
-  void Init(FloatType max_abs_error_val, bool nan_eq_nan_val) {
-    max_abs_error_ = max_abs_error_val;
-    nan_eq_nan_ = nan_eq_nan_val;
-  }
-  FloatType max_abs_error_;
-  bool nan_eq_nan_;
-};
-
-// Implements the Pointee(m) matcher for matching a pointer whose
-// pointee matches matcher m.  The pointer can be either raw or smart.
-template <typename InnerMatcher>
-class PointeeMatcher {
- public:
-  explicit PointeeMatcher(const InnerMatcher& matcher) : matcher_(matcher) {}
-
-  // This type conversion operator template allows Pointee(m) to be
-  // used as a matcher for any pointer type whose pointee type is
-  // compatible with the inner matcher, where type Pointer can be
-  // either a raw pointer or a smart pointer.
-  //
-  // The reason we do this instead of relying on
-  // MakePolymorphicMatcher() is that the latter is not flexible
-  // enough for implementing the DescribeTo() method of Pointee().
-  template <typename Pointer>
-  operator Matcher<Pointer>() const {
-    return Matcher<Pointer>(
-        new Impl<GTEST_REFERENCE_TO_CONST_(Pointer)>(matcher_));
-  }
-
- private:
-  // The monomorphic implementation that works for a particular pointer type.
-  template <typename Pointer>
-  class Impl : public MatcherInterface<Pointer> {
-   public:
-    typedef typename PointeeOf<GTEST_REMOVE_CONST_(  // NOLINT
-        GTEST_REMOVE_REFERENCE_(Pointer))>::type Pointee;
-
-    explicit Impl(const InnerMatcher& matcher)
-        : matcher_(MatcherCast<const Pointee&>(matcher)) {}
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "points to a value that ";
-      matcher_.DescribeTo(os);
-    }
-
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << "does not point to a value that ";
-      matcher_.DescribeTo(os);
-    }
-
-    virtual bool MatchAndExplain(Pointer pointer,
-                                 MatchResultListener* listener) const {
-      if (GetRawPointer(pointer) == NULL)
-        return false;
-
-      *listener << "which points to ";
-      return MatchPrintAndExplain(*pointer, matcher_, listener);
-    }
-
-   private:
-    const Matcher<const Pointee&> matcher_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
-  const InnerMatcher matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(PointeeMatcher);
-};
-
-#if GTEST_HAS_RTTI
-// Implements the WhenDynamicCastTo<T>(m) matcher that matches a pointer or
-// reference that matches inner_matcher when dynamic_cast<T> is applied.
-// The result of dynamic_cast<To> is forwarded to the inner matcher.
-// If To is a pointer and the cast fails, the inner matcher will receive NULL.
-// If To is a reference and the cast fails, this matcher returns false
-// immediately.
-template <typename To>
-class WhenDynamicCastToMatcherBase {
- public:
-  explicit WhenDynamicCastToMatcherBase(const Matcher<To>& matcher)
-      : matcher_(matcher) {}
-
-  void DescribeTo(::std::ostream* os) const {
-    GetCastTypeDescription(os);
-    matcher_.DescribeTo(os);
-  }
-
-  void DescribeNegationTo(::std::ostream* os) const {
-    GetCastTypeDescription(os);
-    matcher_.DescribeNegationTo(os);
-  }
-
- protected:
-  const Matcher<To> matcher_;
-
-  static std::string GetToName() {
-    return GetTypeName<To>();
-  }
-
- private:
-  static void GetCastTypeDescription(::std::ostream* os) {
-    *os << "when dynamic_cast to " << GetToName() << ", ";
-  }
-
-  GTEST_DISALLOW_ASSIGN_(WhenDynamicCastToMatcherBase);
-};
-
-// Primary template.
-// To is a pointer. Cast and forward the result.
-template <typename To>
-class WhenDynamicCastToMatcher : public WhenDynamicCastToMatcherBase<To> {
- public:
-  explicit WhenDynamicCastToMatcher(const Matcher<To>& matcher)
-      : WhenDynamicCastToMatcherBase<To>(matcher) {}
-
-  template <typename From>
-  bool MatchAndExplain(From from, MatchResultListener* listener) const {
-    // TODO(sbenza): Add more detail on failures. ie did the dyn_cast fail?
-    To to = dynamic_cast<To>(from);
-    return MatchPrintAndExplain(to, this->matcher_, listener);
-  }
-};
-
-// Specialize for references.
-// In this case we return false if the dynamic_cast fails.
-template <typename To>
-class WhenDynamicCastToMatcher<To&> : public WhenDynamicCastToMatcherBase<To&> {
- public:
-  explicit WhenDynamicCastToMatcher(const Matcher<To&>& matcher)
-      : WhenDynamicCastToMatcherBase<To&>(matcher) {}
-
-  template <typename From>
-  bool MatchAndExplain(From& from, MatchResultListener* listener) const {
-    // We don't want an std::bad_cast here, so do the cast with pointers.
-    To* to = dynamic_cast<To*>(&from);
-    if (to == NULL) {
-      *listener << "which cannot be dynamic_cast to " << this->GetToName();
-      return false;
-    }
-    return MatchPrintAndExplain(*to, this->matcher_, listener);
-  }
-};
-#endif  // GTEST_HAS_RTTI
-
-// Implements the Field() matcher for matching a field (i.e. member
-// variable) of an object.
-template <typename Class, typename FieldType>
-class FieldMatcher {
- public:
-  FieldMatcher(FieldType Class::*field,
-               const Matcher<const FieldType&>& matcher)
-      : field_(field), matcher_(matcher), whose_field_("whose given field ") {}
-
-  FieldMatcher(const std::string& field_name, FieldType Class::*field,
-               const Matcher<const FieldType&>& matcher)
-      : field_(field),
-        matcher_(matcher),
-        whose_field_("whose field `" + field_name + "` ") {}
-
-  void DescribeTo(::std::ostream* os) const {
-    *os << "is an object " << whose_field_;
-    matcher_.DescribeTo(os);
-  }
-
-  void DescribeNegationTo(::std::ostream* os) const {
-    *os << "is an object " << whose_field_;
-    matcher_.DescribeNegationTo(os);
-  }
-
-  template <typename T>
-  bool MatchAndExplain(const T& value, MatchResultListener* listener) const {
-    return MatchAndExplainImpl(
-        typename ::testing::internal::
-            is_pointer<GTEST_REMOVE_CONST_(T)>::type(),
-        value, listener);
-  }
-
- private:
-  // The first argument of MatchAndExplainImpl() is needed to help
-  // Symbian's C++ compiler choose which overload to use.  Its type is
-  // true_type iff the Field() matcher is used to match a pointer.
-  bool MatchAndExplainImpl(false_type /* is_not_pointer */, const Class& obj,
-                           MatchResultListener* listener) const {
-    *listener << whose_field_ << "is ";
-    return MatchPrintAndExplain(obj.*field_, matcher_, listener);
-  }
-
-  bool MatchAndExplainImpl(true_type /* is_pointer */, const Class* p,
-                           MatchResultListener* listener) const {
-    if (p == NULL)
-      return false;
-
-    *listener << "which points to an object ";
-    // Since *p has a field, it must be a class/struct/union type and
-    // thus cannot be a pointer.  Therefore we pass false_type() as
-    // the first argument.
-    return MatchAndExplainImpl(false_type(), *p, listener);
-  }
-
-  const FieldType Class::*field_;
-  const Matcher<const FieldType&> matcher_;
-
-  // Contains either "whose given field " if the name of the field is unknown
-  // or "whose field `name_of_field` " if the name is known.
-  const std::string whose_field_;
-
-  GTEST_DISALLOW_ASSIGN_(FieldMatcher);
-};
-
-// Implements the Property() matcher for matching a property
-// (i.e. return value of a getter method) of an object.
-//
-// Property is a const-qualified member function of Class returning
-// PropertyType.
-template <typename Class, typename PropertyType, typename Property>
-class PropertyMatcher {
- public:
-  // The property may have a reference type, so 'const PropertyType&'
-  // may cause double references and fail to compile.  That's why we
-  // need GTEST_REFERENCE_TO_CONST, which works regardless of
-  // PropertyType being a reference or not.
-  typedef GTEST_REFERENCE_TO_CONST_(PropertyType) RefToConstProperty;
-
-  PropertyMatcher(Property property, const Matcher<RefToConstProperty>& matcher)
-      : property_(property),
-        matcher_(matcher),
-        whose_property_("whose given property ") {}
-
-  PropertyMatcher(const std::string& property_name, Property property,
-                  const Matcher<RefToConstProperty>& matcher)
-      : property_(property),
-        matcher_(matcher),
-        whose_property_("whose property `" + property_name + "` ") {}
-
-  void DescribeTo(::std::ostream* os) const {
-    *os << "is an object " << whose_property_;
-    matcher_.DescribeTo(os);
-  }
-
-  void DescribeNegationTo(::std::ostream* os) const {
-    *os << "is an object " << whose_property_;
-    matcher_.DescribeNegationTo(os);
-  }
-
-  template <typename T>
-  bool MatchAndExplain(const T&value, MatchResultListener* listener) const {
-    return MatchAndExplainImpl(
-        typename ::testing::internal::
-            is_pointer<GTEST_REMOVE_CONST_(T)>::type(),
-        value, listener);
-  }
-
- private:
-  // The first argument of MatchAndExplainImpl() is needed to help
-  // Symbian's C++ compiler choose which overload to use.  Its type is
-  // true_type iff the Property() matcher is used to match a pointer.
-  bool MatchAndExplainImpl(false_type /* is_not_pointer */, const Class& obj,
-                           MatchResultListener* listener) const {
-    *listener << whose_property_ << "is ";
-    // Cannot pass the return value (for example, int) to MatchPrintAndExplain,
-    // which takes a non-const reference as argument.
-#if defined(_PREFAST_ ) && _MSC_VER == 1800
-    // Workaround bug in VC++ 2013's /analyze parser.
-    // https://connect.microsoft.com/VisualStudio/feedback/details/1106363/internal-compiler-error-with-analyze-due-to-failure-to-infer-move
-    posix::Abort();  // To make sure it is never run.
-    return false;
-#else
-    RefToConstProperty result = (obj.*property_)();
-    return MatchPrintAndExplain(result, matcher_, listener);
-#endif
-  }
-
-  bool MatchAndExplainImpl(true_type /* is_pointer */, const Class* p,
-                           MatchResultListener* listener) const {
-    if (p == NULL)
-      return false;
-
-    *listener << "which points to an object ";
-    // Since *p has a property method, it must be a class/struct/union
-    // type and thus cannot be a pointer.  Therefore we pass
-    // false_type() as the first argument.
-    return MatchAndExplainImpl(false_type(), *p, listener);
-  }
-
-  Property property_;
-  const Matcher<RefToConstProperty> matcher_;
-
-  // Contains either "whose given property " if the name of the property is
-  // unknown or "whose property `name_of_property` " if the name is known.
-  const std::string whose_property_;
-
-  GTEST_DISALLOW_ASSIGN_(PropertyMatcher);
-};
-
-// Type traits specifying various features of different functors for ResultOf.
-// The default template specifies features for functor objects.
-// Functor classes have to typedef argument_type and result_type
-// to be compatible with ResultOf.
-template <typename Functor>
-struct CallableTraits {
-  typedef typename Functor::result_type ResultType;
-  typedef Functor StorageType;
-
-  static void CheckIsValid(Functor /* functor */) {}
-  template <typename T>
-  static ResultType Invoke(Functor f, T arg) { return f(arg); }
-};
-
-// Specialization for function pointers.
-template <typename ArgType, typename ResType>
-struct CallableTraits<ResType(*)(ArgType)> {
-  typedef ResType ResultType;
-  typedef ResType(*StorageType)(ArgType);
-
-  static void CheckIsValid(ResType(*f)(ArgType)) {
-    GTEST_CHECK_(f != NULL)
-        << "NULL function pointer is passed into ResultOf().";
-  }
-  template <typename T>
-  static ResType Invoke(ResType(*f)(ArgType), T arg) {
-    return (*f)(arg);
-  }
-};
-
-// Implements the ResultOf() matcher for matching a return value of a
-// unary function of an object.
-template <typename Callable>
-class ResultOfMatcher {
- public:
-  typedef typename CallableTraits<Callable>::ResultType ResultType;
-
-  ResultOfMatcher(Callable callable, const Matcher<ResultType>& matcher)
-      : callable_(callable), matcher_(matcher) {
-    CallableTraits<Callable>::CheckIsValid(callable_);
-  }
-
-  template <typename T>
-  operator Matcher<T>() const {
-    return Matcher<T>(new Impl<T>(callable_, matcher_));
-  }
-
- private:
-  typedef typename CallableTraits<Callable>::StorageType CallableStorageType;
-
-  template <typename T>
-  class Impl : public MatcherInterface<T> {
-   public:
-    Impl(CallableStorageType callable, const Matcher<ResultType>& matcher)
-        : callable_(callable), matcher_(matcher) {}
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "is mapped by the given callable to a value that ";
-      matcher_.DescribeTo(os);
-    }
-
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << "is mapped by the given callable to a value that ";
-      matcher_.DescribeNegationTo(os);
-    }
-
-    virtual bool MatchAndExplain(T obj, MatchResultListener* listener) const {
-      *listener << "which is mapped by the given callable to ";
-      // Cannot pass the return value (for example, int) to
-      // MatchPrintAndExplain, which takes a non-const reference as argument.
-      ResultType result =
-          CallableTraits<Callable>::template Invoke<T>(callable_, obj);
-      return MatchPrintAndExplain(result, matcher_, listener);
-    }
-
-   private:
-    // Functors often define operator() as non-const method even though
-    // they are actually stateless. But we need to use them even when
-    // 'this' is a const pointer. It's the user's responsibility not to
-    // use stateful callables with ResultOf(), which does't guarantee
-    // how many times the callable will be invoked.
-    mutable CallableStorageType callable_;
-    const Matcher<ResultType> matcher_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };  // class Impl
-
-  const CallableStorageType callable_;
-  const Matcher<ResultType> matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(ResultOfMatcher);
-};
-
-// Implements a matcher that checks the size of an STL-style container.
-template <typename SizeMatcher>
-class SizeIsMatcher {
- public:
-  explicit SizeIsMatcher(const SizeMatcher& size_matcher)
-       : size_matcher_(size_matcher) {
-  }
-
-  template <typename Container>
-  operator Matcher<Container>() const {
-    return MakeMatcher(new Impl<Container>(size_matcher_));
-  }
-
-  template <typename Container>
-  class Impl : public MatcherInterface<Container> {
-   public:
-    typedef internal::StlContainerView<
-         GTEST_REMOVE_REFERENCE_AND_CONST_(Container)> ContainerView;
-    typedef typename ContainerView::type::size_type SizeType;
-    explicit Impl(const SizeMatcher& size_matcher)
-        : size_matcher_(MatcherCast<SizeType>(size_matcher)) {}
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "size ";
-      size_matcher_.DescribeTo(os);
-    }
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << "size ";
-      size_matcher_.DescribeNegationTo(os);
-    }
-
-    virtual bool MatchAndExplain(Container container,
-                                 MatchResultListener* listener) const {
-      SizeType size = container.size();
-      StringMatchResultListener size_listener;
-      const bool result = size_matcher_.MatchAndExplain(size, &size_listener);
-      *listener
-          << "whose size " << size << (result ? " matches" : " doesn't match");
-      PrintIfNotEmpty(size_listener.str(), listener->stream());
-      return result;
-    }
-
-   private:
-    const Matcher<SizeType> size_matcher_;
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
- private:
-  const SizeMatcher size_matcher_;
-  GTEST_DISALLOW_ASSIGN_(SizeIsMatcher);
-};
-
-// Implements a matcher that checks the begin()..end() distance of an STL-style
-// container.
-template <typename DistanceMatcher>
-class BeginEndDistanceIsMatcher {
- public:
-  explicit BeginEndDistanceIsMatcher(const DistanceMatcher& distance_matcher)
-      : distance_matcher_(distance_matcher) {}
-
-  template <typename Container>
-  operator Matcher<Container>() const {
-    return MakeMatcher(new Impl<Container>(distance_matcher_));
-  }
-
-  template <typename Container>
-  class Impl : public MatcherInterface<Container> {
-   public:
-    typedef internal::StlContainerView<
-        GTEST_REMOVE_REFERENCE_AND_CONST_(Container)> ContainerView;
-    typedef typename std::iterator_traits<
-        typename ContainerView::type::const_iterator>::difference_type
-        DistanceType;
-    explicit Impl(const DistanceMatcher& distance_matcher)
-        : distance_matcher_(MatcherCast<DistanceType>(distance_matcher)) {}
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "distance between begin() and end() ";
-      distance_matcher_.DescribeTo(os);
-    }
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << "distance between begin() and end() ";
-      distance_matcher_.DescribeNegationTo(os);
-    }
-
-    virtual bool MatchAndExplain(Container container,
-                                 MatchResultListener* listener) const {
-#if GTEST_HAS_STD_BEGIN_AND_END_
-      using std::begin;
-      using std::end;
-      DistanceType distance = std::distance(begin(container), end(container));
-#else
-      DistanceType distance = std::distance(container.begin(), container.end());
-#endif
-      StringMatchResultListener distance_listener;
-      const bool result =
-          distance_matcher_.MatchAndExplain(distance, &distance_listener);
-      *listener << "whose distance between begin() and end() " << distance
-                << (result ? " matches" : " doesn't match");
-      PrintIfNotEmpty(distance_listener.str(), listener->stream());
-      return result;
-    }
-
-   private:
-    const Matcher<DistanceType> distance_matcher_;
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
- private:
-  const DistanceMatcher distance_matcher_;
-  GTEST_DISALLOW_ASSIGN_(BeginEndDistanceIsMatcher);
-};
-
-// Implements an equality matcher for any STL-style container whose elements
-// support ==. This matcher is like Eq(), but its failure explanations provide
-// more detailed information that is useful when the container is used as a set.
-// The failure message reports elements that are in one of the operands but not
-// the other. The failure messages do not report duplicate or out-of-order
-// elements in the containers (which don't properly matter to sets, but can
-// occur if the containers are vectors or lists, for example).
-//
-// Uses the container's const_iterator, value_type, operator ==,
-// begin(), and end().
-template <typename Container>
-class ContainerEqMatcher {
- public:
-  typedef internal::StlContainerView<Container> View;
-  typedef typename View::type StlContainer;
-  typedef typename View::const_reference StlContainerReference;
-
-  // We make a copy of expected in case the elements in it are modified
-  // after this matcher is created.
-  explicit ContainerEqMatcher(const Container& expected)
-      : expected_(View::Copy(expected)) {
-    // Makes sure the user doesn't instantiate this class template
-    // with a const or reference type.
-    (void)testing::StaticAssertTypeEq<Container,
-        GTEST_REMOVE_REFERENCE_AND_CONST_(Container)>();
-  }
-
-  void DescribeTo(::std::ostream* os) const {
-    *os << "equals ";
-    UniversalPrint(expected_, os);
-  }
-  void DescribeNegationTo(::std::ostream* os) const {
-    *os << "does not equal ";
-    UniversalPrint(expected_, os);
-  }
-
-  template <typename LhsContainer>
-  bool MatchAndExplain(const LhsContainer& lhs,
-                       MatchResultListener* listener) const {
-    // GTEST_REMOVE_CONST_() is needed to work around an MSVC 8.0 bug
-    // that causes LhsContainer to be a const type sometimes.
-    typedef internal::StlContainerView<GTEST_REMOVE_CONST_(LhsContainer)>
-        LhsView;
-    typedef typename LhsView::type LhsStlContainer;
-    StlContainerReference lhs_stl_container = LhsView::ConstReference(lhs);
-    if (lhs_stl_container == expected_)
-      return true;
-
-    ::std::ostream* const os = listener->stream();
-    if (os != NULL) {
-      // Something is different. Check for extra values first.
-      bool printed_header = false;
-      for (typename LhsStlContainer::const_iterator it =
-               lhs_stl_container.begin();
-           it != lhs_stl_container.end(); ++it) {
-        if (internal::ArrayAwareFind(expected_.begin(), expected_.end(), *it) ==
-            expected_.end()) {
-          if (printed_header) {
-            *os << ", ";
-          } else {
-            *os << "which has these unexpected elements: ";
-            printed_header = true;
-          }
-          UniversalPrint(*it, os);
-        }
-      }
-
-      // Now check for missing values.
-      bool printed_header2 = false;
-      for (typename StlContainer::const_iterator it = expected_.begin();
-           it != expected_.end(); ++it) {
-        if (internal::ArrayAwareFind(
-                lhs_stl_container.begin(), lhs_stl_container.end(), *it) ==
-            lhs_stl_container.end()) {
-          if (printed_header2) {
-            *os << ", ";
-          } else {
-            *os << (printed_header ? ",\nand" : "which")
-                << " doesn't have these expected elements: ";
-            printed_header2 = true;
-          }
-          UniversalPrint(*it, os);
-        }
-      }
-    }
-
-    return false;
-  }
-
- private:
-  const StlContainer expected_;
-
-  GTEST_DISALLOW_ASSIGN_(ContainerEqMatcher);
-};
-
-// A comparator functor that uses the < operator to compare two values.
-struct LessComparator {
-  template <typename T, typename U>
-  bool operator()(const T& lhs, const U& rhs) const { return lhs < rhs; }
-};
-
-// Implements WhenSortedBy(comparator, container_matcher).
-template <typename Comparator, typename ContainerMatcher>
-class WhenSortedByMatcher {
- public:
-  WhenSortedByMatcher(const Comparator& comparator,
-                      const ContainerMatcher& matcher)
-      : comparator_(comparator), matcher_(matcher) {}
-
-  template <typename LhsContainer>
-  operator Matcher<LhsContainer>() const {
-    return MakeMatcher(new Impl<LhsContainer>(comparator_, matcher_));
-  }
-
-  template <typename LhsContainer>
-  class Impl : public MatcherInterface<LhsContainer> {
-   public:
-    typedef internal::StlContainerView<
-         GTEST_REMOVE_REFERENCE_AND_CONST_(LhsContainer)> LhsView;
-    typedef typename LhsView::type LhsStlContainer;
-    typedef typename LhsView::const_reference LhsStlContainerReference;
-    // Transforms std::pair<const Key, Value> into std::pair<Key, Value>
-    // so that we can match associative containers.
-    typedef typename RemoveConstFromKey<
-        typename LhsStlContainer::value_type>::type LhsValue;
-
-    Impl(const Comparator& comparator, const ContainerMatcher& matcher)
-        : comparator_(comparator), matcher_(matcher) {}
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "(when sorted) ";
-      matcher_.DescribeTo(os);
-    }
-
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << "(when sorted) ";
-      matcher_.DescribeNegationTo(os);
-    }
-
-    virtual bool MatchAndExplain(LhsContainer lhs,
-                                 MatchResultListener* listener) const {
-      LhsStlContainerReference lhs_stl_container = LhsView::ConstReference(lhs);
-      ::std::vector<LhsValue> sorted_container(lhs_stl_container.begin(),
-                                               lhs_stl_container.end());
-      ::std::sort(
-           sorted_container.begin(), sorted_container.end(), comparator_);
-
-      if (!listener->IsInterested()) {
-        // If the listener is not interested, we do not need to
-        // construct the inner explanation.
-        return matcher_.Matches(sorted_container);
-      }
-
-      *listener << "which is ";
-      UniversalPrint(sorted_container, listener->stream());
-      *listener << " when sorted";
-
-      StringMatchResultListener inner_listener;
-      const bool match = matcher_.MatchAndExplain(sorted_container,
-                                                  &inner_listener);
-      PrintIfNotEmpty(inner_listener.str(), listener->stream());
-      return match;
-    }
-
-   private:
-    const Comparator comparator_;
-    const Matcher<const ::std::vector<LhsValue>&> matcher_;
-
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(Impl);
-  };
-
- private:
-  const Comparator comparator_;
-  const ContainerMatcher matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(WhenSortedByMatcher);
-};
-
-// Implements Pointwise(tuple_matcher, rhs_container).  tuple_matcher
-// must be able to be safely cast to Matcher<tuple<const T1&, const
-// T2&> >, where T1 and T2 are the types of elements in the LHS
-// container and the RHS container respectively.
-template <typename TupleMatcher, typename RhsContainer>
-class PointwiseMatcher {
-  GTEST_COMPILE_ASSERT_(
-      !IsHashTable<GTEST_REMOVE_REFERENCE_AND_CONST_(RhsContainer)>::value,
-      use_UnorderedPointwise_with_hash_tables);
-
- public:
-  typedef internal::StlContainerView<RhsContainer> RhsView;
-  typedef typename RhsView::type RhsStlContainer;
-  typedef typename RhsStlContainer::value_type RhsValue;
-
-  // Like ContainerEq, we make a copy of rhs in case the elements in
-  // it are modified after this matcher is created.
-  PointwiseMatcher(const TupleMatcher& tuple_matcher, const RhsContainer& rhs)
-      : tuple_matcher_(tuple_matcher), rhs_(RhsView::Copy(rhs)) {
-    // Makes sure the user doesn't instantiate this class template
-    // with a const or reference type.
-    (void)testing::StaticAssertTypeEq<RhsContainer,
-        GTEST_REMOVE_REFERENCE_AND_CONST_(RhsContainer)>();
-  }
-
-  template <typename LhsContainer>
-  operator Matcher<LhsContainer>() const {
-    GTEST_COMPILE_ASSERT_(
-        !IsHashTable<GTEST_REMOVE_REFERENCE_AND_CONST_(LhsContainer)>::value,
-        use_UnorderedPointwise_with_hash_tables);
-
-    return MakeMatcher(new Impl<LhsContainer>(tuple_matcher_, rhs_));
-  }
-
-  template <typename LhsContainer>
-  class Impl : public MatcherInterface<LhsContainer> {
-   public:
-    typedef internal::StlContainerView<
-         GTEST_REMOVE_REFERENCE_AND_CONST_(LhsContainer)> LhsView;
-    typedef typename LhsView::type LhsStlContainer;
-    typedef typename LhsView::const_reference LhsStlContainerReference;
-    typedef typename LhsStlContainer::value_type LhsValue;
-    // We pass the LHS value and the RHS value to the inner matcher by
-    // reference, as they may be expensive to copy.  We must use tuple
-    // instead of pair here, as a pair cannot hold references (C++ 98,
-    // 20.2.2 [lib.pairs]).
-    typedef ::testing::tuple<const LhsValue&, const RhsValue&> InnerMatcherArg;
-
-    Impl(const TupleMatcher& tuple_matcher, const RhsStlContainer& rhs)
-        // mono_tuple_matcher_ holds a monomorphic version of the tuple matcher.
-        : mono_tuple_matcher_(SafeMatcherCast<InnerMatcherArg>(tuple_matcher)),
-          rhs_(rhs) {}
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "contains " << rhs_.size()
-          << " values, where each value and its corresponding value in ";
-      UniversalPrinter<RhsStlContainer>::Print(rhs_, os);
-      *os << " ";
-      mono_tuple_matcher_.DescribeTo(os);
-    }
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << "doesn't contain exactly " << rhs_.size()
-          << " values, or contains a value x at some index i"
-          << " where x and the i-th value of ";
-      UniversalPrint(rhs_, os);
-      *os << " ";
-      mono_tuple_matcher_.DescribeNegationTo(os);
-    }
-
-    virtual bool MatchAndExplain(LhsContainer lhs,
-                                 MatchResultListener* listener) const {
-      LhsStlContainerReference lhs_stl_container = LhsView::ConstReference(lhs);
-      const size_t actual_size = lhs_stl_container.size();
-      if (actual_size != rhs_.size()) {
-        *listener << "which contains " << actual_size << " values";
-        return false;
-      }
-
-      typename LhsStlContainer::const_iterator left = lhs_stl_container.begin();
-      typename RhsStlContainer::const_iterator right = rhs_.begin();
-      for (size_t i = 0; i != actual_size; ++i, ++left, ++right) {
-        if (listener->IsInterested()) {
-          StringMatchResultListener inner_listener;
-          // Create InnerMatcherArg as a temporarily object to avoid it outlives
-          // *left and *right. Dereference or the conversion to `const T&` may
-          // return temp objects, e.g for vector<bool>.
-          if (!mono_tuple_matcher_.MatchAndExplain(
-                  InnerMatcherArg(ImplicitCast_<const LhsValue&>(*left),
-                                  ImplicitCast_<const RhsValue&>(*right)),
-                  &inner_listener)) {
-            *listener << "where the value pair (";
-            UniversalPrint(*left, listener->stream());
-            *listener << ", ";
-            UniversalPrint(*right, listener->stream());
-            *listener << ") at index #" << i << " don't match";
-            PrintIfNotEmpty(inner_listener.str(), listener->stream());
-            return false;
-          }
-        } else {
-          if (!mono_tuple_matcher_.Matches(
-                  InnerMatcherArg(ImplicitCast_<const LhsValue&>(*left),
-                                  ImplicitCast_<const RhsValue&>(*right))))
-            return false;
-        }
-      }
-
-      return true;
-    }
-
-   private:
-    const Matcher<InnerMatcherArg> mono_tuple_matcher_;
-    const RhsStlContainer rhs_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
- private:
-  const TupleMatcher tuple_matcher_;
-  const RhsStlContainer rhs_;
-
-  GTEST_DISALLOW_ASSIGN_(PointwiseMatcher);
-};
-
-// Holds the logic common to ContainsMatcherImpl and EachMatcherImpl.
-template <typename Container>
-class QuantifierMatcherImpl : public MatcherInterface<Container> {
- public:
-  typedef GTEST_REMOVE_REFERENCE_AND_CONST_(Container) RawContainer;
-  typedef StlContainerView<RawContainer> View;
-  typedef typename View::type StlContainer;
-  typedef typename View::const_reference StlContainerReference;
-  typedef typename StlContainer::value_type Element;
-
-  template <typename InnerMatcher>
-  explicit QuantifierMatcherImpl(InnerMatcher inner_matcher)
-      : inner_matcher_(
-           testing::SafeMatcherCast<const Element&>(inner_matcher)) {}
-
-  // Checks whether:
-  // * All elements in the container match, if all_elements_should_match.
-  // * Any element in the container matches, if !all_elements_should_match.
-  bool MatchAndExplainImpl(bool all_elements_should_match,
-                           Container container,
-                           MatchResultListener* listener) const {
-    StlContainerReference stl_container = View::ConstReference(container);
-    size_t i = 0;
-    for (typename StlContainer::const_iterator it = stl_container.begin();
-         it != stl_container.end(); ++it, ++i) {
-      StringMatchResultListener inner_listener;
-      const bool matches = inner_matcher_.MatchAndExplain(*it, &inner_listener);
-
-      if (matches != all_elements_should_match) {
-        *listener << "whose element #" << i
-                  << (matches ? " matches" : " doesn't match");
-        PrintIfNotEmpty(inner_listener.str(), listener->stream());
-        return !all_elements_should_match;
-      }
-    }
-    return all_elements_should_match;
-  }
-
- protected:
-  const Matcher<const Element&> inner_matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(QuantifierMatcherImpl);
-};
-
-// Implements Contains(element_matcher) for the given argument type Container.
-// Symmetric to EachMatcherImpl.
-template <typename Container>
-class ContainsMatcherImpl : public QuantifierMatcherImpl<Container> {
- public:
-  template <typename InnerMatcher>
-  explicit ContainsMatcherImpl(InnerMatcher inner_matcher)
-      : QuantifierMatcherImpl<Container>(inner_matcher) {}
-
-  // Describes what this matcher does.
-  virtual void DescribeTo(::std::ostream* os) const {
-    *os << "contains at least one element that ";
-    this->inner_matcher_.DescribeTo(os);
-  }
-
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    *os << "doesn't contain any element that ";
-    this->inner_matcher_.DescribeTo(os);
-  }
-
-  virtual bool MatchAndExplain(Container container,
-                               MatchResultListener* listener) const {
-    return this->MatchAndExplainImpl(false, container, listener);
-  }
-
- private:
-  GTEST_DISALLOW_ASSIGN_(ContainsMatcherImpl);
-};
-
-// Implements Each(element_matcher) for the given argument type Container.
-// Symmetric to ContainsMatcherImpl.
-template <typename Container>
-class EachMatcherImpl : public QuantifierMatcherImpl<Container> {
- public:
-  template <typename InnerMatcher>
-  explicit EachMatcherImpl(InnerMatcher inner_matcher)
-      : QuantifierMatcherImpl<Container>(inner_matcher) {}
-
-  // Describes what this matcher does.
-  virtual void DescribeTo(::std::ostream* os) const {
-    *os << "only contains elements that ";
-    this->inner_matcher_.DescribeTo(os);
-  }
-
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    *os << "contains some element that ";
-    this->inner_matcher_.DescribeNegationTo(os);
-  }
-
-  virtual bool MatchAndExplain(Container container,
-                               MatchResultListener* listener) const {
-    return this->MatchAndExplainImpl(true, container, listener);
-  }
-
- private:
-  GTEST_DISALLOW_ASSIGN_(EachMatcherImpl);
-};
-
-// Implements polymorphic Contains(element_matcher).
-template <typename M>
-class ContainsMatcher {
- public:
-  explicit ContainsMatcher(M m) : inner_matcher_(m) {}
-
-  template <typename Container>
-  operator Matcher<Container>() const {
-    return MakeMatcher(new ContainsMatcherImpl<Container>(inner_matcher_));
-  }
-
- private:
-  const M inner_matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(ContainsMatcher);
-};
-
-// Implements polymorphic Each(element_matcher).
-template <typename M>
-class EachMatcher {
- public:
-  explicit EachMatcher(M m) : inner_matcher_(m) {}
-
-  template <typename Container>
-  operator Matcher<Container>() const {
-    return MakeMatcher(new EachMatcherImpl<Container>(inner_matcher_));
-  }
-
- private:
-  const M inner_matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(EachMatcher);
-};
-
-struct Rank1 {};
-struct Rank0 : Rank1 {};
-
-namespace pair_getters {
-#if GTEST_LANG_CXX11
-using std::get;
-template <typename T>
-auto First(T& x, Rank1) -> decltype(get<0>(x)) {  // NOLINT
-  return get<0>(x);
-}
-template <typename T>
-auto First(T& x, Rank0) -> decltype((x.first)) {  // NOLINT
-  return x.first;
-}
-
-template <typename T>
-auto Second(T& x, Rank1) -> decltype(get<1>(x)) {  // NOLINT
-  return get<1>(x);
-}
-template <typename T>
-auto Second(T& x, Rank0) -> decltype((x.second)) {  // NOLINT
-  return x.second;
-}
-#else
-template <typename T>
-typename T::first_type& First(T& x, Rank0) {  // NOLINT
-  return x.first;
-}
-template <typename T>
-const typename T::first_type& First(const T& x, Rank0) {
-  return x.first;
-}
-
-template <typename T>
-typename T::second_type& Second(T& x, Rank0) {  // NOLINT
-  return x.second;
-}
-template <typename T>
-const typename T::second_type& Second(const T& x, Rank0) {
-  return x.second;
-}
-#endif  // GTEST_LANG_CXX11
-}  // namespace pair_getters
-
-// Implements Key(inner_matcher) for the given argument pair type.
-// Key(inner_matcher) matches an std::pair whose 'first' field matches
-// inner_matcher.  For example, Contains(Key(Ge(5))) can be used to match an
-// std::map that contains at least one element whose key is >= 5.
-template <typename PairType>
-class KeyMatcherImpl : public MatcherInterface<PairType> {
- public:
-  typedef GTEST_REMOVE_REFERENCE_AND_CONST_(PairType) RawPairType;
-  typedef typename RawPairType::first_type KeyType;
-
-  template <typename InnerMatcher>
-  explicit KeyMatcherImpl(InnerMatcher inner_matcher)
-      : inner_matcher_(
-          testing::SafeMatcherCast<const KeyType&>(inner_matcher)) {
-  }
-
-  // Returns true iff 'key_value.first' (the key) matches the inner matcher.
-  virtual bool MatchAndExplain(PairType key_value,
-                               MatchResultListener* listener) const {
-    StringMatchResultListener inner_listener;
-    const bool match = inner_matcher_.MatchAndExplain(
-        pair_getters::First(key_value, Rank0()), &inner_listener);
-    const std::string explanation = inner_listener.str();
-    if (explanation != "") {
-      *listener << "whose first field is a value " << explanation;
-    }
-    return match;
-  }
-
-  // Describes what this matcher does.
-  virtual void DescribeTo(::std::ostream* os) const {
-    *os << "has a key that ";
-    inner_matcher_.DescribeTo(os);
-  }
-
-  // Describes what the negation of this matcher does.
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    *os << "doesn't have a key that ";
-    inner_matcher_.DescribeTo(os);
-  }
-
- private:
-  const Matcher<const KeyType&> inner_matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(KeyMatcherImpl);
-};
-
-// Implements polymorphic Key(matcher_for_key).
-template <typename M>
-class KeyMatcher {
- public:
-  explicit KeyMatcher(M m) : matcher_for_key_(m) {}
-
-  template <typename PairType>
-  operator Matcher<PairType>() const {
-    return MakeMatcher(new KeyMatcherImpl<PairType>(matcher_for_key_));
-  }
-
- private:
-  const M matcher_for_key_;
-
-  GTEST_DISALLOW_ASSIGN_(KeyMatcher);
-};
-
-// Implements Pair(first_matcher, second_matcher) for the given argument pair
-// type with its two matchers. See Pair() function below.
-template <typename PairType>
-class PairMatcherImpl : public MatcherInterface<PairType> {
- public:
-  typedef GTEST_REMOVE_REFERENCE_AND_CONST_(PairType) RawPairType;
-  typedef typename RawPairType::first_type FirstType;
-  typedef typename RawPairType::second_type SecondType;
-
-  template <typename FirstMatcher, typename SecondMatcher>
-  PairMatcherImpl(FirstMatcher first_matcher, SecondMatcher second_matcher)
-      : first_matcher_(
-            testing::SafeMatcherCast<const FirstType&>(first_matcher)),
-        second_matcher_(
-            testing::SafeMatcherCast<const SecondType&>(second_matcher)) {
-  }
-
-  // Describes what this matcher does.
-  virtual void DescribeTo(::std::ostream* os) const {
-    *os << "has a first field that ";
-    first_matcher_.DescribeTo(os);
-    *os << ", and has a second field that ";
-    second_matcher_.DescribeTo(os);
-  }
-
-  // Describes what the negation of this matcher does.
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    *os << "has a first field that ";
-    first_matcher_.DescribeNegationTo(os);
-    *os << ", or has a second field that ";
-    second_matcher_.DescribeNegationTo(os);
-  }
-
-  // Returns true iff 'a_pair.first' matches first_matcher and 'a_pair.second'
-  // matches second_matcher.
-  virtual bool MatchAndExplain(PairType a_pair,
-                               MatchResultListener* listener) const {
-    if (!listener->IsInterested()) {
-      // If the listener is not interested, we don't need to construct the
-      // explanation.
-      return first_matcher_.Matches(pair_getters::First(a_pair, Rank0())) &&
-             second_matcher_.Matches(pair_getters::Second(a_pair, Rank0()));
-    }
-    StringMatchResultListener first_inner_listener;
-    if (!first_matcher_.MatchAndExplain(pair_getters::First(a_pair, Rank0()),
-                                        &first_inner_listener)) {
-      *listener << "whose first field does not match";
-      PrintIfNotEmpty(first_inner_listener.str(), listener->stream());
-      return false;
-    }
-    StringMatchResultListener second_inner_listener;
-    if (!second_matcher_.MatchAndExplain(pair_getters::Second(a_pair, Rank0()),
-                                         &second_inner_listener)) {
-      *listener << "whose second field does not match";
-      PrintIfNotEmpty(second_inner_listener.str(), listener->stream());
-      return false;
-    }
-    ExplainSuccess(first_inner_listener.str(), second_inner_listener.str(),
-                   listener);
-    return true;
-  }
-
- private:
-  void ExplainSuccess(const std::string& first_explanation,
-                      const std::string& second_explanation,
-                      MatchResultListener* listener) const {
-    *listener << "whose both fields match";
-    if (first_explanation != "") {
-      *listener << ", where the first field is a value " << first_explanation;
-    }
-    if (second_explanation != "") {
-      *listener << ", ";
-      if (first_explanation != "") {
-        *listener << "and ";
-      } else {
-        *listener << "where ";
-      }
-      *listener << "the second field is a value " << second_explanation;
-    }
-  }
-
-  const Matcher<const FirstType&> first_matcher_;
-  const Matcher<const SecondType&> second_matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(PairMatcherImpl);
-};
-
-// Implements polymorphic Pair(first_matcher, second_matcher).
-template <typename FirstMatcher, typename SecondMatcher>
-class PairMatcher {
- public:
-  PairMatcher(FirstMatcher first_matcher, SecondMatcher second_matcher)
-      : first_matcher_(first_matcher), second_matcher_(second_matcher) {}
-
-  template <typename PairType>
-  operator Matcher<PairType> () const {
-    return MakeMatcher(
-        new PairMatcherImpl<PairType>(
-            first_matcher_, second_matcher_));
-  }
-
- private:
-  const FirstMatcher first_matcher_;
-  const SecondMatcher second_matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(PairMatcher);
-};
-
-// Implements ElementsAre() and ElementsAreArray().
-template <typename Container>
-class ElementsAreMatcherImpl : public MatcherInterface<Container> {
- public:
-  typedef GTEST_REMOVE_REFERENCE_AND_CONST_(Container) RawContainer;
-  typedef internal::StlContainerView<RawContainer> View;
-  typedef typename View::type StlContainer;
-  typedef typename View::const_reference StlContainerReference;
-  typedef typename StlContainer::value_type Element;
-
-  // Constructs the matcher from a sequence of element values or
-  // element matchers.
-  template <typename InputIter>
-  ElementsAreMatcherImpl(InputIter first, InputIter last) {
-    while (first != last) {
-      matchers_.push_back(MatcherCast<const Element&>(*first++));
-    }
-  }
-
-  // Describes what this matcher does.
-  virtual void DescribeTo(::std::ostream* os) const {
-    if (count() == 0) {
-      *os << "is empty";
-    } else if (count() == 1) {
-      *os << "has 1 element that ";
-      matchers_[0].DescribeTo(os);
-    } else {
-      *os << "has " << Elements(count()) << " where\n";
-      for (size_t i = 0; i != count(); ++i) {
-        *os << "element #" << i << " ";
-        matchers_[i].DescribeTo(os);
-        if (i + 1 < count()) {
-          *os << ",\n";
-        }
-      }
-    }
-  }
-
-  // Describes what the negation of this matcher does.
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    if (count() == 0) {
-      *os << "isn't empty";
-      return;
-    }
-
-    *os << "doesn't have " << Elements(count()) << ", or\n";
-    for (size_t i = 0; i != count(); ++i) {
-      *os << "element #" << i << " ";
-      matchers_[i].DescribeNegationTo(os);
-      if (i + 1 < count()) {
-        *os << ", or\n";
-      }
-    }
-  }
-
-  virtual bool MatchAndExplain(Container container,
-                               MatchResultListener* listener) const {
-    // To work with stream-like "containers", we must only walk
-    // through the elements in one pass.
-
-    const bool listener_interested = listener->IsInterested();
-
-    // explanations[i] is the explanation of the element at index i.
-    ::std::vector<std::string> explanations(count());
-    StlContainerReference stl_container = View::ConstReference(container);
-    typename StlContainer::const_iterator it = stl_container.begin();
-    size_t exam_pos = 0;
-    bool mismatch_found = false;  // Have we found a mismatched element yet?
-
-    // Go through the elements and matchers in pairs, until we reach
-    // the end of either the elements or the matchers, or until we find a
-    // mismatch.
-    for (; it != stl_container.end() && exam_pos != count(); ++it, ++exam_pos) {
-      bool match;  // Does the current element match the current matcher?
-      if (listener_interested) {
-        StringMatchResultListener s;
-        match = matchers_[exam_pos].MatchAndExplain(*it, &s);
-        explanations[exam_pos] = s.str();
-      } else {
-        match = matchers_[exam_pos].Matches(*it);
-      }
-
-      if (!match) {
-        mismatch_found = true;
-        break;
-      }
-    }
-    // If mismatch_found is true, 'exam_pos' is the index of the mismatch.
-
-    // Find how many elements the actual container has.  We avoid
-    // calling size() s.t. this code works for stream-like "containers"
-    // that don't define size().
-    size_t actual_count = exam_pos;
-    for (; it != stl_container.end(); ++it) {
-      ++actual_count;
-    }
-
-    if (actual_count != count()) {
-      // The element count doesn't match.  If the container is empty,
-      // there's no need to explain anything as Google Mock already
-      // prints the empty container.  Otherwise we just need to show
-      // how many elements there actually are.
-      if (listener_interested && (actual_count != 0)) {
-        *listener << "which has " << Elements(actual_count);
-      }
-      return false;
-    }
-
-    if (mismatch_found) {
-      // The element count matches, but the exam_pos-th element doesn't match.
-      if (listener_interested) {
-        *listener << "whose element #" << exam_pos << " doesn't match";
-        PrintIfNotEmpty(explanations[exam_pos], listener->stream());
-      }
-      return false;
-    }
-
-    // Every element matches its expectation.  We need to explain why
-    // (the obvious ones can be skipped).
-    if (listener_interested) {
-      bool reason_printed = false;
-      for (size_t i = 0; i != count(); ++i) {
-        const std::string& s = explanations[i];
-        if (!s.empty()) {
-          if (reason_printed) {
-            *listener << ",\nand ";
-          }
-          *listener << "whose element #" << i << " matches, " << s;
-          reason_printed = true;
-        }
-      }
-    }
-    return true;
-  }
-
- private:
-  static Message Elements(size_t count) {
-    return Message() << count << (count == 1 ? " element" : " elements");
-  }
-
-  size_t count() const { return matchers_.size(); }
-
-  ::std::vector<Matcher<const Element&> > matchers_;
-
-  GTEST_DISALLOW_ASSIGN_(ElementsAreMatcherImpl);
-};
-
-// Connectivity matrix of (elements X matchers), in element-major order.
-// Initially, there are no edges.
-// Use NextGraph() to iterate over all possible edge configurations.
-// Use Randomize() to generate a random edge configuration.
-class GTEST_API_ MatchMatrix {
- public:
-  MatchMatrix(size_t num_elements, size_t num_matchers)
-      : num_elements_(num_elements),
-        num_matchers_(num_matchers),
-        matched_(num_elements_* num_matchers_, 0) {
-  }
-
-  size_t LhsSize() const { return num_elements_; }
-  size_t RhsSize() const { return num_matchers_; }
-  bool HasEdge(size_t ilhs, size_t irhs) const {
-    return matched_[SpaceIndex(ilhs, irhs)] == 1;
-  }
-  void SetEdge(size_t ilhs, size_t irhs, bool b) {
-    matched_[SpaceIndex(ilhs, irhs)] = b ? 1 : 0;
-  }
-
-  // Treating the connectivity matrix as a (LhsSize()*RhsSize())-bit number,
-  // adds 1 to that number; returns false if incrementing the graph left it
-  // empty.
-  bool NextGraph();
-
-  void Randomize();
-
-  std::string DebugString() const;
-
- private:
-  size_t SpaceIndex(size_t ilhs, size_t irhs) const {
-    return ilhs * num_matchers_ + irhs;
-  }
-
-  size_t num_elements_;
-  size_t num_matchers_;
-
-  // Each element is a char interpreted as bool. They are stored as a
-  // flattened array in lhs-major order, use 'SpaceIndex()' to translate
-  // a (ilhs, irhs) matrix coordinate into an offset.
-  ::std::vector<char> matched_;
-};
-
-typedef ::std::pair<size_t, size_t> ElementMatcherPair;
-typedef ::std::vector<ElementMatcherPair> ElementMatcherPairs;
-
-// Returns a maximum bipartite matching for the specified graph 'g'.
-// The matching is represented as a vector of {element, matcher} pairs.
-GTEST_API_ ElementMatcherPairs
-FindMaxBipartiteMatching(const MatchMatrix& g);
-
-struct UnorderedMatcherRequire {
-  enum Flags {
-    Superset = 1 << 0,
-    Subset = 1 << 1,
-    ExactMatch = Superset | Subset,
-  };
-};
-
-// Untyped base class for implementing UnorderedElementsAre.  By
-// putting logic that's not specific to the element type here, we
-// reduce binary bloat and increase compilation speed.
-class GTEST_API_ UnorderedElementsAreMatcherImplBase {
- protected:
-  explicit UnorderedElementsAreMatcherImplBase(
-      UnorderedMatcherRequire::Flags matcher_flags)
-      : match_flags_(matcher_flags) {}
-
-  // A vector of matcher describers, one for each element matcher.
-  // Does not own the describers (and thus can be used only when the
-  // element matchers are alive).
-  typedef ::std::vector<const MatcherDescriberInterface*> MatcherDescriberVec;
-
-  // Describes this UnorderedElementsAre matcher.
-  void DescribeToImpl(::std::ostream* os) const;
-
-  // Describes the negation of this UnorderedElementsAre matcher.
-  void DescribeNegationToImpl(::std::ostream* os) const;
-
-  bool VerifyMatchMatrix(const ::std::vector<std::string>& element_printouts,
-                         const MatchMatrix& matrix,
-                         MatchResultListener* listener) const;
-
-  bool FindPairing(const MatchMatrix& matrix,
-                   MatchResultListener* listener) const;
-
-  MatcherDescriberVec& matcher_describers() {
-    return matcher_describers_;
-  }
-
-  static Message Elements(size_t n) {
-    return Message() << n << " element" << (n == 1 ? "" : "s");
-  }
-
-  UnorderedMatcherRequire::Flags match_flags() const { return match_flags_; }
-
- private:
-  UnorderedMatcherRequire::Flags match_flags_;
-  MatcherDescriberVec matcher_describers_;
-
-  GTEST_DISALLOW_ASSIGN_(UnorderedElementsAreMatcherImplBase);
-};
-
-// Implements UnorderedElementsAre, UnorderedElementsAreArray, IsSubsetOf, and
-// IsSupersetOf.
-template <typename Container>
-class UnorderedElementsAreMatcherImpl
-    : public MatcherInterface<Container>,
-      public UnorderedElementsAreMatcherImplBase {
- public:
-  typedef GTEST_REMOVE_REFERENCE_AND_CONST_(Container) RawContainer;
-  typedef internal::StlContainerView<RawContainer> View;
-  typedef typename View::type StlContainer;
-  typedef typename View::const_reference StlContainerReference;
-  typedef typename StlContainer::const_iterator StlContainerConstIterator;
-  typedef typename StlContainer::value_type Element;
-
-  template <typename InputIter>
-  UnorderedElementsAreMatcherImpl(UnorderedMatcherRequire::Flags matcher_flags,
-                                  InputIter first, InputIter last)
-      : UnorderedElementsAreMatcherImplBase(matcher_flags) {
-    for (; first != last; ++first) {
-      matchers_.push_back(MatcherCast<const Element&>(*first));
-      matcher_describers().push_back(matchers_.back().GetDescriber());
-    }
-  }
-
-  // Describes what this matcher does.
-  virtual void DescribeTo(::std::ostream* os) const {
-    return UnorderedElementsAreMatcherImplBase::DescribeToImpl(os);
-  }
-
-  // Describes what the negation of this matcher does.
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    return UnorderedElementsAreMatcherImplBase::DescribeNegationToImpl(os);
-  }
-
-  virtual bool MatchAndExplain(Container container,
-                               MatchResultListener* listener) const {
-    StlContainerReference stl_container = View::ConstReference(container);
-    ::std::vector<std::string> element_printouts;
-    MatchMatrix matrix =
-        AnalyzeElements(stl_container.begin(), stl_container.end(),
-                        &element_printouts, listener);
-
-    if (matrix.LhsSize() == 0 && matrix.RhsSize() == 0) {
-      return true;
-    }
-
-    if (match_flags() == UnorderedMatcherRequire::ExactMatch) {
-      if (matrix.LhsSize() != matrix.RhsSize()) {
-        // The element count doesn't match.  If the container is empty,
-        // there's no need to explain anything as Google Mock already
-        // prints the empty container. Otherwise we just need to show
-        // how many elements there actually are.
-        if (matrix.LhsSize() != 0 && listener->IsInterested()) {
-          *listener << "which has " << Elements(matrix.LhsSize());
-        }
-        return false;
-      }
-    }
-
-    return VerifyMatchMatrix(element_printouts, matrix, listener) &&
-           FindPairing(matrix, listener);
-  }
-
- private:
-  template <typename ElementIter>
-  MatchMatrix AnalyzeElements(ElementIter elem_first, ElementIter elem_last,
-                              ::std::vector<std::string>* element_printouts,
-                              MatchResultListener* listener) const {
-    element_printouts->clear();
-    ::std::vector<char> did_match;
-    size_t num_elements = 0;
-    for (; elem_first != elem_last; ++num_elements, ++elem_first) {
-      if (listener->IsInterested()) {
-        element_printouts->push_back(PrintToString(*elem_first));
-      }
-      for (size_t irhs = 0; irhs != matchers_.size(); ++irhs) {
-        did_match.push_back(Matches(matchers_[irhs])(*elem_first));
-      }
-    }
-
-    MatchMatrix matrix(num_elements, matchers_.size());
-    ::std::vector<char>::const_iterator did_match_iter = did_match.begin();
-    for (size_t ilhs = 0; ilhs != num_elements; ++ilhs) {
-      for (size_t irhs = 0; irhs != matchers_.size(); ++irhs) {
-        matrix.SetEdge(ilhs, irhs, *did_match_iter++ != 0);
-      }
-    }
-    return matrix;
-  }
-
-  ::std::vector<Matcher<const Element&> > matchers_;
-
-  GTEST_DISALLOW_ASSIGN_(UnorderedElementsAreMatcherImpl);
-};
-
-// Functor for use in TransformTuple.
-// Performs MatcherCast<Target> on an input argument of any type.
-template <typename Target>
-struct CastAndAppendTransform {
-  template <typename Arg>
-  Matcher<Target> operator()(const Arg& a) const {
-    return MatcherCast<Target>(a);
-  }
-};
-
-// Implements UnorderedElementsAre.
-template <typename MatcherTuple>
-class UnorderedElementsAreMatcher {
- public:
-  explicit UnorderedElementsAreMatcher(const MatcherTuple& args)
-      : matchers_(args) {}
-
-  template <typename Container>
-  operator Matcher<Container>() const {
-    typedef GTEST_REMOVE_REFERENCE_AND_CONST_(Container) RawContainer;
-    typedef typename internal::StlContainerView<RawContainer>::type View;
-    typedef typename View::value_type Element;
-    typedef ::std::vector<Matcher<const Element&> > MatcherVec;
-    MatcherVec matchers;
-    matchers.reserve(::testing::tuple_size<MatcherTuple>::value);
-    TransformTupleValues(CastAndAppendTransform<const Element&>(), matchers_,
-                         ::std::back_inserter(matchers));
-    return MakeMatcher(new UnorderedElementsAreMatcherImpl<Container>(
-        UnorderedMatcherRequire::ExactMatch, matchers.begin(), matchers.end()));
-  }
-
- private:
-  const MatcherTuple matchers_;
-  GTEST_DISALLOW_ASSIGN_(UnorderedElementsAreMatcher);
-};
-
-// Implements ElementsAre.
-template <typename MatcherTuple>
-class ElementsAreMatcher {
- public:
-  explicit ElementsAreMatcher(const MatcherTuple& args) : matchers_(args) {}
-
-  template <typename Container>
-  operator Matcher<Container>() const {
-    GTEST_COMPILE_ASSERT_(
-        !IsHashTable<GTEST_REMOVE_REFERENCE_AND_CONST_(Container)>::value ||
-            ::testing::tuple_size<MatcherTuple>::value < 2,
-        use_UnorderedElementsAre_with_hash_tables);
-
-    typedef GTEST_REMOVE_REFERENCE_AND_CONST_(Container) RawContainer;
-    typedef typename internal::StlContainerView<RawContainer>::type View;
-    typedef typename View::value_type Element;
-    typedef ::std::vector<Matcher<const Element&> > MatcherVec;
-    MatcherVec matchers;
-    matchers.reserve(::testing::tuple_size<MatcherTuple>::value);
-    TransformTupleValues(CastAndAppendTransform<const Element&>(), matchers_,
-                         ::std::back_inserter(matchers));
-    return MakeMatcher(new ElementsAreMatcherImpl<Container>(
-                           matchers.begin(), matchers.end()));
-  }
-
- private:
-  const MatcherTuple matchers_;
-  GTEST_DISALLOW_ASSIGN_(ElementsAreMatcher);
-};
-
-// Implements UnorderedElementsAreArray(), IsSubsetOf(), and IsSupersetOf().
-template <typename T>
-class UnorderedElementsAreArrayMatcher {
- public:
-  template <typename Iter>
-  UnorderedElementsAreArrayMatcher(UnorderedMatcherRequire::Flags match_flags,
-                                   Iter first, Iter last)
-      : match_flags_(match_flags), matchers_(first, last) {}
-
-  template <typename Container>
-  operator Matcher<Container>() const {
-    return MakeMatcher(new UnorderedElementsAreMatcherImpl<Container>(
-        match_flags_, matchers_.begin(), matchers_.end()));
-  }
-
- private:
-  UnorderedMatcherRequire::Flags match_flags_;
-  ::std::vector<T> matchers_;
-
-  GTEST_DISALLOW_ASSIGN_(UnorderedElementsAreArrayMatcher);
-};
-
-// Implements ElementsAreArray().
-template <typename T>
-class ElementsAreArrayMatcher {
- public:
-  template <typename Iter>
-  ElementsAreArrayMatcher(Iter first, Iter last) : matchers_(first, last) {}
-
-  template <typename Container>
-  operator Matcher<Container>() const {
-    GTEST_COMPILE_ASSERT_(
-        !IsHashTable<GTEST_REMOVE_REFERENCE_AND_CONST_(Container)>::value,
-        use_UnorderedElementsAreArray_with_hash_tables);
-
-    return MakeMatcher(new ElementsAreMatcherImpl<Container>(
-        matchers_.begin(), matchers_.end()));
-  }
-
- private:
-  const ::std::vector<T> matchers_;
-
-  GTEST_DISALLOW_ASSIGN_(ElementsAreArrayMatcher);
-};
-
-// Given a 2-tuple matcher tm of type Tuple2Matcher and a value second
-// of type Second, BoundSecondMatcher<Tuple2Matcher, Second>(tm,
-// second) is a polymorphic matcher that matches a value x iff tm
-// matches tuple (x, second).  Useful for implementing
-// UnorderedPointwise() in terms of UnorderedElementsAreArray().
-//
-// BoundSecondMatcher is copyable and assignable, as we need to put
-// instances of this class in a vector when implementing
-// UnorderedPointwise().
-template <typename Tuple2Matcher, typename Second>
-class BoundSecondMatcher {
- public:
-  BoundSecondMatcher(const Tuple2Matcher& tm, const Second& second)
-      : tuple2_matcher_(tm), second_value_(second) {}
-
-  template <typename T>
-  operator Matcher<T>() const {
-    return MakeMatcher(new Impl<T>(tuple2_matcher_, second_value_));
-  }
-
-  // We have to define this for UnorderedPointwise() to compile in
-  // C++98 mode, as it puts BoundSecondMatcher instances in a vector,
-  // which requires the elements to be assignable in C++98.  The
-  // compiler cannot generate the operator= for us, as Tuple2Matcher
-  // and Second may not be assignable.
-  //
-  // However, this should never be called, so the implementation just
-  // need to assert.
-  void operator=(const BoundSecondMatcher& /*rhs*/) {
-    GTEST_LOG_(FATAL) << "BoundSecondMatcher should never be assigned.";
-  }
-
- private:
-  template <typename T>
-  class Impl : public MatcherInterface<T> {
-   public:
-    typedef ::testing::tuple<T, Second> ArgTuple;
-
-    Impl(const Tuple2Matcher& tm, const Second& second)
-        : mono_tuple2_matcher_(SafeMatcherCast<const ArgTuple&>(tm)),
-          second_value_(second) {}
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "and ";
-      UniversalPrint(second_value_, os);
-      *os << " ";
-      mono_tuple2_matcher_.DescribeTo(os);
-    }
-
-    virtual bool MatchAndExplain(T x, MatchResultListener* listener) const {
-      return mono_tuple2_matcher_.MatchAndExplain(ArgTuple(x, second_value_),
-                                                  listener);
-    }
-
-   private:
-    const Matcher<const ArgTuple&> mono_tuple2_matcher_;
-    const Second second_value_;
-
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
-  const Tuple2Matcher tuple2_matcher_;
-  const Second second_value_;
-};
-
-// Given a 2-tuple matcher tm and a value second,
-// MatcherBindSecond(tm, second) returns a matcher that matches a
-// value x iff tm matches tuple (x, second).  Useful for implementing
-// UnorderedPointwise() in terms of UnorderedElementsAreArray().
-template <typename Tuple2Matcher, typename Second>
-BoundSecondMatcher<Tuple2Matcher, Second> MatcherBindSecond(
-    const Tuple2Matcher& tm, const Second& second) {
-  return BoundSecondMatcher<Tuple2Matcher, Second>(tm, second);
-}
-
-// Returns the description for a matcher defined using the MATCHER*()
-// macro where the user-supplied description string is "", if
-// 'negation' is false; otherwise returns the description of the
-// negation of the matcher.  'param_values' contains a list of strings
-// that are the print-out of the matcher's parameters.
-GTEST_API_ std::string FormatMatcherDescription(bool negation,
-                                                const char* matcher_name,
-                                                const Strings& param_values);
-
-// Implements a matcher that checks the value of a optional<> type variable.
-template <typename ValueMatcher>
-class OptionalMatcher {
- public:
-  explicit OptionalMatcher(const ValueMatcher& value_matcher)
-      : value_matcher_(value_matcher) {}
-
-  template <typename Optional>
-  operator Matcher<Optional>() const {
-    return MakeMatcher(new Impl<Optional>(value_matcher_));
-  }
-
-  template <typename Optional>
-  class Impl : public MatcherInterface<Optional> {
-   public:
-    typedef GTEST_REMOVE_REFERENCE_AND_CONST_(Optional) OptionalView;
-    typedef typename OptionalView::value_type ValueType;
-    explicit Impl(const ValueMatcher& value_matcher)
-        : value_matcher_(MatcherCast<ValueType>(value_matcher)) {}
-
-    virtual void DescribeTo(::std::ostream* os) const {
-      *os << "value ";
-      value_matcher_.DescribeTo(os);
-    }
-
-    virtual void DescribeNegationTo(::std::ostream* os) const {
-      *os << "value ";
-      value_matcher_.DescribeNegationTo(os);
-    }
-
-    virtual bool MatchAndExplain(Optional optional,
-                                 MatchResultListener* listener) const {
-      if (!optional) {
-        *listener << "which is not engaged";
-        return false;
-      }
-      const ValueType& value = *optional;
-      StringMatchResultListener value_listener;
-      const bool match = value_matcher_.MatchAndExplain(value, &value_listener);
-      *listener << "whose value " << PrintToString(value)
-                << (match ? " matches" : " doesn't match");
-      PrintIfNotEmpty(value_listener.str(), listener->stream());
-      return match;
-    }
-
-   private:
-    const Matcher<ValueType> value_matcher_;
-    GTEST_DISALLOW_ASSIGN_(Impl);
-  };
-
- private:
-  const ValueMatcher value_matcher_;
-  GTEST_DISALLOW_ASSIGN_(OptionalMatcher);
-};
-
-namespace variant_matcher {
-// Overloads to allow VariantMatcher to do proper ADL lookup.
-template <typename T>
-void holds_alternative() {}
-template <typename T>
-void get() {}
-
-// Implements a matcher that checks the value of a variant<> type variable.
-template <typename T>
-class VariantMatcher {
- public:
-  explicit VariantMatcher(::testing::Matcher<const T&> matcher)
-      : matcher_(internal::move(matcher)) {}
-
-  template <typename Variant>
-  bool MatchAndExplain(const Variant& value,
-                       ::testing::MatchResultListener* listener) const {
-    if (!listener->IsInterested()) {
-      return holds_alternative<T>(value) && matcher_.Matches(get<T>(value));
-    }
-
-    if (!holds_alternative<T>(value)) {
-      *listener << "whose value is not of type '" << GetTypeName() << "'";
-      return false;
-    }
-
-    const T& elem = get<T>(value);
-    StringMatchResultListener elem_listener;
-    const bool match = matcher_.MatchAndExplain(elem, &elem_listener);
-    *listener << "whose value " << PrintToString(elem)
-              << (match ? " matches" : " doesn't match");
-    PrintIfNotEmpty(elem_listener.str(), listener->stream());
-    return match;
-  }
-
-  void DescribeTo(std::ostream* os) const {
-    *os << "is a variant<> with value of type '" << GetTypeName()
-        << "' and the value ";
-    matcher_.DescribeTo(os);
-  }
-
-  void DescribeNegationTo(std::ostream* os) const {
-    *os << "is a variant<> with value of type other than '" << GetTypeName()
-        << "' or the value ";
-    matcher_.DescribeNegationTo(os);
-  }
-
- private:
-  static std::string GetTypeName() {
-#if GTEST_HAS_RTTI
-    GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(
-        return internal::GetTypeName<T>());
-#endif
-    return "the element type";
-  }
-
-  const ::testing::Matcher<const T&> matcher_;
-};
-
-}  // namespace variant_matcher
-
-namespace any_cast_matcher {
-
-// Overloads to allow AnyCastMatcher to do proper ADL lookup.
-template <typename T>
-void any_cast() {}
-
-// Implements a matcher that any_casts the value.
-template <typename T>
-class AnyCastMatcher {
- public:
-  explicit AnyCastMatcher(const ::testing::Matcher<const T&>& matcher)
-      : matcher_(matcher) {}
-
-  template <typename AnyType>
-  bool MatchAndExplain(const AnyType& value,
-                       ::testing::MatchResultListener* listener) const {
-    if (!listener->IsInterested()) {
-      const T* ptr = any_cast<T>(&value);
-      return ptr != NULL && matcher_.Matches(*ptr);
-    }
-
-    const T* elem = any_cast<T>(&value);
-    if (elem == NULL) {
-      *listener << "whose value is not of type '" << GetTypeName() << "'";
-      return false;
-    }
-
-    StringMatchResultListener elem_listener;
-    const bool match = matcher_.MatchAndExplain(*elem, &elem_listener);
-    *listener << "whose value " << PrintToString(*elem)
-              << (match ? " matches" : " doesn't match");
-    PrintIfNotEmpty(elem_listener.str(), listener->stream());
-    return match;
-  }
-
-  void DescribeTo(std::ostream* os) const {
-    *os << "is an 'any' type with value of type '" << GetTypeName()
-        << "' and the value ";
-    matcher_.DescribeTo(os);
-  }
-
-  void DescribeNegationTo(std::ostream* os) const {
-    *os << "is an 'any' type with value of type other than '" << GetTypeName()
-        << "' or the value ";
-    matcher_.DescribeNegationTo(os);
-  }
-
- private:
-  static std::string GetTypeName() {
-#if GTEST_HAS_RTTI
-    GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(
-        return internal::GetTypeName<T>());
-#endif
-    return "the element type";
-  }
-
-  const ::testing::Matcher<const T&> matcher_;
-};
-
-}  // namespace any_cast_matcher
-}  // namespace internal
-
-// ElementsAreArray(iterator_first, iterator_last)
-// ElementsAreArray(pointer, count)
-// ElementsAreArray(array)
-// ElementsAreArray(container)
-// ElementsAreArray({ e1, e2, ..., en })
-//
-// The ElementsAreArray() functions are like ElementsAre(...), except
-// that they are given a homogeneous sequence rather than taking each
-// element as a function argument. The sequence can be specified as an
-// array, a pointer and count, a vector, an initializer list, or an
-// STL iterator range. In each of these cases, the underlying sequence
-// can be either a sequence of values or a sequence of matchers.
-//
-// All forms of ElementsAreArray() make a copy of the input matcher sequence.
-
-template <typename Iter>
-inline internal::ElementsAreArrayMatcher<
-    typename ::std::iterator_traits<Iter>::value_type>
-ElementsAreArray(Iter first, Iter last) {
-  typedef typename ::std::iterator_traits<Iter>::value_type T;
-  return internal::ElementsAreArrayMatcher<T>(first, last);
-}
-
-template <typename T>
-inline internal::ElementsAreArrayMatcher<T> ElementsAreArray(
-    const T* pointer, size_t count) {
-  return ElementsAreArray(pointer, pointer + count);
-}
-
-template <typename T, size_t N>
-inline internal::ElementsAreArrayMatcher<T> ElementsAreArray(
-    const T (&array)[N]) {
-  return ElementsAreArray(array, N);
-}
-
-template <typename Container>
-inline internal::ElementsAreArrayMatcher<typename Container::value_type>
-ElementsAreArray(const Container& container) {
-  return ElementsAreArray(container.begin(), container.end());
-}
-
-#if GTEST_HAS_STD_INITIALIZER_LIST_
-template <typename T>
-inline internal::ElementsAreArrayMatcher<T>
-ElementsAreArray(::std::initializer_list<T> xs) {
-  return ElementsAreArray(xs.begin(), xs.end());
-}
-#endif
-
-// UnorderedElementsAreArray(iterator_first, iterator_last)
-// UnorderedElementsAreArray(pointer, count)
-// UnorderedElementsAreArray(array)
-// UnorderedElementsAreArray(container)
-// UnorderedElementsAreArray({ e1, e2, ..., en })
-//
-// UnorderedElementsAreArray() verifies that a bijective mapping onto a
-// collection of matchers exists.
-//
-// The matchers can be specified as an array, a pointer and count, a container,
-// an initializer list, or an STL iterator range. In each of these cases, the
-// underlying matchers can be either values or matchers.
-
-template <typename Iter>
-inline internal::UnorderedElementsAreArrayMatcher<
-    typename ::std::iterator_traits<Iter>::value_type>
-UnorderedElementsAreArray(Iter first, Iter last) {
-  typedef typename ::std::iterator_traits<Iter>::value_type T;
-  return internal::UnorderedElementsAreArrayMatcher<T>(
-      internal::UnorderedMatcherRequire::ExactMatch, first, last);
-}
-
-template <typename T>
-inline internal::UnorderedElementsAreArrayMatcher<T>
-UnorderedElementsAreArray(const T* pointer, size_t count) {
-  return UnorderedElementsAreArray(pointer, pointer + count);
-}
-
-template <typename T, size_t N>
-inline internal::UnorderedElementsAreArrayMatcher<T>
-UnorderedElementsAreArray(const T (&array)[N]) {
-  return UnorderedElementsAreArray(array, N);
-}
-
-template <typename Container>
-inline internal::UnorderedElementsAreArrayMatcher<
-    typename Container::value_type>
-UnorderedElementsAreArray(const Container& container) {
-  return UnorderedElementsAreArray(container.begin(), container.end());
-}
-
-#if GTEST_HAS_STD_INITIALIZER_LIST_
-template <typename T>
-inline internal::UnorderedElementsAreArrayMatcher<T>
-UnorderedElementsAreArray(::std::initializer_list<T> xs) {
-  return UnorderedElementsAreArray(xs.begin(), xs.end());
-}
-#endif
-
-// _ is a matcher that matches anything of any type.
-//
-// This definition is fine as:
-//
-//   1. The C++ standard permits using the name _ in a namespace that
-//      is not the global namespace or ::std.
-//   2. The AnythingMatcher class has no data member or constructor,
-//      so it's OK to create global variables of this type.
-//   3. c-style has approved of using _ in this case.
-const internal::AnythingMatcher _ = {};
-// Creates a matcher that matches any value of the given type T.
-template <typename T>
-inline Matcher<T> A() {
-  return Matcher<T>(new internal::AnyMatcherImpl<T>());
-}
-
-// Creates a matcher that matches any value of the given type T.
-template <typename T>
-inline Matcher<T> An() { return A<T>(); }
-
-// Creates a polymorphic matcher that matches anything equal to x.
-// Note: if the parameter of Eq() were declared as const T&, Eq("foo")
-// wouldn't compile.
-template <typename T>
-inline internal::EqMatcher<T> Eq(T x) { return internal::EqMatcher<T>(x); }
-
-// Constructs a Matcher<T> from a 'value' of type T.  The constructed
-// matcher matches any value that's equal to 'value'.
-template <typename T>
-Matcher<T>::Matcher(T value) { *this = Eq(value); }
-
-template <typename T, typename M>
-Matcher<T> internal::MatcherCastImpl<T, M>::CastImpl(
-    const M& value,
-    internal::BooleanConstant<false> /* convertible_to_matcher */,
-    internal::BooleanConstant<false> /* convertible_to_T */) {
-  return Eq(value);
-}
-
-// Creates a monomorphic matcher that matches anything with type Lhs
-// and equal to rhs.  A user may need to use this instead of Eq(...)
-// in order to resolve an overloading ambiguity.
-//
-// TypedEq<T>(x) is just a convenient short-hand for Matcher<T>(Eq(x))
-// or Matcher<T>(x), but more readable than the latter.
-//
-// We could define similar monomorphic matchers for other comparison
-// operations (e.g. TypedLt, TypedGe, and etc), but decided not to do
-// it yet as those are used much less than Eq() in practice.  A user
-// can always write Matcher<T>(Lt(5)) to be explicit about the type,
-// for example.
-template <typename Lhs, typename Rhs>
-inline Matcher<Lhs> TypedEq(const Rhs& rhs) { return Eq(rhs); }
-
-// Creates a polymorphic matcher that matches anything >= x.
-template <typename Rhs>
-inline internal::GeMatcher<Rhs> Ge(Rhs x) {
-  return internal::GeMatcher<Rhs>(x);
-}
-
-// Creates a polymorphic matcher that matches anything > x.
-template <typename Rhs>
-inline internal::GtMatcher<Rhs> Gt(Rhs x) {
-  return internal::GtMatcher<Rhs>(x);
-}
-
-// Creates a polymorphic matcher that matches anything <= x.
-template <typename Rhs>
-inline internal::LeMatcher<Rhs> Le(Rhs x) {
-  return internal::LeMatcher<Rhs>(x);
-}
-
-// Creates a polymorphic matcher that matches anything < x.
-template <typename Rhs>
-inline internal::LtMatcher<Rhs> Lt(Rhs x) {
-  return internal::LtMatcher<Rhs>(x);
-}
-
-// Creates a polymorphic matcher that matches anything != x.
-template <typename Rhs>
-inline internal::NeMatcher<Rhs> Ne(Rhs x) {
-  return internal::NeMatcher<Rhs>(x);
-}
-
-// Creates a polymorphic matcher that matches any NULL pointer.
-inline PolymorphicMatcher<internal::IsNullMatcher > IsNull() {
-  return MakePolymorphicMatcher(internal::IsNullMatcher());
-}
-
-// Creates a polymorphic matcher that matches any non-NULL pointer.
-// This is convenient as Not(NULL) doesn't compile (the compiler
-// thinks that that expression is comparing a pointer with an integer).
-inline PolymorphicMatcher<internal::NotNullMatcher > NotNull() {
-  return MakePolymorphicMatcher(internal::NotNullMatcher());
-}
-
-// Creates a polymorphic matcher that matches any argument that
-// references variable x.
-template <typename T>
-inline internal::RefMatcher<T&> Ref(T& x) {  // NOLINT
-  return internal::RefMatcher<T&>(x);
-}
-
-// Creates a matcher that matches any double argument approximately
-// equal to rhs, where two NANs are considered unequal.
-inline internal::FloatingEqMatcher<double> DoubleEq(double rhs) {
-  return internal::FloatingEqMatcher<double>(rhs, false);
-}
-
-// Creates a matcher that matches any double argument approximately
-// equal to rhs, including NaN values when rhs is NaN.
-inline internal::FloatingEqMatcher<double> NanSensitiveDoubleEq(double rhs) {
-  return internal::FloatingEqMatcher<double>(rhs, true);
-}
-
-// Creates a matcher that matches any double argument approximately equal to
-// rhs, up to the specified max absolute error bound, where two NANs are
-// considered unequal.  The max absolute error bound must be non-negative.
-inline internal::FloatingEqMatcher<double> DoubleNear(
-    double rhs, double max_abs_error) {
-  return internal::FloatingEqMatcher<double>(rhs, false, max_abs_error);
-}
-
-// Creates a matcher that matches any double argument approximately equal to
-// rhs, up to the specified max absolute error bound, including NaN values when
-// rhs is NaN.  The max absolute error bound must be non-negative.
-inline internal::FloatingEqMatcher<double> NanSensitiveDoubleNear(
-    double rhs, double max_abs_error) {
-  return internal::FloatingEqMatcher<double>(rhs, true, max_abs_error);
-}
-
-// Creates a matcher that matches any float argument approximately
-// equal to rhs, where two NANs are considered unequal.
-inline internal::FloatingEqMatcher<float> FloatEq(float rhs) {
-  return internal::FloatingEqMatcher<float>(rhs, false);
-}
-
-// Creates a matcher that matches any float argument approximately
-// equal to rhs, including NaN values when rhs is NaN.
-inline internal::FloatingEqMatcher<float> NanSensitiveFloatEq(float rhs) {
-  return internal::FloatingEqMatcher<float>(rhs, true);
-}
-
-// Creates a matcher that matches any float argument approximately equal to
-// rhs, up to the specified max absolute error bound, where two NANs are
-// considered unequal.  The max absolute error bound must be non-negative.
-inline internal::FloatingEqMatcher<float> FloatNear(
-    float rhs, float max_abs_error) {
-  return internal::FloatingEqMatcher<float>(rhs, false, max_abs_error);
-}
-
-// Creates a matcher that matches any float argument approximately equal to
-// rhs, up to the specified max absolute error bound, including NaN values when
-// rhs is NaN.  The max absolute error bound must be non-negative.
-inline internal::FloatingEqMatcher<float> NanSensitiveFloatNear(
-    float rhs, float max_abs_error) {
-  return internal::FloatingEqMatcher<float>(rhs, true, max_abs_error);
-}
-
-// Creates a matcher that matches a pointer (raw or smart) that points
-// to a value that matches inner_matcher.
-template <typename InnerMatcher>
-inline internal::PointeeMatcher<InnerMatcher> Pointee(
-    const InnerMatcher& inner_matcher) {
-  return internal::PointeeMatcher<InnerMatcher>(inner_matcher);
-}
-
-#if GTEST_HAS_RTTI
-// Creates a matcher that matches a pointer or reference that matches
-// inner_matcher when dynamic_cast<To> is applied.
-// The result of dynamic_cast<To> is forwarded to the inner matcher.
-// If To is a pointer and the cast fails, the inner matcher will receive NULL.
-// If To is a reference and the cast fails, this matcher returns false
-// immediately.
-template <typename To>
-inline PolymorphicMatcher<internal::WhenDynamicCastToMatcher<To> >
-WhenDynamicCastTo(const Matcher<To>& inner_matcher) {
-  return MakePolymorphicMatcher(
-      internal::WhenDynamicCastToMatcher<To>(inner_matcher));
-}
-#endif  // GTEST_HAS_RTTI
-
-// Creates a matcher that matches an object whose given field matches
-// 'matcher'.  For example,
-//   Field(&Foo::number, Ge(5))
-// matches a Foo object x iff x.number >= 5.
-template <typename Class, typename FieldType, typename FieldMatcher>
-inline PolymorphicMatcher<
-  internal::FieldMatcher<Class, FieldType> > Field(
-    FieldType Class::*field, const FieldMatcher& matcher) {
-  return MakePolymorphicMatcher(
-      internal::FieldMatcher<Class, FieldType>(
-          field, MatcherCast<const FieldType&>(matcher)));
-  // The call to MatcherCast() is required for supporting inner
-  // matchers of compatible types.  For example, it allows
-  //   Field(&Foo::bar, m)
-  // to compile where bar is an int32 and m is a matcher for int64.
-}
-
-// Same as Field() but also takes the name of the field to provide better error
-// messages.
-template <typename Class, typename FieldType, typename FieldMatcher>
-inline PolymorphicMatcher<internal::FieldMatcher<Class, FieldType> > Field(
-    const std::string& field_name, FieldType Class::*field,
-    const FieldMatcher& matcher) {
-  return MakePolymorphicMatcher(internal::FieldMatcher<Class, FieldType>(
-      field_name, field, MatcherCast<const FieldType&>(matcher)));
-}
-
-// Creates a matcher that matches an object whose given property
-// matches 'matcher'.  For example,
-//   Property(&Foo::str, StartsWith("hi"))
-// matches a Foo object x iff x.str() starts with "hi".
-template <typename Class, typename PropertyType, typename PropertyMatcher>
-inline PolymorphicMatcher<internal::PropertyMatcher<
-    Class, PropertyType, PropertyType (Class::*)() const> >
-Property(PropertyType (Class::*property)() const,
-         const PropertyMatcher& matcher) {
-  return MakePolymorphicMatcher(
-      internal::PropertyMatcher<Class, PropertyType,
-                                PropertyType (Class::*)() const>(
-          property,
-          MatcherCast<GTEST_REFERENCE_TO_CONST_(PropertyType)>(matcher)));
-  // The call to MatcherCast() is required for supporting inner
-  // matchers of compatible types.  For example, it allows
-  //   Property(&Foo::bar, m)
-  // to compile where bar() returns an int32 and m is a matcher for int64.
-}
-
-// Same as Property() above, but also takes the name of the property to provide
-// better error messages.
-template <typename Class, typename PropertyType, typename PropertyMatcher>
-inline PolymorphicMatcher<internal::PropertyMatcher<
-    Class, PropertyType, PropertyType (Class::*)() const> >
-Property(const std::string& property_name,
-         PropertyType (Class::*property)() const,
-         const PropertyMatcher& matcher) {
-  return MakePolymorphicMatcher(
-      internal::PropertyMatcher<Class, PropertyType,
-                                PropertyType (Class::*)() const>(
-          property_name, property,
-          MatcherCast<GTEST_REFERENCE_TO_CONST_(PropertyType)>(matcher)));
-}
-
-#if GTEST_LANG_CXX11
-// The same as above but for reference-qualified member functions.
-template <typename Class, typename PropertyType, typename PropertyMatcher>
-inline PolymorphicMatcher<internal::PropertyMatcher<
-    Class, PropertyType, PropertyType (Class::*)() const &> >
-Property(PropertyType (Class::*property)() const &,
-         const PropertyMatcher& matcher) {
-  return MakePolymorphicMatcher(
-      internal::PropertyMatcher<Class, PropertyType,
-                                PropertyType (Class::*)() const &>(
-          property,
-          MatcherCast<GTEST_REFERENCE_TO_CONST_(PropertyType)>(matcher)));
-}
-
-// Three-argument form for reference-qualified member functions.
-template <typename Class, typename PropertyType, typename PropertyMatcher>
-inline PolymorphicMatcher<internal::PropertyMatcher<
-    Class, PropertyType, PropertyType (Class::*)() const &> >
-Property(const std::string& property_name,
-         PropertyType (Class::*property)() const &,
-         const PropertyMatcher& matcher) {
-  return MakePolymorphicMatcher(
-      internal::PropertyMatcher<Class, PropertyType,
-                                PropertyType (Class::*)() const &>(
-          property_name, property,
-          MatcherCast<GTEST_REFERENCE_TO_CONST_(PropertyType)>(matcher)));
-}
-#endif
-
-// Creates a matcher that matches an object iff the result of applying
-// a callable to x matches 'matcher'.
-// For example,
-//   ResultOf(f, StartsWith("hi"))
-// matches a Foo object x iff f(x) starts with "hi".
-// callable parameter can be a function, function pointer, or a functor.
-// Callable has to satisfy the following conditions:
-//   * It is required to keep no state affecting the results of
-//     the calls on it and make no assumptions about how many calls
-//     will be made. Any state it keeps must be protected from the
-//     concurrent access.
-//   * If it is a function object, it has to define type result_type.
-//     We recommend deriving your functor classes from std::unary_function.
-//
-template <typename Callable, typename ResultOfMatcher>
-internal::ResultOfMatcher<Callable> ResultOf(
-    Callable callable, const ResultOfMatcher& matcher) {
-  return internal::ResultOfMatcher<Callable>(
-          callable,
-          MatcherCast<typename internal::CallableTraits<Callable>::ResultType>(
-              matcher));
-  // The call to MatcherCast() is required for supporting inner
-  // matchers of compatible types.  For example, it allows
-  //   ResultOf(Function, m)
-  // to compile where Function() returns an int32 and m is a matcher for int64.
-}
-
-// String matchers.
-
-// Matches a string equal to str.
-inline PolymorphicMatcher<internal::StrEqualityMatcher<std::string> > StrEq(
-    const std::string& str) {
-  return MakePolymorphicMatcher(
-      internal::StrEqualityMatcher<std::string>(str, true, true));
-}
-
-// Matches a string not equal to str.
-inline PolymorphicMatcher<internal::StrEqualityMatcher<std::string> > StrNe(
-    const std::string& str) {
-  return MakePolymorphicMatcher(
-      internal::StrEqualityMatcher<std::string>(str, false, true));
-}
-
-// Matches a string equal to str, ignoring case.
-inline PolymorphicMatcher<internal::StrEqualityMatcher<std::string> > StrCaseEq(
-    const std::string& str) {
-  return MakePolymorphicMatcher(
-      internal::StrEqualityMatcher<std::string>(str, true, false));
-}
-
-// Matches a string not equal to str, ignoring case.
-inline PolymorphicMatcher<internal::StrEqualityMatcher<std::string> > StrCaseNe(
-    const std::string& str) {
-  return MakePolymorphicMatcher(
-      internal::StrEqualityMatcher<std::string>(str, false, false));
-}
-
-// Creates a matcher that matches any string, std::string, or C string
-// that contains the given substring.
-inline PolymorphicMatcher<internal::HasSubstrMatcher<std::string> > HasSubstr(
-    const std::string& substring) {
-  return MakePolymorphicMatcher(
-      internal::HasSubstrMatcher<std::string>(substring));
-}
-
-// Matches a string that starts with 'prefix' (case-sensitive).
-inline PolymorphicMatcher<internal::StartsWithMatcher<std::string> > StartsWith(
-    const std::string& prefix) {
-  return MakePolymorphicMatcher(
-      internal::StartsWithMatcher<std::string>(prefix));
-}
-
-// Matches a string that ends with 'suffix' (case-sensitive).
-inline PolymorphicMatcher<internal::EndsWithMatcher<std::string> > EndsWith(
-    const std::string& suffix) {
-  return MakePolymorphicMatcher(internal::EndsWithMatcher<std::string>(suffix));
-}
-
-// Matches a string that fully matches regular expression 'regex'.
-// The matcher takes ownership of 'regex'.
-inline PolymorphicMatcher<internal::MatchesRegexMatcher> MatchesRegex(
-    const internal::RE* regex) {
-  return MakePolymorphicMatcher(internal::MatchesRegexMatcher(regex, true));
-}
-inline PolymorphicMatcher<internal::MatchesRegexMatcher> MatchesRegex(
-    const std::string& regex) {
-  return MatchesRegex(new internal::RE(regex));
-}
-
-// Matches a string that contains regular expression 'regex'.
-// The matcher takes ownership of 'regex'.
-inline PolymorphicMatcher<internal::MatchesRegexMatcher> ContainsRegex(
-    const internal::RE* regex) {
-  return MakePolymorphicMatcher(internal::MatchesRegexMatcher(regex, false));
-}
-inline PolymorphicMatcher<internal::MatchesRegexMatcher> ContainsRegex(
-    const std::string& regex) {
-  return ContainsRegex(new internal::RE(regex));
-}
-
-#if GTEST_HAS_GLOBAL_WSTRING || GTEST_HAS_STD_WSTRING
-// Wide string matchers.
-
-// Matches a string equal to str.
-inline PolymorphicMatcher<internal::StrEqualityMatcher<std::wstring> > StrEq(
-    const std::wstring& str) {
-  return MakePolymorphicMatcher(
-      internal::StrEqualityMatcher<std::wstring>(str, true, true));
-}
-
-// Matches a string not equal to str.
-inline PolymorphicMatcher<internal::StrEqualityMatcher<std::wstring> > StrNe(
-    const std::wstring& str) {
-  return MakePolymorphicMatcher(
-      internal::StrEqualityMatcher<std::wstring>(str, false, true));
-}
-
-// Matches a string equal to str, ignoring case.
-inline PolymorphicMatcher<internal::StrEqualityMatcher<std::wstring> >
-StrCaseEq(const std::wstring& str) {
-  return MakePolymorphicMatcher(
-      internal::StrEqualityMatcher<std::wstring>(str, true, false));
-}
-
-// Matches a string not equal to str, ignoring case.
-inline PolymorphicMatcher<internal::StrEqualityMatcher<std::wstring> >
-StrCaseNe(const std::wstring& str) {
-  return MakePolymorphicMatcher(
-      internal::StrEqualityMatcher<std::wstring>(str, false, false));
-}
-
-// Creates a matcher that matches any ::wstring, std::wstring, or C wide string
-// that contains the given substring.
-inline PolymorphicMatcher<internal::HasSubstrMatcher<std::wstring> > HasSubstr(
-    const std::wstring& substring) {
-  return MakePolymorphicMatcher(
-      internal::HasSubstrMatcher<std::wstring>(substring));
-}
-
-// Matches a string that starts with 'prefix' (case-sensitive).
-inline PolymorphicMatcher<internal::StartsWithMatcher<std::wstring> >
-StartsWith(const std::wstring& prefix) {
-  return MakePolymorphicMatcher(
-      internal::StartsWithMatcher<std::wstring>(prefix));
-}
-
-// Matches a string that ends with 'suffix' (case-sensitive).
-inline PolymorphicMatcher<internal::EndsWithMatcher<std::wstring> > EndsWith(
-    const std::wstring& suffix) {
-  return MakePolymorphicMatcher(
-      internal::EndsWithMatcher<std::wstring>(suffix));
-}
-
-#endif  // GTEST_HAS_GLOBAL_WSTRING || GTEST_HAS_STD_WSTRING
-
-// Creates a polymorphic matcher that matches a 2-tuple where the
-// first field == the second field.
-inline internal::Eq2Matcher Eq() { return internal::Eq2Matcher(); }
-
-// Creates a polymorphic matcher that matches a 2-tuple where the
-// first field >= the second field.
-inline internal::Ge2Matcher Ge() { return internal::Ge2Matcher(); }
-
-// Creates a polymorphic matcher that matches a 2-tuple where the
-// first field > the second field.
-inline internal::Gt2Matcher Gt() { return internal::Gt2Matcher(); }
-
-// Creates a polymorphic matcher that matches a 2-tuple where the
-// first field <= the second field.
-inline internal::Le2Matcher Le() { return internal::Le2Matcher(); }
-
-// Creates a polymorphic matcher that matches a 2-tuple where the
-// first field < the second field.
-inline internal::Lt2Matcher Lt() { return internal::Lt2Matcher(); }
-
-// Creates a polymorphic matcher that matches a 2-tuple where the
-// first field != the second field.
-inline internal::Ne2Matcher Ne() { return internal::Ne2Matcher(); }
-
-// Creates a polymorphic matcher that matches a 2-tuple where
-// FloatEq(first field) matches the second field.
-inline internal::FloatingEq2Matcher<float> FloatEq() {
-  return internal::FloatingEq2Matcher<float>();
-}
-
-// Creates a polymorphic matcher that matches a 2-tuple where
-// DoubleEq(first field) matches the second field.
-inline internal::FloatingEq2Matcher<double> DoubleEq() {
-  return internal::FloatingEq2Matcher<double>();
-}
-
-// Creates a polymorphic matcher that matches a 2-tuple where
-// FloatEq(first field) matches the second field with NaN equality.
-inline internal::FloatingEq2Matcher<float> NanSensitiveFloatEq() {
-  return internal::FloatingEq2Matcher<float>(true);
-}
-
-// Creates a polymorphic matcher that matches a 2-tuple where
-// DoubleEq(first field) matches the second field with NaN equality.
-inline internal::FloatingEq2Matcher<double> NanSensitiveDoubleEq() {
-  return internal::FloatingEq2Matcher<double>(true);
-}
-
-// Creates a polymorphic matcher that matches a 2-tuple where
-// FloatNear(first field, max_abs_error) matches the second field.
-inline internal::FloatingEq2Matcher<float> FloatNear(float max_abs_error) {
-  return internal::FloatingEq2Matcher<float>(max_abs_error);
-}
-
-// Creates a polymorphic matcher that matches a 2-tuple where
-// DoubleNear(first field, max_abs_error) matches the second field.
-inline internal::FloatingEq2Matcher<double> DoubleNear(double max_abs_error) {
-  return internal::FloatingEq2Matcher<double>(max_abs_error);
-}
-
-// Creates a polymorphic matcher that matches a 2-tuple where
-// FloatNear(first field, max_abs_error) matches the second field with NaN
-// equality.
-inline internal::FloatingEq2Matcher<float> NanSensitiveFloatNear(
-    float max_abs_error) {
-  return internal::FloatingEq2Matcher<float>(max_abs_error, true);
-}
-
-// Creates a polymorphic matcher that matches a 2-tuple where
-// DoubleNear(first field, max_abs_error) matches the second field with NaN
-// equality.
-inline internal::FloatingEq2Matcher<double> NanSensitiveDoubleNear(
-    double max_abs_error) {
-  return internal::FloatingEq2Matcher<double>(max_abs_error, true);
-}
-
-// Creates a matcher that matches any value of type T that m doesn't
-// match.
-template <typename InnerMatcher>
-inline internal::NotMatcher<InnerMatcher> Not(InnerMatcher m) {
-  return internal::NotMatcher<InnerMatcher>(m);
-}
-
-// Returns a matcher that matches anything that satisfies the given
-// predicate.  The predicate can be any unary function or functor
-// whose return type can be implicitly converted to bool.
-template <typename Predicate>
-inline PolymorphicMatcher<internal::TrulyMatcher<Predicate> >
-Truly(Predicate pred) {
-  return MakePolymorphicMatcher(internal::TrulyMatcher<Predicate>(pred));
-}
-
-// Returns a matcher that matches the container size. The container must
-// support both size() and size_type which all STL-like containers provide.
-// Note that the parameter 'size' can be a value of type size_type as well as
-// matcher. For instance:
-//   EXPECT_THAT(container, SizeIs(2));     // Checks container has 2 elements.
-//   EXPECT_THAT(container, SizeIs(Le(2));  // Checks container has at most 2.
-template <typename SizeMatcher>
-inline internal::SizeIsMatcher<SizeMatcher>
-SizeIs(const SizeMatcher& size_matcher) {
-  return internal::SizeIsMatcher<SizeMatcher>(size_matcher);
-}
-
-// Returns a matcher that matches the distance between the container's begin()
-// iterator and its end() iterator, i.e. the size of the container. This matcher
-// can be used instead of SizeIs with containers such as std::forward_list which
-// do not implement size(). The container must provide const_iterator (with
-// valid iterator_traits), begin() and end().
-template <typename DistanceMatcher>
-inline internal::BeginEndDistanceIsMatcher<DistanceMatcher>
-BeginEndDistanceIs(const DistanceMatcher& distance_matcher) {
-  return internal::BeginEndDistanceIsMatcher<DistanceMatcher>(distance_matcher);
-}
-
-// Returns a matcher that matches an equal container.
-// This matcher behaves like Eq(), but in the event of mismatch lists the
-// values that are included in one container but not the other. (Duplicate
-// values and order differences are not explained.)
-template <typename Container>
-inline PolymorphicMatcher<internal::ContainerEqMatcher<  // NOLINT
-                            GTEST_REMOVE_CONST_(Container)> >
-    ContainerEq(const Container& rhs) {
-  // This following line is for working around a bug in MSVC 8.0,
-  // which causes Container to be a const type sometimes.
-  typedef GTEST_REMOVE_CONST_(Container) RawContainer;
-  return MakePolymorphicMatcher(
-      internal::ContainerEqMatcher<RawContainer>(rhs));
-}
-
-// Returns a matcher that matches a container that, when sorted using
-// the given comparator, matches container_matcher.
-template <typename Comparator, typename ContainerMatcher>
-inline internal::WhenSortedByMatcher<Comparator, ContainerMatcher>
-WhenSortedBy(const Comparator& comparator,
-             const ContainerMatcher& container_matcher) {
-  return internal::WhenSortedByMatcher<Comparator, ContainerMatcher>(
-      comparator, container_matcher);
-}
-
-// Returns a matcher that matches a container that, when sorted using
-// the < operator, matches container_matcher.
-template <typename ContainerMatcher>
-inline internal::WhenSortedByMatcher<internal::LessComparator, ContainerMatcher>
-WhenSorted(const ContainerMatcher& container_matcher) {
-  return
-      internal::WhenSortedByMatcher<internal::LessComparator, ContainerMatcher>(
-          internal::LessComparator(), container_matcher);
-}
-
-// Matches an STL-style container or a native array that contains the
-// same number of elements as in rhs, where its i-th element and rhs's
-// i-th element (as a pair) satisfy the given pair matcher, for all i.
-// TupleMatcher must be able to be safely cast to Matcher<tuple<const
-// T1&, const T2&> >, where T1 and T2 are the types of elements in the
-// LHS container and the RHS container respectively.
-template <typename TupleMatcher, typename Container>
-inline internal::PointwiseMatcher<TupleMatcher,
-                                  GTEST_REMOVE_CONST_(Container)>
-Pointwise(const TupleMatcher& tuple_matcher, const Container& rhs) {
-  // This following line is for working around a bug in MSVC 8.0,
-  // which causes Container to be a const type sometimes (e.g. when
-  // rhs is a const int[])..
-  typedef GTEST_REMOVE_CONST_(Container) RawContainer;
-  return internal::PointwiseMatcher<TupleMatcher, RawContainer>(
-      tuple_matcher, rhs);
-}
-
-#if GTEST_HAS_STD_INITIALIZER_LIST_
-
-// Supports the Pointwise(m, {a, b, c}) syntax.
-template <typename TupleMatcher, typename T>
-inline internal::PointwiseMatcher<TupleMatcher, std::vector<T> > Pointwise(
-    const TupleMatcher& tuple_matcher, std::initializer_list<T> rhs) {
-  return Pointwise(tuple_matcher, std::vector<T>(rhs));
-}
-
-#endif  // GTEST_HAS_STD_INITIALIZER_LIST_
-
-// UnorderedPointwise(pair_matcher, rhs) matches an STL-style
-// container or a native array that contains the same number of
-// elements as in rhs, where in some permutation of the container, its
-// i-th element and rhs's i-th element (as a pair) satisfy the given
-// pair matcher, for all i.  Tuple2Matcher must be able to be safely
-// cast to Matcher<tuple<const T1&, const T2&> >, where T1 and T2 are
-// the types of elements in the LHS container and the RHS container
-// respectively.
-//
-// This is like Pointwise(pair_matcher, rhs), except that the element
-// order doesn't matter.
-template <typename Tuple2Matcher, typename RhsContainer>
-inline internal::UnorderedElementsAreArrayMatcher<
-    typename internal::BoundSecondMatcher<
-        Tuple2Matcher, typename internal::StlContainerView<GTEST_REMOVE_CONST_(
-                           RhsContainer)>::type::value_type> >
-UnorderedPointwise(const Tuple2Matcher& tuple2_matcher,
-                   const RhsContainer& rhs_container) {
-  // This following line is for working around a bug in MSVC 8.0,
-  // which causes RhsContainer to be a const type sometimes (e.g. when
-  // rhs_container is a const int[]).
-  typedef GTEST_REMOVE_CONST_(RhsContainer) RawRhsContainer;
-
-  // RhsView allows the same code to handle RhsContainer being a
-  // STL-style container and it being a native C-style array.
-  typedef typename internal::StlContainerView<RawRhsContainer> RhsView;
-  typedef typename RhsView::type RhsStlContainer;
-  typedef typename RhsStlContainer::value_type Second;
-  const RhsStlContainer& rhs_stl_container =
-      RhsView::ConstReference(rhs_container);
-
-  // Create a matcher for each element in rhs_container.
-  ::std::vector<internal::BoundSecondMatcher<Tuple2Matcher, Second> > matchers;
-  for (typename RhsStlContainer::const_iterator it = rhs_stl_container.begin();
-       it != rhs_stl_container.end(); ++it) {
-    matchers.push_back(
-        internal::MatcherBindSecond(tuple2_matcher, *it));
-  }
-
-  // Delegate the work to UnorderedElementsAreArray().
-  return UnorderedElementsAreArray(matchers);
-}
-
-#if GTEST_HAS_STD_INITIALIZER_LIST_
-
-// Supports the UnorderedPointwise(m, {a, b, c}) syntax.
-template <typename Tuple2Matcher, typename T>
-inline internal::UnorderedElementsAreArrayMatcher<
-    typename internal::BoundSecondMatcher<Tuple2Matcher, T> >
-UnorderedPointwise(const Tuple2Matcher& tuple2_matcher,
-                   std::initializer_list<T> rhs) {
-  return UnorderedPointwise(tuple2_matcher, std::vector<T>(rhs));
-}
-
-#endif  // GTEST_HAS_STD_INITIALIZER_LIST_
-
-// Matches an STL-style container or a native array that contains at
-// least one element matching the given value or matcher.
-//
-// Examples:
-//   ::std::set<int> page_ids;
-//   page_ids.insert(3);
-//   page_ids.insert(1);
-//   EXPECT_THAT(page_ids, Contains(1));
-//   EXPECT_THAT(page_ids, Contains(Gt(2)));
-//   EXPECT_THAT(page_ids, Not(Contains(4)));
-//
-//   ::std::map<int, size_t> page_lengths;
-//   page_lengths[1] = 100;
-//   EXPECT_THAT(page_lengths,
-//               Contains(::std::pair<const int, size_t>(1, 100)));
-//
-//   const char* user_ids[] = { "joe", "mike", "tom" };
-//   EXPECT_THAT(user_ids, Contains(Eq(::std::string("tom"))));
-template <typename M>
-inline internal::ContainsMatcher<M> Contains(M matcher) {
-  return internal::ContainsMatcher<M>(matcher);
-}
-
-// IsSupersetOf(iterator_first, iterator_last)
-// IsSupersetOf(pointer, count)
-// IsSupersetOf(array)
-// IsSupersetOf(container)
-// IsSupersetOf({e1, e2, ..., en})
-//
-// IsSupersetOf() verifies that a surjective partial mapping onto a collection
-// of matchers exists. In other words, a container matches
-// IsSupersetOf({e1, ..., en}) if and only if there is a permutation
-// {y1, ..., yn} of some of the container's elements where y1 matches e1,
-// ..., and yn matches en. Obviously, the size of the container must be >= n
-// in order to have a match. Examples:
-//
-// - {1, 2, 3} matches IsSupersetOf({Ge(3), Ne(0)}), as 3 matches Ge(3) and
-//   1 matches Ne(0).
-// - {1, 2} doesn't match IsSupersetOf({Eq(1), Lt(2)}), even though 1 matches
-//   both Eq(1) and Lt(2). The reason is that different matchers must be used
-//   for elements in different slots of the container.
-// - {1, 1, 2} matches IsSupersetOf({Eq(1), Lt(2)}), as (the first) 1 matches
-//   Eq(1) and (the second) 1 matches Lt(2).
-// - {1, 2, 3} matches IsSupersetOf(Gt(1), Gt(1)), as 2 matches (the first)
-//   Gt(1) and 3 matches (the second) Gt(1).
-//
-// The matchers can be specified as an array, a pointer and count, a container,
-// an initializer list, or an STL iterator range. In each of these cases, the
-// underlying matchers can be either values or matchers.
-
-template <typename Iter>
-inline internal::UnorderedElementsAreArrayMatcher<
-    typename ::std::iterator_traits<Iter>::value_type>
-IsSupersetOf(Iter first, Iter last) {
-  typedef typename ::std::iterator_traits<Iter>::value_type T;
-  return internal::UnorderedElementsAreArrayMatcher<T>(
-      internal::UnorderedMatcherRequire::Superset, first, last);
-}
-
-template <typename T>
-inline internal::UnorderedElementsAreArrayMatcher<T> IsSupersetOf(
-    const T* pointer, size_t count) {
-  return IsSupersetOf(pointer, pointer + count);
-}
-
-template <typename T, size_t N>
-inline internal::UnorderedElementsAreArrayMatcher<T> IsSupersetOf(
-    const T (&array)[N]) {
-  return IsSupersetOf(array, N);
-}
-
-template <typename Container>
-inline internal::UnorderedElementsAreArrayMatcher<
-    typename Container::value_type>
-IsSupersetOf(const Container& container) {
-  return IsSupersetOf(container.begin(), container.end());
-}
-
-#if GTEST_HAS_STD_INITIALIZER_LIST_
-template <typename T>
-inline internal::UnorderedElementsAreArrayMatcher<T> IsSupersetOf(
-    ::std::initializer_list<T> xs) {
-  return IsSupersetOf(xs.begin(), xs.end());
-}
-#endif
-
-// IsSubsetOf(iterator_first, iterator_last)
-// IsSubsetOf(pointer, count)
-// IsSubsetOf(array)
-// IsSubsetOf(container)
-// IsSubsetOf({e1, e2, ..., en})
-//
-// IsSubsetOf() verifies that an injective mapping onto a collection of matchers
-// exists.  In other words, a container matches IsSubsetOf({e1, ..., en}) if and
-// only if there is a subset of matchers {m1, ..., mk} which would match the
-// container using UnorderedElementsAre.  Obviously, the size of the container
-// must be <= n in order to have a match. Examples:
-//
-// - {1} matches IsSubsetOf({Gt(0), Lt(0)}), as 1 matches Gt(0).
-// - {1, -1} matches IsSubsetOf({Lt(0), Gt(0)}), as 1 matches Gt(0) and -1
-//   matches Lt(0).
-// - {1, 2} doesn't matches IsSubsetOf({Gt(0), Lt(0)}), even though 1 and 2 both
-//   match Gt(0). The reason is that different matchers must be used for
-//   elements in different slots of the container.
-//
-// The matchers can be specified as an array, a pointer and count, a container,
-// an initializer list, or an STL iterator range. In each of these cases, the
-// underlying matchers can be either values or matchers.
-
-template <typename Iter>
-inline internal::UnorderedElementsAreArrayMatcher<
-    typename ::std::iterator_traits<Iter>::value_type>
-IsSubsetOf(Iter first, Iter last) {
-  typedef typename ::std::iterator_traits<Iter>::value_type T;
-  return internal::UnorderedElementsAreArrayMatcher<T>(
-      internal::UnorderedMatcherRequire::Subset, first, last);
-}
-
-template <typename T>
-inline internal::UnorderedElementsAreArrayMatcher<T> IsSubsetOf(
-    const T* pointer, size_t count) {
-  return IsSubsetOf(pointer, pointer + count);
-}
-
-template <typename T, size_t N>
-inline internal::UnorderedElementsAreArrayMatcher<T> IsSubsetOf(
-    const T (&array)[N]) {
-  return IsSubsetOf(array, N);
-}
-
-template <typename Container>
-inline internal::UnorderedElementsAreArrayMatcher<
-    typename Container::value_type>
-IsSubsetOf(const Container& container) {
-  return IsSubsetOf(container.begin(), container.end());
-}
-
-#if GTEST_HAS_STD_INITIALIZER_LIST_
-template <typename T>
-inline internal::UnorderedElementsAreArrayMatcher<T> IsSubsetOf(
-    ::std::initializer_list<T> xs) {
-  return IsSubsetOf(xs.begin(), xs.end());
-}
-#endif
-
-// Matches an STL-style container or a native array that contains only
-// elements matching the given value or matcher.
-//
-// Each(m) is semantically equivalent to Not(Contains(Not(m))). Only
-// the messages are different.
-//
-// Examples:
-//   ::std::set<int> page_ids;
-//   // Each(m) matches an empty container, regardless of what m is.
-//   EXPECT_THAT(page_ids, Each(Eq(1)));
-//   EXPECT_THAT(page_ids, Each(Eq(77)));
-//
-//   page_ids.insert(3);
-//   EXPECT_THAT(page_ids, Each(Gt(0)));
-//   EXPECT_THAT(page_ids, Not(Each(Gt(4))));
-//   page_ids.insert(1);
-//   EXPECT_THAT(page_ids, Not(Each(Lt(2))));
-//
-//   ::std::map<int, size_t> page_lengths;
-//   page_lengths[1] = 100;
-//   page_lengths[2] = 200;
-//   page_lengths[3] = 300;
-//   EXPECT_THAT(page_lengths, Not(Each(Pair(1, 100))));
-//   EXPECT_THAT(page_lengths, Each(Key(Le(3))));
-//
-//   const char* user_ids[] = { "joe", "mike", "tom" };
-//   EXPECT_THAT(user_ids, Not(Each(Eq(::std::string("tom")))));
-template <typename M>
-inline internal::EachMatcher<M> Each(M matcher) {
-  return internal::EachMatcher<M>(matcher);
-}
-
-// Key(inner_matcher) matches an std::pair whose 'first' field matches
-// inner_matcher.  For example, Contains(Key(Ge(5))) can be used to match an
-// std::map that contains at least one element whose key is >= 5.
-template <typename M>
-inline internal::KeyMatcher<M> Key(M inner_matcher) {
-  return internal::KeyMatcher<M>(inner_matcher);
-}
-
-// Pair(first_matcher, second_matcher) matches a std::pair whose 'first' field
-// matches first_matcher and whose 'second' field matches second_matcher.  For
-// example, EXPECT_THAT(map_type, ElementsAre(Pair(Ge(5), "foo"))) can be used
-// to match a std::map<int, string> that contains exactly one element whose key
-// is >= 5 and whose value equals "foo".
-template <typename FirstMatcher, typename SecondMatcher>
-inline internal::PairMatcher<FirstMatcher, SecondMatcher>
-Pair(FirstMatcher first_matcher, SecondMatcher second_matcher) {
-  return internal::PairMatcher<FirstMatcher, SecondMatcher>(
-      first_matcher, second_matcher);
-}
-
-// Returns a predicate that is satisfied by anything that matches the
-// given matcher.
-template <typename M>
-inline internal::MatcherAsPredicate<M> Matches(M matcher) {
-  return internal::MatcherAsPredicate<M>(matcher);
-}
-
-// Returns true iff the value matches the matcher.
-template <typename T, typename M>
-inline bool Value(const T& value, M matcher) {
-  return testing::Matches(matcher)(value);
-}
-
-// Matches the value against the given matcher and explains the match
-// result to listener.
-template <typename T, typename M>
-inline bool ExplainMatchResult(
-    M matcher, const T& value, MatchResultListener* listener) {
-  return SafeMatcherCast<const T&>(matcher).MatchAndExplain(value, listener);
-}
-
-// Returns a string representation of the given matcher.  Useful for description
-// strings of matchers defined using MATCHER_P* macros that accept matchers as
-// their arguments.  For example:
-//
-// MATCHER_P(XAndYThat, matcher,
-//           "X that " + DescribeMatcher<int>(matcher, negation) +
-//               " and Y that " + DescribeMatcher<double>(matcher, negation)) {
-//   return ExplainMatchResult(matcher, arg.x(), result_listener) &&
-//          ExplainMatchResult(matcher, arg.y(), result_listener);
-// }
-template <typename T, typename M>
-std::string DescribeMatcher(const M& matcher, bool negation = false) {
-  ::std::stringstream ss;
-  Matcher<T> monomorphic_matcher = SafeMatcherCast<T>(matcher);
-  if (negation) {
-    monomorphic_matcher.DescribeNegationTo(&ss);
-  } else {
-    monomorphic_matcher.DescribeTo(&ss);
-  }
-  return ss.str();
-}
-
-#if GTEST_LANG_CXX11
-// Define variadic matcher versions. They are overloaded in
-// gmock-generated-matchers.h for the cases supported by pre C++11 compilers.
-template <typename... Args>
-internal::AllOfMatcher<typename std::decay<const Args&>::type...> AllOf(
-    const Args&... matchers) {
-  return internal::AllOfMatcher<typename std::decay<const Args&>::type...>(
-      matchers...);
-}
-
-template <typename... Args>
-internal::AnyOfMatcher<typename std::decay<const Args&>::type...> AnyOf(
-    const Args&... matchers) {
-  return internal::AnyOfMatcher<typename std::decay<const Args&>::type...>(
-      matchers...);
-}
-
-template <typename... Args>
-internal::ElementsAreMatcher<tuple<typename std::decay<const Args&>::type...>>
-ElementsAre(const Args&... matchers) {
-  return internal::ElementsAreMatcher<
-      tuple<typename std::decay<const Args&>::type...>>(
-      make_tuple(matchers...));
-}
-
-template <typename... Args>
-internal::UnorderedElementsAreMatcher<
-    tuple<typename std::decay<const Args&>::type...>>
-UnorderedElementsAre(const Args&... matchers) {
-  return internal::UnorderedElementsAreMatcher<
-      tuple<typename std::decay<const Args&>::type...>>(
-      make_tuple(matchers...));
-}
-
-#endif  // GTEST_LANG_CXX11
-
-// AllArgs(m) is a synonym of m.  This is useful in
-//
-//   EXPECT_CALL(foo, Bar(_, _)).With(AllArgs(Eq()));
-//
-// which is easier to read than
-//
-//   EXPECT_CALL(foo, Bar(_, _)).With(Eq());
-template <typename InnerMatcher>
-inline InnerMatcher AllArgs(const InnerMatcher& matcher) { return matcher; }
-
-// Returns a matcher that matches the value of an optional<> type variable.
-// The matcher implementation only uses '!arg' and requires that the optional<>
-// type has a 'value_type' member type and that '*arg' is of type 'value_type'
-// and is printable using 'PrintToString'. It is compatible with
-// std::optional/std::experimental::optional.
-// Note that to compare an optional type variable against nullopt you should
-// use Eq(nullopt) and not Optional(Eq(nullopt)). The latter implies that the
-// optional value contains an optional itself.
-template <typename ValueMatcher>
-inline internal::OptionalMatcher<ValueMatcher> Optional(
-    const ValueMatcher& value_matcher) {
-  return internal::OptionalMatcher<ValueMatcher>(value_matcher);
-}
-
-// Returns a matcher that matches the value of a absl::any type variable.
-template <typename T>
-PolymorphicMatcher<internal::any_cast_matcher::AnyCastMatcher<T> > AnyWith(
-    const Matcher<const T&>& matcher) {
-  return MakePolymorphicMatcher(
-      internal::any_cast_matcher::AnyCastMatcher<T>(matcher));
-}
-
-// Returns a matcher that matches the value of a variant<> type variable.
-// The matcher implementation uses ADL to find the holds_alternative and get
-// functions.
-// It is compatible with std::variant.
-template <typename T>
-PolymorphicMatcher<internal::variant_matcher::VariantMatcher<T> > VariantWith(
-    const Matcher<const T&>& matcher) {
-  return MakePolymorphicMatcher(
-      internal::variant_matcher::VariantMatcher<T>(matcher));
-}
-
-// These macros allow using matchers to check values in Google Test
-// tests.  ASSERT_THAT(value, matcher) and EXPECT_THAT(value, matcher)
-// succeed iff the value matches the matcher.  If the assertion fails,
-// the value and the description of the matcher will be printed.
-#define ASSERT_THAT(value, matcher) ASSERT_PRED_FORMAT1(\
-    ::testing::internal::MakePredicateFormatterFromMatcher(matcher), value)
-#define EXPECT_THAT(value, matcher) EXPECT_PRED_FORMAT1(\
-    ::testing::internal::MakePredicateFormatterFromMatcher(matcher), value)
-
-}  // namespace testing
-
-// Include any custom callback matchers added by the local installation.
-// We must include this header at the end to make sure it can use the
-// declarations from this file.
-// Copyright 2015, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// ============================================================
-// An installation-specific extension point for gmock-matchers.h.
-// ============================================================
-//
-// Adds google3 callback support to CallableTraits.
-
-#ifndef GMOCK_INCLUDE_GMOCK_INTERNAL_CUSTOM_GMOCK_MATCHERS_H_
-#define GMOCK_INCLUDE_GMOCK_INTERNAL_CUSTOM_GMOCK_MATCHERS_H_
-#endif  // GMOCK_INCLUDE_GMOCK_INTERNAL_CUSTOM_GMOCK_MATCHERS_H_
-
-#endif  // GMOCK_INCLUDE_GMOCK_GMOCK_MATCHERS_H_
-
-#if GTEST_HAS_EXCEPTIONS
-# include <stdexcept>  // NOLINT
-#endif
-
-namespace testing {
-
-// An abstract handle of an expectation.
-class Expectation;
-
-// A set of expectation handles.
-class ExpectationSet;
-
-// Anything inside the 'internal' namespace IS INTERNAL IMPLEMENTATION
-// and MUST NOT BE USED IN USER CODE!!!
-namespace internal {
-
-// Implements a mock function.
-template <typename F> class FunctionMocker;
-
-// Base class for expectations.
-class ExpectationBase;
-
-// Implements an expectation.
-template <typename F> class TypedExpectation;
-
-// Helper class for testing the Expectation class template.
-class ExpectationTester;
-
-// Base class for function mockers.
-template <typename F> class FunctionMockerBase;
-
-// Protects the mock object registry (in class Mock), all function
-// mockers, and all expectations.
-//
-// The reason we don't use more fine-grained protection is: when a
-// mock function Foo() is called, it needs to consult its expectations
-// to see which one should be picked.  If another thread is allowed to
-// call a mock function (either Foo() or a different one) at the same
-// time, it could affect the "retired" attributes of Foo()'s
-// expectations when InSequence() is used, and thus affect which
-// expectation gets picked.  Therefore, we sequence all mock function
-// calls to ensure the integrity of the mock objects' states.
-GTEST_API_ GTEST_DECLARE_STATIC_MUTEX_(g_gmock_mutex);
-
-// Untyped base class for ActionResultHolder<R>.
-class UntypedActionResultHolderBase;
-
-// Abstract base class of FunctionMockerBase.  This is the
-// type-agnostic part of the function mocker interface.  Its pure
-// virtual methods are implemented by FunctionMockerBase.
-class GTEST_API_ UntypedFunctionMockerBase {
- public:
-  UntypedFunctionMockerBase();
-  virtual ~UntypedFunctionMockerBase();
-
-  // Verifies that all expectations on this mock function have been
-  // satisfied.  Reports one or more Google Test non-fatal failures
-  // and returns false if not.
-  bool VerifyAndClearExpectationsLocked()
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex);
-
-  // Clears the ON_CALL()s set on this mock function.
-  virtual void ClearDefaultActionsLocked()
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) = 0;
-
-  // In all of the following Untyped* functions, it's the caller's
-  // responsibility to guarantee the correctness of the arguments'
-  // types.
-
-  // Performs the default action with the given arguments and returns
-  // the action's result.  The call description string will be used in
-  // the error message to describe the call in the case the default
-  // action fails.
-  // L = *
-  virtual UntypedActionResultHolderBase* UntypedPerformDefaultAction(
-      void* untyped_args, const std::string& call_description) const = 0;
-
-  // Performs the given action with the given arguments and returns
-  // the action's result.
-  // L = *
-  virtual UntypedActionResultHolderBase* UntypedPerformAction(
-      const void* untyped_action, void* untyped_args) const = 0;
-
-  // Writes a message that the call is uninteresting (i.e. neither
-  // explicitly expected nor explicitly unexpected) to the given
-  // ostream.
-  virtual void UntypedDescribeUninterestingCall(
-      const void* untyped_args,
-      ::std::ostream* os) const
-          GTEST_LOCK_EXCLUDED_(g_gmock_mutex) = 0;
-
-  // Returns the expectation that matches the given function arguments
-  // (or NULL is there's no match); when a match is found,
-  // untyped_action is set to point to the action that should be
-  // performed (or NULL if the action is "do default"), and
-  // is_excessive is modified to indicate whether the call exceeds the
-  // expected number.
-  virtual const ExpectationBase* UntypedFindMatchingExpectation(
-      const void* untyped_args,
-      const void** untyped_action, bool* is_excessive,
-      ::std::ostream* what, ::std::ostream* why)
-          GTEST_LOCK_EXCLUDED_(g_gmock_mutex) = 0;
-
-  // Prints the given function arguments to the ostream.
-  virtual void UntypedPrintArgs(const void* untyped_args,
-                                ::std::ostream* os) const = 0;
-
-  // Sets the mock object this mock method belongs to, and registers
-  // this information in the global mock registry.  Will be called
-  // whenever an EXPECT_CALL() or ON_CALL() is executed on this mock
-  // method.
-  // TODO(wan@google.com): rename to SetAndRegisterOwner().
-  void RegisterOwner(const void* mock_obj)
-      GTEST_LOCK_EXCLUDED_(g_gmock_mutex);
-
-  // Sets the mock object this mock method belongs to, and sets the
-  // name of the mock function.  Will be called upon each invocation
-  // of this mock function.
-  void SetOwnerAndName(const void* mock_obj, const char* name)
-      GTEST_LOCK_EXCLUDED_(g_gmock_mutex);
-
-  // Returns the mock object this mock method belongs to.  Must be
-  // called after RegisterOwner() or SetOwnerAndName() has been
-  // called.
-  const void* MockObject() const
-      GTEST_LOCK_EXCLUDED_(g_gmock_mutex);
-
-  // Returns the name of this mock method.  Must be called after
-  // SetOwnerAndName() has been called.
-  const char* Name() const
-      GTEST_LOCK_EXCLUDED_(g_gmock_mutex);
-
-  // Returns the result of invoking this mock function with the given
-  // arguments.  This function can be safely called from multiple
-  // threads concurrently.  The caller is responsible for deleting the
-  // result.
-  UntypedActionResultHolderBase* UntypedInvokeWith(void* untyped_args)
-      GTEST_LOCK_EXCLUDED_(g_gmock_mutex);
-
- protected:
-  typedef std::vector<const void*> UntypedOnCallSpecs;
-
-  typedef std::vector<internal::linked_ptr<ExpectationBase> >
-  UntypedExpectations;
-
-  // Returns an Expectation object that references and co-owns exp,
-  // which must be an expectation on this mock function.
-  Expectation GetHandleOf(ExpectationBase* exp);
-
-  // Address of the mock object this mock method belongs to.  Only
-  // valid after this mock method has been called or
-  // ON_CALL/EXPECT_CALL has been invoked on it.
-  const void* mock_obj_;  // Protected by g_gmock_mutex.
-
-  // Name of the function being mocked.  Only valid after this mock
-  // method has been called.
-  const char* name_;  // Protected by g_gmock_mutex.
-
-  // All default action specs for this function mocker.
-  UntypedOnCallSpecs untyped_on_call_specs_;
-
-  // All expectations for this function mocker.
-  //
-  // It's undefined behavior to interleave expectations (EXPECT_CALLs
-  // or ON_CALLs) and mock function calls.  Also, the order of
-  // expectations is important.  Therefore it's a logic race condition
-  // to read/write untyped_expectations_ concurrently.  In order for
-  // tools like tsan to catch concurrent read/write accesses to
-  // untyped_expectations, we deliberately leave accesses to it
-  // unprotected.
-  UntypedExpectations untyped_expectations_;
-};  // class UntypedFunctionMockerBase
-
-// Untyped base class for OnCallSpec<F>.
-class UntypedOnCallSpecBase {
- public:
-  // The arguments are the location of the ON_CALL() statement.
-  UntypedOnCallSpecBase(const char* a_file, int a_line)
-      : file_(a_file), line_(a_line), last_clause_(kNone) {}
-
-  // Where in the source file was the default action spec defined?
-  const char* file() const { return file_; }
-  int line() const { return line_; }
-
- protected:
-  // Gives each clause in the ON_CALL() statement a name.
-  enum Clause {
-    // Do not change the order of the enum members!  The run-time
-    // syntax checking relies on it.
-    kNone,
-    kWith,
-    kWillByDefault
-  };
-
-  // Asserts that the ON_CALL() statement has a certain property.
-  void AssertSpecProperty(bool property,
-                          const std::string& failure_message) const {
-    Assert(property, file_, line_, failure_message);
-  }
-
-  // Expects that the ON_CALL() statement has a certain property.
-  void ExpectSpecProperty(bool property,
-                          const std::string& failure_message) const {
-    Expect(property, file_, line_, failure_message);
-  }
-
-  const char* file_;
-  int line_;
-
-  // The last clause in the ON_CALL() statement as seen so far.
-  // Initially kNone and changes as the statement is parsed.
-  Clause last_clause_;
-};  // class UntypedOnCallSpecBase
-
-// This template class implements an ON_CALL spec.
-template <typename F>
-class OnCallSpec : public UntypedOnCallSpecBase {
- public:
-  typedef typename Function<F>::ArgumentTuple ArgumentTuple;
-  typedef typename Function<F>::ArgumentMatcherTuple ArgumentMatcherTuple;
-
-  // Constructs an OnCallSpec object from the information inside
-  // the parenthesis of an ON_CALL() statement.
-  OnCallSpec(const char* a_file, int a_line,
-             const ArgumentMatcherTuple& matchers)
-      : UntypedOnCallSpecBase(a_file, a_line),
-        matchers_(matchers),
-        // By default, extra_matcher_ should match anything.  However,
-        // we cannot initialize it with _ as that triggers a compiler
-        // bug in Symbian's C++ compiler (cannot decide between two
-        // overloaded constructors of Matcher<const ArgumentTuple&>).
-        extra_matcher_(A<const ArgumentTuple&>()) {
-  }
-
-  // Implements the .With() clause.
-  OnCallSpec& With(const Matcher<const ArgumentTuple&>& m) {
-    // Makes sure this is called at most once.
-    ExpectSpecProperty(last_clause_ < kWith,
-                       ".With() cannot appear "
-                       "more than once in an ON_CALL().");
-    last_clause_ = kWith;
-
-    extra_matcher_ = m;
-    return *this;
-  }
-
-  // Implements the .WillByDefault() clause.
-  OnCallSpec& WillByDefault(const Action<F>& action) {
-    ExpectSpecProperty(last_clause_ < kWillByDefault,
-                       ".WillByDefault() must appear "
-                       "exactly once in an ON_CALL().");
-    last_clause_ = kWillByDefault;
-
-    ExpectSpecProperty(!action.IsDoDefault(),
-                       "DoDefault() cannot be used in ON_CALL().");
-    action_ = action;
-    return *this;
-  }
-
-  // Returns true iff the given arguments match the matchers.
-  bool Matches(const ArgumentTuple& args) const {
-    return TupleMatches(matchers_, args) && extra_matcher_.Matches(args);
-  }
-
-  // Returns the action specified by the user.
-  const Action<F>& GetAction() const {
-    AssertSpecProperty(last_clause_ == kWillByDefault,
-                       ".WillByDefault() must appear exactly "
-                       "once in an ON_CALL().");
-    return action_;
-  }
-
- private:
-  // The information in statement
-  //
-  //   ON_CALL(mock_object, Method(matchers))
-  //       .With(multi-argument-matcher)
-  //       .WillByDefault(action);
-  //
-  // is recorded in the data members like this:
-  //
-  //   source file that contains the statement => file_
-  //   line number of the statement            => line_
-  //   matchers                                => matchers_
-  //   multi-argument-matcher                  => extra_matcher_
-  //   action                                  => action_
-  ArgumentMatcherTuple matchers_;
-  Matcher<const ArgumentTuple&> extra_matcher_;
-  Action<F> action_;
-};  // class OnCallSpec
-
-// Possible reactions on uninteresting calls.
-enum CallReaction {
-  kAllow,
-  kWarn,
-  kFail,
-};
-
-}  // namespace internal
-
-// Utilities for manipulating mock objects.
-class GTEST_API_ Mock {
- public:
-  // The following public methods can be called concurrently.
-
-  // Tells Google Mock to ignore mock_obj when checking for leaked
-  // mock objects.
-  static void AllowLeak(const void* mock_obj)
-      GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex);
-
-  // Verifies and clears all expectations on the given mock object.
-  // If the expectations aren't satisfied, generates one or more
-  // Google Test non-fatal failures and returns false.
-  static bool VerifyAndClearExpectations(void* mock_obj)
-      GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex);
-
-  // Verifies all expectations on the given mock object and clears its
-  // default actions and expectations.  Returns true iff the
-  // verification was successful.
-  static bool VerifyAndClear(void* mock_obj)
-      GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex);
-
- private:
-  friend class internal::UntypedFunctionMockerBase;
-
-  // Needed for a function mocker to register itself (so that we know
-  // how to clear a mock object).
-  template <typename F>
-  friend class internal::FunctionMockerBase;
-
-  template <typename M>
-  friend class NiceMock;
-
-  template <typename M>
-  friend class NaggyMock;
-
-  template <typename M>
-  friend class StrictMock;
-
-  // Tells Google Mock to allow uninteresting calls on the given mock
-  // object.
-  static void AllowUninterestingCalls(const void* mock_obj)
-      GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex);
-
-  // Tells Google Mock to warn the user about uninteresting calls on
-  // the given mock object.
-  static void WarnUninterestingCalls(const void* mock_obj)
-      GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex);
-
-  // Tells Google Mock to fail uninteresting calls on the given mock
-  // object.
-  static void FailUninterestingCalls(const void* mock_obj)
-      GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex);
-
-  // Tells Google Mock the given mock object is being destroyed and
-  // its entry in the call-reaction table should be removed.
-  static void UnregisterCallReaction(const void* mock_obj)
-      GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex);
-
-  // Returns the reaction Google Mock will have on uninteresting calls
-  // made on the given mock object.
-  static internal::CallReaction GetReactionOnUninterestingCalls(
-      const void* mock_obj)
-          GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex);
-
-  // Verifies that all expectations on the given mock object have been
-  // satisfied.  Reports one or more Google Test non-fatal failures
-  // and returns false if not.
-  static bool VerifyAndClearExpectationsLocked(void* mock_obj)
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(internal::g_gmock_mutex);
-
-  // Clears all ON_CALL()s set on the given mock object.
-  static void ClearDefaultActionsLocked(void* mock_obj)
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(internal::g_gmock_mutex);
-
-  // Registers a mock object and a mock method it owns.
-  static void Register(
-      const void* mock_obj,
-      internal::UntypedFunctionMockerBase* mocker)
-          GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex);
-
-  // Tells Google Mock where in the source code mock_obj is used in an
-  // ON_CALL or EXPECT_CALL.  In case mock_obj is leaked, this
-  // information helps the user identify which object it is.
-  static void RegisterUseByOnCallOrExpectCall(
-      const void* mock_obj, const char* file, int line)
-          GTEST_LOCK_EXCLUDED_(internal::g_gmock_mutex);
-
-  // Unregisters a mock method; removes the owning mock object from
-  // the registry when the last mock method associated with it has
-  // been unregistered.  This is called only in the destructor of
-  // FunctionMockerBase.
-  static void UnregisterLocked(internal::UntypedFunctionMockerBase* mocker)
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(internal::g_gmock_mutex);
-};  // class Mock
-
-// An abstract handle of an expectation.  Useful in the .After()
-// clause of EXPECT_CALL() for setting the (partial) order of
-// expectations.  The syntax:
-//
-//   Expectation e1 = EXPECT_CALL(...)...;
-//   EXPECT_CALL(...).After(e1)...;
-//
-// sets two expectations where the latter can only be matched after
-// the former has been satisfied.
-//
-// Notes:
-//   - This class is copyable and has value semantics.
-//   - Constness is shallow: a const Expectation object itself cannot
-//     be modified, but the mutable methods of the ExpectationBase
-//     object it references can be called via expectation_base().
-//   - The constructors and destructor are defined out-of-line because
-//     the Symbian WINSCW compiler wants to otherwise instantiate them
-//     when it sees this class definition, at which point it doesn't have
-//     ExpectationBase available yet, leading to incorrect destruction
-//     in the linked_ptr (or compilation errors if using a checking
-//     linked_ptr).
-class GTEST_API_ Expectation {
- public:
-  // Constructs a null object that doesn't reference any expectation.
-  Expectation();
-
-  ~Expectation();
-
-  // This single-argument ctor must not be explicit, in order to support the
-  //   Expectation e = EXPECT_CALL(...);
-  // syntax.
-  //
-  // A TypedExpectation object stores its pre-requisites as
-  // Expectation objects, and needs to call the non-const Retire()
-  // method on the ExpectationBase objects they reference.  Therefore
-  // Expectation must receive a *non-const* reference to the
-  // ExpectationBase object.
-  Expectation(internal::ExpectationBase& exp);  // NOLINT
-
-  // The compiler-generated copy ctor and operator= work exactly as
-  // intended, so we don't need to define our own.
-
-  // Returns true iff rhs references the same expectation as this object does.
-  bool operator==(const Expectation& rhs) const {
-    return expectation_base_ == rhs.expectation_base_;
-  }
-
-  bool operator!=(const Expectation& rhs) const { return !(*this == rhs); }
-
- private:
-  friend class ExpectationSet;
-  friend class Sequence;
-  friend class ::testing::internal::ExpectationBase;
-  friend class ::testing::internal::UntypedFunctionMockerBase;
-
-  template <typename F>
-  friend class ::testing::internal::FunctionMockerBase;
-
-  template <typename F>
-  friend class ::testing::internal::TypedExpectation;
-
-  // This comparator is needed for putting Expectation objects into a set.
-  class Less {
-   public:
-    bool operator()(const Expectation& lhs, const Expectation& rhs) const {
-      return lhs.expectation_base_.get() < rhs.expectation_base_.get();
-    }
-  };
-
-  typedef ::std::set<Expectation, Less> Set;
-
-  Expectation(
-      const internal::linked_ptr<internal::ExpectationBase>& expectation_base);
-
-  // Returns the expectation this object references.
-  const internal::linked_ptr<internal::ExpectationBase>&
-  expectation_base() const {
-    return expectation_base_;
-  }
-
-  // A linked_ptr that co-owns the expectation this handle references.
-  internal::linked_ptr<internal::ExpectationBase> expectation_base_;
-};
-
-// A set of expectation handles.  Useful in the .After() clause of
-// EXPECT_CALL() for setting the (partial) order of expectations.  The
-// syntax:
-//
-//   ExpectationSet es;
-//   es += EXPECT_CALL(...)...;
-//   es += EXPECT_CALL(...)...;
-//   EXPECT_CALL(...).After(es)...;
-//
-// sets three expectations where the last one can only be matched
-// after the first two have both been satisfied.
-//
-// This class is copyable and has value semantics.
-class ExpectationSet {
- public:
-  // A bidirectional iterator that can read a const element in the set.
-  typedef Expectation::Set::const_iterator const_iterator;
-
-  // An object stored in the set.  This is an alias of Expectation.
-  typedef Expectation::Set::value_type value_type;
-
-  // Constructs an empty set.
-  ExpectationSet() {}
-
-  // This single-argument ctor must not be explicit, in order to support the
-  //   ExpectationSet es = EXPECT_CALL(...);
-  // syntax.
-  ExpectationSet(internal::ExpectationBase& exp) {  // NOLINT
-    *this += Expectation(exp);
-  }
-
-  // This single-argument ctor implements implicit conversion from
-  // Expectation and thus must not be explicit.  This allows either an
-  // Expectation or an ExpectationSet to be used in .After().
-  ExpectationSet(const Expectation& e) {  // NOLINT
-    *this += e;
-  }
-
-  // The compiler-generator ctor and operator= works exactly as
-  // intended, so we don't need to define our own.
-
-  // Returns true iff rhs contains the same set of Expectation objects
-  // as this does.
-  bool operator==(const ExpectationSet& rhs) const {
-    return expectations_ == rhs.expectations_;
-  }
-
-  bool operator!=(const ExpectationSet& rhs) const { return !(*this == rhs); }
-
-  // Implements the syntax
-  //   expectation_set += EXPECT_CALL(...);
-  ExpectationSet& operator+=(const Expectation& e) {
-    expectations_.insert(e);
-    return *this;
-  }
-
-  int size() const { return static_cast<int>(expectations_.size()); }
-
-  const_iterator begin() const { return expectations_.begin(); }
-  const_iterator end() const { return expectations_.end(); }
-
- private:
-  Expectation::Set expectations_;
-};
-
-
-// Sequence objects are used by a user to specify the relative order
-// in which the expectations should match.  They are copyable (we rely
-// on the compiler-defined copy constructor and assignment operator).
-class GTEST_API_ Sequence {
- public:
-  // Constructs an empty sequence.
-  Sequence() : last_expectation_(new Expectation) {}
-
-  // Adds an expectation to this sequence.  The caller must ensure
-  // that no other thread is accessing this Sequence object.
-  void AddExpectation(const Expectation& expectation) const;
-
- private:
-  // The last expectation in this sequence.  We use a linked_ptr here
-  // because Sequence objects are copyable and we want the copies to
-  // be aliases.  The linked_ptr allows the copies to co-own and share
-  // the same Expectation object.
-  internal::linked_ptr<Expectation> last_expectation_;
-};  // class Sequence
-
-// An object of this type causes all EXPECT_CALL() statements
-// encountered in its scope to be put in an anonymous sequence.  The
-// work is done in the constructor and destructor.  You should only
-// create an InSequence object on the stack.
-//
-// The sole purpose for this class is to support easy definition of
-// sequential expectations, e.g.
-//
-//   {
-//     InSequence dummy;  // The name of the object doesn't matter.
-//
-//     // The following expectations must match in the order they appear.
-//     EXPECT_CALL(a, Bar())...;
-//     EXPECT_CALL(a, Baz())...;
-//     ...
-//     EXPECT_CALL(b, Xyz())...;
-//   }
-//
-// You can create InSequence objects in multiple threads, as long as
-// they are used to affect different mock objects.  The idea is that
-// each thread can create and set up its own mocks as if it's the only
-// thread.  However, for clarity of your tests we recommend you to set
-// up mocks in the main thread unless you have a good reason not to do
-// so.
-class GTEST_API_ InSequence {
- public:
-  InSequence();
-  ~InSequence();
- private:
-  bool sequence_created_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(InSequence);  // NOLINT
-} GTEST_ATTRIBUTE_UNUSED_;
-
-namespace internal {
-
-// Points to the implicit sequence introduced by a living InSequence
-// object (if any) in the current thread or NULL.
-GTEST_API_ extern ThreadLocal<Sequence*> g_gmock_implicit_sequence;
-
-// Base class for implementing expectations.
-//
-// There are two reasons for having a type-agnostic base class for
-// Expectation:
-//
-//   1. We need to store collections of expectations of different
-//   types (e.g. all pre-requisites of a particular expectation, all
-//   expectations in a sequence).  Therefore these expectation objects
-//   must share a common base class.
-//
-//   2. We can avoid binary code bloat by moving methods not depending
-//   on the template argument of Expectation to the base class.
-//
-// This class is internal and mustn't be used by user code directly.
-class GTEST_API_ ExpectationBase {
- public:
-  // source_text is the EXPECT_CALL(...) source that created this Expectation.
-  ExpectationBase(const char* file, int line, const std::string& source_text);
-
-  virtual ~ExpectationBase();
-
-  // Where in the source file was the expectation spec defined?
-  const char* file() const { return file_; }
-  int line() const { return line_; }
-  const char* source_text() const { return source_text_.c_str(); }
-  // Returns the cardinality specified in the expectation spec.
-  const Cardinality& cardinality() const { return cardinality_; }
-
-  // Describes the source file location of this expectation.
-  void DescribeLocationTo(::std::ostream* os) const {
-    *os << FormatFileLocation(file(), line()) << " ";
-  }
-
-  // Describes how many times a function call matching this
-  // expectation has occurred.
-  void DescribeCallCountTo(::std::ostream* os) const
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex);
-
-  // If this mock method has an extra matcher (i.e. .With(matcher)),
-  // describes it to the ostream.
-  virtual void MaybeDescribeExtraMatcherTo(::std::ostream* os) = 0;
-
- protected:
-  friend class ::testing::Expectation;
-  friend class UntypedFunctionMockerBase;
-
-  enum Clause {
-    // Don't change the order of the enum members!
-    kNone,
-    kWith,
-    kTimes,
-    kInSequence,
-    kAfter,
-    kWillOnce,
-    kWillRepeatedly,
-    kRetiresOnSaturation
-  };
-
-  typedef std::vector<const void*> UntypedActions;
-
-  // Returns an Expectation object that references and co-owns this
-  // expectation.
-  virtual Expectation GetHandle() = 0;
-
-  // Asserts that the EXPECT_CALL() statement has the given property.
-  void AssertSpecProperty(bool property,
-                          const std::string& failure_message) const {
-    Assert(property, file_, line_, failure_message);
-  }
-
-  // Expects that the EXPECT_CALL() statement has the given property.
-  void ExpectSpecProperty(bool property,
-                          const std::string& failure_message) const {
-    Expect(property, file_, line_, failure_message);
-  }
-
-  // Explicitly specifies the cardinality of this expectation.  Used
-  // by the subclasses to implement the .Times() clause.
-  void SpecifyCardinality(const Cardinality& cardinality);
-
-  // Returns true iff the user specified the cardinality explicitly
-  // using a .Times().
-  bool cardinality_specified() const { return cardinality_specified_; }
-
-  // Sets the cardinality of this expectation spec.
-  void set_cardinality(const Cardinality& a_cardinality) {
-    cardinality_ = a_cardinality;
-  }
-
-  // The following group of methods should only be called after the
-  // EXPECT_CALL() statement, and only when g_gmock_mutex is held by
-  // the current thread.
-
-  // Retires all pre-requisites of this expectation.
-  void RetireAllPreRequisites()
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex);
-
-  // Returns true iff this expectation is retired.
-  bool is_retired() const
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    return retired_;
-  }
-
-  // Retires this expectation.
-  void Retire()
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    retired_ = true;
-  }
-
-  // Returns true iff this expectation is satisfied.
-  bool IsSatisfied() const
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    return cardinality().IsSatisfiedByCallCount(call_count_);
-  }
-
-  // Returns true iff this expectation is saturated.
-  bool IsSaturated() const
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    return cardinality().IsSaturatedByCallCount(call_count_);
-  }
-
-  // Returns true iff this expectation is over-saturated.
-  bool IsOverSaturated() const
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    return cardinality().IsOverSaturatedByCallCount(call_count_);
-  }
-
-  // Returns true iff all pre-requisites of this expectation are satisfied.
-  bool AllPrerequisitesAreSatisfied() const
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex);
-
-  // Adds unsatisfied pre-requisites of this expectation to 'result'.
-  void FindUnsatisfiedPrerequisites(ExpectationSet* result) const
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex);
-
-  // Returns the number this expectation has been invoked.
-  int call_count() const
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    return call_count_;
-  }
-
-  // Increments the number this expectation has been invoked.
-  void IncrementCallCount()
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    call_count_++;
-  }
-
-  // Checks the action count (i.e. the number of WillOnce() and
-  // WillRepeatedly() clauses) against the cardinality if this hasn't
-  // been done before.  Prints a warning if there are too many or too
-  // few actions.
-  void CheckActionCountIfNotDone() const
-      GTEST_LOCK_EXCLUDED_(mutex_);
-
-  friend class ::testing::Sequence;
-  friend class ::testing::internal::ExpectationTester;
-
-  template <typename Function>
-  friend class TypedExpectation;
-
-  // Implements the .Times() clause.
-  void UntypedTimes(const Cardinality& a_cardinality);
-
-  // This group of fields are part of the spec and won't change after
-  // an EXPECT_CALL() statement finishes.
-  const char* file_;          // The file that contains the expectation.
-  int line_;                  // The line number of the expectation.
-  const std::string source_text_;  // The EXPECT_CALL(...) source text.
-  // True iff the cardinality is specified explicitly.
-  bool cardinality_specified_;
-  Cardinality cardinality_;            // The cardinality of the expectation.
-  // The immediate pre-requisites (i.e. expectations that must be
-  // satisfied before this expectation can be matched) of this
-  // expectation.  We use linked_ptr in the set because we want an
-  // Expectation object to be co-owned by its FunctionMocker and its
-  // successors.  This allows multiple mock objects to be deleted at
-  // different times.
-  ExpectationSet immediate_prerequisites_;
-
-  // This group of fields are the current state of the expectation,
-  // and can change as the mock function is called.
-  int call_count_;  // How many times this expectation has been invoked.
-  bool retired_;    // True iff this expectation has retired.
-  UntypedActions untyped_actions_;
-  bool extra_matcher_specified_;
-  bool repeated_action_specified_;  // True if a WillRepeatedly() was specified.
-  bool retires_on_saturation_;
-  Clause last_clause_;
-  mutable bool action_count_checked_;  // Under mutex_.
-  mutable Mutex mutex_;  // Protects action_count_checked_.
-
-  GTEST_DISALLOW_ASSIGN_(ExpectationBase);
-};  // class ExpectationBase
-
-// Impements an expectation for the given function type.
-template <typename F>
-class TypedExpectation : public ExpectationBase {
- public:
-  typedef typename Function<F>::ArgumentTuple ArgumentTuple;
-  typedef typename Function<F>::ArgumentMatcherTuple ArgumentMatcherTuple;
-  typedef typename Function<F>::Result Result;
-
-  TypedExpectation(FunctionMockerBase<F>* owner, const char* a_file, int a_line,
-                   const std::string& a_source_text,
-                   const ArgumentMatcherTuple& m)
-      : ExpectationBase(a_file, a_line, a_source_text),
-        owner_(owner),
-        matchers_(m),
-        // By default, extra_matcher_ should match anything.  However,
-        // we cannot initialize it with _ as that triggers a compiler
-        // bug in Symbian's C++ compiler (cannot decide between two
-        // overloaded constructors of Matcher<const ArgumentTuple&>).
-        extra_matcher_(A<const ArgumentTuple&>()),
-        repeated_action_(DoDefault()) {}
-
-  virtual ~TypedExpectation() {
-    // Check the validity of the action count if it hasn't been done
-    // yet (for example, if the expectation was never used).
-    CheckActionCountIfNotDone();
-    for (UntypedActions::const_iterator it = untyped_actions_.begin();
-         it != untyped_actions_.end(); ++it) {
-      delete static_cast<const Action<F>*>(*it);
-    }
-  }
-
-  // Implements the .With() clause.
-  TypedExpectation& With(const Matcher<const ArgumentTuple&>& m) {
-    if (last_clause_ == kWith) {
-      ExpectSpecProperty(false,
-                         ".With() cannot appear "
-                         "more than once in an EXPECT_CALL().");
-    } else {
-      ExpectSpecProperty(last_clause_ < kWith,
-                         ".With() must be the first "
-                         "clause in an EXPECT_CALL().");
-    }
-    last_clause_ = kWith;
-
-    extra_matcher_ = m;
-    extra_matcher_specified_ = true;
-    return *this;
-  }
-
-  // Implements the .Times() clause.
-  TypedExpectation& Times(const Cardinality& a_cardinality) {
-    ExpectationBase::UntypedTimes(a_cardinality);
-    return *this;
-  }
-
-  // Implements the .Times() clause.
-  TypedExpectation& Times(int n) {
-    return Times(Exactly(n));
-  }
-
-  // Implements the .InSequence() clause.
-  TypedExpectation& InSequence(const Sequence& s) {
-    ExpectSpecProperty(last_clause_ <= kInSequence,
-                       ".InSequence() cannot appear after .After(),"
-                       " .WillOnce(), .WillRepeatedly(), or "
-                       ".RetiresOnSaturation().");
-    last_clause_ = kInSequence;
-
-    s.AddExpectation(GetHandle());
-    return *this;
-  }
-  TypedExpectation& InSequence(const Sequence& s1, const Sequence& s2) {
-    return InSequence(s1).InSequence(s2);
-  }
-  TypedExpectation& InSequence(const Sequence& s1, const Sequence& s2,
-                               const Sequence& s3) {
-    return InSequence(s1, s2).InSequence(s3);
-  }
-  TypedExpectation& InSequence(const Sequence& s1, const Sequence& s2,
-                               const Sequence& s3, const Sequence& s4) {
-    return InSequence(s1, s2, s3).InSequence(s4);
-  }
-  TypedExpectation& InSequence(const Sequence& s1, const Sequence& s2,
-                               const Sequence& s3, const Sequence& s4,
-                               const Sequence& s5) {
-    return InSequence(s1, s2, s3, s4).InSequence(s5);
-  }
-
-  // Implements that .After() clause.
-  TypedExpectation& After(const ExpectationSet& s) {
-    ExpectSpecProperty(last_clause_ <= kAfter,
-                       ".After() cannot appear after .WillOnce(),"
-                       " .WillRepeatedly(), or "
-                       ".RetiresOnSaturation().");
-    last_clause_ = kAfter;
-
-    for (ExpectationSet::const_iterator it = s.begin(); it != s.end(); ++it) {
-      immediate_prerequisites_ += *it;
-    }
-    return *this;
-  }
-  TypedExpectation& After(const ExpectationSet& s1, const ExpectationSet& s2) {
-    return After(s1).After(s2);
-  }
-  TypedExpectation& After(const ExpectationSet& s1, const ExpectationSet& s2,
-                          const ExpectationSet& s3) {
-    return After(s1, s2).After(s3);
-  }
-  TypedExpectation& After(const ExpectationSet& s1, const ExpectationSet& s2,
-                          const ExpectationSet& s3, const ExpectationSet& s4) {
-    return After(s1, s2, s3).After(s4);
-  }
-  TypedExpectation& After(const ExpectationSet& s1, const ExpectationSet& s2,
-                          const ExpectationSet& s3, const ExpectationSet& s4,
-                          const ExpectationSet& s5) {
-    return After(s1, s2, s3, s4).After(s5);
-  }
-
-  // Implements the .WillOnce() clause.
-  TypedExpectation& WillOnce(const Action<F>& action) {
-    ExpectSpecProperty(last_clause_ <= kWillOnce,
-                       ".WillOnce() cannot appear after "
-                       ".WillRepeatedly() or .RetiresOnSaturation().");
-    last_clause_ = kWillOnce;
-
-    untyped_actions_.push_back(new Action<F>(action));
-    if (!cardinality_specified()) {
-      set_cardinality(Exactly(static_cast<int>(untyped_actions_.size())));
-    }
-    return *this;
-  }
-
-  // Implements the .WillRepeatedly() clause.
-  TypedExpectation& WillRepeatedly(const Action<F>& action) {
-    if (last_clause_ == kWillRepeatedly) {
-      ExpectSpecProperty(false,
-                         ".WillRepeatedly() cannot appear "
-                         "more than once in an EXPECT_CALL().");
-    } else {
-      ExpectSpecProperty(last_clause_ < kWillRepeatedly,
-                         ".WillRepeatedly() cannot appear "
-                         "after .RetiresOnSaturation().");
-    }
-    last_clause_ = kWillRepeatedly;
-    repeated_action_specified_ = true;
-
-    repeated_action_ = action;
-    if (!cardinality_specified()) {
-      set_cardinality(AtLeast(static_cast<int>(untyped_actions_.size())));
-    }
-
-    // Now that no more action clauses can be specified, we check
-    // whether their count makes sense.
-    CheckActionCountIfNotDone();
-    return *this;
-  }
-
-  // Implements the .RetiresOnSaturation() clause.
-  TypedExpectation& RetiresOnSaturation() {
-    ExpectSpecProperty(last_clause_ < kRetiresOnSaturation,
-                       ".RetiresOnSaturation() cannot appear "
-                       "more than once.");
-    last_clause_ = kRetiresOnSaturation;
-    retires_on_saturation_ = true;
-
-    // Now that no more action clauses can be specified, we check
-    // whether their count makes sense.
-    CheckActionCountIfNotDone();
-    return *this;
-  }
-
-  // Returns the matchers for the arguments as specified inside the
-  // EXPECT_CALL() macro.
-  const ArgumentMatcherTuple& matchers() const {
-    return matchers_;
-  }
-
-  // Returns the matcher specified by the .With() clause.
-  const Matcher<const ArgumentTuple&>& extra_matcher() const {
-    return extra_matcher_;
-  }
-
-  // Returns the action specified by the .WillRepeatedly() clause.
-  const Action<F>& repeated_action() const { return repeated_action_; }
-
-  // If this mock method has an extra matcher (i.e. .With(matcher)),
-  // describes it to the ostream.
-  virtual void MaybeDescribeExtraMatcherTo(::std::ostream* os) {
-    if (extra_matcher_specified_) {
-      *os << "    Expected args: ";
-      extra_matcher_.DescribeTo(os);
-      *os << "\n";
-    }
-  }
-
- private:
-  template <typename Function>
-  friend class FunctionMockerBase;
-
-  // Returns an Expectation object that references and co-owns this
-  // expectation.
-  virtual Expectation GetHandle() {
-    return owner_->GetHandleOf(this);
-  }
-
-  // The following methods will be called only after the EXPECT_CALL()
-  // statement finishes and when the current thread holds
-  // g_gmock_mutex.
-
-  // Returns true iff this expectation matches the given arguments.
-  bool Matches(const ArgumentTuple& args) const
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    return TupleMatches(matchers_, args) && extra_matcher_.Matches(args);
-  }
-
-  // Returns true iff this expectation should handle the given arguments.
-  bool ShouldHandleArguments(const ArgumentTuple& args) const
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-
-    // In case the action count wasn't checked when the expectation
-    // was defined (e.g. if this expectation has no WillRepeatedly()
-    // or RetiresOnSaturation() clause), we check it when the
-    // expectation is used for the first time.
-    CheckActionCountIfNotDone();
-    return !is_retired() && AllPrerequisitesAreSatisfied() && Matches(args);
-  }
-
-  // Describes the result of matching the arguments against this
-  // expectation to the given ostream.
-  void ExplainMatchResultTo(
-      const ArgumentTuple& args,
-      ::std::ostream* os) const
-          GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-
-    if (is_retired()) {
-      *os << "         Expected: the expectation is active\n"
-          << "           Actual: it is retired\n";
-    } else if (!Matches(args)) {
-      if (!TupleMatches(matchers_, args)) {
-        ExplainMatchFailureTupleTo(matchers_, args, os);
-      }
-      StringMatchResultListener listener;
-      if (!extra_matcher_.MatchAndExplain(args, &listener)) {
-        *os << "    Expected args: ";
-        extra_matcher_.DescribeTo(os);
-        *os << "\n           Actual: don't match";
-
-        internal::PrintIfNotEmpty(listener.str(), os);
-        *os << "\n";
-      }
-    } else if (!AllPrerequisitesAreSatisfied()) {
-      *os << "         Expected: all pre-requisites are satisfied\n"
-          << "           Actual: the following immediate pre-requisites "
-          << "are not satisfied:\n";
-      ExpectationSet unsatisfied_prereqs;
-      FindUnsatisfiedPrerequisites(&unsatisfied_prereqs);
-      int i = 0;
-      for (ExpectationSet::const_iterator it = unsatisfied_prereqs.begin();
-           it != unsatisfied_prereqs.end(); ++it) {
-        it->expectation_base()->DescribeLocationTo(os);
-        *os << "pre-requisite #" << i++ << "\n";
-      }
-      *os << "                   (end of pre-requisites)\n";
-    } else {
-      // This line is here just for completeness' sake.  It will never
-      // be executed as currently the ExplainMatchResultTo() function
-      // is called only when the mock function call does NOT match the
-      // expectation.
-      *os << "The call matches the expectation.\n";
-    }
-  }
-
-  // Returns the action that should be taken for the current invocation.
-  const Action<F>& GetCurrentAction(
-      const FunctionMockerBase<F>* mocker,
-      const ArgumentTuple& args) const
-          GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    const int count = call_count();
-    Assert(count >= 1, __FILE__, __LINE__,
-           "call_count() is <= 0 when GetCurrentAction() is "
-           "called - this should never happen.");
-
-    const int action_count = static_cast<int>(untyped_actions_.size());
-    if (action_count > 0 && !repeated_action_specified_ &&
-        count > action_count) {
-      // If there is at least one WillOnce() and no WillRepeatedly(),
-      // we warn the user when the WillOnce() clauses ran out.
-      ::std::stringstream ss;
-      DescribeLocationTo(&ss);
-      ss << "Actions ran out in " << source_text() << "...\n"
-         << "Called " << count << " times, but only "
-         << action_count << " WillOnce()"
-         << (action_count == 1 ? " is" : "s are") << " specified - ";
-      mocker->DescribeDefaultActionTo(args, &ss);
-      Log(kWarning, ss.str(), 1);
-    }
-
-    return count <= action_count ?
-        *static_cast<const Action<F>*>(untyped_actions_[count - 1]) :
-        repeated_action();
-  }
-
-  // Given the arguments of a mock function call, if the call will
-  // over-saturate this expectation, returns the default action;
-  // otherwise, returns the next action in this expectation.  Also
-  // describes *what* happened to 'what', and explains *why* Google
-  // Mock does it to 'why'.  This method is not const as it calls
-  // IncrementCallCount().  A return value of NULL means the default
-  // action.
-  const Action<F>* GetActionForArguments(
-      const FunctionMockerBase<F>* mocker,
-      const ArgumentTuple& args,
-      ::std::ostream* what,
-      ::std::ostream* why)
-          GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    if (IsSaturated()) {
-      // We have an excessive call.
-      IncrementCallCount();
-      *what << "Mock function called more times than expected - ";
-      mocker->DescribeDefaultActionTo(args, what);
-      DescribeCallCountTo(why);
-
-      // TODO(wan@google.com): allow the user to control whether
-      // unexpected calls should fail immediately or continue using a
-      // flag --gmock_unexpected_calls_are_fatal.
-      return NULL;
-    }
-
-    IncrementCallCount();
-    RetireAllPreRequisites();
-
-    if (retires_on_saturation_ && IsSaturated()) {
-      Retire();
-    }
-
-    // Must be done after IncrementCount()!
-    *what << "Mock function call matches " << source_text() <<"...\n";
-    return &(GetCurrentAction(mocker, args));
-  }
-
-  // All the fields below won't change once the EXPECT_CALL()
-  // statement finishes.
-  FunctionMockerBase<F>* const owner_;
-  ArgumentMatcherTuple matchers_;
-  Matcher<const ArgumentTuple&> extra_matcher_;
-  Action<F> repeated_action_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(TypedExpectation);
-};  // class TypedExpectation
-
-// A MockSpec object is used by ON_CALL() or EXPECT_CALL() for
-// specifying the default behavior of, or expectation on, a mock
-// function.
-
-// Note: class MockSpec really belongs to the ::testing namespace.
-// However if we define it in ::testing, MSVC will complain when
-// classes in ::testing::internal declare it as a friend class
-// template.  To workaround this compiler bug, we define MockSpec in
-// ::testing::internal and import it into ::testing.
-
-// Logs a message including file and line number information.
-GTEST_API_ void LogWithLocation(testing::internal::LogSeverity severity,
-                                const char* file, int line,
-                                const std::string& message);
-
-template <typename F>
-class MockSpec {
- public:
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-  typedef typename internal::Function<F>::ArgumentMatcherTuple
-      ArgumentMatcherTuple;
-
-  // Constructs a MockSpec object, given the function mocker object
-  // that the spec is associated with.
-  MockSpec(internal::FunctionMockerBase<F>* function_mocker,
-           const ArgumentMatcherTuple& matchers)
-      : function_mocker_(function_mocker), matchers_(matchers) {}
-
-  // Adds a new default action spec to the function mocker and returns
-  // the newly created spec.
-  internal::OnCallSpec<F>& InternalDefaultActionSetAt(
-      const char* file, int line, const char* obj, const char* call) {
-    LogWithLocation(internal::kInfo, file, line,
-                    std::string("ON_CALL(") + obj + ", " + call + ") invoked");
-    return function_mocker_->AddNewOnCallSpec(file, line, matchers_);
-  }
-
-  // Adds a new expectation spec to the function mocker and returns
-  // the newly created spec.
-  internal::TypedExpectation<F>& InternalExpectedAt(
-      const char* file, int line, const char* obj, const char* call) {
-    const std::string source_text(std::string("EXPECT_CALL(") + obj + ", " +
-                                  call + ")");
-    LogWithLocation(internal::kInfo, file, line, source_text + " invoked");
-    return function_mocker_->AddNewExpectation(
-        file, line, source_text, matchers_);
-  }
-
-  // This operator overload is used to swallow the superfluous parameter list
-  // introduced by the ON/EXPECT_CALL macros. See the macro comments for more
-  // explanation.
-  MockSpec<F>& operator()(const internal::WithoutMatchers&, void* const) {
-    return *this;
-  }
-
- private:
-  template <typename Function>
-  friend class internal::FunctionMocker;
-
-  // The function mocker that owns this spec.
-  internal::FunctionMockerBase<F>* const function_mocker_;
-  // The argument matchers specified in the spec.
-  ArgumentMatcherTuple matchers_;
-
-  GTEST_DISALLOW_ASSIGN_(MockSpec);
-};  // class MockSpec
-
-// Wrapper type for generically holding an ordinary value or lvalue reference.
-// If T is not a reference type, it must be copyable or movable.
-// ReferenceOrValueWrapper<T> is movable, and will also be copyable unless
-// T is a move-only value type (which means that it will always be copyable
-// if the current platform does not support move semantics).
-//
-// The primary template defines handling for values, but function header
-// comments describe the contract for the whole template (including
-// specializations).
-template <typename T>
-class ReferenceOrValueWrapper {
- public:
-  // Constructs a wrapper from the given value/reference.
-  explicit ReferenceOrValueWrapper(T value)
-      : value_(::testing::internal::move(value)) {
-  }
-
-  // Unwraps and returns the underlying value/reference, exactly as
-  // originally passed. The behavior of calling this more than once on
-  // the same object is unspecified.
-  T Unwrap() { return ::testing::internal::move(value_); }
-
-  // Provides nondestructive access to the underlying value/reference.
-  // Always returns a const reference (more precisely,
-  // const RemoveReference<T>&). The behavior of calling this after
-  // calling Unwrap on the same object is unspecified.
-  const T& Peek() const {
-    return value_;
-  }
-
- private:
-  T value_;
-};
-
-// Specialization for lvalue reference types. See primary template
-// for documentation.
-template <typename T>
-class ReferenceOrValueWrapper<T&> {
- public:
-  // Workaround for debatable pass-by-reference lint warning (c-library-team
-  // policy precludes NOLINT in this context)
-  typedef T& reference;
-  explicit ReferenceOrValueWrapper(reference ref)
-      : value_ptr_(&ref) {}
-  T& Unwrap() { return *value_ptr_; }
-  const T& Peek() const { return *value_ptr_; }
-
- private:
-  T* value_ptr_;
-};
-
-// MSVC warns about using 'this' in base member initializer list, so
-// we need to temporarily disable the warning.  We have to do it for
-// the entire class to suppress the warning, even though it's about
-// the constructor only.
-
-#ifdef _MSC_VER
-# pragma warning(push)          // Saves the current warning state.
-# pragma warning(disable:4355)  // Temporarily disables warning 4355.
-#endif  // _MSV_VER
-
-// C++ treats the void type specially.  For example, you cannot define
-// a void-typed variable or pass a void value to a function.
-// ActionResultHolder<T> holds a value of type T, where T must be a
-// copyable type or void (T doesn't need to be default-constructable).
-// It hides the syntactic difference between void and other types, and
-// is used to unify the code for invoking both void-returning and
-// non-void-returning mock functions.
-
-// Untyped base class for ActionResultHolder<T>.
-class UntypedActionResultHolderBase {
- public:
-  virtual ~UntypedActionResultHolderBase() {}
-
-  // Prints the held value as an action's result to os.
-  virtual void PrintAsActionResult(::std::ostream* os) const = 0;
-};
-
-// This generic definition is used when T is not void.
-template <typename T>
-class ActionResultHolder : public UntypedActionResultHolderBase {
- public:
-  // Returns the held value. Must not be called more than once.
-  T Unwrap() {
-    return result_.Unwrap();
-  }
-
-  // Prints the held value as an action's result to os.
-  virtual void PrintAsActionResult(::std::ostream* os) const {
-    *os << "\n          Returns: ";
-    // T may be a reference type, so we don't use UniversalPrint().
-    UniversalPrinter<T>::Print(result_.Peek(), os);
-  }
-
-  // Performs the given mock function's default action and returns the
-  // result in a new-ed ActionResultHolder.
-  template <typename F>
-  static ActionResultHolder* PerformDefaultAction(
-      const FunctionMockerBase<F>* func_mocker,
-      typename RvalueRef<typename Function<F>::ArgumentTuple>::type args,
-      const std::string& call_description) {
-    return new ActionResultHolder(Wrapper(func_mocker->PerformDefaultAction(
-        internal::move(args), call_description)));
-  }
-
-  // Performs the given action and returns the result in a new-ed
-  // ActionResultHolder.
-  template <typename F>
-  static ActionResultHolder* PerformAction(
-      const Action<F>& action,
-      typename RvalueRef<typename Function<F>::ArgumentTuple>::type args) {
-    return new ActionResultHolder(
-        Wrapper(action.Perform(internal::move(args))));
-  }
-
- private:
-  typedef ReferenceOrValueWrapper<T> Wrapper;
-
-  explicit ActionResultHolder(Wrapper result)
-      : result_(::testing::internal::move(result)) {
-  }
-
-  Wrapper result_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ActionResultHolder);
-};
-
-// Specialization for T = void.
-template <>
-class ActionResultHolder<void> : public UntypedActionResultHolderBase {
- public:
-  void Unwrap() { }
-
-  virtual void PrintAsActionResult(::std::ostream* /* os */) const {}
-
-  // Performs the given mock function's default action and returns ownership
-  // of an empty ActionResultHolder*.
-  template <typename F>
-  static ActionResultHolder* PerformDefaultAction(
-      const FunctionMockerBase<F>* func_mocker,
-      typename RvalueRef<typename Function<F>::ArgumentTuple>::type args,
-      const std::string& call_description) {
-    func_mocker->PerformDefaultAction(internal::move(args), call_description);
-    return new ActionResultHolder;
-  }
-
-  // Performs the given action and returns ownership of an empty
-  // ActionResultHolder*.
-  template <typename F>
-  static ActionResultHolder* PerformAction(
-      const Action<F>& action,
-      typename RvalueRef<typename Function<F>::ArgumentTuple>::type args) {
-    action.Perform(internal::move(args));
-    return new ActionResultHolder;
-  }
-
- private:
-  ActionResultHolder() {}
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ActionResultHolder);
-};
-
-// The base of the function mocker class for the given function type.
-// We put the methods in this class instead of its child to avoid code
-// bloat.
-template <typename F>
-class FunctionMockerBase : public UntypedFunctionMockerBase {
- public:
-  typedef typename Function<F>::Result Result;
-  typedef typename Function<F>::ArgumentTuple ArgumentTuple;
-  typedef typename Function<F>::ArgumentMatcherTuple ArgumentMatcherTuple;
-
-  FunctionMockerBase() {}
-
-  // The destructor verifies that all expectations on this mock
-  // function have been satisfied.  If not, it will report Google Test
-  // non-fatal failures for the violations.
-  virtual ~FunctionMockerBase()
-        GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-    MutexLock l(&g_gmock_mutex);
-    VerifyAndClearExpectationsLocked();
-    Mock::UnregisterLocked(this);
-    ClearDefaultActionsLocked();
-  }
-
-  // Returns the ON_CALL spec that matches this mock function with the
-  // given arguments; returns NULL if no matching ON_CALL is found.
-  // L = *
-  const OnCallSpec<F>* FindOnCallSpec(
-      const ArgumentTuple& args) const {
-    for (UntypedOnCallSpecs::const_reverse_iterator it
-             = untyped_on_call_specs_.rbegin();
-         it != untyped_on_call_specs_.rend(); ++it) {
-      const OnCallSpec<F>* spec = static_cast<const OnCallSpec<F>*>(*it);
-      if (spec->Matches(args))
-        return spec;
-    }
-
-    return NULL;
-  }
-
-  // Performs the default action of this mock function on the given
-  // arguments and returns the result. Asserts (or throws if
-  // exceptions are enabled) with a helpful call descrption if there
-  // is no valid return value. This method doesn't depend on the
-  // mutable state of this object, and thus can be called concurrently
-  // without locking.
-  // L = *
-  Result PerformDefaultAction(
-      typename RvalueRef<typename Function<F>::ArgumentTuple>::type args,
-      const std::string& call_description) const {
-    const OnCallSpec<F>* const spec =
-        this->FindOnCallSpec(args);
-    if (spec != NULL) {
-      return spec->GetAction().Perform(internal::move(args));
-    }
-    const std::string message =
-        call_description +
-        "\n    The mock function has no default action "
-        "set, and its return type has no default value set.";
-#if GTEST_HAS_EXCEPTIONS
-    if (!DefaultValue<Result>::Exists()) {
-      throw std::runtime_error(message);
-    }
-#else
-    Assert(DefaultValue<Result>::Exists(), "", -1, message);
-#endif
-    return DefaultValue<Result>::Get();
-  }
-
-  // Performs the default action with the given arguments and returns
-  // the action's result.  The call description string will be used in
-  // the error message to describe the call in the case the default
-  // action fails.  The caller is responsible for deleting the result.
-  // L = *
-  virtual UntypedActionResultHolderBase* UntypedPerformDefaultAction(
-      void* untyped_args,  // must point to an ArgumentTuple
-      const std::string& call_description) const {
-    ArgumentTuple* args = static_cast<ArgumentTuple*>(untyped_args);
-    return ResultHolder::PerformDefaultAction(this, internal::move(*args),
-                                              call_description);
-  }
-
-  // Performs the given action with the given arguments and returns
-  // the action's result.  The caller is responsible for deleting the
-  // result.
-  // L = *
-  virtual UntypedActionResultHolderBase* UntypedPerformAction(
-      const void* untyped_action, void* untyped_args) const {
-    // Make a copy of the action before performing it, in case the
-    // action deletes the mock object (and thus deletes itself).
-    const Action<F> action = *static_cast<const Action<F>*>(untyped_action);
-    ArgumentTuple* args = static_cast<ArgumentTuple*>(untyped_args);
-    return ResultHolder::PerformAction(action, internal::move(*args));
-  }
-
-  // Implements UntypedFunctionMockerBase::ClearDefaultActionsLocked():
-  // clears the ON_CALL()s set on this mock function.
-  virtual void ClearDefaultActionsLocked()
-      GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-
-    // Deleting our default actions may trigger other mock objects to be
-    // deleted, for example if an action contains a reference counted smart
-    // pointer to that mock object, and that is the last reference. So if we
-    // delete our actions within the context of the global mutex we may deadlock
-    // when this method is called again. Instead, make a copy of the set of
-    // actions to delete, clear our set within the mutex, and then delete the
-    // actions outside of the mutex.
-    UntypedOnCallSpecs specs_to_delete;
-    untyped_on_call_specs_.swap(specs_to_delete);
-
-    g_gmock_mutex.Unlock();
-    for (UntypedOnCallSpecs::const_iterator it =
-             specs_to_delete.begin();
-         it != specs_to_delete.end(); ++it) {
-      delete static_cast<const OnCallSpec<F>*>(*it);
-    }
-
-    // Lock the mutex again, since the caller expects it to be locked when we
-    // return.
-    g_gmock_mutex.Lock();
-  }
-
- protected:
-  template <typename Function>
-  friend class MockSpec;
-
-  typedef ActionResultHolder<Result> ResultHolder;
-
-  // Returns the result of invoking this mock function with the given
-  // arguments.  This function can be safely called from multiple
-  // threads concurrently.
-  Result InvokeWith(
-      typename RvalueRef<typename Function<F>::ArgumentTuple>::type args)
-      GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-    // const_cast is required since in C++98 we still pass ArgumentTuple around
-    // by const& instead of rvalue reference.
-    void* untyped_args = const_cast<void*>(static_cast<const void*>(&args));
-    scoped_ptr<ResultHolder> holder(
-        DownCast_<ResultHolder*>(this->UntypedInvokeWith(untyped_args)));
-    return holder->Unwrap();
-  }
-
-  // Adds and returns a default action spec for this mock function.
-  OnCallSpec<F>& AddNewOnCallSpec(
-      const char* file, int line,
-      const ArgumentMatcherTuple& m)
-          GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-    Mock::RegisterUseByOnCallOrExpectCall(MockObject(), file, line);
-    OnCallSpec<F>* const on_call_spec = new OnCallSpec<F>(file, line, m);
-    untyped_on_call_specs_.push_back(on_call_spec);
-    return *on_call_spec;
-  }
-
-  // Adds and returns an expectation spec for this mock function.
-  TypedExpectation<F>& AddNewExpectation(const char* file, int line,
-                                         const std::string& source_text,
-                                         const ArgumentMatcherTuple& m)
-      GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-    Mock::RegisterUseByOnCallOrExpectCall(MockObject(), file, line);
-    TypedExpectation<F>* const expectation =
-        new TypedExpectation<F>(this, file, line, source_text, m);
-    const linked_ptr<ExpectationBase> untyped_expectation(expectation);
-    // See the definition of untyped_expectations_ for why access to
-    // it is unprotected here.
-    untyped_expectations_.push_back(untyped_expectation);
-
-    // Adds this expectation into the implicit sequence if there is one.
-    Sequence* const implicit_sequence = g_gmock_implicit_sequence.get();
-    if (implicit_sequence != NULL) {
-      implicit_sequence->AddExpectation(Expectation(untyped_expectation));
-    }
-
-    return *expectation;
-  }
-
- private:
-  template <typename Func> friend class TypedExpectation;
-
-  // Some utilities needed for implementing UntypedInvokeWith().
-
-  // Describes what default action will be performed for the given
-  // arguments.
-  // L = *
-  void DescribeDefaultActionTo(const ArgumentTuple& args,
-                               ::std::ostream* os) const {
-    const OnCallSpec<F>* const spec = FindOnCallSpec(args);
-
-    if (spec == NULL) {
-      *os << (internal::type_equals<Result, void>::value ?
-              "returning directly.\n" :
-              "returning default value.\n");
-    } else {
-      *os << "taking default action specified at:\n"
-          << FormatFileLocation(spec->file(), spec->line()) << "\n";
-    }
-  }
-
-  // Writes a message that the call is uninteresting (i.e. neither
-  // explicitly expected nor explicitly unexpected) to the given
-  // ostream.
-  virtual void UntypedDescribeUninterestingCall(
-      const void* untyped_args,
-      ::std::ostream* os) const
-          GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-    const ArgumentTuple& args =
-        *static_cast<const ArgumentTuple*>(untyped_args);
-    *os << "Uninteresting mock function call - ";
-    DescribeDefaultActionTo(args, os);
-    *os << "    Function call: " << Name();
-    UniversalPrint(args, os);
-  }
-
-  // Returns the expectation that matches the given function arguments
-  // (or NULL is there's no match); when a match is found,
-  // untyped_action is set to point to the action that should be
-  // performed (or NULL if the action is "do default"), and
-  // is_excessive is modified to indicate whether the call exceeds the
-  // expected number.
-  //
-  // Critical section: We must find the matching expectation and the
-  // corresponding action that needs to be taken in an ATOMIC
-  // transaction.  Otherwise another thread may call this mock
-  // method in the middle and mess up the state.
-  //
-  // However, performing the action has to be left out of the critical
-  // section.  The reason is that we have no control on what the
-  // action does (it can invoke an arbitrary user function or even a
-  // mock function) and excessive locking could cause a dead lock.
-  virtual const ExpectationBase* UntypedFindMatchingExpectation(
-      const void* untyped_args,
-      const void** untyped_action, bool* is_excessive,
-      ::std::ostream* what, ::std::ostream* why)
-          GTEST_LOCK_EXCLUDED_(g_gmock_mutex) {
-    const ArgumentTuple& args =
-        *static_cast<const ArgumentTuple*>(untyped_args);
-    MutexLock l(&g_gmock_mutex);
-    TypedExpectation<F>* exp = this->FindMatchingExpectationLocked(args);
-    if (exp == NULL) {  // A match wasn't found.
-      this->FormatUnexpectedCallMessageLocked(args, what, why);
-      return NULL;
-    }
-
-    // This line must be done before calling GetActionForArguments(),
-    // which will increment the call count for *exp and thus affect
-    // its saturation status.
-    *is_excessive = exp->IsSaturated();
-    const Action<F>* action = exp->GetActionForArguments(this, args, what, why);
-    if (action != NULL && action->IsDoDefault())
-      action = NULL;  // Normalize "do default" to NULL.
-    *untyped_action = action;
-    return exp;
-  }
-
-  // Prints the given function arguments to the ostream.
-  virtual void UntypedPrintArgs(const void* untyped_args,
-                                ::std::ostream* os) const {
-    const ArgumentTuple& args =
-        *static_cast<const ArgumentTuple*>(untyped_args);
-    UniversalPrint(args, os);
-  }
-
-  // Returns the expectation that matches the arguments, or NULL if no
-  // expectation matches them.
-  TypedExpectation<F>* FindMatchingExpectationLocked(
-      const ArgumentTuple& args) const
-          GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    // See the definition of untyped_expectations_ for why access to
-    // it is unprotected here.
-    for (typename UntypedExpectations::const_reverse_iterator it =
-             untyped_expectations_.rbegin();
-         it != untyped_expectations_.rend(); ++it) {
-      TypedExpectation<F>* const exp =
-          static_cast<TypedExpectation<F>*>(it->get());
-      if (exp->ShouldHandleArguments(args)) {
-        return exp;
-      }
-    }
-    return NULL;
-  }
-
-  // Returns a message that the arguments don't match any expectation.
-  void FormatUnexpectedCallMessageLocked(
-      const ArgumentTuple& args,
-      ::std::ostream* os,
-      ::std::ostream* why) const
-          GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    *os << "\nUnexpected mock function call - ";
-    DescribeDefaultActionTo(args, os);
-    PrintTriedExpectationsLocked(args, why);
-  }
-
-  // Prints a list of expectations that have been tried against the
-  // current mock function call.
-  void PrintTriedExpectationsLocked(
-      const ArgumentTuple& args,
-      ::std::ostream* why) const
-          GTEST_EXCLUSIVE_LOCK_REQUIRED_(g_gmock_mutex) {
-    g_gmock_mutex.AssertHeld();
-    const int count = static_cast<int>(untyped_expectations_.size());
-    *why << "Google Mock tried the following " << count << " "
-         << (count == 1 ? "expectation, but it didn't match" :
-             "expectations, but none matched")
-         << ":\n";
-    for (int i = 0; i < count; i++) {
-      TypedExpectation<F>* const expectation =
-          static_cast<TypedExpectation<F>*>(untyped_expectations_[i].get());
-      *why << "\n";
-      expectation->DescribeLocationTo(why);
-      if (count > 1) {
-        *why << "tried expectation #" << i << ": ";
-      }
-      *why << expectation->source_text() << "...\n";
-      expectation->ExplainMatchResultTo(args, why);
-      expectation->DescribeCallCountTo(why);
-    }
-  }
-
-  // There is no generally useful and implementable semantics of
-  // copying a mock object, so copying a mock is usually a user error.
-  // Thus we disallow copying function mockers.  If the user really
-  // wants to copy a mock object, they should implement their own copy
-  // operation, for example:
-  //
-  //   class MockFoo : public Foo {
-  //    public:
-  //     // Defines a copy constructor explicitly.
-  //     MockFoo(const MockFoo& src) {}
-  //     ...
-  //   };
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(FunctionMockerBase);
-};  // class FunctionMockerBase
-
-#ifdef _MSC_VER
-# pragma warning(pop)  // Restores the warning state.
-#endif  // _MSV_VER
-
-// Implements methods of FunctionMockerBase.
-
-// Verifies that all expectations on this mock function have been
-// satisfied.  Reports one or more Google Test non-fatal failures and
-// returns false if not.
-
-// Reports an uninteresting call (whose description is in msg) in the
-// manner specified by 'reaction'.
-void ReportUninterestingCall(CallReaction reaction, const std::string& msg);
-
-}  // namespace internal
-
-// The style guide prohibits "using" statements in a namespace scope
-// inside a header file.  However, the MockSpec class template is
-// meant to be defined in the ::testing namespace.  The following line
-// is just a trick for working around a bug in MSVC 8.0, which cannot
-// handle it if we define MockSpec in ::testing.
-using internal::MockSpec;
-
-// Const(x) is a convenient function for obtaining a const reference
-// to x.  This is useful for setting expectations on an overloaded
-// const mock method, e.g.
-//
-//   class MockFoo : public FooInterface {
-//    public:
-//     MOCK_METHOD0(Bar, int());
-//     MOCK_CONST_METHOD0(Bar, int&());
-//   };
-//
-//   MockFoo foo;
-//   // Expects a call to non-const MockFoo::Bar().
-//   EXPECT_CALL(foo, Bar());
-//   // Expects a call to const MockFoo::Bar().
-//   EXPECT_CALL(Const(foo), Bar());
-template <typename T>
-inline const T& Const(const T& x) { return x; }
-
-// Constructs an Expectation object that references and co-owns exp.
-inline Expectation::Expectation(internal::ExpectationBase& exp)  // NOLINT
-    : expectation_base_(exp.GetHandle().expectation_base()) {}
-
-}  // namespace testing
-
-// Implementation for ON_CALL and EXPECT_CALL macros. A separate macro is
-// required to avoid compile errors when the name of the method used in call is
-// a result of macro expansion. See CompilesWithMethodNameExpandedFromMacro
-// tests in internal/gmock-spec-builders_test.cc for more details.
-//
-// This macro supports statements both with and without parameter matchers. If
-// the parameter list is omitted, gMock will accept any parameters, which allows
-// tests to be written that don't need to encode the number of method
-// parameter. This technique may only be used for non-overloaded methods.
-//
-//   // These are the same:
-//   ON_CALL(mock, NoArgsMethod()).WillByDefault(...);
-//   ON_CALL(mock, NoArgsMethod).WillByDefault(...);
-//
-//   // As are these:
-//   ON_CALL(mock, TwoArgsMethod(_, _)).WillByDefault(...);
-//   ON_CALL(mock, TwoArgsMethod).WillByDefault(...);
-//
-//   // Can also specify args if you want, of course:
-//   ON_CALL(mock, TwoArgsMethod(_, 45)).WillByDefault(...);
-//
-//   // Overloads work as long as you specify parameters:
-//   ON_CALL(mock, OverloadedMethod(_)).WillByDefault(...);
-//   ON_CALL(mock, OverloadedMethod(_, _)).WillByDefault(...);
-//
-//   // Oops! Which overload did you want?
-//   ON_CALL(mock, OverloadedMethod).WillByDefault(...);
-//     => ERROR: call to member function 'gmock_OverloadedMethod' is ambiguous
-//
-// How this works: The mock class uses two overloads of the gmock_Method
-// expectation setter method plus an operator() overload on the MockSpec object.
-// In the matcher list form, the macro expands to:
-//
-//   // This statement:
-//   ON_CALL(mock, TwoArgsMethod(_, 45))...
-//
-//   // ...expands to:
-//   mock.gmock_TwoArgsMethod(_, 45)(WithoutMatchers(), nullptr)...
-//   |-------------v---------------||------------v-------------|
-//       invokes first overload        swallowed by operator()
-//
-//   // ...which is essentially:
-//   mock.gmock_TwoArgsMethod(_, 45)...
-//
-// Whereas the form without a matcher list:
-//
-//   // This statement:
-//   ON_CALL(mock, TwoArgsMethod)...
-//
-//   // ...expands to:
-//   mock.gmock_TwoArgsMethod(WithoutMatchers(), nullptr)...
-//   |-----------------------v--------------------------|
-//                 invokes second overload
-//
-//   // ...which is essentially:
-//   mock.gmock_TwoArgsMethod(_, _)...
-//
-// The WithoutMatchers() argument is used to disambiguate overloads and to
-// block the caller from accidentally invoking the second overload directly. The
-// second argument is an internal type derived from the method signature. The
-// failure to disambiguate two overloads of this method in the ON_CALL statement
-// is how we block callers from setting expectations on overloaded methods.
-#define GMOCK_ON_CALL_IMPL_(mock_expr, Setter, call)                          \
-  ((mock_expr).gmock_##call)(::testing::internal::GetWithoutMatchers(), NULL) \
-      .Setter(__FILE__, __LINE__, #mock_expr, #call)
-
-#define ON_CALL(obj, call) \
-  GMOCK_ON_CALL_IMPL_(obj, InternalDefaultActionSetAt, call)
-
-#define EXPECT_CALL(obj, call) \
-  GMOCK_ON_CALL_IMPL_(obj, InternalExpectedAt, call)
-
-#endif  // GMOCK_INCLUDE_GMOCK_GMOCK_SPEC_BUILDERS_H_
-
-#if GTEST_HAS_STD_FUNCTION_
-# include <functional>
-#endif
-
-namespace testing {
-namespace internal {
-
-template <typename F>
-class FunctionMockerBase;
-
-// Note: class FunctionMocker really belongs to the ::testing
-// namespace.  However if we define it in ::testing, MSVC will
-// complain when classes in ::testing::internal declare it as a
-// friend class template.  To workaround this compiler bug, we define
-// FunctionMocker in ::testing::internal and import it into ::testing.
-template <typename F>
-class FunctionMocker;
-
-template <typename R>
-class FunctionMocker<R()> : public
-    internal::FunctionMockerBase<R()> {
- public:
-  typedef R F();
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With() {
-    return MockSpec<F>(this, ::testing::make_tuple());
-  }
-
-  R Invoke() {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple());
-  }
-};
-
-template <typename R, typename A1>
-class FunctionMocker<R(A1)> : public
-    internal::FunctionMockerBase<R(A1)> {
- public:
-  typedef R F(A1);
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With(const Matcher<A1>& m1) {
-    return MockSpec<F>(this, ::testing::make_tuple(m1));
-  }
-
-  R Invoke(A1 a1) {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple(internal::forward<A1>(a1)));
-  }
-};
-
-template <typename R, typename A1, typename A2>
-class FunctionMocker<R(A1, A2)> : public
-    internal::FunctionMockerBase<R(A1, A2)> {
- public:
-  typedef R F(A1, A2);
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With(const Matcher<A1>& m1, const Matcher<A2>& m2) {
-    return MockSpec<F>(this, ::testing::make_tuple(m1, m2));
-  }
-
-  R Invoke(A1 a1, A2 a2) {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple(internal::forward<A1>(a1),
-        internal::forward<A2>(a2)));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3>
-class FunctionMocker<R(A1, A2, A3)> : public
-    internal::FunctionMockerBase<R(A1, A2, A3)> {
- public:
-  typedef R F(A1, A2, A3);
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With(const Matcher<A1>& m1, const Matcher<A2>& m2,
-      const Matcher<A3>& m3) {
-    return MockSpec<F>(this, ::testing::make_tuple(m1, m2, m3));
-  }
-
-  R Invoke(A1 a1, A2 a2, A3 a3) {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple(internal::forward<A1>(a1),
-        internal::forward<A2>(a2), internal::forward<A3>(a3)));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4>
-class FunctionMocker<R(A1, A2, A3, A4)> : public
-    internal::FunctionMockerBase<R(A1, A2, A3, A4)> {
- public:
-  typedef R F(A1, A2, A3, A4);
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With(const Matcher<A1>& m1, const Matcher<A2>& m2,
-      const Matcher<A3>& m3, const Matcher<A4>& m4) {
-    return MockSpec<F>(this, ::testing::make_tuple(m1, m2, m3, m4));
-  }
-
-  R Invoke(A1 a1, A2 a2, A3 a3, A4 a4) {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple(internal::forward<A1>(a1),
-        internal::forward<A2>(a2), internal::forward<A3>(a3),
-        internal::forward<A4>(a4)));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5>
-class FunctionMocker<R(A1, A2, A3, A4, A5)> : public
-    internal::FunctionMockerBase<R(A1, A2, A3, A4, A5)> {
- public:
-  typedef R F(A1, A2, A3, A4, A5);
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With(const Matcher<A1>& m1, const Matcher<A2>& m2,
-      const Matcher<A3>& m3, const Matcher<A4>& m4, const Matcher<A5>& m5) {
-    return MockSpec<F>(this, ::testing::make_tuple(m1, m2, m3, m4, m5));
-  }
-
-  R Invoke(A1 a1, A2 a2, A3 a3, A4 a4, A5 a5) {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple(internal::forward<A1>(a1),
-        internal::forward<A2>(a2), internal::forward<A3>(a3),
-        internal::forward<A4>(a4), internal::forward<A5>(a5)));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6>
-class FunctionMocker<R(A1, A2, A3, A4, A5, A6)> : public
-    internal::FunctionMockerBase<R(A1, A2, A3, A4, A5, A6)> {
- public:
-  typedef R F(A1, A2, A3, A4, A5, A6);
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With(const Matcher<A1>& m1, const Matcher<A2>& m2,
-      const Matcher<A3>& m3, const Matcher<A4>& m4, const Matcher<A5>& m5,
-      const Matcher<A6>& m6) {
-    return MockSpec<F>(this, ::testing::make_tuple(m1, m2, m3, m4, m5, m6));
-  }
-
-  R Invoke(A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6) {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple(internal::forward<A1>(a1),
-        internal::forward<A2>(a2), internal::forward<A3>(a3),
-        internal::forward<A4>(a4), internal::forward<A5>(a5),
-        internal::forward<A6>(a6)));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7>
-class FunctionMocker<R(A1, A2, A3, A4, A5, A6, A7)> : public
-    internal::FunctionMockerBase<R(A1, A2, A3, A4, A5, A6, A7)> {
- public:
-  typedef R F(A1, A2, A3, A4, A5, A6, A7);
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With(const Matcher<A1>& m1, const Matcher<A2>& m2,
-      const Matcher<A3>& m3, const Matcher<A4>& m4, const Matcher<A5>& m5,
-      const Matcher<A6>& m6, const Matcher<A7>& m7) {
-    return MockSpec<F>(this, ::testing::make_tuple(m1, m2, m3, m4, m5, m6, m7));
-  }
-
-  R Invoke(A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6, A7 a7) {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple(internal::forward<A1>(a1),
-        internal::forward<A2>(a2), internal::forward<A3>(a3),
-        internal::forward<A4>(a4), internal::forward<A5>(a5),
-        internal::forward<A6>(a6), internal::forward<A7>(a7)));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7, typename A8>
-class FunctionMocker<R(A1, A2, A3, A4, A5, A6, A7, A8)> : public
-    internal::FunctionMockerBase<R(A1, A2, A3, A4, A5, A6, A7, A8)> {
- public:
-  typedef R F(A1, A2, A3, A4, A5, A6, A7, A8);
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With(const Matcher<A1>& m1, const Matcher<A2>& m2,
-      const Matcher<A3>& m3, const Matcher<A4>& m4, const Matcher<A5>& m5,
-      const Matcher<A6>& m6, const Matcher<A7>& m7, const Matcher<A8>& m8) {
-    return MockSpec<F>(this, ::testing::make_tuple(m1, m2, m3, m4, m5, m6, m7,
-        m8));
-  }
-
-  R Invoke(A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6, A7 a7, A8 a8) {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple(internal::forward<A1>(a1),
-        internal::forward<A2>(a2), internal::forward<A3>(a3),
-        internal::forward<A4>(a4), internal::forward<A5>(a5),
-        internal::forward<A6>(a6), internal::forward<A7>(a7),
-        internal::forward<A8>(a8)));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7, typename A8, typename A9>
-class FunctionMocker<R(A1, A2, A3, A4, A5, A6, A7, A8, A9)> : public
-    internal::FunctionMockerBase<R(A1, A2, A3, A4, A5, A6, A7, A8, A9)> {
- public:
-  typedef R F(A1, A2, A3, A4, A5, A6, A7, A8, A9);
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With(const Matcher<A1>& m1, const Matcher<A2>& m2,
-      const Matcher<A3>& m3, const Matcher<A4>& m4, const Matcher<A5>& m5,
-      const Matcher<A6>& m6, const Matcher<A7>& m7, const Matcher<A8>& m8,
-      const Matcher<A9>& m9) {
-    return MockSpec<F>(this, ::testing::make_tuple(m1, m2, m3, m4, m5, m6, m7,
-        m8, m9));
-  }
-
-  R Invoke(A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6, A7 a7, A8 a8, A9 a9) {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple(internal::forward<A1>(a1),
-        internal::forward<A2>(a2), internal::forward<A3>(a3),
-        internal::forward<A4>(a4), internal::forward<A5>(a5),
-        internal::forward<A6>(a6), internal::forward<A7>(a7),
-        internal::forward<A8>(a8), internal::forward<A9>(a9)));
-  }
-};
-
-template <typename R, typename A1, typename A2, typename A3, typename A4,
-    typename A5, typename A6, typename A7, typename A8, typename A9,
-    typename A10>
-class FunctionMocker<R(A1, A2, A3, A4, A5, A6, A7, A8, A9, A10)> : public
-    internal::FunctionMockerBase<R(A1, A2, A3, A4, A5, A6, A7, A8, A9, A10)> {
- public:
-  typedef R F(A1, A2, A3, A4, A5, A6, A7, A8, A9, A10);
-  typedef typename internal::Function<F>::ArgumentTuple ArgumentTuple;
-
-  MockSpec<F> With(const Matcher<A1>& m1, const Matcher<A2>& m2,
-      const Matcher<A3>& m3, const Matcher<A4>& m4, const Matcher<A5>& m5,
-      const Matcher<A6>& m6, const Matcher<A7>& m7, const Matcher<A8>& m8,
-      const Matcher<A9>& m9, const Matcher<A10>& m10) {
-    return MockSpec<F>(this, ::testing::make_tuple(m1, m2, m3, m4, m5, m6, m7,
-        m8, m9, m10));
-  }
-
-  R Invoke(A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6, A7 a7, A8 a8, A9 a9,
-      A10 a10) {
-    // Even though gcc and MSVC don't enforce it, 'this->' is required
-    // by the C++ standard [14.6.4] here, as the base class type is
-    // dependent on the template argument (and thus shouldn't be
-    // looked into when resolving InvokeWith).
-    return this->InvokeWith(ArgumentTuple(internal::forward<A1>(a1),
-        internal::forward<A2>(a2), internal::forward<A3>(a3),
-        internal::forward<A4>(a4), internal::forward<A5>(a5),
-        internal::forward<A6>(a6), internal::forward<A7>(a7),
-        internal::forward<A8>(a8), internal::forward<A9>(a9),
-        internal::forward<A10>(a10)));
-  }
-};
-
-// Removes the given pointer; this is a helper for the expectation setter method
-// for parameterless matchers.
-//
-// We want to make sure that the user cannot set a parameterless expectation on
-// overloaded methods, including methods which are overloaded on const. Example:
-//
-//   class MockClass {
-//     MOCK_METHOD0(GetName, string&());
-//     MOCK_CONST_METHOD0(GetName, const string&());
-//   };
-//
-//   TEST() {
-//     // This should be an error, as it's not clear which overload is expected.
-//     EXPECT_CALL(mock, GetName).WillOnce(ReturnRef(value));
-//   }
-//
-// Here are the generated expectation-setter methods:
-//
-//   class MockClass {
-//     // Overload 1
-//     MockSpec<string&()> gmock_GetName() { ... }
-//     // Overload 2. Declared const so that the compiler will generate an
-//     // error when trying to resolve between this and overload 4 in
-//     // 'gmock_GetName(WithoutMatchers(), nullptr)'.
-//     MockSpec<string&()> gmock_GetName(
-//         const WithoutMatchers&, const Function<string&()>*) const {
-//       // Removes const from this, calls overload 1
-//       return AdjustConstness_(this)->gmock_GetName();
-//     }
-//
-//     // Overload 3
-//     const string& gmock_GetName() const { ... }
-//     // Overload 4
-//     MockSpec<const string&()> gmock_GetName(
-//         const WithoutMatchers&, const Function<const string&()>*) const {
-//       // Does not remove const, calls overload 3
-//       return AdjustConstness_const(this)->gmock_GetName();
-//     }
-//   }
-//
-template <typename MockType>
-const MockType* AdjustConstness_const(const MockType* mock) {
-  return mock;
-}
-
-// Removes const from and returns the given pointer; this is a helper for the
-// expectation setter method for parameterless matchers.
-template <typename MockType>
-MockType* AdjustConstness_(const MockType* mock) {
-  return const_cast<MockType*>(mock);
-}
-
-}  // namespace internal
-
-// The style guide prohibits "using" statements in a namespace scope
-// inside a header file.  However, the FunctionMocker class template
-// is meant to be defined in the ::testing namespace.  The following
-// line is just a trick for working around a bug in MSVC 8.0, which
-// cannot handle it if we define FunctionMocker in ::testing.
-using internal::FunctionMocker;
-
-// GMOCK_RESULT_(tn, F) expands to the result type of function type F.
-// We define this as a variadic macro in case F contains unprotected
-// commas (the same reason that we use variadic macros in other places
-// in this file).
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_RESULT_(tn, ...) \
-    tn ::testing::internal::Function<__VA_ARGS__>::Result
-
-// The type of argument N of the given function type.
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_ARG_(tn, N, ...) \
-    tn ::testing::internal::Function<__VA_ARGS__>::Argument##N
-
-// The matcher type for argument N of the given function type.
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_MATCHER_(tn, N, ...) \
-    const ::testing::Matcher<GMOCK_ARG_(tn, N, __VA_ARGS__)>&
-
-// The variable for mocking the given method.
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_MOCKER_(arity, constness, Method) \
-    GTEST_CONCAT_TOKEN_(gmock##constness##arity##_##Method##_, __LINE__)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD0_(tn, constness, ct, Method, ...)                       \
-  GMOCK_RESULT_(tn, __VA_ARGS__) ct Method() constness {                     \
-    GTEST_COMPILE_ASSERT_(                                                   \
-        (::testing::tuple_size<tn ::testing::internal::Function<             \
-             __VA_ARGS__>::ArgumentTuple>::value == 0),                      \
-        this_method_does_not_take_0_arguments);                              \
-    GMOCK_MOCKER_(0, constness, Method).SetOwnerAndName(this, #Method);      \
-    return GMOCK_MOCKER_(0, constness, Method).Invoke();                     \
-  }                                                                          \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method() constness {              \
-    GMOCK_MOCKER_(0, constness, Method).RegisterOwner(this);                 \
-    return GMOCK_MOCKER_(0, constness, Method).With();                       \
-  }                                                                          \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                           \
-      const ::testing::internal::WithoutMatchers&,                           \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {         \
-    return ::testing::internal::AdjustConstness_##constness(this)            \
-        ->gmock_##Method();                                                  \
-  }                                                                          \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(0, constness, \
-                                                               Method)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD1_(tn, constness, ct, Method, ...)                        \
-  GMOCK_RESULT_(tn, __VA_ARGS__)                                              \
-  ct Method(GMOCK_ARG_(tn, 1, __VA_ARGS__) gmock_a1) constness {              \
-    GTEST_COMPILE_ASSERT_(                                                    \
-        (::testing::tuple_size<tn ::testing::internal::Function<              \
-             __VA_ARGS__>::ArgumentTuple>::value == 1),                       \
-        this_method_does_not_take_1_argument);                                \
-    GMOCK_MOCKER_(1, constness, Method).SetOwnerAndName(this, #Method);       \
-    return GMOCK_MOCKER_(1, constness, Method)                                \
-        .Invoke(::testing::internal::forward<GMOCK_ARG_(tn, 1, __VA_ARGS__)>( \
-            gmock_a1));                                                       \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      GMOCK_MATCHER_(tn, 1, __VA_ARGS__) gmock_a1) constness {                \
-    GMOCK_MOCKER_(1, constness, Method).RegisterOwner(this);                  \
-    return GMOCK_MOCKER_(1, constness, Method).With(gmock_a1);                \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      const ::testing::internal::WithoutMatchers&,                            \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {          \
-    return ::testing::internal::AdjustConstness_##constness(this)             \
-        ->gmock_##Method(::testing::A<GMOCK_ARG_(tn, 1, __VA_ARGS__)>());     \
-  }                                                                           \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(1, constness,  \
-                                                               Method)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD2_(tn, constness, ct, Method, ...)                        \
-  GMOCK_RESULT_(tn, __VA_ARGS__)                                              \
-  ct Method(GMOCK_ARG_(tn, 1, __VA_ARGS__) gmock_a1,                          \
-            GMOCK_ARG_(tn, 2, __VA_ARGS__) gmock_a2) constness {              \
-    GTEST_COMPILE_ASSERT_(                                                    \
-        (::testing::tuple_size<tn ::testing::internal::Function<              \
-             __VA_ARGS__>::ArgumentTuple>::value == 2),                       \
-        this_method_does_not_take_2_arguments);                               \
-    GMOCK_MOCKER_(2, constness, Method).SetOwnerAndName(this, #Method);       \
-    return GMOCK_MOCKER_(2, constness, Method)                                \
-        .Invoke(::testing::internal::forward<GMOCK_ARG_(tn, 1, __VA_ARGS__)>( \
-                    gmock_a1),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 2, __VA_ARGS__)>( \
-                    gmock_a2));                                               \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      GMOCK_MATCHER_(tn, 1, __VA_ARGS__) gmock_a1,                            \
-      GMOCK_MATCHER_(tn, 2, __VA_ARGS__) gmock_a2) constness {                \
-    GMOCK_MOCKER_(2, constness, Method).RegisterOwner(this);                  \
-    return GMOCK_MOCKER_(2, constness, Method).With(gmock_a1, gmock_a2);      \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      const ::testing::internal::WithoutMatchers&,                            \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {          \
-    return ::testing::internal::AdjustConstness_##constness(this)             \
-        ->gmock_##Method(::testing::A<GMOCK_ARG_(tn, 1, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 2, __VA_ARGS__)>());     \
-  }                                                                           \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(2, constness,  \
-                                                               Method)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD3_(tn, constness, ct, Method, ...)                        \
-  GMOCK_RESULT_(tn, __VA_ARGS__)                                              \
-  ct Method(GMOCK_ARG_(tn, 1, __VA_ARGS__) gmock_a1,                          \
-            GMOCK_ARG_(tn, 2, __VA_ARGS__) gmock_a2,                          \
-            GMOCK_ARG_(tn, 3, __VA_ARGS__) gmock_a3) constness {              \
-    GTEST_COMPILE_ASSERT_(                                                    \
-        (::testing::tuple_size<tn ::testing::internal::Function<              \
-             __VA_ARGS__>::ArgumentTuple>::value == 3),                       \
-        this_method_does_not_take_3_arguments);                               \
-    GMOCK_MOCKER_(3, constness, Method).SetOwnerAndName(this, #Method);       \
-    return GMOCK_MOCKER_(3, constness, Method)                                \
-        .Invoke(::testing::internal::forward<GMOCK_ARG_(tn, 1, __VA_ARGS__)>( \
-                    gmock_a1),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 2, __VA_ARGS__)>( \
-                    gmock_a2),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 3, __VA_ARGS__)>( \
-                    gmock_a3));                                               \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      GMOCK_MATCHER_(tn, 1, __VA_ARGS__) gmock_a1,                            \
-      GMOCK_MATCHER_(tn, 2, __VA_ARGS__) gmock_a2,                            \
-      GMOCK_MATCHER_(tn, 3, __VA_ARGS__) gmock_a3) constness {                \
-    GMOCK_MOCKER_(3, constness, Method).RegisterOwner(this);                  \
-    return GMOCK_MOCKER_(3, constness, Method)                                \
-        .With(gmock_a1, gmock_a2, gmock_a3);                                  \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      const ::testing::internal::WithoutMatchers&,                            \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {          \
-    return ::testing::internal::AdjustConstness_##constness(this)             \
-        ->gmock_##Method(::testing::A<GMOCK_ARG_(tn, 1, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 2, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 3, __VA_ARGS__)>());     \
-  }                                                                           \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(3, constness,  \
-                                                               Method)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD4_(tn, constness, ct, Method, ...)                        \
-  GMOCK_RESULT_(tn, __VA_ARGS__)                                              \
-  ct Method(GMOCK_ARG_(tn, 1, __VA_ARGS__) gmock_a1,                          \
-            GMOCK_ARG_(tn, 2, __VA_ARGS__) gmock_a2,                          \
-            GMOCK_ARG_(tn, 3, __VA_ARGS__) gmock_a3,                          \
-            GMOCK_ARG_(tn, 4, __VA_ARGS__) gmock_a4) constness {              \
-    GTEST_COMPILE_ASSERT_(                                                    \
-        (::testing::tuple_size<tn ::testing::internal::Function<              \
-             __VA_ARGS__>::ArgumentTuple>::value == 4),                       \
-        this_method_does_not_take_4_arguments);                               \
-    GMOCK_MOCKER_(4, constness, Method).SetOwnerAndName(this, #Method);       \
-    return GMOCK_MOCKER_(4, constness, Method)                                \
-        .Invoke(::testing::internal::forward<GMOCK_ARG_(tn, 1, __VA_ARGS__)>( \
-                    gmock_a1),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 2, __VA_ARGS__)>( \
-                    gmock_a2),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 3, __VA_ARGS__)>( \
-                    gmock_a3),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 4, __VA_ARGS__)>( \
-                    gmock_a4));                                               \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      GMOCK_MATCHER_(tn, 1, __VA_ARGS__) gmock_a1,                            \
-      GMOCK_MATCHER_(tn, 2, __VA_ARGS__) gmock_a2,                            \
-      GMOCK_MATCHER_(tn, 3, __VA_ARGS__) gmock_a3,                            \
-      GMOCK_MATCHER_(tn, 4, __VA_ARGS__) gmock_a4) constness {                \
-    GMOCK_MOCKER_(4, constness, Method).RegisterOwner(this);                  \
-    return GMOCK_MOCKER_(4, constness, Method)                                \
-        .With(gmock_a1, gmock_a2, gmock_a3, gmock_a4);                        \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      const ::testing::internal::WithoutMatchers&,                            \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {          \
-    return ::testing::internal::AdjustConstness_##constness(this)             \
-        ->gmock_##Method(::testing::A<GMOCK_ARG_(tn, 1, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 2, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 3, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 4, __VA_ARGS__)>());     \
-  }                                                                           \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(4, constness,  \
-                                                               Method)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD5_(tn, constness, ct, Method, ...)                        \
-  GMOCK_RESULT_(tn, __VA_ARGS__)                                              \
-  ct Method(GMOCK_ARG_(tn, 1, __VA_ARGS__) gmock_a1,                          \
-            GMOCK_ARG_(tn, 2, __VA_ARGS__) gmock_a2,                          \
-            GMOCK_ARG_(tn, 3, __VA_ARGS__) gmock_a3,                          \
-            GMOCK_ARG_(tn, 4, __VA_ARGS__) gmock_a4,                          \
-            GMOCK_ARG_(tn, 5, __VA_ARGS__) gmock_a5) constness {              \
-    GTEST_COMPILE_ASSERT_(                                                    \
-        (::testing::tuple_size<tn ::testing::internal::Function<              \
-             __VA_ARGS__>::ArgumentTuple>::value == 5),                       \
-        this_method_does_not_take_5_arguments);                               \
-    GMOCK_MOCKER_(5, constness, Method).SetOwnerAndName(this, #Method);       \
-    return GMOCK_MOCKER_(5, constness, Method)                                \
-        .Invoke(::testing::internal::forward<GMOCK_ARG_(tn, 1, __VA_ARGS__)>( \
-                    gmock_a1),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 2, __VA_ARGS__)>( \
-                    gmock_a2),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 3, __VA_ARGS__)>( \
-                    gmock_a3),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 4, __VA_ARGS__)>( \
-                    gmock_a4),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 5, __VA_ARGS__)>( \
-                    gmock_a5));                                               \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      GMOCK_MATCHER_(tn, 1, __VA_ARGS__) gmock_a1,                            \
-      GMOCK_MATCHER_(tn, 2, __VA_ARGS__) gmock_a2,                            \
-      GMOCK_MATCHER_(tn, 3, __VA_ARGS__) gmock_a3,                            \
-      GMOCK_MATCHER_(tn, 4, __VA_ARGS__) gmock_a4,                            \
-      GMOCK_MATCHER_(tn, 5, __VA_ARGS__) gmock_a5) constness {                \
-    GMOCK_MOCKER_(5, constness, Method).RegisterOwner(this);                  \
-    return GMOCK_MOCKER_(5, constness, Method)                                \
-        .With(gmock_a1, gmock_a2, gmock_a3, gmock_a4, gmock_a5);              \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      const ::testing::internal::WithoutMatchers&,                            \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {          \
-    return ::testing::internal::AdjustConstness_##constness(this)             \
-        ->gmock_##Method(::testing::A<GMOCK_ARG_(tn, 1, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 2, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 3, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 4, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 5, __VA_ARGS__)>());     \
-  }                                                                           \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(5, constness,  \
-                                                               Method)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD6_(tn, constness, ct, Method, ...)                        \
-  GMOCK_RESULT_(tn, __VA_ARGS__)                                              \
-  ct Method(GMOCK_ARG_(tn, 1, __VA_ARGS__) gmock_a1,                          \
-            GMOCK_ARG_(tn, 2, __VA_ARGS__) gmock_a2,                          \
-            GMOCK_ARG_(tn, 3, __VA_ARGS__) gmock_a3,                          \
-            GMOCK_ARG_(tn, 4, __VA_ARGS__) gmock_a4,                          \
-            GMOCK_ARG_(tn, 5, __VA_ARGS__) gmock_a5,                          \
-            GMOCK_ARG_(tn, 6, __VA_ARGS__) gmock_a6) constness {              \
-    GTEST_COMPILE_ASSERT_(                                                    \
-        (::testing::tuple_size<tn ::testing::internal::Function<              \
-             __VA_ARGS__>::ArgumentTuple>::value == 6),                       \
-        this_method_does_not_take_6_arguments);                               \
-    GMOCK_MOCKER_(6, constness, Method).SetOwnerAndName(this, #Method);       \
-    return GMOCK_MOCKER_(6, constness, Method)                                \
-        .Invoke(::testing::internal::forward<GMOCK_ARG_(tn, 1, __VA_ARGS__)>( \
-                    gmock_a1),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 2, __VA_ARGS__)>( \
-                    gmock_a2),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 3, __VA_ARGS__)>( \
-                    gmock_a3),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 4, __VA_ARGS__)>( \
-                    gmock_a4),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 5, __VA_ARGS__)>( \
-                    gmock_a5),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 6, __VA_ARGS__)>( \
-                    gmock_a6));                                               \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      GMOCK_MATCHER_(tn, 1, __VA_ARGS__) gmock_a1,                            \
-      GMOCK_MATCHER_(tn, 2, __VA_ARGS__) gmock_a2,                            \
-      GMOCK_MATCHER_(tn, 3, __VA_ARGS__) gmock_a3,                            \
-      GMOCK_MATCHER_(tn, 4, __VA_ARGS__) gmock_a4,                            \
-      GMOCK_MATCHER_(tn, 5, __VA_ARGS__) gmock_a5,                            \
-      GMOCK_MATCHER_(tn, 6, __VA_ARGS__) gmock_a6) constness {                \
-    GMOCK_MOCKER_(6, constness, Method).RegisterOwner(this);                  \
-    return GMOCK_MOCKER_(6, constness, Method)                                \
-        .With(gmock_a1, gmock_a2, gmock_a3, gmock_a4, gmock_a5, gmock_a6);    \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      const ::testing::internal::WithoutMatchers&,                            \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {          \
-    return ::testing::internal::AdjustConstness_##constness(this)             \
-        ->gmock_##Method(::testing::A<GMOCK_ARG_(tn, 1, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 2, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 3, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 4, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 5, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 6, __VA_ARGS__)>());     \
-  }                                                                           \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(6, constness,  \
-                                                               Method)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD7_(tn, constness, ct, Method, ...)                        \
-  GMOCK_RESULT_(tn, __VA_ARGS__)                                              \
-  ct Method(GMOCK_ARG_(tn, 1, __VA_ARGS__) gmock_a1,                          \
-            GMOCK_ARG_(tn, 2, __VA_ARGS__) gmock_a2,                          \
-            GMOCK_ARG_(tn, 3, __VA_ARGS__) gmock_a3,                          \
-            GMOCK_ARG_(tn, 4, __VA_ARGS__) gmock_a4,                          \
-            GMOCK_ARG_(tn, 5, __VA_ARGS__) gmock_a5,                          \
-            GMOCK_ARG_(tn, 6, __VA_ARGS__) gmock_a6,                          \
-            GMOCK_ARG_(tn, 7, __VA_ARGS__) gmock_a7) constness {              \
-    GTEST_COMPILE_ASSERT_(                                                    \
-        (::testing::tuple_size<tn ::testing::internal::Function<              \
-             __VA_ARGS__>::ArgumentTuple>::value == 7),                       \
-        this_method_does_not_take_7_arguments);                               \
-    GMOCK_MOCKER_(7, constness, Method).SetOwnerAndName(this, #Method);       \
-    return GMOCK_MOCKER_(7, constness, Method)                                \
-        .Invoke(::testing::internal::forward<GMOCK_ARG_(tn, 1, __VA_ARGS__)>( \
-                    gmock_a1),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 2, __VA_ARGS__)>( \
-                    gmock_a2),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 3, __VA_ARGS__)>( \
-                    gmock_a3),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 4, __VA_ARGS__)>( \
-                    gmock_a4),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 5, __VA_ARGS__)>( \
-                    gmock_a5),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 6, __VA_ARGS__)>( \
-                    gmock_a6),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 7, __VA_ARGS__)>( \
-                    gmock_a7));                                               \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      GMOCK_MATCHER_(tn, 1, __VA_ARGS__) gmock_a1,                            \
-      GMOCK_MATCHER_(tn, 2, __VA_ARGS__) gmock_a2,                            \
-      GMOCK_MATCHER_(tn, 3, __VA_ARGS__) gmock_a3,                            \
-      GMOCK_MATCHER_(tn, 4, __VA_ARGS__) gmock_a4,                            \
-      GMOCK_MATCHER_(tn, 5, __VA_ARGS__) gmock_a5,                            \
-      GMOCK_MATCHER_(tn, 6, __VA_ARGS__) gmock_a6,                            \
-      GMOCK_MATCHER_(tn, 7, __VA_ARGS__) gmock_a7) constness {                \
-    GMOCK_MOCKER_(7, constness, Method).RegisterOwner(this);                  \
-    return GMOCK_MOCKER_(7, constness, Method)                                \
-        .With(gmock_a1, gmock_a2, gmock_a3, gmock_a4, gmock_a5, gmock_a6,     \
-              gmock_a7);                                                      \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      const ::testing::internal::WithoutMatchers&,                            \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {          \
-    return ::testing::internal::AdjustConstness_##constness(this)             \
-        ->gmock_##Method(::testing::A<GMOCK_ARG_(tn, 1, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 2, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 3, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 4, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 5, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 6, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 7, __VA_ARGS__)>());     \
-  }                                                                           \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(7, constness,  \
-                                                               Method)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD8_(tn, constness, ct, Method, ...)                        \
-  GMOCK_RESULT_(tn, __VA_ARGS__)                                              \
-  ct Method(GMOCK_ARG_(tn, 1, __VA_ARGS__) gmock_a1,                          \
-            GMOCK_ARG_(tn, 2, __VA_ARGS__) gmock_a2,                          \
-            GMOCK_ARG_(tn, 3, __VA_ARGS__) gmock_a3,                          \
-            GMOCK_ARG_(tn, 4, __VA_ARGS__) gmock_a4,                          \
-            GMOCK_ARG_(tn, 5, __VA_ARGS__) gmock_a5,                          \
-            GMOCK_ARG_(tn, 6, __VA_ARGS__) gmock_a6,                          \
-            GMOCK_ARG_(tn, 7, __VA_ARGS__) gmock_a7,                          \
-            GMOCK_ARG_(tn, 8, __VA_ARGS__) gmock_a8) constness {              \
-    GTEST_COMPILE_ASSERT_(                                                    \
-        (::testing::tuple_size<tn ::testing::internal::Function<              \
-             __VA_ARGS__>::ArgumentTuple>::value == 8),                       \
-        this_method_does_not_take_8_arguments);                               \
-    GMOCK_MOCKER_(8, constness, Method).SetOwnerAndName(this, #Method);       \
-    return GMOCK_MOCKER_(8, constness, Method)                                \
-        .Invoke(::testing::internal::forward<GMOCK_ARG_(tn, 1, __VA_ARGS__)>( \
-                    gmock_a1),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 2, __VA_ARGS__)>( \
-                    gmock_a2),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 3, __VA_ARGS__)>( \
-                    gmock_a3),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 4, __VA_ARGS__)>( \
-                    gmock_a4),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 5, __VA_ARGS__)>( \
-                    gmock_a5),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 6, __VA_ARGS__)>( \
-                    gmock_a6),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 7, __VA_ARGS__)>( \
-                    gmock_a7),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 8, __VA_ARGS__)>( \
-                    gmock_a8));                                               \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      GMOCK_MATCHER_(tn, 1, __VA_ARGS__) gmock_a1,                            \
-      GMOCK_MATCHER_(tn, 2, __VA_ARGS__) gmock_a2,                            \
-      GMOCK_MATCHER_(tn, 3, __VA_ARGS__) gmock_a3,                            \
-      GMOCK_MATCHER_(tn, 4, __VA_ARGS__) gmock_a4,                            \
-      GMOCK_MATCHER_(tn, 5, __VA_ARGS__) gmock_a5,                            \
-      GMOCK_MATCHER_(tn, 6, __VA_ARGS__) gmock_a6,                            \
-      GMOCK_MATCHER_(tn, 7, __VA_ARGS__) gmock_a7,                            \
-      GMOCK_MATCHER_(tn, 8, __VA_ARGS__) gmock_a8) constness {                \
-    GMOCK_MOCKER_(8, constness, Method).RegisterOwner(this);                  \
-    return GMOCK_MOCKER_(8, constness, Method)                                \
-        .With(gmock_a1, gmock_a2, gmock_a3, gmock_a4, gmock_a5, gmock_a6,     \
-              gmock_a7, gmock_a8);                                            \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      const ::testing::internal::WithoutMatchers&,                            \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {          \
-    return ::testing::internal::AdjustConstness_##constness(this)             \
-        ->gmock_##Method(::testing::A<GMOCK_ARG_(tn, 1, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 2, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 3, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 4, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 5, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 6, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 7, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 8, __VA_ARGS__)>());     \
-  }                                                                           \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(8, constness,  \
-                                                               Method)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD9_(tn, constness, ct, Method, ...)                        \
-  GMOCK_RESULT_(tn, __VA_ARGS__)                                              \
-  ct Method(GMOCK_ARG_(tn, 1, __VA_ARGS__) gmock_a1,                          \
-            GMOCK_ARG_(tn, 2, __VA_ARGS__) gmock_a2,                          \
-            GMOCK_ARG_(tn, 3, __VA_ARGS__) gmock_a3,                          \
-            GMOCK_ARG_(tn, 4, __VA_ARGS__) gmock_a4,                          \
-            GMOCK_ARG_(tn, 5, __VA_ARGS__) gmock_a5,                          \
-            GMOCK_ARG_(tn, 6, __VA_ARGS__) gmock_a6,                          \
-            GMOCK_ARG_(tn, 7, __VA_ARGS__) gmock_a7,                          \
-            GMOCK_ARG_(tn, 8, __VA_ARGS__) gmock_a8,                          \
-            GMOCK_ARG_(tn, 9, __VA_ARGS__) gmock_a9) constness {              \
-    GTEST_COMPILE_ASSERT_(                                                    \
-        (::testing::tuple_size<tn ::testing::internal::Function<              \
-             __VA_ARGS__>::ArgumentTuple>::value == 9),                       \
-        this_method_does_not_take_9_arguments);                               \
-    GMOCK_MOCKER_(9, constness, Method).SetOwnerAndName(this, #Method);       \
-    return GMOCK_MOCKER_(9, constness, Method)                                \
-        .Invoke(::testing::internal::forward<GMOCK_ARG_(tn, 1, __VA_ARGS__)>( \
-                    gmock_a1),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 2, __VA_ARGS__)>( \
-                    gmock_a2),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 3, __VA_ARGS__)>( \
-                    gmock_a3),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 4, __VA_ARGS__)>( \
-                    gmock_a4),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 5, __VA_ARGS__)>( \
-                    gmock_a5),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 6, __VA_ARGS__)>( \
-                    gmock_a6),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 7, __VA_ARGS__)>( \
-                    gmock_a7),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 8, __VA_ARGS__)>( \
-                    gmock_a8),                                                \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 9, __VA_ARGS__)>( \
-                    gmock_a9));                                               \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      GMOCK_MATCHER_(tn, 1, __VA_ARGS__) gmock_a1,                            \
-      GMOCK_MATCHER_(tn, 2, __VA_ARGS__) gmock_a2,                            \
-      GMOCK_MATCHER_(tn, 3, __VA_ARGS__) gmock_a3,                            \
-      GMOCK_MATCHER_(tn, 4, __VA_ARGS__) gmock_a4,                            \
-      GMOCK_MATCHER_(tn, 5, __VA_ARGS__) gmock_a5,                            \
-      GMOCK_MATCHER_(tn, 6, __VA_ARGS__) gmock_a6,                            \
-      GMOCK_MATCHER_(tn, 7, __VA_ARGS__) gmock_a7,                            \
-      GMOCK_MATCHER_(tn, 8, __VA_ARGS__) gmock_a8,                            \
-      GMOCK_MATCHER_(tn, 9, __VA_ARGS__) gmock_a9) constness {                \
-    GMOCK_MOCKER_(9, constness, Method).RegisterOwner(this);                  \
-    return GMOCK_MOCKER_(9, constness, Method)                                \
-        .With(gmock_a1, gmock_a2, gmock_a3, gmock_a4, gmock_a5, gmock_a6,     \
-              gmock_a7, gmock_a8, gmock_a9);                                  \
-  }                                                                           \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                            \
-      const ::testing::internal::WithoutMatchers&,                            \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {          \
-    return ::testing::internal::AdjustConstness_##constness(this)             \
-        ->gmock_##Method(::testing::A<GMOCK_ARG_(tn, 1, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 2, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 3, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 4, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 5, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 6, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 7, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 8, __VA_ARGS__)>(),      \
-                         ::testing::A<GMOCK_ARG_(tn, 9, __VA_ARGS__)>());     \
-  }                                                                           \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(9, constness,  \
-                                                               Method)
-
-// INTERNAL IMPLEMENTATION - DON'T USE IN USER CODE!!!
-#define GMOCK_METHOD10_(tn, constness, ct, Method, ...)                        \
-  GMOCK_RESULT_(tn, __VA_ARGS__)                                               \
-  ct Method(GMOCK_ARG_(tn, 1, __VA_ARGS__) gmock_a1,                           \
-            GMOCK_ARG_(tn, 2, __VA_ARGS__) gmock_a2,                           \
-            GMOCK_ARG_(tn, 3, __VA_ARGS__) gmock_a3,                           \
-            GMOCK_ARG_(tn, 4, __VA_ARGS__) gmock_a4,                           \
-            GMOCK_ARG_(tn, 5, __VA_ARGS__) gmock_a5,                           \
-            GMOCK_ARG_(tn, 6, __VA_ARGS__) gmock_a6,                           \
-            GMOCK_ARG_(tn, 7, __VA_ARGS__) gmock_a7,                           \
-            GMOCK_ARG_(tn, 8, __VA_ARGS__) gmock_a8,                           \
-            GMOCK_ARG_(tn, 9, __VA_ARGS__) gmock_a9,                           \
-            GMOCK_ARG_(tn, 10, __VA_ARGS__) gmock_a10) constness {             \
-    GTEST_COMPILE_ASSERT_(                                                     \
-        (::testing::tuple_size<tn ::testing::internal::Function<               \
-             __VA_ARGS__>::ArgumentTuple>::value == 10),                       \
-        this_method_does_not_take_10_arguments);                               \
-    GMOCK_MOCKER_(10, constness, Method).SetOwnerAndName(this, #Method);       \
-    return GMOCK_MOCKER_(10, constness, Method)                                \
-        .Invoke(::testing::internal::forward<GMOCK_ARG_(tn, 1, __VA_ARGS__)>(  \
-                    gmock_a1),                                                 \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 2, __VA_ARGS__)>(  \
-                    gmock_a2),                                                 \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 3, __VA_ARGS__)>(  \
-                    gmock_a3),                                                 \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 4, __VA_ARGS__)>(  \
-                    gmock_a4),                                                 \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 5, __VA_ARGS__)>(  \
-                    gmock_a5),                                                 \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 6, __VA_ARGS__)>(  \
-                    gmock_a6),                                                 \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 7, __VA_ARGS__)>(  \
-                    gmock_a7),                                                 \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 8, __VA_ARGS__)>(  \
-                    gmock_a8),                                                 \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 9, __VA_ARGS__)>(  \
-                    gmock_a9),                                                 \
-                ::testing::internal::forward<GMOCK_ARG_(tn, 10, __VA_ARGS__)>( \
-                    gmock_a10));                                               \
-  }                                                                            \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                             \
-      GMOCK_MATCHER_(tn, 1, __VA_ARGS__) gmock_a1,                             \
-      GMOCK_MATCHER_(tn, 2, __VA_ARGS__) gmock_a2,                             \
-      GMOCK_MATCHER_(tn, 3, __VA_ARGS__) gmock_a3,                             \
-      GMOCK_MATCHER_(tn, 4, __VA_ARGS__) gmock_a4,                             \
-      GMOCK_MATCHER_(tn, 5, __VA_ARGS__) gmock_a5,                             \
-      GMOCK_MATCHER_(tn, 6, __VA_ARGS__) gmock_a6,                             \
-      GMOCK_MATCHER_(tn, 7, __VA_ARGS__) gmock_a7,                             \
-      GMOCK_MATCHER_(tn, 8, __VA_ARGS__) gmock_a8,                             \
-      GMOCK_MATCHER_(tn, 9, __VA_ARGS__) gmock_a9,                             \
-      GMOCK_MATCHER_(tn, 10, __VA_ARGS__) gmock_a10) constness {               \
-    GMOCK_MOCKER_(10, constness, Method).RegisterOwner(this);                  \
-    return GMOCK_MOCKER_(10, constness, Method)                                \
-        .With(gmock_a1, gmock_a2, gmock_a3, gmock_a4, gmock_a5, gmock_a6,      \
-              gmock_a7, gmock_a8, gmock_a9, gmock_a10);                        \
-  }                                                                            \
-  ::testing::MockSpec<__VA_ARGS__> gmock_##Method(                             \
-      const ::testing::internal::WithoutMatchers&,                             \
-      constness ::testing::internal::Function<__VA_ARGS__>*) const {           \
-    return ::testing::internal::AdjustConstness_##constness(this)              \
-        ->gmock_##Method(::testing::A<GMOCK_ARG_(tn, 1, __VA_ARGS__)>(),       \
-                         ::testing::A<GMOCK_ARG_(tn, 2, __VA_ARGS__)>(),       \
-                         ::testing::A<GMOCK_ARG_(tn, 3, __VA_ARGS__)>(),       \
-                         ::testing::A<GMOCK_ARG_(tn, 4, __VA_ARGS__)>(),       \
-                         ::testing::A<GMOCK_ARG_(tn, 5, __VA_ARGS__)>(),       \
-                         ::testing::A<GMOCK_ARG_(tn, 6, __VA_ARGS__)>(),       \
-                         ::testing::A<GMOCK_ARG_(tn, 7, __VA_ARGS__)>(),       \
-                         ::testing::A<GMOCK_ARG_(tn, 8, __VA_ARGS__)>(),       \
-                         ::testing::A<GMOCK_ARG_(tn, 9, __VA_ARGS__)>(),       \
-                         ::testing::A<GMOCK_ARG_(tn, 10, __VA_ARGS__)>());     \
-  }                                                                            \
-  mutable ::testing::FunctionMocker<__VA_ARGS__> GMOCK_MOCKER_(10, constness,  \
-                                                               Method)
-
-#define MOCK_METHOD0(m, ...) GMOCK_METHOD0_(, , , m, __VA_ARGS__)
-#define MOCK_METHOD1(m, ...) GMOCK_METHOD1_(, , , m, __VA_ARGS__)
-#define MOCK_METHOD2(m, ...) GMOCK_METHOD2_(, , , m, __VA_ARGS__)
-#define MOCK_METHOD3(m, ...) GMOCK_METHOD3_(, , , m, __VA_ARGS__)
-#define MOCK_METHOD4(m, ...) GMOCK_METHOD4_(, , , m, __VA_ARGS__)
-#define MOCK_METHOD5(m, ...) GMOCK_METHOD5_(, , , m, __VA_ARGS__)
-#define MOCK_METHOD6(m, ...) GMOCK_METHOD6_(, , , m, __VA_ARGS__)
-#define MOCK_METHOD7(m, ...) GMOCK_METHOD7_(, , , m, __VA_ARGS__)
-#define MOCK_METHOD8(m, ...) GMOCK_METHOD8_(, , , m, __VA_ARGS__)
-#define MOCK_METHOD9(m, ...) GMOCK_METHOD9_(, , , m, __VA_ARGS__)
-#define MOCK_METHOD10(m, ...) GMOCK_METHOD10_(, , , m, __VA_ARGS__)
-
-#define MOCK_CONST_METHOD0(m, ...) GMOCK_METHOD0_(, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD1(m, ...) GMOCK_METHOD1_(, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD2(m, ...) GMOCK_METHOD2_(, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD3(m, ...) GMOCK_METHOD3_(, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD4(m, ...) GMOCK_METHOD4_(, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD5(m, ...) GMOCK_METHOD5_(, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD6(m, ...) GMOCK_METHOD6_(, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD7(m, ...) GMOCK_METHOD7_(, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD8(m, ...) GMOCK_METHOD8_(, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD9(m, ...) GMOCK_METHOD9_(, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD10(m, ...) GMOCK_METHOD10_(, const, , m, __VA_ARGS__)
-
-#define MOCK_METHOD0_T(m, ...) GMOCK_METHOD0_(typename, , , m, __VA_ARGS__)
-#define MOCK_METHOD1_T(m, ...) GMOCK_METHOD1_(typename, , , m, __VA_ARGS__)
-#define MOCK_METHOD2_T(m, ...) GMOCK_METHOD2_(typename, , , m, __VA_ARGS__)
-#define MOCK_METHOD3_T(m, ...) GMOCK_METHOD3_(typename, , , m, __VA_ARGS__)
-#define MOCK_METHOD4_T(m, ...) GMOCK_METHOD4_(typename, , , m, __VA_ARGS__)
-#define MOCK_METHOD5_T(m, ...) GMOCK_METHOD5_(typename, , , m, __VA_ARGS__)
-#define MOCK_METHOD6_T(m, ...) GMOCK_METHOD6_(typename, , , m, __VA_ARGS__)
-#define MOCK_METHOD7_T(m, ...) GMOCK_METHOD7_(typename, , , m, __VA_ARGS__)
-#define MOCK_METHOD8_T(m, ...) GMOCK_METHOD8_(typename, , , m, __VA_ARGS__)
-#define MOCK_METHOD9_T(m, ...) GMOCK_METHOD9_(typename, , , m, __VA_ARGS__)
-#define MOCK_METHOD10_T(m, ...) GMOCK_METHOD10_(typename, , , m, __VA_ARGS__)
-
-#define MOCK_CONST_METHOD0_T(m, ...) \
-    GMOCK_METHOD0_(typename, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD1_T(m, ...) \
-    GMOCK_METHOD1_(typename, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD2_T(m, ...) \
-    GMOCK_METHOD2_(typename, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD3_T(m, ...) \
-    GMOCK_METHOD3_(typename, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD4_T(m, ...) \
-    GMOCK_METHOD4_(typename, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD5_T(m, ...) \
-    GMOCK_METHOD5_(typename, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD6_T(m, ...) \
-    GMOCK_METHOD6_(typename, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD7_T(m, ...) \
-    GMOCK_METHOD7_(typename, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD8_T(m, ...) \
-    GMOCK_METHOD8_(typename, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD9_T(m, ...) \
-    GMOCK_METHOD9_(typename, const, , m, __VA_ARGS__)
-#define MOCK_CONST_METHOD10_T(m, ...) \
-    GMOCK_METHOD10_(typename, const, , m, __VA_ARGS__)
-
-#define MOCK_METHOD0_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD0_(, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD1_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD1_(, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD2_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD2_(, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD3_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD3_(, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD4_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD4_(, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD5_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD5_(, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD6_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD6_(, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD7_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD7_(, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD8_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD8_(, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD9_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD9_(, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD10_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD10_(, , ct, m, __VA_ARGS__)
-
-#define MOCK_CONST_METHOD0_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD0_(, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD1_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD1_(, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD2_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD2_(, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD3_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD3_(, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD4_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD4_(, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD5_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD5_(, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD6_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD6_(, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD7_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD7_(, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD8_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD8_(, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD9_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD9_(, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD10_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD10_(, const, ct, m, __VA_ARGS__)
-
-#define MOCK_METHOD0_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD0_(typename, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD1_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD1_(typename, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD2_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD2_(typename, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD3_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD3_(typename, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD4_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD4_(typename, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD5_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD5_(typename, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD6_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD6_(typename, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD7_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD7_(typename, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD8_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD8_(typename, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD9_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD9_(typename, , ct, m, __VA_ARGS__)
-#define MOCK_METHOD10_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD10_(typename, , ct, m, __VA_ARGS__)
-
-#define MOCK_CONST_METHOD0_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD0_(typename, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD1_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD1_(typename, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD2_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD2_(typename, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD3_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD3_(typename, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD4_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD4_(typename, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD5_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD5_(typename, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD6_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD6_(typename, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD7_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD7_(typename, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD8_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD8_(typename, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD9_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD9_(typename, const, ct, m, __VA_ARGS__)
-#define MOCK_CONST_METHOD10_T_WITH_CALLTYPE(ct, m, ...) \
-    GMOCK_METHOD10_(typename, const, ct, m, __VA_ARGS__)
-
-// A MockFunction<F> class has one mock method whose type is F.  It is
-// useful when you just want your test code to emit some messages and
-// have Google Mock verify the right messages are sent (and perhaps at
-// the right times).  For example, if you are exercising code:
-//
-//   Foo(1);
-//   Foo(2);
-//   Foo(3);
-//
-// and want to verify that Foo(1) and Foo(3) both invoke
-// mock.Bar("a"), but Foo(2) doesn't invoke anything, you can write:
-//
-// TEST(FooTest, InvokesBarCorrectly) {
-//   MyMock mock;
-//   MockFunction<void(string check_point_name)> check;
-//   {
-//     InSequence s;
-//
-//     EXPECT_CALL(mock, Bar("a"));
-//     EXPECT_CALL(check, Call("1"));
-//     EXPECT_CALL(check, Call("2"));
-//     EXPECT_CALL(mock, Bar("a"));
-//   }
-//   Foo(1);
-//   check.Call("1");
-//   Foo(2);
-//   check.Call("2");
-//   Foo(3);
-// }
-//
-// The expectation spec says that the first Bar("a") must happen
-// before check point "1", the second Bar("a") must happen after check
-// point "2", and nothing should happen between the two check
-// points. The explicit check points make it easy to tell which
-// Bar("a") is called by which call to Foo().
-//
-// MockFunction<F> can also be used to exercise code that accepts
-// std::function<F> callbacks. To do so, use AsStdFunction() method
-// to create std::function proxy forwarding to original object's Call.
-// Example:
-//
-// TEST(FooTest, RunsCallbackWithBarArgument) {
-//   MockFunction<int(string)> callback;
-//   EXPECT_CALL(callback, Call("bar")).WillOnce(Return(1));
-//   Foo(callback.AsStdFunction());
-// }
-template <typename F>
-class MockFunction;
-
-template <typename R>
-class MockFunction<R()> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD0_T(Call, R());
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R()> AsStdFunction() {
-    return [this]() -> R {
-      return this->Call();
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-template <typename R, typename A0>
-class MockFunction<R(A0)> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD1_T(Call, R(A0));
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R(A0)> AsStdFunction() {
-    return [this](A0 a0) -> R {
-      return this->Call(::std::move(a0));
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-template <typename R, typename A0, typename A1>
-class MockFunction<R(A0, A1)> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD2_T(Call, R(A0, A1));
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R(A0, A1)> AsStdFunction() {
-    return [this](A0 a0, A1 a1) -> R {
-      return this->Call(::std::move(a0), ::std::move(a1));
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-template <typename R, typename A0, typename A1, typename A2>
-class MockFunction<R(A0, A1, A2)> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD3_T(Call, R(A0, A1, A2));
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R(A0, A1, A2)> AsStdFunction() {
-    return [this](A0 a0, A1 a1, A2 a2) -> R {
-      return this->Call(::std::move(a0), ::std::move(a1), ::std::move(a2));
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-template <typename R, typename A0, typename A1, typename A2, typename A3>
-class MockFunction<R(A0, A1, A2, A3)> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD4_T(Call, R(A0, A1, A2, A3));
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R(A0, A1, A2, A3)> AsStdFunction() {
-    return [this](A0 a0, A1 a1, A2 a2, A3 a3) -> R {
-      return this->Call(::std::move(a0), ::std::move(a1), ::std::move(a2),
-          ::std::move(a3));
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-template <typename R, typename A0, typename A1, typename A2, typename A3,
-    typename A4>
-class MockFunction<R(A0, A1, A2, A3, A4)> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD5_T(Call, R(A0, A1, A2, A3, A4));
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R(A0, A1, A2, A3, A4)> AsStdFunction() {
-    return [this](A0 a0, A1 a1, A2 a2, A3 a3, A4 a4) -> R {
-      return this->Call(::std::move(a0), ::std::move(a1), ::std::move(a2),
-          ::std::move(a3), ::std::move(a4));
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-template <typename R, typename A0, typename A1, typename A2, typename A3,
-    typename A4, typename A5>
-class MockFunction<R(A0, A1, A2, A3, A4, A5)> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD6_T(Call, R(A0, A1, A2, A3, A4, A5));
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R(A0, A1, A2, A3, A4, A5)> AsStdFunction() {
-    return [this](A0 a0, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5) -> R {
-      return this->Call(::std::move(a0), ::std::move(a1), ::std::move(a2),
-          ::std::move(a3), ::std::move(a4), ::std::move(a5));
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-template <typename R, typename A0, typename A1, typename A2, typename A3,
-    typename A4, typename A5, typename A6>
-class MockFunction<R(A0, A1, A2, A3, A4, A5, A6)> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD7_T(Call, R(A0, A1, A2, A3, A4, A5, A6));
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R(A0, A1, A2, A3, A4, A5, A6)> AsStdFunction() {
-    return [this](A0 a0, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6) -> R {
-      return this->Call(::std::move(a0), ::std::move(a1), ::std::move(a2),
-          ::std::move(a3), ::std::move(a4), ::std::move(a5), ::std::move(a6));
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-template <typename R, typename A0, typename A1, typename A2, typename A3,
-    typename A4, typename A5, typename A6, typename A7>
-class MockFunction<R(A0, A1, A2, A3, A4, A5, A6, A7)> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD8_T(Call, R(A0, A1, A2, A3, A4, A5, A6, A7));
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R(A0, A1, A2, A3, A4, A5, A6, A7)> AsStdFunction() {
-    return [this](A0 a0, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6, A7 a7) -> R {
-      return this->Call(::std::move(a0), ::std::move(a1), ::std::move(a2),
-          ::std::move(a3), ::std::move(a4), ::std::move(a5), ::std::move(a6),
-          ::std::move(a7));
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-template <typename R, typename A0, typename A1, typename A2, typename A3,
-    typename A4, typename A5, typename A6, typename A7, typename A8>
-class MockFunction<R(A0, A1, A2, A3, A4, A5, A6, A7, A8)> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD9_T(Call, R(A0, A1, A2, A3, A4, A5, A6, A7, A8));
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R(A0, A1, A2, A3, A4, A5, A6, A7, A8)> AsStdFunction() {
-    return [this](A0 a0, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6, A7 a7,
-        A8 a8) -> R {
-      return this->Call(::std::move(a0), ::std::move(a1), ::std::move(a2),
-          ::std::move(a3), ::std::move(a4), ::std::move(a5), ::std::move(a6),
-          ::std::move(a7), ::std::move(a8));
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-template <typename R, typename A0, typename A1, typename A2, typename A3,
-    typename A4, typename A5, typename A6, typename A7, typename A8,
-    typename A9>
-class MockFunction<R(A0, A1, A2, A3, A4, A5, A6, A7, A8, A9)> {
- public:
-  MockFunction() {}
-
-  MOCK_METHOD10_T(Call, R(A0, A1, A2, A3, A4, A5, A6, A7, A8, A9));
-
-#if GTEST_HAS_STD_FUNCTION_
-  ::std::function<R(A0, A1, A2, A3, A4, A5, A6, A7, A8, A9)> AsStdFunction() {
-    return [this](A0 a0, A1 a1, A2 a2, A3 a3, A4 a4, A5 a5, A6 a6, A7 a7,
-        A8 a8, A9 a9) -> R {
-      return this->Call(::std::move(a0), ::std::move(a1), ::std::move(a2),
-          ::std::move(a3), ::std::move(a4), ::std::move(a5), ::std::move(a6),
-          ::std::move(a7), ::std::move(a8), ::std::move(a9));
-    };
-  }
-#endif  // GTEST_HAS_STD_FUNCTION_
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(MockFunction);
-};
-
-}  // namespace testing
-
-#endif  // GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_FUNCTION_MOCKERS_H_
-// This file was GENERATED by command:
-//     pump.py gmock-generated-matchers.h.pump
-// DO NOT EDIT BY HAND!!!
-
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements some commonly used variadic matchers.
-
-#ifndef GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_MATCHERS_H_
-#define GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_MATCHERS_H_
-
-#include <iterator>
-#include <sstream>
-#include <string>
-#include <vector>
-
-namespace testing {
-namespace internal {
-
-// The type of the i-th (0-based) field of Tuple.
-#define GMOCK_FIELD_TYPE_(Tuple, i) \
-    typename ::testing::tuple_element<i, Tuple>::type
-
-// TupleFields<Tuple, k0, ..., kn> is for selecting fields from a
-// tuple of type Tuple.  It has two members:
-//
-//   type: a tuple type whose i-th field is the ki-th field of Tuple.
-//   GetSelectedFields(t): returns fields k0, ..., and kn of t as a tuple.
-//
-// For example, in class TupleFields<tuple<bool, char, int>, 2, 0>, we have:
-//
-//   type is tuple<int, bool>, and
-//   GetSelectedFields(make_tuple(true, 'a', 42)) is (42, true).
-
-template <class Tuple, int k0 = -1, int k1 = -1, int k2 = -1, int k3 = -1,
-    int k4 = -1, int k5 = -1, int k6 = -1, int k7 = -1, int k8 = -1,
-    int k9 = -1>
-class TupleFields;
-
-// This generic version is used when there are 10 selectors.
-template <class Tuple, int k0, int k1, int k2, int k3, int k4, int k5, int k6,
-    int k7, int k8, int k9>
-class TupleFields {
- public:
-  typedef ::testing::tuple<GMOCK_FIELD_TYPE_(Tuple, k0),
-      GMOCK_FIELD_TYPE_(Tuple, k1), GMOCK_FIELD_TYPE_(Tuple, k2),
-      GMOCK_FIELD_TYPE_(Tuple, k3), GMOCK_FIELD_TYPE_(Tuple, k4),
-      GMOCK_FIELD_TYPE_(Tuple, k5), GMOCK_FIELD_TYPE_(Tuple, k6),
-      GMOCK_FIELD_TYPE_(Tuple, k7), GMOCK_FIELD_TYPE_(Tuple, k8),
-      GMOCK_FIELD_TYPE_(Tuple, k9)> type;
-  static type GetSelectedFields(const Tuple& t) {
-    return type(get<k0>(t), get<k1>(t), get<k2>(t), get<k3>(t), get<k4>(t),
-        get<k5>(t), get<k6>(t), get<k7>(t), get<k8>(t), get<k9>(t));
-  }
-};
-
-// The following specialization is used for 0 ~ 9 selectors.
-
-template <class Tuple>
-class TupleFields<Tuple, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1> {
- public:
-  typedef ::testing::tuple<> type;
-  static type GetSelectedFields(const Tuple& /* t */) {
-    return type();
-  }
-};
-
-template <class Tuple, int k0>
-class TupleFields<Tuple, k0, -1, -1, -1, -1, -1, -1, -1, -1, -1> {
- public:
-  typedef ::testing::tuple<GMOCK_FIELD_TYPE_(Tuple, k0)> type;
-  static type GetSelectedFields(const Tuple& t) {
-    return type(get<k0>(t));
-  }
-};
-
-template <class Tuple, int k0, int k1>
-class TupleFields<Tuple, k0, k1, -1, -1, -1, -1, -1, -1, -1, -1> {
- public:
-  typedef ::testing::tuple<GMOCK_FIELD_TYPE_(Tuple, k0),
-      GMOCK_FIELD_TYPE_(Tuple, k1)> type;
-  static type GetSelectedFields(const Tuple& t) {
-    return type(get<k0>(t), get<k1>(t));
-  }
-};
-
-template <class Tuple, int k0, int k1, int k2>
-class TupleFields<Tuple, k0, k1, k2, -1, -1, -1, -1, -1, -1, -1> {
- public:
-  typedef ::testing::tuple<GMOCK_FIELD_TYPE_(Tuple, k0),
-      GMOCK_FIELD_TYPE_(Tuple, k1), GMOCK_FIELD_TYPE_(Tuple, k2)> type;
-  static type GetSelectedFields(const Tuple& t) {
-    return type(get<k0>(t), get<k1>(t), get<k2>(t));
-  }
-};
-
-template <class Tuple, int k0, int k1, int k2, int k3>
-class TupleFields<Tuple, k0, k1, k2, k3, -1, -1, -1, -1, -1, -1> {
- public:
-  typedef ::testing::tuple<GMOCK_FIELD_TYPE_(Tuple, k0),
-      GMOCK_FIELD_TYPE_(Tuple, k1), GMOCK_FIELD_TYPE_(Tuple, k2),
-      GMOCK_FIELD_TYPE_(Tuple, k3)> type;
-  static type GetSelectedFields(const Tuple& t) {
-    return type(get<k0>(t), get<k1>(t), get<k2>(t), get<k3>(t));
-  }
-};
-
-template <class Tuple, int k0, int k1, int k2, int k3, int k4>
-class TupleFields<Tuple, k0, k1, k2, k3, k4, -1, -1, -1, -1, -1> {
- public:
-  typedef ::testing::tuple<GMOCK_FIELD_TYPE_(Tuple, k0),
-      GMOCK_FIELD_TYPE_(Tuple, k1), GMOCK_FIELD_TYPE_(Tuple, k2),
-      GMOCK_FIELD_TYPE_(Tuple, k3), GMOCK_FIELD_TYPE_(Tuple, k4)> type;
-  static type GetSelectedFields(const Tuple& t) {
-    return type(get<k0>(t), get<k1>(t), get<k2>(t), get<k3>(t), get<k4>(t));
-  }
-};
-
-template <class Tuple, int k0, int k1, int k2, int k3, int k4, int k5>
-class TupleFields<Tuple, k0, k1, k2, k3, k4, k5, -1, -1, -1, -1> {
- public:
-  typedef ::testing::tuple<GMOCK_FIELD_TYPE_(Tuple, k0),
-      GMOCK_FIELD_TYPE_(Tuple, k1), GMOCK_FIELD_TYPE_(Tuple, k2),
-      GMOCK_FIELD_TYPE_(Tuple, k3), GMOCK_FIELD_TYPE_(Tuple, k4),
-      GMOCK_FIELD_TYPE_(Tuple, k5)> type;
-  static type GetSelectedFields(const Tuple& t) {
-    return type(get<k0>(t), get<k1>(t), get<k2>(t), get<k3>(t), get<k4>(t),
-        get<k5>(t));
-  }
-};
-
-template <class Tuple, int k0, int k1, int k2, int k3, int k4, int k5, int k6>
-class TupleFields<Tuple, k0, k1, k2, k3, k4, k5, k6, -1, -1, -1> {
- public:
-  typedef ::testing::tuple<GMOCK_FIELD_TYPE_(Tuple, k0),
-      GMOCK_FIELD_TYPE_(Tuple, k1), GMOCK_FIELD_TYPE_(Tuple, k2),
-      GMOCK_FIELD_TYPE_(Tuple, k3), GMOCK_FIELD_TYPE_(Tuple, k4),
-      GMOCK_FIELD_TYPE_(Tuple, k5), GMOCK_FIELD_TYPE_(Tuple, k6)> type;
-  static type GetSelectedFields(const Tuple& t) {
-    return type(get<k0>(t), get<k1>(t), get<k2>(t), get<k3>(t), get<k4>(t),
-        get<k5>(t), get<k6>(t));
-  }
-};
-
-template <class Tuple, int k0, int k1, int k2, int k3, int k4, int k5, int k6,
-    int k7>
-class TupleFields<Tuple, k0, k1, k2, k3, k4, k5, k6, k7, -1, -1> {
- public:
-  typedef ::testing::tuple<GMOCK_FIELD_TYPE_(Tuple, k0),
-      GMOCK_FIELD_TYPE_(Tuple, k1), GMOCK_FIELD_TYPE_(Tuple, k2),
-      GMOCK_FIELD_TYPE_(Tuple, k3), GMOCK_FIELD_TYPE_(Tuple, k4),
-      GMOCK_FIELD_TYPE_(Tuple, k5), GMOCK_FIELD_TYPE_(Tuple, k6),
-      GMOCK_FIELD_TYPE_(Tuple, k7)> type;
-  static type GetSelectedFields(const Tuple& t) {
-    return type(get<k0>(t), get<k1>(t), get<k2>(t), get<k3>(t), get<k4>(t),
-        get<k5>(t), get<k6>(t), get<k7>(t));
-  }
-};
-
-template <class Tuple, int k0, int k1, int k2, int k3, int k4, int k5, int k6,
-    int k7, int k8>
-class TupleFields<Tuple, k0, k1, k2, k3, k4, k5, k6, k7, k8, -1> {
- public:
-  typedef ::testing::tuple<GMOCK_FIELD_TYPE_(Tuple, k0),
-      GMOCK_FIELD_TYPE_(Tuple, k1), GMOCK_FIELD_TYPE_(Tuple, k2),
-      GMOCK_FIELD_TYPE_(Tuple, k3), GMOCK_FIELD_TYPE_(Tuple, k4),
-      GMOCK_FIELD_TYPE_(Tuple, k5), GMOCK_FIELD_TYPE_(Tuple, k6),
-      GMOCK_FIELD_TYPE_(Tuple, k7), GMOCK_FIELD_TYPE_(Tuple, k8)> type;
-  static type GetSelectedFields(const Tuple& t) {
-    return type(get<k0>(t), get<k1>(t), get<k2>(t), get<k3>(t), get<k4>(t),
-        get<k5>(t), get<k6>(t), get<k7>(t), get<k8>(t));
-  }
-};
-
-#undef GMOCK_FIELD_TYPE_
-
-// Implements the Args() matcher.
-template <class ArgsTuple, int k0 = -1, int k1 = -1, int k2 = -1, int k3 = -1,
-    int k4 = -1, int k5 = -1, int k6 = -1, int k7 = -1, int k8 = -1,
-    int k9 = -1>
-class ArgsMatcherImpl : public MatcherInterface<ArgsTuple> {
- public:
-  // ArgsTuple may have top-level const or reference modifiers.
-  typedef GTEST_REMOVE_REFERENCE_AND_CONST_(ArgsTuple) RawArgsTuple;
-  typedef typename internal::TupleFields<RawArgsTuple, k0, k1, k2, k3, k4, k5,
-      k6, k7, k8, k9>::type SelectedArgs;
-  typedef Matcher<const SelectedArgs&> MonomorphicInnerMatcher;
-
-  template <typename InnerMatcher>
-  explicit ArgsMatcherImpl(const InnerMatcher& inner_matcher)
-      : inner_matcher_(SafeMatcherCast<const SelectedArgs&>(inner_matcher)) {}
-
-  virtual bool MatchAndExplain(ArgsTuple args,
-                               MatchResultListener* listener) const {
-    const SelectedArgs& selected_args = GetSelectedArgs(args);
-    if (!listener->IsInterested())
-      return inner_matcher_.Matches(selected_args);
-
-    PrintIndices(listener->stream());
-    *listener << "are " << PrintToString(selected_args);
-
-    StringMatchResultListener inner_listener;
-    const bool match = inner_matcher_.MatchAndExplain(selected_args,
-                                                      &inner_listener);
-    PrintIfNotEmpty(inner_listener.str(), listener->stream());
-    return match;
-  }
-
-  virtual void DescribeTo(::std::ostream* os) const {
-    *os << "are a tuple ";
-    PrintIndices(os);
-    inner_matcher_.DescribeTo(os);
-  }
-
-  virtual void DescribeNegationTo(::std::ostream* os) const {
-    *os << "are a tuple ";
-    PrintIndices(os);
-    inner_matcher_.DescribeNegationTo(os);
-  }
-
- private:
-  static SelectedArgs GetSelectedArgs(ArgsTuple args) {
-    return TupleFields<RawArgsTuple, k0, k1, k2, k3, k4, k5, k6, k7, k8,
-        k9>::GetSelectedFields(args);
-  }
-
-  // Prints the indices of the selected fields.
-  static void PrintIndices(::std::ostream* os) {
-    *os << "whose fields (";
-    const int indices[10] = { k0, k1, k2, k3, k4, k5, k6, k7, k8, k9 };
-    for (int i = 0; i < 10; i++) {
-      if (indices[i] < 0)
-        break;
-
-      if (i >= 1)
-        *os << ", ";
-
-      *os << "#" << indices[i];
-    }
-    *os << ") ";
-  }
-
-  const MonomorphicInnerMatcher inner_matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(ArgsMatcherImpl);
-};
-
-template <class InnerMatcher, int k0 = -1, int k1 = -1, int k2 = -1,
-    int k3 = -1, int k4 = -1, int k5 = -1, int k6 = -1, int k7 = -1,
-    int k8 = -1, int k9 = -1>
-class ArgsMatcher {
- public:
-  explicit ArgsMatcher(const InnerMatcher& inner_matcher)
-      : inner_matcher_(inner_matcher) {}
-
-  template <typename ArgsTuple>
-  operator Matcher<ArgsTuple>() const {
-    return MakeMatcher(new ArgsMatcherImpl<ArgsTuple, k0, k1, k2, k3, k4, k5,
-        k6, k7, k8, k9>(inner_matcher_));
-  }
-
- private:
-  const InnerMatcher inner_matcher_;
-
-  GTEST_DISALLOW_ASSIGN_(ArgsMatcher);
-};
-
-// A set of metafunctions for computing the result type of AllOf.
-// AllOf(m1, ..., mN) returns
-// AllOfResultN<decltype(m1), ..., decltype(mN)>::type.
-
-// Although AllOf isn't defined for one argument, AllOfResult1 is defined
-// to simplify the implementation.
-template <typename M1>
-struct AllOfResult1 {
-  typedef M1 type;
-};
-
-template <typename M1, typename M2>
-struct AllOfResult2 {
-  typedef BothOfMatcher<
-      typename AllOfResult1<M1>::type,
-      typename AllOfResult1<M2>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3>
-struct AllOfResult3 {
-  typedef BothOfMatcher<
-      typename AllOfResult1<M1>::type,
-      typename AllOfResult2<M2, M3>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4>
-struct AllOfResult4 {
-  typedef BothOfMatcher<
-      typename AllOfResult2<M1, M2>::type,
-      typename AllOfResult2<M3, M4>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5>
-struct AllOfResult5 {
-  typedef BothOfMatcher<
-      typename AllOfResult2<M1, M2>::type,
-      typename AllOfResult3<M3, M4, M5>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6>
-struct AllOfResult6 {
-  typedef BothOfMatcher<
-      typename AllOfResult3<M1, M2, M3>::type,
-      typename AllOfResult3<M4, M5, M6>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7>
-struct AllOfResult7 {
-  typedef BothOfMatcher<
-      typename AllOfResult3<M1, M2, M3>::type,
-      typename AllOfResult4<M4, M5, M6, M7>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8>
-struct AllOfResult8 {
-  typedef BothOfMatcher<
-      typename AllOfResult4<M1, M2, M3, M4>::type,
-      typename AllOfResult4<M5, M6, M7, M8>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8, typename M9>
-struct AllOfResult9 {
-  typedef BothOfMatcher<
-      typename AllOfResult4<M1, M2, M3, M4>::type,
-      typename AllOfResult5<M5, M6, M7, M8, M9>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8, typename M9, typename M10>
-struct AllOfResult10 {
-  typedef BothOfMatcher<
-      typename AllOfResult5<M1, M2, M3, M4, M5>::type,
-      typename AllOfResult5<M6, M7, M8, M9, M10>::type
-  > type;
-};
-
-// A set of metafunctions for computing the result type of AnyOf.
-// AnyOf(m1, ..., mN) returns
-// AnyOfResultN<decltype(m1), ..., decltype(mN)>::type.
-
-// Although AnyOf isn't defined for one argument, AnyOfResult1 is defined
-// to simplify the implementation.
-template <typename M1>
-struct AnyOfResult1 {
-  typedef M1 type;
-};
-
-template <typename M1, typename M2>
-struct AnyOfResult2 {
-  typedef EitherOfMatcher<
-      typename AnyOfResult1<M1>::type,
-      typename AnyOfResult1<M2>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3>
-struct AnyOfResult3 {
-  typedef EitherOfMatcher<
-      typename AnyOfResult1<M1>::type,
-      typename AnyOfResult2<M2, M3>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4>
-struct AnyOfResult4 {
-  typedef EitherOfMatcher<
-      typename AnyOfResult2<M1, M2>::type,
-      typename AnyOfResult2<M3, M4>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5>
-struct AnyOfResult5 {
-  typedef EitherOfMatcher<
-      typename AnyOfResult2<M1, M2>::type,
-      typename AnyOfResult3<M3, M4, M5>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6>
-struct AnyOfResult6 {
-  typedef EitherOfMatcher<
-      typename AnyOfResult3<M1, M2, M3>::type,
-      typename AnyOfResult3<M4, M5, M6>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7>
-struct AnyOfResult7 {
-  typedef EitherOfMatcher<
-      typename AnyOfResult3<M1, M2, M3>::type,
-      typename AnyOfResult4<M4, M5, M6, M7>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8>
-struct AnyOfResult8 {
-  typedef EitherOfMatcher<
-      typename AnyOfResult4<M1, M2, M3, M4>::type,
-      typename AnyOfResult4<M5, M6, M7, M8>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8, typename M9>
-struct AnyOfResult9 {
-  typedef EitherOfMatcher<
-      typename AnyOfResult4<M1, M2, M3, M4>::type,
-      typename AnyOfResult5<M5, M6, M7, M8, M9>::type
-  > type;
-};
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8, typename M9, typename M10>
-struct AnyOfResult10 {
-  typedef EitherOfMatcher<
-      typename AnyOfResult5<M1, M2, M3, M4, M5>::type,
-      typename AnyOfResult5<M6, M7, M8, M9, M10>::type
-  > type;
-};
-
-}  // namespace internal
-
-// Args<N1, N2, ..., Nk>(a_matcher) matches a tuple if the selected
-// fields of it matches a_matcher.  C++ doesn't support default
-// arguments for function templates, so we have to overload it.
-template <typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher>(matcher);
-}
-
-template <int k1, typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher, k1>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher, k1>(matcher);
-}
-
-template <int k1, int k2, typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher, k1, k2>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher, k1, k2>(matcher);
-}
-
-template <int k1, int k2, int k3, typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher, k1, k2, k3>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher, k1, k2, k3>(matcher);
-}
-
-template <int k1, int k2, int k3, int k4, typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4>(matcher);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5>(matcher);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, int k6, typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5, k6>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5, k6>(matcher);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, int k6, int k7,
-    typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5, k6, k7>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5, k6,
-      k7>(matcher);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, int k6, int k7, int k8,
-    typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5, k6, k7, k8>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5, k6, k7,
-      k8>(matcher);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, int k6, int k7, int k8,
-    int k9, typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5, k6, k7, k8, k9>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5, k6, k7, k8,
-      k9>(matcher);
-}
-
-template <int k1, int k2, int k3, int k4, int k5, int k6, int k7, int k8,
-    int k9, int k10, typename InnerMatcher>
-inline internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5, k6, k7, k8, k9,
-    k10>
-Args(const InnerMatcher& matcher) {
-  return internal::ArgsMatcher<InnerMatcher, k1, k2, k3, k4, k5, k6, k7, k8,
-      k9, k10>(matcher);
-}
-
-// ElementsAre(e_1, e_2, ... e_n) matches an STL-style container with
-// n elements, where the i-th element in the container must
-// match the i-th argument in the list.  Each argument of
-// ElementsAre() can be either a value or a matcher.  We support up to
-// 10 arguments.
-//
-// The use of DecayArray in the implementation allows ElementsAre()
-// to accept string literals, whose type is const char[N], but we
-// want to treat them as const char*.
-//
-// NOTE: Since ElementsAre() cares about the order of the elements, it
-// must not be used with containers whose elements's order is
-// undefined (e.g. hash_map).
-
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<> >
-ElementsAre() {
-  typedef ::testing::tuple<> Args;
-  return internal::ElementsAreMatcher<Args>(Args());
-}
-
-template <typename T1>
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type> >
-ElementsAre(const T1& e1) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type> Args;
-  return internal::ElementsAreMatcher<Args>(Args(e1));
-}
-
-template <typename T1, typename T2>
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type> >
-ElementsAre(const T1& e1, const T2& e2) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type> Args;
-  return internal::ElementsAreMatcher<Args>(Args(e1, e2));
-}
-
-template <typename T1, typename T2, typename T3>
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type> >
-ElementsAre(const T1& e1, const T2& e2, const T3& e3) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type> Args;
-  return internal::ElementsAreMatcher<Args>(Args(e1, e2, e3));
-}
-
-template <typename T1, typename T2, typename T3, typename T4>
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type> >
-ElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type> Args;
-  return internal::ElementsAreMatcher<Args>(Args(e1, e2, e3, e4));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5>
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type> >
-ElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type> Args;
-  return internal::ElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6>
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type,
-        typename internal::DecayArray<T6>::type> >
-ElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5, const T6& e6) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type,
-      typename internal::DecayArray<T6>::type> Args;
-  return internal::ElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5, e6));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7>
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type,
-        typename internal::DecayArray<T6>::type,
-        typename internal::DecayArray<T7>::type> >
-ElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5, const T6& e6, const T7& e7) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type,
-      typename internal::DecayArray<T6>::type,
-      typename internal::DecayArray<T7>::type> Args;
-  return internal::ElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5, e6, e7));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8>
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type,
-        typename internal::DecayArray<T6>::type,
-        typename internal::DecayArray<T7>::type,
-        typename internal::DecayArray<T8>::type> >
-ElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5, const T6& e6, const T7& e7, const T8& e8) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type,
-      typename internal::DecayArray<T6>::type,
-      typename internal::DecayArray<T7>::type,
-      typename internal::DecayArray<T8>::type> Args;
-  return internal::ElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5, e6, e7,
-      e8));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9>
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type,
-        typename internal::DecayArray<T6>::type,
-        typename internal::DecayArray<T7>::type,
-        typename internal::DecayArray<T8>::type,
-        typename internal::DecayArray<T9>::type> >
-ElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5, const T6& e6, const T7& e7, const T8& e8, const T9& e9) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type,
-      typename internal::DecayArray<T6>::type,
-      typename internal::DecayArray<T7>::type,
-      typename internal::DecayArray<T8>::type,
-      typename internal::DecayArray<T9>::type> Args;
-  return internal::ElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5, e6, e7,
-      e8, e9));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10>
-inline internal::ElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type,
-        typename internal::DecayArray<T6>::type,
-        typename internal::DecayArray<T7>::type,
-        typename internal::DecayArray<T8>::type,
-        typename internal::DecayArray<T9>::type,
-        typename internal::DecayArray<T10>::type> >
-ElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5, const T6& e6, const T7& e7, const T8& e8, const T9& e9,
-    const T10& e10) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type,
-      typename internal::DecayArray<T6>::type,
-      typename internal::DecayArray<T7>::type,
-      typename internal::DecayArray<T8>::type,
-      typename internal::DecayArray<T9>::type,
-      typename internal::DecayArray<T10>::type> Args;
-  return internal::ElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5, e6, e7,
-      e8, e9, e10));
-}
-
-// UnorderedElementsAre(e_1, e_2, ..., e_n) is an ElementsAre extension
-// that matches n elements in any order.  We support up to n=10 arguments.
-//
-// If you have >10 elements, consider UnorderedElementsAreArray() or
-// UnorderedPointwise() instead.
-
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<> >
-UnorderedElementsAre() {
-  typedef ::testing::tuple<> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args());
-}
-
-template <typename T1>
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type> >
-UnorderedElementsAre(const T1& e1) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args(e1));
-}
-
-template <typename T1, typename T2>
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type> >
-UnorderedElementsAre(const T1& e1, const T2& e2) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args(e1, e2));
-}
-
-template <typename T1, typename T2, typename T3>
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type> >
-UnorderedElementsAre(const T1& e1, const T2& e2, const T3& e3) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args(e1, e2, e3));
-}
-
-template <typename T1, typename T2, typename T3, typename T4>
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type> >
-UnorderedElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args(e1, e2, e3, e4));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5>
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type> >
-UnorderedElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6>
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type,
-        typename internal::DecayArray<T6>::type> >
-UnorderedElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5, const T6& e6) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type,
-      typename internal::DecayArray<T6>::type> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5,
-      e6));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7>
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type,
-        typename internal::DecayArray<T6>::type,
-        typename internal::DecayArray<T7>::type> >
-UnorderedElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5, const T6& e6, const T7& e7) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type,
-      typename internal::DecayArray<T6>::type,
-      typename internal::DecayArray<T7>::type> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5,
-      e6, e7));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8>
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type,
-        typename internal::DecayArray<T6>::type,
-        typename internal::DecayArray<T7>::type,
-        typename internal::DecayArray<T8>::type> >
-UnorderedElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5, const T6& e6, const T7& e7, const T8& e8) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type,
-      typename internal::DecayArray<T6>::type,
-      typename internal::DecayArray<T7>::type,
-      typename internal::DecayArray<T8>::type> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5,
-      e6, e7, e8));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9>
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type,
-        typename internal::DecayArray<T6>::type,
-        typename internal::DecayArray<T7>::type,
-        typename internal::DecayArray<T8>::type,
-        typename internal::DecayArray<T9>::type> >
-UnorderedElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5, const T6& e6, const T7& e7, const T8& e8, const T9& e9) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type,
-      typename internal::DecayArray<T6>::type,
-      typename internal::DecayArray<T7>::type,
-      typename internal::DecayArray<T8>::type,
-      typename internal::DecayArray<T9>::type> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5,
-      e6, e7, e8, e9));
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10>
-inline internal::UnorderedElementsAreMatcher<
-    ::testing::tuple<
-        typename internal::DecayArray<T1>::type,
-        typename internal::DecayArray<T2>::type,
-        typename internal::DecayArray<T3>::type,
-        typename internal::DecayArray<T4>::type,
-        typename internal::DecayArray<T5>::type,
-        typename internal::DecayArray<T6>::type,
-        typename internal::DecayArray<T7>::type,
-        typename internal::DecayArray<T8>::type,
-        typename internal::DecayArray<T9>::type,
-        typename internal::DecayArray<T10>::type> >
-UnorderedElementsAre(const T1& e1, const T2& e2, const T3& e3, const T4& e4,
-    const T5& e5, const T6& e6, const T7& e7, const T8& e8, const T9& e9,
-    const T10& e10) {
-  typedef ::testing::tuple<
-      typename internal::DecayArray<T1>::type,
-      typename internal::DecayArray<T2>::type,
-      typename internal::DecayArray<T3>::type,
-      typename internal::DecayArray<T4>::type,
-      typename internal::DecayArray<T5>::type,
-      typename internal::DecayArray<T6>::type,
-      typename internal::DecayArray<T7>::type,
-      typename internal::DecayArray<T8>::type,
-      typename internal::DecayArray<T9>::type,
-      typename internal::DecayArray<T10>::type> Args;
-  return internal::UnorderedElementsAreMatcher<Args>(Args(e1, e2, e3, e4, e5,
-      e6, e7, e8, e9, e10));
-}
-
-// AllOf(m1, m2, ..., mk) matches any value that matches all of the given
-// sub-matchers.  AllOf is called fully qualified to prevent ADL from firing.
-
-template <typename M1, typename M2>
-inline typename internal::AllOfResult2<M1, M2>::type
-AllOf(M1 m1, M2 m2) {
-  return typename internal::AllOfResult2<M1, M2>::type(
-      m1,
-      m2);
-}
-
-template <typename M1, typename M2, typename M3>
-inline typename internal::AllOfResult3<M1, M2, M3>::type
-AllOf(M1 m1, M2 m2, M3 m3) {
-  return typename internal::AllOfResult3<M1, M2, M3>::type(
-      m1,
-      ::testing::AllOf(m2, m3));
-}
-
-template <typename M1, typename M2, typename M3, typename M4>
-inline typename internal::AllOfResult4<M1, M2, M3, M4>::type
-AllOf(M1 m1, M2 m2, M3 m3, M4 m4) {
-  return typename internal::AllOfResult4<M1, M2, M3, M4>::type(
-      ::testing::AllOf(m1, m2),
-      ::testing::AllOf(m3, m4));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5>
-inline typename internal::AllOfResult5<M1, M2, M3, M4, M5>::type
-AllOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5) {
-  return typename internal::AllOfResult5<M1, M2, M3, M4, M5>::type(
-      ::testing::AllOf(m1, m2),
-      ::testing::AllOf(m3, m4, m5));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6>
-inline typename internal::AllOfResult6<M1, M2, M3, M4, M5, M6>::type
-AllOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5, M6 m6) {
-  return typename internal::AllOfResult6<M1, M2, M3, M4, M5, M6>::type(
-      ::testing::AllOf(m1, m2, m3),
-      ::testing::AllOf(m4, m5, m6));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7>
-inline typename internal::AllOfResult7<M1, M2, M3, M4, M5, M6, M7>::type
-AllOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5, M6 m6, M7 m7) {
-  return typename internal::AllOfResult7<M1, M2, M3, M4, M5, M6, M7>::type(
-      ::testing::AllOf(m1, m2, m3),
-      ::testing::AllOf(m4, m5, m6, m7));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8>
-inline typename internal::AllOfResult8<M1, M2, M3, M4, M5, M6, M7, M8>::type
-AllOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5, M6 m6, M7 m7, M8 m8) {
-  return typename internal::AllOfResult8<M1, M2, M3, M4, M5, M6, M7, M8>::type(
-      ::testing::AllOf(m1, m2, m3, m4),
-      ::testing::AllOf(m5, m6, m7, m8));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8, typename M9>
-inline typename internal::AllOfResult9<M1, M2, M3, M4, M5, M6, M7, M8, M9>::type
-AllOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5, M6 m6, M7 m7, M8 m8, M9 m9) {
-  return typename internal::AllOfResult9<M1, M2, M3, M4, M5, M6, M7, M8,
-      M9>::type(
-      ::testing::AllOf(m1, m2, m3, m4),
-      ::testing::AllOf(m5, m6, m7, m8, m9));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8, typename M9, typename M10>
-inline typename internal::AllOfResult10<M1, M2, M3, M4, M5, M6, M7, M8, M9,
-    M10>::type
-AllOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5, M6 m6, M7 m7, M8 m8, M9 m9, M10 m10) {
-  return typename internal::AllOfResult10<M1, M2, M3, M4, M5, M6, M7, M8, M9,
-      M10>::type(
-      ::testing::AllOf(m1, m2, m3, m4, m5),
-      ::testing::AllOf(m6, m7, m8, m9, m10));
-}
-
-// AnyOf(m1, m2, ..., mk) matches any value that matches any of the given
-// sub-matchers.  AnyOf is called fully qualified to prevent ADL from firing.
-
-template <typename M1, typename M2>
-inline typename internal::AnyOfResult2<M1, M2>::type
-AnyOf(M1 m1, M2 m2) {
-  return typename internal::AnyOfResult2<M1, M2>::type(
-      m1,
-      m2);
-}
-
-template <typename M1, typename M2, typename M3>
-inline typename internal::AnyOfResult3<M1, M2, M3>::type
-AnyOf(M1 m1, M2 m2, M3 m3) {
-  return typename internal::AnyOfResult3<M1, M2, M3>::type(
-      m1,
-      ::testing::AnyOf(m2, m3));
-}
-
-template <typename M1, typename M2, typename M3, typename M4>
-inline typename internal::AnyOfResult4<M1, M2, M3, M4>::type
-AnyOf(M1 m1, M2 m2, M3 m3, M4 m4) {
-  return typename internal::AnyOfResult4<M1, M2, M3, M4>::type(
-      ::testing::AnyOf(m1, m2),
-      ::testing::AnyOf(m3, m4));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5>
-inline typename internal::AnyOfResult5<M1, M2, M3, M4, M5>::type
-AnyOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5) {
-  return typename internal::AnyOfResult5<M1, M2, M3, M4, M5>::type(
-      ::testing::AnyOf(m1, m2),
-      ::testing::AnyOf(m3, m4, m5));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6>
-inline typename internal::AnyOfResult6<M1, M2, M3, M4, M5, M6>::type
-AnyOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5, M6 m6) {
-  return typename internal::AnyOfResult6<M1, M2, M3, M4, M5, M6>::type(
-      ::testing::AnyOf(m1, m2, m3),
-      ::testing::AnyOf(m4, m5, m6));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7>
-inline typename internal::AnyOfResult7<M1, M2, M3, M4, M5, M6, M7>::type
-AnyOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5, M6 m6, M7 m7) {
-  return typename internal::AnyOfResult7<M1, M2, M3, M4, M5, M6, M7>::type(
-      ::testing::AnyOf(m1, m2, m3),
-      ::testing::AnyOf(m4, m5, m6, m7));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8>
-inline typename internal::AnyOfResult8<M1, M2, M3, M4, M5, M6, M7, M8>::type
-AnyOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5, M6 m6, M7 m7, M8 m8) {
-  return typename internal::AnyOfResult8<M1, M2, M3, M4, M5, M6, M7, M8>::type(
-      ::testing::AnyOf(m1, m2, m3, m4),
-      ::testing::AnyOf(m5, m6, m7, m8));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8, typename M9>
-inline typename internal::AnyOfResult9<M1, M2, M3, M4, M5, M6, M7, M8, M9>::type
-AnyOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5, M6 m6, M7 m7, M8 m8, M9 m9) {
-  return typename internal::AnyOfResult9<M1, M2, M3, M4, M5, M6, M7, M8,
-      M9>::type(
-      ::testing::AnyOf(m1, m2, m3, m4),
-      ::testing::AnyOf(m5, m6, m7, m8, m9));
-}
-
-template <typename M1, typename M2, typename M3, typename M4, typename M5,
-    typename M6, typename M7, typename M8, typename M9, typename M10>
-inline typename internal::AnyOfResult10<M1, M2, M3, M4, M5, M6, M7, M8, M9,
-    M10>::type
-AnyOf(M1 m1, M2 m2, M3 m3, M4 m4, M5 m5, M6 m6, M7 m7, M8 m8, M9 m9, M10 m10) {
-  return typename internal::AnyOfResult10<M1, M2, M3, M4, M5, M6, M7, M8, M9,
-      M10>::type(
-      ::testing::AnyOf(m1, m2, m3, m4, m5),
-      ::testing::AnyOf(m6, m7, m8, m9, m10));
-}
-
-}  // namespace testing
-
-
-// The MATCHER* family of macros can be used in a namespace scope to
-// define custom matchers easily.
-//
-// Basic Usage
-// ===========
-//
-// The syntax
-//
-//   MATCHER(name, description_string) { statements; }
-//
-// defines a matcher with the given name that executes the statements,
-// which must return a bool to indicate if the match succeeds.  Inside
-// the statements, you can refer to the value being matched by 'arg',
-// and refer to its type by 'arg_type'.
-//
-// The description string documents what the matcher does, and is used
-// to generate the failure message when the match fails.  Since a
-// MATCHER() is usually defined in a header file shared by multiple
-// C++ source files, we require the description to be a C-string
-// literal to avoid possible side effects.  It can be empty, in which
-// case we'll use the sequence of words in the matcher name as the
-// description.
-//
-// For example:
-//
-//   MATCHER(IsEven, "") { return (arg % 2) == 0; }
-//
-// allows you to write
-//
-//   // Expects mock_foo.Bar(n) to be called where n is even.
-//   EXPECT_CALL(mock_foo, Bar(IsEven()));
-//
-// or,
-//
-//   // Verifies that the value of some_expression is even.
-//   EXPECT_THAT(some_expression, IsEven());
-//
-// If the above assertion fails, it will print something like:
-//
-//   Value of: some_expression
-//   Expected: is even
-//     Actual: 7
-//
-// where the description "is even" is automatically calculated from the
-// matcher name IsEven.
-//
-// Argument Type
-// =============
-//
-// Note that the type of the value being matched (arg_type) is
-// determined by the context in which you use the matcher and is
-// supplied to you by the compiler, so you don't need to worry about
-// declaring it (nor can you).  This allows the matcher to be
-// polymorphic.  For example, IsEven() can be used to match any type
-// where the value of "(arg % 2) == 0" can be implicitly converted to
-// a bool.  In the "Bar(IsEven())" example above, if method Bar()
-// takes an int, 'arg_type' will be int; if it takes an unsigned long,
-// 'arg_type' will be unsigned long; and so on.
-//
-// Parameterizing Matchers
-// =======================
-//
-// Sometimes you'll want to parameterize the matcher.  For that you
-// can use another macro:
-//
-//   MATCHER_P(name, param_name, description_string) { statements; }
-//
-// For example:
-//
-//   MATCHER_P(HasAbsoluteValue, value, "") { return abs(arg) == value; }
-//
-// will allow you to write:
-//
-//   EXPECT_THAT(Blah("a"), HasAbsoluteValue(n));
-//
-// which may lead to this message (assuming n is 10):
-//
-//   Value of: Blah("a")
-//   Expected: has absolute value 10
-//     Actual: -9
-//
-// Note that both the matcher description and its parameter are
-// printed, making the message human-friendly.
-//
-// In the matcher definition body, you can write 'foo_type' to
-// reference the type of a parameter named 'foo'.  For example, in the
-// body of MATCHER_P(HasAbsoluteValue, value) above, you can write
-// 'value_type' to refer to the type of 'value'.
-//
-// We also provide MATCHER_P2, MATCHER_P3, ..., up to MATCHER_P10 to
-// support multi-parameter matchers.
-//
-// Describing Parameterized Matchers
-// =================================
-//
-// The last argument to MATCHER*() is a string-typed expression.  The
-// expression can reference all of the matcher's parameters and a
-// special bool-typed variable named 'negation'.  When 'negation' is
-// false, the expression should evaluate to the matcher's description;
-// otherwise it should evaluate to the description of the negation of
-// the matcher.  For example,
-//
-//   using testing::PrintToString;
-//
-//   MATCHER_P2(InClosedRange, low, hi,
-//       std::string(negation ? "is not" : "is") + " in range [" +
-//       PrintToString(low) + ", " + PrintToString(hi) + "]") {
-//     return low <= arg && arg <= hi;
-//   }
-//   ...
-//   EXPECT_THAT(3, InClosedRange(4, 6));
-//   EXPECT_THAT(3, Not(InClosedRange(2, 4)));
-//
-// would generate two failures that contain the text:
-//
-//   Expected: is in range [4, 6]
-//   ...
-//   Expected: is not in range [2, 4]
-//
-// If you specify "" as the description, the failure message will
-// contain the sequence of words in the matcher name followed by the
-// parameter values printed as a tuple.  For example,
-//
-//   MATCHER_P2(InClosedRange, low, hi, "") { ... }
-//   ...
-//   EXPECT_THAT(3, InClosedRange(4, 6));
-//   EXPECT_THAT(3, Not(InClosedRange(2, 4)));
-//
-// would generate two failures that contain the text:
-//
-//   Expected: in closed range (4, 6)
-//   ...
-//   Expected: not (in closed range (2, 4))
-//
-// Types of Matcher Parameters
-// ===========================
-//
-// For the purpose of typing, you can view
-//
-//   MATCHER_Pk(Foo, p1, ..., pk, description_string) { ... }
-//
-// as shorthand for
-//
-//   template <typename p1_type, ..., typename pk_type>
-//   FooMatcherPk<p1_type, ..., pk_type>
-//   Foo(p1_type p1, ..., pk_type pk) { ... }
-//
-// When you write Foo(v1, ..., vk), the compiler infers the types of
-// the parameters v1, ..., and vk for you.  If you are not happy with
-// the result of the type inference, you can specify the types by
-// explicitly instantiating the template, as in Foo<long, bool>(5,
-// false).  As said earlier, you don't get to (or need to) specify
-// 'arg_type' as that's determined by the context in which the matcher
-// is used.  You can assign the result of expression Foo(p1, ..., pk)
-// to a variable of type FooMatcherPk<p1_type, ..., pk_type>.  This
-// can be useful when composing matchers.
-//
-// While you can instantiate a matcher template with reference types,
-// passing the parameters by pointer usually makes your code more
-// readable.  If, however, you still want to pass a parameter by
-// reference, be aware that in the failure message generated by the
-// matcher you will see the value of the referenced object but not its
-// address.
-//
-// Explaining Match Results
-// ========================
-//
-// Sometimes the matcher description alone isn't enough to explain why
-// the match has failed or succeeded.  For example, when expecting a
-// long string, it can be very helpful to also print the diff between
-// the expected string and the actual one.  To achieve that, you can
-// optionally stream additional information to a special variable
-// named result_listener, whose type is a pointer to class
-// MatchResultListener:
-//
-//   MATCHER_P(EqualsLongString, str, "") {
-//     if (arg == str) return true;
-//
-//     *result_listener << "the difference: "
-///                     << DiffStrings(str, arg);
-//     return false;
-//   }
-//
-// Overloading Matchers
-// ====================
-//
-// You can overload matchers with different numbers of parameters:
-//
-//   MATCHER_P(Blah, a, description_string1) { ... }
-//   MATCHER_P2(Blah, a, b, description_string2) { ... }
-//
-// Caveats
-// =======
-//
-// When defining a new matcher, you should also consider implementing
-// MatcherInterface or using MakePolymorphicMatcher().  These
-// approaches require more work than the MATCHER* macros, but also
-// give you more control on the types of the value being matched and
-// the matcher parameters, which may leads to better compiler error
-// messages when the matcher is used wrong.  They also allow
-// overloading matchers based on parameter types (as opposed to just
-// based on the number of parameters).
-//
-// MATCHER*() can only be used in a namespace scope.  The reason is
-// that C++ doesn't yet allow function-local types to be used to
-// instantiate templates.  The up-coming C++0x standard will fix this.
-// Once that's done, we'll consider supporting using MATCHER*() inside
-// a function.
-//
-// More Information
-// ================
-//
-// To learn more about using these macros, please search for 'MATCHER'
-// on https://github.com/google/googletest/blob/master/googlemock/docs/
-// CookBook.md
-
-#define MATCHER(name, description)\
-  class name##Matcher {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      gmock_Impl()\
-           {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<>()));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>());\
-    }\
-    name##Matcher() {\
-    }\
-   private:\
-  };\
-  inline name##Matcher name() {\
-    return name##Matcher();\
-  }\
-  template <typename arg_type>\
-  bool name##Matcher::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#define MATCHER_P(name, p0, description)\
-  template <typename p0##_type>\
-  class name##MatcherP {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      explicit gmock_Impl(p0##_type gmock_p0)\
-           : p0(::testing::internal::move(gmock_p0)) {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-      p0##_type const p0;\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<p0##_type>(p0)));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>(p0));\
-    }\
-    explicit name##MatcherP(p0##_type gmock_p0) : \
-        p0(::testing::internal::move(gmock_p0)) {\
-    }\
-    p0##_type const p0;\
-   private:\
-  };\
-  template <typename p0##_type>\
-  inline name##MatcherP<p0##_type> name(p0##_type p0) {\
-    return name##MatcherP<p0##_type>(p0);\
-  }\
-  template <typename p0##_type>\
-  template <typename arg_type>\
-  bool name##MatcherP<p0##_type>::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#define MATCHER_P2(name, p0, p1, description)\
-  template <typename p0##_type, typename p1##_type>\
-  class name##MatcherP2 {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1)\
-           : p0(::testing::internal::move(gmock_p0)), \
-               p1(::testing::internal::move(gmock_p1)) {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-      p0##_type const p0;\
-      p1##_type const p1;\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<p0##_type, p1##_type>(p0, p1)));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>(p0, p1));\
-    }\
-    name##MatcherP2(p0##_type gmock_p0, \
-        p1##_type gmock_p1) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)) {\
-    }\
-    p0##_type const p0;\
-    p1##_type const p1;\
-   private:\
-  };\
-  template <typename p0##_type, typename p1##_type>\
-  inline name##MatcherP2<p0##_type, p1##_type> name(p0##_type p0, \
-      p1##_type p1) {\
-    return name##MatcherP2<p0##_type, p1##_type>(p0, p1);\
-  }\
-  template <typename p0##_type, typename p1##_type>\
-  template <typename arg_type>\
-  bool name##MatcherP2<p0##_type, \
-      p1##_type>::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#define MATCHER_P3(name, p0, p1, p2, description)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type>\
-  class name##MatcherP3 {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2)\
-           : p0(::testing::internal::move(gmock_p0)), \
-               p1(::testing::internal::move(gmock_p1)), \
-               p2(::testing::internal::move(gmock_p2)) {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-      p0##_type const p0;\
-      p1##_type const p1;\
-      p2##_type const p2;\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<p0##_type, p1##_type, p2##_type>(p0, p1, \
-                    p2)));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>(p0, p1, p2));\
-    }\
-    name##MatcherP3(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)) {\
-    }\
-    p0##_type const p0;\
-    p1##_type const p1;\
-    p2##_type const p2;\
-   private:\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type>\
-  inline name##MatcherP3<p0##_type, p1##_type, p2##_type> name(p0##_type p0, \
-      p1##_type p1, p2##_type p2) {\
-    return name##MatcherP3<p0##_type, p1##_type, p2##_type>(p0, p1, p2);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type>\
-  template <typename arg_type>\
-  bool name##MatcherP3<p0##_type, p1##_type, \
-      p2##_type>::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#define MATCHER_P4(name, p0, p1, p2, p3, description)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type>\
-  class name##MatcherP4 {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3)\
-           : p0(::testing::internal::move(gmock_p0)), \
-               p1(::testing::internal::move(gmock_p1)), \
-               p2(::testing::internal::move(gmock_p2)), \
-               p3(::testing::internal::move(gmock_p3)) {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-      p0##_type const p0;\
-      p1##_type const p1;\
-      p2##_type const p2;\
-      p3##_type const p3;\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<p0##_type, p1##_type, p2##_type, \
-                    p3##_type>(p0, p1, p2, p3)));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>(p0, p1, p2, p3));\
-    }\
-    name##MatcherP4(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, \
-        p3##_type gmock_p3) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)) {\
-    }\
-    p0##_type const p0;\
-    p1##_type const p1;\
-    p2##_type const p2;\
-    p3##_type const p3;\
-   private:\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type>\
-  inline name##MatcherP4<p0##_type, p1##_type, p2##_type, \
-      p3##_type> name(p0##_type p0, p1##_type p1, p2##_type p2, \
-      p3##_type p3) {\
-    return name##MatcherP4<p0##_type, p1##_type, p2##_type, p3##_type>(p0, \
-        p1, p2, p3);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type>\
-  template <typename arg_type>\
-  bool name##MatcherP4<p0##_type, p1##_type, p2##_type, \
-      p3##_type>::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#define MATCHER_P5(name, p0, p1, p2, p3, p4, description)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type>\
-  class name##MatcherP5 {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4)\
-           : p0(::testing::internal::move(gmock_p0)), \
-               p1(::testing::internal::move(gmock_p1)), \
-               p2(::testing::internal::move(gmock_p2)), \
-               p3(::testing::internal::move(gmock_p3)), \
-               p4(::testing::internal::move(gmock_p4)) {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-      p0##_type const p0;\
-      p1##_type const p1;\
-      p2##_type const p2;\
-      p3##_type const p3;\
-      p4##_type const p4;\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<p0##_type, p1##_type, p2##_type, p3##_type, \
-                    p4##_type>(p0, p1, p2, p3, p4)));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>(p0, p1, p2, p3, p4));\
-    }\
-    name##MatcherP5(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, \
-        p4##_type gmock_p4) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)) {\
-    }\
-    p0##_type const p0;\
-    p1##_type const p1;\
-    p2##_type const p2;\
-    p3##_type const p3;\
-    p4##_type const p4;\
-   private:\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type>\
-  inline name##MatcherP5<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type> name(p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, \
-      p4##_type p4) {\
-    return name##MatcherP5<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type>(p0, p1, p2, p3, p4);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type>\
-  template <typename arg_type>\
-  bool name##MatcherP5<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type>::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#define MATCHER_P6(name, p0, p1, p2, p3, p4, p5, description)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type>\
-  class name##MatcherP6 {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5)\
-           : p0(::testing::internal::move(gmock_p0)), \
-               p1(::testing::internal::move(gmock_p1)), \
-               p2(::testing::internal::move(gmock_p2)), \
-               p3(::testing::internal::move(gmock_p3)), \
-               p4(::testing::internal::move(gmock_p4)), \
-               p5(::testing::internal::move(gmock_p5)) {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-      p0##_type const p0;\
-      p1##_type const p1;\
-      p2##_type const p2;\
-      p3##_type const p3;\
-      p4##_type const p4;\
-      p5##_type const p5;\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<p0##_type, p1##_type, p2##_type, p3##_type, \
-                    p4##_type, p5##_type>(p0, p1, p2, p3, p4, p5)));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>(p0, p1, p2, p3, p4, p5));\
-    }\
-    name##MatcherP6(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)), \
-        p5(::testing::internal::move(gmock_p5)) {\
-    }\
-    p0##_type const p0;\
-    p1##_type const p1;\
-    p2##_type const p2;\
-    p3##_type const p3;\
-    p4##_type const p4;\
-    p5##_type const p5;\
-   private:\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type>\
-  inline name##MatcherP6<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type> name(p0##_type p0, p1##_type p1, p2##_type p2, \
-      p3##_type p3, p4##_type p4, p5##_type p5) {\
-    return name##MatcherP6<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type, p5##_type>(p0, p1, p2, p3, p4, p5);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type>\
-  template <typename arg_type>\
-  bool name##MatcherP6<p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-      p5##_type>::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#define MATCHER_P7(name, p0, p1, p2, p3, p4, p5, p6, description)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type>\
-  class name##MatcherP7 {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-          p6##_type gmock_p6)\
-           : p0(::testing::internal::move(gmock_p0)), \
-               p1(::testing::internal::move(gmock_p1)), \
-               p2(::testing::internal::move(gmock_p2)), \
-               p3(::testing::internal::move(gmock_p3)), \
-               p4(::testing::internal::move(gmock_p4)), \
-               p5(::testing::internal::move(gmock_p5)), \
-               p6(::testing::internal::move(gmock_p6)) {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-      p0##_type const p0;\
-      p1##_type const p1;\
-      p2##_type const p2;\
-      p3##_type const p3;\
-      p4##_type const p4;\
-      p5##_type const p5;\
-      p6##_type const p6;\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<p0##_type, p1##_type, p2##_type, p3##_type, \
-                    p4##_type, p5##_type, p6##_type>(p0, p1, p2, p3, p4, p5, \
-                    p6)));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>(p0, p1, p2, p3, p4, p5, p6));\
-    }\
-    name##MatcherP7(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5, \
-        p6##_type gmock_p6) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)), \
-        p5(::testing::internal::move(gmock_p5)), \
-        p6(::testing::internal::move(gmock_p6)) {\
-    }\
-    p0##_type const p0;\
-    p1##_type const p1;\
-    p2##_type const p2;\
-    p3##_type const p3;\
-    p4##_type const p4;\
-    p5##_type const p5;\
-    p6##_type const p6;\
-   private:\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type>\
-  inline name##MatcherP7<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type, p6##_type> name(p0##_type p0, p1##_type p1, \
-      p2##_type p2, p3##_type p3, p4##_type p4, p5##_type p5, \
-      p6##_type p6) {\
-    return name##MatcherP7<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type, p5##_type, p6##_type>(p0, p1, p2, p3, p4, p5, p6);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type>\
-  template <typename arg_type>\
-  bool name##MatcherP7<p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-      p5##_type, p6##_type>::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#define MATCHER_P8(name, p0, p1, p2, p3, p4, p5, p6, p7, description)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type>\
-  class name##MatcherP8 {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-          p6##_type gmock_p6, p7##_type gmock_p7)\
-           : p0(::testing::internal::move(gmock_p0)), \
-               p1(::testing::internal::move(gmock_p1)), \
-               p2(::testing::internal::move(gmock_p2)), \
-               p3(::testing::internal::move(gmock_p3)), \
-               p4(::testing::internal::move(gmock_p4)), \
-               p5(::testing::internal::move(gmock_p5)), \
-               p6(::testing::internal::move(gmock_p6)), \
-               p7(::testing::internal::move(gmock_p7)) {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-      p0##_type const p0;\
-      p1##_type const p1;\
-      p2##_type const p2;\
-      p3##_type const p3;\
-      p4##_type const p4;\
-      p5##_type const p5;\
-      p6##_type const p6;\
-      p7##_type const p7;\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<p0##_type, p1##_type, p2##_type, p3##_type, \
-                    p4##_type, p5##_type, p6##_type, p7##_type>(p0, p1, p2, \
-                    p3, p4, p5, p6, p7)));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>(p0, p1, p2, p3, p4, p5, p6, p7));\
-    }\
-    name##MatcherP8(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5, p6##_type gmock_p6, \
-        p7##_type gmock_p7) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)), \
-        p5(::testing::internal::move(gmock_p5)), \
-        p6(::testing::internal::move(gmock_p6)), \
-        p7(::testing::internal::move(gmock_p7)) {\
-    }\
-    p0##_type const p0;\
-    p1##_type const p1;\
-    p2##_type const p2;\
-    p3##_type const p3;\
-    p4##_type const p4;\
-    p5##_type const p5;\
-    p6##_type const p6;\
-    p7##_type const p7;\
-   private:\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type>\
-  inline name##MatcherP8<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type, p6##_type, p7##_type> name(p0##_type p0, \
-      p1##_type p1, p2##_type p2, p3##_type p3, p4##_type p4, p5##_type p5, \
-      p6##_type p6, p7##_type p7) {\
-    return name##MatcherP8<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type, p5##_type, p6##_type, p7##_type>(p0, p1, p2, p3, p4, p5, \
-        p6, p7);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type>\
-  template <typename arg_type>\
-  bool name##MatcherP8<p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-      p5##_type, p6##_type, \
-      p7##_type>::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#define MATCHER_P9(name, p0, p1, p2, p3, p4, p5, p6, p7, p8, description)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type>\
-  class name##MatcherP9 {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-          p6##_type gmock_p6, p7##_type gmock_p7, p8##_type gmock_p8)\
-           : p0(::testing::internal::move(gmock_p0)), \
-               p1(::testing::internal::move(gmock_p1)), \
-               p2(::testing::internal::move(gmock_p2)), \
-               p3(::testing::internal::move(gmock_p3)), \
-               p4(::testing::internal::move(gmock_p4)), \
-               p5(::testing::internal::move(gmock_p5)), \
-               p6(::testing::internal::move(gmock_p6)), \
-               p7(::testing::internal::move(gmock_p7)), \
-               p8(::testing::internal::move(gmock_p8)) {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-      p0##_type const p0;\
-      p1##_type const p1;\
-      p2##_type const p2;\
-      p3##_type const p3;\
-      p4##_type const p4;\
-      p5##_type const p5;\
-      p6##_type const p6;\
-      p7##_type const p7;\
-      p8##_type const p8;\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<p0##_type, p1##_type, p2##_type, p3##_type, \
-                    p4##_type, p5##_type, p6##_type, p7##_type, \
-                    p8##_type>(p0, p1, p2, p3, p4, p5, p6, p7, p8)));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>(p0, p1, p2, p3, p4, p5, p6, p7, p8));\
-    }\
-    name##MatcherP9(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5, p6##_type gmock_p6, p7##_type gmock_p7, \
-        p8##_type gmock_p8) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)), \
-        p5(::testing::internal::move(gmock_p5)), \
-        p6(::testing::internal::move(gmock_p6)), \
-        p7(::testing::internal::move(gmock_p7)), \
-        p8(::testing::internal::move(gmock_p8)) {\
-    }\
-    p0##_type const p0;\
-    p1##_type const p1;\
-    p2##_type const p2;\
-    p3##_type const p3;\
-    p4##_type const p4;\
-    p5##_type const p5;\
-    p6##_type const p6;\
-    p7##_type const p7;\
-    p8##_type const p8;\
-   private:\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type>\
-  inline name##MatcherP9<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type, p6##_type, p7##_type, \
-      p8##_type> name(p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, \
-      p4##_type p4, p5##_type p5, p6##_type p6, p7##_type p7, \
-      p8##_type p8) {\
-    return name##MatcherP9<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type, p5##_type, p6##_type, p7##_type, p8##_type>(p0, p1, p2, \
-        p3, p4, p5, p6, p7, p8);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type>\
-  template <typename arg_type>\
-  bool name##MatcherP9<p0##_type, p1##_type, p2##_type, p3##_type, p4##_type, \
-      p5##_type, p6##_type, p7##_type, \
-      p8##_type>::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#define MATCHER_P10(name, p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, description)\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type, \
-      typename p9##_type>\
-  class name##MatcherP10 {\
-   public:\
-    template <typename arg_type>\
-    class gmock_Impl : public ::testing::MatcherInterface<\
-        GTEST_REFERENCE_TO_CONST_(arg_type)> {\
-     public:\
-      gmock_Impl(p0##_type gmock_p0, p1##_type gmock_p1, p2##_type gmock_p2, \
-          p3##_type gmock_p3, p4##_type gmock_p4, p5##_type gmock_p5, \
-          p6##_type gmock_p6, p7##_type gmock_p7, p8##_type gmock_p8, \
-          p9##_type gmock_p9)\
-           : p0(::testing::internal::move(gmock_p0)), \
-               p1(::testing::internal::move(gmock_p1)), \
-               p2(::testing::internal::move(gmock_p2)), \
-               p3(::testing::internal::move(gmock_p3)), \
-               p4(::testing::internal::move(gmock_p4)), \
-               p5(::testing::internal::move(gmock_p5)), \
-               p6(::testing::internal::move(gmock_p6)), \
-               p7(::testing::internal::move(gmock_p7)), \
-               p8(::testing::internal::move(gmock_p8)), \
-               p9(::testing::internal::move(gmock_p9)) {}\
-      virtual bool MatchAndExplain(\
-          GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-          ::testing::MatchResultListener* result_listener) const;\
-      virtual void DescribeTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(false);\
-      }\
-      virtual void DescribeNegationTo(::std::ostream* gmock_os) const {\
-        *gmock_os << FormatDescription(true);\
-      }\
-      p0##_type const p0;\
-      p1##_type const p1;\
-      p2##_type const p2;\
-      p3##_type const p3;\
-      p4##_type const p4;\
-      p5##_type const p5;\
-      p6##_type const p6;\
-      p7##_type const p7;\
-      p8##_type const p8;\
-      p9##_type const p9;\
-     private:\
-      ::std::string FormatDescription(bool negation) const {\
-        ::std::string gmock_description = (description);\
-        if (!gmock_description.empty())\
-          return gmock_description;\
-        return ::testing::internal::FormatMatcherDescription(\
-            negation, #name, \
-            ::testing::internal::UniversalTersePrintTupleFieldsToStrings(\
-                ::testing::tuple<p0##_type, p1##_type, p2##_type, p3##_type, \
-                    p4##_type, p5##_type, p6##_type, p7##_type, p8##_type, \
-                    p9##_type>(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9)));\
-      }\
-    };\
-    template <typename arg_type>\
-    operator ::testing::Matcher<arg_type>() const {\
-      return ::testing::Matcher<arg_type>(\
-          new gmock_Impl<arg_type>(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9));\
-    }\
-    name##MatcherP10(p0##_type gmock_p0, p1##_type gmock_p1, \
-        p2##_type gmock_p2, p3##_type gmock_p3, p4##_type gmock_p4, \
-        p5##_type gmock_p5, p6##_type gmock_p6, p7##_type gmock_p7, \
-        p8##_type gmock_p8, \
-        p9##_type gmock_p9) : p0(::testing::internal::move(gmock_p0)), \
-        p1(::testing::internal::move(gmock_p1)), \
-        p2(::testing::internal::move(gmock_p2)), \
-        p3(::testing::internal::move(gmock_p3)), \
-        p4(::testing::internal::move(gmock_p4)), \
-        p5(::testing::internal::move(gmock_p5)), \
-        p6(::testing::internal::move(gmock_p6)), \
-        p7(::testing::internal::move(gmock_p7)), \
-        p8(::testing::internal::move(gmock_p8)), \
-        p9(::testing::internal::move(gmock_p9)) {\
-    }\
-    p0##_type const p0;\
-    p1##_type const p1;\
-    p2##_type const p2;\
-    p3##_type const p3;\
-    p4##_type const p4;\
-    p5##_type const p5;\
-    p6##_type const p6;\
-    p7##_type const p7;\
-    p8##_type const p8;\
-    p9##_type const p9;\
-   private:\
-  };\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type, \
-      typename p9##_type>\
-  inline name##MatcherP10<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type, p6##_type, p7##_type, p8##_type, \
-      p9##_type> name(p0##_type p0, p1##_type p1, p2##_type p2, p3##_type p3, \
-      p4##_type p4, p5##_type p5, p6##_type p6, p7##_type p7, p8##_type p8, \
-      p9##_type p9) {\
-    return name##MatcherP10<p0##_type, p1##_type, p2##_type, p3##_type, \
-        p4##_type, p5##_type, p6##_type, p7##_type, p8##_type, p9##_type>(p0, \
-        p1, p2, p3, p4, p5, p6, p7, p8, p9);\
-  }\
-  template <typename p0##_type, typename p1##_type, typename p2##_type, \
-      typename p3##_type, typename p4##_type, typename p5##_type, \
-      typename p6##_type, typename p7##_type, typename p8##_type, \
-      typename p9##_type>\
-  template <typename arg_type>\
-  bool name##MatcherP10<p0##_type, p1##_type, p2##_type, p3##_type, \
-      p4##_type, p5##_type, p6##_type, p7##_type, p8##_type, \
-      p9##_type>::gmock_Impl<arg_type>::MatchAndExplain(\
-      GTEST_REFERENCE_TO_CONST_(arg_type) arg,\
-      ::testing::MatchResultListener* result_listener GTEST_ATTRIBUTE_UNUSED_)\
-          const
-
-#endif  // GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_MATCHERS_H_
-// This file was GENERATED by command:
-//     pump.py gmock-generated-nice-strict.h.pump
-// DO NOT EDIT BY HAND!!!
-
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Implements class templates NiceMock, NaggyMock, and StrictMock.
-//
-// Given a mock class MockFoo that is created using Google Mock,
-// NiceMock<MockFoo> is a subclass of MockFoo that allows
-// uninteresting calls (i.e. calls to mock methods that have no
-// EXPECT_CALL specs), NaggyMock<MockFoo> is a subclass of MockFoo
-// that prints a warning when an uninteresting call occurs, and
-// StrictMock<MockFoo> is a subclass of MockFoo that treats all
-// uninteresting calls as errors.
-//
-// Currently a mock is naggy by default, so MockFoo and
-// NaggyMock<MockFoo> behave like the same.  However, we will soon
-// switch the default behavior of mocks to be nice, as that in general
-// leads to more maintainable tests.  When that happens, MockFoo will
-// stop behaving like NaggyMock<MockFoo> and start behaving like
-// NiceMock<MockFoo>.
-//
-// NiceMock, NaggyMock, and StrictMock "inherit" the constructors of
-// their respective base class.  Therefore you can write
-// NiceMock<MockFoo>(5, "a") to construct a nice mock where MockFoo
-// has a constructor that accepts (int, const char*), for example.
-//
-// A known limitation is that NiceMock<MockFoo>, NaggyMock<MockFoo>,
-// and StrictMock<MockFoo> only works for mock methods defined using
-// the MOCK_METHOD* family of macros DIRECTLY in the MockFoo class.
-// If a mock method is defined in a base class of MockFoo, the "nice"
-// or "strict" modifier may not affect it, depending on the compiler.
-// In particular, nesting NiceMock, NaggyMock, and StrictMock is NOT
-// supported.
-
-#ifndef GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_NICE_STRICT_H_
-#define GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_NICE_STRICT_H_
-
-
-namespace testing {
-
-template <class MockClass>
-class NiceMock : public MockClass {
- public:
-  NiceMock() : MockClass() {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-#if GTEST_LANG_CXX11
-  // Ideally, we would inherit base class's constructors through a using
-  // declaration, which would preserve their visibility. However, many existing
-  // tests rely on the fact that current implementation reexports protected
-  // constructors as public. These tests would need to be cleaned up first.
-
-  // Single argument constructor is special-cased so that it can be
-  // made explicit.
-  template <typename A>
-  explicit NiceMock(A&& arg) : MockClass(std::forward<A>(arg)) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename... An>
-  NiceMock(A1&& arg1, A2&& arg2, An&&... args)
-      : MockClass(std::forward<A1>(arg1), std::forward<A2>(arg2),
-                  std::forward<An>(args)...) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-#else
-  // C++98 doesn't have variadic templates, so we have to define one
-  // for each arity.
-  template <typename A1>
-  explicit NiceMock(const A1& a1) : MockClass(a1) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-  template <typename A1, typename A2>
-  NiceMock(const A1& a1, const A2& a2) : MockClass(a1, a2) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3>
-  NiceMock(const A1& a1, const A2& a2, const A3& a3) : MockClass(a1, a2, a3) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4>
-  NiceMock(const A1& a1, const A2& a2, const A3& a3,
-      const A4& a4) : MockClass(a1, a2, a3, a4) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5>
-  NiceMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5) : MockClass(a1, a2, a3, a4, a5) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6>
-  NiceMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6) : MockClass(a1, a2, a3, a4, a5, a6) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7>
-  NiceMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7) : MockClass(a1, a2, a3, a4, a5,
-      a6, a7) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7, typename A8>
-  NiceMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7, const A8& a8) : MockClass(a1,
-      a2, a3, a4, a5, a6, a7, a8) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7, typename A8, typename A9>
-  NiceMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7, const A8& a8,
-      const A9& a9) : MockClass(a1, a2, a3, a4, a5, a6, a7, a8, a9) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7, typename A8, typename A9, typename A10>
-  NiceMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7, const A8& a8, const A9& a9,
-      const A10& a10) : MockClass(a1, a2, a3, a4, a5, a6, a7, a8, a9, a10) {
-    ::testing::Mock::AllowUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-#endif  // GTEST_LANG_CXX11
-
-  ~NiceMock() {
-    ::testing::Mock::UnregisterCallReaction(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(NiceMock);
-};
-
-template <class MockClass>
-class NaggyMock : public MockClass {
- public:
-  NaggyMock() : MockClass() {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-#if GTEST_LANG_CXX11
-  // Ideally, we would inherit base class's constructors through a using
-  // declaration, which would preserve their visibility. However, many existing
-  // tests rely on the fact that current implementation reexports protected
-  // constructors as public. These tests would need to be cleaned up first.
-
-  // Single argument constructor is special-cased so that it can be
-  // made explicit.
-  template <typename A>
-  explicit NaggyMock(A&& arg) : MockClass(std::forward<A>(arg)) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename... An>
-  NaggyMock(A1&& arg1, A2&& arg2, An&&... args)
-      : MockClass(std::forward<A1>(arg1), std::forward<A2>(arg2),
-                  std::forward<An>(args)...) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-#else
-  // C++98 doesn't have variadic templates, so we have to define one
-  // for each arity.
-  template <typename A1>
-  explicit NaggyMock(const A1& a1) : MockClass(a1) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-  template <typename A1, typename A2>
-  NaggyMock(const A1& a1, const A2& a2) : MockClass(a1, a2) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3>
-  NaggyMock(const A1& a1, const A2& a2, const A3& a3) : MockClass(a1, a2, a3) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4>
-  NaggyMock(const A1& a1, const A2& a2, const A3& a3,
-      const A4& a4) : MockClass(a1, a2, a3, a4) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5>
-  NaggyMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5) : MockClass(a1, a2, a3, a4, a5) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6>
-  NaggyMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6) : MockClass(a1, a2, a3, a4, a5, a6) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7>
-  NaggyMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7) : MockClass(a1, a2, a3, a4, a5,
-      a6, a7) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7, typename A8>
-  NaggyMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7, const A8& a8) : MockClass(a1,
-      a2, a3, a4, a5, a6, a7, a8) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7, typename A8, typename A9>
-  NaggyMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7, const A8& a8,
-      const A9& a9) : MockClass(a1, a2, a3, a4, a5, a6, a7, a8, a9) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7, typename A8, typename A9, typename A10>
-  NaggyMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7, const A8& a8, const A9& a9,
-      const A10& a10) : MockClass(a1, a2, a3, a4, a5, a6, a7, a8, a9, a10) {
-    ::testing::Mock::WarnUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-#endif  // GTEST_LANG_CXX11
-
-  ~NaggyMock() {
-    ::testing::Mock::UnregisterCallReaction(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(NaggyMock);
-};
-
-template <class MockClass>
-class StrictMock : public MockClass {
- public:
-  StrictMock() : MockClass() {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-#if GTEST_LANG_CXX11
-  // Ideally, we would inherit base class's constructors through a using
-  // declaration, which would preserve their visibility. However, many existing
-  // tests rely on the fact that current implementation reexports protected
-  // constructors as public. These tests would need to be cleaned up first.
-
-  // Single argument constructor is special-cased so that it can be
-  // made explicit.
-  template <typename A>
-  explicit StrictMock(A&& arg) : MockClass(std::forward<A>(arg)) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename... An>
-  StrictMock(A1&& arg1, A2&& arg2, An&&... args)
-      : MockClass(std::forward<A1>(arg1), std::forward<A2>(arg2),
-                  std::forward<An>(args)...) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-#else
-  // C++98 doesn't have variadic templates, so we have to define one
-  // for each arity.
-  template <typename A1>
-  explicit StrictMock(const A1& a1) : MockClass(a1) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-  template <typename A1, typename A2>
-  StrictMock(const A1& a1, const A2& a2) : MockClass(a1, a2) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3>
-  StrictMock(const A1& a1, const A2& a2, const A3& a3) : MockClass(a1, a2, a3) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4>
-  StrictMock(const A1& a1, const A2& a2, const A3& a3,
-      const A4& a4) : MockClass(a1, a2, a3, a4) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5>
-  StrictMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5) : MockClass(a1, a2, a3, a4, a5) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6>
-  StrictMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6) : MockClass(a1, a2, a3, a4, a5, a6) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7>
-  StrictMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7) : MockClass(a1, a2, a3, a4, a5,
-      a6, a7) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7, typename A8>
-  StrictMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7, const A8& a8) : MockClass(a1,
-      a2, a3, a4, a5, a6, a7, a8) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7, typename A8, typename A9>
-  StrictMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7, const A8& a8,
-      const A9& a9) : MockClass(a1, a2, a3, a4, a5, a6, a7, a8, a9) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-  template <typename A1, typename A2, typename A3, typename A4, typename A5,
-      typename A6, typename A7, typename A8, typename A9, typename A10>
-  StrictMock(const A1& a1, const A2& a2, const A3& a3, const A4& a4,
-      const A5& a5, const A6& a6, const A7& a7, const A8& a8, const A9& a9,
-      const A10& a10) : MockClass(a1, a2, a3, a4, a5, a6, a7, a8, a9, a10) {
-    ::testing::Mock::FailUninterestingCalls(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
-#endif  // GTEST_LANG_CXX11
-
-  ~StrictMock() {
-    ::testing::Mock::UnregisterCallReaction(
-        internal::ImplicitCast_<MockClass*>(this));
-  }
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(StrictMock);
-};
-
-// The following specializations catch some (relatively more common)
-// user errors of nesting nice and strict mocks.  They do NOT catch
-// all possible errors.
-
-// These specializations are declared but not defined, as NiceMock,
-// NaggyMock, and StrictMock cannot be nested.
-
-template <typename MockClass>
-class NiceMock<NiceMock<MockClass> >;
-template <typename MockClass>
-class NiceMock<NaggyMock<MockClass> >;
-template <typename MockClass>
-class NiceMock<StrictMock<MockClass> >;
-
-template <typename MockClass>
-class NaggyMock<NiceMock<MockClass> >;
-template <typename MockClass>
-class NaggyMock<NaggyMock<MockClass> >;
-template <typename MockClass>
-class NaggyMock<StrictMock<MockClass> >;
-
-template <typename MockClass>
-class StrictMock<NiceMock<MockClass> >;
-template <typename MockClass>
-class StrictMock<NaggyMock<MockClass> >;
-template <typename MockClass>
-class StrictMock<StrictMock<MockClass> >;
-
-}  // namespace testing
-
-#endif  // GMOCK_INCLUDE_GMOCK_GMOCK_GENERATED_NICE_STRICT_H_
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements some actions that depend on gmock-generated-actions.h.
-
-#ifndef GMOCK_INCLUDE_GMOCK_GMOCK_MORE_ACTIONS_H_
-#define GMOCK_INCLUDE_GMOCK_GMOCK_MORE_ACTIONS_H_
-
-#include <algorithm>
-
-
-namespace testing {
-namespace internal {
-
-// Implements the Invoke(f) action.  The template argument
-// FunctionImpl is the implementation type of f, which can be either a
-// function pointer or a functor.  Invoke(f) can be used as an
-// Action<F> as long as f's type is compatible with F (i.e. f can be
-// assigned to a tr1::function<F>).
-template <typename FunctionImpl>
-class InvokeAction {
- public:
-  // The c'tor makes a copy of function_impl (either a function
-  // pointer or a functor).
-  explicit InvokeAction(FunctionImpl function_impl)
-      : function_impl_(function_impl) {}
-
-  template <typename Result, typename ArgumentTuple>
-  Result Perform(const ArgumentTuple& args) {
-    return InvokeHelper<Result, ArgumentTuple>::Invoke(function_impl_, args);
-  }
-
- private:
-  FunctionImpl function_impl_;
-
-  GTEST_DISALLOW_ASSIGN_(InvokeAction);
-};
-
-// Implements the Invoke(object_ptr, &Class::Method) action.
-template <class Class, typename MethodPtr>
-class InvokeMethodAction {
- public:
-  InvokeMethodAction(Class* obj_ptr, MethodPtr method_ptr)
-      : method_ptr_(method_ptr), obj_ptr_(obj_ptr) {}
-
-  template <typename Result, typename ArgumentTuple>
-  Result Perform(const ArgumentTuple& args) const {
-    return InvokeHelper<Result, ArgumentTuple>::InvokeMethod(
-        obj_ptr_, method_ptr_, args);
-  }
-
- private:
-  // The order of these members matters.  Reversing the order can trigger
-  // warning C4121 in MSVC (see
-  // http://computer-programming-forum.com/7-vc.net/6fbc30265f860ad1.htm ).
-  const MethodPtr method_ptr_;
-  Class* const obj_ptr_;
-
-  GTEST_DISALLOW_ASSIGN_(InvokeMethodAction);
-};
-
-// An internal replacement for std::copy which mimics its behavior. This is
-// necessary because Visual Studio deprecates ::std::copy, issuing warning 4996.
-// However Visual Studio 2010 and later do not honor #pragmas which disable that
-// warning.
-template<typename InputIterator, typename OutputIterator>
-inline OutputIterator CopyElements(InputIterator first,
-                                   InputIterator last,
-                                   OutputIterator output) {
-  for (; first != last; ++first, ++output) {
-    *output = *first;
-  }
-  return output;
-}
-
-}  // namespace internal
-
-// Various overloads for Invoke().
-
-// Creates an action that invokes 'function_impl' with the mock
-// function's arguments.
-template <typename FunctionImpl>
-PolymorphicAction<internal::InvokeAction<FunctionImpl> > Invoke(
-    FunctionImpl function_impl) {
-  return MakePolymorphicAction(
-      internal::InvokeAction<FunctionImpl>(function_impl));
-}
-
-// Creates an action that invokes the given method on the given object
-// with the mock function's arguments.
-template <class Class, typename MethodPtr>
-PolymorphicAction<internal::InvokeMethodAction<Class, MethodPtr> > Invoke(
-    Class* obj_ptr, MethodPtr method_ptr) {
-  return MakePolymorphicAction(
-      internal::InvokeMethodAction<Class, MethodPtr>(obj_ptr, method_ptr));
-}
-
-// WithoutArgs(inner_action) can be used in a mock function with a
-// non-empty argument list to perform inner_action, which takes no
-// argument.  In other words, it adapts an action accepting no
-// argument to one that accepts (and ignores) arguments.
-template <typename InnerAction>
-inline internal::WithArgsAction<InnerAction>
-WithoutArgs(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction>(action);
-}
-
-// WithArg<k>(an_action) creates an action that passes the k-th
-// (0-based) argument of the mock function to an_action and performs
-// it.  It adapts an action accepting one argument to one that accepts
-// multiple arguments.  For convenience, we also provide
-// WithArgs<k>(an_action) (defined below) as a synonym.
-template <int k, typename InnerAction>
-inline internal::WithArgsAction<InnerAction, k>
-WithArg(const InnerAction& action) {
-  return internal::WithArgsAction<InnerAction, k>(action);
-}
-
-// The ACTION*() macros trigger warning C4100 (unreferenced formal
-// parameter) in MSVC with -W4.  Unfortunately they cannot be fixed in
-// the macro definition, as the warnings are generated when the macro
-// is expanded and macro expansion cannot contain #pragma.  Therefore
-// we suppress them here.
-#ifdef _MSC_VER
-# pragma warning(push)
-# pragma warning(disable:4100)
-#endif
-
-// Action ReturnArg<k>() returns the k-th argument of the mock function.
-ACTION_TEMPLATE(ReturnArg,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_0_VALUE_PARAMS()) {
-  return ::testing::get<k>(args);
-}
-
-// Action SaveArg<k>(pointer) saves the k-th (0-based) argument of the
-// mock function to *pointer.
-ACTION_TEMPLATE(SaveArg,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_1_VALUE_PARAMS(pointer)) {
-  *pointer = ::testing::get<k>(args);
-}
-
-// Action SaveArgPointee<k>(pointer) saves the value pointed to
-// by the k-th (0-based) argument of the mock function to *pointer.
-ACTION_TEMPLATE(SaveArgPointee,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_1_VALUE_PARAMS(pointer)) {
-  *pointer = *::testing::get<k>(args);
-}
-
-// Action SetArgReferee<k>(value) assigns 'value' to the variable
-// referenced by the k-th (0-based) argument of the mock function.
-ACTION_TEMPLATE(SetArgReferee,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_1_VALUE_PARAMS(value)) {
-  typedef typename ::testing::tuple_element<k, args_type>::type argk_type;
-  // Ensures that argument #k is a reference.  If you get a compiler
-  // error on the next line, you are using SetArgReferee<k>(value) in
-  // a mock function whose k-th (0-based) argument is not a reference.
-  GTEST_COMPILE_ASSERT_(internal::is_reference<argk_type>::value,
-                        SetArgReferee_must_be_used_with_a_reference_argument);
-  ::testing::get<k>(args) = value;
-}
-
-// Action SetArrayArgument<k>(first, last) copies the elements in
-// source range [first, last) to the array pointed to by the k-th
-// (0-based) argument, which can be either a pointer or an
-// iterator. The action does not take ownership of the elements in the
-// source range.
-ACTION_TEMPLATE(SetArrayArgument,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_2_VALUE_PARAMS(first, last)) {
-  // Visual Studio deprecates ::std::copy, so we use our own copy in that case.
-#ifdef _MSC_VER
-  internal::CopyElements(first, last, ::testing::get<k>(args));
-#else
-  ::std::copy(first, last, ::testing::get<k>(args));
-#endif
-}
-
-// Action DeleteArg<k>() deletes the k-th (0-based) argument of the mock
-// function.
-ACTION_TEMPLATE(DeleteArg,
-                HAS_1_TEMPLATE_PARAMS(int, k),
-                AND_0_VALUE_PARAMS()) {
-  delete ::testing::get<k>(args);
-}
-
-// This action returns the value pointed to by 'pointer'.
-ACTION_P(ReturnPointee, pointer) { return *pointer; }
-
-// Action Throw(exception) can be used in a mock function of any type
-// to throw the given exception.  Any copyable value can be thrown.
-#if GTEST_HAS_EXCEPTIONS
-
-// Suppresses the 'unreachable code' warning that VC generates in opt modes.
-# ifdef _MSC_VER
-#  pragma warning(push)          // Saves the current warning state.
-#  pragma warning(disable:4702)  // Temporarily disables warning 4702.
-# endif
-ACTION_P(Throw, exception) { throw exception; }
-# ifdef _MSC_VER
-#  pragma warning(pop)           // Restores the warning state.
-# endif
-
-#endif  // GTEST_HAS_EXCEPTIONS
-
-#ifdef _MSC_VER
-# pragma warning(pop)
-#endif
-
-}  // namespace testing
-
-#endif  // GMOCK_INCLUDE_GMOCK_GMOCK_MORE_ACTIONS_H_
-// Copyright 2013, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: marcus.boerger@google.com (Marcus Boerger)
-
-// Google Mock - a framework for writing C++ mock classes.
-//
-// This file implements some matchers that depend on gmock-generated-matchers.h.
-//
-// Note that tests are implemented in gmock-matchers_test.cc rather than
-// gmock-more-matchers-test.cc.
-
-#ifndef GMOCK_GMOCK_MORE_MATCHERS_H_
-#define GMOCK_GMOCK_MORE_MATCHERS_H_
-
-
-namespace testing {
-
-// Silence C4100 (unreferenced formal
-// parameter) for MSVC
-#ifdef _MSC_VER
-# pragma warning(push)
-# pragma warning(disable:4100)
-#if (_MSC_VER == 1900)
-// and silence C4800 (C4800: 'int *const ': forcing value
-// to bool 'true' or 'false') for MSVC 14
-# pragma warning(disable:4800)
-  #endif
-#endif
-
-// Defines a matcher that matches an empty container. The container must
-// support both size() and empty(), which all STL-like containers provide.
-MATCHER(IsEmpty, negation ? "isn't empty" : "is empty") {
-  if (arg.empty()) {
-    return true;
-  }
-  *result_listener << "whose size is " << arg.size();
-  return false;
-}
-
-// Define a matcher that matches a value that evaluates in boolean
-// context to true.  Useful for types that define "explicit operator
-// bool" operators and so can't be compared for equality with true
-// and false.
-MATCHER(IsTrue, negation ? "is false" : "is true") {
-  return static_cast<bool>(arg);
-}
-
-// Define a matcher that matches a value that evaluates in boolean
-// context to false.  Useful for types that define "explicit operator
-// bool" operators and so can't be compared for equality with true
-// and false.
-MATCHER(IsFalse, negation ? "is true" : "is false") {
-  return !static_cast<bool>(arg);
-}
-
-#ifdef _MSC_VER
-# pragma warning(pop)
-#endif
-
-
-}  // namespace testing
-
-#endif  // GMOCK_GMOCK_MORE_MATCHERS_H_
-
-namespace testing {
-
-// Declares Google Mock flags that we want a user to use programmatically.
-GMOCK_DECLARE_bool_(catch_leaked_mocks);
-GMOCK_DECLARE_string_(verbose);
-GMOCK_DECLARE_int32_(default_mock_behavior);
-
-// Initializes Google Mock.  This must be called before running the
-// tests.  In particular, it parses the command line for the flags
-// that Google Mock recognizes.  Whenever a Google Mock flag is seen,
-// it is removed from argv, and *argc is decremented.
-//
-// No value is returned.  Instead, the Google Mock flag variables are
-// updated.
-//
-// Since Google Test is needed for Google Mock to work, this function
-// also initializes Google Test and parses its flags, if that hasn't
-// been done.
-GTEST_API_ void InitGoogleMock(int* argc, char** argv);
-
-// This overloaded version can be used in Windows programs compiled in
-// UNICODE mode.
-GTEST_API_ void InitGoogleMock(int* argc, wchar_t** argv);
-
-}  // namespace testing
-
-#endif  // GMOCK_INCLUDE_GMOCK_GMOCK_H_
diff --git a/inference-engine/thirdparty/clDNN/common/googletest-fused/gtest/gtest.h b/inference-engine/thirdparty/clDNN/common/googletest-fused/gtest/gtest.h
deleted file mode 100644
index 5d547c09e83771..00000000000000
--- a/inference-engine/thirdparty/clDNN/common/googletest-fused/gtest/gtest.h
+++ /dev/null
@@ -1,21575 +0,0 @@
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-//
-// The Google C++ Testing and Mocking Framework (Google Test)
-//
-// This header file defines the public API for Google Test.  It should be
-// included by any test program that uses Google Test.
-//
-// IMPORTANT NOTE: Due to limitation of the C++ language, we have to
-// leave some internal implementation details in this header file.
-// They are clearly marked by comments like this:
-//
-//   // INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-//
-// Such code is NOT meant to be used by a user directly, and is subject
-// to CHANGE WITHOUT NOTICE.  Therefore DO NOT DEPEND ON IT in a user
-// program!
-//
-// Acknowledgment: Google Test borrowed the idea of automatic test
-// registration from Barthelemy Dagenais' (barthelemy@prologique.com)
-// easyUnit framework.
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_H_
-#define GTEST_INCLUDE_GTEST_GTEST_H_
-
-#include <limits>
-#include <ostream>
-#include <vector>
-
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-//
-// The Google C++ Testing and Mocking Framework (Google Test)
-//
-// This header file declares functions and macros used internally by
-// Google Test.  They are subject to change without notice.
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_INTERNAL_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_INTERNAL_H_
-
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Authors: wan@google.com (Zhanyong Wan)
-//
-// Low-level types and utilities for porting Google Test to various
-// platforms.  All macros ending with _ and symbols defined in an
-// internal namespace are subject to change without notice.  Code
-// outside Google Test MUST NOT USE THEM DIRECTLY.  Macros that don't
-// end with _ are part of Google Test's public API and can be used by
-// code outside Google Test.
-//
-// This file is fundamental to Google Test.  All other Google Test source
-// files are expected to #include this.  Therefore, it cannot #include
-// any other Google Test header.
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PORT_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PORT_H_
-
-// Environment-describing macros
-// -----------------------------
-//
-// Google Test can be used in many different environments.  Macros in
-// this section tell Google Test what kind of environment it is being
-// used in, such that Google Test can provide environment-specific
-// features and implementations.
-//
-// Google Test tries to automatically detect the properties of its
-// environment, so users usually don't need to worry about these
-// macros.  However, the automatic detection is not perfect.
-// Sometimes it's necessary for a user to define some of the following
-// macros in the build script to override Google Test's decisions.
-//
-// If the user doesn't define a macro in the list, Google Test will
-// provide a default definition.  After this header is #included, all
-// macros in this list will be defined to either 1 or 0.
-//
-// Notes to maintainers:
-//   - Each macro here is a user-tweakable knob; do not grow the list
-//     lightly.
-//   - Use #if to key off these macros.  Don't use #ifdef or "#if
-//     defined(...)", which will not work as these macros are ALWAYS
-//     defined.
-//
-//   GTEST_HAS_CLONE          - Define it to 1/0 to indicate that clone(2)
-//                              is/isn't available.
-//   GTEST_HAS_EXCEPTIONS     - Define it to 1/0 to indicate that exceptions
-//                              are enabled.
-//   GTEST_HAS_GLOBAL_STRING  - Define it to 1/0 to indicate that ::string
-//                              is/isn't available
-//   GTEST_HAS_GLOBAL_WSTRING - Define it to 1/0 to indicate that ::wstring
-//                              is/isn't available
-//   GTEST_HAS_POSIX_RE       - Define it to 1/0 to indicate that POSIX regular
-//                              expressions are/aren't available.
-//   GTEST_HAS_PTHREAD        - Define it to 1/0 to indicate that <pthread.h>
-//                              is/isn't available.
-//   GTEST_HAS_RTTI           - Define it to 1/0 to indicate that RTTI is/isn't
-//                              enabled.
-//   GTEST_HAS_STD_WSTRING    - Define it to 1/0 to indicate that
-//                              std::wstring does/doesn't work (Google Test can
-//                              be used where std::wstring is unavailable).
-//   GTEST_HAS_TR1_TUPLE      - Define it to 1/0 to indicate tr1::tuple
-//                              is/isn't available.
-//   GTEST_HAS_SEH            - Define it to 1/0 to indicate whether the
-//                              compiler supports Microsoft's "Structured
-//                              Exception Handling".
-//   GTEST_HAS_STREAM_REDIRECTION
-//                            - Define it to 1/0 to indicate whether the
-//                              platform supports I/O stream redirection using
-//                              dup() and dup2().
-//   GTEST_USE_OWN_TR1_TUPLE  - Define it to 1/0 to indicate whether Google
-//                              Test's own tr1 tuple implementation should be
-//                              used.  Unused when the user sets
-//                              GTEST_HAS_TR1_TUPLE to 0.
-//   GTEST_LANG_CXX11         - Define it to 1/0 to indicate that Google Test
-//                              is building in C++11/C++98 mode.
-//   GTEST_LINKED_AS_SHARED_LIBRARY
-//                            - Define to 1 when compiling tests that use
-//                              Google Test as a shared library (known as
-//                              DLL on Windows).
-//   GTEST_CREATE_SHARED_LIBRARY
-//                            - Define to 1 when compiling Google Test itself
-//                              as a shared library.
-//   GTEST_DEFAULT_DEATH_TEST_STYLE
-//                            - The default value of --gtest_death_test_style.
-//                              The legacy default has been "fast" in the open
-//                              source version since 2008. The recommended value
-//                              is "threadsafe", and can be set in
-//                              custom/gtest-port.h.
-
-// Platform-indicating macros
-// --------------------------
-//
-// Macros indicating the platform on which Google Test is being used
-// (a macro is defined to 1 if compiled on the given platform;
-// otherwise UNDEFINED -- it's never defined to 0.).  Google Test
-// defines these macros automatically.  Code outside Google Test MUST
-// NOT define them.
-//
-//   GTEST_OS_AIX      - IBM AIX
-//   GTEST_OS_CYGWIN   - Cygwin
-//   GTEST_OS_FREEBSD  - FreeBSD
-//   GTEST_OS_FUCHSIA  - Fuchsia
-//   GTEST_OS_HPUX     - HP-UX
-//   GTEST_OS_LINUX    - Linux
-//     GTEST_OS_LINUX_ANDROID - Google Android
-//   GTEST_OS_MAC      - Mac OS X
-//     GTEST_OS_IOS    - iOS
-//   GTEST_OS_NACL     - Google Native Client (NaCl)
-//   GTEST_OS_NETBSD   - NetBSD
-//   GTEST_OS_OPENBSD  - OpenBSD
-//   GTEST_OS_QNX      - QNX
-//   GTEST_OS_SOLARIS  - Sun Solaris
-//   GTEST_OS_SYMBIAN  - Symbian
-//   GTEST_OS_WINDOWS  - Windows (Desktop, MinGW, or Mobile)
-//     GTEST_OS_WINDOWS_DESKTOP  - Windows Desktop
-//     GTEST_OS_WINDOWS_MINGW    - MinGW
-//     GTEST_OS_WINDOWS_MOBILE   - Windows Mobile
-//     GTEST_OS_WINDOWS_PHONE    - Windows Phone
-//     GTEST_OS_WINDOWS_RT       - Windows Store App/WinRT
-//   GTEST_OS_ZOS      - z/OS
-//
-// Among the platforms, Cygwin, Linux, Max OS X, and Windows have the
-// most stable support.  Since core members of the Google Test project
-// don't have access to other platforms, support for them may be less
-// stable.  If you notice any problems on your platform, please notify
-// googletestframework@googlegroups.com (patches for fixing them are
-// even more welcome!).
-//
-// It is possible that none of the GTEST_OS_* macros are defined.
-
-// Feature-indicating macros
-// -------------------------
-//
-// Macros indicating which Google Test features are available (a macro
-// is defined to 1 if the corresponding feature is supported;
-// otherwise UNDEFINED -- it's never defined to 0.).  Google Test
-// defines these macros automatically.  Code outside Google Test MUST
-// NOT define them.
-//
-// These macros are public so that portable tests can be written.
-// Such tests typically surround code using a feature with an #if
-// which controls that code.  For example:
-//
-// #if GTEST_HAS_DEATH_TEST
-//   EXPECT_DEATH(DoSomethingDeadly());
-// #endif
-//
-//   GTEST_HAS_COMBINE      - the Combine() function (for value-parameterized
-//                            tests)
-//   GTEST_HAS_DEATH_TEST   - death tests
-//   GTEST_HAS_TYPED_TEST   - typed tests
-//   GTEST_HAS_TYPED_TEST_P - type-parameterized tests
-//   GTEST_IS_THREADSAFE    - Google Test is thread-safe.
-//   GTEST_USES_POSIX_RE    - enhanced POSIX regex is used. Do not confuse with
-//                            GTEST_HAS_POSIX_RE (see above) which users can
-//                            define themselves.
-//   GTEST_USES_SIMPLE_RE   - our own simple regex is used;
-//                            the above RE\b(s) are mutually exclusive.
-//   GTEST_CAN_COMPARE_NULL - accepts untyped NULL in EXPECT_EQ().
-
-// Misc public macros
-// ------------------
-//
-//   GTEST_FLAG(flag_name)  - references the variable corresponding to
-//                            the given Google Test flag.
-
-// Internal utilities
-// ------------------
-//
-// The following macros and utilities are for Google Test's INTERNAL
-// use only.  Code outside Google Test MUST NOT USE THEM DIRECTLY.
-//
-// Macros for basic C++ coding:
-//   GTEST_AMBIGUOUS_ELSE_BLOCKER_ - for disabling a gcc warning.
-//   GTEST_ATTRIBUTE_UNUSED_  - declares that a class' instances or a
-//                              variable don't have to be used.
-//   GTEST_DISALLOW_ASSIGN_   - disables operator=.
-//   GTEST_DISALLOW_COPY_AND_ASSIGN_ - disables copy ctor and operator=.
-//   GTEST_MUST_USE_RESULT_   - declares that a function's result must be used.
-//   GTEST_INTENTIONAL_CONST_COND_PUSH_ - start code section where MSVC C4127 is
-//                                        suppressed (constant conditional).
-//   GTEST_INTENTIONAL_CONST_COND_POP_  - finish code section where MSVC C4127
-//                                        is suppressed.
-//
-// C++11 feature wrappers:
-//
-//   testing::internal::forward - portability wrapper for std::forward.
-//   testing::internal::move  - portability wrapper for std::move.
-//
-// Synchronization:
-//   Mutex, MutexLock, ThreadLocal, GetThreadCount()
-//                            - synchronization primitives.
-//
-// Template meta programming:
-//   is_pointer     - as in TR1; needed on Symbian and IBM XL C/C++ only.
-//   IteratorTraits - partial implementation of std::iterator_traits, which
-//                    is not available in libCstd when compiled with Sun C++.
-//
-// Smart pointers:
-//   scoped_ptr     - as in TR2.
-//
-// Regular expressions:
-//   RE             - a simple regular expression class using the POSIX
-//                    Extended Regular Expression syntax on UNIX-like platforms
-//                    or a reduced regular exception syntax on other
-//                    platforms, including Windows.
-// Logging:
-//   GTEST_LOG_()   - logs messages at the specified severity level.
-//   LogToStderr()  - directs all log messages to stderr.
-//   FlushInfoLog() - flushes informational log messages.
-//
-// Stdout and stderr capturing:
-//   CaptureStdout()     - starts capturing stdout.
-//   GetCapturedStdout() - stops capturing stdout and returns the captured
-//                         string.
-//   CaptureStderr()     - starts capturing stderr.
-//   GetCapturedStderr() - stops capturing stderr and returns the captured
-//                         string.
-//
-// Integer types:
-//   TypeWithSize   - maps an integer to a int type.
-//   Int32, UInt32, Int64, UInt64, TimeInMillis
-//                  - integers of known sizes.
-//   BiggestInt     - the biggest signed integer type.
-//
-// Command-line utilities:
-//   GTEST_DECLARE_*()  - declares a flag.
-//   GTEST_DEFINE_*()   - defines a flag.
-//   GetInjectableArgvs() - returns the command line as a vector of strings.
-//
-// Environment variable utilities:
-//   GetEnv()             - gets the value of an environment variable.
-//   BoolFromGTestEnv()   - parses a bool environment variable.
-//   Int32FromGTestEnv()  - parses an Int32 environment variable.
-//   StringFromGTestEnv() - parses a string environment variable.
-
-#include <ctype.h>   // for isspace, etc
-#include <stddef.h>  // for ptrdiff_t
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-#ifndef _WIN32_WCE
-# include <sys/types.h>
-# include <sys/stat.h>
-#endif  // !_WIN32_WCE
-
-#if defined __APPLE__
-# include <AvailabilityMacros.h>
-# include <TargetConditionals.h>
-#endif
-
-// Brings in the definition of HAS_GLOBAL_STRING.  This must be done
-// BEFORE we test HAS_GLOBAL_STRING.
-#include <string>  // NOLINT
-#include <algorithm>  // NOLINT
-#include <iostream>  // NOLINT
-#include <sstream>  // NOLINT
-#include <utility>
-#include <vector>  // NOLINT
-
-// Copyright 2015, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// The Google C++ Testing and Mocking Framework (Google Test)
-//
-// This header file defines the GTEST_OS_* macro.
-// It is separate from gtest-port.h so that custom/gtest-port.h can include it.
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PORT_ARCH_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PORT_ARCH_H_
-
-// Determines the platform on which Google Test is compiled.
-#ifdef __CYGWIN__
-# define GTEST_OS_CYGWIN 1
-#elif defined __SYMBIAN32__
-# define GTEST_OS_SYMBIAN 1
-#elif defined _WIN32
-# define GTEST_OS_WINDOWS 1
-# ifdef _WIN32_WCE
-#  define GTEST_OS_WINDOWS_MOBILE 1
-# elif defined(__MINGW__) || defined(__MINGW32__)
-#  define GTEST_OS_WINDOWS_MINGW 1
-# elif defined(WINAPI_FAMILY)
-#  include <winapifamily.h>
-#  if WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP)
-#   define GTEST_OS_WINDOWS_DESKTOP 1
-#  elif WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_PHONE_APP)
-#   define GTEST_OS_WINDOWS_PHONE 1
-#  elif WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
-#   define GTEST_OS_WINDOWS_RT 1
-#  elif WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_TV_TITLE)
-#   define GTEST_OS_WINDOWS_PHONE 1
-#   define GTEST_OS_WINDOWS_TV_TITLE 1
-#  else
-    // WINAPI_FAMILY defined but no known partition matched.
-    // Default to desktop.
-#   define GTEST_OS_WINDOWS_DESKTOP 1
-#  endif
-# else
-#  define GTEST_OS_WINDOWS_DESKTOP 1
-# endif  // _WIN32_WCE
-#elif defined __APPLE__
-# define GTEST_OS_MAC 1
-# if TARGET_OS_IPHONE
-#  define GTEST_OS_IOS 1
-# endif
-#elif defined __FreeBSD__
-# define GTEST_OS_FREEBSD 1
-#elif defined __Fuchsia__
-# define GTEST_OS_FUCHSIA 1
-#elif defined __linux__
-# define GTEST_OS_LINUX 1
-# if defined __ANDROID__
-#  define GTEST_OS_LINUX_ANDROID 1
-# endif
-#elif defined __MVS__
-# define GTEST_OS_ZOS 1
-#elif defined(__sun) && defined(__SVR4)
-# define GTEST_OS_SOLARIS 1
-#elif defined(_AIX)
-# define GTEST_OS_AIX 1
-#elif defined(__hpux)
-# define GTEST_OS_HPUX 1
-#elif defined __native_client__
-# define GTEST_OS_NACL 1
-#elif defined __NetBSD__
-# define GTEST_OS_NETBSD 1
-#elif defined __OpenBSD__
-# define GTEST_OS_OPENBSD 1
-#elif defined __QNX__
-# define GTEST_OS_QNX 1
-#endif  // __CYGWIN__
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PORT_ARCH_H_
-// Copyright 2015, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Injection point for custom user configurations.
-// The following macros can be defined:
-//
-//   Flag related macros:
-//     GTEST_FLAG(flag_name)
-//     GTEST_USE_OWN_FLAGFILE_FLAG_  - Define to 0 when the system provides its
-//                                     own flagfile flag parsing.
-//     GTEST_DECLARE_bool_(name)
-//     GTEST_DECLARE_int32_(name)
-//     GTEST_DECLARE_string_(name)
-//     GTEST_DEFINE_bool_(name, default_val, doc)
-//     GTEST_DEFINE_int32_(name, default_val, doc)
-//     GTEST_DEFINE_string_(name, default_val, doc)
-//
-//   Logging:
-//     GTEST_LOG_(severity)
-//     GTEST_CHECK_(condition)
-//     Functions LogToStderr() and FlushInfoLog() have to be provided too.
-//
-//   Threading:
-//     GTEST_HAS_NOTIFICATION_ - Enabled if Notification is already provided.
-//     GTEST_HAS_MUTEX_AND_THREAD_LOCAL_ - Enabled if Mutex and ThreadLocal are
-//                                         already provided.
-//     Must also provide GTEST_DECLARE_STATIC_MUTEX_(mutex) and
-//     GTEST_DEFINE_STATIC_MUTEX_(mutex)
-//
-//     GTEST_EXCLUSIVE_LOCK_REQUIRED_(locks)
-//     GTEST_LOCK_EXCLUDED_(locks)
-//
-//   Underlying library support features:
-//     GTEST_HAS_CXXABI_H_
-//
-//   Exporting API symbols:
-//     GTEST_API_ - Specifier for exported symbols.
-//
-// ** Custom implementation starts here **
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_CUSTOM_GTEST_PORT_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_CUSTOM_GTEST_PORT_H_
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_CUSTOM_GTEST_PORT_H_
-
-#if !defined(GTEST_DEV_EMAIL_)
-# define GTEST_DEV_EMAIL_ "googletestframework@@googlegroups.com"
-# define GTEST_FLAG_PREFIX_ "gtest_"
-# define GTEST_FLAG_PREFIX_DASH_ "gtest-"
-# define GTEST_FLAG_PREFIX_UPPER_ "GTEST_"
-# define GTEST_NAME_ "Google Test"
-# define GTEST_PROJECT_URL_ "https://github.com/google/googletest/"
-#endif  // !defined(GTEST_DEV_EMAIL_)
-
-#if !defined(GTEST_INIT_GOOGLE_TEST_NAME_)
-# define GTEST_INIT_GOOGLE_TEST_NAME_ "testing::InitGoogleTest"
-#endif  // !defined(GTEST_INIT_GOOGLE_TEST_NAME_)
-
-// Determines the version of gcc that is used to compile this.
-#ifdef __GNUC__
-// 40302 means version 4.3.2.
-# define GTEST_GCC_VER_ \
-    (__GNUC__*10000 + __GNUC_MINOR__*100 + __GNUC_PATCHLEVEL__)
-#endif  // __GNUC__
-
-// Macros for disabling Microsoft Visual C++ warnings.
-//
-//   GTEST_DISABLE_MSC_WARNINGS_PUSH_(4800 4385)
-//   /* code that triggers warnings C4800 and C4385 */
-//   GTEST_DISABLE_MSC_WARNINGS_POP_()
-#if _MSC_VER >= 1400
-# define GTEST_DISABLE_MSC_WARNINGS_PUSH_(warnings) \
-    __pragma(warning(push))                        \
-    __pragma(warning(disable: warnings))
-# define GTEST_DISABLE_MSC_WARNINGS_POP_()          \
-    __pragma(warning(pop))
-#else
-// Older versions of MSVC don't have __pragma.
-# define GTEST_DISABLE_MSC_WARNINGS_PUSH_(warnings)
-# define GTEST_DISABLE_MSC_WARNINGS_POP_()
-#endif
-
-#ifndef GTEST_LANG_CXX11
-// gcc and clang define __GXX_EXPERIMENTAL_CXX0X__ when
-// -std={c,gnu}++{0x,11} is passed.  The C++11 standard specifies a
-// value for __cplusplus, and recent versions of clang, gcc, and
-// probably other compilers set that too in C++11 mode.
-# if __GXX_EXPERIMENTAL_CXX0X__ || __cplusplus >= 201103L || _MSC_VER >= 1900
-// Compiling in at least C++11 mode.
-#  define GTEST_LANG_CXX11 1
-# else
-#  define GTEST_LANG_CXX11 0
-# endif
-#endif
-
-// Distinct from C++11 language support, some environments don't provide
-// proper C++11 library support. Notably, it's possible to build in
-// C++11 mode when targeting Mac OS X 10.6, which has an old libstdc++
-// with no C++11 support.
-//
-// libstdc++ has sufficient C++11 support as of GCC 4.6.0, __GLIBCXX__
-// 20110325, but maintenance releases in the 4.4 and 4.5 series followed
-// this date, so check for those versions by their date stamps.
-// https://gcc.gnu.org/onlinedocs/libstdc++/manual/abi.html#abi.versioning
-#if GTEST_LANG_CXX11 && \
-    (!defined(__GLIBCXX__) || ( \
-        __GLIBCXX__ >= 20110325ul &&  /* GCC >= 4.6.0 */ \
-        /* Blacklist of patch releases of older branches: */ \
-        __GLIBCXX__ != 20110416ul &&  /* GCC 4.4.6 */ \
-        __GLIBCXX__ != 20120313ul &&  /* GCC 4.4.7 */ \
-        __GLIBCXX__ != 20110428ul &&  /* GCC 4.5.3 */ \
-        __GLIBCXX__ != 20120702ul))   /* GCC 4.5.4 */
-# define GTEST_STDLIB_CXX11 1
-#endif
-
-// Only use C++11 library features if the library provides them.
-#if GTEST_STDLIB_CXX11
-# define GTEST_HAS_STD_BEGIN_AND_END_ 1
-# define GTEST_HAS_STD_FORWARD_LIST_ 1
-# if !defined(_MSC_VER) || (_MSC_FULL_VER >= 190023824)
-// works only with VS2015U2 and better
-#   define GTEST_HAS_STD_FUNCTION_ 1
-# endif
-# define GTEST_HAS_STD_INITIALIZER_LIST_ 1
-# define GTEST_HAS_STD_MOVE_ 1
-# define GTEST_HAS_STD_UNIQUE_PTR_ 1
-# define GTEST_HAS_STD_SHARED_PTR_ 1
-# define GTEST_HAS_UNORDERED_MAP_ 1
-# define GTEST_HAS_UNORDERED_SET_ 1
-#endif
-
-// C++11 specifies that <tuple> provides std::tuple.
-// Some platforms still might not have it, however.
-#if GTEST_LANG_CXX11
-# define GTEST_HAS_STD_TUPLE_ 1
-# if defined(__clang__)
-// Inspired by http://clang.llvm.org/docs/LanguageExtensions.html#__has_include
-#  if defined(__has_include) && !__has_include(<tuple>)
-#   undef GTEST_HAS_STD_TUPLE_
-#  endif
-# elif defined(_MSC_VER)
-// Inspired by boost/config/stdlib/dinkumware.hpp
-#  if defined(_CPPLIB_VER) && _CPPLIB_VER < 520
-#   undef GTEST_HAS_STD_TUPLE_
-#  endif
-# elif defined(__GLIBCXX__)
-// Inspired by boost/config/stdlib/libstdcpp3.hpp,
-// http://gcc.gnu.org/gcc-4.2/changes.html and
-// http://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#manual.intro.status.standard.200x
-#  if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 2)
-#   undef GTEST_HAS_STD_TUPLE_
-#  endif
-# endif
-#endif
-
-// Brings in definitions for functions used in the testing::internal::posix
-// namespace (read, write, close, chdir, isatty, stat). We do not currently
-// use them on Windows Mobile.
-#if GTEST_OS_WINDOWS
-# if !GTEST_OS_WINDOWS_MOBILE
-#  include <direct.h>
-#  include <io.h>
-# endif
-// In order to avoid having to include <windows.h>, use forward declaration
-#if GTEST_OS_WINDOWS_MINGW && !defined(__MINGW64_VERSION_MAJOR)
-// MinGW defined _CRITICAL_SECTION and _RTL_CRITICAL_SECTION as two
-// separate (equivalent) structs, instead of using typedef
-typedef struct _CRITICAL_SECTION GTEST_CRITICAL_SECTION;
-#else
-// Assume CRITICAL_SECTION is a typedef of _RTL_CRITICAL_SECTION.
-// This assumption is verified by
-// WindowsTypesTest.CRITICAL_SECTIONIs_RTL_CRITICAL_SECTION.
-typedef struct _RTL_CRITICAL_SECTION GTEST_CRITICAL_SECTION;
-#endif
-#else
-// This assumes that non-Windows OSes provide unistd.h. For OSes where this
-// is not the case, we need to include headers that provide the functions
-// mentioned above.
-# include <unistd.h>
-# include <strings.h>
-#endif  // GTEST_OS_WINDOWS
-
-#if GTEST_OS_LINUX_ANDROID
-// Used to define __ANDROID_API__ matching the target NDK API level.
-#  include <android/api-level.h>  // NOLINT
-#endif
-
-// Defines this to true iff Google Test can use POSIX regular expressions.
-#ifndef GTEST_HAS_POSIX_RE
-# if GTEST_OS_LINUX_ANDROID
-// On Android, <regex.h> is only available starting with Gingerbread.
-#  define GTEST_HAS_POSIX_RE (__ANDROID_API__ >= 9)
-# else
-#  define GTEST_HAS_POSIX_RE (!GTEST_OS_WINDOWS)
-# endif
-#endif
-
-#if GTEST_USES_PCRE
-// The appropriate headers have already been included.
-
-#elif GTEST_HAS_POSIX_RE
-
-// On some platforms, <regex.h> needs someone to define size_t, and
-// won't compile otherwise.  We can #include it here as we already
-// included <stdlib.h>, which is guaranteed to define size_t through
-// <stddef.h>.
-# include <regex.h>  // NOLINT
-
-# define GTEST_USES_POSIX_RE 1
-
-#elif GTEST_OS_WINDOWS
-
-// <regex.h> is not available on Windows.  Use our own simple regex
-// implementation instead.
-# define GTEST_USES_SIMPLE_RE 1
-
-#else
-
-// <regex.h> may not be available on this platform.  Use our own
-// simple regex implementation instead.
-# define GTEST_USES_SIMPLE_RE 1
-
-#endif  // GTEST_USES_PCRE
-
-#ifndef GTEST_HAS_EXCEPTIONS
-// The user didn't tell us whether exceptions are enabled, so we need
-// to figure it out.
-# if defined(_MSC_VER) && defined(_CPPUNWIND)
-// MSVC defines _CPPUNWIND to 1 iff exceptions are enabled.
-#  define GTEST_HAS_EXCEPTIONS 1
-# elif defined(__BORLANDC__)
-// C++Builder's implementation of the STL uses the _HAS_EXCEPTIONS
-// macro to enable exceptions, so we'll do the same.
-// Assumes that exceptions are enabled by default.
-#  ifndef _HAS_EXCEPTIONS
-#   define _HAS_EXCEPTIONS 1
-#  endif  // _HAS_EXCEPTIONS
-#  define GTEST_HAS_EXCEPTIONS _HAS_EXCEPTIONS
-# elif defined(__clang__)
-// clang defines __EXCEPTIONS iff exceptions are enabled before clang 220714,
-// but iff cleanups are enabled after that. In Obj-C++ files, there can be
-// cleanups for ObjC exceptions which also need cleanups, even if C++ exceptions
-// are disabled. clang has __has_feature(cxx_exceptions) which checks for C++
-// exceptions starting at clang r206352, but which checked for cleanups prior to
-// that. To reliably check for C++ exception availability with clang, check for
-// __EXCEPTIONS && __has_feature(cxx_exceptions).
-#  define GTEST_HAS_EXCEPTIONS (__EXCEPTIONS && __has_feature(cxx_exceptions))
-# elif defined(__GNUC__) && __EXCEPTIONS
-// gcc defines __EXCEPTIONS to 1 iff exceptions are enabled.
-#  define GTEST_HAS_EXCEPTIONS 1
-# elif defined(__SUNPRO_CC)
-// Sun Pro CC supports exceptions.  However, there is no compile-time way of
-// detecting whether they are enabled or not.  Therefore, we assume that
-// they are enabled unless the user tells us otherwise.
-#  define GTEST_HAS_EXCEPTIONS 1
-# elif defined(__IBMCPP__) && __EXCEPTIONS
-// xlC defines __EXCEPTIONS to 1 iff exceptions are enabled.
-#  define GTEST_HAS_EXCEPTIONS 1
-# elif defined(__HP_aCC)
-// Exception handling is in effect by default in HP aCC compiler. It has to
-// be turned of by +noeh compiler option if desired.
-#  define GTEST_HAS_EXCEPTIONS 1
-# else
-// For other compilers, we assume exceptions are disabled to be
-// conservative.
-#  define GTEST_HAS_EXCEPTIONS 0
-# endif  // defined(_MSC_VER) || defined(__BORLANDC__)
-#endif  // GTEST_HAS_EXCEPTIONS
-
-#if !defined(GTEST_HAS_STD_STRING)
-// Even though we don't use this macro any longer, we keep it in case
-// some clients still depend on it.
-# define GTEST_HAS_STD_STRING 1
-#elif !GTEST_HAS_STD_STRING
-// The user told us that ::std::string isn't available.
-# error "::std::string isn't available."
-#endif  // !defined(GTEST_HAS_STD_STRING)
-
-#ifndef GTEST_HAS_GLOBAL_STRING
-# define GTEST_HAS_GLOBAL_STRING 0
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-#ifndef GTEST_HAS_STD_WSTRING
-// The user didn't tell us whether ::std::wstring is available, so we need
-// to figure it out.
-// TODO(wan@google.com): uses autoconf to detect whether ::std::wstring
-//   is available.
-
-// Cygwin 1.7 and below doesn't support ::std::wstring.
-// Solaris' libc++ doesn't support it either.  Android has
-// no support for it at least as recent as Froyo (2.2).
-# define GTEST_HAS_STD_WSTRING \
-    (!(GTEST_OS_LINUX_ANDROID || GTEST_OS_CYGWIN || GTEST_OS_SOLARIS))
-
-#endif  // GTEST_HAS_STD_WSTRING
-
-#ifndef GTEST_HAS_GLOBAL_WSTRING
-// The user didn't tell us whether ::wstring is available, so we need
-// to figure it out.
-# define GTEST_HAS_GLOBAL_WSTRING \
-    (GTEST_HAS_STD_WSTRING && GTEST_HAS_GLOBAL_STRING)
-#endif  // GTEST_HAS_GLOBAL_WSTRING
-
-// Determines whether RTTI is available.
-#ifndef GTEST_HAS_RTTI
-// The user didn't tell us whether RTTI is enabled, so we need to
-// figure it out.
-
-# ifdef _MSC_VER
-
-#  ifdef _CPPRTTI  // MSVC defines this macro iff RTTI is enabled.
-#   define GTEST_HAS_RTTI 1
-#  else
-#   define GTEST_HAS_RTTI 0
-#  endif
-
-// Starting with version 4.3.2, gcc defines __GXX_RTTI iff RTTI is enabled.
-# elif defined(__GNUC__) && (GTEST_GCC_VER_ >= 40302)
-
-#  ifdef __GXX_RTTI
-// When building against STLport with the Android NDK and with
-// -frtti -fno-exceptions, the build fails at link time with undefined
-// references to __cxa_bad_typeid. Note sure if STL or toolchain bug,
-// so disable RTTI when detected.
-#   if GTEST_OS_LINUX_ANDROID && defined(_STLPORT_MAJOR) && \
-       !defined(__EXCEPTIONS)
-#    define GTEST_HAS_RTTI 0
-#   else
-#    define GTEST_HAS_RTTI 1
-#   endif  // GTEST_OS_LINUX_ANDROID && __STLPORT_MAJOR && !__EXCEPTIONS
-#  else
-#   define GTEST_HAS_RTTI 0
-#  endif  // __GXX_RTTI
-
-// Clang defines __GXX_RTTI starting with version 3.0, but its manual recommends
-// using has_feature instead. has_feature(cxx_rtti) is supported since 2.7, the
-// first version with C++ support.
-# elif defined(__clang__)
-
-#  define GTEST_HAS_RTTI __has_feature(cxx_rtti)
-
-// Starting with version 9.0 IBM Visual Age defines __RTTI_ALL__ to 1 if
-// both the typeid and dynamic_cast features are present.
-# elif defined(__IBMCPP__) && (__IBMCPP__ >= 900)
-
-#  ifdef __RTTI_ALL__
-#   define GTEST_HAS_RTTI 1
-#  else
-#   define GTEST_HAS_RTTI 0
-#  endif
-
-# else
-
-// For all other compilers, we assume RTTI is enabled.
-#  define GTEST_HAS_RTTI 1
-
-# endif  // _MSC_VER
-
-#endif  // GTEST_HAS_RTTI
-
-// It's this header's responsibility to #include <typeinfo> when RTTI
-// is enabled.
-#if GTEST_HAS_RTTI
-# include <typeinfo>
-#endif
-
-// Determines whether Google Test can use the pthreads library.
-#ifndef GTEST_HAS_PTHREAD
-// The user didn't tell us explicitly, so we make reasonable assumptions about
-// which platforms have pthreads support.
-//
-// To disable threading support in Google Test, add -DGTEST_HAS_PTHREAD=0
-// to your compiler flags.
-#define GTEST_HAS_PTHREAD                                             \
-  (GTEST_OS_LINUX || GTEST_OS_MAC || GTEST_OS_HPUX || GTEST_OS_QNX || \
-   GTEST_OS_FREEBSD || GTEST_OS_NACL || GTEST_OS_NETBSD || GTEST_OS_FUCHSIA)
-#endif  // GTEST_HAS_PTHREAD
-
-#if GTEST_HAS_PTHREAD
-// gtest-port.h guarantees to #include <pthread.h> when GTEST_HAS_PTHREAD is
-// true.
-# include <pthread.h>  // NOLINT
-
-// For timespec and nanosleep, used below.
-# include <time.h>  // NOLINT
-#endif
-
-// Determines if hash_map/hash_set are available.
-// Only used for testing against those containers.
-#if !defined(GTEST_HAS_HASH_MAP_)
-# if defined(_MSC_VER) && (_MSC_VER < 1900)
-#  define GTEST_HAS_HASH_MAP_ 1  // Indicates that hash_map is available.
-#  define GTEST_HAS_HASH_SET_ 1  // Indicates that hash_set is available.
-# endif  // _MSC_VER
-#endif  // !defined(GTEST_HAS_HASH_MAP_)
-
-// Determines whether Google Test can use tr1/tuple.  You can define
-// this macro to 0 to prevent Google Test from using tuple (any
-// feature depending on tuple with be disabled in this mode).
-#ifndef GTEST_HAS_TR1_TUPLE
-# if GTEST_OS_LINUX_ANDROID && defined(_STLPORT_MAJOR)
-// STLport, provided with the Android NDK, has neither <tr1/tuple> or <tuple>.
-#  define GTEST_HAS_TR1_TUPLE 0
-# elif defined(_MSC_VER) && (_MSC_VER >= 1910)
-// Prevent `warning C4996: 'std::tr1': warning STL4002:
-// The non-Standard std::tr1 namespace and TR1-only machinery
-// are deprecated and will be REMOVED.`
-#  define GTEST_HAS_TR1_TUPLE 0
-# elif GTEST_LANG_CXX11 && defined(_LIBCPP_VERSION)
-// libc++ doesn't support TR1.
-#  define GTEST_HAS_TR1_TUPLE 0
-# else
-// The user didn't tell us not to do it, so we assume it's OK.
-#  define GTEST_HAS_TR1_TUPLE 1
-# endif
-#endif  // GTEST_HAS_TR1_TUPLE
-
-// Determines whether Google Test's own tr1 tuple implementation
-// should be used.
-#ifndef GTEST_USE_OWN_TR1_TUPLE
-// We use our own tuple implementation on Symbian.
-# if GTEST_OS_SYMBIAN
-#  define GTEST_USE_OWN_TR1_TUPLE 1
-# else
-// The user didn't tell us, so we need to figure it out.
-
-// We use our own TR1 tuple if we aren't sure the user has an
-// implementation of it already.  At this time, libstdc++ 4.0.0+ and
-// MSVC 2010 are the only mainstream standard libraries that come
-// with a TR1 tuple implementation.  NVIDIA's CUDA NVCC compiler
-// pretends to be GCC by defining __GNUC__ and friends, but cannot
-// compile GCC's tuple implementation.  MSVC 2008 (9.0) provides TR1
-// tuple in a 323 MB Feature Pack download, which we cannot assume the
-// user has.  QNX's QCC compiler is a modified GCC but it doesn't
-// support TR1 tuple.  libc++ only provides std::tuple, in C++11 mode,
-// and it can be used with some compilers that define __GNUC__.
-# if (defined(__GNUC__) && !defined(__CUDACC__) && (GTEST_GCC_VER_ >= 40000) \
-      && !GTEST_OS_QNX && !defined(_LIBCPP_VERSION)) \
-      || (_MSC_VER >= 1600 && _MSC_VER < 1900)
-#  define GTEST_ENV_HAS_TR1_TUPLE_ 1
-# endif
-
-// C++11 specifies that <tuple> provides std::tuple. Use that if gtest is used
-// in C++11 mode and libstdc++ isn't very old (binaries targeting OS X 10.6
-// can build with clang but need to use gcc4.2's libstdc++).
-# if GTEST_LANG_CXX11 && (!defined(__GLIBCXX__) || __GLIBCXX__ > 20110325)
-#  define GTEST_ENV_HAS_STD_TUPLE_ 1
-# endif
-
-# if GTEST_ENV_HAS_TR1_TUPLE_ || GTEST_ENV_HAS_STD_TUPLE_
-#  define GTEST_USE_OWN_TR1_TUPLE 0
-# else
-#  define GTEST_USE_OWN_TR1_TUPLE 1
-# endif
-# endif  // GTEST_OS_SYMBIAN
-#endif  // GTEST_USE_OWN_TR1_TUPLE
-
-// To avoid conditional compilation we make it gtest-port.h's responsibility
-// to #include the header implementing tuple.
-#if GTEST_HAS_STD_TUPLE_
-# include <tuple>  // IWYU pragma: export
-# define GTEST_TUPLE_NAMESPACE_ ::std
-#endif  // GTEST_HAS_STD_TUPLE_
-
-// We include tr1::tuple even if std::tuple is available to define printers for
-// them.
-#if GTEST_HAS_TR1_TUPLE
-# ifndef GTEST_TUPLE_NAMESPACE_
-#  define GTEST_TUPLE_NAMESPACE_ ::std::tr1
-# endif  // GTEST_TUPLE_NAMESPACE_
-
-# if GTEST_USE_OWN_TR1_TUPLE
-// This file was GENERATED by command:
-//     pump.py gtest-tuple.h.pump
-// DO NOT EDIT BY HAND!!!
-
-// Copyright 2009 Google Inc.
-// All Rights Reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Implements a subset of TR1 tuple needed by Google Test and Google Mock.
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_TUPLE_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_TUPLE_H_
-
-#include <utility>  // For ::std::pair.
-
-// The compiler used in Symbian has a bug that prevents us from declaring the
-// tuple template as a friend (it complains that tuple is redefined).  This
-// hack bypasses the bug by declaring the members that should otherwise be
-// private as public.
-// Sun Studio versions < 12 also have the above bug.
-#if defined(__SYMBIAN32__) || (defined(__SUNPRO_CC) && __SUNPRO_CC < 0x590)
-# define GTEST_DECLARE_TUPLE_AS_FRIEND_ public:
-#else
-# define GTEST_DECLARE_TUPLE_AS_FRIEND_ \
-    template <GTEST_10_TYPENAMES_(U)> friend class tuple; \
-   private:
-#endif
-
-// Visual Studio 2010, 2012, and 2013 define symbols in std::tr1 that conflict
-// with our own definitions. Therefore using our own tuple does not work on
-// those compilers.
-#if defined(_MSC_VER) && _MSC_VER >= 1600  /* 1600 is Visual Studio 2010 */
-# error "gtest's tuple doesn't compile on Visual Studio 2010 or later. \
-GTEST_USE_OWN_TR1_TUPLE must be set to 0 on those compilers."
-#endif
-
-// GTEST_n_TUPLE_(T) is the type of an n-tuple.
-#define GTEST_0_TUPLE_(T) tuple<>
-#define GTEST_1_TUPLE_(T) tuple<T##0, void, void, void, void, void, void, \
-    void, void, void>
-#define GTEST_2_TUPLE_(T) tuple<T##0, T##1, void, void, void, void, void, \
-    void, void, void>
-#define GTEST_3_TUPLE_(T) tuple<T##0, T##1, T##2, void, void, void, void, \
-    void, void, void>
-#define GTEST_4_TUPLE_(T) tuple<T##0, T##1, T##2, T##3, void, void, void, \
-    void, void, void>
-#define GTEST_5_TUPLE_(T) tuple<T##0, T##1, T##2, T##3, T##4, void, void, \
-    void, void, void>
-#define GTEST_6_TUPLE_(T) tuple<T##0, T##1, T##2, T##3, T##4, T##5, void, \
-    void, void, void>
-#define GTEST_7_TUPLE_(T) tuple<T##0, T##1, T##2, T##3, T##4, T##5, T##6, \
-    void, void, void>
-#define GTEST_8_TUPLE_(T) tuple<T##0, T##1, T##2, T##3, T##4, T##5, T##6, \
-    T##7, void, void>
-#define GTEST_9_TUPLE_(T) tuple<T##0, T##1, T##2, T##3, T##4, T##5, T##6, \
-    T##7, T##8, void>
-#define GTEST_10_TUPLE_(T) tuple<T##0, T##1, T##2, T##3, T##4, T##5, T##6, \
-    T##7, T##8, T##9>
-
-// GTEST_n_TYPENAMES_(T) declares a list of n typenames.
-#define GTEST_0_TYPENAMES_(T)
-#define GTEST_1_TYPENAMES_(T) typename T##0
-#define GTEST_2_TYPENAMES_(T) typename T##0, typename T##1
-#define GTEST_3_TYPENAMES_(T) typename T##0, typename T##1, typename T##2
-#define GTEST_4_TYPENAMES_(T) typename T##0, typename T##1, typename T##2, \
-    typename T##3
-#define GTEST_5_TYPENAMES_(T) typename T##0, typename T##1, typename T##2, \
-    typename T##3, typename T##4
-#define GTEST_6_TYPENAMES_(T) typename T##0, typename T##1, typename T##2, \
-    typename T##3, typename T##4, typename T##5
-#define GTEST_7_TYPENAMES_(T) typename T##0, typename T##1, typename T##2, \
-    typename T##3, typename T##4, typename T##5, typename T##6
-#define GTEST_8_TYPENAMES_(T) typename T##0, typename T##1, typename T##2, \
-    typename T##3, typename T##4, typename T##5, typename T##6, typename T##7
-#define GTEST_9_TYPENAMES_(T) typename T##0, typename T##1, typename T##2, \
-    typename T##3, typename T##4, typename T##5, typename T##6, \
-    typename T##7, typename T##8
-#define GTEST_10_TYPENAMES_(T) typename T##0, typename T##1, typename T##2, \
-    typename T##3, typename T##4, typename T##5, typename T##6, \
-    typename T##7, typename T##8, typename T##9
-
-// In theory, defining stuff in the ::std namespace is undefined
-// behavior.  We can do this as we are playing the role of a standard
-// library vendor.
-namespace std {
-namespace tr1 {
-
-template <typename T0 = void, typename T1 = void, typename T2 = void,
-    typename T3 = void, typename T4 = void, typename T5 = void,
-    typename T6 = void, typename T7 = void, typename T8 = void,
-    typename T9 = void>
-class tuple;
-
-// Anything in namespace gtest_internal is Google Test's INTERNAL
-// IMPLEMENTATION DETAIL and MUST NOT BE USED DIRECTLY in user code.
-namespace gtest_internal {
-
-// ByRef<T>::type is T if T is a reference; otherwise it's const T&.
-template <typename T>
-struct ByRef { typedef const T& type; };  // NOLINT
-template <typename T>
-struct ByRef<T&> { typedef T& type; };  // NOLINT
-
-// A handy wrapper for ByRef.
-#define GTEST_BY_REF_(T) typename ::std::tr1::gtest_internal::ByRef<T>::type
-
-// AddRef<T>::type is T if T is a reference; otherwise it's T&.  This
-// is the same as tr1::add_reference<T>::type.
-template <typename T>
-struct AddRef { typedef T& type; };  // NOLINT
-template <typename T>
-struct AddRef<T&> { typedef T& type; };  // NOLINT
-
-// A handy wrapper for AddRef.
-#define GTEST_ADD_REF_(T) typename ::std::tr1::gtest_internal::AddRef<T>::type
-
-// A helper for implementing get<k>().
-template <int k> class Get;
-
-// A helper for implementing tuple_element<k, T>.  kIndexValid is true
-// iff k < the number of fields in tuple type T.
-template <bool kIndexValid, int kIndex, class Tuple>
-struct TupleElement;
-
-template <GTEST_10_TYPENAMES_(T)>
-struct TupleElement<true, 0, GTEST_10_TUPLE_(T) > {
-  typedef T0 type;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-struct TupleElement<true, 1, GTEST_10_TUPLE_(T) > {
-  typedef T1 type;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-struct TupleElement<true, 2, GTEST_10_TUPLE_(T) > {
-  typedef T2 type;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-struct TupleElement<true, 3, GTEST_10_TUPLE_(T) > {
-  typedef T3 type;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-struct TupleElement<true, 4, GTEST_10_TUPLE_(T) > {
-  typedef T4 type;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-struct TupleElement<true, 5, GTEST_10_TUPLE_(T) > {
-  typedef T5 type;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-struct TupleElement<true, 6, GTEST_10_TUPLE_(T) > {
-  typedef T6 type;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-struct TupleElement<true, 7, GTEST_10_TUPLE_(T) > {
-  typedef T7 type;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-struct TupleElement<true, 8, GTEST_10_TUPLE_(T) > {
-  typedef T8 type;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-struct TupleElement<true, 9, GTEST_10_TUPLE_(T) > {
-  typedef T9 type;
-};
-
-}  // namespace gtest_internal
-
-template <>
-class tuple<> {
- public:
-  tuple() {}
-  tuple(const tuple& /* t */)  {}
-  tuple& operator=(const tuple& /* t */) { return *this; }
-};
-
-template <GTEST_1_TYPENAMES_(T)>
-class GTEST_1_TUPLE_(T) {
- public:
-  template <int k> friend class gtest_internal::Get;
-
-  tuple() : f0_() {}
-
-  explicit tuple(GTEST_BY_REF_(T0) f0) : f0_(f0) {}
-
-  tuple(const tuple& t) : f0_(t.f0_) {}
-
-  template <GTEST_1_TYPENAMES_(U)>
-  tuple(const GTEST_1_TUPLE_(U)& t) : f0_(t.f0_) {}
-
-  tuple& operator=(const tuple& t) { return CopyFrom(t); }
-
-  template <GTEST_1_TYPENAMES_(U)>
-  tuple& operator=(const GTEST_1_TUPLE_(U)& t) {
-    return CopyFrom(t);
-  }
-
-  GTEST_DECLARE_TUPLE_AS_FRIEND_
-
-  template <GTEST_1_TYPENAMES_(U)>
-  tuple& CopyFrom(const GTEST_1_TUPLE_(U)& t) {
-    f0_ = t.f0_;
-    return *this;
-  }
-
-  T0 f0_;
-};
-
-template <GTEST_2_TYPENAMES_(T)>
-class GTEST_2_TUPLE_(T) {
- public:
-  template <int k> friend class gtest_internal::Get;
-
-  tuple() : f0_(), f1_() {}
-
-  explicit tuple(GTEST_BY_REF_(T0) f0, GTEST_BY_REF_(T1) f1) : f0_(f0),
-      f1_(f1) {}
-
-  tuple(const tuple& t) : f0_(t.f0_), f1_(t.f1_) {}
-
-  template <GTEST_2_TYPENAMES_(U)>
-  tuple(const GTEST_2_TUPLE_(U)& t) : f0_(t.f0_), f1_(t.f1_) {}
-  template <typename U0, typename U1>
-  tuple(const ::std::pair<U0, U1>& p) : f0_(p.first), f1_(p.second) {}
-
-  tuple& operator=(const tuple& t) { return CopyFrom(t); }
-
-  template <GTEST_2_TYPENAMES_(U)>
-  tuple& operator=(const GTEST_2_TUPLE_(U)& t) {
-    return CopyFrom(t);
-  }
-  template <typename U0, typename U1>
-  tuple& operator=(const ::std::pair<U0, U1>& p) {
-    f0_ = p.first;
-    f1_ = p.second;
-    return *this;
-  }
-
-  GTEST_DECLARE_TUPLE_AS_FRIEND_
-
-  template <GTEST_2_TYPENAMES_(U)>
-  tuple& CopyFrom(const GTEST_2_TUPLE_(U)& t) {
-    f0_ = t.f0_;
-    f1_ = t.f1_;
-    return *this;
-  }
-
-  T0 f0_;
-  T1 f1_;
-};
-
-template <GTEST_3_TYPENAMES_(T)>
-class GTEST_3_TUPLE_(T) {
- public:
-  template <int k> friend class gtest_internal::Get;
-
-  tuple() : f0_(), f1_(), f2_() {}
-
-  explicit tuple(GTEST_BY_REF_(T0) f0, GTEST_BY_REF_(T1) f1,
-      GTEST_BY_REF_(T2) f2) : f0_(f0), f1_(f1), f2_(f2) {}
-
-  tuple(const tuple& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_) {}
-
-  template <GTEST_3_TYPENAMES_(U)>
-  tuple(const GTEST_3_TUPLE_(U)& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_) {}
-
-  tuple& operator=(const tuple& t) { return CopyFrom(t); }
-
-  template <GTEST_3_TYPENAMES_(U)>
-  tuple& operator=(const GTEST_3_TUPLE_(U)& t) {
-    return CopyFrom(t);
-  }
-
-  GTEST_DECLARE_TUPLE_AS_FRIEND_
-
-  template <GTEST_3_TYPENAMES_(U)>
-  tuple& CopyFrom(const GTEST_3_TUPLE_(U)& t) {
-    f0_ = t.f0_;
-    f1_ = t.f1_;
-    f2_ = t.f2_;
-    return *this;
-  }
-
-  T0 f0_;
-  T1 f1_;
-  T2 f2_;
-};
-
-template <GTEST_4_TYPENAMES_(T)>
-class GTEST_4_TUPLE_(T) {
- public:
-  template <int k> friend class gtest_internal::Get;
-
-  tuple() : f0_(), f1_(), f2_(), f3_() {}
-
-  explicit tuple(GTEST_BY_REF_(T0) f0, GTEST_BY_REF_(T1) f1,
-      GTEST_BY_REF_(T2) f2, GTEST_BY_REF_(T3) f3) : f0_(f0), f1_(f1), f2_(f2),
-      f3_(f3) {}
-
-  tuple(const tuple& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_), f3_(t.f3_) {}
-
-  template <GTEST_4_TYPENAMES_(U)>
-  tuple(const GTEST_4_TUPLE_(U)& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_),
-      f3_(t.f3_) {}
-
-  tuple& operator=(const tuple& t) { return CopyFrom(t); }
-
-  template <GTEST_4_TYPENAMES_(U)>
-  tuple& operator=(const GTEST_4_TUPLE_(U)& t) {
-    return CopyFrom(t);
-  }
-
-  GTEST_DECLARE_TUPLE_AS_FRIEND_
-
-  template <GTEST_4_TYPENAMES_(U)>
-  tuple& CopyFrom(const GTEST_4_TUPLE_(U)& t) {
-    f0_ = t.f0_;
-    f1_ = t.f1_;
-    f2_ = t.f2_;
-    f3_ = t.f3_;
-    return *this;
-  }
-
-  T0 f0_;
-  T1 f1_;
-  T2 f2_;
-  T3 f3_;
-};
-
-template <GTEST_5_TYPENAMES_(T)>
-class GTEST_5_TUPLE_(T) {
- public:
-  template <int k> friend class gtest_internal::Get;
-
-  tuple() : f0_(), f1_(), f2_(), f3_(), f4_() {}
-
-  explicit tuple(GTEST_BY_REF_(T0) f0, GTEST_BY_REF_(T1) f1,
-      GTEST_BY_REF_(T2) f2, GTEST_BY_REF_(T3) f3,
-      GTEST_BY_REF_(T4) f4) : f0_(f0), f1_(f1), f2_(f2), f3_(f3), f4_(f4) {}
-
-  tuple(const tuple& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_), f3_(t.f3_),
-      f4_(t.f4_) {}
-
-  template <GTEST_5_TYPENAMES_(U)>
-  tuple(const GTEST_5_TUPLE_(U)& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_),
-      f3_(t.f3_), f4_(t.f4_) {}
-
-  tuple& operator=(const tuple& t) { return CopyFrom(t); }
-
-  template <GTEST_5_TYPENAMES_(U)>
-  tuple& operator=(const GTEST_5_TUPLE_(U)& t) {
-    return CopyFrom(t);
-  }
-
-  GTEST_DECLARE_TUPLE_AS_FRIEND_
-
-  template <GTEST_5_TYPENAMES_(U)>
-  tuple& CopyFrom(const GTEST_5_TUPLE_(U)& t) {
-    f0_ = t.f0_;
-    f1_ = t.f1_;
-    f2_ = t.f2_;
-    f3_ = t.f3_;
-    f4_ = t.f4_;
-    return *this;
-  }
-
-  T0 f0_;
-  T1 f1_;
-  T2 f2_;
-  T3 f3_;
-  T4 f4_;
-};
-
-template <GTEST_6_TYPENAMES_(T)>
-class GTEST_6_TUPLE_(T) {
- public:
-  template <int k> friend class gtest_internal::Get;
-
-  tuple() : f0_(), f1_(), f2_(), f3_(), f4_(), f5_() {}
-
-  explicit tuple(GTEST_BY_REF_(T0) f0, GTEST_BY_REF_(T1) f1,
-      GTEST_BY_REF_(T2) f2, GTEST_BY_REF_(T3) f3, GTEST_BY_REF_(T4) f4,
-      GTEST_BY_REF_(T5) f5) : f0_(f0), f1_(f1), f2_(f2), f3_(f3), f4_(f4),
-      f5_(f5) {}
-
-  tuple(const tuple& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_), f3_(t.f3_),
-      f4_(t.f4_), f5_(t.f5_) {}
-
-  template <GTEST_6_TYPENAMES_(U)>
-  tuple(const GTEST_6_TUPLE_(U)& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_),
-      f3_(t.f3_), f4_(t.f4_), f5_(t.f5_) {}
-
-  tuple& operator=(const tuple& t) { return CopyFrom(t); }
-
-  template <GTEST_6_TYPENAMES_(U)>
-  tuple& operator=(const GTEST_6_TUPLE_(U)& t) {
-    return CopyFrom(t);
-  }
-
-  GTEST_DECLARE_TUPLE_AS_FRIEND_
-
-  template <GTEST_6_TYPENAMES_(U)>
-  tuple& CopyFrom(const GTEST_6_TUPLE_(U)& t) {
-    f0_ = t.f0_;
-    f1_ = t.f1_;
-    f2_ = t.f2_;
-    f3_ = t.f3_;
-    f4_ = t.f4_;
-    f5_ = t.f5_;
-    return *this;
-  }
-
-  T0 f0_;
-  T1 f1_;
-  T2 f2_;
-  T3 f3_;
-  T4 f4_;
-  T5 f5_;
-};
-
-template <GTEST_7_TYPENAMES_(T)>
-class GTEST_7_TUPLE_(T) {
- public:
-  template <int k> friend class gtest_internal::Get;
-
-  tuple() : f0_(), f1_(), f2_(), f3_(), f4_(), f5_(), f6_() {}
-
-  explicit tuple(GTEST_BY_REF_(T0) f0, GTEST_BY_REF_(T1) f1,
-      GTEST_BY_REF_(T2) f2, GTEST_BY_REF_(T3) f3, GTEST_BY_REF_(T4) f4,
-      GTEST_BY_REF_(T5) f5, GTEST_BY_REF_(T6) f6) : f0_(f0), f1_(f1), f2_(f2),
-      f3_(f3), f4_(f4), f5_(f5), f6_(f6) {}
-
-  tuple(const tuple& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_), f3_(t.f3_),
-      f4_(t.f4_), f5_(t.f5_), f6_(t.f6_) {}
-
-  template <GTEST_7_TYPENAMES_(U)>
-  tuple(const GTEST_7_TUPLE_(U)& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_),
-      f3_(t.f3_), f4_(t.f4_), f5_(t.f5_), f6_(t.f6_) {}
-
-  tuple& operator=(const tuple& t) { return CopyFrom(t); }
-
-  template <GTEST_7_TYPENAMES_(U)>
-  tuple& operator=(const GTEST_7_TUPLE_(U)& t) {
-    return CopyFrom(t);
-  }
-
-  GTEST_DECLARE_TUPLE_AS_FRIEND_
-
-  template <GTEST_7_TYPENAMES_(U)>
-  tuple& CopyFrom(const GTEST_7_TUPLE_(U)& t) {
-    f0_ = t.f0_;
-    f1_ = t.f1_;
-    f2_ = t.f2_;
-    f3_ = t.f3_;
-    f4_ = t.f4_;
-    f5_ = t.f5_;
-    f6_ = t.f6_;
-    return *this;
-  }
-
-  T0 f0_;
-  T1 f1_;
-  T2 f2_;
-  T3 f3_;
-  T4 f4_;
-  T5 f5_;
-  T6 f6_;
-};
-
-template <GTEST_8_TYPENAMES_(T)>
-class GTEST_8_TUPLE_(T) {
- public:
-  template <int k> friend class gtest_internal::Get;
-
-  tuple() : f0_(), f1_(), f2_(), f3_(), f4_(), f5_(), f6_(), f7_() {}
-
-  explicit tuple(GTEST_BY_REF_(T0) f0, GTEST_BY_REF_(T1) f1,
-      GTEST_BY_REF_(T2) f2, GTEST_BY_REF_(T3) f3, GTEST_BY_REF_(T4) f4,
-      GTEST_BY_REF_(T5) f5, GTEST_BY_REF_(T6) f6,
-      GTEST_BY_REF_(T7) f7) : f0_(f0), f1_(f1), f2_(f2), f3_(f3), f4_(f4),
-      f5_(f5), f6_(f6), f7_(f7) {}
-
-  tuple(const tuple& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_), f3_(t.f3_),
-      f4_(t.f4_), f5_(t.f5_), f6_(t.f6_), f7_(t.f7_) {}
-
-  template <GTEST_8_TYPENAMES_(U)>
-  tuple(const GTEST_8_TUPLE_(U)& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_),
-      f3_(t.f3_), f4_(t.f4_), f5_(t.f5_), f6_(t.f6_), f7_(t.f7_) {}
-
-  tuple& operator=(const tuple& t) { return CopyFrom(t); }
-
-  template <GTEST_8_TYPENAMES_(U)>
-  tuple& operator=(const GTEST_8_TUPLE_(U)& t) {
-    return CopyFrom(t);
-  }
-
-  GTEST_DECLARE_TUPLE_AS_FRIEND_
-
-  template <GTEST_8_TYPENAMES_(U)>
-  tuple& CopyFrom(const GTEST_8_TUPLE_(U)& t) {
-    f0_ = t.f0_;
-    f1_ = t.f1_;
-    f2_ = t.f2_;
-    f3_ = t.f3_;
-    f4_ = t.f4_;
-    f5_ = t.f5_;
-    f6_ = t.f6_;
-    f7_ = t.f7_;
-    return *this;
-  }
-
-  T0 f0_;
-  T1 f1_;
-  T2 f2_;
-  T3 f3_;
-  T4 f4_;
-  T5 f5_;
-  T6 f6_;
-  T7 f7_;
-};
-
-template <GTEST_9_TYPENAMES_(T)>
-class GTEST_9_TUPLE_(T) {
- public:
-  template <int k> friend class gtest_internal::Get;
-
-  tuple() : f0_(), f1_(), f2_(), f3_(), f4_(), f5_(), f6_(), f7_(), f8_() {}
-
-  explicit tuple(GTEST_BY_REF_(T0) f0, GTEST_BY_REF_(T1) f1,
-      GTEST_BY_REF_(T2) f2, GTEST_BY_REF_(T3) f3, GTEST_BY_REF_(T4) f4,
-      GTEST_BY_REF_(T5) f5, GTEST_BY_REF_(T6) f6, GTEST_BY_REF_(T7) f7,
-      GTEST_BY_REF_(T8) f8) : f0_(f0), f1_(f1), f2_(f2), f3_(f3), f4_(f4),
-      f5_(f5), f6_(f6), f7_(f7), f8_(f8) {}
-
-  tuple(const tuple& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_), f3_(t.f3_),
-      f4_(t.f4_), f5_(t.f5_), f6_(t.f6_), f7_(t.f7_), f8_(t.f8_) {}
-
-  template <GTEST_9_TYPENAMES_(U)>
-  tuple(const GTEST_9_TUPLE_(U)& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_),
-      f3_(t.f3_), f4_(t.f4_), f5_(t.f5_), f6_(t.f6_), f7_(t.f7_), f8_(t.f8_) {}
-
-  tuple& operator=(const tuple& t) { return CopyFrom(t); }
-
-  template <GTEST_9_TYPENAMES_(U)>
-  tuple& operator=(const GTEST_9_TUPLE_(U)& t) {
-    return CopyFrom(t);
-  }
-
-  GTEST_DECLARE_TUPLE_AS_FRIEND_
-
-  template <GTEST_9_TYPENAMES_(U)>
-  tuple& CopyFrom(const GTEST_9_TUPLE_(U)& t) {
-    f0_ = t.f0_;
-    f1_ = t.f1_;
-    f2_ = t.f2_;
-    f3_ = t.f3_;
-    f4_ = t.f4_;
-    f5_ = t.f5_;
-    f6_ = t.f6_;
-    f7_ = t.f7_;
-    f8_ = t.f8_;
-    return *this;
-  }
-
-  T0 f0_;
-  T1 f1_;
-  T2 f2_;
-  T3 f3_;
-  T4 f4_;
-  T5 f5_;
-  T6 f6_;
-  T7 f7_;
-  T8 f8_;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-class tuple {
- public:
-  template <int k> friend class gtest_internal::Get;
-
-  tuple() : f0_(), f1_(), f2_(), f3_(), f4_(), f5_(), f6_(), f7_(), f8_(),
-      f9_() {}
-
-  explicit tuple(GTEST_BY_REF_(T0) f0, GTEST_BY_REF_(T1) f1,
-      GTEST_BY_REF_(T2) f2, GTEST_BY_REF_(T3) f3, GTEST_BY_REF_(T4) f4,
-      GTEST_BY_REF_(T5) f5, GTEST_BY_REF_(T6) f6, GTEST_BY_REF_(T7) f7,
-      GTEST_BY_REF_(T8) f8, GTEST_BY_REF_(T9) f9) : f0_(f0), f1_(f1), f2_(f2),
-      f3_(f3), f4_(f4), f5_(f5), f6_(f6), f7_(f7), f8_(f8), f9_(f9) {}
-
-  tuple(const tuple& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_), f3_(t.f3_),
-      f4_(t.f4_), f5_(t.f5_), f6_(t.f6_), f7_(t.f7_), f8_(t.f8_), f9_(t.f9_) {}
-
-  template <GTEST_10_TYPENAMES_(U)>
-  tuple(const GTEST_10_TUPLE_(U)& t) : f0_(t.f0_), f1_(t.f1_), f2_(t.f2_),
-      f3_(t.f3_), f4_(t.f4_), f5_(t.f5_), f6_(t.f6_), f7_(t.f7_), f8_(t.f8_),
-      f9_(t.f9_) {}
-
-  tuple& operator=(const tuple& t) { return CopyFrom(t); }
-
-  template <GTEST_10_TYPENAMES_(U)>
-  tuple& operator=(const GTEST_10_TUPLE_(U)& t) {
-    return CopyFrom(t);
-  }
-
-  GTEST_DECLARE_TUPLE_AS_FRIEND_
-
-  template <GTEST_10_TYPENAMES_(U)>
-  tuple& CopyFrom(const GTEST_10_TUPLE_(U)& t) {
-    f0_ = t.f0_;
-    f1_ = t.f1_;
-    f2_ = t.f2_;
-    f3_ = t.f3_;
-    f4_ = t.f4_;
-    f5_ = t.f5_;
-    f6_ = t.f6_;
-    f7_ = t.f7_;
-    f8_ = t.f8_;
-    f9_ = t.f9_;
-    return *this;
-  }
-
-  T0 f0_;
-  T1 f1_;
-  T2 f2_;
-  T3 f3_;
-  T4 f4_;
-  T5 f5_;
-  T6 f6_;
-  T7 f7_;
-  T8 f8_;
-  T9 f9_;
-};
-
-// 6.1.3.2 Tuple creation functions.
-
-// Known limitations: we don't support passing an
-// std::tr1::reference_wrapper<T> to make_tuple().  And we don't
-// implement tie().
-
-inline tuple<> make_tuple() { return tuple<>(); }
-
-template <GTEST_1_TYPENAMES_(T)>
-inline GTEST_1_TUPLE_(T) make_tuple(const T0& f0) {
-  return GTEST_1_TUPLE_(T)(f0);
-}
-
-template <GTEST_2_TYPENAMES_(T)>
-inline GTEST_2_TUPLE_(T) make_tuple(const T0& f0, const T1& f1) {
-  return GTEST_2_TUPLE_(T)(f0, f1);
-}
-
-template <GTEST_3_TYPENAMES_(T)>
-inline GTEST_3_TUPLE_(T) make_tuple(const T0& f0, const T1& f1, const T2& f2) {
-  return GTEST_3_TUPLE_(T)(f0, f1, f2);
-}
-
-template <GTEST_4_TYPENAMES_(T)>
-inline GTEST_4_TUPLE_(T) make_tuple(const T0& f0, const T1& f1, const T2& f2,
-    const T3& f3) {
-  return GTEST_4_TUPLE_(T)(f0, f1, f2, f3);
-}
-
-template <GTEST_5_TYPENAMES_(T)>
-inline GTEST_5_TUPLE_(T) make_tuple(const T0& f0, const T1& f1, const T2& f2,
-    const T3& f3, const T4& f4) {
-  return GTEST_5_TUPLE_(T)(f0, f1, f2, f3, f4);
-}
-
-template <GTEST_6_TYPENAMES_(T)>
-inline GTEST_6_TUPLE_(T) make_tuple(const T0& f0, const T1& f1, const T2& f2,
-    const T3& f3, const T4& f4, const T5& f5) {
-  return GTEST_6_TUPLE_(T)(f0, f1, f2, f3, f4, f5);
-}
-
-template <GTEST_7_TYPENAMES_(T)>
-inline GTEST_7_TUPLE_(T) make_tuple(const T0& f0, const T1& f1, const T2& f2,
-    const T3& f3, const T4& f4, const T5& f5, const T6& f6) {
-  return GTEST_7_TUPLE_(T)(f0, f1, f2, f3, f4, f5, f6);
-}
-
-template <GTEST_8_TYPENAMES_(T)>
-inline GTEST_8_TUPLE_(T) make_tuple(const T0& f0, const T1& f1, const T2& f2,
-    const T3& f3, const T4& f4, const T5& f5, const T6& f6, const T7& f7) {
-  return GTEST_8_TUPLE_(T)(f0, f1, f2, f3, f4, f5, f6, f7);
-}
-
-template <GTEST_9_TYPENAMES_(T)>
-inline GTEST_9_TUPLE_(T) make_tuple(const T0& f0, const T1& f1, const T2& f2,
-    const T3& f3, const T4& f4, const T5& f5, const T6& f6, const T7& f7,
-    const T8& f8) {
-  return GTEST_9_TUPLE_(T)(f0, f1, f2, f3, f4, f5, f6, f7, f8);
-}
-
-template <GTEST_10_TYPENAMES_(T)>
-inline GTEST_10_TUPLE_(T) make_tuple(const T0& f0, const T1& f1, const T2& f2,
-    const T3& f3, const T4& f4, const T5& f5, const T6& f6, const T7& f7,
-    const T8& f8, const T9& f9) {
-  return GTEST_10_TUPLE_(T)(f0, f1, f2, f3, f4, f5, f6, f7, f8, f9);
-}
-
-// 6.1.3.3 Tuple helper classes.
-
-template <typename Tuple> struct tuple_size;
-
-template <GTEST_0_TYPENAMES_(T)>
-struct tuple_size<GTEST_0_TUPLE_(T) > {
-  static const int value = 0;
-};
-
-template <GTEST_1_TYPENAMES_(T)>
-struct tuple_size<GTEST_1_TUPLE_(T) > {
-  static const int value = 1;
-};
-
-template <GTEST_2_TYPENAMES_(T)>
-struct tuple_size<GTEST_2_TUPLE_(T) > {
-  static const int value = 2;
-};
-
-template <GTEST_3_TYPENAMES_(T)>
-struct tuple_size<GTEST_3_TUPLE_(T) > {
-  static const int value = 3;
-};
-
-template <GTEST_4_TYPENAMES_(T)>
-struct tuple_size<GTEST_4_TUPLE_(T) > {
-  static const int value = 4;
-};
-
-template <GTEST_5_TYPENAMES_(T)>
-struct tuple_size<GTEST_5_TUPLE_(T) > {
-  static const int value = 5;
-};
-
-template <GTEST_6_TYPENAMES_(T)>
-struct tuple_size<GTEST_6_TUPLE_(T) > {
-  static const int value = 6;
-};
-
-template <GTEST_7_TYPENAMES_(T)>
-struct tuple_size<GTEST_7_TUPLE_(T) > {
-  static const int value = 7;
-};
-
-template <GTEST_8_TYPENAMES_(T)>
-struct tuple_size<GTEST_8_TUPLE_(T) > {
-  static const int value = 8;
-};
-
-template <GTEST_9_TYPENAMES_(T)>
-struct tuple_size<GTEST_9_TUPLE_(T) > {
-  static const int value = 9;
-};
-
-template <GTEST_10_TYPENAMES_(T)>
-struct tuple_size<GTEST_10_TUPLE_(T) > {
-  static const int value = 10;
-};
-
-template <int k, class Tuple>
-struct tuple_element {
-  typedef typename gtest_internal::TupleElement<
-      k < (tuple_size<Tuple>::value), k, Tuple>::type type;
-};
-
-#define GTEST_TUPLE_ELEMENT_(k, Tuple) typename tuple_element<k, Tuple >::type
-
-// 6.1.3.4 Element access.
-
-namespace gtest_internal {
-
-template <>
-class Get<0> {
- public:
-  template <class Tuple>
-  static GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(0, Tuple))
-  Field(Tuple& t) { return t.f0_; }  // NOLINT
-
-  template <class Tuple>
-  static GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(0, Tuple))
-  ConstField(const Tuple& t) { return t.f0_; }
-};
-
-template <>
-class Get<1> {
- public:
-  template <class Tuple>
-  static GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(1, Tuple))
-  Field(Tuple& t) { return t.f1_; }  // NOLINT
-
-  template <class Tuple>
-  static GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(1, Tuple))
-  ConstField(const Tuple& t) { return t.f1_; }
-};
-
-template <>
-class Get<2> {
- public:
-  template <class Tuple>
-  static GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(2, Tuple))
-  Field(Tuple& t) { return t.f2_; }  // NOLINT
-
-  template <class Tuple>
-  static GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(2, Tuple))
-  ConstField(const Tuple& t) { return t.f2_; }
-};
-
-template <>
-class Get<3> {
- public:
-  template <class Tuple>
-  static GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(3, Tuple))
-  Field(Tuple& t) { return t.f3_; }  // NOLINT
-
-  template <class Tuple>
-  static GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(3, Tuple))
-  ConstField(const Tuple& t) { return t.f3_; }
-};
-
-template <>
-class Get<4> {
- public:
-  template <class Tuple>
-  static GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(4, Tuple))
-  Field(Tuple& t) { return t.f4_; }  // NOLINT
-
-  template <class Tuple>
-  static GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(4, Tuple))
-  ConstField(const Tuple& t) { return t.f4_; }
-};
-
-template <>
-class Get<5> {
- public:
-  template <class Tuple>
-  static GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(5, Tuple))
-  Field(Tuple& t) { return t.f5_; }  // NOLINT
-
-  template <class Tuple>
-  static GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(5, Tuple))
-  ConstField(const Tuple& t) { return t.f5_; }
-};
-
-template <>
-class Get<6> {
- public:
-  template <class Tuple>
-  static GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(6, Tuple))
-  Field(Tuple& t) { return t.f6_; }  // NOLINT
-
-  template <class Tuple>
-  static GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(6, Tuple))
-  ConstField(const Tuple& t) { return t.f6_; }
-};
-
-template <>
-class Get<7> {
- public:
-  template <class Tuple>
-  static GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(7, Tuple))
-  Field(Tuple& t) { return t.f7_; }  // NOLINT
-
-  template <class Tuple>
-  static GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(7, Tuple))
-  ConstField(const Tuple& t) { return t.f7_; }
-};
-
-template <>
-class Get<8> {
- public:
-  template <class Tuple>
-  static GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(8, Tuple))
-  Field(Tuple& t) { return t.f8_; }  // NOLINT
-
-  template <class Tuple>
-  static GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(8, Tuple))
-  ConstField(const Tuple& t) { return t.f8_; }
-};
-
-template <>
-class Get<9> {
- public:
-  template <class Tuple>
-  static GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(9, Tuple))
-  Field(Tuple& t) { return t.f9_; }  // NOLINT
-
-  template <class Tuple>
-  static GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(9, Tuple))
-  ConstField(const Tuple& t) { return t.f9_; }
-};
-
-}  // namespace gtest_internal
-
-template <int k, GTEST_10_TYPENAMES_(T)>
-GTEST_ADD_REF_(GTEST_TUPLE_ELEMENT_(k, GTEST_10_TUPLE_(T)))
-get(GTEST_10_TUPLE_(T)& t) {
-  return gtest_internal::Get<k>::Field(t);
-}
-
-template <int k, GTEST_10_TYPENAMES_(T)>
-GTEST_BY_REF_(GTEST_TUPLE_ELEMENT_(k,  GTEST_10_TUPLE_(T)))
-get(const GTEST_10_TUPLE_(T)& t) {
-  return gtest_internal::Get<k>::ConstField(t);
-}
-
-// 6.1.3.5 Relational operators
-
-// We only implement == and !=, as we don't have a need for the rest yet.
-
-namespace gtest_internal {
-
-// SameSizeTuplePrefixComparator<k, k>::Eq(t1, t2) returns true if the
-// first k fields of t1 equals the first k fields of t2.
-// SameSizeTuplePrefixComparator(k1, k2) would be a compiler error if
-// k1 != k2.
-template <int kSize1, int kSize2>
-struct SameSizeTuplePrefixComparator;
-
-template <>
-struct SameSizeTuplePrefixComparator<0, 0> {
-  template <class Tuple1, class Tuple2>
-  static bool Eq(const Tuple1& /* t1 */, const Tuple2& /* t2 */) {
-    return true;
-  }
-};
-
-template <int k>
-struct SameSizeTuplePrefixComparator<k, k> {
-  template <class Tuple1, class Tuple2>
-  static bool Eq(const Tuple1& t1, const Tuple2& t2) {
-    return SameSizeTuplePrefixComparator<k - 1, k - 1>::Eq(t1, t2) &&
-        ::std::tr1::get<k - 1>(t1) == ::std::tr1::get<k - 1>(t2);
-  }
-};
-
-}  // namespace gtest_internal
-
-template <GTEST_10_TYPENAMES_(T), GTEST_10_TYPENAMES_(U)>
-inline bool operator==(const GTEST_10_TUPLE_(T)& t,
-                       const GTEST_10_TUPLE_(U)& u) {
-  return gtest_internal::SameSizeTuplePrefixComparator<
-      tuple_size<GTEST_10_TUPLE_(T) >::value,
-      tuple_size<GTEST_10_TUPLE_(U) >::value>::Eq(t, u);
-}
-
-template <GTEST_10_TYPENAMES_(T), GTEST_10_TYPENAMES_(U)>
-inline bool operator!=(const GTEST_10_TUPLE_(T)& t,
-                       const GTEST_10_TUPLE_(U)& u) { return !(t == u); }
-
-// 6.1.4 Pairs.
-// Unimplemented.
-
-}  // namespace tr1
-}  // namespace std
-
-#undef GTEST_0_TUPLE_
-#undef GTEST_1_TUPLE_
-#undef GTEST_2_TUPLE_
-#undef GTEST_3_TUPLE_
-#undef GTEST_4_TUPLE_
-#undef GTEST_5_TUPLE_
-#undef GTEST_6_TUPLE_
-#undef GTEST_7_TUPLE_
-#undef GTEST_8_TUPLE_
-#undef GTEST_9_TUPLE_
-#undef GTEST_10_TUPLE_
-
-#undef GTEST_0_TYPENAMES_
-#undef GTEST_1_TYPENAMES_
-#undef GTEST_2_TYPENAMES_
-#undef GTEST_3_TYPENAMES_
-#undef GTEST_4_TYPENAMES_
-#undef GTEST_5_TYPENAMES_
-#undef GTEST_6_TYPENAMES_
-#undef GTEST_7_TYPENAMES_
-#undef GTEST_8_TYPENAMES_
-#undef GTEST_9_TYPENAMES_
-#undef GTEST_10_TYPENAMES_
-
-#undef GTEST_DECLARE_TUPLE_AS_FRIEND_
-#undef GTEST_BY_REF_
-#undef GTEST_ADD_REF_
-#undef GTEST_TUPLE_ELEMENT_
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_TUPLE_H_
-# elif GTEST_OS_SYMBIAN
-
-// On Symbian, BOOST_HAS_TR1_TUPLE causes Boost's TR1 tuple library to
-// use STLport's tuple implementation, which unfortunately doesn't
-// work as the copy of STLport distributed with Symbian is incomplete.
-// By making sure BOOST_HAS_TR1_TUPLE is undefined, we force Boost to
-// use its own tuple implementation.
-#  ifdef BOOST_HAS_TR1_TUPLE
-#   undef BOOST_HAS_TR1_TUPLE
-#  endif  // BOOST_HAS_TR1_TUPLE
-
-// This prevents <boost/tr1/detail/config.hpp>, which defines
-// BOOST_HAS_TR1_TUPLE, from being #included by Boost's <tuple>.
-#  define BOOST_TR1_DETAIL_CONFIG_HPP_INCLUDED
-#  include <tuple>  // IWYU pragma: export  // NOLINT
-
-# elif defined(__GNUC__) && (GTEST_GCC_VER_ >= 40000)
-// GCC 4.0+ implements tr1/tuple in the <tr1/tuple> header.  This does
-// not conform to the TR1 spec, which requires the header to be <tuple>.
-
-#  if !GTEST_HAS_RTTI && GTEST_GCC_VER_ < 40302
-// Until version 4.3.2, gcc has a bug that causes <tr1/functional>,
-// which is #included by <tr1/tuple>, to not compile when RTTI is
-// disabled.  _TR1_FUNCTIONAL is the header guard for
-// <tr1/functional>.  Hence the following #define is a hack to prevent
-// <tr1/functional> from being included.
-#   define _TR1_FUNCTIONAL 1
-#   include <tr1/tuple>
-#   undef _TR1_FUNCTIONAL  // Allows the user to #include
-                        // <tr1/functional> if they choose to.
-#  else
-#   include <tr1/tuple>  // NOLINT
-#  endif  // !GTEST_HAS_RTTI && GTEST_GCC_VER_ < 40302
-
-// VS 2010 now has tr1 support.
-# elif _MSC_VER >= 1600
-#  include <tuple>  // IWYU pragma: export  // NOLINT
-
-# else  // GTEST_USE_OWN_TR1_TUPLE
-#  include <tr1/tuple>  // IWYU pragma: export  // NOLINT
-# endif  // GTEST_USE_OWN_TR1_TUPLE
-
-#endif  // GTEST_HAS_TR1_TUPLE
-
-// Determines whether clone(2) is supported.
-// Usually it will only be available on Linux, excluding
-// Linux on the Itanium architecture.
-// Also see http://linux.die.net/man/2/clone.
-#ifndef GTEST_HAS_CLONE
-// The user didn't tell us, so we need to figure it out.
-
-# if GTEST_OS_LINUX && !defined(__ia64__)
-#  if GTEST_OS_LINUX_ANDROID
-// On Android, clone() became available at different API levels for each 32-bit
-// architecture.
-#    if defined(__LP64__) || \
-        (defined(__arm__) && __ANDROID_API__ >= 9) || \
-        (defined(__mips__) && __ANDROID_API__ >= 12) || \
-        (defined(__i386__) && __ANDROID_API__ >= 17)
-#     define GTEST_HAS_CLONE 1
-#    else
-#     define GTEST_HAS_CLONE 0
-#    endif
-#  else
-#   define GTEST_HAS_CLONE 1
-#  endif
-# else
-#  define GTEST_HAS_CLONE 0
-# endif  // GTEST_OS_LINUX && !defined(__ia64__)
-
-#endif  // GTEST_HAS_CLONE
-
-// Determines whether to support stream redirection. This is used to test
-// output correctness and to implement death tests.
-#ifndef GTEST_HAS_STREAM_REDIRECTION
-// By default, we assume that stream redirection is supported on all
-// platforms except known mobile ones.
-# if GTEST_OS_WINDOWS_MOBILE || GTEST_OS_SYMBIAN || \
-    GTEST_OS_WINDOWS_PHONE || GTEST_OS_WINDOWS_RT
-#  define GTEST_HAS_STREAM_REDIRECTION 0
-# else
-#  define GTEST_HAS_STREAM_REDIRECTION 1
-# endif  // !GTEST_OS_WINDOWS_MOBILE && !GTEST_OS_SYMBIAN
-#endif  // GTEST_HAS_STREAM_REDIRECTION
-
-// Determines whether to support death tests.
-// Google Test does not support death tests for VC 7.1 and earlier as
-// abort() in a VC 7.1 application compiled as GUI in debug config
-// pops up a dialog window that cannot be suppressed programmatically.
-#if (GTEST_OS_LINUX || GTEST_OS_CYGWIN || GTEST_OS_SOLARIS ||   \
-     (GTEST_OS_MAC && !GTEST_OS_IOS) ||                         \
-     (GTEST_OS_WINDOWS_DESKTOP && _MSC_VER >= 1400) ||          \
-     GTEST_OS_WINDOWS_MINGW || GTEST_OS_AIX || GTEST_OS_HPUX || \
-     GTEST_OS_OPENBSD || GTEST_OS_QNX || GTEST_OS_FREEBSD || \
-     GTEST_OS_NETBSD || GTEST_OS_FUCHSIA)
-# define GTEST_HAS_DEATH_TEST 1
-#endif
-
-// Determines whether to support type-driven tests.
-
-// Typed tests need <typeinfo> and variadic macros, which GCC, VC++ 8.0,
-// Sun Pro CC, IBM Visual Age, and HP aCC support.
-#if defined(__GNUC__) || (_MSC_VER >= 1400) || defined(__SUNPRO_CC) || \
-    defined(__IBMCPP__) || defined(__HP_aCC)
-# define GTEST_HAS_TYPED_TEST 1
-# define GTEST_HAS_TYPED_TEST_P 1
-#endif
-
-// Determines whether to support Combine(). This only makes sense when
-// value-parameterized tests are enabled.  The implementation doesn't
-// work on Sun Studio since it doesn't understand templated conversion
-// operators.
-#if (GTEST_HAS_TR1_TUPLE || GTEST_HAS_STD_TUPLE_) && !defined(__SUNPRO_CC)
-# define GTEST_HAS_COMBINE 1
-#endif
-
-// Determines whether the system compiler uses UTF-16 for encoding wide strings.
-#define GTEST_WIDE_STRING_USES_UTF16_ \
-    (GTEST_OS_WINDOWS || GTEST_OS_CYGWIN || GTEST_OS_SYMBIAN || GTEST_OS_AIX)
-
-// Determines whether test results can be streamed to a socket.
-#if GTEST_OS_LINUX
-# define GTEST_CAN_STREAM_RESULTS_ 1
-#endif
-
-// Defines some utility macros.
-
-// The GNU compiler emits a warning if nested "if" statements are followed by
-// an "else" statement and braces are not used to explicitly disambiguate the
-// "else" binding.  This leads to problems with code like:
-//
-//   if (gate)
-//     ASSERT_*(condition) << "Some message";
-//
-// The "switch (0) case 0:" idiom is used to suppress this.
-#ifdef __INTEL_COMPILER
-# define GTEST_AMBIGUOUS_ELSE_BLOCKER_
-#else
-# define GTEST_AMBIGUOUS_ELSE_BLOCKER_ switch (0) case 0: default:  // NOLINT
-#endif
-
-// Use this annotation at the end of a struct/class definition to
-// prevent the compiler from optimizing away instances that are never
-// used.  This is useful when all interesting logic happens inside the
-// c'tor and / or d'tor.  Example:
-//
-//   struct Foo {
-//     Foo() { ... }
-//   } GTEST_ATTRIBUTE_UNUSED_;
-//
-// Also use it after a variable or parameter declaration to tell the
-// compiler the variable/parameter does not have to be used.
-#if defined(__GNUC__) && !defined(COMPILER_ICC)
-# define GTEST_ATTRIBUTE_UNUSED_ __attribute__ ((unused))
-#elif defined(__clang__)
-# if __has_attribute(unused)
-#  define GTEST_ATTRIBUTE_UNUSED_ __attribute__ ((unused))
-# endif
-#endif
-#ifndef GTEST_ATTRIBUTE_UNUSED_
-# define GTEST_ATTRIBUTE_UNUSED_
-#endif
-
-#if GTEST_LANG_CXX11
-# define GTEST_CXX11_EQUALS_DELETE_ = delete
-#else  // GTEST_LANG_CXX11
-# define GTEST_CXX11_EQUALS_DELETE_
-#endif  // GTEST_LANG_CXX11
-
-// Use this annotation before a function that takes a printf format string.
-#if (defined(__GNUC__) || defined(__clang__)) && !defined(COMPILER_ICC)
-# if defined(__MINGW_PRINTF_FORMAT)
-// MinGW has two different printf implementations. Ensure the format macro
-// matches the selected implementation. See
-// https://sourceforge.net/p/mingw-w64/wiki2/gnu%20printf/.
-#  define GTEST_ATTRIBUTE_PRINTF_(string_index, first_to_check) \
-       __attribute__((__format__(__MINGW_PRINTF_FORMAT, string_index, \
-                                 first_to_check)))
-# else
-#  define GTEST_ATTRIBUTE_PRINTF_(string_index, first_to_check) \
-       __attribute__((__format__(__printf__, string_index, first_to_check)))
-# endif
-#else
-# define GTEST_ATTRIBUTE_PRINTF_(string_index, first_to_check)
-#endif
-
-
-// A macro to disallow operator=
-// This should be used in the private: declarations for a class.
-#define GTEST_DISALLOW_ASSIGN_(type) \
-  void operator=(type const &) GTEST_CXX11_EQUALS_DELETE_
-
-// A macro to disallow copy constructor and operator=
-// This should be used in the private: declarations for a class.
-#define GTEST_DISALLOW_COPY_AND_ASSIGN_(type) \
-  type(type const &) GTEST_CXX11_EQUALS_DELETE_; \
-  GTEST_DISALLOW_ASSIGN_(type)
-
-// Tell the compiler to warn about unused return values for functions declared
-// with this macro.  The macro should be used on function declarations
-// following the argument list:
-//
-//   Sprocket* AllocateSprocket() GTEST_MUST_USE_RESULT_;
-#if defined(__GNUC__) && (GTEST_GCC_VER_ >= 30400) && !defined(COMPILER_ICC)
-# define GTEST_MUST_USE_RESULT_ __attribute__ ((warn_unused_result))
-#else
-# define GTEST_MUST_USE_RESULT_
-#endif  // __GNUC__ && (GTEST_GCC_VER_ >= 30400) && !COMPILER_ICC
-
-// MS C++ compiler emits warning when a conditional expression is compile time
-// constant. In some contexts this warning is false positive and needs to be
-// suppressed. Use the following two macros in such cases:
-//
-// GTEST_INTENTIONAL_CONST_COND_PUSH_()
-// while (true) {
-// GTEST_INTENTIONAL_CONST_COND_POP_()
-// }
-# define GTEST_INTENTIONAL_CONST_COND_PUSH_() \
-    GTEST_DISABLE_MSC_WARNINGS_PUSH_(4127)
-# define GTEST_INTENTIONAL_CONST_COND_POP_() \
-    GTEST_DISABLE_MSC_WARNINGS_POP_()
-
-// Determine whether the compiler supports Microsoft's Structured Exception
-// Handling.  This is supported by several Windows compilers but generally
-// does not exist on any other system.
-#ifndef GTEST_HAS_SEH
-// The user didn't tell us, so we need to figure it out.
-
-# if defined(_MSC_VER) || defined(__BORLANDC__)
-// These two compilers are known to support SEH.
-#  define GTEST_HAS_SEH 1
-# else
-// Assume no SEH.
-#  define GTEST_HAS_SEH 0
-# endif
-
-#define GTEST_IS_THREADSAFE \
-    (GTEST_HAS_MUTEX_AND_THREAD_LOCAL_ \
-     || (GTEST_OS_WINDOWS && !GTEST_OS_WINDOWS_PHONE && !GTEST_OS_WINDOWS_RT) \
-     || GTEST_HAS_PTHREAD)
-
-#endif  // GTEST_HAS_SEH
-
-// GTEST_API_ qualifies all symbols that must be exported. The definitions below
-// are guarded by #ifndef to give embedders a chance to define GTEST_API_ in
-// gtest/internal/custom/gtest-port.h
-#ifndef GTEST_API_
-
-#ifdef _MSC_VER
-# if GTEST_LINKED_AS_SHARED_LIBRARY
-#  define GTEST_API_ __declspec(dllimport)
-# elif GTEST_CREATE_SHARED_LIBRARY
-#  define GTEST_API_ __declspec(dllexport)
-# endif
-#elif __GNUC__ >= 4 || defined(__clang__)
-# define GTEST_API_ __attribute__((visibility ("default")))
-#endif  // _MSC_VER
-
-#endif  // GTEST_API_
-
-#ifndef GTEST_API_
-# define GTEST_API_
-#endif  // GTEST_API_
-
-#ifndef GTEST_DEFAULT_DEATH_TEST_STYLE
-# define GTEST_DEFAULT_DEATH_TEST_STYLE  "fast"
-#endif  // GTEST_DEFAULT_DEATH_TEST_STYLE
-
-#ifdef __GNUC__
-// Ask the compiler to never inline a given function.
-# define GTEST_NO_INLINE_ __attribute__((noinline))
-#else
-# define GTEST_NO_INLINE_
-#endif
-
-// _LIBCPP_VERSION is defined by the libc++ library from the LLVM project.
-#if !defined(GTEST_HAS_CXXABI_H_)
-# if defined(__GLIBCXX__) || (defined(_LIBCPP_VERSION) && !defined(_MSC_VER))
-#  define GTEST_HAS_CXXABI_H_ 1
-# else
-#  define GTEST_HAS_CXXABI_H_ 0
-# endif
-#endif
-
-// A function level attribute to disable checking for use of uninitialized
-// memory when built with MemorySanitizer.
-#if defined(__clang__)
-# if __has_feature(memory_sanitizer)
-#  define GTEST_ATTRIBUTE_NO_SANITIZE_MEMORY_ \
-       __attribute__((no_sanitize_memory))
-# else
-#  define GTEST_ATTRIBUTE_NO_SANITIZE_MEMORY_
-# endif  // __has_feature(memory_sanitizer)
-#else
-# define GTEST_ATTRIBUTE_NO_SANITIZE_MEMORY_
-#endif  // __clang__
-
-// A function level attribute to disable AddressSanitizer instrumentation.
-#if defined(__clang__)
-# if __has_feature(address_sanitizer)
-#  define GTEST_ATTRIBUTE_NO_SANITIZE_ADDRESS_ \
-       __attribute__((no_sanitize_address))
-# else
-#  define GTEST_ATTRIBUTE_NO_SANITIZE_ADDRESS_
-# endif  // __has_feature(address_sanitizer)
-#else
-# define GTEST_ATTRIBUTE_NO_SANITIZE_ADDRESS_
-#endif  // __clang__
-
-// A function level attribute to disable ThreadSanitizer instrumentation.
-#if defined(__clang__)
-# if __has_feature(thread_sanitizer)
-#  define GTEST_ATTRIBUTE_NO_SANITIZE_THREAD_ \
-       __attribute__((no_sanitize_thread))
-# else
-#  define GTEST_ATTRIBUTE_NO_SANITIZE_THREAD_
-# endif  // __has_feature(thread_sanitizer)
-#else
-# define GTEST_ATTRIBUTE_NO_SANITIZE_THREAD_
-#endif  // __clang__
-
-namespace testing {
-
-class Message;
-
-#if defined(GTEST_TUPLE_NAMESPACE_)
-// Import tuple and friends into the ::testing namespace.
-// It is part of our interface, having them in ::testing allows us to change
-// their types as needed.
-using GTEST_TUPLE_NAMESPACE_::get;
-using GTEST_TUPLE_NAMESPACE_::make_tuple;
-using GTEST_TUPLE_NAMESPACE_::tuple;
-using GTEST_TUPLE_NAMESPACE_::tuple_size;
-using GTEST_TUPLE_NAMESPACE_::tuple_element;
-#endif  // defined(GTEST_TUPLE_NAMESPACE_)
-
-namespace internal {
-
-// A secret type that Google Test users don't know about.  It has no
-// definition on purpose.  Therefore it's impossible to create a
-// Secret object, which is what we want.
-class Secret;
-
-// The GTEST_COMPILE_ASSERT_ macro can be used to verify that a compile time
-// expression is true. For example, you could use it to verify the
-// size of a static array:
-//
-//   GTEST_COMPILE_ASSERT_(GTEST_ARRAY_SIZE_(names) == NUM_NAMES,
-//                         names_incorrect_size);
-//
-// or to make sure a struct is smaller than a certain size:
-//
-//   GTEST_COMPILE_ASSERT_(sizeof(foo) < 128, foo_too_large);
-//
-// The second argument to the macro is the name of the variable. If
-// the expression is false, most compilers will issue a warning/error
-// containing the name of the variable.
-
-#if GTEST_LANG_CXX11
-# define GTEST_COMPILE_ASSERT_(expr, msg) static_assert(expr, #msg)
-#else  // !GTEST_LANG_CXX11
-template <bool>
-  struct CompileAssert {
-};
-
-# define GTEST_COMPILE_ASSERT_(expr, msg) \
-  typedef ::testing::internal::CompileAssert<(static_cast<bool>(expr))> \
-      msg[static_cast<bool>(expr) ? 1 : -1] GTEST_ATTRIBUTE_UNUSED_
-#endif  // !GTEST_LANG_CXX11
-
-// Implementation details of GTEST_COMPILE_ASSERT_:
-//
-// (In C++11, we simply use static_assert instead of the following)
-//
-// - GTEST_COMPILE_ASSERT_ works by defining an array type that has -1
-//   elements (and thus is invalid) when the expression is false.
-//
-// - The simpler definition
-//
-//    #define GTEST_COMPILE_ASSERT_(expr, msg) typedef char msg[(expr) ? 1 : -1]
-//
-//   does not work, as gcc supports variable-length arrays whose sizes
-//   are determined at run-time (this is gcc's extension and not part
-//   of the C++ standard).  As a result, gcc fails to reject the
-//   following code with the simple definition:
-//
-//     int foo;
-//     GTEST_COMPILE_ASSERT_(foo, msg); // not supposed to compile as foo is
-//                                      // not a compile-time constant.
-//
-// - By using the type CompileAssert<(bool(expr))>, we ensures that
-//   expr is a compile-time constant.  (Template arguments must be
-//   determined at compile-time.)
-//
-// - The outter parentheses in CompileAssert<(bool(expr))> are necessary
-//   to work around a bug in gcc 3.4.4 and 4.0.1.  If we had written
-//
-//     CompileAssert<bool(expr)>
-//
-//   instead, these compilers will refuse to compile
-//
-//     GTEST_COMPILE_ASSERT_(5 > 0, some_message);
-//
-//   (They seem to think the ">" in "5 > 0" marks the end of the
-//   template argument list.)
-//
-// - The array size is (bool(expr) ? 1 : -1), instead of simply
-//
-//     ((expr) ? 1 : -1).
-//
-//   This is to avoid running into a bug in MS VC 7.1, which
-//   causes ((0.0) ? 1 : -1) to incorrectly evaluate to 1.
-
-// StaticAssertTypeEqHelper is used by StaticAssertTypeEq defined in gtest.h.
-//
-// This template is declared, but intentionally undefined.
-template <typename T1, typename T2>
-struct StaticAssertTypeEqHelper;
-
-template <typename T>
-struct StaticAssertTypeEqHelper<T, T> {
-  enum { value = true };
-};
-
-// Same as std::is_same<>.
-template <typename T, typename U>
-struct IsSame {
-  enum { value = false };
-};
-template <typename T>
-struct IsSame<T, T> {
-  enum { value = true };
-};
-
-// Evaluates to the number of elements in 'array'.
-#define GTEST_ARRAY_SIZE_(array) (sizeof(array) / sizeof(array[0]))
-
-#if GTEST_HAS_GLOBAL_STRING
-typedef ::string string;
-#else
-typedef ::std::string string;
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-#if GTEST_HAS_GLOBAL_WSTRING
-typedef ::wstring wstring;
-#elif GTEST_HAS_STD_WSTRING
-typedef ::std::wstring wstring;
-#endif  // GTEST_HAS_GLOBAL_WSTRING
-
-// A helper for suppressing warnings on constant condition.  It just
-// returns 'condition'.
-GTEST_API_ bool IsTrue(bool condition);
-
-// Defines scoped_ptr.
-
-// This implementation of scoped_ptr is PARTIAL - it only contains
-// enough stuff to satisfy Google Test's need.
-template <typename T>
-class scoped_ptr {
- public:
-  typedef T element_type;
-
-  explicit scoped_ptr(T* p = NULL) : ptr_(p) {}
-  ~scoped_ptr() { reset(); }
-
-  T& operator*() const { return *ptr_; }
-  T* operator->() const { return ptr_; }
-  T* get() const { return ptr_; }
-
-  T* release() {
-    T* const ptr = ptr_;
-    ptr_ = NULL;
-    return ptr;
-  }
-
-  void reset(T* p = NULL) {
-    if (p != ptr_) {
-      if (IsTrue(sizeof(T) > 0)) {  // Makes sure T is a complete type.
-        delete ptr_;
-      }
-      ptr_ = p;
-    }
-  }
-
-  friend void swap(scoped_ptr& a, scoped_ptr& b) {
-    using std::swap;
-    swap(a.ptr_, b.ptr_);
-  }
-
- private:
-  T* ptr_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(scoped_ptr);
-};
-
-// Defines RE.
-
-#if GTEST_USES_PCRE
-// if used, PCRE is injected by custom/gtest-port.h
-#elif GTEST_USES_POSIX_RE || GTEST_USES_SIMPLE_RE
-
-// A simple C++ wrapper for <regex.h>.  It uses the POSIX Extended
-// Regular Expression syntax.
-class GTEST_API_ RE {
- public:
-  // A copy constructor is required by the Standard to initialize object
-  // references from r-values.
-  RE(const RE& other) { Init(other.pattern()); }
-
-  // Constructs an RE from a string.
-  RE(const ::std::string& regex) { Init(regex.c_str()); }  // NOLINT
-
-# if GTEST_HAS_GLOBAL_STRING
-
-  RE(const ::string& regex) { Init(regex.c_str()); }  // NOLINT
-
-# endif  // GTEST_HAS_GLOBAL_STRING
-
-  RE(const char* regex) { Init(regex); }  // NOLINT
-  ~RE();
-
-  // Returns the string representation of the regex.
-  const char* pattern() const { return pattern_; }
-
-  // FullMatch(str, re) returns true iff regular expression re matches
-  // the entire str.
-  // PartialMatch(str, re) returns true iff regular expression re
-  // matches a substring of str (including str itself).
-  //
-  // TODO(wan@google.com): make FullMatch() and PartialMatch() work
-  // when str contains NUL characters.
-  static bool FullMatch(const ::std::string& str, const RE& re) {
-    return FullMatch(str.c_str(), re);
-  }
-  static bool PartialMatch(const ::std::string& str, const RE& re) {
-    return PartialMatch(str.c_str(), re);
-  }
-
-# if GTEST_HAS_GLOBAL_STRING
-
-  static bool FullMatch(const ::string& str, const RE& re) {
-    return FullMatch(str.c_str(), re);
-  }
-  static bool PartialMatch(const ::string& str, const RE& re) {
-    return PartialMatch(str.c_str(), re);
-  }
-
-# endif  // GTEST_HAS_GLOBAL_STRING
-
-  static bool FullMatch(const char* str, const RE& re);
-  static bool PartialMatch(const char* str, const RE& re);
-
- private:
-  void Init(const char* regex);
-
-  // We use a const char* instead of an std::string, as Google Test used to be
-  // used where std::string is not available.  TODO(wan@google.com): change to
-  // std::string.
-  const char* pattern_;
-  bool is_valid_;
-
-# if GTEST_USES_POSIX_RE
-
-  regex_t full_regex_;     // For FullMatch().
-  regex_t partial_regex_;  // For PartialMatch().
-
-# else  // GTEST_USES_SIMPLE_RE
-
-  const char* full_pattern_;  // For FullMatch();
-
-# endif
-
-  GTEST_DISALLOW_ASSIGN_(RE);
-};
-
-#endif  // GTEST_USES_PCRE
-
-// Formats a source file path and a line number as they would appear
-// in an error message from the compiler used to compile this code.
-GTEST_API_ ::std::string FormatFileLocation(const char* file, int line);
-
-// Formats a file location for compiler-independent XML output.
-// Although this function is not platform dependent, we put it next to
-// FormatFileLocation in order to contrast the two functions.
-GTEST_API_ ::std::string FormatCompilerIndependentFileLocation(const char* file,
-                                                               int line);
-
-// Defines logging utilities:
-//   GTEST_LOG_(severity) - logs messages at the specified severity level. The
-//                          message itself is streamed into the macro.
-//   LogToStderr()  - directs all log messages to stderr.
-//   FlushInfoLog() - flushes informational log messages.
-
-enum GTestLogSeverity {
-  GTEST_INFO,
-  GTEST_WARNING,
-  GTEST_ERROR,
-  GTEST_FATAL
-};
-
-// Formats log entry severity, provides a stream object for streaming the
-// log message, and terminates the message with a newline when going out of
-// scope.
-class GTEST_API_ GTestLog {
- public:
-  GTestLog(GTestLogSeverity severity, const char* file, int line);
-
-  // Flushes the buffers and, if severity is GTEST_FATAL, aborts the program.
-  ~GTestLog();
-
-  ::std::ostream& GetStream() { return ::std::cerr; }
-
- private:
-  const GTestLogSeverity severity_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(GTestLog);
-};
-
-#if !defined(GTEST_LOG_)
-
-# define GTEST_LOG_(severity) \
-    ::testing::internal::GTestLog(::testing::internal::GTEST_##severity, \
-                                  __FILE__, __LINE__).GetStream()
-
-inline void LogToStderr() {}
-inline void FlushInfoLog() { fflush(NULL); }
-
-#endif  // !defined(GTEST_LOG_)
-
-#if !defined(GTEST_CHECK_)
-// INTERNAL IMPLEMENTATION - DO NOT USE.
-//
-// GTEST_CHECK_ is an all-mode assert. It aborts the program if the condition
-// is not satisfied.
-//  Synopsys:
-//    GTEST_CHECK_(boolean_condition);
-//     or
-//    GTEST_CHECK_(boolean_condition) << "Additional message";
-//
-//    This checks the condition and if the condition is not satisfied
-//    it prints message about the condition violation, including the
-//    condition itself, plus additional message streamed into it, if any,
-//    and then it aborts the program. It aborts the program irrespective of
-//    whether it is built in the debug mode or not.
-# define GTEST_CHECK_(condition) \
-    GTEST_AMBIGUOUS_ELSE_BLOCKER_ \
-    if (::testing::internal::IsTrue(condition)) \
-      ; \
-    else \
-      GTEST_LOG_(FATAL) << "Condition " #condition " failed. "
-#endif  // !defined(GTEST_CHECK_)
-
-// An all-mode assert to verify that the given POSIX-style function
-// call returns 0 (indicating success).  Known limitation: this
-// doesn't expand to a balanced 'if' statement, so enclose the macro
-// in {} if you need to use it as the only statement in an 'if'
-// branch.
-#define GTEST_CHECK_POSIX_SUCCESS_(posix_call) \
-  if (const int gtest_error = (posix_call)) \
-    GTEST_LOG_(FATAL) << #posix_call << "failed with error " \
-                      << gtest_error
-
-// Adds reference to a type if it is not a reference type,
-// otherwise leaves it unchanged.  This is the same as
-// tr1::add_reference, which is not widely available yet.
-template <typename T>
-struct AddReference { typedef T& type; };  // NOLINT
-template <typename T>
-struct AddReference<T&> { typedef T& type; };  // NOLINT
-
-// A handy wrapper around AddReference that works when the argument T
-// depends on template parameters.
-#define GTEST_ADD_REFERENCE_(T) \
-    typename ::testing::internal::AddReference<T>::type
-
-// Transforms "T" into "const T&" according to standard reference collapsing
-// rules (this is only needed as a backport for C++98 compilers that do not
-// support reference collapsing). Specifically, it transforms:
-//
-//   char         ==> const char&
-//   const char   ==> const char&
-//   char&        ==> char&
-//   const char&  ==> const char&
-//
-// Note that the non-const reference will not have "const" added. This is
-// standard, and necessary so that "T" can always bind to "const T&".
-template <typename T>
-struct ConstRef { typedef const T& type; };
-template <typename T>
-struct ConstRef<T&> { typedef T& type; };
-
-// The argument T must depend on some template parameters.
-#define GTEST_REFERENCE_TO_CONST_(T) \
-  typename ::testing::internal::ConstRef<T>::type
-
-#if GTEST_HAS_STD_MOVE_
-using std::forward;
-using std::move;
-
-template <typename T>
-struct RvalueRef {
-  typedef T&& type;
-};
-#else  // GTEST_HAS_STD_MOVE_
-template <typename T>
-const T& move(const T& t) {
-  return t;
-}
-template <typename T>
-GTEST_ADD_REFERENCE_(T) forward(GTEST_ADD_REFERENCE_(T) t) { return t; }
-
-template <typename T>
-struct RvalueRef {
-  typedef const T& type;
-};
-#endif  // GTEST_HAS_STD_MOVE_
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// Use ImplicitCast_ as a safe version of static_cast for upcasting in
-// the type hierarchy (e.g. casting a Foo* to a SuperclassOfFoo* or a
-// const Foo*).  When you use ImplicitCast_, the compiler checks that
-// the cast is safe.  Such explicit ImplicitCast_s are necessary in
-// surprisingly many situations where C++ demands an exact type match
-// instead of an argument type convertable to a target type.
-//
-// The syntax for using ImplicitCast_ is the same as for static_cast:
-//
-//   ImplicitCast_<ToType>(expr)
-//
-// ImplicitCast_ would have been part of the C++ standard library,
-// but the proposal was submitted too late.  It will probably make
-// its way into the language in the future.
-//
-// This relatively ugly name is intentional. It prevents clashes with
-// similar functions users may have (e.g., implicit_cast). The internal
-// namespace alone is not enough because the function can be found by ADL.
-template<typename To>
-inline To ImplicitCast_(To x) { return x; }
-
-// When you upcast (that is, cast a pointer from type Foo to type
-// SuperclassOfFoo), it's fine to use ImplicitCast_<>, since upcasts
-// always succeed.  When you downcast (that is, cast a pointer from
-// type Foo to type SubclassOfFoo), static_cast<> isn't safe, because
-// how do you know the pointer is really of type SubclassOfFoo?  It
-// could be a bare Foo, or of type DifferentSubclassOfFoo.  Thus,
-// when you downcast, you should use this macro.  In debug mode, we
-// use dynamic_cast<> to double-check the downcast is legal (we die
-// if it's not).  In normal mode, we do the efficient static_cast<>
-// instead.  Thus, it's important to test in debug mode to make sure
-// the cast is legal!
-//    This is the only place in the code we should use dynamic_cast<>.
-// In particular, you SHOULDN'T be using dynamic_cast<> in order to
-// do RTTI (eg code like this:
-//    if (dynamic_cast<Subclass1>(foo)) HandleASubclass1Object(foo);
-//    if (dynamic_cast<Subclass2>(foo)) HandleASubclass2Object(foo);
-// You should design the code some other way not to need this.
-//
-// This relatively ugly name is intentional. It prevents clashes with
-// similar functions users may have (e.g., down_cast). The internal
-// namespace alone is not enough because the function can be found by ADL.
-template<typename To, typename From>  // use like this: DownCast_<T*>(foo);
-inline To DownCast_(From* f) {  // so we only accept pointers
-  // Ensures that To is a sub-type of From *.  This test is here only
-  // for compile-time type checking, and has no overhead in an
-  // optimized build at run-time, as it will be optimized away
-  // completely.
-  GTEST_INTENTIONAL_CONST_COND_PUSH_()
-  if (false) {
-  GTEST_INTENTIONAL_CONST_COND_POP_()
-    const To to = NULL;
-    ::testing::internal::ImplicitCast_<From*>(to);
-  }
-
-#if GTEST_HAS_RTTI
-  // RTTI: debug mode only!
-  GTEST_CHECK_(f == NULL || dynamic_cast<To>(f) != NULL);
-#endif
-  return static_cast<To>(f);
-}
-
-// Downcasts the pointer of type Base to Derived.
-// Derived must be a subclass of Base. The parameter MUST
-// point to a class of type Derived, not any subclass of it.
-// When RTTI is available, the function performs a runtime
-// check to enforce this.
-template <class Derived, class Base>
-Derived* CheckedDowncastToActualType(Base* base) {
-#if GTEST_HAS_RTTI
-  GTEST_CHECK_(typeid(*base) == typeid(Derived));
-#endif
-
-#if GTEST_HAS_DOWNCAST_
-  return ::down_cast<Derived*>(base);
-#elif GTEST_HAS_RTTI
-  return dynamic_cast<Derived*>(base);  // NOLINT
-#else
-  return static_cast<Derived*>(base);  // Poor man's downcast.
-#endif
-}
-
-#if GTEST_HAS_STREAM_REDIRECTION
-
-// Defines the stderr capturer:
-//   CaptureStdout     - starts capturing stdout.
-//   GetCapturedStdout - stops capturing stdout and returns the captured string.
-//   CaptureStderr     - starts capturing stderr.
-//   GetCapturedStderr - stops capturing stderr and returns the captured string.
-//
-GTEST_API_ void CaptureStdout();
-GTEST_API_ std::string GetCapturedStdout();
-GTEST_API_ void CaptureStderr();
-GTEST_API_ std::string GetCapturedStderr();
-
-#endif  // GTEST_HAS_STREAM_REDIRECTION
-// Returns the size (in bytes) of a file.
-GTEST_API_ size_t GetFileSize(FILE* file);
-
-// Reads the entire content of a file as a string.
-GTEST_API_ std::string ReadEntireFile(FILE* file);
-
-// All command line arguments.
-GTEST_API_ std::vector<std::string> GetArgvs();
-
-#if GTEST_HAS_DEATH_TEST
-
-std::vector<std::string> GetInjectableArgvs();
-// Deprecated: pass the args vector by value instead.
-void SetInjectableArgvs(const std::vector<std::string>* new_argvs);
-void SetInjectableArgvs(const std::vector<std::string>& new_argvs);
-#if GTEST_HAS_GLOBAL_STRING
-void SetInjectableArgvs(const std::vector< ::string>& new_argvs);
-#endif  // GTEST_HAS_GLOBAL_STRING
-void ClearInjectableArgvs();
-
-#endif  // GTEST_HAS_DEATH_TEST
-
-// Defines synchronization primitives.
-#if GTEST_IS_THREADSAFE
-# if GTEST_HAS_PTHREAD
-// Sleeps for (roughly) n milliseconds.  This function is only for testing
-// Google Test's own constructs.  Don't use it in user tests, either
-// directly or indirectly.
-inline void SleepMilliseconds(int n) {
-  const timespec time = {
-    0,                  // 0 seconds.
-    n * 1000L * 1000L,  // And n ms.
-  };
-  nanosleep(&time, NULL);
-}
-# endif  // GTEST_HAS_PTHREAD
-
-# if GTEST_HAS_NOTIFICATION_
-// Notification has already been imported into the namespace.
-// Nothing to do here.
-
-# elif GTEST_HAS_PTHREAD
-// Allows a controller thread to pause execution of newly created
-// threads until notified.  Instances of this class must be created
-// and destroyed in the controller thread.
-//
-// This class is only for testing Google Test's own constructs. Do not
-// use it in user tests, either directly or indirectly.
-class Notification {
- public:
-  Notification() : notified_(false) {
-    GTEST_CHECK_POSIX_SUCCESS_(pthread_mutex_init(&mutex_, NULL));
-  }
-  ~Notification() {
-    pthread_mutex_destroy(&mutex_);
-  }
-
-  // Notifies all threads created with this notification to start. Must
-  // be called from the controller thread.
-  void Notify() {
-    pthread_mutex_lock(&mutex_);
-    notified_ = true;
-    pthread_mutex_unlock(&mutex_);
-  }
-
-  // Blocks until the controller thread notifies. Must be called from a test
-  // thread.
-  void WaitForNotification() {
-    for (;;) {
-      pthread_mutex_lock(&mutex_);
-      const bool notified = notified_;
-      pthread_mutex_unlock(&mutex_);
-      if (notified)
-        break;
-      SleepMilliseconds(10);
-    }
-  }
-
- private:
-  pthread_mutex_t mutex_;
-  bool notified_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(Notification);
-};
-
-# elif GTEST_OS_WINDOWS && !GTEST_OS_WINDOWS_PHONE && !GTEST_OS_WINDOWS_RT
-
-GTEST_API_ void SleepMilliseconds(int n);
-
-// Provides leak-safe Windows kernel handle ownership.
-// Used in death tests and in threading support.
-class GTEST_API_ AutoHandle {
- public:
-  // Assume that Win32 HANDLE type is equivalent to void*. Doing so allows us to
-  // avoid including <windows.h> in this header file. Including <windows.h> is
-  // undesirable because it defines a lot of symbols and macros that tend to
-  // conflict with client code. This assumption is verified by
-  // WindowsTypesTest.HANDLEIsVoidStar.
-  typedef void* Handle;
-  AutoHandle();
-  explicit AutoHandle(Handle handle);
-
-  ~AutoHandle();
-
-  Handle Get() const;
-  void Reset();
-  void Reset(Handle handle);
-
- private:
-  // Returns true iff the handle is a valid handle object that can be closed.
-  bool IsCloseable() const;
-
-  Handle handle_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(AutoHandle);
-};
-
-// Allows a controller thread to pause execution of newly created
-// threads until notified.  Instances of this class must be created
-// and destroyed in the controller thread.
-//
-// This class is only for testing Google Test's own constructs. Do not
-// use it in user tests, either directly or indirectly.
-class GTEST_API_ Notification {
- public:
-  Notification();
-  void Notify();
-  void WaitForNotification();
-
- private:
-  AutoHandle event_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(Notification);
-};
-# endif  // GTEST_HAS_NOTIFICATION_
-
-// On MinGW, we can have both GTEST_OS_WINDOWS and GTEST_HAS_PTHREAD
-// defined, but we don't want to use MinGW's pthreads implementation, which
-// has conformance problems with some versions of the POSIX standard.
-# if GTEST_HAS_PTHREAD && !GTEST_OS_WINDOWS_MINGW
-
-// As a C-function, ThreadFuncWithCLinkage cannot be templated itself.
-// Consequently, it cannot select a correct instantiation of ThreadWithParam
-// in order to call its Run(). Introducing ThreadWithParamBase as a
-// non-templated base class for ThreadWithParam allows us to bypass this
-// problem.
-class ThreadWithParamBase {
- public:
-  virtual ~ThreadWithParamBase() {}
-  virtual void Run() = 0;
-};
-
-// pthread_create() accepts a pointer to a function type with the C linkage.
-// According to the Standard (7.5/1), function types with different linkages
-// are different even if they are otherwise identical.  Some compilers (for
-// example, SunStudio) treat them as different types.  Since class methods
-// cannot be defined with C-linkage we need to define a free C-function to
-// pass into pthread_create().
-extern "C" inline void* ThreadFuncWithCLinkage(void* thread) {
-  static_cast<ThreadWithParamBase*>(thread)->Run();
-  return NULL;
-}
-
-// Helper class for testing Google Test's multi-threading constructs.
-// To use it, write:
-//
-//   void ThreadFunc(int param) { /* Do things with param */ }
-//   Notification thread_can_start;
-//   ...
-//   // The thread_can_start parameter is optional; you can supply NULL.
-//   ThreadWithParam<int> thread(&ThreadFunc, 5, &thread_can_start);
-//   thread_can_start.Notify();
-//
-// These classes are only for testing Google Test's own constructs. Do
-// not use them in user tests, either directly or indirectly.
-template <typename T>
-class ThreadWithParam : public ThreadWithParamBase {
- public:
-  typedef void UserThreadFunc(T);
-
-  ThreadWithParam(UserThreadFunc* func, T param, Notification* thread_can_start)
-      : func_(func),
-        param_(param),
-        thread_can_start_(thread_can_start),
-        finished_(false) {
-    ThreadWithParamBase* const base = this;
-    // The thread can be created only after all fields except thread_
-    // have been initialized.
-    GTEST_CHECK_POSIX_SUCCESS_(
-        pthread_create(&thread_, 0, &ThreadFuncWithCLinkage, base));
-  }
-  ~ThreadWithParam() { Join(); }
-
-  void Join() {
-    if (!finished_) {
-      GTEST_CHECK_POSIX_SUCCESS_(pthread_join(thread_, 0));
-      finished_ = true;
-    }
-  }
-
-  virtual void Run() {
-    if (thread_can_start_ != NULL)
-      thread_can_start_->WaitForNotification();
-    func_(param_);
-  }
-
- private:
-  UserThreadFunc* const func_;  // User-supplied thread function.
-  const T param_;  // User-supplied parameter to the thread function.
-  // When non-NULL, used to block execution until the controller thread
-  // notifies.
-  Notification* const thread_can_start_;
-  bool finished_;  // true iff we know that the thread function has finished.
-  pthread_t thread_;  // The native thread object.
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ThreadWithParam);
-};
-# endif  // !GTEST_OS_WINDOWS && GTEST_HAS_PTHREAD ||
-         // GTEST_HAS_MUTEX_AND_THREAD_LOCAL_
-
-# if GTEST_HAS_MUTEX_AND_THREAD_LOCAL_
-// Mutex and ThreadLocal have already been imported into the namespace.
-// Nothing to do here.
-
-# elif GTEST_OS_WINDOWS && !GTEST_OS_WINDOWS_PHONE && !GTEST_OS_WINDOWS_RT
-
-// Mutex implements mutex on Windows platforms.  It is used in conjunction
-// with class MutexLock:
-//
-//   Mutex mutex;
-//   ...
-//   MutexLock lock(&mutex);  // Acquires the mutex and releases it at the
-//                            // end of the current scope.
-//
-// A static Mutex *must* be defined or declared using one of the following
-// macros:
-//   GTEST_DEFINE_STATIC_MUTEX_(g_some_mutex);
-//   GTEST_DECLARE_STATIC_MUTEX_(g_some_mutex);
-//
-// (A non-static Mutex is defined/declared in the usual way).
-class GTEST_API_ Mutex {
- public:
-  enum MutexType { kStatic = 0, kDynamic = 1 };
-  // We rely on kStaticMutex being 0 as it is to what the linker initializes
-  // type_ in static mutexes.  critical_section_ will be initialized lazily
-  // in ThreadSafeLazyInit().
-  enum StaticConstructorSelector { kStaticMutex = 0 };
-
-  // This constructor intentionally does nothing.  It relies on type_ being
-  // statically initialized to 0 (effectively setting it to kStatic) and on
-  // ThreadSafeLazyInit() to lazily initialize the rest of the members.
-  explicit Mutex(StaticConstructorSelector /*dummy*/) {}
-
-  Mutex();
-  ~Mutex();
-
-  void Lock();
-
-  void Unlock();
-
-  // Does nothing if the current thread holds the mutex. Otherwise, crashes
-  // with high probability.
-  void AssertHeld();
-
- private:
-  // Initializes owner_thread_id_ and critical_section_ in static mutexes.
-  void ThreadSafeLazyInit();
-
-  // Per http://blogs.msdn.com/b/oldnewthing/archive/2004/02/23/78395.aspx,
-  // we assume that 0 is an invalid value for thread IDs.
-  unsigned int owner_thread_id_;
-
-  // For static mutexes, we rely on these members being initialized to zeros
-  // by the linker.
-  MutexType type_;
-  long critical_section_init_phase_;  // NOLINT
-  GTEST_CRITICAL_SECTION* critical_section_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(Mutex);
-};
-
-# define GTEST_DECLARE_STATIC_MUTEX_(mutex) \
-    extern ::testing::internal::Mutex mutex
-
-# define GTEST_DEFINE_STATIC_MUTEX_(mutex) \
-    ::testing::internal::Mutex mutex(::testing::internal::Mutex::kStaticMutex)
-
-// We cannot name this class MutexLock because the ctor declaration would
-// conflict with a macro named MutexLock, which is defined on some
-// platforms. That macro is used as a defensive measure to prevent against
-// inadvertent misuses of MutexLock like "MutexLock(&mu)" rather than
-// "MutexLock l(&mu)".  Hence the typedef trick below.
-class GTestMutexLock {
- public:
-  explicit GTestMutexLock(Mutex* mutex)
-      : mutex_(mutex) { mutex_->Lock(); }
-
-  ~GTestMutexLock() { mutex_->Unlock(); }
-
- private:
-  Mutex* const mutex_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(GTestMutexLock);
-};
-
-typedef GTestMutexLock MutexLock;
-
-// Base class for ValueHolder<T>.  Allows a caller to hold and delete a value
-// without knowing its type.
-class ThreadLocalValueHolderBase {
- public:
-  virtual ~ThreadLocalValueHolderBase() {}
-};
-
-// Provides a way for a thread to send notifications to a ThreadLocal
-// regardless of its parameter type.
-class ThreadLocalBase {
- public:
-  // Creates a new ValueHolder<T> object holding a default value passed to
-  // this ThreadLocal<T>'s constructor and returns it.  It is the caller's
-  // responsibility not to call this when the ThreadLocal<T> instance already
-  // has a value on the current thread.
-  virtual ThreadLocalValueHolderBase* NewValueForCurrentThread() const = 0;
-
- protected:
-  ThreadLocalBase() {}
-  virtual ~ThreadLocalBase() {}
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ThreadLocalBase);
-};
-
-// Maps a thread to a set of ThreadLocals that have values instantiated on that
-// thread and notifies them when the thread exits.  A ThreadLocal instance is
-// expected to persist until all threads it has values on have terminated.
-class GTEST_API_ ThreadLocalRegistry {
- public:
-  // Registers thread_local_instance as having value on the current thread.
-  // Returns a value that can be used to identify the thread from other threads.
-  static ThreadLocalValueHolderBase* GetValueOnCurrentThread(
-      const ThreadLocalBase* thread_local_instance);
-
-  // Invoked when a ThreadLocal instance is destroyed.
-  static void OnThreadLocalDestroyed(
-      const ThreadLocalBase* thread_local_instance);
-};
-
-class GTEST_API_ ThreadWithParamBase {
- public:
-  void Join();
-
- protected:
-  class Runnable {
-   public:
-    virtual ~Runnable() {}
-    virtual void Run() = 0;
-  };
-
-  ThreadWithParamBase(Runnable *runnable, Notification* thread_can_start);
-  virtual ~ThreadWithParamBase();
-
- private:
-  AutoHandle thread_;
-};
-
-// Helper class for testing Google Test's multi-threading constructs.
-template <typename T>
-class ThreadWithParam : public ThreadWithParamBase {
- public:
-  typedef void UserThreadFunc(T);
-
-  ThreadWithParam(UserThreadFunc* func, T param, Notification* thread_can_start)
-      : ThreadWithParamBase(new RunnableImpl(func, param), thread_can_start) {
-  }
-  virtual ~ThreadWithParam() {}
-
- private:
-  class RunnableImpl : public Runnable {
-   public:
-    RunnableImpl(UserThreadFunc* func, T param)
-        : func_(func),
-          param_(param) {
-    }
-    virtual ~RunnableImpl() {}
-    virtual void Run() {
-      func_(param_);
-    }
-
-   private:
-    UserThreadFunc* const func_;
-    const T param_;
-
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(RunnableImpl);
-  };
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ThreadWithParam);
-};
-
-// Implements thread-local storage on Windows systems.
-//
-//   // Thread 1
-//   ThreadLocal<int> tl(100);  // 100 is the default value for each thread.
-//
-//   // Thread 2
-//   tl.set(150);  // Changes the value for thread 2 only.
-//   EXPECT_EQ(150, tl.get());
-//
-//   // Thread 1
-//   EXPECT_EQ(100, tl.get());  // In thread 1, tl has the original value.
-//   tl.set(200);
-//   EXPECT_EQ(200, tl.get());
-//
-// The template type argument T must have a public copy constructor.
-// In addition, the default ThreadLocal constructor requires T to have
-// a public default constructor.
-//
-// The users of a TheadLocal instance have to make sure that all but one
-// threads (including the main one) using that instance have exited before
-// destroying it. Otherwise, the per-thread objects managed for them by the
-// ThreadLocal instance are not guaranteed to be destroyed on all platforms.
-//
-// Google Test only uses global ThreadLocal objects.  That means they
-// will die after main() has returned.  Therefore, no per-thread
-// object managed by Google Test will be leaked as long as all threads
-// using Google Test have exited when main() returns.
-template <typename T>
-class ThreadLocal : public ThreadLocalBase {
- public:
-  ThreadLocal() : default_factory_(new DefaultValueHolderFactory()) {}
-  explicit ThreadLocal(const T& value)
-      : default_factory_(new InstanceValueHolderFactory(value)) {}
-
-  ~ThreadLocal() { ThreadLocalRegistry::OnThreadLocalDestroyed(this); }
-
-  T* pointer() { return GetOrCreateValue(); }
-  const T* pointer() const { return GetOrCreateValue(); }
-  const T& get() const { return *pointer(); }
-  void set(const T& value) { *pointer() = value; }
-
- private:
-  // Holds a value of T.  Can be deleted via its base class without the caller
-  // knowing the type of T.
-  class ValueHolder : public ThreadLocalValueHolderBase {
-   public:
-    ValueHolder() : value_() {}
-    explicit ValueHolder(const T& value) : value_(value) {}
-
-    T* pointer() { return &value_; }
-
-   private:
-    T value_;
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(ValueHolder);
-  };
-
-
-  T* GetOrCreateValue() const {
-    return static_cast<ValueHolder*>(
-        ThreadLocalRegistry::GetValueOnCurrentThread(this))->pointer();
-  }
-
-  virtual ThreadLocalValueHolderBase* NewValueForCurrentThread() const {
-    return default_factory_->MakeNewHolder();
-  }
-
-  class ValueHolderFactory {
-   public:
-    ValueHolderFactory() {}
-    virtual ~ValueHolderFactory() {}
-    virtual ValueHolder* MakeNewHolder() const = 0;
-
-   private:
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(ValueHolderFactory);
-  };
-
-  class DefaultValueHolderFactory : public ValueHolderFactory {
-   public:
-    DefaultValueHolderFactory() {}
-    virtual ValueHolder* MakeNewHolder() const { return new ValueHolder(); }
-
-   private:
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(DefaultValueHolderFactory);
-  };
-
-  class InstanceValueHolderFactory : public ValueHolderFactory {
-   public:
-    explicit InstanceValueHolderFactory(const T& value) : value_(value) {}
-    virtual ValueHolder* MakeNewHolder() const {
-      return new ValueHolder(value_);
-    }
-
-   private:
-    const T value_;  // The value for each thread.
-
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(InstanceValueHolderFactory);
-  };
-
-  scoped_ptr<ValueHolderFactory> default_factory_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ThreadLocal);
-};
-
-# elif GTEST_HAS_PTHREAD
-
-// MutexBase and Mutex implement mutex on pthreads-based platforms.
-class MutexBase {
- public:
-  // Acquires this mutex.
-  void Lock() {
-    GTEST_CHECK_POSIX_SUCCESS_(pthread_mutex_lock(&mutex_));
-    owner_ = pthread_self();
-    has_owner_ = true;
-  }
-
-  // Releases this mutex.
-  void Unlock() {
-    // Since the lock is being released the owner_ field should no longer be
-    // considered valid. We don't protect writing to has_owner_ here, as it's
-    // the caller's responsibility to ensure that the current thread holds the
-    // mutex when this is called.
-    has_owner_ = false;
-    GTEST_CHECK_POSIX_SUCCESS_(pthread_mutex_unlock(&mutex_));
-  }
-
-  // Does nothing if the current thread holds the mutex. Otherwise, crashes
-  // with high probability.
-  void AssertHeld() const {
-    GTEST_CHECK_(has_owner_ && pthread_equal(owner_, pthread_self()))
-        << "The current thread is not holding the mutex @" << this;
-  }
-
-  // A static mutex may be used before main() is entered.  It may even
-  // be used before the dynamic initialization stage.  Therefore we
-  // must be able to initialize a static mutex object at link time.
-  // This means MutexBase has to be a POD and its member variables
-  // have to be public.
- public:
-  pthread_mutex_t mutex_;  // The underlying pthread mutex.
-  // has_owner_ indicates whether the owner_ field below contains a valid thread
-  // ID and is therefore safe to inspect (e.g., to use in pthread_equal()). All
-  // accesses to the owner_ field should be protected by a check of this field.
-  // An alternative might be to memset() owner_ to all zeros, but there's no
-  // guarantee that a zero'd pthread_t is necessarily invalid or even different
-  // from pthread_self().
-  bool has_owner_;
-  pthread_t owner_;  // The thread holding the mutex.
-};
-
-// Forward-declares a static mutex.
-#  define GTEST_DECLARE_STATIC_MUTEX_(mutex) \
-     extern ::testing::internal::MutexBase mutex
-
-// Defines and statically (i.e. at link time) initializes a static mutex.
-// The initialization list here does not explicitly initialize each field,
-// instead relying on default initialization for the unspecified fields. In
-// particular, the owner_ field (a pthread_t) is not explicitly initialized.
-// This allows initialization to work whether pthread_t is a scalar or struct.
-// The flag -Wmissing-field-initializers must not be specified for this to work.
-#  define GTEST_DEFINE_STATIC_MUTEX_(mutex) \
-     ::testing::internal::MutexBase mutex = { PTHREAD_MUTEX_INITIALIZER, false }
-
-// The Mutex class can only be used for mutexes created at runtime. It
-// shares its API with MutexBase otherwise.
-class Mutex : public MutexBase {
- public:
-  Mutex() {
-    GTEST_CHECK_POSIX_SUCCESS_(pthread_mutex_init(&mutex_, NULL));
-    has_owner_ = false;
-  }
-  ~Mutex() {
-    GTEST_CHECK_POSIX_SUCCESS_(pthread_mutex_destroy(&mutex_));
-  }
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(Mutex);
-};
-
-// We cannot name this class MutexLock because the ctor declaration would
-// conflict with a macro named MutexLock, which is defined on some
-// platforms. That macro is used as a defensive measure to prevent against
-// inadvertent misuses of MutexLock like "MutexLock(&mu)" rather than
-// "MutexLock l(&mu)".  Hence the typedef trick below.
-class GTestMutexLock {
- public:
-  explicit GTestMutexLock(MutexBase* mutex)
-      : mutex_(mutex) { mutex_->Lock(); }
-
-  ~GTestMutexLock() { mutex_->Unlock(); }
-
- private:
-  MutexBase* const mutex_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(GTestMutexLock);
-};
-
-typedef GTestMutexLock MutexLock;
-
-// Helpers for ThreadLocal.
-
-// pthread_key_create() requires DeleteThreadLocalValue() to have
-// C-linkage.  Therefore it cannot be templatized to access
-// ThreadLocal<T>.  Hence the need for class
-// ThreadLocalValueHolderBase.
-class ThreadLocalValueHolderBase {
- public:
-  virtual ~ThreadLocalValueHolderBase() {}
-};
-
-// Called by pthread to delete thread-local data stored by
-// pthread_setspecific().
-extern "C" inline void DeleteThreadLocalValue(void* value_holder) {
-  delete static_cast<ThreadLocalValueHolderBase*>(value_holder);
-}
-
-// Implements thread-local storage on pthreads-based systems.
-template <typename T>
-class GTEST_API_ ThreadLocal {
- public:
-  ThreadLocal()
-      : key_(CreateKey()), default_factory_(new DefaultValueHolderFactory()) {}
-  explicit ThreadLocal(const T& value)
-      : key_(CreateKey()),
-        default_factory_(new InstanceValueHolderFactory(value)) {}
-
-  ~ThreadLocal() {
-    // Destroys the managed object for the current thread, if any.
-    DeleteThreadLocalValue(pthread_getspecific(key_));
-
-    // Releases resources associated with the key.  This will *not*
-    // delete managed objects for other threads.
-    GTEST_CHECK_POSIX_SUCCESS_(pthread_key_delete(key_));
-  }
-
-  T* pointer() { return GetOrCreateValue(); }
-  const T* pointer() const { return GetOrCreateValue(); }
-  const T& get() const { return *pointer(); }
-  void set(const T& value) { *pointer() = value; }
-
- private:
-  // Holds a value of type T.
-  class ValueHolder : public ThreadLocalValueHolderBase {
-   public:
-    ValueHolder() : value_() {}
-    explicit ValueHolder(const T& value) : value_(value) {}
-
-    T* pointer() { return &value_; }
-
-   private:
-    T value_;
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(ValueHolder);
-  };
-
-  static pthread_key_t CreateKey() {
-    pthread_key_t key;
-    // When a thread exits, DeleteThreadLocalValue() will be called on
-    // the object managed for that thread.
-    GTEST_CHECK_POSIX_SUCCESS_(
-        pthread_key_create(&key, &DeleteThreadLocalValue));
-    return key;
-  }
-
-  T* GetOrCreateValue() const {
-    ThreadLocalValueHolderBase* const holder =
-        static_cast<ThreadLocalValueHolderBase*>(pthread_getspecific(key_));
-    if (holder != NULL) {
-      return CheckedDowncastToActualType<ValueHolder>(holder)->pointer();
-    }
-
-    ValueHolder* const new_holder = default_factory_->MakeNewHolder();
-    ThreadLocalValueHolderBase* const holder_base = new_holder;
-    GTEST_CHECK_POSIX_SUCCESS_(pthread_setspecific(key_, holder_base));
-    return new_holder->pointer();
-  }
-
-  class ValueHolderFactory {
-   public:
-    ValueHolderFactory() {}
-    virtual ~ValueHolderFactory() {}
-    virtual ValueHolder* MakeNewHolder() const = 0;
-
-   private:
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(ValueHolderFactory);
-  };
-
-  class DefaultValueHolderFactory : public ValueHolderFactory {
-   public:
-    DefaultValueHolderFactory() {}
-    virtual ValueHolder* MakeNewHolder() const { return new ValueHolder(); }
-
-   private:
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(DefaultValueHolderFactory);
-  };
-
-  class InstanceValueHolderFactory : public ValueHolderFactory {
-   public:
-    explicit InstanceValueHolderFactory(const T& value) : value_(value) {}
-    virtual ValueHolder* MakeNewHolder() const {
-      return new ValueHolder(value_);
-    }
-
-   private:
-    const T value_;  // The value for each thread.
-
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(InstanceValueHolderFactory);
-  };
-
-  // A key pthreads uses for looking up per-thread values.
-  const pthread_key_t key_;
-  scoped_ptr<ValueHolderFactory> default_factory_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ThreadLocal);
-};
-
-# endif  // GTEST_HAS_MUTEX_AND_THREAD_LOCAL_
-
-#else  // GTEST_IS_THREADSAFE
-
-// A dummy implementation of synchronization primitives (mutex, lock,
-// and thread-local variable).  Necessary for compiling Google Test where
-// mutex is not supported - using Google Test in multiple threads is not
-// supported on such platforms.
-
-class Mutex {
- public:
-  Mutex() {}
-  void Lock() {}
-  void Unlock() {}
-  void AssertHeld() const {}
-};
-
-# define GTEST_DECLARE_STATIC_MUTEX_(mutex) \
-  extern ::testing::internal::Mutex mutex
-
-# define GTEST_DEFINE_STATIC_MUTEX_(mutex) ::testing::internal::Mutex mutex
-
-// We cannot name this class MutexLock because the ctor declaration would
-// conflict with a macro named MutexLock, which is defined on some
-// platforms. That macro is used as a defensive measure to prevent against
-// inadvertent misuses of MutexLock like "MutexLock(&mu)" rather than
-// "MutexLock l(&mu)".  Hence the typedef trick below.
-class GTestMutexLock {
- public:
-  explicit GTestMutexLock(Mutex*) {}  // NOLINT
-};
-
-typedef GTestMutexLock MutexLock;
-
-template <typename T>
-class GTEST_API_ ThreadLocal {
- public:
-  ThreadLocal() : value_() {}
-  explicit ThreadLocal(const T& value) : value_(value) {}
-  T* pointer() { return &value_; }
-  const T* pointer() const { return &value_; }
-  const T& get() const { return value_; }
-  void set(const T& value) { value_ = value; }
- private:
-  T value_;
-};
-
-#endif  // GTEST_IS_THREADSAFE
-
-// Returns the number of threads running in the process, or 0 to indicate that
-// we cannot detect it.
-GTEST_API_ size_t GetThreadCount();
-
-// Passing non-POD classes through ellipsis (...) crashes the ARM
-// compiler and generates a warning in Sun Studio before 12u4. The Nokia Symbian
-// and the IBM XL C/C++ compiler try to instantiate a copy constructor
-// for objects passed through ellipsis (...), failing for uncopyable
-// objects.  We define this to ensure that only POD is passed through
-// ellipsis on these systems.
-#if defined(__SYMBIAN32__) || defined(__IBMCPP__) || \
-     (defined(__SUNPRO_CC) && __SUNPRO_CC < 0x5130)
-// We lose support for NULL detection where the compiler doesn't like
-// passing non-POD classes through ellipsis (...).
-# define GTEST_ELLIPSIS_NEEDS_POD_ 1
-#else
-# define GTEST_CAN_COMPARE_NULL 1
-#endif
-
-// The Nokia Symbian and IBM XL C/C++ compilers cannot decide between
-// const T& and const T* in a function template.  These compilers
-// _can_ decide between class template specializations for T and T*,
-// so a tr1::type_traits-like is_pointer works.
-#if defined(__SYMBIAN32__) || defined(__IBMCPP__)
-# define GTEST_NEEDS_IS_POINTER_ 1
-#endif
-
-template <bool bool_value>
-struct bool_constant {
-  typedef bool_constant<bool_value> type;
-  static const bool value = bool_value;
-};
-template <bool bool_value> const bool bool_constant<bool_value>::value;
-
-typedef bool_constant<false> false_type;
-typedef bool_constant<true> true_type;
-
-template <typename T, typename U>
-struct is_same : public false_type {};
-
-template <typename T>
-struct is_same<T, T> : public true_type {};
-
-
-template <typename T>
-struct is_pointer : public false_type {};
-
-template <typename T>
-struct is_pointer<T*> : public true_type {};
-
-template <typename Iterator>
-struct IteratorTraits {
-  typedef typename Iterator::value_type value_type;
-};
-
-
-template <typename T>
-struct IteratorTraits<T*> {
-  typedef T value_type;
-};
-
-template <typename T>
-struct IteratorTraits<const T*> {
-  typedef T value_type;
-};
-
-#if GTEST_OS_WINDOWS
-# define GTEST_PATH_SEP_ "\\"
-# define GTEST_HAS_ALT_PATH_SEP_ 1
-// The biggest signed integer type the compiler supports.
-typedef __int64 BiggestInt;
-#else
-# define GTEST_PATH_SEP_ "/"
-# define GTEST_HAS_ALT_PATH_SEP_ 0
-typedef long long BiggestInt;  // NOLINT
-#endif  // GTEST_OS_WINDOWS
-
-// Utilities for char.
-
-// isspace(int ch) and friends accept an unsigned char or EOF.  char
-// may be signed, depending on the compiler (or compiler flags).
-// Therefore we need to cast a char to unsigned char before calling
-// isspace(), etc.
-
-inline bool IsAlpha(char ch) {
-  return isalpha(static_cast<unsigned char>(ch)) != 0;
-}
-inline bool IsAlNum(char ch) {
-  return isalnum(static_cast<unsigned char>(ch)) != 0;
-}
-inline bool IsDigit(char ch) {
-  return isdigit(static_cast<unsigned char>(ch)) != 0;
-}
-inline bool IsLower(char ch) {
-  return islower(static_cast<unsigned char>(ch)) != 0;
-}
-inline bool IsSpace(char ch) {
-  return isspace(static_cast<unsigned char>(ch)) != 0;
-}
-inline bool IsUpper(char ch) {
-  return isupper(static_cast<unsigned char>(ch)) != 0;
-}
-inline bool IsXDigit(char ch) {
-  return isxdigit(static_cast<unsigned char>(ch)) != 0;
-}
-inline bool IsXDigit(wchar_t ch) {
-  const unsigned char low_byte = static_cast<unsigned char>(ch);
-  return ch == low_byte && isxdigit(low_byte) != 0;
-}
-
-inline char ToLower(char ch) {
-  return static_cast<char>(tolower(static_cast<unsigned char>(ch)));
-}
-inline char ToUpper(char ch) {
-  return static_cast<char>(toupper(static_cast<unsigned char>(ch)));
-}
-
-inline std::string StripTrailingSpaces(std::string str) {
-  std::string::iterator it = str.end();
-  while (it != str.begin() && IsSpace(*--it))
-    it = str.erase(it);
-  return str;
-}
-
-// The testing::internal::posix namespace holds wrappers for common
-// POSIX functions.  These wrappers hide the differences between
-// Windows/MSVC and POSIX systems.  Since some compilers define these
-// standard functions as macros, the wrapper cannot have the same name
-// as the wrapped function.
-
-namespace posix {
-
-// Functions with a different name on Windows.
-
-#if GTEST_OS_WINDOWS
-
-typedef struct _stat StatStruct;
-
-# ifdef __BORLANDC__
-inline int IsATTY(int fd) { return isatty(fd); }
-inline int StrCaseCmp(const char* s1, const char* s2) {
-  return stricmp(s1, s2);
-}
-inline char* StrDup(const char* src) { return strdup(src); }
-# else  // !__BORLANDC__
-#  if GTEST_OS_WINDOWS_MOBILE
-inline int IsATTY(int /* fd */) { return 0; }
-#  else
-inline int IsATTY(int fd) { return _isatty(fd); }
-#  endif  // GTEST_OS_WINDOWS_MOBILE
-inline int StrCaseCmp(const char* s1, const char* s2) {
-  return _stricmp(s1, s2);
-}
-inline char* StrDup(const char* src) { return _strdup(src); }
-# endif  // __BORLANDC__
-
-# if GTEST_OS_WINDOWS_MOBILE
-inline int FileNo(FILE* file) { return reinterpret_cast<int>(_fileno(file)); }
-// Stat(), RmDir(), and IsDir() are not needed on Windows CE at this
-// time and thus not defined there.
-# else
-inline int FileNo(FILE* file) { return _fileno(file); }
-inline int Stat(const char* path, StatStruct* buf) { return _stat(path, buf); }
-inline int RmDir(const char* dir) { return _rmdir(dir); }
-inline bool IsDir(const StatStruct& st) {
-  return (_S_IFDIR & st.st_mode) != 0;
-}
-# endif  // GTEST_OS_WINDOWS_MOBILE
-
-#else
-
-typedef struct stat StatStruct;
-
-inline int FileNo(FILE* file) { return fileno(file); }
-inline int IsATTY(int fd) { return isatty(fd); }
-inline int Stat(const char* path, StatStruct* buf) { return stat(path, buf); }
-inline int StrCaseCmp(const char* s1, const char* s2) {
-  return strcasecmp(s1, s2);
-}
-inline char* StrDup(const char* src) { return strdup(src); }
-inline int RmDir(const char* dir) { return rmdir(dir); }
-inline bool IsDir(const StatStruct& st) { return S_ISDIR(st.st_mode); }
-
-#endif  // GTEST_OS_WINDOWS
-
-// Functions deprecated by MSVC 8.0.
-
-GTEST_DISABLE_MSC_WARNINGS_PUSH_(4996 /* deprecated function */)
-
-inline const char* StrNCpy(char* dest, const char* src, size_t n) {
-  return strncpy(dest, src, n);
-}
-
-// ChDir(), FReopen(), FDOpen(), Read(), Write(), Close(), and
-// StrError() aren't needed on Windows CE at this time and thus not
-// defined there.
-
-#if !GTEST_OS_WINDOWS_MOBILE && !GTEST_OS_WINDOWS_PHONE && !GTEST_OS_WINDOWS_RT
-inline int ChDir(const char* dir) { return chdir(dir); }
-#endif
-inline FILE* FOpen(const char* path, const char* mode) {
-  return fopen(path, mode);
-}
-#if !GTEST_OS_WINDOWS_MOBILE
-inline FILE *FReopen(const char* path, const char* mode, FILE* stream) {
-  return freopen(path, mode, stream);
-}
-inline FILE* FDOpen(int fd, const char* mode) { return fdopen(fd, mode); }
-#endif
-inline int FClose(FILE* fp) { return fclose(fp); }
-#if !GTEST_OS_WINDOWS_MOBILE
-inline int Read(int fd, void* buf, unsigned int count) {
-  return static_cast<int>(read(fd, buf, count));
-}
-inline int Write(int fd, const void* buf, unsigned int count) {
-  return static_cast<int>(write(fd, buf, count));
-}
-inline int Close(int fd) { return close(fd); }
-inline const char* StrError(int errnum) { return strerror(errnum); }
-#endif
-inline const char* GetEnv(const char* name) {
-#if GTEST_OS_WINDOWS_MOBILE || GTEST_OS_WINDOWS_PHONE || GTEST_OS_WINDOWS_RT
-  // We are on Windows CE, which has no environment variables.
-  static_cast<void>(name);  // To prevent 'unused argument' warning.
-  return NULL;
-#elif defined(__BORLANDC__) || defined(__SunOS_5_8) || defined(__SunOS_5_9)
-  // Environment variables which we programmatically clear will be set to the
-  // empty string rather than unset (NULL).  Handle that case.
-  const char* const env = getenv(name);
-  return (env != NULL && env[0] != '\0') ? env : NULL;
-#else
-  return getenv(name);
-#endif
-}
-
-GTEST_DISABLE_MSC_WARNINGS_POP_()
-
-#if GTEST_OS_WINDOWS_MOBILE
-// Windows CE has no C library. The abort() function is used in
-// several places in Google Test. This implementation provides a reasonable
-// imitation of standard behaviour.
-void Abort();
-#else
-inline void Abort() { abort(); }
-#endif  // GTEST_OS_WINDOWS_MOBILE
-
-}  // namespace posix
-
-// MSVC "deprecates" snprintf and issues warnings wherever it is used.  In
-// order to avoid these warnings, we need to use _snprintf or _snprintf_s on
-// MSVC-based platforms.  We map the GTEST_SNPRINTF_ macro to the appropriate
-// function in order to achieve that.  We use macro definition here because
-// snprintf is a variadic function.
-#if _MSC_VER >= 1400 && !GTEST_OS_WINDOWS_MOBILE
-// MSVC 2005 and above support variadic macros.
-# define GTEST_SNPRINTF_(buffer, size, format, ...) \
-     _snprintf_s(buffer, size, size, format, __VA_ARGS__)
-#elif defined(_MSC_VER)
-// Windows CE does not define _snprintf_s and MSVC prior to 2005 doesn't
-// complain about _snprintf.
-# define GTEST_SNPRINTF_ _snprintf
-#else
-# define GTEST_SNPRINTF_ snprintf
-#endif
-
-// The maximum number a BiggestInt can represent.  This definition
-// works no matter BiggestInt is represented in one's complement or
-// two's complement.
-//
-// We cannot rely on numeric_limits in STL, as __int64 and long long
-// are not part of standard C++ and numeric_limits doesn't need to be
-// defined for them.
-const BiggestInt kMaxBiggestInt =
-    ~(static_cast<BiggestInt>(1) << (8*sizeof(BiggestInt) - 1));
-
-// This template class serves as a compile-time function from size to
-// type.  It maps a size in bytes to a primitive type with that
-// size. e.g.
-//
-//   TypeWithSize<4>::UInt
-//
-// is typedef-ed to be unsigned int (unsigned integer made up of 4
-// bytes).
-//
-// Such functionality should belong to STL, but I cannot find it
-// there.
-//
-// Google Test uses this class in the implementation of floating-point
-// comparison.
-//
-// For now it only handles UInt (unsigned int) as that's all Google Test
-// needs.  Other types can be easily added in the future if need
-// arises.
-template <size_t size>
-class TypeWithSize {
- public:
-  // This prevents the user from using TypeWithSize<N> with incorrect
-  // values of N.
-  typedef void UInt;
-};
-
-// The specialization for size 4.
-template <>
-class TypeWithSize<4> {
- public:
-  // unsigned int has size 4 in both gcc and MSVC.
-  //
-  // As base/basictypes.h doesn't compile on Windows, we cannot use
-  // uint32, uint64, and etc here.
-  typedef int Int;
-  typedef unsigned int UInt;
-};
-
-// The specialization for size 8.
-template <>
-class TypeWithSize<8> {
- public:
-#if GTEST_OS_WINDOWS
-  typedef __int64 Int;
-  typedef unsigned __int64 UInt;
-#else
-  typedef long long Int;  // NOLINT
-  typedef unsigned long long UInt;  // NOLINT
-#endif  // GTEST_OS_WINDOWS
-};
-
-// Integer types of known sizes.
-typedef TypeWithSize<4>::Int Int32;
-typedef TypeWithSize<4>::UInt UInt32;
-typedef TypeWithSize<8>::Int Int64;
-typedef TypeWithSize<8>::UInt UInt64;
-typedef TypeWithSize<8>::Int TimeInMillis;  // Represents time in milliseconds.
-
-// Utilities for command line flags and environment variables.
-
-// Macro for referencing flags.
-#if !defined(GTEST_FLAG)
-# define GTEST_FLAG(name) FLAGS_gtest_##name
-#endif  // !defined(GTEST_FLAG)
-
-#if !defined(GTEST_USE_OWN_FLAGFILE_FLAG_)
-# define GTEST_USE_OWN_FLAGFILE_FLAG_ 1
-#endif  // !defined(GTEST_USE_OWN_FLAGFILE_FLAG_)
-
-#if !defined(GTEST_DECLARE_bool_)
-# define GTEST_FLAG_SAVER_ ::testing::internal::GTestFlagSaver
-
-// Macros for declaring flags.
-# define GTEST_DECLARE_bool_(name) GTEST_API_ extern bool GTEST_FLAG(name)
-# define GTEST_DECLARE_int32_(name) \
-    GTEST_API_ extern ::testing::internal::Int32 GTEST_FLAG(name)
-# define GTEST_DECLARE_string_(name) \
-    GTEST_API_ extern ::std::string GTEST_FLAG(name)
-
-// Macros for defining flags.
-# define GTEST_DEFINE_bool_(name, default_val, doc) \
-    GTEST_API_ bool GTEST_FLAG(name) = (default_val)
-# define GTEST_DEFINE_int32_(name, default_val, doc) \
-    GTEST_API_ ::testing::internal::Int32 GTEST_FLAG(name) = (default_val)
-# define GTEST_DEFINE_string_(name, default_val, doc) \
-    GTEST_API_ ::std::string GTEST_FLAG(name) = (default_val)
-
-#endif  // !defined(GTEST_DECLARE_bool_)
-
-// Thread annotations
-#if !defined(GTEST_EXCLUSIVE_LOCK_REQUIRED_)
-# define GTEST_EXCLUSIVE_LOCK_REQUIRED_(locks)
-# define GTEST_LOCK_EXCLUDED_(locks)
-#endif  // !defined(GTEST_EXCLUSIVE_LOCK_REQUIRED_)
-
-// Parses 'str' for a 32-bit signed integer.  If successful, writes the result
-// to *value and returns true; otherwise leaves *value unchanged and returns
-// false.
-// TODO(chandlerc): Find a better way to refactor flag and environment parsing
-// out of both gtest-port.cc and gtest.cc to avoid exporting this utility
-// function.
-bool ParseInt32(const Message& src_text, const char* str, Int32* value);
-
-// Parses a bool/Int32/string from the environment variable
-// corresponding to the given Google Test flag.
-bool BoolFromGTestEnv(const char* flag, bool default_val);
-GTEST_API_ Int32 Int32FromGTestEnv(const char* flag, Int32 default_val);
-std::string OutputFlagAlsoCheckEnvVar();
-const char* StringFromGTestEnv(const char* flag, const char* default_val);
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PORT_H_
-
-#if GTEST_OS_LINUX
-# include <stdlib.h>
-# include <sys/types.h>
-# include <sys/wait.h>
-# include <unistd.h>
-#endif  // GTEST_OS_LINUX
-
-#if GTEST_HAS_EXCEPTIONS
-# include <stdexcept>
-#endif
-
-#include <ctype.h>
-#include <float.h>
-#include <string.h>
-#include <iomanip>
-#include <limits>
-#include <map>
-#include <set>
-#include <string>
-#include <vector>
-
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-//
-// The Google C++ Testing and Mocking Framework (Google Test)
-//
-// This header file defines the Message class.
-//
-// IMPORTANT NOTE: Due to limitation of the C++ language, we have to
-// leave some internal implementation details in this header file.
-// They are clearly marked by comments like this:
-//
-//   // INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-//
-// Such code is NOT meant to be used by a user directly, and is subject
-// to CHANGE WITHOUT NOTICE.  Therefore DO NOT DEPEND ON IT in a user
-// program!
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_MESSAGE_H_
-#define GTEST_INCLUDE_GTEST_GTEST_MESSAGE_H_
-
-#include <limits>
-
-
-// Ensures that there is at least one operator<< in the global namespace.
-// See Message& operator<<(...) below for why.
-void operator<<(const testing::internal::Secret&, int);
-
-namespace testing {
-
-// The Message class works like an ostream repeater.
-//
-// Typical usage:
-//
-//   1. You stream a bunch of values to a Message object.
-//      It will remember the text in a stringstream.
-//   2. Then you stream the Message object to an ostream.
-//      This causes the text in the Message to be streamed
-//      to the ostream.
-//
-// For example;
-//
-//   testing::Message foo;
-//   foo << 1 << " != " << 2;
-//   std::cout << foo;
-//
-// will print "1 != 2".
-//
-// Message is not intended to be inherited from.  In particular, its
-// destructor is not virtual.
-//
-// Note that stringstream behaves differently in gcc and in MSVC.  You
-// can stream a NULL char pointer to it in the former, but not in the
-// latter (it causes an access violation if you do).  The Message
-// class hides this difference by treating a NULL char pointer as
-// "(null)".
-class GTEST_API_ Message {
- private:
-  // The type of basic IO manipulators (endl, ends, and flush) for
-  // narrow streams.
-  typedef std::ostream& (*BasicNarrowIoManip)(std::ostream&);
-
- public:
-  // Constructs an empty Message.
-  Message();
-
-  // Copy constructor.
-  Message(const Message& msg) : ss_(new ::std::stringstream) {  // NOLINT
-    *ss_ << msg.GetString();
-  }
-
-  // Constructs a Message from a C-string.
-  explicit Message(const char* str) : ss_(new ::std::stringstream) {
-    *ss_ << str;
-  }
-
-#if GTEST_OS_SYMBIAN
-  // Streams a value (either a pointer or not) to this object.
-  template <typename T>
-  inline Message& operator <<(const T& value) {
-    StreamHelper(typename internal::is_pointer<T>::type(), value);
-    return *this;
-  }
-#else
-  // Streams a non-pointer value to this object.
-  template <typename T>
-  inline Message& operator <<(const T& val) {
-    // Some libraries overload << for STL containers.  These
-    // overloads are defined in the global namespace instead of ::std.
-    //
-    // C++'s symbol lookup rule (i.e. Koenig lookup) says that these
-    // overloads are visible in either the std namespace or the global
-    // namespace, but not other namespaces, including the testing
-    // namespace which Google Test's Message class is in.
-    //
-    // To allow STL containers (and other types that has a << operator
-    // defined in the global namespace) to be used in Google Test
-    // assertions, testing::Message must access the custom << operator
-    // from the global namespace.  With this using declaration,
-    // overloads of << defined in the global namespace and those
-    // visible via Koenig lookup are both exposed in this function.
-    using ::operator <<;
-    *ss_ << val;
-    return *this;
-  }
-
-  // Streams a pointer value to this object.
-  //
-  // This function is an overload of the previous one.  When you
-  // stream a pointer to a Message, this definition will be used as it
-  // is more specialized.  (The C++ Standard, section
-  // [temp.func.order].)  If you stream a non-pointer, then the
-  // previous definition will be used.
-  //
-  // The reason for this overload is that streaming a NULL pointer to
-  // ostream is undefined behavior.  Depending on the compiler, you
-  // may get "0", "(nil)", "(null)", or an access violation.  To
-  // ensure consistent result across compilers, we always treat NULL
-  // as "(null)".
-  template <typename T>
-  inline Message& operator <<(T* const& pointer) {  // NOLINT
-    if (pointer == NULL) {
-      *ss_ << "(null)";
-    } else {
-      *ss_ << pointer;
-    }
-    return *this;
-  }
-#endif  // GTEST_OS_SYMBIAN
-
-  // Since the basic IO manipulators are overloaded for both narrow
-  // and wide streams, we have to provide this specialized definition
-  // of operator <<, even though its body is the same as the
-  // templatized version above.  Without this definition, streaming
-  // endl or other basic IO manipulators to Message will confuse the
-  // compiler.
-  Message& operator <<(BasicNarrowIoManip val) {
-    *ss_ << val;
-    return *this;
-  }
-
-  // Instead of 1/0, we want to see true/false for bool values.
-  Message& operator <<(bool b) {
-    return *this << (b ? "true" : "false");
-  }
-
-  // These two overloads allow streaming a wide C string to a Message
-  // using the UTF-8 encoding.
-  Message& operator <<(const wchar_t* wide_c_str);
-  Message& operator <<(wchar_t* wide_c_str);
-
-#if GTEST_HAS_STD_WSTRING
-  // Converts the given wide string to a narrow string using the UTF-8
-  // encoding, and streams the result to this Message object.
-  Message& operator <<(const ::std::wstring& wstr);
-#endif  // GTEST_HAS_STD_WSTRING
-
-#if GTEST_HAS_GLOBAL_WSTRING
-  // Converts the given wide string to a narrow string using the UTF-8
-  // encoding, and streams the result to this Message object.
-  Message& operator <<(const ::wstring& wstr);
-#endif  // GTEST_HAS_GLOBAL_WSTRING
-
-  // Gets the text streamed to this object so far as an std::string.
-  // Each '\0' character in the buffer is replaced with "\\0".
-  //
-  // INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-  std::string GetString() const;
-
- private:
-#if GTEST_OS_SYMBIAN
-  // These are needed as the Nokia Symbian Compiler cannot decide between
-  // const T& and const T* in a function template. The Nokia compiler _can_
-  // decide between class template specializations for T and T*, so a
-  // tr1::type_traits-like is_pointer works, and we can overload on that.
-  template <typename T>
-  inline void StreamHelper(internal::true_type /*is_pointer*/, T* pointer) {
-    if (pointer == NULL) {
-      *ss_ << "(null)";
-    } else {
-      *ss_ << pointer;
-    }
-  }
-  template <typename T>
-  inline void StreamHelper(internal::false_type /*is_pointer*/,
-                           const T& value) {
-    // See the comments in Message& operator <<(const T&) above for why
-    // we need this using statement.
-    using ::operator <<;
-    *ss_ << value;
-  }
-#endif  // GTEST_OS_SYMBIAN
-
-  // We'll hold the text streamed to this object here.
-  const internal::scoped_ptr< ::std::stringstream> ss_;
-
-  // We declare (but don't implement) this to prevent the compiler
-  // from implementing the assignment operator.
-  void operator=(const Message&);
-};
-
-// Streams a Message to an ostream.
-inline std::ostream& operator <<(std::ostream& os, const Message& sb) {
-  return os << sb.GetString();
-}
-
-namespace internal {
-
-// Converts a streamable value to an std::string.  A NULL pointer is
-// converted to "(null)".  When the input value is a ::string,
-// ::std::string, ::wstring, or ::std::wstring object, each NUL
-// character in it is replaced with "\\0".
-template <typename T>
-std::string StreamableToString(const T& streamable) {
-  return (Message() << streamable).GetString();
-}
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_MESSAGE_H_
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-//
-// Google Test filepath utilities
-//
-// This header file declares classes and functions used internally by
-// Google Test.  They are subject to change without notice.
-//
-// This file is #included in gtest/internal/gtest-internal.h.
-// Do not include this header file separately!
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_FILEPATH_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_FILEPATH_H_
-
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-//
-// The Google C++ Testing and Mocking Framework (Google Test)
-//
-// This header file declares the String class and functions used internally by
-// Google Test.  They are subject to change without notice. They should not used
-// by code external to Google Test.
-//
-// This header file is #included by
-// gtest/internal/gtest-internal.h.
-// It should not be #included by other files.
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_STRING_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_STRING_H_
-
-#ifdef __BORLANDC__
-// string.h is not guaranteed to provide strcpy on C++ Builder.
-# include <mem.h>
-#endif
-
-#include <string.h>
-#include <string>
-
-
-namespace testing {
-namespace internal {
-
-// String - an abstract class holding static string utilities.
-class GTEST_API_ String {
- public:
-  // Static utility methods
-
-  // Clones a 0-terminated C string, allocating memory using new.  The
-  // caller is responsible for deleting the return value using
-  // delete[].  Returns the cloned string, or NULL if the input is
-  // NULL.
-  //
-  // This is different from strdup() in string.h, which allocates
-  // memory using malloc().
-  static const char* CloneCString(const char* c_str);
-
-#if GTEST_OS_WINDOWS_MOBILE
-  // Windows CE does not have the 'ANSI' versions of Win32 APIs. To be
-  // able to pass strings to Win32 APIs on CE we need to convert them
-  // to 'Unicode', UTF-16.
-
-  // Creates a UTF-16 wide string from the given ANSI string, allocating
-  // memory using new. The caller is responsible for deleting the return
-  // value using delete[]. Returns the wide string, or NULL if the
-  // input is NULL.
-  //
-  // The wide string is created using the ANSI codepage (CP_ACP) to
-  // match the behaviour of the ANSI versions of Win32 calls and the
-  // C runtime.
-  static LPCWSTR AnsiToUtf16(const char* c_str);
-
-  // Creates an ANSI string from the given wide string, allocating
-  // memory using new. The caller is responsible for deleting the return
-  // value using delete[]. Returns the ANSI string, or NULL if the
-  // input is NULL.
-  //
-  // The returned string is created using the ANSI codepage (CP_ACP) to
-  // match the behaviour of the ANSI versions of Win32 calls and the
-  // C runtime.
-  static const char* Utf16ToAnsi(LPCWSTR utf16_str);
-#endif
-
-  // Compares two C strings.  Returns true iff they have the same content.
-  //
-  // Unlike strcmp(), this function can handle NULL argument(s).  A
-  // NULL C string is considered different to any non-NULL C string,
-  // including the empty string.
-  static bool CStringEquals(const char* lhs, const char* rhs);
-
-  // Converts a wide C string to a String using the UTF-8 encoding.
-  // NULL will be converted to "(null)".  If an error occurred during
-  // the conversion, "(failed to convert from wide string)" is
-  // returned.
-  static std::string ShowWideCString(const wchar_t* wide_c_str);
-
-  // Compares two wide C strings.  Returns true iff they have the same
-  // content.
-  //
-  // Unlike wcscmp(), this function can handle NULL argument(s).  A
-  // NULL C string is considered different to any non-NULL C string,
-  // including the empty string.
-  static bool WideCStringEquals(const wchar_t* lhs, const wchar_t* rhs);
-
-  // Compares two C strings, ignoring case.  Returns true iff they
-  // have the same content.
-  //
-  // Unlike strcasecmp(), this function can handle NULL argument(s).
-  // A NULL C string is considered different to any non-NULL C string,
-  // including the empty string.
-  static bool CaseInsensitiveCStringEquals(const char* lhs,
-                                           const char* rhs);
-
-  // Compares two wide C strings, ignoring case.  Returns true iff they
-  // have the same content.
-  //
-  // Unlike wcscasecmp(), this function can handle NULL argument(s).
-  // A NULL C string is considered different to any non-NULL wide C string,
-  // including the empty string.
-  // NB: The implementations on different platforms slightly differ.
-  // On windows, this method uses _wcsicmp which compares according to LC_CTYPE
-  // environment variable. On GNU platform this method uses wcscasecmp
-  // which compares according to LC_CTYPE category of the current locale.
-  // On MacOS X, it uses towlower, which also uses LC_CTYPE category of the
-  // current locale.
-  static bool CaseInsensitiveWideCStringEquals(const wchar_t* lhs,
-                                               const wchar_t* rhs);
-
-  // Returns true iff the given string ends with the given suffix, ignoring
-  // case. Any string is considered to end with an empty suffix.
-  static bool EndsWithCaseInsensitive(
-      const std::string& str, const std::string& suffix);
-
-  // Formats an int value as "%02d".
-  static std::string FormatIntWidth2(int value);  // "%02d" for width == 2
-
-  // Formats an int value as "%X".
-  static std::string FormatHexInt(int value);
-
-  // Formats a byte as "%02X".
-  static std::string FormatByte(unsigned char value);
-
- private:
-  String();  // Not meant to be instantiated.
-};  // class String
-
-// Gets the content of the stringstream's buffer as an std::string.  Each '\0'
-// character in the buffer is replaced with "\\0".
-GTEST_API_ std::string StringStreamToString(::std::stringstream* stream);
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_STRING_H_
-
-namespace testing {
-namespace internal {
-
-// FilePath - a class for file and directory pathname manipulation which
-// handles platform-specific conventions (like the pathname separator).
-// Used for helper functions for naming files in a directory for xml output.
-// Except for Set methods, all methods are const or static, which provides an
-// "immutable value object" -- useful for peace of mind.
-// A FilePath with a value ending in a path separator ("like/this/") represents
-// a directory, otherwise it is assumed to represent a file. In either case,
-// it may or may not represent an actual file or directory in the file system.
-// Names are NOT checked for syntax correctness -- no checking for illegal
-// characters, malformed paths, etc.
-
-class GTEST_API_ FilePath {
- public:
-  FilePath() : pathname_("") { }
-  FilePath(const FilePath& rhs) : pathname_(rhs.pathname_) { }
-
-  explicit FilePath(const std::string& pathname) : pathname_(pathname) {
-    Normalize();
-  }
-
-  FilePath& operator=(const FilePath& rhs) {
-    Set(rhs);
-    return *this;
-  }
-
-  void Set(const FilePath& rhs) {
-    pathname_ = rhs.pathname_;
-  }
-
-  const std::string& string() const { return pathname_; }
-  const char* c_str() const { return pathname_.c_str(); }
-
-  // Returns the current working directory, or "" if unsuccessful.
-  static FilePath GetCurrentDir();
-
-  // Given directory = "dir", base_name = "test", number = 0,
-  // extension = "xml", returns "dir/test.xml". If number is greater
-  // than zero (e.g., 12), returns "dir/test_12.xml".
-  // On Windows platform, uses \ as the separator rather than /.
-  static FilePath MakeFileName(const FilePath& directory,
-                               const FilePath& base_name,
-                               int number,
-                               const char* extension);
-
-  // Given directory = "dir", relative_path = "test.xml",
-  // returns "dir/test.xml".
-  // On Windows, uses \ as the separator rather than /.
-  static FilePath ConcatPaths(const FilePath& directory,
-                              const FilePath& relative_path);
-
-  // Returns a pathname for a file that does not currently exist. The pathname
-  // will be directory/base_name.extension or
-  // directory/base_name_<number>.extension if directory/base_name.extension
-  // already exists. The number will be incremented until a pathname is found
-  // that does not already exist.
-  // Examples: 'dir/foo_test.xml' or 'dir/foo_test_1.xml'.
-  // There could be a race condition if two or more processes are calling this
-  // function at the same time -- they could both pick the same filename.
-  static FilePath GenerateUniqueFileName(const FilePath& directory,
-                                         const FilePath& base_name,
-                                         const char* extension);
-
-  // Returns true iff the path is "".
-  bool IsEmpty() const { return pathname_.empty(); }
-
-  // If input name has a trailing separator character, removes it and returns
-  // the name, otherwise return the name string unmodified.
-  // On Windows platform, uses \ as the separator, other platforms use /.
-  FilePath RemoveTrailingPathSeparator() const;
-
-  // Returns a copy of the FilePath with the directory part removed.
-  // Example: FilePath("path/to/file").RemoveDirectoryName() returns
-  // FilePath("file"). If there is no directory part ("just_a_file"), it returns
-  // the FilePath unmodified. If there is no file part ("just_a_dir/") it
-  // returns an empty FilePath ("").
-  // On Windows platform, '\' is the path separator, otherwise it is '/'.
-  FilePath RemoveDirectoryName() const;
-
-  // RemoveFileName returns the directory path with the filename removed.
-  // Example: FilePath("path/to/file").RemoveFileName() returns "path/to/".
-  // If the FilePath is "a_file" or "/a_file", RemoveFileName returns
-  // FilePath("./") or, on Windows, FilePath(".\\"). If the filepath does
-  // not have a file, like "just/a/dir/", it returns the FilePath unmodified.
-  // On Windows platform, '\' is the path separator, otherwise it is '/'.
-  FilePath RemoveFileName() const;
-
-  // Returns a copy of the FilePath with the case-insensitive extension removed.
-  // Example: FilePath("dir/file.exe").RemoveExtension("EXE") returns
-  // FilePath("dir/file"). If a case-insensitive extension is not
-  // found, returns a copy of the original FilePath.
-  FilePath RemoveExtension(const char* extension) const;
-
-  // Creates directories so that path exists. Returns true if successful or if
-  // the directories already exist; returns false if unable to create
-  // directories for any reason. Will also return false if the FilePath does
-  // not represent a directory (that is, it doesn't end with a path separator).
-  bool CreateDirectoriesRecursively() const;
-
-  // Create the directory so that path exists. Returns true if successful or
-  // if the directory already exists; returns false if unable to create the
-  // directory for any reason, including if the parent directory does not
-  // exist. Not named "CreateDirectory" because that's a macro on Windows.
-  bool CreateFolder() const;
-
-  // Returns true if FilePath describes something in the file-system,
-  // either a file, directory, or whatever, and that something exists.
-  bool FileOrDirectoryExists() const;
-
-  // Returns true if pathname describes a directory in the file-system
-  // that exists.
-  bool DirectoryExists() const;
-
-  // Returns true if FilePath ends with a path separator, which indicates that
-  // it is intended to represent a directory. Returns false otherwise.
-  // This does NOT check that a directory (or file) actually exists.
-  bool IsDirectory() const;
-
-  // Returns true if pathname describes a root directory. (Windows has one
-  // root directory per disk drive.)
-  bool IsRootDirectory() const;
-
-  // Returns true if pathname describes an absolute path.
-  bool IsAbsolutePath() const;
-
- private:
-  // Replaces multiple consecutive separators with a single separator.
-  // For example, "bar///foo" becomes "bar/foo". Does not eliminate other
-  // redundancies that might be in a pathname involving "." or "..".
-  //
-  // A pathname with multiple consecutive separators may occur either through
-  // user error or as a result of some scripts or APIs that generate a pathname
-  // with a trailing separator. On other platforms the same API or script
-  // may NOT generate a pathname with a trailing "/". Then elsewhere that
-  // pathname may have another "/" and pathname components added to it,
-  // without checking for the separator already being there.
-  // The script language and operating system may allow paths like "foo//bar"
-  // but some of the functions in FilePath will not handle that correctly. In
-  // particular, RemoveTrailingPathSeparator() only removes one separator, and
-  // it is called in CreateDirectoriesRecursively() assuming that it will change
-  // a pathname from directory syntax (trailing separator) to filename syntax.
-  //
-  // On Windows this method also replaces the alternate path separator '/' with
-  // the primary path separator '\\', so that for example "bar\\/\\foo" becomes
-  // "bar\\foo".
-
-  void Normalize();
-
-  // Returns a pointer to the last occurence of a valid path separator in
-  // the FilePath. On Windows, for example, both '/' and '\' are valid path
-  // separators. Returns NULL if no path separator was found.
-  const char* FindLastPathSeparator() const;
-
-  std::string pathname_;
-};  // class FilePath
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_FILEPATH_H_
-// This file was GENERATED by command:
-//     pump.py gtest-type-util.h.pump
-// DO NOT EDIT BY HAND!!!
-
-// Copyright 2008 Google Inc.
-// All Rights Reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Type utilities needed for implementing typed and type-parameterized
-// tests.  This file is generated by a SCRIPT.  DO NOT EDIT BY HAND!
-//
-// Currently we support at most 50 types in a list, and at most 50
-// type-parameterized tests in one type-parameterized test case.
-// Please contact googletestframework@googlegroups.com if you need
-// more.
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_TYPE_UTIL_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_TYPE_UTIL_H_
-
-
-// #ifdef __GNUC__ is too general here.  It is possible to use gcc without using
-// libstdc++ (which is where cxxabi.h comes from).
-# if GTEST_HAS_CXXABI_H_
-#  include <cxxabi.h>
-# elif defined(__HP_aCC)
-#  include <acxx_demangle.h>
-# endif  // GTEST_HASH_CXXABI_H_
-
-namespace testing {
-namespace internal {
-
-// Canonicalizes a given name with respect to the Standard C++ Library.
-// This handles removing the inline namespace within `std` that is
-// used by various standard libraries (e.g., `std::__1`).  Names outside
-// of namespace std are returned unmodified.
-inline std::string CanonicalizeForStdLibVersioning(std::string s) {
-  static const char prefix[] = "std::__";
-  if (s.compare(0, strlen(prefix), prefix) == 0) {
-    std::string::size_type end = s.find("::", strlen(prefix));
-    if (end != s.npos) {
-      // Erase everything between the initial `std` and the second `::`.
-      s.erase(strlen("std"), end - strlen("std"));
-    }
-  }
-  return s;
-}
-
-// GetTypeName<T>() returns a human-readable name of type T.
-// NB: This function is also used in Google Mock, so don't move it inside of
-// the typed-test-only section below.
-template <typename T>
-std::string GetTypeName() {
-# if GTEST_HAS_RTTI
-
-  const char* const name = typeid(T).name();
-#  if GTEST_HAS_CXXABI_H_ || defined(__HP_aCC)
-  int status = 0;
-  // gcc's implementation of typeid(T).name() mangles the type name,
-  // so we have to demangle it.
-#   if GTEST_HAS_CXXABI_H_
-  using abi::__cxa_demangle;
-#   endif  // GTEST_HAS_CXXABI_H_
-  char* const readable_name = __cxa_demangle(name, 0, 0, &status);
-  const std::string name_str(status == 0 ? readable_name : name);
-  free(readable_name);
-  return CanonicalizeForStdLibVersioning(name_str);
-#  else
-  return name;
-#  endif  // GTEST_HAS_CXXABI_H_ || __HP_aCC
-
-# else
-
-  return "<type>";
-
-# endif  // GTEST_HAS_RTTI
-}
-
-#if GTEST_HAS_TYPED_TEST || GTEST_HAS_TYPED_TEST_P
-
-// AssertyTypeEq<T1, T2>::type is defined iff T1 and T2 are the same
-// type.  This can be used as a compile-time assertion to ensure that
-// two types are equal.
-
-template <typename T1, typename T2>
-struct AssertTypeEq;
-
-template <typename T>
-struct AssertTypeEq<T, T> {
-  typedef bool type;
-};
-
-// A unique type used as the default value for the arguments of class
-// template Types.  This allows us to simulate variadic templates
-// (e.g. Types<int>, Type<int, double>, and etc), which C++ doesn't
-// support directly.
-struct None {};
-
-// The following family of struct and struct templates are used to
-// represent type lists.  In particular, TypesN<T1, T2, ..., TN>
-// represents a type list with N types (T1, T2, ..., and TN) in it.
-// Except for Types0, every struct in the family has two member types:
-// Head for the first type in the list, and Tail for the rest of the
-// list.
-
-// The empty type list.
-struct Types0 {};
-
-// Type lists of length 1, 2, 3, and so on.
-
-template <typename T1>
-struct Types1 {
-  typedef T1 Head;
-  typedef Types0 Tail;
-};
-template <typename T1, typename T2>
-struct Types2 {
-  typedef T1 Head;
-  typedef Types1<T2> Tail;
-};
-
-template <typename T1, typename T2, typename T3>
-struct Types3 {
-  typedef T1 Head;
-  typedef Types2<T2, T3> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4>
-struct Types4 {
-  typedef T1 Head;
-  typedef Types3<T2, T3, T4> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5>
-struct Types5 {
-  typedef T1 Head;
-  typedef Types4<T2, T3, T4, T5> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6>
-struct Types6 {
-  typedef T1 Head;
-  typedef Types5<T2, T3, T4, T5, T6> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7>
-struct Types7 {
-  typedef T1 Head;
-  typedef Types6<T2, T3, T4, T5, T6, T7> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8>
-struct Types8 {
-  typedef T1 Head;
-  typedef Types7<T2, T3, T4, T5, T6, T7, T8> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9>
-struct Types9 {
-  typedef T1 Head;
-  typedef Types8<T2, T3, T4, T5, T6, T7, T8, T9> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10>
-struct Types10 {
-  typedef T1 Head;
-  typedef Types9<T2, T3, T4, T5, T6, T7, T8, T9, T10> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11>
-struct Types11 {
-  typedef T1 Head;
-  typedef Types10<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12>
-struct Types12 {
-  typedef T1 Head;
-  typedef Types11<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13>
-struct Types13 {
-  typedef T1 Head;
-  typedef Types12<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14>
-struct Types14 {
-  typedef T1 Head;
-  typedef Types13<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15>
-struct Types15 {
-  typedef T1 Head;
-  typedef Types14<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16>
-struct Types16 {
-  typedef T1 Head;
-  typedef Types15<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17>
-struct Types17 {
-  typedef T1 Head;
-  typedef Types16<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18>
-struct Types18 {
-  typedef T1 Head;
-  typedef Types17<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19>
-struct Types19 {
-  typedef T1 Head;
-  typedef Types18<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20>
-struct Types20 {
-  typedef T1 Head;
-  typedef Types19<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21>
-struct Types21 {
-  typedef T1 Head;
-  typedef Types20<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22>
-struct Types22 {
-  typedef T1 Head;
-  typedef Types21<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23>
-struct Types23 {
-  typedef T1 Head;
-  typedef Types22<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24>
-struct Types24 {
-  typedef T1 Head;
-  typedef Types23<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25>
-struct Types25 {
-  typedef T1 Head;
-  typedef Types24<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26>
-struct Types26 {
-  typedef T1 Head;
-  typedef Types25<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27>
-struct Types27 {
-  typedef T1 Head;
-  typedef Types26<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28>
-struct Types28 {
-  typedef T1 Head;
-  typedef Types27<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29>
-struct Types29 {
-  typedef T1 Head;
-  typedef Types28<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30>
-struct Types30 {
-  typedef T1 Head;
-  typedef Types29<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31>
-struct Types31 {
-  typedef T1 Head;
-  typedef Types30<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32>
-struct Types32 {
-  typedef T1 Head;
-  typedef Types31<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33>
-struct Types33 {
-  typedef T1 Head;
-  typedef Types32<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34>
-struct Types34 {
-  typedef T1 Head;
-  typedef Types33<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35>
-struct Types35 {
-  typedef T1 Head;
-  typedef Types34<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36>
-struct Types36 {
-  typedef T1 Head;
-  typedef Types35<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37>
-struct Types37 {
-  typedef T1 Head;
-  typedef Types36<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38>
-struct Types38 {
-  typedef T1 Head;
-  typedef Types37<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39>
-struct Types39 {
-  typedef T1 Head;
-  typedef Types38<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40>
-struct Types40 {
-  typedef T1 Head;
-  typedef Types39<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41>
-struct Types41 {
-  typedef T1 Head;
-  typedef Types40<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42>
-struct Types42 {
-  typedef T1 Head;
-  typedef Types41<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43>
-struct Types43 {
-  typedef T1 Head;
-  typedef Types42<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42,
-      T43> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44>
-struct Types44 {
-  typedef T1 Head;
-  typedef Types43<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-      T44> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45>
-struct Types45 {
-  typedef T1 Head;
-  typedef Types44<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-      T44, T45> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46>
-struct Types46 {
-  typedef T1 Head;
-  typedef Types45<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-      T44, T45, T46> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47>
-struct Types47 {
-  typedef T1 Head;
-  typedef Types46<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-      T44, T45, T46, T47> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48>
-struct Types48 {
-  typedef T1 Head;
-  typedef Types47<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-      T44, T45, T46, T47, T48> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48, typename T49>
-struct Types49 {
-  typedef T1 Head;
-  typedef Types48<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-      T44, T45, T46, T47, T48, T49> Tail;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48, typename T49, typename T50>
-struct Types50 {
-  typedef T1 Head;
-  typedef Types49<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-      T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-      T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-      T44, T45, T46, T47, T48, T49, T50> Tail;
-};
-
-
-}  // namespace internal
-
-// We don't want to require the users to write TypesN<...> directly,
-// as that would require them to count the length.  Types<...> is much
-// easier to write, but generates horrible messages when there is a
-// compiler error, as gcc insists on printing out each template
-// argument, even if it has the default value (this means Types<int>
-// will appear as Types<int, None, None, ..., None> in the compiler
-// errors).
-//
-// Our solution is to combine the best part of the two approaches: a
-// user would write Types<T1, ..., TN>, and Google Test will translate
-// that to TypesN<T1, ..., TN> internally to make error messages
-// readable.  The translation is done by the 'type' member of the
-// Types template.
-template <typename T1 = internal::None, typename T2 = internal::None,
-    typename T3 = internal::None, typename T4 = internal::None,
-    typename T5 = internal::None, typename T6 = internal::None,
-    typename T7 = internal::None, typename T8 = internal::None,
-    typename T9 = internal::None, typename T10 = internal::None,
-    typename T11 = internal::None, typename T12 = internal::None,
-    typename T13 = internal::None, typename T14 = internal::None,
-    typename T15 = internal::None, typename T16 = internal::None,
-    typename T17 = internal::None, typename T18 = internal::None,
-    typename T19 = internal::None, typename T20 = internal::None,
-    typename T21 = internal::None, typename T22 = internal::None,
-    typename T23 = internal::None, typename T24 = internal::None,
-    typename T25 = internal::None, typename T26 = internal::None,
-    typename T27 = internal::None, typename T28 = internal::None,
-    typename T29 = internal::None, typename T30 = internal::None,
-    typename T31 = internal::None, typename T32 = internal::None,
-    typename T33 = internal::None, typename T34 = internal::None,
-    typename T35 = internal::None, typename T36 = internal::None,
-    typename T37 = internal::None, typename T38 = internal::None,
-    typename T39 = internal::None, typename T40 = internal::None,
-    typename T41 = internal::None, typename T42 = internal::None,
-    typename T43 = internal::None, typename T44 = internal::None,
-    typename T45 = internal::None, typename T46 = internal::None,
-    typename T47 = internal::None, typename T48 = internal::None,
-    typename T49 = internal::None, typename T50 = internal::None>
-struct Types {
-  typedef internal::Types50<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41, T42, T43, T44, T45, T46, T47, T48, T49, T50> type;
-};
-
-template <>
-struct Types<internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types0 type;
-};
-template <typename T1>
-struct Types<T1, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types1<T1> type;
-};
-template <typename T1, typename T2>
-struct Types<T1, T2, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types2<T1, T2> type;
-};
-template <typename T1, typename T2, typename T3>
-struct Types<T1, T2, T3, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types3<T1, T2, T3> type;
-};
-template <typename T1, typename T2, typename T3, typename T4>
-struct Types<T1, T2, T3, T4, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types4<T1, T2, T3, T4> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5>
-struct Types<T1, T2, T3, T4, T5, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types5<T1, T2, T3, T4, T5> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6>
-struct Types<T1, T2, T3, T4, T5, T6, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types6<T1, T2, T3, T4, T5, T6> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7>
-struct Types<T1, T2, T3, T4, T5, T6, T7, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types7<T1, T2, T3, T4, T5, T6, T7> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types8<T1, T2, T3, T4, T5, T6, T7, T8> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types9<T1, T2, T3, T4, T5, T6, T7, T8, T9> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types10<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types11<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types12<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types13<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types14<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types15<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types16<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types17<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types18<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types19<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types20<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types21<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types22<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types23<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types24<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types25<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types26<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types27<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types28<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types29<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types30<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types31<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types32<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types33<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types34<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types35<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types36<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types37<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types38<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types39<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types40<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types41<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, internal::None,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types42<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41, T42> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None, internal::None> {
-  typedef internal::Types43<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41, T42, T43> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None, internal::None> {
-  typedef internal::Types44<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41, T42, T43, T44> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44, T45,
-    internal::None, internal::None, internal::None, internal::None,
-    internal::None> {
-  typedef internal::Types45<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41, T42, T43, T44, T45> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44, T45,
-    T46, internal::None, internal::None, internal::None, internal::None> {
-  typedef internal::Types46<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41, T42, T43, T44, T45, T46> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44, T45,
-    T46, T47, internal::None, internal::None, internal::None> {
-  typedef internal::Types47<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41, T42, T43, T44, T45, T46, T47> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44, T45,
-    T46, T47, T48, internal::None, internal::None> {
-  typedef internal::Types48<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41, T42, T43, T44, T45, T46, T47, T48> type;
-};
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48, typename T49>
-struct Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15,
-    T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29, T30,
-    T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44, T45,
-    T46, T47, T48, T49, internal::None> {
-  typedef internal::Types49<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41, T42, T43, T44, T45, T46, T47, T48, T49> type;
-};
-
-namespace internal {
-
-# define GTEST_TEMPLATE_ template <typename T> class
-
-// The template "selector" struct TemplateSel<Tmpl> is used to
-// represent Tmpl, which must be a class template with one type
-// parameter, as a type.  TemplateSel<Tmpl>::Bind<T>::type is defined
-// as the type Tmpl<T>.  This allows us to actually instantiate the
-// template "selected" by TemplateSel<Tmpl>.
-//
-// This trick is necessary for simulating typedef for class templates,
-// which C++ doesn't support directly.
-template <GTEST_TEMPLATE_ Tmpl>
-struct TemplateSel {
-  template <typename T>
-  struct Bind {
-    typedef Tmpl<T> type;
-  };
-};
-
-# define GTEST_BIND_(TmplSel, T) \
-  TmplSel::template Bind<T>::type
-
-// A unique struct template used as the default value for the
-// arguments of class template Templates.  This allows us to simulate
-// variadic templates (e.g. Templates<int>, Templates<int, double>,
-// and etc), which C++ doesn't support directly.
-template <typename T>
-struct NoneT {};
-
-// The following family of struct and struct templates are used to
-// represent template lists.  In particular, TemplatesN<T1, T2, ...,
-// TN> represents a list of N templates (T1, T2, ..., and TN).  Except
-// for Templates0, every struct in the family has two member types:
-// Head for the selector of the first template in the list, and Tail
-// for the rest of the list.
-
-// The empty template list.
-struct Templates0 {};
-
-// Template lists of length 1, 2, 3, and so on.
-
-template <GTEST_TEMPLATE_ T1>
-struct Templates1 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates0 Tail;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2>
-struct Templates2 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates1<T2> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3>
-struct Templates3 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates2<T2, T3> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4>
-struct Templates4 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates3<T2, T3, T4> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5>
-struct Templates5 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates4<T2, T3, T4, T5> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6>
-struct Templates6 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates5<T2, T3, T4, T5, T6> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7>
-struct Templates7 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates6<T2, T3, T4, T5, T6, T7> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8>
-struct Templates8 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates7<T2, T3, T4, T5, T6, T7, T8> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9>
-struct Templates9 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates8<T2, T3, T4, T5, T6, T7, T8, T9> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10>
-struct Templates10 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates9<T2, T3, T4, T5, T6, T7, T8, T9, T10> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11>
-struct Templates11 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates10<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12>
-struct Templates12 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates11<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13>
-struct Templates13 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates12<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14>
-struct Templates14 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates13<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15>
-struct Templates15 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates14<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16>
-struct Templates16 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates15<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17>
-struct Templates17 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates16<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18>
-struct Templates18 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates17<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19>
-struct Templates19 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates18<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20>
-struct Templates20 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates19<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21>
-struct Templates21 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates20<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22>
-struct Templates22 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates21<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23>
-struct Templates23 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates22<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24>
-struct Templates24 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates23<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25>
-struct Templates25 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates24<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26>
-struct Templates26 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates25<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27>
-struct Templates27 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates26<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28>
-struct Templates28 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates27<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29>
-struct Templates29 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates28<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30>
-struct Templates30 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates29<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31>
-struct Templates31 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates30<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32>
-struct Templates32 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates31<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33>
-struct Templates33 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates32<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34>
-struct Templates34 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates33<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35>
-struct Templates35 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates34<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36>
-struct Templates36 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates35<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37>
-struct Templates37 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates36<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38>
-struct Templates38 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates37<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39>
-struct Templates39 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates38<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40>
-struct Templates40 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates39<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41>
-struct Templates41 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates40<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42>
-struct Templates42 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates41<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-      T42> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43>
-struct Templates43 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates42<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42,
-      T43> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44>
-struct Templates44 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates43<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42,
-      T43, T44> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45>
-struct Templates45 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates44<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42,
-      T43, T44, T45> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45,
-    GTEST_TEMPLATE_ T46>
-struct Templates46 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates45<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42,
-      T43, T44, T45, T46> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45,
-    GTEST_TEMPLATE_ T46, GTEST_TEMPLATE_ T47>
-struct Templates47 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates46<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42,
-      T43, T44, T45, T46, T47> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45,
-    GTEST_TEMPLATE_ T46, GTEST_TEMPLATE_ T47, GTEST_TEMPLATE_ T48>
-struct Templates48 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates47<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42,
-      T43, T44, T45, T46, T47, T48> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45,
-    GTEST_TEMPLATE_ T46, GTEST_TEMPLATE_ T47, GTEST_TEMPLATE_ T48,
-    GTEST_TEMPLATE_ T49>
-struct Templates49 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates48<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42,
-      T43, T44, T45, T46, T47, T48, T49> Tail;
-};
-
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45,
-    GTEST_TEMPLATE_ T46, GTEST_TEMPLATE_ T47, GTEST_TEMPLATE_ T48,
-    GTEST_TEMPLATE_ T49, GTEST_TEMPLATE_ T50>
-struct Templates50 {
-  typedef TemplateSel<T1> Head;
-  typedef Templates49<T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-      T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-      T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42,
-      T43, T44, T45, T46, T47, T48, T49, T50> Tail;
-};
-
-
-// We don't want to require the users to write TemplatesN<...> directly,
-// as that would require them to count the length.  Templates<...> is much
-// easier to write, but generates horrible messages when there is a
-// compiler error, as gcc insists on printing out each template
-// argument, even if it has the default value (this means Templates<list>
-// will appear as Templates<list, NoneT, NoneT, ..., NoneT> in the compiler
-// errors).
-//
-// Our solution is to combine the best part of the two approaches: a
-// user would write Templates<T1, ..., TN>, and Google Test will translate
-// that to TemplatesN<T1, ..., TN> internally to make error messages
-// readable.  The translation is done by the 'type' member of the
-// Templates template.
-template <GTEST_TEMPLATE_ T1 = NoneT, GTEST_TEMPLATE_ T2 = NoneT,
-    GTEST_TEMPLATE_ T3 = NoneT, GTEST_TEMPLATE_ T4 = NoneT,
-    GTEST_TEMPLATE_ T5 = NoneT, GTEST_TEMPLATE_ T6 = NoneT,
-    GTEST_TEMPLATE_ T7 = NoneT, GTEST_TEMPLATE_ T8 = NoneT,
-    GTEST_TEMPLATE_ T9 = NoneT, GTEST_TEMPLATE_ T10 = NoneT,
-    GTEST_TEMPLATE_ T11 = NoneT, GTEST_TEMPLATE_ T12 = NoneT,
-    GTEST_TEMPLATE_ T13 = NoneT, GTEST_TEMPLATE_ T14 = NoneT,
-    GTEST_TEMPLATE_ T15 = NoneT, GTEST_TEMPLATE_ T16 = NoneT,
-    GTEST_TEMPLATE_ T17 = NoneT, GTEST_TEMPLATE_ T18 = NoneT,
-    GTEST_TEMPLATE_ T19 = NoneT, GTEST_TEMPLATE_ T20 = NoneT,
-    GTEST_TEMPLATE_ T21 = NoneT, GTEST_TEMPLATE_ T22 = NoneT,
-    GTEST_TEMPLATE_ T23 = NoneT, GTEST_TEMPLATE_ T24 = NoneT,
-    GTEST_TEMPLATE_ T25 = NoneT, GTEST_TEMPLATE_ T26 = NoneT,
-    GTEST_TEMPLATE_ T27 = NoneT, GTEST_TEMPLATE_ T28 = NoneT,
-    GTEST_TEMPLATE_ T29 = NoneT, GTEST_TEMPLATE_ T30 = NoneT,
-    GTEST_TEMPLATE_ T31 = NoneT, GTEST_TEMPLATE_ T32 = NoneT,
-    GTEST_TEMPLATE_ T33 = NoneT, GTEST_TEMPLATE_ T34 = NoneT,
-    GTEST_TEMPLATE_ T35 = NoneT, GTEST_TEMPLATE_ T36 = NoneT,
-    GTEST_TEMPLATE_ T37 = NoneT, GTEST_TEMPLATE_ T38 = NoneT,
-    GTEST_TEMPLATE_ T39 = NoneT, GTEST_TEMPLATE_ T40 = NoneT,
-    GTEST_TEMPLATE_ T41 = NoneT, GTEST_TEMPLATE_ T42 = NoneT,
-    GTEST_TEMPLATE_ T43 = NoneT, GTEST_TEMPLATE_ T44 = NoneT,
-    GTEST_TEMPLATE_ T45 = NoneT, GTEST_TEMPLATE_ T46 = NoneT,
-    GTEST_TEMPLATE_ T47 = NoneT, GTEST_TEMPLATE_ T48 = NoneT,
-    GTEST_TEMPLATE_ T49 = NoneT, GTEST_TEMPLATE_ T50 = NoneT>
-struct Templates {
-  typedef Templates50<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-      T42, T43, T44, T45, T46, T47, T48, T49, T50> type;
-};
-
-template <>
-struct Templates<NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT> {
-  typedef Templates0 type;
-};
-template <GTEST_TEMPLATE_ T1>
-struct Templates<T1, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT> {
-  typedef Templates1<T1> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2>
-struct Templates<T1, T2, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT> {
-  typedef Templates2<T1, T2> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3>
-struct Templates<T1, T2, T3, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates3<T1, T2, T3> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4>
-struct Templates<T1, T2, T3, T4, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates4<T1, T2, T3, T4> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5>
-struct Templates<T1, T2, T3, T4, T5, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates5<T1, T2, T3, T4, T5> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6>
-struct Templates<T1, T2, T3, T4, T5, T6, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates6<T1, T2, T3, T4, T5, T6> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates7<T1, T2, T3, T4, T5, T6, T7> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates8<T1, T2, T3, T4, T5, T6, T7, T8> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates9<T1, T2, T3, T4, T5, T6, T7, T8, T9> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates10<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates11<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates12<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates13<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates14<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates15<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates16<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates17<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates18<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates19<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates20<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates21<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT> {
-  typedef Templates22<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT> {
-  typedef Templates23<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT> {
-  typedef Templates24<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT> {
-  typedef Templates25<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT> {
-  typedef Templates26<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT> {
-  typedef Templates27<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT> {
-  typedef Templates28<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT> {
-  typedef Templates29<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates30<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates31<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates32<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates33<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates34<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates35<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates36<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, NoneT, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates37<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, NoneT, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates38<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates39<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, NoneT, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates40<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, NoneT, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates41<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, NoneT,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates42<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-      T42> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates43<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-      T42, T43> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44,
-    NoneT, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates44<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-      T42, T43, T44> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44,
-    T45, NoneT, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates45<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-      T42, T43, T44, T45> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45,
-    GTEST_TEMPLATE_ T46>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44,
-    T45, T46, NoneT, NoneT, NoneT, NoneT> {
-  typedef Templates46<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-      T42, T43, T44, T45, T46> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45,
-    GTEST_TEMPLATE_ T46, GTEST_TEMPLATE_ T47>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44,
-    T45, T46, T47, NoneT, NoneT, NoneT> {
-  typedef Templates47<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-      T42, T43, T44, T45, T46, T47> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45,
-    GTEST_TEMPLATE_ T46, GTEST_TEMPLATE_ T47, GTEST_TEMPLATE_ T48>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44,
-    T45, T46, T47, T48, NoneT, NoneT> {
-  typedef Templates48<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-      T42, T43, T44, T45, T46, T47, T48> type;
-};
-template <GTEST_TEMPLATE_ T1, GTEST_TEMPLATE_ T2, GTEST_TEMPLATE_ T3,
-    GTEST_TEMPLATE_ T4, GTEST_TEMPLATE_ T5, GTEST_TEMPLATE_ T6,
-    GTEST_TEMPLATE_ T7, GTEST_TEMPLATE_ T8, GTEST_TEMPLATE_ T9,
-    GTEST_TEMPLATE_ T10, GTEST_TEMPLATE_ T11, GTEST_TEMPLATE_ T12,
-    GTEST_TEMPLATE_ T13, GTEST_TEMPLATE_ T14, GTEST_TEMPLATE_ T15,
-    GTEST_TEMPLATE_ T16, GTEST_TEMPLATE_ T17, GTEST_TEMPLATE_ T18,
-    GTEST_TEMPLATE_ T19, GTEST_TEMPLATE_ T20, GTEST_TEMPLATE_ T21,
-    GTEST_TEMPLATE_ T22, GTEST_TEMPLATE_ T23, GTEST_TEMPLATE_ T24,
-    GTEST_TEMPLATE_ T25, GTEST_TEMPLATE_ T26, GTEST_TEMPLATE_ T27,
-    GTEST_TEMPLATE_ T28, GTEST_TEMPLATE_ T29, GTEST_TEMPLATE_ T30,
-    GTEST_TEMPLATE_ T31, GTEST_TEMPLATE_ T32, GTEST_TEMPLATE_ T33,
-    GTEST_TEMPLATE_ T34, GTEST_TEMPLATE_ T35, GTEST_TEMPLATE_ T36,
-    GTEST_TEMPLATE_ T37, GTEST_TEMPLATE_ T38, GTEST_TEMPLATE_ T39,
-    GTEST_TEMPLATE_ T40, GTEST_TEMPLATE_ T41, GTEST_TEMPLATE_ T42,
-    GTEST_TEMPLATE_ T43, GTEST_TEMPLATE_ T44, GTEST_TEMPLATE_ T45,
-    GTEST_TEMPLATE_ T46, GTEST_TEMPLATE_ T47, GTEST_TEMPLATE_ T48,
-    GTEST_TEMPLATE_ T49>
-struct Templates<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14,
-    T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28, T29,
-    T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43, T44,
-    T45, T46, T47, T48, T49, NoneT> {
-  typedef Templates49<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-      T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-      T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-      T42, T43, T44, T45, T46, T47, T48, T49> type;
-};
-
-// The TypeList template makes it possible to use either a single type
-// or a Types<...> list in TYPED_TEST_CASE() and
-// INSTANTIATE_TYPED_TEST_CASE_P().
-
-template <typename T>
-struct TypeList {
-  typedef Types1<T> type;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48, typename T49, typename T50>
-struct TypeList<Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-    T44, T45, T46, T47, T48, T49, T50> > {
-  typedef typename Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-      T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-      T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-      T41, T42, T43, T44, T45, T46, T47, T48, T49, T50>::type type;
-};
-
-#endif  // GTEST_HAS_TYPED_TEST || GTEST_HAS_TYPED_TEST_P
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_TYPE_UTIL_H_
-
-// Due to C++ preprocessor weirdness, we need double indirection to
-// concatenate two tokens when one of them is __LINE__.  Writing
-//
-//   foo ## __LINE__
-//
-// will result in the token foo__LINE__, instead of foo followed by
-// the current line number.  For more details, see
-// http://www.parashift.com/c++-faq-lite/misc-technical-issues.html#faq-39.6
-#define GTEST_CONCAT_TOKEN_(foo, bar) GTEST_CONCAT_TOKEN_IMPL_(foo, bar)
-#define GTEST_CONCAT_TOKEN_IMPL_(foo, bar) foo ## bar
-
-// Stringifies its argument.
-#define GTEST_STRINGIFY_(name) #name
-
-class ProtocolMessage;
-namespace proto2 { class Message; }
-
-namespace testing {
-
-// Forward declarations.
-
-class AssertionResult;                 // Result of an assertion.
-class Message;                         // Represents a failure message.
-class Test;                            // Represents a test.
-class TestInfo;                        // Information about a test.
-class TestPartResult;                  // Result of a test part.
-class UnitTest;                        // A collection of test cases.
-
-template <typename T>
-::std::string PrintToString(const T& value);
-
-namespace internal {
-
-struct TraceInfo;                      // Information about a trace point.
-class TestInfoImpl;                    // Opaque implementation of TestInfo
-class UnitTestImpl;                    // Opaque implementation of UnitTest
-
-// The text used in failure messages to indicate the start of the
-// stack trace.
-GTEST_API_ extern const char kStackTraceMarker[];
-
-// Two overloaded helpers for checking at compile time whether an
-// expression is a null pointer literal (i.e. NULL or any 0-valued
-// compile-time integral constant).  Their return values have
-// different sizes, so we can use sizeof() to test which version is
-// picked by the compiler.  These helpers have no implementations, as
-// we only need their signatures.
-//
-// Given IsNullLiteralHelper(x), the compiler will pick the first
-// version if x can be implicitly converted to Secret*, and pick the
-// second version otherwise.  Since Secret is a secret and incomplete
-// type, the only expression a user can write that has type Secret* is
-// a null pointer literal.  Therefore, we know that x is a null
-// pointer literal if and only if the first version is picked by the
-// compiler.
-char IsNullLiteralHelper(Secret* p);
-char (&IsNullLiteralHelper(...))[2];  // NOLINT
-
-// A compile-time bool constant that is true if and only if x is a
-// null pointer literal (i.e. NULL or any 0-valued compile-time
-// integral constant).
-#ifdef GTEST_ELLIPSIS_NEEDS_POD_
-// We lose support for NULL detection where the compiler doesn't like
-// passing non-POD classes through ellipsis (...).
-# define GTEST_IS_NULL_LITERAL_(x) false
-#else
-# define GTEST_IS_NULL_LITERAL_(x) \
-    (sizeof(::testing::internal::IsNullLiteralHelper(x)) == 1)
-#endif  // GTEST_ELLIPSIS_NEEDS_POD_
-
-// Appends the user-supplied message to the Google-Test-generated message.
-GTEST_API_ std::string AppendUserMessage(
-    const std::string& gtest_msg, const Message& user_msg);
-
-#if GTEST_HAS_EXCEPTIONS
-
-// This exception is thrown by (and only by) a failed Google Test
-// assertion when GTEST_FLAG(throw_on_failure) is true (if exceptions
-// are enabled).  We derive it from std::runtime_error, which is for
-// errors presumably detectable only at run time.  Since
-// std::runtime_error inherits from std::exception, many testing
-// frameworks know how to extract and print the message inside it.
-class GTEST_API_ GoogleTestFailureException : public ::std::runtime_error {
- public:
-  explicit GoogleTestFailureException(const TestPartResult& failure);
-};
-
-#endif  // GTEST_HAS_EXCEPTIONS
-
-namespace edit_distance {
-// Returns the optimal edits to go from 'left' to 'right'.
-// All edits cost the same, with replace having lower priority than
-// add/remove.
-// Simple implementation of the Wagner-Fischer algorithm.
-// See http://en.wikipedia.org/wiki/Wagner-Fischer_algorithm
-enum EditType { kMatch, kAdd, kRemove, kReplace };
-GTEST_API_ std::vector<EditType> CalculateOptimalEdits(
-    const std::vector<size_t>& left, const std::vector<size_t>& right);
-
-// Same as above, but the input is represented as strings.
-GTEST_API_ std::vector<EditType> CalculateOptimalEdits(
-    const std::vector<std::string>& left,
-    const std::vector<std::string>& right);
-
-// Create a diff of the input strings in Unified diff format.
-GTEST_API_ std::string CreateUnifiedDiff(const std::vector<std::string>& left,
-                                         const std::vector<std::string>& right,
-                                         size_t context = 2);
-
-}  // namespace edit_distance
-
-// Calculate the diff between 'left' and 'right' and return it in unified diff
-// format.
-// If not null, stores in 'total_line_count' the total number of lines found
-// in left + right.
-GTEST_API_ std::string DiffStrings(const std::string& left,
-                                   const std::string& right,
-                                   size_t* total_line_count);
-
-// Constructs and returns the message for an equality assertion
-// (e.g. ASSERT_EQ, EXPECT_STREQ, etc) failure.
-//
-// The first four parameters are the expressions used in the assertion
-// and their values, as strings.  For example, for ASSERT_EQ(foo, bar)
-// where foo is 5 and bar is 6, we have:
-//
-//   expected_expression: "foo"
-//   actual_expression:   "bar"
-//   expected_value:      "5"
-//   actual_value:        "6"
-//
-// The ignoring_case parameter is true iff the assertion is a
-// *_STRCASEEQ*.  When it's true, the string " (ignoring case)" will
-// be inserted into the message.
-GTEST_API_ AssertionResult EqFailure(const char* expected_expression,
-                                     const char* actual_expression,
-                                     const std::string& expected_value,
-                                     const std::string& actual_value,
-                                     bool ignoring_case);
-
-// Constructs a failure message for Boolean assertions such as EXPECT_TRUE.
-GTEST_API_ std::string GetBoolAssertionFailureMessage(
-    const AssertionResult& assertion_result,
-    const char* expression_text,
-    const char* actual_predicate_value,
-    const char* expected_predicate_value);
-
-// This template class represents an IEEE floating-point number
-// (either single-precision or double-precision, depending on the
-// template parameters).
-//
-// The purpose of this class is to do more sophisticated number
-// comparison.  (Due to round-off error, etc, it's very unlikely that
-// two floating-points will be equal exactly.  Hence a naive
-// comparison by the == operation often doesn't work.)
-//
-// Format of IEEE floating-point:
-//
-//   The most-significant bit being the leftmost, an IEEE
-//   floating-point looks like
-//
-//     sign_bit exponent_bits fraction_bits
-//
-//   Here, sign_bit is a single bit that designates the sign of the
-//   number.
-//
-//   For float, there are 8 exponent bits and 23 fraction bits.
-//
-//   For double, there are 11 exponent bits and 52 fraction bits.
-//
-//   More details can be found at
-//   http://en.wikipedia.org/wiki/IEEE_floating-point_standard.
-//
-// Template parameter:
-//
-//   RawType: the raw floating-point type (either float or double)
-template <typename RawType>
-class FloatingPoint {
- public:
-  // Defines the unsigned integer type that has the same size as the
-  // floating point number.
-  typedef typename TypeWithSize<sizeof(RawType)>::UInt Bits;
-
-  // Constants.
-
-  // # of bits in a number.
-  static const size_t kBitCount = 8*sizeof(RawType);
-
-  // # of fraction bits in a number.
-  static const size_t kFractionBitCount =
-    std::numeric_limits<RawType>::digits - 1;
-
-  // # of exponent bits in a number.
-  static const size_t kExponentBitCount = kBitCount - 1 - kFractionBitCount;
-
-  // The mask for the sign bit.
-  static const Bits kSignBitMask = static_cast<Bits>(1) << (kBitCount - 1);
-
-  // The mask for the fraction bits.
-  static const Bits kFractionBitMask =
-    ~static_cast<Bits>(0) >> (kExponentBitCount + 1);
-
-  // The mask for the exponent bits.
-  static const Bits kExponentBitMask = ~(kSignBitMask | kFractionBitMask);
-
-  // How many ULP's (Units in the Last Place) we want to tolerate when
-  // comparing two numbers.  The larger the value, the more error we
-  // allow.  A 0 value means that two numbers must be exactly the same
-  // to be considered equal.
-  //
-  // The maximum error of a single floating-point operation is 0.5
-  // units in the last place.  On Intel CPU's, all floating-point
-  // calculations are done with 80-bit precision, while double has 64
-  // bits.  Therefore, 4 should be enough for ordinary use.
-  //
-  // See the following article for more details on ULP:
-  // http://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
-  static const size_t kMaxUlps = 4;
-
-  // Constructs a FloatingPoint from a raw floating-point number.
-  //
-  // On an Intel CPU, passing a non-normalized NAN (Not a Number)
-  // around may change its bits, although the new value is guaranteed
-  // to be also a NAN.  Therefore, don't expect this constructor to
-  // preserve the bits in x when x is a NAN.
-  explicit FloatingPoint(const RawType& x) { u_.value_ = x; }
-
-  // Static methods
-
-  // Reinterprets a bit pattern as a floating-point number.
-  //
-  // This function is needed to test the AlmostEquals() method.
-  static RawType ReinterpretBits(const Bits bits) {
-    FloatingPoint fp(0);
-    fp.u_.bits_ = bits;
-    return fp.u_.value_;
-  }
-
-  // Returns the floating-point number that represent positive infinity.
-  static RawType Infinity() {
-    return ReinterpretBits(kExponentBitMask);
-  }
-
-  // Returns the maximum representable finite floating-point number.
-  static RawType Max();
-
-  // Non-static methods
-
-  // Returns the bits that represents this number.
-  const Bits &bits() const { return u_.bits_; }
-
-  // Returns the exponent bits of this number.
-  Bits exponent_bits() const { return kExponentBitMask & u_.bits_; }
-
-  // Returns the fraction bits of this number.
-  Bits fraction_bits() const { return kFractionBitMask & u_.bits_; }
-
-  // Returns the sign bit of this number.
-  Bits sign_bit() const { return kSignBitMask & u_.bits_; }
-
-  // Returns true iff this is NAN (not a number).
-  bool is_nan() const {
-    // It's a NAN if the exponent bits are all ones and the fraction
-    // bits are not entirely zeros.
-    return (exponent_bits() == kExponentBitMask) && (fraction_bits() != 0);
-  }
-
-  // Returns true iff this number is at most kMaxUlps ULP's away from
-  // rhs.  In particular, this function:
-  //
-  //   - returns false if either number is (or both are) NAN.
-  //   - treats really large numbers as almost equal to infinity.
-  //   - thinks +0.0 and -0.0 are 0 DLP's apart.
-  bool AlmostEquals(const FloatingPoint& rhs) const {
-    // The IEEE standard says that any comparison operation involving
-    // a NAN must return false.
-    if (is_nan() || rhs.is_nan()) return false;
-
-    return DistanceBetweenSignAndMagnitudeNumbers(u_.bits_, rhs.u_.bits_)
-        <= kMaxUlps;
-  }
-
- private:
-  // The data type used to store the actual floating-point number.
-  union FloatingPointUnion {
-    RawType value_;  // The raw floating-point number.
-    Bits bits_;      // The bits that represent the number.
-  };
-
-  // Converts an integer from the sign-and-magnitude representation to
-  // the biased representation.  More precisely, let N be 2 to the
-  // power of (kBitCount - 1), an integer x is represented by the
-  // unsigned number x + N.
-  //
-  // For instance,
-  //
-  //   -N + 1 (the most negative number representable using
-  //          sign-and-magnitude) is represented by 1;
-  //   0      is represented by N; and
-  //   N - 1  (the biggest number representable using
-  //          sign-and-magnitude) is represented by 2N - 1.
-  //
-  // Read http://en.wikipedia.org/wiki/Signed_number_representations
-  // for more details on signed number representations.
-  static Bits SignAndMagnitudeToBiased(const Bits &sam) {
-    if (kSignBitMask & sam) {
-      // sam represents a negative number.
-      return ~sam + 1;
-    } else {
-      // sam represents a positive number.
-      return kSignBitMask | sam;
-    }
-  }
-
-  // Given two numbers in the sign-and-magnitude representation,
-  // returns the distance between them as an unsigned number.
-  static Bits DistanceBetweenSignAndMagnitudeNumbers(const Bits &sam1,
-                                                     const Bits &sam2) {
-    const Bits biased1 = SignAndMagnitudeToBiased(sam1);
-    const Bits biased2 = SignAndMagnitudeToBiased(sam2);
-    return (biased1 >= biased2) ? (biased1 - biased2) : (biased2 - biased1);
-  }
-
-  FloatingPointUnion u_;
-};
-
-// We cannot use std::numeric_limits<T>::max() as it clashes with the max()
-// macro defined by <windows.h>.
-template <>
-inline float FloatingPoint<float>::Max() { return FLT_MAX; }
-template <>
-inline double FloatingPoint<double>::Max() { return DBL_MAX; }
-
-// Typedefs the instances of the FloatingPoint template class that we
-// care to use.
-typedef FloatingPoint<float> Float;
-typedef FloatingPoint<double> Double;
-
-// In order to catch the mistake of putting tests that use different
-// test fixture classes in the same test case, we need to assign
-// unique IDs to fixture classes and compare them.  The TypeId type is
-// used to hold such IDs.  The user should treat TypeId as an opaque
-// type: the only operation allowed on TypeId values is to compare
-// them for equality using the == operator.
-typedef const void* TypeId;
-
-template <typename T>
-class TypeIdHelper {
- public:
-  // dummy_ must not have a const type.  Otherwise an overly eager
-  // compiler (e.g. MSVC 7.1 & 8.0) may try to merge
-  // TypeIdHelper<T>::dummy_ for different Ts as an "optimization".
-  static bool dummy_;
-};
-
-template <typename T>
-bool TypeIdHelper<T>::dummy_ = false;
-
-// GetTypeId<T>() returns the ID of type T.  Different values will be
-// returned for different types.  Calling the function twice with the
-// same type argument is guaranteed to return the same ID.
-template <typename T>
-TypeId GetTypeId() {
-  // The compiler is required to allocate a different
-  // TypeIdHelper<T>::dummy_ variable for each T used to instantiate
-  // the template.  Therefore, the address of dummy_ is guaranteed to
-  // be unique.
-  return &(TypeIdHelper<T>::dummy_);
-}
-
-// Returns the type ID of ::testing::Test.  Always call this instead
-// of GetTypeId< ::testing::Test>() to get the type ID of
-// ::testing::Test, as the latter may give the wrong result due to a
-// suspected linker bug when compiling Google Test as a Mac OS X
-// framework.
-GTEST_API_ TypeId GetTestTypeId();
-
-// Defines the abstract factory interface that creates instances
-// of a Test object.
-class TestFactoryBase {
- public:
-  virtual ~TestFactoryBase() {}
-
-  // Creates a test instance to run. The instance is both created and destroyed
-  // within TestInfoImpl::Run()
-  virtual Test* CreateTest() = 0;
-
- protected:
-  TestFactoryBase() {}
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(TestFactoryBase);
-};
-
-// This class provides implementation of TeastFactoryBase interface.
-// It is used in TEST and TEST_F macros.
-template <class TestClass>
-class TestFactoryImpl : public TestFactoryBase {
- public:
-  virtual Test* CreateTest() { return new TestClass; }
-};
-
-#if GTEST_OS_WINDOWS
-
-// Predicate-formatters for implementing the HRESULT checking macros
-// {ASSERT|EXPECT}_HRESULT_{SUCCEEDED|FAILED}
-// We pass a long instead of HRESULT to avoid causing an
-// include dependency for the HRESULT type.
-GTEST_API_ AssertionResult IsHRESULTSuccess(const char* expr,
-                                            long hr);  // NOLINT
-GTEST_API_ AssertionResult IsHRESULTFailure(const char* expr,
-                                            long hr);  // NOLINT
-
-#endif  // GTEST_OS_WINDOWS
-
-// Types of SetUpTestCase() and TearDownTestCase() functions.
-typedef void (*SetUpTestCaseFunc)();
-typedef void (*TearDownTestCaseFunc)();
-
-struct CodeLocation {
-  CodeLocation(const std::string& a_file, int a_line)
-      : file(a_file), line(a_line) {}
-
-  std::string file;
-  int line;
-};
-
-// Creates a new TestInfo object and registers it with Google Test;
-// returns the created object.
-//
-// Arguments:
-//
-//   test_case_name:   name of the test case
-//   name:             name of the test
-//   type_param        the name of the test's type parameter, or NULL if
-//                     this is not a typed or a type-parameterized test.
-//   value_param       text representation of the test's value parameter,
-//                     or NULL if this is not a type-parameterized test.
-//   code_location:    code location where the test is defined
-//   fixture_class_id: ID of the test fixture class
-//   set_up_tc:        pointer to the function that sets up the test case
-//   tear_down_tc:     pointer to the function that tears down the test case
-//   factory:          pointer to the factory that creates a test object.
-//                     The newly created TestInfo instance will assume
-//                     ownership of the factory object.
-GTEST_API_ TestInfo* MakeAndRegisterTestInfo(
-    const char* test_case_name,
-    const char* name,
-    const char* type_param,
-    const char* value_param,
-    CodeLocation code_location,
-    TypeId fixture_class_id,
-    SetUpTestCaseFunc set_up_tc,
-    TearDownTestCaseFunc tear_down_tc,
-    TestFactoryBase* factory);
-
-// If *pstr starts with the given prefix, modifies *pstr to be right
-// past the prefix and returns true; otherwise leaves *pstr unchanged
-// and returns false.  None of pstr, *pstr, and prefix can be NULL.
-GTEST_API_ bool SkipPrefix(const char* prefix, const char** pstr);
-
-#if GTEST_HAS_TYPED_TEST || GTEST_HAS_TYPED_TEST_P
-
-// State of the definition of a type-parameterized test case.
-class GTEST_API_ TypedTestCasePState {
- public:
-  TypedTestCasePState() : registered_(false) {}
-
-  // Adds the given test name to defined_test_names_ and return true
-  // if the test case hasn't been registered; otherwise aborts the
-  // program.
-  bool AddTestName(const char* file, int line, const char* case_name,
-                   const char* test_name) {
-    if (registered_) {
-      fprintf(stderr, "%s Test %s must be defined before "
-              "REGISTER_TYPED_TEST_CASE_P(%s, ...).\n",
-              FormatFileLocation(file, line).c_str(), test_name, case_name);
-      fflush(stderr);
-      posix::Abort();
-    }
-    registered_tests_.insert(
-        ::std::make_pair(test_name, CodeLocation(file, line)));
-    return true;
-  }
-
-  bool TestExists(const std::string& test_name) const {
-    return registered_tests_.count(test_name) > 0;
-  }
-
-  const CodeLocation& GetCodeLocation(const std::string& test_name) const {
-    RegisteredTestsMap::const_iterator it = registered_tests_.find(test_name);
-    GTEST_CHECK_(it != registered_tests_.end());
-    return it->second;
-  }
-
-  // Verifies that registered_tests match the test names in
-  // defined_test_names_; returns registered_tests if successful, or
-  // aborts the program otherwise.
-  const char* VerifyRegisteredTestNames(
-      const char* file, int line, const char* registered_tests);
-
- private:
-  typedef ::std::map<std::string, CodeLocation> RegisteredTestsMap;
-
-  bool registered_;
-  RegisteredTestsMap registered_tests_;
-};
-
-// Skips to the first non-space char after the first comma in 'str';
-// returns NULL if no comma is found in 'str'.
-inline const char* SkipComma(const char* str) {
-  const char* comma = strchr(str, ',');
-  if (comma == NULL) {
-    return NULL;
-  }
-  while (IsSpace(*(++comma))) {}
-  return comma;
-}
-
-// Returns the prefix of 'str' before the first comma in it; returns
-// the entire string if it contains no comma.
-inline std::string GetPrefixUntilComma(const char* str) {
-  const char* comma = strchr(str, ',');
-  return comma == NULL ? str : std::string(str, comma);
-}
-
-// Splits a given string on a given delimiter, populating a given
-// vector with the fields.
-void SplitString(const ::std::string& str, char delimiter,
-                 ::std::vector< ::std::string>* dest);
-
-// TypeParameterizedTest<Fixture, TestSel, Types>::Register()
-// registers a list of type-parameterized tests with Google Test.  The
-// return value is insignificant - we just need to return something
-// such that we can call this function in a namespace scope.
-//
-// Implementation note: The GTEST_TEMPLATE_ macro declares a template
-// template parameter.  It's defined in gtest-type-util.h.
-template <GTEST_TEMPLATE_ Fixture, class TestSel, typename Types>
-class TypeParameterizedTest {
- public:
-  // 'index' is the index of the test in the type list 'Types'
-  // specified in INSTANTIATE_TYPED_TEST_CASE_P(Prefix, TestCase,
-  // Types).  Valid values for 'index' are [0, N - 1] where N is the
-  // length of Types.
-  static bool Register(const char* prefix,
-                       const CodeLocation& code_location,
-                       const char* case_name, const char* test_names,
-                       int index) {
-    typedef typename Types::Head Type;
-    typedef Fixture<Type> FixtureClass;
-    typedef typename GTEST_BIND_(TestSel, Type) TestClass;
-
-    // First, registers the first type-parameterized test in the type
-    // list.
-    MakeAndRegisterTestInfo(
-        (std::string(prefix) + (prefix[0] == '\0' ? "" : "/") + case_name + "/"
-         + StreamableToString(index)).c_str(),
-        StripTrailingSpaces(GetPrefixUntilComma(test_names)).c_str(),
-        GetTypeName<Type>().c_str(),
-        NULL,  // No value parameter.
-        code_location,
-        GetTypeId<FixtureClass>(),
-        TestClass::SetUpTestCase,
-        TestClass::TearDownTestCase,
-        new TestFactoryImpl<TestClass>);
-
-    // Next, recurses (at compile time) with the tail of the type list.
-    return TypeParameterizedTest<Fixture, TestSel, typename Types::Tail>
-        ::Register(prefix, code_location, case_name, test_names, index + 1);
-  }
-};
-
-// The base case for the compile time recursion.
-template <GTEST_TEMPLATE_ Fixture, class TestSel>
-class TypeParameterizedTest<Fixture, TestSel, Types0> {
- public:
-  static bool Register(const char* /*prefix*/, const CodeLocation&,
-                       const char* /*case_name*/, const char* /*test_names*/,
-                       int /*index*/) {
-    return true;
-  }
-};
-
-// TypeParameterizedTestCase<Fixture, Tests, Types>::Register()
-// registers *all combinations* of 'Tests' and 'Types' with Google
-// Test.  The return value is insignificant - we just need to return
-// something such that we can call this function in a namespace scope.
-template <GTEST_TEMPLATE_ Fixture, typename Tests, typename Types>
-class TypeParameterizedTestCase {
- public:
-  static bool Register(const char* prefix, CodeLocation code_location,
-                       const TypedTestCasePState* state,
-                       const char* case_name, const char* test_names) {
-    std::string test_name = StripTrailingSpaces(
-        GetPrefixUntilComma(test_names));
-    if (!state->TestExists(test_name)) {
-      fprintf(stderr, "Failed to get code location for test %s.%s at %s.",
-              case_name, test_name.c_str(),
-              FormatFileLocation(code_location.file.c_str(),
-                                 code_location.line).c_str());
-      fflush(stderr);
-      posix::Abort();
-    }
-    const CodeLocation& test_location = state->GetCodeLocation(test_name);
-
-    typedef typename Tests::Head Head;
-
-    // First, register the first test in 'Test' for each type in 'Types'.
-    TypeParameterizedTest<Fixture, Head, Types>::Register(
-        prefix, test_location, case_name, test_names, 0);
-
-    // Next, recurses (at compile time) with the tail of the test list.
-    return TypeParameterizedTestCase<Fixture, typename Tests::Tail, Types>
-        ::Register(prefix, code_location, state,
-                   case_name, SkipComma(test_names));
-  }
-};
-
-// The base case for the compile time recursion.
-template <GTEST_TEMPLATE_ Fixture, typename Types>
-class TypeParameterizedTestCase<Fixture, Templates0, Types> {
- public:
-  static bool Register(const char* /*prefix*/, const CodeLocation&,
-                       const TypedTestCasePState* /*state*/,
-                       const char* /*case_name*/, const char* /*test_names*/) {
-    return true;
-  }
-};
-
-#endif  // GTEST_HAS_TYPED_TEST || GTEST_HAS_TYPED_TEST_P
-
-// Returns the current OS stack trace as an std::string.
-//
-// The maximum number of stack frames to be included is specified by
-// the gtest_stack_trace_depth flag.  The skip_count parameter
-// specifies the number of top frames to be skipped, which doesn't
-// count against the number of frames to be included.
-//
-// For example, if Foo() calls Bar(), which in turn calls
-// GetCurrentOsStackTraceExceptTop(..., 1), Foo() will be included in
-// the trace but Bar() and GetCurrentOsStackTraceExceptTop() won't.
-GTEST_API_ std::string GetCurrentOsStackTraceExceptTop(
-    UnitTest* unit_test, int skip_count);
-
-// Helpers for suppressing warnings on unreachable code or constant
-// condition.
-
-// Always returns true.
-GTEST_API_ bool AlwaysTrue();
-
-// Always returns false.
-inline bool AlwaysFalse() { return !AlwaysTrue(); }
-
-// Helper for suppressing false warning from Clang on a const char*
-// variable declared in a conditional expression always being NULL in
-// the else branch.
-struct GTEST_API_ ConstCharPtr {
-  ConstCharPtr(const char* str) : value(str) {}
-  operator bool() const { return true; }
-  const char* value;
-};
-
-// A simple Linear Congruential Generator for generating random
-// numbers with a uniform distribution.  Unlike rand() and srand(), it
-// doesn't use global state (and therefore can't interfere with user
-// code).  Unlike rand_r(), it's portable.  An LCG isn't very random,
-// but it's good enough for our purposes.
-class GTEST_API_ Random {
- public:
-  static const UInt32 kMaxRange = 1u << 31;
-
-  explicit Random(UInt32 seed) : state_(seed) {}
-
-  void Reseed(UInt32 seed) { state_ = seed; }
-
-  // Generates a random number from [0, range).  Crashes if 'range' is
-  // 0 or greater than kMaxRange.
-  UInt32 Generate(UInt32 range);
-
- private:
-  UInt32 state_;
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(Random);
-};
-
-// Defining a variable of type CompileAssertTypesEqual<T1, T2> will cause a
-// compiler error iff T1 and T2 are different types.
-template <typename T1, typename T2>
-struct CompileAssertTypesEqual;
-
-template <typename T>
-struct CompileAssertTypesEqual<T, T> {
-};
-
-// Removes the reference from a type if it is a reference type,
-// otherwise leaves it unchanged.  This is the same as
-// tr1::remove_reference, which is not widely available yet.
-template <typename T>
-struct RemoveReference { typedef T type; };  // NOLINT
-template <typename T>
-struct RemoveReference<T&> { typedef T type; };  // NOLINT
-
-// A handy wrapper around RemoveReference that works when the argument
-// T depends on template parameters.
-#define GTEST_REMOVE_REFERENCE_(T) \
-    typename ::testing::internal::RemoveReference<T>::type
-
-// Removes const from a type if it is a const type, otherwise leaves
-// it unchanged.  This is the same as tr1::remove_const, which is not
-// widely available yet.
-template <typename T>
-struct RemoveConst { typedef T type; };  // NOLINT
-template <typename T>
-struct RemoveConst<const T> { typedef T type; };  // NOLINT
-
-// MSVC 8.0, Sun C++, and IBM XL C++ have a bug which causes the above
-// definition to fail to remove the const in 'const int[3]' and 'const
-// char[3][4]'.  The following specialization works around the bug.
-template <typename T, size_t N>
-struct RemoveConst<const T[N]> {
-  typedef typename RemoveConst<T>::type type[N];
-};
-
-#if defined(_MSC_VER) && _MSC_VER < 1400
-// This is the only specialization that allows VC++ 7.1 to remove const in
-// 'const int[3] and 'const int[3][4]'.  However, it causes trouble with GCC
-// and thus needs to be conditionally compiled.
-template <typename T, size_t N>
-struct RemoveConst<T[N]> {
-  typedef typename RemoveConst<T>::type type[N];
-};
-#endif
-
-// A handy wrapper around RemoveConst that works when the argument
-// T depends on template parameters.
-#define GTEST_REMOVE_CONST_(T) \
-    typename ::testing::internal::RemoveConst<T>::type
-
-// Turns const U&, U&, const U, and U all into U.
-#define GTEST_REMOVE_REFERENCE_AND_CONST_(T) \
-    GTEST_REMOVE_CONST_(GTEST_REMOVE_REFERENCE_(T))
-
-// ImplicitlyConvertible<From, To>::value is a compile-time bool
-// constant that's true iff type From can be implicitly converted to
-// type To.
-template <typename From, typename To>
-class ImplicitlyConvertible {
- private:
-  // We need the following helper functions only for their types.
-  // They have no implementations.
-
-  // MakeFrom() is an expression whose type is From.  We cannot simply
-  // use From(), as the type From may not have a public default
-  // constructor.
-  static typename AddReference<From>::type MakeFrom();
-
-  // These two functions are overloaded.  Given an expression
-  // Helper(x), the compiler will pick the first version if x can be
-  // implicitly converted to type To; otherwise it will pick the
-  // second version.
-  //
-  // The first version returns a value of size 1, and the second
-  // version returns a value of size 2.  Therefore, by checking the
-  // size of Helper(x), which can be done at compile time, we can tell
-  // which version of Helper() is used, and hence whether x can be
-  // implicitly converted to type To.
-  static char Helper(To);
-  static char (&Helper(...))[2];  // NOLINT
-
-  // We have to put the 'public' section after the 'private' section,
-  // or MSVC refuses to compile the code.
- public:
-#if defined(__BORLANDC__)
-  // C++Builder cannot use member overload resolution during template
-  // instantiation.  The simplest workaround is to use its C++0x type traits
-  // functions (C++Builder 2009 and above only).
-  static const bool value = __is_convertible(From, To);
-#else
-  // MSVC warns about implicitly converting from double to int for
-  // possible loss of data, so we need to temporarily disable the
-  // warning.
-  GTEST_DISABLE_MSC_WARNINGS_PUSH_(4244)
-  static const bool value =
-      sizeof(Helper(ImplicitlyConvertible::MakeFrom())) == 1;
-  GTEST_DISABLE_MSC_WARNINGS_POP_()
-#endif  // __BORLANDC__
-};
-template <typename From, typename To>
-const bool ImplicitlyConvertible<From, To>::value;
-
-// IsAProtocolMessage<T>::value is a compile-time bool constant that's
-// true iff T is type ProtocolMessage, proto2::Message, or a subclass
-// of those.
-template <typename T>
-struct IsAProtocolMessage
-    : public bool_constant<
-  ImplicitlyConvertible<const T*, const ::ProtocolMessage*>::value ||
-  ImplicitlyConvertible<const T*, const ::proto2::Message*>::value> {
-};
-
-// When the compiler sees expression IsContainerTest<C>(0), if C is an
-// STL-style container class, the first overload of IsContainerTest
-// will be viable (since both C::iterator* and C::const_iterator* are
-// valid types and NULL can be implicitly converted to them).  It will
-// be picked over the second overload as 'int' is a perfect match for
-// the type of argument 0.  If C::iterator or C::const_iterator is not
-// a valid type, the first overload is not viable, and the second
-// overload will be picked.  Therefore, we can determine whether C is
-// a container class by checking the type of IsContainerTest<C>(0).
-// The value of the expression is insignificant.
-//
-// In C++11 mode we check the existence of a const_iterator and that an
-// iterator is properly implemented for the container.
-//
-// For pre-C++11 that we look for both C::iterator and C::const_iterator.
-// The reason is that C++ injects the name of a class as a member of the
-// class itself (e.g. you can refer to class iterator as either
-// 'iterator' or 'iterator::iterator').  If we look for C::iterator
-// only, for example, we would mistakenly think that a class named
-// iterator is an STL container.
-//
-// Also note that the simpler approach of overloading
-// IsContainerTest(typename C::const_iterator*) and
-// IsContainerTest(...) doesn't work with Visual Age C++ and Sun C++.
-typedef int IsContainer;
-#if GTEST_LANG_CXX11
-template <class C,
-          class Iterator = decltype(::std::declval<const C&>().begin()),
-          class = decltype(::std::declval<const C&>().end()),
-          class = decltype(++::std::declval<Iterator&>()),
-          class = decltype(*::std::declval<Iterator>()),
-          class = typename C::const_iterator>
-IsContainer IsContainerTest(int /* dummy */) {
-  return 0;
-}
-#else
-template <class C>
-IsContainer IsContainerTest(int /* dummy */,
-                            typename C::iterator* /* it */ = NULL,
-                            typename C::const_iterator* /* const_it */ = NULL) {
-  return 0;
-}
-#endif  // GTEST_LANG_CXX11
-
-typedef char IsNotContainer;
-template <class C>
-IsNotContainer IsContainerTest(long /* dummy */) { return '\0'; }
-
-// Trait to detect whether a type T is a hash table.
-// The heuristic used is that the type contains an inner type `hasher` and does
-// not contain an inner type `reverse_iterator`.
-// If the container is iterable in reverse, then order might actually matter.
-template <typename T>
-struct IsHashTable {
- private:
-  template <typename U>
-  static char test(typename U::hasher*, typename U::reverse_iterator*);
-  template <typename U>
-  static int test(typename U::hasher*, ...);
-  template <typename U>
-  static char test(...);
-
- public:
-  static const bool value = sizeof(test<T>(0, 0)) == sizeof(int);
-};
-
-template <typename T>
-const bool IsHashTable<T>::value;
-
-template<typename T>
-struct VoidT {
-    typedef void value_type;
-};
-
-template <typename T, typename = void>
-struct HasValueType : false_type {};
-template <typename T>
-struct HasValueType<T, VoidT<typename T::value_type> > : true_type {
-};
-
-template <typename C,
-          bool = sizeof(IsContainerTest<C>(0)) == sizeof(IsContainer),
-          bool = HasValueType<C>::value>
-struct IsRecursiveContainerImpl;
-
-template <typename C, bool HV>
-struct IsRecursiveContainerImpl<C, false, HV> : public false_type {};
-
-// Since the IsRecursiveContainerImpl depends on the IsContainerTest we need to
-// obey the same inconsistencies as the IsContainerTest, namely check if
-// something is a container is relying on only const_iterator in C++11 and
-// is relying on both const_iterator and iterator otherwise
-template <typename C>
-struct IsRecursiveContainerImpl<C, true, false> : public false_type {};
-
-template <typename C>
-struct IsRecursiveContainerImpl<C, true, true> {
-  #if GTEST_LANG_CXX11
-  typedef typename IteratorTraits<typename C::const_iterator>::value_type
-      value_type;
-#else
-  typedef typename IteratorTraits<typename C::iterator>::value_type value_type;
-#endif
-  typedef is_same<value_type, C> type;
-};
-
-// IsRecursiveContainer<Type> is a unary compile-time predicate that
-// evaluates whether C is a recursive container type. A recursive container
-// type is a container type whose value_type is equal to the container type
-// itself. An example for a recursive container type is
-// boost::filesystem::path, whose iterator has a value_type that is equal to
-// boost::filesystem::path.
-template <typename C>
-struct IsRecursiveContainer : public IsRecursiveContainerImpl<C>::type {};
-
-// EnableIf<condition>::type is void when 'Cond' is true, and
-// undefined when 'Cond' is false.  To use SFINAE to make a function
-// overload only apply when a particular expression is true, add
-// "typename EnableIf<expression>::type* = 0" as the last parameter.
-template<bool> struct EnableIf;
-template<> struct EnableIf<true> { typedef void type; };  // NOLINT
-
-// Utilities for native arrays.
-
-// ArrayEq() compares two k-dimensional native arrays using the
-// elements' operator==, where k can be any integer >= 0.  When k is
-// 0, ArrayEq() degenerates into comparing a single pair of values.
-
-template <typename T, typename U>
-bool ArrayEq(const T* lhs, size_t size, const U* rhs);
-
-// This generic version is used when k is 0.
-template <typename T, typename U>
-inline bool ArrayEq(const T& lhs, const U& rhs) { return lhs == rhs; }
-
-// This overload is used when k >= 1.
-template <typename T, typename U, size_t N>
-inline bool ArrayEq(const T(&lhs)[N], const U(&rhs)[N]) {
-  return internal::ArrayEq(lhs, N, rhs);
-}
-
-// This helper reduces code bloat.  If we instead put its logic inside
-// the previous ArrayEq() function, arrays with different sizes would
-// lead to different copies of the template code.
-template <typename T, typename U>
-bool ArrayEq(const T* lhs, size_t size, const U* rhs) {
-  for (size_t i = 0; i != size; i++) {
-    if (!internal::ArrayEq(lhs[i], rhs[i]))
-      return false;
-  }
-  return true;
-}
-
-// Finds the first element in the iterator range [begin, end) that
-// equals elem.  Element may be a native array type itself.
-template <typename Iter, typename Element>
-Iter ArrayAwareFind(Iter begin, Iter end, const Element& elem) {
-  for (Iter it = begin; it != end; ++it) {
-    if (internal::ArrayEq(*it, elem))
-      return it;
-  }
-  return end;
-}
-
-// CopyArray() copies a k-dimensional native array using the elements'
-// operator=, where k can be any integer >= 0.  When k is 0,
-// CopyArray() degenerates into copying a single value.
-
-template <typename T, typename U>
-void CopyArray(const T* from, size_t size, U* to);
-
-// This generic version is used when k is 0.
-template <typename T, typename U>
-inline void CopyArray(const T& from, U* to) { *to = from; }
-
-// This overload is used when k >= 1.
-template <typename T, typename U, size_t N>
-inline void CopyArray(const T(&from)[N], U(*to)[N]) {
-  internal::CopyArray(from, N, *to);
-}
-
-// This helper reduces code bloat.  If we instead put its logic inside
-// the previous CopyArray() function, arrays with different sizes
-// would lead to different copies of the template code.
-template <typename T, typename U>
-void CopyArray(const T* from, size_t size, U* to) {
-  for (size_t i = 0; i != size; i++) {
-    internal::CopyArray(from[i], to + i);
-  }
-}
-
-// The relation between an NativeArray object (see below) and the
-// native array it represents.
-// We use 2 different structs to allow non-copyable types to be used, as long
-// as RelationToSourceReference() is passed.
-struct RelationToSourceReference {};
-struct RelationToSourceCopy {};
-
-// Adapts a native array to a read-only STL-style container.  Instead
-// of the complete STL container concept, this adaptor only implements
-// members useful for Google Mock's container matchers.  New members
-// should be added as needed.  To simplify the implementation, we only
-// support Element being a raw type (i.e. having no top-level const or
-// reference modifier).  It's the client's responsibility to satisfy
-// this requirement.  Element can be an array type itself (hence
-// multi-dimensional arrays are supported).
-template <typename Element>
-class NativeArray {
- public:
-  // STL-style container typedefs.
-  typedef Element value_type;
-  typedef Element* iterator;
-  typedef const Element* const_iterator;
-
-  // Constructs from a native array. References the source.
-  NativeArray(const Element* array, size_t count, RelationToSourceReference) {
-    InitRef(array, count);
-  }
-
-  // Constructs from a native array. Copies the source.
-  NativeArray(const Element* array, size_t count, RelationToSourceCopy) {
-    InitCopy(array, count);
-  }
-
-  // Copy constructor.
-  NativeArray(const NativeArray& rhs) {
-    (this->*rhs.clone_)(rhs.array_, rhs.size_);
-  }
-
-  ~NativeArray() {
-    if (clone_ != &NativeArray::InitRef)
-      delete[] array_;
-  }
-
-  // STL-style container methods.
-  size_t size() const { return size_; }
-  const_iterator begin() const { return array_; }
-  const_iterator end() const { return array_ + size_; }
-  bool operator==(const NativeArray& rhs) const {
-    return size() == rhs.size() &&
-        ArrayEq(begin(), size(), rhs.begin());
-  }
-
- private:
-  enum {
-    kCheckTypeIsNotConstOrAReference = StaticAssertTypeEqHelper<
-        Element, GTEST_REMOVE_REFERENCE_AND_CONST_(Element)>::value
-  };
-
-  // Initializes this object with a copy of the input.
-  void InitCopy(const Element* array, size_t a_size) {
-    Element* const copy = new Element[a_size];
-    CopyArray(array, a_size, copy);
-    array_ = copy;
-    size_ = a_size;
-    clone_ = &NativeArray::InitCopy;
-  }
-
-  // Initializes this object with a reference of the input.
-  void InitRef(const Element* array, size_t a_size) {
-    array_ = array;
-    size_ = a_size;
-    clone_ = &NativeArray::InitRef;
-  }
-
-  const Element* array_;
-  size_t size_;
-  void (NativeArray::*clone_)(const Element*, size_t);
-
-  GTEST_DISALLOW_ASSIGN_(NativeArray);
-};
-
-}  // namespace internal
-}  // namespace testing
-
-#define GTEST_MESSAGE_AT_(file, line, message, result_type) \
-  ::testing::internal::AssertHelper(result_type, file, line, message) \
-    = ::testing::Message()
-
-#define GTEST_MESSAGE_(message, result_type) \
-  GTEST_MESSAGE_AT_(__FILE__, __LINE__, message, result_type)
-
-#define GTEST_FATAL_FAILURE_(message) \
-  return GTEST_MESSAGE_(message, ::testing::TestPartResult::kFatalFailure)
-
-#define GTEST_NONFATAL_FAILURE_(message) \
-  GTEST_MESSAGE_(message, ::testing::TestPartResult::kNonFatalFailure)
-
-#define GTEST_SUCCESS_(message) \
-  GTEST_MESSAGE_(message, ::testing::TestPartResult::kSuccess)
-
-// Suppress MSVC warning 4702 (unreachable code) for the code following
-// statement if it returns or throws (or doesn't return or throw in some
-// situations).
-#define GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(statement) \
-  if (::testing::internal::AlwaysTrue()) { statement; }
-
-#define GTEST_TEST_THROW_(statement, expected_exception, fail) \
-  GTEST_AMBIGUOUS_ELSE_BLOCKER_ \
-  if (::testing::internal::ConstCharPtr gtest_msg = "") { \
-    bool gtest_caught_expected = false; \
-    try { \
-      GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(statement); \
-    } \
-    catch (expected_exception const&) { \
-      gtest_caught_expected = true; \
-    } \
-    catch (...) { \
-      gtest_msg.value = \
-          "Expected: " #statement " throws an exception of type " \
-          #expected_exception ".\n  Actual: it throws a different type."; \
-      goto GTEST_CONCAT_TOKEN_(gtest_label_testthrow_, __LINE__); \
-    } \
-    if (!gtest_caught_expected) { \
-      gtest_msg.value = \
-          "Expected: " #statement " throws an exception of type " \
-          #expected_exception ".\n  Actual: it throws nothing."; \
-      goto GTEST_CONCAT_TOKEN_(gtest_label_testthrow_, __LINE__); \
-    } \
-  } else \
-    GTEST_CONCAT_TOKEN_(gtest_label_testthrow_, __LINE__): \
-      fail(gtest_msg.value)
-
-#define GTEST_TEST_NO_THROW_(statement, fail) \
-  GTEST_AMBIGUOUS_ELSE_BLOCKER_ \
-  if (::testing::internal::AlwaysTrue()) { \
-    try { \
-      GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(statement); \
-    } \
-    catch (...) { \
-      goto GTEST_CONCAT_TOKEN_(gtest_label_testnothrow_, __LINE__); \
-    } \
-  } else \
-    GTEST_CONCAT_TOKEN_(gtest_label_testnothrow_, __LINE__): \
-      fail("Expected: " #statement " doesn't throw an exception.\n" \
-           "  Actual: it throws.")
-
-#define GTEST_TEST_ANY_THROW_(statement, fail) \
-  GTEST_AMBIGUOUS_ELSE_BLOCKER_ \
-  if (::testing::internal::AlwaysTrue()) { \
-    bool gtest_caught_any = false; \
-    try { \
-      GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(statement); \
-    } \
-    catch (...) { \
-      gtest_caught_any = true; \
-    } \
-    if (!gtest_caught_any) { \
-      goto GTEST_CONCAT_TOKEN_(gtest_label_testanythrow_, __LINE__); \
-    } \
-  } else \
-    GTEST_CONCAT_TOKEN_(gtest_label_testanythrow_, __LINE__): \
-      fail("Expected: " #statement " throws an exception.\n" \
-           "  Actual: it doesn't.")
-
-
-// Implements Boolean test assertions such as EXPECT_TRUE. expression can be
-// either a boolean expression or an AssertionResult. text is a textual
-// represenation of expression as it was passed into the EXPECT_TRUE.
-#define GTEST_TEST_BOOLEAN_(expression, text, actual, expected, fail) \
-  GTEST_AMBIGUOUS_ELSE_BLOCKER_ \
-  if (const ::testing::AssertionResult gtest_ar_ = \
-      ::testing::AssertionResult(expression)) \
-    ; \
-  else \
-    fail(::testing::internal::GetBoolAssertionFailureMessage(\
-        gtest_ar_, text, #actual, #expected).c_str())
-
-#define GTEST_TEST_NO_FATAL_FAILURE_(statement, fail) \
-  GTEST_AMBIGUOUS_ELSE_BLOCKER_ \
-  if (::testing::internal::AlwaysTrue()) { \
-    ::testing::internal::HasNewFatalFailureHelper gtest_fatal_failure_checker; \
-    GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(statement); \
-    if (gtest_fatal_failure_checker.has_new_fatal_failure()) { \
-      goto GTEST_CONCAT_TOKEN_(gtest_label_testnofatal_, __LINE__); \
-    } \
-  } else \
-    GTEST_CONCAT_TOKEN_(gtest_label_testnofatal_, __LINE__): \
-      fail("Expected: " #statement " doesn't generate new fatal " \
-           "failures in the current thread.\n" \
-           "  Actual: it does.")
-
-// Expands to the name of the class that implements the given test.
-#define GTEST_TEST_CLASS_NAME_(test_case_name, test_name) \
-  test_case_name##_##test_name##_Test
-
-// Helper macro for defining tests.
-#define GTEST_TEST_(test_case_name, test_name, parent_class, parent_id)\
-class GTEST_TEST_CLASS_NAME_(test_case_name, test_name) : public parent_class {\
- public:\
-  GTEST_TEST_CLASS_NAME_(test_case_name, test_name)() {}\
- private:\
-  virtual void TestBody();\
-  static ::testing::TestInfo* const test_info_ GTEST_ATTRIBUTE_UNUSED_;\
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(\
-      GTEST_TEST_CLASS_NAME_(test_case_name, test_name));\
-};\
-\
-::testing::TestInfo* const GTEST_TEST_CLASS_NAME_(test_case_name, test_name)\
-  ::test_info_ =\
-    ::testing::internal::MakeAndRegisterTestInfo(\
-        #test_case_name, #test_name, NULL, NULL, \
-        ::testing::internal::CodeLocation(__FILE__, __LINE__), \
-        (parent_id), \
-        parent_class::SetUpTestCase, \
-        parent_class::TearDownTestCase, \
-        new ::testing::internal::TestFactoryImpl<\
-            GTEST_TEST_CLASS_NAME_(test_case_name, test_name)>);\
-void GTEST_TEST_CLASS_NAME_(test_case_name, test_name)::TestBody()
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_INTERNAL_H_
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-//
-// The Google C++ Testing and Mocking Framework (Google Test)
-//
-// This header file defines the public API for death tests.  It is
-// #included by gtest.h so a user doesn't need to include this
-// directly.
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_DEATH_TEST_H_
-#define GTEST_INCLUDE_GTEST_GTEST_DEATH_TEST_H_
-
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-//
-// The Google C++ Testing and Mocking Framework (Google Test)
-//
-// This header file defines internal utilities needed for implementing
-// death tests.  They are subject to change without notice.
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_DEATH_TEST_INTERNAL_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_DEATH_TEST_INTERNAL_H_
-
-
-#include <stdio.h>
-
-namespace testing {
-namespace internal {
-
-GTEST_DECLARE_string_(internal_run_death_test);
-
-// Names of the flags (needed for parsing Google Test flags).
-const char kDeathTestStyleFlag[] = "death_test_style";
-const char kDeathTestUseFork[] = "death_test_use_fork";
-const char kInternalRunDeathTestFlag[] = "internal_run_death_test";
-
-#if GTEST_HAS_DEATH_TEST
-
-// DeathTest is a class that hides much of the complexity of the
-// GTEST_DEATH_TEST_ macro.  It is abstract; its static Create method
-// returns a concrete class that depends on the prevailing death test
-// style, as defined by the --gtest_death_test_style and/or
-// --gtest_internal_run_death_test flags.
-
-// In describing the results of death tests, these terms are used with
-// the corresponding definitions:
-//
-// exit status:  The integer exit information in the format specified
-//               by wait(2)
-// exit code:    The integer code passed to exit(3), _exit(2), or
-//               returned from main()
-class GTEST_API_ DeathTest {
- public:
-  // Create returns false if there was an error determining the
-  // appropriate action to take for the current death test; for example,
-  // if the gtest_death_test_style flag is set to an invalid value.
-  // The LastMessage method will return a more detailed message in that
-  // case.  Otherwise, the DeathTest pointer pointed to by the "test"
-  // argument is set.  If the death test should be skipped, the pointer
-  // is set to NULL; otherwise, it is set to the address of a new concrete
-  // DeathTest object that controls the execution of the current test.
-  static bool Create(const char* statement, const RE* regex,
-                     const char* file, int line, DeathTest** test);
-  DeathTest();
-  virtual ~DeathTest() { }
-
-  // A helper class that aborts a death test when it's deleted.
-  class ReturnSentinel {
-   public:
-    explicit ReturnSentinel(DeathTest* test) : test_(test) { }
-    ~ReturnSentinel() { test_->Abort(TEST_ENCOUNTERED_RETURN_STATEMENT); }
-   private:
-    DeathTest* const test_;
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(ReturnSentinel);
-  } GTEST_ATTRIBUTE_UNUSED_;
-
-  // An enumeration of possible roles that may be taken when a death
-  // test is encountered.  EXECUTE means that the death test logic should
-  // be executed immediately.  OVERSEE means that the program should prepare
-  // the appropriate environment for a child process to execute the death
-  // test, then wait for it to complete.
-  enum TestRole { OVERSEE_TEST, EXECUTE_TEST };
-
-  // An enumeration of the three reasons that a test might be aborted.
-  enum AbortReason {
-    TEST_ENCOUNTERED_RETURN_STATEMENT,
-    TEST_THREW_EXCEPTION,
-    TEST_DID_NOT_DIE
-  };
-
-  // Assumes one of the above roles.
-  virtual TestRole AssumeRole() = 0;
-
-  // Waits for the death test to finish and returns its status.
-  virtual int Wait() = 0;
-
-  // Returns true if the death test passed; that is, the test process
-  // exited during the test, its exit status matches a user-supplied
-  // predicate, and its stderr output matches a user-supplied regular
-  // expression.
-  // The user-supplied predicate may be a macro expression rather
-  // than a function pointer or functor, or else Wait and Passed could
-  // be combined.
-  virtual bool Passed(bool exit_status_ok) = 0;
-
-  // Signals that the death test did not die as expected.
-  virtual void Abort(AbortReason reason) = 0;
-
-  // Returns a human-readable outcome message regarding the outcome of
-  // the last death test.
-  static const char* LastMessage();
-
-  static void set_last_death_test_message(const std::string& message);
-
- private:
-  // A string containing a description of the outcome of the last death test.
-  static std::string last_death_test_message_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(DeathTest);
-};
-
-// Factory interface for death tests.  May be mocked out for testing.
-class DeathTestFactory {
- public:
-  virtual ~DeathTestFactory() { }
-  virtual bool Create(const char* statement, const RE* regex,
-                      const char* file, int line, DeathTest** test) = 0;
-};
-
-// A concrete DeathTestFactory implementation for normal use.
-class DefaultDeathTestFactory : public DeathTestFactory {
- public:
-  virtual bool Create(const char* statement, const RE* regex,
-                      const char* file, int line, DeathTest** test);
-};
-
-// Returns true if exit_status describes a process that was terminated
-// by a signal, or exited normally with a nonzero exit code.
-GTEST_API_ bool ExitedUnsuccessfully(int exit_status);
-
-// Traps C++ exceptions escaping statement and reports them as test
-// failures. Note that trapping SEH exceptions is not implemented here.
-# if GTEST_HAS_EXCEPTIONS
-#  define GTEST_EXECUTE_DEATH_TEST_STATEMENT_(statement, death_test) \
-  try { \
-    GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(statement); \
-  } catch (const ::std::exception& gtest_exception) { \
-    fprintf(\
-        stderr, \
-        "\n%s: Caught std::exception-derived exception escaping the " \
-        "death test statement. Exception message: %s\n", \
-        ::testing::internal::FormatFileLocation(__FILE__, __LINE__).c_str(), \
-        gtest_exception.what()); \
-    fflush(stderr); \
-    death_test->Abort(::testing::internal::DeathTest::TEST_THREW_EXCEPTION); \
-  } catch (...) { \
-    death_test->Abort(::testing::internal::DeathTest::TEST_THREW_EXCEPTION); \
-  }
-
-# else
-#  define GTEST_EXECUTE_DEATH_TEST_STATEMENT_(statement, death_test) \
-  GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(statement)
-
-# endif
-
-// This macro is for implementing ASSERT_DEATH*, EXPECT_DEATH*,
-// ASSERT_EXIT*, and EXPECT_EXIT*.
-# define GTEST_DEATH_TEST_(statement, predicate, regex, fail) \
-  GTEST_AMBIGUOUS_ELSE_BLOCKER_ \
-  if (::testing::internal::AlwaysTrue()) { \
-    const ::testing::internal::RE& gtest_regex = (regex); \
-    ::testing::internal::DeathTest* gtest_dt; \
-    if (!::testing::internal::DeathTest::Create(#statement, &gtest_regex, \
-        __FILE__, __LINE__, &gtest_dt)) { \
-      goto GTEST_CONCAT_TOKEN_(gtest_label_, __LINE__); \
-    } \
-    if (gtest_dt != NULL) { \
-      ::testing::internal::scoped_ptr< ::testing::internal::DeathTest> \
-          gtest_dt_ptr(gtest_dt); \
-      switch (gtest_dt->AssumeRole()) { \
-        case ::testing::internal::DeathTest::OVERSEE_TEST: \
-          if (!gtest_dt->Passed(predicate(gtest_dt->Wait()))) { \
-            goto GTEST_CONCAT_TOKEN_(gtest_label_, __LINE__); \
-          } \
-          break; \
-        case ::testing::internal::DeathTest::EXECUTE_TEST: { \
-          ::testing::internal::DeathTest::ReturnSentinel \
-              gtest_sentinel(gtest_dt); \
-          GTEST_EXECUTE_DEATH_TEST_STATEMENT_(statement, gtest_dt); \
-          gtest_dt->Abort(::testing::internal::DeathTest::TEST_DID_NOT_DIE); \
-          break; \
-        } \
-        default: \
-          break; \
-      } \
-    } \
-  } else \
-    GTEST_CONCAT_TOKEN_(gtest_label_, __LINE__): \
-      fail(::testing::internal::DeathTest::LastMessage())
-// The symbol "fail" here expands to something into which a message
-// can be streamed.
-
-// This macro is for implementing ASSERT/EXPECT_DEBUG_DEATH when compiled in
-// NDEBUG mode. In this case we need the statements to be executed and the macro
-// must accept a streamed message even though the message is never printed.
-// The regex object is not evaluated, but it is used to prevent "unused"
-// warnings and to avoid an expression that doesn't compile in debug mode.
-#define GTEST_EXECUTE_STATEMENT_(statement, regex)             \
-  GTEST_AMBIGUOUS_ELSE_BLOCKER_                                \
-  if (::testing::internal::AlwaysTrue()) {                     \
-    GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(statement); \
-  } else if (!::testing::internal::AlwaysTrue()) {             \
-    const ::testing::internal::RE& gtest_regex = (regex);      \
-    static_cast<void>(gtest_regex);                            \
-  } else                                                       \
-    ::testing::Message()
-
-// A class representing the parsed contents of the
-// --gtest_internal_run_death_test flag, as it existed when
-// RUN_ALL_TESTS was called.
-class InternalRunDeathTestFlag {
- public:
-  InternalRunDeathTestFlag(const std::string& a_file,
-                           int a_line,
-                           int an_index,
-                           int a_write_fd)
-      : file_(a_file), line_(a_line), index_(an_index),
-        write_fd_(a_write_fd) {}
-
-  ~InternalRunDeathTestFlag() {
-    if (write_fd_ >= 0)
-      posix::Close(write_fd_);
-  }
-
-  const std::string& file() const { return file_; }
-  int line() const { return line_; }
-  int index() const { return index_; }
-  int write_fd() const { return write_fd_; }
-
- private:
-  std::string file_;
-  int line_;
-  int index_;
-  int write_fd_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(InternalRunDeathTestFlag);
-};
-
-// Returns a newly created InternalRunDeathTestFlag object with fields
-// initialized from the GTEST_FLAG(internal_run_death_test) flag if
-// the flag is specified; otherwise returns NULL.
-InternalRunDeathTestFlag* ParseInternalRunDeathTestFlag();
-
-#endif  // GTEST_HAS_DEATH_TEST
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_DEATH_TEST_INTERNAL_H_
-
-namespace testing {
-
-// This flag controls the style of death tests.  Valid values are "threadsafe",
-// meaning that the death test child process will re-execute the test binary
-// from the start, running only a single death test, or "fast",
-// meaning that the child process will execute the test logic immediately
-// after forking.
-GTEST_DECLARE_string_(death_test_style);
-
-#if GTEST_HAS_DEATH_TEST
-
-namespace internal {
-
-// Returns a Boolean value indicating whether the caller is currently
-// executing in the context of the death test child process.  Tools such as
-// Valgrind heap checkers may need this to modify their behavior in death
-// tests.  IMPORTANT: This is an internal utility.  Using it may break the
-// implementation of death tests.  User code MUST NOT use it.
-GTEST_API_ bool InDeathTestChild();
-
-}  // namespace internal
-
-// The following macros are useful for writing death tests.
-
-// Here's what happens when an ASSERT_DEATH* or EXPECT_DEATH* is
-// executed:
-//
-//   1. It generates a warning if there is more than one active
-//   thread.  This is because it's safe to fork() or clone() only
-//   when there is a single thread.
-//
-//   2. The parent process clone()s a sub-process and runs the death
-//   test in it; the sub-process exits with code 0 at the end of the
-//   death test, if it hasn't exited already.
-//
-//   3. The parent process waits for the sub-process to terminate.
-//
-//   4. The parent process checks the exit code and error message of
-//   the sub-process.
-//
-// Examples:
-//
-//   ASSERT_DEATH(server.SendMessage(56, "Hello"), "Invalid port number");
-//   for (int i = 0; i < 5; i++) {
-//     EXPECT_DEATH(server.ProcessRequest(i),
-//                  "Invalid request .* in ProcessRequest()")
-//                  << "Failed to die on request " << i;
-//   }
-//
-//   ASSERT_EXIT(server.ExitNow(), ::testing::ExitedWithCode(0), "Exiting");
-//
-//   bool KilledBySIGHUP(int exit_code) {
-//     return WIFSIGNALED(exit_code) && WTERMSIG(exit_code) == SIGHUP;
-//   }
-//
-//   ASSERT_EXIT(client.HangUpServer(), KilledBySIGHUP, "Hanging up!");
-//
-// On the regular expressions used in death tests:
-//
-//   On POSIX-compliant systems (*nix), we use the <regex.h> library,
-//   which uses the POSIX extended regex syntax.
-//
-//   On other platforms (e.g. Windows or Mac), we only support a simple regex
-//   syntax implemented as part of Google Test.  This limited
-//   implementation should be enough most of the time when writing
-//   death tests; though it lacks many features you can find in PCRE
-//   or POSIX extended regex syntax.  For example, we don't support
-//   union ("x|y"), grouping ("(xy)"), brackets ("[xy]"), and
-//   repetition count ("x{5,7}"), among others.
-//
-//   Below is the syntax that we do support.  We chose it to be a
-//   subset of both PCRE and POSIX extended regex, so it's easy to
-//   learn wherever you come from.  In the following: 'A' denotes a
-//   literal character, period (.), or a single \\ escape sequence;
-//   'x' and 'y' denote regular expressions; 'm' and 'n' are for
-//   natural numbers.
-//
-//     c     matches any literal character c
-//     \\d   matches any decimal digit
-//     \\D   matches any character that's not a decimal digit
-//     \\f   matches \f
-//     \\n   matches \n
-//     \\r   matches \r
-//     \\s   matches any ASCII whitespace, including \n
-//     \\S   matches any character that's not a whitespace
-//     \\t   matches \t
-//     \\v   matches \v
-//     \\w   matches any letter, _, or decimal digit
-//     \\W   matches any character that \\w doesn't match
-//     \\c   matches any literal character c, which must be a punctuation
-//     .     matches any single character except \n
-//     A?    matches 0 or 1 occurrences of A
-//     A*    matches 0 or many occurrences of A
-//     A+    matches 1 or many occurrences of A
-//     ^     matches the beginning of a string (not that of each line)
-//     $     matches the end of a string (not that of each line)
-//     xy    matches x followed by y
-//
-//   If you accidentally use PCRE or POSIX extended regex features
-//   not implemented by us, you will get a run-time failure.  In that
-//   case, please try to rewrite your regular expression within the
-//   above syntax.
-//
-//   This implementation is *not* meant to be as highly tuned or robust
-//   as a compiled regex library, but should perform well enough for a
-//   death test, which already incurs significant overhead by launching
-//   a child process.
-//
-// Known caveats:
-//
-//   A "threadsafe" style death test obtains the path to the test
-//   program from argv[0] and re-executes it in the sub-process.  For
-//   simplicity, the current implementation doesn't search the PATH
-//   when launching the sub-process.  This means that the user must
-//   invoke the test program via a path that contains at least one
-//   path separator (e.g. path/to/foo_test and
-//   /absolute/path/to/bar_test are fine, but foo_test is not).  This
-//   is rarely a problem as people usually don't put the test binary
-//   directory in PATH.
-//
-// TODO(wan@google.com): make thread-safe death tests search the PATH.
-
-// Asserts that a given statement causes the program to exit, with an
-// integer exit status that satisfies predicate, and emitting error output
-// that matches regex.
-# define ASSERT_EXIT(statement, predicate, regex) \
-    GTEST_DEATH_TEST_(statement, predicate, regex, GTEST_FATAL_FAILURE_)
-
-// Like ASSERT_EXIT, but continues on to successive tests in the
-// test case, if any:
-# define EXPECT_EXIT(statement, predicate, regex) \
-    GTEST_DEATH_TEST_(statement, predicate, regex, GTEST_NONFATAL_FAILURE_)
-
-// Asserts that a given statement causes the program to exit, either by
-// explicitly exiting with a nonzero exit code or being killed by a
-// signal, and emitting error output that matches regex.
-# define ASSERT_DEATH(statement, regex) \
-    ASSERT_EXIT(statement, ::testing::internal::ExitedUnsuccessfully, regex)
-
-// Like ASSERT_DEATH, but continues on to successive tests in the
-// test case, if any:
-# define EXPECT_DEATH(statement, regex) \
-    EXPECT_EXIT(statement, ::testing::internal::ExitedUnsuccessfully, regex)
-
-// Two predicate classes that can be used in {ASSERT,EXPECT}_EXIT*:
-
-// Tests that an exit code describes a normal exit with a given exit code.
-class GTEST_API_ ExitedWithCode {
- public:
-  explicit ExitedWithCode(int exit_code);
-  bool operator()(int exit_status) const;
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ExitedWithCode& other);
-
-  const int exit_code_;
-};
-
-# if !GTEST_OS_WINDOWS && !GTEST_OS_FUCHSIA
-// Tests that an exit code describes an exit due to termination by a
-// given signal.
-class GTEST_API_ KilledBySignal {
- public:
-  explicit KilledBySignal(int signum);
-  bool operator()(int exit_status) const;
- private:
-  const int signum_;
-};
-# endif  // !GTEST_OS_WINDOWS
-
-// EXPECT_DEBUG_DEATH asserts that the given statements die in debug mode.
-// The death testing framework causes this to have interesting semantics,
-// since the sideeffects of the call are only visible in opt mode, and not
-// in debug mode.
-//
-// In practice, this can be used to test functions that utilize the
-// LOG(DFATAL) macro using the following style:
-//
-// int DieInDebugOr12(int* sideeffect) {
-//   if (sideeffect) {
-//     *sideeffect = 12;
-//   }
-//   LOG(DFATAL) << "death";
-//   return 12;
-// }
-//
-// TEST(TestCase, TestDieOr12WorksInDgbAndOpt) {
-//   int sideeffect = 0;
-//   // Only asserts in dbg.
-//   EXPECT_DEBUG_DEATH(DieInDebugOr12(&sideeffect), "death");
-//
-// #ifdef NDEBUG
-//   // opt-mode has sideeffect visible.
-//   EXPECT_EQ(12, sideeffect);
-// #else
-//   // dbg-mode no visible sideeffect.
-//   EXPECT_EQ(0, sideeffect);
-// #endif
-// }
-//
-// This will assert that DieInDebugReturn12InOpt() crashes in debug
-// mode, usually due to a DCHECK or LOG(DFATAL), but returns the
-// appropriate fallback value (12 in this case) in opt mode. If you
-// need to test that a function has appropriate side-effects in opt
-// mode, include assertions against the side-effects.  A general
-// pattern for this is:
-//
-// EXPECT_DEBUG_DEATH({
-//   // Side-effects here will have an effect after this statement in
-//   // opt mode, but none in debug mode.
-//   EXPECT_EQ(12, DieInDebugOr12(&sideeffect));
-// }, "death");
-//
-# ifdef NDEBUG
-
-#  define EXPECT_DEBUG_DEATH(statement, regex) \
-  GTEST_EXECUTE_STATEMENT_(statement, regex)
-
-#  define ASSERT_DEBUG_DEATH(statement, regex) \
-  GTEST_EXECUTE_STATEMENT_(statement, regex)
-
-# else
-
-#  define EXPECT_DEBUG_DEATH(statement, regex) \
-  EXPECT_DEATH(statement, regex)
-
-#  define ASSERT_DEBUG_DEATH(statement, regex) \
-  ASSERT_DEATH(statement, regex)
-
-# endif  // NDEBUG for EXPECT_DEBUG_DEATH
-#endif  // GTEST_HAS_DEATH_TEST
-
-// This macro is used for implementing macros such as
-// EXPECT_DEATH_IF_SUPPORTED and ASSERT_DEATH_IF_SUPPORTED on systems where
-// death tests are not supported. Those macros must compile on such systems
-// iff EXPECT_DEATH and ASSERT_DEATH compile with the same parameters on
-// systems that support death tests. This allows one to write such a macro
-// on a system that does not support death tests and be sure that it will
-// compile on a death-test supporting system. It is exposed publicly so that
-// systems that have death-tests with stricter requirements than
-// GTEST_HAS_DEATH_TEST can write their own equivalent of
-// EXPECT_DEATH_IF_SUPPORTED and ASSERT_DEATH_IF_SUPPORTED.
-//
-// Parameters:
-//   statement -  A statement that a macro such as EXPECT_DEATH would test
-//                for program termination. This macro has to make sure this
-//                statement is compiled but not executed, to ensure that
-//                EXPECT_DEATH_IF_SUPPORTED compiles with a certain
-//                parameter iff EXPECT_DEATH compiles with it.
-//   regex     -  A regex that a macro such as EXPECT_DEATH would use to test
-//                the output of statement.  This parameter has to be
-//                compiled but not evaluated by this macro, to ensure that
-//                this macro only accepts expressions that a macro such as
-//                EXPECT_DEATH would accept.
-//   terminator - Must be an empty statement for EXPECT_DEATH_IF_SUPPORTED
-//                and a return statement for ASSERT_DEATH_IF_SUPPORTED.
-//                This ensures that ASSERT_DEATH_IF_SUPPORTED will not
-//                compile inside functions where ASSERT_DEATH doesn't
-//                compile.
-//
-//  The branch that has an always false condition is used to ensure that
-//  statement and regex are compiled (and thus syntactically correct) but
-//  never executed. The unreachable code macro protects the terminator
-//  statement from generating an 'unreachable code' warning in case
-//  statement unconditionally returns or throws. The Message constructor at
-//  the end allows the syntax of streaming additional messages into the
-//  macro, for compilational compatibility with EXPECT_DEATH/ASSERT_DEATH.
-# define GTEST_UNSUPPORTED_DEATH_TEST(statement, regex, terminator) \
-    GTEST_AMBIGUOUS_ELSE_BLOCKER_ \
-    if (::testing::internal::AlwaysTrue()) { \
-      GTEST_LOG_(WARNING) \
-          << "Death tests are not supported on this platform.\n" \
-          << "Statement '" #statement "' cannot be verified."; \
-    } else if (::testing::internal::AlwaysFalse()) { \
-      ::testing::internal::RE::PartialMatch(".*", (regex)); \
-      GTEST_SUPPRESS_UNREACHABLE_CODE_WARNING_BELOW_(statement); \
-      terminator; \
-    } else \
-      ::testing::Message()
-
-// EXPECT_DEATH_IF_SUPPORTED(statement, regex) and
-// ASSERT_DEATH_IF_SUPPORTED(statement, regex) expand to real death tests if
-// death tests are supported; otherwise they just issue a warning.  This is
-// useful when you are combining death test assertions with normal test
-// assertions in one test.
-#if GTEST_HAS_DEATH_TEST
-# define EXPECT_DEATH_IF_SUPPORTED(statement, regex) \
-    EXPECT_DEATH(statement, regex)
-# define ASSERT_DEATH_IF_SUPPORTED(statement, regex) \
-    ASSERT_DEATH(statement, regex)
-#else
-# define EXPECT_DEATH_IF_SUPPORTED(statement, regex) \
-    GTEST_UNSUPPORTED_DEATH_TEST(statement, regex, )
-# define ASSERT_DEATH_IF_SUPPORTED(statement, regex) \
-    GTEST_UNSUPPORTED_DEATH_TEST(statement, regex, return)
-#endif
-
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_DEATH_TEST_H_
-// This file was GENERATED by command:
-//     pump.py gtest-param-test.h.pump
-// DO NOT EDIT BY HAND!!!
-
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Authors: vladl@google.com (Vlad Losev)
-//
-// Macros and functions for implementing parameterized tests
-// in Google C++ Testing and Mocking Framework (Google Test)
-//
-// This file is generated by a SCRIPT.  DO NOT EDIT BY HAND!
-//
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_PARAM_TEST_H_
-#define GTEST_INCLUDE_GTEST_GTEST_PARAM_TEST_H_
-
-
-// Value-parameterized tests allow you to test your code with different
-// parameters without writing multiple copies of the same test.
-//
-// Here is how you use value-parameterized tests:
-
-#if 0
-
-// To write value-parameterized tests, first you should define a fixture
-// class. It is usually derived from testing::TestWithParam<T> (see below for
-// another inheritance scheme that's sometimes useful in more complicated
-// class hierarchies), where the type of your parameter values.
-// TestWithParam<T> is itself derived from testing::Test. T can be any
-// copyable type. If it's a raw pointer, you are responsible for managing the
-// lifespan of the pointed values.
-
-class FooTest : public ::testing::TestWithParam<const char*> {
-  // You can implement all the usual class fixture members here.
-};
-
-// Then, use the TEST_P macro to define as many parameterized tests
-// for this fixture as you want. The _P suffix is for "parameterized"
-// or "pattern", whichever you prefer to think.
-
-TEST_P(FooTest, DoesBlah) {
-  // Inside a test, access the test parameter with the GetParam() method
-  // of the TestWithParam<T> class:
-  EXPECT_TRUE(foo.Blah(GetParam()));
-  ...
-}
-
-TEST_P(FooTest, HasBlahBlah) {
-  ...
-}
-
-// Finally, you can use INSTANTIATE_TEST_CASE_P to instantiate the test
-// case with any set of parameters you want. Google Test defines a number
-// of functions for generating test parameters. They return what we call
-// (surprise!) parameter generators. Here is a summary of them, which
-// are all in the testing namespace:
-//
-//
-//  Range(begin, end [, step]) - Yields values {begin, begin+step,
-//                               begin+step+step, ...}. The values do not
-//                               include end. step defaults to 1.
-//  Values(v1, v2, ..., vN)    - Yields values {v1, v2, ..., vN}.
-//  ValuesIn(container)        - Yields values from a C-style array, an STL
-//  ValuesIn(begin,end)          container, or an iterator range [begin, end).
-//  Bool()                     - Yields sequence {false, true}.
-//  Combine(g1, g2, ..., gN)   - Yields all combinations (the Cartesian product
-//                               for the math savvy) of the values generated
-//                               by the N generators.
-//
-// For more details, see comments at the definitions of these functions below
-// in this file.
-//
-// The following statement will instantiate tests from the FooTest test case
-// each with parameter values "meeny", "miny", and "moe".
-
-INSTANTIATE_TEST_CASE_P(InstantiationName,
-                        FooTest,
-                        Values("meeny", "miny", "moe"));
-
-// To distinguish different instances of the pattern, (yes, you
-// can instantiate it more then once) the first argument to the
-// INSTANTIATE_TEST_CASE_P macro is a prefix that will be added to the
-// actual test case name. Remember to pick unique prefixes for different
-// instantiations. The tests from the instantiation above will have
-// these names:
-//
-//    * InstantiationName/FooTest.DoesBlah/0 for "meeny"
-//    * InstantiationName/FooTest.DoesBlah/1 for "miny"
-//    * InstantiationName/FooTest.DoesBlah/2 for "moe"
-//    * InstantiationName/FooTest.HasBlahBlah/0 for "meeny"
-//    * InstantiationName/FooTest.HasBlahBlah/1 for "miny"
-//    * InstantiationName/FooTest.HasBlahBlah/2 for "moe"
-//
-// You can use these names in --gtest_filter.
-//
-// This statement will instantiate all tests from FooTest again, each
-// with parameter values "cat" and "dog":
-
-const char* pets[] = {"cat", "dog"};
-INSTANTIATE_TEST_CASE_P(AnotherInstantiationName, FooTest, ValuesIn(pets));
-
-// The tests from the instantiation above will have these names:
-//
-//    * AnotherInstantiationName/FooTest.DoesBlah/0 for "cat"
-//    * AnotherInstantiationName/FooTest.DoesBlah/1 for "dog"
-//    * AnotherInstantiationName/FooTest.HasBlahBlah/0 for "cat"
-//    * AnotherInstantiationName/FooTest.HasBlahBlah/1 for "dog"
-//
-// Please note that INSTANTIATE_TEST_CASE_P will instantiate all tests
-// in the given test case, whether their definitions come before or
-// AFTER the INSTANTIATE_TEST_CASE_P statement.
-//
-// Please also note that generator expressions (including parameters to the
-// generators) are evaluated in InitGoogleTest(), after main() has started.
-// This allows the user on one hand, to adjust generator parameters in order
-// to dynamically determine a set of tests to run and on the other hand,
-// give the user a chance to inspect the generated tests with Google Test
-// reflection API before RUN_ALL_TESTS() is executed.
-//
-// You can see samples/sample7_unittest.cc and samples/sample8_unittest.cc
-// for more examples.
-//
-// In the future, we plan to publish the API for defining new parameter
-// generators. But for now this interface remains part of the internal
-// implementation and is subject to change.
-//
-//
-// A parameterized test fixture must be derived from testing::Test and from
-// testing::WithParamInterface<T>, where T is the type of the parameter
-// values. Inheriting from TestWithParam<T> satisfies that requirement because
-// TestWithParam<T> inherits from both Test and WithParamInterface. In more
-// complicated hierarchies, however, it is occasionally useful to inherit
-// separately from Test and WithParamInterface. For example:
-
-class BaseTest : public ::testing::Test {
-  // You can inherit all the usual members for a non-parameterized test
-  // fixture here.
-};
-
-class DerivedTest : public BaseTest, public ::testing::WithParamInterface<int> {
-  // The usual test fixture members go here too.
-};
-
-TEST_F(BaseTest, HasFoo) {
-  // This is an ordinary non-parameterized test.
-}
-
-TEST_P(DerivedTest, DoesBlah) {
-  // GetParam works just the same here as if you inherit from TestWithParam.
-  EXPECT_TRUE(foo.Blah(GetParam()));
-}
-
-#endif  // 0
-
-
-#if !GTEST_OS_SYMBIAN
-# include <utility>
-#endif
-
-// Copyright 2008 Google Inc.
-// All Rights Reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: vladl@google.com (Vlad Losev)
-
-// Type and function utilities for implementing parameterized tests.
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PARAM_UTIL_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PARAM_UTIL_H_
-
-#include <ctype.h>
-
-#include <iterator>
-#include <set>
-#include <utility>
-#include <vector>
-
-// Copyright 2003 Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Authors: Dan Egnor (egnor@google.com)
-//
-// A "smart" pointer type with reference tracking.  Every pointer to a
-// particular object is kept on a circular linked list.  When the last pointer
-// to an object is destroyed or reassigned, the object is deleted.
-//
-// Used properly, this deletes the object when the last reference goes away.
-// There are several caveats:
-// - Like all reference counting schemes, cycles lead to leaks.
-// - Each smart pointer is actually two pointers (8 bytes instead of 4).
-// - Every time a pointer is assigned, the entire list of pointers to that
-//   object is traversed.  This class is therefore NOT SUITABLE when there
-//   will often be more than two or three pointers to a particular object.
-// - References are only tracked as long as linked_ptr<> objects are copied.
-//   If a linked_ptr<> is converted to a raw pointer and back, BAD THINGS
-//   will happen (double deletion).
-//
-// A good use of this class is storing object references in STL containers.
-// You can safely put linked_ptr<> in a vector<>.
-// Other uses may not be as good.
-//
-// Note: If you use an incomplete type with linked_ptr<>, the class
-// *containing* linked_ptr<> must have a constructor and destructor (even
-// if they do nothing!).
-//
-// Bill Gibbons suggested we use something like this.
-//
-// Thread Safety:
-//   Unlike other linked_ptr implementations, in this implementation
-//   a linked_ptr object is thread-safe in the sense that:
-//     - it's safe to copy linked_ptr objects concurrently,
-//     - it's safe to copy *from* a linked_ptr and read its underlying
-//       raw pointer (e.g. via get()) concurrently, and
-//     - it's safe to write to two linked_ptrs that point to the same
-//       shared object concurrently.
-// TODO(wan@google.com): rename this to safe_linked_ptr to avoid
-// confusion with normal linked_ptr.
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_LINKED_PTR_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_LINKED_PTR_H_
-
-#include <stdlib.h>
-#include <assert.h>
-
-
-namespace testing {
-namespace internal {
-
-// Protects copying of all linked_ptr objects.
-GTEST_API_ GTEST_DECLARE_STATIC_MUTEX_(g_linked_ptr_mutex);
-
-// This is used internally by all instances of linked_ptr<>.  It needs to be
-// a non-template class because different types of linked_ptr<> can refer to
-// the same object (linked_ptr<Superclass>(obj) vs linked_ptr<Subclass>(obj)).
-// So, it needs to be possible for different types of linked_ptr to participate
-// in the same circular linked list, so we need a single class type here.
-//
-// DO NOT USE THIS CLASS DIRECTLY YOURSELF.  Use linked_ptr<T>.
-class linked_ptr_internal {
- public:
-  // Create a new circle that includes only this instance.
-  void join_new() {
-    next_ = this;
-  }
-
-  // Many linked_ptr operations may change p.link_ for some linked_ptr
-  // variable p in the same circle as this object.  Therefore we need
-  // to prevent two such operations from occurring concurrently.
-  //
-  // Note that different types of linked_ptr objects can coexist in a
-  // circle (e.g. linked_ptr<Base>, linked_ptr<Derived1>, and
-  // linked_ptr<Derived2>).  Therefore we must use a single mutex to
-  // protect all linked_ptr objects.  This can create serious
-  // contention in production code, but is acceptable in a testing
-  // framework.
-
-  // Join an existing circle.
-  void join(linked_ptr_internal const* ptr)
-      GTEST_LOCK_EXCLUDED_(g_linked_ptr_mutex) {
-    MutexLock lock(&g_linked_ptr_mutex);
-
-    linked_ptr_internal const* p = ptr;
-    while (p->next_ != ptr) {
-      assert(p->next_ != this &&
-             "Trying to join() a linked ring we are already in. "
-             "Is GMock thread safety enabled?");
-      p = p->next_;
-    }
-    p->next_ = this;
-    next_ = ptr;
-  }
-
-  // Leave whatever circle we're part of.  Returns true if we were the
-  // last member of the circle.  Once this is done, you can join() another.
-  bool depart()
-      GTEST_LOCK_EXCLUDED_(g_linked_ptr_mutex) {
-    MutexLock lock(&g_linked_ptr_mutex);
-
-    if (next_ == this) return true;
-    linked_ptr_internal const* p = next_;
-    while (p->next_ != this) {
-      assert(p->next_ != next_ &&
-             "Trying to depart() a linked ring we are not in. "
-             "Is GMock thread safety enabled?");
-      p = p->next_;
-    }
-    p->next_ = next_;
-    return false;
-  }
-
- private:
-  mutable linked_ptr_internal const* next_;
-};
-
-template <typename T>
-class linked_ptr {
- public:
-  typedef T element_type;
-
-  // Take over ownership of a raw pointer.  This should happen as soon as
-  // possible after the object is created.
-  explicit linked_ptr(T* ptr = NULL) { capture(ptr); }
-  ~linked_ptr() { depart(); }
-
-  // Copy an existing linked_ptr<>, adding ourselves to the list of references.
-  template <typename U> linked_ptr(linked_ptr<U> const& ptr) { copy(&ptr); }
-  linked_ptr(linked_ptr const& ptr) {  // NOLINT
-    assert(&ptr != this);
-    copy(&ptr);
-  }
-
-  // Assignment releases the old value and acquires the new.
-  template <typename U> linked_ptr& operator=(linked_ptr<U> const& ptr) {
-    depart();
-    copy(&ptr);
-    return *this;
-  }
-
-  linked_ptr& operator=(linked_ptr const& ptr) {
-    if (&ptr != this) {
-      depart();
-      copy(&ptr);
-    }
-    return *this;
-  }
-
-  // Smart pointer members.
-  void reset(T* ptr = NULL) {
-    depart();
-    capture(ptr);
-  }
-  T* get() const { return value_; }
-  T* operator->() const { return value_; }
-  T& operator*() const { return *value_; }
-
-  bool operator==(T* p) const { return value_ == p; }
-  bool operator!=(T* p) const { return value_ != p; }
-  template <typename U>
-  bool operator==(linked_ptr<U> const& ptr) const {
-    return value_ == ptr.get();
-  }
-  template <typename U>
-  bool operator!=(linked_ptr<U> const& ptr) const {
-    return value_ != ptr.get();
-  }
-
- private:
-  template <typename U>
-  friend class linked_ptr;
-
-  T* value_;
-  linked_ptr_internal link_;
-
-  void depart() {
-    if (link_.depart()) delete value_;
-  }
-
-  void capture(T* ptr) {
-    value_ = ptr;
-    link_.join_new();
-  }
-
-  template <typename U> void copy(linked_ptr<U> const* ptr) {
-    value_ = ptr->get();
-    if (value_)
-      link_.join(&ptr->link_);
-    else
-      link_.join_new();
-  }
-};
-
-template<typename T> inline
-bool operator==(T* ptr, const linked_ptr<T>& x) {
-  return ptr == x.get();
-}
-
-template<typename T> inline
-bool operator!=(T* ptr, const linked_ptr<T>& x) {
-  return ptr != x.get();
-}
-
-// A function to convert T* into linked_ptr<T>
-// Doing e.g. make_linked_ptr(new FooBarBaz<type>(arg)) is a shorter notation
-// for linked_ptr<FooBarBaz<type> >(new FooBarBaz<type>(arg))
-template <typename T>
-linked_ptr<T> make_linked_ptr(T* ptr) {
-  return linked_ptr<T>(ptr);
-}
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_LINKED_PTR_H_
-// Copyright 2007, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-// Google Test - The Google C++ Testing and Mocking Framework
-//
-// This file implements a universal value printer that can print a
-// value of any type T:
-//
-//   void ::testing::internal::UniversalPrinter<T>::Print(value, ostream_ptr);
-//
-// A user can teach this function how to print a class type T by
-// defining either operator<<() or PrintTo() in the namespace that
-// defines T.  More specifically, the FIRST defined function in the
-// following list will be used (assuming T is defined in namespace
-// foo):
-//
-//   1. foo::PrintTo(const T&, ostream*)
-//   2. operator<<(ostream&, const T&) defined in either foo or the
-//      global namespace.
-//
-// However if T is an STL-style container then it is printed element-wise
-// unless foo::PrintTo(const T&, ostream*) is defined. Note that
-// operator<<() is ignored for container types.
-//
-// If none of the above is defined, it will print the debug string of
-// the value if it is a protocol buffer, or print the raw bytes in the
-// value otherwise.
-//
-// To aid debugging: when T is a reference type, the address of the
-// value is also printed; when T is a (const) char pointer, both the
-// pointer value and the NUL-terminated string it points to are
-// printed.
-//
-// We also provide some convenient wrappers:
-//
-//   // Prints a value to a string.  For a (const or not) char
-//   // pointer, the NUL-terminated string (but not the pointer) is
-//   // printed.
-//   std::string ::testing::PrintToString(const T& value);
-//
-//   // Prints a value tersely: for a reference type, the referenced
-//   // value (but not the address) is printed; for a (const or not) char
-//   // pointer, the NUL-terminated string (but not the pointer) is
-//   // printed.
-//   void ::testing::internal::UniversalTersePrint(const T& value, ostream*);
-//
-//   // Prints value using the type inferred by the compiler.  The difference
-//   // from UniversalTersePrint() is that this function prints both the
-//   // pointer and the NUL-terminated string for a (const or not) char pointer.
-//   void ::testing::internal::UniversalPrint(const T& value, ostream*);
-//
-//   // Prints the fields of a tuple tersely to a string vector, one
-//   // element for each field. Tuple support must be enabled in
-//   // gtest-port.h.
-//   std::vector<string> UniversalTersePrintTupleFieldsToStrings(
-//       const Tuple& value);
-//
-// Known limitation:
-//
-// The print primitives print the elements of an STL-style container
-// using the compiler-inferred type of *iter where iter is a
-// const_iterator of the container.  When const_iterator is an input
-// iterator but not a forward iterator, this inferred type may not
-// match value_type, and the print output may be incorrect.  In
-// practice, this is rarely a problem as for most containers
-// const_iterator is a forward iterator.  We'll fix this if there's an
-// actual need for it.  Note that this fix cannot rely on value_type
-// being defined as many user-defined container types don't have
-// value_type.
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_PRINTERS_H_
-#define GTEST_INCLUDE_GTEST_GTEST_PRINTERS_H_
-
-#include <ostream>  // NOLINT
-#include <sstream>
-#include <string>
-#include <utility>
-#include <vector>
-
-#if GTEST_HAS_STD_TUPLE_
-# include <tuple>
-#endif
-
-#if GTEST_HAS_ABSL
-#include "absl/strings/string_view.h"
-#include "absl/types/optional.h"
-#include "absl/types/variant.h"
-#endif  // GTEST_HAS_ABSL
-
-namespace testing {
-
-// Definitions in the 'internal' and 'internal2' name spaces are
-// subject to change without notice.  DO NOT USE THEM IN USER CODE!
-namespace internal2 {
-
-// Prints the given number of bytes in the given object to the given
-// ostream.
-GTEST_API_ void PrintBytesInObjectTo(const unsigned char* obj_bytes,
-                                     size_t count,
-                                     ::std::ostream* os);
-
-// For selecting which printer to use when a given type has neither <<
-// nor PrintTo().
-enum TypeKind {
-  kProtobuf,              // a protobuf type
-  kConvertibleToInteger,  // a type implicitly convertible to BiggestInt
-                          // (e.g. a named or unnamed enum type)
-#if GTEST_HAS_ABSL
-  kConvertibleToStringView,  // a type implicitly convertible to
-                             // absl::string_view
-#endif
-  kOtherType  // anything else
-};
-
-// TypeWithoutFormatter<T, kTypeKind>::PrintValue(value, os) is called
-// by the universal printer to print a value of type T when neither
-// operator<< nor PrintTo() is defined for T, where kTypeKind is the
-// "kind" of T as defined by enum TypeKind.
-template <typename T, TypeKind kTypeKind>
-class TypeWithoutFormatter {
- public:
-  // This default version is called when kTypeKind is kOtherType.
-  static void PrintValue(const T& value, ::std::ostream* os) {
-    PrintBytesInObjectTo(static_cast<const unsigned char*>(
-                             reinterpret_cast<const void*>(&value)),
-                         sizeof(value), os);
-  }
-};
-
-// We print a protobuf using its ShortDebugString() when the string
-// doesn't exceed this many characters; otherwise we print it using
-// DebugString() for better readability.
-const size_t kProtobufOneLinerMaxLength = 50;
-
-template <typename T>
-class TypeWithoutFormatter<T, kProtobuf> {
- public:
-  static void PrintValue(const T& value, ::std::ostream* os) {
-    std::string pretty_str = value.ShortDebugString();
-    if (pretty_str.length() > kProtobufOneLinerMaxLength) {
-      pretty_str = "\n" + value.DebugString();
-    }
-    *os << ("<" + pretty_str + ">");
-  }
-};
-
-template <typename T>
-class TypeWithoutFormatter<T, kConvertibleToInteger> {
- public:
-  // Since T has no << operator or PrintTo() but can be implicitly
-  // converted to BiggestInt, we print it as a BiggestInt.
-  //
-  // Most likely T is an enum type (either named or unnamed), in which
-  // case printing it as an integer is the desired behavior.  In case
-  // T is not an enum, printing it as an integer is the best we can do
-  // given that it has no user-defined printer.
-  static void PrintValue(const T& value, ::std::ostream* os) {
-    const internal::BiggestInt kBigInt = value;
-    *os << kBigInt;
-  }
-};
-
-#if GTEST_HAS_ABSL
-template <typename T>
-class TypeWithoutFormatter<T, kConvertibleToStringView> {
- public:
-  // Since T has neither operator<< nor PrintTo() but can be implicitly
-  // converted to absl::string_view, we print it as a absl::string_view.
-  //
-  // Note: the implementation is further below, as it depends on
-  // internal::PrintTo symbol which is defined later in the file.
-  static void PrintValue(const T& value, ::std::ostream* os);
-};
-#endif
-
-// Prints the given value to the given ostream.  If the value is a
-// protocol message, its debug string is printed; if it's an enum or
-// of a type implicitly convertible to BiggestInt, it's printed as an
-// integer; otherwise the bytes in the value are printed.  This is
-// what UniversalPrinter<T>::Print() does when it knows nothing about
-// type T and T has neither << operator nor PrintTo().
-//
-// A user can override this behavior for a class type Foo by defining
-// a << operator in the namespace where Foo is defined.
-//
-// We put this operator in namespace 'internal2' instead of 'internal'
-// to simplify the implementation, as much code in 'internal' needs to
-// use << in STL, which would conflict with our own << were it defined
-// in 'internal'.
-//
-// Note that this operator<< takes a generic std::basic_ostream<Char,
-// CharTraits> type instead of the more restricted std::ostream.  If
-// we define it to take an std::ostream instead, we'll get an
-// "ambiguous overloads" compiler error when trying to print a type
-// Foo that supports streaming to std::basic_ostream<Char,
-// CharTraits>, as the compiler cannot tell whether
-// operator<<(std::ostream&, const T&) or
-// operator<<(std::basic_stream<Char, CharTraits>, const Foo&) is more
-// specific.
-template <typename Char, typename CharTraits, typename T>
-::std::basic_ostream<Char, CharTraits>& operator<<(
-    ::std::basic_ostream<Char, CharTraits>& os, const T& x) {
-  TypeWithoutFormatter<T, (internal::IsAProtocolMessage<T>::value
-                               ? kProtobuf
-                               : internal::ImplicitlyConvertible<
-                                     const T&, internal::BiggestInt>::value
-                                     ? kConvertibleToInteger
-                                     :
-#if GTEST_HAS_ABSL
-                                     internal::ImplicitlyConvertible<
-                                         const T&, absl::string_view>::value
-                                         ? kConvertibleToStringView
-                                         :
-#endif
-                                         kOtherType)>::PrintValue(x, &os);
-  return os;
-}
-
-}  // namespace internal2
-}  // namespace testing
-
-// This namespace MUST NOT BE NESTED IN ::testing, or the name look-up
-// magic needed for implementing UniversalPrinter won't work.
-namespace testing_internal {
-
-// Used to print a value that is not an STL-style container when the
-// user doesn't define PrintTo() for it.
-template <typename T>
-void DefaultPrintNonContainerTo(const T& value, ::std::ostream* os) {
-  // With the following statement, during unqualified name lookup,
-  // testing::internal2::operator<< appears as if it was declared in
-  // the nearest enclosing namespace that contains both
-  // ::testing_internal and ::testing::internal2, i.e. the global
-  // namespace.  For more details, refer to the C++ Standard section
-  // 7.3.4-1 [namespace.udir].  This allows us to fall back onto
-  // testing::internal2::operator<< in case T doesn't come with a <<
-  // operator.
-  //
-  // We cannot write 'using ::testing::internal2::operator<<;', which
-  // gcc 3.3 fails to compile due to a compiler bug.
-  using namespace ::testing::internal2;  // NOLINT
-
-  // Assuming T is defined in namespace foo, in the next statement,
-  // the compiler will consider all of:
-  //
-  //   1. foo::operator<< (thanks to Koenig look-up),
-  //   2. ::operator<< (as the current namespace is enclosed in ::),
-  //   3. testing::internal2::operator<< (thanks to the using statement above).
-  //
-  // The operator<< whose type matches T best will be picked.
-  //
-  // We deliberately allow #2 to be a candidate, as sometimes it's
-  // impossible to define #1 (e.g. when foo is ::std, defining
-  // anything in it is undefined behavior unless you are a compiler
-  // vendor.).
-  *os << value;
-}
-
-}  // namespace testing_internal
-
-namespace testing {
-namespace internal {
-
-// FormatForComparison<ToPrint, OtherOperand>::Format(value) formats a
-// value of type ToPrint that is an operand of a comparison assertion
-// (e.g. ASSERT_EQ).  OtherOperand is the type of the other operand in
-// the comparison, and is used to help determine the best way to
-// format the value.  In particular, when the value is a C string
-// (char pointer) and the other operand is an STL string object, we
-// want to format the C string as a string, since we know it is
-// compared by value with the string object.  If the value is a char
-// pointer but the other operand is not an STL string object, we don't
-// know whether the pointer is supposed to point to a NUL-terminated
-// string, and thus want to print it as a pointer to be safe.
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-
-// The default case.
-template <typename ToPrint, typename OtherOperand>
-class FormatForComparison {
- public:
-  static ::std::string Format(const ToPrint& value) {
-    return ::testing::PrintToString(value);
-  }
-};
-
-// Array.
-template <typename ToPrint, size_t N, typename OtherOperand>
-class FormatForComparison<ToPrint[N], OtherOperand> {
- public:
-  static ::std::string Format(const ToPrint* value) {
-    return FormatForComparison<const ToPrint*, OtherOperand>::Format(value);
-  }
-};
-
-// By default, print C string as pointers to be safe, as we don't know
-// whether they actually point to a NUL-terminated string.
-
-#define GTEST_IMPL_FORMAT_C_STRING_AS_POINTER_(CharType)                \
-  template <typename OtherOperand>                                      \
-  class FormatForComparison<CharType*, OtherOperand> {                  \
-   public:                                                              \
-    static ::std::string Format(CharType* value) {                      \
-      return ::testing::PrintToString(static_cast<const void*>(value)); \
-    }                                                                   \
-  }
-
-GTEST_IMPL_FORMAT_C_STRING_AS_POINTER_(char);
-GTEST_IMPL_FORMAT_C_STRING_AS_POINTER_(const char);
-GTEST_IMPL_FORMAT_C_STRING_AS_POINTER_(wchar_t);
-GTEST_IMPL_FORMAT_C_STRING_AS_POINTER_(const wchar_t);
-
-#undef GTEST_IMPL_FORMAT_C_STRING_AS_POINTER_
-
-// If a C string is compared with an STL string object, we know it's meant
-// to point to a NUL-terminated string, and thus can print it as a string.
-
-#define GTEST_IMPL_FORMAT_C_STRING_AS_STRING_(CharType, OtherStringType) \
-  template <>                                                           \
-  class FormatForComparison<CharType*, OtherStringType> {               \
-   public:                                                              \
-    static ::std::string Format(CharType* value) {                      \
-      return ::testing::PrintToString(value);                           \
-    }                                                                   \
-  }
-
-GTEST_IMPL_FORMAT_C_STRING_AS_STRING_(char, ::std::string);
-GTEST_IMPL_FORMAT_C_STRING_AS_STRING_(const char, ::std::string);
-
-#if GTEST_HAS_GLOBAL_STRING
-GTEST_IMPL_FORMAT_C_STRING_AS_STRING_(char, ::string);
-GTEST_IMPL_FORMAT_C_STRING_AS_STRING_(const char, ::string);
-#endif
-
-#if GTEST_HAS_GLOBAL_WSTRING
-GTEST_IMPL_FORMAT_C_STRING_AS_STRING_(wchar_t, ::wstring);
-GTEST_IMPL_FORMAT_C_STRING_AS_STRING_(const wchar_t, ::wstring);
-#endif
-
-#if GTEST_HAS_STD_WSTRING
-GTEST_IMPL_FORMAT_C_STRING_AS_STRING_(wchar_t, ::std::wstring);
-GTEST_IMPL_FORMAT_C_STRING_AS_STRING_(const wchar_t, ::std::wstring);
-#endif
-
-#undef GTEST_IMPL_FORMAT_C_STRING_AS_STRING_
-
-// Formats a comparison assertion (e.g. ASSERT_EQ, EXPECT_LT, and etc)
-// operand to be used in a failure message.  The type (but not value)
-// of the other operand may affect the format.  This allows us to
-// print a char* as a raw pointer when it is compared against another
-// char* or void*, and print it as a C string when it is compared
-// against an std::string object, for example.
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-template <typename T1, typename T2>
-std::string FormatForComparisonFailureMessage(
-    const T1& value, const T2& /* other_operand */) {
-  return FormatForComparison<T1, T2>::Format(value);
-}
-
-// UniversalPrinter<T>::Print(value, ostream_ptr) prints the given
-// value to the given ostream.  The caller must ensure that
-// 'ostream_ptr' is not NULL, or the behavior is undefined.
-//
-// We define UniversalPrinter as a class template (as opposed to a
-// function template), as we need to partially specialize it for
-// reference types, which cannot be done with function templates.
-template <typename T>
-class UniversalPrinter;
-
-template <typename T>
-void UniversalPrint(const T& value, ::std::ostream* os);
-
-enum DefaultPrinterType {
-  kPrintContainer,
-  kPrintPointer,
-  kPrintFunctionPointer,
-  kPrintOther,
-};
-template <DefaultPrinterType type> struct WrapPrinterType {};
-
-// Used to print an STL-style container when the user doesn't define
-// a PrintTo() for it.
-template <typename C>
-void DefaultPrintTo(WrapPrinterType<kPrintContainer> /* dummy */,
-                    const C& container, ::std::ostream* os) {
-  const size_t kMaxCount = 32;  // The maximum number of elements to print.
-  *os << '{';
-  size_t count = 0;
-  for (typename C::const_iterator it = container.begin();
-       it != container.end(); ++it, ++count) {
-    if (count > 0) {
-      *os << ',';
-      if (count == kMaxCount) {  // Enough has been printed.
-        *os << " ...";
-        break;
-      }
-    }
-    *os << ' ';
-    // We cannot call PrintTo(*it, os) here as PrintTo() doesn't
-    // handle *it being a native array.
-    internal::UniversalPrint(*it, os);
-  }
-
-  if (count > 0) {
-    *os << ' ';
-  }
-  *os << '}';
-}
-
-// Used to print a pointer that is neither a char pointer nor a member
-// pointer, when the user doesn't define PrintTo() for it.  (A member
-// variable pointer or member function pointer doesn't really point to
-// a location in the address space.  Their representation is
-// implementation-defined.  Therefore they will be printed as raw
-// bytes.)
-template <typename T>
-void DefaultPrintTo(WrapPrinterType<kPrintPointer> /* dummy */,
-                    T* p, ::std::ostream* os) {
-  if (p == NULL) {
-    *os << "NULL";
-  } else {
-    // T is not a function type.  We just call << to print p,
-    // relying on ADL to pick up user-defined << for their pointer
-    // types, if any.
-    *os << p;
-  }
-}
-template <typename T>
-void DefaultPrintTo(WrapPrinterType<kPrintFunctionPointer> /* dummy */,
-                    T* p, ::std::ostream* os) {
-  if (p == NULL) {
-    *os << "NULL";
-  } else {
-    // T is a function type, so '*os << p' doesn't do what we want
-    // (it just prints p as bool).  We want to print p as a const
-    // void*.
-    *os << reinterpret_cast<const void*>(p);
-  }
-}
-
-// Used to print a non-container, non-pointer value when the user
-// doesn't define PrintTo() for it.
-template <typename T>
-void DefaultPrintTo(WrapPrinterType<kPrintOther> /* dummy */,
-                    const T& value, ::std::ostream* os) {
-  ::testing_internal::DefaultPrintNonContainerTo(value, os);
-}
-
-// Prints the given value using the << operator if it has one;
-// otherwise prints the bytes in it.  This is what
-// UniversalPrinter<T>::Print() does when PrintTo() is not specialized
-// or overloaded for type T.
-//
-// A user can override this behavior for a class type Foo by defining
-// an overload of PrintTo() in the namespace where Foo is defined.  We
-// give the user this option as sometimes defining a << operator for
-// Foo is not desirable (e.g. the coding style may prevent doing it,
-// or there is already a << operator but it doesn't do what the user
-// wants).
-template <typename T>
-void PrintTo(const T& value, ::std::ostream* os) {
-  // DefaultPrintTo() is overloaded.  The type of its first argument
-  // determines which version will be picked.
-  //
-  // Note that we check for container types here, prior to we check
-  // for protocol message types in our operator<<.  The rationale is:
-  //
-  // For protocol messages, we want to give people a chance to
-  // override Google Mock's format by defining a PrintTo() or
-  // operator<<.  For STL containers, other formats can be
-  // incompatible with Google Mock's format for the container
-  // elements; therefore we check for container types here to ensure
-  // that our format is used.
-  //
-  // Note that MSVC and clang-cl do allow an implicit conversion from
-  // pointer-to-function to pointer-to-object, but clang-cl warns on it.
-  // So don't use ImplicitlyConvertible if it can be helped since it will
-  // cause this warning, and use a separate overload of DefaultPrintTo for
-  // function pointers so that the `*os << p` in the object pointer overload
-  // doesn't cause that warning either.
-  DefaultPrintTo(
-      WrapPrinterType <
-                  (sizeof(IsContainerTest<T>(0)) == sizeof(IsContainer)) &&
-              !IsRecursiveContainer<T>::value
-          ? kPrintContainer
-          : !is_pointer<T>::value
-                ? kPrintOther
-#if GTEST_LANG_CXX11
-                : std::is_function<typename std::remove_pointer<T>::type>::value
-#else
-                : !internal::ImplicitlyConvertible<T, const void*>::value
-#endif
-                      ? kPrintFunctionPointer
-                      : kPrintPointer > (),
-      value, os);
-}
-
-// The following list of PrintTo() overloads tells
-// UniversalPrinter<T>::Print() how to print standard types (built-in
-// types, strings, plain arrays, and pointers).
-
-// Overloads for various char types.
-GTEST_API_ void PrintTo(unsigned char c, ::std::ostream* os);
-GTEST_API_ void PrintTo(signed char c, ::std::ostream* os);
-inline void PrintTo(char c, ::std::ostream* os) {
-  // When printing a plain char, we always treat it as unsigned.  This
-  // way, the output won't be affected by whether the compiler thinks
-  // char is signed or not.
-  PrintTo(static_cast<unsigned char>(c), os);
-}
-
-// Overloads for other simple built-in types.
-inline void PrintTo(bool x, ::std::ostream* os) {
-  *os << (x ? "true" : "false");
-}
-
-// Overload for wchar_t type.
-// Prints a wchar_t as a symbol if it is printable or as its internal
-// code otherwise and also as its decimal code (except for L'\0').
-// The L'\0' char is printed as "L'\\0'". The decimal code is printed
-// as signed integer when wchar_t is implemented by the compiler
-// as a signed type and is printed as an unsigned integer when wchar_t
-// is implemented as an unsigned type.
-GTEST_API_ void PrintTo(wchar_t wc, ::std::ostream* os);
-
-// Overloads for C strings.
-GTEST_API_ void PrintTo(const char* s, ::std::ostream* os);
-inline void PrintTo(char* s, ::std::ostream* os) {
-  PrintTo(ImplicitCast_<const char*>(s), os);
-}
-
-// signed/unsigned char is often used for representing binary data, so
-// we print pointers to it as void* to be safe.
-inline void PrintTo(const signed char* s, ::std::ostream* os) {
-  PrintTo(ImplicitCast_<const void*>(s), os);
-}
-inline void PrintTo(signed char* s, ::std::ostream* os) {
-  PrintTo(ImplicitCast_<const void*>(s), os);
-}
-inline void PrintTo(const unsigned char* s, ::std::ostream* os) {
-  PrintTo(ImplicitCast_<const void*>(s), os);
-}
-inline void PrintTo(unsigned char* s, ::std::ostream* os) {
-  PrintTo(ImplicitCast_<const void*>(s), os);
-}
-
-// MSVC can be configured to define wchar_t as a typedef of unsigned
-// short.  It defines _NATIVE_WCHAR_T_DEFINED when wchar_t is a native
-// type.  When wchar_t is a typedef, defining an overload for const
-// wchar_t* would cause unsigned short* be printed as a wide string,
-// possibly causing invalid memory accesses.
-#if !defined(_MSC_VER) || defined(_NATIVE_WCHAR_T_DEFINED)
-// Overloads for wide C strings
-GTEST_API_ void PrintTo(const wchar_t* s, ::std::ostream* os);
-inline void PrintTo(wchar_t* s, ::std::ostream* os) {
-  PrintTo(ImplicitCast_<const wchar_t*>(s), os);
-}
-#endif
-
-// Overload for C arrays.  Multi-dimensional arrays are printed
-// properly.
-
-// Prints the given number of elements in an array, without printing
-// the curly braces.
-template <typename T>
-void PrintRawArrayTo(const T a[], size_t count, ::std::ostream* os) {
-  UniversalPrint(a[0], os);
-  for (size_t i = 1; i != count; i++) {
-    *os << ", ";
-    UniversalPrint(a[i], os);
-  }
-}
-
-// Overloads for ::string and ::std::string.
-#if GTEST_HAS_GLOBAL_STRING
-GTEST_API_ void PrintStringTo(const ::string&s, ::std::ostream* os);
-inline void PrintTo(const ::string& s, ::std::ostream* os) {
-  PrintStringTo(s, os);
-}
-#endif  // GTEST_HAS_GLOBAL_STRING
-
-GTEST_API_ void PrintStringTo(const ::std::string&s, ::std::ostream* os);
-inline void PrintTo(const ::std::string& s, ::std::ostream* os) {
-  PrintStringTo(s, os);
-}
-
-// Overloads for ::wstring and ::std::wstring.
-#if GTEST_HAS_GLOBAL_WSTRING
-GTEST_API_ void PrintWideStringTo(const ::wstring&s, ::std::ostream* os);
-inline void PrintTo(const ::wstring& s, ::std::ostream* os) {
-  PrintWideStringTo(s, os);
-}
-#endif  // GTEST_HAS_GLOBAL_WSTRING
-
-#if GTEST_HAS_STD_WSTRING
-GTEST_API_ void PrintWideStringTo(const ::std::wstring&s, ::std::ostream* os);
-inline void PrintTo(const ::std::wstring& s, ::std::ostream* os) {
-  PrintWideStringTo(s, os);
-}
-#endif  // GTEST_HAS_STD_WSTRING
-
-#if GTEST_HAS_ABSL
-// Overload for absl::string_view.
-inline void PrintTo(absl::string_view sp, ::std::ostream* os) {
-  PrintTo(::std::string(sp), os);
-}
-#endif  // GTEST_HAS_ABSL
-
-#if GTEST_LANG_CXX11
-inline void PrintTo(std::nullptr_t, ::std::ostream* os) { *os << "(nullptr)"; }
-#endif  // GTEST_LANG_CXX11
-
-#if GTEST_HAS_TR1_TUPLE || GTEST_HAS_STD_TUPLE_
-// Helper function for printing a tuple.  T must be instantiated with
-// a tuple type.
-template <typename T>
-void PrintTupleTo(const T& t, ::std::ostream* os);
-#endif  // GTEST_HAS_TR1_TUPLE || GTEST_HAS_STD_TUPLE_
-
-#if GTEST_HAS_TR1_TUPLE
-// Overload for ::std::tr1::tuple.  Needed for printing function arguments,
-// which are packed as tuples.
-
-// Overloaded PrintTo() for tuples of various arities.  We support
-// tuples of up-to 10 fields.  The following implementation works
-// regardless of whether tr1::tuple is implemented using the
-// non-standard variadic template feature or not.
-
-inline void PrintTo(const ::std::tr1::tuple<>& t, ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-
-template <typename T1>
-void PrintTo(const ::std::tr1::tuple<T1>& t, ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-
-template <typename T1, typename T2>
-void PrintTo(const ::std::tr1::tuple<T1, T2>& t, ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-
-template <typename T1, typename T2, typename T3>
-void PrintTo(const ::std::tr1::tuple<T1, T2, T3>& t, ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-
-template <typename T1, typename T2, typename T3, typename T4>
-void PrintTo(const ::std::tr1::tuple<T1, T2, T3, T4>& t, ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5>
-void PrintTo(const ::std::tr1::tuple<T1, T2, T3, T4, T5>& t,
-             ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-          typename T6>
-void PrintTo(const ::std::tr1::tuple<T1, T2, T3, T4, T5, T6>& t,
-             ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-          typename T6, typename T7>
-void PrintTo(const ::std::tr1::tuple<T1, T2, T3, T4, T5, T6, T7>& t,
-             ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-          typename T6, typename T7, typename T8>
-void PrintTo(const ::std::tr1::tuple<T1, T2, T3, T4, T5, T6, T7, T8>& t,
-             ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-          typename T6, typename T7, typename T8, typename T9>
-void PrintTo(const ::std::tr1::tuple<T1, T2, T3, T4, T5, T6, T7, T8, T9>& t,
-             ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-          typename T6, typename T7, typename T8, typename T9, typename T10>
-void PrintTo(
-    const ::std::tr1::tuple<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10>& t,
-    ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-#endif  // GTEST_HAS_TR1_TUPLE
-
-#if GTEST_HAS_STD_TUPLE_
-template <typename... Types>
-void PrintTo(const ::std::tuple<Types...>& t, ::std::ostream* os) {
-  PrintTupleTo(t, os);
-}
-#endif  // GTEST_HAS_STD_TUPLE_
-
-// Overload for std::pair.
-template <typename T1, typename T2>
-void PrintTo(const ::std::pair<T1, T2>& value, ::std::ostream* os) {
-  *os << '(';
-  // We cannot use UniversalPrint(value.first, os) here, as T1 may be
-  // a reference type.  The same for printing value.second.
-  UniversalPrinter<T1>::Print(value.first, os);
-  *os << ", ";
-  UniversalPrinter<T2>::Print(value.second, os);
-  *os << ')';
-}
-
-// Implements printing a non-reference type T by letting the compiler
-// pick the right overload of PrintTo() for T.
-template <typename T>
-class UniversalPrinter {
- public:
-  // MSVC warns about adding const to a function type, so we want to
-  // disable the warning.
-  GTEST_DISABLE_MSC_WARNINGS_PUSH_(4180)
-
-  // Note: we deliberately don't call this PrintTo(), as that name
-  // conflicts with ::testing::internal::PrintTo in the body of the
-  // function.
-  static void Print(const T& value, ::std::ostream* os) {
-    // By default, ::testing::internal::PrintTo() is used for printing
-    // the value.
-    //
-    // Thanks to Koenig look-up, if T is a class and has its own
-    // PrintTo() function defined in its namespace, that function will
-    // be visible here.  Since it is more specific than the generic ones
-    // in ::testing::internal, it will be picked by the compiler in the
-    // following statement - exactly what we want.
-    PrintTo(value, os);
-  }
-
-  GTEST_DISABLE_MSC_WARNINGS_POP_()
-};
-
-#if GTEST_HAS_ABSL
-
-// Printer for absl::optional
-
-template <typename T>
-class UniversalPrinter<::absl::optional<T>> {
- public:
-  static void Print(const ::absl::optional<T>& value, ::std::ostream* os) {
-    *os << '(';
-    if (!value) {
-      *os << "nullopt";
-    } else {
-      UniversalPrint(*value, os);
-    }
-    *os << ')';
-  }
-};
-
-// Printer for absl::variant
-
-template <typename... T>
-class UniversalPrinter<::absl::variant<T...>> {
- public:
-  static void Print(const ::absl::variant<T...>& value, ::std::ostream* os) {
-    *os << '(';
-    absl::visit(Visitor{os}, value);
-    *os << ')';
-  }
-
- private:
-  struct Visitor {
-    template <typename U>
-    void operator()(const U& u) const {
-      *os << "'" << GetTypeName<U>() << "' with value ";
-      UniversalPrint(u, os);
-    }
-    ::std::ostream* os;
-  };
-};
-
-#endif  // GTEST_HAS_ABSL
-
-// UniversalPrintArray(begin, len, os) prints an array of 'len'
-// elements, starting at address 'begin'.
-template <typename T>
-void UniversalPrintArray(const T* begin, size_t len, ::std::ostream* os) {
-  if (len == 0) {
-    *os << "{}";
-  } else {
-    *os << "{ ";
-    const size_t kThreshold = 18;
-    const size_t kChunkSize = 8;
-    // If the array has more than kThreshold elements, we'll have to
-    // omit some details by printing only the first and the last
-    // kChunkSize elements.
-    // TODO(wan@google.com): let the user control the threshold using a flag.
-    if (len <= kThreshold) {
-      PrintRawArrayTo(begin, len, os);
-    } else {
-      PrintRawArrayTo(begin, kChunkSize, os);
-      *os << ", ..., ";
-      PrintRawArrayTo(begin + len - kChunkSize, kChunkSize, os);
-    }
-    *os << " }";
-  }
-}
-// This overload prints a (const) char array compactly.
-GTEST_API_ void UniversalPrintArray(
-    const char* begin, size_t len, ::std::ostream* os);
-
-// This overload prints a (const) wchar_t array compactly.
-GTEST_API_ void UniversalPrintArray(
-    const wchar_t* begin, size_t len, ::std::ostream* os);
-
-// Implements printing an array type T[N].
-template <typename T, size_t N>
-class UniversalPrinter<T[N]> {
- public:
-  // Prints the given array, omitting some elements when there are too
-  // many.
-  static void Print(const T (&a)[N], ::std::ostream* os) {
-    UniversalPrintArray(a, N, os);
-  }
-};
-
-// Implements printing a reference type T&.
-template <typename T>
-class UniversalPrinter<T&> {
- public:
-  // MSVC warns about adding const to a function type, so we want to
-  // disable the warning.
-  GTEST_DISABLE_MSC_WARNINGS_PUSH_(4180)
-
-  static void Print(const T& value, ::std::ostream* os) {
-    // Prints the address of the value.  We use reinterpret_cast here
-    // as static_cast doesn't compile when T is a function type.
-    *os << "@" << reinterpret_cast<const void*>(&value) << " ";
-
-    // Then prints the value itself.
-    UniversalPrint(value, os);
-  }
-
-  GTEST_DISABLE_MSC_WARNINGS_POP_()
-};
-
-// Prints a value tersely: for a reference type, the referenced value
-// (but not the address) is printed; for a (const) char pointer, the
-// NUL-terminated string (but not the pointer) is printed.
-
-template <typename T>
-class UniversalTersePrinter {
- public:
-  static void Print(const T& value, ::std::ostream* os) {
-    UniversalPrint(value, os);
-  }
-};
-template <typename T>
-class UniversalTersePrinter<T&> {
- public:
-  static void Print(const T& value, ::std::ostream* os) {
-    UniversalPrint(value, os);
-  }
-};
-template <typename T, size_t N>
-class UniversalTersePrinter<T[N]> {
- public:
-  static void Print(const T (&value)[N], ::std::ostream* os) {
-    UniversalPrinter<T[N]>::Print(value, os);
-  }
-};
-template <>
-class UniversalTersePrinter<const char*> {
- public:
-  static void Print(const char* str, ::std::ostream* os) {
-    if (str == NULL) {
-      *os << "NULL";
-    } else {
-      UniversalPrint(std::string(str), os);
-    }
-  }
-};
-template <>
-class UniversalTersePrinter<char*> {
- public:
-  static void Print(char* str, ::std::ostream* os) {
-    UniversalTersePrinter<const char*>::Print(str, os);
-  }
-};
-
-#if GTEST_HAS_STD_WSTRING
-template <>
-class UniversalTersePrinter<const wchar_t*> {
- public:
-  static void Print(const wchar_t* str, ::std::ostream* os) {
-    if (str == NULL) {
-      *os << "NULL";
-    } else {
-      UniversalPrint(::std::wstring(str), os);
-    }
-  }
-};
-#endif
-
-template <>
-class UniversalTersePrinter<wchar_t*> {
- public:
-  static void Print(wchar_t* str, ::std::ostream* os) {
-    UniversalTersePrinter<const wchar_t*>::Print(str, os);
-  }
-};
-
-template <typename T>
-void UniversalTersePrint(const T& value, ::std::ostream* os) {
-  UniversalTersePrinter<T>::Print(value, os);
-}
-
-// Prints a value using the type inferred by the compiler.  The
-// difference between this and UniversalTersePrint() is that for a
-// (const) char pointer, this prints both the pointer and the
-// NUL-terminated string.
-template <typename T>
-void UniversalPrint(const T& value, ::std::ostream* os) {
-  // A workarond for the bug in VC++ 7.1 that prevents us from instantiating
-  // UniversalPrinter with T directly.
-  typedef T T1;
-  UniversalPrinter<T1>::Print(value, os);
-}
-
-typedef ::std::vector< ::std::string> Strings;
-
-// TuplePolicy<TupleT> must provide:
-// - tuple_size
-//     size of tuple TupleT.
-// - get<size_t I>(const TupleT& t)
-//     static function extracting element I of tuple TupleT.
-// - tuple_element<size_t I>::type
-//     type of element I of tuple TupleT.
-template <typename TupleT>
-struct TuplePolicy;
-
-#if GTEST_HAS_TR1_TUPLE
-template <typename TupleT>
-struct TuplePolicy {
-  typedef TupleT Tuple;
-  static const size_t tuple_size = ::std::tr1::tuple_size<Tuple>::value;
-
-  template <size_t I>
-  struct tuple_element : ::std::tr1::tuple_element<I, Tuple> {};
-
-  template <size_t I>
-  static typename AddReference<
-      const typename ::std::tr1::tuple_element<I, Tuple>::type>::type get(
-      const Tuple& tuple) {
-    return ::std::tr1::get<I>(tuple);
-  }
-};
-template <typename TupleT>
-const size_t TuplePolicy<TupleT>::tuple_size;
-#endif  // GTEST_HAS_TR1_TUPLE
-
-#if GTEST_HAS_STD_TUPLE_
-template <typename... Types>
-struct TuplePolicy< ::std::tuple<Types...> > {
-  typedef ::std::tuple<Types...> Tuple;
-  static const size_t tuple_size = ::std::tuple_size<Tuple>::value;
-
-  template <size_t I>
-  struct tuple_element : ::std::tuple_element<I, Tuple> {};
-
-  template <size_t I>
-  static const typename ::std::tuple_element<I, Tuple>::type& get(
-      const Tuple& tuple) {
-    return ::std::get<I>(tuple);
-  }
-};
-template <typename... Types>
-const size_t TuplePolicy< ::std::tuple<Types...> >::tuple_size;
-#endif  // GTEST_HAS_STD_TUPLE_
-
-#if GTEST_HAS_TR1_TUPLE || GTEST_HAS_STD_TUPLE_
-// This helper template allows PrintTo() for tuples and
-// UniversalTersePrintTupleFieldsToStrings() to be defined by
-// induction on the number of tuple fields.  The idea is that
-// TuplePrefixPrinter<N>::PrintPrefixTo(t, os) prints the first N
-// fields in tuple t, and can be defined in terms of
-// TuplePrefixPrinter<N - 1>.
-//
-// The inductive case.
-template <size_t N>
-struct TuplePrefixPrinter {
-  // Prints the first N fields of a tuple.
-  template <typename Tuple>
-  static void PrintPrefixTo(const Tuple& t, ::std::ostream* os) {
-    TuplePrefixPrinter<N - 1>::PrintPrefixTo(t, os);
-    GTEST_INTENTIONAL_CONST_COND_PUSH_()
-    if (N > 1) {
-    GTEST_INTENTIONAL_CONST_COND_POP_()
-      *os << ", ";
-    }
-    UniversalPrinter<
-        typename TuplePolicy<Tuple>::template tuple_element<N - 1>::type>
-        ::Print(TuplePolicy<Tuple>::template get<N - 1>(t), os);
-  }
-
-  // Tersely prints the first N fields of a tuple to a string vector,
-  // one element for each field.
-  template <typename Tuple>
-  static void TersePrintPrefixToStrings(const Tuple& t, Strings* strings) {
-    TuplePrefixPrinter<N - 1>::TersePrintPrefixToStrings(t, strings);
-    ::std::stringstream ss;
-    UniversalTersePrint(TuplePolicy<Tuple>::template get<N - 1>(t), &ss);
-    strings->push_back(ss.str());
-  }
-};
-
-// Base case.
-template <>
-struct TuplePrefixPrinter<0> {
-  template <typename Tuple>
-  static void PrintPrefixTo(const Tuple&, ::std::ostream*) {}
-
-  template <typename Tuple>
-  static void TersePrintPrefixToStrings(const Tuple&, Strings*) {}
-};
-
-// Helper function for printing a tuple.
-// Tuple must be either std::tr1::tuple or std::tuple type.
-template <typename Tuple>
-void PrintTupleTo(const Tuple& t, ::std::ostream* os) {
-  *os << "(";
-  TuplePrefixPrinter<TuplePolicy<Tuple>::tuple_size>::PrintPrefixTo(t, os);
-  *os << ")";
-}
-
-// Prints the fields of a tuple tersely to a string vector, one
-// element for each field.  See the comment before
-// UniversalTersePrint() for how we define "tersely".
-template <typename Tuple>
-Strings UniversalTersePrintTupleFieldsToStrings(const Tuple& value) {
-  Strings result;
-  TuplePrefixPrinter<TuplePolicy<Tuple>::tuple_size>::
-      TersePrintPrefixToStrings(value, &result);
-  return result;
-}
-#endif  // GTEST_HAS_TR1_TUPLE || GTEST_HAS_STD_TUPLE_
-
-}  // namespace internal
-
-#if GTEST_HAS_ABSL
-namespace internal2 {
-template <typename T>
-void TypeWithoutFormatter<T, kConvertibleToStringView>::PrintValue(
-    const T& value, ::std::ostream* os) {
-  internal::PrintTo(absl::string_view(value), os);
-}
-}  // namespace internal2
-#endif
-
-template <typename T>
-::std::string PrintToString(const T& value) {
-  ::std::stringstream ss;
-  internal::UniversalTersePrinter<T>::Print(value, &ss);
-  return ss.str();
-}
-
-}  // namespace testing
-
-// Include any custom printer added by the local installation.
-// We must include this header at the end to make sure it can use the
-// declarations from this file.
-// Copyright 2015, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// This file provides an injection point for custom printers in a local
-// installation of gTest.
-// It will be included from gtest-printers.h and the overrides in this file
-// will be visible to everyone.
-// See documentation at gtest/gtest-printers.h for details on how to define a
-// custom printer.
-//
-// ** Custom implementation starts here **
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_CUSTOM_GTEST_PRINTERS_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_CUSTOM_GTEST_PRINTERS_H_
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_CUSTOM_GTEST_PRINTERS_H_
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_PRINTERS_H_
-
-namespace testing {
-
-// Input to a parameterized test name generator, describing a test parameter.
-// Consists of the parameter value and the integer parameter index.
-template <class ParamType>
-struct TestParamInfo {
-  TestParamInfo(const ParamType& a_param, size_t an_index) :
-    param(a_param),
-    index(an_index) {}
-  ParamType param;
-  size_t index;
-};
-
-// A builtin parameterized test name generator which returns the result of
-// testing::PrintToString.
-struct PrintToStringParamName {
-  template <class ParamType>
-  std::string operator()(const TestParamInfo<ParamType>& info) const {
-    return PrintToString(info.param);
-  }
-};
-
-namespace internal {
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// Outputs a message explaining invalid registration of different
-// fixture class for the same test case. This may happen when
-// TEST_P macro is used to define two tests with the same name
-// but in different namespaces.
-GTEST_API_ void ReportInvalidTestCaseType(const char* test_case_name,
-                                          CodeLocation code_location);
-
-template <typename> class ParamGeneratorInterface;
-template <typename> class ParamGenerator;
-
-// Interface for iterating over elements provided by an implementation
-// of ParamGeneratorInterface<T>.
-template <typename T>
-class ParamIteratorInterface {
- public:
-  virtual ~ParamIteratorInterface() {}
-  // A pointer to the base generator instance.
-  // Used only for the purposes of iterator comparison
-  // to make sure that two iterators belong to the same generator.
-  virtual const ParamGeneratorInterface<T>* BaseGenerator() const = 0;
-  // Advances iterator to point to the next element
-  // provided by the generator. The caller is responsible
-  // for not calling Advance() on an iterator equal to
-  // BaseGenerator()->End().
-  virtual void Advance() = 0;
-  // Clones the iterator object. Used for implementing copy semantics
-  // of ParamIterator<T>.
-  virtual ParamIteratorInterface* Clone() const = 0;
-  // Dereferences the current iterator and provides (read-only) access
-  // to the pointed value. It is the caller's responsibility not to call
-  // Current() on an iterator equal to BaseGenerator()->End().
-  // Used for implementing ParamGenerator<T>::operator*().
-  virtual const T* Current() const = 0;
-  // Determines whether the given iterator and other point to the same
-  // element in the sequence generated by the generator.
-  // Used for implementing ParamGenerator<T>::operator==().
-  virtual bool Equals(const ParamIteratorInterface& other) const = 0;
-};
-
-// Class iterating over elements provided by an implementation of
-// ParamGeneratorInterface<T>. It wraps ParamIteratorInterface<T>
-// and implements the const forward iterator concept.
-template <typename T>
-class ParamIterator {
- public:
-  typedef T value_type;
-  typedef const T& reference;
-  typedef ptrdiff_t difference_type;
-
-  // ParamIterator assumes ownership of the impl_ pointer.
-  ParamIterator(const ParamIterator& other) : impl_(other.impl_->Clone()) {}
-  ParamIterator& operator=(const ParamIterator& other) {
-    if (this != &other)
-      impl_.reset(other.impl_->Clone());
-    return *this;
-  }
-
-  const T& operator*() const { return *impl_->Current(); }
-  const T* operator->() const { return impl_->Current(); }
-  // Prefix version of operator++.
-  ParamIterator& operator++() {
-    impl_->Advance();
-    return *this;
-  }
-  // Postfix version of operator++.
-  ParamIterator operator++(int /*unused*/) {
-    ParamIteratorInterface<T>* clone = impl_->Clone();
-    impl_->Advance();
-    return ParamIterator(clone);
-  }
-  bool operator==(const ParamIterator& other) const {
-    return impl_.get() == other.impl_.get() || impl_->Equals(*other.impl_);
-  }
-  bool operator!=(const ParamIterator& other) const {
-    return !(*this == other);
-  }
-
- private:
-  friend class ParamGenerator<T>;
-  explicit ParamIterator(ParamIteratorInterface<T>* impl) : impl_(impl) {}
-  scoped_ptr<ParamIteratorInterface<T> > impl_;
-};
-
-// ParamGeneratorInterface<T> is the binary interface to access generators
-// defined in other translation units.
-template <typename T>
-class ParamGeneratorInterface {
- public:
-  typedef T ParamType;
-
-  virtual ~ParamGeneratorInterface() {}
-
-  // Generator interface definition
-  virtual ParamIteratorInterface<T>* Begin() const = 0;
-  virtual ParamIteratorInterface<T>* End() const = 0;
-};
-
-// Wraps ParamGeneratorInterface<T> and provides general generator syntax
-// compatible with the STL Container concept.
-// This class implements copy initialization semantics and the contained
-// ParamGeneratorInterface<T> instance is shared among all copies
-// of the original object. This is possible because that instance is immutable.
-template<typename T>
-class ParamGenerator {
- public:
-  typedef ParamIterator<T> iterator;
-
-  explicit ParamGenerator(ParamGeneratorInterface<T>* impl) : impl_(impl) {}
-  ParamGenerator(const ParamGenerator& other) : impl_(other.impl_) {}
-
-  ParamGenerator& operator=(const ParamGenerator& other) {
-    impl_ = other.impl_;
-    return *this;
-  }
-
-  iterator begin() const { return iterator(impl_->Begin()); }
-  iterator end() const { return iterator(impl_->End()); }
-
- private:
-  linked_ptr<const ParamGeneratorInterface<T> > impl_;
-};
-
-// Generates values from a range of two comparable values. Can be used to
-// generate sequences of user-defined types that implement operator+() and
-// operator<().
-// This class is used in the Range() function.
-template <typename T, typename IncrementT>
-class RangeGenerator : public ParamGeneratorInterface<T> {
- public:
-  RangeGenerator(T begin, T end, IncrementT step)
-      : begin_(begin), end_(end),
-        step_(step), end_index_(CalculateEndIndex(begin, end, step)) {}
-  virtual ~RangeGenerator() {}
-
-  virtual ParamIteratorInterface<T>* Begin() const {
-    return new Iterator(this, begin_, 0, step_);
-  }
-  virtual ParamIteratorInterface<T>* End() const {
-    return new Iterator(this, end_, end_index_, step_);
-  }
-
- private:
-  class Iterator : public ParamIteratorInterface<T> {
-   public:
-    Iterator(const ParamGeneratorInterface<T>* base, T value, int index,
-             IncrementT step)
-        : base_(base), value_(value), index_(index), step_(step) {}
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<T>* BaseGenerator() const {
-      return base_;
-    }
-    virtual void Advance() {
-      value_ = static_cast<T>(value_ + step_);
-      index_++;
-    }
-    virtual ParamIteratorInterface<T>* Clone() const {
-      return new Iterator(*this);
-    }
-    virtual const T* Current() const { return &value_; }
-    virtual bool Equals(const ParamIteratorInterface<T>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      const int other_index =
-          CheckedDowncastToActualType<const Iterator>(&other)->index_;
-      return index_ == other_index;
-    }
-
-   private:
-    Iterator(const Iterator& other)
-        : ParamIteratorInterface<T>(),
-          base_(other.base_), value_(other.value_), index_(other.index_),
-          step_(other.step_) {}
-
-    // No implementation - assignment is unsupported.
-    void operator=(const Iterator& other);
-
-    const ParamGeneratorInterface<T>* const base_;
-    T value_;
-    int index_;
-    const IncrementT step_;
-  };  // class RangeGenerator::Iterator
-
-  static int CalculateEndIndex(const T& begin,
-                               const T& end,
-                               const IncrementT& step) {
-    int end_index = 0;
-    for (T i = begin; i < end; i = static_cast<T>(i + step))
-      end_index++;
-    return end_index;
-  }
-
-  // No implementation - assignment is unsupported.
-  void operator=(const RangeGenerator& other);
-
-  const T begin_;
-  const T end_;
-  const IncrementT step_;
-  // The index for the end() iterator. All the elements in the generated
-  // sequence are indexed (0-based) to aid iterator comparison.
-  const int end_index_;
-};  // class RangeGenerator
-
-
-// Generates values from a pair of STL-style iterators. Used in the
-// ValuesIn() function. The elements are copied from the source range
-// since the source can be located on the stack, and the generator
-// is likely to persist beyond that stack frame.
-template <typename T>
-class ValuesInIteratorRangeGenerator : public ParamGeneratorInterface<T> {
- public:
-  template <typename ForwardIterator>
-  ValuesInIteratorRangeGenerator(ForwardIterator begin, ForwardIterator end)
-      : container_(begin, end) {}
-  virtual ~ValuesInIteratorRangeGenerator() {}
-
-  virtual ParamIteratorInterface<T>* Begin() const {
-    return new Iterator(this, container_.begin());
-  }
-  virtual ParamIteratorInterface<T>* End() const {
-    return new Iterator(this, container_.end());
-  }
-
- private:
-  typedef typename ::std::vector<T> ContainerType;
-
-  class Iterator : public ParamIteratorInterface<T> {
-   public:
-    Iterator(const ParamGeneratorInterface<T>* base,
-             typename ContainerType::const_iterator iterator)
-        : base_(base), iterator_(iterator) {}
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<T>* BaseGenerator() const {
-      return base_;
-    }
-    virtual void Advance() {
-      ++iterator_;
-      value_.reset();
-    }
-    virtual ParamIteratorInterface<T>* Clone() const {
-      return new Iterator(*this);
-    }
-    // We need to use cached value referenced by iterator_ because *iterator_
-    // can return a temporary object (and of type other then T), so just
-    // having "return &*iterator_;" doesn't work.
-    // value_ is updated here and not in Advance() because Advance()
-    // can advance iterator_ beyond the end of the range, and we cannot
-    // detect that fact. The client code, on the other hand, is
-    // responsible for not calling Current() on an out-of-range iterator.
-    virtual const T* Current() const {
-      if (value_.get() == NULL)
-        value_.reset(new T(*iterator_));
-      return value_.get();
-    }
-    virtual bool Equals(const ParamIteratorInterface<T>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      return iterator_ ==
-          CheckedDowncastToActualType<const Iterator>(&other)->iterator_;
-    }
-
-   private:
-    Iterator(const Iterator& other)
-          // The explicit constructor call suppresses a false warning
-          // emitted by gcc when supplied with the -Wextra option.
-        : ParamIteratorInterface<T>(),
-          base_(other.base_),
-          iterator_(other.iterator_) {}
-
-    const ParamGeneratorInterface<T>* const base_;
-    typename ContainerType::const_iterator iterator_;
-    // A cached value of *iterator_. We keep it here to allow access by
-    // pointer in the wrapping iterator's operator->().
-    // value_ needs to be mutable to be accessed in Current().
-    // Use of scoped_ptr helps manage cached value's lifetime,
-    // which is bound by the lifespan of the iterator itself.
-    mutable scoped_ptr<const T> value_;
-  };  // class ValuesInIteratorRangeGenerator::Iterator
-
-  // No implementation - assignment is unsupported.
-  void operator=(const ValuesInIteratorRangeGenerator& other);
-
-  const ContainerType container_;
-};  // class ValuesInIteratorRangeGenerator
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// Default parameterized test name generator, returns a string containing the
-// integer test parameter index.
-template <class ParamType>
-std::string DefaultParamName(const TestParamInfo<ParamType>& info) {
-  Message name_stream;
-  name_stream << info.index;
-  return name_stream.GetString();
-}
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// Parameterized test name overload helpers, which help the
-// INSTANTIATE_TEST_CASE_P macro choose between the default parameterized
-// test name generator and user param name generator.
-template <class ParamType, class ParamNameGenFunctor>
-ParamNameGenFunctor GetParamNameGen(ParamNameGenFunctor func) {
-  return func;
-}
-
-template <class ParamType>
-struct ParamNameGenFunc {
-  typedef std::string Type(const TestParamInfo<ParamType>&);
-};
-
-template <class ParamType>
-typename ParamNameGenFunc<ParamType>::Type *GetParamNameGen() {
-  return DefaultParamName;
-}
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// Stores a parameter value and later creates tests parameterized with that
-// value.
-template <class TestClass>
-class ParameterizedTestFactory : public TestFactoryBase {
- public:
-  typedef typename TestClass::ParamType ParamType;
-  explicit ParameterizedTestFactory(ParamType parameter) :
-      parameter_(parameter) {}
-  virtual Test* CreateTest() {
-    TestClass::SetParam(&parameter_);
-    return new TestClass();
-  }
-
- private:
-  const ParamType parameter_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ParameterizedTestFactory);
-};
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// TestMetaFactoryBase is a base class for meta-factories that create
-// test factories for passing into MakeAndRegisterTestInfo function.
-template <class ParamType>
-class TestMetaFactoryBase {
- public:
-  virtual ~TestMetaFactoryBase() {}
-
-  virtual TestFactoryBase* CreateTestFactory(ParamType parameter) = 0;
-};
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// TestMetaFactory creates test factories for passing into
-// MakeAndRegisterTestInfo function. Since MakeAndRegisterTestInfo receives
-// ownership of test factory pointer, same factory object cannot be passed
-// into that method twice. But ParameterizedTestCaseInfo is going to call
-// it for each Test/Parameter value combination. Thus it needs meta factory
-// creator class.
-template <class TestCase>
-class TestMetaFactory
-    : public TestMetaFactoryBase<typename TestCase::ParamType> {
- public:
-  typedef typename TestCase::ParamType ParamType;
-
-  TestMetaFactory() {}
-
-  virtual TestFactoryBase* CreateTestFactory(ParamType parameter) {
-    return new ParameterizedTestFactory<TestCase>(parameter);
-  }
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(TestMetaFactory);
-};
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// ParameterizedTestCaseInfoBase is a generic interface
-// to ParameterizedTestCaseInfo classes. ParameterizedTestCaseInfoBase
-// accumulates test information provided by TEST_P macro invocations
-// and generators provided by INSTANTIATE_TEST_CASE_P macro invocations
-// and uses that information to register all resulting test instances
-// in RegisterTests method. The ParameterizeTestCaseRegistry class holds
-// a collection of pointers to the ParameterizedTestCaseInfo objects
-// and calls RegisterTests() on each of them when asked.
-class ParameterizedTestCaseInfoBase {
- public:
-  virtual ~ParameterizedTestCaseInfoBase() {}
-
-  // Base part of test case name for display purposes.
-  virtual const std::string& GetTestCaseName() const = 0;
-  // Test case id to verify identity.
-  virtual TypeId GetTestCaseTypeId() const = 0;
-  // UnitTest class invokes this method to register tests in this
-  // test case right before running them in RUN_ALL_TESTS macro.
-  // This method should not be called more then once on any single
-  // instance of a ParameterizedTestCaseInfoBase derived class.
-  virtual void RegisterTests() = 0;
-
- protected:
-  ParameterizedTestCaseInfoBase() {}
-
- private:
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ParameterizedTestCaseInfoBase);
-};
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// ParameterizedTestCaseInfo accumulates tests obtained from TEST_P
-// macro invocations for a particular test case and generators
-// obtained from INSTANTIATE_TEST_CASE_P macro invocations for that
-// test case. It registers tests with all values generated by all
-// generators when asked.
-template <class TestCase>
-class ParameterizedTestCaseInfo : public ParameterizedTestCaseInfoBase {
- public:
-  // ParamType and GeneratorCreationFunc are private types but are required
-  // for declarations of public methods AddTestPattern() and
-  // AddTestCaseInstantiation().
-  typedef typename TestCase::ParamType ParamType;
-  // A function that returns an instance of appropriate generator type.
-  typedef ParamGenerator<ParamType>(GeneratorCreationFunc)();
-  typedef typename ParamNameGenFunc<ParamType>::Type ParamNameGeneratorFunc;
-
-  explicit ParameterizedTestCaseInfo(
-      const char* name, CodeLocation code_location)
-      : test_case_name_(name), code_location_(code_location) {}
-
-  // Test case base name for display purposes.
-  virtual const std::string& GetTestCaseName() const { return test_case_name_; }
-  // Test case id to verify identity.
-  virtual TypeId GetTestCaseTypeId() const { return GetTypeId<TestCase>(); }
-  // TEST_P macro uses AddTestPattern() to record information
-  // about a single test in a LocalTestInfo structure.
-  // test_case_name is the base name of the test case (without invocation
-  // prefix). test_base_name is the name of an individual test without
-  // parameter index. For the test SequenceA/FooTest.DoBar/1 FooTest is
-  // test case base name and DoBar is test base name.
-  void AddTestPattern(const char* test_case_name,
-                      const char* test_base_name,
-                      TestMetaFactoryBase<ParamType>* meta_factory) {
-    tests_.push_back(linked_ptr<TestInfo>(new TestInfo(test_case_name,
-                                                       test_base_name,
-                                                       meta_factory)));
-  }
-  // INSTANTIATE_TEST_CASE_P macro uses AddGenerator() to record information
-  // about a generator.
-  int AddTestCaseInstantiation(const std::string& instantiation_name,
-                               GeneratorCreationFunc* func,
-                               ParamNameGeneratorFunc* name_func,
-                               const char* file, int line) {
-    instantiations_.push_back(
-        InstantiationInfo(instantiation_name, func, name_func, file, line));
-    return 0;  // Return value used only to run this method in namespace scope.
-  }
-  // UnitTest class invokes this method to register tests in this test case
-  // test cases right before running tests in RUN_ALL_TESTS macro.
-  // This method should not be called more then once on any single
-  // instance of a ParameterizedTestCaseInfoBase derived class.
-  // UnitTest has a guard to prevent from calling this method more then once.
-  virtual void RegisterTests() {
-    for (typename TestInfoContainer::iterator test_it = tests_.begin();
-         test_it != tests_.end(); ++test_it) {
-      linked_ptr<TestInfo> test_info = *test_it;
-      for (typename InstantiationContainer::iterator gen_it =
-               instantiations_.begin(); gen_it != instantiations_.end();
-               ++gen_it) {
-        const std::string& instantiation_name = gen_it->name;
-        ParamGenerator<ParamType> generator((*gen_it->generator)());
-        ParamNameGeneratorFunc* name_func = gen_it->name_func;
-        const char* file = gen_it->file;
-        int line = gen_it->line;
-
-        std::string test_case_name;
-        if ( !instantiation_name.empty() )
-          test_case_name = instantiation_name + "/";
-        test_case_name += test_info->test_case_base_name;
-
-        size_t i = 0;
-        std::set<std::string> test_param_names;
-        for (typename ParamGenerator<ParamType>::iterator param_it =
-                 generator.begin();
-             param_it != generator.end(); ++param_it, ++i) {
-          Message test_name_stream;
-
-          std::string param_name = name_func(
-              TestParamInfo<ParamType>(*param_it, i));
-
-          GTEST_CHECK_(IsValidParamName(param_name))
-              << "Parameterized test name '" << param_name
-              << "' is invalid, in " << file
-              << " line " << line << std::endl;
-
-          GTEST_CHECK_(test_param_names.count(param_name) == 0)
-              << "Duplicate parameterized test name '" << param_name
-              << "', in " << file << " line " << line << std::endl;
-
-          test_param_names.insert(param_name);
-
-          test_name_stream << test_info->test_base_name << "/" << param_name;
-          MakeAndRegisterTestInfo(
-              test_case_name.c_str(),
-              test_name_stream.GetString().c_str(),
-              NULL,  // No type parameter.
-              PrintToString(*param_it).c_str(),
-              code_location_,
-              GetTestCaseTypeId(),
-              TestCase::SetUpTestCase,
-              TestCase::TearDownTestCase,
-              test_info->test_meta_factory->CreateTestFactory(*param_it));
-        }  // for param_it
-      }  // for gen_it
-    }  // for test_it
-  }  // RegisterTests
-
- private:
-  // LocalTestInfo structure keeps information about a single test registered
-  // with TEST_P macro.
-  struct TestInfo {
-    TestInfo(const char* a_test_case_base_name,
-             const char* a_test_base_name,
-             TestMetaFactoryBase<ParamType>* a_test_meta_factory) :
-        test_case_base_name(a_test_case_base_name),
-        test_base_name(a_test_base_name),
-        test_meta_factory(a_test_meta_factory) {}
-
-    const std::string test_case_base_name;
-    const std::string test_base_name;
-    const scoped_ptr<TestMetaFactoryBase<ParamType> > test_meta_factory;
-  };
-  typedef ::std::vector<linked_ptr<TestInfo> > TestInfoContainer;
-  // Records data received from INSTANTIATE_TEST_CASE_P macros:
-  //  <Instantiation name, Sequence generator creation function,
-  //     Name generator function, Source file, Source line>
-  struct InstantiationInfo {
-      InstantiationInfo(const std::string &name_in,
-                        GeneratorCreationFunc* generator_in,
-                        ParamNameGeneratorFunc* name_func_in,
-                        const char* file_in,
-                        int line_in)
-          : name(name_in),
-            generator(generator_in),
-            name_func(name_func_in),
-            file(file_in),
-            line(line_in) {}
-
-      std::string name;
-      GeneratorCreationFunc* generator;
-      ParamNameGeneratorFunc* name_func;
-      const char* file;
-      int line;
-  };
-  typedef ::std::vector<InstantiationInfo> InstantiationContainer;
-
-  static bool IsValidParamName(const std::string& name) {
-    // Check for empty string
-    if (name.empty())
-      return false;
-
-    // Check for invalid characters
-    for (std::string::size_type index = 0; index < name.size(); ++index) {
-      if (!isalnum(name[index]) && name[index] != '_')
-        return false;
-    }
-
-    return true;
-  }
-
-  const std::string test_case_name_;
-  CodeLocation code_location_;
-  TestInfoContainer tests_;
-  InstantiationContainer instantiations_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ParameterizedTestCaseInfo);
-};  // class ParameterizedTestCaseInfo
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// ParameterizedTestCaseRegistry contains a map of ParameterizedTestCaseInfoBase
-// classes accessed by test case names. TEST_P and INSTANTIATE_TEST_CASE_P
-// macros use it to locate their corresponding ParameterizedTestCaseInfo
-// descriptors.
-class ParameterizedTestCaseRegistry {
- public:
-  ParameterizedTestCaseRegistry() {}
-  ~ParameterizedTestCaseRegistry() {
-    for (TestCaseInfoContainer::iterator it = test_case_infos_.begin();
-         it != test_case_infos_.end(); ++it) {
-      delete *it;
-    }
-  }
-
-  // Looks up or creates and returns a structure containing information about
-  // tests and instantiations of a particular test case.
-  template <class TestCase>
-  ParameterizedTestCaseInfo<TestCase>* GetTestCasePatternHolder(
-      const char* test_case_name,
-      CodeLocation code_location) {
-    ParameterizedTestCaseInfo<TestCase>* typed_test_info = NULL;
-    for (TestCaseInfoContainer::iterator it = test_case_infos_.begin();
-         it != test_case_infos_.end(); ++it) {
-      if ((*it)->GetTestCaseName() == test_case_name) {
-        if ((*it)->GetTestCaseTypeId() != GetTypeId<TestCase>()) {
-          // Complain about incorrect usage of Google Test facilities
-          // and terminate the program since we cannot guaranty correct
-          // test case setup and tear-down in this case.
-          ReportInvalidTestCaseType(test_case_name, code_location);
-          posix::Abort();
-        } else {
-          // At this point we are sure that the object we found is of the same
-          // type we are looking for, so we downcast it to that type
-          // without further checks.
-          typed_test_info = CheckedDowncastToActualType<
-              ParameterizedTestCaseInfo<TestCase> >(*it);
-        }
-        break;
-      }
-    }
-    if (typed_test_info == NULL) {
-      typed_test_info = new ParameterizedTestCaseInfo<TestCase>(
-          test_case_name, code_location);
-      test_case_infos_.push_back(typed_test_info);
-    }
-    return typed_test_info;
-  }
-  void RegisterTests() {
-    for (TestCaseInfoContainer::iterator it = test_case_infos_.begin();
-         it != test_case_infos_.end(); ++it) {
-      (*it)->RegisterTests();
-    }
-  }
-
- private:
-  typedef ::std::vector<ParameterizedTestCaseInfoBase*> TestCaseInfoContainer;
-
-  TestCaseInfoContainer test_case_infos_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ParameterizedTestCaseRegistry);
-};
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PARAM_UTIL_H_
-// This file was GENERATED by command:
-//     pump.py gtest-param-util-generated.h.pump
-// DO NOT EDIT BY HAND!!!
-
-// Copyright 2008 Google Inc.
-// All Rights Reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: vladl@google.com (Vlad Losev)
-
-// Type and function utilities for implementing parameterized tests.
-// This file is generated by a SCRIPT.  DO NOT EDIT BY HAND!
-//
-// Currently Google Test supports at most 50 arguments in Values,
-// and at most 10 arguments in Combine. Please contact
-// googletestframework@googlegroups.com if you need more.
-// Please note that the number of arguments to Combine is limited
-// by the maximum arity of the implementation of tuple which is
-// currently set at 10.
-
-#ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PARAM_UTIL_GENERATED_H_
-#define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PARAM_UTIL_GENERATED_H_
-
-
-namespace testing {
-
-// Forward declarations of ValuesIn(), which is implemented in
-// include/gtest/gtest-param-test.h.
-template <typename ForwardIterator>
-internal::ParamGenerator<
-  typename ::testing::internal::IteratorTraits<ForwardIterator>::value_type>
-ValuesIn(ForwardIterator begin, ForwardIterator end);
-
-template <typename T, size_t N>
-internal::ParamGenerator<T> ValuesIn(const T (&array)[N]);
-
-template <class Container>
-internal::ParamGenerator<typename Container::value_type> ValuesIn(
-    const Container& container);
-
-namespace internal {
-
-// Used in the Values() function to provide polymorphic capabilities.
-template <typename T1>
-class ValueArray1 {
- public:
-  explicit ValueArray1(T1 v1) : v1_(v1) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray1& other);
-
-  const T1 v1_;
-};
-
-template <typename T1, typename T2>
-class ValueArray2 {
- public:
-  ValueArray2(T1 v1, T2 v2) : v1_(v1), v2_(v2) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray2& other);
-
-  const T1 v1_;
-  const T2 v2_;
-};
-
-template <typename T1, typename T2, typename T3>
-class ValueArray3 {
- public:
-  ValueArray3(T1 v1, T2 v2, T3 v3) : v1_(v1), v2_(v2), v3_(v3) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray3& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4>
-class ValueArray4 {
- public:
-  ValueArray4(T1 v1, T2 v2, T3 v3, T4 v4) : v1_(v1), v2_(v2), v3_(v3),
-      v4_(v4) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray4& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5>
-class ValueArray5 {
- public:
-  ValueArray5(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5) : v1_(v1), v2_(v2), v3_(v3),
-      v4_(v4), v5_(v5) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray5& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6>
-class ValueArray6 {
- public:
-  ValueArray6(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6) : v1_(v1), v2_(v2),
-      v3_(v3), v4_(v4), v5_(v5), v6_(v6) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray6& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7>
-class ValueArray7 {
- public:
-  ValueArray7(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7) : v1_(v1),
-      v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray7& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8>
-class ValueArray8 {
- public:
-  ValueArray8(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7,
-      T8 v8) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray8& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9>
-class ValueArray9 {
- public:
-  ValueArray9(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8,
-      T9 v9) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray9& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10>
-class ValueArray10 {
- public:
-  ValueArray10(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray10& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11>
-class ValueArray11 {
- public:
-  ValueArray11(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6),
-      v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray11& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12>
-class ValueArray12 {
- public:
-  ValueArray12(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5),
-      v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray12& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13>
-class ValueArray13 {
- public:
-  ValueArray13(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13) : v1_(v1), v2_(v2), v3_(v3), v4_(v4),
-      v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11),
-      v12_(v12), v13_(v13) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray13& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14>
-class ValueArray14 {
- public:
-  ValueArray14(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14) : v1_(v1), v2_(v2), v3_(v3),
-      v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10),
-      v11_(v11), v12_(v12), v13_(v13), v14_(v14) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray14& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15>
-class ValueArray15 {
- public:
-  ValueArray15(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15) : v1_(v1), v2_(v2),
-      v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10),
-      v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray15& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16>
-class ValueArray16 {
- public:
-  ValueArray16(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16) : v1_(v1),
-      v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9),
-      v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15),
-      v16_(v16) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray16& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17>
-class ValueArray17 {
- public:
-  ValueArray17(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16,
-      T17 v17) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14),
-      v15_(v15), v16_(v16), v17_(v17) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray17& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18>
-class ValueArray18 {
- public:
-  ValueArray18(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14),
-      v15_(v15), v16_(v16), v17_(v17), v18_(v18) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray18& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19>
-class ValueArray19 {
- public:
-  ValueArray19(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6),
-      v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13),
-      v14_(v14), v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray19& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20>
-class ValueArray20 {
- public:
-  ValueArray20(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5),
-      v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12),
-      v13_(v13), v14_(v14), v15_(v15), v16_(v16), v17_(v17), v18_(v18),
-      v19_(v19), v20_(v20) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray20& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21>
-class ValueArray21 {
- public:
-  ValueArray21(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21) : v1_(v1), v2_(v2), v3_(v3), v4_(v4),
-      v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11),
-      v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16), v17_(v17),
-      v18_(v18), v19_(v19), v20_(v20), v21_(v21) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray21& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22>
-class ValueArray22 {
- public:
-  ValueArray22(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22) : v1_(v1), v2_(v2), v3_(v3),
-      v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10),
-      v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16),
-      v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray22& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23>
-class ValueArray23 {
- public:
-  ValueArray23(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23) : v1_(v1), v2_(v2),
-      v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10),
-      v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16),
-      v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22),
-      v23_(v23) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray23& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24>
-class ValueArray24 {
- public:
-  ValueArray24(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24) : v1_(v1),
-      v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9),
-      v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15),
-      v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21),
-      v22_(v22), v23_(v23), v24_(v24) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray24& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25>
-class ValueArray25 {
- public:
-  ValueArray25(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24,
-      T25 v25) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14),
-      v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20),
-      v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray25& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26>
-class ValueArray26 {
- public:
-  ValueArray26(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14),
-      v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20),
-      v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25), v26_(v26) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray26& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27>
-class ValueArray27 {
- public:
-  ValueArray27(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6),
-      v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13),
-      v14_(v14), v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19),
-      v20_(v20), v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25),
-      v26_(v26), v27_(v27) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray27& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28>
-class ValueArray28 {
- public:
-  ValueArray28(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5),
-      v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12),
-      v13_(v13), v14_(v14), v15_(v15), v16_(v16), v17_(v17), v18_(v18),
-      v19_(v19), v20_(v20), v21_(v21), v22_(v22), v23_(v23), v24_(v24),
-      v25_(v25), v26_(v26), v27_(v27), v28_(v28) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray28& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29>
-class ValueArray29 {
- public:
-  ValueArray29(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29) : v1_(v1), v2_(v2), v3_(v3), v4_(v4),
-      v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11),
-      v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16), v17_(v17),
-      v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22), v23_(v23),
-      v24_(v24), v25_(v25), v26_(v26), v27_(v27), v28_(v28), v29_(v29) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray29& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30>
-class ValueArray30 {
- public:
-  ValueArray30(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30) : v1_(v1), v2_(v2), v3_(v3),
-      v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10),
-      v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16),
-      v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22),
-      v23_(v23), v24_(v24), v25_(v25), v26_(v26), v27_(v27), v28_(v28),
-      v29_(v29), v30_(v30) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray30& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31>
-class ValueArray31 {
- public:
-  ValueArray31(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31) : v1_(v1), v2_(v2),
-      v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10),
-      v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16),
-      v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22),
-      v23_(v23), v24_(v24), v25_(v25), v26_(v26), v27_(v27), v28_(v28),
-      v29_(v29), v30_(v30), v31_(v31) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray31& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32>
-class ValueArray32 {
- public:
-  ValueArray32(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32) : v1_(v1),
-      v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9),
-      v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15),
-      v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21),
-      v22_(v22), v23_(v23), v24_(v24), v25_(v25), v26_(v26), v27_(v27),
-      v28_(v28), v29_(v29), v30_(v30), v31_(v31), v32_(v32) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray32& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33>
-class ValueArray33 {
- public:
-  ValueArray33(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32,
-      T33 v33) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14),
-      v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20),
-      v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25), v26_(v26),
-      v27_(v27), v28_(v28), v29_(v29), v30_(v30), v31_(v31), v32_(v32),
-      v33_(v33) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray33& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34>
-class ValueArray34 {
- public:
-  ValueArray34(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14),
-      v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20),
-      v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25), v26_(v26),
-      v27_(v27), v28_(v28), v29_(v29), v30_(v30), v31_(v31), v32_(v32),
-      v33_(v33), v34_(v34) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray34& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35>
-class ValueArray35 {
- public:
-  ValueArray35(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6),
-      v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13),
-      v14_(v14), v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19),
-      v20_(v20), v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25),
-      v26_(v26), v27_(v27), v28_(v28), v29_(v29), v30_(v30), v31_(v31),
-      v32_(v32), v33_(v33), v34_(v34), v35_(v35) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray35& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36>
-class ValueArray36 {
- public:
-  ValueArray36(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5),
-      v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12),
-      v13_(v13), v14_(v14), v15_(v15), v16_(v16), v17_(v17), v18_(v18),
-      v19_(v19), v20_(v20), v21_(v21), v22_(v22), v23_(v23), v24_(v24),
-      v25_(v25), v26_(v26), v27_(v27), v28_(v28), v29_(v29), v30_(v30),
-      v31_(v31), v32_(v32), v33_(v33), v34_(v34), v35_(v35), v36_(v36) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray36& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37>
-class ValueArray37 {
- public:
-  ValueArray37(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37) : v1_(v1), v2_(v2), v3_(v3), v4_(v4),
-      v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11),
-      v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16), v17_(v17),
-      v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22), v23_(v23),
-      v24_(v24), v25_(v25), v26_(v26), v27_(v27), v28_(v28), v29_(v29),
-      v30_(v30), v31_(v31), v32_(v32), v33_(v33), v34_(v34), v35_(v35),
-      v36_(v36), v37_(v37) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray37& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38>
-class ValueArray38 {
- public:
-  ValueArray38(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38) : v1_(v1), v2_(v2), v3_(v3),
-      v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10),
-      v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16),
-      v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22),
-      v23_(v23), v24_(v24), v25_(v25), v26_(v26), v27_(v27), v28_(v28),
-      v29_(v29), v30_(v30), v31_(v31), v32_(v32), v33_(v33), v34_(v34),
-      v35_(v35), v36_(v36), v37_(v37), v38_(v38) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray38& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39>
-class ValueArray39 {
- public:
-  ValueArray39(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39) : v1_(v1), v2_(v2),
-      v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10),
-      v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16),
-      v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22),
-      v23_(v23), v24_(v24), v25_(v25), v26_(v26), v27_(v27), v28_(v28),
-      v29_(v29), v30_(v30), v31_(v31), v32_(v32), v33_(v33), v34_(v34),
-      v35_(v35), v36_(v36), v37_(v37), v38_(v38), v39_(v39) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray39& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40>
-class ValueArray40 {
- public:
-  ValueArray40(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40) : v1_(v1),
-      v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9),
-      v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15),
-      v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21),
-      v22_(v22), v23_(v23), v24_(v24), v25_(v25), v26_(v26), v27_(v27),
-      v28_(v28), v29_(v29), v30_(v30), v31_(v31), v32_(v32), v33_(v33),
-      v34_(v34), v35_(v35), v36_(v36), v37_(v37), v38_(v38), v39_(v39),
-      v40_(v40) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray40& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41>
-class ValueArray41 {
- public:
-  ValueArray41(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40,
-      T41 v41) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14),
-      v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20),
-      v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25), v26_(v26),
-      v27_(v27), v28_(v28), v29_(v29), v30_(v30), v31_(v31), v32_(v32),
-      v33_(v33), v34_(v34), v35_(v35), v36_(v36), v37_(v37), v38_(v38),
-      v39_(v39), v40_(v40), v41_(v41) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_), static_cast<T>(v41_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray41& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-  const T41 v41_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42>
-class ValueArray42 {
- public:
-  ValueArray42(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-      T42 v42) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14),
-      v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20),
-      v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25), v26_(v26),
-      v27_(v27), v28_(v28), v29_(v29), v30_(v30), v31_(v31), v32_(v32),
-      v33_(v33), v34_(v34), v35_(v35), v36_(v36), v37_(v37), v38_(v38),
-      v39_(v39), v40_(v40), v41_(v41), v42_(v42) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_), static_cast<T>(v41_),
-        static_cast<T>(v42_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray42& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-  const T41 v41_;
-  const T42 v42_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43>
-class ValueArray43 {
- public:
-  ValueArray43(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-      T42 v42, T43 v43) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6),
-      v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13),
-      v14_(v14), v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19),
-      v20_(v20), v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25),
-      v26_(v26), v27_(v27), v28_(v28), v29_(v29), v30_(v30), v31_(v31),
-      v32_(v32), v33_(v33), v34_(v34), v35_(v35), v36_(v36), v37_(v37),
-      v38_(v38), v39_(v39), v40_(v40), v41_(v41), v42_(v42), v43_(v43) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_), static_cast<T>(v41_),
-        static_cast<T>(v42_), static_cast<T>(v43_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray43& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-  const T41 v41_;
-  const T42 v42_;
-  const T43 v43_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44>
-class ValueArray44 {
- public:
-  ValueArray44(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-      T42 v42, T43 v43, T44 v44) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5),
-      v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12),
-      v13_(v13), v14_(v14), v15_(v15), v16_(v16), v17_(v17), v18_(v18),
-      v19_(v19), v20_(v20), v21_(v21), v22_(v22), v23_(v23), v24_(v24),
-      v25_(v25), v26_(v26), v27_(v27), v28_(v28), v29_(v29), v30_(v30),
-      v31_(v31), v32_(v32), v33_(v33), v34_(v34), v35_(v35), v36_(v36),
-      v37_(v37), v38_(v38), v39_(v39), v40_(v40), v41_(v41), v42_(v42),
-      v43_(v43), v44_(v44) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_), static_cast<T>(v41_),
-        static_cast<T>(v42_), static_cast<T>(v43_), static_cast<T>(v44_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray44& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-  const T41 v41_;
-  const T42 v42_;
-  const T43 v43_;
-  const T44 v44_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45>
-class ValueArray45 {
- public:
-  ValueArray45(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-      T42 v42, T43 v43, T44 v44, T45 v45) : v1_(v1), v2_(v2), v3_(v3), v4_(v4),
-      v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10), v11_(v11),
-      v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16), v17_(v17),
-      v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22), v23_(v23),
-      v24_(v24), v25_(v25), v26_(v26), v27_(v27), v28_(v28), v29_(v29),
-      v30_(v30), v31_(v31), v32_(v32), v33_(v33), v34_(v34), v35_(v35),
-      v36_(v36), v37_(v37), v38_(v38), v39_(v39), v40_(v40), v41_(v41),
-      v42_(v42), v43_(v43), v44_(v44), v45_(v45) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_), static_cast<T>(v41_),
-        static_cast<T>(v42_), static_cast<T>(v43_), static_cast<T>(v44_),
-        static_cast<T>(v45_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray45& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-  const T41 v41_;
-  const T42 v42_;
-  const T43 v43_;
-  const T44 v44_;
-  const T45 v45_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46>
-class ValueArray46 {
- public:
-  ValueArray46(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-      T42 v42, T43 v43, T44 v44, T45 v45, T46 v46) : v1_(v1), v2_(v2), v3_(v3),
-      v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10),
-      v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16),
-      v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22),
-      v23_(v23), v24_(v24), v25_(v25), v26_(v26), v27_(v27), v28_(v28),
-      v29_(v29), v30_(v30), v31_(v31), v32_(v32), v33_(v33), v34_(v34),
-      v35_(v35), v36_(v36), v37_(v37), v38_(v38), v39_(v39), v40_(v40),
-      v41_(v41), v42_(v42), v43_(v43), v44_(v44), v45_(v45), v46_(v46) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_), static_cast<T>(v41_),
-        static_cast<T>(v42_), static_cast<T>(v43_), static_cast<T>(v44_),
-        static_cast<T>(v45_), static_cast<T>(v46_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray46& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-  const T41 v41_;
-  const T42 v42_;
-  const T43 v43_;
-  const T44 v44_;
-  const T45 v45_;
-  const T46 v46_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47>
-class ValueArray47 {
- public:
-  ValueArray47(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-      T42 v42, T43 v43, T44 v44, T45 v45, T46 v46, T47 v47) : v1_(v1), v2_(v2),
-      v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9), v10_(v10),
-      v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15), v16_(v16),
-      v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21), v22_(v22),
-      v23_(v23), v24_(v24), v25_(v25), v26_(v26), v27_(v27), v28_(v28),
-      v29_(v29), v30_(v30), v31_(v31), v32_(v32), v33_(v33), v34_(v34),
-      v35_(v35), v36_(v36), v37_(v37), v38_(v38), v39_(v39), v40_(v40),
-      v41_(v41), v42_(v42), v43_(v43), v44_(v44), v45_(v45), v46_(v46),
-      v47_(v47) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_), static_cast<T>(v41_),
-        static_cast<T>(v42_), static_cast<T>(v43_), static_cast<T>(v44_),
-        static_cast<T>(v45_), static_cast<T>(v46_), static_cast<T>(v47_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray47& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-  const T41 v41_;
-  const T42 v42_;
-  const T43 v43_;
-  const T44 v44_;
-  const T45 v45_;
-  const T46 v46_;
-  const T47 v47_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48>
-class ValueArray48 {
- public:
-  ValueArray48(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-      T42 v42, T43 v43, T44 v44, T45 v45, T46 v46, T47 v47, T48 v48) : v1_(v1),
-      v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7), v8_(v8), v9_(v9),
-      v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14), v15_(v15),
-      v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20), v21_(v21),
-      v22_(v22), v23_(v23), v24_(v24), v25_(v25), v26_(v26), v27_(v27),
-      v28_(v28), v29_(v29), v30_(v30), v31_(v31), v32_(v32), v33_(v33),
-      v34_(v34), v35_(v35), v36_(v36), v37_(v37), v38_(v38), v39_(v39),
-      v40_(v40), v41_(v41), v42_(v42), v43_(v43), v44_(v44), v45_(v45),
-      v46_(v46), v47_(v47), v48_(v48) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_), static_cast<T>(v41_),
-        static_cast<T>(v42_), static_cast<T>(v43_), static_cast<T>(v44_),
-        static_cast<T>(v45_), static_cast<T>(v46_), static_cast<T>(v47_),
-        static_cast<T>(v48_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray48& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-  const T41 v41_;
-  const T42 v42_;
-  const T43 v43_;
-  const T44 v44_;
-  const T45 v45_;
-  const T46 v46_;
-  const T47 v47_;
-  const T48 v48_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48, typename T49>
-class ValueArray49 {
- public:
-  ValueArray49(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-      T42 v42, T43 v43, T44 v44, T45 v45, T46 v46, T47 v47, T48 v48,
-      T49 v49) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14),
-      v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20),
-      v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25), v26_(v26),
-      v27_(v27), v28_(v28), v29_(v29), v30_(v30), v31_(v31), v32_(v32),
-      v33_(v33), v34_(v34), v35_(v35), v36_(v36), v37_(v37), v38_(v38),
-      v39_(v39), v40_(v40), v41_(v41), v42_(v42), v43_(v43), v44_(v44),
-      v45_(v45), v46_(v46), v47_(v47), v48_(v48), v49_(v49) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_), static_cast<T>(v41_),
-        static_cast<T>(v42_), static_cast<T>(v43_), static_cast<T>(v44_),
-        static_cast<T>(v45_), static_cast<T>(v46_), static_cast<T>(v47_),
-        static_cast<T>(v48_), static_cast<T>(v49_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray49& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-  const T41 v41_;
-  const T42 v42_;
-  const T43 v43_;
-  const T44 v44_;
-  const T45 v45_;
-  const T46 v46_;
-  const T47 v47_;
-  const T48 v48_;
-  const T49 v49_;
-};
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48, typename T49, typename T50>
-class ValueArray50 {
- public:
-  ValueArray50(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-      T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-      T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-      T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-      T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-      T42 v42, T43 v43, T44 v44, T45 v45, T46 v46, T47 v47, T48 v48, T49 v49,
-      T50 v50) : v1_(v1), v2_(v2), v3_(v3), v4_(v4), v5_(v5), v6_(v6), v7_(v7),
-      v8_(v8), v9_(v9), v10_(v10), v11_(v11), v12_(v12), v13_(v13), v14_(v14),
-      v15_(v15), v16_(v16), v17_(v17), v18_(v18), v19_(v19), v20_(v20),
-      v21_(v21), v22_(v22), v23_(v23), v24_(v24), v25_(v25), v26_(v26),
-      v27_(v27), v28_(v28), v29_(v29), v30_(v30), v31_(v31), v32_(v32),
-      v33_(v33), v34_(v34), v35_(v35), v36_(v36), v37_(v37), v38_(v38),
-      v39_(v39), v40_(v40), v41_(v41), v42_(v42), v43_(v43), v44_(v44),
-      v45_(v45), v46_(v46), v47_(v47), v48_(v48), v49_(v49), v50_(v50) {}
-
-  template <typename T>
-  operator ParamGenerator<T>() const {
-    const T array[] = {static_cast<T>(v1_), static_cast<T>(v2_),
-        static_cast<T>(v3_), static_cast<T>(v4_), static_cast<T>(v5_),
-        static_cast<T>(v6_), static_cast<T>(v7_), static_cast<T>(v8_),
-        static_cast<T>(v9_), static_cast<T>(v10_), static_cast<T>(v11_),
-        static_cast<T>(v12_), static_cast<T>(v13_), static_cast<T>(v14_),
-        static_cast<T>(v15_), static_cast<T>(v16_), static_cast<T>(v17_),
-        static_cast<T>(v18_), static_cast<T>(v19_), static_cast<T>(v20_),
-        static_cast<T>(v21_), static_cast<T>(v22_), static_cast<T>(v23_),
-        static_cast<T>(v24_), static_cast<T>(v25_), static_cast<T>(v26_),
-        static_cast<T>(v27_), static_cast<T>(v28_), static_cast<T>(v29_),
-        static_cast<T>(v30_), static_cast<T>(v31_), static_cast<T>(v32_),
-        static_cast<T>(v33_), static_cast<T>(v34_), static_cast<T>(v35_),
-        static_cast<T>(v36_), static_cast<T>(v37_), static_cast<T>(v38_),
-        static_cast<T>(v39_), static_cast<T>(v40_), static_cast<T>(v41_),
-        static_cast<T>(v42_), static_cast<T>(v43_), static_cast<T>(v44_),
-        static_cast<T>(v45_), static_cast<T>(v46_), static_cast<T>(v47_),
-        static_cast<T>(v48_), static_cast<T>(v49_), static_cast<T>(v50_)};
-    return ValuesIn(array);
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const ValueArray50& other);
-
-  const T1 v1_;
-  const T2 v2_;
-  const T3 v3_;
-  const T4 v4_;
-  const T5 v5_;
-  const T6 v6_;
-  const T7 v7_;
-  const T8 v8_;
-  const T9 v9_;
-  const T10 v10_;
-  const T11 v11_;
-  const T12 v12_;
-  const T13 v13_;
-  const T14 v14_;
-  const T15 v15_;
-  const T16 v16_;
-  const T17 v17_;
-  const T18 v18_;
-  const T19 v19_;
-  const T20 v20_;
-  const T21 v21_;
-  const T22 v22_;
-  const T23 v23_;
-  const T24 v24_;
-  const T25 v25_;
-  const T26 v26_;
-  const T27 v27_;
-  const T28 v28_;
-  const T29 v29_;
-  const T30 v30_;
-  const T31 v31_;
-  const T32 v32_;
-  const T33 v33_;
-  const T34 v34_;
-  const T35 v35_;
-  const T36 v36_;
-  const T37 v37_;
-  const T38 v38_;
-  const T39 v39_;
-  const T40 v40_;
-  const T41 v41_;
-  const T42 v42_;
-  const T43 v43_;
-  const T44 v44_;
-  const T45 v45_;
-  const T46 v46_;
-  const T47 v47_;
-  const T48 v48_;
-  const T49 v49_;
-  const T50 v50_;
-};
-
-# if GTEST_HAS_COMBINE
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// Generates values from the Cartesian product of values produced
-// by the argument generators.
-//
-template <typename T1, typename T2>
-class CartesianProductGenerator2
-    : public ParamGeneratorInterface< ::testing::tuple<T1, T2> > {
- public:
-  typedef ::testing::tuple<T1, T2> ParamType;
-
-  CartesianProductGenerator2(const ParamGenerator<T1>& g1,
-      const ParamGenerator<T2>& g2)
-      : g1_(g1), g2_(g2) {}
-  virtual ~CartesianProductGenerator2() {}
-
-  virtual ParamIteratorInterface<ParamType>* Begin() const {
-    return new Iterator(this, g1_, g1_.begin(), g2_, g2_.begin());
-  }
-  virtual ParamIteratorInterface<ParamType>* End() const {
-    return new Iterator(this, g1_, g1_.end(), g2_, g2_.end());
-  }
-
- private:
-  class Iterator : public ParamIteratorInterface<ParamType> {
-   public:
-    Iterator(const ParamGeneratorInterface<ParamType>* base,
-      const ParamGenerator<T1>& g1,
-      const typename ParamGenerator<T1>::iterator& current1,
-      const ParamGenerator<T2>& g2,
-      const typename ParamGenerator<T2>::iterator& current2)
-        : base_(base),
-          begin1_(g1.begin()), end1_(g1.end()), current1_(current1),
-          begin2_(g2.begin()), end2_(g2.end()), current2_(current2)    {
-      ComputeCurrentValue();
-    }
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<ParamType>* BaseGenerator() const {
-      return base_;
-    }
-    // Advance should not be called on beyond-of-range iterators
-    // so no component iterators must be beyond end of range, either.
-    virtual void Advance() {
-      assert(!AtEnd());
-      ++current2_;
-      if (current2_ == end2_) {
-        current2_ = begin2_;
-        ++current1_;
-      }
-      ComputeCurrentValue();
-    }
-    virtual ParamIteratorInterface<ParamType>* Clone() const {
-      return new Iterator(*this);
-    }
-    virtual const ParamType* Current() const { return current_value_.get(); }
-    virtual bool Equals(const ParamIteratorInterface<ParamType>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      const Iterator* typed_other =
-          CheckedDowncastToActualType<const Iterator>(&other);
-      // We must report iterators equal if they both point beyond their
-      // respective ranges. That can happen in a variety of fashions,
-      // so we have to consult AtEnd().
-      return (AtEnd() && typed_other->AtEnd()) ||
-         (
-          current1_ == typed_other->current1_ &&
-          current2_ == typed_other->current2_);
-    }
-
-   private:
-    Iterator(const Iterator& other)
-        : base_(other.base_),
-        begin1_(other.begin1_),
-        end1_(other.end1_),
-        current1_(other.current1_),
-        begin2_(other.begin2_),
-        end2_(other.end2_),
-        current2_(other.current2_) {
-      ComputeCurrentValue();
-    }
-
-    void ComputeCurrentValue() {
-      if (!AtEnd())
-        current_value_.reset(new ParamType(*current1_, *current2_));
-    }
-    bool AtEnd() const {
-      // We must report iterator past the end of the range when either of the
-      // component iterators has reached the end of its range.
-      return
-          current1_ == end1_ ||
-          current2_ == end2_;
-    }
-
-    // No implementation - assignment is unsupported.
-    void operator=(const Iterator& other);
-
-    const ParamGeneratorInterface<ParamType>* const base_;
-    // begin[i]_ and end[i]_ define the i-th range that Iterator traverses.
-    // current[i]_ is the actual traversing iterator.
-    const typename ParamGenerator<T1>::iterator begin1_;
-    const typename ParamGenerator<T1>::iterator end1_;
-    typename ParamGenerator<T1>::iterator current1_;
-    const typename ParamGenerator<T2>::iterator begin2_;
-    const typename ParamGenerator<T2>::iterator end2_;
-    typename ParamGenerator<T2>::iterator current2_;
-    linked_ptr<ParamType> current_value_;
-  };  // class CartesianProductGenerator2::Iterator
-
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductGenerator2& other);
-
-  const ParamGenerator<T1> g1_;
-  const ParamGenerator<T2> g2_;
-};  // class CartesianProductGenerator2
-
-
-template <typename T1, typename T2, typename T3>
-class CartesianProductGenerator3
-    : public ParamGeneratorInterface< ::testing::tuple<T1, T2, T3> > {
- public:
-  typedef ::testing::tuple<T1, T2, T3> ParamType;
-
-  CartesianProductGenerator3(const ParamGenerator<T1>& g1,
-      const ParamGenerator<T2>& g2, const ParamGenerator<T3>& g3)
-      : g1_(g1), g2_(g2), g3_(g3) {}
-  virtual ~CartesianProductGenerator3() {}
-
-  virtual ParamIteratorInterface<ParamType>* Begin() const {
-    return new Iterator(this, g1_, g1_.begin(), g2_, g2_.begin(), g3_,
-        g3_.begin());
-  }
-  virtual ParamIteratorInterface<ParamType>* End() const {
-    return new Iterator(this, g1_, g1_.end(), g2_, g2_.end(), g3_, g3_.end());
-  }
-
- private:
-  class Iterator : public ParamIteratorInterface<ParamType> {
-   public:
-    Iterator(const ParamGeneratorInterface<ParamType>* base,
-      const ParamGenerator<T1>& g1,
-      const typename ParamGenerator<T1>::iterator& current1,
-      const ParamGenerator<T2>& g2,
-      const typename ParamGenerator<T2>::iterator& current2,
-      const ParamGenerator<T3>& g3,
-      const typename ParamGenerator<T3>::iterator& current3)
-        : base_(base),
-          begin1_(g1.begin()), end1_(g1.end()), current1_(current1),
-          begin2_(g2.begin()), end2_(g2.end()), current2_(current2),
-          begin3_(g3.begin()), end3_(g3.end()), current3_(current3)    {
-      ComputeCurrentValue();
-    }
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<ParamType>* BaseGenerator() const {
-      return base_;
-    }
-    // Advance should not be called on beyond-of-range iterators
-    // so no component iterators must be beyond end of range, either.
-    virtual void Advance() {
-      assert(!AtEnd());
-      ++current3_;
-      if (current3_ == end3_) {
-        current3_ = begin3_;
-        ++current2_;
-      }
-      if (current2_ == end2_) {
-        current2_ = begin2_;
-        ++current1_;
-      }
-      ComputeCurrentValue();
-    }
-    virtual ParamIteratorInterface<ParamType>* Clone() const {
-      return new Iterator(*this);
-    }
-    virtual const ParamType* Current() const { return current_value_.get(); }
-    virtual bool Equals(const ParamIteratorInterface<ParamType>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      const Iterator* typed_other =
-          CheckedDowncastToActualType<const Iterator>(&other);
-      // We must report iterators equal if they both point beyond their
-      // respective ranges. That can happen in a variety of fashions,
-      // so we have to consult AtEnd().
-      return (AtEnd() && typed_other->AtEnd()) ||
-         (
-          current1_ == typed_other->current1_ &&
-          current2_ == typed_other->current2_ &&
-          current3_ == typed_other->current3_);
-    }
-
-   private:
-    Iterator(const Iterator& other)
-        : base_(other.base_),
-        begin1_(other.begin1_),
-        end1_(other.end1_),
-        current1_(other.current1_),
-        begin2_(other.begin2_),
-        end2_(other.end2_),
-        current2_(other.current2_),
-        begin3_(other.begin3_),
-        end3_(other.end3_),
-        current3_(other.current3_) {
-      ComputeCurrentValue();
-    }
-
-    void ComputeCurrentValue() {
-      if (!AtEnd())
-        current_value_.reset(new ParamType(*current1_, *current2_, *current3_));
-    }
-    bool AtEnd() const {
-      // We must report iterator past the end of the range when either of the
-      // component iterators has reached the end of its range.
-      return
-          current1_ == end1_ ||
-          current2_ == end2_ ||
-          current3_ == end3_;
-    }
-
-    // No implementation - assignment is unsupported.
-    void operator=(const Iterator& other);
-
-    const ParamGeneratorInterface<ParamType>* const base_;
-    // begin[i]_ and end[i]_ define the i-th range that Iterator traverses.
-    // current[i]_ is the actual traversing iterator.
-    const typename ParamGenerator<T1>::iterator begin1_;
-    const typename ParamGenerator<T1>::iterator end1_;
-    typename ParamGenerator<T1>::iterator current1_;
-    const typename ParamGenerator<T2>::iterator begin2_;
-    const typename ParamGenerator<T2>::iterator end2_;
-    typename ParamGenerator<T2>::iterator current2_;
-    const typename ParamGenerator<T3>::iterator begin3_;
-    const typename ParamGenerator<T3>::iterator end3_;
-    typename ParamGenerator<T3>::iterator current3_;
-    linked_ptr<ParamType> current_value_;
-  };  // class CartesianProductGenerator3::Iterator
-
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductGenerator3& other);
-
-  const ParamGenerator<T1> g1_;
-  const ParamGenerator<T2> g2_;
-  const ParamGenerator<T3> g3_;
-};  // class CartesianProductGenerator3
-
-
-template <typename T1, typename T2, typename T3, typename T4>
-class CartesianProductGenerator4
-    : public ParamGeneratorInterface< ::testing::tuple<T1, T2, T3, T4> > {
- public:
-  typedef ::testing::tuple<T1, T2, T3, T4> ParamType;
-
-  CartesianProductGenerator4(const ParamGenerator<T1>& g1,
-      const ParamGenerator<T2>& g2, const ParamGenerator<T3>& g3,
-      const ParamGenerator<T4>& g4)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4) {}
-  virtual ~CartesianProductGenerator4() {}
-
-  virtual ParamIteratorInterface<ParamType>* Begin() const {
-    return new Iterator(this, g1_, g1_.begin(), g2_, g2_.begin(), g3_,
-        g3_.begin(), g4_, g4_.begin());
-  }
-  virtual ParamIteratorInterface<ParamType>* End() const {
-    return new Iterator(this, g1_, g1_.end(), g2_, g2_.end(), g3_, g3_.end(),
-        g4_, g4_.end());
-  }
-
- private:
-  class Iterator : public ParamIteratorInterface<ParamType> {
-   public:
-    Iterator(const ParamGeneratorInterface<ParamType>* base,
-      const ParamGenerator<T1>& g1,
-      const typename ParamGenerator<T1>::iterator& current1,
-      const ParamGenerator<T2>& g2,
-      const typename ParamGenerator<T2>::iterator& current2,
-      const ParamGenerator<T3>& g3,
-      const typename ParamGenerator<T3>::iterator& current3,
-      const ParamGenerator<T4>& g4,
-      const typename ParamGenerator<T4>::iterator& current4)
-        : base_(base),
-          begin1_(g1.begin()), end1_(g1.end()), current1_(current1),
-          begin2_(g2.begin()), end2_(g2.end()), current2_(current2),
-          begin3_(g3.begin()), end3_(g3.end()), current3_(current3),
-          begin4_(g4.begin()), end4_(g4.end()), current4_(current4)    {
-      ComputeCurrentValue();
-    }
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<ParamType>* BaseGenerator() const {
-      return base_;
-    }
-    // Advance should not be called on beyond-of-range iterators
-    // so no component iterators must be beyond end of range, either.
-    virtual void Advance() {
-      assert(!AtEnd());
-      ++current4_;
-      if (current4_ == end4_) {
-        current4_ = begin4_;
-        ++current3_;
-      }
-      if (current3_ == end3_) {
-        current3_ = begin3_;
-        ++current2_;
-      }
-      if (current2_ == end2_) {
-        current2_ = begin2_;
-        ++current1_;
-      }
-      ComputeCurrentValue();
-    }
-    virtual ParamIteratorInterface<ParamType>* Clone() const {
-      return new Iterator(*this);
-    }
-    virtual const ParamType* Current() const { return current_value_.get(); }
-    virtual bool Equals(const ParamIteratorInterface<ParamType>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      const Iterator* typed_other =
-          CheckedDowncastToActualType<const Iterator>(&other);
-      // We must report iterators equal if they both point beyond their
-      // respective ranges. That can happen in a variety of fashions,
-      // so we have to consult AtEnd().
-      return (AtEnd() && typed_other->AtEnd()) ||
-         (
-          current1_ == typed_other->current1_ &&
-          current2_ == typed_other->current2_ &&
-          current3_ == typed_other->current3_ &&
-          current4_ == typed_other->current4_);
-    }
-
-   private:
-    Iterator(const Iterator& other)
-        : base_(other.base_),
-        begin1_(other.begin1_),
-        end1_(other.end1_),
-        current1_(other.current1_),
-        begin2_(other.begin2_),
-        end2_(other.end2_),
-        current2_(other.current2_),
-        begin3_(other.begin3_),
-        end3_(other.end3_),
-        current3_(other.current3_),
-        begin4_(other.begin4_),
-        end4_(other.end4_),
-        current4_(other.current4_) {
-      ComputeCurrentValue();
-    }
-
-    void ComputeCurrentValue() {
-      if (!AtEnd())
-        current_value_.reset(new ParamType(*current1_, *current2_, *current3_,
-            *current4_));
-    }
-    bool AtEnd() const {
-      // We must report iterator past the end of the range when either of the
-      // component iterators has reached the end of its range.
-      return
-          current1_ == end1_ ||
-          current2_ == end2_ ||
-          current3_ == end3_ ||
-          current4_ == end4_;
-    }
-
-    // No implementation - assignment is unsupported.
-    void operator=(const Iterator& other);
-
-    const ParamGeneratorInterface<ParamType>* const base_;
-    // begin[i]_ and end[i]_ define the i-th range that Iterator traverses.
-    // current[i]_ is the actual traversing iterator.
-    const typename ParamGenerator<T1>::iterator begin1_;
-    const typename ParamGenerator<T1>::iterator end1_;
-    typename ParamGenerator<T1>::iterator current1_;
-    const typename ParamGenerator<T2>::iterator begin2_;
-    const typename ParamGenerator<T2>::iterator end2_;
-    typename ParamGenerator<T2>::iterator current2_;
-    const typename ParamGenerator<T3>::iterator begin3_;
-    const typename ParamGenerator<T3>::iterator end3_;
-    typename ParamGenerator<T3>::iterator current3_;
-    const typename ParamGenerator<T4>::iterator begin4_;
-    const typename ParamGenerator<T4>::iterator end4_;
-    typename ParamGenerator<T4>::iterator current4_;
-    linked_ptr<ParamType> current_value_;
-  };  // class CartesianProductGenerator4::Iterator
-
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductGenerator4& other);
-
-  const ParamGenerator<T1> g1_;
-  const ParamGenerator<T2> g2_;
-  const ParamGenerator<T3> g3_;
-  const ParamGenerator<T4> g4_;
-};  // class CartesianProductGenerator4
-
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5>
-class CartesianProductGenerator5
-    : public ParamGeneratorInterface< ::testing::tuple<T1, T2, T3, T4, T5> > {
- public:
-  typedef ::testing::tuple<T1, T2, T3, T4, T5> ParamType;
-
-  CartesianProductGenerator5(const ParamGenerator<T1>& g1,
-      const ParamGenerator<T2>& g2, const ParamGenerator<T3>& g3,
-      const ParamGenerator<T4>& g4, const ParamGenerator<T5>& g5)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5) {}
-  virtual ~CartesianProductGenerator5() {}
-
-  virtual ParamIteratorInterface<ParamType>* Begin() const {
-    return new Iterator(this, g1_, g1_.begin(), g2_, g2_.begin(), g3_,
-        g3_.begin(), g4_, g4_.begin(), g5_, g5_.begin());
-  }
-  virtual ParamIteratorInterface<ParamType>* End() const {
-    return new Iterator(this, g1_, g1_.end(), g2_, g2_.end(), g3_, g3_.end(),
-        g4_, g4_.end(), g5_, g5_.end());
-  }
-
- private:
-  class Iterator : public ParamIteratorInterface<ParamType> {
-   public:
-    Iterator(const ParamGeneratorInterface<ParamType>* base,
-      const ParamGenerator<T1>& g1,
-      const typename ParamGenerator<T1>::iterator& current1,
-      const ParamGenerator<T2>& g2,
-      const typename ParamGenerator<T2>::iterator& current2,
-      const ParamGenerator<T3>& g3,
-      const typename ParamGenerator<T3>::iterator& current3,
-      const ParamGenerator<T4>& g4,
-      const typename ParamGenerator<T4>::iterator& current4,
-      const ParamGenerator<T5>& g5,
-      const typename ParamGenerator<T5>::iterator& current5)
-        : base_(base),
-          begin1_(g1.begin()), end1_(g1.end()), current1_(current1),
-          begin2_(g2.begin()), end2_(g2.end()), current2_(current2),
-          begin3_(g3.begin()), end3_(g3.end()), current3_(current3),
-          begin4_(g4.begin()), end4_(g4.end()), current4_(current4),
-          begin5_(g5.begin()), end5_(g5.end()), current5_(current5)    {
-      ComputeCurrentValue();
-    }
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<ParamType>* BaseGenerator() const {
-      return base_;
-    }
-    // Advance should not be called on beyond-of-range iterators
-    // so no component iterators must be beyond end of range, either.
-    virtual void Advance() {
-      assert(!AtEnd());
-      ++current5_;
-      if (current5_ == end5_) {
-        current5_ = begin5_;
-        ++current4_;
-      }
-      if (current4_ == end4_) {
-        current4_ = begin4_;
-        ++current3_;
-      }
-      if (current3_ == end3_) {
-        current3_ = begin3_;
-        ++current2_;
-      }
-      if (current2_ == end2_) {
-        current2_ = begin2_;
-        ++current1_;
-      }
-      ComputeCurrentValue();
-    }
-    virtual ParamIteratorInterface<ParamType>* Clone() const {
-      return new Iterator(*this);
-    }
-    virtual const ParamType* Current() const { return current_value_.get(); }
-    virtual bool Equals(const ParamIteratorInterface<ParamType>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      const Iterator* typed_other =
-          CheckedDowncastToActualType<const Iterator>(&other);
-      // We must report iterators equal if they both point beyond their
-      // respective ranges. That can happen in a variety of fashions,
-      // so we have to consult AtEnd().
-      return (AtEnd() && typed_other->AtEnd()) ||
-         (
-          current1_ == typed_other->current1_ &&
-          current2_ == typed_other->current2_ &&
-          current3_ == typed_other->current3_ &&
-          current4_ == typed_other->current4_ &&
-          current5_ == typed_other->current5_);
-    }
-
-   private:
-    Iterator(const Iterator& other)
-        : base_(other.base_),
-        begin1_(other.begin1_),
-        end1_(other.end1_),
-        current1_(other.current1_),
-        begin2_(other.begin2_),
-        end2_(other.end2_),
-        current2_(other.current2_),
-        begin3_(other.begin3_),
-        end3_(other.end3_),
-        current3_(other.current3_),
-        begin4_(other.begin4_),
-        end4_(other.end4_),
-        current4_(other.current4_),
-        begin5_(other.begin5_),
-        end5_(other.end5_),
-        current5_(other.current5_) {
-      ComputeCurrentValue();
-    }
-
-    void ComputeCurrentValue() {
-      if (!AtEnd())
-        current_value_.reset(new ParamType(*current1_, *current2_, *current3_,
-            *current4_, *current5_));
-    }
-    bool AtEnd() const {
-      // We must report iterator past the end of the range when either of the
-      // component iterators has reached the end of its range.
-      return
-          current1_ == end1_ ||
-          current2_ == end2_ ||
-          current3_ == end3_ ||
-          current4_ == end4_ ||
-          current5_ == end5_;
-    }
-
-    // No implementation - assignment is unsupported.
-    void operator=(const Iterator& other);
-
-    const ParamGeneratorInterface<ParamType>* const base_;
-    // begin[i]_ and end[i]_ define the i-th range that Iterator traverses.
-    // current[i]_ is the actual traversing iterator.
-    const typename ParamGenerator<T1>::iterator begin1_;
-    const typename ParamGenerator<T1>::iterator end1_;
-    typename ParamGenerator<T1>::iterator current1_;
-    const typename ParamGenerator<T2>::iterator begin2_;
-    const typename ParamGenerator<T2>::iterator end2_;
-    typename ParamGenerator<T2>::iterator current2_;
-    const typename ParamGenerator<T3>::iterator begin3_;
-    const typename ParamGenerator<T3>::iterator end3_;
-    typename ParamGenerator<T3>::iterator current3_;
-    const typename ParamGenerator<T4>::iterator begin4_;
-    const typename ParamGenerator<T4>::iterator end4_;
-    typename ParamGenerator<T4>::iterator current4_;
-    const typename ParamGenerator<T5>::iterator begin5_;
-    const typename ParamGenerator<T5>::iterator end5_;
-    typename ParamGenerator<T5>::iterator current5_;
-    linked_ptr<ParamType> current_value_;
-  };  // class CartesianProductGenerator5::Iterator
-
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductGenerator5& other);
-
-  const ParamGenerator<T1> g1_;
-  const ParamGenerator<T2> g2_;
-  const ParamGenerator<T3> g3_;
-  const ParamGenerator<T4> g4_;
-  const ParamGenerator<T5> g5_;
-};  // class CartesianProductGenerator5
-
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6>
-class CartesianProductGenerator6
-    : public ParamGeneratorInterface< ::testing::tuple<T1, T2, T3, T4, T5,
-        T6> > {
- public:
-  typedef ::testing::tuple<T1, T2, T3, T4, T5, T6> ParamType;
-
-  CartesianProductGenerator6(const ParamGenerator<T1>& g1,
-      const ParamGenerator<T2>& g2, const ParamGenerator<T3>& g3,
-      const ParamGenerator<T4>& g4, const ParamGenerator<T5>& g5,
-      const ParamGenerator<T6>& g6)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5), g6_(g6) {}
-  virtual ~CartesianProductGenerator6() {}
-
-  virtual ParamIteratorInterface<ParamType>* Begin() const {
-    return new Iterator(this, g1_, g1_.begin(), g2_, g2_.begin(), g3_,
-        g3_.begin(), g4_, g4_.begin(), g5_, g5_.begin(), g6_, g6_.begin());
-  }
-  virtual ParamIteratorInterface<ParamType>* End() const {
-    return new Iterator(this, g1_, g1_.end(), g2_, g2_.end(), g3_, g3_.end(),
-        g4_, g4_.end(), g5_, g5_.end(), g6_, g6_.end());
-  }
-
- private:
-  class Iterator : public ParamIteratorInterface<ParamType> {
-   public:
-    Iterator(const ParamGeneratorInterface<ParamType>* base,
-      const ParamGenerator<T1>& g1,
-      const typename ParamGenerator<T1>::iterator& current1,
-      const ParamGenerator<T2>& g2,
-      const typename ParamGenerator<T2>::iterator& current2,
-      const ParamGenerator<T3>& g3,
-      const typename ParamGenerator<T3>::iterator& current3,
-      const ParamGenerator<T4>& g4,
-      const typename ParamGenerator<T4>::iterator& current4,
-      const ParamGenerator<T5>& g5,
-      const typename ParamGenerator<T5>::iterator& current5,
-      const ParamGenerator<T6>& g6,
-      const typename ParamGenerator<T6>::iterator& current6)
-        : base_(base),
-          begin1_(g1.begin()), end1_(g1.end()), current1_(current1),
-          begin2_(g2.begin()), end2_(g2.end()), current2_(current2),
-          begin3_(g3.begin()), end3_(g3.end()), current3_(current3),
-          begin4_(g4.begin()), end4_(g4.end()), current4_(current4),
-          begin5_(g5.begin()), end5_(g5.end()), current5_(current5),
-          begin6_(g6.begin()), end6_(g6.end()), current6_(current6)    {
-      ComputeCurrentValue();
-    }
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<ParamType>* BaseGenerator() const {
-      return base_;
-    }
-    // Advance should not be called on beyond-of-range iterators
-    // so no component iterators must be beyond end of range, either.
-    virtual void Advance() {
-      assert(!AtEnd());
-      ++current6_;
-      if (current6_ == end6_) {
-        current6_ = begin6_;
-        ++current5_;
-      }
-      if (current5_ == end5_) {
-        current5_ = begin5_;
-        ++current4_;
-      }
-      if (current4_ == end4_) {
-        current4_ = begin4_;
-        ++current3_;
-      }
-      if (current3_ == end3_) {
-        current3_ = begin3_;
-        ++current2_;
-      }
-      if (current2_ == end2_) {
-        current2_ = begin2_;
-        ++current1_;
-      }
-      ComputeCurrentValue();
-    }
-    virtual ParamIteratorInterface<ParamType>* Clone() const {
-      return new Iterator(*this);
-    }
-    virtual const ParamType* Current() const { return current_value_.get(); }
-    virtual bool Equals(const ParamIteratorInterface<ParamType>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      const Iterator* typed_other =
-          CheckedDowncastToActualType<const Iterator>(&other);
-      // We must report iterators equal if they both point beyond their
-      // respective ranges. That can happen in a variety of fashions,
-      // so we have to consult AtEnd().
-      return (AtEnd() && typed_other->AtEnd()) ||
-         (
-          current1_ == typed_other->current1_ &&
-          current2_ == typed_other->current2_ &&
-          current3_ == typed_other->current3_ &&
-          current4_ == typed_other->current4_ &&
-          current5_ == typed_other->current5_ &&
-          current6_ == typed_other->current6_);
-    }
-
-   private:
-    Iterator(const Iterator& other)
-        : base_(other.base_),
-        begin1_(other.begin1_),
-        end1_(other.end1_),
-        current1_(other.current1_),
-        begin2_(other.begin2_),
-        end2_(other.end2_),
-        current2_(other.current2_),
-        begin3_(other.begin3_),
-        end3_(other.end3_),
-        current3_(other.current3_),
-        begin4_(other.begin4_),
-        end4_(other.end4_),
-        current4_(other.current4_),
-        begin5_(other.begin5_),
-        end5_(other.end5_),
-        current5_(other.current5_),
-        begin6_(other.begin6_),
-        end6_(other.end6_),
-        current6_(other.current6_) {
-      ComputeCurrentValue();
-    }
-
-    void ComputeCurrentValue() {
-      if (!AtEnd())
-        current_value_.reset(new ParamType(*current1_, *current2_, *current3_,
-            *current4_, *current5_, *current6_));
-    }
-    bool AtEnd() const {
-      // We must report iterator past the end of the range when either of the
-      // component iterators has reached the end of its range.
-      return
-          current1_ == end1_ ||
-          current2_ == end2_ ||
-          current3_ == end3_ ||
-          current4_ == end4_ ||
-          current5_ == end5_ ||
-          current6_ == end6_;
-    }
-
-    // No implementation - assignment is unsupported.
-    void operator=(const Iterator& other);
-
-    const ParamGeneratorInterface<ParamType>* const base_;
-    // begin[i]_ and end[i]_ define the i-th range that Iterator traverses.
-    // current[i]_ is the actual traversing iterator.
-    const typename ParamGenerator<T1>::iterator begin1_;
-    const typename ParamGenerator<T1>::iterator end1_;
-    typename ParamGenerator<T1>::iterator current1_;
-    const typename ParamGenerator<T2>::iterator begin2_;
-    const typename ParamGenerator<T2>::iterator end2_;
-    typename ParamGenerator<T2>::iterator current2_;
-    const typename ParamGenerator<T3>::iterator begin3_;
-    const typename ParamGenerator<T3>::iterator end3_;
-    typename ParamGenerator<T3>::iterator current3_;
-    const typename ParamGenerator<T4>::iterator begin4_;
-    const typename ParamGenerator<T4>::iterator end4_;
-    typename ParamGenerator<T4>::iterator current4_;
-    const typename ParamGenerator<T5>::iterator begin5_;
-    const typename ParamGenerator<T5>::iterator end5_;
-    typename ParamGenerator<T5>::iterator current5_;
-    const typename ParamGenerator<T6>::iterator begin6_;
-    const typename ParamGenerator<T6>::iterator end6_;
-    typename ParamGenerator<T6>::iterator current6_;
-    linked_ptr<ParamType> current_value_;
-  };  // class CartesianProductGenerator6::Iterator
-
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductGenerator6& other);
-
-  const ParamGenerator<T1> g1_;
-  const ParamGenerator<T2> g2_;
-  const ParamGenerator<T3> g3_;
-  const ParamGenerator<T4> g4_;
-  const ParamGenerator<T5> g5_;
-  const ParamGenerator<T6> g6_;
-};  // class CartesianProductGenerator6
-
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7>
-class CartesianProductGenerator7
-    : public ParamGeneratorInterface< ::testing::tuple<T1, T2, T3, T4, T5, T6,
-        T7> > {
- public:
-  typedef ::testing::tuple<T1, T2, T3, T4, T5, T6, T7> ParamType;
-
-  CartesianProductGenerator7(const ParamGenerator<T1>& g1,
-      const ParamGenerator<T2>& g2, const ParamGenerator<T3>& g3,
-      const ParamGenerator<T4>& g4, const ParamGenerator<T5>& g5,
-      const ParamGenerator<T6>& g6, const ParamGenerator<T7>& g7)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5), g6_(g6), g7_(g7) {}
-  virtual ~CartesianProductGenerator7() {}
-
-  virtual ParamIteratorInterface<ParamType>* Begin() const {
-    return new Iterator(this, g1_, g1_.begin(), g2_, g2_.begin(), g3_,
-        g3_.begin(), g4_, g4_.begin(), g5_, g5_.begin(), g6_, g6_.begin(), g7_,
-        g7_.begin());
-  }
-  virtual ParamIteratorInterface<ParamType>* End() const {
-    return new Iterator(this, g1_, g1_.end(), g2_, g2_.end(), g3_, g3_.end(),
-        g4_, g4_.end(), g5_, g5_.end(), g6_, g6_.end(), g7_, g7_.end());
-  }
-
- private:
-  class Iterator : public ParamIteratorInterface<ParamType> {
-   public:
-    Iterator(const ParamGeneratorInterface<ParamType>* base,
-      const ParamGenerator<T1>& g1,
-      const typename ParamGenerator<T1>::iterator& current1,
-      const ParamGenerator<T2>& g2,
-      const typename ParamGenerator<T2>::iterator& current2,
-      const ParamGenerator<T3>& g3,
-      const typename ParamGenerator<T3>::iterator& current3,
-      const ParamGenerator<T4>& g4,
-      const typename ParamGenerator<T4>::iterator& current4,
-      const ParamGenerator<T5>& g5,
-      const typename ParamGenerator<T5>::iterator& current5,
-      const ParamGenerator<T6>& g6,
-      const typename ParamGenerator<T6>::iterator& current6,
-      const ParamGenerator<T7>& g7,
-      const typename ParamGenerator<T7>::iterator& current7)
-        : base_(base),
-          begin1_(g1.begin()), end1_(g1.end()), current1_(current1),
-          begin2_(g2.begin()), end2_(g2.end()), current2_(current2),
-          begin3_(g3.begin()), end3_(g3.end()), current3_(current3),
-          begin4_(g4.begin()), end4_(g4.end()), current4_(current4),
-          begin5_(g5.begin()), end5_(g5.end()), current5_(current5),
-          begin6_(g6.begin()), end6_(g6.end()), current6_(current6),
-          begin7_(g7.begin()), end7_(g7.end()), current7_(current7)    {
-      ComputeCurrentValue();
-    }
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<ParamType>* BaseGenerator() const {
-      return base_;
-    }
-    // Advance should not be called on beyond-of-range iterators
-    // so no component iterators must be beyond end of range, either.
-    virtual void Advance() {
-      assert(!AtEnd());
-      ++current7_;
-      if (current7_ == end7_) {
-        current7_ = begin7_;
-        ++current6_;
-      }
-      if (current6_ == end6_) {
-        current6_ = begin6_;
-        ++current5_;
-      }
-      if (current5_ == end5_) {
-        current5_ = begin5_;
-        ++current4_;
-      }
-      if (current4_ == end4_) {
-        current4_ = begin4_;
-        ++current3_;
-      }
-      if (current3_ == end3_) {
-        current3_ = begin3_;
-        ++current2_;
-      }
-      if (current2_ == end2_) {
-        current2_ = begin2_;
-        ++current1_;
-      }
-      ComputeCurrentValue();
-    }
-    virtual ParamIteratorInterface<ParamType>* Clone() const {
-      return new Iterator(*this);
-    }
-    virtual const ParamType* Current() const { return current_value_.get(); }
-    virtual bool Equals(const ParamIteratorInterface<ParamType>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      const Iterator* typed_other =
-          CheckedDowncastToActualType<const Iterator>(&other);
-      // We must report iterators equal if they both point beyond their
-      // respective ranges. That can happen in a variety of fashions,
-      // so we have to consult AtEnd().
-      return (AtEnd() && typed_other->AtEnd()) ||
-         (
-          current1_ == typed_other->current1_ &&
-          current2_ == typed_other->current2_ &&
-          current3_ == typed_other->current3_ &&
-          current4_ == typed_other->current4_ &&
-          current5_ == typed_other->current5_ &&
-          current6_ == typed_other->current6_ &&
-          current7_ == typed_other->current7_);
-    }
-
-   private:
-    Iterator(const Iterator& other)
-        : base_(other.base_),
-        begin1_(other.begin1_),
-        end1_(other.end1_),
-        current1_(other.current1_),
-        begin2_(other.begin2_),
-        end2_(other.end2_),
-        current2_(other.current2_),
-        begin3_(other.begin3_),
-        end3_(other.end3_),
-        current3_(other.current3_),
-        begin4_(other.begin4_),
-        end4_(other.end4_),
-        current4_(other.current4_),
-        begin5_(other.begin5_),
-        end5_(other.end5_),
-        current5_(other.current5_),
-        begin6_(other.begin6_),
-        end6_(other.end6_),
-        current6_(other.current6_),
-        begin7_(other.begin7_),
-        end7_(other.end7_),
-        current7_(other.current7_) {
-      ComputeCurrentValue();
-    }
-
-    void ComputeCurrentValue() {
-      if (!AtEnd())
-        current_value_.reset(new ParamType(*current1_, *current2_, *current3_,
-            *current4_, *current5_, *current6_, *current7_));
-    }
-    bool AtEnd() const {
-      // We must report iterator past the end of the range when either of the
-      // component iterators has reached the end of its range.
-      return
-          current1_ == end1_ ||
-          current2_ == end2_ ||
-          current3_ == end3_ ||
-          current4_ == end4_ ||
-          current5_ == end5_ ||
-          current6_ == end6_ ||
-          current7_ == end7_;
-    }
-
-    // No implementation - assignment is unsupported.
-    void operator=(const Iterator& other);
-
-    const ParamGeneratorInterface<ParamType>* const base_;
-    // begin[i]_ and end[i]_ define the i-th range that Iterator traverses.
-    // current[i]_ is the actual traversing iterator.
-    const typename ParamGenerator<T1>::iterator begin1_;
-    const typename ParamGenerator<T1>::iterator end1_;
-    typename ParamGenerator<T1>::iterator current1_;
-    const typename ParamGenerator<T2>::iterator begin2_;
-    const typename ParamGenerator<T2>::iterator end2_;
-    typename ParamGenerator<T2>::iterator current2_;
-    const typename ParamGenerator<T3>::iterator begin3_;
-    const typename ParamGenerator<T3>::iterator end3_;
-    typename ParamGenerator<T3>::iterator current3_;
-    const typename ParamGenerator<T4>::iterator begin4_;
-    const typename ParamGenerator<T4>::iterator end4_;
-    typename ParamGenerator<T4>::iterator current4_;
-    const typename ParamGenerator<T5>::iterator begin5_;
-    const typename ParamGenerator<T5>::iterator end5_;
-    typename ParamGenerator<T5>::iterator current5_;
-    const typename ParamGenerator<T6>::iterator begin6_;
-    const typename ParamGenerator<T6>::iterator end6_;
-    typename ParamGenerator<T6>::iterator current6_;
-    const typename ParamGenerator<T7>::iterator begin7_;
-    const typename ParamGenerator<T7>::iterator end7_;
-    typename ParamGenerator<T7>::iterator current7_;
-    linked_ptr<ParamType> current_value_;
-  };  // class CartesianProductGenerator7::Iterator
-
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductGenerator7& other);
-
-  const ParamGenerator<T1> g1_;
-  const ParamGenerator<T2> g2_;
-  const ParamGenerator<T3> g3_;
-  const ParamGenerator<T4> g4_;
-  const ParamGenerator<T5> g5_;
-  const ParamGenerator<T6> g6_;
-  const ParamGenerator<T7> g7_;
-};  // class CartesianProductGenerator7
-
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8>
-class CartesianProductGenerator8
-    : public ParamGeneratorInterface< ::testing::tuple<T1, T2, T3, T4, T5, T6,
-        T7, T8> > {
- public:
-  typedef ::testing::tuple<T1, T2, T3, T4, T5, T6, T7, T8> ParamType;
-
-  CartesianProductGenerator8(const ParamGenerator<T1>& g1,
-      const ParamGenerator<T2>& g2, const ParamGenerator<T3>& g3,
-      const ParamGenerator<T4>& g4, const ParamGenerator<T5>& g5,
-      const ParamGenerator<T6>& g6, const ParamGenerator<T7>& g7,
-      const ParamGenerator<T8>& g8)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5), g6_(g6), g7_(g7),
-          g8_(g8) {}
-  virtual ~CartesianProductGenerator8() {}
-
-  virtual ParamIteratorInterface<ParamType>* Begin() const {
-    return new Iterator(this, g1_, g1_.begin(), g2_, g2_.begin(), g3_,
-        g3_.begin(), g4_, g4_.begin(), g5_, g5_.begin(), g6_, g6_.begin(), g7_,
-        g7_.begin(), g8_, g8_.begin());
-  }
-  virtual ParamIteratorInterface<ParamType>* End() const {
-    return new Iterator(this, g1_, g1_.end(), g2_, g2_.end(), g3_, g3_.end(),
-        g4_, g4_.end(), g5_, g5_.end(), g6_, g6_.end(), g7_, g7_.end(), g8_,
-        g8_.end());
-  }
-
- private:
-  class Iterator : public ParamIteratorInterface<ParamType> {
-   public:
-    Iterator(const ParamGeneratorInterface<ParamType>* base,
-      const ParamGenerator<T1>& g1,
-      const typename ParamGenerator<T1>::iterator& current1,
-      const ParamGenerator<T2>& g2,
-      const typename ParamGenerator<T2>::iterator& current2,
-      const ParamGenerator<T3>& g3,
-      const typename ParamGenerator<T3>::iterator& current3,
-      const ParamGenerator<T4>& g4,
-      const typename ParamGenerator<T4>::iterator& current4,
-      const ParamGenerator<T5>& g5,
-      const typename ParamGenerator<T5>::iterator& current5,
-      const ParamGenerator<T6>& g6,
-      const typename ParamGenerator<T6>::iterator& current6,
-      const ParamGenerator<T7>& g7,
-      const typename ParamGenerator<T7>::iterator& current7,
-      const ParamGenerator<T8>& g8,
-      const typename ParamGenerator<T8>::iterator& current8)
-        : base_(base),
-          begin1_(g1.begin()), end1_(g1.end()), current1_(current1),
-          begin2_(g2.begin()), end2_(g2.end()), current2_(current2),
-          begin3_(g3.begin()), end3_(g3.end()), current3_(current3),
-          begin4_(g4.begin()), end4_(g4.end()), current4_(current4),
-          begin5_(g5.begin()), end5_(g5.end()), current5_(current5),
-          begin6_(g6.begin()), end6_(g6.end()), current6_(current6),
-          begin7_(g7.begin()), end7_(g7.end()), current7_(current7),
-          begin8_(g8.begin()), end8_(g8.end()), current8_(current8)    {
-      ComputeCurrentValue();
-    }
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<ParamType>* BaseGenerator() const {
-      return base_;
-    }
-    // Advance should not be called on beyond-of-range iterators
-    // so no component iterators must be beyond end of range, either.
-    virtual void Advance() {
-      assert(!AtEnd());
-      ++current8_;
-      if (current8_ == end8_) {
-        current8_ = begin8_;
-        ++current7_;
-      }
-      if (current7_ == end7_) {
-        current7_ = begin7_;
-        ++current6_;
-      }
-      if (current6_ == end6_) {
-        current6_ = begin6_;
-        ++current5_;
-      }
-      if (current5_ == end5_) {
-        current5_ = begin5_;
-        ++current4_;
-      }
-      if (current4_ == end4_) {
-        current4_ = begin4_;
-        ++current3_;
-      }
-      if (current3_ == end3_) {
-        current3_ = begin3_;
-        ++current2_;
-      }
-      if (current2_ == end2_) {
-        current2_ = begin2_;
-        ++current1_;
-      }
-      ComputeCurrentValue();
-    }
-    virtual ParamIteratorInterface<ParamType>* Clone() const {
-      return new Iterator(*this);
-    }
-    virtual const ParamType* Current() const { return current_value_.get(); }
-    virtual bool Equals(const ParamIteratorInterface<ParamType>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      const Iterator* typed_other =
-          CheckedDowncastToActualType<const Iterator>(&other);
-      // We must report iterators equal if they both point beyond their
-      // respective ranges. That can happen in a variety of fashions,
-      // so we have to consult AtEnd().
-      return (AtEnd() && typed_other->AtEnd()) ||
-         (
-          current1_ == typed_other->current1_ &&
-          current2_ == typed_other->current2_ &&
-          current3_ == typed_other->current3_ &&
-          current4_ == typed_other->current4_ &&
-          current5_ == typed_other->current5_ &&
-          current6_ == typed_other->current6_ &&
-          current7_ == typed_other->current7_ &&
-          current8_ == typed_other->current8_);
-    }
-
-   private:
-    Iterator(const Iterator& other)
-        : base_(other.base_),
-        begin1_(other.begin1_),
-        end1_(other.end1_),
-        current1_(other.current1_),
-        begin2_(other.begin2_),
-        end2_(other.end2_),
-        current2_(other.current2_),
-        begin3_(other.begin3_),
-        end3_(other.end3_),
-        current3_(other.current3_),
-        begin4_(other.begin4_),
-        end4_(other.end4_),
-        current4_(other.current4_),
-        begin5_(other.begin5_),
-        end5_(other.end5_),
-        current5_(other.current5_),
-        begin6_(other.begin6_),
-        end6_(other.end6_),
-        current6_(other.current6_),
-        begin7_(other.begin7_),
-        end7_(other.end7_),
-        current7_(other.current7_),
-        begin8_(other.begin8_),
-        end8_(other.end8_),
-        current8_(other.current8_) {
-      ComputeCurrentValue();
-    }
-
-    void ComputeCurrentValue() {
-      if (!AtEnd())
-        current_value_.reset(new ParamType(*current1_, *current2_, *current3_,
-            *current4_, *current5_, *current6_, *current7_, *current8_));
-    }
-    bool AtEnd() const {
-      // We must report iterator past the end of the range when either of the
-      // component iterators has reached the end of its range.
-      return
-          current1_ == end1_ ||
-          current2_ == end2_ ||
-          current3_ == end3_ ||
-          current4_ == end4_ ||
-          current5_ == end5_ ||
-          current6_ == end6_ ||
-          current7_ == end7_ ||
-          current8_ == end8_;
-    }
-
-    // No implementation - assignment is unsupported.
-    void operator=(const Iterator& other);
-
-    const ParamGeneratorInterface<ParamType>* const base_;
-    // begin[i]_ and end[i]_ define the i-th range that Iterator traverses.
-    // current[i]_ is the actual traversing iterator.
-    const typename ParamGenerator<T1>::iterator begin1_;
-    const typename ParamGenerator<T1>::iterator end1_;
-    typename ParamGenerator<T1>::iterator current1_;
-    const typename ParamGenerator<T2>::iterator begin2_;
-    const typename ParamGenerator<T2>::iterator end2_;
-    typename ParamGenerator<T2>::iterator current2_;
-    const typename ParamGenerator<T3>::iterator begin3_;
-    const typename ParamGenerator<T3>::iterator end3_;
-    typename ParamGenerator<T3>::iterator current3_;
-    const typename ParamGenerator<T4>::iterator begin4_;
-    const typename ParamGenerator<T4>::iterator end4_;
-    typename ParamGenerator<T4>::iterator current4_;
-    const typename ParamGenerator<T5>::iterator begin5_;
-    const typename ParamGenerator<T5>::iterator end5_;
-    typename ParamGenerator<T5>::iterator current5_;
-    const typename ParamGenerator<T6>::iterator begin6_;
-    const typename ParamGenerator<T6>::iterator end6_;
-    typename ParamGenerator<T6>::iterator current6_;
-    const typename ParamGenerator<T7>::iterator begin7_;
-    const typename ParamGenerator<T7>::iterator end7_;
-    typename ParamGenerator<T7>::iterator current7_;
-    const typename ParamGenerator<T8>::iterator begin8_;
-    const typename ParamGenerator<T8>::iterator end8_;
-    typename ParamGenerator<T8>::iterator current8_;
-    linked_ptr<ParamType> current_value_;
-  };  // class CartesianProductGenerator8::Iterator
-
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductGenerator8& other);
-
-  const ParamGenerator<T1> g1_;
-  const ParamGenerator<T2> g2_;
-  const ParamGenerator<T3> g3_;
-  const ParamGenerator<T4> g4_;
-  const ParamGenerator<T5> g5_;
-  const ParamGenerator<T6> g6_;
-  const ParamGenerator<T7> g7_;
-  const ParamGenerator<T8> g8_;
-};  // class CartesianProductGenerator8
-
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9>
-class CartesianProductGenerator9
-    : public ParamGeneratorInterface< ::testing::tuple<T1, T2, T3, T4, T5, T6,
-        T7, T8, T9> > {
- public:
-  typedef ::testing::tuple<T1, T2, T3, T4, T5, T6, T7, T8, T9> ParamType;
-
-  CartesianProductGenerator9(const ParamGenerator<T1>& g1,
-      const ParamGenerator<T2>& g2, const ParamGenerator<T3>& g3,
-      const ParamGenerator<T4>& g4, const ParamGenerator<T5>& g5,
-      const ParamGenerator<T6>& g6, const ParamGenerator<T7>& g7,
-      const ParamGenerator<T8>& g8, const ParamGenerator<T9>& g9)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5), g6_(g6), g7_(g7), g8_(g8),
-          g9_(g9) {}
-  virtual ~CartesianProductGenerator9() {}
-
-  virtual ParamIteratorInterface<ParamType>* Begin() const {
-    return new Iterator(this, g1_, g1_.begin(), g2_, g2_.begin(), g3_,
-        g3_.begin(), g4_, g4_.begin(), g5_, g5_.begin(), g6_, g6_.begin(), g7_,
-        g7_.begin(), g8_, g8_.begin(), g9_, g9_.begin());
-  }
-  virtual ParamIteratorInterface<ParamType>* End() const {
-    return new Iterator(this, g1_, g1_.end(), g2_, g2_.end(), g3_, g3_.end(),
-        g4_, g4_.end(), g5_, g5_.end(), g6_, g6_.end(), g7_, g7_.end(), g8_,
-        g8_.end(), g9_, g9_.end());
-  }
-
- private:
-  class Iterator : public ParamIteratorInterface<ParamType> {
-   public:
-    Iterator(const ParamGeneratorInterface<ParamType>* base,
-      const ParamGenerator<T1>& g1,
-      const typename ParamGenerator<T1>::iterator& current1,
-      const ParamGenerator<T2>& g2,
-      const typename ParamGenerator<T2>::iterator& current2,
-      const ParamGenerator<T3>& g3,
-      const typename ParamGenerator<T3>::iterator& current3,
-      const ParamGenerator<T4>& g4,
-      const typename ParamGenerator<T4>::iterator& current4,
-      const ParamGenerator<T5>& g5,
-      const typename ParamGenerator<T5>::iterator& current5,
-      const ParamGenerator<T6>& g6,
-      const typename ParamGenerator<T6>::iterator& current6,
-      const ParamGenerator<T7>& g7,
-      const typename ParamGenerator<T7>::iterator& current7,
-      const ParamGenerator<T8>& g8,
-      const typename ParamGenerator<T8>::iterator& current8,
-      const ParamGenerator<T9>& g9,
-      const typename ParamGenerator<T9>::iterator& current9)
-        : base_(base),
-          begin1_(g1.begin()), end1_(g1.end()), current1_(current1),
-          begin2_(g2.begin()), end2_(g2.end()), current2_(current2),
-          begin3_(g3.begin()), end3_(g3.end()), current3_(current3),
-          begin4_(g4.begin()), end4_(g4.end()), current4_(current4),
-          begin5_(g5.begin()), end5_(g5.end()), current5_(current5),
-          begin6_(g6.begin()), end6_(g6.end()), current6_(current6),
-          begin7_(g7.begin()), end7_(g7.end()), current7_(current7),
-          begin8_(g8.begin()), end8_(g8.end()), current8_(current8),
-          begin9_(g9.begin()), end9_(g9.end()), current9_(current9)    {
-      ComputeCurrentValue();
-    }
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<ParamType>* BaseGenerator() const {
-      return base_;
-    }
-    // Advance should not be called on beyond-of-range iterators
-    // so no component iterators must be beyond end of range, either.
-    virtual void Advance() {
-      assert(!AtEnd());
-      ++current9_;
-      if (current9_ == end9_) {
-        current9_ = begin9_;
-        ++current8_;
-      }
-      if (current8_ == end8_) {
-        current8_ = begin8_;
-        ++current7_;
-      }
-      if (current7_ == end7_) {
-        current7_ = begin7_;
-        ++current6_;
-      }
-      if (current6_ == end6_) {
-        current6_ = begin6_;
-        ++current5_;
-      }
-      if (current5_ == end5_) {
-        current5_ = begin5_;
-        ++current4_;
-      }
-      if (current4_ == end4_) {
-        current4_ = begin4_;
-        ++current3_;
-      }
-      if (current3_ == end3_) {
-        current3_ = begin3_;
-        ++current2_;
-      }
-      if (current2_ == end2_) {
-        current2_ = begin2_;
-        ++current1_;
-      }
-      ComputeCurrentValue();
-    }
-    virtual ParamIteratorInterface<ParamType>* Clone() const {
-      return new Iterator(*this);
-    }
-    virtual const ParamType* Current() const { return current_value_.get(); }
-    virtual bool Equals(const ParamIteratorInterface<ParamType>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      const Iterator* typed_other =
-          CheckedDowncastToActualType<const Iterator>(&other);
-      // We must report iterators equal if they both point beyond their
-      // respective ranges. That can happen in a variety of fashions,
-      // so we have to consult AtEnd().
-      return (AtEnd() && typed_other->AtEnd()) ||
-         (
-          current1_ == typed_other->current1_ &&
-          current2_ == typed_other->current2_ &&
-          current3_ == typed_other->current3_ &&
-          current4_ == typed_other->current4_ &&
-          current5_ == typed_other->current5_ &&
-          current6_ == typed_other->current6_ &&
-          current7_ == typed_other->current7_ &&
-          current8_ == typed_other->current8_ &&
-          current9_ == typed_other->current9_);
-    }
-
-   private:
-    Iterator(const Iterator& other)
-        : base_(other.base_),
-        begin1_(other.begin1_),
-        end1_(other.end1_),
-        current1_(other.current1_),
-        begin2_(other.begin2_),
-        end2_(other.end2_),
-        current2_(other.current2_),
-        begin3_(other.begin3_),
-        end3_(other.end3_),
-        current3_(other.current3_),
-        begin4_(other.begin4_),
-        end4_(other.end4_),
-        current4_(other.current4_),
-        begin5_(other.begin5_),
-        end5_(other.end5_),
-        current5_(other.current5_),
-        begin6_(other.begin6_),
-        end6_(other.end6_),
-        current6_(other.current6_),
-        begin7_(other.begin7_),
-        end7_(other.end7_),
-        current7_(other.current7_),
-        begin8_(other.begin8_),
-        end8_(other.end8_),
-        current8_(other.current8_),
-        begin9_(other.begin9_),
-        end9_(other.end9_),
-        current9_(other.current9_) {
-      ComputeCurrentValue();
-    }
-
-    void ComputeCurrentValue() {
-      if (!AtEnd())
-        current_value_.reset(new ParamType(*current1_, *current2_, *current3_,
-            *current4_, *current5_, *current6_, *current7_, *current8_,
-            *current9_));
-    }
-    bool AtEnd() const {
-      // We must report iterator past the end of the range when either of the
-      // component iterators has reached the end of its range.
-      return
-          current1_ == end1_ ||
-          current2_ == end2_ ||
-          current3_ == end3_ ||
-          current4_ == end4_ ||
-          current5_ == end5_ ||
-          current6_ == end6_ ||
-          current7_ == end7_ ||
-          current8_ == end8_ ||
-          current9_ == end9_;
-    }
-
-    // No implementation - assignment is unsupported.
-    void operator=(const Iterator& other);
-
-    const ParamGeneratorInterface<ParamType>* const base_;
-    // begin[i]_ and end[i]_ define the i-th range that Iterator traverses.
-    // current[i]_ is the actual traversing iterator.
-    const typename ParamGenerator<T1>::iterator begin1_;
-    const typename ParamGenerator<T1>::iterator end1_;
-    typename ParamGenerator<T1>::iterator current1_;
-    const typename ParamGenerator<T2>::iterator begin2_;
-    const typename ParamGenerator<T2>::iterator end2_;
-    typename ParamGenerator<T2>::iterator current2_;
-    const typename ParamGenerator<T3>::iterator begin3_;
-    const typename ParamGenerator<T3>::iterator end3_;
-    typename ParamGenerator<T3>::iterator current3_;
-    const typename ParamGenerator<T4>::iterator begin4_;
-    const typename ParamGenerator<T4>::iterator end4_;
-    typename ParamGenerator<T4>::iterator current4_;
-    const typename ParamGenerator<T5>::iterator begin5_;
-    const typename ParamGenerator<T5>::iterator end5_;
-    typename ParamGenerator<T5>::iterator current5_;
-    const typename ParamGenerator<T6>::iterator begin6_;
-    const typename ParamGenerator<T6>::iterator end6_;
-    typename ParamGenerator<T6>::iterator current6_;
-    const typename ParamGenerator<T7>::iterator begin7_;
-    const typename ParamGenerator<T7>::iterator end7_;
-    typename ParamGenerator<T7>::iterator current7_;
-    const typename ParamGenerator<T8>::iterator begin8_;
-    const typename ParamGenerator<T8>::iterator end8_;
-    typename ParamGenerator<T8>::iterator current8_;
-    const typename ParamGenerator<T9>::iterator begin9_;
-    const typename ParamGenerator<T9>::iterator end9_;
-    typename ParamGenerator<T9>::iterator current9_;
-    linked_ptr<ParamType> current_value_;
-  };  // class CartesianProductGenerator9::Iterator
-
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductGenerator9& other);
-
-  const ParamGenerator<T1> g1_;
-  const ParamGenerator<T2> g2_;
-  const ParamGenerator<T3> g3_;
-  const ParamGenerator<T4> g4_;
-  const ParamGenerator<T5> g5_;
-  const ParamGenerator<T6> g6_;
-  const ParamGenerator<T7> g7_;
-  const ParamGenerator<T8> g8_;
-  const ParamGenerator<T9> g9_;
-};  // class CartesianProductGenerator9
-
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10>
-class CartesianProductGenerator10
-    : public ParamGeneratorInterface< ::testing::tuple<T1, T2, T3, T4, T5, T6,
-        T7, T8, T9, T10> > {
- public:
-  typedef ::testing::tuple<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10> ParamType;
-
-  CartesianProductGenerator10(const ParamGenerator<T1>& g1,
-      const ParamGenerator<T2>& g2, const ParamGenerator<T3>& g3,
-      const ParamGenerator<T4>& g4, const ParamGenerator<T5>& g5,
-      const ParamGenerator<T6>& g6, const ParamGenerator<T7>& g7,
-      const ParamGenerator<T8>& g8, const ParamGenerator<T9>& g9,
-      const ParamGenerator<T10>& g10)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5), g6_(g6), g7_(g7), g8_(g8),
-          g9_(g9), g10_(g10) {}
-  virtual ~CartesianProductGenerator10() {}
-
-  virtual ParamIteratorInterface<ParamType>* Begin() const {
-    return new Iterator(this, g1_, g1_.begin(), g2_, g2_.begin(), g3_,
-        g3_.begin(), g4_, g4_.begin(), g5_, g5_.begin(), g6_, g6_.begin(), g7_,
-        g7_.begin(), g8_, g8_.begin(), g9_, g9_.begin(), g10_, g10_.begin());
-  }
-  virtual ParamIteratorInterface<ParamType>* End() const {
-    return new Iterator(this, g1_, g1_.end(), g2_, g2_.end(), g3_, g3_.end(),
-        g4_, g4_.end(), g5_, g5_.end(), g6_, g6_.end(), g7_, g7_.end(), g8_,
-        g8_.end(), g9_, g9_.end(), g10_, g10_.end());
-  }
-
- private:
-  class Iterator : public ParamIteratorInterface<ParamType> {
-   public:
-    Iterator(const ParamGeneratorInterface<ParamType>* base,
-      const ParamGenerator<T1>& g1,
-      const typename ParamGenerator<T1>::iterator& current1,
-      const ParamGenerator<T2>& g2,
-      const typename ParamGenerator<T2>::iterator& current2,
-      const ParamGenerator<T3>& g3,
-      const typename ParamGenerator<T3>::iterator& current3,
-      const ParamGenerator<T4>& g4,
-      const typename ParamGenerator<T4>::iterator& current4,
-      const ParamGenerator<T5>& g5,
-      const typename ParamGenerator<T5>::iterator& current5,
-      const ParamGenerator<T6>& g6,
-      const typename ParamGenerator<T6>::iterator& current6,
-      const ParamGenerator<T7>& g7,
-      const typename ParamGenerator<T7>::iterator& current7,
-      const ParamGenerator<T8>& g8,
-      const typename ParamGenerator<T8>::iterator& current8,
-      const ParamGenerator<T9>& g9,
-      const typename ParamGenerator<T9>::iterator& current9,
-      const ParamGenerator<T10>& g10,
-      const typename ParamGenerator<T10>::iterator& current10)
-        : base_(base),
-          begin1_(g1.begin()), end1_(g1.end()), current1_(current1),
-          begin2_(g2.begin()), end2_(g2.end()), current2_(current2),
-          begin3_(g3.begin()), end3_(g3.end()), current3_(current3),
-          begin4_(g4.begin()), end4_(g4.end()), current4_(current4),
-          begin5_(g5.begin()), end5_(g5.end()), current5_(current5),
-          begin6_(g6.begin()), end6_(g6.end()), current6_(current6),
-          begin7_(g7.begin()), end7_(g7.end()), current7_(current7),
-          begin8_(g8.begin()), end8_(g8.end()), current8_(current8),
-          begin9_(g9.begin()), end9_(g9.end()), current9_(current9),
-          begin10_(g10.begin()), end10_(g10.end()), current10_(current10)    {
-      ComputeCurrentValue();
-    }
-    virtual ~Iterator() {}
-
-    virtual const ParamGeneratorInterface<ParamType>* BaseGenerator() const {
-      return base_;
-    }
-    // Advance should not be called on beyond-of-range iterators
-    // so no component iterators must be beyond end of range, either.
-    virtual void Advance() {
-      assert(!AtEnd());
-      ++current10_;
-      if (current10_ == end10_) {
-        current10_ = begin10_;
-        ++current9_;
-      }
-      if (current9_ == end9_) {
-        current9_ = begin9_;
-        ++current8_;
-      }
-      if (current8_ == end8_) {
-        current8_ = begin8_;
-        ++current7_;
-      }
-      if (current7_ == end7_) {
-        current7_ = begin7_;
-        ++current6_;
-      }
-      if (current6_ == end6_) {
-        current6_ = begin6_;
-        ++current5_;
-      }
-      if (current5_ == end5_) {
-        current5_ = begin5_;
-        ++current4_;
-      }
-      if (current4_ == end4_) {
-        current4_ = begin4_;
-        ++current3_;
-      }
-      if (current3_ == end3_) {
-        current3_ = begin3_;
-        ++current2_;
-      }
-      if (current2_ == end2_) {
-        current2_ = begin2_;
-        ++current1_;
-      }
-      ComputeCurrentValue();
-    }
-    virtual ParamIteratorInterface<ParamType>* Clone() const {
-      return new Iterator(*this);
-    }
-    virtual const ParamType* Current() const { return current_value_.get(); }
-    virtual bool Equals(const ParamIteratorInterface<ParamType>& other) const {
-      // Having the same base generator guarantees that the other
-      // iterator is of the same type and we can downcast.
-      GTEST_CHECK_(BaseGenerator() == other.BaseGenerator())
-          << "The program attempted to compare iterators "
-          << "from different generators." << std::endl;
-      const Iterator* typed_other =
-          CheckedDowncastToActualType<const Iterator>(&other);
-      // We must report iterators equal if they both point beyond their
-      // respective ranges. That can happen in a variety of fashions,
-      // so we have to consult AtEnd().
-      return (AtEnd() && typed_other->AtEnd()) ||
-         (
-          current1_ == typed_other->current1_ &&
-          current2_ == typed_other->current2_ &&
-          current3_ == typed_other->current3_ &&
-          current4_ == typed_other->current4_ &&
-          current5_ == typed_other->current5_ &&
-          current6_ == typed_other->current6_ &&
-          current7_ == typed_other->current7_ &&
-          current8_ == typed_other->current8_ &&
-          current9_ == typed_other->current9_ &&
-          current10_ == typed_other->current10_);
-    }
-
-   private:
-    Iterator(const Iterator& other)
-        : base_(other.base_),
-        begin1_(other.begin1_),
-        end1_(other.end1_),
-        current1_(other.current1_),
-        begin2_(other.begin2_),
-        end2_(other.end2_),
-        current2_(other.current2_),
-        begin3_(other.begin3_),
-        end3_(other.end3_),
-        current3_(other.current3_),
-        begin4_(other.begin4_),
-        end4_(other.end4_),
-        current4_(other.current4_),
-        begin5_(other.begin5_),
-        end5_(other.end5_),
-        current5_(other.current5_),
-        begin6_(other.begin6_),
-        end6_(other.end6_),
-        current6_(other.current6_),
-        begin7_(other.begin7_),
-        end7_(other.end7_),
-        current7_(other.current7_),
-        begin8_(other.begin8_),
-        end8_(other.end8_),
-        current8_(other.current8_),
-        begin9_(other.begin9_),
-        end9_(other.end9_),
-        current9_(other.current9_),
-        begin10_(other.begin10_),
-        end10_(other.end10_),
-        current10_(other.current10_) {
-      ComputeCurrentValue();
-    }
-
-    void ComputeCurrentValue() {
-      if (!AtEnd())
-        current_value_.reset(new ParamType(*current1_, *current2_, *current3_,
-            *current4_, *current5_, *current6_, *current7_, *current8_,
-            *current9_, *current10_));
-    }
-    bool AtEnd() const {
-      // We must report iterator past the end of the range when either of the
-      // component iterators has reached the end of its range.
-      return
-          current1_ == end1_ ||
-          current2_ == end2_ ||
-          current3_ == end3_ ||
-          current4_ == end4_ ||
-          current5_ == end5_ ||
-          current6_ == end6_ ||
-          current7_ == end7_ ||
-          current8_ == end8_ ||
-          current9_ == end9_ ||
-          current10_ == end10_;
-    }
-
-    // No implementation - assignment is unsupported.
-    void operator=(const Iterator& other);
-
-    const ParamGeneratorInterface<ParamType>* const base_;
-    // begin[i]_ and end[i]_ define the i-th range that Iterator traverses.
-    // current[i]_ is the actual traversing iterator.
-    const typename ParamGenerator<T1>::iterator begin1_;
-    const typename ParamGenerator<T1>::iterator end1_;
-    typename ParamGenerator<T1>::iterator current1_;
-    const typename ParamGenerator<T2>::iterator begin2_;
-    const typename ParamGenerator<T2>::iterator end2_;
-    typename ParamGenerator<T2>::iterator current2_;
-    const typename ParamGenerator<T3>::iterator begin3_;
-    const typename ParamGenerator<T3>::iterator end3_;
-    typename ParamGenerator<T3>::iterator current3_;
-    const typename ParamGenerator<T4>::iterator begin4_;
-    const typename ParamGenerator<T4>::iterator end4_;
-    typename ParamGenerator<T4>::iterator current4_;
-    const typename ParamGenerator<T5>::iterator begin5_;
-    const typename ParamGenerator<T5>::iterator end5_;
-    typename ParamGenerator<T5>::iterator current5_;
-    const typename ParamGenerator<T6>::iterator begin6_;
-    const typename ParamGenerator<T6>::iterator end6_;
-    typename ParamGenerator<T6>::iterator current6_;
-    const typename ParamGenerator<T7>::iterator begin7_;
-    const typename ParamGenerator<T7>::iterator end7_;
-    typename ParamGenerator<T7>::iterator current7_;
-    const typename ParamGenerator<T8>::iterator begin8_;
-    const typename ParamGenerator<T8>::iterator end8_;
-    typename ParamGenerator<T8>::iterator current8_;
-    const typename ParamGenerator<T9>::iterator begin9_;
-    const typename ParamGenerator<T9>::iterator end9_;
-    typename ParamGenerator<T9>::iterator current9_;
-    const typename ParamGenerator<T10>::iterator begin10_;
-    const typename ParamGenerator<T10>::iterator end10_;
-    typename ParamGenerator<T10>::iterator current10_;
-    linked_ptr<ParamType> current_value_;
-  };  // class CartesianProductGenerator10::Iterator
-
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductGenerator10& other);
-
-  const ParamGenerator<T1> g1_;
-  const ParamGenerator<T2> g2_;
-  const ParamGenerator<T3> g3_;
-  const ParamGenerator<T4> g4_;
-  const ParamGenerator<T5> g5_;
-  const ParamGenerator<T6> g6_;
-  const ParamGenerator<T7> g7_;
-  const ParamGenerator<T8> g8_;
-  const ParamGenerator<T9> g9_;
-  const ParamGenerator<T10> g10_;
-};  // class CartesianProductGenerator10
-
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// Helper classes providing Combine() with polymorphic features. They allow
-// casting CartesianProductGeneratorN<T> to ParamGenerator<U> if T is
-// convertible to U.
-//
-template <class Generator1, class Generator2>
-class CartesianProductHolder2 {
- public:
-CartesianProductHolder2(const Generator1& g1, const Generator2& g2)
-      : g1_(g1), g2_(g2) {}
-  template <typename T1, typename T2>
-  operator ParamGenerator< ::testing::tuple<T1, T2> >() const {
-    return ParamGenerator< ::testing::tuple<T1, T2> >(
-        new CartesianProductGenerator2<T1, T2>(
-        static_cast<ParamGenerator<T1> >(g1_),
-        static_cast<ParamGenerator<T2> >(g2_)));
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductHolder2& other);
-
-  const Generator1 g1_;
-  const Generator2 g2_;
-};  // class CartesianProductHolder2
-
-template <class Generator1, class Generator2, class Generator3>
-class CartesianProductHolder3 {
- public:
-CartesianProductHolder3(const Generator1& g1, const Generator2& g2,
-    const Generator3& g3)
-      : g1_(g1), g2_(g2), g3_(g3) {}
-  template <typename T1, typename T2, typename T3>
-  operator ParamGenerator< ::testing::tuple<T1, T2, T3> >() const {
-    return ParamGenerator< ::testing::tuple<T1, T2, T3> >(
-        new CartesianProductGenerator3<T1, T2, T3>(
-        static_cast<ParamGenerator<T1> >(g1_),
-        static_cast<ParamGenerator<T2> >(g2_),
-        static_cast<ParamGenerator<T3> >(g3_)));
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductHolder3& other);
-
-  const Generator1 g1_;
-  const Generator2 g2_;
-  const Generator3 g3_;
-};  // class CartesianProductHolder3
-
-template <class Generator1, class Generator2, class Generator3,
-    class Generator4>
-class CartesianProductHolder4 {
- public:
-CartesianProductHolder4(const Generator1& g1, const Generator2& g2,
-    const Generator3& g3, const Generator4& g4)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4) {}
-  template <typename T1, typename T2, typename T3, typename T4>
-  operator ParamGenerator< ::testing::tuple<T1, T2, T3, T4> >() const {
-    return ParamGenerator< ::testing::tuple<T1, T2, T3, T4> >(
-        new CartesianProductGenerator4<T1, T2, T3, T4>(
-        static_cast<ParamGenerator<T1> >(g1_),
-        static_cast<ParamGenerator<T2> >(g2_),
-        static_cast<ParamGenerator<T3> >(g3_),
-        static_cast<ParamGenerator<T4> >(g4_)));
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductHolder4& other);
-
-  const Generator1 g1_;
-  const Generator2 g2_;
-  const Generator3 g3_;
-  const Generator4 g4_;
-};  // class CartesianProductHolder4
-
-template <class Generator1, class Generator2, class Generator3,
-    class Generator4, class Generator5>
-class CartesianProductHolder5 {
- public:
-CartesianProductHolder5(const Generator1& g1, const Generator2& g2,
-    const Generator3& g3, const Generator4& g4, const Generator5& g5)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5) {}
-  template <typename T1, typename T2, typename T3, typename T4, typename T5>
-  operator ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5> >() const {
-    return ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5> >(
-        new CartesianProductGenerator5<T1, T2, T3, T4, T5>(
-        static_cast<ParamGenerator<T1> >(g1_),
-        static_cast<ParamGenerator<T2> >(g2_),
-        static_cast<ParamGenerator<T3> >(g3_),
-        static_cast<ParamGenerator<T4> >(g4_),
-        static_cast<ParamGenerator<T5> >(g5_)));
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductHolder5& other);
-
-  const Generator1 g1_;
-  const Generator2 g2_;
-  const Generator3 g3_;
-  const Generator4 g4_;
-  const Generator5 g5_;
-};  // class CartesianProductHolder5
-
-template <class Generator1, class Generator2, class Generator3,
-    class Generator4, class Generator5, class Generator6>
-class CartesianProductHolder6 {
- public:
-CartesianProductHolder6(const Generator1& g1, const Generator2& g2,
-    const Generator3& g3, const Generator4& g4, const Generator5& g5,
-    const Generator6& g6)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5), g6_(g6) {}
-  template <typename T1, typename T2, typename T3, typename T4, typename T5,
-      typename T6>
-  operator ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5, T6> >() const {
-    return ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5, T6> >(
-        new CartesianProductGenerator6<T1, T2, T3, T4, T5, T6>(
-        static_cast<ParamGenerator<T1> >(g1_),
-        static_cast<ParamGenerator<T2> >(g2_),
-        static_cast<ParamGenerator<T3> >(g3_),
-        static_cast<ParamGenerator<T4> >(g4_),
-        static_cast<ParamGenerator<T5> >(g5_),
-        static_cast<ParamGenerator<T6> >(g6_)));
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductHolder6& other);
-
-  const Generator1 g1_;
-  const Generator2 g2_;
-  const Generator3 g3_;
-  const Generator4 g4_;
-  const Generator5 g5_;
-  const Generator6 g6_;
-};  // class CartesianProductHolder6
-
-template <class Generator1, class Generator2, class Generator3,
-    class Generator4, class Generator5, class Generator6, class Generator7>
-class CartesianProductHolder7 {
- public:
-CartesianProductHolder7(const Generator1& g1, const Generator2& g2,
-    const Generator3& g3, const Generator4& g4, const Generator5& g5,
-    const Generator6& g6, const Generator7& g7)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5), g6_(g6), g7_(g7) {}
-  template <typename T1, typename T2, typename T3, typename T4, typename T5,
-      typename T6, typename T7>
-  operator ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5, T6,
-      T7> >() const {
-    return ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5, T6, T7> >(
-        new CartesianProductGenerator7<T1, T2, T3, T4, T5, T6, T7>(
-        static_cast<ParamGenerator<T1> >(g1_),
-        static_cast<ParamGenerator<T2> >(g2_),
-        static_cast<ParamGenerator<T3> >(g3_),
-        static_cast<ParamGenerator<T4> >(g4_),
-        static_cast<ParamGenerator<T5> >(g5_),
-        static_cast<ParamGenerator<T6> >(g6_),
-        static_cast<ParamGenerator<T7> >(g7_)));
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductHolder7& other);
-
-  const Generator1 g1_;
-  const Generator2 g2_;
-  const Generator3 g3_;
-  const Generator4 g4_;
-  const Generator5 g5_;
-  const Generator6 g6_;
-  const Generator7 g7_;
-};  // class CartesianProductHolder7
-
-template <class Generator1, class Generator2, class Generator3,
-    class Generator4, class Generator5, class Generator6, class Generator7,
-    class Generator8>
-class CartesianProductHolder8 {
- public:
-CartesianProductHolder8(const Generator1& g1, const Generator2& g2,
-    const Generator3& g3, const Generator4& g4, const Generator5& g5,
-    const Generator6& g6, const Generator7& g7, const Generator8& g8)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5), g6_(g6), g7_(g7),
-          g8_(g8) {}
-  template <typename T1, typename T2, typename T3, typename T4, typename T5,
-      typename T6, typename T7, typename T8>
-  operator ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5, T6, T7,
-      T8> >() const {
-    return ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5, T6, T7, T8> >(
-        new CartesianProductGenerator8<T1, T2, T3, T4, T5, T6, T7, T8>(
-        static_cast<ParamGenerator<T1> >(g1_),
-        static_cast<ParamGenerator<T2> >(g2_),
-        static_cast<ParamGenerator<T3> >(g3_),
-        static_cast<ParamGenerator<T4> >(g4_),
-        static_cast<ParamGenerator<T5> >(g5_),
-        static_cast<ParamGenerator<T6> >(g6_),
-        static_cast<ParamGenerator<T7> >(g7_),
-        static_cast<ParamGenerator<T8> >(g8_)));
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductHolder8& other);
-
-  const Generator1 g1_;
-  const Generator2 g2_;
-  const Generator3 g3_;
-  const Generator4 g4_;
-  const Generator5 g5_;
-  const Generator6 g6_;
-  const Generator7 g7_;
-  const Generator8 g8_;
-};  // class CartesianProductHolder8
-
-template <class Generator1, class Generator2, class Generator3,
-    class Generator4, class Generator5, class Generator6, class Generator7,
-    class Generator8, class Generator9>
-class CartesianProductHolder9 {
- public:
-CartesianProductHolder9(const Generator1& g1, const Generator2& g2,
-    const Generator3& g3, const Generator4& g4, const Generator5& g5,
-    const Generator6& g6, const Generator7& g7, const Generator8& g8,
-    const Generator9& g9)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5), g6_(g6), g7_(g7), g8_(g8),
-          g9_(g9) {}
-  template <typename T1, typename T2, typename T3, typename T4, typename T5,
-      typename T6, typename T7, typename T8, typename T9>
-  operator ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5, T6, T7, T8,
-      T9> >() const {
-    return ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5, T6, T7, T8,
-        T9> >(
-        new CartesianProductGenerator9<T1, T2, T3, T4, T5, T6, T7, T8, T9>(
-        static_cast<ParamGenerator<T1> >(g1_),
-        static_cast<ParamGenerator<T2> >(g2_),
-        static_cast<ParamGenerator<T3> >(g3_),
-        static_cast<ParamGenerator<T4> >(g4_),
-        static_cast<ParamGenerator<T5> >(g5_),
-        static_cast<ParamGenerator<T6> >(g6_),
-        static_cast<ParamGenerator<T7> >(g7_),
-        static_cast<ParamGenerator<T8> >(g8_),
-        static_cast<ParamGenerator<T9> >(g9_)));
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductHolder9& other);
-
-  const Generator1 g1_;
-  const Generator2 g2_;
-  const Generator3 g3_;
-  const Generator4 g4_;
-  const Generator5 g5_;
-  const Generator6 g6_;
-  const Generator7 g7_;
-  const Generator8 g8_;
-  const Generator9 g9_;
-};  // class CartesianProductHolder9
-
-template <class Generator1, class Generator2, class Generator3,
-    class Generator4, class Generator5, class Generator6, class Generator7,
-    class Generator8, class Generator9, class Generator10>
-class CartesianProductHolder10 {
- public:
-CartesianProductHolder10(const Generator1& g1, const Generator2& g2,
-    const Generator3& g3, const Generator4& g4, const Generator5& g5,
-    const Generator6& g6, const Generator7& g7, const Generator8& g8,
-    const Generator9& g9, const Generator10& g10)
-      : g1_(g1), g2_(g2), g3_(g3), g4_(g4), g5_(g5), g6_(g6), g7_(g7), g8_(g8),
-          g9_(g9), g10_(g10) {}
-  template <typename T1, typename T2, typename T3, typename T4, typename T5,
-      typename T6, typename T7, typename T8, typename T9, typename T10>
-  operator ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5, T6, T7, T8, T9,
-      T10> >() const {
-    return ParamGenerator< ::testing::tuple<T1, T2, T3, T4, T5, T6, T7, T8, T9,
-        T10> >(
-        new CartesianProductGenerator10<T1, T2, T3, T4, T5, T6, T7, T8, T9,
-            T10>(
-        static_cast<ParamGenerator<T1> >(g1_),
-        static_cast<ParamGenerator<T2> >(g2_),
-        static_cast<ParamGenerator<T3> >(g3_),
-        static_cast<ParamGenerator<T4> >(g4_),
-        static_cast<ParamGenerator<T5> >(g5_),
-        static_cast<ParamGenerator<T6> >(g6_),
-        static_cast<ParamGenerator<T7> >(g7_),
-        static_cast<ParamGenerator<T8> >(g8_),
-        static_cast<ParamGenerator<T9> >(g9_),
-        static_cast<ParamGenerator<T10> >(g10_)));
-  }
-
- private:
-  // No implementation - assignment is unsupported.
-  void operator=(const CartesianProductHolder10& other);
-
-  const Generator1 g1_;
-  const Generator2 g2_;
-  const Generator3 g3_;
-  const Generator4 g4_;
-  const Generator5 g5_;
-  const Generator6 g6_;
-  const Generator7 g7_;
-  const Generator8 g8_;
-  const Generator9 g9_;
-  const Generator10 g10_;
-};  // class CartesianProductHolder10
-
-# endif  // GTEST_HAS_COMBINE
-
-}  // namespace internal
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PARAM_UTIL_GENERATED_H_
-
-namespace testing {
-
-// Functions producing parameter generators.
-//
-// Google Test uses these generators to produce parameters for value-
-// parameterized tests. When a parameterized test case is instantiated
-// with a particular generator, Google Test creates and runs tests
-// for each element in the sequence produced by the generator.
-//
-// In the following sample, tests from test case FooTest are instantiated
-// each three times with parameter values 3, 5, and 8:
-//
-// class FooTest : public TestWithParam<int> { ... };
-//
-// TEST_P(FooTest, TestThis) {
-// }
-// TEST_P(FooTest, TestThat) {
-// }
-// INSTANTIATE_TEST_CASE_P(TestSequence, FooTest, Values(3, 5, 8));
-//
-
-// Range() returns generators providing sequences of values in a range.
-//
-// Synopsis:
-// Range(start, end)
-//   - returns a generator producing a sequence of values {start, start+1,
-//     start+2, ..., }.
-// Range(start, end, step)
-//   - returns a generator producing a sequence of values {start, start+step,
-//     start+step+step, ..., }.
-// Notes:
-//   * The generated sequences never include end. For example, Range(1, 5)
-//     returns a generator producing a sequence {1, 2, 3, 4}. Range(1, 9, 2)
-//     returns a generator producing {1, 3, 5, 7}.
-//   * start and end must have the same type. That type may be any integral or
-//     floating-point type or a user defined type satisfying these conditions:
-//     * It must be assignable (have operator=() defined).
-//     * It must have operator+() (operator+(int-compatible type) for
-//       two-operand version).
-//     * It must have operator<() defined.
-//     Elements in the resulting sequences will also have that type.
-//   * Condition start < end must be satisfied in order for resulting sequences
-//     to contain any elements.
-//
-template <typename T, typename IncrementT>
-internal::ParamGenerator<T> Range(T start, T end, IncrementT step) {
-  return internal::ParamGenerator<T>(
-      new internal::RangeGenerator<T, IncrementT>(start, end, step));
-}
-
-template <typename T>
-internal::ParamGenerator<T> Range(T start, T end) {
-  return Range(start, end, 1);
-}
-
-// ValuesIn() function allows generation of tests with parameters coming from
-// a container.
-//
-// Synopsis:
-// ValuesIn(const T (&array)[N])
-//   - returns a generator producing sequences with elements from
-//     a C-style array.
-// ValuesIn(const Container& container)
-//   - returns a generator producing sequences with elements from
-//     an STL-style container.
-// ValuesIn(Iterator begin, Iterator end)
-//   - returns a generator producing sequences with elements from
-//     a range [begin, end) defined by a pair of STL-style iterators. These
-//     iterators can also be plain C pointers.
-//
-// Please note that ValuesIn copies the values from the containers
-// passed in and keeps them to generate tests in RUN_ALL_TESTS().
-//
-// Examples:
-//
-// This instantiates tests from test case StringTest
-// each with C-string values of "foo", "bar", and "baz":
-//
-// const char* strings[] = {"foo", "bar", "baz"};
-// INSTANTIATE_TEST_CASE_P(StringSequence, StringTest, ValuesIn(strings));
-//
-// This instantiates tests from test case StlStringTest
-// each with STL strings with values "a" and "b":
-//
-// ::std::vector< ::std::string> GetParameterStrings() {
-//   ::std::vector< ::std::string> v;
-//   v.push_back("a");
-//   v.push_back("b");
-//   return v;
-// }
-//
-// INSTANTIATE_TEST_CASE_P(CharSequence,
-//                         StlStringTest,
-//                         ValuesIn(GetParameterStrings()));
-//
-//
-// This will also instantiate tests from CharTest
-// each with parameter values 'a' and 'b':
-//
-// ::std::list<char> GetParameterChars() {
-//   ::std::list<char> list;
-//   list.push_back('a');
-//   list.push_back('b');
-//   return list;
-// }
-// ::std::list<char> l = GetParameterChars();
-// INSTANTIATE_TEST_CASE_P(CharSequence2,
-//                         CharTest,
-//                         ValuesIn(l.begin(), l.end()));
-//
-template <typename ForwardIterator>
-internal::ParamGenerator<
-  typename ::testing::internal::IteratorTraits<ForwardIterator>::value_type>
-ValuesIn(ForwardIterator begin, ForwardIterator end) {
-  typedef typename ::testing::internal::IteratorTraits<ForwardIterator>
-      ::value_type ParamType;
-  return internal::ParamGenerator<ParamType>(
-      new internal::ValuesInIteratorRangeGenerator<ParamType>(begin, end));
-}
-
-template <typename T, size_t N>
-internal::ParamGenerator<T> ValuesIn(const T (&array)[N]) {
-  return ValuesIn(array, array + N);
-}
-
-template <class Container>
-internal::ParamGenerator<typename Container::value_type> ValuesIn(
-    const Container& container) {
-  return ValuesIn(container.begin(), container.end());
-}
-
-// Values() allows generating tests from explicitly specified list of
-// parameters.
-//
-// Synopsis:
-// Values(T v1, T v2, ..., T vN)
-//   - returns a generator producing sequences with elements v1, v2, ..., vN.
-//
-// For example, this instantiates tests from test case BarTest each
-// with values "one", "two", and "three":
-//
-// INSTANTIATE_TEST_CASE_P(NumSequence, BarTest, Values("one", "two", "three"));
-//
-// This instantiates tests from test case BazTest each with values 1, 2, 3.5.
-// The exact type of values will depend on the type of parameter in BazTest.
-//
-// INSTANTIATE_TEST_CASE_P(FloatingNumbers, BazTest, Values(1, 2, 3.5));
-//
-// Currently, Values() supports from 1 to 50 parameters.
-//
-template <typename T1>
-internal::ValueArray1<T1> Values(T1 v1) {
-  return internal::ValueArray1<T1>(v1);
-}
-
-template <typename T1, typename T2>
-internal::ValueArray2<T1, T2> Values(T1 v1, T2 v2) {
-  return internal::ValueArray2<T1, T2>(v1, v2);
-}
-
-template <typename T1, typename T2, typename T3>
-internal::ValueArray3<T1, T2, T3> Values(T1 v1, T2 v2, T3 v3) {
-  return internal::ValueArray3<T1, T2, T3>(v1, v2, v3);
-}
-
-template <typename T1, typename T2, typename T3, typename T4>
-internal::ValueArray4<T1, T2, T3, T4> Values(T1 v1, T2 v2, T3 v3, T4 v4) {
-  return internal::ValueArray4<T1, T2, T3, T4>(v1, v2, v3, v4);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5>
-internal::ValueArray5<T1, T2, T3, T4, T5> Values(T1 v1, T2 v2, T3 v3, T4 v4,
-    T5 v5) {
-  return internal::ValueArray5<T1, T2, T3, T4, T5>(v1, v2, v3, v4, v5);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6>
-internal::ValueArray6<T1, T2, T3, T4, T5, T6> Values(T1 v1, T2 v2, T3 v3,
-    T4 v4, T5 v5, T6 v6) {
-  return internal::ValueArray6<T1, T2, T3, T4, T5, T6>(v1, v2, v3, v4, v5, v6);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7>
-internal::ValueArray7<T1, T2, T3, T4, T5, T6, T7> Values(T1 v1, T2 v2, T3 v3,
-    T4 v4, T5 v5, T6 v6, T7 v7) {
-  return internal::ValueArray7<T1, T2, T3, T4, T5, T6, T7>(v1, v2, v3, v4, v5,
-      v6, v7);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8>
-internal::ValueArray8<T1, T2, T3, T4, T5, T6, T7, T8> Values(T1 v1, T2 v2,
-    T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8) {
-  return internal::ValueArray8<T1, T2, T3, T4, T5, T6, T7, T8>(v1, v2, v3, v4,
-      v5, v6, v7, v8);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9>
-internal::ValueArray9<T1, T2, T3, T4, T5, T6, T7, T8, T9> Values(T1 v1, T2 v2,
-    T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9) {
-  return internal::ValueArray9<T1, T2, T3, T4, T5, T6, T7, T8, T9>(v1, v2, v3,
-      v4, v5, v6, v7, v8, v9);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10>
-internal::ValueArray10<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10> Values(T1 v1,
-    T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10) {
-  return internal::ValueArray10<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10>(v1,
-      v2, v3, v4, v5, v6, v7, v8, v9, v10);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11>
-internal::ValueArray11<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10,
-    T11> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11) {
-  return internal::ValueArray11<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10,
-      T11>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12>
-internal::ValueArray12<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-    T12> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12) {
-  return internal::ValueArray12<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13>
-internal::ValueArray13<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12,
-    T13> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12, T13 v13) {
-  return internal::ValueArray13<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14>
-internal::ValueArray14<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12, T13 v13, T14 v14) {
-  return internal::ValueArray14<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13,
-      v14);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15>
-internal::ValueArray15<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8,
-    T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15) {
-  return internal::ValueArray15<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12,
-      v13, v14, v15);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16>
-internal::ValueArray16<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7,
-    T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15,
-    T16 v16) {
-  return internal::ValueArray16<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11,
-      v12, v13, v14, v15, v16);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17>
-internal::ValueArray17<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7,
-    T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15,
-    T16 v16, T17 v17) {
-  return internal::ValueArray17<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10,
-      v11, v12, v13, v14, v15, v16, v17);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18>
-internal::ValueArray18<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6,
-    T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15,
-    T16 v16, T17 v17, T18 v18) {
-  return internal::ValueArray18<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18>(v1, v2, v3, v4, v5, v6, v7, v8, v9,
-      v10, v11, v12, v13, v14, v15, v16, v17, v18);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19>
-internal::ValueArray19<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5,
-    T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14,
-    T15 v15, T16 v16, T17 v17, T18 v18, T19 v19) {
-  return internal::ValueArray19<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19>(v1, v2, v3, v4, v5, v6, v7, v8,
-      v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20>
-internal::ValueArray20<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20> Values(T1 v1, T2 v2, T3 v3, T4 v4,
-    T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13,
-    T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20) {
-  return internal::ValueArray20<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20>(v1, v2, v3, v4, v5, v6, v7,
-      v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21>
-internal::ValueArray21<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21> Values(T1 v1, T2 v2, T3 v3, T4 v4,
-    T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13,
-    T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21) {
-  return internal::ValueArray21<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21>(v1, v2, v3, v4, v5, v6,
-      v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22>
-internal::ValueArray22<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22> Values(T1 v1, T2 v2, T3 v3,
-    T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12,
-    T13 v13, T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20,
-    T21 v21, T22 v22) {
-  return internal::ValueArray22<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22>(v1, v2, v3, v4,
-      v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19,
-      v20, v21, v22);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23>
-internal::ValueArray23<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23> Values(T1 v1, T2 v2,
-    T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12,
-    T13 v13, T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20,
-    T21 v21, T22 v22, T23 v23) {
-  return internal::ValueArray23<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23>(v1, v2, v3,
-      v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19,
-      v20, v21, v22, v23);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24>
-internal::ValueArray24<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24> Values(T1 v1, T2 v2,
-    T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12,
-    T13 v13, T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20,
-    T21 v21, T22 v22, T23 v23, T24 v24) {
-  return internal::ValueArray24<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24>(v1, v2,
-      v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18,
-      v19, v20, v21, v22, v23, v24);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25>
-internal::ValueArray25<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25> Values(T1 v1,
-    T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11,
-    T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19,
-    T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25) {
-  return internal::ValueArray25<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25>(v1,
-      v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17,
-      v18, v19, v20, v21, v22, v23, v24, v25);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26>
-internal::ValueArray26<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-    T26> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-    T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-    T26 v26) {
-  return internal::ValueArray26<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15,
-      v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27>
-internal::ValueArray27<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26,
-    T27> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-    T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-    T26 v26, T27 v27) {
-  return internal::ValueArray27<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14,
-      v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28>
-internal::ValueArray28<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27,
-    T28> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-    T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-    T26 v26, T27 v27, T28 v28) {
-  return internal::ValueArray28<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13,
-      v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27,
-      v28);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29>
-internal::ValueArray29<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-    T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-    T26 v26, T27 v27, T28 v28, T29 v29) {
-  return internal::ValueArray29<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12,
-      v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26,
-      v27, v28, v29);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30>
-internal::ValueArray30<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8,
-    T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16,
-    T17 v17, T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24,
-    T25 v25, T26 v26, T27 v27, T28 v28, T29 v29, T30 v30) {
-  return internal::ValueArray30<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11,
-      v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25,
-      v26, v27, v28, v29, v30);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31>
-internal::ValueArray31<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7,
-    T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15,
-    T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23,
-    T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31) {
-  return internal::ValueArray31<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10,
-      v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24,
-      v25, v26, v27, v28, v29, v30, v31);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32>
-internal::ValueArray32<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7,
-    T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15,
-    T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23,
-    T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31,
-    T32 v32) {
-  return internal::ValueArray32<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32>(v1, v2, v3, v4, v5, v6, v7, v8, v9,
-      v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23,
-      v24, v25, v26, v27, v28, v29, v30, v31, v32);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33>
-internal::ValueArray33<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6,
-    T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15,
-    T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23,
-    T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31,
-    T32 v32, T33 v33) {
-  return internal::ValueArray33<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33>(v1, v2, v3, v4, v5, v6, v7, v8,
-      v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23,
-      v24, v25, v26, v27, v28, v29, v30, v31, v32, v33);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34>
-internal::ValueArray34<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5,
-    T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14,
-    T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21, T22 v22,
-    T23 v23, T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29, T30 v30,
-    T31 v31, T32 v32, T33 v33, T34 v34) {
-  return internal::ValueArray34<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34>(v1, v2, v3, v4, v5, v6, v7,
-      v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22,
-      v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35>
-internal::ValueArray35<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35> Values(T1 v1, T2 v2, T3 v3, T4 v4,
-    T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13,
-    T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21,
-    T22 v22, T23 v23, T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29,
-    T30 v30, T31 v31, T32 v32, T33 v33, T34 v34, T35 v35) {
-  return internal::ValueArray35<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35>(v1, v2, v3, v4, v5, v6,
-      v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21,
-      v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36>
-internal::ValueArray36<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36> Values(T1 v1, T2 v2, T3 v3, T4 v4,
-    T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13,
-    T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21,
-    T22 v22, T23 v23, T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29,
-    T30 v30, T31 v31, T32 v32, T33 v33, T34 v34, T35 v35, T36 v36) {
-  return internal::ValueArray36<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36>(v1, v2, v3, v4,
-      v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19,
-      v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33,
-      v34, v35, v36);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37>
-internal::ValueArray37<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37> Values(T1 v1, T2 v2, T3 v3,
-    T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12,
-    T13 v13, T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20,
-    T21 v21, T22 v22, T23 v23, T24 v24, T25 v25, T26 v26, T27 v27, T28 v28,
-    T29 v29, T30 v30, T31 v31, T32 v32, T33 v33, T34 v34, T35 v35, T36 v36,
-    T37 v37) {
-  return internal::ValueArray37<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37>(v1, v2, v3,
-      v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19,
-      v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33,
-      v34, v35, v36, v37);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38>
-internal::ValueArray38<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38> Values(T1 v1, T2 v2,
-    T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12,
-    T13 v13, T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20,
-    T21 v21, T22 v22, T23 v23, T24 v24, T25 v25, T26 v26, T27 v27, T28 v28,
-    T29 v29, T30 v30, T31 v31, T32 v32, T33 v33, T34 v34, T35 v35, T36 v36,
-    T37 v37, T38 v38) {
-  return internal::ValueArray38<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38>(v1, v2,
-      v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18,
-      v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32,
-      v33, v34, v35, v36, v37, v38);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39>
-internal::ValueArray39<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39> Values(T1 v1, T2 v2,
-    T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12,
-    T13 v13, T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20,
-    T21 v21, T22 v22, T23 v23, T24 v24, T25 v25, T26 v26, T27 v27, T28 v28,
-    T29 v29, T30 v30, T31 v31, T32 v32, T33 v33, T34 v34, T35 v35, T36 v36,
-    T37 v37, T38 v38, T39 v39) {
-  return internal::ValueArray39<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39>(v1,
-      v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17,
-      v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31,
-      v32, v33, v34, v35, v36, v37, v38, v39);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40>
-internal::ValueArray40<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40> Values(T1 v1,
-    T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11,
-    T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19,
-    T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25, T26 v26, T27 v27,
-    T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33, T34 v34, T35 v35,
-    T36 v36, T37 v37, T38 v38, T39 v39, T40 v40) {
-  return internal::ValueArray40<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15,
-      v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29,
-      v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41>
-internal::ValueArray41<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40,
-    T41> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-    T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-    T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-    T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41) {
-  return internal::ValueArray41<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40, T41>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14,
-      v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28,
-      v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42>
-internal::ValueArray42<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41,
-    T42> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-    T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-    T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-    T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-    T42 v42) {
-  return internal::ValueArray42<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40, T41, T42>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13,
-      v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27,
-      v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41,
-      v42);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43>
-internal::ValueArray43<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42,
-    T43> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-    T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-    T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-    T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-    T42 v42, T43 v43) {
-  return internal::ValueArray43<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40, T41, T42, T43>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12,
-      v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26,
-      v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40,
-      v41, v42, v43);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44>
-internal::ValueArray44<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-    T44> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8, T9 v9,
-    T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16, T17 v17,
-    T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24, T25 v25,
-    T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32, T33 v33,
-    T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40, T41 v41,
-    T42 v42, T43 v43, T44 v44) {
-  return internal::ValueArray44<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40, T41, T42, T43, T44>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11,
-      v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25,
-      v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39,
-      v40, v41, v42, v43, v44);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45>
-internal::ValueArray45<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-    T44, T45> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7, T8 v8,
-    T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15, T16 v16,
-    T17 v17, T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23, T24 v24,
-    T25 v25, T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31, T32 v32,
-    T33 v33, T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39, T40 v40,
-    T41 v41, T42 v42, T43 v43, T44 v44, T45 v45) {
-  return internal::ValueArray45<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40, T41, T42, T43, T44, T45>(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10,
-      v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24,
-      v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38,
-      v39, v40, v41, v42, v43, v44, v45);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46>
-internal::ValueArray46<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-    T44, T45, T46> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7,
-    T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15,
-    T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23,
-    T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31,
-    T32 v32, T33 v33, T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39,
-    T40 v40, T41 v41, T42 v42, T43 v43, T44 v44, T45 v45, T46 v46) {
-  return internal::ValueArray46<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40, T41, T42, T43, T44, T45, T46>(v1, v2, v3, v4, v5, v6, v7, v8, v9,
-      v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23,
-      v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37,
-      v38, v39, v40, v41, v42, v43, v44, v45, v46);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47>
-internal::ValueArray47<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-    T44, T45, T46, T47> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6, T7 v7,
-    T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15,
-    T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23,
-    T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31,
-    T32 v32, T33 v33, T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39,
-    T40 v40, T41 v41, T42 v42, T43 v43, T44 v44, T45 v45, T46 v46, T47 v47) {
-  return internal::ValueArray47<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40, T41, T42, T43, T44, T45, T46, T47>(v1, v2, v3, v4, v5, v6, v7, v8,
-      v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23,
-      v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37,
-      v38, v39, v40, v41, v42, v43, v44, v45, v46, v47);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48>
-internal::ValueArray48<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-    T44, T45, T46, T47, T48> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5, T6 v6,
-    T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14, T15 v15,
-    T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21, T22 v22, T23 v23,
-    T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29, T30 v30, T31 v31,
-    T32 v32, T33 v33, T34 v34, T35 v35, T36 v36, T37 v37, T38 v38, T39 v39,
-    T40 v40, T41 v41, T42 v42, T43 v43, T44 v44, T45 v45, T46 v46, T47 v47,
-    T48 v48) {
-  return internal::ValueArray48<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40, T41, T42, T43, T44, T45, T46, T47, T48>(v1, v2, v3, v4, v5, v6, v7,
-      v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22,
-      v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36,
-      v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48, typename T49>
-internal::ValueArray49<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-    T44, T45, T46, T47, T48, T49> Values(T1 v1, T2 v2, T3 v3, T4 v4, T5 v5,
-    T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13, T14 v14,
-    T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21, T22 v22,
-    T23 v23, T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29, T30 v30,
-    T31 v31, T32 v32, T33 v33, T34 v34, T35 v35, T36 v36, T37 v37, T38 v38,
-    T39 v39, T40 v40, T41 v41, T42 v42, T43 v43, T44 v44, T45 v45, T46 v46,
-    T47 v47, T48 v48, T49 v49) {
-  return internal::ValueArray49<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40, T41, T42, T43, T44, T45, T46, T47, T48, T49>(v1, v2, v3, v4, v5, v6,
-      v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21,
-      v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35,
-      v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49);
-}
-
-template <typename T1, typename T2, typename T3, typename T4, typename T5,
-    typename T6, typename T7, typename T8, typename T9, typename T10,
-    typename T11, typename T12, typename T13, typename T14, typename T15,
-    typename T16, typename T17, typename T18, typename T19, typename T20,
-    typename T21, typename T22, typename T23, typename T24, typename T25,
-    typename T26, typename T27, typename T28, typename T29, typename T30,
-    typename T31, typename T32, typename T33, typename T34, typename T35,
-    typename T36, typename T37, typename T38, typename T39, typename T40,
-    typename T41, typename T42, typename T43, typename T44, typename T45,
-    typename T46, typename T47, typename T48, typename T49, typename T50>
-internal::ValueArray50<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
-    T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25, T26, T27, T28,
-    T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39, T40, T41, T42, T43,
-    T44, T45, T46, T47, T48, T49, T50> Values(T1 v1, T2 v2, T3 v3, T4 v4,
-    T5 v5, T6 v6, T7 v7, T8 v8, T9 v9, T10 v10, T11 v11, T12 v12, T13 v13,
-    T14 v14, T15 v15, T16 v16, T17 v17, T18 v18, T19 v19, T20 v20, T21 v21,
-    T22 v22, T23 v23, T24 v24, T25 v25, T26 v26, T27 v27, T28 v28, T29 v29,
-    T30 v30, T31 v31, T32 v32, T33 v33, T34 v34, T35 v35, T36 v36, T37 v37,
-    T38 v38, T39 v39, T40 v40, T41 v41, T42 v42, T43 v43, T44 v44, T45 v45,
-    T46 v46, T47 v47, T48 v48, T49 v49, T50 v50) {
-  return internal::ValueArray50<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11,
-      T12, T13, T14, T15, T16, T17, T18, T19, T20, T21, T22, T23, T24, T25,
-      T26, T27, T28, T29, T30, T31, T32, T33, T34, T35, T36, T37, T38, T39,
-      T40, T41, T42, T43, T44, T45, T46, T47, T48, T49, T50>(v1, v2, v3, v4,
-      v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19,
-      v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33,
-      v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47,
-      v48, v49, v50);
-}
-
-// Bool() allows generating tests with parameters in a set of (false, true).
-//
-// Synopsis:
-// Bool()
-//   - returns a generator producing sequences with elements {false, true}.
-//
-// It is useful when testing code that depends on Boolean flags. Combinations
-// of multiple flags can be tested when several Bool()'s are combined using
-// Combine() function.
-//
-// In the following example all tests in the test case FlagDependentTest
-// will be instantiated twice with parameters false and true.
-//
-// class FlagDependentTest : public testing::TestWithParam<bool> {
-//   virtual void SetUp() {
-//     external_flag = GetParam();
-//   }
-// }
-// INSTANTIATE_TEST_CASE_P(BoolSequence, FlagDependentTest, Bool());
-//
-inline internal::ParamGenerator<bool> Bool() {
-  return Values(false, true);
-}
-
-# if GTEST_HAS_COMBINE
-// Combine() allows the user to combine two or more sequences to produce
-// values of a Cartesian product of those sequences' elements.
-//
-// Synopsis:
-// Combine(gen1, gen2, ..., genN)
-//   - returns a generator producing sequences with elements coming from
-//     the Cartesian product of elements from the sequences generated by
-//     gen1, gen2, ..., genN. The sequence elements will have a type of
-//     tuple<T1, T2, ..., TN> where T1, T2, ..., TN are the types
-//     of elements from sequences produces by gen1, gen2, ..., genN.
-//
-// Combine can have up to 10 arguments. This number is currently limited
-// by the maximum number of elements in the tuple implementation used by Google
-// Test.
-//
-// Example:
-//
-// This will instantiate tests in test case AnimalTest each one with
-// the parameter values tuple("cat", BLACK), tuple("cat", WHITE),
-// tuple("dog", BLACK), and tuple("dog", WHITE):
-//
-// enum Color { BLACK, GRAY, WHITE };
-// class AnimalTest
-//     : public testing::TestWithParam<tuple<const char*, Color> > {...};
-//
-// TEST_P(AnimalTest, AnimalLooksNice) {...}
-//
-// INSTANTIATE_TEST_CASE_P(AnimalVariations, AnimalTest,
-//                         Combine(Values("cat", "dog"),
-//                                 Values(BLACK, WHITE)));
-//
-// This will instantiate tests in FlagDependentTest with all variations of two
-// Boolean flags:
-//
-// class FlagDependentTest
-//     : public testing::TestWithParam<tuple<bool, bool> > {
-//   virtual void SetUp() {
-//     // Assigns external_flag_1 and external_flag_2 values from the tuple.
-//     tie(external_flag_1, external_flag_2) = GetParam();
-//   }
-// };
-//
-// TEST_P(FlagDependentTest, TestFeature1) {
-//   // Test your code using external_flag_1 and external_flag_2 here.
-// }
-// INSTANTIATE_TEST_CASE_P(TwoBoolSequence, FlagDependentTest,
-//                         Combine(Bool(), Bool()));
-//
-template <typename Generator1, typename Generator2>
-internal::CartesianProductHolder2<Generator1, Generator2> Combine(
-    const Generator1& g1, const Generator2& g2) {
-  return internal::CartesianProductHolder2<Generator1, Generator2>(
-      g1, g2);
-}
-
-template <typename Generator1, typename Generator2, typename Generator3>
-internal::CartesianProductHolder3<Generator1, Generator2, Generator3> Combine(
-    const Generator1& g1, const Generator2& g2, const Generator3& g3) {
-  return internal::CartesianProductHolder3<Generator1, Generator2, Generator3>(
-      g1, g2, g3);
-}
-
-template <typename Generator1, typename Generator2, typename Generator3,
-    typename Generator4>
-internal::CartesianProductHolder4<Generator1, Generator2, Generator3,
-    Generator4> Combine(
-    const Generator1& g1, const Generator2& g2, const Generator3& g3,
-        const Generator4& g4) {
-  return internal::CartesianProductHolder4<Generator1, Generator2, Generator3,
-      Generator4>(
-      g1, g2, g3, g4);
-}
-
-template <typename Generator1, typename Generator2, typename Generator3,
-    typename Generator4, typename Generator5>
-internal::CartesianProductHolder5<Generator1, Generator2, Generator3,
-    Generator4, Generator5> Combine(
-    const Generator1& g1, const Generator2& g2, const Generator3& g3,
-        const Generator4& g4, const Generator5& g5) {
-  return internal::CartesianProductHolder5<Generator1, Generator2, Generator3,
-      Generator4, Generator5>(
-      g1, g2, g3, g4, g5);
-}
-
-template <typename Generator1, typename Generator2, typename Generator3,
-    typename Generator4, typename Generator5, typename Generator6>
-internal::CartesianProductHolder6<Generator1, Generator2, Generator3,
-    Generator4, Generator5, Generator6> Combine(
-    const Generator1& g1, const Generator2& g2, const Generator3& g3,
-        const Generator4& g4, const Generator5& g5, const Generator6& g6) {
-  return internal::CartesianProductHolder6<Generator1, Generator2, Generator3,
-      Generator4, Generator5, Generator6>(
-      g1, g2, g3, g4, g5, g6);
-}
-
-template <typename Generator1, typename Generator2, typename Generator3,
-    typename Generator4, typename Generator5, typename Generator6,
-    typename Generator7>
-internal::CartesianProductHolder7<Generator1, Generator2, Generator3,
-    Generator4, Generator5, Generator6, Generator7> Combine(
-    const Generator1& g1, const Generator2& g2, const Generator3& g3,
-        const Generator4& g4, const Generator5& g5, const Generator6& g6,
-        const Generator7& g7) {
-  return internal::CartesianProductHolder7<Generator1, Generator2, Generator3,
-      Generator4, Generator5, Generator6, Generator7>(
-      g1, g2, g3, g4, g5, g6, g7);
-}
-
-template <typename Generator1, typename Generator2, typename Generator3,
-    typename Generator4, typename Generator5, typename Generator6,
-    typename Generator7, typename Generator8>
-internal::CartesianProductHolder8<Generator1, Generator2, Generator3,
-    Generator4, Generator5, Generator6, Generator7, Generator8> Combine(
-    const Generator1& g1, const Generator2& g2, const Generator3& g3,
-        const Generator4& g4, const Generator5& g5, const Generator6& g6,
-        const Generator7& g7, const Generator8& g8) {
-  return internal::CartesianProductHolder8<Generator1, Generator2, Generator3,
-      Generator4, Generator5, Generator6, Generator7, Generator8>(
-      g1, g2, g3, g4, g5, g6, g7, g8);
-}
-
-template <typename Generator1, typename Generator2, typename Generator3,
-    typename Generator4, typename Generator5, typename Generator6,
-    typename Generator7, typename Generator8, typename Generator9>
-internal::CartesianProductHolder9<Generator1, Generator2, Generator3,
-    Generator4, Generator5, Generator6, Generator7, Generator8,
-    Generator9> Combine(
-    const Generator1& g1, const Generator2& g2, const Generator3& g3,
-        const Generator4& g4, const Generator5& g5, const Generator6& g6,
-        const Generator7& g7, const Generator8& g8, const Generator9& g9) {
-  return internal::CartesianProductHolder9<Generator1, Generator2, Generator3,
-      Generator4, Generator5, Generator6, Generator7, Generator8, Generator9>(
-      g1, g2, g3, g4, g5, g6, g7, g8, g9);
-}
-
-template <typename Generator1, typename Generator2, typename Generator3,
-    typename Generator4, typename Generator5, typename Generator6,
-    typename Generator7, typename Generator8, typename Generator9,
-    typename Generator10>
-internal::CartesianProductHolder10<Generator1, Generator2, Generator3,
-    Generator4, Generator5, Generator6, Generator7, Generator8, Generator9,
-    Generator10> Combine(
-    const Generator1& g1, const Generator2& g2, const Generator3& g3,
-        const Generator4& g4, const Generator5& g5, const Generator6& g6,
-        const Generator7& g7, const Generator8& g8, const Generator9& g9,
-        const Generator10& g10) {
-  return internal::CartesianProductHolder10<Generator1, Generator2, Generator3,
-      Generator4, Generator5, Generator6, Generator7, Generator8, Generator9,
-      Generator10>(
-      g1, g2, g3, g4, g5, g6, g7, g8, g9, g10);
-}
-# endif  // GTEST_HAS_COMBINE
-
-# define TEST_P(test_case_name, test_name) \
-  class GTEST_TEST_CLASS_NAME_(test_case_name, test_name) \
-      : public test_case_name { \
-   public: \
-    GTEST_TEST_CLASS_NAME_(test_case_name, test_name)() {} \
-    virtual void TestBody(); \
-   private: \
-    static int AddToRegistry() { \
-      ::testing::UnitTest::GetInstance()->parameterized_test_registry(). \
-          GetTestCasePatternHolder<test_case_name>(\
-              #test_case_name, \
-              ::testing::internal::CodeLocation(\
-                  __FILE__, __LINE__))->AddTestPattern(\
-                      GTEST_STRINGIFY_(test_case_name), \
-                      GTEST_STRINGIFY_(test_name), \
-                      new ::testing::internal::TestMetaFactory< \
-                          GTEST_TEST_CLASS_NAME_(\
-                              test_case_name, test_name)>()); \
-      return 0; \
-    } \
-    static int gtest_registering_dummy_ GTEST_ATTRIBUTE_UNUSED_; \
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(\
-        GTEST_TEST_CLASS_NAME_(test_case_name, test_name)); \
-  }; \
-  int GTEST_TEST_CLASS_NAME_(test_case_name, \
-                             test_name)::gtest_registering_dummy_ = \
-      GTEST_TEST_CLASS_NAME_(test_case_name, test_name)::AddToRegistry(); \
-  void GTEST_TEST_CLASS_NAME_(test_case_name, test_name)::TestBody()
-
-// The optional last argument to INSTANTIATE_TEST_CASE_P allows the user
-// to specify a function or functor that generates custom test name suffixes
-// based on the test parameters. The function should accept one argument of
-// type testing::TestParamInfo<class ParamType>, and return std::string.
-//
-// testing::PrintToStringParamName is a builtin test suffix generator that
-// returns the value of testing::PrintToString(GetParam()).
-//
-// Note: test names must be non-empty, unique, and may only contain ASCII
-// alphanumeric characters or underscore. Because PrintToString adds quotes
-// to std::string and C strings, it won't work for these types.
-
-#define INSTANTIATE_TEST_CASE_P(prefix, test_case_name, generator, ...)        \
-  static ::testing::internal::ParamGenerator<test_case_name::ParamType>        \
-      gtest_##prefix##test_case_name##_EvalGenerator_() {                      \
-    return generator;                                                          \
-  }                                                                            \
-  static ::std::string gtest_##prefix##test_case_name##_EvalGenerateName_(     \
-      const ::testing::TestParamInfo<test_case_name::ParamType>& info) {       \
-    return ::testing::internal::GetParamNameGen<test_case_name::ParamType>(    \
-        __VA_ARGS__)(info);                                                    \
-  }                                                                            \
-  static int gtest_##prefix##test_case_name##_dummy_ GTEST_ATTRIBUTE_UNUSED_ = \
-      ::testing::UnitTest::GetInstance()                                       \
-          ->parameterized_test_registry()                                      \
-          .GetTestCasePatternHolder<test_case_name>(                           \
-              #test_case_name,                                                 \
-              ::testing::internal::CodeLocation(__FILE__, __LINE__))           \
-          ->AddTestCaseInstantiation(                                          \
-              #prefix, &gtest_##prefix##test_case_name##_EvalGenerator_,       \
-              &gtest_##prefix##test_case_name##_EvalGenerateName_, __FILE__,   \
-              __LINE__)
-
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_PARAM_TEST_H_
-// Copyright 2006, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-//
-// Google C++ Testing and Mocking Framework definitions useful in production code.
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_PROD_H_
-#define GTEST_INCLUDE_GTEST_GTEST_PROD_H_
-
-// When you need to test the private or protected members of a class,
-// use the FRIEND_TEST macro to declare your tests as friends of the
-// class.  For example:
-//
-// class MyClass {
-//  private:
-//   void PrivateMethod();
-//   FRIEND_TEST(MyClassTest, PrivateMethodWorks);
-// };
-//
-// class MyClassTest : public testing::Test {
-//   // ...
-// };
-//
-// TEST_F(MyClassTest, PrivateMethodWorks) {
-//   // Can call MyClass::PrivateMethod() here.
-// }
-//
-// Note: The test class must be in the same namespace as the class being tested.
-// For example, putting MyClassTest in an anonymous namespace will not work.
-
-#define FRIEND_TEST(test_case_name, test_name)\
-friend class test_case_name##_##test_name##_Test
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_PROD_H_
-// Copyright 2008, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: mheule@google.com (Markus Heule)
-//
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_TEST_PART_H_
-#define GTEST_INCLUDE_GTEST_GTEST_TEST_PART_H_
-
-#include <iosfwd>
-#include <vector>
-
-namespace testing {
-
-// A copyable object representing the result of a test part (i.e. an
-// assertion or an explicit FAIL(), ADD_FAILURE(), or SUCCESS()).
-//
-// Don't inherit from TestPartResult as its destructor is not virtual.
-class GTEST_API_ TestPartResult {
- public:
-  // The possible outcomes of a test part (i.e. an assertion or an
-  // explicit SUCCEED(), FAIL(), or ADD_FAILURE()).
-  enum Type {
-    kSuccess,          // Succeeded.
-    kNonFatalFailure,  // Failed but the test can continue.
-    kFatalFailure      // Failed and the test should be terminated.
-  };
-
-  // C'tor.  TestPartResult does NOT have a default constructor.
-  // Always use this constructor (with parameters) to create a
-  // TestPartResult object.
-  TestPartResult(Type a_type,
-                 const char* a_file_name,
-                 int a_line_number,
-                 const char* a_message)
-      : type_(a_type),
-        file_name_(a_file_name == NULL ? "" : a_file_name),
-        line_number_(a_line_number),
-        summary_(ExtractSummary(a_message)),
-        message_(a_message) {
-  }
-
-  // Gets the outcome of the test part.
-  Type type() const { return type_; }
-
-  // Gets the name of the source file where the test part took place, or
-  // NULL if it's unknown.
-  const char* file_name() const {
-    return file_name_.empty() ? NULL : file_name_.c_str();
-  }
-
-  // Gets the line in the source file where the test part took place,
-  // or -1 if it's unknown.
-  int line_number() const { return line_number_; }
-
-  // Gets the summary of the failure message.
-  const char* summary() const { return summary_.c_str(); }
-
-  // Gets the message associated with the test part.
-  const char* message() const { return message_.c_str(); }
-
-  // Returns true iff the test part passed.
-  bool passed() const { return type_ == kSuccess; }
-
-  // Returns true iff the test part failed.
-  bool failed() const { return type_ != kSuccess; }
-
-  // Returns true iff the test part non-fatally failed.
-  bool nonfatally_failed() const { return type_ == kNonFatalFailure; }
-
-  // Returns true iff the test part fatally failed.
-  bool fatally_failed() const { return type_ == kFatalFailure; }
-
- private:
-  Type type_;
-
-  // Gets the summary of the failure message by omitting the stack
-  // trace in it.
-  static std::string ExtractSummary(const char* message);
-
-  // The name of the source file where the test part took place, or
-  // "" if the source file is unknown.
-  std::string file_name_;
-  // The line in the source file where the test part took place, or -1
-  // if the line number is unknown.
-  int line_number_;
-  std::string summary_;  // The test failure summary.
-  std::string message_;  // The test failure message.
-};
-
-// Prints a TestPartResult object.
-std::ostream& operator<<(std::ostream& os, const TestPartResult& result);
-
-// An array of TestPartResult objects.
-//
-// Don't inherit from TestPartResultArray as its destructor is not
-// virtual.
-class GTEST_API_ TestPartResultArray {
- public:
-  TestPartResultArray() {}
-
-  // Appends the given TestPartResult to the array.
-  void Append(const TestPartResult& result);
-
-  // Returns the TestPartResult at the given index (0-based).
-  const TestPartResult& GetTestPartResult(int index) const;
-
-  // Returns the number of TestPartResult objects in the array.
-  int size() const;
-
- private:
-  std::vector<TestPartResult> array_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(TestPartResultArray);
-};
-
-// This interface knows how to report a test part result.
-class TestPartResultReporterInterface {
- public:
-  virtual ~TestPartResultReporterInterface() {}
-
-  virtual void ReportTestPartResult(const TestPartResult& result) = 0;
-};
-
-namespace internal {
-
-// This helper class is used by {ASSERT|EXPECT}_NO_FATAL_FAILURE to check if a
-// statement generates new fatal failures. To do so it registers itself as the
-// current test part result reporter. Besides checking if fatal failures were
-// reported, it only delegates the reporting to the former result reporter.
-// The original result reporter is restored in the destructor.
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-class GTEST_API_ HasNewFatalFailureHelper
-    : public TestPartResultReporterInterface {
- public:
-  HasNewFatalFailureHelper();
-  virtual ~HasNewFatalFailureHelper();
-  virtual void ReportTestPartResult(const TestPartResult& result);
-  bool has_new_fatal_failure() const { return has_new_fatal_failure_; }
- private:
-  bool has_new_fatal_failure_;
-  TestPartResultReporterInterface* original_reporter_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(HasNewFatalFailureHelper);
-};
-
-}  // namespace internal
-
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_TEST_PART_H_
-// Copyright 2008 Google Inc.
-// All Rights Reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_TYPED_TEST_H_
-#define GTEST_INCLUDE_GTEST_GTEST_TYPED_TEST_H_
-
-// This header implements typed tests and type-parameterized tests.
-
-// Typed (aka type-driven) tests repeat the same test for types in a
-// list.  You must know which types you want to test with when writing
-// typed tests. Here's how you do it:
-
-#if 0
-
-// First, define a fixture class template.  It should be parameterized
-// by a type.  Remember to derive it from testing::Test.
-template <typename T>
-class FooTest : public testing::Test {
- public:
-  ...
-  typedef std::list<T> List;
-  static T shared_;
-  T value_;
-};
-
-// Next, associate a list of types with the test case, which will be
-// repeated for each type in the list.  The typedef is necessary for
-// the macro to parse correctly.
-typedef testing::Types<char, int, unsigned int> MyTypes;
-TYPED_TEST_CASE(FooTest, MyTypes);
-
-// If the type list contains only one type, you can write that type
-// directly without Types<...>:
-//   TYPED_TEST_CASE(FooTest, int);
-
-// Then, use TYPED_TEST() instead of TEST_F() to define as many typed
-// tests for this test case as you want.
-TYPED_TEST(FooTest, DoesBlah) {
-  // Inside a test, refer to TypeParam to get the type parameter.
-  // Since we are inside a derived class template, C++ requires use to
-  // visit the members of FooTest via 'this'.
-  TypeParam n = this->value_;
-
-  // To visit static members of the fixture, add the TestFixture::
-  // prefix.
-  n += TestFixture::shared_;
-
-  // To refer to typedefs in the fixture, add the "typename
-  // TestFixture::" prefix.
-  typename TestFixture::List values;
-  values.push_back(n);
-  ...
-}
-
-TYPED_TEST(FooTest, HasPropertyA) { ... }
-
-#endif  // 0
-
-// Type-parameterized tests are abstract test patterns parameterized
-// by a type.  Compared with typed tests, type-parameterized tests
-// allow you to define the test pattern without knowing what the type
-// parameters are.  The defined pattern can be instantiated with
-// different types any number of times, in any number of translation
-// units.
-//
-// If you are designing an interface or concept, you can define a
-// suite of type-parameterized tests to verify properties that any
-// valid implementation of the interface/concept should have.  Then,
-// each implementation can easily instantiate the test suite to verify
-// that it conforms to the requirements, without having to write
-// similar tests repeatedly.  Here's an example:
-
-#if 0
-
-// First, define a fixture class template.  It should be parameterized
-// by a type.  Remember to derive it from testing::Test.
-template <typename T>
-class FooTest : public testing::Test {
-  ...
-};
-
-// Next, declare that you will define a type-parameterized test case
-// (the _P suffix is for "parameterized" or "pattern", whichever you
-// prefer):
-TYPED_TEST_CASE_P(FooTest);
-
-// Then, use TYPED_TEST_P() to define as many type-parameterized tests
-// for this type-parameterized test case as you want.
-TYPED_TEST_P(FooTest, DoesBlah) {
-  // Inside a test, refer to TypeParam to get the type parameter.
-  TypeParam n = 0;
-  ...
-}
-
-TYPED_TEST_P(FooTest, HasPropertyA) { ... }
-
-// Now the tricky part: you need to register all test patterns before
-// you can instantiate them.  The first argument of the macro is the
-// test case name; the rest are the names of the tests in this test
-// case.
-REGISTER_TYPED_TEST_CASE_P(FooTest,
-                           DoesBlah, HasPropertyA);
-
-// Finally, you are free to instantiate the pattern with the types you
-// want.  If you put the above code in a header file, you can #include
-// it in multiple C++ source files and instantiate it multiple times.
-//
-// To distinguish different instances of the pattern, the first
-// argument to the INSTANTIATE_* macro is a prefix that will be added
-// to the actual test case name.  Remember to pick unique prefixes for
-// different instances.
-typedef testing::Types<char, int, unsigned int> MyTypes;
-INSTANTIATE_TYPED_TEST_CASE_P(My, FooTest, MyTypes);
-
-// If the type list contains only one type, you can write that type
-// directly without Types<...>:
-//   INSTANTIATE_TYPED_TEST_CASE_P(My, FooTest, int);
-
-#endif  // 0
-
-
-// Implements typed tests.
-
-#if GTEST_HAS_TYPED_TEST
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// Expands to the name of the typedef for the type parameters of the
-// given test case.
-# define GTEST_TYPE_PARAMS_(TestCaseName) gtest_type_params_##TestCaseName##_
-
-// The 'Types' template argument below must have spaces around it
-// since some compilers may choke on '>>' when passing a template
-// instance (e.g. Types<int>)
-# define TYPED_TEST_CASE(CaseName, Types) \
-  typedef ::testing::internal::TypeList< Types >::type \
-      GTEST_TYPE_PARAMS_(CaseName)
-
-# define TYPED_TEST(CaseName, TestName) \
-  template <typename gtest_TypeParam_> \
-  class GTEST_TEST_CLASS_NAME_(CaseName, TestName) \
-      : public CaseName<gtest_TypeParam_> { \
-   private: \
-    typedef CaseName<gtest_TypeParam_> TestFixture; \
-    typedef gtest_TypeParam_ TypeParam; \
-    virtual void TestBody(); \
-  }; \
-  bool gtest_##CaseName##_##TestName##_registered_ GTEST_ATTRIBUTE_UNUSED_ = \
-      ::testing::internal::TypeParameterizedTest< \
-          CaseName, \
-          ::testing::internal::TemplateSel< \
-              GTEST_TEST_CLASS_NAME_(CaseName, TestName)>, \
-          GTEST_TYPE_PARAMS_(CaseName)>::Register(\
-              "", ::testing::internal::CodeLocation(__FILE__, __LINE__), \
-              #CaseName, #TestName, 0); \
-  template <typename gtest_TypeParam_> \
-  void GTEST_TEST_CLASS_NAME_(CaseName, TestName)<gtest_TypeParam_>::TestBody()
-
-#endif  // GTEST_HAS_TYPED_TEST
-
-// Implements type-parameterized tests.
-
-#if GTEST_HAS_TYPED_TEST_P
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// Expands to the namespace name that the type-parameterized tests for
-// the given type-parameterized test case are defined in.  The exact
-// name of the namespace is subject to change without notice.
-# define GTEST_CASE_NAMESPACE_(TestCaseName) \
-  gtest_case_##TestCaseName##_
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-//
-// Expands to the name of the variable used to remember the names of
-// the defined tests in the given test case.
-# define GTEST_TYPED_TEST_CASE_P_STATE_(TestCaseName) \
-  gtest_typed_test_case_p_state_##TestCaseName##_
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE DIRECTLY.
-//
-// Expands to the name of the variable used to remember the names of
-// the registered tests in the given test case.
-# define GTEST_REGISTERED_TEST_NAMES_(TestCaseName) \
-  gtest_registered_test_names_##TestCaseName##_
-
-// The variables defined in the type-parameterized test macros are
-// static as typically these macros are used in a .h file that can be
-// #included in multiple translation units linked together.
-# define TYPED_TEST_CASE_P(CaseName) \
-  static ::testing::internal::TypedTestCasePState \
-      GTEST_TYPED_TEST_CASE_P_STATE_(CaseName)
-
-# define TYPED_TEST_P(CaseName, TestName) \
-  namespace GTEST_CASE_NAMESPACE_(CaseName) { \
-  template <typename gtest_TypeParam_> \
-  class TestName : public CaseName<gtest_TypeParam_> { \
-   private: \
-    typedef CaseName<gtest_TypeParam_> TestFixture; \
-    typedef gtest_TypeParam_ TypeParam; \
-    virtual void TestBody(); \
-  }; \
-  static bool gtest_##TestName##_defined_ GTEST_ATTRIBUTE_UNUSED_ = \
-      GTEST_TYPED_TEST_CASE_P_STATE_(CaseName).AddTestName(\
-          __FILE__, __LINE__, #CaseName, #TestName); \
-  } \
-  template <typename gtest_TypeParam_> \
-  void GTEST_CASE_NAMESPACE_(CaseName)::TestName<gtest_TypeParam_>::TestBody()
-
-# define REGISTER_TYPED_TEST_CASE_P(CaseName, ...) \
-  namespace GTEST_CASE_NAMESPACE_(CaseName) { \
-  typedef ::testing::internal::Templates<__VA_ARGS__>::type gtest_AllTests_; \
-  } \
-  static const char* const GTEST_REGISTERED_TEST_NAMES_(CaseName) \
-      GTEST_ATTRIBUTE_UNUSED_ = \
-          GTEST_TYPED_TEST_CASE_P_STATE_(CaseName).VerifyRegisteredTestNames( \
-              __FILE__, __LINE__, #__VA_ARGS__)
-
-// The 'Types' template argument below must have spaces around it
-// since some compilers may choke on '>>' when passing a template
-// instance (e.g. Types<int>)
-# define INSTANTIATE_TYPED_TEST_CASE_P(Prefix, CaseName, Types) \
-  bool gtest_##Prefix##_##CaseName GTEST_ATTRIBUTE_UNUSED_ = \
-      ::testing::internal::TypeParameterizedTestCase<CaseName, \
-          GTEST_CASE_NAMESPACE_(CaseName)::gtest_AllTests_, \
-          ::testing::internal::TypeList< Types >::type>::Register(\
-              #Prefix, \
-              ::testing::internal::CodeLocation(__FILE__, __LINE__), \
-              &GTEST_TYPED_TEST_CASE_P_STATE_(CaseName), \
-              #CaseName, GTEST_REGISTERED_TEST_NAMES_(CaseName))
-
-#endif  // GTEST_HAS_TYPED_TEST_P
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_TYPED_TEST_H_
-
-// Depending on the platform, different string classes are available.
-// On Linux, in addition to ::std::string, Google also makes use of
-// class ::string, which has the same interface as ::std::string, but
-// has a different implementation.
-//
-// You can define GTEST_HAS_GLOBAL_STRING to 1 to indicate that
-// ::string is available AND is a distinct type to ::std::string, or
-// define it to 0 to indicate otherwise.
-//
-// If ::std::string and ::string are the same class on your platform
-// due to aliasing, you should define GTEST_HAS_GLOBAL_STRING to 0.
-//
-// If you do not define GTEST_HAS_GLOBAL_STRING, it is defined
-// heuristically.
-
-namespace testing {
-
-// Silence C4100 (unreferenced formal parameter) and 4805
-// unsafe mix of type 'const int' and type 'const bool'
-#ifdef _MSC_VER
-# pragma warning(push)
-# pragma warning(disable:4805)
-# pragma warning(disable:4100)
-#endif
-
-
-// Declares the flags.
-
-// This flag temporary enables the disabled tests.
-GTEST_DECLARE_bool_(also_run_disabled_tests);
-
-// This flag brings the debugger on an assertion failure.
-GTEST_DECLARE_bool_(break_on_failure);
-
-// This flag controls whether Google Test catches all test-thrown exceptions
-// and logs them as failures.
-GTEST_DECLARE_bool_(catch_exceptions);
-
-// This flag enables using colors in terminal output. Available values are
-// "yes" to enable colors, "no" (disable colors), or "auto" (the default)
-// to let Google Test decide.
-GTEST_DECLARE_string_(color);
-
-// This flag sets up the filter to select by name using a glob pattern
-// the tests to run. If the filter is not given all tests are executed.
-GTEST_DECLARE_string_(filter);
-
-// This flag causes the Google Test to list tests. None of the tests listed
-// are actually run if the flag is provided.
-GTEST_DECLARE_bool_(list_tests);
-
-// This flag controls whether Google Test emits a detailed XML report to a file
-// in addition to its normal textual output.
-GTEST_DECLARE_string_(output);
-
-// This flags control whether Google Test prints the elapsed time for each
-// test.
-GTEST_DECLARE_bool_(print_time);
-
-// This flags control whether Google Test prints UTF8 characters as text.
-GTEST_DECLARE_bool_(print_utf8);
-
-// This flag specifies the random number seed.
-GTEST_DECLARE_int32_(random_seed);
-
-// This flag sets how many times the tests are repeated. The default value
-// is 1. If the value is -1 the tests are repeating forever.
-GTEST_DECLARE_int32_(repeat);
-
-// This flag controls whether Google Test includes Google Test internal
-// stack frames in failure stack traces.
-GTEST_DECLARE_bool_(show_internal_stack_frames);
-
-// When this flag is specified, tests' order is randomized on every iteration.
-GTEST_DECLARE_bool_(shuffle);
-
-// This flag specifies the maximum number of stack frames to be
-// printed in a failure message.
-GTEST_DECLARE_int32_(stack_trace_depth);
-
-// When this flag is specified, a failed assertion will throw an
-// exception if exceptions are enabled, or exit the program with a
-// non-zero code otherwise. For use with an external test framework.
-GTEST_DECLARE_bool_(throw_on_failure);
-
-// When this flag is set with a "host:port" string, on supported
-// platforms test results are streamed to the specified port on
-// the specified host machine.
-GTEST_DECLARE_string_(stream_result_to);
-
-// The upper limit for valid stack trace depths.
-const int kMaxStackTraceDepth = 100;
-
-namespace internal {
-
-class AssertHelper;
-class DefaultGlobalTestPartResultReporter;
-class ExecDeathTest;
-class NoExecDeathTest;
-class FinalSuccessChecker;
-class GTestFlagSaver;
-class StreamingListenerTest;
-class TestResultAccessor;
-class TestEventListenersAccessor;
-class TestEventRepeater;
-class UnitTestRecordPropertyTestHelper;
-class WindowsDeathTest;
-class FuchsiaDeathTest;
-class UnitTestImpl* GetUnitTestImpl();
-void ReportFailureInUnknownLocation(TestPartResult::Type result_type,
-                                    const std::string& message);
-
-}  // namespace internal
-
-// The friend relationship of some of these classes is cyclic.
-// If we don't forward declare them the compiler might confuse the classes
-// in friendship clauses with same named classes on the scope.
-class Test;
-class TestCase;
-class TestInfo;
-class UnitTest;
-
-// A class for indicating whether an assertion was successful.  When
-// the assertion wasn't successful, the AssertionResult object
-// remembers a non-empty message that describes how it failed.
-//
-// To create an instance of this class, use one of the factory functions
-// (AssertionSuccess() and AssertionFailure()).
-//
-// This class is useful for two purposes:
-//   1. Defining predicate functions to be used with Boolean test assertions
-//      EXPECT_TRUE/EXPECT_FALSE and their ASSERT_ counterparts
-//   2. Defining predicate-format functions to be
-//      used with predicate assertions (ASSERT_PRED_FORMAT*, etc).
-//
-// For example, if you define IsEven predicate:
-//
-//   testing::AssertionResult IsEven(int n) {
-//     if ((n % 2) == 0)
-//       return testing::AssertionSuccess();
-//     else
-//       return testing::AssertionFailure() << n << " is odd";
-//   }
-//
-// Then the failed expectation EXPECT_TRUE(IsEven(Fib(5)))
-// will print the message
-//
-//   Value of: IsEven(Fib(5))
-//     Actual: false (5 is odd)
-//   Expected: true
-//
-// instead of a more opaque
-//
-//   Value of: IsEven(Fib(5))
-//     Actual: false
-//   Expected: true
-//
-// in case IsEven is a simple Boolean predicate.
-//
-// If you expect your predicate to be reused and want to support informative
-// messages in EXPECT_FALSE and ASSERT_FALSE (negative assertions show up
-// about half as often as positive ones in our tests), supply messages for
-// both success and failure cases:
-//
-//   testing::AssertionResult IsEven(int n) {
-//     if ((n % 2) == 0)
-//       return testing::AssertionSuccess() << n << " is even";
-//     else
-//       return testing::AssertionFailure() << n << " is odd";
-//   }
-//
-// Then a statement EXPECT_FALSE(IsEven(Fib(6))) will print
-//
-//   Value of: IsEven(Fib(6))
-//     Actual: true (8 is even)
-//   Expected: false
-//
-// NB: Predicates that support negative Boolean assertions have reduced
-// performance in positive ones so be careful not to use them in tests
-// that have lots (tens of thousands) of positive Boolean assertions.
-//
-// To use this class with EXPECT_PRED_FORMAT assertions such as:
-//
-//   // Verifies that Foo() returns an even number.
-//   EXPECT_PRED_FORMAT1(IsEven, Foo());
-//
-// you need to define:
-//
-//   testing::AssertionResult IsEven(const char* expr, int n) {
-//     if ((n % 2) == 0)
-//       return testing::AssertionSuccess();
-//     else
-//       return testing::AssertionFailure()
-//         << "Expected: " << expr << " is even\n  Actual: it's " << n;
-//   }
-//
-// If Foo() returns 5, you will see the following message:
-//
-//   Expected: Foo() is even
-//     Actual: it's 5
-//
-class GTEST_API_ AssertionResult {
- public:
-  // Copy constructor.
-  // Used in EXPECT_TRUE/FALSE(assertion_result).
-  AssertionResult(const AssertionResult& other);
-
-#if defined(_MSC_VER) && _MSC_VER < 1910
-  GTEST_DISABLE_MSC_WARNINGS_PUSH_(4800 /* forcing value to bool */)
-#endif
-
-  // Used in the EXPECT_TRUE/FALSE(bool_expression).
-  //
-  // T must be contextually convertible to bool.
-  //
-  // The second parameter prevents this overload from being considered if
-  // the argument is implicitly convertible to AssertionResult. In that case
-  // we want AssertionResult's copy constructor to be used.
-  template <typename T>
-  explicit AssertionResult(
-      const T& success,
-      typename internal::EnableIf<
-          !internal::ImplicitlyConvertible<T, AssertionResult>::value>::type*
-          /*enabler*/ = NULL)
-      : success_(success) {}
-
-#if defined(_MSC_VER) && _MSC_VER < 1910
-  GTEST_DISABLE_MSC_WARNINGS_POP_()
-#endif
-
-  // Assignment operator.
-  AssertionResult& operator=(AssertionResult other) {
-    swap(other);
-    return *this;
-  }
-
-  // Returns true iff the assertion succeeded.
-  operator bool() const { return success_; }  // NOLINT
-
-  // Returns the assertion's negation. Used with EXPECT/ASSERT_FALSE.
-  AssertionResult operator!() const;
-
-  // Returns the text streamed into this AssertionResult. Test assertions
-  // use it when they fail (i.e., the predicate's outcome doesn't match the
-  // assertion's expectation). When nothing has been streamed into the
-  // object, returns an empty string.
-  const char* message() const {
-    return message_.get() != NULL ?  message_->c_str() : "";
-  }
-  // TODO(vladl@google.com): Remove this after making sure no clients use it.
-  // Deprecated; please use message() instead.
-  const char* failure_message() const { return message(); }
-
-  // Streams a custom failure message into this object.
-  template <typename T> AssertionResult& operator<<(const T& value) {
-    AppendMessage(Message() << value);
-    return *this;
-  }
-
-  // Allows streaming basic output manipulators such as endl or flush into
-  // this object.
-  AssertionResult& operator<<(
-      ::std::ostream& (*basic_manipulator)(::std::ostream& stream)) {
-    AppendMessage(Message() << basic_manipulator);
-    return *this;
-  }
-
- private:
-  // Appends the contents of message to message_.
-  void AppendMessage(const Message& a_message) {
-    if (message_.get() == NULL)
-      message_.reset(new ::std::string);
-    message_->append(a_message.GetString().c_str());
-  }
-
-  // Swap the contents of this AssertionResult with other.
-  void swap(AssertionResult& other);
-
-  // Stores result of the assertion predicate.
-  bool success_;
-  // Stores the message describing the condition in case the expectation
-  // construct is not satisfied with the predicate's outcome.
-  // Referenced via a pointer to avoid taking too much stack frame space
-  // with test assertions.
-  internal::scoped_ptr< ::std::string> message_;
-};
-
-// Makes a successful assertion result.
-GTEST_API_ AssertionResult AssertionSuccess();
-
-// Makes a failed assertion result.
-GTEST_API_ AssertionResult AssertionFailure();
-
-// Makes a failed assertion result with the given failure message.
-// Deprecated; use AssertionFailure() << msg.
-GTEST_API_ AssertionResult AssertionFailure(const Message& msg);
-
-}  // namespace testing
-
-// Includes the auto-generated header that implements a family of generic
-// predicate assertion macros. This include comes late because it relies on
-// APIs declared above.
-// Copyright 2006, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-// This file is AUTOMATICALLY GENERATED on 01/02/2018 by command
-// 'gen_gtest_pred_impl.py 5'.  DO NOT EDIT BY HAND!
-//
-// Implements a family of generic predicate assertion macros.
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_PRED_IMPL_H_
-#define GTEST_INCLUDE_GTEST_GTEST_PRED_IMPL_H_
-
-
-namespace testing {
-
-// This header implements a family of generic predicate assertion
-// macros:
-//
-//   ASSERT_PRED_FORMAT1(pred_format, v1)
-//   ASSERT_PRED_FORMAT2(pred_format, v1, v2)
-//   ...
-//
-// where pred_format is a function or functor that takes n (in the
-// case of ASSERT_PRED_FORMATn) values and their source expression
-// text, and returns a testing::AssertionResult.  See the definition
-// of ASSERT_EQ in gtest.h for an example.
-//
-// If you don't care about formatting, you can use the more
-// restrictive version:
-//
-//   ASSERT_PRED1(pred, v1)
-//   ASSERT_PRED2(pred, v1, v2)
-//   ...
-//
-// where pred is an n-ary function or functor that returns bool,
-// and the values v1, v2, ..., must support the << operator for
-// streaming to std::ostream.
-//
-// We also define the EXPECT_* variations.
-//
-// For now we only support predicates whose arity is at most 5.
-
-// GTEST_ASSERT_ is the basic statement to which all of the assertions
-// in this file reduce.  Don't use this in your code.
-
-#define GTEST_ASSERT_(expression, on_failure) \
-  GTEST_AMBIGUOUS_ELSE_BLOCKER_ \
-  if (const ::testing::AssertionResult gtest_ar = (expression)) \
-    ; \
-  else \
-    on_failure(gtest_ar.failure_message())
-
-
-// Helper function for implementing {EXPECT|ASSERT}_PRED1.  Don't use
-// this in your code.
-template <typename Pred,
-          typename T1>
-AssertionResult AssertPred1Helper(const char* pred_text,
-                                  const char* e1,
-                                  Pred pred,
-                                  const T1& v1) {
-  if (pred(v1)) return AssertionSuccess();
-
-  return AssertionFailure() << pred_text << "("
-                            << e1 << ") evaluates to false, where"
-                            << "\n" << e1 << " evaluates to " << v1;
-}
-
-// Internal macro for implementing {EXPECT|ASSERT}_PRED_FORMAT1.
-// Don't use this in your code.
-#define GTEST_PRED_FORMAT1_(pred_format, v1, on_failure)\
-  GTEST_ASSERT_(pred_format(#v1, v1), \
-                on_failure)
-
-// Internal macro for implementing {EXPECT|ASSERT}_PRED1.  Don't use
-// this in your code.
-#define GTEST_PRED1_(pred, v1, on_failure)\
-  GTEST_ASSERT_(::testing::AssertPred1Helper(#pred, \
-                                             #v1, \
-                                             pred, \
-                                             v1), on_failure)
-
-// Unary predicate assertion macros.
-#define EXPECT_PRED_FORMAT1(pred_format, v1) \
-  GTEST_PRED_FORMAT1_(pred_format, v1, GTEST_NONFATAL_FAILURE_)
-#define EXPECT_PRED1(pred, v1) \
-  GTEST_PRED1_(pred, v1, GTEST_NONFATAL_FAILURE_)
-#define ASSERT_PRED_FORMAT1(pred_format, v1) \
-  GTEST_PRED_FORMAT1_(pred_format, v1, GTEST_FATAL_FAILURE_)
-#define ASSERT_PRED1(pred, v1) \
-  GTEST_PRED1_(pred, v1, GTEST_FATAL_FAILURE_)
-
-
-
-// Helper function for implementing {EXPECT|ASSERT}_PRED2.  Don't use
-// this in your code.
-template <typename Pred,
-          typename T1,
-          typename T2>
-AssertionResult AssertPred2Helper(const char* pred_text,
-                                  const char* e1,
-                                  const char* e2,
-                                  Pred pred,
-                                  const T1& v1,
-                                  const T2& v2) {
-  if (pred(v1, v2)) return AssertionSuccess();
-
-  return AssertionFailure() << pred_text << "("
-                            << e1 << ", "
-                            << e2 << ") evaluates to false, where"
-                            << "\n" << e1 << " evaluates to " << v1
-                            << "\n" << e2 << " evaluates to " << v2;
-}
-
-// Internal macro for implementing {EXPECT|ASSERT}_PRED_FORMAT2.
-// Don't use this in your code.
-#define GTEST_PRED_FORMAT2_(pred_format, v1, v2, on_failure)\
-  GTEST_ASSERT_(pred_format(#v1, #v2, v1, v2), \
-                on_failure)
-
-// Internal macro for implementing {EXPECT|ASSERT}_PRED2.  Don't use
-// this in your code.
-#define GTEST_PRED2_(pred, v1, v2, on_failure)\
-  GTEST_ASSERT_(::testing::AssertPred2Helper(#pred, \
-                                             #v1, \
-                                             #v2, \
-                                             pred, \
-                                             v1, \
-                                             v2), on_failure)
-
-// Binary predicate assertion macros.
-#define EXPECT_PRED_FORMAT2(pred_format, v1, v2) \
-  GTEST_PRED_FORMAT2_(pred_format, v1, v2, GTEST_NONFATAL_FAILURE_)
-#define EXPECT_PRED2(pred, v1, v2) \
-  GTEST_PRED2_(pred, v1, v2, GTEST_NONFATAL_FAILURE_)
-#define ASSERT_PRED_FORMAT2(pred_format, v1, v2) \
-  GTEST_PRED_FORMAT2_(pred_format, v1, v2, GTEST_FATAL_FAILURE_)
-#define ASSERT_PRED2(pred, v1, v2) \
-  GTEST_PRED2_(pred, v1, v2, GTEST_FATAL_FAILURE_)
-
-
-
-// Helper function for implementing {EXPECT|ASSERT}_PRED3.  Don't use
-// this in your code.
-template <typename Pred,
-          typename T1,
-          typename T2,
-          typename T3>
-AssertionResult AssertPred3Helper(const char* pred_text,
-                                  const char* e1,
-                                  const char* e2,
-                                  const char* e3,
-                                  Pred pred,
-                                  const T1& v1,
-                                  const T2& v2,
-                                  const T3& v3) {
-  if (pred(v1, v2, v3)) return AssertionSuccess();
-
-  return AssertionFailure() << pred_text << "("
-                            << e1 << ", "
-                            << e2 << ", "
-                            << e3 << ") evaluates to false, where"
-                            << "\n" << e1 << " evaluates to " << v1
-                            << "\n" << e2 << " evaluates to " << v2
-                            << "\n" << e3 << " evaluates to " << v3;
-}
-
-// Internal macro for implementing {EXPECT|ASSERT}_PRED_FORMAT3.
-// Don't use this in your code.
-#define GTEST_PRED_FORMAT3_(pred_format, v1, v2, v3, on_failure)\
-  GTEST_ASSERT_(pred_format(#v1, #v2, #v3, v1, v2, v3), \
-                on_failure)
-
-// Internal macro for implementing {EXPECT|ASSERT}_PRED3.  Don't use
-// this in your code.
-#define GTEST_PRED3_(pred, v1, v2, v3, on_failure)\
-  GTEST_ASSERT_(::testing::AssertPred3Helper(#pred, \
-                                             #v1, \
-                                             #v2, \
-                                             #v3, \
-                                             pred, \
-                                             v1, \
-                                             v2, \
-                                             v3), on_failure)
-
-// Ternary predicate assertion macros.
-#define EXPECT_PRED_FORMAT3(pred_format, v1, v2, v3) \
-  GTEST_PRED_FORMAT3_(pred_format, v1, v2, v3, GTEST_NONFATAL_FAILURE_)
-#define EXPECT_PRED3(pred, v1, v2, v3) \
-  GTEST_PRED3_(pred, v1, v2, v3, GTEST_NONFATAL_FAILURE_)
-#define ASSERT_PRED_FORMAT3(pred_format, v1, v2, v3) \
-  GTEST_PRED_FORMAT3_(pred_format, v1, v2, v3, GTEST_FATAL_FAILURE_)
-#define ASSERT_PRED3(pred, v1, v2, v3) \
-  GTEST_PRED3_(pred, v1, v2, v3, GTEST_FATAL_FAILURE_)
-
-
-
-// Helper function for implementing {EXPECT|ASSERT}_PRED4.  Don't use
-// this in your code.
-template <typename Pred,
-          typename T1,
-          typename T2,
-          typename T3,
-          typename T4>
-AssertionResult AssertPred4Helper(const char* pred_text,
-                                  const char* e1,
-                                  const char* e2,
-                                  const char* e3,
-                                  const char* e4,
-                                  Pred pred,
-                                  const T1& v1,
-                                  const T2& v2,
-                                  const T3& v3,
-                                  const T4& v4) {
-  if (pred(v1, v2, v3, v4)) return AssertionSuccess();
-
-  return AssertionFailure() << pred_text << "("
-                            << e1 << ", "
-                            << e2 << ", "
-                            << e3 << ", "
-                            << e4 << ") evaluates to false, where"
-                            << "\n" << e1 << " evaluates to " << v1
-                            << "\n" << e2 << " evaluates to " << v2
-                            << "\n" << e3 << " evaluates to " << v3
-                            << "\n" << e4 << " evaluates to " << v4;
-}
-
-// Internal macro for implementing {EXPECT|ASSERT}_PRED_FORMAT4.
-// Don't use this in your code.
-#define GTEST_PRED_FORMAT4_(pred_format, v1, v2, v3, v4, on_failure)\
-  GTEST_ASSERT_(pred_format(#v1, #v2, #v3, #v4, v1, v2, v3, v4), \
-                on_failure)
-
-// Internal macro for implementing {EXPECT|ASSERT}_PRED4.  Don't use
-// this in your code.
-#define GTEST_PRED4_(pred, v1, v2, v3, v4, on_failure)\
-  GTEST_ASSERT_(::testing::AssertPred4Helper(#pred, \
-                                             #v1, \
-                                             #v2, \
-                                             #v3, \
-                                             #v4, \
-                                             pred, \
-                                             v1, \
-                                             v2, \
-                                             v3, \
-                                             v4), on_failure)
-
-// 4-ary predicate assertion macros.
-#define EXPECT_PRED_FORMAT4(pred_format, v1, v2, v3, v4) \
-  GTEST_PRED_FORMAT4_(pred_format, v1, v2, v3, v4, GTEST_NONFATAL_FAILURE_)
-#define EXPECT_PRED4(pred, v1, v2, v3, v4) \
-  GTEST_PRED4_(pred, v1, v2, v3, v4, GTEST_NONFATAL_FAILURE_)
-#define ASSERT_PRED_FORMAT4(pred_format, v1, v2, v3, v4) \
-  GTEST_PRED_FORMAT4_(pred_format, v1, v2, v3, v4, GTEST_FATAL_FAILURE_)
-#define ASSERT_PRED4(pred, v1, v2, v3, v4) \
-  GTEST_PRED4_(pred, v1, v2, v3, v4, GTEST_FATAL_FAILURE_)
-
-
-
-// Helper function for implementing {EXPECT|ASSERT}_PRED5.  Don't use
-// this in your code.
-template <typename Pred,
-          typename T1,
-          typename T2,
-          typename T3,
-          typename T4,
-          typename T5>
-AssertionResult AssertPred5Helper(const char* pred_text,
-                                  const char* e1,
-                                  const char* e2,
-                                  const char* e3,
-                                  const char* e4,
-                                  const char* e5,
-                                  Pred pred,
-                                  const T1& v1,
-                                  const T2& v2,
-                                  const T3& v3,
-                                  const T4& v4,
-                                  const T5& v5) {
-  if (pred(v1, v2, v3, v4, v5)) return AssertionSuccess();
-
-  return AssertionFailure() << pred_text << "("
-                            << e1 << ", "
-                            << e2 << ", "
-                            << e3 << ", "
-                            << e4 << ", "
-                            << e5 << ") evaluates to false, where"
-                            << "\n" << e1 << " evaluates to " << v1
-                            << "\n" << e2 << " evaluates to " << v2
-                            << "\n" << e3 << " evaluates to " << v3
-                            << "\n" << e4 << " evaluates to " << v4
-                            << "\n" << e5 << " evaluates to " << v5;
-}
-
-// Internal macro for implementing {EXPECT|ASSERT}_PRED_FORMAT5.
-// Don't use this in your code.
-#define GTEST_PRED_FORMAT5_(pred_format, v1, v2, v3, v4, v5, on_failure)\
-  GTEST_ASSERT_(pred_format(#v1, #v2, #v3, #v4, #v5, v1, v2, v3, v4, v5), \
-                on_failure)
-
-// Internal macro for implementing {EXPECT|ASSERT}_PRED5.  Don't use
-// this in your code.
-#define GTEST_PRED5_(pred, v1, v2, v3, v4, v5, on_failure)\
-  GTEST_ASSERT_(::testing::AssertPred5Helper(#pred, \
-                                             #v1, \
-                                             #v2, \
-                                             #v3, \
-                                             #v4, \
-                                             #v5, \
-                                             pred, \
-                                             v1, \
-                                             v2, \
-                                             v3, \
-                                             v4, \
-                                             v5), on_failure)
-
-// 5-ary predicate assertion macros.
-#define EXPECT_PRED_FORMAT5(pred_format, v1, v2, v3, v4, v5) \
-  GTEST_PRED_FORMAT5_(pred_format, v1, v2, v3, v4, v5, GTEST_NONFATAL_FAILURE_)
-#define EXPECT_PRED5(pred, v1, v2, v3, v4, v5) \
-  GTEST_PRED5_(pred, v1, v2, v3, v4, v5, GTEST_NONFATAL_FAILURE_)
-#define ASSERT_PRED_FORMAT5(pred_format, v1, v2, v3, v4, v5) \
-  GTEST_PRED_FORMAT5_(pred_format, v1, v2, v3, v4, v5, GTEST_FATAL_FAILURE_)
-#define ASSERT_PRED5(pred, v1, v2, v3, v4, v5) \
-  GTEST_PRED5_(pred, v1, v2, v3, v4, v5, GTEST_FATAL_FAILURE_)
-
-
-
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_PRED_IMPL_H_
-
-namespace testing {
-
-// The abstract class that all tests inherit from.
-//
-// In Google Test, a unit test program contains one or many TestCases, and
-// each TestCase contains one or many Tests.
-//
-// When you define a test using the TEST macro, you don't need to
-// explicitly derive from Test - the TEST macro automatically does
-// this for you.
-//
-// The only time you derive from Test is when defining a test fixture
-// to be used in a TEST_F.  For example:
-//
-//   class FooTest : public testing::Test {
-//    protected:
-//     void SetUp() override { ... }
-//     void TearDown() override { ... }
-//     ...
-//   };
-//
-//   TEST_F(FooTest, Bar) { ... }
-//   TEST_F(FooTest, Baz) { ... }
-//
-// Test is not copyable.
-class GTEST_API_ Test {
- public:
-  friend class TestInfo;
-
-  // Defines types for pointers to functions that set up and tear down
-  // a test case.
-  typedef internal::SetUpTestCaseFunc SetUpTestCaseFunc;
-  typedef internal::TearDownTestCaseFunc TearDownTestCaseFunc;
-
-  // The d'tor is virtual as we intend to inherit from Test.
-  virtual ~Test();
-
-  // Sets up the stuff shared by all tests in this test case.
-  //
-  // Google Test will call Foo::SetUpTestCase() before running the first
-  // test in test case Foo.  Hence a sub-class can define its own
-  // SetUpTestCase() method to shadow the one defined in the super
-  // class.
-  static void SetUpTestCase() {}
-
-  // Tears down the stuff shared by all tests in this test case.
-  //
-  // Google Test will call Foo::TearDownTestCase() after running the last
-  // test in test case Foo.  Hence a sub-class can define its own
-  // TearDownTestCase() method to shadow the one defined in the super
-  // class.
-  static void TearDownTestCase() {}
-
-  // Returns true iff the current test has a fatal failure.
-  static bool HasFatalFailure();
-
-  // Returns true iff the current test has a non-fatal failure.
-  static bool HasNonfatalFailure();
-
-  // Returns true iff the current test has a (either fatal or
-  // non-fatal) failure.
-  static bool HasFailure() { return HasFatalFailure() || HasNonfatalFailure(); }
-
-  // Logs a property for the current test, test case, or for the entire
-  // invocation of the test program when used outside of the context of a
-  // test case.  Only the last value for a given key is remembered.  These
-  // are public static so they can be called from utility functions that are
-  // not members of the test fixture.  Calls to RecordProperty made during
-  // lifespan of the test (from the moment its constructor starts to the
-  // moment its destructor finishes) will be output in XML as attributes of
-  // the <testcase> element.  Properties recorded from fixture's
-  // SetUpTestCase or TearDownTestCase are logged as attributes of the
-  // corresponding <testsuite> element.  Calls to RecordProperty made in the
-  // global context (before or after invocation of RUN_ALL_TESTS and from
-  // SetUp/TearDown method of Environment objects registered with Google
-  // Test) will be output as attributes of the <testsuites> element.
-  static void RecordProperty(const std::string& key, const std::string& value);
-  static void RecordProperty(const std::string& key, int value);
-
- protected:
-  // Creates a Test object.
-  Test();
-
-  // Sets up the test fixture.
-  virtual void SetUp();
-
-  // Tears down the test fixture.
-  virtual void TearDown();
-
- private:
-  // Returns true iff the current test has the same fixture class as
-  // the first test in the current test case.
-  static bool HasSameFixtureClass();
-
-  // Runs the test after the test fixture has been set up.
-  //
-  // A sub-class must implement this to define the test logic.
-  //
-  // DO NOT OVERRIDE THIS FUNCTION DIRECTLY IN A USER PROGRAM.
-  // Instead, use the TEST or TEST_F macro.
-  virtual void TestBody() = 0;
-
-  // Sets up, executes, and tears down the test.
-  void Run();
-
-  // Deletes self.  We deliberately pick an unusual name for this
-  // internal method to avoid clashing with names used in user TESTs.
-  void DeleteSelf_() { delete this; }
-
-  const internal::scoped_ptr< GTEST_FLAG_SAVER_ > gtest_flag_saver_;
-
-  // Often a user misspells SetUp() as Setup() and spends a long time
-  // wondering why it is never called by Google Test.  The declaration of
-  // the following method is solely for catching such an error at
-  // compile time:
-  //
-  //   - The return type is deliberately chosen to be not void, so it
-  //   will be a conflict if void Setup() is declared in the user's
-  //   test fixture.
-  //
-  //   - This method is private, so it will be another compiler error
-  //   if the method is called from the user's test fixture.
-  //
-  // DO NOT OVERRIDE THIS FUNCTION.
-  //
-  // If you see an error about overriding the following function or
-  // about it being private, you have mis-spelled SetUp() as Setup().
-  struct Setup_should_be_spelled_SetUp {};
-  virtual Setup_should_be_spelled_SetUp* Setup() { return NULL; }
-
-  // We disallow copying Tests.
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(Test);
-};
-
-typedef internal::TimeInMillis TimeInMillis;
-
-// A copyable object representing a user specified test property which can be
-// output as a key/value string pair.
-//
-// Don't inherit from TestProperty as its destructor is not virtual.
-class TestProperty {
- public:
-  // C'tor.  TestProperty does NOT have a default constructor.
-  // Always use this constructor (with parameters) to create a
-  // TestProperty object.
-  TestProperty(const std::string& a_key, const std::string& a_value) :
-    key_(a_key), value_(a_value) {
-  }
-
-  // Gets the user supplied key.
-  const char* key() const {
-    return key_.c_str();
-  }
-
-  // Gets the user supplied value.
-  const char* value() const {
-    return value_.c_str();
-  }
-
-  // Sets a new value, overriding the one supplied in the constructor.
-  void SetValue(const std::string& new_value) {
-    value_ = new_value;
-  }
-
- private:
-  // The key supplied by the user.
-  std::string key_;
-  // The value supplied by the user.
-  std::string value_;
-};
-
-// The result of a single Test.  This includes a list of
-// TestPartResults, a list of TestProperties, a count of how many
-// death tests there are in the Test, and how much time it took to run
-// the Test.
-//
-// TestResult is not copyable.
-class GTEST_API_ TestResult {
- public:
-  // Creates an empty TestResult.
-  TestResult();
-
-  // D'tor.  Do not inherit from TestResult.
-  ~TestResult();
-
-  // Gets the number of all test parts.  This is the sum of the number
-  // of successful test parts and the number of failed test parts.
-  int total_part_count() const;
-
-  // Returns the number of the test properties.
-  int test_property_count() const;
-
-  // Returns true iff the test passed (i.e. no test part failed).
-  bool Passed() const { return !Failed(); }
-
-  // Returns true iff the test failed.
-  bool Failed() const;
-
-  // Returns true iff the test fatally failed.
-  bool HasFatalFailure() const;
-
-  // Returns true iff the test has a non-fatal failure.
-  bool HasNonfatalFailure() const;
-
-  // Returns the elapsed time, in milliseconds.
-  TimeInMillis elapsed_time() const { return elapsed_time_; }
-
-  // Returns the i-th test part result among all the results. i can range from 0
-  // to total_part_count() - 1. If i is not in that range, aborts the program.
-  const TestPartResult& GetTestPartResult(int i) const;
-
-  // Returns the i-th test property. i can range from 0 to
-  // test_property_count() - 1. If i is not in that range, aborts the
-  // program.
-  const TestProperty& GetTestProperty(int i) const;
-
- private:
-  friend class TestInfo;
-  friend class TestCase;
-  friend class UnitTest;
-  friend class internal::DefaultGlobalTestPartResultReporter;
-  friend class internal::ExecDeathTest;
-  friend class internal::TestResultAccessor;
-  friend class internal::UnitTestImpl;
-  friend class internal::WindowsDeathTest;
-  friend class internal::FuchsiaDeathTest;
-
-  // Gets the vector of TestPartResults.
-  const std::vector<TestPartResult>& test_part_results() const {
-    return test_part_results_;
-  }
-
-  // Gets the vector of TestProperties.
-  const std::vector<TestProperty>& test_properties() const {
-    return test_properties_;
-  }
-
-  // Sets the elapsed time.
-  void set_elapsed_time(TimeInMillis elapsed) { elapsed_time_ = elapsed; }
-
-  // Adds a test property to the list. The property is validated and may add
-  // a non-fatal failure if invalid (e.g., if it conflicts with reserved
-  // key names). If a property is already recorded for the same key, the
-  // value will be updated, rather than storing multiple values for the same
-  // key.  xml_element specifies the element for which the property is being
-  // recorded and is used for validation.
-  void RecordProperty(const std::string& xml_element,
-                      const TestProperty& test_property);
-
-  // Adds a failure if the key is a reserved attribute of Google Test
-  // testcase tags.  Returns true if the property is valid.
-  // TODO(russr): Validate attribute names are legal and human readable.
-  static bool ValidateTestProperty(const std::string& xml_element,
-                                   const TestProperty& test_property);
-
-  // Adds a test part result to the list.
-  void AddTestPartResult(const TestPartResult& test_part_result);
-
-  // Returns the death test count.
-  int death_test_count() const { return death_test_count_; }
-
-  // Increments the death test count, returning the new count.
-  int increment_death_test_count() { return ++death_test_count_; }
-
-  // Clears the test part results.
-  void ClearTestPartResults();
-
-  // Clears the object.
-  void Clear();
-
-  // Protects mutable state of the property vector and of owned
-  // properties, whose values may be updated.
-  internal::Mutex test_properites_mutex_;
-
-  // The vector of TestPartResults
-  std::vector<TestPartResult> test_part_results_;
-  // The vector of TestProperties
-  std::vector<TestProperty> test_properties_;
-  // Running count of death tests.
-  int death_test_count_;
-  // The elapsed time, in milliseconds.
-  TimeInMillis elapsed_time_;
-
-  // We disallow copying TestResult.
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(TestResult);
-};  // class TestResult
-
-// A TestInfo object stores the following information about a test:
-//
-//   Test case name
-//   Test name
-//   Whether the test should be run
-//   A function pointer that creates the test object when invoked
-//   Test result
-//
-// The constructor of TestInfo registers itself with the UnitTest
-// singleton such that the RUN_ALL_TESTS() macro knows which tests to
-// run.
-class GTEST_API_ TestInfo {
- public:
-  // Destructs a TestInfo object.  This function is not virtual, so
-  // don't inherit from TestInfo.
-  ~TestInfo();
-
-  // Returns the test case name.
-  const char* test_case_name() const { return test_case_name_.c_str(); }
-
-  // Returns the test name.
-  const char* name() const { return name_.c_str(); }
-
-  // Returns the name of the parameter type, or NULL if this is not a typed
-  // or a type-parameterized test.
-  const char* type_param() const {
-    if (type_param_.get() != NULL)
-      return type_param_->c_str();
-    return NULL;
-  }
-
-  // Returns the text representation of the value parameter, or NULL if this
-  // is not a value-parameterized test.
-  const char* value_param() const {
-    if (value_param_.get() != NULL)
-      return value_param_->c_str();
-    return NULL;
-  }
-
-  // Returns the file name where this test is defined.
-  const char* file() const { return location_.file.c_str(); }
-
-  // Returns the line where this test is defined.
-  int line() const { return location_.line; }
-
-  // Return true if this test should not be run because it's in another shard.
-  bool is_in_another_shard() const { return is_in_another_shard_; }
-
-  // Returns true if this test should run, that is if the test is not
-  // disabled (or it is disabled but the also_run_disabled_tests flag has
-  // been specified) and its full name matches the user-specified filter.
-  //
-  // Google Test allows the user to filter the tests by their full names.
-  // The full name of a test Bar in test case Foo is defined as
-  // "Foo.Bar".  Only the tests that match the filter will run.
-  //
-  // A filter is a colon-separated list of glob (not regex) patterns,
-  // optionally followed by a '-' and a colon-separated list of
-  // negative patterns (tests to exclude).  A test is run if it
-  // matches one of the positive patterns and does not match any of
-  // the negative patterns.
-  //
-  // For example, *A*:Foo.* is a filter that matches any string that
-  // contains the character 'A' or starts with "Foo.".
-  bool should_run() const { return should_run_; }
-
-  // Returns true iff this test will appear in the XML report.
-  bool is_reportable() const {
-    // The XML report includes tests matching the filter, excluding those
-    // run in other shards.
-    return matches_filter_ && !is_in_another_shard_;
-  }
-
-  // Returns the result of the test.
-  const TestResult* result() const { return &result_; }
-
- private:
-#if GTEST_HAS_DEATH_TEST
-  friend class internal::DefaultDeathTestFactory;
-#endif  // GTEST_HAS_DEATH_TEST
-  friend class Test;
-  friend class TestCase;
-  friend class internal::UnitTestImpl;
-  friend class internal::StreamingListenerTest;
-  friend TestInfo* internal::MakeAndRegisterTestInfo(
-      const char* test_case_name,
-      const char* name,
-      const char* type_param,
-      const char* value_param,
-      internal::CodeLocation code_location,
-      internal::TypeId fixture_class_id,
-      Test::SetUpTestCaseFunc set_up_tc,
-      Test::TearDownTestCaseFunc tear_down_tc,
-      internal::TestFactoryBase* factory);
-
-  // Constructs a TestInfo object. The newly constructed instance assumes
-  // ownership of the factory object.
-  TestInfo(const std::string& test_case_name,
-           const std::string& name,
-           const char* a_type_param,   // NULL if not a type-parameterized test
-           const char* a_value_param,  // NULL if not a value-parameterized test
-           internal::CodeLocation a_code_location,
-           internal::TypeId fixture_class_id,
-           internal::TestFactoryBase* factory);
-
-  // Increments the number of death tests encountered in this test so
-  // far.
-  int increment_death_test_count() {
-    return result_.increment_death_test_count();
-  }
-
-  // Creates the test object, runs it, records its result, and then
-  // deletes it.
-  void Run();
-
-  static void ClearTestResult(TestInfo* test_info) {
-    test_info->result_.Clear();
-  }
-
-  // These fields are immutable properties of the test.
-  const std::string test_case_name_;     // Test case name
-  const std::string name_;               // Test name
-  // Name of the parameter type, or NULL if this is not a typed or a
-  // type-parameterized test.
-  const internal::scoped_ptr<const ::std::string> type_param_;
-  // Text representation of the value parameter, or NULL if this is not a
-  // value-parameterized test.
-  const internal::scoped_ptr<const ::std::string> value_param_;
-  internal::CodeLocation location_;
-  const internal::TypeId fixture_class_id_;   // ID of the test fixture class
-  bool should_run_;                 // True iff this test should run
-  bool is_disabled_;                // True iff this test is disabled
-  bool matches_filter_;             // True if this test matches the
-                                    // user-specified filter.
-  bool is_in_another_shard_;        // Will be run in another shard.
-  internal::TestFactoryBase* const factory_;  // The factory that creates
-                                              // the test object
-
-  // This field is mutable and needs to be reset before running the
-  // test for the second time.
-  TestResult result_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(TestInfo);
-};
-
-// A test case, which consists of a vector of TestInfos.
-//
-// TestCase is not copyable.
-class GTEST_API_ TestCase {
- public:
-  // Creates a TestCase with the given name.
-  //
-  // TestCase does NOT have a default constructor.  Always use this
-  // constructor to create a TestCase object.
-  //
-  // Arguments:
-  //
-  //   name:         name of the test case
-  //   a_type_param: the name of the test's type parameter, or NULL if
-  //                 this is not a type-parameterized test.
-  //   set_up_tc:    pointer to the function that sets up the test case
-  //   tear_down_tc: pointer to the function that tears down the test case
-  TestCase(const char* name, const char* a_type_param,
-           Test::SetUpTestCaseFunc set_up_tc,
-           Test::TearDownTestCaseFunc tear_down_tc);
-
-  // Destructor of TestCase.
-  virtual ~TestCase();
-
-  // Gets the name of the TestCase.
-  const char* name() const { return name_.c_str(); }
-
-  // Returns the name of the parameter type, or NULL if this is not a
-  // type-parameterized test case.
-  const char* type_param() const {
-    if (type_param_.get() != NULL)
-      return type_param_->c_str();
-    return NULL;
-  }
-
-  // Returns true if any test in this test case should run.
-  bool should_run() const { return should_run_; }
-
-  // Gets the number of successful tests in this test case.
-  int successful_test_count() const;
-
-  // Gets the number of failed tests in this test case.
-  int failed_test_count() const;
-
-  // Gets the number of disabled tests that will be reported in the XML report.
-  int reportable_disabled_test_count() const;
-
-  // Gets the number of disabled tests in this test case.
-  int disabled_test_count() const;
-
-  // Gets the number of tests to be printed in the XML report.
-  int reportable_test_count() const;
-
-  // Get the number of tests in this test case that should run.
-  int test_to_run_count() const;
-
-  // Gets the number of all tests in this test case.
-  int total_test_count() const;
-
-  // Returns true iff the test case passed.
-  bool Passed() const { return !Failed(); }
-
-  // Returns true iff the test case failed.
-  bool Failed() const { return failed_test_count() > 0; }
-
-  // Returns the elapsed time, in milliseconds.
-  TimeInMillis elapsed_time() const { return elapsed_time_; }
-
-  // Returns the i-th test among all the tests. i can range from 0 to
-  // total_test_count() - 1. If i is not in that range, returns NULL.
-  const TestInfo* GetTestInfo(int i) const;
-
-  // Returns the TestResult that holds test properties recorded during
-  // execution of SetUpTestCase and TearDownTestCase.
-  const TestResult& ad_hoc_test_result() const { return ad_hoc_test_result_; }
-
- private:
-  friend class Test;
-  friend class internal::UnitTestImpl;
-
-  // Gets the (mutable) vector of TestInfos in this TestCase.
-  std::vector<TestInfo*>& test_info_list() { return test_info_list_; }
-
-  // Gets the (immutable) vector of TestInfos in this TestCase.
-  const std::vector<TestInfo*>& test_info_list() const {
-    return test_info_list_;
-  }
-
-  // Returns the i-th test among all the tests. i can range from 0 to
-  // total_test_count() - 1. If i is not in that range, returns NULL.
-  TestInfo* GetMutableTestInfo(int i);
-
-  // Sets the should_run member.
-  void set_should_run(bool should) { should_run_ = should; }
-
-  // Adds a TestInfo to this test case.  Will delete the TestInfo upon
-  // destruction of the TestCase object.
-  void AddTestInfo(TestInfo * test_info);
-
-  // Clears the results of all tests in this test case.
-  void ClearResult();
-
-  // Clears the results of all tests in the given test case.
-  static void ClearTestCaseResult(TestCase* test_case) {
-    test_case->ClearResult();
-  }
-
-  // Runs every test in this TestCase.
-  void Run();
-
-  // Runs SetUpTestCase() for this TestCase.  This wrapper is needed
-  // for catching exceptions thrown from SetUpTestCase().
-  void RunSetUpTestCase() { (*set_up_tc_)(); }
-
-  // Runs TearDownTestCase() for this TestCase.  This wrapper is
-  // needed for catching exceptions thrown from TearDownTestCase().
-  void RunTearDownTestCase() { (*tear_down_tc_)(); }
-
-  // Returns true iff test passed.
-  static bool TestPassed(const TestInfo* test_info) {
-    return test_info->should_run() && test_info->result()->Passed();
-  }
-
-  // Returns true iff test failed.
-  static bool TestFailed(const TestInfo* test_info) {
-    return test_info->should_run() && test_info->result()->Failed();
-  }
-
-  // Returns true iff the test is disabled and will be reported in the XML
-  // report.
-  static bool TestReportableDisabled(const TestInfo* test_info) {
-    return test_info->is_reportable() && test_info->is_disabled_;
-  }
-
-  // Returns true iff test is disabled.
-  static bool TestDisabled(const TestInfo* test_info) {
-    return test_info->is_disabled_;
-  }
-
-  // Returns true iff this test will appear in the XML report.
-  static bool TestReportable(const TestInfo* test_info) {
-    return test_info->is_reportable();
-  }
-
-  // Returns true if the given test should run.
-  static bool ShouldRunTest(const TestInfo* test_info) {
-    return test_info->should_run();
-  }
-
-  // Shuffles the tests in this test case.
-  void ShuffleTests(internal::Random* random);
-
-  // Restores the test order to before the first shuffle.
-  void UnshuffleTests();
-
-  // Name of the test case.
-  std::string name_;
-  // Name of the parameter type, or NULL if this is not a typed or a
-  // type-parameterized test.
-  const internal::scoped_ptr<const ::std::string> type_param_;
-  // The vector of TestInfos in their original order.  It owns the
-  // elements in the vector.
-  std::vector<TestInfo*> test_info_list_;
-  // Provides a level of indirection for the test list to allow easy
-  // shuffling and restoring the test order.  The i-th element in this
-  // vector is the index of the i-th test in the shuffled test list.
-  std::vector<int> test_indices_;
-  // Pointer to the function that sets up the test case.
-  Test::SetUpTestCaseFunc set_up_tc_;
-  // Pointer to the function that tears down the test case.
-  Test::TearDownTestCaseFunc tear_down_tc_;
-  // True iff any test in this test case should run.
-  bool should_run_;
-  // Elapsed time, in milliseconds.
-  TimeInMillis elapsed_time_;
-  // Holds test properties recorded during execution of SetUpTestCase and
-  // TearDownTestCase.
-  TestResult ad_hoc_test_result_;
-
-  // We disallow copying TestCases.
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(TestCase);
-};
-
-// An Environment object is capable of setting up and tearing down an
-// environment.  You should subclass this to define your own
-// environment(s).
-//
-// An Environment object does the set-up and tear-down in virtual
-// methods SetUp() and TearDown() instead of the constructor and the
-// destructor, as:
-//
-//   1. You cannot safely throw from a destructor.  This is a problem
-//      as in some cases Google Test is used where exceptions are enabled, and
-//      we may want to implement ASSERT_* using exceptions where they are
-//      available.
-//   2. You cannot use ASSERT_* directly in a constructor or
-//      destructor.
-class Environment {
- public:
-  // The d'tor is virtual as we need to subclass Environment.
-  virtual ~Environment() {}
-
-  // Override this to define how to set up the environment.
-  virtual void SetUp() {}
-
-  // Override this to define how to tear down the environment.
-  virtual void TearDown() {}
- private:
-  // If you see an error about overriding the following function or
-  // about it being private, you have mis-spelled SetUp() as Setup().
-  struct Setup_should_be_spelled_SetUp {};
-  virtual Setup_should_be_spelled_SetUp* Setup() { return NULL; }
-};
-
-#if GTEST_HAS_EXCEPTIONS
-
-// Exception which can be thrown from TestEventListener::OnTestPartResult.
-class GTEST_API_ AssertionException
-    : public internal::GoogleTestFailureException {
- public:
-  explicit AssertionException(const TestPartResult& result)
-      : GoogleTestFailureException(result) {}
-};
-
-#endif  // GTEST_HAS_EXCEPTIONS
-
-// The interface for tracing execution of tests. The methods are organized in
-// the order the corresponding events are fired.
-class TestEventListener {
- public:
-  virtual ~TestEventListener() {}
-
-  // Fired before any test activity starts.
-  virtual void OnTestProgramStart(const UnitTest& unit_test) = 0;
-
-  // Fired before each iteration of tests starts.  There may be more than
-  // one iteration if GTEST_FLAG(repeat) is set. iteration is the iteration
-  // index, starting from 0.
-  virtual void OnTestIterationStart(const UnitTest& unit_test,
-                                    int iteration) = 0;
-
-  // Fired before environment set-up for each iteration of tests starts.
-  virtual void OnEnvironmentsSetUpStart(const UnitTest& unit_test) = 0;
-
-  // Fired after environment set-up for each iteration of tests ends.
-  virtual void OnEnvironmentsSetUpEnd(const UnitTest& unit_test) = 0;
-
-  // Fired before the test case starts.
-  virtual void OnTestCaseStart(const TestCase& test_case) = 0;
-
-  // Fired before the test starts.
-  virtual void OnTestStart(const TestInfo& test_info) = 0;
-
-  // Fired after a failed assertion or a SUCCEED() invocation.
-  // If you want to throw an exception from this function to skip to the next
-  // TEST, it must be AssertionException defined above, or inherited from it.
-  virtual void OnTestPartResult(const TestPartResult& test_part_result) = 0;
-
-  // Fired after the test ends.
-  virtual void OnTestEnd(const TestInfo& test_info) = 0;
-
-  // Fired after the test case ends.
-  virtual void OnTestCaseEnd(const TestCase& test_case) = 0;
-
-  // Fired before environment tear-down for each iteration of tests starts.
-  virtual void OnEnvironmentsTearDownStart(const UnitTest& unit_test) = 0;
-
-  // Fired after environment tear-down for each iteration of tests ends.
-  virtual void OnEnvironmentsTearDownEnd(const UnitTest& unit_test) = 0;
-
-  // Fired after each iteration of tests finishes.
-  virtual void OnTestIterationEnd(const UnitTest& unit_test,
-                                  int iteration) = 0;
-
-  // Fired after all test activities have ended.
-  virtual void OnTestProgramEnd(const UnitTest& unit_test) = 0;
-};
-
-// The convenience class for users who need to override just one or two
-// methods and are not concerned that a possible change to a signature of
-// the methods they override will not be caught during the build.  For
-// comments about each method please see the definition of TestEventListener
-// above.
-class EmptyTestEventListener : public TestEventListener {
- public:
-  virtual void OnTestProgramStart(const UnitTest& /*unit_test*/) {}
-  virtual void OnTestIterationStart(const UnitTest& /*unit_test*/,
-                                    int /*iteration*/) {}
-  virtual void OnEnvironmentsSetUpStart(const UnitTest& /*unit_test*/) {}
-  virtual void OnEnvironmentsSetUpEnd(const UnitTest& /*unit_test*/) {}
-  virtual void OnTestCaseStart(const TestCase& /*test_case*/) {}
-  virtual void OnTestStart(const TestInfo& /*test_info*/) {}
-  virtual void OnTestPartResult(const TestPartResult& /*test_part_result*/) {}
-  virtual void OnTestEnd(const TestInfo& /*test_info*/) {}
-  virtual void OnTestCaseEnd(const TestCase& /*test_case*/) {}
-  virtual void OnEnvironmentsTearDownStart(const UnitTest& /*unit_test*/) {}
-  virtual void OnEnvironmentsTearDownEnd(const UnitTest& /*unit_test*/) {}
-  virtual void OnTestIterationEnd(const UnitTest& /*unit_test*/,
-                                  int /*iteration*/) {}
-  virtual void OnTestProgramEnd(const UnitTest& /*unit_test*/) {}
-};
-
-// TestEventListeners lets users add listeners to track events in Google Test.
-class GTEST_API_ TestEventListeners {
- public:
-  TestEventListeners();
-  ~TestEventListeners();
-
-  // Appends an event listener to the end of the list. Google Test assumes
-  // the ownership of the listener (i.e. it will delete the listener when
-  // the test program finishes).
-  void Append(TestEventListener* listener);
-
-  // Removes the given event listener from the list and returns it.  It then
-  // becomes the caller's responsibility to delete the listener. Returns
-  // NULL if the listener is not found in the list.
-  TestEventListener* Release(TestEventListener* listener);
-
-  // Returns the standard listener responsible for the default console
-  // output.  Can be removed from the listeners list to shut down default
-  // console output.  Note that removing this object from the listener list
-  // with Release transfers its ownership to the caller and makes this
-  // function return NULL the next time.
-  TestEventListener* default_result_printer() const {
-    return default_result_printer_;
-  }
-
-  // Returns the standard listener responsible for the default XML output
-  // controlled by the --gtest_output=xml flag.  Can be removed from the
-  // listeners list by users who want to shut down the default XML output
-  // controlled by this flag and substitute it with custom one.  Note that
-  // removing this object from the listener list with Release transfers its
-  // ownership to the caller and makes this function return NULL the next
-  // time.
-  TestEventListener* default_xml_generator() const {
-    return default_xml_generator_;
-  }
-
- private:
-  friend class TestCase;
-  friend class TestInfo;
-  friend class internal::DefaultGlobalTestPartResultReporter;
-  friend class internal::NoExecDeathTest;
-  friend class internal::TestEventListenersAccessor;
-  friend class internal::UnitTestImpl;
-
-  // Returns repeater that broadcasts the TestEventListener events to all
-  // subscribers.
-  TestEventListener* repeater();
-
-  // Sets the default_result_printer attribute to the provided listener.
-  // The listener is also added to the listener list and previous
-  // default_result_printer is removed from it and deleted. The listener can
-  // also be NULL in which case it will not be added to the list. Does
-  // nothing if the previous and the current listener objects are the same.
-  void SetDefaultResultPrinter(TestEventListener* listener);
-
-  // Sets the default_xml_generator attribute to the provided listener.  The
-  // listener is also added to the listener list and previous
-  // default_xml_generator is removed from it and deleted. The listener can
-  // also be NULL in which case it will not be added to the list. Does
-  // nothing if the previous and the current listener objects are the same.
-  void SetDefaultXmlGenerator(TestEventListener* listener);
-
-  // Controls whether events will be forwarded by the repeater to the
-  // listeners in the list.
-  bool EventForwardingEnabled() const;
-  void SuppressEventForwarding();
-
-  // The actual list of listeners.
-  internal::TestEventRepeater* repeater_;
-  // Listener responsible for the standard result output.
-  TestEventListener* default_result_printer_;
-  // Listener responsible for the creation of the XML output file.
-  TestEventListener* default_xml_generator_;
-
-  // We disallow copying TestEventListeners.
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(TestEventListeners);
-};
-
-// A UnitTest consists of a vector of TestCases.
-//
-// This is a singleton class.  The only instance of UnitTest is
-// created when UnitTest::GetInstance() is first called.  This
-// instance is never deleted.
-//
-// UnitTest is not copyable.
-//
-// This class is thread-safe as long as the methods are called
-// according to their specification.
-class GTEST_API_ UnitTest {
- public:
-  // Gets the singleton UnitTest object.  The first time this method
-  // is called, a UnitTest object is constructed and returned.
-  // Consecutive calls will return the same object.
-  static UnitTest* GetInstance();
-
-  // Runs all tests in this UnitTest object and prints the result.
-  // Returns 0 if successful, or 1 otherwise.
-  //
-  // This method can only be called from the main thread.
-  //
-  // INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-  int Run() GTEST_MUST_USE_RESULT_;
-
-  // Returns the working directory when the first TEST() or TEST_F()
-  // was executed.  The UnitTest object owns the string.
-  const char* original_working_dir() const;
-
-  // Returns the TestCase object for the test that's currently running,
-  // or NULL if no test is running.
-  const TestCase* current_test_case() const
-      GTEST_LOCK_EXCLUDED_(mutex_);
-
-  // Returns the TestInfo object for the test that's currently running,
-  // or NULL if no test is running.
-  const TestInfo* current_test_info() const
-      GTEST_LOCK_EXCLUDED_(mutex_);
-
-  // Returns the random seed used at the start of the current test run.
-  int random_seed() const;
-
-  // Returns the ParameterizedTestCaseRegistry object used to keep track of
-  // value-parameterized tests and instantiate and register them.
-  //
-  // INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-  internal::ParameterizedTestCaseRegistry& parameterized_test_registry()
-      GTEST_LOCK_EXCLUDED_(mutex_);
-
-  // Gets the number of successful test cases.
-  int successful_test_case_count() const;
-
-  // Gets the number of failed test cases.
-  int failed_test_case_count() const;
-
-  // Gets the number of all test cases.
-  int total_test_case_count() const;
-
-  // Gets the number of all test cases that contain at least one test
-  // that should run.
-  int test_case_to_run_count() const;
-
-  // Gets the number of successful tests.
-  int successful_test_count() const;
-
-  // Gets the number of failed tests.
-  int failed_test_count() const;
-
-  // Gets the number of disabled tests that will be reported in the XML report.
-  int reportable_disabled_test_count() const;
-
-  // Gets the number of disabled tests.
-  int disabled_test_count() const;
-
-  // Gets the number of tests to be printed in the XML report.
-  int reportable_test_count() const;
-
-  // Gets the number of all tests.
-  int total_test_count() const;
-
-  // Gets the number of tests that should run.
-  int test_to_run_count() const;
-
-  // Gets the time of the test program start, in ms from the start of the
-  // UNIX epoch.
-  TimeInMillis start_timestamp() const;
-
-  // Gets the elapsed time, in milliseconds.
-  TimeInMillis elapsed_time() const;
-
-  // Returns true iff the unit test passed (i.e. all test cases passed).
-  bool Passed() const;
-
-  // Returns true iff the unit test failed (i.e. some test case failed
-  // or something outside of all tests failed).
-  bool Failed() const;
-
-  // Gets the i-th test case among all the test cases. i can range from 0 to
-  // total_test_case_count() - 1. If i is not in that range, returns NULL.
-  const TestCase* GetTestCase(int i) const;
-
-  // Returns the TestResult containing information on test failures and
-  // properties logged outside of individual test cases.
-  const TestResult& ad_hoc_test_result() const;
-
-  // Returns the list of event listeners that can be used to track events
-  // inside Google Test.
-  TestEventListeners& listeners();
-
- private:
-  // Registers and returns a global test environment.  When a test
-  // program is run, all global test environments will be set-up in
-  // the order they were registered.  After all tests in the program
-  // have finished, all global test environments will be torn-down in
-  // the *reverse* order they were registered.
-  //
-  // The UnitTest object takes ownership of the given environment.
-  //
-  // This method can only be called from the main thread.
-  Environment* AddEnvironment(Environment* env);
-
-  // Adds a TestPartResult to the current TestResult object.  All
-  // Google Test assertion macros (e.g. ASSERT_TRUE, EXPECT_EQ, etc)
-  // eventually call this to report their results.  The user code
-  // should use the assertion macros instead of calling this directly.
-  void AddTestPartResult(TestPartResult::Type result_type,
-                         const char* file_name,
-                         int line_number,
-                         const std::string& message,
-                         const std::string& os_stack_trace)
-      GTEST_LOCK_EXCLUDED_(mutex_);
-
-  // Adds a TestProperty to the current TestResult object when invoked from
-  // inside a test, to current TestCase's ad_hoc_test_result_ when invoked
-  // from SetUpTestCase or TearDownTestCase, or to the global property set
-  // when invoked elsewhere.  If the result already contains a property with
-  // the same key, the value will be updated.
-  void RecordProperty(const std::string& key, const std::string& value);
-
-  // Gets the i-th test case among all the test cases. i can range from 0 to
-  // total_test_case_count() - 1. If i is not in that range, returns NULL.
-  TestCase* GetMutableTestCase(int i);
-
-  // Accessors for the implementation object.
-  internal::UnitTestImpl* impl() { return impl_; }
-  const internal::UnitTestImpl* impl() const { return impl_; }
-
-  // These classes and functions are friends as they need to access private
-  // members of UnitTest.
-  friend class ScopedTrace;
-  friend class Test;
-  friend class internal::AssertHelper;
-  friend class internal::StreamingListenerTest;
-  friend class internal::UnitTestRecordPropertyTestHelper;
-  friend Environment* AddGlobalTestEnvironment(Environment* env);
-  friend internal::UnitTestImpl* internal::GetUnitTestImpl();
-  friend void internal::ReportFailureInUnknownLocation(
-      TestPartResult::Type result_type,
-      const std::string& message);
-
-  // Creates an empty UnitTest.
-  UnitTest();
-
-  // D'tor
-  virtual ~UnitTest();
-
-  // Pushes a trace defined by SCOPED_TRACE() on to the per-thread
-  // Google Test trace stack.
-  void PushGTestTrace(const internal::TraceInfo& trace)
-      GTEST_LOCK_EXCLUDED_(mutex_);
-
-  // Pops a trace from the per-thread Google Test trace stack.
-  void PopGTestTrace()
-      GTEST_LOCK_EXCLUDED_(mutex_);
-
-  // Protects mutable state in *impl_.  This is mutable as some const
-  // methods need to lock it too.
-  mutable internal::Mutex mutex_;
-
-  // Opaque implementation object.  This field is never changed once
-  // the object is constructed.  We don't mark it as const here, as
-  // doing so will cause a warning in the constructor of UnitTest.
-  // Mutable state in *impl_ is protected by mutex_.
-  internal::UnitTestImpl* impl_;
-
-  // We disallow copying UnitTest.
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(UnitTest);
-};
-
-// A convenient wrapper for adding an environment for the test
-// program.
-//
-// You should call this before RUN_ALL_TESTS() is called, probably in
-// main().  If you use gtest_main, you need to call this before main()
-// starts for it to take effect.  For example, you can define a global
-// variable like this:
-//
-//   testing::Environment* const foo_env =
-//       testing::AddGlobalTestEnvironment(new FooEnvironment);
-//
-// However, we strongly recommend you to write your own main() and
-// call AddGlobalTestEnvironment() there, as relying on initialization
-// of global variables makes the code harder to read and may cause
-// problems when you register multiple environments from different
-// translation units and the environments have dependencies among them
-// (remember that the compiler doesn't guarantee the order in which
-// global variables from different translation units are initialized).
-inline Environment* AddGlobalTestEnvironment(Environment* env) {
-  return UnitTest::GetInstance()->AddEnvironment(env);
-}
-
-// Initializes Google Test.  This must be called before calling
-// RUN_ALL_TESTS().  In particular, it parses a command line for the
-// flags that Google Test recognizes.  Whenever a Google Test flag is
-// seen, it is removed from argv, and *argc is decremented.
-//
-// No value is returned.  Instead, the Google Test flag variables are
-// updated.
-//
-// Calling the function for the second time has no user-visible effect.
-GTEST_API_ void InitGoogleTest(int* argc, char** argv);
-
-// This overloaded version can be used in Windows programs compiled in
-// UNICODE mode.
-GTEST_API_ void InitGoogleTest(int* argc, wchar_t** argv);
-
-namespace internal {
-
-// Separate the error generating code from the code path to reduce the stack
-// frame size of CmpHelperEQ. This helps reduce the overhead of some sanitizers
-// when calling EXPECT_* in a tight loop.
-template <typename T1, typename T2>
-AssertionResult CmpHelperEQFailure(const char* lhs_expression,
-                                   const char* rhs_expression,
-                                   const T1& lhs, const T2& rhs) {
-  return EqFailure(lhs_expression,
-                   rhs_expression,
-                   FormatForComparisonFailureMessage(lhs, rhs),
-                   FormatForComparisonFailureMessage(rhs, lhs),
-                   false);
-}
-
-// The helper function for {ASSERT|EXPECT}_EQ.
-template <typename T1, typename T2>
-AssertionResult CmpHelperEQ(const char* lhs_expression,
-                            const char* rhs_expression,
-                            const T1& lhs,
-                            const T2& rhs) {
-  if (lhs == rhs) {
-    return AssertionSuccess();
-  }
-
-  return CmpHelperEQFailure(lhs_expression, rhs_expression, lhs, rhs);
-}
-
-// With this overloaded version, we allow anonymous enums to be used
-// in {ASSERT|EXPECT}_EQ when compiled with gcc 4, as anonymous enums
-// can be implicitly cast to BiggestInt.
-GTEST_API_ AssertionResult CmpHelperEQ(const char* lhs_expression,
-                                       const char* rhs_expression,
-                                       BiggestInt lhs,
-                                       BiggestInt rhs);
-
-// The helper class for {ASSERT|EXPECT}_EQ.  The template argument
-// lhs_is_null_literal is true iff the first argument to ASSERT_EQ()
-// is a null pointer literal.  The following default implementation is
-// for lhs_is_null_literal being false.
-template <bool lhs_is_null_literal>
-class EqHelper {
- public:
-  // This templatized version is for the general case.
-  template <typename T1, typename T2>
-  static AssertionResult Compare(const char* lhs_expression,
-                                 const char* rhs_expression,
-                                 const T1& lhs,
-                                 const T2& rhs) {
-    return CmpHelperEQ(lhs_expression, rhs_expression, lhs, rhs);
-  }
-
-  // With this overloaded version, we allow anonymous enums to be used
-  // in {ASSERT|EXPECT}_EQ when compiled with gcc 4, as anonymous
-  // enums can be implicitly cast to BiggestInt.
-  //
-  // Even though its body looks the same as the above version, we
-  // cannot merge the two, as it will make anonymous enums unhappy.
-  static AssertionResult Compare(const char* lhs_expression,
-                                 const char* rhs_expression,
-                                 BiggestInt lhs,
-                                 BiggestInt rhs) {
-    return CmpHelperEQ(lhs_expression, rhs_expression, lhs, rhs);
-  }
-};
-
-// This specialization is used when the first argument to ASSERT_EQ()
-// is a null pointer literal, like NULL, false, or 0.
-template <>
-class EqHelper<true> {
- public:
-  // We define two overloaded versions of Compare().  The first
-  // version will be picked when the second argument to ASSERT_EQ() is
-  // NOT a pointer, e.g. ASSERT_EQ(0, AnIntFunction()) or
-  // EXPECT_EQ(false, a_bool).
-  template <typename T1, typename T2>
-  static AssertionResult Compare(
-      const char* lhs_expression,
-      const char* rhs_expression,
-      const T1& lhs,
-      const T2& rhs,
-      // The following line prevents this overload from being considered if T2
-      // is not a pointer type.  We need this because ASSERT_EQ(NULL, my_ptr)
-      // expands to Compare("", "", NULL, my_ptr), which requires a conversion
-      // to match the Secret* in the other overload, which would otherwise make
-      // this template match better.
-      typename EnableIf<!is_pointer<T2>::value>::type* = 0) {
-    return CmpHelperEQ(lhs_expression, rhs_expression, lhs, rhs);
-  }
-
-  // This version will be picked when the second argument to ASSERT_EQ() is a
-  // pointer, e.g. ASSERT_EQ(NULL, a_pointer).
-  template <typename T>
-  static AssertionResult Compare(
-      const char* lhs_expression,
-      const char* rhs_expression,
-      // We used to have a second template parameter instead of Secret*.  That
-      // template parameter would deduce to 'long', making this a better match
-      // than the first overload even without the first overload's EnableIf.
-      // Unfortunately, gcc with -Wconversion-null warns when "passing NULL to
-      // non-pointer argument" (even a deduced integral argument), so the old
-      // implementation caused warnings in user code.
-      Secret* /* lhs (NULL) */,
-      T* rhs) {
-    // We already know that 'lhs' is a null pointer.
-    return CmpHelperEQ(lhs_expression, rhs_expression,
-                       static_cast<T*>(NULL), rhs);
-  }
-};
-
-// Separate the error generating code from the code path to reduce the stack
-// frame size of CmpHelperOP. This helps reduce the overhead of some sanitizers
-// when calling EXPECT_OP in a tight loop.
-template <typename T1, typename T2>
-AssertionResult CmpHelperOpFailure(const char* expr1, const char* expr2,
-                                   const T1& val1, const T2& val2,
-                                   const char* op) {
-  return AssertionFailure()
-         << "Expected: (" << expr1 << ") " << op << " (" << expr2
-         << "), actual: " << FormatForComparisonFailureMessage(val1, val2)
-         << " vs " << FormatForComparisonFailureMessage(val2, val1);
-}
-
-// A macro for implementing the helper functions needed to implement
-// ASSERT_?? and EXPECT_??.  It is here just to avoid copy-and-paste
-// of similar code.
-//
-// For each templatized helper function, we also define an overloaded
-// version for BiggestInt in order to reduce code bloat and allow
-// anonymous enums to be used with {ASSERT|EXPECT}_?? when compiled
-// with gcc 4.
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-
-#define GTEST_IMPL_CMP_HELPER_(op_name, op)\
-template <typename T1, typename T2>\
-AssertionResult CmpHelper##op_name(const char* expr1, const char* expr2, \
-                                   const T1& val1, const T2& val2) {\
-  if (val1 op val2) {\
-    return AssertionSuccess();\
-  } else {\
-    return CmpHelperOpFailure(expr1, expr2, val1, val2, #op);\
-  }\
-}\
-GTEST_API_ AssertionResult CmpHelper##op_name(\
-    const char* expr1, const char* expr2, BiggestInt val1, BiggestInt val2)
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-
-// Implements the helper function for {ASSERT|EXPECT}_NE
-GTEST_IMPL_CMP_HELPER_(NE, !=);
-// Implements the helper function for {ASSERT|EXPECT}_LE
-GTEST_IMPL_CMP_HELPER_(LE, <=);
-// Implements the helper function for {ASSERT|EXPECT}_LT
-GTEST_IMPL_CMP_HELPER_(LT, <);
-// Implements the helper function for {ASSERT|EXPECT}_GE
-GTEST_IMPL_CMP_HELPER_(GE, >=);
-// Implements the helper function for {ASSERT|EXPECT}_GT
-GTEST_IMPL_CMP_HELPER_(GT, >);
-
-#undef GTEST_IMPL_CMP_HELPER_
-
-// The helper function for {ASSERT|EXPECT}_STREQ.
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-GTEST_API_ AssertionResult CmpHelperSTREQ(const char* s1_expression,
-                                          const char* s2_expression,
-                                          const char* s1,
-                                          const char* s2);
-
-// The helper function for {ASSERT|EXPECT}_STRCASEEQ.
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-GTEST_API_ AssertionResult CmpHelperSTRCASEEQ(const char* s1_expression,
-                                              const char* s2_expression,
-                                              const char* s1,
-                                              const char* s2);
-
-// The helper function for {ASSERT|EXPECT}_STRNE.
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-GTEST_API_ AssertionResult CmpHelperSTRNE(const char* s1_expression,
-                                          const char* s2_expression,
-                                          const char* s1,
-                                          const char* s2);
-
-// The helper function for {ASSERT|EXPECT}_STRCASENE.
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-GTEST_API_ AssertionResult CmpHelperSTRCASENE(const char* s1_expression,
-                                              const char* s2_expression,
-                                              const char* s1,
-                                              const char* s2);
-
-
-// Helper function for *_STREQ on wide strings.
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-GTEST_API_ AssertionResult CmpHelperSTREQ(const char* s1_expression,
-                                          const char* s2_expression,
-                                          const wchar_t* s1,
-                                          const wchar_t* s2);
-
-// Helper function for *_STRNE on wide strings.
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-GTEST_API_ AssertionResult CmpHelperSTRNE(const char* s1_expression,
-                                          const char* s2_expression,
-                                          const wchar_t* s1,
-                                          const wchar_t* s2);
-
-}  // namespace internal
-
-// IsSubstring() and IsNotSubstring() are intended to be used as the
-// first argument to {EXPECT,ASSERT}_PRED_FORMAT2(), not by
-// themselves.  They check whether needle is a substring of haystack
-// (NULL is considered a substring of itself only), and return an
-// appropriate error message when they fail.
-//
-// The {needle,haystack}_expr arguments are the stringified
-// expressions that generated the two real arguments.
-GTEST_API_ AssertionResult IsSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const char* needle, const char* haystack);
-GTEST_API_ AssertionResult IsSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const wchar_t* needle, const wchar_t* haystack);
-GTEST_API_ AssertionResult IsNotSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const char* needle, const char* haystack);
-GTEST_API_ AssertionResult IsNotSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const wchar_t* needle, const wchar_t* haystack);
-GTEST_API_ AssertionResult IsSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const ::std::string& needle, const ::std::string& haystack);
-GTEST_API_ AssertionResult IsNotSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const ::std::string& needle, const ::std::string& haystack);
-
-#if GTEST_HAS_STD_WSTRING
-GTEST_API_ AssertionResult IsSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const ::std::wstring& needle, const ::std::wstring& haystack);
-GTEST_API_ AssertionResult IsNotSubstring(
-    const char* needle_expr, const char* haystack_expr,
-    const ::std::wstring& needle, const ::std::wstring& haystack);
-#endif  // GTEST_HAS_STD_WSTRING
-
-namespace internal {
-
-// Helper template function for comparing floating-points.
-//
-// Template parameter:
-//
-//   RawType: the raw floating-point type (either float or double)
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-template <typename RawType>
-AssertionResult CmpHelperFloatingPointEQ(const char* lhs_expression,
-                                         const char* rhs_expression,
-                                         RawType lhs_value,
-                                         RawType rhs_value) {
-  const FloatingPoint<RawType> lhs(lhs_value), rhs(rhs_value);
-
-  if (lhs.AlmostEquals(rhs)) {
-    return AssertionSuccess();
-  }
-
-  ::std::stringstream lhs_ss;
-  lhs_ss << std::setprecision(std::numeric_limits<RawType>::digits10 + 2)
-         << lhs_value;
-
-  ::std::stringstream rhs_ss;
-  rhs_ss << std::setprecision(std::numeric_limits<RawType>::digits10 + 2)
-         << rhs_value;
-
-  return EqFailure(lhs_expression,
-                   rhs_expression,
-                   StringStreamToString(&lhs_ss),
-                   StringStreamToString(&rhs_ss),
-                   false);
-}
-
-// Helper function for implementing ASSERT_NEAR.
-//
-// INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-GTEST_API_ AssertionResult DoubleNearPredFormat(const char* expr1,
-                                                const char* expr2,
-                                                const char* abs_error_expr,
-                                                double val1,
-                                                double val2,
-                                                double abs_error);
-
-// INTERNAL IMPLEMENTATION - DO NOT USE IN USER CODE.
-// A class that enables one to stream messages to assertion macros
-class GTEST_API_ AssertHelper {
- public:
-  // Constructor.
-  AssertHelper(TestPartResult::Type type,
-               const char* file,
-               int line,
-               const char* message);
-  ~AssertHelper();
-
-  // Message assignment is a semantic trick to enable assertion
-  // streaming; see the GTEST_MESSAGE_ macro below.
-  void operator=(const Message& message) const;
-
- private:
-  // We put our data in a struct so that the size of the AssertHelper class can
-  // be as small as possible.  This is important because gcc is incapable of
-  // re-using stack space even for temporary variables, so every EXPECT_EQ
-  // reserves stack space for another AssertHelper.
-  struct AssertHelperData {
-    AssertHelperData(TestPartResult::Type t,
-                     const char* srcfile,
-                     int line_num,
-                     const char* msg)
-        : type(t), file(srcfile), line(line_num), message(msg) { }
-
-    TestPartResult::Type const type;
-    const char* const file;
-    int const line;
-    std::string const message;
-
-   private:
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(AssertHelperData);
-  };
-
-  AssertHelperData* const data_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(AssertHelper);
-};
-
-}  // namespace internal
-
-// The pure interface class that all value-parameterized tests inherit from.
-// A value-parameterized class must inherit from both ::testing::Test and
-// ::testing::WithParamInterface. In most cases that just means inheriting
-// from ::testing::TestWithParam, but more complicated test hierarchies
-// may need to inherit from Test and WithParamInterface at different levels.
-//
-// This interface has support for accessing the test parameter value via
-// the GetParam() method.
-//
-// Use it with one of the parameter generator defining functions, like Range(),
-// Values(), ValuesIn(), Bool(), and Combine().
-//
-// class FooTest : public ::testing::TestWithParam<int> {
-//  protected:
-//   FooTest() {
-//     // Can use GetParam() here.
-//   }
-//   virtual ~FooTest() {
-//     // Can use GetParam() here.
-//   }
-//   virtual void SetUp() {
-//     // Can use GetParam() here.
-//   }
-//   virtual void TearDown {
-//     // Can use GetParam() here.
-//   }
-// };
-// TEST_P(FooTest, DoesBar) {
-//   // Can use GetParam() method here.
-//   Foo foo;
-//   ASSERT_TRUE(foo.DoesBar(GetParam()));
-// }
-// INSTANTIATE_TEST_CASE_P(OneToTenRange, FooTest, ::testing::Range(1, 10));
-
-template <typename T>
-class WithParamInterface {
- public:
-  typedef T ParamType;
-  virtual ~WithParamInterface() {}
-
-  // The current parameter value. Is also available in the test fixture's
-  // constructor. This member function is non-static, even though it only
-  // references static data, to reduce the opportunity for incorrect uses
-  // like writing 'WithParamInterface<bool>::GetParam()' for a test that
-  // uses a fixture whose parameter type is int.
-  const ParamType& GetParam() const {
-    GTEST_CHECK_(parameter_ != NULL)
-        << "GetParam() can only be called inside a value-parameterized test "
-        << "-- did you intend to write TEST_P instead of TEST_F?";
-    return *parameter_;
-  }
-
- private:
-  // Sets parameter value. The caller is responsible for making sure the value
-  // remains alive and unchanged throughout the current test.
-  static void SetParam(const ParamType* parameter) {
-    parameter_ = parameter;
-  }
-
-  // Static value used for accessing parameter during a test lifetime.
-  static const ParamType* parameter_;
-
-  // TestClass must be a subclass of WithParamInterface<T> and Test.
-  template <class TestClass> friend class internal::ParameterizedTestFactory;
-};
-
-template <typename T>
-const T* WithParamInterface<T>::parameter_ = NULL;
-
-// Most value-parameterized classes can ignore the existence of
-// WithParamInterface, and can just inherit from ::testing::TestWithParam.
-
-template <typename T>
-class TestWithParam : public Test, public WithParamInterface<T> {
-};
-
-// Macros for indicating success/failure in test code.
-
-// ADD_FAILURE unconditionally adds a failure to the current test.
-// SUCCEED generates a success - it doesn't automatically make the
-// current test successful, as a test is only successful when it has
-// no failure.
-//
-// EXPECT_* verifies that a certain condition is satisfied.  If not,
-// it behaves like ADD_FAILURE.  In particular:
-//
-//   EXPECT_TRUE  verifies that a Boolean condition is true.
-//   EXPECT_FALSE verifies that a Boolean condition is false.
-//
-// FAIL and ASSERT_* are similar to ADD_FAILURE and EXPECT_*, except
-// that they will also abort the current function on failure.  People
-// usually want the fail-fast behavior of FAIL and ASSERT_*, but those
-// writing data-driven tests often find themselves using ADD_FAILURE
-// and EXPECT_* more.
-
-// Generates a nonfatal failure with a generic message.
-#define ADD_FAILURE() GTEST_NONFATAL_FAILURE_("Failed")
-
-// Generates a nonfatal failure at the given source file location with
-// a generic message.
-#define ADD_FAILURE_AT(file, line) \
-  GTEST_MESSAGE_AT_(file, line, "Failed", \
-                    ::testing::TestPartResult::kNonFatalFailure)
-
-// Generates a fatal failure with a generic message.
-#define GTEST_FAIL() GTEST_FATAL_FAILURE_("Failed")
-
-// Define this macro to 1 to omit the definition of FAIL(), which is a
-// generic name and clashes with some other libraries.
-#if !GTEST_DONT_DEFINE_FAIL
-# define FAIL() GTEST_FAIL()
-#endif
-
-// Generates a success with a generic message.
-#define GTEST_SUCCEED() GTEST_SUCCESS_("Succeeded")
-
-// Define this macro to 1 to omit the definition of SUCCEED(), which
-// is a generic name and clashes with some other libraries.
-#if !GTEST_DONT_DEFINE_SUCCEED
-# define SUCCEED() GTEST_SUCCEED()
-#endif
-
-// Macros for testing exceptions.
-//
-//    * {ASSERT|EXPECT}_THROW(statement, expected_exception):
-//         Tests that the statement throws the expected exception.
-//    * {ASSERT|EXPECT}_NO_THROW(statement):
-//         Tests that the statement doesn't throw any exception.
-//    * {ASSERT|EXPECT}_ANY_THROW(statement):
-//         Tests that the statement throws an exception.
-
-#define EXPECT_THROW(statement, expected_exception) \
-  GTEST_TEST_THROW_(statement, expected_exception, GTEST_NONFATAL_FAILURE_)
-#define EXPECT_NO_THROW(statement) \
-  GTEST_TEST_NO_THROW_(statement, GTEST_NONFATAL_FAILURE_)
-#define EXPECT_ANY_THROW(statement) \
-  GTEST_TEST_ANY_THROW_(statement, GTEST_NONFATAL_FAILURE_)
-#define ASSERT_THROW(statement, expected_exception) \
-  GTEST_TEST_THROW_(statement, expected_exception, GTEST_FATAL_FAILURE_)
-#define ASSERT_NO_THROW(statement) \
-  GTEST_TEST_NO_THROW_(statement, GTEST_FATAL_FAILURE_)
-#define ASSERT_ANY_THROW(statement) \
-  GTEST_TEST_ANY_THROW_(statement, GTEST_FATAL_FAILURE_)
-
-// Boolean assertions. Condition can be either a Boolean expression or an
-// AssertionResult. For more information on how to use AssertionResult with
-// these macros see comments on that class.
-#define EXPECT_TRUE(condition) \
-  GTEST_TEST_BOOLEAN_(condition, #condition, false, true, \
-                      GTEST_NONFATAL_FAILURE_)
-#define EXPECT_FALSE(condition) \
-  GTEST_TEST_BOOLEAN_(!(condition), #condition, true, false, \
-                      GTEST_NONFATAL_FAILURE_)
-#define ASSERT_TRUE(condition) \
-  GTEST_TEST_BOOLEAN_(condition, #condition, false, true, \
-                      GTEST_FATAL_FAILURE_)
-#define ASSERT_FALSE(condition) \
-  GTEST_TEST_BOOLEAN_(!(condition), #condition, true, false, \
-                      GTEST_FATAL_FAILURE_)
-
-// Macros for testing equalities and inequalities.
-//
-//    * {ASSERT|EXPECT}_EQ(v1, v2): Tests that v1 == v2
-//    * {ASSERT|EXPECT}_NE(v1, v2): Tests that v1 != v2
-//    * {ASSERT|EXPECT}_LT(v1, v2): Tests that v1 < v2
-//    * {ASSERT|EXPECT}_LE(v1, v2): Tests that v1 <= v2
-//    * {ASSERT|EXPECT}_GT(v1, v2): Tests that v1 > v2
-//    * {ASSERT|EXPECT}_GE(v1, v2): Tests that v1 >= v2
-//
-// When they are not, Google Test prints both the tested expressions and
-// their actual values.  The values must be compatible built-in types,
-// or you will get a compiler error.  By "compatible" we mean that the
-// values can be compared by the respective operator.
-//
-// Note:
-//
-//   1. It is possible to make a user-defined type work with
-//   {ASSERT|EXPECT}_??(), but that requires overloading the
-//   comparison operators and is thus discouraged by the Google C++
-//   Usage Guide.  Therefore, you are advised to use the
-//   {ASSERT|EXPECT}_TRUE() macro to assert that two objects are
-//   equal.
-//
-//   2. The {ASSERT|EXPECT}_??() macros do pointer comparisons on
-//   pointers (in particular, C strings).  Therefore, if you use it
-//   with two C strings, you are testing how their locations in memory
-//   are related, not how their content is related.  To compare two C
-//   strings by content, use {ASSERT|EXPECT}_STR*().
-//
-//   3. {ASSERT|EXPECT}_EQ(v1, v2) is preferred to
-//   {ASSERT|EXPECT}_TRUE(v1 == v2), as the former tells you
-//   what the actual value is when it fails, and similarly for the
-//   other comparisons.
-//
-//   4. Do not depend on the order in which {ASSERT|EXPECT}_??()
-//   evaluate their arguments, which is undefined.
-//
-//   5. These macros evaluate their arguments exactly once.
-//
-// Examples:
-//
-//   EXPECT_NE(Foo(), 5);
-//   EXPECT_EQ(a_pointer, NULL);
-//   ASSERT_LT(i, array_size);
-//   ASSERT_GT(records.size(), 0) << "There is no record left.";
-
-#define EXPECT_EQ(val1, val2) \
-  EXPECT_PRED_FORMAT2(::testing::internal:: \
-                      EqHelper<GTEST_IS_NULL_LITERAL_(val1)>::Compare, \
-                      val1, val2)
-#define EXPECT_NE(val1, val2) \
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperNE, val1, val2)
-#define EXPECT_LE(val1, val2) \
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperLE, val1, val2)
-#define EXPECT_LT(val1, val2) \
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperLT, val1, val2)
-#define EXPECT_GE(val1, val2) \
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperGE, val1, val2)
-#define EXPECT_GT(val1, val2) \
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperGT, val1, val2)
-
-#define GTEST_ASSERT_EQ(val1, val2) \
-  ASSERT_PRED_FORMAT2(::testing::internal:: \
-                      EqHelper<GTEST_IS_NULL_LITERAL_(val1)>::Compare, \
-                      val1, val2)
-#define GTEST_ASSERT_NE(val1, val2) \
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperNE, val1, val2)
-#define GTEST_ASSERT_LE(val1, val2) \
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperLE, val1, val2)
-#define GTEST_ASSERT_LT(val1, val2) \
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperLT, val1, val2)
-#define GTEST_ASSERT_GE(val1, val2) \
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperGE, val1, val2)
-#define GTEST_ASSERT_GT(val1, val2) \
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperGT, val1, val2)
-
-// Define macro GTEST_DONT_DEFINE_ASSERT_XY to 1 to omit the definition of
-// ASSERT_XY(), which clashes with some users' own code.
-
-#if !GTEST_DONT_DEFINE_ASSERT_EQ
-# define ASSERT_EQ(val1, val2) GTEST_ASSERT_EQ(val1, val2)
-#endif
-
-#if !GTEST_DONT_DEFINE_ASSERT_NE
-# define ASSERT_NE(val1, val2) GTEST_ASSERT_NE(val1, val2)
-#endif
-
-#if !GTEST_DONT_DEFINE_ASSERT_LE
-# define ASSERT_LE(val1, val2) GTEST_ASSERT_LE(val1, val2)
-#endif
-
-#if !GTEST_DONT_DEFINE_ASSERT_LT
-# define ASSERT_LT(val1, val2) GTEST_ASSERT_LT(val1, val2)
-#endif
-
-#if !GTEST_DONT_DEFINE_ASSERT_GE
-# define ASSERT_GE(val1, val2) GTEST_ASSERT_GE(val1, val2)
-#endif
-
-#if !GTEST_DONT_DEFINE_ASSERT_GT
-# define ASSERT_GT(val1, val2) GTEST_ASSERT_GT(val1, val2)
-#endif
-
-// C-string Comparisons.  All tests treat NULL and any non-NULL string
-// as different.  Two NULLs are equal.
-//
-//    * {ASSERT|EXPECT}_STREQ(s1, s2):     Tests that s1 == s2
-//    * {ASSERT|EXPECT}_STRNE(s1, s2):     Tests that s1 != s2
-//    * {ASSERT|EXPECT}_STRCASEEQ(s1, s2): Tests that s1 == s2, ignoring case
-//    * {ASSERT|EXPECT}_STRCASENE(s1, s2): Tests that s1 != s2, ignoring case
-//
-// For wide or narrow string objects, you can use the
-// {ASSERT|EXPECT}_??() macros.
-//
-// Don't depend on the order in which the arguments are evaluated,
-// which is undefined.
-//
-// These macros evaluate their arguments exactly once.
-
-#define EXPECT_STREQ(s1, s2) \
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperSTREQ, s1, s2)
-#define EXPECT_STRNE(s1, s2) \
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperSTRNE, s1, s2)
-#define EXPECT_STRCASEEQ(s1, s2) \
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperSTRCASEEQ, s1, s2)
-#define EXPECT_STRCASENE(s1, s2)\
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperSTRCASENE, s1, s2)
-
-#define ASSERT_STREQ(s1, s2) \
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperSTREQ, s1, s2)
-#define ASSERT_STRNE(s1, s2) \
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperSTRNE, s1, s2)
-#define ASSERT_STRCASEEQ(s1, s2) \
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperSTRCASEEQ, s1, s2)
-#define ASSERT_STRCASENE(s1, s2)\
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperSTRCASENE, s1, s2)
-
-// Macros for comparing floating-point numbers.
-//
-//    * {ASSERT|EXPECT}_FLOAT_EQ(val1, val2):
-//         Tests that two float values are almost equal.
-//    * {ASSERT|EXPECT}_DOUBLE_EQ(val1, val2):
-//         Tests that two double values are almost equal.
-//    * {ASSERT|EXPECT}_NEAR(v1, v2, abs_error):
-//         Tests that v1 and v2 are within the given distance to each other.
-//
-// Google Test uses ULP-based comparison to automatically pick a default
-// error bound that is appropriate for the operands.  See the
-// FloatingPoint template class in gtest-internal.h if you are
-// interested in the implementation details.
-
-#define EXPECT_FLOAT_EQ(val1, val2)\
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperFloatingPointEQ<float>, \
-                      val1, val2)
-
-#define EXPECT_DOUBLE_EQ(val1, val2)\
-  EXPECT_PRED_FORMAT2(::testing::internal::CmpHelperFloatingPointEQ<double>, \
-                      val1, val2)
-
-#define ASSERT_FLOAT_EQ(val1, val2)\
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperFloatingPointEQ<float>, \
-                      val1, val2)
-
-#define ASSERT_DOUBLE_EQ(val1, val2)\
-  ASSERT_PRED_FORMAT2(::testing::internal::CmpHelperFloatingPointEQ<double>, \
-                      val1, val2)
-
-#define EXPECT_NEAR(val1, val2, abs_error)\
-  EXPECT_PRED_FORMAT3(::testing::internal::DoubleNearPredFormat, \
-                      val1, val2, abs_error)
-
-#define ASSERT_NEAR(val1, val2, abs_error)\
-  ASSERT_PRED_FORMAT3(::testing::internal::DoubleNearPredFormat, \
-                      val1, val2, abs_error)
-
-// These predicate format functions work on floating-point values, and
-// can be used in {ASSERT|EXPECT}_PRED_FORMAT2*(), e.g.
-//
-//   EXPECT_PRED_FORMAT2(testing::DoubleLE, Foo(), 5.0);
-
-// Asserts that val1 is less than, or almost equal to, val2.  Fails
-// otherwise.  In particular, it fails if either val1 or val2 is NaN.
-GTEST_API_ AssertionResult FloatLE(const char* expr1, const char* expr2,
-                                   float val1, float val2);
-GTEST_API_ AssertionResult DoubleLE(const char* expr1, const char* expr2,
-                                    double val1, double val2);
-
-
-#if GTEST_OS_WINDOWS
-
-// Macros that test for HRESULT failure and success, these are only useful
-// on Windows, and rely on Windows SDK macros and APIs to compile.
-//
-//    * {ASSERT|EXPECT}_HRESULT_{SUCCEEDED|FAILED}(expr)
-//
-// When expr unexpectedly fails or succeeds, Google Test prints the
-// expected result and the actual result with both a human-readable
-// string representation of the error, if available, as well as the
-// hex result code.
-# define EXPECT_HRESULT_SUCCEEDED(expr) \
-    EXPECT_PRED_FORMAT1(::testing::internal::IsHRESULTSuccess, (expr))
-
-# define ASSERT_HRESULT_SUCCEEDED(expr) \
-    ASSERT_PRED_FORMAT1(::testing::internal::IsHRESULTSuccess, (expr))
-
-# define EXPECT_HRESULT_FAILED(expr) \
-    EXPECT_PRED_FORMAT1(::testing::internal::IsHRESULTFailure, (expr))
-
-# define ASSERT_HRESULT_FAILED(expr) \
-    ASSERT_PRED_FORMAT1(::testing::internal::IsHRESULTFailure, (expr))
-
-#endif  // GTEST_OS_WINDOWS
-
-// Macros that execute statement and check that it doesn't generate new fatal
-// failures in the current thread.
-//
-//   * {ASSERT|EXPECT}_NO_FATAL_FAILURE(statement);
-//
-// Examples:
-//
-//   EXPECT_NO_FATAL_FAILURE(Process());
-//   ASSERT_NO_FATAL_FAILURE(Process()) << "Process() failed";
-//
-#define ASSERT_NO_FATAL_FAILURE(statement) \
-    GTEST_TEST_NO_FATAL_FAILURE_(statement, GTEST_FATAL_FAILURE_)
-#define EXPECT_NO_FATAL_FAILURE(statement) \
-    GTEST_TEST_NO_FATAL_FAILURE_(statement, GTEST_NONFATAL_FAILURE_)
-
-// Causes a trace (including the given source file path and line number,
-// and the given message) to be included in every test failure message generated
-// by code in the scope of the lifetime of an instance of this class. The effect
-// is undone with the destruction of the instance.
-//
-// The message argument can be anything streamable to std::ostream.
-//
-// Example:
-//   testing::ScopedTrace trace("file.cc", 123, "message");
-//
-class GTEST_API_ ScopedTrace {
- public:
-  // The c'tor pushes the given source file location and message onto
-  // a trace stack maintained by Google Test.
-
-  // Template version. Uses Message() to convert the values into strings.
-  // Slow, but flexible.
-  template <typename T>
-  ScopedTrace(const char* file, int line, const T& message) {
-    PushTrace(file, line, (Message() << message).GetString());
-  }
-
-  // Optimize for some known types.
-  ScopedTrace(const char* file, int line, const char* message) {
-    PushTrace(file, line, message ? message : "(null)");
-  }
-
-#if GTEST_HAS_GLOBAL_STRING
-  ScopedTrace(const char* file, int line, const ::string& message) {
-    PushTrace(file, line, message);
-  }
-#endif
-
-  ScopedTrace(const char* file, int line, const std::string& message) {
-    PushTrace(file, line, message);
-  }
-
-  // The d'tor pops the info pushed by the c'tor.
-  //
-  // Note that the d'tor is not virtual in order to be efficient.
-  // Don't inherit from ScopedTrace!
-  ~ScopedTrace();
-
- private:
-  void PushTrace(const char* file, int line, std::string message);
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(ScopedTrace);
-} GTEST_ATTRIBUTE_UNUSED_;  // A ScopedTrace object does its job in its
-                            // c'tor and d'tor.  Therefore it doesn't
-                            // need to be used otherwise.
-
-// Causes a trace (including the source file path, the current line
-// number, and the given message) to be included in every test failure
-// message generated by code in the current scope.  The effect is
-// undone when the control leaves the current scope.
-//
-// The message argument can be anything streamable to std::ostream.
-//
-// In the implementation, we include the current line number as part
-// of the dummy variable name, thus allowing multiple SCOPED_TRACE()s
-// to appear in the same block - as long as they are on different
-// lines.
-//
-// Assuming that each thread maintains its own stack of traces.
-// Therefore, a SCOPED_TRACE() would (correctly) only affect the
-// assertions in its own thread.
-#define SCOPED_TRACE(message) \
-  ::testing::ScopedTrace GTEST_CONCAT_TOKEN_(gtest_trace_, __LINE__)(\
-    __FILE__, __LINE__, (message))
-
-
-// Compile-time assertion for type equality.
-// StaticAssertTypeEq<type1, type2>() compiles iff type1 and type2 are
-// the same type.  The value it returns is not interesting.
-//
-// Instead of making StaticAssertTypeEq a class template, we make it a
-// function template that invokes a helper class template.  This
-// prevents a user from misusing StaticAssertTypeEq<T1, T2> by
-// defining objects of that type.
-//
-// CAVEAT:
-//
-// When used inside a method of a class template,
-// StaticAssertTypeEq<T1, T2>() is effective ONLY IF the method is
-// instantiated.  For example, given:
-//
-//   template <typename T> class Foo {
-//    public:
-//     void Bar() { testing::StaticAssertTypeEq<int, T>(); }
-//   };
-//
-// the code:
-//
-//   void Test1() { Foo<bool> foo; }
-//
-// will NOT generate a compiler error, as Foo<bool>::Bar() is never
-// actually instantiated.  Instead, you need:
-//
-//   void Test2() { Foo<bool> foo; foo.Bar(); }
-//
-// to cause a compiler error.
-template <typename T1, typename T2>
-bool StaticAssertTypeEq() {
-  (void)internal::StaticAssertTypeEqHelper<T1, T2>();
-  return true;
-}
-
-// Defines a test.
-//
-// The first parameter is the name of the test case, and the second
-// parameter is the name of the test within the test case.
-//
-// The convention is to end the test case name with "Test".  For
-// example, a test case for the Foo class can be named FooTest.
-//
-// Test code should appear between braces after an invocation of
-// this macro.  Example:
-//
-//   TEST(FooTest, InitializesCorrectly) {
-//     Foo foo;
-//     EXPECT_TRUE(foo.StatusIsOK());
-//   }
-
-// Note that we call GetTestTypeId() instead of GetTypeId<
-// ::testing::Test>() here to get the type ID of testing::Test.  This
-// is to work around a suspected linker bug when using Google Test as
-// a framework on Mac OS X.  The bug causes GetTypeId<
-// ::testing::Test>() to return different values depending on whether
-// the call is from the Google Test framework itself or from user test
-// code.  GetTestTypeId() is guaranteed to always return the same
-// value, as it always calls GetTypeId<>() from the Google Test
-// framework.
-#define GTEST_TEST(test_case_name, test_name)\
-  GTEST_TEST_(test_case_name, test_name, \
-              ::testing::Test, ::testing::internal::GetTestTypeId())
-
-// Define this macro to 1 to omit the definition of TEST(), which
-// is a generic name and clashes with some other libraries.
-#if !GTEST_DONT_DEFINE_TEST
-# define TEST(test_case_name, test_name) GTEST_TEST(test_case_name, test_name)
-#endif
-
-// Defines a test that uses a test fixture.
-//
-// The first parameter is the name of the test fixture class, which
-// also doubles as the test case name.  The second parameter is the
-// name of the test within the test case.
-//
-// A test fixture class must be declared earlier.  The user should put
-// the test code between braces after using this macro.  Example:
-//
-//   class FooTest : public testing::Test {
-//    protected:
-//     virtual void SetUp() { b_.AddElement(3); }
-//
-//     Foo a_;
-//     Foo b_;
-//   };
-//
-//   TEST_F(FooTest, InitializesCorrectly) {
-//     EXPECT_TRUE(a_.StatusIsOK());
-//   }
-//
-//   TEST_F(FooTest, ReturnsElementCountCorrectly) {
-//     EXPECT_EQ(a_.size(), 0);
-//     EXPECT_EQ(b_.size(), 1);
-//   }
-
-#define TEST_F(test_fixture, test_name)\
-  GTEST_TEST_(test_fixture, test_name, test_fixture, \
-              ::testing::internal::GetTypeId<test_fixture>())
-
-// Returns a path to temporary directory.
-// Tries to determine an appropriate directory for the platform.
-GTEST_API_ std::string TempDir();
-
-#ifdef _MSC_VER
-#  pragma warning(pop)
-#endif
-
-}  // namespace testing
-
-// Use this function in main() to run all tests.  It returns 0 if all
-// tests are successful, or 1 otherwise.
-//
-// RUN_ALL_TESTS() should be invoked after the command line has been
-// parsed by InitGoogleTest().
-//
-// This function was formerly a macro; thus, it is in the global
-// namespace and has an all-caps name.
-int RUN_ALL_TESTS() GTEST_MUST_USE_RESULT_;
-
-inline int RUN_ALL_TESTS() {
-  return ::testing::UnitTest::GetInstance()->Run();
-}
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_H_
diff --git a/inference-engine/thirdparty/clDNN/common/googletest-fused/version.txt b/inference-engine/thirdparty/clDNN/common/googletest-fused/version.txt
deleted file mode 100644
index 78d13dd5cae94c..00000000000000
--- a/inference-engine/thirdparty/clDNN/common/googletest-fused/version.txt
+++ /dev/null
@@ -1,4 +0,0 @@
-Source:             https://github.com/google/googletest.git
-Latest commit:      78b49242d1def8c7fd884b81901f589c7b3d8db2
-Latest commit date: 2018-07-24
-Latest fuse date:   2018-07-24
\ No newline at end of file
diff --git a/inference-engine/thirdparty/clDNN/tests/CMakeLists.txt b/inference-engine/thirdparty/clDNN/tests/CMakeLists.txt
index 00011e54075fa6..583a88ce1ab898 100644
--- a/inference-engine/thirdparty/clDNN/tests/CMakeLists.txt
+++ b/inference-engine/thirdparty/clDNN/tests/CMakeLists.txt
@@ -55,17 +55,6 @@ file(GLOB __CLDNN_Sources__module_tests
     "${__CLDNN_Directory__module_tests}/*.cpp"
   )
 
-set(__CLDNN_Directory__gtest           "${CLDNN__GTEST_DIR}")
-set(__CLDNN_Label__gtest               "google test framework")
-file(GLOB __CLDNN_Sources__gtest
-    "${__CLDNN_Directory__gtest}/*.cc"
-  )
-
-if(UNIX)
-    set_source_files_properties(${__CLDNN_Sources__gtest} PROPERTIES
-        COMPILE_FLAGS "-Wno-undef")
-endif()
-
 if (MSVC)
   file(GLOB __CLDNN_Sources__natvis
     "${CMAKE_CURRENT_SOURCE_DIR}/float16.natvis"
@@ -84,7 +73,6 @@ set(__CLDNN_AllSources
     ${__CLDNN_Sources__test_cases}
     ${__CLDNN_Sources__test_utils}
     ${__CLDNN_Sources__module_tests}
-    ${__CLDNN_Sources__gtest}
     ${__CLDNN_Sources__natvis}
   )
 
@@ -94,7 +82,6 @@ source_group("${__CLDNN_Label__main}"         FILES ${__CLDNN_Sources__main})
 source_group("${__CLDNN_Label__test_cases}"   FILES ${__CLDNN_Sources__test_cases})
 source_group("${__CLDNN_Label__test_utils}"   FILES ${__CLDNN_Sources__test_utils})
 source_group("${__CLDNN_Label__module_tests}" FILES ${__CLDNN_Sources__module_tests})
-source_group("${__CLDNN_Label__gtest}"        FILES ${__CLDNN_Sources__gtest})
 
 # ===================================== Include/Link directories =======================================
 
@@ -111,9 +98,6 @@ include_directories(
     "${__CLDNN_Directory__ks_common}"
   )
 
-include_directories(SYSTEM
-    "${CLDNN__GTEST_DIR}")
-
 # =================================== Link targets and dependencies ====================================
 
 # Tests executable.
@@ -129,7 +113,7 @@ set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY PROJECT_LABEL "${CLDNN_BUILD
 set_property(TARGET "${CLDNN_BUILD__PROJ}" PROPERTY OUTPUT_NAME   "${CLDNN_BUILD__PROJ_OUTPUT_NAME}")
 
 # Set library dependencies
-target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE "${CLDNN_BUILD__PROJ__clDNN}")
+target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE "${CLDNN_BUILD__PROJ__clDNN}" gtest gmock)
 
 if(WIN32)
   target_link_libraries("${CLDNN_BUILD__PROJ}" PRIVATE setupapi)
diff --git a/inference-engine/thirdparty/clDNN/tests/module_tests/test_uqr_distribution.cpp b/inference-engine/thirdparty/clDNN/tests/module_tests/test_uqr_distribution.cpp
index 0b85f0b1c1c973..1a87db1cab38b3 100644
--- a/inference-engine/thirdparty/clDNN/tests/module_tests/test_uqr_distribution.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/module_tests/test_uqr_distribution.cpp
@@ -39,7 +39,7 @@ struct uniform_quantized_real_distribution_test : ::testing::Test
 };
 
 using uniform_quantized_real_distribution_test_types = ::testing::Types<void, float, double, long double>;
-TYPED_TEST_CASE(uniform_quantized_real_distribution_test, uniform_quantized_real_distribution_test_types);
+TYPED_TEST_SUITE(uniform_quantized_real_distribution_test, uniform_quantized_real_distribution_test_types);
 
 TYPED_TEST(uniform_quantized_real_distribution_test, param_construct_default)
 {
diff --git a/inference-engine/thirdparty/clDNN/tests/module_tests/usm_memory_test.cpp b/inference-engine/thirdparty/clDNN/tests/module_tests/usm_memory_test.cpp
index c67b2516fc78bf..5a00cb56e4e4e3 100644
--- a/inference-engine/thirdparty/clDNN/tests/module_tests/usm_memory_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/module_tests/usm_memory_test.cpp
@@ -104,11 +104,11 @@ TEST_P(ctor_test, basic) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(cldnn_usm, ctor_test, ::testing::ValuesIn(std::vector<usm_test_params>{
+INSTANTIATE_TEST_SUITE_P(cldnn_usm, ctor_test, ::testing::ValuesIn(std::vector<usm_test_params>{
     usm_test_params{ allocation_type::usm_host},
     usm_test_params{ allocation_type::usm_shared},
     usm_test_params{ allocation_type::usm_device},
-}), );
+}));
 
 class copy_and_read_buffer : public BaseUSMTest {};
 TEST_P(copy_and_read_buffer, basic) {
@@ -177,11 +177,11 @@ TEST_P(copy_and_read_buffer, basic) {
 
 }
 
-INSTANTIATE_TEST_CASE_P(cldnn_usm, copy_and_read_buffer, ::testing::ValuesIn(std::vector<usm_test_params>{
+INSTANTIATE_TEST_SUITE_P(cldnn_usm, copy_and_read_buffer, ::testing::ValuesIn(std::vector<usm_test_params>{
         usm_test_params{ allocation_type::usm_host },
         usm_test_params{ allocation_type::usm_shared },
         usm_test_params{ allocation_type::usm_device },
-}), );
+}));
 
 class fill_buffer : public BaseUSMTest {};
 TEST_P(fill_buffer, DISABLED_basic) {
@@ -256,8 +256,8 @@ TEST_P(fill_buffer, DISABLED_basic) {
 
 }
 
-INSTANTIATE_TEST_CASE_P(cldnn_usm, fill_buffer, ::testing::ValuesIn(std::vector<usm_test_params>{
+INSTANTIATE_TEST_SUITE_P(cldnn_usm, fill_buffer, ::testing::ValuesIn(std::vector<usm_test_params>{
     usm_test_params{ allocation_type::usm_host },
         usm_test_params{ allocation_type::usm_shared },
         usm_test_params{ allocation_type::usm_device },
-}), );
+}));
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/binary_convolution_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/binary_convolution_gpu_test.cpp
index e0ac110a817a8b..d0b0504cb77174 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/binary_convolution_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/binary_convolution_gpu_test.cpp
@@ -245,7 +245,7 @@ TEST_P(binary_convolution_test, conv) {
 }
 
 // Batch, groups, IC, IW, IH, OC, OW, OH, KH, KW, SH, SW, PH, PW
-INSTANTIATE_TEST_CASE_P(BinaryConvTest, binary_convolution_test, ::testing::Values(
+INSTANTIATE_TEST_SUITE_P(BinaryConvTest, binary_convolution_test, ::testing::Values(
         TestParams{1, 1,  16,2,2,   4,2,2, 3,3, 1,1, 1,1, -1.0f, data_types::f32, "small"},
         TestParams{1, 1,  17,2,2,   4,2,2, 3,3, 1,1, 1,1, -1.0f, data_types::f32, "small"},
         TestParams{1, 1,  17,2,2,   4,2,2, 3,3, 1,1, 1,1,  0.0f, data_types::f32, "small"},
@@ -300,7 +300,7 @@ INSTANTIATE_TEST_CASE_P(BinaryConvTest, binary_convolution_test, ::testing::Valu
         TestParams{1, 1,  264,24,42, 192,24, 42, 1,1, 1,1, 0,0, -1.0f, data_types::f16, "conv5_5_sep_BIN"}, // back_bone_seq_conv5_5_sep_BIN
         TestParams{1, 1,  192,12,21, 208,12, 21, 1,1, 1,1, 0,0, -1.0f, data_types::f16, "conv5_6_sep_BIN"}, // back_bone_seq_conv5_6_sep_BIN
         TestParams{1, 1,  208,12,21,  88,12, 21, 1,1, 1,1, 0,0, -1.0f, data_types::f16, "conv6_sep_BN"} // back_bone_seq_conv6_sep_BN
-),);
+));
 
 template <typename T>
 static void set_binary_values(cldnn::memory::ptr mem, std::vector<T> args) {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/cache_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/cache_test.cpp
index 469d0b3103f1f8..e43fc33078f610 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/cache_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/cache_test.cpp
@@ -339,7 +339,7 @@ TEST_P(cache_version_test, update) {
         .test();
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke,
     cache_version_test,
     testing::Values(cache_version::version_1, cache_version::version_1_2, cache_version::version_2),
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/concatenation_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/concatenation_gpu_test.cpp
index 10f051b9ff435c..ca80202bb5c718 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/concatenation_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/concatenation_gpu_test.cpp
@@ -612,7 +612,7 @@ TEST_P(concat_gpu_4d_f16, fs_b_yx_fsv32) {
     ASSERT_NO_FATAL_FAILURE(test(format::fs_b_yx_fsv32));
 }
 
-INSTANTIATE_TEST_CASE_P(smoke,
+INSTANTIATE_TEST_SUITE_P(smoke,
                         concat_gpu_4d_f16,
                         concat_gpu_all_params,
                         concat_gpu::PrintToStringParamName);
@@ -625,7 +625,7 @@ TEST_P(concat_gpu_4d_i8, b_fs_yx_fsv16) {
     ASSERT_NO_FATAL_FAILURE(test(format::b_fs_yx_fsv16));
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_low_precision,
+INSTANTIATE_TEST_SUITE_P(smoke_low_precision,
                         concat_gpu_4d_i8,
                         concat_gpu_all_params,
                         concat_gpu::PrintToStringParamName);
@@ -634,7 +634,7 @@ TEST_P(concat_gpu_4d_u8, b_fs_yx_fsv32) {
     ASSERT_NO_FATAL_FAILURE(test(format::b_fs_yx_fsv32));
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_low_precision,
+INSTANTIATE_TEST_SUITE_P(smoke_low_precision,
                         concat_gpu_4d_u8,
                         concat_gpu_all_params,
                         concat_gpu::PrintToStringParamName);
@@ -753,7 +753,7 @@ TEST_P(concat_id_conv_gpu_4d_f16, input_order_opt_b_fs_yx_fsv16) {
     ASSERT_NO_FATAL_FAILURE(test(format::b_fs_yx_fsv16));
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_low_precision,
+INSTANTIATE_TEST_SUITE_P(smoke_low_precision,
                         concat_id_conv_gpu_4d_f16,
                         ::testing::Values(
                             TestParamType_concat(2, { 2, 32 }, 2, 1),
@@ -768,7 +768,7 @@ TEST_P(concat_id_conv_gpu_4d_i8, input_order_opt_b_fs_yx_fsv16) {
     ASSERT_NO_FATAL_FAILURE(test(format::b_fs_yx_fsv16));
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_low_precision,
+INSTANTIATE_TEST_SUITE_P(smoke_low_precision,
                         concat_id_conv_gpu_4d_i8,
                         ::testing::Values(
                             TestParamType_concat(2, { 2, 32 }, 2, 1),
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/convolution_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/convolution_gpu_test.cpp
index 658bd6625f3403..6ea4ff706607fe 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/convolution_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/convolution_gpu_test.cpp
@@ -5005,7 +5005,7 @@ struct convolution_general_gpu : public ::testing::TestWithParam<TestParamType_g
     }
 };
 
-INSTANTIATE_TEST_CASE_P(convolution_gpu_test,
+INSTANTIATE_TEST_SUITE_P(convolution_gpu_test,
                         convolution_gpu_fs_byx_fsv32,
                         ::testing::Values(
                                 // Filter size, Input features, Stride, Output padding, With bias
@@ -5278,7 +5278,7 @@ struct convolution_gpu_fs_byx_fsv32_crop : public ::testing::TestWithParam<TestP
     }
 };
 
-INSTANTIATE_TEST_CASE_P(convolution_gpu_with_crop,
+INSTANTIATE_TEST_SUITE_P(convolution_gpu_with_crop,
     convolution_gpu_fs_byx_fsv32_crop,
     ::testing::Values(
         TestParamType_convolution_gpu_with_crop(1, 14, 24, 1, 0, true)
@@ -5742,7 +5742,7 @@ void blockedFormatZeroCheck(cldnn::memory::ptr out_mem) {
     }
 }
 struct convolution_gpu_block_layout3D : public convolution_gpu_block_layout {};
-INSTANTIATE_TEST_CASE_P(convolution_gpu_block3D,
+INSTANTIATE_TEST_SUITE_P(convolution_gpu_block3D,
                         convolution_gpu_block_layout3D,
                         ::testing::Values(
                                 TestParamType_convolution_gpu_block_layout(1, 3, 10, 1, 1, 0, false, 16),
@@ -6196,7 +6196,7 @@ TEST_P(convolution_gpu_block_layout3D, bfzyx_bsv16_fsv16_fp32_fused_ops)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(convolution_gpu_block,
+INSTANTIATE_TEST_SUITE_P(convolution_gpu_block,
                         convolution_gpu_block_layout,
                         ::testing::Values(
                                 TestParamType_convolution_gpu_block_layout(16, 64, 64, 1, 1, 0, false, 0),
@@ -6635,7 +6635,7 @@ TEST_P(convolution_gpu_block_layout, bfyx_bsv16_fsv16_fp32_fused_ops)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(convolution_depthwise_gpu_fs_b_yx_fsv32,
+INSTANTIATE_TEST_SUITE_P(convolution_depthwise_gpu_fs_b_yx_fsv32,
                         convolution_depthwise_gpu,
                         ::testing::Values(
                                 // Input size, Filter size Y, Filter size X, groups, Stride, Output padding, With bias
@@ -6777,7 +6777,7 @@ TEST_P(convolution_depthwise_gpu, depthwise_conv_fs_b_yx_fsv32)
                 }
 }
 
-INSTANTIATE_TEST_CASE_P(convolution_depthwise_gpu_b_fs_yx_fsv16,
+INSTANTIATE_TEST_SUITE_P(convolution_depthwise_gpu_b_fs_yx_fsv16,
                         convolution_depthwise_gpu_fsv16,
                         ::testing::Values(
                             // Input size, Filter size Y, Filter size X, groups, Stride, Output padding, With bias
@@ -7125,7 +7125,7 @@ TEST_P(convolution_depthwise_gpu_bfyx, depthwise_conv_bfyx)
 }
 
 
-INSTANTIATE_TEST_CASE_P(convolution_depthwise_gpu_bfyx,
+INSTANTIATE_TEST_SUITE_P(convolution_depthwise_gpu_bfyx,
                         convolution_depthwise_gpu_bfyx,
                         ::testing::Values(
                             // Input size, Filter size Y, Filter size X, groups, Stride, Output padding, With bias
@@ -7162,7 +7162,7 @@ INSTANTIATE_TEST_CASE_P(convolution_depthwise_gpu_bfyx,
                         ),
                         convolution_depthwise_gpu::PrintToStringParamName);
 
-INSTANTIATE_TEST_CASE_P(convolution_grouped_fsv4_fsv16,
+INSTANTIATE_TEST_SUITE_P(convolution_grouped_fsv4_fsv16,
                         convolution_grouped_gpu,
                         ::testing::Values(
                             // Input X size, Input Y size, Input Z size, Input features, Output features,
@@ -7449,7 +7449,7 @@ TEST_P(convolution_grouped_gpu, base) {
                     }
 }
 
-INSTANTIATE_TEST_CASE_P(conv_fp16_cases,
+INSTANTIATE_TEST_SUITE_P(conv_fp16_cases,
                         convolution_general_gpu,
                         ::testing::Values(
                             // Input X size, Input Y size, Input Z size, Input features, Output features,
@@ -8298,7 +8298,7 @@ TEST_P(convolution_random_smoke_test, u8s8f32_fsv4_input) {
     ASSERT_NO_FATAL_FAILURE(test.run_random(GetParam()));
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     basic,
     convolution_random_smoke_test,
     testing::ValuesIn(
@@ -8343,7 +8343,7 @@ TEST_P(convolution_random_all_test, s8s8f32_scale) {
     ASSERT_NO_FATAL_FAILURE(test.run_random(GetParam()));
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     DISABLED_basic,
     convolution_random_all_test,
     testing::ValuesIn(
@@ -8633,7 +8633,7 @@ TEST_P(convolution_test, CONVOLUTION) {
     run_single_test();
 }
 
-INSTANTIATE_TEST_CASE_P(DISABLED_CONVOLUTION,
+INSTANTIATE_TEST_SUITE_P(DISABLED_CONVOLUTION,
                         convolution_test,
                         ::testing::ValuesIn(convolution_test::generate_all_test_params()),
                         tests::generic_test::custom_param_name_functor());
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/crop_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/crop_gpu_test.cpp
index e5b6ed490a9e3c..70ff25a32104fe 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/crop_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/crop_gpu_test.cpp
@@ -1235,10 +1235,10 @@ static std::vector<size_t> batches = {1, 8, 16, 17};
 static std::vector<size_t> in_features = {18, 24, 32};
 static std::vector<size_t> crop_features = {4, 8, 12, 17};
 
-INSTANTIATE_TEST_CASE_P(crop_test, crop_gpu,
+INSTANTIATE_TEST_SUITE_P(crop_test, crop_gpu,
                         ::testing::Combine(
                                 ::testing::ValuesIn(batches),
                                 ::testing::ValuesIn(in_features),
                                 ::testing::ValuesIn(crop_features),
                                 ::testing::ValuesIn(formats)
-                                ), );
+                                ));
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/cum_sum_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/cum_sum_gpu_test.cpp
index 0020c53076a293..24cf8229ed6667 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/cum_sum_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/cum_sum_gpu_test.cpp
@@ -207,7 +207,7 @@ namespace {
     };
     std::vector<bool> variants = {false, true};
 }
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         axis_0,
         cum_sum_gpu,
         ::testing::Combine(
@@ -221,9 +221,9 @@ INSTANTIATE_TEST_CASE_P(
             ::testing::ValuesIn(axes[0]),
             ::testing::ValuesIn(variants),
             ::testing::ValuesIn(variants)
-            ), );
+            ));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         axis_1,
         cum_sum_gpu,
         ::testing::Combine(
@@ -237,9 +237,9 @@ INSTANTIATE_TEST_CASE_P(
             ::testing::ValuesIn(axes[1]),
             ::testing::ValuesIn(variants),
             ::testing::ValuesIn(variants)
-            ), );
+            ));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         axis_2,
         cum_sum_gpu,
         ::testing::Combine(
@@ -253,9 +253,9 @@ INSTANTIATE_TEST_CASE_P(
             ::testing::ValuesIn(axes[2]),
             ::testing::ValuesIn(variants),
             ::testing::ValuesIn(variants)
-            ), );
+            ));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         axis_3,
         cum_sum_gpu,
         ::testing::Combine(
@@ -269,9 +269,9 @@ INSTANTIATE_TEST_CASE_P(
             ::testing::ValuesIn(axes[3]),
             ::testing::ValuesIn(variants),
             ::testing::ValuesIn(variants)
-            ), );
+            ));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         axis_4,
         cum_sum_gpu,
         ::testing::Combine(
@@ -285,9 +285,9 @@ INSTANTIATE_TEST_CASE_P(
             ::testing::ValuesIn(axes[4]),
             ::testing::ValuesIn(variants),
             ::testing::ValuesIn(variants)
-            ), );
+            ));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
         axis_5,
         cum_sum_gpu,
         ::testing::Combine(
@@ -301,7 +301,7 @@ INSTANTIATE_TEST_CASE_P(
             ::testing::ValuesIn(axes[5]),
             ::testing::ValuesIn(variants),
             ::testing::ValuesIn(variants)
-            ), );
+            ));
 
 // FIXME: This test fails on some driver versions. Looks like UB in impl or driver issue
 TEST(cum_sum_gpu_f16, DISABLED_basic_1d) {
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/deconvolution_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/deconvolution_gpu_test.cpp
index eb82022c8b389e..f64f876744bfb8 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/deconvolution_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/deconvolution_gpu_test.cpp
@@ -2394,7 +2394,7 @@ struct deconvolution_random_test_params {
 template <typename T>
 struct typed_comparator {
     static ::testing::AssertionResult compare(const char* lhs_expr, const char* rhs_expr, T ref, T val) {
-        return ::testing::internal::EqHelper<false>::Compare(lhs_expr, rhs_expr, ref, val);
+        return ::testing::internal::EqHelper::Compare(lhs_expr, rhs_expr, ref, val);
     }
 };
 
@@ -2800,7 +2800,7 @@ TEST_P(deconvolution_random_test, basic) {
     run();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke, deconvolution_random_test, testing::ValuesIn(
+INSTANTIATE_TEST_SUITE_P(smoke, deconvolution_random_test, testing::ValuesIn(
     deconvolution_random_test_params_generator()
     .add_smoke_2d(data_types::f32, data_types::f32, data_types::f32, format::bfyx, format::any)
     .add_smoke_3d(data_types::f32, data_types::f32, data_types::f32, format::bfzyx, format::any)
@@ -2816,7 +2816,7 @@ INSTANTIATE_TEST_CASE_P(smoke, deconvolution_random_test, testing::ValuesIn(
     .add_smoke_3d(data_types::i8, data_types::i8, data_types::f32, format::b_fs_zyx_fsv16, format::b_fs_zyx_fsv16)
 ), deconvolution_random_test_params::print_params);
 
-INSTANTIATE_TEST_CASE_P(DISABLED_extended, deconvolution_random_test, testing::ValuesIn(
+INSTANTIATE_TEST_SUITE_P(DISABLED_extended, deconvolution_random_test, testing::ValuesIn(
     deconvolution_random_test_params_generator()
     .add_extra_2d(data_types::f32, data_types::f32, data_types::f32, format::bfyx, format::any)
     .add_extra_3d(data_types::f32, data_types::f32, data_types::f32, format::bfzyx, format::any)
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/depth_concatenate_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/depth_concatenate_gpu_test.cpp
index 116d6ba21f4d2b..19b147fbf93db4 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/depth_concatenate_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/depth_concatenate_gpu_test.cpp
@@ -1282,7 +1282,7 @@ TEST_P(depth_concatenate_test, DEPTHCONCATENATE) {
     run_single_test();
 }
 
-INSTANTIATE_TEST_CASE_P(DISABLED_DEPTHCONCATENATE,
+INSTANTIATE_TEST_SUITE_P(DISABLED_DEPTHCONCATENATE,
                         depth_concatenate_test,
                         ::testing::ValuesIn(depth_concatenate_test::generate_all_test_params()),
                         depth_concatenate_test::custom_param_name);
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/detection_output_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/detection_output_test.cpp
index 5d85613c1c5025..b406730a288de0 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/detection_output_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/detection_output_test.cpp
@@ -722,7 +722,7 @@ class detection_output_test : public ::testing::Test {
 };
 
 typedef ::testing::Types<float, FLOAT16> detection_output_test_types;
-TYPED_TEST_CASE(detection_output_test, detection_output_test_types);
+TYPED_TEST_SUITE(detection_output_test, detection_output_test_types);
 
 TYPED_TEST(detection_output_test, test_setup_basic) {
     this->setup_basic();
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/eltwise_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/eltwise_gpu_test.cpp
index 016f4b98c69bbc..2fe220ec04fb35 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/eltwise_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/eltwise_gpu_test.cpp
@@ -3321,7 +3321,7 @@ TEST_P(eltwise_same_input_test, random) {
     execute_same_input(param, true);
 }
 
-INSTANTIATE_TEST_CASE_P(eltwise_same_input,
+INSTANTIATE_TEST_SUITE_P(eltwise_same_input,
                         eltwise_same_input_test,
                         testing::ValuesIn(
                             eltwise_same_input_param_generator()
@@ -3329,7 +3329,7 @@ INSTANTIATE_TEST_CASE_P(eltwise_same_input,
                             .simple_params(data_types::f32, format::b_fs_yx_fsv32, format::b_fs_yx_fsv32)
                             .simple_params(data_types::f32, format::b_fs_yx_fsv16, format::b_fs_yx_fsv16)
                             .simple_params(data_types::f32, format::fs_b_yx_fsv32, format::fs_b_yx_fsv32)
-                        ), );
+                        ));
 
 // mode, input type, input sizes
 using eltwise_test_params = std::tuple<eltwise_mode, data_types, std::vector<std::vector<int32_t>>>;
@@ -3490,12 +3490,12 @@ static std::vector<std::vector<std::vector<int32_t>>> inputs = {
         {{1, 32, 1, 1, 1}, {8, 32, 3, 4, 5}},
 };
 
-INSTANTIATE_TEST_CASE_P(eltwise, eltwise_test,
+INSTANTIATE_TEST_SUITE_P(eltwise, eltwise_test,
                         ::testing::Combine(
                                 ::testing::ValuesIn(modes),
                                 ::testing::ValuesIn(types),
                                 ::testing::ValuesIn(inputs)
-                                ), );
+                                ));
 
 class eltwise_test_6d : public eltwise_test {};
 TEST_P(eltwise_test_6d, bfwzyx) {
@@ -3572,12 +3572,12 @@ static std::vector<std::vector<std::vector<int32_t>>> inputs_6d = {
         {{1, 32, 1, 1, 1, 7}, {8, 32, 4, 5, 6, 7}},
 };
 
-INSTANTIATE_TEST_CASE_P(eltwise, eltwise_test_6d,
+INSTANTIATE_TEST_SUITE_P(eltwise, eltwise_test_6d,
                         ::testing::Combine(
                                 ::testing::ValuesIn(modes),
                                 ::testing::ValuesIn(types),
                                 ::testing::ValuesIn(inputs_6d)
-                                ), );
+                                ));
 
 class eltwise_test_mixed_precision : public eltwise_test {};
 TEST_P(eltwise_test_mixed_precision, fsv16) {
@@ -3653,12 +3653,12 @@ TEST_P(eltwise_test_mixed_precision, fsv16) {
 
 static std::vector<data_types> mixed_types = {data_types::i8, data_types::u8};
 
-INSTANTIATE_TEST_CASE_P(eltwise, eltwise_test_mixed_precision,
+INSTANTIATE_TEST_SUITE_P(eltwise, eltwise_test_mixed_precision,
                         ::testing::Combine(
                                 ::testing::ValuesIn(modes),
                                 ::testing::ValuesIn(mixed_types),
                                 ::testing::ValuesIn(inputs)
-                                ), );
+                                ));
 
 
 struct eltwise_layout_test_params {
@@ -3748,7 +3748,7 @@ TEST_P(eltwise_test_mixed_layout, mixed_layout) {
     }
 }
 
-INSTANTIATE_TEST_CASE_P(eltwise, eltwise_test_mixed_layout,
+INSTANTIATE_TEST_SUITE_P(eltwise, eltwise_test_mixed_layout,
                         ::testing::ValuesIn(std::vector<eltwise_layout_test_params>{
                             eltwise_layout_test_params{CASE_ELTWISE_TEST1},
                             eltwise_layout_test_params{CASE_ELTWISE_TEST2},
@@ -3759,7 +3759,7 @@ INSTANTIATE_TEST_CASE_P(eltwise, eltwise_test_mixed_layout,
                             eltwise_layout_test_params{CASE_ELTWISE_TEST7},
                             eltwise_layout_test_params{CASE_ELTWISE_TEST8},
                             eltwise_layout_test_params{CASE_ELTWISE_TEST9},
-                        }), );
+                        }));
 
 //
 struct eltwise_random_test_params {
@@ -3937,7 +3937,7 @@ TEST_P(eltwise_random_test, random) {
     execute_compare(param, true);
 }
 
-INSTANTIATE_TEST_CASE_P(eltwise_smoke_fsv4,
+INSTANTIATE_TEST_SUITE_P(eltwise_smoke_fsv4,
                         eltwise_random_test,
                         testing::ValuesIn(
                             eltwise_random_test_param_generator()
@@ -3949,4 +3949,4 @@ INSTANTIATE_TEST_CASE_P(eltwise_smoke_fsv4,
                             .simple_params(data_types::f16, format::b_fs_yx_fsv4, format::b_fs_yx_fsv4)
                             .simple_params(data_types::i8, format::b_fs_yx_fsv4, format::b_fs_yx_fsv4)
                             .simple_params(data_types::u8, format::b_fs_yx_fsv4, format::b_fs_yx_fsv4)
-                        ), );
+                        ));
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/fully_connected_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/fully_connected_gpu_test.cpp
index b96ca63bb2e0e9..da33fd428eeddb 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/fully_connected_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/fully_connected_gpu_test.cpp
@@ -1068,7 +1068,7 @@ TEST_P(fully_connected_random_test_f32, basic) {
     run_test();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke,
+INSTANTIATE_TEST_SUITE_P(smoke,
                         fully_connected_random_test_f32,
                         ::testing::Combine(
                             ::testing::Values(1, 2),
@@ -1078,9 +1078,9 @@ INSTANTIATE_TEST_CASE_P(smoke,
                             ::testing::Values(3, 32),
                             ::testing::Values(format::bfyx, format::yxfb),
                             ::testing::Values(format::any),
-                            ::testing::Values("")), );
+                            ::testing::Values("")));
 
-INSTANTIATE_TEST_CASE_P(smoke_bfyx_batched,
+INSTANTIATE_TEST_SUITE_P(smoke_bfyx_batched,
                         fully_connected_random_test_f32,
                         ::testing::Combine(
                             ::testing::Values(2, 8),
@@ -1090,13 +1090,13 @@ INSTANTIATE_TEST_CASE_P(smoke_bfyx_batched,
                             ::testing::Values(3, 32),
                             ::testing::Values(format::bfyx),
                             ::testing::Values(format::bfyx),
-                            ::testing::Values("")), );
+                            ::testing::Values("")));
 
 TEST_P(fully_connected_random_test_f16, basic) {
     run_test();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke_bfyx_b2,
+INSTANTIATE_TEST_SUITE_P(smoke_bfyx_b2,
                         fully_connected_random_test_f16,
                         ::testing::Combine(
                             // Batch 1 is disabled due to sporadic failures in `fully_connected_gpu_bs_f_bsv16_b1`
@@ -1108,9 +1108,9 @@ INSTANTIATE_TEST_CASE_P(smoke_bfyx_b2,
                             ::testing::Values(3, 32),
                             ::testing::Values(format::bfyx),
                             ::testing::Values(format::any),
-                            ::testing::Values("")), );
+                            ::testing::Values("")));
 
-INSTANTIATE_TEST_CASE_P(smoke,
+INSTANTIATE_TEST_SUITE_P(smoke,
                         fully_connected_random_test_f16,
                         ::testing::Combine(
                             ::testing::Values(1, 2),
@@ -1120,9 +1120,9 @@ INSTANTIATE_TEST_CASE_P(smoke,
                             ::testing::Values(3, 32),
                             ::testing::Values(format::yxfb),
                             ::testing::Values(format::any),
-                            ::testing::Values("")), );
+                            ::testing::Values("")));
 
-INSTANTIATE_TEST_CASE_P(smoke_bfyx_batched,
+INSTANTIATE_TEST_SUITE_P(smoke_bfyx_batched,
                         fully_connected_random_test_f16,
                         ::testing::Combine(
                             ::testing::Values(2, 8),
@@ -1132,7 +1132,7 @@ INSTANTIATE_TEST_CASE_P(smoke_bfyx_batched,
                             ::testing::Values(3, 32),
                             ::testing::Values(format::bfyx),
                             ::testing::Values(format::bfyx),
-                            ::testing::Values("")), );
+                            ::testing::Values("")));
 
 
 template <typename InputT, typename WeightsT, typename BiasT, typename OutputT>
@@ -1168,7 +1168,7 @@ TEST_P(fully_connected_random_test_f32_3d, basic) {
     run_test();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke,
+INSTANTIATE_TEST_SUITE_P(smoke,
                         fully_connected_random_test_f32_3d,
                         ::testing::Combine(
                             ::testing::Values(1,3),
@@ -1178,9 +1178,9 @@ INSTANTIATE_TEST_CASE_P(smoke,
                             ::testing::Values(1,3,16),
                             ::testing::Values(format::bfyx),
                             ::testing::Values(format::any),
-                            ::testing::Values("")), );
+                            ::testing::Values("")));
 
-INSTANTIATE_TEST_CASE_P(smoke_big,
+INSTANTIATE_TEST_SUITE_P(smoke_big,
                         fully_connected_random_test_f32_3d,
                         ::testing::Combine(
                             ::testing::Values(3),
@@ -1190,13 +1190,13 @@ INSTANTIATE_TEST_CASE_P(smoke_big,
                             ::testing::Values(17, 32),
                             ::testing::Values(format::bfyx),
                             ::testing::Values(format::any),
-                            ::testing::Values("")), );
+                            ::testing::Values("")));
 
 TEST_P(fully_connected_random_test_f16_3d, basic) {
     run_test();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke,
+INSTANTIATE_TEST_SUITE_P(smoke,
                         fully_connected_random_test_f16_3d,
                         ::testing::Combine(
                             ::testing::Values(1,3),
@@ -1206,13 +1206,13 @@ INSTANTIATE_TEST_CASE_P(smoke,
                             ::testing::Values(1,3,16),
                             ::testing::Values(format::bfyx),
                             ::testing::Values(format::any),
-                            ::testing::Values("")), );
+                            ::testing::Values("")));
 
 TEST_P(fully_connected_random_test_i8_3d, basic) {
     run_test();
 }
 
-INSTANTIATE_TEST_CASE_P(smoke,
+INSTANTIATE_TEST_SUITE_P(smoke,
                         fully_connected_random_test_i8_3d,
                         ::testing::Combine(
                             ::testing::Values(1,3),
@@ -1222,9 +1222,9 @@ INSTANTIATE_TEST_CASE_P(smoke,
                             ::testing::Values(1,3,16),
                             ::testing::Values(format::bfyx),
                             ::testing::Values(format::any),
-                            ::testing::Values("")), );
+                            ::testing::Values("")));
 
-INSTANTIATE_TEST_CASE_P(smoke_big,
+INSTANTIATE_TEST_SUITE_P(smoke_big,
                         fully_connected_random_test_i8_3d,
                         ::testing::Combine(
                             ::testing::Values(1,3),
@@ -1234,7 +1234,7 @@ INSTANTIATE_TEST_CASE_P(smoke_big,
                             ::testing::Values(17, 32),
                             ::testing::Values(format::bfyx),
                             ::testing::Values(format::any),
-                            ::testing::Values("")), );
+                            ::testing::Values("")));
 
 
 struct quantization_t {
@@ -1523,7 +1523,7 @@ TEST_P(fully_connected_u8_f32_test, random) {
     run_random_test();
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     basic,
     fully_connected_i8_i8_test,
     testing::Combine(
@@ -1537,7 +1537,7 @@ INSTANTIATE_TEST_CASE_P(
     fully_connected_i8_i8_test::PrintToStringParamName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     basic,
     fully_connected_i8_u8_test,
     testing::Combine(
@@ -1551,7 +1551,7 @@ INSTANTIATE_TEST_CASE_P(
     fully_connected_i8_u8_test::PrintToStringParamName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     basic,
     fully_connected_i8_f32_test,
     testing::Combine(
@@ -1565,7 +1565,7 @@ INSTANTIATE_TEST_CASE_P(
     fully_connected_i8_f32_test::PrintToStringParamName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     basic,
     fully_connected_u8_i8_test,
     testing::Combine(
@@ -1579,7 +1579,7 @@ INSTANTIATE_TEST_CASE_P(
     fully_connected_u8_i8_test::PrintToStringParamName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     basic,
     fully_connected_u8_u8_test,
     testing::Combine(
@@ -1593,7 +1593,7 @@ INSTANTIATE_TEST_CASE_P(
     fully_connected_u8_u8_test::PrintToStringParamName
 );
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     basic,
     fully_connected_u8_f32_test,
     testing::Combine(
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/fusings_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/fusings_gpu_test.cpp
index a3a88b7b2e8c69..0fea8be6648aa4 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/fusings_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/fusings_gpu_test.cpp
@@ -629,7 +629,7 @@ TEST_P(conv_fp32_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_activation, ::testing::ValuesIn(std::vector<bc_test_params>{
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_activation, ::testing::ValuesIn(std::vector<bc_test_params>{
                                                                            bc_test_params{CASE_CONV_FP32_1, 2, 3},
                                                                            bc_test_params{CASE_CONV_FP32_2, 2, 3},
                                                                            bc_test_params{CASE_CONV_FP32_3, 2, 3},
@@ -639,7 +639,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_activation, ::testing::ValuesIn(s
                                                                            bc_test_params{CASE_CONV_FP16_4, 2, 3},
                                                                            bc_test_params{CASE_CONV_FP16_4, 2, 3},
                                                                            bc_test_params{CASE_CONV_FP16_4, 2, 3},
-}), );
+}));
 
 
 class conv_fp32_scale : public ConvFusingTest {};
@@ -658,7 +658,7 @@ TEST_P(conv_fp32_scale, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_scale,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_scale,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                              // bc_test_params{CASE_CONV_FP32_1, 2, 3},
                                              bc_test_params{CASE_CONV_FP32_2, 2, 3},
@@ -671,7 +671,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_scale,
                                              bc_test_params{CASE_CONV_FP16_3, 2, 3},
                                              bc_test_params{CASE_CONV_FP16_4, 2, 3},
                                              bc_test_params{CASE_CONV_FP16_10, 2, 3},
-                                             }), );
+                                             }));
 
 class conv_fp32_bias : public ConvFusingTest {};
 TEST_P(conv_fp32_bias, basic) {
@@ -688,7 +688,7 @@ TEST_P(conv_fp32_bias, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_bias,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_bias,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                              bc_test_params{CASE_CONV_FP32_1, 2, 3},
                                              bc_test_params{CASE_CONV_FP32_2, 2, 3},
@@ -701,7 +701,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_bias,
                                              bc_test_params{CASE_CONV_FP16_3, 2, 3},
                                              bc_test_params{CASE_CONV_FP16_4, 2, 3},
                                              bc_test_params{CASE_CONV_FP16_10, 2, 3},
-                                             }), );
+                                             }));
 
 class conv_fp32_prelu_eltwise : public ConvFusingTest {};
 TEST_P(conv_fp32_prelu_eltwise, basic_sum) {
@@ -815,7 +815,7 @@ TEST_P(conv_fp32_prelu_eltwise, vector_ops_mixed_types) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_prelu_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_prelu_eltwise,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                              // bc_test_params{CASE_CONV_FP32_1, 2, 4},
                                              bc_test_params{CASE_CONV_FP32_2, 2, 4},
@@ -826,7 +826,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_prelu_eltwise,
                                              bc_test_params{CASE_CONV_FP16_2, 2, 4},
                                              bc_test_params{CASE_CONV_FP16_3, 2, 4},
                                              bc_test_params{CASE_CONV_FP16_4, 2, 4},
-                                             }), );
+                                             }));
 
 class conv_fp32_multi_eltwise_2 : public ConvFusingTest {};
 TEST_P(conv_fp32_multi_eltwise_2, basic) {
@@ -851,7 +851,7 @@ TEST_P(conv_fp32_multi_eltwise_2, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_2,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_multi_eltwise_2,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                              bc_test_params{CASE_CONV_FP32_2, 2, 4},
                                              bc_test_params{CASE_CONV_FP32_3, 2, 4},
@@ -860,7 +860,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_2,
                                              bc_test_params{CASE_CONV_FP16_2, 2, 4},
                                              bc_test_params{CASE_CONV_FP16_3, 2, 4},
                                              bc_test_params{CASE_CONV_FP16_4, 2, 4},
-                                             }), );
+                                             }));
 
 
 class conv_fp32_multi_eltwise_2_clamp : public ConvFusingTest {};
@@ -887,7 +887,7 @@ TEST_P(conv_fp32_multi_eltwise_2_clamp, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_2_clamp,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_multi_eltwise_2_clamp,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                              bc_test_params{CASE_CONV_FP32_2, 2, 5},
                                              bc_test_params{CASE_CONV_FP32_3, 2, 5},
@@ -896,7 +896,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_2_clamp,
                                              bc_test_params{CASE_CONV_FP16_2, 2, 5},
                                              bc_test_params{CASE_CONV_FP16_3, 2, 5},
                                              bc_test_params{CASE_CONV_FP16_4, 2, 5},
-                                             }), );
+                                             }));
 
 
 class conv_fp32_multi_eltwise_4_clamp : public ConvFusingTest {};
@@ -927,7 +927,7 @@ TEST_P(conv_fp32_multi_eltwise_4_clamp, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_4_clamp,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_multi_eltwise_4_clamp,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                              bc_test_params{CASE_CONV_FP32_2, 2, 7},
                                              bc_test_params{CASE_CONV_FP32_3, 2, 7},
@@ -936,7 +936,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_4_clamp,
                                              bc_test_params{CASE_CONV_FP16_2, 2, 7},
                                              bc_test_params{CASE_CONV_FP16_3, 2, 7},
                                              bc_test_params{CASE_CONV_FP16_4, 2, 7},
-                                             }), );
+                                             }));
 
 
 class conv_fp32_multi_eltwise_3_fusing : public ConvFusingTest {};
@@ -964,7 +964,7 @@ TEST_P(conv_fp32_multi_eltwise_3_fusing, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_3_fusing,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_multi_eltwise_3_fusing,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                              bc_test_params{CASE_CONV_FP32_2, 2, 5},
                                              bc_test_params{CASE_CONV_FP32_3, 2, 5},
@@ -973,7 +973,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_3_fusing,
                                              bc_test_params{CASE_CONV_FP16_2, 2, 5},
                                              bc_test_params{CASE_CONV_FP16_3, 2, 5},
                                              bc_test_params{CASE_CONV_FP16_4, 2, 5},
-                                             }), );
+                                             }));
 
 
 
@@ -999,7 +999,7 @@ TEST_P(conv_fp32_multi_eltwise_quantization, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_quantization,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_multi_eltwise_quantization,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                              bc_test_params{CASE_CONV_FP32_2, 4, 5},
                                              bc_test_params{CASE_CONV_FP32_4, 4, 5},
@@ -1007,7 +1007,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_quantization,
                                              bc_test_params{CASE_CONV_FP16_2, 4, 5},
                                              bc_test_params{CASE_CONV_FP16_3, 4, 5},
                                              bc_test_params{CASE_CONV_FP16_4, 4, 5},
-                                             }), );
+                                             }));
 
 
 class conv_fp32_multi_eltwise_concat : public ConvFusingTest {};
@@ -1035,7 +1035,7 @@ TEST_P(conv_fp32_multi_eltwise_concat, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_concat,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_multi_eltwise_concat,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                              bc_test_params{CASE_CONV_FP32_2, 5, 5},
                                              bc_test_params{CASE_CONV_FP32_3, 5, 5},
@@ -1044,7 +1044,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_multi_eltwise_concat,
                                              bc_test_params{CASE_CONV_FP16_2, 5, 5},
                                              bc_test_params{CASE_CONV_FP16_3, 5, 5},
                                              bc_test_params{CASE_CONV_FP16_4, 5, 5},
-                                             }), );
+                                             }));
 
 class conv_fp32_eltwise_b_fs_zyx_fsv16 : public ConvFusingTest {};
 
@@ -1082,7 +1082,7 @@ TEST_P(conv_fp32_swish, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_swish,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_swish,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 // bc_test_params{CASE_CONV_FP32_1, 2, 4},
                                 bc_test_params{CASE_CONV_FP32_2, 2, 4},
@@ -1093,7 +1093,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_swish,
                                 bc_test_params{CASE_CONV_FP16_2, 2, 4},
                                 bc_test_params{CASE_CONV_FP16_3, 2, 4},
                                 bc_test_params{CASE_CONV_FP16_4, 2, 4},
-                        }), );
+                        }));
 
 TEST_P(conv_fp32_eltwise_b_fs_zyx_fsv16, splitted_vector_ops) {
     auto p = GetParam();
@@ -1119,7 +1119,7 @@ TEST_P(conv_fp32_eltwise_b_fs_zyx_fsv16, splitted_vector_ops) {
     //  execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_eltwise_b_fs_zyx_fsv16,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_eltwise_b_fs_zyx_fsv16,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_FP32_6, 2, 3},
                                 bc_test_params{CASE_CONV_FP32_7, 2, 3},
@@ -1135,7 +1135,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_eltwise_b_fs_zyx_fsv16,
                                 bc_test_params{CASE_CONV_FP16_9, 2, 3},
                                 bc_test_params{CASE_CONV_FP16_11, 2, 3},
                                 bc_test_params{CASE_CONV_FP16_12, 2, 3},
-                        }), );
+                        }));
 
 class conv_fp32_quantize_u8_first_conv : public ConvFusingTest {};
 TEST_P(conv_fp32_quantize_u8_first_conv, basic) {
@@ -1157,10 +1157,10 @@ TEST_P(conv_fp32_quantize_u8_first_conv, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_quantize_u8_first_conv,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_quantize_u8_first_conv,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_FP32_14, 2, 3},
-                        }), );
+                        }));
 
 class conv_fp32_quantize_u8 : public ConvFusingTest {};
 TEST_P(conv_fp32_quantize_u8, basic) {
@@ -1181,7 +1181,7 @@ TEST_P(conv_fp32_quantize_u8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_quantize_u8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_quantize_u8,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 // For now only b_fs_yx_fsv16 supports this case
                                 bc_test_params{CASE_CONV_FP32_2, 2, 3},
@@ -1189,7 +1189,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_quantize_u8,
 
                                 bc_test_params{CASE_CONV_FP16_2, 2, 3},
                                 bc_test_params{CASE_CONV_FP16_3, 2, 3},
-                        }), );
+                        }));
 
 class conv_fp32_scale_quantize_i8 : public ConvFusingTest {};
 TEST_P(conv_fp32_scale_quantize_i8, basic) {
@@ -1214,7 +1214,7 @@ TEST_P(conv_fp32_scale_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_scale_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_scale_quantize_i8,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 // For now only b_fs_yx_fsv16 supports this case
                                 bc_test_params{CASE_CONV_FP32_2, 2, 4},
@@ -1222,7 +1222,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_scale_quantize_i8,
 
                                 bc_test_params{CASE_CONV_FP16_2, 2, 4},
                                 bc_test_params{CASE_CONV_FP16_3, 2, 4},
-                        }), );
+                        }));
 
 class conv_fp32_scale_activation_quantize_i8 : public ConvFusingTest {};
 TEST_P(conv_fp32_scale_activation_quantize_i8, basic) {
@@ -1246,7 +1246,7 @@ TEST_P(conv_fp32_scale_activation_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_scale_activation_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_scale_activation_quantize_i8,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 // For now only b_fs_yx_fsv16 supports this case
                                 bc_test_params{CASE_CONV_FP32_2, 2, 5},
@@ -1254,7 +1254,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_scale_activation_quantize_i8,
 
                                 bc_test_params{CASE_CONV_FP16_2, 2, 5},
                                 bc_test_params{CASE_CONV_FP16_3, 2, 5},
-                        }), );
+                        }));
 
 class conv_fp32_scale_activation_quantize_u8_eltwise_fp32 : public ConvFusingTest {};
 TEST_P(conv_fp32_scale_activation_quantize_u8_eltwise_fp32, basic) {
@@ -1279,7 +1279,7 @@ TEST_P(conv_fp32_scale_activation_quantize_u8_eltwise_fp32, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_scale_activation_quantize_u8_eltwise_fp32,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_scale_activation_quantize_u8_eltwise_fp32,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 // For now only b_fs_yx_fsv16 supports this case
                                 bc_test_params{CASE_CONV_FP32_2, 2, 6},
@@ -1287,7 +1287,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_scale_activation_quantize_u8_eltw
 
                                 bc_test_params{CASE_CONV_FP16_2, 2, 6},
                                 bc_test_params{CASE_CONV_FP16_3, 2, 6},
-                        }), );
+                        }));
 
 class conv_fp32_scale_activation_quantize_i8_activation : public ConvFusingTest {};
 TEST_P(conv_fp32_scale_activation_quantize_i8_activation, basic) {
@@ -1312,14 +1312,14 @@ TEST_P(conv_fp32_scale_activation_quantize_i8_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_scale_activation_quantize_i8_activation,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_scale_activation_quantize_i8_activation,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_FP32_2, 2, 6},
                                 bc_test_params{CASE_CONV_FP32_3, 2, 6},
 
                                 bc_test_params{CASE_CONV_FP16_2, 2, 6},
                                 bc_test_params{CASE_CONV_FP16_3, 2, 6},
-                        }), );
+                        }));
 
 
 class conv_fp32_scale_activation_quantize_i8_eltwise_fp32_quantize_i8 : public ConvFusingTest {};
@@ -1350,11 +1350,11 @@ TEST_P(conv_fp32_scale_activation_quantize_i8_eltwise_fp32_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_scale_activation_quantize_i8_eltwise_fp32_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_scale_activation_quantize_i8_eltwise_fp32_quantize_i8,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_FP32_2, 2, 7},
                                 bc_test_params{CASE_CONV_FP32_3, 2, 7},
-                        }), );
+                        }));
 
 class conv_fp32_activation_eltwise_in_u8_fp32 : public WeightsPrimitiveFusingTest<bc_test_params> {};
 TEST_P(conv_fp32_activation_eltwise_in_u8_fp32, basic) {
@@ -1372,7 +1372,7 @@ TEST_P(conv_fp32_activation_eltwise_in_u8_fp32, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_activation_eltwise_in_u8_fp32,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_activation_eltwise_in_u8_fp32,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 // bc_test_params{CASE_CONV_FP32_1, 2, 4}, - eltwise fusing not supported
                                 bc_test_params{CASE_CONV_FP32_2, 2, 4},
@@ -1384,7 +1384,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_activation_eltwise_in_u8_fp32,
                                 // bc_test_params{CASE_CONV_FP32_8, 2, 4}, - unknown bug
                                 bc_test_params{CASE_CONV_FP32_9, 2, 4},
                                 bc_test_params{CASE_CONV_FP32_10, 2, 4},
-                        }), );
+                        }));
 
 class conv_fp32_activation_eltwise_diff_sizes : public ConvEltwTest {};
 TEST_P(conv_fp32_activation_eltwise_diff_sizes, basic) {
@@ -1402,7 +1402,7 @@ TEST_P(conv_fp32_activation_eltwise_diff_sizes, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_activation_eltwise_diff_sizes,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp32_activation_eltwise_diff_sizes,
                         ::testing::ValuesIn(std::vector<conv_eltw_test_params>{
                                 conv_eltw_test_params{CASE_CONV_ELTW_FP32_1, 2, 4},
                                 conv_eltw_test_params{CASE_CONV_ELTW_FP32_2, 2, 4},
@@ -1412,7 +1412,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp32_activation_eltwise_diff_sizes,
                                 conv_eltw_test_params{CASE_CONV_ELTW_FP32_6, 2, 4},
                                 conv_eltw_test_params{CASE_CONV_ELTW_FP32_7, 3, 4},
                                 conv_eltw_test_params{CASE_CONV_ELTW_FP32_8, 3, 4},
-                        }), );
+                        }));
 
 class conv_scale_activation_eltwise_fp32_quantize_i8 : public ConvEltwTest {};
 TEST_P(conv_scale_activation_eltwise_fp32_quantize_i8, basic) {
@@ -1437,7 +1437,7 @@ TEST_P(conv_scale_activation_eltwise_fp32_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_scale_activation_eltwise_fp32_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_scale_activation_eltwise_fp32_quantize_i8,
                         ::testing::ValuesIn(std::vector<conv_eltw_test_params>{
                                 conv_eltw_test_params{CASE_CONV_ELTW_FP32_1, 2, 6},
                                 conv_eltw_test_params{CASE_CONV_ELTW_FP32_2, 2, 6},
@@ -1447,7 +1447,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_scale_activation_eltwise_fp32_quantize
                                 conv_eltw_test_params{CASE_CONV_ELTW_FP32_6, 3, 6},
                                 conv_eltw_test_params{CASE_CONV_ELTW_FP32_7, 3, 6},
                                 conv_eltw_test_params{CASE_CONV_ELTW_FP32_8, 3, 6},
-                        }), );
+                        }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* -------------------------------------- binary convolution cases ------------------------------------- */
@@ -1467,10 +1467,10 @@ TEST_P(conv_bin_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_bin_activation,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_bin_activation,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                             bc_test_params{CASE_BIN_CONV1, 2, 3},
-                                            }), );
+                                            }));
 
 class conv_bin_scale_activation : public ConvFusingTest {};
 TEST_P(conv_bin_scale_activation, basic) {
@@ -1487,11 +1487,11 @@ TEST_P(conv_bin_scale_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_bin_scale_activation,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_bin_scale_activation,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                             bc_test_params{CASE_BIN_CONV1, 2, 4},
                             bc_test_params{CASE_BIN_CONV2, 2, 4},
-                                            }), );
+                                            }));
 
 class conv_bin_quantize_bin : public ConvFusingTest {};
 TEST_P(conv_bin_quantize_bin, channel_wise_quantize) {
@@ -1528,11 +1528,11 @@ TEST_P(conv_bin_quantize_bin, blob_wise_quantize) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_bin_quantize_bin,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_bin_quantize_bin,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                             bc_test_params{CASE_BIN_CONV1, 2, 3},
                             bc_test_params{CASE_BIN_CONV2, 2, 3},
-                                            }), );
+                                            }));
 
 class conv_bin_scale_conv_dw : public ConvFusingTest {};
 TEST_P(conv_bin_scale_conv_dw, dw_kernel_3x3_stride2) {
@@ -1573,11 +1573,11 @@ TEST_P(conv_bin_scale_conv_dw, dw_kernel_3x3_stride1) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_bin_scale_conv_dw,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_bin_scale_conv_dw,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                             bc_test_params{CASE_BIN_CONV2, 3, 4},
                             bc_test_params{CASE_BIN_CONV3, 3, 4},
-                                            }), );
+                                            }));
 
 class conv_bin_scale_conv_dw_prelu : public ConvFusingTest {};
 TEST_P(conv_bin_scale_conv_dw_prelu, dw_kernel_3x3_stride2) {
@@ -1624,11 +1624,11 @@ TEST_P(conv_bin_scale_conv_dw_prelu, dw_kernel_3x3_stride1) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_bin_scale_conv_dw_prelu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_bin_scale_conv_dw_prelu,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                             bc_test_params{CASE_BIN_CONV2, 3, 5},
                             bc_test_params{CASE_BIN_CONV3, 3, 5},
-                                            }), );
+                                            }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ---------------------------------------- INT8 convolution cases ------------------------------------- */
@@ -1664,7 +1664,7 @@ TEST_P(conv_int8_scale, fp16_scale_out) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_scale,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 3},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 3},
@@ -1685,7 +1685,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale,
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 3},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 3},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 3},
-                        }), );
+                        }));
 
 class conv_int8_eltwise : public ConvFusingTest {};
 TEST_P(conv_int8_eltwise, fp16_eltwise_out) {
@@ -1703,7 +1703,7 @@ TEST_P(conv_int8_eltwise, fp16_eltwise_out) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_eltwise,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 3},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 3},
@@ -1724,7 +1724,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_eltwise,
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 3},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 3},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 3},
-                        }), );
+                        }));
 
 class conv_int8_scale_shift_swish : public ConvFusingTest {};
 TEST_P(conv_int8_scale_shift_swish, basic) {
@@ -1748,7 +1748,7 @@ TEST_P(conv_int8_scale_shift_swish, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_shift_swish,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_scale_shift_swish,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 8},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 8},
@@ -1769,7 +1769,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_shift_swish,
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 8},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 8},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 8},
-                        }), );
+                        }));
 
 class conv_int8_prelu_eltwise : public ConvFusingTest {};
 TEST_P(conv_int8_prelu_eltwise, basic) {
@@ -1814,7 +1814,7 @@ TEST_P(conv_int8_prelu_eltwise, fsv16) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_prelu_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_prelu_eltwise,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 4},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 4},
@@ -1839,7 +1839,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_prelu_eltwise,
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 4},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 4},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 4},
-                        }), );
+                        }));
 
 class conv_int8_activation_eltwise_quantize : public ConvFusingTest {};
 TEST_P(conv_int8_activation_eltwise_quantize, fsv16) {
@@ -1900,7 +1900,7 @@ TEST_P(conv_int8_activation_eltwise_quantize, fsv32) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_activation_eltwise_quantize,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_activation_eltwise_quantize,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 5},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 5},
@@ -1914,7 +1914,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_activation_eltwise_quantize,
                                 bc_test_params{CASE_CONV_S8S8_4, 2, 5},
                                 bc_test_params{CASE_CONV_S8S8_7, 2, 5},
                                 bc_test_params{CASE_CONV_S8S8_8, 2, 5},
-                        }), );
+                        }));
 
 class conv_int8_activation_eltwise : public ConvFusingTest {};
 TEST_P(conv_int8_activation_eltwise, fsv16) {
@@ -1965,7 +1965,7 @@ TEST_P(conv_int8_activation_eltwise, fsv32) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_activation_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_activation_eltwise,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 4},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 4},
@@ -1979,7 +1979,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_activation_eltwise,
                                 bc_test_params{CASE_CONV_S8S8_4, 2, 4},
                                 bc_test_params{CASE_CONV_S8S8_7, 2, 4},
                                 bc_test_params{CASE_CONV_S8S8_8, 2, 4},
-                        }), );
+                        }));
 
 class conv_int8_quantize_u8 : public ConvFusingTest {};
 TEST_P(conv_int8_quantize_u8, per_channel) {
@@ -2018,7 +2018,7 @@ TEST_P(conv_int8_quantize_u8, per_tensor) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_quantize_u8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_quantize_u8,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 3},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 3},
@@ -2040,7 +2040,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_quantize_u8,
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 3},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 3},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 3},
-                        }), );
+                        }));
 
 class conv_int8_scale_quantize_i8 : public ConvFusingTest {};
 TEST_P(conv_int8_scale_quantize_i8, basic) {
@@ -2065,7 +2065,7 @@ TEST_P(conv_int8_scale_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_scale_quantize_i8,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 4},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 4},
@@ -2090,7 +2090,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_quantize_i8,
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 4},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 4},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 4},
-                        }), );
+                        }));
 
 class conv_int8_scale_quantize_i8_conv_b_fs_yx_fsv4_int8 : public ConvFusingTest {};
 TEST_P(conv_int8_scale_quantize_i8_conv_b_fs_yx_fsv4_int8, basic) {
@@ -2115,10 +2115,10 @@ TEST_P(conv_int8_scale_quantize_i8_conv_b_fs_yx_fsv4_int8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_quantize_i8_conv_b_fs_yx_fsv4_int8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_scale_quantize_i8_conv_b_fs_yx_fsv4_int8,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{ CASE_CONV_S8S8_11, 2, 4 },
-                        }), );
+                        }));
 
 class conv_int8_relu_quantize : public ConvFusingTest {};
 TEST_P(conv_int8_relu_quantize, i8) {
@@ -2160,7 +2160,7 @@ TEST_P(conv_int8_relu_quantize, u8) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_relu_quantize,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_relu_quantize,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 4},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 4},
@@ -2181,7 +2181,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_relu_quantize,
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 4},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 4},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 4},
-                        }), );
+                        }));
 
 class conv_int8_scale_activation_quantize_i8 : public ConvFusingTest {};
 TEST_P(conv_int8_scale_activation_quantize_i8, basic) {
@@ -2205,7 +2205,7 @@ TEST_P(conv_int8_scale_activation_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 5},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 5},
@@ -2226,7 +2226,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8,
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 5},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 5},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 5},
-                        }), );
+                        }));
 
 class conv_int8_scale_activation_quantize_i8_eltwise_fp32 : public ConvFusingTest {};
 TEST_P(conv_int8_scale_activation_quantize_i8_eltwise_fp32, basic) {
@@ -2251,7 +2251,7 @@ TEST_P(conv_int8_scale_activation_quantize_i8_eltwise_fp32, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8_eltwise_fp32,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8_eltwise_fp32,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 6},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 6},
@@ -2272,7 +2272,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8_eltw
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 6},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 6},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 6},
-                        }), );
+                        }));
 
 class conv_int8_scale_activation_quantize_i8_activation : public ConvFusingTest {};
 TEST_P(conv_int8_scale_activation_quantize_i8_activation, basic) {
@@ -2297,7 +2297,7 @@ TEST_P(conv_int8_scale_activation_quantize_i8_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8_activation,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8_activation,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 6},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 6},
@@ -2318,7 +2318,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8_acti
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 6},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 6},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 6},
-                        }), );
+                        }));
 
 
 class conv_int8_scale_activation_quantize_i8_eltwise_fp32_quantize_i8 : public ConvFusingTest {};
@@ -2350,7 +2350,7 @@ TEST_P(conv_int8_scale_activation_quantize_i8_eltwise_fp32_quantize_i8, DISABLED
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8_eltwise_fp32_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8_eltwise_fp32_quantize_i8,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 7},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 7},
@@ -2371,7 +2371,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_activation_quantize_i8_eltw
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 7},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 7},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 7},
-                        }), );
+                        }));
 
 class conv_int8_scale_prelu_quantize_i8_eltwise_fp32_quantize_i8_vec : public ConvFusingTest {};
 TEST_P(conv_int8_scale_prelu_quantize_i8_eltwise_fp32_quantize_i8_vec, vector_ops) {
@@ -2438,13 +2438,13 @@ TEST_P(conv_int8_scale_prelu_quantize_i8_eltwise_fp32_quantize_i8_vec, vector_op
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_scale_prelu_quantize_i8_eltwise_fp32_quantize_i8_vec,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_scale_prelu_quantize_i8_eltwise_fp32_quantize_i8_vec,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_3, 2, 7},
                                 bc_test_params{CASE_CONV_U8S8_5, 2, 7},
                                 bc_test_params{CASE_CONV_S8S8_3, 2, 7},
                                 bc_test_params{CASE_CONV_S8S8_5, 2, 7},
-                        }), );
+                        }));
 
 class conv_int8_asymmetric_weights : public ConvFusingTest {};
 TEST_P(conv_int8_asymmetric_weights, basic) {
@@ -2492,7 +2492,7 @@ TEST_P(conv_int8_asymmetric_weights, basic) {
     compare(network_not_fused, network_fused, p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_asymmetric_weights,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_asymmetric_weights,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 2},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 2},
@@ -2511,7 +2511,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_asymmetric_weights,
                                 bc_test_params{CASE_CONV3D_S8S8_2, 2, 2},
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 2},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 2},
-                        }), );
+                        }));
 
 class conv_int8_asymmetric_data : public ConvFusingTest {};
 TEST_P(conv_int8_asymmetric_data, basic) {
@@ -2559,7 +2559,7 @@ TEST_P(conv_int8_asymmetric_data, basic) {
     compare(network_not_fused, network_fused, p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_asymmetric_data,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_asymmetric_data,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 3},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 3},
@@ -2580,7 +2580,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_asymmetric_data,
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 3},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 3},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 3},
-                        }), );
+                        }));
 
 class conv_int8_asymmetric_data_and_weights : public ConvFusingTest {};
 TEST_P(conv_int8_asymmetric_data_and_weights, basic) {
@@ -2630,7 +2630,7 @@ TEST_P(conv_int8_asymmetric_data_and_weights, basic) {
     compare(network_not_fused, network_fused, p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_asymmetric_data_and_weights,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_int8_asymmetric_data_and_weights,
                         ::testing::ValuesIn(std::vector<bc_test_params>{
                                 bc_test_params{CASE_CONV_U8S8_1, 2, 3},
                                 bc_test_params{CASE_CONV_U8S8_2, 2, 3},
@@ -2651,7 +2651,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_int8_asymmetric_data_and_weights,
                                 bc_test_params{CASE_CONV3D_S8S8_3, 2, 3},
                                 bc_test_params{CASE_CONV3D_S8S8_4, 2, 3},
                                 bc_test_params{CASE_CONV3D_S8S8_5, 2, 3},
-                        }), );
+                        }));
 
 
 class conv_i8_activation_eltwise_diff_sizes : public ConvEltwTest {};
@@ -2670,14 +2670,14 @@ TEST_P(conv_i8_activation_eltwise_diff_sizes, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_i8_activation_eltwise_diff_sizes,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_i8_activation_eltwise_diff_sizes,
                         ::testing::ValuesIn(std::vector<conv_eltw_test_params>{
                                 conv_eltw_test_params{CASE_CONV_ELTW_i8_1, 3, 4},
                                 conv_eltw_test_params{CASE_CONV_ELTW_i8_2, 2, 4},
                                 conv_eltw_test_params{CASE_CONV_ELTW_i8_3, 2, 4},
                                 conv_eltw_test_params{CASE_CONV_ELTW_i8_4, 2, 4},
                                 conv_eltw_test_params{CASE_CONV_ELTW_i8_5, 3, 4},
-                        }), );
+                        }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ----------------------------------- Force convolution kernel cases ---------------------------------- */
@@ -2724,9 +2724,9 @@ TEST_P(conv_fp16_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp16_activation, ::testing::ValuesIn(std::vector<bc_force_kernel_params>{
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp16_activation, ::testing::ValuesIn(std::vector<bc_force_kernel_params>{
                                                                            bc_force_kernel_params{CASE_CONV_FP16_13, 2, 3, "convolution_gpu_fs_byx_fsv32"},
-                                                                           }), );
+                                                                           }));
 
 
 class conv_fp16_scale : public ConvFusingForceKernelTest {};
@@ -2745,10 +2745,10 @@ TEST_P(conv_fp16_scale, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, conv_fp16_scale,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, conv_fp16_scale,
                         ::testing::ValuesIn(std::vector<bc_force_kernel_params>{
                                              bc_force_kernel_params{CASE_CONV_FP16_13, 2, 3, "convolution_gpu_fs_byx_fsv32"},
-                                             }), );
+                                             }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ---------------------------------------- FC cases --------------------------------------------------- */
@@ -2769,14 +2769,14 @@ TEST_P(fc_fp32_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, fc_fp32_activation, ::testing::ValuesIn(std::vector<bc_test_params>{
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, fc_fp32_activation, ::testing::ValuesIn(std::vector<bc_test_params>{
                                                                             bc_test_params{ CASE_FC_FP32_1, 2, 3 },
                                                                             bc_test_params{ CASE_FC_FP32_2, 2, 3 },
                                                                             bc_test_params{ CASE_FC_FP32_3, 2, 3 },
                                                                             bc_test_params{ CASE_FC_FP32_3D_1, 2, 3 },
                                                                             bc_test_params{ CASE_FC_FP32_3D_2, 2, 3 },
                                                                             bc_test_params{ CASE_FC_FP32_3D_3, 2, 3 },
-}), );
+}));
 
 class fc_fp32_bias : public FCFusingTest {};
 TEST_P(fc_fp32_bias, basic) {
@@ -2793,11 +2793,11 @@ TEST_P(fc_fp32_bias, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, fc_fp32_bias, ::testing::ValuesIn(std::vector<bc_test_params>{
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, fc_fp32_bias, ::testing::ValuesIn(std::vector<bc_test_params>{
                                                                             bc_test_params{ CASE_FC_FP32_1, 2, 3 },
                                                                             bc_test_params{ CASE_FC_FP32_2, 2, 3 },
                                                                             bc_test_params{ CASE_FC_FP32_3, 2, 3 },
-}), );
+}));
 
 class fc_int8_scale : public FCFusingTest {};
 TEST_P(fc_int8_scale, basic) {
@@ -2830,7 +2830,7 @@ TEST_P(fc_int8_scale, fp16_scale_out) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, fc_int8_scale,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, fc_int8_scale,
     ::testing::ValuesIn(std::vector<bc_test_params>{
                         bc_test_params{ CASE_FC_U8S8_1, 2, 3 },
                         bc_test_params{ CASE_FC_U8S8_2, 2, 3 },
@@ -2838,7 +2838,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, fc_int8_scale,
                         bc_test_params{ CASE_FC_U8S8_3D_1, 2, 3 },
                         bc_test_params{ CASE_FC_U8S8_3D_2, 2, 3 },
                         bc_test_params{ CASE_FC_U8S8_3D_3, 2, 3 },
-                        }), );
+                        }));
 
 class fc_int8_quantize_u8 : public FCFusingTest {};
 TEST_P(fc_int8_quantize_u8, basic) {
@@ -2859,7 +2859,7 @@ TEST_P(fc_int8_quantize_u8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu_fc, fc_int8_quantize_u8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu_fc, fc_int8_quantize_u8,
     ::testing::ValuesIn(std::vector<bc_test_params>{
         bc_test_params{CASE_FC_U8S8_1, 2, 3},
         bc_test_params{CASE_FC_U8S8_2, 2, 3},
@@ -2867,7 +2867,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu_fc, fc_int8_quantize_u8,
         bc_test_params{ CASE_FC_U8S8_3D_1, 2, 3 },
         bc_test_params{ CASE_FC_U8S8_3D_2, 2, 3 },
         bc_test_params{ CASE_FC_U8S8_3D_3, 2, 3 },
-        }), );
+        }));
 
 class fc_int8_scale_quantize_i8 : public FCFusingTest {};
 TEST_P(fc_int8_scale_quantize_i8, basic) {
@@ -2889,7 +2889,7 @@ TEST_P(fc_int8_scale_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, fc_int8_scale_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, fc_int8_scale_quantize_i8,
     ::testing::ValuesIn(std::vector<bc_test_params>{
         bc_test_params{CASE_FC_U8S8_1, 2, 4},
         bc_test_params{CASE_FC_U8S8_2, 2, 4},
@@ -2897,7 +2897,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, fc_int8_scale_quantize_i8,
         bc_test_params{ CASE_FC_U8S8_3D_1, 2, 4 },
         bc_test_params{ CASE_FC_U8S8_3D_2, 2, 4 },
         bc_test_params{ CASE_FC_U8S8_3D_3, 2, 4 },
-        }), );
+        }));
 
 
 
@@ -2923,7 +2923,7 @@ TEST_P(fc_int8_scale_activation_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, fc_int8_scale_activation_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, fc_int8_scale_activation_quantize_i8,
     ::testing::ValuesIn(std::vector<bc_test_params>{
         bc_test_params{CASE_FC_U8S8_1, 2, 5},
         bc_test_params{CASE_FC_U8S8_2, 2, 5},
@@ -2936,7 +2936,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, fc_int8_scale_activation_quantize_i8,
         bc_test_params{ CASE_FC_FP32_3D_1, 3, 5 },
         bc_test_params{ CASE_FC_FP32_3D_2, 3, 5 },
         bc_test_params{ CASE_FC_FP32_3D_3, 3, 5 },
-        }), );
+        }));
 
 
 /* ----------------------------------------------------------------------------------------------------- */
@@ -2981,14 +2981,14 @@ TEST_P(gemm_3in_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, gemm_3in_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, gemm_3in_quantize_i8,
     ::testing::ValuesIn(std::vector<gemm_test_params>{
                         gemm_test_params{ CASE_GEMM_3IN_FP16_1, 4, 5 },
                         gemm_test_params{ CASE_GEMM_3IN_S8S8_1, 4, 5 },
                         gemm_test_params{ CASE_GEMM_3IN_S8S8_2, 4, 5 },
                         gemm_test_params{ CASE_GEMM_3IN_S8S8_3, 4, 5 },
                       //gemm_test_params{ CASE_GEMM_3IN_FP32_1, 4, 5 },
-}), );
+}));
 
 class gemm_2in_quantize_u8 : public GemmFusingTest {};
 TEST_P(gemm_2in_quantize_u8, basic) {
@@ -3008,14 +3008,14 @@ TEST_P(gemm_2in_quantize_u8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, gemm_2in_quantize_u8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, gemm_2in_quantize_u8,
     ::testing::ValuesIn(std::vector<gemm_test_params>{
                         gemm_test_params{ CASE_GEMM_2IN_FP16_1, 3, 4 },
                         gemm_test_params{ CASE_GEMM_2IN_U8U8_1, 3, 4 },
                         gemm_test_params{ CASE_GEMM_2IN_U8U8_2, 3, 4 },
                         gemm_test_params{ CASE_GEMM_2IN_U8U8_3, 3, 4 },
                       //gemm_test_params{ CASE_GEMM_2IN_FP32_1, 3, 4 },
-}), );
+}));
 
 class gemm_2in_scale : public GemmFusingTest {};
 TEST_P(gemm_2in_scale, basic) {
@@ -3046,14 +3046,14 @@ TEST_P(gemm_2in_scale, fp16_scale_out) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, gemm_2in_scale,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, gemm_2in_scale,
     ::testing::ValuesIn(std::vector<gemm_test_params>{
                         gemm_test_params{ CASE_GEMM_2IN_FP32_1, 3, 4 },
                         gemm_test_params{ CASE_GEMM_2IN_FP16_1, 3, 4 },
                         gemm_test_params{ CASE_GEMM_2IN_U8U8_1, 3, 4 },
                         gemm_test_params{ CASE_GEMM_2IN_U8U8_2, 3, 4 },
                         gemm_test_params{ CASE_GEMM_2IN_U8U8_3, 3, 4 },
-}), );
+}));
 
 class gemm_2in_act_scale_quantize_i8 : public GemmFusingTest {};
 TEST_P(gemm_2in_act_scale_quantize_i8, basic) {
@@ -3076,13 +3076,13 @@ TEST_P(gemm_2in_act_scale_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, gemm_2in_act_scale_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, gemm_2in_act_scale_quantize_i8,
     ::testing::ValuesIn(std::vector<gemm_test_params>{
                         gemm_test_params{ CASE_GEMM_2IN_FP32_1, 3, 6 },
                         gemm_test_params{ CASE_GEMM_2IN_FP16_1, 3, 6 },
                         gemm_test_params{ CASE_GEMM_2IN_U8S8_1, 3, 6 },
                         gemm_test_params{ CASE_GEMM_2IN_S8U8_1, 3, 6 },
-}), );
+}));
 
 class gemm_2in_act_scale_quantize_eltwise_i8 : public GemmFusingTest {};
 TEST_P(gemm_2in_act_scale_quantize_eltwise_i8, basic) {
@@ -3107,13 +3107,13 @@ TEST_P(gemm_2in_act_scale_quantize_eltwise_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, gemm_2in_act_scale_quantize_eltwise_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, gemm_2in_act_scale_quantize_eltwise_i8,
     ::testing::ValuesIn(std::vector<gemm_test_params>{
                         gemm_test_params{ CASE_GEMM_ELTWISE_2IN_FP32_1, 3, 7 },
                         gemm_test_params{ CASE_GEMM_ELTWISE_2IN_FP16_1, 3, 7 },
                         gemm_test_params{ CASE_GEMM_ELTWISE_2IN_U8S8_1, 3, 7 },
                         gemm_test_params{ CASE_GEMM_ELTWISE_2IN_S8U8_1, 3, 7 },
-}), );
+}));
 
 class gemm_2in_act_scale_eltwise : public GemmFusingTest {};
 TEST_P(gemm_2in_act_scale_eltwise, basic) {
@@ -3150,13 +3150,13 @@ TEST_P(gemm_2in_act_scale_eltwise, broadcast_eltwise) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, gemm_2in_act_scale_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, gemm_2in_act_scale_eltwise,
     ::testing::ValuesIn(std::vector<gemm_test_params>{
                         gemm_test_params{ CASE_GEMM_ELTWISE_2IN_FP32_1, 3, 6 },
                         gemm_test_params{ CASE_GEMM_ELTWISE_2IN_FP16_1, 3, 6 },
                         gemm_test_params{ CASE_GEMM_ELTWISE_2IN_U8S8_1, 3, 6 },
                         gemm_test_params{ CASE_GEMM_ELTWISE_2IN_S8U8_1, 3, 6 },
-}), );
+}));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ---------------------------------------- Resample cases --------------------------------------------- */
@@ -3214,7 +3214,7 @@ TEST_P(resample_quantize, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, resample_quantize,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, resample_quantize,
     ::testing::ValuesIn(std::vector<resample_test_params>{
                         resample_test_params{ CASE_RESAMPLE_FP32_1, 2, 3 },
                         resample_test_params{ CASE_RESAMPLE_FP32_2, 2, 3 },
@@ -3237,7 +3237,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, resample_quantize,
                         // resample_test_params{ CASE_RESAMPLE_FP16_7, 2, 3 },
                         // resample_test_params{ CASE_RESAMPLE_FP16_8, 2, 3 },
                         // resample_test_params{ CASE_RESAMPLE_FP16_9, 2, 3 },
-}), );
+}));
 
 class resample_scale_activation_eltwise : public ResamplePrimitiveFusingTest {};
 TEST_P(resample_scale_activation_eltwise, basic) {
@@ -3256,7 +3256,7 @@ TEST_P(resample_scale_activation_eltwise, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, resample_scale_activation_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, resample_scale_activation_eltwise,
     ::testing::ValuesIn(std::vector<resample_test_params>{
                         resample_test_params{ CASE_RESAMPLE_FP32_1, 2, 5 },
                         resample_test_params{ CASE_RESAMPLE_FP32_2, 2, 5 },
@@ -3293,7 +3293,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, resample_scale_activation_eltwise,
                         resample_test_params{ CASE_RESAMPLE_U8_2, 2, 5 },
                         resample_test_params{ CASE_RESAMPLE_U8_3, 2, 5 },
                         resample_test_params{ CASE_RESAMPLE_U8_4, 2, 5 },
-}), );
+}));
 
 class resample_quantize_concat : public ResamplePrimitiveFusingTest {};
 TEST_P(resample_quantize_concat, along_f) {
@@ -3320,7 +3320,7 @@ TEST_P(resample_quantize_concat, along_f) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, resample_quantize_concat,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, resample_quantize_concat,
     ::testing::ValuesIn(std::vector<resample_test_params>{
                         resample_test_params{ CASE_RESAMPLE_FP32_1, 3, 6 },
                         resample_test_params{ CASE_RESAMPLE_FP32_2, 3, 6 },
@@ -3353,7 +3353,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, resample_quantize_concat,
 
                         resample_test_params{ CASE_RESAMPLE_U8_3, 3, 6 },
                         resample_test_params{ CASE_RESAMPLE_U8_4, 3, 6 },
-}), );
+}));
 
 class resample_scale_concat : public ResamplePrimitiveFusingTest {};
 TEST_P(resample_scale_concat, along_f) {
@@ -3376,7 +3376,7 @@ TEST_P(resample_scale_concat, along_f) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, resample_scale_concat,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, resample_scale_concat,
     ::testing::ValuesIn(std::vector<resample_test_params>{
                         resample_test_params{ CASE_RESAMPLE_FP32_1, 3, 6 },
                         resample_test_params{ CASE_RESAMPLE_FP32_2, 3, 6 },
@@ -3413,7 +3413,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, resample_scale_concat,
                         resample_test_params{ CASE_RESAMPLE_U8_2, 3, 6 },
                         resample_test_params{ CASE_RESAMPLE_U8_3, 3, 6 },
                         resample_test_params{ CASE_RESAMPLE_U8_4, 3, 6 },
-}), );
+}));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* --------------------------------------- MVN cases --------------------------------------------------- */
@@ -3502,7 +3502,7 @@ TEST_P(mvn_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, mvn_activation,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, mvn_activation,
     ::testing::ValuesIn(std::vector<mvn_test_params>{
                         mvn_test_params{ CASE_MVN_F32_1, 2, 3 },
                         mvn_test_params{ CASE_MVN_F32_2, 2, 3 },
@@ -3524,7 +3524,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, mvn_activation,
                         mvn_test_params{ CASE_MVN_U8_4, 2, 3 },
                         mvn_test_params{ CASE_MVN_3D_U8_1, 2, 3 },
                         mvn_test_params{ CASE_MVN_3D_U8_2, 2, 3 },
-}), );
+}));
 
 class mvn_scale_quantize_i8 : public MVNFusingTest {};
 TEST_P(mvn_scale_quantize_i8, basic) {
@@ -3546,7 +3546,7 @@ TEST_P(mvn_scale_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, mvn_scale_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, mvn_scale_quantize_i8,
     ::testing::ValuesIn(std::vector<mvn_test_params>{
         // Full fusing for fp input not supported yet, it may lead to output padding and non-optimal kernel
         // mvn_test_params{ CASE_MVN_F32_1, 2, 4 },
@@ -3569,7 +3569,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, mvn_scale_quantize_i8,
         mvn_test_params{ CASE_MVN_U8_4, 2, 4 },
         mvn_test_params{ CASE_MVN_3D_U8_1, 2, 4 },
         mvn_test_params{ CASE_MVN_3D_U8_2, 2, 4 },
-}), );
+}));
 
 class mvn_scale_activation_eltwise_fp32_quantize_i8 : public MVNFusingTest {};
 TEST_P(mvn_scale_activation_eltwise_fp32_quantize_i8, basic) {
@@ -3594,7 +3594,7 @@ TEST_P(mvn_scale_activation_eltwise_fp32_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, mvn_scale_activation_eltwise_fp32_quantize_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, mvn_scale_activation_eltwise_fp32_quantize_i8,
     ::testing::ValuesIn(std::vector<mvn_test_params>{
         // Full using for fp input not supported yet, it may lead to output padding and non-optimal kernel
         // mvn_test_params{ CASE_MVN_F32_1, 2, 7 },
@@ -3629,7 +3629,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, mvn_scale_activation_eltwise_fp32_quantize_
         mvn_test_params{ CASE_MVN_3D_U8_3, 2, 6 },
         mvn_test_params{ CASE_MVN_3D_U8_4, 2, 6 },
         mvn_test_params{ CASE_MVN_3D_U8_5, 3, 6 },
-}), );
+}));
 
 class mvn_eltwise : public MVNFusingTest {};
 TEST_P(mvn_eltwise, basic) {
@@ -3645,7 +3645,7 @@ TEST_P(mvn_eltwise, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, mvn_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, mvn_eltwise,
     ::testing::ValuesIn(std::vector<mvn_test_params>{
         mvn_test_params{ CASE_MVN_I8_5, 2, 3 },
         mvn_test_params{ CASE_MVN_I8_6, 2, 3 },
@@ -3665,7 +3665,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, mvn_eltwise,
         mvn_test_params{ CASE_MVN_3D_U8_3, 2, 3 },
         mvn_test_params{ CASE_MVN_3D_U8_4, 2, 3 },
         mvn_test_params{ CASE_MVN_3D_U8_5, 3, 3 },
-}), );
+}));
 
 class mvn_eltwise_f16 : public MVNFusingTest {};
 TEST_P(mvn_eltwise_f16, basic) {
@@ -3681,11 +3681,11 @@ TEST_P(mvn_eltwise_f16, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, mvn_eltwise_f16,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, mvn_eltwise_f16,
     ::testing::ValuesIn(std::vector<mvn_test_params>{
         mvn_test_params{ CASE_MVN_I8_6, 2, 3 },
         mvn_test_params{ CASE_MVN_U8_2, 2, 3 },
-}), );
+}));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ---------------------------------------- LRN cases -------------------------------------------------- */
@@ -3811,7 +3811,7 @@ TEST_P(lrn_fp32_quantize_u8_scale_activation, per_channel) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         lrn_fp32_quantize_u8_scale_activation,
                         ::testing::ValuesIn(std::vector<lrn_test_params>{
                             // InputDataType = FP32   OutputDataType = FP32
@@ -3835,7 +3835,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             lrn_test_params{CASE_LRN_FP32_TO_FP16_4, 2, 5, lrn_norm_region_across_channel, "lrn_gpu_across_channel_multiple_features"},
                             lrn_test_params{CASE_LRN_FP32_TO_FP16_5, 2, 5, lrn_norm_region_across_channel, "lrn_gpu_across_channel_multiple_features_fsv16"},
 
-                        }), );
+                        }));
 
 class lrn_fp32_quantize_i8_scale_activation : public LrnFusingTest {};
 TEST_P(lrn_fp32_quantize_i8_scale_activation, basic) {
@@ -3862,7 +3862,7 @@ TEST_P(lrn_fp32_quantize_i8_scale_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         lrn_fp32_quantize_i8_scale_activation,
                         ::testing::ValuesIn(std::vector<lrn_test_params>{
                             // InputDataType = FP32   OutputDataType = INT8
@@ -3878,7 +3878,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
 
                             // InputDataType = FP16   OutputDataType = INT8/UINT8 can't be tested for now, because quantize
                             // primitive doesn't support input type FP16 while fusing (prepare_quantization.cpp :114 -> prepare_primitive_fusing.cpp :474)
-                        }), );
+                        }));
 
 class lrn_fp32_scale_activation_quantize_u8 : public LrnFusingTest {};
 TEST_P(lrn_fp32_scale_activation_quantize_u8, basic) {
@@ -3905,7 +3905,7 @@ TEST_P(lrn_fp32_scale_activation_quantize_u8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         lrn_fp32_scale_activation_quantize_u8,
                         ::testing::ValuesIn(std::vector<lrn_test_params>{
                             // InputDataType = FP32   OutputDataType = UINT8
@@ -3918,7 +3918,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             lrn_test_params{CASE_LRN_FP32_3, 2, 5, lrn_norm_region_within_channel, "lrn_within_channel_byxf_opt"},
                             lrn_test_params{CASE_LRN_FP32_4, 2, 5, lrn_norm_region_across_channel, "lrn_gpu_across_channel_multiple_features"},
                             lrn_test_params{CASE_LRN_FP32_5, 2, 5, lrn_norm_region_across_channel, "lrn_gpu_across_channel_multiple_features_fsv16"},
-                        }), );
+                        }));
 
 class lrn_fp16_scale_activation : public LrnFusingTest {};
 TEST_P(lrn_fp16_scale_activation, basic) {
@@ -3940,7 +3940,7 @@ TEST_P(lrn_fp16_scale_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         lrn_fp16_scale_activation,
                         ::testing::ValuesIn(std::vector<lrn_test_params>{
                             // InputDataType = FP16   OutputDataType = FP16
@@ -3952,7 +3952,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             lrn_test_params{CASE_LRN_FP16_3, 2, 4, lrn_norm_region_within_channel, "lrn_within_channel_byxf_opt"},
                             lrn_test_params{CASE_LRN_FP16_4, 2, 4, lrn_norm_region_across_channel, "lrn_gpu_across_channel_multiple_features"},
                             lrn_test_params{CASE_LRN_FP16_5, 2, 4, lrn_norm_region_across_channel, "lrn_gpu_across_channel_multiple_features_fsv16"},
-                        }), );
+                        }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* -------------------------------- Activation cases --------------------------------------------------- */
@@ -4062,7 +4062,7 @@ TEST_P(activation_quantize_i8, per_channel) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     activation_quantize_i8,
     ::testing::ValuesIn(std::vector<activation_test_params>{
@@ -4080,9 +4080,9 @@ INSTANTIATE_TEST_CASE_P(
         activation_test_params{CASE_ACTIVATION_3D_F32_0, 2, 3, "activation_ref"},
         activation_test_params{CASE_ACTIVATION_3D_F32_1, 2, 3, "activation_ref"},
         activation_test_params{CASE_ACTIVATION_3D_F32_2, 2, 3, "activation_ref"},
-    }), );
+    }));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     DISABLED_fusings_gpu,
     activation_quantize_i8,
     ::testing::ValuesIn(std::vector<activation_test_params>{
@@ -4091,7 +4091,7 @@ INSTANTIATE_TEST_CASE_P(
         activation_test_params{CASE_ACTIVATION_F32_7, 2, 3, "activation_ref"},     // FIXME - accuracy bug
         activation_test_params{CASE_ACTIVATION_3D_F32_3, 2, 3, "activation_ref"},  // FIXME - accuracy bug
         activation_test_params{CASE_ACTIVATION_3D_F32_5, 2, 3, "activation_ref"},  // FIXME - accuracy bug
-    }), );
+    }));
 
 class activation_scale_activation_quantize_u8 : public ActivationFusingTest {};
 TEST_P(activation_scale_activation_quantize_u8, basic) {
@@ -4130,7 +4130,7 @@ TEST_P(activation_scale_activation_quantize_u8, per_channel) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     activation_scale_activation_quantize_u8,
     ::testing::ValuesIn(std::vector<activation_test_params>{
@@ -4151,14 +4151,14 @@ INSTANTIATE_TEST_CASE_P(
         activation_test_params{CASE_ACTIVATION_3D_F32_0, 2, 5, "activation_ref"},
         activation_test_params{CASE_ACTIVATION_3D_F32_1, 2, 5, "activation_ref"},
         activation_test_params{CASE_ACTIVATION_3D_F32_2, 2, 5, "activation_ref"},
-    }), );
+    }));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     DISABLED_fusings_gpu,
     activation_scale_activation_quantize_u8,
     ::testing::ValuesIn(std::vector<activation_test_params>{
         activation_test_params{CASE_ACTIVATION_3D_F32_5, 2, 5, "activation_ref"},  // FIXME - accuracy bug
-    }), );
+    }));
 
 class activation_scale_activation : public ActivationFusingTest {};
 TEST_P(activation_scale_activation, basic) {
@@ -4174,7 +4174,7 @@ TEST_P(activation_scale_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     activation_scale_activation,
     ::testing::ValuesIn(std::vector<activation_test_params>{
@@ -4227,15 +4227,15 @@ INSTANTIATE_TEST_CASE_P(
         activation_test_params{CASE_ACTIVATION_I8_1, 2, 4, "activation_ref"},
         activation_test_params{CASE_ACTIVATION_I8_2, 2, 4, "activation_ref"},
         activation_test_params{CASE_ACTIVATION_3D_I8_1, 2, 4, "activation_ref"}
-    }), );
+    }));
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     DISABLED_fusings_gpu,
     activation_scale_activation,
     ::testing::ValuesIn(std::vector<activation_test_params>{
         activation_test_params{CASE_ACTIVATION_3D_F32_4, 2, 4, "activation_ref"},  // FIXME - accuracy bug
         activation_test_params{CASE_ACTIVATION_3D_F32_5, 2, 4, "activation_ref"},  // FIXME - accuracy bug
-    }), );
+    }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* --------------------------------------- Deconvolution cases ----------------------------------------- */
@@ -4373,7 +4373,7 @@ TEST_P(deconv_actv, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_actv,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, deconv_actv,
     ::testing::ValuesIn(std::vector<deconv_test_params>{
         deconv_test_params{ CASE_DECONV_FP32_1, 2, 3 },
         deconv_test_params{ CASE_DECONV_FP32_2, 2, 3 },
@@ -4448,7 +4448,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_actv,
         deconv_test_params{ CASE_DECONV_S8S8_3D_6, 2, 3 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_7, 2, 3 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_8, 2, 3 },
-}), );
+}));
 
 
 class deconv_bias : public DeconvolutionFusingTest {};
@@ -4467,7 +4467,7 @@ TEST_P(deconv_bias, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_bias,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, deconv_bias,
     ::testing::ValuesIn(std::vector<deconv_test_params>{
         deconv_test_params{ CASE_DECONV_FP32_1, 2, 3 },
         deconv_test_params{ CASE_DECONV_FP32_2, 2, 3 },
@@ -4500,7 +4500,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_bias,
         deconv_test_params{ CASE_DECONV_S8S8_3D_1, 2, 3 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_2, 2, 3 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_3, 2, 3 },
-}), );
+}));
 
 class deconv_scale : public DeconvolutionFusingTest {};
 TEST_P(deconv_scale, basic) {
@@ -4531,7 +4531,7 @@ TEST_P(deconv_scale, fp16_scale_out) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_scale,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, deconv_scale,
     ::testing::ValuesIn(std::vector<deconv_test_params>{
         deconv_test_params{ CASE_DECONV_U8S8_1, 2, 3 },
         deconv_test_params{ CASE_DECONV_U8S8_2, 2, 3 },
@@ -4568,7 +4568,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_scale,
         deconv_test_params{ CASE_DECONV_S8S8_3D_6, 2, 3 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_7, 2, 3 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_8, 2, 3 },
-}), );
+}));
 
 class deconv_actv_eltw_actv : public DeconvolutionFusingTest {};
 TEST_P(deconv_actv_eltw_actv, basic) {
@@ -4588,7 +4588,7 @@ TEST_P(deconv_actv_eltw_actv, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_actv_eltw_actv,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, deconv_actv_eltw_actv,
     ::testing::ValuesIn(std::vector<deconv_test_params>{
         // Some fusings disabled under deconvolution -> convolution optimization
         deconv_test_params{ CASE_DECONV_FP32_1, 3, 5 },
@@ -4664,7 +4664,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_actv_eltw_actv,
         deconv_test_params{ CASE_DECONV_S8S8_3D_6, 2, 5 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_7, 2, 5 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_8, 2, 5 },
-}), );
+}));
 
 class deconv_scale_actv_quant_i8 : public DeconvolutionFusingTest {};
 TEST_P(deconv_scale_actv_quant_i8, basic) {
@@ -4688,7 +4688,7 @@ TEST_P(deconv_scale_actv_quant_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_scale_actv_quant_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, deconv_scale_actv_quant_i8,
     ::testing::ValuesIn(std::vector<deconv_test_params>{
         deconv_test_params{ CASE_DECONV_FP32_1, 2, 5 },
         deconv_test_params{ CASE_DECONV_FP32_2, 2, 5 },
@@ -4765,7 +4765,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_scale_actv_quant_i8,
         deconv_test_params{ CASE_DECONV_S8S8_3D_6, 2, 5 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_7, 2, 5 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_8, 2, 5 },
-}), );
+}));
 
 class deconv_scale_actv_quant_u8_eltw_scale_actv_quant_i8 : public DeconvolutionFusingTest {};
 TEST_P(deconv_scale_actv_quant_u8_eltw_scale_actv_quant_i8, basic) {
@@ -4799,7 +4799,7 @@ TEST_P(deconv_scale_actv_quant_u8_eltw_scale_actv_quant_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_scale_actv_quant_u8_eltw_scale_actv_quant_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, deconv_scale_actv_quant_u8_eltw_scale_actv_quant_i8,
     ::testing::ValuesIn(std::vector<deconv_test_params>{
         deconv_test_params{ CASE_DECONV_FP32_1, 2, 9 },
         deconv_test_params{ CASE_DECONV_FP32_2, 2, 9 },
@@ -4874,7 +4874,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_scale_actv_quant_u8_eltw_scale_actv_
         deconv_test_params{ CASE_DECONV_S8S8_3D_6, 2, 9 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_7, 2, 9 },
         deconv_test_params{ CASE_DECONV_S8S8_3D_8, 2, 9 },
-}), );
+}));
 
 class deconv_scale_activation_quantize_i8_eltwise_quantize_u8 : public ConvEltwTest {};
 TEST_P(deconv_scale_activation_quantize_i8_eltwise_quantize_u8, basic) {
@@ -4903,7 +4903,7 @@ TEST_P(deconv_scale_activation_quantize_i8_eltwise_quantize_u8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_scale_activation_quantize_i8_eltwise_quantize_u8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, deconv_scale_activation_quantize_i8_eltwise_quantize_u8,
                         ::testing::ValuesIn(std::vector<conv_eltw_test_params>{
                                 conv_eltw_test_params{CASE_DECONV_ELTW_FP32_1, 2, 7},
                                 conv_eltw_test_params{CASE_DECONV_ELTW_FP32_2, 2, 7},
@@ -4920,7 +4920,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_scale_activation_quantize_i8_eltwise
                                 conv_eltw_test_params{CASE_DECONV_ELTW_i8_4, 2, 7},
                                 conv_eltw_test_params{CASE_DECONV_ELTW_i8_5, 2, 7},
 
-                        }), );
+                        }));
 
 
 class deconv_activation_eltwise_diff_sizes : public ConvEltwTest {};
@@ -4939,7 +4939,7 @@ TEST_P(deconv_activation_eltwise_diff_sizes, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_activation_eltwise_diff_sizes,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, deconv_activation_eltwise_diff_sizes,
                         ::testing::ValuesIn(std::vector<conv_eltw_test_params>{
                                 conv_eltw_test_params{CASE_DECONV_ELTW_FP32_1, 2, 4},
                                 conv_eltw_test_params{CASE_DECONV_ELTW_FP32_2, 2, 4},
@@ -4955,7 +4955,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, deconv_activation_eltwise_diff_sizes,
                                 conv_eltw_test_params{CASE_DECONV_ELTW_i8_3, 2, 4},
                                 conv_eltw_test_params{CASE_DECONV_ELTW_i8_4, 2, 4},
                                 conv_eltw_test_params{CASE_DECONV_ELTW_i8_5, 2, 4},
-                        }), );
+                        }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* --------------------------------------- Pooling cases ----------------------------------------------- */
@@ -5086,7 +5086,7 @@ TEST_P(pooling_f32_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         pooling_f32_activation,
                         ::testing::ValuesIn(std::vector<pooling_test_params>{
                             pooling_test_params{CASE_POOLING_F32_1, 2, 3, pooling_mode::max, ""},
@@ -5103,7 +5103,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             pooling_test_params{CASE_POOLING_I8_1, 2, 3, pooling_mode::average, ""},
                             pooling_test_params{CASE_POOLING_I8_2, 2, 3, pooling_mode::max, ""},
                             pooling_test_params{CASE_POOLING_I8_2, 2, 3, pooling_mode::average, ""},
-                        }), );
+                        }));
 
 class pooling_f32_scale : public PoolingFusingTest {};
 TEST_P(pooling_f32_scale, basic) {
@@ -5132,7 +5132,7 @@ TEST_P(pooling_f32_scale, fp16_scale_out) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         pooling_f32_scale,
                         ::testing::ValuesIn(std::vector<pooling_test_params>{
                             pooling_test_params{CASE_POOLING_F32_1, 2, 3, pooling_mode::max, ""},
@@ -5147,7 +5147,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             pooling_test_params{CASE_POOLING_I8_1, 2, 3, pooling_mode::average, ""},
                             pooling_test_params{CASE_POOLING_I8_2, 2, 3, pooling_mode::max, ""},
                             pooling_test_params{CASE_POOLING_I8_2, 2, 3, pooling_mode::average, ""},
-                        }), );
+                        }));
 
 class pooling_scale_activation_quantize : public PoolingFusingTest {};
 TEST_P(pooling_scale_activation_quantize, basic) {
@@ -5207,7 +5207,7 @@ TEST_P(pooling_scale_activation_quantize, per_channel) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                          pooling_scale_activation_quantize,
                          ::testing::ValuesIn(std::vector<pooling_test_params>{
                             // Input type: FP32
@@ -5245,13 +5245,13 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             pooling_test_params{CASE_POOLING_U8_5, 2, 5, pooling_mode::max, "pooling_gpu_int8_ref"},
                             pooling_test_params{CASE_POOLING_U8_6, 2, 5, pooling_mode::average, "pooling_gpu_int8_ref"},
                             pooling_test_params{CASE_POOLING_U8_6, 2, 5, pooling_mode::max, "pooling_gpu_int8_ref"},
-                        }), );
+                        }));
 
-INSTANTIATE_TEST_CASE_P(DISABLED_fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(DISABLED_fusings_gpu,
                          pooling_scale_activation_quantize,
                          ::testing::ValuesIn(std::vector<pooling_test_params>{
                             pooling_test_params{CASE_POOLING_F32_3, 2, 5, pooling_mode::average, "pooling_gpu_average_opt"},  //currently not enabled, fusing not upported
-                        }), );
+                        }));
 
 class pooling_scale_activation : public PoolingFusingTest {};
 TEST_P(pooling_scale_activation, basic) {
@@ -5282,7 +5282,7 @@ TEST_P(pooling_scale_activation, eltwise_mul) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         pooling_scale_activation,
                         ::testing::ValuesIn(std::vector<pooling_test_params>{
                             // Input type: F32
@@ -5396,7 +5396,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             pooling_test_params{CASE_POOLING_U8_FP16_5, 2, 4, pooling_mode::max, "pooling_gpu_int8_ref"},
                             pooling_test_params{CASE_POOLING_U8_FP16_6, 2, 4, pooling_mode::average, "pooling_gpu_int8_ref"},
                             pooling_test_params{CASE_POOLING_U8_FP16_6, 2, 4, pooling_mode::max, "pooling_gpu_int8_ref"},
-                     }), );
+                     }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* -------------------------------- DepthToSpace cases ------------------------------------------------- */
@@ -5461,7 +5461,7 @@ TEST_P(depth_to_space_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     depth_to_space_quantize_i8,
     ::testing::ValuesIn(std::vector<depth_to_space_test_params>{
@@ -5469,7 +5469,7 @@ INSTANTIATE_TEST_CASE_P(
         depth_to_space_test_params{CASE_DEPTH_TO_SPACE_F32_2, 2, 3},
         depth_to_space_test_params{CASE_DEPTH_TO_SPACE_F16_1, 2, 3},
         depth_to_space_test_params{CASE_DEPTH_TO_SPACE_F16_2, 2, 3},
-    }), );
+    }));
 
 class depth_to_space_scale_act_eltwise_quantize_u8 : public DepthToSpaceFusingsTest {};
 TEST_P(depth_to_space_scale_act_eltwise_quantize_u8, basic) {
@@ -5492,7 +5492,7 @@ TEST_P(depth_to_space_scale_act_eltwise_quantize_u8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     depth_to_space_scale_act_eltwise_quantize_u8,
     ::testing::ValuesIn(std::vector<depth_to_space_test_params>{
@@ -5504,7 +5504,7 @@ INSTANTIATE_TEST_CASE_P(
         depth_to_space_test_params{CASE_DEPTH_TO_SPACE_U8_2, 2, 6},
         depth_to_space_test_params{CASE_DEPTH_TO_SPACE_I8_1, 2, 6},
         depth_to_space_test_params{CASE_DEPTH_TO_SPACE_I8_2, 2, 6},
-    }), );
+    }));
 
 
 class depth_to_space_scale_act_eltw : public DepthToSpaceFusingsTest {};
@@ -5523,7 +5523,7 @@ TEST_P(depth_to_space_scale_act_eltw, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     depth_to_space_scale_act_eltw,
     ::testing::ValuesIn(std::vector<depth_to_space_test_params>{
@@ -5535,7 +5535,7 @@ INSTANTIATE_TEST_CASE_P(
         depth_to_space_test_params{CASE_DEPTH_TO_SPACE_U8_2, 2, 5},
         depth_to_space_test_params{CASE_DEPTH_TO_SPACE_I8_1, 2, 5},
         depth_to_space_test_params{CASE_DEPTH_TO_SPACE_I8_2, 2, 5},
-    }), );
+    }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* -------------------------------- SpaceToDepth cases ------------------------------------------------- */
@@ -5600,7 +5600,7 @@ TEST_P(space_to_depth_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     space_to_depth_quantize_i8,
     ::testing::ValuesIn(std::vector<space_to_depth_params>{
@@ -5608,7 +5608,7 @@ INSTANTIATE_TEST_CASE_P(
         space_to_depth_params{CASE_SPACE_TO_DEPTH_F32_2, 2, 3},
         space_to_depth_params{CASE_SPACE_TO_DEPTH_F16_1, 2, 3},
         space_to_depth_params{CASE_SPACE_TO_DEPTH_F16_2, 2, 3},
-    }), );
+    }));
 
 class space_to_depth_scale_act_eltwise_quantize_u8 : public SpaceToDepthFusingsTest {};
 TEST_P(space_to_depth_scale_act_eltwise_quantize_u8, basic) {
@@ -5631,7 +5631,7 @@ TEST_P(space_to_depth_scale_act_eltwise_quantize_u8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     space_to_depth_scale_act_eltwise_quantize_u8,
     ::testing::ValuesIn(std::vector<space_to_depth_params>{
@@ -5643,7 +5643,7 @@ INSTANTIATE_TEST_CASE_P(
         space_to_depth_params{CASE_SPACE_TO_DEPTH_U8_2, 2, 6},
         space_to_depth_params{CASE_SPACE_TO_DEPTH_I8_1, 2, 6},
         space_to_depth_params{CASE_SPACE_TO_DEPTH_I8_2, 2, 6},
-    }), );
+    }));
 
 
 class space_to_depth_scale_act_eltw : public SpaceToDepthFusingsTest {};
@@ -5662,7 +5662,7 @@ TEST_P(space_to_depth_scale_act_eltw, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     space_to_depth_scale_act_eltw,
     ::testing::ValuesIn(std::vector<space_to_depth_params>{
@@ -5674,7 +5674,7 @@ INSTANTIATE_TEST_CASE_P(
         space_to_depth_params{CASE_SPACE_TO_DEPTH_U8_2, 2, 5},
         space_to_depth_params{CASE_SPACE_TO_DEPTH_I8_1, 2, 5},
         space_to_depth_params{CASE_SPACE_TO_DEPTH_I8_2, 2, 5},
-    }), );
+    }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ------------------------------------------ Gather cases --------------------------------------------- */
@@ -5779,7 +5779,7 @@ TEST_P(gather_quantize, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, gather_quantize,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, gather_quantize,
     ::testing::ValuesIn(std::vector<gather_test_params>{
                         gather_test_params{ CASE_GATHER_FP32_1, 2, 3 },
                         gather_test_params{ CASE_GATHER_FP32_2, 2, 3 },
@@ -5804,7 +5804,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, gather_quantize,
                         gather_test_params{ CASE_GATHER_5D_FP16_3, 2, 3 },
                         gather_test_params{ CASE_GATHER_5D_FP16_4, 2, 3 },
                         gather_test_params{ CASE_GATHER_5D_FP16_5, 2, 3 },
-}), );
+}));
 
 class gather_scale_activation : public GatherPrimitiveFusingTest {};
 TEST_P(gather_scale_activation, basic) {
@@ -5822,7 +5822,7 @@ TEST_P(gather_scale_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, gather_scale_activation,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, gather_scale_activation,
     ::testing::ValuesIn(std::vector<gather_test_params>{
                         gather_test_params{ CASE_GATHER_FP32_1, 2, 4 },
                         gather_test_params{ CASE_GATHER_FP32_2, 2, 4 },
@@ -5847,7 +5847,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, gather_scale_activation,
                         gather_test_params{ CASE_GATHER_5D_FP16_3, 2, 4 },
                         gather_test_params{ CASE_GATHER_5D_FP16_4, 2, 4 },
                         gather_test_params{ CASE_GATHER_5D_FP16_5, 2, 4 },
-}), );
+}));
 
 
 /* ----------------------------------------------------------------------------------------------------- */
@@ -5956,7 +5956,7 @@ TEST_P(scatter_update_quantize, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_update_quantize,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, scatter_update_quantize,
     ::testing::ValuesIn(std::vector<scatter_update_test_params>{
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_FP32_1, 2, 3 },
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_FP32_2, 2, 3 },
@@ -5982,7 +5982,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_update_quantize,
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_5D_FP16_3, 2, 3 },
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_5D_FP16_4, 2, 3 },
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_5D_FP16_5, 2, 3 },
-}), );
+}));
 
 class scatter_update_scale_activation : public ScatterUpdatePrimitiveFusingTest {};
 TEST_P(scatter_update_scale_activation, basic) {
@@ -6000,7 +6000,7 @@ TEST_P(scatter_update_scale_activation, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_update_scale_activation,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, scatter_update_scale_activation,
     ::testing::ValuesIn(std::vector<scatter_update_test_params>{
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_FP32_1, 2, 4 },
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_FP32_2, 2, 4 },
@@ -6026,7 +6026,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_update_scale_activation,
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_5D_FP16_3, 2, 4 },
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_5D_FP16_4, 2, 4 },
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_5D_FP16_5, 2, 4 },
-}), );
+}));
 
 class scatter_update_scale_activation_eltwise : public ScatterUpdatePrimitiveFusingTest {};
 TEST_P(scatter_update_scale_activation_eltwise, basic) {
@@ -6046,7 +6046,7 @@ TEST_P(scatter_update_scale_activation_eltwise, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_update_scale_activation_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, scatter_update_scale_activation_eltwise,
                         ::testing::ValuesIn(std::vector<scatter_update_test_params> {
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_FP32_1, 3, 5 },
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_FP32_2, 3, 5 },
@@ -6073,7 +6073,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_update_scale_activation_eltwise,
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_5D_FP16_4, 3, 5 },
                         scatter_update_test_params{ CASE_SCATTER_UPDATE_5D_FP16_5, 3, 5 },
 
-}), );
+}));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ------------------------------------------ ScatterElementsUpdate cases --------------------------------------------- */
@@ -6176,7 +6176,7 @@ TEST_P(scatter_elements_update_quantize, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_elements_update_quantize,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, scatter_elements_update_quantize,
     ::testing::ValuesIn(std::vector<scatter_elements_update_test_params>{
                         scatter_elements_update_test_params{ CASE_SCATTER_ELEMENTS_UPDATE_FP32_1, 2, 3 },
                         scatter_elements_update_test_params{ CASE_SCATTER_ELEMENTS_UPDATE_FP32_2, 2, 3 },
@@ -6191,7 +6191,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_elements_update_quantize,
                         scatter_elements_update_test_params{ CASE_SCATTER_ELEMENTS_UPDATE_5D_FP16_1, 2, 3 },
                         scatter_elements_update_test_params{ CASE_SCATTER_ELEMENTS_UPDATE_5D_FP16_2, 2, 3 },
                         scatter_elements_update_test_params{ CASE_SCATTER_ELEMENTS_UPDATE_5D_FP16_3, 2, 3 },
-}), );
+}));
 
 class scatter_elements_update_scale_activation_eltwise : public ScatterElementsUpdatePrimitiveFusingTest {};
 TEST_P(scatter_elements_update_scale_activation_eltwise, basic) {
@@ -6211,7 +6211,7 @@ TEST_P(scatter_elements_update_scale_activation_eltwise, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_elements_update_scale_activation_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, scatter_elements_update_scale_activation_eltwise,
     ::testing::ValuesIn(std::vector<scatter_elements_update_test_params>{
                         scatter_elements_update_test_params{ CASE_SCATTER_ELEMENTS_UPDATE_FP32_1, 2, 5 },
                         scatter_elements_update_test_params{ CASE_SCATTER_ELEMENTS_UPDATE_FP32_2, 2, 5 },
@@ -6226,7 +6226,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_elements_update_scale_activation_el
                         scatter_elements_update_test_params{ CASE_SCATTER_ELEMENTS_UPDATE_5D_FP16_1, 2, 5 },
                         scatter_elements_update_test_params{ CASE_SCATTER_ELEMENTS_UPDATE_5D_FP16_2, 2, 5 },
                         scatter_elements_update_test_params{ CASE_SCATTER_ELEMENTS_UPDATE_5D_FP16_3, 2, 5 },
-}), );
+}));
 
 /* ------------------------------------------------------------------------------------------------------------ */
 /* ---------------------------------------- PERMUTE FUSE cases -------------------------------------------------- */
@@ -6357,7 +6357,7 @@ TEST_P(permute_activation_scale_eltwise, basic) {
         execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_activation_scale_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, permute_activation_scale_eltwise,
                         ::testing::ValuesIn(std::vector<permute_params> {
                             permute_params{CASE_PERMUTE_F32_0, 2, 5},
                             permute_params{CASE_PERMUTE_F32_1, 2, 5},
@@ -6431,7 +6431,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_activation_scale_eltwise,
                             permute_params{CASE_PERMUTE_TILE_8x8_4x4_FSV16_5D_1, 2, 5},
                             permute_params{CASE_PERMUTE_TILE_8x8_4x4_FSV16_5D_2, 2, 5},
                             permute_params{CASE_PERMUTE_TILE_8x8_4x4_FSV16_5D_3, 2, 5},
-                        }), );
+                        }));
 
 class permute_quant_u8: public PermuteFusingTest {};
 TEST_P(permute_quant_u8, basic) {
@@ -6451,14 +6451,14 @@ TEST_P(permute_quant_u8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_quant_u8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, permute_quant_u8,
                         ::testing::ValuesIn(std::vector<permute_params> {
                             permute_params{CASE_PERMUTE_F32_0, 2, 3},
                             permute_params{CASE_PERMUTE_F32_1, 2, 3},
 
                             permute_params{CASE_PERMUTE_F16_0, 2, 3},
                             permute_params{CASE_PERMUTE_F16_1, 2, 3},
-                        }), );
+                        }));
 
 class permute_scale_actv_eltw_scale_actv_quant_i8: public PermuteFusingTest {};
 TEST_P(permute_scale_actv_eltw_scale_actv_quant_i8, basic) {
@@ -6486,7 +6486,7 @@ TEST_P(permute_scale_actv_eltw_scale_actv_quant_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_scale_actv_eltw_scale_actv_quant_i8,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, permute_scale_actv_eltw_scale_actv_quant_i8,
                         ::testing::ValuesIn(std::vector<permute_params> {
                             permute_params{CASE_PERMUTE_F32_0, 2, 8},
                             permute_params{CASE_PERMUTE_F32_1, 2, 8},
@@ -6536,7 +6536,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_scale_actv_eltw_scale_actv_quant_i8
                             permute_params{CASE_PERMUTE_U8_3D_1, 2, 8},
                             permute_params{CASE_PERMUTE_U8_3D_2, 2, 8},
                             permute_params{CASE_PERMUTE_U8_3D_3, 2, 8},
-                        }), );
+                        }));
 
 class permute_scale_eltwise_actv_scale_actv: public PermuteFusingTest {};
 TEST_P(permute_scale_eltwise_actv_scale_actv, basic) {
@@ -6560,7 +6560,7 @@ TEST_P(permute_scale_eltwise_actv_scale_actv, basic) {
         execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_scale_eltwise_actv_scale_actv,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, permute_scale_eltwise_actv_scale_actv,
                         ::testing::ValuesIn(std::vector<permute_params> {
                             permute_params{CASE_PERMUTE_F32_0, 2, 7},
                             permute_params{CASE_PERMUTE_F32_1, 2, 7},
@@ -6634,7 +6634,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_scale_eltwise_actv_scale_actv,
                             permute_params{CASE_PERMUTE_TILE_8x8_4x4_FSV16_5D_1, 2, 7},
                             permute_params{CASE_PERMUTE_TILE_8x8_4x4_FSV16_5D_2, 2, 7},
                             permute_params{CASE_PERMUTE_TILE_8x8_4x4_FSV16_5D_3, 2, 7},
-                        }), );
+                        }));
 
 /* ------------------------------------------------------------------------------------------------------------ */
 /* ---------------------------- PERMUTE FUSE REDUNDANT REORDER cases ------------------------------------------ */
@@ -6727,7 +6727,7 @@ TEST_P(permute_redundant_reorder, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_redundant_reorder,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, permute_redundant_reorder,
                             ::testing::ValuesIn(std::vector<permute_reorder_params> {
                                 permute_reorder_params{CASE_PERMUTE_REORDER_F32_0, 3, 4},
                                 permute_reorder_params{CASE_PERMUTE_REORDER_F32_1, 3, 4},
@@ -6759,7 +6759,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_redundant_reorder,
                                 permute_reorder_params{CASE_PERMUTE_REORDER_TILED_F16_9, 3, 4},
                                 permute_reorder_params{CASE_PERMUTE_REORDER_TILED_F16_10, 3, 4},
                                 permute_reorder_params{CASE_PERMUTE_REORDER_TILED_F16_11, 3, 4},
-                                }),);
+                                }));
 
 class permute_act_reorder : public PermuteReorderFusingTest {};
 
@@ -6776,7 +6776,7 @@ TEST_P(permute_act_reorder, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_act_reorder,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, permute_act_reorder,
                             ::testing::ValuesIn(std::vector<permute_reorder_params> {
                                 permute_reorder_params{CASE_PERMUTE_REORDER_F32_0, 3, 5},
                                 permute_reorder_params{CASE_PERMUTE_REORDER_F32_1, 3, 5},
@@ -6802,7 +6802,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, permute_act_reorder,
                                 permute_reorder_params{CASE_PERMUTE_REORDER_TILED_F16_9, 3, 5},
                                 permute_reorder_params{CASE_PERMUTE_REORDER_TILED_F16_10, 3, 5},
                                 permute_reorder_params{CASE_PERMUTE_REORDER_TILED_F16_11, 3, 5},
-                                }),);
+                                }));
 
 class NormalizeFusingTest : public ::BaseFusingTest<normalize_test_params> {
 public:
@@ -6840,12 +6840,12 @@ TEST_P(normalize_i8_quantize, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         normalize_i8_quantize,
                         ::testing::ValuesIn(std::vector<normalize_test_params>{
                             normalize_test_params{CASE_NORMALIZE_I8_1, false, 2, 3},
                             normalize_test_params{CASE_NORMALIZE_I8_1, true, 2, 3},
-                        }), );
+                        }));
 
 class normalize_i8_float : public NormalizeFusingTest {};
 TEST_P(normalize_i8_float, basic) {
@@ -6863,12 +6863,12 @@ TEST_P(normalize_i8_float, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         normalize_i8_float,
                         ::testing::ValuesIn(std::vector<normalize_test_params>{
                             normalize_test_params{CASE_NORMALIZE_I8_1, false, 2, 4},
                             normalize_test_params{CASE_NORMALIZE_I8_1, true, 2, 4},
-                        }), );
+                        }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ---------------------------------------- BatchToSpace cases ----------------------------------------- */
@@ -6935,7 +6935,7 @@ TEST_P(batch_to_space_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     batch_to_space_quantize_i8,
     ::testing::ValuesIn(std::vector<batch_to_space_test_params> {
@@ -6943,7 +6943,7 @@ INSTANTIATE_TEST_CASE_P(
         batch_to_space_test_params{CASE_BATCH_TO_SPACE_F32_2, 2, 3},
         batch_to_space_test_params{CASE_BATCH_TO_SPACE_F16_1, 2, 3},
         batch_to_space_test_params{CASE_BATCH_TO_SPACE_F16_2, 2, 3},
-    }), );
+    }));
 
 class batch_to_space_scale_act_eltwise_quantize_u8 : public BatchToSpaceFusingsTest {};
 TEST_P(batch_to_space_scale_act_eltwise_quantize_u8, basic) {
@@ -6966,7 +6966,7 @@ TEST_P(batch_to_space_scale_act_eltwise_quantize_u8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     batch_to_space_scale_act_eltwise_quantize_u8,
     ::testing::ValuesIn(std::vector<batch_to_space_test_params> {
@@ -6978,7 +6978,7 @@ INSTANTIATE_TEST_CASE_P(
         batch_to_space_test_params{CASE_BATCH_TO_SPACE_U8_2, 2, 6},
         batch_to_space_test_params{CASE_BATCH_TO_SPACE_I8_1, 2, 6},
         batch_to_space_test_params{CASE_BATCH_TO_SPACE_I8_2, 2, 6},
-    }), );
+    }));
 
 class batch_to_space_scale_act_eltw : public BatchToSpaceFusingsTest {};
 TEST_P(batch_to_space_scale_act_eltw, basic) {
@@ -6996,7 +6996,7 @@ TEST_P(batch_to_space_scale_act_eltw, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     batch_to_space_scale_act_eltw,
     ::testing::ValuesIn(std::vector<batch_to_space_test_params> {
@@ -7008,7 +7008,7 @@ INSTANTIATE_TEST_CASE_P(
         batch_to_space_test_params{CASE_BATCH_TO_SPACE_U8_2, 2, 5},
         batch_to_space_test_params{CASE_BATCH_TO_SPACE_I8_1, 2, 5},
         batch_to_space_test_params{CASE_BATCH_TO_SPACE_I8_2, 2, 5},
-    }), );
+    }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ---------------------------------------- SpaceToBatch cases ----------------------------------------- */
@@ -7076,7 +7076,7 @@ TEST_P(space_to_batch_quantize_i8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     space_to_batch_quantize_i8,
     ::testing::ValuesIn(std::vector<space_to_batch_test_params> {
@@ -7084,7 +7084,7 @@ INSTANTIATE_TEST_CASE_P(
         space_to_batch_test_params{CASE_SPACE_TO_BATCH_F32_2, 2, 3},
         space_to_batch_test_params{CASE_SPACE_TO_BATCH_F16_1, 2, 3},
         space_to_batch_test_params{CASE_SPACE_TO_BATCH_F16_2, 2, 3},
-    }), );
+    }));
 
 class space_to_batch_scale_act_eltwise_quantize_u8 : public SpaceToBatchFusingsTest {};
 TEST_P(space_to_batch_scale_act_eltwise_quantize_u8, basic) {
@@ -7107,7 +7107,7 @@ TEST_P(space_to_batch_scale_act_eltwise_quantize_u8, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     space_to_batch_scale_act_eltwise_quantize_u8,
     ::testing::ValuesIn(std::vector<space_to_batch_test_params> {
@@ -7119,7 +7119,7 @@ INSTANTIATE_TEST_CASE_P(
         space_to_batch_test_params{CASE_SPACE_TO_BATCH_U8_2, 2, 6},
         space_to_batch_test_params{CASE_SPACE_TO_BATCH_I8_1, 2, 6},
         space_to_batch_test_params{CASE_SPACE_TO_BATCH_I8_2, 2, 6},
-    }), );
+    }));
 
 
 class space_to_batch_scale_act_eltw : public SpaceToBatchFusingsTest {};
@@ -7138,7 +7138,7 @@ TEST_P(space_to_batch_scale_act_eltw, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     fusings_gpu,
     space_to_batch_scale_act_eltw,
     ::testing::ValuesIn(std::vector<space_to_batch_test_params> {
@@ -7150,7 +7150,7 @@ INSTANTIATE_TEST_CASE_P(
         space_to_batch_test_params{CASE_SPACE_TO_BATCH_U8_2, 2, 5},
         space_to_batch_test_params{CASE_SPACE_TO_BATCH_I8_1, 2, 5},
         space_to_batch_test_params{CASE_SPACE_TO_BATCH_I8_2, 2, 5},
-    }), );
+    }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ---------------------------------------- Eltwise cases -------------------------------------------------- */
@@ -7268,7 +7268,7 @@ TEST_P(eltwise_quantize, i8_per_channel) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         eltwise_quantize,
                         ::testing::ValuesIn(std::vector<eltwise_test_params>{
                             eltwise_test_params{CASE_ELTWISE_FP16_1, 3, 4},
@@ -7301,7 +7301,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             eltwise_test_params{CASE_ELTWISE_FP32_6, 3, 4},
                             eltwise_test_params{CASE_ELTWISE_I8_4, 3, 4},
                             eltwise_test_params{CASE_ELTWISE_U8_4, 3, 4},
-                        }), );
+                        }));
 
 class eltwise_const_path : public EltwiseFusingTest {};
 TEST_P(eltwise_const_path, not_fuse_to_const_eltwise) {
@@ -7318,7 +7318,7 @@ TEST_P(eltwise_const_path, not_fuse_to_const_eltwise) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         eltwise_const_path,
                         ::testing::ValuesIn(std::vector<eltwise_test_params>{
                             eltwise_test_params{CASE_ELTWISE_FP16_3, 2, 3},
@@ -7327,7 +7327,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             eltwise_test_params{CASE_ELTWISE_FP32_6, 2, 3},
                             eltwise_test_params{CASE_ELTWISE_I8_4, 2, 3},
                             eltwise_test_params{CASE_ELTWISE_U8_4, 2, 3},
-                        }), );
+                        }));
 
 class eltwise_fp32_fsv16 : public EltwiseFusingTest {};
 TEST_P(eltwise_fp32_fsv16, add) {
@@ -7364,12 +7364,12 @@ TEST_P(eltwise_fp32_fsv16, add_per_element) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         eltwise_fp32_fsv16,
                         ::testing::ValuesIn(std::vector<eltwise_test_params>{
                             eltwise_test_params{CASE_ELTWISE_FP16_3, 3, 5},
                             eltwise_test_params{CASE_ELTWISE_FP32_3, 3, 5},
-                        }), );
+                        }));
 
 class eltwise_fp32_fsv32 : public EltwiseFusingTest {};
 TEST_P(eltwise_fp32_fsv32, add) {
@@ -7406,13 +7406,13 @@ TEST_P(eltwise_fp32_fsv32, add_per_element) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         eltwise_fp32_fsv32,
                         ::testing::ValuesIn(std::vector<eltwise_test_params>{
                             // There's no optimized eltwise kernel yet for fsv32 layout that supports fused_ops
                             // So only activation is fused via legacy mechanism
                             eltwise_test_params{CASE_ELTWISE_FP16_4, 4, 5},
-                        }), );
+                        }));
 
 class eltwise_fp32_fsv4 : public EltwiseFusingTest {};
 TEST_P(eltwise_fp32_fsv4, add) {
@@ -7449,14 +7449,14 @@ TEST_P(eltwise_fp32_fsv4, add_per_element) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         eltwise_fp32_fsv4,
                         ::testing::ValuesIn(std::vector<eltwise_test_params>{
                             eltwise_test_params{CASE_ELTWISE_FP32_5, 3, 5},
                             eltwise_test_params{CASE_ELTWISE_FP32_6, 3, 5},
                             eltwise_test_params{CASE_ELTWISE_I8_4,   3, 5},
                             eltwise_test_params{CASE_ELTWISE_U8_4,   3, 5},
-                        }), );
+                        }));
 
 //
 class eltwise_fp32_fused_prims : public EltwiseFusingTest {};
@@ -7502,7 +7502,7 @@ TEST_P(eltwise_fp32_fused_prims, eltwise_activation_with_broadcast) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         eltwise_fp32_fused_prims,
                         ::testing::ValuesIn(std::vector<eltwise_test_params>{
                             eltwise_test_params{CASE_ELTWISE_FP16_1, 3, 5},
@@ -7534,7 +7534,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             eltwise_test_params{CASE_ELTWISE_FP32_6, 3, 5},
                             eltwise_test_params{CASE_ELTWISE_I8_4, 3, 5},
                             eltwise_test_params{CASE_ELTWISE_U8_4, 3, 5},
-                        }), );
+                        }));
 
 class eltwise_fp32_scale : public EltwiseFusingTest {};
 TEST_P(eltwise_fp32_scale, 6d) {
@@ -7550,11 +7550,11 @@ TEST_P(eltwise_fp32_scale, 6d) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         eltwise_fp32_scale,
                         ::testing::ValuesIn(std::vector<eltwise_test_params>{
                             eltwise_test_params{CASE_ELTWISE_FP32_4, 3, 4},
-                        }), );
+                        }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ---------------------------------------- Scale cases ------------------------------------------------ */
@@ -7660,13 +7660,13 @@ TEST_P(scale_basic, bias_act_quantize) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         scale_basic,
                         ::testing::ValuesIn(std::vector<scale_test_params>{
                             scale_test_params{CASE_SCALE_FP32_1, 2, 4},
                             scale_test_params{CASE_SCALE_FP32_2, 2, 4},
                             scale_test_params{CASE_SCALE_FP32_3, 2, 4},
-                        }), );
+                        }));
 
 class eltwise_no_pitches_same_dims_quantize : public EltwiseFusingTest {};
 TEST_P(eltwise_no_pitches_same_dims_quantize, quantize_f32_output) {
@@ -7685,7 +7685,7 @@ TEST_P(eltwise_no_pitches_same_dims_quantize, quantize_f32_output) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         eltwise_no_pitches_same_dims_quantize,
                         ::testing::ValuesIn(std::vector<eltwise_test_params>{
                             eltwise_test_params{CASE_ELTWISE_FP16_1, 3, 4},
@@ -7694,7 +7694,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             eltwise_test_params{CASE_ELTWISE_FP32_1, 3, 4},
                             eltwise_test_params{CASE_ELTWISE_FP32_2, 3, 4},
                             eltwise_test_params{CASE_ELTWISE_FP32_3, 3, 4},
-                        }), );
+                        }));
 
 class eltwise_activation : public EltwiseFusingTest {};
 TEST_P(eltwise_activation, basic) {
@@ -7721,7 +7721,7 @@ TEST_P(eltwise_activation, fp16_out) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         eltwise_activation,
                         ::testing::ValuesIn(std::vector<eltwise_test_params>{
                             eltwise_test_params{CASE_ELTWISE_FP16_1, 3, 4},
@@ -7736,7 +7736,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             eltwise_test_params{CASE_ELTWISE_FP16_FP32_1, 3, 4},
                             eltwise_test_params{CASE_ELTWISE_FP16_FP32_2, 3, 4},
                             eltwise_test_params{CASE_ELTWISE_FP16_FP32_3, 3, 4}
-                        }), );
+                        }));
 
 /* ----------------------------------------------------------------------------------------------------- */
 /* ---------------------------------------- Reduce cases ----------------------------------------------- */
@@ -7871,7 +7871,7 @@ TEST_P(reduce_eltwise_activation_quantize, per_channel) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         reduce_eltwise_activation_quantize,
                         ::testing::ValuesIn(std::vector<reduce_test_params>{
                             reduce_test_params{CASE_REDUCE_F32_0, 2, 5, reduce_mode::mean, {reduce::along_x, reduce::along_f, reduce::along_y, reduce::along_b}, true, "reduce_gpu_b_fs_yx_fsv16"},
@@ -7941,7 +7941,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             reduce_test_params{CASE_REDUCE_U8_4, 2, 5, reduce_mode::sum, {reduce::along_b}, true, "reduce_gpu_b_fs_yx_fsv16"},
                             reduce_test_params{CASE_REDUCE_U8_0, 2, 5, reduce_mode::max, {reduce::along_f}, true, "reduce_gpu_b_fs_yx_fsv16"},
                             reduce_test_params{CASE_REDUCE_U8_4, 2, 5, reduce_mode::mean, {reduce::along_x}, true, "reduce_gpu_b_fs_yx_fsv16"}
-                        }), );
+                        }));
 
 class reduce_scale_activation : public ReduceFusingTest {};
 TEST_P(reduce_scale_activation, basic) {
@@ -7968,7 +7968,7 @@ TEST_P(reduce_scale_activation, per_channel) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu,
                         reduce_scale_activation,
                         ::testing::ValuesIn(std::vector<reduce_test_params>{
                             reduce_test_params{CASE_REDUCE_F32_0, 2, 4, reduce_mode::max, {reduce::along_x, reduce::along_y, reduce::along_b}, true, "reduce_gpu_b_fs_yx_fsv16"},
@@ -7986,9 +7986,9 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu,
                             reduce_test_params{CASE_REDUCE_F16_2, 2, 4, reduce_mode::mean, {reduce::along_x, reduce::along_y}, true, "reduce_ref"},
                             reduce_test_params{CASE_REDUCE_F16_0, 2, 4, reduce_mode::min, {reduce::along_x}, true, "reduce_gpu_b_fs_yx_fsv16"},
                             reduce_test_params{CASE_REDUCE_F16_0, 2, 4, reduce_mode::sum, {reduce::along_x}, true, "reduce_gpu_b_fs_yx_fsv16"},
-                        }), );
+                        }));
 
-INSTANTIATE_TEST_CASE_P(DISABLED_fusings_gpu,
+INSTANTIATE_TEST_SUITE_P(DISABLED_fusings_gpu,
                         reduce_eltwise_activation_quantize,
                         ::testing::ValuesIn(std::vector<reduce_test_params>{
                             // No layout format available for quantize/scale
@@ -7998,7 +7998,7 @@ INSTANTIATE_TEST_CASE_P(DISABLED_fusings_gpu,
                             reduce_test_params{CASE_REDUCE_I32_3, 2, 4, reduce_mode::sum, {reduce::along_x}, true, "reduce_ref"},
                             reduce_test_params{CASE_REDUCE_I8_3, 2, 4, reduce_mode::mean, {reduce::along_x}, true, "reduce_ref"},
                             reduce_test_params{CASE_REDUCE_U8_3, 2, 4, reduce_mode::l2, {reduce::along_x}, true, "reduce_ref"}
-                        }), );
+                        }));
 
 
 /* ----------------------------------------------------------------------------------------------------- */
@@ -8108,7 +8108,7 @@ TEST_P(scatter_nd_update_quantize, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_nd_update_quantize,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, scatter_nd_update_quantize,
     ::testing::ValuesIn(std::vector<scatter_nd_update_test_params>{
         scatter_nd_update_test_params{ CASE_SCATTER_ND_UPDATE_FP16_4D_1, 2, 3 },
         scatter_nd_update_test_params{ CASE_SCATTER_ND_UPDATE_FP16_4D_2, 2, 3 },
@@ -8151,7 +8151,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_nd_update_quantize,
         scatter_nd_update_test_params{ CASE_SCATTER_ND_UPDATE_FP32_6D_4, 2, 3 },
         scatter_nd_update_test_params{ CASE_SCATTER_ND_UPDATE_FP32_6D_5, 2, 3 },
         scatter_nd_update_test_params{ CASE_SCATTER_ND_UPDATE_FP32_6D_6, 2, 3 },
-}), );
+}));
 
 class scatter_nd_update_scale_activation_eltwise : public ScatterNDUpdatePrimitiveFusingTest {};
 TEST_P(scatter_nd_update_scale_activation_eltwise, basic) {
@@ -8172,7 +8172,7 @@ TEST_P(scatter_nd_update_scale_activation_eltwise, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_nd_update_scale_activation_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, scatter_nd_update_scale_activation_eltwise,
     ::testing::ValuesIn(std::vector<scatter_nd_update_test_params>{
         scatter_nd_update_test_params{ CASE_SCATTER_ND_UPDATE_FP16_4D_1, 2, 5 },
         scatter_nd_update_test_params{ CASE_SCATTER_ND_UPDATE_FP16_4D_2, 2, 5 },
@@ -8217,7 +8217,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, scatter_nd_update_scale_activation_eltwise,
         scatter_nd_update_test_params{ CASE_SCATTER_ND_UPDATE_FP32_6D_4, 2, 5 },
         scatter_nd_update_test_params{ CASE_SCATTER_ND_UPDATE_FP32_6D_5, 2, 5 },
         scatter_nd_update_test_params{ CASE_SCATTER_ND_UPDATE_FP32_6D_6, 2, 5 },
-}), );
+}));
 
 
 /* ----------------------------------------------------------------------------------------------------- */
@@ -8325,7 +8325,7 @@ TEST_P(gather_nd_quantize, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, gather_nd_quantize,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, gather_nd_quantize,
     ::testing::ValuesIn(std::vector<gather_nd_test_params>{
         gather_nd_test_params{ CASE_GATHER_ND_FP16_4D_1, 2, 3 },
         gather_nd_test_params{ CASE_GATHER_ND_FP16_4D_2, 2, 3 },
@@ -8358,7 +8358,7 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, gather_nd_quantize,
         gather_nd_test_params{ CASE_GATHER_ND_FP32_6D_2, 2, 3 },
         gather_nd_test_params{ CASE_GATHER_ND_FP32_6D_3, 2, 3 },
         gather_nd_test_params{ CASE_GATHER_ND_FP32_6D_4, 2, 3 },
-}), );
+}));
 
 class gather_nd_activation_scale_eltwise : public GatherNDPrimitiveFusingTest {};
 TEST_P(gather_nd_activation_scale_eltwise, basic) {
@@ -8379,7 +8379,7 @@ TEST_P(gather_nd_activation_scale_eltwise, basic) {
     execute(p);
 }
 
-INSTANTIATE_TEST_CASE_P(fusings_gpu, gather_nd_activation_scale_eltwise,
+INSTANTIATE_TEST_SUITE_P(fusings_gpu, gather_nd_activation_scale_eltwise,
     ::testing::ValuesIn(std::vector<gather_nd_test_params>{
         gather_nd_test_params{ CASE_GATHER_ND_FP16_4D_1, 2, 5 },
         gather_nd_test_params{ CASE_GATHER_ND_FP16_4D_2, 2, 5 },
@@ -8412,4 +8412,4 @@ INSTANTIATE_TEST_CASE_P(fusings_gpu, gather_nd_activation_scale_eltwise,
         gather_nd_test_params{ CASE_GATHER_ND_FP32_6D_2, 2, 5 },
         gather_nd_test_params{ CASE_GATHER_ND_FP32_6D_3, 2, 5 },
         gather_nd_test_params{ CASE_GATHER_ND_FP32_6D_4, 2, 5 },
-}), );
+}));
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/gemm_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/gemm_gpu_test.cpp
index ae1ca356924dcf..cb02adf490c736 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/gemm_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/gemm_gpu_test.cpp
@@ -3537,27 +3537,27 @@ class GemmBaseTest : public ::testing::TestWithParam<gemm_params> {
 class gemm_int8_transposition_tests : public ::GemmBaseTest<gemm_base_test_params, int8_t, uint8_t, float, float, int32_t> {};
 TEST_P(gemm_int8_transposition_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_int8_transposition_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_int8_transposition_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_INT8_NN_TRANSPOSITION, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_NT_TRANSPOSITION, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_TN_TRANSPOSITION, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_TT_TRANSPOSITION, "gemm_mmad_int8" },
-}), );
+}));
 
 class gemm_int8_broadcast_tests : public ::GemmBaseTest<gemm_base_test_params, int8_t, uint8_t, float, float, int32_t> {};
 TEST_P(gemm_int8_broadcast_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_int8_broadcast_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_int8_broadcast_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_INT8_BROADCAST_1, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_BROADCAST_2, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_BROADCAST_3, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_BROADCAST_4, "gemm_mmad_int8" },
-}), );
+}));
 
 class gemm_int8_leftovers_tests : public ::GemmBaseTest<gemm_base_test_params, int8_t, uint8_t, float, float, int32_t> {};
 TEST_P(gemm_int8_leftovers_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_int8_leftovers_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_int8_leftovers_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_INT8_LEFTOVERS_1, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_LEFTOVERS_2, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_LEFTOVERS_3, "gemm_mmad_int8" },
@@ -3570,124 +3570,124 @@ INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_int8_leftovers_tests, ::testing::ValuesIn
                         gemm_base_test_params{ CASE_GEMM_INT8_LEFTOVERS_10, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_LEFTOVERS_11, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_LEFTOVERS_12, "gemm_mmad_int8" },
-}), );
+}));
 
 class gemm_int8_combo_tests : public ::GemmBaseTest<gemm_base_test_params, int8_t, uint8_t, float, float, int32_t> {};
 TEST_P(gemm_int8_combo_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_int8_combo_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_int8_combo_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_INT8_COMBO_1, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_COMBO_2, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_COMBO_3, "gemm_mmad_int8" },
                         gemm_base_test_params{ CASE_GEMM_INT8_COMBO_4, "gemm_mmad_int8" },
-}), );
+}));
 
 class gemm_int8_slm_combo_tests : public ::GemmBaseTest<gemm_base_test_params, int8_t, uint8_t, float, float, int32_t> {};
 TEST_P(gemm_int8_slm_combo_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_int8_slm_combo_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_int8_slm_combo_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_INT8_SLM_COMBO_1, "gemm_mmad_int8_slm" },
                         gemm_base_test_params{ CASE_GEMM_INT8_SLM_COMBO_2, "gemm_mmad_int8_slm" },
                         gemm_base_test_params{ CASE_GEMM_INT8_SLM_COMBO_3, "gemm_mmad_int8_slm" },
                         gemm_base_test_params{ CASE_GEMM_INT8_SLM_COMBO_4, "gemm_mmad_int8_slm" },
-}), );
+}));
 
 class gemm_fp32_tiled_nn_tests : public ::GemmBaseTest<gemm_base_test_params, float, float, float, float, float> {};
 TEST_P(gemm_fp32_tiled_nn_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_fp32_tiled_nn_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_fp32_tiled_nn_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NN_1, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NN_2, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NN_3, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NN_4, "gemm_tiled_opt" },
-}), );
+}));
 
 class gemm_fp32_tiled_nt_tests : public ::GemmBaseTest<gemm_base_test_params, float, float, float, float, float> {};
 TEST_P(gemm_fp32_tiled_nt_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_fp32_tiled_nt_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_fp32_tiled_nt_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NT_1, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NT_2, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NT_3, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NT_4, "gemm_tiled_opt" },
-}), );
+}));
 
 class gemm_fp32_tiled_tn_tests : public ::GemmBaseTest<gemm_base_test_params, float, float, float, float, float> {};
 TEST_P(gemm_fp32_tiled_tn_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_fp32_tiled_tn_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_fp32_tiled_tn_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_TN_1, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_TN_2, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_TN_3, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_TN_4, "gemm_tiled_opt" },
-}), );
+}));
 
 class gemm_fp32_tiled_tt_tests : public ::GemmBaseTest<gemm_base_test_params, float, float, float, float, float> {};
 TEST_P(gemm_fp32_tiled_tt_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_fp32_tiled_tt_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_fp32_tiled_tt_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_TT_1, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_TT_2, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_TT_3, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_TT_4, "gemm_tiled_opt" },
-}), );
+}));
 
 class gemm_fp32_tiled_nn_broadcast_tests : public ::GemmBaseTest<gemm_base_test_params, float, float, float, float, float> {};
 TEST_P(gemm_fp32_tiled_nn_broadcast_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_fp32_tiled_nn_broadcast_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_fp32_tiled_nn_broadcast_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NN_BROADCAST_1, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NN_BROADCAST_2, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NN_BROADCAST_3, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP32_TILED_NN_BROADCAST_4, "gemm_tiled_opt" },
-}), );
+}));
 
 class gemm_fp16_tiled_nn_tests : public ::GemmBaseTest<gemm_base_test_params, FLOAT16, FLOAT16, FLOAT16, FLOAT16, FLOAT16> {};
 TEST_P(gemm_fp16_tiled_nn_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_fp16_tiled_nn_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_fp16_tiled_nn_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NN_1, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NN_2, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NN_3, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NN_4, "gemm_tiled_opt" },
-}), );
+}));
 
 class gemm_fp16_tiled_nt_tests : public ::GemmBaseTest<gemm_base_test_params, FLOAT16, FLOAT16, FLOAT16, FLOAT16, FLOAT16> {};
 TEST_P(gemm_fp16_tiled_nt_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_fp16_tiled_nt_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_fp16_tiled_nt_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NT_1, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NT_2, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NT_3, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NT_4, "gemm_tiled_opt" },
-}), );
+}));
 
 class gemm_fp16_tiled_tn_tests : public ::GemmBaseTest<gemm_base_test_params, FLOAT16, FLOAT16, FLOAT16, FLOAT16, FLOAT16> {};
 TEST_P(gemm_fp16_tiled_tn_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_fp16_tiled_tn_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_fp16_tiled_tn_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_TN_1, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_TN_2, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_TN_3, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_TN_4, "gemm_tiled_opt" },
-}), );
+}));
 
 class gemm_fp16_tiled_tt_tests : public ::GemmBaseTest<gemm_base_test_params, FLOAT16, FLOAT16, FLOAT16, FLOAT16, FLOAT16> {};
 TEST_P(gemm_fp16_tiled_tt_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_fp16_tiled_tt_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_fp16_tiled_tt_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_TT_1, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_TT_2, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_TT_3, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_TT_4, "gemm_tiled_opt" },
-}), );
+}));
 
 class gemm_fp16_tiled_nn_broadcast_tests : public ::GemmBaseTest<gemm_base_test_params, FLOAT16, FLOAT16, FLOAT16, FLOAT16, FLOAT16> {};
 TEST_P(gemm_fp16_tiled_nn_broadcast_tests, basic) { auto p = GetParam(); execute(p); }
 
-INSTANTIATE_TEST_CASE_P(gemm_gpu, gemm_fp16_tiled_nn_broadcast_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
+INSTANTIATE_TEST_SUITE_P(gemm_gpu, gemm_fp16_tiled_nn_broadcast_tests, ::testing::ValuesIn(std::vector <gemm_base_test_params> {
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NN_BROADCAST_1, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NN_BROADCAST_2, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NN_BROADCAST_3, "gemm_tiled_opt" },
                         gemm_base_test_params{ CASE_GEMM_FP16_TILED_NN_BROADCAST_4, "gemm_tiled_opt" },
-}), );
+}));
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_dynamic_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_dynamic_gpu_test.cpp
index c8a037bfbf157e..b25f483384f7bb 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_dynamic_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/lstm_dynamic_gpu_test.cpp
@@ -527,8 +527,8 @@ struct lstm_dynamic_single_layer_test : public ::testing::Test
 
 };
 typedef ::testing::Types<float, FLOAT16> lstm_dynamic_test_types;
-TYPED_TEST_CASE(lstm_dynamic_single_layer_test, lstm_dynamic_test_types);
-TYPED_TEST_CASE(lstm_dynamic_input_layer_test, lstm_dynamic_test_types);
+TYPED_TEST_SUITE(lstm_dynamic_single_layer_test, lstm_dynamic_test_types);
+TYPED_TEST_SUITE(lstm_dynamic_input_layer_test, lstm_dynamic_test_types);
 
 /*
 ----------------------------------------------
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/mvn_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/mvn_gpu_test.cpp
index 5632a19cfd4ebe..3437ea51f27ba0 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/mvn_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/mvn_gpu_test.cpp
@@ -695,20 +695,20 @@ struct mvn_test_case_generator : std::vector<mvn_basic_test_params> {
     }
 };
 
-INSTANTIATE_TEST_CASE_P(smoke,
+INSTANTIATE_TEST_SUITE_P(smoke,
                         mvn_random_test,
                         testing::ValuesIn(mvn_test_case_generator()
                                               .smoke_tests(format::b_fs_yx_fsv16, data_types::i8)
-                                              .smoke_tests(format::b_fs_yx_fsv16, data_types::u8)), );
+                                              .smoke_tests(format::b_fs_yx_fsv16, data_types::u8)));
 
-INSTANTIATE_TEST_CASE_P(zyx,
+INSTANTIATE_TEST_SUITE_P(zyx,
                         mvn_random_test,
                         testing::ValuesIn(mvn_test_case_generator()
                                               .zyx_tests(format::b_fs_zyx_fsv16, data_types::i8)
-                                              .zyx_tests(format::b_fs_zyx_fsv16, data_types::u8)), );
+                                              .zyx_tests(format::b_fs_zyx_fsv16, data_types::u8)));
 
-INSTANTIATE_TEST_CASE_P(extended,
+INSTANTIATE_TEST_SUITE_P(extended,
                         mvn_random_test,
                         testing::ValuesIn(mvn_test_case_generator()
                                               .extended_tests(format::b_fs_yx_fsv16, data_types::i8)
-                                              .extended_tests(format::b_fs_yx_fsv16, data_types::u8)), );
+                                              .extended_tests(format::b_fs_yx_fsv16, data_types::u8)));
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/non_max_suppression_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/non_max_suppression_test.cpp
index 7ca881a7b2747b..e192b0e20fd826 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/non_max_suppression_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/non_max_suppression_test.cpp
@@ -86,7 +86,7 @@ struct non_max_suppression_basic : public testing::Test {
 };
 
 using nms_types = testing::Types<float, half_t>;
-TYPED_TEST_CASE(non_max_suppression_basic, nms_types);
+TYPED_TEST_SUITE(non_max_suppression_basic, nms_types);
 
 TYPED_TEST(non_max_suppression_basic, basic) {
     auto& engine = tests::get_test_engine();
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/permute_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/permute_gpu_test.cpp
index 7fb709d9845984..5c5aadfc54b856 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/permute_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/permute_gpu_test.cpp
@@ -1718,7 +1718,7 @@ void TiledPermuteTest::run_test(const std::vector<cldnn::tensor::value_type>& si
 
 class permute_tile_fsv_4d: public TiledPermuteTest {};
 
-INSTANTIATE_TEST_CASE_P(, permute_tile_fsv_4d,
+INSTANTIATE_TEST_SUITE_P(, permute_tile_fsv_4d,
     ::testing::ValuesIn(std::vector<TiledPermuteParam> {
         // b_fs_yx_fsv16
         // normal cases
@@ -1762,7 +1762,7 @@ INSTANTIATE_TEST_CASE_P(, permute_tile_fsv_4d,
         // fy_not_aligned
         {{1, 4 - 3, 4 - 1, 2}, format::b_fs_yx_fsv4},
         {{1, 4 - 1, 4 - 3, 2}, format::b_fs_yx_fsv4},
-    }),);
+    }));
 
 TEST_P(permute_tile_fsv_4d, f16) {
     auto p = GetParam();
@@ -1791,7 +1791,7 @@ TEST_P(permute_tile_fsv_4d, i64) {
 
 class permute_tile_fsv_5d: public TiledPermuteTest {};
 
-INSTANTIATE_TEST_CASE_P(, permute_tile_fsv_5d,
+INSTANTIATE_TEST_SUITE_P(, permute_tile_fsv_5d,
     ::testing::ValuesIn(std::vector<TiledPermuteParam> {
         // b_fs_zyx_fsv16
         // normal cases
@@ -1822,7 +1822,7 @@ INSTANTIATE_TEST_CASE_P(, permute_tile_fsv_5d,
         {{1, 32 - 1, 32 - 7, 2, 2}, format::b_fs_zyx_fsv32},
         {{1, 32 - 7, 32 - 9, 2, 2}, format::b_fs_zyx_fsv32},
         {{1, 32 - 9, 32 - 15, 2, 2}, format::b_fs_zyx_fsv32},
-    }),);
+    }));
 
 TEST_P(permute_tile_fsv_5d, f16) {
     auto p = GetParam();
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/pooling_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/pooling_gpu_test.cpp
index 009490ac681d44..3169dc5467bbec 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/pooling_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/pooling_gpu_test.cpp
@@ -2715,7 +2715,7 @@ TEST_P(pooling_random_test, avg_u8) {
     ASSERT_NO_FATAL_FAILURE(test_case.run_random(GetParam()));
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_low_precision_2d_spatial,
     pooling_random_test,
     testing::Combine(testing::Values(1, 2),
@@ -2732,7 +2732,7 @@ INSTANTIATE_TEST_CASE_P(
                      testing::Values(false, true)),
                     testing::internal::DefaultParamName<pooling_random_test_params>);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_low_precision_3d_spatial,
     pooling_random_test,
     testing::Combine(testing::Values(1, 2),
@@ -2746,7 +2746,7 @@ INSTANTIATE_TEST_CASE_P(
                      testing::Values(false, true)),
                     testing::internal::DefaultParamName<pooling_random_test_params>);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     batched_low_precision,
     pooling_random_test,
     testing::Combine(
@@ -2831,7 +2831,7 @@ TEST_P(pooling_random_test_fp16_fp32, max_fp32) {
     ASSERT_NO_FATAL_FAILURE(test_case.run_random(GetParam()));
 }
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     smoke_low_precision,
     pooling_random_test_fp16_fp32,
     testing::Combine(testing::Values(1, 2),
@@ -3760,7 +3760,7 @@ TEST_P(pooling_test, POOLING)
     run_single_test();
 }
 
-INSTANTIATE_TEST_CASE_P(DISABLED_POOLING,
+INSTANTIATE_TEST_SUITE_P(DISABLED_POOLING,
                         pooling_test,
                         ::testing::Combine(::testing::ValuesIn(pooling_test::generate_generic_test_params()),
                                            ::testing::ValuesIn(pooling_test::generate_specific_test_params())),
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/pyramid_roi_align_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/pyramid_roi_align_gpu_test.cpp
index a4e168ee9a018c..8fb70a6b4ba300 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/pyramid_roi_align_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/pyramid_roi_align_gpu_test.cpp
@@ -20,7 +20,7 @@ struct pyramid_roi_align_typed_test : testing::Test {
 };
 using pyramid_roi_align_types = testing::Types<float, half_t>;
 
-TYPED_TEST_CASE(pyramid_roi_align_typed_test, pyramid_roi_align_types);
+TYPED_TEST_SUITE(pyramid_roi_align_typed_test, pyramid_roi_align_types);
 
 TYPED_TEST(pyramid_roi_align_typed_test, smoke_4levels) {
     using Type = typename pyramid_roi_align_typed_test<TypeParam>::Type;
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/reduce_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/reduce_gpu_test.cpp
index 77a3e06d40c1dc..f443eefba583f1 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/reduce_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/reduce_gpu_test.cpp
@@ -559,7 +559,7 @@ TEST_P(general_reduce_gpu_i8_f32, base) { execute(); }
 class general_reduce_gpu_f32_f32 : public ReduceTestBase<data_types::f32, data_types::f32> {};
 TEST_P(general_reduce_gpu_f32_f32, base) { execute(); }
 
-INSTANTIATE_TEST_CASE_P(reduce_gpu_b_fs_yx_fsv16_i8_i8,
+INSTANTIATE_TEST_SUITE_P(reduce_gpu_b_fs_yx_fsv16_i8_i8,
                         general_reduce_gpu_i8_i8,
                         ::testing::Values(
                             TestParamType_general_reduce_gpu(2, 12, 1, 1, 3, 2, format::b_fs_yx_fsv16, reduce_mode::logical_or, {reduce::along_y, reduce::along_x}, "reduce_gpu_b_fs_yx_fsv16", false, data_types::i8, true, data_types::i8),
@@ -571,7 +571,7 @@ INSTANTIATE_TEST_CASE_P(reduce_gpu_b_fs_yx_fsv16_i8_i8,
                             ),
                             general_reduce_gpu::PrintToStringParamName);
 
- INSTANTIATE_TEST_CASE_P(reduce_gpu_b_fs_yx_fsv16_i8_f32,
+ INSTANTIATE_TEST_SUITE_P(reduce_gpu_b_fs_yx_fsv16_i8_f32,
                         general_reduce_gpu_i8_f32,
                         ::testing::Values(
                             TestParamType_general_reduce_gpu(3, 3, 1, 1, 3, 2, format::b_fs_yx_fsv16, reduce_mode::sum, {reduce::along_x, reduce::along_y, reduce::along_f, reduce::along_b}, "reduce_gpu_b_fs_yx_fsv16", false, data_types::i8, false, data_types::f32),
@@ -611,7 +611,7 @@ INSTANTIATE_TEST_CASE_P(reduce_gpu_b_fs_yx_fsv16_i8_i8,
                             ),
                             general_reduce_gpu::PrintToStringParamName);
 
- INSTANTIATE_TEST_CASE_P(reduce_gpu_b_fs_yx_fsv16_f32_f32,
+ INSTANTIATE_TEST_SUITE_P(reduce_gpu_b_fs_yx_fsv16_f32_f32,
                         general_reduce_gpu_f32_f32,
                         ::testing::Values(
                             TestParamType_general_reduce_gpu(7, 3, 1, 1, 13, 11, format::b_fs_yx_fsv16, reduce_mode::mean, {reduce::along_x, reduce::along_y, reduce::along_f, reduce::along_b}, "reduce_gpu_b_fs_yx_fsv16", false, data_types::f32, false, data_types::f32),
@@ -654,7 +654,7 @@ INSTANTIATE_TEST_CASE_P(reduce_gpu_b_fs_yx_fsv16_i8_i8,
                         ),
                         general_reduce_gpu::PrintToStringParamName);
 
- INSTANTIATE_TEST_CASE_P(reduce_gpu_ref_f32_f32,
+ INSTANTIATE_TEST_SUITE_P(reduce_gpu_ref_f32_f32,
                         general_reduce_gpu_f32_f32,
                         ::testing::Values(
                             TestParamType_general_reduce_gpu(2, 4, 4, 5, 8, 8, format::bfwzyx, reduce_mode::mean, {reduce::along_f, reduce::along_y, reduce::along_w, reduce::along_z}, "reduce_ref", false, data_types::f32, false, data_types::f32),
@@ -710,7 +710,7 @@ INSTANTIATE_TEST_CASE_P(reduce_gpu_b_fs_yx_fsv16_i8_i8,
                             TestParamType_general_reduce_gpu(17, 4, 1, 1, 12, 15, format::bfyx, reduce_mode::mean, {reduce::along_b}, "reduce_ref", true, data_types::f32, false, data_types::f32)
                         ), general_reduce_gpu::PrintToStringParamName);
 
-INSTANTIATE_TEST_CASE_P(DISABLED_reduce_gpu_ref_f32_f32,
+INSTANTIATE_TEST_SUITE_P(DISABLED_reduce_gpu_ref_f32_f32,
                         general_reduce_gpu_f32_f32,
                         ::testing::Values(
                             TestParamType_general_reduce_gpu(1, 7, 1, 1, 4, 3,format::bfyx, reduce_mode::mean, {reduce::along_b, reduce::along_f, reduce::along_y}, "reduce_ref", false, data_types::f32, false, data_types::f32),
@@ -1775,7 +1775,7 @@ TEST_P(general_reduce_gpu_xy_f32, base) { execute(); }
 class general_reduce_gpu_xy_i8 : public ReduceXYWithBigTensorTestBase<data_types::i8, data_types::i8> {};
 TEST_P(general_reduce_gpu_xy_i8, base) { execute(); }
 
-INSTANTIATE_TEST_CASE_P(reduce_gpu_b_fs_yx_fsv16_xy_f32,
+INSTANTIATE_TEST_SUITE_P(reduce_gpu_b_fs_yx_fsv16_xy_f32,
                         general_reduce_gpu_xy_f32,
                         ::testing::Values(
                             TestParamType_general_reduce_gpu(1, 32, 1, 1,  18,  18, format::b_fs_yx_fsv16, reduce_mode::max, {reduce::along_x, reduce::along_y}, "reduce_gpu_b_fs_yx_fsv16", false, data_types::f32, true, data_types::f32),
@@ -1783,7 +1783,7 @@ INSTANTIATE_TEST_CASE_P(reduce_gpu_b_fs_yx_fsv16_xy_f32,
                         ),
                         general_reduce_gpu::PrintToStringParamName);
 
-INSTANTIATE_TEST_CASE_P(reduce_gpu_b_fs_yx_fsv16_xy_i8,
+INSTANTIATE_TEST_SUITE_P(reduce_gpu_b_fs_yx_fsv16_xy_i8,
                         general_reduce_gpu_xy_i8,
                         ::testing::Values(
                             TestParamType_general_reduce_gpu(1, 32, 1, 1,  18,  18, format::b_fs_yx_fsv16, reduce_mode::max, {reduce::along_x, reduce::along_y}, "reduce_gpu_b_fs_yx_fsv16", false, data_types::i8, true, data_types::i8),
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/reorder_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/reorder_gpu_test.cpp
index 6f0b6a3dfc1eac..fbaa020500b54c 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/reorder_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/reorder_gpu_test.cpp
@@ -2389,7 +2389,7 @@ TEST_P(reorder_test, REORDER)
     run_single_test();
 }
 
-INSTANTIATE_TEST_CASE_P(DISABLED_REORDER,
+INSTANTIATE_TEST_SUITE_P(DISABLED_REORDER,
                         reorder_test,
                         ::testing::ValuesIn(reorder_test::generate_specific_test_params()),
                         tests::generic_test::custom_param_name_functor());
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/resample_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/resample_gpu_test.cpp
index ca7da7f9e83013..d401d6ede4a6bf 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/resample_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/resample_gpu_test.cpp
@@ -746,7 +746,7 @@ struct resample_random_test_param_generator : std::vector<resample_random_test_p
 
 };
 
-INSTANTIATE_TEST_CASE_P(smoke_resample,
+INSTANTIATE_TEST_SUITE_P(smoke_resample,
                         resample_random_test,
                         testing::ValuesIn(
                             resample_random_test_param_generator()
@@ -759,7 +759,7 @@ INSTANTIATE_TEST_CASE_P(smoke_resample,
                             .smoke_params(data_types::f16, format::b_fs_yx_fsv16, format::b_fs_yx_fsv16)
                             .smoke_params(data_types::i8, format::b_fs_yx_fsv16, format::b_fs_yx_fsv16)
                             .smoke_params(data_types::u8, format::b_fs_yx_fsv16, format::b_fs_yx_fsv16)
-                        ), );
+                        ));
 
 
 /////////////////////////////////////////////////////////////////////////
@@ -944,20 +944,20 @@ TEST_P(caffe_resample_random_test, random) {
     execute_compare(param, true);
 }
 
-INSTANTIATE_TEST_CASE_P(caffe_smoke_caffe_fsv16,
+INSTANTIATE_TEST_SUITE_P(caffe_smoke_caffe_fsv16,
                         caffe_resample_random_test,
                         testing::ValuesIn(
                             caffe_resample_random_test_param_generator()
                             .smoke_params(data_types::f32, format::b_fs_yx_fsv16, format::b_fs_yx_fsv16)
                             .smoke_params(data_types::f16, format::b_fs_yx_fsv16, format::b_fs_yx_fsv16)
-                        ), );
+                        ));
 
-INSTANTIATE_TEST_CASE_P(caffe_smoke_caffe_fsv32,
+INSTANTIATE_TEST_SUITE_P(caffe_smoke_caffe_fsv32,
                         caffe_resample_random_test,
                         testing::ValuesIn(
                             caffe_resample_random_test_param_generator()
                             .smoke_params(data_types::f16, format::fs_b_yx_fsv32, format::fs_b_yx_fsv32)
-                        ), );
+                        ));
 
 TEST(resample_gpu, interpolate_in2x2x3x2_nearest1) {
     //  Input  : 2x2x3x2
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/softmax_gpu_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/softmax_gpu_test.cpp
index 2ecc932b41e72d..40ca8c5a9a06bd 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/softmax_gpu_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/softmax_gpu_test.cpp
@@ -998,7 +998,7 @@ TEST_P(softmax_test, SOFTMAX)
     run_single_test();
 }
 
-INSTANTIATE_TEST_CASE_P(DISABLED_SOFTMAX,
+INSTANTIATE_TEST_SUITE_P(DISABLED_SOFTMAX,
     softmax_test,
     ::testing::Combine(::testing::ValuesIn(softmax_test::generate_generic_test_params()), ::testing::ValuesIn(softmax_test::generate_specific_test_params())),
     softmax_test::custom_param_name);
diff --git a/inference-engine/thirdparty/clDNN/tests/test_cases/topology_test.cpp b/inference-engine/thirdparty/clDNN/tests/test_cases/topology_test.cpp
index 42eae68031cb9e..d22af574bfb28a 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_cases/topology_test.cpp
+++ b/inference-engine/thirdparty/clDNN/tests/test_cases/topology_test.cpp
@@ -481,7 +481,7 @@ TEST_P(topology_test, TOPOLOGY)
      }
 }
 
-INSTANTIATE_TEST_CASE_P(DISABLED_TOPOLOGY,
+INSTANTIATE_TEST_SUITE_P(DISABLED_TOPOLOGY,
     topology_test,
     ::testing::Combine( ::testing::ValuesIn(topology_test::generate_all_output_layouts()),
                         ::testing::ValuesIn(topology_test::all_generator_vectors<3>())),
diff --git a/inference-engine/thirdparty/clDNN/tests/test_utils/network_test.h b/inference-engine/thirdparty/clDNN/tests/test_utils/network_test.h
index efc9d7b6cbfd41..9fcf27ae319f0d 100644
--- a/inference-engine/thirdparty/clDNN/tests/test_utils/network_test.h
+++ b/inference-engine/thirdparty/clDNN/tests/test_utils/network_test.h
@@ -30,7 +30,7 @@ namespace tests {
 template <typename T>
 struct typed_comparator {
     static ::testing::AssertionResult compare(const char* lhs_expr, const char* rhs_expr, T ref, T val) {
-        return ::testing::internal::EqHelper<false>::Compare(lhs_expr, rhs_expr, ref, val);
+        return ::testing::internal::EqHelper::Compare(lhs_expr, rhs_expr, ref, val);
     }
 };
 
diff --git a/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_stress_tests.cpp b/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_stress_tests.cpp
index 636fc6f6094274..2f7f74ce63072e 100644
--- a/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_stress_tests.cpp
+++ b/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_stress_tests.cpp
@@ -151,7 +151,7 @@ TEST_P(MvncStressTests, FullCycleOfWork101Times) {
 
 }
 
-INSTANTIATE_TEST_CASE_P(MvncTestsCommon,
+INSTANTIATE_TEST_SUITE_P(MvncTestsCommon,
                         MvncStressTests,
                         ::testing::ValuesIn(myriadProtocols),
                         PrintToStringParamName());
diff --git a/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_common.cpp b/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_common.cpp
index 1a3bca457f9a0e..11c8c58c7116a5 100644
--- a/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_common.cpp
+++ b/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_common.cpp
@@ -539,22 +539,22 @@ TEST_P(MvncInference, DISABLED_DoOneIterationOfInference) {
 }
 
 
-INSTANTIATE_TEST_CASE_P(MvncTestsCommon,
+INSTANTIATE_TEST_SUITE_P(MvncTestsCommon,
                         MvncOpenDevice,
                         ::testing::ValuesIn(myriadProtocols),
                         PrintToStringParamName());
 
-INSTANTIATE_TEST_CASE_P(MvncTestsCommon,
+INSTANTIATE_TEST_SUITE_P(MvncTestsCommon,
                         MvncLoggingTests,
                         ::testing::ValuesIn(myriadProtocols),
                         PrintToStringParamName());
 
-INSTANTIATE_TEST_CASE_P(MvncTestsCommon,
+INSTANTIATE_TEST_SUITE_P(MvncTestsCommon,
                         MvncGraphAllocations,
                         ::testing::ValuesIn(myriadProtocols),
                         PrintToStringParamName());
 
-INSTANTIATE_TEST_CASE_P(MvncTestsCommon,
+INSTANTIATE_TEST_SUITE_P(MvncTestsCommon,
                         MvncInference,
                         ::testing::ValuesIn(myriadProtocols),
                         PrintToStringParamName());
diff --git a/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_usb.cpp b/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_usb.cpp
index 87544a6e0c6e15..5e98b1a77e4261 100644
--- a/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_usb.cpp
+++ b/inference-engine/thirdparty/movidius/mvnc/tests/mvnc_tests_usb.cpp
@@ -242,7 +242,7 @@ TEST_P(MvncDevicePlatform, OpenAndClose) {
 
 }
 
-INSTANTIATE_TEST_CASE_P(MvncTestsPlatform,
+INSTANTIATE_TEST_SUITE_P(MvncTestsPlatform,
                         MvncDevicePlatform,
                         ::testing::ValuesIn(myriadPlatforms),
                         PrintToStringParamName());
diff --git a/inference-engine/thirdparty/movidius/tests/XLink/XLink_common_tests.cpp b/inference-engine/thirdparty/movidius/tests/XLink/XLink_common_tests.cpp
index 81dc53ec4d26e4..7fc215c090a438 100644
--- a/inference-engine/thirdparty/movidius/tests/XLink/XLink_common_tests.cpp
+++ b/inference-engine/thirdparty/movidius/tests/XLink/XLink_common_tests.cpp
@@ -374,56 +374,56 @@ TEST_P(XLinkOpenStreamTests, DISABLED_CannotOpenStreamsMoreThanMemoryOnDevice) {
 // Initialization of XLinkCommonTests
 //------------------------------------------------------------------------------
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     XLinkCommon,
     XLinkBootTests,
     Combine(Values(X_LINK_USB_VSC, X_LINK_PCIE),
             Values(X_LINK_ANY_PLATFORM)),
     XLinkBootTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     XLinkCommon,
     XLinkConnectTests,
     Combine(Values(X_LINK_USB_VSC, X_LINK_PCIE),
             Values(X_LINK_ANY_PLATFORM)),
     XLinkBootTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     XLinkCommon,
     XLinkFindFirstSuitableDevicePlatformTests,
     Combine(Values(X_LINK_USB_VSC),
             Values(X_LINK_MYRIAD_2, X_LINK_MYRIAD_X, X_LINK_ANY_PLATFORM)),
     XLinkBootTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     XLinkCommonPCIE,
     XLinkFindFirstSuitableDevicePlatformTests,
     Combine(Values(X_LINK_PCIE),
             Values(X_LINK_ANY_PLATFORM)),
     XLinkBootTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     XLinkCommon,
     XLinkFindFirstSuitableDeviceTests,
     Combine(Values(X_LINK_USB_VSC, X_LINK_PCIE),
             Values(X_LINK_ANY_PLATFORM)),
     XLinkBootTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     XLinkCommon,
     XLinkResetAllTests,
     Combine(Values(X_LINK_USB_VSC, X_LINK_PCIE),
             Values(X_LINK_ANY_PLATFORM)),
     XLinkBootTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     XLinkCommon,
     XLinkResetRemoteTests,
     Combine(Values(X_LINK_USB_VSC, X_LINK_PCIE),
             Values(X_LINK_ANY_PLATFORM)),
     XLinkBootTests::getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(
+INSTANTIATE_TEST_SUITE_P(
     XLinkCommon,
     XLinkOpenStreamTests,
     Combine(Values(X_LINK_USB_VSC, X_LINK_PCIE),
diff --git a/tests/stress_tests/memcheck_tests/tests.cpp b/tests/stress_tests/memcheck_tests/tests.cpp
index f939869dea6779..c8cefc9575650a 100644
--- a/tests/stress_tests/memcheck_tests/tests.cpp
+++ b/tests/stress_tests/memcheck_tests/tests.cpp
@@ -102,7 +102,7 @@ TEST_P(MemCheckTestSuite, infer_request_inference) {
 }
 // tests_pipelines/tests_pipelines.cpp
 
-INSTANTIATE_TEST_CASE_P(MemCheckTests, MemCheckTestSuite,
+INSTANTIATE_TEST_SUITE_P(MemCheckTests, MemCheckTestSuite,
                         ::testing::ValuesIn(
                                 generateTestsParams({"devices", "models"})),
                         getTestCaseName);
diff --git a/tests/stress_tests/memleaks_tests/tests.cpp b/tests/stress_tests/memleaks_tests/tests.cpp
index b00f24f8552aa2..22c1812cb742c4 100644
--- a/tests/stress_tests/memleaks_tests/tests.cpp
+++ b/tests/stress_tests/memleaks_tests/tests.cpp
@@ -120,15 +120,15 @@ TEST_P(MemLeaksTestSuite, infer_request_inference) {
 }
 // tests_pipelines/tests_pipelines.cpp
 
-INSTANTIATE_TEST_CASE_P(MemLeaksTests, MemLeaksTestSuiteNoModel,
+INSTANTIATE_TEST_SUITE_P(MemLeaksTests, MemLeaksTestSuiteNoModel,
                         ::testing::ValuesIn(generateTestsParams({"processes", "threads", "iterations", "devices"})),
                         getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(MemLeaksTests, MemLeaksTestSuiteNoDevice,
+INSTANTIATE_TEST_SUITE_P(MemLeaksTests, MemLeaksTestSuiteNoDevice,
                         ::testing::ValuesIn(generateTestsParams({"processes", "threads", "iterations", "models"})),
                         getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(MemLeaksTests, MemLeaksTestSuite,
+INSTANTIATE_TEST_SUITE_P(MemLeaksTests, MemLeaksTestSuite,
                         ::testing::ValuesIn(
                                 generateTestsParams({"processes", "threads", "iterations", "devices", "models"})),
                         getTestCaseName);
diff --git a/tests/stress_tests/unittests/tests.cpp b/tests/stress_tests/unittests/tests.cpp
index 43d594563ba8ed..1a13732da091f0 100644
--- a/tests/stress_tests/unittests/tests.cpp
+++ b/tests/stress_tests/unittests/tests.cpp
@@ -77,14 +77,14 @@ TEST_P(UnitTestSuite, infer_request_inference_full_pipeline) {
 }
 // tests_pipelines/tests_pipelines_full_pipeline.cpp
 
-INSTANTIATE_TEST_CASE_P(StressUnitTests, UnitTestSuiteNoModel,
+INSTANTIATE_TEST_SUITE_P(StressUnitTests, UnitTestSuiteNoModel,
                         ::testing::ValuesIn(generateTestsParams({"processes", "threads", "iterations", "devices"})),
                         getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(StressUnitTests, UnitTestSuiteNoDevice,
+INSTANTIATE_TEST_SUITE_P(StressUnitTests, UnitTestSuiteNoDevice,
                         ::testing::ValuesIn(generateTestsParams({"processes", "threads", "iterations", "models"})),
                         getTestCaseName);
 
-INSTANTIATE_TEST_CASE_P(StressUnitTests, UnitTestSuite,
+INSTANTIATE_TEST_SUITE_P(StressUnitTests, UnitTestSuite,
                         ::testing::ValuesIn(generateTestsParams({"processes", "threads", "iterations", "devices", "models"})),
                         getTestCaseName);
diff --git a/thirdparty/CMakeLists.txt b/thirdparty/CMakeLists.txt
index 2c6b656408b6f6..9b74b65ca1f2de 100644
--- a/thirdparty/CMakeLists.txt
+++ b/thirdparty/CMakeLists.txt
@@ -2,9 +2,61 @@
 # SPDX-License-Identifier: Apache-2.0
 #
 
-add_subdirectory(ittapi)
-add_subdirectory(itt_collector)
+add_subdirectory(ittapi EXCLUDE_FROM_ALL)
+add_subdirectory(itt_collector EXCLUDE_FROM_ALL)
 add_subdirectory(xbyak EXCLUDE_FROM_ALL)
 add_subdirectory(zlib EXCLUDE_FROM_ALL)
 add_subdirectory(cnpy EXCLUDE_FROM_ALL)
+
+#
+# Google Tests framework
+#
+
+function(add_gtest_libraries)
+    set(gtest_force_shared_crt ON CACHE BOOL "disable static CRT for google test")
+    if (UNIX)
+        ie_add_compiler_flags(-Wno-undef)
+        if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+            ie_add_compiler_flags(-Wno-deprecated-copy)
+        endif()
+    endif()
+
+    set(BUILD_SHARED_LIBS OFF)
+    set(INSTALL_GTEST OFF)
+    add_subdirectory(gtest EXCLUDE_FROM_ALL)
+
+    get_target_property(gtest_include_dirs gtest INTERFACE_INCLUDE_DIRECTORIES)
+    set_target_properties(gtest PROPERTIES INTERFACE_SYSTEM_INCLUDE_DIRECTORIES "${gtest_include_dirs}")
+
+    get_target_property(gmock_include_dirs gtest INTERFACE_INCLUDE_DIRECTORIES)
+    set_target_properties(gmock PROPERTIES INTERFACE_SYSTEM_INCLUDE_DIRECTORIES "${gmock_include_dirs};${gmock_SOURCE_DIR}/include")
+
+    # If we have specified /Z7 option, remove -Zi option which comes from gtest
+    if (WIN32)
+        set(gtest_targets "gtest;gtest_main;gmock;gmock_main")
+        foreach(target_name ${gtest_targets})
+            if(TARGET "${target_name}")
+                get_target_property(_target_cxx_flags ${target_name} COMPILE_FLAGS)
+                if(_target_cxx_flags)
+                    if(CMAKE_CXX_FLAGS_DEBUG MATCHES ".+/Z7.+" OR CMAKE_CXX_FLAGS_RELWITHDEBINFO MATCHES ".+/Z7.+")
+                        string(REPLACE "-Zi" " " _target_cxx_flags ${_target_cxx_flags})
+                        message(STATUS "Removing -Zi flag from target " ${target_name})
+                        set_target_properties(${target_name} PROPERTIES COMPILE_FLAGS "${_target_cxx_flags}")
+                    endif()
+                endif()
+            endif()
+        endforeach()
+    endif()
+    set_target_properties(gtest gtest_main gmock gmock_main
+                          PROPERTIES FOLDER thirdparty)
+endfunction()
+
+if(NGRAPH_UNIT_TEST_ENABLE OR ENABLE_TESTS)
+    add_gtest_libraries()
+endif()
+
+#
+# Export 3rdparty components
+#
+
 openvino_developer_export_targets(COMPONENT openvino_common TARGETS xbyak)
diff --git a/thirdparty/gtest b/thirdparty/gtest
new file mode 160000
index 00000000000000..f8bd7cebfb3e6a
--- /dev/null
+++ b/thirdparty/gtest
@@ -0,0 +1 @@
+Subproject commit f8bd7cebfb3e6aeefb1d1d93e46a461471e15f57

From ad997410c87333520cccd5834aa68e1de41d7cb9 Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Tue, 22 Jun 2021 17:43:17 +0300
Subject: [PATCH 163/178] Port to master (#6279)

* add single_layer_test for Interpolate-1 (#6133)

* Fixed InferenceEngineConfig.cmake usage in include() (#6136)

* Deprecated API updates (#6252)

* Fixed links to OMZ / DL Streamer (#6257)

* Added doxygen deprecation for LowLatency

Co-authored-by: Elena Gvozdeva <elena.gvozdeva@intel.com>
---
 README.md                                     |  2 +-
 docs/IE_DG/API_Changes.md                     | 36 ++++++++++
 docs/benchmarks/performance_benchmarks_faq.md | 26 ++++----
 docs/get_started/get_started_linux.md         |  2 +-
 docs/get_started/get_started_macos.md         |  2 +-
 docs/get_started/get_started_raspbian.md      |  4 +-
 docs/get_started/get_started_windows.md       |  2 +-
 docs/index.md                                 |  8 +--
 .../install_guides/installing-openvino-apt.md |  6 +-
 .../installing-openvino-linux.md              |  4 +-
 .../installing-openvino-macos.md              |  2 +-
 .../installing-openvino-raspbian.md           |  4 +-
 .../installing-openvino-windows.md            |  2 +-
 .../install_guides/installing-openvino-yum.md |  2 +-
 .../dldt_optimization_guide.md                |  4 +-
 .../templates/InferenceEngineConfig.cmake.in  | 10 +++
 .../include/ie_transformations.hpp            |  3 +-
 .../single_layer/interpolate.hpp              | 23 +++++++
 .../src/single_layer/interpolate.cpp          | 65 +++++++++++++++++++
 tests/stress_tests/README.md                  |  2 +-
 tests/stress_tests/scripts/get_testdata.py    |  2 +-
 tools/benchmark/README.md                     |  2 +-
 22 files changed, 173 insertions(+), 40 deletions(-)

diff --git a/README.md b/README.md
index 30c314e76de9b3..7ce2e9f998a1f6 100644
--- a/README.md
+++ b/README.md
@@ -42,7 +42,7 @@ Please report questions, issues and suggestions using:
 ---
 \* Other names and brands may be claimed as the property of others.
 
-[Open Model Zoo]:https://github.com/opencv/open_model_zoo
+[Open Model Zoo]:https://github.com/openvinotoolkit/open_model_zoo
 [Inference Engine]:https://software.intel.com/en-us/articles/OpenVINO-InferEngine
 [Model Optimizer]:https://software.intel.com/en-us/articles/OpenVINO-ModelOptimizer
 [nGraph]:https://docs.openvinotoolkit.org/latest/openvino_docs_nGraph_DG_DevGuide.html
diff --git a/docs/IE_DG/API_Changes.md b/docs/IE_DG/API_Changes.md
index 2534a4a6c3856a..df04ddd6789f07 100644
--- a/docs/IE_DG/API_Changes.md
+++ b/docs/IE_DG/API_Changes.md
@@ -10,10 +10,14 @@ The sections below contain detailed list of changes made to the Inference Engine
 
 ### Deprecated API
 
+ **InferenceEngine::Parameter**
+
  * InferenceEngine::Parameter(const std::shared_ptr<ngraph::Variant>&)
  * InferenceEngine::Parameter(std::shared_ptr<ngraph::Variant>& var)
  * std::shared_ptr<ngraph::Variant> InferenceEngine::Parameter::asVariant() const
  * InferenceEngine::Parameter::operator std::shared_ptr<ngraph::Variant>() const
+
+ **GPU plugin configuration keys**
  * KEY_CLDNN_NV12_TWO_INPUTS GPU plugin option. Use KEY_GPU_NV12_TWO_INPUTS instead
  * KEY_CLDNN_PLUGIN_PRIORITY GPU plugin option. Use KEY_GPU_PLUGIN_PRIORITY instead
  * KEY_CLDNN_PLUGIN_THROTTLE GPU plugin option. Use KEY_GPU_PLUGIN_THROTTLE instead
@@ -24,6 +28,38 @@ The sections below contain detailed list of changes made to the Inference Engine
  * KEY_TUNING_MODE GPU plugin option
  * KEY_TUNING_FILE GPU plugin option
 
+ **InferenceEngine::IInferRequest**
+ * IInferRequest interface is deprecated, use InferRequest wrapper:
+  * Constructor for InferRequest from IInferRequest:: Ptr is deprecated
+  * Cast operator for InferRequest to IInferRequest shared pointer is deprecated
+
+ **InferenceEngine::ICNNNetwork**
+ * ICNNNetwork interface is deprecated by means of deprecation of all its methods, use CNNNetwork wrapper
+  * CNNNetwork methods working with ICNNNetwork are deprecated:
+  * Cast to ICNNNetwork shared pointer
+  * Cast to reference to ICNNNetwork interface
+  * Constructor from ICNNNetwork shared pointer
+
+ **InferenceEngine::IExecutableNetwork**
+ * IExecutableNetwork is deprecated, use ExecutableNetwork wrappers:
+  * Constructor of ExecutableNetwork from IExecutableNetwork shared pointer is deprecated
+ * The following ExecutableNetwork methods are deprecated:
+  * ExecutableNetwork::reset
+  * Cast operator to IExecutableNetwork shared pointer
+  * ExecutableNetwork::CreateInferRequestPtr - use ExecutableNetwork::CreateInferRequest instead
+
+ **Extensions API**
+ * InferenceEngine::make_so_pointer which is used to create Extensions library is replaced by std::make_shared<Extension>(..)
+ * InferenceEngine::IExtension::Release is deprecated with no replacement
+ * Use IE_DEFINE_EXTENSION_CREATE_FUNCTION helper macro instead of explicit declaration of CreateExtension function, which create extension.
+
+ **Other changes**
+ * Version::ApiVersion structure is deprecated, Inference Engine does not have API version anymore
+ * LowLatency - use lowLatency2 instead
+ * CONFIG_KEY(DUMP_EXEC_GRAPH_AS_DOT) - use InferenceEngine::ExecutableNetwork::GetExecGraphInfo::serialize() instead
+ * Core::ImportNetwork with no device - pass device name explicitly.
+ * details::InferenceEngineException - use InferenceEngine::Exception and its derivatives instead.
+
 ## 2021.3
 
 ### New API
diff --git a/docs/benchmarks/performance_benchmarks_faq.md b/docs/benchmarks/performance_benchmarks_faq.md
index f48c3cf38fde52..a89d0fc07c3e90 100644
--- a/docs/benchmarks/performance_benchmarks_faq.md
+++ b/docs/benchmarks/performance_benchmarks_faq.md
@@ -6,7 +6,7 @@ The following questions and answers are related to [performance benchmarks](./pe
 New performance benchmarks are typically published on every `major.minor` release of the Intel® Distribution of OpenVINO™ toolkit.
 
 #### 2. Where can I find the models used in the performance benchmarks?
-All of the models used are included in the toolkit's [Open Model Zoo](https://github.com/opencv/open_model_zoo) GitHub repository. 
+All of the models used are included in the toolkit's [Open Model Zoo](https://github.com/openvinotoolkit/open_model_zoo) GitHub repository. 
 
 #### 3. Will there be new models added to the list used for benchmarking?
 The models used in the performance benchmarks were chosen based on general adoption and usage in deployment scenarios. We're continuing to add new models that support a diverse set of workloads and usage.
@@ -21,23 +21,23 @@ All of the performance benchmarks were generated using the open-sourced tool wit
 The image size used in the inference depends on the network being benchmarked. The following table shows the list of input sizes for each network model.
 |   **Model**																														 |   **Public Network**                    |     **Task**                | **Input Size** (Height x Width)   |
 |------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------|-----------------------------|-----------------------------------|
-|    [bert-large-uncased-whole-word-masking-squad](https://github.com/opencv/open_model_zoo/tree/develop/models/intel/bert-large-uncased-whole-word-masking-squad-int8-0001)   | 	BERT-large	|question / answer	|384|
-|    [deeplabv3-TF](https://github.com/opencv/open_model_zoo/tree/master/models/public/deeplabv3)                                    |	  DeepLab v3 Tf	                       |semantic segmentation	     |    513x513                          |
+|    [bert-large-uncased-whole-word-masking-squad](https://github.com/openvinotoolkit/open_model_zoo/tree/develop/models/intel/bert-large-uncased-whole-word-masking-squad-int8-0001)   | 	BERT-large	|question / answer	|384|
+|    [deeplabv3-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/deeplabv3)                                    |	  DeepLab v3 Tf	                       |semantic segmentation	     |    513x513                          |
 |    [densenet-121-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/densenet-121-tf)                  | 	  Densenet-121 Tf	                   |classification	    |    224x224                 |
-|    [facenet-20180408-102900-TF](https://github.com/opencv/open_model_zoo/tree/master/models/public/facenet-20180408-102900)        |    FaceNet TF                           |    face recognition         |    160x160                        |
-|    [faster_rcnn_resnet50_coco-TF](https://github.com/opencv/open_model_zoo/tree/master/models/public/faster_rcnn_resnet50_coco)    |    Faster RCNN Tf                       |    object detection           |    600x1024					     |
+|    [facenet-20180408-102900-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/facenet-20180408-102900)        |    FaceNet TF                           |    face recognition         |    160x160                        |
+|    [faster_rcnn_resnet50_coco-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/faster_rcnn_resnet50_coco)    |    Faster RCNN Tf                       |    object detection           |    600x1024					     |
 |    [googlenet-v1-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/googlenet-v1-tf)				     |    GoogLeNet_ILSVRC-2012                |    classification           |    224x224				  |
-|    [inception-v3-TF](https://github.com/opencv/open_model_zoo/tree/master/models/public/googlenet-v3)								 |    Inception v3 Tf                      |    classification           |    299x299				  |
-|    [mobilenet-ssd-CF](https://github.com/opencv/open_model_zoo/tree/master/models/public/mobilenet-ssd)						     |    SSD (MobileNet)_COCO-2017_Caffe      |    object detection         |    300x300						 |
+|    [inception-v3-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/googlenet-v3)								 |    Inception v3 Tf                      |    classification           |    299x299				  |
+|    [mobilenet-ssd-CF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/mobilenet-ssd)						     |    SSD (MobileNet)_COCO-2017_Caffe      |    object detection         |    300x300						 |
 |    [mobilenet-v1-1.0-224-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/mobilenet-v1-1.0-224-tf)  |    MobileNet v1 Tf                      |    classification    |    224x224                        |
-|    [mobilenet-v2-1.0-224-TF](https://github.com/opencv/open_model_zoo/tree/master/models/public/mobilenet-v2-1.0-224)			     |    MobileNet v2 Tf                      |    classification           |    224x224						 |
+|    [mobilenet-v2-1.0-224-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/mobilenet-v2-1.0-224)			     |    MobileNet v2 Tf                      |    classification           |    224x224						 |
 |    [mobilenet-v2-pytorch](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/mobilenet-v2-pytorch )		 |    Mobilenet V2 PyTorch                 |    classification           |    224x224					     |
-|    [resnet-18-pytorch](https://github.com/opencv/open_model_zoo/tree/master/models/public/resnet-18-pytorch)		  			     |    ResNet-18 PyTorch                    |    classification           |    224x224						 |
+|    [resnet-18-pytorch](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/resnet-18-pytorch)		  			     |    ResNet-18 PyTorch                    |    classification           |    224x224						 |
 |    [resnet-50-pytorch](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/resnet-50-pytorch)              |    ResNet-50 v1 PyTorch                 |    classification           |    224x224                        |
-|    [resnet-50-TF](https://github.com/opencv/open_model_zoo/tree/master/models/public/resnet-50-tf)								 |    ResNet-50_v1_ILSVRC-2012             |    classification           |    224x224						 |
-|    [se-resnext-50-CF](https://github.com/opencv/open_model_zoo/tree/master/models/public/se-resnext-50)						     |    Se-ResNext-50_ILSVRC-2012_Caffe      |    classification           |    224x224						 |
-|    [squeezenet1.1-CF](https://github.com/opencv/open_model_zoo/tree/master/models/public/squeezenet1.1)						     |    SqueezeNet_v1.1_ILSVRC-2012_Caffe    |    classification           |    227x227						 |
-|    [ssd300-CF](https://github.com/opencv/open_model_zoo/tree/master/models/public/ssd300)										     |    SSD (VGG-16)_VOC-2007_Caffe          |    object detection         |    300x300						 |
+|    [resnet-50-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/resnet-50-tf)								 |    ResNet-50_v1_ILSVRC-2012             |    classification           |    224x224						 |
+|    [se-resnext-50-CF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/se-resnext-50)						     |    Se-ResNext-50_ILSVRC-2012_Caffe      |    classification           |    224x224						 |
+|    [squeezenet1.1-CF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/squeezenet1.1)						     |    SqueezeNet_v1.1_ILSVRC-2012_Caffe    |    classification           |    227x227						 |
+|    [ssd300-CF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/ssd300)										     |    SSD (VGG-16)_VOC-2007_Caffe          |    object detection         |    300x300						 |
 |    [yolo_v3-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/yolo-v3-tf)                            | 	  TF Keras YOLO v3 Modelset            |	 object detection	      |    300x300                        |
 |    [yolo_v4-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/yolo-v4-tf)                            | 	  Yolo-V4 TF                           |	 object detection	     |    608x608                        |
 |    [ssd_mobilenet_v1_coco-TF](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/ssd_mobilenet_v1_coco)   |    ssd_mobilenet_v1_coco                |    object detection         |    300x300                        |
diff --git a/docs/get_started/get_started_linux.md b/docs/get_started/get_started_linux.md
index d64d63ed2fccf9..b7b8bd470693ad 100644
--- a/docs/get_started/get_started_linux.md
+++ b/docs/get_started/get_started_linux.md
@@ -227,7 +227,7 @@ You must have a model that is specific for you inference task. Example model typ
 - Custom (Often based on SSD)
 
 Options to find a model suitable for the OpenVINO™ toolkit are:
-- Download public and Intel's pre-trained models from the [Open Model Zoo](https://github.com/opencv/open_model_zoo) using [Model Downloader tool](@ref omz_tools_downloader).
+- Download public and Intel's pre-trained models from the [Open Model Zoo](https://github.com/openvinotoolkit/open_model_zoo) using [Model Downloader tool](@ref omz_tools_downloader).
 - Download from GitHub*, Caffe* Zoo, TensorFlow* Zoo, etc.
 - Train your own model.
         
diff --git a/docs/get_started/get_started_macos.md b/docs/get_started/get_started_macos.md
index a15240a1c9b9c4..c58cd418bcf698 100644
--- a/docs/get_started/get_started_macos.md
+++ b/docs/get_started/get_started_macos.md
@@ -211,7 +211,7 @@ You must have a model that is specific for you inference task. Example model typ
 - Custom (Often based on SSD)
 
 Options to find a model suitable for the OpenVINO™ toolkit are:
-- Download public and Intel's pre-trained models from the [Open Model Zoo](https://github.com/opencv/open_model_zoo) using the [Model Downloader tool](@ref omz_tools_downloader). 
+- Download public and Intel's pre-trained models from the [Open Model Zoo](https://github.com/openvinotoolkit/open_model_zoo) using the [Model Downloader tool](@ref omz_tools_downloader). 
 - Download from GitHub*, Caffe* Zoo, TensorFlow* Zoo, and other resources.
 - Train your own model.
         
diff --git a/docs/get_started/get_started_raspbian.md b/docs/get_started/get_started_raspbian.md
index 5f3baf87d2f638..d810958d723f4c 100644
--- a/docs/get_started/get_started_raspbian.md
+++ b/docs/get_started/get_started_raspbian.md
@@ -13,7 +13,7 @@ On Raspbian* OS, the OpenVINO™ toolkit consists of the following components:
 
 > **NOTE**:
 > * The OpenVINO™ package for Raspberry* does not include the [Model Optimizer](../MO_DG/Deep_Learning_Model_Optimizer_DevGuide.md). To convert models to Intermediate Representation (IR), you need to install it separately to your host machine.
-> * The package does not include the Open Model Zoo demo applications. You can download them separately from the [Open Models Zoo repository](https://github.com/opencv/open_model_zoo).
+> * The package does not include the Open Model Zoo demo applications. You can download them separately from the [Open Models Zoo repository](https://github.com/openvinotoolkit/open_model_zoo).
 
 In addition, [code samples](../IE_DG/Samples_Overview.md) are provided to help you get up and running with the toolkit.
 
@@ -43,7 +43,7 @@ The primary tools for deploying your models and applications are installed to th
 The OpenVINO™ workflow on Raspbian* OS is as follows:
 1. **Get a pre-trained model** for your inference task. If you want to use your model for inference, the model must be converted to the `.bin` and `.xml` Intermediate Representation (IR) files, which are used as input by Inference Engine. On Raspberry PI, OpenVINO™ toolkit includes only the Inference Engine module. The Model Optimizer is not supported on this platform. To get the optimized models you can use one of the following options:
    
-   * Download public and Intel's pre-trained models from the [Open Model Zoo](https://github.com/opencv/open_model_zoo) using [Model Downloader tool](@ref omz_tools_downloader).
+   * Download public and Intel's pre-trained models from the [Open Model Zoo](https://github.com/openvinotoolkit/open_model_zoo) using [Model Downloader tool](@ref omz_tools_downloader).
     <br> For more information on pre-trained models, see [Pre-Trained Models Documentation](@ref omz_models_group_intel)
    
    * Convert a model using the Model Optimizer from a full installation of Intel® Distribution of OpenVINO™ toolkit on one of the supported platforms. Installation instructions are available:
diff --git a/docs/get_started/get_started_windows.md b/docs/get_started/get_started_windows.md
index 253af476efb186..fa6680d30b92df 100644
--- a/docs/get_started/get_started_windows.md
+++ b/docs/get_started/get_started_windows.md
@@ -211,7 +211,7 @@ You must have a model that is specific for you inference task. Example model typ
 - Custom (Often based on SSD)
 
 Options to find a model suitable for the OpenVINO™ toolkit are:
-- Download public and Intel's pre-trained models from the [Open Model Zoo](https://github.com/opencv/open_model_zoo) using the [Model Downloader tool](@ref omz_tools_downloader).
+- Download public and Intel's pre-trained models from the [Open Model Zoo](https://github.com/openvinotoolkit/open_model_zoo) using the [Model Downloader tool](@ref omz_tools_downloader).
 - Download from GitHub*, Caffe* Zoo, TensorFlow* Zoo, and other resources.
 - Train your own model.
         
diff --git a/docs/index.md b/docs/index.md
index ca6a1fa6b2ecdb..76da992f162087 100644
--- a/docs/index.md
+++ b/docs/index.md
@@ -94,12 +94,12 @@ Intel® Distribution of OpenVINO™ toolkit includes the following components:
 - [Open Model Zoo](@ref omz_models_group_intel)     
     - [Demos](@ref omz_demos): Console applications that provide robust application templates to help you implement specific deep learning scenarios.
     - Additional Tools: A set of tools to work with your models including [Accuracy Checker Utility](@ref omz_tools_accuracy_checker) and [Model Downloader](@ref omz_tools_downloader).
-    - [Documentation for Pretrained Models](@ref omz_models_group_intel): Documentation for pre-trained models that are available in the [Open Model Zoo repository](https://github.com/opencv/open_model_zoo).
-- Deep Learning Streamer (DL Streamer): Streaming analytics framework, based on GStreamer, for constructing graphs of media analytics components. DL Streamer can be installed by the Intel® Distribution of OpenVINO™ toolkit installer. Its open-source version is available on [GitHub](https://github.com/opencv/gst-video-analytics). For the DL Streamer documentation, see:
+    - [Documentation for Pretrained Models](@ref omz_models_group_intel): Documentation for pre-trained models that are available in the [Open Model Zoo repository](https://github.com/openvinotoolkit/open_model_zoo).
+- Deep Learning Streamer (DL Streamer): Streaming analytics framework, based on GStreamer, for constructing graphs of media analytics components. DL Streamer can be installed by the Intel® Distribution of OpenVINO™ toolkit installer. Its open-source version is available on [GitHub](https://github.com/openvinotoolkit/dlstreamer_gst). For the DL Streamer documentation, see:
     - [DL Streamer Samples](@ref gst_samples_README)
     - [API Reference](https://openvinotoolkit.github.io/dlstreamer_gst/)
-    - [Elements](https://github.com/opencv/gst-video-analytics/wiki/Elements)
-    - [Tutorial](https://github.com/opencv/gst-video-analytics/wiki/DL%20Streamer%20Tutorial)
+    - [Elements](https://github.com/openvinotoolkit/dlstreamer_gst/wiki/Elements)
+    - [Tutorial](https://github.com/openvinotoolkit/dlstreamer_gst/wiki/DL-Streamer-Tutorial)
 - [OpenCV](https://docs.opencv.org/master/) : OpenCV* community version compiled for Intel® hardware
 - [Intel® Media SDK](https://software.intel.com/en-us/media-sdk) (in Intel® Distribution of OpenVINO™ toolkit for Linux only)
 
diff --git a/docs/install_guides/installing-openvino-apt.md b/docs/install_guides/installing-openvino-apt.md
index 1bd734bd856969..982d71102b37a3 100644
--- a/docs/install_guides/installing-openvino-apt.md
+++ b/docs/install_guides/installing-openvino-apt.md
@@ -14,7 +14,7 @@ The following components are installed with the OpenVINO runtime package:
 |-----------|------------|
 | [Inference Engine](../IE_DG/Deep_Learning_Inference_Engine_DevGuide.md)| The engine that runs a deep learning model. It includes a set of libraries for an easy inference integration into your applications. |
 | [OpenCV*](https://docs.opencv.org/master/) | OpenCV* community version compiled for Intel® hardware. |
-| Deep Learning Streamer (DL Streamer) | Streaming analytics framework, based on GStreamer, for constructing graphs of media analytics components. For the DL Streamer documentation, see [DL Streamer Samples](@ref gst_samples_README), [API Reference](https://openvinotoolkit.github.io/dlstreamer_gst/), [Elements](https://github.com/opencv/gst-video-analytics/wiki/Elements), [Tutorial](https://github.com/opencv/gst-video-analytics/wiki/DL%20Streamer%20Tutorial). |
+| Deep Learning Streamer (DL Streamer) | Streaming analytics framework, based on GStreamer, for constructing graphs of media analytics components. For the DL Streamer documentation, see [DL Streamer Samples](@ref gst_samples_README), [API Reference](https://openvinotoolkit.github.io/dlstreamer_gst/), [Elements](https://github.com/openvinotoolkit/dlstreamer_gst/wiki/Elements), [Tutorial](https://github.com/openvinotoolkit/dlstreamer_gst/wiki/DL-Streamer-Tutorial). |
 
 ## Included with Developer Package
 
@@ -28,8 +28,8 @@ The following components are installed with the OpenVINO developer package:
 | [Sample Applications](../IE_DG/Samples_Overview.md)           | A set of simple console applications demonstrating how to use the Inference Engine in your applications. |
 | [Demo Applications](@ref omz_demos) | A set of console applications that demonstrate how you can use the Inference Engine in your applications to solve specific use cases. |
 | Additional Tools                                   | A set of tools to work with your models including [Accuracy Checker utility](@ref omz_tools_accuracy_checker), [Post-Training Optimization Tool Guide](@ref pot_README), [Model Downloader](@ref omz_tools_downloader) and other  |
-| [Documentation for Pre-Trained Models ](@ref omz_models_group_intel)                                   | Documentation for the pre-trained models available in the [Open Model Zoo repo](https://github.com/opencv/open_model_zoo).  |
-| Deep Learning Streamer (DL Streamer)   | Streaming analytics framework, based on GStreamer\*, for constructing graphs of media analytics components. For the DL Streamer documentation, see [DL Streamer Samples](@ref gst_samples_README), [API Reference](https://openvinotoolkit.github.io/dlstreamer_gst/), [Elements](https://github.com/opencv/gst-video-analytics/wiki/Elements), [Tutorial](https://github.com/opencv/gst-video-analytics/wiki/DL%20Streamer%20Tutorial). |
+| [Documentation for Pre-Trained Models ](@ref omz_models_group_intel)                                   | Documentation for the pre-trained models available in the [Open Model Zoo repo](https://github.com/openvinotoolkit/open_model_zoo).  |
+| Deep Learning Streamer (DL Streamer)   | Streaming analytics framework, based on GStreamer\*, for constructing graphs of media analytics components. For the DL Streamer documentation, see [DL Streamer Samples](@ref gst_samples_README), [API Reference](https://openvinotoolkit.github.io/dlstreamer_gst/), [Elements](https://github.com/openvinotoolkit/dlstreamer_gst/wiki/Elements), [Tutorial](https://github.com/openvinotoolkit/dlstreamer_gst/wiki/DL-Streamer-Tutorial). |
 
 ## Set up the Repository
 ### Install the GPG key for the repository
diff --git a/docs/install_guides/installing-openvino-linux.md b/docs/install_guides/installing-openvino-linux.md
index a78fa8fc43d7a1..4a7dd77c506ba0 100644
--- a/docs/install_guides/installing-openvino-linux.md
+++ b/docs/install_guides/installing-openvino-linux.md
@@ -28,8 +28,8 @@ The Intel® Distribution of OpenVINO™ toolkit for Linux\*:
 | [Inference Engine Code Samples](../IE_DG/Samples_Overview.md)           | A set of simple console applications demonstrating how to utilize specific OpenVINO capabilities in an application and how to perform specific tasks, such as loading a model, running inference, querying specific device capabilities, and more. |
 | [Demo Applications](@ref omz_demos)           | A set of simple console applications that provide robust application templates to help you implement specific deep learning scenarios. |
 | Additional Tools                                   | A set of tools to work with your models including [Accuracy Checker utility](@ref omz_tools_accuracy_checker), [Post-Training Optimization Tool Guide](@ref pot_README), [Model Downloader](@ref omz_tools_downloader) and other  |
-| [Documentation for Pre-Trained Models ](@ref omz_models_group_intel)                                   | Documentation for the pre-trained models available in the [Open Model Zoo repo](https://github.com/opencv/open_model_zoo).  |
-| Deep Learning Streamer (DL Streamer)   | Streaming analytics framework, based on GStreamer, for constructing graphs of media analytics components. For the DL Streamer documentation, see [DL Streamer Samples](@ref gst_samples_README), [API Reference](https://openvinotoolkit.github.io/dlstreamer_gst/), [Elements](https://github.com/opencv/gst-video-analytics/wiki/Elements), [Tutorial](https://github.com/opencv/gst-video-analytics/wiki/DL%20Streamer%20Tutorial). |
+| [Documentation for Pre-Trained Models ](@ref omz_models_group_intel)                                   | Documentation for the pre-trained models available in the [Open Model Zoo repo](https://github.com/openvinotoolkit/open_model_zoo).  |
+| Deep Learning Streamer (DL Streamer)   | Streaming analytics framework, based on GStreamer, for constructing graphs of media analytics components. For the DL Streamer documentation, see [DL Streamer Samples](@ref gst_samples_README), [API Reference](https://openvinotoolkit.github.io/dlstreamer_gst/), [Elements](https://github.com/openvinotoolkit/dlstreamer_gst/wiki/Elements), [Tutorial](https://github.com/openvinotoolkit/dlstreamer_gst/wiki/DL-Streamer-Tutorial). |
 
 **Could Be Optionally Installed**
 
diff --git a/docs/install_guides/installing-openvino-macos.md b/docs/install_guides/installing-openvino-macos.md
index d878eac5c3a84a..e4a225bdab05bd 100644
--- a/docs/install_guides/installing-openvino-macos.md
+++ b/docs/install_guides/installing-openvino-macos.md
@@ -29,7 +29,7 @@ The following components are installed by default:
 | [Sample Applications](../IE_DG/Samples_Overview.md)                                                                                | A set of simple console applications demonstrating how to use the Inference Engine in your applications. |
 | [Demos](@ref omz_demos)                                   | A set of console applications that demonstrate how you can use the Inference Engine in your applications to solve specific use-cases  |
 | Additional Tools                                   | A set of tools to work with your models including [Accuracy Checker utility](@ref omz_tools_accuracy_checker), [Post-Training Optimization Tool Guide](@ref pot_README), [Model Downloader](@ref omz_tools_downloader) and other  |
-| [Documentation for Pre-Trained Models ](@ref omz_models_group_intel)                                   | Documentation for the pre-trained models available in the [Open Model Zoo repo](https://github.com/opencv/open_model_zoo)  |
+| [Documentation for Pre-Trained Models ](@ref omz_models_group_intel)                                   | Documentation for the pre-trained models available in the [Open Model Zoo repo](https://github.com/openvinotoolkit/open_model_zoo)  |
 
 **Could Be Optionally Installed**
 
diff --git a/docs/install_guides/installing-openvino-raspbian.md b/docs/install_guides/installing-openvino-raspbian.md
index 61cff12e424760..338beda73c813a 100644
--- a/docs/install_guides/installing-openvino-raspbian.md
+++ b/docs/install_guides/installing-openvino-raspbian.md
@@ -28,7 +28,7 @@ The OpenVINO toolkit for Raspbian OS is an archive with pre-installed header fil
 
 > **NOTE**:
 > * The package does not include the [Model Optimizer](../MO_DG/Deep_Learning_Model_Optimizer_DevGuide.md). To convert models to Intermediate Representation (IR), you need to install it separately to your host machine.
-> * The package does not include the Open Model Zoo demo applications. You can download them separately from the [Open Models Zoo repository](https://github.com/opencv/open_model_zoo).
+> * The package does not include the Open Model Zoo demo applications. You can download them separately from the [Open Models Zoo repository](https://github.com/openvinotoolkit/open_model_zoo).
 
 ## Development and Target Platforms
 
@@ -166,7 +166,7 @@ Read the next topic if you want to learn more about OpenVINO workflow for Raspbe
 
 If you want to use your model for inference, the model must be converted to the .bin and .xml Intermediate Representation (IR) files that are used as input by Inference Engine. OpenVINO™ toolkit support on Raspberry Pi only includes the Inference Engine module of the Intel® Distribution of OpenVINO™ toolkit. The Model Optimizer is not supported on this platform. To get the optimized models you can use one of the following options:
 
-* Download public and Intel's pre-trained models from the [Open Model Zoo](https://github.com/opencv/open_model_zoo) using [Model Downloader tool](@ref omz_tools_downloader).
+* Download public and Intel's pre-trained models from the [Open Model Zoo](https://github.com/openvinotoolkit/open_model_zoo) using [Model Downloader tool](@ref omz_tools_downloader).
 
    For more information on pre-trained models, see [Pre-Trained Models Documentation](@ref omz_models_group_intel)
 
diff --git a/docs/install_guides/installing-openvino-windows.md b/docs/install_guides/installing-openvino-windows.md
index 054950292b6337..8e3b6ece81fb4d 100644
--- a/docs/install_guides/installing-openvino-windows.md
+++ b/docs/install_guides/installing-openvino-windows.md
@@ -62,7 +62,7 @@ The following components are installed by default:
 |[Inference Engine Samples](../IE_DG/Samples_Overview.md)                             |A set of simple console applications demonstrating how to use Intel's Deep Learning Inference Engine in your applications.  |
 | [Demos](@ref omz_demos)                                   | A set of console applications that demonstrate how you can use the Inference Engine in your applications to solve specific use-cases  |
 | Additional Tools                                   | A set of tools to work with your models including [Accuracy Checker utility](@ref omz_tools_accuracy_checker), [Post-Training Optimization Tool Guide](@ref pot_README), [Model Downloader](@ref omz_tools_downloader) and other  |
-| [Documentation for Pre-Trained Models ](@ref omz_models_group_intel)                                   | Documentation for the pre-trained models available in the [Open Model Zoo repo](https://github.com/opencv/open_model_zoo)  |
+| [Documentation for Pre-Trained Models ](@ref omz_models_group_intel)                                   | Documentation for the pre-trained models available in the [Open Model Zoo repo](https://github.com/openvinotoolkit/open_model_zoo)  |
 
 **Could Be Optionally Installed**
 
diff --git a/docs/install_guides/installing-openvino-yum.md b/docs/install_guides/installing-openvino-yum.md
index c326cb93a0f6c1..ae34d202293b4f 100644
--- a/docs/install_guides/installing-openvino-yum.md
+++ b/docs/install_guides/installing-openvino-yum.md
@@ -16,7 +16,7 @@ The following components are installed with the OpenVINO runtime package:
 |-----------|------------|
 | [Inference Engine](../IE_DG/Deep_Learning_Inference_Engine_DevGuide.md)| The engine that runs a deep learning model. It includes a set of libraries for an easy inference integration into your applications. |
 | [OpenCV*](https://docs.opencv.org/master/) | OpenCV* community version compiled for Intel® hardware. |
-| Deep Learning Stream (DL Streamer) | Streaming analytics framework, based on GStreamer, for constructing graphs of media analytics components. For the DL Streamer documentation, see [DL Streamer Samples](@ref gst_samples_README), [API Reference](https://openvinotoolkit.github.io/dlstreamer_gst/), [Elements](https://github.com/opencv/gst-video-analytics/wiki/Elements), [Tutorial](https://github.com/opencv/gst-video-analytics/wiki/DL%20Streamer%20Tutorial). |
+| Deep Learning Stream (DL Streamer) | Streaming analytics framework, based on GStreamer, for constructing graphs of media analytics components. For the DL Streamer documentation, see [DL Streamer Samples](@ref gst_samples_README), [API Reference](https://openvinotoolkit.github.io/dlstreamer_gst/), [Elements](https://github.com/openvinotoolkit/dlstreamer_gst/wiki/Elements), [Tutorial](https://github.com/openvinotoolkit/dlstreamer_gst/wiki/DL-Streamer-Tutorial). |
 
 ## Set up the Repository
 
diff --git a/docs/optimization_guide/dldt_optimization_guide.md b/docs/optimization_guide/dldt_optimization_guide.md
index 9ece7fec93a628..3d7b41e7f8ccef 100644
--- a/docs/optimization_guide/dldt_optimization_guide.md
+++ b/docs/optimization_guide/dldt_optimization_guide.md
@@ -142,7 +142,7 @@ CPU plugin completely relies on the Intel&reg; Math Kernel Library for Deep Neur
 
 The only hint you can get from that is how the major primitives are accelerated (and you cannot change this). For example, on the Core machines, you should see variations of the `jit_avx2` when inspecting the <a href="#performance-counters">internal inference performance counters</a> (and additional '_int8' postfix for [int8 inference](../IE_DG/Int8Inference.md)). If you are an advanced user, you can further trace the CPU execution with (see <a href="#vtune-examples">Intel&reg; VTune&trade;</a>).
 
-Internally, the Inference Engine has a threading abstraction level, which allows for compiling the [open source version](https://github.com/opencv/dldt) with either Intel&reg; Threading Building Blocks (Intel&reg; TBB) which is now default, or OpenMP* as an alternative parallelism solution. When using inference on the CPU, this is particularly important to align threading model with the rest of your application (and any third-party libraries that you use) to avoid oversubscription. For more information, see <a href="#note-on-app-level-threading">Note on the App-Level Threading</a> section.
+Internally, the Inference Engine has a threading abstraction level, which allows for compiling the [open source version](https://github.com/openvinotoolkit/openvino) with either Intel&reg; Threading Building Blocks (Intel&reg; TBB) which is now default, or OpenMP* as an alternative parallelism solution. When using inference on the CPU, this is particularly important to align threading model with the rest of your application (and any third-party libraries that you use) to avoid oversubscription. For more information, see <a href="#note-on-app-level-threading">Note on the App-Level Threading</a> section.
 
  Since R1 2019, the OpenVINO&trade; toolkit comes pre-compiled with Intel TBB,
  so any  OpenMP* API or environment settings (like `OMP_NUM_THREADS`) has no effect.
@@ -348,7 +348,7 @@ If you are building an app-level pipeline with third-party components like GStre
 - As explained in the <a href="#cpu-checklist">CPU Checklist</a> section, by default the Inference Engine uses Intel TBB as a parallel engine. Thus, any OpenVINO-internal threading (including CPU inference) uses the same threads pool, provided by the TBB. But there are also other threads in your application, so oversubscription is possible at the application level:
 - The rule of thumb is that you should try to have the overall number of active threads in your application equal to the number of cores in your machine. Keep in mind the spare core(s) that the OpenCL driver under the GPU plugin might also need.
 - One specific workaround to limit the number of threads for the Inference Engine is using the [CPU configuration options](../IE_DG/supported_plugins/CPU.md).
-- To avoid further oversubscription, use the same threading model in all modules/libraries that your application uses. Notice that third party components might bring their own threading. For example, using Inference Engine which is now compiled with the TBB by default might lead to [performance troubles](https://www.threadingbuildingblocks.org/docs/help/reference/appendices/known_issues/interoperability.html) when mixed in the same app with another computationally-intensive library, but compiled with OpenMP. You can try to compile the [open source version](https://github.com/opencv/dldt) of the Inference Engine to use the OpenMP as well. But notice that in general, the TBB offers much better composability, than other threading solutions.
+- To avoid further oversubscription, use the same threading model in all modules/libraries that your application uses. Notice that third party components might bring their own threading. For example, using Inference Engine which is now compiled with the TBB by default might lead to [performance troubles](https://www.threadingbuildingblocks.org/docs/help/reference/appendices/known_issues/interoperability.html) when mixed in the same app with another computationally-intensive library, but compiled with OpenMP. You can try to compile the [open source version](https://github.com/openvinotoolkit/openvino) of the Inference Engine to use the OpenMP as well. But notice that in general, the TBB offers much better composability, than other threading solutions.
 - If your code (or third party libraries) uses GNU OpenMP, the Intel&reg; OpenMP (if you have recompiled Inference Engine with that) must be initialized first. This can be achieved by linking your application with the Intel OpenMP instead of GNU OpenMP, or using `LD_PRELOAD` on Linux* OS.
 
 ### Letting the Inference Engine Accelerate Image Pre-processing/Conversion <a name="image-preprocessing"></a>
diff --git a/inference-engine/cmake/templates/InferenceEngineConfig.cmake.in b/inference-engine/cmake/templates/InferenceEngineConfig.cmake.in
index cb4c62442c5ed6..261edbf3d730f3 100644
--- a/inference-engine/cmake/templates/InferenceEngineConfig.cmake.in
+++ b/inference-engine/cmake/templates/InferenceEngineConfig.cmake.in
@@ -29,6 +29,11 @@
 # Common functions
 #
 
+if(NOT DEFINED CMAKE_FIND_PACKAGE_NAME)
+    set(CMAKE_FIND_PACKAGE_NAME InferenceEngine)
+    set(_need_package_name_reset ON)
+endif()
+
 # we have to use our own version of find_dependency because of support cmake 3.7
 macro(_ie_find_dependency dep)
     set(cmake_fd_quiet_arg)
@@ -138,3 +143,8 @@ unset(IE_PACKAGE_PREFIX_DIR)
 set_and_check(InferenceEngine_INCLUDE_DIRS "@PACKAGE_IE_INCLUDE_DIR@")
 
 check_required_components(${CMAKE_FIND_PACKAGE_NAME})
+
+if(_need_package_name_reset)
+    unset(CMAKE_FIND_PACKAGE_NAME)
+    unset(_need_package_name_reset)
+endif()
diff --git a/inference-engine/include/ie_transformations.hpp b/inference-engine/include/ie_transformations.hpp
index 0867b02972c471..62b92a995781f3 100644
--- a/inference-engine/include/ie_transformations.hpp
+++ b/inference-engine/include/ie_transformations.hpp
@@ -16,6 +16,7 @@
 namespace InferenceEngine {
 
 /**
+ * @deprecated Use InferenceEngine::lowLatency2 instead
  * @brief The transformation finds all TensorIterator layers in the network, processes all back
  * edges that describe a connection between Result and Parameter of the TensorIterator body,
  * and inserts ReadValue layer between Parameter and the next layers after this Parameter,
@@ -50,7 +51,6 @@ namespace InferenceEngine {
  *    network->infer (...)  // Using stored states, calculating new values for states.
  *
  * @param network A network to apply LowLatency transformation
- * *
  */
 
 INFERENCE_ENGINE_DEPRECATED("This transformation will be removed in 2023.1. "
@@ -84,7 +84,6 @@ INFERENCE_ENGINE_API_CPP(void) LowLatency(InferenceEngine::CNNNetwork& network);
           If "true", then the transformation inserts Constant before ReadValue operation.
           If "false, then the transformation leaves existed initializing subgraph for ReadValue operation.
  * Loop operation by a given number. Does not affect TensorIterators.
- * *
  */
 INFERENCE_ENGINE_API_CPP(void) lowLatency2(InferenceEngine::CNNNetwork& network,
                                            bool use_const_initializer = true);
diff --git a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/interpolate.hpp b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/interpolate.hpp
index 07e86512ee0ea9..541f3c4fa16a73 100644
--- a/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/interpolate.hpp
+++ b/inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/single_layer/interpolate.hpp
@@ -51,4 +51,27 @@ class InterpolateLayerTest : public testing::WithParamInterface<InterpolateLayer
     void SetUp() override;
 };
 
+//Interpolate-1 test
+typedef std::tuple<
+        InferenceEngine::Precision,        // Net precision
+        InferenceEngine::Precision,        // Input precision, output is the same
+        InferenceEngine::Layout,           // Input layout, output is the same
+        InferenceEngine::SizeVector,       // Input shapes
+        InferenceEngine::SizeVector,       // Target shapes
+        std::string,                       // InterpolateMode
+        ngraph::AxisSet,                   // Axes
+        bool,                              // AntiAlias
+        std::vector<size_t>,               // Pads
+        LayerTestsUtils::TargetDevice      // Device name
+> Interpolate1LayerTestParams;
+
+class Interpolate1LayerTest : public testing::WithParamInterface<Interpolate1LayerTestParams>,
+                             virtual public LayerTestsUtils::LayerTestsCommon {
+public:
+    static std::string getTestCaseName(testing::TestParamInfo<Interpolate1LayerTestParams> obj);
+
+protected:
+    void SetUp() override;
+};
+
 }  // namespace LayerTestsDefinitions
diff --git a/inference-engine/tests/functional/shared_test_classes/src/single_layer/interpolate.cpp b/inference-engine/tests/functional/shared_test_classes/src/single_layer/interpolate.cpp
index 37db7bdeeab173..77d22a1af3a5cb 100644
--- a/inference-engine/tests/functional/shared_test_classes/src/single_layer/interpolate.cpp
+++ b/inference-engine/tests/functional/shared_test_classes/src/single_layer/interpolate.cpp
@@ -94,4 +94,69 @@ void InterpolateLayerTest::SetUp() {
     const ngraph::ResultVector results{std::make_shared<ngraph::opset3::Result>(interpolate)};
     function = std::make_shared<ngraph::Function>(results, params, "interpolate");
 }
+
+//Interpolate-1
+
+std::string Interpolate1LayerTest::getTestCaseName(testing::TestParamInfo<Interpolate1LayerTestParams> obj) {
+    InferenceEngine::Precision netPrecision;
+    InferenceEngine::Precision dataPrecision;
+    InferenceEngine::Layout dataLayout;
+    InferenceEngine::SizeVector inputShapes, targetShapes;
+    std::string mode;
+    ngraph::AxisSet axes;
+    bool antialias;
+    std::vector<size_t> pads;
+    std::string targetDevice;
+    std::tie(netPrecision, dataPrecision, dataLayout, inputShapes, targetShapes,
+             mode, axes, antialias, pads, targetDevice) = obj.param;
+
+    std::ostringstream result;
+    result << "IS=" << CommonTestUtils::vec2str(inputShapes) << "_";
+    result << "TS=" << CommonTestUtils::vec2str(targetShapes) << "_";
+    result << "InterpolateMode=" << mode << "_";
+    result << "Antialias=" << antialias << "_";
+    result << "PB=" << CommonTestUtils::vec2str(pads) << "_";
+    result << "PE=" << CommonTestUtils::vec2str(pads) << "_";
+    result << "Axes=" << CommonTestUtils::vec2str(axes.to_vector()) << "_";
+    result << "netPRC=" << netPrecision.name() << "_";
+    result << "PRC=" << dataPrecision.name() << "_";
+    result << "Layout=" << dataLayout << "_";
+    result << "trgDev=" << targetDevice;
+    return result.str();
+}
+
+void Interpolate1LayerTest::SetUp() {
+    std::vector<size_t> inputShape, targetShape;
+    auto netPrecision = InferenceEngine::Precision::UNSPECIFIED;
+    InferenceEngine::Precision dataPrecision;
+    InferenceEngine::Layout dataLayout;
+    std::string mode;
+    ngraph::AxisSet axes;
+    bool antialias;
+    std::vector<size_t> pads;
+    std::tie(netPrecision, dataPrecision, dataLayout, inputShape, targetShape,
+             mode, axes, antialias, pads, targetDevice) = this->GetParam();
+
+    auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
+    auto params = ngraph::builder::makeParams(ngPrc, {inputShape});
+
+    auto sizesConst = ngraph::opset3::Constant(ngraph::element::Type_t::i64, {targetShape.size()}, targetShape);
+    auto sizesInput = std::make_shared<ngraph::opset3::Constant>(sizesConst);
+
+    bool align_corners = true;
+
+    ngraph::op::v0::InterpolateAttrs interpolateAttributes;
+    interpolateAttributes.axes = axes;
+    interpolateAttributes.mode = mode;
+    interpolateAttributes.align_corners = align_corners;
+    interpolateAttributes.antialias = antialias;
+    interpolateAttributes.pads_begin = pads;
+    interpolateAttributes.pads_end = pads;
+
+    auto interpolate = std::make_shared<ngraph::op::v0::Interpolate>(params[0], sizesInput, interpolateAttributes);
+
+    const ngraph::ResultVector results{std::make_shared<ngraph::opset1::Result>(interpolate)};
+    function = std::make_shared<ngraph::Function>(results, params, "interpolate");
+}
+
 }  // namespace LayerTestsDefinitions
diff --git a/tests/stress_tests/README.md b/tests/stress_tests/README.md
index 865dc22428ebe4..a99146ad180b4f 100644
--- a/tests/stress_tests/README.md
+++ b/tests/stress_tests/README.md
@@ -87,4 +87,4 @@ grep -rh ./MemCheckTests-logs -e ".*<model " | sed -e "s/.*<model /<model /" | s
 
 [VDP-shared-folders]: https://wiki.ith.intel.com/display/DLSDK/VDP+shared+folders
 [gtest-parallel]: https://github.com/google/gtest-parallel
-[open_model_zoo]: https://github.com/opencv/open_model_zoo
\ No newline at end of file
+[open_model_zoo]: https://github.com/openvinotoolkit/open_model_zoo
\ No newline at end of file
diff --git a/tests/stress_tests/scripts/get_testdata.py b/tests/stress_tests/scripts/get_testdata.py
index 77bf2c1a52de53..cae9de7bec032a 100755
--- a/tests/stress_tests/scripts/get_testdata.py
+++ b/tests/stress_tests/scripts/get_testdata.py
@@ -124,7 +124,7 @@ def main():
         if os.path.exists(str(omz_path)):
             shutil.rmtree(str(omz_path))
         cmd = 'git clone --single-branch --branch develop' \
-              ' https://github.com/opencv/open_model_zoo {omz_path}'.format(omz_path=omz_path)
+              ' https://github.com/openvinotoolkit/open_model_zoo {omz_path}'.format(omz_path=omz_path)
         run_in_subprocess(cmd)
 
     # prepare virtual environment and install requirements
diff --git a/tools/benchmark/README.md b/tools/benchmark/README.md
index 50caa430fd470f..16491d6c9b208c 100644
--- a/tools/benchmark/README.md
+++ b/tools/benchmark/README.md
@@ -189,4 +189,4 @@ Throughput: 73.28 FPS
 ## See Also
 * [Using Inference Engine Samples](./docs/IE_DG/Samples_Overview.md)
 * [Model Optimizer](./docs/MO_DG/Deep_Learning_Model_Optimizer_DevGuide.md)
-* [Model Downloader](https://github.com/opencv/open_model_zoo/tree/2018/model_downloader)
+* [Model Downloader](https://github.com/openvinotoolkit/open_model_zoo/tree/2018/model_downloader)

From 27ae3ec433bccd7972b492923133e8b2ea83a3f6 Mon Sep 17 00:00:00 2001
From: Anton Chetverikov <Anton.Chetverikov@intel.com>
Date: Tue, 22 Jun 2021 18:28:37 +0300
Subject: [PATCH 164/178] Add documentation on how to convert RCAN model to
 master branch (#6180)

* Add documentation for PyTorch RCAN model

* Update instruction

* Add new file to list

* Update instruction

* Update instruction

* cherry-pick changes from release branch

* Apply review feedback

* Fix merge conflict
---
 .../Convert_Model_From_PyTorch.md             |  1 +
 .../pytorch_specific/Convert_RCAN.md          | 31 +++++++++++++++++++
 docs/doxygen/ie_docs.xml                      |  1 +
 3 files changed, 33 insertions(+)
 create mode 100644 docs/MO_DG/prepare_model/convert_model/pytorch_specific/Convert_RCAN.md

diff --git a/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_PyTorch.md b/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_PyTorch.md
index a03df559291a06..086d1de2e20324 100644
--- a/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_PyTorch.md
+++ b/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_PyTorch.md
@@ -25,6 +25,7 @@ It is not a full list of models that can be converted to ONNX\* and to IR.
 * F3Net topology can be converted using [Convert PyTorch\* F3Net to the IR](pytorch_specific/Convert_F3Net.md) instruction.
 * QuartzNet topologies from [NeMo project](https://github.com/NVIDIA/NeMo) can be converted using [Convert PyTorch\* QuartzNet to the IR](pytorch_specific/Convert_QuartzNet.md) instruction.
 * YOLACT topology can be converted using [Convert PyTorch\* YOLACT to the IR](pytorch_specific/Convert_YOLACT.md) instruction.
+* [RCAN](https://github.com/yulunzhang/RCAN) topologies can be converted using [Convert PyTorch\* RCAN to the IR](pytorch_specific/Convert_RCAN.md) instruction.
 
 ## Export PyTorch\* Model to ONNX\* Format <a name="export-to-onnx"></a>
 
diff --git a/docs/MO_DG/prepare_model/convert_model/pytorch_specific/Convert_RCAN.md b/docs/MO_DG/prepare_model/convert_model/pytorch_specific/Convert_RCAN.md
new file mode 100644
index 00000000000000..8fdefe128a6ecb
--- /dev/null
+++ b/docs/MO_DG/prepare_model/convert_model/pytorch_specific/Convert_RCAN.md
@@ -0,0 +1,31 @@
+# Convert PyTorch* RCAN to the Intermediate Representation {#openvino_docs_MO_DG_prepare_model_convert_model_pytorch_specific_Convert_RCAN}
+
+[RCAN](https://github.com/yulunzhang/RCAN): Image Super-Resolution Using Very Deep Residual Channel Attention Networks
+
+## Download and Convert the Model to ONNX*
+
+To download the pre-trained model or train the model yourself, refer to the
+[instruction](https://github.com/yulunzhang/RCAN/blob/master/README.md) in the RCAN model repository. Firstly,
+convert the model to ONNX\* format. Create and run the script with the following content in the root
+directory of the model repository:
+```python
+from argparse import Namespace
+
+import torch
+
+from RCAN_TestCode.code.model.rcan import RCAN
+
+config = Namespace(n_feats=64, n_resblocks=4, n_resgroups=2, reduction=16, scale=[2], data_train='DIV2K', res_scale=1,
+                   n_colors=3, rgb_range=255)
+net = RCAN(config)
+net.eval()
+dummy_input = torch.randn(1, 3, 360, 640)
+torch.onnx.export(net, dummy_input, 'RCAN.onnx')
+```
+The script generates the ONNX\* model file RCAN.onnx. You can find more information about model parameters (`n_resblocks`, `n_resgroups`, and others) in the model repository and use different values of them. The model conversion was tested with the repository hash commit `3339ebc59519c3bb2b5719b87dd36515ec7f3ba7`.
+
+## Convert ONNX* RCAN Model to IR
+
+```sh
+./mo.py --input_model RCAN.onnx
+```
diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index 008b0d051646b1..7e426edde06d50 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -60,6 +60,7 @@ limitations under the License.
                             <tab type="user" title="Convert PyTorch* RNN-T Model " url="@ref openvino_docs_MO_DG_prepare_model_convert_model_pytorch_specific_Convert_RNNT"/>
                             <tab type="user" title="Convert PyTorch* YOLACT Model" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_pytorch_specific_Convert_YOLACT"/>
                             <tab type="user" title="Convert PyTorch* F3Net Model" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_pytorch_specific_Convert_F3Net"/>
+                            <tab type="user" title="Convert PyTorch* RCAN Model" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_pytorch_specific_Convert_RCAN">
                             </tab>
                         </tab>
                         <tab type="user" title="Model Optimizations Techniques" url="@ref openvino_docs_MO_DG_prepare_model_Model_Optimization_Techniques"/>

From e00cee2fc616cb80babe1ef09ea96f65e687fd74 Mon Sep 17 00:00:00 2001
From: Anna Khakimova <anna.khakimova@intel.com>
Date: Tue, 22 Jun 2021 18:46:23 +0300
Subject: [PATCH 165/178] Pre-processing: Split and Merge kernels refactoring.
 (#6205)

* * Split and Merge kernel refactoring

* * SFINAE: replace condition compilation macro with std::enable_if
---
 .../ie_preprocess_gapi_kernels_neon.cpp       |  75 +-
 .../ie_preprocess_gapi_kernels_neon.hpp       | 101 +--
 .../ie_preprocess_gapi_kernels_avx2.cpp       |  74 +-
 .../ie_preprocess_gapi_kernels_avx2.hpp       | 100 +--
 .../ie_preprocess_gapi_kernels_avx512.cpp     |  75 +-
 .../ie_preprocess_gapi_kernels_avx512.hpp     |  95 +--
 .../ie_preprocess_gapi_kernels_sse42.cpp      | 111 +--
 .../ie_preprocess_gapi_kernels_sse42.hpp      |  94 +--
 .../ie_preprocess_gapi_kernels.cpp            | 767 ++++++------------
 .../ie_preprocess_gapi_kernels_simd_impl.hpp  | 478 ++++-------
 .../fluid_preproc/common/fluid_tests.cpp      |  20 +
 11 files changed, 564 insertions(+), 1426 deletions(-)

diff --git a/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.cpp b/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.cpp
index 8290fda25ed8ea..b0b3281748ed53 100644
--- a/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.cpp
+++ b/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.cpp
@@ -29,67 +29,6 @@ namespace InferenceEngine {
 namespace gapi {
 namespace kernels {
 namespace neon {
-
-void mergeRow_8UC2(const uint8_t in0[], const uint8_t in1[],
-                   uint8_t out[], int length) {
-    mergeRow_8UC2_Impl(in0, in1, out, length);
-}
-
-void mergeRow_8UC3(const uint8_t in0[], const uint8_t in1[],
-                   const uint8_t in2[], uint8_t out[], int length) {
-    mergeRow_8UC3_Impl(in0, in1, in2, out, length);
-}
-
-void mergeRow_8UC4(const uint8_t in0[], const uint8_t in1[], const uint8_t in2[],
-                   const uint8_t in3[], uint8_t out[], int length) {
-    mergeRow_8UC4_Impl(in0, in1, in2, in3, out, length);
-}
-
-void mergeRow_32FC2(const float in0[], const float in1[],
-                    float out[], int length) {
-    mergeRow_32FC2_Impl(in0, in1, out, length);
-}
-
-void mergeRow_32FC3(const float in0[], const float in1[], const float in2[],
-                    float out[], int length) {
-    mergeRow_32FC3_Impl(in0, in1, in2, out, length);
-}
-
-void mergeRow_32FC4(const float in0[], const float in1[],
-                    const float in2[], const float in3[],
-                    float out[], int length) {
-    mergeRow_32FC4_Impl(in0, in1, in2, in3, out, length);
-}
-
-void splitRow_8UC2(const uint8_t in[], uint8_t out0[],
-                   uint8_t out1[], int length) {
-    splitRow_8UC2_Impl(in, out0, out1, length);
-}
-
-void splitRow_8UC3(const uint8_t in[], uint8_t out0[],
-                   uint8_t out1[], uint8_t out2[], int length) {
-    splitRow_8UC3_Impl(in, out0, out1, out2, length);
-}
-
-void splitRow_8UC4(const uint8_t in[], uint8_t out0[], uint8_t out1[],
-                   uint8_t out2[], uint8_t out3[], int length) {
-    splitRow_8UC4_Impl(in, out0, out1, out2, out3, length);
-}
-
-void splitRow_32FC2(const float in[], float out0[], float out1[], int length) {
-    splitRow_32FC2_Impl(in, out0, out1, length);
-}
-
-void splitRow_32FC3(const float in[], float out0[], float out1[],
-                    float out2[], int length) {
-    splitRow_32FC3_Impl(in, out0, out1, out2, length);
-}
-
-void splitRow_32FC4(const float in[], float out0[], float out1[],
-                    float out2[], float out3[], int length) {
-    splitRow_32FC4_Impl(in, out0, out1, out2, out3, length);
-}
-
 void calcRowArea_8U(uchar dst[], const uchar *src[], const Size& inSz,
                     const Size& outSz, Q0_16 yalpha, const MapperUnit8U &ymap,
                     int xmaxdf, const short xindex[], const Q0_16 xalpha[],
@@ -693,6 +632,20 @@ template void nv12ToRgbRowImpl(neon_tag, const uint8_t** y_rows, const uint8_t*
 
 template void i420ToRgbRowImpl(neon_tag, const uint8_t** y_rows, const uint8_t* u_row,
                                const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+template void splitRowImpl<neon_tag, uint8_t, 2>(neon_tag, const uint8_t* in, std::array<uint8_t*, 2>& outs, const int length);
+template void splitRowImpl<neon_tag, float, 2>(neon_tag, const float* in, std::array<float*, 2>& outs, const int length);
+template void splitRowImpl<neon_tag, uint8_t, 3>(neon_tag, const uint8_t* in, std::array<uint8_t*, 3>& outs, const int length);
+template void splitRowImpl<neon_tag, float, 3>(neon_tag, const float* in, std::array<float*, 3>& outs, const int length);
+template void splitRowImpl<neon_tag, uint8_t, 4>(neon_tag, const uint8_t* in, std::array<uint8_t*, 4>& outs, const int length);
+template void splitRowImpl<neon_tag, float, 4>(neon_tag, const float* in, std::array<float*, 4>& outs, const int length);
+
+template void mergeRowImpl<neon_tag, uint8_t, 2>(neon_tag, const std::array<const uint8_t*, 2>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<neon_tag, float, 2>(neon_tag, const std::array<const float*, 2>& ins, float* out, const int length);
+template void mergeRowImpl<neon_tag, uint8_t, 3>(neon_tag, const std::array<const uint8_t*, 3>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<neon_tag, float, 3>(neon_tag, const std::array<const float*, 3>& ins, float* out, const int length);
+template void mergeRowImpl<neon_tag, uint8_t, 4>(neon_tag, const std::array<const uint8_t*, 4>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<neon_tag, float, 4>(neon_tag, const std::array<const float*, 4>& ins, float* out, const int length);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.hpp b/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.hpp
index 4e0b82a62596c7..5f6d1b783db2e8 100644
--- a/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.hpp
+++ b/inference-engine/src/preprocessing/arm_neon/ie_preprocess_gapi_kernels_neon.hpp
@@ -92,87 +92,6 @@ void calcRowLinear_32F(float       *dst[],
                        const Size& inSz,
                        const Size& outSz,
                                int lpi);
-
-//----------------------------------------------------------------------
-
-void mergeRow_8UC2(const uint8_t in0[],
-                   const uint8_t in1[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_8UC3(const uint8_t in0[],
-                   const uint8_t in1[],
-                   const uint8_t in2[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_8UC4(const uint8_t in0[],
-                   const uint8_t in1[],
-                   const uint8_t in2[],
-                   const uint8_t in3[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_32FC2(const float in0[],
-                    const float in1[],
-                          float out[],
-                            int length);
-
-void mergeRow_32FC3(const float in0[],
-                    const float in1[],
-                    const float in2[],
-                          float out[],
-                            int length);
-
-void mergeRow_32FC4(const float in0[],
-                    const float in1[],
-                    const float in2[],
-                    const float in3[],
-                          float out[],
-                            int length);
-
-void splitRow_8UC2(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                             int length);
-
-void splitRow_8UC3(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                         uint8_t out2[],
-                             int length);
-
-void splitRow_8UC4(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                         uint8_t out2[],
-                         uint8_t out3[],
-                             int length);
-
-void splitRow_32FC2(const float in[],
-                          float out0[],
-                          float out1[],
-                            int length);
-
-void splitRow_32FC3(const float in[],
-                          float out0[],
-                          float out1[],
-                          float out2[],
-                            int length);
-
-void splitRow_32FC4(const float in[],
-                          float out0[],
-                          float out1[],
-                          float out2[],
-                          float out3[],
-                            int length);
-
-void calculate_i420_to_rgb(const  uchar **srcY,
-                           const  uchar *srcU,
-                           const  uchar *srcV,
-                                  uchar **dstRGBx,
-                                    int width);
-
 }  // namespace neon
 
 template<typename isa_tag_t, typename T>
@@ -192,6 +111,26 @@ void i420ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* u_row,
 
 extern template void i420ToRgbRowImpl(neon_tag, const uint8_t** y_rows, const uint8_t* u_row,
                                       const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+template<typename isa_tag_t, typename T, int chs>
+void splitRowImpl(isa_tag_t, const T* in, std::array<T*, chs>& outs, const int length);
+
+extern template void splitRowImpl<neon_tag, uint8_t, 2>(neon_tag, const uint8_t* in, std::array<uint8_t*, 2>& outs, const int length);
+extern template void splitRowImpl<neon_tag, float, 2>(neon_tag, const float* in, std::array<float*, 2>& outs, const int length);
+extern template void splitRowImpl<neon_tag, uint8_t, 3>(neon_tag, const uint8_t* in, std::array<uint8_t*, 3>& outs, const int length);
+extern template void splitRowImpl<neon_tag, float, 3>(neon_tag, const float* in, std::array<float*, 3>& outs, const int length);
+extern template void splitRowImpl<neon_tag, uint8_t, 4>(neon_tag, const uint8_t* in, std::array<uint8_t*, 4>& outs, const int length);
+extern template void splitRowImpl<neon_tag, float, 4>(neon_tag, const float* in, std::array<float*, 4>& outs, const int length);
+
+template<typename isa_tag_t, typename T, int chs>
+void mergeRowImpl(isa_tag_t, const std::array<const T*, chs>& ins, T* out, const int length);
+
+extern template void mergeRowImpl<neon_tag, uint8_t, 2>(neon_tag, const std::array<const uint8_t*, 2>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<neon_tag, float, 2>(neon_tag, const std::array<const float*, 2>& ins, float* out, const int length);
+extern template void mergeRowImpl<neon_tag, uint8_t, 3>(neon_tag, const std::array<const uint8_t*, 3>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<neon_tag, float, 3>(neon_tag, const std::array<const float*, 3>& ins, float* out, const int length);
+extern template void mergeRowImpl<neon_tag, uint8_t, 4>(neon_tag, const std::array<const uint8_t*, 4>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<neon_tag, float, 4>(neon_tag, const std::array<const float*, 4>& ins, float* out, const int length);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.cpp b/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.cpp
index aefd2e2dfc33ce..d5b3d2664f465f 100644
--- a/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.cpp
+++ b/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.cpp
@@ -47,66 +47,6 @@ namespace kernels {
 
 namespace avx {
 
-void mergeRow_8UC2(const uint8_t in0[], const uint8_t in1[],
-                   uint8_t out[], int length) {
-    mergeRow_8UC2_Impl(in0, in1, out, length);
-}
-
-void mergeRow_8UC3(const uint8_t in0[], const uint8_t in1[],
-                   const uint8_t in2[], uint8_t out[], int length) {
-    mergeRow_8UC3_Impl(in0, in1, in2, out, length);
-}
-
-void mergeRow_8UC4(const uint8_t in0[], const uint8_t in1[], const uint8_t in2[],
-                   const uint8_t in3[], uint8_t out[], int length) {
-    mergeRow_8UC4_Impl(in0, in1, in2, in3, out, length);
-}
-
-void mergeRow_32FC2(const float in0[], const float in1[],
-                    float out[], int length) {
-    mergeRow_32FC2_Impl(in0, in1, out, length);
-}
-
-void mergeRow_32FC3(const float in0[], const float in1[], const float in2[],
-                    float out[], int length) {
-    mergeRow_32FC3_Impl(in0, in1, in2, out, length);
-}
-
-void mergeRow_32FC4(const float in0[], const float in1[],
-                    const float in2[], const float in3[],
-                    float out[], int length) {
-    mergeRow_32FC4_Impl(in0, in1, in2, in3, out, length);
-}
-
-void splitRow_8UC2(const uint8_t in[], uint8_t out0[],
-                   uint8_t out1[], int length) {
-    splitRow_8UC2_Impl(in, out0, out1, length);
-}
-
-void splitRow_8UC3(const uint8_t in[], uint8_t out0[],
-                   uint8_t out1[], uint8_t out2[], int length) {
-    splitRow_8UC3_Impl(in, out0, out1, out2, length);
-}
-
-void splitRow_8UC4(const uint8_t in[], uint8_t out0[], uint8_t out1[],
-                   uint8_t out2[], uint8_t out3[], int length) {
-    splitRow_8UC4_Impl(in, out0, out1, out2, out3, length);
-}
-
-void splitRow_32FC2(const float in[], float out0[], float out1[], int length) {
-    splitRow_32FC2_Impl(in, out0, out1, length);
-}
-
-void splitRow_32FC3(const float in[], float out0[], float out1[],
-                    float out2[], int length) {
-    splitRow_32FC3_Impl(in, out0, out1, out2, length);
-}
-
-void splitRow_32FC4(const float in[], float out0[], float out1[],
-                    float out2[], float out3[], int length) {
-    splitRow_32FC4_Impl(in, out0, out1, out2, out3, length);
-}
-
 void calcRowArea_8U(uchar dst[], const uchar *src[], const Size& inSz,
                     const Size& outSz, Q0_16 yalpha, const MapperUnit8U &ymap,
                     int xmaxdf, const short xindex[], const Q0_16 xalpha[],
@@ -562,6 +502,20 @@ template void nv12ToRgbRowImpl(avx2_tag, const uint8_t** y_rows, const uint8_t*
 
 template void i420ToRgbRowImpl(avx2_tag, const uint8_t** y_rows, const uint8_t* u_row,
                                const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+template void splitRowImpl<avx2_tag, uint8_t, 2>(avx2_tag, const uint8_t* in, std::array<uint8_t*, 2>& outs, const int length);
+template void splitRowImpl<avx2_tag, float, 2>(avx2_tag, const float* in, std::array<float*, 2>& outs, const int length);
+template void splitRowImpl<avx2_tag, uint8_t, 3>(avx2_tag, const uint8_t* in, std::array<uint8_t*, 3>& outs, const int length);
+template void splitRowImpl<avx2_tag, float, 3>(avx2_tag, const float* in, std::array<float*, 3>& outs, const int length);
+template void splitRowImpl<avx2_tag, uint8_t, 4>(avx2_tag, const uint8_t* in, std::array<uint8_t*, 4>& outs, const int length);
+template void splitRowImpl<avx2_tag, float, 4>(avx2_tag, const float* in, std::array<float*, 4>& outs, const int length);
+
+template void mergeRowImpl<avx2_tag, uint8_t, 2>(avx2_tag, const std::array<const uint8_t*, 2>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<avx2_tag, float, 2>(avx2_tag, const std::array<const float*, 2>& ins, float* out, const int length);
+template void mergeRowImpl<avx2_tag, uint8_t, 3>(avx2_tag, const std::array<const uint8_t*, 3>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<avx2_tag, float, 3>(avx2_tag, const std::array<const float*, 3>& ins, float* out, const int length);
+template void mergeRowImpl<avx2_tag, uint8_t, 4>(avx2_tag, const std::array<const uint8_t*, 4>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<avx2_tag, float, 4>(avx2_tag, const std::array<const float*, 4>& ins, float* out, const int length);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.hpp b/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.hpp
index 512121b0f3dc76..5784abe0ea1fca 100644
--- a/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.hpp
+++ b/inference-engine/src/preprocessing/cpu_x86_avx2/ie_preprocess_gapi_kernels_avx2.hpp
@@ -106,84 +106,8 @@ void calcRowLinear_32F(float *dst[],
                        const Size & inSz,
                        const Size & outSz,
                        int    lpi);
-
-//----------------------------------------------------------------------
-
-
-void mergeRow_8UC2(const uint8_t in0[],
-                   const uint8_t in1[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_8UC3(const uint8_t in0[],
-                   const uint8_t in1[],
-                   const uint8_t in2[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_8UC4(const uint8_t in0[],
-                   const uint8_t in1[],
-                   const uint8_t in2[],
-                   const uint8_t in3[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_32FC2(const float in0[],
-                    const float in1[],
-                          float out[],
-                            int length);
-
-void mergeRow_32FC3(const float in0[],
-                    const float in1[],
-                    const float in2[],
-                          float out[],
-                            int length);
-
-void mergeRow_32FC4(const float in0[],
-                    const float in1[],
-                    const float in2[],
-                    const float in3[],
-                          float out[],
-                            int length);
-
-void splitRow_8UC2(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                             int length);
-
-void splitRow_8UC3(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                         uint8_t out2[],
-                             int length);
-
-void splitRow_8UC4(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                         uint8_t out2[],
-                         uint8_t out3[],
-                             int length);
-
-void splitRow_32FC2(const float in[],
-                          float out0[],
-                          float out1[],
-                            int length);
-
-void splitRow_32FC3(const float in[],
-                          float out0[],
-                          float out1[],
-                          float out2[],
-                            int length);
-
-void splitRow_32FC4(const float in[],
-                          float out0[],
-                          float out1[],
-                          float out2[],
-                          float out3[],
-                            int length);
 }  // namespace avx
 
-
 template<typename isa_tag_t, typename T>
 void chanToPlaneRowImpl(isa_tag_t, const T* in, const int chan, const int chs, T* out, const int length);
 
@@ -192,7 +116,7 @@ extern template void chanToPlaneRowImpl(avx2_tag, const float*   in, const int c
 
 template<typename isa_tag_t>
 void nv12ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* uv_row,
-                             uint8_t** out_rows, const int buf_width);
+                      uint8_t** out_rows, const int buf_width);
 
 extern template void nv12ToRgbRowImpl(avx2_tag, const uint8_t** y_rows,
                                       const uint8_t* uv_row, uint8_t** out_rows,
@@ -200,10 +124,30 @@ extern template void nv12ToRgbRowImpl(avx2_tag, const uint8_t** y_rows,
 
 template<typename isa_tag_t>
 void i420ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* u_row,
-                             const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+                      const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
 
 extern template void i420ToRgbRowImpl(avx2_tag, const uint8_t** y_rows, const uint8_t* u_row,
                                       const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+template<typename isa_tag_t, typename T, int chs>
+void splitRowImpl(isa_tag_t, const T* in, std::array<T*, chs>& outs, const int length);
+
+extern template void splitRowImpl<avx2_tag, uint8_t, 2>(avx2_tag, const uint8_t* in, std::array<uint8_t*, 2>& outs, const int length);
+extern template void splitRowImpl<avx2_tag, float, 2>(avx2_tag, const float* in, std::array<float*, 2>& outs, const int length);
+extern template void splitRowImpl<avx2_tag, uint8_t, 3>(avx2_tag, const uint8_t* in, std::array<uint8_t*, 3>& outs, const int length);
+extern template void splitRowImpl<avx2_tag, float, 3>(avx2_tag, const float* in, std::array<float*, 3>& outs, const int length);
+extern template void splitRowImpl<avx2_tag, uint8_t, 4>(avx2_tag, const uint8_t* in, std::array<uint8_t*, 4>& outs, const int length);
+extern template void splitRowImpl<avx2_tag, float, 4>(avx2_tag, const float* in, std::array<float*, 4>& outs, const int length);
+
+template<typename isa_tag_t, typename T, int chs>
+void mergeRowImpl(isa_tag_t, const std::array<const T*, chs>& ins, T* out, const int length);
+
+extern template void mergeRowImpl<avx2_tag, uint8_t, 2>(avx2_tag, const std::array<const uint8_t*, 2>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<avx2_tag, float, 2>(avx2_tag, const std::array<const float*, 2>& ins, float* out, const int length);
+extern template void mergeRowImpl<avx2_tag, uint8_t, 3>(avx2_tag, const std::array<const uint8_t*, 3>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<avx2_tag, float, 3>(avx2_tag, const std::array<const float*, 3>& ins, float* out, const int length);
+extern template void mergeRowImpl<avx2_tag, uint8_t, 4>(avx2_tag, const std::array<const uint8_t*, 4>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<avx2_tag, float, 4>(avx2_tag, const std::array<const float*, 4>& ins, float* out, const int length);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.cpp b/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.cpp
index 78b74f532d62a6..ec8b52d3aad2df 100644
--- a/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.cpp
+++ b/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.cpp
@@ -41,66 +41,6 @@ namespace kernels {
 
 namespace avx512 {
 
-void mergeRow_8UC2(const uint8_t in0[], const uint8_t in1[],
-                   uint8_t out[], int length) {
-    mergeRow_8UC2_Impl(in0, in1, out, length);
-}
-
-void mergeRow_8UC3(const uint8_t in0[], const uint8_t in1[],
-                   const uint8_t in2[], uint8_t out[], int length) {
-    mergeRow_8UC3_Impl(in0, in1, in2, out, length);
-}
-
-void mergeRow_8UC4(const uint8_t in0[], const uint8_t in1[], const uint8_t in2[],
-                   const uint8_t in3[], uint8_t out[], int length) {
-    mergeRow_8UC4_Impl(in0, in1, in2, in3, out, length);
-}
-
-void mergeRow_32FC2(const float in0[], const float in1[],
-                    float out[], int length) {
-    mergeRow_32FC2_Impl(in0, in1, out, length);
-}
-
-void mergeRow_32FC3(const float in0[], const float in1[], const float in2[],
-                    float out[], int length) {
-    mergeRow_32FC3_Impl(in0, in1, in2, out, length);
-}
-
-void mergeRow_32FC4(const float in0[], const float in1[],
-                    const float in2[], const float in3[],
-                    float out[], int length) {
-    mergeRow_32FC4_Impl(in0, in1, in2, in3, out, length);
-}
-
-void splitRow_8UC2(const uint8_t in[], uint8_t out0[],
-                   uint8_t out1[], int length) {
-    splitRow_8UC2_Impl(in, out0, out1, length);
-}
-
-void splitRow_8UC3(const uint8_t in[], uint8_t out0[],
-                   uint8_t out1[], uint8_t out2[], int length) {
-    splitRow_8UC3_Impl(in, out0, out1, out2, length);
-}
-
-void splitRow_8UC4(const uint8_t in[], uint8_t out0[], uint8_t out1[],
-                   uint8_t out2[], uint8_t out3[], int length) {
-    splitRow_8UC4_Impl(in, out0, out1, out2, out3, length);
-}
-
-void splitRow_32FC2(const float in[], float out0[], float out1[], int length) {
-    splitRow_32FC2_Impl(in, out0, out1, length);
-}
-
-void splitRow_32FC3(const float in[], float out0[], float out1[],
-                    float out2[], int length) {
-    splitRow_32FC3_Impl(in, out0, out1, out2, length);
-}
-
-void splitRow_32FC4(const float in[], float out0[], float out1[],
-                    float out2[], float out3[], int length) {
-    splitRow_32FC4_Impl(in, out0, out1, out2, out3, length);
-}
-
 void calcRowArea_8U(uchar dst[], const uchar *src[], const Size& inSz,
                     const Size& outSz, Q0_16 yalpha, const MapperUnit8U &ymap,
                     int xmaxdf, const short xindex[], const Q0_16 xalpha[],
@@ -632,7 +572,6 @@ void calcRowLinear_32F(float *dst[],
                                int  lpi) {
     calcRowLinear_32FC1(dst, src0, src1, alpha, mapsx, beta, inSz, outSz, lpi);
 }
-
 }  // namespace avx512
 
 template void chanToPlaneRowImpl(avx512_tag, const uint8_t* in, const int chan, const int chs, uint8_t* out, const int length);
@@ -642,6 +581,20 @@ template void nv12ToRgbRowImpl(avx512_tag, const uint8_t** y_rows, const uint8_t
 
 template void i420ToRgbRowImpl(avx512_tag, const uint8_t** y_rows, const uint8_t* u_row,
                                const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+template void splitRowImpl<avx512_tag, uint8_t, 2>(avx512_tag, const uint8_t* in, std::array<uint8_t*, 2>& outs, const int length);
+template void splitRowImpl<avx512_tag, float, 2>(avx512_tag, const float* in, std::array<float*, 2>& outs, const int length);
+template void splitRowImpl<avx512_tag, uint8_t, 3>(avx512_tag, const uint8_t* in, std::array<uint8_t*, 3>& outs, const int length);
+template void splitRowImpl<avx512_tag, float, 3>(avx512_tag, const float* in, std::array<float*, 3>& outs, const int length);
+template void splitRowImpl<avx512_tag, uint8_t, 4>(avx512_tag, const uint8_t* in, std::array<uint8_t*, 4>& outs, const int length);
+template void splitRowImpl<avx512_tag, float, 4>(avx512_tag, const float* in, std::array<float*, 4>& outs, const int length);
+
+template void mergeRowImpl<avx512_tag, uint8_t, 2>(avx512_tag, const std::array<const uint8_t*, 2>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<avx512_tag, float, 2>(avx512_tag, const std::array<const float*, 2>& ins, float* out, const int length);
+template void mergeRowImpl<avx512_tag, uint8_t, 3>(avx512_tag, const std::array<const uint8_t*, 3>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<avx512_tag, float, 3>(avx512_tag, const std::array<const float*, 3>& ins, float* out, const int length);
+template void mergeRowImpl<avx512_tag, uint8_t, 4>(avx512_tag, const std::array<const uint8_t*, 4>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<avx512_tag, float, 4>(avx512_tag, const std::array<const float*, 4>& ins, float* out, const int length);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.hpp b/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.hpp
index 8d2778781a35b4..23da40a80892a9 100644
--- a/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.hpp
+++ b/inference-engine/src/preprocessing/cpu_x86_avx512/ie_preprocess_gapi_kernels_avx512.hpp
@@ -106,83 +106,8 @@ void calcRowLinear_32F(float *dst[],
                        const Size & inSz,
                        const Size & outSz,
                        int    lpi);
-
-//----------------------------------------------------------------------
-
-void mergeRow_8UC2(const uint8_t in0[],
-                   const uint8_t in1[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_8UC3(const uint8_t in0[],
-                   const uint8_t in1[],
-                   const uint8_t in2[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_8UC4(const uint8_t in0[],
-                   const uint8_t in1[],
-                   const uint8_t in2[],
-                   const uint8_t in3[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_32FC2(const float in0[],
-                    const float in1[],
-                          float out[],
-                            int length);
-
-void mergeRow_32FC3(const float in0[],
-                    const float in1[],
-                    const float in2[],
-                          float out[],
-                            int length);
-
-void mergeRow_32FC4(const float in0[],
-                    const float in1[],
-                    const float in2[],
-                    const float in3[],
-                          float out[],
-                            int length);
-
-void splitRow_8UC2(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                             int length);
-
-void splitRow_8UC3(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                         uint8_t out2[],
-                             int length);
-
-void splitRow_8UC4(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                         uint8_t out2[],
-                         uint8_t out3[],
-                             int length);
-
-void splitRow_32FC2(const float in[],
-                          float out0[],
-                          float out1[],
-                            int length);
-
-void splitRow_32FC3(const float in[],
-                          float out0[],
-                          float out1[],
-                          float out2[],
-                            int length);
-
-void splitRow_32FC4(const float in[],
-                          float out0[],
-                          float out1[],
-                          float out2[],
-                          float out3[],
-                            int length);
 }  // namespace avx512
 
-
 template<typename isa_tag_t, typename T>
 void chanToPlaneRowImpl(isa_tag_t, const T* in, const int chan, const int chs, T* out, const int length);
 
@@ -200,6 +125,26 @@ void i420ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* u_row,
 
 extern template void i420ToRgbRowImpl(avx512_tag, const uint8_t** y_rows, const uint8_t* u_row,
                                       const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+template<typename isa_tag_t, typename T, int chs>
+void splitRowImpl(isa_tag_t, const T* in, std::array<T*, chs>& outs, const int length);
+
+extern template void splitRowImpl<avx512_tag, uint8_t, 2>(avx512_tag, const uint8_t* in, std::array<uint8_t*, 2>& outs, const int length);
+extern template void splitRowImpl<avx512_tag, float, 2>(avx512_tag, const float* in, std::array<float*, 2>& outs, const int length);
+extern template void splitRowImpl<avx512_tag, uint8_t, 3>(avx512_tag, const uint8_t* in, std::array<uint8_t*, 3>& outs, const int length);
+extern template void splitRowImpl<avx512_tag, float, 3>(avx512_tag, const float* in, std::array<float*, 3>& outs, const int length);
+extern template void splitRowImpl<avx512_tag, uint8_t, 4>(avx512_tag, const uint8_t* in, std::array<uint8_t*, 4>& outs, const int length);
+extern template void splitRowImpl<avx512_tag, float, 4>(avx512_tag, const float* in, std::array<float*, 4>& outs, const int length);
+
+template<typename isa_tag_t, typename T, int chs>
+void mergeRowImpl(isa_tag_t, const std::array<const T*, chs>& ins, T* out, const int length);
+
+extern template void mergeRowImpl<avx512_tag, uint8_t, 2>(avx512_tag, const std::array<const uint8_t*, 2>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<avx512_tag, float, 2>(avx512_tag, const std::array<const float*, 2>& ins, float* out, const int length);
+extern template void mergeRowImpl<avx512_tag, uint8_t, 3>(avx512_tag, const std::array<const uint8_t*, 3>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<avx512_tag, float, 3>(avx512_tag, const std::array<const float*, 3>& ins, float* out, const int length);
+extern template void mergeRowImpl<avx512_tag, uint8_t, 4>(avx512_tag, const std::array<const uint8_t*, 4>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<avx512_tag, float, 4>(avx512_tag, const std::array<const float*, 4>& ins, float* out, const int length);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.cpp b/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.cpp
index 4823e33537141a..d3d448b9279dac 100644
--- a/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.cpp
+++ b/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.cpp
@@ -1267,103 +1267,6 @@ void calcRowArea_CVKL_U8_SSE42(const uchar  * src[],
 }
 
 #endif  // CVKL
-//------------------------------------------------------------------------------
-
-void mergeRow_8UC2(const uint8_t in0[],
-                   const uint8_t in1[],
-                         uint8_t out[],
-                             int length) {
-    mergeRow_8UC2_Impl(in0, in1, out, length);
-}
-
-void mergeRow_8UC3(const uint8_t in0[],
-                   const uint8_t in1[],
-                   const uint8_t in2[],
-                         uint8_t out[],
-                             int length) {
-    mergeRow_8UC3_Impl(in0, in1, in2, out, length);
-}
-
-void mergeRow_8UC4(const uint8_t in0[],
-                   const uint8_t in1[],
-                   const uint8_t in2[],
-                   const uint8_t in3[],
-                         uint8_t out[],
-                             int length) {
-    mergeRow_8UC4_Impl(in0, in1, in2, in3, out, length);
-}
-
-void mergeRow_32FC2(const float in0[],
-                    const float in1[],
-                          float out[],
-                            int length) {
-    mergeRow_32FC2_Impl(in0, in1, out, length);
-}
-
-void mergeRow_32FC3(const float in0[],
-                    const float in1[],
-                    const float in2[],
-                          float out[],
-                            int length) {
-    mergeRow_32FC3_Impl(in0, in1, in2, out, length);
-}
-
-void mergeRow_32FC4(const float in0[],
-                    const float in1[],
-                    const float in2[],
-                    const float in3[],
-                          float out[],
-                            int length) {
-    mergeRow_32FC4_Impl(in0, in1, in2, in3, out, length);
-}
-
-void splitRow_8UC2(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                             int length) {
-    splitRow_8UC2_Impl(in, out0, out1, length);
-}
-
-void splitRow_8UC3(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                         uint8_t out2[],
-                             int length) {
-    splitRow_8UC3_Impl(in, out0, out1, out2, length);
-}
-
-void splitRow_8UC4(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                         uint8_t out2[],
-                         uint8_t out3[],
-                             int length) {
-    splitRow_8UC4_Impl(in, out0, out1, out2, out3, length);
-}
-
-void splitRow_32FC2(const float in[],
-                          float out0[],
-                          float out1[],
-                            int length) {
-    splitRow_32FC2_Impl(in, out0, out1, length);
-}
-
-void splitRow_32FC3(const float in[],
-                          float out0[],
-                          float out1[],
-                          float out2[],
-                            int length) {
-    splitRow_32FC3_Impl(in, out0, out1, out2, length);
-}
-
-void splitRow_32FC4(const float in[],
-                          float out0[],
-                          float out1[],
-                          float out2[],
-                          float out3[],
-                            int length) {
-    splitRow_32FC4_Impl(in, out0, out1, out2, out3, length);
-}
 
 template void chanToPlaneRowImpl(sse42_tag, const uint8_t* in, const int chan, const int chs, uint8_t* out, const int length);
 template void chanToPlaneRowImpl(sse42_tag, const float* in, const int chan, const int chs, float* out, const int length);
@@ -1372,6 +1275,20 @@ template void nv12ToRgbRowImpl(sse42_tag, const uint8_t** y_rows, const uint8_t*
 
 template void i420ToRgbRowImpl(sse42_tag, const uint8_t** y_rows, const uint8_t* u_row,
                                const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+template void splitRowImpl<sse42_tag, uchar, 2>(sse42_tag, const uint8_t* in, std::array<uint8_t*, 2>& outs, const int length);
+template void splitRowImpl<sse42_tag, float, 2>(sse42_tag, const float* in, std::array<float*, 2>& outs, const int length);
+template void splitRowImpl<sse42_tag, uchar, 3>(sse42_tag, const uint8_t* in, std::array<uint8_t*, 3>& outs, const int length);
+template void splitRowImpl<sse42_tag, float, 3>(sse42_tag, const float* in, std::array<float*, 3>& outs, const int length);
+template void splitRowImpl<sse42_tag, uchar, 4>(sse42_tag, const uint8_t* in, std::array<uint8_t*, 4>& outs, const int length);
+template void splitRowImpl<sse42_tag, float, 4>(sse42_tag, const float* in, std::array<float*, 4>& outs, const int length);
+
+template void mergeRowImpl<sse42_tag, uchar, 2>(sse42_tag, const std::array<const uint8_t*, 2>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<sse42_tag, float, 2>(sse42_tag, const std::array<const float*, 2>& ins, float* out, const int length);
+template void mergeRowImpl<sse42_tag, uchar, 3>(sse42_tag, const std::array<const uint8_t*, 3>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<sse42_tag, float, 3>(sse42_tag, const std::array<const float*, 3>& ins, float* out, const int length);
+template void mergeRowImpl<sse42_tag, uchar, 4>(sse42_tag, const std::array<const uint8_t*, 4>& ins, uint8_t* out, const int length);
+template void mergeRowImpl<sse42_tag, float, 4>(sse42_tag, const std::array<const float*, 4>& ins, float* out, const int length);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.hpp b/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.hpp
index 8726013357e0fa..6dfa51515e796c 100644
--- a/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.hpp
+++ b/inference-engine/src/preprocessing/cpu_x86_sse42/ie_preprocess_gapi_kernels_sse42.hpp
@@ -106,80 +106,6 @@ void calcRowLinear_32F(float *dst[],
                  const Size & outSz,
                        int    lpi);
 
-//----------------------------------------------------------------------
-
-void mergeRow_8UC2(const uint8_t in0[],
-                   const uint8_t in1[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_8UC3(const uint8_t in0[],
-                   const uint8_t in1[],
-                   const uint8_t in2[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_8UC4(const uint8_t in0[],
-                   const uint8_t in1[],
-                   const uint8_t in2[],
-                   const uint8_t in3[],
-                         uint8_t out[],
-                             int length);
-
-void mergeRow_32FC2(const float in0[],
-                    const float in1[],
-                          float out[],
-                            int length);
-
-void mergeRow_32FC3(const float in0[],
-                    const float in1[],
-                    const float in2[],
-                          float out[],
-                            int length);
-
-void mergeRow_32FC4(const float in0[],
-                    const float in1[],
-                    const float in2[],
-                    const float in3[],
-                          float out[],
-                            int length);
-
-void splitRow_8UC2(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                             int length);
-
-void splitRow_8UC3(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                         uint8_t out2[],
-                             int length);
-
-void splitRow_8UC4(const uint8_t in[],
-                         uint8_t out0[],
-                         uint8_t out1[],
-                         uint8_t out2[],
-                         uint8_t out3[],
-                             int length);
-
-void splitRow_32FC2(const float in[],
-                          float out0[],
-                          float out1[],
-                            int length);
-
-void splitRow_32FC3(const float in[],
-                          float out0[],
-                          float out1[],
-                          float out2[],
-                            int length);
-
-void splitRow_32FC4(const float in[],
-                          float out0[],
-                          float out1[],
-                          float out2[],
-                          float out3[],
-                            int length);
-
 template<typename isa_tag_t, typename T>
 void chanToPlaneRowImpl(isa_tag_t, const T* in, const int chan, const int chs,
                         T* out, const int length);
@@ -199,6 +125,26 @@ void i420ToRgbRowImpl(isa_tag_t, const uint8_t** y_rows, const uint8_t* u_row,
 
 extern template void i420ToRgbRowImpl(sse42_tag, const uint8_t** y_rows, const uint8_t* u_row,
                                       const uint8_t* v_row, uint8_t** out_rows, const int buf_width);
+
+template<typename isa_tag_t, typename T, int chs>
+void splitRowImpl(isa_tag_t, const T* in, std::array<T*, chs>& outs, const int length);
+
+extern template void splitRowImpl<sse42_tag, uint8_t, 2>(sse42_tag, const uint8_t* in, std::array<uint8_t*, 2>& outs, const int length);
+extern template void splitRowImpl<sse42_tag, float, 2>(sse42_tag, const float* in, std::array<float*, 2>& outs, const int length);
+extern template void splitRowImpl<sse42_tag, uint8_t, 3>(sse42_tag, const uint8_t* in, std::array<uint8_t*, 3>& outs, const int length);
+extern template void splitRowImpl<sse42_tag, float, 3>(sse42_tag, const float* in, std::array<float*, 3>& outs, const int length);
+extern template void splitRowImpl<sse42_tag, uint8_t, 4>(sse42_tag, const uint8_t* in, std::array<uint8_t*, 4>& outs, const int length);
+extern template void splitRowImpl<sse42_tag, float, 4>(sse42_tag, const float* in, std::array<float*, 4>& outs, const int length);
+
+template<typename isa_tag_t, typename T, int chs>
+void mergeRowImpl(isa_tag_t, const std::array<const T*, chs>& ins, T* out, const int length);
+
+extern template void mergeRowImpl<sse42_tag, uint8_t, 2>(sse42_tag, const std::array<const uint8_t*, 2>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<sse42_tag, float, 2>(sse42_tag, const std::array<const float*, 2>& ins, float* out, const int length);
+extern template void mergeRowImpl<sse42_tag, uint8_t, 3>(sse42_tag, const std::array<const uint8_t*, 3>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<sse42_tag, float, 3>(sse42_tag, const std::array<const float*, 3>& ins, float* out, const int length);
+extern template void mergeRowImpl<sse42_tag, uint8_t, 4>(sse42_tag, const std::array<const uint8_t*, 4>& ins, uint8_t* out, const int length);
+extern template void mergeRowImpl<sse42_tag, float, 4>(sse42_tag, const std::array<const float*, 4>& ins, float* out, const int length);
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels.cpp b/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels.cpp
index 4cf4fffd74be9f..ffc62705c99ab0 100644
--- a/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels.cpp
+++ b/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels.cpp
@@ -42,396 +42,10 @@
 
 namespace InferenceEngine {
 namespace gapi {
-namespace kernels {
 
-template<typename T, int chs> static
-void mergeRow(const std::array<const uint8_t*, chs>& ins, uint8_t* out, int length) {
-// AVX512 implementation of wide universal intrinsics is slower than AVX2.
-// It is turned off until the cause isn't found out.
-#if 0
-#ifdef HAVE_AVX512
-    if (with_cpu_x86_avx512f()) {
-        if (std::is_same<T, uint8_t>::value && chs == 2) {
-            avx512::mergeRow_8UC2(ins[0], ins[1], out, length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 3) {
-            avx512::mergeRow_8UC3(ins[0], ins[1], ins[2], out, length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 4) {
-            avx512::mergeRow_8UC4(ins[0], ins[1], ins[2], ins[3], out, length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 2) {
-            avx512::mergeRow_32FC2(reinterpret_cast<const float*>(ins[0]),
-                                   reinterpret_cast<const float*>(ins[1]),
-                                   reinterpret_cast<float*>(out), length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 3) {
-            avx512::mergeRow_32FC3(reinterpret_cast<const float*>(ins[0]),
-                                   reinterpret_cast<const float*>(ins[1]),
-                                   reinterpret_cast<const float*>(ins[2]),
-                                   reinterpret_cast<float*>(out), length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 4) {
-            avx512::mergeRow_32FC4(reinterpret_cast<const float*>(ins[0]),
-                                   reinterpret_cast<const float*>(ins[1]),
-                                   reinterpret_cast<const float*>(ins[2]),
-                                   reinterpret_cast<const float*>(ins[3]),
-                                   reinterpret_cast<float*>(out), length);
-            return;
-        }
-    }
-#endif  // HAVE_AVX512
-#endif
-
-#ifdef HAVE_AVX2
-    if (with_cpu_x86_avx2()) {
-        if (std::is_same<T, uint8_t>::value && chs == 2) {
-            avx::mergeRow_8UC2(ins[0], ins[1], out, length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 3) {
-            avx::mergeRow_8UC3(ins[0], ins[1], ins[2], out, length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 4) {
-            avx::mergeRow_8UC4(ins[0], ins[1], ins[2], ins[3], out, length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 2) {
-            avx::mergeRow_32FC2(reinterpret_cast<const float*>(ins[0]),
-                                reinterpret_cast<const float*>(ins[1]),
-                                reinterpret_cast<float*>(out), length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 3) {
-            avx::mergeRow_32FC3(reinterpret_cast<const float*>(ins[0]),
-                                reinterpret_cast<const float*>(ins[1]),
-                                reinterpret_cast<const float*>(ins[2]),
-                                reinterpret_cast<float*>(out), length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 4) {
-            avx::mergeRow_32FC4(reinterpret_cast<const float*>(ins[0]),
-                                reinterpret_cast<const float*>(ins[1]),
-                                reinterpret_cast<const float*>(ins[2]),
-                                reinterpret_cast<const float*>(ins[3]),
-                                reinterpret_cast<float*>(out), length);
-            return;
-        }
-    }
-#endif  // HAVE_AVX2
-
-#ifdef HAVE_SSE
-    if (with_cpu_x86_sse42()) {
-        if (std::is_same<T, uint8_t>::value && chs == 2) {
-            mergeRow_8UC2(ins[0], ins[1], out, length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 3) {
-            mergeRow_8UC3(ins[0], ins[1], ins[2], out, length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 4) {
-            mergeRow_8UC4(ins[0], ins[1], ins[2], ins[3], out, length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 2) {
-            mergeRow_32FC2(reinterpret_cast<const float*>(ins[0]),
-                           reinterpret_cast<const float*>(ins[1]),
-                           reinterpret_cast<float*>(out), length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 3) {
-            mergeRow_32FC3(reinterpret_cast<const float*>(ins[0]),
-                           reinterpret_cast<const float*>(ins[1]),
-                           reinterpret_cast<const float*>(ins[2]),
-                           reinterpret_cast<float*>(out), length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 4) {
-            mergeRow_32FC4(reinterpret_cast<const float*>(ins[0]),
-                           reinterpret_cast<const float*>(ins[1]),
-                           reinterpret_cast<const float*>(ins[2]),
-                           reinterpret_cast<const float*>(ins[3]),
-                           reinterpret_cast<float*>(out), length);
-            return;
-        }
-    }
-#endif  // HAVE_SSE
-
-#ifdef HAVE_NEON
-    if (std::is_same<T, uint8_t>::value && chs == 2) {
-        neon::mergeRow_8UC2(ins[0], ins[1], out, length);
-        return;
-    }
-
-    if (std::is_same<T, uint8_t>::value && chs == 3) {
-        neon::mergeRow_8UC3(ins[0], ins[1], ins[2], out, length);
-        return;
-    }
-
-    if (std::is_same<T, uint8_t>::value && chs == 4) {
-        neon::mergeRow_8UC4(ins[0], ins[1], ins[2], ins[3], out, length);
-        return;
-    }
-
-    if (std::is_same<T, float>::value && chs == 2) {
-        neon::mergeRow_32FC2(reinterpret_cast<const float*>(ins[0]),
-                             reinterpret_cast<const float*>(ins[1]),
-                             reinterpret_cast<float*>(out), length);
-        return;
-    }
-
-    if (std::is_same<T, float>::value && chs == 3) {
-        neon::mergeRow_32FC3(reinterpret_cast<const float*>(ins[0]),
-                             reinterpret_cast<const float*>(ins[1]),
-                             reinterpret_cast<const float*>(ins[2]),
-                             reinterpret_cast<float*>(out), length);
-        return;
-    }
-
-    if (std::is_same<T, float>::value && chs == 4) {
-        neon::mergeRow_32FC4(reinterpret_cast<const float*>(ins[0]),
-                             reinterpret_cast<const float*>(ins[1]),
-                             reinterpret_cast<const float*>(ins[2]),
-                             reinterpret_cast<const float*>(ins[3]),
-                             reinterpret_cast<float*>(out), length);
-        return;
-    }
-#endif  // HAVE_NEON
-
-    const T* insT[chs];
-    for (int c = 0; c < chs; c++) {
-        insT[c] = reinterpret_cast<const T*>(ins[c]);
-    }
-    auto outT = reinterpret_cast<T*>(out);
-
-    for (int x = 0; x < length; x++) {
-        for (int c = 0; c < chs; c++) {
-            outT[chs*x + c] = insT[c][x];
-        }
-    }
-}
-
-template<typename T, int chs> static
-void splitRow(const uint8_t* in, std::array<uint8_t*, chs>& outs, int length) {
-#ifdef HAVE_AVX512
-    if (with_cpu_x86_avx512f()) {
-        if (std::is_same<T, uint8_t>::value && chs == 2) {
-            avx512::splitRow_8UC2(in, outs[0], outs[1], length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 3) {
-            avx512::splitRow_8UC3(in, outs[0], outs[1], outs[2], length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 4) {
-            avx512::splitRow_8UC4(in, outs[0], outs[1], outs[2], outs[3], length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 2) {
-            avx512::splitRow_32FC2(reinterpret_cast<const float*>(in),
-                                   reinterpret_cast<float*>(outs[0]),
-                                   reinterpret_cast<float*>(outs[1]),
-                                   length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 3) {
-            avx512::splitRow_32FC3(reinterpret_cast<const float*>(in),
-                                   reinterpret_cast<float*>(outs[0]),
-                                   reinterpret_cast<float*>(outs[1]),
-                                   reinterpret_cast<float*>(outs[2]),
-                                   length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 4) {
-            avx512::splitRow_32FC4(reinterpret_cast<const float*>(in),
-                                   reinterpret_cast<float*>(outs[0]),
-                                   reinterpret_cast<float*>(outs[1]),
-                                   reinterpret_cast<float*>(outs[2]),
-                                   reinterpret_cast<float*>(outs[3]),
-                                   length);
-            return;
-        }
-    }
-#endif  // HAVE_AVX512
-
-#ifdef HAVE_AVX2
-
-    if (with_cpu_x86_avx2()) {
-        if (std::is_same<T, uint8_t>::value && chs == 2) {
-            avx::splitRow_8UC2(in, outs[0], outs[1], length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 3) {
-            avx::splitRow_8UC3(in, outs[0], outs[1], outs[2], length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 4) {
-            avx::splitRow_8UC4(in, outs[0], outs[1], outs[2], outs[3], length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 2) {
-            avx::splitRow_32FC2(reinterpret_cast<const float*>(in),
-                                reinterpret_cast<float*>(outs[0]),
-                                reinterpret_cast<float*>(outs[1]),
-                                length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 3) {
-            avx::splitRow_32FC3(reinterpret_cast<const float*>(in),
-                                reinterpret_cast<float*>(outs[0]),
-                                reinterpret_cast<float*>(outs[1]),
-                                reinterpret_cast<float*>(outs[2]),
-                                length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 4) {
-            avx::splitRow_32FC4(reinterpret_cast<const float*>(in),
-                                reinterpret_cast<float*>(outs[0]),
-                                reinterpret_cast<float*>(outs[1]),
-                                reinterpret_cast<float*>(outs[2]),
-                                reinterpret_cast<float*>(outs[3]),
-                                length);
-            return;
-        }
-    }
-#endif  // HAVE_AVX2
-
-#ifdef HAVE_SSE
-    if (with_cpu_x86_sse42()) {
-        if (std::is_same<T, uint8_t>::value && chs == 2) {
-            splitRow_8UC2(in, outs[0], outs[1], length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 3) {
-            splitRow_8UC3(in, outs[0], outs[1], outs[2], length);
-            return;
-        }
-
-        if (std::is_same<T, uint8_t>::value && chs == 4) {
-            splitRow_8UC4(in, outs[0], outs[1], outs[2], outs[3], length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 2) {
-            splitRow_32FC2(reinterpret_cast<const float*>(in),
-                           reinterpret_cast<float*>(outs[0]),
-                           reinterpret_cast<float*>(outs[1]),
-                           length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 3) {
-            splitRow_32FC3(reinterpret_cast<const float*>(in),
-                           reinterpret_cast<float*>(outs[0]),
-                           reinterpret_cast<float*>(outs[1]),
-                           reinterpret_cast<float*>(outs[2]),
-                           length);
-            return;
-        }
-
-        if (std::is_same<T, float>::value && chs == 4) {
-            splitRow_32FC4(reinterpret_cast<const float*>(in),
-                           reinterpret_cast<float*>(outs[0]),
-                           reinterpret_cast<float*>(outs[1]),
-                           reinterpret_cast<float*>(outs[2]),
-                           reinterpret_cast<float*>(outs[3]),
-                           length);
-            return;
-        }
-    }
-#endif  // HAVE_SSE
-
-#ifdef HAVE_NEON
-    if (std::is_same<T, uint8_t>::value && chs == 2) {
-        neon::splitRow_8UC2(in, outs[0], outs[1], length);
-        return;
-    }
-
-    if (std::is_same<T, uint8_t>::value && chs == 3) {
-        neon::splitRow_8UC3(in, outs[0], outs[1], outs[2], length);
-        return;
-    }
-
-    if (std::is_same<T, uint8_t>::value && chs == 4) {
-        neon::splitRow_8UC4(in, outs[0], outs[1], outs[2], outs[3], length);
-        return;
-    }
-
-    if (std::is_same<T, float>::value && chs == 2) {
-        neon::splitRow_32FC2(reinterpret_cast<const float*>(in),
-                             reinterpret_cast<float*>(outs[0]),
-                             reinterpret_cast<float*>(outs[1]),
-                             length);
-        return;
-    }
-
-    if (std::is_same<T, float>::value && chs == 3) {
-        neon::splitRow_32FC3(reinterpret_cast<const float*>(in),
-                             reinterpret_cast<float*>(outs[0]),
-                             reinterpret_cast<float*>(outs[1]),
-                             reinterpret_cast<float*>(outs[2]),
-                             length);
-        return;
-    }
-
-    if (std::is_same<T, float>::value && chs == 4) {
-        neon::splitRow_32FC4(reinterpret_cast<const float*>(in),
-                             reinterpret_cast<float*>(outs[0]),
-                             reinterpret_cast<float*>(outs[1]),
-                             reinterpret_cast<float*>(outs[2]),
-                             reinterpret_cast<float*>(outs[3]),
-                             length);
-        return;
-    }
-#endif  // HAVE_NEON
-
-    auto inT = reinterpret_cast<const T*>(in);
-
-    T* outsT[chs];
-    for (int c = 0; c < chs; c++) {
-        outsT[c] = reinterpret_cast<T*>(outs[c]);
-    }
-
-    for (int x = 0; x < length; x++) {
-        for (int c = 0; c < chs; c++) {
-            outsT[c][x] = inT[chs*x + c];
-        }
-    }
-}
+//using namespace kernels;
 
+namespace kernels {
 namespace {
 
 struct fp_16_t {
@@ -583,167 +197,107 @@ bool is_cv_type_in_list(const int type_id) {
 }
 
 namespace {
-
 using merge_supported_types = typelist<uint8_t, int8_t, uint16_t, int16_t, int32_t, float, fp_16_t>;
 
-template<int chs>
-struct typed_merge_row {
-    using p_f = void (*)(const std::array<const uint8_t*, chs>& ins, uint8_t* out, int length);
-
-    template <typename type>
-    p_f operator()(type_to_type<type> ) { return mergeRow<type, chs>; }
-
-    p_f operator()(type_to_type<fp_16_t> ) {
-        static_assert(sizeof(fp_16_t) == sizeof(fp_16_t::v),
-                "fp_16_t should be a plain wrap over FP16 implementation type");
-        return mergeRow<decltype(fp_16_t::v), chs>;
-    }
-};
-
-}  // namespace
-
-GAPI_FLUID_KERNEL(FMerge2, Merge2, false) {
-    static const int LPI = 4;
-    static const int Window = 1;
-    static void run(const cv::gapi::fluid::View& a,
-                    const cv::gapi::fluid::View& b,
-                          cv::gapi::fluid::Buffer& out) {
-        GAPI_DbgAssert(is_cv_type_in_list<merge_supported_types>(out.meta().depth));
-
-        const auto rowFunc = type_dispatch<merge_supported_types>(out.meta().depth, cv_type_id{}, typed_merge_row<2>{}, nullptr);
-        for (int l = 0; l < out.lpi(); l++) {
-            rowFunc({a.InLineB(l), b.InLineB(l)}, out.OutLineB(l), a.length());
+template<typename T, int chs>
+void mergeRowImpl(scalar_tag, const std::array<const T*, chs>& ins, T* out, const int length) {
+    for (int x = 0; x < length; ++x) {
+        for (int c = 0; c < chs; ++c) {
+            out[chs * x + c] = ins[c][x];
         }
     }
-};
+}
 
-GAPI_FLUID_KERNEL(FMerge3, Merge3, false) {
-    static const int LPI = 4;
-    static const int Window = 1;
-    static void run(const cv::gapi::fluid::View& a,
-                    const cv::gapi::fluid::View& b,
-                    const cv::gapi::fluid::View& c,
-                          cv::gapi::fluid::Buffer& out) {
-        GAPI_DbgAssert(is_cv_type_in_list<merge_supported_types>(out.meta().depth));
+template<typename isa_tag_t, int chs>
+struct typed_merge_row {
+    using p_f = void (*)(const std::array<const uint8_t*, chs>& ins, uint8_t* out, const int length);
 
-        const auto rowFunc = type_dispatch<merge_supported_types>(out.meta().depth, cv_type_id{}, typed_merge_row<3>{}, nullptr);
-        for (int l = 0; l < out.lpi(); l++) {
-            rowFunc({a.InLineB(l), b.InLineB(l), c.InLineB(l)}, out.OutLineB(l), a.length());
-        }
+    template <typename type>
+    typename std::enable_if<std::is_same<isa_tag_t, scalar_tag>::value ||
+            (!std::is_same<isa_tag_t, scalar_tag>::value && !std::is_same<type, uint8_t>::value &&
+             !std::is_same<type, float>::value), p_f>::type
+    operator()(type_to_type<type> ) {
+        return [](const std::array<const uint8_t*, chs>& ins, uint8_t* out, const int length) {
+            const auto inT = reinterpret_cast<const std::array<const type*, chs>&>(ins);
+            auto outT = reinterpret_cast<type*>(out);
+            scalar_tag t;
+            mergeRowImpl<type, chs>(t, inT, outT, length);
+        };
     }
-};
 
-GAPI_FLUID_KERNEL(FMerge4, Merge4, false) {
-    static const int LPI = 4;
-    static const int Window = 1;
-    static void run(const cv::gapi::fluid::View& a,
-                    const cv::gapi::fluid::View& b,
-                    const cv::gapi::fluid::View& c,
-                    const cv::gapi::fluid::View& d,
-                          cv::gapi::fluid::Buffer& out) {
-        GAPI_DbgAssert(is_cv_type_in_list<merge_supported_types>(out.meta().depth));
-
-        const auto rowFunc = type_dispatch<merge_supported_types>(out.meta().depth, cv_type_id{}, typed_merge_row<4>{}, nullptr);
-        for (int l = 0; l < out.lpi(); l++) {
-            rowFunc({a.InLineB(l), b.InLineB(l), c.InLineB(l), d.InLineB(l)}, out.OutLineB(l), a.length());
-        }
+    template<typename tag = isa_tag_t>
+    typename std::enable_if<!std::is_same<tag, scalar_tag>::value, p_f>::type
+    operator()(type_to_type<uint8_t>) {
+        return [](const std::array<const uint8_t*, chs>& ins, uint8_t* out, const int length) {
+            tag t;
+            mergeRowImpl<tag, uint8_t, chs>(t, ins, out, length);
+        };
     }
-};
-
-
-namespace {
-using split_supported_types = typelist<uint8_t, int8_t, uint16_t, int16_t, int32_t, float, fp_16_t>;
-
-template<int chs>
-struct typed_split_row {
-    using p_f = void (*)(const uint8_t* in, std::array<uint8_t*, chs>& outs, int length);
-
-    template <typename type>
-    p_f operator()(type_to_type<type> ) { return splitRow<type, chs>; }
 
-    p_f operator()(type_to_type<fp_16_t> ) {
-        static_assert(sizeof(fp_16_t) == sizeof(fp_16_t::v),
-                "fp_16_t should be a plain wrap over FP16 implementation type");
-        return splitRow<decltype(fp_16_t::v), chs>;
+    template<typename tag = isa_tag_t>
+    typename std::enable_if<!std::is_same<tag, scalar_tag>::value, p_f>::type
+    operator()(type_to_type<float>) {
+        return [](const std::array<const uint8_t*, chs>& ins, uint8_t* out, const int length) {
+            const auto inT = reinterpret_cast<const std::array<const float*, chs>&>(ins);
+            auto outT = reinterpret_cast<float*>(out);
+            tag t;
+            mergeRowImpl<tag, float, chs>(t, inT, outT, length);
+        };
     }
 };
 
 }  // namespace
 
-GAPI_FLUID_KERNEL(FSplit2, Split2, false) {
-    static const int LPI = 4;
-    static const int Window = 1;
-    static void run(const cv::gapi::fluid::View  & in,
-                          cv::gapi::fluid::Buffer& out1,
-                          cv::gapi::fluid::Buffer& out2) {
-        GAPI_DbgAssert(2 == in.meta().chan);
-        GAPI_DbgAssert(1 == out1.meta().chan);
-        GAPI_DbgAssert(1 == out2.meta().chan);
-        GAPI_DbgAssert(in.meta().depth == out1.meta().depth);
-        GAPI_DbgAssert(in.meta().depth == out2.meta().depth);
-        GAPI_DbgAssert(is_cv_type_in_list<split_supported_types>(in.meta().depth));
+namespace {
+using split_supported_types = typelist<uint8_t, int8_t, uint16_t, int16_t, int32_t, float, fp_16_t>;
 
-        const auto rowFunc = type_dispatch<split_supported_types>(in.meta().depth, cv_type_id{}, typed_split_row<2>{}, nullptr);
-        for (int i = 0, lpi = out1.lpi(); i < lpi; i++) {
-            std::array<uint8_t*, 2> outs = {out1.OutLineB(i), out2.OutLineB(i)};
-            rowFunc(in.InLineB(i), outs, in.length());
+template<typename T, int chs>
+void splitRowImpl(scalar_tag, const T* in, std::array<T*, chs>& outs, const int length) {
+    for (int x = 0; x < length; ++x) {
+        for (int c = 0; c < chs; ++c) {
+            outs[c][x] = in[chs * x + c];
         }
     }
-};
-
-GAPI_FLUID_KERNEL(FSplit3, Split3, false) {
-    static const int LPI = 4;
-    static const int Window = 1;
-    static void run(const cv::gapi::fluid::View  & in,
-                          cv::gapi::fluid::Buffer& out1,
-                          cv::gapi::fluid::Buffer& out2,
-                          cv::gapi::fluid::Buffer& out3) {
-        GAPI_DbgAssert(3 == in.meta().chan);
-        GAPI_DbgAssert(1 == out1.meta().chan);
-        GAPI_DbgAssert(1 == out2.meta().chan);
-        GAPI_DbgAssert(1 == out3.meta().chan);
-        GAPI_DbgAssert(in.meta().depth == out1.meta().depth);
-        GAPI_DbgAssert(in.meta().depth == out2.meta().depth);
-        GAPI_DbgAssert(in.meta().depth == out3.meta().depth);
+}
 
-        GAPI_DbgAssert(is_cv_type_in_list<split_supported_types>(in.meta().depth));
+template<typename isa_tag_t, int chs>
+struct typed_split_row {
+    using p_f = void (*)(const uint8_t* in, std::array<uint8_t*, chs>& outs, const int length);
 
-        const auto rowFunc = type_dispatch<split_supported_types>(in.meta().depth, cv_type_id{}, typed_split_row<3>{}, nullptr);
-        for (int i = 0, lpi = out1.lpi(); i < lpi; i++) {
-            std::array<uint8_t*, 3> outs = {out1.OutLineB(i), out2.OutLineB(i),
-                                            out3.OutLineB(i)};
-            rowFunc(in.InLineB(i), outs, in.length());
-        }
+    template <typename type>
+    typename std::enable_if<std::is_same<isa_tag_t, scalar_tag>::value ||
+            (!std::is_same<isa_tag_t, scalar_tag>::value && !std::is_same<type, uint8_t>::value &&
+             !std::is_same<type, float>::value), p_f>::type
+    operator()(type_to_type<type> ) {
+        return [](const uint8_t* in, std::array<uint8_t*, chs>& outs, const int length) {
+            const auto inT = reinterpret_cast<const type*>(in);
+            auto outT = reinterpret_cast<std::array<type*, chs>&>(outs);
+            scalar_tag t;
+            splitRowImpl<type, chs>(t, inT, outT, length);
+        };
     }
-};
 
-GAPI_FLUID_KERNEL(FSplit4, Split4, false) {
-    static const int LPI = 4;
-    static const int Window = 1;
-    static void run(const cv::gapi::fluid::View  & in,
-                          cv::gapi::fluid::Buffer& out1,
-                          cv::gapi::fluid::Buffer& out2,
-                          cv::gapi::fluid::Buffer& out3,
-                          cv::gapi::fluid::Buffer& out4) {
-        GAPI_DbgAssert(4 == in.meta().chan);
-        GAPI_DbgAssert(1 == out1.meta().chan);
-        GAPI_DbgAssert(1 == out2.meta().chan);
-        GAPI_DbgAssert(1 == out3.meta().chan);
-        GAPI_DbgAssert(1 == out4.meta().chan);
-        GAPI_DbgAssert(in.meta().depth == out1.meta().depth);
-        GAPI_DbgAssert(in.meta().depth == out2.meta().depth);
-        GAPI_DbgAssert(in.meta().depth == out3.meta().depth);
-        GAPI_DbgAssert(in.meta().depth == out4.meta().depth);
-        GAPI_DbgAssert(is_cv_type_in_list<split_supported_types>(in.meta().depth));
+    template<typename tag = isa_tag_t>
+    typename std::enable_if<!std::is_same<tag, scalar_tag>::value, p_f>::type
+    operator()(type_to_type<uint8_t>) {
+        return [](const uint8_t* in, std::array<uint8_t*, chs>& outs, const int length) {
+            tag t;
+            splitRowImpl<tag, uint8_t, chs>(t, in, outs, length);
+        };
+    }
 
-        const auto rowFunc = type_dispatch<split_supported_types>(in.meta().depth, cv_type_id{}, typed_split_row<4>{}, nullptr);
-        for (int i = 0, lpi = out1.lpi(); i < lpi; i++) {
-            std::array<uint8_t*, 4> outs = {out1.OutLineB(i), out2.OutLineB(i),
-                                            out3.OutLineB(i), out4.OutLineB(i)};
-            rowFunc(in.InLineB(i), outs, in.length());
-        }
+    template<typename tag = isa_tag_t>
+    typename std::enable_if<!std::is_same<tag, scalar_tag>::value, p_f>::type
+    operator()(type_to_type<float>) {
+        return [](const uint8_t* in, std::array<uint8_t*, chs>& outs, const int length) {
+            const auto inT = reinterpret_cast<const float*>(in);
+            auto outT = reinterpret_cast<std::array<float*, chs>&>(outs);
+            tag t;
+            splitRowImpl<tag, float, chs>(t, inT, outT, length);
+        };
     }
 };
+}  // namespace
 
 //----------------------------------------------------------------------
 using isas_set = typelist<
@@ -1005,36 +559,179 @@ GAPI_FLUID_KERNEL(FI420toRGB, I420toRGB, false) {
         rowFunc(y_rows, u_row, v_row, out_rows, buf_width);
     }
 };
+
+GAPI_FLUID_KERNEL(FSplit2, Split2, false) {
+    static const int LPI = 4;
+    static const int Window = 1;
+    static void run(const cv::gapi::fluid::View & in,
+        cv::gapi::fluid::Buffer & out1,
+        cv::gapi::fluid::Buffer & out2) {
+        GAPI_DbgAssert(2 == in.meta().chan);
+        GAPI_DbgAssert(1 == out1.meta().chan);
+        GAPI_DbgAssert(1 == out2.meta().chan);
+        GAPI_DbgAssert(in.meta().depth == out1.meta().depth);
+        GAPI_DbgAssert(in.meta().depth == out2.meta().depth);
+        GAPI_DbgAssert(is_cv_type_in_list<split_supported_types>(in.meta().depth));
+
+        const auto rowFunc = type_dispatch<split_supported_types>(in.meta().depth, cv_type_id{}, typed_split_row<isa_tag_t, 2>{}, nullptr);
+        for (int i = 0, lpi = out1.lpi(); i < lpi; i++) {
+            std::array<uint8_t*, 2> outs = { out1.OutLineB(i), out2.OutLineB(i) };
+            rowFunc(in.InLineB(i), outs, in.length());
+        }
+    }
+};
+
+GAPI_FLUID_KERNEL(FSplit3, Split3, false) {
+    static const int LPI = 4;
+    static const int Window = 1;
+    static void run(const cv::gapi::fluid::View & in,
+        cv::gapi::fluid::Buffer & out1,
+        cv::gapi::fluid::Buffer & out2,
+        cv::gapi::fluid::Buffer & out3) {
+        GAPI_DbgAssert(3 == in.meta().chan);
+        GAPI_DbgAssert(1 == out1.meta().chan);
+        GAPI_DbgAssert(1 == out2.meta().chan);
+        GAPI_DbgAssert(1 == out3.meta().chan);
+        GAPI_DbgAssert(in.meta().depth == out1.meta().depth);
+        GAPI_DbgAssert(in.meta().depth == out2.meta().depth);
+        GAPI_DbgAssert(in.meta().depth == out3.meta().depth);
+
+        GAPI_DbgAssert(is_cv_type_in_list<split_supported_types>(in.meta().depth));
+
+        const auto rowFunc = type_dispatch<split_supported_types>(in.meta().depth, cv_type_id{}, typed_split_row<isa_tag_t, 3>{}, nullptr);
+        for (int i = 0, lpi = out1.lpi(); i < lpi; i++) {
+            std::array<uint8_t*, 3> outs = { out1.OutLineB(i), out2.OutLineB(i),
+                                            out3.OutLineB(i) };
+            rowFunc(in.InLineB(i), outs, in.length());
+        }
+    }
+};
+
+GAPI_FLUID_KERNEL(FSplit4, Split4, false) {
+    static const int LPI = 4;
+    static const int Window = 1;
+    static void run(const cv::gapi::fluid::View & in,
+        cv::gapi::fluid::Buffer & out1,
+        cv::gapi::fluid::Buffer & out2,
+        cv::gapi::fluid::Buffer & out3,
+        cv::gapi::fluid::Buffer & out4) {
+        GAPI_DbgAssert(4 == in.meta().chan);
+        GAPI_DbgAssert(1 == out1.meta().chan);
+        GAPI_DbgAssert(1 == out2.meta().chan);
+        GAPI_DbgAssert(1 == out3.meta().chan);
+        GAPI_DbgAssert(1 == out4.meta().chan);
+        GAPI_DbgAssert(in.meta().depth == out1.meta().depth);
+        GAPI_DbgAssert(in.meta().depth == out2.meta().depth);
+        GAPI_DbgAssert(in.meta().depth == out3.meta().depth);
+        GAPI_DbgAssert(in.meta().depth == out4.meta().depth);
+        GAPI_DbgAssert(is_cv_type_in_list<split_supported_types>(in.meta().depth));
+
+        const auto rowFunc = type_dispatch<split_supported_types>(in.meta().depth, cv_type_id{}, typed_split_row<isa_tag_t, 4>{}, nullptr);
+        for (int i = 0, lpi = out1.lpi(); i < lpi; i++) {
+            std::array<uint8_t*, 4> outs = { out1.OutLineB(i), out2.OutLineB(i),
+                                            out3.OutLineB(i), out4.OutLineB(i) };
+            rowFunc(in.InLineB(i), outs, in.length());
+        }
+    }
+};
+
+GAPI_FLUID_KERNEL(FMerge2, Merge2, false) {
+    static const int LPI = 4;
+    static const int Window = 1;
+    static void run(const cv::gapi::fluid::View & a,
+        const cv::gapi::fluid::View & b,
+        cv::gapi::fluid::Buffer & out) {
+        GAPI_DbgAssert(is_cv_type_in_list<merge_supported_types>(out.meta().depth));
+
+        const auto rowFunc = type_dispatch<merge_supported_types>(out.meta().depth, cv_type_id{}, typed_merge_row<isa_tag_t, 2>{}, nullptr);
+        for (int l = 0; l < out.lpi(); l++) {
+            rowFunc({ a.InLineB(l), b.InLineB(l) }, out.OutLineB(l), a.length());
+        }
+    }
+};
+
+GAPI_FLUID_KERNEL(FMerge3, Merge3, false) {
+    static const int LPI = 4;
+    static const int Window = 1;
+    static void run(const cv::gapi::fluid::View & a,
+        const cv::gapi::fluid::View & b,
+        const cv::gapi::fluid::View & c,
+        cv::gapi::fluid::Buffer & out) {
+        GAPI_DbgAssert(is_cv_type_in_list<merge_supported_types>(out.meta().depth));
+
+        const auto rowFunc = type_dispatch<merge_supported_types>(out.meta().depth, cv_type_id{}, typed_merge_row<isa_tag_t, 3>{}, nullptr);
+        for (int l = 0; l < out.lpi(); l++) {
+            rowFunc({ a.InLineB(l), b.InLineB(l), c.InLineB(l) }, out.OutLineB(l), a.length());
+        }
+    }
+};
+
+GAPI_FLUID_KERNEL(FMerge4, Merge4, false) {
+    static const int LPI = 4;
+    static const int Window = 1;
+    static void run(const cv::gapi::fluid::View & a,
+        const cv::gapi::fluid::View & b,
+        const cv::gapi::fluid::View & c,
+        const cv::gapi::fluid::View & d,
+        cv::gapi::fluid::Buffer & out) {
+        GAPI_DbgAssert(is_cv_type_in_list<merge_supported_types>(out.meta().depth));
+
+        const auto rowFunc = type_dispatch<merge_supported_types>(out.meta().depth, cv_type_id{}, typed_merge_row<isa_tag_t, 4>{}, nullptr);
+        for (int l = 0; l < out.lpi(); l++) {
+            rowFunc({ a.InLineB(l), b.InLineB(l), c.InLineB(l), d.InLineB(l) }, out.OutLineB(l), a.length());
+        }
+    }
+};
 };
 
 namespace {
-struct ColorConversionISA {
+struct CC_and_MergeISA {
     cv::gapi::GKernelPackage& pckg;
 
-    ColorConversionISA(cv::gapi::GKernelPackage& _pckg) : pckg(_pckg) {}
+    CC_and_MergeISA(cv::gapi::GKernelPackage& _pckg) : pckg(_pckg) {}
 
     template<typename isa_tag_t>
     bool operator()(type_to_type<isa_tag_t>) {
         pckg.include<typename choose_impl<isa_tag_t>::FI420toRGB>();
         pckg.include<typename choose_impl<isa_tag_t>::FNV12toRGB>();
         pckg.include<typename choose_impl<isa_tag_t>::FChanToPlane>();
+        pckg.include<typename choose_impl<isa_tag_t>::FMerge2>();
+        pckg.include<typename choose_impl<isa_tag_t>::FMerge3>();
+        pckg.include<typename choose_impl<isa_tag_t>::FMerge4>();
+        //at the moment type_dispatch requires something to be returned by the lambda
+        return true;
+    }
+};
+
+struct SplitISA {
+    cv::gapi::GKernelPackage& pckg;
+
+    SplitISA(cv::gapi::GKernelPackage& _pckg) : pckg(_pckg) {}
+
+    template<typename isa_tag_t>
+    bool operator()(type_to_type<isa_tag_t>) {
+        pckg.include<typename choose_impl<isa_tag_t>::FSplit2>();
+        pckg.include<typename choose_impl<isa_tag_t>::FSplit3>();
+        pckg.include<typename choose_impl<isa_tag_t>::FSplit4>();
         //at the moment type_dispatch requires something to be returned by the lambda
         return true;
     }
 };
 }  //namespace
 
-cv::gapi::GKernelPackage FColorConversionChooseISA() {
+cv::gapi::GKernelPackage FKernelsChooseISA() {
     // At the moment AVX512 implementation of wide universal intrinsics is slower than AVX2.
     // So, disable it for now.
     using isas = remove_t<isas_set, avx512_tag>;
 
-    cv::gapi::GKernelPackage pckg;
-    ColorConversionISA ctpISA{pckg};
+    cv::gapi::GKernelPackage pckg1, pckg2;
+    CC_and_MergeISA ccISA{ pckg1 };
+    SplitISA sISA{ pckg2 };
 
-    type_dispatch<isas>(is_isa_present{}, ctpISA, false);
+    type_dispatch<isas>(is_isa_present{}, ccISA, false);
+    type_dispatch<isas_set>(is_isa_present{}, sISA, false);
 
-    return pckg;
+    return combine(pckg1, pckg2);
 }
 
 //----------------------------------------------------------------------
@@ -2601,7 +2298,7 @@ using namespace kernels;
 
 cv::gapi::GKernelPackage preprocKernels() {
     return combine(
-        FColorConversionChooseISA(),
+        FKernelsChooseISA(),
         cv::gapi::kernels
         <FScalePlanes
         , FScalePlanes4
@@ -2612,12 +2309,6 @@ cv::gapi::GKernelPackage preprocKernels() {
         , FUpscalePlaneArea32f
         , FScalePlaneArea8u
         , FScalePlaneArea32f
-        , FMerge2
-        , FMerge3
-        , FMerge4
-        , FSplit2
-        , FSplit3
-        , FSplit4
         , FConvertDepth
         , FSubC
         , FDivC
diff --git a/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_simd_impl.hpp b/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_simd_impl.hpp
index e559b83b87e405..62ccb39dca925c 100644
--- a/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_simd_impl.hpp
+++ b/inference-engine/src/preprocessing/ie_preprocess_gapi_kernels_simd_impl.hpp
@@ -18,360 +18,204 @@ namespace gapi {
 
 namespace kernels {
 
-CV_ALWAYS_INLINE void mergeRow_8UC2_Impl(const uint8_t in0[], const uint8_t in1[],
-                                         uint8_t out[], int length) {
-    int l = 0;
-
-#if MANUAL_SIMD
-    constexpr int nlanes = v_uint8::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_uint8 r0, r1;
-        r0 = vx_load(&in0[l]);
-        r1 = vx_load(&in1[l]);
-        v_store_interleave(&out[2*l], r0, r1);
-    }
-
-    // to think about how to remove those ifs
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
-    }
-#endif
-
-    for (; l < length; ++l) {
-        out[2*l + 0] = in0[l];
-        out[2*l + 1] = in1[l];
-    }
-}
-
-CV_ALWAYS_INLINE void mergeRow_8UC3_Impl(const uint8_t in0[], const uint8_t in1[],
-                                         const uint8_t in2[], uint8_t out[], int length) {
-    int l = 0;
-
-#if MANUAL_SIMD
-    constexpr int nlanes = v_uint8::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_uint8 r0, r1, r2;
-        r0 = vx_load(&in0[l]);
-        r1 = vx_load(&in1[l]);
-        r2 = vx_load(&in2[l]);
-        v_store_interleave(&out[3*l], r0, r1, r2);
-    }
-
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
-    }
-#endif
-
-    for (; l < length; ++l) {
-        out[3*l + 0] = in0[l];
-        out[3*l + 1] = in1[l];
-        out[3*l + 2] = in2[l];
-    }
-}
-
-CV_ALWAYS_INLINE void mergeRow_8UC4_Impl(const uint8_t in0[], const uint8_t in1[],
-                                         const uint8_t in2[], const uint8_t in3[],
-                                         uint8_t out[], int length) {
-    int l = 0;
-
-#if MANUAL_SIMD
-    constexpr int nlanes = v_uint8::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_uint8 r0, r1, r2, r3;
-        r0 = vx_load(&in0[l]);
-        r1 = vx_load(&in1[l]);
-        r2 = vx_load(&in2[l]);
-        r3 = vx_load(&in3[l]);
-        v_store_interleave(&out[4*l], r0, r1, r2, r3);
-    }
-
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
-    }
-#endif
-
-    for (; l < length; ++l) {
-        out[4*l + 0] = in0[l];
-        out[4*l + 1] = in1[l];
-        out[4*l + 2] = in2[l];
-        out[4*l + 3] = in3[l];
-    }
-}
-
-CV_ALWAYS_INLINE void mergeRow_32FC2_Impl(const float in0[], const float in1[],
-                                          float out[], int length) {
-    int l = 0;
+template <typename VecT, typename T>
+CV_ALWAYS_INLINE void mergeRowC2_Impl(const T in0[], const T in1[],
+                                      T out[], const int length) {
+    int x = 0;
 
 #if MANUAL_SIMD
-    constexpr int nlanes = v_float32::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_float32 r0, r1;
-        r0 = vx_load(&in0[l]);
-        r1 = vx_load(&in1[l]);
-        v_store_interleave(&out[2*l], r0, r1);
-    }
+    constexpr int nlanes = VecT::nlanes;
+    GAPI_DbgAssert(length >= nlanes);
+
+    VecT r0, r1;
+    for (; length >= nlanes;) {
+        for (; x <= length - nlanes; x += nlanes) {
+            r0 = vx_load(&in0[x]);
+            r1 = vx_load(&in1[x]);
+            v_store_interleave(&out[2*x], r0, r1);
+        }
 
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
+        if (x < length) {
+            x = length - nlanes;
+            continue;
+        }
+        break;
     }
 #endif
 
-    for (; l < length; ++l) {
-        out[2*l + 0] = in0[l];
-        out[2*l + 1] = in1[l];
+    for (; x < length; ++x) {
+        out[2*x + 0] = in0[x];
+        out[2*x + 1] = in1[x];
     }
 }
 
-CV_ALWAYS_INLINE void mergeRow_32FC3_Impl(const float in0[], const float in1[], const float in2[],
-                                          float out[], int length) {
-    int l = 0;
+template <typename VecT, typename T>
+CV_ALWAYS_INLINE void mergeRowC3_Impl(const T in0[], const T in1[],
+                                      const T in2[], T out[], const int length) {
+    int x = 0;
 
 #if MANUAL_SIMD
-    constexpr int nlanes = v_float32::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_float32 r0, r1, r2;
-        r0 = vx_load(&in0[l]);
-        r1 = vx_load(&in1[l]);
-        r2 = vx_load(&in2[l]);
-        v_store_interleave(&out[3*l], r0, r1, r2);
-    }
+    constexpr int nlanes = VecT::nlanes;
+    GAPI_DbgAssert(length >= nlanes);
+
+    VecT r0, r1, r2;
+    for (; length >= nlanes;) {
+        for (; x <= length - nlanes; x += nlanes) {
+            r0 = vx_load(&in0[x]);
+            r1 = vx_load(&in1[x]);
+            r2 = vx_load(&in2[x]);
+            v_store_interleave(&out[3*x], r0, r1, r2);
+        }
 
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
+        if (x < length) {
+            x = length - nlanes;
+            continue;
+        }
+        break;
     }
 #endif
 
-    for (; l < length; ++l) {
-        out[3*l + 0] = in0[l];
-        out[3*l + 1] = in1[l];
-        out[3*l + 2] = in2[l];
+    for (; x < length; ++x) {
+        out[3*x + 0] = in0[x];
+        out[3*x + 1] = in1[x];
+        out[3*x + 2] = in2[x];
     }
 }
 
-CV_ALWAYS_INLINE void mergeRow_32FC4_Impl(const float in0[], const float in1[],
-                                          const float in2[], const float in3[],
-                                          float out[], int length) {
-    int l = 0;
+template <typename VecT, typename T>
+CV_ALWAYS_INLINE void mergeRowC4_Impl(const T in0[], const T in1[],
+                                      const T in2[], const T in3[],
+                                      T out[], const int length) {
+    int x = 0;
 
 #if MANUAL_SIMD
-    constexpr int nlanes = v_float32::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_float32 r0, r1, r2, r3;
-        r0 = vx_load(&in0[l]);
-        r1 = vx_load(&in1[l]);
-        r2 = vx_load(&in2[l]);
-        r3 = vx_load(&in3[l]);
-        v_store_interleave(&out[4*l], r0, r1, r2, r3);
-    }
+    constexpr int nlanes = VecT::nlanes;
+    GAPI_DbgAssert(length >= nlanes);
+
+    VecT r0, r1, r2, r3;
+    for (; length >= nlanes;) {
+        for (; x <= length - nlanes; x += nlanes) {
+            r0 = vx_load(&in0[x]);
+            r1 = vx_load(&in1[x]);
+            r2 = vx_load(&in2[x]);
+            r3 = vx_load(&in3[x]);
+            v_store_interleave(&out[4* x], r0, r1, r2, r3);
+        }
 
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
+        if (x < length) {
+            x = length - nlanes;
+            continue;
+        }
+        break;
     }
 #endif
 
-    for (; l < length; ++l) {
-        out[4*l + 0] = in0[l];
-        out[4*l + 1] = in1[l];
-        out[4*l + 2] = in2[l];
-        out[4*l + 3] = in3[l];
+    for (; x < length; ++x) {
+        out[4*x + 0] = in0[x];
+        out[4*x + 1] = in1[x];
+        out[4*x + 2] = in2[x];
+        out[4*x + 3] = in3[x];
     }
 }
-
 //------------------------------------------------------------------------------
-
-CV_ALWAYS_INLINE void splitRow_8UC2_Impl(const uint8_t in[], uint8_t out0[],
-                                         uint8_t out1[], int length) {
-    int l = 0;
-
-#if MANUAL_SIMD
-    constexpr int nlanes = v_uint8::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_uint8 r0, r1;
-        v_load_deinterleave(&in[2*l], r0, r1);
-        vx_store(&out0[l], r0);
-        vx_store(&out1[l], r1);
-    }
-
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
-    }
-#endif
-
-    for (; l < length; ++l) {
-        out0[l] = in[2*l + 0];
-        out1[l] = in[2*l + 1];
-    }
-}
-
-CV_ALWAYS_INLINE void splitRow_8UC3_Impl(const uint8_t in[], uint8_t out0[],
-                                         uint8_t out1[], uint8_t out2[], int length) {
-    int l = 0;
-
-#if MANUAL_SIMD
-    constexpr int nlanes = v_uint8::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-         v_uint8 r0, r1, r2;
-         v_load_deinterleave(&in[3*l], r0, r1, r2);
-         vx_store(&out0[l], r0);
-         vx_store(&out1[l], r1);
-         vx_store(&out2[l], r2);
-    }
-
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
-    }
-#endif
-
-    for (; l < length; ++l) {
-        out0[l] = in[3*l + 0];
-        out1[l] = in[3*l + 1];
-        out2[l] = in[3*l + 2];
-    }
-}
-
-CV_ALWAYS_INLINE void splitRow_8UC4_Impl(const uint8_t in[], uint8_t out0[], uint8_t out1[],
-                                         uint8_t out2[], uint8_t out3[], int length) {
-    int l = 0;
-
-#if MANUAL_SIMD
-    constexpr int nlanes = v_uint8::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_uint8 r0, r1, r2, r3;
-        v_load_deinterleave(&in[4*l], r0, r1, r2, r3);
-        vx_store(&out0[l], r0);
-        vx_store(&out1[l], r1);
-        vx_store(&out2[l], r2);
-        vx_store(&out3[l], r3);
-    }
-
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
-    }
-#endif
-
-    for (; l < length; ++l) {
-        out0[l] = in[4*l + 0];
-        out1[l] = in[4*l + 1];
-        out2[l] = in[4*l + 2];
-        out3[l] = in[4*l + 3];
-    }
-}
-
-CV_ALWAYS_INLINE void splitRow_32FC2_Impl(const float in[], float out0[],
-                                float out1[], int length) {
-    int l = 0;
+template <typename VecT, typename T>
+CV_ALWAYS_INLINE void splitRowC2_Impl(const T in[], T out0[],
+                                      T out1[], const int length) {
+    int x = 0;
 
 #if MANUAL_SIMD
-    constexpr int nlanes = v_float32::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_float32 r0, r1;
-        v_load_deinterleave(&in[2*l], r0, r1);
-        vx_store(&out0[l], r0);
-        vx_store(&out1[l], r1);
-    }
+    constexpr int nlanes = VecT::nlanes;
+    GAPI_DbgAssert(length >= nlanes);
+
+    VecT r0, r1;
+    for (; length >= nlanes;) {
+        for (; x <= length - nlanes; x += nlanes) {
+            v_load_deinterleave(&in[2*x], r0, r1);
+            vx_store(&out0[x], r0);
+            vx_store(&out1[x], r1);
+        }
 
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
+        if (x < length) {
+            x = length - nlanes;
+            continue;
+        }
+        break;
     }
-
 #endif
 
-    for (; l < length; ++l) {
-        out0[l] = in[2*l + 0];
-        out1[l] = in[2*l + 1];
+    for (; x < length; ++x) {
+        out0[x] = in[2*x + 0];
+        out1[x] = in[2*x + 1];
     }
 }
 
-CV_ALWAYS_INLINE void splitRow_32FC3_Impl(const float in[], float out0[], float out1[],
-                                          float out2[], int length) {
-    int l = 0;
+template <typename VecT, typename T>
+CV_ALWAYS_INLINE void splitRowC3_Impl(const T in[], T out0[],
+                                      T out1[], T out2[], const int length) {
+    int x = 0;
 
 #if MANUAL_SIMD
-    constexpr int nlanes = v_float32::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_float32 r0, r1, r2;
-        v_load_deinterleave(&in[3*l], r0, r1, r2);
-        vx_store(&out0[l], r0);
-        vx_store(&out1[l], r1);
-        vx_store(&out2[l], r2);
-    }
+    constexpr int nlanes = VecT::nlanes;
+    GAPI_DbgAssert(length >= nlanes);
+
+    VecT r0, r1, r2;
+    for (; length >= nlanes;) {
+        for (; x <= length - nlanes; x += nlanes) {
+             v_load_deinterleave(&in[3*x], r0, r1, r2);
+             vx_store(&out0[x], r0);
+             vx_store(&out1[x], r1);
+             vx_store(&out2[x], r2);
+        }
 
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
+        if (x < length) {
+            x = length - nlanes;
+            continue;
+        }
+        break;
     }
 #endif
 
-    for (; l < length; ++l) {
-        out0[l] = in[3*l + 0];
-        out1[l] = in[3*l + 1];
-        out2[l] = in[3*l + 2];
+    for (; x < length; ++x) {
+        out0[x] = in[3*x + 0];
+        out1[x] = in[3*x + 1];
+        out2[x] = in[3*x + 2];
     }
 }
 
-CV_ALWAYS_INLINE void splitRow_32FC4_Impl(const float in[], float out0[], float out1[],
-                                          float out2[], float out3[], int length) {
-    int l = 0;
+template <typename VecT, typename T>
+CV_ALWAYS_INLINE void splitRowC4_Impl(const T in[], T out0[], T out1[],
+                                      T out2[], T out3[], const int length) {
+    int x = 0;
 
 #if MANUAL_SIMD
-    constexpr int nlanes = v_float32::nlanes;
-
-    cycle:
-    for (; l <= length - nlanes; l += nlanes) {
-        v_float32 r0, r1, r2, r3;
-        v_load_deinterleave(&in[4*l], r0, r1, r2, r3);
-        vx_store(&out0[l], r0);
-        vx_store(&out1[l], r1);
-        vx_store(&out2[l], r2);
-        vx_store(&out3[l], r3);
-    }
+    constexpr int nlanes = VecT::nlanes;
+    GAPI_DbgAssert(length >= nlanes);
+
+    VecT r0, r1, r2, r3;
+    for (; length >= nlanes;) {
+        for (; x <= length - nlanes; x += nlanes) {
+            v_load_deinterleave(&in[4*x], r0, r1, r2, r3);
+            vx_store(&out0[x], r0);
+            vx_store(&out1[x], r1);
+            vx_store(&out2[x], r2);
+            vx_store(&out3[x], r3);
+        }
 
-    if (l < length && length >= nlanes) {
-        l = length - nlanes;
-        goto cycle;
+        if (x < length) {
+            x = length - nlanes;
+            continue;
+        }
+        break;
     }
 #endif
 
-    for (; l < length; ++l) {
-        out0[l] = in[4*l + 0];
-        out1[l] = in[4*l + 1];
-        out2[l] = in[4*l + 2];
-        out3[l] = in[4*l + 3];
+    for (; x < length; ++x) {
+        out0[x] = in[4*x + 0];
+        out1[x] = in[4*x + 1];
+        out2[x] = in[4*x + 2];
+        out3[x] = in[4*x + 3];
     }
 }
-
 //------------------------------------------------------------------------------
 
 CV_ALWAYS_INLINE void uvToRGBuv(const v_uint8& u, const v_uint8& v,
@@ -880,6 +724,38 @@ CV_ALWAYS_INLINE void chanToPlaneRowImpl(isa_tag_t, const T* in, const int chan,
         out[x] = in[x*chs + chan];
     }
 }
+
+template<typename isa_tag_t, typename T, int chs>
+CV_ALWAYS_INLINE void splitRowImpl(isa_tag_t, const T* in, std::array<T*, chs>& outs, const int length) {
+    static_assert(chs > 1 && chs < 5, "This number of channels isn't supported.");
+
+    if (chs == 2) {
+        splitRowC2_Impl<vector_type_of_t<isa_tag_t, T>, T>(in, outs[0], outs[1], length);
+        return;
+    } else if (chs == 3) {
+        splitRowC3_Impl<vector_type_of_t<isa_tag_t, T>, T>(in, outs[0], outs[1], outs[2], length);
+        return;
+    } else {
+        splitRowC4_Impl<vector_type_of_t<isa_tag_t, T>, T>(in, outs[0], outs[1], outs[2], outs[3], length);
+        return;
+    }
+}
+
+template<typename isa_tag_t, typename T, int chs>
+CV_ALWAYS_INLINE void mergeRowImpl(isa_tag_t, const std::array<const T*, chs>& ins, T* out, const int length) {
+    static_assert(chs > 1 && chs < 5, "This number of channels isn't supported.");
+
+    if (chs == 2) {
+        mergeRowC2_Impl<vector_type_of_t<isa_tag_t, T>, T>(ins[0], ins[1], out, length);
+        return;
+    } else if (chs == 3) {
+        mergeRowC3_Impl<vector_type_of_t<isa_tag_t, T>, T>(ins[0], ins[1], ins[2], out, length);
+        return;
+    } else {
+        mergeRowC4_Impl<vector_type_of_t<isa_tag_t, T>, T>(ins[0], ins[1], ins[2], ins[3], out, length);
+        return;
+    }
+}
 }  // namespace kernels
 }  // namespace gapi
 }  // namespace InferenceEngine
diff --git a/inference-engine/tests_deprecated/fluid_preproc/common/fluid_tests.cpp b/inference-engine/tests_deprecated/fluid_preproc/common/fluid_tests.cpp
index c7300214c7ba86..f883e7223487a8 100644
--- a/inference-engine/tests_deprecated/fluid_preproc/common/fluid_tests.cpp
+++ b/inference-engine/tests_deprecated/fluid_preproc/common/fluid_tests.cpp
@@ -98,10 +98,30 @@ cv::String typeToString(int type)
     case CV_8UC2  : return "CV_8UC2";
     case CV_8UC3  : return "CV_8UC3";
     case CV_8UC4  : return "CV_8UC4";
+    case CV_16FC1 : return "CV_16FC1";
+    case CV_16FC2 : return "CV_16FC2";
+    case CV_16FC3 : return "CV_16FC3";
+    case CV_16FC4 : return "CV_16FC4";
     case CV_32FC1 : return "CV_32FC1";
     case CV_32FC2 : return "CV_32FC2";
     case CV_32FC3 : return "CV_32FC3";
     case CV_32FC4 : return "CV_32FC4";
+    case CV_8SC1  : return "CV_8SC1";
+    case CV_8SC2  : return "CV_8SC2";
+    case CV_8SC3  : return "CV_8SC3";
+    case CV_8SC4  : return "CV_8SC4";
+    case CV_16SC1 : return "CV_16SC1";
+    case CV_16SC2 : return "CV_16SC2";
+    case CV_16SC3 : return "CV_16SC3";
+    case CV_16SC4 : return "CV_16SC4";
+    case CV_16UC1 : return "CV_16UC1";
+    case CV_16UC2 : return "CV_16UC2";
+    case CV_16UC3 : return "CV_16UC3";
+    case CV_16UC4 : return "CV_16UC4";
+    case CV_32SC1 : return "CV_32SC1";
+    case CV_32SC2 : return "CV_32SC2";
+    case CV_32SC3 : return "CV_32SC3";
+    case CV_32SC4 : return "CV_32SC4";
     }
     CV_Assert(!"ERROR: unsupported type!");
     return nullptr;

From bfdc6ce7a82b22274f745a9aa00cd58e8d701f1c Mon Sep 17 00:00:00 2001
From: Yegor Kruglov <yegor.kruglov@intel.com>
Date: Tue, 22 Jun 2021 19:05:48 +0300
Subject: [PATCH 166/178] [BERT-NER] Document model support (#6238)

* added docs

* up to date with release branch

* delete print
---
 .../Convert_Model_From_PyTorch.md             |  1 +
 .../pytorch_specific/Convert_Bert_ner.md      | 55 +++++++++++++++++++
 docs/doxygen/ie_docs.xml                      |  1 +
 3 files changed, 57 insertions(+)
 create mode 100644 docs/MO_DG/prepare_model/convert_model/pytorch_specific/Convert_Bert_ner.md

diff --git a/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_PyTorch.md b/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_PyTorch.md
index 086d1de2e20324..0898fd7e2225f7 100644
--- a/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_PyTorch.md
+++ b/docs/MO_DG/prepare_model/convert_model/Convert_Model_From_PyTorch.md
@@ -26,6 +26,7 @@ It is not a full list of models that can be converted to ONNX\* and to IR.
 * QuartzNet topologies from [NeMo project](https://github.com/NVIDIA/NeMo) can be converted using [Convert PyTorch\* QuartzNet to the IR](pytorch_specific/Convert_QuartzNet.md) instruction.
 * YOLACT topology can be converted using [Convert PyTorch\* YOLACT to the IR](pytorch_specific/Convert_YOLACT.md) instruction.
 * [RCAN](https://github.com/yulunzhang/RCAN) topologies can be converted using [Convert PyTorch\* RCAN to the IR](pytorch_specific/Convert_RCAN.md) instruction.
+* [BERT_NER](https://github.com/kamalkraj/BERT-NER) can be converted using [Convert PyTorch* BERT-NER to the IR](pytorch_specific/Convert_Bert_ner.md) instruction.
 
 ## Export PyTorch\* Model to ONNX\* Format <a name="export-to-onnx"></a>
 
diff --git a/docs/MO_DG/prepare_model/convert_model/pytorch_specific/Convert_Bert_ner.md b/docs/MO_DG/prepare_model/convert_model/pytorch_specific/Convert_Bert_ner.md
new file mode 100644
index 00000000000000..cc4e78dcb077a8
--- /dev/null
+++ b/docs/MO_DG/prepare_model/convert_model/pytorch_specific/Convert_Bert_ner.md
@@ -0,0 +1,55 @@
+# Convert PyTorch* BERT-NER to the Intermediate Representation {#openvino_docs_MO_DG_prepare_model_convert_model_pytorch_specific_Convert_Bert_ner}
+
+## Download and Convert the Model to ONNX*
+
+To download a pre-trained model or train the model yourself, refer
+to the [instruction](https://github.com/kamalkraj/BERT-NER/blob/dev/README.md) in the
+BERT-NER model repository. The model with config files is stored in the `out_base` directory.
+
+To convert the model to ONNX* format, create and run the script with the following content in the root
+directory of the model repository. If you download the pre-trained model, you need
+to download [`bert.py`](https://github.com/kamalkraj/BERT-NER/blob/dev/bert.py) to run the script.
+The instruction was tested with the repository hash commit `e5be564156f194f1becb0d82aeaf6e762d9eb9ed`.
+
+```python
+import torch
+
+from bert import Ner
+
+ner = Ner("out_base")
+
+input_ids, input_mask, segment_ids, valid_positions = ner.preprocess('Steve went to Paris')
+input_ids = torch.tensor([input_ids], dtype=torch.long, device=ner.device)
+input_mask = torch.tensor([input_mask], dtype=torch.long, device=ner.device)
+segment_ids = torch.tensor([segment_ids], dtype=torch.long, device=ner.device)
+valid_ids = torch.tensor([valid_positions], dtype=torch.long, device=ner.device)
+
+ner_model, tknizr, model_config = ner.load_model("out_base")
+
+with torch.no_grad():
+    logits = ner_model(input_ids, segment_ids, input_mask, valid_ids)
+torch.onnx.export(ner_model,
+                  (input_ids, segment_ids, input_mask, valid_ids),
+                  "bert-ner.onnx",
+                  input_names=['input_ids', 'segment_ids', 'input_mask', 'valid_ids'],
+                  output_names=['output'],
+                  dynamic_axes={
+                      "input_ids": {0: "batch_size"},
+                      "segment_ids": {0: "batch_size"},
+                      "input_mask": {0: "batch_size"},
+                      "valid_ids": {0: "batch_size"},
+                      "output": {0: "output"}
+                  },
+                  opset_version=11,
+                  )
+```
+
+The script generates ONNX* model file `bert-ner.onnx`.
+
+## Convert ONNX* BERT-NER model to IR
+
+```bash
+python mo.py --input_model bert-ner.onnx --input "input_mask[1 128],segment_ids[1 128],input_ids[1 128]"
+```
+
+where `1` is `batch_size` and `128` is `sequence_length`.
\ No newline at end of file
diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index 7e426edde06d50..b50832954d4a55 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -61,6 +61,7 @@ limitations under the License.
                             <tab type="user" title="Convert PyTorch* YOLACT Model" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_pytorch_specific_Convert_YOLACT"/>
                             <tab type="user" title="Convert PyTorch* F3Net Model" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_pytorch_specific_Convert_F3Net"/>
                             <tab type="user" title="Convert PyTorch* RCAN Model" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_pytorch_specific_Convert_RCAN">
+                            <tab type="user" title="Convert PyTorch* BERT-NER Model" url="@ref openvino_docs_MO_DG_prepare_model_convert_model_pytorch_specific_Convert_Bert_ner"/>
                             </tab>
                         </tab>
                         <tab type="user" title="Model Optimizations Techniques" url="@ref openvino_docs_MO_DG_prepare_model_Model_Optimization_Techniques"/>

From 861d89c9888da18a04c214f6b4bb687be092002c Mon Sep 17 00:00:00 2001
From: hyunback kim <hyunback.kim@intel.com>
Date: Wed, 23 Jun 2021 13:10:04 +0900
Subject: [PATCH 167/178] [IE CLDNN] Use single primitive reduce when
 axis=1(feature) (#4548)

---
 inference-engine/src/cldnn_engine/cldnn_engine.cpp         | 5 +----
 .../core/actual_kernels/resample/resample_kernel_opt.cpp   | 7 +++++++
 .../core/cl_kernels/reduce_gpu_b_fs_yx_fsv16.cl            | 6 +++---
 3 files changed, 11 insertions(+), 7 deletions(-)

diff --git a/inference-engine/src/cldnn_engine/cldnn_engine.cpp b/inference-engine/src/cldnn_engine/cldnn_engine.cpp
index e3e67d95f152f1..5b3b90eb832597 100644
--- a/inference-engine/src/cldnn_engine/cldnn_engine.cpp
+++ b/inference-engine/src/cldnn_engine/cldnn_engine.cpp
@@ -134,11 +134,8 @@ cldnn::device_info clDNNEngine::GetDeviceInfo(const std::map<std::string, std::s
 template<typename T>
 static bool disableReduceDecomposition(const std::shared_ptr<const ngraph::Node> node) {
     if (auto op = std::dynamic_pointer_cast<const T>(node)) {
-        auto reduction_axes = op->get_reduction_axes().to_vector();
-        bool reduce_along_f = op->get_reduction_axes().size() == 1 && std::count(reduction_axes.begin(), reduction_axes.end(), 1) != 0;
         bool fp16_batch_not_1 = op->get_element_type() == ngraph::element::f16 && op->input(0).get_shape()[0] != 1;
-        bool can_use_reduce = !reduce_along_f && !fp16_batch_not_1;
-        return can_use_reduce;
+        return !fp16_batch_not_1;
     }
     return false;
 }
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/resample/resample_kernel_opt.cpp b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/resample/resample_kernel_opt.cpp
index 7773696b3141a7..388f64d05565ca 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/resample/resample_kernel_opt.cpp
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/actual_kernels/resample/resample_kernel_opt.cpp
@@ -22,6 +22,8 @@ ParamsKey ResampleKernelOpt::GetSupportedKey() const {
     ParamsKey k;
     k.EnableInputDataType(Datatype::F16);
     k.EnableInputDataType(Datatype::F32);
+    k.EnableInputDataType(Datatype::UINT8);
+    k.EnableInputDataType(Datatype::INT8);
     k.EnableOutputDataType(Datatype::F16);
     k.EnableOutputDataType(Datatype::F32);
     k.EnableOutputDataType(Datatype::UINT8);
@@ -84,6 +86,11 @@ bool ResampleKernelOpt::Validate(const Params& p, const optional_params& o) cons
 
     const auto& input = params.inputs[0];
 
+    if ((input.GetDType() == Datatype::UINT8 || input.GetDType() == Datatype::INT8) &&
+        params.resampleType != ResampleType::NEAREST_NEIGHBOR &&
+        params.resampleType != ResampleType::BILINEAR_INTERP)
+        return false;
+
     if (input.GetLayout() != DataLayout::fs_b_yx_fsv32 && input.GetLayout() != DataLayout::b_fs_yx_fsv16)
         return false;
 
diff --git a/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/reduce_gpu_b_fs_yx_fsv16.cl b/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/reduce_gpu_b_fs_yx_fsv16.cl
index 84cf2021fb6eca..62ef44901b75d4 100644
--- a/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/reduce_gpu_b_fs_yx_fsv16.cl
+++ b/inference-engine/thirdparty/clDNN/kernel_selector/core/cl_kernels/reduce_gpu_b_fs_yx_fsv16.cl
@@ -256,7 +256,7 @@ uint offset = batch_out * input_batch_pitch + ((feature_out + FSV - 1) / FSV) *
             for (uint yi = y_out; yi < y_max_val; ++yi) {
                 for (uint xi = x_out; xi < x_max_val; ++xi) {
                     INPUT_VEC input = (INPUT_VEC)(INPUT_INIT_VAL);
-                    #if (REDUCE_MAX_MODE || REDUCE_MIN_MODE || REDUCE_PROD_MODE || REDUCE_AND_MODE || REDUCE_LOG_SUM_EXP_MODE) && REDUCE_FEATURE && (INPUT0_FEATURE_NUM % FSV != 0)
+                    #if REDUCE_FEATURE && (INPUT0_FEATURE_NUM % FSV != 0)
                         if (fi + FSV <= INPUT0_FEATURE_NUM)
                             input = BLOCK_READ(data, offset);
                         else
@@ -273,7 +273,7 @@ uint offset = batch_out * input_batch_pitch + ((feature_out + FSV - 1) / FSV) *
                 #if INPUT0_SIZE_X % READ_OFFSET != 0
                     for (uint xi = x_leftover_start; xi < x_leftover_end; ++xi) {
                         INPUT0_TYPE leftovers = INIT_VAL;
-                        #if (REDUCE_MAX_MODE || REDUCE_MIN_MODE || REDUCE_PROD_MODE || REDUCE_AND_MODE || REDUCE_LOG_SUM_EXP_MODE) && REDUCE_FEATURE && (INPUT0_FEATURE_NUM % FSV != 0)
+                        #if REDUCE_FEATURE && (INPUT0_FEATURE_NUM % FSV != 0)
                             if (fi + FSV <= INPUT0_FEATURE_NUM)
                                 leftovers = DT_INPUT_BLOCK_READ(data, offset);
                             else
@@ -341,7 +341,7 @@ uint offset = batch_out * input_batch_pitch + ((feature_out + FSV - 1) / FSV) *
             for (uint yi = y_out; yi < y_max_val; ++yi) {
                 for (uint xi = x_out; xi < x_max_val; ++xi) {
                     INPUT_VEC input = (INPUT_VEC)(INPUT_INIT_VAL);
-                    #if (REDUCE_MAX_MODE || REDUCE_MIN_MODE || REDUCE_PROD_MODE || REDUCE_AND_MODE || REDUCE_LOG_SUM_EXP_MODE) && REDUCE_FEATURE && (INPUT0_FEATURE_NUM % FSV != 0)
+                    #if REDUCE_FEATURE && (INPUT0_FEATURE_NUM % FSV != 0)
                         if (fi + FSV <= INPUT0_FEATURE_NUM)
                             input = BLOCK_READ(data, offset);
                         else

From 4a4c3e8ec9e6e53695ee40d9037c6502a879ef0f Mon Sep 17 00:00:00 2001
From: Mikhail Nosov <mikhail.nosov@intel.com>
Date: Wed, 23 Jun 2021 09:33:50 +0300
Subject: [PATCH 168/178] [DOC] Model caching feature overview (#5519)

* Docs: Model caching feature overview

* Update docs/IE_DG/Intro_to_Performance.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Apply suggestions from code review

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

* Review comments
- Moved code examples to snippets
- Added link to Model Caching overview from "Inference Engine Developer Guide"
- Few minor changes

* Update docs/IE_DG/Intro_to_Performance.md

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>

Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
---
 docs/IE_DG/Intro_to_Performance.md         |  6 ++
 docs/IE_DG/Model_caching_overview.md       | 65 ++++++++++++++++++++++
 docs/doxygen/ie_docs.xml                   |  1 +
 docs/img/caching_enabled.png               |  3 +
 docs/img/caching_times.png                 |  3 +
 docs/snippets/InferenceEngine_Caching0.cpp | 17 ++++++
 docs/snippets/InferenceEngine_Caching1.cpp | 13 +++++
 docs/snippets/InferenceEngine_Caching2.cpp | 14 +++++
 docs/snippets/InferenceEngine_Caching3.cpp | 20 +++++++
 9 files changed, 142 insertions(+)
 create mode 100644 docs/IE_DG/Model_caching_overview.md
 create mode 100644 docs/img/caching_enabled.png
 create mode 100644 docs/img/caching_times.png
 create mode 100644 docs/snippets/InferenceEngine_Caching0.cpp
 create mode 100644 docs/snippets/InferenceEngine_Caching1.cpp
 create mode 100644 docs/snippets/InferenceEngine_Caching2.cpp
 create mode 100644 docs/snippets/InferenceEngine_Caching3.cpp

diff --git a/docs/IE_DG/Intro_to_Performance.md b/docs/IE_DG/Intro_to_Performance.md
index 0c9457ed4bfd79..48d1ea5c56cff0 100644
--- a/docs/IE_DG/Intro_to_Performance.md
+++ b/docs/IE_DG/Intro_to_Performance.md
@@ -31,6 +31,12 @@ input images to achieve optimal throughput. However, high batch size also comes
 latency penalty. So, for more real-time oriented usages, lower batch sizes (as low as a single input) are used.
 Refer to the [Benchmark App](../../inference-engine/samples/benchmark_app/README.md) sample, which allows latency vs. throughput measuring.
 
+## Using Caching API for first inference latency optimization
+Since with the 2021.4 release, Inference Engine provides an ability to enable internal caching of loaded networks.
+This can significantly reduce load network latency for some devices at application startup.
+Internally caching uses plugin's Export/ImportNetwork flow, like it is done for [Compile tool](../../inference-engine/tools/compile_tool/README.md), using the regular ReadNetwork/LoadNetwork API.
+Refer to the [Model Caching Overview](Model_caching_overview.md) for more detailed explanation.
+
 ## Using Async API
 To gain better performance on accelerators, such as VPU, the Inference Engine uses the asynchronous approach (see
 [Integrating Inference Engine in Your Application (current API)](Integrate_with_customer_application_new_API.md)).
diff --git a/docs/IE_DG/Model_caching_overview.md b/docs/IE_DG/Model_caching_overview.md
new file mode 100644
index 00000000000000..25ae7387c244f3
--- /dev/null
+++ b/docs/IE_DG/Model_caching_overview.md
@@ -0,0 +1,65 @@
+# Model Caching Overview {#openvino_docs_IE_DG_Model_caching_overview}
+
+## Introduction
+
+As described in [Inference Engine Introduction](inference_engine_intro.md), common application flow consists of the following steps:
+
+1. **Create Inference Engine Core object**
+
+2. **Read the Intermediate Representation** - Read an Intermediate Representation file into an object of the `InferenceEngine::CNNNetwork`
+
+3. **Prepare inputs and outputs**
+
+4. **Set configuration** Pass device-specific loading configurations to the device
+
+5. **Compile and Load Network to device** - Use the `InferenceEngine::Core::LoadNetwork()` method with specific device
+
+6. **Set input data**
+
+7. **Execute**
+
+Step #5 can potentially perform several time-consuming device-specific optimizations and network compilations,
+and such delays can lead to bad user experience on application startup. To avoid this, some devices offer
+Import/Export network capability, and it is possible to either use [Compile tool](../../inference-engine/tools/compile_tool/README.md)
+or enable model caching to export compiled network automatically. Reusing cached networks can significantly reduce load network time.
+
+
+## Set "CACHE_DIR" config option to enable model caching
+
+To enable model caching, the application must specify the folder where to store cached blobs. It can be done like this
+
+
+@snippet snippets/InferenceEngine_Caching0.cpp part0
+
+With this code, if device supports Import/Export network capability, cached blob is automatically created inside the `myCacheFolder` folder
+CACHE_DIR config is set to the Core object. If device does not support Import/Export capability, cache is just not created and no error is thrown
+
+Depending on your device, total time for loading network on application startup can be significantly reduced.
+Please also note that very first LoadNetwork (when cache is not yet created) takes slightly longer time to 'export' compiled blob into a cache file
+![caching_enabled]
+
+## Even faster: use LoadNetwork(modelPath)
+
+In some cases, applications do not need to customize inputs and outputs every time. Such applications always
+call `cnnNet = ie.ReadNetwork(...)`, then `ie.LoadNetwork(cnnNet, ..)` and it can be further optimized.
+For such cases, more convenient API to load network in one call is introduced in the 2021.4 release.
+
+@snippet snippets/InferenceEngine_Caching1.cpp part1
+
+With enabled model caching, total load time is even smaller - in case that ReadNetwork is optimized as well
+
+@snippet snippets/InferenceEngine_Caching2.cpp part2
+
+![caching_times]
+
+
+## Advanced examples
+
+Not every device supports network import/export capability, enabling of caching for such devices do not have any effect.
+To check in advance if a particular device supports model caching, your application can use the following code:
+
+@snippet snippets/InferenceEngine_Caching3.cpp part3
+
+
+[caching_enabled]: ../img/caching_enabled.png
+[caching_times]: ../img/caching_times.png
diff --git a/docs/doxygen/ie_docs.xml b/docs/doxygen/ie_docs.xml
index b50832954d4a55..19a87a1e11e97c 100644
--- a/docs/doxygen/ie_docs.xml
+++ b/docs/doxygen/ie_docs.xml
@@ -285,6 +285,7 @@ limitations under the License.
                 <tab type="user" title="Inference Engine API Changes History" url="@ref openvino_docs_IE_DG_API_Changes"/>
                 <tab type="user" title="Inference Engine Memory primitives" url="@ref openvino_docs_IE_DG_Memory_primitives"/>
                 <tab type="user" title="Inference Engine Device Query API" url="@ref openvino_docs_IE_DG_InferenceEngine_QueryAPI"/>
+                <tab type="user" title="Inference Engine Model Caching" url="@ref openvino_docs_IE_DG_Model_caching_overview"/>
                 <tab type="usergroup" title="Inference Engine Extensibility Mechanism" url="@ref openvino_docs_IE_DG_Extensibility_DG_Intro">
                     <tab type="user" title="Extension Library" url="@ref openvino_docs_IE_DG_Extensibility_DG_Extension"/>
                     <tab type="user" title="Custom Operations" url="@ref openvino_docs_IE_DG_Extensibility_DG_AddingNGraphOps"/>
diff --git a/docs/img/caching_enabled.png b/docs/img/caching_enabled.png
new file mode 100644
index 00000000000000..f8a898764e17ee
--- /dev/null
+++ b/docs/img/caching_enabled.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:488a7a47e5086a6868c22219bc9d58a3508059e5a1dc470f2653a12552dea82f
+size 36207
diff --git a/docs/img/caching_times.png b/docs/img/caching_times.png
new file mode 100644
index 00000000000000..11d9c8b088f9f6
--- /dev/null
+++ b/docs/img/caching_times.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2eed189f9cb3d30fe13b4ba4515edd4e6da5d01545660e65fa8a33d945967281
+size 28894
diff --git a/docs/snippets/InferenceEngine_Caching0.cpp b/docs/snippets/InferenceEngine_Caching0.cpp
new file mode 100644
index 00000000000000..5311a3d0bb681c
--- /dev/null
+++ b/docs/snippets/InferenceEngine_Caching0.cpp
@@ -0,0 +1,17 @@
+#include <ie_core.hpp>
+
+int main() {
+using namespace InferenceEngine;
+    std::string modelPath = "/tmp/myModel.xml";
+    std::string device = "GNA";
+    std::map<std::string, std::string> deviceConfig;
+//! [part0]
+    InferenceEngine::Core ie;                                 // Step 1: create Inference engine object
+    ie.SetConfig({{CONFIG_KEY(CACHE_DIR), "myCacheFolder"}}); // Step 1b: Enable caching
+    auto cnnNet = ie.ReadNetwork(modelPath);                  // Step 2: ReadNetwork
+    //...                                                     // Step 3: Prepare inputs/outputs
+    //...                                                     // Step 4: Set device configuration
+    ie.LoadNetwork(cnnNet, device, deviceConfig);             // Step 5: LoadNetwork
+//! [part0]
+return 0;
+}
diff --git a/docs/snippets/InferenceEngine_Caching1.cpp b/docs/snippets/InferenceEngine_Caching1.cpp
new file mode 100644
index 00000000000000..3c9d0c5b22d558
--- /dev/null
+++ b/docs/snippets/InferenceEngine_Caching1.cpp
@@ -0,0 +1,13 @@
+#include <ie_core.hpp>
+
+int main() {
+using namespace InferenceEngine;
+    std::string modelPath = "/tmp/myModel.xml";
+    std::string device = "GNA";
+    std::map<std::string, std::string> deviceConfig;
+//! [part1]
+    InferenceEngine::Core ie;                                 // Step 1: create Inference engine object
+    ie.LoadNetwork(modelPath, device, deviceConfig);          // Step 2: LoadNetwork by model file path
+//! [part1]
+return 0;
+}
diff --git a/docs/snippets/InferenceEngine_Caching2.cpp b/docs/snippets/InferenceEngine_Caching2.cpp
new file mode 100644
index 00000000000000..aaf4b33c10da90
--- /dev/null
+++ b/docs/snippets/InferenceEngine_Caching2.cpp
@@ -0,0 +1,14 @@
+#include <ie_core.hpp>
+
+int main() {
+using namespace InferenceEngine;
+    std::string modelPath = "/tmp/myModel.xml";
+    std::string device = "GNA";
+    std::map<std::string, std::string> deviceConfig;
+//! [part2]
+    InferenceEngine::Core ie;                                  // Step 1: create Inference engine object
+    ie.SetConfig({{CONFIG_KEY(CACHE_DIR), "myCacheFolder"}});  // Step 1b: Enable caching
+    ie.LoadNetwork(modelPath, device, deviceConfig);           // Step 2: LoadNetwork by model file path
+//! [part2]
+return 0;
+}
diff --git a/docs/snippets/InferenceEngine_Caching3.cpp b/docs/snippets/InferenceEngine_Caching3.cpp
new file mode 100644
index 00000000000000..ce91a798552c79
--- /dev/null
+++ b/docs/snippets/InferenceEngine_Caching3.cpp
@@ -0,0 +1,20 @@
+#include <ie_core.hpp>
+
+int main() {
+using namespace InferenceEngine;
+    std::string modelPath = "/tmp/myModel.xml";
+    std::string deviceName = "GNA";
+    std::map<std::string, std::string> deviceConfig;
+    InferenceEngine::Core ie;
+//! [part3]
+    // Get list of supported metrics
+    std::vector<std::string> keys = ie.GetMetric(deviceName, METRIC_KEY(SUPPORTED_METRICS));
+
+    // Find 'IMPORT_EXPORT_SUPPORT' metric in supported metrics
+    auto it = std::find(keys.begin(), keys.end(), METRIC_KEY(IMPORT_EXPORT_SUPPORT));
+
+    // If metric 'IMPORT_EXPORT_SUPPORT' exists, check it's value
+    bool cachingSupported = (it != keys.end()) && ie.GetMetric(deviceName, METRIC_KEY(IMPORT_EXPORT_SUPPORT));
+//! [part3]
+    return 0;
+}

From 337abf1458ac7c7b349798a2e358f8eb784c1d6d Mon Sep 17 00:00:00 2001
From: Vladimir Paramuzov <vladimir.paramuzov@intel.com>
Date: Wed, 23 Jun 2021 09:34:42 +0300
Subject: [PATCH 169/178] [IE CLDNN] Fixed invalid stream reference in
 network_output if network is deleted (#6281)

---
 .../thirdparty/clDNN/api/cldnn/graph/network.hpp     | 12 +++++++-----
 inference-engine/thirdparty/clDNN/src/network.cpp    |  4 ++++
 2 files changed, 11 insertions(+), 5 deletions(-)

diff --git a/inference-engine/thirdparty/clDNN/api/cldnn/graph/network.hpp b/inference-engine/thirdparty/clDNN/api/cldnn/graph/network.hpp
index a6829259d79aba..61134cacc2faa5 100644
--- a/inference-engine/thirdparty/clDNN/api/cldnn/graph/network.hpp
+++ b/inference-engine/thirdparty/clDNN/api/cldnn/graph/network.hpp
@@ -36,8 +36,8 @@ struct network_output {
     memory::ptr get_memory() const {
         // TODO: in_order queue doesn't create proper output event in some cases which leads to syncronization issues with user app
         // So call finish for associated stream to enusre that the output data is ready.
-        if (_stream.get_queue_type() == queue_types::in_order) {
-            _stream.finish();
+        if (_stream->get_queue_type() == queue_types::in_order) {
+            _stream->finish();
         } else {
             _event->wait();
         }
@@ -47,8 +47,8 @@ struct network_output {
 private:
     event::ptr _event;
     memory::ptr _result;
-    stream& _stream;
-    network_output(event::ptr evt, memory::ptr mem, stream& stream) : _event(evt), _result(mem), _stream(stream) {}
+    stream::ptr _stream;
+    network_output(event::ptr evt, memory::ptr mem, stream::ptr stream) : _event(evt), _result(mem), _stream(stream) {}
     friend struct network;
 };
 
@@ -110,6 +110,8 @@ struct network {
 
     stream& get_stream() const;
 
+    stream::ptr get_stream_ptr() const;
+
     /// @brief Return internal network id.
     uint32_t get_id();
 
@@ -144,7 +146,7 @@ struct network {
 
     /// @brief Returns @ref network_output object for particular @p output. Can't be called before network execution
     network_output get_output(const primitive_id& output_id) const {
-        return network_output(get_primitive_event(output_id), get_output_memory(output_id), get_stream());
+        return network_output(get_primitive_event(output_id), get_output_memory(output_id), get_stream_ptr());
     }
 
     /// @brief Returns the list of @ref event for the primitives that were executed in network.
diff --git a/inference-engine/thirdparty/clDNN/src/network.cpp b/inference-engine/thirdparty/clDNN/src/network.cpp
index 14cfd2791a50de..b828261ff1cb62 100644
--- a/inference-engine/thirdparty/clDNN/src/network.cpp
+++ b/inference-engine/thirdparty/clDNN/src/network.cpp
@@ -69,6 +69,10 @@ stream& network::get_stream() const {
     return _impl->get_stream();
 }
 
+stream::ptr network::get_stream_ptr() const {
+    return _impl->get_stream_ptr();
+}
+
 uint32_t network::get_id() {
     return _impl->get_id();
 }

From 2f432980a08dcf0996d0da5a2da86941e365dd84 Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Wed, 23 Jun 2021 09:53:18 +0300
Subject: [PATCH 170/178] Moved gflags, preprocessing dependencies to root
 (#6309)

---
 .gitmodules                                   | 16 +--
 inference-engine/CMakeLists.txt               | 97 ++++++-------------
 inference-engine/samples/CMakeLists.txt       | 22 ++---
 inference-engine/samples/thirdparty/gflags    |  1 -
 .../src/inference_engine/CMakeLists.txt       |  2 +-
 .../src/legacy_api/CMakeLists.txt             |  2 +-
 .../CMakeLists.txt                            |  2 +-
 .../offline_transformations/CMakeLists.txt    |  2 +-
 .../src/preprocessing/CMakeLists.txt          |  4 +-
 inference-engine/src/snippets/CMakeLists.txt  |  2 +-
 .../src/transformations/CMakeLists.txt        |  2 +-
 inference-engine/thirdparty/CMakeLists.txt    |  7 --
 .../thirdparty/movidius/XLink/CMakeLists.txt  |  2 +-
 .../thirdparty/movidius/mvnc/CMakeLists.txt   |  3 +-
 tests/stress_tests/CMakeLists.txt             | 19 ++--
 tests/time_tests/CMakeLists.txt               |  3 -
 .../src/timetests_helper/CMakeLists.txt       | 20 ++--
 thirdparty/CMakeLists.txt                     | 59 ++++++++++-
 .../thirdparty => thirdparty}/ade             |  0
 .../thirdparty => thirdparty}/fluid/README.md |  0
 .../thirdparty => thirdparty}/fluid/check.sh  |  0
 .../fluid/checksum.sh                         |  0
 .../fluid/checksum.txt                        |  0
 .../fluid/modules/gapi/CMakeLists.txt         |  0
 .../modules/gapi/cmake/DownloadADE.cmake      |  0
 .../fluid/modules/gapi/cmake/init.cmake       |  0
 .../fluid/modules/gapi/cmake/standalone.cmake |  0
 .../fluid/modules/gapi/doc/00-root.markdown   |  0
 .../modules/gapi/doc/01-background.markdown   |  0
 .../fluid/modules/gapi/doc/10-hld-overview.md |  0
 .../modules/gapi/doc/20-kernel-api.markdown   |  0
 .../gapi/doc/30-implementation.markdown       |  0
 .../modules/gapi/doc/dot/kernel_hierarchy.dot |  0
 .../fluid/modules/gapi/doc/pics/demo.jpg      |  0
 .../modules/gapi/doc/pics/gapi_scheme.png     |  0
 .../gapi/doc/pics/kernel_hierarchy.png        |  0
 .../modules/gapi/doc/pics/render_example.png  |  0
 .../fluid/modules/gapi/doc/slides/README.md   |  0
 .../modules/gapi/doc/slides/gapi_overview.org |  0
 .../fluid/modules/gapi/doc/slides/get_sty.sh  |  0
 .../modules/gapi/doc/slides/ocv_logo.eps      |  0
 .../modules/gapi/include/opencv2/gapi.hpp     |  0
 .../gapi/include/opencv2/gapi/core.hpp        |  0
 .../gapi/include/opencv2/gapi/cpu/core.hpp    |  0
 .../include/opencv2/gapi/cpu/gcpukernel.hpp   |  0
 .../gapi/include/opencv2/gapi/cpu/imgproc.hpp |  0
 .../gapi/include/opencv2/gapi/cpu/video.hpp   |  0
 .../gapi/include/opencv2/gapi/fluid/core.hpp  |  0
 .../opencv2/gapi/fluid/gfluidbuffer.hpp       |  0
 .../opencv2/gapi/fluid/gfluidkernel.hpp       |  0
 .../include/opencv2/gapi/fluid/imgproc.hpp    |  0
 .../gapi/include/opencv2/gapi/garg.hpp        |  0
 .../gapi/include/opencv2/gapi/garray.hpp      |  0
 .../include/opencv2/gapi/gasync_context.hpp   |  0
 .../gapi/include/opencv2/gapi/gcall.hpp       |  0
 .../gapi/include/opencv2/gapi/gcommon.hpp     |  0
 .../gapi/include/opencv2/gapi/gcompiled.hpp   |  0
 .../include/opencv2/gapi/gcompiled_async.hpp  |  0
 .../include/opencv2/gapi/gcompoundkernel.hpp  |  0
 .../include/opencv2/gapi/gcomputation.hpp     |  0
 .../opencv2/gapi/gcomputation_async.hpp       |  0
 .../gapi/include/opencv2/gapi/gframe.hpp      |  0
 .../gapi/include/opencv2/gapi/gkernel.hpp     |  0
 .../gapi/include/opencv2/gapi/gmat.hpp        |  0
 .../gapi/include/opencv2/gapi/gmetaarg.hpp    |  0
 .../gapi/include/opencv2/gapi/gopaque.hpp     |  0
 .../gapi/include/opencv2/gapi/gproto.hpp      |  0
 .../gapi/include/opencv2/gapi/gpu/core.hpp    |  0
 .../include/opencv2/gapi/gpu/ggpukernel.hpp   |  0
 .../gapi/include/opencv2/gapi/gpu/imgproc.hpp |  0
 .../gapi/include/opencv2/gapi/gscalar.hpp     |  0
 .../gapi/include/opencv2/gapi/gstreaming.hpp  |  0
 .../gapi/include/opencv2/gapi/gtransform.hpp  |  0
 .../include/opencv2/gapi/gtype_traits.hpp     |  0
 .../gapi/include/opencv2/gapi/gtyped.hpp      |  0
 .../gapi/include/opencv2/gapi/imgproc.hpp     |  0
 .../gapi/include/opencv2/gapi/infer.hpp       |  0
 .../opencv2/gapi/infer/bindings_ie.hpp        |  0
 .../gapi/include/opencv2/gapi/infer/ie.hpp    |  0
 .../include/opencv2/gapi/infer/ie/util.hpp    |  0
 .../gapi/include/opencv2/gapi/infer/onnx.hpp  |  0
 .../include/opencv2/gapi/infer/parsers.hpp    |  0
 .../gapi/include/opencv2/gapi/media.hpp       |  0
 .../gapi/include/opencv2/gapi/ocl/core.hpp    |  0
 .../include/opencv2/gapi/ocl/goclkernel.hpp   |  0
 .../gapi/include/opencv2/gapi/ocl/imgproc.hpp |  0
 .../include/opencv2/gapi/opencv_includes.hpp  |  0
 .../gapi/include/opencv2/gapi/operators.hpp   |  0
 .../gapi/include/opencv2/gapi/own/assert.hpp  |  0
 .../gapi/include/opencv2/gapi/own/convert.hpp |  0
 .../gapi/include/opencv2/gapi/own/cvdefs.hpp  |  0
 .../gapi/include/opencv2/gapi/own/exports.hpp |  0
 .../gapi/include/opencv2/gapi/own/mat.hpp     |  0
 .../include/opencv2/gapi/own/saturate.hpp     |  0
 .../gapi/include/opencv2/gapi/own/scalar.hpp  |  0
 .../gapi/include/opencv2/gapi/own/types.hpp   |  0
 .../include/opencv2/gapi/plaidml/core.hpp     |  0
 .../opencv2/gapi/plaidml/gplaidmlkernel.hpp   |  0
 .../include/opencv2/gapi/plaidml/plaidml.hpp  |  0
 .../gapi/include/opencv2/gapi/render.hpp      |  0
 .../include/opencv2/gapi/render/render.hpp    |  0
 .../opencv2/gapi/render/render_types.hpp      |  0
 .../gapi/include/opencv2/gapi/rmat.hpp        |  0
 .../gapi/include/opencv2/gapi/s11n.hpp        |  0
 .../gapi/include/opencv2/gapi/s11n/base.hpp   |  0
 .../include/opencv2/gapi/streaming/cap.hpp    |  0
 .../include/opencv2/gapi/streaming/desync.hpp |  0
 .../include/opencv2/gapi/streaming/format.hpp |  0
 .../include/opencv2/gapi/streaming/meta.hpp   |  0
 .../include/opencv2/gapi/streaming/source.hpp |  0
 .../gapi/include/opencv2/gapi/util/any.hpp    |  0
 .../opencv2/gapi/util/compiler_hints.hpp      |  0
 .../opencv2/gapi/util/copy_through_move.hpp   |  0
 .../include/opencv2/gapi/util/optional.hpp    |  0
 .../gapi/include/opencv2/gapi/util/throw.hpp  |  0
 .../include/opencv2/gapi/util/type_traits.hpp |  0
 .../gapi/include/opencv2/gapi/util/util.hpp   |  0
 .../include/opencv2/gapi/util/variant.hpp     |  0
 .../gapi/include/opencv2/gapi/video.hpp       |  0
 .../gapi/misc/python/pyopencv_gapi.hpp        |  0
 .../modules/gapi/misc/python/shadow_gapi.hpp  |  0
 .../gapi/misc/python/test/test_gapi_core.py   |  0
 .../misc/python/test/test_gapi_imgproc.py     |  0
 .../gapi/misc/python/test/test_gapi_infer.py  |  0
 .../python/test/test_gapi_sample_pipelines.py |  0
 .../misc/python/test/test_gapi_streaming.py   |  0
 .../gapi/perf/common/gapi_core_perf_tests.cpp |  0
 .../gapi/perf/common/gapi_core_perf_tests.hpp |  0
 .../perf/common/gapi_core_perf_tests_inl.hpp  |  0
 .../perf/common/gapi_imgproc_perf_tests.cpp   |  0
 .../perf/common/gapi_imgproc_perf_tests.hpp   |  0
 .../common/gapi_imgproc_perf_tests_inl.hpp    |  0
 .../perf/common/gapi_render_perf_tests.cpp    |  0
 .../perf/common/gapi_render_perf_tests.hpp    |  0
 .../common/gapi_render_perf_tests_inl.hpp     |  0
 .../perf/common/gapi_video_perf_tests.cpp     |  0
 .../perf/common/gapi_video_perf_tests.hpp     |  0
 .../perf/common/gapi_video_perf_tests_inl.hpp |  0
 .../perf/cpu/gapi_core_perf_tests_cpu.cpp     |  0
 .../perf/cpu/gapi_core_perf_tests_fluid.cpp   |  0
 .../perf/cpu/gapi_imgproc_perf_tests_cpu.cpp  |  0
 .../cpu/gapi_imgproc_perf_tests_fluid.cpp     |  0
 .../perf/cpu/gapi_video_perf_tests_cpu.cpp    |  0
 .../perf/gpu/gapi_core_perf_tests_gpu.cpp     |  0
 .../perf/gpu/gapi_imgproc_perf_tests_gpu.cpp  |  0
 .../internal/gapi_compiler_perf_tests.cpp     |  0
 .../fluid/modules/gapi/perf/perf_bench.cpp    |  0
 .../fluid/modules/gapi/perf/perf_main.cpp     |  0
 .../fluid/modules/gapi/perf/perf_precomp.hpp  |  0
 .../render/gapi_render_perf_tests_ocv.cpp     |  0
 .../modules/gapi/samples/api_example.cpp      |  0
 .../modules/gapi/samples/api_ref_snippets.cpp |  0
 .../modules/gapi/samples/draw_example.cpp     |  0
 .../modules/gapi/samples/dynamic_graph.cpp    |  0
 .../modules/gapi/samples/gaze_estimation.cpp  |  0
 .../gapi/samples/infer_ie_onnx_hybrid.cpp     |  0
 .../modules/gapi/samples/infer_single_roi.cpp |  0
 .../modules/gapi/samples/infer_ssd_onnx.cpp   |  0
 .../gapi/samples/kernel_api_snippets.cpp      |  0
 .../gapi/samples/privacy_masking_camera.cpp   |  0
 .../modules/gapi/samples/slides_blur_gapi.cpp |  0
 .../modules/gapi/samples/slides_sobel_cv.cpp  |  0
 .../gapi/samples/slides_sobel_gapi.cpp        |  0
 .../modules/gapi/samples/text_detection.cpp   |  0
 .../gapi/scripts/measure_privacy_masking.py   |  0
 .../fluid/modules/gapi/src/api/README.md      |  0
 .../fluid/modules/gapi/src/api/ft_render.cpp  |  0
 .../fluid/modules/gapi/src/api/ft_render.hpp  |  0
 .../modules/gapi/src/api/ft_render_priv.hpp   |  0
 .../fluid/modules/gapi/src/api/garray.cpp     |  0
 .../fluid/modules/gapi/src/api/gbackend.cpp   |  0
 .../modules/gapi/src/api/gbackend_priv.hpp    |  0
 .../fluid/modules/gapi/src/api/gcall.cpp      |  0
 .../fluid/modules/gapi/src/api/gcall_priv.hpp |  0
 .../modules/gapi/src/api/gcomputation.cpp     |  0
 .../gapi/src/api/gcomputation_priv.hpp        |  0
 .../fluid/modules/gapi/src/api/gframe.cpp     |  0
 .../fluid/modules/gapi/src/api/ginfer.cpp     |  0
 .../fluid/modules/gapi/src/api/gkernel.cpp    |  0
 .../fluid/modules/gapi/src/api/gmat.cpp       |  0
 .../fluid/modules/gapi/src/api/gnode.cpp      |  0
 .../fluid/modules/gapi/src/api/gnode.hpp      |  0
 .../fluid/modules/gapi/src/api/gnode_priv.hpp |  0
 .../fluid/modules/gapi/src/api/gopaque.cpp    |  0
 .../fluid/modules/gapi/src/api/gorigin.cpp    |  0
 .../fluid/modules/gapi/src/api/gorigin.hpp    |  0
 .../fluid/modules/gapi/src/api/gproto.cpp     |  0
 .../modules/gapi/src/api/gproto_priv.hpp      |  0
 .../fluid/modules/gapi/src/api/grunarg.cpp    |  0
 .../fluid/modules/gapi/src/api/gscalar.cpp    |  0
 .../modules/gapi/src/api/kernels_core.cpp     |  0
 .../modules/gapi/src/api/kernels_imgproc.cpp  |  0
 .../gapi/src/api/kernels_nnparsers.cpp        |  0
 .../gapi/src/api/kernels_streaming.cpp        |  0
 .../modules/gapi/src/api/kernels_video.cpp    |  0
 .../fluid/modules/gapi/src/api/media.cpp      |  0
 .../modules/gapi/src/api/ocv_mask_creator.hpp |  0
 .../fluid/modules/gapi/src/api/operators.cpp  |  0
 .../fluid/modules/gapi/src/api/render.cpp     |  0
 .../fluid/modules/gapi/src/api/render_ocv.cpp |  0
 .../fluid/modules/gapi/src/api/render_ocv.hpp |  0
 .../modules/gapi/src/api/render_priv.hpp      |  0
 .../fluid/modules/gapi/src/api/rmat.cpp       |  0
 .../fluid/modules/gapi/src/api/s11n.cpp       |  0
 .../fluid/modules/gapi/src/backends/README.md |  0
 .../gapi/src/backends/common/gbackend.hpp     |  0
 .../src/backends/common/gcompoundbackend.cpp  |  0
 .../src/backends/common/gcompoundkernel.cpp   |  0
 .../gapi/src/backends/common/gmetabackend.cpp |  0
 .../gapi/src/backends/common/gmetabackend.hpp |  0
 .../src/backends/common/serialization.cpp     |  0
 .../src/backends/common/serialization.hpp     |  0
 .../gapi/src/backends/cpu/gcpubackend.cpp     |  0
 .../gapi/src/backends/cpu/gcpubackend.hpp     |  0
 .../gapi/src/backends/cpu/gcpucore.cpp        |  0
 .../gapi/src/backends/cpu/gcpuimgproc.cpp     |  0
 .../gapi/src/backends/cpu/gcpukernel.cpp      |  0
 .../gapi/src/backends/cpu/gcpuvideo.cpp       |  0
 .../gapi/src/backends/cpu/gnnparsers.cpp      |  0
 .../gapi/src/backends/cpu/gnnparsers.hpp      |  0
 .../gapi/src/backends/fluid/gfluidbackend.cpp |  0
 .../gapi/src/backends/fluid/gfluidbackend.hpp |  0
 .../gapi/src/backends/fluid/gfluidbuffer.cpp  |  0
 .../src/backends/fluid/gfluidbuffer_priv.hpp  |  0
 .../gapi/src/backends/fluid/gfluidcore.cpp    |  0
 .../gapi/src/backends/fluid/gfluidimgproc.cpp |  0
 .../fluid/gfluidimgproc_func.dispatch.cpp     |  0
 .../src/backends/fluid/gfluidimgproc_func.hpp |  0
 .../fluid/gfluidimgproc_func.simd.hpp         |  0
 .../gapi/src/backends/fluid/gfluidutils.hpp   |  0
 .../gapi/src/backends/ie/bindings_ie.cpp      |  0
 .../gapi/src/backends/ie/giebackend.cpp       |  0
 .../gapi/src/backends/ie/giebackend.hpp       |  0
 .../src/backends/ie/giebackend/giewrapper.cpp |  0
 .../src/backends/ie/giebackend/giewrapper.hpp |  0
 .../modules/gapi/src/backends/ie/util.hpp     |  0
 .../gapi/src/backends/ocl/goclbackend.cpp     |  0
 .../gapi/src/backends/ocl/goclbackend.hpp     |  0
 .../gapi/src/backends/ocl/goclcore.cpp        |  0
 .../gapi/src/backends/ocl/goclcore.hpp        |  0
 .../gapi/src/backends/ocl/goclimgproc.cpp     |  0
 .../gapi/src/backends/ocl/goclimgproc.hpp     |  0
 .../gapi/src/backends/ocl/goclkernel.cpp      |  0
 .../gapi/src/backends/onnx/gonnxbackend.cpp   |  0
 .../gapi/src/backends/onnx/gonnxbackend.hpp   |  0
 .../src/backends/plaidml/gplaidmlbackend.cpp  |  0
 .../src/backends/plaidml/gplaidmlbackend.hpp  |  0
 .../src/backends/plaidml/gplaidmlcore.cpp     |  0
 .../src/backends/plaidml/plaidml_util.hpp     |  0
 .../gapi/src/backends/render/ft_render.cpp    |  0
 .../gapi/src/backends/render/ft_render.hpp    |  0
 .../src/backends/render/ft_render_priv.hpp    |  0
 .../gapi/src/backends/render/grenderocv.cpp   |  0
 .../gapi/src/backends/render/grenderocv.hpp   |  0
 .../src/backends/render/grenderocvbackend.cpp |  0
 .../src/backends/render/grenderocvbackend.hpp |  0
 .../backends/streaming/gstreamingbackend.cpp  |  0
 .../backends/streaming/gstreamingbackend.hpp  |  0
 .../backends/streaming/gstreamingkernel.hpp   |  0
 .../fluid/modules/gapi/src/compiler/README.md |  0
 .../modules/gapi/src/compiler/gcompiled.cpp   |  0
 .../gapi/src/compiler/gcompiled_priv.hpp      |  0
 .../modules/gapi/src/compiler/gcompiler.cpp   |  0
 .../modules/gapi/src/compiler/gcompiler.hpp   |  0
 .../gapi/src/compiler/gislandmodel.cpp        |  0
 .../gapi/src/compiler/gislandmodel.hpp        |  0
 .../modules/gapi/src/compiler/gmodel.cpp      |  0
 .../modules/gapi/src/compiler/gmodel.hpp      |  0
 .../modules/gapi/src/compiler/gmodel_priv.hpp |  0
 .../gapi/src/compiler/gmodelbuilder.cpp       |  0
 .../gapi/src/compiler/gmodelbuilder.hpp       |  0
 .../modules/gapi/src/compiler/gobjref.hpp     |  0
 .../modules/gapi/src/compiler/gstreaming.cpp  |  0
 .../gapi/src/compiler/gstreaming_priv.hpp     |  0
 .../gapi/src/compiler/passes/dump_dot.cpp     |  0
 .../modules/gapi/src/compiler/passes/exec.cpp |  0
 .../gapi/src/compiler/passes/helpers.cpp      |  0
 .../gapi/src/compiler/passes/helpers.hpp      |  0
 .../gapi/src/compiler/passes/intrin.cpp       |  0
 .../gapi/src/compiler/passes/islands.cpp      |  0
 .../gapi/src/compiler/passes/kernels.cpp      |  0
 .../modules/gapi/src/compiler/passes/meta.cpp |  0
 .../gapi/src/compiler/passes/passes.hpp       |  0
 .../src/compiler/passes/pattern_matching.cpp  |  0
 .../src/compiler/passes/pattern_matching.hpp  |  0
 .../compiler/passes/perform_substitution.cpp  |  0
 .../gapi/src/compiler/passes/streaming.cpp    |  0
 .../src/compiler/passes/transformations.cpp   |  0
 .../gapi/src/compiler/transactions.hpp        |  0
 .../modules/gapi/src/executor/conc_queue.hpp  |  0
 .../modules/gapi/src/executor/gapi_itt.hpp    |  0
 .../modules/gapi/src/executor/gasync.cpp      |  0
 .../modules/gapi/src/executor/gexecutor.cpp   |  0
 .../modules/gapi/src/executor/gexecutor.hpp   |  0
 .../gapi/src/executor/gstreamingexecutor.cpp  |  0
 .../gapi/src/executor/gstreamingexecutor.hpp  |  0
 .../gapi/src/executor/gtbbexecutor.cpp        |  0
 .../gapi/src/executor/gtbbexecutor.hpp        |  0
 .../modules/gapi/src/executor/last_value.hpp  |  0
 .../fluid/modules/gapi/src/logger.hpp         |  0
 .../fluid/modules/gapi/src/precomp.hpp        |  0
 .../test/common/gapi_compoundkernel_tests.cpp |  0
 .../gapi/test/common/gapi_core_tests.cpp      |  0
 .../gapi/test/common/gapi_core_tests.hpp      |  0
 .../gapi/test/common/gapi_core_tests_inl.hpp  |  0
 .../gapi/test/common/gapi_imgproc_tests.cpp   |  0
 .../gapi/test/common/gapi_imgproc_tests.hpp   |  0
 .../test/common/gapi_imgproc_tests_inl.hpp    |  0
 .../gapi/test/common/gapi_operators_tests.cpp |  0
 .../gapi/test/common/gapi_operators_tests.hpp |  0
 .../test/common/gapi_operators_tests_inl.hpp  |  0
 .../test/common/gapi_parsers_tests_common.hpp |  0
 .../gapi/test/common/gapi_render_tests.cpp    |  0
 .../gapi/test/common/gapi_render_tests.hpp    |  0
 .../test/common/gapi_render_tests_inl.hpp     |  0
 .../gapi/test/common/gapi_tests_common.hpp    |  0
 .../gapi/test/common/gapi_tests_helpers.hpp   |  0
 .../gapi/test/common/gapi_video_tests.cpp     |  0
 .../gapi/test/common/gapi_video_tests.hpp     |  0
 .../test/common/gapi_video_tests_common.hpp   |  0
 .../gapi/test/common/gapi_video_tests_inl.hpp |  0
 .../gapi/test/cpu/gapi_core_tests_cpu.cpp     |  0
 .../gapi/test/cpu/gapi_core_tests_fluid.cpp   |  0
 .../gapi/test/cpu/gapi_imgproc_tests_cpu.cpp  |  0
 .../test/cpu/gapi_imgproc_tests_fluid.cpp     |  0
 .../gapi_ocv_stateful_kernel_test_utils.hpp   |  0
 .../cpu/gapi_ocv_stateful_kernel_tests.cpp    |  0
 .../test/cpu/gapi_operators_tests_cpu.cpp     |  0
 .../test/cpu/gapi_operators_tests_fluid.cpp   |  0
 .../gapi/test/cpu/gapi_render_tests_cpu.cpp   |  0
 .../gapi/test/cpu/gapi_video_tests_cpu.cpp    |  0
 .../executor/gtbbexecutor_internal_tests.cpp  |  0
 .../modules/gapi/test/gapi_array_tests.cpp    |  0
 .../modules/gapi/test/gapi_async_test.cpp     |  0
 .../gapi/test/gapi_basic_hetero_tests.cpp     |  0
 .../gapi/test/gapi_compile_args_tests.cpp     |  0
 .../modules/gapi/test/gapi_desc_tests.cpp     |  0
 .../test/gapi_fluid_parallel_rois_test.cpp    |  0
 .../gapi/test/gapi_fluid_resize_test.cpp      |  0
 .../modules/gapi/test/gapi_fluid_roi_test.cpp |  0
 .../modules/gapi/test/gapi_fluid_test.cpp     |  0
 .../gapi/test/gapi_fluid_test_kernels.cpp     |  0
 .../gapi/test/gapi_fluid_test_kernels.hpp     |  0
 .../modules/gapi/test/gapi_frame_tests.cpp    |  0
 .../gapi/test/gapi_gcompiled_tests.cpp        |  0
 .../gapi/test/gapi_gcomputation_tests.cpp     |  0
 .../fluid/modules/gapi/test/gapi_gpu_test.cpp |  0
 .../gapi/test/gapi_graph_meta_tests.cpp       |  0
 .../modules/gapi/test/gapi_kernel_tests.cpp   |  0
 .../modules/gapi/test/gapi_mock_kernels.hpp   |  0
 .../modules/gapi/test/gapi_opaque_tests.cpp   |  0
 .../gapi/test/gapi_plaidml_pipelines.cpp      |  0
 .../modules/gapi/test/gapi_planar_test.cpp    |  0
 .../gapi/test/gapi_sample_pipelines.cpp       |  0
 .../modules/gapi/test/gapi_scalar_tests.cpp   |  0
 .../modules/gapi/test/gapi_smoke_test.cpp     |  0
 .../gapi/test/gapi_transform_tests.cpp        |  0
 .../modules/gapi/test/gapi_typed_tests.cpp    |  0
 .../modules/gapi/test/gapi_util_tests.cpp     |  0
 .../gapi/test/gpu/gapi_core_tests_gpu.cpp     |  0
 .../gapi/test/gpu/gapi_imgproc_tests_gpu.cpp  |  0
 .../test/gpu/gapi_operators_tests_gpu.cpp     |  0
 .../gapi/test/infer/gapi_infer_ie_test.cpp    |  0
 .../gapi/test/infer/gapi_infer_onnx_test.cpp  |  0
 .../gapi/test/infer/gapi_infer_tests.cpp      |  0
 .../test/internal/gapi_int_backend_tests.cpp  |  0
 .../test/internal/gapi_int_dynamic_graph.cpp  |  0
 .../test/internal/gapi_int_executor_tests.cpp |  0
 .../gapi/test/internal/gapi_int_garg_test.cpp |  0
 .../test/internal/gapi_int_gmetaarg_test.cpp  |  0
 .../internal/gapi_int_gmodel_builder_test.cpp |  0
 .../internal/gapi_int_island_fusion_tests.cpp |  0
 .../test/internal/gapi_int_island_tests.cpp   |  0
 .../gapi_int_pattern_matching_test.cpp        |  0
 .../gapi_int_perform_substitution_test.cpp    |  0
 .../test/internal/gapi_int_proto_tests.cpp    |  0
 .../internal/gapi_int_recompilation_test.cpp  |  0
 .../test/internal/gapi_int_vectorref_test.cpp |  0
 .../test/internal/gapi_transactions_test.cpp  |  0
 .../gapi/test/opencl_kernels_test_gapi.hpp    |  0
 .../gapi/test/own/conc_queue_tests.cpp        |  0
 .../gapi/test/own/gapi_types_tests.cpp        |  0
 .../test/own/last_written_value_tests.cpp     |  0
 .../fluid/modules/gapi/test/own/mat_tests.cpp |  0
 .../modules/gapi/test/own/scalar_tests.cpp    |  0
 .../gapi/test/render/ftp_render_test.cpp      |  0
 .../test/render/gapi_render_tests_ocv.cpp     |  0
 .../gapi/test/rmat/rmat_integration_tests.cpp |  0
 .../gapi/test/rmat/rmat_test_common.hpp       |  0
 .../modules/gapi/test/rmat/rmat_tests.cpp     |  0
 .../gapi/test/rmat/rmat_view_tests.cpp        |  0
 .../gapi/test/s11n/gapi_s11n_tests.cpp        |  0
 .../test/s11n/gapi_sample_pipelines_s11n.cpp  |  0
 .../test/streaming/gapi_streaming_tests.cpp   |  0
 .../fluid/modules/gapi/test/test_main.cpp     |  0
 .../fluid/modules/gapi/test/test_precomp.hpp  |  0
 .../modules/gapi/test/util/any_tests.cpp      |  0
 .../modules/gapi/test/util/optional_tests.cpp |  0
 .../modules/gapi/test/util/variant_tests.cpp  |  0
 .../fluid/revision.txt                        |  0
 .../thirdparty => thirdparty}/fluid/update.sh |  0
 thirdparty/gflags                             |  1 +
 .../ocv/opencv_hal.hpp                        |  0
 .../ocv/opencv_hal_avx.hpp                    |  0
 .../ocv/opencv_hal_avx512.hpp                 |  0
 .../ocv/opencv_hal_intrin.hpp                 |  0
 .../ocv/opencv_hal_intrin_forward.hpp         |  0
 .../ocv/opencv_hal_neon.hpp                   |  0
 .../ocv/opencv_hal_sse.hpp                    |  0
 409 files changed, 127 insertions(+), 139 deletions(-)
 delete mode 160000 inference-engine/samples/thirdparty/gflags
 rename {inference-engine/thirdparty => thirdparty}/ade (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/README.md (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/check.sh (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/checksum.sh (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/checksum.txt (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/CMakeLists.txt (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/cmake/DownloadADE.cmake (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/cmake/init.cmake (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/cmake/standalone.cmake (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/00-root.markdown (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/01-background.markdown (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/10-hld-overview.md (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/20-kernel-api.markdown (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/30-implementation.markdown (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/dot/kernel_hierarchy.dot (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/pics/demo.jpg (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/pics/gapi_scheme.png (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/pics/kernel_hierarchy.png (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/pics/render_example.png (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/slides/README.md (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/slides/gapi_overview.org (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/slides/get_sty.sh (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/doc/slides/ocv_logo.eps (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/core.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/cpu/core.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/cpu/gcpukernel.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/cpu/imgproc.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/cpu/video.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/fluid/core.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/fluid/gfluidbuffer.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/fluid/gfluidkernel.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/fluid/imgproc.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/garg.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/garray.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gasync_context.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gcall.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gcommon.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gcompiled.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gcompiled_async.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gcompoundkernel.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gcomputation.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gcomputation_async.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gframe.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gkernel.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gmat.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gmetaarg.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gopaque.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gproto.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gpu/core.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gpu/ggpukernel.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gpu/imgproc.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gscalar.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gstreaming.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gtransform.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gtype_traits.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/gtyped.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/imgproc.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/infer.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/infer/bindings_ie.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/infer/ie.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/infer/ie/util.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/infer/onnx.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/infer/parsers.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/media.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/ocl/core.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/ocl/goclkernel.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/ocl/imgproc.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/opencv_includes.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/operators.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/own/assert.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/own/convert.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/own/cvdefs.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/own/exports.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/own/mat.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/own/saturate.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/own/scalar.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/own/types.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/plaidml/core.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/plaidml/gplaidmlkernel.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/plaidml/plaidml.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/render.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/render/render.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/render/render_types.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/rmat.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/s11n.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/s11n/base.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/streaming/cap.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/streaming/desync.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/streaming/format.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/streaming/meta.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/streaming/source.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/util/any.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/util/compiler_hints.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/util/copy_through_move.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/util/optional.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/util/throw.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/util/type_traits.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/util/util.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/util/variant.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/include/opencv2/gapi/video.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/misc/python/pyopencv_gapi.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/misc/python/shadow_gapi.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/misc/python/test/test_gapi_core.py (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/misc/python/test/test_gapi_imgproc.py (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/misc/python/test/test_gapi_infer.py (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/misc/python/test/test_gapi_sample_pipelines.py (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/misc/python/test/test_gapi_streaming.py (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_core_perf_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_core_perf_tests.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_core_perf_tests_inl.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests_inl.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_render_perf_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_render_perf_tests.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_render_perf_tests_inl.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_video_perf_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_video_perf_tests.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/common/gapi_video_perf_tests_inl.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/cpu/gapi_core_perf_tests_cpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/cpu/gapi_core_perf_tests_fluid.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/cpu/gapi_imgproc_perf_tests_cpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/cpu/gapi_imgproc_perf_tests_fluid.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/cpu/gapi_video_perf_tests_cpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/gpu/gapi_core_perf_tests_gpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/gpu/gapi_imgproc_perf_tests_gpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/internal/gapi_compiler_perf_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/perf_bench.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/perf_main.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/perf_precomp.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/perf/render/gapi_render_perf_tests_ocv.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/api_example.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/api_ref_snippets.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/draw_example.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/dynamic_graph.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/gaze_estimation.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/infer_ie_onnx_hybrid.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/infer_single_roi.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/infer_ssd_onnx.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/kernel_api_snippets.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/privacy_masking_camera.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/slides_blur_gapi.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/slides_sobel_cv.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/slides_sobel_gapi.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/samples/text_detection.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/scripts/measure_privacy_masking.py (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/README.md (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/ft_render.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/ft_render.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/ft_render_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/garray.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gbackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gbackend_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gcall.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gcall_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gcomputation.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gcomputation_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gframe.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/ginfer.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gkernel.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gmat.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gnode.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gnode.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gnode_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gopaque.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gorigin.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gorigin.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gproto.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gproto_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/grunarg.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/gscalar.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/kernels_core.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/kernels_imgproc.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/kernels_nnparsers.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/kernels_streaming.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/kernels_video.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/media.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/ocv_mask_creator.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/operators.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/render.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/render_ocv.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/render_ocv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/render_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/rmat.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/api/s11n.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/README.md (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/common/gbackend.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/common/gcompoundbackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/common/gcompoundkernel.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/common/gmetabackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/common/gmetabackend.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/common/serialization.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/common/serialization.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/cpu/gcpubackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/cpu/gcpubackend.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/cpu/gcpucore.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/cpu/gcpuimgproc.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/cpu/gcpukernel.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/cpu/gcpuvideo.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/cpu/gnnparsers.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/cpu/gnnparsers.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/fluid/gfluidbackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/fluid/gfluidbackend.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/fluid/gfluidbuffer.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/fluid/gfluidbuffer_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/fluid/gfluidcore.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/fluid/gfluidimgproc.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.dispatch.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.simd.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/fluid/gfluidutils.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ie/bindings_ie.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ie/giebackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ie/giebackend.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ie/giebackend/giewrapper.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ie/giebackend/giewrapper.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ie/util.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ocl/goclbackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ocl/goclbackend.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ocl/goclcore.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ocl/goclcore.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ocl/goclimgproc.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ocl/goclimgproc.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/ocl/goclkernel.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/onnx/gonnxbackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/onnx/gonnxbackend.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/plaidml/gplaidmlbackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/plaidml/gplaidmlbackend.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/plaidml/gplaidmlcore.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/plaidml/plaidml_util.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/render/ft_render.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/render/ft_render.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/render/ft_render_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/render/grenderocv.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/render/grenderocv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/render/grenderocvbackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/render/grenderocvbackend.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/streaming/gstreamingbackend.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/streaming/gstreamingbackend.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/backends/streaming/gstreamingkernel.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/README.md (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gcompiled.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gcompiled_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gcompiler.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gcompiler.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gislandmodel.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gislandmodel.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gmodel.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gmodel.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gmodel_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gmodelbuilder.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gmodelbuilder.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gobjref.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gstreaming.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/gstreaming_priv.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/dump_dot.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/exec.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/helpers.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/helpers.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/intrin.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/islands.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/kernels.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/meta.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/passes.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/pattern_matching.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/pattern_matching.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/perform_substitution.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/streaming.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/passes/transformations.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/compiler/transactions.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/executor/conc_queue.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/executor/gapi_itt.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/executor/gasync.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/executor/gexecutor.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/executor/gexecutor.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/executor/gstreamingexecutor.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/executor/gstreamingexecutor.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/executor/gtbbexecutor.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/executor/gtbbexecutor.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/executor/last_value.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/logger.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/src/precomp.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_compoundkernel_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_core_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_core_tests.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_core_tests_inl.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_imgproc_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_imgproc_tests.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_imgproc_tests_inl.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_operators_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_operators_tests.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_operators_tests_inl.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_parsers_tests_common.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_render_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_render_tests.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_render_tests_inl.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_tests_common.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_tests_helpers.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_video_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_video_tests.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_video_tests_common.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/common/gapi_video_tests_inl.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/cpu/gapi_core_tests_cpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/cpu/gapi_core_tests_fluid.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/cpu/gapi_imgproc_tests_cpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/cpu/gapi_imgproc_tests_fluid.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/cpu/gapi_ocv_stateful_kernel_test_utils.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/cpu/gapi_ocv_stateful_kernel_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/cpu/gapi_operators_tests_cpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/cpu/gapi_operators_tests_fluid.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/cpu/gapi_render_tests_cpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/cpu/gapi_video_tests_cpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/executor/gtbbexecutor_internal_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_array_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_async_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_basic_hetero_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_compile_args_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_desc_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_fluid_parallel_rois_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_fluid_resize_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_fluid_roi_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_fluid_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_fluid_test_kernels.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_fluid_test_kernels.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_frame_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_gcompiled_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_gcomputation_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_gpu_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_graph_meta_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_kernel_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_mock_kernels.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_opaque_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_plaidml_pipelines.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_planar_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_sample_pipelines.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_scalar_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_smoke_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_transform_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_typed_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gapi_util_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gpu/gapi_core_tests_gpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gpu/gapi_imgproc_tests_gpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/gpu/gapi_operators_tests_gpu.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/infer/gapi_infer_ie_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/infer/gapi_infer_onnx_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/infer/gapi_infer_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_backend_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_dynamic_graph.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_executor_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_garg_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_gmetaarg_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_gmodel_builder_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_island_fusion_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_island_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_pattern_matching_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_perform_substitution_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_proto_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_recompilation_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_int_vectorref_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/internal/gapi_transactions_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/opencl_kernels_test_gapi.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/own/conc_queue_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/own/gapi_types_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/own/last_written_value_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/own/mat_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/own/scalar_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/render/ftp_render_test.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/render/gapi_render_tests_ocv.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/rmat/rmat_integration_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/rmat/rmat_test_common.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/rmat/rmat_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/rmat/rmat_view_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/s11n/gapi_s11n_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/s11n/gapi_sample_pipelines_s11n.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/streaming/gapi_streaming_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/test_main.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/test_precomp.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/util/any_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/util/optional_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/modules/gapi/test/util/variant_tests.cpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/revision.txt (100%)
 rename {inference-engine/thirdparty => thirdparty}/fluid/update.sh (100%)
 create mode 160000 thirdparty/gflags
 rename {inference-engine/thirdparty => thirdparty}/ocv/opencv_hal.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/ocv/opencv_hal_avx.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/ocv/opencv_hal_avx512.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/ocv/opencv_hal_intrin.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/ocv/opencv_hal_intrin_forward.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/ocv/opencv_hal_neon.hpp (100%)
 rename {inference-engine/thirdparty => thirdparty}/ocv/opencv_hal_sse.hpp (100%)

diff --git a/.gitmodules b/.gitmodules
index 586502884302ac..c2cb47a6158025 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -1,15 +1,7 @@
-[submodule "inference-engine/thirdparty/ade"]
-	path = inference-engine/thirdparty/ade
-	url = https://github.com/opencv/ade.git
-	ignore = dirty
 [submodule "inference-engine/thirdparty/mkl-dnn"]
 	path = inference-engine/thirdparty/mkl-dnn
 	url = https://github.com/openvinotoolkit/oneDNN.git
 	ignore = dirty
-[submodule "inference-engine/samples/thirdparty/gflags"]
-	path = inference-engine/samples/thirdparty/gflags
-	url = https://github.com/gflags/gflags.git
-	ignore = dirty
 [submodule "thirdparty/xbyak"]
 	path = thirdparty/xbyak
 	url = https://github.com/herumi/xbyak.git
@@ -18,6 +10,14 @@
 	path = thirdparty/zlib/zlib
 	url = https://github.com/madler/zlib.git
 	ignore = dirty
+[submodule "thirdparty/ade"]
+	path = thirdparty/ade
+	url = https://github.com/opencv/ade.git
+	ignore = dirty
+[submodule "thirdparty/gflags"]
+	path = thirdparty/gflags
+	url = https://github.com/gflags/gflags.git
+	ignore = dirty
 [submodule "thirdparty/gtest"]
 	path = thirdparty/gtest
 	url = https://github.com/openvinotoolkit/googletest.git
diff --git a/inference-engine/CMakeLists.txt b/inference-engine/CMakeLists.txt
index 1ac7fd8bf62b4a..a11aac8d8a24de 100644
--- a/inference-engine/CMakeLists.txt
+++ b/inference-engine/CMakeLists.txt
@@ -13,55 +13,28 @@ include(cmake/features.cmake)
 # resolving dependencies for the project
 include(cmake/dependencies.cmake)
 
-function(ie_developer_export_targets)
-    openvino_developer_export_targets(COMPONENT inference_engine TARGETS ${ARGN})
-endfunction()
-
-function(ie_developer_export)
-    set(all_dev_targets gflags ie_libraries)
-    foreach(component IN LISTS openvino_export_components)
-        export(TARGETS ${${component}} NAMESPACE IE::
-               APPEND FILE "${CMAKE_BINARY_DIR}/${component}_dev_targets.cmake")
-        list(APPEND all_dev_targets ${${component}})
-    endforeach()
-
-    add_custom_target(ie_dev_targets ALL DEPENDS ${all_dev_targets})
-endfunction()
-
 add_subdirectory(thirdparty)
-
 add_subdirectory(src)
-
 add_subdirectory(ie_bridges/c)
 
-if(ENABLE_TESTS)
-    add_subdirectory(tests_deprecated)
-    add_subdirectory(tests)
+if(ENABLE_PYTHON)
+    add_subdirectory(ie_bridges/python)
 endif()
 
 add_subdirectory(tools)
+add_subdirectory(samples)
 
-function(ie_build_samples)
-    # samples should be build with the same flags as from OpenVINO package,
-    # so unset all flags
-    foreach(var CMAKE_CXX_FLAGS CMAKE_C_FLAGS CMAKE_CXX_STANDARD
-                CMAKE_EXE_LINKER_FLAGS CMAKE_POLICY_DEFAULT_CMP0063
-                CMAKE_CXX_VISIBILITY_PRESET CMAKE_C_VISIBILITY_PRESET
-                CMAKE_VISIBILITY_INLINES_HIDDEN CMAKE_POSITION_INDEPENDENT_CODE
-                THREADS_PREFER_PTHREAD_FLAG X86_64 X86 ARM AARCH64 LINUX
-                MINGW64 CMAKE_BUILD_TYPE CMAKE_MACOSX_RPATH)
-        unset(${var})
-    endforeach()
-    include("${IEDevScripts_DIR}/compile_flags/sanitizer.cmake")
-    add_subdirectory(samples)
-endfunction()
+if(ENABLE_TESTS)
+    add_subdirectory(tests_deprecated)
+    add_subdirectory(tests)
+endif()
 
-# gflags and format_reader targets are kept inside of samples directory and
-# they must be built even if samples build is disabled (required for tests and tools).
-ie_build_samples()
+#
+# Coverage
+#
 
-if(ENABLE_PYTHON)
-    add_subdirectory(ie_bridges/python)
+if(ENABLE_COVERAGE)
+    include(cmake/coverage.cmake)
 endif()
 
 #
@@ -70,19 +43,7 @@ endif()
 
 # install C++ samples
 
-ie_cpack_add_component(cpp_samples DEPENDS core)
-
-install(DIRECTORY ../thirdparty/zlib
-        DESTINATION ${IE_CPACK_IE_DIR}/samples/cpp/thirdparty
-        COMPONENT cpp_samples
-        USE_SOURCE_PERMISSIONS
-        PATTERN .clang-format EXCLUDE)
-
-install(DIRECTORY ../thirdparty/cnpy
-        DESTINATION ${IE_CPACK_IE_DIR}/samples/cpp/thirdparty
-        COMPONENT cpp_samples
-        USE_SOURCE_PERMISSIONS
-        PATTERN .clang-format EXCLUDE)
+ie_cpack_add_component(cpp_samples DEPENDS cpp_samples_deps core)
 
 if(UNIX)
     install(DIRECTORY samples/
@@ -142,7 +103,7 @@ endif()
 # Developer package
 #
 
-openvino_developer_export_targets(COMPONENT openvino_common TARGETS format_reader gflags ie_samples_utils)
+openvino_developer_export_targets(COMPONENT openvino_common TARGETS format_reader ie_samples_utils)
 
 # for Template plugin
 if(NGRAPH_INTERPRETER_ENABLE)
@@ -153,36 +114,34 @@ function(ie_generate_dev_package_config)
     # dummy check that OpenCV is here
     find_package(OpenCV QUIET)
 
-    ie_developer_export()
+    set(all_dev_targets gflags ie_libraries)
+    foreach(component IN LISTS openvino_export_components)
+        export(TARGETS ${${component}} NAMESPACE IE::
+            APPEND FILE "${CMAKE_BINARY_DIR}/${component}_dev_targets.cmake")
+        list(APPEND all_dev_targets ${${component}})
+    endforeach()
+    add_custom_target(ie_dev_targets ALL DEPENDS ${all_dev_targets})
 
     configure_package_config_file("${InferenceEngine_SOURCE_DIR}/cmake/templates/InferenceEngineDeveloperPackageConfig.cmake.in"
-                                "${CMAKE_BINARY_DIR}/InferenceEngineDeveloperPackageConfig.cmake"
-                                INSTALL_DESTINATION share # not used
-                                PATH_VARS "OpenVINO_MAIN_SOURCE_DIR;IE_MAIN_SOURCE_DIR;gflags_BINARY_DIR"
-                                NO_CHECK_REQUIRED_COMPONENTS_MACRO)
+                                  "${CMAKE_BINARY_DIR}/InferenceEngineDeveloperPackageConfig.cmake"
+                                  INSTALL_DESTINATION share # not used
+                                  PATH_VARS "OpenVINO_MAIN_SOURCE_DIR;IE_MAIN_SOURCE_DIR"
+                                  NO_CHECK_REQUIRED_COMPONENTS_MACRO)
 
     configure_file("${IE_MAIN_SOURCE_DIR}/cmake/templates/InferenceEngineConfig-version.cmake.in"
-                "${CMAKE_BINARY_DIR}/InferenceEngineDeveloperPackageConfig-version.cmake"
-                @ONLY)
+                   "${CMAKE_BINARY_DIR}/InferenceEngineDeveloperPackageConfig-version.cmake"
+                   @ONLY)
 endfunction()
 
 ie_generate_dev_package_config()
 
-#
-# Coverage
-#
-
-if(ENABLE_COVERAGE)
-    include(cmake/coverage.cmake)
-endif()
-
 #
 # Add extra modules
 #
 
 function(register_extra_modules)
     # post export
-    ie_developer_export_targets(inference_engine)
+    openvino_developer_export_targets(COMPONENT inference_engine TARGETS inference_engine)
     openvino_developer_export_targets(COMPONENT ngraph TARGETS ${NGRAPH_LIBRARIES})
 
     set(InferenceEngineDeveloperPackage_DIR "${CMAKE_CURRENT_BINARY_DIR}/build-modules")
diff --git a/inference-engine/samples/CMakeLists.txt b/inference-engine/samples/CMakeLists.txt
index 7924c56779dcdf..fe3952f5779789 100644
--- a/inference-engine/samples/CMakeLists.txt
+++ b/inference-engine/samples/CMakeLists.txt
@@ -110,19 +110,17 @@ if(NOT DEFINED CMAKE_CXX_STANDARD)
 endif()
 ####################################
 
-set (GFLAGS_IS_SUBPROJECT TRUE)
-set (HAVE_SYS_STAT_H 1)
-set (HAVE_INTTYPES_H 1)
-set (INTTYPES_FORMAT C99)
-set (BUILD_TESTING OFF)
-
-if(EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/thirdparty/gflags")
+if(EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/thirdparty/gflags" AND
+    NOT DEFINED IE_MAIN_SOURCE_DIR)
     function(add_gflags)
-        if(NOT WIN32)
-            set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-all")
-            set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wno-all")
-        endif()
+        # common gflags settings
+        set(GFLAGS_IS_SUBPROJECT TRUE)
+        set(HAVE_SYS_STAT_H 1)
+        set(HAVE_INTTYPES_H 1)
+        set(INTTYPES_FORMAT C99)
+        set(BUILD_TESTING OFF)
         set(BUILD_SHARED_LIBS OFF)
+
         add_subdirectory(thirdparty/gflags EXCLUDE_FROM_ALL)
         set_target_properties(gflags_nothreads_static PROPERTIES FOLDER thirdparty)
     endfunction()
@@ -154,7 +152,7 @@ elseif(EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/common/opencv_c_wrapper")
 endif()
 
 # samples build can be switched off during whole IE build
-if (IE_MAIN_SOURCE_DIR AND NOT ENABLE_SAMPLES)
+if (DEFINED IE_MAIN_SOURCE_DIR AND NOT ENABLE_SAMPLES)
     return()
 endif()
 
diff --git a/inference-engine/samples/thirdparty/gflags b/inference-engine/samples/thirdparty/gflags
deleted file mode 160000
index 46f73f88b18aee..00000000000000
--- a/inference-engine/samples/thirdparty/gflags
+++ /dev/null
@@ -1 +0,0 @@
-Subproject commit 46f73f88b18aee341538c0dfc22b1710a6abedef
diff --git a/inference-engine/src/inference_engine/CMakeLists.txt b/inference-engine/src/inference_engine/CMakeLists.txt
index e8ed1a5c4c38f4..4a5577e08328f3 100644
--- a/inference-engine/src/inference_engine/CMakeLists.txt
+++ b/inference-engine/src/inference_engine/CMakeLists.txt
@@ -219,7 +219,7 @@ export(TARGETS ${TARGET_NAME} NAMESPACE IE::
 
 # Export for developer package
 
-ie_developer_export_targets(${TARGET_NAME}_plugin_api)
+openvino_developer_export_targets(COMPONENT inference_engine TARGETS ${TARGET_NAME}_plugin_api)
 
 # install TBB
 
diff --git a/inference-engine/src/legacy_api/CMakeLists.txt b/inference-engine/src/legacy_api/CMakeLists.txt
index 8eae82bd288b02..4cae8b7c6b1d7e 100644
--- a/inference-engine/src/legacy_api/CMakeLists.txt
+++ b/inference-engine/src/legacy_api/CMakeLists.txt
@@ -75,7 +75,7 @@ set_target_properties(${TARGET_NAME} ${TARGET_NAME}_obj
 
 # developer package
 
-ie_developer_export_targets(${TARGET_NAME})
+openvino_developer_export_targets(COMPONENT inference_engine TARGETS ${TARGET_NAME})
 
 # install
 
diff --git a/inference-engine/src/low_precision_transformations/CMakeLists.txt b/inference-engine/src/low_precision_transformations/CMakeLists.txt
index 5d0dfc04fcfe3b..c6306dbc08f067 100644
--- a/inference-engine/src/low_precision_transformations/CMakeLists.txt
+++ b/inference-engine/src/low_precision_transformations/CMakeLists.txt
@@ -45,7 +45,7 @@ set_target_properties(${TARGET_NAME} PROPERTIES INTERPROCEDURAL_OPTIMIZATION_REL
 
 # developer package
 
-ie_developer_export_targets(${TARGET_NAME})
+openvino_developer_export_targets(COMPONENT inference_engine TARGETS ${TARGET_NAME})
 
 # install
 
diff --git a/inference-engine/src/offline_transformations/CMakeLists.txt b/inference-engine/src/offline_transformations/CMakeLists.txt
index cde9fd43f13d49..fb80fe3b6eb584 100644
--- a/inference-engine/src/offline_transformations/CMakeLists.txt
+++ b/inference-engine/src/offline_transformations/CMakeLists.txt
@@ -29,4 +29,4 @@ add_cpplint_target(${TARGET_NAME}_cpplint FOR_TARGETS ${TARGET_NAME})
 
 # developer package
 
-ie_developer_export_targets(${TARGET_NAME})
+openvino_developer_export_targets(COMPONENT inference_engine TARGETS ${TARGET_NAME})
diff --git a/inference-engine/src/preprocessing/CMakeLists.txt b/inference-engine/src/preprocessing/CMakeLists.txt
index f9548339d18b33..fe890eafa497f4 100644
--- a/inference-engine/src/preprocessing/CMakeLists.txt
+++ b/inference-engine/src/preprocessing/CMakeLists.txt
@@ -113,7 +113,7 @@ target_compile_definitions(${TARGET_NAME}_obj PRIVATE IMPLEMENT_PREPROC_PLUGIN
                                                       $<TARGET_PROPERTY:inference_engine,INTERFACE_COMPILE_DEFINITIONS>
                                                       $<TARGET_PROPERTY:fluid,INTERFACE_COMPILE_DEFINITIONS>)
 
-target_include_directories(${TARGET_NAME}_obj SYSTEM PRIVATE "${IE_MAIN_SOURCE_DIR}/thirdparty/ocv"
+target_include_directories(${TARGET_NAME}_obj SYSTEM PRIVATE $<TARGET_PROPERTY:ocv_hal,INTERFACE_INCLUDE_DIRECTORIES>
                                                              $<TARGET_PROPERTY:fluid,INTERFACE_INCLUDE_DIRECTORIES>
                                                              $<TARGET_PROPERTY:inference_engine_plugin_api,INTERFACE_INCLUDE_DIRECTORIES>)
 
@@ -178,7 +178,7 @@ set_target_properties(${TARGET_NAME} ${TARGET_NAME}_obj ${TARGET_NAME}_s
 
 # developer package
 
-ie_developer_export_targets(${TARGET_NAME})
+openvino_developer_export_targets(COMPONENT inference_engine TARGETS ${TARGET_NAME})
 
 # install
 
diff --git a/inference-engine/src/snippets/CMakeLists.txt b/inference-engine/src/snippets/CMakeLists.txt
index e64a3536b1025a..2fc29be7096ffd 100644
--- a/inference-engine/src/snippets/CMakeLists.txt
+++ b/inference-engine/src/snippets/CMakeLists.txt
@@ -48,7 +48,7 @@ set_target_properties(${TARGET_NAME} PROPERTIES INTERPROCEDURAL_OPTIMIZATION_REL
 
 # developer package
 
-ie_developer_export_targets(${TARGET_NAME})
+openvino_developer_export_targets(COMPONENT inference_engine TARGETS ${TARGET_NAME})
 
 # install
 
diff --git a/inference-engine/src/transformations/CMakeLists.txt b/inference-engine/src/transformations/CMakeLists.txt
index d13a4da2c7c12f..0cb9123ffea2b8 100644
--- a/inference-engine/src/transformations/CMakeLists.txt
+++ b/inference-engine/src/transformations/CMakeLists.txt
@@ -45,7 +45,7 @@ set_target_properties(${TARGET_NAME} PROPERTIES INTERPROCEDURAL_OPTIMIZATION_REL
 
 # developer package
 
-ie_developer_export_targets(${TARGET_NAME})
+openvino_developer_export_targets(COMPONENT inference_engine TARGETS ${TARGET_NAME})
 
 # install
 
diff --git a/inference-engine/thirdparty/CMakeLists.txt b/inference-engine/thirdparty/CMakeLists.txt
index f028ebdd8ab5d2..789c3d2c2af49d 100644
--- a/inference-engine/thirdparty/CMakeLists.txt
+++ b/inference-engine/thirdparty/CMakeLists.txt
@@ -62,15 +62,8 @@ else()
     target_include_directories(pugixml INTERFACE "$<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/pugixml/src>")
 endif()
 
-add_subdirectory(ade EXCLUDE_FROM_ALL)
-add_subdirectory(fluid/modules/gapi)
-
-set_target_properties(ade fluid PROPERTIES FOLDER thirdparty)
-
 # developer package
 
-openvino_developer_export_targets(COMPONENT openvino_common TARGETS ade fluid)
-
 if (NOT USE_SYSTEM_PUGIXML)
     set_target_properties(pugixml PROPERTIES FOLDER thirdparty)
     openvino_developer_export_targets(COMPONENT openvino_common TARGETS pugixml)
diff --git a/inference-engine/thirdparty/movidius/XLink/CMakeLists.txt b/inference-engine/thirdparty/movidius/XLink/CMakeLists.txt
index 9b365d8da3961f..6cdd8b865f2499 100644
--- a/inference-engine/thirdparty/movidius/XLink/CMakeLists.txt
+++ b/inference-engine/thirdparty/movidius/XLink/CMakeLists.txt
@@ -42,4 +42,4 @@ endif()
 set_property(TARGET ${TARGET_NAME} PROPERTY C_STANDARD 99)
 
 # TODO: remove once all options are migrated
-ie_developer_export_targets(${TARGET_NAME})
+openvino_developer_export_targets(COMPONENT inference_engine_vpu TARGETS ${TARGET_NAME})
diff --git a/inference-engine/thirdparty/movidius/mvnc/CMakeLists.txt b/inference-engine/thirdparty/movidius/mvnc/CMakeLists.txt
index 31a22c740351a8..3ab1a9299d3816 100644
--- a/inference-engine/thirdparty/movidius/mvnc/CMakeLists.txt
+++ b/inference-engine/thirdparty/movidius/mvnc/CMakeLists.txt
@@ -72,9 +72,8 @@ if(NOT WIN32)
 endif()
 
 # TODO: remove once all options are migrated
-ie_developer_export_targets(${TARGET_NAME})
+openvino_developer_export_targets(COMPONENT inference_engine_vpu TARGETS ${TARGET_NAME})
 
 if(ENABLE_TESTS AND ENABLE_MYRIAD_MVNC_TESTS)
     add_subdirectory(tests)
 endif()
-
diff --git a/tests/stress_tests/CMakeLists.txt b/tests/stress_tests/CMakeLists.txt
index 386971ffbda0dd..cdbb9b88850a54 100644
--- a/tests/stress_tests/CMakeLists.txt
+++ b/tests/stress_tests/CMakeLists.txt
@@ -9,24 +9,19 @@ if (CMAKE_BUILD_TYPE STREQUAL "")
     set(CMAKE_BUILD_TYPE "Release")
 endif()
 
-set (GFLAGS_IS_SUBPROJECT TRUE)
-set (HAVE_SYS_STAT_H 1)
-set (HAVE_INTTYPES_H 1)
-set (INTTYPES_FORMAT C99)
-set (BUILD_TESTING OFF)
-
 find_package(InferenceEngineDeveloperPackage REQUIRED)
 
 set(OpenVINO_MAIN_SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/../../")
 
-if(EXISTS "${OpenVINO_MAIN_SOURCE_DIR}/inference-engine/samples/thirdparty/gflags")
+if(EXISTS "${OpenVINO_MAIN_SOURCE_DIR}/thirdparty/gflags")
     function(add_gflags)
-        if(NOT WIN32)
-            set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-all")
-            set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wno-all")
-        endif()
+        set(GFLAGS_IS_SUBPROJECT TRUE)
+        set(HAVE_SYS_STAT_H 1)
+        set(HAVE_INTTYPES_H 1)
+        set(INTTYPES_FORMAT C99)
+        set(BUILD_TESTING OFF)
         set(BUILD_SHARED_LIBS OFF)
-        add_subdirectory(${OpenVINO_MAIN_SOURCE_DIR}/inference-engine/samples/thirdparty/gflags
+        add_subdirectory(${OpenVINO_MAIN_SOURCE_DIR}/thirdparty/gflags
                          ${CMAKE_CURRENT_BINARY_DIR}/gflags_build
                          EXCLUDE_FROM_ALL)
         set_target_properties(gflags_nothreads_static PROPERTIES FOLDER thirdparty)
diff --git a/tests/time_tests/CMakeLists.txt b/tests/time_tests/CMakeLists.txt
index e4d4beb94473bc..bbc46bf2b2ee20 100644
--- a/tests/time_tests/CMakeLists.txt
+++ b/tests/time_tests/CMakeLists.txt
@@ -21,9 +21,6 @@ if (NOT InferenceEngine_FOUND)
     if (NOT InferenceEngine_FOUND)
         # Search OpenVINO Inference Engine via InferenceEngineDeveloperPackage_DIR
         # in order to provide backward compatibility with old OpenVINO packages
-        set (HAVE_SYS_STAT_H 1)
-        set (HAVE_INTTYPES_H 1)
-        set (INTTYPES_FORMAT C99)
         find_package(InferenceEngineDeveloperPackage REQUIRED)
     endif()
 endif()
diff --git a/tests/time_tests/src/timetests_helper/CMakeLists.txt b/tests/time_tests/src/timetests_helper/CMakeLists.txt
index 75808ac98e583e..e67c5d0ea8d25c 100644
--- a/tests/time_tests/src/timetests_helper/CMakeLists.txt
+++ b/tests/time_tests/src/timetests_helper/CMakeLists.txt
@@ -8,21 +8,15 @@ file (GLOB SRC *.cpp)
 add_library(${TARGET_NAME} STATIC ${SRC})
 target_include_directories(${TARGET_NAME} PUBLIC "${CMAKE_SOURCE_DIR}/include")
 
-set (GFLAGS_IS_SUBPROJECT TRUE)
-set (BUILD_TESTING OFF)
-
-set (HAVE_SYS_STAT_H 1)
-set (HAVE_INTTYPES_H 1)
-set (INTTYPES_FORMAT C99)
-
-if(EXISTS "${OpenVINO_MAIN_SOURCE_DIR}/inference-engine/samples/thirdparty/gflags")
+if(EXISTS "${OpenVINO_MAIN_SOURCE_DIR}/thirdparty/gflags")
     function(add_gflags)
-        if(NOT WIN32)
-            set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-all")
-            set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wno-all")
-        endif()
+        set(GFLAGS_IS_SUBPROJECT TRUE)
+        set(HAVE_SYS_STAT_H 1)
+        set(HAVE_INTTYPES_H 1)
+        set(INTTYPES_FORMAT C99)
+        set(BUILD_TESTING OFF)
         set(BUILD_SHARED_LIBS OFF)
-        add_subdirectory(${OpenVINO_MAIN_SOURCE_DIR}/inference-engine/samples/thirdparty/gflags
+        add_subdirectory(${OpenVINO_MAIN_SOURCE_DIR}/thirdparty/gflags
                          ${CMAKE_CURRENT_BINARY_DIR}/gflags_build
                          EXCLUDE_FROM_ALL)
         set_target_properties(gflags_nothreads_static PROPERTIES FOLDER thirdparty)
diff --git a/thirdparty/CMakeLists.txt b/thirdparty/CMakeLists.txt
index 9b74b65ca1f2de..0c307f9eed37dd 100644
--- a/thirdparty/CMakeLists.txt
+++ b/thirdparty/CMakeLists.txt
@@ -2,12 +2,50 @@
 # SPDX-License-Identifier: Apache-2.0
 #
 
+if(ENABLE_LTO)
+    set(CMAKE_INTERPROCEDURAL_OPTIMIZATION_RELEASE ON)
+endif()
+
 add_subdirectory(ittapi EXCLUDE_FROM_ALL)
 add_subdirectory(itt_collector EXCLUDE_FROM_ALL)
-add_subdirectory(xbyak EXCLUDE_FROM_ALL)
 add_subdirectory(zlib EXCLUDE_FROM_ALL)
 add_subdirectory(cnpy EXCLUDE_FROM_ALL)
 
+add_subdirectory(xbyak EXCLUDE_FROM_ALL)
+openvino_developer_export_targets(COMPONENT openvino_common TARGETS xbyak)
+
+#
+# Fluid, G-API, OpenCV HAL
+#
+
+add_library(ocv_hal INTERFACE)
+target_include_directories(ocv_hal INTERFACE "${CMAKE_CURRENT_SOURCE_DIR}/ocv")
+
+add_subdirectory(ade EXCLUDE_FROM_ALL)
+add_subdirectory(fluid/modules/gapi EXCLUDE_FROM_ALL)
+set_target_properties(ade fluid PROPERTIES FOLDER thirdparty)
+openvino_developer_export_targets(COMPONENT openvino_common TARGETS ade fluid)
+
+#
+# Gflags
+#
+
+function(add_gflags)
+    # common gflags settings
+    set(GFLAGS_IS_SUBPROJECT TRUE)
+    set(HAVE_SYS_STAT_H 1)
+    set(HAVE_INTTYPES_H 1)
+    set(INTTYPES_FORMAT C99)
+    set(BUILD_TESTING OFF)
+    set(BUILD_SHARED_LIBS OFF)
+
+    add_subdirectory(gflags EXCLUDE_FROM_ALL)
+    set_target_properties(gflags_nothreads_static PROPERTIES FOLDER thirdparty)
+    openvino_developer_export_targets(COMPONENT openvino_common TARGETS gflags)
+endfunction()
+
+add_gflags()
+
 #
 # Google Tests framework
 #
@@ -56,7 +94,22 @@ if(NGRAPH_UNIT_TEST_ENABLE OR ENABLE_TESTS)
 endif()
 
 #
-# Export 3rdparty components
+# Install
 #
 
-openvino_developer_export_targets(COMPONENT openvino_common TARGETS xbyak)
+ie_cpack_add_component(cpp_samples_deps)
+
+install(DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/gflags
+        DESTINATION ${IE_CPACK_IE_DIR}/samples/cpp/thirdparty
+        COMPONENT cpp_samples_deps
+        USE_SOURCE_PERMISSIONS)
+
+install(DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/zlib
+        DESTINATION ${IE_CPACK_IE_DIR}/samples/cpp/thirdparty
+        COMPONENT cpp_samples_deps
+        USE_SOURCE_PERMISSIONS)
+
+install(DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/cnpy
+        DESTINATION ${IE_CPACK_IE_DIR}/samples/cpp/thirdparty
+        COMPONENT cpp_samples_deps
+        USE_SOURCE_PERMISSIONS)
diff --git a/inference-engine/thirdparty/ade b/thirdparty/ade
similarity index 100%
rename from inference-engine/thirdparty/ade
rename to thirdparty/ade
diff --git a/inference-engine/thirdparty/fluid/README.md b/thirdparty/fluid/README.md
similarity index 100%
rename from inference-engine/thirdparty/fluid/README.md
rename to thirdparty/fluid/README.md
diff --git a/inference-engine/thirdparty/fluid/check.sh b/thirdparty/fluid/check.sh
similarity index 100%
rename from inference-engine/thirdparty/fluid/check.sh
rename to thirdparty/fluid/check.sh
diff --git a/inference-engine/thirdparty/fluid/checksum.sh b/thirdparty/fluid/checksum.sh
similarity index 100%
rename from inference-engine/thirdparty/fluid/checksum.sh
rename to thirdparty/fluid/checksum.sh
diff --git a/inference-engine/thirdparty/fluid/checksum.txt b/thirdparty/fluid/checksum.txt
similarity index 100%
rename from inference-engine/thirdparty/fluid/checksum.txt
rename to thirdparty/fluid/checksum.txt
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/CMakeLists.txt b/thirdparty/fluid/modules/gapi/CMakeLists.txt
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/CMakeLists.txt
rename to thirdparty/fluid/modules/gapi/CMakeLists.txt
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/cmake/DownloadADE.cmake b/thirdparty/fluid/modules/gapi/cmake/DownloadADE.cmake
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/cmake/DownloadADE.cmake
rename to thirdparty/fluid/modules/gapi/cmake/DownloadADE.cmake
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/cmake/init.cmake b/thirdparty/fluid/modules/gapi/cmake/init.cmake
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/cmake/init.cmake
rename to thirdparty/fluid/modules/gapi/cmake/init.cmake
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/cmake/standalone.cmake b/thirdparty/fluid/modules/gapi/cmake/standalone.cmake
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/cmake/standalone.cmake
rename to thirdparty/fluid/modules/gapi/cmake/standalone.cmake
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/00-root.markdown b/thirdparty/fluid/modules/gapi/doc/00-root.markdown
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/00-root.markdown
rename to thirdparty/fluid/modules/gapi/doc/00-root.markdown
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/01-background.markdown b/thirdparty/fluid/modules/gapi/doc/01-background.markdown
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/01-background.markdown
rename to thirdparty/fluid/modules/gapi/doc/01-background.markdown
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/10-hld-overview.md b/thirdparty/fluid/modules/gapi/doc/10-hld-overview.md
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/10-hld-overview.md
rename to thirdparty/fluid/modules/gapi/doc/10-hld-overview.md
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/20-kernel-api.markdown b/thirdparty/fluid/modules/gapi/doc/20-kernel-api.markdown
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/20-kernel-api.markdown
rename to thirdparty/fluid/modules/gapi/doc/20-kernel-api.markdown
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/30-implementation.markdown b/thirdparty/fluid/modules/gapi/doc/30-implementation.markdown
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/30-implementation.markdown
rename to thirdparty/fluid/modules/gapi/doc/30-implementation.markdown
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/dot/kernel_hierarchy.dot b/thirdparty/fluid/modules/gapi/doc/dot/kernel_hierarchy.dot
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/dot/kernel_hierarchy.dot
rename to thirdparty/fluid/modules/gapi/doc/dot/kernel_hierarchy.dot
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/pics/demo.jpg b/thirdparty/fluid/modules/gapi/doc/pics/demo.jpg
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/pics/demo.jpg
rename to thirdparty/fluid/modules/gapi/doc/pics/demo.jpg
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/pics/gapi_scheme.png b/thirdparty/fluid/modules/gapi/doc/pics/gapi_scheme.png
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/pics/gapi_scheme.png
rename to thirdparty/fluid/modules/gapi/doc/pics/gapi_scheme.png
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/pics/kernel_hierarchy.png b/thirdparty/fluid/modules/gapi/doc/pics/kernel_hierarchy.png
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/pics/kernel_hierarchy.png
rename to thirdparty/fluid/modules/gapi/doc/pics/kernel_hierarchy.png
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/pics/render_example.png b/thirdparty/fluid/modules/gapi/doc/pics/render_example.png
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/pics/render_example.png
rename to thirdparty/fluid/modules/gapi/doc/pics/render_example.png
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/slides/README.md b/thirdparty/fluid/modules/gapi/doc/slides/README.md
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/slides/README.md
rename to thirdparty/fluid/modules/gapi/doc/slides/README.md
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/slides/gapi_overview.org b/thirdparty/fluid/modules/gapi/doc/slides/gapi_overview.org
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/slides/gapi_overview.org
rename to thirdparty/fluid/modules/gapi/doc/slides/gapi_overview.org
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/slides/get_sty.sh b/thirdparty/fluid/modules/gapi/doc/slides/get_sty.sh
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/slides/get_sty.sh
rename to thirdparty/fluid/modules/gapi/doc/slides/get_sty.sh
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/doc/slides/ocv_logo.eps b/thirdparty/fluid/modules/gapi/doc/slides/ocv_logo.eps
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/doc/slides/ocv_logo.eps
rename to thirdparty/fluid/modules/gapi/doc/slides/ocv_logo.eps
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/core.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/core.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/core.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/core.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/core.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/core.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/core.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/core.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/gcpukernel.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/gcpukernel.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/gcpukernel.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/gcpukernel.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/imgproc.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/imgproc.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/imgproc.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/imgproc.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/video.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/video.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/video.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/cpu/video.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/core.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/core.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/core.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/core.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/gfluidbuffer.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/gfluidbuffer.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/gfluidbuffer.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/gfluidbuffer.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/gfluidkernel.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/gfluidkernel.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/gfluidkernel.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/gfluidkernel.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/imgproc.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/imgproc.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/imgproc.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/fluid/imgproc.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/garg.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/garg.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/garg.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/garg.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/garray.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/garray.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/garray.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/garray.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gasync_context.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gasync_context.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gasync_context.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gasync_context.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcall.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcall.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcall.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcall.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcommon.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcommon.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcommon.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcommon.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompiled.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompiled.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompiled.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompiled.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompiled_async.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompiled_async.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompiled_async.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompiled_async.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompoundkernel.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompoundkernel.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompoundkernel.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcompoundkernel.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcomputation.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcomputation.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcomputation.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcomputation.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcomputation_async.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcomputation_async.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcomputation_async.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gcomputation_async.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gframe.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gframe.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gframe.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gframe.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gkernel.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gkernel.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gkernel.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gkernel.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gmat.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gmat.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gmat.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gmat.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gmetaarg.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gmetaarg.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gmetaarg.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gmetaarg.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gopaque.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gopaque.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gopaque.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gopaque.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gproto.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gproto.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gproto.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gproto.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/core.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/core.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/core.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/core.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/ggpukernel.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/ggpukernel.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/ggpukernel.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/ggpukernel.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/imgproc.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/imgproc.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/imgproc.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gpu/imgproc.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gscalar.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gscalar.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gscalar.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gscalar.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gstreaming.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gstreaming.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gstreaming.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gstreaming.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtransform.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtransform.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtransform.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtransform.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtype_traits.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtype_traits.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtype_traits.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtype_traits.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtyped.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtyped.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtyped.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/gtyped.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/imgproc.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/imgproc.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/imgproc.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/imgproc.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/bindings_ie.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/bindings_ie.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/bindings_ie.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/bindings_ie.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/ie.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/ie.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/ie.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/ie.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/ie/util.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/ie/util.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/ie/util.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/ie/util.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/onnx.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/onnx.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/onnx.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/onnx.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/parsers.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/parsers.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/parsers.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/infer/parsers.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/media.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/media.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/media.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/media.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/core.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/core.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/core.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/core.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/goclkernel.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/goclkernel.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/goclkernel.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/goclkernel.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/imgproc.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/imgproc.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/imgproc.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/ocl/imgproc.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/opencv_includes.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/opencv_includes.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/opencv_includes.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/opencv_includes.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/operators.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/operators.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/operators.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/operators.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/assert.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/assert.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/assert.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/assert.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/convert.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/convert.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/convert.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/convert.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/cvdefs.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/cvdefs.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/cvdefs.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/cvdefs.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/exports.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/exports.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/exports.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/exports.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/mat.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/mat.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/mat.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/mat.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/saturate.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/saturate.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/saturate.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/saturate.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/scalar.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/scalar.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/scalar.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/scalar.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/types.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/types.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/types.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/own/types.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/core.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/core.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/core.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/core.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/gplaidmlkernel.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/gplaidmlkernel.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/gplaidmlkernel.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/gplaidmlkernel.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/plaidml.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/plaidml.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/plaidml.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/plaidml/plaidml.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/render.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/render.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/render.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/render.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/render/render.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/render/render.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/render/render.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/render/render.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/render/render_types.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/render/render_types.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/render/render_types.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/render/render_types.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/rmat.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/rmat.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/rmat.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/rmat.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/s11n.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/s11n.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/s11n.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/s11n.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/s11n/base.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/s11n/base.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/s11n/base.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/s11n/base.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/cap.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/cap.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/cap.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/cap.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/desync.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/desync.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/desync.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/desync.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/format.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/format.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/format.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/format.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/meta.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/meta.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/meta.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/meta.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/source.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/source.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/source.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/streaming/source.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/any.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/any.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/any.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/any.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/compiler_hints.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/compiler_hints.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/compiler_hints.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/compiler_hints.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/copy_through_move.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/copy_through_move.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/copy_through_move.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/copy_through_move.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/optional.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/optional.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/optional.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/optional.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/throw.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/throw.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/throw.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/throw.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/type_traits.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/type_traits.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/type_traits.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/type_traits.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/util.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/util.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/util.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/util.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/variant.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/variant.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/variant.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/util/variant.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/video.hpp b/thirdparty/fluid/modules/gapi/include/opencv2/gapi/video.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/include/opencv2/gapi/video.hpp
rename to thirdparty/fluid/modules/gapi/include/opencv2/gapi/video.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/misc/python/pyopencv_gapi.hpp b/thirdparty/fluid/modules/gapi/misc/python/pyopencv_gapi.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/misc/python/pyopencv_gapi.hpp
rename to thirdparty/fluid/modules/gapi/misc/python/pyopencv_gapi.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/misc/python/shadow_gapi.hpp b/thirdparty/fluid/modules/gapi/misc/python/shadow_gapi.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/misc/python/shadow_gapi.hpp
rename to thirdparty/fluid/modules/gapi/misc/python/shadow_gapi.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_core.py b/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_core.py
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_core.py
rename to thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_core.py
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_imgproc.py b/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_imgproc.py
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_imgproc.py
rename to thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_imgproc.py
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_infer.py b/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_infer.py
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_infer.py
rename to thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_infer.py
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_sample_pipelines.py b/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_sample_pipelines.py
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_sample_pipelines.py
rename to thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_sample_pipelines.py
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_streaming.py b/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_streaming.py
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_streaming.py
rename to thirdparty/fluid/modules/gapi/misc/python/test/test_gapi_streaming.py
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests.cpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests.cpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests.hpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests.hpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests_inl.hpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests_inl.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests_inl.hpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_core_perf_tests_inl.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests.cpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests.cpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests.hpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests.hpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests_inl.hpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests_inl.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests_inl.hpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_imgproc_perf_tests_inl.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests.cpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests.cpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests.hpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests.hpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests_inl.hpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests_inl.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests_inl.hpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_render_perf_tests_inl.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests.cpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests.cpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests.hpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests.hpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests_inl.hpp b/thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests_inl.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests_inl.hpp
rename to thirdparty/fluid/modules/gapi/perf/common/gapi_video_perf_tests_inl.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/cpu/gapi_core_perf_tests_cpu.cpp b/thirdparty/fluid/modules/gapi/perf/cpu/gapi_core_perf_tests_cpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/cpu/gapi_core_perf_tests_cpu.cpp
rename to thirdparty/fluid/modules/gapi/perf/cpu/gapi_core_perf_tests_cpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/cpu/gapi_core_perf_tests_fluid.cpp b/thirdparty/fluid/modules/gapi/perf/cpu/gapi_core_perf_tests_fluid.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/cpu/gapi_core_perf_tests_fluid.cpp
rename to thirdparty/fluid/modules/gapi/perf/cpu/gapi_core_perf_tests_fluid.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/cpu/gapi_imgproc_perf_tests_cpu.cpp b/thirdparty/fluid/modules/gapi/perf/cpu/gapi_imgproc_perf_tests_cpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/cpu/gapi_imgproc_perf_tests_cpu.cpp
rename to thirdparty/fluid/modules/gapi/perf/cpu/gapi_imgproc_perf_tests_cpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/cpu/gapi_imgproc_perf_tests_fluid.cpp b/thirdparty/fluid/modules/gapi/perf/cpu/gapi_imgproc_perf_tests_fluid.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/cpu/gapi_imgproc_perf_tests_fluid.cpp
rename to thirdparty/fluid/modules/gapi/perf/cpu/gapi_imgproc_perf_tests_fluid.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/cpu/gapi_video_perf_tests_cpu.cpp b/thirdparty/fluid/modules/gapi/perf/cpu/gapi_video_perf_tests_cpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/cpu/gapi_video_perf_tests_cpu.cpp
rename to thirdparty/fluid/modules/gapi/perf/cpu/gapi_video_perf_tests_cpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/gpu/gapi_core_perf_tests_gpu.cpp b/thirdparty/fluid/modules/gapi/perf/gpu/gapi_core_perf_tests_gpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/gpu/gapi_core_perf_tests_gpu.cpp
rename to thirdparty/fluid/modules/gapi/perf/gpu/gapi_core_perf_tests_gpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/gpu/gapi_imgproc_perf_tests_gpu.cpp b/thirdparty/fluid/modules/gapi/perf/gpu/gapi_imgproc_perf_tests_gpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/gpu/gapi_imgproc_perf_tests_gpu.cpp
rename to thirdparty/fluid/modules/gapi/perf/gpu/gapi_imgproc_perf_tests_gpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/internal/gapi_compiler_perf_tests.cpp b/thirdparty/fluid/modules/gapi/perf/internal/gapi_compiler_perf_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/internal/gapi_compiler_perf_tests.cpp
rename to thirdparty/fluid/modules/gapi/perf/internal/gapi_compiler_perf_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/perf_bench.cpp b/thirdparty/fluid/modules/gapi/perf/perf_bench.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/perf_bench.cpp
rename to thirdparty/fluid/modules/gapi/perf/perf_bench.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/perf_main.cpp b/thirdparty/fluid/modules/gapi/perf/perf_main.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/perf_main.cpp
rename to thirdparty/fluid/modules/gapi/perf/perf_main.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/perf_precomp.hpp b/thirdparty/fluid/modules/gapi/perf/perf_precomp.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/perf_precomp.hpp
rename to thirdparty/fluid/modules/gapi/perf/perf_precomp.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/perf/render/gapi_render_perf_tests_ocv.cpp b/thirdparty/fluid/modules/gapi/perf/render/gapi_render_perf_tests_ocv.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/perf/render/gapi_render_perf_tests_ocv.cpp
rename to thirdparty/fluid/modules/gapi/perf/render/gapi_render_perf_tests_ocv.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/api_example.cpp b/thirdparty/fluid/modules/gapi/samples/api_example.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/api_example.cpp
rename to thirdparty/fluid/modules/gapi/samples/api_example.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/api_ref_snippets.cpp b/thirdparty/fluid/modules/gapi/samples/api_ref_snippets.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/api_ref_snippets.cpp
rename to thirdparty/fluid/modules/gapi/samples/api_ref_snippets.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/draw_example.cpp b/thirdparty/fluid/modules/gapi/samples/draw_example.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/draw_example.cpp
rename to thirdparty/fluid/modules/gapi/samples/draw_example.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/dynamic_graph.cpp b/thirdparty/fluid/modules/gapi/samples/dynamic_graph.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/dynamic_graph.cpp
rename to thirdparty/fluid/modules/gapi/samples/dynamic_graph.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/gaze_estimation.cpp b/thirdparty/fluid/modules/gapi/samples/gaze_estimation.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/gaze_estimation.cpp
rename to thirdparty/fluid/modules/gapi/samples/gaze_estimation.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/infer_ie_onnx_hybrid.cpp b/thirdparty/fluid/modules/gapi/samples/infer_ie_onnx_hybrid.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/infer_ie_onnx_hybrid.cpp
rename to thirdparty/fluid/modules/gapi/samples/infer_ie_onnx_hybrid.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/infer_single_roi.cpp b/thirdparty/fluid/modules/gapi/samples/infer_single_roi.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/infer_single_roi.cpp
rename to thirdparty/fluid/modules/gapi/samples/infer_single_roi.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/infer_ssd_onnx.cpp b/thirdparty/fluid/modules/gapi/samples/infer_ssd_onnx.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/infer_ssd_onnx.cpp
rename to thirdparty/fluid/modules/gapi/samples/infer_ssd_onnx.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/kernel_api_snippets.cpp b/thirdparty/fluid/modules/gapi/samples/kernel_api_snippets.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/kernel_api_snippets.cpp
rename to thirdparty/fluid/modules/gapi/samples/kernel_api_snippets.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/privacy_masking_camera.cpp b/thirdparty/fluid/modules/gapi/samples/privacy_masking_camera.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/privacy_masking_camera.cpp
rename to thirdparty/fluid/modules/gapi/samples/privacy_masking_camera.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/slides_blur_gapi.cpp b/thirdparty/fluid/modules/gapi/samples/slides_blur_gapi.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/slides_blur_gapi.cpp
rename to thirdparty/fluid/modules/gapi/samples/slides_blur_gapi.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/slides_sobel_cv.cpp b/thirdparty/fluid/modules/gapi/samples/slides_sobel_cv.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/slides_sobel_cv.cpp
rename to thirdparty/fluid/modules/gapi/samples/slides_sobel_cv.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/slides_sobel_gapi.cpp b/thirdparty/fluid/modules/gapi/samples/slides_sobel_gapi.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/slides_sobel_gapi.cpp
rename to thirdparty/fluid/modules/gapi/samples/slides_sobel_gapi.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/samples/text_detection.cpp b/thirdparty/fluid/modules/gapi/samples/text_detection.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/samples/text_detection.cpp
rename to thirdparty/fluid/modules/gapi/samples/text_detection.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/scripts/measure_privacy_masking.py b/thirdparty/fluid/modules/gapi/scripts/measure_privacy_masking.py
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/scripts/measure_privacy_masking.py
rename to thirdparty/fluid/modules/gapi/scripts/measure_privacy_masking.py
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/README.md b/thirdparty/fluid/modules/gapi/src/api/README.md
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/README.md
rename to thirdparty/fluid/modules/gapi/src/api/README.md
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/ft_render.cpp b/thirdparty/fluid/modules/gapi/src/api/ft_render.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/ft_render.cpp
rename to thirdparty/fluid/modules/gapi/src/api/ft_render.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/ft_render.hpp b/thirdparty/fluid/modules/gapi/src/api/ft_render.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/ft_render.hpp
rename to thirdparty/fluid/modules/gapi/src/api/ft_render.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/ft_render_priv.hpp b/thirdparty/fluid/modules/gapi/src/api/ft_render_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/ft_render_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/api/ft_render_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/garray.cpp b/thirdparty/fluid/modules/gapi/src/api/garray.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/garray.cpp
rename to thirdparty/fluid/modules/gapi/src/api/garray.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gbackend.cpp b/thirdparty/fluid/modules/gapi/src/api/gbackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gbackend.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gbackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gbackend_priv.hpp b/thirdparty/fluid/modules/gapi/src/api/gbackend_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gbackend_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/api/gbackend_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gcall.cpp b/thirdparty/fluid/modules/gapi/src/api/gcall.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gcall.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gcall.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gcall_priv.hpp b/thirdparty/fluid/modules/gapi/src/api/gcall_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gcall_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/api/gcall_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gcomputation.cpp b/thirdparty/fluid/modules/gapi/src/api/gcomputation.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gcomputation.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gcomputation.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gcomputation_priv.hpp b/thirdparty/fluid/modules/gapi/src/api/gcomputation_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gcomputation_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/api/gcomputation_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gframe.cpp b/thirdparty/fluid/modules/gapi/src/api/gframe.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gframe.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gframe.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/ginfer.cpp b/thirdparty/fluid/modules/gapi/src/api/ginfer.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/ginfer.cpp
rename to thirdparty/fluid/modules/gapi/src/api/ginfer.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gkernel.cpp b/thirdparty/fluid/modules/gapi/src/api/gkernel.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gkernel.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gkernel.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gmat.cpp b/thirdparty/fluid/modules/gapi/src/api/gmat.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gmat.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gmat.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gnode.cpp b/thirdparty/fluid/modules/gapi/src/api/gnode.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gnode.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gnode.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gnode.hpp b/thirdparty/fluid/modules/gapi/src/api/gnode.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gnode.hpp
rename to thirdparty/fluid/modules/gapi/src/api/gnode.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gnode_priv.hpp b/thirdparty/fluid/modules/gapi/src/api/gnode_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gnode_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/api/gnode_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gopaque.cpp b/thirdparty/fluid/modules/gapi/src/api/gopaque.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gopaque.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gopaque.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gorigin.cpp b/thirdparty/fluid/modules/gapi/src/api/gorigin.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gorigin.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gorigin.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gorigin.hpp b/thirdparty/fluid/modules/gapi/src/api/gorigin.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gorigin.hpp
rename to thirdparty/fluid/modules/gapi/src/api/gorigin.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gproto.cpp b/thirdparty/fluid/modules/gapi/src/api/gproto.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gproto.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gproto.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gproto_priv.hpp b/thirdparty/fluid/modules/gapi/src/api/gproto_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gproto_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/api/gproto_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/grunarg.cpp b/thirdparty/fluid/modules/gapi/src/api/grunarg.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/grunarg.cpp
rename to thirdparty/fluid/modules/gapi/src/api/grunarg.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/gscalar.cpp b/thirdparty/fluid/modules/gapi/src/api/gscalar.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/gscalar.cpp
rename to thirdparty/fluid/modules/gapi/src/api/gscalar.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/kernels_core.cpp b/thirdparty/fluid/modules/gapi/src/api/kernels_core.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/kernels_core.cpp
rename to thirdparty/fluid/modules/gapi/src/api/kernels_core.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/kernels_imgproc.cpp b/thirdparty/fluid/modules/gapi/src/api/kernels_imgproc.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/kernels_imgproc.cpp
rename to thirdparty/fluid/modules/gapi/src/api/kernels_imgproc.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/kernels_nnparsers.cpp b/thirdparty/fluid/modules/gapi/src/api/kernels_nnparsers.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/kernels_nnparsers.cpp
rename to thirdparty/fluid/modules/gapi/src/api/kernels_nnparsers.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/kernels_streaming.cpp b/thirdparty/fluid/modules/gapi/src/api/kernels_streaming.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/kernels_streaming.cpp
rename to thirdparty/fluid/modules/gapi/src/api/kernels_streaming.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/kernels_video.cpp b/thirdparty/fluid/modules/gapi/src/api/kernels_video.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/kernels_video.cpp
rename to thirdparty/fluid/modules/gapi/src/api/kernels_video.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/media.cpp b/thirdparty/fluid/modules/gapi/src/api/media.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/media.cpp
rename to thirdparty/fluid/modules/gapi/src/api/media.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/ocv_mask_creator.hpp b/thirdparty/fluid/modules/gapi/src/api/ocv_mask_creator.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/ocv_mask_creator.hpp
rename to thirdparty/fluid/modules/gapi/src/api/ocv_mask_creator.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/operators.cpp b/thirdparty/fluid/modules/gapi/src/api/operators.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/operators.cpp
rename to thirdparty/fluid/modules/gapi/src/api/operators.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/render.cpp b/thirdparty/fluid/modules/gapi/src/api/render.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/render.cpp
rename to thirdparty/fluid/modules/gapi/src/api/render.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/render_ocv.cpp b/thirdparty/fluid/modules/gapi/src/api/render_ocv.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/render_ocv.cpp
rename to thirdparty/fluid/modules/gapi/src/api/render_ocv.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/render_ocv.hpp b/thirdparty/fluid/modules/gapi/src/api/render_ocv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/render_ocv.hpp
rename to thirdparty/fluid/modules/gapi/src/api/render_ocv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/render_priv.hpp b/thirdparty/fluid/modules/gapi/src/api/render_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/render_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/api/render_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/rmat.cpp b/thirdparty/fluid/modules/gapi/src/api/rmat.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/rmat.cpp
rename to thirdparty/fluid/modules/gapi/src/api/rmat.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/api/s11n.cpp b/thirdparty/fluid/modules/gapi/src/api/s11n.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/api/s11n.cpp
rename to thirdparty/fluid/modules/gapi/src/api/s11n.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/README.md b/thirdparty/fluid/modules/gapi/src/backends/README.md
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/README.md
rename to thirdparty/fluid/modules/gapi/src/backends/README.md
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/gbackend.hpp b/thirdparty/fluid/modules/gapi/src/backends/common/gbackend.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/gbackend.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/common/gbackend.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/gcompoundbackend.cpp b/thirdparty/fluid/modules/gapi/src/backends/common/gcompoundbackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/gcompoundbackend.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/common/gcompoundbackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/gcompoundkernel.cpp b/thirdparty/fluid/modules/gapi/src/backends/common/gcompoundkernel.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/gcompoundkernel.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/common/gcompoundkernel.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/gmetabackend.cpp b/thirdparty/fluid/modules/gapi/src/backends/common/gmetabackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/gmetabackend.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/common/gmetabackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/gmetabackend.hpp b/thirdparty/fluid/modules/gapi/src/backends/common/gmetabackend.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/gmetabackend.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/common/gmetabackend.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/serialization.cpp b/thirdparty/fluid/modules/gapi/src/backends/common/serialization.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/serialization.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/common/serialization.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/serialization.hpp b/thirdparty/fluid/modules/gapi/src/backends/common/serialization.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/common/serialization.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/common/serialization.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpubackend.cpp b/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpubackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpubackend.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/cpu/gcpubackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpubackend.hpp b/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpubackend.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpubackend.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/cpu/gcpubackend.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpucore.cpp b/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpucore.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpucore.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/cpu/gcpucore.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpuimgproc.cpp b/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpuimgproc.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpuimgproc.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/cpu/gcpuimgproc.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpukernel.cpp b/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpukernel.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpukernel.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/cpu/gcpukernel.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpuvideo.cpp b/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpuvideo.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gcpuvideo.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/cpu/gcpuvideo.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gnnparsers.cpp b/thirdparty/fluid/modules/gapi/src/backends/cpu/gnnparsers.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gnnparsers.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/cpu/gnnparsers.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gnnparsers.hpp b/thirdparty/fluid/modules/gapi/src/backends/cpu/gnnparsers.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/cpu/gnnparsers.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/cpu/gnnparsers.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbackend.cpp b/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbackend.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbackend.hpp b/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbackend.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbackend.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbackend.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbuffer.cpp b/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbuffer.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbuffer.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbuffer.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbuffer_priv.hpp b/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbuffer_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbuffer_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidbuffer_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidcore.cpp b/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidcore.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidcore.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidcore.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc.cpp b/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.dispatch.cpp b/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.dispatch.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.dispatch.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.dispatch.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.hpp b/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.simd.hpp b/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.simd.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.simd.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidimgproc_func.simd.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidutils.hpp b/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidutils.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidutils.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/fluid/gfluidutils.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/bindings_ie.cpp b/thirdparty/fluid/modules/gapi/src/backends/ie/bindings_ie.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/bindings_ie.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/ie/bindings_ie.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend.cpp b/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/ie/giebackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend.hpp b/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/ie/giebackend.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend/giewrapper.cpp b/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend/giewrapper.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend/giewrapper.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/ie/giebackend/giewrapper.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend/giewrapper.hpp b/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend/giewrapper.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/giebackend/giewrapper.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/ie/giebackend/giewrapper.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/util.hpp b/thirdparty/fluid/modules/gapi/src/backends/ie/util.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ie/util.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/ie/util.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclbackend.cpp b/thirdparty/fluid/modules/gapi/src/backends/ocl/goclbackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclbackend.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/ocl/goclbackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclbackend.hpp b/thirdparty/fluid/modules/gapi/src/backends/ocl/goclbackend.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclbackend.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/ocl/goclbackend.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclcore.cpp b/thirdparty/fluid/modules/gapi/src/backends/ocl/goclcore.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclcore.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/ocl/goclcore.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclcore.hpp b/thirdparty/fluid/modules/gapi/src/backends/ocl/goclcore.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclcore.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/ocl/goclcore.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclimgproc.cpp b/thirdparty/fluid/modules/gapi/src/backends/ocl/goclimgproc.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclimgproc.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/ocl/goclimgproc.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclimgproc.hpp b/thirdparty/fluid/modules/gapi/src/backends/ocl/goclimgproc.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclimgproc.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/ocl/goclimgproc.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclkernel.cpp b/thirdparty/fluid/modules/gapi/src/backends/ocl/goclkernel.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/ocl/goclkernel.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/ocl/goclkernel.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/onnx/gonnxbackend.cpp b/thirdparty/fluid/modules/gapi/src/backends/onnx/gonnxbackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/onnx/gonnxbackend.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/onnx/gonnxbackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/onnx/gonnxbackend.hpp b/thirdparty/fluid/modules/gapi/src/backends/onnx/gonnxbackend.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/onnx/gonnxbackend.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/onnx/gonnxbackend.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlbackend.cpp b/thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlbackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlbackend.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlbackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlbackend.hpp b/thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlbackend.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlbackend.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlbackend.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlcore.cpp b/thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlcore.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlcore.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/plaidml/gplaidmlcore.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/plaidml/plaidml_util.hpp b/thirdparty/fluid/modules/gapi/src/backends/plaidml/plaidml_util.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/plaidml/plaidml_util.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/plaidml/plaidml_util.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/ft_render.cpp b/thirdparty/fluid/modules/gapi/src/backends/render/ft_render.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/ft_render.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/render/ft_render.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/ft_render.hpp b/thirdparty/fluid/modules/gapi/src/backends/render/ft_render.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/ft_render.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/render/ft_render.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/ft_render_priv.hpp b/thirdparty/fluid/modules/gapi/src/backends/render/ft_render_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/ft_render_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/render/ft_render_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/grenderocv.cpp b/thirdparty/fluid/modules/gapi/src/backends/render/grenderocv.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/grenderocv.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/render/grenderocv.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/grenderocv.hpp b/thirdparty/fluid/modules/gapi/src/backends/render/grenderocv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/grenderocv.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/render/grenderocv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/grenderocvbackend.cpp b/thirdparty/fluid/modules/gapi/src/backends/render/grenderocvbackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/grenderocvbackend.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/render/grenderocvbackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/grenderocvbackend.hpp b/thirdparty/fluid/modules/gapi/src/backends/render/grenderocvbackend.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/render/grenderocvbackend.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/render/grenderocvbackend.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingbackend.cpp b/thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingbackend.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingbackend.cpp
rename to thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingbackend.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingbackend.hpp b/thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingbackend.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingbackend.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingbackend.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingkernel.hpp b/thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingkernel.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingkernel.hpp
rename to thirdparty/fluid/modules/gapi/src/backends/streaming/gstreamingkernel.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/README.md b/thirdparty/fluid/modules/gapi/src/compiler/README.md
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/README.md
rename to thirdparty/fluid/modules/gapi/src/compiler/README.md
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gcompiled.cpp b/thirdparty/fluid/modules/gapi/src/compiler/gcompiled.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gcompiled.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gcompiled.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gcompiled_priv.hpp b/thirdparty/fluid/modules/gapi/src/compiler/gcompiled_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gcompiled_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gcompiled_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gcompiler.cpp b/thirdparty/fluid/modules/gapi/src/compiler/gcompiler.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gcompiler.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gcompiler.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gcompiler.hpp b/thirdparty/fluid/modules/gapi/src/compiler/gcompiler.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gcompiler.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gcompiler.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gislandmodel.cpp b/thirdparty/fluid/modules/gapi/src/compiler/gislandmodel.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gislandmodel.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gislandmodel.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gislandmodel.hpp b/thirdparty/fluid/modules/gapi/src/compiler/gislandmodel.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gislandmodel.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gislandmodel.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gmodel.cpp b/thirdparty/fluid/modules/gapi/src/compiler/gmodel.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gmodel.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gmodel.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gmodel.hpp b/thirdparty/fluid/modules/gapi/src/compiler/gmodel.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gmodel.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gmodel.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gmodel_priv.hpp b/thirdparty/fluid/modules/gapi/src/compiler/gmodel_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gmodel_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gmodel_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gmodelbuilder.cpp b/thirdparty/fluid/modules/gapi/src/compiler/gmodelbuilder.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gmodelbuilder.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gmodelbuilder.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gmodelbuilder.hpp b/thirdparty/fluid/modules/gapi/src/compiler/gmodelbuilder.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gmodelbuilder.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gmodelbuilder.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gobjref.hpp b/thirdparty/fluid/modules/gapi/src/compiler/gobjref.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gobjref.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gobjref.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gstreaming.cpp b/thirdparty/fluid/modules/gapi/src/compiler/gstreaming.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gstreaming.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gstreaming.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gstreaming_priv.hpp b/thirdparty/fluid/modules/gapi/src/compiler/gstreaming_priv.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/gstreaming_priv.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/gstreaming_priv.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/dump_dot.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/dump_dot.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/dump_dot.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/dump_dot.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/exec.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/exec.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/exec.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/exec.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/helpers.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/helpers.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/helpers.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/helpers.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/helpers.hpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/helpers.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/helpers.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/helpers.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/intrin.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/intrin.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/intrin.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/intrin.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/islands.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/islands.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/islands.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/islands.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/kernels.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/kernels.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/kernels.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/kernels.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/meta.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/meta.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/meta.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/meta.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/passes.hpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/passes.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/passes.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/passes.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/pattern_matching.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/pattern_matching.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/pattern_matching.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/pattern_matching.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/pattern_matching.hpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/pattern_matching.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/pattern_matching.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/pattern_matching.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/perform_substitution.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/perform_substitution.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/perform_substitution.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/perform_substitution.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/streaming.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/streaming.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/streaming.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/streaming.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/transformations.cpp b/thirdparty/fluid/modules/gapi/src/compiler/passes/transformations.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/passes/transformations.cpp
rename to thirdparty/fluid/modules/gapi/src/compiler/passes/transformations.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/compiler/transactions.hpp b/thirdparty/fluid/modules/gapi/src/compiler/transactions.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/compiler/transactions.hpp
rename to thirdparty/fluid/modules/gapi/src/compiler/transactions.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/executor/conc_queue.hpp b/thirdparty/fluid/modules/gapi/src/executor/conc_queue.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/executor/conc_queue.hpp
rename to thirdparty/fluid/modules/gapi/src/executor/conc_queue.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/executor/gapi_itt.hpp b/thirdparty/fluid/modules/gapi/src/executor/gapi_itt.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/executor/gapi_itt.hpp
rename to thirdparty/fluid/modules/gapi/src/executor/gapi_itt.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/executor/gasync.cpp b/thirdparty/fluid/modules/gapi/src/executor/gasync.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/executor/gasync.cpp
rename to thirdparty/fluid/modules/gapi/src/executor/gasync.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/executor/gexecutor.cpp b/thirdparty/fluid/modules/gapi/src/executor/gexecutor.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/executor/gexecutor.cpp
rename to thirdparty/fluid/modules/gapi/src/executor/gexecutor.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/executor/gexecutor.hpp b/thirdparty/fluid/modules/gapi/src/executor/gexecutor.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/executor/gexecutor.hpp
rename to thirdparty/fluid/modules/gapi/src/executor/gexecutor.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/executor/gstreamingexecutor.cpp b/thirdparty/fluid/modules/gapi/src/executor/gstreamingexecutor.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/executor/gstreamingexecutor.cpp
rename to thirdparty/fluid/modules/gapi/src/executor/gstreamingexecutor.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/executor/gstreamingexecutor.hpp b/thirdparty/fluid/modules/gapi/src/executor/gstreamingexecutor.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/executor/gstreamingexecutor.hpp
rename to thirdparty/fluid/modules/gapi/src/executor/gstreamingexecutor.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/executor/gtbbexecutor.cpp b/thirdparty/fluid/modules/gapi/src/executor/gtbbexecutor.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/executor/gtbbexecutor.cpp
rename to thirdparty/fluid/modules/gapi/src/executor/gtbbexecutor.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/executor/gtbbexecutor.hpp b/thirdparty/fluid/modules/gapi/src/executor/gtbbexecutor.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/executor/gtbbexecutor.hpp
rename to thirdparty/fluid/modules/gapi/src/executor/gtbbexecutor.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/executor/last_value.hpp b/thirdparty/fluid/modules/gapi/src/executor/last_value.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/executor/last_value.hpp
rename to thirdparty/fluid/modules/gapi/src/executor/last_value.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/logger.hpp b/thirdparty/fluid/modules/gapi/src/logger.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/logger.hpp
rename to thirdparty/fluid/modules/gapi/src/logger.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/src/precomp.hpp b/thirdparty/fluid/modules/gapi/src/precomp.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/src/precomp.hpp
rename to thirdparty/fluid/modules/gapi/src/precomp.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_compoundkernel_tests.cpp b/thirdparty/fluid/modules/gapi/test/common/gapi_compoundkernel_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_compoundkernel_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_compoundkernel_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_core_tests.cpp b/thirdparty/fluid/modules/gapi/test/common/gapi_core_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_core_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_core_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_core_tests.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_core_tests.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_core_tests.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_core_tests.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_core_tests_inl.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_core_tests_inl.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_core_tests_inl.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_core_tests_inl.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests.cpp b/thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests_inl.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests_inl.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests_inl.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_imgproc_tests_inl.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests.cpp b/thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests_inl.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests_inl.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests_inl.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_operators_tests_inl.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_parsers_tests_common.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_parsers_tests_common.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_parsers_tests_common.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_parsers_tests_common.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_render_tests.cpp b/thirdparty/fluid/modules/gapi/test/common/gapi_render_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_render_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_render_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_render_tests.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_render_tests.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_render_tests.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_render_tests.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_render_tests_inl.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_render_tests_inl.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_render_tests_inl.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_render_tests_inl.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_tests_common.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_tests_common.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_tests_common.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_tests_common.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_tests_helpers.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_tests_helpers.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_tests_helpers.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_tests_helpers.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests.cpp b/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_video_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_video_tests.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests_common.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests_common.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests_common.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_video_tests_common.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests_inl.hpp b/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests_inl.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/common/gapi_video_tests_inl.hpp
rename to thirdparty/fluid/modules/gapi/test/common/gapi_video_tests_inl.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_core_tests_cpu.cpp b/thirdparty/fluid/modules/gapi/test/cpu/gapi_core_tests_cpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_core_tests_cpu.cpp
rename to thirdparty/fluid/modules/gapi/test/cpu/gapi_core_tests_cpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_core_tests_fluid.cpp b/thirdparty/fluid/modules/gapi/test/cpu/gapi_core_tests_fluid.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_core_tests_fluid.cpp
rename to thirdparty/fluid/modules/gapi/test/cpu/gapi_core_tests_fluid.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_imgproc_tests_cpu.cpp b/thirdparty/fluid/modules/gapi/test/cpu/gapi_imgproc_tests_cpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_imgproc_tests_cpu.cpp
rename to thirdparty/fluid/modules/gapi/test/cpu/gapi_imgproc_tests_cpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_imgproc_tests_fluid.cpp b/thirdparty/fluid/modules/gapi/test/cpu/gapi_imgproc_tests_fluid.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_imgproc_tests_fluid.cpp
rename to thirdparty/fluid/modules/gapi/test/cpu/gapi_imgproc_tests_fluid.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_ocv_stateful_kernel_test_utils.hpp b/thirdparty/fluid/modules/gapi/test/cpu/gapi_ocv_stateful_kernel_test_utils.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_ocv_stateful_kernel_test_utils.hpp
rename to thirdparty/fluid/modules/gapi/test/cpu/gapi_ocv_stateful_kernel_test_utils.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_ocv_stateful_kernel_tests.cpp b/thirdparty/fluid/modules/gapi/test/cpu/gapi_ocv_stateful_kernel_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_ocv_stateful_kernel_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/cpu/gapi_ocv_stateful_kernel_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_operators_tests_cpu.cpp b/thirdparty/fluid/modules/gapi/test/cpu/gapi_operators_tests_cpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_operators_tests_cpu.cpp
rename to thirdparty/fluid/modules/gapi/test/cpu/gapi_operators_tests_cpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_operators_tests_fluid.cpp b/thirdparty/fluid/modules/gapi/test/cpu/gapi_operators_tests_fluid.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_operators_tests_fluid.cpp
rename to thirdparty/fluid/modules/gapi/test/cpu/gapi_operators_tests_fluid.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_render_tests_cpu.cpp b/thirdparty/fluid/modules/gapi/test/cpu/gapi_render_tests_cpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_render_tests_cpu.cpp
rename to thirdparty/fluid/modules/gapi/test/cpu/gapi_render_tests_cpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_video_tests_cpu.cpp b/thirdparty/fluid/modules/gapi/test/cpu/gapi_video_tests_cpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/cpu/gapi_video_tests_cpu.cpp
rename to thirdparty/fluid/modules/gapi/test/cpu/gapi_video_tests_cpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/executor/gtbbexecutor_internal_tests.cpp b/thirdparty/fluid/modules/gapi/test/executor/gtbbexecutor_internal_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/executor/gtbbexecutor_internal_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/executor/gtbbexecutor_internal_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_array_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_array_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_array_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_array_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_async_test.cpp b/thirdparty/fluid/modules/gapi/test/gapi_async_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_async_test.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_async_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_basic_hetero_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_basic_hetero_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_basic_hetero_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_basic_hetero_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_compile_args_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_compile_args_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_compile_args_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_compile_args_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_desc_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_desc_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_desc_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_desc_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_parallel_rois_test.cpp b/thirdparty/fluid/modules/gapi/test/gapi_fluid_parallel_rois_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_parallel_rois_test.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_fluid_parallel_rois_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_resize_test.cpp b/thirdparty/fluid/modules/gapi/test/gapi_fluid_resize_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_resize_test.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_fluid_resize_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_roi_test.cpp b/thirdparty/fluid/modules/gapi/test/gapi_fluid_roi_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_roi_test.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_fluid_roi_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_test.cpp b/thirdparty/fluid/modules/gapi/test/gapi_fluid_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_test.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_fluid_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_test_kernels.cpp b/thirdparty/fluid/modules/gapi/test/gapi_fluid_test_kernels.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_test_kernels.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_fluid_test_kernels.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_test_kernels.hpp b/thirdparty/fluid/modules/gapi/test/gapi_fluid_test_kernels.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_fluid_test_kernels.hpp
rename to thirdparty/fluid/modules/gapi/test/gapi_fluid_test_kernels.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_frame_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_frame_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_frame_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_frame_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_gcompiled_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_gcompiled_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_gcompiled_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_gcompiled_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_gcomputation_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_gcomputation_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_gcomputation_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_gcomputation_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_gpu_test.cpp b/thirdparty/fluid/modules/gapi/test/gapi_gpu_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_gpu_test.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_gpu_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_graph_meta_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_graph_meta_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_graph_meta_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_graph_meta_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_kernel_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_kernel_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_kernel_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_kernel_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_mock_kernels.hpp b/thirdparty/fluid/modules/gapi/test/gapi_mock_kernels.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_mock_kernels.hpp
rename to thirdparty/fluid/modules/gapi/test/gapi_mock_kernels.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_opaque_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_opaque_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_opaque_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_opaque_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_plaidml_pipelines.cpp b/thirdparty/fluid/modules/gapi/test/gapi_plaidml_pipelines.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_plaidml_pipelines.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_plaidml_pipelines.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_planar_test.cpp b/thirdparty/fluid/modules/gapi/test/gapi_planar_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_planar_test.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_planar_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_sample_pipelines.cpp b/thirdparty/fluid/modules/gapi/test/gapi_sample_pipelines.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_sample_pipelines.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_sample_pipelines.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_scalar_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_scalar_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_scalar_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_scalar_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_smoke_test.cpp b/thirdparty/fluid/modules/gapi/test/gapi_smoke_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_smoke_test.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_smoke_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_transform_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_transform_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_transform_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_transform_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_typed_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_typed_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_typed_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_typed_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gapi_util_tests.cpp b/thirdparty/fluid/modules/gapi/test/gapi_util_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gapi_util_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/gapi_util_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gpu/gapi_core_tests_gpu.cpp b/thirdparty/fluid/modules/gapi/test/gpu/gapi_core_tests_gpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gpu/gapi_core_tests_gpu.cpp
rename to thirdparty/fluid/modules/gapi/test/gpu/gapi_core_tests_gpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gpu/gapi_imgproc_tests_gpu.cpp b/thirdparty/fluid/modules/gapi/test/gpu/gapi_imgproc_tests_gpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gpu/gapi_imgproc_tests_gpu.cpp
rename to thirdparty/fluid/modules/gapi/test/gpu/gapi_imgproc_tests_gpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/gpu/gapi_operators_tests_gpu.cpp b/thirdparty/fluid/modules/gapi/test/gpu/gapi_operators_tests_gpu.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/gpu/gapi_operators_tests_gpu.cpp
rename to thirdparty/fluid/modules/gapi/test/gpu/gapi_operators_tests_gpu.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/infer/gapi_infer_ie_test.cpp b/thirdparty/fluid/modules/gapi/test/infer/gapi_infer_ie_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/infer/gapi_infer_ie_test.cpp
rename to thirdparty/fluid/modules/gapi/test/infer/gapi_infer_ie_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/infer/gapi_infer_onnx_test.cpp b/thirdparty/fluid/modules/gapi/test/infer/gapi_infer_onnx_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/infer/gapi_infer_onnx_test.cpp
rename to thirdparty/fluid/modules/gapi/test/infer/gapi_infer_onnx_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/infer/gapi_infer_tests.cpp b/thirdparty/fluid/modules/gapi/test/infer/gapi_infer_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/infer/gapi_infer_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/infer/gapi_infer_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_backend_tests.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_backend_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_backend_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_backend_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_dynamic_graph.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_dynamic_graph.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_dynamic_graph.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_dynamic_graph.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_executor_tests.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_executor_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_executor_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_executor_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_garg_test.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_garg_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_garg_test.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_garg_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_gmetaarg_test.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_gmetaarg_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_gmetaarg_test.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_gmetaarg_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_gmodel_builder_test.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_gmodel_builder_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_gmodel_builder_test.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_gmodel_builder_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_island_fusion_tests.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_island_fusion_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_island_fusion_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_island_fusion_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_island_tests.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_island_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_island_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_island_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_pattern_matching_test.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_pattern_matching_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_pattern_matching_test.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_pattern_matching_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_perform_substitution_test.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_perform_substitution_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_perform_substitution_test.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_perform_substitution_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_proto_tests.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_proto_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_proto_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_proto_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_recompilation_test.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_recompilation_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_recompilation_test.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_recompilation_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_vectorref_test.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_int_vectorref_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_int_vectorref_test.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_int_vectorref_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_transactions_test.cpp b/thirdparty/fluid/modules/gapi/test/internal/gapi_transactions_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/internal/gapi_transactions_test.cpp
rename to thirdparty/fluid/modules/gapi/test/internal/gapi_transactions_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/opencl_kernels_test_gapi.hpp b/thirdparty/fluid/modules/gapi/test/opencl_kernels_test_gapi.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/opencl_kernels_test_gapi.hpp
rename to thirdparty/fluid/modules/gapi/test/opencl_kernels_test_gapi.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/own/conc_queue_tests.cpp b/thirdparty/fluid/modules/gapi/test/own/conc_queue_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/own/conc_queue_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/own/conc_queue_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/own/gapi_types_tests.cpp b/thirdparty/fluid/modules/gapi/test/own/gapi_types_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/own/gapi_types_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/own/gapi_types_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/own/last_written_value_tests.cpp b/thirdparty/fluid/modules/gapi/test/own/last_written_value_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/own/last_written_value_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/own/last_written_value_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/own/mat_tests.cpp b/thirdparty/fluid/modules/gapi/test/own/mat_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/own/mat_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/own/mat_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/own/scalar_tests.cpp b/thirdparty/fluid/modules/gapi/test/own/scalar_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/own/scalar_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/own/scalar_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/render/ftp_render_test.cpp b/thirdparty/fluid/modules/gapi/test/render/ftp_render_test.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/render/ftp_render_test.cpp
rename to thirdparty/fluid/modules/gapi/test/render/ftp_render_test.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/render/gapi_render_tests_ocv.cpp b/thirdparty/fluid/modules/gapi/test/render/gapi_render_tests_ocv.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/render/gapi_render_tests_ocv.cpp
rename to thirdparty/fluid/modules/gapi/test/render/gapi_render_tests_ocv.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/rmat/rmat_integration_tests.cpp b/thirdparty/fluid/modules/gapi/test/rmat/rmat_integration_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/rmat/rmat_integration_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/rmat/rmat_integration_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/rmat/rmat_test_common.hpp b/thirdparty/fluid/modules/gapi/test/rmat/rmat_test_common.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/rmat/rmat_test_common.hpp
rename to thirdparty/fluid/modules/gapi/test/rmat/rmat_test_common.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/rmat/rmat_tests.cpp b/thirdparty/fluid/modules/gapi/test/rmat/rmat_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/rmat/rmat_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/rmat/rmat_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/rmat/rmat_view_tests.cpp b/thirdparty/fluid/modules/gapi/test/rmat/rmat_view_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/rmat/rmat_view_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/rmat/rmat_view_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/s11n/gapi_s11n_tests.cpp b/thirdparty/fluid/modules/gapi/test/s11n/gapi_s11n_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/s11n/gapi_s11n_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/s11n/gapi_s11n_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/s11n/gapi_sample_pipelines_s11n.cpp b/thirdparty/fluid/modules/gapi/test/s11n/gapi_sample_pipelines_s11n.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/s11n/gapi_sample_pipelines_s11n.cpp
rename to thirdparty/fluid/modules/gapi/test/s11n/gapi_sample_pipelines_s11n.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/streaming/gapi_streaming_tests.cpp b/thirdparty/fluid/modules/gapi/test/streaming/gapi_streaming_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/streaming/gapi_streaming_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/streaming/gapi_streaming_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/test_main.cpp b/thirdparty/fluid/modules/gapi/test/test_main.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/test_main.cpp
rename to thirdparty/fluid/modules/gapi/test/test_main.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/test_precomp.hpp b/thirdparty/fluid/modules/gapi/test/test_precomp.hpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/test_precomp.hpp
rename to thirdparty/fluid/modules/gapi/test/test_precomp.hpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/util/any_tests.cpp b/thirdparty/fluid/modules/gapi/test/util/any_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/util/any_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/util/any_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/util/optional_tests.cpp b/thirdparty/fluid/modules/gapi/test/util/optional_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/util/optional_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/util/optional_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/modules/gapi/test/util/variant_tests.cpp b/thirdparty/fluid/modules/gapi/test/util/variant_tests.cpp
similarity index 100%
rename from inference-engine/thirdparty/fluid/modules/gapi/test/util/variant_tests.cpp
rename to thirdparty/fluid/modules/gapi/test/util/variant_tests.cpp
diff --git a/inference-engine/thirdparty/fluid/revision.txt b/thirdparty/fluid/revision.txt
similarity index 100%
rename from inference-engine/thirdparty/fluid/revision.txt
rename to thirdparty/fluid/revision.txt
diff --git a/inference-engine/thirdparty/fluid/update.sh b/thirdparty/fluid/update.sh
similarity index 100%
rename from inference-engine/thirdparty/fluid/update.sh
rename to thirdparty/fluid/update.sh
diff --git a/thirdparty/gflags b/thirdparty/gflags
new file mode 160000
index 00000000000000..e171aa2d15ed9e
--- /dev/null
+++ b/thirdparty/gflags
@@ -0,0 +1 @@
+Subproject commit e171aa2d15ed9eb17054558e0b3a6a413bb01067
diff --git a/inference-engine/thirdparty/ocv/opencv_hal.hpp b/thirdparty/ocv/opencv_hal.hpp
similarity index 100%
rename from inference-engine/thirdparty/ocv/opencv_hal.hpp
rename to thirdparty/ocv/opencv_hal.hpp
diff --git a/inference-engine/thirdparty/ocv/opencv_hal_avx.hpp b/thirdparty/ocv/opencv_hal_avx.hpp
similarity index 100%
rename from inference-engine/thirdparty/ocv/opencv_hal_avx.hpp
rename to thirdparty/ocv/opencv_hal_avx.hpp
diff --git a/inference-engine/thirdparty/ocv/opencv_hal_avx512.hpp b/thirdparty/ocv/opencv_hal_avx512.hpp
similarity index 100%
rename from inference-engine/thirdparty/ocv/opencv_hal_avx512.hpp
rename to thirdparty/ocv/opencv_hal_avx512.hpp
diff --git a/inference-engine/thirdparty/ocv/opencv_hal_intrin.hpp b/thirdparty/ocv/opencv_hal_intrin.hpp
similarity index 100%
rename from inference-engine/thirdparty/ocv/opencv_hal_intrin.hpp
rename to thirdparty/ocv/opencv_hal_intrin.hpp
diff --git a/inference-engine/thirdparty/ocv/opencv_hal_intrin_forward.hpp b/thirdparty/ocv/opencv_hal_intrin_forward.hpp
similarity index 100%
rename from inference-engine/thirdparty/ocv/opencv_hal_intrin_forward.hpp
rename to thirdparty/ocv/opencv_hal_intrin_forward.hpp
diff --git a/inference-engine/thirdparty/ocv/opencv_hal_neon.hpp b/thirdparty/ocv/opencv_hal_neon.hpp
similarity index 100%
rename from inference-engine/thirdparty/ocv/opencv_hal_neon.hpp
rename to thirdparty/ocv/opencv_hal_neon.hpp
diff --git a/inference-engine/thirdparty/ocv/opencv_hal_sse.hpp b/thirdparty/ocv/opencv_hal_sse.hpp
similarity index 100%
rename from inference-engine/thirdparty/ocv/opencv_hal_sse.hpp
rename to thirdparty/ocv/opencv_hal_sse.hpp

From f72e2481ffe60046fbf3778a78c7366c8d21ce55 Mon Sep 17 00:00:00 2001
From: Andrei Molotkov <andrei.molotkov@intel.com>
Date: Wed, 23 Jun 2021 10:00:18 +0300
Subject: [PATCH 171/178] [IE CLDNN] Improving prepare quantization pass
 (#5679)

---
 .../graph_optimizer/prepare_quantization.cpp  | 1290 ++++++++---------
 .../reverse_optional_nodes_outputs.cpp        |    4 +-
 .../clDNN/src/include/pass_manager.h          |   15 +-
 3 files changed, 611 insertions(+), 698 deletions(-)

diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_quantization.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_quantization.cpp
index 35d31d3a71d211..277676f87ac717 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_quantization.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/prepare_quantization.cpp
@@ -20,415 +20,412 @@
 #include <memory>
 #include <vector>
 
-void prepare_quantization::prepare_packed_quantize(program_impl& p) {
-    auto itr = p.get_processing_order().begin();
-    while (itr != p.get_processing_order().end()) {
-        auto node_itr = itr++;
-        auto &node = (*node_itr);
+template<typename T>
+bool check_binarization(memory::ptr mem_input_low, memory::ptr mem_input_high, program_impl& p) {
+    bool is_binarization = true;
+    const auto& stream = p.get_stream();
+    mem_lock<T> data_input_low_lock{mem_input_low, stream};
+    mem_lock<T> data_input_high_lock{mem_input_high, stream};
+    auto data_input_low = data_input_low_lock.data();
+    auto data_input_high = data_input_high_lock.data();
+    const size_t number_mem_layout_elements = mem_input_high->get_layout().count();
+    for (size_t i = 0; i < number_mem_layout_elements; i++) {
+        if (data_input_high[i] != data_input_low[i]) {
+            is_binarization = false;
+            break;
+        }
+    }
+    return is_binarization;
+}
 
-        program_helpers::do_for_types<quantize>(*node, [&](quantize_node& quantize_node) {
-            if (quantize_node.is_output())
-                return;
 
-            auto levels = quantize_node.get_primitive()->levels;
+void  prepare_quantization::prepare_scale_shift_opt(program_impl &p, quantize_node& quantize_node) {
+    const auto& stream = p.get_stream();
+    program_node &input_low_node = quantize_node.get_dependency(1);
+    program_node &input_high_node = quantize_node.get_dependency(2);
+    program_node &output_low_node = quantize_node.get_dependency(3);
+    program_node &output_high_node = quantize_node.get_dependency(4);
 
-            program_node &input_low_node = quantize_node.get_dependency(1);
-            program_node &input_high_node = quantize_node.get_dependency(2);
+    if (!input_low_node.is_type<data>() || !input_high_node.is_type<data>() ||
+        !output_low_node.is_type<data>() || !output_high_node.is_type<data>()) {
+        return;
+    }
 
-            if (!input_low_node.is_type<data>() || !input_high_node.is_type<data>()) {
-                return;
+    auto &input_low = input_low_node.as<data>();
+    auto &input_high = input_high_node.as<data>();
+    auto &output_low = output_low_node.as<data>();
+    auto &output_high = output_high_node.as<data>();
+
+    auto mem_input_low = input_low.get_attached_memory_ptr();
+    auto mem_input_high = input_high.get_attached_memory_ptr();
+    auto mem_output_low = output_low.get_attached_memory_ptr();
+    auto mem_output_high = output_high.get_attached_memory_ptr();
+
+    auto scales_layout = mem_input_low->get_layout();
+    scales_layout.size = tensor::max(scales_layout.size, mem_input_high->get_layout().size);
+    scales_layout.size = tensor::max(scales_layout.size, mem_output_low->get_layout().size);
+    scales_layout.size = tensor::max(scales_layout.size, mem_output_high->get_layout().size);
+
+    auto mem_input_scale  = p.get_engine().allocate_memory(scales_layout, false);
+    auto mem_input_shift  = p.get_engine().allocate_memory(scales_layout, false);
+    auto mem_output_scale = p.get_engine().allocate_memory(scales_layout, false);
+    auto mem_output_shift = p.get_engine().allocate_memory(scales_layout, false);
+
+    auto get_offset_safe = [](const layout& l, const tensor& idx) -> int {
+        auto sizes = l.size;
+        auto pitches = l.get_pitches();
+
+        return (idx.batch[0] % sizes.batch[0])*pitches.batch[0]
+                        + (idx.feature[0] % sizes.feature[0])*pitches.feature[0]
+                        + (idx.spatial[1] % sizes.spatial[1])*pitches.spatial[1]
+                        + (idx.spatial[0] % sizes.spatial[0])*pitches.spatial[0];
+    };
+
+    auto lock_memory = [&stream] (memory::ptr memory, std::function<void(std::size_t, float)>& set_data,
+                                  std::function<float(size_t)>& get_data) {
+        switch (memory->get_layout().data_type) {
+            case data_types::f32: {
+                std::shared_ptr<mem_lock<float>> data_lock_ptr = std::make_shared<mem_lock<float>>(memory, stream);
+                float* data = data_lock_ptr->data();
+                set_data = [data] (size_t idx, float value) {
+                    data[idx] = value;
+                };
+                get_data = [data] (size_t idx) {
+                    return data[idx];
+                };
+                return std::pair<std::shared_ptr<mem_lock<float>>, std::shared_ptr<mem_lock<uint16_t>>>(data_lock_ptr, nullptr);
+            }
+            case data_types::f16: {
+                std::shared_ptr<mem_lock<uint16_t>> data_lock_ptr = std::make_shared<mem_lock<uint16_t>>(memory, stream);
+                uint16_t* data = data_lock_ptr->data();
+                set_data = [data] (size_t idx, float value) {
+                    data[idx] = float_to_half(value);
+                };
+                get_data = [data] (size_t idx) {
+                    return half_to_float(data[idx]);
+                };
+                return std::pair<std::shared_ptr<mem_lock<float>>, std::shared_ptr<mem_lock<uint16_t>>>(nullptr, data_lock_ptr);
             }
+            default:
+                throw std::runtime_error("prepare_quantization: Unsupported precision of quantize output values");
+        }
+    };
+
+    std::function<void(size_t, float)> set_data_input_low;
+    std::function<float(size_t)> get_data_input_low;
+    auto input_low_locked_memory = lock_memory(mem_input_low, set_data_input_low, get_data_input_low);
+
+    std::function<void(size_t, float)> set_data_input_high;
+    std::function<float(size_t)> get_data_input_high;
+    auto input_high_locked_memory = lock_memory(mem_input_high, set_data_input_high, get_data_input_high);
+
+    std::function<void(size_t, float)> set_data_output_low;
+    std::function<float(size_t)> get_data_output_low;
+    auto output_low_locked_memory = lock_memory(mem_output_low, set_data_output_low, get_data_output_low);
+
+    std::function<void(size_t, float)> set_data_output_high;
+    std::function<float(size_t)> get_data_output_high;
+    auto output_high_locked_memory = lock_memory(mem_output_high, set_data_output_high, get_data_output_high);
+
+    std::function<void(std::size_t, float)> set_data_input_scale;
+    std::function<float(size_t)> get_data_input_scale;
+    auto input_scale_locked_memory = lock_memory(mem_input_scale, set_data_input_scale, get_data_input_scale);
+
+    std::function<void(size_t, float)> set_data_input_shift;
+    std::function<float(size_t)> get_data_input_shift;
+    auto input_shift_locked_memory = lock_memory(mem_input_shift, set_data_input_shift, get_data_input_shift);
+
+    std::function<void(size_t, float)> set_data_output_scale;
+    std::function<float(size_t)> get_data_output_scale;
+    auto output_scale_locked_memory = lock_memory(mem_output_scale, set_data_output_scale, get_data_output_scale);
+
+    std::function<void(size_t, float)> set_data_output_shift;
+    std::function<float(size_t)> get_data_output_shift;
+    auto output_shift_locked_memory = lock_memory(mem_output_shift, set_data_output_shift, get_data_output_shift);
+
+    bool has_negative_scales = false;
+    bool need_post_scale = false;
+    bool need_post_shift = false;
+    int levels = quantize_node.get_primitive()->levels;
+
+    for (int b = 0; b < scales_layout.size.batch[0]; b++) {
+        for (int f = 0; f < scales_layout.size.feature[0]; f++) {
+            for (int y = 0; y < scales_layout.size.spatial[1]; y++) {
+                for (int x = 0; x < scales_layout.size.spatial[0]; x++) {
+                    auto idx = cldnn::tensor(format::bfyx, {b, f, y, x}, 0);
+                    auto s_offset = scales_layout.get_linear_offset(idx);
+                    float in_lo = get_data_input_low(get_offset_safe(mem_input_low->get_layout(), idx));
+                    float in_hi = get_data_input_high(get_offset_safe(mem_input_high->get_layout(), idx));
+
+                    float out_lo = get_data_output_low(get_offset_safe(mem_output_low->get_layout(), idx));
+                    float out_hi = get_data_output_high(get_offset_safe(mem_output_high->get_layout(), idx));
+                    set_data_input_scale(s_offset, (static_cast<float>(levels) - 1.f) / (in_hi - in_lo));
+                    set_data_input_shift(s_offset, - in_lo * (static_cast<float>(levels) - 1.f) / (in_hi - in_lo));
+                    set_data_output_scale(s_offset, (out_hi - out_lo) / (static_cast<float>(levels) - 1.f));
+                    set_data_output_shift(s_offset, out_lo);
+
+                    if (get_data_output_scale(s_offset) != 1.0f) {
+                        need_post_scale = true;
+                    }
+                    if (get_data_output_shift(s_offset) != 0.0f) {
+                        need_post_shift = true;
+                    }
+                    if (get_data_input_scale(s_offset) < 0.0f) {
+                        has_negative_scales = true;
+                    }
+                }
+            }
+        }
+    }
 
-            auto &input_low = input_low_node.as<data>();
-            auto &input_high = input_high_node.as<data>();
+    bool need_pre_shift = false;
+    bool per_tensor_in_scale = true;
+    bool per_tensor_in_shift = true;
+    bool per_tensor_in_range = true;
+    bool per_tensor_out_scale = true;
+    bool per_tensor_out_shift = true;
+    float in_scale_val = get_data_input_scale(0);
+    float in_shift_val = get_data_input_shift(0);
+    float out_scale_val = get_data_output_scale(0);
+    float out_shift_val = get_data_output_shift(0);
+    float in_lo_val = get_data_input_low(0);
+    float in_hi_val = get_data_input_high(0);
+    for (size_t i = 0; i < scales_layout.count(); i++) {
+        if (in_scale_val != get_data_input_scale(i))
+            per_tensor_in_scale = false;
+        if (in_shift_val != get_data_input_shift(i))
+            per_tensor_in_shift = false;
+        if (out_scale_val != get_data_output_scale(i))
+            per_tensor_out_scale = false;
+        if (out_shift_val != get_data_output_shift(i))
+            per_tensor_out_shift = false;
+        if (get_data_input_shift(i) != 0.0f)
+            need_pre_shift = true;
+
+        if (in_lo_val != get_data_input_low(i % mem_input_low->get_layout().count()) ||
+            in_hi_val != get_data_input_high(i % mem_input_high->get_layout().count()))
+            per_tensor_in_range = false;
+    }
 
-            auto mem_input_low = input_low.get_attached_memory_ptr();
-            auto mem_input_high = input_high.get_attached_memory_ptr();
+    if (has_negative_scales) {
+        return;
+    }
 
-            auto output_dt = quantize_node.get_output_layout().data_type;
+    auto in_scale_prim = std::make_shared<data>(quantize_node.id() + "_in_scale", mem_input_scale);
+    auto in_shift_prim = std::make_shared<data>(quantize_node.id() + "_in_shift", mem_input_shift);
+    auto out_scale_prim = std::make_shared<data>(quantize_node.id() + "_output_scale", mem_output_scale);
+    auto out_shift_prim = std::make_shared<data>(quantize_node.id() + "_output_shift", mem_output_shift);
+    auto& in_scale_node = p.get_or_create(in_scale_prim);
+    auto& in_shift_node = p.get_or_create(in_shift_prim);
+    auto& out_scale_node = p.get_or_create(out_scale_prim);
+    auto& out_shift_node = p.get_or_create(out_shift_prim);
+
+    auto& inputs = p.get_inputs();
+
+    inputs.push_back(&in_scale_node);
+    inputs.push_back(&in_shift_node);
+    inputs.push_back(&out_scale_node);
+    inputs.push_back(&out_shift_node);
+
+    p.add_connection(in_scale_node, quantize_node);
+    p.add_connection(in_shift_node, quantize_node);
+    p.add_connection(out_scale_node, quantize_node);
+    p.add_connection(out_shift_node, quantize_node);
+    quantize_node.add_memory_dependency(in_scale_node.id());
+    quantize_node.add_memory_dependency(in_shift_node.id());
+    quantize_node.add_memory_dependency(out_scale_node.id());
+    quantize_node.add_memory_dependency(out_shift_node.id());
+    p.get_processing_order().insert(&quantize_node, &in_shift_node);
+    p.get_processing_order().insert(&quantize_node, &in_scale_node);
+    p.get_processing_order().insert(&quantize_node, &out_shift_node);
+    p.get_processing_order().insert(&quantize_node, &out_scale_node);
+
+    quantize_node.set_scale_shift_opt();
+
+    if (need_post_scale) {
+        quantize_node.set_need_post_scale();
+    }
 
-            auto& stream = p.get_stream();
+    if (need_post_shift) {
+        quantize_node.set_need_post_shift();
+    }
 
-            if (levels == 2) {
-                bool is_binarization = true;
-                switch (mem_input_high->get_layout().data_type) {
-                    case data_types::f32: {
-                        mem_lock<float> data_input_low_lock{mem_input_low, stream};
-                        mem_lock<float> data_input_high_lock{mem_input_high, stream};
-                        auto data_input_low = data_input_low_lock.data();
-                        auto data_input_high = data_input_high_lock.data();
+    if (need_pre_shift) {
+        quantize_node.set_need_pre_shift();
+    }
 
-                        for (size_t i = 0; i < mem_input_high->get_layout().count(); i++) {
-                            if (data_input_high[i] != data_input_low[i]) {
-                                is_binarization = false;
-                                break;
-                            }
-                        }
-                        break;
-                    }
-                    case data_types::f16: {
-                        mem_lock<uint16_t> data_input_low_lock{mem_input_low, stream};
-                        mem_lock<uint16_t> data_input_high_lock{mem_input_high, stream};
-                        auto data_input_low = data_input_low_lock.data();
-                        auto data_input_high = data_input_high_lock.data();
-
-                        for (size_t i = 0; i < mem_input_high->get_layout().count(); i++) {
-                            if (data_input_high[i] != data_input_low[i]) {
-                                is_binarization = false;
-                                break;
-                            }
-                        }
-                        break;
-                    }
-                    default:
-                        CLDNN_ERROR_MESSAGE(node->id(), "prepare_quantization: Unsupported precision of quantize inputs");
-                }
+    if (per_tensor_in_scale) {
+        quantize_node.set_per_tensor_input_scale();
+        quantize_node.set_input_scale_val(in_scale_val);
+    }
 
-                if (is_binarization) {
-                    output_dt = data_types::bin;
-                }
-            }
+    if (per_tensor_in_shift && need_pre_shift) {
+        quantize_node.set_per_tensor_input_shift();
+        quantize_node.set_input_shift_val(in_shift_val);
+    }
+
+    auto out_dt = quantize_node.get_output_layout().data_type;
+    bool need_clamp = levels != 256 || (out_dt != data_types::u8 && out_dt != data_types::i8);
+    if (need_clamp) {
+        quantize_node.set_need_clamp();
+    }
 
-            quantize_node.typed_desc()->output_data_type = optional_data_type{output_dt};
-            quantize_node.recalc_output_layout();
-        });
+    if (per_tensor_in_range) {
+        quantize_node.set_per_tensor_input_range();
+        quantize_node.set_input_lo_val(in_lo_val);
+        quantize_node.set_input_hi_val(in_hi_val);
+    }
+    if (per_tensor_out_scale) {
+        quantize_node.set_per_tensor_output_scale();
+        quantize_node.set_output_scale_val(out_scale_val);
+    }
+
+    if (per_tensor_out_shift) {
+        quantize_node.set_per_tensor_output_shift();
+        quantize_node.set_output_shift_val(out_shift_val);
     }
 }
 
-void prepare_quantization::prepare_scale_shift_opt(program_impl &p) {
-    auto itr = p.get_processing_order().begin();
-    while (itr != p.get_processing_order().end()) {
-        auto node_itr = itr++;
-        auto &node = (*node_itr);
+void prepare_quantization::handle_quantize_node(program_impl& p, quantize_node& quantize_node) {
+    if (quantize_node.get_primitive()->levels == 2) {
+        prepare_packed_quantize(p, quantize_node);
+    } else if (quantize_node.get_primitive()->levels <= 256 && !quantize_node.get_scale_shift_opt() && !quantize_node.is_constant()) {
+        prepare_scale_shift_opt(p, quantize_node);
+    }
+}
 
-        program_helpers::do_for_types<quantize>(*node, [&](quantize_node& quantize_node) {
-            auto levels = quantize_node.get_primitive()->levels;
-            if (levels == 2 || levels > 256 || quantize_node.get_scale_shift_opt() || quantize_node.is_constant())
-                return;
+void prepare_quantization::prepare_packed_quantize(program_impl& p, quantize_node& quantize_node) {
+    program_node &input_low_node = quantize_node.get_dependency(1);
+    program_node &input_high_node = quantize_node.get_dependency(2);
 
-            auto& stream = p.get_stream();
+    if (quantize_node.is_output() || !input_low_node.is_type<data>() || !input_high_node.is_type<data>()) {
+        return;
+    }
 
-            program_node &input_low_node = quantize_node.get_dependency(1);
-            program_node &input_high_node = quantize_node.get_dependency(2);
-            program_node &output_low_node = quantize_node.get_dependency(3);
-            program_node &output_high_node = quantize_node.get_dependency(4);
+    auto &input_low = input_low_node.as<data>();
+    auto &input_high = input_high_node.as<data>();
 
-            if (!input_low_node.is_type<data>() || !input_high_node.is_type<data>() ||
-                !output_low_node.is_type<data>() || !output_high_node.is_type<data>()) {
-                return;
-            }
+    auto mem_input_low = input_low.get_attached_memory_ptr();
+    auto mem_input_high = input_high.get_attached_memory_ptr();
 
-            auto &input_low = input_low_node.as<data>();
-            auto &input_high = input_high_node.as<data>();
-            auto &output_low = output_low_node.as<data>();
-            auto &output_high = output_high_node.as<data>();
-
-            auto mem_input_low = input_low.get_attached_memory_ptr();
-            auto mem_input_high = input_high.get_attached_memory_ptr();
-            auto mem_output_low = output_low.get_attached_memory_ptr();
-            auto mem_output_high = output_high.get_attached_memory_ptr();
-
-            auto scales_layout = mem_input_low->get_layout();
-            scales_layout.size = tensor::max(scales_layout.size, mem_input_high->get_layout().size);
-            scales_layout.size = tensor::max(scales_layout.size, mem_output_low->get_layout().size);
-            scales_layout.size = tensor::max(scales_layout.size, mem_output_high->get_layout().size);
-
-            auto mem_input_scale  = p.get_engine().allocate_memory(scales_layout, false);
-            auto mem_input_shift  = p.get_engine().allocate_memory(scales_layout, false);
-            auto mem_output_scale = p.get_engine().allocate_memory(scales_layout, false);
-            auto mem_output_shift = p.get_engine().allocate_memory(scales_layout, false);
-
-            auto get_offset_safe = [](layout l, tensor idx) -> int {
-                auto sizes = l.size;
-                auto pitches = l.get_pitches();
-
-                int offset = (idx.batch[0] % sizes.batch[0])*pitches.batch[0]
-                             + (idx.feature[0] % sizes.feature[0])*pitches.feature[0]
-                             + (idx.spatial[1] % sizes.spatial[1])*pitches.spatial[1]
-                             + (idx.spatial[0] % sizes.spatial[0])*pitches.spatial[0];
-                return offset;
-            };
-
-            bool has_negative_scales = false;
-            bool need_post_scale = false;
-            bool need_post_shift = false;
-            bool need_pre_shift = false;
-            auto out_dt = quantize_node.get_output_layout().data_type;
-            bool need_clamp = levels != 256 || (out_dt != data_types::u8 && out_dt != data_types::i8);
-            bool per_tensor_in_scale = true;
-            bool per_tensor_in_shift = true;
-            bool per_tensor_in_range = true;
-            bool per_tensor_out_scale = true;
-            bool per_tensor_out_shift = true;
-            float in_scale_val = 0.0f;
-            float in_shift_val = 0.0f;
-            float out_scale_val = 0.0f;
-            float out_shift_val = 0.0f;
-            float in_lo_val = 0.0f;
-            float in_hi_val = 0.0f;
-            switch (mem_output_high->get_layout().data_type) {
-                case data_types::f32: {
-                    // TODO [LOW PRECISION]: Output low/high values can be removed.
-                    mem_lock<float> data_input_low_lock{mem_input_low, stream};
-                    mem_lock<float> data_input_high_lock{mem_input_high, stream};
-                    mem_lock<float> data_output_low_lock{mem_output_low, stream};
-                    mem_lock<float> data_output_high_lock{mem_output_high, stream};
-                    mem_lock<float> data_input_scale_lock{mem_input_scale, stream};
-                    mem_lock<float> data_input_shift_lock{mem_input_shift, stream};
-                    mem_lock<float> data_output_scale_lock{mem_output_scale, stream};
-                    mem_lock<float> data_output_shift_lock{mem_output_shift, stream};
-
-                    auto data_input_low = data_input_low_lock.data();
-                    auto data_input_high = data_input_high_lock.data();
-                    auto data_output_low = data_output_low_lock.data();
-                    auto data_output_high = data_output_high_lock.data();
-                    auto data_input_scale = data_input_scale_lock.data();
-                    auto data_input_shift = data_input_shift_lock.data();
-                    auto data_output_scale = data_output_scale_lock.data();
-                    auto data_output_shift = data_output_shift_lock.data();
-
-                    for (int b = 0; b < scales_layout.size.batch[0]; b++) {
-                        for (int f = 0; f < scales_layout.size.feature[0]; f++) {
-                            for (int y = 0; y < scales_layout.size.spatial[1]; y++) {
-                                for (int x = 0; x < scales_layout.size.spatial[0]; x++) {
-                                    auto idx = cldnn::tensor(format::bfyx, {b, f, y, x}, 0);
-                                    auto s_offset = scales_layout.get_linear_offset(idx);
-                                    auto in_lo = data_input_low[get_offset_safe(mem_input_low->get_layout(), idx)];
-                                    auto in_hi = data_input_high[get_offset_safe(mem_input_high->get_layout(), idx)];
-
-                                    auto out_lo = data_output_low[get_offset_safe(mem_output_low->get_layout(), idx)];
-                                    auto out_hi = data_output_high[get_offset_safe(mem_output_high->get_layout(), idx)];
-                                    data_input_scale[s_offset] = (static_cast<float>(levels) - 1) / (in_hi - in_lo);
-                                    data_input_shift[s_offset] = - in_lo * (static_cast<float>(levels) - 1) / (in_hi - in_lo);
-                                    data_output_scale[s_offset] = (out_hi - out_lo) / (static_cast<float>(levels) - 1);
-                                    data_output_shift[s_offset] = out_lo;
-
-                                    if (data_output_scale[s_offset] != 1.0f) {
-                                        need_post_scale = true;
-                                    }
-                                    if (data_output_shift[s_offset] != 0.0f) {
-                                        need_post_shift = true;
-                                    }
-                                    if (data_input_scale[s_offset] < 0.0f) {
-                                        has_negative_scales = true;
-                                    }
-                                }
-                            }
-                        }
-                    }
+    bool is_binarization = true;
+    switch (mem_input_high->get_layout().data_type) {
+        case data_types::f32: {
+            is_binarization = check_binarization<float>(mem_input_low, mem_input_high, p);
+            break;
+        }
+        case data_types::f16: {
+            is_binarization = check_binarization<uint16_t>(mem_input_low, mem_input_high, p);
+            break;
+        }
+        default:
+            CLDNN_ERROR_MESSAGE(quantize_node.id(), "prepare_quantization: Unsupported precision of quantize inputs");
+    }
 
-                    in_scale_val = data_input_scale[0];
-                    in_shift_val = data_input_shift[0];
-                    out_scale_val = data_output_scale[0];
-                    out_shift_val = data_output_shift[0];
-                    in_lo_val = data_input_low[0];
-                    in_hi_val = data_input_high[0];
-
-                    for (size_t i = 0; i < scales_layout.count(); i++) {
-                        if (in_scale_val != data_input_scale[i])
-                            per_tensor_in_scale = false;
-                        if (in_shift_val != data_input_shift[i])
-                            per_tensor_in_shift = false;
-                        if (out_scale_val != data_output_scale[i])
-                            per_tensor_out_scale = false;
-                        if (out_shift_val != data_output_shift[i])
-                            per_tensor_out_shift = false;
-                        if (data_input_shift[i] != 0.0f)
-                            need_pre_shift = true;
-
-                        if (in_lo_val != data_input_low[i % mem_input_low->get_layout().count()] ||
-                            in_hi_val != data_input_high[i % mem_input_high->get_layout().count()])
-                            per_tensor_in_range = false;
-                    }
-                    break;
-                }
-                case data_types::f16: {
-                    mem_lock<uint16_t> data_input_low_lock{mem_input_low, stream};
-                    mem_lock<uint16_t> data_input_high_lock{mem_input_high, stream};
-                    mem_lock<uint16_t> data_output_low_lock{mem_output_low, stream};
-                    mem_lock<uint16_t> data_output_high_lock{mem_output_high, stream};
-                    mem_lock<uint16_t> data_input_scale_lock{mem_input_scale, stream};
-                    mem_lock<uint16_t> data_input_shift_lock{mem_input_shift, stream};
-                    mem_lock<uint16_t> data_output_scale_lock{mem_output_scale, stream};
-                    mem_lock<uint16_t> data_output_shift_lock{mem_output_shift, stream};
-
-                    auto data_input_low = data_input_low_lock.data();
-                    auto data_input_high = data_input_high_lock.data();
-                    auto data_output_low = data_output_low_lock.data();
-                    auto data_output_high = data_output_high_lock.data();
-                    auto data_input_scale = data_input_scale_lock.data();
-                    auto data_input_shift = data_input_shift_lock.data();
-                    auto data_output_scale = data_output_scale_lock.data();
-                    auto data_output_shift = data_output_shift_lock.data();
-
-                    for (int b = 0; b < scales_layout.size.batch[0]; b++) {
-                        for (int f = 0; f < scales_layout.size.feature[0]; f++) {
-                            for (int y = 0; y < scales_layout.size.spatial[1]; y++) {
-                                for (int x = 0; x < scales_layout.size.spatial[0]; x++) {
-                                    auto idx = cldnn::tensor(format::bfyx, {b, f, y, x}, 0);
-                                    auto s_offset = scales_layout.get_linear_offset(idx);
-                                    auto in_lo = half_to_float(data_input_low[get_offset_safe(mem_input_low->get_layout(), idx)]);
-                                    auto in_hi = half_to_float(data_input_high[get_offset_safe(mem_input_high->get_layout(), idx)]);
-
-                                    auto out_lo = half_to_float(data_output_low[get_offset_safe(mem_output_low->get_layout(), idx)]);
-                                    auto out_hi = half_to_float(data_output_high[get_offset_safe(mem_output_high->get_layout(), idx)]);
-                                    data_input_scale[s_offset] = float_to_half((static_cast<float>(levels) - 1) / (in_hi - in_lo));
-                                    data_input_shift[s_offset] = float_to_half(- in_lo * (static_cast<float>(levels) - 1) / (in_hi - in_lo));
-                                    data_output_scale[s_offset] = float_to_half((out_hi - out_lo) / (static_cast<float>(levels) - 1));
-                                    data_output_shift[s_offset] = float_to_half(out_lo);
-
-                                    if (half_to_float(data_output_scale[s_offset]) != 1.0f) {
-                                        need_post_scale = true;
-                                    }
-                                    if (half_to_float(data_output_shift[s_offset]) != 0.0f) {
-                                        need_post_shift = true;
-                                    }
-                                    if (half_to_float(data_input_scale[s_offset]) < 0.0f) {
-                                        has_negative_scales = true;
-                                    }
-                                }
-                            }
-                        }
-                    }
+    auto output_dt = quantize_node.get_output_layout().data_type;
+    if (is_binarization) {
+        output_dt = data_types::bin;
+    }
 
-                    in_scale_val = half_to_float(data_input_scale[0]);
-                    in_shift_val = half_to_float(data_input_shift[0]);
-                    out_scale_val = half_to_float(data_output_scale[0]);
-                    out_shift_val = half_to_float(data_output_shift[0]);
-                    in_lo_val = half_to_float(data_input_low[0]);
-                    in_hi_val = half_to_float(data_input_high[0]);
-                    for (size_t i = 0; i < scales_layout.count(); i++) {
-                        if (in_scale_val != half_to_float(data_input_scale[i]))
-                            per_tensor_in_scale = false;
-                        if (in_shift_val != half_to_float(data_input_shift[i]))
-                            per_tensor_in_shift = false;
-                        if (out_scale_val != half_to_float(data_output_scale[i]))
-                            per_tensor_out_scale = false;
-                        if (out_shift_val != half_to_float(data_output_shift[i]))
-                            per_tensor_out_shift = false;
-                        if (half_to_float(data_input_shift[i]) != 0.0f)
-                            need_pre_shift = true;
-
-                        if (in_lo_val != half_to_float(data_input_low[i % mem_input_low->get_layout().count()]) ||
-                            in_hi_val != half_to_float(data_input_high[i % mem_input_high->get_layout().count()]))
-                            per_tensor_in_range = false;
-                    }
-                    break;
-                }
-                default:
-                    throw std::runtime_error("prepare_quantization: Unsupported precision of quantize output values");
-            }
+    quantize_node.typed_desc()->output_data_type = optional_data_type{output_dt};
+    quantize_node.recalc_output_layout();
+}
 
-            if (has_negative_scales) {
-                return;
-            }
+void prepare_quantization::prepare_dequantize_merge(program_impl& p, eltwise_node& eltwise_node) {
+    for (size_t i = 1; i < eltwise_node.get_dependencies().size(); i++) {
+        if (!eltwise_node.get_dependency(i).is_type<data>()) {
+            return;
+        }
+    }
 
-            auto in_scale_prim = std::make_shared<data>(quantize_node.id() + "_in_scale", mem_input_scale);
-            auto in_shift_prim = std::make_shared<data>(quantize_node.id() + "_in_shift", mem_input_shift);
-            auto out_scale_prim = std::make_shared<data>(quantize_node.id() + "_output_scale", mem_output_scale);
-            auto out_shift_prim = std::make_shared<data>(quantize_node.id() + "_output_shift", mem_output_shift);
-            auto& in_scale_node = p.get_or_create(in_scale_prim);
-            auto& in_shift_node = p.get_or_create(in_shift_prim);
-            auto& out_scale_node = p.get_or_create(out_scale_prim);
-            auto& out_shift_node = p.get_or_create(out_shift_prim);
-
-            auto& inputs = p.get_inputs();
-
-            inputs.push_back(&in_scale_node);
-            inputs.push_back(&in_shift_node);
-            inputs.push_back(&out_scale_node);
-            inputs.push_back(&out_shift_node);
-
-            p.add_connection(in_scale_node, quantize_node);
-            p.add_connection(in_shift_node, quantize_node);
-            p.add_connection(out_scale_node, quantize_node);
-            p.add_connection(out_shift_node, quantize_node);
-            quantize_node.add_memory_dependency(in_scale_node.id());
-            quantize_node.add_memory_dependency(in_shift_node.id());
-            quantize_node.add_memory_dependency(out_scale_node.id());
-            quantize_node.add_memory_dependency(out_shift_node.id());
-            p.get_processing_order().insert(&quantize_node, &in_shift_node);
-            p.get_processing_order().insert(&quantize_node, &in_scale_node);
-            p.get_processing_order().insert(&quantize_node, &out_shift_node);
-            p.get_processing_order().insert(&quantize_node, &out_scale_node);
-
-            quantize_node.set_scale_shift_opt();
-
-            if (need_post_scale) {
-                quantize_node.set_need_post_scale();
-            }
+    auto get_scale_shift_mem = [](const cldnn::eltwise_node& eltw, size_t dep_id) -> memory::ptr {
+        if (dep_id >= eltw.get_dependencies().size())
+            CLDNN_ERROR_MESSAGE(eltw.id(), "Invalid dependency id in dequantize optimization");
 
-            if (need_post_shift) {
-                quantize_node.set_need_post_shift();
-            }
+        return eltw.get_dependency(dep_id).as<data>().get_attached_memory_ptr();
+    };
 
-            if (need_pre_shift) {
-                quantize_node.set_need_pre_shift();
-            }
+    const auto& eltw_mode = eltwise_node.get_primitive()->mode;
+    if (eltw_mode != eltwise_mode::sum && eltw_mode != eltwise_mode::prod)
+        return;
 
-            if (per_tensor_in_scale) {
-                quantize_node.set_per_tensor_input_scale();
-                quantize_node.set_input_scale_val(in_scale_val);
-            }
+    auto& input = eltwise_node.input();
+    const auto& stream = p.get_stream();
 
-            if (per_tensor_in_shift && need_pre_shift) {
-                quantize_node.set_per_tensor_input_shift();
-                quantize_node.set_input_shift_val(in_shift_val);
-            }
+    for (auto& user : input.get_users()) {
+        if (user == &eltwise_node)
+            continue;
 
-            if (need_clamp) {
-                quantize_node.set_need_clamp();
-            }
+        if (!user->is_type<eltwise>() || user->get_dependencies().size() != eltwise_node.get_dependencies().size())
+            continue;
 
-            if (per_tensor_in_range) {
-                quantize_node.set_per_tensor_input_range();
-                quantize_node.set_input_lo_val(in_lo_val);
-                quantize_node.set_input_hi_val(in_hi_val);
-            }
-            if (per_tensor_out_scale) {
-                quantize_node.set_per_tensor_output_scale();
-                quantize_node.set_output_scale_val(out_scale_val);
-            }
+        auto& eltwise_dep = user->as<eltwise>();
+        if (eltwise_dep.get_primitive()->mode != eltwise_node.get_primitive()->mode)
+            continue;
 
-            if (per_tensor_out_shift) {
-                quantize_node.set_per_tensor_output_shift();
-                quantize_node.set_output_shift_val(out_shift_val);
+        bool valid_scale_node = true;
+        for (size_t i = 1; i < eltwise_dep.get_dependencies().size(); i++) {
+            if (!eltwise_dep.get_dependency(i).is_type<data>()) {
+                valid_scale_node = false;
             }
-        });
-    }
-}
-
-void prepare_quantization::remove_fake_reorders(program_impl& p) {
-    auto itr = p.get_processing_order().begin();
-    while (itr != p.get_processing_order().end()) {
-        auto node_itr = itr++;
-        auto &node = (*node_itr);
+        }
 
-        if (!node->is_type<reorder>() || !node->is_in_data_flow())
+        if (!valid_scale_node)
             continue;
 
-        if (node->get_users().size() != 1 || node->get_dependencies().size() != 1)
-            continue;
+        bool same_params = true;
+        for (size_t i = 1; i < eltwise_node.get_dependencies().size(); i++) {
+            auto mem0 = get_scale_shift_mem(eltwise_dep, i);
+            auto mem1 = get_scale_shift_mem(eltwise_node, i);
 
-        auto &usr = node->get_users().front();
-        auto &dep = node->get_dependency(0);
-        if (!(usr->is_type<convolution>() && usr->get_dependency(1).get_output_layout().data_type == data_types::i8) ||
-            !dep.is_input() ||
-            dep.get_output_layout().data_type != data_types::u8 ||
-            (node->get_output_layout().data_type != data_types::f32 && node->get_output_layout().data_type != data_types::f16) ||
-            dep.get_output_layout().format != node->get_output_layout().format ||
-            dep.get_output_layout().size != node->get_output_layout().size)
-            continue;
+            mem_lock<uint8_t> mem0_lock{mem0, stream};
+            mem_lock<uint8_t> mem1_lock{mem1, stream};
+            auto ptr0 = mem0_lock.data();
+            auto ptr1 = mem1_lock.data();
 
-        p.replace_all_usages(*node, dep);
-        p.add_optimized_primitive_info(node->id());
-        p.remove_all_connections(*node);
-        p.remove_if_dangling(*node);
+            for (size_t j = 0; j < mem0->get_layout().bytes_count(); j++) {
+                if (ptr0[j] != ptr1[j]) {
+                    same_params = false;
+                    break;
+                }
+            }
+        }
+
+        if (same_params) {
+            while (!eltwise_node.get_dependencies().empty()) {
+                auto& dep = eltwise_node.get_dependency(0);
+                p.remove_connection(dep, eltwise_node);
+                p.remove_if_dangling(dep);
+            }
+            p.add_optimized_primitive_info(eltwise_node.id(), {user->id()});
+            p.replace_all_usages(eltwise_node, *user);
+        }
     }
 }
 
+void prepare_quantization::remove_fake_reorders(program_impl& p, reorder_node& reorder_node) {
+    if (!reorder_node.is_in_data_flow() || reorder_node.get_users().size() != 1 || reorder_node.get_dependencies().size() != 1) {
+        return;
+    }
+
+    auto &usr = reorder_node.get_users().front();
+    auto &dep = reorder_node.get_dependency(0);
+    if (!(usr->is_type<convolution>() && usr->get_dependency(1).get_output_layout().data_type == data_types::i8) ||
+        !dep.is_input() ||
+        dep.get_output_layout().data_type != data_types::u8 ||
+        (reorder_node.get_output_layout().data_type != data_types::f32 && reorder_node.get_output_layout().data_type != data_types::f16) ||
+        dep.get_output_layout().format != reorder_node.get_output_layout().format ||
+        dep.get_output_layout().size != reorder_node.get_output_layout().size)
+        return;
+
+    p.replace_all_usages(reorder_node, dep);
+    p.add_optimized_primitive_info(reorder_node.id());
+    p.remove_all_connections(reorder_node);
+    p.remove_if_dangling(reorder_node);
+}
+
 template<typename W_T, typename AZP_T>
-void fill_compensation_typed(W_T* w, AZP_T* azp, W_T* wzp, float* comp, int GS, int OC, int IC, int KS) {
+void fill_compensation_typed(W_T* w, AZP_T* azp, W_T* wzp, float* comp, const int GS, const int OC, const int IC, const int KS) {
     for (int g = 0; g < GS; g++) {
         for (int oc = 0; oc < OC; oc++) {
             float c = 0.f;
@@ -445,381 +442,294 @@ void fill_compensation_typed(W_T* w, AZP_T* azp, W_T* wzp, float* comp, int GS,
                     }
                 }
             }
-
             comp[g*OC + oc] = -c;
         }
     }
 }
 
-void prepare_quantization::prepare_asymmetric_quantization(program_impl &p) {
-    auto itr = p.get_processing_order().begin();
-    while (itr != p.get_processing_order().end()) {
-        auto node_itr = itr++;
-        auto& node = (*node_itr);
-
-        auto& stream = p.get_stream();
+void prepare_quantization::prepare_asymmetric_quantization(program_impl &p, convolution_node& convolution_node) {
+    // Detects if given eltwise node performs zero point subtraction
+    auto is_zero_point_node = [](const eltwise_node& node) -> bool {
+        auto prim = node.get_primitive();
 
-        // Detects if given eltwise node performs zero point subtraction
-        auto is_zero_point_node = [](eltwise_node& node) -> bool {
-            auto prim = node.get_primitive();
+        if (node.get_dependencies().size() != 2 || prim->mode != eltwise_mode::sub)
+            return false;
 
-            if (node.get_dependencies().size() != 2 || prim->mode != eltwise_mode::sub)
-                return false;
+        if (node.get_users().size() != 1)
+            return false;
 
-            if (node.get_users().size() != 1)
-                return false;
+        auto in0_layout = node.get_dependency(0).get_output_layout();
+        auto in1_layout = node.get_dependency(1).get_output_layout();
 
-            auto in0_layout = node.get_dependency(0).get_output_layout();
-            auto in1_layout = node.get_dependency(1).get_output_layout();
+        if (!node.get_dependency(1).is_type<data>())
+            return false;
 
-            if (!node.get_dependency(1).is_type<data>())
-                return false;
+        // Check if sub inputs are quantized
+        if (in0_layout.data_type != data_types::u8 && in0_layout.data_type != data_types::i8)
+            return false;
 
-            // Check if sub inputs are quantized
-            if (in0_layout.data_type != data_types::u8 && in0_layout.data_type != data_types::i8)
-                return false;
+        // Zero point must have the same type as quantized data
+        if (in0_layout.data_type != in1_layout.data_type)
+            return false;
 
-            // Zero point must have the same type as quantized data
-            if (in0_layout.data_type != in1_layout.data_type)
-                return false;
+        return true;
+    };
 
-            return true;
-        };
+    const auto& stream = p.get_stream();
+    auto fill_compensation = [&](int groups, const memory::ptr w, const memory::ptr azp, const memory::ptr wzp, memory::ptr compensation) {
+        const auto& wl = w->get_layout();
 
-        auto fill_compensation = [&](int groups, const memory::ptr w, const  memory::ptr azp, const memory::ptr wzp, memory::ptr compensation) {
-            auto wl = w->get_layout();
+        const int GS = groups;
+        const int OC = wl.size.batch[0] / GS;
+        const int IC = wl.size.feature[0];  // already divided by GS
+        const int KS = wl.size.spatial[0]*wl.size.spatial[1]*wl.size.spatial[2];
 
-            int GS = groups;
-            int OC = wl.size.batch[0] / GS;
-            int IC = wl.size.feature[0];  // already divided by GS
-            int KS = wl.size.spatial[0]*wl.size.spatial[1]*wl.size.spatial[2];
+        const auto& w_dt = wl.data_type;
+        const auto& azp_dt = azp->get_layout().data_type;
 
-            auto w_dt = wl.data_type;
-            auto azp_dt = azp->get_layout().data_type;
+        mem_lock<float> comp_lock{compensation, stream};
 
-            mem_lock<float> comp_lock{compensation, stream};
-
-            if (w_dt == data_types::u8 && azp_dt == data_types::u8) {
-                mem_lock<uint8_t> w_lock(w, stream);
-                mem_lock<uint8_t> azp_lock(azp, stream);
-                if (wzp) {
-                    mem_lock<uint8_t> wzp_lock(wzp, stream);
-                    fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
-                } else {
-                    fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<uint8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
-                }
-            } else if (w_dt == data_types::i8 && azp_dt == data_types::u8) {
-                mem_lock<int8_t> w_lock(w, stream);
-                mem_lock<uint8_t> azp_lock(azp, stream);
-                if (wzp) {
-                    mem_lock<int8_t> wzp_lock(wzp, stream);
-                    fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
-                } else {
-                    fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<int8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
-                }
-            } else if (w_dt == data_types::i8 && azp_dt == data_types::i8) {
-                mem_lock<int8_t> w_lock(w, stream);
-                mem_lock<int8_t> azp_lock(azp, stream);
-                if (wzp) {
-                    mem_lock<int8_t> wzp_lock(wzp, stream);
-                    fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
-                } else {
-                    fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<int8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
-                }
-            } else if (w_dt == data_types::u8 && azp_dt == data_types::i8) {
-                mem_lock<uint8_t> w_lock(w, stream);
-                mem_lock<int8_t> azp_lock(azp, stream);
-                if (wzp) {
-                    mem_lock<uint8_t> wzp_lock(wzp, stream);
-                    fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
-                } else {
-                    fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<uint8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
-                }
+        if (w_dt == data_types::u8 && azp_dt == data_types::u8) {
+            mem_lock<uint8_t> w_lock(w, stream);
+            mem_lock<uint8_t> azp_lock(azp, stream);
+            if (wzp) {
+                mem_lock<uint8_t> wzp_lock(wzp, stream);
+                fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
+            } else {
+                fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<uint8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
             }
-        };
-
-        auto asymmetric_convolution_f = [&](convolution_node& convolution_node) {
-            auto& in0 = convolution_node.get_dependency(0);
-            auto& in1 = convolution_node.get_dependency(1);
-
-            bool asymmetric_data = in0.is_type<eltwise>() && is_zero_point_node(in0.as<eltwise>());
-            bool asymmetric_weights = in1.is_type<eltwise>() && is_zero_point_node(in1.as<eltwise>());
-
-            if (!asymmetric_data && !asymmetric_weights)
-                return;
-
-            // Input that doesn't match asymmetric pattern should be quantized
-            // Cases like fp32 input + i8 weights + i8 w_zp are not supported
-            if (!asymmetric_data &&
-                in0.get_output_layout().data_type != data_types::u8 &&
-                in0.get_output_layout().data_type != data_types::i8)
-                return;
-
-            if (!asymmetric_weights &&
-                in1.get_output_layout().data_type != data_types::u8 &&
-                in1.get_output_layout().data_type != data_types::i8)
-                return;
-
-            auto old_conv_prim = convolution_node.get_primitive();
-
-            // Split is not supported
-            if (old_conv_prim->weights.size() > 1)
-                return;
-
-
-            primitive_id input = old_conv_prim->input[0];
-            std::vector<primitive_id> weights = old_conv_prim->weights;
-            std::vector<primitive_id> w_zero_points = {};
-            std::vector<primitive_id> a_zero_points = {};
-            std::vector<primitive_id> compensation = {};
-
-            cldnn::program_node* new_input = &in0;
-            cldnn::program_node* new_weights = &in1;
-            cldnn::program_node* new_bias = !old_conv_prim->bias.empty() ? &convolution_node.get_dependency(2) : nullptr;
-            cldnn::program_node* new_a_zp = nullptr;
-            cldnn::program_node* new_w_zp = nullptr;
-            cldnn::program_node* new_compenstation = nullptr;
-
-            bool need_compensation = false;
-
-            auto output_size = convolution_node.get_output_layout().size;
-            int ofm = in1.get_output_layout().size.batch[0];
-            int ifm = in0.get_output_layout().size.feature[0];
-            int ofm_aligned = ((ofm + 31) / 32) * 32;
-            int ifm_aligned = ((ifm + 31) / 32) * 32;
-            int groups = static_cast<int>(convolution_node.get_groups());
-
-            if (asymmetric_data) {
-                new_input = &in0.get_dependency(0);
-                new_a_zp = &in0.get_dependency(1);
-
-                auto l = layout{new_a_zp->get_output_layout().data_type, format::bfyx, tensor{1, ifm_aligned, 1, 1}};
-                int s = new_a_zp->get_output_layout().size.feature[0];
-                auto azp_aligned = p.get_engine().allocate_memory(l);
-                auto old_ptr = new_a_zp->as<data>().get_attached_memory_ptr();
-                mem_lock<int8_t> new_data{azp_aligned, stream};
-                mem_lock<int8_t> old_data{old_ptr, stream};
-                for (int i = 0; i < ifm_aligned; i++) {
-                    new_data.data()[i] = old_data.data()[i % s];
-                }
-                new_a_zp->as<data>().attach_memory(azp_aligned);
-
-                input = new_input->id();
-                a_zero_points.push_back(new_a_zp->id());
-                need_compensation = true;
+        } else if (w_dt == data_types::i8 && azp_dt == data_types::u8) {
+            mem_lock<int8_t> w_lock(w, stream);
+            mem_lock<uint8_t> azp_lock(azp, stream);
+            if (wzp) {
+                mem_lock<int8_t> wzp_lock(wzp, stream);
+                fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
+            } else {
+                fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<int8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
             }
-
-            if (asymmetric_weights) {
-                new_weights = &in1.get_dependency(0);
-                new_w_zp = &in1.get_dependency(1);
-
-                auto l = layout{new_w_zp->get_output_layout().data_type, format::bfyx, tensor{ofm_aligned, 1, 1, 1}};
-                int s = new_w_zp->get_output_layout().size.batch[0];
-                auto wzp_aligned = p.get_engine().allocate_memory(l);
-                auto old_ptr = new_w_zp->as<data>().get_attached_memory_ptr();
-                mem_lock<int8_t> new_data{wzp_aligned, stream};
-                mem_lock<int8_t> old_data{old_ptr, stream};
-                for (int i = 0; i < ofm_aligned; i++) {
-                    new_data.data()[i] = old_data.data()[i % s];
-                }
-                new_w_zp->as<data>().attach_memory(wzp_aligned);
-
-                weights = { new_weights->id() };
-                w_zero_points.push_back(new_w_zp->id());
+        } else if (w_dt == data_types::i8 && azp_dt == data_types::i8) {
+            mem_lock<int8_t> w_lock(w, stream);
+            mem_lock<int8_t> azp_lock(azp, stream);
+            if (wzp) {
+                mem_lock<int8_t> wzp_lock(wzp, stream);
+                fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
+            } else {
+                fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<int8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
             }
-
-            if (need_compensation) {
-                auto l = layout{data_types::f32, format::bfyx, tensor{1, ofm_aligned, 1, 1}};
-                auto data_to_allocate = p.get_engine().allocate_memory(l);
-                auto w = new_weights->as<data>().get_attached_memory_ptr();
-                auto azp = asymmetric_data ? new_a_zp->as<data>().get_attached_memory_ptr() : nullptr;
-                auto wzp = asymmetric_weights ? new_w_zp->as<data>().get_attached_memory_ptr() : nullptr;
-                fill_compensation(groups, w, azp, wzp, data_to_allocate);
-
-                auto compensation_prim = std::make_shared<data>(convolution_node.id() + "_compensation", data_to_allocate);
-                new_compenstation = &p.get_or_create(compensation_prim);
-                p.get_inputs().push_back(new_compenstation);
-                compensation.push_back(new_compenstation->id());
+        } else if (w_dt == data_types::u8 && azp_dt == data_types::i8) {
+            mem_lock<uint8_t> w_lock(w, stream);
+            mem_lock<int8_t> azp_lock(azp, stream);
+            if (wzp) {
+                mem_lock<uint8_t> wzp_lock(wzp, stream);
+                fill_compensation_typed(w_lock.data(), azp_lock.data(), wzp_lock.data(), comp_lock.data(), GS, OC, IC, KS);
+            } else {
+                fill_compensation_typed(w_lock.data(), azp_lock.data(), static_cast<uint8_t*>(nullptr), comp_lock.data(), GS, OC, IC, KS);
             }
+        }
+    };
 
-            // Collect dependencies of a new convolution node
-            std::vector<program_node*> dependencies = {new_input, new_weights};
-            if (new_bias)
-                dependencies.push_back(new_bias);
-            if (new_w_zp)
-                dependencies.push_back(new_w_zp);
-            if (new_a_zp)
-                dependencies.push_back(new_a_zp);
-            if (new_compenstation)
-                dependencies.push_back(new_compenstation);
-
-            auto new_conv_prim = std::make_shared<convolution>(
-                        convolution_node.id() + "_asymmetric",
-                        input,
-                        weights,
-                        old_conv_prim->bias,
-                        w_zero_points,
-                        a_zero_points,
-                        compensation,
-                        old_conv_prim->groups,
-                        *old_conv_prim->output_data_type,
-                        old_conv_prim->stride,
-                        old_conv_prim->input_offset,
-                        old_conv_prim->dilation,
-                        output_size,
-                        old_conv_prim->grouped_weights_shape,
-                        old_conv_prim->output_padding);
-
-            auto& new_conv_node = p.get_or_create(new_conv_prim);
-
-            // Replace old conv node with the new one. New node has correct users, but dependencies don't match primitive parameters,
-            // so replace it with the vector collected earlier
-            p.replace(convolution_node, new_conv_node);
-            if (need_compensation) {
-                p.get_processing_order().insert(&new_conv_node, new_compenstation);
-                new_compenstation->users.push_back(&new_conv_node);
-            }
-            new_conv_node.dependencies = dependencies;
+    auto& in0 = convolution_node.get_dependency(0);
+    auto& in1 = convolution_node.get_dependency(1);
 
-            // Remove sub operations from the graph and set correct users for zero points and inputs
-            if (asymmetric_data) {
-                if (!new_a_zp || !new_input)
-                    CLDNN_ERROR_MESSAGE(new_conv_node.id(), "Unexpected nullptr in asymmetric quantization for activations optimization");
+    bool asymmetric_data = in0.is_type<eltwise>() && is_zero_point_node(in0.as<eltwise>());
+    bool asymmetric_weights = in1.is_type<eltwise>() && is_zero_point_node(in1.as<eltwise>());
 
-                auto& zp_users = new_a_zp->users;
-                auto& in_users = new_input->users;
-                // Erase sub node from input and zero point users...
-                zp_users.erase(std::remove(zp_users.begin(), zp_users.end(), &in0), zp_users.end());
-                in_users.erase(std::remove(in_users.begin(), in_users.end(), &in0), in_users.end());
+    if (!asymmetric_data && !asymmetric_weights)
+        return;
 
-                // ...because now the user is new convolution node
-                new_a_zp->users.push_back(&new_conv_node);
-                new_input->users.push_back(&new_conv_node);
+    // Input that doesn't match asymmetric pattern should be quantized
+    // Cases like fp32 input + i8 weights + i8 w_zp are not supported
+    if (!asymmetric_data &&
+        in0.get_output_layout().data_type != data_types::u8 &&
+        in0.get_output_layout().data_type != data_types::i8)
+        return;
 
-                p.add_optimized_primitive_info(in0.id(), {new_conv_node.id()});
+    if (!asymmetric_weights &&
+        in1.get_output_layout().data_type != data_types::u8 &&
+        in1.get_output_layout().data_type != data_types::i8)
+        return;
 
-                // Remove sub node on activations
-                in0.dependencies.clear();
-                in0.users.clear();
-                p.remove_if_dangling(in0);
-            }
+    auto old_conv_prim = convolution_node.get_primitive();
 
-            if (asymmetric_weights) {
-                if (!new_w_zp || !new_weights)
-                    CLDNN_ERROR_MESSAGE(new_conv_node.id(), "Unexpected nullptr in asymmetric quantization for weights optimization");
+    // Split is not supported
+    if (old_conv_prim->weights.size() > 1)
+        return;
 
-                auto& zp_users = new_w_zp->users;
-                auto& wei_users = new_weights->users;
-                // Erase sub node from weights and zero point users...
-                zp_users.erase(std::remove(zp_users.begin(), zp_users.end(), &in1), zp_users.end());
-                wei_users.erase(std::remove(wei_users.begin(), wei_users.end(), &in1), wei_users.end());
 
-                // ...because now the user is new convolution node
-                new_weights->users.push_back(&new_conv_node);
-                new_w_zp->users.push_back(&new_conv_node);
+    primitive_id input = old_conv_prim->input[0];
+    std::vector<primitive_id> a_zero_points = {};
 
-                p.add_optimized_primitive_info(in1.id(), {new_conv_node.id()});
+    cldnn::program_node* new_input = &in0;
+    cldnn::program_node* new_a_zp = nullptr;
+    cldnn::program_node* new_w_zp = nullptr;
 
-                // Remove sub node on weights
-                in1.dependencies.clear();
-                in1.users.clear();
-                p.remove_if_dangling(in1);
-            }
+    bool need_compensation = false;
 
-            new_conv_node.recalc_output_layout();
-        };
+    auto output_size = convolution_node.get_output_layout().size;
+    int ofm = in1.get_output_layout().size.batch[0];
+    int ifm = in0.get_output_layout().size.feature[0];
+    int ofm_aligned = ((ofm + 31) / 32) * 32;
+    int ifm_aligned = ((ifm + 31) / 32) * 32;
 
-        program_helpers::do_for_types<convolution>(*node,
-                                                    asymmetric_convolution_f);
-    }
-}
+    if (asymmetric_data) {
+        new_input = &in0.get_dependency(0);
+        new_a_zp = &in0.get_dependency(1);
 
-void prepare_quantization::prepare_dequantize_merge(program_impl &p) {
-    auto itr = p.get_processing_order().begin();
-    while (itr != p.get_processing_order().end()) {
-        auto &node = (*itr++);
+        auto l = layout{new_a_zp->get_output_layout().data_type, format::bfyx, tensor{1, ifm_aligned, 1, 1}};
+        int s = new_a_zp->get_output_layout().size.feature[0];
+        auto azp_aligned = p.get_engine().allocate_memory(l);
+        auto old_ptr = new_a_zp->as<data>().get_attached_memory_ptr();
+        mem_lock<int8_t> new_data{azp_aligned, stream};
+        mem_lock<int8_t> old_data{old_ptr, stream};
+        for (int i = 0; i < ifm_aligned; i++) {
+            new_data.data()[i] = old_data.data()[i % s];
+        }
+        new_a_zp->as<data>().attach_memory(azp_aligned);
 
-        if (node->is_output())
-            continue;
+        input = new_input->id();
+        a_zero_points.push_back(new_a_zp->id());
+        need_compensation = true;
+    }
 
-        program_helpers::do_for_types<eltwise>(*node, [&p](eltwise_node& node) {
-            for (size_t i = 1; i < node.get_dependencies().size(); i++) {
-                if (!node.get_dependency(i).is_type<data>()) {
-                    return;
-                }
-            }
+    std::vector<primitive_id> w_zero_points = {};
+    std::vector<primitive_id> weights = old_conv_prim->weights;
+    cldnn::program_node* new_weights = &in1;
+    if (asymmetric_weights) {
+        new_weights = &in1.get_dependency(0);
+        new_w_zp = &in1.get_dependency(1);
+
+        auto l = layout{new_w_zp->get_output_layout().data_type, format::bfyx, tensor{ofm_aligned, 1, 1, 1}};
+        int s = new_w_zp->get_output_layout().size.batch[0];
+        auto wzp_aligned = p.get_engine().allocate_memory(l);
+        auto old_ptr = new_w_zp->as<data>().get_attached_memory_ptr();
+        mem_lock<int8_t> new_data{wzp_aligned, stream};
+        mem_lock<int8_t> old_data{old_ptr, stream};
+        for (int i = 0; i < ofm_aligned; i++) {
+            new_data.data()[i] = old_data.data()[i % s];
+        }
+        new_w_zp->as<data>().attach_memory(wzp_aligned);
 
-            auto get_scale_shift_mem = [](const eltwise_node& eltw, size_t dep_id) -> memory::ptr {
-                if (dep_id >= eltw.get_dependencies().size())
-                    CLDNN_ERROR_MESSAGE(eltw.id(), "Invalid dependency id in dequantize optimization");
+        weights = { new_weights->id() };
+        w_zero_points.push_back(new_w_zp->id());
+    }
 
-                return eltw.get_dependency(dep_id).as<data>().get_attached_memory_ptr();
-            };
+    std::vector<primitive_id> compensation = {};
+    cldnn::program_node* new_compenstation = nullptr;
+    if (need_compensation) {
+        auto l = layout{data_types::f32, format::bfyx, tensor{1, ofm_aligned, 1, 1}};
+        auto data_to_allocate = p.get_engine().allocate_memory(l);
+        auto w = new_weights->as<data>().get_attached_memory_ptr();
+        auto azp = asymmetric_data ? new_a_zp->as<data>().get_attached_memory_ptr() : nullptr;
+        auto wzp = asymmetric_weights ? new_w_zp->as<data>().get_attached_memory_ptr() : nullptr;
+        int groups = static_cast<int>(convolution_node.get_groups());
+        fill_compensation(groups, w, azp, wzp, data_to_allocate);
+
+        auto compensation_prim = std::make_shared<data>(convolution_node.id() + "_compensation", data_to_allocate);
+        new_compenstation = &p.get_or_create(compensation_prim);
+        p.get_inputs().push_back(new_compenstation);
+        compensation.push_back(new_compenstation->id());
+    }
 
-            auto eltw_mode = node.get_primitive()->mode;
-            if (eltw_mode != eltwise_mode::sum && eltw_mode != eltwise_mode::prod)
-                return;
+    // Collect dependencies of a new convolution node
+    std::vector<program_node*> dependencies = {new_input, new_weights};
+    cldnn::program_node* new_bias = !old_conv_prim->bias.empty() ? &convolution_node.get_dependency(2) : nullptr;
+    if (new_bias)
+        dependencies.push_back(new_bias);
+    if (new_w_zp)
+        dependencies.push_back(new_w_zp);
+    if (new_a_zp)
+        dependencies.push_back(new_a_zp);
+    if (new_compenstation)
+        dependencies.push_back(new_compenstation);
+
+    auto new_conv_prim = std::make_shared<convolution>(
+                convolution_node.id() + "_asymmetric",
+                input,
+                weights,
+                old_conv_prim->bias,
+                w_zero_points,
+                a_zero_points,
+                compensation,
+                old_conv_prim->groups,
+                *old_conv_prim->output_data_type,
+                old_conv_prim->stride,
+                old_conv_prim->input_offset,
+                old_conv_prim->dilation,
+                output_size,
+                old_conv_prim->grouped_weights_shape,
+                old_conv_prim->output_padding);
+
+    auto& new_conv_node = p.get_or_create(new_conv_prim);
+
+    // Replace old conv node with the new one. New node has correct users, but dependencies don't match primitive parameters,
+    // so replace it with the vector collected earlier
+    p.replace(convolution_node, new_conv_node);
+    if (need_compensation) {
+        p.get_processing_order().insert(&new_conv_node, new_compenstation);
+        new_compenstation->users.push_back(&new_conv_node);
+    }
+    new_conv_node.dependencies = dependencies;
 
-            auto& input = node.input();
-            auto& stream = p.get_stream();
+    // Remove sub operations from the graph and set correct users for zero points and inputs
+    if (asymmetric_data) {
+        if (!new_a_zp || !new_input)
+            CLDNN_ERROR_MESSAGE(new_conv_node.id(), "Unexpected nullptr in asymmetric quantization for activations optimization");
 
-            for (auto& user : input.get_users()) {
-                if (user == &node)
-                    continue;
+        auto& zp_users = new_a_zp->users;
+        auto& in_users = new_input->users;
+        // Erase sub node from input and zero point users...
+        zp_users.erase(std::remove(zp_users.begin(), zp_users.end(), &in0), zp_users.end());
+        in_users.erase(std::remove(in_users.begin(), in_users.end(), &in0), in_users.end());
 
-                if (!user->is_type<eltwise>() || user->get_dependencies().size() != node.get_dependencies().size())
-                    continue;
+        // ...because now the user is new convolution node
+        new_a_zp->users.push_back(&new_conv_node);
+        new_input->users.push_back(&new_conv_node);
 
-                auto& eltwise_dep = user->as<eltwise>();
-                if (eltwise_dep.get_primitive()->mode != node.get_primitive()->mode)
-                    continue;
+        p.add_optimized_primitive_info(in0.id(), {new_conv_node.id()});
 
-                bool valid_scale_node = true;
-                for (size_t i = 1; i < eltwise_dep.get_dependencies().size(); i++) {
-                    if (!eltwise_dep.get_dependency(i).is_type<data>()) {
-                        valid_scale_node = false;
-                    }
-                }
+        // Remove sub node on activations
+        in0.dependencies.clear();
+        in0.users.clear();
+        p.remove_if_dangling(in0);
+    }
 
-                if (!valid_scale_node)
-                    continue;
+    if (asymmetric_weights) {
+        if (!new_w_zp || !new_weights)
+            CLDNN_ERROR_MESSAGE(new_conv_node.id(), "Unexpected nullptr in asymmetric quantization for weights optimization");
 
-                bool same_params = true;
-                for (size_t i = 1; i < node.get_dependencies().size(); i++) {
-                    auto mem0 = get_scale_shift_mem(eltwise_dep, i);
-                    auto mem1 = get_scale_shift_mem(node, i);
+        auto& zp_users = new_w_zp->users;
+        auto& wei_users = new_weights->users;
+        // Erase sub node from weights and zero point users...
+        zp_users.erase(std::remove(zp_users.begin(), zp_users.end(), &in1), zp_users.end());
+        wei_users.erase(std::remove(wei_users.begin(), wei_users.end(), &in1), wei_users.end());
 
-                    mem_lock<uint8_t> mem0_lock{mem0, stream};
-                    mem_lock<uint8_t> mem1_lock{mem1, stream};
-                    auto ptr0 = mem0_lock.data();
-                    auto ptr1 = mem1_lock.data();
+        // ...because now the user is new convolution node
+        new_weights->users.push_back(&new_conv_node);
+        new_w_zp->users.push_back(&new_conv_node);
 
-                    for (size_t j = 0; j < mem0->get_layout().bytes_count(); j++) {
-                        if (ptr0[j] != ptr1[j]) {
-                            same_params = false;
-                            break;
-                        }
-                    }
-                }
+        p.add_optimized_primitive_info(in1.id(), {new_conv_node.id()});
 
-                if (same_params) {
-                    while (!node.get_dependencies().empty()) {
-                        auto& dep = node.get_dependency(0);
-                        p.remove_connection(dep, node);
-                        p.remove_if_dangling(dep);
-                    }
-                    p.add_optimized_primitive_info(node.id(), {user->id()});
-                    p.replace_all_usages(node, *user);
-                }
-            }
-        });
+        // Remove sub node on weights
+        in1.dependencies.clear();
+        in1.users.clear();
+        p.remove_if_dangling(in1);
     }
+
+    new_conv_node.recalc_output_layout();
 }
 
 void prepare_quantization::run(program_impl& p) {
-    prepare_packed_quantize(p);
-    prepare_scale_shift_opt(p);
-    prepare_dequantize_merge(p);
-    remove_fake_reorders(p);
-    prepare_asymmetric_quantization(p);
+    auto itr = p.get_processing_order().begin();
+    while (itr != p.get_processing_order().end()) {
+        auto &node = (*itr++);
+        if (node->is_type<quantize>()) {
+            handle_quantize_node(p, node->as<quantize>());
+        } else if (node->is_type<eltwise>()) {
+            prepare_dequantize_merge(p, node->as<eltwise>());
+        } else if (node->is_type<reorder>()) {
+            remove_fake_reorders(p, node->as<reorder>());
+        } else if (node->is_type<convolution>()) {
+            prepare_asymmetric_quantization(p, node->as<convolution>());
+        }
+    }
 }
diff --git a/inference-engine/thirdparty/clDNN/src/graph_optimizer/reverse_optional_nodes_outputs.cpp b/inference-engine/thirdparty/clDNN/src/graph_optimizer/reverse_optional_nodes_outputs.cpp
index 7b39a83f7b598e..fc26305a75267b 100644
--- a/inference-engine/thirdparty/clDNN/src/graph_optimizer/reverse_optional_nodes_outputs.cpp
+++ b/inference-engine/thirdparty/clDNN/src/graph_optimizer/reverse_optional_nodes_outputs.cpp
@@ -17,9 +17,7 @@ using namespace cldnn;
     the processing order was valid).
 */
 void reverse_optional_nodes_outputs::run(program_impl& p) {
-    auto node_itr = p.get_processing_order().begin();
-    while (node_itr != p.get_processing_order().end()) {
-        auto& node = (*node_itr++);
+    for (auto& node : p.get_processing_order()) {
         if (node->is_type<lstm_dynamic_timeloop>()) {
             auto& typed_node = node->as<lstm_dynamic_timeloop>();
             typed_node.reverse_optional_outputs_connections();
diff --git a/inference-engine/thirdparty/clDNN/src/include/pass_manager.h b/inference-engine/thirdparty/clDNN/src/include/pass_manager.h
index 0047c7bec52347..f726e1ca1efbe3 100644
--- a/inference-engine/thirdparty/clDNN/src/include/pass_manager.h
+++ b/inference-engine/thirdparty/clDNN/src/include/pass_manager.h
@@ -9,12 +9,16 @@
 #include "split_inst.h"
 #include "lstm_inst.h"
 #include "lstm_dynamic_inst.h"
+#include "quantize_inst.h"
+#include "eltwise_inst.h"
+#include "convolution_inst.h"
 #include <string>
 #include <vector>
 #include <memory>
 #include <list>
 #include <utility>
 #include <set>
+#include <functional>
 
 #include <fstream>
 
@@ -149,11 +153,12 @@ class prepare_quantization : public base_pass {
 
 private:
     void run(program_impl& p) override;
-    void prepare_packed_quantize(program_impl& p);
-    void prepare_scale_shift_opt(program_impl& p);
-    void prepare_dequantize_merge(program_impl& p);
-    void remove_fake_reorders(program_impl& p);
-    void prepare_asymmetric_quantization(program_impl& p);
+    void handle_quantize_node(program_impl& p, quantize_node& quantize_node);
+    void prepare_packed_quantize(program_impl& p, quantize_node& quantize_node);
+    void prepare_dequantize_merge(program_impl& p, eltwise_node& eltwise_node);
+    void remove_fake_reorders(program_impl& p, reorder_node& reorder_node);
+    void prepare_asymmetric_quantization(program_impl& p, convolution_node& convolution_node);
+    void prepare_scale_shift_opt(program_impl &p, quantize_node& quantize_node);
 };
 
 class prepare_conv_eltw_fusing : public base_pass {

From c483cdced6c3a7c11a6949822f24de7d2cfffdfe Mon Sep 17 00:00:00 2001
From: LuweiZhou <luwei.zhou@intel.com>
Date: Wed, 23 Jun 2021 15:50:52 +0800
Subject: [PATCH 172/178] Revise swish (#5983)

* Update Swish OP description.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>

* Use RTTI to declare/define NGraph Swish OP.
Add input element type check when constructing Swish OP.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>

* Add Swish into activation serialization test list.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>

* Add Swish into IE CPU plugin activation single layer test suit.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>

* Add Swish NGraph backend and visitor API tests.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>

* Add Swish unsupported parameter data type test cases.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>

* Update the Swish OP visistor API to use typed test.

Signed-off-by: Luwei Zhou <luwei.zhou@intel.com>
---
 docs/ops/activation/Swish_4.md                |  3 +-
 .../serialization/single_layer/activation.cpp |  1 +
 .../single_layer_tests/activation.cpp         |  3 +-
 ngraph/core/include/ngraph/op/swish.hpp       |  3 +-
 ngraph/core/src/op/swish.cpp                  |  8 +-
 ngraph/test/CMakeLists.txt                    |  2 +
 ngraph/test/backend/swish.in.cpp              | 75 +++++++++++++++++++
 ngraph/test/type_prop/swish.cpp               | 14 ++++
 ngraph/test/visitors/op/swish.cpp             | 11 +++
 9 files changed, 115 insertions(+), 5 deletions(-)
 create mode 100644 ngraph/test/backend/swish.in.cpp
 create mode 100644 ngraph/test/visitors/op/swish.cpp

diff --git a/docs/ops/activation/Swish_4.md b/docs/ops/activation/Swish_4.md
index 1a8b7d1b51a4f9..04f17390a16783 100644
--- a/docs/ops/activation/Swish_4.md
+++ b/docs/ops/activation/Swish_4.md
@@ -9,7 +9,8 @@
 **Detailed description**
 
 *Swish* operation is introduced in this [article](https://arxiv.org/abs/1710.05941).
-It performs element-wise activation function on a given input tensor, based on the following mathematical formula:
+
+*Swish* is a smooth, non-monotonic function. The non-monotonicity property of *Swish* distinguishes itself from most common activation functions. It performs element-wise activation function on a given input tensor,  based on the following mathematical formula:
 
 \f[
 Swish(x) = x\cdot \sigma(\beta x) = x \left(1 + e^{-(\beta x)}\right)^{-1}
diff --git a/inference-engine/tests/functional/inference_engine/serialization/single_layer/activation.cpp b/inference-engine/tests/functional/inference_engine/serialization/single_layer/activation.cpp
index fb0c8acd71b806..371808cbbc5246 100644
--- a/inference-engine/tests/functional/inference_engine/serialization/single_layer/activation.cpp
+++ b/inference-engine/tests/functional/inference_engine/serialization/single_layer/activation.cpp
@@ -52,6 +52,7 @@ const std::map<ActivationTypes, std::vector<std::vector<float>>> activationTypes
         {Ceiling,               {}},
         {Mish,                  {}},
         {HSwish,                {}},
+        {Swish,                 {{0.3f}}},
         {SoftPlus,              {}},
         {HSigmoid,              {}},
         {RoundHalfToEven,       {}},
diff --git a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
index 87f986f980fffc..1ad6c8912f70a5 100644
--- a/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/activation.cpp
@@ -59,7 +59,8 @@ const std::map<ActivationTypes, std::vector<std::vector<float>>> activationTypes
         {RoundHalfAwayFromZero, {}},
         {Erf,                   {}},
         {GeluErf,               {}},
-        {GeluTanh,              {}}
+        {GeluTanh,              {}},
+        {Swish,                 {{0.4f}}}
 };
 
 // List of operations that should be tested also with integer precision
diff --git a/ngraph/core/include/ngraph/op/swish.hpp b/ngraph/core/include/ngraph/op/swish.hpp
index a9fed2785e66a9..144b99846ee993 100644
--- a/ngraph/core/include/ngraph/op/swish.hpp
+++ b/ngraph/core/include/ngraph/op/swish.hpp
@@ -20,8 +20,7 @@ namespace ngraph
             class NGRAPH_API Swish : public ngraph::op::Op
             {
             public:
-                static constexpr NodeTypeInfo type_info{"Swish", 4};
-                const NodeTypeInfo& get_type_info() const override { return type_info; }
+                NGRAPH_RTTI_DECLARATION;
                 Swish() = default;
 
                 /// \brief Constructs an Swish operation.
diff --git a/ngraph/core/src/op/swish.cpp b/ngraph/core/src/op/swish.cpp
index 5b3b7175950d7a..9f5a4a0c8e3d28 100644
--- a/ngraph/core/src/op/swish.cpp
+++ b/ngraph/core/src/op/swish.cpp
@@ -14,7 +14,7 @@
 using namespace std;
 using namespace ngraph;
 
-constexpr NodeTypeInfo op::v4::Swish::type_info;
+NGRAPH_RTTI_DEFINITION(op::v4::Swish, "Swish", 4);
 
 op::v4::Swish::Swish(const Output<Node>& arg)
     : Op({arg})
@@ -43,6 +43,12 @@ void op::v4::Swish::validate_and_infer_types()
                           "Swish must have 1 or 2 inputs, but it has: ",
                           inputs_count);
 
+    NODE_VALIDATION_CHECK(this,
+                          get_input_element_type(0).is_real(),
+                          "Swish input tensor must be floating point type(",
+                          get_input_element_type(0),
+                          ").");
+
     if (inputs_count == 2)
     {
         NODE_VALIDATION_CHECK(this,
diff --git a/ngraph/test/CMakeLists.txt b/ngraph/test/CMakeLists.txt
index f6d12b6483d4f6..df92b547132fa3 100644
--- a/ngraph/test/CMakeLists.txt
+++ b/ngraph/test/CMakeLists.txt
@@ -300,6 +300,7 @@ set(SRC
     visitors/op/squeeze.cpp
     visitors/op/sqrt.cpp
     visitors/op/strided_slice.cpp
+    visitors/op/swish.cpp
     visitors/op/tanh.cpp
     visitors/op/topk.cpp
     visitors/op/transpose.cpp
@@ -476,6 +477,7 @@ set(MULTI_TEST_SRC
     backend/squared_difference.in.cpp
     backend/squeeze.in.cpp
     backend/subtract.in.cpp
+    backend/swish.in.cpp
     backend/tan.in.cpp
     backend/tanh.in.cpp
     backend/tile.in.cpp
diff --git a/ngraph/test/backend/swish.in.cpp b/ngraph/test/backend/swish.in.cpp
new file mode 100644
index 00000000000000..549b6678b50d63
--- /dev/null
+++ b/ngraph/test/backend/swish.in.cpp
@@ -0,0 +1,75 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#include "gtest/gtest.h"
+#include "ngraph/ngraph.hpp"
+#include "util/engine/test_engines.hpp"
+#include "util/test_case.hpp"
+#include "util/test_control.hpp"
+
+using namespace std;
+using namespace ngraph;
+
+static string s_manifest = "${MANIFEST}";
+using TestEngine = test::ENGINE_CLASS_NAME(${BACKEND_NAME});
+
+NGRAPH_TEST(${BACKEND_NAME}, swish_2D_with_beta0_6)
+{
+    Shape in_shape{2, 4};
+    element::Type et = element::f32;
+    auto beta = 0.6f;
+
+    auto args0 = make_shared<op::Parameter>(et, in_shape);
+    auto args1 = make_shared<op::Parameter>(et, Shape{});
+    auto swish = make_shared<op::v4::Swish>(args0, args1);
+    auto f = make_shared<Function>(swish, ParameterVector{args0, args1});
+
+    vector<vector<float>> in_vec{vector<float>{0.4, -5.7, -6, 3, -0.9, 23, 5, 3.3} , vector<float>{beta}};
+    vector<float> out_vec{in_vec[0]};
+    std::transform(out_vec.begin(), out_vec.end(), out_vec.begin(), [&beta](float x) -> float { return (x / (1.0f + std::exp(x * beta * -1.0f)));});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>(in_vec);
+    test_case.add_expected_output<float>(in_shape, out_vec);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, swish_2D_without_beta)
+{
+    Shape in_shape{2, 3};
+    element::Type et = element::f32;
+
+    auto args0 = make_shared<op::Parameter>(et, in_shape);
+    auto swish = make_shared<op::v4::Swish>(args0);
+    auto f = make_shared<Function>(swish, ParameterVector{args0});
+
+    vector<float> in_vec{1, 8, -8, 17, -0.5, -1};
+    vector<float> out_vec{in_vec};
+    std::transform(out_vec.begin(), out_vec.end(), out_vec.begin(), [](float x) -> float { return (x / (1.0f + std::exp(x * -1.0f)));});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_input<float>(in_vec);
+    test_case.add_expected_output<float>(in_shape, out_vec);
+    test_case.run();
+}
+
+NGRAPH_TEST(${BACKEND_NAME}, swish_4D_with_beta0_33)
+{
+    Shape in_shape{2, 2, 1, 2};
+    element::Type et = element::f32;
+    auto beta = 0.33f;
+
+    auto args0 = make_shared<op::Parameter>(et, in_shape);
+    auto args1 = make_shared<op::Parameter>(et, Shape{});
+    auto swish = make_shared<op::v4::Swish>(args0, args1);
+    auto f = make_shared<Function>(swish, ParameterVector{args0, args1});
+
+    vector<vector<float>> in_vec{vector<float>{0.1, 0.6, 20, -7, -5.3, 3.5, -9, 11} , vector<float>{beta}};
+    vector<float> out_vec{in_vec[0]};
+    std::transform(out_vec.begin(), out_vec.end(), out_vec.begin(), [&beta](float x) -> float { return (x / (1.0f + std::exp(x * beta * -1.0f)));});
+
+    auto test_case = test::TestCase<TestEngine>(f);
+    test_case.add_multiple_inputs<float>(in_vec);
+    test_case.add_expected_output<float>(in_shape, out_vec);
+    test_case.run();
+}
diff --git a/ngraph/test/type_prop/swish.cpp b/ngraph/test/type_prop/swish.cpp
index bdb45ed1292a36..324d8f46a59cc7 100644
--- a/ngraph/test/type_prop/swish.cpp
+++ b/ngraph/test/type_prop/swish.cpp
@@ -81,3 +81,17 @@ TEST(type_prop, swish_2_inputs)
     ASSERT_TRUE(swish_func->get_output_partial_shape(0).same_scheme(data->get_output_shape(0)));
     ASSERT_TRUE(swish_func->get_output_partial_shape(0).rank().is_static());
 }
+
+TEST(type_prop, swish_incompatible_type_boolean)
+{
+    auto data = make_shared<op::Parameter>(element::boolean, Shape{1, 3, 6});
+    auto beta = make_shared<op::Parameter>(element::f32, Shape{});
+    ASSERT_THROW(make_shared<op::v4::Swish>(data, beta);, ngraph::NodeValidationFailure);
+}
+
+TEST(type_prop, swish_incompatible_types_u32)
+{
+    auto data = make_shared<op::Parameter>(element::f32, Shape{1, 3, 6});
+    auto beta = make_shared<op::Parameter>(element::u32, Shape{});
+    ASSERT_THROW(make_shared<op::v4::Swish>(data, beta);, ngraph::NodeValidationFailure);
+}
diff --git a/ngraph/test/visitors/op/swish.cpp b/ngraph/test/visitors/op/swish.cpp
new file mode 100644
index 00000000000000..0cb39f353cddaf
--- /dev/null
+++ b/ngraph/test/visitors/op/swish.cpp
@@ -0,0 +1,11 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#include "unary_ops.hpp"
+
+using Type = ::testing::Types<UnaryOperatorType<ngraph::op::v4::Swish, element::f32>>;
+
+INSTANTIATE_TYPED_TEST_CASE_P(visitor_without_atrribute,
+                              UnaryOperatorVisitor,
+                              Type,
+                              UnaryOperatorTypeName);

From afe033a39aef75daa3b8071d084c1cf719153735 Mon Sep 17 00:00:00 2001
From: Ilya Lavrenov <ilya.lavrenov@intel.com>
Date: Wed, 23 Jun 2021 10:51:32 +0300
Subject: [PATCH 173/178] Added debug suffix for ngraph dependencies (#6303)

---
 ngraph/CMakeLists.txt | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/ngraph/CMakeLists.txt b/ngraph/CMakeLists.txt
index 55e218daa2c924..4c598cf7b84897 100644
--- a/ngraph/CMakeLists.txt
+++ b/ngraph/CMakeLists.txt
@@ -129,6 +129,8 @@ set(NGRAPH_FORWARD_CMAKE_ARGS
     -DCMAKE_CXX_EXTENSIONS:BOOL=${CMAKE_CXX_EXTENSIONS}
     -DCMAKE_EXPORT_COMPILE_COMMANDS:BOOL=${CMAKE_EXPORT_COMPILE_COMMANDS}
     -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=${CMAKE_POSITION_INDEPENDENT_CODE}
+    -DCMAKE_DEBUG_POSTFIX=${CMAKE_DEBUG_POSTFIX}
+    -DCMAKE_RELEASE_POSTFIX=${CMAKE_RELEASE_POSTFIX}
     )
 
 if(CMAKE_TOOLCHAIN_FILE)

From 3307b1f17ce2dbb8206f7dfc0012100b7c2fab44 Mon Sep 17 00:00:00 2001
From: Nadezhda Ageeva <nadezhda.ageeva@intel.com>
Date: Wed, 23 Jun 2021 11:20:20 +0300
Subject: [PATCH 174/178] [GNA]: Cleanup policy (#6216)

---
 .../src/gna_plugin/gna_graph_compiler.cpp     |  12 +-
 .../src/gna_plugin/gna_graph_compiler.hpp     |   3 -
 .../src/gna_plugin/gna_plugin.cpp             |  13 +-
 .../src/gna_plugin/gna_plugin.hpp             |   8 -
 .../src/gna_plugin/gna_plugin_policy.hpp      |  96 -----------
 .../gna_plugin/optimizer/gna_pass_manager.cpp | 155 +++---------------
 .../gna_plugin/optimizer/gna_pass_manager.hpp |  17 --
 .../unit/engines/gna/gna_matcher.cpp          |   5 +-
 .../unit/engines/gna/gna_matcher.hpp          |   6 -
 .../gna/layers/gna_align_filter2_tests.cpp    |  53 ++----
 10 files changed, 37 insertions(+), 331 deletions(-)
 delete mode 100644 inference-engine/src/gna_plugin/gna_plugin_policy.hpp

diff --git a/inference-engine/src/gna_plugin/gna_graph_compiler.cpp b/inference-engine/src/gna_plugin/gna_graph_compiler.cpp
index bf44e437af0e6a..e48595d6c272f3 100644
--- a/inference-engine/src/gna_plugin/gna_graph_compiler.cpp
+++ b/inference-engine/src/gna_plugin/gna_graph_compiler.cpp
@@ -34,7 +34,6 @@
 #include "layers/gna_crop_layer.hpp"
 #include "layers/gna_fake_quantize_layer.hpp"
 #include "round_float_define.hpp"
-#include "gna_plugin_policy.hpp"
 #include "gna_groups.hpp"
 #include "backend/gna_limitations.hpp"
 
@@ -62,10 +61,6 @@ void GNAGraphCompiler::setGNAFlagsPtr(std::shared_ptr<GNAPluginNS::GNAFlags> gna
     this->gnaFlags = std::move(gnaFlagsPtr);
 }
 
-void GNAGraphCompiler::setPolicy(GNAPluginNS::Policy policyToSet) {
-    this->policy = policyToSet;
-}
-
 intel_dnn_component_t * GNAGraphCompiler::find_first_unused_input(InferenceEngine::CNNLayerPtr current) {
     if (current->insData.empty())
         return nullptr;
@@ -547,10 +542,7 @@ void GNAGraphCompiler::finalizeConvolution2DPrimitive(InferenceEngine::CNNLayerP
     auto effectiveInputWidth = in_width;
     auto effectiveInputHeight = in_height;
 
-    if (policy.cnn2dInputPaddingSupported) {
-        effectiveInputWidth += convolution._padding_x * 2;
-        effectiveInputHeight += convolution._padding_y * 2;
-    } else if (convolution._padding_x != 0 || convolution._padding_y != 0 ||
+    if (convolution._padding_x != 0 || convolution._padding_y != 0 ||
         convolution._pads_end.at(X_AXIS) != 0 || convolution._pads_end.at(Y_AXIS) != 0) {
         THROW_GNA_LAYER_EXCEPTION(layer) << "Convolution's input padding is not supported";
     }
@@ -1641,7 +1633,7 @@ void GNAGraphCompiler::ConcatAlignFilterPrimitive(InferenceEngine::CNNLayerPtr l
     uint32_t num_rows_copied = 0;
     // in case of left alignment succeed, but due to number of elements not multiple of 8 we need to insert align_filter
     // we are improving it by inserting copy layer of size that covers most of elements - remained max of 32x31 affine filter
-    if (policy.ConcatAlignmentPolicy == Policy::ConcatAlignment::FAST &&  0 == numRowsPadded && ALIGN(num_rows_in, 32) > 32) {
+    if (0 == numRowsPadded && ALIGN(num_rows_in, 32) > 32) {
         // can we use copy at all
         num_rows_copied = ALIGN(num_rows_in, 32) - 32;
 
diff --git a/inference-engine/src/gna_plugin/gna_graph_compiler.hpp b/inference-engine/src/gna_plugin/gna_graph_compiler.hpp
index 148c34a82ff5a7..d761d91739269c 100644
--- a/inference-engine/src/gna_plugin/gna_graph_compiler.hpp
+++ b/inference-engine/src/gna_plugin/gna_graph_compiler.hpp
@@ -27,7 +27,6 @@
 #include "backend/gna_limitations.hpp"
 #include "gna_device.hpp"
 #include "gna_data_types.hpp"
-#include "gna_plugin_policy.hpp"
 
 namespace GNAPluginNS {
 class GNAGraphCompiler {
@@ -36,7 +35,6 @@ class GNAGraphCompiler {
     std::shared_ptr<GNAPluginNS::gna_memory_type> gnamem;
     std::shared_ptr<GNAPluginNS::InputDesc> inputDesc;
     std::shared_ptr<GNAPluginNS::GNAFlags> gnaFlags;
-    Policy policy;
 
     // layers with extra storage for connections and additional
     // non trivial processing
@@ -64,7 +62,6 @@ class GNAGraphCompiler {
     void setDNNPtr(std::shared_ptr<GNAPluginNS::backend::AMIntelDNN> dnnPtr);
     void setInputDescPtr(std::shared_ptr<GNAPluginNS::InputDesc> inputDescPtr);
     void setGNAFlagsPtr(std::shared_ptr<GNAPluginNS::GNAFlags> gnaFlagsPtr);
-    void setPolicy(GNAPluginNS::Policy policy);
 
     void fillMemoryConnections(std::unordered_map<std::string,
             std::vector<InferenceEngine::CNNLayerPtr>> &memoryPairs);
diff --git a/inference-engine/src/gna_plugin/gna_plugin.cpp b/inference-engine/src/gna_plugin/gna_plugin.cpp
index 01ba39a638a0b4..d6944f0c621fd4 100644
--- a/inference-engine/src/gna_plugin/gna_plugin.cpp
+++ b/inference-engine/src/gna_plugin/gna_plugin.cpp
@@ -738,7 +738,7 @@ void GNAPlugin::LoadNetwork(CNNNetwork & _network) {
     // network optimisation phases
     int passIdx = 0;
     auto run_passes = [&] (const CNNNetwork& network, bool runBeforeCopy, bool lowPrecision) {
-        auto passes = make_shared<PassManager>(PassManagerSettings{policy, runBeforeCopy, lowPrecision}, network);
+        auto passes = make_shared<PassManager>(PassManagerSettings{runBeforeCopy, lowPrecision}, network);
         passes->registerPass<RemoveConstPass>();
         passes->registerPass<UnrollTIPass>();
         passes->registerPass<RemoveConstPass>();
@@ -765,13 +765,7 @@ void GNAPlugin::LoadNetwork(CNNNetwork & _network) {
         passes->registerPass<FlattenTrivialConcatPass>();
         passes->registerPass<InsertConcatAligningFilterPass>();
         passes->registerPass<ReorderConcatInputsPass>();
-        if (policy.PermutePolicy != Policy::Permute::DISABLED) {
-            passes->registerPass<ReversePermutationsPass>();
-        }
-        if (policy.NHWCToNCHWPolicy != Policy::NHWCToNCHW::DISABLED) {
-            passes->registerPass<RemovePermutationsNHWCToNCHWPass>();
-        }
-
+        passes->registerPass<RemovePermutationsNHWCToNCHWPass>();
         passes->registerPass<InsertIdentityLayerPass>();
         passes->registerPass<BreakFusingOfOutputLayersPass>();
         passes->registerPass<InsertDiagonalLayerPass>();
@@ -840,9 +834,6 @@ void GNAPlugin::LoadNetwork(CNNNetwork & _network) {
 
     auto sortedNet = CNNNetSortTopologicallyEx(newNet, make_fuzed_order);
 
-    // passing policy to compiler
-    graphCompiler.setPolicy(policy);
-
     if (sortedNet.empty()) {
         THROW_GNA_EXCEPTION << "Sorted network is empty";
     }
diff --git a/inference-engine/src/gna_plugin/gna_plugin.hpp b/inference-engine/src/gna_plugin/gna_plugin.hpp
index 1b37439eead68e..2ce54f4f9ab292 100644
--- a/inference-engine/src/gna_plugin/gna_plugin.hpp
+++ b/inference-engine/src/gna_plugin/gna_plugin.hpp
@@ -21,7 +21,6 @@
 #include "backend/am_intel_dnn.hpp"
 #include "gna_data_types.hpp"
 #include "gna_graph_compiler.hpp"
-#include "gna_plugin_policy.hpp"
 #include "gna_plugin_log.hpp"
 #include "gna_plugin_config.hpp"
 #include <legacy/ie_util_internal.hpp>
@@ -69,8 +68,6 @@ class GNAPlugin : public InferenceEngine::IInferencePlugin {
 
     intel_dnn_number_type_t output_type = kDnnInt;
 
-    GNAPluginNS::Policy policy;
-
 #if GNA_LIB_VER == 2
     void createRequestConfigsForGnaModels();
 #endif
@@ -158,11 +155,6 @@ class GNAPlugin : public InferenceEngine::IInferencePlugin {
     INFERENCE_ENGINE_DEPRECATED("Use InferRequest::QueryState instead")
     std::vector<InferenceEngine::IVariableStateInternal::Ptr>  QueryState();
 
-     /**
-      * test-wise API
-      */
-     void SetPolicy(GNAPluginNS::Policy p) {policy = p;}
-
      /**
       * QueryMetrics API
       */
diff --git a/inference-engine/src/gna_plugin/gna_plugin_policy.hpp b/inference-engine/src/gna_plugin/gna_plugin_policy.hpp
deleted file mode 100644
index 0611ca0993c73c..00000000000000
--- a/inference-engine/src/gna_plugin/gna_plugin_policy.hpp
+++ /dev/null
@@ -1,96 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#pragma once
-
-#include <ostream>
-namespace GNAPluginNS {
-/**
- * @brief policy agregates various settings that cannot be tweak using configuration options right now,
- * and essential to keep test coverage for options both in on and off cases
- */
-class Policy {
- public:
-    /**
-    * @brief for scaleshift substitution, weight tiling simplify final graph but have extra weights overhead
-    * if not defined scaleshift broadcast will result in creating multiple diagonal layers instead of weight tiling
-    */
-    enum class ScaleShift {
-        WEIGHTS_TILING,
-        /**
-         * GNA has limited amount of batch so even existed topologies cannot be substituted with only batching,
-         * this option combines batch and weights tiling
-         */
-        BATCH_AND_WEIGHTS_TILING,
-        DIAGLAYER_TILING
-    } ScaleShiftPolicy = ScaleShift::WEIGHTS_TILING;
-
-    /**
-     * Policy on whether to substitute permute layers or not
-     */
-    enum class Permute {
-        DISABLED,
-        AUTO_PERMUTE
-    } PermutePolicy = Permute::DISABLED;
-
-    enum class FlattenTrivialConcatConversion {
-        DISABLED,
-        ENABLED
-    } ConcatConversionPolicy = FlattenTrivialConcatConversion::ENABLED;
-
-    enum class ConcatAlignment {
-        DISABLED,
-        DISABLED_FOR_FP32,
-        ENABLED,
-        FAST
-    } ConcatAlignmentPolicy = ConcatAlignment::FAST;
-
-    /**
-    * Policy to support --disable_nhwc_to_nchw option in MO
-    */
-    enum class NHWCToNCHW {
-        DISABLED,
-        REMOVE_LAST,
-        REMOVE_ALL
-    } NHWCToNCHWPolicy = NHWCToNCHW::REMOVE_ALL;
-
- /**
- * @brief trim of gna diagonal affine layer maximum elements number
- */
-    class GNAAffineDiagonal {
-    public:
-        enum : uint32_t {
-            UNLIMIT,
-            // gna limit this to be OxFFFF
-            LIMITED_TO_DEFAULT_GNA2_65536 = 65536 - 64
-        };
-        uint32_t limitedTo = LIMITED_TO_DEFAULT_GNA2_65536;
-    } GNAAffineDiagonalPolicy;
-
-    bool cnn2dInputPaddingSupported = false;
-};
-
-inline std::ostream& operator<<(std::ostream& os, Policy::ScaleShift policy) {
-    switch (policy) {
-        case Policy::ScaleShift::WEIGHTS_TILING   : os << "WEIGHTS_TILING";    break;
-        case Policy::ScaleShift::BATCH_AND_WEIGHTS_TILING: os << "BATCH_AND_WEIGHTS_TILING"; break;
-        case Policy::ScaleShift::DIAGLAYER_TILING : os << "DIAGLAYER_TILING";  break;
-        default    : os.setstate(std::ios_base::failbit);
-    }
-    return os;
-}
-
-inline std::ostream& operator<<(std::ostream& os, Policy::ConcatAlignment policy) {
-    switch (policy) {
-        case Policy::ConcatAlignment::DISABLED   : os << "DISABLED";    break;
-        case Policy::ConcatAlignment::DISABLED_FOR_FP32   : os << "DISABLED_FOR_FP32";    break;
-        case Policy::ConcatAlignment::ENABLED   : os << "ENABLED";    break;
-        case Policy::ConcatAlignment::FAST   : os << "FAST";    break;
-        default    : os.setstate(std::ios_base::failbit);
-    }
-    return os;
-}
-
-
-}  // namespace GNAPluginNS
diff --git a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
index 5b0fbcd297d400..f16645ae6cb2ad 100644
--- a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
+++ b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.cpp
@@ -2,7 +2,6 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "gna_plugin_policy.hpp"
 #include <vector>
 #include <string>
 #include <memory>
@@ -613,99 +612,6 @@ void SubstitutePReluPass::run() {
     }
 }
 
-void ReversePermutationsPass::run() {
-    OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "ReversePermutationsPass");
-    std::function<CNNLayerPtr(CNNLayerPtr, std::function<bool(CNNLayerPtr)>)> prevLayerSkipCertain
-        = [&prevLayerSkipCertain](CNNLayerPtr layer, std::function<bool(CNNLayerPtr)> shouldSkip) -> CNNLayerPtr {
-        if (CNNNetHasPrevLayer(layer.get())) {
-            return nullptr;
-        }
-        auto prev = CNNNetPrevLayer(layer);
-
-        if (!shouldSkip(prev)) return prevLayerSkipCertain(prev, shouldSkip);
-
-        return prev;
-    };
-
-    std::function<CNNLayerPtr(CNNLayerPtr)> nextLayerSkipReshape = [&nextLayerSkipReshape](CNNLayerPtr layer) -> CNNLayerPtr {
-        if (layer->outData.empty()) {
-            return nullptr;
-        }
-        if (getInputTo(layer->outData.front()).size() != 1) {
-            return nullptr;
-        }
-        auto next = getInputTo(layer->outData.front()).begin()->second;
-
-        if (LayerInfo(next).isNonFunctional()) return nextLayerSkipReshape(next);
-
-        return next;
-    };
-
-    auto prevConv = [&prevLayerSkipCertain](CNNLayerPtr layer) -> CNNLayerPtr {
-        return prevLayerSkipCertain(layer, [] (CNNLayerPtr l2) {
-            return
-                LayerInfo(l2).isNonFunctional() ||
-                LayerInfo(l2).isPooling() ||
-                LayerInfo(l2).isActivation();
-        });
-    };
-
-    std::unordered_set<std::string> affineWithPermutedWeights;
-    std::list<CNNLayerPtr> permutationstoRemove;
-
-    for (auto & l : *pLayers) {
-        if (!LayerInfo(l).isPermute()) {
-            continue;
-        }
-
-        auto layerOrder = l->GetParamAsInts("order");
-
-        if (layerOrder != std::vector<int>({0, 3, 2, 1})) {
-            THROW_GNA_EXCEPTION << "Unsupported permute layer: " << l->name << ", order: was " << l->GetParamAsString("order") <<
-                               ", but support order is 0,3,2,1";
-        }
-
-        // search for it's input convolution
-        auto prev = prevConv(l);
-
-        // pooling no used in speech models without convolution
-        if (!prev) {
-            THROW_GNA_EXCEPTION << "Unsupported permute layer: " << l->name << " no valid input to that layer";
-        }
-
-        // we can remove that permutation if it is input to ScaleShift or FC layer
-        auto next = nextLayerSkipReshape(l);
-        if (!next || !LayerInfo(next).isFullyConnected()) {
-            THROW_GNA_EXCEPTION << "Unsupported permute layer: " << l->name << " no valid output of that layer";
-        }
-
-        permutationstoRemove.push_back(l);
-
-        // removing that permutation layer and saving information about affine
-        affineWithPermutedWeights.insert(next->name);
-    }
-
-    for (auto && toRemove : permutationstoRemove) {
-        CNNNetworkRemoveLayer(toRemove);
-    }
-
-    // search for conv->affine sequences
-    for (auto & l : *pLayers) {
-        if (!LayerInfo(l).isFullyConnected() || 0 != affineWithPermutedWeights.count(l->name)) {
-            continue;
-        }
-        // found an affine layer that not involved in permutations removing
-        // searching whether it has direct input from convolution
-        auto prevConvLayer = prevConv(l);
-        if (!prevConvLayer) continue;
-
-        auto directPrev = CNNNetPrevLayer(l);
-
-        // TODO : make new permute
-        CNNNetworkInsertLayer(l, directPrev, CNNLayerPtr(nullptr));
-    }
-}
-
 void RemovePermutationsNHWCToNCHWPass::run() {
     OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "RemovePermutationsNHWCToNCHWPass");
     std::set<CNNLayerPtr> permutations_to_remove;
@@ -720,7 +626,7 @@ void RemovePermutationsNHWCToNCHWPass::run() {
 
         if (prev == nullptr || next == nullptr) continue;
 
-        if (LayerInfo(prev).isPermute() && getPassManager()->getPolicy().NHWCToNCHWPolicy == Policy::NHWCToNCHW::REMOVE_ALL) {
+        if (LayerInfo(prev).isPermute()) {
             permutations_to_remove.insert(prev);
         }
 
@@ -1040,9 +946,6 @@ void FlattenTrivialConcatPass::run() {
     // 1, 1, 5, 3 then for axis 0, 1, 2 the change will be made and inputs will be reshaped to 1, 15,
     // but for shape 2, 1, 5, 3 only axis 0 is valid and inputs will reshape to 1, 30
     auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(pLayers->front());
-    if (getPassManager()->getPolicy().ConcatConversionPolicy == Policy::FlattenTrivialConcatConversion::DISABLED) return;
-    if (getPassManager()->getPolicy().ConcatAlignmentPolicy == Policy::ConcatAlignment::DISABLED) return;
-    if (getPassManager()->getPolicy().ConcatAlignmentPolicy == Policy::ConcatAlignment::DISABLED_FOR_FP32 && !quantized) return;
 
     auto getLayerByIndex = [](int idx, ConcatLayer* concatLayer) {
         auto input = concatLayer->insData[idx];
@@ -1118,14 +1021,6 @@ void FlattenTrivialConcatPass::run() {
 void InsertConcatAligningFilterPass::run() {
     OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "InsertConcatAligningFilterPass");
     auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(pLayers->front());
-
-    if (getPassManager()->getPolicy().ConcatAlignmentPolicy == Policy::ConcatAlignment::DISABLED) {
-        return;
-    }
-    // aligning specific not required in fp32 mode
-    if (getPassManager()->getPolicy().ConcatAlignmentPolicy == Policy::ConcatAlignment::DISABLED_FOR_FP32 && !quantized) {
-        return;
-    }
     // currently concat layer only supports 2 bytes in int16 and int8 mode. In fp32 mode this no necessary but usefull for testing
     const int bytesPerConcatElement = 2;
 
@@ -1244,10 +1139,6 @@ void InsertConcatAligningFilterPass::run() {
 void ReorderConcatInputsPass::run() {
     OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "ReorderConcatInputsPass");
     auto quantized = InferenceEngine::getInjectedData<QuantizedLayerParams>(pLayers->front());
-    // aligning specific not required in fp32 mode
-    if (getPassManager()->getPolicy().ConcatAlignmentPolicy == Policy::ConcatAlignment::DISABLED_FOR_FP32 && !quantized) {
-        return;
-    }
     int numOfLinkLayers = 0;
 
     for (auto& l : *pLayers) {
@@ -1461,9 +1352,6 @@ static InferenceEngine::Blob::Ptr tileBlob(Blob::Ptr& blob, size_t TileTo) {
 
 void EltwiseSplitOverChannelsPass::run() {
     OV_ITT_SCOPED_TASK(itt::domains::GNA_LT, "EltwiseSplitOverChannelsPass");
-    if (getPassManager()->getPolicy().GNAAffineDiagonalPolicy.limitedTo == Policy::GNAAffineDiagonal::UNLIMIT) {
-        return;
-    }
 
     for (auto & l : *pLayers) {
         if (!LayerInfo(l).isEltwise()) {
@@ -1478,7 +1366,8 @@ void EltwiseSplitOverChannelsPass::run() {
         auto oData = l->outData.front();
         auto out_width = GetDataDimSize(oData, DataDimName::W);
         auto totalElementsForOutput = details::product(oData->getDims().begin(), oData->getDims().end());
-        auto maxAffineElements = getPassManager()->getPolicy().GNAAffineDiagonalPolicy.limitedTo;
+         // gna limit this to be OxFFFF
+        auto maxAffineElements = 65536 - 64;
         if (totalElementsForOutput <= maxAffineElements) {
             continue;
         }
@@ -1629,31 +1518,25 @@ void SubstituteScaleShiftBroadCastPass::run() {
         }
 
         gnalog() << "Substitution ScaleShift broadcast for layer: " << l->name << "\n";
-        // approach 1 - weights tiling
-        if (getPassManager()->getPolicy().ScaleShiftPolicy == Policy::ScaleShift::WEIGHTS_TILING) {
-            if (nElements % scaleShift->_weights->size()) {
-                THROW_GNA_EXCEPTION << "Cannot tile weights for layer: " << l->name << ", due to weights size not GCD of dims product";
-            }
-            scaleShift->_weights = tileBlob(scaleShift->_weights, nElements);
-            if (scaleShift->_biases) {
-                if (nElements % scaleShift->_biases->size()) {
-                    THROW_GNA_EXCEPTION << "Cannot tile biases for layer: " << l->name << ", due to biases size not GCD of dims product";
-                }
-                scaleShift->_biases = tileBlob(scaleShift->_biases, nElements);
+        if (nElements % scaleShift->_weights->size()) {
+            THROW_GNA_EXCEPTION << "Cannot tile weights for layer: " << l->name << ", due to weights size not GCD of dims product";
+        }
+        scaleShift->_weights = tileBlob(scaleShift->_weights, nElements);
+        if (scaleShift->_biases) {
+            if (nElements % scaleShift->_biases->size()) {
+                THROW_GNA_EXCEPTION << "Cannot tile biases for layer: " << l->name << ", due to biases size not GCD of dims product";
             }
+            scaleShift->_biases = tileBlob(scaleShift->_biases, nElements);
+        }
 
-            auto tensor = InferenceEngine::TensorDesc(insData->getTensorDesc());
-            tensor.reshape(SizeVector{ batchSize, nElements }, Layout::NC);
-            auto reshapeName = scaleShift->name + "_input_" + std::to_string(0) + "_reshape";
-            auto reshape = CNNNetworkCreateReshape(tensor, reshapeName, quantized);
-            auto layer_before_scale_shift = getCreatorLayer(insData);
+        auto tensor = InferenceEngine::TensorDesc(insData->getTensorDesc());
+        tensor.reshape(SizeVector{ batchSize, nElements }, Layout::NC);
+        auto reshapeName = scaleShift->name + "_input_" + std::to_string(0) + "_reshape";
+        auto reshape = CNNNetworkCreateReshape(tensor, reshapeName, quantized);
+        auto layer_before_scale_shift = getCreatorLayer(insData);
 
-            CNNNetworkInsertLayer(layer_before_scale_shift.lock(), l, reshape);
-            gnalog() << "\tInserted " << reshapeName << " between " << layer_before_scale_shift.lock()->name << " and " << l->name << std::endl;
-        } else {
-            THROW_GNA_EXCEPTION << "Not implemented substitution of scaleshift broadcast policy of "
-                                << getPassManager()->getPolicy().ScaleShiftPolicy <<  "using layers tiling, layer: " << l->name;
-        }
+        CNNNetworkInsertLayer(layer_before_scale_shift.lock(), l, reshape);
+        gnalog() << "\tInserted " << reshapeName << " between " << layer_before_scale_shift.lock()->name << " and " << l->name << std::endl;
     }
 }
 
diff --git a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.hpp b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.hpp
index af98288396e8c4..97e1e942e5c844 100644
--- a/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.hpp
+++ b/inference-engine/src/gna_plugin/optimizer/gna_pass_manager.hpp
@@ -8,7 +8,6 @@
 #include <string>
 #include <map>
 #include <ie_common.h>
-#include "gna_plugin_policy.hpp"
 
 namespace GNAPluginNS {
 /**
@@ -29,7 +28,6 @@ class IPassManager {
 public:
     virtual ~IPassManager() = default;
     virtual int &getIntVar(std::string name) = 0;
-    virtual const Policy &getPolicy() const = 0;
     virtual const bool& isLowPrecision() const = 0;
     virtual InferenceEngine::CNNNetwork &getNetwork() = 0;
 };
@@ -75,17 +73,6 @@ DECL_PASS(InsertIdentityLayer);
  */
 DECL_PASS(SubstituteScaleShiftBroadCast);
 
-/**
- * @brief GNA convolution layers have deinterleaved layout, while affine one doesn't
- * so between convolution and affine layers permute layers need to be inserted,
- * current MO approach is to insert such permutations
- * since GNA-HW already support conv->affine in permuted for, this pass inverses MO behavior
- * so its remove permutations of certain form conv->conv, and between conv->affine
- * and insert permutation between conv->affine if they are missed in IR
- * @param layers
- */
-DECL_PASS(ReversePermutations);
-
 /**
  * @brief Pass support --disable_nhwc_to_nchw option in MO
  * @param layers
@@ -220,7 +207,6 @@ DECL_PASS(MoveFakeQuantizeLayerIntoQuantParams);
 DECL_PASS(TransposeWeightsFromNCHWToNHWC);
 
 struct PassManagerSettings {
-    Policy policy;
     /// @brief whether to run passes before copy
     bool runBeforeCopy;
     bool lowPrecision;
@@ -245,9 +231,6 @@ class PassManager : public IPassManager, public std::enable_shared_from_this<Pas
     int & getIntVar(std::string name) override {
         return intMap[name];
     }
-    const Policy & getPolicy() const override {
-        return settings.policy;
-    }
     const bool& isLowPrecision() const override {
         return settings.lowPrecision;
     }
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.cpp b/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.cpp
index f6d611faffe9a9..7b328aa9e97c49 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.cpp
@@ -107,7 +107,6 @@ void GNAPropagateMatcher :: match() {
     try {
         // matching gna propagate forward call.
         GNAPlugin plugin(_env.config);
-        plugin.SetPolicy(_env.policy);
         size_t inputSize = 10;
         size_t outputSize = 10;
         InputsDataMap inputsInfo;
@@ -508,7 +507,7 @@ void GNAPluginCreationMatcher :: match() {
 void GNAPluginAOTMatcher :: match() {
     // matching gna_propagate forward call.
     MockICNNNetwork net;
-    
+
     size_t weightsSize = 656384;
     auto weights = make_shared_blob<uint8_t >({ Precision::U8, {weightsSize}, Layout::C });
     weights->allocate();
@@ -744,7 +743,7 @@ void GNAQueryStateMatcher :: match() {
         auto weights = make_shared_blob<uint8_t>({ Precision::U8, {weightsSize}, Layout::C });
         weights->allocate();
         fillWeights(weights);
-        
+
         InferenceEngine::Core core;
         InferenceEngine::CNNNetwork network;
         ASSERT_NO_THROW_IE_EXCEPTION(network = core.ReadNetwork(_env.model, weights));
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.hpp b/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.hpp
index 16074bf4ba6bc3..da05e64aa0a921 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.hpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/gna_matcher.hpp
@@ -23,7 +23,6 @@
 #include <cpp/ie_cnn_network.h>
 
 #include <backend/dnn_types.h>
-#include <gna_plugin_policy.hpp>
 #include <backend/gna_types.h>
 #include <gna/gna_config.hpp>
 #include <gna_plugin.hpp>
@@ -95,7 +94,6 @@ class GnaPluginTestEnvironment {
     bool exportNetworkOnly = false;
     std::function<void (InferenceEngine::CNNNetwork &)> cb;
     std::map<std::string, std::string> config;
-    GNAPluginNS::Policy policy;
     bool matchThrows = false;
     uint32_t proc_type = static_cast<intel_gna_proc_t>(GNA_SOFTWARE & GNA_HARDWARE);
     std::string importedModelFileName;
@@ -168,10 +166,6 @@ class GNATestConfigurability : public GNATestBase{
         _env.config[GNA_CONFIG_KEY(DEVICE_MODE)] = GNA_CONFIG_VALUE(SW_FP32);
         return *dynamic_cast<T*>(this);
     }
-    T & withPolicy(GNAPluginNS::Policy::ConcatAlignment concatAlignmentPolicy) {
-        _env.policy.ConcatAlignmentPolicy = concatAlignmentPolicy;
-        return *dynamic_cast<T*>(this);
-    }
     T & withGNADeviceMode(std::string value) {
         _env.config[GNA_CONFIG_KEY(DEVICE_MODE)] = value;
         return *dynamic_cast<T*>(this);
diff --git a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_align_filter2_tests.cpp b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_align_filter2_tests.cpp
index 75fdc98b934934..2181706cb2cb58 100644
--- a/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_align_filter2_tests.cpp
+++ b/inference-engine/tests_deprecated/unit/engines/gna/layers/gna_align_filter2_tests.cpp
@@ -11,7 +11,7 @@
 #include <ie_precision.hpp>
 #include "../gna_matcher.hpp"
 
-using GNAAlignFilterTestParams  = std::tuple<InferenceEngine::Precision, GNAPluginNS::Policy::ConcatAlignment, std::size_t, std::size_t>;
+using GNAAlignFilterTestParams  = std::tuple<InferenceEngine::Precision, std::size_t, std::size_t>;
 using namespace GNAPluginNS;
 
 class GNAAlignFilterTest : public GNATest<>,
@@ -20,11 +20,9 @@ class GNAAlignFilterTest : public GNATest<>,
 
     static std::string getTestName(const testing::TestParamInfo<GNAAlignFilterTestParams>& params) {
         std::string test_name;
-        if (std::get<1>(params.param) == GNAPluginNS::Policy::ConcatAlignment::FAST) {
-            test_name += "fast_";
-        }
-        test_name += "concat_of(" + std::to_string(std::get<2>(params.param));
-        test_name += "_" + std::to_string(std::get<3>(params.param));
+        test_name += "fast_";
+        test_name += "concat_of(" + std::to_string(std::get<1>(params.param));
+        test_name += "_" + std::to_string(std::get<2>(params.param));
         test_name += ")_on_";
         test_name += std::get<0>(params.param).name();
         return test_name;
@@ -34,10 +32,9 @@ class GNAAlignFilterTest : public GNATest<>,
 
     InferenceEngine::Precision precision = InferenceEngine::Precision::FP32;
     std::size_t concat_inputs[2];
-    GNAPluginNS::Policy::ConcatAlignment alignmentPolicy;
 
     void SetUp() override {
-        std::tie(precision, alignmentPolicy, concat_inputs[0], concat_inputs[1]) = GetParam();
+        std::tie(precision, concat_inputs[0], concat_inputs[1]) = GetParam();
     }
 
     std::shared_ptr<ngraph::Function> getNgraphModel() {
@@ -83,40 +80,18 @@ TEST_P(GNAAlignFilterTest, concatWith_2_Inputs_Small_mem_footprint) {
         return getFastAffineFilterParams(sz).second;
     };
 
-    switch(alignmentPolicy) {
-        case  Policy::ConcatAlignment::ENABLED : {
-            //align first input by 8
-            auto firstFilter = ALIGN(concat_inputs[0], 8) * concat_inputs[0];
-            //align first input by 8
-            auto extraLeftElementsForSecond = concat_inputs[0] + 32 - ALIGN(concat_inputs[0], 32);
-
-            auto secondFilter = ALIGN(concat_inputs[1], 8) * (extraLeftElementsForSecond + concat_inputs[1]);
-
-            expected_affine_size = firstFilter + secondFilter;
-            break;
-        }
-        case   Policy::ConcatAlignment::FAST  : {
-
-            expected_copy_layers = getNumCopyElements(concat_inputs[0]);
-            expected_affine_size = getsNumFilterWeights(concat_inputs[0]);
+    expected_copy_layers = getNumCopyElements(concat_inputs[0]);
+    expected_affine_size = getsNumFilterWeights(concat_inputs[0]);
 
-            // calculation size for second filter
-            auto offset = ALIGN(concat_inputs[0], 32) - 32;
-            auto zerolen = concat_inputs[0] - offset;
-            auto second_output_len = zerolen + concat_inputs[1];
+    // calculation size for second filter
+    auto offset = ALIGN(concat_inputs[0], 32) - 32;
+    auto zerolen = concat_inputs[0] - offset;
+    auto second_output_len = zerolen + concat_inputs[1];
 
-            expected_affine_size += second_output_len  * ALIGN(concat_inputs[1], 8);
-            break;
-        }
-
-        default : {
-            FAIL() << "unsupported align policy: " << alignmentPolicy;
-        }
-    }
+    expected_affine_size += second_output_len  * ALIGN(concat_inputs[1], 8);
 
     assert_that().onInferNgraphModel(ngraf)
         .inNotCompactMode()
-        .withPolicy(alignmentPolicy)
         .withGNAConfig(std::string(GNA_CONFIG_KEY(SCALE_FACTOR)) + "_0", 1.0f)
         .withGNAConfig(std::string(GNA_CONFIG_KEY(SCALE_FACTOR)) + "_1", 1.0f)
         .withGNAConfig(GNA_CONFIG_KEY(PRECISION), precision.name())
@@ -152,7 +127,6 @@ TEST_P(GNAAlignFilterTest, concatWith_2_Inputs_accurate) {
             .gna()
             .propagate_forward()
             .onCPU()
-            .withPolicy(alignmentPolicy)
             .called_with()
             .input(ngraf->get_parameters().at(0)->get_name(), input_data[0])
             .input(ngraf->get_parameters().at(1)->get_name(), input_data[1])
@@ -161,7 +135,6 @@ TEST_P(GNAAlignFilterTest, concatWith_2_Inputs_accurate) {
         assert_that().onInferNgraphModel(ngraf)
             .inNotCompactMode()
             .gna()
-            .withPolicy(alignmentPolicy)
             .withGNAConfig(std::string(GNA_CONFIG_KEY(SCALE_FACTOR)) + "_0", 1.0f)
             .withGNAConfig(std::string(GNA_CONFIG_KEY(SCALE_FACTOR)) + "_1", 1.0f)
             .withGNAConfig(GNA_CONFIG_KEY(PRECISION), "I16")
@@ -175,8 +148,6 @@ INSTANTIATE_TEST_SUITE_P(
     GNAAlignFilterTest,
     testing::Combine(
     testing::Values(InferenceEngine::Precision::FP32, InferenceEngine::Precision::I16),
-    //fast or not fast alignment policy
-    testing::Values(GNAPluginNS::Policy::ConcatAlignment::FAST, GNAPluginNS::Policy::ConcatAlignment::ENABLED),
     // Size of first Split layer output
     testing::Values(31, 49),
     // Size of second Split layer output

From 240e1950258d4c8fdd2743d8aafb7687f8123385 Mon Sep 17 00:00:00 2001
From: Yuan Hu <yuan2.hu@intel.com>
Date: Wed, 23 Jun 2021 17:37:28 +0800
Subject: [PATCH 175/178] fix Round tests (#6294)

issue:test_mode always use one type

Signed-off-by: Hu, Yuan2 <yuan2.hu@intel.com>
---
 ngraph/test/visitors/op/round.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ngraph/test/visitors/op/round.cpp b/ngraph/test/visitors/op/round.cpp
index 159d07edd332d7..f99e39252976f7 100644
--- a/ngraph/test/visitors/op/round.cpp
+++ b/ngraph/test/visitors/op/round.cpp
@@ -22,7 +22,7 @@ void static test_mode(opset5::Round::RoundMode mode)
 {
     NodeBuilder::get_ops().register_factory<opset5::Round>();
     auto data = make_shared<op::Parameter>(element::f32, Shape{200});
-    auto round = make_shared<opset5::Round>(data, opset5::Round::RoundMode::HALF_TO_EVEN);
+    auto round = make_shared<opset5::Round>(data, mode);
     NodeBuilder builder(round);
     auto g_round = as_type_ptr<opset5::Round>(builder.create());
 

From d690ac687161f9fa7d1a40cdc32d41714b5c9990 Mon Sep 17 00:00:00 2001
From: iliya mironov <iliya.mironov@intel.com>
Date: Wed, 23 Jun 2021 13:18:33 +0300
Subject: [PATCH 176/178] Add yolov4 converting tutorial (#6212)

* Add yolov4 converting tutorial

* Update doc

* Update doc
---
 .../Convert_YOLO_From_Tensorflow.md           | 51 +++++++++++++++----
 1 file changed, 41 insertions(+), 10 deletions(-)

diff --git a/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_YOLO_From_Tensorflow.md b/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_YOLO_From_Tensorflow.md
index 653165576ce125..eb2a1611e0251e 100644
--- a/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_YOLO_From_Tensorflow.md
+++ b/docs/MO_DG/prepare_model/convert_model/tf_specific/Convert_YOLO_From_Tensorflow.md
@@ -1,18 +1,49 @@
 # Converting YOLO* Models to the Intermediate Representation (IR) {#openvino_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_YOLO_From_Tensorflow}
 
-This tutorial explains how to convert real-time object detection YOLOv1\*, YOLOv2\*, and YOLOv3\* public models to the Intermediate Representation (IR). All YOLO\* models are originally implemented in the DarkNet\* framework and consist of two files:
+This document explains how to convert real-time object detection YOLOv1\*, YOLOv2\*, YOLOv3\* and YOLOv4\* public models to the Intermediate Representation (IR). All YOLO\* models are originally implemented in the DarkNet\* framework and consist of two files:
 * `.cfg` file with model configurations  
 * `.weights` file with model weights
 
 Depending on a YOLO model version, the Model Optimizer converts it differently:
 
-- YOLOv3 has several implementations. This tutorial uses a TensorFlow implementation of YOLOv3 model, which can be directly converted to the IR.
+- YOLOv4 must be first converted from Keras\* to TensorFlow 2\*.
+- YOLOv3 has several implementations. This tutorial uses a TensorFlow implementation of YOLOv3 model, which can be directly converted to an IR.
 - YOLOv1 and YOLOv2 models must be first converted to TensorFlow\* using DarkFlow\*.
 
+## <a name="yolov4-to-ir"></a>Convert YOLOv4 Model to IR
+
+This section explains how to convert the YOLOv4 Keras\* model from the [https://github.com/Ma-Dan/keras-yolo4](https://github.com/Ma-Dan/keras-yolo4]) repository to an IR. To convert the YOLOv4 model, follow the instructions below:
+
+1. Download YOLOv4 weights from [yolov4.weights](https://drive.google.com/open?id=1cewMfusmPjYWbrnuJRuKhPMwRe_b9PaT).
+
+2. Clone the repository with the YOLOv4 model.
+```sh
+git clone https://github.com/Ma-Dan/keras-yolo4.git
+```
+
+3. Convert the model to the TensorFlow 2\* format. Save the code below to the `converter.py` file in the same folder as you downloaded `yolov4.weights` and run it.
+```python
+from keras-yolo4.model import Mish
+
+model = tf.keras.models.load_model('yolo4_weight.h5', custom_objects={'Mish': Mish})
+tf.saved_model.save(model, 'yolov4')
+```
+
+```sh
+python converter.py 
+```
+
+4. Run Model Optimizer to converter the model from the TensorFlow 2 format to an IR:
+
+> **NOTE:** Before you run the convertion, make sure you have installed all the Model Optimizer dependencies for TensorFlow 2.
+```sh
+python mo.py --saved_model_dir yolov4 --output_dir models/IRs --input_shape [1,608,608,3] --model_name yolov4 
+```
+
 ## <a name="yolov3-to-ir"></a>Convert YOLOv3 Model to IR
 
-On GitHub*, you can find several public versions of TensorFlow YOLOv3 model implementation. This tutorial explains how to convert YOLOv3 model from
-the [https://github.com/mystic123/tensorflow-yolo-v3](https://github.com/mystic123/tensorflow-yolo-v3) repository (commit ed60b90) to IR , but the process is similar for other versions of TensorFlow YOLOv3 model.
+On GitHub*, you can find several public versions of TensorFlow YOLOv3 model implementation. This section explains how to convert YOLOv3 model from
+the [https://github.com/mystic123/tensorflow-yolo-v3](https://github.com/mystic123/tensorflow-yolo-v3) repository (commit ed60b90) to an IR , but the process is similar for other versions of TensorFlow YOLOv3 model.
 
 ### <a name="yolov3-overview"></a>Overview of YOLOv3 Model Architecture
 Originally, YOLOv3 model includes feature extractor called `Darknet-53` with three branches at the end that make detections at three different scales. These branches must end with the YOLO `Region` layer.
@@ -45,7 +76,7 @@ python3 convert_weights_pb.py --class_names coco.names --data_format NHWC --weig
 ```sh
 python3 convert_weights_pb.py --class_names coco.names --data_format NHWC --weights_file yolov3-tiny.weights --tiny
 ```
-At this step, you may receive a warning like `WARNING:tensorflow:Entity <...> could not be transformed and will be executed as-is.`. To workaround this issue, switch to gast 0.2.2 with the following command:
+At this step, you may receive a warning like `WARNING:tensorflow:Entity <...> could not be transformed and will be executed as-is.`. To work around this issue, switch to gast 0.2.2 with the following command:
 ```sh
 pip3 install --user gast==0.2.2
 ```
@@ -55,7 +86,7 @@ If you have YOLOv3 weights trained for an input image with the size different fr
 python3 convert_weights_pb.py --class_names coco.names --data_format NHWC --weights_file yolov3_608.weights --size 608
 ```
 
-### Convert YOLOv3 TensorFlow Model to the IR
+### Convert YOLOv3 TensorFlow Model to IR
 
 To solve the problems explained in the <a href="#yolov3-overview">YOLOv3 architecture overview</a> section, use the `yolo_v3.json` or `yolo_v3_tiny.json` (depending on a model) configuration file with custom operations located in the `<OPENVINO_INSTALL_DIR>/deployment_tools/model_optimizer/extensions/front/tf` repository.
 
@@ -79,7 +110,7 @@ It consists of several attributes:<br>
 where:
 - `id` and `match_kind` are parameters that you cannot change.
 - `custom_attributes` is a parameter that stores all the YOLOv3 specific attributes:
-    - `classes`, `coords`, `num`, and `masks` are attributes that you should copy from the configuration file
+    - `classes`, `coords`, `num`, and `masks` are attributes that you should copy from the configuration 
     file that was used for model training. If you used DarkNet officially shared weights,
     you can use `yolov3.cfg` or `yolov3-tiny.cfg` configuration file from https://github.com/pjreddie/darknet/tree/master/cfg. Replace the default values in `custom_attributes` with the parameters that
     follow the `[yolo]` titles in the configuration file.
@@ -87,7 +118,7 @@ where:
     - `entry_points` is a node name list to cut off the model and append the Region layer with custom attributes specified above.
 
 
-To generate the IR of the YOLOv3 TensorFlow model, run:<br>
+To generate an IR of the YOLOv3 TensorFlow model, run:<br>
 ```sh
 python3 mo_tf.py                                                   \
 --input_model /path/to/yolo_v3.pb                                  \
@@ -96,7 +127,7 @@ python3 mo_tf.py                                                   \
 --output_dir <OUTPUT_MODEL_DIR>
 ```
 
-To generate the IR of the YOLOv3-tiny TensorFlow model, run:<br>
+To generate an IR of the YOLOv3-tiny TensorFlow model, run:<br>
 ```sh
 python3 mo_tf.py                                                        \
 --input_model /path/to/yolo_v3_tiny.pb                                  \
@@ -179,4 +210,4 @@ The model was trained with input values in the range `[0,1]`. OpenVINO&trade; to
 * `--transformations_config` adds missing `Region` layers to the model. In the IR, the `Region` layer has name `RegionYolo`.
 For other applicable parameters, refer to [Convert Model from TensorFlow](../Convert_Model_From_TensorFlow.md).
 
-> **NOTE:** The color channel order (RGB or BGR) of an input data should match the channel order of the model training dataset. If they are different, perform the `RGB<->BGR` conversion specifying the command-line parameter: `--reverse_input_channels`. Otherwise, inference results may be incorrect. For more information about the parameter, refer to **When to Reverse Input Channels** section of [Converting a Model Using General Conversion Parameters](../Converting_Model_General.md).
+> **NOTE:** The color channel order (RGB or BGR) of an input data should match the channel order of the model training dataset. If they are different, perform the `RGB<->BGR` conversion specifying the command-line parameter: `--reverse_input_channels`. Otherwise, inference results may be incorrect. For more information about the parameter, refer to **When to Reverse Input Channels** section of [Converting a Model Using General Conversion Parameters](../Converting_Model_General.md).
\ No newline at end of file

From fcc07f80cd51b68ef760c002a6aef604adae619f Mon Sep 17 00:00:00 2001
From: Egor Shulman <egor.shulman@intel.com>
Date: Wed, 23 Jun 2021 14:22:10 +0300
Subject: [PATCH 177/178] [IE CPU] Reimplement extension nodes via MKLDNNNode
 API (#5784)

---
 .../src/mkldnn_plugin/cpu_types.h             |  23 +-
 .../src/mkldnn_plugin/mkldnn_node.cpp         |  21 +
 .../src/mkldnn_plugin/mkldnn_node.h           |  44 +-
 .../src/mkldnn_plugin/nodes/bucketize.cpp     | 242 -------
 .../nodes/ctc_greedy_decoder.cpp              | 183 -----
 .../nodes/ctc_greedy_decoder_seq_len.cpp      | 203 ------
 .../src/mkldnn_plugin/nodes/ctc_loss.cpp      | 302 --------
 .../src/mkldnn_plugin/nodes/cum_sum.cpp       | 271 -------
 .../mkldnn_plugin/nodes/detectionoutput.cpp   | 663 ------------------
 .../nodes/detectionoutput_onnx.cpp            | 402 -----------
 .../src/mkldnn_plugin/nodes/gather_tree.cpp   | 184 -----
 .../src/mkldnn_plugin/nodes/grn.cpp           |  91 ---
 .../src/mkldnn_plugin/nodes/list_tbl.hpp      |  21 -
 .../src/mkldnn_plugin/nodes/log_softmax.cpp   | 136 ----
 .../nodes/mkldnn_bucketize_node.cpp           | 218 ++++++
 .../nodes/mkldnn_bucketize_node.h             |  43 ++
 .../nodes/mkldnn_concat_node.cpp              |   2 +-
 .../nodes/mkldnn_ctc_greedy_decoder_node.cpp  | 167 +++++
 .../nodes/mkldnn_ctc_greedy_decoder_node.h    |  32 +
 ...mkldnn_ctc_greedy_decoder_seq_len_node.cpp | 170 +++++
 .../mkldnn_ctc_greedy_decoder_seq_len_node.h  |  35 +
 .../nodes/mkldnn_ctc_loss_node.cpp            | 279 ++++++++
 .../nodes/mkldnn_ctc_loss_node.h              |  32 +
 .../nodes/mkldnn_cum_sum_node.cpp             | 279 ++++++++
 .../mkldnn_plugin/nodes/mkldnn_cum_sum_node.h |  50 ++
 .../nodes/mkldnn_def_conv_node.cpp            |   2 +-
 .../nodes/mkldnn_detection_output_node.cpp    | 601 ++++++++++++++++
 .../nodes/mkldnn_detection_output_node.h      |  86 +++
 .../mkldnn_plugin/nodes/mkldnn_eltwise_node.h |   1 -
 .../mkldnn_embedding_bag_offset_sum_node.cpp  |   2 +-
 .../mkldnn_embedding_bag_packed_sum_node.cpp  |   2 +-
 .../mkldnn_embedding_segments_sum_node.cpp    |   2 +-
 ...mental_detectron_detection_output_node.cpp | 369 ++++++++++
 ...rimental_detectron_detection_output_node.h |  46 ++
 ...n_generate_proposals_single_image_node.cpp | 429 ++++++++++++
 ...ron_generate_proposals_single_image_node.h |  50 ++
 ...ntal_detectron_priorgridgenerator_node.cpp |  95 +++
 ...mental_detectron_priorgridgenerator_node.h |  46 ++
 ...tal_detectron_roifeatureextractor_node.cpp | 413 +++++++++++
 ...ental_detectron_roifeatureextractor_node.h |  41 ++
 ...n_experimental_detectron_topkrois_node.cpp |  82 +++
 ...dnn_experimental_detectron_topkrois_node.h |  40 ++
 ... => mkldnn_extract_image_patches_node.cpp} | 278 ++++----
 ...pp => mkldnn_extract_image_patches_node.h} |  36 +-
 .../nodes/mkldnn_gather_elements_node.cpp     |   2 +-
 .../nodes/mkldnn_gather_nd_node.cpp           |   2 +-
 .../nodes/mkldnn_gather_node.cpp              |   4 +-
 .../nodes/mkldnn_gather_tree_node.cpp         | 148 ++++
 .../nodes/mkldnn_gather_tree_node.h           |  38 +
 .../mkldnn_plugin/nodes/mkldnn_grn_node.cpp   |  81 +++
 .../src/mkldnn_plugin/nodes/mkldnn_grn_node.h |  30 +
 .../nodes/mkldnn_log_softmax_node.cpp         | 116 +++
 .../nodes/mkldnn_log_softmax_node.h           |  34 +
 .../mkldnn_plugin/nodes/mkldnn_math_node.cpp  |  10 +-
 .../mkldnn_plugin/nodes/mkldnn_math_node.h    |   1 -
 .../nodes/mkldnn_non_max_suppression_node.cpp | 406 +++++++++++
 .../nodes/mkldnn_non_max_suppression_node.h   | 102 +++
 .../nodes/mkldnn_proposal_node.cpp            | 198 ++++++
 .../nodes/mkldnn_proposal_node.h              |  42 ++
 .../mkldnn_plugin/nodes/mkldnn_range_node.cpp | 140 ++++
 .../mkldnn_plugin/nodes/mkldnn_range_node.h   |  34 +
 .../nodes/mkldnn_reorg_yolo_node.cpp          |  93 +++
 .../nodes/mkldnn_reorg_yolo_node.h            |  30 +
 .../nodes/mkldnn_reverse_sequence_node.cpp    | 182 +++++
 .../nodes/mkldnn_reverse_sequence_node.h      |  38 +
 .../nodes/mkldnn_softmax_node.cpp             |   2 +-
 .../mkldnn_plugin/nodes/mkldnn_topk_node.cpp  | 478 +++++++++++++
 .../mkldnn_plugin/nodes/mkldnn_topk_node.h    | 114 +++
 .../nodes/mkldnn_transpose_node.cpp           |   2 +-
 .../nodes/non_max_suppression.cpp             | 464 ------------
 .../nodes/priorgridgenerator_onnx.cpp         | 121 ----
 .../src/mkldnn_plugin/nodes/proposal.cpp      | 227 ------
 .../src/mkldnn_plugin/nodes/proposal_onnx.cpp | 450 ------------
 .../src/mkldnn_plugin/nodes/range.cpp         | 164 -----
 .../src/mkldnn_plugin/nodes/reorg_yolo.cpp    |  99 ---
 .../mkldnn_plugin/nodes/reverse_sequence.cpp  | 209 ------
 .../nodes/roifeatureextractor_onnx.cpp        | 433 ------------
 .../src/mkldnn_plugin/nodes/topk.cpp          | 572 ---------------
 .../src/mkldnn_plugin/nodes/topkrois_onnx.cpp | 101 ---
 .../extract_image_patches.cpp                 |   2 +-
 80 files changed, 6169 insertions(+), 5714 deletions(-)
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/bucketize.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/ctc_greedy_decoder.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/ctc_greedy_decoder_seq_len.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/ctc_loss.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/cum_sum.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/detectionoutput.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/detectionoutput_onnx.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/gather_tree.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/grn.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/log_softmax.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_bucketize_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_bucketize_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_seq_len_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_seq_len_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_loss_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_loss_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_cum_sum_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_cum_sum_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_detection_output_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_detection_output_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_detection_output_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_detection_output_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_generate_proposals_single_image_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_generate_proposals_single_image_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_priorgridgenerator_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_priorgridgenerator_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_roifeatureextractor_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_roifeatureextractor_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_topkrois_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_topkrois_node.h
 rename inference-engine/src/mkldnn_plugin/nodes/{extract_image_patches.cpp => mkldnn_extract_image_patches_node.cpp} (66%)
 rename inference-engine/src/mkldnn_plugin/nodes/{extract_image_patches.hpp => mkldnn_extract_image_patches_node.h} (64%)
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_tree_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_tree_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_grn_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_grn_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_log_softmax_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_log_softmax_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_non_max_suppression_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_non_max_suppression_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_proposal_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_proposal_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_range_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_range_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_reorg_yolo_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_reorg_yolo_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_reverse_sequence_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_reverse_sequence_node.h
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_topk_node.cpp
 create mode 100644 inference-engine/src/mkldnn_plugin/nodes/mkldnn_topk_node.h
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/non_max_suppression.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/priorgridgenerator_onnx.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/proposal.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/proposal_onnx.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/range.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/reorg_yolo.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/reverse_sequence.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/roifeatureextractor_onnx.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/topk.cpp
 delete mode 100644 inference-engine/src/mkldnn_plugin/nodes/topkrois_onnx.cpp

diff --git a/inference-engine/src/mkldnn_plugin/cpu_types.h b/inference-engine/src/mkldnn_plugin/cpu_types.h
index d7f55446024c8c..e5bc8af0b5c745 100644
--- a/inference-engine/src/mkldnn_plugin/cpu_types.h
+++ b/inference-engine/src/mkldnn_plugin/cpu_types.h
@@ -64,7 +64,28 @@ enum Type {
     Reference,
     ShuffleChannels,
     DFT,
-    Math
+    Math,
+    CTCLoss,
+    Bucketize,
+    CTCGreedyDecoder,
+    CTCGreedyDecoderSeqLen,
+    CumSum,
+    DetectionOutput,
+    ExperimentalDetectronDetectionOutput,
+    LogSoftmax,
+    TopK,
+    GatherTree,
+    GRN,
+    Range,
+    Proposal,
+    ReorgYolo,
+    ReverseSequence,
+    ExperimentalDetectronTopKROIs,
+    ExperimentalDetectronROIFeatureExtractor,
+    ExperimentalDetectronPriorGridGenerator,
+    ExperimentalDetectronGenerateProposalsSingleImage,
+    ExtractImagePatches,
+    NonMaxSuppression
 };
 
 enum Algorithm {
diff --git a/inference-engine/src/mkldnn_plugin/mkldnn_node.cpp b/inference-engine/src/mkldnn_plugin/mkldnn_node.cpp
index e2e2a3276b8c78..e46c7a7b0bdf9e 100644
--- a/inference-engine/src/mkldnn_plugin/mkldnn_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/mkldnn_node.cpp
@@ -203,6 +203,27 @@ static const InferenceEngine::details::caseless_unordered_map<std::string, Type>
         { "SoftPlus", Math},
         { "Softsign", Math},
         { "Tan", Math},
+        { "CTCLoss", CTCLoss},
+        { "Bucketize", Bucketize},
+        { "CTCGreedyDecoder", CTCGreedyDecoder},
+        { "CTCGreedyDecoderSeqLen", CTCGreedyDecoderSeqLen},
+        { "CumSum", CumSum},
+        { "DetectionOutput", DetectionOutput},
+        { "ExperimentalDetectronDetectionOutput", ExperimentalDetectronDetectionOutput},
+        { "LogSoftmax", LogSoftmax},
+        { "TopK", TopK},
+        { "GatherTree", GatherTree},
+        { "GRN", GRN},
+        { "Range", Range},
+        { "Proposal", Proposal},
+        { "ReorgYolo", ReorgYolo},
+        { "ReverseSequence", ReverseSequence},
+        { "ExperimentalDetectronTopKROIs", ExperimentalDetectronTopKROIs},
+        { "ExperimentalDetectronROIFeatureExtractor", ExperimentalDetectronROIFeatureExtractor},
+        { "ExperimentalDetectronPriorGridGenerator", ExperimentalDetectronPriorGridGenerator},
+        { "ExperimentalDetectronGenerateProposalsSingleImage", ExperimentalDetectronGenerateProposalsSingleImage},
+        { "ExtractImagePatches", ExtractImagePatches},
+        { "NonMaxSuppressionIEInternal", NonMaxSuppression}
 };
 
 Type TypeFromName(const std::string type) {
diff --git a/inference-engine/src/mkldnn_plugin/mkldnn_node.h b/inference-engine/src/mkldnn_plugin/mkldnn_node.h
index e5f86f03ea0c4a..29618d51fdbaf5 100644
--- a/inference-engine/src/mkldnn_plugin/mkldnn_node.h
+++ b/inference-engine/src/mkldnn_plugin/mkldnn_node.h
@@ -129,7 +129,7 @@ static std::string NameFromType(Type type) {
         case EmbeddingBagPackedSum:
             return "EmbeddingBagPackedSum";
         case EmbeddingBagOffsetsSum:
-            return "EmbeddingBagPackedSum";
+            return "EmbeddingBagOffsetsSum";
         case Gather:
             return "Gather";
         case GatherElements:
@@ -150,6 +150,48 @@ static std::string NameFromType(Type type) {
             return "DFT";
         case Math:
             return "Math";
+        case CTCLoss:
+            return "CTCLoss";
+        case Bucketize:
+            return "Bucketize";
+        case CTCGreedyDecoder:
+            return "CTCGreedyDecoder";
+        case CTCGreedyDecoderSeqLen:
+            return "CTCGreedyDecoderSeqLen";
+        case CumSum:
+            return "CumSum";
+        case DetectionOutput:
+            return "DetectionOutput";
+        case ExperimentalDetectronDetectionOutput:
+            return "ExperimentalDetectronDetectionOutput";
+        case LogSoftmax:
+            return "LogSoftmax";
+        case TopK:
+            return "TopK";
+        case GatherTree:
+            return "GatherTree";
+        case GRN:
+            return "GRN";
+        case Range:
+            return "Range";
+        case Proposal:
+            return "Proposal";
+        case ReorgYolo:
+            return "ReorgYolo";
+        case ReverseSequence:
+            return "ReverseSequence";
+        case ExperimentalDetectronTopKROIs:
+            return "ExperimentalDetectronTopKROIs";
+        case ExperimentalDetectronROIFeatureExtractor:
+            return "ExperimentalDetectronROIFeatureExtractor";
+        case ExperimentalDetectronPriorGridGenerator:
+            return "ExperimentalDetectronPriorGridGenerator";
+        case ExperimentalDetectronGenerateProposalsSingleImage:
+            return "ExperimentalDetectronGenerateProposalsSingleImage";
+        case ExtractImagePatches:
+            return "ExtractImagePatches";
+        case NonMaxSuppression:
+            return "NonMaxSuppression";
         default:
             return "Unknown";
     }
diff --git a/inference-engine/src/mkldnn_plugin/nodes/bucketize.cpp b/inference-engine/src/mkldnn_plugin/nodes/bucketize.cpp
deleted file mode 100644
index febdf1a8dfd0f2..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/bucketize.cpp
+++ /dev/null
@@ -1,242 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <cmath>
-#include <string>
-#include <vector>
-#include <array>
-#include <cassert>
-#include <algorithm>
-#include <limits>
-#include "ie_parallel.hpp"
-#include <ngraph/opsets/opset3.hpp>
-
-using namespace MKLDNNPlugin;
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-class BucketizeImpl : public ExtLayerBase {
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            const auto bucketsize = std::dynamic_pointer_cast<const ngraph::opset3::Bucketize>(op);
-            if (!bucketsize) {
-                errorMessage = "Only opset3 Bucketize operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    std::string errorPrefix;
-
-public:
-    explicit BucketizeImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            errorPrefix = "Bucketize layer with name '" + op->get_friendly_name() + "' ";
-            const auto bucketsize = std::dynamic_pointer_cast<const ngraph::opset3::Bucketize>(op);
-
-            if (op->get_input_size() != 2 || op->get_output_size() != 1) {
-                IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
-            }
-
-            // check one attribute
-            with_right = bucketsize->get_with_right_bound();
-
-            // check precisions for input and output tensors
-            input_precision = details::convertPrecision(op->get_input_element_type(INPUT_TENSOR_PORT));
-            if (input_precision != Precision::FP32 && input_precision != Precision::I32 &&
-                input_precision != Precision::I64) {
-                input_precision = Precision::FP32;
-            }
-            boundaries_precision = details::convertPrecision(op->get_input_element_type(INPUT_BINS_PORT));
-            if (boundaries_precision != Precision::FP32 && boundaries_precision != Precision::I32 &&
-                boundaries_precision != Precision::I64) {
-                boundaries_precision = Precision::FP32;
-            }
-            output_precision = details::convertPrecision(op->get_output_element_type(OUTPUT_TENSOR_PORT));
-            if (output_precision != Precision::I32 && output_precision != Precision::I64) {
-                output_precision = Precision::I32;
-            }
-
-            // check dimensions of input tensors
-            SizeVector input_tensor_dims = op->get_input_shape(INPUT_TENSOR_PORT);
-            if (input_tensor_dims.size() < 1) {
-                IE_THROW() << errorPrefix << " has incorrect dimensions of the input.";
-            }
-            SizeVector input_bin_dims = op->get_input_shape(INPUT_BINS_PORT);
-            if (input_bin_dims.size() != 1) {
-                IE_THROW() << errorPrefix << " has incorrect dimensions of the boundaries tensor.";
-            }
-            if (input_bin_dims[0] != 0) {
-                with_bins = true;
-            }
-            num_bin_values = input_bin_dims[0];
-
-            num_values = std::accumulate(input_tensor_dims.begin(), input_tensor_dims.end(), size_t(1), std::multiplies<size_t>());
-
-            addConfig(op, {{TensorDescCreatorTypes::ncsp, input_precision},
-                           {TensorDescCreatorTypes::ncsp, boundaries_precision}},
-                          {{TensorDescCreatorTypes::ncsp, output_precision}});
-        }
-        catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs, ResponseDesc *resp) noexcept override {
-        auto precision_mask = getPrecisionMask(input_precision, boundaries_precision, output_precision);
-
-        switch (precision_mask) {
-        case getPrecisionMask(Precision::FP32, Precision::FP32, Precision::I32):
-            bucketize<PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::I32>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::FP32, Precision::FP32, Precision::I64):
-            bucketize<PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::I64>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::FP32, Precision::I32, Precision::I32):
-            bucketize<PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::I32>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::FP32, Precision::I32, Precision::I64):
-            bucketize<PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::I64>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::FP32, Precision::I64, Precision::I32):
-            bucketize<PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::I32>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::FP32, Precision::I64, Precision::I64):
-            bucketize<PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::I64>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I32, Precision::FP32, Precision::I32):
-            bucketize<PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::I32>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I32, Precision::FP32, Precision::I64):
-            bucketize<PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::I64>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I32, Precision::I32, Precision::I32):
-            bucketize<PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::I32>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I32, Precision::I32, Precision::I64):
-            bucketize<PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::I64>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I32, Precision::I64, Precision::I32):
-            bucketize<PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::I32>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I32, Precision::I64, Precision::I64):
-            bucketize<PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::I64>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I64, Precision::FP32, Precision::I32):
-            bucketize<PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::I32>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I64, Precision::FP32, Precision::I64):
-            bucketize<PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::FP32>::value_type,
-                PrecisionTrait<Precision::I64>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I64, Precision::I32, Precision::I32):
-            bucketize<PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::I32>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I64, Precision::I32, Precision::I64):
-            bucketize<PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::I32>::value_type,
-                PrecisionTrait<Precision::I64>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I64, Precision::I64, Precision::I32):
-            bucketize<PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::I32>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        case getPrecisionMask(Precision::I64, Precision::I64, Precision::I64):
-            bucketize<PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::I64>::value_type,
-                PrecisionTrait<Precision::I64>::value_type>(inputs[0], inputs[1], outputs[0]);
-            break;
-        default:
-            return GENERAL_ERROR;
-        }
-
-        return OK;
-    }
-
-private:
-    template <typename T, typename T_BOUNDARIES, typename T_IND>
-    void bucketize(Blob::Ptr input, Blob::Ptr boundaries, Blob::Ptr output) {
-        const auto *input_data = input->cbuffer().as<const T *>();
-        const auto *boundaries_data = boundaries->cbuffer().as<const T_BOUNDARIES *>();
-        auto *output_data = output->buffer().as<T_IND *>();
-
-        if (with_bins == false) {
-            memset(output_data, 0, num_values * sizeof(T_IND));
-            return;
-        }
-
-        // boundaries are assumed to be sorted and to have unique elements
-        parallel_for(num_values, [&](size_t ind) {
-            T value = input_data[ind];
-            if (with_right) {
-                auto low = std::lower_bound(boundaries_data, boundaries_data + num_bin_values, value);
-                output_data[ind] = static_cast<T_IND>(low - boundaries_data);
-            } else {
-                auto up = std::upper_bound(boundaries_data, boundaries_data + num_bin_values, value);
-                output_data[ind] = static_cast<T_IND>(up - boundaries_data);
-            }
-        });
-    }
-
-    const size_t INPUT_TENSOR_PORT = 0;
-    const size_t INPUT_BINS_PORT = 1;
-    const size_t OUTPUT_TENSOR_PORT = 0;
-
-    size_t num_values = 0;
-    size_t num_bin_values = 0;
-    bool with_right = false;
-    bool with_bins = false;
-
-    Precision input_precision;
-    Precision boundaries_precision;
-    Precision output_precision;
-};
-
-REG_FACTORY_FOR(BucketizeImpl, Bucketize);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/ctc_greedy_decoder.cpp b/inference-engine/src/mkldnn_plugin/nodes/ctc_greedy_decoder.cpp
deleted file mode 100644
index 0ba6ca7e960230..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/ctc_greedy_decoder.cpp
+++ /dev/null
@@ -1,183 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-#include "ie_parallel.hpp"
-#include <ngraph/op/ctc_greedy_decoder.hpp>
-#include <nodes/common/tensor_desc_creator.h>
-
-#include <string>
-#include <vector>
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-class CTCGreedyDecoderImpl: public ExtLayerBase {
-public:
-    static bool isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            auto greedyDecOp = ngraph::as_type_ptr<const ngraph::op::v0::CTCGreedyDecoder>(op);
-            if (!greedyDecOp) {
-                errorMessage = "Node is not an instance of the CTCGreedyDecoder operation from operation set v0.";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-
-        return true;
-    }
-
-    explicit CTCGreedyDecoderImpl(const std::shared_ptr<ngraph::Node>& op) : mergeRepeated_(true) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            std::string errPrefix = "CTCGreedyDecoder layer with name '" + op->get_friendly_name() + "' ";
-            if (op->get_input_size() != 2)
-                IE_THROW() << errPrefix << "has invalid number of input edges: " << op->get_input_size();
-            if (op->get_output_size() != 1)
-                IE_THROW() << errPrefix << "has invalid number of outputs edges: " << op->get_output_size();
-
-            if (op->get_input_shape(DATA_INDEX)[0] != op->get_input_shape(SEQUENCE_LENGTH_INDEX)[0] &&
-                    op->get_input_shape(DATA_INDEX)[1] != op->get_input_shape(SEQUENCE_LENGTH_INDEX)[1])
-                IE_THROW() << errPrefix << "has invalid input shapes.";
-
-            Precision inDataPrecision = details::convertPrecision(op->get_input_element_type(DATA_INDEX));
-            if (inDataPrecision != Precision::FP32 && inDataPrecision != Precision::BF16)
-                IE_THROW() << errPrefix << "has unsupported 'data' input precision: " << inDataPrecision;
-
-            Precision seqLenPrecision = details::convertPrecision(op->get_input_element_type(SEQUENCE_LENGTH_INDEX));
-            if (seqLenPrecision != Precision::FP32 && seqLenPrecision != Precision::BF16)
-                IE_THROW() << errPrefix << "has unsupported 'sequence_length' input precision: " << seqLenPrecision;
-
-            auto greedyDecOp = ngraph::as_type_ptr<const ngraph::op::v0::CTCGreedyDecoder>(op);
-            mergeRepeated_ = greedyDecOp->get_ctc_merge_repeated();
-
-            addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                           {TensorDescCreatorTypes::ncsp, Precision::FP32}},
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-            throw;
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs,
-                       ResponseDesc *resp) noexcept override {
-        const float* probabilities = inputs[DATA_INDEX]->cbuffer().as<const float*>() +
-            inputs[DATA_INDEX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        const float* sequenceMask = inputs[SEQUENCE_LENGTH_INDEX]->cbuffer().as<const float*>() +
-            inputs[SEQUENCE_LENGTH_INDEX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        float* outputSequences = outputs[0]->buffer().as<float*>() +
-            outputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-        const size_t T = inputs[DATA_INDEX]->getTensorDesc().getDims()[0];
-        const size_t B = inputs[DATA_INDEX]->getTensorDesc().getDims()[1];
-        const int C = inputs[DATA_INDEX]->getTensorDesc().getDims()[2];
-        const size_t BC = B * C;
-        const size_t CB1 = C * (B - 1);
-
-        const int blankIndex = C - 1;
-
-        std::vector<size_t> sequenceLengths(B, 0);
-        parallel_for(B, [&](size_t b) {
-            size_t t = 0;
-            for (; t < T; t++) {
-                if (sequenceMask[B * t + b] == 0.f)
-                    break;
-            }
-            sequenceLengths[b] = t;
-        });
-
-        size_t workAmount = 0;
-        for (size_t b = 0; b < B; b++) {
-            workAmount += sequenceLengths[b];
-        }
-
-        // Parallelization could not be made directly by T due to output index depends on merged classes and
-        // blank index, thus could not be shared between threads. Better to divide operation on two steps.
-        // At the first stage find the maximum index. At second stage merge if needed.
-        // Such approach makes parallelization more efficient.
-        auto threadBody = [&](const int ithr, const int nthr) {
-            size_t start(0lu), end(0lu);
-            splitter(workAmount, nthr, ithr, start, end);
-            if (start >= end)
-                return;
-            size_t tStart = 0lu, bStart = 0lu;
-            for (; bStart < B; bStart++) {
-                tStart += sequenceLengths[bStart];
-                if (tStart >= start) {
-                    tStart = start - (tStart - sequenceLengths[bStart]);
-                    break;
-                }
-            }
-
-            size_t workCounter = start;
-
-            for (size_t b = bStart; b < B; ++b) {
-                size_t outputIndex = b * T + tStart;
-                const float* probs = probabilities + b * C + BC * tStart;
-                size_t sequenceLength = sequenceLengths[b];
-
-                for (size_t t = tStart; t < sequenceLength; ++t) {
-                    int maxClassIdx = 0;
-
-                    float maxProb = probs[0];
-                    ++probs;
-
-                    for (int c = 1; c < C; ++c, ++probs) {
-                        if (*probs > maxProb) {
-                            maxClassIdx = c;
-                            maxProb = *probs;
-                        }
-                    }
-                    probs += CB1;
-                    outputSequences[outputIndex++] = static_cast<float>(maxClassIdx);
-
-                    if (++workCounter >= end) {
-                        return;
-                    }
-                }
-                tStart = 0lu;
-            }
-        }; // thread body
-
-        parallel_nt(0, threadBody);
-
-        parallel_for(B, [&](size_t b) {
-            int prevClassIdx = -1;
-            size_t outputIndex = b * T;
-            const size_t sequenceLength = sequenceLengths[b];
-            float* shiftedOut = outputSequences + b * T;
-            for (size_t t = 0; t < sequenceLength; ++t) {
-                if (*shiftedOut < blankIndex &&
-                        !(mergeRepeated_ && *shiftedOut == prevClassIdx)) {
-                    outputSequences[outputIndex++] = *shiftedOut;
-                }
-                prevClassIdx = *shiftedOut;
-                shiftedOut++;
-            }
-            std::fill(outputSequences + outputIndex, outputSequences + (b + 1) * T, -1.f);
-        });
-
-        return OK;
-    }
-
-private:
-    const size_t DATA_INDEX = 0lu;
-    const size_t SEQUENCE_LENGTH_INDEX = 1lu;
-    bool mergeRepeated_;
-};
-
-REG_FACTORY_FOR(CTCGreedyDecoderImpl, CTCGreedyDecoder);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/ctc_greedy_decoder_seq_len.cpp b/inference-engine/src/mkldnn_plugin/nodes/ctc_greedy_decoder_seq_len.cpp
deleted file mode 100644
index c60684ee0af3f8..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/ctc_greedy_decoder_seq_len.cpp
+++ /dev/null
@@ -1,203 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-#include "ie_parallel.hpp"
-#include <ngraph/op/ctc_greedy_decoder_seq_len.hpp>
-#include <nodes/common/tensor_desc_creator.h>
-
-#include <string>
-#include <vector>
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-class CTCGreedyDecoderSeqLenImpl: public ExtLayerBase {
-public:
-    static bool isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            auto greedyDecOp = ngraph::as_type_ptr<const ngraph::op::v6::CTCGreedyDecoderSeqLen>(op);
-            if (!greedyDecOp) {
-                errorMessage = "Node is not an instance of the CTCGreedyDecoderSeqLen operation from operation set v6.";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-
-        return true;
-    }
-
-    explicit CTCGreedyDecoderSeqLenImpl(const std::shared_ptr<ngraph::Node>& op) : mergeRepeated_(true) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            std::string errPrefix = "CTCGreedyDecoderSeqLen layer with name '" + op->get_friendly_name() + "' ";
-            if (op->get_input_size() < 2 || op->get_input_size() > 3)
-                IE_THROW() << errPrefix << "has invalid number of input edges: " << op->get_input_size();
-            if (op->get_output_size() != 2)
-                IE_THROW() << errPrefix << "has invalid number of outputs edges: " << op->get_output_size();
-
-            if (op->get_input_shape(DATA_INDEX)[0] != op->get_input_shape(SEQUENCE_LENGTH_INDEX)[0])
-                IE_THROW() << errPrefix << "has invalid input shapes.";
-
-            Precision inDataPrecision = details::convertPrecision(op->get_input_element_type(DATA_INDEX));
-            if (inDataPrecision != Precision::FP32 && inDataPrecision != Precision::BF16)
-                IE_THROW() << errPrefix << "has unsupported 'data' input precision: " << inDataPrecision;
-
-            Precision seqLenPrecision = details::convertPrecision(op->get_input_element_type(SEQUENCE_LENGTH_INDEX));
-            if (seqLenPrecision != Precision::I32 && seqLenPrecision != Precision::I64)
-                IE_THROW() << errPrefix << "has unsupported 'sequence_length' input precision: " << seqLenPrecision;
-
-            auto greedyDecOp = ngraph::as_type_ptr<const ngraph::op::v6::CTCGreedyDecoderSeqLen>(op);
-            mergeRepeated_ = greedyDecOp->get_merge_repeated();
-
-            if (op->get_input_size() == BLANK_INDEX) {
-                addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                               {TensorDescCreatorTypes::ncsp, Precision::I32}},
-                              {{TensorDescCreatorTypes::ncsp, Precision::I32},
-                               {TensorDescCreatorTypes::ncsp, Precision::I32}});
-            } else {
-                Precision blIdxPrecision = details::convertPrecision(op->get_input_element_type(BLANK_INDEX));
-                if (blIdxPrecision != Precision::I32 && blIdxPrecision != Precision::I64)
-                    IE_THROW() << errPrefix << "has unsupported 'blank_index' input precision: " << blIdxPrecision;
-
-                addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                               {TensorDescCreatorTypes::ncsp, Precision::I32},
-                               {TensorDescCreatorTypes::ncsp, Precision::I32}},
-                              {{TensorDescCreatorTypes::ncsp, Precision::I32},
-                               {TensorDescCreatorTypes::ncsp, Precision::I32}});
-            }
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-            throw;
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs,
-                       ResponseDesc *resp) noexcept override {
-        const float* probabilities = inputs[DATA_INDEX]->cbuffer().as<const float*>() +
-            inputs[DATA_INDEX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        const int* sequenceLengths = inputs[SEQUENCE_LENGTH_INDEX]->cbuffer().as<const int*>() +
-            inputs[SEQUENCE_LENGTH_INDEX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        int* decodedClasses = outputs[DECODED_CLASSES_INDEX]->buffer().as<int*>() +
-            outputs[DECODED_CLASSES_INDEX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        int* decodedClassesLength = outputs[DECODED_CLASSES_LENGTH_INDEX]->buffer().as<int*>() +
-            outputs[DECODED_CLASSES_LENGTH_INDEX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-        const auto& inDims = inputs[DATA_INDEX]->getTensorDesc().getDims();
-        const size_t B = inDims[0];
-        const size_t T = inDims[1];
-        const int C = inDims[2];
-        const size_t TC = T * C;
-
-        int blankIndex = C - 1;
-        if (inputs.size() > BLANK_INDEX)
-            blankIndex = (inputs[BLANK_INDEX]->cbuffer().as<const int*>() +
-                inputs[BLANK_INDEX]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0];
-
-        size_t workAmount = 0;
-        for (size_t b = 0; b < B; b++) {
-            if (sequenceLengths[b] > T) {
-                if (resp) {
-                    std::string errorMsg = errPrefix
-                        + ". Sequence length " + std::to_string(sequenceLengths[b])
-                        + " cannot be greater than according decoded classes dimension size "
-                        + std::to_string(outputs[DECODED_CLASSES_INDEX]->getTensorDesc().getDims()[1]);
-                    errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-                }
-                return PARAMETER_MISMATCH;
-            }
-            workAmount += sequenceLengths[b];
-        }
-        // Parallelization could not be made directly by T due to output index depends on merged classes and
-        // blank index, thus could not be shared between threads. Better to divide operation on two steps.
-        // At the first stage find the maximum index. At second stage merge if needed.
-        // Such approach makes parallelization more efficient.
-        auto threadBody = [&](const int ithr, const int nthr) {
-            size_t start(0lu), end(0lu);
-            splitter(workAmount, nthr, ithr, start, end);
-            if (start >= end)
-                return;
-            size_t tStart = 0lu, bStart = 0lu;
-            for (; bStart < B; bStart++) {
-                tStart += sequenceLengths[bStart];
-                if (tStart >= start) {
-                    tStart = start - (tStart - sequenceLengths[bStart]);
-                    break;
-                }
-            }
-
-            size_t workCounter = start;
-
-            for (size_t b = bStart; b < B; ++b) {
-                size_t outputIndex = b * T + tStart;
-                const float* probs = probabilities + b * TC + C * tStart;
-                const size_t actualSeqLen = sequenceLengths[b];
-
-                for (size_t t = tStart; t < actualSeqLen; ++t) {
-                    int maxClassIdx = 0;
-                    float maxProb = probs[0];
-                    probs++;
-
-                    for (int c = 1; c < C; c++, probs++) {
-                        if (*probs > maxProb) {
-                            maxClassIdx = c;
-                            maxProb = *probs;
-                        }
-                    }
-                    decodedClasses[outputIndex++] = maxClassIdx;
-
-                    if (++workCounter >= end) {
-                        return;
-                    }
-                }
-                tStart = 0lu;
-            }
-        }; // thread body
-
-        parallel_nt(0, threadBody);
-
-        parallel_for(B, [&](size_t b) {
-            int prevClassIdx = -1;
-            size_t outputIndex = b * T;
-            const size_t actualSeqLen = sequenceLengths[b];
-            int* shiftedOut = decodedClasses + b * T;
-
-            for (size_t t = 0; t < actualSeqLen; ++t) {
-                if (*shiftedOut != blankIndex &&
-                        !(mergeRepeated_ && *shiftedOut == prevClassIdx)) {
-                    decodedClasses[outputIndex++] = *shiftedOut;
-                }
-                prevClassIdx = *shiftedOut;
-                shiftedOut++;
-            }
-            std::fill(decodedClasses + outputIndex, decodedClasses + (b + 1) * T, -1);
-            decodedClassesLength[b] = outputIndex - b * T;
-        });
-
-        return OK;
-    }
-
-private:
-    const size_t DATA_INDEX = 0lu;
-    const size_t SEQUENCE_LENGTH_INDEX = 1lu;
-    const size_t BLANK_INDEX = 2lu;
-    const size_t DECODED_CLASSES_INDEX = 0lu;
-    const size_t DECODED_CLASSES_LENGTH_INDEX = 1lu;
-    bool mergeRepeated_;
-    std::string errPrefix;
-};
-
-REG_FACTORY_FOR(CTCGreedyDecoderSeqLenImpl, CTCGreedyDecoderSeqLen);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/ctc_loss.cpp b/inference-engine/src/mkldnn_plugin/nodes/ctc_loss.cpp
deleted file mode 100644
index 84d6b55a1a47e9..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/ctc_loss.cpp
+++ /dev/null
@@ -1,302 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-#include "ie_parallel.hpp"
-#include <ngraph/op/ctc_loss.hpp>
-#include <nodes/common/tensor_desc_creator.h>
-
-#include <cmath>
-
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-class CTCLossImpl : public ExtLayerBase {
-public:
-    static bool isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            auto ctcLossOp = ngraph::as_type_ptr<const ngraph::op::v4::CTCLoss>(op);
-            if (!ctcLossOp) {
-                errorMessage = "Node is not an instance of the CTCLoss operation from operation set v4.";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-
-        return true;
-    }
-
-    explicit CTCLossImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            _logPrefix = std::string("CTCLoss layer with name '") + op->get_friendly_name() + "'";
-
-            if (op->get_input_size() != 4 && op->get_input_size() != 5)
-                IE_THROW() << _logPrefix << " has invalid inputs number.";
-
-            auto ctcLossOp = ngraph::as_type_ptr<const ngraph::op::v4::CTCLoss>(op);
-            _ctcMergeRepeated = ctcLossOp->get_ctc_merge_repeated();
-            _preprocessCollapseRepeated = ctcLossOp->get_preprocess_collapse_repeated();
-            _unique = ctcLossOp->get_unique();
-
-            std::vector<DataConfigurator> inDataConfigurators;
-            inDataConfigurators.push_back({TensorDescCreatorTypes::ncsp, Precision::FP32});
-            for (int i = 1; i < op->get_input_size(); i++) {
-                inDataConfigurators.push_back({TensorDescCreatorTypes::ncsp, Precision::I32});
-            }
-            addConfig(op, inDataConfigurators,
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-            throw;
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs,
-                       std::vector<Blob::Ptr>& outputs,
-                       ResponseDesc *resp) noexcept override {
-        StatusCode returnCode = OK;
-
-        const float* logits = inputs[0]->cbuffer().as<const float*>() +
-            inputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        const int* logitsLength = inputs[1]->cbuffer().as<const int*>() +
-            inputs[1]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        const int* labels = inputs[2]->cbuffer().as<const int*>() +
-            inputs[2]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        const int* labelsLength = inputs[3]->cbuffer().as<const int*>() +
-            inputs[3]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        float* dstData = outputs[0]->buffer().as<float*>() +
-            outputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-        const auto& logitsShape = inputs[0]->getTensorDesc().getDims();
-        const size_t batchNum = logitsShape[0];
-        const size_t maxTime = logitsShape[1];
-        const size_t classesNum = logitsShape[2];
-
-        int blankIndex = classesNum - 1;
-        if (inputs.size() > 4) {
-            blankIndex = inputs[4]->cbuffer().as<const int*>()[0];
-        }
-
-        std::vector<int> decodedTargetLenB(batchNum, 0);
-        std::vector<std::vector<int>> targetDB(batchNum);
-        std::vector<std::vector<std::vector<float>>> logProbabilitiesB(batchNum);
-        std::vector<std::string> errorMsgB(parallel_get_max_threads());
-
-        auto threadBody_1 = [&](const int ithr, const int nthr) {
-            size_t start(0lu), end(0lu);
-            splitter(batchNum, nthr, ithr, start, end);
-            if (start >= end)
-                return;
-
-            for (size_t b = start; b < end; b++) {
-                if (logitsLength[b] < 0 || labelsLength[b] < 0 || logitsLength[b] > maxTime || labelsLength[b] > logitsLength[b]) {
-                    errorMsgB[ithr] = _logPrefix + ". Logit length cannot be greater than max sequence length. "
-                        + "Label length cannot be greater than a logit length"
-                        + " and both cannot be negative.\nMaxSeqLen: "
-                        + std::to_string(maxTime) + "; Logit len: " + std::to_string(logitsLength[b])
-                        + "; Label len: " + std::to_string(labelsLength[b]);
-                    returnCode = GENERAL_ERROR;
-                    return;
-                }
-                const size_t actualLogitLen = logitsLength[b];
-                const size_t actualTargetLen = labelsLength[b];
-                size_t decodedTargetLen = 0lu;
-
-                // Decoding target: merge repeated characters if preprocess_collapse_repeated == True,
-                // find unique elemnts if unique == True.
-                // Inserts blanks before each index and a blank at the end.
-                const int* target = &labels[b * maxTime];
-                targetDB[b].resize(actualTargetLen * 2 + 1);
-                auto& targetD = targetDB[b];
-                if (_unique) {
-                    std::unordered_set<int> uniqVals;
-                    for (size_t t = 0lu; t < actualTargetLen; t++) {
-                        if (uniqVals.find(target[t]) != uniqVals.end()) {
-                            continue;
-                        }
-                        uniqVals.insert(target[t]);
-                        targetD[decodedTargetLen++] = blankIndex;
-                        targetD[decodedTargetLen++] = target[t];
-                    }
-                    targetD[decodedTargetLen++] = blankIndex;
-                } else if (_preprocessCollapseRepeated) {
-                    auto prevValue = target[0];
-                    targetD[decodedTargetLen++] = blankIndex;
-                    targetD[decodedTargetLen++] = target[0];
-                    for (size_t t = 1lu; t < actualTargetLen; t++) {
-                        if (target[t] == prevValue) {
-                            continue;
-                        }
-                        targetD[decodedTargetLen++] = blankIndex;
-                        targetD[decodedTargetLen++] = prevValue = target[t];
-                    }
-                    targetD[decodedTargetLen++] = blankIndex;
-                } else {
-                    for (size_t t = 0lu; t < actualTargetLen; t++) {
-                        targetD[decodedTargetLen++] = blankIndex;
-                        targetD[decodedTargetLen++] = target[t];
-                    }
-                    targetD[decodedTargetLen++] = blankIndex;
-                }
-                decodedTargetLenB[b] = decodedTargetLen;
-
-                auto& logProbabilities = logProbabilitiesB[b];
-                logProbabilities.resize(actualLogitLen);
-                for (size_t ll = 0; ll < actualLogitLen; ll++) {
-                    logProbabilities[ll].resize(decodedTargetLen);
-                }
-            } // for batch
-        }; // threadBody_1
-
-        parallel_nt(0, threadBody_1);
-        if (returnCode != OK) {
-            std::string resErr("");
-            for (auto& err : errorMsgB) {
-                if (!err.empty())
-                    resErr += err + "\n";
-                resErr.copy(resp->msg, sizeof(resp->msg) - 1);
-            }
-            return returnCode;
-        }
-
-        const size_t TC = maxTime * classesNum;
-
-        size_t workAmount2 = 0lu;
-        for (size_t b = 0; b < batchNum; b++) {
-            workAmount2 += logitsLength[b];
-        }
-
-        auto threadBody_2 = [&](const int ithr, const int nthr) {
-            size_t start(0lu), end(0lu);
-            size_t sB(0lu), sT(0lu);
-            splitter(workAmount2, nthr, ithr, start, end);
-            if (start >= end)
-                return;
-            int64_t cw = 0, st = start;
-            for (; sB < batchNum; sB++) {
-                cw += logitsLength[sB];
-                if (cw >= st) {
-                    sT = logitsLength[sB] + st - cw;
-                    break;
-                }
-            }
-            size_t workCounter = start;
-
-            for (size_t b = sB; b < batchNum; b++) {
-                const size_t actualLogitLen = logitsLength[b];
-                const size_t decodedTargetLen = decodedTargetLenB[b];
-                auto& logProbabilities = logProbabilitiesB[b];
-                auto& targetD = targetDB[b];
-
-                double expSum = 0.0;
-                size_t btcT = b * TC + sT * classesNum;
-                // logProbabilities = logSoftmax = logits[b][t][c] - ln(sum_c(exp(logits[b][t])))
-                for (size_t t = sT; t < actualLogitLen; t++) {
-                    expSum = 0.0;
-                    for (size_t c = 0lu; c < classesNum; c++) {
-                        expSum += std::exp(logits[btcT + c]);
-                    }
-                    for (size_t s = 0lu; s < decodedTargetLen; s++) {
-                        logProbabilities[t][s] = logits[btcT + targetD[s]] - std::log(expSum);
-                    }
-                    btcT += classesNum;
-                    if (++workCounter >= end) {
-                        return;
-                    }
-                }
-                sT = 0lu;
-            }  // for batch
-        }; // threadBody_2
-
-        parallel_nt(0, threadBody_2);
-
-        const auto float_inf = std::numeric_limits<float>::infinity();
-
-        auto sumLogs = [&float_inf](float log1, float log2) {
-            if (log1 == -float_inf) {
-                return log2;
-            } else if (log2 == -float_inf) {
-                return log1;
-            } else {
-                if (log1 > log2)
-                    return log1 + std::log1pf(std::exp(log2 - log1));
-                else
-                    return log2 + std::log1pf(std::exp(log1 - log2));
-            }
-        };
-
-        auto threadBody_3 = [&](const int ithr, const int nthr) {
-            size_t start(0lu), end(0lu);
-            splitter(batchNum, nthr, ithr, start, end);
-            if (start >= end)
-                return;
-
-            // As per Connectionist Temporal Classification - Labeling Unsegmented Sequence Data with Recurrent Neural Networks:
-            // Graves et al., 2016, paragraph 4.1 (10)
-            for (size_t b = start; b < end; b++) {
-                auto& targetD = targetDB[b];
-                auto& logProbabilities = logProbabilitiesB[b];
-                const int actualLogitLen = logitsLength[b];
-                const int decodedTargetLen = decodedTargetLenB[b];
-                std::vector<std::vector<float>> logBwd(decodedTargetLen, std::vector<float>(actualLogitLen, -float_inf));
-                for (int s = decodedTargetLen - 2; s < decodedTargetLen; s++)
-                    logBwd[s][actualLogitLen - 1] = 0.f;
-
-                for (int t = actualLogitLen - 2; t >= 0; t--) {
-                    const int t_1 = t + 1;
-                    for (int s = std::max(0, decodedTargetLen - (2 * (actualLogitLen - t)));
-                            s < std::min(decodedTargetLen, 2 * (t_1)); s++) {
-                        if (_ctcMergeRepeated || targetD[s] == blankIndex) {
-                            logBwd[s][t] = sumLogs(logBwd[s][t],
-                                logBwd[s][t_1] + logProbabilities[t_1][s]);
-                        }
-
-                        if (s + 1 < decodedTargetLen) {
-                            logBwd[s][t] = sumLogs(logBwd[s][t],
-                                logBwd[s + 1][t_1] + logProbabilities[t_1][s + 1]);
-                        }
-
-                        if (s + 2 < decodedTargetLen) {
-                            if (targetD[s] != blankIndex && (!_ctcMergeRepeated || (targetD[s] != targetD[s + 2]))) {
-                                logBwd[s][t] = sumLogs(logBwd[s][t],
-                                    logBwd[s + 2][t_1] + logProbabilities[t_1][s + 2]);
-                            }
-                        }
-                    }
-                }
-
-                logBwd[0][0] += logProbabilities[0][0];
-                logBwd[1][0] += logProbabilities[0][(decodedTargetLen > 1) ? 1 : 0];
-
-                dstData[b] = -sumLogs(logBwd[0][0], logBwd[1][0]);
-            } // for batch
-        }; // threadBody_3
-
-        parallel_nt(0, threadBody_3);
-
-        return returnCode;
-    } // execute
-
-protected:
-    bool _ctcMergeRepeated;
-    bool _preprocessCollapseRepeated;
-    bool _unique;
-
-    std::string _logPrefix;
-};
-
-REG_FACTORY_FOR(CTCLossImpl, CTCLoss);
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/cum_sum.cpp b/inference-engine/src/mkldnn_plugin/nodes/cum_sum.cpp
deleted file mode 100644
index 8940527713cd36..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/cum_sum.cpp
+++ /dev/null
@@ -1,271 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "list.hpp"
-#include "base.hpp"
-
-#include <string>
-#include <vector>
-#include "ie_parallel.hpp"
-#include "ie_precision.hpp"
-#include <ngraph/opsets/opset1.hpp>
-#include <ngraph/opsets/opset3.hpp>
-#include <ie_ngraph_utils.hpp>
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-class CumSumImpl: public ExtLayerBase {
-    enum { CUM_SUM_DATA, AXIS, numOfInputs };
-    bool exclusive;
-    bool reverse;
-    size_t numOfDims;
-    size_t axis = 0;
-    std::vector<size_t> shape;
-
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            const auto cumsum = std::dynamic_pointer_cast<const ngraph::opset3::CumSum>(op);
-            if (!cumsum) {
-                errorMessage = "Only opset3 CumSum operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-public:
-    explicit CumSumImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            layerName = op->get_friendly_name();
-            if ((op->get_input_size() != numOfInputs && op->get_input_size() != (numOfInputs - 1)) || op->get_output_size() != 1)
-                IE_THROW() << "CumSum layer with name '" << layerName << "' has incorrect number of input/output edges!";
-
-            const auto &dataShape = op->get_input_shape(CUM_SUM_DATA);
-            if (dataShape.size() < 1) {
-                IE_THROW() << "CumSum layer with name '" << layerName << "' doesn't support 'data' input tensor with rank: " << dataShape.size();
-            }
-            numOfDims = dataShape.size();
-
-            const auto cumsum = std::dynamic_pointer_cast<const ngraph::opset3::CumSum>(op);
-            exclusive = cumsum->is_exclusive();
-            reverse = cumsum->is_reverse();
-
-            auto dataPrecision = details::convertPrecision(cumsum->get_input_element_type(CUM_SUM_DATA));
-            if (dataPrecision != Precision::I8 && dataPrecision != Precision::U8 && dataPrecision != Precision::I16 && dataPrecision != Precision::I32 &&
-                dataPrecision != Precision::FP32 && dataPrecision != Precision::I64 && dataPrecision != Precision::U64 && dataPrecision != Precision::BF16)
-                IE_THROW() << "CumSum layer with name '" << layerName << "' has unsupported 'data' input precision: " << dataPrecision.name();
-
-            if (cumsum->get_input_size() == numOfInputs) {
-                const auto& axisTensorPrec = details::convertPrecision(cumsum->get_input_element_type(AXIS));
-                if (axisTensorPrec != Precision::I32 && axisTensorPrec != Precision::I64)
-                    IE_THROW() << "CumSum layer with name '" << layerName << "' has unsupported 'axis' input precision: " << axisTensorPrec.name();
-
-                if (!ngraph::is_scalar(cumsum->get_input_shape(AXIS)))
-                    IE_THROW() << "CumSum layer with name '" << layerName << "' doesn't support 'axis' input tensor with non scalar rank";
-            }
-
-            if (dataShape != cumsum->get_output_shape(0))
-                IE_THROW() << "CumSum layer with name '" << layerName << "' has different 'data' input and output dimensions";
-
-            shape = dataShape;
-
-            std::vector<DataConfigurator> inDataConfigurators;
-            if (dataPrecision == Precision::BF16)
-                dataPrecision = Precision::FP32;
-            inDataConfigurators.push_back({TensorDescCreatorTypes::ncsp, dataPrecision});
-            if (op->get_input_size() > 1)
-                inDataConfigurators.push_back({TensorDescCreatorTypes::ncsp, Precision::I32});
-            addConfig(op, inDataConfigurators, {{TensorDescCreatorTypes::ncsp, dataPrecision}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs, ResponseDesc *resp) noexcept override {
-        if (inputs.size() == numOfInputs)
-            axis = getAxis(inputs[AXIS], inputs[CUM_SUM_DATA]);
-
-        const auto &dataPrecision = inputs[CUM_SUM_DATA]->getTensorDesc().getPrecision();
-        switch (dataPrecision) {
-            case Precision::I8   : { execImpl<int8_t>(inputs[CUM_SUM_DATA], outputs[0]); break; }
-            case Precision::U8   : { execImpl<uint8_t>(inputs[CUM_SUM_DATA], outputs[0]); break; }
-            case Precision::I16  : { execImpl<int16_t>(inputs[CUM_SUM_DATA], outputs[0]); break; }
-            case Precision::I32  : { execImpl<int32_t>(inputs[CUM_SUM_DATA], outputs[0]); break; }
-            case Precision::FP32 : { execImpl<float>(inputs[CUM_SUM_DATA], outputs[0]); break; }
-            case Precision::I64  : { execImpl<int64_t>(inputs[CUM_SUM_DATA], outputs[0]); break; }
-            case Precision::U64  : { execImpl<uint64_t>(inputs[CUM_SUM_DATA], outputs[0]); break; }
-            default : {
-                if (resp) {
-                    std::string errorMsg = "CumSum layer with name '" + layerName + "' has unsupported 'data' input precision: " + dataPrecision.name();
-                    errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-                }
-                return GENERAL_ERROR;
-            }
-        }
-        return OK;
-    }
-
-private:
-    template <typename dataType>
-    void execImpl(const Blob::CPtr& _input, const Blob::Ptr& _output) {
-        const auto *input = _input->cbuffer().as<const dataType *>() + _input->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        auto *output = _output->buffer().as<dataType *>() + _output->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        const std::vector<size_t> strides = _input->getTensorDesc().getBlockingDesc().getStrides();
-
-        if (reverse) {
-            if (exclusive) {
-                cumSum<true, true, dataType>(input, output, strides);
-            } else {
-                cumSum<true, false, dataType>(input, output, strides);
-            }
-        } else {
-            if (exclusive) {
-                cumSum<false, true, dataType>(input, output, strides);
-            } else {
-                cumSum<false, false, dataType>(input, output, strides);
-            }
-        }
-    }
-
-    template <bool reverse, bool exclusive, typename dataType>
-    void cumSum(const dataType *input, dataType *output, const std::vector<size_t> &strides) {
-        SizeVector iterationRange(numOfDims - 1);
-        size_t j = 0;
-        for (size_t i = 0; i < shape.size(); i++) {
-            if (i == axis)
-                continue;
-            iterationRange[j++] = shape[i];
-        }
-        size_t work_amount_dst = std::accumulate(iterationRange.begin(), iterationRange.end(), 1, std::multiplies<size_t>());
-        parallel_nt(0, [&](const int ithr, const int nthr) {
-            size_t start = 0, end = 0;
-            SizeVector counters(numOfDims - 1, 0);
-            splitter(work_amount_dst, nthr, ithr, start, end);
-
-            parallelItInit(start, counters, iterationRange);
-
-            for (size_t iwork = start; iwork < end; ++iwork) {
-                std::vector<size_t> forStartOffset(numOfDims);
-                forStartOffset[axis] = 0;
-                for (size_t offsetIdx = 0, countersIdx = 0; offsetIdx < numOfDims; ++offsetIdx) {
-                    if (offsetIdx == axis) {
-                        continue;
-                    }
-                    forStartOffset[offsetIdx] = counters[countersIdx++];
-                }
-
-                size_t startOffset = getStartOffset(forStartOffset, strides);
-
-                const dataType *inputStart = input + startOffset;
-                dataType *outputStart = output + startOffset;
-
-                size_t offset = strides[axis];
-                if (reverse) {
-                    if (exclusive) {
-                        outputStart[offset*(shape[axis] - 1)] = 0;
-                        for (int64_t i = shape[axis] - 2; i >= 0; i--) {
-                            outputStart[i*offset] = inputStart[(i+1)*offset] + outputStart[(i+1)*offset];
-                        }
-                    } else {
-                        outputStart[offset*(shape[axis] - 1)] = inputStart[offset * (shape[axis] - 1)];
-                        for (int64_t i = shape[axis] - 2; i >= 0; i--) {
-                            outputStart[i*offset] = inputStart[i*offset] + outputStart[(i+1)*offset];
-                        }
-                    }
-                } else {
-                    if (exclusive) {
-                        outputStart[0] = 0;
-                        for (size_t i = 1; i < shape[axis]; i++) {
-                            outputStart[i*offset] = inputStart[(i-1)*offset] + outputStart[(i-1)*offset];
-                        }
-                    } else {
-                        outputStart[0] = inputStart[0];
-                        for (size_t i = 1; i < shape[axis]; i++) {
-                            outputStart[i*offset] = inputStart[i*offset] + outputStart[(i-1)*offset];
-                        }
-                    }
-                }
-
-                parallelItStep(counters, iterationRange);
-            }
-        });
-    }
-
-    void parallelItInit(size_t start, std::vector<size_t>& counters, const std::vector<size_t>& iterationRange) {
-        auto itCounter = counters.rbegin();
-        auto itWork = iterationRange.rbegin();
-        while (itCounter != counters.rend() && itWork != iterationRange.rend()) {
-            *itCounter = start % *itWork;
-            start /= *itWork;
-            ++itCounter;
-            ++itWork;
-        }
-    }
-
-    inline void parallelItStep(std::vector<size_t>& counters, const std::vector<size_t>& iterationRange) {
-        auto itCounter = counters.rbegin();
-        auto itWork = iterationRange.rbegin();
-
-        while (itCounter != counters.rend() && itWork != iterationRange.rend()) {
-            *itCounter = (*itCounter + 1) % *itWork;
-            if (*itCounter != 0) {
-                break;
-            }
-            ++itCounter;
-            ++itWork;
-        }
-    }
-
-    inline size_t getStartOffset(const std::vector<size_t> &forStartOffset, const std::vector<size_t>& strides) const {
-        size_t startOffset = 0;
-        for (size_t idx = 0; idx < forStartOffset.size(); ++idx) {
-            startOffset += forStartOffset[idx] * strides[idx];
-        }
-        return startOffset;
-    }
-
-    size_t getAxis(const Blob::CPtr& _axis, const Blob::CPtr& _data) const {
-        const auto& axisPrecision = _axis->getTensorDesc().getPrecision();
-        const int64_t dataShapeSize = static_cast<int64_t>(_data->getTensorDesc().getDims().size());
-        int64_t axisValueFromBlob;
-        switch (axisPrecision) {
-            case Precision::I32 : {
-                const auto *axisPtr = _axis->cbuffer().as<const int32_t *>();
-                axisValueFromBlob = static_cast<int64_t>(axisPtr[0]);
-                break;
-            }
-            case Precision::I64 : {
-                const auto *axisPtr = _axis->cbuffer().as<const int64_t *>();
-                axisValueFromBlob = axisPtr[0];
-                break;
-            }
-            default : {
-                IE_THROW() << "CumSum layer with name '" << layerName << "'  doesn't support 'axis' input with precision: " << axisPrecision.name();
-            }
-        }
-        if (axisValueFromBlob < -dataShapeSize || axisValueFromBlob > dataShapeSize - 1)
-            IE_THROW() << "CumSum layer with name '" << layerName << "'  has axis with a value out of range: " << axisValueFromBlob;
-        return axisValueFromBlob >= 0 ? axisValueFromBlob : (axisValueFromBlob + dataShapeSize);
-    }
-
-private:
-    std::string layerName;
-};
-
-REG_FACTORY_FOR(CumSumImpl, CumSum);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
\ No newline at end of file
diff --git a/inference-engine/src/mkldnn_plugin/nodes/detectionoutput.cpp b/inference-engine/src/mkldnn_plugin/nodes/detectionoutput.cpp
deleted file mode 100644
index bd3b1da8fc878c..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/detectionoutput.cpp
+++ /dev/null
@@ -1,663 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <cfloat>
-#include <vector>
-#include <cmath>
-#include <string>
-#include <utility>
-#include <algorithm>
-#include "caseless.hpp"
-#include "ie_parallel.hpp"
-#include "common/tensor_desc_creator.h"
-#include <ngraph/op/detection_output.hpp>
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-template <typename T>
-static bool SortScorePairDescend(const std::pair<float, T>& pair1,
-                                 const std::pair<float, T>& pair2) {
-    return pair1.first > pair2.first;
-}
-
-class DetectionOutputImpl: public ExtLayerBase {
-public:
-    bool isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            auto doOp = ngraph::as_type_ptr<const ngraph::op::v0::DetectionOutput>(op);
-            if (!doOp) {
-                errorMessage = "Node is not an instance of the DetectionOutput from the operations set v0.";
-                return false;
-            }
-            if (!details::CaselessEq<std::string>()(doOp->get_attrs().code_type, "caffe.PriorBoxParameter.CENTER_SIZE") &&
-                    !details::CaselessEq<std::string>()(doOp->get_attrs().code_type, "caffe.PriorBoxParameter.CORNER")) {
-                errorMessage = "Unsupported code_type attribute.";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    explicit DetectionOutputImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-            if (op->get_input_size() != 3 && op->get_input_size() != 5)
-                IE_THROW() <<  "Invalid number of input edges.";
-
-            if (op->get_output_size() != 1)
-                IE_THROW() << "Invalid number of output edges.";
-
-            auto doOp = ngraph::as_type_ptr<const ngraph::op::v0::DetectionOutput>(op);
-            auto attributes = doOp->get_attrs();
-
-            _num_classes = attributes.num_classes;
-            _background_label_id = attributes.background_label_id;
-            _top_k = attributes.top_k;
-            _variance_encoded_in_target = attributes.variance_encoded_in_target;
-            _keep_top_k = attributes.keep_top_k[0];
-            _nms_threshold = attributes.nms_threshold;
-            _confidence_threshold = attributes.confidence_threshold;
-            _share_location = attributes.share_location;
-            _clip_before_nms = attributes.clip_before_nms;
-            _clip_after_nms = attributes.clip_after_nms;
-            _decrease_label_id = attributes.decrease_label_id;
-            _normalized = attributes.normalized;
-            _image_height = attributes.input_height;
-            _image_width = attributes.input_width;
-            _prior_size = _normalized ? 4 : 5;
-            _offset = _normalized ? 0 : 1;
-            _num_loc_classes = _share_location ? 1 : _num_classes;
-
-            with_add_box_pred = op->get_input_size() == 5;
-            _objectness_score = attributes.objectness_score;
-
-            _code_type = (details::CaselessEq<std::string>()(attributes.code_type, "caffe.PriorBoxParameter.CENTER_SIZE") ?
-                CodeType::CENTER_SIZE : CodeType::CORNER);
-
-            _num_priors = static_cast<int>(op->get_input_shape(idx_priors).back() / _prior_size);
-            _priors_batches = op->get_input_shape(idx_priors).front() != 1;
-
-            if (_num_priors * _num_loc_classes * 4 != static_cast<int>(op->get_input_shape(idx_location)[1]))
-                IE_THROW() << "Number of priors must match number of location predictions ("
-                                   << _num_priors * _num_loc_classes * 4 << " vs "
-                                   << op->get_input_shape(idx_location)[1] << ")";
-
-            if (_num_priors * _num_classes != static_cast<int>(op->get_input_shape(idx_confidence).back()))
-                IE_THROW() << "Number of priors must match number of confidence predictions.";
-
-            if (_decrease_label_id && _background_label_id != 0)
-                IE_THROW() << "Cannot use decrease_label_id and background_label_id parameter simultaneously.";
-
-            _num = static_cast<int>(op->get_input_shape(idx_confidence)[0]);
-
-            _decoded_bboxes.resize(_num * _num_classes * _num_priors * 4);
-            _buffer.resize(_num * _num_classes * _num_priors);
-            _indices.resize(_num * _num_classes * _num_priors);
-            _detections_count.resize(_num * _num_classes);
-            _bbox_sizes.resize(_num * _num_classes * _num_priors);
-            _num_priors_actual.resize(_num);
-
-            const auto &confSize = op->get_input_shape(idx_confidence);
-            _reordered_conf.resize(std::accumulate(confSize.begin(), confSize.end(), 1, std::multiplies<size_t>()));
-
-            std::vector<DataConfigurator> inDataConfigurators(op->get_input_size(), {TensorDescCreatorTypes::ncsp, Precision::FP32});
-            addConfig(op, inDataConfigurators,
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-            throw;
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs,
-                       ResponseDesc *resp) noexcept override {
-        float *dst_data = outputs[0]->buffer();
-
-        const float *loc_data    = inputs[idx_location]->buffer().as<const float *>();
-        const float *conf_data   = inputs[idx_confidence]->buffer().as<const float *>();
-        const float *prior_data  = inputs[idx_priors]->buffer().as<const float *>();
-        const float *arm_conf_data = inputs.size() > 3 ? inputs[idx_arm_confidence]->buffer().as<const float *>() : nullptr;
-        const float *arm_loc_data = inputs.size() > 4 ? inputs[idx_arm_location]->buffer().as<const float *>() : nullptr;
-
-        const int N = inputs[idx_confidence]->getTensorDesc().getDims()[0];
-
-        float *decoded_bboxes_data = _decoded_bboxes.data();
-        float *reordered_conf_data = _reordered_conf.data();
-        float *bbox_sizes_data     = _bbox_sizes.data();
-        int *detections_data       = _detections_count.data();
-        int *buffer_data           = _buffer.data();
-        int *indices_data          = _indices.data();
-        int *num_priors_actual     = _num_priors_actual.data();
-
-        for (int n = 0; n < N; ++n) {
-            const float *ppriors = prior_data;
-            const float *prior_variances = prior_data + _num_priors*_prior_size;
-            if (_priors_batches) {
-                ppriors += _variance_encoded_in_target ? n*_num_priors*_prior_size : 2*n*_num_priors*_prior_size;
-                prior_variances += _variance_encoded_in_target ? 0 : 2*n*_num_priors*_prior_size;
-            }
-
-            if (_share_location) {
-                const float *ploc = loc_data + n*4*_num_priors;
-                float *pboxes = decoded_bboxes_data + n*4*_num_priors;
-                float *psizes = bbox_sizes_data + n*_num_priors;
-
-                if (with_add_box_pred) {
-                    const float *p_arm_loc = arm_loc_data + n*4*_num_priors;
-                    decodeBBoxes(ppriors, p_arm_loc, prior_variances, pboxes, psizes, num_priors_actual, n, _offset, _prior_size);
-                    decodeBBoxes(pboxes, ploc, prior_variances, pboxes, psizes, num_priors_actual, n, 0, 4, false);
-                } else {
-                    decodeBBoxes(ppriors, ploc, prior_variances, pboxes, psizes, num_priors_actual, n, _offset, _prior_size);
-                }
-            } else {
-                for (int c = 0; c < _num_loc_classes; ++c) {
-                    if (c == _background_label_id) {
-                        continue;
-                    }
-                    const float *ploc = loc_data + n*4*_num_loc_classes*_num_priors + c*4;
-                    float *pboxes = decoded_bboxes_data + n*4*_num_loc_classes*_num_priors + c*4*_num_priors;
-                    float *psizes = bbox_sizes_data + n*_num_loc_classes*_num_priors + c*_num_priors;
-                    if (with_add_box_pred) {
-                        const float *p_arm_loc = arm_loc_data + n*4*_num_loc_classes*_num_priors + c*4;
-                        decodeBBoxes(ppriors, p_arm_loc, prior_variances, pboxes, psizes, num_priors_actual, n, _offset, _prior_size);
-                        decodeBBoxes(pboxes, ploc, prior_variances, pboxes, psizes, num_priors_actual, n, 0, 4, false);
-                    } else {
-                        decodeBBoxes(ppriors, ploc, prior_variances, pboxes, psizes, num_priors_actual, n, _offset, _prior_size);
-                    }
-                }
-            }
-        }
-
-        if (with_add_box_pred) {
-            for (int n = 0; n < N; ++n) {
-                for (int p = 0; p < _num_priors; ++p) {
-                    if (arm_conf_data[n*_num_priors*2 + p * 2 + 1] < _objectness_score) {
-                        for (int c = 0; c < _num_classes; ++c) {
-                            reordered_conf_data[n*_num_priors*_num_classes + c*_num_priors + p] = c == _background_label_id ? 1.0f : 0.0f;
-                        }
-                    } else {
-                        for (int c = 0; c < _num_classes; ++c) {
-                            reordered_conf_data[n*_num_priors*_num_classes + c*_num_priors + p] = conf_data[n*_num_priors*_num_classes + p*_num_classes + c];
-                        }
-                    }
-                }
-            }
-        } else {
-            for (int n = 0; n < N; ++n) {
-                for (int c = 0; c < _num_classes; ++c) {
-                    for (int p = 0; p < _num_priors; ++p) {
-                        reordered_conf_data[n*_num_priors*_num_classes + c*_num_priors + p] = conf_data[n*_num_priors*_num_classes + p*_num_classes + c];
-                    }
-                }
-            }
-        }
-
-        memset(detections_data, 0, N*_num_classes*sizeof(int));
-
-        for (int n = 0; n < N; ++n) {
-            int detections_total = 0;
-
-            if (!_decrease_label_id) {
-                // Caffe style
-                parallel_for(_num_classes, [&](int c) {
-                    if (c != _background_label_id) {  // Ignore background class
-                        int *pindices    = indices_data + n*_num_classes*_num_priors + c*_num_priors;
-                        int *pbuffer     = buffer_data + c*_num_priors;
-                        int *pdetections = detections_data + n*_num_classes + c;
-
-                        const float *pconf = reordered_conf_data + n*_num_classes*_num_priors + c*_num_priors;
-                        const float *pboxes;
-                        const float *psizes;
-                        if (_share_location) {
-                            pboxes = decoded_bboxes_data + n*4*_num_priors;
-                            psizes = bbox_sizes_data + n*_num_priors;
-                        } else {
-                            pboxes = decoded_bboxes_data + n*4*_num_classes*_num_priors + c*4*_num_priors;
-                            psizes = bbox_sizes_data + n*_num_classes*_num_priors + c*_num_priors;
-                        }
-
-                        nms_cf(pconf, pboxes, psizes, pbuffer, pindices, *pdetections, num_priors_actual[n]);
-                    }
-                });
-            } else {
-                // MXNet style
-                int *pindices = indices_data + n*_num_classes*_num_priors;
-                int *pbuffer = buffer_data;
-                int *pdetections = detections_data + n*_num_classes;
-
-                const float *pconf = reordered_conf_data + n*_num_classes*_num_priors;
-                const float *pboxes = decoded_bboxes_data + n*4*_num_loc_classes*_num_priors;
-                const float *psizes = bbox_sizes_data + n*_num_loc_classes*_num_priors;
-
-                nms_mx(pconf, pboxes, psizes, pbuffer, pindices, pdetections, _num_priors);
-            }
-
-            for (int c = 0; c < _num_classes; ++c) {
-                detections_total += detections_data[n*_num_classes + c];
-            }
-
-            if (_keep_top_k > -1 && detections_total > _keep_top_k) {
-                std::vector<std::pair<float, std::pair<int, int>>> conf_index_class_map;
-
-                for (int c = 0; c < _num_classes; ++c) {
-                    int detections = detections_data[n*_num_classes + c];
-                    int *pindices = indices_data + n*_num_classes*_num_priors + c*_num_priors;
-
-                    float *pconf  = reordered_conf_data + n*_num_classes*_num_priors + c*_num_priors;
-
-                    for (int i = 0; i < detections; ++i) {
-                        int idx = pindices[i];
-                        conf_index_class_map.push_back(std::make_pair(pconf[idx], std::make_pair(c, idx)));
-                    }
-                }
-
-                std::sort(conf_index_class_map.begin(), conf_index_class_map.end(),
-                          SortScorePairDescend<std::pair<int, int>>);
-                conf_index_class_map.resize(_keep_top_k);
-
-                // Store the new indices.
-                memset(detections_data + n*_num_classes, 0, _num_classes * sizeof(int));
-
-                for (size_t j = 0; j < conf_index_class_map.size(); ++j) {
-                    int label = conf_index_class_map[j].second.first;
-                    int idx = conf_index_class_map[j].second.second;
-                    int *pindices = indices_data + n * _num_classes * _num_priors + label * _num_priors;
-                    pindices[detections_data[n*_num_classes + label]] = idx;
-                    detections_data[n*_num_classes + label]++;
-                }
-            }
-        }
-
-        const int num_results = outputs[0]->getTensorDesc().getDims()[2];
-        const int DETECTION_SIZE = outputs[0]->getTensorDesc().getDims()[3];
-        if (DETECTION_SIZE != 7) {
-            return NOT_IMPLEMENTED;
-        }
-
-        int dst_data_size = 0;
-        if (_keep_top_k > 0)
-            dst_data_size = N * _keep_top_k * DETECTION_SIZE * sizeof(float);
-        else if (_top_k > 0)
-            dst_data_size = N * _top_k * _num_classes * DETECTION_SIZE * sizeof(float);
-        else
-            dst_data_size = N * _num_classes * _num_priors * DETECTION_SIZE * sizeof(float);
-
-        if (dst_data_size > outputs[0]->byteSize()) {
-            return OUT_OF_BOUNDS;
-        }
-        memset(dst_data, 0, dst_data_size);
-
-        int count = 0;
-        for (int n = 0; n < N; ++n) {
-            const float *pconf   = reordered_conf_data + n * _num_priors * _num_classes;
-            const float *pboxes  = decoded_bboxes_data + n*_num_priors*4*_num_loc_classes;
-            const int *pindices  = indices_data + n*_num_classes*_num_priors;
-
-            for (int c = 0; c < _num_classes; ++c) {
-                for (int i = 0; i < detections_data[n*_num_classes + c]; ++i) {
-                    int idx = pindices[c*_num_priors + i];
-
-                    dst_data[count * DETECTION_SIZE + 0] = static_cast<float>(n);
-                    dst_data[count * DETECTION_SIZE + 1] = static_cast<float>(_decrease_label_id ? c-1 : c);
-                    dst_data[count * DETECTION_SIZE + 2] = pconf[c*_num_priors + idx];
-
-                    float xmin = _share_location ? pboxes[idx*4 + 0] :
-                                 pboxes[c*4*_num_priors + idx*4 + 0];
-                    float ymin = _share_location ? pboxes[idx*4 + 1] :
-                                 pboxes[c*4*_num_priors + idx*4 + 1];
-                    float xmax = _share_location ? pboxes[idx*4 + 2] :
-                                 pboxes[c*4*_num_priors + idx*4 + 2];
-                    float ymax = _share_location ? pboxes[idx*4 + 3] :
-                                 pboxes[c*4*_num_priors + idx*4 + 3];
-
-                    if (_clip_after_nms) {
-                        xmin = (std::max)(0.0f, (std::min)(1.0f, xmin));
-                        ymin = (std::max)(0.0f, (std::min)(1.0f, ymin));
-                        xmax = (std::max)(0.0f, (std::min)(1.0f, xmax));
-                        ymax = (std::max)(0.0f, (std::min)(1.0f, ymax));
-                    }
-
-                    dst_data[count * DETECTION_SIZE + 3] = xmin;
-                    dst_data[count * DETECTION_SIZE + 4] = ymin;
-                    dst_data[count * DETECTION_SIZE + 5] = xmax;
-                    dst_data[count * DETECTION_SIZE + 6] = ymax;
-
-                    ++count;
-                }
-            }
-        }
-
-        if (count < num_results) {
-            // marker at end of boxes list
-            dst_data[count * DETECTION_SIZE + 0] = -1;
-        }
-
-        return OK;
-    }
-
-private:
-    const int idx_location = 0;
-    const int idx_confidence = 1;
-    const int idx_priors = 2;
-    const int idx_arm_confidence = 3;
-    const int idx_arm_location = 4;
-
-    int _num_classes = 0;
-    int _background_label_id = 0;
-    int _top_k = 0;
-    int _variance_encoded_in_target = 0;
-    int _keep_top_k = 0;
-    int _code_type = 0;
-
-    bool _share_location    = false;
-    bool _clip_before_nms   = false;  // clip bounding boxes before nms step
-    bool _clip_after_nms    = false;  // clip bounding boxes after nms step
-    bool _decrease_label_id = false;
-
-    bool with_add_box_pred = false;
-
-    int _image_width = 0;
-    int _image_height = 0;
-    int _prior_size = 4;
-    bool _normalized = true;
-    int _offset = 0;
-
-    float _nms_threshold = 0.0f;
-    float _confidence_threshold = 0.0f;
-    float _objectness_score = 0.0f;
-
-    int _num = 0;
-    int _num_loc_classes = 0;
-    int _num_priors = 0;
-    bool _priors_batches = false;
-
-    enum CodeType {
-        CORNER = 1,
-        CENTER_SIZE = 2,
-    };
-
-    void decodeBBoxes(const float *prior_data, const float *loc_data, const float *variance_data,
-                      float *decoded_bboxes, float *decoded_bbox_sizes, int* num_priors_actual, int n, const int& offs, const int& pr_size,
-                      bool decodeType = true); // after ARM = false
-
-    void nms_cf(const float *conf_data, const float *bboxes, const float *sizes,
-                int *buffer, int *indices, int &detections, int num_priors_actual);
-
-    void nms_mx(const float *conf_data, const float *bboxes, const float *sizes,
-                int *buffer, int *indices, int *detections, int num_priors_actual);
-
-    std::vector<float> _decoded_bboxes;
-    std::vector<int> _buffer;
-    std::vector<int> _indices;
-    std::vector<int> _detections_count;
-    std::vector<float> _reordered_conf;
-    std::vector<float> _bbox_sizes;
-    std::vector<int> _num_priors_actual;
-};
-
-struct ConfidenceComparator {
-    explicit ConfidenceComparator(const float* conf_data) : _conf_data(conf_data) {}
-
-    bool operator()(int idx1, int idx2) {
-        if (_conf_data[idx1] > _conf_data[idx2]) return true;
-        if (_conf_data[idx1] < _conf_data[idx2]) return false;
-        return idx1 < idx2;
-    }
-
-    const float* _conf_data;
-};
-
-static inline float JaccardOverlap(const float *decoded_bbox,
-                                   const float *bbox_sizes,
-                                   const int idx1,
-                                   const int idx2) {
-    float xmin1 = decoded_bbox[idx1*4 + 0];
-    float ymin1 = decoded_bbox[idx1*4 + 1];
-    float xmax1 = decoded_bbox[idx1*4 + 2];
-    float ymax1 = decoded_bbox[idx1*4 + 3];
-
-    float xmin2 = decoded_bbox[idx2*4 + 0];
-    float ymin2 = decoded_bbox[idx2*4 + 1];
-    float xmax2 = decoded_bbox[idx2*4 + 2];
-    float ymax2 = decoded_bbox[idx2*4 + 3];
-
-    if (xmin2 > xmax1 || xmax2 < xmin1 || ymin2 > ymax1 || ymax2 < ymin1) {
-        return 0.0f;
-    }
-
-    float intersect_xmin = (std::max)(xmin1, xmin2);
-    float intersect_ymin = (std::max)(ymin1, ymin2);
-    float intersect_xmax = (std::min)(xmax1, xmax2);
-    float intersect_ymax = (std::min)(ymax1, ymax2);
-
-    float intersect_width  = intersect_xmax - intersect_xmin;
-    float intersect_height = intersect_ymax - intersect_ymin;
-
-    if (intersect_width <= 0 || intersect_height <= 0) {
-        return 0.0f;
-    }
-
-    float intersect_size = intersect_width * intersect_height;
-    float bbox1_size = bbox_sizes[idx1];
-    float bbox2_size = bbox_sizes[idx2];
-
-    return intersect_size / (bbox1_size + bbox2_size - intersect_size);
-}
-
-void DetectionOutputImpl::decodeBBoxes(const float *prior_data,
-                                       const float *loc_data,
-                                       const float *variance_data,
-                                       float *decoded_bboxes,
-                                       float *decoded_bbox_sizes,
-                                       int* num_priors_actual,
-                                       int n,
-                                       const int& offs,
-                                       const int& pr_size,
-                                       bool decodeType) {
-    num_priors_actual[n] = _num_priors;
-    if (!_normalized && decodeType) {
-        int num = 0;
-        for (; num < _num_priors; ++num) {
-            float batch_id = prior_data[num * pr_size + 0];
-            if (batch_id == -1.f) {
-                num_priors_actual[n] = num;
-                break;
-            }
-        }
-    }
-    parallel_for(num_priors_actual[n], [&](int p) {
-        float new_xmin = 0.0f;
-        float new_ymin = 0.0f;
-        float new_xmax = 0.0f;
-        float new_ymax = 0.0f;
-
-        float prior_xmin = prior_data[p*pr_size + 0 + offs];
-        float prior_ymin = prior_data[p*pr_size + 1 + offs];
-        float prior_xmax = prior_data[p*pr_size + 2 + offs];
-        float prior_ymax = prior_data[p*pr_size + 3 + offs];
-
-        float loc_xmin = loc_data[4*p*_num_loc_classes + 0];
-        float loc_ymin = loc_data[4*p*_num_loc_classes + 1];
-        float loc_xmax = loc_data[4*p*_num_loc_classes + 2];
-        float loc_ymax = loc_data[4*p*_num_loc_classes + 3];
-
-        if (!_normalized) {
-            prior_xmin /= _image_width;
-            prior_ymin /= _image_height;
-            prior_xmax /= _image_width;
-            prior_ymax /= _image_height;
-        }
-
-        if (_code_type == CodeType::CORNER) {
-            if (_variance_encoded_in_target) {
-                // variance is encoded in target, we simply need to add the offset predictions.
-                new_xmin = prior_xmin + loc_xmin;
-                new_ymin = prior_ymin + loc_ymin;
-                new_xmax = prior_xmax + loc_xmax;
-                new_ymax = prior_ymax + loc_ymax;
-            } else {
-                new_xmin = prior_xmin + variance_data[p*4 + 0] * loc_xmin;
-                new_ymin = prior_ymin + variance_data[p*4 + 1] * loc_ymin;
-                new_xmax = prior_xmax + variance_data[p*4 + 2] * loc_xmax;
-                new_ymax = prior_ymax + variance_data[p*4 + 3] * loc_ymax;
-            }
-        } else if (_code_type == CodeType::CENTER_SIZE) {
-            float prior_width    =  prior_xmax - prior_xmin;
-            float prior_height   =  prior_ymax - prior_ymin;
-            float prior_center_x = (prior_xmin + prior_xmax) / 2.0f;
-            float prior_center_y = (prior_ymin + prior_ymax) / 2.0f;
-
-            float decode_bbox_center_x, decode_bbox_center_y;
-            float decode_bbox_width, decode_bbox_height;
-
-            if (_variance_encoded_in_target) {
-                // variance is encoded in target, we simply need to restore the offset predictions.
-                decode_bbox_center_x = loc_xmin * prior_width  + prior_center_x;
-                decode_bbox_center_y = loc_ymin * prior_height + prior_center_y;
-                decode_bbox_width  = std::exp(loc_xmax) * prior_width;
-                decode_bbox_height = std::exp(loc_ymax) * prior_height;
-            } else {
-                // variance is encoded in bbox, we need to scale the offset accordingly.
-                decode_bbox_center_x = variance_data[p*4 + 0] * loc_xmin * prior_width + prior_center_x;
-                decode_bbox_center_y = variance_data[p*4 + 1] * loc_ymin * prior_height + prior_center_y;
-                decode_bbox_width    = std::exp(variance_data[p*4 + 2] * loc_xmax) * prior_width;
-                decode_bbox_height   = std::exp(variance_data[p*4 + 3] * loc_ymax) * prior_height;
-            }
-
-            new_xmin = decode_bbox_center_x - decode_bbox_width  / 2.0f;
-            new_ymin = decode_bbox_center_y - decode_bbox_height / 2.0f;
-            new_xmax = decode_bbox_center_x + decode_bbox_width  / 2.0f;
-            new_ymax = decode_bbox_center_y + decode_bbox_height / 2.0f;
-        }
-
-        if (_clip_before_nms) {
-            new_xmin = (std::max)(0.0f, (std::min)(1.0f, new_xmin));
-            new_ymin = (std::max)(0.0f, (std::min)(1.0f, new_ymin));
-            new_xmax = (std::max)(0.0f, (std::min)(1.0f, new_xmax));
-            new_ymax = (std::max)(0.0f, (std::min)(1.0f, new_ymax));
-        }
-
-        decoded_bboxes[p*4 + 0] = new_xmin;
-        decoded_bboxes[p*4 + 1] = new_ymin;
-        decoded_bboxes[p*4 + 2] = new_xmax;
-        decoded_bboxes[p*4 + 3] = new_ymax;
-
-        decoded_bbox_sizes[p] = (new_xmax - new_xmin) * (new_ymax - new_ymin);
-    });
-}
-
-void DetectionOutputImpl::nms_cf(const float* conf_data,
-                          const float* bboxes,
-                          const float* sizes,
-                          int* buffer,
-                          int* indices,
-                          int& detections,
-                          int num_priors_actual) {
-    int count = 0;
-    for (int i = 0; i < num_priors_actual; ++i) {
-        if (conf_data[i] > _confidence_threshold) {
-            indices[count] = i;
-            count++;
-        }
-    }
-
-    int num_output_scores = (_top_k == -1 ? count : (std::min)(_top_k, count));
-
-    std::partial_sort_copy(indices, indices + count,
-                           buffer, buffer + num_output_scores,
-                           ConfidenceComparator(conf_data));
-
-    for (int i = 0; i < num_output_scores; ++i) {
-        const int idx = buffer[i];
-
-        bool keep = true;
-        for (int k = 0; k < detections; ++k) {
-            const int kept_idx = indices[k];
-            float overlap = JaccardOverlap(bboxes, sizes, idx, kept_idx);
-            if (overlap > _nms_threshold) {
-                keep = false;
-                break;
-            }
-        }
-        if (keep) {
-            indices[detections] = idx;
-            detections++;
-        }
-    }
-}
-
-void DetectionOutputImpl::nms_mx(const float* conf_data,
-                          const float* bboxes,
-                          const float* sizes,
-                          int* buffer,
-                          int* indices,
-                          int* detections,
-                          int num_priors_actual) {
-    int count = 0;
-    for (int i = 0; i < num_priors_actual; ++i) {
-        float conf = -1;
-        int id = 0;
-        for (int c = 1; c < _num_classes; ++c) {
-            float temp = conf_data[c*_num_priors + i];
-            if (temp > conf) {
-                conf = temp;
-                id = c;
-            }
-        }
-
-        if (id > 0 && conf >= _confidence_threshold) {
-            indices[count++] = id*_num_priors + i;
-        }
-    }
-
-    int num_output_scores = (_top_k == -1 ? count : (std::min)(_top_k, count));
-
-    std::partial_sort_copy(indices, indices + count,
-                           buffer, buffer + num_output_scores,
-                           ConfidenceComparator(conf_data));
-
-    for (int i = 0; i < num_output_scores; ++i) {
-        const int idx = buffer[i];
-        const int cls = idx/_num_priors;
-        const int prior = idx%_num_priors;
-
-        int &ndetection = detections[cls];
-        int *pindices = indices + cls*_num_priors;
-
-        bool keep = true;
-        for (int k = 0; k < ndetection; ++k) {
-            const int kept_idx = pindices[k];
-            float overlap = 0.0f;
-            if (_share_location) {
-                overlap = JaccardOverlap(bboxes, sizes, prior, kept_idx);
-            } else {
-                overlap = JaccardOverlap(bboxes, sizes, cls*_num_priors + prior, cls*_num_priors + kept_idx);
-            }
-            if (overlap > _nms_threshold) {
-                keep = false;
-                break;
-            }
-        }
-        if (keep) {
-            pindices[ndetection++] = prior;
-        }
-    }
-}
-
-REG_FACTORY_FOR(DetectionOutputImpl, DetectionOutput);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/detectionoutput_onnx.cpp b/inference-engine/src/mkldnn_plugin/nodes/detectionoutput_onnx.cpp
deleted file mode 100644
index fefcee872cea4f..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/detectionoutput_onnx.cpp
+++ /dev/null
@@ -1,402 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <cassert>
-#include <cfloat>
-#include <vector>
-#include <cmath>
-#include <string>
-#include <utility>
-#include <algorithm>
-#include "ie_parallel.hpp"
-#include "common/tensor_desc_creator.h"
-#include <ngraph/op/experimental_detectron_detection_output.hpp>
-
-
-namespace {
-struct Indexer {
-  const std::vector<int> dims_;
-  int total_{1};
-
-  explicit Indexer(const std::vector<int>& dims) : dims_(dims) {
-      total_ = 1;
-      for (size_t i = 0; i < dims_.size(); ++i) {
-          total_ *= dims_[i];
-      }
-  }
-
-  int operator()(const std::vector<int>& idx) const {
-      int flat_idx = 0;
-      assert(idx.size() == dims_.size());
-      for (size_t i = 0; i < dims_.size(); ++i) {
-          assert(0 <= idx[i] && idx[i] < dims_[i]);
-          flat_idx = flat_idx * dims_[i] + idx[i];
-      }
-      assert(flat_idx < total_);
-      return flat_idx;
-  }
-};
-}  // namespace
-
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-static
-void refine_boxes(const float* boxes, const float* deltas, const float* weights, const float* scores,
-                  float* refined_boxes, float* refined_boxes_areas, float* refined_scores,
-                  const int rois_num, const int classes_num,
-                  const float img_H, const float img_W,
-                  const float max_delta_log_wh,
-                  float coordinates_offset) {
-    Indexer box_idx({rois_num, 4});
-    Indexer delta_idx({rois_num, classes_num, 4});
-    Indexer score_idx({rois_num, classes_num});
-
-    Indexer refined_box_idx({classes_num, rois_num, 4});
-    Indexer refined_score_idx({classes_num, rois_num});
-
-    for (int roi_idx = 0; roi_idx < rois_num; ++roi_idx) {
-        float x0 = boxes[box_idx({roi_idx, 0})];
-        float y0 = boxes[box_idx({roi_idx, 1})];
-        float x1 = boxes[box_idx({roi_idx, 2})];
-        float y1 = boxes[box_idx({roi_idx, 3})];
-
-        if (x1 - x0 <= 0 || y1 - y0 <= 0) {
-            continue;
-        }
-
-        // width & height of box
-        const float ww = x1 - x0 + coordinates_offset;
-        const float hh = y1 - y0 + coordinates_offset;
-        // center location of box
-        const float ctr_x = x0 + 0.5f * ww;
-        const float ctr_y = y0 + 0.5f * hh;
-
-        for (int class_idx = 1; class_idx < classes_num; ++class_idx) {
-            const float dx = deltas[delta_idx({roi_idx, class_idx, 0})] / weights[0];
-            const float dy = deltas[delta_idx({roi_idx, class_idx, 1})] / weights[1];
-            const float d_log_w = deltas[delta_idx({roi_idx, class_idx, 2})] / weights[2];
-            const float d_log_h = deltas[delta_idx({roi_idx, class_idx, 3})] / weights[3];
-
-            // new center location according to deltas (dx, dy)
-            const float pred_ctr_x = dx * ww + ctr_x;
-            const float pred_ctr_y = dy * hh + ctr_y;
-            // new width & height according to deltas d(log w), d(log h)
-            const float pred_w = std::exp((std::min)(d_log_w, max_delta_log_wh)) * ww;
-            const float pred_h = std::exp((std::min)(d_log_h, max_delta_log_wh)) * hh;
-
-            // update upper-left corner location
-            float x0_new = pred_ctr_x - 0.5f * pred_w;
-            float y0_new = pred_ctr_y - 0.5f * pred_h;
-            // update lower-right corner location
-            float x1_new = pred_ctr_x + 0.5f * pred_w - coordinates_offset;
-            float y1_new = pred_ctr_y + 0.5f * pred_h - coordinates_offset;
-
-            // adjust new corner locations to be within the image region,
-            x0_new = std::max<float>(0.0f, x0_new);
-            y0_new = std::max<float>(0.0f, y0_new);
-            x1_new = std::max<float>(0.0f, x1_new);
-            y1_new = std::max<float>(0.0f, y1_new);
-
-            // recompute new width & height
-            const float box_w = x1_new - x0_new + coordinates_offset;
-            const float box_h = y1_new - y0_new + coordinates_offset;
-
-            refined_boxes[refined_box_idx({class_idx, roi_idx, 0})] = x0_new;
-            refined_boxes[refined_box_idx({class_idx, roi_idx, 1})] = y0_new;
-            refined_boxes[refined_box_idx({class_idx, roi_idx, 2})] = x1_new;
-            refined_boxes[refined_box_idx({class_idx, roi_idx, 3})] = y1_new;
-
-            refined_boxes_areas[refined_score_idx({class_idx, roi_idx})] = box_w * box_h;
-
-            refined_scores[refined_score_idx({class_idx, roi_idx})] = scores[score_idx({roi_idx, class_idx})];
-        }
-    }
-}
-
-template <typename T>
-static bool SortScorePairDescend(const std::pair<float, T>& pair1,
-                                 const std::pair<float, T>& pair2) {
-    return pair1.first > pair2.first;
-}
-
-
-struct ConfidenceComparator {
-    explicit ConfidenceComparator(const float* conf_data) : _conf_data(conf_data) {}
-
-    bool operator()(int idx1, int idx2) {
-        if (_conf_data[idx1] > _conf_data[idx2]) return true;
-        if (_conf_data[idx1] < _conf_data[idx2]) return false;
-        return idx1 < idx2;
-    }
-
-    const float* _conf_data;
-};
-
-static inline float JaccardOverlap(const float *decoded_bbox,
-                                   const float *bbox_sizes,
-                                   const int idx1,
-                                   const int idx2,
-                                   const float coordinates_offset = 1) {
-    float xmin1 = decoded_bbox[idx1 * 4 + 0];
-    float ymin1 = decoded_bbox[idx1 * 4 + 1];
-    float xmax1 = decoded_bbox[idx1 * 4 + 2];
-    float ymax1 = decoded_bbox[idx1 * 4 + 3];
-
-    float xmin2 = decoded_bbox[idx2 * 4 + 0];
-    float ymin2 = decoded_bbox[idx2 * 4 + 1];
-    float ymax2 = decoded_bbox[idx2 * 4 + 3];
-    float xmax2 = decoded_bbox[idx2 * 4 + 2];
-
-    if (xmin2 > xmax1 || xmax2 < xmin1 || ymin2 > ymax1 || ymax2 < ymin1) {
-        return 0.0f;
-    }
-
-    float intersect_xmin = (std::max)(xmin1, xmin2);
-    float intersect_ymin = (std::max)(ymin1, ymin2);
-    float intersect_xmax = (std::min)(xmax1, xmax2);
-    float intersect_ymax = (std::min)(ymax1, ymax2);
-
-    float intersect_width  = intersect_xmax - intersect_xmin + coordinates_offset;
-    float intersect_height = intersect_ymax - intersect_ymin + coordinates_offset;
-
-    if (intersect_width <= 0 || intersect_height <= 0) {
-        return 0.0f;
-    }
-
-    float intersect_size = intersect_width * intersect_height;
-    float bbox1_size = bbox_sizes[idx1];
-    float bbox2_size = bbox_sizes[idx2];
-
-    return intersect_size / (bbox1_size + bbox2_size - intersect_size);
-}
-
-
-static void nms_cf(const float* conf_data,
-                          const float* bboxes,
-                          const float* sizes,
-                          int* buffer,
-                          int* indices,
-                          int& detections,
-                          const int boxes_num,
-                          const int pre_nms_topn,
-                          const int post_nms_topn,
-                          const float confidence_threshold,
-                          const float nms_threshold) {
-    int count = 0;
-    for (int i = 0; i < boxes_num; ++i) {
-        if (conf_data[i] > confidence_threshold) {
-            indices[count] = i;
-            count++;
-        }
-    }
-
-    int num_output_scores = (pre_nms_topn == -1 ? count : (std::min)(pre_nms_topn, count));
-
-    std::partial_sort_copy(indices, indices + count,
-                           buffer, buffer + num_output_scores,
-                           ConfidenceComparator(conf_data));
-
-    detections = 0;
-    for (int i = 0; i < num_output_scores; ++i) {
-        const int idx = buffer[i];
-
-        bool keep = true;
-        for (int k = 0; k < detections; ++k) {
-            const int kept_idx = indices[k];
-            float overlap = JaccardOverlap(bboxes, sizes, idx, kept_idx);
-            if (overlap > nms_threshold) {
-                keep = false;
-                break;
-            }
-        }
-        if (keep) {
-            indices[detections] = idx;
-            detections++;
-        }
-    }
-
-    detections = (post_nms_topn == -1 ? detections : (std::min)(post_nms_topn, detections));
-}
-
-
-class ExperimentalDetectronDetectionOutputImpl: public ExtLayerBase {
-private:
-    const int INPUT_ROIS {0};
-    const int INPUT_DELTAS {1};
-    const int INPUT_SCORES {2};
-    const int INPUT_IM_INFO {3};
-
-    const int OUTPUT_BOXES {0};
-    const int OUTPUT_CLASSES {1};
-    const int OUTPUT_SCORES {2};
-
-public:
-    bool isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            auto doOp = ngraph::as_type_ptr<const ngraph::op::v6::ExperimentalDetectronDetectionOutput>(op);
-            if (!doOp) {
-                errorMessage = "Node is not an instance of the ExperimentalDetectronDetectionOutput from the operations set v6.";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    explicit ExperimentalDetectronDetectionOutputImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-            auto doOp = ngraph::as_type_ptr<const ngraph::op::v6::ExperimentalDetectronDetectionOutput>(op);
-            auto attributes = doOp->get_attrs();
-
-            score_threshold_ = attributes.score_threshold;
-            nms_threshold_ = attributes.nms_threshold;
-            max_delta_log_wh_ = attributes.max_delta_log_wh;
-            classes_num_ = attributes.num_classes;
-            max_detections_per_class_ = attributes.post_nms_count;
-            max_detections_per_image_ = attributes.max_detections_per_image;
-            class_agnostic_box_regression_ = attributes.class_agnostic_box_regression;
-            deltas_weights_ = attributes.deltas_weights;
-
-            std::vector<DataConfigurator> inDataConfigurators(op->get_input_size(), {TensorDescCreatorTypes::ncsp, Precision::FP32});
-
-            addConfig(op, inDataConfigurators,
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                           {TensorDescCreatorTypes::ncsp, Precision::I32},
-                           {TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-            throw;
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs,
-                       ResponseDesc *resp) noexcept override {
-        const int rois_num = inputs[INPUT_ROIS]->getTensorDesc().getDims()[0];
-        assert(classes_num_ == static_cast<int>(inputs[INPUT_SCORES]->getTensorDesc().getDims()[1]));
-        assert(4 * classes_num_ == static_cast<int>(inputs[INPUT_DELTAS]->getTensorDesc().getDims()[1]));
-
-        const auto* boxes = inputs[INPUT_ROIS]->buffer().as<const float *>();
-        const auto* deltas = inputs[INPUT_DELTAS]->buffer().as<const float *>();
-        const auto* scores = inputs[INPUT_SCORES]->buffer().as<const float *>();
-        const auto* im_info = inputs[INPUT_IM_INFO]->buffer().as<const float *>();
-
-        auto* output_boxes = outputs[OUTPUT_BOXES]->buffer().as<float *>();
-        auto* output_scores = outputs[OUTPUT_SCORES]->buffer().as<float *>();
-        auto* output_classes = outputs[OUTPUT_CLASSES]->buffer().as<int32_t *>();
-
-        const float img_H = im_info[0];
-        const float img_W = im_info[1];
-
-        // Apply deltas.
-        std::vector<float> refined_boxes(classes_num_ * rois_num * 4, 0);
-        std::vector<float> refined_scores(classes_num_ * rois_num, 0);
-        std::vector<float> refined_boxes_areas(classes_num_ * rois_num, 0);
-        Indexer refined_box_idx({classes_num_, rois_num, 4});
-        Indexer refined_score_idx({classes_num_, rois_num});
-
-        refine_boxes(boxes, deltas, &deltas_weights_[0], scores,
-                     &refined_boxes[0], &refined_boxes_areas[0], &refined_scores[0],
-                     rois_num, classes_num_,
-                     img_H, img_W,
-                     max_delta_log_wh_,
-                     1.0f);
-
-        // Apply NMS class-wise.
-        std::vector<int> buffer(rois_num, 0);
-        std::vector<int> indices(classes_num_ * rois_num, 0);
-        std::vector<int> detections_per_class(classes_num_, 0);
-        int total_detections_num = 0;
-
-        for (int class_idx = 1; class_idx < classes_num_; ++class_idx) {
-            nms_cf(&refined_scores[refined_score_idx({class_idx, 0})],
-                   &refined_boxes[refined_box_idx({class_idx, 0, 0})],
-                   &refined_boxes_areas[refined_score_idx({class_idx, 0})],
-                   &buffer[0],
-                   &indices[total_detections_num],
-                   detections_per_class[class_idx],
-                   rois_num,
-                   -1,
-                   max_detections_per_class_,
-                   score_threshold_,
-                   nms_threshold_);
-            total_detections_num += detections_per_class[class_idx];
-        }
-
-        // Leave only max_detections_per_image_ detections.
-        // confidence, <class, index>
-        std::vector<std::pair<float, std::pair<int, int>>> conf_index_class_map;
-
-        int indices_offset = 0;
-        for (int c = 0; c < classes_num_; ++c) {
-            int n = detections_per_class[c];
-            for (int i = 0; i < n; ++i) {
-                int idx = indices[indices_offset + i];
-                float score = refined_scores[refined_score_idx({c, idx})];
-                conf_index_class_map.push_back(std::make_pair(score, std::make_pair(c, idx)));
-            }
-            indices_offset += n;
-        }
-
-        assert(max_detections_per_image_ > 0);
-        if (total_detections_num > max_detections_per_image_) {
-            std::partial_sort(conf_index_class_map.begin(),
-                              conf_index_class_map.begin() + max_detections_per_image_,
-                              conf_index_class_map.end(),
-                              SortScorePairDescend<std::pair<int, int>>);
-            conf_index_class_map.resize(max_detections_per_image_);
-            total_detections_num = max_detections_per_image_;
-        }
-
-        // Fill outputs.
-        memset(output_boxes, 0, max_detections_per_image_ * 4 * sizeof(output_boxes[0]));
-        memset(output_scores, 0, max_detections_per_image_ * sizeof(output_scores[0]));
-        memset(output_classes, 0, max_detections_per_image_ * sizeof(output_classes[0]));
-
-        int i = 0;
-        for (const auto & detection : conf_index_class_map) {
-            float score = detection.first;
-            int cls = detection.second.first;
-            int idx = detection.second.second;
-            output_boxes[4 * i + 0] = refined_boxes[refined_box_idx({cls, idx, 0})];
-            output_boxes[4 * i + 1] = refined_boxes[refined_box_idx({cls, idx, 1})];
-            output_boxes[4 * i + 2] = refined_boxes[refined_box_idx({cls, idx, 2})];
-            output_boxes[4 * i + 3] = refined_boxes[refined_box_idx({cls, idx, 3})];
-            output_scores[i] = score;
-            output_classes[i] = cls;
-            ++i;
-        }
-
-        return OK;
-    }
-
-private:
-    float score_threshold_;
-    float nms_threshold_;
-    float max_delta_log_wh_;
-    int classes_num_;
-    int max_detections_per_class_;
-    int max_detections_per_image_;
-    bool class_agnostic_box_regression_;
-    std::vector<float> deltas_weights_;
-};
-
-
-
-REG_FACTORY_FOR(ExperimentalDetectronDetectionOutputImpl, ExperimentalDetectronDetectionOutput);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/gather_tree.cpp b/inference-engine/src/mkldnn_plugin/nodes/gather_tree.cpp
deleted file mode 100644
index 4ea74721adca49..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/gather_tree.cpp
+++ /dev/null
@@ -1,184 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-#include <ngraph/op/gather_tree.hpp>
-#include <nodes/common/tensor_desc_creator.h>
-#include <utils/general_utils.h>
-
-#include <cmath>
-#include <limits>
-#include <cfloat>
-#include <string>
-#include <vector>
-#include <cassert>
-#include <algorithm>
-#include "ie_parallel.hpp"
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-class GatherTreeImpl: public ExtLayerBase {
-public:
-    static bool isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            auto gatherElementsOp = ngraph::as_type_ptr<const ngraph::op::v1::GatherTree>(op);
-            if (!gatherElementsOp) {
-                errorMessage = "Node is not an instance of the GatherTree operation from operation set v1.";
-                return false;
-            }
-
-            auto precision = op->get_input_element_type(GATHER_TREE_STEP_IDX);
-            if (!MKLDNNPlugin::one_of(precision, ngraph::element::f32, ngraph::element::i32))
-                precision = ngraph::element::f32;
-            if (op->get_input_element_type(GATHER_TREE_PARENT_IDX) != precision ||
-                    op->get_input_element_type(GATHER_TREE_MAX_SEQ_LEN) != precision ||
-                    op->get_input_element_type(GATHER_TREE_END_TOKEN) != precision ||
-                    op->get_output_element_type(0) != precision) {
-                errorMessage = "Node has incorrect input/output data precision. Must be the same.";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-
-        return true;
-    }
-
-    explicit GatherTreeImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            std::string errorPrefix = std::string("Node GatherTree with name '") + op->get_friendly_name() + "'";
-            if (op->get_input_size() != 4)
-                IE_THROW() << errorPrefix << " has incorrect number of input edges.";
-            if (op->get_output_size() != 1)
-                IE_THROW() << errorPrefix << " has incorrect number of output edges.";
-
-            precision = details::convertPrecision(op->get_input_element_type(GATHER_TREE_STEP_IDX));
-            if (!MKLDNNPlugin::one_of(precision, Precision::FP32, Precision::I32))
-                precision = Precision::FP32;
-
-            if (op->get_input_shape(GATHER_TREE_STEP_IDX).size() != 3)
-                IE_THROW() << errorPrefix << " step_idx vector should be 3 dimension";
-            if (op->get_input_shape(GATHER_TREE_PARENT_IDX).size() != 3)
-                IE_THROW() << errorPrefix << " parent_idx vector should be 3 dimension";
-            if (op->get_input_shape(GATHER_TREE_MAX_SEQ_LEN).size() != 1)
-                IE_THROW() << errorPrefix << " max_seq_len vector should be 1 dimension";
-            if (op->get_input_shape(GATHER_TREE_END_TOKEN).size() != 0)
-                IE_THROW() << errorPrefix << " end_token should be 1 dimension";
-
-            addConfig(op, {{TensorDescCreatorTypes::ncsp, precision},
-                           {TensorDescCreatorTypes::ncsp, precision},
-                           {TensorDescCreatorTypes::ncsp, precision},
-                           {TensorDescCreatorTypes::ncsp, precision}},
-                          {{TensorDescCreatorTypes::ncsp, precision}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-            throw;
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs, ResponseDesc *resp) noexcept override {
-        if (precision == Precision::FP32)
-            return execute_impl<float  >(inputs, outputs, resp);
-        else
-            return execute_impl<int32_t>(inputs, outputs, resp);
-    }
-
-    template<typename DATA_T>
-    StatusCode execute_impl(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs, ResponseDesc *resp) noexcept {
-        const auto *step_idx = inputs[GATHER_TREE_STEP_IDX]->cbuffer().as<DATA_T *>() +
-            inputs[GATHER_TREE_STEP_IDX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        const auto * const parent_idx = inputs[GATHER_TREE_PARENT_IDX]->cbuffer().as<DATA_T *>() +
-            inputs[GATHER_TREE_PARENT_IDX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        const size_t parent_idx_size = inputs[GATHER_TREE_PARENT_IDX]->size()
-            - inputs[GATHER_TREE_PARENT_IDX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        const auto *max_seq_len = inputs[GATHER_TREE_MAX_SEQ_LEN]->cbuffer().as<DATA_T *>() +
-            inputs[GATHER_TREE_MAX_SEQ_LEN]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        auto end_token = (inputs[GATHER_TREE_END_TOKEN]->cbuffer().as<DATA_T *>() +
-            inputs[GATHER_TREE_END_TOKEN]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0];
-        auto * final_idx = outputs[0]->cbuffer().as<DATA_T *>() +
-            outputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-        SizeVector step_idx_dims = inputs[GATHER_TREE_STEP_IDX]->getTensorDesc().getDims();
-        SizeVector parent_idx_dims = inputs[GATHER_TREE_PARENT_IDX]->getTensorDesc().getDims();
-        SizeVector max_seq_len_dims = inputs[GATHER_TREE_MAX_SEQ_LEN]->getTensorDesc().getDims();
-        SizeVector final_idx_dims = outputs[0]->getTensorDesc().getDims();
-        int32_t max_time = step_idx_dims[0];
-        const size_t batch_size = step_idx_dims[1];
-        const size_t beam_width = step_idx_dims[2];
-        const size_t bb_size = batch_size * beam_width;
-
-        if (max_time != static_cast<int32_t>(parent_idx_dims[0]) || max_time != static_cast<int32_t>(final_idx_dims[0]) ||
-            batch_size != parent_idx_dims[1] || batch_size != final_idx_dims[1] || batch_size != max_seq_len_dims[0] ||
-            beam_width != parent_idx_dims[2] || beam_width != final_idx_dims[2]) {
-            if (resp) {
-                std::string errorMsg = "Input/Output tensors dimensions mismatch";
-                errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-            }
-            return PARAMETER_MISMATCH;
-        }
-
-        bool incorrect_result = false;
-        parallel_for2d(batch_size, beam_width, [&](size_t batch, size_t beam) {
-            int32_t max_sequence_in_beam = std::min<int32_t>(max_time, static_cast<int32_t>(max_seq_len[batch]));
-            if (max_sequence_in_beam > 0) {
-                int32_t time, idx = (max_time - 1) * bb_size + batch * beam_width;
-                for (time = (max_time - 1); time >= max_sequence_in_beam; time--, idx -= bb_size)
-                    final_idx[idx + beam] = end_token;
-
-                for (int32_t parent = static_cast<int32_t>(beam); time >= 0; time--, idx -= bb_size) {
-                    if (parent < 0
-                            || parent >= static_cast<int32_t>(beam_width)
-                            || idx + parent >= parent_idx_size) {
-                        incorrect_result = true;
-                        break;
-                    }
-                    final_idx[idx + beam] = step_idx[idx + parent];
-                    parent = static_cast<int32_t>(parent_idx[idx + parent]);
-                }
-
-                bool finished = false;
-                auto *final = &final_idx[batch * beam_width + beam];
-                for (time = 0; time < max_sequence_in_beam; time++, final += bb_size) {
-                    if (finished)
-                        (*final) = end_token;
-                    else if ((*final) == end_token)
-                        finished = true;
-                }
-            }
-        });
-
-        if (incorrect_result) {
-            if (resp) {
-                std::string errorMsg = "Wrong parent index, result is incorrect";
-                errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-            }
-            return OUT_OF_BOUNDS;
-        }
-
-        return OK;
-    }
-
-private:
-    static const size_t GATHER_TREE_STEP_IDX = 0;
-    static const size_t GATHER_TREE_PARENT_IDX = 1;
-    static const size_t GATHER_TREE_MAX_SEQ_LEN = 2;
-    static const size_t GATHER_TREE_END_TOKEN = 3;
-
-    InferenceEngine::Precision precision;
-};
-
-REG_FACTORY_FOR(GatherTreeImpl, GatherTree);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/grn.cpp b/inference-engine/src/mkldnn_plugin/nodes/grn.cpp
deleted file mode 100644
index 6ee077fd52ff1e..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/grn.cpp
+++ /dev/null
@@ -1,91 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <cmath>
-#include <string>
-#include <vector>
-#include "ie_parallel.hpp"
-#include <ngraph/opsets/opset1.hpp>
-
-using namespace MKLDNNPlugin;
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-class GRNImpl: public ExtLayerBase {
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            const auto grn = std::dynamic_pointer_cast<const ngraph::opset1::GRN>(op);
-            if (!grn) {
-                errorMessage = "Only opset1 GRN operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    std::string errorPrefix;
-
-public:
-    explicit GRNImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            errorPrefix = "GRN layer with name '" + op->get_friendly_name() + "'";
-            const auto grn = std::dynamic_pointer_cast<const ngraph::opset1::GRN>(op);
-
-            if (op->get_input_size() != 1 || op->get_output_size() != 1)
-                IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
-
-            bias = grn->get_bias();
-
-            addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32, false, 0}},
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32, false, 0}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs,
-                       ResponseDesc *resp) noexcept override {
-        float* src_data = inputs[0]->buffer();
-        float* dst_data = outputs[0]->buffer();
-
-        SizeVector dims = inputs[0]->getTensorDesc().getDims();
-
-        int N = static_cast<int>((dims.size() > 0) ? dims[0] : 1);
-        int C = static_cast<int>((dims.size() > 1) ? dims[1] : 1);
-        int H = static_cast<int>((dims.size() > 2) ? dims[2] : 1);
-        int W = static_cast<int>((dims.size() > 3) ? dims[3] : 1);
-
-        parallel_for3d(N, H, W, [&](int b, int h, int w) {
-            double variance = 0;
-            for (int c = 0; c < C; c++) {
-                variance += std::pow(src_data[b*C*H*W + c*H*W + h*W + w], 2);
-            }
-            variance = std::pow(variance + bias, 0.5f);
-            for (int c = 0; c < C; c++) {
-                dst_data[b*C*H*W + c*H*W + h*W + w] = src_data[b*C*H*W + c*H*W + h*W + w] / static_cast<float>(variance);
-            }
-        });
-        return OK;
-    }
-
-private:
-    float bias = 1.0f;
-};
-
-REG_FACTORY_FOR(GRNImpl, GRN);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/list_tbl.hpp b/inference-engine/src/mkldnn_plugin/nodes/list_tbl.hpp
index d06cefa7985ac2..d005c1e16b630d 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/list_tbl.hpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/list_tbl.hpp
@@ -7,24 +7,3 @@
 # define MKLDNN_EXTENSION_NODE(__prim, __type)
 #endif
 
-MKLDNN_EXTENSION_NODE(CTCLossImpl, CTCLoss);
-MKLDNN_EXTENSION_NODE(ExperimentalDetectronTopKROIsImpl, ExperimentalDetectronTopKROIs);
-MKLDNN_EXTENSION_NODE(ExtractImagePatchesImpl, ExtractImagePatches);
-MKLDNN_EXTENSION_NODE(ReverseSequenceImpl, ReverseSequence);
-MKLDNN_EXTENSION_NODE(DetectionOutputImpl, DetectionOutput);
-MKLDNN_EXTENSION_NODE(ExperimentalDetectronDetectionOutputImpl, ExperimentalDetectronDetectionOutput);
-MKLDNN_EXTENSION_NODE(LogSoftmaxImpl, LogSoftmax);
-MKLDNN_EXTENSION_NODE(ReorgYoloImpl, ReorgYolo);
-MKLDNN_EXTENSION_NODE(ExperimentalDetectronROIFeatureExtractorImpl, ExperimentalDetectronROIFeatureExtractor);
-MKLDNN_EXTENSION_NODE(ExperimentalDetectronGenerateProposalsSingleImageImpl, ExperimentalDetectronGenerateProposalsSingleImage);
-MKLDNN_EXTENSION_NODE(NonMaxSuppressionImpl, NonMaxSuppressionIEInternal);
-MKLDNN_EXTENSION_NODE(TopKImpl, TopK);
-MKLDNN_EXTENSION_NODE(ExperimentalDetectronPriorGridGeneratorImpl, ExperimentalDetectronPriorGridGenerator);
-MKLDNN_EXTENSION_NODE(GRNImpl, GRN);
-MKLDNN_EXTENSION_NODE(BucketizeImpl, Bucketize);
-MKLDNN_EXTENSION_NODE(CTCGreedyDecoderImpl, CTCGreedyDecoder);
-MKLDNN_EXTENSION_NODE(CTCGreedyDecoderSeqLenImpl, CTCGreedyDecoderSeqLen);
-MKLDNN_EXTENSION_NODE(ProposalImpl, Proposal);
-MKLDNN_EXTENSION_NODE(RangeImpl, Range);
-MKLDNN_EXTENSION_NODE(GatherTreeImpl, GatherTree);
-MKLDNN_EXTENSION_NODE(CumSumImpl, CumSum);
diff --git a/inference-engine/src/mkldnn_plugin/nodes/log_softmax.cpp b/inference-engine/src/mkldnn_plugin/nodes/log_softmax.cpp
deleted file mode 100644
index 337549e3434be0..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/log_softmax.cpp
+++ /dev/null
@@ -1,136 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <cmath>
-#include <limits>
-#include <cfloat>
-#include <string>
-#include <vector>
-#include <cassert>
-#include "ie_parallel.hpp"
-#include <ngraph/opsets/opset5.hpp>
-
-using namespace MKLDNNPlugin;
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-class LogSoftmaxImpl: public ExtLayerBase {
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            const auto logSoftMax = std::dynamic_pointer_cast<const ngraph::opset5::LogSoftmax>(op);
-            if (!logSoftMax) {
-                errorMessage = "Only opset5 LogSoftmax operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-public:
-    explicit LogSoftmaxImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            errorPrefix = "LogSoftmax layer with name '" + op->get_friendly_name() + "'";
-            const auto logSoftMax = std::dynamic_pointer_cast<const ngraph::opset5::LogSoftmax>(op);
-
-            if (op->get_input_size() != 1 || op->get_output_size() != 1)
-                IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
-
-            SizeVector dims = op->get_input_shape(0);
-            if (!dims.size())
-                dims = SizeVector(1, 1);
-            int axis = logSoftMax->get_axis();
-            if (axis < 0)
-                axis += dims.size();
-
-            if (dims.size() < static_cast<size_t>((size_t)(1) + axis))
-                IE_THROW() << errorPrefix << " has incorrect input parameters dimensions and axis number!";
-
-            int j;
-            for (j = dims.size() - 1; j >= 0; j--) {
-                if (dims[j] != 1) break;
-            }
-            if (j == axis) is_last_dim = true;
-
-            for (int i = 0; i < axis; i++)
-                axis_step *= dims[i];
-            reduced_axis_size = dims[axis];
-            for (size_t i = (axis + 1); i < dims.size(); i++)
-                reduced_axis_stride *= dims[i];
-
-            addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs, ResponseDesc *resp) noexcept override {
-        const float *src_data = inputs[0]->cbuffer().as<float *>() +
-            inputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        float* dst_data = outputs[0]->buffer().as<float *>() +
-            outputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-        if (is_last_dim) {
-            parallel_for(axis_step, [&](size_t i) {
-                const float *src_dataPtr = &src_data[i * reduced_axis_size];
-                float *dst_dataPtr = &dst_data[i * reduced_axis_size];
-
-                float reduce_prod = 0.0f;
-                const float max = *std::max_element(src_dataPtr, src_dataPtr + reduced_axis_size);
-                for (size_t j = 0; j < reduced_axis_size; ++j)
-                    reduce_prod += expf(src_dataPtr[j] - max);
-
-                reduce_prod = logf(reduce_prod);
-                for (size_t j = 0; j < reduced_axis_size; ++j)
-                    dst_dataPtr[j] = src_dataPtr[j] - max - reduce_prod;
-            });
-        } else {
-            parallel_for2d(axis_step, reduced_axis_stride, [&](size_t k, size_t i) {
-                const float *src_dataPtr = &src_data[k * reduced_axis_stride * reduced_axis_size + i];
-                float *dst_dataPtr = &dst_data[k * reduced_axis_stride * reduced_axis_size + i];
-
-                float reduce_prod = 0.0f;
-                float max = std::numeric_limits<float>::min();
-                for (size_t j = 0; j < reduced_axis_size; ++j) {
-                    if (src_dataPtr[j * reduced_axis_stride] > max)
-                        max = src_dataPtr[j * reduced_axis_stride];
-                }
-
-                for (size_t j = 0; j < reduced_axis_size; ++j)
-                    reduce_prod += expf(src_dataPtr[j * reduced_axis_stride] - max);
-
-                reduce_prod = logf(reduce_prod);
-                for (size_t j = 0; j < reduced_axis_size; ++j)
-                    dst_dataPtr[j * reduced_axis_stride] = src_dataPtr[j * reduced_axis_stride] - max - reduce_prod;
-            });
-        }
-
-        return OK;
-    }
-
-private:
-    size_t reduced_axis_size;
-    size_t reduced_axis_stride = 1;
-    size_t axis_step = 1;
-    bool is_last_dim = false;
-
-    std::string errorPrefix;
-};
-
-REG_FACTORY_FOR(LogSoftmaxImpl, LogSoftmax);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_bucketize_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_bucketize_node.cpp
new file mode 100644
index 00000000000000..c6c327a1993f3d
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_bucketize_node.cpp
@@ -0,0 +1,218 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+#include <algorithm>
+
+#include <ngraph/opsets/opset3.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_bucketize_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNBucketizeNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto bucketsize = std::dynamic_pointer_cast<const ngraph::opset3::Bucketize>(op);
+        if (!bucketsize) {
+            errorMessage = "Only opset3 Bucketize operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNBucketizeNode::MKLDNNBucketizeNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+                                     MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "Bucketize layer with name '" + op->get_friendly_name() + "' ";
+    const auto bucketsize = std::dynamic_pointer_cast<const ngraph::opset3::Bucketize>(op);
+
+    if (getOriginalInputsNumber() != 2 || getOriginalOutputsNumber() != 1) {
+        IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
+    }
+
+    // check one attribute
+    with_right = bucketsize->get_with_right_bound();
+
+    // check dimensions of input tensors
+    SizeVector input_tensor_dims = op->get_input_shape(INPUT_TENSOR_PORT);
+    if (input_tensor_dims.size() < 1) {
+        IE_THROW() << errorPrefix << " has incorrect dimensions of the input.";
+    }
+    SizeVector input_bin_dims = op->get_input_shape(INPUT_BINS_PORT);
+    if (input_bin_dims.size() != 1) {
+        IE_THROW() << errorPrefix << " has incorrect dimensions of the boundaries tensor.";
+    }
+    if (input_bin_dims[0] != 0) {
+        with_bins = true;
+    }
+    num_bin_values = input_bin_dims[0];
+
+    num_values = std::accumulate(input_tensor_dims.begin(), input_tensor_dims.end(), size_t(1), std::multiplies<size_t>());
+}
+
+void MKLDNNBucketizeNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    // check precisions for input and output tensors
+    input_precision = getOriginalInputPrecisionAtPort(INPUT_TENSOR_PORT);
+    if (input_precision != Precision::FP32 && input_precision != Precision::I32 &&
+        input_precision != Precision::I64) {
+        input_precision = Precision::FP32;
+    }
+    boundaries_precision = getOriginalInputPrecisionAtPort(INPUT_BINS_PORT);
+    if (boundaries_precision != Precision::FP32 && boundaries_precision != Precision::I32 &&
+        boundaries_precision != Precision::I64) {
+        boundaries_precision = Precision::FP32;
+    }
+    output_precision = getOriginalOutputPrecisionAtPort(OUTPUT_TENSOR_PORT);
+    if (output_precision != Precision::I32 && output_precision != Precision::I64) {
+        output_precision = Precision::I32;
+    }
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, input_precision},
+                          {TensorDescCreatorTypes::ncsp, boundaries_precision}},
+                         {{TensorDescCreatorTypes::ncsp, output_precision}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNBucketizeNode::execute(mkldnn::stream strm) {
+    auto precision_mask = getPrecisionMask(input_precision, boundaries_precision, output_precision);
+
+    switch (precision_mask) {
+        case getPrecisionMask(Precision::FP32, Precision::FP32, Precision::I32):
+            bucketize<PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type>();
+            break;
+        case getPrecisionMask(Precision::FP32, Precision::FP32, Precision::I64):
+            bucketize<PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type>();
+            break;
+        case getPrecisionMask(Precision::FP32, Precision::I32, Precision::I32):
+            bucketize<PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type>();
+            break;
+        case getPrecisionMask(Precision::FP32, Precision::I32, Precision::I64):
+            bucketize<PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type>();
+            break;
+        case getPrecisionMask(Precision::FP32, Precision::I64, Precision::I32):
+            bucketize<PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type>();
+            break;
+        case getPrecisionMask(Precision::FP32, Precision::I64, Precision::I64):
+            bucketize<PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I32, Precision::FP32, Precision::I32):
+            bucketize<PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I32, Precision::FP32, Precision::I64):
+            bucketize<PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I32, Precision::I32, Precision::I32):
+            bucketize<PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I32, Precision::I32, Precision::I64):
+            bucketize<PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I32, Precision::I64, Precision::I32):
+            bucketize<PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I32, Precision::I64, Precision::I64):
+            bucketize<PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I64, Precision::FP32, Precision::I32):
+            bucketize<PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I64, Precision::FP32, Precision::I64):
+            bucketize<PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::FP32>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I64, Precision::I32, Precision::I32):
+            bucketize<PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I64, Precision::I32, Precision::I64):
+            bucketize<PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I64, Precision::I64, Precision::I32):
+            bucketize<PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::I32>::value_type>();
+            break;
+        case getPrecisionMask(Precision::I64, Precision::I64, Precision::I64):
+            bucketize<PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type,
+                    PrecisionTrait<Precision::I64>::value_type>();
+            break;
+        default:
+            IE_THROW() << errorPrefix << " has unsupported precision: " << precision_mask;
+    }
+}
+
+template <typename T, typename T_BOUNDARIES, typename T_IND>
+void MKLDNNBucketizeNode::bucketize() {
+    const auto *input_data = reinterpret_cast<const T *>(getParentEdgeAt(0)->getMemoryPtr()->GetPtr());
+    const auto *boundaries_data = reinterpret_cast<const T_BOUNDARIES *>(getParentEdgeAt(1)->getMemoryPtr()->GetPtr());
+    auto *output_data = reinterpret_cast<T_IND *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+
+    if (!with_bins) {
+        memset(output_data, 0, num_values * sizeof(T_IND));
+        return;
+    }
+
+    // boundaries are assumed to be sorted and to have unique elements
+    parallel_for(num_values, [&](size_t ind) {
+        T value = input_data[ind];
+        if (with_right) {
+            auto low = std::lower_bound(boundaries_data, boundaries_data + num_bin_values, value);
+            output_data[ind] = static_cast<T_IND>(low - boundaries_data);
+        } else {
+            auto up = std::upper_bound(boundaries_data, boundaries_data + num_bin_values, value);
+            output_data[ind] = static_cast<T_IND>(up - boundaries_data);
+        }
+    });
+}
+
+bool MKLDNNBucketizeNode::created() const {
+    return getType() == Bucketize;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNBucketizeNode, Bucketize)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_bucketize_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_bucketize_node.h
new file mode 100644
index 00000000000000..472e6aee3cfb03
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_bucketize_node.h
@@ -0,0 +1,43 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNBucketizeNode : public MKLDNNNode {
+public:
+    MKLDNNBucketizeNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    template <typename T, typename T_BOUNDARIES, typename T_IND>
+    void bucketize();
+
+    const size_t INPUT_TENSOR_PORT = 0;
+    const size_t INPUT_BINS_PORT = 1;
+    const size_t OUTPUT_TENSOR_PORT = 0;
+
+    size_t num_values = 0;
+    size_t num_bin_values = 0;
+    bool with_right = false;
+    bool with_bins = false;
+
+    InferenceEngine::Precision input_precision;
+    InferenceEngine::Precision boundaries_precision;
+    InferenceEngine::Precision output_precision;
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_concat_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_concat_node.cpp
index ba760cae535806..4990a658d61f1c 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_concat_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_concat_node.cpp
@@ -33,7 +33,7 @@ namespace {
 
 bool MKLDNNConcatNode::isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
     try {
-        auto concatOp = ngraph::as_type_ptr<const ngraph::op::v0::Concat>(op);
+        const auto concatOp = ngraph::as_type_ptr<const ngraph::op::v0::Concat>(op);
         if (!concatOp) {
             errorMessage = "Node is not an instance of the Concat operation.";
             return false;
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_node.cpp
new file mode 100644
index 00000000000000..34c9aaf191e697
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_node.cpp
@@ -0,0 +1,167 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+
+#include <ngraph/op/ctc_greedy_decoder.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_ctc_greedy_decoder_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNCTCGreedyDecoderNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto greedyDecOp = ngraph::as_type_ptr<const ngraph::op::v0::CTCGreedyDecoder>(op);
+        if (!greedyDecOp) {
+            errorMessage = "Node is not an instance of the CTCGreedyDecoder operation from operation set v0.";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNCTCGreedyDecoderNode::MKLDNNCTCGreedyDecoderNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "CTCGreedyDecoder layer with name '" + op->get_friendly_name() + "' ";
+    if (getOriginalInputsNumber() != 2)
+        IE_THROW() << errorPrefix << "has invalid number of input edges: " << getOriginalInputsNumber();
+    if (getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << "has invalid number of outputs edges: " << getOriginalOutputsNumber();
+
+    if (op->get_input_shape(DATA_INDEX)[0] != op->get_input_shape(SEQUENCE_LENGTH_INDEX)[0] &&
+        op->get_input_shape(DATA_INDEX)[1] != op->get_input_shape(SEQUENCE_LENGTH_INDEX)[1])
+        IE_THROW() << errorPrefix << "has invalid input shapes.";
+
+    auto greedyDecOp = ngraph::as_type_ptr<const ngraph::op::v0::CTCGreedyDecoder>(op);
+    mergeRepeated = greedyDecOp->get_ctc_merge_repeated();
+}
+
+void MKLDNNCTCGreedyDecoderNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    Precision inDataPrecision = getOriginalInputPrecisionAtPort(DATA_INDEX);
+    if (inDataPrecision != Precision::FP32 && inDataPrecision != Precision::BF16)
+        IE_THROW() << errorPrefix << "has unsupported 'data' input precision: " << inDataPrecision;
+
+    Precision seqLenPrecision = getOriginalInputPrecisionAtPort(SEQUENCE_LENGTH_INDEX);
+    if (seqLenPrecision != Precision::FP32 && seqLenPrecision != Precision::BF16)
+        IE_THROW() << errorPrefix << "has unsupported 'sequence_length' input precision: " << seqLenPrecision;
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNCTCGreedyDecoderNode::execute(mkldnn::stream strm) {
+    const float* probabilities = reinterpret_cast<const float *>(getParentEdgeAt(DATA_INDEX)->getMemoryPtr()->GetPtr());
+    const float* sequenceMask = reinterpret_cast<const float *>(getParentEdgeAt(SEQUENCE_LENGTH_INDEX)->getMemoryPtr()->GetPtr());
+    float* outputSequences = reinterpret_cast<float *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+
+    const size_t T = getParentEdgeAt(DATA_INDEX)->getDims()[0];
+    const size_t B = getParentEdgeAt(DATA_INDEX)->getDims()[1];
+    const int C = getParentEdgeAt(DATA_INDEX)->getDims()[2];
+    const size_t BC = B * C;
+    const size_t CB1 = C * (B - 1);
+
+    const int blankIndex = C - 1;
+
+    std::vector<size_t> sequenceLengths(B, 0);
+    parallel_for(B, [&](size_t b) {
+        size_t t = 0;
+        for (; t < T; t++) {
+            if (sequenceMask[B * t + b] == 0.f)
+                break;
+        }
+        sequenceLengths[b] = t;
+    });
+
+    size_t workAmount = 0;
+    for (size_t b = 0; b < B; b++) {
+        workAmount += sequenceLengths[b];
+    }
+
+    // Parallelization could not be made directly by T due to output index depends on merged classes and
+    // blank index, thus could not be shared between threads. Better to divide operation on two steps.
+    // At the first stage find the maximum index. At second stage merge if needed.
+    // Such approach makes parallelization more efficient.
+    auto threadBody = [&](const int ithr, const int nthr) {
+        size_t start(0lu), end(0lu);
+        splitter(workAmount, nthr, ithr, start, end);
+        if (start >= end)
+            return;
+        size_t tStart = 0lu, bStart = 0lu;
+        for (; bStart < B; bStart++) {
+            tStart += sequenceLengths[bStart];
+            if (tStart >= start) {
+                tStart = start - (tStart - sequenceLengths[bStart]);
+                break;
+            }
+        }
+
+        size_t workCounter = start;
+
+        for (size_t b = bStart; b < B; ++b) {
+            size_t outputIndex = b * T + tStart;
+            const float* probs = probabilities + b * C + BC * tStart;
+            size_t sequenceLength = sequenceLengths[b];
+
+            for (size_t t = tStart; t < sequenceLength; ++t) {
+                int maxClassIdx = 0;
+
+                float maxProb = probs[0];
+                ++probs;
+
+                for (int c = 1; c < C; ++c, ++probs) {
+                    if (*probs > maxProb) {
+                        maxClassIdx = c;
+                        maxProb = *probs;
+                    }
+                }
+                probs += CB1;
+                outputSequences[outputIndex++] = static_cast<float>(maxClassIdx);
+
+                if (++workCounter >= end) {
+                    return;
+                }
+            }
+            tStart = 0lu;
+        }
+    }; // thread body
+
+    parallel_nt(0, threadBody);
+
+    parallel_for(B, [&](size_t b) {
+        int prevClassIdx = -1;
+        size_t outputIndex = b * T;
+        const size_t sequenceLength = sequenceLengths[b];
+        float* shiftedOut = outputSequences + b * T;
+        for (size_t t = 0; t < sequenceLength; ++t) {
+            if (*shiftedOut < blankIndex &&
+                !(mergeRepeated && *shiftedOut == prevClassIdx)) {
+                outputSequences[outputIndex++] = *shiftedOut;
+            }
+            prevClassIdx = *shiftedOut;
+            shiftedOut++;
+        }
+        std::fill(outputSequences + outputIndex, outputSequences + (b + 1) * T, -1.f);
+    });
+}
+
+bool MKLDNNCTCGreedyDecoderNode::created() const {
+    return getType() == CTCGreedyDecoder;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNCTCGreedyDecoderNode, CTCGreedyDecoder)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_node.h
new file mode 100644
index 00000000000000..26554ae7333dca
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_node.h
@@ -0,0 +1,32 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNCTCGreedyDecoderNode : public MKLDNNNode {
+public:
+    MKLDNNCTCGreedyDecoderNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    const size_t DATA_INDEX = 0lu;
+    const size_t SEQUENCE_LENGTH_INDEX = 1lu;
+    bool mergeRepeated;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_seq_len_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_seq_len_node.cpp
new file mode 100644
index 00000000000000..0eccdbfa1b5b07
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_seq_len_node.cpp
@@ -0,0 +1,170 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+
+#include <ngraph/op/ctc_greedy_decoder_seq_len.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_ctc_greedy_decoder_seq_len_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNCTCGreedyDecoderSeqLenNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto greedyDecOp = ngraph::as_type_ptr<const ngraph::op::v6::CTCGreedyDecoderSeqLen>(op);
+        if (!greedyDecOp) {
+            errorMessage = "Node is not an instance of the CTCGreedyDecoderSeqLen operation from operation set v6.";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNCTCGreedyDecoderSeqLenNode::MKLDNNCTCGreedyDecoderSeqLenNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "CTCGreedyDecoderSeqLen layer with name '" + op->get_friendly_name() + "' ";
+    if (getOriginalInputsNumber() < 2 || getOriginalInputsNumber() > 3)
+        IE_THROW() << errorPrefix << "has invalid number of input edges: " << getOriginalInputsNumber();
+    if (getOriginalOutputsNumber() != 2)
+        IE_THROW() << errorPrefix << "has invalid number of outputs edges: " << getOriginalOutputsNumber();
+
+    if (op->get_input_shape(DATA_INDEX)[0] != op->get_input_shape(SEQUENCE_LENGTH_INDEX)[0])
+        IE_THROW() << errorPrefix << "has invalid input shapes.";
+
+    auto greedyDecOp = ngraph::as_type_ptr<const ngraph::op::v6::CTCGreedyDecoderSeqLen>(op);
+    mergeRepeated = greedyDecOp->get_merge_repeated();
+}
+
+void MKLDNNCTCGreedyDecoderSeqLenNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    Precision inDataPrecision = getOriginalInputPrecisionAtPort(DATA_INDEX);
+    if (inDataPrecision != Precision::FP32 && inDataPrecision != Precision::BF16)
+        IE_THROW() << errorPrefix << "has unsupported 'data' input precision: " << inDataPrecision;
+
+    Precision seqLenPrecision = getOriginalInputPrecisionAtPort(SEQUENCE_LENGTH_INDEX);
+    if (seqLenPrecision != Precision::I32 && seqLenPrecision != Precision::I64)
+        IE_THROW() << errorPrefix << "has unsupported 'sequence_length' input precision: " << seqLenPrecision;
+
+    std::vector<DataConfigurator> inDataConf;
+    inDataConf.reserve(getOriginalInputsNumber());
+    inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::FP32);
+    for (int i = 1; i < getOriginalInputsNumber(); ++i)
+        inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::I32);
+
+    addSupportedPrimDesc(inDataConf,
+                         {{TensorDescCreatorTypes::ncsp, Precision::I32},
+                          {TensorDescCreatorTypes::ncsp, Precision::I32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNCTCGreedyDecoderSeqLenNode::execute(mkldnn::stream strm) {
+    const float* probabilities = reinterpret_cast<const float *>(getParentEdgeAt(DATA_INDEX)->getMemoryPtr()->GetPtr());
+    const int* sequenceLengths = reinterpret_cast<const int *>(getParentEdgeAt(SEQUENCE_LENGTH_INDEX)->getMemoryPtr()->GetPtr());
+    int* decodedClasses =  reinterpret_cast<int *>(getChildEdgesAtPort(DECODED_CLASSES_INDEX)[0]->getMemoryPtr()->GetPtr());
+    int* decodedClassesLength = reinterpret_cast<int *>(getChildEdgesAtPort(DECODED_CLASSES_LENGTH_INDEX)[0]->getMemoryPtr()->GetPtr());
+
+    const size_t B = getParentEdgeAt(DATA_INDEX)->getDims()[0];;
+    const size_t T = getParentEdgeAt(DATA_INDEX)->getDims()[1];;
+    const int C = getParentEdgeAt(DATA_INDEX)->getDims()[2];;
+    const size_t TC = T * C;
+
+    int blankIndex = C - 1;
+    if (inDims.size() > BLANK_INDEX)
+        blankIndex = (reinterpret_cast<const int  *>(getParentEdgeAt(BLANK_INDEX)->getMemoryPtr()->GetPtr()))[0];
+
+    size_t workAmount = 0;
+    for (size_t b = 0; b < B; b++) {
+        if (sequenceLengths[b] > T) {
+            std::string errorMsg = errorPrefix
+                                   + ". Sequence length " + std::to_string(sequenceLengths[b])
+                                   + " cannot be greater than according decoded classes dimension size "
+                                   + std::to_string(getChildEdgesAtPort(DECODED_CLASSES_INDEX)[0]->getDims()[1]);
+            IE_THROW() << errorMsg;
+        }
+        workAmount += sequenceLengths[b];
+    }
+    // Parallelization could not be made directly by T due to output index depends on merged classes and
+    // blank index, thus could not be shared between threads. Better to divide operation on two steps.
+    // At the first stage find the maximum index. At second stage merge if needed.
+    // Such approach makes parallelization more efficient.
+    auto threadBody = [&](const int ithr, const int nthr) {
+        size_t start(0lu), end(0lu);
+        splitter(workAmount, nthr, ithr, start, end);
+        if (start >= end)
+            return;
+        size_t tStart = 0lu, bStart = 0lu;
+        for (; bStart < B; bStart++) {
+            tStart += sequenceLengths[bStart];
+            if (tStart >= start) {
+                tStart = start - (tStart - sequenceLengths[bStart]);
+                break;
+            }
+        }
+
+        size_t workCounter = start;
+
+        for (size_t b = bStart; b < B; ++b) {
+            size_t outputIndex = b * T + tStart;
+            const float* probs = probabilities + b * TC + C * tStart;
+            const size_t actualSeqLen = sequenceLengths[b];
+
+            for (size_t t = tStart; t < actualSeqLen; ++t) {
+                int maxClassIdx = 0;
+                float maxProb = probs[0];
+                probs++;
+
+                for (int c = 1; c < C; c++, probs++) {
+                    if (*probs > maxProb) {
+                        maxClassIdx = c;
+                        maxProb = *probs;
+                    }
+                }
+                decodedClasses[outputIndex++] = maxClassIdx;
+
+                if (++workCounter >= end) {
+                    return;
+                }
+            }
+            tStart = 0lu;
+        }
+    }; // thread body
+
+    parallel_nt(0, threadBody);
+
+    parallel_for(B, [&](size_t b) {
+        int prevClassIdx = -1;
+        size_t outputIndex = b * T;
+        const size_t actualSeqLen = sequenceLengths[b];
+        int* shiftedOut = decodedClasses + b * T;
+
+        for (size_t t = 0; t < actualSeqLen; ++t) {
+            if (*shiftedOut != blankIndex &&
+                !(mergeRepeated && *shiftedOut == prevClassIdx)) {
+                decodedClasses[outputIndex++] = *shiftedOut;
+            }
+            prevClassIdx = *shiftedOut;
+            shiftedOut++;
+        }
+        std::fill(decodedClasses + outputIndex, decodedClasses + (b + 1) * T, -1);
+        decodedClassesLength[b] = outputIndex - b * T;
+    });
+}
+
+bool MKLDNNCTCGreedyDecoderSeqLenNode::created() const {
+    return getType() == CTCGreedyDecoderSeqLen;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNCTCGreedyDecoderSeqLenNode, CTCGreedyDecoderSeqLen)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_seq_len_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_seq_len_node.h
new file mode 100644
index 00000000000000..b1d5ab6d9ffef3
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_greedy_decoder_seq_len_node.h
@@ -0,0 +1,35 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNCTCGreedyDecoderSeqLenNode : public MKLDNNNode {
+public:
+    MKLDNNCTCGreedyDecoderSeqLenNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    const size_t DATA_INDEX = 0lu;
+    const size_t SEQUENCE_LENGTH_INDEX = 1lu;
+    const size_t BLANK_INDEX = 2lu;
+    const size_t DECODED_CLASSES_INDEX = 0lu;
+    const size_t DECODED_CLASSES_LENGTH_INDEX = 1lu;
+    bool mergeRepeated;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_loss_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_loss_node.cpp
new file mode 100644
index 00000000000000..b355dcaefcd4b0
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_loss_node.cpp
@@ -0,0 +1,279 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <cmath>
+
+#include <ngraph/op/ctc_loss.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_ctc_loss_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNCTCLossNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto ctcLossOp = ngraph::as_type_ptr<const ngraph::op::v4::CTCLoss>(op);
+        if (!ctcLossOp) {
+            errorMessage = "Node is not an instance of the CTCLoss operation from operation set v4.";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNCTCLossNode::MKLDNNCTCLossNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+                                     MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = std::string("CTCLoss layer with name '") + op->get_friendly_name() + "'";
+
+    if (getOriginalInputsNumber() != 4 && getOriginalInputsNumber() != 5)
+        IE_THROW() << errorPrefix << " has invalid inputs number.";
+
+    auto ctcLossOp = ngraph::as_type_ptr<const ngraph::op::v4::CTCLoss>(op);
+    ctcMergeRepeated = ctcLossOp->get_ctc_merge_repeated();
+    preprocessCollapseRepeated = ctcLossOp->get_preprocess_collapse_repeated();
+    unique = ctcLossOp->get_unique();
+}
+
+void MKLDNNCTCLossNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    std::vector<DataConfigurator> inDataConf;
+    inDataConf.reserve(getOriginalInputsNumber());
+    inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::FP32);
+    for (int i = 1; i < getOriginalInputsNumber(); ++i)
+        inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::I32);
+
+    addSupportedPrimDesc(inDataConf,
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNCTCLossNode::execute(mkldnn::stream strm) {
+    StatusCode returnCode = OK;
+
+    const float* logits = reinterpret_cast<const float *>(getParentEdgeAt(0)->getMemoryPtr()->GetPtr());
+    const int* logitsLength = reinterpret_cast<const int *>(getParentEdgeAt(1)->getMemoryPtr()->GetPtr());
+    const int* labels = reinterpret_cast<const int *>(getParentEdgeAt(2)->getMemoryPtr()->GetPtr());
+    const int* labelsLength = reinterpret_cast<const int *>(getParentEdgeAt(3)->getMemoryPtr()->GetPtr());
+    float* dstData = reinterpret_cast<float *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+
+    const size_t batchNum = getParentEdgeAt(0)->getDims()[0];
+    const size_t maxTime = getParentEdgeAt(0)->getDims()[1];
+    const size_t classesNum = getParentEdgeAt(0)->getDims()[2];
+
+    int blankIndex = classesNum - 1;
+    if (inDims.size() > 4) {
+        blankIndex = reinterpret_cast<const int *>(getParentEdgeAt(4)->getMemoryPtr()->GetPtr())[0];
+    }
+
+    std::vector<int> decodedTargetLenB(batchNum, 0);
+    std::vector<std::vector<int>> targetDB(batchNum);
+    std::vector<std::vector<std::vector<float>>> logProbabilitiesB(batchNum);
+    std::vector<std::string> errorMsgB(parallel_get_max_threads());
+
+    auto threadBody_1 = [&](const int ithr, const int nthr) {
+        size_t start(0lu), end(0lu);
+        splitter(batchNum, nthr, ithr, start, end);
+        if (start >= end)
+            return;
+
+        for (size_t b = start; b < end; b++) {
+            if (logitsLength[b] < 0 || labelsLength[b] < 0 || logitsLength[b] > maxTime || labelsLength[b] > logitsLength[b]) {
+                errorMsgB[ithr] = errorPrefix + ". Logit length cannot be greater than max sequence length. "
+                                  + "Label length cannot be greater than a logit length"
+                                  + " and both cannot be negative.\nMaxSeqLen: "
+                                  + std::to_string(maxTime) + "; Logit len: " + std::to_string(logitsLength[b])
+                                  + "; Label len: " + std::to_string(labelsLength[b]);
+                returnCode = GENERAL_ERROR;
+                return;
+            }
+            const size_t actualLogitLen = logitsLength[b];
+            const size_t actualTargetLen = labelsLength[b];
+            size_t decodedTargetLen = 0lu;
+
+            // Decoding target: merge repeated characters if preprocess_collapse_repeated == True,
+            // find unique elemnts if unique == True.
+            // Inserts blanks before each index and a blank at the end.
+            const int* target = &labels[b * maxTime];
+            targetDB[b].resize(actualTargetLen * 2 + 1);
+            auto& targetD = targetDB[b];
+            if (unique) {
+                std::unordered_set<int> uniqVals;
+                for (size_t t = 0lu; t < actualTargetLen; t++) {
+                    if (uniqVals.find(target[t]) != uniqVals.end()) {
+                        continue;
+                    }
+                    uniqVals.insert(target[t]);
+                    targetD[decodedTargetLen++] = blankIndex;
+                    targetD[decodedTargetLen++] = target[t];
+                }
+                targetD[decodedTargetLen++] = blankIndex;
+            } else if (preprocessCollapseRepeated) {
+                auto prevValue = target[0];
+                targetD[decodedTargetLen++] = blankIndex;
+                targetD[decodedTargetLen++] = target[0];
+                for (size_t t = 1lu; t < actualTargetLen; t++) {
+                    if (target[t] == prevValue) {
+                        continue;
+                    }
+                    targetD[decodedTargetLen++] = blankIndex;
+                    targetD[decodedTargetLen++] = prevValue = target[t];
+                }
+                targetD[decodedTargetLen++] = blankIndex;
+            } else {
+                for (size_t t = 0lu; t < actualTargetLen; t++) {
+                    targetD[decodedTargetLen++] = blankIndex;
+                    targetD[decodedTargetLen++] = target[t];
+                }
+                targetD[decodedTargetLen++] = blankIndex;
+            }
+            decodedTargetLenB[b] = decodedTargetLen;
+
+            auto& logProbabilities = logProbabilitiesB[b];
+            logProbabilities.resize(actualLogitLen);
+            for (size_t ll = 0; ll < actualLogitLen; ll++) {
+                logProbabilities[ll].resize(decodedTargetLen);
+            }
+        } // for batch
+    }; // threadBody_1
+
+    parallel_nt(0, threadBody_1);
+    if (returnCode != OK) {
+        std::string resErr("");
+        for (auto& err : errorMsgB) {
+            if (!err.empty())
+                resErr += err + "\n";
+        }
+        IE_THROW() << resErr;
+    }
+
+    const size_t TC = maxTime * classesNum;
+
+    size_t workAmount2 = 0lu;
+    for (size_t b = 0; b < batchNum; b++) {
+        workAmount2 += logitsLength[b];
+    }
+
+    auto threadBody_2 = [&](const int ithr, const int nthr) {
+        size_t start(0lu), end(0lu);
+        size_t sB(0lu), sT(0lu);
+        splitter(workAmount2, nthr, ithr, start, end);
+        if (start >= end)
+            return;
+        int64_t cw = 0, st = start;
+        for (; sB < batchNum; sB++) {
+            cw += logitsLength[sB];
+            if (cw >= st) {
+                sT = logitsLength[sB] + st - cw;
+                break;
+            }
+        }
+        size_t workCounter = start;
+
+        for (size_t b = sB; b < batchNum; b++) {
+            const size_t actualLogitLen = logitsLength[b];
+            const size_t decodedTargetLen = decodedTargetLenB[b];
+            auto& logProbabilities = logProbabilitiesB[b];
+            auto& targetD = targetDB[b];
+
+            double expSum = 0.0;
+            size_t btcT = b * TC + sT * classesNum;
+            // logProbabilities = logSoftmax = logits[b][t][c] - ln(sum_c(exp(logits[b][t])))
+            for (size_t t = sT; t < actualLogitLen; t++) {
+                expSum = 0.0;
+                for (size_t c = 0lu; c < classesNum; c++) {
+                    expSum += std::exp(logits[btcT + c]);
+                }
+                for (size_t s = 0lu; s < decodedTargetLen; s++) {
+                    logProbabilities[t][s] = logits[btcT + targetD[s]] - std::log(expSum);
+                }
+                btcT += classesNum;
+                if (++workCounter >= end) {
+                    return;
+                }
+            }
+            sT = 0lu;
+        }  // for batch
+    }; // threadBody_2
+
+    parallel_nt(0, threadBody_2);
+
+    const auto float_inf = std::numeric_limits<float>::infinity();
+
+    auto sumLogs = [&float_inf](float log1, float log2) {
+        if (log1 == -float_inf) {
+            return log2;
+        } else if (log2 == -float_inf) {
+            return log1;
+        } else {
+            if (log1 > log2)
+                return log1 + std::log1pf(std::exp(log2 - log1));
+            else
+                return log2 + std::log1pf(std::exp(log1 - log2));
+        }
+    };
+
+    auto threadBody_3 = [&](const int ithr, const int nthr) {
+        size_t start(0lu), end(0lu);
+        splitter(batchNum, nthr, ithr, start, end);
+        if (start >= end)
+            return;
+
+        // As per Connectionist Temporal Classification - Labeling Unsegmented Sequence Data with Recurrent Neural Networks:
+        // Graves et al., 2016, paragraph 4.1 (10)
+        for (size_t b = start; b < end; b++) {
+            auto& targetD = targetDB[b];
+            auto& logProbabilities = logProbabilitiesB[b];
+            const int actualLogitLen = logitsLength[b];
+            const int decodedTargetLen = decodedTargetLenB[b];
+            std::vector<std::vector<float>> logBwd(decodedTargetLen, std::vector<float>(actualLogitLen, -float_inf));
+            for (int s = decodedTargetLen - 2; s < decodedTargetLen; s++)
+                logBwd[s][actualLogitLen - 1] = 0.f;
+
+            for (int t = actualLogitLen - 2; t >= 0; t--) {
+                const int t_1 = t + 1;
+                for (int s = std::max(0, decodedTargetLen - (2 * (actualLogitLen - t)));
+                     s < std::min(decodedTargetLen, 2 * (t_1)); s++) {
+                    if (ctcMergeRepeated || targetD[s] == blankIndex) {
+                        logBwd[s][t] = sumLogs(logBwd[s][t],
+                                               logBwd[s][t_1] + logProbabilities[t_1][s]);
+                    }
+
+                    if (s + 1 < decodedTargetLen) {
+                        logBwd[s][t] = sumLogs(logBwd[s][t],
+                                               logBwd[s + 1][t_1] + logProbabilities[t_1][s + 1]);
+                    }
+
+                    if (s + 2 < decodedTargetLen) {
+                        if (targetD[s] != blankIndex && (!ctcMergeRepeated || (targetD[s] != targetD[s + 2]))) {
+                            logBwd[s][t] = sumLogs(logBwd[s][t],
+                                                   logBwd[s + 2][t_1] + logProbabilities[t_1][s + 2]);
+                        }
+                    }
+                }
+            }
+
+            logBwd[0][0] += logProbabilities[0][0];
+            logBwd[1][0] += logProbabilities[0][(decodedTargetLen > 1) ? 1 : 0];
+
+            dstData[b] = -sumLogs(logBwd[0][0], logBwd[1][0]);
+        } // for batch
+    }; // threadBody_3
+
+    parallel_nt(0, threadBody_3);
+}
+
+bool MKLDNNCTCLossNode::created() const {
+    return getType() == CTCLoss;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNCTCLossNode, CTCLoss)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_loss_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_loss_node.h
new file mode 100644
index 00000000000000..b46ff413e829be
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_ctc_loss_node.h
@@ -0,0 +1,32 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNCTCLossNode : public MKLDNNNode {
+public:
+    MKLDNNCTCLossNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    bool ctcMergeRepeated;
+    bool preprocessCollapseRepeated;
+    bool unique;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_cum_sum_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_cum_sum_node.cpp
new file mode 100644
index 00000000000000..3f6c8f903482ce
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_cum_sum_node.cpp
@@ -0,0 +1,279 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#include "list.hpp"
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+
+#include <ngraph/opsets/opset1.hpp>
+#include <ngraph/opsets/opset3.hpp>
+#include "ie_parallel.hpp"
+#include "ie_precision.hpp"
+#include <ie_ngraph_utils.hpp>
+#include "mkldnn_cum_sum_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNCumSumNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto cumsum = std::dynamic_pointer_cast<const ngraph::opset3::CumSum>(op);
+        if (!cumsum) {
+            errorMessage = "Only opset3 CumSum operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNCumSumNode::MKLDNNCumSumNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "CumSum layer with name '" + op->get_friendly_name() + "' ";
+
+    if ((getOriginalInputsNumber() != numOfInputs && getOriginalInputsNumber() != (numOfInputs - 1)) || getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
+
+    const auto &dataShape = op->get_input_shape(CUM_SUM_DATA);
+    if (dataShape.size() < 1) {
+        IE_THROW() << errorPrefix << " doesn't support 'data' input tensor with rank: " << dataShape.size();
+    }
+    numOfDims = dataShape.size();
+
+    const auto cumsum = std::dynamic_pointer_cast<const ngraph::opset3::CumSum>(op);
+    exclusive = cumsum->is_exclusive();
+    reverse = cumsum->is_reverse();
+
+    if (getOriginalInputsNumber() == numOfInputs) {
+        if (!ngraph::is_scalar(cumsum->get_input_shape(AXIS)))
+            IE_THROW() << errorPrefix << " doesn't support 'axis' input tensor with non scalar rank";
+    }
+
+    if (dataShape != cumsum->get_output_shape(0))
+        IE_THROW() << errorPrefix << " has different 'data' input and output dimensions";
+
+    shape = dataShape;
+}
+
+void MKLDNNCumSumNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    dataPrecision = getOriginalInputPrecisionAtPort(CUM_SUM_DATA);
+    if (dataPrecision != Precision::I8 && dataPrecision != Precision::U8 && dataPrecision != Precision::I16 && dataPrecision != Precision::I32 &&
+        dataPrecision != Precision::FP32 && dataPrecision != Precision::I64 && dataPrecision != Precision::U64 && dataPrecision != Precision::BF16)
+        IE_THROW() << errorPrefix << " has unsupported 'data' input precision: " << dataPrecision.name();
+
+    if (getOriginalInputsNumber() == numOfInputs) {
+        const auto &axisTensorPrec = getOriginalInputPrecisionAtPort(AXIS);
+        if (axisTensorPrec != Precision::I32 && axisTensorPrec != Precision::I64)
+            IE_THROW() << errorPrefix << " has unsupported 'axis' input precision: " << axisTensorPrec.name();
+    }
+
+    std::vector<DataConfigurator> inDataConf;
+    inDataConf.reserve(getOriginalInputsNumber());
+    inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, dataPrecision);
+    for (int i = 1; i < getOriginalInputsNumber(); ++i)
+        inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::I32);
+
+    addSupportedPrimDesc(inDataConf,
+                         {{TensorDescCreatorTypes::ncsp, dataPrecision}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNCumSumNode::execute(mkldnn::stream strm) {
+    if (inDims.size() == numOfInputs)
+        axis = getAxis(getParentEdgeAt(AXIS)->getBlob(), getParentEdgeAt(CUM_SUM_DATA)->getBlob());
+
+    switch (dataPrecision) {
+        case Precision::I8   : {
+            exec<int8_t>();
+            break;
+        }
+        case Precision::U8   : {
+            exec<uint8_t>();
+            break;
+        }
+        case Precision::I16  : {
+            exec<int16_t>();
+            break;
+        }
+        case Precision::I32  : {
+            exec<int32_t>();
+            break;
+        }
+        case Precision::FP32 : {
+            exec<float>();
+            break;
+        }
+        case Precision::I64  : {
+            exec<int64_t>();
+            break;
+        }
+        case Precision::U64  : {
+            exec<uint64_t>();
+            break;
+        }
+        default : {
+            std::string errorMsg = errorPrefix + " has unsupported 'data' input precision: " + dataPrecision.name();
+            IE_THROW() << errorMsg;
+        }
+    }
+}
+
+
+template <typename dataType>
+void MKLDNNCumSumNode::exec() {
+    const auto *input = reinterpret_cast<const dataType *>(getParentEdgeAt(CUM_SUM_DATA)->getMemoryPtr()->GetPtr());
+    auto *output = reinterpret_cast<dataType *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+    const std::vector<size_t> strides = getParentEdgeAt(CUM_SUM_DATA)->getDesc().getBlockingDesc().getStrides();
+
+    if (reverse) {
+        if (exclusive) {
+            cumSum<true, true, dataType>(input, output, strides);
+        } else {
+            cumSum<true, false, dataType>(input, output, strides);
+        }
+    } else {
+        if (exclusive) {
+            cumSum<false, true, dataType>(input, output, strides);
+        } else {
+            cumSum<false, false, dataType>(input, output, strides);
+        }
+    }
+}
+
+template <bool reverse, bool exclusive, typename dataType>
+void MKLDNNCumSumNode::cumSum(const dataType *input, dataType *output, const std::vector<size_t> &strides) {
+    SizeVector iterationRange(numOfDims - 1);
+    size_t j = 0;
+    for (size_t i = 0; i < shape.size(); i++) {
+        if (i == axis)
+            continue;
+        iterationRange[j++] = shape[i];
+    }
+    size_t work_amount_dst = std::accumulate(iterationRange.begin(), iterationRange.end(), 1, std::multiplies<size_t>());
+    parallel_nt(0, [&](const int ithr, const int nthr) {
+        size_t start = 0, end = 0;
+        SizeVector counters(numOfDims - 1, 0);
+        splitter(work_amount_dst, nthr, ithr, start, end);
+
+        parallelItInit(start, counters, iterationRange);
+
+        for (size_t iwork = start; iwork < end; ++iwork) {
+            std::vector<size_t> forStartOffset(numOfDims);
+            forStartOffset[axis] = 0;
+            for (size_t offsetIdx = 0, countersIdx = 0; offsetIdx < numOfDims; ++offsetIdx) {
+                if (offsetIdx == axis) {
+                    continue;
+                }
+                forStartOffset[offsetIdx] = counters[countersIdx++];
+            }
+
+            size_t startOffset = getStartOffset(forStartOffset, strides);
+
+            const dataType *inputStart = input + startOffset;
+            dataType *outputStart = output + startOffset;
+
+            size_t offset = strides[axis];
+            if (reverse) {
+                if (exclusive) {
+                    outputStart[offset*(shape[axis] - 1)] = 0;
+                    for (int64_t i = shape[axis] - 2; i >= 0; i--) {
+                        outputStart[i*offset] = inputStart[(i+1)*offset] + outputStart[(i+1)*offset];
+                    }
+                } else {
+                    outputStart[offset*(shape[axis] - 1)] = inputStart[offset * (shape[axis] - 1)];
+                    for (int64_t i = shape[axis] - 2; i >= 0; i--) {
+                        outputStart[i*offset] = inputStart[i*offset] + outputStart[(i+1)*offset];
+                    }
+                }
+            } else {
+                if (exclusive) {
+                    outputStart[0] = 0;
+                    for (size_t i = 1; i < shape[axis]; i++) {
+                        outputStart[i*offset] = inputStart[(i-1)*offset] + outputStart[(i-1)*offset];
+                    }
+                } else {
+                    outputStart[0] = inputStart[0];
+                    for (size_t i = 1; i < shape[axis]; i++) {
+                        outputStart[i*offset] = inputStart[i*offset] + outputStart[(i-1)*offset];
+                    }
+                }
+            }
+
+            parallelItStep(counters, iterationRange);
+        }
+    });
+}
+
+void MKLDNNCumSumNode::parallelItInit(size_t start, std::vector<size_t>& counters, const std::vector<size_t>& iterationRange) {
+    auto itCounter = counters.rbegin();
+    auto itWork = iterationRange.rbegin();
+    while (itCounter != counters.rend() && itWork != iterationRange.rend()) {
+        *itCounter = start % *itWork;
+        start /= *itWork;
+        ++itCounter;
+        ++itWork;
+    }
+}
+
+inline void MKLDNNCumSumNode::parallelItStep(std::vector<size_t>& counters, const std::vector<size_t>& iterationRange) {
+    auto itCounter = counters.rbegin();
+    auto itWork = iterationRange.rbegin();
+
+    while (itCounter != counters.rend() && itWork != iterationRange.rend()) {
+        *itCounter = (*itCounter + 1) % *itWork;
+        if (*itCounter != 0) {
+            break;
+        }
+        ++itCounter;
+        ++itWork;
+    }
+}
+
+inline size_t MKLDNNCumSumNode::getStartOffset(const std::vector<size_t> &forStartOffset, const std::vector<size_t>& strides) const {
+    size_t startOffset = 0;
+    for (size_t idx = 0; idx < forStartOffset.size(); ++idx) {
+        startOffset += forStartOffset[idx] * strides[idx];
+    }
+    return startOffset;
+}
+
+size_t MKLDNNCumSumNode::getAxis(const Blob::CPtr& _axis, const Blob::CPtr& _data) const {
+    const auto& axisPrecision = _axis->getTensorDesc().getPrecision();
+    const int64_t dataShapeSize = static_cast<int64_t>(_data->getTensorDesc().getDims().size());
+    int64_t axisValueFromBlob;
+    switch (axisPrecision) {
+        case Precision::I32 : {
+            const auto *axisPtr = _axis->cbuffer().as<const int32_t *>();
+            axisValueFromBlob = static_cast<int64_t>(axisPtr[0]);
+            break;
+        }
+        case Precision::I64 : {
+            const auto *axisPtr = _axis->cbuffer().as<const int64_t *>();
+            axisValueFromBlob = axisPtr[0];
+            break;
+        }
+        default : {
+            IE_THROW() << errorPrefix << "  doesn't support 'axis' input with precision: " << axisPrecision.name();
+        }
+    }
+    if (axisValueFromBlob < -dataShapeSize || axisValueFromBlob > dataShapeSize - 1)
+        IE_THROW() << errorPrefix << "  has axis with a value out of range: " << axisValueFromBlob;
+    return axisValueFromBlob >= 0 ? axisValueFromBlob : (axisValueFromBlob + dataShapeSize);
+}
+
+bool MKLDNNCumSumNode::created() const {
+    return getType() == CumSum;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNCumSumNode, CumSum)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_cum_sum_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_cum_sum_node.h
new file mode 100644
index 00000000000000..794d6bc73f1722
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_cum_sum_node.h
@@ -0,0 +1,50 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNCumSumNode : public MKLDNNNode {
+public:
+    MKLDNNCumSumNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    template <typename dataType>
+    void exec();
+
+    template <bool reverse, bool exclusive, typename dataType>
+    void cumSum(const dataType *input, dataType *output, const std::vector<size_t> &strides);
+
+    void parallelItInit(size_t start, std::vector<size_t>& counters, const std::vector<size_t>& iterationRange);
+
+    inline void parallelItStep(std::vector<size_t>& counters, const std::vector<size_t>& iterationRange);
+
+    inline size_t getStartOffset(const std::vector<size_t> &forStartOffset, const std::vector<size_t>& strides) const;
+
+    size_t getAxis(const InferenceEngine::Blob::CPtr& _axis, const InferenceEngine::Blob::CPtr& _data) const;
+
+    enum { CUM_SUM_DATA, AXIS, numOfInputs };
+    bool exclusive;
+    bool reverse;
+    size_t numOfDims;
+    size_t axis = 0;
+    std::vector<size_t> shape;
+
+    InferenceEngine::Precision dataPrecision;
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_def_conv_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_def_conv_node.cpp
index dde4d960c5897e..a2fae182a52f70 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_def_conv_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_def_conv_node.cpp
@@ -741,7 +741,7 @@ struct jit_uni_def_conv_kernel_f32 : public jit_uni_def_conv_kernel, public jit_
 
 bool MKLDNNDeformableConvolutionNode::isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
     try {
-        auto defConvNode = ngraph::as_type_ptr<const ngraph::op::v1::DeformableConvolution>(op);
+        const auto defConvNode = ngraph::as_type_ptr<const ngraph::op::v1::DeformableConvolution>(op);
         if (!defConvNode) {
             errorMessage = "Node is not an instance of DeformableConvolution form the operation set v1.";
             return false;
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_detection_output_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_detection_output_node.cpp
new file mode 100644
index 00000000000000..4b8c695a987315
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_detection_output_node.cpp
@@ -0,0 +1,601 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+
+#include <ngraph/op/detection_output.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_detection_output_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+template <typename T>
+static bool SortScorePairDescend(const std::pair<float, T>& pair1,
+                                 const std::pair<float, T>& pair2) {
+    return pair1.first > pair2.first;
+}
+
+bool MKLDNNDetectionOutputNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto doOp = ngraph::as_type_ptr<const ngraph::op::v0::DetectionOutput>(op);
+        if (!doOp) {
+            errorMessage = "Node is not an instance of the DetectionOutput from the operations set v0.";
+            return false;
+        }
+        if (!details::CaselessEq<std::string>()(doOp->get_attrs().code_type, "caffe.PriorBoxParameter.CENTER_SIZE") &&
+            !details::CaselessEq<std::string>()(doOp->get_attrs().code_type, "caffe.PriorBoxParameter.CORNER")) {
+            errorMessage = "Unsupported code_type attribute: " + doOp->get_attrs().code_type;
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNDetectionOutputNode::MKLDNNDetectionOutputNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "DetectionOutput layer with name '" + op->get_friendly_name() + "' ";
+
+    if (getOriginalInputsNumber() != 3 && getOriginalInputsNumber() != 5)
+        IE_THROW() << errorPrefix <<  " has incorrect number of input edges.";
+
+    if (getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << " has incorrect number of output edges.";
+
+    auto doOp = ngraph::as_type_ptr<const ngraph::op::v0::DetectionOutput>(op);
+    auto attributes = doOp->get_attrs();
+
+    _num_classes = attributes.num_classes;
+    _background_label_id = attributes.background_label_id;
+    _top_k = attributes.top_k;
+    _variance_encoded_in_target = attributes.variance_encoded_in_target;
+    _keep_top_k = attributes.keep_top_k[0];
+    _nms_threshold = attributes.nms_threshold;
+    _confidence_threshold = attributes.confidence_threshold;
+    _share_location = attributes.share_location;
+    _clip_before_nms = attributes.clip_before_nms;
+    _clip_after_nms = attributes.clip_after_nms;
+    _decrease_label_id = attributes.decrease_label_id;
+    _normalized = attributes.normalized;
+    _image_height = attributes.input_height;
+    _image_width = attributes.input_width;
+    _prior_size = _normalized ? 4 : 5;
+    _offset = _normalized ? 0 : 1;
+    _num_loc_classes = _share_location ? 1 : _num_classes;
+
+    with_add_box_pred = getOriginalInputsNumber() == 5;
+    _objectness_score = attributes.objectness_score;
+
+    _code_type = (details::CaselessEq<std::string>()(attributes.code_type, "caffe.PriorBoxParameter.CENTER_SIZE") ?
+                  CodeType::CENTER_SIZE : CodeType::CORNER);
+
+    _num_priors = static_cast<int>(op->get_input_shape(idx_priors).back() / _prior_size);
+    _priors_batches = op->get_input_shape(idx_priors).front() != 1;
+
+    if (_num_priors * _num_loc_classes * 4 != static_cast<int>(op->get_input_shape(idx_location)[1]))
+        IE_THROW() << errorPrefix << " has incorrect number of priors must match number of location predictions ("
+                   << _num_priors * _num_loc_classes * 4 << " vs "
+                   << op->get_input_shape(idx_location)[1] << ")";
+
+    if (_num_priors * _num_classes != static_cast<int>(op->get_input_shape(idx_confidence).back()))
+        IE_THROW() << " has incorrect number of priors must match number of confidence predictions.";
+
+    if (_decrease_label_id && _background_label_id != 0)
+        IE_THROW() << errorPrefix << " cannot use decrease_label_id and background_label_id parameter simultaneously.";
+
+    _num = static_cast<int>(op->get_input_shape(idx_confidence)[0]);
+
+    _decoded_bboxes.resize(_num * _num_classes * _num_priors * 4);
+    _buffer.resize(_num * _num_classes * _num_priors);
+    _indices.resize(_num * _num_classes * _num_priors);
+    _detections_count.resize(_num * _num_classes);
+    _bbox_sizes.resize(_num * _num_classes * _num_priors);
+    _num_priors_actual.resize(_num);
+
+    const auto &confSize = op->get_input_shape(idx_confidence);
+    _reordered_conf.resize(std::accumulate(confSize.begin(), confSize.end(), 1, std::multiplies<size_t>()));
+}
+
+void MKLDNNDetectionOutputNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    std::vector<DataConfigurator> inDataConf;
+    inDataConf.reserve(getOriginalInputsNumber());
+    for (int i = 0; i < getOriginalInputsNumber(); ++i)
+        inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::FP32);
+
+    addSupportedPrimDesc(inDataConf,
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNDetectionOutputNode::execute(mkldnn::stream strm) {
+    float *dst_data = reinterpret_cast<float *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+
+    const float *loc_data    = reinterpret_cast<const float *>(getParentEdgeAt(idx_location)->getMemoryPtr()->GetPtr());
+    const float *conf_data   = reinterpret_cast<const float *>(getParentEdgeAt(idx_confidence)->getMemoryPtr()->GetPtr());
+    const float *prior_data  = reinterpret_cast<const float *>(getParentEdgeAt(idx_priors)->getMemoryPtr()->GetPtr());
+    const float *arm_conf_data = inDims.size() > 3 ?
+            reinterpret_cast<const float *>(getParentEdgeAt(idx_arm_confidence)->getMemoryPtr()->GetPtr()) : nullptr;
+    const float *arm_loc_data = inDims.size() > 4 ?
+            reinterpret_cast<const float *>(getParentEdgeAt(idx_arm_location)->getMemoryPtr()->GetPtr()) : nullptr;
+
+    const int N = getParentEdgeAt(idx_confidence)->getDims()[0];
+
+    float *decoded_bboxes_data = _decoded_bboxes.data();
+    float *reordered_conf_data = _reordered_conf.data();
+    float *bbox_sizes_data     = _bbox_sizes.data();
+    int *detections_data       = _detections_count.data();
+    int *buffer_data           = _buffer.data();
+    int *indices_data          = _indices.data();
+    int *num_priors_actual     = _num_priors_actual.data();
+
+    for (int n = 0; n < N; ++n) {
+        const float *ppriors = prior_data;
+        const float *prior_variances = prior_data + _num_priors*_prior_size;
+        if (_priors_batches) {
+            ppriors += _variance_encoded_in_target ? n*_num_priors*_prior_size : 2*n*_num_priors*_prior_size;
+            prior_variances += _variance_encoded_in_target ? 0 : 2*n*_num_priors*_prior_size;
+        }
+
+        if (_share_location) {
+            const float *ploc = loc_data + n*4*_num_priors;
+            float *pboxes = decoded_bboxes_data + n*4*_num_priors;
+            float *psizes = bbox_sizes_data + n*_num_priors;
+
+            if (with_add_box_pred) {
+                const float *p_arm_loc = arm_loc_data + n*4*_num_priors;
+                decodeBBoxes(ppriors, p_arm_loc, prior_variances, pboxes, psizes, num_priors_actual, n, _offset, _prior_size);
+                decodeBBoxes(pboxes, ploc, prior_variances, pboxes, psizes, num_priors_actual, n, 0, 4, false);
+            } else {
+                decodeBBoxes(ppriors, ploc, prior_variances, pboxes, psizes, num_priors_actual, n, _offset, _prior_size);
+            }
+        } else {
+            for (int c = 0; c < _num_loc_classes; ++c) {
+                if (c == _background_label_id) {
+                    continue;
+                }
+                const float *ploc = loc_data + n*4*_num_loc_classes*_num_priors + c*4;
+                float *pboxes = decoded_bboxes_data + n*4*_num_loc_classes*_num_priors + c*4*_num_priors;
+                float *psizes = bbox_sizes_data + n*_num_loc_classes*_num_priors + c*_num_priors;
+                if (with_add_box_pred) {
+                    const float *p_arm_loc = arm_loc_data + n*4*_num_loc_classes*_num_priors + c*4;
+                    decodeBBoxes(ppriors, p_arm_loc, prior_variances, pboxes, psizes, num_priors_actual, n, _offset, _prior_size);
+                    decodeBBoxes(pboxes, ploc, prior_variances, pboxes, psizes, num_priors_actual, n, 0, 4, false);
+                } else {
+                    decodeBBoxes(ppriors, ploc, prior_variances, pboxes, psizes, num_priors_actual, n, _offset, _prior_size);
+                }
+            }
+        }
+    }
+
+    if (with_add_box_pred) {
+        for (int n = 0; n < N; ++n) {
+            for (int p = 0; p < _num_priors; ++p) {
+                if (arm_conf_data[n*_num_priors*2 + p * 2 + 1] < _objectness_score) {
+                    for (int c = 0; c < _num_classes; ++c) {
+                        reordered_conf_data[n*_num_priors*_num_classes + c*_num_priors + p] = c == _background_label_id ? 1.0f : 0.0f;
+                    }
+                } else {
+                    for (int c = 0; c < _num_classes; ++c) {
+                        reordered_conf_data[n*_num_priors*_num_classes + c*_num_priors + p] = conf_data[n*_num_priors*_num_classes + p*_num_classes + c];
+                    }
+                }
+            }
+        }
+    } else {
+        for (int n = 0; n < N; ++n) {
+            for (int c = 0; c < _num_classes; ++c) {
+                for (int p = 0; p < _num_priors; ++p) {
+                    reordered_conf_data[n*_num_priors*_num_classes + c*_num_priors + p] = conf_data[n*_num_priors*_num_classes + p*_num_classes + c];
+                }
+            }
+        }
+    }
+
+    memset(detections_data, 0, N*_num_classes*sizeof(int));
+
+    for (int n = 0; n < N; ++n) {
+        int detections_total = 0;
+
+        if (!_decrease_label_id) {
+            // Caffe style
+            parallel_for(_num_classes, [&](int c) {
+                if (c != _background_label_id) {  // Ignore background class
+                    int *pindices    = indices_data + n*_num_classes*_num_priors + c*_num_priors;
+                    int *pbuffer     = buffer_data + c*_num_priors;
+                    int *pdetections = detections_data + n*_num_classes + c;
+
+                    const float *pconf = reordered_conf_data + n*_num_classes*_num_priors + c*_num_priors;
+                    const float *pboxes;
+                    const float *psizes;
+                    if (_share_location) {
+                        pboxes = decoded_bboxes_data + n*4*_num_priors;
+                        psizes = bbox_sizes_data + n*_num_priors;
+                    } else {
+                        pboxes = decoded_bboxes_data + n*4*_num_classes*_num_priors + c*4*_num_priors;
+                        psizes = bbox_sizes_data + n*_num_classes*_num_priors + c*_num_priors;
+                    }
+
+                    nms_cf(pconf, pboxes, psizes, pbuffer, pindices, *pdetections, num_priors_actual[n]);
+                }
+            });
+        } else {
+            // MXNet style
+            int *pindices = indices_data + n*_num_classes*_num_priors;
+            int *pbuffer = buffer_data;
+            int *pdetections = detections_data + n*_num_classes;
+
+            const float *pconf = reordered_conf_data + n*_num_classes*_num_priors;
+            const float *pboxes = decoded_bboxes_data + n*4*_num_loc_classes*_num_priors;
+            const float *psizes = bbox_sizes_data + n*_num_loc_classes*_num_priors;
+
+            nms_mx(pconf, pboxes, psizes, pbuffer, pindices, pdetections, _num_priors);
+        }
+
+        for (int c = 0; c < _num_classes; ++c) {
+            detections_total += detections_data[n*_num_classes + c];
+        }
+
+        if (_keep_top_k > -1 && detections_total > _keep_top_k) {
+            std::vector<std::pair<float, std::pair<int, int>>> conf_index_class_map;
+
+            for (int c = 0; c < _num_classes; ++c) {
+                int detections = detections_data[n*_num_classes + c];
+                int *pindices = indices_data + n*_num_classes*_num_priors + c*_num_priors;
+
+                float *pconf  = reordered_conf_data + n*_num_classes*_num_priors + c*_num_priors;
+
+                for (int i = 0; i < detections; ++i) {
+                    int idx = pindices[i];
+                    conf_index_class_map.push_back(std::make_pair(pconf[idx], std::make_pair(c, idx)));
+                }
+            }
+
+            std::sort(conf_index_class_map.begin(), conf_index_class_map.end(),
+                      SortScorePairDescend<std::pair<int, int>>);
+            conf_index_class_map.resize(_keep_top_k);
+
+            // Store the new indices.
+            memset(detections_data + n*_num_classes, 0, _num_classes * sizeof(int));
+
+            for (size_t j = 0; j < conf_index_class_map.size(); ++j) {
+                int label = conf_index_class_map[j].second.first;
+                int idx = conf_index_class_map[j].second.second;
+                int *pindices = indices_data + n * _num_classes * _num_priors + label * _num_priors;
+                pindices[detections_data[n*_num_classes + label]] = idx;
+                detections_data[n*_num_classes + label]++;
+            }
+        }
+    }
+
+    const int num_results = getChildEdgesAtPort(0)[0]->getDims()[2];
+    const int DETECTION_SIZE = getChildEdgesAtPort(0)[0]->getDims()[3];
+    if (DETECTION_SIZE != 7) {
+        IE_THROW() << NOT_IMPLEMENTED;
+    }
+
+    int dst_data_size = 0;
+    if (_keep_top_k > 0)
+        dst_data_size = N * _keep_top_k * DETECTION_SIZE * sizeof(float);
+    else if (_top_k > 0)
+        dst_data_size = N * _top_k * _num_classes * DETECTION_SIZE * sizeof(float);
+    else
+        dst_data_size = N * _num_classes * _num_priors * DETECTION_SIZE * sizeof(float);
+
+    if (dst_data_size > getChildEdgesAtPort(0)[0]->getBlob()->byteSize()) {
+        IE_THROW() << OUT_OF_BOUNDS;
+    }
+    memset(dst_data, 0, dst_data_size);
+
+    int count = 0;
+    for (int n = 0; n < N; ++n) {
+        const float *pconf   = reordered_conf_data + n * _num_priors * _num_classes;
+        const float *pboxes  = decoded_bboxes_data + n*_num_priors*4*_num_loc_classes;
+        const int *pindices  = indices_data + n*_num_classes*_num_priors;
+
+        for (int c = 0; c < _num_classes; ++c) {
+            for (int i = 0; i < detections_data[n*_num_classes + c]; ++i) {
+                int idx = pindices[c*_num_priors + i];
+
+                dst_data[count * DETECTION_SIZE + 0] = static_cast<float>(n);
+                dst_data[count * DETECTION_SIZE + 1] = static_cast<float>(_decrease_label_id ? c-1 : c);
+                dst_data[count * DETECTION_SIZE + 2] = pconf[c*_num_priors + idx];
+
+                float xmin = _share_location ? pboxes[idx*4 + 0] :
+                             pboxes[c*4*_num_priors + idx*4 + 0];
+                float ymin = _share_location ? pboxes[idx*4 + 1] :
+                             pboxes[c*4*_num_priors + idx*4 + 1];
+                float xmax = _share_location ? pboxes[idx*4 + 2] :
+                             pboxes[c*4*_num_priors + idx*4 + 2];
+                float ymax = _share_location ? pboxes[idx*4 + 3] :
+                             pboxes[c*4*_num_priors + idx*4 + 3];
+
+                if (_clip_after_nms) {
+                    xmin = (std::max)(0.0f, (std::min)(1.0f, xmin));
+                    ymin = (std::max)(0.0f, (std::min)(1.0f, ymin));
+                    xmax = (std::max)(0.0f, (std::min)(1.0f, xmax));
+                    ymax = (std::max)(0.0f, (std::min)(1.0f, ymax));
+                }
+
+                dst_data[count * DETECTION_SIZE + 3] = xmin;
+                dst_data[count * DETECTION_SIZE + 4] = ymin;
+                dst_data[count * DETECTION_SIZE + 5] = xmax;
+                dst_data[count * DETECTION_SIZE + 6] = ymax;
+
+                ++count;
+            }
+        }
+    }
+
+    if (count < num_results) {
+        // marker at end of boxes list
+        dst_data[count * DETECTION_SIZE + 0] = -1;
+    }
+}
+
+struct ConfidenceComparator {
+    explicit ConfidenceComparator(const float* conf_data) : _conf_data(conf_data) {}
+
+    bool operator()(int idx1, int idx2) {
+        if (_conf_data[idx1] > _conf_data[idx2]) return true;
+        if (_conf_data[idx1] < _conf_data[idx2]) return false;
+        return idx1 < idx2;
+    }
+
+    const float* _conf_data;
+};
+
+static inline float JaccardOverlap(const float *decoded_bbox,
+                                   const float *bbox_sizes,
+                                   const int idx1,
+                                   const int idx2) {
+    float xmin1 = decoded_bbox[idx1*4 + 0];
+    float ymin1 = decoded_bbox[idx1*4 + 1];
+    float xmax1 = decoded_bbox[idx1*4 + 2];
+    float ymax1 = decoded_bbox[idx1*4 + 3];
+
+    float xmin2 = decoded_bbox[idx2*4 + 0];
+    float ymin2 = decoded_bbox[idx2*4 + 1];
+    float xmax2 = decoded_bbox[idx2*4 + 2];
+    float ymax2 = decoded_bbox[idx2*4 + 3];
+
+    if (xmin2 > xmax1 || xmax2 < xmin1 || ymin2 > ymax1 || ymax2 < ymin1) {
+        return 0.0f;
+    }
+
+    float intersect_xmin = (std::max)(xmin1, xmin2);
+    float intersect_ymin = (std::max)(ymin1, ymin2);
+    float intersect_xmax = (std::min)(xmax1, xmax2);
+    float intersect_ymax = (std::min)(ymax1, ymax2);
+
+    float intersect_width  = intersect_xmax - intersect_xmin;
+    float intersect_height = intersect_ymax - intersect_ymin;
+
+    if (intersect_width <= 0 || intersect_height <= 0) {
+        return 0.0f;
+    }
+
+    float intersect_size = intersect_width * intersect_height;
+    float bbox1_size = bbox_sizes[idx1];
+    float bbox2_size = bbox_sizes[idx2];
+
+    return intersect_size / (bbox1_size + bbox2_size - intersect_size);
+}
+
+void MKLDNNDetectionOutputNode::decodeBBoxes(const float *prior_data,
+                                       const float *loc_data,
+                                       const float *variance_data,
+                                       float *decoded_bboxes,
+                                       float *decoded_bbox_sizes,
+                                       int* num_priors_actual,
+                                       int n,
+                                       const int& offs,
+                                       const int& pr_size,
+                                       bool decodeType) {
+    num_priors_actual[n] = _num_priors;
+    if (!_normalized && decodeType) {
+        int num = 0;
+        for (; num < _num_priors; ++num) {
+            float batch_id = prior_data[num * pr_size + 0];
+            if (batch_id == -1.f) {
+                num_priors_actual[n] = num;
+                break;
+            }
+        }
+    }
+    parallel_for(num_priors_actual[n], [&](int p) {
+        float new_xmin = 0.0f;
+        float new_ymin = 0.0f;
+        float new_xmax = 0.0f;
+        float new_ymax = 0.0f;
+
+        float prior_xmin = prior_data[p*pr_size + 0 + offs];
+        float prior_ymin = prior_data[p*pr_size + 1 + offs];
+        float prior_xmax = prior_data[p*pr_size + 2 + offs];
+        float prior_ymax = prior_data[p*pr_size + 3 + offs];
+
+        float loc_xmin = loc_data[4*p*_num_loc_classes + 0];
+        float loc_ymin = loc_data[4*p*_num_loc_classes + 1];
+        float loc_xmax = loc_data[4*p*_num_loc_classes + 2];
+        float loc_ymax = loc_data[4*p*_num_loc_classes + 3];
+
+        if (!_normalized) {
+            prior_xmin /= _image_width;
+            prior_ymin /= _image_height;
+            prior_xmax /= _image_width;
+            prior_ymax /= _image_height;
+        }
+
+        if (_code_type == CodeType::CORNER) {
+            if (_variance_encoded_in_target) {
+                // variance is encoded in target, we simply need to add the offset predictions.
+                new_xmin = prior_xmin + loc_xmin;
+                new_ymin = prior_ymin + loc_ymin;
+                new_xmax = prior_xmax + loc_xmax;
+                new_ymax = prior_ymax + loc_ymax;
+            } else {
+                new_xmin = prior_xmin + variance_data[p*4 + 0] * loc_xmin;
+                new_ymin = prior_ymin + variance_data[p*4 + 1] * loc_ymin;
+                new_xmax = prior_xmax + variance_data[p*4 + 2] * loc_xmax;
+                new_ymax = prior_ymax + variance_data[p*4 + 3] * loc_ymax;
+            }
+        } else if (_code_type == CodeType::CENTER_SIZE) {
+            float prior_width    =  prior_xmax - prior_xmin;
+            float prior_height   =  prior_ymax - prior_ymin;
+            float prior_center_x = (prior_xmin + prior_xmax) / 2.0f;
+            float prior_center_y = (prior_ymin + prior_ymax) / 2.0f;
+
+            float decode_bbox_center_x, decode_bbox_center_y;
+            float decode_bbox_width, decode_bbox_height;
+
+            if (_variance_encoded_in_target) {
+                // variance is encoded in target, we simply need to restore the offset predictions.
+                decode_bbox_center_x = loc_xmin * prior_width  + prior_center_x;
+                decode_bbox_center_y = loc_ymin * prior_height + prior_center_y;
+                decode_bbox_width  = std::exp(loc_xmax) * prior_width;
+                decode_bbox_height = std::exp(loc_ymax) * prior_height;
+            } else {
+                // variance is encoded in bbox, we need to scale the offset accordingly.
+                decode_bbox_center_x = variance_data[p*4 + 0] * loc_xmin * prior_width + prior_center_x;
+                decode_bbox_center_y = variance_data[p*4 + 1] * loc_ymin * prior_height + prior_center_y;
+                decode_bbox_width    = std::exp(variance_data[p*4 + 2] * loc_xmax) * prior_width;
+                decode_bbox_height   = std::exp(variance_data[p*4 + 3] * loc_ymax) * prior_height;
+            }
+
+            new_xmin = decode_bbox_center_x - decode_bbox_width  / 2.0f;
+            new_ymin = decode_bbox_center_y - decode_bbox_height / 2.0f;
+            new_xmax = decode_bbox_center_x + decode_bbox_width  / 2.0f;
+            new_ymax = decode_bbox_center_y + decode_bbox_height / 2.0f;
+        }
+
+        if (_clip_before_nms) {
+            new_xmin = (std::max)(0.0f, (std::min)(1.0f, new_xmin));
+            new_ymin = (std::max)(0.0f, (std::min)(1.0f, new_ymin));
+            new_xmax = (std::max)(0.0f, (std::min)(1.0f, new_xmax));
+            new_ymax = (std::max)(0.0f, (std::min)(1.0f, new_ymax));
+        }
+
+        decoded_bboxes[p*4 + 0] = new_xmin;
+        decoded_bboxes[p*4 + 1] = new_ymin;
+        decoded_bboxes[p*4 + 2] = new_xmax;
+        decoded_bboxes[p*4 + 3] = new_ymax;
+
+        decoded_bbox_sizes[p] = (new_xmax - new_xmin) * (new_ymax - new_ymin);
+    });
+}
+
+void MKLDNNDetectionOutputNode::nms_cf(const float* conf_data,
+                                 const float* bboxes,
+                                 const float* sizes,
+                                 int* buffer,
+                                 int* indices,
+                                 int& detections,
+                                 int num_priors_actual) {
+    int count = 0;
+    for (int i = 0; i < num_priors_actual; ++i) {
+        if (conf_data[i] > _confidence_threshold) {
+            indices[count] = i;
+            count++;
+        }
+    }
+
+    int num_output_scores = (_top_k == -1 ? count : (std::min)(_top_k, count));
+
+    std::partial_sort_copy(indices, indices + count,
+                           buffer, buffer + num_output_scores,
+                           ConfidenceComparator(conf_data));
+
+    for (int i = 0; i < num_output_scores; ++i) {
+        const int idx = buffer[i];
+
+        bool keep = true;
+        for (int k = 0; k < detections; ++k) {
+            const int kept_idx = indices[k];
+            float overlap = JaccardOverlap(bboxes, sizes, idx, kept_idx);
+            if (overlap > _nms_threshold) {
+                keep = false;
+                break;
+            }
+        }
+        if (keep) {
+            indices[detections] = idx;
+            detections++;
+        }
+    }
+}
+
+void MKLDNNDetectionOutputNode::nms_mx(const float* conf_data,
+                                 const float* bboxes,
+                                 const float* sizes,
+                                 int* buffer,
+                                 int* indices,
+                                 int* detections,
+                                 int num_priors_actual) {
+    int count = 0;
+    for (int i = 0; i < num_priors_actual; ++i) {
+        float conf = -1;
+        int id = 0;
+        for (int c = 1; c < _num_classes; ++c) {
+            float temp = conf_data[c*_num_priors + i];
+            if (temp > conf) {
+                conf = temp;
+                id = c;
+            }
+        }
+
+        if (id > 0 && conf >= _confidence_threshold) {
+            indices[count++] = id*_num_priors + i;
+        }
+    }
+
+    int num_output_scores = (_top_k == -1 ? count : (std::min)(_top_k, count));
+
+    std::partial_sort_copy(indices, indices + count,
+                           buffer, buffer + num_output_scores,
+                           ConfidenceComparator(conf_data));
+
+    for (int i = 0; i < num_output_scores; ++i) {
+        const int idx = buffer[i];
+        const int cls = idx/_num_priors;
+        const int prior = idx%_num_priors;
+
+        int &ndetection = detections[cls];
+        int *pindices = indices + cls*_num_priors;
+
+        bool keep = true;
+        for (int k = 0; k < ndetection; ++k) {
+            const int kept_idx = pindices[k];
+            float overlap = 0.0f;
+            if (_share_location) {
+                overlap = JaccardOverlap(bboxes, sizes, prior, kept_idx);
+            } else {
+                overlap = JaccardOverlap(bboxes, sizes, cls*_num_priors + prior, cls*_num_priors + kept_idx);
+            }
+            if (overlap > _nms_threshold) {
+                keep = false;
+                break;
+            }
+        }
+        if (keep) {
+            pindices[ndetection++] = prior;
+        }
+    }
+}
+
+bool MKLDNNDetectionOutputNode::created() const {
+    return getType() == DetectionOutput;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNDetectionOutputNode, DetectionOutput)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_detection_output_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_detection_output_node.h
new file mode 100644
index 00000000000000..dbf9bde760907c
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_detection_output_node.h
@@ -0,0 +1,86 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNDetectionOutputNode : public MKLDNNNode {
+public:
+    MKLDNNDetectionOutputNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    const int idx_location = 0;
+    const int idx_confidence = 1;
+    const int idx_priors = 2;
+    const int idx_arm_confidence = 3;
+    const int idx_arm_location = 4;
+
+    int _num_classes = 0;
+    int _background_label_id = 0;
+    int _top_k = 0;
+    int _variance_encoded_in_target = 0;
+    int _keep_top_k = 0;
+    int _code_type = 0;
+
+    bool _share_location    = false;
+    bool _clip_before_nms   = false;  // clip bounding boxes before nms step
+    bool _clip_after_nms    = false;  // clip bounding boxes after nms step
+    bool _decrease_label_id = false;
+
+    bool with_add_box_pred = false;
+
+    int _image_width = 0;
+    int _image_height = 0;
+    int _prior_size = 4;
+    bool _normalized = true;
+    int _offset = 0;
+
+    float _nms_threshold = 0.0f;
+    float _confidence_threshold = 0.0f;
+    float _objectness_score = 0.0f;
+
+    int _num = 0;
+    int _num_loc_classes = 0;
+    int _num_priors = 0;
+    bool _priors_batches = false;
+
+    enum CodeType {
+        CORNER = 1,
+        CENTER_SIZE = 2,
+    };
+
+    void decodeBBoxes(const float *prior_data, const float *loc_data, const float *variance_data,
+                      float *decoded_bboxes, float *decoded_bbox_sizes, int* num_priors_actual, int n, const int& offs, const int& pr_size,
+                      bool decodeType = true); // after ARM = false
+
+    void nms_cf(const float *conf_data, const float *bboxes, const float *sizes,
+                int *buffer, int *indices, int &detections, int num_priors_actual);
+
+    void nms_mx(const float *conf_data, const float *bboxes, const float *sizes,
+                int *buffer, int *indices, int *detections, int num_priors_actual);
+
+    std::vector<float> _decoded_bboxes;
+    std::vector<int> _buffer;
+    std::vector<int> _indices;
+    std::vector<int> _detections_count;
+    std::vector<float> _reordered_conf;
+    std::vector<float> _bbox_sizes;
+    std::vector<int> _num_priors_actual;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_eltwise_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_eltwise_node.h
index 6b565370917db7..34e95d45ae06e8 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_eltwise_node.h
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_eltwise_node.h
@@ -124,4 +124,3 @@ class MKLDNNEltwiseNode : public MKLDNNNode {
 };
 
 }  // namespace MKLDNNPlugin
-
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_bag_offset_sum_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_bag_offset_sum_node.cpp
index c8810e4444b2a5..f59b69b023d99c 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_bag_offset_sum_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_bag_offset_sum_node.cpp
@@ -13,7 +13,7 @@ using namespace InferenceEngine;
 
 bool MKLDNNEmbeddingBagOffsetSumNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
     try {
-        auto embBagOffsetSumOp = ngraph::as_type_ptr<const ngraph::op::v3::EmbeddingBagOffsetsSum>(op);
+        const auto embBagOffsetSumOp = ngraph::as_type_ptr<const ngraph::op::v3::EmbeddingBagOffsetsSum>(op);
         if (!embBagOffsetSumOp) {
             errorMessage = "Node is not an instance of the EmbeddingBagOffsetsSum operation from opset v3.";
             return false;
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_bag_packed_sum_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_bag_packed_sum_node.cpp
index 4d1b808b502fb5..3318e1089faeed 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_bag_packed_sum_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_bag_packed_sum_node.cpp
@@ -13,7 +13,7 @@ using namespace InferenceEngine;
 
 bool MKLDNNEmbeddingBagPackedSumNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
     try {
-        auto embBagPackedSumOp = ngraph::as_type_ptr<const ngraph::op::v3::EmbeddingBagPackedSum>(op);
+        const auto embBagPackedSumOp = ngraph::as_type_ptr<const ngraph::op::v3::EmbeddingBagPackedSum>(op);
         if (!embBagPackedSumOp) {
             errorMessage = "Node is not an instance of the EmbeddingBagPackedSum operation from opset v3.";
             return false;
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_segments_sum_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_segments_sum_node.cpp
index 798feecf7bd062..82eae04dcc2193 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_segments_sum_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_embedding_segments_sum_node.cpp
@@ -13,7 +13,7 @@ using namespace InferenceEngine;
 
 bool MKLDNNEmbeddingSegmentsSumNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
     try {
-        auto embBagSegSumOp = ngraph::as_type_ptr<const ngraph::op::v3::EmbeddingSegmentsSum>(op);
+        const auto embBagSegSumOp = ngraph::as_type_ptr<const ngraph::op::v3::EmbeddingSegmentsSum>(op);
         if (!embBagSegSumOp) {
             errorMessage = "Node is not an instance of the EmbeddingSegmentsSum operation from opset v3.";
             return false;
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_detection_output_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_detection_output_node.cpp
new file mode 100644
index 00000000000000..fe2362003f377a
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_detection_output_node.cpp
@@ -0,0 +1,369 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+
+#include <ngraph/op/experimental_detectron_detection_output.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_experimental_detectron_detection_output_node.h"
+
+
+struct Indexer {
+    const std::vector<int> dims_;
+    int total_{1};
+
+    explicit Indexer(const std::vector<int>& dims) : dims_(dims) {
+        total_ = 1;
+        for (size_t i = 0; i < dims_.size(); ++i) {
+            total_ *= dims_[i];
+        }
+    }
+
+    int operator()(const std::vector<int>& idx) const {
+        int flat_idx = 0;
+        assert(idx.size() == dims_.size());
+        for (size_t i = 0; i < dims_.size(); ++i) {
+            assert(0 <= idx[i] && idx[i] < dims_[i]);
+            flat_idx = flat_idx * dims_[i] + idx[i];
+        }
+        assert(flat_idx < total_);
+        return flat_idx;
+    }
+};
+
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+static
+void refine_boxes(const float* boxes, const float* deltas, const float* weights, const float* scores,
+                  float* refined_boxes, float* refined_boxes_areas, float* refined_scores,
+                  const int rois_num, const int classes_num,
+                  const float img_H, const float img_W,
+                  const float max_delta_log_wh,
+                  float coordinates_offset) {
+    Indexer box_idx({rois_num, 4});
+    Indexer delta_idx({rois_num, classes_num, 4});
+    Indexer score_idx({rois_num, classes_num});
+
+    Indexer refined_box_idx({classes_num, rois_num, 4});
+    Indexer refined_score_idx({classes_num, rois_num});
+
+    for (int roi_idx = 0; roi_idx < rois_num; ++roi_idx) {
+        float x0 = boxes[box_idx({roi_idx, 0})];
+        float y0 = boxes[box_idx({roi_idx, 1})];
+        float x1 = boxes[box_idx({roi_idx, 2})];
+        float y1 = boxes[box_idx({roi_idx, 3})];
+
+        if (x1 - x0 <= 0 || y1 - y0 <= 0) {
+            continue;
+        }
+
+        // width & height of box
+        const float ww = x1 - x0 + coordinates_offset;
+        const float hh = y1 - y0 + coordinates_offset;
+        // center location of box
+        const float ctr_x = x0 + 0.5f * ww;
+        const float ctr_y = y0 + 0.5f * hh;
+
+        for (int class_idx = 1; class_idx < classes_num; ++class_idx) {
+            const float dx = deltas[delta_idx({roi_idx, class_idx, 0})] / weights[0];
+            const float dy = deltas[delta_idx({roi_idx, class_idx, 1})] / weights[1];
+            const float d_log_w = deltas[delta_idx({roi_idx, class_idx, 2})] / weights[2];
+            const float d_log_h = deltas[delta_idx({roi_idx, class_idx, 3})] / weights[3];
+
+            // new center location according to deltas (dx, dy)
+            const float pred_ctr_x = dx * ww + ctr_x;
+            const float pred_ctr_y = dy * hh + ctr_y;
+            // new width & height according to deltas d(log w), d(log h)
+            const float pred_w = std::exp((std::min)(d_log_w, max_delta_log_wh)) * ww;
+            const float pred_h = std::exp((std::min)(d_log_h, max_delta_log_wh)) * hh;
+
+            // update upper-left corner location
+            float x0_new = pred_ctr_x - 0.5f * pred_w;
+            float y0_new = pred_ctr_y - 0.5f * pred_h;
+            // update lower-right corner location
+            float x1_new = pred_ctr_x + 0.5f * pred_w - coordinates_offset;
+            float y1_new = pred_ctr_y + 0.5f * pred_h - coordinates_offset;
+
+            // adjust new corner locations to be within the image region,
+            x0_new = std::max<float>(0.0f, x0_new);
+            y0_new = std::max<float>(0.0f, y0_new);
+            x1_new = std::max<float>(0.0f, x1_new);
+            y1_new = std::max<float>(0.0f, y1_new);
+
+            // recompute new width & height
+            const float box_w = x1_new - x0_new + coordinates_offset;
+            const float box_h = y1_new - y0_new + coordinates_offset;
+
+            refined_boxes[refined_box_idx({class_idx, roi_idx, 0})] = x0_new;
+            refined_boxes[refined_box_idx({class_idx, roi_idx, 1})] = y0_new;
+            refined_boxes[refined_box_idx({class_idx, roi_idx, 2})] = x1_new;
+            refined_boxes[refined_box_idx({class_idx, roi_idx, 3})] = y1_new;
+
+            refined_boxes_areas[refined_score_idx({class_idx, roi_idx})] = box_w * box_h;
+
+            refined_scores[refined_score_idx({class_idx, roi_idx})] = scores[score_idx({roi_idx, class_idx})];
+        }
+    }
+}
+
+template <typename T>
+static bool SortScorePairDescend(const std::pair<float, T>& pair1,
+                                 const std::pair<float, T>& pair2) {
+    return pair1.first > pair2.first;
+}
+
+
+struct ConfidenceComparator {
+    explicit ConfidenceComparator(const float* conf_data) : _conf_data(conf_data) {}
+
+    bool operator()(int idx1, int idx2) {
+        if (_conf_data[idx1] > _conf_data[idx2]) return true;
+        if (_conf_data[idx1] < _conf_data[idx2]) return false;
+        return idx1 < idx2;
+    }
+
+    const float* _conf_data;
+};
+
+static inline float JaccardOverlap(const float *decoded_bbox,
+                                   const float *bbox_sizes,
+                                   const int idx1,
+                                   const int idx2,
+                                   const float coordinates_offset = 1) {
+    float xmin1 = decoded_bbox[idx1 * 4 + 0];
+    float ymin1 = decoded_bbox[idx1 * 4 + 1];
+    float xmax1 = decoded_bbox[idx1 * 4 + 2];
+    float ymax1 = decoded_bbox[idx1 * 4 + 3];
+
+    float xmin2 = decoded_bbox[idx2 * 4 + 0];
+    float ymin2 = decoded_bbox[idx2 * 4 + 1];
+    float ymax2 = decoded_bbox[idx2 * 4 + 3];
+    float xmax2 = decoded_bbox[idx2 * 4 + 2];
+
+    if (xmin2 > xmax1 || xmax2 < xmin1 || ymin2 > ymax1 || ymax2 < ymin1) {
+        return 0.0f;
+    }
+
+    float intersect_xmin = (std::max)(xmin1, xmin2);
+    float intersect_ymin = (std::max)(ymin1, ymin2);
+    float intersect_xmax = (std::min)(xmax1, xmax2);
+    float intersect_ymax = (std::min)(ymax1, ymax2);
+
+    float intersect_width  = intersect_xmax - intersect_xmin + coordinates_offset;
+    float intersect_height = intersect_ymax - intersect_ymin + coordinates_offset;
+
+    if (intersect_width <= 0 || intersect_height <= 0) {
+        return 0.0f;
+    }
+
+    float intersect_size = intersect_width * intersect_height;
+    float bbox1_size = bbox_sizes[idx1];
+    float bbox2_size = bbox_sizes[idx2];
+
+    return intersect_size / (bbox1_size + bbox2_size - intersect_size);
+}
+
+
+static void nms_cf(const float* conf_data,
+                   const float* bboxes,
+                   const float* sizes,
+                   int* buffer,
+                   int* indices,
+                   int& detections,
+                   const int boxes_num,
+                   const int pre_nms_topn,
+                   const int post_nms_topn,
+                   const float confidence_threshold,
+                   const float nms_threshold) {
+    int count = 0;
+    for (int i = 0; i < boxes_num; ++i) {
+        if (conf_data[i] > confidence_threshold) {
+            indices[count] = i;
+            count++;
+        }
+    }
+
+    int num_output_scores = (pre_nms_topn == -1 ? count : (std::min)(pre_nms_topn, count));
+
+    std::partial_sort_copy(indices, indices + count,
+                           buffer, buffer + num_output_scores,
+                           ConfidenceComparator(conf_data));
+
+    detections = 0;
+    for (int i = 0; i < num_output_scores; ++i) {
+        const int idx = buffer[i];
+
+        bool keep = true;
+        for (int k = 0; k < detections; ++k) {
+            const int kept_idx = indices[k];
+            float overlap = JaccardOverlap(bboxes, sizes, idx, kept_idx);
+            if (overlap > nms_threshold) {
+                keep = false;
+                break;
+            }
+        }
+        if (keep) {
+            indices[detections] = idx;
+            detections++;
+        }
+    }
+
+    detections = (post_nms_topn == -1 ? detections : (std::min)(post_nms_topn, detections));
+}
+
+bool MKLDNNExperimentalDetectronDetectionOutputNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto doOp = ngraph::as_type_ptr<const ngraph::op::v6::ExperimentalDetectronDetectionOutput>(op);
+        if (!doOp) {
+            errorMessage = "Node is not an instance of the ExperimentalDetectronDetectionOutput from the operations set v6.";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNExperimentalDetectronDetectionOutputNode::MKLDNNExperimentalDetectronDetectionOutputNode
+                        (const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+                         MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+    auto doOp = ngraph::as_type_ptr<const ngraph::op::v6::ExperimentalDetectronDetectionOutput>(op);
+    auto attributes = doOp->get_attrs();
+
+    score_threshold_ = attributes.score_threshold;
+    nms_threshold_ = attributes.nms_threshold;
+    max_delta_log_wh_ = attributes.max_delta_log_wh;
+    classes_num_ = attributes.num_classes;
+    max_detections_per_class_ = attributes.post_nms_count;
+    max_detections_per_image_ = attributes.max_detections_per_image;
+    class_agnostic_box_regression_ = attributes.class_agnostic_box_regression;
+    deltas_weights_ = attributes.deltas_weights;
+}
+
+void MKLDNNExperimentalDetectronDetectionOutputNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    std::vector<DataConfigurator> inDataConf;
+    inDataConf.reserve(getOriginalInputsNumber());
+    for (int i = 0; i < getOriginalInputsNumber(); ++i)
+        inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::FP32);
+
+    addSupportedPrimDesc(inDataConf,
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::I32},
+                          {TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNExperimentalDetectronDetectionOutputNode::execute(mkldnn::stream strm) {
+    const int rois_num = getParentEdgeAt(INPUT_ROIS)->getDims()[0];
+    assert(classes_num_ == static_cast<int>(getParentEdgeAt(INPUT_SCORES)->getDims()[1]));
+    assert(4 * classes_num_ == static_cast<int>(getParentEdgeAt(INPUT_DELTAS)->getDims()[1]));
+
+    const auto* boxes = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_ROIS)->getMemoryPtr()->GetPtr());
+    const auto* deltas = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_DELTAS)->getMemoryPtr()->GetPtr());
+    const auto* scores = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_SCORES)->getMemoryPtr()->GetPtr());
+    const auto* im_info = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_IM_INFO)->getMemoryPtr()->GetPtr());
+
+    auto* output_boxes = reinterpret_cast<float *>(getChildEdgesAtPort(OUTPUT_BOXES)[0]->getMemoryPtr()->GetPtr());
+    auto* output_scores = reinterpret_cast<float *>(getChildEdgesAtPort(OUTPUT_SCORES)[0]->getMemoryPtr()->GetPtr());
+    auto* output_classes = reinterpret_cast<int32_t *>(getChildEdgesAtPort(OUTPUT_CLASSES)[0]->getMemoryPtr()->GetPtr());
+
+    const float img_H = im_info[0];
+    const float img_W = im_info[1];
+
+    // Apply deltas.
+    std::vector<float> refined_boxes(classes_num_ * rois_num * 4, 0);
+    std::vector<float> refined_scores(classes_num_ * rois_num, 0);
+    std::vector<float> refined_boxes_areas(classes_num_ * rois_num, 0);
+    Indexer refined_box_idx({classes_num_, rois_num, 4});
+    Indexer refined_score_idx({classes_num_, rois_num});
+
+    refine_boxes(boxes, deltas, &deltas_weights_[0], scores,
+                 &refined_boxes[0], &refined_boxes_areas[0], &refined_scores[0],
+                 rois_num, classes_num_,
+                 img_H, img_W,
+                 max_delta_log_wh_,
+                 1.0f);
+
+    // Apply NMS class-wise.
+    std::vector<int> buffer(rois_num, 0);
+    std::vector<int> indices(classes_num_ * rois_num, 0);
+    std::vector<int> detections_per_class(classes_num_, 0);
+    int total_detections_num = 0;
+
+    for (int class_idx = 1; class_idx < classes_num_; ++class_idx) {
+        nms_cf(&refined_scores[refined_score_idx({class_idx, 0})],
+               &refined_boxes[refined_box_idx({class_idx, 0, 0})],
+               &refined_boxes_areas[refined_score_idx({class_idx, 0})],
+               &buffer[0],
+               &indices[total_detections_num],
+               detections_per_class[class_idx],
+               rois_num,
+               -1,
+               max_detections_per_class_,
+               score_threshold_,
+               nms_threshold_);
+        total_detections_num += detections_per_class[class_idx];
+    }
+
+    // Leave only max_detections_per_image_ detections.
+    // confidence, <class, index>
+    std::vector<std::pair<float, std::pair<int, int>>> conf_index_class_map;
+
+    int indices_offset = 0;
+    for (int c = 0; c < classes_num_; ++c) {
+        int n = detections_per_class[c];
+        for (int i = 0; i < n; ++i) {
+            int idx = indices[indices_offset + i];
+            float score = refined_scores[refined_score_idx({c, idx})];
+            conf_index_class_map.push_back(std::make_pair(score, std::make_pair(c, idx)));
+        }
+        indices_offset += n;
+    }
+
+    assert(max_detections_per_image_ > 0);
+    if (total_detections_num > max_detections_per_image_) {
+        std::partial_sort(conf_index_class_map.begin(),
+                          conf_index_class_map.begin() + max_detections_per_image_,
+                          conf_index_class_map.end(),
+                          SortScorePairDescend<std::pair<int, int>>);
+        conf_index_class_map.resize(max_detections_per_image_);
+        total_detections_num = max_detections_per_image_;
+    }
+
+    // Fill outputs.
+    memset(output_boxes, 0, max_detections_per_image_ * 4 * sizeof(output_boxes[0]));
+    memset(output_scores, 0, max_detections_per_image_ * sizeof(output_scores[0]));
+    memset(output_classes, 0, max_detections_per_image_ * sizeof(output_classes[0]));
+
+    int i = 0;
+    for (const auto & detection : conf_index_class_map) {
+        float score = detection.first;
+        int cls = detection.second.first;
+        int idx = detection.second.second;
+        output_boxes[4 * i + 0] = refined_boxes[refined_box_idx({cls, idx, 0})];
+        output_boxes[4 * i + 1] = refined_boxes[refined_box_idx({cls, idx, 1})];
+        output_boxes[4 * i + 2] = refined_boxes[refined_box_idx({cls, idx, 2})];
+        output_boxes[4 * i + 3] = refined_boxes[refined_box_idx({cls, idx, 3})];
+        output_scores[i] = score;
+        output_classes[i] = cls;
+        ++i;
+    }
+}
+
+bool MKLDNNExperimentalDetectronDetectionOutputNode::created() const {
+    return getType() == ExperimentalDetectronDetectionOutput;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNExperimentalDetectronDetectionOutputNode, ExperimentalDetectronDetectionOutput)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_detection_output_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_detection_output_node.h
new file mode 100644
index 00000000000000..2df28ce5c4983b
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_detection_output_node.h
@@ -0,0 +1,46 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNExperimentalDetectronDetectionOutputNode : public MKLDNNNode {
+public:
+    MKLDNNExperimentalDetectronDetectionOutputNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    const int INPUT_ROIS {0};
+    const int INPUT_DELTAS {1};
+    const int INPUT_SCORES {2};
+    const int INPUT_IM_INFO {3};
+
+    const int OUTPUT_BOXES {0};
+    const int OUTPUT_CLASSES {1};
+    const int OUTPUT_SCORES {2};
+
+    float score_threshold_;
+    float nms_threshold_;
+    float max_delta_log_wh_;
+    int classes_num_;
+    int max_detections_per_class_;
+    int max_detections_per_image_;
+    bool class_agnostic_box_regression_;
+    std::vector<float> deltas_weights_;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_generate_proposals_single_image_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_generate_proposals_single_image_node.cpp
new file mode 100644
index 00000000000000..255f8443765660
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_generate_proposals_single_image_node.cpp
@@ -0,0 +1,429 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "base.hpp"
+
+#include <cstring>
+#include <cassert>
+#include <cmath>
+#include <string>
+#include <vector>
+#include <utility>
+#include <algorithm>
+
+#if defined(HAVE_AVX2)
+#include <immintrin.h>
+#endif
+
+#include <ngraph/op/experimental_detectron_generate_proposals.hpp>
+#include "ie_parallel.hpp"
+#include "common/cpu_memcpy.h"
+#include "mkldnn_experimental_detectron_generate_proposals_single_image_node.h"
+
+namespace {
+struct Indexer4d {
+    int dim3_;
+    int dim23_;
+    int dim123_;
+
+    explicit Indexer4d(int dim0, int dim1, int dim2, int dim3):
+            dim3_(dim3), dim23_(dim2 * dim3), dim123_(dim1 * dim2 * dim3) {
+        (void)dim0;
+    }
+
+    int operator()(int i, int j, int k, int n) const {
+        return  i * dim123_ + j * dim23_ + k * dim3_ + n;
+    }
+};
+}  // namespace
+
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+static
+void refine_anchors(const float* deltas, const float* scores, const float* anchors,
+                    float* proposals, const int anchors_num, const int bottom_H,
+                    const int bottom_W, const float img_H, const float img_W,
+                    const float min_box_H, const float min_box_W,
+                    const float max_delta_log_wh,
+                    float coordinates_offset) {
+    Indexer4d delta_idx(anchors_num, 4, bottom_H, bottom_W);
+    Indexer4d score_idx(anchors_num, 1, bottom_H, bottom_W);
+    Indexer4d proposal_idx(bottom_H, bottom_W, anchors_num, 5);
+    Indexer4d anchor_idx(bottom_H, bottom_W, anchors_num, 4);
+
+    parallel_for2d(bottom_H, bottom_W, [&](int h, int w) {
+        for (int anchor = 0; anchor < anchors_num; ++anchor) {
+            int a_idx = anchor_idx(h, w, anchor, 0);
+            float x0 = anchors[a_idx + 0];
+            float y0 = anchors[a_idx + 1];
+            float x1 = anchors[a_idx + 2];
+            float y1 = anchors[a_idx + 3];
+
+            const float dx = deltas[delta_idx(anchor, 0, h, w)];
+            const float dy = deltas[delta_idx(anchor, 1, h, w)];
+            const float d_log_w = deltas[delta_idx(anchor, 2, h, w)];
+            const float d_log_h = deltas[delta_idx(anchor, 3, h, w)];
+
+            const float score = scores[score_idx(anchor, 0, h, w)];
+
+            // width & height of box
+            const float ww = x1 - x0 + coordinates_offset;
+            const float hh = y1 - y0 + coordinates_offset;
+            // center location of box
+            const float ctr_x = x0 + 0.5f * ww;
+            const float ctr_y = y0 + 0.5f * hh;
+
+            // new center location according to deltas (dx, dy)
+            const float pred_ctr_x = dx * ww + ctr_x;
+            const float pred_ctr_y = dy * hh + ctr_y;
+            // new width & height according to deltas d(log w), d(log h)
+            const float pred_w = std::exp(std::min(d_log_w, max_delta_log_wh)) * ww;
+            const float pred_h = std::exp(std::min(d_log_h, max_delta_log_wh)) * hh;
+
+            // update upper-left corner location
+            x0 = pred_ctr_x - 0.5f * pred_w;
+            y0 = pred_ctr_y - 0.5f * pred_h;
+            // update lower-right corner location
+            x1 = pred_ctr_x + 0.5f * pred_w - coordinates_offset;
+            y1 = pred_ctr_y + 0.5f * pred_h - coordinates_offset;
+
+            // adjust new corner locations to be within the image region,
+            x0 = std::max<float>(0.0f, std::min<float>(x0, img_W - coordinates_offset));
+            y0 = std::max<float>(0.0f, std::min<float>(y0, img_H - coordinates_offset));
+            x1 = std::max<float>(0.0f, std::min<float>(x1, img_W - coordinates_offset));
+            y1 = std::max<float>(0.0f, std::min<float>(y1, img_H - coordinates_offset));
+
+            // recompute new width & height
+            const float box_w = x1 - x0 + coordinates_offset;
+            const float box_h = y1 - y0 + coordinates_offset;
+
+            int p_idx = proposal_idx(h, w, anchor, 0);
+            proposals[p_idx + 0] = x0;
+            proposals[p_idx + 1] = y0;
+            proposals[p_idx + 2] = x1;
+            proposals[p_idx + 3] = y1;
+            proposals[p_idx + 4] = (min_box_W <= box_w) * (min_box_H <= box_h) * score;
+        }
+    });
+}
+
+static void unpack_boxes(const float* p_proposals, float* unpacked_boxes, int pre_nms_topn) {
+    parallel_for(pre_nms_topn, [&](size_t i) {
+        unpacked_boxes[0*pre_nms_topn + i] = p_proposals[5*i + 0];
+        unpacked_boxes[1*pre_nms_topn + i] = p_proposals[5*i + 1];
+        unpacked_boxes[2*pre_nms_topn + i] = p_proposals[5*i + 2];
+        unpacked_boxes[3*pre_nms_topn + i] = p_proposals[5*i + 3];
+        unpacked_boxes[4*pre_nms_topn + i] = p_proposals[5*i + 4];
+    });
+}
+
+static
+void nms_cpu(const int num_boxes, int is_dead[],
+             const float* boxes, int index_out[], int* const num_out,
+             const int base_index, const float nms_thresh, const int max_num_out,
+             float coordinates_offset) {
+    const int num_proposals = num_boxes;
+    int count = 0;
+
+    const float* x0 = boxes + 0 * num_proposals;
+    const float* y0 = boxes + 1 * num_proposals;
+    const float* x1 = boxes + 2 * num_proposals;
+    const float* y1 = boxes + 3 * num_proposals;
+
+    std::memset(is_dead, 0, num_boxes * sizeof(int));
+
+#if defined(HAVE_AVX2)
+    __m256  vc_fone = _mm256_set1_ps(coordinates_offset);
+    __m256i vc_ione = _mm256_set1_epi32(1);
+    __m256  vc_zero = _mm256_set1_ps(0.0f);
+
+    __m256 vc_nms_thresh = _mm256_set1_ps(nms_thresh);
+#endif
+
+    for (int box = 0; box < num_boxes; ++box) {
+        if (is_dead[box])
+            continue;
+
+        index_out[count++] = base_index + box;
+        if (count == max_num_out)
+            break;
+
+        int tail = box + 1;
+
+#if defined(HAVE_AVX2)
+        __m256 vx0i = _mm256_set1_ps(x0[box]);
+        __m256 vy0i = _mm256_set1_ps(y0[box]);
+        __m256 vx1i = _mm256_set1_ps(x1[box]);
+        __m256 vy1i = _mm256_set1_ps(y1[box]);
+
+        __m256 vA_width  = _mm256_sub_ps(vx1i, vx0i);
+        __m256 vA_height = _mm256_sub_ps(vy1i, vy0i);
+        __m256 vA_area   = _mm256_mul_ps(_mm256_add_ps(vA_width, vc_fone), _mm256_add_ps(vA_height, vc_fone));
+
+        for (; tail <= num_boxes - 8; tail += 8) {
+            __m256i *pdst = reinterpret_cast<__m256i*>(is_dead + tail);
+            __m256i  vdst = _mm256_loadu_si256(pdst);
+
+            __m256 vx0j = _mm256_loadu_ps(x0 + tail);
+            __m256 vy0j = _mm256_loadu_ps(y0 + tail);
+            __m256 vx1j = _mm256_loadu_ps(x1 + tail);
+            __m256 vy1j = _mm256_loadu_ps(y1 + tail);
+
+            __m256 vx0 = _mm256_max_ps(vx0i, vx0j);
+            __m256 vy0 = _mm256_max_ps(vy0i, vy0j);
+            __m256 vx1 = _mm256_min_ps(vx1i, vx1j);
+            __m256 vy1 = _mm256_min_ps(vy1i, vy1j);
+
+            __m256 vwidth  = _mm256_add_ps(_mm256_sub_ps(vx1, vx0), vc_fone);
+            __m256 vheight = _mm256_add_ps(_mm256_sub_ps(vy1, vy0), vc_fone);
+            __m256 varea = _mm256_mul_ps(_mm256_max_ps(vc_zero, vwidth), _mm256_max_ps(vc_zero, vheight));
+
+            __m256 vB_width  = _mm256_sub_ps(vx1j, vx0j);
+            __m256 vB_height = _mm256_sub_ps(vy1j, vy0j);
+            __m256 vB_area   = _mm256_mul_ps(_mm256_add_ps(vB_width, vc_fone), _mm256_add_ps(vB_height, vc_fone));
+
+            __m256 vdivisor = _mm256_sub_ps(_mm256_add_ps(vA_area, vB_area), varea);
+            __m256 vintersection_area = _mm256_div_ps(varea, vdivisor);
+
+            __m256 vcmp_0 = _mm256_cmp_ps(vx0i, vx1j, _CMP_LE_OS);
+            __m256 vcmp_1 = _mm256_cmp_ps(vy0i, vy1j, _CMP_LE_OS);
+            __m256 vcmp_2 = _mm256_cmp_ps(vx0j, vx1i, _CMP_LE_OS);
+            __m256 vcmp_3 = _mm256_cmp_ps(vy0j, vy1i, _CMP_LE_OS);
+            __m256 vcmp_4 = _mm256_cmp_ps(vc_nms_thresh, vintersection_area, _CMP_LT_OS);
+
+            vcmp_0 = _mm256_and_ps(vcmp_0, vcmp_1);
+            vcmp_2 = _mm256_and_ps(vcmp_2, vcmp_3);
+            vcmp_4 = _mm256_and_ps(vcmp_4, vcmp_0);
+            vcmp_4 = _mm256_and_ps(vcmp_4, vcmp_2);
+
+            _mm256_storeu_si256(pdst, _mm256_blendv_epi8(vdst, vc_ione, _mm256_castps_si256(vcmp_4)));
+        }
+#endif
+
+        for (; tail < num_boxes; ++tail) {
+            float res = 0.0f;
+
+            const float x0i = x0[box];
+            const float y0i = y0[box];
+            const float x1i = x1[box];
+            const float y1i = y1[box];
+
+            const float x0j = x0[tail];
+            const float y0j = y0[tail];
+            const float x1j = x1[tail];
+            const float y1j = y1[tail];
+
+            if (x0i <= x1j && y0i <= y1j && x0j <= x1i && y0j <= y1i) {
+                // overlapped region (= box)
+                const float x0 = std::max<float>(x0i, x0j);
+                const float y0 = std::max<float>(y0i, y0j);
+                const float x1 = std::min<float>(x1i, x1j);
+                const float y1 = std::min<float>(y1i, y1j);
+
+                // intersection area
+                const float width  = std::max<float>(0.0f,  x1 - x0 + coordinates_offset);
+                const float height = std::max<float>(0.0f,  y1 - y0 + coordinates_offset);
+                const float area   = width * height;
+
+                // area of A, B
+                const float A_area = (x1i - x0i + coordinates_offset) * (y1i - y0i + coordinates_offset);
+                const float B_area = (x1j - x0j + coordinates_offset) * (y1j - y0j + coordinates_offset);
+
+                // IoU
+                res = area / (A_area + B_area - area);
+            }
+
+            if (nms_thresh < res)
+                is_dead[tail] = 1;
+        }
+    }
+
+    *num_out = count;
+}
+
+
+static
+void fill_output_blobs(const float* proposals, const int* roi_indices,
+                       float* rois, float* scores,
+                       const int num_proposals, const int num_rois, const int post_nms_topn) {
+    const float *src_x0 = proposals + 0 * num_proposals;
+    const float *src_y0 = proposals + 1 * num_proposals;
+    const float *src_x1 = proposals + 2 * num_proposals;
+    const float *src_y1 = proposals + 3 * num_proposals;
+    const float *src_score = proposals + 4 * num_proposals;
+
+    parallel_for(num_rois, [&](size_t i) {
+        int index = roi_indices[i];
+        rois[i * 4 + 0] = src_x0[index];
+        rois[i * 4 + 1] = src_y0[index];
+        rois[i * 4 + 2] = src_x1[index];
+        rois[i * 4 + 3] = src_y1[index];
+        scores[i] = src_score[index];
+    });
+
+    if (num_rois < post_nms_topn) {
+        for (int i = 4 * num_rois; i < 4 * post_nms_topn; i++) {
+            rois[i] = 0.f;
+        }
+        for (int i = num_rois; i < post_nms_topn; i++) {
+            scores[i] = 0.f;
+        }
+    }
+}
+
+bool MKLDNNExperimentalDetectronGenerateProposalsSingleImageNode::isSupportedOperation
+            (const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto proposalOp = ngraph::as_type_ptr<const ngraph::op::v6::ExperimentalDetectronGenerateProposalsSingleImage>(op);
+        if (!proposalOp) {
+            errorMessage = "Node is not an instance of the Proposal from the operations set v0.";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNExperimentalDetectronGenerateProposalsSingleImageNode::MKLDNNExperimentalDetectronGenerateProposalsSingleImageNode
+        (const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+                MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    auto proposalOp = ngraph::as_type_ptr<const ngraph::op::v6::ExperimentalDetectronGenerateProposalsSingleImage>(op);
+    auto proposalAttrs = proposalOp->get_attrs();
+
+    min_size_ = proposalAttrs.min_size;
+    nms_thresh_ = proposalAttrs.nms_threshold;
+    pre_nms_topn_ = proposalAttrs.pre_nms_count;
+    post_nms_topn_ = proposalAttrs.post_nms_count;
+
+    coordinates_offset = 0.0f;
+
+    roi_indices_.resize(post_nms_topn_);
+}
+
+void MKLDNNExperimentalDetectronGenerateProposalsSingleImageNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNExperimentalDetectronGenerateProposalsSingleImageNode::execute(mkldnn::stream strm) {
+    try {
+        if (inDims.size() != 4 || outDims.size() != 2) {
+            IE_THROW() << "Incorrect number of input or output edges!";
+        }
+
+        size_t anchor_dims_size = 1;
+        for (size_t i = 0; i < getParentEdgeAt(INPUT_ANCHORS)->getDims().ToSizeVector().size(); i++) {
+            anchor_dims_size *= getParentEdgeAt(INPUT_ANCHORS)->getDims().ToSizeVector()[i];
+        }
+
+        size_t deltas_dims_size = 1;
+        for (size_t i = 0; i < getParentEdgeAt(INPUT_DELTAS)->getDims().ToSizeVector().size(); i++) {
+            deltas_dims_size *= getParentEdgeAt(INPUT_DELTAS)->getDims().ToSizeVector()[i];
+        }
+        if (anchor_dims_size != deltas_dims_size)
+            IE_THROW() << "'Anchors' blob size for ONNXProposal is incompatible with 'deltas' blob size!";
+
+        size_t score_dims_size = 1;
+        for (size_t i = 0; i < getParentEdgeAt(INPUT_SCORES)->getDims().ToSizeVector().size(); i++) {
+            score_dims_size *= getParentEdgeAt(INPUT_SCORES)->getDims().ToSizeVector()[i];
+        }
+        if (deltas_dims_size != (4 * score_dims_size))
+            IE_THROW() << "'Deltas' blob size for ONNXProposal is incompatible with 'scores' blob size!";
+
+        // Prepare memory
+        const float *p_deltas_item  = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_DELTAS)->getMemoryPtr()->GetPtr());
+        const float *p_scores_item  = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_SCORES)->getMemoryPtr()->GetPtr());
+        const float *p_anchors_item = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_ANCHORS)->getMemoryPtr()->GetPtr());
+        const float *p_img_info_cpu = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_IM_INFO)->getMemoryPtr()->GetPtr());
+
+        float *p_roi_item       = reinterpret_cast<float *>(getChildEdgesAtPort(OUTPUT_ROIS)[0]->getMemoryPtr()->GetPtr());
+        float *p_roi_score_item = reinterpret_cast<float *>(getChildEdgesAtPort(OUTPUT_SCORES)[0]->getMemoryPtr()->GetPtr());
+
+        const int anchors_num = getParentEdgeAt(INPUT_SCORES)->getDims()[0];
+
+        // bottom shape: (num_anchors) x H x W
+        const int bottom_H = getParentEdgeAt(INPUT_DELTAS)->getDims()[1];
+        const int bottom_W = getParentEdgeAt(INPUT_DELTAS)->getDims()[2];
+
+        // input image height & width
+        const float img_H = p_img_info_cpu[0];
+        const float img_W = p_img_info_cpu[1];
+
+        // scale factor for height & width
+
+        // minimum box width & height
+        const float min_box_H = min_size_;
+        const float min_box_W = min_size_;
+
+        // number of all proposals = num_anchors * H * W
+        const int num_proposals = anchors_num * bottom_H * bottom_W;
+
+        // number of top-n proposals before NMS
+        const int pre_nms_topn = std::min<int>(num_proposals, pre_nms_topn_);
+
+        // number of final RoIs
+        int num_rois = 0;
+
+        // enumerate all proposals
+        //   num_proposals = num_anchors * H * W
+        //   (x1, y1, x2, y2, score) for each proposal
+        // NOTE: for bottom, only foreground scores are passed
+        struct ProposalBox {
+            float x0;
+            float y0;
+            float x1;
+            float y1;
+            float score;
+        };
+        std::vector<ProposalBox> proposals_(num_proposals);
+        std::vector<float> unpacked_boxes(5 * pre_nms_topn);
+        std::vector<int> is_dead(pre_nms_topn);
+
+        // Execute
+        int batch_size = 1;  // inputs[INPUT_DELTAS]->getTensorDesc().getDims()[0];
+        for (int n = 0; n < batch_size; ++n) {
+            refine_anchors(p_deltas_item, p_scores_item, p_anchors_item,
+                           reinterpret_cast<float *>(&proposals_[0]), anchors_num, bottom_H,
+                           bottom_W, img_H, img_W,
+                           min_box_H, min_box_W,
+                           static_cast<const float>(log(1000. / 16.)),
+                           1.0f);
+            std::partial_sort(proposals_.begin(), proposals_.begin() + pre_nms_topn, proposals_.end(),
+                              [](const ProposalBox &struct1, const ProposalBox &struct2) {
+                                  return (struct1.score > struct2.score);
+                              });
+
+            unpack_boxes(reinterpret_cast<float *>(&proposals_[0]), &unpacked_boxes[0], pre_nms_topn);
+            nms_cpu(pre_nms_topn, &is_dead[0], &unpacked_boxes[0], &roi_indices_[0], &num_rois, 0,
+                    nms_thresh_, post_nms_topn_, coordinates_offset);
+            fill_output_blobs(&unpacked_boxes[0], &roi_indices_[0], p_roi_item, p_roi_score_item,
+                              pre_nms_topn, num_rois, post_nms_topn_);
+        }
+    } catch (const std::exception &e) {
+        std::string errorMsg = e.what();
+        IE_THROW() << errorMsg;
+    }
+}
+
+bool MKLDNNExperimentalDetectronGenerateProposalsSingleImageNode::created() const {
+    return getType() == ExperimentalDetectronGenerateProposalsSingleImage;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNExperimentalDetectronGenerateProposalsSingleImageNode, ExperimentalDetectronGenerateProposalsSingleImage)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_generate_proposals_single_image_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_generate_proposals_single_image_node.h
new file mode 100644
index 00000000000000..b2f5f0bcd89fe1
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_generate_proposals_single_image_node.h
@@ -0,0 +1,50 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNExperimentalDetectronGenerateProposalsSingleImageNode : public MKLDNNNode {
+public:
+    MKLDNNExperimentalDetectronGenerateProposalsSingleImageNode(const std::shared_ptr<ngraph::Node>& op,
+        const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    // Inputs:
+    //      rois, shape [n, 4]
+    //      rois_probs, shape [n]
+    // Outputs:
+    //      top_rois, shape [max_rois, 4]
+
+    const int INPUT_IM_INFO {0};
+    const int INPUT_ANCHORS {1};
+    const int INPUT_DELTAS {2};
+    const int INPUT_SCORES {3};
+    const int OUTPUT_ROIS {0};
+    const int OUTPUT_SCORES {1};
+
+    float min_size_;
+    int pre_nms_topn_;
+    int post_nms_topn_;
+    float nms_thresh_;
+    float coordinates_offset;
+
+    std::vector<int> roi_indices_;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_priorgridgenerator_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_priorgridgenerator_node.cpp
new file mode 100644
index 00000000000000..b5d073a0b3552e
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_priorgridgenerator_node.cpp
@@ -0,0 +1,95 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "base.hpp"
+
+#include <string>
+
+#include <ngraph/opsets/opset6.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_experimental_detectron_priorgridgenerator_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNExperimentalDetectronPriorGridGeneratorNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto priorGridGen = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronPriorGridGenerator>(op);
+        if (!priorGridGen) {
+            errorMessage = "Only opset6 ExperimentalDetectronPriorGridGenerator operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNExperimentalDetectronPriorGridGeneratorNode::MKLDNNExperimentalDetectronPriorGridGeneratorNode
+        (const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+                MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "ExperimentalDetectronPriorGridGenerator layer with name '" + op->get_friendly_name() + "'";
+    const auto priorGridGen = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronPriorGridGenerator>(op);
+    if (getOriginalInputsNumber() != 3 || getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
+
+    if (op->get_input_shape(INPUT_PRIORS).size() != 2 ||
+        op->get_input_shape(INPUT_FEATUREMAP).size() != 4 ||
+        op->get_input_shape(INPUT_IMAGE).size() != 4)
+        IE_THROW() << errorPrefix << " has unsupported input shape";
+
+    const auto &attr = priorGridGen->get_attrs();
+    grid_w_ = attr.w;
+    grid_h_ = attr.h;
+    stride_h_ = attr.stride_y;
+    stride_w_ = attr.stride_x;
+}
+
+void MKLDNNExperimentalDetectronPriorGridGeneratorNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNExperimentalDetectronPriorGridGeneratorNode::execute(mkldnn::stream strm) {
+    const int num_priors_ = getParentEdgeAt(INPUT_PRIORS)->getDims()[0];
+    assert(getParentEdgeAt(INPUT_PRIORS)->getDims()[1] == 4);
+
+    // Execute
+    const int layer_width = grid_w_ ? grid_w_ : getParentEdgeAt(INPUT_FEATUREMAP)->getDims()[3];
+    const int layer_height = grid_h_ ? grid_h_ : getParentEdgeAt(INPUT_FEATUREMAP)->getDims()[2];
+    const float step_w = stride_w_ ? stride_w_ : static_cast<float>(getParentEdgeAt(INPUT_IMAGE)->getDims()[3]) / layer_width;
+    const float step_h = stride_h_ ? stride_h_ : static_cast<float>(getParentEdgeAt(INPUT_IMAGE)->getDims()[2]) / layer_height;
+
+    const auto *bottom_data_0 = reinterpret_cast<const float *>(getParentEdgeAt(0)->getMemoryPtr()->GetPtr());
+    auto *top_data_0 = reinterpret_cast<float *>(getChildEdgesAtPort(OUTPUT_ROIS)[0]->getMemoryPtr()->GetPtr());
+
+    for (int h = 0; h < layer_height; ++h) {
+        for (int w = 0; w < layer_width; ++w) {
+            for (int s = 0; s < num_priors_; ++s) {
+                top_data_0[0] = bottom_data_0[4 * s + 0] + step_w * (w + 0.5f);
+                top_data_0[1] = bottom_data_0[4 * s + 1] + step_h * (h + 0.5f);
+                top_data_0[2] = bottom_data_0[4 * s + 2] + step_w * (w + 0.5f);
+                top_data_0[3] = bottom_data_0[4 * s + 3] + step_h * (h + 0.5f);
+                top_data_0 += 4;
+            }
+        }
+    }
+}
+
+bool MKLDNNExperimentalDetectronPriorGridGeneratorNode::created() const {
+    return getType() == ExperimentalDetectronPriorGridGenerator;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNExperimentalDetectronPriorGridGeneratorNode, ExperimentalDetectronPriorGridGenerator)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_priorgridgenerator_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_priorgridgenerator_node.h
new file mode 100644
index 00000000000000..9ef117f44e65f7
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_priorgridgenerator_node.h
@@ -0,0 +1,46 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNExperimentalDetectronPriorGridGeneratorNode : public MKLDNNNode {
+public:
+    MKLDNNExperimentalDetectronPriorGridGeneratorNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    // Inputs:
+    //      priors, shape [n, 4]
+    //      [feature_map], shape [b, c, h, w]
+    //      [im_data], shape [b, 3, im_h, im_w]
+    // Outputs:
+    //      priors_grid, shape [m, 4]
+
+    const int INPUT_PRIORS {0};
+    const int INPUT_FEATUREMAP {1};
+    const int INPUT_IMAGE {2};
+
+    const int OUTPUT_ROIS {0};
+
+    int grid_w_;
+    int grid_h_;
+    float stride_w_;
+    float stride_h_;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_roifeatureextractor_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_roifeatureextractor_node.cpp
new file mode 100644
index 00000000000000..94e7f033a95548
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_roifeatureextractor_node.cpp
@@ -0,0 +1,413 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+#include <algorithm>
+
+#include <ngraph/opsets/opset6.hpp>
+#include "ie_parallel.hpp"
+#include "common/cpu_memcpy.h"
+#include "mkldnn_experimental_detectron_roifeatureextractor_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+// implementation taken from Caffe2
+template <typename T>
+struct PreCalc {
+    int pos1;
+    int pos2;
+    int pos3;
+    int pos4;
+    T w1;
+    T w2;
+    T w3;
+    T w4;
+};
+
+template <typename T>
+void pre_calc_for_bilinear_interpolate(
+        const int height,
+        const int width,
+        const int pooled_height,
+        const int pooled_width,
+        const int iy_upper,
+        const int ix_upper,
+        T roi_start_h,
+        T roi_start_w,
+        T bin_size_h,
+        T bin_size_w,
+        int roi_bin_grid_h,
+        int roi_bin_grid_w,
+        std::vector<PreCalc<T>>& pre_calc) {
+    int pre_calc_index = 0;
+    for (int ph = 0; ph < pooled_height; ph++) {
+        for (int pw = 0; pw < pooled_width; pw++) {
+            for (int iy = 0; iy < iy_upper; iy++) {
+                const T yy = roi_start_h + ph * bin_size_h +
+                             static_cast<T>(iy + .5f) * bin_size_h /
+                             static_cast<T>(roi_bin_grid_h);  // e.g., 0.5, 1.5
+                for (int ix = 0; ix < ix_upper; ix++) {
+                    const T xx = roi_start_w + pw * bin_size_w +
+                                 static_cast<T>(ix + .5f) * bin_size_w /
+                                 static_cast<T>(roi_bin_grid_w);
+
+                    T x = xx;
+                    T y = yy;
+                    // deal with: inverse elements are out of feature map boundary
+                    if (y < -1.0 || y > height || x < -1.0 || x > width) {
+                        // empty
+                        PreCalc<T> pc;
+                        pc.pos1 = 0;
+                        pc.pos2 = 0;
+                        pc.pos3 = 0;
+                        pc.pos4 = 0;
+                        pc.w1 = 0;
+                        pc.w2 = 0;
+                        pc.w3 = 0;
+                        pc.w4 = 0;
+                        pre_calc.at(pre_calc_index) = pc;
+                        pre_calc_index += 1;
+                        continue;
+                    }
+
+                    if (y <= 0) {
+                        y = 0;
+                    }
+                    if (x <= 0) {
+                        x = 0;
+                    }
+
+                    int y_low = static_cast<int>(y);
+                    int x_low = static_cast<int>(x);
+                    int y_high = 0;
+                    int x_high = 0;
+
+                    if (y_low >= height - 1) {
+                        y_high = y_low = height - 1;
+                        y = (T)y_low;
+                    } else {
+                        y_high = y_low + 1;
+                    }
+
+                    if (x_low >= width - 1) {
+                        x_high = x_low = width - 1;
+                        x = (T)x_low;
+                    } else {
+                        x_high = x_low + 1;
+                    }
+
+                    T ly = y - y_low;
+                    T lx = x - x_low;
+                    T hy = static_cast<T>(1) - ly, hx = static_cast<T>(1) - lx;
+                    T w1 = hy * hx, w2 = hy * lx, w3 = ly * hx, w4 = ly * lx;
+
+                    // save weights and indeces
+                    PreCalc<T> pc;
+                    pc.pos1 = y_low * width + x_low;
+                    pc.pos2 = y_low * width + x_high;
+                    pc.pos3 = y_high * width + x_low;
+                    pc.pos4 = y_high * width + x_high;
+                    pc.w1 = w1;
+                    pc.w2 = w2;
+                    pc.w3 = w3;
+                    pc.w4 = w4;
+                    pre_calc[pre_calc_index] = pc;
+
+                    pre_calc_index += 1;
+                }
+            }
+        }
+    }
+}
+
+template <typename T>
+void ROIAlignForward_cpu_kernel(
+        const int nthreads,
+        const T* bottom_data,
+        const T& spatial_scale,
+        const int channels,
+        const int height,
+        const int width,
+        const int pooled_height,
+        const int pooled_width,
+        const int sampling_ratio,
+        const T* bottom_rois,
+        const bool aligned,
+        T* top_data) {
+    int roi_cols = 4;
+
+    int n_rois = nthreads / channels / pooled_width / pooled_height;
+    // (n, c, ph, pw) is an element in the pooled output
+    parallel_for(n_rois, [&](size_t n) {
+        int index_n = n * channels * pooled_width * pooled_height;
+
+        // roi could have 4 or 5 columns
+        const T* offset_bottom_rois = bottom_rois + n * roi_cols;
+        int roi_batch_ind = 0;
+        if (roi_cols == 5) {
+            roi_batch_ind = static_cast<int>(offset_bottom_rois[0]);
+            offset_bottom_rois++;
+        }
+
+        T offset = aligned ? (T)0.5 : (T)0.0;
+        // Do not using rounding; this implementation detail is critical
+        T roi_start_w = offset_bottom_rois[0] * spatial_scale - offset;
+        T roi_start_h = offset_bottom_rois[1] * spatial_scale - offset;
+        T roi_end_w = offset_bottom_rois[2] * spatial_scale - offset;
+        T roi_end_h = offset_bottom_rois[3] * spatial_scale - offset;
+
+        // Force malformed ROIs to be 1x1
+        T roi_width = (std::max)(roi_end_w - roi_start_w, (T)1.);
+        T roi_height = (std::max)(roi_end_h - roi_start_h, (T)1.);
+        T bin_size_h = static_cast<T>(roi_height) / static_cast<T>(pooled_height);
+        T bin_size_w = static_cast<T>(roi_width) / static_cast<T>(pooled_width);
+
+        // We use roi_bin_grid to sample the grid and mimic integral
+        int roi_bin_grid_h = (sampling_ratio > 0)
+                             ? sampling_ratio
+                             : static_cast<int>(ceil(roi_height / pooled_height));  // e.g., = 2
+        int roi_bin_grid_w =
+                (sampling_ratio > 0) ? sampling_ratio : static_cast<int>(ceil(roi_width / pooled_width));
+
+        // We do average (integral) pooling inside a bin
+        const T count = static_cast<T>(roi_bin_grid_h * roi_bin_grid_w);  // e.g. = 4
+
+        // we want to precalculate indeces and weights shared by all chanels,
+        // this is the key point of optimiation
+        std::vector<PreCalc<T>> pre_calc(
+                roi_bin_grid_h * roi_bin_grid_w * pooled_width * pooled_height);
+        pre_calc_for_bilinear_interpolate(
+                height,
+                width,
+                pooled_height,
+                pooled_width,
+                roi_bin_grid_h,
+                roi_bin_grid_w,
+                roi_start_h,
+                roi_start_w,
+                bin_size_h,
+                bin_size_w,
+                roi_bin_grid_h,
+                roi_bin_grid_w,
+                pre_calc);
+
+        for (int c = 0; c < channels; c++) {
+            int index_n_c = index_n + c * pooled_width * pooled_height;
+            const T* offset_bottom_data =
+                    bottom_data + (roi_batch_ind * channels + c) * height * width;
+            int pre_calc_index = 0;
+
+            for (int ph = 0; ph < pooled_height; ph++) {
+                for (int pw = 0; pw < pooled_width; pw++) {
+                    int index = index_n_c + ph * pooled_width + pw;
+
+                    T output_val = 0.;
+                    for (int iy = 0; iy < roi_bin_grid_h; iy++) {
+                        for (int ix = 0; ix < roi_bin_grid_w; ix++) {
+                            PreCalc<T> pc = pre_calc[pre_calc_index];
+                            output_val += pc.w1 * offset_bottom_data[pc.pos1] +
+                                          pc.w2 * offset_bottom_data[pc.pos2] +
+                                          pc.w3 * offset_bottom_data[pc.pos3] +
+                                          pc.w4 * offset_bottom_data[pc.pos4];
+
+                            pre_calc_index += 1;
+                        }
+                    }
+                    output_val /= count;
+
+                    top_data[index] = output_val;
+                }  // for pw
+            }  // for ph
+        }  // for c
+    });
+}
+
+
+void redistribute_rois(const float* rois, int* level_ids,
+                       const int num_rois, const int levels_num) {
+    const float canonical_scale = 224.0f;
+    const int canonical_level = 2;
+
+    for (int i = 0; i < num_rois; ++i) {
+        const float x0 = rois[4 * i + 0];
+        const float y0 = rois[4 * i + 1];
+        const float x1 = rois[4 * i + 2];
+        const float y1 = rois[4 * i + 3];
+
+        int target_level = levels_num;
+        float area = (x1 - x0) * (y1 - y0);
+        if (area > 0) {
+            area = std::sqrt(area) / canonical_scale;
+            area = std::log2(area + 1e-6f);
+            target_level = static_cast<int>(std::floor(area + canonical_level));
+            target_level = (std::max)(0, (std::min)(levels_num - 1, target_level));
+        }
+
+        level_ids[i] = target_level;
+    }
+}
+
+
+void reord(const float* src_data, const int* ranks, const int n, const int step, float* dst_data,
+             int* dst_mapping) {
+    std::iota(dst_mapping, dst_mapping + n, 0);
+    std::sort(dst_mapping, dst_mapping + n, [&ranks](size_t i1, size_t i2) {return ranks[i1] < ranks[i2];});
+    for (int i = 0; i < n; ++i) {
+        const int j = dst_mapping[i];
+        assert(0 <= j && j < n);
+        cpu_memcpy(dst_data + i * step, src_data + j * step, sizeof(float) * step);
+    }
+}
+
+void split_points(const std::vector<int>& ids, std::vector<int>& rois_per_level, const int levels_num) {
+    rois_per_level.clear();
+    rois_per_level.resize(levels_num, 0);
+    for (size_t i = 0; i < ids.size(); ++i) {
+        assert(0 <= ids[i] && ids[i] < levels_num);
+        rois_per_level[ids[i]]++;
+    }
+    for (int i = 1; i < levels_num; ++i) {
+        rois_per_level[i] += rois_per_level[i - 1];
+    }
+    rois_per_level.insert(rois_per_level.begin(), 0);
+}
+
+
+void reorder_rois(const float *rois, const int* ids, int* mapping, const int rois_num,
+                  float * reordered_rois, std::vector<int>& rois_per_level, const int levels_num) {
+    rois_per_level.clear();
+    rois_per_level.resize(levels_num, 0);
+    for (int i = 0; i < rois_num; ++i) {
+        assert(0 <= ids[i] && ids[i] < levels_num);
+        rois_per_level[ids[i]]++;
+    }
+    for (int i = 1; i < levels_num; ++i) {
+        rois_per_level[i] += rois_per_level[i - 1];
+    }
+    rois_per_level.insert(rois_per_level.begin(), 0);
+
+    std::vector<int> level_counter = rois_per_level;
+
+    for (int i = 0; i < rois_num; ++i) {
+        const int level = ids[i];
+        assert(level < levels_num);
+        const int j = level_counter[level];
+        assert(0 <= j && j < rois_num);
+        reordered_rois[j * 4 + 0] = rois[i * 4 + 0];
+        reordered_rois[j * 4 + 1] = rois[i * 4 + 1];
+        reordered_rois[j * 4 + 2] = rois[i * 4 + 2];
+        reordered_rois[j * 4 + 3] = rois[i * 4 + 3];
+        level_counter[level]++;
+    }
+}
+
+bool MKLDNNExperimentalDetectronROIFeatureExtractorNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto roiFeatureExtractor = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronROIFeatureExtractor>(op);
+        if (!roiFeatureExtractor) {
+            errorMessage = "Only opset6 ExperimentalDetectronROIFeatureExtractor operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNExperimentalDetectronROIFeatureExtractorNode::MKLDNNExperimentalDetectronROIFeatureExtractorNode
+        (const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+                MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    const auto roiFeatureExtractor = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronROIFeatureExtractor>(op);
+    const auto &attr = roiFeatureExtractor->get_attrs();
+    output_dim_ = attr.output_size;
+    pyramid_scales_ = attr.pyramid_scales;
+    sampling_ratio_ = attr.sampling_ratio;
+    aligned_ = attr.aligned;
+    pooled_height_ = output_dim_;
+    pooled_width_ = output_dim_;
+}
+
+void MKLDNNExperimentalDetectronROIFeatureExtractorNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    std::vector<DataConfigurator> inDataConf;
+    inDataConf.reserve(getOriginalInputsNumber());
+    for (int i = 0; i < getOriginalInputsNumber(); ++i)
+        inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::FP32);
+
+    addSupportedPrimDesc(inDataConf,
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNExperimentalDetectronROIFeatureExtractorNode::execute(mkldnn::stream strm) {
+    const int levels_num = inDims.size() - INPUT_FEATURES_START;
+    const int num_rois = getParentEdgeAt(INPUT_ROIS)->getDims()[0];
+    const int channels_num = getParentEdgeAt(INPUT_FEATURES_START)->getDims()[1];
+    const int feaxels_per_roi = pooled_height_ * pooled_width_ * channels_num;
+
+    auto *input_rois = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_ROIS)->getMemoryPtr()->GetPtr());
+    auto *output_rois_features = reinterpret_cast<float *>(getChildEdgesAtPort(OUTPUT_ROI_FEATURES)[0]->getMemoryPtr()->GetPtr());
+    float *output_rois = nullptr;
+    if (OUTPUT_ROIS < outDims.size()) {
+        output_rois = reinterpret_cast<float *>(getChildEdgesAtPort(OUTPUT_ROIS)[0]->getMemoryPtr()->GetPtr());
+    }
+
+    std::vector<int> level_ids(num_rois, 0);
+    redistribute_rois(input_rois, reinterpret_cast<int *>(&level_ids[0]), num_rois, levels_num);
+
+    std::vector<float> reordered_rois(4 * num_rois, 0);
+    std::vector<int> original_rois_mapping(num_rois, 0);
+    reord(input_rois, &level_ids[0], num_rois, 4, &reordered_rois[0], &original_rois_mapping[0]);
+
+    std::vector<int> rois_per_level;
+    split_points(level_ids, rois_per_level, levels_num + 1);
+
+    std::vector<float> output_rois_features_temp(feaxels_per_roi * num_rois, 0);
+    for (int i = 0; i < levels_num; ++i) {
+        const int level_rois_offset = rois_per_level[i];
+        const int level_rois_num = rois_per_level[i + 1] - level_rois_offset;
+        if (level_rois_num > 0) {
+            auto *featuremap = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_FEATURES_START + i)->getMemoryPtr()->GetPtr());
+            const int featuremap_height = getParentEdgeAt(INPUT_FEATURES_START + i)->getDims()[2];
+            const int featuremap_width = getParentEdgeAt(INPUT_FEATURES_START + i)->getDims()[3];
+            ROIAlignForward_cpu_kernel<float>(feaxels_per_roi * level_rois_num,
+                                              featuremap,
+                                              1.0f / pyramid_scales_[i],
+                                              channels_num,
+                                              featuremap_height,
+                                              featuremap_width,
+                                              pooled_height_,
+                                              pooled_width_,
+                                              sampling_ratio_,
+                                              &reordered_rois[4 * level_rois_offset],
+                                              aligned_,
+                                              &output_rois_features_temp[feaxels_per_roi * level_rois_offset]);
+        }
+    }
+
+    std::vector<int> dummy_mapping(num_rois, 0);
+    reord(&output_rois_features_temp[0], &original_rois_mapping[0], num_rois, feaxels_per_roi,
+            output_rois_features, &dummy_mapping[0]);
+    if (output_rois != nullptr) {
+        cpu_memcpy(output_rois, input_rois, 4 * num_rois * sizeof(float));
+    }
+}
+
+bool MKLDNNExperimentalDetectronROIFeatureExtractorNode::created() const {
+    return getType() == ExperimentalDetectronROIFeatureExtractor;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNExperimentalDetectronROIFeatureExtractorNode, ExperimentalDetectronROIFeatureExtractor)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_roifeatureextractor_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_roifeatureextractor_node.h
new file mode 100644
index 00000000000000..bfcb9061f26fbe
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_roifeatureextractor_node.h
@@ -0,0 +1,41 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNExperimentalDetectronROIFeatureExtractorNode : public MKLDNNNode {
+public:
+    MKLDNNExperimentalDetectronROIFeatureExtractorNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    const int INPUT_ROIS {0};
+    const int INPUT_FEATURES_START {1};
+
+    const int OUTPUT_ROI_FEATURES {0};
+    const int OUTPUT_ROIS {1};
+
+    int output_dim_ = 0;
+    int pooled_height_ = 0;
+    int pooled_width_ = 0;
+    std::vector<int64_t> pyramid_scales_;
+    int sampling_ratio_ = 0;
+    bool aligned_ = false;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_topkrois_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_topkrois_node.cpp
new file mode 100644
index 00000000000000..d543658f78e724
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_topkrois_node.cpp
@@ -0,0 +1,82 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+#include <algorithm>
+
+#include <ngraph/opsets/opset6.hpp>
+#include "ie_parallel.hpp"
+#include "common/cpu_memcpy.h"
+#include "mkldnn_experimental_detectron_topkrois_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNExperimentalDetectronTopKROIsNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto topKROI = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronTopKROIs>(op);
+        if (!topKROI) {
+            errorMessage = "Only opset6 ExperimentalDetectronTopKROIs operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNExperimentalDetectronTopKROIsNode::MKLDNNExperimentalDetectronTopKROIsNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "ExperimentalDetectronTopKROIs layer with name '" + op->get_friendly_name() + "'";
+    const auto topKROI = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronTopKROIs>(op);
+    if (getOriginalInputsNumber() != 2 || getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
+
+    if (op->get_input_shape(INPUT_ROIS).size() != 2 || op->get_input_shape(INPUT_PROBS).size() != 1)
+        IE_THROW() << errorPrefix << " has nsupported input shape";
+
+    max_rois_num_ = topKROI->get_max_rois();
+}
+
+void MKLDNNExperimentalDetectronTopKROIsNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNExperimentalDetectronTopKROIsNode::execute(mkldnn::stream strm) {
+    const int input_rois_num = getParentEdgeAt(INPUT_ROIS)->getDims()[0];
+    const int top_rois_num = (std::min)(max_rois_num_, input_rois_num);
+
+    auto *input_rois = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_ROIS)->getMemoryPtr()->GetPtr());
+    auto *input_probs = reinterpret_cast<const float *>(getParentEdgeAt(INPUT_PROBS)->getMemoryPtr()->GetPtr());
+    auto *output_rois = reinterpret_cast<float *>(getChildEdgesAtPort(OUTPUT_ROIS)[0]->getMemoryPtr()->GetPtr());
+
+    std::vector<size_t> idx(input_rois_num);
+    iota(idx.begin(), idx.end(), 0);
+    // FIXME. partial_sort is enough here.
+    sort(idx.begin(), idx.end(), [&input_probs](size_t i1, size_t i2) {return input_probs[i1] > input_probs[i2];});
+
+    for (int i = 0; i < top_rois_num; ++i) {
+        cpu_memcpy(output_rois + 4 * i, input_rois + 4 * idx[i], 4 * sizeof(float));
+    }
+}
+
+bool MKLDNNExperimentalDetectronTopKROIsNode::created() const {
+    return getType() == ExperimentalDetectronTopKROIs;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNExperimentalDetectronTopKROIsNode, ExperimentalDetectronTopKROIs)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_topkrois_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_topkrois_node.h
new file mode 100644
index 00000000000000..76171de71e473c
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_experimental_detectron_topkrois_node.h
@@ -0,0 +1,40 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNExperimentalDetectronTopKROIsNode : public MKLDNNNode {
+public:
+    MKLDNNExperimentalDetectronTopKROIsNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    // Inputs:
+    //      rois, shape [n, 4]
+    //      rois_probs, shape [n]
+    // Outputs:
+    //      top_rois, shape [max_rois, 4]
+
+    const int INPUT_ROIS {0};
+    const int INPUT_PROBS {1};
+
+    const int OUTPUT_ROIS {0};
+    int max_rois_num_;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/extract_image_patches.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_extract_image_patches_node.cpp
similarity index 66%
rename from inference-engine/src/mkldnn_plugin/nodes/extract_image_patches.cpp
rename to inference-engine/src/mkldnn_plugin/nodes/mkldnn_extract_image_patches_node.cpp
index b0f0aa5d327ed8..d4c5d3037962b0 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/extract_image_patches.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_extract_image_patches_node.cpp
@@ -1,22 +1,22 @@
-// Copyright (C) 2020-2021 Intel Corporation
+// Copyright (C) 2018-2021 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "extract_image_patches.hpp"
-#include "caseless.hpp"
-#include "ie_parallel.hpp"
-#include "list.hpp"
-#include <cpu/x64/jit_generator.hpp>
+#include "base.hpp"
+
 #include <cstring>
 #include <string>
 #include <cmath>
+
 #include <ngraph/opsets/opset3.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_extract_image_patches_node.h"
+#include "list.hpp"
+#include <cpu/x64/jit_generator.hpp>
+#include "caseless.hpp"
 
 using namespace MKLDNNPlugin;
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
+using namespace InferenceEngine;
 
 using details::CaselessEq;
 
@@ -266,11 +266,11 @@ struct jit_extract_image_patches_kernel : public jit_uni_extract_image_patches_k
         align(64);
         L(gather_index_table);
         for (int32_t i = 0; i < vlen / sizeof(int32_t); i++)
-                dd(i * jpp.SW * jpp.dtype_size);
+            dd(i * jpp.SW * jpp.dtype_size);
     }
 };
 
-bool ExtractImagePatchesImpl::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+bool MKLDNNExtractImagePatchesNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
     try {
         const auto extImgPatcher = std::dynamic_pointer_cast<const ngraph::opset3::ExtractImagePatches>(op);
         if (!extImgPatcher) {
@@ -292,140 +292,141 @@ bool ExtractImagePatchesImpl::isSupportedOperation(const std::shared_ptr<ngraph:
     return true;
 }
 
-ExtractImagePatchesImpl::ExtractImagePatchesImpl(const std::shared_ptr<ngraph::Node>& op) {
-    try {
-        std::string errorMessage;
-        if (!isSupportedOperation(op, errorMessage)) {
-            IE_THROW(NotImplemented) << errorMessage;
-        }
+MKLDNNExtractImagePatchesNode::MKLDNNExtractImagePatchesNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
 
-        errorPrefix = "ExtractImagePatches layer with name '" + op->get_friendly_name() + "' ";
-        const auto extImgPatcher = std::dynamic_pointer_cast<const ngraph::opset3::ExtractImagePatches>(op);
+    errorPrefix = "ExtractImagePatches layer with name '" + op->get_friendly_name() + "' ";
+    const auto extImgPatcher = std::dynamic_pointer_cast<const ngraph::opset3::ExtractImagePatches>(op);
 
-        if (op->get_input_size() != 1 || op->get_output_size() != 1)
-                IE_THROW() << errorPrefix << "has incorrect number of input or output edges!"
-                           << " Input: " << op->get_input_size() << "; Output: " << op->get_output_size();
-
-        if (op->get_input_shape(0).size() != 4)
-                IE_THROW() << errorPrefix << "must have 4D input tensor. Actual: " << op->get_input_shape(0).size();
-
-        if (op->get_output_shape(0).size() != 4)
-            IE_THROW() << errorPrefix << "must have 4D output tensor. Actual: " << op->get_output_shape(0).size();
-
-        const auto precision = details::convertPrecision(op->get_input_element_type(0));
-            if (_supported_precisions_sizes.find(precision.size()) == _supported_precisions_sizes.end())
-                IE_THROW() << errorPrefix << "has unsupported precision: " << precision.name();
-
-        auto ksizes = extImgPatcher->get_sizes();
-        auto strides = extImgPatcher->get_strides();
-        auto rates = extImgPatcher->get_rates();
-        if (extImgPatcher->get_auto_pad() == ngraph::op::PadType::VALID) {
-            _auto_pad = ExtImgPatcherPadType::VALID;
-        } else if (extImgPatcher->get_auto_pad() == ngraph::op::PadType::SAME_LOWER) {
-            _auto_pad = ExtImgPatcherPadType::SAME_LOWER;
-        } else if (extImgPatcher->get_auto_pad() == ngraph::op::PadType::SAME_UPPER) {
-            _auto_pad = ExtImgPatcherPadType::SAME_UPPER;
-        } else {
-            IE_THROW() << errorPrefix << "has unsupported pad type: " << extImgPatcher->get_auto_pad();
-        }
+    if (getOriginalInputsNumber() != 1 || getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << "has incorrect number of input or output edges!"
+                   << " Input: " << getOriginalInputsNumber() << "; Output: " << getOriginalOutputsNumber();
+
+    if (op->get_input_shape(0).size() != 4)
+        IE_THROW() << errorPrefix << "must have 4D input tensor. Actual: " << op->get_input_shape(0).size();
 
-        if (ksizes.size() != 2 || strides.size() != 2 || rates.size() != 2)
-            IE_THROW() << errorPrefix << "must have the following attributes with shape {2}: sizes, strides, rates.";
-        _ksizes.clear();
-        _strides.clear();
-        _rates.clear();
-        for (const auto& x :  ksizes) {
-            if (x < 0)
-                IE_THROW() << "Kernel sizes must be non-negative, got '" << x << "'.";
-            _ksizes.push_back(static_cast<size_t>(x));
+    if (op->get_output_shape(0).size() != 4)
+        IE_THROW() << errorPrefix << "must have 4D output tensor. Actual: " << op->get_output_shape(0).size();
+
+    auto ksizes = extImgPatcher->get_sizes();
+    auto strides = extImgPatcher->get_strides();
+    auto rates = extImgPatcher->get_rates();
+    if (extImgPatcher->get_auto_pad() == ngraph::op::PadType::VALID) {
+        _auto_pad = ExtImgPatcherPadType::VALID;
+    } else if (extImgPatcher->get_auto_pad() == ngraph::op::PadType::SAME_LOWER) {
+        _auto_pad = ExtImgPatcherPadType::SAME_LOWER;
+    } else if (extImgPatcher->get_auto_pad() == ngraph::op::PadType::SAME_UPPER) {
+        _auto_pad = ExtImgPatcherPadType::SAME_UPPER;
+    } else {
+        IE_THROW() << errorPrefix << "has unsupported pad type: " << extImgPatcher->get_auto_pad();
+    }
+
+    if (ksizes.size() != 2 || strides.size() != 2 || rates.size() != 2)
+        IE_THROW() << errorPrefix << "must have the following attributes with shape {2}: sizes, strides, rates.";
+    _ksizes.clear();
+    _strides.clear();
+    _rates.clear();
+    for (const auto& x :  ksizes) {
+        if (x < 0)
+            IE_THROW() << "Kernel sizes must be non-negative, got '" << x << "'.";
+        _ksizes.push_back(static_cast<size_t>(x));
+    }
+    for (const auto& x :  strides) {
+        if (x < 0)
+            IE_THROW() << "Strides must be non-negative, got '" << x << "'.";
+        _strides.push_back(static_cast<size_t>(x));
+    }
+    for (const auto& x :  rates) {
+        if (x < 0)
+            IE_THROW() << "Rates must be non-negative, got '" << x << "'.";
+        _rates.push_back(static_cast<size_t>(x));
+    }
+
+    SizeVector in_dims = op->get_input_shape(0);
+    _pad_left = 0;
+    _pad_top = 0;
+    jit_extract_image_patches_params jpp;
+    jpp.need_padding = false;
+    if (_auto_pad != ExtImgPatcherPadType::VALID) {
+        const size_t iheight = in_dims[2];
+        const size_t iwidth = in_dims[3];
+        const int64_t ihStep = _ksizes[0] + (_rates[0] - 1) * (_ksizes[0] - 1);
+        const int64_t iwStep = _ksizes[1] + (_rates[1] - 1) * (_ksizes[1] - 1);
+
+        int64_t PW = (std::ceil(1.f * iwidth/_strides[1]) - 1) * _strides[1] + iwStep - iwidth;
+        int64_t PH = (std::ceil(1.f * iheight/_strides[0]) - 1) * _strides[0] + ihStep - iheight;
+
+        int64_t increment_sign = 0;
+        if (_auto_pad == ExtImgPatcherPadType::SAME_LOWER) {
+            increment_sign = 1;
+        } else if (_auto_pad == ExtImgPatcherPadType::SAME_UPPER) {
+            increment_sign = -1;
         }
-        for (const auto& x :  strides) {
-            if (x < 0)
-                IE_THROW() << "Strides must be non-negative, got '" << x << "'.";
-            _strides.push_back(static_cast<size_t>(x));
+
+        if ((PW > 0) && (PW < iwStep)) {
+            _pad_left = static_cast<size_t>((PW + increment_sign * (PW % 2)) / 2);
+            jpp.need_padding = true;
         }
-        for (const auto& x :  rates) {
-            if (x < 0)
-                IE_THROW() << "Rates must be non-negative, got '" << x << "'.";
-            _rates.push_back(static_cast<size_t>(x));
+        if ((PH > 0) && (PH < ihStep)) {
+            _pad_top = static_cast<size_t>((PH + increment_sign * (PH % 2)) / 2);
+            jpp.need_padding = true;
         }
+    }
 
-        SizeVector in_dims = op->get_input_shape(0);
-        _pad_left = 0;
-        _pad_top = 0;
-        jit_extract_image_patches_params jpp;
-        jpp.need_padding = false;
-        if (_auto_pad != ExtImgPatcherPadType::VALID) {
-            const size_t iheight = in_dims[2];
-            const size_t iwidth = in_dims[3];
-            const int64_t ihStep = _ksizes[0] + (_rates[0] - 1) * (_ksizes[0] - 1);
-            const int64_t iwStep = _ksizes[1] + (_rates[1] - 1) * (_ksizes[1] - 1);
-
-            int64_t PW = (std::ceil(1.f * iwidth/_strides[1]) - 1) * _strides[1] + iwStep - iwidth;
-            int64_t PH = (std::ceil(1.f * iheight/_strides[0]) - 1) * _strides[0] + ihStep - iheight;
-
-            int64_t increment_sign = 0;
-            if (_auto_pad == ExtImgPatcherPadType::SAME_LOWER) {
-                increment_sign = 1;
-            } else if (_auto_pad == ExtImgPatcherPadType::SAME_UPPER) {
-                increment_sign = -1;
-            }
+    jpp.IW = in_dims[3];
+    SizeVector out_dims = op->get_output_shape(0);
+    jpp.OH = out_dims[2];
+    jpp.OW = out_dims[3];
+    jpp.KH = _ksizes[0];
+    jpp.KW = _ksizes[1];
+    jpp.SH = _strides[0];
+    jpp.SW = _strides[1];
+    jpp.dtype_size = getOriginalInputPrecisionAtPort(0).size();
+    jpp.block_size = 1;
+
+    if (mayiuse(x64::avx512_common)) {
+        jpp.block_size = cpu_isa_traits<x64::avx512_common>::vlen / jpp.dtype_size;
+        extract_image_patches_kernel.reset(new jit_extract_image_patches_kernel<x64::avx512_common>(jpp));
+    } else if (mayiuse(x64::avx2)) {
+        jpp.block_size = cpu_isa_traits<x64::avx2>::vlen / jpp.dtype_size;
+        extract_image_patches_kernel.reset(new jit_extract_image_patches_kernel<x64::avx2>(jpp));
+    } else if (mayiuse(x64::sse41)) {
+        jpp.block_size = cpu_isa_traits<x64::sse41>::vlen / jpp.dtype_size;
+        extract_image_patches_kernel.reset(new jit_extract_image_patches_kernel<x64::sse41>(jpp));
+    }
 
-            if ((PW > 0) && (PW < iwStep)) {
-                _pad_left = static_cast<size_t>((PW + increment_sign * (PW % 2)) / 2);
-                jpp.need_padding = true;
-            }
-            if ((PH > 0) && (PH < ihStep)) {
-                _pad_top = static_cast<size_t>((PH + increment_sign * (PH % 2)) / 2);
-                jpp.need_padding = true;
-            }
-        }
+    if (extract_image_patches_kernel)
+        extract_image_patches_kernel->create_ker();
+}
 
-        jpp.IW = in_dims[3];
-        SizeVector out_dims = op->get_output_shape(0);
-        jpp.OH = out_dims[2];
-        jpp.OW = out_dims[3];
-        jpp.KH = _ksizes[0];
-        jpp.KW = _ksizes[1];
-        jpp.SH = _strides[0];
-        jpp.SW = _strides[1];
-        jpp.dtype_size = precision.size();
-        jpp.block_size = 1;
-
-        if (mayiuse(x64::avx512_common)) {
-            jpp.block_size = cpu_isa_traits<x64::avx512_common>::vlen / jpp.dtype_size;
-            extract_image_patches_kernel.reset(new jit_extract_image_patches_kernel<x64::avx512_common>(jpp));
-        } else if (mayiuse(x64::avx2)) {
-            jpp.block_size = cpu_isa_traits<x64::avx2>::vlen / jpp.dtype_size;
-            extract_image_patches_kernel.reset(new jit_extract_image_patches_kernel<x64::avx2>(jpp));
-        } else if (mayiuse(x64::sse41)) {
-            jpp.block_size = cpu_isa_traits<x64::sse41>::vlen / jpp.dtype_size;
-            extract_image_patches_kernel.reset(new jit_extract_image_patches_kernel<x64::sse41>(jpp));
-        }
+void MKLDNNExtractImagePatchesNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
 
-        if (extract_image_patches_kernel)
-            extract_image_patches_kernel->create_ker();
+    precision = getOriginalInputPrecisionAtPort(0);
+    if (_supported_precisions_sizes.find(precision.size()) == _supported_precisions_sizes.end())
+        IE_THROW() << errorPrefix << "has unsupported precision: " << precision.name();
 
-        addConfig(op, {{TensorDescCreatorTypes::ncsp, precision}},
-                      {{TensorDescCreatorTypes::ncsp, precision}});
-    } catch (InferenceEngine::Exception &ex) {
-        errorMsg = ex.what();
-    }
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, precision}},
+                         {{TensorDescCreatorTypes::ncsp, precision}},
+                         impl_desc_type::ref_any);
 }
 
-StatusCode ExtractImagePatchesImpl::execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs, ResponseDesc *resp) noexcept {
-    const char *src_data = inputs[0]->cbuffer().as<const char *>() +
-            inputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-    char *dst_data = outputs[0]->buffer().as<char *>() +
-            outputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-    const size_t dtype_size = inputs[0]->getTensorDesc().getPrecision().size();
+void MKLDNNExtractImagePatchesNode::execute(mkldnn::stream strm) {
+    const char *src_data = reinterpret_cast<const char *>(getParentEdgeAt(0)->getMemoryPtr()->GetPtr());
+    char *dst_data = reinterpret_cast<char *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+    const size_t dtype_size = getOriginalInputPrecisionAtPort(0).size();
 
-    const auto& inDims = inputs[0]->getTensorDesc().getDims();
+    const auto& inDims = getParentEdgeAt(0)->getDims().ToSizeVector();
     const size_t IC = inDims[1];
     const size_t IH = inDims[2];
     const size_t IW = inDims[3];
 
-    const auto& outDims = outputs[0]->getTensorDesc().getDims();
+    const auto& outDims = getChildEdgesAtPort(0)[0]->getDims().ToSizeVector();
     const size_t OB = outDims[0];
     const size_t OH = outDims[2];
     const size_t OW = outDims[3];
@@ -435,8 +436,8 @@ StatusCode ExtractImagePatchesImpl::execute(std::vector<Blob::Ptr>& inputs, std:
     const size_t RH = _rates[0], RW = _rates[1];
     const size_t PT = _pad_top, PL = _pad_left;
 
-    const std::vector<size_t> istrides = inputs[0]->getTensorDesc().getBlockingDesc().getStrides();
-    const std::vector<size_t> ostrides = outputs[0]->getTensorDesc().getBlockingDesc().getStrides();
+    const std::vector<size_t> istrides = getParentEdgeAt(0)->getDesc().getBlockingDesc().getStrides();
+    const std::vector<size_t> ostrides = getChildEdgesAtPort(0)[0]->getDesc().getBlockingDesc().getStrides();
     const std::vector<size_t> ostrides_partial = {ostrides[0], KW * IC * ostrides[1], IC * ostrides[1], ostrides[1]};
 
     if (extract_image_patches_kernel) {
@@ -471,7 +472,7 @@ StatusCode ExtractImagePatchesImpl::execute(std::vector<Blob::Ptr>& inputs, std:
             const size_t iw_hpad = std::ceil((IW - 1.f * iw_start) / SW) > OW ? OW : std::ceil((IW - 1.f * iw_start) / SW);
 
             char *my_dst_ptr = dst_data +
-                    (ob * ostrides_partial[0] + kh * ostrides_partial[1] + kw * ostrides_partial[2] + ic * ostrides_partial[3]) * dtype_size;
+                               (ob * ostrides_partial[0] + kh * ostrides_partial[1] + kw * ostrides_partial[2] + ic * ostrides_partial[3]) * dtype_size;
             const char *my_src_ptr = src_data + (ob * istrides[0] + ic * istrides[1] + ih_start * istrides[2] + iw_start) * dtype_size;
 
             size_t num_bytes_to_set = ih_lpad * OW * dtype_size;
@@ -480,14 +481,14 @@ StatusCode ExtractImagePatchesImpl::execute(std::vector<Blob::Ptr>& inputs, std:
 
             const char* src_ptr_h_stop = my_src_ptr + ih_hpad * SH * IW * dtype_size;
             for (const char *src_h_ptr = my_src_ptr + ih_lpad * SH * IW * dtype_size;
-                src_h_ptr < src_ptr_h_stop; src_h_ptr += SH * IW * dtype_size) {
+                 src_h_ptr < src_ptr_h_stop; src_h_ptr += SH * IW * dtype_size) {
                 num_bytes_to_set = iw_lpad * dtype_size;
                 memset(my_dst_ptr, 0, num_bytes_to_set);
                 my_dst_ptr += num_bytes_to_set;
 
                 const char* src_ptr_w_stop = src_h_ptr + iw_hpad * SW * dtype_size;
                 for (const char* src_w_ptr = src_h_ptr + iw_lpad * SW * dtype_size;
-                    src_w_ptr < src_ptr_w_stop; src_w_ptr += SW * dtype_size) {
+                     src_w_ptr < src_ptr_w_stop; src_w_ptr += SW * dtype_size) {
                     num_bytes_to_set = dtype_size;
                     memcpy(my_dst_ptr, src_w_ptr, num_bytes_to_set);
                     my_dst_ptr += num_bytes_to_set;
@@ -500,11 +501,12 @@ StatusCode ExtractImagePatchesImpl::execute(std::vector<Blob::Ptr>& inputs, std:
             memset(my_dst_ptr, 0, num_bytes_to_set);
         });
     }
-    return OK;
 }
 
-const std::set<size_t> ExtractImagePatchesImpl::_supported_precisions_sizes = {1, 2, 4};
+const std::set<size_t> MKLDNNExtractImagePatchesNode::_supported_precisions_sizes = {1, 2, 4};
+
+bool MKLDNNExtractImagePatchesNode::created() const {
+    return getType() == ExtractImagePatches;
+}
 
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
+REG_MKLDNN_PRIM_FOR(MKLDNNExtractImagePatchesNode, ExtractImagePatches)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/extract_image_patches.hpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_extract_image_patches_node.h
similarity index 64%
rename from inference-engine/src/mkldnn_plugin/nodes/extract_image_patches.hpp
rename to inference-engine/src/mkldnn_plugin/nodes/mkldnn_extract_image_patches_node.h
index 8ed62fbca89b0d..2990b12d08f2e3 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/extract_image_patches.hpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_extract_image_patches_node.h
@@ -1,16 +1,16 @@
-// Copyright (C) 2021 Intel Corporation
+// Copyright (C) 2018-2021 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
+
 #pragma once
 
-#include "base.hpp"
+#include <ie_common.h>
+#include <mkldnn_node.h>
+#include <string>
+#include <memory>
 #include <vector>
-#include <set>
-#include <cassert>
 
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
+namespace MKLDNNPlugin {
 
 struct jit_extract_image_patches_params {
     size_t IW;
@@ -40,12 +40,17 @@ struct jit_uni_extract_image_patches_kernel {
     virtual ~jit_uni_extract_image_patches_kernel() {}
 };
 
-
-class ExtractImagePatchesImpl : public ExtLayerBase {
+class MKLDNNExtractImagePatchesNode : public MKLDNNNode {
 public:
-    explicit ExtractImagePatchesImpl(const std::shared_ptr<ngraph::Node>& op);
-    StatusCode execute(std::vector<Blob::Ptr>&, std::vector<Blob::Ptr>&, ResponseDesc*) noexcept override;
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+    MKLDNNExtractImagePatchesNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
 
 private:
     enum class ExtImgPatcherPadType {
@@ -63,12 +68,9 @@ class ExtractImagePatchesImpl : public ExtLayerBase {
     static const std::set<size_t> _supported_precisions_sizes;
 
     ExtImgPatcherPadType _auto_pad;
+    InferenceEngine::Precision precision;
 
     std::string errorPrefix;
 };
 
-REG_FACTORY_FOR(ExtractImagePatchesImpl, ExtractImagePatches);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_elements_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_elements_node.cpp
index eabd4f52aac8b2..e3e14e356912db 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_elements_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_elements_node.cpp
@@ -18,7 +18,7 @@ using namespace InferenceEngine;
 
 bool MKLDNNGatherElementsNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
     try {
-        auto gatherElementsOp = ngraph::as_type_ptr<const ngraph::op::v6::GatherElements>(op);
+        const auto gatherElementsOp = ngraph::as_type_ptr<const ngraph::op::v6::GatherElements>(op);
         if (!gatherElementsOp) {
             errorMessage = "Node is not an instance of the GatherElements operation from operation set v6.";
             return false;
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_nd_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_nd_node.cpp
index 3e858dd309d8ca..ee7623f9b4810b 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_nd_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_nd_node.cpp
@@ -18,7 +18,7 @@ using namespace InferenceEngine;
 
 bool MKLDNNGatherNDNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
     try {
-        auto gatherElementsOp = ngraph::as_type_ptr<const ngraph::op::v5::GatherND>(op);
+        const auto gatherElementsOp = ngraph::as_type_ptr<const ngraph::op::v5::GatherND>(op);
         if (!gatherElementsOp) {
             errorMessage = "Node is not an instance of the GatherND operation from operation set v5.";
             return false;
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_node.cpp
index 3bd50aadf3357e..ade92f6a4a0060 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_node.cpp
@@ -15,13 +15,13 @@ using namespace InferenceEngine;
 
 bool MKLDNNGatherNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
     try {
-        auto gatherOp = ngraph::as_type_ptr<const ngraph::op::v7::Gather>(op);
+        const auto gatherOp = ngraph::as_type_ptr<const ngraph::op::v7::Gather>(op);
         if (!gatherOp) {
             errorMessage = "Only opset7 Gather operation is supported";
             return false;
         }
 
-        auto axesOp = gatherOp->get_input_node_shared_ptr(GATHER_AXIS);
+        const auto axesOp = gatherOp->get_input_node_shared_ptr(GATHER_AXIS);
         if (!ngraph::as_type_ptr<const ngraph::op::Constant>(axesOp)) {
             errorMessage = "Only Constant operation on 'axis' input is supported";
             return false;
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_tree_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_tree_node.cpp
new file mode 100644
index 00000000000000..ce396446df2418
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_tree_node.cpp
@@ -0,0 +1,148 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+#include <cmath>
+
+#include <ngraph/op/gather_tree.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_gather_tree_node.h"
+#include <utils/general_utils.h>
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNGatherTreeNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto gatherElementsOp = ngraph::as_type_ptr<const ngraph::op::v1::GatherTree>(op);
+        if (!gatherElementsOp) {
+            errorMessage = "Node is not an instance of the GatherTree operation from operation set v1.";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNGatherTreeNode::MKLDNNGatherTreeNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = std::string("Node GatherTree with name '") + op->get_friendly_name() + "'";
+    if (op->get_input_size() != 4)
+        IE_THROW() << errorPrefix << " has incorrect number of input edges.";
+    if (op->get_output_size() != 1)
+        IE_THROW() << errorPrefix << " has incorrect number of output edges.";
+
+    if (op->get_input_shape(GATHER_TREE_STEP_IDX).size() != 3)
+        IE_THROW() << errorPrefix << " step_idx vector should be 3 dimension";
+    if (op->get_input_shape(GATHER_TREE_PARENT_IDX).size() != 3)
+        IE_THROW() << errorPrefix << " parent_idx vector should be 3 dimension";
+    if (op->get_input_shape(GATHER_TREE_MAX_SEQ_LEN).size() != 1)
+        IE_THROW() << errorPrefix << " max_seq_len vector should be 1 dimension";
+    if (op->get_input_shape(GATHER_TREE_END_TOKEN).size() != 0)
+        IE_THROW() << errorPrefix << " end_token should be 1 dimension";
+}
+
+void MKLDNNGatherTreeNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    precision = getOriginalInputPrecisionAtPort(GATHER_TREE_STEP_IDX);
+    if (!MKLDNNPlugin::one_of(precision, Precision::FP32, Precision::I32))
+        precision = Precision::FP32;
+
+    if (getOriginalInputPrecisionAtPort(GATHER_TREE_PARENT_IDX)  != precision ||
+        getOriginalInputPrecisionAtPort(GATHER_TREE_MAX_SEQ_LEN) != precision ||
+        getOriginalInputPrecisionAtPort(GATHER_TREE_END_TOKEN)   != precision ||
+        getOriginalOutputPrecisionAtPort(0)                 != precision) {
+            IE_THROW() << errorPrefix << " has incorrect input/output data precision. Must be the same.";
+    }
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, precision},
+                            {TensorDescCreatorTypes::ncsp, precision},
+                            {TensorDescCreatorTypes::ncsp, precision},
+                            {TensorDescCreatorTypes::ncsp, precision}},
+                         {{TensorDescCreatorTypes::ncsp, precision}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNGatherTreeNode::execute(mkldnn::stream strm) {
+    if (precision == Precision::FP32)
+        return gatherTreeKernel<float>();
+    else
+        return gatherTreeKernel<int32_t>();
+}
+
+template<typename DATA_T>
+void MKLDNNGatherTreeNode::gatherTreeKernel() noexcept {
+    const auto *step_idx = reinterpret_cast<DATA_T *>(getParentEdgeAt(GATHER_TREE_STEP_IDX)->getMemoryPtr()->GetPtr());
+    const auto * const parent_idx = reinterpret_cast<DATA_T *>(getParentEdgeAt(GATHER_TREE_PARENT_IDX)->getMemoryPtr()->GetPtr());
+    const size_t parent_idx_size = getParentEdgeAt(GATHER_TREE_PARENT_IDX)->getDims().size()
+                                   - getParentEdgeAt(GATHER_TREE_PARENT_IDX)->getDesc().getBlockingDesc().getOffsetPadding();
+    const auto *max_seq_len = reinterpret_cast<DATA_T *>(getParentEdgeAt(GATHER_TREE_MAX_SEQ_LEN)->getMemoryPtr()->GetPtr());
+    auto end_token = (reinterpret_cast<DATA_T *>(getParentEdgeAt(GATHER_TREE_END_TOKEN)->getMemoryPtr()->GetPtr()))[0];
+    auto * final_idx = reinterpret_cast<DATA_T *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+
+    SizeVector step_idx_dims = getParentEdgeAt(GATHER_TREE_STEP_IDX)->getDims().ToSizeVector();
+    SizeVector parent_idx_dims = getParentEdgeAt(GATHER_TREE_PARENT_IDX)->getDims().ToSizeVector();
+    SizeVector max_seq_len_dims = getParentEdgeAt(GATHER_TREE_MAX_SEQ_LEN)->getDims().ToSizeVector();
+    SizeVector final_idx_dims = getChildEdgesAtPort(0)[0]->getDims().ToSizeVector();
+    int32_t max_time = step_idx_dims[0];
+    const size_t batch_size = step_idx_dims[1];
+    const size_t beam_width = step_idx_dims[2];
+    const size_t bb_size = batch_size * beam_width;
+
+    if (max_time != static_cast<int32_t>(parent_idx_dims[0]) || max_time != static_cast<int32_t>(final_idx_dims[0]) ||
+        batch_size != parent_idx_dims[1] || batch_size != final_idx_dims[1] || batch_size != max_seq_len_dims[0] ||
+        beam_width != parent_idx_dims[2] || beam_width != final_idx_dims[2]) {
+        std::string errorMsg = "Input/Output tensors dimensions mismatch";
+        IE_THROW() << errorMsg;
+    }
+
+    bool incorrect_result = false;
+    parallel_for2d(batch_size, beam_width, [&](size_t batch, size_t beam) {
+        int32_t max_sequence_in_beam = std::min<int32_t>(max_time, static_cast<int32_t>(max_seq_len[batch]));
+        if (max_sequence_in_beam > 0) {
+            int32_t time, idx = (max_time - 1) * bb_size + batch * beam_width;
+            for (time = (max_time - 1); time >= max_sequence_in_beam; time--, idx -= bb_size)
+                final_idx[idx + beam] = end_token;
+
+            for (int32_t parent = static_cast<int32_t>(beam); time >= 0; time--, idx -= bb_size) {
+                if (parent < 0 || parent >= static_cast<int32_t>(beam_width) || idx + parent >= parent_idx_size) {
+                    incorrect_result = true;
+                    break;
+                }
+                final_idx[idx + beam] = step_idx[idx + parent];
+                parent = static_cast<int32_t>(parent_idx[idx + parent]);
+            }
+
+            bool finished = false;
+            auto *final = &final_idx[batch * beam_width + beam];
+            for (time = 0; time < max_sequence_in_beam; time++, final += bb_size) {
+                if (finished)
+                    (*final) = end_token;
+                else if ((*final) == end_token)
+                    finished = true;
+            }
+        }
+    });
+
+    if (incorrect_result) {
+        std::string errorMsg = "Wrong parent index, result is incorrect";
+        IE_THROW() << errorMsg;
+    }
+}
+
+bool MKLDNNGatherTreeNode::created() const {
+    return getType() == GatherTree;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNGatherTreeNode, GatherTree)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_tree_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_tree_node.h
new file mode 100644
index 00000000000000..63f34fe6d6e685
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_gather_tree_node.h
@@ -0,0 +1,38 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNGatherTreeNode : public MKLDNNNode {
+public:
+    MKLDNNGatherTreeNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+    template<typename DATA_T>
+    void gatherTreeKernel() noexcept;
+
+        private:
+    static const size_t GATHER_TREE_STEP_IDX = 0;
+    static const size_t GATHER_TREE_PARENT_IDX = 1;
+    static const size_t GATHER_TREE_MAX_SEQ_LEN = 2;
+    static const size_t GATHER_TREE_END_TOKEN = 3;
+
+    InferenceEngine::Precision precision;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_grn_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_grn_node.cpp
new file mode 100644
index 00000000000000..0dbe8dee59ea51
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_grn_node.cpp
@@ -0,0 +1,81 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "base.hpp"
+
+#include <string>
+
+#include <ngraph/opsets/opset1.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_grn_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNGRNNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto grn = std::dynamic_pointer_cast<const ngraph::opset1::GRN>(op);
+        if (!grn) {
+            errorMessage = "Only opset1 GRN operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNGRNNode::MKLDNNGRNNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "GRN layer with name '" + op->get_friendly_name() + "'";
+    const auto grn = std::dynamic_pointer_cast<const ngraph::opset1::GRN>(op);
+
+    if (getOriginalInputsNumber() != 1 || getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
+
+    bias = grn->get_bias();
+}
+
+void MKLDNNGRNNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32, false, 0}},
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32, false, 0}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNGRNNode::execute(mkldnn::stream strm) {
+    const float* src_data = reinterpret_cast<const float *>(getParentEdgeAt(0)->getMemoryPtr()->GetPtr());
+    float* dst_data = reinterpret_cast<float *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+
+    SizeVector dims = getParentEdgeAt(0)->getDims().ToSizeVector();
+
+    int N = static_cast<int>((dims.size() > 0) ? dims[0] : 1);
+    int C = static_cast<int>((dims.size() > 1) ? dims[1] : 1);
+    int H = static_cast<int>((dims.size() > 2) ? dims[2] : 1);
+    int W = static_cast<int>((dims.size() > 3) ? dims[3] : 1);
+
+    parallel_for3d(N, H, W, [&](int b, int h, int w) {
+        double variance = 0;
+        for (int c = 0; c < C; c++) {
+            variance += std::pow(src_data[b*C*H*W + c*H*W + h*W + w], 2);
+        }
+        variance = std::pow(variance + bias, 0.5f);
+        for (int c = 0; c < C; c++) {
+            dst_data[b*C*H*W + c*H*W + h*W + w] = src_data[b*C*H*W + c*H*W + h*W + w] / static_cast<float>(variance);
+        }
+    });
+}
+
+bool MKLDNNGRNNode::created() const {
+    return getType() == GRN;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNGRNNode, GRN)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_grn_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_grn_node.h
new file mode 100644
index 00000000000000..8fe8d9d75b04e7
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_grn_node.h
@@ -0,0 +1,30 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNGRNNode : public MKLDNNNode {
+public:
+    MKLDNNGRNNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    float bias = 1.0f;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_log_softmax_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_log_softmax_node.cpp
new file mode 100644
index 00000000000000..5750f8517b0096
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_log_softmax_node.cpp
@@ -0,0 +1,116 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <cmath>
+
+#include <ngraph/opsets/opset5.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_log_softmax_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNLogSoftmaxNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto logSoftMax = std::dynamic_pointer_cast<const ngraph::opset5::LogSoftmax>(op);
+        if (!logSoftMax) {
+            errorMessage = "Only opset5 LogSoftmax operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNLogSoftmaxNode::MKLDNNLogSoftmaxNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+                                     MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "LogSoftmax layer with name '" + op->get_friendly_name() + "'";
+    const auto logSoftMax = std::dynamic_pointer_cast<const ngraph::opset5::LogSoftmax>(op);
+
+    if (getOriginalInputsNumber() != 1 || getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
+
+    SizeVector dims = op->get_input_shape(0);
+    if (!dims.size())
+        dims = SizeVector(1, 1);
+    int axis = logSoftMax->get_axis();
+    if (axis < 0)
+        axis += dims.size();
+
+    if (dims.size() < static_cast<size_t>((size_t)(1) + axis))
+        IE_THROW() << errorPrefix << " has incorrect input parameters dimensions and axis number!";
+
+    int j;
+    for (j = dims.size() - 1; j >= 0; j--) {
+        if (dims[j] != 1) break;
+    }
+    if (j == axis) isLastDim = true;
+
+    for (int i = 0; i < axis; i++)
+        axisStep *= dims[i];
+    reducedAxisSize = dims[axis];
+    for (size_t i = (axis + 1); i < dims.size(); i++)
+        reducedAxisStride *= dims[i];
+}
+
+void MKLDNNLogSoftmaxNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNLogSoftmaxNode::execute(mkldnn::stream strm) {
+    const float *srcData = reinterpret_cast<const float *>(getParentEdgeAt(0)->getMemoryPtr()->GetPtr());
+    float* dstData = reinterpret_cast<float *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+
+    if (isLastDim) {
+        parallel_for(axisStep, [&](size_t i) {
+            const float *srcDataPtr = &srcData[i * reducedAxisSize];
+            float *dstDataPtr = &dstData[i * reducedAxisSize];
+
+            float reduceProd = 0.0f;
+            const float max = *std::max_element(srcDataPtr, srcDataPtr + reducedAxisSize);
+            for (size_t j = 0; j < reducedAxisSize; ++j)
+                reduceProd += expf(srcDataPtr[j] - max);
+
+            reduceProd = logf(reduceProd);
+            for (size_t j = 0; j < reducedAxisSize; ++j)
+                dstDataPtr[j] = srcDataPtr[j] - max - reduceProd;
+        });
+    } else {
+        parallel_for2d(axisStep, reducedAxisStride, [&](size_t k, size_t i) {
+            const float *srcDataPtr = &srcData[k * reducedAxisStride * reducedAxisSize + i];
+            float *dstDataPtr = &dstData[k * reducedAxisStride * reducedAxisSize + i];
+
+            float reduceProd = 0.0f;
+            float max = std::numeric_limits<float>::min();
+            for (size_t j = 0; j < reducedAxisSize; ++j) {
+                if (srcDataPtr[j * reducedAxisStride] > max)
+                    max = srcDataPtr[j * reducedAxisStride];
+            }
+
+            for (size_t j = 0; j < reducedAxisSize; ++j)
+                reduceProd += expf(srcDataPtr[j * reducedAxisStride] - max);
+
+            reduceProd = logf(reduceProd);
+            for (size_t j = 0; j < reducedAxisSize; ++j)
+                dstDataPtr[j * reducedAxisStride] = srcDataPtr[j * reducedAxisStride] - max - reduceProd;
+        });
+    }
+}
+
+bool MKLDNNLogSoftmaxNode::created() const {
+    return getType() == LogSoftmax;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNLogSoftmaxNode, LogSoftmax)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_log_softmax_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_log_softmax_node.h
new file mode 100644
index 00000000000000..456d7321efcdc4
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_log_softmax_node.h
@@ -0,0 +1,34 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNLogSoftmaxNode : public MKLDNNNode {
+public:
+    MKLDNNLogSoftmaxNode(const std::shared_ptr<ngraph::Node>& op,
+        const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    size_t reducedAxisSize;
+    size_t reducedAxisStride = 1;
+    size_t axisStep = 1;
+    bool isLastDim = false;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_math_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_math_node.cpp
index ecfa4fbbd32468..908686bf6df1eb 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_math_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_math_node.cpp
@@ -43,17 +43,17 @@ MKLDNNMathNode::MKLDNNMathNode(const std::shared_ptr<ngraph::Node>& op, const mk
     }
 
     initializers[op->get_type_info()](op, *this);
-
-    size_t sizeVector = op->get_input_size();
-    inDataConf.reserve(sizeVector);
-    for (int i = 0; i < sizeVector; ++i)
-        inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::FP32);
 }
 
 void MKLDNNMathNode::initSupportedPrimitiveDescriptors() {
     if (!supportedPrimitiveDescriptors.empty())
         return;
 
+    std::vector<DataConfigurator> inDataConf;
+    inDataConf.reserve(getOriginalInputsNumber());
+    for (int i = 0; i < getOriginalInputsNumber(); ++i)
+        inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::FP32);
+
     addSupportedPrimDesc(inDataConf,
                          {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
                          impl_desc_type::ref_any);
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_math_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_math_node.h
index a91cb3ae373d9c..28260dc476ec54 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_math_node.h
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_math_node.h
@@ -28,7 +28,6 @@ class MKLDNNMathNode : public MKLDNNNode {
     float beta = 0.0f;
     float gamma = 0.0f;
 
-    std::vector<DataConfigurator> inDataConf;
     std::string errorPrefix;
 };
 
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_non_max_suppression_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_non_max_suppression_node.cpp
new file mode 100644
index 00000000000000..093127eada5f9a
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_non_max_suppression_node.cpp
@@ -0,0 +1,406 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <cmath>
+#include <string>
+#include <vector>
+#include <algorithm>
+#include <utility>
+#include <queue>
+
+#include "mkldnn_non_max_suppression_node.h"
+#include "ie_parallel.hpp"
+#include <ngraph_ops/nms_ie_internal.hpp>
+#include "utils/general_utils.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNNonMaxSuppressionNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto nms = std::dynamic_pointer_cast<const ngraph::op::internal::NonMaxSuppressionIEInternal>(op);
+        if (!nms) {
+            errorMessage = "Only internal NonMaxSuppression operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNNonMaxSuppressionNode::MKLDNNNonMaxSuppressionNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+        std::string errorMessage;
+        if (!isSupportedOperation(op, errorMessage)) {
+            IE_THROW(NotImplemented) << errorMessage;
+        }
+
+        errorPrefix = "NMS layer with name '" + op->get_friendly_name() + "' ";
+        const auto nms = std::dynamic_pointer_cast<const ngraph::op::internal::NonMaxSuppressionIEInternal>(op);
+
+        if (getOriginalInputsNumber() < 2 || getOriginalInputsNumber() > 6)
+            IE_THROW() << errorPrefix << "has incorrect number of input edges: " << getOriginalInputsNumber();
+
+        if (getOriginalOutputsNumber() < 1 || getOriginalOutputsNumber() > 3)
+            IE_THROW() << errorPrefix << "has incorrect number of output edges: " << getOriginalOutputsNumber();
+
+        boxEncodingType = nms->m_center_point_box ? boxEncoding::CENTER : boxEncoding::CORNER;
+
+        sort_result_descending = nms->m_sort_result_descending;
+
+        const SizeVector &boxes_dims = op->get_input_shape(NMS_BOXES);
+        num_batches = boxes_dims[0];
+        num_boxes = boxes_dims[1];
+        if (boxes_dims.size() != 3)
+            IE_THROW() << errorPrefix << "has unsupported 'boxes' input rank: " << boxes_dims.size();
+        if (boxes_dims[2] != 4)
+            IE_THROW() << errorPrefix << "has unsupported 'boxes' input 3rd dimension size: " << boxes_dims[2];
+
+        const SizeVector &scores_dims = op->get_input_shape(NMS_SCORES);
+        num_classes = scores_dims[1];
+        if (scores_dims.size() != 3)
+            IE_THROW() << errorPrefix << "has unsupported 'scores' input rank: " << scores_dims.size();
+
+        if (num_batches != scores_dims[0])
+            IE_THROW() << errorPrefix << " num_batches is different in 'boxes' and 'scores' inputs";
+        if (num_boxes != scores_dims[2])
+            IE_THROW() << errorPrefix << " num_boxes is different in 'boxes' and 'scores' inputs";
+
+        numFiltBox.resize(num_batches);
+        for (auto & i : numFiltBox)
+            i.resize(num_classes);
+
+        inputShape_MAXOUTPUTBOXESPERCLASS = op->get_input_shape(NMS_MAXOUTPUTBOXESPERCLASS);
+        inputShape_IOUTHRESHOLD = op->get_input_shape(NMS_IOUTHRESHOLD);
+        inputShape_SCORETHRESHOLD = op->get_input_shape(NMS_SCORETHRESHOLD);
+        if (getOriginalInputsNumber() > NMS_SOFTNMSSIGMA) {
+            inputShape_SOFTNMSSIGMA = op->get_input_shape(NMS_SOFTNMSSIGMA);
+        }
+
+        outputShape_SELECTEDINDICES = op->get_output_shape(NMS_SELECTEDINDICES);
+        outputShape_SELECTEDSCORES = op->get_output_shape(NMS_SELECTEDSCORES);
+
+        const SizeVector &valid_outputs_dims = op->get_input_shape(NMS_VALIDOUTPUTS);
+        if (valid_outputs_dims.size() != 1)
+            IE_THROW() << errorPrefix << "has unsupported 'valid_outputs' output rank: " << valid_outputs_dims.size();
+        if (valid_outputs_dims[0] != 1)
+            IE_THROW() << errorPrefix << "has unsupported 'valid_outputs' output 1st dimension size: " << valid_outputs_dims[1];
+}
+
+void MKLDNNNonMaxSuppressionNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    const std::vector<Precision> supportedFloatPrecision = {Precision::FP32, Precision::BF16};
+    const std::vector<Precision> supportedIntOutputPrecision = {Precision::I32, Precision::I64};
+
+    checkPrecision(getOriginalInputPrecisionAtPort(NMS_BOXES), supportedFloatPrecision, "boxes", inType);
+    checkPrecision(getOriginalInputPrecisionAtPort(NMS_SCORES), supportedFloatPrecision, "scores", inType);
+    checkPrecision(getOriginalInputPrecisionAtPort(NMS_VALIDOUTPUTS), supportedIntOutputPrecision, "valid_outputs", outType);
+
+    const std::vector<Precision> supportedPrecision = {Precision::I16, Precision::U8, Precision::I8, Precision::U16, Precision::I32,
+                                                       Precision::U32, Precision::I64, Precision::U64};
+
+    check1DInput(inputShape_MAXOUTPUTBOXESPERCLASS, supportedPrecision, "max_output_boxes_per_class", NMS_MAXOUTPUTBOXESPERCLASS);
+    check1DInput(inputShape_IOUTHRESHOLD, supportedFloatPrecision, "iou_threshold", NMS_IOUTHRESHOLD);
+    check1DInput(inputShape_SCORETHRESHOLD, supportedFloatPrecision, "score_threshold", NMS_SCORETHRESHOLD);
+
+    if (getOriginalInputsNumber() > NMS_SOFTNMSSIGMA) {
+        check1DInput(inputShape_SOFTNMSSIGMA, supportedFloatPrecision, "soft_nms_sigma", NMS_SOFTNMSSIGMA);
+    }
+
+    checkOutput(outputShape_SELECTEDINDICES, supportedIntOutputPrecision, "selected_indices", NMS_SELECTEDINDICES);
+    checkOutput(outputShape_SELECTEDSCORES, supportedFloatPrecision, "selected_scores", NMS_SELECTEDSCORES);
+
+    std::vector<DataConfigurator> inDataConf;
+    inDataConf.reserve(getOriginalInputsNumber());
+    for (int i = 0; i < getOriginalInputsNumber(); ++i) {
+        Precision inPrecision = i == NMS_MAXOUTPUTBOXESPERCLASS ? Precision::I32 : Precision::FP32;
+        inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, inPrecision);
+    }
+
+    std::vector<DataConfigurator> outDataConf;
+    outDataConf.reserve(getOriginalOutputsNumber());
+    for (int i = 0; i < getOriginalOutputsNumber(); ++i) {
+        Precision outPrecision = i == NMS_SELECTEDSCORES ? Precision::FP32 : Precision::I32;
+        outDataConf.emplace_back(TensorDescCreatorTypes::ncsp, outPrecision);
+    }
+
+    addSupportedPrimDesc(inDataConf, outDataConf, impl_desc_type::ref_any);
+}
+
+void MKLDNNNonMaxSuppressionNode::execute(mkldnn::stream strm) {
+    const float *boxes = reinterpret_cast<const float *>(getParentEdgeAt(NMS_BOXES)->getMemoryPtr()->GetPtr());
+    const float *scores = reinterpret_cast<const float *>(getParentEdgeAt(NMS_SCORES)->getMemoryPtr()->GetPtr());
+
+    max_output_boxes_per_class = outDims.size() > NMS_SELECTEDSCORES ? 0 : num_boxes;
+    if (inDims.size() > NMS_MAXOUTPUTBOXESPERCLASS) {
+        max_output_boxes_per_class = reinterpret_cast<int *>(getParentEdgeAt(NMS_MAXOUTPUTBOXESPERCLASS)->getMemoryPtr()->GetPtr())[0];
+    }
+
+    if (max_output_boxes_per_class == 0)
+        return;
+
+    iou_threshold = outDims.size() > NMS_SELECTEDSCORES ? 0.0f : 1.0f;
+    if (inDims.size() > NMS_IOUTHRESHOLD)
+        iou_threshold = reinterpret_cast<float *>(getParentEdgeAt(NMS_IOUTHRESHOLD)->getMemoryPtr()->GetPtr())[0];
+
+    score_threshold = 0.0f;
+    if (inDims.size() > NMS_SCORETHRESHOLD)
+        score_threshold = reinterpret_cast<float *>(getParentEdgeAt(NMS_SCORETHRESHOLD)->getMemoryPtr()->GetPtr())[0];
+
+    soft_nms_sigma = 0.0f;
+    if (inDims.size() > NMS_SOFTNMSSIGMA)
+        soft_nms_sigma = reinterpret_cast<float *>(getParentEdgeAt(NMS_SOFTNMSSIGMA)->getMemoryPtr()->GetPtr())[0];
+    scale = 0.0f;
+    if (soft_nms_sigma > 0.0) {
+        scale = -0.5 / soft_nms_sigma;
+    }
+
+    int *selected_indices = reinterpret_cast<int *>(getChildEdgesAtPort(NMS_SELECTEDINDICES)[0]->getMemoryPtr()->GetPtr());
+
+    float *selected_scores = nullptr;
+    if (outDims.size() > NMS_SELECTEDSCORES)
+        selected_scores = reinterpret_cast<float *>(getChildEdgesAtPort(NMS_SELECTEDSCORES)[0]->getMemoryPtr()->GetPtr());
+
+    int *valid_outputs = nullptr;
+    if (outDims.size() > NMS_VALIDOUTPUTS)
+        valid_outputs = reinterpret_cast<int *>(getChildEdgesAtPort(NMS_VALIDOUTPUTS)[0]->getMemoryPtr()->GetPtr());
+
+    auto boxesStrides = getParentEdgeAt(NMS_BOXES)->getDesc().getBlockingDesc().getStrides();
+    auto scoresStrides = getParentEdgeAt(NMS_SCORES)->getDesc().getBlockingDesc().getStrides();
+
+    std::vector<filteredBoxes> filtBoxes(max_output_boxes_per_class * num_batches * num_classes);
+
+    if (soft_nms_sigma == 0.0f) {
+        nmsWithoutSoftSigma(boxes, scores, boxesStrides, scoresStrides, filtBoxes);
+    } else {
+        nmsWithSoftSigma(boxes, scores, boxesStrides, scoresStrides, filtBoxes);
+    }
+
+    size_t startOffset = numFiltBox[0][0];
+    for (size_t b = 0; b < numFiltBox.size(); b++) {
+        size_t batchOffset = b*num_classes*max_output_boxes_per_class;
+        for (size_t c = (b == 0 ? 1 : 0); c < numFiltBox[b].size(); c++) {
+            size_t offset = batchOffset + c*max_output_boxes_per_class;
+            for (size_t i = 0; i < numFiltBox[b][c]; i++) {
+                filtBoxes[startOffset + i] = filtBoxes[offset + i];
+            }
+            startOffset += numFiltBox[b][c];
+        }
+    }
+    filtBoxes.resize(startOffset);
+
+    // need more particular comparator to get deterministic behaviour
+    // escape situation when filtred boxes with same score have different position from launch to launch
+    if (sort_result_descending) {
+        parallel_sort(filtBoxes.begin(), filtBoxes.end(),
+                      [](const filteredBoxes& l, const filteredBoxes& r) {
+                          return (l.score > r.score) ||
+                                 (l.score ==  r.score && l.batch_index < r.batch_index) ||
+                                 (l.score ==  r.score && l.batch_index == r.batch_index && l.class_index < r.class_index) ||
+                                 (l.score ==  r.score && l.batch_index == r.batch_index && l.class_index == r.class_index && l.box_index < r.box_index);
+                      });
+    }
+
+    const size_t selectedBoxesNum = getChildEdgesAtPort(NMS_SELECTEDINDICES)[0]->getDims()[0];
+    const size_t validOutputs = std::min(filtBoxes.size(), selectedBoxesNum);
+
+    int selectedIndicesStride = getChildEdgesAtPort(NMS_SELECTEDINDICES)[0]->getDesc().getBlockingDesc().getStrides()[0];
+    int *selectedIndicesPtr = selected_indices;
+    float *selectedScoresPtr = selected_scores;
+
+    size_t idx = 0lu;
+    for (; idx < validOutputs; idx++) {
+        selectedIndicesPtr[0] = filtBoxes[idx].batch_index;
+        selectedIndicesPtr[1] = filtBoxes[idx].class_index;
+        selectedIndicesPtr[2] = filtBoxes[idx].box_index;
+        selectedIndicesPtr += selectedIndicesStride;
+        if (outDims.size() > NMS_SELECTEDSCORES) {
+            selectedScoresPtr[0] = static_cast<float>(filtBoxes[idx].batch_index);
+            selectedScoresPtr[1] = static_cast<float>(filtBoxes[idx].class_index);
+            selectedScoresPtr[2] = static_cast<float>(filtBoxes[idx].score);
+            selectedScoresPtr += selectedIndicesStride;
+        }
+    }
+    std::fill(selectedIndicesPtr, selectedIndicesPtr + (selectedBoxesNum - idx) * selectedIndicesStride, -1);
+    if (outDims.size() > NMS_SELECTEDSCORES) {
+        std::fill(selectedScoresPtr, selectedScoresPtr + (selectedBoxesNum - idx) * selectedIndicesStride, -1.f);
+    }
+    if (outDims.size() > NMS_VALIDOUTPUTS)
+        *valid_outputs = static_cast<int>(validOutputs);
+}
+
+bool MKLDNNNonMaxSuppressionNode::created() const {
+    return getType() == NonMaxSuppression;
+}
+
+float MKLDNNNonMaxSuppressionNode::intersectionOverUnion(const float *boxesI, const float *boxesJ) {
+    float yminI, xminI, ymaxI, xmaxI, yminJ, xminJ, ymaxJ, xmaxJ;
+    if (boxEncodingType == boxEncoding::CENTER) {
+        //  box format: x_center, y_center, width, height
+        yminI = boxesI[1] - boxesI[3] / 2.f;
+        xminI = boxesI[0] - boxesI[2] / 2.f;
+        ymaxI = boxesI[1] + boxesI[3] / 2.f;
+        xmaxI = boxesI[0] + boxesI[2] / 2.f;
+        yminJ = boxesJ[1] - boxesJ[3] / 2.f;
+        xminJ = boxesJ[0] - boxesJ[2] / 2.f;
+        ymaxJ = boxesJ[1] + boxesJ[3] / 2.f;
+        xmaxJ = boxesJ[0] + boxesJ[2] / 2.f;
+    } else {
+        //  box format: y1, x1, y2, x2
+        yminI = (std::min)(boxesI[0], boxesI[2]);
+        xminI = (std::min)(boxesI[1], boxesI[3]);
+        ymaxI = (std::max)(boxesI[0], boxesI[2]);
+        xmaxI = (std::max)(boxesI[1], boxesI[3]);
+        yminJ = (std::min)(boxesJ[0], boxesJ[2]);
+        xminJ = (std::min)(boxesJ[1], boxesJ[3]);
+        ymaxJ = (std::max)(boxesJ[0], boxesJ[2]);
+        xmaxJ = (std::max)(boxesJ[1], boxesJ[3]);
+    }
+
+    float areaI = (ymaxI - yminI) * (xmaxI - xminI);
+    float areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);
+    if (areaI <= 0.f || areaJ <= 0.f)
+        return 0.f;
+
+    float intersection_area =
+            (std::max)((std::min)(ymaxI, ymaxJ) - (std::max)(yminI, yminJ), 0.f) *
+            (std::max)((std::min)(xmaxI, xmaxJ) - (std::max)(xminI, xminJ), 0.f);
+    return intersection_area / (areaI + areaJ - intersection_area);
+}
+
+void MKLDNNNonMaxSuppressionNode::nmsWithSoftSigma(const float *boxes, const float *scores, const SizeVector &boxesStrides,
+                                                             const SizeVector &scoresStrides, std::vector<filteredBoxes> &filtBoxes) {
+    auto less = [](const boxInfo& l, const boxInfo& r) {
+        return l.score < r.score || ((l.score == r.score) && (l.idx > r.idx));
+    };
+
+    auto coeff = [&](float iou) {
+        const float weight = std::exp(scale * iou * iou);
+        return iou <= iou_threshold ? weight : 0.0f;
+    };
+
+    parallel_for2d(num_batches, num_classes, [&](int batch_idx, int class_idx) {
+        std::vector<filteredBoxes> fb;
+        const float *boxesPtr = boxes + batch_idx * boxesStrides[0];
+        const float *scoresPtr = scores + batch_idx * scoresStrides[0] + class_idx * scoresStrides[1];
+
+        std::priority_queue<boxInfo, std::vector<boxInfo>, decltype(less)> sorted_boxes(less);
+        for (int box_idx = 0; box_idx < num_boxes; box_idx++) {
+            if (scoresPtr[box_idx] > score_threshold)
+                sorted_boxes.emplace(boxInfo({scoresPtr[box_idx], box_idx, 0}));
+        }
+
+        fb.reserve(sorted_boxes.size());
+        if (sorted_boxes.size() > 0) {
+            while (fb.size() < max_output_boxes_per_class && !sorted_boxes.empty()) {
+                boxInfo currBox = sorted_boxes.top();
+                float origScore = currBox.score;
+                sorted_boxes.pop();
+
+                bool box_is_selected = true;
+                for (int idx = static_cast<int>(fb.size()) - 1; idx >= currBox.suppress_begin_index; idx--) {
+                    float iou = intersectionOverUnion(&boxesPtr[currBox.idx * 4], &boxesPtr[fb[idx].box_index * 4]);
+                    currBox.score *= coeff(iou);
+                    if (iou >= iou_threshold) {
+                        box_is_selected = false;
+                        break;
+                    }
+                    if (currBox.score <= score_threshold)
+                        break;
+                }
+
+                currBox.suppress_begin_index = fb.size();
+                if (box_is_selected) {
+                    if (currBox.score == origScore) {
+                        fb.push_back({ currBox.score, batch_idx, class_idx, currBox.idx });
+                        continue;
+                    }
+                    if (currBox.score > score_threshold) {
+                        sorted_boxes.push(currBox);
+                    }
+                }
+            }
+        }
+        numFiltBox[batch_idx][class_idx] = fb.size();
+        size_t offset = batch_idx*num_classes*max_output_boxes_per_class + class_idx*max_output_boxes_per_class;
+        for (size_t i = 0; i < fb.size(); i++) {
+            filtBoxes[offset + i] = fb[i];
+        }
+    });
+}
+
+void MKLDNNNonMaxSuppressionNode::nmsWithoutSoftSigma(const float *boxes, const float *scores, const SizeVector &boxesStrides,
+                                                                const SizeVector &scoresStrides, std::vector<filteredBoxes> &filtBoxes) {
+    int max_out_box = static_cast<int>(max_output_boxes_per_class);
+    parallel_for2d(num_batches, num_classes, [&](int batch_idx, int class_idx) {
+        const float *boxesPtr = boxes + batch_idx * boxesStrides[0];
+        const float *scoresPtr = scores + batch_idx * scoresStrides[0] + class_idx * scoresStrides[1];
+
+        std::vector<std::pair<float, int>> sorted_boxes;
+        for (int box_idx = 0; box_idx < num_boxes; box_idx++) {
+            if (scoresPtr[box_idx] > score_threshold)
+                sorted_boxes.emplace_back(std::make_pair(scoresPtr[box_idx], box_idx));
+        }
+
+        int io_selection_size = 0;
+        if (sorted_boxes.size() > 0) {
+            parallel_sort(sorted_boxes.begin(), sorted_boxes.end(),
+                          [](const std::pair<float, int>& l, const std::pair<float, int>& r) {
+                              return (l.first > r.first || ((l.first == r.first) && (l.second < r.second)));
+                          });
+            int offset = batch_idx*num_classes*max_output_boxes_per_class + class_idx*max_output_boxes_per_class;
+            filtBoxes[offset + 0] = filteredBoxes(sorted_boxes[0].first, batch_idx, class_idx, sorted_boxes[0].second);
+            io_selection_size++;
+            for (size_t box_idx = 1; (box_idx < sorted_boxes.size()) && (io_selection_size < max_out_box); box_idx++) {
+                bool box_is_selected = true;
+                for (int idx = io_selection_size - 1; idx >= 0; idx--) {
+                    float iou = intersectionOverUnion(&boxesPtr[sorted_boxes[box_idx].second * 4], &boxesPtr[filtBoxes[offset + idx].box_index * 4]);
+                    if (iou >= iou_threshold) {
+                        box_is_selected = false;
+                        break;
+                    }
+                }
+
+                if (box_is_selected) {
+                    filtBoxes[offset + io_selection_size] = filteredBoxes(sorted_boxes[box_idx].first, batch_idx, class_idx, sorted_boxes[box_idx].second);
+                    io_selection_size++;
+                }
+            }
+        }
+        numFiltBox[batch_idx][class_idx] = io_selection_size;
+    });
+}
+
+void MKLDNNNonMaxSuppressionNode::checkPrecision(const Precision prec, const std::vector<Precision> precList,
+                                                           const std::string name, const std::string type) {
+    if (std::find(precList.begin(), precList.end(), prec) == precList.end())
+        IE_THROW() << errorPrefix << "has unsupported '" << name << "' " << type << " precision: " << prec;
+}
+
+void MKLDNNNonMaxSuppressionNode::check1DInput(const SizeVector& dims, const std::vector<Precision> precList,
+                                                         const std::string name, const size_t port) {
+    checkPrecision(getOriginalInputPrecisionAtPort(port), precList, name, inType);
+
+    if (dims.size() != 0 && dims.size() != 1)
+        IE_THROW() << errorPrefix << "has unsupported '" << name << "' input rank: " << dims.size();
+    if (dims.size() == 1)
+        if (dims[0] != 1)
+            IE_THROW() << errorPrefix << "has unsupported '" << name << "' input 1st dimension size: " << dims[0];
+}
+
+void MKLDNNNonMaxSuppressionNode::checkOutput(const SizeVector& dims, const std::vector<Precision> precList,
+                                                        const std::string name, const size_t port) {
+    checkPrecision(getOriginalOutputPrecisionAtPort(port), precList, name, outType);
+
+    if (dims.size() != 2)
+        IE_THROW() << errorPrefix << "has unsupported '" << name << "' output rank: " << dims.size();
+    if (dims[1] != 3)
+        IE_THROW() << errorPrefix << "has unsupported '" << name << "' output 2nd dimension size: " << dims[1];
+}
+
+
+REG_MKLDNN_PRIM_FOR(MKLDNNNonMaxSuppressionNode, NonMaxSuppression)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_non_max_suppression_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_non_max_suppression_node.h
new file mode 100644
index 00000000000000..4651da1f2e795c
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_non_max_suppression_node.h
@@ -0,0 +1,102 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+#include <string>
+#include <memory>
+#include <vector>
+
+using namespace InferenceEngine;
+
+namespace MKLDNNPlugin {
+
+class MKLDNNNonMaxSuppressionNode : public MKLDNNNode {
+public:
+    MKLDNNNonMaxSuppressionNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+    struct filteredBoxes {
+        float score;
+        int batch_index;
+        int class_index;
+        int box_index;
+        filteredBoxes() = default;
+        filteredBoxes(float _score, int _batch_index, int _class_index, int _box_index) :
+                score(_score), batch_index(_batch_index), class_index(_class_index), box_index(_box_index) {}
+    };
+
+    struct boxInfo {
+        float score;
+        int idx;
+        int suppress_begin_index;
+    };
+
+    float intersectionOverUnion(const float *boxesI, const float *boxesJ);
+
+    void nmsWithSoftSigma(const float *boxes, const float *scores, const SizeVector &boxesStrides,
+                          const SizeVector &scoresStrides, std::vector<filteredBoxes> &filtBoxes);
+
+    void nmsWithoutSoftSigma(const float *boxes, const float *scores, const SizeVector &boxesStrides,
+                             const SizeVector &scoresStrides, std::vector<filteredBoxes> &filtBoxes);
+
+private:
+    // input
+    const size_t NMS_BOXES = 0;
+    const size_t NMS_SCORES = 1;
+    const size_t NMS_MAXOUTPUTBOXESPERCLASS = 2;
+    const size_t NMS_IOUTHRESHOLD = 3;
+    const size_t NMS_SCORETHRESHOLD = 4;
+    const size_t NMS_SOFTNMSSIGMA = 5;
+
+    // output
+    const size_t NMS_SELECTEDINDICES = 0;
+    const size_t NMS_SELECTEDSCORES = 1;
+    const size_t NMS_VALIDOUTPUTS = 2;
+
+    enum class boxEncoding {
+        CORNER,
+        CENTER
+    };
+    boxEncoding boxEncodingType = boxEncoding::CORNER;
+    bool sort_result_descending = true;
+
+    size_t num_batches;
+    size_t num_boxes;
+    size_t num_classes;
+
+    size_t max_output_boxes_per_class = 0lu;
+    float iou_threshold = 0.0f;
+    float score_threshold = 0.0f;
+    float soft_nms_sigma = 0.0f;
+    float scale = 1.f;
+
+    SizeVector inputShape_MAXOUTPUTBOXESPERCLASS;
+    SizeVector inputShape_IOUTHRESHOLD;
+    SizeVector inputShape_SCORETHRESHOLD;
+    SizeVector inputShape_SOFTNMSSIGMA;
+
+    SizeVector outputShape_SELECTEDINDICES;
+    SizeVector outputShape_SELECTEDSCORES;
+
+    std::string errorPrefix;
+
+    std::vector<std::vector<size_t>> numFiltBox;
+    const std::string inType = "input", outType = "output";
+
+    void checkPrecision(const Precision prec, const std::vector<Precision> precList, const std::string name, const std::string type);
+    void check1DInput(const SizeVector& dims, const std::vector<Precision> precList, const std::string name, const size_t port);
+    void checkOutput(const SizeVector& dims, const std::vector<Precision> precList, const std::string name, const size_t port);
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_proposal_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_proposal_node.cpp
new file mode 100644
index 00000000000000..584960373aeb2e
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_proposal_node.cpp
@@ -0,0 +1,198 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+
+#include <ngraph/op/proposal.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_proposal_node.h"
+
+static std::vector<float> generate_anchors(proposal_conf &conf) {
+    auto base_size = conf.base_size_;
+    auto coordinates_offset = conf.coordinates_offset;
+    auto round_ratios = conf.round_ratios;
+
+    auto num_ratios = conf.ratios.size();
+    auto ratios = conf.ratios.data();
+
+    auto num_scales = conf.scales.size();
+    auto scales = conf.scales.data();
+
+    std::vector<float> anchors(num_scales * num_ratios * 4);
+    auto anchors_ptr = anchors.data();
+
+    // base box's width & height & center location
+    const float base_area = static_cast<float>(base_size * base_size);
+    const float half_base_size = base_size * 0.5f;
+    const float center = 0.5f * (base_size - coordinates_offset);
+
+    // enumerate all transformed boxes
+    for (int ratio = 0; ratio < num_ratios; ++ratio) {
+        // transformed width & height for given ratio factors
+        float ratio_w;
+        float ratio_h;
+        if (round_ratios) {
+            ratio_w = std::roundf(std::sqrt(base_area / ratios[ratio]));
+            ratio_h = std::roundf(ratio_w * ratios[ratio]);
+        } else {
+            ratio_w = std::sqrt(base_area / ratios[ratio]);
+            ratio_h = ratio_w * ratios[ratio];
+        }
+
+        float * const p_anchors_wm = anchors_ptr + 0 * num_ratios * num_scales + ratio * num_scales;
+        float * const p_anchors_hm = anchors_ptr + 1 * num_ratios * num_scales + ratio * num_scales;
+        float * const p_anchors_wp = anchors_ptr + 2 * num_ratios * num_scales + ratio * num_scales;
+        float * const p_anchors_hp = anchors_ptr + 3 * num_ratios * num_scales + ratio * num_scales;
+
+        for (int scale = 0; scale < num_scales; ++scale) {
+            // transformed width & height for given scale factors
+            const float scale_w = 0.5f * (ratio_w * scales[scale] - coordinates_offset);
+            const float scale_h = 0.5f * (ratio_h * scales[scale] - coordinates_offset);
+
+            // (x1, y1, x2, y2) for transformed box
+            p_anchors_wm[scale] = center - scale_w;
+            p_anchors_hm[scale] = center - scale_h;
+            p_anchors_wp[scale] = center + scale_w;
+            p_anchors_hp[scale] = center + scale_h;
+
+            if (conf.shift_anchors) {
+                p_anchors_wm[scale] -= half_base_size;
+                p_anchors_hm[scale] -= half_base_size;
+                p_anchors_wp[scale] -= half_base_size;
+                p_anchors_hp[scale] -= half_base_size;
+            }
+        }
+    }
+    return anchors;
+}
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNProposalNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto proposal0Op = ngraph::as_type_ptr<const ngraph::op::v0::Proposal>(op);
+        const auto proposal4Op = ngraph::as_type_ptr<const ngraph::op::v4::Proposal>(op);
+        if (!proposal0Op && !proposal4Op) {
+            errorMessage = "Node is not an instance of the Proposal from the operations set v0 or v4.";
+            return false;
+        }
+        auto proposalOp = std::dynamic_pointer_cast<const ngraph::op::v0::Proposal>(op);
+        if (proposalOp->get_attrs().framework != "tensorflow" && !proposalOp->get_attrs().framework.empty()) {
+            errorMessage = "Unsupported framework attribute: " + proposalOp->get_attrs().framework;
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNProposalNode::MKLDNNProposalNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    auto proposalOp = std::dynamic_pointer_cast<const ngraph::op::v0::Proposal>(op);
+    auto proposalAttrs = proposalOp->get_attrs();
+
+    conf.feat_stride_ = proposalAttrs.feat_stride;
+    conf.base_size_ = proposalAttrs.base_size;
+    conf.min_size_ = proposalAttrs.min_size;
+    conf.pre_nms_topn_ = proposalAttrs.pre_nms_topn;
+    conf.post_nms_topn_ = proposalAttrs.post_nms_topn;
+    conf.nms_thresh_ = proposalAttrs.nms_thresh;
+    conf.box_coordinate_scale_ = proposalAttrs.box_coordinate_scale;
+    conf.box_size_scale_ = proposalAttrs.box_size_scale;
+    conf.scales = proposalAttrs.scale;
+    conf.ratios = proposalAttrs.ratio;
+    conf.normalize_ = proposalAttrs.normalize;
+    conf.clip_before_nms = proposalAttrs.clip_before_nms;
+    conf.clip_after_nms = proposalAttrs.clip_after_nms;
+    conf.anchors_shape_0 = conf.ratios.size() * conf.scales.size();
+
+    if (proposalAttrs.framework == "tensorflow") {
+        conf.coordinates_offset = 0.0f;
+        conf.initial_clip = true;
+        conf.shift_anchors = true;
+        conf.round_ratios = false;
+        conf.swap_xy = true;
+    } else {
+        conf.coordinates_offset = 1.0f;
+        conf.initial_clip = false;
+        conf.shift_anchors = false;
+        conf.round_ratios = true;
+        conf.swap_xy = false;
+    }
+
+    anchors = generate_anchors(conf);
+    roi_indices.resize(conf.post_nms_topn_);
+
+    store_prob = op->get_output_size() == 2;
+}
+
+void MKLDNNProposalNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    if (store_prob) {
+        addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                              {TensorDescCreatorTypes::ncsp, Precision::FP32},
+                              {TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                             {{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                              {TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                             impl_desc_type::ref_any);
+    } else {
+        addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                              {TensorDescCreatorTypes::ncsp, Precision::FP32},
+                              {TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                             {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                             impl_desc_type::ref_any);
+    }
+}
+
+void MKLDNNProposalNode::execute(mkldnn::stream strm) {
+    try {
+        const float* probabilitiesData = reinterpret_cast<const float *>(getParentEdgeAt(PROBABILITIES_IN_IDX)->getMemoryPtr()->GetPtr());
+        const float* anchorsData = reinterpret_cast<const float *>(getParentEdgeAt(ANCHORS_IN_IDX)->getMemoryPtr()->GetPtr());
+        const float* imgInfoData = reinterpret_cast<const float *>(getParentEdgeAt(IMG_INFO_IN_IDX)->getMemoryPtr()->GetPtr());
+        float* outRoiData = reinterpret_cast <float *>(getChildEdgesAtPort(ROI_OUT_IDX)[0]->getMemoryPtr()->GetPtr());
+        float* outProbData = nullptr;
+        if (store_prob)
+            outProbData = reinterpret_cast <float *>(getChildEdgesAtPort(PROBABILITIES_OUT_IDX)[0]->getMemoryPtr()->GetPtr());
+
+        auto inProbDims = getParentEdgeAt(0)->getDims().ToSizeVector();
+        const size_t imgInfoSize = getParentEdgeAt(2)->getDims()[0];
+
+        // input image height & width
+        const float imgHeight = imgInfoData[0];
+        const float imgWidth = imgInfoData[1];
+        if (!std::isnormal(imgHeight) || !std::isnormal(imgWidth) || (imgHeight < 0.f) || (imgWidth < 0.f)) {
+            IE_THROW() << "Proposal operation image info input must have positive image height and width.";
+        }
+
+        // scale factor for height & width
+        const float scaleHeight = imgInfoData[2];
+        const float scaleWidth = imgInfoSize == 4 ? imgInfoData[3] : scaleHeight;
+        if (!std::isfinite(scaleHeight) || !std::isfinite(scaleWidth) || (scaleHeight < 0.f) || (scaleWidth < 0.f)) {
+            IE_THROW() << "Proposal operation image info input must have non negative scales.";
+        }
+
+        InferenceEngine::Extensions::Cpu::XARCH::proposal_exec(probabilitiesData, anchorsData, inProbDims,
+                {imgHeight, imgWidth, scaleHeight, scaleWidth}, anchors.data(), roi_indices.data(), outRoiData, outProbData, conf);
+    } catch (const InferenceEngine::Exception& e) {
+        std::string errorMsg = e.what();
+        IE_THROW() << errorMsg;
+    }
+}
+
+bool MKLDNNProposalNode::created() const {
+    return getType() == Proposal;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNProposalNode, Proposal)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_proposal_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_proposal_node.h
new file mode 100644
index 00000000000000..4fdb333b25921b
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_proposal_node.h
@@ -0,0 +1,42 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+#include "proposal_imp.hpp"
+
+using proposal_conf = InferenceEngine::Extensions::Cpu::proposal_conf;
+
+namespace MKLDNNPlugin {
+
+class MKLDNNProposalNode : public MKLDNNNode {
+public:
+    MKLDNNProposalNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    const size_t PROBABILITIES_IN_IDX = 0lu;
+    const size_t ANCHORS_IN_IDX = 1lu;
+    const size_t IMG_INFO_IN_IDX = 2lu;
+    const size_t ROI_OUT_IDX = 0lu;
+    const size_t PROBABILITIES_OUT_IDX = 1lu;
+
+    proposal_conf conf;
+    std::vector<float> anchors;
+    std::vector<int> roi_indices;
+    bool store_prob;  // store blob with proposal probabilities
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_range_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_range_node.cpp
new file mode 100644
index 00000000000000..33e625fce6f88a
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_range_node.cpp
@@ -0,0 +1,140 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "base.hpp"
+
+#include <string>
+
+#include <ngraph/opsets/opset1.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_range_node.h"
+#include <utils/general_utils.h>
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNRangeNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        if (!MKLDNNPlugin::one_of(op->get_type_info(), ngraph::op::v0::Range::type_info, ngraph::op::v4::Range::type_info)) {
+            errorMessage = "Only opset1 and opset4 Range operation is supported";
+            return false;
+        }
+        if (std::dynamic_pointer_cast<const ngraph::opset1::Constant>(op->get_input_node_shared_ptr(RANGE_START)) == nullptr ||
+            std::dynamic_pointer_cast<const ngraph::opset1::Constant>(op->get_input_node_shared_ptr(RANGE_LIMIT)) == nullptr ||
+            std::dynamic_pointer_cast<const ngraph::opset1::Constant>(op->get_input_node_shared_ptr(RANGE_DELTA)) == nullptr) {
+            errorMessage = "Only const inputs for Range operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNRangeNode::MKLDNNRangeNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "Range layer with name '" + op->get_friendly_name() + "'";
+
+    if (getOriginalInputsNumber() != 3 || getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
+
+    SizeVector start_dims = op->get_input_shape(RANGE_START);
+    if (ngraph::shape_size(start_dims) != 1)
+        IE_THROW() << errorPrefix << " has start scalar with more than 1 value";
+
+    SizeVector limit_dims = op->get_input_shape(RANGE_LIMIT);
+    if (ngraph::shape_size(limit_dims) != 1)
+        IE_THROW() << errorPrefix << " has limit scalar with more than 1 value";
+
+    SizeVector delta_dims = op->get_input_shape(RANGE_DELTA);
+    if (ngraph::shape_size(delta_dims) != 1)
+        IE_THROW() << errorPrefix << " has delta scalar with more than 1 value";
+
+    SizeVector dst_dims = op->get_output_shape(0);
+    if (dst_dims.size() > 1)
+        IE_THROW() << errorPrefix << " has unsupported rank for output: " << dst_dims.size();
+}
+
+void MKLDNNRangeNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    std::vector<DataConfigurator> inDataConf;
+    std::vector<DataConfigurator> outDataConf;
+
+    if (!(getOriginalInputPrecisionAtPort(RANGE_START) == Precision::I32 &&
+            getOriginalInputPrecisionAtPort(RANGE_LIMIT) == Precision::I32 &&
+            getOriginalInputPrecisionAtPort(RANGE_DELTA) == Precision::I32 &&
+            getOriginalOutputPrecisionAtPort(0)     == Precision::I32) &&
+        !(getOriginalInputPrecisionAtPort(RANGE_START) == Precision::FP32 &&
+            getOriginalInputPrecisionAtPort(RANGE_LIMIT) == Precision::FP32 &&
+            getOriginalInputPrecisionAtPort(RANGE_DELTA) == Precision::FP32 &&
+            getOriginalOutputPrecisionAtPort(0) == Precision::FP32)) {
+        inDataConf.reserve(getOriginalInputsNumber());
+        for (int i = 0; i < getOriginalInputsNumber(); ++i)
+            inDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::FP32);
+        outDataConf.reserve(1);
+        outDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::FP32);
+        addSupportedPrimDesc(inDataConf, outDataConf, impl_desc_type::ref_any);
+    } else {
+        inDataConf.reserve(getOriginalInputsNumber());
+        for (int i = 0; i < getOriginalInputsNumber(); ++i)
+            inDataConf.emplace_back(TensorDescCreatorTypes::ncsp);
+        outDataConf.reserve(1);
+        outDataConf.emplace_back(TensorDescCreatorTypes::ncsp);
+        addSupportedPrimDesc(inDataConf, outDataConf, impl_desc_type::ref_any);
+    }
+}
+
+void MKLDNNRangeNode::execute(mkldnn::stream strm) {
+    StatusCode retcode = OK;
+    switch (getParentEdgeAt(0)->getDesc().getPrecision()) {
+        case Precision::FP32:
+            retcode = rangeKernel<float>();
+            break;
+        case Precision::I32:
+            retcode = rangeKernel<int32_t>();
+            break;
+        default:
+            IE_THROW() << "Incorrect output precision. Only FP32 and I32 are supported!";
+    }
+    if (retcode == PARAMETER_MISMATCH) {
+        std::string errorMsg = "Range indexes exceeds data tensor dimension";
+        IE_THROW() << errorMsg;
+    }
+}
+
+template <typename data_t>
+InferenceEngine::StatusCode MKLDNNRangeNode::rangeKernel() noexcept {
+    size_t dst_size = (getChildEdgesAtPort(0)[0]->getDims())[0];
+    data_t* dst_data = reinterpret_cast<data_t *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+    data_t start = reinterpret_cast<const data_t *>(getParentEdgeAt(RANGE_START)->getMemoryPtr()->GetPtr())[0];
+    data_t limit = reinterpret_cast<const data_t *>(getParentEdgeAt(RANGE_LIMIT)->getMemoryPtr()->GetPtr())[0];
+    data_t delta = reinterpret_cast<const data_t *>(getParentEdgeAt(RANGE_DELTA)->getMemoryPtr()->GetPtr())[0];
+    size_t work_amount_dst = static_cast<size_t>(std::floor(std::abs((limit - start) / delta)));
+    if (work_amount_dst != dst_size)
+        return PARAMETER_MISMATCH;
+
+    parallel_nt(0, [&](const int ithr, const int nthr) {
+        size_t iwork = 0, end = 0;
+        splitter(work_amount_dst, nthr, ithr, iwork, end);
+        data_t dst_value = start + iwork * delta;
+
+        for (; iwork < end; ++iwork, dst_value += delta) {
+            dst_data[iwork] = dst_value;
+        }
+    });
+    return OK;
+}
+
+bool MKLDNNRangeNode::created() const {
+    return getType() == Range;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNRangeNode, Range)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_range_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_range_node.h
new file mode 100644
index 00000000000000..b5584be6aa949c
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_range_node.h
@@ -0,0 +1,34 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNRangeNode : public MKLDNNNode {
+public:
+    MKLDNNRangeNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+    template <typename data_t>
+    InferenceEngine::StatusCode rangeKernel() noexcept;
+private:
+    static const size_t RANGE_START = 0;
+    static const size_t RANGE_LIMIT = 1;
+    static const size_t RANGE_DELTA = 2;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reorg_yolo_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reorg_yolo_node.cpp
new file mode 100644
index 00000000000000..3db7470e92fba9
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reorg_yolo_node.cpp
@@ -0,0 +1,93 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "base.hpp"
+
+#include <string>
+
+#include <ngraph/opsets/opset2.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_reorg_yolo_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNReorgYoloNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto reorgYolo = std::dynamic_pointer_cast<const ngraph::opset2::ReorgYolo>(op);
+        if (!reorgYolo) {
+            errorMessage = "Only opset2 ReorgYolo operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNReorgYoloNode::MKLDNNReorgYoloNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+        MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = std::string(op->get_type_name()) + " node with name '" + op->get_friendly_name() + "'";
+    if (getOriginalInputsNumber() != 1 || getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
+
+    const auto reorgYolo = std::dynamic_pointer_cast<const ngraph::opset2::ReorgYolo>(op);
+    const auto strides = reorgYolo->get_strides();
+    if (strides.empty())
+        IE_THROW() << errorPrefix << " has empty strides";
+    stride = strides[0];
+}
+
+void MKLDNNReorgYoloNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNReorgYoloNode::execute(mkldnn::stream strm) {
+    const auto *src_data = reinterpret_cast<const float *>(getParentEdgeAt(0)->getMemoryPtr()->GetPtr());
+    auto *dst_data = reinterpret_cast<float *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+
+    int IW = (getParentEdgeAt(0)->getDesc().getDims().size() > 3) ? getParentEdgeAt(0)->getDims()[3] : 1;
+    int IH = (getParentEdgeAt(0)->getDesc().getDims().size() > 2) ? getParentEdgeAt(0)->getDims()[2] : 1;
+    int IC = (getParentEdgeAt(0)->getDesc().getDims().size() > 1) ? getParentEdgeAt(0)->getDims()[1] : 1;
+    int B  = (getParentEdgeAt(0)->getDesc().getDims().size() > 0) ? getParentEdgeAt(0)->getDims()[0] : 1;
+
+    int ic_off = IC / (stride * stride);
+    int ih_off = IH * stride;
+    int iw_off = IW * stride;
+    for (int b = 0; b < B; b++) {
+        for (int ic = 0; ic < IC; ic++) {
+            for (int ih = 0; ih < IH; ih++) {
+                for (int iw = 0; iw < IW; iw++) {
+                    int dstIndex = b * IC * IH * IW + ic * IH * IW + ih * IW + iw;
+
+                    int oc = ic % ic_off;
+                    int offset = ic / ic_off;
+
+                    int ow = iw * stride + offset % stride;
+                    int oh = ih * stride + offset / stride;
+
+                    int srcIndex = b * ic_off * ih_off * iw_off + oc * ih_off * iw_off + oh * iw_off + ow;
+
+                    dst_data[dstIndex] = src_data[srcIndex];
+                }
+            }
+        }
+    }
+}
+
+bool MKLDNNReorgYoloNode::created() const {
+    return getType() == ReorgYolo;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNReorgYoloNode, ReorgYolo)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reorg_yolo_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reorg_yolo_node.h
new file mode 100644
index 00000000000000..b88f19010e0491
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reorg_yolo_node.h
@@ -0,0 +1,30 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNReorgYoloNode : public MKLDNNNode {
+public:
+    MKLDNNReorgYoloNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    int stride;
+
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reverse_sequence_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reverse_sequence_node.cpp
new file mode 100644
index 00000000000000..5f6e6083e90c4a
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reverse_sequence_node.cpp
@@ -0,0 +1,182 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "base.hpp"
+
+#include <string>
+#include <vector>
+
+#include <ngraph/opsets/opset1.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_reverse_sequence_node.h"
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNReverseSequenceNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto revSeq = std::dynamic_pointer_cast<const ngraph::opset1::ReverseSequence>(op);
+        if (!revSeq) {
+            errorMessage = "Only opset1 ReverseSequence operation is supported";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNReverseSequenceNode::MKLDNNReverseSequenceNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+                                         MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+
+    errorPrefix = "ReverseSequence layer with name '" + op->get_friendly_name() + "'";
+    const auto revSeq = std::dynamic_pointer_cast<const ngraph::opset1::ReverseSequence>(op);
+
+    if (getOriginalInputsNumber() != 2 || getOriginalOutputsNumber() != 1)
+        IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
+
+    src_dims = op->get_input_shape(REVERSESEQUENCE_DATA);
+
+    SizeVector seq_lengths_dims = op->get_input_shape(REVERSESEQUENCE_LENGTHS);
+    if (seq_lengths_dims.size() != 1)
+        IE_THROW() << errorPrefix << " has incorrect 2nd input rank: " << seq_lengths_dims.size();
+
+    SizeVector dst_dims = op->get_output_shape(0);
+    if (src_dims.size() != dst_dims.size())
+        IE_THROW() << errorPrefix << " has incorrect number of input/output sizes!";
+
+    for (size_t i = 0; i < dst_dims.size(); i++) {
+        if (src_dims[i] != dst_dims[i])
+            IE_THROW() << errorPrefix << " has incorrect number of input/output dimension!";
+    }
+
+    seq_axis = revSeq->get_sequence_axis();
+
+    if (seq_axis < 0 || seq_axis >= static_cast<int>(src_dims.size()))
+        IE_THROW() << errorPrefix << " has incorrect 'seq_axis' parameters dimensions and axis number!";
+
+    batch_axis = revSeq->get_batch_axis();
+
+    if (batch_axis < 0 || batch_axis >= static_cast<int>(src_dims.size()))
+        IE_THROW() << errorPrefix << " has incorrect 'batch_axis' parameters dimensions and axis number!";
+
+    if (seq_lengths_dims[0] != dst_dims[batch_axis])
+        IE_THROW() << errorPrefix << " has incorrect 'seq_lengths_dims' parameters dimension!";
+
+    srcStrides.resize(src_dims.size());
+    srcStrides[srcStrides.size() - 1] = 1;
+    for (int i = srcStrides.size() - 2; i >= 0; i--) {
+        srcStrides[i] = srcStrides[i + 1] * src_dims[i + 1];
+    }
+
+    work_amount_dst = srcStrides[0] * src_dims[0];
+}
+
+void MKLDNNReverseSequenceNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    lengthsPrecision = getOriginalInputPrecisionAtPort(REVERSESEQUENCE_LENGTHS);
+    if (lengthsPrecision != Precision::I32 && lengthsPrecision != Precision::FP32)
+        lengthsPrecision = Precision::I32;
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, lengthsPrecision}},
+                         {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNReverseSequenceNode::execute(mkldnn::stream strm) {
+    size_t i;
+    const float *src_data = reinterpret_cast<const float *>(getParentEdgeAt(REVERSESEQUENCE_DATA)->getMemoryPtr()->GetPtr());
+    float* dst_data = reinterpret_cast<float *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+
+    switch (getParentEdgeAt(REVERSESEQUENCE_LENGTHS)->getDesc().getPrecision()) {
+        case Precision::FP32: {
+            float *seq_lengths_data = reinterpret_cast<float *>(getParentEdgeAt(REVERSESEQUENCE_LENGTHS)->getMemoryPtr()->GetPtr());
+            for (i = 0; i < src_dims[batch_axis]; i++) {
+                if (static_cast<int32_t>(seq_lengths_data[i]) > static_cast<int>(src_dims[seq_axis])) {
+                    std::string errorMsg = "Incorrect input 'seq_lengths' values!";
+                    IE_THROW() << errorMsg;
+                }
+            }
+
+            parallel_nt(0, [&](const int ithr, const int nthr) {
+                size_t i, start = 0, end = 0, src_idx = 0;
+                SizeVector counters(src_dims.size(), 0);
+                splitter(work_amount_dst, nthr, ithr, start, end);
+                for (int j = src_dims.size() - 1, i = start; j >= 0; j--) {
+                    counters[j] = i % src_dims[j];
+                    i /= src_dims[j];
+                }
+
+                for (size_t iwork = start; iwork < end; ++iwork) {
+                    for (i = 0, src_idx = 0; i < src_dims.size(); ++i) {
+                        size_t idx = counters[i];
+                        if (static_cast<int>(i) == seq_axis &&
+                            static_cast<int>(idx) < static_cast<int32_t>(seq_lengths_data[counters[batch_axis]])) {
+                            idx = static_cast<int32_t>(seq_lengths_data[counters[batch_axis]]) - idx - 1;
+                        }
+                        src_idx += idx * srcStrides[i];
+                    }
+                    dst_data[iwork] = src_data[src_idx];
+                    for (int j = src_dims.size() - 1; j >= 0; j--) {
+                        counters[j] = (counters[j] + 1) % src_dims[j];
+                        if (counters[j] != 0) break;
+                    }
+                }
+            });
+        }
+        break;
+        case Precision::I32: {
+            int32_t *seq_lengths_data = reinterpret_cast<int32_t *>(getParentEdgeAt(REVERSESEQUENCE_LENGTHS)->getMemoryPtr()->GetPtr());
+            for (i = 0; i < src_dims[batch_axis]; i++) {
+                if (seq_lengths_data[i] > static_cast<int>(src_dims[seq_axis])) {
+                    std::string errorMsg = "Incorrect input 'seq_lengths' values!";
+                    IE_THROW() << errorMsg;
+                }
+            }
+
+            parallel_nt(0, [&](const int ithr, const int nthr) {
+                size_t i, start = 0, end = 0, src_idx = 0;
+                SizeVector counters(src_dims.size(), 0);
+                splitter(work_amount_dst, nthr, ithr, start, end);
+                for (int j = src_dims.size() - 1, i = start; j >= 0; j--) {
+                    counters[j] = i % src_dims[j];
+                    i /= src_dims[j];
+                }
+
+                for (size_t iwork = start; iwork < end; ++iwork) {
+                    for (i = 0, src_idx = 0; i < src_dims.size(); ++i) {
+                        size_t idx = counters[i];
+                        if (static_cast<int>(i) == seq_axis &&
+                            static_cast<int>(idx) < seq_lengths_data[counters[batch_axis]]) {
+                            idx = seq_lengths_data[counters[batch_axis]] - idx - 1;
+                        }
+                        src_idx += idx * srcStrides[i];
+                    }
+                    dst_data[iwork] = src_data[src_idx];
+                    for (int j = src_dims.size() - 1; j >= 0; j--) {
+                        counters[j] = (counters[j] + 1) % src_dims[j];
+                        if (counters[j] != 0) break;
+                    }
+                }
+            });
+        }
+        break;
+        default:
+            IE_THROW() << "ReverseSequence layer does not support "
+                        << getParentEdgeAt(REVERSESEQUENCE_LENGTHS)->getDesc().getPrecision()  << " precision";
+    }
+}
+
+bool MKLDNNReverseSequenceNode::created() const {
+    return getType() == ReverseSequence;
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNReverseSequenceNode, ReverseSequence)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reverse_sequence_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reverse_sequence_node.h
new file mode 100644
index 00000000000000..4b3cf056c63afa
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_reverse_sequence_node.h
@@ -0,0 +1,38 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNReverseSequenceNode : public MKLDNNNode {
+public:
+    MKLDNNReverseSequenceNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+    void initSupportedPrimitiveDescriptors() override;
+    void createPrimitive() override {};
+    void execute(mkldnn::stream strm) override;
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept;
+
+private:
+    const size_t REVERSESEQUENCE_DATA = 0;
+    const size_t REVERSESEQUENCE_LENGTHS = 1;
+
+    int seq_axis;
+    int batch_axis;
+    InferenceEngine::SizeVector src_dims;
+    InferenceEngine::SizeVector srcStrides;
+    size_t work_amount_dst;
+
+    InferenceEngine::Precision lengthsPrecision;
+    std::string errorPrefix;
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_softmax_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_softmax_node.cpp
index 6d4c9a27dc4d8b..53dda785e69115 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_softmax_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_softmax_node.cpp
@@ -14,7 +14,7 @@ using namespace InferenceEngine;
 
 MKLDNNSoftMaxNode::MKLDNNSoftMaxNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng, MKLDNNWeightsSharing::Ptr &cache) :
         MKLDNNNode(op, eng, cache) {
-    auto softmaxOp = ngraph::as_type_ptr<ngraph::op::v1::Softmax>(op);
+    const auto softmaxOp = ngraph::as_type_ptr<ngraph::op::v1::Softmax>(op);
     if (softmaxOp) {
         axis = softmaxOp->get_axis();
     } else {
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_topk_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_topk_node.cpp
new file mode 100644
index 00000000000000..1c78c44b48df5a
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_topk_node.cpp
@@ -0,0 +1,478 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <cmath>
+
+#include <ngraph/op/topk.hpp>
+#include "ie_parallel.hpp"
+#include "mkldnn_topk_node.h"
+#include "utils/general_utils.h"
+
+#if defined(HAVE_SSE) || defined(HAVE_AVX2) || defined(HAVE_AVX512F)
+#include <immintrin.h>
+#endif
+
+using namespace MKLDNNPlugin;
+using namespace InferenceEngine;
+
+bool MKLDNNTopKNode::isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
+    try {
+        const auto topKOp = ngraph::as_type_ptr<const ngraph::op::v1::TopK>(op);
+        if (!topKOp) {
+            errorMessage = "Node is not an instance of the TopK from the operations set v1 or v3";
+            return false;
+        }
+        if (topKOp->get_mode() != ngraph::op::TopKMode::MAX &&
+            topKOp->get_mode() != ngraph::op::TopKMode::MIN) {
+            errorMessage = "Unsupported mode.";
+            return false;
+        }
+        if (!MKLDNNPlugin::one_of(topKOp->get_sort_type(), ngraph::op::TopKSortType::NONE,
+                                  ngraph::op::TopKSortType::SORT_VALUES,
+                                  ngraph::op::TopKSortType::SORT_INDICES)) {
+            errorMessage = "Unsupported sort type.";
+            return false;
+        }
+    } catch (...) {
+        return false;
+    }
+    return true;
+}
+
+MKLDNNTopKNode::MKLDNNTopKNode(const std::shared_ptr<ngraph::Node>& op, const mkldnn::engine& eng,
+                                     MKLDNNWeightsSharing::Ptr &cache) : MKLDNNNode(op, eng, cache) {
+    std::string errorMessage;
+    if (!isSupportedOperation(op, errorMessage)) {
+        IE_THROW(NotImplemented) << errorMessage;
+    }
+    auto topK1Op = ngraph::as_type_ptr<ngraph::op::v1::TopK>(op);
+
+    SizeVector dstDims = topK1Op->get_output_shape(TOPK_VALUE);
+    src_dims = topK1Op->get_input_shape(TOPK_DATA);
+
+    axis = topK1Op->get_axis();
+
+    if (topK1Op->get_mode() == ngraph::op::TopKMode::MAX)
+        mode_max = true;
+    else
+        mode_max = false;
+
+    if (topK1Op->get_sort_type() == ngraph::op::TopKSortType::SORT_VALUES)
+        sort_value = true;
+    else
+        sort_value = false;
+
+    int j;
+    for (j = src_dims.size() - 1; j >= 0; j--) {
+        if (src_dims[j] != 1) break;
+    }
+    if (static_cast<size_t>(j) == axis) is_last_dim = true;
+
+    for (size_t i = 0; i < axis; i++) {
+        axis_step *= src_dims[i];
+    }
+    axis_dim = src_dims[axis];
+    for (size_t i = (axis + 1); i < src_dims.size(); i++) {
+        axis_stride *= src_dims[i];
+    }
+    dim = static_cast<int>(src_dims[axis]);
+    before_num = count(src_dims, 0, axis);
+}
+
+void MKLDNNTopKNode::initSupportedPrimitiveDescriptors() {
+    if (!supportedPrimitiveDescriptors.empty())
+        return;
+
+    std::vector<DataConfigurator> outDataConf;
+    outDataConf.reserve(getOriginalOutputsNumber());
+    outDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::FP32);
+    for (int i = 1; i < getOriginalOutputsNumber(); ++i)
+        outDataConf.emplace_back(TensorDescCreatorTypes::ncsp, Precision::I32);
+
+    addSupportedPrimDesc({{TensorDescCreatorTypes::ncsp, Precision::FP32},
+                          {TensorDescCreatorTypes::ncsp, Precision::I32}},
+                         outDataConf,
+                         impl_desc_type::ref_any);
+}
+
+void MKLDNNTopKNode::execute(mkldnn::stream strm) {
+    const float *src = reinterpret_cast<const float *>(getParentEdgeAt(TOPK_DATA)->getMemoryPtr()->GetPtr());
+    src_k = reinterpret_cast<int *>(getParentEdgeAt(TOPK_K)->getMemoryPtr()->GetPtr())[0];
+    float* dst_data = nullptr;
+    int* dst_idx = nullptr;
+
+    if (outDims.size() == 1) {
+        if (getOriginalOutputPrecisionAtPort(0) == Precision::FP32) {
+            dst_data = reinterpret_cast<float *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+        } else {
+            dst_idx = reinterpret_cast<int *>(getChildEdgesAtPort(0)[0]->getMemoryPtr()->GetPtr());
+        }
+        SizeVector dstDims = getChildEdgesAtPort(0)[0]->getDims().ToSizeVector();
+
+        if (dstDims[axis] != static_cast<size_t>(src_k)) {
+            std::string errorMsg = "Output tensor dimension mismatch";
+            IE_THROW() << errorMsg;
+        }
+    } else if (outDims.size() == 2) {
+        dst_data = reinterpret_cast<float *>(getChildEdgesAtPort(TOPK_VALUE)[0]->getMemoryPtr()->GetPtr());
+        SizeVector dst_data_dims = getChildEdgesAtPort(TOPK_VALUE)[0]->getDims().ToSizeVector();
+
+        dst_idx = reinterpret_cast<int *>(getChildEdgesAtPort(TOPK_INDEX)[0]->getMemoryPtr()->GetPtr());
+        SizeVector dst_idx_dims = getChildEdgesAtPort(TOPK_INDEX)[0]->getDims().ToSizeVector();
+
+        if (dst_idx_dims[axis] != static_cast<size_t>(src_k) || dst_data_dims[axis] != static_cast<size_t>(src_k)) {
+            std::string errorMsg = "Output tensors dimension mismatch";
+            IE_THROW() << errorMsg;
+        }
+    } else {
+        std::string errorMsg = "Output tensors amount mismatch";
+        IE_THROW() << errorMsg;
+    }
+
+    if (src_dims[axis] < static_cast<size_t>(src_k))
+        src_k = src_dims[axis];
+
+    SizeVector in_dims = getParentEdgeAt(TOPK_DATA)->getDims().ToSizeVector();
+
+    if (src_k == 1) {
+        if (is_last_dim) {
+            if (mode_max)
+                top1<std::greater>(src, dst_data, dst_idx, in_dims);
+            else
+                top1<std::less>(src, dst_data, dst_idx, in_dims);
+        } else {
+            if (mode_max)
+                top1_axis<cmpgt_ps, std::greater>(src, dst_data, dst_idx, in_dims);
+            else
+                top1_axis<cmplt_ps, std::less>(src, dst_data, dst_idx, in_dims);
+        }
+    } else {
+        if (is_last_dim) {
+            if (mode_max)
+                topk<std::greater>(src, dst_data, dst_idx, in_dims);
+            else
+                topk<std::less>(src, dst_data, dst_idx, in_dims);
+        } else {
+            if (mode_max)
+                topk_axis<cmpgt_ps, std::greater>(src, dst_data, dst_idx, in_dims);
+            else
+                topk_axis<cmplt_ps, std::less>(src, dst_data, dst_idx, in_dims);
+        }
+    }
+}
+
+bool MKLDNNTopKNode::created() const {
+    return getType() == TopK;
+}
+
+template <class Compare1, template <typename> class Compare2>
+void MKLDNNTopKNode::top1_axis(const float* src_data, float* dst_data, int* dst_idx, SizeVector in_dims) {
+    int after_num = count(in_dims, axis + 1, in_dims.size());
+    int first_index = 0;
+
+#if defined(HAVE_SSE) || defined(HAVE_AVX2) || defined(HAVE_AVX512F)
+    parallel_for2d(before_num, after_num / block_size, [&](int i0, int ib1) {
+            int s_index = i0 * dim * after_num + ib1 * block_size;
+            vec_type_f vmax_val = _mm_uni_loadu_ps(src_data + s_index);
+            vec_type_i vindex_max_val = _mm_uni_setzero_si();
+            for (int i2 = 1; i2 < dim; i2++) {
+                s_index += after_num;
+                vec_type_f vsrc = _mm_uni_loadu_ps(src_data + s_index);
+                vmask_type vmask = Compare1::cmp_ps(vsrc, vmax_val);
+                vmax_val = _mm_uni_blendv_ps(vmax_val, vsrc, vmask);
+
+                vec_type_i vindex_cur_val = _mm_uni_set1_epi32(i2);
+#if defined(HAVE_AVX512F)
+                vindex_max_val = _mm512_mask_blend_epi32(vmask, vindex_max_val, vindex_cur_val);
+#else
+                vindex_max_val = _mm_uni_blendv_epi8(vindex_max_val, vindex_cur_val, _mm_uni_castps_si(vmask));
+#endif
+            }
+            if (dst_data)
+                _mm_uni_storeu_ps(dst_data + i0 * after_num + ib1 * block_size, vmax_val);
+            if (dst_idx)
+                _mm_uni_storeu_si(reinterpret_cast<vec_type_i*>(dst_idx + i0 * after_num + ib1 * block_size), vindex_max_val);
+        });
+        first_index = after_num / block_size * block_size;
+#endif
+    int rest = after_num - first_index;
+    parallel_for2d(before_num, rest, [&](int i0, int i1) {
+        int index_max_val = 0;
+        int s_index = i0 * dim * after_num + first_index + i1;
+        float max_val = src_data[s_index];
+        for (int i2 = 1; i2 < dim; i2++) {
+            s_index += after_num;
+            if (Compare2<float>()(src_data[s_index], max_val)) {
+                max_val = src_data[s_index];
+                index_max_val = i2;
+            }
+        }
+        if (dst_data)
+            dst_data[i0 * after_num + first_index + i1] = max_val;
+        if (dst_idx)
+            dst_idx[i0 * after_num + first_index + i1] = index_max_val;
+    });
+}
+
+template <template <typename> class Compare>
+void MKLDNNTopKNode::top1(const float* src_data, float* dst_data, int* dst_idx, SizeVector in_dims) {
+    parallel_for(before_num, [&](int i0) {
+        int index_max_val = 0;
+        int s_index = i0 * dim;
+        float max_val = src_data[s_index];
+        for (int i1 = 1; i1 < dim; i1++) {
+            s_index++;
+            if (Compare<float>()(src_data[s_index], max_val)) {
+                max_val = src_data[s_index];
+                index_max_val = i1;
+            }
+        }
+        if (dst_data)
+            dst_data[i0] = max_val;
+        if (dst_idx)
+            dst_idx[i0] = index_max_val;
+    });
+}
+
+template <class Compare1, template <typename> class Compare2>
+void MKLDNNTopKNode::topk_axis(const float* src_data, float* dst_data, int* dst_idx, SizeVector in_dims) {
+    int after_num = count(in_dims, axis + 1, in_dims.size());
+    int first_index = 0;
+
+#if defined(HAVE_SSE) || defined(HAVE_AVX2) || defined(HAVE_AVX512F)
+    if (src_k < count_vec) {
+            parallel_for2d(before_num, after_num / block_size, [&](int i0, int ib1) {
+#if defined(HAVE_AVX512F)
+                const int N = 32;
+                vec_type_f vmax_values[N];
+                vec_type_i vmax_indexes[N];
+#else
+                const int N = 16;
+                vec_type_f vmax_values[N];
+                vec_type_i vmax_indexes[N];
+#endif
+                vec_type_f vtmp;
+                vec_type_i vtmp_indexes;
+                vmask_type vmask;
+                int s_index = i0 * dim * after_num + ib1 * block_size;
+
+                auto vswap_func = [&](int index1, int index2) {
+                    vtmp = vmax_values[index1];
+                    vmax_values[index1] = _mm_uni_blendv_ps(vmax_values[index1], vmax_values[index2], vmask);
+                    vmax_values[index2] = _mm_uni_blendv_ps(vmax_values[index2], vtmp, vmask);
+
+                    vtmp_indexes = vmax_indexes[index1];
+#if defined(HAVE_AVX512F)
+                    vmax_indexes[index1] = _mm512_mask_blend_epi32(vmask, vmax_indexes[index1], vmax_indexes[index2]);
+                    vmax_indexes[index2] = _mm512_mask_blend_epi32(vmask, vmax_indexes[index2], vtmp_indexes);
+#else
+                    vmax_indexes[index1] = _mm_uni_blendv_epi8(vmax_indexes[index1], vmax_indexes[index2], _mm_uni_castps_si(vmask));
+                    vmax_indexes[index2] = _mm_uni_blendv_epi8(vmax_indexes[index2], vtmp_indexes, _mm_uni_castps_si(vmask));
+#endif
+                };
+
+                for (int i2 = 0; i2 < src_k; i2++) {
+                    vmax_values[i2] = _mm_uni_loadu_ps(src_data + s_index);
+                    vmax_indexes[i2] = _mm_uni_set1_epi32(i2);
+                    s_index += after_num;
+                }
+                for (int i2 = 0; i2 < src_k - 1; i2++) {
+                    for (int i3 = src_k - 1; i3 > i2; i3--) {
+                        vmask = Compare1::cmp_ps(vmax_values[i3], vmax_values[i3 - 1]);
+#if defined(HAVE_AVX512F)
+                        if (vmask)
+                            vswap_func(i3, i3 - 1);
+#else
+                        int swap = _mm_uni_movemask_ps(vmask);
+                        if (swap)
+                            vswap_func(i3, i3 - 1);
+#endif
+                    }
+                }
+                for (int i2 = src_k; i2 < dim; i2++) {
+                    vmax_values[src_k] = _mm_uni_loadu_ps(src_data + s_index);
+                    vmax_indexes[src_k] = _mm_uni_set1_epi32(i2);
+                    for (int i3 = src_k; i3 > 0; i3--) {
+                        vmask = Compare1::cmp_ps(vmax_values[i3], vmax_values[i3 - 1]);
+#if defined(HAVE_AVX512F)
+                        if (vmask)
+                            vswap_func(i3, i3 - 1);
+                        else
+                            break;
+#else
+                        int swap = _mm_uni_movemask_ps(vmask);
+                        if (swap)
+                            vswap_func(i3, i3 - 1);
+                        else
+                            break;
+#endif
+                    }
+                    s_index += after_num;
+                }
+                if (!sort_value) {
+                    for (int i2 = 0; i2 < src_k - 1; i2++) {
+                        for (int i3 = src_k - 1; i3 > i2; i3--) {
+                            vmask = _mm_uni_cmpgt_i32(vmax_indexes[i3 - 1], vmax_indexes[i3]);
+#if defined(HAVE_AVX512F)
+                            if (vmask)
+                                vswap_func(i3, i3 - 1);
+                            else
+                                break;
+#else
+                            int swap = _mm_uni_movemask_ps(vmask);
+                            if (swap)
+                                vswap_func(i3, i3 - 1);
+                            else
+                                break;
+#endif
+                        }
+                    }
+                }
+                if (dst_data) {
+                    for (int i2 = 0; i2 < src_k; i2++)
+                        _mm_uni_storeu_ps(dst_data + (i0 * src_k + i2) * after_num + ib1 * block_size, vmax_values[i2]);
+                }
+                if (dst_idx) {
+                    for (int i2 = 0; i2 < src_k; i2++)
+                        _mm_uni_storeu_si(reinterpret_cast<vec_type_i*>(dst_idx + (i0 * src_k + i2) * after_num + ib1 * block_size), vmax_indexes[i2]);
+                }
+            });
+            first_index = after_num / block_size * block_size;
+        }
+#endif
+    int rest = after_num - first_index;
+    parallel_for2d(before_num, rest, [&](int i0, int i1) {
+        std::vector<float> max_values(src_k + 1);
+        std::vector<int> max_indexes(src_k + 1);
+        float tmp_value;
+        int tmp_index;
+        int s_index = i0 * dim * after_num + first_index + i1;
+
+        auto swap_func = [&](int index1, int index2) {
+            tmp_value = max_values[index1];
+            max_values[index1] = max_values[index2];
+            max_values[index2] = tmp_value;
+
+            tmp_index = max_indexes[index1];
+            max_indexes[index1] = max_indexes[index2];
+            max_indexes[index2] = tmp_index;
+        };
+
+        for (int i2 = 0; i2 < src_k; i2++) {
+            max_values[i2] = src_data[s_index];
+            max_indexes[i2] = i2;
+            s_index += after_num;
+        }
+        for (int i2 = 0; i2 < src_k - 1; i2++) {
+            for (int i3 = src_k - 1; i3 > i2; i3--) {
+                if (Compare2<float>()(max_values[i3], max_values[i3 - 1])) {
+                    swap_func(i3, i3 - 1);
+                }
+            }
+        }
+        for (int i2 = src_k; i2 < dim; i2++) {
+            max_values[src_k] = src_data[s_index];
+            max_indexes[src_k] = i2;
+            for (int i3 = src_k; i3 > 0; i3--) {
+                if (Compare2<float>()(max_values[i3], max_values[i3 - 1]))
+                    swap_func(i3, i3 - 1);
+                else
+                    break;
+            }
+            s_index += after_num;
+        }
+        if (!sort_value) {
+            for (int i2 = 0; i2 < src_k - 1; i2++) {
+                for (int i3 = src_k - 1; i3 > i2; i3--) {
+                    if (std::greater<int>()(max_indexes[i3 - 1], max_indexes[i3])) {
+                        swap_func(i3, i3 - 1);
+                    }
+                }
+            }
+        }
+        if (dst_data) {
+            for (int i2 = 0; i2 < src_k; i2++)
+                dst_data[i0 * src_k * after_num + i2 * after_num + first_index + i1] = max_values[i2];
+        }
+        if (dst_idx) {
+            for (int i2 = 0; i2 < src_k; i2++)
+                dst_idx[i0 * src_k * after_num + i2 * after_num + first_index + i1] = max_indexes[i2];
+        }
+    });
+}
+
+template <template <typename> class Compare>
+void MKLDNNTopKNode::topk(const float* src_data, float* dst_data, int* dst_idx, SizeVector in_dims) {
+    parallel_for(before_num, [&](int i0) {
+        std::vector<float> max_values(src_k + 1);
+        std::vector<int> max_indexes(src_k + 1);
+        float tmp_value;
+        int tmp_index;
+        int s_index = i0 * dim;
+
+        auto swap_func = [&](int index1, int index2) {
+            tmp_value = max_values[index1];
+            max_values[index1] = max_values[index2];
+            max_values[index2] = tmp_value;
+
+            tmp_index = max_indexes[index1];
+            max_indexes[index1] = max_indexes[index2];
+            max_indexes[index2] = tmp_index;
+        };
+
+        for (int i2 = 0; i2 < src_k; i2++) {
+            max_values[i2] = src_data[s_index];
+            max_indexes[i2] = i2;
+            s_index++;
+        }
+        for (int i2 = 0; i2 < src_k - 1; i2++) {
+            for (int i3 = src_k - 1; i3 > i2; i3--) {
+                if (Compare<float>()(max_values[i3], max_values[i3 - 1])) {
+                    swap_func(i3, i3 - 1);
+                }
+            }
+        }
+        for (int i2 = src_k; i2 < dim; i2++) {
+            max_values[src_k] = src_data[s_index];
+            max_indexes[src_k] = i2;
+            for (int i3 = src_k; i3 > 0; i3--) {
+                if (Compare<float>()(max_values[i3], max_values[i3 - 1]))
+                    swap_func(i3, i3 - 1);
+                else
+                    break;
+            }
+            s_index++;
+        }
+        if (!sort_value) {
+            for (int i2 = 0; i2 < src_k - 1; i2++) {
+                for (int i3 = src_k - 1; i3 > i2; i3--) {
+                    if (std::greater<int>()(max_indexes[i3 - 1], max_indexes[i3])) {
+                        swap_func(i3, i3 - 1);
+                    }
+                }
+            }
+        }
+        if (dst_data) {
+            for (int i2 = 0; i2 < src_k; i2++)
+                dst_data[i0 * src_k + i2] = max_values[i2];
+        }
+        if (dst_idx) {
+            for (int i2 = 0; i2 < src_k; i2++)
+                dst_idx[i0 * src_k + i2] = max_indexes[i2];
+        }
+    });
+}
+
+inline int MKLDNNTopKNode::count(SizeVector dims, size_t start_ind, size_t end_ind) {
+    size_t count = 1;
+    for (size_t i = start_ind; i < end_ind; i++)
+        count *= dims[i];
+    return static_cast<int>(count);
+}
+
+inline int MKLDNNTopKNode::count(SizeVector dims, size_t start_ind) {
+    return count(dims, start_ind, dims.size());
+}
+
+REG_MKLDNN_PRIM_FOR(MKLDNNTopKNode, TopK)
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_topk_node.h b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_topk_node.h
new file mode 100644
index 00000000000000..b8594c40363731
--- /dev/null
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_topk_node.h
@@ -0,0 +1,114 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#pragma once
+
+#include "ie_common.h"
+#include <ie_common.h>
+#include <mkldnn_node.h>
+
+namespace MKLDNNPlugin {
+
+class MKLDNNTopKNode : public MKLDNNNode {
+public:
+    MKLDNNTopKNode(const std::shared_ptr<ngraph::Node> &op, const mkldnn::engine &eng,
+                   MKLDNNWeightsSharing::Ptr &cache);
+
+    void getSupportedDescriptors() override {};
+
+    void initSupportedPrimitiveDescriptors() override;
+
+    void createPrimitive() override {};
+
+    void execute(mkldnn::stream strm) override;
+
+    bool created() const override;
+
+    static bool isSupportedOperation(const std::shared_ptr<ngraph::Node> &op, std::string &errorMessage) noexcept;
+
+#if defined(HAVE_AVX512F)
+    const int block_size = 16;
+    typedef __m512 vec_type_f;
+    typedef __m512i vec_type_i;
+    typedef __mmask16 vmask_type;
+#elif defined(HAVE_AVX2)
+    const int block_size = 8;
+    typedef __m256 vec_type_f;
+    typedef __m256i vec_type_i;
+    typedef __m256 vmask_type;
+#elif defined(HAVE_SSE)
+    const int block_size = 4;
+    typedef __m128 vec_type_f;
+    typedef __m128i vec_type_i;
+    typedef __m128 vmask_type;
+#else
+    typedef float vec_type_f;
+    typedef int vmask_type;
+#endif
+
+    struct cmpgt_ps {
+        static inline vmask_type cmp_ps(const vec_type_f _Left, const vec_type_f _Right) {
+#if defined(HAVE_SSE) || defined(HAVE_AVX2) || defined(HAVE_AVX512F)
+            return _mm_uni_cmpgt_ps(_Left, _Right);
+#else
+            return _Left > _Right ? _Left : _Right;
+#endif
+        }
+    };
+
+    struct cmplt_ps {
+        static inline vmask_type cmp_ps(const vec_type_f _Left, const vec_type_f _Right) {
+#if defined(HAVE_SSE) || defined(HAVE_AVX2) || defined(HAVE_AVX512F)
+            return _mm_uni_cmpgt_ps(_Right, _Left);
+#else
+            return _Right > _Left ? _Right : _Left;
+#endif
+        }
+    };
+
+    template<class Compare1, template<typename> class Compare2>
+    void top1_axis(const float *src_data, float *dst_data, int *dst_idx, InferenceEngine::SizeVector in_dims);
+
+    template<template<typename> class Compare>
+    void top1(const float *src_data, float *dst_data, int *dst_idx, InferenceEngine::SizeVector in_dims);
+
+    template<class Compare1, template<typename> class Compare2>
+    void topk_axis(const float *src_data, float *dst_data, int *dst_idx, InferenceEngine::SizeVector in_dims);
+
+    template<template<typename> class Compare>
+    void topk(const float *src_data, float *dst_data, int *dst_idx, InferenceEngine::SizeVector in_dims);
+
+private:
+    const size_t TOPK_DATA = 0;
+    const size_t TOPK_K = 1;
+    const size_t TOPK_VALUE = 0;
+    const size_t TOPK_INDEX = 1;
+
+    InferenceEngine::SizeVector src_dims;
+    size_t axis;
+    size_t axis_dim;
+    size_t axis_stride = 1;
+    size_t axis_step = 1;
+    bool is_last_dim = false;
+    int src_k = 1;
+
+    bool sort_value = false;
+    bool mode_max = true;
+
+    int dim, before_num;
+
+    std::string errorPrefix;
+
+#if defined(HAVE_AVX512F)
+    const int count_vec = 32;
+#elif defined(HAVE_SSE) || defined(HAVE_AVX2)
+    const int count_vec = 16;
+#endif
+
+    inline int count(InferenceEngine::SizeVector dims, size_t start_ind, size_t end_ind);
+
+    inline int count(InferenceEngine::SizeVector dims, size_t start_ind = 0);
+};
+
+}  // namespace MKLDNNPlugin
diff --git a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_transpose_node.cpp b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_transpose_node.cpp
index eaf93abd611cbe..49bc1bb695dd1b 100644
--- a/inference-engine/src/mkldnn_plugin/nodes/mkldnn_transpose_node.cpp
+++ b/inference-engine/src/mkldnn_plugin/nodes/mkldnn_transpose_node.cpp
@@ -19,7 +19,7 @@ using namespace InferenceEngine;
 
 bool MKLDNNTransposeNode::isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
     try {
-        auto transposeOp = ngraph::as_type_ptr<const ngraph::op::v1::Transpose>(op);
+        const auto transposeOp = ngraph::as_type_ptr<const ngraph::op::v1::Transpose>(op);
         if (!transposeOp) {
             errorMessage = "Node is not an instance of the Transpose operation.";
             return false;
diff --git a/inference-engine/src/mkldnn_plugin/nodes/non_max_suppression.cpp b/inference-engine/src/mkldnn_plugin/nodes/non_max_suppression.cpp
deleted file mode 100644
index b5884a234cbb2e..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/non_max_suppression.cpp
+++ /dev/null
@@ -1,464 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <cmath>
-#include <string>
-#include <vector>
-#include <cassert>
-#include <algorithm>
-#include <utility>
-#include <queue>
-#include "ie_parallel.hpp"
-#include <ngraph_ops/nms_ie_internal.hpp>
-#include "utils/general_utils.h"
-#include <ie_ngraph_utils.hpp>
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using namespace MKLDNNPlugin;
-
-class NonMaxSuppressionImpl: public ExtLayerBase {
-public:
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            const auto nms = std::dynamic_pointer_cast<const ngraph::op::internal::NonMaxSuppressionIEInternal>(op);
-            if (!nms) {
-                errorMessage = "Only internal NonMaxSuppression operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    explicit NonMaxSuppressionImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            errorPrefix = "NMS layer with name '" + op->get_friendly_name() + "' ";
-            const auto nms = std::dynamic_pointer_cast<const ngraph::op::internal::NonMaxSuppressionIEInternal>(op);
-
-            if (nms->get_input_size() < 2 || nms->get_input_size() > 6)
-                IE_THROW() << errorPrefix << "has incorrect number of input edges: " << nms->get_input_size();
-
-            if (nms->get_output_size() < 1 || nms->get_output_size() > 3)
-                IE_THROW() << errorPrefix << "has incorrect number of output edges: " << nms->get_output_size();
-
-            boxEncodingType = nms->m_center_point_box ? boxEncoding::CENTER : boxEncoding::CORNER;
-
-            sort_result_descending = nms->m_sort_result_descending;
-
-            const std::vector<Precision> supportedFloatPrecision = {Precision::FP32, Precision::BF16};
-            const std::vector<Precision> supportedIntOutputPrecision = {Precision::I32, Precision::I64};
-
-            checkPrecision(op->get_input_element_type(NMS_BOXES), supportedFloatPrecision, "boxes", inType);
-            const SizeVector &boxes_dims = op->get_input_shape(NMS_BOXES);
-            num_batches = boxes_dims[0];
-            num_boxes = boxes_dims[1];
-            if (boxes_dims.size() != 3)
-                IE_THROW() << errorPrefix << "has unsupported 'boxes' input rank: " << boxes_dims.size();
-            if (boxes_dims[2] != 4)
-                IE_THROW() << errorPrefix << "has unsupported 'boxes' input 3rd dimension size: " << boxes_dims[2];
-
-            checkPrecision(op->get_input_element_type(NMS_SCORES), supportedFloatPrecision, "scores", inType);
-            const SizeVector &scores_dims = op->get_input_shape(NMS_SCORES);
-            num_classes = scores_dims[1];
-            if (scores_dims.size() != 3)
-                IE_THROW() << errorPrefix << "has unsupported 'scores' input rank: " << scores_dims.size();
-
-            if (num_batches != scores_dims[0])
-                IE_THROW() << errorPrefix << " num_batches is different in 'boxes' and 'scores' inputs";
-            if (num_boxes != scores_dims[2])
-                IE_THROW() << errorPrefix << " num_boxes is different in 'boxes' and 'scores' inputs";
-
-            numFiltBox.resize(num_batches);
-            for (size_t i = 0; i < numFiltBox.size(); i++)
-                numFiltBox[i].resize(num_classes);
-
-            const std::vector<Precision> supportedPrecision = {Precision::I16, Precision::U8, Precision::I8, Precision::U16, Precision::I32,
-                                                               Precision::U32, Precision::I64, Precision::U64};
-            check1DInput(op, supportedPrecision, "max_output_boxes_per_class", NMS_MAXOUTPUTBOXESPERCLASS);
-            check1DInput(op, supportedFloatPrecision, "iou_threshold", NMS_IOUTHRESHOLD);
-            check1DInput(op, supportedFloatPrecision, "score_threshold", NMS_SCORETHRESHOLD);
-
-            if (op->get_input_size() > NMS_SOFTNMSSIGMA) {
-                check1DInput(op, supportedFloatPrecision, "soft_nms_sigma", NMS_SOFTNMSSIGMA);
-            }
-
-            checkOutput(op, supportedIntOutputPrecision, "selected_indices", NMS_SELECTEDINDICES);
-            checkOutput(op, supportedFloatPrecision, "selected_scores", NMS_SELECTEDSCORES);
-            checkPrecision(op->get_input_element_type(NMS_VALIDOUTPUTS), supportedIntOutputPrecision, "valid_outputs", outType);
-            const SizeVector &valid_outputs_dims = op->get_input_shape(NMS_VALIDOUTPUTS);
-            if (valid_outputs_dims.size() != 1)
-                IE_THROW() << errorPrefix << "has unsupported 'valid_outputs' output rank: " << valid_outputs_dims.size();
-            if (valid_outputs_dims[0] != 1)
-                IE_THROW() << errorPrefix << "has unsupported 'valid_outputs' output 1st dimension size: " << valid_outputs_dims[1];
-
-            LayerConfig config;
-            for (size_t i = 0; i < op->get_input_size(); i++) {
-                DataConfig inConfig;
-
-                Precision inPrecision = i == NMS_MAXOUTPUTBOXESPERCLASS ? Precision::I32 : Precision::FP32;
-                const SizeVector& inDims = op->get_input_shape(i);
-                inConfig.desc = TensorDesc(inPrecision, inDims, InferenceEngine::TensorDesc::getLayoutByDims(inDims));
-                config.inConfs.push_back(inConfig);
-            }
-            for (size_t i = 0; i < op->get_output_size(); i++) {
-                DataConfig outConfig;
-
-                Precision outPrecision = i == NMS_SELECTEDSCORES ? Precision::FP32 : Precision::I32;
-                const SizeVector& outDims = op->get_output_shape(i);
-                outConfig.desc = TensorDesc(outPrecision, outDims, InferenceEngine::TensorDesc::getLayoutByDims(outDims));
-                config.outConfs.push_back(outConfig);
-            }
-
-            config.dynBatchSupport = false;
-            confs.push_back(config);
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    float intersectionOverUnion(const float *boxesI, const float *boxesJ) {
-        float yminI, xminI, ymaxI, xmaxI, yminJ, xminJ, ymaxJ, xmaxJ;
-        if (boxEncodingType == boxEncoding::CENTER) {
-            //  box format: x_center, y_center, width, height
-            yminI = boxesI[1] - boxesI[3] / 2.f;
-            xminI = boxesI[0] - boxesI[2] / 2.f;
-            ymaxI = boxesI[1] + boxesI[3] / 2.f;
-            xmaxI = boxesI[0] + boxesI[2] / 2.f;
-            yminJ = boxesJ[1] - boxesJ[3] / 2.f;
-            xminJ = boxesJ[0] - boxesJ[2] / 2.f;
-            ymaxJ = boxesJ[1] + boxesJ[3] / 2.f;
-            xmaxJ = boxesJ[0] + boxesJ[2] / 2.f;
-        } else {
-            //  box format: y1, x1, y2, x2
-            yminI = (std::min)(boxesI[0], boxesI[2]);
-            xminI = (std::min)(boxesI[1], boxesI[3]);
-            ymaxI = (std::max)(boxesI[0], boxesI[2]);
-            xmaxI = (std::max)(boxesI[1], boxesI[3]);
-            yminJ = (std::min)(boxesJ[0], boxesJ[2]);
-            xminJ = (std::min)(boxesJ[1], boxesJ[3]);
-            ymaxJ = (std::max)(boxesJ[0], boxesJ[2]);
-            xmaxJ = (std::max)(boxesJ[1], boxesJ[3]);
-        }
-
-        float areaI = (ymaxI - yminI) * (xmaxI - xminI);
-        float areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);
-        if (areaI <= 0.f || areaJ <= 0.f)
-            return 0.f;
-
-        float intersection_area =
-            (std::max)((std::min)(ymaxI, ymaxJ) - (std::max)(yminI, yminJ), 0.f) *
-            (std::max)((std::min)(xmaxI, xmaxJ) - (std::max)(xminI, xminJ), 0.f);
-        return intersection_area / (areaI + areaJ - intersection_area);
-    }
-
-    struct filteredBoxes {
-        float score;
-        int batch_index;
-        int class_index;
-        int box_index;
-        filteredBoxes() = default;
-        filteredBoxes(float _score, int _batch_index, int _class_index, int _box_index) :
-                      score(_score), batch_index(_batch_index), class_index(_class_index), box_index(_box_index) {}
-    };
-
-    struct boxInfo {
-        float score;
-        int idx;
-        int suppress_begin_index;
-    };
-
-    void nmsWithSoftSigma(const float *boxes, const float *scores, const SizeVector &boxesStrides, const SizeVector &scoresStrides,
-                          std::vector<filteredBoxes> &filtBoxes) {
-        auto less = [](const boxInfo& l, const boxInfo& r) {
-            return l.score < r.score || ((l.score == r.score) && (l.idx > r.idx));
-        };
-
-        auto coeff = [&](float iou) {
-            const float weight = std::exp(scale * iou * iou);
-            return iou <= iou_threshold ? weight : 0.0f;
-        };
-
-        parallel_for2d(num_batches, num_classes, [&](int batch_idx, int class_idx) {
-            std::vector<filteredBoxes> fb;
-            const float *boxesPtr = boxes + batch_idx * boxesStrides[0];
-            const float *scoresPtr = scores + batch_idx * scoresStrides[0] + class_idx * scoresStrides[1];
-
-            std::priority_queue<boxInfo, std::vector<boxInfo>, decltype(less)> sorted_boxes(less);
-            for (int box_idx = 0; box_idx < num_boxes; box_idx++) {
-                if (scoresPtr[box_idx] > score_threshold)
-                    sorted_boxes.emplace(boxInfo({scoresPtr[box_idx], box_idx, 0}));
-            }
-
-            fb.reserve(sorted_boxes.size());
-            if (sorted_boxes.size() > 0) {
-                while (fb.size() < max_output_boxes_per_class && !sorted_boxes.empty()) {
-                    boxInfo currBox = sorted_boxes.top();
-                    float origScore = currBox.score;
-                    sorted_boxes.pop();
-
-                    bool box_is_selected = true;
-                    for (int idx = static_cast<int>(fb.size()) - 1; idx >= currBox.suppress_begin_index; idx--) {
-                        float iou = intersectionOverUnion(&boxesPtr[currBox.idx * 4], &boxesPtr[fb[idx].box_index * 4]);
-                        currBox.score *= coeff(iou);
-                        if (iou >= iou_threshold) {
-                            box_is_selected = false;
-                            break;
-                        }
-                        if (currBox.score <= score_threshold)
-                            break;
-                    }
-
-                    currBox.suppress_begin_index = fb.size();
-                    if (box_is_selected) {
-                        if (currBox.score == origScore) {
-                            fb.push_back({ currBox.score, batch_idx, class_idx, currBox.idx });
-                            continue;
-                        }
-                        if (currBox.score > score_threshold) {
-                            sorted_boxes.push(currBox);
-                        }
-                    }
-                }
-            }
-            numFiltBox[batch_idx][class_idx] = fb.size();
-            size_t offset = batch_idx*num_classes*max_output_boxes_per_class + class_idx*max_output_boxes_per_class;
-            for (size_t i = 0; i < fb.size(); i++) {
-                filtBoxes[offset + i] = fb[i];
-            }
-        });
-    }
-
-    void nmsWithoutSoftSigma(const float *boxes, const float *scores, const SizeVector &boxesStrides, const SizeVector &scoresStrides,
-                             std::vector<filteredBoxes> &filtBoxes) {
-        int max_out_box = static_cast<int>(max_output_boxes_per_class);
-        parallel_for2d(num_batches, num_classes, [&](int batch_idx, int class_idx) {
-            const float *boxesPtr = boxes + batch_idx * boxesStrides[0];
-            const float *scoresPtr = scores + batch_idx * scoresStrides[0] + class_idx * scoresStrides[1];
-
-            std::vector<std::pair<float, int>> sorted_boxes;
-            for (int box_idx = 0; box_idx < num_boxes; box_idx++) {
-                if (scoresPtr[box_idx] > score_threshold)
-                    sorted_boxes.emplace_back(std::make_pair(scoresPtr[box_idx], box_idx));
-            }
-
-            int io_selection_size = 0;
-            if (sorted_boxes.size() > 0) {
-                parallel_sort(sorted_boxes.begin(), sorted_boxes.end(),
-                              [](const std::pair<float, int>& l, const std::pair<float, int>& r) {
-                                    return (l.first > r.first || ((l.first == r.first) && (l.second < r.second)));
-                                });
-                int offset = batch_idx*num_classes*max_output_boxes_per_class + class_idx*max_output_boxes_per_class;
-                filtBoxes[offset + 0] = filteredBoxes(sorted_boxes[0].first, batch_idx, class_idx, sorted_boxes[0].second);
-                io_selection_size++;
-                for (size_t box_idx = 1; (box_idx < sorted_boxes.size()) && (io_selection_size < max_out_box); box_idx++) {
-                    bool box_is_selected = true;
-                    for (int idx = io_selection_size - 1; idx >= 0; idx--) {
-                        float iou = intersectionOverUnion(&boxesPtr[sorted_boxes[box_idx].second * 4], &boxesPtr[filtBoxes[offset + idx].box_index * 4]);
-                        if (iou >= iou_threshold) {
-                            box_is_selected = false;
-                            break;
-                        }
-                    }
-
-                    if (box_is_selected) {
-                        filtBoxes[offset + io_selection_size] = filteredBoxes(sorted_boxes[box_idx].first, batch_idx, class_idx, sorted_boxes[box_idx].second);
-                        io_selection_size++;
-                    }
-                }
-            }
-            numFiltBox[batch_idx][class_idx] = io_selection_size;
-        });
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs, ResponseDesc *resp) noexcept override {
-        const float *boxes = inputs[NMS_BOXES]->cbuffer().as<const float *>() + inputs[NMS_BOXES]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        const float *scores = inputs[NMS_SCORES]->cbuffer().as<const float *>() + inputs[NMS_SCORES]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-        max_output_boxes_per_class = outputs.size() > NMS_SELECTEDSCORES ? 0 : num_boxes;
-        if (inputs.size() > NMS_MAXOUTPUTBOXESPERCLASS) {
-            max_output_boxes_per_class = (inputs[NMS_MAXOUTPUTBOXESPERCLASS]->cbuffer().as<int *>() +
-                                          inputs[NMS_MAXOUTPUTBOXESPERCLASS]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0];
-        }
-
-        if (max_output_boxes_per_class == 0)
-            return OK;
-
-        iou_threshold = outputs.size() > NMS_SELECTEDSCORES ? 0.0f : 1.0f;
-        if (inputs.size() > NMS_IOUTHRESHOLD)
-            iou_threshold = (inputs[NMS_IOUTHRESHOLD]->cbuffer().as<float *>() +
-                             inputs[NMS_IOUTHRESHOLD]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0];
-
-        score_threshold = 0.0f;
-        if (inputs.size() > NMS_SCORETHRESHOLD)
-            score_threshold = (inputs[NMS_SCORETHRESHOLD]->cbuffer().as<float *>() +
-                               inputs[NMS_SCORETHRESHOLD]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0];
-
-        soft_nms_sigma = 0.0f;
-        if (inputs.size() > NMS_SOFTNMSSIGMA)
-            soft_nms_sigma = (inputs[NMS_SOFTNMSSIGMA]->cbuffer().as<float *>() +
-                              inputs[NMS_SOFTNMSSIGMA]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0];
-        scale = 0.0f;
-        if (soft_nms_sigma > 0.0) {
-            scale = -0.5 / soft_nms_sigma;
-        }
-
-        int *selected_indices = outputs[NMS_SELECTEDINDICES]->buffer().as<int *>() +
-                                outputs[NMS_SELECTEDINDICES]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-        float *selected_scores = nullptr;
-        if (outputs.size() > NMS_SELECTEDSCORES)
-            selected_scores = outputs[NMS_SELECTEDSCORES]->buffer().as<float *>() +
-                              outputs[NMS_SELECTEDSCORES]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-        int *valid_outputs = nullptr;
-        if (outputs.size() > NMS_VALIDOUTPUTS)
-            valid_outputs = outputs[NMS_VALIDOUTPUTS]->buffer().as<int *>() +
-                            outputs[NMS_VALIDOUTPUTS]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-        const SizeVector &boxesStrides = inputs[NMS_BOXES]->getTensorDesc().getBlockingDesc().getStrides();
-        const SizeVector &scoresStrides = inputs[NMS_SCORES]->getTensorDesc().getBlockingDesc().getStrides();
-
-        std::vector<filteredBoxes> filtBoxes(max_output_boxes_per_class * num_batches * num_classes);
-
-        if (soft_nms_sigma == 0.0f) {
-            nmsWithoutSoftSigma(boxes, scores, boxesStrides, scoresStrides, filtBoxes);
-        } else {
-            nmsWithSoftSigma(boxes, scores, boxesStrides, scoresStrides, filtBoxes);
-        }
-
-        size_t startOffset = numFiltBox[0][0];
-        for (size_t b = 0; b < numFiltBox.size(); b++) {
-            size_t batchOffset = b*num_classes*max_output_boxes_per_class;
-            for (size_t c = (b == 0 ? 1 : 0); c < numFiltBox[b].size(); c++) {
-                size_t offset = batchOffset + c*max_output_boxes_per_class;
-                for (size_t i = 0; i < numFiltBox[b][c]; i++) {
-                    filtBoxes[startOffset + i] = filtBoxes[offset + i];
-                }
-                startOffset += numFiltBox[b][c];
-            }
-        }
-        filtBoxes.resize(startOffset);
-
-        // need more particular comparator to get deterministic behaviour
-        // escape situation when filtred boxes with same score have different position from launch to launch
-        if (sort_result_descending) {
-            parallel_sort(filtBoxes.begin(), filtBoxes.end(),
-                          [](const filteredBoxes& l, const filteredBoxes& r) {
-                                return (l.score > r.score) ||
-                                (l.score ==  r.score && l.batch_index < r.batch_index) ||
-                                (l.score ==  r.score && l.batch_index == r.batch_index && l.class_index < r.class_index) ||
-                                (l.score ==  r.score && l.batch_index == r.batch_index && l.class_index == r.class_index && l.box_index < r.box_index);
-                            });
-        }
-
-        const size_t selectedBoxesNum = outputs[NMS_SELECTEDINDICES]->getTensorDesc().getDims()[0];
-        const size_t validOutputs = std::min(filtBoxes.size(), selectedBoxesNum);
-
-        int selectedIndicesStride = outputs[NMS_SELECTEDINDICES]->getTensorDesc().getBlockingDesc().getStrides()[0];
-        int *selectedIndicesPtr = selected_indices;
-        float *selectedScoresPtr = selected_scores;
-
-        size_t idx = 0lu;
-        for (; idx < validOutputs; idx++) {
-            selectedIndicesPtr[0] = filtBoxes[idx].batch_index;
-            selectedIndicesPtr[1] = filtBoxes[idx].class_index;
-            selectedIndicesPtr[2] = filtBoxes[idx].box_index;
-            selectedIndicesPtr += selectedIndicesStride;
-            if (outputs.size() > NMS_SELECTEDSCORES) {
-                selectedScoresPtr[0] = static_cast<float>(filtBoxes[idx].batch_index);
-                selectedScoresPtr[1] = static_cast<float>(filtBoxes[idx].class_index);
-                selectedScoresPtr[2] = static_cast<float>(filtBoxes[idx].score);
-                selectedScoresPtr += selectedIndicesStride;
-            }
-        }
-        std::fill(selectedIndicesPtr, selectedIndicesPtr + (selectedBoxesNum - idx) * selectedIndicesStride, -1);
-        if (outputs.size() > NMS_SELECTEDSCORES) {
-            std::fill(selectedScoresPtr, selectedScoresPtr + (selectedBoxesNum - idx) * selectedIndicesStride, -1.f);
-        }
-        if (outputs.size() > NMS_VALIDOUTPUTS)
-            *valid_outputs = static_cast<int>(validOutputs);
-
-        return OK;
-    }
-
-private:
-    // input
-    const size_t NMS_BOXES = 0;
-    const size_t NMS_SCORES = 1;
-    const size_t NMS_MAXOUTPUTBOXESPERCLASS = 2;
-    const size_t NMS_IOUTHRESHOLD = 3;
-    const size_t NMS_SCORETHRESHOLD = 4;
-    const size_t NMS_SOFTNMSSIGMA = 5;
-
-    // output
-    const size_t NMS_SELECTEDINDICES = 0;
-    const size_t NMS_SELECTEDSCORES = 1;
-    const size_t NMS_VALIDOUTPUTS = 2;
-
-    enum class boxEncoding {
-        CORNER,
-        CENTER
-    };
-    boxEncoding boxEncodingType = boxEncoding::CORNER;
-    bool sort_result_descending = true;
-
-    size_t num_batches;
-    size_t num_boxes;
-    size_t num_classes;
-
-    size_t max_output_boxes_per_class = 0lu;
-    float iou_threshold = 0.0f;
-    float score_threshold = 0.0f;
-    float soft_nms_sigma = 0.0f;
-    float scale = 1.f;
-
-    std::string errorPrefix;
-
-    std::vector<std::vector<size_t>> numFiltBox;
-    const std::string inType = "input", outType = "output";
-
-    void checkPrecision(const ngraph::element::Type &ngPrec, const std::vector<Precision> precList, const std::string name, const std::string type) {
-        const auto prec = details::convertPrecision(ngPrec);
-        if (std::find(precList.begin(), precList.end(), prec) == precList.end())
-            IE_THROW() << errorPrefix << "has unsupported '" << name << "' " << type << " precision: " << prec;
-    }
-
-    void check1DInput(const std::shared_ptr<ngraph::Node>& op, const std::vector<Precision> precList, const std::string name, const size_t port) {
-        checkPrecision(op->get_input_element_type(port), precList, name, inType);
-
-        const SizeVector &dims = op->get_input_shape(port);
-        if (dims.size() != 0 && dims.size() != 1)
-            IE_THROW() << errorPrefix << "has unsupported '" << name << "' input rank: " << dims.size();
-        if (dims.size() == 1)
-            if (dims[0] != 1)
-                IE_THROW() << errorPrefix << "has unsupported '" << name << "' input 1st dimension size: " << dims[0];
-    }
-
-    void checkOutput(const std::shared_ptr<ngraph::Node>& op, const std::vector<Precision> precList, const std::string name, const size_t port) {
-        checkPrecision(op->get_output_element_type(port), precList, name, outType);
-
-        const SizeVector &dims = op->get_output_shape(port);
-        if (dims.size() != 2)
-            IE_THROW() << errorPrefix << "has unsupported '" << name << "' output rank: " << dims.size();
-        if (dims[1] != 3)
-            IE_THROW() << errorPrefix << "has unsupported '" << name << "' output 2nd dimension size: " << dims[1];
-    }
-};
-
-REG_FACTORY_FOR(NonMaxSuppressionImpl, NonMaxSuppressionIEInternal);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/priorgridgenerator_onnx.cpp b/inference-engine/src/mkldnn_plugin/nodes/priorgridgenerator_onnx.cpp
deleted file mode 100644
index 662c24086ee172..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/priorgridgenerator_onnx.cpp
+++ /dev/null
@@ -1,121 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-#include <algorithm>
-#include <cassert>
-#include <vector>
-#include <ngraph/opsets/opset6.hpp>
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-const int INPUT_PRIORS {0};
-const int INPUT_FEATUREMAP {1};
-const int INPUT_IMAGE {2};
-
-const int OUTPUT_ROIS {0};
-
-class ExperimentalDetectronPriorGridGeneratorImpl: public ExtLayerBase {
-private:
-    // Inputs:
-    //      priors, shape [n, 4]
-    //      [feature_map], shape [b, c, h, w]
-    //      [im_data], shape [b, 3, im_h, im_w]
-    // Outputs:
-    //      priors_grid, shape [m, 4]
-
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            const auto priorGridGen = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronPriorGridGenerator>(op);
-            if (!priorGridGen) {
-                errorMessage = "Only opset6 ExperimentalDetectronPriorGridGenerator operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    std::string errorPrefix;
-
-public:
-    explicit ExperimentalDetectronPriorGridGeneratorImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            errorPrefix = "ExperimentalDetectronPriorGridGenerator layer with name '" + op->get_friendly_name() + "'";
-            const auto priorGridGen = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronPriorGridGenerator>(op);
-            if (op->get_input_size() != 3 || op->get_output_size() != 1)
-                IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
-
-            if (op->get_input_shape(INPUT_PRIORS).size() != 2 ||
-                op->get_input_shape(INPUT_FEATUREMAP).size() != 4 ||
-                    op->get_input_shape(INPUT_IMAGE).size() != 4)
-                IE_THROW() << errorPrefix << " has unsupported input shape";
-
-            const auto &attr = priorGridGen->get_attrs();
-            grid_w_ = attr.w;
-            grid_h_ = attr.h;
-            stride_h_ = attr.stride_y;
-            stride_w_ = attr.stride_x;
-
-            addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                           {TensorDescCreatorTypes::ncsp, Precision::FP32},
-                           {TensorDescCreatorTypes::ncsp, Precision::FP32}},
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs,
-                       ResponseDesc *resp) noexcept override {
-        const int num_priors_ = inputs[INPUT_PRIORS]->getTensorDesc().getDims()[0];
-        assert(inputs[INPUT_PRIORS]->getTensorDesc().getDims()[1] == 4);
-
-        // Execute
-        const int layer_width = grid_w_ ? grid_w_ : inputs[INPUT_FEATUREMAP]->getTensorDesc().getDims()[3];
-        const int layer_height = grid_h_ ? grid_h_ : inputs[INPUT_FEATUREMAP]->getTensorDesc().getDims()[2];
-        const float step_w = stride_w_ ? stride_w_ : static_cast<float>(inputs[INPUT_IMAGE]->getTensorDesc().getDims()[3]) / layer_width;
-        const float step_h = stride_h_ ? stride_h_ : static_cast<float>(inputs[INPUT_IMAGE]->getTensorDesc().getDims()[2]) / layer_height;
-
-        const auto *bottom_data_0 = inputs[0]->buffer().as<const float *>();
-        auto *top_data_0 = outputs[OUTPUT_ROIS]->buffer().as<float *>();
-
-        for (int h = 0; h < layer_height; ++h) {
-            for (int w = 0; w < layer_width; ++w) {
-                for (int s = 0; s < num_priors_; ++s) {
-                    top_data_0[0] = bottom_data_0[4 * s + 0] + step_w * (w + 0.5f);
-                    top_data_0[1] = bottom_data_0[4 * s + 1] + step_h * (h + 0.5f);
-                    top_data_0[2] = bottom_data_0[4 * s + 2] + step_w * (w + 0.5f);
-                    top_data_0[3] = bottom_data_0[4 * s + 3] + step_h * (h + 0.5f);
-                    top_data_0 += 4;
-                }
-            }
-        }
-
-        return OK;
-    }
-
-private:
-    int grid_w_;
-    int grid_h_;
-    float stride_w_;
-    float stride_h_;
-};
-
-
-REG_FACTORY_FOR(ExperimentalDetectronPriorGridGeneratorImpl, ExperimentalDetectronPriorGridGenerator);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/proposal.cpp b/inference-engine/src/mkldnn_plugin/nodes/proposal.cpp
deleted file mode 100644
index 7dd69d7a0071ea..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/proposal.cpp
+++ /dev/null
@@ -1,227 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <string>
-#include <cmath>
-#include <vector>
-
-#include "common/tensor_desc_creator.h"
-#include "proposal_imp.hpp"
-#include <ngraph/op/proposal.hpp>
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-static
-std::vector<float> generate_anchors(proposal_conf &conf) {
-    auto base_size = conf.base_size_;
-    auto coordinates_offset = conf.coordinates_offset;
-    auto round_ratios = conf.round_ratios;
-
-    auto num_ratios = conf.ratios.size();
-    auto ratios = conf.ratios.data();
-
-    auto num_scales = conf.scales.size();
-    auto scales = conf.scales.data();
-
-    std::vector<float> anchors(num_scales * num_ratios * 4);
-    auto anchors_ptr = anchors.data();
-
-    // base box's width & height & center location
-    const float base_area = static_cast<float>(base_size * base_size);
-    const float half_base_size = base_size * 0.5f;
-    const float center = 0.5f * (base_size - coordinates_offset);
-
-    // enumerate all transformed boxes
-    for (int ratio = 0; ratio < num_ratios; ++ratio) {
-        // transformed width & height for given ratio factors
-        float ratio_w;
-        float ratio_h;
-        if (round_ratios) {
-            ratio_w = std::roundf(std::sqrt(base_area / ratios[ratio]));
-            ratio_h = std::roundf(ratio_w * ratios[ratio]);
-        } else {
-            ratio_w = std::sqrt(base_area / ratios[ratio]);
-            ratio_h = ratio_w * ratios[ratio];
-        }
-
-        float * const p_anchors_wm = anchors_ptr + 0 * num_ratios * num_scales + ratio * num_scales;
-        float * const p_anchors_hm = anchors_ptr + 1 * num_ratios * num_scales + ratio * num_scales;
-        float * const p_anchors_wp = anchors_ptr + 2 * num_ratios * num_scales + ratio * num_scales;
-        float * const p_anchors_hp = anchors_ptr + 3 * num_ratios * num_scales + ratio * num_scales;
-
-        for (int scale = 0; scale < num_scales; ++scale) {
-            // transformed width & height for given scale factors
-            const float scale_w = 0.5f * (ratio_w * scales[scale] - coordinates_offset);
-            const float scale_h = 0.5f * (ratio_h * scales[scale] - coordinates_offset);
-
-            // (x1, y1, x2, y2) for transformed box
-            p_anchors_wm[scale] = center - scale_w;
-            p_anchors_hm[scale] = center - scale_h;
-            p_anchors_wp[scale] = center + scale_w;
-            p_anchors_hp[scale] = center + scale_h;
-
-            if (conf.shift_anchors) {
-                p_anchors_wm[scale] -= half_base_size;
-                p_anchors_hm[scale] -= half_base_size;
-                p_anchors_wp[scale] -= half_base_size;
-                p_anchors_hp[scale] -= half_base_size;
-            }
-        }
-    }
-    return anchors;
-}
-
-class ProposalImpl : public ExtLayerBase {
-public:
-    bool isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            auto proposal0Op = ngraph::as_type_ptr<const ngraph::op::v0::Proposal>(op);
-            auto proposal4Op = ngraph::as_type_ptr<const ngraph::op::v4::Proposal>(op);
-            if (!proposal0Op && !proposal4Op) {
-                errorMessage = "Node is not an instance of the Proposal from the operations set v0 or v4.";
-                return false;
-            }
-            auto proposalOp = std::dynamic_pointer_cast<const ngraph::op::v0::Proposal>(op);
-            // [NM] TODO: Enable after fix Issue: 53750
-            // if (proposalOp->get_attrs().framework != "tensorflow" && !proposalOp->get_attrs().framework.empty()) {
-            //     errorMessage = "Unsupported framework attribute: " + proposalOp->get_attrs().framework;
-            //     return false;
-            // }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    explicit ProposalImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            auto proposalOp = std::dynamic_pointer_cast<const ngraph::op::v0::Proposal>(op);
-            auto proposalAttrs = proposalOp->get_attrs();
-
-            conf.feat_stride_ = proposalAttrs.feat_stride;
-            conf.base_size_ = proposalAttrs.base_size;
-            conf.min_size_ = proposalAttrs.min_size;
-            conf.pre_nms_topn_ = proposalAttrs.pre_nms_topn;
-            conf.post_nms_topn_ = proposalAttrs.post_nms_topn;
-            conf.nms_thresh_ = proposalAttrs.nms_thresh;
-            conf.box_coordinate_scale_ = proposalAttrs.box_coordinate_scale;
-            conf.box_size_scale_ = proposalAttrs.box_size_scale;
-            conf.scales = proposalAttrs.scale;
-            conf.ratios = proposalAttrs.ratio;
-            conf.normalize_ = proposalAttrs.normalize;
-            conf.clip_before_nms = proposalAttrs.clip_before_nms;
-            conf.clip_after_nms = proposalAttrs.clip_after_nms;
-            conf.anchors_shape_0 = conf.ratios.size() * conf.scales.size();
-
-            if (proposalAttrs.framework == "tensorflow") {
-                conf.coordinates_offset = 0.0f;
-                conf.initial_clip = true;
-                conf.shift_anchors = true;
-                conf.round_ratios = false;
-                conf.swap_xy = true;
-            } else {
-                conf.coordinates_offset = 1.0f;
-                conf.initial_clip = false;
-                conf.shift_anchors = false;
-                conf.round_ratios = true;
-                conf.swap_xy = false;
-            }
-
-            anchors = generate_anchors(conf);
-            roi_indices.resize(conf.post_nms_topn_);
-
-            store_prob = op->get_output_size() == 2;
-            if (store_prob) {
-                addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                               {TensorDescCreatorTypes::ncsp, Precision::FP32},
-                               {TensorDescCreatorTypes::ncsp, Precision::FP32}},
-                              {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                               {TensorDescCreatorTypes::ncsp, Precision::FP32}});
-            } else {
-                addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                               {TensorDescCreatorTypes::ncsp, Precision::FP32},
-                               {TensorDescCreatorTypes::ncsp, Precision::FP32}},
-                              {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-            }
-        } catch (const InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-            throw;
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr> &inputs, std::vector<Blob::Ptr> &outputs,
-                       ResponseDesc *resp) noexcept override {
-        try {
-            const float* probabilitiesData = inputs[PROBABILITIES_IN_IDX]->cbuffer().as<const float*>() +
-                inputs[PROBABILITIES_IN_IDX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-            const float* anchorsData = inputs[ANCHORS_IN_IDX]->cbuffer().as<const float*>() +
-                inputs[ANCHORS_IN_IDX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-            const float* imgInfoData = inputs[IMG_INFO_IN_IDX]->cbuffer().as<const float*>() +
-                inputs[IMG_INFO_IN_IDX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-            float* outRoiData = outputs[ROI_OUT_IDX]->buffer().as<float*>() +
-                outputs[ROI_OUT_IDX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-            float* outProbData = nullptr;
-            if (store_prob)
-                outProbData = outputs[PROBABILITIES_OUT_IDX]->buffer().as<float*>() +
-                    outputs[PROBABILITIES_OUT_IDX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-            auto inProbDims = inputs[0]->getTensorDesc().getDims();
-            const size_t imgInfoSize = inputs[2]->getTensorDesc().getDims()[0];
-
-            // input image height & width
-            const float imgHeight = imgInfoData[0];
-            const float imgWidth = imgInfoData[1];
-            if (!std::isnormal(imgHeight) || !std::isnormal(imgWidth) || (imgHeight < 0.f) || (imgWidth < 0.f)) {
-                IE_THROW() << "Proposal operation image info input must have positive image height and width.";
-            }
-
-            // scale factor for height & width
-            const float scaleHeight = imgInfoData[2];
-            const float scaleWidth = imgInfoSize == 4 ? imgInfoData[3] : scaleHeight;
-            if (!std::isfinite(scaleHeight) || !std::isfinite(scaleWidth) || (scaleHeight < 0.f) || (scaleWidth < 0.f)) {
-                IE_THROW() << "Proposal operation image info input must have non negative scales.";
-            }
-
-            XARCH::proposal_exec(probabilitiesData, anchorsData, inProbDims,
-                    {imgHeight, imgWidth, scaleHeight, scaleWidth}, anchors.data(), roi_indices.data(), outRoiData, outProbData, conf);
-
-            return OK;
-        } catch (const InferenceEngine::Exception& e) {
-            if (resp) {
-                std::string errorMsg = e.what();
-                errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-            }
-            return GENERAL_ERROR;
-        }
-    }
-
-private:
-    const size_t PROBABILITIES_IN_IDX = 0lu;
-    const size_t ANCHORS_IN_IDX = 1lu;
-    const size_t IMG_INFO_IN_IDX = 2lu;
-    const size_t ROI_OUT_IDX = 0lu;
-    const size_t PROBABILITIES_OUT_IDX = 1lu;
-
-    proposal_conf conf;
-    std::vector<float> anchors;
-    std::vector<int> roi_indices;
-    bool store_prob;  // store blob with proposal probabilities
-};
-
-REG_FACTORY_FOR(ProposalImpl, Proposal);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/proposal_onnx.cpp b/inference-engine/src/mkldnn_plugin/nodes/proposal_onnx.cpp
deleted file mode 100644
index c36f47f68c62ce..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/proposal_onnx.cpp
+++ /dev/null
@@ -1,450 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <cstring>
-#include <cassert>
-#include <cmath>
-#include <string>
-#include <vector>
-#include <utility>
-#include <algorithm>
-#if defined(HAVE_AVX2)
-#include <immintrin.h>
-#endif
-#include "ie_parallel.hpp"
-#include <ngraph/op/experimental_detectron_generate_proposals.hpp>
-#include "common/tensor_desc_creator.h"
-
-
-namespace {
-struct Indexer4d {
-  int dim3_;
-  int dim23_;
-  int dim123_;
-
-  explicit Indexer4d(int dim0, int dim1, int dim2, int dim3):
-      dim3_(dim3), dim23_(dim2 * dim3), dim123_(dim1 * dim2 * dim3) {
-      (void)dim0;
-  }
-
-  int operator()(int i, int j, int k, int n) const {
-      return  i * dim123_ + j * dim23_ + k * dim3_ + n;
-  }
-};
-}  // namespace
-
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-static
-void refine_anchors(const float* deltas, const float* scores, const float* anchors,
-                    float* proposals, const int anchors_num, const int bottom_H,
-                    const int bottom_W, const float img_H, const float img_W,
-                    const float min_box_H, const float min_box_W,
-                    const float max_delta_log_wh,
-                    float coordinates_offset) {
-    Indexer4d delta_idx(anchors_num, 4, bottom_H, bottom_W);
-    Indexer4d score_idx(anchors_num, 1, bottom_H, bottom_W);
-    Indexer4d proposal_idx(bottom_H, bottom_W, anchors_num, 5);
-    Indexer4d anchor_idx(bottom_H, bottom_W, anchors_num, 4);
-
-    parallel_for2d(bottom_H, bottom_W, [&](int h, int w) {
-            for (int anchor = 0; anchor < anchors_num; ++anchor) {
-                int a_idx = anchor_idx(h, w, anchor, 0);
-                float x0 = anchors[a_idx + 0];
-                float y0 = anchors[a_idx + 1];
-                float x1 = anchors[a_idx + 2];
-                float y1 = anchors[a_idx + 3];
-
-                const float dx = deltas[delta_idx(anchor, 0, h, w)];
-                const float dy = deltas[delta_idx(anchor, 1, h, w)];
-                const float d_log_w = deltas[delta_idx(anchor, 2, h, w)];
-                const float d_log_h = deltas[delta_idx(anchor, 3, h, w)];
-
-                const float score = scores[score_idx(anchor, 0, h, w)];
-
-                // width & height of box
-                const float ww = x1 - x0 + coordinates_offset;
-                const float hh = y1 - y0 + coordinates_offset;
-                // center location of box
-                const float ctr_x = x0 + 0.5f * ww;
-                const float ctr_y = y0 + 0.5f * hh;
-
-                // new center location according to deltas (dx, dy)
-                const float pred_ctr_x = dx * ww + ctr_x;
-                const float pred_ctr_y = dy * hh + ctr_y;
-                // new width & height according to deltas d(log w), d(log h)
-                const float pred_w = std::exp(std::min(d_log_w, max_delta_log_wh)) * ww;
-                const float pred_h = std::exp(std::min(d_log_h, max_delta_log_wh)) * hh;
-
-                // update upper-left corner location
-                x0 = pred_ctr_x - 0.5f * pred_w;
-                y0 = pred_ctr_y - 0.5f * pred_h;
-                // update lower-right corner location
-                x1 = pred_ctr_x + 0.5f * pred_w - coordinates_offset;
-                y1 = pred_ctr_y + 0.5f * pred_h - coordinates_offset;
-
-                // adjust new corner locations to be within the image region,
-                x0 = std::max<float>(0.0f, std::min<float>(x0, img_W - coordinates_offset));
-                y0 = std::max<float>(0.0f, std::min<float>(y0, img_H - coordinates_offset));
-                x1 = std::max<float>(0.0f, std::min<float>(x1, img_W - coordinates_offset));
-                y1 = std::max<float>(0.0f, std::min<float>(y1, img_H - coordinates_offset));
-
-                // recompute new width & height
-                const float box_w = x1 - x0 + coordinates_offset;
-                const float box_h = y1 - y0 + coordinates_offset;
-
-                int p_idx = proposal_idx(h, w, anchor, 0);
-                proposals[p_idx + 0] = x0;
-                proposals[p_idx + 1] = y0;
-                proposals[p_idx + 2] = x1;
-                proposals[p_idx + 3] = y1;
-                proposals[p_idx + 4] = (min_box_W <= box_w) * (min_box_H <= box_h) * score;
-            }
-    });
-}
-
-static void unpack_boxes(const float* p_proposals, float* unpacked_boxes, int pre_nms_topn) {
-    parallel_for(pre_nms_topn, [&](size_t i) {
-        unpacked_boxes[0*pre_nms_topn + i] = p_proposals[5*i + 0];
-        unpacked_boxes[1*pre_nms_topn + i] = p_proposals[5*i + 1];
-        unpacked_boxes[2*pre_nms_topn + i] = p_proposals[5*i + 2];
-        unpacked_boxes[3*pre_nms_topn + i] = p_proposals[5*i + 3];
-        unpacked_boxes[4*pre_nms_topn + i] = p_proposals[5*i + 4];
-    });
-}
-
-static
-void nms_cpu(const int num_boxes, int is_dead[],
-             const float* boxes, int index_out[], int* const num_out,
-             const int base_index, const float nms_thresh, const int max_num_out,
-             float coordinates_offset) {
-    const int num_proposals = num_boxes;
-    int count = 0;
-
-    const float* x0 = boxes + 0 * num_proposals;
-    const float* y0 = boxes + 1 * num_proposals;
-    const float* x1 = boxes + 2 * num_proposals;
-    const float* y1 = boxes + 3 * num_proposals;
-
-    std::memset(is_dead, 0, num_boxes * sizeof(int));
-
-#if defined(HAVE_AVX2)
-    __m256  vc_fone = _mm256_set1_ps(coordinates_offset);
-    __m256i vc_ione = _mm256_set1_epi32(1);
-    __m256  vc_zero = _mm256_set1_ps(0.0f);
-
-    __m256 vc_nms_thresh = _mm256_set1_ps(nms_thresh);
-#endif
-
-    for (int box = 0; box < num_boxes; ++box) {
-        if (is_dead[box])
-            continue;
-
-        index_out[count++] = base_index + box;
-        if (count == max_num_out)
-            break;
-
-        int tail = box + 1;
-
-#if defined(HAVE_AVX2)
-        __m256 vx0i = _mm256_set1_ps(x0[box]);
-        __m256 vy0i = _mm256_set1_ps(y0[box]);
-        __m256 vx1i = _mm256_set1_ps(x1[box]);
-        __m256 vy1i = _mm256_set1_ps(y1[box]);
-
-        __m256 vA_width  = _mm256_sub_ps(vx1i, vx0i);
-        __m256 vA_height = _mm256_sub_ps(vy1i, vy0i);
-        __m256 vA_area   = _mm256_mul_ps(_mm256_add_ps(vA_width, vc_fone), _mm256_add_ps(vA_height, vc_fone));
-
-        for (; tail <= num_boxes - 8; tail += 8) {
-            __m256i *pdst = reinterpret_cast<__m256i*>(is_dead + tail);
-            __m256i  vdst = _mm256_loadu_si256(pdst);
-
-            __m256 vx0j = _mm256_loadu_ps(x0 + tail);
-            __m256 vy0j = _mm256_loadu_ps(y0 + tail);
-            __m256 vx1j = _mm256_loadu_ps(x1 + tail);
-            __m256 vy1j = _mm256_loadu_ps(y1 + tail);
-
-            __m256 vx0 = _mm256_max_ps(vx0i, vx0j);
-            __m256 vy0 = _mm256_max_ps(vy0i, vy0j);
-            __m256 vx1 = _mm256_min_ps(vx1i, vx1j);
-            __m256 vy1 = _mm256_min_ps(vy1i, vy1j);
-
-            __m256 vwidth  = _mm256_add_ps(_mm256_sub_ps(vx1, vx0), vc_fone);
-            __m256 vheight = _mm256_add_ps(_mm256_sub_ps(vy1, vy0), vc_fone);
-            __m256 varea = _mm256_mul_ps(_mm256_max_ps(vc_zero, vwidth), _mm256_max_ps(vc_zero, vheight));
-
-            __m256 vB_width  = _mm256_sub_ps(vx1j, vx0j);
-            __m256 vB_height = _mm256_sub_ps(vy1j, vy0j);
-            __m256 vB_area   = _mm256_mul_ps(_mm256_add_ps(vB_width, vc_fone), _mm256_add_ps(vB_height, vc_fone));
-
-            __m256 vdivisor = _mm256_sub_ps(_mm256_add_ps(vA_area, vB_area), varea);
-            __m256 vintersection_area = _mm256_div_ps(varea, vdivisor);
-
-            __m256 vcmp_0 = _mm256_cmp_ps(vx0i, vx1j, _CMP_LE_OS);
-            __m256 vcmp_1 = _mm256_cmp_ps(vy0i, vy1j, _CMP_LE_OS);
-            __m256 vcmp_2 = _mm256_cmp_ps(vx0j, vx1i, _CMP_LE_OS);
-            __m256 vcmp_3 = _mm256_cmp_ps(vy0j, vy1i, _CMP_LE_OS);
-            __m256 vcmp_4 = _mm256_cmp_ps(vc_nms_thresh, vintersection_area, _CMP_LT_OS);
-
-            vcmp_0 = _mm256_and_ps(vcmp_0, vcmp_1);
-            vcmp_2 = _mm256_and_ps(vcmp_2, vcmp_3);
-            vcmp_4 = _mm256_and_ps(vcmp_4, vcmp_0);
-            vcmp_4 = _mm256_and_ps(vcmp_4, vcmp_2);
-
-            _mm256_storeu_si256(pdst, _mm256_blendv_epi8(vdst, vc_ione, _mm256_castps_si256(vcmp_4)));
-        }
-#endif
-
-        for (; tail < num_boxes; ++tail) {
-            float res = 0.0f;
-
-            const float x0i = x0[box];
-            const float y0i = y0[box];
-            const float x1i = x1[box];
-            const float y1i = y1[box];
-
-            const float x0j = x0[tail];
-            const float y0j = y0[tail];
-            const float x1j = x1[tail];
-            const float y1j = y1[tail];
-
-            if (x0i <= x1j && y0i <= y1j && x0j <= x1i && y0j <= y1i) {
-                // overlapped region (= box)
-                const float x0 = std::max<float>(x0i, x0j);
-                const float y0 = std::max<float>(y0i, y0j);
-                const float x1 = std::min<float>(x1i, x1j);
-                const float y1 = std::min<float>(y1i, y1j);
-
-                // intersection area
-                const float width  = std::max<float>(0.0f,  x1 - x0 + coordinates_offset);
-                const float height = std::max<float>(0.0f,  y1 - y0 + coordinates_offset);
-                const float area   = width * height;
-
-                // area of A, B
-                const float A_area = (x1i - x0i + coordinates_offset) * (y1i - y0i + coordinates_offset);
-                const float B_area = (x1j - x0j + coordinates_offset) * (y1j - y0j + coordinates_offset);
-
-                // IoU
-                res = area / (A_area + B_area - area);
-            }
-
-            if (nms_thresh < res)
-                is_dead[tail] = 1;
-        }
-    }
-
-    *num_out = count;
-}
-
-
-static
-void fill_output_blobs(const float* proposals, const int* roi_indices,
-                       float* rois, float* scores,
-                       const int num_proposals, const int num_rois, const int post_nms_topn) {
-    const float *src_x0 = proposals + 0 * num_proposals;
-    const float *src_y0 = proposals + 1 * num_proposals;
-    const float *src_x1 = proposals + 2 * num_proposals;
-    const float *src_y1 = proposals + 3 * num_proposals;
-    const float *src_score = proposals + 4 * num_proposals;
-
-    parallel_for(num_rois, [&](size_t i) {
-        int index = roi_indices[i];
-        rois[i * 4 + 0] = src_x0[index];
-        rois[i * 4 + 1] = src_y0[index];
-        rois[i * 4 + 2] = src_x1[index];
-        rois[i * 4 + 3] = src_y1[index];
-        scores[i] = src_score[index];
-    });
-
-    if (num_rois < post_nms_topn) {
-        for (int i = 4 * num_rois; i < 4 * post_nms_topn; i++) {
-            rois[i] = 0.f;
-        }
-        for (int i = num_rois; i < post_nms_topn; i++) {
-            scores[i] = 0.f;
-        }
-    }
-}
-
-
-class ExperimentalDetectronGenerateProposalsSingleImageImpl : public ExtLayerBase {
-private:
-    const int INPUT_IM_INFO {0};
-    const int INPUT_ANCHORS {1};
-    const int INPUT_DELTAS {2};
-    const int INPUT_SCORES {3};
-    const int OUTPUT_ROIS {0};
-    const int OUTPUT_SCORES {1};
-
-public:
-    bool isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            auto proposalOp = ngraph::as_type_ptr<const ngraph::op::v6::ExperimentalDetectronGenerateProposalsSingleImage>(op);
-            if (!proposalOp) {
-                errorMessage = "Node is not an instance of the Proposal from the operations set v0.";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    explicit ExperimentalDetectronGenerateProposalsSingleImageImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            auto proposalOp = ngraph::as_type_ptr<const ngraph::op::v6::ExperimentalDetectronGenerateProposalsSingleImage>(op);
-            auto proposalAttrs = proposalOp->get_attrs();
-
-            min_size_ = proposalAttrs.min_size;
-            nms_thresh_ = proposalAttrs.nms_threshold;
-            pre_nms_topn_ = proposalAttrs.pre_nms_count;
-            post_nms_topn_ = proposalAttrs.post_nms_count;
-
-            coordinates_offset = 0.0f;
-
-            roi_indices_.resize(post_nms_topn_);
-            addConfig(op,
-                      {{TensorDescCreatorTypes::ncsp, Precision::FP32}, {TensorDescCreatorTypes::ncsp, Precision::FP32},
-                       {TensorDescCreatorTypes::ncsp, Precision::FP32}, {TensorDescCreatorTypes::ncsp, Precision::FP32}},
-                      {{TensorDescCreatorTypes::ncsp, Precision::FP32}, {TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-            throw;
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr> &inputs, std::vector<Blob::Ptr> &outputs,
-                       ResponseDesc *resp) noexcept override {
-        try {
-            if (inputs.size() != 4 || outputs.size() != 2) {
-                IE_THROW() << "Incorrect number of input or output edges!";
-            }
-
-            size_t anchor_dims_size = 1;
-            for (size_t i = 0; i < inputs[INPUT_ANCHORS]->getTensorDesc().getDims().size(); i++) {
-                anchor_dims_size *= inputs[INPUT_ANCHORS]->getTensorDesc().getDims()[i];
-            }
-
-            size_t deltas_dims_size = 1;
-            for (size_t i = 0; i < inputs[INPUT_DELTAS]->getTensorDesc().getDims().size(); i++) {
-                deltas_dims_size *= inputs[INPUT_DELTAS]->getTensorDesc().getDims()[i];
-            }
-            if (anchor_dims_size != deltas_dims_size)
-                IE_THROW() << "'Anchors' blob size for ONNXProposal is incompatible with 'deltas' blob size!";
-
-            size_t score_dims_size = 1;
-            for (size_t i = 0; i < inputs[INPUT_SCORES]->getTensorDesc().getDims().size(); i++) {
-                score_dims_size *= inputs[INPUT_SCORES]->getTensorDesc().getDims()[i];
-            }
-            if (deltas_dims_size != (4 * score_dims_size))
-                IE_THROW() << "'Deltas' blob size for ONNXProposal is incompatible with 'scores' blob size!";
-
-            // Prepare memory
-            const float* p_deltas_item = inputs[INPUT_DELTAS]->buffer();
-            const float* p_scores_item = inputs[INPUT_SCORES]->buffer();
-            const float* p_anchors_item = inputs[INPUT_ANCHORS]->buffer();
-            const float* p_img_info_cpu = inputs[INPUT_IM_INFO]->buffer();
-
-            float* p_roi_item = outputs[OUTPUT_ROIS]->buffer();
-            float* p_roi_score_item = outputs[OUTPUT_SCORES]->buffer();
-
-            const int anchors_num = inputs[INPUT_SCORES]->getTensorDesc().getDims()[0];
-
-            // bottom shape: (num_anchors) x H x W
-            const int bottom_H = inputs[INPUT_DELTAS]->getTensorDesc().getDims()[1];
-            const int bottom_W = inputs[INPUT_DELTAS]->getTensorDesc().getDims()[2];
-
-            // input image height & width
-            const float img_H = p_img_info_cpu[0];
-            const float img_W = p_img_info_cpu[1];
-
-            // scale factor for height & width
-
-            // minimum box width & height
-            const float min_box_H = min_size_;
-            const float min_box_W = min_size_;
-
-            // number of all proposals = num_anchors * H * W
-            const int num_proposals = anchors_num * bottom_H * bottom_W;
-
-            // number of top-n proposals before NMS
-            const int pre_nms_topn = std::min<int>(num_proposals, pre_nms_topn_);
-
-            // number of final RoIs
-            int num_rois = 0;
-
-            // enumerate all proposals
-            //   num_proposals = num_anchors * H * W
-            //   (x1, y1, x2, y2, score) for each proposal
-            // NOTE: for bottom, only foreground scores are passed
-            struct ProposalBox {
-                float x0;
-                float y0;
-                float x1;
-                float y1;
-                float score;
-            };
-            std::vector<ProposalBox> proposals_(num_proposals);
-            std::vector<float> unpacked_boxes(5 * pre_nms_topn);
-            std::vector<int> is_dead(pre_nms_topn);
-
-            // Execute
-            int batch_size = 1;  // inputs[INPUT_DELTAS]->getTensorDesc().getDims()[0];
-            for (int n = 0; n < batch_size; ++n) {
-                refine_anchors(p_deltas_item, p_scores_item, p_anchors_item,
-                               reinterpret_cast<float *>(&proposals_[0]), anchors_num, bottom_H,
-                               bottom_W, img_H, img_W,
-                               min_box_H, min_box_W,
-                               static_cast<const float>(log(1000. / 16.)),
-                               1.0f);
-                std::partial_sort(proposals_.begin(), proposals_.begin() + pre_nms_topn, proposals_.end(),
-                                  [](const ProposalBox& struct1, const ProposalBox& struct2) {
-                                      return (struct1.score > struct2.score);
-                                  });
-
-                unpack_boxes(reinterpret_cast<float *>(&proposals_[0]), &unpacked_boxes[0], pre_nms_topn);
-                nms_cpu(pre_nms_topn, &is_dead[0], &unpacked_boxes[0], &roi_indices_[0], &num_rois, 0,
-                        nms_thresh_, post_nms_topn_, coordinates_offset);
-                fill_output_blobs(&unpacked_boxes[0], &roi_indices_[0], p_roi_item, p_roi_score_item,
-                                  pre_nms_topn, num_rois, post_nms_topn_);
-            }
-
-            return OK;
-        } catch (const std::exception& e) {
-            if (resp) {
-                std::string errorMsg = e.what();
-                errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-            }
-            return GENERAL_ERROR;
-        }
-    }
-
-private:
-    float min_size_;
-    int pre_nms_topn_;
-    int post_nms_topn_;
-    float nms_thresh_;
-    float coordinates_offset;
-
-    std::vector<int> roi_indices_;
-};
-
-REG_FACTORY_FOR(ExperimentalDetectronGenerateProposalsSingleImageImpl, ExperimentalDetectronGenerateProposalsSingleImage);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/range.cpp b/inference-engine/src/mkldnn_plugin/nodes/range.cpp
deleted file mode 100644
index 30de35f6c72781..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/range.cpp
+++ /dev/null
@@ -1,164 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <cmath>
-#include <string>
-#include <vector>
-#include <cassert>
-#include "ie_parallel.hpp"
-#include <ngraph/opsets/opset1.hpp>
-#include <utils/general_utils.h>
-
-using namespace MKLDNNPlugin;
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-
-
-class RangeImpl: public ExtLayerBase {
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            if (!MKLDNNPlugin::one_of(op->get_type_info(), ngraph::op::v0::Range::type_info, ngraph::op::v4::Range::type_info)) {
-                errorMessage = "Only opset1 and opset4 Range operation is supported";
-                return false;
-            }
-            if (std::dynamic_pointer_cast<const ngraph::opset1::Constant>(op->get_input_node_shared_ptr(RANGE_START)) == nullptr ||
-                std::dynamic_pointer_cast<const ngraph::opset1::Constant>(op->get_input_node_shared_ptr(RANGE_LIMIT)) == nullptr ||
-                    std::dynamic_pointer_cast<const ngraph::opset1::Constant>(op->get_input_node_shared_ptr(RANGE_DELTA)) == nullptr) {
-                errorMessage = "Only const inputs for Range operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    std::string errorPrefix;
-
-public:
-    explicit RangeImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            errorPrefix = "Range layer with name '" + op->get_friendly_name() + "'";
-
-            if (op->get_input_size() != 3 || op->get_output_size() != 1)
-                IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
-
-            SizeVector start_dims = op->get_input_shape(RANGE_START);
-            if (ngraph::shape_size(start_dims) != 1)
-                IE_THROW() << errorPrefix << " has start scalar with more than 1 value";
-
-            SizeVector limit_dims = op->get_input_shape(RANGE_LIMIT);
-            if (ngraph::shape_size(limit_dims) != 1)
-                IE_THROW() << errorPrefix << " has limit scalar with more than 1 value";
-
-            SizeVector delta_dims = op->get_input_shape(RANGE_DELTA);
-            if (ngraph::shape_size(delta_dims) != 1)
-                IE_THROW() << errorPrefix << " has delta scalar with more than 1 value";
-
-            SizeVector dst_dims = op->get_output_shape(0);
-            if (dst_dims.size() > 1)
-                IE_THROW() << errorPrefix << " has unsupported rank for output: " << dst_dims.size();
-
-            if (!(details::convertPrecision(op->get_input_element_type(RANGE_START)) == Precision::I32 &&
-                  details::convertPrecision(op->get_input_element_type(RANGE_LIMIT)) == Precision::I32 &&
-                  details::convertPrecision(op->get_input_element_type(RANGE_DELTA)) == Precision::I32 &&
-                  details::convertPrecision(op->get_output_element_type(0)) == Precision::I32) &&
-                !(details::convertPrecision(op->get_input_element_type(RANGE_START)) == Precision::FP32 &&
-                  details::convertPrecision(op->get_input_element_type(RANGE_LIMIT)) == Precision::FP32 &&
-                  details::convertPrecision(op->get_input_element_type(RANGE_DELTA)) == Precision::FP32 &&
-                  details::convertPrecision(op->get_output_element_type(0)) == Precision::FP32)) {
-                      addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                                     {TensorDescCreatorTypes::ncsp, Precision::FP32},
-                                     {TensorDescCreatorTypes::ncsp, Precision::FP32}},
-                                    {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-            } else {
-                addConfig(op, {{TensorDescCreatorTypes::ncsp},
-                               {TensorDescCreatorTypes::ncsp},
-                               {TensorDescCreatorTypes::ncsp}},
-                              {{TensorDescCreatorTypes::ncsp}});
-            }
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs, ResponseDesc *resp) noexcept override {
-        StatusCode retcode = OK;
-        switch (outputs[0]->getTensorDesc().getPrecision()) {
-        case Precision::FP32: {
-            retcode = range((inputs[RANGE_START]->cbuffer().as<float *>() +
-                             inputs[RANGE_START]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0],
-                            (inputs[RANGE_LIMIT]->cbuffer().as<float *>() +
-                             inputs[RANGE_LIMIT]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0],
-                            (inputs[RANGE_DELTA]->cbuffer().as<float *>() +
-                             inputs[RANGE_DELTA]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0], outputs[0]);
-        }
-        break;
-        case Precision::I32: {
-            retcode = range((inputs[RANGE_START]->cbuffer().as<int32_t *>() +
-                             inputs[RANGE_START]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0],
-                            (inputs[RANGE_LIMIT]->cbuffer().as<int32_t *>() +
-                             inputs[RANGE_LIMIT]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0],
-                            (inputs[RANGE_DELTA]->cbuffer().as<int32_t *>() +
-                             inputs[RANGE_DELTA]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0], outputs[0]);
-        }
-        break;
-        default:
-            if (resp) {
-                std::string errorMsg = "Incorrect output precision. Only FP32 and I32 are supported!";
-                errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-            }
-            retcode = GENERAL_ERROR;
-        }
-        if (resp && retcode == PARAMETER_MISMATCH) {
-            std::string errorMsg = "Range indexes exceeds data tensor dimension";
-            errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-        }
-        return retcode;
-    }
-
-private:
-    static const size_t RANGE_START = 0;
-    static const size_t RANGE_LIMIT = 1;
-    static const size_t RANGE_DELTA = 2;
-
-    template <typename data_t>
-    StatusCode range(data_t start, data_t limit, data_t delta, Blob::Ptr output);
-};
-
-template <typename data_t>
-StatusCode RangeImpl::range(data_t start, data_t limit, data_t delta, Blob::Ptr output) {
-    size_t dst_size = (output->getTensorDesc().getDims())[0];
-    data_t* dst_data = output->cbuffer().as<data_t *>() +
-                       output->getTensorDesc().getBlockingDesc().getOffsetPadding();
-    size_t work_amount_dst = static_cast<size_t>(std::floor(std::abs((limit - start) / delta)));
-    if (work_amount_dst != dst_size)
-        return PARAMETER_MISMATCH;
-
-    parallel_nt(0, [&](const int ithr, const int nthr) {
-        size_t iwork = 0, end = 0;
-        splitter(work_amount_dst, nthr, ithr, iwork, end);
-        data_t dst_value = start + iwork * delta;
-
-        for (; iwork < end; ++iwork, dst_value += delta) {
-            dst_data[iwork] = dst_value;
-        }
-    });
-    return OK;
-}
-REG_FACTORY_FOR(RangeImpl, Range);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/reorg_yolo.cpp b/inference-engine/src/mkldnn_plugin/nodes/reorg_yolo.cpp
deleted file mode 100644
index ff705fc63b03ad..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/reorg_yolo.cpp
+++ /dev/null
@@ -1,99 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-#include <vector>
-#include <ngraph/opsets/opset2.hpp>
-
-using namespace MKLDNNPlugin;
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-class ReorgYoloImpl: public ExtLayerBase {
-public:
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            const auto reorgYolo = std::dynamic_pointer_cast<const ngraph::opset2::ReorgYolo>(op);
-            if (!reorgYolo) {
-                errorMessage = "Only opset2 ReorgYolo operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    explicit ReorgYoloImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            errorPrefix = std::string(op->get_type_name()) + " node with name '" + op->get_friendly_name() + "'";
-            if (op->get_input_size() != 1 || op->get_output_size() != 1)
-                IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
-
-            const auto reorgYolo = std::dynamic_pointer_cast<const ngraph::opset2::ReorgYolo>(op);
-            const auto strides = reorgYolo->get_strides();
-            if (strides.empty())
-                IE_THROW() << errorPrefix << " has empty strides";
-            stride = strides[0];
-
-            addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32}},
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs,
-                       ResponseDesc *resp) noexcept override {
-        const auto *src_data = inputs[0]->cbuffer().as<const float *>();
-        auto *dst_data = outputs[0]->buffer().as<float *>();
-
-        int IW = (inputs[0]->getTensorDesc().getDims().size() > 3) ? inputs[0]->getTensorDesc().getDims()[3] : 1;
-        int IH = (inputs[0]->getTensorDesc().getDims().size() > 2) ? inputs[0]->getTensorDesc().getDims()[2] : 1;
-        int IC = (inputs[0]->getTensorDesc().getDims().size() > 1) ? inputs[0]->getTensorDesc().getDims()[1] : 1;
-        int B = (inputs[0]->getTensorDesc().getDims().size() > 0) ? inputs[0]->getTensorDesc().getDims()[0] : 1;
-
-        int ic_off = IC / (stride * stride);
-        int ih_off = IH * stride;
-        int iw_off = IW * stride;
-        for (int b = 0; b < B; b++) {
-            for (int ic = 0; ic < IC; ic++) {
-                for (int ih = 0; ih < IH; ih++) {
-                    for (int iw = 0; iw < IW; iw++) {
-                        int dstIndex = b * IC * IH * IW + ic * IH * IW + ih * IW + iw;
-
-                        int oc = ic % ic_off;
-                        int offset = ic / ic_off;
-
-                        int ow = iw * stride + offset % stride;
-                        int oh = ih * stride + offset / stride;
-
-                        int srcIndex = b * ic_off * ih_off * iw_off + oc * ih_off * iw_off + oh * iw_off + ow;
-
-                        dst_data[dstIndex] = src_data[srcIndex];
-                    }
-                }
-            }
-        }
-        return OK;
-    }
-
-private:
-    int stride;
-
-    std::string errorPrefix;
-};
-
-REG_FACTORY_FOR(ReorgYoloImpl, ReorgYolo);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/reverse_sequence.cpp b/inference-engine/src/mkldnn_plugin/nodes/reverse_sequence.cpp
deleted file mode 100644
index 85ab2b8c414b5c..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/reverse_sequence.cpp
+++ /dev/null
@@ -1,209 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <cmath>
-#include <string>
-#include <vector>
-#include <cassert>
-#include <algorithm>
-#include "ie_parallel.hpp"
-#include <ngraph/opsets/opset1.hpp>
-
-using namespace MKLDNNPlugin;
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-class ReverseSequenceImpl: public ExtLayerBase {
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            const auto revSeq = std::dynamic_pointer_cast<const ngraph::opset1::ReverseSequence>(op);
-            if (!revSeq) {
-                errorMessage = "Only opset1 ReverseSequence operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    std::string errorPrefix;
-
-public:
-    explicit ReverseSequenceImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            errorPrefix = "ReverseSequence layer with name '" + op->get_friendly_name() + "'";
-            const auto revSeq = std::dynamic_pointer_cast<const ngraph::opset1::ReverseSequence>(op);
-
-            if (op->get_input_size() != 2 || op->get_output_size() != 1)
-                IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
-
-            src_dims = op->get_input_shape(REVERSESEQUENCE_DATA);
-
-            Precision lengthsPrecision = details::convertPrecision(op->get_input_element_type(REVERSESEQUENCE_LENGTHS));
-            if (lengthsPrecision != Precision::I32 && lengthsPrecision != Precision::FP32)
-                lengthsPrecision = Precision::I32;
-
-            SizeVector seq_lengths_dims = op->get_input_shape(REVERSESEQUENCE_LENGTHS);
-            if (seq_lengths_dims.size() != 1)
-                IE_THROW() << errorPrefix << " has incorrect 2nd input rank: " << seq_lengths_dims.size();
-
-            SizeVector dst_dims = op->get_output_shape(0);
-            if (src_dims.size() != dst_dims.size())
-                IE_THROW() << errorPrefix << " has incorrect number of input/output sizes!";
-
-            for (size_t i = 0; i < dst_dims.size(); i++) {
-                if (src_dims[i] != dst_dims[i])
-                    IE_THROW() << errorPrefix << " has incorrect number of input/output dimension!";
-            }
-
-            seq_axis = revSeq->get_sequence_axis();
-
-            if (seq_axis < 0 || seq_axis >= static_cast<int>(src_dims.size()))
-                IE_THROW() << errorPrefix << " has incorrect 'seq_axis' parameters dimensions and axis number!";
-
-            batch_axis = revSeq->get_batch_axis();
-
-            if (batch_axis < 0 || batch_axis >= static_cast<int>(src_dims.size()))
-                IE_THROW() << errorPrefix << " has incorrect 'batch_axis' parameters dimensions and axis number!";
-
-            if (seq_lengths_dims[0] != dst_dims[batch_axis])
-                IE_THROW() << errorPrefix << " has incorrect 'seq_lengths_dims' parameters dimension!";
-
-            srcStrides.resize(src_dims.size());
-            srcStrides[srcStrides.size() - 1] = 1;
-            for (int i = srcStrides.size() - 2; i >= 0; i--) {
-                srcStrides[i] = srcStrides[i + 1] * src_dims[i + 1];
-            }
-
-            work_amount_dst = srcStrides[0] * src_dims[0];
-
-            addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                           {TensorDescCreatorTypes::ncsp, lengthsPrecision}},
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs, ResponseDesc *resp) noexcept override {
-        size_t i;
-        const float *src_data = inputs[REVERSESEQUENCE_DATA]->cbuffer().as<const float *>() +
-                                inputs[REVERSESEQUENCE_DATA]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        float* dst_data = outputs[0]->cbuffer().as<float *>() +
-                          outputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-
-        switch (inputs[REVERSESEQUENCE_LENGTHS]->getTensorDesc().getPrecision()) {
-            case Precision::FP32: {
-                float *seq_lengths_data = inputs[REVERSESEQUENCE_LENGTHS]->cbuffer().as<float *>() +
-                                          inputs[REVERSESEQUENCE_LENGTHS]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-                for (i = 0; i < src_dims[batch_axis]; i++) {
-                    if (static_cast<int32_t>(seq_lengths_data[i]) > static_cast<int>(src_dims[seq_axis])) {
-                        if (resp) {
-                            std::string errorMsg = "Incorrect input 'seq_lengths' values!";
-                            errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-                        }
-                        return PARAMETER_MISMATCH;
-                    }
-                }
-
-                parallel_nt(0, [&](const int ithr, const int nthr) {
-                    size_t i, start = 0, end = 0, src_idx = 0;
-                    SizeVector counters(src_dims.size(), 0);
-                    splitter(work_amount_dst, nthr, ithr, start, end);
-                    for (int j = src_dims.size() - 1, i = start; j >= 0; j--) {
-                        counters[j] = i % src_dims[j];
-                        i /= src_dims[j];
-                    }
-
-                    for (size_t iwork = start; iwork < end; ++iwork) {
-                        for (i = 0, src_idx = 0; i < src_dims.size(); ++i) {
-                            size_t idx = counters[i];
-                            if (static_cast<int>(i) == seq_axis &&
-                                    static_cast<int>(idx) < static_cast<int32_t>(seq_lengths_data[counters[batch_axis]])) {
-                                idx = static_cast<int32_t>(seq_lengths_data[counters[batch_axis]]) - idx - 1;
-                            }
-                            src_idx += idx * srcStrides[i];
-                        }
-                        dst_data[iwork] = src_data[src_idx];
-                        for (int j = src_dims.size() - 1; j >= 0; j--) {
-                            counters[j] = (counters[j] + 1) % src_dims[j];
-                            if (counters[j] != 0) break;
-                        }
-                    }
-                });
-            }
-            break;
-            case Precision::I32: {
-                int32_t *seq_lengths_data = inputs[REVERSESEQUENCE_LENGTHS]->cbuffer().as<int32_t *>() +
-                                            inputs[REVERSESEQUENCE_LENGTHS]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-                for (i = 0; i < src_dims[batch_axis]; i++) {
-                    if (seq_lengths_data[i] > static_cast<int>(src_dims[seq_axis])) {
-                        if (resp) {
-                            std::string errorMsg = "Incorrect input 'seq_lengths' values!";
-                            errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-                        }
-                        return PARAMETER_MISMATCH;
-                    }
-                }
-
-                parallel_nt(0, [&](const int ithr, const int nthr) {
-                    size_t i, start = 0, end = 0, src_idx = 0;
-                    SizeVector counters(src_dims.size(), 0);
-                    splitter(work_amount_dst, nthr, ithr, start, end);
-                    for (int j = src_dims.size() - 1, i = start; j >= 0; j--) {
-                        counters[j] = i % src_dims[j];
-                        i /= src_dims[j];
-                    }
-
-                    for (size_t iwork = start; iwork < end; ++iwork) {
-                        for (i = 0, src_idx = 0; i < src_dims.size(); ++i) {
-                            size_t idx = counters[i];
-                            if (static_cast<int>(i) == seq_axis &&
-                                    static_cast<int>(idx) < seq_lengths_data[counters[batch_axis]]) {
-                                idx = seq_lengths_data[counters[batch_axis]] - idx - 1;
-                            }
-                            src_idx += idx * srcStrides[i];
-                        }
-                        dst_data[iwork] = src_data[src_idx];
-                        for (int j = src_dims.size() - 1; j >= 0; j--) {
-                            counters[j] = (counters[j] + 1) % src_dims[j];
-                            if (counters[j] != 0) break;
-                        }
-                    }
-                });
-            }
-            break;
-            default:
-                return GENERAL_ERROR;
-        }
-
-        return OK;
-    }
-
-private:
-    const size_t REVERSESEQUENCE_DATA = 0;
-    const size_t REVERSESEQUENCE_LENGTHS = 1;
-
-    int seq_axis;
-    int batch_axis;
-    SizeVector src_dims;
-    SizeVector srcStrides;
-    size_t work_amount_dst;
-};
-
-REG_FACTORY_FOR(ReverseSequenceImpl, ReverseSequence);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/roifeatureextractor_onnx.cpp b/inference-engine/src/mkldnn_plugin/nodes/roifeatureextractor_onnx.cpp
deleted file mode 100644
index 102a9bbd6cbbfa..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/roifeatureextractor_onnx.cpp
+++ /dev/null
@@ -1,433 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-// There are some code snippets in this file.
-// Original source file is avaialble here (Copyright (c) 2018 Facebook, MIT License):
-// https://github.com/facebookresearch/maskrcnn-benchmark/blob/master/maskrcnn_benchmark/csrc/cpu/ROIAlign_cpu.cpp
-//
-
-#include "base.hpp"
-#include <cassert>
-#include <cmath>
-#include <vector>
-#include <string>
-#include <algorithm>
-#include "ie_parallel.hpp"
-#include "common/cpu_memcpy.h"
-#include <ngraph/opsets/opset6.hpp>
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-// implementation taken from Caffe2
-template <typename T>
-struct PreCalc {
-  int pos1;
-  int pos2;
-  int pos3;
-  int pos4;
-  T w1;
-  T w2;
-  T w3;
-  T w4;
-};
-
-template <typename T>
-void pre_calc_for_bilinear_interpolate(
-    const int height,
-    const int width,
-    const int pooled_height,
-    const int pooled_width,
-    const int iy_upper,
-    const int ix_upper,
-    T roi_start_h,
-    T roi_start_w,
-    T bin_size_h,
-    T bin_size_w,
-    int roi_bin_grid_h,
-    int roi_bin_grid_w,
-    std::vector<PreCalc<T>>& pre_calc) {
-  int pre_calc_index = 0;
-  for (int ph = 0; ph < pooled_height; ph++) {
-    for (int pw = 0; pw < pooled_width; pw++) {
-      for (int iy = 0; iy < iy_upper; iy++) {
-        const T yy = roi_start_h + ph * bin_size_h +
-            static_cast<T>(iy + .5f) * bin_size_h /
-                static_cast<T>(roi_bin_grid_h);  // e.g., 0.5, 1.5
-        for (int ix = 0; ix < ix_upper; ix++) {
-          const T xx = roi_start_w + pw * bin_size_w +
-              static_cast<T>(ix + .5f) * bin_size_w /
-                  static_cast<T>(roi_bin_grid_w);
-
-          T x = xx;
-          T y = yy;
-          // deal with: inverse elements are out of feature map boundary
-          if (y < -1.0 || y > height || x < -1.0 || x > width) {
-            // empty
-            PreCalc<T> pc;
-            pc.pos1 = 0;
-            pc.pos2 = 0;
-            pc.pos3 = 0;
-            pc.pos4 = 0;
-            pc.w1 = 0;
-            pc.w2 = 0;
-            pc.w3 = 0;
-            pc.w4 = 0;
-            pre_calc.at(pre_calc_index) = pc;
-            pre_calc_index += 1;
-            continue;
-          }
-
-          if (y <= 0) {
-            y = 0;
-          }
-          if (x <= 0) {
-            x = 0;
-          }
-
-          int y_low = static_cast<int>(y);
-          int x_low = static_cast<int>(x);
-          int y_high = 0;
-          int x_high = 0;
-
-          if (y_low >= height - 1) {
-            y_high = y_low = height - 1;
-            y = (T)y_low;
-          } else {
-            y_high = y_low + 1;
-          }
-
-          if (x_low >= width - 1) {
-            x_high = x_low = width - 1;
-            x = (T)x_low;
-          } else {
-            x_high = x_low + 1;
-          }
-
-          T ly = y - y_low;
-          T lx = x - x_low;
-          T hy = static_cast<T>(1) - ly, hx = static_cast<T>(1) - lx;
-          T w1 = hy * hx, w2 = hy * lx, w3 = ly * hx, w4 = ly * lx;
-
-          // save weights and indeces
-          PreCalc<T> pc;
-          pc.pos1 = y_low * width + x_low;
-          pc.pos2 = y_low * width + x_high;
-          pc.pos3 = y_high * width + x_low;
-          pc.pos4 = y_high * width + x_high;
-          pc.w1 = w1;
-          pc.w2 = w2;
-          pc.w3 = w3;
-          pc.w4 = w4;
-          pre_calc[pre_calc_index] = pc;
-
-          pre_calc_index += 1;
-        }
-      }
-    }
-  }
-}
-
-template <typename T>
-void ROIAlignForward_cpu_kernel(
-    const int nthreads,
-    const T* bottom_data,
-    const T& spatial_scale,
-    const int channels,
-    const int height,
-    const int width,
-    const int pooled_height,
-    const int pooled_width,
-    const int sampling_ratio,
-    const T* bottom_rois,
-    const bool aligned,
-    T* top_data) {
-  int roi_cols = 4;
-
-  int n_rois = nthreads / channels / pooled_width / pooled_height;
-  // (n, c, ph, pw) is an element in the pooled output
-  parallel_for(n_rois, [&](size_t n) {
-    int index_n = n * channels * pooled_width * pooled_height;
-
-    // roi could have 4 or 5 columns
-    const T* offset_bottom_rois = bottom_rois + n * roi_cols;
-    int roi_batch_ind = 0;
-    if (roi_cols == 5) {
-      roi_batch_ind = static_cast<int>(offset_bottom_rois[0]);
-      offset_bottom_rois++;
-    }
-
-    T offset = aligned ? (T)0.5 : (T)0.0;
-    // Do not using rounding; this implementation detail is critical
-    T roi_start_w = offset_bottom_rois[0] * spatial_scale - offset;
-    T roi_start_h = offset_bottom_rois[1] * spatial_scale - offset;
-    T roi_end_w = offset_bottom_rois[2] * spatial_scale - offset;
-    T roi_end_h = offset_bottom_rois[3] * spatial_scale - offset;
-
-    // Force malformed ROIs to be 1x1
-    T roi_width = (std::max)(roi_end_w - roi_start_w, (T)1.);
-    T roi_height = (std::max)(roi_end_h - roi_start_h, (T)1.);
-    T bin_size_h = static_cast<T>(roi_height) / static_cast<T>(pooled_height);
-    T bin_size_w = static_cast<T>(roi_width) / static_cast<T>(pooled_width);
-
-    // We use roi_bin_grid to sample the grid and mimic integral
-    int roi_bin_grid_h = (sampling_ratio > 0)
-        ? sampling_ratio
-        : static_cast<int>(ceil(roi_height / pooled_height));  // e.g., = 2
-    int roi_bin_grid_w =
-        (sampling_ratio > 0) ? sampling_ratio : static_cast<int>(ceil(roi_width / pooled_width));
-
-    // We do average (integral) pooling inside a bin
-    const T count = static_cast<T>(roi_bin_grid_h * roi_bin_grid_w);  // e.g. = 4
-
-    // we want to precalculate indeces and weights shared by all chanels,
-    // this is the key point of optimiation
-    std::vector<PreCalc<T>> pre_calc(
-        roi_bin_grid_h * roi_bin_grid_w * pooled_width * pooled_height);
-    pre_calc_for_bilinear_interpolate(
-        height,
-        width,
-        pooled_height,
-        pooled_width,
-        roi_bin_grid_h,
-        roi_bin_grid_w,
-        roi_start_h,
-        roi_start_w,
-        bin_size_h,
-        bin_size_w,
-        roi_bin_grid_h,
-        roi_bin_grid_w,
-        pre_calc);
-
-    for (int c = 0; c < channels; c++) {
-      int index_n_c = index_n + c * pooled_width * pooled_height;
-      const T* offset_bottom_data =
-          bottom_data + (roi_batch_ind * channels + c) * height * width;
-      int pre_calc_index = 0;
-
-      for (int ph = 0; ph < pooled_height; ph++) {
-        for (int pw = 0; pw < pooled_width; pw++) {
-          int index = index_n_c + ph * pooled_width + pw;
-
-          T output_val = 0.;
-          for (int iy = 0; iy < roi_bin_grid_h; iy++) {
-            for (int ix = 0; ix < roi_bin_grid_w; ix++) {
-              PreCalc<T> pc = pre_calc[pre_calc_index];
-              output_val += pc.w1 * offset_bottom_data[pc.pos1] +
-                  pc.w2 * offset_bottom_data[pc.pos2] +
-                  pc.w3 * offset_bottom_data[pc.pos3] +
-                  pc.w4 * offset_bottom_data[pc.pos4];
-
-              pre_calc_index += 1;
-            }
-          }
-          output_val /= count;
-
-          top_data[index] = output_val;
-        }  // for pw
-      }  // for ph
-    }  // for c
-  });
-}
-
-
-void redistribute_rois(const float* rois, int* level_ids,
-                       const int num_rois, const int levels_num) {
-    const float canonical_scale = 224.0f;
-    const int canonical_level = 2;
-
-    for (int i = 0; i < num_rois; ++i) {
-        const float x0 = rois[4 * i + 0];
-        const float y0 = rois[4 * i + 1];
-        const float x1 = rois[4 * i + 2];
-        const float y1 = rois[4 * i + 3];
-
-        int target_level = levels_num;
-        float area = (x1 - x0) * (y1 - y0);
-        if (area > 0) {
-            area = std::sqrt(area) / canonical_scale;
-            area = std::log2(area + 1e-6f);
-            target_level = static_cast<int>(std::floor(area + canonical_level));
-            target_level = (std::max)(0, (std::min)(levels_num - 1, target_level));
-        }
-
-        level_ids[i] = target_level;
-    }
-}
-
-
-void reorder(const float* src_data, const int* ranks, const int n, const int step, float* dst_data,
-             int* dst_mapping) {
-    std::iota(dst_mapping, dst_mapping + n, 0);
-    std::sort(dst_mapping, dst_mapping + n, [&ranks](size_t i1, size_t i2) {return ranks[i1] < ranks[i2];});
-    for (int i = 0; i < n; ++i) {
-        const int j = dst_mapping[i];
-        assert(0 <= j && j < n);
-        cpu_memcpy(dst_data + i * step, src_data + j * step, sizeof(float) * step);
-    }
-}
-
-void split_points(const std::vector<int>& ids, std::vector<int>& rois_per_level, const int levels_num) {
-    rois_per_level.clear();
-    rois_per_level.resize(levels_num, 0);
-    for (size_t i = 0; i < ids.size(); ++i) {
-        assert(0 <= ids[i] && ids[i] < levels_num);
-        rois_per_level[ids[i]]++;
-    }
-    for (int i = 1; i < levels_num; ++i) {
-        rois_per_level[i] += rois_per_level[i - 1];
-    }
-    rois_per_level.insert(rois_per_level.begin(), 0);
-}
-
-
-void reorder_rois(const float *rois, const int* ids, int* mapping, const int rois_num,
-                  float * reordered_rois, std::vector<int>& rois_per_level, const int levels_num) {
-    rois_per_level.clear();
-    rois_per_level.resize(levels_num, 0);
-    for (int i = 0; i < rois_num; ++i) {
-        assert(0 <= ids[i] && ids[i] < levels_num);
-        rois_per_level[ids[i]]++;
-    }
-    for (int i = 1; i < levels_num; ++i) {
-        rois_per_level[i] += rois_per_level[i - 1];
-    }
-    rois_per_level.insert(rois_per_level.begin(), 0);
-
-    std::vector<int> level_counter = rois_per_level;
-
-    for (int i = 0; i < rois_num; ++i) {
-        const int level = ids[i];
-        assert(level < levels_num);
-        const int j = level_counter[level];
-        assert(0 <= j && j < rois_num);
-        reordered_rois[j * 4 + 0] = rois[i * 4 + 0];
-        reordered_rois[j * 4 + 1] = rois[i * 4 + 1];
-        reordered_rois[j * 4 + 2] = rois[i * 4 + 2];
-        reordered_rois[j * 4 + 3] = rois[i * 4 + 3];
-        level_counter[level]++;
-    }
-}
-
-class ExperimentalDetectronROIFeatureExtractorImpl: public ExtLayerBase {
-private:
-    const int INPUT_ROIS {0};
-    const int INPUT_FEATURES_START {1};
-
-    const int OUTPUT_ROI_FEATURES {0};
-    const int OUTPUT_ROIS {1};
-
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            const auto roiFeatureExtractor = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronROIFeatureExtractor>(op);
-            if (!roiFeatureExtractor) {
-                errorMessage = "Only opset6 ExperimentalDetectronROIFeatureExtractor operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-public:
-    explicit ExperimentalDetectronROIFeatureExtractorImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            const auto roiFeatureExtractor = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronROIFeatureExtractor>(op);
-            const auto &attr = roiFeatureExtractor->get_attrs();
-            output_dim_ = attr.output_size;
-            pyramid_scales_ = attr.pyramid_scales;
-            sampling_ratio_ = attr.sampling_ratio;
-            aligned_ = attr.aligned;
-            pooled_height_ = output_dim_;
-            pooled_width_ = output_dim_;
-
-            std::vector<DataConfigurator> inDataConfigurators(op->get_input_size(), DataConfigurator{TensorDescCreatorTypes::ncsp, Precision::FP32});
-            addConfig(op, inDataConfigurators,
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                           {TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs,
-                       ResponseDesc *resp) noexcept override {
-        const int levels_num = inputs.size() - INPUT_FEATURES_START;
-        const int num_rois = inputs[INPUT_ROIS]->getTensorDesc().getDims()[0];
-        const int channels_num = inputs[INPUT_FEATURES_START]->getTensorDesc().getDims()[1];
-        const int feaxels_per_roi = pooled_height_ * pooled_width_ * channels_num;
-
-        auto *input_rois = inputs[INPUT_ROIS]->buffer().as<const float *>();
-        auto *output_rois_features = outputs[OUTPUT_ROI_FEATURES]->buffer().as<float *>();
-        float *output_rois = nullptr;
-        if (OUTPUT_ROIS < static_cast<int>(outputs.size())) {
-            output_rois = outputs[OUTPUT_ROIS]->buffer().as<float *>();
-        }
-
-        std::vector<int> level_ids(num_rois, 0);
-        redistribute_rois(input_rois, reinterpret_cast<int *>(&level_ids[0]), num_rois, levels_num);
-
-        std::vector<float> reordered_rois(4 * num_rois, 0);
-        std::vector<int> original_rois_mapping(num_rois, 0);
-        reorder(input_rois, &level_ids[0], num_rois, 4, &reordered_rois[0], &original_rois_mapping[0]);
-
-        std::vector<int> rois_per_level;
-        split_points(level_ids, rois_per_level, levels_num + 1);
-
-        std::vector<float> output_rois_features_temp(feaxels_per_roi * num_rois, 0);
-        for (int i = 0; i < levels_num; ++i) {
-            const int level_rois_offset = rois_per_level[i];
-            const int level_rois_num = rois_per_level[i + 1] - level_rois_offset;
-            if (level_rois_num > 0) {
-                auto *featuremap = inputs[INPUT_FEATURES_START + i]->buffer().as<const float *>();
-                const int featuremap_height = inputs[INPUT_FEATURES_START + i]->getTensorDesc().getDims()[2];
-                const int featuremap_width = inputs[INPUT_FEATURES_START + i]->getTensorDesc().getDims()[3];
-                ROIAlignForward_cpu_kernel<float>(feaxels_per_roi * level_rois_num,
-                    featuremap,
-                    1.0f / pyramid_scales_[i],
-                    channels_num,
-                    featuremap_height,
-                    featuremap_width,
-                    pooled_height_,
-                    pooled_width_,
-                    sampling_ratio_,
-                    &reordered_rois[4 * level_rois_offset],
-                    aligned_,
-                    &output_rois_features_temp[feaxels_per_roi * level_rois_offset]);
-            }
-        }
-
-        std::vector<int> dummy_mapping(num_rois, 0);
-        reorder(&output_rois_features_temp[0], &original_rois_mapping[0], num_rois, feaxels_per_roi,
-                output_rois_features, &dummy_mapping[0]);
-        if (output_rois != nullptr) {
-            cpu_memcpy(output_rois, input_rois, 4 * num_rois * sizeof(float));
-        }
-
-        return OK;
-    }
-
-private:
-    int output_dim_ = 0;
-    int pooled_height_ = 0;
-    int pooled_width_ = 0;
-    std::vector<int64_t> pyramid_scales_;
-    int sampling_ratio_ = 0;
-    bool aligned_ = false;
-};
-
-REG_FACTORY_FOR(ExperimentalDetectronROIFeatureExtractorImpl, ExperimentalDetectronROIFeatureExtractor);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/topk.cpp b/inference-engine/src/mkldnn_plugin/nodes/topk.cpp
deleted file mode 100644
index 31cd81b410dab5..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/topk.cpp
+++ /dev/null
@@ -1,572 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-
-#include <cmath>
-#include <limits>
-#include <cfloat>
-#include <string>
-#include <vector>
-#include <cassert>
-#include <functional>
-
-#include "ie_parallel.hpp"
-#include <ngraph/op/topk.hpp>
-#include "common/tensor_desc_creator.h"
-#include "utils/general_utils.h"
-#if defined(HAVE_SSE) || defined(HAVE_AVX2) || defined(HAVE_AVX512F)
-#include <immintrin.h>
-#endif
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-class TopKImpl: public ExtLayerBase {
-public:
-    bool isSupportedOperation(const std::shared_ptr<const ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            auto topKOp = ngraph::as_type_ptr<const ngraph::op::v1::TopK>(op);
-            if (!topKOp) {
-                errorMessage = "Node is not an instance of the TopK from the operations set v1 or v3";
-                return false;
-            }
-            if (topKOp->get_mode() != ngraph::op::TopKMode::MAX &&
-                    topKOp->get_mode() != ngraph::op::TopKMode::MIN) {
-                errorMessage = "Unsupported mode.";
-                return false;
-            }
-            if (!MKLDNNPlugin::one_of(topKOp->get_sort_type(), ngraph::op::TopKSortType::NONE,
-                    ngraph::op::TopKSortType::SORT_VALUES,
-                    ngraph::op::TopKSortType::SORT_INDICES)) {
-                errorMessage = "Unsupported sort type.";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    explicit TopKImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-                IE_THROW(NotImplemented) << errorMessage;
-            }
-            auto topK1Op = ngraph::as_type_ptr<ngraph::op::v1::TopK>(op);
-
-            SizeVector dstDims = topK1Op->get_output_shape(TOPK_VALUE);
-            src_dims = topK1Op->get_input_shape(TOPK_DATA);
-
-            axis = topK1Op->get_axis();
-
-            if (topK1Op->get_mode() == ngraph::op::TopKMode::MAX)
-                mode_max = true;
-            else
-                mode_max = false;
-
-            if (topK1Op->get_sort_type() == ngraph::op::TopKSortType::SORT_VALUES)
-                sort_value = true;
-            else
-                sort_value = false;
-
-            int j;
-            for (j = src_dims.size() - 1; j >= 0; j--) {
-                if (src_dims[j] != 1) break;
-            }
-            if (static_cast<size_t>(j) == axis) is_last_dim = true;
-
-            for (size_t i = 0; i < axis; i++) {
-                axis_step *= src_dims[i];
-            }
-            axis_dim = src_dims[axis];
-            for (size_t i = (axis + 1); i < src_dims.size(); i++) {
-                axis_stride *= src_dims[i];
-            }
-            dim = static_cast<int>(src_dims[axis]);
-            before_num = count(src_dims, 0, axis);
-
-            if (topK1Op->get_output_size() == 1) {
-                addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                               {TensorDescCreatorTypes::ncsp, Precision::I32}},
-                              {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-            } else {
-                addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                               {TensorDescCreatorTypes::ncsp, Precision::I32}},
-                              {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                               {TensorDescCreatorTypes::ncsp, Precision::I32}});
-            }
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-            throw;
-        }
-    }
-
-#if defined(HAVE_AVX512F)
-    const int block_size = 16;
-    typedef __m512 vec_type_f;
-    typedef __m512i vec_type_i;
-    typedef __mmask16 vmask_type;
-#elif defined(HAVE_AVX2)
-    const int block_size = 8;
-    typedef __m256 vec_type_f;
-    typedef __m256i vec_type_i;
-    typedef __m256 vmask_type;
-#elif defined(HAVE_SSE)
-    const int block_size = 4;
-    typedef __m128 vec_type_f;
-    typedef __m128i vec_type_i;
-    typedef __m128 vmask_type;
-#else
-    typedef float vec_type_f;
-    typedef int vmask_type;
-#endif
-
-    struct cmpgt_ps {
-        static inline vmask_type cmp_ps(const vec_type_f _Left, const vec_type_f _Right) {
-#if defined(HAVE_SSE) || defined(HAVE_AVX2) || defined(HAVE_AVX512F)
-            return _mm_uni_cmpgt_ps(_Left, _Right);
-#else
-            return _Left > _Right ? _Left : _Right;
-#endif
-        }
-    };
-
-    struct cmplt_ps {
-        static inline vmask_type cmp_ps(const vec_type_f _Left, const vec_type_f _Right) {
-#if defined(HAVE_SSE) || defined(HAVE_AVX2) || defined(HAVE_AVX512F)
-            return _mm_uni_cmpgt_ps(_Right, _Left);
-#else
-            return _Right > _Left ? _Right : _Left;
-#endif
-        }
-    };
-
-    template <class Compare1, template <typename> class Compare2>
-    void top1_axis(const float* src_data, float* dst_data, int* dst_idx, SizeVector in_dims) {
-        int after_num = count(in_dims, axis + 1, in_dims.size());
-        int first_index = 0;
-
-#if defined(HAVE_SSE) || defined(HAVE_AVX2) || defined(HAVE_AVX512F)
-        parallel_for2d(before_num, after_num / block_size, [&](int i0, int ib1) {
-            int s_index = i0 * dim * after_num + ib1 * block_size;
-            vec_type_f vmax_val = _mm_uni_loadu_ps(src_data + s_index);
-            vec_type_i vindex_max_val = _mm_uni_setzero_si();
-            for (int i2 = 1; i2 < dim; i2++) {
-                s_index += after_num;
-                vec_type_f vsrc = _mm_uni_loadu_ps(src_data + s_index);
-                vmask_type vmask = Compare1::cmp_ps(vsrc, vmax_val);
-                vmax_val = _mm_uni_blendv_ps(vmax_val, vsrc, vmask);
-
-                vec_type_i vindex_cur_val = _mm_uni_set1_epi32(i2);
-#if defined(HAVE_AVX512F)
-                vindex_max_val = _mm512_mask_blend_epi32(vmask, vindex_max_val, vindex_cur_val);
-#else
-                vindex_max_val = _mm_uni_blendv_epi8(vindex_max_val, vindex_cur_val, _mm_uni_castps_si(vmask));
-#endif
-            }
-            if (dst_data)
-                _mm_uni_storeu_ps(dst_data + i0 * after_num + ib1 * block_size, vmax_val);
-            if (dst_idx)
-                _mm_uni_storeu_si(reinterpret_cast<vec_type_i*>(dst_idx + i0 * after_num + ib1 * block_size), vindex_max_val);
-        });
-        first_index = after_num / block_size * block_size;
-#endif
-        int rest = after_num - first_index;
-        parallel_for2d(before_num, rest, [&](int i0, int i1) {
-            int index_max_val = 0;
-            int s_index = i0 * dim * after_num + first_index + i1;
-            float max_val = src_data[s_index];
-            for (int i2 = 1; i2 < dim; i2++) {
-                s_index += after_num;
-                if (Compare2<float>()(src_data[s_index], max_val)) {
-                    max_val = src_data[s_index];
-                    index_max_val = i2;
-                }
-            }
-            if (dst_data)
-                dst_data[i0 * after_num + first_index + i1] = max_val;
-            if (dst_idx)
-                dst_idx[i0 * after_num + first_index + i1] = index_max_val;
-        });
-    }
-
-    template <template <typename> class Compare>
-    void top1(const float* src_data, float* dst_data, int* dst_idx, SizeVector in_dims) {
-        parallel_for(before_num, [&](int i0) {
-            int index_max_val = 0;
-            int s_index = i0 * dim;
-            float max_val = src_data[s_index];
-            for (int i1 = 1; i1 < dim; i1++) {
-                s_index++;
-                if (Compare<float>()(src_data[s_index], max_val)) {
-                    max_val = src_data[s_index];
-                    index_max_val = i1;
-                }
-            }
-            if (dst_data)
-                dst_data[i0] = max_val;
-            if (dst_idx)
-                dst_idx[i0] = index_max_val;
-        });
-    }
-
-    template <class Compare1, template <typename> class Compare2>
-    void topk_axis(const float* src_data, float* dst_data, int* dst_idx, SizeVector in_dims) {
-        int after_num = count(in_dims, axis + 1, in_dims.size());
-        int first_index = 0;
-
-#if defined(HAVE_SSE) || defined(HAVE_AVX2) || defined(HAVE_AVX512F)
-        if (src_k < count_vec) {
-            parallel_for2d(before_num, after_num / block_size, [&](int i0, int ib1) {
-#if defined(HAVE_AVX512F)
-                const int N = 32;
-                vec_type_f vmax_values[N];
-                vec_type_i vmax_indexes[N];
-#else
-                const int N = 16;
-                vec_type_f vmax_values[N];
-                vec_type_i vmax_indexes[N];
-#endif
-                vec_type_f vtmp;
-                vec_type_i vtmp_indexes;
-                vmask_type vmask;
-                int s_index = i0 * dim * after_num + ib1 * block_size;
-
-                auto vswap_func = [&](int index1, int index2) {
-                    vtmp = vmax_values[index1];
-                    vmax_values[index1] = _mm_uni_blendv_ps(vmax_values[index1], vmax_values[index2], vmask);
-                    vmax_values[index2] = _mm_uni_blendv_ps(vmax_values[index2], vtmp, vmask);
-
-                    vtmp_indexes = vmax_indexes[index1];
-#if defined(HAVE_AVX512F)
-                    vmax_indexes[index1] = _mm512_mask_blend_epi32(vmask, vmax_indexes[index1], vmax_indexes[index2]);
-                    vmax_indexes[index2] = _mm512_mask_blend_epi32(vmask, vmax_indexes[index2], vtmp_indexes);
-#else
-                    vmax_indexes[index1] = _mm_uni_blendv_epi8(vmax_indexes[index1], vmax_indexes[index2], _mm_uni_castps_si(vmask));
-                    vmax_indexes[index2] = _mm_uni_blendv_epi8(vmax_indexes[index2], vtmp_indexes, _mm_uni_castps_si(vmask));
-#endif
-                };
-
-                for (int i2 = 0; i2 < src_k; i2++) {
-                    vmax_values[i2] = _mm_uni_loadu_ps(src_data + s_index);
-                    vmax_indexes[i2] = _mm_uni_set1_epi32(i2);
-                    s_index += after_num;
-                }
-                for (int i2 = 0; i2 < src_k - 1; i2++) {
-                    for (int i3 = src_k - 1; i3 > i2; i3--) {
-                        vmask = Compare1::cmp_ps(vmax_values[i3], vmax_values[i3 - 1]);
-#if defined(HAVE_AVX512F)
-                        if (vmask)
-                            vswap_func(i3, i3 - 1);
-#else
-                        int swap = _mm_uni_movemask_ps(vmask);
-                        if (swap)
-                            vswap_func(i3, i3 - 1);
-#endif
-                    }
-                }
-                for (int i2 = src_k; i2 < dim; i2++) {
-                    vmax_values[src_k] = _mm_uni_loadu_ps(src_data + s_index);
-                    vmax_indexes[src_k] = _mm_uni_set1_epi32(i2);
-                    for (int i3 = src_k; i3 > 0; i3--) {
-                        vmask = Compare1::cmp_ps(vmax_values[i3], vmax_values[i3 - 1]);
-#if defined(HAVE_AVX512F)
-                        if (vmask)
-                            vswap_func(i3, i3 - 1);
-                        else
-                            break;
-#else
-                        int swap = _mm_uni_movemask_ps(vmask);
-                        if (swap)
-                            vswap_func(i3, i3 - 1);
-                        else
-                            break;
-#endif
-                    }
-                    s_index += after_num;
-                }
-                if (!sort_value) {
-                    for (int i2 = 0; i2 < src_k - 1; i2++) {
-                        for (int i3 = src_k - 1; i3 > i2; i3--) {
-                            vmask = _mm_uni_cmpgt_i32(vmax_indexes[i3 - 1], vmax_indexes[i3]);
-#if defined(HAVE_AVX512F)
-                            if (vmask)
-                                vswap_func(i3, i3 - 1);
-                            else
-                                break;
-#else
-                            int swap = _mm_uni_movemask_ps(vmask);
-                            if (swap)
-                                vswap_func(i3, i3 - 1);
-                            else
-                                break;
-#endif
-                        }
-                    }
-                }
-                if (dst_data) {
-                    for (int i2 = 0; i2 < src_k; i2++)
-                        _mm_uni_storeu_ps(dst_data + (i0 * src_k + i2) * after_num + ib1 * block_size, vmax_values[i2]);
-                }
-                if (dst_idx) {
-                    for (int i2 = 0; i2 < src_k; i2++)
-                        _mm_uni_storeu_si(reinterpret_cast<vec_type_i*>(dst_idx + (i0 * src_k + i2) * after_num + ib1 * block_size), vmax_indexes[i2]);
-                }
-            });
-            first_index = after_num / block_size * block_size;
-        }
-#endif
-        int rest = after_num - first_index;
-        parallel_for2d(before_num, rest, [&](int i0, int i1) {
-            std::vector<float> max_values(src_k + 1);
-            std::vector<int> max_indexes(src_k + 1);
-            float tmp_value;
-            int tmp_index;
-            int s_index = i0 * dim * after_num + first_index + i1;
-
-            auto swap_func = [&](int index1, int index2) {
-                tmp_value = max_values[index1];
-                max_values[index1] = max_values[index2];
-                max_values[index2] = tmp_value;
-
-                tmp_index = max_indexes[index1];
-                max_indexes[index1] = max_indexes[index2];
-                max_indexes[index2] = tmp_index;
-            };
-
-            for (int i2 = 0; i2 < src_k; i2++) {
-                max_values[i2] = src_data[s_index];
-                max_indexes[i2] = i2;
-                s_index += after_num;
-            }
-            for (int i2 = 0; i2 < src_k - 1; i2++) {
-                for (int i3 = src_k - 1; i3 > i2; i3--) {
-                    if (Compare2<float>()(max_values[i3], max_values[i3 - 1])) {
-                        swap_func(i3, i3 - 1);
-                    }
-                }
-            }
-            for (int i2 = src_k; i2 < dim; i2++) {
-                max_values[src_k] = src_data[s_index];
-                max_indexes[src_k] = i2;
-                for (int i3 = src_k; i3 > 0; i3--) {
-                    if (Compare2<float>()(max_values[i3], max_values[i3 - 1]))
-                        swap_func(i3, i3 - 1);
-                    else
-                        break;
-                }
-                s_index += after_num;
-            }
-            if (!sort_value) {
-                for (int i2 = 0; i2 < src_k - 1; i2++) {
-                    for (int i3 = src_k - 1; i3 > i2; i3--) {
-                        if (std::greater<int>()(max_indexes[i3 - 1], max_indexes[i3])) {
-                            swap_func(i3, i3 - 1);
-                        }
-                    }
-                }
-            }
-            if (dst_data) {
-                for (int i2 = 0; i2 < src_k; i2++)
-                    dst_data[i0 * src_k * after_num + i2 * after_num + first_index + i1] = max_values[i2];
-            }
-            if (dst_idx) {
-                for (int i2 = 0; i2 < src_k; i2++)
-                    dst_idx[i0 * src_k * after_num + i2 * after_num + first_index + i1] = max_indexes[i2];
-            }
-        });
-    }
-
-    template <template <typename> class Compare>
-    void topk(const float* src_data, float* dst_data, int* dst_idx, SizeVector in_dims) {
-        parallel_for(before_num, [&](int i0) {
-            std::vector<float> max_values(src_k + 1);
-            std::vector<int> max_indexes(src_k + 1);
-            float tmp_value;
-            int tmp_index;
-            int s_index = i0 * dim;
-
-            auto swap_func = [&](int index1, int index2) {
-                tmp_value = max_values[index1];
-                max_values[index1] = max_values[index2];
-                max_values[index2] = tmp_value;
-
-                tmp_index = max_indexes[index1];
-                max_indexes[index1] = max_indexes[index2];
-                max_indexes[index2] = tmp_index;
-            };
-
-            for (int i2 = 0; i2 < src_k; i2++) {
-                max_values[i2] = src_data[s_index];
-                max_indexes[i2] = i2;
-                s_index++;
-            }
-            for (int i2 = 0; i2 < src_k - 1; i2++) {
-                for (int i3 = src_k - 1; i3 > i2; i3--) {
-                    if (Compare<float>()(max_values[i3], max_values[i3 - 1])) {
-                        swap_func(i3, i3 - 1);
-                    }
-                }
-            }
-            for (int i2 = src_k; i2 < dim; i2++) {
-                max_values[src_k] = src_data[s_index];
-                max_indexes[src_k] = i2;
-                for (int i3 = src_k; i3 > 0; i3--) {
-                    if (Compare<float>()(max_values[i3], max_values[i3 - 1]))
-                        swap_func(i3, i3 - 1);
-                    else
-                        break;
-                }
-                s_index++;
-            }
-            if (!sort_value) {
-                for (int i2 = 0; i2 < src_k - 1; i2++) {
-                    for (int i3 = src_k - 1; i3 > i2; i3--) {
-                        if (std::greater<int>()(max_indexes[i3 - 1], max_indexes[i3])) {
-                            swap_func(i3, i3 - 1);
-                        }
-                    }
-                }
-            }
-            if (dst_data) {
-                for (int i2 = 0; i2 < src_k; i2++)
-                    dst_data[i0 * src_k + i2] = max_values[i2];
-            }
-            if (dst_idx) {
-                for (int i2 = 0; i2 < src_k; i2++)
-                    dst_idx[i0 * src_k + i2] = max_indexes[i2];
-            }
-        });
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs, ResponseDesc *resp) noexcept override {
-        const float *src = inputs[TOPK_DATA]->cbuffer().as<float *>() +
-            inputs[TOPK_DATA]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-        src_k = (inputs[TOPK_K]->cbuffer().as<int *>() +
-            inputs[TOPK_K]->getTensorDesc().getBlockingDesc().getOffsetPadding())[0];
-        float* dst_data = nullptr;
-        int* dst_idx = nullptr;
-
-        if (outputs.size() == 1) {
-            if (outputs[0]->getTensorDesc().getPrecision() == Precision::FP32) {
-                dst_data = outputs[0]->buffer().as<float *>() +
-                    outputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-            } else {
-                dst_idx = outputs[0]->buffer().as<int *>() +
-                    outputs[0]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-            }
-            SizeVector dstDims = outputs[0]->getTensorDesc().getDims();
-
-            if (dstDims[axis] != static_cast<size_t>(src_k)) {
-                if (resp) {
-                    std::string errorMsg = "Output tensor dimension mismatch";
-                    errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-                }
-                return PARAMETER_MISMATCH;
-            }
-        } else if (outputs.size() == 2) {
-            dst_data = outputs[TOPK_VALUE]->buffer().as<float *>() +
-                outputs[TOPK_VALUE]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-            SizeVector dst_data_dims = outputs[TOPK_VALUE]->getTensorDesc().getDims();
-
-            dst_idx = outputs[TOPK_INDEX]->buffer().as<int *>() +
-                outputs[TOPK_INDEX]->getTensorDesc().getBlockingDesc().getOffsetPadding();
-            SizeVector dst_idx_dims = outputs[TOPK_INDEX]->getTensorDesc().getDims();
-
-            if (dst_idx_dims[axis] != static_cast<size_t>(src_k) || dst_data_dims[axis] != static_cast<size_t>(src_k)) {
-                if (resp) {
-                    std::string errorMsg = "Output tensors dimension mismatch";
-                    errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-                }
-                return PARAMETER_MISMATCH;
-            }
-        } else {
-            if (resp) {
-                std::string errorMsg = "Output tensors amount mismatch";
-                errorMsg.copy(resp->msg, sizeof(resp->msg) - 1);
-            }
-            return PARAMETER_MISMATCH;
-        }
-
-        if (src_dims[axis] < static_cast<size_t>(src_k))
-            src_k = src_dims[axis];
-
-        SizeVector in_dims = inputs[TOPK_DATA]->getTensorDesc().getDims();
-
-        if (src_k == 1) {
-            if (is_last_dim) {
-                if (mode_max)
-                    top1<std::greater>(src, dst_data, dst_idx, in_dims);
-                else
-                    top1<std::less>(src, dst_data, dst_idx, in_dims);
-            } else {
-                if (mode_max)
-                    top1_axis<cmpgt_ps, std::greater>(src, dst_data, dst_idx, in_dims);
-                else
-                    top1_axis<cmplt_ps, std::less>(src, dst_data, dst_idx, in_dims);
-            }
-        } else {
-            if (is_last_dim) {
-                if (mode_max)
-                    topk<std::greater>(src, dst_data, dst_idx, in_dims);
-                else
-                    topk<std::less>(src, dst_data, dst_idx, in_dims);
-            } else {
-                if (mode_max)
-                    topk_axis<cmpgt_ps, std::greater>(src, dst_data, dst_idx, in_dims);
-                else
-                    topk_axis<cmplt_ps, std::less>(src, dst_data, dst_idx, in_dims);
-            }
-        }
-
-        return OK;
-    }
-
-private:
-    const size_t TOPK_DATA = 0;
-    const size_t TOPK_K = 1;
-    const size_t TOPK_VALUE = 0;
-    const size_t TOPK_INDEX = 1;
-
-    SizeVector src_dims;
-    size_t axis;
-    size_t axis_dim;
-    size_t axis_stride = 1;
-    size_t axis_step = 1;
-    bool is_last_dim = false;
-    int src_k = 1;
-
-    bool sort_value = false;
-    bool mode_max = true;
-
-    int dim, before_num;
-
-#if defined(HAVE_AVX512F)
-    const int count_vec = 32;
-#elif defined(HAVE_SSE) || defined(HAVE_AVX2)
-    const int count_vec = 16;
-#endif
-
-    inline int count(SizeVector dims, size_t start_ind, size_t end_ind) {
-        size_t count = 1;
-        for (size_t i = start_ind; i < end_ind; i++)
-            count *= dims[i];
-        return static_cast<int>(count);
-    }
-
-    inline int count(SizeVector dims, size_t start_ind = 0) {
-        return count(dims, start_ind, dims.size());
-    }
-};
-
-REG_FACTORY_FOR(TopKImpl, TopK);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/src/mkldnn_plugin/nodes/topkrois_onnx.cpp b/inference-engine/src/mkldnn_plugin/nodes/topkrois_onnx.cpp
deleted file mode 100644
index e4203ef564f2a1..00000000000000
--- a/inference-engine/src/mkldnn_plugin/nodes/topkrois_onnx.cpp
+++ /dev/null
@@ -1,101 +0,0 @@
-// Copyright (C) 2018-2021 Intel Corporation
-// SPDX-License-Identifier: Apache-2.0
-//
-
-#include "base.hpp"
-#include <algorithm>
-#include <cassert>
-#include <vector>
-#include "common/cpu_memcpy.h"
-#include <ngraph/opsets/opset6.hpp>
-
-using MKLDNNPlugin::TensorDescCreatorTypes;
-
-namespace InferenceEngine {
-namespace Extensions {
-namespace Cpu {
-
-class ExperimentalDetectronTopKROIsImpl: public ExtLayerBase {
-private:
-    // Inputs:
-    //      rois, shape [n, 4]
-    //      rois_probs, shape [n]
-    // Outputs:
-    //      top_rois, shape [max_rois, 4]
-
-    const int INPUT_ROIS {0};
-    const int INPUT_PROBS {1};
-
-    const int OUTPUT_ROIS {0};
-
-    bool isSupportedOperation(const std::shared_ptr<ngraph::Node>& op, std::string& errorMessage) noexcept {
-        try {
-            const auto topKROI = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronTopKROIs>(op);
-            if (!topKROI) {
-                errorMessage = "Only opset6 ExperimentalDetectronTopKROIs operation is supported";
-                return false;
-            }
-        } catch (...) {
-            return false;
-        }
-        return true;
-    }
-
-    std::string errorPrefix;
-
-public:
-    explicit ExperimentalDetectronTopKROIsImpl(const std::shared_ptr<ngraph::Node>& op) {
-        try {
-            std::string errorMessage;
-            if (!isSupportedOperation(op, errorMessage)) {
-              IE_THROW(NotImplemented) << errorMessage;
-            }
-
-            errorPrefix = "ExperimentalDetectronTopKROIs layer with name '" + op->get_friendly_name() + "'";
-            const auto topKROI = std::dynamic_pointer_cast<const ngraph::opset6::ExperimentalDetectronTopKROIs>(op);
-            if (op->get_input_size() != 2 || op->get_output_size() != 1)
-                IE_THROW() << errorPrefix << " has incorrect number of input/output edges!";
-
-            if (op->get_input_shape(INPUT_ROIS).size() != 2 || op->get_input_shape(INPUT_PROBS).size() != 1)
-                IE_THROW() << errorPrefix << " has nsupported input shape";
-
-            max_rois_num_ = topKROI->get_max_rois();
-
-            addConfig(op, {{TensorDescCreatorTypes::ncsp, Precision::FP32},
-                           {TensorDescCreatorTypes::ncsp, Precision::FP32}},
-                          {{TensorDescCreatorTypes::ncsp, Precision::FP32}});
-        } catch (InferenceEngine::Exception &ex) {
-            errorMsg = ex.what();
-        }
-    }
-
-    StatusCode execute(std::vector<Blob::Ptr>& inputs, std::vector<Blob::Ptr>& outputs,
-                       ResponseDesc *resp) noexcept override {
-        const int input_rois_num = inputs[INPUT_ROIS]->getTensorDesc().getDims()[0];
-        const int top_rois_num = (std::min)(max_rois_num_, input_rois_num);
-
-        auto *input_rois = inputs[INPUT_ROIS]->buffer().as<const float *>();
-        auto *input_probs = inputs[INPUT_PROBS]->buffer().as<const float *>();
-        auto *output_rois = outputs[OUTPUT_ROIS]->buffer().as<float *>();
-
-        std::vector<size_t> idx(input_rois_num);
-        iota(idx.begin(), idx.end(), 0);
-        // FIXME. partial_sort is enough here.
-        sort(idx.begin(), idx.end(), [&input_probs](size_t i1, size_t i2) {return input_probs[i1] > input_probs[i2];});
-
-        for (int i = 0; i < top_rois_num; ++i) {
-            cpu_memcpy(output_rois + 4 * i, input_rois + 4 * idx[i], 4 * sizeof(float));
-        }
-
-        return OK;
-    }
-
-private:
-    int max_rois_num_;
-};
-
-REG_FACTORY_FOR(ExperimentalDetectronTopKROIsImpl, ExperimentalDetectronTopKROIs);
-
-}  // namespace Cpu
-}  // namespace Extensions
-}  // namespace InferenceEngine
diff --git a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/extract_image_patches.cpp b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/extract_image_patches.cpp
index 20edee2aad70d4..58e532a7f1cc8e 100755
--- a/inference-engine/tests/functional/plugin/cpu/single_layer_tests/extract_image_patches.cpp
+++ b/inference-engine/tests/functional/plugin/cpu/single_layer_tests/extract_image_patches.cpp
@@ -43,7 +43,7 @@ class ExtractImagePatchesLayerCPUTest : public testing::WithParamInterface<extra
         ngraph::op::PadType pad_type;
         InferenceEngine::Precision netPrecision;
         std::tie(inputShape, kernel, strides, rates, pad_type, netPrecision, inPrc, outPrc, inLayout, targetDevice) = basicParamsSet;
-        selectedType = std::string("unknown_") + netPrecision.name();
+        selectedType = std::string("ref_any_") + netPrecision.name();
 
         auto ngPrc = FuncTestUtils::PrecisionUtils::convertIE2nGraphPrc(netPrecision);
         auto inputNode = std::make_shared<ngraph::opset6::Parameter>(ngPrc, ngraph::Shape(inputShape));

From 19b551e068694a1ed3c85846fed559d9fb036b54 Mon Sep 17 00:00:00 2001
From: Egor Duplensky <egor.duplenskii@intel.com>
Date: Wed, 23 Jun 2021 14:26:27 +0300
Subject: [PATCH 178/178] [CPU] Update mkl-dnn version with fix for false
 positive assert (#5936)

---
 inference-engine/thirdparty/mkl-dnn | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/inference-engine/thirdparty/mkl-dnn b/inference-engine/thirdparty/mkl-dnn
index e0381c369fc3be..08072cad466a0f 160000
--- a/inference-engine/thirdparty/mkl-dnn
+++ b/inference-engine/thirdparty/mkl-dnn
@@ -1 +1 @@
-Subproject commit e0381c369fc3bed487b0dcfef7e9fcb2e0aea575
+Subproject commit 08072cad466a0fddeddd26b9ecf393fb37df617a